A TAXONOMIC REVISION OF THE GENUS CEIBA MILL. (BOMBACACEAE)

GIBBS, P. & J. SEMIR (2003). A taxonomic revision of the genus Ceiba Mill. (Bombacaceae). Anales Jard Bot. Madrid 60(2): 259-300. In this taxonomic revision of Ceiba Mill, (in which we include Chorisia Kunth) we recognize 17 species, seven of which we group in the C. insignis species aggregate. One new species is described, C. lupuna P.E. Gibbs & Semir from Peru, which is referred to the C. insignis agg., and one new subspecies, C. aesculifolia subsp. parvifolia (Rose) P.E. Gibbs & Semir is recognized. Distribution maps are provided for 16 species (that for C. pentandra is restricted to the New World), and six species are illustrated.


INTRODUCTION
perimental Station of the Institute Agronomico (Fazenda Santa Eliza), Campinas (Sao This revision of the genus Ceiba represents Paulo). We did not solve this particular probthe tardy completion of taxonomic studies we lem until much later but sporadic taxonomic initiated with the genus Chorisia in the early studies with herbarium material and species 1980s. At that time our main research interest represented in the splendid collection of trees was in the breeding system of Chorisia spe-of known provenance at the Fazenda Santa ciosa and related species, and our taxonomic Eliza led to a gradual understanding of this enquiry began because of doubts we had con-group, and our conviction that the genus Chocerning the identity of pale yellow flowered risia should be merged with Ceiba (cf. GIBBS trees listed as "Chorisia insignis HBK" in cul -& al., 1988). However, our taxonomic retivation at the horticultural section of the Ex-search with Ceiba went into abeyance for a number of years whilst the first author was involved with studies on the reproductive biology of species of the Brazilian cerrado and Argentine Chaco vegetation, and second author undertook a doctoral thesis on the genus Lychnophora Mart. (Asteraceae).
Fortunately, over the intervening years, with one notable exception, few other people have taken an interest in the taxonomy of species of Ceiba or Chorisia, despite the great beauty of the flowers of these trees, and some confusion concerning their identity. The exception was a paper by RAVENNA (1998). The principle objective of RAVENNA (1998) was to accept Gibbs, Semir and da Cruz's view that Chorisia should be merged with Ceiba, and then to effect all possible recombinations of Chorisa species under Ceiba. No attempt was made to study type material of the species of Chorisia to be recombined, and unsurprisingly, of the seven new combinations effected in mis paper, two were superfluous. In addition, two new species were proposed by Ravenna (both rejected in the present revision), together with an unworkable key which purported to identify five of the 10 species considered in his study. Why the key was restricted to just these five taxa was not explained. The paper was produced in the author's own xeroxed series 'Onira Botanical Leaflets' rather than published in a peer-reviewed journal, and so is difficult to access, but since care was taken to ensure that copies were sent to some major botanical institutions, e.g. Royal Botanic Gardens, Kew, effective publication was achieved. RAVENNA (1988) claimed that GIBBS & AL. (1988) "believed that Ch. crispiflora HBK, Ch. speciosa St. Hil.. Juss. & Camb., Ch. ventricosa Nees & Mart., Ch. integrifolia Ulbr. and Ch. incana Rob. are referable to Ch. insignis. The writer's own experience with living material of these and other species in their native habit, cannot allow him to accept Gibbs & al. opinion". This implies we treated all of these taxa as synonyms of Chorisia insignis, which in this paper we recombined as Ceiba insignis. In fact we commented that these species "form a polymorphic complex or aggregate species... Our detailed studies are not yet complete but it appears that all other species of Chorisia... are referable to it" [i.e. to such a species aggregate]. Mindful of the adage "fools rush in where angels fear to tread", we concluded: "Whether any of these taxa can be maintained at a specific or subspecific level must await the outcome of our analysis of the Ceiba insignis complex." In the event, we recognize here the Ceiba insignis aggregate with seven component species, including a new species C. lupuna, and we treat Ch. integrifolia as a synonym of C. insignis s.s., and Ch. incana as a synonym of C. ventricosa. GIBBS & al. (1988) provided a taxonomic history of the genera Ceiba and Chorisia, and a discussion of their alleged differential characters, particularly the nature of the stamen tube, which led us to believe that Chorisia should not be separated from Ceiba. To provide an introduction to this revision, we here include a summary of the views put forward in that earlier publication.
Additional generic names were subsequently proposed by diverse authors but none received widespread acceptance, and the modern view of Ceiba and related genera was cast by SCHUMANN in his treatment of the group for Martius's Flora brasiliensis (1886) and subsequently his monograph in Engler and Prantl's Pflanzenfamilien (1890). In the tribe Adansoniae Benth., Schumann recognized four genera: Adansonia, Bombax, Chorisia and Ceiba. The broadly based Bombax recognized here was subsequently split by various authors, most notably by ROBYNS (1963). SCHUMANN (1890) recognized Ceiba with three sections: sect. Campylanthera with four species from tropical America including C. samauma and C. pentandra, sect. Eriodendron, with one very different species, C. rivieri (now at Spirothera), and sect. Erione, again with a disparate group of taxa including C. rosea from C America, and C. pubiflora and C. jasminodora from Brazil. Except for recognition of the genus Spirotheca by ULBRICH (1914), which involved the transfer of Ceiba rivieiri to this new genus, Schumann's view of the genera Ceiba and Chorisia has remained essentially intact until the proposal to unite these genera by GIBBS & al. (1988).
Our reasons for this proposal are discussed in detail in the paper cited above. Essentially they are based on the fact that the diagnostic character between Chorisia vs. Ceiba, viz. that species of the former have a lower stami-nal tube with a corona-like structure of staminal appendages, and with the upper staminal filaments fused to form a tube, whilst species of Ceiba, lack such appendages and divide to give 5 filaments, in fact breaks down with intermediate conditions. Strictly applied, the distinction between Chorisia and Ceiba leads to such species as Chorisia speciosa and the closely related (and interfertile) Ceiba pubiflora, and similarly, Chorisia glaziovii and the closely related Ceiba erianthos, being placed in separate genera, as DAWSON (1944) and SANTOS (1964) noted. We have also produced fertile hybrids between Ceiba erianthos X Chorisia speciosa and C. erianthos X Chorisia chodatii. Again as discussed by GIBBS & al. (1988), pollen morphology divides these taxa not according to Chorisia vs. Ceiba, but a large group of both these genera, with essentially peritreme grains, are distinct from some four or so species with distinctly oblate grains. We used this striking pollen difference to recognize two sections with the united genus Ceiba.
