A revision of Spondias L. (Anacardiaceae) in the Neotropics

Abstract As part of an ongoing study of Anacardiaceae subfamily Spondioideae, the ten native and one introduced species of Spondias in the Neotropics are revised. The genus is circumscribed. Three new species, Spondias admirabilis, Spondias expeditionaria, and Spondias globosa, are described and illustrated; a key to the taxa found in the Neotropics and distribution maps are provided. The Paleotropical species and allied genera are reviewed. Diagnostic character sets include leaf architecture, habit, flower morphology, and gross fruit morphology. Notes on the ecology and economic botany of the species are provided.


Introduction
Spondias L. is a genus of fruit trees that comprises 18 species native to tropical America and Asia, and Madagascar. It is the type genus of the subfamily Spondioideae Takht. emend. Pell & J. D. Mitch., which is indicated by molecular systematic work currently under way as being the more basal (but possibly polyphyletic) group of a monophyletic Anacardiaceae sister to the Burseraceae (Pell 2004;Mitchell et al. 2006;Pell et al. 2011). Th is corroborates previous conclusions by Wannan and Quinn (e.g., 1990), who noted that the endocarp structure of Spondioideae is similar to that of Canarium (Burseraceae), by Bachelier and Endress (2009), who noted that those fl oral morphology and anatomy characters shared by Burseraceae and Anacardiaceae are usually found in the Spondioideae (as Spondiadoideae), and by Terrazas (1994), whose results from cladistic analyses of wood and leaf anatomical characters and rbcL sequences also showed the group as basal.
Th e subfamily is characterized by consistently obdiplostemonous fl owers (sensu Bachelier and Endress 2009) and by usually separate styles, 4-5-carpellate ovaries, apical ovule insertion, thick exocarp, and a usually bony endocarp composed of a mass of strongly lignifi ed and irregularly oriented sclerenchyma, often but not always operculate.

Taxonomic history
Generic limits and defi nitions in the Anacardiaceae have been re-drawn several times during the past century and still need some work; revision of Spondias has required further re-examination of subfamily Spondioideae and led to the discovery of at least one new genus from Madagascar (Mitchell et al. 2006) and a few possible additional segregate genera.
Spondias was one of the fi rst genera of Anacardiaceae described by Linnaeus (1737: 365), with the type species S. mombin published in 1753. Inexplicably, he later published two names homotypic with S. mombin (and therefore illegitimate), S. myrobalanus (Linnaeus 1759a: 1036) and S. lutea (Linnaeus 1762: 613). Furthermore, he published the name S. myrobalanus a second time, in Flora jamaicensis (Linnaeus 1759b: 16), in this instance an illegitimate homonym homotypic with S. purpurea. Th e syntypes for the latter S. myrobalanus served as the basis for S. purpurea, for which Bornstein (1989) selected a lectotype. Despite its early establishment and economic importance, numerous taxa have been assigned to Spondias that are no longer accepted by most authors. Th is fact is illustrated by Kostermans (1991) who in his treament of Asian Spondias, referred the excluded species to 15 diff erent genera, three of these not even in the Spondioideae as currently defi ned. Bentham and Hooker (1862) were the fi rst to formulate groupings of genera in the Anacardiaceae, dividing the family into two tribes, the Anacardieae and Spondieae [sic]. Subsequently, Marchand (1869) published the tribe Spondiadeae (as Spondieae) and was the fi rst to formulate a relatively modern concept of Spondias, in which he included Evia Blume, Cytheraea Wight & Arn., and Wirtgenia Jung. ex Hassk. (nom. illegit., non Wirtgenia Sch. Bip.). On the other hand, of the taxa he either accepted in Spondias or recognized as synonyms of species in the genus, four are considered here to belong to other genera (see excluded species section).
Two treatments of Asian Spondioideae took opposite extremes in the circumscription of the genus. In their revision of tropical Asian Spondias, Airy Shaw and Forman (1967) lumped Allospondias and Solenocarpus with a rather broadly defi ned Spondias, but this would leave Spondias without a single synapomorphy and in fact joins taxa with disparate character states: simple vs. bipinnately compound leaves, unicarpellate vs. multicarpellate ovaries, and leafl ets with or without an intramarginal vein. Ding Hou's (1978) treatment of the family for Flora Malesiana followed Airy Shaw and Forman's broad circumscription.
In contrast, Kostermans (1981Kostermans ( , 1991 defi ned the genera of the Spondioideae rather narrowly, maintaining Allospondias and Solenocarpus, transferring Spondias philippinensis (Elmer) Airy Shaw and Forman to the latter, describing the new genus Haplospondias, and formally returning the South Pacifi c species Spondias dulcis Parkinson into the pre-existing genus Evia Commerson ex Blume emend. Kosterm. Pending completion of a molecular systematic-based generic revision of the subfamily, we comment briefl y on these in turn: We agree with Kostermans that Allospondias lakonensis (Pierre) Stapf [syn. : Spondias lakonensis Pierre] should be kept out of Spondias, on the basis of the former's lack of an intramarginal vein and presence of perpendicular epimedial tertiary veins, highly branched sclereids terminating the FEVs (freely ending leaf veinlets), styles connivent at anthesis and stigmas extrorse on the developing fruit, lack of a fi brous matrix on the endocarp, and presence of parenchyma-fi lled chambers in the endocarp. Kostermans' circumscription of Allospondias encompassed Spondias laxifl ora (Kurz) Airy Shaw & Forman; we concur that it does not belong in Spondias because of its eucamptodromous secondary venation (vs. intramarginal vein), versatile (vs. fi xed) anthers, and endocarp apertures not covered by the fi brous matrix, but we reserve judgment on its placement in Allospondias: while A. laxifl ora and A. lakonensis (the type species of Allospondias) share an apert calyx, valvate corolla, and papillate stigmas, and both lack a fi brous matrix outside the endocarp, A. lakonensis has the styles continuous with the lateral lobes on the pistil and connivent or appearing connate apically, the stigmas spathulate, while A. laxifl ora has the styles free and not continuous with the lateral lobes on the pistil, the stigmas capitate or discoid. Moreover, the fruit in A. lakonensis is radially symmetrical and 4-5-lobed and has parenchyma-fi lled chambers, while in A. laxifl ora it is strongly oblique and unlobed and lacks parenchyma-fi lled chambers. We agree provisionally with Kostermans that Solenocarpus indicus Wight & Arn. does not belong in Spondias. While like Spondias it has admedial tertiary veins originating near the margin, an intramarginal vein, apert calyx (imbricate in S. purpurea), and valvate corolla (quincuncial in S. purpurea), in contrast to Spondias s.s. it has a single fl abellate style, single stigma and unicarpellate gynoecium, and it lacks a fi brous matrix on the endocarp. Moreover, in Solenocarpus the sclereids in the mesocarp occur only in a layer just outside the endocarp, while in Spondias they essentially sheath the resin canals, which are scattered throughout the mesocarp (Wannan and Quinn 1990). We agree with Kostermans that Spondias philippinensis should be removed from Spondias on the basis of its eucamptodromous secondary venation (vs. intramarginal vein), single narrowly fl abellate style, single stigma, unicarpellate ovary, and strongly oblique fruit, and we tentatively agree with Kostermans' placement of this species in Solenocarpus, because although S. philippinensis lacks an intramarginal vein, its fl ower morphology is very similar to that of Solenocarpus indicus: apert calyx, valvate corolla, single narrowly fl abellate style, and unicarpellate gynoecium. We agree that monotypic Haplospondias brandisiana is distinct from Spondias; it has simple leaves without an intramarginal vein, and a single style with an oblique (and possibly bilobed) stigma. Kostermans (1991) published Spondias tonkinensis Kost. and later considered it to be a taxonomic synonym of S. petelotii (Tardieu-Blot) Kost. (syn.: Dracontomelon petelotii Tardieu-Blot; Kostermans 1992), but our generic assignment of this taxon is as yet uncertain. It has festooned-brochidodromous leafl et venation with raisedmarsupiform (pouch-like) domatia in the axils of the secondary veins, calyx lobes open in bud but their bases imbricate, petals imbricate, anther connective glandular (as in Cyrtocarpa), styles connivent at anthesis, and stigmas slightly extrorse and slightly ellipsoid. Each of these character states can be found elsewhere in the Spondioideae but by no means in this combination. Placement of Spondias dulcis in Evia was based on the endocarp with spinose projections penetrating the mesocarp (vs. a simple fi brous matrix) and the woody pedicel partially embedded in the basicrescent developing fruit. In light of the fact that S. dulcis shares similar fl ower morphology as well as the intramarginal vein and other aspects of leaf architecture with the American species of Spondias, these fruit characters could be seen as derived within Spondias, rather than warranting recognition at generic rank. Placement of Spondias bipinnata presents a challenge. Like Spondias, it has an apert calyx, valvate corolla, the endocarp with a fi brous matrix, and a 5-locular stone. Unlike Spondias, it has bipinnate leaves, lacks an intramarginal vein, and has pubescent styles; other aspects of its morphology are shared with Allospondias lakonensis, including virtually identical leafl et venation and connivent styles.
More recently, the genus Attilaea was described from the Yucatán of Mexico by Martínez and Ramos-Álvarez (2007. Attilaea is similar to Spondias because of its intramarginal vein and red fl owers (like in S. purpurea), but it diff ers by the climbing habit, the 2-carpellate, unilocular gynoecium (vs. (3-4)5-carpellate, 4-5-locular), and the single seed per fruit (vs. [4][5]. At species rank, Marchand (1869) placed S. mombin as a synonym of the nomen illegitimum S. lutea L.; this error was not corrected until 1926 by Fawcett and Rendle. In addition, he recognized the illegitimate S. dulcis G. Forst. (a synonym of S. dulcis Parkinson), and described var. amara, including in it several taxonomic entities referred by Kostermans (1991) and the present authors to S. acuminata, S. pinnata, and S. malayana.
In Flora brasiliensis, Engler (1876) kept S. mombin as a synonym of the nomen illegitimum S. lutea, put S. venulosa as a variety of S. purpurea, and described S. macrocarpa.
In Monographie phanerogamerum, Engler (1883) placed Warmingia paucifl ora Engl. in synonymy with Spondias purpurea (which he had placed in Burseraceae in 1876). He also placed Spondias macrocarpa as well as the Asian taxon S. acida as varieties of S. dulcis. He raised S. purpurea var. venulosa to specifi c rank, and recognized S. mangifera, a name considered by Kostermans (1991) as illegitimate under S. pinnata.
In the Flora of Jamaica, Fawcett and Rendle (1926) were perhaps the fi rst to critically examine the early literature and place Spondias lutea in synonymy with S. monbin [sic], but they erroneously placed the Eastern Brazilian S. macrocarpa in synonymy with S. cytherea (=S. dulcis). Th ey illustrated both staminate and pistillate fl owers of S. purpurea.
In the Flora of the Lesser Antilles, Bornstein (1989) lectotypifi ed S. mombin and S. purpurea; he also researched the typifi cation of S. lutea and S. myrobalanus (Syst. Nat. Treated as synonym of S. pinnata in Flora of China (Min & Barfod, 2008) S. admirabilis J. D. Mitch. & Daly 10, 2: 1036. 1759.) and explained why these are both nomina illegitima, being based on the same types as S. mombin. He also pointed out that S. purpurea is a nomen novum for the illegitimate homonyn for Spondias myrobalanus L. (Fl. jamaic. 16. 1759).
Th e eighteen species of Spondias accepted in the present treatment, their geographic distributions, and their principal synonyms are summarized in Table 1.

