DELTA home

The genera of Leguminosae-Caesalpinioideae and Swartzieae

L. Watson and M.J. Dallwitz

This document is reproduced from: Watson, L. and Dallwitz, M.J. (1983). The genera of Leguminosae-Caesalpinioideae: anatomy, morphology, classification and keys. 95 pages. (Research School of Biological Sciences, Australian National University, Canberra.)



The Swartzieae are probably best referred to the subfamily Papilionoideae (see Cowan and Polhill 1981). However, it is appropriate to cover them here, given their long association with Caesalpinioideae in the taxonomic literature and bearing in mind that they may usefully be included together in keys. In the context of Caesalpinioideae, the Swartzieae form a coherent group showing more similarities with Caesalpinieae than with Detarieae.

The following classification of Caesalpinioideae summarises the results of studies conducted over a number of years (see Watson 1981). The groups have been derived and tested using the variety of numerical methods available in the TAXON package of programs at the CSIRO. Division of Computing Research, Canberra (Ross 1982). Since numerical taxonomy continues to be pursued in some quarters using hopelessly inadequate computer programs, it is worth emphasizing the necessity of employing suitable methods. The more obvious taxonomic assemblages, among caesalpinioid legumes and elsewhere, will tend to emerge no matter how crude the technique; but ability to handle directly the various kinds of real taxonomic data (numerical, binary, disordered multistate, ordered multistate), and to make proper allowances for missing data and logically inapplicable values, are obviously minimal requirements of any classificatory method offered for professional use (cf. Watson, Williams and Lance 1967). Of the mixed-data classificatory programs available under TAXON, by far the best for general taxonomic purposes is MACINF (formerly MULTBET), which is an agglomerative program employing a Shannon-type information statistic. The classification of Caesalpinioideae set out here has been synthesised anew with the aid of three MACINF analyses, applied to the latest (1982) version of the data and employing different maskings of characters and taxa to test the robustness of groups. Further improvements in the descriptive data, including corrections, addition of some new characters and inclusion of several extra taxa have in fact led to no significant changes in the overall pattern of generic relationships established by earlier analyses (cf. Watson 1981). Group descriptions have been prepared with the help of the program GCOM (formerly GROUPER).

This classification differs from the treatment of Caesalpinioideae given in Polhill and Raven (1981; prepared by R. M. Polhill, R. S. Cowan, J. E. Vidal, H. S. Irwin, R. C. Barneby, R. D. Wunderlin, K. Larsen and S. S. Larsen), in a number of respects. In particular, the Amherstieae are not recognised here as a tribe, and the lower level groupings are in general more inclusive. However, in most important respects the schemes are very similar, especially by contrast with classifications on offer until recently. The differences reflect differences of interpretation, probably stemming principally from differences over taxonomic intent.

The primary aim here, as elsewhere in Watson’s work, has been to prepare comparative descriptions by compilation and original observation, then detect patterns of variation in terms of correlations among character states, and display them in useful fashion in an appropriate classificatory framework. The value of taxonomic classification resides, in facilitating useful generalisations about the accumulated data; in offering the hope of reliably predicting the likely distributions of character states among taxa for which the data arc as yet incomplete; and (c) in contributing to the generation and testing of ideas about patterns of variation likely to be found in characters not yet contemplated systematically, notably by suggesting suitable samples for future work. The latter aspect is nicely illustrated in the present context by the interaction between our exploratory exercise in classifying from first principles (Pettigrew and Watson 1977) and Graham and Barker’s subsequent work on pollen morphology (1981).

From this standpoint, the retention of Detarieae and Amherstieae as separate tribes seems undesirable (cf. Watson 1981), and two informal groupings of the genera (Detarieae, subgroups 1 and 2) are presented instead. These certainly have some substance in terms of character trends, but comparisons of the alternative group descriptions with the generic descriptions of Daniella, Eperua and Eurypetalum (for example) will show that the distinction is fairly marginal. The two minor subsets of genera within subgroup 1 (la and lb) seem fairly clear, although the assignments of poorly-known genera are inevitably dubious. The minor subsets within subgroup 2 (2a, 2b and 2c) on the other hand, are very feeble indeed. No doubt they reflect to some extent real low-level similarity patterns among the genera, but they are presented here mainly to illustrate the intermingling of Amherstieae and Detarieae in a recent MACINF analysis. It appears that in this circle of affinity there may have been repetitive transference of function from perianth to bracteoles, and the Amherstieae do not seem to constitute a satisfactorily coherent group distinct from the other genera of Detarieae subgroup 2 at any hierarchical level. However, for ease of cross-reference with Cowan and Polhill’s (1981) classification into Detarieae and Amherstieae. their assignments are indicated in the character list (see ‘character’ 134, p. 6).