The taxonomy of Ceiba presents some special problems. Most species flower in the leafless condition so that many specimens are either of flowers or leaves. Many Ceibas are large forest trees with flowers of difficult access, so that often specimens have been prepared from old fallen flowers collected on the forest floor. Moreover, since many species have crepuscular or nocturnal anthesis, so that flowers collected after midday the following day are already entering senescence. Add to which most species have fleshy flowers with delicate colour variations which, if not dried carefully, e.g. using aluminum corrugates and heated presses, lose their colour and partially their form. These comments are not made to excuse poor taxonomy, but rather to plead for careful future collecting with this group, and to stress the value of colour photos, and perhaps to justify the rather broad approach we have taken with several species, e.g. C. aesculifolia, and C. pentandra. It will be surprising if our treatment will represent the last word on these taxa. This revision is based on herbarium specimens from the following herbaria: BAB,  BAF, BHCB, BRAD, CAS, CEN, CEPEC,  COL, CPAP, DS, E, EAC, ESA, F, FHO, G,  GUA, HUEFS, HXBH, INPA, IPA, K, LIL,  LL, MBM, MO, NY, OXF P, R, RB, S, SJRP,  SP, SPF, TEX, U, UEC, US, WIS, although we have not necessarily seen all material in any particular herbarium. Where possible we have also studied species either in the field (C. glaziovii, C. jasminodora, C. pubiflora, C. speciosa), or in arboreta (C. erianthos, C. samauma).

Growth habit
All ceibas are trees, commonly 5-20 m but in some species with imposing height, e.g. Ceiba pentandra, at 30-50 m, is a canopy emergent of the varzea forest in Amazonian Brazil and riverine forest in W Africa. This species may also present well developed buttresses. At the other extreme, C. jasminodora is often encountered as a treelet of 1.5-2 m in upland, rocky campo rupestre habitats in Minas Gerais. Most ceibas present aculeate spines on trunk and branches, and in some species (particularly C. chodatii, C. pubiflora, C. speciosa of the C. insignis agg.) the trunk may be markedly ventricose (hence the common names barriguda, palo borracho) but individuals of the same species may vary with respect to spines and trunk form.

Leaves
All species have digitate leaves, commonly 5-7 foliolate, usually with a long, slender petiole and diminuitive petiolules, commonly more or less lanceolate, chartaceous to coriaceous, with margin entire or dentate. Serration of the margin may be variable within species, but is consistent in e.g. the aptly named C. aesculifolia, and consistently absent in e.g. C. erianthos, C. jasminodora, C. samauma and C. schottii. Again, C. jas-minodora is distinctive in having 3-foliolate leaves, usually with a distinct mucron. Indumentum, when present, of simple or stellate hairs, sometimes varying between individuals of the same species.

Calyx
Usually more or less campanulate, robust to fleshy, usually glabrous externally (with few exceptions, particularly some specimens of C. aesculifolia, and C. soluta) and densely villous within, opening to give 4-5 irregular lobes. Dispersed nectaries are distributed on the inner surface of the calyx.

Corolla
Commonly with 5 rather fleshy, more or less linear-oblong to spathulate petals which may be held erect to give a funnelform (C. chodatii) or more or less campanulate (C. boliviano) flower, but in most species with petals spreading, and in some e.g. C. jasminodora, C. aesculifolia, markedly reflexed.
Petal colour and markings are important characters in this genus, although usually lost with herbarium specimens (for which, in many cases, the flowers have been collected after having fallen to the ground), and sparingly commented upon by collectors, making colour photographs invaluable for understanding the species. A common feature is for the petal limb, whether white, ivory or pink to magenta in colour, to have distally sparse dark, usually deep carmine, flecks or striations, which may merge below, but with the claw pale yellow. Thus, C. boliviano has strikingly profuse and anastamosing carmine striations on a white-pink petal background, whilst C. pubiflora is very variable in this respect, ranging from delicate pale pink petals with few flecks, to dense pink-lilac with many dark striations which may merge below. The ivory-pale yellow petals of C. chodatii present very few flecks. The white petals of C. glaziovii present variable carmine flecks or striations, whilst the whitish petals of C. erianthos have a conspicuous central carmine zone. The petals of our new species C. lupuna are a uniformly dense red, with the usual yel-low claw, whilst those of the morphologically rather similar C. speciosa and C. crispiflora have magenta petals with sparse carmine flecks or striations. In old or badly pressed specimens, Ceiba flowers of diverse species appear yellowish brown with dark brown markings, and the yellow claw becomes white.
In most species, the petals are shortly sericeous pubescent externally and mostly glabrous within, although the distal inner surface may also have hairs, especially towards the margins. In the species of sect. Campylanthera (the C. aesculifolia complex, C. schottii) this external pubescence has a tan colour, whilst in C. erianthos and C. boliviano the petals are notably white villous-tomentose externally.

Androecium
The androecium in Ceiba species provides important characters. Superficially, in most species, seemingly five stamens are united by their filaments to form a lower tube, which surrounds the pistil, and which then splits to give five separate filaments and anthers. The situation is structurally much more complex, and reflects the amazing diversity of stamen filament-anther fission and fusion in the Malvales as partially documented in the study by VAN HEEL (1966).