Comments on phylogeny
It is exceedingly diffi cult to purify and amplify DNA even from fresh leaf samples of Spondias (S. Pell and A. Miller, pers. comm.); to date, sequences have been obtained from S. globosa, S. mombin, S. purpurea, S. testudinis, and S. tuberosa, but not from any taxa native to the Paleotropics other than S. dulcis, S. malayana and S. pinnata. Current data suggest that the subfamily Spondioideae is divided into two clades; the smaller and more basal of these contains Spondias along with Allospondias, Dracontomelon, Pegia, and Pseudospondias. Neotropical Spondias is sister to Asian Spondias (Weeks et al. 2014).
Vessels diff use-porous to rarely slightly semi-ring-porous with a vessel density ranging from 2-14/mm 2 (visible without a hand lens), and varying from 121-357 μm diam, mostly solitary or in groups of 2-8, round to slightly oval in outline. Parenchyma paratracheal, vasicentric, sometimes aliform. Rays heterogeneous (homogeneous in S. tuberosa, Paula and Alves 1980), Kribs Type II or III, rhombic crystals of calcium oxalate often present. Fibers mostly libriform and septate. Axial and radial gum ducts and resin canals present. Terrazas (1999) observed that the distinguishing characteristic of Spondias wood is the large diameter of the vessels (often >150 μm) relative to other Anacardiaceae genera.
Th e wood of the recently described Anacardiaceae genus Attilaea (Martínez and Ramos-Álvarez, 2007), which is endemic to the Yucatán Peninsula, has been compared with that of the sympatric and morphologically similar S. purpurea (Gómez 2009,

Leaf architecture
Th e present work uses the terminology in Ellis et al. (2009), and the genus and species descriptions contain details on leaf architecture. Although the presence of an intramarginal vein (Fig. 1) on leafl ets is characteristic of Spondias (except present in Attilaea abalak and Solenocarpus indicus), the genus presents a wealth of other characters of its leaf architecture that are variable and taxonomically useful (see below the key to the Neotropical species based on leaf(let) characters). Many macromorphological leaf characters and virtually every major character of the leafl et venation and symmetry shows at least some variation among species; the following characters have proven most useful: (1) number of leafl et pairs; (2)  Breeding systems, floral morphology, flower and fruit anatomy Flowers of all species except S. purpurea are morphologically hermaphroditic (Bachelier and Endress 2009); S. purpurea is dioecious and the fl owers display relatively strong sexual dimorphism for the duration of the fl ower: the pistillate fl owers have staminodes where microspore mother cells degenerate during meiosis, and in a tropical dry forest in Morelos, Mexico the staminate fl owers develop a month earlier than the pistillate ones (Hernández-Martínez et al. 1999). In the other species of Spondias occurring in the Neotropics, we have observed strong protandry, such that on a given infl orescence one fi nds fl owers with dehiscing anthers and an underdeveloped gynoecium at early anthesis, and fl owers with fully-developed pistil and passed anthers at late anthesis (see Fig. 11); it is likely that most if not all reports of staminate fl owers are in fact simply hermaphrodite fl owers in early anthesis. In S. mombin, Lozano (1986) reported hermaphroditic, staminate, and pistillate fl owers on the same plant; she determined that in some fl owers the anthers produced sterile pollen, and she interpreted the poorly developed pistil in some fl owers as representing staminate fl owers with suppressed gynoecia. Nadia et al. (2007) concluded that S. tuberosa is andromonoecious; Bawa (1974) concluded that S. mombin is monoecious and strongly self-incompatible while Stacy et al. (1996) found that species to have functionally hermaphroditic fl owers, and Croat (1974a) reported S. radlkoferi to have mostly bisexual or rarely pistillate fl owers.
Floral morphology and anatomy of Spondias have been studied by Lozano (1986b;S. mombin), Bachelier and Endress (2009;S. dulcis and S. purpurea), and Hernández-Martínez et al. (1999;S. purpurea). Th e fl owers of Spondias are pentamerous and isomerous. Th e sepals are free although early aestivation is quincuncial. Corolla aestivation is valvate, and at anthesis the petals are free and refl exed or patent ((sub)erect only in S. purpurea). Th e fl owers are obdiplostemonous (the outer whorl of stamens opposite the carpels); the antepetalous stamens are shorter at all stages. Th e anthers are dorsifi xed on the lower half in S. dulcis but dorsifi xed at the base in S. purpurea; all species are longitudinally dehiscent most of their length. All species have an intrastaminal, annular, secretory disk. Th e styles are longer than the ovary at fi rst; they are massive and connivent in S. dulcis, but in S. purpurea they are shorter than the ovary, more discrete and arranged around the periphery of the pistil apex. As in most Anacardiaceae examined to date, the pollen tube grows down a furrow, then channels into separate papilla-lined stylar canals that bypass the micropyle by passing a bridge (ponticulus) through a dorsal outgrowth of the funicle. Th e stigmas are papillate; they are capitate in S. purpurea but linear and oblique in S. dulcis and the remaining Neotropical species. Th e ovules are crassinucellar, bitegmic, and syntropous (= apotropous), with a long funicle.
As in other Spondioideae, the exocarp is very thick. Th e homology of the fi brous matrix in Spondias is unresolved; it was considered by Lozano (1986a) to be derived from the inner mesocarp, and by Wannan and Quinn (1990) to be part of the endocarp. As conspicuous and economically important as the fruits of Spondias are, however, fruit anatomy is poorly studied. An exception is the work of Wannan and Quinn (1990), who used anatomical patterns of the endocarp in Anacardiaceae to detect generic affi nities in the family and to characterize subfamily Spondioideae and distinguish it from the other tribes recognized by Engler (1883) that are now joined to form the only other subfamily, Anacardioideae (Pell 2004, Mitchell et al. 2006, Wannan 2006, Pell et al. 2011. Wannan and Quinn (1990) observed that in Spondioideae the sclereids in the endocarp are arranged in an irregular way, without the histologically discrete layers characteristic of the remainder of the family. Within Spondioideae, they found considerable diff erences among genera in the histology of the mesocarp, which ranges from completely parenchymatous in Dracontomelon to completely lignifi ed in Pleiogynium; in Spondias pinnata, sclereids essentially sheath the resin canals, which are scattered throughout the mesocarp.

