VOLUME 68 1981 OF THE mn I I -i»j^_ Published © Missouri Botanical Garden 1982 MiseoUAf BOTANrSKC CONTENTS 204 Allorge, L. (See Markgraf, F., P. Boiteau & L. Allorge) 677 Almeda, Frank, Jr. The Mexican and Central American Species of Adelobotiys (Melastomataceae) AvERETT, John E. (See D'Arcy, W. G., Johnnie L. Gentry, Jr. & John E. Averett) Announcements 1982 AETFAT Congress 226 227 Annual Midwest Botany Graduate Student Meeting 504 Errata Flora Mesoamericana 231 228 230 125 254 254 524 The 1981 Jesse M. Greenman Award Ayala F., Franklin. Four New Species of Dioscorea from Amazonian Peru Bawa, K. S. & J. H. Beach. Evolution of Sexual Systems in Flower- ing Plants Beach, J. H. (See Bawa, K. S. & J. H. Beach) Blackwell, Will H. & Martha J. Powell. A Preliminary Note on Pollination m the Chenopodiaceae Boiteau, P. (See Markgraf, F., P. Boiteau & L. Allorge) 677 ^^'yf.^^'' ^"^^^. ^- Cortella de. (See Ormond, Wilma T., Maria Ceha B. Pinheiro & Alicia R. Cortella de Castells) 514 Clayton, W. D. Evolution and Distribution of Grasses 5 Cohen, Dan. (See Dafni, Amots, Dan Cohen & Immanuel Noy-Meir) .... 652 Connor, H. E. Evolution of Reproductive Systems in the Gramineae .. 48 Croat, Thomas B. A Revision of Syngonium (Araceae) 565 Dafni, Amots, Dan Cohen & Imanuel Noy-Meir. Life-Cycle Vari- ation m Geophytes 652 Dahlgren Rolf & Peter Goldblatt. A Note on the Rediscovery of Argyrolobium involucratum (Thunb.) Harv. and the Generic Border- line Between Argyrolobium and Melolobium (Fabaceae-Crotalarieae) 558 ^ "^Panan^' ^' ^ ^^"^ ^^^''^' ""^ Hernandia (Hernandiaceae) from D'Arcy, W G., Johnnie L. Gentry, Jr. & John E.'averett" Recognition of Brachistus (Solanaceae) , 226 D'Arcy, W. G. (See Hammel, Barry E. «fe W. G. D'Arcy) 213 Davidse, Gerrit. Chromosome Number of Miscellaneous Angiosperms 222 M^ _■ Davidse, Gerrit. Setaria vahifolia (Swallen) Davidse, a New Combi- nation (Gramineae: Panicoideae) 226 deWet, J. M. J. Grasses and the Culture History of Man : 87 Dillon, Michael O. Three New Species of Floiirensia (Asteraceae- 668 Heliantheae) from South America 105 Engel, John J. Haplomitrium monoicum, A Remarkable New Species of Calobryales (Hepaticae) from New Caledonia, Together with a Reclassification of Subg. Haplomitrium Eyde, Richard H. Reproductive Structures and Evolution in Ludwigia (Onagraceae). III. Vasculature, Nectaries, Conclusions 379 Foster, Robin. (See Gentry, Alwyn H. & Robin Foster) 122 Gentry, Alwyn H. & Robin Foster. A New Peruvian Styloceras (Buxaceae): Discovery of a Phytogeographical Missing Link 122 Gentry, Alwyn H. New Species and a New Combination in Palmae, 112 218 226 f Theaceae, Araliaceae, Apocynaceae, and Bignoniaceae from the Choco and Amazonian Peru .- FoRERo, Enrique. New Species of Connams (Connaraceae) from Peru Gentry, Johnnie L., Jr. (See D'Arcy, W. G., Johnnie L. Gentry, Jr. & John E. Averett) Goldblatt, Peter. Chromosome Cytology of Bruniaceae ...„_^_ 546 Goldblatt, Peter. Chromosome Numbers in Legumes II 1:^ 551 Goldblatt, Peter. Notes on the Cytology and Distribution of Anap- aUna, Tritoniopsis, and Sparaxis, Cape Iridaceae 562 Goldblatt, Peter. Systematics and Biology of Homeria 413 Goldblatt, Peter. Systematics, Phylogeny and Evolution of Dietes (Iridaceae) 132 4 Goldblatt, Peter. (See Dahlgren, Rolf & Peter Goldblatt) 558 Gould, Lucille. Frank Walton Gould, 1913-1981 Hammel, Barry E. & W. G. D'Arcy. New Taxa from the Uplands of Panama 213 Heinrich, Bernd. The Energetics of Pollination 370 Iltis, Hugh H. Studies in the Capparidaceae XV: Capparis pana- mensis, n. sp. 681 Keil, David J. Pectis linifolia (Compositae: Tageteae) Added to the Flora of Panama KooPMAN, Karl F. The Distributional Patterns of New World Nectar- Feeding Bats Levin, Donald A. Dispersal Versus Gene Flow in Plants 233 352 Wilbur 154 Maas, Paul J. M. On the True Identity of Lagenanthus parviflorus Ewan (Gentianaceae) Markgraf, F., p. Boiteau & L. Allorge. Two New Species of 685 Bonafousia (Apocynaceae) from Panama and Colombia-Ecuador .... 677 McPherson, Gordon. Studies in Ipomoea (Convolvulaceae) L The Arborescens Group Neff, John L. (See Simpson, Beryl B. & John L. Neff) --. Noy-Meir, Imanuel. (See Dafni, Amots, Dan Cohen & Imanuel Noy- Meir) Ormond, Wilma T., Maria Celia B. Pinheiro & Alicia R. Cortella DE Castells. a Contribution to the Floral Biology and Reproduc- 527 301 652 live System of Coiiroupita guianensis Aubl. (Lecythidaceae) 514 172 514 Pilz, George E. Sapotaceae of Panama Pinheiro, Maria Celia B. (See Ormond, Wilma T., Maria Celia B. Pmheiro & Alicia R. Cortella de Castells) Pohl, Richard W. Evolution and Systematics of the Gramineae- The Twenty-sixth Systematics Symposium Pohl Richard W. Validation of the Name Aulonemia patriae Pohl (Gramineae: Bambusoideae) . . Powell, Martha J. (See Blackwell, Will H. & Martha J. Powell) 524 1 225 Slmpson, Beryl B. & John L. Neff. Floral Rewards: Alternatives to Pollen and Nectar 301 661 15 Smith, Gerald L. New Taxa in Piptocarpha R. Br. (Vernonieae: Compositae) SODERSTROM, Thomas R. Some Evolutionary Trends'in the'sam- busoideae (Poaceae) Stebbins, G. Ledyard. Coevolution of Grasses and Herbivores ... 75 Steiner, Kim E. Nectarivory and Potential Pollination by a Neo- tropical Marsupial ^ ^^^ ^'^''wiih^P.r^ri ■ i^f ^''^Ph'^^' Aspects of Bird-Flower Coevolution, with Particular Reference to Central America Systematics Symposia: Evolution of Reproductive Systems in the Gramineae: The Twenty- sixth Systematics Symposium . ^ytematics Symposium 233 323 1 Wilbur, Robert L. Additional Panamanian Species of Bui^cistem (Campanulaceae: Lobelioideae) i^urnuisura WiLLSON, Mary F. On the Evolution of ComplexLife Cycles in Plants' A Review and an Ecological Perspective ....... 167 154 275 k r^ THE m f-r.-f. ^ * Jt* JME68 1981 NUMBER 1 JOHN S. LCHMANN BUILDING, MISSOURI BOTANICAL GAf M CONTENTS MISSOURI BOTAHICAC EVOLUTION AND SYSTE^J^IC^ 9fot^**^ GRAMINEAE: The Twenty-sixth Systematics Sympo^nv^^^ Richard W. Pohl 1 4 Frank Walton Gould, 1913-1981 •*''°B/^f/^'*t?owW -" 4 Evolution and Distribution of Grasses W. D. Clayton 5 Some Evolutionary Trends in the Bambusoideae (Poaceae) Thomas R. Soderstrom 15 Evolution of Reproductive Systems in the Gramineae H. E. Connor — . 48 Coevolution of Grasses and Herbivores G. Ledyard Stebbins 75 Grasses and the Culture History of Man 7. M. J. deWet 87 Three New Species of Flourensia (Asteraceac-Heliantheae) from South America Michael O. Dillon - 105 New Species and a New Combination in Palmae, Theaceae, Araliaceae, Apocynaceae, and Bignoniaceae from the Choco and Amazonian Peru Alwyn H. Gentry 1 12 A New Peruvian Styloceras (Buxaceae): Discovery of a Phytogeographical Missing Link Ahvyn H. Gentry & Robin Foster 122 (Contents continued on back cover) VOLUME 68 1981 NUMBER 1 OF THE n mi MB The AXXALS contains papers, primarily in systematic botany, contributed from the M hould write the editor for information concerning arrangements for pub- lishing in the AxxALS. Editowal Committee GERRn Davidse, Editor Missouri Botanical Cardeii JOHX D. DWVEH Mi.s.souri Botanical Garden 6 St. Louis Vniversittj Peter Goldblatt Missouri Botanical Garden Published four times a year by the Missouri Botanical Garden, St. Louis, Missouri 63110. ISSN 0026-6493 ^" ^^Bo^^Sm^T' n "'"\^^" '^'^'"^" ^'^^ ''f 'he Annuls. i-.U. Hox 368, 1041 \ew Hampshire. Lawre-nce, Kansas 66044 Subscription price is $4.5 per volun.e U.S.. Canada, and Mexico, $50 all other countries. Four issues per volun.e. Second class postage paid at Lawrence. Kansas 66044 © Missouri Botanical Garden 1981 OF THE VOLUME 68 1981 NUMBER 1 EVOLUTION AND SYSTEMATICS OF THE GRAMINEAE THE TWENTY-SIXTH SYSTEMATICS SYMPOSIUM Richard W. Pohl^ M Garden 19-20 October 1979 with partial support from National Science Foun- dation Grant DEB78- 10180. Speakers at the Symposium were W. D. Clayton, Kew; H. E. Connor, Christchurch; Thomas R. Soderstrom, Washington, D.C.; G. Ledyard Stebbins, Davis; J. M. J. deWet, Urbana. Their contributions are published here. Other speakers, whose papers are not published in this issue, were Hugh H. litis. University of Maize — The Incredible TransforrriHtif Wi Madison: "From Teosinte to W sity. College Station: "Evolution in the Genus Bouteloua Lag," and G. Davidse, Missouri Botanical Garden: "A Review of Chromosome Numbers in the Gramineae." This was the first time in the Symposium's history that a specific plant family was chosen as a topic. The unrivaled economic importance of the grasses and their prominence in all major phytogeographic regions of the world seemed to be adequate reasons for this choice. The purpose of the Symposium was to review and reinterpret selected topics dealing with the evolution and systematics of the Gramineae. In the western world, the history of the classification of the grasses began with the treatment of the group by Theophrastus (fourth century b.c). In the first book in his Enquiry Into Plants (Hort, 1916), Theophrastus divided all plants into four groups: trees, shrubs, undershrubs, and herbs. Book VII of his treatment was entitled "Of Herbaceous Plants: Cereals, Pulses, and Summer Crops." In- cluded in this chapter were wheat, barley, one-seeded wheat, rice-wheat, millet, and Italian millet. Sesamum indicum of the Pedaliaceae was included. This treat- ment was obviously utilitarian, and does not define the family as a distinct and inclusive unit. ' Department of Botany, Iowa State University, Ames. Iowa 50011 * Deceased 1 1 March 1981. See second article of this issue. Ann. Missouri Box. Gard. 68: 1-3. 1981. 0026-6493/8 1 /OOO 1 -0003/$0 . 45/0 2 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 The Herbalists generally utilized the major divisions of the Theophrastian classification. There was a tendency to group together the grasslike herbs— Jun- caceae, Gramineae, and Cyperaceae — but none of the Herbalist authors erected a category which includes only grasses. Such oddities as Ustilago, Ricinus, Tri- glochin, Sparganium, Typha, and Equisetum were added from time to time. Tournefort (1700), in the early eighteenth century, listed many common grass genera, but included such interlopers as Ricinus. Jussieu (1789) was the first to apply the name Gramineae to the grasses, and to erect a group that included only grasses. His system of classification included 13 unnamed groups, based primarily on the number of stamens, styles, and florets in the spikelets. In 1833, Kunth described 13 tribes of grasses, many of them still in current usage. His treatment was followed essentially by Endlicher (1836-1840) and Steudel (1855). Bentham & Hooker (1883) published a classification for the Gramineae, dividing the family into two series, the Paniceae and Poaceae, both originally described by Robert Brown (1814). These series corresponded more or less to the two subfamilies, Panicoideae and Festucoideae, later utilized by Hitchcock. Hackel (1887), in Die natiirlichen Pflanzenfamilien, utilized an identical sys- tem, dividing the two series on the basis of the point of disarticulation. The most prevalent system of classification, especially in the Western Hemisphere, during much of the twentieth century is that of Alfred Spear Hitchcock. The Hitchcock system is roughly the same as that of Bentham and Hooker, but in the reverse order. The two large subfamilies given by Hitchcock are of very different phyletic merit. The Panicoideae, with a very few deletions, is a good natural unit, with great internal consistency in a plethora of characters. The Festucoideae of Hitch- cock, on the other hand, was a virtual morass of convergent elements, which could not be resolved by use of the traditional characteristics of the inflorescence and spikelet. It was the task of twentieth century agrostology to restudy this group and to segregate its numerous divergent elements. Early work on the cy- tology of the grasses by Avdulov (1931) and the anatomy of leaf epidermis and cross-section by Prat (1932) indicated strongly that the subfamily Festucoideae should be dismembered and its disparate elements reassigned to more coherent units. These early publications stimulated a multitude of studies on many aspects of grass anatomy, cytology, and physiology which have greatly clarified our un- derstanding of the phylogeny of the family. As a result of the new interest in grass classification, a symposium on The Natural Classification of the Gramineae Montreal We making event. In this symposium, discussions of the grass embryo by John Reed- W W Al-Aish, of grass serology by Fairbrothers and Johnson, and of general mor- phology by C. E. Hubbard, and others, were heard. A final summation by G. Ledyard Stebbins, Jr. suggested a new classification for the Gramineae (Stebbins & Crampton, 1961), resulting in a near total dismemberment of the subfamily Festucoideae. The new classification, which has since undergone minor modifi- cations, was a monumental synthesis of diverse data from many fields of inves- tigation. Current classifications usually list six subfamilies, mostly based upon 1981] POHL—SYSTEMATICS SYMPOSIUM 3 anatomical, cytological, and physiological bases, rather than the traditional inflo- rescence and spikelet characters. The new systems have greatly enhanced internal coherence and predictive value. However, the subfamilies and tribes so estab- lished may not be the most convenient for routine identification of specimens, and for this purpose, conventional artificial keys need to be designed and em- ployed. There are many occasions, however, when simple anatomical investi- gations, quickly performed on fresh or dried material with the aid of a razor blade, will immediately reveal the proper assignment of an unknown with greater certainty than inflorescence or spikelet characters. All systems of classification for the Gramineae, be they conventional or mod- ern and synthetic, rely on characteristics of extant grasses, and in that sense, we are stifl faced with the '^abominable mystery" of the origins of the Gramineae. The recent investigations of the surface anatomy of fossil grass anthoecia by Thomasson (1979), indicate that we can now trace the phyletic lines of at least some grasses (the stipoids) back to Miocene ancestors and forecast the introduc- tion of the fourth dimension of time into our understanding of the grass family. Literature Cited AvDULOV, N. P, 1931. Karyo-systematische Untersuchung der Familie Gramineen. Bull. Appl. Bot. Genet. PI. Breed. Suppl. 44: 1-428, Bentham, G. & J. D. Hooker. 1883. Genera Plantarum. Vol. 3. L. Reeve & Co., London. Brown, R. 1814. General Remarks, Geographical and Systematical, on the Botany of Terra Aus- tralis. London. G. and W. Nichol. [Vol. 2, of A Voyage to Terra Australis.] Endlicher, S. L. 1836-1840. Genera Plantarum. Beck Universitatis Bibliopolam. Wien. Hackel, E. 1887. Gramineae. In A. Eneler & K. Prantl feditorsV Die natiirlirher II, 2: 1-98. Wilhelm Engelman, Leipzig. Pflanzenfamil Hort, a. (translator). 1916. Theophrastus. Enquiry Into Plants. 2 vols. William Heinemann, London. JussiEU, A. L. DE. 1789. Genera Plantarum. Herissant et Theophilum Barrois, Paris. Kunth, C. S. 1833. Enumeratio Plantarum. Vol. 1. Agrostographia Synoptica. J. G. Cottae, Tu- bingen. Prat, H. 1932. L^epiderme des Graminees. Ann. Sci. Nat. Bot., ser. 10, 14: 117-324. Stebbins, G. L. & B. Crampton. 1961. A suggested revision of the grass genera of temperate North America. Recent Adv. Bot. I: 133-145. Univ. of Toronto Press, Toronto. Steudel, E. G. 1855. Synopsis Plantarum Glumacearum. Vol. 1. Gramineae. J. B. Metzler, Stutt- gart. Thomasson, J. R. 1979. Late Cenozoic grasses and other angiosperms from Kansas. Nebraska, and Colorado: Biostratigraphy and relationships to living taxa. Kansas Geol. Surv. Bull. 218: 1-68. Tournefort. J. P. DE. 1700. Institutiones Rei Herbariae. Ed. 1.3 vols. Typographia Regia, Paris. FRANK WALTON GOULD, 1913-1981 One of the leading agrostologists in the world and a speaker at the Twenty- Walton 11 March 1981 after an illness with brain cancer. It seems most appropriate to include a short biographical sketch in his memory with this Symposium. His widow, Lucile Gould, has kindly provided the biographical information.— Editor. Frank Walton Gould was born in May- ville, North Dakota, 25 July 1913, son of a geographer, with a brother and sister 2 and 4 years older. His third-grade year was spent at his father's hometown, Sheridan, New York, while his father completed work for his Master's Degree at the University of Michigan. From that time until he completed college, he lived where his father was even- tually head of the geography department at what is now Northern Illinois University, Frank's alma mater. He missed celebrating the 50th anniversary of his DeKalb Town- ship High School graduation by 2 months. In 1935-1936 he completed work for his Master's Degree in Botany at the Univer- sity Wiscons Dr. Norman C. Fassett, surveying the prairie remnants of Dane County. In 1941 he completed his Ph.D. at the University of California at Berkeley under Dr. Lincoln Constance, monographing the liliaceous genus Camassia. During his graduate assistant work there with Dr. G. Ledyard Steb- bins he became intrigued with the study of grasses— E/v/n//^, etc. Also there he married his 40-year partner, Lucile. His first professorship was filling a year's vacancy at St. George, Utah; then 2 years at Compton Junior College, California. In southern California he continued his taxonomic studies of grasses with the aid of Dr. Carl Epiing at U.C.L.A. In 1944 he became curator of the herbarium at the University of Arizona in Tucson. After 5 years he, his wife, and 3 daughters moved to College Station where he curated the Tracy Herbarium at Texas A&M University for 31 years, retiring as distinguished professor emeritus in 1979. However, a few years before his "retirement" he determined to pursue publi- cation of a manual of the grasses of Mexico. He had much of the research done and had a good start on the manuscript at the time of his death. In addition to some 80 publications, he authored 4 grass manuals: Southwestern USA, Texas Coastal Bend, Texas, Baja California (in press), and the well- known textbook. Grass Systcmatlcs (now being revised).— Lucile Gould. Ann. Missouri Bot. Gard. 68: 4. 1981. EVOLUTION AND DISTRIBUTION OF GRASSES W. D. Clayton^ Abstract Recent developments in grass taxonomy give a new insight into their classification, and point to a phylogenetic sequence which maps differences in their internal metabolism. Corroboratory fossil evidence is unfortunately exceedingly meager, but it can be supplemented by examining the impli- cations of present-day distributions. The subfamilies are distributed in worldwide climatic belts, but two-thirds of the genera are confined to single continents. Obviously, the genera are poor travellers, so how did the grasses become so widespread? Much depends on the probability of transoceanic transport. The evidence is inconclusive, but it seems likely that the tropical subfamilies spread during the first half of the Tertiary when the maximum water gap was 1200 km. Species distributions are likewise influenced by climatic differentiation and continental isolation. But they sometimes reveal the intervention of other factors, particularly the disruptive effect of climatic change in the Pleisto- cene. Data from the Afro-montane flora do not support the proposition that species from adjacent, but contrasting, ecological environments are distributed independently. Nor does a mapping of en- demics encourage the concept of discrete centers. Classification Fundamental to most biological disciplines is the need to identify the organ- isms they work with, and the primary purpose of taxonomy is to satisfy this need by devising a standardized nomenclature, differential diagnoses and techniques for identification. One of the powerful concepts it has developed for achieving these aims is that of hierarchical classification. In grasses this classification has traditionally been based upon similarities in spikelet structure. This provides an eminently practical system, but suffers from the defect that spikelets are subject to a good deal of parallel evolution. Yet the method should not be despised, for 80 years ago Stapf (1897) appreciated, on morphological grounds alone, that the Poeae and Eragrostideae were not partic- ularly closely related. Nevertheless, even when parallel evolution was recog- nized, there was no criterion by which the true kinship could be determined. The first revolution in classification occurred in the 1930s, when the taxonomic significance of leaf anatomy began to be appreciated. There was shown to be a major division between temperate and tropical tribes, and several subdivisions among the latter. With these findings it was possible to resolve many of the problems posed by parallel evolution, and to propose modifications with some confidence that the system was approaching closer to an expression of natural relationships. The second revolution started in the 1960s, when it was shown that differences in leaf anatomy could be correlated with differences in photosynthetic pathway. It is now becoming clear that, at the higher levels, our classification maps vari- ations in the internal metabolism of the plants. Taxonomy, which set out to devise a practical filing system for cataloging plants, has thus acquired a classification which constitutes a source of biological evidence in its own right. To unravel its meaning is inevitably speculative, but * Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AE, Great Britain Ann. Missouri Bot. Gard. 68: 5-14. 1981. 0026-6493/8 1 /0005-00 1 4/$0 1.15/0 6 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 GO N D WA NALAN D E U R AMER IC A Commelin ales ? FOR EST SAVANNA STEPPE Figure I. Suggested relationships among the major groups of grasses. A = Arundinelleae, I = Isachneae, O = Oryzeae, R = Aristideae. C4 metabolism is indicated by stippling; it is divided into the MS and PS types of Brown (1977). speculation is no bad thing provided we understand it to be no more than a working hypothesis, which gives coherence to a complex web of detail and grants some insight into the probable processes at work. Phylogeny We know, from the doctrine of evolution, that similarity between organisms is not accidental, but is due, in the first instance, to the chain of inheritance that links them together. We cannot determine this genealogical tree directly, but a good natural classification maps degrees of overall similarity, and should therefore contain sufficient information for us to deduce the main outlines of phylogeny. Let us then look at our classification in this light (Fig. 1). The origin of the grasses is unknown, but most authorities relate them to the Commelinales in general and the Flagellariaceae in particular; at the very least there seems to be agreement that the most likely relatives are to be found in the tropical forests. Now the Bambusoideae, a subfamily defined by certain anatom- ical peculiarities such as fusoid and arm cells, is both tropical and primitive. Primitive in the sense that some genera have incomplete suppression of axillary >9^n CLAYTON— EVOLUTION AND DISTRIBUTION OF GRASSES 7 buds in the inflorescence and spikelets, and the flowers may retain trimerous symmetry. Among them is a group of small tribes (e.g., Olyreae) known as the bamboo allies, which paradoxically combine the primitive and baroque. This collection of curios, apparently the relics of ancient departures from the main- stream of grass evolution, gives some hint of the diversity that must once have existed in the ancestral stock. They are mostly insignificant broad-leaved inhab- itants of the rainforest ground layer, which are often mistaken for other forest families such as Zingiberaceae. By contrast their near relatives the true bamboos have become successful competitors in this environment by developing woody tissue and adopting the form of trees. A flaw in this argument is the anemophilous pollen grain of all grasses (Page, 1978), for it seems incompatible with a forest origin. However, Oryzeae is a peripheral tribe of the subfamily, and its predilec- tion for open marshland hints that the primitive habitat might have been glades and margins rather than the forest itself. However, the future of the grasses lay, not in the forest, but in the tract between forest and desert that we know as savanna. The most likely candidates for the grasses which first moved into this environment belong to the subfamily Arundinoideae, for this is also primitive; primitive in the sense that it lacks dis- tinctive features, but seems rather to represent the lowest common denominator of subsequent subfamilies. Its notoriously difficult taxonomy suggests that it is now reduced to dismembered fragments around a core which has become extinct. Subfamily Chloridoideae is easy to place for it abuts onto Arundinoideae, and indeed the boundary between the two is difficult to establish. It ushers in a new development, the Kranz syndrome. This is a set of anatomical characters asso- ciated with C4 metabolism, in effect an extra loop in the photosynthetic cycle (the basic form of which is known as Q), that renders it more efficient in high tem- peratures (Teeri & Stowe, 1976; Tieszen et al., 1979). Subfamily Panicoideae presents more problems. It has also evolved the Kranz syndrome, in fact two different versions of it (the MS and PS types of Brown, 1977), but some genera of Paniceae still retain the older non-Kranz anatomy and C3 metabolism. Moreover, there are no direct links between the main tribes, Paniceae and Andropogoneae. The situation is still rather confused, but a ten- tative arrangement is shown in Fig. 1. Finally, subfamily Pooideae seems to represent a new venture for the grasses, an adaptation to cold climates and invasion of the temperate steppes, it has retained the primitive C3 pathway, for the more advanced form offers no advan- tage in a cool climate. This suggests a relationship with Arundinoideae, a few of whose genera have shown their ability to penetrate deeply into the temperate zone. The Time Scale Phylogeny would be simple if the sequence of events could be dated, but unfortunately the fossil record of grasses is exceedingly meager. However, there are some scraps of evidence relating to three main phases in the development of the grassland ecosystem. The broad-leaved bamboo allies are a very minor constituent of the forest 8 ANNALS OF THE MISSOURI BOTANICAL GARDEN IVol. 68 ground layer and the woody bamboos themselves are thought to have been a later secondary adaptation. Therefore, whether we regard the closed rainforest as a cradle or museum, the grasses are unlikely to have been of much significance until they established themselves under the lighter canopy of marginal woodland, and assumed something approaching the familiar narrow-leaved life form. We may tentatively associate the onset of this process with the Paleocene, when unambiguous grass pollen first appears in the record (Muller, in litt.). Thereafter the unique potential adaptability of their life form, with its basal tillering, struc- tural combination of stem and sheath, intercalary meristems and substitution of leaf sheaths for the floral envelope, enabled the grasses to move into progressively drier and more open habitats, securing an increasingly important role in the under- story of the low interrupted tropical woodland and bushland which presumably occupied this environment. Development of the distinctive savanna physiognomy was aided by fire, for fires started by lightning seem always to have been a feature of the savanna environment (Komarek, 1972); burning is certainly of long standing for the savanna trees have evolved a number of distinctive fire-resistant features. Equally certainly the grass life form is unharmed by fires which destroy many of its competitors. In fact, savanna is an ecosystem in which grass provides the fuel that ensures its own survival. The second phase of development followed the introduction of herbivores to the system. Contemporary grass-eating mammals have high-crowned teeth resis- tant to the abrasive properties of the silica bodies in grass leaves, and such teeth first appear among fossil mammals of the Middle Oligocene (Gregory, 1971; Webb, 1977). It can be inferred that grassland had emerged as a vegetation type by this time, and that grasses were undergoing further modifications to mitigate the effect of predation. The mammals responded by reciprocal adaptation, and in doing so they achieved a kind of symbiosis with the grasses, which were able to attract and sustain a level of grazing pressure sufficient to cripple their com- petitors. The Oligocene also provides the first grass fossils, some spikelets iden- tified as Stipa (MacGinitie, 1953), indicating that at last some modern genera were extant by this time. The third phase was the association of the grasslands with man, who extended and transformed them to support his grazing herds, and to provide the cereal crops upon which he relies so heavily for subsistence. This phase is documented in the archaeological record, but is not germane to the present discussion. The main point to emerge from the scanty fossil record is that grasses were in existence during the early Tertiary when the continents were in the final stages of separation, and that their history must take this factor into account. The po- sition of the continents at this time has been discussed in detail by Raven & Axelrod (1974) and Raven (1979), but the salient points may be summarized very briefly as follows. Africa and South America were some 800 km apart in the Paleocene, though there were probably intermediate oceanic islands, and moving gradually towards their present separation of 2,500 km. Madagascar-India was already separated from Africa; India parted company with Madagascar in the Paleocene, meeting Asia in the Middle Oligocene. There were probably feasible indirect routs for seed dispersal between Africa, Eurasia, North America and South America during most of the first half of the Tertiary. In addition there was 1981] CLAYTON— EVOLUTION AND DISTRIBUTION OF GRASSES 9 a direct temperate climate connection between South America, Antarctica and Australia until the Eocene (the connection between South Africa and Antarctica was severed in the late Cretaceous). Beyond these snippets of fossil evidence, we must make what we can of a kind of fossil that is still available to us, and that is the present-day distribution of grasses. The Distribution of Subfamilies Hartley (1958a, 1958b, 1973) and Hartley & Slater (1960) have produced a series of maps showing the distribution of tribes and subfamilies. Their findings may be briefly summarized as follows: Paniceae. — Tropical, humid equatorial zone; center in South America. Andropogoneae. — Tropical, seasonal rainfall zone; centers in India and South- east Asia. Chloridoideae. — Tropical, particularly the dry belts of Cancer and Capricorn. Pooideae. — Temperate, mainly northern but also in South America. Hartley took his data from a large number of published Floras, calculating the percentage contribution which each group made to the total grass flora. This has the advantage that there is no pressing need to reconcile differences in taxonomic treatments between the various Floras. It yields a useful measure of relative importance, but percentages can sometimes be misleading. Cross (1980) has pro- duced a set of maps using simple species counts, which reveal a number of discrepancies between tribal distribution measured by species abundance, and Hartley's method of relative importance. For example, the Pooideae have a high percentage figure in the north temperate zone where they are undiluted by other groups, but are actually most prolific in the Mediterranean region. She provides additional data on Bambusoideae (tropical) and Arundinoideae (southern sub- tropical). From the foregoing it is evident that the major taxa have sorted themselves into worldwide climatic belts. Most of these are more or less sympatric in the tropics, but the Pooideae are radically different, having a predominantly north temperate distribution, and the Arundinoideae display a southern subtropical, possibly relict (Darlington, 1965), pattern. In short there are no surprises, for the distribution of major taxa is entirely consistent with taxonomic prediction. Distribution of Genera To examine this problem the world was divided into 25 areas, and the grass genera scored as present or absent in each. The resulting data matrix was then sorted by cluster analysis (Clayton, 1975). Disregarding 87 genera whose distribution was essentially worldwide, there were found to be seven clusters of continental extent which together accounted for 450 genera. The remaining 98 genera were shared in various ways between adjacent clusters; some of these genera may have been naturalized introductions (known adventives were excluded from the outset), but well over half were rep- resented by different species in the two parts of their range. They linked the 10 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Voi . 68 clusters into two chains: Eurasia, North America, temperate South America; and tropical America, Africa, tropical Asia, Australia. The two chains are predictable, and strengthen the view that Pooideae are geographically distinct from the rest of the family. However, the discovery that two-thirds of the genera are confined to single continents is disturbing, for it demonstrates that genera are poor travellers, and raises the question of how the subfamilies achieved their worldwide distribution. The ability of grasses to cross short stretches of sea is not in doubt, but there is no reliable evidence either way on the subject of longer journeys. Transoceanic voyages are certainly very difficult, for most of the wind or water currents have a sufficient north-south component to carry the disseminules into a different cli- matic zone, and the problem of establishing themselves in the face of native competition is known to be formidable. Moreover the rapidity with which nu- merous weedy grasses are becoming naturalized in alien territories strongly sug- gests that these plants, though admirably equipped for dispersal, could not cross the oceans until modern commerce provided a vector. Of course, the vegetation of oceanic islands testifies that disseminules do arrive by some means or other. However, their flora is usually conspicuously unbalanced by comparison with the mainland, showing clear signs of derivation from a limited number of haphazard immigrants. There is no evidence of this phenomenon if we compare the grass floras of continents. To the contrary, they seem rather broadly based; the major subtribes of Andropogoneae, for example, show much the same pantropical distribution as the tribe itself. It can also be argued that the low probability of successful transoceanic dispersal may have been offset by the enormous time span available. This prop- osition can be partially tested, for it implies that the older tribes have the best chance of achieving a wide distribution, and should therefore contain propor- tionately more of the 87 widespread genera. In fact the reverse is true; the tribes believed to be younger contain a higher proportion of widespread genera. It seems that the distribution of the older tribes is conditioned by their rate of extinction, rather than their rate of dispersal. The inconsistency between the worldwide distribution of higher taxa, and a demonstrable reluctance of genera to cross the present day oceans, can be re- solved by invoking continental drift. It is envisaged that the tropical subfamilies spread across the world during the first half of the Tertiary, the greatest water gap being some 1,200 km across the Atlantic; formidable enough, but nothing like such a strain on plausibility as the present width of the ocean. The older Arun- dinoids were driven outwards by the greater photosynthetic efficiency, in tropical latitudes, of the newer Kranz subfamilies. In the north temperate zone they gave rise to the Pooideae, and were all but replaced by them, the Pooids subsequently dispersing southwards along mountain chains. In the south temperate zone the Arundinoideae have survived, reaching Australia (together perhaps with some cool tolerant Kranz genera) via Antarctica. During the latter half of the Tertiary genera continued to evolve, but in comparative isolation upon the increasingly separate continents. It is of interest to note that on the average widespread genera are much larger than single-continent genera, an obvious consequence of the larger area involved. 1981] CLAYTON— EVOLUTION AND DISTRIBUTION OF GRASSES 11 But they are still significantly larger when calculated on a species per continent basis, an effect presumably related to their greater age. The contrast between large and small genera suggests that two evolutionary strategies may be involved; the surviving, and by implication successful, older genera securely entrenched in their preferred environments which they exploit by complementary speciation, while the younger genera explore new and possibly labile environments by com- petitive replacement. Distribution of Species Species distributions are commonly demonstrated by means of selected ex- amples. Whether the patterns found are applicable to the bulk of the species is left open to conjecture, but there is sufficient of a probabilistic element in plant distributions to cast suspicion on small samples. It is clearly better to work with very large samples, such as the whole of the Gramineae, but this is beset by three main problems. Firstly, there is the matter of taxonomy, for little faith can be placed in results if the species themselves are in a state of taxonomic or nomenclatural confusion. Fortunately the grasses have attracted more taxonomic attention than any other family of comparable size, and we are now in a position to put together a tolerably accurate catalogue for most of the world except South America. Another facet of this problem is that it is sometimes very difficult to decide whether a species should rightfully be included as a native or not. The second problem stems from interrelationships between the distributions of species, of habitats and of communities. This is largely a matter of scale, depending on the purpose of the investigation. The present intention is not to examine the ecological sorting of species into communities under the influence of local habitat factors, but to look beyond this to the overall species pools from which these communities are recruited. The sampling unit should therefore be fairly large in order to find the major patterns, while averaging out the finer detail. It has to be assumed that a full spectrum of habitats is repeated within each of these units, a reasonable assumption insofar as the common catenary sequence of soil types is concerned, but not altogether true of unusual habitats. Thirdly, there are difficulties caused by the large number of species involved, and the often rudimentary nature of available distribution records. The problem becomes more tractable if data collection is inherently simple. The method adopted was borrowed from ecology, and consisted of scoring each species as present or absent in ''quadrats'' of country size. The resulting data matrix was first sorted by cluster analysis (a minimum spanning tree), and then refined by inspection. It was satisfying to find that so crude a method actually worked, and that the results were fairly robust even when quite large adjustments were made to the data. It was found that species are not arranged in a random mosaic, but are su- perimposed to such an extent that a limited number of generalized patterns suffice to describe their distribution; it seems likely that this is because they have become entrained with the dominant life form of the major vegetation formations. The individual distributions are not coincident but concentric, as might be expected 12 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 from the differing degrees of adaptability that exist between species. Consequent- ly the cumulative patterns (phytochoria) have no definite boundaries, but fade away gradually at the edges, and are best mapped by some form of contouring. They also tend to overlap each other to a greater or lesser extent, an effect which may be exaggerated among grasses, which have a great capacity for infiltrating the serai stages of neighboring formations from which competition would other- wise exclude them. Nevertheless, it is important to appreciate that phytochoria do not have mutually exclusive territories in the manner of ecological commu- nities; indeed, the latter are often synthesized from the members of several trans- gressive floristic elements. Another feature of species distributions is that two or more subsets may rest within the geographical limits of a wider pattern, thus affording the basis for a heirarchical classification. Three principal categories are usually recognized: Kiniidom. Generally corresponds to the major land masses, and reflects the effect of continental isolation. Reijion. Generally corresponds to the major vegetation formations, and reflects the effect of climatic differentiation. Domain. In many ways the most interesting category, for it often bears little relation to ecological factors, but seems instead to represent a legacy of dis- equilibrium left by past historic events, notably the disruptive effect of cli- matic changes during the Pleistocene. Regional studies have been made by Clayton & Hepper (1974), Clayton & Panigrahi (1974) and Cope (1977). These have subsequently been consolidated in a series of maps showing the distribution of grass species throughout the Old World (Clayton & Cope, 1980a) and in North America (Clayton & Cope, 1980b). THE AFROMONTANE GRASSES In working out the distribution of species it has been tacitly assumed that the floras of contrasting, but adjacent, habitats are not distributed independently. This proposition can be tested on data from the African mountains, for it is widely held that the flora of these high islandlike habitats is distributed differently from that of the surrounding lowlands, and constitutes a distinct Afromontane Region. However, when the grasses of the highland Afromontane Region (Clayton, 1976) and lowland Ethiopian Domain (Clayton & Cope, 1980a) are compared, they are found to have much the same geographical extent. There is thus evidence to suggest that the factors determining the present limits of these two floras have been broadly similar, despite the manifest difference in habitats, and that the geographical case for separating them is weak. Separation of the Afromontane Region has been strongly influenced by the taxonomic affinity of its flora to temperate genera, implying that its original source was quite different from that of the surrounding tropical lowland flora. Presum- ably its precursors entered from the north. The precursors of the South African montane flora, of similar taxonomic affinity, must have passed along the same route, but the two floras are now represented by quite different species. They have evidently been evolving in situ for a considerable time and their ultimate 1981] CLAYTON— EVOLUTION AND DISTRIBUTION OF GRASSES 13 origin is not particularly relevant to their present distribution. Although evidence from taxonomic relationships is a valuable aid to understanding the past, it is often associated with an extended time span. We should therefore be wary of accepting it in the same context as the, largely Pleistocene, events which have shaped the boundaries of present-day phytochoria. ENDEMIC SPECIES Endemic species have always had a particular fascination for phytogeogra- phers, and they are frequently invoked as a guide to the center of origin of genera of even whole floras. However, when endemics, in the sense of species confined to a single country, are plotted on a map (Clayton & Cope, 1980a), they are found to be surprisingly common. Evidently the notion of a limited number of compact endemic centers cannot be entertained. In fact, endemics are most abundant in the southern tips of continents, where isolation is probably a major factor. But they are also frequent in all the major orogenic zones (the flat pediplain of the Congo-Zambezi watershed is a notable exception), and there is a low background count almost everywhere. It seems that we are confronted with two contrasting environments. The mountains offer a great variety of niches, and can accommodate a wealth of species; moreover, they are buffered against the worst effects of climatic change, for the vegetation can adjust by shifting its altitudinal zonation upward or down- ward. On the other hand, the plains offer a relatively uniform environment where evolution proceeds by competitive replacement, and where climatic change en- tails a major disturbance. In short, the distribution of endemics gives little support to simplistic inter- pretations in terms of centers of origin or refugia. Certainly the species-rich moun- tains may constitute a valuable reservoir of genetic diversity, but it is likely that evolution has proceeded just as rapidly on the plains, though leaving no relicts to mark its course. Conclusion The history of the grasses is not a simple matter, for it is compounded from the evolutionary thrust of competition, selection pressures exerted by the envi- ronment, the degree of isolation imposed by shifting geographical configurations, and disruptive migrations induced by climatic change. Nor is it amenable to a wholly deterministic approach, for many of these factors operate in a probabilistic fashion. Nevertheless, I have tried to show that a coherent story can be pieced together from existing knowledge of taxonomy and distributions. I should em- phasize that it is but a working hypothesis, for there are all too many gaps that must, for the moment, be bridged by conjecture rather than fact. Literature Cited Brown, W. V. 1977. The Kranz syndrome and its subtypes in grass systematics. Mem. Torrey Bot Club 23(3): 1-97, Clayton, W. D. 1975. Chorology of the genera of Gramineae. Kew Bull. 30: 111-132. . 1976. The chorology of African mountain grasses. Kew Bull. 31: 273-288. 14 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 171. & T. A. Cope. 1980a. The chorology of Old World species of Gramineae. Kew Bull. 35: 135- & . 1980b. The chorology of North American species of Gramineae. Kew Bull. 35 567-576. — & F. N. Heppfr. 1974. Computer-aided chorology of West African grasses. Kew Bull. 29 213-234. & G. Panigrahl 1974. Computer-aided chorology of Indian grasses. Kew Bull. 29: 669-686. Cope, T. A, 1977. Computer-aided chorology of Middle Eastern grasses. Kew Bull. 31: 819-828. Cross, R. A. 1980. Distribution of sub-families of Gramineae in the Old World. Kew Bull. 35: 279- 289. Darlington, P. J. 1965. Biogeography of the Southern End of the World. Harvard Univ. Press, Cambridge, Massachusetts. Gregory, J. T. 1971. Speculation on the significance of fossil vertebrates for the antiquity of the Great Plains of North America. Abh. Hess. Landesamtes Bodenforsch. 60: 64-72. Hartley, W. 1958a. Studies on the origin, evolution and distribution of the Gramineae. I, The tribe Andropogoneae. Austral. J. Bot. 6: 116-128. . 1958b. II, The tribe Paniceae. Austral. J. Bot. 6: 343-357. . 1973. V, The sub-family Festucoideae. Austral. J. Bot. 21: 201-234. & C. Slater. 1960. Ill, The tribes of the sub-family Eragrostoideae. Austral. J. Bot. 8: 256- 276. KoMAREK, E. V. 1972. Lightning and fire ecology in Africa. Proc. Annual Tall Timbers Fire Ecol. II: 473-511. MacGinitie, H. D. 1953. Fossil plants of the Florissant beds, Colorado. Publ. Carnegie Inst. Wash. 599: 1-198. Page. J. S. 1978. A scanning electron microscope survey of grass pollen. Kew Bull. 32: 313-319. Raven, P. H. 1979. Plate tectonics and southern hemisphere biogeography. Pp. 3-24, K. Lars & L. B. Holm-Nielsen, Tropical Botany. Academic Press, New York. & D. I. AxELROD. 1974. Angiosperm biogeography and past continental movements. Ann. Missouri Bot. Gard. 61: 539-673. Stapf, O. 1897. Gramineae. In W. T. Thiselton-Dyer, Flora Capensis. Vol. 7: 310-750, 760-765. Teeri, J. A. & L. G. Stowe. 1976, Climatic patterns and the distribution of C4 grasses in North America. Oecologia 23: 1-12. TiESZEN, L. L., M. M. Senyimba, S. K. Imbamba & J. H. Troughton. 1979. The distribution of C;j and C4 grasses and carbon isotope discrimination along an altitudinal and moisture gradient in Kenya. Oecologia 37: 337-350. Webb, S. D. 1977. A history of savanna vertebrates in the New World. I. North America. Annual Rev. Ecol. Syst. 8: 355-380. SOME EVOLUTIONARY TRENDS IN THE BAMBUSOIDEAE (POACEAE)^ Thomas R. Soderstrom^ Abstract Bamboos, which have woody culms, and several genera with herbaceous culms share a similar type of leaf anatomy and epidermis. Various other morphological features, including a distinctive type of seedling, also indicate a close relationship that supports the grouping of these genera together into a single subfamily, the Bambusoideae. A review of the chromosome numbers in the subfamily reveals that the herbaceous members are mostly diploid while the woody ones are polyploid, with jc = 12 the basic number. Such evidence corroborates the hypothesis that bamboos have been derived from herbaceous ancestors. Most bamboos flower infrequently and have had far less opportunity for floral evolution than have the herbaceous members, which flower continuously or at least seasonally. The most primitive inflorescences have therefore been retained in the bamboos while highly special- ized ones have developed in the herbaceous members. A knowledge of the former type of inflores- cence is useful in an interpretation of the latter. With this in mind, the inflorescence of the herbaceous bambusoid grass, Streptochaeta, has been reexamined. This genus has long been considered to be the most primitive grass, in great part due to the presence of three large structures in the spikelet, thought to be primitive lodicules and two structures interpreted as a primitive, two-parted palea. Comparison of Streptochaeta with other members of the Bambusoideae suggests that the spikelet in fact lacks both lodicules and palea and that these structures represent instead bracts on different axes of a highly modified pseudospikelet. Such a pseudospikelet is comparable to that of a bamboo. While Streptochaeta may be considered primitive in its herbaceous nature and possession of a pseudospike- let, it must be regarded as advanced in other features, among them the lack of lodicules. No one member of the subfamily can be considered most primitive. The least advanced inflorescences are retained in bamboos, such as Bamhusa, while the most primitive growth form occurs in herbaceous genera like Streptochaeta, Streptogyna, and Pharus. The significant trends of evolution within the subfamily become apparent, however, only when all of the genera — woody and herbaceous — are considered together. The bambusoid line — with its complex leaf anatomy and epidermis — is itself specialized and not to be regarded as the precursor of the other groups of grasses. The major, large, natural groups of grasses can be determined by relatively few factors that are of basic importance in the family, among them the internal structure of the leaf and features of the leaf epidermis and embryo. Other char- acters, such as chromosome number and seedling type, in correlation with these result in the recognition of six or seven major groups that most agrostologists agree upon as being natural. These are commonly known as the arundinoid (phragmitoid), bambusoid, centostecoid, chloridoid (eragrostoid), oryzoid, pani- coid, and pooid (festucoid). Often these groups are called subfamilies. Odd genera here and there have sometimes been considered of equal importance so that in the literature we find whole subfamilies based upon single genera: Anomoch- looideae and Micrairoideae are examples. The bambusoid group is best known by its Gargantuan members, the tree grasses, such as Dendrocalamus giganteus of northern Burma, whose culms may * The illustrations, except for Fig. 7, were prepared by Alice R. Tangerini, to whom I am indeed indebted for her skillful rendition of a multitude of new and often difficult structures. Figure 7 was drawn by Mrs. Gesina Berendina Threlkeld as part of a series of illustrations of Ceylonese bamboos, to be published as a unit at a later date. 2 Department of Botany, Smithsonian Institution, Washington, D.C. 20560. Ann. Missouri Bot. Card. 68: 15-47. 1981. 0026-6493/8 1/00 1 5-0047/$3 .45/0 16 ANNALS OF THE MISSOURI BOTANICAL GARDEN |VoL. 68 reach 30-35 m in height and 25-30 cm in diameter. The clumps of such bamboos reach enormous proportions and the plants are known for their propensity to continue growing vegetatively, even for decades, before they flower. Size, com- plexity of body, and rarity of flowering have contributed to the neglect that bam- boos have received from systematists and the taxonomic confusion that has long plagued them. Not all bamboos, however, are so large as DenJrocalamus. At the other end of the scale we find Arundinaria py^maea, whose wiry culms reach no more than half a meter in height, but like the bigger members also branch at the nodes and flower seldom. Regardless of size, both Dendrocalamus and Arundinaria share a number of features that are common throughout the bamboos: a specific type of leaf anatomy and epidermis; distinctive seedling; flowers with three lodicules, often six sta- mens, and three stigmas; and fruit with a small embryo and linear hilum. Many other grasses of small stature that inhabit shaded forests also possess these basic features, which demonstrate a natural relationship. Besides their smaller stature, they do not have such complex branching as the bamboos nor such long-lived culms, and they commonly flower throughout the year or at least seasonally. These grasses, with the bamboos, constitute the subfamily Bambusoideae. We refer to the one group of Bambusoideae as ^ ^herbaceous bambusoid grass- es" and to the other as "woody bambusoid grasses," or simply bamboos. When we speak of ^^woody," we mean this in the sense of being hard, not like the stem of a dicotyledonous tree that produces secondary xylem. The bamboo culm is characterized by collateral bundles embedded in parenchymatous tissue, with the bundles toward the inside more separated from one another than those toward the periphery, which are smaller and occur very close together. A cap of fiber strands usually occurs on both sides of the bundle, and this abundant scleren- chyma in the culm, along with thick-walled and lignified ground tissue, accounts for the hardness of the culm. Some additional strengthening may be contributed by the silica that is present in the epidermis. Detailed studies on the anatomy of the bamboo culm have been carried out by Grosser and collaborators (Grosser, 1970; Grosser & Liese, 1971; Grosser & Zamuco, 1971). All bambusoid grasses grow in association with woody vegetation and are never components of prairies or grasslands, being most common in temperate woodlands or tropical forests or, if herbaceous, in the shaded understory of warm forests. They are usually dependent on humidity, at least during the growing season; those few that grow in drier regions or areas with a dry season may lose their leaves during this period. Bambusoid grasses are most abundant in the world's tropics and subtropics, but a few members are found in temperate-cold areas of both hemispheres. They occur between 46"" north latitude and 47° south latitude and from sea level to as high as 4,000 m elevation, and are found on all continents except Europe. The bambusoid grasses represent the most widespread and diverse assemblage of genera and species within the family. So distinctive and natural is the group that Tzvelev (1976), in his recent system of classification of grasses, recognized it as one of his two subfamilies, the other being the Pooi- deae. Following are some of the basic characteristics of the Bambusoideae, a more detailed elaboration of which is given in Calderon & Soderstrom (1980). 1981J SODERSTROM— BAMBUSOIDEAE 17 Bambusoideae General morphology: perennials, herbaceous or woody, rhizomatous; culms when woody branched at the nodes; leaf blades usually flat, broad, lanceolate or linear-lanceolate, articulate with the sheath by a petiole that orients the blade in different positions; blades with tessellate venation strongly or weakly manifest; flowering seasonal or occurring infrequently; inflorescences of different types, often of complex systems of partial inflorescences of limited branching or unlim- ited branching and production of spikelets or pseudospikelets; spikelets or pseu- dospikelets 1-many-flowered, without glumes or with 1-several ''transitional'' glumes; lemmae 3-many-nerved, awnless or only seldom awned and then the awn not geniculate; palea 2-many-nerved, keeled or rounded dorsally, excep- tionally bifid; lodicules generally 3 (0-6, rarely many), usually large and with hairs of different types and well-developed vascularization; stamens 3-6 (rarely 2, occasionally many), sometimes partially fused or monadelphous; stigmas 2 or 3; fruit usually a caryopsis, sometimes an achene or fleshy; hilum linear, almost as long as the fruit; embryo small in comparison to the fruit. Seedling: coleoptile usually short and not elevated from the caryopsis; first one to several leaves usually bladeless or with a reduced blade; first expanded blade usually broad, ovate or lanceolate, horizontal in position. Leaf anatomy: leaf blade with a conspicuous midrib containing a complex vascular system of several bundles in 2 rows, strongly developed sclerenchyma and ground tissue; mesophyll with cells arranged in horizontal layers parallel to the epidermis, not radiate; chlorenchyma composed of arm cells and translucent fusoid cells at each side of the vascular bundle and in between layers of arm cells; bundle sheaths always double and well developed, the outer sheath with very few chloroplasts; transverse veinlets connecting the longitudinal vascular bundles. Leaf epidermis: with short cells in pairs or sometimes in rows over the veins; silica bodies usually cross shaped, saddle shaped, of olyroid type or of interme- diate forms; microhairs nearly always present, bicellular with both cells of about the same length and with a rounded apex, or sometimes of 3 or 4 cells, papillae common and abundant on the long cells and overarching the stomata; long cells with thick, sinuous walls; stomata usually with low dome-shaped or sometimes triangular subsidiary cells. The herbaceous bambusoid grasses, which are fewer in number than bamboos but have received more attention in recent years, are more clearly understood than the bamboos. We are able to delimit the genera of herbaceous bambusoid grasses and are able to recognize related groups of genera, or tribes. While this is also true for some bamboos, for the most part generic limits are still poorly understood and consequently tribes are not yet recognized for most. We presently recognize eight tribes of herbaceous bambusoid grasses, the following six occurring in the American tropics: Anomochloeae, Olyreae, Pari- aneae, Phareae, Streptochaeteae, and Streptogyneae. Only two of these are rep- resented outside of the Americas. The Phareae contains two genera, Pharus of the American tropics and Leptaspis of the Old World tropics, and the Strepto- gyneae is monotypic, with one species of Streptogyna occurring in the New 18 ANNALS OF THb: MISSOURI BOTANICAL GARDEN [Vol. 68 World and one species in Africa, southern India and Ceylon. The other American tribes vary in size, with the Anomochloeae and Streptochaeteae each monotypic (Ano/nochloa and Streptochacta), the Parianeae with two genera {Eremitis, Par- iana), and Olyreae with 15 {Arherella, Cryptochloa, Diandrolyra, Ekmanochloa, Froesiochloa, Lithachne, Mnclurolyra, Mniochloa, Olyra, Piresia, Raddia, Rad- diella. Rchia, Rcitzia. Strephium). A key to the herbaceous American tribes appears in Calderon & Soderstrom (1980). The tribe Atractocarpeae is represented only in Africa, and contains the gen- era Guaduella and Puelia, while the monotypic tribe Buergersiochloeae is known only from the genus Buergersiochloa in New Guinea. The herbaceous bambusoid grasses are most abundant in the New World where they reach their greatest diversification and comprise well over one hundred species, many still unde- scribed. There are 17 described genera of bamboos in the New World alone, of which only two occur in Asia, Arundinaria and Bamhusa. The others are all endemic to the American continent. An equal or greater number of bamboo genera occur in the Old World, with many endemic to Madagascar, surpassing the number that occur in all of Africa. The trends of evolution in the subfamily become apparent only when we study all the genera together, herbaceous and woody. Previous systems of classification have often obscured these patterns, for the woody genera were placed in a single tribe, Bambuseae, and the herbaceous genera were scattered in widely unrelated tribes, such as the Hordeae and Paniceae, that pertain to other subfamilies. A survey of many bambusoid seedlings shows that they are of an unique type, with certain features not found in the seedlings of other grasses. The fact that the seedlings of both herbaceous and woody members are similar strengthens the argument that these genera are closely related, which corroborates the conclu- sions based on studies of the leaf anatomy and epidermis. The development of the woody habit and emphasis on vegetative reproduction in the bamboos has been accompanied by a decrease in flowering. Most bamboos bloom only at long intervals, these sometimes as much as 120 years. This has in effect prevented gene interchange in bamboos and arrested evolution of the in- florescence. On the other hand the herbaceous members, which bloom throughout the year or at least seasonally, in comparison with the bamboos have had count- less generations of flowers and continuous opportunity for evolution of the flow- ering system. Among the herbaceous members, we therefore encounter highly specialized inflorescences and spikelets while in the bamboos we find them to be more primitive and less specialized. The herbaceous bambusoid genus, Streptochaeta, long considered to be the most primitive of all grasses, can now be reinterpreted in light of the above ideas. A study of its seedlings confirms that the genus is bambusoid. A knowledge of chromosome numbers in the subfamilies suggests that the somatic number of In = 22 in this genus represents a diploid based on a derived basic number. Analysis of the so-called spikelet of the genus and comparison of it with the inflorescence of a primitive bamboo reveals that the flowering unit is in fact a highly modified pseudospikelet. While Streptochaeta does indeed have primitive features, our study of other 1981] SODERSTROM— BAMBUSOIDEAE 19 bambusoid seedlings and chromosomes shows that two other genera must also be primitive. These genera, Streptogyna and Pharus, have peculiar, apparently undifferentiated, seedlings, and are diploids, based on the primitive number of X = 12. Their present-day distribution also points to an archaic existence, for one snecies of StreotoQvna occurs in the New World and one in the Old World. World World The herbaceous bambusoid genus, Puelia, which is endemic to Africa, also is a diploid based on ,v = 12. We thus find a concentration of primitive herbaceous types in present-day tropical Africa, with the most highly specialized herbaceous World World polyploid bamboos, however, are found in Asia. The Bambusoid Seedling (Figs. 1-3) Studies of the embryo in grasses have been so numerous and interpretations of the component structures so diverse that a voluminous literature exists. The reader is referred to Brown (1960), who reviewed the major papers and interpre- tations of each author. I do not, however, agree with the conclusions that he personally drew. His paper was followed in close sequence by at least two others that do not agree with his findings either: these are Guignard (1961) and Negbi & Roller (1962). The grass embryo is of special interest because it has unique structures whose homologies have been the subject of much debate. Most studies have been made on cultivated grasses such as wheat {Triticum) and maize {Zea), an unfortunate situation as these grasses have been so modified by man. I agree with Reeder (1953) who felt that more primitive grasses should be used for study of these structures; he himself chose one, Streptochaeta, for his study of the coleoptile. The grass embryo, together with the endosperm, and surrounding wall struc- tures comprise the grass fruit, which is generally a caryopsis. In rare, specialized cases, such as the bamboo genus Melocanna (Stapf, 1904), the endosperm is lacking in the mature fruit and the pericarp, along with the scutellum, becomes fleshy. We can imagine the grass plant in its earliest stages as an axis, the basic part being the embryonal axis with the main seat of differentiation the node where the scutellum is attached. We may call this the first node of the grass plant and it occurs within the embryo. The axis continues upward and the next node marks the origin of the coleoptile. These two nodes and the internode between them are important in our discussion as these structures are unique to grasses and not really comparable to embryos of other monocotyledons. The scutellum is a flat organ that is specialized for absorption of nutrients from the food reserve or endosperm. It remains within the fruit as the other parts become visible at germination. Opposite the scutellum there is sometimes present, as a nonvascularized outgrowth, a structure called the epiblast. The coleoptile is a sheathing structure and is generally closed; it protects the growing apex of the embryo, which it completely encloses. At germination the growing apex, or plu- mule, pushes through it. 20 ANNALS OF THE MISSOURI BOTANICAL GARDEN |Vol. 68 Below the point of attachment of the scutellum is an elongated zone of tissue, the radicle, enclosed in a sheathing structure called the coleorhiza. The coleorhiza terminates in suspensor cells. The scutellum is generally accepted to be a cotyledon. For this reason the node where it is attached is called by some the '^cotyledonary node," but I prefer the term ^^scutellar node/' as employed by Avery (1930), as it is descriptive rather than interpretive. By the same reasoning Avery's term, "coleoptilar node," for the next node is appropriate. According to some authors (Negbi & Roller, 1962), the epiblast represents a second cotyledon. However, it may be nothing more than an outgrowth of the coleorhiza and, furthermore, it is not always present. For practical purposes, since a scutellum and coleoptile are always present, let us use them as markers of the first and second nodes of the embryonal axis, the region in between being the first internode of the plant's axis as accepted by Avery. This internode is somewhat specialized, as we might expect, and each succeeding internode becomes anatomically more like those of the culm. Since the time of Celakovsky (1897), this region has often been referred to as the mesocotyl, an interpretive designation that relies on a theory that regards this as an elongated node belonging neither to epicotyl nor hypocotyl. The plumule, enclosed within the coleoptile, consists of numerous nodes and internodes, the nodes bearing the embryonic leaves. Its axis is a continuation of the embryonal axis; its first internode, the one just above the coleoptile (which is the second internode of the axis), is called a "transition internode'' by Avery since it has an anatomical structure intermediate between that of the first inter- node and the next higher one, the third. Upon germination the plumule elongates and pushes through the coleoptile, which is a closed sheathing structure that is generally two-nerved. The second internode may elongate greatly at this time, elevating the remainder of the plumule. At the same time the radicle elongates and pushes through the coleorhiza. Some authors consider the coleorhiza to be a modified primary root, especially since it terminates in suspensor cells, which are characteristic of that organ. If the coleorhiza is indeed homologous to the primary root, and this seems reasonable to me, the structure normally referred to as ''primary root'' is an adventitious one. To avoid the use of an interpretive term in identifying this structure that emerges through the coleorhiza, I am using the word ''radicle." The structure of the embryo, its size relative to the endosperm, structure of the starch, and type of seedling have all provided useful characters in grass systematics. The most important papers on this subject, especially the embryo, are those of Yakovlev (1950) and Reeder (1957, 1962). Numerous publications by Kuwabara, commencing with his 1960 paper in English, have shown the systematic significance of grass seedlings, a topic dis- cussed more recently in detail by Hoshikawa (1969). The latter author studied over 200 grass species in 88 genera and found that, as with embryo types, seedling types could also be used to delimit major natural groupings within the family. Unfortunately, the terminology of Hoshikawa is not clear and is sometimes misleading. For example, the coleoptilar node (as I am using it) is referred to by him as the "cotyledonary node," a term often used for the first or scutellar node. 1981] SODERSTROM— BAMBUSOIDEAE 21 Figure 1. Seedlings of Bambusoideae. aff Puntarenas Linhares, Soderstrom & Sucre 1901,— h. Embryo of Streptochaeta spicata just beginning to emerge . Streptochaeta spicata, same collection as b.— d. Lithachne pauciftora, Puerto Rico, Mayagiiez, Soderstrom 1801. —e. Detail of germinating fruit of Lithachne pauciftora, same collection as d.— f. Ochlandra stridula, Ceylon, near Anandara, Soderstrom 2563. 22 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 For the scutellar node he coins the term, "transitionary node." In spite of these drawbacks, his scheme is interesting and of value. He points out differences that are quite useful in differentiating seedlings, such as points of origin of adventitious roots and their relative rates of development, and elongation or suppression of the first internode (his "mesocotyl") and second internode. Based on the study of seedlings of seven bamboo species, Hoshikawa (1969) defined the bambusoid seedling as follows, his terminology following mine in parentheses: the first internode (mesocotyl) does not elongate, adventitious roots are lacking from both the scutellar (transitionary) node and coleoptilar (cotyle- donary) node, and the blades of the leaves (leaves) from the lowest nodes are entirely suppressed or are only weakly developed. Unaware of this paper, Calderon and I (1973) defined the bambusoid seedling on the basis of an herbaceous genus, Maclurolyra. We stated that the coleoptile is short and not elevated above the caryopsis by an internode (i.e., the first internode does not elongate), the first two to several leaves lack blades or the blades are reduced, and the first expanded blade is broad, ovate -lanceolate, and horizontal in position. We also pointed out the similarity of bambusoid to oryzoid grasses in the presence of reduced blades in the first leaves of the new shoot, but that in the latter group the first expanded blade is linear and ascending rather than horizontal in position. I have now examined the seedlings of several more herbaceous and woody bamboos, including the genera Streptochaeta, Streptogyna, and Pharus, all of which belong to different tribes and have been considered as archaic and related to bamboos. All of the seedlings were collected from beneath parent plants in the field, with the exception of Streptogyna. In that case I collected the fruits in Espirito Santo, Brazil, on 16 March 1972, and planted them in pots at the Jardim Botanico in Rio de Janeiro on April 13. Seedlings were collected at intervals of about one month and preserved in FAA (90 parts of 50 percent EtOH : 5 parts glacial acetic acid : 5 parts formalin). The following descriptions of the seedlings (whether of herbaceous or woody genera, as indicated) may be compared with the illustrations that appear in Figs. 1-3. SEEDLING DESCRIPTIONS 1. Arundinaria gigantea (Fig. 2f) woody A short coleoptile is followed by two sheaths and a leaf with a small ovate, more or less horizontal, blade. 2. Aulonemia aff. aristulata (Fig. la) woody The first and second internodes do not elongate; the coleoptile remains very short, followed by two short sheaths, a longer sheath, and the first leaf with a small, oval horizontal blade. 3. Diandrolyra sp. (Fig. 2e) herbaceous The short coleoptile is followed by two sheaths, and then a leaf with a small, horizontal, ovate blade. ifiora (Figs. Id-e) herbaceous In one seedling the coleoptile is short, followed by two sheaths, and a leaf with an ovoid-lanceolate blade horizontal in position. In the other seedling an 1981] SODERSTROM— BAMBUSOIDEAE 23 Figure 2. Seedlings of Bambusoideae. — a. Pharus sp., Brazil. Bahia, Calderon 2171. — b-c. Raddia sp., Brazil, Rio de Janeiro, Soderstrom, Sucre, & Calderon 1858. Olyra loretensis. Colombia, Leticia, Soderstrom 1429 Diandrolyra sp., Brazil, Rio de Janeiro, Soderstrom & Sucre 1935.— f. Arundinaria gigantea, USA, Maryland, McClure, bamboo garden introduction no. 2762, 24 ANNALS OF THE MISSOURI BOTANICAL GARDEN (Vol. 68 elongated portion precedes a node bearing some roots that pierce the coleoptile that covers it. Here it is the second internode that has elongated. The node bearing the first leaf is covered by the elongated coleoptile. 5. Ochlandra stridida (Fig. If) woody The coleoptile is short, followed by two sheaths and then several closely overlapping sheaths, each bearing a small, more or less horizontal-ascending blade. Germination of a bud, apparently in the axil of the coleoptile, gives rise immediately to a second shoot similar to the first. Quick germination of further buds produces a miniature clump of new shoots at the seedling stage. 6. Olyra lore tens is (Fig. 2d) herbaceous The short coleoptile is followed by two sheaths without blades; the following leaf bears a small, ovate, more or less horizontally positioned blade. 7. Pharus sp. (Fig. 2a) herbaceous The first internode does not elongate, the second hardly so or sometimes extending for some length, with the coleoptile elongating concomitantly. The third node produces a leaf with expanded blade, which is broad, ovate, and horizontally positioned, but without reduced blades preceding it. 8. Raddia sp. (Figs. 2b-c) herbaceous The short coleoptile is followed by one leaf with a tiny erect blade and a second one with the first expanded blade, which is broad-lanceolate and hori- zontal. 9. Strcptochaeta spicata (Figs. Ib-c) herbaceous The first internode does not elongate; the second internode is very short. There are up to three leaves without blades or with reduced blades that precede the first leaf with an expanded one, which is broad and horizontally positioned. The short coleoptile is 5-nerved. 10. Streptogyna americana (Figs. 3a-n) herbaceous The first internode does not elongate; the second internode elongates, as does the coleoptile. The third node produces a root and leaf with narrow, ascending, fully developed blade. There are no reduced blades nor is the first blade broad and horizontal. The basal portion of the coleoptile is thickened and appears to be fused to the second internode but careful dissection reveals them to be free from one another. There are two strong nerves in the coleoptile and sometimes a faint line that appears to be a median nerve, although this may only be an artifact. However, the coleoptile is mucronate, a condition not found in any of the other bambusoid grasses I have studied. Figure 3. Fruit and seedling of Streptogyna americana. — a. Caryopsis showing relation of embryo (at base) to endosperm. — b. Caryopsis showing long linear hilum. — c. Base of caryopsis enlarged to show embryo. — d. Seedling 5-6 weeks old breaking through covering bracts, showing coleoptile (co), radicle (r), and first leaf (1,). — e. Opposite side of d showing seedling in relation to the persistent rachilla internode. — f. Seedling 8 weeks old. — g. Detail of fat region of embryo emergence, showing the coleoptile, epiblast (ep), coleorhiza, radicle, and scutellum (s), covered by the pericarp. — h. Enlargement of g showing coleorhiza (cr). — i. Seedling 3 months old. The mature elongated style is shown here, but broken off or shortened in the other drawings. — j. Upper part of coleoptile off. (x6). — k. Detail of i at region of embryo emergence. — 1. Detail of i at region of embryo emergence, from side of persistent rachilla internode (ri). — m. Seedling 4 months old showing 3 developed leaves. — n. Ligule of first leaf of f. [Drawings a and b based on Swollen 5089 from Obidos, Para, 1981] SODERSTROM— BAMBUSOIDEAE 25 Brazil. The seedlings are all taken from plants cultivated at the Jardim Botanico, Rio de Janeiro, Bra- zil, from fruits collected by Soderstrom & Sucre 1906 in Brazil, Espirito Santo, Reserva Florestal de Linhares, 16 March 1972. All fruits were planted 13 April 1972: d was removed at the end of May, f on June 14, i on July 16, and m on August 16.] 26 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 BAMBUSOID SEEDLING DEFINED On the basis of published findings and the additional seedlings described here, we can make the following definition of the bambusoid seedling: The first internode does not elongate; the second internode is usually short with the coleoptile remaining short, but occasionally it elongates with the co- leoptile elongating as well; the first node or first several above the coleoptilar node bear leaves that lack a blade or the blade is reduced; the first expanded blade is usually broad, ovate-lanceolate and positioned horizontally; adventitious roots are usually lacking but may occur at both the scutellar and coleoptilar nodes; a bud may be produced at the coleoptilar node. While this is a generalized description, we find that the seedlings of two genera in our study depart from this. Pharus and Streptogyna both lack reduced blades; in the former the first blade is expanded, large and ovate, and in the latter it is narrow and ascending. The principal features of bambusoid grass seedlings then are the nonelongation of the first internode and general lack of roots at the scutellar and coleoptilar nodes. In the oryzoid and pooid grasses, for example, there is an elongation of the first internode and production of roots at the coleoptilar node. V Chromosomes (Fig. 4) Chromosome counts in the Bambusoideae are by no means numerous, yet there is a sufficient number to allow us to discern some general patterns and form some postulations. The following account is based on data from the literature and from counts made on plants that we have collected in the field or have in culti- vation. In each case I am citing the single reference for the count, or a represen- tative one when there are more. The differences in somatic numbers in the Bambusoideae are great, ranging from 2n = 14 in Olyra fasciculata (Morisset, in litt.) to 2« =72 in Dendrocahi- mus giganteus (Gould & Soderstrom, 1974). That the former is an herbaceous member and the latter a gigantic woody one is particularly interesting as, in general, the lower, diploid numbers are found in the herbaceous genera and higher, polyploid numbers in the woody. A basic number of x = 12 is found in a few of the herbaceous bamboos and the tetraploid complement, 2fi = 48, among many of the woody bamboos. In diploid form, 2n = 24, this number occurs in genera of three tribes of herbaceous bambusoid grasses: Streptogyna (Kammacher et al., 1973) of the Streptogyneae; Leptaspis (Tateoka, 1958) and Pharus (Pohl & Davidse, 1971) of the Phareae; and Puelia (Dujardin, 1978) of the Atractocarpeae. Hsu (1972) gave this number for two woody bamboos from Taiwan, Bamhusa oldhamii and B. stenostachya. The tetraploid number of 2/7 =48 occurs throughout the bamboos in widely unrelated taxa, such as Anmdimiria gigantea (Gould, 1960) from the United States, Neurolepis (Gould & Soderstrom, 1970) from Andean South America, Chimonohamhusa (Mehra & Kalia, 1976) from the Himalayas, Indocalamus (Jan- aki Ammal, 1945) of Ceylon, and Shihataea (Okamura & Kondo, 1963) from Japan. Hexaploids oi 2n = 72 have been recorded for several woody bamboos, es- pecially those largest in stature, including Oxytenanthera ahyssinica (Reeder & 1981] SODERSTROM— BAMBUSOIDEAE 27 E 3 n u 3 E re o (0 J5 ^ (A re o o o re c ^ T5 re Q. c O) 0) 0) ■ ^^ OQO o X) E PQ 4> s: (A X) E E o O E o U LU 28 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Singh, 1967) of Africa and Ochlandra (Janaki Ammal, 1945) from India. However, there is a noticeable concentration of hexaploids in the mountains of northern Burma and western China. The hexaploid bamboos that are native to this region include Bamhusa polymorpha (Janaki Ammal, 1945), B. tulda (Christopher & Abraham, 1971), Cephalostachyum pergracUe (Sarkar et al., 1977a), Dendroca- lamus hmndisii (Janaki Ammal, (Mehra Sharma, 1972), D. longispcithus (Janaki Ammal, 1945), D. strictus (Richarria & Kotval, 1940), and Gigantochloa macrostachya (Sarkar et al., 1977b). The hexa- ploid bamboos from this region include the largest species, Dendrocalamus gi- ganteus, and some of the most widely planted and economically important, such as Bamhusa tulda and B. polymorpha. It is interesting to point out that Janaki Ammal (1950, 1954) found this region to contain the highest polyploid races oi Rhododendron and Buddleia. Darlington (1973) speaks of this region, between the headwaters of the Yangtze and Salween rivers, as the most diversified and fioraliy richest in the world. He feels that in colonizing new habitats in this territory new polyploid races and species are produced and survive in abundance. He regards this as a special situation in which extreme or peripheral ecological conditions confront the species in what is geographically the interior of its range. World World of Bamhusa, have a somatic number of 2ai = 46. This number has been found in three species of the subgenus so far examined: B. capitata (Gould & Soderstrom, 1967), B. chacoensis (Quari'n, 1977), and B. paraguayana (Quari'n, 1977). We and Swallenochloa. The expected tetraploid number of 2ai =48 was given by Janaki Ammal (1959) for the species of Chusquea that she studied, while Pohl & Davidse (1971) found 2« = 40 for the species of Chusquea and Swallenochloa that they examined. An even lower number, 2n = 36, was reported by Virkki (1963). The greatest variation of somatic numbers within a group occurs in the tribe Olyreae, which contains 15 described genera of herbaceous bambusoid grasses. The tribe is endemic to tropical America except for the single weedy species, Olyra latifoUa, which has become naturalized elsewhere. Somatic numbers of 22 and 20 are the most common in the tribe, but 18 and 14 have also been reported, as well as 30, 40, and 44. A basic number of a^ = 12 has not been found in any member of this tribe. In the genus Olyra itself we find n = 11 to be common: in O. latifolia we have diploids of 2« =22 (Davidse & Pohl, 1972a) and tetraploids oi 2n =44 (Davidse & Pohl, 1974), while in another widespread weedy species, O. micran- tha, we have only 2/2 = 40 (Gould & Soderstrom, 1967). Olyra taquara is a diploid, with 2n = 20, and O.fasciculata a diploid with 2n ^ 14 (Gould & Soder- strom, 1967). Other genera in the tribe include Cryptochloa, with a report of 2n = 22 for C. concinna (Davidse & Pohl, 1974) and 2a? = 20 for an undescribed species {Soderstrom 1380) from Colombia (Gould in litt.). Other olyroid genera with a basic number of « = 11 are Lithachne, where both the diploid, Iti = 22 (Pohl 1981] SODERSTROM— BAMBUSOIDEAE 29 & Davidse, 1971), and tetraploid, 2n = 44 (Quarin, 1977), have been found; Mac- lurolyra tecta, In = 22 (Calderon & Soderstrom, 1973); and Piresia, In = 22 (Gould & Soderstrom, 1967). A basic number of 10 has been found in Raddiella esenbeckii (as R, nana) by Davidse & Pohl (1972b) and Rehia (as Bulhulus) nervata by Gould & Soderstrom (1967). In the herbaceous tribe Parianeae, In = 44 has been found for a species of Pariana {Calderon & Dressier 2136) (J. Hunziker, 1978, in litt.) and Tn = 12 for P. parvispica (Pohl, 1972), which contrasts with a basic number of x = 12 implied in the 2/2 = 48 count reported by Reeder et al. (1969) for P. stenolemma. In our own cultivated specimens of Eremitis, Royce Oliver (pers. comm.) has recently found the two species examined to be high polyploids, with somatic numbers over 60. In the tribe Streptochaeteae, both species of the single genus, Streptochaeta, have been found to be diploids, with In =22. This count was reported for 5. sodiroana (Pohl & Davidse, 1971) and S. spicata (Valencia, 1962). We still do not have counts for two of the tribes of herbaceous genera: An- omochloeae of South America and Buergersiochloeae of New Guinea. We also lack counts for the majority of woody genera. While the number of chromosomes in the set is important and further counts will be useful in understanding trends of evolution in the subfamily, karyotype analyses should also be undertaken. One such analysis was presented by Daker (1968), who investigated at Kew the cultivated material of the herbaceous bam- boo, Diandrolyra hicolor. He found only 18 chromosomes in the set and his illustration shows that they are asymmetric. Another such analysis was made by Virkki (1963), who found 18 pairs of metacentric chromosomes in the bamboo, Swallenochloa (as Chusquea) subtessellata. In his study of Streptochaeta spi- cata, Valencia stated that of the eleven pairs of chromosomes he observed at metaphase, the position of the centromere was central in four pairs, while in the others the arms were slightly unequal. Streptochaeta A NEW SPECIES One of the most unusual of the herbaceous bambusoid grasses is the genus Streptochaeta, whose remarkable ''spikelet'' has long been the basis of investi- gation and speculation. The genus was named by Schrader and published by Nees von Esenbeck, who examined the type specimen in the Berlin herbarium and published the name, 5. spicata, in 1829. The type specimen came from Felisberto in the state of Bahia, Brazil. In his treatment of the bamboos of that country, Nees (1835) allocated the genera to three groups, one of which included Strep- tochaeta by itself. The following year he made it the basis of a tribe, Strepto- chaeteae (Nees, 1836). The genus is found only in tropical America, with the most common species — the type — occurring from southern Mexico to northern Argentina. It is a medium- sized plant with broad, oval leaves generally 2-4.5 cm wide. A less widespread, but not uncommon, species is S. sodiroana, which ranges from Guatemala, Hon- duras, Belize, Costa Rica, and Panama to lowland Ecuador. It is a larger plant 30 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 with oval leaves that reach as much as 8 cm across and with smaller but more numerous spikelets in the inflorescence. Until now, only two species have been known in the genus. However, a third can now be added. In 1972, during a collecting trip through eastern Brazil my colleague, Dimitri Sucre, and I located a population of Streptochaeta plants quite distinct from the known species. The genus was immediately recognized by the distinctive spikelets, although they differed from the known ones in size and other features. But the narrow, lanceolate blades presented an especially striking con- trast to those of the other two species, which have among the broadest and most ovate blades of any grass. A study of the new species raised again the nature of the so-called ''spikelef' in this genus, an interpretation of which I present in the discussion that follows its formal description. I am naming the new species, Streptochaeta angustifolia, in obvious reference to its distinctive blades. Streptochaeta angustifolia Soderstrom, sp. nov. type: Brazil, Espirito Santo, Mun. Cachoeiro de Itapemirim, 10 km from Cachoeiro toward Alegre, 20°47'S, 4r09'W, elev. ca. 90 m, Soderstrom & Sucre 1969 (RB, holotype; CEPLAC, INPA, K, MO, P, US, isotypes).— Figs. 5-6. Gramen perenne, usque ad 70 cm altum. Laminae symmetricae, lanceolatae, 10-15 cm longae, 0.5-1.8 cm latae. Inflorescenlia subspicata, 8-1 1 cm longa, 6-8 spiculis in spiram dispositis. Spiculae 1.0-1.5 cm longae, cum 1 1 bracteis spiraliter et verticillatim dispositis. Stamina 6, filamentis adnatis; antherae ca. 5.4 mm longae, apicaliter exsertae. Ovarium fusiforme, 3 stigmatibus, ca. 5 mm longis, non-plumosis, apicaliter exsertis. Perennial with a knotty, short-rhizomatous base of sympodial habit. Culms unbranched, erect, hollow with thick walls, 40-70 cm tall, with 4-8 dark, prom- inent nodes, the internodes 4-14 cm long, in some instances many in succession shortened and resulting in a fascicle of leaves with the sheaths strongly overlap- ping. Leaves evenly distributed along the culm, acuminate at the tip, symmetric at the base, 10-15 cm long, 0.5-1.8 cm wide, scaberulous on the upper surface, glabrous or hirtellous on the lower, the margins entire, the midrib prominent on the upper and lower surfaces, the primary nerves 3 or 4 on each side of the midrib, manifest only on the lower surface, connected by manifest transverse veinlets. Sheaths strongly ribbed, pale green becoming brown, glabrous over the back, ciliate on the upper margins. Petiole lacking, the juncture of blade and sheath a smooth, dark band of tissue covered by cilia abaxially. Inflorescence terminating the culm with a second one occasionally produced from a bud in the axil of the uppermost leaf, 8-11 cm long, subspicate with 6-8 ''spikelets'' ar- ranged spirally on the axis; peduncle 10-13 cm long, pale green, glabrous, be- coming flocculose toward the summit, the axis flocculose, exserted well beyond its subtending leaf, the leaf blade sometimes reduced or lacking. Spikelets falling entire at maturity, 1.0-1.5 cm long with an awn 3-4 times as long; axis of the spikelet bearing numerous more or less spirally arranged bracts: bracts I-V green- ish stramineous, membranous, short, empty, deeply dentate, I and II with 3 or 4 thick nerves, one extending into an awn, 1.3-3.6 mm long, positioned on the side of the spikelet toward the main axis of the inflorescence, near to each other but not overlapping; bracts III-V about the same size and shape, 4-5 mm long, with deeply dentate margins, and with 5 or 6 thick nerves, only slightly overlap- 1981] SODERSTROM— BAMBUSOIDEAE 31 Figure 5. Streptochaeta an^ustifolia.—?i. Habit (x!/^).— b. Midregion of leaf showing summit of sheath and upper surface of bhide (x4.5). — c. Midregion of leaf showing summit of sheath and lower surface of blade (x5). — d. Rhizome system with culm base (x1). — e. Portion of rachis enlarged (X 1.5). [All drawings based on Soderstrom & Sucre 1969, Brazil, Espirito Santo, mun. Cachoeiro de Itapemirim.J 32 ANNALS OF THE MISSOURI BOTANICAL GARDEN IVoL. 68 ping at the base; bract VI subtending a flowering axis and separated from bracts I_V by a curved glabrous internode, the bract ca. 1-1.4 cm long, elongate, co- riaceous, lanceolate, concave, glabrous, 12-nerved, rounded on the back except for a flattened portion at the base from which extends a small, downward-pointing beak, narrowed above and forming a long, slender, tendrillike, coiled awn, 3-5 cm long; bracts VII and VIII lanceolate-triangular, the summits spreading away from each other, 8- or 9-nerved, the bases imbricate, coriaceous, 8.5-10.3 mm long; bract IX lacking; bracts X, XI, and XII a trimerous whorl embracing the androecium and gynoecium, convolute, coriaceous, 12-15-nerved, 12-14.5 mm long. Stamens 6, the filaments fused and forming a delicate tube around the gynoecium at anthesis, the anthers yellow, 5.4 mm long, exserted through the apex of the spikelet, the free part of the filaments above ca. 2 mm long, these attached to the anthers about !4 from the base; ovary fusiform, ca. 4 mm long, the style to 3 mm long, the stigmas 3, nonplumose, ca. 5 mm long. The inflorescence of Streptochaeta, including our new species, bears a num- ber of short-pedicellate spikelets on an axis more or less spirally arranged. For S, spkata. Arber (1929) reported a 2/5 phyllotaxy and Page (1951) a 3/8 phyllotaxy. In that species there are 8-11 spikelets on the axis, in S. sodiroana up to 100, and in the new species, 6-8. In all of them the axis terminates in an aborted portion, sometimes represented only by a tuft of hairs. The so-called spikelet (Fig. 6) of Streptochaeta is short-pedicellate and bears numerous bracts, again arranged around the axis in a more or less spiral to whorled, but not distichous, fashion. The first two are small and few-nerved, placed side by side, with margins free from one another, and facing the axis; these have thick nerves and are membranous between the nerves. The following three bracts are of the same consistency with thick nerves, deeply dentate mar- gins, but are a little larger than the first two; they are somewhat overlapping at the base. The sixth and succeeding bracts are coriaceous, elongate, many-nerved, and curved. Bract VI bears a long coiled and twisted terminal awn. At maturity the awns of the spikelets become entangled and all the spikelets usually fall or are carried away together. Above bract VI are two elongate bracts with overlapping bases and narrowed summits that are falcate and point away from each other. Above this pair are three more elongate and coriaceous bracts that form a whorl around the repro- ductive organs; their margins are overlapping at the base. All of these bracts, the sixth and succeeding ones, form essentially a hard, more or less tubular, structure that surrounds the reproductive organs. The androecium is composed of six stamens whose filaments are fused at the base. At anthesis the tube elongates and the anthers are thrust out through the top of the spikelet, with the extremely thin, delicate, and transparent tube sur- rounding the ovary. The anthers are attached to the filaments about one-quarter from the base and do not hang from the spikelet in a versatile manner. The Figure 6. Streptochaeta angustifolia.—z. Pseudospikelet(x4.5).— b. Series of bracts (I -5) from the base of the pseudospikelel (x6).— c. Pseudospikelet with basal bracts 1-5 removed and showing bracts 7 and 8, whose bases are overlapping (x4. 5),— d. Bract 6 with long coiled awn (x4. 5).— e. Back portion of the base of bract 6 showing region where embryo exits at germination, — f. Bracts 10-12 (x6). — g. Bracts 7 and 8 (x6). Bract 9, which exists in other species, has not been found here. — h. 1981] SODERSTROM— BAMBUSOIDEAE 33 12 7 8 Ovary with long style and 3 stigmas, surrounded by the thin, fused filaments of the 6 stamens (x4.5). [All drawings based on Soderstrom & Sucre 1969, Brazil, Espirito Santo, Mun. Cachoeiro de Ita- pemirim.] 34 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 gynoecium consists of an elongate ovary, a long style, and three long, nonplumose stigmas that also exit through the summit of the spikelet. After fertilization the ovary develops into a fruit with a long, linear hilum and small embryo. It is tightly enclosed within the hard bracts, all of which remain together as a unit through germination. At germination the embryo pushes through the noncoriaceous bases of the upper bracts and through a special tissue at the base of bract VI. ANALYSIS OF StREPTOCHAETA SPIKELET Several important anatomical and morphological studies have been carried out on the genus, the most significant being those of Arber (1929, 1934), Metcalfe (1960), and Page (1947, Metcalfe showed that the leaf is of the bambusoid type, with its complex system of vascular bundles in the keel, the nonradiate chlorenchyma made up of arm cells, and the presence of fusoid cells. The epidermis is also of the bambusoid type, and so is the seedling, as I have shown. Furthermore, the chromosome number of 2/j = 22 corresponds to that of other herbaceous bambusoid grasses of the New World, especially members of the tribes Olyreae and Parianeae. These New World bam- busoid grasses, based on numbers derived from .v = 12, have highly evolved and often specialized inflorescences, a pattern shared by Streptochaeta. There is no reason to concur with Butzin (1965), who made this genus the basis of a mono- typic subfamily. The nondistichous arrangement of the spikelets on the inflorescence axis and similar arrangement of the multiple bracts on the spikelet axis have confused botanists since the day the genus was described. The genus has been the subject of investigation by numerous morphologists, and many papers have been pub- lished with widely divergent hypotheses concerning the interpretation of the spikelet. Celakov (1890), Arber (1929), and Page (1951). All authors have agreed that Streptochaeta is a primitive grass, perhaps the most so in the family. Until now interpretations of the spikelet of Streptochaeta have been overly influenced, in my opinion, by the nondistichous arrangement of its many parts, in a family where distichy is the rule. Twisting and torsion are prevalent in this taxon, however, and one can see the twisted inflorescence axis with the naked eye. Such a twisting could camouflage a distichous arrangement and bring the parts into a somewhat spiral phyllotaxy. This twisting was probably repeated in the axis of the spikelet, at least historically, and this, with shortening of inter- nodes, could account for some parts appearing to be in whorls. Within the Bam- busoideae nondistichous placement of parts is not altogether unusual; as exam- ples, I refer to the glumes of the male spikelets of Pariana placed side by side and to the spiral arrangement of spikelet bracts (glumes) in the inflorescence of the bamboo, Melocanna, as reported by Petrova (1973). The nonbambusoid grass, Micraira, is well known for the spiral arrangement of its leaves. Page (1951) found that buds would sometimes form in the axils of bracts I-V in S. spicata, and Arber (1929) encountered them in some bracts in S. sodiroana. While such buds do not normally develop further, even when present, Page found 1981] SODERSTROM— BAMBUSOIDEAE 35 that in rare instances the bud subtended by bract V would develop partially. In such a case the bract itself elongated, became coriaceous, and produced a coiled awn just like that of bract VI, to which it may thus be considered homologous. Bracts I-VI are therefore homologous structures. Except for their nondistichous placement on the axis, then, we can consider bracts I-V to be homologous structures that are potentially gemmiferous and are homologous to bract VI, which is long, coriaceous, many-nerved, long-awned and sometimes interpreted as the fertile lemma. It subtends further structures that precede the reproductive organs. These further structures include a pair (sometimes a third developing) with the bases overlapping, sometimes interpreted as two halves of a palea, followed by a whorl of three, universally interpreted as lodicules. There is no problem in the interpretation of the reproductive organs them- selves. The gynoecium bears three, long nonplumose stigmas. Three stigmas are commonly found in bamboos, such as Anmdinaria and Bambusa, and non- plumose stigmas are to be found in such genera as Eremitis, Anomochloa, Strep- togyna, and Pharus. The androecium consists of six stamens whose filaments are fused at the base and at anthesis extend into a thin, delicate, and translucent tube that surrounds the ovary. Six stamens are of frequent occurrence in the Bam- busoideae and are to be found in such genera as Bambusa and Elytrostachys of the New World and Melocanna of the Old. The anthers are thrust through the opening at the top of the spikelet as are the stigmas. They are attached about one-quarter from the base and do not hang in a versatile manner as the anthers of many wind-pollinated grasses. The staminal tube formed by the filaments is extremely thin and fragile, a condition that could only occur in a closed and protected environment, such as that afforded by the whorl of three coriaceous structures that completely envelop and protect it. The fusion of filaments into a staminal tube occurs throughout the Bambu- soideae, and to my knowledge six stamens are always involved. We find this in the herbaceous bambusoid genus, Froesiochloa, as well as in all or some species of the following bamboos: Dendrocalamus, Gigantochloa, Oxytenanthera, and Schizostachyum. In all of these cases the anthers exit through the apex of the spikelet and the tube is enclosed within hardened scales. Thus, stamens of Streptochaeta are unlike those of most grasses in that the anthers are thrust through the top of the spikelet rather than laterally from it where they can hang in a versatile condition. Whatever the mechanism of polli- nation may be in this genus, the position of the anthers and the nonplumose nature of the stigmas speak against wind as the agent of pollen transfer, a situation not unknown in the subfamily (Soderstrom & Calderon, 1971). In most grasses that are wind-pollinated, the bracts (usually lemma and palea) that embrace the reproductive organs spread apart and allow the stamens and stigmas to extend laterally, and after fertilization they again close and protect the developing fruit. The pushing apart of the lemma and palea is brought about by the lodicules which at this moment become swollen and turgid, thus forcing apart these structures. After fertilization they become flaccid and the lemma and palea come together again. Dobrotvorskaya (1962), who has made extensive studies on lodicules, consid- 36 ANNALS OF THH MISSOURI BOTANICAL GARDEN [Vol. 68 ers this to be their primary function. She also feels that they play a role in protecting the ovary and regulating its water metabolism. In Strcptochaeta the anthers do not exit laterally, which we can simply ob- serve or deduce from the fused filaments, a condition that would not allow this. The nonplumose stigmas also corroborate the fact that we are not dealing with a wind-pollinated grass in which laterally exiting, versatile anthers are the rule. The primary function of lodicules in Strcptochaeta, then, does not exist. In other bambusoid grasses where there is a staminal tube and the anthers exit terminally, there are no lodicules. In all species of Gigantochloa the filaments form a tube and lodicules are lacking. I have not found lodicules in Frocsiochloa\ and in those species of Schizostachyum that have a staminal tube there are none. In other cases, where the filaments are free but are enclosed in hard scales that do not open and the anthers exit terminally, the lodicules are likewise absent. Such is the case in Anomochloa and Bambusa atra. It seems reasonable to assume that Strcptochaeta has followed the same course of evolution as other bambusoid grasses, and so with the development of the staminal tube and terminal exiting of the anthers there would also be loss of lodicule function and therefore loss of the lodicules themselves. Furthermore, the structures that have been called lodicules are so unlike these organs in any grass, bambusoid or otherwise, that it is difficult to accept them as such. Rather, they may be bracts that are homologous to the ''palea bracts,'' which in turn are homologous to bracts I-VI. Dobrotvorskaya (1962) presented data from her study of these 'iodicules'' in Strcptochaeta that corroborates this hypothesis, even though she did not question their true nature. She referred to the genus as the most primitive grass and ac- cepted the three large scales as lodicules. She pointed out that they protected the delicate staminal tube and resembled bracts to a great degree, not corresponding to usual lodicules, which are small scales. Dobrotvorskaya found that the cells in the lower part of the Strcptochaeta lodicules reached gigantic size and she compared these to similar cells found in the inner epidermis of the lemma and palea of grasses in the tribe Hordeae. She also observed similar large cells in the lemma and palea of Anthoxanthum odoratum, an especially interesting obser- vation, for in that species the ovary is enclosed in a hardened lemma and palea and lodicules are lacking. Thus, the anatomical evidence presented by Dobrot- vorskaya favors the interpretation that the lodicules of Strcptochaeta are bracts homologous to lemmas (or lemma and palea) that protect the ovary, have a water- storage function, and peiform a role in the process of floral development. At this point we should refer to teratological specimens that were studied by Page (1951). She found that in cases where the bud of bract V developed, the bract itself elongated and became similar to bract VI, complete with a long, coiled awn. At the same time the following two bracts, VII and VIII, elongated and produced small, coiled awns, thus indicating an homology between these bracts and all of the preceding ones. Page also presented evidence to show that bracts VII and VIII were on a different axis from bracts I-VI. A further important point is that in the early stages of development a ridge always develops opposite and above bracts VII and VIII. On rare occasions, in 5. spicata. Page (1951) found that the ridge developed into a full-grown bract and 1981] SODERSTROM— BAMBUSOIDEAE 37 completed a whorl of three. Some authors, such as Arber (1929), argued that bracts VII and VIII represented halves of a palea, and the interpretation of these two as members of an outer perianth was the basis of Celakovsky's theory in 1889 that the grass palea originated from two outer perianth parts. Page, however, showed that all three had separate origins, a fact that thus negated these hypoth- eses. If the three large bracts that surround the reproductive organs are not lodicules and the two (or three) below them are not halves of a palea, what do they rep- resent? By their size, many-nerved condition, and coriaceous nature, they most resemble bract VI, except for the coiled awn, and I have already pointed out that the "palea bracts" sometimes do, in fact, develop coiled awns that indicate ho- mology with bract VI. I have already suggested that the whorled arrangement of bracts may be brought about by twisting of the axis and shortening of internodes, conditions that appear to be the rule in the inflorescence of this genus. THE PSEUDOSPIKELET OF BAMBOOS Two items are especially pertinent to my interpretation of the Streptochaeta spikelet: (1) bracts I-V are potentially gemmiferous, and (2) the genus is a member of the Bambusoideae. On the first point rests the assumption that we are not dealing with a spikelet but an inflorescence branch, and on the second the opinion that we should interpret such a branch by analogy to similar structures within related bambusoid grasses. The numerous, potentially gemmiferous, bracts that we find in the spikelet of Streptochaeta, represent a condition that is common in bamboos. Kurz (1876: 262) was, to my knowledge, the first to draw attention to the structure, but it was McClure (1934) who analyzed it in some Chinese species of Schizostachyum, and gave further definition to it in later works (1966, 1973). As defined by McClure, a pseudospikelet is a spikeletlike branch of an indeterminately branching inflo- rescence, and I interpret this to be the condition that we encounter in Strepto- chaeta. 1 shall henceforth use the term pseudospikelet for the flowering unit in this genus and interpret it by analogy to the pseudospikelet that characterizes many bamboos, rather than to the spikelet that characterizes most other grasses. As an example of the pseudospikelet, I have illustrated the inflorescence of Bamhusa atra (Figs. 7a-b), a bamboo native to the Moluccas. As is typical of plants with pseudospikelets, the leaves of the branch, prior to flowering, become progressively smaller until they are represented by sheaths only, each small sheath subtending a bud that may grow into a short flowering axis that is like a spikelet in appearance. While this axis terminates in a spikelet, with bracts (lem- mas) that subtend flowers, there are additional bracts below this that subtend prophyllate branch buds instead of being empty like usual glumes. Such a primary pseudospikelet is diagrammed in Fig. 8d. Here the whole axis is subtended by a bract and bears a prophyllum at its first node. A structure comparable to the ordinary spikelet consists of the upper bracts that subtend floral axes (palea and flower) and two empty bracts (glumes) below this series. Below these empty bracts occur several more bracts like the empty ones in all ways except that each subtends a prophyllate bud. This primary pseudospikelet, as shown in Fig. 8c, 38 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Figure 7. Bamhusa atra. — a. Young inflorescence branch (xVi). — b. Mature inflorescence branch (xVi). Leaf complement (xVi). — d. Upper portion of culm leaf from inside to show ligule (x!4). — e. Upper portion of culm leaf from outside (x!4). — f. Culm leaf in place (xl4). — g. Ligule and auricle of foliage leaf (x3). [All drawings based on fresh material of Soderstrom & Kulatunge 1600 from plant cultivated at Peradeniya, Ceylon.] Figure 8. Flowering systems of Bambusa atra and Streptochaeta angustifolia. — a. Flowering branch of Bambusa atra. — b. Schematic diagram of a. Primary pseudospikelet of Bambusa a tra . Schematic diagram of a primary pseudospikelet such as c. Primary pseudospikelet of 1981] SODERSTROM— BAMBUSOIDEAE 39 Bambusa atra with secondary pseudospikelets developing in the axils of the basal bracts. — f. Sche- matic diagram of a primary pseudospikelet with early secondary pseudospikelets, such as e. — g. Mature clusters of pseudospikelets oi Bambusa atra. — h. Schematic diagram of a fully mature pseu- dospikelet cluster such as g. — i. Representation of pseudospikelet of Streptochaeta angustifolia. — j. Schematic and interpretative diagram of the pseudospikelet in i. [Solid triangle indicates a primary pseudospikelet; open circle with male and female signs indicates fertile floret; solid circle indicates a bud; double-barred flag represents a prophyllum; curved line with hanging tip represents leaf sheath and blade, respectively; curved line alone represents a sheath or subtending bract; broken line rep- resents hypothetical missing structure; wiggly line represents terminal aborted part of axis.] 40 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 looks like an ordinary, many-flowered spikelet, but its indeterminate nature be- comes apparent upon germination of the lower buds, which themselves develop into flowering axes, or secondary pseudospikelets, as shown in Fig. 8e and dia- grammed in Fig. 8f. These secondary pseudospikelets themselves can produce further, tertiary, pseudospikelets, the final result being a cluster of spikelets of various orders, as shown in Fig. 8g and diagrammed in Fig. 8h. In the case of Banihusa atra, which is represented in Fig. 8, the primary pseudospikelets are, of course, the first to develop, but within each spikelet maturation of the florets proceeds acropetally, with the uppermost ones remaining incomplete or aborted. The number of bracts on a floral axis and the number that are empty or that subtend buds (floral or branch) are variable, as is the quantity of pseudospikelets ultimately produced. In some species (such as Banihusa glaucescens) only a few secondary pseudospikelets develop, while in others many are formed in each order and result in a large and dense sphere (as in species of Dendrocalamus and Oxytcnanthera). Subtending bracts and prophylla in the inflorescence are common in bamboos that bear pseudospikelets. The development of the usual spikelet of determinate growth very likely came about by the loss of subtending bracts, prophylla, and some of the buds. Without much imagination we can derive the spicate inflores- cence of Strcptogyna from a theoretical type like that shown in Fig. 8b by loss of the subtending bracts and in each floral axis loss of the prophyllum, loss of the prophyllate buds in the lower bracts (glumes), and the retention of flowering buds in the upper ones (lemmas). In some cases, such as the bamboo Chusquea, four empty bracts generally precede the fertile one (lemma), which subtends a flower. McClure (1973) has referred to these empty bracts as ^'transitional glumes" to distinguish them from ''glumes,'' which in the ordinary spikelet means the two empty bracts that precede the lemma(s). The grass panicle can be derived from an elongation of each flowering axis, as shown in Fig. 8h, loss of subtending bracts and prophylla, and of buds from the lower bracts. This would result in determinate spikelets with empty glumes and fertile florets, each pedicellate, and arranged in a panicle, which is common to most grasses. THE PSEUDOSPIKELET OF StREPTOCHAETA If we regard the flowering unit of Streptochaeta as a pseudospikelet and accept the parts as spirally or verticillately arranged due to variations in twisting of the axes and telescoping of the internodes, we can compare it to a fully de- veloped pseudospikelet, as found in Bambusa atra (Figs. 7b, 8a-h). Such a com- parison allows the following interpretation (cf. Figs. 8i-j): Bracts I-VI are arranged on the axis (2) of the primary pseudospikelet which aborts at the tip, repeating abortion of the apex of the main inflorescence axis (1). The bract that theoretically would subtend the axis (2) and the prophyllum of this axis have been lost through evolution. The bud subtended by bract VI always germinates and produces a new axis (3), which, theoretically, is the axis of a secondary pseudospikelet. Bract VI is the first of the several coriaceous bracts that surround the reproductive parts and its long, coiled awn later assists 1981] SODERSTROM— BAMBUSOIDEAE 41 in the dispersal of the fruit that it ultimately harbors. (At this point it is interesting to recall that when the bud subtended by bract V develops, that bract takes on a form and function like bract VI.) Page has given evidence to show that bracts VII and VIII (and therefore IX, which follows) are on an axis separate from bracts X-XII, so I have diagrammed them (Fig. 8j) as on another axis (4), and assumed that the tip of this axis aborted through evolution and that its prophyllum has likewise been lost. Assuming that the branching pattern repeats itself, bract IX theoretically subtends axis (4); this subtending bract sometimes develops. Axis (4), which belongs to the tertiary pseudospikelet, bears three bracts, X-XII, which previous investigators have considered to be lodicules. This axis conceiv- ably could terminate in the flower but if we assume the branching pattern to be consistent throughout, we must assume that the tip of axis (4) aborted in the course of evolution and that bract XII subtends the floral axis. This axis is sur- rounded and protected by bracts X-XII and the prophyllum (palea) has been lost through evolution as well as the lodicules. The absence of lodicules, androecium of 6 stamens, and gynoecium with 3 stigmas corresponds to the condition we find in Bambusa atra, with which we have compared the pseudospikelet of Strepto- chaeta. When we reexamine bracts I-XII, it becomes apparent that there are three sets of different kinds of bracts. Bracts I-VI are small and scalelike except when subtending a bud that germinates, as is always the case with bract VI and oc- casionally with bract V. The next set of bracts are VII and VIII, with IX some- times developed; these two or three are long, coriaceous and many-nerved with falcate tips. Bracts X-XII are also long, coriaceous, many-nerved, and with long, but erect, acute tips. Each set of bracts, similar among themselves but with some differences between each other, pertains to a separate axis: bracts I-VI to the axis of the primary pseudospikelet, VII-IX to that of the secondary pseudospike- let, and X-XII to that of a tertiary pseudospikelet. What has commonly been referred to as the spikelet in Streptochaeta is thus a highly modified branching system made up of three orders of pseudospikelets. Streptochaeta is the only herbaceous bambusoid grass that has retained the pseudospikelet, with the possible exception of Anomochloa and the African gen- era, which I have not yet studied. But like other herbaceous bambusoid grasses its inflorescence has become highly specialized, resulting in a greatly telescoped and modified branching system. I would agree with previous investigators that Streptochaeta is among the most primitive of grasses. However, I base this opinion not on its possession of three lodicules and a two-parted palea, structures that I reason are not even present, but on its herbaceous nature and retention of a pseudospikelet. The Primitive Bambusoid Grass The major natural groups of grasses as we recognize them today (e.g., arun- dinoid, pooid, oryzoid, bambusoid) doubtless became differentiated early in the evolution of the family and derived from a form adapted for wind pollination. Various extant grasses possess certain features that we presume to be primitive in that they are common to the larger group, monocotyledons, to which grasses belong. We assume that these features, such as three (possibly six) lodicules or 42 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Voi . 