Based on histological sections and cleared tissues (clearing the androecium with concentrated KOH-fuchsin solution, which provides translucent tissues with deeply staining veins, was particularly helpful) our interpretation of the androecium in Ceiba species is as follows. Two whorls with (internally) 5 and (externally) 10 filaments are present. In the lower staminal tube which is present in all species, these are represented by 15 vascular traces in five groups of three. In species of the C. insignis agg., which possess a distinct "coronalike" whorl of five bifid, staminal appendages, each appendage receives a curved deviation of the outer two vascular traces, which then rejoin the central traces to ascend the upper staminal tube, and enter the collar of "five" seemingly bithecate anthers. These ac-tually derive from 15 monothecal anthers, each with the interior thecum reduced to connective tissue. In species which have five free stamen filaments above the level of the staminal appendages (or the vestiges of these), each filament presents three fused vascular traces, and an "anther" with two functional thecae, again with the central thecum reduced to connective tissue. In C. trischistandra, the triple nature of the staminal filaments is revealed as each "filament" finally divides to furnish three monothecate anthers. In C. soluta such fission extends to the level of the staminal "corona" to give 15 staminal filaments, although the actual number seems to be variable in the few specimens available.
In C. glaziovii the staminal appendages are not bifid, whilst in C. erianthos they are represented merely by a hairy swelling, as SAN-TOS (1964) noted and illustrated. In other species of sect. Ceiba, the staminal appendages may form a disc-like structure (C.jasminodora), or a truncate ledge (C. pentandra). In species of sect. Campylanthera, the staminal appendages in C. aesculifolia and C. soluta lack a vascular loop, and are absent (or represented by an "articulation") in C. schottii.

Fruit and seeds
In all species the fruit is a rotund to ellipsoidal, 5-valvate capsule, with a mostly smooth exterior, and in which the endocarp develops into a white cotton-fibered mass (hence common names "painera", "pochote") which surrounds the many seeds. When the capsule valves fall away, this cottony kapok aids in the wind dispersal of the entangled seeds. Seeds are round to pyriform to reniform, usually largish, 5-10 mm, with testa dark brown to black, with matt to smooth surface. CYTOLOGY BAKER & BAKER (1968) reported chromosome numbers for diverse genera of the Bombacaceae, including Chorisia speciosa (= Ceiba speciosa) with 2n = 72, and 26 counts for C. pentandra s.l. of 2n = 72-88, from provenances in Jamaica, Guyana, W Africa and Indonesia. All counts were with root tip material, either from acetocarmine squash preparations, or haematoxylin stained sections. Da Cruz (GIBBS & ai, 1988), using aceto-orcein squash preparations of anther material, established chromosome numbers of n = 43 for Ceiba insignis, C. erianthos, C. glaziovii, C. jasminodora, C. pentandra, C. pubiflora and C. speciosa. Interestingly, da Cruz noted that root tips of germinating seeds of C. speciosa gave counts of In = 69-87 showing that aneusomatic divisions may occur, and perhaps explaining the cytological variation reported by BAKER & BAKER (1968). These consistent counts of n = 43 for Ceiba species, contrasted with counts of n = 46 for Pachira aquatica Aubl., n = 44 for Bombax longiflorum (Mart. & Zucc.) K. Schum. and n = 46 for Spirotheca passifloroides Cuatrec.

REPRODUCTIVE BIOLOGY
Most Ceiba species have nocturnal anthesis and C. pentandra (GRIBEL & ai, 1999) and C. erianthos (observations by J. Semir) are known to be bat-pollinated. This also likely to be the case in C. glaziovii, C. boliviano, C. insignis, C. samauma, C. ventricosa, and C. trischistandra, but field observations are lacking. C. erianthos and C. glaziovii have particularly copious nectar. C. chodatii also has crepuscular anthesis but with sparse nectar, and is probably pollinated by sphingid moths (we have seen sphingids visiting flowers of cultivated trees in Rosario, Argentina), as may be the case with C. jasminodora and C. schottii. C. speciosa (and probably C. crispiflora) is pollinated by diurnal butterflies which beat their wings against the anthers or stigma as they probe between the staminal appendages for nectar. Hummingbirds are frequent but ineffective visitors to this species since they do not touch the anthers/ stigma. However, C. pubiflora, with its resupinate anthers and somewhat more copious nectar, seems to be adapted for pollination by humming birds, which are certainly frequent visitors to its flowers. It is notable that the corona-like staminal appendages, when they occur, serve to restrict access to the nectar. Thus they are present in all moth, butterfly or hummingbird pollinated species, but absent or reduced in most, but not all, bat or potentially bat-pollinated species.
The Bombacaceae present a cluster of taxa [see GIBBS & BIANCHI (1999) for review] with late-acting self-incompatibility (LSI) sensu SEAVEY & BAWA (1986). In this system, selfed flowers uniformly fail to form fruits although self pollen tubes grow to the ovary and penetrate ovules. In the genus Ceiba, LSI has been studied in the species C. chodatii and C. speciosa (GIBBS & BIANCHI, 1993) and C. pentandra (GRIBEL & ai, 1999). HABITAT Most species of Ceiba are restricted to seasonally dry woodlands, including C. erianthos, which occurs in the coastal restinga of SE Brazil, but almost always associated with rocky outcrops, as is C. jasminodora in the "campos rupestres" of Minas Gerais. C. pubiflora seems to have a marked but not exclusive preference for calcareous soils in its distribution from Mato Grosso Sul to the caatingas of northern Minas Gerais and Bahia. However, some species, such as C. samauma and C. speciosa, seem to be able to occupy both dry seasonal forest and humid river valleys, whilst C. pentandra shows even greater tolerance, occurring in seasonally flooded lowland forest in Amazonia, but also in mesic habitats in Central America. This same tolerance seems to apply to C. pentandra in West Africa, where BAKER (1965) reported it to be rare in evergreen rain forest but very common in moist, semideciduous forests, and commented: "it is a common constituent of the gallery forests which line the river banks as these lead out from the true forest into the savanna woodlands and the true savannas". As far as we are aware, only C. lupuna occurs exlusively in humid forest.  NICOLSON, 1979).