Pollen
Pollen descriptions are available for only three of the species found in the Neotropics: S. mombin, S. purpurea, and S. radlkoferi (see Table 2); moreover, all published studies of Spondias pollen have utilized only light microscopy. Studies are needed for more of the species, and SEM studies are needed for examining exine characters.
Th e shape is spheroidal to (sub)prolate. Like most Anacardiaceae, Spondias pollen is tricolporate. Th e exine is semi-tectate, and microstriate to microreticulate. Quantitative values are apparently variable within species and generally not diagnostic except for nexine thickness (for S. radlkoferi) and possibly P/E ratio.

Hybridization and intermediates
Th ere is a great deal of circumstantial evidence of hybridization in neotropical Spondias, almost all of it implicating S. mombin as one of the putative parents and most reported cases occurring near the range edges of one or both putative parents or where one or both may have been introduced. Only one of the putative hybrids has been formally recognized, Spondias × robe Urban from Haiti (Urban 1929). Th e type collection, Ekman 12532 (A, GH, NY, US) as well as one collection each from Cuba (Schafer 1525, NY), the Dominican Republic (Zanoni & Jiménez 44521), and Costa Rica (Grayum 11809, NY) combine the small, obovate leafl ets and pink corolla of S. purpurea with the much more lax infl orescence and very diff erent remaining fl ower morphology of S. mombin.
Phylogeographic studies of Spondias purpurea and both sympatric and allopatric populations of S. mombin (Miller andSchaal 2005, Miller 2008) suggest ongoing hybridization between these two species in southern Mesoamerica, although Miller (2008) notes that more studies are needed to determine the relative roles of hybridization versus incomplete lineage sorting.
In northeastern Brazil, local people recognize a variant of Spondias locally referred to as "umbu cajá," which some have conjectured might be a hybrid between S. mombin and S. tuberosa. Almeida et al. (2007) conducted a genetic study of this putative hybrid but could fi nd no evidence in the chromosome banding patterns. Th is entity may be simply a cultivated race of the regionally popular S. tuberosa.
Conversely, in northwestern Costa Rica isozyme studies of individuals of Spondias found some that were morphologically indistinguishable from S. mombin but whose isozyme bands strongly suggested they were hybrids between S. mombin and S. radlkoferi (e.g., Moran et al. 6293, NY); R. Moran and J. Hamrick, unpublished results).
In southwestern Amazonia, where S. mombin, S. globosa, and S. testudinis are native and S. purpurea and S. dulcis are sparsely cultivated, there is a distinct entity recognized by local people, who call it "cajá açu" ('large cajá'). Th is may be a hybrid between S. mombin and S. testudinis. Th e fruits are oblong and lenticellate like S. testudinis, but larger. As in S. mombin, the lateral leafl ets are relatively broadly oblique-elliptic (vs. obliquely oblanceolate to narrowly elliptic in S. testudinis), with the margin mostly entire but with a few of the laciniate teeth characteristic of S. testudinis. Th is entity has multiple admedial tertiaries arising from the secondaries, whereas in S. testudinis (as in S. globosa) the admedial tertiaries are long and composite while in S. mombin they are random-reticulate.
Th ere is morphological evidence of hybridization between Spondias mombin and S. globosa. Th e former is widespread and widely cultivated from S Mexico to SE Brazil and eastern central Bolivia, while S. globosa is more restricted, occurring in W Amazonia and disjunct in Zulia, Venezuela (see Figs 12,13). In SW Amazonia, the two are usually easy to distinguish, but in Amazonian Ecuador, Colombia, and Peru one frequently fi nds individuals that appear to be intermediate between the two. Typical S. globosa has (depressed-)globose fruits, leafl ets with composite admedial tertiaries running from the intramarginal vein as well as from the secondaries, and marginal ultimate venation lacking a fi mbrial vein and incompletely looped, while S. mombin has broadly oblong fruits, tertiary venation mostly irregular-reticulate with some admedial branching from the secondaries, and the marginal ultimate venation consisting of a fi mbrial vein. We have observed quite a number of specimens (see Index of Specimens Examined) with the fruits of S. mombin and the leafl et architecture of S. globosa (although we have not been able to check the marginal ultimate venation of all the specimens).

Seedlings
Th ere has been some confusion in the literature about terminology related to germination patterns and seedling morphology. Here we use the terminology of Duke (1965) and agree with him that the terms hypogeal and epigeal have been misapplied to seedlings that are cryptocotylar (the cotyledons enclosed in the germinating seed) and phanerocotylar (the cotyledons exposed).
Th e species of Spondias occurring in the Neotropics all have hypogeal germination. Th e seedlings are phanerocotylar (although Macías-Rodríguez and Pérez-Jiménez [* Our morphology for S. purpurea disagrees with that of Duke (1965), who reported the fi rst eophylls as opposite and trifoliolate, but as we have a voucher to examine, we suggest that his illustration labeled as S. purpurea is misidentifi ed.] (1994) report S. purpurea as cryptocotylar). A tap root and hypocotyl emerge from one end of the fruit; the hypocotyl emerges, curved at fi rst, then carries cotyledons aloft; the cotyledons are opposite, linear, green, sessile, somewhat fl eshy; fi rst eophylls opposite (S. purpurea usually alternate); lateral leafl ets of eophylls usually ovate (lanceolate in S. testudinis, oblanceolate in S. purpurea), the margin serrate (usually sparsely and regularly so), the teeth sometimes laciniate (S. purpurea, S. radlkoferi, S. testudinis). Th e patterns are summarized in Table 3.