Subgroups 1–4 of the Caesalpinieae are barely detectable, but reflect a general pattern picked out in many analyses (cf. Watson 1981). There is the interesting problem, in this circle of affinity, of balancing data on rather ill-defined but widely quoted features of general morphology (degrees of zygomorphy, etc.) against less ambiguous but less widely recorded anatomical characters (druses, stomatal configurations, etc.). These groups are not random assortments of genera, but comparisons of generic descriptions (e.g., of Haematoxylum, Jacqueshuberia, Schizolobium, Stahlia} with the alternatives offered by groups 1 and 4 vividly illustrate the problems posed by taxa showing varying degrees of intermediacy, and by intra-taxon variability involving critical features. This is not the stuff of stable taxonomic classification, and there is no good case for formal nomenclatural recognition of such tenuous assemblages. Polhill and Vidal (1981) also opted for informal subgroups within the Caesalpinieae, and given the differences in classificatory approach, it is neither surprising nor particularly significant that their interpretation differs somewhat from this one.

This classification has been derived by phenetic methods, critically examined from a phenetic standpoint, and the groups of Caesalpinioideae are presented in a sequence which purports to display degrees of phenetic relatedness. There is no evidence to show that effective taxonomy depends upon introducing phylogenetic considerations (whether supposedly ‘analytical’, or frankly speculative) into the group-forming process; and there has been no phylogenetic input into the present operation, except in so far as acceptance of the fact of evolution may influence questions of comparative morphology and character definition. The valuable, predictive characteristics of taxonomic classifications apparently depend upon the extent to which groupings reflect evolutionary lineages. However, there is no evidence that taxonomic systems erected on the shifting sands of a priori phylogenetic speculation are ever convincingly informative about anything, least of all about patterns of evolutionary divergences. The generality of evolution was deduced by Lamarck and Darwin from the facts of (phenetic) classifications, and the (phenetic) classifications of George Bentham frequently prove superior, in all respects, to those of his phylogenetic- ally educated successors, despite the fact that he had less observational information to work with. The Caesalpinioideae are a case in point: cf. Bentham (1865), Watson and Pettigrew (1977). the system set out here, and the one adopted by Polhill and Raven (1981), by contrast with the classifications of Taubert (1894), Melchior (1956) and Hutchinson (1964). In fact, phenetic classification remains the only practicable method of proven worth available to plant taxonomists.

Some of the genera of Caesalpinioideae are still inadequately described, and the process of filling out the data and making corrections will continue indefinitely. Meanwhile, by contrast with most plant, groups of comparable size, the organised descriptive data now available for classifying this group are very extensive. There have been major advances recently regarding the quality and quantity of the descriptive information, and the descriptions on which this classifi­cation is based represent a laborious attempt to include as much as possible of the accessible comparative data. The descriptions themselves should provide abundant scope for testing alternative classificatory approaches. It will be interesting to see whether the burgeoning industry of ‘phylogenetic analysis’ will prove capable of making anything of them, assuming such techniques are applicable at all to real taxonomic problems as exemplified by these plants. The results of computational studies already conducted have not been accepted blindly: attempts at re-arranging the genera within Caesalpinieae and Detarieae, by intuition and using other people’s groups, have failed to reveal any stronger patterns that those shown here, and it seems fair to conclude that clear groups do not exist. However, the family Leguminosae needs overall reassessment in terms of comprehensive samples and properly comparative data. Such an operation could throw new light on the classification of genera currently referred to the subfamily Caesalpinioideae.

IIn the following group descriptions, a number in parentheses indicates the number of genera exhibiting a character state. This number may be followed by a slash and another number indicating the number of genera for which the character is recorded (and applicable). For example, (7/26) indicates that 7 genera exhibit the character state, out of the 26 genera for which data on the character are recorded.


Swartzieae (11 genera)

Trees (11 genera), shrubs (6 genera), rarely lianas (1). Leaves sometimes unifoliolate or simple (6 genera); otherwise pinnate (8 genera), then usually imparipinnate (7/8); the leaflets petiolulate, stipels sometimes present. Stomata paracytic (3 genera) or not (6 genera), abaxial epidermal walls straight; with (4) or without (5) transcurrent veins; druses absent from the mesophyll (9/9), phloem transfer cells absent. Wood storied (10/10). Inflorescences racemose, racemose-paniculate or flowers solitary. Flowers not pentamerous; calyx entire in bud, opening irregularly into segments or into regular lobes; corolla of 0–5 petals, sometimes with vestigial petals; where ascertainable, corolla usually imbricate-descending (papilionaceous; 6/7). Staminodes absent; stamens usually 10 or more, usually (10/11) equal, free or (rarely) cohering via hairs; anthers basifixed (7) or dorsifixed (5). Ovary usually stipitate (10/11); seeds sometimes arillate (3/8), with starch, no amyloid, radicle inflexed. Mainly South American (9 genera) and African (4 genera).