Trees, in some species of considerable size, and sometimes with a ventricose trunk, both trunk and branches often with stout aculeate spines. Leaves alternate, compound-digitate, with a longish petiole, with 3-5(8) leaflets; leaflets elliptic, lanceolate, or oblanceolate, serrate or entire, usually attenuate, acute to acuminate, rarely obtuse; both surfaces usually glabrous, occasionally sparsely simple or stellate hairy. Inflorescences in few-flowered fascicles or flowers solitary, with very caducous bracts and bracteoles. Flowers 18-145 mm, with 5 usually spreading or sometimes funnelform or campanulate petals. Pedicels usually stout. Calyx thick-fleshy, campanulate, opening irregularly to give 3-5 lobes, externally usually glabrous, internally densely villous-pubescent, with dispersed nectaries over the internal surface. Petals connate to the staminal tube at the base, usually oblong-spathulate, with entire or undulate margin, usually mostly glabrous internally, sparsely to densely whitish to brownish sericeous externally, white to pinkish to magenta or red, usually with the claw yellowish, limb with inner surface frequently blotched with carmine striations. Staminal filaments fused to form a tube around the ovary, sometimes with a corona-like whorl of staminal appendages. Staminal tube usually dividing to give five free filaments; but in some species these filaments remain fused above the staminal appendages to give an upper staminal tube (insignis agg.), whilst in two species, the five initial filaments split to give three filaments terminating in monothecate anthers: in C. trichistandra this division occurs in the upper part of the common filament, whereas in C. soluta it occurs just above the staminal appendages. The ovary superior, usually pyriform, five loculate-carpellate, with axile placentation and many ovules. Style usually slender, white and glabrous (hairy in C. samauma), terminating in a globose somewhat lobed stigma, which may be white to red. Fruit an elongate capsule with the seeds embedded in dense cotton-downy fibres originating from the endocarp. Seeds large, 5-10 mm, roundish to pyriform to reniform, dark brownish-black.
Three species occur in Mexico and Central America, and 13 species are distributed in South America. C. pentandra is the only species which extends outwith South-Central America and the Caribbean Islands, occurring in W Africa, where it is probably native, and also in India, SE Asia and the Pacific, to which areas it was most likely introduced by man.
flowering 9 -C -glaziovU We group the following sequence of seven species in an aggregate species sensu HEY-  (6): 63 a number of species (binomials). The compo-(1890) nent binomials are in taxonomic terms close-ly related and difficult to discriminate. Their distinguishing characters, although less pronounced and perhaps fewer in number than those which serve to distinguish between other species, are, however, constant and the species appear to be effectively isolated from one another". With one exception, the component species of the C. insignis agg. are characterized by the presence of an entire staminal tube which terminates in a collar of anthers.
Species of this aggregate extend in a more or less U-shaped arc of semi-deciduous vegetation from NE-SE-SW Brazil, Paraguay, the Argentine piedmont area as far south as Tucuman, Bolivia, and NW Peru to S Ecuador. This kind of distribution parallels that of the "Pleistocenic Arc" of seasonal woodlands as defined by PRADO & GIBBS (1993), PENNING-TON & al. (2000), but with the complication that whilst most of the species of the C. insignis agg., e.g. C. insignis s.s., C. chodatii, C. incana and C. pubiflora, and C. ventricosa are certainly resticted to seasonally dry habitats, the widely distributed C. speciosa and also C. lupuna, can be found in moist riverine forest in Peru and in W Brazil (Acre, Rondonia).
One species referred to this aggregate, C. pubiflora, has free stamens. This species occurs from Paraguay to centre-west-NE Brazil (Mato Grosso, Goias to Minas Gerais and Bahia), especially on calcareous soils. Ceiba pubiflora shares the prominent coronalike staminal appendages of the C. insignis agg. but either immediately, or some 5-10 mm above the corona, five separate staminal filaments diverge. It is likely that all members of the C. insignis aggregate are interfertile. Certainly C. speciosa x C. chodatii form fertile hybrids, some of which are commonly cultivated as street trees in Argentina, and C. pubiflora x C. speciosa can also cross. Moreover, occasional specimens of C. speciosa and C. chodatii are encountered with flowers with the upper staminal tube partially split to give five filaments with separate anthers. However, it is notable that the component species diverge in timing of anthesis, and as a consequence, pollinator type. 1. Stamens with 5 free, usually resupinate filaments arising directly from, or some 5-10 mm above, the staminal appendages 7. C. pubiflora -Stamens above the staminal appendages united into a tube which has a collar of 5 sinuous, anthers (occasionally this tube with some fission distally and so anthers free) 2 2. Petals ivory to pale yellow, whitish, or white suffused pink 3 -Petals deep pink-magenta, or red, at least distally 5 3. Rower rather funnel-form with ivory-pale yellow petals held erect, staminal appendages whitish, glabrous 6. C. chodatii -Flower rather stellate with the white to whitepinkish petals spreading, staminal appendages usually hairy 4 4. Petals narrowly spathulate or oblong, with only slightly undulate margin, white to suffused pink distally, yellowish towards the base; staminal tube below the staminal appendages pale and glabrous, staminal appendages white or reddish, glabrous to sparsely hairy... 1. C. insignis -Petals narrowly oblong with markedly undulate-crespate margin, whitish with dark reddish flecks distally, these merging towards the base; lower staminal tube reddish and cinerous sericeous, staminal appendages dark red, densely hairy 5. C. ventricosa 5. Petals uniformly deep red distally, pale yellow at the base 2. C. lupuna -Petals deep pink-magenta distally, with many dark red striations mid-length, pale yellow towards the base 6 6. Staminal tube below the staminal appendages glabrous; petals rather spathulate or broadly oblong, usually more than 15 mm wide, margin only slightly undulate 3. C. speciosa -Staminal tube below the staminal appendages sericeous; petals narrowly oblong, usually less than 15 mm wide, with markedly undulate-crespate margin 4. C. crispiflora This species has a problematical history, not least because in 1900 Hicken identified trees of C. chodatii cultivated at La Recoleta in Buenos Aires as Chorisia insignis HBK, and largely as a consequence, the name C. insignis has been widely given to specimens collected in Argentina, Bolivia and Peru as well as to cultivated trees, giving rise to considerable taxonomic confusion. We here identify C. insignis s.s. with a white-pinkish flowered species of dry, deciduous woodland in NW Peru-Ecuador. We base our view on the following considerations.
In the Humboldt and Bonpland collections at P there is only one specimen referable to C. insignis which surprisingly does not have a locality but simply bears (apparently in Kunth's hand by comparison with examples in BURDET, 1976, p. 145) the name "Chorisia insignis". It is very likely that this specimen in Paris, which has the flower dissected with calyx, staminal tube and individual petals displayed, as well as leaves, provided the basis for the illustration of Chorisia insignis of the protologue, and we therefore have no hesitation in choosing it as the lectotype of C. insignis. However, from the reference to a fruit in his description, Kunth evidently also had access to either additional specimens, or to Bonpland's notes on this taxon.