Ecology
A comprehensive review of Spondias ecology in tropical America will not be attempted here; some taxon-specifi c notes are provided under each species. Th e majority of the ten native species grow primarily in tropical lowland moist forests below 1,000 m elevation, with two exceptions. Spondias purpurea naturally occurs in (semi-)deciduous forests, and S. tuberosa grows in semi-arid deciduous forests called caatinga arbórea in NE Brazil. Th e latter has water-storing tuberous roots, while S. purpurea stores water in its trunk (Borchert 1994); both remain leafl ess for long periods of time in the dry season. Based on the limited amount of research on their fl oral biology, Spondias species are pollinated primarily by Hymenoptera (mostly bees, some wasp species) (Roubik et al. 1986, Nadia et al. 2007); Carneiro and Martins (2012) found that fl owers of S. mombin were pollinated primarily by Scaptotrigona aff . tubiba and Africanized honey bees due to their abundance, behavior, and ability to visit a large number of fl owers. Th e fl eshy drupes of Spondias are ingested or expectorated, and the unit of dispersal is the endocarp or stone. Th e mesocarp has a high sugar content (Riba-Hernández et al. 2003), and the fruits are dispersed by a wide diversity of vertebrates, such as medium-sized to large birds (e.g., orioles to chachalacas); various mammals, such as bats, primates, rodents (e.g., agoutis), deer, peccaries, tapirs, coatis, kinkajous, coyotes, and foxes; reptiles (such as ctenosaurs and tortoises) (van der Pijl 1957, Goodwin and Greenhall 1961, Hernández-Camacho and Cooper 1976, Gardner 1977, Husson 1978, Freese and Oppenheimer 1981, Janzen 1985, Fleming 1987, Moskovits and Bjorndal 1990, Handley et al. 1991, Mandujano et al. 1994, Griz and Machado 2001, Vinke et al. 2008).
Janzen Martin 1982, Janzen 1985) hypothesized that Spondias fruits are among those whose dispersal has been adversely impacted by the extinction of the Pleistocene megafauna (e.g,. gomphotheres, tree sloths, horses, camels), because they are less eff ectively dispersed (i.e., distance and germination rates diminished).; Spondias fruits can be important in the diet of certain vertebrate species (especially during periods of fruit scarcity) and in forest succession or recovery. Some Spondias species, such as S. mombin and S. radlkoferi, are dispersed into pastures and large clearings and become common in second-growth forests (Janzen 1985).
Th e gum exudates of some Spondias species can be an important constituent of the diet of monkeys such as tamarins (Garber 1985). Bruchid beetles are the primary seed predators of at least two species of Spondias (S. radlkoferi and S. mombin (Janzen 1980). Several species of fungal endophytes have been identifi ed in the leafl ets of Spondias mombin, and these appear to play a role in the production of certain secondary metabolites (Rodrigues and Samuels 1999).

Economic botany
Spondias has a history of use going back at least as far as 6500 B.C., in the Tehuacán Valley of Mexico (Smith 1967). Four species of Spondias are economically important in tropical America: S. dulcis, S. mombin, S. purpurea and S. tuberosa. Th e economic botany of the commonly cultivated species of Spondias has been reviewed extensively in several publications, therefore we present a brief overview after each species with reference to important publications. Th e vernacular names of Spondias species cited in the economic botany literature are often misleading because of incorrect synonymies, erroneous species identifi cations, uncritical use of previous literature, and faulty equation of common names and scientifi c names. Approximately 180 common names have been linked to Spondias purpurea (Miller and Schaal 2005), while Morton (1981) listed 96 diff erent common names for S. mombin. In Morton (1981) and Lim (2012), for example, Spondias purpurea and S. tuberosa are confused. For this reason, it is essential to both use and cite vouchers in economic botany and in ethnobotany.

Conservation
Following the categories and criteria established by the IUCN (http://www.iucnredlist. org/technical-documents/categories-and-criteria/2001-categories-criteria), based on observations in the fi eld and in herbaria, we consider three species of Spondias to be at risk of extinction in the short-and medium-term: S. admirabilis (Endangered), S. expeditionaria (Critically Endangered), and S. macrocarpa (Vulnerable). Four species are rather broadly cultivated (S. dulcis, S. mombin, S. purpurea and S. tuberosa) and so run no risk of extinction, although their genetic diversity maybe becoming compromised due to habitat destruction. Spondias radlkoferi and S. venulosa are cultivated on a limited scale and range. Type. Spondias mombin L. Description. Small to large trees (rarely shrubby and broadly branching), usually hermaphroditic (except S. purpurea). Simple, thick plank buttresses to 100 cm high sometimes present. Outer bark brown or usually gray, densely to broadly fi ssured, sometimes thick, usually rough, often with raised lenticels, rarely (some S. mombin and S. purpurea) with large, corky, tooth-like projections. Inner bark usually broadly striate (white and rose, red, orange, or brown). Resin viscous and usually clear or less often cloudy (S. globosa). Trichomes of various simple hairs and capitate glandular hairs. Leaves alternate, aggregated toward branch tips, sometimes deciduous (sometimes facultatively so), usually imparipinnate, petiolulate; leafl ets (sub)opposite, the apex often apiculate; margin entire/cren(ul)ate/serr(ul)ate (Figs 2-4), the sinuses often glandular; intramarginal vein present (see Figs 1-10 for leafl et venation); intercostal tertiaries usually irregular-reticulate and often branching admedially, areoles often poorly developed; freely ending veinlets dendritic, often highly branched, leafl ets often glandular-punctate. Infl orescences produced either before leaf fl ush (e.g., S. purpurea) or during leaf fl ush, terminal or axillary, paniculate (reduced to a pseudoracemose botryoid in S. purpurea, Barfod 1988); infl orescence bracts subtending secondary axes early-caducous; pedicel articulated. Flowers usually hermaphroditic but strongly protandrous (dioecious in S. purpurea, Fig. 14), pentamerous, calyx green (often red in S. purpurea), shorter than the disk, lobes usually apert (quincuncial at the base and valvate at the apex where they meet in S. purpurea, Bachelier and Endress 2009); corolla valvate, petals yellowish-white (usually red in S. purpurea); patent or strongly refl exed at anthesis (spreading in S. purpurea), often terminating in a swollen and strongly infl exed apiculum, the margin papillate; stamens (8)10, in two unequal series, outer series opposite the carpels, fi laments linear to subulate, glabrous, anthers dorsifi xed, yellow; disk intrastaminal, annular, summit and outer margin variously shaped (Fig.  15), usually yellow or less often pink (some S. purpurea) or purple (some S. radlkoferi); pistillode (S. purpurea only) reduced to 3-5 subulate styles; at anthesis the ovary usually rudimentary, developing after the anthers dehisce (Fig. 11), pistil (3-) 5-locular, surmounted by (3-)5 short, free styles (S. purpurea) or usually the styles (broadly) subulate and separate at least half the length of the pistil, connivent at least basally, stigmas usually extrorse (introrse in S. purpurea and S. radlkoferi), discoid to linear; each locule with one apical ovule, the ovules crassinucellar and bitegmic, styles separating as fruit develops. Fruit ( Fig. 16) a (1) 3-5-seeded, green, yellow, orange, or red drupe, oblong, (depressed-)globose, obovoid, or ellipsoid, remnants of styles widely separated and often visible in developing fruits; exocarp usually smooth (densely lenticellate in S. testudinis); mesocarp fl eshy, edible, sweet and sour, endocarp bony and enveloped by a fi brous matrix (with spiny projections and growing over apex of pedicel in S. dulcis), at maturity the fruit often breaking away at articulation of pedicel. Pollen (based on Roubik and Moreno P. 1991): 41-56 × 22-44 μ, prolate or subprolate, pores 6-15 μ diam. Seedlings: germination hypogeal, tap root and hypocotyl emerge from blunt end of fruit; hypocotyl emerges, curved at fi rst, then carries cotyledons aloft; phanerocotylar; cotyledons opposite, linear, green, sessile, somewhat fl eshy; fi rst eophylls opposite and imparipinnate (S. testudinis, Fig. 17H) or trifoliolate (pers. obs., Vogel 1980, Garwood 2009). Chromosomes: n = 16, 2n = 32 (Simmonds 1954, Bawa 1973, Mehra 1976, Guerra 1986, Pedrosa et al. 1999. Of the ca. 18 species in Spondias, ten are native to the New World, distributed from Mexico to southern Brazil , one is native to Madagascar, and seven are native to Asia and the South Pacifi c, from Malesia (sensu Flora Malesiana) to tropical China, Sri Lanka, Indochina, Th ailand, India (except extreme north), Myanmar (Burma), Solomon Islands east to Polynesia. Spondias dulcis is cultivated in tropical America and the Antilles; S. mombin and S. purpurea are both introduced throughout Tropical West Africa and Asia (and in the West Indies, where they are not found in primary vegetation and may not be native); S. mombin is often adventive in Tropical West Africa.
Some of the species native to the Neotropics have restricted distributions. Spondias testudinis is restricted to southwestern Amazonia, and S. admirabilis and S. expeditionaria both are known from very few localities in Brazil's Atlantic Coastal Forest, while the other two Atlantic Forest species, S. macrocarpa and especially S. venulosa, are somewhat more broadly distributed in that region; both S. expeditionaria and S. macrocarpa are rare as well. Th e natural distribution of S. tuberosa is the arid caatinga vegetation of Northeastern Brazil, and S. globosa is a Western Amazon species. Spondias radlkoferi ranges from Mexico through Central America to Colombia and NW Venezuela; there is an unconfi rmed report from Los Ríos in W Ecuador (Dodson 8837, MO). Spondias purpurea is native to N Mexico through Central America and may be native to SW Ecuador. Spondias mombin is native to moist forests through much of northern South America, although it is uncertain whether the populations in Brazil's Atlantic Coastal Forest are native.
Leafl et venation: Fimbrial vein absent; secondary veins 7-13 pairs, rather straight, spacing irregular, angle decreasing toward apex and often increasing toward base, insertion decurrent; some intersecondaries present and parallel to secondaries; intercostal tertiaries admedially ramifi ed, with frequent composite admedials and very little reticulation; areolation barely developed (mostly at tertiary rank); FEVs 4-5+-branched, somewhat dichotomous, terminating in tracheoid idioblasts; marginal ultimate vena-tion incompletely looped; on abaxial side the midvein often very narrowly prominulous and usually sunk in a groove, secondary veins usually fl at (sometimes prominulous) but drying discolorous, densely pubescent at base; on adaxial side the midvein narrowly prominulous to prominent, remainder of the veins fl at, those above secondary rank obscure, sparsely pubescent on the midvein at base, glabrescent distally.
Distribution. Restricted to Mata Atlântica (Atlantic coastal forests) complex in the state of Rio de Janeiro, Brazil.
Ecology. Spondias admirabilis is a small to relatively large tree of semi-deciduous slope forests, moist forests, secondary forests, closed shrubby vegetation, and open rocky areas, between 100-300 m. It is known to fl ower Apr-Nov and fruit Oct-June.
Etymology. Th e specifi c epithet (Latin for "remarkable" or "astonishing") expresses our surprise at discovering a new species of Spondias dispersed among collections of two other species from a region that is relatively well-known. Conservation status. We propose to classify this species as Endangered, with the following justifi cation: It is restricted to Rio de Janeiro state, it is a lowland species in a region where the lowland forests are highly fragmented, and it is not common, considering that it is represented by relatively few collections in a region that has been rather well sampled botanically.