68 perianth parts, six stamens, and three stigmas, were present in the primitive form that preceded all the major groups. Chromosome numbers in multiples of six are characteristic of many grasses, including Oryza and Luziola of the Oryzoideae, Centostcca and Lophatherum of the centostecoid group, Amndo of the Arundinoideae, and many genera of the Bambusoideae. We may postulate that six is indeed the basic number in the subfamily Bambusoideae, although no count of 2/2 = 12 has ever been recorded. To my knowledge, none of the original bambusoid grasses that were diploids of 2/2 = 12 now exists. Probably .v = 12 occurred long ago through doubling and thus brought about this new basic number of polyploid origin. Stebbins (1971: 190-191) pointed out that in ''old, mature, or declining polyploid complexes cer- tain segments initiate new series of polyploid numbers in which the basic number that is multiplied is not the original basic number of the complex [in our case, jc ^ 6] but some multiple of it [.v ^ 12]/' These new series are called ''secondary cycles of polyploidy." At the original tetraploid level of 2a/ = 24, there was a period of diversification and differentiation before the higher polyploids appeared. Herbaceous genera such as Streptogyna, Pharus, Leptaspis, and Puelici may represent original tetra- ploids that were adapted to forest conditions. While most such genera have since become extinct, these few — particularly the first three — may have survived be- cause of their adaptation to more disturbed forest conditions than the rest and a more widespread distribution made possible by excellent dispersal mechanisms. Among the Bambusoideae we consider the above genera, with 2// = 24, to be diploids, based on the basic number of 12 as opposed to the theoretical basic number of 6. The primitive grass leaf probably had an anatomy most like the present-day pooid type, with a relatively undifferentiated mesophyll and simple epidermis lacking bicellular microhairs. The various modifications that took place in the bambusoid leaf, such as the development of arm cells and fusoid cells, may have been derived from that as an adaptation to the shaded, humid forest con- ditions which characterize the habitat of the herbaceous bambusoid genera. Bamboos, whose leaf anatomy and epidermis differ little from those of the herbaceous members, must have evolved from herbaceous ancestors. Bamboos are mostly polyploids, based on x - 12, usually tetraploids but in some cases hexaploids. Development of the large woody habit may have been in response to competition with the tree, the major growth form of the forest. The process that led from herbaceous to woody involved an increase in the chromosome number, or polyploidy, development of complex branching at the nodes, an overall in- crease in size, and emphasis on vegetative growth, such as the development of a strong rhizome system. This particular stress on vegetative growth with greater longevity of the in- dividual culm and increase in size by branching in the bamboos led, at the same time, to a decrease in flowering. Vegetative reproduction became the dominant condition in bamboos, at the expense of sexual reproduction, which now occurred only at intervals, these sometimes as long as 120 years. With flowering so infre- quent, the simultaneous occurrence of this event in two species became too remote a possibility for gene exchange between them to play part in further evolution. By their diminished flowering and the loss of active inflorescence evo- lution, the bamboos became in essence guardians of ancient flowering systems, 1981] SODERSTROM— BAMBUSOIDEAE 43 long since lost in all other grasses — including the herbaceous bamboos — that flow- er every year. We find such a primitive type of inflorescence in bamboo genera Hke Bam- busa, Dendrocalamus, and Oxytenanthera, where the leaves of a vegetative branch become progressively smaller until they consist only of sheaths that sub- tend reproductive buds. These buds develop into indeterminate spikeletlike branches called '^pseudospikelets/' In the most primitive form, such as in Bam- busa, the flowers in these pseudospikelets are complete, with three well-devel- oped lodicules, six stamens, and three stigmas. From such a pseudospikelet with complete flowers, we can follow the hypothetical evolution of the usual grass spikelet and its arrangement in a "raceme" or ''panicle.'' Such a development must have come about by the loss of subtending bracts, prophylla, and buds, which are common in the pseudospikelet, and a decrease in the number of all parts of the flower. We would predict that in the most primitive herbaceous bambusoid grass the leaves of the axis, as in Bambusa, would become progressively smaller until they were reduced to sheaths that subtend pseudospikelets. Except for the lack of subtending bracts, the closest we come to this is Streptochaeta, where the pseu- dospikelet is highly modified. The inflorescence of Streptogyna is also primitive in its spicate arrangement, although subtending bracts are not present and the spikelets are of the usual determinate type. In both of these genera, however, the spikelets are bisexual, the former with an unspeciaHzed androecium of six sta- mens and gynoecium of three stigmas, and the latter with an advanced androe- cium of three stamens and gynoecium sometimes of two stigmas. The inflores- cence of Anomochloa may be unspeciaHzed and certainly begs analysis; however, the flower is highly specialized in its lack of lodicules, four stamens, and single nonplumose stigma. The many-flowered, bisexual spikelets of the African genera, Puelia and Guaduella, which are arranged in panicles, also need further study. The most highly evolved inflorescences of herbaceous bambusoid grasses are those that are monoecious with one-flowered spikelets. Four tribes exhibit this condition: Buergersiochloeae, Olyreae, Parianeae, and Phareae. Although Raven & Axelrod (1974: 594) speak of the olyroid grasses as relatively unspeciaHzed, they are, on the contrary, the most specialized in the subfamily. Genera of this tribe have not only reached the monoecious state but within the same plant there is often great differentiation between male and female inflorescences. The tribe, with variable chromosome complements derived from basic numbers lower than 12, appears to be in a state of active evolution. The only herbaceous bambusoid grasses that are diploids based on x = 12 are found in the tribes Phareae, Strep- togyneae, and Atractocarpeae. All of these tribes are also present in Africa: the World World one in Africa and Asia; and the third contains two genera — Guaduella and Puelia West of these lacks bicellular microhairs, which are commonly found in bambusoid grasses. The seedlings of those studied — Pharus and Streptogyna — do not have reduced leaves and are the most unspeciaHzed in the subfamily. We do not, then, have a single most primitive bambusoid grass but rather several, each possessing certain primitive features. The most unspeciaHzed inflo- 44 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 rescences of pseudospikelets are found in the polyploid woody genera like Bam- husa, while the primitive herbaceous growth form occurs in diploid genera like Streptogyna, Both of these have chromosome complements based on the prim- itive basic number of 12. Streptochaeta, long regarded as the most primitive grass, cannot hold this title alone. It may indeed be regarded as primitive in its herbaceous condition, modified pseudospikelet, and flower that contains six stamens and three stigmas. But its lack of lodicules, nonplumose stigmas, seedlings with reduced leaves, and a chromosome complement based on a derived basic number of a = II, all show advancement. I do not yet have enough information to postulate the origin of bamboos, but I feel that they have developed from herbaceous stock and have a close and common ancestry. The genera are still not well understood and valid phytogeo- graphical comparisons cannot yet be made. We do know, however, that Africa is poor in bamboos, with few genera and species represented, while Madagascar is the home of several distinct genera. The chromosome numbers of American bamboos tend to be lower while those of Asia are higher, with the greatest con- centration of hexaploids occurring in northern Burma and western China. Clearly we must now study the herbaceous bambusoid grasses of tropical West Africa and the bamboos that occur on Madagascar. An understanding of these may offer further clues regarding the earliest development of the subfamily and aid us in our interpretation of its evolution. Literature Cited Arber, a. 1929. Studies in the Gramineae. VI. I. Streptochaeta. 2. Anomochloa. 3. Ichnanthus. Ann. Bol. (London) 43; 35-53. . 1934. The Gramineae: A Study of Cereal, Bamboo, and Grass. The University Press, Cam- bridge. Avery, G. S., Jr. 1930. Comparative anatomy and morphology of embryos and seedlings of maize, oats, and wheat. Bot. Gaz. (Crawfordsville) 89: 1-39. Brown, W. V. 1%0. The morphology of the grass embryo. Phytomorphology 10: 215-233. BuTZiN, F. 1965. Neue Untersuchungen iiber die Bliite der Gramineae. Inaugural-Dissertation, Ber- lin. Calderon, C. E. & T. R. Soderstrom. 1973. Morphological and anatomical considerations of the grass subfamily Bambusoideae, based on the new genus Maclurolyra. Smithsonian Contr. Bot. i 1: i-iii, 1-55, 24 fi^s. & . 1980. The genera of Bambusoideae (Poaceae) of the American continent: keys and comments. Smithsonian Contr. Bot. 44: i-iv, 1-27. u ■ ■ ■ 4 Celakovsky, L. J. 1889. Uber den Ahrchenbau der Brasilianischen Grasgattung Streptochaeta Schrader. Sitzungsber. Konigl. Bohm. Ges. Wiss. Prag, Math.-Naturwiss. CI. 3: 14-42, table 2. [Unpublished English translation by N4rs. Geti Saad, prepared in 1974 for the Smithsonian In- stitution and National Science Foundation, Washington, D.C.] . 1897. Ueber die Homologien des Grasembryos. Bot. Zeitung, 2. Abt. 55: 141-174, pL 4. Christopher. J. & A. Abraham. 1971. Studies on the cytology and phytogeny of South Indian grasses. I. Subfamilies Bambusoideae, Oryzoideae, Arundinoideae and Festucoideae. Cytologia 36: 579-594. Daker, M. G. 1968. Karyotype analysis of Diandrohra bicolor Stapf (Gramineae). Kew Bull. 21: 433^34. Darlington, C. D. 1973. Chromosome Botany and the Origins of Cultivated Plants. Allen & Unwin, London. Davidse, G. & R. W. PoHL. 1972a. Chromosome numbers and notes on some Central American grasses. Canad. J. Bot. 50: 273-283. & -. 1972b. Chromosome number, meiotic behavior, and notes on some grasses from Central America and the West Indies. Canad. J. Bol. 50: 1441-1452. 19811 SODERSTROM— BAMBUSOIDEAE 45 & . 1974. Chromosome numbers, meiotic behavior, and notes on tropical American grasses (Gramineae). Canad. J. Bot. 52: 317-328. DoBROTVORSKAYA, A. V. 1962. Morfologlchesklye osobennosti lodikul u nekotorykh predstavitelej semejstva zlakovykh. Trudy Botanicheskogo Instituta im. V. L. Komarova Akademii Nauk SSSR, seriya 7 (Morfologiya i Anatomiya Rastenii) 5: 148-165. [Morphological peculiarities of lodicules in some representatives of the grass family. Unpublished English translation by M. Kovanda prepared in 1971. Washington, D.C.J DoELL, J. C. 1880. Tribe 1 1, Bambusaceae. In C. F. P. von Martius (editor). Flora Brasiliensis. Vol. 2(3): 161-220, pis, 44-56 [fascicle 83]. Munich, Vienna. DujARDiN, M. 1978. Chromosome numbers of some tropical African grasses from western Zaire. Canad. J. Bot. 56: 2138-2152. Gould, F. W. 1%0. Chromosome numbers in southwestern grasses. IT Amer. J. Bot. 47: 873-877. & T. R. SoDERSTROM. 1967. Chromosome numbers of tropical American grasses. Amer. J. Bot. 54: 676-683. — & . 1970. Chromosome numbers of some Mexican and Colombian grasses. Canad. J. Bot. 48: 1633-1639. — & . 1974. Chromosome numbers of some Ceylonese grasses. Canad. J. Bot. 52: 1075- 1090. Grosser, D. 1970. Beitrag zur Histologic und Klassifikation Asiatischer Bambusarten. Dissertation, Universitat Hamburg. 321 pp., 23 tables. & W. LiESE. 1971 . On the anatomy of Asian bamboos, with special reference to their vascular bundles. Wood Sci. Techn. 5: 290-312. — & G. I. Zamuco, Jr. 1971. Anatomy of some bamboo species in the Philippines. Philipp. J. Sci. 100: 57-73, 8 pis, GuiGNARD, J. L. 1961. Recherches sur I'embryogenie des Graminees; rapports des Graminees avec Tautres Monocotyledones. Ann. Sci. Nat. Bot., ser. 12, 2: 491-610. Hackee, E. 1890. Fine zweite Art von Streptochaeta\ St. sodlroana n. sp. Oesterr. Bot. Z. 40: 1 1 1- 114. HosHiKAWA, K. 1969. Underground organs of the seedlings and the systemalics of Gramineae. Bot. Gaz, (Crawfordsville) 130: 192-203. Hsu, C.-C. 1972. Preliminary chromosome studies on the vascular plants of Taiwan (V). Cytotax- nomy on some monocotyledons. Taiwania 17: 48-65. Janaki Ammal, E. K. 1945. In C. D. Darlington & E. K. Janaki Ammal. Chromosome Atlas of Cultivated Plants. Allen & Unwin, London. . 1950. Polyploidy in the genus Rhododendron. Rhododendron Year Book 5: 92-98. . 1954. The cyto-geography of the genus Buddleia in Asia. Sci. & Cult. 19: 578-581. . 1959. A cytosystematic survey of Bambuseae. I. Bull. Bot. Surv. India 1: 78-84. Kammacher, p., G. Anoma, E. Adjanohoun & L. Ake Assi. 1973. Nombres chromosomiques de Graminees de Cote-dTvoire. Candollea 28: 191-217. KuRZ, S. 1876. Bamboo and its uses. Indian Forester 1: 219-269, pis. I-II; 1(4): 335-362, pis. HI- IV. KuwABARA, Y. 1%0. The first seedling leaf in grass systematics. J. Jap. Bot. 35: 139-145. McClure, F. a. 1934. The inflorescence in Scluzostachyuni Nees. J. Wash, Acad. Sci. 24: 541- 548. . 1966. The Bamboos — A Fresh Perspective. Harvard Univ. Press, Cambridge, Massachusetts. . 1973, Genera of bamboos native to the New World (Gramineae: Bambusoideae). Smithsonian Contr. Bot. 9: i-xii, 1-148, 48 figs. (Edited by T. R. Soderstrom.] Mehra, p. N. & V. Kalia. 1976. In lOPB chromsome number reports LIV, Taxon 25: 631-649. & M. L. Sharma. 1972. In lOPB chromosome number reports XXXVI. Taxon 21: 340-341. Metcalfe, C. R. 1960. Anatomy of the Monocotyledons. Vol. 1. Gramineae. Oxford Univ. Press, London. Nees von Esenbeck, C. G. D. 1829. Agrostologia Brasiliensis . . . J. G. Cottae, Stuttgart. [This represents Volume 2, Part 1, Gramineae, of Flora Brasiliensis seu enumeratis . . . , edited by C. F. P. von Martius.] . 1835. Bambuseae Brasilienses. Recensuit, et alias in India orientali provenientes adjecit. Linnaea 9: 461-494. — . 1836. Genera [of grasses]. Pp. 378-383, in J. Lindley, A Natural System of Botany; or, A Systematic View of the Organisation, Natural Affinities, and Geographical Distribution, of the Whole Vegetable Kingdom ... Ed. 2. Longman, Rees, Orme, Brown, Green, and Longman, London. Negbi, M. & D. KoLLER. 1962. Homologies in the grass embryo — a re-evaluation. Phytomorphology 12: 289-296. Okamura, H. & S. KoNDO. 1%3. Karyotype analysis of Bambusaceae (I). Rep. Fuji Bamboo Gard. 8: 35-39. 46 ANNALS OF THE MISSOURI BOTANICAL GARDEN |Voi . 68 Page, V. M. 1947. Leaf anatomy of Streptochaeta and the relation of this genus to the bamboos. Bull. Torrey Bot. Club 74: 232-239. . 1951. Morphology of the spikeiet of Streptochaeta. Bull. Torrey Bot. Club 78: ll-'il. Petrova, L. R. 1973. Morfologia reproduktivnykh organov nekotorykh vidov Bambusoideae (k filogenii podsemejstva). Byull. Muskovsk. Obshch. Isp. Prir., Otd. Biol. 78: 113-123. [Morphol- ogy of the reproductive organs in some Bambusoideae; toward the phylogeny of the subfamily. Unpublished English translation by M. Kovanda, prepared in 1974, Prague. Copy in Smithsonian Libraries, Washington, D.Cl PoHL, R. W. 1972. New taxa of Hierochloe, Pariana, and Triplasis from Costa Rica. Iowa State J. Res. 47: 71-78. & G. Davidse. 1971. Chromosome numbers of Costa Rican grasses. Brittonia 23: 293-324. QuARiN, C. 1977. Recuentos cromosomicos en gramineas de Argentina subtropical. Hickenia 1: 73-78. Raven, P. H. & D. L Axelrod. 1974. Angiosperm biogeography and past continental movements. Ann. Missouri Bot. Garden 6; 539-673. Reeder, J. R. 1953. The embryo of Streptochaeta and its bearing on the homology of the coleoptile. Amer. J. Bot. 40: 77-80. . 1957. The embryo in grass systematics. Amer. J. Bot. 44: 756-768. . 1962, The bambusoid embryo: a reappraisal. Amer, J. Bot. 49: 639-641. & D. N. Singh. 1967. In lOPB chromosome number reports XL Taxon 16: 215-222. , T. R. Soderstrom & C. E. Calderon. 1969. In lOPB chromosome number reports XXIL Taxon 18: 441-442. RiCHHARiA, R. & J. KoTVAi.. 1940. Chromosomc number in bamboo {Dendrocahunus strictus). Indian J. Agric. Sci. 10: 1033. Sarkar, a, K., R, Mallick, N. Dltta & U. Chatterjee. 1977a. //; lOPB chromosome number reports LVII. Taxon 26: 448^50. et al. 1977b. /// lOPB chromosome number reports LVIII. Taxon 26: 565. Soderstrom, T. R. & C. E. Calderon. 1971. Insect pollination in tropical rain forest grasses. Biotropica 3: 1-6. Stapf, O. 1904. On the fruit of Melocanna hamhusoides, Trin., an endospermless, viviparous genus of Bambuseae. Bot. Trans. Linn. Soc. London. Bot. 6: 401-425, pLs. 45-47, Stebbins, G. L. 1971. Chromosomal Evolution of Higher Plants. Edward Arnold Ltd., London. Tateoka, T. 1958. Somatic chromosomes of Leptasp'is and Strepto^yna (Poaceae). Nature 182: 1619-1620. TzvELEV, N. N. 1976. Zlaki SSSR. Akademiya Nauk SSSR Botanicheskii Institut im. V. L. Koma- rova, Leningrad. [Poaceae of the USSR. V. L. Komarov Botanical Institute of the Academy of Sciences of the USSR.] Valencia, J. I%2. Los cromosomas de Streptochaeta spicata Schrad. (Gramineae). Darwiniana 12: 379-383. Virkki, N. 1963. Eighteen chromosome pairs in an American bamboo, Chusquea suhtesseUata Hitchc. J. Agric. Univ. Puerto Rico 47: 98-101. Yakovlev, M. S. 1950. Struktura endosperma i zarodysha zlakov kak sistematicheskii priznak. Moskva, Leningrad: Akademiya Nauk SSSR, Trudy Botanicheskogo Instituta im. V. L. Komaro- va, seriya 7 (Morfologiya i Anatomiya Rastenii) I: 121-218. [Endosperm and embryo structure of grasses as a taxonomic feature. Moscow, Leningrad: Transactions of the V. L. Komarov Bo- tanical Institute of the Academy of Sciences of the USSR, series 7 (Morphology and Anatomy of Plants) 1: 121-218. Unpublished English translation by the Indian National Scientific Docu- mentation Centre, prepared in 1975.] 1981J SODERSTROM— BAMBUSOIDEAE 47 Index to Taxa Anomochloa Brongniart 18, 35, 36, 41, 43 Anomochloeae C. E. Hubbard in Hutchinson 17, 18, 29 r Anomochlooideae Pilger ex Polztal 15 Anthoxanthum odoratum L. 36 Arberella Soderstrom & Calderon 18 Arundinaria Michaux 16. 18, 27, 35 giganiea (Walter) Muhlenberg 22, 23, 26 pygmaea (Miquel) Ascherson & Graebner 16 Arundinoideae Tateoka 42 Arundo L. 42 Atractocarpeae Jacques-Felix ined. 18, 26, 43 Aulonemia aff. aristulata (Doell) McCIure 21, 22 Bambusa Retzius corr. Schreber 18, 27, 35, 43, 44 Bambusa subgen. Guadua (Kunth) Hackel 27. 28 atra Lindley 36-41 capitata Trinius 28 chacoensis Rojas 28 ghiucescens (Willdenow) Siebold ex Holttum 40 oldhamii Munro 26 paraguayana (Doell) Bertoni 28 polymorpha Munro 28 stenostachya Hackel 26 tulda Roxburgh 28 Bambuseae Kunth 18 Bambusoideae Nees 15-17, 21. 23, 26, 27, 34, 35, 37,42 Buergersiochloa Pilger 18 Buergersiochloeae Blake 18, 29, 43 Bulhulus Swallen 29 Centosteca Desvaux 42 Cephalostachyum 27 pergracile Munro 28 Chimonohambusa Makino 26, 27 Chusquea Kunth 27. 28, 40 subtessellata Hitchcock 29 Cryptochloa Swallen in Woodson & Schery 18, 28 concinna (Hooker f.) Swallen 28 Dendrocalamus Nees 16, 27, 35, 40, 43 brandisii (Munro) Kurz 28 giganteus Munro 15, 26, 28 hamihonii Nees & Arnott ex Munro 28 longispathus (Kurz) Kurz 28 strictus (Roxburgh) Nees 28 Diandrolyra Stapf 18, 22, 23 bicolor Stapf 29 Ekmanochloa Hitchcock 18 Elytrostachys McCIure 35 Eremitis Doell in Martius 18, 29, 35 Froesiochloa G. A. Black 18, 35, 36 Gigantochloa Kurz 27, 35, 36 macrostachya Kurz 28 Guadua Kunth 28 Guaduella Franchet 18, 43 Hordeae Kunth 18, 36 Indocalamus Nakai 26, 27 Leptaspis R. Brown 17, 19, 26, 42, 43 Lithachne Palisot de Beauvois 18, 28 Lophatherum Brongniart in Duperry 42 Luziola A. L. Jussieu 42 Maclurolvra Calderon & Soderstrom 22 tecta Calderon & Soderstrom 18, 29 Melocanna Trinius 19, 34, 35 Micraira F. Mueller 34 Micrairoideae Pilger 15 Mniochloa Chase 18 Neurolepis Meisner 26, 27 Ochlandra Thwaites 27, 28 stridula Thwaites 21, 24 Ohra L. 18 fasckulata Trinius 26, 28 latifolia L. 28 loretensis Mez 23, 24 m'urantha H. B. K. 28 taquara Swallen 28 Olyreae Kunth 17, 18, 27, 28, 34, 43 Oryza L. 42 Oryzoideae Parodi ined. 42 Oxytenanthera Munro 27, 35, 40, 43 abyssinica (A. Richard) Munro 26 Paniceae Kunth 18 Pariana Fusee-Aublet 18, 27. 29, 34 parvLspica Pohl 29 stenolemma Tutin 29 Parianeae (Hackel) C. E. Hubbard in Hutchinson 17. 18, 29, 34,43 Phareae Stapf in Thiselton-Dyer 17, 26, 27, 43 Pharus P. Browne 17, 19, 22-24, 26, 35, 42, 43 Phyllostachys Siebold & Zuccanni 27 Piresia Swallen 18, 29 Pooideae 16 Puelia Franchet 18, 19, 26, 27, 42, 43 Raddia A. Bertoloni 18, 23, 24 Raddiclla Swallen in Maguire et al. 18, 29 esenbeckii (Steudel) Calderon & Soderstrom 29 nana (Doell) Swallen 29 Rehia Fijten 18, 29 nervata (Swallen) Fijten 29 Reitzia Swallen 18 Sasa Makino & Shibala 27 Schizostachyum Nees 35-37 Shibataea Makino 26, 27 Strephium Schrader ex Nees 18 Streptochaeta Schrader ex Nees 18, 19, 22, 29, 30, 32, 34-41, 43, 44 angustifolia Soderstrom 30-33, 38, 39 sodiroana Hackel 29, 32, 34 spicata Schrader ex Nees 21, 24, 29. 32. 34. 36 Streptochaeteae Nees in Lindley 17, 18, 29, 44 Streptogyna Palisot de Beauvois 17, 22, 26. 27, 35, 40, 42-44 americana C. E. Hubbard 24, 25 Streptogyneae C. E. Hubbard ex Calderon & So- derstrom 17, 26, 43 Swallenochloa McCIure 27, 28 subtessellata (Hitchcock) McCIure 29 Thamnocalamus Munro 27 Tnticum L. 19 pauciflora (Swartz) Palisot de Beauvois 21, 22 Zea L. 19 EVOLUTION OF REPRODUCTIVE SYSTEMS IN THE GRAMINEAE 1 H. E. Connor^ Abstract From a simple hermaphrodite flower, and from a complex incompalibihty system unique among the flowering plants, several breeding systems have evolved in the Gramineae. Self-compatibility is the most commonplace variant and following this mutation, habitual or facultative cleistogamy is a simple evolutionary step. Separation of the sexes to different plants as in dioecism is relatively uncommon (ca. 20 genera), and gynodioecism is much less frequent still (3-4 genera). Both are seen as escape pathways from self-compatibility; the development of such pathways is discussed. Sepa- ration of the sexes to separate flowers as in monoecism is relatively common and with the variants andromonoecism and gynomonoecism is the most substantial departure from hermaphroditism in the family. These states are also interpreted as responses to self-compatibility; and though they do not generate cross-fertilization, they assist its evolution. Pathways for the evolution of these breeding systems are described. Apomixis and the breeding system best suited are discussed. Reproductive biology in the Gramineae begins at the transition in the shoot apex from leaf production to the initiation of inflorescence primordia and the later development of floral structures. These have been well described for numerous grasses (Barnard, 1955, 1957, 1964; Bonnett, 1966; Sharman, 1960), and are me- diated by photoperiod. The review of Evans (1964) elegantly reveals data on the interplay of daylength, temperature, and vernalization on inflorescence devel- opment. Floral induction and initiation may occur in the season of flowering (Evans, 1964), or in the season preceding inflorescence emergence (Mark, 1965; Hodgson, 1966). Inflorescence emergence is temperature or daylength dependent (Cooper, 1952; Connor, 1963; Heslop-Harrison, 1961). Temperature also controls anthesis, and later the release of pollen from anthers. Photoperiod, however, has other effects on the reproductive cycle. It may, for example, affect the frequency of cleistogamy in facultatively cleistogamous grasses (Langer & Wilson, 1965), or in a facultative apomict the frequency of apomictic or sexual embryo sacs (Knox, 1967; Knox & Heslop-Harrison, 1966), or depress maleness (Heslop-Harrison, 1959), or reduce the number of florets in the male inflorescence oi Zea mays (Moss & Heslop-Harrison, 1968), or control protandry and protogyny (Emerson, 1924). Although these environmental influences are considerable, genetic influences may cause transient male and female sterility, or restore lost fertility in part or in whole, or promote redistribution of the sex-forms. A combination of both genetic and physiological factors may introduce difficulties into an interpretation of the breeding system of any grass. At the International Symposium on Reproduction in Flowering Plants held at Christchurch, New Zealand, in 1979, 1 presented a survey of the breeding systems known in the Gramineae (Connor, 1980). This present paper is complementary ' 1 am grateful to colleagues and friends in several countries for helpful discussion, and especially to D. W. Clayton, Kew, T, R. Soderstrom, Smithsonian Institution, Ana M. Anton, Universidad Nacional de Cordoba, Elizabeth Edgar and C. J. Webb, Botany Division, DSIR. ^ Botany Division, Department of Scientific and Industrial Research, Christchurch, New Zealand. Ann. MrssouRi Rot. Card. 68: 48-74. 1981. 0026-M93/8 1 /0048-0074/S2 . 8 VO 1981] CONNOR— REPRODUCTIVE SYSTEMS IN THE GRAMINEAE 49 to that; the bibliographies of both papers supplement each other. Here, as there, I follow the organization of tribes proposed by Hubbard (1973) for the grasses. The Flower, the Spikelet, and the Inflorescence The primitive grass flower is postulated as having three bracts and lodicules, six stamens, and a 1- or 3-locular tristigmatic gynoecium (Schuster, 1910; Arber, 1934; Clifford, 1961). The spikelet is considered to have been many flowered, and the inflorescence is thought to have comprised many of these spikelets in a simple more- or less-branched, terminal panicle. Associated with these characteristics is the further postulate of entomophily (Schuster, 1910; Clifford, 1961; Stebbins, 1956, 1974), From these very reasonable postulates there is a general evolution in the flower by reduction to two bracts and lodicules, three stamens, two stigmas and one ovule, though this combination is by no means universal. Associated features that are interpreted as part of an anemophilous syndrome are: rapidly elongating staminal filaments; easily exserted stigmata; readily dispersed pollen; a single ovule per flower; and the preservation of a self-incompatibility system. Within the flower some further reductions are detectable, e.g., lodicules may be lacking or be so small as to be ineffective. Absence of lodicules is frequently associated with protogyny in hermaphrodite flowers but may also occur in mon- oecious and dioecious taxa; see summary in Connor (1980). Stamens may be reduced to two or one per flower; this state seems randomly distributed among the tribes and is independent of what seems its logical correlative — cleistogamy (Connor, 1980). Among spikelets there are some trends in common. Geminate dimorphic spikelets of the form one sessile and the other pedicelled may be found in such different tribes as the Phareae and Olyreae (Bambusoideae) and in the Andro- pogoneae and Maydeae (Panicoideae). This form is frequently associated with a discrimination such that male elements alone are in pedicelled spikelets and fe- male organs in sessile ones. A single hermaphrodite floret borne terminally in a spikelet of two or more florets may be found in the Bambuseae, Arundinelleae, Aveneae, Paniceae, and Andropogoneae. The ultimate reduction is to one floret per spikelet, and this is found in such differing tribes as Phareae, Agrostideae, and Sporoboleae. The precise evolutionary phases displayed in inflorescence development are not topics for this paper except as they display relationships to sex-form. Among monoecious genera there is a slight trend towards axillary female inflorescences and terminal male inflorescences, e.g., Humhertochloa, Hydrochloa, Luziola, and Zea. Axillary inflorescences are a feature of some herbaceous members of the Bambusoideae; in such genera as Arherella, Raddia, and Strephium they occur at almost all the nodes, but in Olyra axillary branches occur at uppermost nodes only (Calderon & Soderstrom, 1973; Soderstrom & Calderon, 1979a). These genera are also monoecious but the axillary inflorescences are of mixed male and female spikelets. Clayton (1969) detects in the Andropogoneae an evolution towards numerous smaller axillary inflorescences along with the development of the spathe and the 50 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 TaHI E \. Reprodi jctive systems in the ti ribes of grasses ; + - present. H ermaphro- Tribe ditism Monoecism^ Dioecism^ Apomixis Clei stogamy Arundineae + + + + Chlorideae + + + + + Eragrostideae -1- + + + + Festuceae + +*• +^ +f^ -h Paniceae -*- + + + + Andropogoneae + + + + Aveneae + + + Hordeeae + + + + Pappophoreae + -1- + + " Including andro- and gynomonoecism. '' Including gynodioecism. " Mostly contributed by Poa. spatheole; these form false panicles and it is supposed that they developed from an apparently primitive terminal panicle. As a whole, most of these differences reflect morphological adaptations that were ventured upon in the grasses and have become widespread, but it is im- possible to indicate what has failed to evolve or failed to succeed. This point is well developed by Clifford (1961) in discussing the possible number of symmet- rical arrangements of floral organs that can be derived by reduction, and the number actually present in flowers in the family. Diversity of Reproductive Systems The grasses display a wide variety of breeding systems that define their genetic architecture; these range from self-incompatible hermaphroditism through to dioecism. Hermaphrodite flowers are the most common, being absent from about 20 dioecious genera, and about 50 monoecious ones. The array of floral systems includes a marked andromonoecism and a lesser frequency of gynomonoecism; gynodioecism is rare. Heterostyly is unknown. The extreme form of self-fertilization is seen in habitual cleistogamy, but autogamy or geitonogamy occurs in plants that flower chasmogamically, and self- compatibility is spread widely among tribes. Apomixis is recorded in about 30 genera, but apomixis, like autogamy and geitonogamy, is probably much under- estimated. An attempt at measuring the diversity of reproductive systems among the tribes, based on relatively simple criteria such as the presence of dioecism, mon- oecism, and apomixis (Table 1), indicates that all the major systems are present in very few tribes. The most diverse display lies in five tribes: Arundineae, Chlo- rideae, Eragrostideae, Festuceae, and Paniceae, and a further group of four tribes Andropogoneae, Aveneae, Hordeeae, and Pappophoreae is nearly as variable. However, apomixis is known in only one species in the Hordeeae; dioecism is known only in monotypic Sohnsia in the Pappophoreae; these data unduly im- balance Table 1. Deficiencies in the description of the flowers in many genera 1981] CONNOR— REPRODUCTIVE SYSTEMS IN THE GRAMINEAE 51 make it difficult to interpret the breeding system present, and may be the cause of further imbalance in Table 1 . Among the tribes, and hermaphroditism apart, the most frequently occurring breeding systems are monoecism, including andro- and gynomonoecism, which is known in about 25 tribes, and cleistogamy which is known in about 20 tribes. Reproductive Systems and Their Evolution I propose to consider the reproductive systems in the family, and their pos- sible, or even probable, evolution. This will be fraught with difficulties, not the least of which will be problems of finding true relatives, i.e., the correct placing of some genera in tribes, or the incomplete descriptions of flowers of many gen- era. Occam's razor will have to be used to prevent a plethora of inconsequential pathways. The assumption that hermaphroditism is basic to the grass flower seems in- contestable. Any deviation from hermaphroditism that results in separation of the two sexes — the evolution of various kinds of monoecism and dioecism — is one major departure. It is relatively easy to derive dioecism by postulating a gene for male sterility and another for female sterility, but valid genetic bases for their establishment, and for their maintenance, in natural populations are required as a variety of models show (Charlesworth & Charlesworth, 1978a, 1978b; Lloyd, 1974b, 1975a, 1976; Ross & Weir, 1976). The mutants must find the genetic milieu suitable for fixation or for the development of polymorphism, and develop in dioecism the linkages necessary to control segregation for sex. Similarly, the evolution of self-compatibility from self-incompatibility is easily described, but self-incompatibility must be selected against for self-compatibility to become fixed in a population. Habitual cleistogamy presumes self-compatibility, or at worst, the simultaneous evolution of self-compatibility and of changes in the flowering process. I have chosen to present the steps that are envisaged in the evolution of any particular breeding system always as steps forward towards that system, i.e., as a progression. That regression is possible, e.g., from dioecism to monoecism by restoration of the alternate sex, is not commented upon unless the evidence suggests that this may have significantly occurred. separation of sexes The male and female organs may be separated from the close company of hermaphroditism in a variety of ways, but the simplest expression lies in dioecism and the more complex expression in the various forms of monoecism. Dioecism is less frequent than monoecism in the Gramineae. MONOECISM, ANDROMONOECISM, AND GYNOMONOECISM Monoecious plants are those where both sexes occur on the one plant but wholly or partly in different flowers; in monoecism male and female flowers occur on the same plant; in andromonoecism hermaphrodite (perfect) flowers and male flowers occur on the same plant; in gynomonoecism hermaphrodite flowers and 52 ANNAI.S OF THE MISSOURI BOTANICAL GARDEN [Vol . 6K Table 2. Monoecism in the tribes of grasses; tribes in the order of Hubbard (1973). Monoecism^ Atractocarpeae 3,'' Phareae I Olyreae I Buergersiochloeae Parianeae 1 Centotheceae 3,'* Oryzeae 1 Phyllorachideae 2 I Isachneae 2, ,.^ -) b Aveneae 2,^ Eragrostideae 3,^ Paniceae 2,*^ Lecomielleae 2,^ Andropogoneae 2,** Maydeae 2 Andromonoecism* Bambuseae 3,*^ Arundineae 3,** Danthonieae 3,** Arundinelleae 2,^ Isachneae*' Hordeeae 3 Aveneae 2,3,** Phalarideae 2,3,** Pappophoreae 3,*" (Eragrostideae) Chlorideae 2,** Zoysieae 1 Paniceae*' Andropogonei B 1 Gynomonoecism a (Olyreae) Centotheceae'' Isachneae'' Festuceae 3,*' Eragrostideae' Andropogoneae*" l-flowered. 2 = 2-flowered, 3 = 3- or more-flowered in the genera in which the monoecious states are known. *• Sex differentiation within spikelet. ' Sex differentiation both within and between spikelets. female flowers occur on the same plant. The distribution of these systems among the tribes is shown in Table 2 where the tribes are arranged in the linear order of Hubbard (1973). Because the step from hermaphroditism to andromonoecism or to gynomon- oecism demands a single loss of fertility in each, and because both states are seen as pathways to monoecism, they may be considered first. Gynomonoecism Gynomonoecism is not frequent in the grasses; it is probably only successful in eragrostoid Munroa (Anton & Hunziker, 1978), but is known in a total of 8 genera among the six tribes listed in Table 2, viz., Centotheca, Coelachne, Dian- drolyra, Eriochrysis, Hetcranthoecia, Munroa, Piresia, Poa, (full references in Connor, 1980). For its origin, gynomonoecism demands a loss of male fertility in some flowers, but there must be a decrease in inbreeding to compensate for the reduction in male fertility (Charlesworth & Charlesworth, 1978b). Of the genera listed, there is evidence of self-compatibility only in Poa. Another feature of gynomonoecism is that relatively fewer anthers must pro- vide pollen for a relatively greater number of ovules, a contrast with andromon- 1981] CONNOR— RHPRODUCTIVE SYSTEMS IN THE GRAMINEAE 53 oecism (see p. 58). Anton (1978) gave some examples of the distribution of her- maphrodite and female flowers in Argentinian Poa, which show that there may be as few as 1 anther per ovule and up to 1.7 anthers per ovule. Andromonoecism Andromonoecism, hermaphrodite and male flowers on one plant, is wide- spread. This system is expressed in two major aspects; in one the two sex-forms occur within a single spikelet, and in the other differentiation for sex-form occurs between spikelets. Andromonoecism is associated with the trend towards a single hermaphrodite terminal floret, a syndrome which is found in several tribes, but is at its fullest development in the biflowered spikelets of the large tribes Andropogoneae, Arun- dinelleae, and Paniceae. Andropogoneae.— ThQ Andropogoneae could well be the graveyard of those who wish to interpret the variety of andromonoecism expressed there! At the generic level one may choose between the 20 genera advocated by Roberty (1960) and a current estimate of about 80 genera. In this tribe spikelets are usually paired, one pedicelled and the other sessile, and each is biflowered. The lower floret is usually neuter and the upper her- maphrodite [in abbreviated form S(OH) + P(OH)]^ but the andromonoecious form [S(OH) + P(0M)1 is a common alternative. The range of sex-forms in Table 3 shows that andromonoecism may be expressed in several ways. The essential first step in the evolution of andromonoecism is a loss of some female fertiHty; ovules develop a low genetic value, and those flowers become male. This step is said not to be easily established, and the view of Charles- worth & Charlesworth (1978b) is that, within the limits of their model, andro- monoecism is unlikely to evolve as an outbreeding system. The evolution of andromonoecism in the Andropogoneae would best be seen to have developed along the following pathway. There must first have been the evolution towards biflowered spikelets with hermaphrodite florets; and the de- velopment of paired spikelets. This was followed by the lower florets becoming male through the loss of female fertility. A subsequent mutation affecting the male fertility of those same lower florets would leave an apical hermaphrodite flower in each of the paired spikelets, and thus allow one modal form: S(OH) + P(OH). A further loss of female fertility, this time in the pedicelled spikelet only, would give rise to the andromonoecious form: S(OH) + P(OM). Such progres- sions would be slow to evolve. The time at which the male and female sterility mutations arise during the evolution of the modal sex-forms allows the development of some of the andro- monoecious variants: S(MH) + P(MH) of Andropterum where female sterility alone has operated in the basal flowers; S(MH) + P(OM) of Sehima where the sessile spikelets show an uncompleted series of gene actions; and S(MH) + P(MM) of Rohynsiochloa with even fewer mutations. 3 S = Sessile spikelet, P = pedicelled spikelet; H = hermaphrodite; M = male; F = female; O = neuter. Florets of a spikelet are included by brackets and in the order lower, upper. Symbols linked by + indicate presence on the one plant. 54 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Table 3. Sex-forms in the Andropogoneae. H = hermaphrodite; M = male; F = female; O neuter. If triads are present there is usually an extra pedicelled spikelet. Homogamous pairs, where present, occur below the pairs shown here, and all flowers are M or O. Based on data from Jacques- Felix, Hubbard, Stapf, Pilger, Clayton. Sessile* O H M H M H M H M H M F O F M M M F M M Spikelet Pedicelled^ Andromonoecism O M OO O M M M M H Monoecism 00 O M OF M F FF Elionurus Kerriochloa Sehima Rohynsiochloa Andropterum Triplopogon j\genium Sclerandrium Lophopogon Spikelet Sessile^ OH O H OH OH Pedicelled^ Gynomonoecism O F Eriochrysis Hermaphroditism 00 OH Cleistachne B hide a Erianthus a In each pair the left entry represents lower floret and right the upper floret of spikelet. Internal evidence for the evolution of andromonoecism in this way in the Andropogoneae is sparse; pairs of spikelets in which both flowers of each are hermaphrodite is reported as a mutant in Sorghum, where a dominant gene con- (W 1%5) closest approximation to paired spikelets of totally hermaphrodite flowers is in Andropterum-. S(MH) + P(MH). The main evidence seems to lie in the presence of basal male florets, even in sessile spikelets (Table 3), indicating that the initial loss of female fertility occurred there. In the development of the spikelet the upper floret grows faster and is always larger (Bonnett, 1966). Paniceae and Arundinelleae .—The Paniceae show a more general trend towards a reduced form of hermaphroditism because often the upper floret of a spikelet is a lone hermaphrodite flower. In the biflowered Arundinelleae there is a common pattern of hermaphroditism in the upper floret, and various levels of emasculation and defeminization in the rph 1964) mutation pathway to andromonoecism because in only the lower floret is there any loss of sexuality. Self-fertilization or self-compatibility is not very widely documented in the Andropogoneae, even with the inclusion of genera where cleistogamy is recorded (Connor, 1980). For the Paniceae the information is about equal, though Pas- palum (Burson, 1979) and Lasiacis (Davidse, 1979) may be more firmly included as self-fertile genera. In the Arundinelleae there is no information on the com- patibility relationship of any taxon; however, the anthers in Arundinella pumila, as described by Bor (1955), are very small and suggestive of cleistogamy, but anther length dimorphism occurs in Arundinella, e.g., anthers in male florets in A. holcoides and A. ciliata are about half as long as those in hermaphrodite 1981] CONNOR— RFPROnUCTIVH SYSTEMS IN THE GRAMINEAE 55 flowers. Anther length dimorphism is a feature of andromonoecism (Connor, 1980), but its relationship to self-fertilization or self-incompatibility, or to any aspect of pollination is unknown. On the whole, there may be sufficient self- compatibility to argue the development of andromonoecism in the ways outlined if andromonoecism aids in increased cross-fertilization. Monoecism Monoecism, separate male and female flowers on one plant, is exclusive to some tribes (Table 2). No other sex-form occurs in Buergersiochloeae, Lecom- telleae, Maydeae, Parianeae, Phareae, or Phyllorachideae. Monoecism, together with other sex-forms, occurs in: Andropogoneae, Atractocarpeae, Aveneae, Centotheceae, Isachneae, Olyreae, Oryzeae, and Paniceae. In general, a 1- or 2-flowered spikelet is a characteristic associate of monoecism. In the Atractocarpeae, Isachneae, Aveneae, and some Paniceae monoecism occurs as separate male and female flowers in the one spikelet — perhaps the lowest form of strict monoecism. In the zizanioid Oryzeae, and in many instances in the Andropogoneae, there is separation of spikelets of different sex (see table 3 in Connor, 1980). Among the bambusoid tribes Olyreae, Phareae, Parianeae, Buergersiochloeae, and Phyllorachideae the arrangement of sexually differentiated spikelets in an inflorescence is varied; thus there may be: (i) sexes mixed in an inflorescence, often in paired spikelets with sessile female and pedicelled male as in Maclurolyra or Phams; (ii) sexes segregated on one inflorescence, females above and males below as in Buergersiochloa ; (iii) sexes completely segregated in pairs of racemes as in Ekmanochloa; (iv) sexes completely segregated with male flowers terminal above, and axillary females below as in Humhertochloa. Internal evidence that a gynomonoecious pathway to monoecism is at least possible in the Olyreae, is available from the work of Soderstrom & Calderon (1974) who showed that pedicelled spikelets with an hermaphrodite organization, and sessile female spikelets occur in Diandrolym, and Bahian Piresia. The ped- icellate "hermaphrodite" spikelets, however, fall soon after pollen is shed; ef- fectively, monoecism is achieved. Regardless of this precise pattern, the gyno- monoecious pathway is evident. Charlesworth & Charlesworth (1978b) describe this system as involving firstly a reduction in male fertility in some hermaphrodite flowers to produce female flowers and thus gynomonoecism, followed by a re- duction in female fertility of the hermaphrodite flowers to produce male flowers, a sequence H^H + F^M + F.^ Correlatively with the sterilities there must be a large reduction in inbreeding. Monoecism is seen by the Charlesworths as an uncommon development in flowering plants, and they emphasize that it may ■' H = Hermaphrodite, M = male, F = female; symbols linked by + indicate presence on the one plant. 56 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 well be necessary for two or more mutations to reduce female fertility to such an extent that it will yield a male flower. There is no linkage between sterility genes. Redistribution of male and female flowers into the patterns indicated above would seem to flow easily from a simple, terminal inflorescence of flowers of both sexes to the advanced form of "terminal male and axillary female" inflores- cences. This redistribution could be developing during the sequence of mutations eliminating ovule production of those flowers that were still hermaphrodite. Not every known sex-form is accounted for: e.g., in Lithachne the axillary inflores- cence may contain one apical female-flowered spikelet and several male-flowered spikelets below (Hitchcock & Chase, 1917; Soderstrom & Calderon, 1974), and in PhyUorachis the terminal inflorescence bears male flowers above and female flowers below, while the axillary inflorescence is totally female (Hubbard, 1939). These examples and others represent possible phases in the development of mon- oecism; it is unreasonable to expect evolution conveniently to match every model. The tribe Maydeae is strictly monoecious; here male inflorescences are found above the female ones, there are sessile and pedicelled spikelets in unisexual pairs, and a tendency towards axillary inflorescences of female flowers. The tribe is alternatively included in the strongly andromonoecious Andropogoneae; any projected pathway to monoecism would seem to be that through andromonoe- cism, followed by redistribution of male flowers to terminal inflorescences, and with hermaphrodite flowers being distributed below the males, or to axillary po- sitions. A second mutation causing male sterility in the hermaphrodite flowers of axillary inflorescences would produce female flowers only. In the Zizanieae, or zizanioid Oryzeae, males are present in terminal inflo- rescences, and females in leaf axils below as in Luziola, or are distributed on one inflorescence as female flowers above the male flowers as in Zizania. The path- way to monoecism is expected to be the same as that for the bambusoid tribes, though there is no internal supporting evidence; and Zizania latifolia by not conforming exactly to the "female above, males below" diagnosis but in having inflorescences less perfectly arranged (Bannikova, 1976), in fact offers some evi- dence that sex-forms are being redistributed. The Zizanieae are self-fertile (East, Monoec (We genetics as the other monoecious forms. Between-floret differentiation occurs in the Atractocarpeae, Centotheceae, Eragrostideae, Isachneae and Lecomtelleae, m one or iwo genera or tne Andropogoneae, and Argentmian Hierochloe (Ave- neae). In both the Centotheceae and the Atractocarpeae there are several florets per spikelet. The sex-forms in the Centotheceae range from hermaphroditism through gynomonoecism to monoecism as in: Orthoclada H; Centotheca H -h F; Cal- deronella M -H F and Zeugites M + F. In Zeugites and Caldemnella the lowest floret only is female; but in Centotheca the upper floret is female. Neuter flowers may be present in some members of the Centotheceae, so that in Chevalierella the lowest floret is hermaphrodite and the upper neuter; in Chasmanthium the two lowermost florets are neuter and the remainder hermaphrodite. The internal evidence from Centotheca would point to a gynomonoecious origin of monoe- cism. 1981] CONNOR— REPRODUCTIVH SYSTEMS IN THE GRAMINEAE 57 Blepharidachne (Eragrostideae) has three monoecious species and a fourth, B. kingii, which is hermaphrodite. The pathway here to monoecism is via andro- monoecism. Only one seed is formed in the four-flowered spikelet — that in the penultimate floret which has a functional gynoecium. Anthers may number 1 or 2 or 3 depending on the species (Hunziker & Anton, 1979). In Atractocarpa and Puelia the female flower is apical in the spikelet, the flowers below being male and/or neuter; most related bambusoid genera have hermaphrodite flowers. Hier- ochloe (Aveneae) in Argentina is monoecious (De Paula, 1975), but most species in that genus are andromonoecious. Andromonoecism is the appropriate pathway to monoecism in these genera. In the group of related tribes Isachneae, Lecomtelleae, and Paniceae, sex differentiation mostly occurs within the spikelet: usually the lower flower is male and the upper female, though in Lecomtella there are some spikelets with two male florets and others with one male and one female flower. In the Isachneae because gynomonoecism is present as in Coelachne and Heteranthoecia (p. 51), and andromonoecism in some species of Isachne (Bor, 1952), there is internal evidence of two possible pathways to monoecism. Lecomtella, with an excess of male flowers in some spikelets (Stapf, 1927), may represent what Charlesworth & Charlesworth (1978b) refer to as a reduction in female fertility through a mu- tation increasing the ratio of male to female flowers, making the bisexual form more malelike. Such a character is indicative of an andromonoecious origin. But the Lecomtelleae is a tribe distant from the main stream of evolution in the Paniceae (Brown, 1977), and it would be unwise to transfer too much influence from it to the Paniceae. In the Paniceae, Chamaeraphis and Pseudoraphis are monoecious. An andromonoecious pathway is indicated for members of these tribes. Hygrochloa, aquatic, monoecious, newly recognized, and the only genus in the Paniceae with male and female spikelets separately distributed on the inflorescence, is of remote affinity in the tribe according to Lazarides (1979). Redistribution of sex-form is associated with the most highly developed forms of monoecism; the change from a shoot with a terminal inflorescence to one incorporating axillary inflorescences as well, seems one opportunity in the evo- lution of monoecism. It necessitates a particular morphogenesis, but it seems to be the next most successful evolutionary step in inflorescence development. It is also a very deliberate step towards the possibility of dichogamy — one feature that can be associated with attempts to promote allogamy. The evolutionary pathways to monoecism backed by genetic models of Charlesworth & Charlesworth (1978a, 1978b) seem not unreasonable even though they are heavily weighted in the direction of the avoidance of inbreeding by promotion at least of some way of ensuring outcrossing. Among the monoecious tribes, self-compatibility is known in the Andropo- goneae, Aveneae, Eragrostideae, Maydeae, and Paniceae, but nothing is recorded + of the compatibilities of the Buergersiochloeae, Centotheceae, Olyreae, Pari- aneae, Phareae, and Phyllorachideae (see list in Connor, 1980). Self-compatibility and the associated inbreeding depression are seen by Charlesworth & Charles- worth (1978a, 1978b) to be essential to the pathways they outlined. There is sufficient self-compatibility to satisfy that requirement. Should any of these mon- 58 ANNAl.S OF THR MISSOURI BOTANICAL (JARDHN [Voi . 68 oecious genera be self-incompatible, alternative genetic explanations must be sought. Excess Maleness ''Chez beaucoup d'especes tropicales et surtout chez les Andropogoneae, comme leur nom Tindique, il y a androphilie c'est-a-dire nette predominance des elements males'' (Jacques-Felix, 1962). Clifford (1961) might see these extra anthers as offsetting a reduction to three stamens per flower, because for a wind- pollinating group this common reduction is seen to be disadvantageous. Cruden (1977) and other authors would refei to this in general as producing very high pollen-ovule ratios. Charlesworth & Charlesworth (1978b) might see this predom- inance of male flowers as part of a projected sequence towards monoecism and/ or dioecism where at some stage plants are ''more maleUke'' in that they possess many male flowers. The loss of the ovule function suggests that pollen is of high genetic value and that there have been substantial increases in male fertility; both these are anticipated from their models. In the Andropogoneae the size of the increase in male fertility is measurable in some taxa in two ways; (i) the number of solely male flowers relative to her- maphrodite flowers and (ii) the relative sizes of the anthers found in both. Neither of these is the pollen-ovule ratio of Cruden though they contain its essentials. Differences in size of anthers, depending in which flower they occur, are given for Dihcteropogony Exotheca, and Hyperthalia by Connor (1980). Euclasta con- dylotricha is also described as dimorphic for anther size, those of the male flower being much longer than those of the hermaphrodite flower (Jacques-Felix, 1962); I am unable to find any absolute measurements. There must be other equivalent data for the tribe though I am unaware of them. The relative number of anthers per ovule in a pair of spikelets can be readily calculated from data in Table 3. For many genera with the form S(OH) + P(OM) there are three anthers in the male flower and three in the hermaphrodite flower to 1 ovule, i.e., 6 a : 1 ov. From the form S(OF) + P(OM) the ratio is 3 a : 1 ov, as is the ratio in forms with pairs S(OH) + P(OH). Where there are pairs of homogamous male spikelets subtending the "fertile" part of an inflorescence, there is an immediate increase in anther: ovule ratio. Thus in a genus like Elymcmdni in one raceme with three ovules there may be something of the order of 28 to 46 anthers — a range of 9 a : I ov to 15 a : 1 ov (see fig. 240 in Jacques-Felix, 1962). Where anther size polymorphism occurs as in Hyperthelia edidis (Hubbard, 1950 as Hyparrhenia) each ovule in a raceme is accompanied by three anthers up to 9 mm long in the sessile spikelet, six anthers up to 19 mm long in the pedicelled spikelets, and six anthers up to 4 mm long in the homogamous pair. In //. colohantha, the upper raceme of the pair is reduced to one pedicelled, male spikelet (Clayton, 1967); the ratio here becomes 18 a: I ov for the two racemes. Clayton did not describe anther sizes. The data in Weatherwax (1926) for Oriental Maydeae give anther : ovule ratios for: Chionachne 12-20 a : 1 ov, Coix 180-600 a : 1 ov, Polytoca 30-60 a : 1 ov. Among the tropical Bambusoideae monoecism is well established in the Oly- reae; Soderstrom & Calderon (1979a) and Calderon & Soderstrom (1973) indicate I98I] CONNOR— RFPRODUCTIVE SYSTEMS IN THE GRAMINEAE 59 the frequencies of male and female flowers in some genera. In Maclurolyra for example, the terminal inflorescence bears one female flower and close by 4-6 males giving 12-18 a: 1 ov, and below these may be several pairs of spikelets with one female and one male flower each, i.e., 3 a: 1 ov. Bulbulus is similar; (this genus has been renamed Rehia by Fijten, 1975). In Arberella flaccida the ratio is 15-24 a: 1 ov, in A. costaricensis 24-30 a: 1 ov, in A, dressleri 18-24 a : 1 ov. In Strephium and Raddia — where the two sexes are borne on separate inflo- rescences — there are many more male flowers than female in Strephium, but fewer male flowers than female inflorescences in Raddia. Pariana, the monotypic genus of the Parianeae, may bear 75-105 a: 1 ov, or even more. For cross-fertilization pollen must firstly reach a receptive stigma, only then can the compatibility specificity be expressed. Wind-pollination is a very unspe- cialized pollination system and in the Gramineae is linked to an efficient and specialized incompatibility system. High pollen : ovule ratios are characteristic of anemophily. The Andropogoneae, Olyreae, and Maydeae that have been dis- cussed, all possess high or relatively high, anther: ovule ratios (the precise pollen : ovule ratios I am unable to calculate). Yet the Maydeae are self-fertile (East, 1940) and so are many species of the Andropogoneae. For the tropical Olyreae, wind pollination is very difficult because of the extreme stillness of their forest floor habitat (Davis & Richards, 1933; Whitehead, 1969; Soderstrom & Calderon, 1971, 1979b); further, the leaves cover or protect the abundant axillary inflorescences, limiting what wind action there is. For the herbaceous bambusoids, therefore, the excess of male flowers may be the guar- antor of pollination by an inadequate anemophily or by the secondary develop- ment of entomophily. This explanation cannot be applied to the Andropogoneae of the savanna and of tropical areas elsewhere as far as I know. The Andropogoneae may just need abundant pollen to overcome problems of effective wind pollination; if so, they differ significantly in this respect from many other grasses. DIOECISM Dioecism is not a major phenomenon in the Gramineae, as dioecious species are known from 20 oligotypic genera only. Among the tribes, dioecism is found in the Aeluropodeae, Arundineae, Chlorideae, Eragrostideae, Festuceae, Pani- ceae, and Pappophoreae (Table 4). Gynodioecism, a specialized form of dioecism, if included here does not alter the number of tribes with dioecious taxa. Two tribes are significantly dioecious, the Aeluropodeae and the Chlorideae; in the former there are five dioecious genera and in the latter there are seven. Elsewhere one or two genera in large tribes is characteristic, e.g., Spinifex and Zygochloa are the only two dioecious genera in the Paniceae (Blake, 1941), and only Neer- agrostis (Nicora, 1962) and Scleropogon (Reeder, 1969) in the Eragrostideae. In most genera with dioecious taxa the spikelets are multiflowered; exceptions are the two biflowered panicoid genera Spinifex and Zygochloa; none is single- flowered. For the Arundineae where Gynerium is dioecious and Lamprothyrsus has 60 ANNALS OF THH MISSOURI BOTANICAL GARDEN [Vol . 68 Tabll 4. Dioecious genera; references in Connor (1980). Aeluropodeae: AUolepis, Distichlis, Jouvea, Monanthochloe, Reederochloa. Arundineae: Cortaderia}' Gynerium, Lamprothyrsus. Chlorideae: Bouteloua,^ Buchloe,^ Buchlomimus, Cyclostachya, OpiziOf^ Pringleochloa,^ Soderstrotyiia.^ Eragrostiocae: Neeragrostis, Scleropogon. Festuceae: Festuca (subgenus Leucopoa),'' Poa,''-^' Paniceae; Spinifex, Zygochloa. Pappophoreae: Sohnsia. " Occasionally monoecious. ** Also gynodioecious. ^ Correcting entry in table 1 of Connor (1980). been so at some time — and gynodioecism is active in Cortaderia (Connor, 1974) a gynodioecious pathway to dioecism is evident. Andromonoecism is also present in the tribe in Phm^mites and Goss\veilerochloa\ Gossweilerochloa, amonotypic genus from Angola has lower florets hermaphrodite and upper florets male (Ren- voize, 1979). Neither it nor Phragmites persuades me that a route through an- dromonoecism to monoecism and thence to dioecism is likely in the tribe. The suggested pathway towards dioecism in the Arundineae via gynodioecism involves firstly the establishment of male sterility gene(s) — usually recessive — in some hermaphrodite plants giving rise to hermaphrodite and female plants, i.e., gynodioecism. This is followed by a mutation — usually dominant — for female sterility which acts in the hermaphrodites thus giving male plants. Such a com- bination of alternating gene action will lead to sets of linked genes controlling sex, and to heterogametic male plants. Charlesworth & Charlesworth (1978a) emphasize that this process is a long and extremely slow one. Also important is that self-compatibility and a high level of self-fertilization occur in the group where evolution of dioecism is taking place, and that there is some increase in pollen output by the males. In New Zealand species of Cortaderia, at least, there is abundant self-compatibility in hermaphrodite plants (Connor, 1974). The tribe Chlorideae offers perhaps the greatest concentration of dioecious genera, but Reeder (1969) pointed out that with the exception of Buchlomimus and Cyclostachya, plants of SwcTi/ot', Opizia, Pringleochloa, and Soderstromia may at times be monoecious. Systems currently known in the Chlorideae, apart from dioecism, include hermaphroditism, andromonoecism, and possibly gyno- dioecism in Bouteloua chondrosioides (Reeder & Reeder, 1966; Reeder, 1969). On the basis that monoecism is recurrent among the dioecious taxa (Reeder, 1969), the probable pathway to dioecism for the Chlorideae is through monoe- cism, that is, along pathways to monoecism that have already been discussed (p. 54). It is clear that this pathway is much longer than the gynodioecious one, necessitating firstly the development of andromonoecism as the consequence of reduction in female fertility in some flowers, and subsequently a further reduction in male fertility of the remaining hermaphrodite flowers to produce female flowers and thus the monoecious state. As was discussed earlier (p. 52), low genetic values in the ovules of the hermaphrodite flowers that become male is the major factor. Subsequent steps producing more malelike and more femalelike pheno- types ultimately yield the male and female plants of a dioecious population, Dioe- 1981] CONNOR— REPRODUCTIVE SYSTEMS IN THE GRAMINEAE 61 cism originating in this way will be controlled by sets of linked genes and be the result of an extremely slow evolutionary process. Genetic conditions and selec- tion pressures under which such systems could evolve are detailed by Charles- worth & Charlesworth (1978b). In taxa following this sort of pathway, males and females with traces of the activity, not just the morphology, of the opposite sex may be found (Lloyd, 1975b). In the Chlorideae cleistogamy is common and may be one of the self-fertilizing systems being selected against by dioecism. In the closely related Eragrostideae there are two dioecious genera, Neer- agrostis and Scleropogon, There is internal evidence from within the tribe to indicate a pathway towards dioecism, and the proposition of an escape from inbreeding is quite strong, because self-compatibility and cleistogamy are well established there (see table 8 in Connor, 1980). Ectrosia, which is commonly cleistogamous, may have a large number of male flowers — making it andromon- oecious — or those flowers may be neuter (Hubbard, 1936). Munroa is gynomon- oecious in most species; Blepharidachne. on the other hand, is hermaphrodite in one species and monoecious in three others; on these bases, pathways through monoecism are probable for the dioecious genera. The gynomonoecious origin of dioecism differs only in initial detail from the andromonoecious pathway just described for the Chlorideae. Initially a male ste- rility gene acts in some flowers to produce plants with female and hermaphrodite flowers. Subsequent steps invoke female sterility in the hermaphrodite flowers to produce monoecious plants; from then on the process is that described above. The difficulties alluded to by Charlesworth & Charlesworth (1978b) are unchanged by the chronology of genes reducing fertility. In the Aeluropodeae five genera (or more) are dioecious (Table 4). Yet there is no internal evidence to suggest a pathway to dioecism; Aeluropus itself is hermaphrodite. The Pappophoreae are distinguished on the other hand by her- maphroditism except for the monotypic dioecious genus Sohnsia for which no clear pathway to dioecism is indicated from the floral forms of other genera in the tribe; what is clear is that cleistogamy is a feature of many pappophorean genera (Connor, 1980). For Poa (Festuceae) pathways to dioecism are not difficult to suggest, though none is explicit from internal evidence. In Poa there are gynodioecious species which would suggest this direct pathway, but gynomonoecism is also present and suggests that the much longer monoecious pathway is possible. The choice may lie in recognizing that two pathways could operate. In Festuca, subgenus Leu- copoa, there is sporadic dioecism; no particular pathway is evident. In the Paniceae Spinifex, and its segregate Zygochloa, are the only dioecious genera; they are also the only dioecious genera in the subfamily Panicoideae. The floral arrangement in the biflowered Paniceae — lower floret male or very often neuter, upper floret hermaphrodite — indicates andromonoecism or alternatively hermaphroditism. A possible pathway to dioecism could involve either monoe- cism or gynodioecism. Monoecism is a low frequency phenomenon in the tribe, occurring in Hygrochloa, Chamaeraphis and Pseudoraphis only. Gynodioecism is unknown. The gynodioecious pathway is simpler than the monoecious one because the action of the necessary genes seems more direct, but simplicity is unlikely to be the guarantor of an evolutionary pathway. 62 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 The pathways that have been discussed as being the most probable in the evolution of monoecism and dioecism depend to a large extent on the presence of self-compatibility and self-fertilization. This may appear to suggest that self- compatibility was, or is, very widespread in the family. It is. But if the prede- cessors of 20 dioecious genera and of 45-50 monoecious genera evolved as a reaction to self-compatibility, that number is very small relative to the 600-700 genera in the family. Willson (1979), who does not accept that the advantages of obligate outcross- ing '\ . . present as powerful and all pervasive a force as seems to be commonly assumed/' described an ecological gradient towards dioecism. Sexual selection along one gradient is seen ultimately in an '\ . . overwhelming reproductive suc- cess of some individuals functioning as say, males; selection then would favor the allocation of food resources entirely to male flowers and increase their success still further.'' Genetic advantage to dioecism is not excluded, and Willson sug- gests that both genetic diversity from recombination, and sexual selection were contributors to the evolution of dioecism. HERMAPHRODITISM Variation in the breeding system of hermaphrodite plants is achieved at the genetic level by mutations affecting compatibility relationships, and at another level by changes in the timing of presentation of pollen and of receptive stigmata. Dichogamy does not invariably guarantee a within-plant delay in the maturation of male- and female-bearing organs or in their presentation, often because of the sequential emergence of inflorescences on the one plant (see Burton, 1974). An important evolutionary step in the family is the development of self-com- patibility, a step independent of changes in levels of ploidy (Lundqvist, 1975). Self-compatibility has been identified in about 100 genera in the family (Connor, 1 980) . CLEISTOGAMY Self-fertilization is extensively used as a normal method of reproduction es- pecially in annuals (Stebbins, 1957), and in the specialized form of cleistogamy, self-fertilization has been reported in 18 tribes and from about 70 genera (table 8 in Connor, 1980). This is most probably an underestimate if one judges from the illustrations of so many flowers, and from the anther sizes given in some descriptions. Mostly cleistogamy is facultative, though some taxa are predomi- nantly cleistogamic and some few are known only in the cleistogamic form. Some tribes have greater concentrations of cleistogamous genera than others, e.g., in the Eragrostideae 11 genera bear cleistogamic flowers, and there are 9 genera in the Paniceae, 7 in the Andropogoneae, and 6 in each of the Aveneae, Dantho- nieae, and Festuceae. The evolution of cleistogamy depends initially on the presence of self-com- patibility. Morphological features often associated with nonanthesis, e.g., re- duced anther and lodicule sizes, or even the loss of lodicules, are probably in- terpretable as precocious development (Lord, 1979). In addition to cleistogamous flowers in terminal inflorescences, cleistogamous inflorescences may be hidden I981J CONNOR— RKPkOnUCTIVE SYSTHMS IN THE GRAMINEAE 63 in leaf axils — the cleistogenes of Chase (1918) which 1 prefer to call clandestine inflorescences or spikelets. Clandestine inflorescences are smaller than aerially borne inflorescences, and in the most reduced form are single spikelets only; they are probably precociously developed, because in Stipa leucotricha Dyksterhuis (1945) could record them in plants about six months old. In the same species he found that clandestine spikelets were advanced in seed setting before anthesis had begun in aerial inflo- rescences. This specialized form as clandestine spikelets is more advanced than facul- tative cleistogamy in aerial inflorescences. In all known examples they are a secondary form of seed setting, but usually produce in total fewer seeds than aerially borne inflorescences, though as Hubbard (1933) pointed out, clandestine spikelets may occur in all leaf axils of Cleistochloa ; in Microlaena polynoda very many leaf sheaths are swollen with spikelets (Connor & Matthews, 1977). Clandestine axillary spikelets are known in 13 genera, viz., Aristida. Calyp- tochloa, Cleistochloa, Cleistogenes, Cottea, Danthonia, Diplachne, Enneapo- gon, Microlaena, Muhlenhergia, Sieglingia, Stipa, Triplasis; full references are in Connor (1980). Aerial inflorescences in all these genera bear both cleistogamic and chasmogamous florets, except Cleistochloa where only chasmogamic flowers are known (Hubbard, 1933). At another level of development, cleistogamy is found in subterranean spike- lets; these spikelets are most probably more highly evolved than those that arise from the conversion of an axillary vegetative bud to a reproductive one as in clandestine spikelets. Subterranean spikelets are reported from four genera only: Chloris (Chlorideae), Amphicarpum and Paspalum (Paniceae) and Eremitis (Par- ianeae). In Eremitis these subterranean spikelets were only recently discovered (Soderstrom & Calderon, 1974), but the well-known annual Amphicarpum purshii was subject to experiment by McNamara & Quinn (1977) who showed the high level of dependence by this taxon on subterranean spikelets. Cleistogamy was suggested by Clifford (1961) as one possible response fol- lowing a change from entomophily to anemophily in the Gramineae. For cleis- togamy to develop the sole requirement is that it must be preceded by a mutation to self-compatibility; this mutation and the development of ancillary features of the cleistogamic habit were unlikely to be simultaneous. Cleistogamy occurs in plants at both high and low latitudes, at high and low altitudes, in forest and grassland, in annuals and perennials, in archaic tribes and in advanced tribes, in tribes significantly associated with monoecism and dioecism and in tribes that are not. Cleistogamy is known to be a response to differences in daylength, to be a response to soil moisture variation, to co-occur simulta- neously with chasmogamy in the same plant, to have differing modes of polli- nation, and to occur before or after inflorescence emergence. Many of these responses coincide with ecological demands, but ecological or ecophysiological interpretations do not totally account for all occurrences of cleistogamy and cer- tainly not for simultaneous cleistogamy and chasmogamy on the one plant. Cleistogamy, like other breeding systems, is one evolutionary possibility open to the grasses; they adopted it as a specialized self-compatibility system and later 64 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 adapted it in the form of axillary clandestine spikelets, and to a lesser extent in subterranean spikelets. Its frequency is a measure of its success. APOMIXIS Departures from and variants on hermaphroditism are pathways open to the grasses, and a departure from sexual reproduction yet another. I do not propose to discuss proliferation though it is important at high latitudes, and intend only very briefly to discuss apomixis because one either discusses embryological de- tail, or environmental control, or the genetics of inheritance, or invariability in the next generation. All these topics have been most adequately treated elsewhere (Stebbins, 1941; Gustafsson, 1946-1947; Nygren, 1954, 1967; Battaglia, 1963; Connor, 1980). The distribution of apomixis in the grasses (Table 5) is of no significant value in any systematic interpretation of tribes or lesser ranks. In the Andropogoneae, Paniceae, and closely related tribes, apomixis has been reported in about 20 genera, something of the order of 10% of the genera. It is significantly absent, or unrecorded, in the Bambusoideae as interpreted by Soderstrom & Calderon (1979b). The Arundinoideae, which always seems central to any discussion on the family, includes two apomictic genera; there are three in the Festucoideae, and three in the Chloridoideae-Eragrostoideae. At the genetic level, somatic apospory is more frequent in the Panicoideae than elsewhere, but neither it nor gonial apospory (diplospory) is exclusive to any subdivision of the family. The absence of pollen stimulation, i.e., nonpseu- dogamy. is known only in the five nonpanicoid genera: Calamagrostis, Carta- deria, Lamprothyrsus, Nardus, Poa; autonomous apomixis is a more independent state than pseudogamy but possesses little scope for versatility. Apomixis is associated with monoecism in Tripsacum only, and with dioecism in the special cases where females alone are now known — Cortaderia and Lam- prothyrsus. Apomixis occurs in self-incompatible and self-compatible hermaph- roditic flowers. In the Andropogoneae it has often replaced andromonoecism. Apomixis is clearly a recent development in the Gramineae and coincident with the universal observations of hybridization and polyploidy. In many genera apomixis is readily reversible with sexual reproduction be- cause apomixis is rarely totally obligate. Such a capability seems more advan- tageous than that in the nearest approximation, that is, in the switch from cleis- togamic to chasmogamic flowering. Whether the reversals from dioecism as reported for the Chlorideae by Reeder (1969), or any other reversal, is as frequent or as balanced as the sexual-asexual switch among apomicts is unknown. Response to the Breeding Systems Is there an optimal breeding system for the Gramineae? The question contains a teleological element, but teleology is necessary to allow the formulation of satisfactory hypotheses. Theory suggests two possibilities for maximum hetero- zygosity, for recombinants, and for maximum capacity to react to major envi- ronmental changes; these are dioecism and multiallelic incompatibility genes — I981J COSINOR— RHPRODLCTIVE SYSTEVIS IN THE CJRAMINEAE 65 1 ABLE 5. Apomixis in the tribes of grasses; tribes in the order of Hi Libbard (1973); references in Connor (1980). -h - present; - absent; SI - self-incompatible; SC - self-compatible; MS - male sterile. Apospory T^ -^— Pseu- Tribe Gonial Somatic dogamy Genera Sex-Form Arundineae + 2; CortaJeria, Lamprothyrsus in MS only Hordeeae + + 1; Agropyron hermaphrodite; SC Fesluceae + + +/- 1 ; Poa hermaphrodite or MS Aveneae + -f 1 ; Hierochloe andromonoecious Agrostideae + +/- 1; Calamagrostis hermaphrodite; SI, SC Nardeae + I; Nardus hermaphrodite or MS Eragrostideae + + 1 ; Eragrostis hermaphrodite Chlorideae + + 2; Boiiteloua, Chloris hermaphrodite; SI Paniceae +^ + + 9; Cenchrus, Punicum, etc. andromonoecious or hermaphrodite Anthephoreae + 1; Anthephora andromonoecious Andropogoneae +^ + + 9; Bothriochloci, Dichcmthium, etc, 1 ; Tripsacum andromonoecious Maydeae + + monoecious a Reported in Paspalum only. ^ Reported in Saccharum only. the S genes. Other systems seem less adequate on a variety of grounds. Hetero- styly is absent. In the evolution of the breeding systems discussed so far, there is a strong element of the avoidance of self-fertilization especially through monoecism, in- cluding andromonoecism and gynomonoecism, and protandry and protogyny in hermaphrodite flowers. Monoecism of itself is not necessarily a breeding system that promotes the avoidance of self-fertilization because self-fertile taxa will re- main self-fertile after the evolution of monoecism. Monoecism increases the op- portunity for dichogamy but does not generate it. Andromonoecism is established in one of the more successful groups of the world's grasses, the Panicoideae; the andromonoecious habit has no special merit as a breeding system unless increased male fertility, measured in terms of pollen production, is demanded by the genetic situation. In the Andropogoneae the tendency to become exceedingly male, and consequently with fewer seed-bearing florets, is one of the significant enigmas of the tribe (Clayton, 1967; Olorode & Baquar, 1976; Connor, 1980). Charlesworth & Charlesworth (1978b) do not see andromonoecism evolving as an outbreeding device. Gynomonoecism. occuring only sporadically in five or six tribes, has only been studied as a breeding system in Poa annua (Ellis et al., 1971; Ellis, 1974). Without strong dichogamy, self-fertilization is unlikely to be totally avoided. Gy- nomonoecism has its main significance as one point on an evolutionary pathway to monoecism, but the evidence from within the tribes in which monoecism oc- curs, does not suggest that any of the highly evolved monoecious sex-forms is related to currently gynomonoecious groups except in Diandrolyra (see p. 54). Hermaphroditism is very often associated with self-compatibility, and exper- imentally verified self-fertilization without significant inbreeding depression has 66 ANNALS OF THfc: MISSOURI BOIANICAL GARDEN [Voi . 68 been established in 45 genera; to these must be added those genera where cleis- togamy is easily verified, bringing the total to about 100 genera (see Connor, 1980). In many of these genera there are self-incompatible species as well, e.g., in Bromiis\ Lolium, Phaluris, and Poa. It is sufficient to indicate that the mutation to self-compatibility is, or was, a frequent development, and because this derived self-compatibility is present in simple and advanced tribes, it may be assumed that the tendency to mutate is not much younger than the tribes themselves. Apomixis is generally interpreted as a system lacking evolutionary advantage, but as it is more frequently facultative than obligate, there are opportunities for the formation of new recombinants. Should mitotic recombination as discussed by Stern (1936), Evans & Paddock (1976), and Harrison & Carpenter (1977) be of widespread occurrence, some of the opprobrium attached to apomixis may be withdrawn, and this system may be admitted into the group of genetically diverse reproductive systems. Best suited to anemophily as the pollen dispersal system, and to the require- ment for the maximum number of recombinants, are dioecism and multiallelic incompatibility. DIOECISM Dioecism leaves little option but for obligate cross-pollination; the assumption is usually made that all pollinations are compatible. In strict dioecism each sex- form makes an equal genetic contribution to the next generation, but there must also be: (i) chromosomal sex determination, or its equivalent, (ii) an efficient pollen transfer mechanism, and (iii) a normal sex-ratio. Should there be an eco- logical differentiation favoring one sex over the other as shown by Freeman et al. (1976) for Distichlis spicata and for other families by Lloyd & Webb (1978), then the distribution of the sex-forms, both in number and place, may affect the efficiency of dioecism. Dioecism in the Gramineae is at a low frequency — about 20 genera from seven tribes. The genetics of dioecism is unexplored in the family. One count of males and females in Biuhloe dactyloides, that of Schaffner (1920), whose results in- dicated .50M : .50F in the field, this together with a small sample of seed giving .50M : ,50F, constitute the major published facts of dioecism. Voigt et al. (1975) indicated that with selection a high frequency of female plants could be obtained in B, dactyloides but presented no details. Based on quadrat counts of inflores- cences, frequencies in the rhizomatous grass Distichlis spicata are: .49M : .5 IF, but the sexes are not distributed at random (Freeman et al. 1976). In Bouteloua chondrosioides .31M: .69F was obtained from seed (Reeder 8l Reeder, 1966); this is a very wide departure from .50M : .50F. Should B. chondrosioides be gynodioecious these valuable results are uninterpretable because of the uncer- tainty about the sex-form of the seed parents. There is no doubt, however, that there is an excess of female plants — 57M and 118F — among the three families raised from seed; Lloyd (1974a) discusses the topic of female predominant sex- ratios. Differentiation associated with dioecism is better known. There are, for ex- ample, more florets per spikelets in male plants than in females in Jouvea and 1981] CONNOR— REPRODUCriVH SYSTEMS IN FHE GRAMINEAE 67 Monanthochloe (Villamil, 1969; Weatherwax, 1939), and the greater number of male florets is assumed to be related to the supply of pollen to female spikelets that are hidden among the leaves. In Jouvea the males are borne well above the leaves, but in Monanthochloe the male flowers, hke female flowers, are hidden. The fit to the model is imperfect. The less specialized form of dioecism, gynodioecism, is restricted to two genera, Cortaderia and Poa, but there are three genera if Bouteloua is admitted. In Poa there is an array of species (see summary in Connor, 1980); in Cortaderia there are about 20 gynodioecious species and 5 that are apomictic (Connor, 1974; Philipson, 1978; Costas-Lippmann, 1979). In Cortaderia hermaphrodite plants are heterogametic. Some hermaphrodites are self-compatible, but South American Cortaderia selloana behaves for the greater part as a dioecious species. Hermaphrodites are self-incompatible, and at most produce a few, poorly germinable seeds; their contribution to subsequent generations is chiefly as pollen parents. Natural populations do not depart from .50M : .50F. The genetics of male sterility are complex, though the behavior in natural populations is simple; thus .50M : .50F is reproducible in nature because female plants are pollinated by plants themselves derived from crosses of the same kind, and therefore heterozygous. Both Lloyd and I (Connor, 1974; Lloyd, 1976) have shown how .50M : .50F is maintained in natural populations. Male-fertility is dominant over male-sterility, and it is possible in experiments to produce populations of male plants. In some experimental crosses, however, the frequency of females may be greatly in excess of the 50% so commonly found in my experiments and in natural populations (Connor, 1974), and reach 80%. I have no ready genetic model to fit these data, but am attracted on the one hand to explanation for imbalanced sex ratios based on meiotic drive (see Sweeney & Barr, 1978; Hastings & Wood, 1978), and on the other, to cytoplasmic interaction. This discussion of a subdioecious species is to present data of the kind that is needed for dioecism. Elsewhere I have discussed the self-compatible and strict- ly gynodioecious New Zealand species of Cortaderia (Connor, 1965, 1974). The small number of dioecious taxa and their distribution among tribes may be an indication that dioecism has been found wanting as an evolutionary pos- sibility within the family. Relative to other families where dioecism is present, dioecism in grasses — about 3% of the genera — is at a low frequency. The greatest concentration of dioecious grasses is in the New World (Reeder, 1969), and it will be interesting to see if the ecological selection for dioecism suggested by Willson (1979) can be invoked to explain any features associated with this pattern; some features of distribution of the taxa are clearly related to major ecological char- acters. Gynodioecism is so rare in the family as to not merit much attention as an optimal breeding system. On a pathway to dioecism, such as is discussed by Charlesworth & Charlesworth (1978a), Lloyd (1974a, 1974b, 1975a, 1976), Ross (1970, 1978), Ross & Shaw (1971), Ross & Weir (1976), and Webb (1979), gyno- dioecious grass is of some evolutionary interest; as a result of occasional but nonpersistent male sterility in many species it is of some interest to plant breed- ers. 68 ANNAI.S OF THH MISSOURI BOTANICAL GARDEN [Vol. 68 INCOMPATIBILITY Lundqvist in 1954 demonstrated the genetics of self-compatibility in Secede cereale, and in later papers he showed that the system occurred widely enough in the family to be considered the grass or S-Z system (Lundqvist, 1955, 1962, 1965, 1969). Work in BrizUy Phahiris and other grasses has been in full agreement (Hayman, 1956; Murray, 1974); in Lolium Hayward & Wright (1971) and Spoor (1976) obtained some inconsistent results, but Cornish et al. (1979a) obtained an orthodox fit to the grass system. Incompatibility in the Gramineae is unique and its characters include: (i) 3- nucleate pollen (Brewbaker, 1957); (ii) dry stigmas (Heslop-Harrison & Shivanna, 1977); (iii) action for the most part, but not exclusively, on the stigmatic surface; (iv) two independent multi-allelic S loci {S and Z) with functional dependence in a complementary and cooperative action such that the presence of an identical specificity, e.g., SiZ^ in both the pollen and the stigma, results in incompatibility (Lundqvist, 1975). Most of these characters are those recorded where incompat- ibility is sporophytically controlled, but incompatibility in the Gramineae is ga- metophytic. Figures 8 and 9 in de Nettancourt (1977) clearly demonstrate the compatibilities and incompatibilities of the system. Lundqvist (1975, et praec.) considered that the S-Z system had arisen by duplication of a one-locus system. Pandey (1977, 1980), who considers that the complementary incompatibility system is a secondary one derived from the prim- itive one-locus multiallelic system, accepts duplication of an 5-gene as one major part of the evolution of the two-loci system. He envisages firstly the breakdown of the original self-incompatibility and the development of self-compatibility. This is followed by the reintroduction of self-incompatibility genes through recurrent inter- or intraspecific hybridization, and after selection against S competition and dominance to allow these fundamental characters of this system to evolve, a duplicate, independently acting self-incompatibility locus is acquired. Self-incom- patibility has reevolved, though the characteristics of the system are now novel and may be less efficient than the original system. Charlesworth & Charlesworth (1979) consider that the grass system is most likely to have arisen when '\ . . a variety of (allelic) specificities is initially present without causing an incompati- bility reaction, and that selection acts on loci (other than the S locus) which affect the chance that pollen which matches an allele in the stigma-bearing plant will be rejected.'' Lundqvist (1975), de Nettancourt (1977), and Pandey (1977) emphasize the benefit of the bifactorial system of the grasses over the monofactorial system common to many other families. Because the number of specificities in a popu- lation is the product of the number of segregating alleles in each locus, the two- loci system ^\ . , confers on the population a reduction of cross-incompatibility between plants and a minimized risk of random loss of members of its valuable pool of incompatibility genes" (Lundqvist, 1975). Recent contributions to the genetics of the two loci are from Charlesworth (1979) and from Cornish et al. (1979b). The grass incompatibility system prevents fertilization between genotypes identical in incompatibility alleles, and like other incompatibility systems protects 1981] CONNOR— REPRODUCTIVE SYSTEMS IN THE GRAMINEAE 69 and regenerates itself. Advantages are reflected in recombination and reconsti- tution; and for the Gramineae if these quaHties are desirable, they are achieved by the retention of a versatile, advanced incompatibility system. This efficient incompatibihty system has other attendants. The grasses show what Stebbins (1974) called an extreme adaptation for wind-pollination and cross- fertilization: lodicules controlled by weather; elongation of the stamen filament leading to anthers being presented in a new orientation, pollen quickly shed into the air, abundant, light, easily dispersed; large easily exserted stigmata. Yet, none of these attributes prevents (i) a high level of compatibility between sibs where seed dispersal results in a very narrow distribution, (ii) the mutation self-incom- patibility to self-compatibility, (iii) the evolution of the diverse floral array seen in the family. Pandey (1977) even sees the pseudocompatibility and the high level of sib- compatibility that arise from the two-loci incompatibility system, as extremely advantageous in long-distance dispersal from a single diaspore; both these char- acters are usually interpreted as the least efficient aspects of the two-loci com- plementary system. The accidental arrival of a second diaspore from a long dis- tance, and a second accident is as likely as the first, will restore the incompatibility system. The reevolution of a self-incompatibility system from derived self-compati- bility as described by Pandey (1977, 1980) is of very low likelihood if the taxa had long been self-compatible. Added to this, there is no known self-incompatibility system in the grasses that could have had a two-loci complementary system as its progenitor; if such a system were to evolve, it would be of the less efficient one-locus Tradescantia type. For those self-compatible grasses where an out- breeding system would be advantageous and which selection might favor, the choice is among various forms of monoecism and dioecism. The proposals of Charlesworth & Charlesworth (1978a, 1978b) for the evolution of these states demand the presence of high levels of self-compatibility; these proposals accord with the potentialities indicated by Pandey (1977, 1980). Conclusion Is there an optimal breeding system for the Gramineae? Self-incompatible hermaphrodite flowers in many-flowered spikelets, in inflorescences of many spikelets would seem a very efficient option; such a combination of characters would annually yield a large number of seeds of varied genotype, and no sec- ondary characters would be involved. This state is reached in its fullest expression in the Festuceae, and in some related tribes in the Festucoideae, such as Aveneae, Bromeae, and Hordeeae. Such a conclusion is valid only within the narrow con- fines of the questions being examined. The conclusion makes no allowance for any other factors preceding seed setting, e.g., floral induction, or consequent upon seed setting, e.g., seed dispersal; it makes no allowance for any ecological considerations, e.g., duration of growing season; it makes no allowance for life- history, e.g., annuals or perennials. And it makes no allowance for the natural distribution of the tribes of grasses on the face of the earth. 70 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Voi . 68 The Festucoideae are not especially well known for alternative reproductive systems, self-compatibility excepted; apomixis is known in Calamagrostis, Hier- ochloe, Poci, and very slightly in Agropyron\ there is very sparse dioecism, and little monoecism, much of it in Poa, Members of this subfamily show that, al- though apparently possessing many of the characters that would seem the most advantageous, other breeding systems have evolved if even only sporadically, or are now the residuum of evolution. Any analysis leading to the conclusion favoring the festucoids as possessing the most efficient breeding system in the grasses could be a biased one, biased by the overwhelming evidence from temperate grasses. That so many grasses differ from the festucean model reflects the fact that the adaptive values for grasses include more than the requirement of a self-generating incompatibility system, and indicates that one seed in an andromonoecious spathate raceme, or monoecious axillary paniculodium, may well have equal or superior adaptive values in the tropics and subtropics. The Gramineae, if viewed solely by the array of breeding systems found there, is an advanced family in the Monocotyledonae. Stebbins (1974) found that the grasses are ''. . . the climax of flowering-plant evolution . . . ,'' but his conclusion was logically based on an examination of all the qualities in the family. Bews (1929) was astonished that such an unspecialized system of pollination could be associated with so successful a family. The grasses show relative conservatism in the arrangement of the flower; in the development of the breeding systems associated with those flowers there is relatively less conservatism, though never flamboyance. The major reproductive systems known in the Plant Kingdom are represented except heterostyly, spo- rophytic control of incompatibility, and overt insect pollination, though ento- mophily probably occurs in the tropics. Although there are some unexpected features of the incompatibility system (Lundqvist, 1954, et seq.), and some unique conditions at pollination (Heslop- Harrison, 1979, 1980), self-incompatibility is basic to the family. For an her- maphroditic, anemophilous, self-incompatible group, optional ways of ensuring seed setting are those leading towards (i) self-compatibility in demanding eco- logical conditions; (ii) production of abundant pollen to ensure cross-fertilization in competitive genetic conditions; (iii) espousal of apomixis in conditions of polyploidy, and of hybridism. Under self-compatibility as described in (i), initial significant inbreeding depression would be the major disadvantage; for the cross- pollination proposition an excess of flowers that are pollen producers only (an- dromonoecism) places a low genetic value on ovules and leads to their loss from flowers; and apomixis has always been selected against, resulting in the faculta- tive rather than the obligate form. There was an early and frequent loss of the incompatibility system without polyploidy as a direct cause, yet self-compatibility is an essential prerequisite only for the development of cleistogamy. Dioecism could evolve from self-com- patible progenitors, as could monoecism. For self-compatible grasses an escape is dioecism; monoecism does not guarantee that surcease. Hermaphroditism, alone or in some monoecious combination, sustains most of the world's grasses, and those grasses that are the sustenance of the world. 1981] CONNOR— REPRODUCTIVE SYSTEMS IN THE GRAMINEAE 71 LiTERATURK CllED Anton. A. M. 1978. Notas crilicas sobre gramineas de Argentina. III. Contribucion a) conocimienlo de la sexualidad en Poa. Bol. Acad. Nac. Ci. 52: 267-275. & A. T. HuNZiKER. 1978. El genero Munroa (Poaceae): Sinopsis morfologica y laxonomica. Bol. Acad. Nac. Ci. 52: 229-252. Arber, a. 1934. The Gramineae. Cambridge Univ. Press, Cambridge. 480 pp. Bannikova, V. A. 1976. Morphology of panicle and peculiarities of flowering of Zizania latifoliu (Griseb.) Stapf. Bot. 2urn. (Moscow & Leningrad) 61: 990-993. Barnard, C. 1955. Histogenesis of the inflorescence and flower of Triticum acstivum L. Austral. J. Bot. 3: 1-20. . 1957. Floral histogenesis in the monocotyledons. I. The Gramineae. Austral. J. Bot. 5: 1-20. . 1964. Form and structure. Pp. 47-72, in C. Barnard (editor). Grasses and Grasslands. Mac- millan & Co., London. 269 pp. BATTACii lA, E. 1963. Apomixis. Pp. 221-264, in P. Maheshwari (editor). Recent Advances in the Embryology of Angiosperms. Int. Soc. Plant Morphologists, Univ. of Dehli. Bews, J. W. 1929. The World's Grasses. Their Differentiation, Distribution, Economics and Ecol- ogy. Longmans, Green & Co., London. 408 pp. Bi AKE, S. T. 1941. New genera of Australian grasses. Univ. Queensland Dept. Biol. Pap. 1 (No. 19): 1-12. Bonnett, O. T. 1966. Inflorescences of maize, wheat, rye, barley, and oats. Univ. Illinois Agric, Exp. Sta, Bull. 721: 1-105. BoR, N. L. 1952. Notes on Asiatic grasses: X. Kew Bull. 1952: 321-323. . 1955. Notes on Asiatic grasses: XXIV. Kew Bull. 1955: 377-414. Brewbaker, J. L. 1957. Pollen cytology and self-incompatibility systems in plants. J. Hered. 48: 271-277. Brown, W. V. 1977. The Kranz syndrome and its subtypes in grass systematics. Mem. Torrey Bot. Club 23(3): 1-97. BuRSON. B. L. 1979. Cytogenetics of Paspalum urvillei x P. intermedium and P. dilatatum x P. paniculatum hybrids. Crop Sci. 19: 534-538. Burton, G. W. 1974. Factors affecting pollen movement and natural crossing in pearl millet. Crop Sci. 14: 802-805. Calderon, C. E. & T. R. Soderstrom. 1973. Morphological and anatomical considerations of the grass subfamily Bambusoideae based on the new genus Maclurolyra. Smithsonian Contr. Bot. 11: 1-55. Chareesworth, B. & D. Charlesworth. 1978a. A model for the evolution of dioecy and gyno- dioecy. Amer. Naturalist 112: 975-997. & , 1978b. Population genetics of partial male-sterility apd the evolution of monoecy and dioecy. Heredity 41: 137-153. & . 1979. The evolution and breakdown of .S-allele systems. Heredity 43: 41-55. Charlesworih, D. 1979. Some properties of populations with multilocus homomorphic gameto- phytic incompatibility systems. Heredity 43: 19-25. Chase, A. 1918. Axillary cleistogenes in some American grasses. Amer, J. Bot. 5; 254-258. Clayton. W. D. 1967. Studies in the Gramineae: XII. Kew Bull, 20: 433-^49. . 1969. A revision of the genus Hyparrhenia. Kew Bull., Add. Ser. 2. 1% pp. Clifford, H. T. 1961. Floral evolution in the family Gramineae. Evolution 15: 455-460. Connor, H. E. 1963. Growth of flowering stems in five New Zealand tussock grasses. New Zealand J. Bot. 1: 149-165. . I%5. Breeding systems in New Zealand grasses. VL New Zealand J. Bot. 3: 233-242. . 1974. Breeding systems in Cortaderia (Gramineae). Evolution 27: 663-678. , 1980. Breeding systems in the grasses — a survey. New Zealand J. Bot. 17: 547-574. & B. A. MAirHHWS. 1977. Breeding systems in New Zealand grasses. VOL New Zealand J. Bol. 15: 531-534. Cooper, J. P. 1952. Studies on growth and development in Lolium. IIL Influence of season and latitude on ear emergence. J. Ecol. 40: 352-379. Cornish, M. A., M. D. Hayward & M. J. Lawrence. 1979a. Self-incompatibility in ryegrass. I. Genetic control in diploid Lolium perenne L. Heredity 43: 95-106. , & . 1979b. Self-incompatibility in ryegrass. II. The joint segregation of S and Z in Lolium perenne L. Heredity 43: 129-136. CoSTAS-LiPPMANN, M. 1979. Embryogeny of Cortaderia selloana and C. jubata (Gramineae). Bot. Gaz. (Crawfordsville) 140: 393-397. 72 ANNAI.S OF THE MISSOURI BOTANICAL GARDEN [Vol . 68 CRUDtN, R. W. 1977. Pollen-ovule ratios: a conservative indicator of breeding systems in flowering plants. Evolution 31: 32^6. Davidse, G. 1979. A systematic study of the genus Lasiacis (Gramineae:Paniceae). Ann. Missouri Bot. Gard. 65: 1133-1254. Davis, T. A. W. & P. W. Ric mards. 1933. The vegetation of Morahalli Creek British Guiana: an ecological study of a limited area of tropical rain forest. Pt 1. J. Ecol. 21: 350-384. Dt- Paula, M. E. 1975. Las especies del gcnero Hierochhe (Gramineae) de Argentina y Chile. Darwiniana 19: 422^57. Dyksterhuis, E. J. 1945. Axillary cleistogenes in Stipa leucotricha and their role in nature. Ecology 26: 195-199. East, E. M. 1940. The distribution of self-slcrility in the flowering plants. Proc. Amer. Phil. Soc. 82: 449-518. Ellis, W. M. 1974. The breeding system and variation in populations of Poa annua L. Evolution 27: 656-662. , B. T. O. Lee & D, M. Calder. 1971. A biometric analysis of populations o{ Poa annua L. Evolution 25: 29-37. Emerson, R. A. 1924. Control of flowering in leosinte. J. Heredity 15: 41-48. Evans, D. A. & E. F. Paddock. 1976. Comparisons of somatic crossing over frequency in Nicotiana tahacum and three other crop species. Canad. J. Genet. Cytol. 18: 57-65. Evans, L. T. 1964. Reproduction. Pp. 126-153, in C. Barnard (editor), Grasses and Grasslands. Macmillan & Co., London. 269 pp. FijTEN, F. 1975. A taxonomic revision o( Bucr^ersiochloa Pilg. (Gramineae). Blumea 22: 415-418. Freeman, D. C, L. G. Ki ikom & K. T. Harper. 1976. Differential resource utilization by the sexes of dioecious plants. Science 193: 597-599. Gustafsson, a. 1946-1947. Apomixis in Higher Plants. C. W. K. Gleerup, Lund. 370 pp. Harrison, B. J. & R. Carpenter. 1977. Somatic crossing-over in Antirrhinum majus. Heredity 38; 169-189. Hastings, R. J. & R. J. Wood. 1978. Meiotic drive at the D (M'') locus and fertility in the mosquito, Aedes aei^ypli (L.). Genelica 49: 159-163. Hayman, D. L. 1956. The genetical control of incompatibility in Phalaris coeruh'scens Desf. Aus- tral. J. Biol. Sci. 9: 321-331. Havward, M. D. & a. J. Wright. 1971. The genetic control of incompatibility in LoUum pcrenne L. Genetica42: 414^21. Heslop-Harrison, J. 1959. Photoperiod and fertility in RotthoeUia exaltata L.f, Ann. Bot. (London) 23: 345-349. , l%I. The function of the glume pit and the control of cleistogamy in Bothriochtoa dccipiem V (Hack.) C. E. Hubbard. Phylomorphology II: 378-383. . 1979. Aspects of the structure, cytochemistry and germination of the pollen of rye (Secale cereale L.). Ann. Bot. (London) 44 Suppl. 46 pp. . 1980. The pollen-stigma interaction in grasses: a review. New Zealand J. Bot. 17: 537-546. Hesiof-Harrison, Y, & K. R. Shivanna. 1977. The receptive surface of the Angiosperm stigma. Ann. Bot. (London) 41: 1233-1258. Hitchcock, A. S. & A. Chase. 1917. Grasses of the West Indies. Contr. U.S. Natl. Herb. 18: 261- 471. Hodgson, H. J. 1966. Floral initiation in Alaskan Gramineae. Bot. Gaz. (Crawfordsville) 127: 64- 70. Hubbard. C. E. 1933. Clcisfochloa suhjuncca C. E. Hubbard. Hooker's Icon. PI. tab. 3209. . 1936. Ectrosia leporina R.Bi. Hooker's Icon. PI. (ah. 3312, . 1939. Phyllorachis sa^^ittata Trimen. Hooker's Icon. PI. tab. 3386. -^. 1950. Hyparrhenia edulis C. E. Hubbard. Hooker's Icon. PI. tab. 3495 A et B. . 1973. Gramineae. Pp. 504-508, //; A Dictionary of the Flowering Plants and Ferns. Ed. 8, Revised by H. K. Airy Shaw. Cambridge Univ. Press, Cambridge. HuN/iKER, A. T. & A. M. Anton. 1979. A synoptical revision of Blepharidachnc (Poaceae). Brit- tonia31: 446-^53. Jacques-Felix, H. 1962. Les Graminees (Poaceae) D'Afrique Tropicale. 1. Generalites, Classifi- cation, Description des Genres. Bulletin Scientifique No. 8, Institut de Recherches Agronomiques Tropicales et des Cultures Vivrieres. 345 pp. Knox, R. B. 1967. Apomixis: seasonal and population differences in a grass. Science 157: 325-326. & J. Heslop-Harrison. 1966. Control of pollen fertility through the agency of the light re- gime in the grass Dichanthium aristatum, Phyton (Horn) 11: 256-267. Langer, R. H. M. & D. Wilson. I%5. Environmental control of cleistogamy in prairie grass {Bromus unioloides H. B. K.). New Phytol. 64: 80-85. 1981] CONNOR— RHFROOUCTIVE SYSTHMS IN THE GRAMINEAE 73 Lazarides, M, 1979. Hygrochloa, a new genus of aquatic grasses from the Northern Territory. Bru- nonia2: 85-91. Lloyd, D. G. 1974a. Female-predominant \AA 1974b. The genetic contributions of individual males and females in dioecious and gynodi- oecious Angiosperms. Heredity 32: 45-51. 1975a. The maintenance of gynodioecy and androdioecy in Angiosperms. Genetica 45: 325- 339. 1975b. Breeding systems in CotuUi. III. Dioecious populations. New Phylol. 74: 109-123. — , 1976. The transmission of genes via pollen and ovules in gynodioecious Angiosperms. Theor. Populat. Biol. 9: 299-316. & C. J. Webb. 1978. Secondary sex characters in plants. Bot. Rev. (Lancaster) 43: 177-216 ■ m Lord, E. 1979. The development of cleistogamous and chasmogamous flowers in Lamium aniplex- icaule (Labiatae): an example of heteroblastic inflorescence development. Bot. Gaz. (Crawfords- ville) 140: 39-50. LuNDQViST, A. 1954. Studies on self-sterility in rye, Sccalc cereale L. Hereditas 40: 278-294. . 1955. Genetics of self-incompatibility in Festuca pratensis Huds. Hereditas 41: 518-520. . 1962. Self-incompatibility in diploid Hordeum hulhosum L. Hereditas 48: 138-152. . 1965. Self-incompatibility in DactyUs aschcrsoniana Graebn. Hereditas 54: 70-87. . 1969. Self-incompatibility in DactyUs glomcrata L. Hereditas 61: 353-360. . 1975. Complex self-incompatibility systems in Angiosperms. Proc. Roy. Soc. London, Ser. B., Biol. Sci. 188: 235-245. McNamara. J. & J. A. QuiNN. 1977. Resource allocation and reproduction in populations of Am- phicarpum purshii (Gramineae). Amer. J. Bot. 64: 17-23. Mark, A. F. 1965. Flowering, seeding and seedling establishment of narrow-leaved snow-tussock, Chionochloa ligida. New Zealand J. Bot. 3: 180-193. Moss. G. I. & J. Hhslop-Harrison. 1968. Photoperiod and pollen sterility in maize. Ann. Bot. (London) 32: 833-846. Murray, B. G. 1974. Breeding systems and floral biology in the genus Briza^ Heredity 33: 285-292. Nettancourt, D. dp 1977. Incompatibility in Angiosperms. Monographs on Theoretical and Ap- plied Genetics 3. Springer-Verlag, Berlin. 230 pp. Nicora, E. G. 1962. Revalidacion del genero de Gramineas Necragrosds de la flora Norteamericana. Revisla Argent. Agron. 29: 1-11. Nygren, a. 1954. Apomixis in the Angiosperms. II. Bot. Rev. (Lancaster) 20: 577-649. , 1967, Apomixis in the Angiosperms. in W. Ruhland, (editor), Handbuch der Pflanzenphisio- logie. Vol. 18: 551-596. OroRODE, O. & S. R. Baquar. 1976. The Hypanhenia involucrata-H. suhplumosa (Gramineae) complex in Nigeria: Morphological and Cytological characterisation. Bot. J. Linn. Soc. 72: 211- 222. Pandey, K. K. 1977. Origin of complementary incompatibility systems in flowering plants. Theor. Appl. Genet. 49: 101-109. . 1979. Long-distance dispersal and self-incompatibility. New Zealand J. Bot. 17: 225-226. 1980, Evolution of incompatibility systems in plants: origin of ^'independent" and ^'com- plementary" control of incompatibility in Angiosperms. New^ Phytol. 84: 381-400. Philipson. M. N. 1978. Apomixis in Cortaderia juhata (Gramineae). New Zealand J. Bot. 16: 45- 59. Phipps, J. B. 1964. Studies in the Arundinelleae (Gramineae). L Classification of the taxa occurring in Bechuanaland, the Rhodesias and Nyasaland, and Mocambique. Kirkia 4: 87-124. Reeder, J. R. 1969. Las gramineas dioicas de Mexico. Bol. Soc. Bot. Mexico 30: 121-126. & C. G. Reeder. 1966. Notes on Mexican grasses IV. Dioecy in BouteUnui chondrosioides. Briltonia 18: 188-191. Renvoize, S, a. 1979. A new genus in the tribe Arundineae (Gramineae). Kew Bull. 33: 525-527. Roberty, G. I960. Monographic systematique des Andropogonees du globe. Boissiera 9: 1^55. Ross, M. D. 1970. Evolution of dioecy from gynodioecy. Evolution 24: 827-828. . 1978. The evolution of gynodioecy and subdioecy. Evolution 32: 174-188. — & R. F. Shaw. 1971. Maintenance of male-sterility in plant populations. Heredity 26: 1-8. & B. S. Weir. 1976. Maintenance of males and females in hermaphrodite populations and the evolution of dioecy. Evolution 30: 425^41. ScHAFFNER, J. H. 1920. The dioecious nature of buffalo-grass. Bull. Torrey Bot. Club 47: 119-124. Schuster, J. 1910. Ober die Morphologic der Grasbliite. Elora 100: 213-266, Sharman, B, C. 1960. Development of the inflorescence and spikelets of Anthoxanihum odoratum L. New Phytol. 59: 60-64. Soderstrom, T. R. & C. E. Calderon. 1971. Insect pollination in tropical rain forest grasses. Biotropica 3: 1-16. 74 ANNALS OF THK MISSOURI BOTANICAL GARDLN [Voi . 68 & . 1974. Primitive forest grasses and evolution of the Bamhusoideae. Biotropica 6: 14I-L^3. & . 1979a. ArhereUa (Poaceae:Bambusoideae): a new genus from tropical America. Brittonia3L 433^45. & . 1979b. A commentary on the bamboos (Poaceae:Bambusoideae). Biotropica 11: 161-172. Spoor, W. 1976. Self-incompatibility in Lolium perenne L. Heredity 37: 417^21. Stapf , (). 1927. Lcconitella mcula^uscciriensis A. Camus, Hooker's Icon. PL tab. 3123, Stebbins, G. L. 1941. Apomixis in the Angiospcrms. Bot. Rev. (Lancaster) 7: 507-542. . 1956. Cytogenetics and evolution of the grass family, Amer. J. Bot. 43: 890-905. . 1957. Self-fertilization and population variability in the higher plants. Amer. Naturalist 91: 337-354, . 1974. Flowering Plants: Evolution Above the Species Level. The Belknap Press of Harvard Univ. Press, Cambridge, Massachusetts. 399 pp. Stern, C. 1936. Somatic crossing over and segregation in Drosophila melufw^astcr. Genetics 21: 625-730. Sweeney. T. L. & A. R. B.\rr. 1978. Sex ratio distortion caused by meiotic drive in a mosquito, Cidc.x pipiens L. Genetics 88: 427-446. ViLLAMii , C. B. 1%9. El genero Monanthochloe (Gramineae). Estudios morfologicos y taxonomicos con especial referenda a la especie argentina. Kuriziana 5: 369-391. VoiGT, P. W., W. R. Kneebone, J. R. Harlan & R. M. Ahring. 1975. Registration of Texoka buffalograss. Crop Sci. 15: 885. Weatherwax, p. 1926. Comparative morphology of the Oriental Maydeae. Indiana Univ. Stud. 13(73): 1-18. . 1939. The morphology and phylogenetic position of the genus Jouvea (Gramineae). Bull. Torrey Bot. Club 66: 315-325. Webb, C. J. 1979. Breeding systems and the evolution of dioecy in New Zealand apioid Umbelliferae. Evolution 33: 662-672. Webster, O. J. 1%5. Genetic studies in Sorghum vulgarc Pers. Crop Sci. 5: 207-210. Weir, C. E. & H. M. Dale. 1960. A developmental study of wild rice, Zizania aqtuitica L. Canad. J. Bot. 38: 719-739. Whiiehead, D. R. 1%9. Wind pollination in the Angiosperms: evolutionary and environmental considerations. Evolution 23: 28-35. WiLLSON, M. F. 1979. Sexual selection in plants. Amer. Naturalist 113: 777-790. COEVOLUTION OF GRASSES AND HERBIVORES G. Ledyard Stebbins* Abstract The appearance of fossilized silica bodies derived from the leaf epidermis of grasses and of mammalian fossils having high-crowned teeth was nearly simultaneous in lower to middle Eocene strata of Patagonia, where these fossils are associated with dry land sediments that indicate the presence of savannas containing shrubs and traversed by rivers that provided mesic habitats. In North America, the earliest clearly identified grass fossils are stipoid fruits of lower Miocene age, while the oldest mammals having high-crowned teeth are rhinoceroses of Miocene age. The abundant stipoid fruits known from the Miocene and Pliocene Epochs in the central United States indicate that the earliest Miocene species were quite different from modern counterparts, but that early Pliocene species have modern counterparts in the pampas of South America. During the Pleistocene, stipoid grasses ceased to be dominant elements of North American grasslands, being replaced by grasses belonging to the tribes Chlorideae and Andropogoneae. This change was associated with the appear- ance of a drier, more continental climate and with the appearance of bison and sheep on the North American plains. The evolutionary significance of these coordinated changes is discussed. A fundamental tenet of modern evolutionary theory is that, at least with re- spect to visible morphological and anatomical characteristics, rates of evolution reflect to a large degree rates of change through time in the nature of population- environment interactions (Dobzhansky et al., 1977). Rapid evolution is often as- sociated with successful responses to rapid and complex changes in the environ- ment. The environmental changes that stimulate most strongly rapid evolution involve both the physical and biotic environment. Often the changes in chmate and similar physical factors trigger off changes in structure and behavior on the part of those organisms that have the most direct relationships with the external environment, and these in turn affect other organisms of a given community. An excellent example of rapid evolution as a result of this complex web of physical and organismic changes is the evolution of grasses and the herbivorous mammals that depend upon them. Paleobotanists and vertebrate paleontologists from both North America and South America have now accumulated data that can be synthesized into an account of this coevolution which, though admittedly partial, is nevertheless highly convincing. The very different history of events occurring on these two continents, plus the remarkable synchrony on each con- tinent between the evolution of grasses and mammalian herbivores shows clearly the close interdependence between the evolution of these two very different kinds of organisms. The Kinds of Evidence That are Avah able The most important kind of evidence from which the story can be pieced together is provided by the fossil record. Fortunately, this record is remarkably good in both North and South America, including remains of both grasses and mammals. The grass fossils that are the most diagnostic are fruits that clearly belong to the tribe Stipeae, The pioneering monograph of these fruits prepared by Elias (1942) has now been supplemented by more modern and thorough re- 1 Department of Genetics, University of California. Davis, California 95616 Ann. Missouri Bot. Gard. 68: 75-86. 1981. 0026-6493/8 1/0075-0086/$ 1 .35/0 76 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 search on the part of Thomasson (1978a, 1978b, 1979). The remarkable features of these fossils are both their abundance and the details of cellular structure that can be seen in them. Moreover, in contrast to most other fossils of angiosperms, they reveal characters of form that are usually used by systematists dealing with their modern counterparts as diagnostic of species and genera. Next in importance are fossilized bits of leaf epidermis. To my knowledge, the only definite fossils of this kind are those discovered by Litke (1968) in coal deposits of eastern Germany. In them, cell outlines are perfectly preserved, so that identification to tribe is unquestionable. They show that at least one tribe of the family, the Oryzeae, was already differentiated by the end of the Eocene Epoch. Of a similar character, but harder to identify, are the much more abundant bits of organic silica known as opalines, that occur in some sediments. These microscopic particles are derived from epidermal cells of grasses that have be- come fully impregnated with silica. When well preserved, they are diagnostic of subfamilies, as shown by Prat (1932) and others. Most reports of these fossils are from sediments of Pliocene age, but much older examples, some of them dating from the middle or early Eocene, have been reported from Patagonia, South America (Frenguelli, 1930; Teruggi, 1955; Bertoldi de Pomar, 1971, 1972). The pollen of Poaceae is highly distinctive as to the family, but of little value for the recognition of its subdivisions. It does not become abundant until the Miocene Epoch, when evidence from fruits and epidermal fragments indicates that the family was already fully differentiated. For this reason, palynological fossils are among the least valuable for reconstructing the evolutionary history of the family. Leaves that on the basis of superficial appearance have been as- signed to the family and given such generic names as Poacites are even less helpful. Recent studies of some of them on the basis of microscopic cellular structure has shown that they belong to other families of monocotyledons (Litke, 1 968) . The supreme value of fossils of mammalian herbivores lies in the fact that the parts that are most commonly preserved — teeth and hoofs — are highly diagnostic of both their morphological relationships and their mode of life. High-crowned (hypsodont) teeth having complex patterns of enamel evolved in response to pressures exerted by a diet of hard, siliceous leaves, particularly those of grasses. Long slender legs that terminate in hoofs made up of one or two much enlarged toes evolved in response to rapid running or leaping in country that was at least partly open and free of dense forests. Consequently, hypsodont teeth and well- developed hoofs can be expected to be associated with fossil remains of grasses that are adapted to open country. The frequent presence of this coexistence, or coadaptive syndrome, will become evident from the facts to be presented below. Direct fossil evidence can be supplemented by indirect evidence of several kinds. Particularly useful is that derived from detailed studies of sediments (Pas- cual & Odreman, 1971; Andreis, 1972; Spaletti & Mazzoni, 1978). The presence of saline deposits and gypsum in beds that are clearly terrestrial rather than marine in origin indicates at least seasonal drought. Other characteristics, such as deposits resulting from showers of volcanic ash, and the nature of indications of animal activity, such as burrows of worms and nests of scarab beetles, can also be helpful. 1981] STEBHlNS—(iR ASS HERBIVORE COEVOLU FION 77 Another kind of indirect evidence can be obtained from comparative studies of successions of fossils through time. Because of their high diagnostic value, mammalian fossils are particularly helpful for deciding when a succession rep- resents progressive evolution in situ, and when a change in faunas is due to migration from another region. Although successions of plant fossils are rarely complete enough to lead toward decisions of this kind, an exception is provided by the fruits of Poaceae, tribe Stipeae, that have been uncovered on the plains of the central United States. The combined study of these grass fossils and the mammalian remains associated with them provide the best opportunity for ana- lyzing coevolution of terrestrial plants and animals over a long period of time that exists anywhere in the world. Indirect evidence can also be obtained from geology, particularly the discipline of plate tectonics, that provides indications of past continental movements. This subject has been carefully reviewed by Raven & Axelrod (1974). Nevertheless, great caution must be used in applying this evidence to problems of plant and animal dispersal. Even with respect to organisms that are least able to be trans- ported over long stretches of ocean, such as mammals, one cannot assume au- tomatically that a land bridge is necessary to transport them from one continent to another. As reviewed by Patterson & Pascual (1968, 1972), a combination of comparative fossil evidence and that related to past continental movements in- dicates that during the end of the Cretaceous and the beginning of the Tertiary Period, several groups of mammals must have reached South America by migra- tion over water. Since most seeds of angiosperms can be transported over long distances as easily or more easily than can mammals, the entrance of angiosperm groups into South America during the same period by occasional long distance migration can by no means be ruled out. Finally, modern patterns of geographical distribution can supplement those provided by fossils. These must also be interpreted with caution, and associated with fossil distribution patterns whenever possible. With respect to grasses, the inference is reasonable that the migration of Stipeae, for which a fossil record is available, was accompanied by migrations of grasses belonging to other tribes, which left no record because both their vegetative parts and their seeds are too perishable. Since phytogenies involving polyploidy usually proceed from lower to higher levels (Stebbins, 1971, 1980) inferences about past patterns of distri- bution can often be obtained by comparing the distribution patterns of diploids and related polyploids (Stebbins, 1947, 1950, 1971). Relevani Features of Climatic Chancre and continenl al movements The record of coevolution between grasses and herbivores spans most of the Tertiary and Quaternary Periods. It begins in southern South America (Patagonia) in the middle of the Eocene Epoch, about 45 million years ago, and continues until the retreat of the glaciers in the post-Pleistocene and the consequent post- glacial xerothermic stage permitted the establishment of open xeric grasslands in the west central part of the United States, plus adjacent Canada and Mexico, less than 10,000 years ago. Climatic and orogenic changes throughout this long period are relevant to the problem. 78 ANNALS OF THE MISSOURI BOTANICAL GARDHN [Voi . 68 The story of these changes has been reviewed so many times that it does not need to be repeated here. North and South America underwent a gradual cooHng of climate, accompanied by the rise of the western cordillera, that increased habitat diversity, and brought to the central regions of the continents the conti- nental climate that now exists in them, characterized by cold winters and hot summers. Of particular interest, however, are the differences between the two continents in the intensity of these effects. The differences will therefore be reviewed. In the first place, semiarid conditions, including the appearance of open sa- vannas dotted with widely spaced trees and shrubs, appeared in South America about 20 million years before their appearance in North America. On both con- tinents, the Paleocene Epoch was characterized by continuous forests perhaps interspersed by small enclaves of open country. In the middle of the Eocene, however, evidence from the Mustersan strata in the Province of Chubut, Argen- tina at 45° south latitude, indicates the presence of widespread savanna conditions, as is described in more detail below. Although direct evidence for the age of these strata is lacking, Marshall et al. (1977) have estimated their age as middle Eocene, about 48 million years old, on the basis of stratigraphic evidence plus a potassium- argon date of 35 million years for a sample from the younger Deseadan formation. At this time, evidence from the fossil record of both plants and animals indicates that central North America was still covered by more or less continuous forests. Second, the greater land mass of North America caused that continent to respond to the uplift of mountains by acquiring a much greater degree of conti- nentality than that which prevails in South America. This difference, which is evident to anyone who has travelled over the two continents, is clearly shown by comparisons between the different climate diagrams presented by Walter & Lieth (1967). Its relation to the timing of environmental changes is highly important. Equable climates persisted in central and eastern South America long after they had disappeared from corresponding parts of North America. This means that plant species or genera, arriving in South America during the Tertiary Period, had a greater change for persisting in a relatively unchanged condition than did their relatives that remained in North America. In many respects, the present, or rather pre-human conditions prevailing on the pampas of Argentina, Uruguay and southern Brazil during the Miocene and early Pliocene epochs probably re- semble more the conditions that prevailed on the Great Plains of North America during the Miocene and early Pliocene epochs than those that prevail there today. Another important difference is with respect to the biogeographic connections that existed on the two continents. South America was an island continent from the end of the Cretaceous Period until the end of the Tertiary. It received im- migrations from other continents only via occasional long distance dispersal, except for a possible closer link with Antarctica. North America, on the other hand, was intermittently connected with Eurasia by land bridges that permitted extensive transmigrations of flora and fauna. Consequently, the early Tertiary mammalian fauna of South America was decidedly unbalanced, being derived from a few accidental introductions (Patterson & Pascual, 1972). Balance was achieved by extensive adaptive radiation from this small number of immigrant stocks. On the other hand, the fauna of North America was constantly receiving immigrants from Eurasia, so that the fauna of the Northern hemisphere was at 1981] STEBBINS— GRASS-HERBIVORE COEVOLUTION 79 all times balanced, and contained several different potential competitors for each new ecological niche that was opened up as a result of environmental change. There is every reason for believing that the same factors produced comparable differences in the evolution of grasses and other angiosperms. Finally, the two continents were affected very differently by the Pleistocene glaciation. In North America, the Great Plains, that during the Miocene and Pliocene epochs were the scene of coevolution of grasses and mammals, during the Pleistocene were ice covered over the northern part, and farther south became tundra, or coniferous forest (Davis, 1976). On the other hand, glaciation in South America was confined to the high Andes and the southern portion of Patagonia. The extensive grasslands that form the pampas of Argentina, Uruguay and south- ern Brazil were little affected by this change. Three factors, therefore, would have contributed to the longer persistence and slower evolution of plains grasses in South America as compared to North Amer- ica: (I) the longer persistence of equable climates in its temperate regions: (2) the lack of immigration and therefore of new levels of competition during most of the Tertiary Period: and (3) the comparatively mild influence of the Pleistocene gla- ciation. These factors were, however, balanced by another factor that greatly stimulated both evolution and extinction. This was the massive immigration of northern mammals into South America at the end of the Pliocene Epoch, as a result of the rise of the Panamanian land bridge. As is explained below, this immigration was most probably accompanied by a similar immigration of northern grasses and other angiospermous herbs. The Succession of Events in the Tertiary History of Grasses and Herbivores Prior to the Eocene Epoch, no evidence exists for the occurrence either of grasses (Poaceae) or of mammals having teeth adapted to feeding upon them. Primitive ungulates (Condylarths) were widespread during the Paleocene and ap- parently originated in the Upper Cretaceous (Patterson & Pascual, 1972), but these forms all had low-crowned teeth that lacked complex patterns of enamel. They were almost certainly browsers. The earliest mammals having hypsodont teeth were the family Archaeohyra- cidae, of the order Notungulata, that appeared first in the early Eocene strata of Patagonia. They were followed in the middle Eocene by the Notohippidae, be- longing to the same order. In the latter family, the evolution of hooflike feet was superficially similar to the much later evolution of hoofs in North American horses. Careful analyses of the land-borne sediments in which these remains were found makes highly probable the conclusion that the animals lived in open sa- vannas, dotted with trees and shrubs, having an aspect not unlike that of the pampas of temperate South America when first visited by Europeans (Andreis, 1972; Spaletti & Mazzoni, 1978). In these strata opalines that were derived from epidermal cells of grass leaves, though scarce, were undoubtedly present (Fren- guelli, 1930; Teruggi, 1955). Unfortunately, their state of preservation is such that they cannot be identified as to subfamily or tribe. Contemporary Eocene strata of North America contain fossils of herbivorous mammals, including Condylarths 80 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 and primitive ungulates, but these all have characteristics associated with brows- ing rather than grazing. The origin of both these Eocene grasses and grazers of temperate South America is in doubt. The Notungulates may have evolved from the more primitive Condylarths that are found abundantly in Paleocene strata, in response to selec- tive pressures produced by the changing habitats. Could the grasses have also evolved from more primitive, forest-loving groups? This hypothesis is certainly plausible, in view of the present distribution of primitive grasses in the forests of tropical South America. This region is a center of diversity not only for bambusoid grasses having relatively low chromosome numbers, (Soderstrom & Calderon, 1974), but also for genera such as Streptochaetu and Anomochloa, that are either archaic or appear to have diverged in their own unusual direction from primitive ancestors. One cannot, however, conclude from these facts that the Gramineae as a whole originated in South America. Bambuseae having similar low chromosome numbers exist also in West Africa, and the subfamilies of nonbambusoid grasses that had the largest number of unspecialized characteristics, particularly the Arundinoideae, tribe Danthonieae, are far better developed in Africa than in South America. Furthermore, the isolation of South America during the Eocene would make highly improbable the emigration of Poaceae from that continent to the other continents where they certainly existed during the middle of the Tertiary Period. The existence of the subfamily Oryzoideae in Europe during the Upper Eocene (Litke, 1968) is highly significant in this connnection. Consequently, I believe that Poaceae most probably entered South America from the Old World, possibly Africa, during the Paleocene Epoch or the uppermost Cretaceous. Al- though Africa and South America were already well separated during this time, islands may well have existed between them in the South Atlantic. These could have served as stepping stones for long distance migration. This hypothesis has been advanced by Raven & Axelrod (1975) for many elements of the South American biota, both flora and fauna. During the Oligocene, the coevolution of grasses and herbivorous grazers in South America continued along the lines that were begun during the Eocene. Several groups of Notungulates became even more diversified and specialized for life in open savannas. The accompanying evolution of grasses has left no record, except that the scattered presence of opaline siliceous bodies suggests that they also were actively evolving. In North America, the earliest evidence that Poaceae were present comes from the Florissant deposits of uppermost Eocene or earliest Oligocene (MacGinitie, 1953; Beetle, 1958; see Epis & Chapin 1975 for accurate dating). These are fruits of two different kinds, one of them belonging to the tribe Stipeae, and perhaps the genus Stipa itself, and the other possibly to the genus Phalaris. Both of these genera belong to the subfamily Pooideae, but to different tribes, Stipeae and Phalarideae, respectively. The existence of such genera, that are end branches of different lines of evolution within the subfamily, suggests that the Pooideae were already well differentiated at that time. This conclusion agrees well with the completely modern character of the pooid fragments of epidermis found by Litke (1968) in Miocene coal-bearing strata of Germany. Both Stipa and Phalaris most probably reached North America from Eurasia. 1981] STEBBINS— GRASS-HERBIVORE COFVOLLTION 81 The probable origin of subfamily Pooideae (^Testucoideae") in temperate Eurasia has been emphasized by Hartley (1973) on the basis of modern distribution pat- terns. The two largest genera of the tribe Stipeae, Stipa and Oryzopsis, both have pronounced centers of diversity in central Asia (Komarov, 1934). The species of {Phalaris) are also primarily Eurasian. Consequently, the earliest known grasses that inhabited open savannas in North America appear to have acquired their distinctive characteristics by a course of evolution that took place during the Eocene Epoch in Eurasia. Even though temperate South America is at present a center of diversity for the tribe Stipeae, it probably was not the center of origin for the tribe. The paleosoils of Eocene and Oligocene age in temperate South America, including those of the temperate pampas where Stipeae are now abundant, have been ex- amined as carefully as have the Miocene deposits of North America. The absence of stipoid fossils from South American deposits is, therefore, as significant as their abundance in North America. Stipoid grasses most probably migrated from North to South America contemporaneously with the ungulates of North Amer- ican origin, such as camels and horses. Incredibly rich deposits of stipoid fruits of ^'seeds'' have been found in Mio- cene and Pliocene deposits of central North America, extending from Nebraska and Kansas west to Colorado, and southward to New Mexico. They provide what is probably the best fossil sequence of herbaceous angiosperms anywhere in the world. Even though the stratigraphic succession, as worked out by Elias and others, has recently been questioned (Thomasson, 1979), a phylogenetic succes- sion of fruit forms can still be recognized. The known sequence begins in the early to middle Miocene with Berriochloa primaeva Elias, which apparently has no modern counterpart, and ends in the Lower Pliocene with a series of well- differentiated forms that can be assigned to the modern genera Stipa. Oryzopsis, Piptochaetium and NassellcL Stipoid grasses evolved contemporaneously with two groups of grazing un- gulates, horses and pronghorn ^^antelopes.'' The Eocene and Oligocene repre- sentatives of the horse line that evolved in central North America were small to medium-sized in stature, had low-crowned teeth and feet consisting of three or more digits (Romer, 1966; Simpson, 1951). During the Miocene, successive genera of horses increased considerably in size. At the same time, they radiated adap- tively into two different lines. One of them led to Anchitherium and Hypohippus, which became relatively large, but retained teeth equipped for browsing (Simp- son, 1951). In the other, represented in the Miocene by Merychippus, adaptation to grazing evolved through the acquisition of teeth having more complex patterns of enamel, accompanied by cement. These horses also ran more firmly on their middle toe and acquired a springing gait more like that of modern horses. In the earlier horses and the Anchitherium-Hypohippus line, the hind part of the foot consisted of a flexible pad or cushion, much as in the feet of modern dogs or camels. The presence of this pad prevented the animal from flexing its toes or hoofs, an action of horses that is essential for fast running of a relatively large animal. Consequently, an adaptive change correlated with the altered teeth was the loss of this pad, and the evolution of a springing motion during running, based upon the flexion of the middle toe. Merychippus and its descendants became 82 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 adapted to more open savannas, that were increasing in extent during the Miocene epoch. The shift in diet was probably a result of selective pressure brought about both by the increased abundance of grasses relative to the foliage of shrubs and trees, and intense competition for the latter from other browsers including tapirs, rhinoceroses, chalicotheres and camels, as well as Anchitherium-type horses. The peifection of teeth adapted for grazing most probably coincided with the evolution of firmer, more siliceous leaves on the part of the grasses that grew in the more open areas. At the end of the Miocene, a second adaptive radiation took place. Three closely related lines of this radiation retained the three-toed hoofs of Merychip- piis, but acquired teeth having higher crowns. The three terminal forms, Nanip- pus, Neohipparion and Hipparion, differed from each other in body size and details of tooth structure. Of these, Nanippus and Hipparion persisted through- out the Pliocene. In the fourth line, represented during the Pliocene by Pliohippus and later by modern horses (Equus), the changes in teeth were somewhat similar, but those of legs and toes were more profound. The legs became longer and more slender, and the side toes were lost, producing the single large hoof characteristic of modern horses. A highly significant fact about Pliocene evolution of horses, worked out in detail for the Great Basin by Shotwell (1961), was the contemporaneous and largely sympatric existence of two quite different kinds of horses, Hipparion and Pliohippus. Shotwell infers that they occupied somewhat different habitats; Hip- parion living in savannas or parklands containing many trees and shrubs, while Pliohippus occupied the more open areas. He suggests that the side toes provided added traction in dodging movements while the animal was escaping from pred- ators in savannas or parkland, while the presence of a single large toe made running in a straight line more efficient for a horse that lived chiefly or entirely in open country. Comparing the distribution throughout the Pliocene of Hipparion and Pliohippus, Shotwell concludes that the spread of Pliohippus at the expense of Hipparion coincided closely with spread of open grassland, replacing savanna or parkland. Apparently, the direct response to the increasingly dry climate was with respect to the flora, but that the horses, with little or no further evolution, altered their distribution in response to the change in vegetation. Thomasson, reviewing the change in vegetation and in horse distribution in the Great Plains of the central states, concludes that the same interacting succession of events took place there. Unfortunately, grass fruits have not been found in deposits younger than early Pliocene, so that the changes in flora during the last five million years cannot be followed. The changes that took place must be inferred from comparisons be- tween the fossil fruits dating from the earliest Pliocene and the composition of the modern floras. These are as follows. In the South American pampas the stipoid genera Stipa, Nassella and Piptochactium, not known as fossils, are now abundant and represented by a large number of species. In North America, on the other hand, stipoid grasses are much less abundant than they were during the early Pliocene. The genus Stipa is represented chiefly by S. comata, S. spartea, S. viridula and S. rohusta, all of which have fruits that differ considerably from the Pliocene fossils. A fact to be noted is that the first three of these species are 1981] STEBBINS— GRASS-HERBIVORE COEVOLUTION 83 polyploids. The chromosome number of both S. comata and S, spartea is In = 44 to 46, depending upon the presence or absence of a small extra pair, perhaps of B-type chromosomes (Stebbins & Love, 1941). That of S. viridula is In = 82 (Johnson & Rogler, 1943). Since the modern fruits of these species are consid- erably larger than any of the Pliocene fossils, one can suspect that their polyploidy arose during the later Pliocene, Pleistocene, or even more recently. The other two genera that are known from fossils of Pliocene age, Nassella and Piptochaetium, are now absent from the central and western grasslands. Piptochaetium survives in the arid mountains of the southwest, as well as in the form of two species that are usually placed in the genus Stipa, S. avenaceay which occurs in rocky woods throughout the eastern United States, and S. av- enacoides, native to open, savannalike woodlands in Florida. North of temperate South America, Nassella is represented only by a few localized species in Mexico and Central America. The stipoid grasses have been largely replaced in the Great Plains by genera belonging to three completely different tribes; the bluestems (Andropogon, s.l.), belonging to the tribe Andropogoneae; switchgrass {Panicum virgatum), tribe Paniceae; grama grass {Bouteloua curtipendula and B. gracilis) and buffalo grass {Biichloe dactyloides), the latter two both of the tribe Chlorideae. When and why did this revolutionary change take place? No definite answer can be given to either of these questions. The answer to the question ^'when?'' is made difficult by the absence of grass fossils dating from the Upper Pliocene and Pleistocene epochs. If, as seems likely, stipoid grasses were associated chiefly with savannas and grasses belonging to other tribes were predominant on open plains, then the gradual restriction of savanna habitats, as can be inferred from the distribution of Hipparion and other fossil mammals, must have been associated with a similar decrease in abundance of stipoids, or at least of those species that occur in the early Pliocene deposits. This would mean that by the end of the Pliocene epoch, the genera Berriochloa, Piptochae- tium and Nassella had already disappeared from most of the central plains. Nevertheless, they probably persisted in some parts of the area as late as early Pleistocene. This inference is based upon my belief that their invasion of the South American pampas was simultaneous with the transcontinental migration of northern ungulates, particularly horses (Equidae) and Camelidae. This migration is usually dated as immediately following the appearance of the Panamanian land bridge, at the end of the Pliocene or the beginning of the Pleistocene epoch. At the beginning of the Pleistocene glaciations, therefore, the composition of the grass flora of the central plains was probably already different from that recorded in lower Pliocene deposits. Such a change would be expected as a result of increasing aridity and progressively colder winters, which is evident from the composition of woody floras from throughout the western United States (Axelrod, 1975). Species of the genera Stipa and Oryzopsis were probably there, as they are today, as well as Agropyron and Elymus, which probably entered North America from Eurasia during the Pliocene (Stebbins, 1974). Species of the genus Panicum, perhaps descendants of the Mio-Pliocene P. elegans and relatives of the modern P. virgatum, were probably present. Nevertheless, good reasons exist for believing that three of the dominant elements of the modern grass flora, 84 ANNAI.S OF THE MISSOURI BOTANICAL GARDHN [Voi . 68 the bluestems {Andropogon or Schizachyriiim), indian grass {Sorghastnim), gra- ma grasses {Bouteloua) and buffalo grass {Buchloe), were either absent or un- common. As discussed elsewhere (Stebbins, 1974) the species of these groups that now dominate the plains are polyploids that appear to have close diploid relatives either in the southern United States or in Mexico. The polyploid com- plexes to which they belong have distribution patterns that suggest youthfulness (Stebbins, 1971). I believe, therefore, that these species are of post-Pleistocene origin. Throughout the Pliocene epoch, the mammalian herbivores that dominated the Great Plains belonged to the same evolutionary lines that had been evolving there during the Miocene: horses, pronghorn "antelopes" and camels. They were becoming larger, and gave rise to a number of branches off their main evolution- ary lines. Nevertheless, their effect upon the plains grasses must have been sim- ilar to that exerted by their Miocene ancestors, though perhaps more intense. The Pleistocene glaciations, plus the profound effects that they exerted on both climate and soil for hundreds of miles south of the ice margin itself, must have altered greatly the central and western grasslands. Immediately south of the ice margin, grasslands were replaced by tundra and spruce forest, even in areas that are now short grass plains, such as north and central Nebraska (Davis, 1976). During drier periods, the accumulation of loess must have altered greatly soil conditions, contributing further to the displacement of floras. During the Pleistocene epoch, the fauna of herbivores was altered by the arrival of bison and sheep from Eurasia. These animals graze more closely and densely than do horses and pronghorns, which dominated the plains before their arrival. They would be particularly destructive to long-leaved caespitose or bunch grasses, such as most species of Stipa and perennial Triticeae. On the other hand, rhizomatous, sod-forming grasses, such as Bouteloua and Buchloe, would be much more resistant to their grazing pressure. For these reasons, I believe that the grassland vegetation of the central and western United States, as recorded by the botanists who first visited the area during the nineteenth century, consists of recent plant communities that were put together after the retreat of the last ice sheet. Some of their contemporary species may have arrived or at least spread during comparatively modern times, as a result of burning and other activities on the part of the Indians (Gleason & Cron- quist, 1964: 206-208). Coevolution of grasses and herbivores has progressed con- tinuously on the North American grasslands ever since they first appeared as savannalike openings in the forest during the early part of the Tertiary period. Conclusion In spite of large gaps in the fossil record, available information permits a partial synthesis that reveals close connections between the evolution of large mammalian grazing herbivores and the grasses upon which they fed. This coevo- lution began during the Eocene Epoch, 45 to 55 million years ago, and has con- tinued up to the present. Further research studies, conducted in a variety of disciplines, will surely fill in and amplify the picture. Paleosoils need to be ex- amined more carefully, with particular reference to the possible occurrence in 1981] STEBBINS— GRASS-HERBIVORE COEVOLUTION 85 Whenever enough preserved, they could be examined by agrostologists who might be able to determine at least approximately their systematic affinities. More information is needed about the stipoid fruits found in the Great Plains area, particularly their total distribution in both space and time. At the same time, careful morphological comparisons between the fruits of species belonging to modern stipoid genera will help greatly in the classification of the fossil forms, and for estimating the amounts of difference between them. A synopsis of morphological and histolog- ical characteristics of stipoid fruits on a worldwide basis will be necessary for interpreting such comparisons. Cytogenetic and biochemical comparisons of dif- ferent modern species and genera will be a necessary prelude to the interpretation of evolutionary relationships. In addition, the effects of grazing animals upon the species composition need to be better understood. Research workers in the discipline of range management have conducted many experiments to show that different intensities of grazing affect greatly the species composition of range lands. In addition, trampling by hoofs of the grazers exerts additional effects, particularly during wet seasons. This kind of damage is quite different depending upon the kinds of hoofs that the different species of grazers possess. Other factors are the different sizes of the flocks and herds that are characteristic of different species, and the tendency of some of them to travel long distances, while others remain for relatively long periods of time in the same area. I am not aware of any careful investigations even of the comparative effect on range lands of different kinds of modern do- mestic animals, such as horses and cattle, although ranchers and range specialists are well aware of the existence of differences, and anecdotal accounts of them are numerous. In this entire field, the opportunities for collaborative research are very great. They can be directed toward a highly valuable aim: interpreting the evolution of two of the world's major food resources— grasslands and grazing ani- mals. Mankind has depended upon them for his existence ever since our remote ancestors ventured onto the savannas and began a new mode of existence. Literature Cited Andreis, R. R. 1972. Paleosuelos de la formacion Musters (Eoceno medio), Laguna del Mate, Prov. de Chubut., Rep. Argentina. Revista Asoc. Argent. Mineral., Petrol., Sedimentol. 3: 91-97. AxELROD, D. I. 1975. Evolution and biogeography of Madrean-Tethyan sclerophyll vegetation. Ann. Missouri Bot. Card. 62: 280-334. Beetle, A. A. 1958. Piptochaetium and Phalaris in the fossil record. Bull. Torrey Bot. Club 85: 179-181. Bertoi-DI de Pomar, H. 1971. Ensayo de clasificacion morfologica de los silicofitolitos. Ameghi- niana 8: 317-328. . 1972. Opalo organogenico en sedimentos superficiales de la llanura santafesina. Ameghiniana 9: 265-279. Brues, C.T. & B. B. Brues. 1909. A new fossil grass from the Miocene of Colorado. Bull. Wisconsin Nat. Hist. Soc. 6: 170-171. Davis, M. B. 1976. Pleistocene biogeography of temperate deciduous forests. Geoscience and Man 13: 13-26. Dobzhansky, T., F. J. Ayala, G. L. Stebbins & J. W. Valentine. 1977. Evolution. W. H. Freeman, San Francisco. Elias, M. K. 1942. Tertiary prairie grasses and other herbs from the high plains. Special Pap. Geol. Soc. Amer. 41: 1-176. Epis, R. C. & C. E. Chapin. 1975. Geomorphic and tectonic implications of the post-Laramide, late 86 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Voi . 68 Eocene erosion surface in the soulhern Rocky Mountains. Pp. 45-74, in B. F. Curtis (editor), Cenozoic History of the Southern Rocky Mountains. Geological Soc. Amer., Boulder. Colorado! Frhnguelli, J. 1930. Particulas de silice organizada en el loess y en los limos pampeanos. Selulas siliceas de Grammeas. Anales Soc. Ci. Santa Fe 2: 1-47. Gleason, H. a. & A. Cronquist. 1964. The Natural Geography of Plants. Columbia Univ. Press, New York. Harti EY. W. 1973. Studies on the origin, evohition and distribution of the Gramineae V. The subfamily Festucoideae. Austral. J. Hot. 21: 201-234. Johnson, B. L. & G. A. Rogler. 1943. A cytotaxonomic study of an intergeneric hybrid between Oryzopsis hymenoiJes and Stipa viriJula. Amer. J. Bot. 30: 49-56. KoMARov, V. L. (editor). 1934. Flora USSR. Vol. 2. Akademiia Nauk., Leningrad. LiTKE, R. 1%8. Uber den Nachweis tertiarer Gramineen. Monatsber. Deutsch. Akad. Wiss. Berlin 10: 462-471. MacGinitie, H. D. 1953. Fossil plants of the Florissant beds, Colorado. Publ. Carnegie Inst. Wash. 599: 1-198. Marshall, L. G., R. Pascual, G. H. Curtis & R. E, Drake. 1977. South American geochron- ology: radiometric time scale for middle to late Tertiary mammal-bearing horizons in Patagonia. Science 195: 1325-1328. Pascual, R. & O. E. Odreman. 1971. Evolucion de los communidades de los vertebrados del Terciario aregenlino. Los aspectos paleozoogeograficos y paleoclimaticos relacionados. Ameghi- niana 8: 372-412. Patti RsoN, B. & R. Pascual. 1%8. The fossil mammal fauna of South America. Quart. Rev. Biol 43: 409-151. & ■ 197-. The fossil mammal fauna of South America. Pp. 247-309, in A. Keast., F. Erk & B. Glass (editors). Evolution, Mammals and Southern Continents. State Univ. New York Press, Albany. Prat, H. 1932. L'epiderme des Graminees. Etude anatomique et systematique. Ann. Sci. Nat Bot ser. 10, 14: 117-324. Ravln, P. H. & D. I. Axhlroo. 1974. Angiosperm biogeography and past continental movements. Ann. Missouri Bot. Gard. 61: 539-673. & . 1975. History of the flora and fauna of Latin America. Amer. Sci. 63: 420-429. Romkr, a. S. 1966. Vertebrate Paleontology. Ed. 3. Univ. Chicago Press, Chicago. Shotwlll, J. A. 1961, Late Tertiary biogeography of horses in the northern Great Basin. J. Pa- leontol. 35: 203-217. Simpson, G. G. 1951. Horses. Oxford Univ. Press, New York. Sooerstrom, T. R. & C. E. Caldlron. 1974. Primitive forest grasses and the evolution of the Bambusoideae. Biotropica 6: 141-153. Spaletti, L. A. & M. M. Mazzoni. 1978. Sedimentologia del grupo Sarmiento en un pei-fil ubicado al sudeste del Lago Colhue Huapi Provincia de Chubut. Obra Cent. Mus. La Plata (Areentina) 4: 261-283. Stebbins, G. L. 1947. The origin of the complex Bromus cahnatus and its phytogeographic impli- cations. Contr. Gray Herb. 165: 42-55. . 1950. Variation and Evolution in Plants. Columbia Univ. Press, New York. . 1971, Chromosomal Evolution in Higher Plants. E. Arnold, London. . 1974. The role of polyploidy in the evolution of North American grasslands. Taxon 24: 91- 106. . 1980. Polyploidy in plants: Unsolved problems and prospects. Pp. 495-520, in W. H. Lewis (editor). Polyploidy: Biological Relevance. Plenum Press, New York. & R. M. Love. 1941. A cytological study of California forage grasses. Amer. J. Bot. 28: 371- 383. Teruggi, M. E. 1955. Algunas observaciones microscopicas sobre vidrio volcanico y opalo organ- ogeno en sedimentos pampianos. Notas Mus. La Plata, Geol. 18(66): 17-26. Thomasson, J, R. 1978a. Epidermal patterns of the lemma in some fossil and living grasses and their phylogenetic significance. Science 199: 975-977. . 1978b. Observations on the characteristics of the lemma and palea of the late Cenozoic grass Pdnicum elegans. Amer. J. Bot. 65: 34-39. . 1979. Late Cenozoic grasses and other angiosperms from Kansas, Nebraska and Colorado. Biostratigraphy and relationships to living taxa. Kansas Geol. Surv. Bull. 218: 1-68. Wai ler H. & H. LiETH. 1967. Klimadiagramme Weltallas. G. Fischer, Jena. GRASSES AND THE CULTURE HISTORY OF MAN J. M. J. deWet' Abstract The beginnings of plant and animal husbandry are lost in antiquity. It is not possible to determine from the available archaeological record when plant domestication was initiated. Changes in pheno- type, however, known to be associated with cereal species under domestication, and preserved m the archaeological record indicate that growing crops was an established way of life some 10,000 years ago. There are four major domesticated cereal complexes. Three evolved in the Old World and one in the New World. Wheat, barley, rye and oats are Near Eastern cereals, and spread across Eurasia early during the history of agriculture. Rice is the principal cereal of South Asia, sorghum and pearl millet the major cereals of the African savanna, and maize is a domesticate of Mesoamenca. Why the shift from hunter-gatherer to settled agriculturist occurred during the culture history of man is not known. Food production may have been initiated when man was faced with a gradual reduction in productivity of effort required to maintain accepted standards of living, traditional group size, and social organization. Once initiated, population pressures in particular will tend to demand agriculture. Hunter-gatherers live in equilibrium with the environment and have little lasting effects on nature. Farming, by its very nature, destroys the natural environment. Habitats are permanently altered, and a return to hunting and food gathering becomes impossible. Survival of civilized man has become absolutely dependent on cereal agriculture. Overpopulation, depletion of resources, planetary pol- lution, and the social ills of cities are penalties we have to pay for the pleasures of an abundant and stable food supply. Grasses have been playing a principal role in shaping the culture history of man every since he became sapient. They are basic to human life. The staple food of the great majority of mankind comes from grasses and they provide food for the grazing animals from which man derives most of his protein. The Gra- mineae is a relatively young but successful family. It includes an estimated 600 genera and 8,000 species that are widely distributed across the world. Grasses occur on all continents, including Antarctica, and are absent only from regions that are too barren or too cold to support the growth of flowering plants. The fossil record of grasses is meager, but there is good evidence to suggest that they emerged as a distinct family during late Cretaceous times when the flowering plants were spreading throughout the world. By early Miocene grasslands prob- ably were assuming a prominent place in the earth's vegetation, and it is estimated that at present almost one-quarter of the world's plant cover is composed of grass (Barnard, 1969). The evolution of the family has been strongly influenced by herbivorous mam- mals. Early ungulates were probably browsing, rather than grazing animals. Non- ruminant kinds of Artiodactyla appear in the fossil record from the early Tertiary. Ruminant forms, particularly the Bovinae, arose during the Miocene and have coevolved with grasses ever since. The transverse intercalary growth zone in leaves and above culm nodes, the short internodes of aerial stems of annual grasses, and the tufted habit of perennial grasses during vegetative growth, are adaptations to withstand grazing. Man appeared on the scene well after the Gramineae and grazing ungulates became widely dispersed. Hominid evolution dates back to the Pliocene, but the » Crop Evolution Laboratory, Department of Agronomy, University of Illinois, Urbana, Illinois 61801. Ann. Missouri Bot. Garo. 68: 87-104. 1981. (K)26-6493/8 1 /0087-0 1 04/$ 1 .95/0 88 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Voi . 68 genus Homo appears in the fossil record only since the Pleistocene, and our own species is probably less than 100,000 years old (Isaac & Leakey, 1979). Man started as a hunter and gatherer of plants for food. He competed with animals for the available supply of plant food, and hunted some of these same animals to satisfy his craving for meat. Gradually man learned how to control the food resources in the areas where he lived. Selected animals were protected from their natural enemies, and populations of favored food plants were increased through planting. Plant and animal domestications were initiated, and man started on a path of rapid social evolution. The Domestication Process The beginnings of domestication are lost in antiquity. It is not possible to determine from the available archaeological record when plant or animal husband- ry was first practiced. But, changes in phenotype, known to be associated with particular plant and animal species under domestication and preserved in the archaeological record, indicate that growing crops and herding animals were es- tablished ways of life some 10,000 years ago (Higgs, 1972). The transition from hunting-gathering to plant and animal husbandry occurred in different ways in different parts of the world. Different plant and animal species were domesticated at different times and places across the range of their intensive exploitation by man. Full pastoral nomadism is practiced only in the Old World, and only in areas where growing crops is impossible or totally unreliable as a constant source of food. In the wet tropics and high Andes of South America where grasses are not abundant, cultivation of tubers made settled agriculture possible. On the plains of Africa and Eurasia, and on the highlands of Mesoam- erica, grass cultivation provided the staple food for a sedentary way of life. Domesticated plants depend on man for suitable habitats. This dependence on man-disturbed habitats came about through selection pressures associated with harvesting and sowing (deWet & Harlan, 1975). Weeds are similarly adapted to permanently man-disturbed habitats. The major difference is that weeds are usu- ally spontaneous in this habitat while domesticated taxa depend on man for prop- agation. The ecological boundaries between the wild, weed and domesticated W Weed We 'flexci Fouta-Djalon highlands of Guinea it is a cultivated cereal (Porteres, 1951), and in Angola the species aggressively invades cultivated fields where it is often protected and harvested as a crop (deWel, 1977). Domesticated species may behave as weeds when they have not completely lost the ability of natural seed dispersal. True fonio, Digitaria e.xilis (Kippist) Stapf is widely cultivated in West Africa (Porteres, 1955). Its wild progenitor is not not known. But, the cultigen commonly escapes and occurs as a weed in the fields of other crops, where it is often harvested. Cultivated cereal species are annuals, and their closest wild relatives are ag- gressive annual colonizers of disturbed habitats. The only exception is Bromiis mango Desv. (mango). This biannual was cultivated in Chiloe province of central 1981] DE WET— GRASSES AND CULTURE 89 Chile, until the eighteenth century, when it was replaced by the introduced Old World cereals wheat and barley (Gay, 1865; Molina, 1782; Cruz, 1972). Today B. man^'o is known only as a wild plant, distributed in Central Chile and adjacent Argentina (Parodi & Hernandez, 1964). Colonizing ability is essential in domestication. Propagation by man of se- lected genotypes, however, constitutes domestication. Sowing of seeds harvested from a planted population results in loss of seed dormancy, an increase in seedling vigor, and eventually the inability of the cultigen to successfully invade natural habitats (deWet, 1975, 1979). Phenotypic traits associated with sowing of annual seed crops are uniform population maturity, and commonly also an increase in fruit size. Harvesting in association with sowing leads to a reduction in natural seed dispersal ability, and uniform individual plant maturity (Harlan et al., 1973). Phenotypic characteristics associated with harvesting are usually persistence of spikelets or florets on the inflorescence after maturity, and either reduction of inflorescence-bearing culms or synchronized tillering. Fully domesticated cereal species depend on man for a suitable habitat as well as seed dispersal. The do- mestication process continues as long as the cereal is harvested and sown in successive generations. Selection pressures associated with harvesting and sowing are commonly su- perimposed on pressures induced by man in selecting for traits to satisfy his fancies. This leads to rapid evolution under domestication. Genetic drift, selec- tion, and isolation allow for phenotypically different kinds of a domesticated species to be grown by different groups of people for the same or different uses. Cultivated grain sorghum. Sorghum hicolor (L.) Moench, as an example, is widely grown in Africa and Asia, morphologically variable, and frequently clas- sified into 28 species with 165 botanical varieties and several hundred recognized phenotypes (Snowden, 1936). Genetically, these taxa are conspecific and repre- sent little more than selections that are being maintained by man to suit his fancies and needs (deWet, 1978). Phenotypic changes associated with cereal domesti- cation are often grotesque. The spike of wild pearl millet, Pennisetum america- num (L.) Leeke is at most 10 cm long (Brunken, 1977), while some cultivars have spikes over one meter long (Brunken et al., 1977). Even more spectacular are the differences that distinguish maize {Zea mays L.) from its closest wild relative Z. mays subsp. mexicana (Schrad.) litis (teosinte). Maize and teosinte are genetically conspecific, and evidence presented by Beadle (1977) suggests that teosinte is wild maize. These two taxa differ con- spicuously from one another in four basic characteristics of the female inflores- cence. First, the inflorescence is distichous in teosinte and polystichous in maize. Second, female spikelets are solitary at a rachis node in teosinte and paired in maize. Third, female spikelets of teosinte are individually sunken into indurated cavities on the rachis, each of which is closed by an indurated outer glume to form a fruitcase. In maize the paired spikelets are located in shallow, more or less indurated cupules. Fourth, fruitcases disarticulate at maturity in teosinte but cupules are persistent in maize. These opposing traits are not simple genetic alternatives (Galinat, 1975). Intermediate expressions of tunicate alleles deter- mine induration, recessive alleles of a complex genetic system restores fertility to the rudimentary spikelet of a female pair, and recessive alleles of at least two 90 ANNALS OF THH MISSOURI BOTANICAL GARDEN [Vol 68 genes change the distichous spike of teosinte into the polystichous ear of maize (deWet, 1979). Phcnotypic changes associated with domestication are those selectively fa- vored by man. The three principal innovations of cereal domestication, loss of natural seed dispersal ability, uniform population maturity, and uniform individual plant maturity facilitate harvesting, and increase the percentage of harvestable seed. Phenotypic changes characteristic of individual cereal species under do- mestication further increase yield, enhance threshing, or are designed to suit the fancies of the cultivator. Increase in inflorescence size commonly accompanies a shift from wild to domesticated. Cultivated maize from the Coxcatlan culture phase in the Tehuacan Valley of Mexico had less than 100 kernels per ear. Less th:m two millennin Inter kernel number had increased to over 600 Der ear. Domesticated Cereals The principal use of grass by man is as cereals. Grasses are also highly valued as feed for livestock, and serve man in many other ways. Landscaping the habitat we live and play in is hardly complete without a lawn. There are tuif grasses selected for house lawns, golf courses, parks, and sport fields. Grasses are also planted as ornamentals. No tropical garden is complete without a clump of bam- boo, and Cortaderia seUoatui (Schult.) Ashers. & Graebn. (pampas) and Mis- canthus sinensis Anderss. (eulalia) are popular lawn ornamentals in warm and temperate regions. Grasses provide us with food other than cereal grains. Sugar is extracted from the stems of domesticated Saccharum officinamm L. (sugar cane). In the Far East Zizania latifolia Turcz. (water rice) is grown as a vegetable. Fungus infests the lower leafbases and the swollen stems are eaten. The young shoots of Sctaria palmifoUa (Koen.) Slapf G^ngle rice) are eaten in New Guinea, and the well-known bamboo shoots of commerce are harvested from species of Bambusch Dendrocalamus, and Phyllostachys. Grasses are also used as fuel, timber, roofing material, and as material from which paper, mats, and containers are made. Two species, Cymhopoi^on citratus (DC. ex Nees) Stapf (lemon grass) and Vetiveria zizcinioides (L.) Nash (vetiver), are commercially grown for their essential oils that are used in the perfume and pharmaceutical industries. Grasses serve man in many ways. But, it is their use as cereals that help shape his culture history. The caryopses of most grasses are edible, and at least 300 species are known to be harvested in the wild as cereals. Among these, 35 species belonging to 20 genera are, or were at one time grown as cereals (deWet, 1979). Brittle grass, Sctaria geniculata (Lam.) Beauv., is known as a cultivated cereal only in an archaeological context from Mexico (Callen, 1965, 1967). Canary grass, Phalaris canariensis L., is grown as a food for caged birds rather than as food for man. American wild rice, Zizania aquatica L., has been harvested commer- cially in the wild for centuries, but it is only during the last decade that this cereal has been successfully cultivated (deWet & Oelke, 1978). Ethiopian oats, Avena abyssinica Hochst., is not consciously sown but is accidentally planted and har- vested with other cereals. It is an obligate weed that has lost the ability of natural seed dispersal (Rajhathy & Thomas, 1974). The biannual Bromus mango (mango) was cultivated in central Chile until the eighteenth century, when it was replaced 1981] DE WET— GRASSES AND CULTURE 91 by wheat and barley. The remaining 30 cereal species are still cultivated, al- though several of them as minor crops. The major cereals are wheat {Triticum spp.), rice {Oryza sativa L.), maize {Zea mays), sorghum {Sorghum bicolor), and pearl millet {Pemusetum americanum) in order of their importance as human food. It is estimated that in 1976 some 413 metric tons of wheat, 344 metric tons of rice, and 335 metric tons of maize were produced by the world's cereal farmers. There are four major domesticated cereal complexes, each with a distinct geographic region of origin (Harlan, 1976). Three complexes evolved in the Old World and one in the New World. The small grains developed in the Near East, with wheat eventually becoming dominant across Eurasia, except for South Asia where rice is the principal cereal. The African savanna provided the world with sorghum and pearl millet, and maize is a domesticate of the highlands of Me- soamerica. The archaeological record suggests that southwestern Asia is the old- est region of cereal domestication, and that the knowledge of agriculture may have spread from here across the Old World and eventually to the New World (Carter, 1977). It is doubtful, however, that cereal agriculture evolved only once. Near Eastern cereals did become widespread across temperate Eurasia and Med- iterranean Africa soon after they became domesticated. But, there is no evidence to suggest that the cereal complexes in South Asia, the African savanna, or Mesoamerica were not independently domesticated. NEAR EASTERN COMPLEX Three species of wheat, Triticum monococcum L. (einkorn), T. turgidum L. (emmer) and T, x aestivum L. (bread wheat), barley {Hordcum vulgare L.), oats {Avena sativa L.) and rye {Secale cereale L.) are Near Eastern cereal domesti- cates. The Natufian of Palestine is the first known culture in the Near East that was equipped to extensively harvest and process small grains (Redman, 1977). Common tools used by the Natufian include grinding stones, stone mortars, and sickles with sheen on their stone blades. Extensive harvesting of cereals demand some kind of sickle, and since the florets of wild wheats and barley are tightly enclosed by glumes, processing is necessary to thresh the grain free. Remains of wild barley and wild einkorn appear in the archaeological record at Tell Mureybit (Syria) dating back to between 10,050 and 9542 B.P. (Renfrew, 1969). This probably is the period of initial cereal cultivation in the Near East. Loss of natural seed dispersal ability is commonly accepted as indicating domes- tication of cereals in an archaeological context. However, cereals may have been cultivated for many generations before this domesticated phenotype became es- tablished. Domesticated wheat and barley, totally dependent on man for seed dispersal, appear in the archaeological records between 9500 and 8500 B.P. from numerous occupation sites (Harlan, 1977) extending from Turkey to southwest- ern Syria and Palestine. The practice of growing wheat and barley reached Greece by 8000 B.P., and during the next 2,000 years spread along the valleys of the Danube and Rhine to the Netherlands, and along the Mediterranean to France. Bread wheat, a strictly domesticated taxon, appears around 8000 B.P. in such scattered settlements as Knossos in Greece, Hacilar and ^atal Hiiyiik in Turkey, Tel Ramad in Palestine, and Tepe Sabz in Iraq (Renfrew, 1973). Domesticated oats and rye are rare or absent among archaeological plant 92 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Voi 68 remains dating back more than 7,000 years, although wild oats in particular seems to have been commonly harvested for at least 1,000 years before the species became domesticated. Harlan (1977) suggests that these two cereals were added to the Near Eastern complex as secondary crops. They invaded cultivated fields as weeds, were harvested, and eventually became domesticated. Rye is wild in Anatolia and Transcaucasia, and Evans (1976) proposes that the weed race evolved after the spread of cereal cultivation into these regions. When or where rye was first cultivated is not known. Oat is wild in the Mediterranean Basin and widespread in the Near East (Baum, 1977). Cultivated oat is reported from Greece by the late eighth millennium B.P., but the crop only became important some five millennia later and then in Central Europe (Holden, 1976). There is no evidence for an incipient food production period in Europe before the introduction of domesticated Near Eastern cereals (Waterbolk, 1968). Lake- dwelling settlements such as at Niederwil in northeastern Switzerland, dating back to around 3700-3625 B.C., are characterized by fully domesticated barley and bread wheat, and Chenopodium album L., a species probably harvested in the wild as a pseudocereal (vanZeist & Casparie, 1974). Two other cereals, Pan- icuni nuliaceum L. (proso millet) and Setaria italica (L.) Beauv. (foxtail millet) occur in Swiss Lake Dwelling sites (Neuweiler, 1946). Wild foxtail millet occurs across Eurasia, but wild proso millet is known only from Central Asia (deWet et al., 1979; Kitagawa, 1979). It is possible that foxtail millet was independently domesticated in Europe and the Far East, but it seems likely that Panicum was introduced to Europe as a domesticated cereal (Werth, 1937). A single cereal, Avcna strigosa Schrebn. (sand oats) is a truly European domesticate. Wild sand oats occurs in the western Mediterranean region and was probably domesticated as a secondary crop in Iberia. By the end of the prehistoric period, settled farming based on plant and animal husbandry, was an established way of life in Europe (Waterbolk, 1968). The spread of wheat and barley east into the Indus valley and into China was later than west into Europe. They became important cereals during the Harappa- Mohenjo Daro period between 2500 and 1700 B.C. (Vishnu-Mittre, 1968; Allchin, 1969), and along the lower Indus Valley at Chirand between 2500 and 1800 B.C. (Vishnu-Mittre, 1974). These cereals did not reach China until about 3500 B.P. (Ho, 1969). Cereals were not the only field crops domesticated in the Near East. Almost as important as cereals are peas, Pisum sativum L. (Waines, 1975), broad beans, Vicia faha L. (Landizinsky, 1975; Zohary, 1977) and lentils. Lens culinaris L. (Ladinzinsky, 1979). These pulses appear in the archaeological record of the Near East as early as cereals. They spread with the cereals across temperate Eurasia, and together made possible the early civilizations of Mesopotamia, Egypt, Greece, Rome, and Europe. ASIAN COMPLEX Rice, Oryza sativa L., is the principal cereal of South and East Asia. Minor cereals belonging to the Asian complex are CoLx lacryma-johi L. (jobs tears) and Di^itaria cruciata (Nees) A. Camus (raishan) from Assam, Echinochloa colona (L.) Link (shama), Brachiaria ramosa (L.) Stapf (anda horra) and Panicum su- 1981) DE WET— GRASSES AND CULTURE 93 matrense Roth, ex Roem. & Schult. (sawan) from South India, Paspcdum scroh- iculatum L. (khodo millet) and Setaria ghiuca (L.) Beau v. (korali) from Central India, Echinochloa frumentacea (Roxb.) Link. (Japanese millet) from the Far East, and Digitaria sangiiinalis (L.) Scop, (manna) from Kashir. Manna was also widely collected as a wild cereal in southern Europe until historical times. It is known to be an important cultivated cereal only in Kashmir and the Caucasus. Proso millet [Panicum miliaccum) and foxtail millet {Setaria italica) are the only known Asiatic cereals that were grown in Europe during prehistoric times. The oldest known cereal agriculture in India was practiced by the Harrapans from the Punjab and the Indus Valley. They were growing wheat, barley, and sorghum by 4500 B.P. (Vishnu-Mittre, 1977). These are introduced cereals, wheat and barley from the Near East and sorghum from Africa. Indigenous rice was added to this complex during the late fifth millennium B.P., and other native cereals were locally domesticated during the Indian Bronze age (3700-3000 B.P.) or later. It is also during the Bronze age that African Eleusine coracana (L.) Gaertn. (finger millet) was introduced to South India. The oldest known Asiatic cereals are rice (Oryza saliva), proso millet (Pan- icum miliaceiim), and foxtail millet {Setaria italica). The two millets are ancient cereals from across Eurasia. They were grown in Central Europe some five mil- lennia ago (Heer, 1886) and were widespread in Yang-shao sites from southern Shensi and Shansi provinces in China dating back at least six millennia (Ho, 1977). Wild foxtail millet occurs across temperate Eurasia and could have been independently domesticated in Europe and China (deWet et al., 1979). Wild Pan- icum miliaccum is known only from Central Asia (Kitagawa, 1937, 1979), and proso millet seems to be truly a Chinese domesticate. If this is true, the presence of proso millet in Europe by 5000 B.P., and possibly in Greece by 8000 B.P. (Hopf, 1962), is surprising. In Europe this millet is commonly associated with wheat and barley. These Near Eastern cereals, however, did not reach China until 3500 B.P. (Ho, 1969). It seems possible that wild Panicum miliaccum is or was at one time more widely distributed than is presently known, and as is true of foxtail millet, was taken into cultivation independently in Europe and Asia. Rice first occurs in the archaeological record of China around 5700 B.P., but outside the assumed Neolithic nuclear area of plant domestication (Ho, 1977). It appears as a cereal in India more than a millennium later. Rice is known from around 4300 B.P. in Harappan sites from Rangpur and is widespread along the lower Indus Valley and southern India about 1,000 years later (Vishnu-Mittre, 1977). Solheim (1971) reports that rice impressions, dating back to around 5000 B.P., are present on pottery from Non Nok Tha in Thailand. It is not possible to determine whether these impressions are of wild or cultivated rice. Wild Oryza sativa is native across South Asia (Oka, 1974). The practice of cereal cultivation was probably introduced to northwestern India from the Near East, and from here or perhaps Africa to southern India some 5,000 years ago. Gorman (1977) proposes that rice was domesticated, to- gether with several rootcrops, in southeastern Asia, and suggests that the initial domestication dates back some nine millennia. Chin (1971), however, concludes that rice farming dates from around 5500 B.P. at Non Nok Tha and Ban Chiang in Thailand, essentially the same date that is suggested by Ho (1977) for rice 94 ANNAI.S OF THF MISSOURI BOTANICAL GARDEN [Vol. 68 cultivation in China. Chang (1976) suggests that rice was first cultivated in ancient India, and that the wide dispersal of this cultigen from this nuclear area led to the formation of three ecogeographic races, indica, sinica, and javanica. It seems equally likely, however, that rice was independently taken into cultivation in at least these three general regions, and that this noncentric (Harlan, 1971) domes- tication of rice allowed for the development of ecogeographic complexes. Rice is the principal cereal in South Asia. The other Asiatic cereals are im- portant in more localized regions. Proso millet {Panicum miliaceum) and foxtail millet {Sctaria itciUca) are extensively grown in Central Asia and India. Jobs tears {Coix lacryma-jobi) is cultivated in Assam and adjacent Burma, and the Philip- pines (Arora, 1977). Domesticated races of Coix from these two regions are distinctly different, and this crop may have been independently domesticated in northeastern India and the Philippines. Raishan {Digitaria cruciata) seems to be an endemic, and possible recent crop of the Khasi hills in Assam (Singh & Arora, 1972). Japanese millet {Echinochloa frumentacea) and the Indian shama {E. co- lona) may represent complexes of the same cereal species (Yabuno, 1966). Khodo millet {Paspalum scrohicidatum) has been widely grown across the Indian plains at least since the beginning of the Indian Bronze age (Kajale, 1977). Setaria glauca (korali) is a Kharif crop in central India, and Brahiaria ramosa (anda korra) is little more than an encouraged weed in South India. As in the Near East, pulses played an important part in the history of agri- culture in South Asia and the Far East. The soybean, Glycine max (L.) Merr., is an early domesticate from China (Hymowitz & Newell, 1977). Chickpea, Cicer aricntum L., black gram, Vigna mungo (L.) Hepper, pigeon pea, Cajanus cajan (L.) Hutch., are native crops of South Asia (Ladizinsky & Adler, 1976; Dana, 1976). The winged-bean, Psophocarpus tetragonolohus (L.) DC, is extensively grown in the tropics of Southeast Asia (Hymowitz & Boyd, 1977), and the mung bean, Vigna radiata (L.) Wikz., is important across the Orient for its nutritious seeds, edible pods, and young sprouts. AFRICAN COMPLEX Cereals grown during prehistoric times in Africa north of the Sahara are typ- ically Near Eastern in origin. Wheat and barley reached the lower Nile Valley by the middle of the seventh millennium B.P. (Wendorf & Schild, 1976) and were staple food crops of ancient Egypt (Wonig, 1886). This is not surprising. North Africa has a Mediterranean climate, and the Nile Valley floods during summer when crops from the African savanna can be grown (deWet & Huckabay, 1967). Near Eastern cereals reached the highlands of Ethiopia, probably not later than the beginning of the Christian era. But, here they were grown together with native African cereals. The West African savanna and East African highlands produced a cereal ag- riculture independent from that of the Near East. Native African domesticated cereals include Pcnnisetufu americanam (pearl millet) from the Sahel, Eleusine tefiZuQc) Trotter (tef) defl D. ihurua Stapf (black fonio), Oryza glahernma Steud. (African rice), and Sor- ghum hicolor (sorghum) from the savanna of western Africa. 19811 DH WET— GRASSHS ANH CULTURE 95 The antiquity of native cereal cultivation in Africa is not known. Archaeolog- ical remains of finger millet from Ethiopia are estimated to date back five millennia (Hilu et al., 1979). This archaeological race of Eleusine coracana has lost the ability of natural seed dispersal, and cereal agriculture in Ethiopia must be sub- stantially older. Wild Eleusine coracana is distributed across the highlands of East Africa, but this cereal was probably domesticated in Ethiopia (Hilu & deWet, 1976). Finger millet reached South India around 3500 B.P. (Vishnu-Mittre, 1977). Remains of pearl millet dating back to the third millennium B.P. were uncovered at a lake edge settlement in Mauritania (Munson, 1976). A sequence from gath- ering wild grasses to growing pearl millet is obvious in this settlement. However, it is unlikely that Pennisetum americanuni was domesticated at the western edge of the Sahel. Wild pearl millet occurs in the central Sahel and highlands of the central Sahara, the region where this cereal was probably first cultivated (Brunken et al., 1977). Pearl millet was introduced to India and became widely cultivated in semiarid parts of South Asia more or less at the same time as its probable spread across the arid savanna of Africa (Vishnu-Mittre, 1969, 1971). The known archaeological record of sorghum, the most important native cereal, dates back in Africa for only two millennia (Connah, 1967; Clark & Stemler, 1975). But, it was an important crop in India as early as three millennia ago (Vishnu-Mittre, 1974). It is, however, an African rather than Indian cereal domesticate (deWet & Harlan, 1971). Wild Sorghum hicolor extends across the African savanna and is commonly harvested as a cereal (deWet, 1978). Archaeological evidence of early cultivation of other native African cereals are almost completely lacking. The minor African cereals remain more or less endemic to the regions of their probable first cultivation. Wild Eragrostis pilosa (L.) P. Beauv., the probable wild ancestor of tef, is widely distributed and ex- tensively harvested in East Africa (Barth, 1821-1865), but tef is grown as a cereal only on the Ethiopian highlands. Animal fonio is an endemic crop of the Fouta- Djalon in Guinea, while true fonio, black fonio, and African rice are grown as cereals only in the West African savanna (deWet, 1977). African rice is primarily a crop of the Niger delta. It never became a major crop in the wet forest east of the Bandama river. Here the principal food is indigenous yams (Baker, 1962). The paleontological record shows that the Sahara was substantially wetter some 8,000 years ago than it is now, and it is known that around 7000 B.P. people with cattle, sheep and goats were camping along edges of shallow lakes and were probably harvesting wild cereals in areas that are now desert (Clark, 1976). During the next several millennia the Sahara became progressively drier, and by 4000 B.P. the desert extended across most of North Africa. It was probably these nomadic herdsmen, who migrated south into the present savanna, that eventually domesticated the wild cereals they used to harvest. Pearl millet and sorghum are the principal and possibly oldest domesticated cereals of the African savanna. They, together with cultivated indigenous yams and the pulses, Voandzeia suh- terranea (L.) Thauars. (bambara groundnut) and Vigna ungulata (L.) Walp (cow pea), made a settled way of life possible. These early farmers made possible cultures that eventually produced the magnificant terracotta and bronzes of Benin around the beginning of the Christian era (Shaw, 1976). Sorghum, pennisetum, cow^ pea, and the bambara groundnut spread east and south along the savanna. 96 ANNAL.S OF THE MISSOURI BOIANICAL GARDEN [Voi . 68 and they together with finger millet and Dolichos lahlah L. (an East African pulse) fed the iron age cuUures that flourished across southern Africa until the eighteenth century (Pagan, 1967; Robinson, 1966; Summers, 1958). NEW WORLD COMPLEX Five cereal species were domesticated in the New World. Brittle grass, Se- taria gcnicidata (Lam.) Beau v., was extensively harvested in northwestern Mexico and on the Mexican Central Plateau during prehistoric times (Callen, 1965). This species constituted an important part of the diet of the inhabitants of El Riego cave in Tamaulipas some 7,000 years ago. What is of interest, is that the grains of this cereal steadily increased in size during the 2,000 years it was used as a food, suggesting to Callen (1967) that it must have been cultivated. It was eventually abandoned as a cereal after the introduction of maize to the region and is not known to have been grown during historical times. Sauwi {Panicum sonorum Beal) is an endemic cultivated cereal of the Warihio Indians in southwestern Chihuahua (Gentry, 1942). It is grown among maize, harvested, and the grains are ground into flour and either mixed with milk to produce a drink, or mixed with maize flour to make pinole. Wild Panicum so- norum extends from southern Arizona to Honduras. The cultivated race retains the ability of natural seed dispersal. Escaped cultivated kinds, recognized by grains that are larger than those of wild races, are widespread around Alamos in Sonora, and around archaeological ruins in Nyarit and Jalisco, suggesting that sauwi may have been more widely cultivated before the introduction of corn to northeastern Mexico. It is surprising that Father Kino (1684-1685), a missionary and early explorer of northwestern Mexico did not mention this native millet in his writings. Today the species is valued more as a fodder for livestock than as a cereal grain. Mango {Bromus man^o) is the only known domesticated cereal that is not an annual. Claudio Gay (1865) was one of a few South American botanists who actually saw this cereal in relatively extensive cultivation. Gay writes that before the conquest, the people of central Chile made a kind of bread without yeast that they called covque, and that this bread was made from a native cereal known as mango. He visited central Chile in 1837 and found mango growing in two fields in the department of Castro. The species is biannual. At the time of his visit to this part of Chile, livestock were allowed to graze on the fields during the first season of growth. Plants were protected from livestock the next year and pro- duced a cereal crop in the fall. Mango was harvested, threshed, and the grain was ground into flour that was used to make bread or chicha. Bread and chicha made from mango were reported to be inferior to that made from the wheat and apples that were extensively grown around Castro by the early nineteenth century. Gay (1865) mentions a possible second cereal that used to be grown in Chile. It resembled barley, except that the grains were smaller, and it was harvested while the plants were green to prevent the inflorescence from shattering. Laet (1633) describes a cereal called teca, similar to barley, with stems like oats and grains a little smaller than those of rye. Whether teca is the second cereal de- scribed by Gay is impossible to ascertain with certainty (Parodi & Hernandez, 1981J DE WET— (JRASSHS AND CULTURE 97 1964). Later, Molina (1782) talks about two native cereals in Chile, ''el Magu/' a species of rye, and '1a Tuca/' a kind of barley. He unfortunately saw neither of these cereals in cuhivation. They were already replaced by wheat and barley when his natural history of Chile was written. Ball (1884) reports that the Arau- cano Indians of Bahia Blanca in Argentina use Bromus unioloides H.B.K. (5. catharticus Vahl) as a wild cereal. It is possible that this wild cereal represents the teca of Laet (1633) and tuca of Molina (1782). The only New World cereal of present day importance is Zea mays (maize). It was domesticated in Mesoamerica, became widely distributed across the New World by the time of conquest by Europeans, and has since been dispersed across all tropical and warm temperate farming regions of the Old World. Wherever it is adapted, maize is replacing native African and Asiatic cereals except for rice. Maize is unique in female inflorescence morphology among grasses. The paired female spikelets are arranged in more or less indurated cupules along and around a central rachis in usually eight or more rows. The closest wild relative of maize, Z. mays subsp. mcxicana (teosinte) has a distichous female inflores- cence. The oldest indisputable maize known in the archaeological record comes from the Coxcatlan culture phase of the Tehuacan Valley in Mexico (Mangelsdorf, MacNeish & Galinal, 1967a). This maize dates back some 7,000 years and is assumed to be wild maize by Mangelsdorf and his coworkers (Mangelsdorf, 1974). Beadle (1977) points out, however, that this maize lacks the ability of natural seed dispersal and must have been cultivated. The antiquity of cereal agriculture in the New World is not known. MacNeish (1971) suggests that Barranca horticulture started in Tehuacan Valley between 7000 and 5000 B.C., and Niederberger ( 1979) points out that fully sedentary com- munities existed in the Basin of Mexico at least since the sixth millennium B.C. Be that as it may, maize was grown on the Mexican plateau some 7,000 years ago, and if teosinte is wild maize, which is almost certain, it must have been taken into cultivation considerably earlier than during the Coxcatlan culture phase of Tehuacan. Teosinte does not occur in the arid Tehuacan Valley, but is distrib- uted along the western escarpment of the mountains and the wetter parts of the Mexican plateau (Wilkes, 1977). From its center of domestication in Mesoamerica maize spread rapidly across the Americas. It is known to have been cultivated in New Mexico and Arizona not later than 4000 B.P. (Mangelsdorf, Dick & Camara-Hernandez, 1967), and also reached northwestern Mexico as a cereal some 4,000 years ago (Mangelsdorf, MacNeish & Galinat, 1967b). Maize was grown in Ecuador during the early for- mative stage dating back some 5,000 years (Zevallos et al., 1977). In the Americas maize evolved with beans {Phaseolus vulgaris L. and P, lunatus L.), squash {Cucurhita spp.), and amaranths as staple foods (Kaplan, 1965; Gentry, 1969; Baudet, 1977). Only in the wet tropics and high Andes are tubers more important than maize. Indeed, maize is such an excellent cereal, with such good nutritious qualities and a wide range of adaptations, that its do- mestication probably excluded several other potential cereals from being taken into cultivation. It is known that maize replaced Setaria geniculata in Mexico and Bromus mango in Chile as planted cereals. 98 ANNAIS OF THE MISSOURI BOTANICAL GARDHN |Voi . 68 Indian wild rice {Zlzania aquatica) is a newly domesticated cereal. The species is widely distributed in temperate North America from the Dakotas east to the Atlantic coast and south to Florida and Texas. In the northeast, and in Wisconsin, Minnesota, and adjacent Canada wild rice is extensively harvested. Hofstrand (1970) estimates that some 40,000 acres of natural stands are com- mercially harvested in Minnesota and Wisconsin alone. The range of wild rice has probably been extended by sowing ever since it was first used as a cereal. It is easy to establish along shores of shallow lakes. Attempts to grow Zizania successfully, however, have until recently failed. The first serious attempts to grow wild rice commercially in paddies date back less than two decades (Oelke et al., 1973). Since 1971 production of wild rice in man- made habitats exceeds that harvested from natural stands in Minnesota (Brunson, 1972). Success in growing Zizania followed the discovery of a population with poor seed dispersal ability. It is estimated that in wild stands some 90% of grain escapes the harvester due to natural seed dispersal. Less than 50% of the grain is naturally dispersed in improved races grown commercially (deWet & Oelke, 1978). Origins and Consequences of Food Production The shift from hunting and food gathering to animal and plant husbandry ranks with the industrial revolution as one of the great achievements of man. Hunting- gathering is not necessarily a difficult way of life. Gathering wild food is less labor demanding than growing the same plants for food (Bronson, 1977). Bushmen and other present-day nomadic hunter-gatherers devote at most a few hours a day to subsistance activities (Lee, 1972). However, a nomadic way of life limits cultural evolution. Settled communities demand a regular food supply in the area of set- tlement. This is best achieved through agriculture. Why the shift from hunter-gatherer to settled agriculturist occurred during the culture history of man is not known. Man must have had a basic knowledge of plant cultivation long before he actually started to domesticate plants by con- sciously sowing what was harvested from a planted population. The Indians of the Great Basin of western North America were specialized harvesters of wild cereals, who sowed to increase the population density of the fields to be har- vested. Similarly, gardens of wild food plants are often maintained around tem- porary settlements of nomadic herdsmen. Sowing experiments are continued for a few generations and then abandoned. Plant and animal husbandry evolved over several millennia, and settled farm- ing is a relatively recent innovation of man that does not date back much beyond 10,000 years. The possible reasons why farming did not evolve earlier during the culture history of man are discussed by Bronson (1977). He suggests four possible explanations for this delay in food production. The first is labor saving. As already pointed out, gathering plant food and hunting are less labor intensive than farming. Second, time was required for reliable cultivated crops to evolve. Food crops probably were grown for millennia before evidence of domestication became obvious in the archaeological record. Third, farming de- I98I] DE WFT— GRASSES AND CULTURE 99 veloped as an adaptive response to increase in population numbers. Plant hus- bandry was not necessary when populations were small (Cohen, 1977). Fourth, farming involves risk. The hunter-gatherer had to invest a substantial amount of labor and resources in growing food crops without a guarantee of success. Crop failure is still common and often leads to famine. Smith (1972) suggests that food production was initiated on a number of oc- casions by different groups of hunter-gatherers when faced with a gradual reduc- tion in productivity of effort required to maintain the "culturally approved stan- dard of living and the traditional group size and social organization.'' Once initiated, population pressures in particular will tend to demand and intensify agricultural activities. A combination of these and other factors probably led to the beginnings of agriculture. From a botanical point of view, certain species lend themselves to domestication while others are almost impossible to domesticate (deWet & Har- lan, 1975; deWet, 1979). American wild rice, Zizania aquatica, does not readily adapt to man-made habitats. Although it was extensively sown in natural habitats, it was not domesticated by Amerinds (Dore, 1969; deWet & Oelke, 1978). Only natural colonizers are readily cultivated as annual seed crops. Planting of seeds harvested from previously man-sown populations may have first become impor- tant to preserve selected traits. It seems likely that teosinte was taken into cul- tivation to preserve a newly discovered tunicate allele. The oldest known maize from Tehuacan is distinctly tunicate. It has soft, papery glumes. This mutation not only induces some stability to the rachis of teosinte, and thus facilitates harvesting, but also greatly enhances threshing the grain from the otherwise in- durated glumes. The common characteristic of domesticated cereals, their in- ability of natural seed dispersal, may also have been a trait that was consciously selected by man. Harvesting by sickle (Wilke et al., 1972) in association with sowing (Harlan et al., 1973), however, automatically leads to a gradual loss of mechanisms for efficient natural seed dispersal. Be that as it may, wherever grasses were extensively harvested and agriculture was possible, selected species that were preadapted to withstand habitat disturbances by man eventually became domesticated. Cereal agriculture seems to be oldest in Southwest Asia. Carter (1977) argues that the idea of plant and animal husbandry spread slowly from this nuclear center across the Old World and eventually to the Americas. Conclusive evidence for a diffusion of the knowledge of agriculture between the Old and New Worlds is lacking, however, and it seems more likely that agricuhure evolved independently in several regions of the Old and New Worlds. Different species were domesti- cated at different times in the Near East, Far East, African savanna, and New World without evidence of exchange of crops among centers during the earlier stages of domestication. Civilizations developed independently in the Old and New Worlds until late historical times. Accidental contacts may have occurred between Europe and. the east coast of North America, and between the west coast of both Mesoamerica and South America and Asia (Carter, 1977; Lathrap, 1977). However, even if true, cultural exchange must have been minimal. It was not until the fifteenth century that Old World crops were brought to the New 100 ANNALS OF THE MISSOURI BOTANICAL GARDLN [Voi . 68 World in exchange for such important New World crops as maize, potato, sweet potato, garden beans, tomato, and peppers. Cultural contact between the Near East and Europe is obvious in the archae- ological record at least since 8000 B.P. Near Eastern agriculture also spread to ancient Egypt where wheat and barley became the basis on which dynastic Egypt was built. Near Eastern influences probably reached the African savanna civili- zations around the seventh century A.D. with the expansion of Islam across North Africa (Lewicke, 1974). Sorghum races that evolved in India after some 3,000 years of isolation were reintroduced to Africa during this time (Harlan & Stemler, 1976). Contact of the Near East with the Far East occurred around 1500 B.C. when wheat and barley reached China (Ho, 1969). Food production has advantages over food gathering. However, several so- cieties never adopted agriculture. The aborigines of Australia, the bushmen of southern Africa and several jungle tribes in Africa, Asia and the Americas, even today shun a settled way of life. Comparing these hunter-gatherer societies with farmers, it is obvious that food production has substantially accelerated the social evolution of man during the 400 or so generations since the beginnings of cereal cultivation (Childe, 1936). What are the consequences of food production? Rousseau (1755) argues that agriculture made organized work a necessity, forests were destroyed to make way for crops, individual ownership of property was introduced, and slavery developed. Survival of "civilized'" man became dependent on cereal agriculture. Overpopulation, depletion of the world's resources, planetary pollution, and the social ills of cities are penalties we have to pay for the pleasure of an abundant and stable food supply. The social consequences of village life are discussed by Smith (1972), and fall outside the scope of this manuscript. More important from the point of view of a botanist is the impact of farming on the immediate environment. Food produc- tion permits a greatly increased population density per unit area over that char- acteristic for hunter-gatherers. Cipolla (1964) estimates that the world's popula- tion was between five and ten million some 10,000 years ago when man first practiced conscious plant and animal husbandry. Hunter-gatherers live in equi- librium with the environment and have little lasting effects on nature. Farming by its very nature destroys the natural environment. Habitats are permanently altered, and a return to hunting and food gathering becomes impossible. Food production, for better or for worse, is here to stay. We must learn to cope with the pleasures as well as ills of civilization. Weeds and domesticates evolved in the permanently disturbed man-made hab- itat. Domesticates are cared for, and we depend on them for our food supplies. They are carefully selected for total fitness in specific environments, and their population sizes are carefully controlled by man. Weeds are spontaneous in the man-disturbed habitat. Great effort is required to maintain their population growth, lest they totally colonize the habitat in which man grows his crops. Man, the ultimate weed, must learn to control his own population growth. Food pro- duction can keep pace with the human population explosion for the immediate future. However, planetary pollution, shortages of natural resources, and the 1981] DH WHT— GRASSES AND CULTURE 101 numerous social ills that accompany overpopulation are problems that will in- tensify. Only when these problems are solved can man hope to again live in harmony with his environment. LiTEZRATURH CiTED Allchin, F. R. 1969, Early cultivated plants in India and Pakastan. Pp. 317-322, in P. J, Ucko & G.W. Dimbleby (editors). The Domestication and Exploitation of Plants and Animals. Aldine Publ. Co., Chicago. Arora, R. K. 1977. Job^s-tears {Coi.x lacryma-johi) — a minor food and fodder crop of northeastern India. Econ. Bot. 31: 358-366. Baker, H. G. 1962. Comments on the thesis that there was a major centre of plant domestication near the headwaters of the river Niger. J. African Hist. 3: 229-233. Ball, J. 1884. Contributions to the flora of North Patagonia and adjacent territory. J. Linn. Soc. Bot. 21: 203-240. Barnard, C. (editor). 1969. Grasses and Grasslands. MacMillan and Co., London. Barth, M. 1821-1865. Reisen und Entdeckungen in Nord und Central Africa. 5 vols. P. Perthes, Gotha. Baudft. J. C. 1977. Origine et classification des especes cultivees du genre Phaseolus. Bull. Soc. Roy. Bol. Belgique. 110: 65-76. Baum, B. R. 1977. Oats: Wild and Cultivated |A monograph of the genus Avena L. (Poaceae)]. Canad. Dept. Agric. Biosyst. Inst. Monogr. 14. Thorn Press, Ottawa. Beadle, G. W. 1977. The origins of Zea mays. Pp. 615-635, in C. A. Reed (editor). Origins of Agriculture. Moulon Publ., Ihe Hague. Bronson, B. 1977. The earliest farming: Demography as cause and consequence. Pp. 23^8, in C. A. Reed (editor). Origins of Agriculture. Moulon Publ., The Hague. Brunken, J. N. 1977. A systematic study oi Pcnnisctuni sect. Pennisetum (Gramineae). Amer. J. Bot. 64; 161-176. , J. M. J. deWet & J. R. Harlan. 1977. The morphology and domestication of pearl millet. Econ. Bot. 31: 163-174. Brunson, E. 1972. The taming of wild rice. Minn. Vol. 35: 44-51. Callen, E. O. 1965. Food habits of some Pre-Columbian Mexican Indians. Econ. Bot. 19: 335-343. . 1967. The first New World cereal. Amer. Antiquity 32: 535-538. Carter. G. F. 1977. A hypothesis suggesting a single origin of agriculture. Pp. 89-133, in C.A. Reed (editor). Origins of Agriculture. Mouton Publ., The Hague. Chang, T. 1976. The origin, evolution, cultivation, dissemination and diversification of Asiatic and African rices. Euphytica 25: 425^41. Childe, V. G. 1936. Man Makes Himself. Watts, London. Chin, You-di. 1971. Archaeological evidence for rice in Thailand. Silpakon 15: 39-44. CiPOLLA, C. M. 1964. The Economic History of World Population. Harmondsworth, London. Clark. J. D. 1976. Prehistoric populations and pressures favoring plant domestication in Africa. Pp. 67-105, in J. R. Harlan. J. M. J. deWet & A. B. L. Stemler (editors). Origins of African Plant Domestication. Mouton Publ., The Hague. & A, B. L. Stemler. 1975. Early domesticated sorghum from Central Sudan. Nature 254: 589-59 1 . Cohen, M. N. 1977. Population pressure and the origins of agriculture: An archaeological example from the coast of Peru, Pp. 135-177, in C. A. Reed (editor), Origins of Agriculture. Mouton Publ., The Hague. CoNNAH, G. 1967. Progress report on archaeological work in Bornu 1964-1977 with particular ref- erence to the excavations at Daima mound. Northern Hist. Res. Scheme, 2nd Interim Report, Zaria, Nigeria. Cruz, W. A. 1972. Bromus mangiK a disappearing cereal. Idesia 2: 127-131. Dana, S. 1976. Origin, evolution and distribution of some grain legumes. Indian J. Genet. PI. Breed. 36: 143-145. deWet, J. M. J. 1975. Evolution dynamics of cereal domestication. Bull. Torrey Bot. Club 102: 307-312. . 1977. Domestication of African cereals. African Econ. Hist. 3 (Spring): 15-32. . 1978. Systematics and evolution oi Sorghum sect. Sorghum (Gramineae). Amer. J. Bot. 65: 477-484. — . 1979. Principles of evolution and cereal domestication. Pp. 269-282, in A. C. Zeven & A. M. van Harten (editors), Broadening the Genetic Base of Crops. PUDOC, Wageningen. 102 ANNALS OF THE MISSOURI BOTANICAL GARDLN [Vol. 68 — & J. R. Harlan. 1971. The origin and domestication of Sorghum hicolor. Econ. Bot. 25: 128-135. — & . 1975. Weeds and domesticates. Evolution in the man-made habitat. Econ. BoL 29: 99-107. — & J. P. HucKABAY. 1967. The origins of Sorghum hicolor. IL Distribution and domestication. Evolution 21: 787-802. & E. A. Oei KE. 1978. Domestication of American wild rice (Zizania aquatica L., Grami- neae). J. Agric. Tradil. Bot. Appl. 25: 67-84. , L. L. Oestry-Stidd & J. L Cubero. 1979. Origins and evolution of foxtail millets (Setaria ifalica). J. Agric, Tradit. Bot. Appl. 26: 53-64. DoRE, W. G. 1969. Wild rice. Canada Dept. Agric. Res. Publ. 1393: 1-84. Evans, G.M. 1976. Rye— Secale cereah (Gr^mincdQ-TviUcin^c). Pp, 108-1 1 1, m N. W. Simmonds (editor). Evolution of Crop Plants. Longman, London. Pagan, B. M. 1967. Iron Age Cultures in Zambia. Chatto and Windus, London. Galinat, W. C. 1975. The evolutionary emergence of maize. Bull. Torrey Bot. Club 102: 313-324. Gay, C. 1865. Historia Fisica y Politica de Chile— Agricultura Chilena. 2 vols. Paris. Gentry, H. S. 1942. Rio Mayo plants. Publ. Carnegie Inst. Wash. 527: 1-328. . 1969. Origin of the common bean, Phaseohis vulgaris. Econ. Bot. 23: 56-59. Gorman, C. 1977. A priori models and Thai prehistory: A reconstruction of agriculture in South- Hague. Publ Harlan, J. R. 1971. Agricultural origins. Centers and noncenters. Science 174: 468^74. . 1976. Plant and animal distribution in relation to domestication. Philos. Trans., Ser. B 274: 13-25. 1977, The origins of cereal agriculture in the Old Worid. Pp. 357-383, in C. A. Reed (editor), Origins of Agriculture. Mouton Publ., The Hague. — & A. B. L. Stemler. 1976. The races of Sorghum in Africa. Pp. 465^78, in J. R. Harian, J. M. J. deWet & A. B. L. Stemler (editors). Origins of African Plant Domestication. Mouton Publ., The Hague. , J. M. J. deWet & E. G. Price. 1973. Comparative evolution of cereals. Evolution 27: 311- 325. Helr, O. 1886. Die Pflanzen der Pfahlbauten. Neujahrsblatt der Nalurforschenden Gessellschaft, Zu- rich. HiGGS, E. S. 1972. Papers in Economic Prehistory. Cambridge Univ. Press, Cambridge. Hn u, K. W. & J. M. J. deWet. 1976. Domestication of Eleusine coracana (L.) Gartner. Econ. Bot. 30: 199-208. & J. R. Harlan. 1979. Archaeobotanical studies of Eleusine coracana suhsp. cor- acana (finger millet). Amcr. J. Bot. 66: 330-333 Ho, Ping- 1 1. I%9. The Loes and the Origins of Chinese Agriculture. Chinese Univ., Hong Kong. . 1977. The indigenous origins of Chinese agriculture. Pp. 413-484, in C. A. Reed (editor), Origins of Agriculture, Mouton Publ., The Hague. Hofstrand, R, H. 1970. Wild ricing. Nat. Hist. 79: 50-55. HoLDEN, J. H. W. 1976. Oiits—Avena ssp. (Gramineae-Aveneae). Pp. 86-90, in N. W. Simmonds (editor). Evolution of Crop Plants. Longman, London. HoPF, M. 1962. Bericht uber die Untersuchung von Samen und Holzkohleresten von der Agrissa- Maghula aus der prakermischen bis mitlelbronzezeitlichen Schichten. Pp. 101-110, in V, Mi- lojcic, J. Boessne & M. Hopf (editors), Die Deutschen Ausgrabungen auf der Agrissa-Maghula in Thessalien. Bonn. HvMowiTZ, T. & J. Boyd. 1977. Origin, ethnobotany and agricultural potential of the winged-bean — Psophocarpus tetragonolohus. Econ. Bot. 31: 180-188. & C. A. Newell. 1977. Current thoughts on origins, present status and future of soybeans. Pp. 197-209, //; D. S. Seigler (editor). Crop Resources. Academic Press, New York. Isaac, G. & E. F. Leakey. 1979. Human Ancestors. Scientific American Publ., Freeman and Co., San Francisco. Kajale, M. D. 1977. On the botanical findings from excavations at Daimabad, a Chalcolithic site in western Maharashtra, India. Curr. Sci. 46: 818-819. Kaplan, L. 1965. Archaeology and domestication in American Phaseolus (Beans). Econ. Bot. 19: 358-368. Kino, E. F. 1684-1685. First from the Gulf to the Pacific: The diary of the Kino Atondo pennisular expedition— December 14, 1684-January 13, 1685. Transcribed and translated, and edited by W. M. Mathes. 1969. Dawsons, Los Angeles. KiTAGAWA. M. 1937. Contributio ad cognitionem florae Manshuricae. X. Bot. Mag. (Tokyo) S|: 150-157. 1981] DE WFT— GRASSES AND CULTURE 103 1979. Neo-Lineamente Florae Manshuricae. J. Cramer, Vaduz. Landizinskv, G. 1975. On the origin of the broadbean, Viciafaha L. Israel J. Bol. 24; 80-88, . 1979, Origin oflentil and its wild gene pool. Euphytica 28: 179-188. & A. Adier. 1976. Origin of chickpea, Cicer arientium L. Euphytica 25: 211-218. Laet, J. DE. 1633. Novus Orbis seu Descriptionis Indiae Occidentalis. El Libro XII. Apud. Elzei- viros, Lugd., Batavia. Lathrap, D. W. 1977. Our father the Cayman, our mother the Gourd: Spinden revisited, or a unitary model for the emergence of agricukure in the New World. Pp. 713-751, in C. A. Reed (editor). Origins of Agriculture. Moulon Publ., The Hague. Lee, R. B. 1972. Work effort, group structure, and land-use in contemporary hunter-gatherers. Pp. 177-185, in P. J. Ucko, R, Tringham & G. W. Dimbleby (editors), Man, Settlement and Ur- banism, Schenkman, Cambridge, Massachusetts. Lewicke, T. 1974. African Food in the Middle Ages. Cambridge Univ. Press, Cambridge. MacNeish, R. S. 1971. Speculation about how and why food production and village life developed in the Tehuacan Valley, Mexico. Archaeology 24: 307-315. Mangelsdorf, p. C. 1974. Corn, Its Origin, Evolution and Improvement. Belknap, Harvard Univ. Press, Cambridge. , H. W. Dick & J. Camara-Hernandez. 1967. Bat cave revisited. Bot. Mus. Leaf!. 22: 1- 31. , R. S. MacNeish & W. C. Gai inat. 1967a. Prehistoric wild and cultivated maize. Pp. 178 ^ ... ^^ w _ - ^ . . -_- — — — — — — — — — 200, in D. S. Byers (editor). The Prehistory of the Tehuacan Valley. Vol. 1. Univ. Texas Press, Austin. & . 1967b. Prehistoric maize, leosinte and Tripsacum from Tamaulipas, Mex- ico, Bot. Mus. Leafl. 22: 33-62. Molina, G. 1. 1782. Saggio Sulla Storia Naturale de Chile. Bologna, Italia. MuNSON, P. J. 1976. Archaeological data on the origins of cultivation in the southwestern Sahara and their implications for West Africa. Pp. 187-209, in J. R. Harlan, J. M. J. deWet & A. B. L. Stemler (editors), Origins of African Plant Domestication. Mouton Publ., The Hague. Neuwhii ER, E. 1946. Nachtrage Ungeschichtlichen Pflanzen. Vierteljahrschr. Natuif. Ges. Zurich 91: 122-136. NiEDERBERGER, C. 1979. Early sedentary economy in the basin of Mexico. Science 203: 131-142. Oelke, E. p.. W. a. El liot, M. E. Kernkamp & D. W. Noetzel. 1973. Commercial production of wild rice. Univ. Minn. Agric. Ext. Serv. Folder 284. Oka, Hiko-ichi. 1974. Experimental studies on the origin of cultivated rice. Genetics 78: 475^86. Parodi, L. R. & J. C. Hernandez. 1964. El Mango, cereal extinguido en cultivo, sobre vive en estada salvaje. Ci. & Invest. 20: 543-549. Porteres, R. 1951. Une cereale mineure cultivee dans TOuest-Africain (Brachinria deflexa C. E. Hubbard var. sativa nov. var.). L'Agron. Trop. 6: 39^2. 1955. Les cereales mineures du genre Di^itaria en Afrique et Europe. J. Agric. Trop. Bol. Appl. 2: 349-675. Rajhathy, T. & H. Thomas. 1974. Cytogenetics of oats {Avena L.). Miscel. Publ. Genet. Soc. Canada 2: 1-90. Redman, C. L. 1977. Man, domestication, and culture in Southwestern Asia. Pp. 523-541, in C. A. Reed (editor). Origins of Agriculture. Mouton Publ.. The Hague. Renfrew, J. M. 1969. The archaeological evidence for the domestication of plants: Methods and problems. Pp. 149-172, in P. J. Ucko & G. W. Dimbleby (editors), The Domestication and Exploitation of Plants and Animals. Aldine Publ. Co., Chicago. 1973. Paleobotany. The Prehistoric Food Plants of the Near East and Europe. Columbia Univ. Press, New York. Robinson, K. R. 1966. The Leopard's Kopje culture, its position in the Iron Age of Southern Rhodesia. S. African Archaeol. Bull. 21: 5-51. Rousseau, J. 1755. Discours sur Torigine et les fondements de Tinegalite parmi les hommes. Paris. Shaw, T. 1976. Early crops in Africa: A review of the evidence. Pp. 107-153, in J. R. Harlan, J. M. J. deWet & A. B. L. Stemler (editors). Origins of African Plant Domestication. Mouton Publ., The Hague. Singh. H. B. & R. K. Arora. 1972. Raishan (Di^itaria sp.) — A minor millet of the Khasi Hills, India. Econ. Bot. 26: 376-380. Smith, P. L. 1972. The consequences of food production. Addison Wesley Module in Anthropology 31: 1-38. Snowden, J. D. 1936. The Cultivated Races of Sori^hum. Allard and Son, London. SoLHEiM, W. G. 1971. New light on a forgotten past. Natl. Geogr. Mag. 139: 330-339. Summers, R. 1958. Inyanga. Cambridge Univ. Press, Cambridge. 104 ANNAIS or IHH MISSOURI BOIANICAI. GARDKN [Vol . 68 vanZeist, W. & W. A. Caspakie. 1974. Niederwil, a paleobotanical study of a Swiss neolithic lake shore settlement. Geol. Mijnbouw 53: 415-428. ViSHNU-MiTTRE. 1968. ProtohistoHc records of agriculture in India. Trans. Bose Res. Inst. 31: 17- 106. . 1969. Remains of rice and millet. Pp. 229-235, in H. D. Sakalia ct al. (editors), Excavations at Ahar (Tambavati). Poona. . 1971. Ancient plant economy at Hallur. Pp. 125-133, in M, S. Nagaraja Rao (editor). Proto- historic Cultures of the Tungabhadra Valley. Dharwar. . 1974. The beginnings of agriculture: Palaebotanical evidence in India. Pp. 3-30, in Sir Joseph Hutchinson (editor). Evolutionary Studies in World Crops. Diversity and Change in the Indian Subcontinent. Cambridge Univ. Press, Cambridge. . 1977. Changing economy in ancient India. Pp. 569-588. in C. A. Reed (editor). Origins of Agriculture. Mouton Publ., The Hague. Waines, J. G. 1975. The biosystematics and domestication of peas {Pisum L.). Bull. Torrey Bot. Club 102: 388-395. Waterboik, H. T. 1968. Food production in prehistoric Europe. Science 162; 1093-1102. Wendorf, F. & R. SCHILD. 1976. The use of ground grain during the late Paleolithic of the lower Nile valley, Egypt. Pp. 269-288, in J. R. Harlan, J. M. J. deWet & A. B. L. Stemler (editors). Origins of African Plant Domestication. Mouton Pub!., The Hague. Wkrth, E. 1937. Zur Geographic und Geschichtc der Hirzen. Angew. Bot. 19: 42-88. WiiKE, P, J., R. Bettinger, R. F. King & J. F. O'Connell. 1972. Harvest selection and domes- tication in seed plants. Antiquity 46: 203-208 . Wilkes, H. G. 1977. Hybridization of maize and tenosinte, in Mexico and Guatemala and the improvement of maize. Econ. Bot. 31: 254-293. WoNiG, F. 1886. Die Pflanzen in alten Aegypten. Albert Heitz, Leipzig. Yabuno, T. 1966. Biosystematic study of the genus Echinochloa (Gramineae). Jan. J. Bot 19" 277- 323. Zevali OS M., C, W, C. Galinat, D, W. Latmrap, E. R. Leng, J. G. Marcos & K. M. Ki umpp. 1977. The San Pablo corn kernel and its friends. Science 196: 385-389. ZoHARV, D. 1977. Comments on the origin of cultivated broad bean, Viciafaha L. Israel J. Bot 26* 39^2. THREE NEW SPECIES OF FLOURENSIA (ASTERACEAE-HELIANTHEAE) FROM SOUTH AMERICA 1 Michael O. Dillon^ Abstract Three new species of Flourcnsia (Asleraceae-Helianlheae) from South America are described t . peruviana Dillon and F. pofycephala Dillon from central and southern Peru respectively: and F hlakcana Dillon from northcentral Argentina. The genus Flourensia is represented by at least four species in Peru, two of these described for the first time within this paper. Flourcnsia macrophylla Blake is represented by scattered populations in valleys of various rivers draining into the Pacific along the western slope or Cordillera Occidental from northern to central Peru (2,500-3,500 m). Flourensia angustifolia (DC.) Blake is found in intermontane valleys associated with the tributaries of the Rio Perene and Rio Huallaga in the Cordillera Central (1,700-3,300 m). Flourensia peruviana Dillon and F. polycephala Dillon are distributed in dry sites in valleys with eastern drainage from southcentral and southern Peru. Flourensia heterolepis Blake is represented by scattered populations in the Cordillera Real of southeastern Bo- livia (ca. 2,700 m), a distance of ca. 725 km from the southeastern Peruvian taxa. Each species is readily distinguished by a complement of morphological char- acters and a distinct geographical distribution. The distributional pattern of these taxa corresponds with regions postulated to have undergone a series of humid- arid cycles during the Quaternary, which drastically and repeatedly altered vege- tation patterns (Vuilleumier, 1971; Simpson, 1975). While it is difficult to accu- rately determine when and how these taxa attained their present distributions, a rather recent radiation is suggested. A similar pattern of species distribution is exhibited in other taxa occupying various habitats in the Peruvian Andes (Simp- son, 1975). An additional example is found in Teconia (Bignoniaceae) (Gentry, 1979) which has unique taxa in each of the major valleys similarly occupied by different Flourensia species. The following key compares the salient differences between all the Flourensia species of Peru and adjacent Bolivia. Figure 1 illustrates the distribution of these species. la. Leaves shallowly denticulate. 2a. Leaves oval to oblong-oval, the apex obtuse to subobtuse F. macrophylla 2b. Leaves lanceolate to narrowly-elliptic, the apex acute F. angustifolia lb. Leaves strictly entire. 3a. Outer phyllaries 2-3 mm long, the inner 3-5 mm long, all ca. 1.5 mm wide F. peruviana ^ I thank Dr. W. G, D'Arcy for providing the Latin descriptions, and Dr. A. Gentry for reviewing the manuscript. I wish to acknowledge the support of field work by an N.S.F. predoctoral dissertation improvement grant and the N.S.F, Flora of Peru grant. Illustrations were prepared by Elizabeth Liebman, Exhibition Department, Field Museum of Natural History. - Botany Department, Field Museum of Natural History, Roosevelt Road at Lakeshore Drive, Chicago, Illinois 60605, U.S.A. Ann. Missouri Bot. Card. 68: 105-111. 1981. 0026-6493/81/0105-01 1 1/$0.85/0 106 ANNAI S OFTHt, MISSOURI BOTANICAL GAROHN [Vol . t,H Flourensia macrophylla angustifol la FiGURt 1. Distribution map of Flourensia species in Peru and adjacent Bolivia 3b. Outer phyllaries 5-9 mm long, the inner 8-10 mm 4a. Inflorescences 2-5-flowered, terminal and axi 4b. Inflorej; s 4-8-flowered, cy long, all (l-)2-3(-3,5) mm wide, ary (central Bolivia) F. hetcrolcpis _._ F, polycephala Flourensia polycephala Dillon, sp. nov. — Fig. 2. Species haec ab F. hetcrolcpis Blake dift'ert capitulis numcrosis, inflorescentia cymosa, phyllariis extimis (5-)7-8 mm longis, ca. 1 mm latis, intimis rhombeo-ovatis, 8-9 mm longis, 3.0-3.5 mm latis. floribus radii 10-13, pappo saepe articulato. Shrub to 4 m; branchlets striate, resinous. Leaves lanceolate, to lance-elliptic, (8-)10-l2(-14) cm long, 1.5-2.5(-3.0) cm wide, entire, acute or rarely obtuse, broadly acuminate, the margins strigillose; petioles 2-5 mm long. Inflorescence 4-8-flowered, cymose-paniculate; peduncles (l-)3-5(-7) cm long. Capitula 1.0- 1,5 cm wide (excluding the ray florets), 1.5-2,0 cm high; involucre graduated, 3-seriate; outer phyllaries linear-lanceolate, (5-)7-9 mm long, ca. 1 mm wide, 198IJ Dl[±ON—FL()LHf-:\SIA 107 Fk.ure 2. Flourens'ui polycephala Dillon, x-ki. [After Marin 231 (F).| acute, glandular, keeled, the inner phyllaries rhombic, 8-9 mm long, 3.0-3.5 mm wide, attenuate, laterally chartaceous, glandular, keeled; paleas ca. 9 mm long, ca. 3 mm wide, acute to obtuse, erose, glandular; ray florets 10-13, the ligules oblong, ca. 3 cm long, 5-6 mm wide, the tube ca. 5 mm long; disc florets 20-30, ca. 7 mm long, cylindric-campanulate. Achenes obconical, 8-9 mm long, ca. 3 mm wide, depressed ovate in cross-section, villous; pappus of 2 awns, ca. 3 mm long, readily disarticulating, squamellae absent. Mari isotype). Flourcnsia polycephala differs from F. heterolepis Blake in having more nu- 108 ANNAl.S OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 merous capitula and cymose inflorescences. It differs from its nearest geographic neighbor, F, peruviana, in having longer phyllaries and more ray florets. This species is known from dry, rocky slopes in the quehradas associated with the Rio Apurimac and Rio Urubamba in southeastern Peru (2,700- 3,200 m) (Fig. 1). Flowering Jan. -Apr. Additional Specimens Examined: Peru, apurimac: Grau, Orepeza Valley, Vargas 9784 (F, UC). Cuzco: EUenherg WOO (US). Huasao, Herrera 3098 (US). Urubamba, lower end of Quebrada Pu- mahuanco, ca. 2-4 km NW of Urubamba, litis et al. 854 (US). Urubamba, Lamalha 53 (LPS); Hda. Urco, Vilconota Valley, Vargas 638 (MO). Rumichaca, Vargas 9597 (LIL). Chicon Canyon, Vargas 11053 (UC). Flourensia peruviana Dillon, sp. nov. — Fig. 3. Species haec ab F. polycephala Dillon differl involucro minore, phyllariis reduclis aliler isdem, foliis magniludine formaque similiter F. angustifolia (DC.) Blake, cognata approximatissiam, sed differt folia Integra. Pappus acheniorum intcrdum artisis disarticulalis duo 3 mm longis confectus. Shrub to 2 m; branchlets striate, resinous. Leaves lanceolate to oblong-lan- ceolate, (5-)7-10(-12) cm long, (l.O-)l. 5-2,0 (-3.0) cm wide, entire, acute to obtuse, broadly acuminate, the margins strigillose; petioles (2-)4-8(-10) mm long. Inflorescence 4-8-flowered, cymose-paniculate; peduncles 0.5-5.0 cm long. Ca- pitula 7-10 mm wide (excluding ray florets), ca. 10 mm high; involucre graduated, 2-3-scriate; outer phyllaries lanceolate, ca. 2 mm long, ca. 1 mm wide, acute, glandular, keeled, the inner phyllaries elliptic-obovate, ca. 5 mm long, ca. 1.5 mm wide, laterally chartaceous, glandular, keeled, acute; paleas ca. 9 mm long, rounded apically, erose, glandular; ray florets 8-10, the ligules oblong to oval, 10-18 mm long, ca. 5 mm wide, the tube ca. 5 mm long; disc florets 20-40, ca. 7 mm long, cylindric-campanulate. Achenes obconical, ca. 10 mm long, 2-3 mm wide, depressed ovate in cross-section, villous; pappus of 2 slender awns, ca. 3 mm long, readily disarticulating, squamellae absent. Type: Peru, huancavelica: Checcyancu a 4 km E of Conaica, 3000-3500 m, 14 Mar. 1971, Tovar 193 (US, holotype; LPS, USM, F, isotypes). Flourensia peruviana differs from F. polycephala in possessing a much small- er involucre with the phyllaries reduced in size, but essentially the same shape. Its leaves are comparable in size and shape to F, angustifolia (DC.) Blake, its nearest geographic neighbor, but differ in having strictly entire margins. This species is known from southcentral Peru on dry, rocky slopes in the quebradas associated with the Rio Mantaro (1,700-3,500 m) (Fig. I). Flowering Mar. -Apr. Additional specimens examined: Peru, ayacucho: La Mejorada to Ayacucho, KM 15, Ochoa 574 (GH). Huamanga, Chaquihuaycco, arriba de Ayacucho, Tovar 5491 (USM), Chanchara. Rio Cachi, Tovar 5589 (USM). Alrededores de Ayacucho. Tovar 5709 (USM). huancavelica: Taya- chaja, between Izuchaca and Mariscal, Tovar 1378 (LPS). Another new species of Flourensia from Argentina may also conveniently be described here. Flourensia blakeana Dillon, sp. nov. — Fig. 4. Frutox 0.4-1.0 m alto, ramoso, cortice cana ad nigra, ramulis nigris. Folia (l.5-)2.0-3.5(-4.7) cm longa, (0.3-)0.4-0.8(-l.l) cm lata, anguste oblongo-elliptica, acuta, acuminata, Integra, marginibus slrigillosis; petiolis 1-3 mm longis. Capitula 1-4, terminales axillaresque, cymosa (5-)7-IO(-14) mm 1981J DILLON— /■7,r>t//^/:.V.VM 109 FiGURK 3. Flourensia peruviana Dillon, x>4. [After Tovar 193 (F).] 110 ANNALS OFTHH MISSOURI BOTANICAL GARDLN [Vol . 68 Figure 4. Flourensia hlakcana Dillon, x%. [After Dillon & Rodriguez 560 (F).] lata. 6-11 mm alta, pedunculis 1^ cm longis, involucre 2-seriato. phyllariis aequalibiis (3-)4-6(-7) mm longis, ca. 1 mm latis, extimis linearo-lanceolatis, intimis anguste rhombeis, attenuatis, herbaceis, nigris, basim strigillosis, paleis 5-6 mm longis, obtusis, nigellis. Flores radii ca. 8, disci ca. 25, corollis ca. 4,5 mm longis, cylindricis-campanulatis. Achenia obconica, ca. 6 mm longa, ca. 2 mm lata, villosa, pappi arislis duo ca. 3.5 mm longis persistentibus confecti, squamellis carentes. Chromosomalum Humerus n = 18. M 1 .0 m tall, the bark gray to black; branchJets black. 5(^ )0.4-0.8(-l.l) cm wide, acute to acuminate, entire, the margins strigillose; petioles 1-3 mm long. Inft 1-4-flowered; peduncles 1-4 cm long. Capitida (5-)7 1981] DlLLOr<—FL()U/J 3th M. patula and M v M. patula in the Fh M. natida and M. ns> Ml also differs in a glabrous often elongate inflorescence, and usually shorter more M M M Mt a spathaceously split calyx, reduced few-flowered inflorescence and smaller bracts and bracteoles. It is possible that collections from intermediate areas would M. nseudouatuhi and M 4 * Mi lower Amazonia. LilLRATURE ClThD Bailey, L. H. & H. E. Moore, Jr. 1949. Palms uncertain anJ new. Gentes Herb. 8: 91-205. Brown, K. S., Jr. 1975. Geographical patterns of evolution in neotropical lepidoptera. Systematics and derivation of known and new Heliconiini (Nymphalidae: Nymphalinae). J. Fntomol. (B) 44: 201-242. Burret, M. 1932. Die Palmcngattungen Martinezia und Aiphancs. Notizbl. Bol. Gart. Berlin-Dah- lem II: 557-577. , 1937. Plantae Duqueanae. Notizbl. Bot. Gart. Berlin-Dahlcm 13: 489^95. . 1940. Palmae. //; F. L. E. Diels. Neue Arten aus Ecuador III. Notizbl. Bot. Gart. Berlin- Dahlem 15: 23-38. DuGAND. A. 1944. Palmas Colombianas 11. Caldasia 2: 443-458. Gentry, A. H. 1974. Notes on Panamanian Apocynaceae. Ann. Missouri Bol. Gard. 61: 891-900. . 1978. Florislic knowledge and needs in Pacific Tropical America. Brittonia 30: 134-153. — . 1979. Extinction and conservation of plant species in tropical America: a phytogeographical perspective. Pp. 110-126, in J. Hedberg (editor), Systematic Botany, Plant Utilization and Bio- sphere Conservation. Almquist & Wiksell, Stockholm. . 1981. Phytogeographic patterns in northwest South America and southern Central America as evidence for a Choco refugium. Chap. 10, in G. T. Prance (editor), The Biological Model of Diversification in the Tropics. Columbia Univ. Press, New York. KoBUSKi. C. E. 1941. Studies in the Theaceae, VIII. A synopsis of the genus Freziera. J. Arnold Arbor. 22: 457^96. Markgraf, F. 1939. Die amerikanischen Tabernaemontanoideen. Notizbl. Bot. Gart. Berlin-Dahlem 14: 151-184. Moore, H. E. 1951. Various new palms. Gentes Herb. 8: 223-238. Rao, A. S. 1956. A revision of Rauvolfia with particular reference to the American species. Ann. Missouri Bot. Gard. 43: 253-354. Woodson, R. E., Jr. 1933. Studies in the Apocynaceae IV. The American genera of Echitoide I. AUomark^rafia; II. Mcsechites: III. Mandevilla; IV. Macrosiphonia. Ann. Missouri Bot. Gard. 20: 605-790. A NEW PERUVIAN STYLOCERAS (BUXACEAE): DISCOVERY OF A PHYTOGEOGRAPHICAL MISSING LINK 1 Alwyn H. Gentry^ and Robin Foster^ Abstract A new species of Sfyloceras (Buxaceae), S. brokawii A. Gentry & R. Foster, is described from lowland Amazonian Peru. This is the first non-Andean species of the genus and provides significant new phytogeographical evidence as to the origins of this remarkable and isolated genus. Styloceras brokawii A. Gentry & R. Foster, sp. nov. — Fig. 1. Frutex dioecius. Folia alterna elliptica, acuta vel breviacuminala, ad basim acuta. Infloresccntia masculina spica axillaris, gracilis, floribus solitariis, distantibus, bracteatis, sessilibus, antheris ca. 14, sessilibus. Flores feminei solitarii, pedicellati, calyce 4-5-lobato, ovario puberulo, stigmatibus 2(-3) magnis. Fructus globosus, stigmatibus persistentibus. Dioecious (rarely monoecious) shrub 2-6 m tall; branchlets angulate when young, becoming subterete, glabrous, striate. Leaves alternate, elliptic, the apex acute to short acuminate, acute to abruptly attenuate at the base. (7-) 10-20 cm long, (3.5-)5-9(-10) cm wide, entire, somewhat 3-veined from above the base, 4-5 secondary nerves on each side, these plane above and prominulous belov/, tertiary veins obscure, glabrous, the petiole 1-3 cm long. Male inflorescence, axillary, pendant, spicate, slender, 2-6 cm long, the flowers rather distant, each flower subtended by a ciliate-tTiargined ovate bract ca. 1 mm long, the flowers sessile, the anthers ca. 14, sessile, bilocular, 1.5-2 mm long, slightly curved. Female flowers solitary, pedicellate, the pedicel 7-9 mm long, glabrous, with several very inconspicuous minute bracts near the base, the calyx 4-5-lobed, the lobes triangular, reflexed, the corolla absent, the ovary subsessile, truncate, 2-parted, minutely puberulous, the apex prolonged into 2 (rarely 3) large divergent stigmas over 1 cm long, the tips recurved, yellow when fresh, the base of the stigmas slightly contracted into the ca. 2 mm long style. Fruit globose, fleshy, yellow, 2-3 cm in diameter, the 2 stigmas persistent as 2 subapical horns 1.8-2 cm long, their bases separated by 1-2 cm, tardily dehiscent to expose gelatinous material surrounding 2-3 dark blue seeds. Type: Peru, madre de dios: Prov. Manu, forest behind Manu settlement, 350 m alt, shrub 3-5 m, male catkins white, dioecious, 4 Aug. 1973, R, Foster, /y. Brokaw & N, Brokaw 2474 (MO, holotype; F, isotype). Additional Collections Examined: Phru. madre de digs: in forest behind Manu settlement, 4 Aug. 1973, Foster et al. 2471 (F, MO). Shintuya. forest I km up small stream from Rio Madre de Dios. 8 Aug. 1974, Foster et al. 3/ IS (F). Rio Alto Madre de Dios, halfway between Shintuya and Manu settlements, forest near chacra of Sr. Carpio, 10-11 Aug. 1974, Foster el al. 3229 (F), 3249 » Supported by the Flora of Peru grant (DEB 75-20325 and DEB 79-05078) from the National Science Foundation. ^ Missouri Botanical Garden, Post Office Box 299, St. Louis, Missouri 63166. ^ University of Chicago, Department of Biology, Chicago. Illinois 60637. Ann. Missouri Bot. Gard. 68: 122-124. 1981. 0026-6493/8 1 /0 1 22-0 1 24/$0.45/0 1981] GKN IKY & FOSThlR—STYUX'KRAS 123 1 Figure I. Styloceras hrokawii A. Gentry & R. Foster. Fruit and female flowers {Gentry et al. 27222). (F). Parque Nacional del Manu, Rio Manu, vicinity of Cocha Cashu Biological Station, 14 Sep. 1976. Foster & Russell 5035 {¥). Rio Palotoa, tributary of Alto Madre de Dios, NW of Shintuya, called Rio Pantiacolla on maps, 500 m, 26 Aug. 1978, Foster c^ Terhori^h 6712 (F). Cocha Cashu Camp, Manu National Park, trails 4, 5, and 9, 380 m. 21 Oct. 1979, Gentry et al. 27070 (AMAZ, F, MO. USM). Across river from Cocha Cashu camp, Manu Park, 380^20 m, 23 Oct. 1979, Gentry et al. 27185 (MO. USM), 27221 (MO, USM), 27222 (AMAZ, F, MO, USM). The known range of S, hrokawii is between 350 and 600 m in the shade of closed seasonally inundated forest and noninundated upland forest. It seems most abundant in intermediate areas of alluvial soil subject to rare inundations. At the type locality this species is abundant with up to 10-30 individuals per hectare. In other collection sites, individuals were at densities of one or less per hectare. Its peak of reproductive activity occurs during the drier months of July, August, and September when the forest is partially deciduous. Most full-size individuals were in reproductive condition at the time of collection, but plants in bud and with ripe fruit could be found at the same time. There appeared to be more male than female plants, and females were usually larger. One collection (Gentry ct iiL 27222) has a female flower at the base of the predominantly male inflorescence. This species is named in honor of Howard P. Brokaw who found the plants from which the first specimens were made and in recognition of the contributions of him and his family to the cause of biological conservation in the neotropics. There are only three accepted species of the remarkable and isolated genus Styloceras which is sometimes recognized as constituting a distinct family (Airy Shaw, 1966). Styloceras hrokawii is most similar to S, laurifolium (Willd.) H.B.K. which ranges from central Peru to the Colombian and Venezuelan Andes. That species differs in a denser male inflorescence, glabrous ovary, and thicker 124 ANNALS OF THE MISSOURI BOTANICAL GARDEN (Vol. 68 leaves with more prominent venation. Styloccras kunthianum Juss. of the upper Pastaza valley in Ecuador, which was erroneously synonymized with S, lauri- folium in the Flora of Peru, differs in being monoecious with much shorter in- florescences. Styloccras columnare Mull.-Arg. of the Sorata area of Bolivia differs most notably in having the styles united into a basal column. Styloccras is phytogeographically, as well as taxonomically, problematical, and 5. hrokani'i provides important new phytogeographical insight. Buxaceae is an ancient, basically Laurasian group, and Styloccras is the only South American genus (one West Indian species of Buxus also reaches the Venezuelan coastal Cordillera). Citing the traditional grouping of Styloccras and African Notobuxus (now usually merged with Buxus, Gentry, 1978) as tribe Stylocereae, Raven & Axelrod (1974) suggest that nevertheless the family already may have been pres- ent in West Gondwanaland in the Cretaceous when more or less direct migration between Africa and South America was still possible. All of the other species of Styloccras'dVQ found at much higher elevations and the restriction of such a pu- tatively old and archaic genus as Styloccras to the geologically young Andes has seemed anomalous. The new species is the first non-Andean Styloccras, sug- gesting a possible reinterpretation of the genus as an old tropical lowland forest one which successfully invaded the upper slopes of the Andes relatively recently. This interpretation is also supported by arrival of Styloccras in the palynological record of the Colombian Cordillera Oriental only about 1.2 million years ago (van der Hammen, 1974), well after the establishment of suitable montane habitat in that region. On the basis of its phytogeography we predict that S. hrokawii will prove the least advanced member of Styloccras, a prediction which would be borne out if contraction of the inflorescence and fusion of the style bases prove to be specialized characters as would be expected. Literature Cited Airy Shaw, H. K. 1966. A Dictionary of the Flowering Plants and Ferns. Ed. 7. University Press, Cambridge. Gentry, A. H. 1978. Buxaceae. In Flora of Panama. Ann. Missouri Bot. Card. 65: 5-8. Hammen, T. van der. 1974. The Pleistocene changes of vegetation and climate in tropical South America. J. Biogeogr, I: 3-26. Raven, P. H. & D. I. Axei rod. 1974. Angiosperm hiogeography and past continental movements. Ann. Missouri Bot. Card. 61: 539-673. FOUR NEW SPECIES OF DIOSCOREA FROM AMAZONIAN PERU 1 By Franklin Ayala F/ Abstract Four new species of Dioscorea from northeastern Peru are described: D. revilhw F. Ayala. D tamshiyacuensis F. Ayala, D. claytonii F. Ayala, and D. schunkci F. Ayala & T. Clayton. During a recent study of the genus Dioscorea L. in northeastern Peru (Ayala, 1979) the following four new species were discovered. Dioscorea revillae F. Ayala, sp. nov. — Fig. 1. Herba sinistrorsum volubilis ca. 1.7 mm vix crassa lineis 10 vel plus proxime nodos praedila; folia 7-10 cm longa, 3.5-4.0 cm lata, alternata, ovata vel ovata-lanceolata, apice acuta, basi acuta, margine integro; venae laterales 5 eis externissimis bifidis; petiolus gracilis basi dilatatus, 0.3-0.5 mm crassus, 4^.8 cm longus; inflorescentia spiciformis axillaris, ramis nuliis, 17-20 cm longa; flores luteo-virides solitarii, 2-3 mm longi, ca. 3.2 mm in diametro, bractcis duabus basi floris utriusque locatis; sepala oblonga, ca. 2.3 mm longa, ca. 0.8 mm lata, apice involuta apiculata, basi acuta; stamina 6, filamentis cuneatis vel crassis; antherae subglobosae, longitudinaliter dehiscentes. Herbaceous vine; stem sinistrorse, 1,7 mm thick, somewhat tetragonal, with 10 or more transverse lines close to the nodes; leaves 7-10 cm long, 3.5^.0 cm wide, alternate, ovate to ovate-lanceolate, acute at the base and apex, the margin entire, the nerves 5, the most external bifid; petiole slender, dilated at the base, 0.3-0.5 mm thick, 4^.8 cm long; inflorescence a spike, axillary, unbranched, 17-20 cm long; flowers yellowish green, solitary, 2.3 mm long, 3.2 mm in diam- eter, with 2 bracts at the base of each flower; sepals oblong, 2.3 mm long, 0.8 mm wide, the apex involute, apiculate, acute at the base; stamens 6, longitudinally dehiscent, the anthers subglobose, the filament cuneate and thick. Type: Peru, loreto: Provincia Maynas, Distrito Iquitos, boca del Rio Na- nay, 1 Nov. 1976, Juan Revilla 1730 (AMAZ, holotype; F, MO, USM, isotypes). This species is related to Dioscorea paraguayensis R. Knuth which was placed by Knuth (1924: 72) in subgenus Helmia (Knuth) Benth., section Spliaer- antha Uline. It differs in having 5 nerves rather than 7. The petiole of D. para- guayensis is 2.5 cm long and the rachis of the inflorescence is 7 cm long and branched, while the petiole of the new species is 4.8 cm long with the rachis reaching 20 cm in length and unbranched. Dioscorea tamshiyacuensis F. Ayala, sp. nov. — Fig. 2, Herba volubilis caule gracili, 0.2-0.4 mm crasso, inermi, glabro, sinistrorsum volubili lenticellis violaceis ornato; folia integra glabra, elliptico-oblonga ad oblongo-lanceolata, acuminata, basi rotun- ^ I thank J. Dwyer, Alwyn H. Gentry, and Jose Maria Arroyo for assistance with the Latin de- scriptions. This study was supported in part by a LASPAU graduate Fellowship to study at Washington University, St. Louis and the Missouri Botanical Garden. ^ Director, Herbarium Amazonense (AMAZ), Universidad Nacional de la Amazonia Peruana, Aptdo. 421 Iquitos, Peru. Ann. Missouri Bot. Gard. 68: 125-131. 1981. ()()26-6493/8 1 /0 1 25-0 131/0. 85/0 126 ANNAI.S OF THE MISSOURI BOTANICAL GARDHN [Vol . f>S c Figure I. Dioscorea revilluc F. Ayala. at base of petal (xI4).— D. Petal (xI4).— E. (AMAZ).l A. Habit (x^/,o).—B. Flower (x 14).— C. Stamen adnate Stamen (xl4).— F.-G. Torus (xl4). [After ReviUa 1730 I981J AY M.A—DIOSCORFJA 127 Figure 2. Dioscorea tamshiyacuensis F. Ayala. — A. Habit (x'^/n). — B. Flower (xSVs). — C. Petal (xl2).— D. Stamen (xl5Vr>). |After Ayala 564 (AMAZ). data, 9-1 1 cm longa. 4-5 cm lata, nervis 3, violaceis; peliolus 2-3 cm longus; inflorescentia spicifor- mis, fasciculata, rhachi diminuto-alata 12-20 cm longa: flores sessiles, 0.15-0.20 cm longi, 0.12-0.15 cm lati; sepala viridia ovato-lanceolata, 0.8-2 mm longa, 0,6-1 mm lata; petala violacea, ovata, intror- sum volubilia, 1.0-2.5 mm longa, 1.0-1.2 mm lata; perianthium campanulato-rotundatum; stamina 128 ANNALS OF THE MISSOURI BOTANICAL GARDLN [Vol. 68 Figure 3. Dioscorea cluytonii F. Ayala. — A. Habit [xV-x). — B. Flower (x7'/2). — C. Bract (x4!/2).— D. Stamen (xl5). [After Klu^ 4345 (MO).j 6, antheris lutcis, filamcntis brevibus; flores bracteis duabus praedili. bractea utraque 0.1 mm longa, lanceolata vel ovato-lanceolata; radix lignosa. Herbaceous vine; roots woody; stems slender, 0.3-0.4 mm thick, unarmed, glabrous, sinistrorse, with violet lenticels; leaves entire, glabrous, 9-11 cm long, 4-5 cm wide, elliptic-oblong to oblong-lanceolate, the apex acuminate, the base rounded, the nerves 3, violet; petiole 2-3 cm long; inflorescence a spike, fascic- 19811 AY ALA— niO.SCOREA 129 B C D E F Figure 4. Dioscorea schunkei h. Ayala & T. Clayton. — A. Habit (x-M). — B. Inflorescence (x%).— C.-D. Flower (x7'/2).—E. Bract {x4'/2).-^I-.-G. Stamen (x 13). [M\qv Schunke 3825 {UO).] ulate, the rachis minutely winged, 12-20 cm long; flowers sessile, 1.5-2.0 cm long, 1.2-1.5 cm wide, with 2 bracts at the base of each flower, the bracts 0.1 mm long, lanceolate to ovate-lanceolate; sepals green, ovate-lanceolate, 0.8-2 mm long, 0,6-1 mm wide; petals violet, ovate, introrse, 1-2.5 mm long, 1-1.2 mm wide; perianth campanulate-rotate; stamens 6, the anthers yellowish, the fila- ments short. 130 ANNALS OF THE MISSOURI BOTANICAL GARDEN IVoi . 68 Type: Peru, loreto: Provincia Maynas, Tamshiyacu, 8 Mar. 1974, F. Ayala 564 (MO, holotype; AMAZ, isotype). This species is related to D. apurimacensis R. Knuth which was placed by Knuth (1931: 94) in subgenus Eudioscorea Pax, section Cryptantha Uline. It differs from D. apurimacensis in twining sinistrorsely rather than dextrorsely and in having three nerves rather than five. Dioscorea claytonii F. Ayala, sp. nov. — Fig. 3. Herbadextrorsum volubilis, ca. 2.5-3 mmcrassa,foliisalternatis, 9-1 1 cm remotis, cordiformibus, campylodromis; folia basi late cordata, apice acuminata, cuspidata, 2.5-7.5 cm longa, 2.3-5.5 cm lata, nervis 7-9, penultimis bifidis vel ramosis; petiolus glaber, 2-4 cm longus, 0.4 mm latus; inflo- rescentia axillaris, racemosa, rhachidi 6.5-8.0 cm longa, hirsuta; flores campanulati; perianthium 4 cm longum, 4.2 mm latum, pedicellus 1.2-2.8 cm longus, glaber; stamina fertilia 6, I mm longa, filamentis latis, 0.7 mm longis; anthera capitata, 0.3 mm longa; capsulae apice rotundatae, 2.0 cm longae, 1-1.3 cm latae; semina undique ala membranacea cincta, 0.2-0.5 cm longa, 0.2-0,3 cm lata. Vine; stem dextrorse, 2.5-3 mm thick; leaves alternate, separated by 9-11 cm, cordiform, campylodromus, broadly cordate at the base, the apex acuminate or cuspidate, 2.5-7.5 cm long, 2.3-5.5 cm wide, with 7-9 nerves, the penultimate nerves bifid or branched; petiole glabrous, 2.0^.0 cm long, 0.4 mm wide; inflo- rescence an axillary raceme, the rachis 6.5-8 cm long, hirsute; flowers campan- ulate, the perianth 4 mm long, 4.2 mm wide, the pedicel 1 .2-2,8 cm long, glabrous; stamens 6, 1 mm long, the filaments broad, 0.7 mm long, the anthers capitate, 0.3 mm long; capsule apically rounded, 2 cm long, 1-1.3 cm wide, the seeds winged, 0.2-0.5 cm long, 0.2-0.3 cm wide. Type: Peru, san martin: Provincia Mariscal Caceres, Distrito Juan Jui, Alto Rio Huallaga, 400-800 m, forest. May 1936, Klu^ 4345 (MO, holotype; phototype and fragm., AMAZ). This species is related to Dioscorea midtispicata R. Knuth which was de- scribed in subgenus Helmia (Knuth) Benth., section Centrostemon Griseb. (Knuth, 1916). It differs especially in having a long and broad filament rather than a short and thin one. Dioscorea schunkei F. Ayala & T. Clayton, sp. nov. — Fig. 4. Herba sinistrorsum volubilis, ca. 7-8 m alta, caiile glabro, 2 mm lato; folia glabra, alternala, trifoliala, papyracea, segmentum folialum oblongo-elliplica, apice acuta, basi attenuate acuta, 4-6 cm longa. 1.8-2.2 cm lata: petiolus 3 cm longus, I mm latus. Racemus rhachidi 9 mm longa; flores campanulati 4 mm lati, contorti» involuti: pedicellus 1 mm longus, bracteis duabus basi floris utrique locatis; bractea 1.2-2.5 mm longa, acuminata, caudata; stamina fertilia 6, 1.3 cm longa, 0.2 mm crassa, antheris 0.3 mm longis. Vine, sinestrorse, 7-8 m high; stem glabrous, 2 mm wide; leaves glabrous, alternate, trifoliate, papyraceous, the leaflets oblong-elliptic, apically acute, acutely attenuate at the base, 4-6 cm long, 1.8-2.2 cm wide, the petiole 3 cm long, 1 mm wide; inflorescence a raceme, the rachis 9 cm long; flowers campan- ulate, 4 mm wide, contorted, involute, the peduncle 1 mm long with 2 bracts at the base, the bracts 1.2-2.5 mm long, acuminate, caudate; stamens 6, 1.3 mm long, 0.2 mm thick, the anthers 0.3 mm long. 1981] A Y A \.A—DIOSCORfiA 131 Type: Peru, san martin: Provincia Marsical Caceres, Distrito Tocache Nue- vo, en bosque alto camino a Shunte, 26 Feb. 1970, J. Schunke Vigo 3825 (MO, holotype: photocopy and fragm., AMAZ). //c Trifoli than being glabrous and having three stamens. The species was independently recognized as new by the late Temple Clayton of the University of Minnesota, as indicated by inclusion of his name as co- author. Literature Cited Ayala, F. 1979. Dioscoreaceae del Departmenlo de Lorelo. Unpublished doctoral thesis, Univer- sidad de Trujillo, Peru. Knuth, R. 1916. Dioscoreaceae. In T. Herzog, Die von Dr. Th. Herzog auf seiner zweiten Rcise Pflanzen 29: 55-56. . 1924. Dioscoreaceae. In A. Engler, Das Pflanzenreich IV. 43 (Heft 87); 1-387. . 1931. Dioscoreaceae novae. Repert. Spec. Nov. Regni Veg. 29: 92-96. SYSTEMATICS, PHYLOGENY AND EVOLUTION OF DIETES (IRIDACEAE)^ Petlr Goldblatt^ Abstract Dieles is a member of Iridaceae-Iridoideae and is probably the most primitive member of the Old World tribe Irideae. It shares characteristics of both Iris and the African genus Moraea and is most likely close to the ancestral stock that gave rise to these more speciaUzed genera. Dietes compri six species, five African and one remarkable disjunct on Lord Howe Island between Australia and New Zealand, D. rohinsoniana. The South African D, hicolor appears more closely related to D, rohinsoniana than to the other African species, and these two share several characters primitive in the genus. The remaining four African species include the wide-ranging D. iridioiJes, which extends from the southern Cape to Ethiopia, and three more localized eastern southern African species. Dietes is a small genus of Iridaceae, closely allied to the African genus Moraea and to the widespread Northern Hemisphere genus Iris. Six species are currently recognized, five African and one restricted to Lord Lowe Island in the Tasman Sea between Australia and New Zealand (Fig. 1). In spite of the extraordinary disjunction in the distribution, there seems no doubt that Dietes is a natural genus, all the species sharing unique vegetative and floral structures and a basic chromosome number of a* = 10. History The name Dietes was first proposed by R. A. Salisbury in 1812, but as pub- lished it was nomenclaturally invalid, lacking description or reference to a pre- viously published generic description. Salisbury chose the name Dietes to indi- cate what he believed to be its dual affinities to both Iris and Moraea, a point of view fully supported here. Few authors accepted the genus initially, though Sweet (1830, 1839) put forward several combinations, unfortunately never validating the genus with a description. Dietes appears again in the literature in 1846 in an article by Spae concerning the species currently called D. hicolor. Spae used both Moraea hicolor and Dietes hicolor in the title, but apparently regarded Moraea as the correct generic name. Later, in 1852 when Spae again wrote about D. hicolor, he unambiguously placed it in Moraea, Dietes was only fully accepted in 1866 by Klatt, who provided a complete generic description. Klatt acknowledged Salisbury as the source of the name and recognized three species in the genus, D. hicolor, D. catenulata, and D. com- pressa. The last two are, in my opinion, the same species and conspecific with the much earlier D. iridioides. It was also Klatt who realized that the newly described Australasian species Iris rohinsoniana F. Muell. belonged in Dietes (Klatt, 1882). * This research was supported by grant DEB 78-10655 from the U.S. National Science Founda- tion. I thank Mrs. A. A. Mauve-Obermeyer for her helpful suggestions and collaboration in this project and Margo Branch for preparation of the illustrations. ^ B. A. Krukoff, Curator of African Botany, Missouri Botanical Garden, Post Office Box 299, St. Louis, Missouri 63166. Ann. Missouri Bot. Garo. 68: 132-153. 1981. 0026-6493/8 1/0132-OI53/$2. 35/0 1981] GOLDBI.ATT— /)//:r/:\S' 133 FiGURF I. Worldwide range of Dietes, showing the distribution of the five African species in southern and eastern Africa and the single species on Lord Howe Island. Salisbury was not the first to consider Dietes a distinct genus. Medikus in 1790 published Naron which, though confusingly described, is clearly a new genus accommodating only the Linnaean species Moraea iridioides, although under the superfluous name N. orientale. Naron was never accepted and appar- ently on only one occasion is the genus again mentioned, by Moench in 1794 where he proposed the corrected combination TV. indiodeum (L.) Moench. Naron was completely overlooked subsequently, although a perfectly valid earlier syn- onym for Dietes, even typified by the same species under a different name. Dietes did not receive general acceptance during the nineteenth century, al- though Klatt (1882, 1895) continued to recognize the genus. Instead, J. G. Baker's view of Dietes as a subgenus of Moraea prevailed and the genus was so treated in the important works. Flora Capensis (Baker, 1896) and Flora of Tropical Africa (Baker, 1898). In 1928 Brown raised Dietes once more to generic rank, 134 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 and he described two new African species, D. ^randiflora and D. prolongata, the latter now considered merely a minor variant of D. iridioides. From this time, Dietes was consistently recognized in southern Africa for the genus, as when Gerstner in 1943 described the Natal forest species D, hutcheriana. Brown (1928) was also responsible for changing the name of D. iridioides to D. vegeta based on Monica vcgeta L., a species dating from 1762 which he believed to be typified by the same figure that is the type of M. irioides L. (1767). As explained at length (Barnard & Goldblatt, 1975), this was incorrect, Moraea vegeta is in fact the type species of Moraea and is not a Dietes, An important contribution to the understanding of Dietes was made by A. A. Obermeyer who, in a series of four beautifully illustrated articles in Flowering Plants of Africa (Obermeyer, 1967a, 1967b, 1968a, 1968b), reviewed current knowledge of Dietes. She contributed valuable new data on the biology of the genus based on her own observations and those of R. G. Strey, the Natal botanist. Obermeyer also described a new species, D. flavida, bringing the number of species in the genus to six. Dietes is now widely known in many parts of the world as a horticultural subject. Although not the earliest name for the genus, Dietes is correct as it has been conserved against Naron (Goldblatt, 1973; Brummitt, 1978) in the interests of nomenclatural stability. Relationships Mor to which of the two is more closely related, but there is perhaps a consensus that Moraea genera. 1 believe it may be closer to the truth to consider Dietes a primitive genus in the Iridioideae, and ancestral to Moraea and Iris. Dietes appears to me to have a combination of the more unspecialized features of both Iris and Moraea, It is an evergreen herb of areas with a fairly equable climate. It has a thick persistent rhizome, a fan of equitant, tough, long-lived leaves, and a freely branching scape which in Z). robinsoniana and D. hicolor is paniclelike. The flower consists of free, spreading, clawed tepals; typically entirely free stamens (the filaments are joined basally in a form of D. iridioides); and flattened petaloid style branches each with the transverse stigmas and paired crests typical of both Iris and Moraea, Iris differs in several respects: most species are deciduous; the inflorescence is usually reduced rather than freely branched; the perianth is usually united to form a tube; and the tepals are dimorphic with large pendant outer tepals and smaller, usually erect inner tepals. Important features shared with Dietes are: the rhizome, clearly primitive in Iris; the basically isobilateral leaf (specialized in some species and subgenera and square or bifacial); free stamens; and spathe valves free to base. In Moraea the rootstock is always a corm, the leaf fundamentally bifacial (occasionally terete), and the deciduous condition is normal. In the flower, tepals may be subequal, but if so, both whorls are usually strongly reflexed (more often the inner tepals are smaller, sometimes erect, and occasionally lacking); a peri- anth tube is found in two species, but free tepals are basic to the genus; the I9S1J GOLDlil.WJ—DIETES 135 Table 1. Comparison of critical characteristics of Dietes, Iris, and Moraea. Character Evergreen Rhizome Free tepals Free stamens Free spathe valves Isobilateral leaf Subequal tepals Dietes + Taxa Iris (+)- +(-) + + +(-) Moraea + (-) ( + ) + ^ The unspecialized condition of each organ is listed with presence indicated by plus, absence by minus. Exceptions, usually derived, are in parentheses. filaments are always contiguous at the base and with minor exceptions partially united; and the spathe valves of the majority of species are united at least in the lower part. The similarities and differences between Dietes, Moraea, and Iris are sum- marized in Table 1, where it is evident that Iris and Dietes share more unspe- cialized (i.e., primitive) features than do Moraea and Dietes. It seems reasonable, therefore, to consider Dietes ancestral to Iris and to Moraea, with Moraea on balance probably having diverged somewhat farther than Iris from the ancestral type. On this line of reasoning, Dietes emerges as the basal genus of Old World Iridoideae and close to the ancestral stock which gave rise to Iris and its allies in the Northern Hemisphere (Hermodactylis, Belamcanda and /m -segregate genera Iridodictyon, Juno, Xiphium, etc.) and to Ferraria, Moraea, and its cor- miferous allies in Africa {Galaxia, Homeria, Hexaglottis, Gynandriris, etc.). In an earlier article in which 1 proposed this hypothesis (Goldblatt, 1976), I suggested a classification in which these Old World Iridoideae be grouped in a single tribe Irideae, subdivided as follows: Dietes, Iris, Hermodactylis, and Belamcanda be grouped in one subtribe Iridinae; the corm-bearing, bifacial leafed Moraea, Ga- laxia, Homeria, Hexaglottis, and Gynandriris be placed in another, Homeriinae Goldbl.; and Ferraria with an isobilateral leaf and a distinctive type of corm in a third, Ferrariinae Goldbl. It seems to me this treatment reasonably, though not perfectly, reflects what is known to date about the relationships of the genera of Old World Iridoideae. Dietes exhibits some similarities with two New World iridoid genera, Neo- marica and Triniezia. These genera also comprise evergreen, mainly forest species, and the rootstock in Neomarica is a creeping, persistent rhizome. Neo- niarica and Trimezia, as well as several other bulb-bearing New World Iridoideae, have flattened, rather petallike style branches, though usually narrower and small- er than those in the Old World Iridoidea. The New World genera differ consis- tently in one significant feature, the inner tepals, often elaborately folded, are nectiferous. In contrast, Old World species have flat inner tepals which rarely produce nectar, a function here of the outer tepals. It seems logical to group all New World Iridoideae in one tribe, the earliest name for which would seem to be Tigridieae. 136 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Table 2. Chromosome numbers in Dietes. New counts are indicated in bold type. Species D. robinsoniuna D. hicolor (as '' Moraea iridioides var. hicolor^*) D. iridioides (as D. ve^eta (L.) N.E.Br.) (as '^Monica iridioides vdv. Johnsonii^') (as **M. iridioides var. mcleyii") (as D. prolongata N.E.Br.) (as M. iridioides) D, flavida D. ^randijiora D, hutcheriana Diploid Number 20 40 40 40 20 20 20 40 20 20 20 20 20 20 Reference or Collection Data Goldblatl(1979) Goldblatt (1971); Chimphamba (1974) Sharma& Sharma (1%0) Goldblatt 2884 (MO), South AtYica, Cape, Kap R. valley. Goldblatt 3732 (MO), South Africa, Cape, Swellendam distr. Goldblatt (1971) (three localities) Sakai (1952); Sharma & Sharma (1960) Sharma & Sharma (1960) Riley (1962) Banerjee & Sharma (1971) Admiraal s.n. (PRE), South Africa, Natal, Josini Dam. Goldblatt (1971); Chimphamba (1974) Bayliss 7014 (MO), South Africa, Cape. Riebeek East. Goldblatt (1971) Cytology Basic chromosome number in Dietes is x = 10 (Table 2). Four of the five African species are diploid, 2n = 20, while D. bicolor is tetraploid with 2ji = 40 (Goldblatt, 1971; Chimphamba, 1974). Dietes robinsoniana, counted from un- vouchered seed obtained from the Royal Botanic Gardens, Kew, also has In ^ 20, as does a hybrid D. rohinsoniana x D. iridioides raised by M. Boussard, Verdun, France (Goldblatt, 1979). Evidently, one of the named cultivars of Di- etes, D. iridioides ''mcleyii'' is polyploid, In = 40 (Sharma & Sharma, I960). Karyotypes of all species are similar and comprise large metacentrics and submetacentrics and smaller acrocentrics. Size differences between matching pairs are small, and the range from largest to smallest is between 7 /xm and 4 /xm. Chromosome numbers in Dietes are listed in Table 2, which includes previously published data as well as the first report for D, flavida, and several original counts for species already known cytologically. Methods employed in obtaining counts have been described elsewhere (Goldblatt, 1978, 1979). Evolution To summarize from the previous section, Dietes is seen as a primitive member of Iridoideae, and the most primitive genus of the Old World members of this subfamily. The species are all indigenous to equable habitats, either shady forest or forest margins, or along streams and other wet places. In Africa the genus is probably relatively ancient, probably dating back to Paleogene time, when the African climate was generally far more equable than it is at present. Extant species of Dietes are perhaps best viewed as relicts now restricted in distribution to well-watered habitats. 1981] GOLDBLATJ—DIETES 137 The single species on Lord Howe Island, which seems to be the most primitive in the genus, is clearly a relict, and is probably more like ancestral Dietes stock than is any living African species. Thus it probably has considerable antiquity on Lord Howe Island, a continental fragment that once had direct overland connec- tions with Australia. It might be a relict of a once more widespread Australian group that has become extinct on the mainland. It is one of the puzzles of plant geography how Dietes reached Australasia, where it is the only member of the tribe Irideae. The most reasonable explanation seems to me long distance dis- persal from Africa, probably well before the Pleistocene. In Paleogene time, when it seems likely that Dietes existed, Australia was well separated from Africa, but India occupied an intermediate position (Raven & Axelrod, 1974) and perhaps afforded a way-station for Dietes. The evolution of Iris and its allies on the one hand and Moraea and its relatives on the other probably began in the mid-Oligocene as world climates began to deteriorate and habitats increased for strongly seasonal and deciduous forms like Iris and Moraea, The closure of the Tethys Sea in mid-Miocene time brought Africa-Arabia into contact with Eurasia, which would have greatly fa- cilitated plant migration from Africa northward (Raven & Axelrod, 1978). It seems likely that the ancestors of Iris moved into Eurasia at about this time and began to spread and radiate from here throughout the Northern Hemisphere. By con- trast, Moraea and its allies remained in Africa, sometimes growing near areas inhabited by Dietes, but always in drier habitats into which their dry-season dormancy allowed extensive radiation. Evolution and Subgeneric Relationships Within Dietes there seems to be one fundamental difference between the inflorescences of Z). rohinsoniana and D. hicolor compared with those of the remaining species. The inflorescence of D. rohinsoniana with its much ramified paniclelike structure and large green bracts seems most likely to be closest to the ancestral and thus primitive type. Dietes bicolor is similar in inflorescence ar- rangement, but the plants and inflorescences are much smaller, the branching pattern more irregular, and the bracts smaller, and often dry. Occasionally branching may be suppressed in this species so that one or two upper nodes bear only a bract without an axillary branch. The inflorescence of the other four species can in no way be called a panicle; the stem branches very irregularly, but bears several sheathing bracts along its upper length from the axils of which fertile branches, or stolons, may later be initiated. Occasionally stems may even be entirely unbranched, but sheathing bracts are always present indicating sites of suppressed, but potential, branching. It seems reasonable to consider this second type of inflorescence specialized by reduction from the panicle type in D. rohinsoniana and Z). bicolor, A second character that unites D. rohinsoniana and D. bicolor is the capsule which is similar in both, being globose in shape with a truncated, flat apex. The capsule of D. hicolor is much smaller, but in other respects very like that of D. rohinsoniana. The capsules of the other species vary to some extent, but are generally elongated and tend to be oblong in shape. They seem more similar to 138 ANNAI.S OF THE MISSOURI BOTANICAL GARDEN [Voi . 68 one another, despite differences in dehiscence, than to those of D. whinsonkum or D. bicolor. It seems unreasonable to propose subgeneric divisions in so small a genus as Dietc's, but it is important, I think, to emphasize the similarities between D. robinsoniana and D. bicolor relative to the other four species and to indicate their apparent primitive position in the genus. I propose an informal grouping, Panicidatae, for these two species and a second, Pauciramosae, for the remain- ing. Among the Pauciramosae, a general morphological similarity in all vegetative and floral features suggests that the species of the group are closely related. They are probably of much more recent origin. The large flower of D. grandiflora, extending to stamens and style branches, sets this species somewhat apart, though the similarities of this streamside and open-woodland species to the forest- dwelling D. iridioides are many, and it is sometimes difficult to tell them apart from dried material. Dietes butchcriana, although distinct, appears to be a lo- calized derivative of the less specialized D. iridioides, adapted to the mist forests of Natal. The localized and disjunct D.flavida, very difficult to distinguish from D. iridioides, appears also to have been derived from the more widespread D. iridioides, and is adapted to forest margin situations. Morphology ROOTSTOCK Dietes has a thick, tough, fibrous to woody creeping rhizome which persists for several years. It bears a fan of equitant leaves at its apex. Side branches are produced from lateral buds which grow eventually to form new plants, initially grouped in a clump with the original plant. LEAVES The leaves of Dietes are, like most Iridaceae, isobilateral, and linear to en- siform, and they are arranged in a distichous fan. The leaves are thick, leathery, and fairly long-lived so that plants are evergreen. Except in D. bicolor, there is no discrete central vein, although several large veins run together in the lower part of the leaf. In D. bicolor two or more large median veins generally run very close together, simulating in appearance a midvein. Leaflike structures without a free lamina are considered bracts and are discussed in the following section. SCAPE The flowering stem, or scape, simply referred to in the text as the stem, is erect, fairly thick, and usually as high or higher than the leaves. The stem bears leaves on the lower nodes which decrease in size upwards. Upper nodes bear bractlike leaves that are entirely sheathing and lack a lamina. They closely re- semble the spathe valves of the inflorescence, except in D. bicolor in which the stem bracts are smaller and often dry. The stem is usually branched in the upper part, either forming a distinct paniclelike inflorescence as in D. robinsoniana and D. bicolor (Fig. 4) or forming 1981] GOLDBLATT— /)/Z:TO 139 an Irregular, rather lax type of inflorescence (Fig. 5), In D. iridioides stems are sometimes unbranched, but typically only in young individuals. At all the branch points of the stem, the subtending stem bracts are paired, a feature apparently not usual in Iridaceae, but known also in Pilansia (Ixioideae) and Bohartia (Sis- yrinchioideae) (Lewis, 1954). INFLORESCENCE The ultimate inflorescences of Dietes are the terminal, so-called rhipidia typical of Iridaceae. Two opposed bracts, called spathes, enclose a several-flow- ered inflorescence in which individual flowers are borne one by one several days or weeks apart. The flowers are raised on a stiff pedicel which is characteristically covered with a light brown pubescence on one side. This may be difficult to see in older flowers or in fruiting plants, but is present in all species. The spathes are herbaceous but stiff in texture, with the inner longer than the outer. The apices are obtuse to emarginate and usually turn brown in mature inflorescences. Indi- vidual rhipidia may be solitary, grouped in a lax arrangement, or together form a panicle as detailed in the section dealing with the stem. FLOWERS The flowers are generally /m-like, large, pale colored, and usually have con- spicuous nectar guides at the base of the limbs of the outer tepals. The tepals are free, unguiculate, and differentiated into a larger outer and smaller inner whorl. The claws of the tepals are ascending, and the limbs spread horizontally. The claws of the outer tepals have a basal nectary and a median line of pubescence (a beard), and they are usually spotted or striped yellow to orange. The inner tepals are seldom marked, but there are conspicuous brown marks on the claws of the inner tepals of D. grandiflora (Fig. 8). ANDROECIUM The three filaments are free except in forms of D. iridioides and are held apart from one another. They are slender with a slightly broadened base. The anthers, held against the style branches, are extrorse, shortly tailed, and have a small apical appendage. In the southern Cape forms of D. iridioides the filaments are joined in the lower half and have a bulbous base. In plants occurring to the east the broadened bases of the filaments are contiguous but free. GYNOLCIUM The ovary is large, green, and typically exserted from the spathes, though partly included in D. rohinsoniana and sometimes in D, bicolor. The style, itself slender and short, divides to form three large flattened petaloid branches opposite the outer tepals. Each style branch is inclined and lies against a tepal claw. The stigma is a transverse, single or bilobed membranous structure located near the top of the style branch. The style branch bifurcates above the stigma to form a pair of petaloid appendages called crests. 140 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 A B C D E F ^^6 Figure 2. Ripe capsules of Dietes, all approximately life size, dehisced). — B. D. hicolor. —A. D. robinsoniuna (partially C. D. irkHoides (dehiscent and indehiscent forms).— D. D. Havida.—E. D. hutcherianu. — F. D. ^nindiftora. FRUITS The fruit is typically a loculicidally dehiscent capsule in Dietes. The degree of dehiscence varies, and fruits of D. butcher'uina are indehiscent, while those of D. flavida split irregularly according to Obermeyer (1968b). The capsules are tough, thick-walled structures and are distinctive in each species (Fig. 2). In D. rohinsoniana and D. hicolor the capsules are erect, globose, with a flat truncated apex and partially dehiscent from the apex to about the middle. In D. grandiftora and D. iridioides the capsules are oblong and ridged, and in the latter usually conspicuously rostrate. According to Obermeyer, the capsule of D. gnmdiflora is fully dehiscent, but in D. iridioides the capsules lie on the ground and decay but do not split. My own experience is contrary; I have observed D. iridioides in cultivation and in the field and have noted mature capsules splitting at least to the midline (Fig. 2B). Collections from the southern Cape also have partly de- 1981] GOLDBLATT— /)//:TO 141 hiscent capsules. Most collections from elsewhere in Africa have capsules closed or split only near the apex, and it seems that there is some degree of variation in this character in D. iridioides. The capsules of D. flavida and D. hutcheriana are distinctive in their smooth surface, somewhat globose rather than oblong shape, and in being pendulous (Figs. 2C-D), SEEDS Seeds of all species of Dictes are large, somewhat irregular in shape, and distinctly flattened. Systematics Dietes Salisb. ex Klatt, Linnaea 34: 583. 1866, nom. cons. vs. Naron Medikus. TYPE species: D. compressa (L.f.) Klatt, nom. illeg. superf. pro Moraca iridioides L. = Dietes iridioides (L.) Klatt. Naron Medikus, Hist. & Comment. Acad. Elect. Sci. Theod.-Palat. 6: 419. 1790, nom. rej. vs. Dietes Salisb. ex Klatt. 1866. type species: N, orientale Medikus, nom. illeg. superf. pro Monica iridioides L. = Dietes iridioides (L.) Klatt. Dietes Salisb., Trans. Hort. Soc. London 1: 307. 1812, nom. nudum. Moraea Miller sensu Linnaeus, Syst. Nat., ed. 12, 2: 78. 1767; sensu Miller, Gard. Diet., ed. 8. 1768 et sensu auct., pro parte. Iris series Dietes Salisb. ex Baker, J. Linn. Soc. Bol. 16: 147. 1878. type species: /. compressa L.f. Moraea subgen. Dictes Salisb. ex Baker, Handb. Irid. 48. 1892; Fl. Cap. 6: II. 1896. type species: M. iridioides L. Plants medium to large, perennial, evergreen, herbs. Rootstock a thick, fi- brous creeping rhizome, persisting for several years. Leaves several, distichous at the apex of the rhizome, tough, leathery, equitant, linear to ensiform. Stems usually erect, bearing leaves at the lower nodes and sheathing spathelike bracts at the upper nodes; branching irregularly in the upper half or forming a distinct many-branched panicle; stem bracts paired at the branch points. Inflorescences enclosed in paired, opposed sheathing, bractlike spathes, the outer spathe small- er, the margins free to the base; flowers several per spathe, borne one at a time on stiff pedicels which are characteristically pubescent on the outer surface. Flowers large, pale colored with nectar guides at the base of the limb of the outer tepals, the claw of the outer tepal bearded to papillate in midline; tepals free to the base, the outer whorl larger, both whorls of tepals with an ascending claw and outspread limb. Filaments either free, filiform with a slightly expanded base or broad based and contiguous or united in the lower part; anthers linear, held against the style branches. Ovary green, terete, included or exserted from the spathes; .s7>7<:\v short, dividing to form 3 branches, style branches large, flattened and petallike, bearing transverse stigma lobes on the abaxial face above which the branch bifurcates, forming paired crests. Fruit a many-seeded capsule, either indehiscent, or partly to entirely splitting along locule septa; seeds large, some- what irregular in outline, but depressed. Basic chromosome number: x = 10. Number of species: 6. Distribution: forests, forest margins, and streamsides, east, central and southeast tropical Africa, coastal southern Africa, and Lord Howe Island, Aus- tralasia; Fig. 1. 142 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol . 68 A^J S Figure 3. Dietes rohinsoniana. Flower xl; stamen and style branches x2. 1. r. Artificial Key to Dietes Leaves 30-50 mm or more at the widest point. 2. Stem forming a regular branching panicle in the upper part; outer tepals ca. 45 mm long; stamens ca. 15 mm long _.____. 1. D. rohinsoniana 2'. Stem forming an irregularly branched inflorescence; outer tepals 30-32 mm long; sta- mens ca. 10 mm long 5. £). hutcheriana Leaves less than 25 mm at the widest point. 3. Stem bracts 9-20 mm long, often brown, obviously paired ___ 2. D. hicolor y. Stem bracts 25-50 mm long, usually green, not obviously paired and the inner much smaller, or evidently lacking. 4. Outer tepals 45-60 mm long; style branches excluding crests 12-20 mm long; inner tepals with brown markings on the claw __ 6. D. grandijlora 4'. Outer tepals 24^0 mm long; style branches 7-9 mm long; inner tepals not marked. 5. Flowers white with a yellow nectar guide and blue to blue-flushed style branch- es; capsule erect, rough walled, usually furrowed and with a conspiduous beak (filaments sometimes united in the lower halO 3. D, iridioides 5'. Flowers pale yellow with an orange brown nectar guide and yellow (occasion- ally blue-flushed) style branches; capsule pendant, smooth walled and without a beak (filaments always free) 4. D.ftavida 1981] GOL D B [. ATT— PftTES 143 Group PANICULATAE 1. Dietes robinsoniana (F. Muell.) Klatt, Erganz. (Abh. Naturf. Ges. Halle 15: 374. 1882)40. 1882.— Fig. 3. Iris robinsoniana F. Muell., Fragment. Phytogr. Austral. 7: 153, 1871. type: Lord Howe Island, Moore s.n. (K, holotype). Moraea robinsoniana (F. Muell.) Benth. & Muell., Fl. Austral. 6: 409. 1873. Plants 1-1.5 m high. Leaves about as long as the inflorescences, linear-ensi- form, the largest to 5 cm at the widest point. Stem erect, elliptical in section, to 15 mm thick, bearing small leaves at the lower nodes and spathelike bracts at the upper nodes; inflorescence much branched, forming the upper fifth of the stem, the nodes of the inflorescence with soft-textured, paired bracts, ranging in length from 5 cm at the lower to 2 cm at the upper nodes, the inner bract small, the outer inflated. Spathes 35^0 mm long, obtuse to emarginate, the outer V3-V2 the inner. Flowers white with orange nectar guides on the outer tepals, opening ca. 10 A.M., fading in late afternoon; outer tepals 45 mm long, the claw ca. 15 mm long, the limb horizontal, ca. 30 mm wide; inner tepals slightly smaller. Filaments to 10 mm long; anthers 5-6 mm long. Ovary ca. 8 mm long, included in the spathes; style ca. 4 mm long, the branches ca. 10 mm long, ca. 8 mm wide; crests to 5 mm long, obtuse. Capsule globose, truncate at the apex, to 30 mm long, ca. 25 mm wide, dehiscent only in the upper half. Chromosome number: 2/2 = 20. Flowering time: spring and summer. Distribution: forest and forest margins, Lord Howe Island, between Australia and New Zealand; Fig, 1, I have noted in the earlier sections of this treatment that Dietes robinsoniana, the only non- African member of the genus, is correctly placed in Dietes, It seems to have, in fact, a combination of more unspecialized characteristics than any other species and is thus regarded as the most primitive in the genus. However, it is not altogether isolated and appears to be more closely related to the eastern South African D. bicolor than the latter is to the remaining African species. 1. Dietes bicolor (Steud.) Sweet ex Klatt, Linnaea 34: 584. 1863.— Fig. 4. Moraea bicolor Steud., Nom. Bot., ed. 2. 2: 159. 1841. type: South Africa, Cape, exact locality not known, illustr. in Bot. Reg. tab. 1404. Iris bicolor Lindl., Bot, Reg. tab. 1404. 1831. horn, illcg. non Miller, Card. Diet., ed. 8. Iris no. 13. 1768. Dietes bicolor Sweet, Hort. Brit., ed. 3, 66L 1839, nom. inval. {irietes nom. nudum). Dietes bicolor Spae, Ann. Gand. 2: tab. 70. 1846, nom. inval. (not accepted by author). Plants 80-120 cm high. Leaves 50-100 cm long, linear, pale green, with a distinct, usually double central vein, 6-12 mm wide. Stem erect, bearing short leaves on the lower nodes and short paired opposite bractlike structures on the upper nodes; stem bracts herbaceous or dry and brown, 9-20 mm long, acute, the margins free to the base. Spathes herbaceous, 34^5 mm long, the outer Vi- V3 the inner, the margins free to the base, the apices obtuse to emarginate. Flow- ers yellow, usually with a dark brown nectar guide on the outer tepals, the guide sometimes lacking and the tepal claw dotted; outer tepals 35 mm long, the claw ca. 12 mm long, bearded and speckled orange, the limb horizontal, to 23 mm 144 ANNALS OF THE MISSOURI BOTANICAL GARDEN (Vol. 6R /, / ^h A \ { i Si /. 's) // ^ (I A^S. FiGURF 4. Morphology and distribution of Dietes bicolor. Flowering branch xO.5; flower and outer tepal xl; ovary, stamen and style branches xl.5. wide; inner tepals to 33 mm long, ca. 18 mm wide, horizontal. Filaments 6 mm long; anthers 4-8 mm long. Ovary ca. 10 mm long; slyle ca. 2 mm long, the branches 8-10 mm long, to 9 mm wide; crests ca. 5 mm long. Capsule to 25 mm long, globose-truncate, dehiscing only in the upper half. Chromosome number In 40. 1981] GOLDBLATT— D/£:re.V 145 Flowering time: spring-summer (Aug. -Feb.). Distribution: along streams and vleis, the eastern Cape between Grahamstown and East London; Fig. 4, Dietes bicolor, the only polyploid species of the genus, is seen as a relict, now confined to moist situations in a limited area of the eastern Cape, South Africa. It is taxonomically isolated and is almost certainly more closely related to the Lord Howe Island species, D. rohinsoniana, than to any African species. There has been some nomenclatural confusion over the authorship of the combination Dietes hicolor and over the original authority of the basionym. Lind- ley described Iris hicolor in 1831, but this name is illegitimate, being a later homonym. The epithet, used by Steudel in Moraea as A/, bicolor, is considered valid only from this date (1841) when M. bicolor is considered a new name. Sweet's combination D. bicolor (1839) is invalid because Dietes was not at this time a vaUdly described genus. Spae also used the name D. bicolor in 1846, but clearly did not accept this combination and instead appears to have regarded the name Moraea bicolor as correct for the species. Dietes bicolor is consequently only to be accepted from 1863 when Klatt provided a valid description for Dietes. and then made the combination accepted here. Dietes bicolor is a valuable ornamental that is widely cultivated today. Plants may be very free flowering, and although each flower lasts only one day, plants usually produce flowers almost every day for months. Under unsuitable cultural conditions, D. bicolor may fail to bloom for years, while apparently healthy and producing foliage only. 3. Dietes iridioides (L.) Sweet ex Klatt, Th. Durand & Schinz, Consp. Fl. Afr. 5: 156. 1895.— Fig. 5. Moraea iridioides L., Mant. PI. 28. 1767. type: South Africa, locality not known, cult. Chelsea Physic Card., illustr. Miller, Fig. PI. tab, 238, fig. I (lectotype). Naron iridioideum (L.) Moench, Meth. PI. 627. 1794, Dieres iridioides (L.) Sweet, Hort. Brit., ed. 2, 497. 1830, nom. inval. (Dietes nom. nudum). Naron orientate Medikus, Hist. & Comment. Acad. Elect. Sci. Theod.-Palat. 6: 419. 1790, nom. illeg. superf. pro Moraea iridioides L. Ferraria hlanda Salisb., Prodr. 42, 1896, nom. illeg. supeif. pro M. iridioides L. Dietes iridifolia Salisb., Trans. Hort. Soc. London 1: 307. 1812, nom. inval. (Dietes nom. nudum) et superf. pro Moraea iridioides L. Iris moraeoides Ker, Bot. Mag. sub tab. 1407. 1811, nom. nov. pro Moraea iridioides L. /. compressa L.f., Suppl. PI. 98. 1781; Thunb., Diss. Irid. no. 12. 1782. type: South Africa, Cape, near Zeekorivier, Thunberg s.n. (Herb. Thunherg 1117 UPS, lectotype). Dietes compressa (L.f.) Klatt, Linnaea 34: 584. 1863. Moraea catenulata Lindl., Bot. Reg. tab. 1074, 1827. type: Mauritius, R. Barclay, illustr. Bot. Reg. tab. 1074 (lectotype). Dietes catenulata (Lindl.) Sweet ex Klalt, Linnaea 34: 585. 1863. D. catenulata (Lindl.) Sweet, Hort. Brit., ed. 2, 497. 1830, nom. inval. (Dietes nom. nudum). Iris crassifolia Lodd., Bot. Cab. tab. 1861. 1832, nom. nudum. /. crassifolia G. Don, Hort. Brit., ed. 3, 661. 1839, nom. nudum. Moraea vegeta L. sensu Miller, Card. Diet., ed. 8. 1768. Dietes vegeta (L.) N. E. Br. sensu N. E. Br., J. Linn. Soc. Bot. 48: 36. 1928. -A/, vegeta L. Moraea iridioides var. prolongata Baker, Fl. Cap. 6: 26. 1896. type; not cited but probably Natal, Inanda, Wood 1341 (K). Dietes prolongata (Baker) N. E. Br.. J. Linn. Soc. Bot. 48: 37. 1928. Dietes prolongata var. galpinii N. E. Br., J. Linn. Soc. Bot. 48: 37. 1928. type: South Africa, Transvaal, near Barberton, Galpin 1206 (K, lectotype). 146 ANNA[.S OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Figure 5. Morphology and distribution of Dietes iridioides. Fruiting branch xO.5; flower xl; ovary, stamen and style branches xl.5. 1981] GOi.Dm.A] V—DIETES 147 Plants (15-)30-60 cm high. Leaves 25^0(-60) mm long, 6-15(-25) mm wide, linear-ensiform. Stem bearing short leaves below, and sheathing, bractlike leaves above, irregularly branching; stem bracts 25-30 mm long, often dry and brownish; old inflorescences often producing long stolons which bear a fan of leaves distally that eventually root. Spathes 35-50(-55) mm long, the outer ca. Vi the inner, obtuse-emarginate at the apex. Flowers white with yellow nectar guides on the outer tepals, the claws of the outer and often the inner tepals orange dotted, the style branches blue or white, flushed with blue; outer tepals 24-35 mm long, the claw ca. 16 mm long, heavily ciliate in midline, papillate, the limb spreading to recurved, 12-16 mm wide; inner tepals 24-28 mm long, 9-12 mm wide, spreading- recurved. Filaments 5-9 mm long, free or united or contiguous in the lower 3 mm; anthers 3-6 mm long. Ovary 8-15 mm long, lightly ridged; style 2-3 mm long, the branches 7-9 mm long, 4-6 mm wide; crests ca. 5 mm long. Capsule ovoid-cylindric, usually rostrate, 20-30 mm long, 14 mm in diameter. Chromo- some number: 2n = 20. Flowering time: sporadic during spring and summer, blooming earliest in areas of winter rainfall. Distribution: evergreen forests, from the southern Cape near Riviersonderend throughout eastern southern Africa and northwards through Rhodesia, Malawi, Zambia, eastern Zaire, Tanzania, Uganda, to Kenya; Fig. 5. Dietes iridioides is the most widespread and common species of the genus, extending almost from the southern tip of Africa to Kenya. It is easy to distin- guish from related species by its relatively small white flower with violet style branches, and its rostrate, cylindrical capsule, which is frequently indehiscent. Following Obermeyer (1968a), I include as synonyms D. prolongata, a species recognized by N. E. Brown, as well as its variety, var. galpinii. The name D. pro- longata was given to stolon-producing plants of D. vegeta, said to have unmarked outer petals. Stolon production is common in D. iridioides and may occur on any plant with old inflorescences. Dietes iridioides was for several years known as D, vegeta (L.) N. E. Brown, following Brown's (1928) erroneous conclusion that Linnaeus' s Moraea vegeta (1762) was to be typified by the illustration published in Miller's Figures of Plants {tab. 238, fig. /), which is the type of Moraea iridioides (and thus of D. iridioides (1767). The reasons for Brown's error and the restoration of the name D, iri- dioides for the species have been described at length elsewhere (Barnard & Gold- blatt, 1975). Moraea vegeta is the type species of Moraea and is currently re- garded as the correct name for the Cape species also sometimes known as M. tristis (L.f.) Ker (Goldblatt, 1976). There is an interesting pattern of variation in D, iridioides. Plants in the southern part of its range, in the southern Cape as far east as the Humansdorp district, have filaments united in the lower half. The filament column is distinc- tively swollen and bulbous towards the base. North and east of these populations, in the eastern Cape, essentially a summer rainfall area, plants have filaments with a similar bulbous base, but the individual filaments are free, but contiguous. I have seen few plants from the rest of the range, but other reports indicate that the filaments are narrower, with only slightly expanded bases, and they are en- 148 ANNAI,S OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Figure 6. Dietes flavida. Flower and fruits xl; ovary, stamen and style branches xl.5 tirely free. The pattern is apparently continuous. There seems no need to express this situation in a formal taxonomic way, but the condition of united filaments in the southern Cape plants indicates that they comprise a clearly recognizable form of D, iridioides. 4. Dietes flavida Obermeyer, Fl, PI. Africa 149: tab. 1488. 1967. type: South Africa, Natal, Jozini Dam, Admiraal 5293 (PRE, holotype). — Fig. 6. Plants 50-70 cm high. Leaves 30-50 cm long, linear-ensiform, 15-22 mm wide. Stem bearing leaves below and reduced sheathing spathelike leaves above; branching irregularly; stem bracts 30-50 mm long. Spathes (40-)45-50 mm long, the outer ca. V2 inner, the apices obtuse to emarginate. Flowers pale yellow, with brown nectar guides on the outer tepals and spotted on the claw; outer tepals 30^0 mm long, the claw ca. 15 mm long, the limb horizontal, 15-17 mm wide; inner tepals smaller, to 38 mm long. Filaments 4-6 mm long, broadened at the base; anthers 5-6 mm. Ovary 10-14 mm long; style 2-3 mm long, the branches ca. 8 mm long, 3-4 mm wide; crests 5-10 mm, acute. Capsule ovoid, pendulous, smooth, 30-35 mm long, ca. 12 mm in diameter, dehiscing irregularly. Chromo- some number: In 20. Flowering time: sporadically during the summer months. Distribution: along forest margins and lightly shaded areas, Lebombo Moun- tains of northern Natal and Swaziland, also recorded from the Baviaans Kloof Mountains in the eastern Cape; Fig. 7. Dietes flavida is closely related to the widespread D. iridioides. and is easily confused with this species. When living the two are readily distinguished by flower color as well as leaf characteristics. Dietes flavida has cream to yellow 1981] GOLDBLATT— /:>/Er£:5 149 y^/-^ Figure 7. Morphology of Dietes butcheriana, and distribution of D. hutcheriana, D. flaviJa and D. grandiflora. Flower and fruits xl; ovary, stamen and style branches xl.5. flowers with brown-spotted nectar guides, the style branches are usually also cream or rarely very lightly flushed with purple, and the leaves have a grayish waxy covering. Dietes iridioides, in contrast, has white flowers with style branch- es conspicuously blue-purple flushed, and dark green leaves. When dry these 150 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Voi . 68 characters are lost, and fruits are needed for accurate determination. The fruits of D. flavida are oblong-ovoid, smooth, pendulous, and dehisce irregularly in- stead of along the carpel sutures. The fruits of D. iridioides are typically oblong and distinctly beaked, have a somewhat fissured surface, and are either indehis- cent, or they split along the carpel sutures from the apex downwards for some distance. Dictes flavida is characteristic of more open sites than the shade-loving D. iridioides, and it is reported to occur along forest edges, on cliffs, and along streams. When described by Obermeyer in 1967, D. flavida was thought to be restricted to the southern Lebombo Mountains in Natal and Swaziland. It has now been collected by R. D. Bayliss in the Baviaans Kloof Mountains of the southeastern Cape. This range disjunction is remarkable. Further collecting may bring to light some stations between these two extremes, but it is clear that this is nevertheless a true disjunction. 5. Dietes butcheriana Gerstner, J. S. African Bot. 9: 149. 1943. types: South Africa, Natal, Nkandia Forest, Zululand, Gerstner 601 (NH, lectotype); Ger- stner 4159 (NH, syntype); Obermeyer, FI. PI. Africa 149: tab. 1487. 1967. Fig. 7. Plants 50-120 cm high. Leaves larger than the stem, to 100(-120) cm long, ensiform, 30^8 mm wide. Stem thick, often somewhat flexuous, 50-60 cm long, with small leaves on the lower nodes and sheathing spathelike bracts on the upper nodes; stem bracts 30-60 mm long; branches relatively short. Spathes 45-60 mm long, the outer ca. Vi the inner, acute, obtuse or emarginate at the apex. Fknvers white with yellow nectar guides on the outer tepals, the claws of the inner and outer tepals speckled orange, the style branches white; outer tepals 30-35 mm long, the claw 12 mm long, bearded, the limb spreading, 15-20 mm wide; inner tepals 25-35 mm long, to 12 mm wide, the limb spreading. Filaments 6 mm long; anthers 4^.5 mm long. Ovary ca. 10 mm long; style ca. 3 mm long, the branches 8-9 mm long, 6-8 mm wide; crests 7-10 mm long. Capsule subglobose, 25-35 mm long, 20 mm in diameter, subpendulous, indehiscent, smooth when ripe. Chromosome number: 2/t - 20. Flowering time: sporadic, mainly spring and summer. Distribution: in deep shade, in mist-belt forests, Natal and Zululand; Fig. 7. Dietes butcheriana, a Natal endemic of moist forest areas, is closely related to D. iridioides and may be regarded as a specialized derivative of this widespread African species. Dietes butcheriana can easily be distinguished by its very broad leaves, ca. 30-50 cm wide, white flowers without purple or blue-flushed style branches, and globose, smooth, indehiscent capsules. 6. Dietes grandiflora N. E. Br., J. Linn. Soc. Bot. 48: 35. 1928. type: South Africa, Cape, Kentani Div., Pef^ler 484 (K, lectotype; BOL, isolectotype). Fig. 8. Plants tall, 1-1.5 m high. Leaves 75-100 mm long, linear, 10-15(-20) mm wide. Stem branched irregularly, bearing short leaves at the lower nodes and spathelike bracts from the upper nodes; stem bracts 25-50 mm long, entirely 1981J G O L D B 1 , A VT—DIE TES 151 Figure 8. Dictes grcmdifloni. Flower xl; stamen and style branches x2; fruit x2. sheathing, herbaceous, the apices obtuse apiculate. Spathcs 50-80 mm long, the outer about V2 the inner, the margins free to the base, the apices obtuse or emarginate. Flowers white with yellow nectar guides and a yellow beard on the outer tepals, the inner tepals marked dark brown towards the base, the style branches pale mauve; outer tepals 45-60 mm long, the claw 20-26 mm long, with a dense yellow beard down the midline, the limb horizontal to recurved, 25-35 mm wide; inner tepals (36-)40-45 mm long, to 25 mm wide, the limb spread 152 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol.68 horizontally. Filaments 10-13 mm long, tapering from a broad base to the apex, contiguous at the base; anthers 7-10 mm long. Ovary 13-16 mm long; style ca. 5 mm long, the branches 12-20 mm long, 6-8 mm wide; crests 12-15 mm long, erect. Capsule 28^5(-50) mm long, ridged and furrowed vertically, cylindrical. Chromosome number 2n = 20. Flowering time: sporadic, during spring and summer. Distribution: forest margins and especially along streams, the Eastern Cape from King William's Town through the Transkei to Natal; Fig. 7. Dietes grandiflora stands out as the largest-flowered species of the genus. It is related to the more widespread D. iridioides but is easily distinguishable by its height and large flowers, which last three days in contrast to those of D. iridioides which last a single day. The flowers of D. grandiflora are conspicuously marked with a heavy yellow beard on the claws of the outer tepals, while the inner tepals have dark brown markings. This species is widely cultivated in areas of tropical to subtropical climates, and it will stand light frost. Once established, the plants are extremely persistent even when completely neglected. However, more and larger flowers and a longer flowering season reward those who give the plant better care. Literature Cited Baker, J. G. 18%. Irideae. In W. T. Thiselton-Dyer (editor), Flora Capensis 6: 9-171. Reeve & Co., London. 1898. Irideae. In W. T. Thiselton-Dyer (editor). Flora of Tropical Africa 7: 337-376. Reeve & Co., London. Banerjee, J. & A. K. Sharma. 1971. A cytotaxonomical analysis of several genera of the family Iridaceae. PI. Sci. 3: 14-29. Barnard, T. T. & P. Goldblatt. 1975. A reappraisal of the application of specific epithets of the type species of Moraea and Dietes (Iridaceae). Taxon 24: 125-131. Brown, N. E. 1928. The Iridaceae of Thunberg's Herbarium. J. Linn. Soc. Bot. 48: 34-42. Brummitt, R. K. 1978. Report of the committee for Spermatophyta, 20. Taxon 27: 285-289. Chimphamba, B. B. 1974. Karyotype analysis in Moraea and Dietes. Cytologia 39: 525-529. Gerstner, F. J. 1943. Dietes butcheriana Gerstner. J. S. African Bot. 9: 149-151. Goi-DBi ATT, P. 197L Cytological and morphological studies in the southern African Iridaceae. J. S. African Bot. 37: 317^60. — . 1973. Proposal for the conservation of the generic name Dietes Salisb. (1812) against Naron Medik. (1796) (Iridaceae). Taxon 22: 504-505. — . 1976. A revision of Moraea (Iridaceae) in the winter rainfall region of southern Africa. Ann. Missouri Bot. Gard. 63: 657-786. . 1978. A contribution to cytology in Cornales. Ann. Missouri Bot. Gard. 65: 650-655. . 1979. Preliminary cytology of Australasian Iridaceae. Ann. Missouri Bot. Gard. 66: 851-855. Ki ATT, F. W. 1866. Revisio Iridearum: Dietes. Linnaea 34: 583-586. . 1882. Erganzungen und Berichtigungen zu Baker's Systema Iridacearum. Halle. 1895. Irideae. In Th. Durand & H. Schinz (editors), Conspectus Florae Africanae 5: 143- 230. Charles Vande Weghe, Bruxelles. Lewis, G. J. 1954. Some aspects of the morphology, phylogeny and taxonomy of the South African Iridaceae. Ann. S. African Mus. 40: 15-113. Medikus, F. C. 1790. Naron. Hist. & Comment. Acad. Elect. Sci. Theod.-Palat. 6: 419. MoENCH, C. 1794. Methodus Plantas. Marburg. Libraria Academiae. Obermeyer, a. a. 1967a- Dietes butcheriana. Fl. PL Africa 38: tab. 1487. . !%7b. Dietes ftavida. Fl. PI. Africa 38: tab. 1488. . l%8a. Dietes vegeta. Fl. PI. Africa 39; tab. 1524. . l%8b. Dietes bicolor. Fl. PI. Africa 39: tab. 1525. Raven, P. H. & D. 1. Axelrod. 1974. Patterns of angiosperm distribution in the light of continental drift. Ann. Missouri Bot. Gard. 61: 539-673. 1981] GOLDBLATT— /^/tT£3 153 & . 1978. Origin and relationships of the California flora. Univ. California Publ. Bot. 72: 1-134. Rii KY, H. p. 1962. Chromosome studies in some South African monocotyledons. Canad. J. Genet. Cytol. 4: 50-55. Sakai, B. 1952. Zytologische Untersuchungen bei Iridaceae I, Uber die Karyotypen verschiedener Arten der Unterfamilie Iridioideae. Cytologia 17: 104-1 1 1. Salisbury, R. A. 1812. On the cultivation of rare plants, especially such as have been introduced since the death of Mr. Philip Miller. Trans. Hort. Soc. London 1: 261-360. Sharma, A. K. & A. Sharma. 1%0. Cytology of some members of the family Iridaceae. Cytologia 26: 274-285. Spae, D. 1846. Dieles bicolor. Ann. Soc. Roy. Agric. Gand 2: 233-234. . 1852. Moraea hicolor. Fl. Serres Jard. Eur., ser. 1, 7: tab. 744. Sweet, R. 1830. Hortus Britannicus. Ed. 2. Ridgeway, London. . 1839. Hortus Britannicus. Ed. 3. Ridgeway, London. ADDITIONS TO THE ERICACEAE OF PANAMA^ Robert L. Wilbur^ and James L. Luteyn^ AbS I RACT The following laxa of Ericaceae are newly described from Panama: Cavendishia aberrans Lu- teyn, C. arizonensis Luteyn, C. chiriquiensis var. huUata Luteyn, C. fortunensis Luteyn, Didonica panamcnsis Luteyn & Wilbur, Disteri^ma hammelil Wilbur & Luteyn, Lateropora tuhulifera Wilbur & Luteyn, Macleania megahracteata Wilbur & Luteyn and Themistoclesia revoluta Wilbur & Lu- teyn. Cavendishia quereme (H.B.K.) Bentham & Hooker is reported for the first time from Panama, being previously known only from Costa Rica and Colombia. Vaccinium dissimile Blake is now known from Colon Province and the range of Vaccinium jefense Luteyn & Wilbur is extended to include Code Province. Keys to the Central American representatives of the genera Distengma, Lateropora, and Theniistoclesia are provided. It may be surprising to many that, although an account of the Ericaceae of Panama was published only recently (Wilbur & Luteyn, 1978) based upon a study of all collections available in the American herbaria with significant holdings of Panamanian plants, a paper would be published so soon increasing by 15% the known ericad flora of that small county. All of these additions are based upon collections previously unavailable to us, and almost all are from collections made In only recently explored areas. It is safe to predict that many additional species await discovery in the extensive areas of Panama that are botanically either unex- plored or little explored, and it is to be hoped that funds for continued exploration will be forthcoming. Obviously the botanical exploration of this biogeographically critical area is far from completed. Cavendishia aberrans Luteyn, sp. nov. Frutex epiphyticus. Folia ovalo-lanceolata, 13-15 cm longa, 4.5-6 cm lata; petioli 15-17 mm longi. Inflorescentia racemosa, floribus 20; rhachides 12 cm longae; bracteae florales oblongae, 7-10 mm longae; pedicelli 19-20 mm longi. Calyx 6-7 mm longus; lobi calycis 1 mm longi calloso-incrassati. Corolla 41-43 mm longa. Stamina 34-36 mm longa, filamentis alternatim 1-2 mm et 5-6 mm longis, antheris tubulis inclusis alternatim 29-30 mm et 33-35 mm longis, poris 3 mm longis. Epiphytic shrub\ mature branches subterete, bluntly angled, glabrous, brown when dry. Leaves ovate-lanceolate, 13-15 cm long, 4.5-6 cm broad, basally ob- tuse, apically long but broadly acuminate, tan or olive brown when dry, glabrous, the margins slightly revolute, 3-plinerved, the nerves arising from the base, the midrib impressed above and raised beneath, the lateral veins plane above and raised beneath, the veinlets reticulate and slightly raised on both surfaces; petioles subterete, rugose, 15-17 mm long, 2.5 mm in diameter, glabrous, glaucous. In- florescence viscid, 20-flowered, the lowest 3-4 floral bracts sterile; rachis 12 cm long, 3.5 mm in diameter at the base, bluntly angled, glabrous but with scattered ' Assisted by National Science Foundation Grant DEB 77-0430 (W. G. D'Arcy, principal in- vestigator). Grateful acknowledgement is made to the National Science Foundation (DEB 76-10185 and DEB 77-13455) for its support which made this study possible. We are extremely grateful to Lie. Mireya D. Correa A. of the Universidad de Panama and William D'Arcy of the Missouri Botanical Garden for making these collections available to us. ^ Department of Botany, Duke University, Durham, North Carolina 27706. ^ The New York Botanical Garden, Bronx, New York 10458. Ann. Missouri Bot. Gard. 68: 154-166. 1981. 0026-6493/8 1 /0 1 54-0 1 66/$ 1 .45/0 I981J WILBUR & [.UTEYN— ERICACEAE 155 globular glands; floral bracts conspicuously striate, oblong, basally truncate, api- cally rounded, 7-10 mm long, 5-6 mm broad, each margin with 10-20 angular or globular sessile glands, some of these often caducous; pedicels subterete, gla- brous, 19-20 mm long, 1-1.2 mm in diameter, "purple," provided at the distal tip with 6-8 flesh-colored ovoid-angular short-stipitate glands each less than 0.5 mm in diameter; bracteoles continuous (not articulate) with the pedicel, subop- posite, ovate, 1.5 mm long, ca. 1 mm broad at the base, the margins completely callose thickened. Flowers with the calyx glabrous, 6-7 mm long, the hypanthium 2 mm long, capanulate, 5 (weakly 10)-ribbed, basally apophysate with the rim undulate or turned upward, "purple," the limb spreading, noticeably constricted at the base, "green," 3-4 mm long, the lobes triangular, obtuse, 1 mm long, ca. 3 mm wide at the base, erect after anthesis, completely callose thickened without and onto the limb, only the margins callose thickened within, the sinuses flat or broadly rounded; corolla glabrous, 41-43 mm long, 6 mm in diameter, "waxy red," the lobes triangular, obtuse, 2 mm long; stamens of equal length overall, 34-36 mm long, the filaments pilose along the margins and inner surfaces, alter- nately 1-2 mm and 5-6 mm long, the anthers including the tubules alternately 29- 30 mm and 33-35 mm long, the thecae 5-6 mm and 6-6.5 mm long, the tubules apically dehiscent by short clefts ca. 3 mm long with ragged and slightly flaring margins; style sigmoid, glabrous, 42^5 mm long. Berry not seen. Type: Panama, cocle: Near saw mill 8 km N of El Cope, 28 km NW of Penonome, very wet cloud forest, 600-750 m, 9 Jan. 1977, Maas, Berg & Dress- ier 2774 (U, holotype, photo NY). L Cavendishia aberrans is isolated within the genus, differing from all other species by its extremely long anther tubules and short dehiscence clefts. Species of Cavendishia usually have anther tubules and thecae of approximately equal length; however, the new species has tubules six times longer than the thecae. Also, species of Cavendishia normally have dehiscence clefts about half the tubule length; the new species has clefts only 3 mm long, being Vy-V;, the length of the tubules. In the long tubules in relation to the thecae and short dehiscence clefts, Cav- endishia aherrans resembles the genus Plutarchia A. C. Smith, an Andean genus of Colombia, which may thus strengthen an already suspected relationship be- tween these genera. Species of Plutarchia may have filaments of alternate lengths, although only slightly unequal, and floral bracts which reach 10 mm, thus superficially resembling Cavendishia. However, the floral bracts of Plutarchia are never showy. Plutarchia also differs in consistently smaller leaves (with only one exception leaves are less than 5 cm long); calyx hypanthia which are often angled or winged (never in Cavendishia) and never apophysate (common in Cav- endishia); and most important, in the lack of development of the various types of glands on the calyces and bracts of so many cavendishias. Cavendishia arizonensis Luteyn, sp. nov. Folia elliplica, ovato-elliplica vel obovata, 7-13.5 cm longa, 2-6 cm lata, bullata; petioli 7 mm longi. Inflorescentia racemosa, floribus 82; rhachides 27 cm longae; bracteae florales oblongae, 17-22 mm longae; pedicelli 12-15 mm longi. Calyx glaber, hypanthio dense glanduloso, nonapophysato; 156 ANNALS OF THE MISSOURI BOTANICAL GARDLN (Voi . 68 limbus campanulatus, lobis calloso-incrassatis et glandulosis secus margines inclusis 3.5-4 mm longus. Corolla immatura. Shrub; mature branches subterete, striate, glabrous, yellowish brown when dry, with minute, black stipitate, globular glands ca. 0,1 mm in diameter. Leaves elliptic, ovate-elliptic to sometimes obovate, 7-13.5 cm long, 2-6 cm broad, ba- sally cuneate, apically acute or short acuminate, olive brown above and tan be- neath when dry, glabrous but with scattered, elevated, reddish globular glands 0.1 mm in diameter along the upper surface, and scattered red glandular fimbriae on the lower surface, 5-plinerved, the veins all arising from the base, the midrib and lateral nerves deeply and conspicuously impressed above, raised beneath causing the leaves to be longitudinally compressed and bullate, the veinlets im- pressed above, raised beneath; petioles subterete, flattened above, rugose, ca. 7 mm long, 4 mm in diameter, glabrous. Inflorescence racemose, viscid, 82-flow- ered; rachis subterete, bluntly angled, glabrous, 27 cm long, 6 mm in diameter at the base, the proximal 9-10 cm densely provided with globular glands, the distal portion without glands; floral bracts glabrous, oblong, 17-22 mm long, ca. 11 mm broad, basally clasping, apically rounded, marginally with a few globular glands; pedicels subterete, bluntly angled or striate, glabrous, 12-15 mm long, 1.5-2 mm in diameter, with scattered globular glands 0.2-0.3 mm in diameter, especially at the distal tip; bracteoles basal, broadly ovate, ca. 4 mm long, 3.5 mm broad, marginally glandular callose-thickened at the distal tip and flanked by oblong callose thickenings on each side (not globular glands). Flowers with the calyx glabrous, viscid, 6-6.5 mm long, the hypanthium cylindric, rugose, 2.5 mm long, basally nonapophysate, covered by globular, angular glands to 0.5 mm in diam- eter, the limb campanulate, 3.5^ mm long including the lobes, without glands, the lobes triangular, 1.5 mm long, marginally glandular callose-thickened, the sinuses broadly rounded; corolla immature, cylindric, 18 mm long, 5-6 mm in diameter, 'Mavender,'' pilose, the distal third especially dense near the tip; sta- mens equal, 16 mm long, the filaments distinct, alternately 2 mm and 5 mm long, short pilose on the ventral side of the distal half, the anthers including the tubules alternately 12 mm and 15 mm long, the thecae 6 mm long, the dehiscence clefts 5 mm long; style 17 mm long. Beny not seen. Type: Panama, veraguas: N of Santa Fe, summit of Cerro Arizona, heath- like terrain formed by the dense crowns of elfin forest trees, 4700 ft, 10 Sep. 1978, Hammel4733 (MO, holotype). In Wilbur & Luteyn (1978) Cavendishia arizonensis keys to C. panamensis to which it is most closely related. However, the differences, in combination, make the new species a very distinctive plant. Cavendishia arizonensis has a longer rachis (27 cm vs. 8-17 cm) with many more flowers (82 vs. 1 1-37); shorter floral bracts (17-22 mm vs. 20-50 mm), pedicels (10-15 mm vs. 11-19 mm) and corollas (18-20 mm vs. 25-35 mm); bracteoles with callose-thickened glands (not globular glands); and leaves which are strikingly bullate and have short acute tips (not flat and long acuminate to caudate-acuminate). Also, in C. panamensis, the floral bracts are proportionally longer covering the pedicels, calyx and lower half of the corolla at anthesis, whereas they cover only the pedicels and calyx in the new species. One other interesting note about the new species is that the inflo- 1981] WILBUR & LUTEYN— ERICACEAE 157 rescence and especially the calyces are covered by dead ants or ant-body parts. This is probably the result of ants being attracted to the glandular exudate (for feeding purposes?) and then getting stuck and being unable to escape the highly sticky substance. This phenomenon is frequently observed on species of Cav- endishia which secrete a viscid latex (e.g., C. lactiviscida, C. ciliata, C pana- mensis, etc.). Cavendishia chiriquiensis var. buUata Luteyn, var. nov. A var. chiriquiensi foliis valde bullatis, rhachidibus longioribus, bracteolarum el hypanlhii forma, corollarum pubescentia, et distributione geographica diftert. Type. Panama, cocle: 7 km N of El Cope, near Rivera Sawmill, 70-850 m, 10 Sep. 1977, Folsom 5239 (MO, holotype). Other Specimens Examined: Panama. cuiRiQuf: Cerro Colorado, road along top, 1500-1750 m, Folsom et al. 4694 (MO). Cerro Colorado, cloud forest on continental divide, 1200-1500 m, Mori &. Dressier 7786 (MO, NY). 28 km from Ri'o San Felix bridge, 1500 m, Sullivan 280 (DUKE, MO). Cavendishia chiriquiensis A. C. Smith has been collected from three geo- graphically and altitudinally separated areas — the Boquete region at 1,700-1,900 m (type location of var. chiriquiensis), the Cerro Colorado area at 1,200-1,7.^0 m, and now the region of El Cope at 700-850 m. Cavendishia chiriquiensis var. hulhita is restricted to Cerro Colorado and El Cope and differs from the typical variety primarily in the characters mentioned in the diagnosis and in several supporting character-features noted in Table 1. The Cerro Colorado populations themselves are somewhat intermediate in leaf size, shape and venation, in rachis length, in floral bract length, in calyx lobes, and in corolla pubescence. However, the overriding characters of bullate leaves and the usually longer rachises give it and the El Cope population a very different appearance, and one which merits varietal recognition. One collection of var. bullata from Cerro Colorado, Mori & Dressier 7786, was annotated as C. chiriquiensis in 1976 and was so cited in the treatment for the Flora of Panama (Wilbur & Luteyn, 1978). Many more collections of both varieties are needed from areas between Boquete, Cerro Colorado and El Cope to understand the variation which is only now coming to light. Cavendishia fortunensis Luteyn, sp. nov. Frutex epiphyticus. Folia elliptica, (5-)7-l I cm longa, 2^ cm lata; petioli 3-5 mm longi. Inflo- rescentia racemosa, floribus 20; rhachides II cm longae; bracteae florales oblongae vel ovalo- ellipticae, 21-23 mm longae, marginibus glandulosae; pcdicelli 5-7 mm longi; bracteolae oblongo- oblanceolatae, 12-14 mm longae, marginibus glandulosae. Calyx glaber, hypanthio dense glanduloso; limbus campanulatus, lobis calloso-incrassatis. Corolla 19 mm longa. Stamina 17 mm longa, filamentis alternatim 2.5 mm et 6 mm longis, antheris tubulis inckisis alternatim 13 mm et 16.5 mm longis, poris 5-6 mm longis. Epiphytic shrub; branches subterete or bluntly angled, slightly striate, gla- brous, with scattered globular glands, reddish brown when dry. Leaves elliptic, (5-)7-ll cm long, 2-4 cm broad, basally obtuse or narrowly rounded, apically acuminate, sometimes abruptly short caudate-acuminate, reddish brown when dry, glabrous but with scattered, elevated, black globular glands O.I mm in di- ameter along the upper leaf surface, these often caducous leaving a reddish 158 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Table 1. Comparison of CavenJishia chiriquiensis var. chiriquiensis and var. bullata. Var. chiriquiensis Var. bullata Boquete Cerro Colorado El Cope Leaves ovate; caudale- acuminale; 4.5-8 cm long; 3-5- plinerved, flat, not buUate elliptic; caudate- acuminate; 4.5-6 cm long; 3-5- plinerved, 3 nerves usually impressed lanceolate; caudate- acuminate; 7-9 cm long; 5-plinerved, 5 nerves deeply impressed Inflorescence 6-9(-ll)-flowered (7-)10-12-flowered 1 9- 11 -flowered; but the lowest 2 nodes often sterile Rachis 1 1.5-4.5 cm long; globular glands rare; sometimes flexuous (2.5-)6-7.5 cm long; globular glands often dense in the basal 1 cm; nonflexuous 10.5 cm long; globular glands over the entire length; nonflexuous Floral bracts (length) 15-25 mm 17-20 mm 20-26 mm Bracteoles linear to linear- lanceolate oblong-ovate ovate to oblong- ovate Hypanthjum slightly apophysate apophysate apophysate Calyx lobes apiculate; entirely to nearly entirely callose thickened 0.5-0.75 mm; almost entirely callose thickened 1.5 mm; marginally callose thickened Corolla 4 1 pilose distal in the V^ glabrous, rarely weakly puberulous in the distal Vi 4 glabrous r ''punctate'' scar, 5(-7)-plinerved, the innermost lateral nerves arising slightly above the base, the midrib and lateral nerves deeply impressed above and con- spicuously raised beneath causing the leaves to be bullate, the veinlets slightly impressed above and raised beneath; petioles subterete, rugose, 3-5 mm long, 2,5-3 mm in diameter, hirsute, with globular glands 0,2 mm in diameter in the distal portion. Inflorescence viscid, ca. 20-flowered, the lowest few nodes sterile; rachis subterete, bluntly angled, striate, glabrous, at least 11 cm long (the upper portion still in bud) and 3 mm in diameter at the base, with globular or angular glands ca. 0.2 mm long scattered along its length; floral bracts oblong to oval- elliptic, glabrous, rose, basally narrowed, truncate and clasping, apically rounded, 21-23 mm long, 10-11 mm broad, marginally crisped and with 12-20 dark red globular glands 0.1-0.2 mm in diameter on each side; pedicels subterete, swollen distally, striate, glabrous, 5-7 mm long, 0.75 mm in diameter, with globular glands concentrated at the distal swollen portion; bracteoles oblong-oblanceolate, 12-14 mm long, 6-8 mm broad, located midway along the pedicel and clasping or nearly concealing the entire calyx and lower Vs of the corolla at anthesis, the margins cripsed and with dark red globular glands scattered along the edge. Flowers with the calyx glabrous, ca. 3,5-5 mm long, the hypanthium cylindric, obscurely ribbed, 1.5-2 mm long, basally truncate, covered by globular or angular glands 0.2 mm in diameter, the limb spreading-campanulate, 2-3 mm long including the lobes, both covered by globular glands, the lobes broadly triangular, 1 mm long, 1981] WILBUR & I.UTHYN— KRICACEAE 159 completely glandular callose-thickened the distal %, erect after anthesis, the si- nuses flat to broadly rounded; corolla ca. 19 mm long, 4 mm in diameter, slightly narrowed distally, glabrous without but sparsely pilose within, pink, the lobes triangular, acute to obtuse, 1 mm long; stamens 17 mm long, the filaments distinct, alternately 2.5 mm and 6 mm long, the short filaments sparsely pilose dorsally at the distal tips, the long filaments densely pilose ventrally in the distal half, the anthers including the tubules alternately ca. 13 mm and 16.5 mm long, the thecae ca. 7 mm long, the dehiscence pores 5-6 mm long; style ca. 19 mm long, glabrous. Beny not seen. Type: Panama, chiriqui: E del sitio de presa en Fortuna, 6 Mayo 1976 Mendozci 338 (DUKE, holotype; PMA, isotype, not seen; photo NY). In the key to Cavendishia from Panama (Wilbur & Luteyn, 1978), this new species would key closest to C. chiriquiensis but would have some characters of C panamensis. Upon closer examination, however, it is seen to be phenetically most similar to C. pseudo-stenophylla, differing most conspicuously in its elliptic (not linear-elliptic) leaves, glabrous bracts and corollas, and shorter bracteoles and corollas. These four species form a close and distinct group within the genus and are restricted to the mountains of Veraguas and Chiriqui provinces (see also Luteyn, 1976, for further discussion of this group). Cavendishia quereme (H.B.K.) Bentham & Hooker This species, recently collected in Panama, was known previously only from the Central Valley of Costa Rica and the region of Queremal in western Colombia. The Panamanian collections are morphologically similar to the specimens from Costa Rica and Colombia. Specimens Examined: Panama, chiriqui: Camino a Soledad, SO del campamento Fortuna (sitio de presa) desde la region de la finca Pitti, Correa et al. 2211 (MO). Camino hacia la finca Landau. NE del campamento de Fortuna (Hornito), sitio de presa, 1100 m, Correa et al, 2365 (DUKE). Al este del sitio de presa en Fortuna, Mendozci et al. 112 (DUKE). Didonica panamensis Luteyn & Wilbur, sp. nov. Frutex epiphyticus. Folia elliptica vel ovato-lanceolata, 4-7 cm longa, (1.5-)2-3.5(-5) cm lata, basi attenuato, apice acuminata. Inflorescentia axillaris, racemosa, floribus 3-4(-5); pedicelli 15-25 mm longi. Hypanthium ad pedicellum articulatum, 3^ mm longum; limbus calycis 7-9 mm longus lobis inclusis; lobi apiculati, 1 mm longi. Corolla campanulato-cylindrica, 12-13 mm longa. Stamina 10, 12-14 mm longa; filamenta 4-5 mm longa; anlherae 10-12 mm longae tubulis inclusis; tubuli 4.5- 5.5 mm longi. Stylus 18-19 mm longus. \ I Epiphytic shrub, glabrous except for the filaments; mature stems terete or subterete, grayish brown, the bark adherent or exfoliating in thin longitudinal strips, the immature stems bluntly angled, reddish. Leaves elliptic to ovate-Ian- ceolate, 4-7 cm long and (1.5-)2-3.5(-5) cm broad, basally slightly tapering and attenuate, apically acuminate, sometimes abruptly so, the margins obscurely and remotely crenate, each crenation tipped by a tiny reddish brown gland, abun- dantly provided beneath with minute glandular fimbriae arising from concave depressions in the leaf surface, 3(5)-plinerved, the midrib conspicuously im- pressed above, elevated beneath, the lateral nerves arising slightly above the base and slightly elevated on both surfaces; pseudostipules ca. 1.5 mm long; petiole subterete, rugose, 6-12(-l5) mm long, 2-3 mm wide, flanked the entire length by 160 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 the attenuate leaf blade. Inflorescence axillary, racemose, 3-4(-5)-flowered; rach- is subterete, striate, 3-6 mm long and 2 mm broad at anthesis but extending to 13 mm; floral bracts ovate, acuminate, 1.5 mm long, 1.5 mm broad, marginally glandular-fimbriate; pedicels terete, striate, 15-25 mm long, ca. 1 mm in diameter, expanding to 2 mm in diameter at the distal tip, articulate with the hypanthium; bracteoles ovate, sharply acuminate to awl shaped, ca. 1 mm long, ca. 1 mm broad, located V3 of the way up the pedicel, marginally glandular-fimbriate. Flow- ers with the calyx 11-12 mm long, the hypanthium subcylindric or slightly spread- ing, slightly rugose, 3-4 mm long, ca. 4 mm in diameter, the limb spreading, somewhat campanulate, slightly rugose, 7-9 mm long including the lobes, 1 1-13 mm in diameter at the distal tip, the lobes barely differentiated, broadly ovate, apiculate, ca, 1 mm long, 6 mm broad at the base; corolla broadly campanulate- cylindric, 12-13 mm long, 12-13 mm broad, pale yellowish green, surface of the lower half drying smooth, the upper half slightly papillate, the lobes triangular, 3-4 mm long, 4-5 mm broad, erect; stamens 10, equal, 12-14 mm long, the filaments distinct, 4-5 mm long, marginally and dorsally long pilose and with glandular-fimbriate trichomes dorsally along the upper half of the filament and connective, the anthers including the tubules 10-12 mm long, slightly incurved at the base, the thecae minutely granular, 5.5-6.5 mm long, the tubules rigid, spreading, 4.5-5.5 mm long, dehiscing latrorsely by ovate-elliptic clefts 2-3 mm long, the inner surface moderately provided with apicular trichomes to 1 mm long; style glabrous, 18-19 mm long, exserted. Berry not seen. Type: Panama, cocle: Continental divide N of Penonome on road to Co- clesito between Llano Grande and Cascajal, in forest remnant, 1600 ft, 2 May 1979, Hammel 7223 (MO, holotype). Other Specimen Examined: Panama, cocle: Continental divide ridge, Coclesito Road, Hammel 2537 (MO). Didonica punamensis is the second species in this recently described genus (Luteyn & Wilbur, 1977). Its stamens agree exactly with those of /). penduJa Luteyn & Wilbur, the type of the genus, providing further support for this generic distinction. The new species differs significantly, however, from /). pcnduhi in characters which modify the generic description: (1) the inflorescences of D. panamensis are racemose, being composed of 3-5 flowers, not of solitary flowers as D. pendula; (2) the pedicel and calyx of D. panamensis are articulate, not continuous as in D, pendula. Other specific differences are the shorter pedicels of D. panamensis (15-25 mm vs. 15-18 cm for D, pendula), and the glandular trichomes on the filaments and connectives of D. panamensis. The dis- covery of this new species has not, unfortunately, given us any new insight into the evolutionary relationships of Didonica with any other vaccinoid genus. Disterigma hammelii Wilbur & Luteyn, sp. nov. Frutex epiphyticus; ramulis pilosulis. Petioli 0.5-1 mm longi, glabri; lamina coriacea, glabra inlegra, elliptica vel lance-elliptica, (4-)5-7(-8) mm longa, 1 .5-2.2(-3.0) mm lata, Hypanthium glabrum, 1.5-1.8mmlongum, 1.0-1.4 mmdiametro: lobi calycis4,erecti,acuti, lanceolati, 2.5-3 mmlongi.ciliati. Corolla subrosa; tubus cylindraceus, 10-12 mm longus, 2.5-3.2 mm diametro; lobi 4, leflexi, 2.2-3 mm longi, lanceolati, pubescenti. Stamina 8; filamenta 8-9 mm longa, antherae lanceolatae, papillatae 1.2-1.5 mm longae, tubuli antherarum 2, 2.5-2.8 mm longi. 1981] WILBUR Si LUTEYN— ERICACEAE 161 Straggly epiphytic shrub] branchlets irregularly ridged and grooved, the recent growth reddish brown and densely spreading pilosulose with tawny to reddish brown trichomes 0.4-1.0 mm long, the older growth dull brownish to grayish and only moderately pilosulose with hyaline to tawny trichomes; internodes short, almost all less than !4 the length of the subtending leaves and usually 1-2 mm long. Leaves coriaceous, entire, elliptic to lance-elliptic, (4.0-)5-7(-8) mm long, 1.5-2.2(-3.0) mm wide, mostly about 2.5 times as long as wide, apically acute in general outline but the actual apex obtuse, gradually tapering basally into the petiole, glabrate on both surfaces and margins, rarely with a few appressed, reddish glandular trichomes on the lower surface, the midvein slightly elevated beneath but the venation otherwise indistinct, apparently dark green above and pale green beneath; petioles 0.5-1 mm long, glabrous or nearly so. Inflorescence of solitary flowers in the axils of foliage leaves, subtended by several (5-8) ovate to broadly elliptical, tawny to dull reddish brown, scarious bracts mostly to 0.8- 1.2 mm long; bracteoles 2, broadly clasping, ovate to broadly oblong, broadly rounded apically, scarious, pale greenish to yellowish, striate, 2-2.7 mm long, borne just beneath and embracing the hypanthium; pedicels ca. 0.5 mm long and obscured from view by the bracts and bracteoles. Flowers with the hypanthium terete, campanulate to obconic, glabrous to sparingly pilosulose, pale greenish, 1.5-1.8 mm long, 1.0-1.4 mm in diameter, the calyx lobes 4, strongly ascendent, narrowly lanceolate, acute, with incurved margins, 2.5-3 mm long, entire, ciliate for the distal V^-Vi but otherwise glabrous; corolla ''pale pink," the tube cylindric, 10-12 mm long, 2.5-3.2 mm in diameter, nearly glabrous but usually with a few appressed glandular trichomes ca. 0.1 mm long just above the middle, the lobes 4, ascendent to more typically reflexed at maturity, 2.2-3 mm long, lanceolate, acute, distally spreading short-pubescent with hyaline trichomes; stamens 8, about as long or slightly longer than the corolla tube, the filaments 8-9 mm long, flattened, sparingly hyaline pilosulose with trichomes 0.2-0.6 mm long, the an- thers lanceolate, the thecae minutely papillate 1.2-1.5 mm long, each tapering into a separate, slender tubule 2.5-2.8 mm long; style glabrous, exserted beyond the corolla tube, 13-14 mm long. Berry not seen. Type: Panama, chiriqui-bocas del toro: Border trail along continental divide, ca. 5 mi NE of Boquete near Cerro Pate Macho above Palo Alto, 7100 ft, 23 May 1979, Hammel 7390 (DUKE, holotype). This species is named in honor of Mr. Barry Hammel who spent approximately 18 months collecting for the Missouri Botanical Garden in Panama. We are most appreciative of his special attention to the ericads during his botanical explora- tions. Disterigma hanunelii belongs to a small group of disterigmas which normally have a scandent, ''wiry'' epiphytic habit, short internodes, and narrowly lanceo- late to nearly linear leaves less than 5 mm wide. This group ranges from western Panama to central Peru and includes D. agathosmoides (Weddell) Niedenzu, D. weberhaueri Hoerold, D, panamensis Standley, D. luteynii Wilbur, and D. ham- melii. From this group the new species may be distinguished by its longer corolla and filaments, and by its exserted style. The description of this species of Disterigma dramatically illustrates the im- pact of recent field work in Central America. A synopsis of the Central American 162 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Species was published seven years ago (Wilbur, 1974) summarizing all that was known of the genus in Central America at that time (three species, including one therein described as new). This is the third new species described since then and the sixth species that we recognize from Central America. We have seen a small, incomplete specimen from the ''summit of Cerro Arizona, north of Santa Fe, Veraguas'' {Hummel 4742, MO) which is undescribed. It has a 4-parted calyx and elliptic leaves about 20 mm long and half as wide lacking an apiculus. We await fuller material before formally describing it. A key distinguishing the six Central American species follows. Key to the Central American Species of Distehgma 1. Lobes of the calyx and corolla 3; corolla deep red D. trimerum Wilbur & Luteyn r. Lobes of the calyx and corolla 4; corolla white or pale greenish. 2. Thecae of anthers both emptying into a single tubule L5 mm long; leaves mostly more than 1 cm long and wide, with an apicule 1-2 mm long __ D, utleyorum Wilbur & Luteyn 2'. Thecae of anthers each emptying into its own tubule 2 mm long or longer; leaves mostly 1 cm or less long and wide, nonapiculate. 3. Leaf blades 4 mm wide or wider; calyx lobes eciliate; calyx tube 4-angled in cross- section; Guatemala, Costa Rica, Panama (ChiriquO and northern South America D. humholdtii (Klotzsch) Niedenzu 3'. Leaf blades 1.5-4.5 mm wide; calyx lobes ciliate; calyx tube terete or nearly so in cross-section; endemic to Panama. 4. Calyx lobes 2.5 mm long or longer; corolla tube 10 mm long or longer; corolla lobes 2.2 long or longer; filaments more than 7 mm long; Chiriqui-Bocas del Toro, Panama D. hammeUi Wilbur & Luteyn 4'. Calyx lobes 2.5 mm long or less; corolla lube 6 mm long or less; corolla lobes less than 2 mm long: filaments less than 6 mm long; Panama and Darien, Panama. 5. Calyx lobes 2-2.5 mm long; leaves typically widest at or near the middle. usually linear, mostly ca. 4 times as long as wide; Darien . D. panamense Standley 5'. Calyx lobes 1.5-1.7 mm long; leaves typically widest decidedly above the middle, oblanceolate, mostly ca. 3 limes as long as wide; Panama _._ D, lutc\nU Wilbur Lateropora tubulifera Wilbur & Luteyn, sp. nov. Frutex ca. 1.5 m alius. Folia coriacea. elliptica, 2.5-3.8 cm longa, 1.2-1.8 cm lata; petioli glabri, 3^ mm longi. Inflorescentia axillaris, corymboso-racemosa, 2-3 cm longa; bracleae ovatae vel ob- longae, 1.5-2.5 mm longae, 1.5-2 mm latae, ciliatae; pedicelli 10-18 mm longi, glabri; bracleoli ciliali, 2-2.2 mm longi, 1.5-1.8 mm lati, ciliali. Hypanlhium ad pedicellum articulalum, campanulatum, 5- angulatum, 3-3.5 mm longum, glahrum; lobi calycis 5, erecti, triangulares, aculi, glabri, 1.0-L5 mm longi. Corolla urceolata, 3.5-5 mm longa, extus glabra, inlus villosa; lobi corallae 5, triangulares, aculi, 1-1.2 mm longi, inlus villosi. Stamina 10; filamenta 2-2.2 mm longa, supra pilosa; thecae granulosae, ca. 2.5-3.0 mm longae; lubuli ca. 0.8 mm longi. Shrub 1.5 m tall; immature branchlets moderately thick, 2-2.5 mm in diameter, glabrous, dark purplish, striate to irregularly angulate, the mature branches gray- ish to brownish. Leaves coriaceous, elliptic, 2.5-3.8 cm long, 1.2-1.8 cm wide, apically acute to obtuse and acutely tapering to the base, entire with an elevated or slightly thickened margin, glabrous or glabrate beneath, indistinctly pinnately 3^ M axillary, corymbosely racemose, (l-)3-5-flowered, 2-3 cm long; rachis 1-1.5 cm long, glabrous; bracts 3-4, broadly oblong to orbicular, basal, peduncular, short- ciliate, 1-3 mm long, ca. 1.5-2.2 mm wide; floral bracts ovate to oblong, 1.5-2.5 mm long, 1.5-2 mm wide, glabrous except for short cilia, slightly keeled and 1981] WILBUR & LUTEYN— ERICACEAE 163 short-apiculate, irregularly narrowly hyaline margined; pedicels slightly clavate, 10-18 mm long, medially ca. 1 mm in diameter, glabrous except for an incon- spicuous fringe of glandular trichomes 0.2-0.4 mm long borne distally just beneath the sharply delimited disarticulation groove; bracteoles subopposite to alternate, ovate to oblong, 2-2.2 mm long, 1.5-1.8 mm wide, ciliate, irregularly hyaline margined, appressed, located in the lower proximal third. Flowers with the hy- panthium campanulate, clearly 5-angulate in the lower half with 5 rounded ridges ascending between the calyx lobes, 3-3.5 mm long, 3-3.2 mm in diameter at anthesis, glabrous, disarticulating from the pedicel by a clearly marked groove, the calyx limb somewhat flaring or at least wider than the hypanthium, glabrous, 0.7-1.0 mm long, the calyx lobes 5, erect, deltoid, acute, 1.0-1.5 mm long, basally 1.8-2.2 mm wide, glabrous; corolla urceolate, 3.5-5 mm long, "pink," externally glabrous, internally densely pilose to tangled villous in the distal half with white trichomes 0.6-1.0 mm long; corolla lobes 5, triangular, acute, 1-1.2 mm long, glabrous externally and densely tangled villous internally; stamens 10, the fila- ments flattened, 2.0-2.2 mm long, basally less than 0.5 mm wide, distally attached for about V^ the length from just above the strongly incurved base to the thecae, densely pilose distally with white trichomes 0.5-1.2 mm long, the anthers con- spicuously granular, 2.5-3 mm long including the strongly incurved base of V4-V3 the length of the thecae but excluding the smooth, ca. 0.8 mm long tubules, dehiscing by a slit extending from the apex of the tubule to the pronounced basal curve. Type: Panama, chiriqui: Cerro Hornito, in cloud forest, 6000 ft, 8 May 1978, Hummel 3069 (MO, holotype). The specific epithet of this species refers to the tubules of the anthers which are at least twice as long as those of the other two known species. The leaf blades are less than 4 cm long and their smallness is in great contrast to the leaf blades of the other two species of this endemic Panamanian genus. A key to the genus follows. Key to the Genus Lateropora 1. Calyx lohes 2-3 mm long. 4-5 mm wide basally; corolla lobes ca. 3 mm long; petioles 5-15 mm long; leaves 6-13 cm long, 4-9 cm wide L. ovata A. C. Smith r. Calyx lobes 2 mm long or less, 2.2 mm wide or less basally: corolla lobes less than 2 mm long; petioles 5 mm long or less; leaves 2.5-8.5 cm long, 1.2-4.5 cm wide. 2. Leaves when young appressed strigillose beneath with trichomes 0.2-0.3 mm long; tubules of the anthers 0.3 mm long or less; leaf blades 3-8.5 cm long; Veraguas L. sanuifccnsis Wilbur & Luteyn 2'. Leaves glabrous beneath; tubules of the anthers ca. 0.8 mm long; leaf blades 2.5-3.8 cm long; Chiriqui L. luhiilifera Wilbur & Luteyn Macleania megabracteolata Wilbur & Luteyn, sp. nov. Frutex epiphyticus. Lamina foliorum coriacea, elliptica, (5-)7-!0(-l2.5) cm longa, (2.7-)3-4.5 (-5.7) cm lata, apice acuta vel acuminata, basi cuneata vel rotunda, 5-7-plinervia. Intlorescentia axillaris, racemosa. 4-7 cm longa, floribus 4-7; bracteae oblongae vel spatulatae, 9-13 mm longae, integrae, ciliatae; bracteolae 2. ellipticae, aculae. 13-16 mm longae. ca. 5 mm latae; pedicelli arti- culati, 5-11 mm longi. Hypanthium angulatum. 5-7 mm altum; limbus calycis ca. 2 mm altus; lobi calycis 5. 3.5^ mm longi, acuti, glanduloso-puberuli. Corolla 12-14 mm longa, glanduloso-puberula; tubus cylindricus, roseatus, 10-12 mm longus, 4.5-5 mm diamctro; lobi 5, acuti, 2-2.5 mm longi. Stamina 10; filamenta distincta, 3-3.5 mm longa; thecae granulosae. ca. 3 mm longae; tubuli 2, 5-6 mm longi; pori obliqui introrsi, 0.6 mm longi. 164 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Epiphytic shrub; stems strongly and coarsely ridged and grooved, glabrous or nearly so, brownish. Leaves coriaceous, elliptic, (5-)7-10(-12.5) cm long, (2.7-)3-4(-5.7) cm wide, basally rounded or broadly tapering, apically acuminate with the tip 1 cm or more long or acute, entire and slightly revolute, glabrous above or sparingly to moderately spreading short-pubescent or puberulent on the very base of the midvein, glabrate beneath with punctate depressions marking former trichome attachments or, especially in younger leaves, with minute, red- dish brown appressed glandular trichomes 0.1-0.2 mm long, 5-7-plinerved, the secondary veins arising at or near the base with the midvein proximally promi- nently elevated and impressed distally while elevated beneath; petioles stout, rugose, 3 Inflorescence axillary, borne near the end of the branches, racemose, 4-7-flowered, the rachis, pedicels, bracts, calyces and corollas moderately provided with reddish brown appressed glandular trichomes 0.1-0.2 mm long: rachis striate or irregularly angled, 3-7 cm long, the lower 1 .5-2 cm enveloped by ovate to oblong bracts 5-8 mm long; pedicels striate, 5-11 mm long, markedly broadened and articulate just beneath the hypanthium; floral bracts oblong to spatulate, 9-13 mm long, entire, apically rounded; brac- teoles 2, borne just beneath the hypanthium, elliptic, acute, 13-16 mm long, ca. 5 mm wide. Flowers with the hypanthium basally strongly angulate or pentagonal from the prominently intersinal ridges in the lower half, 5-7 mm long, the calyx 5-4 2^ roseate, 10-12 mm long, 4.5-5.5 mm in diameter, the lobes triangular-deltoid, acute, 2-2.5 mm long, reportedly pale greenish; stamens 10, the filaments distinct, 3-3.5 mm long, spreading short-pubescent with slender trichomes 0.2-0.4 mm long, the anthers coarsely granular, ca, 3 mm long, each with two completely separate cylindrical tubules 5-6 mm long opening by an apical, introrse oval cleft; style glabrous, the stigma punctiform. Fruit unknown. Type: Panama, chiriqui: E de la presa en Fortuna, 22 Feb. 1976, MendozUy Mendieta & Mayo 191 (DUKE, holotype; PMA, isotype, 2 isotypes to be dis- tributed). This is the third known species of Macleania found north of South America in which each anther sac terminates in a distinct tubule. The bracteoles of the two previously known species, M. rupestris (H.B.K.) A. C. Smith [including M. glabra (Klotzsch) Hoerold] and A/, talamancensis Wilbur & Luteyn, are less than 4 mm long while those of M. megahracteolata are longer than 12 mm. Themistoclesia revoluta Wilbur & Luteyn, sp. nov. Frutex epiphylicus. Lamina foliorum pinnatinervia. elliptica vel oblanceolala, 5-6.5(-7.5) cm longa, 9-14 mm lata, basi cuneata, apice oblusa vel rotundata, Integra, revoluta, supra et subtus pubescentia; pelioli 3-4 mm longi, puberuli. Racemi axillares, 5-10 cm longi; bracleae 2.5^ mm longae; pedicelii 0.9-1.6 cm longi; bracteoli appressi, 2-3 mm longi. Hypanthium ad pedicellum non articulatum, 5-angulatum, pubescens; iimbus calycis 1-1.2 mm longus, lobi ca. 1 mm longi. Corolla 5-angulata, 8-9 mm longa; tubus glabrus, 5-5.5 mm longus; lobi erecti, 2.5-3 mm longi, ciliati. Stamina 10, 6-8 mm longa; filamenta ciliata, 1.5-2 mm longa; antherae 2.5-3 mm longae; tubuli antherarum 2, distincli, 5-6 mm longi. Stylus 8-9 mm longus, glabrus. Epiphytic shrubs; stems ridged and grooved but becoming terete, light brown to dun colored, minutely puberulent or short-pubescent. Leaves spirally ar- 1981] WILBUR & LUTEYN— ERICACEAE 165 ranged, narrowly elliptic to oblanceolate, 5-6.5(-7.5) cm long, 9-14 mm wide, basally cuneate, apically obtuse to rounded, entire, strongly revolute, moderately to densely short-pubescent on both the upper and lower surfaces, sparingly to moderately beset with glandular, reddish brown, appressed trichomes 0.1-0.2 mm long, the venation pinnate, the midvein prominently impressed above and elevated beneath, the secondary and tertiary veins when dry elevated above and moderately so beneath; petiole dorsally compressed and grooved, 3^ mm long, inconspicuously puberulous to short-pubescent. Inflorescence borne at distal tip of the branches, axillary, racemose, IO-16-flowered; bracts 6-8, basal, appressed, lanceolate to narrowly triangular, acute, 3.5-5 mm long, finely short-pubescent abaxially; rachis ridged or striate, 5-10 cm long, densely short-pubescent with hyaline trichomes 0.2-0.5 mm long; floral bracts 2.5-4 mm long, narrowly tri- angular to narrowly lanceolate; pedicels 0.9-1.6 cm long, nonarticulate with the hypanthium, sharply striate, moderately to densely spreading short-pubescent with the slender, hyaline trichomes 0.2-0.3 mm long; bracteoles appressed, nar- rowly triangular, acute, spreading short-pubescent, keeled, 2-3 mm long, usually medial or submedial, alternate. Flowers with the hypanthium obpyramidal, strongly 5-angulate or winged, the wings alternate with the lobes, moderately spreading short-pubescent throughout, ca. 5 mm long, ca. 5 mm in diameter at the apex, the calyx limb 1-1.2 mm long, the calyx lobes depressed-triangular, acute to apiculate, ca. 1 mm long; corolla pentagonal, narrowly 5-winged opposite the lobes, 8-9 mm long overall, the tube tapering from the base to the apex, glabrous, 5-5.5 mm long, the lobes erect, 2.5-3 mm long, sparingly ciliate along the midrib, strongly reflexed at anthesis exposing the stamens and style; stamens 10, 6-8 mm long, alternately slightly unequal, the filaments equal, 1.5-2 mm long, sparsely ciliate and glandular-fimbriate, the anther thecae 2.5-3 mm long, basally tapering into a short-setose appendix, closely coherent basally due to the inter- locking of the antheridial grooves; tubules 2 per anther, separate, twice as long as the thecae or 5-6 mm long, tapering upward and opening by a slender introrse slit ca. 1.5 mm long; style 8-9 mm long, glabrous, the stigma punctiform. Fruit unknown. Type: Panama, chiriqui; Bajo-Fortuna, 19 Mar. 1976, Mendoza, Mendieta & Mayo 258 (DUKE, holotype; PMA, isotype). Additional Specimen Examined: Panama, chiriqui: Desembocadura del Rio Hornitos, Mendozo et al. 239 (PMA). This species is unlike any other Central American species of Themistoclesia in its narrow, strongly revolute leaves and long-pedicelled flowers borne in elon- gate racemes. It is most closely related to 7. cutucuensis A. C. Smith from eastern Ecuador and adjacent Peru, but differs in its densely scabrous inflores- cences (including rachises, pedicels, bracts and calyces); its strongly revolute leaves with longer petioles, pinnate venation, and tapering (not subcordate) bases; and in its glandular anther connectives. A key to the Central American species of Themistoclesia follows. Key to the Central American Species of Themistoclesia m 1. Staminal tubules I per anther with a single introrse cleft; hypanthium distinctly 5-winged; leaves apically obtuse T. pentandra Sleumer 166 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 r. Staminal tubules 2 per anther, the tubules either separate or united but with 2 introrse clefts; hypanthium bluntly 5-angled, not winged; leaves apically acute or acuminate. 2. Corolla 2.5-3.5 mm long; leaves less than 1.9 cm long, cuneate 2'. Corolla 5-9 mm long; leaves more than 2 cm long. T. smithiana (Standley) Sleumer 3. Corolla narrowly 5-winged, 8-9 mm long; leaves 5-7 cm long, strongly revolute; flowers in !0-16-flowered racemes T. revoluta Wilbur & Luteyn 3'. Corolla not winged, 5-6 mm long; leaves 2-2.5 cm long, not at all revolute; flowers solitary. 4. Bracteoles ovate to hemispheric, completely concealing the glabrous pedicel; corolla cylindric, pilosulose throughout; stamens 5 mm long, the filaments pilosulose, the anther basally minutely appendiculate _ T. costaricensis Luteyn & Wilbur 4'. Bracteoles awl shaped, never concealing the pilosulose pedicel; corolla urceo- late, pilosulose distally; stamens 2.5 mm long, the filaments glabrous, the an- ther without a basal appendage T. horquetensis Luteyn & Wilbur Vaccinium dissimile Blake This peculiar creeping vine is easily overlooked and, at the time the account of the Ericaceae was prepared for the Flora of Panama, it was unknown from the Province of Colon. Specimen Examined: Panama, colon: S approach of Cerro Bruja from Rio Escandaloso, Hum- mel 3123 (MO). Vaccinium jefense Luteyn & Wilbur When described, this species was thought to be a narrow endemic in the vicinity of Cerro Jefe, an approximately 1,000 meter elevation mountainous area just east of the Canal Zone. It is now known from several collections made along the Continental Divide in Code Province. Specimens Examined: Panama, cocle: Summit of Alto Calvario, 900 m, Folsom & Robinson 2402 (MO); Folsom 2685 (MO). 7 km N of Llano Grande on road to Coclesito, 1700 ft. Hummel 1942 (MO). Near Sawmill above El Cope, 3000 ft, Hammel 2363 (MO). Literature Cited Luteyn, J. L. 1976. A revision of the genus Cavendishia (Vacciniaceae) in Mexico and Central America. Mem. New York Bot. Card. 28(3): 1-138. & R. L. Wilbur. 1977. New genera and species of Ericaceae (Vaccinieae) from Costa Rica and Panama. Brittonia 29: 255-276. Wilbur, R. L. 1974. The Central American species of the genus Disterigma (Ericaceae; Vaccinieae). Bull. Torrey Bot. Club 101: 245-249. & J. L. Luteyn. 1978. Ericaceae. In R. E. Woodson, Jr. & R. W. Schery, Flora of Panama. Ann. Missouri Bot. Gard. 65: 27-144. ADDITIONAL PANAMANIAN SPECIES OF BVRMEISTERA (CAMPANULACEAE: LOBELIOIDEAE^ Robert L. Wilbur^ Abstract Burmeistera hammelii Wilbur and B. mcvaughii Wilbur are newly described from Panama. Bur- meistera obtusifolia, previously known only from Costa Rica, is recorded for the first time from Panama. A key to all Panamanian species of Burmeistera is included. Barry Hammel spent approximately eighteen months as the Missouri Botan- ical Garden's collector in Panama. He proved himself to be a very zealous col- lector and his particular attention to the genus Burmeistera. for which I am most appreciative, resulted in the following additions to the known lobeliad flora of Panama. Since the two previously undescribed species of Burmeistera and the range extension of a Costa Rican species into Panama are a significant increase in the number of species known from Panama, it seems desirable to include a new key to facilitate their determinations. Species numbered 1-16 are described in Wilbur (1976) while X1-X3 are discussed after the present key. 1, Median cauline leaves more than 8 limes as long as wide, narrowly linear 1 . B. chiriquiensis Wilbur r. Median cauline leaves less than 6 times as long as wide, ovate, lanceolate, elliptic or at least not narrowly linear. 2. Anthers externally pilose with golden trichomes; plants vegetatively entirely glabrous 8. B. microphylla Donnell Smith 2'. Anthers either glabrous externally or with white or sordid, grayish trichomes; plants vegetatively either glabrous or pubescent. 3. Flowers 2.9 cm long or less. 4. Filament tube 10 mm long or more; hypanthium 3-6 mm long; corolla yellow or greenish. 5. Corolla yellow; calyx lobes acute, deltoid: filament puberulous 11. B. parviflora F. E. Wimmer ex Standley 5'. Corolla greenish; calyx lobes obtuse, oblong; filament tube distally glabrous 9. B. morii Wilbur 4'. Filament tube less than 9 mm long; hypanthium ca. 3 mm long; corolla purplish 7. B, kirkbridei Wilbur 3'. Flowers 3 cm long or longer. 6. Hypanthium in anthesis cylindric to campanulate or urceolate, basally somewhat rounded and with the sides more or less parallel. 7. Leaves lanceolate to lance-ovate, gradually long-acuminate from below the middle, 2.5 cm wide or less; plants vegetatively entirely glabrous 6. B. ^lauca (F. E. Wimmer) Gleason 7'. Leaves elliptic, ovate or broadly lanceolate, usually wider than 2.5 cm; plants vegetatively usually somewhat puberulent or scabcrulent. 8. Calyx lobes 5 mm long or less. 9. Leaf margins conspicuously denticulate or serrulate with (4-)5-9 teeth per cm XI. B. hammelii Wilbur 9' . Leaf margins inconspicuously serrulate with 3 or fewer teeth per cm. 10. Corolla puberulous; calyx lobes and hypanthium apiculate to puberulous; berry red 5. B. dukei Wilbur ' Grateful acknowledgment is due to the National Science Foundation (DEB 76-10185). Also assisted by National Science Foundation Grant DEB 77-0430 (W. G. D'Arcy, principal investigator). ^ Department of Botany, Duke University, Durham, North Carolina 27706. Ann. Missouri Bot. Gard. 68: 167-171. 1981. 0026-6493/81/0167-0171/$0.65/0 168 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 10'. Corolla glabrous; calyx lobes and hypanthium glabrous; berry while or greenish turning blackish. 1 1. Upper corolla lobes less than 12 mm long and the 2 lateral lobes about 7 mm long; corolla greenish; anthers externally puberulous 12. B. pirrensis Wilbur ir. Upper corolla lobes 18 mm long or longer and the 2 lateral lobes 12-14 mm long; corolla dark red; anthers externally glabrous _ ___ 3. B. darienensis Wilbur 8'. Calyx lobes usually more than 5 mm long. 12. Two lower anthers densely bearded apically with a dense tangle of trichomes; leaves mostly 5 cm or more wide; plants of Darien _ ___ 10. B. panamensis Wilbur 12'. Two lower anthers sparingly fringed apically or glabrous with not more than a sparse fringe of trichomes; leaves mostly less than 5 cm wide; plants from west of Darien. 13. Corolla green or bronzy; hypanthium and calyx lobes green, drying green or stramineous, rarely suffused with purple or maroon; berry greatly inflated with a thin, bladdery wall, 2-4 cm in diameter and 2.5-5 cm long 16. B. vulgans F. E. Wimmer 13'. Corolla predominantly purplish or green with a deep suffusion of maroon or dark purple; hypanthium and calyx lobes purplish and drying dark purplish or blackish; berry fleshy and with a substantial, thick wall, l-2(-2.5) cm in diameter and l-2(-3) cm long 2. B. cyclostigmata Donnell Smith 6'. Hypanthium in anthesis obconic, acutely tapering at the base with the sides strongly divergent. 14. Leaf margins conspicuously denticulate or serrulate with 6 or more teeth per cm X2. B. mcvaughii "^Wbwv 14'. Leaf margins entire to inconspicuously serrulate with 4 or fewer teeth per cm. 15. Calyx lobes 5 mm long or more. 16. Stems densely spreading hirsute with the trichomes visible with- out magnification; calyx lobes spreading or even reflexed, 5-10 mm long, 2.5 mm wide or wider .___ X3. B. obtusifolia F. E. Wimmer 16'. Stems glabrous or inconspicuously appressed pubescent; calyx lobes ascendent, (7-)l 1-20 mm long, 2 mm or less wide — 13. B. tenuiflora Donnell Smith 15'. Calyx lobes less than 4 mm long. 17. Corolla tube puberulent or inconspicuously appressed short-pu- bescent. 18. Pedicels and hypanthia glabrous to moderately spiculate or appressed short-pubescent or even strigillose; Panama east of Volcan Chiriqui 4. B, dendrophila F. E. Wimmer 18'. Pedicels and hypanthia moderately to densely spreading short-pubescent; Panama west of Volcan Chiriqui 15. B. utleyi Wilbur 17'. Corolla tube glabrous. 19. Filaments distally glabrous; anther tube externally glabrous; upper corolla lobes 7 mm long 14. B. toroensis Wilbur 19'. Filaments distally puberulous; anther tube externally pu- berulous; upper corolla lobes 10 mm long or longer 4. B. dendrophila F. E. Wimmer XI. Burmeistra hammelii Wilbur, type: Panama, Darien, NE slope of Cerro Hamm MO Planta erecta vel scandenta, ca. 1 m alta; ramuli puberuli vel glabrati. Lamina foliorum ovata vel elliptica, 7-19 cm longa, 3.5-9 cm lata, ca. 2-plo longiora quam lata, acuta vel acuminata ad apicem et basi rotundata vel cuneata, margine serrulata 4-9 serrulatis per cm; venulosi infra strigillosi; petioli puberuli, 0.8-4 cm longi. Flores 3^ cm longi; pedicelli glabri, 3^ cm longi. Hypanthium cylindricum 1981] WI LB U R^BURMEISTERA 169 vel obconicum, glabrum, 4-5 mm longum, 3-4 mm latum; lobi calycis deltoidi vel oblongi, acuti vel obtusi, denticulati, glabri, 2-3 mm longi. Corolla glabra, virenta; tubus 1.5-2 cm longus, ca. 2-3 mm diametro; lobi linear! vel oblongi, superiori 12-16(-18) mm longi, laterati 7-9 mm longi, inferiori 5- 6.5 mm longi. Filamenta glabra vel puberula, 22-27 mm longa; tubus anthearum appressi-pilosus, 4- 6 mm longus. Bacca globosa, 5-6 mm longa, 6-8 mm lata. Erect to scandent herb to 1 m tall; branches inconspicuously puberulous to glabrate. Principal leaves with blades ovate to broadly elliptic, 7-19 cm long, 3.5-9 cm wide, ca. 2 times as long as wide, apically acute to more typically acuminate, basally rounded to somewhat cuneate, marginally conspicuously ser- rate throughout with ca. 4-6(-9) teeth/cm each ca. 1-1.5 mm long, glabrous on the upper surface and densely strigillose along the veins beneath; petioles 0.8-4 cm long, sparingly to moderately but always inconspicuously puberulous. Flow- ers 3^ cm long at anthesis; pedicels solitary in the axils of the reduced (1.5^ cm long) upper leaves and thus racemous, strongly ascendent to wide spreading, glabrous, 3^ cm long, 1-2 mm in diameter, ebracteolate; hypanthium in anthesis cylindric to obconic, glabrous, basally acute to rounded, 4-5 mm long, 3-4 mm in diameter medially; calyx lobes deltoid to oblongish, acute to obtuse, denticu- late, glabrous, 2-3 mm long, 1.5-2 mm wide basally; corolla glabrous, green, the tube 1.5-2 cm long, 2-3 mm in diameter, the lobes linear to narrowly oblong, the upper lobes falcate, 12-16(-18) mm long, the lateral lobes 7-9 mm long, the lowermost lobe 5-6.5 mm long; filament tube glabrous to puberulous, 22-27 mm long, the anthers externally appressed-pilosulose, 4-6 mm long, the lowest an- thers slightly pilose-fringed. Berries globose, 5-6 mm long, 6-8 mm in diameter. This species is known only from the type collection made on the Cerro Sapo in Darien. It is named for the collector, Barry Hammel, who through his particular attention to the genus Bunneistera, has considerably advanced our understanding of the Panamanian representatives. X2. Burmeistera mcvaughii Wilbur, type: Panama, Chiriqui-Bocas del Toro, bor- der trail along continental divide ca. 5 mi NE of Boquete near Cerro Pate Macho above Palo Alto along trail above 6800 ft, Hammel 7399 (DUKE, holotype; MO, plus 6 specimens to be distributed, isotypes). Planta suffrulicosa; ramuli puberuli. Lamina foliorum elliptica, 6-12(-l6) cm longa, 3-5(-7) cm lata, 1.5-2(-3) plo longiora quam lata, acuta vel acuminata ad apicem el basi acuta vel rotunda, margine serrulata vel denticulata, 6-8 denticulatis per cm; petioli puberuli, 2-3 cm longi; folia juvenilia pinnatifida, 1-3 dissecta. Flores 3.2-4 cm longa; pedicelli puberuli, 4-7 cm longi. Hypanthium ob- conicum, puberulum, costatum, 5-7 mm longum, 3-5 mm latum; lobi calycis linear vel lanceolati, serrulati, puberuli, 6-8 mm longi, 1-1.2 mm lati. Corolla puberula, virenta; tubus 12-18 mm longus; lobi oblongi vel deltoidi, superiori 10-13 mm longi, laterali 6-7 mm longi, inferiori ca. 5-7 mm longi. Filamenta 22-27 mm longa, puberula; tubus anthearum puberuli. Bacca cylindrica, ca. 15 mm longa, 10-12 mm lata. Terrestrial, scandent shruhlet up to 2.5 m long; stems inconspicuously and minutely puberulous above with trichomes 0.1 mm long or less, glabrate below. Principal leaves elliptic, 6-12(-16) cm long, 3-5(-7) cm wide, mostly 1.5-2(-3) times as long as wide, apically acute to more typically acuminate, basally acute to rounded, marginally conspicuously serrulate or denticulate throughout with ca. 6-8 teeth/cm each ca. 0.2-0.4 mm long, inconspicuously and sparingly mi- croscopically puberulous on the veins above or glabrate, moderately and incon- 170 ANNALS OF- THE MISSOURI BOTANICAL GARDEN IVoi.. 68 spicuously microscopically puberulous on the veins beneath, the veins forming a conspicuous reticulum with a more or less distinct marginal vein ca. 1 mm from the edge; juvenile leaves greatly dissected, 1-2-pinnatifid; petioles mostly 2-3 cm long, sparingly puberulous to glabrate. Flowers 3.2^ cm long; pedicels solitary in the upper leaf axils, ascendent, straight to somewhat curving, puberulous, ebracteolate, 4-7 cm long; hypanthium at anthesis obconic, basally acute, mod- erately to densely but microscopically puberulent, the trichomes less than 0.1 mm long, slightly curved and mostly restricted to the 10 proximal vertical costae, 5-7 mm long, distally 3-5 mm in diameter; calyx lobes linear to narrowly lan- ceolate, acute, inconspicuously serrulate, ascendent, inconspicuously minutely puberulous on both surfaces with hyaline trichomes 0.1 mm long, 6-8(-9) mm long, medially 1-1.2 mm wide; corolla moderately to densely but inconspicuously puberulous with hyaline trichomes 0.1 mm long or less, pale green, the tube 12- 18 mm long, the lobes falcately oblong to deltoid, acute, the 2 upper lobes 10-13 mm long, the 2 lateral lobes 6-7 mm long, the lowermost lobe 5-7 mm long; filament tube exserted, 22-27 mm long, distally moderately spreading puberulous, the anthers 3-5 mm long, moderately to densely spreading puberulous, apically with the two shortest anthers with a short pilose fringe. Berries broadly cylindric, ca. 15 mm long, 10-12 mm in diameter. This species seemingly most closely resembles Burmeistera dendrophihi F. E. Wimmer and B. utleyi Wilbur which are also known only from the mountainous border area between the Panamanian provinces of Chiriqui and Bocas del Toro. Burmeistera mcvaughii differs from both of these species in its more broadly elliptical leaves and pronouncedly serrulate to denticulate margin and its narrowly lanceolate to linear, acute calyx lobes. Barry Hammel, who has had the most field experience with these species, has observed that B, mcvaughii occurs at somewhat higher elevations and in more exposed, mist-swept habitats. The species is named, as a small token of esteem, for Professor Rogers Mc- Vaugh who has recently retired from his role as Professor of Botany and Curator of Vascular Plants at the University of Michigan. Among his numerous botanical achievements, he has long been the leading student of the American Campanu- laceae; hence it is most fitting that his labors of more than forty years ago in producing the basic treatment of the lobeliads for the North American Flora be recalled by those of us who have been attracted to these plants not only by their beauty but by the availability of his eminently useful account. All wish him well in his efforts to complete the extremely ambitious Flora Novogaliciami upon which he has devoted so much effort during the past three decades. To complete it in a manner satisfying his own high standards will require a long, healthy and happy period of "retirement" and we all wish him that. The orthography of the specific epithet is not in accord with Recommendation 73C4 of the International Code; it is doubtful if Dr. McVaugh would approve such a rendition of his patro- nymic. Other Specimens Examined: Panama, chiriqui'-bocas del toro: In forest near top of continental divide ridge along trail towards Cerro Pate Macho, above Palo Alto, ca. 4 mi NE of Boquete, 6800 ft. Hummel 6490 (DUKE. MO). 1981] Wll.aVR^BURMEISTERA 171 X3. Burmeistera obtusifolia F. E. Wimmer This is the first collection of this species seen from Panama, Previously Bur- meistera obtusifolia was known only from the northern slopes of the Cordillera de Talamanca in Costa Rica and apparently also from the Volcan Turrialba, The presence of this species in western Panama is a range extension of approximately 290 km and is another lobeliad now reported from western Panama, all of which were previously unknown from any area closer than the central Talamancas of Costa Rica. These recent extensions include, in addition to B, obtusifolia , B. par- viflora F. E. Wimmer ex Standley, Centropogon gutierrezH (Planch. & Oerst.) F. E. Wimmer and C smithii F. E. Wimmer. There are differences in both vegetative and floral pubescence and in floral dimensions between the Costa Rican and Panamanian collections but their importance can properly only be evaluated after additional collections have been made. Specimens Examined: Panama. cmRigui-BOCAS del toro: Border trail along continental divide ca. 5 mi NE of Boquete near Cerro Pate Macho above Palo Alto, 7100 ft, Hammel 7370 (DUKE, MO, and specimens to be distributed). Literature Cited Wii BUR, R. L. 1976. Campanulaceae. /// R. E. Woodson, Jr. & R. W. Schery, Flora of Panama Ann. Missouri Bot. Card. 63: 593-655. SAPOTACEAE OF PANAMA 1 George E. Pilz^ Abstract The Flora of Panama Sapotaceae treatment by Blackwell is updated. Pouteria con^esiifolia Pilz and P. leptopedicellata Pilz are newly described and the combination Pouteria buenaventurensis (Aubr.) Pilz is made. As now revised, the Panamanian flora includes 44 species in 11 genera, an increase from the 24 species in 6 genera known to Blackwell. This revision is based on collections made in the past decade. Trees or shrubs with milky latex. Leaves alternate, rarely opposite; nodes mostly trilacunar, the traces evident on the leaf scar; blades simple, entire; pu- bescence of 2-armed hairs, one arm sometimes obsolete. Flowers solitary or clustered in axils or at recently defoliated nodes, rarely cauliflorous, perfect, rarely unisexual, actinomorphic; sepals in one or two whorls, rarely spiral, free or rarely united; corolla gamopetalous, the lobes simple or divided, usually as many as sepals; stamens generally as many as the corolla lobes and opposite them, the staminodes sometimes present between the corolla lobes, variously developed; pistil syncarpous, the style simple, sometimes obscurely lobed at the summit, the ovary superior, 1-14-loculed, the carpels uniovulate. Fruit a berry; exocarp usually fleshy, often becoming sclerotic; seeds 1-several, the testa hard, smooth, often shiny, the attachment area (scar) lateral or basal, variously devel- oped, often rough and duller in color, the endosperm either copious on either side of the flat foliaceous cotyledons, or scanty, or absent with thick fleshy cot- yledons. Scarcely 12 years have elapsed since Blackwell (1968) published the treatment of Sapotaceae for the Flora oj Panama. Blackwell treated 24 species in 6 genera, while the present study treats 44 species in 1 1 genera. Recent collections, partic- ularly from moist forest areas, have significantly increased the number of taxa known from Panama, and they have also permited the first descriptions of flowers and/or fruits for several species. Still, all too many species have inadequate series of specimens, and collectors can greatly assist future monographers by seeking flowers and fruits of all but the most common cultivated species. Determination of geographical distribution and morphological variation of most species awaits further collection of specimens. Collectors can also make valuable contributions by preserving viable seeds and/or developing flowers for cytological study. Sapotaceae are almost complete- ly unknown cytologically, particularly American species. Patterns seen in a few chromosome counts that have been made of Old World taxa look interesting, and a more complete record might prove valuable in determining the natural affinities of certain species. ^ Assisted by National Science Foundation Grant DEB 77-04300 (W. G. D'Arcy, principal in vestigator). This work was initiated while the author held a National Endowment for the Arts Post doctoral Curatorial Traineeship at the Missouri Botanical Garden. Their financial support and en couragement is gratefully acknowledged. * The Polytechnic, Department of Biology, Ibadan, Nigeria. Ann. Missouri Box. Card. 68: 172-203. 1981. 0026-6493/81/0172 -0203/$3.35/0 1981] F^ILZ— SAPOTACEAE 173 Full descriptions are included in this treatment only for those taxa not treated by Blackwell (1968). Likewise, complete lists of synonyms are not included. The nomenclature used for American Sapotaceae has been in considerable flux, and it is by no means stabilized even now. Over 200 names have been proposed for the 44 species treated in this study. Nearly half of the later synonyms are no- menclatural, reflecting the widely different concept of genera held by various authorities. More complete lists of synonyms may be found in papers cited after the generic descriptions. As a taxonomic result of this study I am describing two new species of Pou- teria and transferring RichardcUa huenaventurensis to Pouteria. Literature: Aubreville, A. 1965. Sapotacees. Adansonia, mem. L 1-157. . 1972. Sapotaceae. /// The botany of the Guayana Highland, Part IX. Mem Memoire 11: 1-262. W. H, 1968. Sapotaceae. In Flora of Panama. Ann. Missouri Bot Card. 55: 145-169. Williams. 1967. Sapotaceae. /// Flora of Guatemala. Fieldiana, Bot. 24(3): 211-244. a. Calyx biseriate, with two distinct series of 3/3 or 4/4 persistent sepals. b. Calyx of 3/3 sepals; fertile stamens 6; seed scar basilateral and several times longer than broad 5, Manilkara bb. Calyx of 4/4 sepals; fertile stamens 8; seed scar nearly basal and suborbicular 8. Mimusops aa. Calyx with a single cycle or series of sepals, if distinctly biseriate then 2/2. c. Flowers with well-developed staminodes. d. Seed scar small, basal or basilateral, never reaching the middle of the seed; stamens inserted at top o\^ the throat in the sinuses between the corolla lobes: corolla lobes with lateral appendages or without appendages. e. Corolla lobes with lateral appendages; petiole 0.2-2 cm long. f. Seed with endosperm: plants unarmed: ovary usually glabrous; style 1.2-2 mm long: apex of the fruit tapering to a persistent style 3. Dipholis ff. Seed without endosperm; plants often with spines; ovary usually pubescent; style 3-7 mm long: apex of the fruit rounded or truncate 1. Bumel'ui ee. Corolla lobes without appendages: petioles 3-9 cm long, usually V2 the length of the blade or more 6. Mastichodendron dd. Seed scar lateral, usually extending past the middle of the seed; stamen insertion various; corolla without dorsal or lateral appendages, g. Leaves striate, the lateral veins usually less than 1 mm apart; seeds with some endosperm 7. MicrophoUs gg. Leaves with 8-20(-50) lateral veins, usually 4 mm or more apart: seed without endosperm. h. Sepals united for V2 their length or more 11- Synsepaliou hh. Sepals essentially free, imbricate or quincuncial 10. Pouteria cc. Flowers with staminodes irregularly present or rudimentary or lacking. i. Seed without endosperm. j. Seed scar lateral, linear, 3 mm or less in width; corolla lobes generally as long as the tube. 174 ANNALS OF THE MISSOURI BOl ANICAL GARDEN [Voi . 68 k. Ovary 1-3-loculed; leaves often glaucous, the lateral veins 13-18 pairs, weak or obscure beneath 9. NeoAytliece kk. Ovary (4-)5(-6)-loculed; leaves not glaucous, the lateral veins 7-11 pairs, well developed 10. Poutcria ij. Seed scar lateral, covering Vi of seed, elliptic or oblong, 1 cm or more wide; corolla lobes more than twice as long as the tube 4. Ehieohima ii. Seed with endosperm 2. ChiysophyUuin 1. BUMELIA Bumelia Swartz, Prodr. Veg. Ind. Occ. 49. 1788, nom. cons, type: B. retusa Swartz. Shrubs or trees usually armed with spines. Leaves alternate, exstipulate; blades firm. Flowers subsessile to pedicellate; sepals 5, uniseriate; corolla lobes 5, each lobe entire or divided into 3 segments; stamens epipetalous, the filaments attached at the level of the sinuses, the staminodes alternating with the stamens, petaloid, entire to erose or lacinate; ovary 5-loculed. Fruit baccate, rounded to retuse at the apex; seeds solitary, the seed scar subbasal, small, scarcely longer than broad, the endosperm absent, the cotyledons fleshy. Bumelia is a genus of about 25 species found in tropical and warm America. Literature: Cronquist, A. 1945. Studies in the Sapotaceae, III. Dipholis and Bumelia. J. Arnold Arbor. 26: 435-471 {Bumelia, 445-471). 1. Bumelia persimilis Hemsley subsp. persimilis. Biol. Centr.-Amer., Bot. 2: 298. 1882. type; Mexico, Vera Cruz, Orizaba, Botleri 989 (F, US, isotypes). Shrub or tree to 18 m. Leaves 3-12 cm long, 1-5 cm wide, usually elliptic, acute to obtuse; primary lateral veins 10-30 pairs; petiole 2-10 mm. Flowers several to numerous in the axils; pedicels 3-6 mm long; sepals (4-)5(-6), 1.8-3.7 mm long; corolla 4.5-6 mm long, the lobes 5, twice as long as the tube; filaments attached at the level of the sinuses; ovary 5-locuIed. Fruit fleshy, smooth, 1.2- 2.5 cm long, 1-2 cm wide, 1-seeded; seed scar smaU, subbasal (Blackwell, 1968). This variety ranges from Mexico through Central America to Venezuela. coci e: El Valle de Anton along Rio Indio trail. Hunter <.i Allen 298 (EAP, F). i.os santos: Los Santos, Lao 321 (MO). Panama: Chepo, Klm^e 12 (US). 2. CHRYSOPHYLLUM Chrysophyllum L., Sp. PI. 192. 1753. type: C. cainito L. Trees or shrubs lacking spines. Leaves alternate, exstipulate; blades firm to coriaceous, the secondary lateral veins often parallel to the primary series. Flow- ers pedicellate; sepals 5, uniseriate; corolla lobes 5, lacking appendages; stamens epipetalous, the filaments variously attached from near the base of the corolla tube to the level of the sinuses, the staminodes absent or (rarely in individual flowers) I or more irregularly developed in the corolla sinuses; ovary 4-12-loc- 1981] PILZ— SAPOTACEAE 175 uled. Fruit baccate, often edible; seeds 1 -several, the seed scar large, lateral or basilateral, variously developed, linear to covering nearly half the seed surface, the endosperm copious, the cotyledons thin and foliaceous. Chrysophyllum is a pantropical genus of about 80 species. Literature: Aubrevilie, A. 1961. Notes sur les Sapotacees Africaines et Sud-Americaines. Adansonia, n.s., 1: 6-38 {Chrysophyllum, 9-13). Cronquist, A. 1945. Studies in the Sapotaceae, I. The North American species of ChrysophyUunu Bull. Torrey Bot. Club 72: 191-204. . 1946. Studies in the Sapotaceae, V. The South American species of ChrysophyUunu Bull, Torrey Bot. Club 73: 286-311. a. Corolla lobes equal lo or longer than the lube: sepals to 1 mm long. b. Seed scar lateral, nearly as long as the seed, 4-7 mm wide; stigma lobes mostly 7- 12; fruit 4-10-seeded 1. C. cainito bb. Seed scar basilateral, nearly as broad as long, not reaching much past middle of the seed or shorter; stigma lobes (4-)5(-6); fruit l-seeded 4. C. tnexkanum aa. Corolla lobes generally less than half as long as the tube: sepals 1-4 mm long. c. Sepals 1-2 mm long; seed scar basilateral, not reaching much past middle of seed, usually shorter 5. C. panamense cc. Sepals 3-4 mm long. d. Staminal filaments attached near the middle of the corolla tube; pedicel to 5 mm long; petiole 1-3 cm long; seed scar lateral, nearly entire length of seed 2. C. excelsum dd. Staminal filaments attached to the top of the corolla tube; pedicel 4-20 mm long; petiole 1 cm or less in length; seed scar presumably basilateral, reaching only middle of seed 3. C. hirsutiini 1. Chrysophyllum cainito L., Sp. PI. 192. 1753. Tree to 30 m. Leaves 10-16 cm long, 5-8 cm wide, elliptic to oblong, short- acuminate; primary lateral veins 10-18 pairs; petioles 1-2.5 cm long. Flowers numerous in the axils; pedicels 5-16 mm long; sepals (4-)5(-6), ca. 1 mm long; corolla 3-5 mm long, the lobes (4-)5(-6), as long as the tube or slightly longer; filaments attached at the level of the sinuses; ovary 4-12-loculed. Fruit fleshy, subglobose, 3-10 cm broad, 4-10-seeded; seed scar lateral, extending nearly as long as the seed, elliptic, 4-7 mm broad (Blackwell, 1968). ChrysophyUum cainito ranges over Mexico through Central America to north- ern South America and from Florida through the West Indies. It is widely culti- vated and is probably indigenous only to the West Indies. BOCAS DEL TORO: Changuinola Valley, Dunlap 24 (F). Almiranle, road to Chiriqui, McDaniel 5074 (MO). Chiriqui Lagoon, Wedel 2523 (MO), canal zone: Ancon Hill, Allen 2672 (EAP). Madden Dam and along Azote Caballo Road near Alajuela, Docl^'e 16573 (MO, UC). Upper Chilibre River, 1/2-1 mi below Chilibre, Seihert 1505 (MO), chiriqui: E of Gualaca, Allen 5033 (EAP, MO). Progreso, Cooper & Slater 247 (F), 264 (F). Burica Peninsula, San Bartolo Limite, 12 mi W of Puerto Armuelles, Croat 22168 (MO), cocle: Floor of El Valle de Anton, .4//^// 2747 (EAP, F). darien: Rio Sabana above Santa Fe, Duke 14104 (MO). Campamento Buena Vista, Rio Chucunaque above confluence with Rio Tuquesa, Stern et al. 853 (MO, UC). herrera: Ocu, Allen 3647 (EAP, MO), los santos: Tonosi, Duke 12488 (MO). Panama: Panamerican Highway ca. halfway between El Llano and Rio Momoni, Duke 5527 (MO). Chepo, Klu^i^e 49 (F). Taboga Island, Woodson et al 1537 (,M0). san 176 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 BLAs: Mainland opposite Playon Chico, Gentry 6J97 (MO), veraguas: Cerro Tute, region W of Santa Fe, Allen 4441 (EAP, MO). Coiba Island, D^^yer 2331 (MO). El Embulsadero, 8 mi W of Santiago. Tyson 6075 (MO). 2. Chrysophyilum excelsum Huber, Bol. Mus. Paraense Hist. Nat. 3: 55. 1902. type: Brazil, specimens collected from a cultivated tree at Para, Huher 3016 (G, not seen; F, US, isotypes). Tree to 30 m. Leaves with petioles 1-4 cm long; blades elliptic to elliptic- obovate, thin to firm, rounded to mucronate, 6-30 cm long, 4-12 cm wide, gla- brate; primary lateral veins 9-14 pairs, the secondary and tertiary ones forming an evident reticulum. Flowers 5-10 at recently defoliated nodes; pedicels 3-5 mm long, strigose; sepals 5, ovate, 3-3.5 mm long, 3-4 mm wide, rufous sericeous- strigose; corolla 3.8^.3 mm long, the lobes 5, fringed-ciliolate, ca. 14 as long as the tube; filaments attached near the middle of the tube; ovary 5-loculed, the style 0.5-2.5 mm long, glabrous. Fruit yellow, globose, 3.5-5 cm thick; seeds 4- 5, ovoid, about 25 mm long, 13-14 mm wide, 7 mm thick, the seed scar lateral, as long as the seed, ovate, 2-5 mm broad. This species occurs in the upper Amazon Basin, Brazil, Venezuela, and Pan- ama. Leaves of the Panamanian material are generally thicker and smaller than those found on South American plants, but the venation pattern is very similar and otherwise the plants are not distinctive. Panama: Near archeological site at edge of Madden Lake, Gentry & Tyson 5040 (MO). 3. Chrysophyilum hirsutum Cronquist, Bull. Torrey Bot. Club 72: 198. 1945. type: Costa Rica, Alajuela, Villa Quesada, San Carlos, edge of forest, Austin Smith 1776 (A, not seen; F, isotype). Shruh or tree to 10 m tall. Leaves with petioles to 1 cm long; blades generally elliptic, thin, rounded to abruptly acuminate, 6-11 cm long, 2.5-5 cm wide, sparsely rufous-hirsutulose, glabrate; primary lateral veins 10-15 pairs, the sec- ondary and tertiary ones inconspicuous. Flowers 2-6 in the axils; pedicels 4-20 mm long, sericeous-strigose; sepals 5, acute to rounded, 2-4 mm long, 1.5-2.5 mm wide, densely rufous pubescent; corolla greenish white, 4.5-7 mm long, rufous pubescent on the back of the lobes and distal portion of the tube, the lobes (4-)5, V3~V2 as long as tube; filaments to 0.5 mm long, attached at the level of the sinuses; ovary densely rufous pubescent, 5-loculed. Fruit little known, but ap- parently 1 -seeded and over 2 cm long. This species occurs in Costa Rica and Panama. 600 The pedicels of Panamanian material are two to three times longer than those rwi coion: Agua Clara rainfall station, Santa Rita Ridge, Foster 1739 (MO). Santa Rita Ridge, Gentry & Dwyer 4800 (MO). Panama: Cerro Azul, Lao & Holdridge 34 (MO). 1981] PILZ— SAPOTACEAE 177 4. Chrysophylium mexicanum Brand, ex Standley, Contr. U.S. Natl. Herb. 23: 1114. 1924. type: Mexico, Veracruz, Zacuapan, Purpus 7679 (US, holotype; MO, NY, UC, isotypes). Shrub or tree to 20 m tall. Leaves with petioles 4-10 mm long; blades generally elliptic, thin, acuminate to acute, 5-14 cm long, 2-6 cm wide, rufous to pale pubescent, glabrate; primary lateral veins 9-14(-20) pairs, prominent to very weak, the secondary and tertiary ones weak and often obscure. Flowers 1-15 (-20) in the axils; pedicels 2-5 mm long, pale to rufous pubescent; sepals 5, ovate, ca. 1 mm long, 1 mm wide, rufous or pale pubescent; corolla greenish white, 2- 2.5(-3.5) mm long, glabrous, the lobes (4-)5, as long as the tube or slightly longer; filaments 0.1-0.5 mm long, attached near the level of the sinuses; ovary rufous pubescent, 5-6-loculed, the stigma 0.1-0.4 mm long, 5-6-lobed. Fruit ellipsoid, 1.5-2 cm long, 1 cm thick, smooth, glabrate; seeds solitary, ovoid, ca. 12 mm long, 7-8 mm thick, the seed scar basilateral, extending to about the middle of the seed, 3-4 mm broad. This species ranges from Mexico through Central America to Colombia. Chrysophylium mexicanum is highly variable as to leaf size, pubescence, and degree of prominence of the lateral veins, but these differences are not sufficiently correlated to justify the recognition of subspecies or varieties. cocle: El Valle de Anton, Croat 25282 (MO); Lao 286 (MO). Behind Club Campestre, Duke 13266 (MO), darien: Hills near Pidiaque, Duke 8047 (MO). Rio Sabana above Santa Fe, Duke 14108 (MO). 5. Chrysophylium panamense Pittier, Contr. U.S. Natl. Herb. 18: 165, 1916. type: Panama, Pittier 4005 (US, holotype; NY, isotype). Tree to 15 m. Leaves 15-25 cm long, 6-11 cm wide, elliptic to elliptic-obovate, acuminate; primary lateral veins 10-20 pairs; petioles 1-2.5 cm long. Flowers several to numerous in the axils; pedicels 3-7 mm long; sepals (4-)5(-6), 1-2 mm long; corolla 3.5-6 mm long, the lobes {4-)5(-6), less than Vi the length of the tube; filaments attached at the level of the sinuses; ovary 4-12-loculed. Fruit fleshy, subglobose to ovoid, to 2 cm broad, 3-8-seeded; seed scar basilateral, extending to the middle of the seed or slightly beyond (Blackwell, 1968). This species occurs in Costa Rica, Panama, and Venezuela (Amazonas). BOCAS DEL TORO: Almirantc, Cooper 353 (F). Without locality, Wedel 230 (F, MO), canal zone: Barro Colorado Island, Aviles 969 (F); Bailey & Bailey 397 (F); Croat 8012 (MO); Ebinger 210 (MO); Gangham 591 (F); Haxes sm. (F, MO); Shattuck 778 (F), 969 (MO), 1024 (F, MO); Starry 82 (F), 118 (F); ^Netmore & Abbe 169 (F); Zetek 3810 (EAP, F, MO), 4327 (F), 4330 (F). Sirri River, Trinidad Basin, Pittier 4005 (NY, US), colon; Between Salud and Boca del Rio Indio, Howell 28 (MO). 2- 3 mi up Rio Guanche, Kennedy & Foster 2153a (MO). Rio Viejo, 4 km NE of Puerto Pilon, Nee 7178 (MO), darien: La Boca de Pirre, Bristan 1271 (MO). Panama: El Llano-Carti Road, 2.3 km N of Panamerican Highway, Nee & Dwyer 9239 (MO). 3. DIPHOLIS ifi A. DC. 178 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Trees or shrubs lacking spines or thorns. Leaves alternate, exstipulate; blades coriaceous. Flowers small, subsessile to pedicellate; sepals 5, uniseriate; corolla lobes 5, each lobe divided into 3 segments; stamens epipetalous, the filaments attached at the level of the sinuses, the staminodes alternating with the stamens, petaloid, often erose, fimbriate or lacinate; ovary 5-loculed. Fruit baccate, abruptly tapering to a short persistent style; seeds solitary, the seed scar basal, rarely basilateral, small, scarcely longer than broad, the endosperm copious, the cotyledons thin and foliaceous. Dipholis is a tropical North American genus of about 15 species found prin- cipally in the Greater Antilles. Literature: Cronquist, A. 1945. Studies in the Sapotaceae, 111. Dipholis and Bumelia, J Arnold Arbor. 26: 435-471 {Dipholis. 435-445), 1. Dipholis minutiflora Pittier, Contr. U.S. Natl. Herb. 13: 464. 1912. type: Costa Rica, Dota Mts., forests of El Copey, Tonduz 11935 (US, holotype; NY, isotype). Tree to 30 m. Leaves 5-20 cm long, 2-10 cm wide, elliptic-oblanceolate to elliptic-obovate, rounded to acute; primary lateral veins 10-20 pairs; petiole 5-18 mm long. Flowers numerous at defoliated nodes; pedicels 3-10 mm long; sepals 5(-8), 1.5-2.2 mm long; corolla 4^ tube; filaments attached at the level of the sinuses; ovary 5-loculed. Fruit 1.5- 2.5 cm long, 1-2 cm wide, the flesh scant, wrinkled when dry; seed solitary, the seed scar subbasal, broadly elliptic, 3-5 mm long (Blackwell, 1968). This species ranges from Mexico to Panama. CHiRiQui: Cerro Punta, Allen 1564 (EAP, F, MO). Quebrada Velo, Finca Lerida, Allen 4676 (EAP, MO). Valley of the upper Rio Chiriqui Viejo, White I09h (F, MO); White & White I (MO). Bajo Mono, mouth of Quebrada Chiquero, along Rio Caldera, Woodson et al. 995 (F, MO). Panama: Hills above Campana, Allen 1314 (EAP, F, MO). Cerro Jefe, Gentry 4867 (MO). 4. ELAEOLUMA Elaeoluma Baillon, Hist, PI. 11; 293. 1892. type: E, schoniburi^kiana (Miq.) Bail- Ion. Shrubs or small trees lacking spines. Leaves alternate, exstipulate; blades firm to coriaceous, the primary lateral veins generally few. Flowers short pedi- cellate; sepals 5, uniseriate; corolla lobes 5, lacking appendages; stamens epi- petalous, the filaments attached to a very short corolla tube, the staminodes alternating with the stamens, rudimentary, the number irregular, often completely absent; ovary (2-)3(-"5)-loculed. Fruit baccate; seeds solitary, the seed scar lat- eral, linear to very broad, the endosperm absent, the cotyledons fleshy. Elaeoluma is a genus of 3 species found chiefly in northern South America. 1981] PILZ— SAFK)TACEAH 179 1. Elaeoluma glabrescens (Mart. & Eichl.) Aubreville, Adansonia, n.s., 1: 26. 1961. Lucunia glabrescens Mart. & Eichl. in Mart., Fl. Bras. 7: 72. 1863. type: Brazil, Rio Negro between Barcellos and San Gabriel, Spruce 2029 (photo MO; NY, P, isotypes). Tree of unknown height. Leaves with petioles 10-15 mm long; blades ovate- elliptic to oblong-elliptic, firm to coriaceous, acuminate, 7-15 cm long, 3-7 cm wide, glabrous; primary lateral veins 9-12 pairs, slender, obscure above, slightly raised and evident beneath, the secondary and tertiary ones closely anastomosing. Flowers 5-10 in the axils or at recently defoliated nodes; pedicels 3-5 mm long; sepals 5(-6), ovate, 2-6 mm long, glabrate; corolla glabrous, 5-6 mm long, the lobes 5, elliptic, 3-5 times longer than the tube; filaments to 1.5 mm long, attached to the very short tube, the staminodes not observed; ovary pilose to minutely tomentose, (2-)3(-5)-loculed, the style 1-2 mm long, glabrous. Fruit obovoid to subglobose, smooth, glabrous, 2-4.5 cm long, 1-3 cm wide; seeds solitary, ovoid to obovoid, to 3 cm long, 2.5 cm wide, the seed scar covering about half the seed, ventral, extending the entire length. This species occurs in Amazonian Brazil, Peru, and Panama. Elaeoluma glabrescens is closely related to species of Neoxythece, but it is readily distinguished by its broad, ventral seed scar. The number of ovary locules has been used to separate these genera, but I have found it to be too variable to be useful in identifying specimens. SAN FU as: Chucunaque, 2-10 mi above the Cuna-Darien boundary, Duke 8565 (MO). 5. MANILKARA Manilkara Adanson, Fam. PI. 2: 166. 1763, nom. cons, type: M. kauki (L.) Dubard. Trees lacking spines. Leaves alternate, exstipulate; blades coriaceous; pri- mary lateral veins numerous, fine. Flowers long-pedicellate; sepals 6, biseriate; corolla lobes 6, each lobe often divided into 3 segments; stamens epipetalous, the filaments attached at the level of the sinuses, the staminodes alternating with the stamens, petaloid, rarely replaced by functional stamens; ovary 6-16-loculed. Fruit baccate; seeds 1-several, ovoid to laterally compressed, the seed scar bas- ilateral, oval to linear, the endosperm copious, the cotyledons thin and foliaceous. Manilkara is a pantropical genus of about 50 species. Literature: 4 Cronquist, A, 1945. Studies in the Sapotaceae, IV. The North American species of Manilkara. Bull. Torrey Bot. Club 72: 550-562. a. Corolla lobes with broad dorsal appendages, appearing to be 18 in number. b. Flowers 3-12 per axillary fascicle; pedicel glabrous; fruit smooth or slightly rough- ened; seed solitary, the scar 5 mm broad, basilateral, barely reaching middle of seed 1. M . hidentuld 180 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 bb. Flowers solitary in the leaf axils; pedicel tomentulose to glabrate; fruit mealy rough- ened; seeds several, the scar 2 mm broad, linear, from near the base to well beyond the middle of the seed 3. M. meridionalis aa. Corolla lobes entire or merely tridentate at the apex, appearing to be 6 in number. c. Flowers 2-5 per axillary fascicle; pedicel strigillose; corolla tube less than half as long as the lobes; seed scar basilateral, linear, barely reaching the middle of the seed 2. M. chicle cc. Flowers solitary in the leaf axils; pedicel rufous-tomentulose; corolla tube more than half as long as the lobes; seed scar linear, from near the base to well beyond the middle of the seed 4. W. zxipota 1, Manilkara bidentata (A. DC.) Chevalier, Rev. Int. Bot. Appl. Arg. Trop. 12: 270. 1932. Mimusops bidentata A. DC. in DC, Prodr. 8: 204. 1844. type: French Guiana, Richard (not seen). Tree to 35 m. Leaves 6-30 cm long, 3-12 cm wide, obovate to elliptic, obtuse to emarginate; petioles 1.5-4 cm long. Flowers 3-12 in the axils or at recently defoliated nodes; pedicels 1.2-3 cm long; sepals 4-6 mm long; corolla 5-7 mm long, the tube ca. V5 of the total length, the lobes with 2 dorsal appendages as broad as and slightly longer than the lobes; ovary 6-10-loculed. Fruit smooth or slightly roughened, subglobose, 2-3.5 cm long, 1-seeded; seed scar basilateral, barely reaching the middle of the seed, ca. 5 mm broad (Blackwell, 1968). This species occurs in Hispaniola, Puerto Rico, Lesser Antilles, northern South America and Panama. BOCAS DEL TORo: Ri'o San Pedro, between the Rio Cana and Rio Calovebora, Gordon 13 (MO). Without locahty, Co.\ ,\.n. (US), canal zone: Hills around Gatiin, Pittier 2699 (US). Panama: Chepo, Khi^e 55 (EAP, US), san blas: Hills near Puerto Obaldia, Pittier 4318 (NY, US), 4384 (US). 2. Manilkara chicle (Pittier) Gilly, Trop. Woods 73: 14. 1943. Achras chicle Pittier, J. Wash. Acad. Sci. 9: 436. 1919. type: Guatemala, Izabal, Vega Grande, near Los Amates, Pittier 8537 (US, holotype; NY. isotype). Tree to 40 m. Leaves 8-26 cm long, 3-8 cm wide, oblanceolate to elliptic, rounded to acuminate; petioles 1-3.5 cm long. Flowers 2-5 in the axils; pedicels 0.5-3 cm long; sepals 5-9 mm long; corolla 5.5-9 mm long, the tube V6-^/i of the total length, the lobes without dorsal appendages; ovary 7-9-loculed. Fruit mealy roughened, subglobose, 2-4 cm long, 4-6-seeded; seed scar basilateral, barely reaching middle of the seed, narrow, 2 mm wide (Blackwell, 1968). This species ranges from Mexico through Central America to Colombia. CANAL zone: Ancon Hill, Standley 26384 (US), cocle: La Pintada, Leon 22 (MO), colon: Santa Rita Ridge, Gentry 6566 (MO). Santa Rita Ridge Road, 17 km from Boyd-Roosevelt Highway, Mori & Kallunki503I (MO), darien: Pinas, Duke 10653 (MO). Patino, Pittier 5698 (VS), LOS Santos: Ca. Vi mi S of Pedasi, Stimson 5294 (MO, UC). Panama: Trapiche Island, Allen 2607 (EAP). San Jose Island, Johnston 342 (MO), 753 (MO, US). Cerro Penon, 3 km S of Alcalde Diaz, Nee 8867 (MO). Around Alhajuela, Chagres Valley, Pittier 3457 (US). 3. Manilkara meridionalis Gilly, Trop. Woods 73: 12. 1943. type: Costa Rica, Punta Arenas, Esparta, Biolley 17308 (US, holotype; NY, isotype). 1981] PILZ SAF'OTACEAE 181 Tree to 30 m. Leaves 5-13 cm long, 1.5-4.5 cm wide, elliptic to narrowly obovate, rounded to acuminate; petioles 1-2.7 cm long. Flowers solitary in the axils; pedicels 1-2.3 cm long; sepals 6-9 mm long; corolla 7-10 mm long, the tube '/i-% of the total length, the lobes with 2 dorsal appendages broader than but about equal in length to the lobes; ovary 9-12-loculed. Fruit mealy roughened, subglobose, 3-3.5 cm long, several seeded; seed scar basilateral, extending past the middle of the seed, linear, 2 mm wide (Blackwell, 1968). This species ranges from Mexico through Central America to Colombia and Venezuela. Panama: Taboga Island, StuiuUey 27099 (US); Woodson et al. 1455 (MO, NYj. Without locality, Hayes 793 (NY). 4. Manilkara zapota (L.) van Royen, Blumea 7: 410. 1953. Achias zapota L., Sp. PI. 2: 1190. 1753, based on Sapota fnictii ovato, majori Plumier. Sapola fritctu ovato, majori Plumier, Nov. PI. Amer. Gen. 43, tab. 4. 1703. type: Piumier's plate, for discussion see Moore & Stearn, Taxon 16: 382-395, 1967. Tree to 40 m. Leaves 4-15 cm long, 1.5-6 cm wide, elliptic to oblong-elliptic, rounded to acuminate; petioles 0.8-3 cm long. Flowers solitary in the axils; ped- icels 1.2-2.5 cm long; sepals 6-10 mm long; corolla 6-11 mm long, the tube Vi- % of the total length, the lobes without dorsal appendages; ovary 10-12-loculed. Fruit mealy roughened, subglobose, to 10 cm in diameter, (4-)8-12-seeded; seed scar basilateral, extending well past the middle of the seed, linear, 2 mm wide (Blackwell, 1968). This species ranges from Mexico through Central America to northern South America and from Florida through the West Indies. It is widely cultivated and many collections represent introduced plants. CANAL zone: Summit Gardens, Croat 6760 (MO); Mori & Kallunki 1875. 4532 (both MO). Balboa, Standley 27121 (MO, US), 30860 (US). 6. MASTICHODENDRON Mastichodendron Cronquist, Lloydia9: 245. 1946. type: M.foetidissimum (Jacq.) Cronquist. Trees lacking spines. Leaves alternate to subopposite, exstipulate; blades firm, the primary lateral veins few, curved. Flowers pedicellate; sepals 5, uni- seriate; corolla lobes 5, lacking appendages; stamens epipetalous, the filaments attached near the level of the sinuses, the staminodes alternating with the sta- mens, ovate to deltoid or lanceolate, not petaloid; ovary 5-loculed. Fruit baccate; seeds solitary, the seed scar basilateral, circular to lanceolate, not extending past the middle of the seed, the endosperm copious, the cotyledons thin and folia- ceous. Mastichodendron is a tropical North American genus of about 6 species. 182 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Literature: Cronquist, A. 1946. Studies in the Sapotaceae, II. Survey of the North American genera. Lloydia 9: 241-292 (Mastichodendron, 244-252). 1. Mastichodendron capiri (A. DC.) Cronquist var. tempisque (Pittier) Cronquist, Lloydia 9: 250. 1946. Sideroxyhm leinpisque Pittier. Contr. U.S. Natl. Herb. 13: 461. 1912. type: El Salvador. La Laguna de Santa Tecla, Pittier 1917 (US). Tree to 25 m. Leaves 6-15 cm long, 3-7 cm wide, generally elliptic, acuminate to rounded (emarginate); primary lateral veins 8-15 pairs; petioles 3-9 cm long. Flowers numerous at defoliated nodes; pedicels 3-8 mm long; sepals 5, 1.5-3.2 mm long; corolla 5-8 mm long, the lobes 5, twice as long as the tube; filaments attached at the level of the sinuses; ovary 5-loculed. Fruit yellow, 2.5^ cm long, 2-3 cm wide, 1-seeded, the flesh scant, wrinkled when dry; seed scar basilateral, lanceolate to subglobose, 6-9 mm long, 4-6 mm wide (Blackwell, 1968). This species ranges from Mexico through Central America to Panama. cocle: Penonome. Williams 421 (F, NY, US), los santos: Corozal de Macaracas, Lao ')88 (MO). 7. MICROPHOLIS Micropholis Pierre, Not. Bot. Sapot. 37. 1891. type: M. rugosa (Swartz) Pierre. Trees or shrubs lacking spines. Leaves alternate, exstipulate; blades firm, appearing striate, the primary lateral veins numerous, fine, crowded, nearly straight. Flowers pedicellate; sepals 5, uniseriate; corolla lobes 5, lacking ap- pendages; stamens epipetalous, the filaments attached near the level of the si- nuses, the staminodes alternating with the stamens, lanceolate to subpetaloid; ovary 5-loculed. Fruit baccate; seeds 1-several, the seed scar lateral, linear, the endosperm present but not copious, the cotyledons fleshy. Micronholis is a ironical American pf^nnti nf s^ihnnt ^n Qt^^nVc Literature: Cronquist, A. 1946. Studies in the Sapotaceae, II. Survey of the North American genera. Lloydia 9: 241-292 {Micropholis, 252-257). Micropholis Matuda Tree to 20 m. Leaves with petioles 8-10 mm long, highly canaliculate; blades narrowly elliptic to elliptic-obovate, thin, abruptly acuminate, 10-15 cm long, 3- 5 cm wide, essentially glabrous; primary lateral veins very numerous and fine, 0.2-0.3 mm apart. Flowers 5-10(-30) in the axils or at recently defoliated nodes; pedicels 4-8 mm long, rufous puberulent; sepals 5(-6), broadly ovate, 2.5-3.5 mm long, 2.5 mm wide, rufous pubescent abaxially, glabrous on the inner surface; 1981J ]>ILZ— SAPOTACEAh: 183 corolla white, glabrous, 5-6.5 mm long, the lobes 5, ovate, as long as or slightly longer than the tube; filaments 1-1.8 mm long, attached at the level of the sinuses, the staminodes 5, lance-subulate, 1.5-3 mm long; ovary rufous pubescent, 5- loculed, the style 5-6 mm long, glabrous. Fruit broadly ellipsoidal to pyriform, 2-3.5 cm long; seeds solitary, ellipsoidal, laterally compressed, 20-25 mm long, the seed scar linear, extending the length of the seed, to 5 mm wide. This species occurs in Mexico, Belize and Panama. The disjunct distribution of Micropholis mexicana is noteworthy, and Stand- ley & Williams (Fieldiana, Bot. 24: 232. 1967) mention an apparently undescribed Panamanian species. However, I find no specific differences in the material from Panama. PANAMA: Cerro Jefe, Foster & Kennedy 1871 (MO). El Llano-Carti road, 12 km from Panamer- ican Highway, Mori et ul. 4668 (MO). 8. MIMUSOPS Mimusops L., Sp. PI. 349. 1753. type: M. elengi L. Trees lacking spines. Leaves alternate; stipules caducous; blades glabrous or soon glabrate. Flowers pedicellate; sepals 8, biseriate; corolla lobes 8, each lobe divided into 3 segments; stamens epipetalous, the filaments attached at the level of the sinuses, the staminodes alternating with the stamens, usually simple, lan- ceolate or ligulate, densely pilose abaxially and along the margins; ovary usually 8-loculed. Fruit baccate; seeds 1-several, ovoid to laterally compressed, the seed scar basal to basilateral, scarcely longer than broad, the endosperm copious, the cotyledons thin and foliaceous. A genus of about 40 species found throughout the Old World tropics. Two species are cultivated in Panama. a. Pedicels 40-70 mm long; leaves thick, the petioles 1-1.5 cm long, the primary lateral veins 8-10 pairs 1. M. co/umersonii aa. Pedicels 7-8 mm long; leaves thin, the petioles 2-2.6 cm long, the primary laterals 15-25 pairs 2. M. eleni^i 1. Mimusops 1904. Imhricaria conunersonii G. Don, Gen. Syst. 4: 35. 1838. type: Madagascar, Conimerson (P). Tree to 20 m. Leaves with petioles 1-1.5 cm long; blades obovate to elliptic, thick, rounded to emarginate, (7-)10-13(-20) cm long, (4-)6-8(-ll) cm wide, densely rufous pubescent, soon glabrate; primary lateral veins 8-10 pairs, the I secondary ones parallel to the primary laterals and nearly the same size, the tertiary laterals irregularly anastomosing. Flowers 1-3 in the axils or at recently defoliated nodes; pedicels 4-7 cm long, stout, rufous pubescent; sepals 7-12 mm long, lanceolate, densely rufous pubescent; corolla glabrous, 7-13 mm long, the lobes 6-11 mm long with 2 dorsal appendages frequently divided 2-5 times, the tube 1-2 mm long; filaments 0.5-3.5 mm long, attached at the level of the sinuses, the staminodes lanceolate, 4-6 mm long, rufous pubescent; ovary densely rufous 184 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Voi . 68 pubescent, 8-loculed, the style 5-6 mm long. Fruit globose, 3-5 cm in diameter; seeds 2-8, laterally compressed, 2-2.5 cm long, the seed scar basal, nearly cir- cular, 5 mm broad. This species occurs in Madagascar and the Comores. It is cultivated in Panama and elsewhere in the tropics. The edible pulp of the fruit is reputedly mealy and sweet. CANAL zone: Summit Garden, Ha\es 305 (US), vhraguas: Divisa, Lao 4 (MO). 2. Mimusops elengi L., Sp. PI. 349. 1753. Tree to 15 m. Leaves with petioles 2-2,6 cm long; blades elliptic to elliptic- obovate, thin, 5-12 cm long, 4-6 cm wide, sparsely rufous pubescent, soon gla- brate; primary lateral veins 15-25 pairs, fine, the secondary ones parallel to the primary laterals and nearly the same size, the tertiary laterals irregularly anas- tomosing. Flowers 1-3 in the axils; pedicels 7-8 mm long, rufous pubescent; sepals 6-7 mm long, triangular, pale rufous pubescent; corolla white, glabrous, 8-12 mm long, the lobes 6-9 mm long with 2 dorsal appendages of about equal length, the tube 2^ mm long; filaments 1-3 mm long, attached at the level of the sinuses, the staminodes lanceolate, 4-6 mm long, pubescent; ovary densely pu- bescent, 8-loculed, the style 3-4 mm long. Fruit ovoid, smooth, 2-3 cm long, 1.5-2 cm thick; seeds solitary, ovoid, 1.5-2 cm long, the seed scar basilateral, nearly circular, 3 mm broad. This species is from southeast Asia. It is cultivated in Panama and elsewhere in the tropics. It is a good shade tree with sweetly scented small white flowers. canal zone: Coco Solo Navy Reservation, Coffev s.tu (MO). Cultivated street tree. Nee 14035 (MO). Fort Clayton, Tyson & Blum 2005 (MO). Curundu, Tyson 3460. 3482 (both MO). Panama: Panama City, Lao 455 (MO). Near Hotel Panama, Nee 11492 (MO). 9. NEOXYTHECE Neoxythece Aubreville & Pellegrin, Adansonia, n.s., 1: 16. 1961. type: N, elegans (A. DC.) Aubreville. Trees or shrubs lacking spines. Leaves alternate, exstipulate; blades firm to coriaceous, the primary lateral veins few, often obscure. Flowers short-pedicel- late; sepals 5, uniseriate; corolla lobes 5, lacking appendages; stamens epipet- alous, the filaments variously attached from near the base of the tube to the level of the sinuses, the staminodes alternating with the stamens, rudimentary, the number irregular, often completely absent; ovary 2(-3)-loculed. Fruit baccate; seeds solitary, the seed scar lateral, linear, usually extending nearly the entire length of the seed, the endosperm absent, the cotyledons fleshy. Neoxythece is a tropical American genus of about 10 species found chiefly in northern South America. Literature: Cronquist, A. 1946. Studies in the Sapotaceae, VI. Miscellaneous notes. Bu Torrey Bot. Club 73: 465-471 {Oxythece, 467-468). 1981] PILZ— SAPOTACEAE 185 a. Fruit constricted basally to a stipe 8-15 mm long, not beaked, yellow orange; leaves gla- brous, rarely pale pubescent beneath 1. N. dura aa. Fruit with a prominent beak to 1 cm long, not stipitate, purple black; leaves chestnut brown puberulent beneath 2. N. ma^uirei 1. Neoxythece dura (Eyma) Aubreville & Pellegrin, Adansonia, n.s., 1: 17. 1961. PouterUi dura Eyma, Recueil Trav. Bot. Neerl. 33: 187. 1936. type: Guyana, Schomhur^k 910 (not seen). Tree to 8 m. Leaves with petioles 10-15 mm long; blades obovate to elliptic- obovate, firm to coriaceous, acute to rarely rounded, acuminate, 4-8 cm long, 2- 3 cm wide, glabrous above, rarely pale puberulent beneath, glabrate; primary lateral veins 13-18 pairs, obscure, the secondary and tertiary ones obscure. Flow- ers 2-7 in the axils or at recently defoliated nodes; pedicels 5 mm long; sepals 5, ovate, 1-2 mm long, 1-2 mm wide, abaxially rufous pubescent; corolla glabrous, 3 mm long, the lobes 5, ovate, as long as the tube; filaments to 0.5 mm long, attached near the middle of the tube, the staminodes 0-5, minute, deltoid; ovary hirsute, 2-loculed, the style 3 mm long. Fruit constricted basally to a stipe 8-15 mm long, subglobose above the stipe, yellow orange, rufous pubescent, 3-5 cm long, 2-2.5 cm broad; seeds solitary, ovoid, laterally compressed, to 2.5 cm long, the seed scar linear, extending nearly the entire length of the seed, 1-2 mm wide. This species occurs in Brazil (Amazonas), Surinam, Guyana, Venezuela, Co- lombia, and Panama. Neoyxthece dura is very closely related to /V. maguirei, but its yellow orange, basally constricted fruits, even when they are young, readily distinguish it from the purple black fruits of A^. maguirei, darien: Without locality, Duke & Bristan 8336 (MO). Panama: Road between El Llano and Carti-Tupile, 12 mi above Panamerican Highway, Liesner 1328 (MO, NY, UC). 2. Neoxythece maguirei Aubreville, Mem. New York Bot. Gard. 23: 223. 1972. type: Venezuela, Amazonas, Rio Guainia, infrequent in Caiio Pimichin below Pimichin, Maguire & Wurdack 35614 (NY, holotype; NY, isotype). Shrub or small tree to 6 m. Leaves with petioles 10-15 mm long, highly canaliculate; blades obovate to obovate-elliptic, coriaceous, rounded, often cus- pidate, 6-8 cm long, 3-3.5 cm wide, glabrous above, distinctly chestnut brown puberulent beneath; primary lateral veins 14-18 pairs, slender, the secondary and tertiary ones obscure. Flowers 1-7 at recently defoliated nodes; pedicels 4-5 mm long; sepals 5, ovate, 1.5-2 mm long, 1.5-2 mm wide, abaxially sparsely pubes- cent; corolla white, glabrous, 3-3.5 mm long, the lobes 5, ovate, ca. as long as the tube; filaments to 0.5 mm long, attached near the middle of the tube, the staminodes not observed; ovary hirsute, 2-loculed, the style 0.5 mm long. Fruit with a prominent beak to 1 cm long, ovoid, black purple, rufous pubescent, 2.5- 3 cm long, 1-1.5 cm wide; seeds solitary, ellipsoidal, 17 mm long, 8 mm wide, 5 mm thick, the seed scar linear, extending nearly the entire length of the seed, 1-1.5 mm wide. 186 ANNAl.S OF THE MISSOURI BOTANICAL GARDEN (Voi . 68 This species occurs in Venezuela (Amazonas) and Panama. Neoxythece maguirei is similar and presumably closely related to /V. dura. Panama: El Llano-Carti Highway, 17-20 km N of El Llano, Dressier 4630 (MO). El Llano-Carti road, km 12, Croat 26045 (MO), 10. POUTERIA Pouteria Aublet, Hist. PI. Guiane 1: 85. pL 33, 1775. type: P, gidanensis Aublet. Trees or shrubs lacking spines. Leaves alternate or rarely subopposite, ex- stipulate; blades membranous to coriaceous, the primary lateral veins generally few, strongly arcuate near the margin. Flowers sessile to long-pedicellate; sepals 4-12, spiral or in 1 or 2 series; corolla lobes 4-6, lacking appendages; stamens epipetalous, occasionally abortive, the filaments variously attached from near the base of the tube to the level of the sinuses, the staminodes alternating with the stamens, petaloid to rudimentary; ovary 1-10-loculed. Fruit fleshy, occasionally sclerotic; seeds 1-several, the seed scar lateral, linear to very broad, the endo- sperm absent, the cotyledons fleshy. Pouteria is a genus of perhaps 100 species found chiefly in tropical America. Pouteria has been viewed quite differently by various workers. Baehni (1942) considered it in a very broad sense and included over 300 distinct species. More recently Baehni (1965) split the genus somewhat, but his list of generic synonyms still includes over twenty names, and a list of probable synonyms adds twenty more. Aubreville (1965, 1972) and Lundell (1976) have adopted much narrower generic concepts. Following Aubreville's nomenclature the Panamanian Pouteria would be split into no fewer than nine genera. Blackwell (1968) followed the system proposed by Cronquist (1946), and I follow roughly the same generic limits for American Sapotaceae, particularly Pouteria, This is done with the knowledge that several segregates of Pouteria are probably sound and deserve recognition at least at the subgeneric level. Oth- erwise we are left with a large but still poorly defined genus. Unfortunately many of the segregate genera themselves are poorly defined, being based on relatively few specimens. As more material is collected, the generic limits may have to be shifted. This is particularly true among American species. Large series of spec- imens exist for relatively few species, and the number of species only known from the type collection is large. If Panamanian Pouteria can be used as an example, 7 of 25 species are known from only the type specimen or type locality. Similarly many species lack complete descriptions of flowers and/or fruits. In this treatment flowers of three Pouteria species are adequately described for the first time. Aubreville (1965, 1972) has proposed the most complete scheme for dividing Pouteria (sensu lato) into its narrower generic units. To aid those unfamiliar with his many important works, which are exclusively in French, I have included in this portion of the treatment the name adopted by Aubreville. When it is certain, this name has been placed in the synonymy along with the basionym. A few species are known from such scanty material that correct placement in Aubre- I981J PILZ— SAfKJFACEAE 187 ville^s scheme is uncertain. It should be noted that only 3 of the 25 Panamanian Pouteria would be retained in this genus by Aubreville. Literature: Aubreville, A. 1961. Notes sur des Pouteriees Americaines. Adansonia, n.s., 1: 150-191. Baehni, C. 1942. Memoires sur les Sapotacees, II. Le Genre Pouteria, Candollea 9: HI ^16. Cronquist, A. 1946. Studies in the Sapotaceae, II. Survey of the North American genera. Lloydia 9: 241-292 {Pouteria, 257-291). Lundell, C, L. 1976. Studies of American plants, XII. Wrightia 5: 241-259 (Sa- potaceae, 252-256). a. Sepals 8-12, evidently spiralled, increasing in size centripetally. b. Fruit fleshy. c. Leaves with 20-50 primary lateral veins: pedicels lo 2 mm long; sepals often emarginate: corolla lobes as long as the tube 19. P, sapota cc. Leaves with 12-20 primary lateral veins; pedicels 4-6 mm long; sepals entire or scarcely emarginate; corolla lobes much shorter than the tube 13. /^. fossicola bb. Fruit woody. d. Leaves with 8-13 primary lateral veins; sepals thick and fleshy 8. P. cooped dd. Leaves with 13-20 primary lateral veins; sepals thin 20. P. sclerocarpa aa. Sepals 4-7, imbricate or quincuncial, about equal in length. e. Flowers sessile, the pedicels 2 mm long or less. f. Sepals 5-7; filaments attached at the level of the sinuses between the corolla lobes (unknown in P. chiricana). g. Corolla 6-8 mm long; sepals 6 or 7; corolla tube longer than the corolla lobes 20. P. sclerocarpa Corolla 2.5-5.5 mm long; sepals 5; corolla tube shorter th:^n the corolla lobes. h. Secondary lateral veins fine, close, sinuous but regularly disposed 23. P, suhrotata hh. Secondary lateral veins rather coarse, irregularly disposed. I. Fruit ovoid to subglobose: sepals 2.5-5.5 mm long 10. P. durlandii ii. Fruit ellipsoid, narrowed to a stipitate base; sepals 2-2.6 mm long 6. P. chiricana ff. Sepals 4(-5); filaments attached at the middle or base of the corolla tube (un- known in P. samhuensis). j. Fruit fleshy. k. Fruit glabrous 2. P. caimito kk. Fruit conspicuously hairy 18. P. samhuensis ij. Fruit woody or gall-like. I. Fruit covered with hairy processes to 2 cm long; sepals 7-9 mm long; corolla 6-9 mm long 16. P. nci^lecta II. Fruit glabrous to pubescent, but no long hairy processes; sepals 2.5- 4 mm long; corolla 2.5-4 mm long 22. P. stylosa ee. Flowers pedicellate, at least some pedicels 3 mm or more in length. m. Corolla at least 7 mm long; largest sepals at least 4 mm long. n. Corolla glabrous externally; pedicels 3-8 mm long. o. Sepals 4. p. Primary lateral veins of the leaves well marked, easily distin- guished from the secondary ones; petioles 1.5-2.5 cm long; style 5-6 mm long 5. P, carahohensis pp. Primary lateral veins difficult to distinguish from the secondary ones; petioles I-I.5 cm long; style 6-11 mm long 9. P, dominii^ensis 1 88 ANNALS OF THE MISSOURI BOTANICAL GARDEN (Vol. 68 00. Sepals 5-6 1. /*. huenaventurensis nn. Corolla sparsely sericeous-strigose externally; pedicels 8-15 mm long 4. P. campechiana mm. Corolla less than 6.5 mm long; sepals less than 3.5 mm long. q. Ovary with 1-3 locules. r. Leaves glabrous or essentially so 25. F, unilocuhiris rr. Leaves densely pubescent, at least beneath. s. Pubescence pale; leaves with 9-14 primary lateral veins 24. P. tarapotensis ss. Pubescence rufous; leaves with 15-22 primary lateral veins 3, P, calistophylla qq. Ovary with 4 or more locules. t. Sepals and corolla lobes generally 4-merous. u. Leaves finely pale sericeous beneath . 12. P. euryphylla uu. Leaves sparsely pubescent to glabrous. v. Leaves 15-25 cm long, 8-15 cm wide; petiole 20-30 mm long; pedicel 6-10 mm long; style 3^ mm long __ 14. P. leptopedicelluta vv. Leaves 5-13 cm long, 2-5 cm wide; petiole 4-17 mm long; pedicel 2-5 mm long; style 1-1.5 mm long 1\. P. stipitula tt. Sepals and corolla lobes generally 5-merous. w. Leaves crowded at ends of stout branches; flowers and fruits borne on branches 7 mm or more in diameter ___ 7. P. con^estifolia WW. Leaves scattered along slender branches; flowers and fruits usually borne on branches less than 5 mm in diameter. X. Petiole flattened above, broadly canaliculate; fruit usu- ally with 1 seed __ II. P. en^lcri XX. Petiole nearly terete, narrowly canaliculate; fruit usually with 5 seeds. y. Fruit mealy roughened, even when young 15. P. lucentifoliii yy. Fruit nearly smooth, slightly low-luberculate \1 . P. pentasperma 1. Pouteria huenaventurensis (Aubreville) Pilz, comb. nov. Richaniella huenaventurensis Aubreville, Adansonia, n.s., 7: 146. 1967. type: Colombia, Deparl- mento del Valle, Rio Yurumangui, Cuatrecasas 15821 (US). Tree to 30 m. Leaves with petioles 1-2 cm long; blades elliptic to elliptic- oblanceolate, coriaceous, rounded to obtuse and occasionally abruptly acumi- nate, 8-18(-20)cmlong, 4.5-6(-ll)cm wide, sparsely rufous pubescent, glabrate; primary lateral veins 8-12 pairs, the secondary and tertiary laterals prominent, irregular, often branched, much of the venation perpendicular to the previous series. Flowers 1-3 in the axils and at recently defoliated nodes; pedicels 3^ mm long; sepals 5-6, ovate, 3-6 mm wide, the outer 2-2.5 mm long, the inner 4-5 mm long, densely sericeous except on the margins and the lower portion of the adaxial surface; corolla greenish white, glabrous, cylindric, 7-12 mm long, the lobes 5, rounded, about V2 as long as the tube; filaments 3 mm long, attached at the level of the sinuses or slightly below, the staminodes lanceolate, 2-2.5 mm long; ovary densely pubescent, (4-)5-loculed, the style 8-10 mm long. Fruit to 4 cm long, globose, mealy roughened, brown; seeds solitary, to 2.8 cm long, globose, the seed scar covering Vi-% of the seed surface. This species occurs in Colombia and Panama. 1981] pil^Z— SAPOTACEAE 189 Although the plants are morphologically very similar, it must be noted that the Colombian specimens are all from low elevations (Departmento del Valle), while the Panamanian material was collected at elevations of 350 to 1,000 m. colon: Santa Rita Ridge road, 14 km from Boyd-Roosevelt Highway. Mori & Kallunki 4907 (MO). Panama: Cerro Jefe, D^^yer 9495 (F, MO); Gentry 4874, 6144 (both MO). N of Goofy Lake, Folsom et al. 1959 (MO). 16-20 km above Panamerican Highway on road from El Llano to Carti- Tupile, Kennedy 2709 (MO). 9-20.7 km above Panamerican Highway on road from El Llano to Carti, Mori & Kallun'ki4686, 51 13. 5151 (all MO). 2. Pouteria caimito (Ruiz & Pavon) Radlkofer, Sitzungsber. Math.-Phys. CI. Kon- igl. Bayer. Akad. Wiss. Miinchen 12: 333. 1882; Aubreville, Adansonia, n.s., 1: 154. 1961. Achras caimito Ruiz & Pavon, Fl. Peruv, 3: 18, pL 240. 1802. type: Peru, Ruiz. & Pavon (F, isotype). Tree to 30 m. Leaves 5-24 cm long, 2-9 cm wide, obovate-oblanceolate to elliptic, acuminate, rarely acute; petioles 5-15 mm long. Flowers 1-5 per axil, subsessile to sessile; sepals 4(-5), 3^.5 mm long; corolla 5-8 mm long, the lobes 4(-5), about as long as the tube; filaments attached near the middle of the tube; ovary 4(-6)-loculed. Fruit yellow to brown, fleshy, 5-10 cm long, 4-8 cm wide, 1-4-seeded; seed scar extending the entire length of the seed, 3-4 mm wide (Blackwell, 1968). This species occurs in Peru and Brazil and through northern South America to Panama. It is occasionally cultivated. colon: Salud Hills, Lao & Holdrid^c 192 (MO, tentative identification of sterile specimen). darien: Sambii River, Pittier 5555 (F, US). 3. Pouteria calistophylla (Standley) Baehni, Candollea 9: 419. 1942. Liicuma calistophylla Standley, Publ. Field Mus. Nat. Hist., Bot. Ser. 4: 252. 1929. type: Panama, Cooper 481 (F, holotype; K, NY, US, isotypes). Tree ca. 20 m tall. Leaves 10-22 cm long, 5-10 cm wide, obovate or elliptic- obovate, acuminate; primary lateral veins 15-22 pairs; petioles 1-3 cm long. Flowers imperfectly known; pedicel ca. 5 mm long; sepals 5, ca. 2 mm long; corolla 5-lobed; filaments attached at the level of the sinuses: ovary probably 2-loculed. Fruit unknown (Blackwell, 1968). Known definitely only from the type collection. BOCAS DEL TORo: Cricamola Valley, Cooper 481 (F, K, NY, US). 4. Pouteria campechiana (H.B.K.) Baehni, Candollea 9: 398. 1942. Lucuma campechiana H.B.K., Nov. Gen. Sp. PI. 3: 240. 1819. type: Mexico, near Campeche, Humboldt & Bonphmd (photo F). Richardia campechiana (H.B.K.) Pierre, Not. Bot. Sapot. 20. 1890; Aubreville, Adansonia, n.s., I: 175. 1961. Radlkoferella i^lahrifolia (Pittier) Aubreville, Mem. New York Bot. Card. 23: 206. 1972. Lucuma glahrifolia Pittier, Contr. U.S. Natl. Herb. 20: 481. 1922. type: Panama, Pittier 6542 (US, holotype: EAP, F, NY, isotypes). 190 ANNALS OF THH MISSOURI BOIANICAL GARDEN IVoi . 68 Tree to 30 m. Leaves 10-35 cm long, 4-10 cm wide, elliptic to narrowly obovate, acuminate to rarely rounded; primary lateral veins 12-20 pairs; petioles 1-3 cm long. Flowers (l-)2-4(-9) per axil; pedicels 8-15 mm long; sepals (4-)5 (-6), 5-10(-12) mm long; corolla 7-14 mm long, the lobes (4-)5(-7), as long as the tube or somewhat shorter; filaments attached slightly below the level of the sinuses; ovary (4-)5(-10)-loculed. Fruit yellow at maturity, fleshy and edible, to 7 cm long, 7 cm wide, 1-4-seeded; seed scar extending the entire length of the seed, 1-2 cm broad (Blackwell, 1968). This species ranges from Mexico to Amazonas, Brazil. It is widely cultivated and many collections represent introduced plants. Aubreville (1961, 1972) placed plants of this species into two closely related genera as Richardella campechiana and Radlkoferella gfahrifolia. The variability in the number of flower parts, particularly the number of sepals and ovary locules, as well as leaf size and shape, and fruit characteristics has resulted in the proposal of at least ten allied taxa. Segregation of this species into related species or even subspecies would be artificial until extensive monographic studies are undertaken. CANAL zone: Ancon, Piper 6027 (F, US). Barro Colorado Island, Zetek 5562 (EAP, MO), coci E-: Penonome, WUUanis 56 (US), colon: Donoso, Holdridi^e 6202 (MO), darien: Casaya Island, Duke 10374 (MO). Pinas, Duke 105H9 (MO). Pinogana, Pittier6542 (EAP, F, NY, US). 2 mi E of Santa Fe, Tyson et al. 4839 (MO). LOS Santos: Punta Mala, Croat 9751 (MO). 16 mi S of Macaracas at Quebrada Bejiico, Tyson et al. 3092 (MO). Panama: Trapiche Island, Allen 2609, 2627 (both EAP, F); Miller 1875. 1902 (both US). Chepillo Island, Duke 10320 (MO). Saboga Island, Duke 10343 (MO). Espiritu Santo Island, Duke 10454 (MO). San Jose Island, Erianson 201. 235 (both US), 396 (EAP, US); Johnston 526, 733, 1171 (all MO, US); Miller 1928 (US). Rio Pasiga, Gentry 2310 (MO). Chichebre. Chepo, Hoklruli^e 6497 (MO). Pacheca Island, Tyson et al. 5601, 5604 (both MO), veraguas: Bahia Honda, Pueblo Nuevo, Barclay 2827 (MO). 5. Pouteria carabobensis Pittier, Contr. Fl. Venez. 12. 1921. type: Venezuela, hills of Guaremales, road from Puerto Cabello to San Felipe, Pittier 8921 (US, holotype, photo MO; NY, isotype). r Tree to 25 m. Leaves with petioles 15-25 mm long; blades elliptic to elliptic- obovate, thin to firm, obtuse to acute, often abruptly acuminate, 12-22 cm long, 4-10 cm wide, very sparsely gray pubescent to glabrous; primary lateral veins 11-13 pairs, thin, the secondary laterals curved, oblique to the primary ones, those near the margin straighter and nearly perpendicular, the tertiary laterals very fine, anastomosing. Flowers 4-7 in the axils and at recently defoliated nodes; pedicels 3-8 mm long; sepals 4, ovate, 4 mm long, 4-5 mm wide, the outer ones pubescent on the abaxial surface, the inner ones and adaxial surfaces glabrous; corolla papillose, glabrous, 7-13 mm long, the lobes 4-5(-6), ovate, nearly equal to much shorter than the tube; filaments 1-2 mm long, attached at the level of the sinuses, the staminodes broadly lanceolate, 1.5-3 mm long; ovary pale pilose, 4-loculed, the style glabrous, 5-6 mm long. Fruit reputedly edible, oth( unknown. rwi This species occurs in Venezuela and Panama. Plants of this species have highly variable flowers. I have seen only flowers with 4 sepals, but the corollas may be 4-6-merous on the same branch. 1981] PILZ— SAPOTACEAH 191 darien: Rio Tuquese, al middle Tuquesa Mining Company camp called Charco Peje, riverside, Mori 7101 (MO), 6. Pouteria chiricana (Standley) Baehni, Candollea 9: 420. 1942. Lucuma chiricana Standley, Publ. Field Mus, Nat. Hist., Bot. Ser. 4: 251. 1929. type: Panama, Cooper & Slater 254 (F, holotype: NY, US, isotypes). Tree to 30 m. Leaves 8-16 cm long, 2.5-6 cm wide, elliptic to oblong, acu- minate; petioles 6-17 mm long. Flowers subsessile in the axils; sepals 5, 2.2-2.6 mm long; corolla 3-3.2 mm long, the lobes 5, about twice as long as the tube; ovary 4-loculed. Fruit fleshy, 3-3.5 cm long, 2-2.5 cm wide, 1-seeded; seed scar extending the length of the seed, ca. 8 mm wide (Blackwell, 1968). This species occurs in Costa Rica and Panama. Flowers with stamens have not been collected for Pouteria chiricana. Other Central American Pouteria known to possess heteromorphic flowers are P. amyJalina, P, durlandii, and P. stipitata. BOCAS DEL TORo: Almirante region, Cooper 445 (F, NY), 457 (EAF, F, NY), chiriqui: Progreso, Cooper & Slater 230, 254 (both F, NY, US). Burica Peninsula, 9 mi S of Puerto Armuelles, Croat 22107 (MO, NY). 7. Pouteria congestifolia Pilz.*^ type: Panama, Allen 3426 (MO, holotype; EAP, F, isotypes). Tree 30 m tall. Leaves crowded at the ends of stout branches; petioles 2-3 cm long; blades oblanceolate, firm, 15-20 cm long, to 5 cm wide, mucronate to shallowly emarginate, sparsely pilose; primary lateral veins 15-20 pairs, the sec- ondary ones somewhat sinuous, generally perpendicular to the primary laterals, the tertiary ones highly reticulate, nearly as prominent as the secondaries. Flow- ers 2-4 in the axils or at recently defoliated nodes; pedicels 4-10 mm long, rufous pubescent; sepals 5, ovate, 3-4 mm long, 2.5-3 mm wide, densely rufous seri- ceous externally except for the margin of the inner sepals, the adaxial surface of all sepals glabrous; corolla green, 4-5 mm long, the 5 lobes ciliolate, sparsely pubescent, about as long as the tube; filaments 1.5-2 mm long, attached near base of the tube, the staminodes lanceolate, 0.7 mm long; ovary 1-1.5 mm high, densely rufous pilose, 4-5-loculed, the style 1.5 mm tall, 5-lobed. Fruit unknown. Known only from two collections. They are curious specimens that do not seem to be closely related to any species of my acquaintance. The combination of a 5-merous flower with filaments inserted near the base of the corolla tube and with well-developed staminodes is not frequently observed in American Sapo- ■' Pouteria congestifolia Pilz, sp. nov. Arbor 25 m alta. Folia ad ramulorum apicem dense con- gesta; petioli 2-3 cm longi: laminae oblanceolata, firmae, 15-20 cm longa, 4-5 cm lata, apice mu- cronata vel emarginata: costae 15-20 jugalae. Flores 1~A ad axillam foliorum persistentium vel de- lapsorum fasciculati; pedicelli rufo-pubescentes, 4-10 mm longi; sepala 5, ovata, 3^ mm longa, 2.5- 3 mm lata, dorso rufo-sericeo, intus glabro; corolla 4-5 mm longa: lobi 5, ciliolati, tubo aequilongi; filamenta 1.5-2 mm longa, fere ad basin tubi affixa; staminodia lanceolata, 0.7 mm longa; ovarium 1- 1.5 mm altus, dense pilosum, 4-5-loculare, cum stylo 1.5 mm longum. Bacca ignota. 192 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Figure L Pouieria con^estifolia Pilz.— A. Habit (x^/io).— B. Flower (x7).— C. Corolla frag- ment (x7). [After Allen 3426 (MO). J 1981] PILZ— SAPOTACEAE 193 taceae. Further generic speculation awaits the collection of fruits from this species. The epithet congestijolia draws attention to the manner in which the leaves are crowded at the ends of stout branches. CHiRigui: East of the Fortuna dam site, Mendozci 370 (MO), cocle: El Valle de Anton, 1000 m, Allen 3426 (EAP, F, MO). 8. Pouteria cooperi Cronquist, Lloydia 9: 291, 1946. type: Panama, Cooper 499 (NY, holotype; EAP, F, US, isotypes). Calocarpuni cooperi (Cronquist) Lundell, Wrighlia 5: 252. 1976. Tree ca. 15 m talK Leaves 8-17 cm long, 3-7 cm wide, elliptic or narrowly obovate, acuminate; primary lateral veins 8-13 pairs; petioles 1-3 cm long. Flow- ers subsessile; calyx of 10 or more thick, fleshy, broad sepals to 5 mm long: other flower parts unknown. Fruit reputedly woody (Blackwell, 1968). This species is known only from the type collection, BOCAS DEL TORo: Almirantc region, Cricamola Valley, Cooper 499 (EAP, F, NY, US). 9. Pouteria dominigensis (Gaertner f.) Baehni var. dominigensis, Lloydia 9: 278. 1946. Luciima dofiiifii^cnse Gaertner f., Fruct. 3: 131. 1807, rvpt: Haiti, Poiteau, not seen. Radlkoferelhi domingensis (Gaertner f.) Pierre, Not. Bot. Sapot. 21. 1890; Auhreville. Adansonia, n.s,, 1: 185. 1961. Tree to 10 m. Leaves 4-13 cm long, 2-6 cm wide, obovate-oblanceolate, rounded, rarely acute; primary lateral veins 10-20 pairs; petioles 1-1.5 cm long. Flowers usually several per axil; pedicels 3-8 mm long; sepals 4, 4-9 mm long; corolla 8-16 mm long, the lobes (5-)6, about as long as the tube; filaments at- tached at the level of the sinuses; ovary (5-)6(-8)-locu]ed. Fruit yellow, fleshy, 2-5 cm long, 3-6 cm wide, 1-several-seeded; seed scar variable in length, 3-10 mm broad (Blackwell, 1968). This species occurs in West Indies and southern Florida. It is cultivated in the Canal Zone. CANAL zone: Without locality, Joliansen 36 (F, NY, US). Ancon, Mell s.n. (F, NY). Balboa, Standley 26894 (F, MO, US), 30859 (F, US). 10. Pouteria durlandii (Standley) Baehni, Candollea 9: 422. 1942. Lucuma durlandii Standley, Trop. Woods 4: 5. 1925. ivpf: Guatemala, Departmento Pelen, El Paso, Durland sM. (US, holotype; F, isotype). Peteniodendron durlandii (Standley) Lundell, Wrightia 5: 254. 1975. Paralahatia durlandii (Standley) Aubreville, Adansonia, n.s., 3: 21. 1963. Tree to 25 m. Leaves with petioles 1-3 cm long; blades obovate to oblanceo- late, firm, obtuse to abruptly acuminate, 10-25 cm long, 4-9 cm wide, essentially glabrous above, pale pubescent beneath, glabrate; primary lateral veins 8-13 pairs, the secondary and tertiary ones equally prominent with raised veins be- 194 ANNALS OF THE MISSOURI BOTANICAL GARDLN [Voi . 68 neath. Flowers 1-3 in the axils or at recently defoliated nodes, often heteromor- phic and unisexual, the pistillate flowers generally smaller than the staminate ones; pedicels 1-2 mm long, densely rufous pubescent; sepals 5, ovate, 2.5-4 mm long, 2-2.5 mm wide, sparsely to densely rufous pubescent; corolla greenish white, glabrous, 3-5.5 mm long, the lobes 5, ovate, equal in length to twice as long as the tube; filaments 0.5-1.5 mm long, attached at the level of the sinuses, the staminodes petaloid to lanceolate, 0.5-2 mm long; ovary densely rufous pu- bescent, 2(-5)-loculed, the style 0.8-1.5 mm long. Fruit yellow, subglobose, densely rufous pubescent, glabrate, to 2.5 cm long; seeds 1-3, ovoid to subglo- bose, to 2 cm long, the seed scar extending nearly the entire length of the seed, 4-11 mm broad, covering V^-V2 of the seed. This species occurs in Mexico, Guatemala, Belize, Honduras, Costa Rica, and Panama. Peteniodendron Lundell (1975) is distinguished by its heteromorphic unisexual flower, 2-3-loculate ovary, pubescent fruits, and large seed scars. None of the above characteristics are unique to Peteniodendron and within the highly retic- ulated Poutereae Peteniodendron does not, at this time, deserve generic recog- nition. Other Central American Ponteria known to possess heteromorphic uni- sexual flowers are P. amydalina, P> chiricana, and P. stipitata. Only young fruiting specimens have been collected in Panama, and their iden- tity is problematic. Panamanian material differs from typical Ponteria durhmdii in generally having 5-loculed ovaries, although ovaries with 2-5 locules were observed on the same branch. DAKihN: Without locality, Duke 8357 (MO). Pine, Duke & Bristan 248 (MO). II, Pouteria engleri Eyma, Recueil Trav. Bot. Neerl. 33: 178. 1936. type: French Guiana, Melinon sji, (P, holotype; K, isotype). Nenuiluma engleri (Eyma) Aubreville & Pellegrin, Adansonia, n.s., 1:31. 1961. Tree to 20 m. Leaves with petioles 10-l5(-25) mm long; blades obovate to elliptic-ovate, firm, rounded, mucronate, 6-15(-20) cm long, 3-8(-13) cm wide, very sparsely pale pubescent, glabrate; primary lateral veins 7-11 pairs, distinct, the secondary and tertiary ones obscure above, evident beneath, sinuous, areoles incomplete to lacking. Flowers l-10(-20) in the axils of the leaves; pedicels 4-7 mm long, sparsely rufous pubescent; sepals 5, ovate, 2-2.2 mm long, 2 mm wide, sparsely rufous pubescent, glabrate; corolla greenish white, glabrous, 3.5-4 mm long, the lobes 5, ovate, about as long as the tube; filaments 1 mm long, attached near the middle or base of the tube, the staminodes rare, variable in number, deltoid, minute; ovary pale pubescent, 4-5-loculed, the style 2 mm long. Fruit ellipsoidal, 18-23 mm long, smooth, glabrous; seeds 1-5, ovoid, laterally com- pressed, 14-18 mm long, the seed scar linear, extending the entire length of the seed, 2.5 mm wide. This species occurs in Surinam, Guyana, French Guiana, and Panama. colon: Maria Chiquita, HoUlriclge 6524 (MO). Panama: El Llano to Carli, Correa et al. 1838 (MO, tentative identification of a badly parasitized specimen). 198iJ PILZ— SAPOTACEAE 195 12. Pouteria euryphylla (Standley) Baehni, Candollea 9: 249. 1942. Lucuma euryphylla Standley, Publ. Field Mus. Nat. Hist., Bot. Ser. 4: 252. 1929. type: Panama, Cooper 611 (F, holotype; NY, isotype). Tree ca. 15 m. Leaves 10-22 cm long, 4-12 cm wide, elliptic or obovate- elliptic, acuminate; primary lateral veins 9-13 pairs; petioles 1.5-6cm long. Flon^- ers few per axil; pedicels 4-5 mm long; sepals 4, ca. 3.5 mm long; corolla 3.7-4.5 mm long, the lobes 4(-5), about as long as the tube; filaments attached at about the middle of the tube; ovary 4-loculed. Fruit unknown (Blackwell, 1968). Known only from the type collection. Most likely related to four species from Brazil and northern South America that are placed by Aubreville in Pseudola- hatia. BOCAS DEL TORo: Almirante region, Buena Vista Camp, Cooper 611 (F, NY). 13. Pouteria fossicola Cronquist, Lloydia 9: 289. 1946. type: Panama, Canal Zone, Barro Colorado Island, north shore near Pearson terminal, Salvozxi 999 (A, not seen). Calovarpum fosslcolum (Cronquist) Lundeil, Wrightia 5: 252. 1976. C. horueanum Standley & L. O. Williams, herbarium name, never validly published. Tree to 30 m. Leaves 8-30 cm long, 4-13 cm wide, obovate, short-acuminate to obtuse; primary lateral veins 12-20 pairs; petioles 1-4.5 cm long. Flowers 1- 5 in the axils of the leaves or at recently defoliated nodes; pedicels 4-6 mm long; sepals (6-)8(-9), spirally arranged, to 7 mm long; corolla white, densely pubescent except for the margins of the lobes and the base of the tube, cylindric, 9-11 mm long, the tube comprising M of the total length, the lobes 5, ovate; filaments attached at the level of the sinuses or slightly below, 1-1.5 mm long, the stami- nodes lanceolate, 1-1.2 mm long; ovary 5-loculed. Frail yellow green to grayish white, fleshy, 8-15 cm long, 4-9 cm wide, l(-2)-seeded; seed scar extending the entire length of the seed, 3-4 cm broad (Blackwell, 1968). This species occurs in Panama and southeastern Costa Rica. As presently circumscribed, this species, once known only from Barro Col- orado Island (Blackwell, 1968), extends to Puntarenas, Costa Rica, Allen 5753 and 6636 (both EAP). CANAi zone: Barro Colorado Island, Bangham 583 (F); Zetek 3870 (F, MO), cocle: El Valle de Anton, Lao 278 (MO). El Valle de Anton, 600 m, cultivated tree. Nee . Achinium ad 7 mm longum, alis ca. dimidio latitudino corpi, aristis validis, triangularibus, ca. % longitudino corpi. Tree to 25 m tall, 40 cm d.b.h.; trunk gray, smooth; wood white; young stems woolly with yellow brown hairs. Leaves alternate, broadly lanceolate; blade to 60 cm long, 30 cm wide, the base shallowly cordate or truncate, but acuminate near the inflorescence, often oblique, the margin with ca. 5 short teeth per cm, tomentose below, woolly on the veins, scabrous above; petioles 10-12 cm long, Infl^ 20 cm across; pedicels 1-10 cm long, woolly. Heads globose, truncate at the base, the involucre 1.5-2 cm wide; involucral bracts approximating 2 whorls, the outer bracts wider and shorter than the inner more lanceolate bracts; paleas 10-12 mm long, folded around the floret; ray florets fertile, bright orange yellow, ca. 25 per head, the corolla to 20 mm long, abruptly expanded from the short, pubescent tube (2^ mm long) into a lanceolate ligule to 4 mm wide; disc florets fertile, yellow green, the corolla 6-7 mm long, glandular pubescent on the lower '/3, cylindrical, slightly and gradually expanded towards the top, the lobes shallow, less than 1 mm long; anthers black, ca. 3 mm long; style branches spatulate, pubescent at the tips. Achene body black, to 7 mm long, the wings ca. V2 the width of the body, one of them ciliate, the 2 stout triangulate awns ca. % the length of the body. ^ This work is part of the continuing study of the Panamanian fiora by the Missouri Botanical Garden Support was provided by the National Science Foundation Grant DEB 79-22192 (W. G. D'Arcy, principal investigator) and by the Center for Field Research. ^ Department of Botany, Duke University, Durham, North Carolina 27706. ^ Missouri Botanical Garden, Post Office Box 299, St. Louis, Missouri 63166, ^ The field work was undertaken by the authors with the assistance of John E. Averett, University of Missouri, St. Louis, and four Earthwatch volunteers: Elaine C, Hill, Susan Schwartz, Helen M. and Oliver Wolcott. Housing and other help was provided by Maximo Menendez Caballero and by the Panama Audubon Society. Ann. Missouri Bot. Card. 68: 213-217. 1981. 0026-6493/8 1/02 1 3-02 1 7/$0.65/0 214 ANNAI.S OF- THH MISSOURI BOTANICAL GAROKN [Vol . f>8 .-■''3 p--'j ■■■* tt *■ ■■.- m Figure I. Verhesina haruensis Hammel & D'Arcy. lAfter Hummel et al. 6449 (MO),] A. Habit (x!/2).— B. Disc lloret (x4) Tvph: Panama, chiriqui; Volcan Baru, E slope along road to summit near Potrero Muleto, 10400 ft, Hammel et al. 6449 (MO, holotype; F, K, PMA, TEX, isotypes). 1981] HAMMEL & irARCY— NEW PANAMANIAN TAXA 215 Other Specimens Seen: Panama, chiriqui': E slope of Volcan Baru in wet forest N of Potrero Muleto along road to summit, 10200 ft, Hummel 5643 (MO). E slope of Volcan Baru in wet forest N of Potrero Muleto along road to summit, 10400 ft. Hummel 7490 (MO). In the Flora of Panama treatment of the Compositae (D'Arcy, 1975) individ- uals of this species will key out to Verhesinafuscasiccans D'Arcy which is known only from central Panama. This species differs from V. fuscasiccans most ob- viously in overall size of the plants and in leaf shape. We saw no flowering material of V. haruensis under 15 m tall. The large cordate- to truncate-based leaf blades with densely woolly petioles are quite distinct from those of V. fus- casiccans which are long acuminate into a winged, slightly pubescent petiole. The flower heads of V. haruensis are about twice as large as those of the material of V. fuscasiccans studied, including its type. Verbesina haruensis is common along several draws and slopes at about 3,000 m in high rainfall areas on the east and north side of the volcano. From a distance we saw the plants widely scattered along a few deep draws but we also found a number of almost pure stands of this magnificent composite tree. Flowering col- lections have been made between November and May and it seems likely that this tree blooms in all months of the year. Only the tallest plants were found in flower. Sicyos chiriquensis Hammel & D Arcy, sp. nov. — Fig. 2. Labruscae monoeciae, foliis simplicibus 3-5 lobatis basim profunde cordatis. Inflorescentia stami- nata axillaris, racemosa, de inflorescentiis foemineis separata; inflorescentia pistillata axillaris capi- tata. Flos staminatus 5-merus campanulatus 4-6 mm longus, 3^ mm latus, lobis calycis linearibus 1 mm longis, staminibus 5, filamentis connatis sed apice leviter divergentibus. antheris reniformibus, prope medio insertis. Flos pistillatus 5 merus salverformis, perianthio perianthiis staminatis minori, ca. 3 mm longo, stigma 2 vel 3 lobata. Fructus in fasciculis ad 4 cm diametro relentes, maturi samaroidei elliptici ca. 2 cm longi, 1.5 cm lati papyracei praeter partem seminis, cristis binis centralis longitudinalis semine oblectis margineque irregulariter serratim, dentibus late triangularis 2-4 mm longis, aristis recurvatis teclis. Slender monoecious vines; stems glabrous to slightly pubescent when young. Leaves simple, pentagonal in outline, 5-15 cm long, equally wide, primarily 3-lobed to % the distance to the midrib, the lower 2 lobes with 1 or 2 secondary enations, the leaf base deeply cordate and hastate, the margin undulate dentate with ca. 3 teeth per cm, the blade scabrous above with scattered short hairs, more numerous toward the margin, hirsute below, prominently so on the veins; petioles 1-22 cm long, pubescent along one side; tendrils 1-2 branched. Staminate inflorescence axillary, racemose, 2-6 cm long, few(ca. 10)-flowered, floriferous in the upper V^, the pedicels to 2 cm long. Flowers yellow green, campanulate, 4-6 mm long, 3-4 mm wide; lobes of the calyx 5, linear, 1 mm long; corolla 5-merous, merging imperceptibly with the calyx tube, divided to % the length, the lobes broadly lanceolate, nearly erect, incurved, densely papillose-pubescent inside, 3-7-nerved; stamens 5, the filaments united into a column equalling or slightly exceeding the floral tube, free and divergent for ca. 1 mm at the summit, the anthers reniform, attached near the middle. Pistillate inflorescence axillary, capitate, 15-20 flowers sessile on the end of a peduncle ca. 2 cm long; perianth smaller than that of the staminate flowers, ca. 3 mm long, 5-merous, salverform; 216 ANNALS OF THE MISSOURI BOTANICAL GAKDHN [Vol . 68 Figure 2. Sicyos chiriquensis Hammel & D'Arcy. — A. Habit (x!/2), — B. Fruit (x2). [After Hammel et al. 6823 (MO). J calyx lobes linear, 1 mm long; corolla lobes lanceolate-triangular, 2 mm long, papillose inside; stigma 2- or 3-lobed, the flat elliptic ovary with a low, longitudinal ridge on the center of each face, these ridges and the margin of the ovary toothed, the teeth capped with 4 recurved barbs. Fruits samaralike, tan to lustrous brown, 1981] HAMMEL & D ARCY— NEW PANAMANIAN TAXA 217 ca. 2 cm long, 1.5 cm wide, paper thin except in the area of the seed, the central longitudinal ridges over the seed and margin of the samara irregularly serrate, the teeth broadly triangular, 2-4 mm long, capped with recurved barbs, the fruits held by these barbs in pendulous clusters to 4 cm in diameter; seed adherent to the fruit wall. Type: Panama, chiriqui: Along small stream in backyard of Audubon cabin, Bambito, 6200 ft, Hammel et aL 6823 (MO, holotype; DUKE, F, K, MICH, PMA, US, isotypes), Sicyos chiriquensis is quite distinct from other Panamanian Cucurbitaceae in its paper-thin samaroid fruits. Although these represent an extreme otherwise unknown in the genus, somewhat flattened fruits and armed fruits with recurved barbs on the spines do occur. This complex of characters suggests that animal and perhaps torrent dispersal have been important in the evolution of Sicyos. To the casual observer Sicyos chiriquensis might easily appear sterile while in full flower and fruit. The small green flowers and leaflike clusters of fruits are very inconspicuous. The type locality has been visited by many biologists in recent years but this appears to be the plant's first gathering. The collection was made in April on the wet northwest side of Volcan Baru near Cerro Punta. LiTERAiURE Cited DArcy, W. G. 1975 [1976]. Compositae. In R. E. Woodson, Jr. & R. W. Schery, Flora of Panama. Ann. Missouri Bot. Garden 62: 835-1322. NEW SPECIES OF CONNARUS (CONNARACEAE) FROM PERU 1 Enrique Forero Abstract Two new species of Connaraceae from Peru — Connarus hracteoso-villosus Forero and Connarus elsue Forero — are described and illustrated. In the course of studying the Neotropical species of Connaraceae I have found several novelties. This paper includes two interesting species of Connarus from Amazonian Peru. Connarus bracteoso-villosus Forero, sp. nov. — Fig. 1. Species insignis axibus inflorescentiae dense villosis et bracteis elongatis (usque 0.5 cm longis) dense villosis ab congeneribus distincta. Shrub 1 m tall, branchlets grey; conspicuously lenticellate, the lenticels small. Leaves imparipinnate, 7-9 foliolate; pulvinulus 5-7 mm long; petiole cylindric, glabrous, (7-)8 cm long; rachis cylindric, glabrous, (7.5-)11.5 cm long; leaflets 5-11.5 cm long, 2.5-4.5 cm wide, elliptic, chartaceous, glabrous, rounded to angustate at the base, acuminate at the apex; petiolule 4-5 mm long; midrib somewhat impressed above, prominent below, the lateral veins 9-10 pairs, di- verging from the midrib at angles of 70°, anastomosing diffusely near the margin, slightly prominent below. Inflorescence paniculate, congested; rachilla up to 20 cm long, characteristically densely villous; bracts characteristically long, up to 5 mm, densely villous. Flowers pedicellate; pedicel 5 mm long, villous, thick; sepals 3-3.3 mm long, 1.8-2 mm wide, ovate-elliptic, conspicuously punctate, villous without, villous towards the apex within, the apex acute, densely villous; petals 4-4.2 mm long, 1.3 mm wide, subspatulate, densely glandulose without, glandulose in the upper half within, the margin glandulose, and with 1-3 dots in some (usually 2) of the petals, the rest epunctate; stamens 10, 5 short, 2.2 mm long, 5 long, 3 mm long, glandulose, the tube ca. 0,5 mm long, the anthers globose, ca. 0,4 mm diameter; ovary densely villous. Fruit unknown. Type: Peru, depto. huanuco: Provincia Pachitea, Distrito Honoria, Bosque Nacional de Iparia, region de ^'bosque seco tropical' ' a lo largo del Rio Pachitea cerca del campamento Miel de Abeja (1 km arriba del pueblo de Tournavista o unos 20 km arriba de la confluencia con el Rio Ucayalf), en bosque bajo en la carretera Miel de Abejas, arbusto 1 m, flores amarillas palidas, 5 Jun. 1967, Schunke 2040 (F, holotype). * Acknowledgments are due lo the John Simon Guggenheim Memorial Foundation for financial support. Margaret Kurzius prepared the excellent illustration of Connarus hracteoso-villosus. ^ Profesor Asociado, Instituto de Ciencias Naturales, Universidad Nacional, Aparlado 7495, Bogota, Colombia. Ann. Missouri Bot. Card. 68: 218-221. 1981. 0026-6493/8 1/02 1 8-022 1/$0.55/0 1981] FO R ERO~CC)NNA R US 219 2 mm Figure 1. Conarus hracteoso-viUosus Forero, Schunke 2040 (F).^A. Habit (xVi). — B. Flower (x6). — C. Flower section (x6). — D. Petal (x7). This species is known so far only from the type locality in the Department of Huanuco, Peru, occurring in low forest at altitudes between 300 and 400 m. The affinities of this very distinct species are uncertain. It can be easily dis- tinguished from any other species of the genus by the very long bracts which are densely villous and by the densely villous pubescence of the inflorescence. 220 ANNALS OF THE MISSOURI BOTANICAL GARDLN [Vol . 68 NEGATfVE No. 102 1 1 PLANTS OF PERU DF.PT, SAN HARTIN: Prov. S*n H*rtln: Ro«d from Puente Colonbla to Shaoala. «inn^ Mo Mayo. Alt. 280 Shrub 3 ■. tall on tteap bank in full »un rrulta graen. ,..181S2S2 VnLD UUSBUU OF MTUtAL HISTORr CoB. Tlnotli7 Plowmu -_ J- !>ita: 30 April 1976 Figure 2. Holotype of Connarus elsae Forero, Plowman 6017 (F). (Photo: Dr. Maria Lehron Luleyn, New York Botanical Garden.) Connarus elsae Forero, sp. nov. — Fig. 2. Ab speciehus affinis C. puncfatus Planch, petalis epunctatis foliolorum basium angustata vel rotundata, stipite fructis brevioribus differt. 1981] FORllRO—CONNARUS 221 Small tree or shrub, up to 8 m tall; branchlets puberulous, conspicuously lenticellate, the lenticels small. Leaves imparipinnate, (3-)5(-7)-foliolate; pulvin- ulus 3-5 mm long; petiole glabrous, (2-)3-7,5 cm long; rachis glabrous, (l-)2.5- 11.5 cm long; leaflets (5-)7-12(-15) cm long, (2.4-)2.7-5 cm wide, elliptic or narrowly elliptic, chaitaceous, glabrous, attenuate or somewhat rounded at the base, acuminate at the apex; petiolule 2-5 mm long; midrib impressed above, prominent below, the lateral veins 8-10 pairs, diverging from the midrib at angles of 45°-60°, anastomosing irregularly near the margin, slightly prominulous below. Inflorescence paniculate; panicules fasciculate in groups of 1-3 per axil; rachilla (2-)4-9(-15) cm long, tomentose; bracts conspicuous, black dotted. Flowers ped- icellate; pedicel 3^ mm long, puberulous; sepals 2.5-3 mm long, 1-1.5 mm wide, ovate, sparsely punctate, puberulous without, glabrate within, the apex obtuse; petals 3.5-5 mm long, 1.3-1.5 mm wide, obovate-lanceolate, epunctaie, sparsely puberulous and glandulose without, sparsely glandulose within; stamens 10, 5 short, 1.5-2 mm long, 5 long, 2-3.2 mm long, with few glandular hairs towards the apex, the anthers globose, 3-4 mm in diameter. Follicle stipitate; stipe 2-5 mm long; fruit 2-2.5 cm long, 1.4-1.7 cm wide, glabrous or glabrate, outside, densely puberulous inside; calyx in fruit persistent. Type: Peru. dept. san martin: Provincia San Martin, road from Puente Colombia to Shapaja, along Rio Mayo, shrub 3 m tall on steep bank in full sun, fruits green, 30 Apr. 1976, Plowman 6017 (F, holotype). Additional Collections Examined: Peru, san MARrfw; Juan Jui, alto rio Hualiaga, Oct. 1934, ^^///^' 3855 (BM, F, GH, K, MO, NY, US). Tarapoto, Apr. 1856, Spnae s.n. (K). Alto Rio Hualiaga, Dec. 1929, Williams 5670 (F, US), 5733 (F, G). Rio Mayo, Tarapoto, 14 Dec. 1929, Williams 6209 (BM, F), 6214 (F). Rumizapa, near Tarapoto, Dec. 1929, Williams 6808 (F, US). This species is known from the Department of San Martin, Peru, where it occurs above 280 m elevation. It has often been confused with Connarus punc- tatus from which it differs by the epunctate petals, the shorter stipe, and also by the less numerous and somewhat smaller leaflets. NOTES CHROMOSOME NUMBERS OF MISCELLANEOUS ANGIOSPERMS Chromosome numbers are reported in Table 1 for 20 collections representing 19 species. First reports are indicated by an asterisk preceding the name. Counts agreeing with those previously reported by other authors are not discussed any further. Aponogetonaceae Raven (1975) has suggested a base number of x = 8 for the family. Apono- i^eton crispus is the third chromosomally known species based on this number, whereas the high polyploid A. natans has been reported to have the apparently aneuploid numbers Iti = 76, 78, 92 (Misra, 1972; Sharma & Chatterjee, 1967). ErIOCAUI ACEAE This family remains very poorly known cytologically with only the genus Enociuthn previously known. Raven (1975) suggested a base number v = 8 for Eriocaiiloii, although numbers based on .v = 9 and 10 have also been reported for four species (Cave, 1967; Erlandsson, 1942; Hedberg 8l Hedberg, 1977; Mehra & Sachdeva, 1971). The two current counts for Syngonanthus provide further support for a base number x - 8 for the family but also suggest that the family is cytologically diverse and well worth intensive cytological investigation. Gesneriaceae All the numbers here reported for species of this family agree with numbers previously reported for the respective genera. However, the report of tetraploidy 1971) although not previously unknown in the genus. Wiehle Gramineae Ergrostis hypnoides, previously known from a single diploid count in the tropical portion of its range (Davidse & Pohl, 1972) is now shown to be tetraploid in the temperate portion of its range. This suggests a tropical origin for this species but more extensive sampling is obviously necessary. Solanaceae The chromosome number for the Panamanian endemic Bnmfelsia clnycr agrees with that previously reported for five other species and provides furthe evidence for a base number of a = 11 for the genus (Plowman, 1973). Ann. Missouri Bot. Card. 68: 222-223. 1981. I 1981] NOTES 223 Tabi h I, Chromosome numbers and voucher information for miscellaneous angiosperms. Aponogetonaceae * Apono^eton crispus Thunb. n = 24. Sri Lanka: Central Province, Nuwara Eliya District, Horton Plains, Davidse 7616 (MO). BUTOMACEAE Linmocharis jiava (L.) Buchn. n = 10. Venezuela: Guarico, ca. 50 km N of San Fernando de Apure, Davidse, Af^ostini iSc A^ostini 3820 (MO). Eriocaui aceae ""Sxngonanthus caulcscens (Poir.) Ruhl. n = 13. Venezuela: Guarico, ca. 32 km SSE of Calabozo, Davidse 3771 (MO). *S\ngonanfhus xeranthemoides (Bong.) Ruhl. /; = 16. Venezuela: Apure, between the Rio Cin- aruco and the Galeras de Cinaruco, Davidse & Gonzalez I2279A (MO). Gesneriaceae *Achinienes erecta (Lam.) H. P. Fuchs. n = 22. Jamaica: Portland Parish, MuriePs Rock, between Section and Hardware Gap, Davidse & Proctor 3244 (MO). Achimenes hmgiflora DC. n = \\. Nicaragua: Managua, 27 km S of Managua along Hwy. 8, Davidse & Pohl 2382 (MO). Alloplectus tetragorius (Oerst.) Hanst. n = 9. Cosla Rica: Cartago, 2 km E of Muneco, Davidse & Pohl 1679 (MO). Columnea linearis Oerst. 2n = 18. Nicaragua: Chontales, 8 km E of Villa Somoza along Hwy. 7, Davidse 2734 (MO). *Corvtoplectus congestus (Find.) Wiehler. n = 9. Venezuela: Merida, 66 km NE of Merida along " the Merida- Azulita Rd., Davidse 3239 (MO). Gesneria aucaulis L. w = 14. Jamaica: St. Catherine Parish, 2.1 mi SE of Bog Walk, Davidse 3270 (MO). Kohleria hirsuta (H.B.K.) Regel. // = 13. Venezuela: Distrito Federal, 7.0 km SW of Carayaca, Davidse 2893 (MO). Kohleria tuhiflora (Cav.) Hanst. n = 13. Nicaragua: Rivas, La Cuesta, 3 km NE of San Juan del Sur, Davidse & Pohl 2284 (ISC, MO). Kohleria tuhiflora (Cav.) Hanst. n = 13. Venezuela: Barinas, 25 km NW of the Merida inter- section (just outside of Barinas) along road to Merida, Davidse 3187 (MO). ""Rhxtidophxllnm tomentosuni (L.) Mart, ex G. Don. // = 14. Jamaica: St. Thomas Parish, just NW of Cedar Valley along road to Arntully, Davidse 3262 (MO). Sinningia incarnata (Aubl.) D. Denh. /; = L3. El Sai vador: Libertad, 3 km E of La Libertad along Hwy. 2, Davidse <^ Pohl 2056 (MO). Gramineae ""Eragrostis hypnoides (Lam.) B.S.P. // = 20. United States: Missouri, St. Charles Co., ca. 1 mi SW of the bridge at the intersection of U.S. Hwy. 40 and the Missouri River, Davidse 3553 (LSC, MO). Eragrostis pectinacea (Michx.) Beauv. /; = 30. United States: Missouri, St. Charles Co., ca. 1 mi SW of the bridge at the intersection of U.S. Hwy. 40 and the Missouri River, Davidse 3554 (ISC, MO). Sporoholiis asper (Michx.) Kunth. // = 27. United Si ates: St. Louis, Davidse 3671 (MO). Lauraceae Cassxthafilliforniis L. n = 24. Belize: Belize District, -M mi in from Western Hwy. on Fergerson Bank, Dwyer 12803 (MO). Solanaceae *Brunfelsia dwyeri D'Arcy. In - 22. Panama: Panama, Cerro Jefe, Gentry 4883 (MO). * First report or different count. LiTERAi URE Cited Cave, M. S. 1967. In Documented chromosome numbers of plants. Madrono 19: 134-L36. Davidse, G. & R. W. Pohl. 1972. Chromosome numbers, meiotic behavior, and notes on some grasses from Central America and the West Indies. Canad. J. Bot. 50: 1441-1452. 224 ANNALS OF THE MISSOURI BOTANICAL GARDLN [Vol. 68 Erlandsson, S. 1942. The chromosome numbers of three Eriocaulon species. Arkiv. Bot. 30B(1): Hedberg, I. & O. Hedberg. 1977. Chromosome numbers of afroalpine and afromontane angio- sperms. Bot. Not. 130: 1-24. Mehra. F. N. & S. K. Sachdeva. 1971. In lOPB chromosome reports XXXIII. Taxon 20: 609- 614. MisRA, M. p. 1972: Cytological studies in some Indian Potumo^eton and Apono^eton species. Bull. Bot. Soc. Bengal 26: 47-51. Plowman, T, C. 1973. The South American species of BrunfelsUi (Solanaceae). Ph. D. thesis. Harvard Univ., Cambridge, Massachusetts. Sharma, A. K. & T. Chatterjee. 1967. Cytotaxonomy of Helobiae with special reference to the mode of evolution. Cytologia 32: 286-307. Raven, P. H. 1975. The bases of angiosperm phylogeny: cytology. Ann. Missouri Bot. Card. 62; 724-764. WiEHi PR, H. 1971. Chromosome numbers in some American Gesneriaceae. Baileya 18: 118-120. Gerrit Davidsc, Missouri Botanical Garden, Post Office Box 299, St, Louis, Missouri 63166, A NEW SPECIES OF HERNANDIA (HERNANDIACEAE) FROM PANAMA Hernandia hammelii D'Arcy, spec. nov. Arbor 15-20 m alta, ligno leni, virguli crassis, cicatrices petiolorum circulares conspicuas ferenti, ramunculis parum gracilioribus nigris siccantibus; folia ovata vel eiliptica, 6-8 cm longa, 2.5^ cm lata, apice breve accuminata, basim rotundata, glabra, coriacea, costa conspicua, nervis lateralibus irregularibus, arcuatis, 2-5 utrinque supra inconspicuis subtus evidentibus siccantibus, pari basali oppositi, petiolis rectis, gracilis, angulatis siccantibus; flores bracteolis anguste ovatis subtentis. Flo- res fructusque evoluti non visi. Type: Panama, cocle province: continental divide on road to Coclesito, 1600 ft, Hammel 7205 (MO). Tree 15-20 m tall; wood soft, white; young branches with conspicuous circular leaf scars, the twigs drying slightly narrower, dark. Leaves ovate or elliptical, 6- 8 cm long, 2.5^ cm wide, apically short acuminate, blunt, basally rounded, glabrous, coriaceous, the costa conspicuous, the lateral nerves irregular, 2-5 on each side, arcuate, ascending, obscure above, drying evident beneath, the basal pair opposite; petioles slender, drying angled, 2-3 cm long. Inflorescence cymose, bracteate, mostly covered with minute grayish trichomes, the flowers subtended by small, narrowly ovate, costate, caducous bracteoles. Flowers with a basal cupule which completely envelopes the young developing fruit. ■ This species is singular in its small leaves which somewhat resemble those of H. cuhensis Griseb. of Cuba, but in that species the leaves are much broader and the petioles much longer and more widely spaced. This is the only species of Hernandia known on the American mainland with such small, crowded and uniform-appearing leaves. The bracteoles in this species are much like those in other hernandias, but may be sooner caducous. The cupule at the base of the developing fruit appears to envelop the fruit to a much greater degree than in other species. How it is placed in mature fruit is unknown. Ann. Missouri Bot. Card. 68: 224-225. 1981. 1981] NOTES 225 The type locality is an area of poorly collected disappearing wet montane forest of low elevation in central Panama. Assisted by National Science Foundation Grant DEB 79-22192. W, G. D'Arcy, Missouri Botanical Garden, Post Offi Mis PECTIS LINIFOLIA (COMPOSITAE: TAGETEAE) ADDED TO THE FLORA OF PANAMA In the treatment of Pectis for the Flora of Panama. Keil (1975) noted that the widespread P, linifolia L. was known to occur both to the north and to the south of Panama but had never been collected in that country. A recent collection of P. linifolia var. linifolia has added this species to the known flora of Panama: Colon: Punta Chame, D'Arcy 10218 (MO, OS). This taxon would key to P, elongata var. oerstediana in the Flora of Panama treatment. These taxa can be easily distinguished, however. Pectis linifolia var. linifolia has a pappus of stiff, reflexed smooth awns, and its herbage is not scent- ed. Pectis elongata var. oerstediana has a pappus of slender erect scabrous bristles, and the herbage has a strong odor resembling that of lemons or stinkbugs. Synonymy, descriptions, illustrations and a range map for Pectis linifolia have been published by Keil (1978), Literature Cited Keil, D. J. 1975. VI. Tageteae. In R. E. Woodson, Jr. & R. W. Schery and collaborators, Flora of Panama. Ann. Missouri Bot. Gard. 62: 1220-1241. , 1978. Revision of Pectis sect. Pectidium (Compositae: Tageteae). Rhodora 80: 135-146. — David y. Keil, Biological Sciences Department, California Polytechnic State University, San Luis Obispo, California 93407, Ann. Missouri Bot. Gard. 68: 225. 1981. VALIDATION OF THE NAME AULONEMIA PATRIAE POHL (GRAMINEAE: BAMBUSOIDEAE) Aulonemia patriae Pohl (1980: 68), a new species of bamboo, was described on the basis of vegetative and fruiting material from the same colony, at Alto del Roble, Costa Rica. Since the fruiting material was long past maturity, and lacked good foliage, I indicated that two specimens, one vegetative and the other fruiting, from this colony should be regarded as syntypes. This procedure seemed to me to be available under Article 7.7 of the International Code of Botanical Nomen- clature (Stafleu et al., 1978). Other taxonomists, including Dr. Stafleu, feel that it is obligatory to designate one of the specimens as a holotype or lectotype, although neither one displays all of the differential characteristics of the species. I therefore designate the following specimen as the lectotype of A. patriae Pohl, Ann. Missouri Bot. Gard. 68: 225-226. 1981. 226 ANNALS OF THE MISSOURI BOTANICAL GARDEZN IVoi . 68 following the procedure described in the Guide for the Determination of Types, 4C(Stafleuet al., 1978). Lectotype: Costa Rica. prov. heredia: Alto del Roble, Pohl & Gahel 13577 (ISC). Literature Cited Pohl, R. W. 1980. Flora Costaricensis: Family #15 Gramineae. Fieldiana (Botany), n.s. 4: 1-608. Si AhLEU, F. A. et aL (editors). 1978. International Code of Botanical Nomenclature. Regnum Veg. 97: 1-456. Bohn, Scheltena & Hokkema, Utrecht. Richard W. PoliL Department of Botany, Ion a Slate University, Ames, Iowa 50011. SETARIA VARIIFOLIA (SWALLEN) DAVIDSE, A NEW COMBINATION (GRAMINEAE: PANICOIDEAE) Panicum variifolium Swallen, Publ. Carnegie Inst. Wash. 436: 345. 1934 is a distinctive species of the Yucatan Peninsula that should be treated as a species of Setaria, Setaria variifolia (Swallen) Davidse, comb, nov., has a single bristle subtending the terminal spikelet on each branch and occasionally some of the lower spikelets as well. This leaves no doubt as to its generic affinity with Setaria. Its slender, elliptic, acuminate spikelets with the upper floret only weakly rugose suggests a close relationship with subgenus Ptychophyllum. However, the blades appear to be flat rather than plicate as in Ptychophyllum. This plus the condensed, sparingly branched, few-flowered inflorescence suggests affinity to subgenus Paiirochaetium . Further observations, especially field studies on the form of the blades, should eventually clarify these difficulties. — Gerrii Da\icise, Missouri Botanical Garden, Post Office Box 299, St. Louis, Missouri 63166. Ann. Missouri Bot. Card. 68: 226. 1981. RECOGNITION OF BRACHISTUS (SOLANACEAE) The genus Brachistus Miers was reduced to a section of Witheringia L'Her. by Hunziker (1969). Further consideration of this group of plants, especially with regard to the calyx, which is dentate or lobate instead of truncate and hence has quite different vasculature, has led to belief by ourselves and by Hunziker that Brachistus should be restored to generic rank. Support for this concept will appear in a later publication. Brachistus, in our view, embraces only the species listed here and not the many others mistakenly placed in the genus in the past. Ann. Missouri Bot. Gard. 68: 226. 1981. 1981] NOTES 227 Miers, Ann. Mag. Nat. Hist., ser. 2, 3: 262. 1849. lectotype: B. ifolius (H.B.K.) Miers, based on Witheringia stramonifolia H.B.K. Pubescent trees or shrubs. Inflorescences usually many-flowered, sessile or short pedunculate. Flowers small, 5-merous; calyx 5-dentate or lobate, the lobes sometimes short; corolla 5-lobed, the lobes longer than the tube; anthers oblong, sometimes apiculate. Fruit a berry, subglobose, enclosed one-half or more by the accrescent calyx; seeds fabiform. (Adapted from Hunziker, 1969). Brachistus stramonifolius (H.B.K.) Miers, Ann. Mag. Nat. Hist., ser. 2, 3: 263. 1849. Southern Mexico to Panama. Brachistus nelsonii (Fern) D'Arcy, J. L. Gentry & Averett, comb. nov. Athenaea nelsonii Fern., Proc. Amer. Acad. Arts 35: 567. 1900. type: Mexico, Chiapas, Nelson 3395 (GH, holotype; US, isotype). Southern Mexico to Panama. Brachistus affinis (Morton) D'Arcy, J. L. Gentry & Averett, comb. nov. Athenaea affinis Morton, Contr. Univ. Mich. Herb. 4: 24. 1940. type: Belize, Lundell 6452 (US, holotype; A, F, US, isotypes). Belize. Key to Species a. Leaves acute or obtuse to nearly truncate at the base B. stramonifolius aa. Leaves cordate at the base. b. Pubescence of long, glandular hairs, at least some of the hairs 1 mm long; corolla marked with purple B. nelsonii bb. Pubescence of short, eglandular hairs, the hairs less than 1 mm long; corolla not marked with purple B. affinis Literature Cited Hunziker, A. T. 1969. Estudios sobre Solanaceae V. Contribucion al conocimiento de Capsicum y generos afines {Witheringia, Acnistus, Athenaea. etc.) Kurtziana 5: 101-179. — W. G. D'Arcy, Missouri Botanical Garden, Post Office Box 299, St. Louis, Missouri 63166, U.S.A., Johnnie L. Gentry, Jr., University of Arkansas Museum, 338 Hotz Hall, Fayetteville, Arkansas, 72701 , U.S.A. and John E. Averett, f Missouri, 8001 Natural Bridge Road, St. Louis, Mis U.S.A. 1982 AETFAT CONGRESS The 1982 AETFAT Congress will take place from 19-22 January 1982 at the CSIR Conference Centre, Pretoria, Republic of South Africa. The congress is organized by the Botanical Research Institute of the Department of Agriculture and Fisheries, Secretariat of the "Association pour TEtude Taxonomique de la Flora d'Afrique Tropicale" (AETFAT) and the "South African Association of Ann. Missouri Bot. Gard. 68: 227-228. 1981. 228 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Botanists'' (SAAB), assisted by the Symposium Secretariat of the '^Council for Scientific and Industrial Research'' (CSIR). A symposium on 'The Origin, Evolution and Migrations of African Floras'' will be a major feature of the congress. The aims of this symposium are to supply a brief overview of the present state of knowledge on the evolution of African floras and vegetation, particularly in the light of the movements of continents and of climatic changes: to link these facts with the distribution, composition and diversity of present-day floras and vegetation in an attempt to unravel their past history. Whereas only the more recent fossil deposits have a strong bearing on present-day floras, the accounts on older deposits are being deliberately included to be able to present as complete a picture as possible, and to reflect current research activity in this field, A number of the topics included in the proposed symposium were discussed during the Symposium held at the Missouri Botanical Garden in 1977 on the phytogeography of Africa. The proceedings of this symposium were published in the ANNALS OF THE MISSOURI BOTANICAL GARDEN Vol. 65: 367-589 (1978) and have stimulated general interest. It is felt that these topics could well be approached from a more regional point of view in 1982 so that more detailed information will become available for many of the African countries. All inquiries concerning the congress should be addressed to: The Symposium Secretariat S.229, CSIR, P.O. Box 395, Pretoria 0001, South Africa. Telephone: International + 27 12 86-9211, extension 2077 (Elsie Coetzee) or 2063 (Ann van Dyk). Telex: SA 3-630. FLORA MESOAMERICANA The Missouri Botanical Garden, the Instituto de Biologia of the Universidad Nacional Autonoma de Mexico and the British Museum (Natural History) are collaborating to produce a synoptic Flora of the estimated 16,000 vascular plants of Mesoamerica. This Flora will serve as a concise identification manual of Meso- american plants. Such a project has been discussed in the botanical community for many years and is of such a magnitude that it would, if it were to be successful, require the cooperation of many institutions within Mesoamerica and outside of it. With this basic premise the three organizing institutions began discussions for a Flora Mesoamericana Project in 1979 and appointed G. Davidse (MBG), M. Sousa (UNAM), A. O. Chater (BM) and J. C. Humphries (BM) as organizers and editors of this project. From these initial discussions a plan of action was developed. In early 1980 a broad outline of the plan of action was distributed to individuals and institutions from Mesoamerica, the U.S. and Europe his- torically active in Mesoamerican botanical studies. Comments were solicited and invitations were extended to attend an organizational and planning meeting sponsored by the three organizing institutions in St. Louis, 14-15 July 1980. Twenty-five persons representing 13 institutions attended: British Museum Ann. Missouri Bot. Card. 68: 228-229. 1981. 1981) NOTES 229 (Natural History), California Academy of Sciences, Colegio Superior de Agri- cultura Tropical, Duke University, Field Museum of Natural History, Herbario Nacional de Colombia, Herbario Nacional de Nicaragua, Instituto Politecnico Nacional, Missouri Botanical Garden, National Science Foundation, Universidad Nacional Autonoma de Honduras, Universidad Nacional Autonoma de Mexico and Universidad de Panama. General approval and support were given to the need for such a project and wide agreement was reached on matters of procedure. The Flora will be published in Spanish by the Universidad Nacional Autonoma de Mexico and will appear in seven volumes over a period of sixteen years. As much of the Flora as possible will be written by experts, the remainder by staff of the organizing institutions. It will summarize existing in- formation and incorporate much new research. Keys and descriptions will be included at all levels. Types and selected specimens will be cited, and sufficient synonymy will be given to correlate the Flora with previously published work in the area. A concise statement about distribution and ecology will also be included. In general the style of the Flora will be very similar to that of Flora Europaea, Me Wh draw the exact boundaries for the Flora was largely pragmatic. Having the Panamanian-Colombian frontier serve as the southern boundary seems to be logically dictated on a geographical basis. In addition, since the Flora of Panama ends at that border and since the new Flora de Colombia project will adequately cover the area east of this border, this boundary seems to be a logical one. The northern boundary clearly should be in the Isthmus of Tehuantepec region. A very good phytogeographical one seems to be follow- ing (M. Sousa and D. Breedlove, pers. comm.): from the Gulf of Tehuantepec along the Tehuantepec River, to the Sierra Mixe to the 300 m contour, west along the Sierra Mixe and the Sierra de Juarez to the Papaloapan River, and along this river to the Gulf of Mexico. However, it was strongly and unanim- ously agreed at the St. Louis meeting that such a boundary would be highly impractical for specimen handling and distribution plotting and that using the Veracruz-Oaxaca and Tabasco-Chiapas borders would at once be practical for specimen handling and at the same time include virtually all of truly tropical Mexico. It was realized that a case could well be made for extending the northern boundary into Veracruz and perhaps into Tamaulipas. However, a significant number of basically northern species would then be included (Rzedowski, pers. comm.). Thus a practical political boundary was chosen with the provision that for taxa extending just beyond this boundary and which seem to have a truly southern tropical derivation, authors could decide to include these taxa in the Flora on a case-by-case basis. A concurrent collecting program has been started by the three organizing institutions in conjunction with other botanists and institutions in Mesoamerica. It is aimed at significantly strengthening current attempts to collect in poorly or never-collected areas before the surviving natural vegetation is completely destroyed. A knowledge of these areas is necessary for comprehensive coverage 230 ANNALS OF THH MISSOURI BOTANICAL GARDEN [Vol. 68 in the Flora. Specialist collecting by persons contributing family treatments is also scheduled to the extent possible by available funds. When completed, the Flora will comprise a data base for the floristics and taxonomy of the area. From such a firm taxonomic framework it will be possible to plan more effectively detailed taxonomic studies in the area, as well as broader programs in conservation and land use. Additionally, we foresee that the production of local and special purpose floras of individual countries or areas and covering such topics as economic and medicinal uses will be a major spin-off from the Flora Mesoamericana, The organizers and editors welcome inquiries and correspondence concerning any aspect of the project — Gerrit Davidse (Organizer and Editor), Missouri Botanical Garden, P.O.B. 299, St. Louis, MO 63166, U.S.A.; Mario Sousa (Organizer and Editor). Instituto de Biologia, Herbario Nacional, U.N. A.M., Apartado Postal 70-367, Mexico 20, D.F., Mexico; C. J. Humphries (Organizer), and A. O. Chater (Editor), Botany Department, British Museum (Natural History), Cromwell Road, London SW7 5BD, U.K. THE 1981 JESSE M. GREENMAN AWARD The 1981 Jesse M. Greenman Award has been won by William R. Buck for his publication '^A Generic Revision of the Entodontaceae'' (J. Hattori Bot. Lab. 48: 71-159. 1980.). This monographic study is based on a Ph.D. dissertation from the Department of Botany, University of Michigan, Ann Arbor. The Greenman Award, a cash prize of $250, is presented each year by the Alumni Association of the Missouri Botanical Garden. It recognizes the paper judged best in vascular plant or bryophyte systematics based on a doctoral dissertation which was published during the previous year. Papers published during 1981 are now being considered for the 15th annual award, which will be presented in the summer of 1982. Reprints of such papers should be sent to: Greenman Award Committee, Department of Botany, Missouri Botanical Garden, P.O. Box 299, St. Louis, MO 63166, U.S.A. In order to be considered for the 1982 award, reprints must be received by 1 July 1982. Ann. Missouri Bot. Card. 68: 230. 1981. I98IJ NOTES 231 ERRATA Thorp, Robbin W. 1979. Structural, behavioral, and physiological adaptations of bees (Apoidea) for collecting pollen. Ann. Missouri Bot. Gard. 66: 788- 812. P. 791, in Table 1, should read Hylaeinae, not Hyaelinae. P. 797, Fig. 12 caption, should read Paratetrapedia (Paratetrapedia) not ^ay;- malopsis, P. 809, Grinfeld, E. K. should read 1962, vol. 41 not 1969, vol. 48. P. 811, Pasteels, J. M. & J. J. Pasteels should read 1977, Arch. Biol. 88: 441- 468, not 1979 (in press). P. 811, Thorp, R. W. & D. L. Briggs should read 1980, 53: 166-170, not 1979 (in press). I thank J. Neff, Austin, Texas, for bringing the first errors to my attention. — Rohhin W. Thorp, Department of Entomology, University of Ccdifornia, Davis, California 95615. Ann. Missouri Bot. Gard. 68: 231. 1981. The previous issue of the Annals of the Missouri Botanical Garden, Vol. 67, No. 4, pp. 819-1069, was published on 29 June 1981. Index to plant chromosome numbers 1978 Chromosomal information has become increas- ingly important in recent years in studies of plant systematics and evolution. Nearly all modern tax- onomic monographs and revisions include chromosomal data and use this information in analyzing species or generic relationships or evolu- tionary patterns. Chromosome counts accumulate at the rate of over 5,000 a year, scattered in the literature. They are reported in brief articles, as major compilations for families or genera, and in revisions and monographs in numerous botanical and other jour- nals and in books. Index to Plant Chromosome Numbers 1975-1978 is the most recent volume in the series of catalogues of plant chromosome numbers that was begun 25 years ago as a means of bringing together this scattered in- formation and making it readily available. The information for the Index is compiled by an international committee of 24 and collated and edited by Peter Goldblatt, Missouri Botanical Garden. It includes some 25,000 reports published from 1975 through -^0 and will be indispensible to those engaged in a variety of botanical studies ranging from systematics and evolution to plant breeding, forestry, and horticulture. Counts for all plant groups— algae, fungi, rvophytes, pteridophytes, and spermatophytes— are included and arranged alphabetically by family, genus, and species. Each count includes the name of the taxon as used in the original report, the number reported, and reference to the original report. References are contained in a bibliography of about 2,000 citations. Published as Monographs in Systematic Botany from the Missouri Botanical Garden Vol. 5. vii + 553 pages. September 1981. Paper. r Order your postpaid copy from: Department Eleven Missouri Botanical Garden P.O. Box 299 St. Louis, MO 63166 U.S.A. Pnce: $15,00. postpaid. quest^ed h ?^ ^^^ ^^ ^^^^^ ^^ ^° "^^ ^^^ ^^'^^^ ^^^^ ^^'^^ °^ ^ photocopy of it. Please include information re- quirinfi •*' '*" ^^^ ^^^^^ on separate paper. Orders should be prepaid: a $1 .00 fee will be added to orders re- g invoicing, and no shipments are made until payment is received. Pleas e send copy(ies) of Index to Plant Chromosome Numbers 1975-1978 @ $15.00 each to lame n Payment enclosed. Addresi n Send invoice ($1 .00 fee will be added to total). PoM*J Code Country ^ I ^ "■ {Contents continued from front cover) Four New Species of Dioscorea from Amazonian Peru Franklin Ayala F ^ ^^^ Systematics, Phylogeny and Evolution of Dietes (Iridaceae) Peter Goldblatt ^ 13- Additions to the Ericaceae of Panama Robert L. Wilbur c& James L. Liiteyn 1— ^^ ^-^^ Additional Panamanian Species of Burmeistera (Campanulaceae: Lobelioi- deae) Robert L. Wilbur _: I-- ^^^ Sapotaceae of Panama George E. Pilz - The Mexican and Central American Species of Adelobotrys (Melasto- mataceae) Frank Almeda, Jr. .._ __:__ . ^ i New Taxa from the Uplands of Western Panama Barry E. Hammel & W. G. D'Arcy ......... ._._...... ^^^ I New Species of Connarus (Connaraceae) from Peru Enrique Forero 5 NOTES Chromosome Numbers of Miscellaneous Angiosperms Gerrit ^ Davidse ^^ ^^ ^^ ^^ ^^^ ^^ ^^ ^^ ^^ ^^ ^^ ^^^ ■ V -WW ^^ T^ ^^ ^*^ ^^* ^^ ^^ -^K ^^ ^^ •^ ^^ ^^ ^^ A New Species of Hernandia (Hernandiaceae) from Panama W^- ^' D'Arcj ... : - ^^ i Pectis linifolia (Compositae: Tageteae) Added to the Flora of Panama David J. Keil ^ Validation of the Name Aulonemia patriae Pohl (Gramineae: Bambusoideae) Richard W. Pohl i Setaria variifolia (Swallen) Davidse, a New Combination (Gramineae: Panicoideae) Gerrit Davidse - " + Recognition of Brachistus (Solanaceae) W. G. D'Arcy, Johnnie I- ^^^ Gentry, Jr. & John E. Averett *" i 227 1982 AETFAT Congress ' 228 Flora Mesoamericana " 230 The 1981 Jesse M. Greenman Award 231 Errata _ FTHE PLUME 68 1981 NUMBER 2 ■•-;v..- *V\'--.'^ ■'■■:■■ ^:'---- RDE I ^ The > CONTENTS REPRODUCTIVE STRATEGIES IN PLANTS AND ANIMALS Twenty-seventh Systematics Symposium. CJJspersal Versus Gene Flow in Plants Donald A. Levin 233 Evolution of Sexual Systems in Flowering Plants K. S. Bmva and J' H. Beach 254 ^n the Evolution of Complex Life Cycles in Plants: a Review and an Ecological Perspective Mary F. Willson Eloral Rewards: Alternatives to Pollen and Nectar Beryl B. Simpson and 275 John L. Neff ^^^ Geographical Aspects of Bird-Flower Coevolution, with Particular Refer- ence to Central America F. Gary Stiles he Distributional Patterns of New Worid Nectar-Feeding Bats 323 Karl F. Ko opman The Ene rgetics of Pollination Bernd Heinrich 352 370 S r VOLUME 68 1981 NUMBER 2 5 OF THE mm The Annals contains papers, primarily in systematic botany, contributed from the Missouri Botanical Garden. Papers origi- nating outside the Garden will also be accepted. Authors should write the editor for information concerning arrangements for pub- lishing in the Annals. Editorial Conlmittee Gerrit Davidse, Editor Missouri Botanical Garden John D. Dwyer Missouri Botanical Garden ir St. Louts University J Peter Goldblatt Missouri Botanical Garden Nancy Morin Missouri Botanical Garden Published four times a year by the Missouri Botanical Garden, St. Louis, Missouri 63110. ISSN 0026-6493 For subscription information contact the Business Office of the Annals, P.O. Box 368, 1041 New Hampshire, Lawrence, Kansas 66044. Subscription price is $45 per volume U.S., Canada, and Mexico, $50 all other countries. Four issues per volume. Second class postage paid at Lawrence, Kansas 66044 © Missouri Botanical Garden 1982 OF THE VOLUME 68 1981 NUMBER 2 DISPERSAL VERSUS GENE FLOW IN PLANTS I Donald A. Levin^ Abstract The notions that gene flow in plants is restricted and that neighborhood size and area are small are based upon data on pollination, and pollen and seed dispersal. However, neighborhood size and area estimates from these data incorporate several assumptions: (1) Pollinator flight distance is repre- sentative of pollen dispersal distance; (2) All pollen picked up by pollinators from one plant is deposited on the next one visited; (3) PoUinator flights between plants are random in direction; (4) Pollen-pistil compatibility is independent of the proximity of egg and pollen parents; (5) Seed viability is indepen- dent of the proximity of egg and pollen parents; (6) Seed production is independent of the proximity of seeds to the source; (7) The fitness of plants is independent of the distance between egg and pollen parent. These assumptions have been found to be unwarranted in many instances leading to the following conclusions: (1) Pollen carry-over may be considerable; (2) Pollinator flight sequences have a directional component; (3) Pollen-pistil compatibility may be lower among near-neighbors than among moderately spaced plants; (4) Seed abortion may be higher following self and near-neighbor crosses than following wide crosses; (5) Seed-set may be lower following crosses of neighboring plants than widely spaced plants; (6) Seed and seedling mortality may increase as the seed source is ap- proached; (7) Fitness of offspring from distant crosses may be superior to that from self or near- neighbor crosses; (8) Gene flow over some distance may be higher in natural populations than in crops. There is abundant reason to believe that dispersal data underestimate gene flow. As a con- sequence, we may conclude that neighborhood sizes and areas are larger, there is less potential for random differentiation, there is less isolation by distance, there is less potential for geographical differentiation, and that stronger selection is needed to foster and maintain local differences than judged from dispersal data alone. Nevertheless, even if gene flow distances were twice as large as we now think, the spatial scale of gene dispersal: (1) is still small enough to allow substantial differen- tiation over short distances with moderate selective differentials, and (2) is too small to be a major cohesive force within a species. In 1969, Ehrlich & Raven challenged the prevailing view that plant species were Mendelian populations integrated by gene flow. They argued that gene flow probably is much too restricted to homogenize anything but local gene pools. They pointed to selection as the primary cohesive and disruptive force in evo- lution. Selection would determine the influence of gene flow, often counterbal- ancing it in the course and maintenance of microgeographic differentiation. I This study was supported in part by National Science Foundation Grant DEB 78-23654. 1 am indebted to John Endler and Verne Grant for suggesting several ways to improve the paper. '' Department of Botany, University of Texas, Austin, Texas 78712. Ann. Missouri Bot. Card. 68: 233-253. 1981. 0026-6493/81/0233-0253/$02. 15/0 234 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol, 68 The general model of restricted gene flow has received substantial support from the crop and native plant literature on pollinator foraging behavior, and pollen and seed dispersal (Levin & Kerster, 1974; Levin, 1979). For the most part, pollinators fly from a plant to one of its near neighbors, suggesting that cross- pollinations are principally between neighboring plants. Animal- and wind-borne pollen dispersal tends to assume a leptokurtic distribution, with most pollen being deposited within a few meters of the source. In most species, pollen is rarely transported beyond 1000 m. Seeds also tend to be distributed narrowly about the source and are rarely carried beyond 200 m. Data from pollinator foraging behavior, and pollen and seed dispersal may be incorporated into population genetic models to obtain a first approximation of the breeding structure of populations (Levin & Kerster, 1974). In the process, a number of simplifying assumptions typically are made. These are as follows: (I) pollen picked up by a pollinator is deposited on the next plant visited; (2) polli- nator flights between plants is random in direction; (3) pollen-pistil compatibility, seed viability and seed-set are independent of pollen to egg parent distances; (4) seed and seedling survivorship are independent of the proximity to the seed source; (5) populations have no seed pool; (6) interpopulation gene flow is inde- pendent of population size and patterns of plant distribution. The major purpose of this paper is to test these assumptions. I will show that dispersal data underestimate the expanse of gene flow, but that the breeding structure of populations is still sufficiently restricted in space to permit selective differentiation over short distances. Pollinator Foraging Behavior and Pollen Flow Pollen flow distances would be distributed as pollinator flight distances were all pollen collected from one plant deposited on the next visited by a pollinator. Given the inefficiency of the pollen transport mechanism, it seems likely that there would be carry-over to a second, third or fourth plant. Thomson & Plo- wright (1980) were able to assess the functional form and extent of pollen carry- over in the bumblebee-pollinated Erythronium americanum by introducing a sin- gle red-pollen flower into a bees foraging sequence, and examining successively visited yellow-pollen flowers. Pollen usually was deposited on 7 or more flowers. In one sequence, pollen was deposited on several flowers through the 20th and then on numbers 23, 25, 29, and 54. In most sequences maximum pollen loads were deposited on the second flower or subsequent ones; and the deposition schedule was platykurtic. Thomson & Plowright also analyzed carry-over in bumblebee-pollinated CUntonia borealis and Diervilla lonicera using emasculated flowers. Pollen usually was carried beyond the fifth flower (often to the eighth). Tn many runs the initial fall-off of pollen loads roughly approaches a negative exponential function. Before considering the extent to which pollen carry-over increases the dis- tance pollen is transported, it is necessary to consider whether pollinators forage at random with regard to direction in successive flights. Zimmerman (1979) argued that in a population where the probability of revisiting any specific flower was small, pollinators should forage at random with regard to the direction of suc- cessive moves; i.e., angles of departures should have a uniform distribution. He 1981] LEVIN— DISPERSAL VERSUS GENE FLOW 235 Table L Pollen dispersal as influenced by pollen carry-over and flight directionality. Direction Constant" Carry-over Schedule D = D = 0.71 D 1 A. V = 50%, II = 20%, III = 15%, IV = 10%, V = 5% = 40%, II = 10%, III = 10%, IV = 10%, V = 10% 0%. VII = \0% 1.35 1.57 B. I 1.74 2.12 C. I = 20%, II = 10% 20%, IV = 10% 20%, VI 10%, VII 10% 1.84 2.45 D. I XI E. I 20%, III = 20% 20%, VII 0% 0% 10%, XIII 50%. V = 0% 20% 10%, XII = 10% IQ% 2.25 1.59 3.30 3.67 a random flights; D = 1 unidirectional flights; D = 0.71 observed directionality. ^ Roman numerals refer to plant number in a sequence. 2.00 3.10 3.70 5.80 6.00 Direction Constant x Mean Flight Distance = Mean Pollen Dispersal Distance; D = refers to observed that the bumblebee Bombus flavifrons forages randomly with respect to direction on Polemonium foliosissimum. Likewise, Gill & Wolf (1977) reported that the departure direction for sunbirds (Nectarinia sp.) feeding on Leonotis nepetifoUa was usually independent of the arrival direction. On the other hand, strong directional components to pollinator foraging have been reported in the case of Bombus sp. working Aconitum columbianum and Delphinium nelsonii (Pyke, 1978), Bombus on Armeria maritima and Limonium vulgare (Woodell, (Wadd IS bees and butterflies on Lvthru llifera on artificial flowers (Wadd 1971), and Apis of successive flights increases the distance pollen is carried, if there is pollen carry-over. The specific effects of pollen carry-over and directionality on pollen dispersal are best described with a series of simple models. Consider first a case of pollen carry-over with no directional component to foraging. The mean pollen dispersal Vi] Vlr th 1 X [S proportion pollen deposited plant in a random foraging bout is over increases pollen dispersal beyond pollinator flight moves is shown in Table 1. The percentage increase is dependent on the carry-over schedule, the more liberal the schedule the larger the effect. Note that regardless of the schedule, carry-over increases the mean pollen dispersal distance several percent. The effect of directionality of pollen dispersal distance is most easily dem- onstrated if we assume that pollinators fly in a straight line. With random polli- nation, the i^^ plant is (i^*^) x (mean flight distance) from the location of the pollen donor. With all flights in the same direction, the exponent of the i^^ plant is K so that the i^^ plant is (i') x (mean flight distance) from the pollen source. The mean pollen dispersal distance with unidirectionality is the (mean pollinator flight dis- (s :th plant x i'). Let us assume that 50% of the pollen from a plant is deposited on the first plant, 20% on the second, 15% on the third, 10% on the fourth, and 5% on the fifth plant visited. The mean dispersal distance with this carry-over schedule is twice that of the 236 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 mean flight distance. With random foraging, the dispersal mean is 35% greater than the flight mean. The effect of directionality when combined with other carry- over schedules is summarized in Table 1. Having considered the consequences of carry-over within the context of no directionality and complete directionality, let us turn to a more realistic scenario, one involving moderate directionality as observed in natural populations. With a positive correlation in arrival and departure directions, the mean distance of the i*^ plant from the pollen source is between i*^^ and P x (mean flight distance), the exponent increasing as the correlation increases (Levin et al., 1971). The exponents based on field observations vary from 0.69 and 0.72 for honeybees and butterflies, respectively, on Lythrum saVwaria (Levin et al., 1971) to 0.85 for bumblebees on Aconitum and Delphinium (Pyke, 1978). Using 0.71 as a repre- sentative exponent, I calculated the effect of moderate directionality coupled with various carry-over schedules on pollen dispersal. The results are presented in Table L The mean pollen dispersal distance is increased by a factor of 1.5 to more than 3. Thus pollen may move substantially farther than we would surmise from pollinator foraging data. Carry-over and directionality open the breeding structure of populations. Thus far we have seen evidence for and consequences of pollen carry-over and flight directionality. It is important to recognize that neither carry-over sched- ules nor the magnitude of directionality are fixed attributes of a plant-pollinator relationship. Rather, both are dependent on the amount of floral reward; and this has some interesting implications for pollen dispersal. As noted earlier, Thomson & Plowright (1980) documented carry-over schedules by analyzing pollen depo- sition levels in relation to the number of pollinator moves from the pollen source. In addition to employing unmanipulated flowers, they provided bees alternately with nectar-enriched and nectar-drained flowers of Diervilla lonicera. Taking run and sequence position into account, enriched flowers received significantly more pollen than drained ones. Pollen deposition per flower, or more precisely the fraction of the load deposited, is a simple function of the time spent on the two flower types; significantly more time is spent on enriched flowers. Heinrich (1979b) also studied the relationship between time and floral reward. He analyzed foraging behavior of Bomhus temcola workers in a field of TrifoUum repens in which sectors had been screened for two days to exclude foragers. On the av- erage, bees foraging in the rich clover patches visited 2.95 heads per minute compared to 18.9 heads per minute in the depleted clover patches. Bees in the rich clover probed into 11.6 florets/head, vs. 2.3 florets/head in the depleted clover. It is likely that much more pollen was deposited per head in the rich patches than in the depleted ones. Correlatively, there was probably much less carry-over in the rich patches. Bees also modify their directionality in response to resource quality. Pyke (1978) showed that the mean angular deviations of Bomhus arrivals and depar- tures on inflorescences of Aconitum columhianum and Delphinium nelsonii in- creased as the number of flowers visited per inflorescence increases. Thus as nectar reward per inflorescence increases, there is less correlation between suc- cessive flights, and pollinators will tend to stay in an area. Heinrich (1979b) monitored the angular deviation of Bomhus on successively visited clover heads I981J LEVIN— DISPERSAL VERSUS GENE FLOW 237 in rich and depleted patches. In the rich patches bees had high angular deviations, whereas bees tended to pass through patches where resources per flower were low. This aspect of foraging behavior is consistent with that described for pred- ators feeding on prey (cf. Pyke et al., 1977). It is possible that directionality may vary as a function of plant density. Pol- linators might move more directly and rapidly through a population if the energy gain : expenditure ratio were relatively low because of large interplant distances. If pollinators are found to move with greater directionality as density declines, this behavior would compound the effect of low density on flight distance and thus on pollen flow. The lower the density the greater is the mean flight distance (Levin & Kerster, 1974). If a sparse population also had small plants, then the level of directionality and flight distance would be high, and jointly result in unusually (for the species) broad pollen dispersal. Bee flight distance as well as directionality varies with resource qualty. In the case oi Bomhus flavifrons workers on Aconitum columbianum, the mean distance moved by bees from one inflorescence to another decreases toward an asymptote as the number of flowers on an inflorescence increases (Pyke, 1978). Pyke regards the numbers of flowers visited per inflorescence as an indicator of resource qual- ity, so that as quality declines flight distance increases. Bombus teiricola foraging on TrifoUum repens displays the same general behavior. The distances of inter- head moves in depleted patches was approximately twice that in rich patch, i.e., patches that were screened for two days (Heinrich, 1979a). On the average, bees foraging in the depleted patches rejected (without landing or probing) 27% of the flower heads they approached. When they entered rich patches, they rarely (0.4%) rejected heads they approached. The foraging movements (viz. directionahty and distance) described by Pyke (1978) and Heinrich (1979b) follow patterns predicted by mathematical models based on optimal foraging. Accordingly, we would expect other types of bees and indeed other categories of pollinators to forage in a similar manner. Given that pollinators respond to resource quality as the Bombus species, pollen flow will be considerably greater in depauperate populations than in rich ones. Since this condition will most likely vary through time, the same population may be both depauperate and rich, and correlatively have different pollen dispersal profiles over time. When nectar levels are very low due to temperature or moisture stress (Percival, 1965), pollen flow may be relatively broad for a considerable period. If pollinators respond to single flowers or inflorescences in predictable ways, they probably would also respond to the entire plant in the same ways, were the sum of the open flowers at some time the measure of quality. If site quality is poor, plants would produce few flowers per unit time, and have relatively few open flowers at any one time. A pollinator may "treat" this population as nutri- tionally depauperate because the energy gain per plant is small, and as a conse- quence move farther and in a more directional manner than if the plants had numerous flowers. Accordingly, pollen dispersal would be more widespread than usual in a marginal site. On the other hand site quality may be unusually good, and plants may be unusually florificent. Then pollen flow might be more restricted than usual. The discussion thus far has considered only pollination of single plant species. 238 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 However, pollinators are not completely flower constant in natural communities. There is usually some switching of species by lepidopterans, hummingbirds, bum- blebees, and to a lesser extent by honeybees (Baker & Hurd, 1968; Free, 1970; Proctor & Yeo, 1973). Individual bumblebees typically major on plant species and minor on one or a few others (Heinrich, 1976, 1979a). A minor species may differ manifestly from the major in its floral architecture and correlatively have its pollen transported in a position different from that of the major. Accordingly, pollen from a minor plant is apt to be transported to other minor plants even though several majors may intervene. Even if floral structures were similar, some pollen of the minor probably would be transported to another minor. By virtue of the fact that minor plants only intermittently are visited, cross-pollination among them often will be between widely spaced conspecifics, the distances being much larger than if the species were a major. Carry-over of minor pollen would augment the effect. Thus by virtue of being treated as a minor, the breeding structure of a species may be substantially enlarged. A given plant species may not be treated as a minor by all bees in a population. In that case, the pollen dispersal profile will be a composite of a typical leptokurtic distribution and a collection of relatively long distance events residing in the tail of that distribution. Potential Versus Actual Gene Flow Whereas the foraging behavior of pollinators suggests where pollen may be transported, the actual dispersal of pollen and hence pollen-mediated gene flow may be determined by the dispersion of genetic markers relative to a central point in a population. If we introduce a homozygous dominant genetic marker into a population, allow pollen to be dispersed, collect seed at various distances from the source, and determine the percentage hybrid plants as a function of distance, we would have a reasonably accurate measure of gene flow in that population. If we observed pollinator flight distances from the marker, we would then be able to compare flight and gene flow distances. Were all pollen deposited on the first plant visited, then pollinator flight and gene flow distributions would be concor- dant. Pollen carry-over would result in the divergence of the two distributions, the greater the carry-over the more discordant the distributions. In 1973 I began an investigation of gene flow in natural populations of Phlox drummondii, a spring-flowering annual of central Texas. I introduced a core of ca. 100 plants of a cultivar (Twinkle) into an extensive population near Gonzales, Texas. The cultivar was homozygous for a dominant corolla lobe character. Seeds were collected every two days in 1 m wide concentric rings around the core, and were progeny tested in the greenhouse. The principal pollinator of P. drum- mondii was Battus philinor. This and other lepidopterans were scored for inter- plant flight distance. Data for three episodes of pollination and gene flow are presented in Figure 1. Each pollination episode was scored over several days 3 to 4 weeks prior to each seed collection. Since the seeds mature in 3 to 4 weeks, the amount and distribution of hybrids will be a function of the foraging behavior we observed. The pollinator flight distribution was highly leptokurtic, with most flights being 2 m or less. A small percentage of the flights exceeded 10 m (Fig. I). The mean flight distance was ca. 2.75 m. In contrast, the gene flow pattern was less leptokurtic, and had more events beyond 10 m. Mean gene flow 1981] LEVIN— DISPERSAL VERSUS GENE FLOW 239 70 60 50 40 30 20 X:2 73 I 23456789 I0>I0 X ^4.33 2345 6789 IO>IO o 70 60 50 40 30 20 10 X = 2.57 (XL UJ cr < I 23456 789 IO>IO X = 4.04 23456 789 IO>IO 70 60 50 40 30 20 10 xo.oo I 2345 6789 10 > 10 DISTANCE (M) X- 4,66 234 56 78 910 DISTANCE (M) FiGURH I. Pollinator flight distance and gene flow distance at three periods of the growing season in a population of Phlox druminondii (Gonzales, Texas). distance was about 4.3 m, which was about 1.5 times the mean flight distance. This difference must be due to pollen carry-over. A disparity between pollinator flight and gene flow distributions also has been observed in Lupinus texensis by Schaal (1980). She placed 7 plants homozygous for an electrophoretic marker into a synthetic population of 91 plants lacking this marker. Bees were allowed to forage for 3 days and their flight distances were scored. Thereafter all plants were returned to the greenhouse, and seeds were collected subsequently from the field bee-exposed flowers. Mean pollinator flight distance was 0.97 m vs. 1.82 m for the gene flow distance. The level of kurtosis was significantly higher in the flight distribution. The differences in means and kurtosis are the result of pollen carry-over. Differential Crossability It is tempting to assume that parental distances are distributed as are the distances of pollen grains. However, pollen grains from afl plants do not have an equal probability of effecting fertilization in a given plant. The self-incompatibility (S) locus may preclude certain types of crosses. Pollen which shares a self-com- patibility allele with another plant is excluded from breeding with it (de Nettan- 240 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 court, 1977). Given the richness of S-alleles in natural populations, plants sharing S-alleles are likely to be related. This means that in populations where seed dispersal is highly restricted, and thus where neighboring plants are often related, a distant pollen grain is more likely to effect fertilization than a pollen grain from a neighboring plant. The success of pollen, or its fertilization ability, may be a function of the similarity of the pollen genotype to the stigma genotype. Pfahler (1965) analyzed the fertilization ability of maize pollen using pollen mixtures containing 50% pol- len from each of two parents. The two pollen types are equally successful if 50% of the progeny have a marker from one parent. When pollen from a source related to the female was mixed with pollen from unrelated sources in various 2-way combinations, between 33% and 43% of the fertilizations involved the related pollen. Thus related pollen was at a disadvantage, and negative assortative mating ensued. Since self-pollen often is competitively inferior to outcross pollen in mixtures (cf. Mulcahy, 1975), it is likely that the disadvantage of related outcross pollen vis-a-vis unrelated pollen is a common phenomenon. Given that pollen-pistil compatibility might vary with relatedness, might it also vary with distance in species with narrow seed and pollen dispersal? We sought the relationship between pollen-pistil compatibility and plant distance in Phlox drummondii. Pollen and seed dispersal in this species typically is a few meters or less, and populations display significant heterozygote deficiency relative to Hardy- Weinberg expectations (Levin, 1977). These observations suggest that populations undergo moderate inbreeding. Phloxes were collected as seedlings along 35 m linear transects in each of 14 populations. The seedlings were grown to maturity in the greenhouse. Crosses were made within and between populations. Crossability was considered in terms of the spatial relationships of the pollen and egg parents which were known. On the average, pollen grain germination increased with distance up to 10 m. Addi- tional distances up to 35 m were not accompanied by overall changes in germi- nation percentage. Pollen germination averaged about 17% with near-neighbor pollinations vs. ca. 20% with pollinations involving plants 10 m away or more. Pollen from parents in neighboring populations displayed germination rates sim- ilar to those found in intrapopulation crosses beyond 10 m. Pollen from sources greater than 20 km away performed substantially below that of pollen from within populations or from neighboring populations, mean germination being ca. 19% and 14%, respectively. Thus pollen germination rates are highest when mating plants are moderate distances apart. The crossability of two plants depends not only on pollen-pistil compatibility, but also upon the vigor of the developing seed. The developing seed is sensitive to genetic disharmonies brought about by inbreeding. High levels of seed abortion following inbreeding are well known in numerous predominantly outcrossing species (Crumpacker, 1967; Franklin, 1970), Of particular interest within the context of this discussion is whether abortion is dependent on the distance between mating plants. If relatedness is a function of distance, we would expect abortion levels between neighboring plants to be higher than that between plants several meters apart. In populations of Phlox drununondii, seed abortion decreased with interparent distances up to about 10 1981] LEVIN— DISPERSAL VERSUS GENE FLOW 241 m, and remained at that level as distances increased to 35 m. Abortion following crosses of 1 m or less averaged 18% vs. 14% following crosses of 10 m and more. It seems likely that higher abortion in the progeny of crosses between neighbors reflects their overall relatedness, because abortion levels increase substantially with self-fertilization. Crosses between neighboring populations did not yield low- er abortion levels than that observed at interparent distances beyond 10 m within populations. Crosses between populations more than 20 km apart did yield sig- nificantly higher abortion rates (19% vs. 14%). Distance-dependent abortion rates also have been described within Picea glauca (Coles & Fowler, 1976). Crosses between plants less than 100 m apart yielded 28% less sound seed than crosses at greater distances. Selfing produces little sound seed. Presumably some plants near each other were relatives, so that inbreeding varied as a function of distance. Seed-set in plants is dependent upon pollen-pistil compatibility and seed vi- ability. In Phlox drummondii, seed-set tends to increase as the distance between plants within a population increases. This pattern is evident across populations. Seed-set from crosses between near neighbors averages 53% compared to 65% for crosses between plants at least 10 m apart. Crosses between neighboring populations have seed-sets similar to that of crosses between plants 20 m apart (x = 63%). However, crosses between populations more than 20 km apart yielded lower seed-set (x = 55%) than crosses between neighboring populations. Thus the general pattern is one of an intermediate optimum, more than a few meters and less than 20 km. Seed-set following crosses between parents separated by various distances has also been analyzed in Delphinium nelsonii (Price & Waser, 1979). For the most part, 10 m crosses gave higher seed-set in the two study populations over two years than crosses over I m, 3 m, 30 m, 100 m and 1000 m. The pattern of an intermediate optimum also has been described in Stylidium. In S, elongatum and S. confluens. the percentage seed-set was greater in crosses separated by 40 to 60 km than in crosses of smaller or greater distances (Banyard & James, 1979). Since embryo abortion often follows self-fertilization in these species, it seems likely that reduced seed-sets in intrapopulation crosses were a manifestation of inbreeding depression. A decline in seed-set between ''distant'' populations in Delphinium and Stylidium may be due to genome incompatibility fostered by divergent local adaptations. As we study fine scale crossing relationships in natural populations, it is ev- ident that the facility with which plants interbreed is dependent on their spatial relationships which are correlated with their genetic relationships. On the average, neighboring plants in outcrossing species seem less likely to successfully inter- breed than plants tens to hundreds of meters apart. As a consequence, the breed- ing structure of populations may be more open than we would imagine. Differential Viability The breeding structure of populations needs to be considered in terms of the resulting standing crop as well as the zygote or seed population. As the breeding structure of the seed population is shaped by genotype-dependent mortality dur- ing development, so the breeding structure of the adult population is shaped by 242 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 genotype-dependent mortality throughout the vegetative phase of the life cycle. In essence, there is a potential and realized breeding structure. The former de- scribes the mating pattern, the latter integrates the mating pattern and mortality. Products of outcrossing typically have higher viability than products of selfing during seed development and as established plants (Allard et al., 1968; Stern & Roche, 1974; Janossy & Lupton, 1976). In conifers the incidence of seed abortion (Koski, 1971, 1973; Birshir & Pepper, 1978), and defective or slow growing seed- lings (Franklin, 1970; Koski, 1973) is 2 to 5 times higher with selfing than with outcrossing. The genetic load per zygote averages more than 8 lethal equivalents in some species of Pinus and Pseudotsuga (Franklin, 1972; Sorensen, 1969; Ko- ski, 1973), Although zygote populations of many species are often somewhat in- bred, the level may diminish through time as a result of differential mortality. For example, in Eucalyptus pauciflora the average outcrossing rate for seeds is 63% compared to 76% for seedlings (Phillips & Brown, 1977). Different outcrossing rates are seen even among seeds stored for different periods. In Eucalyptus delegatus, outcrossing rates in old seeds was 85% versus 66% in the most recent collection (Moran & Brown, 1980). Differences in seedling quality as a function of interplant distances have been found when sought. One component of seedling quality, epicotyl length was stud- ied in the progeny of crosses of various distances in Picea glauca. Epicotyls of progeny of near-neighbor crosses were shorter than those from distantly spaced parents (Coles & Fowler, 1976). These data suggest that neighboring plants were related, especially since the epicotyls of self-pollinated seedlings were 24% short- er than those from long-distance outcrosses. Differences in epicotyl length as a function of distance have obvious fitness implications in that the larger seedlings within a cohort usually have higher survivorship than the smaller ones (Harper, 1977). The effects of interparent distance on seedling survivorship have been de- scribed in Delphinium nelsonii (Price & Waser, 1979). Seedlings from crosses between plants 10 m apart had higher survivorship in the native sites than seed- lings from crosses between plants 1 m, 100 m, and 1000 m apart, respectively. h Self-seedlings had the lowest survivorship. Price & Waser conclude that seed- lings from crosses between neighboring plants may be suffering inbreeding depression, whereas the seedlings of wide crosses may be suffering genomic incompatibility. Survivorship within a cohort is most likely to be genotype-dependent with mixed selfing and outcrossing. In species of inbreeding annuals, it is common to find an excess of heterozygotes relative to expectations based on the mating system (Allard et al., 1968; Clegg et al., 1978). Comparisons of seedlings and adult genotypes from natural populations of Avena harbata showed that hetero- zygotes at one of 3 esterase loci had about a 30% advantage over the homozygotes (Clegg & Allard, 1973). In Avcna sativa, heterozygotes for two loci governing crown rust reaction had a 50% advantage over homozygotes (Fatunla & Frey, 1980). On the other hand, no substantial viability advantage was found in single or multilocus heterozygotes in experimental barley populations (Clegg et al., 1978). Genotype frequencies in parents and progeny have been analyzed in Eu- calyptus pauciflora (Phillips & Brown, 1977) and £. delegatcnsis (Moran & 1981] LEVIN— DISPERSAL VERSUS GENE FLOW 243 Brown, 1980), Heterozygotes appear to have about a 20% advantage in the for- mer, and a 50% advantage in the latter. In perennials changes in heterozygosity through time may be inferred from genotype frequencies in different age classes all sampled at one time. Using this approach, Schaal & Levin (1976) found a substantial increase in heterozygosity through time in Liatris cylindracea. Dif- ferential survivorship in favor of heterozygotes effectively opens the realized breeding structure of populations because survivors are weighted in favor of the progeny of unrelated plants. Seed Dispersal Distributions of seed dispersal distances have been described for numerous species with various adaptations for dispersal (cf. Levin & Kerster, 1974; Levin, 1979). For the most part, seeds remain in the vicinity of the seed source. The long-distance component of seed dispersal distributions is poorly understood and difficult to document. It is likely that our impressions of seed dispersal are too conservative owing to our ignorance of the dispersal curve tail (Grant, 1980). Seed dispersal typically is measured in terms of absolute distance. However, a mean dispersal distance of 25 m in a population of mesophytic climax tree species does not have the same impHcations as it does for a small prairie herb. For the tree, 25 m would be only 2 or 3 plant diameters (canopies) away, whereas for the herb 25 m would be hundreds of plants from the source. Accordingly, parents and offspring would be near-neighbors in the case of the tree, but widely spaced in the case of the herb. It also follows that half-sibs borne on a single plants will be near-neighbors in the case of the tree, but well dispersed in the case of the herb. Near-neighbor pollination in the former would then result in inbreed- ing, whereas in the latter it would not. Also, it follows that the gene pool of the tree species would be little homogenized by 25 m dispersal whereas that of the herb would be well homogenized. We may conclude that in this example seed dispersal or seed-mediated gene flow effectively is much greater in the herb than the tree. The numbers of juxtaposed plant canopies over which seeds are dispersed is easily estimated. Phlox pilosa seeds are scattered by capsule dehiscence an av- erage of 1.2 m with a maximum of 3.6 m (Levin & Kerster, 1968). The diameter of a mature plant in a typical site is about 3 dm. Accordingly, seeds are dispersed an average of 4 plant diameters and a maximum of 10 diameters. Seeds of Liatris aspera are dispersed an average 2.5 m with a maximum of 9 m (Levin & Kerster, 1969). The diameter of a robust adult is about 20 dm. Accordingly, seeds are dispersed an average of 12 plant diameters and as far as 45 diameters. In Euca- lyptus re g nans, wind-borne seeds travel a mean distance of about 40 m and as far as 125 m (Gilbert, 1958). With plant diameters of 8 m, seeds are being dispersed an average of 5 plant diameters and as far as 15 diameters. These examples are representative of late successional plants and point to the effectively restricted distribution of seeds therein. Mean seed dispersal distances for weedy plants with a well-developed plumose pappus are not available. However, from Sheldon & Burrows's (1973) calculations of dispersability based upon propagule's terminal velocity and resistance coefficients, we may infer that in strong winds seeds of 244 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 such weeds as Cirsium cirvense and Sonchus arvcnsis are transported an average of 20 plant diameters. Plumed propagules of Salix repens and Popidus tremula also may be carried an average of 15 diameters or more. The so-called weedier plants thus are ''placing" their offspring farther away than are the late succes- sional species. The distribution of seed relative to a source may be altered by distance- and density-responsive seed predators as proposed by Janzen (1970). He argued that *'no matter how large the seed crop in a given year, or how far the seed from the parent, density-responsive predators will pursue seeds and seedlings until their density is so low that search is no longer profitable.'' He also postulated that "if seeds are placed or planted at various distances from a parent tree at low density . . . , their survival to well developed sapling stages should increase with distance from the parent." Janzen's postulates have received support from seed predation patterns in Sterculia apatula (Janzen, 1972), Schcclea rostrata (Wilson & Janzen, 1972), Juglans nigra (Stapanian & Smith, 1978), Andira inermis (Jan- zen et al., 1976), and Datura discolor (O'Dowd & Hay, 1980), and seedling mortality patterns in Casearia nitida (Howe & Primack, 1975) and Casearia corymbosa (Howe, 1977). Density-dependent and proximity-dependent predation introduce a differential between potential seed dispersal and actual seed dispersal, the mean and variance of the latter being greater. Since predation tends to in- crease distance between parents and offspring, and between half-sibs, the result is an opening of the breeding structure of populations. Interpopulation Gene Flow The extent of interpopulation gene exchange in nature is unknown. On the other hand, plant breeders have determined distances sufficient to preclude in- terpopulation contamination of varieties. In many outcrossing crops, more than 5% of the seeds will be hybrid if the distance between populations is about 10 m; and more than \% of the seeds will be hybrid if the distance is about 500 m (Kernick, 1961; International Crop Improvement Association, 1963). We might be tempted to assume that gene flow in crops is representative of that in natural populations. However, there are several considerations which suggest that gene flow between natural populations may be higher than that in crops over the same distance. First, there is the matter of population size. Natural populations typi- cally are smaller and more patchy in their structure than crop plots. The level of interpopulation hybridization is a function of population size or varietal mass, so that hybridization between populations will increase as population size declines (Fryxell, 1956; Bateman, 1947; Crane & Mather, 1943; Williams & Evans, 1935). Another consideration is the spatial relationships of populations. Plant breeders employ a donor population and one to several recipient populations. All pollen can be traced to that donor population. In nature, a population may receive pollen from several neighboring populations. Even if the levels were low from each source, they are additive, and the sum might have a considerable impact on the genetic structure of the recipient population. Finally, there is the matter of the pollinator. Most animal-pollinated crop plants are serviced primarily by hon- eybees, whose foraging area is considerably less than other bees and lepidopter- 1981] LEVIN— DISPERSAL VERSUS GENE FLOW 245 ans (Heinrich, 1975; Free, 1970). Accordingly, hybridization is lower than might be the case were the same plants subjected to natural pollinators. The extent to which gene flow fosters the convergence of populations or restricts their divergence depends not only on the frequency of hybridization but also upon the adaptedness of hybrids. The more fit the hybrids the greater the importance of each hybridization event. Are interpopulation hybrids inferior or superior to indigenous plants? In predominantly outbreeding crop species, the vigor and fecundity of hybrids typically increase as the genetic distance between parental strains increases until some critical level of divergence is reached when interactions at a few loci counterbalance the effect of heterozygosity. In maize, heterosis is a positive function of the level of divergence of strains except for the most divergent ones (Moll et al., 1965). In Nicotiana (Matzinger & Wernsman, 1967) and Gossypium (Mariani & AvieH, 1973), heterosis increases with diver- gence to the level of related species. In natural population systems, maximum heterosis also is associated with moderate levels of divergence. Heterosis has been documented in interracial crosses of Norway spruce (Nilsson, 1974), Douglas fir (Orr-Ewing, 1969), and loblolly pine (Woessner, 1972; Owino & Zobel, 1977). Greater heterosis and higher fertility were obtained in the progeny of crosses between Liriodcndron tidipifcra of different populations (often only a few miles apart) than in crosses within the same population (Carpenter & Guard, 1950). Heterosis in Mimulus is best developed in hybrids between populations which have undergone moderate degrees of divergence regardless of their geographical relationships (Vickery, 1978). Beltran & James (1974) demonstated heterosis in hybrids between chro- mosomally homozygous populations of Isotoma petraea, the more inbred the populations the greater the vigor of their hybrids. No correlation was found be- tween heterosis and the level of population divergence. The Isotoma study is of particular interest because it shows that the genetic structure of populations, as shaped by population size and breeding structure, affects the relative quality of distant pollen parents. Our knowledge of interpopulation hybrids is based principally upon crosses between populations tens to hundreds of miles apart, and upon performance trials in greenhouses, gardens, or plantations. Of principal concern here are hybrids between neighboring populations or those within pollination range, and how those hybrids would fare in natural populations. Neighboring populations may be well differentiated as a result of selective differentials or random drift. If populations are different by virtue of differential selection and hybrids are intermediate to their parents, then interpopulation hybrids may be ill-fit in either population; it is unlikely they will be better adapted than local residents. On the other hand, if populations have small effective sizes, have diverged as a result of stochastic processes, and are genetically depauperate, then interpopulation hybrids may be superior (heterotic) to local residents in both the egg and pollen parent popula- tions. Genetic Differentiation In practice we estimate gene flow in plants from the distributions of pollen We 246 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 in one form or another to estimate gene flow. Whereas dispersal of pollen and seeds has some general predictive value with regard to gene flow, dispersal per se does not guarantee gene flow. This point recently was elaborated upon by Endler (1977, 1979) with respect to animals. He concludes that dispersal over- estimates gene flow because of the complexities of animal behavior and their life history attributes. In the case of plants I have reached the opposite conclusion. Estimates of gene flow based upon dispersal of pollinators, pollen grains, and seeds tend to underestimate actual gene flow distances because they do not take into consideration pollen carry-over and flight directionality, proximity-dependent cross-compatibility and seed-set, and proximity- and density-dependent seed pre- dation and seedling survivorship, seed pools, genotype-dependent survivorship, and the fitness of interpopulation hybrids. The implications of wider gene dis- persal are best understood within the framework of breeding structure and selec- tion models which provide numerical and spatial scales upon which patterns of gene flow may be interpreted. The breeding structure of populations may be considered in terms of Wright's (1940, 1943, 1946, 1951) neighborhood model. The neighborhood is the area from which any two parents could have come with equal probability. The effective size of a neighborhood is equivalent to the number of reproducing individuals in a circle whose radius is equivalent to twice the standard deviation of the gene dispersal distance (Wright, 1946). Strictly speaking this circle includes 87% of the parents of individuals at its center. The neighborhood size in a plant with pollen and seed dispersal may be described as Ne = 6.3dr(SpiV2Np + Xsj^NJ where p refers to pollen dispersal distance, Np number of pollen dispersal observations, s the seed dispersal distance, Ns the number of seed dispersal observations, and r the proportion of outcross progeny (Levin & Kerster, 1969). The area of a neighborhood is N^./d, where d is the genetically effective density. Genetically effective density is approximately the density of flowering plants. The neighbor- hood describes the scale of gene flow and thus sets the lower limit on the area that can respond to spatially defined disruptive selection (Slatkin, 1973; May et al., 1975; Endler, 1977). Estimates of neighborhood size and area for several representative herbs and trees are enumerated in Tables 2 and 3. Neighborhood sizes vary from 5 in the facultatively cleistogamic Lithospennum caroUniense to 547 in Viola pedata. Neighborhood areas also vary widely among species, ranging from less than 20 m^ in Viola hlanda and Phlox pilosa to over 30,000 m^ in Pinus elliottii. There is also considerable variation within species. The neighborhood sizes of herbs and trees fall within the same range, but the neighborhood areas of trees are consid- erably larger. The size of the neighborhood determines the level of inbreeding and penchant for genetic drift between sectors of population. For most species in Table 2, the size of the neighborhood is sufficient (>50) to preclude substantial differentiation via genetic drift within a continuous population (Wright, 1943). These values are calculated from dispersal data alone and thus are apt to be underestimates of actual neighborhood sizes. The area of neighborhoods in plants, especially herbs, is sufficiently small to permit marked differentiation in response to selection over short to moderate distances. If mean gene flow distances were greater than that of dispersal, the 1981] LEVIN— DISPERSAL VERSUS GENE FLOW 247 Table 2. Neighborhood size and gene flow-selection balance in herbs. Species Phlox pilosa Liatris aspera Liatris cvlindracea Viola peJata Viola pensylvanica Lithospermum caroli- niense N, 310 5 / a 205-547 2.5 m 3.6 m Cline width<- A A 'c 0.1 m 1000 m Coeffi- cient of Selec- tion'i Reference 75-282 1.6 m 5.1 m 30-191 1.5 m 4.7 m 165 12.5 m 11.5 m 2.3 m 7.3 m 17.9 m 71.1 m 67.4 m 90.8 m 7.9 m 8.2 m 0.03 Levin & Kerster, 1%8 0.02 Levin & Kerster, 1%9 0.05 Schaal & Levin, 1978 19.4 m 95.2 m 0.06 Beattie & Culver. 1979 20.1m 97.6 m 0.07 Beattie & Culver, 1979 1.4 m 4.5 m 11.0m 65.4 m 0.03 Kerster & Levin, 1968 square root of mean squared dispersal distance. characteristic length scale of variation of gene frequencies; s = 0.10 0.10; A refers to width of environment transition. For gene substitution over 30 m. neighborhood area would increase. Then there would be less isolation by distance within and between populations, and less potential for selective differentiation over short distances, than would be possible with the areas enumerated in Tables 2 and 3. The relationship between gene flow and selection in shaping local variation patterns in a heterogeneous environment and the spatial scale of such variation is best considered in terms of a one-dimensional gene flow scale /, the square root of the mean squared dispersal distance, rather than neighborhood area (Fish- er, 1950; Slatkin, 1973; Endler, 1977). The gene flow distance, /, is related to the neighborhood radius by r /V2 by distance between two subpopulations, and the more rapidly and fully these aggregates may respond to local selective differentials. Knowing /, we may cal- culate the minimum distance over which a population may respond to selection, assuming an abrupt selective change between adjacent environments as in End- ler' s (1977) ecotone model. This is referred to as the characteristic scale length Table 3. Neighborhood size and gene flow-selection balance in trees. Species N,. l^ \ I ^ Pinus cembroides 11 17.0 m Finns eUiottii 365 67.8 m Cedrus athmtica 207 74.3 m Pseudotsuf^'ii taxlfoiia 26 18.8 m Cline width^ A = 0.1 m A = 1000 m Coeffi- cient of Selec- tion" 53.8 m 131.8 m 342.0 m 214.4 m 525.3 m 959.7 m 234.9 m 575.5 m 911.4 m 59.4 m 145.5 m 365.4 m 0.02 0.25 0.31 0.02 a / = square root of mean squared dispersal distance. /, characteristic length scale of variation of gene frequencies; .v 0.10 ^ s = 0.10: A refers to width of environment transition. ^ For gene substitution o\qy 300 m. Reference J. Wright, 1953 Wanget al., 1960 J. Wright, 1953 J. Wright, 1953 248 ANNALS OF THE MISSOURI BOTANICAL GARDEN IVoL. 68 of variation of gene frequencies (/,) and is defined as I^ = Us, where s is the maximum difference in fitness between homozygotes in two environments (Slat- kin, 1973). Regardless of the spatial heterogeneity in selection pressures, gene frequencies would not vary significantly over a distance less than this length. If the environment changes on a scale less than this length, gene frequencies will respond to selection intensities averaged over the characteristic length. When the scale of variation in the environment is greater than this length, populations can respond to a heterogeneous environment and differentiate into distinctive units with clines between them (May et al., 1975; Endler, 1977). The characteristic scale lengths, assuming s ^ 0. 10, for the plants under con- sideration vary from 4.7 m for Liatris aspcra to 234 m for Cedrus atlantica (Tables 2 and 3). If the selective differential between the environments is less, the distances would be greater and vice versa. As we might expect, populations of herbs are better able to respond to local environmental heterogeneity than trees. Were the intensity of selection reduced by an order of magnitude, the characteristic scale length would increase roughly 3 times. If the mean gene flow distance were twice the dispersal distance (/ in Tables 2 and 3), then the char- acteristic scale lengths shown in Tables 2 and 3 would have to be doubled. Consider next the width of clines between two habitat types which are jux- taposed, so that the transition occurs over a very short distance, less than the characteristic scale length. Suppose that genotypes AA, Aa, and aa have fitness of 1 + ^, I and 1 - s, respectively, in one habitat, and 1 - s, 1, and 1 + s, respectively, in the other. The width of a cline from p = 1 to p = is 2.45 l^ (May et al., 1975). Using 0. 10 as the value of s (as above), we find that the width of clines varies from 1 1 to 20 m for herbs (Table 2) and from 132 to 575 m for trees (Table 3). It is evident that marked differentiation, in this case gene sub- stitution, can occur over relatively short distances in response to moderate se- lective differentials. This is the consequence of restricted gene flow. If the actual gene flow distances were twice those described (/ in Tables 2 and 3), then the width of the cline would double. The transition between environments may be gradual instead of abrupt. Sup- pose that one homozygote is best fit for one environment and the other homo- zygote the other, that the fitnesses of homozygotes varied over a 1000 m gradient, and that the fitness relationships are as given in the previous example. The width of the cline is 2.40(/(."^A)"^ where A is the transition distance over which both homozygote fitnesses are changing relative to one another (May et al., 1975). The width for the herbs is between 71 and 98 m (Table 2), and for the trees is between 342 and 960 m (Table 3). This illustration shows that substantial differentiation may occur over moderate distances along environmental gradients. Were the actual gene flow distances twice those inferred from dispersal distances, the clines would be twice as wide. Finally let us ask the intensity of selection that would be necessary to bring about gene substitution (p = 1 .0 to p = 0) over a span of 30 m for herbs and 300 m for trees. The value of s may be estimated from the following relationship: maximum slope of gene frequencies ~Vs/3/, where the slope is the difference in gene frequencies/distance and / is the root mean square of the migration distance as above (Slatkin & Maruyama, 1975). The selection coefficients necessary to 1981] LEVIN— DISPERSAL VERSUS GENE FLOW 249 bring about gene substitution over 30 m in herbs are surprisingly small, less than .07 (Table 2). If the actual gene flow distances were greater than dispersal dis- tances, the intensity of selection necessary for gene substitution would be greater. The selection coefficient necessary to bring about gene substitution over 300 m in trees is weak in some species and strong in others (Table 3). If the actual gene flow distances were greater than dispersal suggests, substitution would not be possible in Cedrus atlantica and Pinus elliottii. The spread of an advantageous gene across populations also is dependent upon gene flow. It is intuitively obvious that the greater the distances over which genes move per generation the more rapid the spread of the advantageous gene. Fisher (1937) found that the advance of advantageous genes or velocity of the wave of gene increase along a linear habitat can be described as v = cr(2m), where a is the standard deviation for gene flow distance and ni is the selective advantage. The standard deviation for the plant species considered earlier may be obtained from Tables 2 and 3, recognizing that the gene flow distance I ^ 2xt. The spread of an advantageous gene in Phlox pilosciy assuming that the selective advantage is s = 0.10, is 0.16 m per generation. In Cedrus atlantica, the rate is 7.4 m per generation. Even if the selective advantage is large the spread is slow. If 5 = 0.5, the rate of spread is still only 0.8 m per generation in Phlox and 37 m in Cedrus. The progress of the advantageous gene would be at least four times greater in a two-dimensional population with the same gene flow characteristics (Wright, 1969). In the case of a linear array of discontinuous populations, the advantageous genes would progress in a wavelike fashion, with the velocity being a function of interpopulation gene flow and time to fixation within populations (Slatkin, 1976). The velocity of the wave of an advantageous gene may be estimated from the tabulations of Slatkin (1976). Consider a rather realistic set of circumstances in which populations are 1 km apart, where the population size (N) equals 500, adjacent populations exchange 1 individual per generation (m = 1), and the ad- vantage of a gene is 10%. Under these circumstances, the rate of spread is about 20 m per generation. If gene flow were reduced by an order of magnitude (m = 0. 1), the velocity would be about 1 1 m per generation. Although gene exchange is most likely to occur between adjacent populations, there may be some between distant populations. These long-distance events would substantially increase the rate of spread of an advantageous gene. Return- ing to a linear array of populations and the tabulations of Slatkin (1976), and assuming that on the average 0.01 individual moves between populations 10 steps apart with populations of 500 and a selective advantage of 10%, the rate of spread would be about 30 m per generation. This compares with 20 m per generation with gene flow between adjacent populations. Models of gene migration in a one-dimensional array pertain to population systems distributed along rivers, shorelines, and other such linear habitats. Most plant population systems are distributed in two dimensions. The rate of spread of an advantageous gene in a two-dimensional array has not been described. However, it will be greater than in a one-dimensional array because there are more pathways over which a gene could move between populations (Slatkin, 1976). The rate of spread is determined primarily by the lag time associated with 250 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 the movement of genes between populations, i.e., the time between gene entry and its attainment of a frequency such that it might be transported by a migrant (seed or pollen). Let us assume that in a two-dimensional array the level of gene exchange between populations increases by a factor of two, so that there are 2 migrants per generation between populations of 500 {s = 0. 10). The rate of spread of an advantageous gene will be 23 m per generation compared to 20 m when there was only 1 migrant per generation. If the number of migrants is doubled in the aforementioned model of gene exchange between populations 10 steps re- moved (from 0.01 to 0.02 individual per generation), the rate of spread increases from 30 m per generation to 52 m. If the number of migrants increases by a factor of 5 due to 2-dimensional gene exchange, the rate of spread increases to 86 m per generation. The progress of an advantageous gene through a population system remains moderate even with very liberal gene-flow estimates. Depending upon whether a plant was predominantly selfing or outcrossing and on the means of seed dispersal, the wave of advance of a beneficial gene at high frequency probably would range from less than 5 m to more than 50 m per generation. At these rates a beneficial gene would not have time to dominate much beyond the region in which it originated. Using 25 m per generation for illustrative purposes, it would take 4000 years for a high frequency wave to move 100 km in an annual plant. Even if the rate were 100 m per generation, it would take 1000 years for a wave to cover this distance or over an area about 10,000 km'^ This constitutes only a small portion of the range of most species. The rate of spread will be even slower in perennials, because the rate is measured in terms of distance per generation. The rate of spread in a short-lived perennial herb (generation time of 10 years) would be one-tenth that described, so that the range of distances per year would range from less than 0.5 to around 5 m. For a long- lived perennial the rate would be reduced by an additional 50% or more. Clearly, with the biological processes we are envisioning, novel adaptive mutations origi- nating in one population cannot rise to prevalence throughout a significant por- tion of a species range. in summary, 1 have argued that estimates of gene flow within and among populations which are based upon the movement of pollinators, pollen, and seeds tend to be too small. On the other hand, the scale of dispersal is so small that even if gene flow distances were twice those of dispersal, differentiation within populations under moderate disruptive selection pressure, and interpopulation divergence with very weak selective differentials would still be possible. Indeed populations more than a few kilometers apart may be completely isolated by distance and thus free to diverge along avenues independent of those taken by other populations. Gene flow appears so restricted that novel adaptive mutations would be confined to relatively small portions of species ranges. The view of Ehrlich & Raven (1969) that gene flow is not a prime integrating force within species still seems valid. Indeed, if cohesion does occur as a con- sequence of gene exchange, ostensibly it is within continuous subpopulations, discontinuous subpopulations, or tightly knit clusters of neighboring populations. These narrowly circumscribed ensembles thus are the effective units of evolution. Literature Cited Allard, R. W., S. K. Jain & P. L. Workman. 1968. The genetics of inbreeding species. Advances Genet. 14: 55-\?^\. 1981 J LEVIN— DISPERSAL VERSUS GENE FLOW 251 Baker, H. G. & P. D. Hurd. 1968. Intrafloral ecology. Annual Rev. Entomol. 13: 385^14. Banyard, B. J. & S. H. James. 1979. Biosystematic studies in the Stylidium crassifolium species complex (Stylidae). Austral. J. Bot. 27: ll-YI . Bateman, A. J. 1947. Contamination of seed crops. L Insect pollination. J. Genet. 48: 257-275. Beattie, a. J. & D. C. Culver. 1979. Neighborhood size in Viola, Evolution 33: 1225-1229. Beltran, L C. & S. H. James. 1974. Complex hybridity in Isotoma petraea. IV. Heterosis in interpopulational hybrids. Austral. J. Bot. 22: 251-264. BiRSHiR, J. & W. D. Pepper. 1978. Estimation of number of embryonic lethal alleles in conifers. I. Self-pollinated seed. Silvae Genet. 27; 50-54. Carpenter, I. W. & A. T. Guard. 1950. Some effects of cross-pollination on seed production and hybrid vigor of tuliptrees. J. Forest. 48: 852-855. Clegg, M. T. & R. W. Allard. 1973. Viability and fecundity selection in the slender wild oat. Arena barbata L. Science 181: 667-668. , A. L. Kahler & R. W, Allard. 1978. Estimation of life cycle components of selection in an experimental plant population. Genetics 89: 765-792. Coles, J. F. & D, P. Fowler. 1976. Inbreeding in neighboring trees in two white spruce populations. Silvae Genet. 25: 29-34. Crane, M, B. & K. Mather. 1943. The natural cross-pollination of crop plants with particular reference to the radish. J. Appl. Biol. 30: 301-308. Crumpacker, D. W. 1967. Genetic loads in maize (Zea mays L.) and other cross-fertilized plants and animals. Evol. Biol. 1: 306-424. Ehrlich, p. R. & P. H. Raven. 1969. Differentiation of populations. Science 165: 1228-1232. Endler, J. A. 1977. Geographic Variation, Speciation, and Clines. Princeton Univ. Press, Princeton, New Jersey. . 1979. Gene flow and life history patterns. Genetics 93: 263-284. Fatunla, T. & K. J. Frey. 1980. Analysis of genetic changes in radiated and non-radiated bulk oat {Arena satira L.) populations. Theor. Appl. Genet. 56: 199-202. Fischer, R. A. 1937. The wave of advance of advantageous genes. Ann. Eugen. 7: 355-369. . 1950. Gene frequencies in a cline determined by selection and diffusion. Biometrics 6: 353- 361. Franklin, E. C. 1970. Survey of mutant forms and inbreeding depression in species of the family Pinaceae. Southeast Forest Expt. Sta. USDA For. Serv. Res. Pap. SE-61:I-21. . 1972. Genetic load in loblolly pine. Amer. Naturalist 106: 262-265. Free. J. B. 1970. Insect Pollination of Crops. Academic Press, New York. Fryxell, p. a. 1956. Effect of varietal mass on the percentage of outcrossing in Gossypium hir- sutum. J. Hered. 57: 299-301. Gilbert. J. M. 1958. Forest succession in the Florentine Valley, Tasmania. Proc. Royal Soc. Tas- mania 93: 129-151. Gill, F. B. & L. L. Wolf. 1977. Nonrandom foraging by sunbirds in patchy environment. Ecology 58: 1284-1296. Grant, V. 1980. Gene flow and the homogeneity of species populations. Biol. Zbl. 99: 157-169. Harper, J. L. 1977. Population Biology of Plants. Academic Press, New York. Harrington, J. F. 1972. Seed storage and longevity. Pp. 145-245, in T. T. Kozlowski (editor). Seed Biology. Vol. 3. Academic Press, New York. Heinrich, B. 1975. Energetics of pollination. Ann. Rev. Ecol. Syst. 6: 139-170. . 1976. The foraging specializations of individual bumblebees. Ecol. Monogr. 46: 105-128. . 1979a. ''Majoring" and "minoring" by foraging bumblebees, Bombus ragans: An experi- mental analysis. Ecology 60: 245-255. . 1979b. Resource heterogeneity and patterns of movement in foraging bumblebees. Oecologia 40: 235-245. Howe, H. F. 1977. Bird activity and seed dispersal of a tropical tree. Ecology 58: 539-550. & R. Primack. 1975. Differential seed dispersal by birds of the tree Casearia nitida (Fla- courtiaceae). Biotropica 7: 278-283. International Crop Improvement Association. 1963. Minimum Seed Certification Standards. Publ. 20. Janossy, a. & F. G. H. Lupton (editors). 1976. Heterosis in Plant Breeding. Elsevier, New York. Janzen, D. H. 1970. Herbivores and the number of tree species in tropical American forests. Amer. Naturalist 104: 501-528. . 1972. Escape in space by Sterculia apetala seeds from the bug Dysdercus fasciatus in a Costa Rican deciduous forest. Ecology 53: 350-361. , G. A. Miller. J. Hackforth-James, C. M. Pond, K. Hooper & D. P. Janos. 1976. Two Costa Rican bat-generated seed shadows of Andira inermis (Leguminosae). Ecology 57: 1068- 1075. Kernick, M. D. 1961. Seed production of specific crops. Pp. 181-547, in Agricultural and Horti- cultural Seeds. FAO Agricultural Studies No. 55. 252 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 KtRSTHR, H. W. & D. A. Levin. 1968. Neighborhood size in Lithospenuum caroliniense. Genetics 60: 577-587. KosKi, V. 1971. Embryonic lethals of Picca abics and Finns sylvestris, Comm. Inst. For. Fenn. 75: 1-30. . 1973. On self-pollination, genetic load, and subsequent inbreeding in some conifers. Comm. Inst. For. Fenn, 78: 1-42. Lkvin, D. A. 1977. The organization of genetic diversity in Phlox dnunmondii. Evolution 31: 477- 494. . 1979. The nature of plant species. Science 204: 381-384. ■ & H. W. Kerster. 1968. Local gene dispersal in Phlox pilosa. Evolution 22: 130-139. & . 1969. Density-dependent gene dispersal in Liatris. Amer. Naturalist 103: 61-74. & . 1974. Gene flow in seed plants. Evol. Biol. 7: 139-220. , & M. NiEDZLEK. 1971. Pollinator flight directionality and its effect on pollen flow. Evolution 25: 113-118. Mariani, A. & E. AviELL 1973. Heterosis during the early phases of growth in intraspecific and interspecific crosses of cotton. Crop Sci. 13: 15-18. Matzinger, D. F. & E. A. Wernsman. 1967. Genetic diversity and heterosis in Nicotiana. I. Interspecific cross. Der Ziichter 37: 186-191. May, R. M., J. A. Endler & R. E. McMurtrie. 1975. Gene frequency clines in the presence of selection opposed by genetic drift. Amer. Naturalist 109: 659-676. Moi L, R. H., J. H. LoNNQUiST, V. FoRTUNo & E. C. Johnson. 1965. The relationship of heterosis and genetic divergence in maize. Genetics 52: 139-144. Moran, G. F. & A. H. D. Brown. 1980. Temporal heterogeneity of outcrossing rates in alpine ash {Eucalyptus delegatus R. T. Bak,). Theor. Appl. Genet. 57: 101-105. Muu AHY, D. L. (editor). 1975. Gamete Competition in Plants and Animals. North Holland Publ. Co., Amsterdam. Nagylaki, T. 1976. Clines with variable migration. Genetics 83: 867-886. Nettancourt. D. de 1977. Incompatibility in Angiosperms. Springer-Verlag, Berlin. NiLSSoN, B. D. 1974. Heterosis in an intraspecific hybridization experiment in Norway spruce, Proc. Joint lUFRO Meeting S.02.04: 1-3. O'DowD, D. J. & M. Hay. 1980, Mutualism between harvester ants and a desert ephemeral: Seed escape from rodents. Ecology 61: 531-540. Orr-Ewing, A. L. 1969. Racial crossing in Douglas fir. Proc. Working Group on Quantitative Genetics Section 22 lUFRO North Carolina State University, Raleigh. OwiNO, F. & B. Zobel. 1977. Genotype x environment interaction and genotype stability in loblolly pine. III. Heterosis and heterosis x environment interaction. Silvae Genet. 23: 114-116. Percival, M. 1965. Floral Biology. Pergamon Press. New York. PhAHLER. P. L. 1965. Fertilization ability of maize pollen grains. I. Pollen sources. Genetics 52: 513-520. Phillips, M. A. & A. H. D. Brown. 1977. Mating system and hybridily in Eucalyptus pauciflora. Austral. J. Biol. Sci. 30: 337-344. Price, M. V. & N. M. Waser. 1979. Pollen dispersal and optimal outcrossing in Delphinium nelsoni. Nature 277: 294-297. Proctor, M. & P. Yeo. 1973. The Pollination of Flowers. Taplinger Publ. Co., New York. Pyke, G. H. 1978. Optimal foraging: Movement patterns of bumblebees between inflorescences. Theor. Populat. Biol. 13: 72-98. , H. R. PuLLiAM & E. L. Charnov. 1977. Optimal foraging: A selective review of theory and tests. Quart. Rev. Biol. 52: 137-154. Roberts, E. H. 1972. Viability of Seeds. Chapman and HalK London. Schaai_, B. a. 1980. Measurement of gene flow In Lupinus texensis. Nature 284: 450^51. & D. A. Levin. 1976. The demographic genetics of Liatris cylindracea L. Amer. Naturalist 110: 191-206. & . 1978. Morphological differentiation and neighborhood size in Liatris cylindracea. Amer. J. Bot. 65: 923-928. Sheldon, J. C. & F. M. Burrows. 1973. The dispersal effectiveness of the achene-pappus units of selected Compositae in steady winds with convection. New Phytol. 72: 665-675. Slatkin, M. 1973. Gene flow and selection in a cline. Genetics 75: 733-756. . 1976. The rate of spread of an advantageous allele in a subdivided population. Pp. 767-780 in S. Karlin and E. Ncvo (editors), Population Genetics and Ecology. Academic Press, New York. & T. Maruyama. 1975. Genetic drift in a dine. Genetics 81: 209-222. SoRENSEN, F. 1969. Embryonic genetic load in coastal Douglas-fir, Pseudotsuga menziesii var. menziesii. Amer. Naturalist 103: 389-398. 1981] LEVIN— DISPERSAL VERSUS GENE FLOW 253 Stapanian, M. a. & C. C. Smith. 1978. A model for seed scatterhoarding: Coevolution of fox squirrels and black walnuts. Ecology 59: 884-896. Stern, K. & L. Roche. 1974. Genetics of Forest Ecosystems. Springer- Verlag, Berlin. Thomson. J. D. & R. C. Plowright. 1980. Pollen carryover, nectar rewards, and pollinator be- havior with special reference to DierviUa lonicera. Oecologia 46: 68-74. ViCKERY. R. K. 1978. Case studies in the evolution of species complexes in Minudus. Evol. Biol. 11: 405^07. Waddington, K. D. 1979. Flight patterns of three species of sweat bees (Halictidae) foraging at Convolvulus arvensis, J. Kansas Entomol. Soc. 52: 751-758. . 1980. Flight patterns of foraging bees relative to density of artificial flowers and distribution of nectar. Oecologia 44: 199-204. Wang, C. W., T. O. Perry & A. G. Johnson. 1960. Pollen dispersion of slash pine {Pinus elliottii Engelm.) with special reference to seed orchard management. Silvae Genet. 9: 78-86. Williams, R. D. & G. Evans. 1935. The efficiency of spatial isolation in maintaining the purity of red clover. Welsh J. Agric. II: 164-171. Wilson, D. E. & D. H. Janzen. 1972. Predation on Schcelca palm seeds by brucid beetles: Seed density and distance from the parent palm. Ecology 53: 954-959. Woessner, R. a. 1972. Crossing among loblolly pines indigenous to different areas as a means of genetic improvement. Silvae Genet. 21: 35-39. Woodell, S. R. J. 1978. Directionality in bumblebees in relation to environmental factors. Pp. 31- 39, in A. J. Richards (editor). The Pollination of Flowers by Insects. Academic Press, New York. Wright, J. W. 1953. Pollen dispersion studies: Some practical applications. J. Forest. 51: 1 14-1 18. Wright, S. 1940. Breeding structure of populations in relation to speciation. Amer. Naturalist 74: 232-248. . 1943. Isolation by distance. Genetics 28: 114-138. . 1946. Isolation by distance under diverse mating systems. Genetics 31: 39-59. . 1951. The genetic structure of populations. Ann. Eugen. 15: 323-354. . 1969. Evolution and Genetics of Populations. Vol. 2. Univ. of Chicago Press, Chicago. Zimmerman, M. 1979. Optimal foraging: a case for random movement. Oecologia 43: 261-267. EVOLUTION OF SEXUAL SYSTEMS IN FLOWERING PLANTS 1 K. S. Bawa^ and J. H. Beach^ Abstract The diversity of sexual systems in plants has been generally attributed to selection for an optimal amount of genetic recombination. However, sexual systems such as hermaphroditism (including het- erostyly), monoecism, andromonoecism, gynomonoecism, dioecism, androdioecism, and gynodioe- cism may also be viewed as different patterns of relative resource allocation to paternal and maternal functions to optimize paternal and maternal reproductive success in different ways. These different patterns may arise in large part in response to reproductive competition resulting from sexual selec- tion. But the efficacy of sexual selection in zoophilous species is mainly determined by pollinator behavior. It follows then that the evolution of a particular sexual system must be influenced by the dynamics of the pollination system. The role of pollinators in the evolution of sexual systems is examined by considering several types of interactions between flowers and pollinators. The role of cost-sharing between paternal and maternal functions in pollinator attraction is stressed in the evo- lution of hermaphroditism. Andromonoecism is considered in terms of loss of pistils in that part of the flower crop which is produced merely to attract pollinators and/or to fulfill male function. In the evolution of andromonoecism to monoecism, the role of stamens of hermaphroditic flowers in the functional integrity of the pollination system is evaluated. The importance of long mouth parts of pollinators to promote compatible pollinations in the evolution of heterostyly is pointed out. The evolution of protogyny is considered in relation to long inhabitation of pollinators in flowers and inflorescences. The evolution of dioecism is examined in relation to the ability of pollinators to respond to minor changes in floral resources thereby altering the patterns of pollen donation and pollen receipt. Finally, the importance of stamens in hermaphroditic plants in attracting pollen col- lecting bees is emphasized in the maintenance of androdioecism. The development of a general hypothesis to explain the diversity of sexual systems will require not only a comprehensive knowledge of pollination ecology but also a revision of the sexual system classification that will take into account functional gender rather than intrinsic gender estimates based solely on morphology. The flowering plants display a wide variety of sexual systems ranging from obligate selfing in association with self-compatibility to obligate outcrossing in conjunction with self-incompatibility (Darlington, 1958; Grant, 1958; Lewis & John, 1963; Mather, 1973; Solbrig, 1976; Jain, 1976; de Nettancourt, 1977). In addition, genetic recombination may be partially or completely circumvented by facultative or obligate apomixis (Stebbins, 1950). Superimposed upon these ge- netic systems are such temporal and morphological mechanisms as protandry, protogyny, heterostyly, monoecism, andromonoecism, gynomonoecism, dioe- cism, gynodioecism, and androdioecism, that are also presumed to regulate the level of outcrossing (see Darwin, 1877; Miiller, 1883; Mather, 1940; Lewis, 1942; Baker, 1959; Jain, 1961; Lloyd, 1975a; Ross, 1970; Charlesworth & Charlesworth, 1978; Charlesworth & Charlesworth, 1979; Ganders, 1979). ^ This work was supported in part by NSF grants DEB 75-21018 and DEB 77-25558. Discussions with Herbert Baker, Ric Charnov, Bill Haber, John Kress, David Lloyd, Richard Primack, Peter Raven, and Colin Webb helped in the clarification of several ideas and in improvement of the manu- script. Paul Opler allowed us to use photographs taken by him. ' Department of Biology, University of Massachusetts, Boston, Massachusetts 02125. ^ Department of Botany, University of Massachusetts, Amherst, Massachusetts 01003. Present address: Department of Zoology, University of Florida. Gainesville, Florida 3261 1. Ann. Missouri Box. Card. 68: 254-274. 1981. 0026-6493/81/0254-0274/$02. 15/0 1981] BAWA & BEACH— EVOLUTION OF SEXUAL SYSTEMS 255 Darwin (1877) was the first to comprehensively document and explain the diversity of sexual systems in plants. His work, including his studies of the effects of self- and cross-fertilization on the fitness of plants (Darwin, 1876), had a major impact on current ideas about the evolution of sexual systems. After the devel- opment of the synthetic theory of evolution, the genetic implications of the di- versity in sexual systems emerged as a major issue in the evolutionary biology of plants, and selective pressure for an optimal amount of recombination came to be viewed as the essential force in the evolution of sexual systems (Stebbins, 1958; Darlington, 1958; Grant, 1958; Baker, 1959; Lewis & John, 1963; Mather, 1973). In turn, patterns of plant sexuality came to be regarded as outcrossing mechanisms and regulators of genetic recombination. Recently, models have been proposed for the evolution of some sexual sys- tems that do not invoke outcrossing as the main selective force. Charnov et al. (1979) and Maynard Smith (1978) have postulated the evolution of hermaphro- ditism, dioecism, and gynodioecism in terms of optimal allocation of resources to male and female reproduction. Janzen (1977) has commented upon the effect of optimal mate selection on the evolution of monoecism and dioecism. The evolution of dioecism has also been examined in the context of sexual selection (Willson, 1979; Charnov, 1979; Bawa, 1980a; Givnish, 1980), dispersal by avian frugivores (Bawa, 1980a; Givnish, 1980; see also Lloyd, 1981), foraging behavior of pollinators (Beach & Bawa, 1980; Beach, 1981), and disruptive selection re- sulting from differential utilization of habitats by male and female plants (Free- man et al., 1980). Pleiotropic effects of male sterility gene have been implicated in the evolution of gynodioecism in Plantago lanceolata (Krohne et al., 1980). There are two major difficulties with the general explanation that selective pressure for outcrossing or an optimal amount of recombination underlies the diversity of sexual systems. First, the argument might explain the evolution of self- versus cross-fertilization but does not account for the tremendous diversity of sexual systems, almost all of which facilitate outcrossing (Willson, 1979). It is possible that different sexual systems result in different levels of outcrossing, but there is no evidence that as one moves from andromonoecism and gynomonoe- cism to dioecism, one moves along a consistent gradient of increasing cross- pollination. In fact, several andromonoecious, monoecious, and gynodioecious species are known to be self-incompatible (see below). Second, it has been dem- onstrated that the ability to self- or cross-fertilize, by itself, is often not a good indicator of the level of recombination in natural populations (see Allard, 1965; Jain, 1976), because the level of recombination is determined not only by the sexual system but also by the mechanics of crossing-over, linkage (Darlington, 1958; Lewis & John, 1963), the foraging behavior of pollinators and seed dispersal agents (Levin & Kerster, 1974), and selection against inbreeding (Jain, 1976). Our discussion of the evolution of flowering plant sexual systems is developed with repeated emphasis of some basic ecological differences between paternal and maternal reproductive success (Horovitz & Harding, 1972; Charnov, 1979; Lloyd, 1979a, 1980b; Willson, 1979). Our approach is based on two proposals. The first is that paternal reproductive success is limited by a plant's ability to disperse pollen to conspecific stigmas, whereas maternal success is usually limited by the amount of nutritional resources available for developing embryos, seeds. 256 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol . 68 and fruits (Bateman, 1948; Charnov, 1979; Lloyd, 1979a). Thus, paternal and maternal reproductive success may be optimized in different ways. The second proposal is that conspecific pollen movement, in particular patterns of dispersal and receipt, is greatly constrained by the type of pollinator or pollination system. We suggest that; (a) sexual systems such as hermaphroditism (including hetero- styly), mont^ecism, andromonoecism, gynomonoecism, dioecism, androdioecism, and gynodioecism be viewed as different patterns of relative resource allocation to paternal and maternal functions to optimize paternal and maternal reproductive success in dissimilar ways (see also Charnov et al., 1976; Lloyd, 1979a); (b) that these different patterns arise mainly in response to reproductive competition re- sulting from sexual selection (Willson, 1979); and (c) that the evolution of a particular pattern is constrained largely by the dynamics of the pollination system. The last point, being new and a crucial element in our proposal, needs elaboration. Although sexual selection influences the relative allocation of resources to male and female functions (Charnov, 1979; Willson, 1979), the actual distribution of these resources in the form of male, female, and bisexual flowers is largely determined by the ecology of the pollination system. In the case of biotic polli- nation, this is a result of the foraging behavior of pollinators determining the pattern of pollen removal and pollen receipt, and consequently the effective role of flowers as pollen donors and pollen receivers (Willson & Price, 1977; Pyke, 1978). But the foraging behavior of pollinators itself is influenced by selection in plants for variation in floral rewards in space and time to optimize the movement of pollinators and thereby pollen flow. Variation in floral rewards may be achieved through changes in the relative proportions of male, female, and bisexual flowers, or of pollen donors and pollen receivers, because male flowers may only produce pollen or pollen as well as nectar, whereas female flowers generally secrete only nectar except in cases involving mimicry (Baker, 1976; Bawa, 1980b). Many sexual systems may simply represent such variations that have coevolved with the foraging behavior of pollinators. Our treatment of sexual systems here em- phasizes such coevolution and the role of pollinators in differentially influencing paternal and maternal reproductive success. In addition to the putative ecological and energetic advantages of different patterns of floral sexuality that we review below, a number of attempts have been made to elucidate the adaptiveness of hermaphroditic organisms (in the broad sense, to include, for example, monoecious plants) over unisexual individuals (Baker, 1967; Maynard Smith, 1978; Heath, 1977; Charnov, 1979; and see Lloyd, 1981, for several additional references). Although these proposals are relevant to the adaptive nature of hermaphroditism vs. unisexuality at the individual level, we limit our discussion here to the ecology of floral sexuality. We have avoided the use of the term breeding system throughout the paper in favor of sexual system. This seems more appropriate as it does not carry the implication that plant gender is the sole result of selection for a certain level of genetic variability. The sexual system does include those factors that directly influence the level of outcrossing, but we view the distribution of male and female functions in space and in time, and the ecological interactions among individuals that mate with each other, as being primarily the result of the coevolution between sexual partners and also between flowers and pollinators. 1981] BAWA & BEACH— EVOLUTION OF SEXUAL SYSTEMS 257 Table L Standard classification of flowering plant sexual systems as currently used.^-^ A. Systems based on the spatial distribution of male and female reproductive organs. L Sexually monomorphic^ systems characterized by only one gender class of individuals. 1. Hermaphroditism: Plants bear only bisexual flowers. 2. Monoecism: Plants bear male and female flowers. 3. Andromonoecism: Plants bear bisexual and male flowers. 4. Gynomonoecism: Plants bear bisexual and female flowers. II. Sexually dimorphic species characterized by two gender classes of individuals. "* 1. Dioecism: Plants bear either male or female flowers. 2. Gynodioccism: Plants bear either female or bisexual flowers. 3. Androdioecism: Plants bear either male or bisexual flowers. B. Systems based on the temporal distribution of male and female organs. 1. Protandry: Pollen removed from the anthers before stigmas attain receptivity. 2. Protogyny: Stigmas become receptive before anthers release pollen. C. Systems based on the presence or absence of self-incompatibility alleles. 1. Self-incompatibility: Plants polymorphic with respect to the presence of self-incompatibility alleles; pollinations involving pollen and stigma sharing the same self-incompatibility alleles, including self pollinations, result in no fruit set. 2. Self-compatibility : Plants monomorphic and without the presence of self-incompatibility alleles; all pollinations, including self-pollinations, result in fruit set. D. Systems based on variation in style and stamen length. 1. Distyly: Two types of individuals that bear different forms of flowers, pin flowers with long styles and short stamens and thrum flowers with short styles and long stamens. Self-pollination and pollination within the morphs generally incompatible. 2. Tristyly: Three types of individuals that bear long-, mid-, or short-style flowers. Anthers occupy two out of the possible three positions, for example, long-style flowers have anthers at the short and mid position, mid-style flowers have anthers at the short and long position, and so on. Compatible pollinations result from crosses involving stigmas and anthers at the same level. '^ In addition to the systems described below, there exist other systems such as cleistogamy and various forms of apomixis (see Stebbins, 1950). *' The systems described below are not mutually exclusive. ^ The use of monomorphic and dimorphic follows that of Lloyd (1972a, 1979a. 1980a). These terms should not be confused with their application elsewhere, usually to describe floral heteromorphism based on variation in style and stamen length (Ganders. 1979). *" Plants of sexually dimorphic species may exhibit considerable variation in sex expression, especially in gynodioecious species in which individuals with bisexual flowers may be partially or fully female-sterile. Classification of Sexual Systems Lloyd (1980a) has pointed out difficulties with the existing classification of flowering plant sexual systems. The descriptive terms used by taxonomists and ecologists alike, being derived from Linnaeus's (1737) artificial classification of flowering plants based on sexual systems, are typological, qualitative, and defined by arbitrary limits. Because of this and for additional reasons (see Discussion) the available terminology neither adequately describes patterns of sexuality in plants, nor their effective gender. The work of Lloyd (1980a) in establishing quantitative measures of plant gender is of great value. However, for purposes of our initial discussion, we will use the traditional sexual system categories as shown in Table 1. Sexually Monomorphic Systems Hermaphroditism, — We use the term hermaphrodite in a restricted sense to designate those species with simultaneously bisexual flowers. Most flowering plants have only bisexual flowers (Yampolsky & Yampolsky, 1922; Lloyd, 1981) and have their pollen distributed by a diverse array of biotic 258 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 and abiotic agents (Faegri & van der Pijl, 1971). The ecological and evolutionary significance of bisexuality was emphasized by Baker & Hurd (1968), who sug- gested that the coevolution of hermaphroditic flowers with animal pollination might be an important advancement by early angiosperms since pollen-producing and pollen-receiving organs in the same flower allowed for efficient simultaneous deposition and removal of pollen. Baker & Hurd (1968) and Crepet (1979) have argued that since the original attraction of visiting insects for flowers was the presence of pollen for food, there would have been no incentive for pollinators to visit a female flower, giving an advantage to plants with hermaphroditic flowers. Charnov et al. (1976) and Maynard Smith (1978) have proposed that bisexual flowers sometimes represent the optimal use of energetic resources available for reproduction, since the fixed costs associated with male and female functions would be shared. Such costs would include, for example, bracts subtending flow- ers, pedicels supporting flowers, flower parts that serve in pollinator attraction, and nectar rewards. It is significant that in contrast to sexually dimorphic species, hermaphroditic species generally have large, showy flowers (Bawa & Opler, 1975; Bawa, 1980a). Dioecism is poorly represented in the Araceae and Palmae in which the energetic investment into large bracts (spathes) and inflorescence stalks (spadices) presum- ably far exceeds the investment into very small male and female flowers. These correlations are consistent with the hypothesis that whenever common costs of male and female functions are large relative to the costs of the production of the two types of gametes, hermaphroditism may be favored over unisexuality (Heath, 1977; see also Lloyd, 1979a). Note, however, that Heath (1977) proposed the hypothesis to explain the evolution of hermaphroditism vs. unisexuality in ani- mals and that he defined hermaphroditism in a broad sense to cover monoecious as well as hermaphroditic species. It is difficult to assign a single most important selective force to the evolu- tionary rise and maintenance of bisexual flowers because, ecologically, hermaph- roditism encompasses a diverse group of plants. The production of functional male and female gametes does not mean that either the flower or the individual plant contributes to the next generation equally via the male and female pathways (Horovitz & Harding, 1972; Lloyd, 1979a; Willson, 1979). Willson & Rathcke (1974) and Willson & Price (1977) have provided evidence that in milkweeds {Ascleplas spp.) an increase in the number of flowers in an inflorescence results in a greater genetic contribution via pollen to the next generation, but not via ovules. In addition, phenomena such as dichogamy, self-incompatibility, and het- erostyly make hermaphroditic species a complex assemblage of plants. Andromonoecism . — Andromonoecism has been reported in species pollinated by bats (Heithaus et al., 1974), bees (Bell & Lindsey, 1978; Symon, 1979; Bertin, 1981), bees and flies (Primack & Lloyd, 1980), hummingbirds (Bertin, 1981), and moths (Bawa, unpublished data). In grasses, andromonoecious species are wind pollinated (Connor, 1979). The evolution of andromonoecism has been generally ascribed to selective pressure for increasing cross-fertilization (e.g., Heithaus et al., 1974), but this explanation is incomplete for several reasons (Primack & Lloyd, 1980), including the existence of self-incompatibility in several andromonoecious species (Zapata 1981J BAWA & BEACH— EVOLUTION OF SEXUAL SYSTEMS 259 & Arroyo, 1978; Bawa, unpublished data), and thus other selective forces for the evolution of the sexual system should be considered. It is a common observation that many hermaphroditic plants generally bear many more flowers than the number of fruits that are matured (see Bawa, 1974, for fruit/flower ratios in several species). Those flowers that do not set seed may function to either attract pollinators and/or to disperse pollen (Willson & Price, 1977; Lloyd, 1979a). Andromonoecism can simply be regarded as representing the situation in which nonfunctional pistils are aborted prior to flowering in those flowers that are destined to serve male or attraction functions (Zapata & Arroyo, 1978). However, it is noteworthy that andromonoecism, though widely distrib- uted, is relatively rare as compared to hermaphroditism, whereas the phenome- non of ''excess'' flower production is very common in flowering plants. The question then arises as to the significance of pistils in hermaphroditic flowers that largely act as pollen donors. Three possibihties might be considered. First, in many species, especially those with extremely specialized pollination mecha- nisms, e.g., Apocynaceae and Orchidaceae, the abortion of pistils could disrupt the pollination system by structurally perturbing the floral morphology. Second, the abortion of pistils in many flowers before pollination could restrict the efficacy of selection on progeny acting through control over pollen germination, pollen tube growth, and embryo and fruit abortion. Third, pistils may not be aborted in most species because there is no predictability before pollination with respect to the fate of flowers as pollen donors and pollen recipients (Lloyd, 1980b). Additional factors in the evolution of andromonoecism have been recently explored by Primack & Lloyd (1980). Gynomonoecism. — As compared to andromonoecism, gynomonoecism is re- stricted in its distribution, being known in less than half a dozen families (Yam- polsky & Yampolsky, 1922). The Compositae contains the greatest number of gynomonoecious taxa (Loyd, 1979a). Unfortunately, unlike andromonoecism, detailed information about the poUination biology of gynomonoecious species is not available. In the Compositae gynomonoecism results from the sterilization of stamens in the ray (peripheral) florets of the inflorescence. By being petaloid the ray florets enhance the attractiveness of the inflorescence consisting of small flowers. This gives the inflorescence a flowerlike structure, and makes it a func- tional pollination unit. It is possible that in this family the selective pressure for attractive petaloid ray florets has led to the sterilization of stamens; the large size of male-sterile florets in the Compositae, and of male- and female-sterile flowers in Viburnum and other genera raises the possibility that the resources expended in stamens and pistils may be easily reallocated to other floral organs such as petals. Indeed, by attracting pollinators, ray florets in the Compositae influence the level of outcrossing. However, a consideration of gynomonoecism as a pol- lination rather than an outbreeding system makes it easier to explain why the vast majority of gynomonoecious species occur in the Compositae. Lloyd (1979a) suggests two other factors to account for the evolution of gy- nomonoecism in the Compositae, One is that the increase in the number of ''pol- liniferous'' flowers in the capitulum would result in neither greater floral display nor an increase in the number of visits by pollinators since it is the capitulum rather than the individual flowers that functions as the unit of attraction. The 260 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Other is that in a capitulum composed of uniovulate flowers, gynomonoecism may be the only way to increase the number of seeds without increasing the number of "polliniferous" flowers. Monoecism.— The sexual specialization of flowers represented by andromon- oecism and gynomonoecism is carried a step further in monoecious taxa that are characterized by the presence of male and female flowers on the same plant. Contrary to the popular viewpoint (Yampolsky & Yampolsky, 1922; Grant, 1951; Stebbins, 1951; Faegri & van der Pijl, 1971; Charlesworth & Charlesworth, 1979) monoecism is not confined to wind-pollinated plants. In tropical forests a large number of zoophilous species are monoecious. The vast majority of monoecious species in a dry deciduous and a wet evergreen forest in the lowlands of Costa Rica are insect pollinated, a few are hummingbird pollinated (Bawa, un- published observations). It has been argued that monoecism has evolved as a result of the selective advantages of cross-fertilization (e.g., Maynard Smith, 1978), though Godley (1955) has demonstrated the presence of self-incompatibility in several species. We propose that evolution of monoecism represents the continuation of the trend seen in andromonoecious species towards the specialization of flowers into pollen donors and pollen recipients which is due in part to sexual selection and in part to the mechanics of pollination. Monoecism can arise in one step from andro- monoecism by the sterilization or abortion of the stamens in hermaphroditic flow- ers. In relation to the role of pollinators in the evolution of monoecism, we discuss the conditions under which andromonoecism might evolve into monoecism and the conditions under which andromonoecism remains stable. Selection for the sterilization or abortion of stamens in hermaphroditic flowers may occur under two conditions. The first condition is when pollen in the flowers interferes with the deposition of incoming cross-pollen (Bawa, 1980a). Such in- terference is likely when the pollination mechanism is imprecise and flowers are small and closely aggregated in an inflorescence, e.g., Araceae, Euphorbiaceae, Moraceae, Palmae, and others. In many taxa of these families, not only is there spatial separation, but also temporal separation of male and female flowers, sug- gesting perhaps the role of interference as well as sexual selection in the spatial and temporal differentiation of male and female functions. Interference is also likely when pollen is picked up from and deposited on the inflorescence during the same foraging trip, as is the case in some protogynous, monoecious aroids. In other words, if the male and female phases cannot be separated in time in the same flower, they might be separated in space, and by further evolution in space and time. It is noteworthy that many andromonoecious species have large flowers, when contrasted with monoecious species, and, furthermore, these flowers are not very closely aggregated in inflorescences. But in andromono- ecious Umbelliferae the flowers are small, clustered together, and even oc- cur in dense capitate heads (Webb, 1980). However, the temporal separation of male and female phases in bisexual flowers, as well as at the level of inflores- cences and individuals, is very pronounced (Miiller, 1883; Cruden & Hermann- Parker, 1977). Second, male sterility in hermaphroditic flowers may evolve when the optimal conditions for female reproductive success (pollen receipt) and for male repro- 1981] BAWA & BEACH— EVOLUTION OF SEXUAL SYSTEMS 261 Figure 1. Hermaphroditic flowers of Bomhacopsis quinata (a), Bauhinia paulctia (b). Pithc- cellohium saman (c), and Capparis sp. (d), to show the importance of stamens in the maintenance of andromonoecism (see text for details). All species, except the particular species of PitheceUohium shown here, are andromonoecious; however, andromonoecism has been found by us in some species of PitheceUohium closely similar to this particular species. The species are from lowland dry deciduous forest in Costa Rica; Photographs (a) and (b) are by P. A. Opler. ductive success (pollen donation) are vastly different and strongly influenced by the position of male and female flowers (Heslop-Harrison, 1972). For example, in zoophilous plants where pollinators typically forage from the bottom towards the top in one-day inflorescences, one may expect female flowers at the bottom and male at the top of the inflorescence, as is the case in many monoecious species of the Euphorbiaceae. Under certain conditions pollinators may select against the loss of sterilization of stamens in the hermaphroditic flowers of andromonoecious species; they there- by impose constraints upon the evolution of andromonoecism towards monoe- cism. For example, in those cases where the stamens play a large role in the integrity of the pollination system, there would be strong selection against their loss. In andromonoecious species such as Bauhinia paulctia, Bomhacopsis quin- ata. and Capparis pittcri the loss of stamens would destroy the integrity of the flowers or the attractiveness of the inflorescence (Fig. 1). In many mimosoid legumes, the organization of the flowers is largely dependent upon stamens (Fig. Ic) so that one would not expect the andromonoecious species (known to occur in Albizzia, Calliandra, and Pithcccllobium, W, Haber, pers. comm.) to evolve 262 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Short Style or 'Thrum' Long Style or 'Pin' Male Female Figure 2. Flower forms of distylous species and derived dioecious laxa. The thin arrows be- tween the upper pair of flowers indicate the poUinations that result in fertilization in distylous plants. 1981] BAWA & BEACH— EVOLUTION OF SEXUAL SYSTEMS 263 into monoecious taxa. In Solanum, another genus where andromonoecism is common, stamens may not only play a part in the attraction of the pollinators, but also offer the sole reward to the pollinators since the flowers contain no nectar (Anderson, 1980; W. Haber, pers. comm.). In andromonoecious Lepto- spermum of the Myrtaceae stamens again may be crucial in attracting the flower visitors (Primack & Lloyd, 1980). For Aesculus pavia, Bertin (1981) suggested the role of pollen as a food reward in preventing the evolution of andromonoecism to monoecism. It is not certain if monoecism generally evolves from andromonoecism. There is little discussion of different evolutionary pathways to monoecism in the liter- ature (but see Lloyd, 1972a, 1972b, 1975b). Regardless of the evolutionary path- ways involved, selective forces other than those associated with pollination may have also played a role in the evolution of this sexual system. Heterostyly, — The flower heteromorphisms characteristic of distyly and tri- styly have from the earliest study been recognized as structural adaptations to promote disassortative pollination, i.e., the movement of pollen between incom- patibility groups (Darwin, 1877). Heterostyly was probably the first sexual system to be recognized as partly an ecological phenomenon. Distyly, the most common expression of heterostyly is usually associated with gamopetalous, tubular flowers and pollination by relatively long-tongued lepidop- terans, hymenopterans, or hummingbirds. The significance of this tripartite re- lationship among the sexual system, flower morphology, and mode of pollination lies in the fact that the efficacy of pollen transfer from short stamens to short styles (or long stamens to long styles) is contingent upon the deposition of pollen at different locations on the mouth parts of the pollinators. The corolla tube must be relatively long and narrow to allow only restricted access by nectar-seeking probes in order to assure the accurate localization of pollen deposition on the vector, thus promoting subsequent pollen transfer between stamens and styles of the same length. Most heterostylous species are self-incompatible (Ganders, 1979). The evo- lution of the self-incompatibility system, which prevents both self-fertilization and mating between plants of the same flower form, probably occurred before the rise of the associated floral hetermorphisms (Ganders, 1979). It is most likely that the morphological features of distyly and tristyly, as part of the pollination system, arose as a response to the appearance of a limited number of incompat- ibility groups in order to increase disassortative (compatible) pollination and thus a plant's reproductive output (Ganders, 1979; Beach & Kress, 1980). Clearly, both the function and the adaptive basis for the evolution of heterostyly can only be understood by considering the breeding system as an ecological phenomenon: an adaptation to manipulate pollinator movements and pollen flow. The role of pollinators in the evolution of sexual systems is demonstrated more markedly by the conversion of distyly into dioecism (Fig. 2). This change. The heavy vertical arrows represent the evolutionary pathways that have given rise to unisexual flowers. Vestigial styles and stamens are not shown in unisexual flowers. In most cases the change from distyly to dioecy is accompanied by a reduction in the size of the corolla tube. (See Beach & Bawa, 1980, for details) 264 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 If from one outbreeding system to another has occurred in several genera in the families Boraginaceae and Rubiaceae (Baker, 1958; Lloyd, 1979b; Beach & Bawa, 1980). In every known case where dioecism has evolved in this way, the ancestral long-style form has become female, and the short-style plants have become males (Fig. 2). This would not be the result if selection for unisexuality was unrelated to the ancestral flower condition. Beach & Bawa (1980) have proposed that such a switch from distyly to dioecism is most likely the result of a form of pollinator- mediated selection for femaleness in the long-style and maleness in the short- style form. In Coussarea tahimancana Standley (Rubiaceae), the evolution of dioecism is probably the result of a switch in the pollinating fauna from moths that are the characteristic pollinators of the genus Coussarea in Costa Rica, to short-tongued bees that are incapable of reaching the lower floral organs (short- styles and short-stamens). Lloyd (1979b) and Beach & Bawa (1980) discuss ad- ditional genetic and ecological features of the evolution of dioecism from distyly. Protandry and Protogyny. — The terms protandry and protogyny have been used at the level of the individual, for example, to describe changes in sex expres- sion between reproductive seasons in perennial monoecious plants (Frankel & Galun, 1977). In contrast, the terms are conventionally reserved by zoologists to describe changes in sex expression that occur only once during the life of an organism (Heath, 1977). Protandry and protogyny are used here with reference to plants that display dichogamy within a single reproductive season and, more specifically, that usually undergo a male-female or female-male transition within a period of a few days, or as little as a few hours. The terms are also applied here with reference to the relative timing of male and female functions within a flower or at most an inflorescence. Dichogamy is generally assumed to be an adaptation to limit self-pollination and to promote cross-pollination (Miiller, 1883; Proctor & Yeo, 1972; but see Onyekwelu & Harper, 1979; Lovett-Doust, 1980). Undoubtedly, differences in the timing of anther dehiscence and stigma receptivity influence the amount of incoming and outgoing pollen. However, if selective pressure for outcrossing was the only factor involved in the evolution of protandry and protogyny, one would expect these systems to occur in almost equal frequencies and to be distributed randomly throughout the flowering plants. But protandry is far more common than protogyny (Burtt, 1977), and, as discussed below, protogyny seems to be largely confined to certain taxonomic groups and pollination modes. Protandry should be very common in flowering plants for two reasons. First, intrasexual selection or intraspecific competition for mates should promote the dispersal of a plant's pollen before conspecific stigmas have received pollen from other genotypes (see also Webb, 1981). At the same time, selection should favor the receptivity of the stigma when the pollinators have removed pollen from a diverse array of genotypes. On these considerations alone, one may expect protandry to be an almost universal feature of flowering plants, but since this is not the case, the factors that result in the evolution of protogyny will be explored below. Second, the fact that pollen but not the ovules undergo dispersal makes the conditions for the evolution of protogyny more stringent than those for protandry. Consider, for example, a population consisting of outcrossing individuals in which the hermaphroditic flowers that open on a given day last only for that day (e.g.. 1981] BAWA & BEACH— EVOLUTION OF SEXUAL SYSTEMS 265 sunrise to sunset). In such a population protogyny can not evolve unless one presumes the pollinators are carrying substantial amounts of pollen from the foraging undertaken in the previous day. However, in the same population all plants can be simultaneously protandrous because the stigmas can be matured later after some pollen has been deposited on the bodies of the pollinators. As the longevity of the flowers increases, the conditions for the evolution of proto- gyny become less stringent. Thus in hermaphroditic species, protogyny might only evolve when the processes of pollen receipt and pollen donation in a flower are extended over one daylight period. As discussed below, the flowers of many protogynous species indeed do extend beyond one day. Although extensive data on flower longevity for hermaphroditic angiosperms as a whole are lacking, flowers last one day (i.e., one daylight period) in the vast majority of hermaphroditic plants in the tropics. In many of these species almost all pollen is removed within a few hours after anthesis in early morning, but the peak in nectar production is W W the stigmas as pollinators continue to forage for several hours after the pollen has been removed. The late deposition of pollen is also indicated by slight exsertion of stigmas in some of these species in late morning. Although direct evidence for protandry is lacking in these species, they certainly are not, and cannot be, protogynous under the given conditions of flower longevity and the pollinator foraging behavior. The differences between male and female gametes in dispersal imposes an additional requirement to the evolution of protogyny. Because of the reasons outlined above, unlike protandry, the operation of protogyny is dependent upon some plants being in the male and others in the female phase at a given time. This is usually brought about by asynchronous development of flowers and inflo- rescences, as for example in the species of the Annonaceae, Araceae (Fig. 3), Cyclanthaceae, Moraceae, and Palmae. In other species, for example in Persea gratissima, plants are dimorphic with respect to the timing of male and female phases: in one type the flowers in the female stage open in the morning and then close in the afternoon to reopen in the male stage the following afternoon; in the other type the flowers open in the afternoon in the female stage and then in the male stage in the following morning (Stout, 1924). The evolution of protogyny can be traced to three aspects of pollination bi- ology. First, protogyny has coevolved in conjunction with several specific life- history aspects of pollinators. In many species of Magnoliaceae, Annonaceae, Araceae, Cyclanthaceae, and Palmae (among others), protogyny is associated with cantharophily (Faegri & van der Fiji, 1971; Bawa & Beach, unpubUshed observations). The pollinating beetles fly in the late afternoon or early evening to flowers (or inflorescences) while carrying pollen from other conspecific plants and then crawl into some type of enclosure formed by spathes, bracts, or perianth parts and while doing so. deposit pollen on stigmatic regions; after spending the night and most of the following diurnal period in the enclosures, the beetles emerge to fly to another flower or inflorescence usually on a different individual (Fig. 3). Pollen is released just prior to the beetles' departure. The important aspect here is not the pollination by beetles per se, but: (I) the time of beetle flight behavior, (2) the long residency of the pollinators in floral structures, and 266 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol . 68 FiGURF-: 3. Flower of Cymhopeialum sp. (a) and an inflorescence of Dieffenbachia sp. (b). A petal of the Cymbopetalum flower has been removed to show the exudate on the receptive stigmas and the tightly packed stamens surrounding the gynoecium. The inflorescence of Dieffenbachia bears female flowers at the bottom, the portion shown to be completely enclosed by the spathe, and male flowers at the top, the portion shown to be exposed. Both the species are from a tropical lowland wet evergreen forest in Costa Rica. (3) one foraging trip every 24 hours. Recall that in Ficus, a genus in which protogyny is an universal feature of the monoecious species, the prolonged in- habitation of wasps is also associated with just one trip between the pollinated and the to-be-pollinated figs (Ramirez, 1969). It is apparent that the type of beetle pollination in the families mentioned above, and pollination by fig wasps could not operate and would not evolve in association with protandry. Second, selective pressure against the clogging of stigmas by a plant's own pollen may result in the evolution of protogyny. The self-pollen may interfere with the deposition of incoming pollen and/or compete with it for germination sites (Bawa & Opler, 1975). The possibility of clogging increases when the flowers are closely aggregated (Burtt, 1977) and the pollination mechanism is imprecise. There are no observations or data that relate the amount of self-pollen received by the stigma (after the termination of the female phase in protogynous species) to the precision of pollination in either the protogynous or the nonprotogynous taxa. This explanation is different from the traditional explanation that seeks the evolution of protogyny in selective pressures favoring outcrossing, because it removes the difficulty of explaining the occurrence of protogyny in self-incom- patible species (see for example Burtt, 1977). Clogging in self-incompatible as well as self-compatible species decreases the amount of incoming pollen that can be deposited on the stigma, as well as the amount of outgoing pollen. Lloyd & 1981] BAWA & BEACH— EVOLUTION OF SEXUAL SYSTEMS 267 Yates (1981) have used a similar explanation to account for the evolution of protandry in Wahlenhergia albomarginata. Third, uncertainty of cross-pollination may also select for protogyny. Pollen from the anthers of the same flower could be used for pollination if the initial effort in securing cross-pollination were to fail. Protandry may offer no such possibility. It is notable that protogyny is quite common among herbaceous plants that flower early in spring in the north temperate zone, when the conditions for cross-pollination are unpredictable (Schemske et al., 1978). The consideration of the evolution of protandry and protogyny is complicated by the fact that the plants can be protandrous at the level of the individual and protogynous at the level of flowers and inflorescences or vice versa. For example, most species in the Umbellifprae are protandrous, and the species in which pro- togyny has been reported have the first order umbels consisting only or mostly of male flowers (Bell & Lindsey, 1978). Thus the so-called protogynous species are, in most cases, actually protandrous. From the evolutionary viewpoint it is the individual-level phenomena that are of interest. However, our phenological knowledge of differential maturation of male and female parts is in most cases restricted to flowers or inflorescences. Sexually Dimorphic Systems Dioecism. — Dioecism is widespread in flowering plants. Many attempts have been made to explain its evolutionary basis, and until recently, most evolutionary models have dealt with the genetic benefits of outcrossing as the selective force of most importance (Charlesworth & Charlesworth, 1978; Maynard Smith, 1978; and references therein). Although there is some empirical evidence to support the outcrossing argument (Lloyd, 1981), several alternative models to explain the evolution of dioecism have been proposed (Charnov et al., 1976; Charnov, 1979; Willson, 1979; Bawa, 1980a; Givnish, 1980; Beach, 1981). We will not review all of these recently published proposals, but rather briefly examine those models that deal with the consequences of pollinator foraging behavior on the evolution of the sexual systems. Most dioecious plants are insect pollinated (Bawa, 1980a), although in the north temperate zone there seems to be an association between dioecism and wind pollination (Freeman et al., 1980). Among the zoophilous taxa, an unusually large number of species are pollinated by small bees or flies (Bawa & Opler, 1975). This correlation between dioecism and pollination by small insects occurs at the taxonomic as well as at the community level of organization (Bawa, 1980a), Three attempts have been made to elucidate the ecological and evolutionary basis of this correlation. According to Givnish (1980) pollination by small opportunistic insects is in- efficient in the sense that they mediate little interplant movement of pollen. Thus, increases in male reproductive effort (in originally hermaphroditic plants) do not result in corresponding increases in paternal fitness. In contrast, Givnish argues that an increase in maternal reproductive investment (e.g., the maturation of more fruits) should result in a disproportionate gain in female fitness. As a consequence, he proposes that female individuals could successfully invade an hermaphroditic 268 ANNALS OF THE MISSOURI BOTANICAL GARDEN |Vol. 68 population J and then create conditions favorable for the evolution and establish- ment of males. Bawa (1980a) and Beach (1981) have argued that pollinators such as small generalist bees respond dramatically to changes in floral displays. Thus an in- crease in flower number on an hermaphrodite may lead to a disproportionate increase in male fitness because plants with larger floral displays may either attract more pollinators and thereby disperse more pollen, or be visited earlier in the day and thereby transmit more genes via pollen than via ovules. The variation in flower number may result from intrasexual competition or may be a part of the normal variation in a natural outcrossing population. As a consequence of this variation and the type of pollinator-mediated selection described here, males would be established. Females may become established when individuals with a smaller number of flowers increase their fitness due to resources saved from reduced pollen dispersal costs (see Bawa, 1980a, for other factors leading to the establishment of females). Lloyd (1981) has suggested that dioecism is more likely to arise in species serviced by small promiscuous insects because pollination by such pollinators would result in a high level of selfing in self-compatible species. Regardless of which factor has contributed more to the observed correlation, the involvement of pollinators in the evolution of dioecism cannot be denied. Gyfwdioecism, — The evolution of gynodioecism involves the establishment of male-sterile mutants in a population consisting of hermaphrodites. There is evi- dence that selective pressure for outcrossing is responsible for the spread of such mutants (Lloyd, 1981, and references therein). Thus, at this time, pollinator-me- diated selection cannot be invoked to explain the evolution of gynodioecism, though it is noteworthy that the majority of gynodioecious species are also pol- linated by small insects (D. G. Lloyd, pers. comm.). Gynodioecism usually evolves into dioecism by the gradual loss of female fertility of the hermaphrodites (Lloyd, 1975a; Charlesworth & Charlesworth, 1978). Selective pressures underlying the conversion of hermaphrodites into males are not fully understood; however, according to Charlesworth & Charles- worth (1978), an increase in pollen production is a requirement. If production were equated with dispersal, increased dispersal could result from pollinator- mediated intrasexual competition for the females, especially if the gain in fitness from increased dispersal outweighed the loss of fitness due to the elimination of female functions. It is noteworthy that in the Umbelliferae, dioecious species have a higher male/female flower ratio than the gynodioecious species (Webb & Lloyd, 1980). It is possible that an increase in male flower number in dioecious species has resulted from intrasexual competition, Androdioecism. — The establishment of males in a population consisting of hermaphrodites results in the evolution of androdioecism. It is generally assumed that selective pressure for outcrossing does not result in the evolution of andro- dioecism because in a selfing population, the ovules of hermaphrodites are not readily available to male plants (Lloyd, 1975a; Charlesworth & Charlesworth, 1978). In the absence of selfing, the pollen production of males must be more than twice that of hermaphrodites in order for androdioecism to evolve. Such a dramatic increase in pollen production (and dispersal) may be possible when 1981] BAWA & BEACH— EVOLUTION OF SEXUAL SYSTEMS 269 pollination is effected by pollen collecting visitors (see also Ross, 1980). Andro- dioecism has been reported in some species of Solarium (Symon, 1979). The flowers of Solanum produce no nectar; pollen is the only reward to pollinators, which are pollen-collecting bees. We would predict that most additional examples of the evolution of dioecism from andromonoecism are likely to be reported from bee-pollinated species. Discussion We systems stems from a multitude of interactions between plants and pollinators: (1) ability of pollinators to respond to minor changes in floral resources, thereby altering the patterns of pollen donation and pollen receipt (evolution of dioecism); (2) need in plants to retain structures crucial to the integrity of the pollination system (maintenance of andromonoecism, androdioecism); (3) a single foraging trip by the pollinators to flowers or inflorescences associated with long inhabi- tation in these structures to find mates, avoid predators, or gather food (evolution of protogyny); (4) interference between pollen removal and pollen receipt in plants with small flowers pollinated by unspecialized insects (evolution of monoecism); and (5) precise deposition of pollen on the long mouth parts of pollinators to promote compatible pollinations (evolution of heterostyly). Although we have cited examples where other factors may have been more important, we do not deny the role of outcrossing in the evolution of sexual systems. Different selective forces may operate at different levels. Selection for outbreeding or for an optimal amount of recombination may explain why plants are cross- or self-fertilized, while sexual selection, including pollinator-mediated selection, may explain why outcrossing is achieved in different ways, or why some species are hermaphro- ditic, others andromonoecious, monoecious, gynomonoecious, dioecious, gyno- dioecious, or androdioecious. We have so far considered only spatial and temporal patterns of floral sex- uality. It is, however, worth emphasizing that the evolution of self-compatibility and self-incompatibility too is not independent of pollination events. In several taxa, the evolution of self-compatibility has been traced to the paucity of polli- nators due to inclement weather (Hagerup, 1951), competition for pollinators (Levin, 1972), changes in pollinator fauna (W. B. Haber & G. W. Frankie, pers. comm.), and the traplining behavior of pollinators that precludes the necessity for physiological self-incompatibility (W. J. Kress, pers. comm.). Self-incompat- ibility or the ability to discrimate between self- and cross-pollen has evolved only in angiosperms. The flowering plants are also unique in the sense that only in this group does a large number of diverse pollen genotypes land on the stigma as a result of animal pollination (Mulcahy, 1979). In gymnosperms, for example, very few pollen grains reach the pollination chamber (Stern & Roche, 1974). Fisher (1958: 143) was the first to consider the theoretical significance of discrimination against different pollen genotypes in the context of sexual selection. He cited the work of Jones (1928) on G, factors in maize to underscore the fact that the discrimination can also involve pollen from different genotypes within the same species, and is not necessarily restricted to self- versus cross-pollen. We have argued that the evolution of sexual systems is constrained by the 270 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 way the pollinators interact with flowers. But, except in a few instances, we have been unable to predict the type of sexual system that would coevolve with a particular feature of pollination. The tremendous diversity of plant sexual systems and their secondary modifications, the wide variety of pollinators and their di- verse behaviors, and the lack of general knowledge about the ecological relation- ships between sexual systems and pollinators make it difficult to develop a general hypothesis. In addition, we have considered sexual systems within the existing classification, but the classification is based on the morphological description of sexual systems, and is inadequate for several reasons. First, the purely morphological description masks a considerable amount of quantitative variation within different systems. A number of recent studies (Bawa, 1974; Zapata & Arroyo, 1978; Willson & Price, 1977; Schemske, 1980) have quantitatively demonstrated what has been widely observed in hermaphroditic flowering plants: in many species only a small minority of flowers function to produce seeds and fruits. When traditional sexual system criteria are used to evaluate morphological or intrinsic gender estimates, all these species are clas- sified as hermaphroditic on the basis of potential flower function or preanthesis gender. However, if the actual performance of the flowers is taken into account with estimates of functional gender or effective gender, we must conclude that since most flowers function at most as pollen donors, the sexual system of most hermaphroditic species would be more accurately described as andromonoecious. Thus, when gender estimates which include postfertilization events, or at least the probability of male and female function, are contrasted to prefertilization estimates, we note that the same species could be characterized to have two different systems under the existing classification. A second inadequacy of our current classification is that it falls short of fully describing the temporal distribution of sexual function in a species. The consid- eration of the temporal dimension changes the properties of the sexual system deduced from morphological grounds alone. A monoecious plant that matures male flowers first and female flowers several days after the male phase does not have the same sexual system as a plant in which male and female flowers mature more or less at the same time. Furthermore, the plant is neither protandrous in the same way as a plant with hermaphrodite flowers, nor is it dioecious, as such plants are sometimes described in the absence of information concerning the temporal sequence of male and female flowers (Bawa, 1977). Finally, the morphological classification does not take into account the way the system functions. Faegri & van der PijI (1971), in discussing such well-known examples as the heads of Compositae, Dipsacaceae, and some Leguminosae, point out that in many cases the morphological differences between flowers and inflorescences are in themselves irrelevant in pollination ecology. More impor- tant, in the context of this discussion, these differences can also be irrelevant in the characterization of sexual systems. Consider, for example, the two scarab beetle-pollinated genera, the hermaphroditic Cymbopetalum (Annonaceae) and, the monoecious Dieffenbachia (Araceae) (Fig. 3). The beetles visit these species in the manner described above under Protandry and Protogyny, In brief, incom- ing beetles bearing pollen from other individuals inhabit the protogynous flowers {Cymbopetalum) or inflorescences {Dieffenbachia) for about 24 hours and then 1981] BAWA & BEACH— EVOLUTION OF SEXUAL SYSTEMS 27 1 after the flower or inflorescence completes the male phase, leave in search of additional plants carrying pollen from the flowers or inflorescences just inhabited. The pistils in the Cymbopetalum flowers receive pollen in the same way as the female flowers of the Dieffenhachia inflorescence, and similarly the stamens re- lease pollen in the same way as the male flowers of the aroid. Thus, from a functional view the Cymbopetalum flower is an analogue of Dieffenhachia inflo- rescence. Current sexual system classification obscures the role of the monecious inflorescence as a functional unit. In such a situation, then, it does not seem particularly useful to debate the selective forces responsible for the retention of primary hermaphroditism or the evolution of monoecism unless the dynamics of the pollination biology are taken into consideration. We conclude with the following: 1 . The sexual systems of species are fundamentally linked to the pollination biology of the plants and in many instances can only be understood within the context of the pollination system. We suggest, therefore, that further theoretical considerations as to the adaptive nature of sexual systems must consider in more detail the reproductive ecology of the plants. 2. The taxonomy of sexual systems is largely determined by the type of gender estimates taken at the level of the individual. Intrinsic gender estimates and those based on morphological appearance are not as valuable as estimates of functional or effective gender for determining how floral sexuality actually functions and similarly for illuminating variation in effective gender between conspecific plants. The documentation of this variation is of great utility for understanding the se- lective forces and evolutionary pathways of sexual system evolution. 3. The temporal dimension of plant sexuality is greatly underestimated by current classification schemes which are largely based on spatial features of plant gender. 4. The morphological distinction between flowers, inflorescences, and even larger groups of flowers such as the totality of flowers in a tree canopy is main- tained for most general purposes, but it must also be realized that these units of attraction and/or function might be irrelevant as far as the pollinators are con- cerned. Consequently, our sexual-system classification is to an extent arbitrary, as class limits are defined on the basis of morphological features and not on the basis of actual function. 5. Finally, we conclude that viewing plant sexual systems with vague reference to the regulation of genetic recombination is unlikely to account fully for the evolution of sexual systems and that the key to understanding them lies in con- sidering patterns of sexuality as means of optimizing male and female reproduc- tive success in different ways within the constraints imposed by the pollination system. Literature Cited Allard, R. W. 1965. Genetic systems associated with colonizing ability in predominantly self- pollinating species. Pp. 49-76, in H. G. Baker & G. L. Stebbins (editors), Genetics of Colonizing Species. Academic Press, New York. Anderson. G. J. 1980. Dioecious Solanum species of hermaphroditic origin is an example of a broad convergence. Nature 282: 836-838. 272 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Baker, H. G. 1958. Studies in the reproductive biology of West African Rubiaceae. J. W. African Sci. Assoc. 4: 9-24. . 1959. Reproductive methods as factors in speciation in flowering plants. Cold Spring Harbor Symp. Quant. Biol. 24: 177-191. — . 1967. Support for Baker's law — as a rule. Evolution 21: 853-856. — . 1976. ''Mistake'' pollination as a reproductive system with special reference to the Carica- ceae. Pp. 161-169, in J. Hurley & B. T. Styles (editors). Tropical Trees: Variation, Breeding and Conservation. Academic Press, New York. — & P. H. HuRD, Jr. 1968. Intrafloral ecology. Annual Rev. Entomol. 13: 385^14. Bateman, a. J. 1948. Intrasexual selection in Drosophila. Heredity 2: 349-369. Bawa, K. S. 1974. Breeding systems of tree species of a lowland tropical community. Evolution 28:85-92. . 1977. The reproductive biology of Cupania guatemalensis Radlk. (Sapindaceae). Evolution 31: 52-63. . 1980a. Evolution of dioecy in flowering plants. Annual Rev. Ecol. Syst. 11: 15-39. . 1980b. Mimicry of male by female flowers and intrasexual competition for pollinators in Jacaratia dolichatda (D. Smith) Woodson (Caricaceae). Evolution 34: 467-474. & P. A. OPLtR. 1975. Dioecism in tropical forest trees. Evolution 29: 167-179. Blach, J. H. 1981. Pollinator foraging and the evolution of dioecy. Amer. Naturalist 118: 572-577. & K. S. Bawa. 1980. Role of pollinators in the evolution of dioecy from distyly. Evolution 34: 1138-1142. — & W. J. Kress. 1980. Sporophyte versus gametophyte: a note on the origin of self-incom- patiblity in flowering plants. Syst. Bot. 5: 1-5. Bell, C. R. & A. H. Lindsey. 1978. The umbel as a reproductive unit in the Apiaceae. Pp. 739- 747, in Actes du 2eme Symposium International Sur les Ombelliferes (Perpignan, 1977). Con- tributions pluridiciplinaires a la systematique. Beriin, R. I. 1981. The ecology of sex expression in red buckeye. Submitted for publication. Burtt, B. L. 1977. Aspects of diversification in the capitulum. Pp. 42-59, in V. H. Heywood, J. B. Harborne & B. L. Turner (editors). The Biology and Chemistry of the Compositae. Academic Press, New York. Charlesworth, B. & D. Charlesworth, 1978. A model for the evolution of dioecy and gyno- dioecy. Amer. Naturalist 112: 957-997. Charlesworth, D. & B. Cilarlesworth. 1979. The evolutionary genetics of sexual systems in flowering plants, Proc, Roy. Soc. London, Ser. B, Biol. Sci. 205: 513-530. Charnov, E. L. 1979. Simultaneous hermaphroditism and sexual selection. Proc. Natl. Acad. U.S.A. 76: 2480-2484. , J. Maynard Smith & J. J. Bull. 1976. Why be an hermaphrodite? Nature 263: 125-126. Connor, H. E. 1979. Breeding systems in the grasses: a survey. New Zealand J. Bot. 17: 547-574. Crepet, W. L. 1979. Insect pollination: a paleontological perspective. BioScience 29: 102-108. Cruden, R. W. & S. M. Herman-Parker. 1977. Temporal dioecism: an alternative to dioecism? Evolution 31: 863-866. Darlington, C. D. 1958. The Evolution of Genetic Systems. Oliver and Boyd, London. Darwin, C. 1876. The Effects of Cross and Self Fertilisation in the Vegetable Kingdom. John Murray, London. . 1877. The Different Forms of Flowers on Plants of the Same Species. John Murray, London. Faegri, K. & L. van der Pijl. 1971. The Principles of Pollination Ecology. Pergamon Press, New York. Fisher, R. A. 1958. The genetical Theory of Natural Selection. Ed. 2. Dover Publications, New York. Frankel, R. & E. Galun. 1977. Pollination Mechanisms, Reproduction and Plant Breeding. Spring- er-Verlag, New York. Frankie, G. W., W. Habf r, P. A. Opi er & K. S. Bawa. 1981. Characteristics and organization of the large bee pollination system in the Costa Rican dry forest. /// C. E. Jones & R. J. Little (editors), Handbook of Pollination Biology. Van Nostrand Reinhold Co., New York (in press). Freeman, D. C, K. T. Harper & W. Kent Ostler. 1980. Ecology of plant dioecy in the inter- mountain region of western North America and California. Oecologia 44: 410^17. Ganders, F. R. 1979. The biology of heterostyly. New Zealand J. Bot. 17: 607-635. GiVNiSH, T. J. 1980. Ecological constraints on the evolution of breeding systems in seed plants: dioecy and dispersal in the Gymnosperms. Evolution 34: 959-972. GoDLEY, E. J. 1955. Monoecy and incompatibility. Nature 176: 1176-1177. Grant, V. 1951. The fertilization of flowers. Sci. Amer. 184: 52-56. 1981] BAWA & BEACH— EVOLUTION OF SEXUAL SYSTEMS 273 . 1958. The regulation of recombination in plants. Cold Spring Harbor Symp. Quant. Biol. 23: 337-363. Haghruh, O. 1951. Pollination in the Faroes — in spite of rain and the poverty of insects. Biol. Meddel. Kongel. Danske Vidensk. Selsk. 18(15): 1-48. Heath, D. J. 1977. Simultaneous hermaphroditism; cost and benefit. J. Theor. Biol. 64: 363-373. Heithaus, E. R., p. a. Opler & H. G. Baker. 1974. Bat activity and the pollination of Bauhinia pauletia: plant-pollinator coevolution. Ecology 55: 412-419. Heslop-Harrison, J. 1972. Sexuality of angiosperms. Pp. 133-289, in F. C. Steward (editor). Plant Physiology: A Treatise. Vol. 6 C. Academic Press, New York. HoRoviTZ, A. & J. Harding. 1972. The concept of male outcrossing in hermaphrodite higher plants. Heredity 29: 223-236. Jain, S. K. 1961 . On the possible adaptive significance of male sterility in predominantly inbreeding populations. Genetics 46: 1237-1240. . 1976. The evolution of inbreeding in plants. Annual Rev. Ecol. Syst. 7: 469^95. Janzen, D. H. 1977. A note on optimal mate selection by plants. Amer. Naturalist HI: 365-371. Jones, D. F. 1928. Selective Fertilization. Univ. Chicago Press, Chicago. Krohne, D. T., L Baker & H. G. Baker. 1980. The maintenance of the gynodioecious breeding system in Plantago hmceolata L. Amer. Midi. Naturalist 103: 269-279. Levin, D. A. 1972. Competition for pollinator service: a stimulus for the evolution of autogamy. Evolution 26: 668-669. & H. W. Kerster. 1974. Gene flow in seed plants. Evol. Biol. 7: 139-220. Lewis, D. 1942. The evolution of sex in flowering plants. Biol. Rev. Cambridge Philos. Soc. 17:46- 67. Lewis, K. R. & B. John. 1963. Chromosome Marker. Churchill, London. Linnaeus, C. 1737. Genera Plantarum. Ed. 1. [Ed. 5 (1754) facsimile seen (1960)J. Engleman (Cra- mer), Weinheim. Lloyd, D. G. 1972a. Breeding systems in Cotula L. (Compositae, Anthemideae) 1. The array of monoclinous and diclinous systems. New Phytol. 71: 1181-1194. . 1972b. Breeding systems in Cotula L. (Compositae, Anthemideae) IL Monoecious popula- tions. New Phytol. 71: 1195-1202. — . 1975a. The maintenance of gynodioecy and androdioecy in angiosperms. Genetica 45: 325- 339. — . 1975b. Breeding systems in Cotula L. IV. Reversion from dioecy to monoecy. New Phytol. 74: 125-145. — . 1979a. Parental strategies in angiosperms. New Zealand J. Bot. 17: 595-606. — . 1979b. The evolution of heterostylous populations towards dioecy. PI. Syst. Evol. 131: 71- 80. — . 1980a. Sexual strategies in plants IH. A quantitative method for describing the gender of plants. New Zealand J. Bot. 18: 103-108. — . 1980b. Sexual strategies in plants. I. An hypothesis of serial adjustment of maternal invest- ment during one reproductive season. New Phytol. 86: 69-79. — . 1981. The selection of combined versus separate sexes in seed plants. Amer. Naturalist, in press. & J. M. A. Yates. 1981. Inlrasexual selection and the segregation of pollen and stigmas in Wahlenhergia albomarginata (Campanulaceae) and other hermaphrodite plants. Manuscript sub- mitted to Evolution. LovETT-DousT, J. 1980. Floral sex ratios in andromonoecious UmbcUiferae. New Phytol. 85: 265- 273. Mather, K. 1940. Outbreeding and separation of the sexes. Nature 145: 484-486. . 1973. Genetical Structure of Populations. Chapman and Hall, London. Maynard Smith, J. 1978. The Evolution of Sex. Cambridge Univ. Press, Cambridge. MuLCAHv, D. L. 1979. The rise of the angiosperms: a genecological factor. Science 206: 20-23. MiJLLER, H. 1883. The Fertilization of Flowers. Macmillan, London. Nettancourt, D. de. 1977. Incompatibility in Angiosperms. Springer-Verlag, New York. Onyekwelu, S. S. & J. L. Harper. 1979. Sex ratio and niche differentiation in spinach {Spinacia oleracea L.). Nature 282: 609-611. Primack, R. B. & D. G. Lloyd. 1980. Andromonoecy in the New Zealand montane shrub Manuka, Leptospermum scoparium (Myrtaceae). Amer. J. Bot. 67: 361-368. Proctor, M. & P. Yeo. 1972. The Pollination of Flowers. Taplinger Publ. Co., New York. Pyke, G. H. 1978. Optimal foraging in bumblebees and coevolution with their plants. Oecologia 36: 281-293. 274 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Ramirez, B. W. 1969. Fig wasps: mechanism of pollen tranfer. Science 163: 580-581. Ross, M. D. 1970. Evolution of dioecy from gynodioecy. Evolution 24: 827-828. . 1980. The evolution and decay of overdominance during the evolution of gynodioecy, sub- dioecy, and dioecy. Amer. Naturalist 116: 607-620. ScHEMSKE, D. W. 1980. Evolution of floral display in the orchid Brassavola nodosa. Evolution 34: 489^93. , M. F. WiLLSON, M. N. Melampy, L. J. Miller, L. Verner, K. M. Schemske & L. B. Best. 1978. Flowering ecology of some spring woodland herbs. Ecology 59: 351-366. SoLBRiG, O. T. 1976. On the relative advantages of cross- and self-pollination. Ann. Missouri Bot. Card. 63: 262-276. Stebbins, G. L. 1950. Variation and Evolution in Plants. Columbia Univ. Press, New York. . 1951. Natural selection and the differentiation of angiosperm families. Evolution 5: 299-324. . 1958. Longevity, habitat and the release of genetic variability in higher plants. Cold Spring Harbor Symp. Quant. Biol. 23: 365-378. Stern, K. & L. Roche. 1974. Genetics of Forest Ecosystems. Springer-Verlag, New York. Stout, A. B. 1924. The flower mechanisms of avocados with reference to pollination and production of fruit. J. New York Bot. Card. 25: 1-7. Symon, D. E. 1979. Sex forms in Solanum (Solanaceae) and the role of pollen collecting insects. Pp. 385-397, /// J. G. Hawkes, R. N. Lester & A. D. Skelding (editors), The Biology and Taxonomy of the Solanaceae. Academic Press, New York. Webb, C. J. 1980. The status of New Zealand Actinotus (UmbeUiferae). New Zealand J. Bot. 18: 343-345. . 1981. Andromonoecism, protandry, and sexual selection in UmbeUiferae. New Zealand J. Bot. (in press). & D. G. Lloyd. 1980. Sex ratios in New Zealand apioid UmbeUiferae. New Zealand J. Bot. 18: 121-126. WiLLSON, M. F. 1979. Sexual selection in plants. Amer. Naturalist 113: 777-790. & B. J. Rathcke. 1974. Adaptive design of the floral display in Asclcpias syriaca L. Amer. Midi. Naturalist 92: 47-57. — & P. W. Price. 1977. The evolution of inflorescence size in Asclcpias (Asclepiadaceae). Evolution 31: 495-511. Yampolsky, C. & H. Yampolskv. 1922. Distribution of sex forms in the phanerogamic flora. Bibliogr. Genet. 3: 1-62. Zapata, T. R. & M. T. K. Arroyo. 1978. Plant reproductive ecology of a secondary deciduous tropical forest in Venezuela. Biotropica 10: 221-230. ON THE EVOLUTION OF COMPLEX LIFE CYCLES IN PLANTS: A REVIEW AND AN ECOLOGICAL PERSPECTIVE 1 Mary F. Willson^ Abstract Complex life cycles and alternation of generations are characteristic of many plants, a diploid sporophyte typically alternating with a haploid gametophyte. The prominence of each generation varies greatly among taxa. Purely phylogenetic or morphogenetic explanations of these differences are unsatisfying, as are those based solely on population fitness. Existing adaptational explanations seek selective advantages in diploidy and in sexual reproduction, but these explanations leave much to be explained — i.e., the existence of asexuality and of haploid organisms. Much of the existing variation in life cycles can be explained by selection on reproductive rates to meet the ecological problems of dispersal, colonization, niche preemption and exploitation, and mate competition. Place- ment of complex life cycles of plants in an ecological framework will, I hope, encourage specific studies exploring possible adaptive aspects and limitations on the evolution of life cycles in plants with different life histories. A prominent feature of most textbook introductions to various major groups of plants is the description of reputedly typical life cycles of selected forms. Attention is focused on the alternation of haploid and diploid phases (or gener- ations) of the cycle, the relative ''dominance" of one phase or another, the timing of meiosis and fertilization, the occurrence of asexual reproduction, and which phase is the dispersal or resting stage. Most of the available published literature on plant life cycles is directed toward elucidating developmental, morphological or physiological and, finally, phylogenetic patterns. A rather small number of papers (see below) addresses the possible selective advantages or disadvantages of reproducing sexually or asexually, of being haploid or diploid, or of having more than one morphological generation. Here, I use the word generation to refer to both the alternating diploid and haploid entities and to any intervening, asexually produced, progenies. So far as I can determine, virtually nothing has been written on the ecology (aside from genetics) of complex life cycles in plants. Several botanical reviewers have claimed that all of the ideas summarized here regarding the ecology and evolution of plant life cycles are "old hat/' They should be, perhaps, but no one has shown me evidence that they are. This review is presented in hopes of stimulating discussion and research that include an eco- logical perspective. I begin by summarizing some apparently classic life cycles of selected plants (broadly defined to include fungi, but excluding prokaryotes). There follows a review of available hypotheses attempting to explain the evolution of such cycles in plants and a critique of these hypotheses. Finally, I review some of the liter- * 1 am most grateful to the following friends and colleagues who provided references, thoughtful criticism, and encouragement: C. K. Augspurger, B. L. Benner, R. I. Bertin, N. Burley, S. E. Franson, A. Hartgerinck. L. Hoffman, T. D. Lee, M. Lynch, B. McPheron, H. Michaels, P, W. Price, P. J. Regal, D. W. Schemske, and J. N. Thompson. G. C. Williams graciously contributed his perception and insight toward improvement of the penultimate draft. ^ Department of Ecology. Ethology, and Evolution, University of Illinois, Vivarium, 606 E. Healey Street, Champaign, Illinois 61820. Ann. Missouri Bot. Card. 68: 275-300, 1981. 0026-6493/8 1/0275-0300/$02.65/0 276 ANNALS OF THE MISSOURI BOTANICAL GARDKN [Vol. 68 ature on complex life cycles in animals in hopes of unearthing some possible approaches useful in interpreting plant life cycles and close by posing some eco- logical suggestions about plant life cycles that, I think, provide a useful beginning in understanding the evolution of these cycles. Life Cycle Patterns in Plants Plant life cycles are many and varied and there are numerous and divers ways of classifying them (e.g.. Chapman & Chapman, 1961; Lindenmayer, 1964; Scott & Tngold, 1955). Although in most major taxa some species have become entirely asexual, it is generally considered that all plants exhibit a fundamental alternation of generations. The basic pattern is a 2N sporophyte (producing spores) alter- nating with a IN gametophyte (producing gametes). Relative dominance of one generation over another can be considered in terms of conspicuousness of "body size" and/or in terms of duration. Because an in- conspicuous dormant zygote may persist longer than its ecologically active par- ent(s), T prefer to use body size as an index of dominance. However, if asexually produced progeny are viewed as part of an extended "evolutionary individual" (Janzen, 1977, 1979a), body size and duration may converge in meaning. [Janzen was challenged by Addicott (1979) specifically for aphids (which are obviously not plants, but the idea is relevant!), in which parthenogenesis does not neces- sitate invariant offspring identical to the parent. While the point is made that genetic identity of asexual parents and offspring cannot be assumed, this does not alter the general idea that asexually produced young are more similar than those produced sexually and the clone is still the unit of selection (Janzen, 1 979b).] Some multicellular organisms have potential for large increases in body size (and may also expand by asexual means); other multicellular organisms and unicellular ones may "substitute" asexual multiplication for an increase in body size and prolong the duration of one or both generations by this means. I am concerned in this paper mainly with plants in which both sporophyte and gametophyte are active, functioning (i.e., not resting or dormant) organisms at appropriate times of the life cycle. Therefore, algae (such as Fucus and most diatoms) that exist primarily as diploids, with meiosis during gametogenesis (as is the case in most animals, except certain Foraminifera that alternate haploid/ diploid generations (Grell, 1967; Ghiselin, 1974), are of interest mostly as con- trasts. Slime molds, as well as some yeasts and other fungi (Raper, 1966b), exist mainly as diploids. Meiosis produces IN spores that germinate into swimming cells; these fuse to form a zygote that eventually grows (sometimes several to- gether) into a new organism. The seed plants are diploid-dominant also; the IN gametophytes are tiny and wholly dependent on the sporophyte. In an ecological sense, although not morphologically or genetically, the seed plants may be con- sidered to lack alternation of generations. On the other hand, certain plants exist primarily as haploids, with meiosis upon germination of the 2N zygote. Many freshwater green algae, typified by Chlamydomonas, have a 2N zygote, often dormant, that meiotically forms veg- etative IN cells, which divide mitotically to form either gametes or more vege- tative cells. Among the fungi, most phycomycetes are basically haploid (Burnett, 1976). Two hyphae fuse to form a2N zygospore, which is a multinucleate resistant 1981] WILLSON— PLANT LIFE CYCLES 277 phase. This undergoes meiosis to form spores that grow into the macroscopically visible mold. Between these two extremes are many plants that function actively in two (usually) phases. The gametophyte is more conspicuous in some, the sporophyte in others, and in a few the generations are similar in conspicuousness. Most bryophytes are conspicuous as persistent, IN gametophytes. These form gametes by mitosis, the sperm swims to the egg, which lies in a protective jacket of cells, and the resultant zygote grows into a 2N sporophyte, still attached to the (female) parent gametophyte. The sporophyte, generally less conspicuous than the gametophyte, appears seasonally and produces IN spores meiotically; these disperse and grow into new gametophytes. The gametophyte is considered to be the independent generation and is longer lived. Vegetative propagation, especially by gametophytes, is apparently common. The 2N sporophyte of the lower vascular plants is more apparent than the tiny, often subterranean, gametophytes. The sporophyte meiotically produces spores that grow into gametophytes that produce gametes by mitosis. Although in some fern populations (Farrar, 1967) no sporophyte is known to exist, the fern sporophyte typically develops on a small IN gametophyte, which may eventually disappear. Spores have half the chromosome complement of the sporophyte, but high levels of polyploidy permit, in some species, recombination between dupli- cated, unlinked loci, which produces varied gametes on a gametophyte from a single spore (Chapman et al., 1979; Lloyd, 1974b). A third pattern of alternation, occurring in several green, brown, and red algae, involves free-living, multicellular haploid and diploid phases. The phases may be similar in external morphology (isomorphic, as in Ulva) or dissimilar (heteromorphic, as in '"Derbcsicf' and others) (Bold & Wynne, 1978; Fritsch, 1942; Wynne, 1969; Kung-Chu, 1959; Dube, 1967). Each phase may reproduce asexually, forming more individuals of the same phase. The isomorphic forms may grow side by side on the same rocks, thus apparently occupying the same spatial environment; this seems to be the case for Ulva. Ectocarpus (a brown alga) is intriguing in that it is reportedly isomorphic in warm seas but appears to have a reduced haploid phase in cold waters (Bell & Woodcock, 1968). Prasiola stipitata forms spores in the upper intertidal zone and gametes in the lower zone; curiously, the spores are diploid and the lower portion of the ''gametophyte" is also while its upper portion is haploid (Friedmann, 1959). Red algae have varied and complex life histories that are not well understood (P. S. Dixon, 1973). Some consist of three generations: a haploid phase and two diploid phases, one of which is small and attached to the haploid, followed by a free-living 2N phase that is isomorphic or heteromorphic with the haploid, de- pending on the species. In others the haploid phase grows on the free diploid, and in still others the free-living diploid seems to be lost altogether, meiotic products from the attached diploids growing directly into a new haploid. Many more variants are likely to occur (e.g., West & Norris, 1966). The aquatic mold AUomyccs, unusual among the phycomycetes (Darlington, 1958), has isomorphic generations, the haploid phase or gametophyte producing male and female gametes by mitosis on the same filament. These merge to form a 2N phase that can resist desiccation and may produce more 2N individuals by mitosis or IN individuals by meiosis. 278 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Ascomycetes and basidiomycetes are unusual; they can exist in two different conditions not defined by chromosome number. The 2N zygote is borne on its parent and is highly ephemeral, immediately forming IN spores by meiosis. The spores disperse and eventually grow into a new IN generation of hyphae. These fungi resemble most phycomycetes in that the diploid phase is much reduced. However, two hyphae (of different mating strains) can join together to form binucleate cells; this dikaryotic condition may persist for some time. Nuclear status is technically haploid although the organism possesses a double comple- ment of chromosomes. Thus these fungi exist primarily in a IN or in a IN + IN phase. In summary: some plants produce gametes meiotically; the IN generation is invariably much less conspicuous than the 2N generation. Other plants produce gametes mitotically. Tn some of these the 2N generation is inconspicuous. But in others the diploid generation has a period of growth and becomes "apparent" (sensu Feeny, 1976). The apparency of two generations presumably has an adap- tive basis, but just how a functional alternation of generations could be adaptive seems to be largely unexplored. Existing adaptational hypotheses applied to plant life cycles address primarily the evolution of diploidy and the evolution of sex. Furthermore, several nonadaptational hypotheses can be found, although Allsop (1966) suggested that "in general terms, the entire life cycle may represent an evolutionary adaptation . . . ," and Bonner (1965) noted that both phases of the life cycle are undoubtedly adaptive. EXISTING HYPOTHESES SPECIFIC TO PLANT LIFE CYCLES (A) PhYI OGENETIC AND MORPHOGENETIC EXPLANATIONS A general trend in the plant kingdom from dominance of the gametophyte to dominance of the sporophyte is often noted (e.g., Chamberlain, 1935; Fuller, 1955; Bonner, 1965). Even taking into account the branching, nonlinear pathways of plant phylogeny, such trends can only be descriptive. The phylogenetic ideas regarding alternation of generations reviewed by Wahl (1965) and Roe (1975) are largely of historic interest. In any case, explanations for the phylogenetic preem- inence of, for instance, the angiosperms, are legion (Mulcahy, 1979; Cavalier- Smith, 1978). Morphogenetic "explanations" invoke genetic and developmental events as causes of morphological conditions (e.g.. Bell, 1979). It is sometimes suggested that the function of sexuality in algae is the formation of resistant stages. Although sex may indeed precede the resistant phase and in some species the two events may have become tightly linked, it is absurd to suppose that sexuality evolved so that resistant phases could be formed. Asexual modes of reproduction could equally lead to a resistant, dormant stage and do in some organisms (Blackman, 1974; Drebes, 1977). Haploid and diploid conditions are associated with game- tophyte and sporophyte conditions, respectively, and this relationship is some- times taken to be causal (i.e., ploidy controls phenotype), rather than descriptive and associative, according to Bold & Wynne (1978). Nevertheless, haploid spo- rophytes are known in some algae {Viva, Cladophora, Laminariales) and both haploid and diploid gametophytes occur in Ectocarpus. Furthermore, bryophyte 1981] WILLSON— PLANT LIFE CYCLES 279 and fern gametophytes and sporophytes can apparently exist in either ploidy level (Watson, 1971; Bold, 1967; Bell & Woodcock, 1968) and bryophyte sporophytes can be induced by wounding the gametophyte (Crum, 1973). However, the re- productive potential of such individuals is not indicated (but see Hoxmark, 1975, on Ulva); if they cannot reproduce at a rate competitive with their normal con- freres, they must be regarded as ecological "sports." However true such mech- anistic links might be, they ultimately cannot provide an evolutionary, adaptational, explanation for life-cycle variations. (B) Explanations dependent on population fitness "The evolutionary significance of diploidy to higher organisms resides, there- fore, in the greater flexibility which it confers on their populations" (Stebbins, 1960: 213; and see Bonner, 1965). Of necessity, the "selective advantage of evo- lutionary flexibility" in haploid organisms must then be less (Stebbins, I960). In this sense, evolutionary flexibility is a population characteristic, not an individual one, and thus an indirect consequence of normal Darwinian selection. Such flex- ibility may indeed exist, however, but may be best interpreted as a population consequence of individual selection as discussed in (C) below. Similarly, the flushing of the deleterious genes from a population is a conse- quence of the expression of a haploid generation. From the point of view of an individual, the loss of some proportion of its genes may seem unnecessarily det- rimental, since many of them are deleterious only in the haploid state; if zygote formation followed meiosis with little delay, those "deleterious" genes might even have beneficial effects in the diploid. Furthermore, only genes that are "turned on" in a haploid phase could be affected (Bonner, 1965). A model for the evolution of sexual reproduction as a repair mechanism (Walker, 1978) possesses some attractive features. However, it depends explicitly on population, not individual, fitness (Williams & Walker, 1978) and is not con- sidered further here (but see Dougherty, 1955; Bernstein et al., 1981). (C) Adaptational explanations (1) Diploidy as an adaptation. — Generally, diploidy is thought to buffer an individual, in some circumstances at least, from the effects of deleterious mutants and to offer the possibility of heterosis (e.g.. Crow & Kimura, 1965; Stebbins, I960; Raper, 1966a; Raper & Flexer, 1970; refs. in Levin & Funderburg, 1979; Rehfeldt & Lester, 1969). Sexual reproduction by diploids also releases genetic variation immediately; this is not true for haploids, which typically must first make a diploid entity as a means of releasing variation (Ghiselin, 1974). Adams & Hansche (1974) consider that these may be secondary effects that arose through time rather than initial advantages of diploidy. In addition, alternative alleles may switch on or off to deal with a fluctuating environment. Diploidy should therefore prevail in particularly variable, unpredictable, or difficult circumstances. These possibilities often seem to provide a standard "explanation" for the evolution of diploid dominance. They necessitate the assumption that all other patterns must occur in relatively benign environments in which the putalively delicate haploid is not at risk. Lewis & Wolpert (1979) suggest that diploids have spare copies of 280 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 all genes, which can be modified for new functions and thus easily expand the genome at a much faster rate than haploids, where an original gene must be duplicated before it can be modified. Diploids can therefore more easily evolve the complex gene control mechanisms that characterize complex multicellular organisms. (Lewis and Wolpert further explain sex as an adaptation that preserves diploidy.) Although diploidy may be a preadaptation that permits (through buff- ering, etc.) the evolution of complex developmental sequences and complicated organisms, this obviously cannot explain the origin of diploidy (Stebbins, 1960; Cavalier-Smith, 1978). Cavalier-Smith (1978) made the fascinating suggestion that diploidy in eu- karyote life cycles may have evolved when rapid growth is less important or larger cell size and lower growth rates are actually advantageous (see Willson & Burley, in press). The volume of 2N nuclei and cells is commonly twice that of haploids and rates of cell division and development are strongly inversely cor- related with DNA content (though other factors can be involved also, Bennett, 1974; Price & Bachmann, 1976; Price et al., 1973). Bennett (1971, 1972, 1974) argued explicitly that short generation times require low nuclear DNA content and that high DNA content necessitates a longer generation time. This idea is likely to be involved with haploid/diploid life cycles (Cavalier-Smith, 1978). Fur- thermore, IN cells have higher surface/volume ratios, and enzymes active at the cell surfaces may also contribute to faster haploid growth (Weiss et al., 1975; Adams & Hansche, 1974). Such differences in growth and development are only sometimes apparent when conspecific cells of different ploidy (:^2N) are com- pared (e.g., D'Amato, 1977; Willson & Burley, in press). However, because there is often variation among conspecific individuals in the amount of DNA, we should not yet reject the notion that alternating generations differing in ploidy may have different growth potential. Evans's (1968) interpretation of Durrant's (1962) ex- periment on the induction by fertilizer treatments of heritable changes in Linum is tantalizing: plants with intermediate amounts of nuclear DNA have very plastic growth responses; when grown in conditions inducing large size, the offspring have increased DNA content. Jones (1975) cited evidence from a grass {Lolium) that the frequency of accessory or B chromosomes increased significantly with increasing plant density and competition. Intraspecific, clinal variation in DNA content may occur also in Picea sitchcnsis (Burley, 1965; Mitsche, 1971; Moir & Fox, 1977; Mergen & Thielges, 1967; Sparrow et al., 1972; but see Dhillon et al., 1978) and some other conifers (Mergen & Thielges, 1967; Mitsche, 1968). Geographic variation in frequencies of B chromosomes is not uncommon (e.g., Sparrow et al., 1952; Bosemark et al., 1956a, 1956b; Frost, 1957; Muntzing, 1957; Teoh & Jones, 1978). It would be interesting to learn if this kind of variability occurs in species with alternating generations. Although B chromosomes apparently carry no major genes themselves, they can affect gene activity and rates of crossing-over on regular chromosomes (Rees, 1972; Jones, 1975, 1976). Altogether, as little as 10% of the eukaryotic genome may exist as unique-sequence DNA that codes for RNA and functional polypep- tides (Jones, 1976); much DNA exists as repetitive sequences (more than 10*' repeated sequences in some cases (Lewin, 1974) and up to 95% of total nuclear DNA content (Flavell et al., 1974) in B chromosomes or other conditions, some 1981] WILLSON— PLANT LIFE CYCLES 281 f r of which are genetically inactive and even non-Mendelian. The preferential trans- mission of B chromosomes to functional gamete nuclei in both male and female of several angiosperms (Stebbins, 1971) deserves special attention. Although Stebbins (1966) interpreted certain changes in DNA content in terms of possible genetic consequences (e.g., sequential gene action on long chromosomes, reduc- tion of chromosome number to reduce genetic variability, and stabilizing new genotypes in polyploids), it seems reasonable to consider nongenetic conse- quences of DNA content as well (see also Hinegardner, 1976; Bennett, 1971, 1972). (2) Sex as an adaptation to environmental uncertainty. — Having a sexual generation may be adaptive to changing conditions (Ghiselin, 1974; Williams, 1975; Emlen, 1973). These authors argue that sex is adaptive when production of a varied progeny enhances reproductive success of the parent. Thus, sexual re- production should occur where or when the environment is unpredictable. Be- cause sex involves recombination, it often results in a more varied array of off- spring, which increases the likelihood that some will be able to successfully cope with environmental changes. Without disputing the possible advantages of sex, which have often been debated, I would like to provisionally accept the notion that sex is adaptive in physically and especially biologically varying environments (Williams, 1975; Levin, 1975; Ghiselin, 1974; Treisman, 1976; Glesener & Tilman, 1978; Warner, 1978; Calow, 1978b; Hamilton, 1980; and others). The timing of sex is also commonly interpreted in terms of environmental change, imminent or just past. Considering spores to represent sex and recom- bination, Bonner (1958) says that ''the spores, when they emerge, have previously undergone recombination (or do so upon germination)." Bonner, Williams, and Ghiselin all provide examples of asexual generations repeating through benign seasons and a sexual generation intervening when the environment is soon to change or has just done so. CRITIQUE OF EXISTING ADAPTATIONAL HYPOTHESES First, the/supposed delicacy of IN entities (relative to those of 2N) is some- what debatable. In bryophytes and green algae the haploid gametophyte is the persistent form; in some algae, IN and 2N generations have similar gross mor- phology and may live side by side. Furthermore, haploid cells in vitro are some- times less sensitive to stresses (putative mutagens, fungicides) than expected (Metzger-Freed, 1974; Henriques et al., 1977; Upshall et al., 1977). And a growing literature on diploidization and gene silencing in polyploids (e.g., Ferris & Whitt, 1977a, 1977b; Garcia-Olmeda et al., 1978) and differential chromosome elimina- tion (Collins et al., 1978) suggests that buffering functions may not necessarily be central. Conversely, the arbitrary suppression of one X chromosome, possibly as a means of dosage compensation (Lyon, 1974; Monk & Kathuria, 1977; Luc- chesi, 1978; Epstein et al., 1978), in female mammals suggests that partial hap- loidy may not be deleterious. Certain parasitic protozoans reduce their chromo- some number by means of zygotic meiosis and restore the 2N quantity of DNA just in time for mitotic gametogenesis (Canning & Morgan, 1975). And haploid males occur regularly in certain groups of rotifers, arachnids, and insects (White, 282 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 1973). In fact, the tendency of many organisms, both plant and animal to protect the (2N) zygote in some way could be interpreted to signal its great vulnerability to assorted environmental hazards (just as resistance of the spore is often inter- preted). Even in diploid organisms there is variation in amounts of DNA present (Price. 1976; Cavalier-Smith, 1978), even among conspecific individuals (Jones, 1975). Second, there is increasing evidence (from insects, Rasch et al., 1977; bryo- phytes, Longton, 1976; ferns, Klekowski & Baker, 1966; Ulva, Hoxmark & Nord- by, 1974) that DNA synthesis in originally haploid individuals restores the 2N quantity of DNA (but, of course, not heterozygosity and perhaps not reproductive performance; Hoxmark, 1975). Third, fitness of isogenic haploid and diploid forms of Saccharomyccs grown in competition with each other were similar when dosages of enzymes active in the cell interior differed with ploidy. But IN cells out-competed 2N cells when dosages of surface-active enzymes differed. These experiments suggest no ad- vantage to diploidy per se under these experimental conditions (Adams & Hansche, 1974; Weiss et al., 1975). Cavalier-Smith (1978) argues that haploidy may indeed by favored where rapid growth and development are advantageous. Taken together, such evidence suggests that the perils of haploidy may be minor or, at least, could be avoided in many ways. (It may still be true, of course, that a switch from diploidy to haploidy could be difficult once diploidy is entrenched in a group, because many biological features may then come to depend on the diploid conditions and would be disrupted.) The timing of reproduction must depend on divers factors differing among populations. The production of variability actually occurs twice in a life cycle characterized by alternation of generations, once when meiosis occurs and again when gametes join to form a zygote (see below). For a sexual generation, mate- finding may be a critical problem to overcome, and physiological changes and special mating structures may entail costs in energy or nutrients (Daly, 1978; Calow, 1978b; Muenchow, 1978; Solbrig, 1979). Biological factors including pop- ulation density and sex ratio are also potentially relevant. A constellation of factors related to the probability of juvenile survival and/or to costs and risks encountered by reproducing adults must be considered with respect to the timing of any phase of reproduction: the availability of suitable (biotic or abiotic) dis- persal agents and of "safe sites" (Harper, 1977) for establishment, the risks of "prcdation" (or herbivory) on young plants of the next generation, the intensity of competition from ecologically similar species (e.g., Schaffer, 1977), the time required and the resources available for maturation (Boyce, 1979), and so on. Predation risks may be particularly relevant to uni- (or few-) cellular phases subject to capture by size-selective predators. Fungal infection of sori may select against a prolonged season of spore production in Chondms crlspus (Prince & Kingsbury, 1973). Therefore, aside from a broad correlation of sex with certain kinds of environmental uncertainty, I conclude that existing explanations for the timing of alternation of generations are probably inadequate. Furthermore, while existing hypotheses aim at explaining the existence of a sexual or diploid portion of the life cycle, they do not suggest why a free-living haploid phase is sometimes retained. Nor have I found many suggestions in the 1981] WILLSON—PLANT LIFE CYCLES 283 botanical literature that help explain the relative body sizes of gametophyte and sporophyte or which phase is the dispersing phase, the resistant phase, or the growing multiplicative stages. To begin answering such questions, I present a selective survey of literature, largely zoological, that bears on the ecology and evolution of complex life cycles. From this I hope to glean suggestions that foster the understanding of complex life cycles in plants. Complex Life Cycles in Animals Animal life cycles can be broadly classified into two categories. Simple life cycles are characterized by direct development from young to adult. Simple cycles are found among most vertebrates (except many amphibians and some fishes), anthozoan coelenterates, some free-living flatworms and some annelids, hemimetabolous insects, and a few other invertebrates scattered in many taxa. This is not to say that the niches of young and old individuals might not differ greatly, but the transition is relatively gradual. By analogy, seed plants could be viewed as possessing a simple life cycle, since the gametophytic phase has no life of its own. In contrast, complex life cycles exhibit two (or more) distinct phases with very different ecologies and very different morphologies and behav- ior. There are two kinds: (1) Every surviving individual passes through each phase. Reproduction usually is performed by one phase of the cycle. Typical examples are the anuran tadpole that metamorphoses into a frog or toad and caterpillars that are transformed into butterflies. Complex life cycles of this sort are characteristic of most major animal phyla except chordates, in which only protochordates, some fishes, and amphibians conform to this pattern. (2) Other animals with complex life cycles, such as aphids, rotifers, and cladocerans, and many parasites, are more similar to the 'iower'' plants: several asexual genera- tions may be followed by one in which sexual reproduction occurs. Each phys- iologically defined individual usually exists in only one phase, although sometimes females may switch from asexual to sexual reproduction and back. In any event, reproduction occurs in both phases rather than in one. Istock's (1967) seminal paper focused on the ecology of complex life cycles, particularly of the first sort, noting that distinct phases of the life cycle are largely independent of each other in terms of morphological and behavioral adaptations but are ecologically dependent in that each furnishes individuals to the other part of the cycle. He argued that changes in adaptation of the different phases of the life cycle are not likely to proceed at equal rates and, as a consequence, one phase will sooner or later be unable to supply the other phase with enough in- dividuals, and the population will gradually spiral to extinction. In short, complex life cycles would seem to be evolutionarily unstable. Istock's dilemma lies in the conflict between ecological rationalization and the palpable fact that complex life cycles are extremely common. Considering just the insects, as many as 85% of the species may have complex life cycles (E. McLeod, pers. comm.), indicating that extensive adaptive radiation may accompany life-cycle complexity — which hardly suggests the brink of extinction. Slade & Wassersug (1975) later showed that instability is not a necessary feature of complex life cycles. Gill's (1978) study of the red-spotted newt {No- 284 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 tophthalmiis viridcscens) indicates that adult reproductive failure is common and a small number of adults can produce enough young to maintain many local populations. This suggests that the ecological dependence of population levels at different phases of the life cycle may be loose and that a good deal of variation in recruitment rates can be tolerated. Furthermore, larval and adult phases can not be evolutionarily independent— the life history pattern is a unit (Strathmann, 1974). Ewing (1977) describes a genetic model for the stable maintenance of cyclic haploid and diploid phases. Indeed, Slade and Wassersug argue that complex life cycles are adaptive, noting (as had Istock) that seasonal (or other frequent) changes in the environment provide temporary changes in resource abundance upon which creatures may capitalize. Fluctuating environments and/or a colonizing or opportunistic life his- tory (e.g., Lewontin, 1965; Hutchinson, 1951) may select for high reproductive capacity of individual parents and possibly also a life-cycle stage with the capacity to exploit ephemeral resources as a means of increasing parental reproductive output. Furthermore, Slade and Wassersug suggest that a second (or third) life- cycle stage may open new means of dispersal— as a means of reaching new sites of ephemerally available resources. Bryant (1969) also argued for the adaptiveness of complex life cycles in insects, especially in spatially heterogeneous environ- ments. Some evidence is available to support the idea that complex life cycles are adaptive. I divide a series of examples into two "models" for descriptive pur- poses and convenience (see also Wilbur, 1980). THE DISPERSAL MODEL In effect, adult salamanders are viewed as a dispersal phase of the life cycle, metamorphosis from larva to adult often occurring in larvae that are less suc- cessful than other individuals in exploiting the aquatic larval habitat (Wilbur & Collins, 1973) or that happen to live in ephemeral ponds (Rose & Armentrout, 1976). Larval phases seem to be prolonged in areas where the terrestrial envi- ronment is unsuitable for adults and when the aquatic environment is free from major predators (Sprules, 1974; Bruce, 1979). Sexual maturity may be reached by individuals with larval morphology. This condition is not necessarily associ- ated with an early achievement of reproductive status and, in fact, may be as- sociated with delayed reproduction (Gould, 1977). Delay can have far-reaching consequences because the age of first reproduction may be a critical factor in determining the rate at which a genotype contributes genes to future generations (Cole, 1954; Lewontin, 1965). Goin et al. (1968) note that salamanders inhabiting permanent waters generally have higher DNA content (and slower growth) than those adapted to exploit temporary pools, a suggestion in line with that of Cav- alier-Smith (1978). The red-spotted newt is unusual in having a tripartite life cycle: the aquatic larva transforms into a terrestrial "eft," which eventually returns to water as an adult. Efts are lacking in populations occupying ponds that are suitable year- round. Efts grow and mature more slowly than fully aquatic individuals (Healy, 1973), so the advantage of having an eft stage must outweigh this potential dis- 1981] WILLSON— PLANT LIFE CYCLES 285 advantage. Healy (1975) suggests it may reduce the chance of capture by certain pond predators. Perhaps even more importantly, the eft may be a specific ad- aptation for the colonization of ponds, particularly beaver ponds, that are geo- logically temporary — lasting for less time than it takes for a population of newts to saturate it (Gill, 1978). Some salamanders have lost the free-living larval stage and may fall under the next model. Others have lost the adult, dispersing phase, and the 'larva" be- comes reproductive. I do not know of any salamanders in which the larval stage is the main disperser, but larvae are often the principal dispersal phase of many aquatic invertebrates, such as tunicates (Ghiselin, 1974), and a number of animal parasites, such as trematodes. Holometabolous insects clearly fit the dispersal model as well, adults commonly being the dispersal phase. THE RESOURCE-EXPLOITATION MODEL In contrast to salamanders, reproduction by larval anurans is unknown, and only the tadpole stage is ever deleted from the life cycle. Anuran larvae are highly specialized for feeding and rapid growth, and tadpoles are often found in ephem- eral ponds very rich in nutrients (Wassersug, 1974, 1975). The larval stage may be lost by species for which no suitable larval environments are available or if the risks of predation outweigh the advantages of rapid growth (Wilbur, 1980). Similarly, planktonic larval stages of invertebrates are often suppressed if poten- tial larval habitats fluctuate wildly (Calow, 1978a). In this model, one phase of the life cycle is viewed as a means of capitalizing opportunistically on rich re- sources. It may also have been a means of reducing competition between juveniles and adults. In some cases, such as sessile marine invertebrates, asexual multi- plication coloniality may be associated with preemption of space (Jackson, 1977). In some respects rotifers are similar to anurans but with the addition of asexual multiplication of individuals. Most rotifers reproduce rapidly and asexually through the summer (Birky & Gilbert, 1971) when resources are high and, even- tually, a sexual generation (King & Snell, 1977) produces overwintering eggs. Rapid asexual reproduction is considered to be an adaptation of opportunistic species that annually recolonize the ponds and lakes in which they live (Birky & Gilbert, 1971). It is likely that different asexual clones have differing capacities for converting food into offspring, thus exhibiting different rates of increase — with the result that the potential for producing sexual young at the end of the season must differ greatly among genetic lineages (see e.g., Snell, 1979, for ro- tifers; Shick et al., 1979, for sea anemones; Turkington et al., 1979, and Tur- kington & Harper, 1979, for seed plants). A complex life cycle of an aphid is described in some detail in the next three paragraphs, because more ecologically relevant details seem to be available. This case seems to combine elements of both dispersal and resource exploitation (see also Bryant, 1969, on holometabolous insects). Myzus persicae, an aphid, reproduces parthenogentically through the summer, first on Prumis spp. hosts, later and for a longer time on a wide diversity of herbaceous host plants (Newton et al., 1953; van Emden et al., 1969). Repro- duction is asexual, although some recombination occurs and offspring are variable 286 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 (Addicott, 1979), though perhaps less so than if they were produced sexually. At the end of summer when populations are presumably high, a sexual phase occurs and overwintering, sexually produced, eggs are deposited on Prunus hosts. The sexual phase is unusual in that females produce males and winged females. These females return to Prunus hosts and asexually produce several wingless daughters that mate with the males. Winged aphids produce fewer and smaller young than wingless morphs (A. F. G. Dixon, 1973) and mature earlier (Davies & Landis, 1951), The sexual grandmothers thus obtain multiple granddaughters, centrally located on the winter host (Gould, 1977), Changes in the host plant may be advantageous in summer because the nu- trient content of Prunus leaves drops dramatically (A, F. G, Dixon, 1973) and average aphid fecundity, which is very sensitive to soluble nitrogen availability (van Emden et al., 1969), drops to about V22 of what it was (Davies & Landis, 1951; van Emden et al., 1969). This suggests that, like tadpoles, the asexual summer aphids on herbaceous host may be a way of exploiting an ephemeral flush of resources. Aphid mortality is high when dispersing to new hosts, but the ephemerality of resources on any one host necessitates dispersal (Kennedy & Stroyan, 1959). A return to Prunus in the fall may be less related to the senes- cence of herbage (since aphids seem to like senescing tissue) than to an early and reliable availability of Prunus leafbuds in the spring (B. McPheron, pers. comm.), despite the possibly higher predation there (van Emden et al., 1969; A. F. G. Dixon, 1973). Some populations, in regions with no unfavorable season or no Prunus, have entirely asexual reproduction (Broadbent & Heathcote, 1955; Blackman, 1974). Overwintering eggs do not ultimately require sexuality for their production inasmuch as two families closely related to aphids produce resistant eggs asexually (Blackman, 1974). The timing of the sexual phase of the cycle may be related both to high population density at the end of a season of intensive multiplication of individuals and to environmental uncertainty. Reproduction by parthenogenetic aphids is rapid, not as a result of individual fecundity but of phenomenally rapid maturation (Kennedy & Stroyan, 1959; Gould, 1977). The generations are "telescoped," and each grandmother may contain within her body the embryos for the next two generations of aphid. Gould considers the acceleration of maturation and the rapid turnover of generations to be adaptive in reaching and exploiting ephemeral, patchy resources — it is a way of generating large numbers of dispersing young (e.g., Lewontin, 1965; Cole, 1954) to colonize and exploit new patches (see also Ehrendorfer, 1965). Further advantages of asexual reproduction, commonly discussed for plants, may lie in the possibility of establishment of a single propagule (Ehrendorfer, 1965; Allard, 1965; Baker, 1965, 1967; Lloyd, 1974a, 1974b; Holbrook- Walker & Lloyd, 1973; Singh & Roy, 1977) after disersal has been accomplished — thus enhancing the likelihood of success in dispersal. This selective survey of complex life cycles and reproductive life histories of animals suggests several ecological principles of possible relevance to plant life cycles. (I) In some instances, all parthenogenetically produced daughters of a single female can be viewed as extensions of that female, in terms of exploiting the environment. (2) Asexual reproduction (in various forms) and high rates of 1981] WILLSON— PLANT LIFE CYCLES 287 reproduction may be adaptations for colonization and for outcompeting other colonists for available resources. At least one phase of a complex life cycle may be viewed as a means of exploiting ephemeral resources. (3) A dispersal phase is associated with fluctuating environments as a means of escaping to a more reliable patch or as a colonization strategy, enhancing the survival of the off- spring. Whether dispersal or exploitation is more important will depend on the population in question. Clearly, several of these principles could apply to any one kind of organism. These ecological ideas, taken largely from studies of animal life cycles, and others discussed below, are surely germane to plants. In the next section some of these applications are developed. Toward an Ecology of Plant Life Cycles For ease of presentation I have segregated this discussion under headings but do not mean to imply that these pieces are unrelated. Nevertheless, it is possible to argue that various life-history features have gradually become associated; that is, they are not intrinsically linked (Drew, 1955). sexual and asexual generations The environmental uncertainty hypothesis seems to be the best available hy- pothesis for the presence of a sexual generation. What then remains is to explain the persistence of an asexual phase in complex life cycles and the timing of each phase. A series of asexual generations probably can be viewed best as a process of self-multiplication. The successful asexual parent, which is well adapted to its circumstances, produces equally well-adapted young rather than variants, some of which would be adaptively inferior in an unvarying environment. To the extent that some form of recombination may be possible in the asexual phase (as in endomeiosis or automixis in certain insects, White, 1973), or if mutations occur, the progeny may not be completely invariant. (Recombinations among genes of a single individual may be considered sexual processes, but I prefer, in the present context, to label as sexual those reproductive processes that can involve two individuals.) Asexual reproduction (here used to exclude vegetative propagation by rhi- zomes, stolons, etc.) can often occur more rapidly and less expensively than sexual (Calow et al., 1979; Congdon et al., 1978; Walker, 1979; Whittier, 1970), although this may not always be true (Lamb & Willey, 1979). Coulter (1914), Scott & Ingold (1955), and Ghiselin (1974) hint that the life cycle might be timed around the period favorable for "vegetative^" growth. It is likely, even, that the timing of sex is governed more by the advantages of asexual reproduction during seasons favorable for multiplication than by proximity of sexual reproduction to environmental uncertainty— in short, the advantages of asexual reproduction in suitable seasons could delay sexual reproduction until the end of that period. The costs of mating will also influence the timing of sex. There is evidence for certain algal species that increased light or nutrients may induce sexuality directly or indirectly, which hints at resource limitation of sexual activity for 288 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 those species (Coleman, 1962; Sager & Granick, 1954). Added carbon sources facilitate production of apogamous sporophytes (not from a fertilized egg) in Pteridium ferns (Whittier, 1964; Bell, 1979); could similar effects be found for sexually produced sporophytes? Most experiments in induction of sexual activity in plants have been concerned with proximate triggers rather than nutritive con- dition of the plants that might determine receptivity to those stimuli. We need to know why particular stimuli (such as N depletion) are appropriate signals to certain species and how the receptivity varies with condition of the organism. Asexual reproduction is probably adaptive in colonizing and in exploiting ephemeral, patchy resources. Burnett (1976: 550) briefly mentions this possibility of fungi, as does Robinson (1967: 29) for parasitic basidiomycetes. The life his- tories of many freshwater green algae would seem to conform to this pattern. HAPLOID AND DIPLOID GENERATIONS By analogy with animals possessing complex life cycles, it seems likely that gametophytes and sporophytes (as well as asexual and sexual generations) tend to occupy different environments. White (1973: 751) says that alternation of gen- erations presupposes two alternative ecological niches. How the niches for plants might differ seems less evident than for animals, in which habitats and/or re- sources are distinct, but that obviously does not mean that niche differences do not exist. Spatial differentiation of gametophytes and sporophytes is not evident in Ulva, bryophytes, or some ferns, but some seasonal or physiological segre- gation would be possible, as may be true for Ectocarpus (Chapman & Chapman, 1973), Cladophora (Fritsch, 1935), BatrachospermumlAudoninella, and others (Bold & Wynne, 1978: 486). Lubchenco & Cubit (1980) relate heteromorphic phases to herbivory pressures, upright stages accomplishing high rates of growth and reproduction in the absence of predation and crustose stages surviving pe- riods of heavy grazing. However, such morphological differences are not always associated with differences in ploidy and can sometimes be induced by the phys- ical environment (e.g., Saccorhiza polyscides, Norton, 1969), The red alga Chondrus crispus is relatively well studied; it apparently exhibits extraordinary variability in seasonal patterns of reproduction by each isomorphic generation, both regionally and between habitats (Prince & Kingsbury, 1973; Mathieson & Burns, 1975; Chen & McLachlan, 1972; Taylor & Chen, 1973; Mathieson & Prince, 1973; Hehre & Mathieson, 1970). Such variability could provide an excellent system for exploring possible adaptive values of seasonal patterns. Unlike the kelps and Postelsia, which depend on seasonally opened patches for settlement (see below), Chondrus spores settle in quiet microsites protected by other plants (Prince & Kingsbury, 1973) and maturation may be slow (Chen & McLachlan, 1972; Mathieson & Burns, 1975). Strong seasonal patterns of reproduction (e.g., Longton & Greene, 1969; Tal- lis, 1959) and marked seasonal differences in sporophyte production (e.g., Greene, 1960; Jones, 1947; Arnell, 1905) are known among many groups of closely related mosses, for example, although the ecological basis for such patterns seems to be unstudied. Ecological comparisons of such species might well elucidate the adaptive basis of seasonal differences in reproductive schedules and the timing of each generation. 1981] WILLSON— PLANT LIFE CYCLES 289 If niche differences between gametophyte and sporophyte indeed exist, we can begin to search for possible advantages accruing from the exploitation of different environments. The '"models" derived from animals suggest that two likely factors are resource exploitation and dispersal. However, the question still remains as to why it is often entities of different ploidy levels that serve different functions and why, instead, plants with complex life cycles have not evolved, for instance, a 2N ''larva" to exploit certain conditions. In fact, it seems that a few algae do exhibit heteromorphic juvenile forms that transform, without a repro- ductive event, into the adult of that generation (Fritsch, 1942; Allsop, 1966). But the existence of such forms only changes the question to why are they not more common. In short, the possible existence of niche differences does not explain why those niches are often exploited by a haploid and a diploid generation. (In some cases of heteromorphic life cycles, ploidy levels are not regularly associated with a particular morph; Lubchenco & Cubit, 1980.) Haploids can commonly grow and multiply (asexually) faster than diploids (Cavalier-Smith, 1978; Adams & Hansche, 1974; Weiss et al., 1975), owing to their smaller volume and relatively greater surface area. Therefore they may be suitable entities whenever rapid multiplication is advantageous. If food or space resources are limiting, selection may favor rapid multiplication as a means of preempting the resource and outcompeting lineages that multiply more slowly (see also Cohen, 1977). Such a tactic can be advantageous only up to the point where sibling competition outweighs nonsibling competition. Two factors may alleviate sibling competition: (1) Siblings disperse before sibling competition be- comes too intense. (2) Siblings can sequester resources and keep them from other sibling groups. Then if sibling competition becomes intense, death of some sib- lings might release enough resources that the success of remaining siblings is compensatingly enhanced. Rapid multiplication may also be advantageous if the season suitable for multiplication is short but selection favors production of large progenies. Many offspring improve the success of dispersal (see below). Finally, if there is competition for the privilege of participating in fertilization, there is selection for being well represented in the gamete pool at the time that fertilization occurs. If haploids can replicate faster than diploids (either for intrinsic reasons such as cell size or because the environment they exploit permits them to do so), then a haploid generation may be an evolutionary means of generating numerous gametes to increase the probability that the diploid of one generation will be the parent of many diploids in the next generation. In this case loss of genes that are deleterious in the haploid state may be more than compensated, particularly if release from sibling competition among the haploids allows the remainder to multiply still more. Diploidy is generally considered to have evolved after haploidy, and therefore it might be thought that the only thing to be explained is diploidy itself. As we have seen, there exist several suggestions about the origin and maintenance of diploidy; the other side of the question is, why retain the haploid? Furthermore, even if haploidy were the primitive condition, it does not follow that extant organisms were originally haploid and secondarily evolved a diploid generation. Diploidy is well established in most major algal groups and goes hand-in-hand with the condition of eukaryosis (Raper & Flexer, 1970). What needs to be ex- 290 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 plained here is not so much the existence of the diploid, but rather the balance between haploid and diploid phases. Differences in potential growth rates of the two phases seem a likely basis for differential niche exploitation. To the extent that cells modify their DNA content and nuclear size (Jones, 1975; Cavalier- Smith, 1978), such differences may be either decreased or increased. The peculiar institutions of prolonged dikaryosis or heterokaryosis found among the fungi attracts special consideration (e.g., Raper, 1966b). In a number of species the dikaryon or heterokaryon grows vegetatively, independent of the haploid phase. In many smuts and some yeasts, the dikaryon is not only well developed and long lasting, it has quite different nutritional requirements and/or host specificity than the haploid phase (Raper, 1966b). Thus niche differences are clear. However, the functional significance (in terms of evolutionary adaptation) of dikaryosis compared to true diploidy seems unclear, and Raper (1971) calls them "functional substitutes." Furthermore, many asexual fungi recombine so- matically in the dikaryotic phase, forming haploid recombinant products similar to those of sexual reproduction (Raper, 1966b). However, although the genes of dikaryons may usually be expressed as if they were truly diploid and the genes of one nucleus may often compensate for a deficiency in the other (e.g., Medina, 1977), such complementation is not always observed (e.g., Fincham et al., 1979; Lewis & Vakeria, 1977; Medina, 1977; Senathirajah & Lewis, 1975; Roberts, 1964; Pontecorvo, 1963). Darlington (1958) suggests that dikaryosis reduces the costs of reproduction by permitting one fully developed mycelium to fertilize another; the invading nucleus divides upon entry and its descendants move along the receiving mycelium, so that the whole mycelium becomes dikaryotic from a single fertilization event. Furthermore, the genetic character of the dikaryotic mycelium, at least in Ascomycetes, can adjust directly to environmental changes (often created by its own activities; Burnett, 1976) by differentially changing the relative numbers of each type of nucleus (Darlington, 1958). Selection may favor small size of nuclei (and thus IN) to facilitate division and movement along the hypha (Cavalier-Smith, 1978). In addition, there is some evidence from several basidiomycetes that dikar- yons can grow faster (in the lab) than their uninucleate forms (e.g., Croft & Simchen, 1965; Simchen & Jinks, 1964). A rapidly expanding, dikaryotic myce- lium might be advantageous in reaching and preempting such substrates as lignin (B. Benner, pers. comm.). Many basidiomycetes, including these, feed on lignin (Webster, 1970), which persists in the forest soil for a long time and is relatively common (Robinson, 1967). The sugar and cellulose substrates commonly used by other fungi are more ephemeral than lignin, and these fungi have contrasting adaptations that emphasize rapid dispersal and waiting for the resource to become available (Robinson, 1967). However, how general is the occurrence of faster // responsible, seem to be unknown. An intriguing but unexplored possiblity is that of conflict between the nuclei Will two different sources may differ, and it is conceivable that one nucleus competes with the other for resources. A winning nucleus could replicate faster and come to dominate the hypha and perhaps future reproduction in that heterokaryote. 1981] WILLSON— PLANT LIFE CYCLES 291 Mechanisms such as synchrony of nuclear division would keep such ''cheating'' from occurring; but the control of synchrony and its adaptive function need to be elucidated. The persistance of dikaryotic states in the face of such potentially destabilizing tendencies deserves attention. DISPERSAL Either the haploid or the diploid phase, or both, may disperse. The IN stage disperses in many lower plants, although it is the 2N phase in Fucus and both in some green algae. The diploid phase is dispersed in seed plants and in animals. (For purposes of this essay, I am excluding pollen dispersal for seed plants be- cause that is not directly involved with establishment of a new individual.) I suspect that dispersal occurs whenever in the cycle it is possible and expedient. Expediency is controlled by a variety of factors, including the availability of proper currents or other vectors that aid dispersal and the presence of suitable safe sites for colonization (Armstrong, 1976). Typically, the dispersal state oc- cupies more continuous habitat than the establisher, whose habitat is often patchy. A good example may be provided by the kelps (S. E. Franson, pers. comm.), which release their spores in winter, when storms create new openings for es- tablishment in the rocky subtidal zone (Dayton, 1975). The tiny gametophytes can settle in these sites, claiming them for the far larger sporophytes that grow, initially, on the parent and later take its place. The annual brown alga Postelsia palmaeformis also depends on continual re-creation of new habitat openings for survival (Paine, 1979). The first sporophytes germinate in February and March, when colonization sites are most available, and grow rapidly, partly as an escape from grazing by chitons (Paine, 1979). Spores are produced through the summer and usually settle near the parent. Short-distance dispersal coupled with an un- usual ability of this alga to clear substrate and make it available for its later siblings is apparently a means for this alga to preempt space and thus to maintain the colony through time (Dayton, 1973). Dispersal is typically a high-risk tactic for colonization — a shotgun search for safe sites (see Hamilton & May, 1977; Strathmann, 1974); because mortality of dispersing propagules is likely to be high, large numbers of such units are com- monly produced (e.g., Kennedy, 1975; Ehrendorfer, 1965). The greater the ad- vantage of large numbers, the smaller each propagule may be (to some limit) (see e.g.. Smith & Fretwell, 1974). This means that dispersal should usually be pre- ceded by a multiplicative phase, particularly one of high fecundity. In some cases, as in many animal parasites, fecundity is enhanced by asexual reproduction of 'iarvae/' as an adaptation to the uncertainties of dispersal (Kennedy, 1975). Larval reproduction exploits resources not directly available to the parent and thus effectively increases the parental fecundity. The life histories and dispersal ecologies of many green algae and mosses with unusual distributions of habitats may exhibit especially strong selection for high fecundity [e.g., algae saprophytic on nitrogenous wastes or oozing tree sap, or those colonizing the backs of turtles (Prescott, 1968) or sloths, and mosses that specialize to areas burned by hot, slow fires (Southron, 1976) or to the dung of herbivores or carnivores (Crum, 1973)]. 292 ANNAI.S OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 However, if these species use sit-and-wait tactics, with long dormancies of the dispersing propagules, the effects on fecundity should be smaller. The astronom- ical differences in spore production of different mosses — differing by as much as 8 X 10^ (Crum, 1973)— are surely related to the probability of juvenile survival and availability of safe sites. Dispersing propagules tend to be small, although spore size in closely related ferns is related to ploidy levels (Walker, 1979), and apogamous fern spores tend to be fewer and larger than those of allied sexual species (Bell, 1979). RELATIVE DOMINANCE OF THE GENERATIONS For purposes of discussion, I deal first with multicellular plants that have the evolutionary potential for significant increases in body size of the physiologically defined individual. Typically we find a contrast in body size of the larger parent and the smaller, dispersing offspring. Dispersal units tend to be small, not only because offspring are usually initially smaller than their parent, but for several ecological reasons. First, to the extent that the success of dispersal is enhanced by production of numerous propagules, the parental resources available for reproduction should be divided as finely as possible, consistent with other constraints on propagule size. Second, they may be carried about more easily by currents or other vectors. Third, perhaps they may have more potentially suitable safe sites in which to land and become established. Fecundity is often broadly correlated with body size, larger bodies producing more young than smaller ones of the same species; this is true for many inver- tebrates, fishes, amphibians, reptiles, and even mammals, as well as many seed plants (e.g., Werner, 1975; Leverich & Levin, 1979; Harper «& White, 1974). Not only can large bodies often capture more nutrients, perhaps having a competitive edge over other plants (Grime, 1977), they can also structurally carry more off- spring (e.g., Calow, 1978b). Fucus vesiculosus, for example, has a minimum size for reproduction; if that minimum is not reached in time for the sexual season, sexual reproduction is delayed until the following season. Large plants can bear over 3200 receptacles for egg development. Juvenile mortality is very high, due to washing away and mollusc grazing; adults may live as much as 4-5 years in sheltered locations (Knight & Parke, 1950). Selection for production of numerous offspring is presumably strong, the reproductive season is long, and body size clearly affects "litter" size. In addition, the very small gametophyte of some leptosporangiate ferns may have eliminated production of multiple zygotes and multiple sporophytes (Buch- holz. 1922), potentially reducing both total output and the means available to a female of choosing among potential fathers for her offspring. So we might predict that large body size ("dominance") may be characteristic of the life-cycle generation subject to the greater intensity of selection for high fecundity (and capturing resources). This could be the generation that precedes the dispersal phase and/or the one whose offspring begin exploitation and preemp- tion of an ephemeral environmental patch. At the same time, there is likely to be a cost associated with growth (e.g., in terms of increased risk of mortality; Sa- rukhan, 1977), which is one source of limitation on the achievement of large body 1981] WILLSON— PLANT LIFE CYCLES 293 size. Other limits on achieving a large body size exist (Littler & Littler, 1980) and include grazing pressure (Lubchenco & Cubit, 1980). Other reproductive functions may also affect body size. The height of the sporophyte in mosses and the lower vascular plants often increases greatly at maturity (Crum, 1973) and may increase dispersal potential of the airborne spores (Coulter, 1914). On the other hand, gametophytes may be most successful if small, thereby increasing access to water for gamete movement (Coulter, 1914). Given that the sporophyte is borne by the gametophyte, rather than vice versa, perhaps for the above reasons, in mosses another factor may be involved. The I moss sporophyte is supported and often nourished by the gametophyte (Thomas et al., 1978; Bell & Woodcock, 1968; Crum, 1973; but compare Bold, 1940). Physiological dependence of sporophyte on gametophyte suggests that bryo- phytes may often have difficulty in getting enough ''food." Even though many sporophytes are independent in terms of carbon supply (Bold, 1940), much nu- trition is apparently garnered by gametophyte leaves (Watson, 1971; Tamm, 1964; Clymo, 1963) and gametophyte growth is seemingly correlated with moisture availability (Pitkin, 1975; Tallis, 1959). A conspicuous gametophyte is probably a means of feeding a growing sporophyte; a striking reduction of gametophyte size in mosses, such as Ephememm (Crum, 1973), and some ferns (Lloyd, 1974a) may indicate that nutrient capture is less critical than claiming a site for settle- ment. Finally, the risk of damage by herbivores may vary with size, and change in size is one evolutionary means of reducing such risks. For unicellular plants (and others with strict size limitations on physiological individuals), the fecundity and nutrient capture arguments may be germane from the viewpoint of evolutionary individuals. Even though body size of physiological individuals remains small, high fecundity and resource acquisition may be achieved through asexual multiplication (especially of haploids?). A limit may be placed on the extent of such multiplication if local sibhng competition reduces the reproductive ability of each individual. Body size and fecundity also have potential effects on the generation of vari- able offspring through sexual reproduction. If meiosis follows fertilization without intervening cell division, as happens in many algae (Round, 1973; Fritsch, 1945), only two possible recombinations can be realized (Svedelius, 1929). The more cells are dividing meiotically (up to a point), the closer the theoretical maximum number of variants can be approached (see also Ghiselin, 1974: 72; Burnett, 1976: 550). This argument was also presented for red algae by Searles (1980). Thus, physiological individuals producing large litters and evolutionary individuals that achieve the same fecundity in different ways both generate variable progenies. For species with both sporophyte and gametophyte generations, variability is produced at two points in the cycle (neglecting somatic mutations and recombi- nations), when meiosis produces segregant combinations and when those join in new combinations to form a zygote (Ghiselin, 1974). Conclusion The interpretation of plant life cycles as ecological ^^strategies'' remains un- satisfactorily general and incomplete for many reasons. I have suggested that complex life cycles may be adaptive in dispersal and colonization, in competitive 294 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 interactions and resource preemption, in competition for mating privileges, and in increasing fecundity to increase the likelihood of success in the first three endeavors. By calling attention to some likely kinds of interpretations, I hope to encourage both tests of these ideas and the generation of new and more specific ecological hypotheses relevant to complex life cycles in plants. The incredible array of life cycle variants seems bound to be linked to an equally fascinating variety of ecological conditions — an array perhaps unmatched in animals. LiTERATURF. CiTED Adams, J. & P. E. Hansche. 1974. Population studies in microorganisms. I. Evolution of diploidy in Saccharomyces ccrevisiac. Genetics 76: 327-338. Addicott, J. F. 1979. On the population biology of aphids. Amer. Naturalist 114: 760-764. Allard, R. W. 1965. Genetic systems associated with colonizing ability in predominantly self- pollinated species. Pp. 50-75, In H. G, Baker & G. L. Stebbins (editors). The Genetics of Colonizing Species. Academic Press, New York. Allsop, a. 1966. Developmental stages and life histories in the lower green plants. Pp. 64-87, in E. G. Cutter (editor). Trends in Plant Morphogenesis. Wiley, New York. Armstrong, R, A. 1976. Fugitive species: experiments with fungi and some theoretical consider- ations. Ecology 57: 953-963. Arnell, H. W. 1905. Phaenological observation of mosses. Bryologist 8: 41^4. Baker, H. G. 1965. Characteristics and modes of origin of weeds. Pp. 147-168, in H. G. Baker & G. L. Stebbins (editors). The Genetics of Colonizing Species. Academic Press, New York. . 1967. Support for Baker's Law — as a rule. Evolution 21: 853-856. Bell, P.R. 1979. The contribution of the ferns to an understanding of the life cycles of vascular plants. Pp. 58-85, in A. F. Dyer (editor). The Experimental Biology of Ferns. Academic Press, New York. & C. L. F. Woodcock. 1968. The Diversity of Green Plants. Arnold, London. Bennett, M. D. 1971. The duration of meiosis. Proc. Roy. Soc. London, Ser. B, Biol. Sci. 178: 277-299. . 1972. Nuclear DNA and minimum generation time in herbaceous plants. Proc. Roy. Soc. London. Ser. B, Biol. Sci. 181: 109-135. — . 1974. Nuclear characters in plants. Brookhaven Symp. Biol. 25: 344-366. & J. B. Smith. 1972. The effects of polyploidy on meiotic duration and pollen development in cereal anthers. Proc. Roy. Soc. London, Ser. B, Biol. Sci. 181: 81-107. Bernstein. H., G. S. Bvers, & R. E. Michod. 1981. Evolution of sexual reproduction: importance of DNA repair, complementation, and variation. Amer. Naturalist 117: 537-549. BiRKY, C. W. & J. J. Gilbert. 1971, Parthenogenesis in rotifers: the control of sexual and asexual reproduction. Amer. Zool. 11: 245-266. Bi ACKMAN, R. L. 1974. Life-cycle variation in Myzus persicae (Sulz.) (Hom., Aphididae) in different parts of the world, in relation to genotype and environment. Bull. Entomol. Res. 63: 595-607. BoLO, H. C. 1940. The nutrition of the sporophyte in the Musci. Amer. J. Bot. 27: 318-322. . 1967. Morphology of Plants. Fd. 2. Harper & Row, New York. & M. J. Wynne. 1978. Introduction to the Algae. Prentice Hall, Englewood Cliffs, New Jersey. Bonner, J. T. 1958. The relation of spore formation to recombination. Amer. Naturalist 92: 193- 200. 1965. Size and Cycle. Princeton Univ. Press, Princeton, New Jersey. Bosemark. N. O. 1956a. On accessory chromosomes in Festuca pratensis. IIL Frequency and geographical distribution of plants with accessory chromosomes. Hereditas 42: 189-210. . 1956b. Cytogenetics of accessory chromosomes in Phleum populations. Hereditas 42: 443- 466. Boyce, M. S. 1979. Seasonality and patterns of natural selection for life histories. Amer. Naturalist 114: 569-583. Broadbent, L. & G. D. Heathcote. 1955. Sources of overwintering Myzus persicae (Sulzer) in England. PI. Pathol. 4: 135-137. Bruce, R. C. 1979. Evolution of paedomorphosis in salamanders of the genus Gyrinophilus. Evo- lution 33: 998-1000. Bryant, E. H. 1969. A system favoring the evolution of holometabolous development. Ann. Ento- mol. Soc. Amer. 62: 1087-1091. 1981] WILLSON— PLANT LIFE CYCLES 295 BucHHOLZ, J. T. 1922. Developmental selection in vascular plants. But. Gaz. (Crawfordsville) 73: 249-286. BuRLEY, J. 1965. Karyotype analysis of Sitka spruce, Picea sitchensis (Bong.) Carr. Silvae Genet. 14: 127-132. Burnett, J. H. 1976. Fundamentals of Mycology. Ed. 2. Edward Arnold, London. Calow. p. 1978a. The evolution of life-cycle strategies in fresh-water gastropods. Malacologia 17: 351-364. . 1978b. Life Cycles: An Evolutionary Approach to the Physiology of Reproduction, Devel- opment, and Aging. Chapman & Hall, London. , M. Beveridge, & R. SiBLY. 1979. Heads and tails: adaptational aspects of asexual repro- duction in freshwater triclads. Amer. Zool. 19: 715-727. Canning, E. U. & K. Morgan. 1975. DNA synthesis, reduction, and elimination during life cycles of the eimeriine coccidian, Eimcria tenella and the hemogregarine, Hepatozoon domer^uci. Exp. Parasitol. 38: 217-227. Cavalier-Smith, T. 1978. Nuclear volume control by nucleoskeletal DNA, selection for cell volume and cell growth rate, and the selection of the DNA C-value paradox. J. Cell Sci. 34: 247-278. Chamberlain, C. J. 1935. Gymnosperms: Structure and Evolution. Univ. of Chicago Press, Chi- cago. Chapman, D. J. & V. J. Chapman. 1961. Life histories in the algae. Ann. Bot. (London), n.s. 25: 547-561. Chapman, R. H., E. J. Klekowski & R. K. Selander. 1979. Homoeologous heterozygosity and recombination in the fern Pteridium aquilinum. Science 204: 1207-1209. Chapman, V. J. & D. J, Chapman. 1973. The Algae. Ed. 2. Macmillan, New York. Chen, L. C.-M. & J. McLachlan. 1972. The life history of Chondrus crispus in culture. Canad. J. Bot. 50: 1055-1060. Clymo, R. S. 1963. Ion exchange in Sphagnum and its relation to bog ecology. Ann. Bot. (London), n.s. 27: 309-324. Cohen, J. 1977. Reproduction. Butterworths, London. Cole. L. C. 1954. The population consequence of life history phenomena. Quart. Rev. Biol. 29: 103-137. Coleman, A. W. 1962. Sexuality. Pp. 711-729, /// R. A. Lewis (editor). Physiology and Biochemistry of Algae. Academic Press, London. Collins, O. R., C. D. Therrien & D. A. Betteri y. 1978. Genetical and cytological evidence for chromosome elimination in a true slime mold, Didvmium iridis. Amer. J. Bot. 65: 660-670. Congdon, J. D., L. J. ViTT & N. H. Hadley. 1978. Parental investment: comparative reproductive energetics in bisexual and unisexual lizards, genus Cneniidophorus. Amer. Naturalist 112: 509- 521. Coulter, J. M. 1914. Evolution of Sex in Plants. Univ. of Chicago Press, Chicago. Croft, J. H. & G. Simchen. 1965. Natural variation among monokaryons of CoUybia velutipcs, Amer. Naturalist 99: 451^61. Crow, J. F. & M. Kimura. 1965. Evolution in sexual and asexual populations. Amer. Naturalist 99: 439^50. Crum, H. 1973. Mosses of the Great Lakes forest. Contr. Univ. Michigan Herb. 10: 1^04. Daly, M. 1978. The cost of mating. Amer. Naturalist 112: 771-774. D'Amato, F, 1977. Nuclear Cytology in Relation to Development. Cambridge Univ. Press, Cam- bridge. Darlington, C. D. 1958. Evolution of Genetic Systems. Ed. 2. Oliver & Boyd, Edinburgh. Davies, E. W. & B. J. Landis. 1951. Life history of the green peach aphid on peach and its relation to the aphid problem in potatoes in Washington. J. Econ. Entomol. 44: 586-590. Dayton, P. K. 1973. Dispersion, dispersal, and persistence of the annual intertidal alga, Postelsia paUnaeformis Ruprecht. Ecology 54: 432-438. . 1975. Experimental evaluation of ecological dominance in a rocky intertidal algal community. Ecol. Monogr. 45: 137-159. Dhillon, S. S., G. p. Berlyn & J. P. Mitsche. 1978. Nuclear DNA content in populations of Pinus rigida. Amer. J. Bot. 65: 192-196. Dixon, A. F. G. 1973. Biology of Aphids. Edward Arnold, London. Dixon, P. S. 1973. Biology of the Rhodophyta. Hafner, New York. Dougherty, E. C. 1955. Comparative evolution and the origin of sexuality. Syst. Zool. 4: 145-169, 190. Drebes, G. 1977. Sexuality. Bot. Monogr. 13: 250-283. Drew, K. M. 1955. Life histories in the algae with special reference to the Chlorophyta, Phaeophyta and Rhodophyta. Biol. Rev. Cambridge Philos. Soc. 30: 343-390. 296 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 DuBE, M. l%7. On the life history of Monostroma fuscum (Postels and Ruprecht) Wiltrock. J. PhycoL 3: 64-73. DuRRANT, A. 1962. The environmental induction of heritable change in Linum. Heredity 17: 27-61. Ehrendorfer, F. 1965. Dispersal mechanisms, genetic systems, and colonizing abilities in some flowering plant families. Pp. 331-351, in H. G. Baker & G. L, Stebbins (editors), The Genetics of Colonizing Species. Academic Press, New York. Emden, H. F. van, V. F. Eastop, R. D. Hughes & M. J. Way. 1969. The ecology of Myzus persicae. Annual Rev. Entomol. 14: 197-270. Emlen, J. M. 1973. Ecology: An Evolutionary Approach. Addison-Wesley Publ. Co., Reading, Massachusetts. Epstein, C. J., S. Smith, B. Travis & G. Tucker. 1978. Both X chromosomes function before visible X-chromosome inactivalion in female mouse embryos. Nature 274: 500-503. Evans, G. M. 1968. Nuclear changes in flax. Heredity 23: 25-38. EwiNG, E. p. 1977. Selection at the haploid and diploid phases: cyclical variation. Genetics 87: 195- 208. Farrar, D. R. 1967. Gametophytes of four tropical fern genera reproducing independently of their sporophytes in the Southern Appalachians. Science 155: 1266-1267. Feeny, p. 1976. Plant apparency and chemical defense. Recent Advances Phytochem. 10: 1^0. Ferris, S, D. &. G. S. Whitt. 1977a. Loss of duplicate gene expression after polyploidization. Nature 265: 258-260. & . 1977b. Duplicate gene expression in diploid and tetraploid loaches (Cypriniformes, Cobitidae). Biochem. Genet. 15: 1097-1112. Fincham, J. R. S., P. R. Day & A. Radford. 1979. Fungal Genetics. Ed. 4. Blackwell, Oxford. Flavell, R. B., M. D. Bennett, J. B. Smith & D. B. Smith. 1974. Genome size and the proportion of repeated nucleotide sequence DNA in plants. Biochem. Genet. 12: 257-269. Friedmann, T. 1959. Structure, life-history, and sex determination of Prasiola stipitata Subr. Ann. Bot. (London), n.s. 23: 571-594. Fritsch, F. E. 1935, 1945. The Structure and Reproduction of the Algae. Vol. 1, 2. Cambridge Univ. Press, Cambridge. , 1942. IL The algal life-cycle. Ann. Bot. (London), n.s. 6: 53-563. Frost, S. 1957. The inheritance of the accessory chromosomes in Centaurea scahiosa. Hereditas 43: 405^21. Fuller, H. J. 1955. The Plant World. Ed. 3. Holt, New York. Garcia-Olmeda, F., p. Carbonero, C. Aragoncillo & C. Salcedo. 1978. Loss of redundant gene expression after polyploidization in plants. Experientia 34: 332-333. Ghiselin, M. T. 1974. The Economy of Nature and the Evolution of Sex. Univ. of California Press, Berkeley. Gill, D. E. 1978. The metapopulation ecology of the red-spotted newt, Notophthalmus viridescens (Rafinesque). Ecol. Monogr. 48: 145-166. Glesener, R. R. & D. Tilman. 1978. Sexuality and the components of environmental uncertainty: clues from geographic parthenogenesis in terrestrial animals. Amer. Naturalist 112: 659-673. Goin,0. B.,C. J. Goin &K. Bachmann. 1968. DNA and amphibian life history. Copeia 1968: 532- 540. Gould, S. J. 1977. Ontogeny and Phylogeny. Belknap Press, Cambridge, Massachusetts. Greene, S. W. 1960. The maturation cycle, or the stages of development of gametangia and capsules in mosses. Trans. Brit. Bryol. Soc. 3: 736-745. Grell, K. B. 1967. Sexual reproduction in protozoa. Pp. 147-213, in T. T, Chen (editor), Research in Protozoology. Vol. 2. Pergamon Press, Oxford. Grime, J. P. 1977. Evidence for the existence of three primary strategies in plants and its relevance to ecological and evolutionary theory. Amer. Naturalist 111: 1169-1194. Hamilton, W. D. 1980. Sex versus non-sex versus parasite. Oikos 35: 282-290. & R. M. May. 1977. Dispersal in stable habitats. Nature 269: 578-581. Harper, J. L. 1977. Population Biology of Plants. Academic Press, New York. & J. White. 1974. The demography of plants. Annual Rev. Ecol. Syst. 5: 419-463. Healy, W, R. 1973. Life history variation and the growth of juvenile Notophthalmus viridescens from Massachusetts. Copeia 1973: 641-647. -. 1975. Breeding and postlarval migration of the red-spotted newt, Notophthalmus viridescens, in Massachusetts. Ecology 56: 673-680. Hehre, E. J. & A. C. Mathieson. 1970. Investigation of New England marine algae. III. Com- position, seasonal occurrences and reproductive periodicity of the marine Rhodophyceae in New Hampshire. Rhodora 72: 194-239. Henriques, J. A. P., R. Chanet, D. Averbeck & E. Moustacchi. 1977. Lethality and '^petite'' mutation induced by the photoaddition of 8-methoxypsoralen in yeast. Molec. Gen. Genet. 158: 63-72. 1981] WILLSON— PLANT LIFE CYCLES 297 HiNEGARDNER, R. 1976. Evolution in genome size. Pp. 179-199, /// F. J. Ayala (editor). Molecular Evolution. Sinauer Assoc, Sunderland, Massachusetts. Holbrook-Walker, S. G. & R. M. Lloyd. 1973. Reproductive biology and gametophyte mor- phology of the Hawaiian fern genus Sadleria (Blechnaceae) relative to habitat diversity and propensity for colonization. Bot. J. Linn. Soc. 67: 157-174. HoxMARK, R. C. 1975. Protein variation during the life cycle of the alga Viva midahUis. Pp. 279- 284, in D. L. Mulcahy (editor), Gametic Competition in Plants and Animals. North-Holland Publ. Co., Amsterdam. & 0. NoRDBY. 1974. Haploid meiosis as a regular phenomenon in the life cycle of Ulva mutahilis. Hereditas 76: 239-250. Hutchinson. G. E. 1951. Copepodology for the ornithologist. Ecology 32: 51\-511. IsTOCK, C. A. 1967. The evolution of complex life cycle phenomena: an ecological perspective. Evolution 21: 592-605. Jackson, J. B. C. 1977. Competition on marine hard substrate: the adaptive significance of solitary and colonial strategies. Amer. Naturalist 111: 743-767. Janzen, D. H. 1977. What are dandelions and aphids? Amer. Naturalist 1 1 1: 586-589. . 1979a. Janzen's reply. Amer. Naturalist 114: 156-157. . 1979b. Reply to Addicolt. Amer. Naturalist 114: 763-764. Jones, E. W. 1947. The time of fruiting of Uloia bmchii Hornsch. and U. crispa Brid. Trans. Brit. Bryol. Soc. 1: 20-21. Jones, R. N. 1975. B-chromosome systems in flowering plants and animal species. Int. Rev. Cytol, 40: 1-100. . 1976. Genome organization in higher plants. Chromosomes Today 5: 1 17-130. Kennedy, C. R. 1975. Ecological Animal Parasitology. Wiley, New York. Kennedy, J. S. & H. L. G. Stroyan. 1959. Biology of aphids. Annual Rev. Entomol. 4: 139-160. King, C. E. & T. W. Snell. 1977. Sexual recombination in rotifers. Heredity 30: 357-360. Klekowski, E. J. & H. G. Baker. 1966. Evolutionary significance of polyploidy in the Pteridophyta. Science 153: 305-307. Knight, M. & M. Parke. 1950. A biological study of Fucus vesicidosus L. and F. serratus L. J. Mar. Biol. Assoc. U.K. 29: 439-514. Kung-Chu Fan. 1959. Studies on the life histories of marine algae. L Codiolum petrocelidis and Spongomorpha coalita. Bull. Torrey Bot. Club 86: 1-12. Lamb, R. Y. & R. B. Willev. 1979. Are parthenogenetic and related bisexual insects equal in fertility? Evolution 33: 774-775. Leverich, W. J. & D. A. Levin. 1979. Age-specific survivorship and reproduction in Phlox drum- mondil. Amer. Naturalist 113: 881-903. Levin, D. A. 1975. Pest pressure and recombination systems in plants. Amer. Naturalist 109: 437- 452. & S. W. Funderburg. 1979. Genome size in angiosperms: temperate vs. tropical species. Amer. Naturalist 114: 784-795. Lewin, B. 1974. Gene Expression. Vol. 2. Wiley, New York. Lewis, D. & D. Vakeria. 1977. Resistance of p-fluorophenylalanine in diploid/haploid dikaryosis: dominance modifier gene explained as a controller of hybrid multimer formation. Genet. Res. 30: 31-43. Lewis, J. & L. Wolpert. 1979. Diploidy, evolution and sex. J. Theor. Biol. 78: 425^38. Lewontin, R. C. 1965. Selection for colonizing ability. Pp. 77-91, /// H. G. Baker & G. L. Stebbins (editors). The Genetics of Colonizing Species. Academic Press, New York. Lindenmayer, a. 1964. Life cycles as hierarchical relations. Pp. 416^70, in J. R. Gregg & F. T. C. Harris (editors). Form and Strategy in Science. Reidel, Dordrecht, Netherlands. Littler, M. M. & D. S. Littler. 1980. The evolution of thallus form and survival strategies in benthic marine macroalgae: field and laboratory tests of a functional form model. Amer. Natu- ralist 116: 25-44. Lloyd. R. M. 1974a. Mating systems and genetic load in pioneer and non-pioneer Hawaiian Pteri- dophyta. Bot. J. Linn. Soc. 69: 23-35. 1974b. Reproductive biology and evolution in the Pteridophyta. Ann. Missouri Bol. Gard. 61: 318-331. Longton, R. E. 1976. Reproductive biology and evolutionary potential in bryophytes. J. Hattori Bot. Lab. 41: 205-223. & S. W. Greene. 1969. The growth and reproductive cycle of Pleurozinm schreberi (Brid.) Mitt. Ann. Bot. (London), n.s. 33: 83-105. LuBCHENCO, J. & J. Cubit. 1980. Heteromorphic life histories of certain marine algae as adaptations to variations in herbivory. Ecology 61: 676-687. LuccHESi, J. C. 1978. Gene dosage compensation and the evolution of sex chromosomes. Science 202: 711-716. 298 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 , & J. S. Prince. 1973. Ecology of Chondrus crispus Stackhouse. Pp. 53-77 in M. J. Harvey & J. McLachlan (editors), Chondrus crispus. Nova Scotia Inst. Sci. Halifax, N.S. Lyon, M. F. 1974. Review lecture: mechanisms and evolutionary origins of variable X-chromosome activity in mammals. Proc. Roy. Soc. London, Ser. B, Biol. Sci. 187: 243-268. MATHihsoN, A. C. & R. L. Burns. 1975. Ecological studies of economic red algae. V. Growth and reproduction of natural and harvested populations of Chondrus crispus Stackhouse in New Hampshire. J. Exp. Mar. Biol. Ecol. 17: 137-156. Mkdina, J. R. 1977. Continuous variation of genie dosage in phycomyces. Genet. Res. 30: 21 1-220. Mergen, F. & B. A. Thielges. 1967. Intraspecific variation in nuclear volume in four conifers. Evolution 21: 720-724. Metzger-Freed, L. 1974. An analysis of survival in haploid and diploid cell cultures after exposures to ICR acridine half-mustard compounds mutagenic for bacteria. Proc. Natl. Acad. U.S.A. 71: 4416^420. MiTSCHE, J. P. 1968. Quantitative study of intraspecific variation of DNA per cell in Picea glauca and Pinus hanksiana. Canad. J. Genet. Cytol. 10: 590-600. . 1971. Intraspecific variation of DNA per cell between Picea sitchensis (Bong.) Carr. prove- nances. Chromosoma 32: 343-352. MoiR, R. B. & D. P. Fox. 1977. Supernumerary chromosome distribution in provenances of Picea sitchensis (Bong.) Carr. Silvae Genet. 26: 26-33. Monk, M. & H. Kathuria. 1977. Dosage compensation for an X-linked gene in preimplantation mouse embryos. Nature 270: 599-601. Muenchow, G. 1978. A note on the timing of sex in asexual/sexual organisms. Amer. Naturalist 112: 774-779. MuLCAHY, D. L. 1979. The rise of the angiosperms: a genecological factor. Science 206: 20-23. MuNTZiNG, A. 1957. Frequency of accessory chromosomes in rye strains from Iran and Korea. Hereditas 43: 682-685. Newton, J. H., M. A. Palmer & G. M. List. 1953. Fall migration of aphids with special references to the green peach aphid. J. Econ. Entomol. 46: 667-670. Norton, T. A. 1969. Growth form and environment in Saccorhiza polyscides. J. Mar. Biol. Assoc. U.K. 49: 1025-1045. Paine, R. T. 1979. Disaster, catastrophe, and local persistence of the sea palm Postelsia pahnae- formis. Science 205: 685-687. Pitkin, P. H. 1975. Variability and seasonality of the growth of some corticolous pjeurocarpous mosses. J. Bryol. 8: 337-356. Pontfcorvo, G. 1963. Microbial genetics: retrospect and prospect. Proc. Roy. Soc. London, Ser. B, Biol. Sci. 158: 1-23. Prescott, G. W. 1%8. The Algae: A Review. Houghton-Mifflin, Boston. Price, H. J. 1976. Evolution of DNA content in higher plants. Bot. Rev. (Lancaster) 42: 27-52. & K. Bachmann. 1976. Mitotic cycle time and DNA content in annual and perennial Mi- croserlnae (Compositae: Cichoriaceae). PI. Syst. Evol. 126: 323-330. , A, H. Sparrow & A. F. Nauman. 1973. Correlations between nuclear volume, cell volume and DNA content in meristematic cells of herbaceous angiosperms. Experientia 29: 1028-1029. Prince, J. S. & J. M. Kingsbury. 1973. The ecology of Chondrus crispus at Plymouth, Massachu- setts. II. Field studies. Amer. J. Bot. 60: 964-975. Rarer, J. R. 1966a. Genetics of Sexuality in Higher Fungi. Ronald, New York. . 1966b. Life cycles, basic patterns of sexuality, and sexual mechanism. Pp. 473-511, in G. C. Ainsworth & A. S. Sussman (editors). The Fungi. Vol. 2. Academic Press, New York. . 1971. Growth and reproduction of fungi. Pp. 167-230. in F. C. Steward (editor). Plant Physiology: A Treatise. Vol. 6A. Academic Press, New York. Raper, J. R. & a. S. Flexer. 1970. The road to diploidy with emphasis on a detour. Symp. Soc. Gen. Microbiol. 20: 401-432. Rasch, E. M., J. D. Cassidy & R. C. King. 1977. Evidence for dosage compensation in parthe- nogenetic Hymenoptera. Chromosoma 59: 323-340. Rees, H. 1972. DNA in higher plants. Brookhaven Symp. Biol. 23: 394-417. Rehfeldt, G. E. & D. T. Lester. 1969. Specialization and flexibility in genetic systems of forest trees. Silvae Genet. 18: 118-123. Roberts, C. G. 1964. Complementation in balanced heterokaryons and heterozygous diploids of Aspergillus nidulans. Genet. Res. 5: 211-229. Robinson, R. K. 1967. Ecology of Fungi. English Universities Press, London. Roe, K. E. 1975. Origin of the alternation of generations in plants; recombination of the traditional theories. Biologist 57: 1-13. Rose, F. L. & D. Armentrout. 1976. Adaptive strategies oi Amhystoma tigrinum Green inhabiting the Llano Estacado of West Texas. J. Anim. Ecol. 45: 713-729. Round, F. E. 1973. The Biology of the Algae, Ed. 2. St. Martins, New York. 1981] WILLSON— PLANT LIFE CYCLES 299 Sager, R. & S. Granick. 1954. Nutritional control of Chlamvdomonas reinhardi. J. Gen. Physiol. 37: 729-742. Sarukhan, J. 1977. On selective pressures and energy allocation in populations of Ranunculus repens L., /?. bulbosus L. and R, acris L. Ann. Missouri Bot. Gard. 63: 290-308. ScHAFFER, W. M. 1977. Some observations in the evolution of reproductive rale and competitive ability in flowering plants. Theor. Populat. Biol. 11: 90-104. Scott, D. H. & C. T. Ingold. 1955. Flowerless Plants. Ed. 12. A. & C. Black, London. Searles, R. B. 1980. The strategy of the red algal life history. Amer. Naturalist 115: 113-120. Senathirajah, S. & D. Lewis. 1975. Resistance to amino acid analogues in Coprinus: dominance modifier genes and dominance reversal in dikaryons and diploids. Genet. Res. Cambridge 25: 95- 107. Shick, J. M., R. J. Hoffmann & A. N. Lamb. 1979. Asexual reprodution, population structure, and genotype-environment interaction in sea anemones. Amer. Zool. 19: 699-713. Simchen, G. & J. L. Jinks. 1964. The determination of dikaryotic growth rate in the basidiomycete Schizophyllum commune: a biometrical analysis. Heredity 19: 629-649. Singh, V. P. & S. K. Roy. 1977. Mating systems and distributions in some tropical ferns. Ann. Bot. (London), n.s. 41: 1055-1060. Slade, N. a. & R. J. Wassersug. 1975. On the evolution of complex life cycles. Evolution 29: 568-57 1 . Smith, C. C. & S. D. Fretwell. 1974. The optimal balance between size and number of offspring. Amer. Naturalist 108: 499-506. Snell, T. W. 1979. Intraspecific competition and population structure in rotifers. Ecology 60: 494- 502. SoLBRiG, O. T. 1979. A cost-benefit analysis of recombination in plants. Pp. 114-130. in Topics in Plant Population Biology. Columbia Univ. Press, New York. Southron, A. L. D. 1976. Bryophyte recolonization of burnt ground with particular references to Funaria hygrometrica. L Factors affecting the pattern of recolonization. J. Bryol. 9: 63-80. Sparrow, A. H., V. Pond & R. C. Sparrow. 1952. Distribution and behavior of supernumerary chromosomes during microsporogenesis in a population of Trillium erectum L. Amer. Naturalist 86: 277-292. , H. J. Price & A. G. Underbrink. 1972. A survey of DNA content per cell and per chromosome of prokaryotic and eukaryotic organisms: some evolutionary considerations. Brook- haven Symp. Biol. 23: 451-493. Sprules, W. G. 1974. The adaptive significance of paedogenesis in North American species of Ambystoma (Amphibia: Caudata): an hypothesis. Canad. J. Zool. 52: 393-400. Stebbins, G. L. 1960. The comparative evolution of genetic systems. Pp. 197-226, in S. Tax (editor). Evolution After Darwin. Vol. 1. Univ. of Chicago Press, Chicago. . 1966. Chromosomal variation and evolution. Science 152: 1463-1469. . 1971. Chromosomal Evolution in Higher Plants. Addison Wesley, Reading, Massachusetts. Strathmann, R. 1974. The spread of sibling larvae of sedentary marine invertebrates. Amer. Nat- uralist 108: 29^4. SvEDELius, N. 1929. An evaluation of the structural evidences for genetic relationships in plants: algae. Proc. Int. Congr. PI. Sci. 1: 457-471. Tallis, J. H. 1959. Studies in the biology and ecology of Rhacomitrium lanuginosuni Brid. IL Growth, reproduction and physiology. J. EcoL 47: 325-350. Tamm, C, O. 1964. Growth oi Hylocomium splcndcns in relation to tree canopy. Bryologist 67: 423- 426. Taylor, A. R. A. & L. C.-M. Chen. 1973. The biology of Chondrus crispus Stackhouse: system- atics, morphology, and life history. Pp. 1-19, in M. J. Harvey & J. McLachlan (editors), Chon- drus crispus. Nova Scotia Inst. Sci., Halifax, Nova Scotia. Teoh, S. B. & R. N. Jones. 1978. B chromosome selection and fitness in rye. Heredity 41: 35-48. Thomas, R. J., D. S. Stanton, D. H. Longendorfer & M. E. Farr. 1978. Physiological evalu- ation of the nutritional autonomy of a hornwort sporophyte. Bot. Gaz. (Crawfordsville) 139: 306- 311. Treisman, M. 1976. The evolution of sexual reproduction: a model which assumes individual se- lection. J. Theor. Biol. 60: 421-431. Turkington, R. & J. L. Harper. 1979. The growth, distribution and neighbour relationships of Trifolium repens in a permanent pasture. J. Ecol. 67: 245-254. , M. A. Cahn, a. Vardy & J. L. Harper. 1979. The growth, distribution and neighbour relationships of Trifolium repens in a permanent pasture. J. Ecol. 67: 231-243. Upshall, a., B. G1DDINGS& I. D. MoRTiMORE. 1977. The use of benlate for distinguishing between haploid and diploid strains of Aspergillus nidulans and Aspergillus terreus. J. Gen. Microbiol. 150: 413^18. Wahl, H. a. 1965. Alternation of generations — again. Turtox News 43: 206-209, 248-251. 300 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Walkkr, T. 1978. The evolution of sexual reproduction as a repair mechanism. Pts. 1 and 2. Acta Biolheor. 27: 133-158. Walker. T. G. 1979. The cytogenetics of ferns. Pp.87-132» in A. F. Dyer (editor). The Experimental Biology of Ferns. Academic Press, New York. Warner, R. R. 1978. Sexual-asexual evolutionary equilibrium? Amer. Naturalist 112: 960-962. Wassf.rsug, R. J. 1974. Evolution of anuran life cycles. Science 185: 377-378. . 1975. The adaptive significance of the tadpole stage with comments on the maintenance of complex life cycles in anurans. Amer. Zool. 15: 405-417. Watson, E. V. 1971. The Structure and Life of Bryophytes. Ed. 3. Hutchinson, London. Webster, J. 1970. Introduction to Fungi. Cambridge Univ. Press, Cambridge. Weiss, R. L., J. R. Kukora & J. Adams. 1975. The relationship between enzyme activity, cell geometry and fitness in Saccharomyaw cerevisiae, Proc. Natl. Acad. U.S.A. 72: 794-798. Werner, P. A. 1975. Predictions of fate from rosette size in teasels (Dipsacus fuUorum L.). Oeco- logia 20: 197-201. West» J. A. & R. E. NoRRis. 1966. Unusual phenomena in the life histoires of Floridcae in culture. J. Phycol. 2: 54-57. White, M. J. D. 1973. Animal Cytology and Evolution. Ed. 3. Cambridge Univ. Press. Cambridge. Whiitier, D. P. 1964. The influence of cultural condition in the induction of apogamy in Ptcridiufu gametophyles. Amer. J. Bot. 51: 730-736. -. 1970. The rate of gametophyte maturation in sexual and apogamous species of ferns. Phy- lomorphology 20: 30-35. Wilbur, H. M. 1980. Complex life cycles. Annual Rev. Ecol. Syst. II: 67-94. & J. P. Collins. 1973. Ecological aspects of amphibian metamorphosis. Science 182: 1305- 1314. Williams. G. C. 1975. Sex and Evolution. Princeton Univ. Press, Princeton, New Jersey. Williams, R. M. & I. Walker. 1978. The evolution of sexual reproduction as a repair mechanism. Part II. Mathematical treatment of the wheel model and its significance for real systems. Acta Biotheor. 27: 159-184. WiLLSON. M. F, & N. BuRLEY. In press. Mate choice in plants: mechanisms and consequences. Princeton Univ. Press, Princeton, New Jersey. Wynne, M. J. 1969. Life history and systematic studies of some Pacific North American Phaeo- phyceae (brown algae). Univ. Calif. Publ. Bot. 50: 1-88. FLORAL REWARDS: ALTERNATIVES TO POLLEN AND NECTAR 1 Beryl B. Simpson^ and John L. Neff^ Abstract In angiosperms selection has led to the utilization of numerous substances other than pollen and nectar that help to insure repeated visitation by pollinating animals. Here, we group the various substances into nonnutritive and nutritive rewards and discuss within each group the specific kinds that occur. In our discussion of nutritive rewards, we emphasize floral oils, lipids produced by one of two types of specialized secretory organs called elaiophores and which serve as nutritive rewards for certain New World anthophorine bees. Although discovered only within the last 15 years, the syndrome of oil production now appears to be one of the most widespread kinds of floral rewards. We report here for the first time the occurrence of oil production in the Solanaceae {Nierembergia). It is apparent that oil production has evolved independently many times, but plants which produce oils that are collected by female anthophorine bees show similarities in the chemistry of the oils and the types of structures that produce them. It is not clear whether other groups of plants reported to produce oils but which are not pollinated by anthophorine bees possess an analogous system or not. Floral rewards can be considered any component of a flower or inflorescence that is used by animals and, because of this use, insures repeated visitation that will lead to pollination. Without doubt, pollen and nectar are the primary rewards offered by flowers to visiting animals in order to buy their services as pollinating agents. Of the two, nectar is sought by a wider array of animals than pollen. On the other hand, pollen is the primary reward for which bees, probably the single most important group of pollinators, visit flowers. The role of pollen and nectar in the attraction of potential pollinators has been appreciated for hundreds of years, but we have only recently begun to realize the complex nature of these two rewards. Nectar, for example, formerly considered to be a simple sugar solution, has been shown to consist of a variety of chemicals dissolved, or sus- pended, in an aqueous solution. These range from mixtures of one to three com- mon sugars (glucose, sucrose and fructose) to more complex sugar solutions (Percival, 1961) or combinations of sugars, free amino acids, 'Witamins,"' lipids, and other compounds (Baker & Baker, 1975; Baker, 1978). The complex chemical nature of pollen has been realized for a century (refs, in Barbier, 1971), but only in the last twenty years have researchers begun to explore the varied nature of specific enzymes contained in the pollen walls and their possible roles in incom- patibility reactions (Stanley & Linskens, 1974). These same enzymes may play a role in pollen recognition by specific pollinators. The chemistry of pollen is in fact so complex that it has been impossible to provide a precise description of pollen chemistry that is all-inclusive. The continued elucidations of the intricate nature of these common rewards has spurred studies of pollination biology and provided an impetus for the investigation or reinvestigation of other floral re- wards. * We thank S. Yankowski, M. J. Mann and S. Braden for assistance in the preparation of spec- imens for Figs. 1-13. ^ Department of Botany, University of Texas, Austin, Texas 78712. ^ 7307 Running Rope, Austin, Texas 78731. Ann. Missouri Bot. Card. 68: 301-322. 1981. 0026-6493/81/0301-0322/$02.35/0 302 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Table L Floral rewards other than nuptual nectar and functional pollen Nonnutritive rewards Incidental by-products of floral structure Floral trichomes used in nest construction Sleeping places Heat sources Mating sites Products actively secreted as rewards for potential pollinators Nest materials (resins, waxes, or chemical mixtures) Sexual atlractants Nutritive rewards Brood places (larval nutrition) Adult nutrition Food tissues (food scales, food bodies, sweet tissues, pseudopollen) Nonfertile ''food'* pollen Stigmatic secretions Fatty oils (lipids) Here we will concentrate on floral rewards other than nectar and pollen in the traditional senses of pollen as functional gametes and nectar as a primarily sugar- dominated water solution secreted from one of the numerous kinds of floral nec- taries (Fahn, 1952, 1979; Kartashova, 1965), Most of the rewards we discuss have been derived from totally different portions of the flowers or inflorescences, but we will include rewards that have been derived from pollen or nectar but which are now functionally or chemically distinct. In our discussion, we have divided alternative rewards into two groups. The first group includes those used by pollinators for purposes other than nutrition and the second, those which serve primarily as food sources for adults, larvae, or both. Table 1 lists the kinds of rewards within each of these categories. Of the rewards in the latter group, we will emphasize lipids most heavily, primarily those of the Krameriaceae and Malpighiaceae. groups with which we have been working for several years. Nonnutritive Floral Rewards STRUCTURES WHICH SECONDARILY SERVE AS FLORAL REWARDS Nest Construction. — In several cases animals, primarily insects, visit flowers for something which appears to be unrelated to pollination per se. In other words, the structures used appear to have an adaptive function not directly related to the attraction of pollen vectors. The relationships between the flower and the visitor in these instances is often so casual and/or the probability of pollen transfer so low, that there has been no selection for the enhancement of the association, and consequently no further modification of the structures used. An example of this type of association is the collection of floral trichomes for use in nest con- struction by some bees. We have observed bees of the genus Anthidium visiting Krameria and Lurrea flowers in order to clip trichomes from the surface of the ovaries. The position of the bees indicates that they can effect some pollination while engaging in this activity. Generally, however, these bees use vegetative trichomes, and there appears to be no selection for an increase in the abundance 198]] SIMPSON & NEFF— FLORAL REWARDS 303 of ovarian trichomes as a reward to encourage such visitation. Similarly, during the collection of petals or pieces of petals of some flowers, leaf cutter bees oc- casionally may effect pollination, but the relationship is always a very casual one. Sleeping Places and Heat Sources. — Flowers, as a result of their shapes and behavior, can be used by either male or female (but primarily by male) bees as sleeping places. In some instances the selection of a particular flower species in which to spend the night is quite rigid; in other cases, any funnel, tube, or dish- shaped flower that closes for the night apparently will do. The use of flowers as sleeping places is quite common, but rarely reported even though the movement into and out of flowers undoubtedly leads to occasional pollination (Linsley et al., 1956; personal observations). Nevertheless, there appears to be no selection for the reinforcement of this relationship, presumably because the females of the species which sleep in the flowers usually collect nectar and/or pollen from open flowers during the day and serve, much more efficiently, as pollinators. Resembling the use of flowers as sleeping places is their use as ''heaters." Particularly in the Arctic and at high elevations where ambient air temperatures are quite low, dish-shaped flowers can collect heat within the bowl and provide energy necessary for insect activity (Hocking & Sharplin, 1965; Kevan, 1972; Smith, 1975). Several researchers have postulated that selection has increased the ability of some flowers to absorb or concentrate heat (Hocking & Sharplin, 1965; Kevan, 1972; Smith, 1975) because they attract insects that can serve as pollinating agents. SUBSTANCES PRODUCED SPECIFICALLY TO SERVE AS FLORAL REWARDS Nest Construction. — The production of chemical substances by plants that are gathered by bees for use in nest construction is well known (Krombein, 1967; Grigarick & Stange, 1968; Iwata, 1976). The substances are generally resins ex- uded from the stems of plants, including conifers, legumes, mangroves, and species of the Euphorbiaceae. These exudates are believed to serve a primary function within the plant as deterrents to predation (Berryman, 1972). A novel case of resin production as a ' 'floral' ' reward has been recently investigated by Armbruster & Webster (1979) in Dalechampia (Euphorbiaceae). Terpenes se- creted by glands on the pseudanthium, a highly reduced inflorescence, of several species of the genus (Armbruster, in correspondence), attract female euglossine, anthidiine, and trigonine bees that visit the inflorescences to gather the resin and, while doing so, deposit and pick up pollen. In actuality the secretory gland is a vegetative structure, but the nature of the inflorescence is such that the entire structure with its subtending bracts functions as a flower. Some species of Clusia (Guttiferae) have been reported also to have flowers that secrete a sticky substance collected by bees, presumably for use in nest construction (Armbruster & Webster, 1979; Armbruster, in correspondence). Florally produced waxes of Maxillaria divaricata, M. veriferum and M. flavo- viride (Orchidaceae) collected by female bees as nest-construction material can apparently also serve as pollinator rewards (Porsch, 1905; van der Pijl & Dodson, 1 966) . Other cases that may eventually be shown to involve the production of non- 304 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 nutritive rewards include Ericaceae that produce concentrations of waxes on the back of the stamens (Dorr, 1980), oily exudates of Eria vulpina (Orchidaceae) flowers (Kirchner, 1925), and the secretions of the anthers of Mouriri (Melasto- mataceae, Buchmann, 1978, but see lipids below). Sexual Attractants. — Deceit, or the trickery of flowers to lure pollinating in- sects to themselves by mimicking food, brood places, or female insects is not considered here to be an actual reward. In Ophrys (Orchidaceae, cf. Bergstrom, 1978), flowers have been shown to produce scents that attract male bees, but they do not appear to provide any actual reward for the individuals which land on, and attempt to copulate with, the flowers. Other plant species, however, produce sexual attractants that are collected in appreciable quantities by polli- nating male bees. In these cases the chemicals can be considered true rewards. The most publicized examples of the production of sexual attractants and their collection involve members of the Orchidaceae and euglossine bees (Dodson & Frymire, 1961a, 1961b; Vogel, 1963, 1966a; van der PijI & Dodson, 1966; Dodson et ah, 1969), The oils, now known to be primarily monoterpenes, are collected only by males which land on flowers and brush patches of secretory tissue with hairs on the tarsi of the front legs. The oils are absorbed onto the plumose hairs and then transferred to the hind legs where they diffuse into highly vascularized regions inside the tibia (Vogel, 1963, 1966a). Despite years of investigation, the use of the collected oils has not yet been conclusively proved (cf. Williams, in press, for a thorough review of orchid-euglossine relationships). After it was established that the male bees were not gnawing on the petals, investigators hypothesized that the oils might contain scents that mimicked odors of females or nests (VogeK 1966a), but this idea was discounted when it was shown that the bees were collecting chemical substances (Dodson & Frymire, 1961b; Dodson et al., 1969). A second hypothesis proposed by Vogel (1966b) was that the floral oils were used by the male bees in the production of pheromones, Williams (1980, in press) has recently elaborated on this hypothesis and proposed on the basis of preliminary chemical results that the males modify the floral monoterpenes in the legs, transport them within the body, and then again chem- ically alter the compounds in the mandibular glands. The mandibular gland se- cretions are species specific pheromones. At the same time that Vogel put forth his second hypothesis, Dodson (1966) suggested that the oils were metabolically important for the male bees and prolonged their lives. This idea is now, however, generally discounted (Williams, in press), A final theory about the use of the oils later proposed by Dodson (1975) was that they were used by specific males to attract other males into Icks that subsequently attracted females with which they would mate. While the failure of field researchers to observe leks as a common phenomenon argues against this idea, it has not yet been disproved. The associations between euglossine bees and orchids is often very precise and appears to be a highly coevolved system involving precise mixtures of floral compounds and very species specific taxa of bees. The syndrome is not, however, limited to the Orchidaceae. The collection of floral scents by male euglossine bees has been reported in the Araceae {Spathiphyllum, Anthuriunu and Xantho- soma, Dodson, 1966), Gesneriaceae {Gloxinia. Vogel, 1966b; Drynionia. Wil- liams, in press), Solanaceae {Cyphomandra, Williams, in press), Euphorbiaceae 1981J SIMPSON & NEFF— FLORAL REWARDS 305 {Dalechampia, Armbruster & Webster, 1979), and Haemodoraceae {Xiphidium, Buchmann, 1978, 1980). In Dalechampia (Armbruster & Webster, 1979) a some- what unusual situation pertains in that the oils are produced by a gland on the pseudanthium rather than by a gland within a flower. Williams (in press) also mentions some Cyclanthaceae that may use floral volatile oils as rewards for pollinating insects. Nutritive Substances Floral substances consumed by animals that can serve as pollinators range from various kinds of flower tissues to complex secretory products. In our dis- cussion of nutritive floral rewards, we will consider first unspecialized tissues of flowers or inflorescences, then modified tissues, and finally, particular secretory products. The actual tissues of a flower or an inflorescence can be consumed by animals that play varying roles in the pollination of the plants on which they feed. As in the case of other types of floral reward-pollinator interactions, these associations cover the spectrum from casual encounters to obligately interdependent relation- ships. BROOD PLACES (LARVAL FOOD SOURCES) We will consider first brood place associations involving tissues that do not appear to have been modified for any particular nutritive function. However, it has been shown that the adults which oviposit in the flowers serve as pollinators while searching for, and ovipositing in, host plants. The larvae themselves rarely effect pollination as they are usually confined to a flower or inflorescence during development and are generally mobile only after the periods of anthesis and stigma receptivity have ended. When we talk of brood place-pollinator relation- ships, we are not speaking of simple parasitism of flowers and the developing ovules such as occurs with bruchids and legumes (Zacher, 1952) or tephritid flies and species of the Compositae (Christenson & Foote, 1960). While adults of these insect groups are often associated with the flowers of the species on which they oviposit and are usually quite specific in their choice of host plant, they do not constitute significant pollinators in terms of the number of visits per flower per unit time or in terms of amounts of pollen carried from flower to flower. Our use of brood place as a floral reward is restricted to cases in which the adults have been shown to be one of the most important, or the only, pollinator of the flowers involved. Perhaps the two most famous examples of plants dependent on ovipositing adults as pollinators are the Yucca (Agavaceae) — yucca moth {Tegiticula spp.) and the Ficus (Moraceae) — fig wasp (Blastophagidae) associations. The intricate relationships between these taxa and their pollinators have fascinated botanists for many years and have been described numerous times (e.g., Proctor & Yeo, 1972; Faegri & van der Pijl, 1979). In both cases it is now known that pollen, a traditional floral reward is gathered, but not consumed, by the females of both the yucca moths and fig wasps. In Yucca a pollen ball is gathered by a female moth from a flower or series of flowers. She then carries the completed ball to 306 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 another flower and begins to oviposit. She usually interrupts the ovipositing pro- cess to climb to the top of the style and force the pollen ball into the stigmatic cavity. The larvae which hatch feed on the developing seeds (Riley, 1872; Powell & Mackie, 1966). Unlike the yucca system in which one species of moth pollinates the vast majority of Yucca species, figs are pollinated by species-specific female wasps. The tiny females become dusted with pollen (active packing of pollen into pockets on the thorax and coxae also occur in many taxa, Ramirez, 1969) when they leave the synconium in which they developed and mated. After entering a different synconium in search of oviposition sites, these females pollinate nu- merous flowers. The larvae develop in, and feed on, the ovarian tissues of female fig flowers. The jackfruit, Artocarpus hcterophyllus (Artocarpaceae), likewise appears to have adapted to the use of brood-place seeking females as pollinators. Van der Pijl (1953) reported that clusters of male flowers in which females of the genus Dettopsomvia (Drosophilidae: Diptera) oviposit are enlarged relative to other members of the genus and provide a medium of floral tissue for larval develop- ment. A similar association between flies of the genus Athcrigona (Anthomyi- deae: Diptera) and Alocasia puhem (Araceae) was also reported by van der Pijl (1953). Other plant species, with perfect rather than unisexual flowers, also use floral tissue as larval food as a reward for pollinating female insects. Thrips that oviposit in flowers have been suggested as the pollinators of different species of Calluna, Phyllode, and Erica (Ericaceae) by Hagerup & Hagerup (1953), but Haslerud's (1974) results indicate that the female insects effect little or no polli- nation except, perhaps, in Erica. In Trollius europaeus (Ranunculaceae), larvae of its major pollinator Chiastochaeta trollii (Thysanoptera) consume receptacular tissue and developing ovules later in the season (Hagerup & Peterson, 1956). Recently, Brantjes (1976a, 1976b) carefully described the relationship between Mclamirium album (Caryophyllaceae) and its principal pollinator, female Hadena bicruris, a noctuid moth. The females visit the flowers for nectar and to lay their eggs in the receptacle. After hatching, the larvae consume an appreciable portion of the potential seed crop. In this particular case Brantjes concluded that there is a precarious balance between pollinator service and simple seed predation. In Thuranthos (Liliaceae) an equally hazardous system exists. The two species of this African genus appear to depend on noctuid moths (i.e., Diaphone eumela) for pollination (Stirton, 1976). The adult females, which feed on nectar of open flowers, oviposit on young, unopened buds at the apex of the flowering inflores- cences. The developing, voracious larvae crawl down the rachises, consuming ovaries and maturing fruits. In contrast to plant species which are pollinated primarily by insects feeding on completely unspecialized floral parts, are several taxa that have a proliferation of certain tissues as food sources for the larvae of pollinating insects. These plants are often obligately dependent upon their pollinating-parasitizing visitors. One of the cases in which this sort of syndrome is most dramatically exhibited is in the Hydnoraceae. Both of the genera of this family, Hydnora and Proso- panchc are obligate root parasites of other angiosperms and both appear to de- pend upon beetles for successful pollination. None of the species of the family has leaves. Flowers and fruits are the only portions of the plants borne above 1981] SIMPSON & NEFF— FLORAL REWARDS 307 ground. In Prosopanche americana, nitidulid beetles {Neopocadius nitiduloidcs) and weevils (primarily Oxycoiynus hydnorae) feed as adults first on the outer walls of the perianth and, when the flowers open, on the staminal column inner perianth walls. Once they have crawled into the flowers, the insects become temporarily trapped, but continue feeding and simultaneously oviposit into the inner perianth walls. Occasionally, a female which has arrived dusted with pollen will crawl into the lower chamber of the flower and contact the flat stigmatic surface. The anthers of Prosopanche dehisce a day or so after the flower opens, dusting females who have completed oviposition. The insects then crawl or chew their way out of the flower (Simpson & Neff, 1977). During the period of fruit maturation, the larvae of these groups feed on the particularly thick layers of floral parenchyma. Bruch (1923) described a similar sequence involving the same beetles for the second member of the genus, P, burmeisteri. Hydnora, the other genus in the family, also appears to be pollinated by beetles, but the accounts of both Marloth (1907) and Vogel (1954) indicate that it attracts carrion beetles and flies because of its fetid odor and red-purple color. It is not clear if the white, fatty structure inside the flower described by Marloth (1907) functions simply as a source of an odor attractant, or if it acts as a food body. Similar structures in Prosopanche americana, which lack a strong odor, are involved in the floral trap mechanism. ADULT FOOD SOURCES While floral or inflorescence tissue may be enlarged for larval nutrition, there appear to be no proved cases in which there is a special type of tissue produced as a source of food for the larvae of potential pollinators. For tissues other than pollen that serve as nutritive rewards for adult animals, this is not the case. The pandanaceous genus Freycinetia has fleshy bracts surrounding the flowers. These ''food bracts'' were initially hypothesized by Porsch (1930) to serve as food for birds and bats. Recently, Cox (1980) has shown that flying foxes feed on the bracts and simultaneously pollinate the flowers of the dioecious plants. Likewise, Baker (1978) reported that the fleshy sepals of male flowers of the palm Bactris major are eaten by pollen-carrying beetles, and Purseglove (1968) proposed that the staminodes in cacao flowers {Theohroma cacao, Sterculiaceae), are pierced by pollinating certopogonid midges. In several species of Araceae, parts of the spadixes have been modified into food tissues (Faegri & van der Fiji, 1979) that are gnawed by pollinating beetles. These tissues are on the low^er part of spadix in AmorphophaUus variabilis and form projections above the female flowers in Typhonium trilobaium (van der Fiji, 1953). A final example is the sweet corolla of Madhuca (Sapotaceae) species (Faegri & van der Fiji, 1979). Beach (in correspondence) has found that species of Bactris (Falmae) have glandular trichomes on the inflorescence rachises that are consumed by scarab beetles (Cyclocephala) while they mate on, and also pollinate, the flowers. About thirty years ago, Grant (1950) demonstrated that the flowers of Caly- canthus occidentalis (Calycanthaceae) are primarily pollinated by a nitidulid bee- tle (Coleopterus truncatiis) that feeds on the tepal tips that have become modified into food bodies. McCormack (1975) later demonstrated that many species of beetles are initially attracted to the flowers by a complex array of volatile com- 308 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 pounds which may mimic fungal odors, rather than by the food bodies themselves. Rickson's (1979) study of the composition of the food bodies showed that they were rich in protein with low levels of starch and lipids. Starchy food bodies serve as rewards for pollinating beetles in several species of Nymphaea (Nym- phaeaceae) (review in Schneider, 1979). In the Orchidaceae a number of cases of specialized food tissues have been reported, but many of these appear to need substantiation. The reports indicate that some orchids have a sugar-rich food tissue that must be pierced in order to obtain the sweet fluids (van der PijI «& Dodson, 1966). The aspect of these cells is similar to that of normal epidermal nectaries, but the nectar is not secreted. Among genera reported to have this type of tissue are Cattleya, Epidcndmm, and SohraUa. In Diuris a similar tissue forms a ring on the top of the receptacle (Coleman, 1932). Various workers have indicated that many Orchis species may also have a type of tissue that must be pierced to obtain a sweet liquid (Knoll, 1956; van der Pijl & Dodson. 1966), but other investigators have concluded that the spurs of the flowers are empty and attract pollinators only by deceit. Other species of the Orchidaceae have been reported to have trichomes called food hairs that are collected or consumed by pollinating bees. At least two species of Polystachyci (Porsch, 1906; Beck, 1914) have trichomes on the labellum that have been hypothesized to serve this function. However, Vogel (1978), after observing P. pohcqiiinL decided that the trichomes are "empty," of no food value, and serve as pollen mimics rather than as a food source. In Maxillaria mfescens studied by Porsch (1905), the floral trichomes were shown to contain starch and lipids. which they indicate they have evidence of food hairs being consumed by polli- nating bees. A final purported food tissue in the Orchidaceae reported by Beck (1912) is in the form of food scales on labellum of Vanilla planifoHa which appear to contain sugars and starches. Van der Pijl & Dodson (1966) also list Coelogyne and Cynihidium as having scales or hairs that are grazed by bees. Pseiidopollen. Ma We mention here that the term pseudopoUen has been used in two somewhat different ways. Van der Pijl & Dodson (1966) defined pseudopoUen as a pollenlike mass of cells that results from the disintegration of multicellular trichomes. In the cases (see also Porsch, 1909; Dodson & Frymire, 1961b), the they cite of Maxillaria (see also Porsch, 1909; Dodson & Frymire, pollenlike cells appear to contain starch and serve as an actual nutritional reward for bees that collect them. Porsch (1909) cited a similar case in Rondcletia (Ru- biaceae). Vogel (1978) later used the term pseudopoUen for trichomes that mimic pollen and attract pollinators by deceit. Among groups that use pollen-mimicking trichomes in this way are members of the Commelinaceae {Tradescantia and Commelinantia) and orchids such as Calopogon. Following VogeKs usage no reward is, of course, obtained. Nonfunctional, Dimorphic Pollen. — A number of species of angiosperms have dimorphic pollen associated with a dimorphism in pollen function. One form of the pollen serves as the male gametophyte, the other, sterile form, serves as a reward for pollinators. Vogel (1978) described several cases of plant species which have sterile pollen that is used as a mimic of pollen to lure potential pollinators into appropriately visiting the flowers. However, here we are concerned with a 1981] SIMPSON & NEFF— FLORAL REWARDS 309 derivitive of functional pollen that serves as a reward in its own right. Classical cases of food pollen occur in Cassia (Leguminpsae, Tischler, 1917) and Melas- w toma (Melatomataceae, Forbes, 1882). However, the extent of the phenomenon in these large genera is unknown, and it has not been investigated recently. We have been unable to verify the presence of dimorphic pollen in any of the species of Cassia we have examined. More recent cases of dimorphic pollen have been reported in Tripogandra (Commelinaceae, Lee, 1961), Tctracera spp. (Dilleniaceae, Kubitzi & Baretta- Kuipers, 1969), and Lecythis pisonis and Couropita guianensis (Lecythidaceae, Mori et al., 1980). In Tripogandra there is only the supposition that the two pollen forms serve different functions, and in Tetracera the sterile pollen is found in perfect flowers of New World species that are labeled as being androdioecious, but which are functionally dioecious. The sterile pollen type could, therefore, be simply the remnant of the former, sexual flower, a form of deception, or an actual reward. Within the Lecythidaceae, there is ample documentation of large, polli- nating bees feeding on the sterile form of the pollen while becoming dusted on the back with functional pollen. Secretions Other than Sugar from Floral Nectaries. — The locations and types of floral nectaries have been thoroughly discussed by Fahn (1952) and Kartashova (1965). However, sugar-dominated secretions from structures other than typical floral nectaries can serve the same function. The lapping of stigmatic fluid from Ephedra campylopoda (Porsch, 1910), Gnetum (van der Pijl, 1953), and some palms (e.g., Chamerops humilis, personal observation) by insects has been linked with pollination. Copious, sweet stigmatic secretions of Anthurium (Araceae) that serve to attract pollinators have been reported and illustrated by Dauman (1930) and Croat (1980). All of these secretions are predominantly sugar solutions, but apparently can also contain amino acids as well (Baker, 1978). According to Martin (1969) and Fahn (1979), secretions of wet stigmas are usually composed primarily of oil and amino acids with small amounts of sugar. While it is apparent that more chemical analyses are needed, it is possible that there has been a selective modification of the composition of stigmatic secretions that are used as pollinator rewards. Secretions from extrafloral structures that are parts of a compound inflores- cence that functions as a flower (e.g., in the Araceae and the Euphorbiaceae) can also serve as ''floral'' rewards. Oils. — The last nonpollen and nectar reward that we want to discuss is floral oil. This group of florally secreted chemicals was only recently recognized, but since its first report (Vogel, 1969) has been intensively studied by Vogel (1974) and our laboratory (Simpson et al., 1977, 1979; Seigler et al., 1978; Neff& Simp- son, 1981). The term floral oils, it should be pointed out, is now used only for nonvolatile oils, not the essential oils that serve as odor attractants (although these may be mixed with floral oils), or as sexual attractants. It is becoming increasingly apparent that floral oils are one of the most widespread alternatives to pollen and nectar used as rewards for flower-visiting insects. We will, there- fore, discuss floral oils in more detail than the other rewards mentioned above. In particular we will look at the taxonomic distribution of oil-secreting flowers, the structures of the organs that produce the oils, and the nature of the compounds 310 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Figures 1-3. Types of elaiophores, or oil-secreting floral glands. — 1. An epithelial elaiophore shown in cross-section, showing the highly differentiated layer of epidermal secretory cells covered by a common cuticle (x 180). — 2. A portion of an epithelial elaiophore of Knimeria gruyi (Krameri- aceae) viewed under SEM (xl81). See Fig. 4 for a view of an entire elaiophore. — 3. Trichome elaiophores of Trimezia sp. (Iridaceae) from Goias, Brazil (x62). produced. We will also elaborate upon the collection of the oils by specialized bees and indicate what is known about oil use. It has been known for at least 200 years that various angiosperms such as the members of the Malpigh Krameria have large glandular structures on the flowers. Before 1969, it was assumed that these structures were nectaries or remnants of functional nectaries. Vogel (1969) was the first to dem- onstrate in convincing detail that structures which he named elaiophores secreted lipids rather than nectar. In his later treatment (1974), he provided a detailed description of two kinds of oil-secreting organs, trichome and epithelial elaio- phores, and listed five angiosperm families in which he thought they occurred. Trichome elaiophores (Fig. 3) are glandular trichomes that secrete lipids. They usually occur in patches on the corolla, but they can also occur on the stamens or ovary bases. The apical cell (or cells) of each trichome secretes oil that collects under the surrounding cuticle or in spaces between the trichomes. Vogel (1974) found evidence of these elaiophores in the Iridaceae, Orchidaceae, and Scroph- ulariaceae. Epithelial elaiophores (Figs. 1, 2) are areas of glandular tissue with lipid-se- creting epidermal cells. In this case the oils accumulate under the collective cuticle. Vogel (1974) listed three families, Orchidaceae, Malpighiaceae, and Kra- meriaceae, with taxa that appeared to have this type of elaiophore. On the basis of anatomical evidence, Vogel (1974) hypothesized that the oils were physically secreted from both types of elaiophores through pores in the cuticle. The oils seemed to be forced through the pores by female anthophorine bees equipped with a scraper of bristles on the front and/or mid legs or sopped up by pads of absorbant hairs on the forelegs. He diagrammed (1974: 468) what he believed to be the motions of the legs of Centris during the process of forcing oils through the pores of an epithelial elaiophore of Stigmaphyllon (Malpighiaceae). In addition to his morphological studies of elaiophores, Vogel (1974) called attention to the fact that these oils are collected only by certain female bees of 1981] SIMPSON & NEFF— FLORAL REWARDS 3 1 1 the family Anthophoridae. Both the production of floral lipids and their collection appeared to be restricted to the New World although he indicated two genera of Scrophulariaceae {Diascia and Bowkeria) native to South Africa which may have flowers that secrete oils. Nothing is known of the polHnators of either genus. In several instances Vogel documented oil production with various chemical tests and personally observed oil collection by anthophorines in South America, In other cases, such as Krameria, he postulated oil production and its collection on the basis of floral anatomy and distributional records of bees. Vogel ascertained for several genera that the secretory products were lipids by showing their lack of miscibility with water. His subsequent use of thin layer chromatography demonstrated that the secretions contained mixtures of several compounds. In most cases one or two of the components was present in greater quantities than the others. For Calceolaria he carried the analyses further and, in collaboration with Dr. F. Caesar (Vogel, 1974: 88-121), investigated the struc- tures of the compounds involved. They concluded that the lipids of Calceolaria pavonii consisted primarily of diglycerides (which they called monoglycerides) with a )8-hydroxy fatty acid and an acetate attached to each glycerol backbone. They also reported small amounts of free fatty acids in the mixtures. Finally, Vogel hypothesized that the oils are used in place of nectar as the liquid component of the larval provisions of the solitary anthophorine bees. He examined the nests of several species, including fresh nests of Tapinotaspis cae- rulea and older, somewhat degraded, nests of a Centris and analyzed their con- tents. In addition to lipids he found traces of sugars that included fructose, su- crose, and di- and triglycerides. Only triglycerides are uncommon in plant nectars. Nevertheless, because the sugars were present in such small concentrations, Vogel (1974) concluded that the oils were a replacement for nectar in the larval food. Since 1974, Vogel has reported three genera in two additional families, the Primulaceae and Cucurbitaceae (Vogel, 1976a, 1976b), which appear to have tri- chome elaiophores. In contrast to all of the groups reported before, one of these genera, Lysimachia, is almost entirely temperate in distribution. Both of the members of the Cucurbitaceae {Momordica and Thladiantha) are restricted to the Old World tropics. Moreover, entirely different groups of bees from those in the New World tropics, species of Macropis (Melittidae) and Ctenoplectra (Cten- oplectridae) reportedly collect the oils of these genera. We have been investigating the phenomenon of floral oil secretion in the Krameriaceae and other New World groups for several years. In many cases our studies have confirmed Vogel's observations and conclusions. However, our data differ in some cases from his and we have been able to add to his observations. We have, for example, recently confirmed that the Solanaceae contains at least one genus, Nierembergia, that produces floral oils (Seigler, Simpson and Neff, in preparation). Nierembergia gracilis in Argentina is visited primarily by oil- collecting anthophorines {Tapinotaspis spp. and Centris spp.). Our chemical analyses of the extracts of the portions of the petals with secretory trichomes have shown that they produce, among other things, the same types of oils as other oil flowers. Most of our studies, however, have centered around Krameria and Centris, 312 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol.68 its primary pollinator. Kramcria is the sole genus in the Krameriaceae, a small group of about 15 species that has been variously aligned with the Leguminosae and the Polygalaceae. A synthesis of morphological and anatomical data suggests that a placement in the Polygalales is most realistic (Simpson & Skvarla, 1981). The flowers of all Kramcria species are zygomorphic with five, separate, showy sepals. The five (or four) petals are reduced, two or three of them form a flag above the superior ovary and the remaining two have become lens-shaped glands flanking the ovary (Fig. 4). The glandular petals are 2-4 mm in diameter and can contain a milligram of oil per flower (not 0.9 mg per gland as reported in Simpson et al., 1977). In this case, therefore, entire petals have been modified into elaio- phores. We have examined in detail, by light and scanning electron microscopy, the structure of the elaiophores of the Krameriaceae (Figs. 1,2,4) and Malpighiaceae. In neither case did we find evidence of pores in the cuticle. Schnepf (1969) earlier found no pores in the cuticle of the trichome elaiophores of Calceolaria. Unvis- ited flowers, particularly those of the Krameriaceae, do not have free oils on the gland surface. Observations of glands after bee visitation clearly shows that dur- ing the collection process, female Centris rupture the cuticles (Fig. 5) while scraping the surface. We have also examined the glands of Kramcria using transmission electron microscopy (Simpson & Johnson, in preparation). The only previous work of the ultrastructure of oil-secreting glands was carried out by Schnepf (1969). He sec- tioned and described the multicellular apices of oil-secreting trichomes of Cal- ceolaria rugosa. He pointed out the large quantities of endoplasmic reticulum (ER) in the dense cytoplasm and described what he termed complexes of endo- plasmic reticulum and leucoplasts. Dictyosomes were especially noticeable in young secretory cells. The outer cell wall of the secretory cells was uneven, but generally thick compared to the walls of the nonsecreting stalk cells of the gland. He hypothesized that the oils were somehow able to penetrate through the thinner parts of the cell wall and collect under the cuticle. Despite his efforts, Schnepf (1969) was unable to locate large areas of oil accumulation within the cells or any apparent precursors of the oils. Our studies showed many of the same features described by Schnepf. How- ever, it should be pointed out that dense cytoplasm and relatively large amounts of ER are characteristic of plant secretory cells in general (Fahn, 1979). We have also noticed what appear to be leucoplasts in the cytoplasm and evidence, par- ticularly in young cells, of vesicle formation by the ER. With the exception of normal lipid droplets, we have also been unable to pin-point areas of lipid ac- cumulation within the cells. Fresh sections cut with a freeze microtome and stained immediately with Sudan black, a stain specific for lipids, showed a dis- tribution of the stain throughout the cytoplasm. In older cells, there is a con- spicuous shrinking of the cytoplasm from the outer cell walls producing a space between the plasmalemma and the cell wall. Schnepf found no such pulling away in the cells of Calceolaria. He did, however, find a similar structure in the se- cretory glands of Salvia pra tense (Schnepf, 1972). The glands of Salvia are not oil glands and Schnepf postulated that the material which accumulated in the snace between the nlasmalemma was mucilaee. We have stained fresh, freeze- \ 1981J SIMPSON & NEFF— FLORAL REWARDS 313 Figures 4-5. Epithelial elaiophores of Krameria cuspidata before (Fig. 4) and after (Fig. 5) visitation by an oil-collecting Centris. The cuticle under which the oils collect (Fig. 4) has been ruptured (Fig. 5) by the scraping movements of the fore and mid legs of the bees. Both x20. sectioned material with both ruthenium red (indicative of aqueous materials) and Sudan black. The space between the plasmalemma and the cell wall stained with neither. We must conclude therefore, that, unlike tissues of oil seeds that accu- mulate oil within the cells (e.g., Sinapis, Rest & Vaughn, 1972), oil-secreting cells transport the oils as soon as they are manufactured, or store very small quantities at a time uniformly throughout the cell. We know that oil secretion begins before the flowers open because unopened buds are forced open by bees who scrape the glands. Likewise, it appears that even after the cuticle is ruptured by an initial visit, the glands continue secreting, at least through the initial part of anthesis. Our chemical analyses of the lipid secretions of Krameria, several Malpigh- iaceae, Iridaceae, and Nierembergia showed that the principal components of the oils differ from those reported by Vogel (1974) for Calceolaria. Vogel and his collaborator concluded that the majority of the oil they analyzed was in the form of a glyceride. While we found traces of glycerides in some samples (Seigler et al., 1978), we estimate that about 90% of the lipids is in the form of free fatty acids. Free fatty acids are relatively rare in plant tissues. Moreover, the fatty acids have an acetyl group in the (i position, an unusual position for substitutions in fatty acids. We should mention, however, that the /3-acetoxy fatty acids we have found (saturated acids with chain lengths of C16, C18, and C20), are the same as the fatty acids which Vogel found to be constituents of his glycerides. Analyses of epithelial elaiophore secretions of Malplghia glabra and Mascagnla macroptera (Malpighiaceae) and trichome elaiophore secretions of Trimezia sp. (Iridaceae) and Nierembergia all showed that they, like Krameria, contained /3-acetoxy fatty acids. Nierembergia, however, contained a wider array of lipids and phenolics than the other species. We also analyzed the oils of Lysimachia ciliata proposed by Vogel (1978) to 3 14 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Voi . 68 be an oil-producing species that is collected by female Macropis. Like the neotropical taxa visited by anthophorine bees, this species seems to contain free acetoxy fatty acids, but the extracts contain many other compounds as well. The principal components of the oils appear to be terpenes. It is, there- fore, not clear that the system in Lysimachia is strictly comparable to that of We World Buchmann (1978; Buchmann & Buchmann, 1981) recently reported that fe- Mouri myrtilloidcs (Melastomataceae) in Panama and he analyzed the secretions of glandular areas on the connectives of the anthers. His analyses demonstrated an array of classes of chemicals in the secretions, including fatty acids, amino acids, short chain acids, glycosides, and saponins. On the basis of the studies carried Mour the same syndrome of oil production as the other New World taxa Vogel (1974) and we have examined. The extracts he examined did contain numerous fatty acids, but he did not indicate if they were substituted or not. In addition, many of the components that he found in the oils of this species appear to be rather toxic (as food) to insects. His hypothesis (in press) that the bees which collect Mouriri oils may have to detoxify them suggests a use different from that of the oils of the other plants studied. Finally, Buchmann & Buchmann (1981) observed Trigona bees collecting Mouriri oils and placing them in the scopae separate from pollen. Trigona has never been reported collecting floral oils from any of the flowers previously investigated. Consequently, it appears that Mouriri and its relatives have a system of floral rewards different from that currently considered as floral oils. Because of conflicting reports, we also tried to determine if the glandular secretions of Kramcria contained sugars. Analyses of nest contents of oil-col- lecting bees have shown that some contain appreciable amounts of glucose and fructose (Simpson and Neff, unpublished). It is known that adult oil-collecting bees visit plants other than their oil hosts for nectar, and it has consequently been tacitly assumed that any sugars in the larval provisions came from nonoil plants. If there are sugars in the elaiophore secretions, they may account for the sugars found in nest provisions. Percival (1961) and Baker (1978) have reported the presence of sugars in the "nectar" of several species of supposed oil flowers of the Malpighiaceae. Since Vogel (1974; 531) hypothesized that the elaiophores of the Malpighiaceae are derived from extrafloral nectaries, sugars might logically be expected to still be present in their secretions. However, if sugars are present in large quantities in the elaiophore secretions, Vogel's suggestion that the oils are used instead of nectar is unfounded. What is necessary to determine is the quantity of sugars in elaiophores if they are present. Small quantities of sugars are found in most plant secretions such as gums, latexes, etc. Likewise, lipids are commonly found in nectar (Baker & Baker, 1975), but their quantities are usually very small relative to the total solution. We therefore analyzed simple gland extracts, material from squashes of entire glands, extracts of macerated glands, and extracts of macerated calyx lobes (of Malpighiaceous species from which the glands had been removed) of species including Kramcria lanccolata, Malpighia glabra, and Stigmaphyllon sp. (ex- 1981] SIMPSON & NEFF— FLORAL REWARDS 315 amined by both Percival and Baker). Whenever extracts were made, over fifty glands or calyx lobes were used. The plant material was extracted with distilled water, and the decanted extracts evaporated to dryness. The residue which re- mained was taken up in a very small quantity of methanol and spotted on cellulose 300 N plates (Stahl, 1969: 814) and visualized with standard reagents. In no case have we been able to find any traces of sugars. Further study is obviously re- quired, but clearly sugars are not significant components of the mature elaiophore secretions of the species we have examined. When oil collection was first reported by Vogel (1969), and its presence in anthophorine nests confirmed (Vogel, 1974), it seemed assured that the primary role of the oils was to serve as a larval food source. The fact that nectar is now known to be present in significant quantities in some of these nests (Neff and Simpson, in press), raises some doubts about nutrition as the sole use of the oils. Floral oils should provide a concentrated energy source that could allow foragers to show a higher energy profit than bees foraging just for pollen or nectar. Avail- able data on this point is sparse and inconclusive. Raw (1979) found that Centris dirrhoda foraging for oil and pollen on Malpighia punicifolia in Jamaica had very high floral visitation rates (41.8 flowers per minute) and was estimated to be able to complete a pollen-oil foraging run in seventeen minutes. In our studies, prin- cipally on Krameria species and several different species of the Malpighiaceae, we typically find much lower floral visitation rates (3.4-7.5 flowers per minute on Krameria and 5.0-19,2 on Malpighia glabra). We have also not observed foragers which collect pollen from their oil hosts. In addition, individual oil-foraging bouts are quite prolonged (occasionally over 30 minutes). As yet, it is thus impossible to formulate an energy budget for any oil-collecting bees, but we are planning to attempt this in the immediate future. Since floral oils may have qualities other than a high caloric value, we have proposed a number of alternative, but not mutually exclusive, possibilities for floral oil use. Alternative explanations we have proposed are: that the oils are incorporated into the nest linings, that the oils serve a fungicidal, bacterial or anti-predator function, or that the oils help to prevent water from being absorbed into the nest provisions (this last was sug- gested to us by Jerry Rozen). To date the only one of these we have been able to test is the possible fungicidal activity. We have tested (with the help of Robert Slocum) Krameria gland extracts against three species of fungi {Aspergillus fla- vus, Saccharomyces cerevisiae, and Fusarium sp.) known to be pathenogenic in solitary bee nests (Batra et al., 1973). The tests proved to be negative. Conse- quently, while many short chain fatty acids have a fungicidal activity (Wyss et al., 1945), those of Krameria appear to have no such effect. In fact, in the agar cultures on which we placed filter paper discs impregnated with Krameria oils, the fungi appeared to be fully capable of metabolizing them. We have not yet been able to test the other hypotheses, but think that it is unlikely that the bees are incorporating the lipids into the nest lining because closely related anthophorines which do not collect oils construct virtually iden- tical kinds of nests. Moreover, pollen is frequently incorporated into the scopal loads with the oils, yet pollen is not part of the nest linings of the taxa we have examined. At the present time therefore, it still appears most likely that the oils are used as one of the primary larval metabolites. The nutritional hypothesis gains support from the recent finding that the larvae of certain species of Anthophora 316 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 FiGURFS 6-9. Oil-collecting structures of female anthophorine bees.— 6. Right forebasitarsus of Tetrapedio mama, Tetrapcdiini {x79). — 7. Right forebasitarsus of an undescribed Tapinotaspis, {Tapinorhina) Exomalopsini (x65).— 8. Tarsus, left midlcg (distal portion) of Tapinotaspis {Tapino- taspis) ihalyhaca, Exomalopsini (x4.1) showing the brushlike collecting hairs of this species.— 9. Basitarsus, left foreleg of ParatetrapeJia nuicsta. Exomalopsini (x83). consume the maternally secreted fatty lining of their larval cells (Norden et al., 1980). Anthophora is a genus of nonoil collecting bees closely related to the oil- collecting anthophorines. The oil-collecting structures of female Anthophorinae exhibit a wide array of morphologies ranging from long, sickle-shaped, relatively straight rows of simple setae to hoodlike cups of highly modified setae or pads of finely branched hairs 1981J SIMPSON & NEFF— FLORAL REWARDS 317 Figures 10-13. Forelegs of species of Centris species (Anthophorini, Centridini) showing the variations of the oil-collecting structures present in this genus. — 10. Tibia and tarsus of a Centris aff. tricolor which collects on flowers with epithelial elaiophores (x24). — 11. Left basitarsus of C. aff. autrani which collects oils from flowers with trichome elaiophores such as Calceolaria spp. (xl21). — 12. Left basitarsus of C. versicolor, a collector of oils from species of the Malpighiaceae, all of which have epithelial elaiophores (x62). — 13. Left basitarsus of a nonoil-collecting Centris, C, pallida showing the loss of the oil-collecting structures (x62). (Figs. 6-13 and Neff & Simpson, in press). As pointed out by Vogel (1974), there is a good correlation between the type of collecting structure and the type of elaiophore of the plants visited. Species of bees with only pads of hairs on the basitarsi visit only flowers with trichome elaiophores, presumably because they are incapable of scraping and rupturing the cuticles of epithelial elaiophores. On the other hand, species with scraping combs can, and do, visit plants with either type of elaiophores. We have examined in detail the collecting structures of oil-collecting antho- phorines (Neff & Simpson, 1981). These studies suggest that oil-collecting 3 18 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 may have arisen only once, or at most twice, in this family and that collection on plant species with trichome elaiophores preceded that of collection from epithelial elaiophores. Within Centris (Figs. 10-13), however, there is an ancestral asso- ciation with epithelial elaiophores, particularly those of Malpighiaceae, indepen- dent losses of the ability to collect oils (with a corresponding reduction or loss of the collecting structures), and a variety of modifications of the oil-collecting apparatus related to the radiation in the spectrum of oil-producing hosts (Figs. 6- 13). We have studied in some detail the foraging behavior of marked females of Centris utripes in mapped populations of Krameria hinceolata in Austin, Texas. While oils are frequently incorporated into the scopal loads of pollen foraging bees, extensive observations indicate that oil foragers rarely carry significant amounts of pollen in their scopae. In this case at least, pollen foraging, which invariably involves plants other than Krameria, must occur after Krameria vis- itation. Oil foragers usually show extreme site constancy in our study area. One bee that was followed for 34.5 minutes, visited 1 19 consecutive Krameria flowers and buds within an area of 10 m^ However, some individual foragers move on a regular basis between populations 50 to 100 m apart. The same marked female bees were observed day after day in the same populations of Krameria which are composed of what appear to be 10 to 25 flowering individuals. Controlled pollinations have shown that Krameria lanceolata is self-compatible but not au- togamous. Within a plant, therefore, geitonogamy appears to be unavoidable as individual bees usually visit a high proportion of the open flowers on any partic- ular plant. However, individual bees may frequently approach, but rarely revisit, flowers they have recently visited on a given plant. We have performed a series of preliminary experiments to help to understand what are the specific attractants of Krameria flowers and how female Centris are able to recognize flowers they have recently visited. The experiments have in- volved removal of various floral parts as well as exposure of gland secretions on filter paper discs to foraging bees. Observations of the antennae dipping activities of female Centris indicates that the primary short-range cues are olfactory, al- though the isolated gland secretions on filter paper never attract female bees. Scent production, at least to a human nose, seems to be localized in the elaio- phores in K. lanceolata and K. grayi, yet, typical oil gathering motions of female Centris were still observed when either the elaiophores or the stamens were removed from flowers. In the former case the bees simply went through the scraping motions with the legs grazing only the sides of the ovary. Current evi- dence suggests that a hierarchy of cues is involved which includes floral form and color, an odor (volatile oils) mixed with the fatty oils, and scent marking, either passive or active, by the bees themselves. Further experiments are planned to unravel the sequential effects of these factors. In the Austin area Krameria lanceolata is the only native oil plant. To study the relative constancy of female Centris when presented with two potential oil Mai nigh Malpigl in central Texas, but it is planted in Austin as an ornamental. Centris atripes, our primary study bee, is quite widespread in the southwestern United States and throughout Mexico and does occur naturally in areas where Malpighia glabra is 1981] SIMPSON & NEFF— FLORAL REWARDS 319 native. We assume, however, that the bees with which we were working were naive to Malpighia since we have found none planted near the research station. After a few minutes exposure to the potted plants, the Centris foragers switched from Krameria to Malpighia and proceeded to scrape repeatedly all the open Malpighia flowers. After this initial active bout of foraging, the bees began to move back and forth between Krameria and Malpighia, These preliminary ex- periments suggest that these bees lack a strong innate preference for a particular oil host, at least among flowers of similar size and color. It is therefore possible that, as in many other pollination systems, constancy in foraging may be mediated by levels of resource availabihty. Knowledge of the blooming patterns of oil plants in a community should be particularly enlightening in view of the fact that Centris and its relatives are major pollinators of many tropical species (Frankie et al., 1976). Our own work in the arid scrub vegetation in northwestern Argentina indicates that a single oil-col- lecting species, Centris brethesi, is one of the dominant pollinators of the entire community, and its interactions with various pollen, nectar, and oil hosts have led to fairly complicated phenological patterns. These kinds of interactions should be even more interesting in more mesic tropical environments where both the diversity of oil collecting bees and oil flowers increase markedly. We therefore plan to pursue studies of oil plants as a requisite component of the pollination of tropical communities. Conclusions While pollen and nectar are without doubt the primary floral rewards, selection has promoted the use of numerous alternative rewards in an impressive number of plant species. We conclude, as have others, that selection favored the elabo- ration or use of these rewards because plants which possessed them were able to capture a segment of the pollinator community not used by other plants in a community and/or achieve greater constancy of visitation. If the use of alternative rewards insures greater constancy and, presumably, more effective pollination, we might ask why more plants have not turned to them. We believe the answer is two-fold. First, many alternative rewards are more expensive (energetically) than pollen and nectar, and, second, the use of an alternative reward often locks both the plant and the pollinator into a one-to-one relationship. While in the short run such specificity might be advantageous, it provides a situation conducive to relatively rapid extinction. It is likely that in such systems as those involving orchids and euglossines or figs and fig wasps, that coevolutionary radiation can occur rapidly, but that extinction also eliminates many species pairs. For oil plants the situation is somewhat more complex since oil species have now radiated and oil-collecting bees do not appear to be species specific in their choice of oil hosts. If several oil plants co-occur within a community, they now appear to partition their pollinator fauna along the lines often observed for plants offering traditional rewards, namely, by temporal, spatial, or size displacements. Oil flowers also often have more options than some other plants which offer alternative rewards because their pollen can be, and is occasionally, used as a reward by oil-collecting bees. Consequently, they are not locked into a system involving a single, specialized reward. 320 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 If we ask why bees have turned to the collection of oils, the answer is not straightforward and final resolution of the question must await the accumulation of more data on the life histories of oil-collecting bees. If, however, larval nutri- tion is the primary use of the collected oils, it appears likely that their high energy content per unit weight relative to carbohydrates makes its use profitable. The discovery that non-oil-collecting anthophorine bees secrete maternally synthe- sized fatty oils from the Dufour's glands which are subsequently fed upon by the larvae provides support to the idea that fats are a superior food. On the basis of present reports, few groups of flower-visiting bees appear to be capable of me- tabolizing high lipid diets. More work is needed to assess the utilization of lipids by other groups of bees and to determine the degree of dependency of oil-col- lecting anthophorines on floral oils. LiTERATURF. CiTED Armbrustlr, W. S. & G. L. Webster. 1979. Pollination of two species of Daleclwnipia (Euphor- biaceae) in Mexico by Euglossine bees. Biotropica 1 1: 278-283. Baker, H. G. 1978. Chemical aspects of the pollination biology of woody plants in the tropics. Pp. 57-82, in Tomlinson, P. B. & M. H. Zimmermann (editors). Tropical Trees as Living Systems. Cambridge Univ. Press, New York. & I. Baker. 1975. Studies of nectar-constitution and pollinator-plant coevolution. Pp. 100- 140, in L. E. Gilbert & P. H. Raven (editors), Coevolution of Animals and Plants. Univ. Texas Press. Austin. Barbier, M. 1971. Chemistry and biochemistry of pollens. Progr. Phytochem. 2: 1-33. Batra, L. R., S. W. T. Batr.\, & G. E. Bohart. 1973. The mycoflora of domesticated and wild bees (Apoidea). Mycopathol. Mycol. Appl. 49: 13-44. Beck von Mannagetta, G. R. 1912. Die Futterschuppen der Bliiten von Vanilla planifolia Andr. Sitzungsber. Kaiserl. Akad. Wiss., Math.-Naturwiss. CI., Abt. 1 121: 510-521. . 1914. Die Pollennachahmung in der Bliiten der Orchideengattung Eria. Sitzungsber. Kaiserl. Akad. Wiss., Math.-Naturwiss. CI., Abt. I 123: 1033-2047. Bergstrom. G. 1978. Role of volatile chemicals in Ophry s-poWmeXor interactions. Pp. 207-23K in J. B. Harbornc (editor). Biochemical Aspects of Plant and Animal Coevolution. Academic Press, New York. Berryman, a. A. 1972. Resistence of conifers to invasion by bark beetle-fungus associations. BioScience 22: 598-602. BRANrjES, N. B. M. 1976a. Riddles around the pollination of Melandrium album (Mill.) Garcke (Caryophyllaceae) during the oviposition by Hadcna bicruris Hufn. (Noctuidae, Lepidoptera), I. Proc. Koninkl. Ned. Akad. Wctensch., Ser. C, Biol. Med. Sci. 79: 1-12. ^ . 1976b. Riddles around the pollination of Melandrium album (Mill.) Garcke (Caryophyllaceae) during the oviposition by Hadcna bicruris Hufn. (Noctuidae, Lepidoptera), II. Proc. Koninkl. Ned. Akad. Wetensch., Ser. C, Biol. Med. Sci. 79: 127-141. BucHMANN, S. L. 1978. Vibratile pollination in angiosperms with poricidally dehiscent anthers. Ph.D. thesis, Univ. California, Davis. . 1980. Preliminary anlhecological observations on Xiphidium cacruleum (Monocotyledonae: Haemodoraceae) in Panama. J. Kansas Entomol. Soc. 53: 685-699. — & M, D. BucHMANN. 1981. Anthecology of Mouriri myrtilloidcs (Melastomataceae: Memecyleae), an oil flower in Panama. Biotropica 13 (Supplement): 7-24. Bruc H, C. 1923. Coleopteros fertilizadores de "Trosopanche Burmeisteri" de Bary. Physis 7: 82- 88. Christenson, L. D. & R. H. Foote. 1960. Biology of fruit flies. Annual Rev. Entomol. 5: 171-192. Coleman, E. 1932. Pollination of Diuris pedunculata R. Br. Victoria Naturalist 49: 179-186. Cox. P. A. 1980. Flying fox pollination and the maintenance of unisexuality in populations of Frey- cinctia rcincckci (Pandanaceae). Abstr. Second Internal. Congr. Syst. Evoi. Biol., Univ. British Columbia, Vancouver, July 17-24, 1980. p. 169. Croat, T. B. 1980. Flowering of the neotropical genus Anihurium (Araceae). Amer. J. Bot. 67: 888- 904. Daumann. E. 1930. Nektarabscheidung in der Bliitenregion einiger Araceen. Planta 12: 38-48. DoDSON, C. H. 1966. Ethology of some bees of the tribe Euglossini (Hymenoptera: Apidae). J. Kansas Entomol. Soc. 39: 607-629. 1981J SIMPSON & NEFF— FLORAL REWARDS 321 — . 1975. Coevolution of orchids and bees. Pp. 91-99, in L. E. Gilbert & P. H. Raven (editors), Coevolution of Animals and Plants. Univ. Texas Press, Austin. — & G. P. Frymire. 1961a. Preliminary studies in the genus Stanhopea (Orchidaceae). Ann. Missouri Bot. Gard. 48: 137-172. — & . 1961b. Natural pollination of orchids. Missouri Bot. Gard. Bull. 49: 133-152. — , R. L. Dressler, H. G. Hills, R. M. Adams, & N. H. Williams. 1969. Biologically active compounds in orchid fragrances. Science 164: 1234-1249. Dorr, L. J. 1980. The reproductive biology and pollination ecology of Zenobia (Ericaceae). M.A. thesis, Univ. North Carolina, Chapel Hill. Faegri, K. & L. VAN DER PijL. 1979. Principles of Pollination Ecology. Pergamon Press, New York. Fahn, A. 1952. On the structure of floral nectaries. Bot. Gaz. (Crawfordsville) 113: 464-470. . 1979. Secretory Tissues in Plants. Academic Press, New York. Forbes, H. O. 1882. Two kinds of stamens with different functions in the same flower. Nature 26: 386. Frankie, G. W., p. a. Opler & K. Bawa. 1976. Foraging behaviour of solitary bees: implications for outcrossing of a neotropical forest tree species. J. Ecol. 64: 1049-1057. Grant, V. 1950. The pollination of Calycanthus occidentalis. Amer. J. Bot. 37: 294-297. Grigarick, H. a. & L. A. Stance. 1968. Pollen collecting bees of the Anthidiini of California. Bull. Calif. Insect Surv. 9: 1-113. Hagerup, E. & O. Hagerup. 1953. Thrips pollination of Erica tetraUx. New Phytol. 52: 1-7. Hagerup, O. & V. Peterson. 1950. Botanisk Atlas. Kobenhavn, Munksgaard. Haslerud, H-D. 1974. Pollination of some Ericaceae in Norway. Norwegian J. Bot. 21: 21 1-216. Hocking, B. & C. D. Sharplin. 1965. Flower basking by Arctic insects. Nature 206: 215. Iwata, K. 1976. Evolution of Instinct. Comparative Ethology of Hymenoptera. Amerind Press, New Delhi. Kartashova, N. N. 1965. Stroenie i funktsiya nektarnikov tsvetka dvidolnikh rastenii. (Structure and Function of Nectaries). Tomask Univ. Press. Leningrad. Kevan, P. G. 1972. Sun-tracking solar furnaces in high Arctic flowers: significance for pollen and insects. Science 189: 723-726. Kirchner, V. 1925. Uber die sogenannten Pollenblumen und die Ausbeutestoffe der Bliiten. Flora 118: 312-330. Knoll, F. 1956. Die Biologic der Bluten. Springer- Verlag, Berlin. Krombein, K. V. 1967. Trap-nesting Wasps and Bees. Smithsonian Institution Press, Washington, D.C. KuBiTZi, K. & T. Baretta-Kuipers. 1969. Pollendimorphie und Androdiozie bei Tetracera (Dil- leniaceae). Naturwissenschaften 56: 219-220. Lee, R. E. 1961. Pollen dimorphism in Tripogandra grandiflora, Baileya 9: 53-56. LiNSLEV, E. G., J. W. MacSwain & R. F. Smith. 1956. Biological observations on Xenoglossa fulva Smith with some generalizations on biological characters of other eucerine bees (Hyme- noptera: Anthophoridae). Bull. S. Calif. Acad. Sci. 54: 128-141. Marloth, R. 1907. Hydnora africana. Trans. S. African Philos. Soc. 16: 465^68. Martin, F. W. 1969. Compounds from stigmas often species, Amer. J. Bot. 56: 1023-1027. McCoRMACK, J. H. 1975. Beetle pollination of Calycanthus floridus L.: pollinator behavior as a function of volatile oils. Ph.D. thesis, Univ. of Connecticut, Storrs. Mori, S. A., J. E. Orchard & G. T. Prance. 1980, Intrafloral pollen differentiation in the New World Lecythidaceae, subfamily Lecythioideae. Science 209: 400^03. Neff, J. L. & B. B. Simpson. 1981. Oil collecting structures in the Anthophoridae (Hymenop- tera): Morphology, function, and use in systematics. J. Kansas Entomol. Soc. 54; 95-123. Norden, B., S. W. T. Batra, H. M. Fales, A. Hefetz & G. J. Shaw. 1980. Anthophora bees: Unusual glycerides from maternal Dufour's glands serve as larval food and cell lining. Science 207: 1095-1097. Percival, M. 1961. Types of nectar in angiosperms. New Phytol. 60: 235-281. PijL, L. van der. 1953. On the flower biology of some plants from Java with general remarks on fly-traps (species of Annona, Artocarpus, Typhonium, Gnetum, Arisaema, and Abroma), Ann. Bogor. 1: 77-99. & C. H. DoDSON. 1966. Orchid Flowers: Their Pollination and Evolution, Univ. Miami Press, Coral Gables. PoRSCH, O. 1905. Beitrage zur ''histologischen" Bliitenbiologie. T Oesterr. Bot. Z. 55: 165-173, 227-235, 253-260. . 1906. Beitrage zur ''histologischen'' Bliitenbiologie. IL Oesterr. Bot. Z. 56: 41-47, 83-95, 125-143, 176-185. — . 1909. Neuere Untersuchungen ijber die Insektenanlockungsmittel der Orchideenblute. Mitt. Naturwiss. Vereines Steierrmark 45: 346-370. 322 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 — . 1910. Ephedra camp\lopoJa C. A. Mey. eine entomophile Gymnosperme. Ber. Deutsch. Bot. Ges. 28: 404^12. 1930. Kritische Quellenstudien iiber Blumen besuch durch Vogel. V. Biol. Gen. 6: 135-246. Powell, J. A. & R. A. Mackie. 1966. Biological interrelationships of moths and Yucca whippici (Lepidoplera: Gelechiidae, Blaslopasidae, Prodoxidae). Univ. Calif. Publ. Entomol. 42: 1-46. Proctor, M. & P. Yeo. 1972. The Pollination of Flowers. Taplinger, New York. PuRSEGLOVE, J» W. 1968. Tropical Dicotyledons. Longman, London. Rami'rez, B. W. 1969. Fig wasps: mechanism of pollen transfer. Science 163: 580-58 L Raw, a. 1979. CetUris dirrhoda (Anthophoridae), the bee visiting West Indian cherry flowers {Mal- pi^hia panicifolia). Revista Biol. Trop. 27: 203-205. Rest. J. A. & J. G. Vaughn. 1972. The development of protein and oil bodies in the seed of Sinapis alba L.. Planta 105: 245-262. RiCKSON, F. R. 1979. Ultrastructural development of the beetle food tissue of Calycanthus flowers. Amer. J. Bot. 66: 80-86. Riley, C. V. 1872. Insects shaped by the needs of flowers. In Anonymous. Report in Amer. Assoc. Adv. Sci., Dubuque, Iowa.. Aug. 1872 Meeting. Nature 6: 444. Schneider, E. L. 1979. Pollination biology of the Nymphaeaceae. Proc. IV Int. Symposium on Pollination. Maryland Agric. Exp. Sta. Spec. Misc. Publ. 1: 419-429. ScHNEPF, E. 1969. Uber den Feinbau von Oldriisen II. Die Drusenhaare in Calceohiria-BXuiQn. Protoplasma 67: 195-203. . 1972. Tubulares endoplasmatisches Reticulum in Driisen mit lopophilen ausscheidungen von Ficus, Ledum und Salvia. Biochem. Physiol. Pflanzen 163: 113-125. Seigler, D., B. B. Simpson, C. Martin & J. L. Neff. 1978. Free 3-acetoxyfatty acids in floral glands of Krameria species. Phytochemistry 17: 995-996. Simpson, B. B. & J. L. Neff. 1977. Frosopis as a niche component of plants and animals. Vascular plants. Pp. 124-132, in B. B. Simpson (editor), Mesquite. Its Biology in Two Desert Ecosyst Dowden, Hutchinson and Ross, Inc., Stroudsburg, PA. & J. J. Skvarla. 1981. Pollen morphology and ultrastructure of Krameria (Krameriaceae): Utility in questions of intrafamilial and interfamilial classification. Amer. J. Bot. 68: 277-294. — , J. L. Neff & D. Seigler. 1977. Krameria, free fatty acids and oil-collecting bees. Nature 267: 150-151. — , D. S. Seigler & J. L. Neff. 1979. Lipids from the floral glands of Krameria. Biochem. Syst. Ecol. 7: 193-194. Smith, A. P. 1975. Insect pollination and heliotropism in Oritrophium lemnophilum (Compositae) of the Andean paramos. Biotropica 7: 284-286. StahL, E. 1969. Thin-layer Chromatography. Ed. 2. Springer-Verlag. New York. Stanley, R. G. & H. F. Linskens. 1974. Pollen. Biology, Biochemistry, Management. Springer- Verlag. New York. Stirton, C. H. 1976. Thuranthos: notes on generic status, morphology, phenology and pollination biology. Bothalia 12: 161-165. Tishler, W. 1917. Pollenbiologische Studien. Z. Bot. 9: 417^88. Vogel, S. 1954. Bliitenbiologische Typen als Elemente der Sippengliederung. Bot. Stud. I: 1-177. . 1963. Das sexuelle Anlockungsprinzip der Catasetinen- und Stanhopeen-BIiiten und die wahre Funktion ihres sogenannten Futlergewebes. Oesterr. Bot. Z. 110: 308-337. — . 1966a. Scent organs of orchid flowers and their relation to insect pollination. Proc. Fifth World Orchid Conference, Long Beach, Calif. 235-259. — . 1966b, Parfiimsammelnde Bienen als Bestauber von Orchidaceen und Gloxinia. Oesterr. Bot. Z, 113: 302-361. — . 1969. Flowers offering fatty oil instead of nectar. XI Int. Bot. Congr. Abstr. p. 229. , 1974. Olblumen und olsammelnde Bienen. Akad. Wiss. Math.-Naturwiss. Kl. Trop. Pflan- zenwelt 7: 1-267. — . 1976a. Oil collecting bees of the Old World and their flowers. Paper presented at the XV Int. Congr. Entomol., 19-27 (Aug. 21) Aug. 1976.. Washington, D.C. ■ > — . 1976b. Lysimachia: Olblumen der Holarktis. Naturwissenschaften 63: 44-45. . 1978. Evolutionary shifts from reward to deception in pollen flowers. Pp. 89-96, in A. J. Richards (editor). The Pollination of Flowers by Insects. Academic Press, New York. Williams, N. H. 1980. Why male euglossine bees visit orchid flowers. Abstr. II Internat. Congr. Syst. Evol. Biol., Vancouver, B.C., Canada, 17-24 July, 1980. p. 395. . In press. The biology of orchids and euglossine bees. In Orchid Biology: Reviews and Perspectives. IL Cornell Univ. Press, New York. Wvss, 0., B. J. LuDwiG & R. R. Joiner. 1945, The fungicidal action of fatty acids and related compounds. Arch. Biochem. 7: 415^25. Zacher, F. 1952. Die Nahrpflanzen der sammenkafer. Z. Angew. Entomol. 33: 460-480. GEOGRAPHICAL ASPECTS OF BIRD-FLOWER COEVOLUTION, WITH PARTICULAR REFERENCE TO CENTRAL AMERICA 1 F. Gary Stiles^ Abstract The overall objective is to compare the ecological impact of bird-flower coevolution in different geographical areas. However, it is first necessary to define the parameters of such coevolution in broader terms than those of the traditional '^syndrome of ornithophily," which focuses very narrowly on some aspects of floral morphology. I recognize three distinct components of flower function: attraction, reward, and filtering mechanisms, and discuss their functioning in an ecological context, and as they relate to the genetic system or ''pollination unit" of the plant. Then I turn to nectar- feeding birds, and discuss not only morphological, but ecological and behavioral specializations to flowers as a food source. These discussions develop explicitly my criteria for detecting and evaluating bird-flower coevolution. The different groups of birds known to feed regularly (as opposed to oppor- tunistically) on nectar are then compared according to these criteria, to determine their relative degrees of specialization for, and dependence upon, a high-nectar diet. Different groups are found to vary widely in their degrees of specialization for flower-feeding, and it is evident that bird-flower coevolution has followed very different courses, and led to widely divergent ecological systems in different geographical areas. By any criteria the hummingbirds are the most specialized avian nec- tarivores, although they are approached in this regard by some members of certain passerine groups, notably among the sunbirds. Several groups of passerine nectarivores also occur with the humming- birds in many New World areas; these groups show low to moderate degrees of specialization for nectarivory, either as pollinators or as parasites on the hummingbird-flower system. The New World tropics thus present a wide range of specializations for flower-feeding in their avifauna, and represent a particularly interesting area for study. Patterns of ornithophily and nectarivory are thus examined in detail for this area, concentrating specifically on Southern Central America, especially Costa Rica. The altitudinal and geographical distributions of the two main groups of hummingbirds, the hermits and nonhermits, are found to differ, as are the taxonomic and ecological affinities of their primary foodplants. The hermits are most numerous in wet lowlands and the adjacent foothills, and are primarily associated with large monocotyledonous herbs, notably Heliconia. The nonhermits reach their greatest taxonomic and ecological diversity in the lower middle elevations, and are the only group present at high elevations; they seem to have coevolved with the flowers of a variety of dicot families, and the bromeliads among the monocots. Passerine nectarivores occur primarily as parasites on the hummingbird-flower system (Coerebidae) and are important as pollinators only in seasonally dry areas when the hummingbirds are poorly represented. Within the last ten to fifteen years the study of poHination has passed from a purely botanical pursuit to an extremely active field of ecology. This is especially true with respect to bird pollination systems: a resurgence of interest in the foraging ecology, energetics, and social behavior of nectarivorous birds is leading to a new and broader appreciation of their role as pollinators. This, in turn, is one of the major catalysts in the continuing emergence of a more balanced and ^ My ideas on pollination biology have been sharpened by discussion with many people, including OTS faculty and students, too numerous to mention here. However, I acknowledge a particular debt in this regard to H. G. Baker, G. W. Frankie, B. Heinrich, P. A. Opler, S. Salas, and L. L. Wolf. My studies in Costa Rica and a recent trip to Venezuela have been financed by the Vicerectoria de Investigacion, Universidad de Costa Rica; by CONICIT; and, earlier, by the American Museum of Natural History. P. H. Raven provided several helpful suggestions and references for the manuscript. for which F. L. G. Sanders supplied criticism. Finally, I thank Missouri Botanical Garden and Gerrit Davidse for the opportunity to participate in this symposium. ^ Escuela de Biologia. Universidad de Costa Rica. Ciudad Universitaria ''Rodrigo Facio,'' Costa Rica, Central America. Ann. Missouri Bot. Gard. 68: 323-351. 1981. 0026-6493/8 1/0323-0351/502.95/0 324 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 profound view of plant-pollinator coevolution. This new outlook is still evolving rapidly, and rather than attempt a theoretical synthesis, I will apply some of these new points of view to a geographical survey of bird-flower coevolution. My specific objective is to compare the kinds and degrees of coadaptations between birds and flowers in different geographical areas, hoping thereby to arrive at a better understanding of the ecological and evolutionary impact of these coad- aptations in different biota. By coadaptation, I am referring to the degree of ecological interdependence of bird and flower, as well as to the degree of mor- phological correspondence (as embodied by the well-known "syndrome of orni- thophily"). The most highly coevolved system, then, is one in which the flower is completely dependent on the bird for pollination, and the bird requires the energy provided by the flower for breeding and molt: reproduction and perhaps even survival of either would be impossible without the other. Such coadaptation makes the species involved peculiarly vulnerable— what affects one will perforce affect the other— but it also makes possible the occupation of new ecological niches or adaptive zones (e.g., winter breeding in Ribes speciosum and Calypte anna, cf. Stiles, 1973). In this paper I shall first review the kinds of coadaptations that can occur between birds and flowers, considering behavioral, physiological, and ecological as well as morphological parameters. I shall thereby develop criteria by which to compare the bird-flower coadaptations of different continents. I shall then discuss in more detail bird-flower coadaptations of the New World tropics, particularly Central America. Hopefully this analysis will lead to a deeper understanding of the ecological impact of these systems, and of the evolution of the groups of birds and flowers involved. The Syndrome of Ornithophily Revisited Until recently most studies of bird-flower coadaptations were devoted to dem- onstrating the existence of pollination by birds, and to elucidating the floral mech- anisms involved. These adaptations are summarized in the well-known "syn- drome of ornithophily" which represents the culmination of many years of observation and controversy (Faegri and van der Pijl, 1966), as well as a con- venient starting point for the present discussion. Two things are immediately obvious from this "syndrome": the emphasis is narrowly focused upon the flower itself, ignoring many other aspects of the plant's biology that might affect bird visitation and pollination. Also, virtually nothing is said about possible special- izations of birds for flower visitation. In this paper I shall attempt to take a broader and less one-sided view of bird- flower coevolution. Elsewhere (Stiles, 1978b) I have discussed in some detail the various ways in which parameters of plant populations, and of the plant itself, can impinge upon pollination biology, with particular reference to bird pollination. For other discussions of this general theme, see Gadgil & Solbrig, 1972; Stiles, 1975; Wilbur, 1976; Pitelka, 1977; Bawa & Opler, 1977. Here I wish to emphasize aspects of floral biology per se, but without ignoring this broader context. First, some definitions may be in order. Faegri & van der Pijl (1966) distinguish between "attraction unit" (those features that attract a pollinator from some distance away to the flower or inflorescence), and the "pollination unit" (within 1981] STILES— BIRD-FLOWER COEVOLUTION 325 which pollination occurs). In reality two distinct phenomena are being confused by these authors* use of "attraction unit": the attraction of the pollinators' at- tention to the flower, and the provision of a real or simulated reward that actually induces the animal to visit the flower. I restrict the term "attraction unit" (or component) to those mechanisms subserving the former function, and use "re- ward unit" (or component) for those accomplishing the latter. Clearly the func- tioning of the attraction component must be intimately related to the sensory biology of the flower visitors, while the reward component must correspond to some nutritional, sexual, or other need of the pollinator. Although this distinction may be blurred in some flowers (see e.g., Simpson, this symposium), it is clear enough in bird flowers: bright colors form the basis of the attraction component; nectar, of the reward component. Another function that I choose to distinguish is the restriction of access to the reward to a narrow segment of the potential spectrum of visitors attracted. The term "exclusion mechanisms" has been used here, but I prefer the term "filtering component" (or mechanism). At least in bird flowers, such limitations of the visitor spectrum are often not absolute exclusions but relative ones, based upon pollinator energetics: a flower not worth exploiting to some pollinators under some conditions may become profitable under others (e.g., Heinrich, 1975; Wolf et al., 1975, 1976). Finally, the term "pollination unit" can be made more explicit by relating it to the breeding system of the plant (cf. Bawa, 1974; Cruden, 1977) and genetic structure of the plant population (e.g.. Price & Waser, 1978). The "optimum pollination unit" might be defined as the minimum spatial separation of flowers over which movement of a given amount of pollen will produce maximum seed set (including seed viability, cf. Levin, this symposium). The "minimum polli- nation unit" would be the minimum separation of flowers (in terms of being on the same inflorescence, plant, or clone as well as physical distance) required for pollination to occur at afl (see Stiles, 1978b). Only for obligately self-fertilizing flowers would the two terms be synonymous (the unit would be one flower in either case). ATTRACTION In ornithophilous flowers attraction is primarily, perhaps exclusively by color, though odor cannot be ruled out entirely in some cases (Stiles, 1976). The chief requisite for attraction is thus conspicuousness to birds, which will reflect the properties of avian vision and habitat features. Birds have their greatest spectral sensitivity and finest hue discrimination towards the long wavelength end of the visual spectrum (reviews in Sillman, 1974; Stiles, 1976), although the complex interplay of cone pigments and colored oil droplets makes the situation a great deal more complex than in mammals (e.g., Bowmaker, 1977). Moreover, hum- mingbirds at least may be able to perceive ultraviolet "colors" (Goldsmith, 1980); the role of this ability in foraging remains to be studied. The prevalence of ultra- violet patterns and nectar guides in relatively nectar-poor entomophilous flowers may make it possible for hummingbirds to avoid these flowers on sight. Certainly the vast majority of bird-pollinated flowers feature long wavelength colors in the attraction unit. The occurrence of other colors, or of contrasting colors, may add to conspicuousness against particular backgrounds (see Stiles, 1976). Another 326 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 factor may be the colors of the birds themselves: various authors have noted correspondences between the display colors of certain (mostly Old World) nec- tarivorous birds and some of their preferred flowers (Faegri & van der Pijl, 1966). Most often the colorful birds are males, who may show a hormone-mediated higher responsiveness to these colors (Ducker, 1970; Morton, 1979). The manner in which the colors are displayed is related to a number of factors. The flower may carry the attractive colors if it is long-lived and/or in an open habitat. Tn many ornithophilous species immature flowers or bracts are also col- orful, adding to the size and longevity of the visual signal. In other species, the need for a conspicuous, long-lived signal has been solved by the evolution of a large, colorful inflorescence (e.g., Heliconia, Costus. many bromeliads). The flowers in such cases are often ephemeral and inconspicuous except insofar as they contrast with the inflorescence, probably adaptations to help protect them from destructive nectar thieves (cf. Stiles, 1975, 1976). A variation on this theme Is the incorporation of red or red-spotted leaves into the attraction unit (Jones & Rich, 1972; Stiles, 1978b). REWARDS Nectar is the only floral reward regularly offered to birds. Only one question- able case of regular pollen-eating has been reported for hummingbirds (Carpenter, 1976a), although it is common in parrots of the subfamily Loriinae (Churchill & Christensen, 1970; Forshaw, 1973). However, the flowers involved are often eaten by the parrots, whose role as pollinators — and hence participants in coevolution- ary relationships with plants — remains to be studied in detail (cf. Paton & Ford, 1977; Ford et al., 1979). Compared with insect-pollinated flowers, bird-pollinated flowers usually show higher nectar volumes, slightly to markedly lower concentrations, and consid- erably higher sugar production overall (Baker, 1975; Heinrich, 1975; Stiles, 1975, 1978b; Opler, 1981; Cruden et al., 1981). This reflects the high energy require- ments of the birds, and both laboratory and field studies suggest that energetic criteria are the most important determinants of flower choice in nectarivorous birds (Hainsworth & Wolf, 1976; Stiles, 1976). The three dominant sugars in nectar, sucrose, glucose, and fructose are energetically equivalent, but sucrose usually predominates in nectars of hummingbird flowers (Stiles, 1976; Baker & Baker, 1981). Old World bird-flowers generally have nectars low in sucrose; their passerine pollinators may have difficulty digesting this sugar since experiments with European Starlings (a member of a family containing a number of facultative nectarivores) demonstrate that they cannot maintain weight on a high-sucrose diet (Schulcr, 1977). Hummingbirds usually prefer sucrose-rich nectars in the field and under appropriate laboratory conditions (Stiles, 1976), but will use Old World bird-flowers freely in gardens, doubtless because of their high nectar vol- umes (Stiles, 1973, 1976). Other components of nectar include amino acids, lipids, and polysaccharides (Baker & Baker, 1975). Probably none of these is nutritionally essential to nectar- feeding birds, which have access to such substances in insects and fruit, unlike most insect pollinators (cf. Baker & Baker, 1975; Gilbert, 1975; Ford & Paton, 1981] STILES— BIRD-FLOWER COEVOLUTION 327 Table I . Rates of nectar extraction and feeding preferences among hummingbirds visiting two species of Heliconia in a Costa Rican second-growth area.^ Heliconia Species Heliconia latispatha Heliconia imhricata Hummingbird Species Slope of Ex- traction Line'' # Visits^ Slope of Ex- traction Line # Visits Amazilia tzactl Thalurania furcata Chalybura urochrysia Phacthornis superciliosus 0.097 0.125 0.252 0.322 182 70 36 13 0.176 0.119 0.112 0.252 119 201 149 88 ^ Observations made June-Aug. 1971, 1972, 1974 in an area ca. 3 ha in size, of old second growth; H. imhricata was about Wz times as abundant (in terms of numbers of flowers) as H. latispatha overall. ^ Slope of the line of time spent probing flowers {y axis) vs. nectar extracted (.v axis): the lower the slope, the more nectar can be extracted in a given time. Based on 35 or more observations except C urochrysia-H. latispatha (N ^ 21) and P. superciliosus-H. latispatha (N = 13). ^ Number of times a bird of a given species was seen visiting flowers of a given plant species during census walks through the study area. Each species shows a highly significant preference (P < 0.01 by chi-square test) for the flower at which it can most quickly extract nectar. 1976a). At low concentrations amino acids are not detected by hummingbirds; at high concentrations, they are rejected (Hainsworth & Wolf, 1976). This should place a selective premium on reducing amino acid concentrations in nectar, and Baker & Baker (1975) have indeed found that tropical bird-flowers have nectars with very low amino acid content. The more recently evolved North American hummingbird flowers have higher amino acid concentrations in their nectars — but still markedly lower than those of the presumably ancestral bee-pollinated species (Grant & Grant, 1968; Baker & Baker, 1975). FILTERING MECHANISMS I will discuss two sorts of mechanisms here; those tending to restrict visitation by nonpollinating animals, especially destructive ones; and those that enhance specificity of flower choice among potential bird pollinators. Insect visitation may be reduced in bird flowers by a variety of mechanisms: red color (in addition to its conspicuousness to birds, this color is relatively inconspicuous to some bees, although probably not to many butterflies (Raven, 1972), dilute nectar (Baker, 1975; Bolten & Feinsinger, 1977), appropriate timing of nectar production (Stiles, 1975), and perhaps nectar composition. If taste con- ditioning occurs in insects as well as hummingbirds, the production of high-su- crose nectar may in itself favor hummingbird visitation (cf. Stiles, 1976). The hard floral parts and large nectar-sexual sphere distance mentioned in Table 1 more likely evolved as protection against destructive nectar thieves than against the ''hard beaks" of pollinating birds. Opler (1981) demonstrated a direct cor- relation between nectar flow and flower weight among a large number of tropical plant species: the increase in weight coming largely from harder and thicker protective tissues in the perianth, as well as a longer corolla tube. In humming- 328 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Voi . 68 birds and presumably sunbirds the birds do not thrust the bill any further into the flower than necessary to enable the extensible tongue to reach the nectar; the tongue grooves fill with nectar most effectively in a confined space if the tongue is extended beyond the bill tip (Hainsworth, 1973; Schlamowitz et al., 1976). Restricting visitation to a few of the potential pollinators can promote effi- ciency of pollination by reducing the frequency of mixed pollen loads. Specificity of flower choice, or "flower constancy," is important at the level of the individual pollinator; degree of specificity probably reflects the relative energetic profit- abilities of the flowers available to each forager (Heinrich, 1976). Amount of nectar available, efficiency of nectar extraction, and cost of transport between flowers are probably the most important factors determining profitability at this level (Wolf et al., 1975). Different lengths and/or curvatures of corolla tubes can affect the extraction efficiencies of different hummingbird species in relation to differences (often subtle) in bill morphology; this in turn may strongly influence flower choice by the birds (for an example involving Heliconia, see Table 1: in each case the bird strongly prefers the species of Heliconia from which it can most efficiently extract nectar). Tubular flowers have the further effect of forcing the bird to orient its bill in a particular way when probing the flower, especially when bill and corolla are curved; this in turn facilitates placing pollen on a specific part of the bird, which can reduce mixing of pollen loads even on a relatively nonspecific pollinator (clearly an advantage for a rare plant species). This can open the way for a variety of highly specific morphological coadaptations (cf. Stiles, 1975; Brown & Kodric-Brown, 1979). Bird flowers mostly fall into two broad morphological groups: tube and brush. Tubular flowers are in many respects the most highly evolved and the most likely to enter into specific coevolutionary relationships. Brush flowers generally house the nectar in a cup or short tube, out from which a "brush" of stamens extends. Visitors seeking nectar (or pollen) are liberally dusted with pollen, which is brought more or less haphazardly in contact with the stigmas. Many bee- and bat-pollinated flowers are of similar construction, and it may be that brush flowers simply represent a generalized adaptation to pollinators large in relation to flower size. As such they would be very ineffective filters — any visitor in a given size range could pollinate the flower. Brush flowers may, in effect, be specialized for pollination by a wide spectrum of taxonomically diverse visitors. For instance, the red stamens of Calliandra spp. (Mimosaceae) may represent an adaptation for drawing birds into an essentially chiropterophilous syndrome. Many regular pollinators of brush flowers are quite unspecialized in their flower visitation, such that pollination occurs more or less haphazardly as the animal blunders about in or on the flower. Such "mess and soil" pollination (Faegri & van der Pijl, 1966) may be carried out by bats, nonflying mammals (Carpenter, 1976a; Sussman &. Raven, 1978), and birds such as lories. Nectar in brush flowers is available to any visitor that does not mind getting dusted with pollen; the potential for exclusive coevolutionary relationships is low. Pollinator specificity can be enhanced if interspecific competition occurs be- tween potential pollinators. Aggressive behavior and/or territoriality of dominant individuals or species can result in partial or complete exclusion of subordinate individuals and/or species from flowers that they might otherwise visit (Lyon, 1981 1 STILES— BIRD-FLOWER COEVOLUTION 329 1976; Wolf et al., 1976). The wide range in nectar production among sympatric hummingbird flowers may function in part as a resource gradient along which different hummingbirds might specialize according to their dominance status and energy needs (Feinsinger & Colwell, 1978; Stiles, 1978b). POLLINATION UNITS The production of relatively large amounts of nectar and protective tissue makes bird pollination quite expensive on a per-flower basis. Thus bird pollination will probably evolve only where the advantages, in terms of enhanced pollination, outweigh these expenses (see review in Stiles, 1978b). The pollination strategy of an ornithophilous plant amounts to making enough nectar available to attract avian visitors, while adjusting their movements between flowers to the size of the optimal (or minimal) pollination unit. This entails adjusting nectar availability in terms of the proportion of the pollinator's energy requirements that can be sat- isfied by a single flower, shoot, or clone, and thus determining the degree of interplant movement (Heinrich & Raven, 1972). Bearing in mind that they but represent the ends of a continuum, we can distinguish two extreme foraging tactics of nectarivorous birds: route foraging or traplining, and territoriality (Feinsinger, 1976; Stiles & Wolf, 1979). The former is ideal for promoting cross-pollination, but if the plants are widely scattered many flowers may be missed. Territoriality restricts pollen flow to and from the defended area, but its effect on cross-pollination depends on the effectiveness of the defense, and on whether the territory itself contains one or many plants (Linhart, 1973; Ray et al., 1981). Within the territory, systematic foraging of the resident may maximize the proportion of flowers visited (Gill & Wolf, 1975; Stiles, 1978b). Whether territorial or traplining pollinators are favored will depend ulti- mately upon the plant's spatial dispersion and breeding system, and various pa- rameters of morphology and phenology can be varied to favor one or the other type of pollinator, as has occurred in Heliconia (Stiles, 1975, 1979). Avian Adaptations for Flower-Feeding I now wish to treat the various possible avian specializations for visiting and (at least in some cases) pollinating flowers. Although virtually all of these adap- tations have been discussed previously, I know of no recent attempt to bring them together to present, in effect, a coherent ''syndrome of anthophily'' of the birds. This would seem to be a necessary first step in assessing the degree of specialization for flower visiting of any given species or group; this in turn could provide an indication of the overall evolutionary development of bird-flower coad- aptations in any given region. The features listed in my ''syndrome of anthophily" (Table 2) stress relative degree of specialization; a bird may be considered more or less specialized for flower visiting according to the extent to which it shows any particular array of features to a greater or lesser extent than do its nonan- thophilous closest relatives or putative ancestors (where known). This procedure may somewhat deemphasize absolute degree of specialization in those cases where a group is in some way preadapted for flower visiting. For instance, small size may be viewed as a specialization for flower visitation (see below). The 330 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 Table 2. Specialization for flower-feeding in birds: the '^syndrome of anthophily. 1 •> Characteristics of Birds Characteristics of Flowers 1. Small body size; usually less than 20 grams 1, Most flowers small, nectar content usually 200 fil or less 2. Bill usually slender, often long and/or curved, 2. Nectar deep-seated, often at end of long and/ or curved tube; hard flower parts (to be sur- passed by nectar-robbers) matching flower corollas; occasionally hooked, awl-pointed, etc.; nasal operculum well de- veloped 3. Tongue tip grooved, fringed, and/or capable 3. Nectar relatively dilute, low viscosity, often of rolling into tube to take up nectar by cap- in deep-seated chamber with narrow entrance illarity, tongue extensible beyond bill tip; pa- pillate tip for pollen feeders 4. Gut with extensible crop for storage and rapid 4. Nectar with low amino acid content, main pro tein source insects. absorption of nectar, esophagus and intestinal openings close together, leaving stomach as a diverticulum into which insects pass, but little nectar (not necessary for pollen feeders) 5. Agility to reach flowers, especially hovering 5. Flower hangs free or faces outward or down flight, sometimes large or strong feet to cling to inflorescence to discourage insects 6. Aggressive behavior and often feeding terri- 6. Flowers stationary, visible, highly defensible tonality well developed whenever flowers suf- ficiently abundant resource 7. Excellent spatial memory; can recall location 7. Flowers stationary, seasonal, patchy in distri- of flowers last visited in a dense array, or of flower clumps over wide areas and for long time periods bution 8. Wide-ranging; seasonal movements on basis 8. Spatial and temporal variations in blooming of flower abundance shifts 9. Breeding and/or molt closely tied to flower 9. Regularly recurring peak(s) of flowering each abundance; nectar a necessary source of en- ergy to meet expenses year, permitting birds to schedule their peak energy demands accordingly Tennessee Warbler {Vennivora percgrina), a frequent flower visitor, is sufficient- ly small to be considered moderately specialized in this respect (8-9 g). How- ever, its closest relatives in the genus Vennivora weigh scarcely more (8-11 g) but, like most of the family Parulidae, rarely or never visit flowers* Hence, the degree of actual specialization in size of V. peregrina is at best slight. Many of the features listed in Table 2 are straightforward and well known, others may require some explanation. Small size (nearly always under 20 g) is common to virtually all specialized avian nectarivores, but within this size range several selective factors can operate: larger size favors social dominance, but reduces the number of flower species that can be profitably exploited (cf. Lyon, 1976; Wolf et al., 1975); small size may favor utilization of torpor as an energy- saving device (Hainsworth et al., 1977; Brown et al., 1978). With regard to lo- comotion, hovering ability is probably the ultimate specialization in that it permits exploitation of fr^e-hanging blossoms, which give the greatest protection against insect visitation (Faegri & van der Pijl, 1966). However, for nonhovering species, the ability to cling acrobatically to flowers and inflorescences might be manifested in terms of larger feet than nonnectarivorous relatives, especially if these are 1981] STILES— BIRD-FLOWER COEVOLUTION 33 1 typical foliage-gleaning types. In the hummingbirds, where most species hover to feed, the ability to perch while feeding might reduce the costs of nectar ex- traction, especially for larger species (Wolf et al., 1972, 1975). I know of no small (<4 g) hummingbird that regularly clings to flowers to feed, but many medium- to-large species do, especially in the highlands (Wolf et al., 1976; Stiles, unpub- lished data). It is probably no coincidence that Eutoxeres spp., by far the largest hermit hummingbirds (10-13 vs. 2V2~1 g), are also the only ones to regularly cling to flowers to feed and have extraordinarily powerful feet. The nasal operculum is a fleshy flap that largely covers the nostrils in nectar- feeders; it presumably prevents nectar and pollen from clogging the nasal passages. It may serve as an index of nectarivory in some cases, such as the Coerebidae: it is highly developed in Diglossa and Coereba, which are highly nectarivorous; and but slightly developed in most other members of the family, who are but occasional, facultative nectarivores (see below). The bills of nectarivorous birds are important not only as indicators of flower- feeding per se; they also reflect the manner of nectar extraction (e.g., whether by piercing, mashing or probing), as well as the diversity and specificity of bird- flower coadaptations that may occur within a region or community — and indi- rectly, the relative age of the bird-flower association there (e.g., comparing North American vs. neotropical bird-flower communities: cf. Snow & Snow, 1972, 1980; Brown & Kodric-Brown, 1979). However, low bill-flower diversity may exist in a relatively old bird-flower association if specificity per se is not highly advan- tageous (see below and Paton & Ford, 1977). Specializations of the gut for nectar-feeding are relatively clear-cut, although in such groups as the Dicaeidae, adaptations for nectarivory may overlap with, or be subordinate to, those for frugivory (Docters van Leeuwen, 1954). Brush- tipped tongues occur in several groups that feed on fruit juices as well as (or instead of) nectar (e.g., the Zosteropidae and Coerebidae). No particular gross morphological specialization appears in the gut of the Loriinae: the nectar taken is mainly absorbed in the crop (present in parrots generally), and pollen digestion appears within the capacities of the usual psittacine gut apparatus (Churchill & Christensen, 1970; Forshaw, 1973), although enzymatic specializations may exist. I know of no features of plumage or integument that could be unequivocally interpreted as specializations for flower visitation: feathers in general have an ideal texture for pollen transport (Faegri & van der Pijl, 1966). The bright colors of some (but by no means all) nectar-feeding birds may have evolved in part as aggressive signals in relation to feeding territories, but there is little to suggest that they have evolved to match the flowers they feed on; if anything, the reverse may be true in some cases (e.g., Morton, 1979). It is precisely in evaluating behavioral specialization for flower feeding that one encounters the greatest difficulties: not only is there much variation between and among different groups, but critical data are scarce or lacking in many cases. The ecological characteristics of flowers as a food source — stationary, conspic- uous, renewable, more or less repeatable from one year to the next — allow some deductions about possible behavioral specializations. The aggressiveness and fre- quent feeding territoriality of nectar-feeding birds follow from these features (Gill, 1971; Stiles, 1973; Carpenter, 1978; Wolf & Wolf, 1976, etc.). By contrast, rela- 332 ANNALS OF THE MISSOURI BOTANICAL GARDEN IVoL. 68 lively few data are available on spatial memory of nectarivores, especially in relation to other birds (Gass, 1978). Flower-feeding does not seem to be correlated with promiscuous or polygynous mating systems, contrary to my earlier predic- tion based upon hummingbirds (Stiles, 1973). Rather, the kinds of social behavior (flocking vs. solitary, monogamous vs. promiscuous, etc.) may affect foraging patterns and thus selection for bird pollination in plants of different phenologies and growth habits (see below and Stiles, 1978b). The Major Groups of Flower-Visiting Birds Table 3 represents an attempt to compare semiquantitatively the principal groups of flower-feeding birds in terms of their degree of specialization for (and dependence upon) a nectar (or pollen) diet. I use a scale running from = no particular specialization, to 3 = relatively highly specialized, as compared to non- nectarivorous near relatives or putative ancestors (known or hypothetical). The results of this analysis are expressed in terms of an approximate mean degree of specialization for the group in question, and a corresponding figure for the most specialized species in each group. Obviously these figures are an oversimplifi- cation: the various criteria used (Table 2) are not necessarily equivalent (at least in any quantitative sense), and specialization in different ways can lead to a similar overall mean. Moreover, the species within each group can exhibit a wide range of specialization according to any given criterion. For many species (in- deed, for most species of many groups) published information is inadequate for an accurate assessment of specialization, and the value(s) presented represent simply my best guess, if anything, 1 have probably been too conservative in judging specialization according to certain criteria (e.g., annual cycles); more detailed study of the group in question might indicate rather a higher degree of specialization than I have assigned (at least for some species). Nevertheless, provided due caution is exercised, I think that these results are useful in com- paring the relative specializations of the different groups of nectar-feeding birds, at least to a first approximation. This in turn will facilitate comparison of the total spectra of nectarivorous birds on each continent or major biogeographical region, and the kind and degree of bird-flower coevolution likely within each. Several conclusions emerge from this analysis. First, by virtually every cri- terion the hummingbirds are the most specialized avian nectarivores. The most specialized hummingbirds are tightly tied to flowers in nearly every aspect of their biology and are often highly coevolved with a small number of flower species. Such species as Ensifera ensifera are totally dependent on specific flower species (in this case, Passiflora mixta: Snow & Snow, 1980) for critical energy supplies, just as the flowers require the bird for successful pollination. The ex- treme degree of bill-corolla exclusiveness in such cases is simply an indication of specialization in numerous other aspects of the biology of bird and plant. Similar but less extreme degrees of specialization are frequent in the group (e.g.. Stiles, 1973; Stiles & Wolf, 1979; Wolf & Stiles 1970). Some hummingbird species, although highly dependent upon flowers, obtain nectar almost exclusively by piercing corollas (e.g., Heliothryx harroti); thus their potential for forming co- evolutionary relationships with particular flower species is low. It is also worth 1981] STILES— BIRD-FLOWER COEVOLUTION 333 O C > I o a O E 00 c o o N a. V5 Q 03 < o 2 (1> c o a Q jd C 2'^ E-2 P y -"^ C/5 ■- C cd s^ .2 O I E o o O .^ c o s I O > a, 5 y cd O c/3 •- I c Cd cd C O < C/3 O &o 2 o T, r O ^.2 C I o o c o ^i I C/5 o rt Q ^H a c o TD -a J!:2 I 3 on >. OJ > c« a. '5 a D o on ^ OO <^J r^ r^ <"* 00 f^. •■ I ^— __ O (^ r*^ rj ^y^ m m r^j O r J 00 — ri V-) 00 nO r- r) I I r^ I I O "^ I r-i rn r^ r<-i m I I f^ I m I c« c« OJ O -J :d c O 00 c E E o X 00 cd p OJ h2 w^ f*^ 00 r^ rj m I r*^ m ri I ro f*^ r*^ m r^j I <^j cd 00 'n ,^ C to a> n I I I rj I I I r\ ^ ^ W r^. (N 00 00 OC '^ I I 4- ri ri ri f*"* r\ rr. m r*^ 1 1 1 1 1 rj — rM r^ r4 r^ I I r4 I I C o iS <^^ <^J -^r ON ON r-^ f*"* r^ (A I ^ ■w a. 1) ^ cd a. o :d :d c o X Z Q g N ^ N cd a> Q U rj r. + m I r^ m I r^i r^ I on c/5 L- oj cd o 6 c cd =3 :y: E < >2 3 ^ C^ OJ Q a. ^ •T3 -^ (N t3 :5 ^ 0. fin a cd N 2 a o o <1> "cd o E II * r C o ^1 M <« cd 2 etf 5-2 > x; cd >. op ^.-S II Cd (/5 c o s 3 00 E H cd % ^ C/D y5 c cd > cd cd o 00 cd tin ^ E •>< E cd ^ cd cd V5 a, 3 O 00 r^ cd •O Jd 3 *- cd B: -c ,^ E o cd cd 3 C .ii a> oc c •^ E o o c ^ cd Cu 00^ a E 00 l> IS cd cd Q cd c o cd > cd o E o o "3 c u 3 cd 2 C/5 00 U^ ON ON - f*-^ rj ..ON ON (^ "" ^ ^ -^ — Q (U ^^ CQ c Cd > ON as ^ cd O O D. a. 3 u^ o Cd ON ^ ^ ^ On O "" ON O ^ 'B a u OJ 3 3E c c O o 1> ON oo cd at fr ^ Cd O • — c a Q t/5 CQ »/^ ON O Cd (/I o o 'ob'^ 00 2 . ON 2; c/^ [X ^ — 00 ON ON ON \D U 00 ON 00 0) '^ cd Ci. 3 00.$^ 00 c 3 .:2 ^ =3 ii o o c 2 CL > .. cd c O \C ON ON — c o c i/5 Cd *r5 o m ^ r^ * Cd c o c ^ "^ r^ ON 3 < :/5 ON cC o Cd 2 V3 c a> 00 on . O C — o cd X TO Cd o ON Cu E OJ E cd H 334 ANNALS OF THE MISSOURI BOTANICAL GARDEN [Vol. 68 emphasizing that bill-corolla morphologies are useful, but hardly infallible guides to bird-flower coevolution. Extreme bill types need not always indicate highly exclusive relationships: consider those species of Heliconia (the '"pogonantha group" of Stiles, 1979) that show a rather high degree of morphological coad- aptation with the Sicklebills {Eutoxeres spp.), yet are often pollinated mostly by — and are often critical food resources for — hermits of the genus Phaethornis (Stiles & Wolf, 1979). Conversely, some hummingbirds of relatively generalized bill type (straight, ca. 20 mm long) can become quite highly coevolved with particular flowers by virtue of their ecological situation (e.g.. Stiles, 1973; Wolf et al., 1976). Thus a realistic evaluation of bird-flower coevolution often requires detailed ecological data, and these are available for relatively few groups. in most respects the sunbirds (Nectariniidae) qualify as the next most highly specialized group of avian nectarivores. The spate of recent studies on sunbirds by Wolf, Gill, and their coworkers in Africa have demonstrated that these birds are highly adapted for flower visitation in many aspects of their foraging and aggressive behavior (e.g.. Gill & Wolf, 1975, 1977, 1978; Wolf et al., 1975). Information on breeding and annual cycles is relatively sparse, but tends to indicate that at least some species approach the degree of specialization in many hummingbirds (cf. Skead, 1967; Wolf & Wolf, 1976, and included refer- ences). However, there appear to be few highly exclusive coevolutionary rela- tionships between specific sunbird species and flowers. Most sunbird-flowers can be exploited and pollinated by several sunbird species (Skead, 1967; Gill & Wolf, 1978). Possible exceptions include certain high montane ^uwhir d-Lohelia associ- ations (cf. Chapin, 1954) and the Arachnothera-Musa superba association in In- dia. In the latter instance, the very long-billed, dull-colored Arachnothera sun- birds evidently pollinate widely dispersed riparian clumps of Musa in almost exactly the same manner as hermit hummingbirds pollinate certain species of Heliconia (P. Davidar, pers. comm.; cf. Stiles, 1975, and below). Slightly less specialized overall are the honeyeaters (Meliphagidae), but within this group there is a wide range of variation, from species as specialized as most sunbirds (e.g., Promerops spp., Philydonyris novaehollandae: Broekhuysen, 1959; Paton & Ford, 1977) to a number of species that rarely or never visit flowers at all (Officer, 1964). It is thought that Australia was originally colonized by a slightly to moderately specialized nectarivore stock, which subsequently radiated to fill many nonnectarivore niches in a depauperate avifauna (Austin, 1961 ; Keast, 1976). A great range of adaptations also occur in the monophyletic Drepanididae of Hawaii, from highly specialized nectarivores to species adapted to a variety of totally different niches. In this case the ancestral form was probably a car- dueline finch, and the highly specialized nectarivores represent an end point, rather than a starting point of diversification. Unfortunately, the most specialized Hawaiian nectarivores {Drepanis spp.) are now extinct (cf. Amadon, 1950; Bald- win, 1952; Carpenter, 1976b, 1978; Raikow, 1976). The case of the Dicaeidae is exceptional, as many species have evidently formed a tight coevolutionary rela- tionship with certain mistletoes, involving not only pollination but also seed dis- persal; adaptations to the latter appear to have taken precedence over those to the former (Docters van Leeuwen, 1954). In no other case known to me are the pollinators of a plant also its regular dispersers. I98I] STILES— BIRD-FLOWER COEVOLUTION 335 The coerebids or honeycreepers are a polyphyletic group, some members of which are probably descended from emberizine finches (the genus Diglossa, the flower-piercers), some from the tanagers (Dacnis, Chlorophanes, and related genera), perhaps some from the wood-warblers; the Bananaquit (Coereba) is of uncertain affinities (Beecher, 1951b; Skutch, 1962; R. W. Storer, pers. comm.). The most highly nectarivorous members of the group {Diglossa, Coereba) are quite highly specialized and dependent upon floral nectar as an energy source, but they are essentially parasites on hummingbird-flower systems, usually {Coe- reba) or virtually always (Diglossa) piercing the corollas of the flowers they visit (e.g., Colwell, 1973; Colwell et al., 1974). Coereba is a legitimate pollinator of a few plant species (Feinsinger et al., 1979). The remainder of the Coerebidae are at best facultative