Based on specimens from the the Chamaya-Tomependa area, C. insignis is a species of semi-deciduous woodlands in dry valleys which has whitish-pink flushed flowers. SCHUMANN (1886), in his description of Chorisia insignis in Martius' Flora Brasiliensis, repeated the Chamaya and Tomependa localities but added a Spruce collection from Tarapoto, some 250 km to the East, in San Martin province. Since duplicates of this collection were widely distributed, it comprises an important source for C. insignis s.s. but cannot of course be considered as type material. Despite its location in the upper valley of the Rio Huallaga, the Tarapoto area is a noted enclave of savanna vegetation where even some Brazilian cerrado species are found (pers. comm., T.R. Pennington) so that it is perhaps not surprising that C. insignis, essentially a dry woodland species of the W side of the Maran6n river system, also occurs there. The Spruce 3928 Tarapoto specimen at K has a more extensive label than most duplicates of this collection, and states "flores rosei" and that the plants occurred in "praeraptis saxosis". Although some labels refer to the flowers of C. insignis from Ecuador and Peru as simply 'white', the photos of this species by Lewis (collection Lewis & al. 3308,near Catamayo,Ecuador), and also that of Gentry (Gentry & al. 22688 near Pucara, in NW Peru) show petals with a distinct pinkish tinge. There is no great problem therefore in including the Spruce Tarapoto collection in C. insignis s.s., despite its occurrence at a site well to the E of the Maran6n river, and reference to 'pink' flowers.
In RAVENNA'S (1998) key to five species of Ceiba, C. insignis was keyed out together with C. speciosa as: "Corolla pink to bright purple-pink", and then distinguished from C. speciosa with a series of vegetative charac-ters ranging from mature tree height 45 m (insignis) vs. 25 m (speciosa), and "thorns not perfectly conical, oblique or curved, 30-55 mm, leaf margins repand, or teeth diminutive, less than 1 mm. Anthers 4-5 mm" (insignis) vs. "thorns perpendicular to bark surface, perfectly conical, 8-15 mm. Leaflets coarsely dentate, teeth oblique 1-3 mm. Anthers 7-8 mm." (speciosa). This interpretation of C. insignis is obviously at variance with that outlined above. However, since no exsiccata were cited it is difficult to understand Ravenna's concept of this species.
Chorisia integrifolia is another problematical taxon. Ulbrich described his species as having 'white flowers', and the type specimen is a Weberbauer collection from near Ja6n (Catamarca), which is close to the type locality of C. insignis at Chamaya, and other localities of 'white-pinkish flowered' C. insignis collections at Pucara by Gentry. Ulbrich's insistence on the diagnostic entire-margined leaves for his C. integrifolia is unhelpful since entire to minutely denticulate margined leaves are common in C. insignis s.s and other taxa of this aggregate. We therefore refer Chorisia integrifolia to C. insignis s.s. RAVENNA (1998) reported seeing trees of C. integrifolia between Pucara and Jaen but was apparently unaware that this locality is close to the type locality of C. insignis, or that this latter species is also characteristic of the "dry montane plains of the province of Jaen" to which he refers C. integrifolia. Unfortunately, RAVENNA (1998) did not include C. integrifolia in his key to species of the C. insignis group, so that his interpretation of this species is again uncertain.
Ulbrich also subsequently identified a collection by Raimondi from Junfn: Chanchamayo to his C. integrifolia. Whilst this specimen certainly has entire-margined leaves, the narrow petals with undulate margins are similar to those of the red-flowered C. lupuna (see below), and the locality on the Rio Mantaro is also in accord with this. We therefore refer this Raimondi specimen to C. lupuna.
In an exchange of letters, the late Al Gentry made the following comments concerning the red-flowered Chorisia which we have here described as Ceiba lupuna: "My big problem concerns what name to put on the red-flowered Chorisia that is common around Pucallpa in Amazonian Peru. We have two species of Chorisia growing sympatrically in this area. One has bright magenta flowers with rather sericeous petals and I am assuming this is correctly called C. insignis HBK [C. speciosa on our interpretation]. The other has bright red flowers, black spines on the trunk and has petals when dried more wooly-tannish pubescent outside". Another letter commented: "I recall once flying from Puerto Bermudez to Pucallpa and in the area near Puerto Bermudez seeing both red and pink Chorisia. Nearer Pucallpa there was only the red flowered ones". And with regard to ecology of these taxa Gentry noted: "My guess is that the red and pink flowered species sort themselves out according to soil fertility more than precipitation. The pink flowered form grows in rich soil forests and the red flowered form in more highly leached forests on lateritic soils". This species, with deep red flowers, is most likely to have diurnal anthesis but pollinator type unknown.  Trees 10-20 m, usually with somewhat swollen, aculeate trunk. Leaflets rather chartaceous, usually serrate. Pedicels 10-15 mm long rather stout. Petals 70-105 x 25-35 mm, spathulate to oblong, externally sericeous, internally glabrous, margin somewhat undulate, dark pink-magenta distally, usually with dark striations midlength, base white to yellow. Staminal tube to the appendages glabrous, appendage lobes usually dark reddish, densely cinerous hairy, sometimes pinkish and glabrous; upper staminal tube white flushed pink, sometimes splitting distally to give 5 short filaments. Stigma white or pinkish. Fruit an ellipsoidal to pyriform capsule, c. 12-18 x 9-12 cm.