Spondias dulcis
Leafl et venation: Fimbrial vein absent; secondary veins 12-20 pairs, usually darker than blade abaxially, usually straight and nearly perpendicular to midvein, spacing regular or sometimes decreasing toward base, angle increasing toward apex and base, insertion on midvein decurrent; intersecondaries ca. 1 per pair of secondaries and parallel to them, long and straight; intercostal tertiaries few, principally admedially branching parallel to secondaries but some irregular-reticulate, also some admedial tertiaries branching from intramarginal vein; quaternaries irregular-reticulate, FEVs highly branched, dendritic, tracheoid idioblasts absent; marginal secondary present; on abaxial side the midvein prominulous to prominent, secondaries fl at; on adaxial side the midvein prominulous, secondary veins impressed to prominulous.
Distribution. Broadly cultivated in lowland moist forest regions throughout the Neotropics.
Ecology. Given this species' broad distribution, its known phenology is broken down by region. Conservation status. Considering that this taxon is native to Asia/Oceania and rather widely cultivated in tropical America, it can be considered of Least Concern, at least for the Neotropics.
Discussion. According to Smith (1985), the type specimen was made from a plant cultivated in Mauritius but grown from seed brought by Commerson from Tahiti in 1768. Th e earliest eff ective publication of Spondias dulcis is by Parkinson (J.voy. South Seas, 1773), a botanical artist who accompanied Banks and Solander in Captain Cooks's fi rst expedition to the Southern oceans. A later publication of this name by G. Forster (Pl. esc. 33. 1786) is considered an isonym as it is based on the same type as Parkinson's name.
Spondias cytherea Sonn. was once considered the earliest valid name of this species, on the basis that the names in Parkinson's publication were considered invalidly published (Airy Shaw and Forman 1967). A more recent examination of Parkinson's work has shown S. dulcis to be a valid name and therefore to have priority over S. cytherea (Fosberg 1960). Following this argument, Smith (1985) lectotypifi ed S. dulcis with a collection considered the voucher of Parkinson's illustration of this species (and therefore a typotype).
Leafl et venation: Fimbrial vein absent; secondary veins 7-14 pairs, spreading, the spacing and angle sometimes irregular, the angle increasing toward the base, insertion excurrent or abruptly decurrent; some perpendicular inter-secondaries present, also sometimes some perpendicular epimedial tertiaries present; intercostal tertiary veins irregular-reticulate with some admedial branching; areoles poorly developed, mostly at tertiary rank, FEVs 4+-branched, dendritic, terminating in tracheoid idioblasts; marginal ultimate venation incompletely looped; on abaxial side the midvein and secondary veins narrowly prominent, tertiaries prominulous and slightly darker than the lamina, the midvein with sparse trichomes, rest of blade with scattered trichomes of all three types; on adaxial side the midvein narrowly prominent, the secondary and tertiary veins prominulous to fl at, the midvein and secondary veins with dense to sparse trichomes and sometimes scattered glandular hairs.
Distribution. Mata Atlântica (Atlantic coastal forests) complex in the states of Minas Gerais and especially Espírito Santo, Brazil.
Ecology. Th ree of the four known collections were made in pasture, a coff ee plantation, and a secondary forest, so although it appears to be rare, the species is likely adapted to disturbed conditions. It has been collected in fl ower in Oct-Dec and in fruit in March.  Conservation status. We propose to classify this species as Critically Endangered, with the following justifi cations: (1) we are aware of only four herbarium collections from only two distinct localities; (2) it is evidently rare considering how few collections have been made; and (3) there is very little forest remaining in these collecting localities.
Spondias expeditionaria appears as S. macrocarpa in Árvores Brasileiras (Lorenzi 1998), Brazilian Trees (Lorenzi 2002), and Brazilian Fruits and Cultivated Exotics , all published before discovery of the new species. Th e plates show habit, bark, wood, a fl owering branchlet, and mature fruits and endocarps. Diagnosis. Canopy or emergent tree to 40 m tall, inner bark red with narrow white striations; similar to S. mombin because of the similar indumentum, the infl orescences highly branched, disk short and thick, and fruits of similar size; S. globosa diff ers by the outer bark lacking spinose projections (vs. corky, tubercular, or spinose projections), intersecondary veins parallel to secondaries and strong, often reaching intramarginal vein (vs. intersecondaries reticulating and weak, not usually reaching intramarginal vein), intercostal tertiaries arising at or near intramarginal vein (vs. intercostal tertiaries primarily irregular-reticulate), fruits globose to perdepressed-ovoid, rarely very slightly oblong of obovoid (vs. oblong or less often ellipsoid or slightly oblong-ovoid).
Leafl et venation: Fimbrial vein absent; secondary veins in 9-15 pairs, straight to slightly arcuate, the spacing decreasing toward apex and base, the angle decreasing toward the apex and increasing toward base, insertion on midvein decurrent; intersecondaries occasional, parallel to secondaries and almost reaching the intramarginal vein; intercostal tertiaries few, most of them arising from near the intramarginal vein and forming strong composite admedials parallel to secondaries, with some irregular reticulation; quaternaries irregular-reticulate and freely admedially ramifi ed; areolation usually at quaternary rank, FEVs 5+-branched, dendritic, tracheoid idioblasts absent; marginal ultimate venation usually looped (sometimes incompletely); on abaxial side all veins narrowly prominent or sometimes the secondaries and tertaries prominulous, occasionally discolorous; on adaxial side all veins narrowly prominulous to almost fl at or occasionally all but the midvein slightly impressed; on both surfaces the midvein with scattered longer hairs and bristles near the base and (sub)glabrous distally, sometimes glabrescent.
Distribution. Spondias globosa is a western Amazon element, apparently disjunct to Zulia and Barinas in western Venezuela.
Ecology. Th is is very much a lowland taxon, ranging only between 100-350(500) m elevation. It is most often found in formations such as fl oodplain forests or tahuampa forest on poorly drained, periodically or seasonally inundated soils, although it has been reported from a range of soils including not just black alluvial soils but also oxisols, lateritic soils, and red and yellow clay soils. Apart from fl ooded formations it is found in primary forests in well-drained soils, including on undulating or hilly terrain. Occasionally it grows in secondary forest, bamboo-dominated forest, or rarely pasture or shrubby disturbed vegetation.
In SW Amazonia, this species is known to fl ower in Sep-Nov and fruit Oct-May, but in NW Amazonia the collection data indicate that it can be found fl owering and fruiting all year.
Th e yellow-footed tortoise, Geochelone (Chelonoidis) denticulata, has been observed dispersing the fruits of S. globosa (as S. venulosa) in Amazonian Peru andColombia (Rodríguez-Bayona andRylander 1985 andStevenson et al. 2007, respectively). Th ese fruits comprise an important food source in times of scarcity for primates such as the woolly monkey (Stevenson 2005). Conservation status. Th is taxon is widespread in Amazonia and can be considered of Least Concern except in Zulia, Venezuela (Maracaibo watershed), where very little lowland forest remains and where it has been collected only once.
Discussion. Although Spondias globosa is geographically sympatric with S. mombin in many localities, the two appear to have undergone niche partitioning: in interviews with forest residents in the middle Ucayali and upper Purus rivers, they readily recognized the two as distinct taxa long before botanists came to the same conclusion, pointing out not only diff erences in the bark and fruits but also that S. globosa tends to keep to the fl oodplains (vs. terra fi rme) and fl owers and fruits later in any given locality. In the middle Ucayali the prevailing common name for S. globosa is "uvos colorado," referring to its mostly red (versus usually pale pink) inner bark.
Distribution. Spondias macrocarpa is native to moist upland forests of the Mata Atlântica Complex, in southern Bahia, Rio de Janeiro, Espírito Santo, and extreme southeastern Minas Gerais.
Ecology. Th is species appears to be rare where it does occur. It has been recorded in mussununga forest (dense forest with discontinuous canopy 8-15 m high, on level terrain, in sandy soils (spodosols) that are often poorly drained) (Stefano et al. 200,NY), and tabuleiro forest (dense forest with continuous canopy 20-25 m high, on level terrain, in sandy clay soils (oxisols)) (Stefano et al. 225,NY). Th e species is known to fl ower Jun-Feb and to fruit Mar-Apr (Jun).  Conservation status. We classify this species as "Vulnerable," although it is relatively widespread in Atlantic Forest Complex of Brazil, because of the relatively small number of known specimens represented in a well-collected region. It should be noted that many of the collections are rather old and may be from localities that are no longer forested.
Discussion. Th e species referred to as Spondias macrocarpa in Árvores Brasileiras (Lorenzi 1998), Brazilian Trees (Lorenzi 2002), and Brazilian Fruits and Cultivated Exotics ) is in fact Spondias expeditionaria; these books were published before discovery of the latter. Indeed, S. macrocarpa most closely resembles S. expeditionaria; a comparison of the two appears under the latter species.
Leafl et venation: Fimbrial vein present; secondary vein pairs 10-16, mostly straight, the spacing slightly decreasing near base only, the angle decreasing toward apex and increasing toward base, insertion decurrent; inter-secondaries present, average <1 per pair of secondaries and parallel to them, longer than halfway to margin but zig-zagging; intercostal tertiaries irregular-reticulate and admedially ramifi ed, quaternaries irregular-reticulate and admedially ramifi ed, areolation at tertiary and quaternary ranks; FEVs highly branched, dendritic, some slight thickening; fi mbrial vein present; on abaxial side the midvein and secondaries prominent (secondaries rarely prominulous to fl at), glabrous except midvein and secondary veins sometimes sparsely pubescent; on adaxial side the midvein prominulous or rarely fl at, secondaries prominulous to impressed, glabrous.
Distribution. Spondias mombin is widely cultivated in the moist tropics, but it is native in Mexico south to SE Brazil; it may be native to E Brazil but this is uncertain.
Ecology. Native populations of S. mombin occur in tropical moist to semi-deciduous forests as well as gallery forests and forest islands in savannas, less often in fl oodplain forests (e.g., Little 8092, NY; see Peters and Hammond 1990); one collection is from white-sand dunes (Prance & Silva 24231, NY). In Central America, S. mombin has been considered a relatively early-successional species (Nason and Hamrick 1997). In drier or more open conditions, the bark tends to be thicker and to produce spinose projections; reportedly it insulates against fi re damage to the cambium (Pinard and Huff man 1997).
Given this species' broad distribution, its known phenology is broken down by region. Mexico In Rio de Janeiro state, Brazil, S. mombin is evergreen (Rodrigues and Samuels 1999), whereas in other parts of its range, such as C Panama and NW Costa Rica , it can be facultatively deciduous for up to two months (Croat 1974a andJanzen 1985, respectively). In Guanacaste Province, NW Costa Rica, the species fl owers toward the end of the 6-month long dry season (late April to early May). In C Panama, fl owering can range Feb-May, and the local period of fl owering is ca. two months, with the trees of any given population hightly synchronized (Adler and Kielpinski 2000). At that same site, fruits required approx. fi ve months to mature, and they ripened Jul-Oct with a peak in Aug-Sep. Th e fruiting season tends to be highly regular, but fruit production varies greatly among years (Milton et al. 2005).
Common names. As noted, Morton (1981) listed 96 diff erent common names for S. mombin. Th is species has been recorded as being called jobo in Belize, Cuba, the Dominican Republic, northern Colombia, Ecuador (Esmeraldas), Guatemala, Nicaragua, Mexico (Oaxaca, Veracruz), Panama, and Puerto Rico, and called hog plum in Belize, Jamaica, Tortola, and Trinidad and Tobago. Th e species is generally called taperebá (taperibá) in Brazilian Amazonia but more commonly called cajá or cajazeiro in the rest of Brazil (Ducke 1946, Smith et al. 2007). Other common names include the following. Bolivia Economic botany. Th e range and habitats of introduced S. mombin overlap signifi cantly. Th e species has a long pre-Columbian history of use (e.g., Ducke 1946); carbonized endocarps are abundant in middens of the extinct Marajoara culture of Marajó Island at the mouth of the Amazon (Roosevelt 1991, as Spondias lutea), and it is described and illustrated (as caia) in Frei Cristóvão de Lisbôa's História dos Animais e Árvores do Maranhão (1968, fascimile of ca. 1625 manuscript). It continues to be an important plant resource in Amazonia (Smith et al. 2007). In the West Indies, it was probably introduced, as suggested by its occurrence primarily in disturbed areas. It is well-established as an invasive species in tropical West African forests and savannas (Ghazanfar 1989). It is more commonly cultivated in tropical Africa than S and SE Asia (Kostermans 1991).
Most collections that cite uses note the edible fruit. Th e species can be dominant in some periodically fl ooded riverine habitats (Peters and Hammond 1990), and individual trees can be highly productive, producing up to 10,000 fruits per tree (Adler and Kielpinski 2000). Th e second most reported observation is that the fruits of S. mombin are eaten by game animals and in these cases S. mombin serves as a "waiting tree" where locals go to hunt in the fruiting season. Th e animals include Ateles (Croat 12291, NY), Alouatta and Cebus monkeys (Balée 3380 and 2569, NY, respectively) as well as yellow-footed tortoises and pacas (Balée 2033, NY), tapirs (Ayres and Ayres 1979), and toucans (Miller et al. 217,NY). Ayoka et al. (2008) provided a useful review of the economic and traditional uses of the species. Th e primary use of the species is for its fruits, reportedly high in vitamins C and B1 (Keshinro 1985, Bora et al. 1991. Th e pulp is stewed, or made into preserves, or used to prepare juices and alcoholic beverages (fermented or for fl avoring) (Cavalcante 1976; one fermented product is referred to in Brazil as vinho de taperebá. (Severo et al. 2007) Th e juice is available in restaurants and foodstores in Brazil, and the frozen pulp is commercialized throughout the country. Th e tree is commonly planted as a living fence or in home gardens, or planted for shade and food for livestock (Morton 1981). In the Dominican Republic the fruits are fed to pigs (Zanoni et al. 3028,NY).
Other important uses of Spondias mombin are in traditional medicine (see review in Duke et al. 2009), both in its native range and where it has been introduced. Th e ethnobotanical literature and herbarium specimen labels provide many accounts of the uses of its roots, leaves, fl owers, fruits (rarely) and especially bark for medicinal purposes, to treat myriad medical problems such as wounds, fever (Balée 2569, NY), dysentery, vaginal bleeding, genital ulcers, respiratory conditions, intestinal and digestive ailments, (Morton 1981;Grenand et al. 2004), malaria (Milliken 1997), leishmaniasis (Fleury 1991), colds (Yuncker 17086, NY), and as a contraceptive and abortifacient (Offi ah andAnyanwu 1989, Schultes andRaff auf 1990). Most often the preparations are infusions or decoctions that are either ingested or applied to the aff ected area. In Amazonia, the inner bark is ground into a powder and used as a disinfectant for wounds, and the boiled powder is used as an oral disinfectant (Nelson 749, 785, NY). In Nicaragua, a decoction of bark and leaves is used to treat malaria, diarrhea, infections, skin rashes, and sores (Coe 2275, MO).
Conservation status. We consider this species to be of Least Concern because of its broad range and often large populations, moreover it is widely cultivated.