Aldina, Baphiopsis, Bocoa, Candolleodendron, Cordyla, Exostyles, Harleyodendron, Lecointea, Mildbraediodendron, Swartzia, Zollernia


I. Caesalpinieae sensu lato (49 genera)

Trees or shrubs (a few scandent), sometimes armed with prickles or spines. Leaves pinnate or (30 genera) bipinnate, with spiral phyllotaxy; often (19) with non-paracytic stomata, no hairs with expanded/embedded feet, nearly always without conspicuous phloem transfer cells in the minor veins, mesophyll hardly ever exhibiting secretory cavities. Stipules (where present and recorded) interpetiolar. Wood infrequently storied (7/26).

Inflorescences not cymose. Bracteoles often (35) lacking. Flowers often pentamerous through calyx, corolla and androecium. Calyx polysepalous to gamosepalous, regular to very zygomorphic, often (20) not imbricate; corolla present, actinomorphic to zygomorphic but rarely with vestigial petals. Androecial members usually 10, nearly always all free (48), anthers usually dorsifixed, dehiscing by lateral to introrse slits. Pollen often with a margocolpus (11/30), the tectum nearly always punctate or reticulate. Disc nearly always absent. Ovary usually (43) central and free.

Seeds endospermic (23) or not, amyloid absent. Cotyledons mostly of types 2 or 3. Basic chromosome numbers often 13 or 14 (16 genera), less often 12 (9 genera).

More genera in South America (29) than in Africa (21), and with stronger representation than Detarieae (tribe IV) in Madagascar.

Subgroup 1 (21 genera)

Trees or (often) shrubs, occasionally scandent, not infrequently armed. Leaves with rachides adaxially grooved or (more often) not; leaflets often many, those of pinnate leaves opposite or alternate, often sessile; abaxial epidermis usually scarcely staining with safranin, capitate glandular hairs sometimes present, stomata usually not paracytic; hardly ever exhibiting transcurrent veins, the minor veins often without sclerenchyma, druses nearly always present in the mesophyll. Floral tube tending to be relatively shorter than in subgroup 4 (usually polysepalous, the segments free to the rim of the hypanthium); calyx often not imbricate, usually covering the petals in bud. Calyx and corolla usually very irregular. Fruit a two-valved pod or (often) indehiscent, often winged.

Acrocarpus, Aprevalia, Bussea, Caesalpinia, Cercidium, Colvillea, Conzattia, Cordeauxia, Delonix. Hoffmanseggia, Lemuropisum, Lophocarpinia, Mezoneuron, Moldenhauera, Parkinsonia, Peltophorum, Pterolobium, Stenodrepanum, Stuhlmannia, Wagatea, Zuccagnia

Subgroup 2 (4 genera)

Intermediate between subgroups 1 and 4, but usually with short-shoots, sometimes two to several superimposed axillary buds (2/2), bipinnate leaves, small unisexual and usually non-pentamerous flowers; 3-6 sepals and petals, calyx not covering corolla in bud, 4-12 stamens, fruit often indehiscent or dehiscing atypically, seeds endospermic.

Arcoa, Gleditsia, Gymnocladus, Tetrapterocarpon

Subgroup 3 (1 genus)

Intermediate between subgroups 1 and 4, characterised by a peculiar gynoecium and fruit, and 2n = 20.


Subgroup 4 (23 genera)

Mainly trees, less often (6 genera) shrubs, unarmed (except sometimes in Haematoxylum and Arcoa). Leaves nearly always with adaxially grooved rachides; the leaflets often few, those of pinnate leaves generally opposite, petiolulate; abaxial epidermis nearly always clearly staining with safranin, without capitate trichomes, the stomata paracytic; the minor veins with sclerenchyma, sometimes with trans- current veins, druses nearly always absent from the mesophyll. Floral tube tending to be relatively longer than in subgroup 1, the calyx often gamosepalous above the rim of the hypanthium; calyx imbricate, often not covering the corolla in bud; calyx and corolla mostly regular to slightly irregular. Fruit a two-valved pod, less often indehiscent, sometimes a follicle; not winged.

Arapatiella, Batesia, Burkea, Campsiandra, Cenostigma, Chidlowia, Dimorphandra, Diptychandra, Erythrophloeum, Haematoxylum, Jacqueshuberia, Kaoue, Melanoxylon, Mora, Pachyelasma, Recordoxylon, Schizolobium, Selerolobium, Stachyothyrsus, Stahlia, Sympetalandra, Tachigalia, Vouacapoua

II. Cercideae (Bauhinieae) (11 genera)

Trees or shrubs, lianas or vines, with or without tendrils. Leaves bifoliolate with sessile leaflets, or simple to bilobed (often mucronate between the lobes), phyllotaxy spiral or distichous. Stomata usually paracytic (5/6), abaxial epidermal walls nearly always straight; the minor bundles accompanied by sclerenchyma, usually (7/8) having transcurrent veins; minor veins without phloem transfer cells; druses usually (8/8) present in the mesophyll. Wood storied (313).

Gamosepalous, the calyx open or closed apically before anthesis, entire, dentate, spathaceous or valvately 2–5 lobed. Hypanthium present or absent. Corolla slightly to very zygomorphic (absent in Brenierea). Androecium parts 10 (rarely fewer), usually unequal (9/10), staminodes present (6/11) or absent, the filaments partially connate or not, anthers versatile. Ovary stipitate, adnate to hypanthium or free.

Seeds usually (8/10) endospermic, without amyloid, often (8/10) arillate.

Basic chromosome numbers 13 (2 genera), 14 (5 genera), and 7 (1 genus).

Taxa best included in Bauhinia (cf. Wunderlin, Larsen and Larsen 1981) in parentheses:

Adenolobus, (Barklya), Bauhinia, (Bracteolanthus), Brenierea, Cercis, Griffonia, (Lasiobema, Lysiphyllum, Phanera, Piliostigma)

III. Cassieae (20 genera)

Trees, shrubs (occasionally climbing) or herbs. Leaves usually imparipinnate (15 genera), sometimes unifoliate (6), rarely phyllodes; phyllotaxy spiral or distichous; petiolules not twisted; abaxial epidermis often scarcely staining with safranin (9/15), stomata usually paracytic (10/14), hooked trichomes often (12/14) and epidermal crystals sometimes (6/16) present, hairs with expanded/embedded feet absent, adaxial interveinal epidermal walls usually straight. Mesophyll without secretory cavities; druses sometimes present (4 genera), usually absent; the minor veins accompanied by fibres (15/15), transcurrent veins often present (10/14), phloem of minor veins with or without conspicuous transfer cells. Wood often storied (7/9).

Inflorescence often cymose (16), bracteoles present. Flowers not pentamerous through calyx, corolla and androecium; calyx imbricate, hypanthium absent or very short; androecium parts usually fewer than 10 (17 genera), rarely more (2 genera), staminodes present; filaments nearly always attached at base of connective (and anthers non-versatile), anther dehiscence often via basal or apical pores (14 genera); intrastaminal disc sometimes present.

Seeds usually non-endospermic (8/10), without amyloid (10). Basic chromosome numbers 4 (1 genus), 12 (2 genera) and 14 (5 genera).

Androcalymma, Apuleia, Baudouinia, Cassia, Ceratonia, Dansera, Dialium, Dicorynia, Distemonanthus, Duparquetia, Eligmocarpus, Kalappia, Koompassia, Labichea, Martiodendron, Mendoravia, Petalostylis, Storckielia, Uittienia, Zenia

IV. Detarieae (including Amherstieae) (81 genera)

Unarmed trees and shrubs. Leaves pinnate or (in 22 genera) bi-foliate, often (24) with twisted petiolules, stipules when present (and recorded) intrapetiolar, phyllotaxy spiral or distichous; abaxial epidermis nearly always (69) readily staining with safranin, often having hairs with expanded/embedded feet (25), not infrequently with epidermal crystals (17); nearly always having conspicuous phloem transfer cells (57/61), sometimes with unaligned mesophyll fibres and/or sclereids (12), often with secretory mesophyll cavities (gland-dotted). Wood sometimes with intercellular canals, sometimes storied (13/35).

Bracteoles nearly always present, small, imbricate and caducous to large, valvate and showy. Flowers sometimes distichous (17), usually (75) not pentamerous throughout calyx, corolla and androecium; calyx usually imbricate, sometimes very reduced; corolla usually somewhat to very zygomorphic, often with vestigial petals (32/65), sometimes absent; stamens 2-many, often (45) synandrous, anther dehiscence latrorse (11/11); pollen tectum striate (24), reticulate (18) or punctate (16); ovary usually excentric/adnate (51), but often central/free (29).

Seeds not endospermic (29/29); amyloid nearly always present (43/48). Cotyledons mainly of types 3–4.

Basic chromosome number usually 12 (21/27).

More genera in Africa (53) than in South America, and seemingly less well represented in Madagascar than are the Caesalpinieae (tribe I)

Subgroup 1 (22 genera)

Trees. The leaflets often alternate (12/18), often with a continuous marginal nerve (S/11); the mesophyll usually gland-dotted (16/18) with epithelium-lined secretory cavities (15/16). Wood usually with secretory canals (13/16). Bracteoles small, not valvate. Flowers mostly small, hypanthium often absent (11/20); corolla often (13 genera) absent. Pollen tectum punctate (8) or reticulate (5); ovary often sessile (10/20), usually central and completely free of adnation (16/20). Stamens usually 10 (19 genera), no staminodes. Fruits mostly indehiscent (17/22), less often a 2-valved pod (usually without twisting valves); seed sometimes (5/17) without amyloid. Basic chromosome numbers other than 12 in 5/10 genera.

1a (12 genera). Leaves bifoliolate (7 genera) or with few leaflets; epidermal crystals rare (2/10). Wood with scattered gum ducts. Flowers spirally disposed, calyx imbricate, corolla usually absent (9/12); stamen number sometimes other than 10 (5/12). Germination hypogeal in 2/6 genera.

Augouardia, Colophospermum, Gossweilerodendron, Guibourtia, Hardwickea, Hymenaea, Kingiodendron, Oxystigma, Peltogyne, Prioria, Stemonocoleus, Trachylobium

1b (10 genera). Leaves normally not bifoliolate, with few to (6 genera) many leaflets; epidermal crystals common (5/7). Wood with gum ducts in tangential lines. Flowers distichous and calyx subvalvate (in all 8 well known genera), corolla variously reduced but present in 7/10 genera; stamens nearly always 10. Germination epigeal (8/8).

Apaloxylon, Baikiaea, Bathiaea, Copaifera, Detarium, Gilletiodendron. Hylodendron, Sindora, Sindoropsis, Tessmannia

Subgroup 2 (59 genera)

Trees and (in about 20 genera) shrubs. The leaflets nearly always opposite (47/50), usually without a continuous marginal nerve (where recorded: 19/21), usually without mesophyll secretory cavities (gland-dots absent). Wood nearly always without secretory canals. Bracteoles large and enclosing the flowers (34) to small, valvate (24) or not, flowers often large and showy; hypanthium present (51/54); corolla early always present; staminodes often present (about 20 genera); pollen tectum usually striate (24) or reticulate (13), rarely punctate). Ovary nearly always stipitate (53/55), usually excentric/adnate (6/54); fruit usually a 2-valved pod (48/50), usually with twisting/enrolling valves (26/37), rarely indehiscent (6/50); seeds with amyloid. Basic chromosome number nearly always 12 (15/16).

2a (31 genera). Afzelia, Brachystegia, Brodriguesia, Cryptosepalum, Cynometra, Daniella, Endertia, Eperua, Eurypetalum, Gilbertiodendron, Goniorrachis, Hymenostegia, Intsia, Julbernardia, Lebruniodendron, Leonardoxa, Leucostegane, Loesenera, Macrolobium, Maniltoa, Neochevalierodendron, Oddoniodendron, Paramacrolobium, Plagiosiphon, Saraca, Schizoscyphus, Schotia, Scorodophloeus, Talbotiella, Tamarindus, Zenkerella

2b (6 genera). Aphanocalyx, Michelsonia, Microberlinia, Monopetalanthus, Pellegriniodendron, Tetraberlinia

2c (22 genera). Amherstia, Anthonotha, Berlinia, Brachycylix, Brownea, Browneopsis, Crudia, Dicymbe, Didelotia, Elizabetha, Englerodendron, Heterostemon, Humboldtia, Isoberlinia, Librevillea, Lysidice, Paloue, Paloveopsis, Phyllocarpus, Polystemonanthus, Pseudomacrolobium, Thylacanthus

V. Unclassified (2 genera)

Umtiza. Anatomical data required.

Poeppigea. Showing a number of similarities with Cassieae, including distichous phyllotaxy, leaves imparipinnate, leaflets opposite, with phloem transfer cells, adaxial epidermal walls straight, abaxial epidermis scarcely staining with safranin; corolla ‘pseudo-papilionaceous’; conspicuous disc present; no amyloid; but 10 fertile stamens and no staminodes, campanulate hypanthium, etc.