Ceiba speciosa
Flowering January-May. Dry semi-deciduous woodland and wet forest. An extensive area from NE and SE Brazil and also W Brazil, N Argentina, Paraguay, Bolivia, S and C Peru ( fig. 1). This species is also widely cultivated in Argentina and Brazil as an ornamental. ), continue to flutter their wings as they probe the flower, such that these contact either the collar of anthers or stigma. C. speciosa flowers are also visited by humming birds, which mostly fail to contact the anthers, and nocturnal visits by large Phyllostomatus hastatus bats have also been reported by TADDEI (1977).  Tree 10 m or more with sometimes swollen, aculeate trunk. Leaflets denticulate. Pedicels 15-23 mm long. Petals c. 85 x 6-10 mm, very narrowly oblong, margin markedly undulate-crespate, sericeous externally, glabrous internally, dark pinkish-magenta with few striations distally, yellowish at the base. Staminal column c. 15 mm to the appendages, cinereous sericeous; appendage lobes hairy, dark red, staminal tube above the appendages glabrous. Stigma white or pink. Fruit an ellipsoidal to pyriform capsule c. 15 X 10 cm.
Illustrations. SANTOS (1967:9 fig. 2), SAN-TOS (1969, fig. 4), Very similar to C. speciosa in most features, but the narrow, markedly undulate margined petals, and pubescent lower staminal tube, resemble C. ventricosa (see below). Apparently restricted to atlantic forest and restinga around Rio de Janeiro and so partially sympatric with C. speciosa. DE CANDOLLE (1824: 480) treated C. ventricosa (see below) as a synonym of C. crispiflora, which is not surprising since both species have narrow, undulate-crespate petals. Indeed, the flowers of C. crispiflora resemble what might be expected in a hybrid between C. speciosa x C. ventricosa, but we have absolutely no evidence of such hybrid status.

Examined specimens
BRAZIL. MINAS GERAIS: Carangola, 20°40'S, 42°01W. 14-  Trees 10 m or more with swollen, aculeate trunk. Leaflets somewhat chartaceous, glabrous, usually entire, sometimes obscurely denticulate. Pedicels c. 20-30 mm long. Petals c. 60 x 8-10 mm, narrowly oblong with a markedly undulate-crespate margin, externally sericeous, internally glabrous, white to cream, with sparse dark reddish flecks distally, these merging more uniformly towards the base, becoming chocolate-brown in old flowers and dried specimens. Staminal column densely reddish-sericeous up to and including the 5 staminal appendages which are prominently bifid, then a glabrous, whitish staminal tube to the 5 sinuous anthers. Stigma somewhat 5-lobed, white. Fruit not seen. Harley (in litt.) notes that Prince Maximillian von Wied probably travelled upstream from near Jequie", on the Rio de Contas and headed north, spending the night at the locality Cabe?a de Boi (which no longer exists) and then another night near where he probably collected Chorisa ventricosa, before reaching Santa Ines on the Rio Jiquirica (= Tiquirica). Although the natural vegetation is much devasted today, this is an area of of deciduous dry forest. Since no specimen of Chorisia ventricosa has been located in the von Wied herbarium at BR, we choose as an epitype for this species dos Santos 1579. in Martius' Flora Brasiliensis. The similarity between these two species is particularly marked in herbarium specimens which have lost flower colour. Given differences in artistic style, the illustration of C. incana (ROBYNS, 1967) is remarkably similar to that of C. ventricosa in NEES & MARTIUS (1823, tab. 9).
Presumably with its pallid flowers, C. ventricosa is another species with crepuscular anthesis and nocturnal pollinators.  Cultivated trees in Campinas (Brazil) and Rosario (Argentina), the latter possibly of hybrid origin, open their flowers in the late afternoon, and it is likely that C. chodatii, with its erect pallid colored petals is another species with crepuscular-nocturnal pollinators. However, given that the species has a distribution largely outwith that of flower-visiting bats, and the nectar supply is sparse, these are likely to sphingid moths.
Flowers rather variable in size, and also include forms ranging from pale pink petal with very few striations, to others flushed dark pink-lilac and with distinct dark, winecoloured striations which tend to coalesce. C. pubiflora has diurnal anthesis. Flowers on trees in Bahia, and also in cultivation in Sao Paulo, were observed to be frequently visited, and so probably pollinated, by humming birds. RAVENNA (1998) described his C. jabaibana with entire margined leaflets, and pink, 7-8 cm flowers with petals externally tomentose. The important details of the androecium were described as: "Columna staminum circ. 30 mm longa, laevis, ad basin 3 mm lata, apicem versus sensim angustiora. Filamenta oblique patentia, 13-15 mm longa. Antherae oblongae, modice flexuosae, 9-10 mm longae, 1.5-1.8 mm latae".
We have studied the type material of C. jaibana, and also had the opportunity to study some 11 carefully annotated collections of Ceiba trees recently made at our suggestion in the Montes Claros-Jaiba-Janauba area of Minas Gerais by E.R. Salviani and J. Dutilh. The holotype consists of a twig with a single, broken flower, with two remaining stamens. The petals are similar to those of C. pubiflora, but the basal staminal tube does not have a collar of staminal appendages as occurs in that species, but rather a swelling, thus resembling C. erianthos. The paratype has leaves with an entire margin and an open fruit with kapok. It does not convincingly appear to correspond to the same plant as the holotypus, but could well be a specimen of C. pubiflora.
Illustrations. Fig. 2. BRITTEN & BAKER (1896) distinguished C. mandoni from C. boliviano with the former having more leaflets (6-7 vs. 3-4) which are slightly more long petiolulate and with more marked serrations, and the staminal appendages much nearer the apex of the staminal tube rather than near the centre as in C. boliviano. However, in material seen by us these characters are variable, sometimes between flowers of the same specimen, and so are unreliable to delimit species. The description of KUNTZE'S (1898) Xylon tunariensis, particularly the anfractuose anthers and Bolivian provenance, accord with C. boliviano.

Illustrations. SANTOS
The specimen Glaziou 18145a, cited by Kuntze for his Xylon glaziovii, was collected from a tree cultivated in Rio de Janeiro, which, acording to the label, apparently originated from a tree cultivated in Ipiranga, Sao Paulo. Confusingly, Glaziou (1913) cited his 18145a for both C. glaziovii and C. erianthos.
Flowering March-July. Restinga and dry coastal forest, usually on rocky outcrops. SE and E Brazil (Rio de Janeiro, E Minas Gerais, Espirito Santo and Bahia) ( fig. 4).
Crepuscular anthesis with bat pollination. The densely tomentose petals of this species are distinctive although some care is needed with herbarium material to distinguish from specimens of C. glaziovii.