Spondias purpurea
Leafl et venation: Fimbrial vein absent; secondary veins 5-10 pairs, usually arcuate, spacing decreasing toward apex and increasing toward base, angle acute and slightly irregular but increasing toward apex and decreasing toward base, insertion excurrent or decurrent; intersecondaries average <1 per intercostal area, perpendicular to midvein and slightly acrofl exed; epimedial tertiaries 1 per intercostal area, perpendicular to midvein; intercostal tertiary fabric irregular-reticulate and strongly admedially ramifi ed; quaternaries irregular-reticulate and freely ramifi ed, areolation at tertiary and quaternary ranks, FEVs highly branched, dendritic, with some terminal thickening; marginal ultimate venation mostly looped; on abaxial side the midvein and secondaries fl at to prominulous, glabrous except midvein often sparsely pubescent toward the base, the surface sparsely to densely micro-pustulate, on adaxial side the midvein prominulous and secondaries fl at to prominulous, glabrous except the midvein sparsely to densely pubescent.
Distribution. Spondias purpurea naturally grows in tropical deciduous forests from NW Mexico to Panama and possibly N Colombia; its cultivated distribution extends well beyond its native distribution, and its true range is complicated by its long association with humans. In some areas where the species has been assumed to be native, some evidence suggests that it has been introduced; for example, Janzen (1985) reported that S. purpurea and S. mombin co-occurred in Guanacaste Province in Costa Rica, but he observed that the former was usually found near roads and trails. Miller (Miller and Schaal 2005) and others have noted that S. purpurea produces fruit parthenocarpically where it is not native, and it is known that the species is propagated asexually in many places, but the observations of Janzen and others suggests that it can reproduce sexually outside its native range.
Ecology. Spondias purpurea grows in highly seasonal tropical (semi-)deciduous forests (Miller and Knouft 2006), and it appears to thrive best in dry conditions; for example, at Chamela Biological Reserve in Jalisco, Mexico, it was much less abundant in semi-deciduous forest than in adjacent deciduous forest (Mandujano et al. 1994), where it is sympatric with Bursera spp., Cyrtocarpa procera and Comocladia sp. (Lott 2002). Still, it thrives in a broad range of habitats and soil types, and its cultivated range is primarily in more humid environments. It has been found (probably cultivated) at elevations up to 1200 m in both Mexico and Ecuador. Given this species' broad distribution, its known phenology is broken down by region. Mexico: fl owering Dec-Aug, fruiting May-Sep; Central America: fl owering Dec-Sep, fruiting Mar-Oct; NW South America west of the Andes: fl owering Feb-Oct; remainder of N South America: fl owering (Jan) Sep-Nov.
Several studies in Mexico have shown that this species is markedly deciduous and fl owers when leafl ess, while the fruits mature near the interface of the dry/leafl ess and wetter/leafy seasons. In Sinaloa, the trees were leafl ess from Jan-May, fl owering in Feb-Mar and fruiting in Jun, and in Puebla the trees were leafl ess Jan-Apr, fl owering Dec-Jan and fruiting Apr-May (Cuevas 1994), while in Chamela (Jalisco) the trees were leafl ess Nov-Jun, fl owering in Feb and the fruits maturing in May (Bullock and Solís-Magallanes 1990).
Spondias purpurea dispersal was studied in detail by Mandujano et al. (1994) in a deciduous forest in Chamela, Mexico. Th ey found that the major dispersers were white-tailed deer, collared peccaries (see also Martínez-Romero and Mandujano 1995), coatis, gray foxes, chachalacas, orioles, and ctenosaurs. At Chamela there was no evidence of bat dispersal, although Lobova et al. (2009) cite references containing observations of this species being dispersed by several species of bats. White-faced capuchin monkeys were observed dispersing fruits of S. purpurea at Santa Rosa National Park in Costa Rica (Freese 1977).
Economic botany. Some cultivated populations preserve genetic diversity of the species which may have been lost from wild populations (Miller and Schaal 2005); this is due to the highly fragmented and reduced extent of tropical dry forests in Mexico and Central America (Mooney et al. 1995). In addition to being introduced and possibly naturalized in tropical America, it has also been introduced in the Paleotropics, especially in the Philippines (not coincidentally a former Spanish colony).
Most cultivated populations of S. purpurea are apparently parthenocarpic (Juliano 1932, Miller andSchaal 2005); the major means of propagation is by stem cuttings (Cuevas 1994, Macía-Barco andBarfod 2000). Th e species is sometimes grown in large orchards, but it is most commonly planted as a living fence or as individual fruit trees in home gardens.

Spondias radlkoferi
Leafl et venation: Intramarginal vein present or sometimes appearing to have a marginal secondary, occasionally hidden by revolute margin. Secondary veins in 5-12 pairs, often arcuate but straight near base, spacing decreasing toward apex, the angle almost uniform but decreasing toward apex; insertion on midvein abruptly decurrent or less often excurrent; some inter-secondaries present, 0-1 per pair of secondaries and usually perpendicular to the midvein, long and reticulating or basifl exed; epimedial tertiaries present, short, parallel to secondaries or perpendicular to midvein, reticulating; intercostal tertiaries alternate-percurrent and irregular-reticulate with some admedial branching; quaternaries irregular-reticulate and freely ramifi ed, areolation at tertiary or quaternary ranks, FEVs 2-3-branched, dendritic, terminating in highly branched sclereids; on abaxial side the midvein and secondaries prominulous to prominent and discolorous, higher-order veins fl at to prominulous, on adaxial surface the midvein narrowly prominulous and the secondaries and higher-order veins fl attened to slightly impressed, on both sides the midvein and secondary veins often with dense to scattered curved hairs, rest of surfaces glabrous or with sparse to scattered hairs, glabrescent with age except abaxial surface usually with hairy-tuft domatia in the axils of secondary veins.
Distribution. Spondias radlkoferi has been recorded from Mexico (Mexico State) S to NW Colombia and Venezuela (Zulia), with one record from Los Ríos in Ecuador. As noted above, diff erent leafl et forms are associated with diff erent parts of its range.
Ecology. Th is species is rather versatile ecologically, growing in primary to secondary formations or even roadsides, in seasonally dry tropical forest, tall evergreen forest, and pluvial forest, on limestone, black clay, and reddish brown stony soils. It occurs on slopes and in valleys at elevations ranging from 10-1000 m.
Given this species' relatively broad distribution, its known phenology is broken down by region. Mexico: fl owering Mar-May, fruiting May-Dec; Central America: fl owering Dec-Jul, fruiting May-Jan.
In central Panama, populations of S. radlkoferi often fl ower 4-6 weeks later than S. mombin populations (Croat 1974a). Th e species is known to fl ower Apr-Jul (peaking in May-Jun) and to fruit Sep-Dec with a peak in Oct-Nov. Croat and others have suggested the fruit becomes an important source of food for mammalian species in times of food scarcity (Croat 1974a, b) and/or forest fragmentation (spider monkeys; Chaves et al. 2012).
As with the other Spondias species whose dispersal has been documented, the fruits of S. radlkoferi are often dispersed by frugivorous bats (Bonaccorso 1978, Bonaccorso and Humphrey 1984, Medellín and Gaona 1991; the endocarps of this species accounted for 50.5% of diaspores collected beneath the leaf tents of the tent-making bat Artibeus watsoni (Melo et al. 2009). Conservation status. We consider this species to be of Least Concern because of its broad range and relatively large populations in Central America, S Mexico and Colombia, although the population(s) in the heavily deforested region of Zulia, Venezuela may be endangered.
Discussion. Spondias radlkoferi most closely resembles S. mombin because of its usually densely fi ssured bark, leaves 3-14-jugate, the midvein of the leafl et usually prominent abaxially, and the petals glabrous abaxially. Th e former can be distinguished by the intramarginal secondary vein (sometimes (sub)marginal) (vs. always removed from the margin); the costal secondary veins usually distinctly arcuate with excurrent insertion on midvein, sometimes with hairy tuft domatia in the axils abaxially (vs. essentially straight to very slightly arcuate with decurrent insertion on midvein, without hairy tuft domatia); the pedicel 0.7-2.5 (3.5) mm long (vs. 2-4.5 mm); and the fruit maturing green (rarely orange), obovoid with abruptly short-acuminate apex (vs. maturing yellow or orange(-brown), oblong to ellipsoid to globose, apex rounded to truncate).
Leafl et venation: Fimbrial vein absent; secondary veins in 10-20 pairs, straight, attenuate at both ends, spacing uniform, angle nearly perpendicular, insertion on midvein abruptly decurrent; intersecondaries and epimedial tertiaries absent; intercostal tertiaries with some irregular-reticulate veins but primarily composite-admedial departing from secondaries or the intramarginal vein, usually attenuate at both ends; quaternaries irregular-reticulate and freely ramifi ed, areolation at tertiary and quaternary ranks, FEVs 3-4+-branched, dendritic, terminating in only slightly thickened tracheoid idioblasts (see Fig. 3, p. 253 in Silva 1973); marginal ultimate venation looped; on both surfaces the midvein narrowly prominulous (seldom prominent abaxially or fl at adaxially), secondaries and higher-order veins fl at to prominulous on both sides, sometimes impressed adaxially, the surface glabrous or when young with dense fl exuous hairs, often glabrescent except hairs persisting on midvein and toward leafl et base; adaxial side glabrous or when young with scattered fl exuous hairs, glabrescent with age.
Distribution. Th e native range of Spondias tuberosa is from Maranhão E to Paraíba and south to Minas Gerais in Brazil; it is also cultivated in SE Brazil (see below).
Ecology. Th is species is a constituent of the arid caatinga in NE Brazil, known to fl ower in Aug-Mar and fruit Oct-May (Machado et al. 1997, plus data from exsiccatae). Nadia et al. (2007) suggested that this species can be andromonoecious, and × 1.3-2 cm); more signifi cantly, the stony endocarp is overlain by a thinner fi brous matrix, moreover the fi ve peri-apical pores are subequal in size (vs. four small and one larger), larger than in S. tuberosa, and oblong (vs. circular), and the pyrene has four (vs. 1) keels or trabeculae (beams)(based on Mattos Silva et al. 2299, NY). Th e fresh fruit is ca. 4-4.7 × 3.2-3.4 cm and the pyrene 2.7-2.8 × 1.8-1.9 cm ).
On the other hand, the vegetative and fl oral morphology of the material referred to umbu-cajá is almost indistinguishable from those of S. tuberosa, although the leaflets tend to be larger and more broadly ovate, and the apex to be narrowly acuminate. Th e revolute leafl et base with rather dense long hairs more closely resembles that of S. venulosa. We are not aware of any diff erences in habit or in bark morphology.
Th e fruits of umbu-cajá are consistently obovoid, but this is within the range of variation of S. tuberosa s.s.; the fruit surface is sparsely lenticellate, which we have not observed in S. tuberosa s.s. (Marlon Marchado, pers. comm., 4/2013).
In his book on Brazilian fruits, Lorenzi et al. (2006) treated umbu-cajá (e.g., Lorenzi 6074, NY) as a separate entity from S. tuberosa, observing that it produces only sterile seeds, that it is propagated only by cuttings, and that it is known only in cultivation in Bahia, Alagoas, and Pernambuco states. Silva Júnior et al. (2004) suggested that umbu-cajá is a hybrid between S. tuberosa and S. mombin, but Almeida et al. (2007) examined karyology and genomic in situ hybridization of Spondias spp. and concluded that the umbu-cajá (1) is not of hybrid origin and (2) is distinct from both putative parents. In the present treatment, this entity is considered a variant of S. tuberosa unless and until further genetic/molecular investigations suggest otherwise.
Distribution. Spondias venulosa ranges from east-central Bahia to southern Rio de Janeiro and extreme southeastern Minas Gerais; cultivated as far south as Campinas and São Paulo city in São Paulo state.

Acknowledgments
Alejandra Vasco helped develop the techniques that yielded the revealing leaf clearings. Bobbi Angell executed the superb illustrations. Maria Cristina Martínez-Habibe produced the fi ne distribution maps. John Pruski tracked down and sent scans of obscure literature. Harri Lorenzi shared images and specimens. Allison Miller provided valuable comments and specimens of wild and cultivated populations of Spondias purpurea. Susan Pell provided unpublished phylogeny data on Spondias and Spondioideae. Wayt Th omas provided useful specimens and silica-dried leaf material of Spondias species from Eastern Brazil. Robbin Moran and J. Hamrick provided insights on hybridization between Spondias mombin and S. radlkoferi. Geovane Siqueira generously provided logistical support in the Reserva Natural Vale and sent invaluable duplicates from the CVRD herbarium. Cássia Sakuragui provided valuable logistical support in Rio de Janeiro and Espírito Santo states. Scott Mori and Carol Gracie took informative photographs of their Spondias collections. Julien Bachelier provided valuable discussions about the developmental anatomy of Spondias fl owers. Th e following herbaria generously sent gifts and/or loans: BM, CAY, CEPEC, G, GH, K, L, LSU, MO, P, K, R, RB, TEX, U, US, W. Part of the junior author's research was supported by NSF grant no. 0918600. Finally, we thank the anonymous reviewer for his/her careful edits and helpful suggestions. Pedro Acevedo showed great patience, persistence and precision in his intensive reviews of the manuscript.