The small cream flowers of this species may open at dusk, and so probably with moth pollination.
Flowering April-July. Dry valleys of Pacific coast. S Ecuador and N Peru (fig. 3).
A very striking species with its red calyx, densely white tomentose petals, and further fission of the five stamen filaments. Curiously the original collection of this species was from a cultivated tree in Lima city, and its transference to Ceiba by BAKHUIZEN  was also based on flowers from cultivated trees in Java (Indonesia). The species was still unknown in its natural habitat when MACBRIDE (1956), in the Flora of Peru, noted: "Type from a garden in Lima; also in Java, country of origin unknown." C. trischistandra is another species presumably with crepuscular anthesis and bat pollination. x 25-50 mm, narrowly elliptical to oblanceolate, base cuneate to somewhat truncate, apex acuminate, margin entire or obscurely dentic-ulate, glabrous, petiolule 5-30 mm long. Flowers fasciculate on the stems. Pedicel c. 20 mm long. Calyx 12-15 x 11-13 mm, campanulate, glabrous. Petals 22-46 x 6-13 mm, white or pinkish, externally densely sericeous sometimes with brownish hairs, sparsely so at the overlapping margins. Staminal tube variable: either c. 3.5 mm, with an abrupt, truncate termination, and then with 5 free white or markedly pinkish filaments arising directly or nearly so, or tube tapering for 5-14 mm before splitting into 5 filaments, with no staminodial appendages; anthers small, markedly anfractuose. Ovary pyriform with stigma usually exserted above the anthers. Fruit ellipsoidal to pyriform capsule 15-18x8-10 cm.
Flowering August to September. Distribution effectively pan-tropical but almost certainly introduced in Asia and the Pacific, although probably native, at least in part, in W Africa (cf. BAKER, 1965). We have not attempted to study this species in any detail outwith the Americas (fig. 6).
Illustrations. ADAMS (1972: 151, fig. 56); LORENZI (1992:60). NICOLSON (1979) provided an extensive commentary on the typification of the genera Bombax, Ceiba and Cochlospermum. He concluded that Bombax pentandrum L., which was based on a mixture of (mostly) asiatic but also some New World elements, must be typified by one of the illustrations in Rheede Hort. Malab., and lectotypified this species with figure 50 in this work. The illustration proposed by NICOLSON (1979) clearly depicts flowers of C. pentandra, and a plant with narrowly elliptical leaflets, but most leaves have eight (or more) leaflets instead of the 5-7 common in this species.  provided a detailed synonymy for Ceiba pentandra (L.) Gaertn., and he distinguished two varieties, var. caribaea (DC) Bakh., which included the New World and African forms, and var. indica (DC) Bakh. (= var. pentandra) for the asiatic plants. However, BAKER (1965) pointed out that most authors who recognize infra-specific, or even specific, variants for C. pentandra, distinguish between the New World and African forms. Both  and BAKER (1965), on the basis of their detailed studies of this taxon, concluded that C. pentandra should be recognized as a single rather polymorphic species, a view which we adopt here.
In Amazonian Brazil, C. pentandra often occurs as a tall emergent tree of varzea vegetation, although it also thrives (sometimes al-most certainly as planted specimens) in seasonally dry areas in Central America. This ability to grow both in riverine flood plain conditions and also in mesic, seasonally dry areas is found in some other species of diverse families, e.g. Triplaris gardneriana Wedd. (Polygonaceae), Pachira aquatica (Bombaceae) [J. Ratter, pers. comm.].
Even as represented in the in the New World, C. pentandra is another very variable Ceiba species, possibly with distinct infraspe- cific taxa. Plants with 'typical' leaves, i.e. narrowly oblong to narrowly elliptical, with distinct petiolules c. 10 mm, which correspond to Bombax cumanense (represented by a sterile specimen in the Humboldt & Bonpland herbarium at P), and which occur in Venezuela, and the Caribbean (e.g. Zanoni & al. 10817 from the Dominican Republic) look rather different from others with oblanceolate leaves which taper to short, 2-5 mm petiolules, and which correspond to Bombax mompoxense (again represented by a sterile Humboldt & Bonpland specimen at P) and characterised by, e.g. Gentry & Renteria 24534 from the Choco in Colombia.
And there are also striking differences in petal size and colour (white to distinctly pink), and the size and form of the staminal tube, and whether the free filaments are white or markedly pink. Two kinds of lower staminal tube are found: one has the tube around 6 mm and tapering to the level where the 5 free filaments arise. Examples occur in both S America and the Caribbean, e.g. Little 9431 (Colombia) and Bodim 8005 (Puerto Rico). The other has a much shorter tube, around 2-3 mm which terminates abruptly to give a truncate ledge from which the free filaments arise. This form is mainly restricted to S America, e.g. Froes 1997 (Maranhao, Brazil) and Krukoff 5646 (Acre, Brazil). However, some specimens, e.g. Zanoni 10817 (Dominican Republic) and Hermans 700 (Cuba) combine these features with a short truncate tube from which arises a short tapering tube which then gives rise to the 5 free stamens.
Given the general lack of specimens with both leaves and flowers, and the apparent absence of consistent geographical correlation with e.g. flower size and length of staminal tube, and also recognizing the long anthropomorphic interaction with this species, which has certainly influenced its distribution and possibly some morphological attributes, we prefer at this time to maintain a single variable species.
Ceiba pentandra is a species with nocturnal anthesis, copious nectar, and bat pollination. In the Manaus area of central Amazonian Brazil some four species of bats were ob-served to visit this species (see GRIBEL & al., 1999). In this area, flowers also have diverse diurnal morning visitors (birds, monkeys) seeking residual nectar, but since the styles abscise by 10 am, it is the nocturnal pollinators which produce fruits. BAKER & HARRIS (1964) also reported bat pollination for C. pentandra in W Africa. The nocturnal anthesis, and the imposing height of the trees, means that most herbarium specimens consist of old flowers that have fallen to the ground. Trees 15 m or more, branches with broadbased spines but usually relatively few on trunk which may be buttressed at the base. Leaves 5 foliolate, petiole c. 50-95 mm long; leaflets 33-120 x 28-55 mm, ellipticaloblanceolate, apex acuminate, base cuneate, margin entire, glabrous, petiolule 5-7 mm long. Flowers axillary, borne singly or in fascicles, rather funnelform. Pedicel c. 15 mm long, stout. Calyx 43-67 x 17-25 mm, cylindrical-funnelform, with 5 conspicuous teeth up to 9 mm, glabrous externally, densely villous within. Petals 100-220 x 17-22 mm, oblong-spathulate, whitish but with dense golden brown long sericeous-villous indumentum externally, internally uniformly short sericeous. Staminal tube basally 35-80 mm, dividing without the presence of staminodial appendages or swelling to give 5 filaments 45-90 mm, bearing short setae but appearing glabrous, with large, up to 18 mm orange, anfractuose anthers. Ovary subglobose, with a long slender style which is densely hairy as it emerges from the staminal tube, becoming glabrous distally, stigma shortly 5-lobed. borne at same level or somewhat above anthers. Fruit ellipsoidal-pyriform capsule 15-18 x 8 cm.
Flowering December-March(May). Humid and riverine forest. Extending from Bolivia and Peru to Amazonian Brazil ( fig. 7).
Illustrations. Fig. 9. A striking species with distinctive entire, mucronate leaflets and cylindrical calyx. Again crepuscular or nocturnal anthesis is likely, and since the petals are held erect, rather than reflexed as in C. aesculifolia, perhaps sphingids are the pollinators.
Flowering March to July (see comments below). Dry hillsides, semi-deciduous woodland. Northern Mexico southwards to Central America ( fig. 8).
As recognized here, C. aesculifolia is a very variable species which is widely distributed from Campeche to Sonora in Mexico, and Central America. There is wide variation in e.g. calyx size, length of the stamens in comparison with the petals, pubescence, and even flowering time (most examples seen flower between March to July but some August to January flowering specimens have been seen) and one would anticipate that some of the numerous species names which have been published for this group (see synonymy below) may comprise valid taxa. There are certainly striking differences in petal size and shape (particularly whether the apex is rounded or acute), and degree of exsertion of the stamens, and also pubescence (used to delimit C. tomentosa but note that e.g. Hinton 13878 has young leaves with simple and stellate hairs, but older leaves only with sparse hairs on the midrib). However, we have been unable to correlate any character differences consistently with distribution patterns.
Again the fact that C. aesculifolia usually flowers in a leafless state, with nocturnal anthesis {Bullock 1583 comments anthesis at 20.15 h) and bat pollination (BAKER & al., 1971), means that most flowers are collected the following day as they are wilting. As a consequence, the quality of some of the  Illustrations. Fig. 9.
Although Pringle 5300, the type of E. tomentosum, has not been located, it is likely to be similar to other collections from this locality, e.g. Pringle 4733 and Pringle 9685, both of which we refer to C. aesculifolia. Eriodendron acuminatum was described from a fruiting specimen, and flowering characters were derived from Sesse & Mogino, icon 94 (cf. WHITE & al, 1998). ROSE (1905), on making the transfer as Ceiba acuminata commented: 'Type locality: Hacienda San Miguel, Chihuahua, collected by Dr. E. Palmer. This species known only from the type collection and has never been collected in flower. It must be near C. tomentosum and with this material in hand it is difficult to separate them. The flowers may well show good specific differences". ROSE (1905)  Tall trees with aculeate trunk and flattish, spreading crown. Leaves described as digitate, but not seen by us. Pedicel very stout, 10 mm long. Calyx also very robust, c. 30 x 30 mm, orange-brown hairy without, very densely villous pubescent within. Petals c. 140 x 35 mm, spathulate, with a white inner surface, and golden brown hairy externally, reflexed and curling back. Staminal tube c. 20 mm, with a corona of five, bifid, densely hairy scales, with (10)15 slender, free filaments, c. 105 mm, each bearing a c. 7 mm anfractuose monthecate anther. Ovary c. 8 mm, pyriform. with a long, slender style which exceeds the level of the anthers by 10 mm or so, but the form of the stigma unknown. Fruit not seen.
Until the recent collection by Hughes & al. 1690, this remarkable species was only known from the type collection which consists of fallen flowers. The description above is derived partly from that of Donnell Smith, but floral details are mostly from the Hughes specimen. These latter flowers, although evidently larger than the Donnell Smith collection, agree in most respects with type description except in one intriguing detail: the original description for this species refers to 10-12 filiform staminal filaments, and Donnell Smith further emphasises this point in his additional comments, noting that freshly fallen flowers were the only ones accessible, and that the staminal column "partite to the annulus into double the number of branches is exceptional for the genus". The type specimen at BM and a duplicate collection at K both have floral fragments with 10 staminal filaments, whereas both flowers of the Hughes & al. 1690 collection seen by us have 15 filaments.
Given the close resemblance of the type specimens of Chorisia soluta to the very variable C. aesculifolia, which also extends to Guatemala, and given the fact that C. aesculifolia also has some very large flowered specimens with a robust calyx, we were initially inclined to treat the type specimen of C. soluta simply as an odd double-filamented variant of C. aesculifolia, especially since it was based on a single tree. This was also the interpretation of STANDLEY & STEYERMARK (1949) who treated Chorisia soluta as a synonym of Ceiba aesculifolia. However, the Hughes & al. 1690 collection shows that other specimens occur with a similarly multi-filamented androecium comparable to the Donnell-Smith collection, although the number of free filaments seems to rather variable. In these circumstances, despite its evident affinity with C. aesculifolia, we maintain Ceiba soluta as a species.

EXCLUDED OR DOUBTFUL NAMES
There is obviously insufficient data here to determine whether this description is indeed a new taxon of Ceiba. Some comment as to whether the staminal tube is entire or with five free filaments would have been helpful. We have not seen the apparently fragmentary type specimen. Pending further collections, we assume, from the locality and description of white flowers, that this material represents rather malformed trees of C. insignis, which also occurs in Piura. Ceiba phaeosantha K. Schum. in Mart, (ed.), Fl. Bras. 12 (3): 214 (1886). Schumann commented "species mihi non visi" and based this species on Eriodendron phaeosantha Decne., J. Soc. Hort. Paris 4: