The flowers of Duckesia verrucosa have (22–)25 stamens of two types, including five fertile antepetalous that each alternate with a group of 4–5 sterile on less stout filaments of varying lengths. One stamen per sterile group is usually as long or longer (clearly the medial antesepalous stamen when five per group) than the fertile stamens and the rest shorter. The filaments are subulate, complanate, basally very short-connate, proximally short-papillose, and distally smoother (especially so on the fertile stamens). Abnormally short filaments (1/3–1/2 of the normal length) lacking anthers rarely substitute for sterile stamens; these are likely the occasional staminodes reported by Cuatrecasas (1961). The filaments each contain a vascular bundle that continues into the connective protrusion for nearly its entire length, and is distally branched in the fertile but unbranched in the sterile anthers. Anthers are dorsifixed below midlength, glabrous, lack stomata, caducous, and have elongate connective protrusions, which are especially thickened and fleshy in the fertile anthers. The fertile anthers are tetrasporangiate, with a pair of larger sporangia (pollen sacs) positioned dorsally at the level of the filament insertion, and a pair at the base that are slightly smaller and nearly perpendicular to the dorsal pair (Fig. 1A, B). Dehiscence of the four thecae is by flap-like valves that open from the ventral side for each lateral sporangium and move dorsally (outward), or for a basal sporangium move from distal to proximal (drop downward; Fig. 1B). The sterile anthers consist of connectives with protrusions that are thinner than those of the fertile anthers and have little trace of sporangia (Fig. 1F); those connectives can be prolonged and misshapen due to packing in bud and form filling around the fertile anthers. The 20 stamens of Duckesia liesneri comb. nov. are of two types with five long antesepalous and 15 short. The filaments are subulate, complanate, connate for up to 1/3 of their length, smooth, and contain a vascular bundle that continues into the connective protrusion and is usually unbranched (rarely with short stubby branches distally). The anthers are all fertile, tetrasporangiate and similar to D. verrucosa; however, the pairs of sporangia lack size differentiation and are close together rather than well-separated (Fig. 1C). In bud, both species have stamens with straight filaments that are longer than the anthers.

Figure 1. 

Stamen structure of Humiriaceae. A Duckesia verrucosa tetrasporangiate anther, dorsal B Duckesia verrucosa tetrasporangiate anther, lateral C Duckesia liesneri tetrasporangiate anther, lateral D Endopleura uchi disporangiate anther, lateral E Endopleura uchi tetrasporangiate anther, lateral F Duckesia verrucosa sterile anther, lateral G Humiriastrum cuspidatum disporangiate anther, lateral H Humiriastrum cuspidatum androecium, dorsal I Endopleura uchi tetrasporangiate anther, lateral J Humiriastrum dentatum disporangiate anther, lateral K Humiriastrum dentatum disporangiate anther, ventral L Humiriastrum diguense disporangiate anther, lateral M Humiria balsamifera var. imbaimadaiensis stamen cluster with 2 of 3 types, dorsal. f = filament attachment location. Sources: A, B, F Ducke 2108 C Henderson 933 D, E, I Assunção 605 G, H Cid et al. 4264 J, K Hatschbach 56145 L Quizhpe et al. 612 M Wurdack 4814 (all US.)

The flowers of Endopleura have 20–22 stamens (to 30, fide Cuatrecasas 1961) of two types. Five antepetalous and five antesepalous long stamens are tetrasporangiate with separated lateral (dorsal) and basal pairs of sporangia (Fig. 1E, I). The 10+ short anthers alternating with the long type are usually disporangiate by missing the lateral pair, or rarely with no (sterile) or three sporangia (Fig. 1D). Filaments are subulate, complanate, basally very short-connate, and short-papillose nearly their entire length; each contains a vascular bundle that continues into the connective protrusion where it is unbranched or distally branched (branching does not appear correlated with stamen type). In bud the filaments are straight with slightly sinuous tips for the longer type (likely due to packing), and as long or longer than the anthers. Anthers are dorsifixed below midlength, glabrous, versatile, lack stomata, caducous, and variously shaped due to distortions of the usually thickened, elongate connective protrusions from packing in bud (e.g., Fig. 1D with impressions of adjacent sporangia indenting the connective protrusion). The four thecae and dehiscence lines resemble those of Duckesia.

The flowers of Humiria have 20 stamens of three types, including five long antesepalous, 10 short and adjacent to antesepalous, and five antepetalous of intermediate length. Two (Cuatrecasas 1961) or three (Narayana and Rao 1973b) anther types have been previously reported, and we observed three types across multiple species and varieties examined. The filaments are subulate, complanate, and connate up to half their length to form a smooth androecial tube. The free tips are smooth dorsally but otherwise papillose along their margins where the projections interdigitate, and more sparingly short-papillose ventrally. The filaments of H. balsamifera var. imbaimadaiensis each contain a vascular bundle that continues into connective protrusion for nearly its entire length and is unbranched. Anthers are disporangiate, dorsifixed at or near their base, versatile, lack stomata, and caducous (Fig. 1M). The thickened, elongate apical protrusions are glabrous and variously shaped according to their type and packing in bud; the protrusions of the long stamens wrap over and conform to the stigmas. The two sporangia are coarsely hirsute, slightly divergent, and each dehiscent from the ventral side via a dorsal hinge (opening outward).

The flowers of Humiriastrum (Urb.) Cuatrec. have 20 stamens of two types, including 10 long antesepalous and antepetalous, and 10 short that are adjacent to the long (Fig. 1H). The filaments are subulate, complanate, basally short- to half-connate, and smooth to sparsely short-papillose; each contains a vascular bundle that continues into the connective protrusion for half to nearly all its length, and is unbranched (H. diguense) or branched (H. dentatum). In bud, the filaments are straight and longer than the anthers. Anthers are disporangiate, dorsifixed below midlength, delicately attached, glabrous, lack stomata, caducous, and have large, thickened apical protrusions, which for the long stamens can wrap over the stigmas with tips folded inward. Cuatrecasas (1961: 49) described the “thecae of anthers basal” in Humiriastrum versus “inferolateral” in Sacoglottis and Schistostemon. However, this distinction is blurred as both orientations occur in Humiriastrum. In H. dentatum, the sole species studied by Narayana and Rao (1973a), the two sporangia are slightly divergent (V-shaped), and dehiscent from the ventral side in the manner of the other disporangiate genera (Fig. 1J–K); this configuration also characterizes H. glaziovii. Most of the remaining species in the genus have variations on (sub)parallel (not divergent) basal sporangia. For example, the sporangia of H. cuspidatum (Fig. 1G) are small, positioned at the anther base so as to be perpendicular to the connective long axis, and each dehiscent by a valve opening distal to proximal (drops downward). In H. diguense, the sporangia of the long stamens are oriented parallel to the connective long axis, and each open from the dorsal side via a ventral hinge (Fig. 1L); the short stamens have smaller, often asymmetrically paired sporangia. This dehiscence configuration is opposite of that found in H. dentatum, although it also differs in other details such as sporangia not divergent.

The flowers of Hylocarpa (examined with LM from a single, young bud and additional fragments) have 20 stamens of two types, including 12 fertile and the rest sterile (30 stamens with 5–15 fertile according to Cuatrecasas 1961), whose exact arrangement was not determined. The filaments are straight, not obviously complanate, shorter than the anthers, and smooth; the papillose filament ornamentation reported by Cuatrecasas (1961) may occur later in development. Fertile anthers are disporangiate, dorsifixed, attached in their proximal third, glabrous, and have large, carnose, clavate connectives that are unlike the tapering apical protrusions in the other genera; sterile anthers are slightly larger with connectives that are more irregularly shaped (Fig. 3G). The basal sporangia are small and divergent on the V-shaped connective base.

The flowers of Sacoglottis have 10 stamens of two types, including five long antesepalous and five short antepetalous. The filaments are subulate, complanate, basally connate to form an androecial tube, and smooth; each contains a vascular bundle that continues into the connective protrusion for nearly its entire length and distally usually bears short stubby branches. In bud, the filaments are straight and longer than the anthers. The filament width and length of connation (1/10–2/3 of total length) vary between species and can be diagnostic (see Table 2). Anthers are disporangiate, dorsifixed near their midlength or just below, versatile via a delicate attachment, glabrous, lack stomata, and caducous (Fig. 2A, B). The two sporangia are slightly divergent (V-shaped) with dehiscence lines on the ventral side that extend over both upper and lower shoulders. The stomium of each sporangium is thus eccentric with the elliptic flap-like valve longer than its dorsal hinge which is much thinner than the thickened valve lip (Fig. 2B, C). The anther connective has a thick apical protrusion, which in bud for the short stamens is compressed against and takes on the shape of the bases of the long stamens, leading to a degree of dimorphism between the two anther-types from this space filling effect. Thin staminode-like processes were rarely observed between the stamens of Sacoglottis guianensis (Fig. 2D). The anthers of S. gabonensis are the largest in the genus (to 1.5 mm long for the long stamens; 2 mm fide Cuatrecasas 1961), but structurally otherwise resemble the other species. We reexamined the limited remaining floral material on the type of S. maguirei (anthers were not seen) and could confirm it has a Sacoglottis bauplan with 10 stamens of two alternating types; its filaments resemble those of the other species and are short, wide, and connate half their length. This species otherwise vegetatively more closely resembles small-leaved species of Humiriastrum.

The flowers of Schistostemon (Urb.) Cuatr. have 20 stamens of three types, including five long trifurcate antesepalous, 10 short and adjacent to the antesepalous, and five of medium length antepetalous, and have 30 anthers (Fig. 2E, F). The filaments resemble those of Sacoglottis and are subulate, complanate, basally connate to form an androecial tube, and smooth. However, the five long stamens have a trigonous (in transverse section) free portion of the filament that then distally trifurcates, with each branch bearing an anther such that each stamen has three anthers. The lateral (dorsal) branches in the plane of the staminal tube are shorter, and the central (ventral) branch is longer and bent slightly inward (Fig. 2G). The five trifurcate stamens of S. oblongifolius have three co-lateral vascular bundles per filament that diverge at the branched apex with one bundle serving each anther, while the remaining 15 stamens each have a single bundle; each vascular bundle continues into the connective protrusion to nearly the tip and distally usually bears short, stubby branches (Fig. 3E). The disporangiate anthers of Schistostemon are very similar in morphology to those of Sacoglottis, including in details of attachment, space-filling protrusion dimorphism, branched vasculature in the apical protrusion, slight sporangia divergence (V-shaped), and dehiscence by valves that open from the ventral side and move dorsally (outward). However, the lateral anthers of each trifurcate stamen are slightly reduced in size relative to the central one.

Figure 2. 

Stamen structure of Humiriaceae. A Sacoglottis perryi stamen cluster of 2 types, ventral B Sacoglottis perryi stamen cluster of 2 types, dorsal C Sacoglottis perryi short-stamen anther with open stomium and pollen D Sacoglottis guianensis androecium with interstaminal staminodes (st), dorsal E Schistostemon macrophyllus stamen cluster of 3 types, ventral F Schistostemon macrophyllus stamen cluster of 3 types, dorsal G Schistostemon oblongifolius trifurcate filament tip, dorsal. Sources: A–C Tripp 2984 D Carvalho et al. 4396 E, F Maas et al. 6577 G Maas et al. 6804 (all US).

The flowers of Vantanea vary in stamen number across the species, ranging from 15–28 in V. depleta to >200 in V. maculicarpa Sabatier & J. Engel (fide Engel and Sabatier 2018), and vary in length from 3 mm in V. spiritu-sancti to 35 mm in V. guianensis Aubl. Stamen number can also vary within collections and species, and the positions of different types (lengths) relative to sepals and petals is indeterminate with higher numbers. The filaments are much longer than the anthers, thin, subcomplanate, of multiple slightly different lengths within a flower, usually distally sinuous in bud which is retained at anthesis, and can be basally connate to form discrete whorls or less ordered. In V. peruviana, an inner cluster is differentiated by greater connation and papillose filaments, while the outer stamens are less connate and smooth. In V. micrantha, stamen whorls are poorly differentiated and all filaments are smooth. In species with relatively few, short stamens (i.e., V. depleta, V. spiritu-sancti), the filaments are clearly a single file and complanate; each contains a vascular bundle that continues into the connective protrusion for half its length and is unbranched. Anthers are tetrasporangiate with two thecae, each containing an internal septum that delineates two pollen sacs per theca. They are dorsifixed at midlength to lower third, versatile, glabrous, lack stomata, and caducous; dehiscence is by a single longitudinal slit per theca that opens widely (Fig. 3A–D). Thecae are variable in size, shape, symmetry, and orientation among the species, including elongate, extending most of the length of the anther and lateral (V. compacta), or shorter (to subovate), limited to the lower half of the anther, and slightly divergent (V-shaped in V. depleta). A theca can be (mono)symmetric, or asymmetric due to size differences between the two constituent sporangia or lateral shifts of them relative to each other. The connective protrusions are thickened and elongate, with some variation according to species such that larger protrusions appear correlated with smaller sporangia (see Fig. 3A–D). In V. spiritu-sancti, the 20 stamens are clearly of three alternating lengths in the manner of Schistostemon including five long antesepalous, 10 short and adjacent to the antesepalous, and five of medium length antepetalous. Its subulate filaments are the shortest in the genus at 2–3 mm, basally connate to 1 mm, distally slightly sinuous, and minutely papillose. The anthers are similar in structure to other Vantanea spp., with four sporangia, internal septa, and dehiscence by longitudinal slits (Fig. 3D). In V. depleta we found a variable stamen number with 22–28 for Hammel & Trainer 12954, and verified a lower range of 15–18 for the type collection, McPherson & Stockwell 10892, as had been reported by McPherson (1988). Its stamen structure of alternating lengths with the longest approximately antesepalous resembles that of V. spiritu-sancti, although less fixed.

Figure 3. 

Stamen structure and anatomy of Humiriaceae. A Vantanea compacta anther, lateral B Vantanea compacta anther, dorsal C Vantanea depleta anther, lateral D Vantanea spiritu-sancti anther, lateral E Schistostemon oblongifolius clearing of androecium with 3 stamen types, ventral F Vantanea spiritu-sancti longitudinal section of ovary with 2 superposed ovules per locule; lower left ovule partial G Hylocarpa heterocarpa anthers (left to right): disporangiate dorsal, ventral; sterile, dorsal. Sources: A, B Hatschbach 21265 C Hammel & Trainer 12954 D, F Silva et al. 1436 E Maas et al. 6804 G Ducke [JBRJ-30137] (all US).

Foliar glands were found to be present on 64/65 accepted species and 22 infraspecific taxa of Humiriaceae (85 taxa examined, 100% presence; Appendix 1, Fig. 4). The remaining species not seen, recently described Vantanea maculicarpa Sabatier & J. Engel, also has scattered abaxial glands reported (Engel and Sabatier 2018). Gland distribution ranges from relatively regular within a species, to variable between collections or even leaves on the same specimen. The two basic positional types include: (1) adaxial leaf base (basilaminar) glands as 1–3(7) more or less symmetric pairs spanning the midvein (Fig. 4B, C, F), and (2) embedded laminar glands as rows (Fig. 4A, D, J) or scattered on the abaxial (rarely also adaxial) surface (Fig. 4I). Both types of glands are circular to elliptic, sunken (especially the laminar glands) or slightly raised, have rims that are variously thickened, an outer corona free of stomata, and smooth surfaces without pores (Fig. 4H, K; 5C). In herbarium specimens the gland epidermis can be detached from the underlying gland and appear pale or whitish, which likely indicates a fluid-filled reserves pocket in the living plants (Fig. 4F, J). The extremes of gland size range from relatively large, elliptic (1.3–1.4 × 0.8–0.9 mm on Hylocarpa heterocarpa; Fig. 4I) to tiny, circular (<0.1 mm diameter on Sacoglottis maguirei) structures. The laminar glands are most frequently arranged in a row near or on the margin (Fig. 4D, J) but they can also be in a row at a lesser distance from the midvein (Fig. 4A), or more rarely widely scattered. The glands away from the margins are often clearly associated with higher order veins (i.e., secondaries, intersecondaries) or otherwise intercostal. The thickened lamina often obscure higher order venation that is likely associated with intercostal glands. The glands are very sparse and/or difficult to distinguish in Duckesia and some Humiriastrum (i.e., H. melanocarpum [Cuatrec] Cuatrec., H. columbianum [Cuatrec.] Cuatrec.), and many other taxa have only a subset of gland positions (i.e., either adaxial basilaminar or abaxial laminar). For example, the scattered abaxial laminar glands in Duckesia verrucosa are only a few per specimen with most leaves lacking glands. Damage to basilaminar glands, such as appearing to be chewed, was sporadically observed and may be due to insect predation.

In Sacoglottis the paired adaxial glands usually occur near the base of the leaf blade (Fig. 4F), and details of their position, size, and shape can differ between species. The circular basilaminar glands of S. gabonensis, when present (lacking on some leaves), are unusual in appearing positionally where the first proximal teeth would be along the crenate margin. However, they lack setae associated with teeth and resemble typical basilaminar glands. The other species of Sacoglottis have basal glands (except clearly absent in S. maguirei) that are more typically basilaminar and adaxial rather than marginal. Most of the other foliar glands in Sacoglottis are tiny, in a row near the abaxial margin, and without thickened rims (Fig. 5C), although they can be variable in size, position, and number among species. Sacoglottis guianensis has considerable intra- and inter-specimen variation in the number of foliar glands. The glands of Sacoglottis maguirei are obvious on young unexpanded leaves, and then become nearly invisible small pits on mature leaves. The deciduous stipules in Sacoglottis are morphologically similar to the marginal glandular setae (Fig. 5A, E).

The leaf margins are either untoothed and entire (Humiria, Vantanea), or have glandular setae at regular intervals associated with teeth of varying degrees of prominence (remaining taxa). The non-entire margins are variously described herein (Appendix 1) and in the literature as crenate(-serrate), crenulate(-serrate), or serrate; they are rarely subentire and nearly smooth with little evidence of setae. This terminology reflects a continuum among species in the height and curvature of their small teeth. Most species have asymmetric teeth that are shallow and rounded (crenate), and more rarely have pointed projections with strongly concave distal flanks (serrate). Humiriastrum melanocarpum with leaves described as having “the margin slightly crenate with small glands” (Cuatrecasas 1961: 154) also has typical non-entire margins but more prominent darkened attachments for the setae. The setae appear deciduous in almost all species, usually rapidly so from young expanding leaves, and leave scars with sometimes darkened, glandular-appearing points of attachment to the leaf that can be used to infer their distribution in the absence of rarely collected young leaves (Fig. 4L, M). The setae have fairly uniform morphology with modest variation in size, although those of Duckesia verrucosa are distinctive in smooth cylindrical shape which is not collapsed on drying and greater persistence (Fig. 4E). In Humiria the abaxial glands are in a (sub)marginal row, 0–0.2 mm from the edge, and associated with slight sinuses so as to form a margin best described as repand or slightly undulate (Fig. 4J). This distinctive margin type clearly does not have setae or associated teeth. The few glands at the leaf base of Humiria fruticosa Cuatrec. are precociously exposed (Fig. 4G) prior to full unrolling of the young leaf (supervolute-involute ptyxis), and present an aspect not seen in other Humiria taxa which otherwise all have their usually more distal marginal glands hidden in tight involute laminar rolls. The 14 varieties of Humiria balsamifera Aubl. span considerable variation in leaf features including in marginal gland density and lamina size.

Figure 4. 

Extrafloral nectary and leaf margin diversity of Humiriaceae. A Vantanea depleta laminar glands, abaxial B Duckesia liesneri basilaminar glands, adaxial C Schistostemon oblongifolius basilaminar glands, adaxial D Humiriastrum ottohuberi laminar glands, abaxial E Duckesia verrucosa robust seta at margin F Sacoglottis guianensis basilaminar glands, adaxial G Humiria fruticosa shoot tip with marginal glands exposed on expanding new leaf H Humiria fruticosa marginal gland I Hylocarpa heterocarpa laminar gland, abaxial J Humiria balsamifera var. minarum dense row of marginal glands, abaxial K Duckesia verrucosa laminar gland, abaxial L Schistostemon retusus darkened scar from deciduous seta M Schistostemon retusus intact seta at margin. g = gland, s = seta scar. Sources: A Mori & Kallunki 4889 B Liesner 22589 C Maas et al. 6804 D Maguire 34912 E, K Ducke 2108 F Jansen-Jacobs et al. 1898 G, H Steyermark 103255 I Ducke [JBRJ-30137] J Mexia 5815 L Redden 3372 M Cuatrecasas 7203 (all US).

Figure 5. 

Micromorphology of Sacoglottis perryi. A Marginal glandular seta B basilaminar gland, adaxial C laminar gland near margin, abaxial D sepal tip, inner side with terminal gland E paired glandular stipules and petiole scar F pollen inside sporangium G stigma with ephemeral lobes intact and showing secretion H stigma lobes shredded showing thin walls I gynoecium with diagnostic hirsute ovary J glandular disc with erose margin. Sources: A–C Gillespie 2810 D–J Tripp 2984 (all US).

The 11 species of Sacoglottis are compared in Table 2 (gland details in Appendix 1), which highlights important diagnostic characters such as inflorescence structure (peduncle and bracts), sepal glands, ovary vestiture, and fruit shape. Especially significant is the densely hirsute ovary, a feature that only occurs in S. perryi (Fig. 5I) and Central American S. trichogyna Cuatrec. All other species in the genus have glabrous ovaries. Besides ovary vestiture, no other aspect of morphology suggests a closer relationship between S. perryi and S. trichogyna, and their fruit structures are very different (Herrera et al. 2010: fig. 2N). The leaves of S. perryi are the second smallest in the genus after diminutive S. maguirei, and have a very pronounced drip tip and an shallowly crenate margin (Fig. 6I, 7A). Sacoglottis perryi possesses the full suite of glandular structures known for the family, including small abaxial and basilaminar glands, rapidly deciduous marginal setae, thickened stipules similar to the setae, gland-tipped sepals, a floral disc, and thickened connectives (Fig. 5A–E, J). The minute stigmas appear to secrete and consist of five hollow (presumably fluid-filled in life) thin-walled lobes that rapidly degrade atop a sturdy, persistent style (Fig. 5G, H). This unusual morphology appears typical for the genus and the family in general. The small, subspheroidal, tricolporate, microreticulate pollen of S. perryi (Fig. 5F) resembles that of other species in the genus (see Bove and Melhem 2000). Four species of Sacoglottis are now known from Guyana including S. amazonica Mart., S. cydonioides Cuatrec., S. guianensis, and S. perryi. The fruit shape of S. perryi is subglobose and intermediate between globose S. cydonioides and more elongate S. amazonica and S. guianensis. The fruits of McDowell 2993 are more elongate (length:width ratio 1.37–1.63:1, mean = 1.47, SD = 0.108, n = 4) than those typical for S. perryi, but in other characters the collection agrees with the species. The endocarp morphology of S. perryi does not resemble that of a putatively undescribed Sacoglottis only known from well-weathered ocean drift endocarps (van der Ham et al. 2015).

The biogeography of Sacoglottis includes disjunctions with clear long-distance dispersal to Africa for S. gabonensis and to Isla del Coco far off the Pacific coast of Costa Rica for endemic S. holdridgei Cuatrec., and there is direct evidence of contemporary and fossil sea drift (Herrera et al. 2012, 2014; van der Ham et al. 2015). Mostly Amazonian Sacoglottis amazonica occurs in Delta Amacuro, Venezuela and reaches the nearby island of Trinidad. Otherwise the genus and indeed the entire family oddly lack a West Indian presence (Acevedo-Rodríguez and Strong 2012). Sacoglottis perryi is geographically relatively isolated from other species in the genus. It only overlaps with S. guianensis based on a single collection (Gillespie 2939, US) from the vicinity of Imbaimadai that has typical elongate fruit and a long-pedunculate inflorescence. Otherwise, S. guianensis occurs further south in the Essequibo River watershed (i.e., southern Pakaraima Mountains and Acari Mountains on the border with Brazil) and is broadly distributed elsewhere in northern South America. We reexamined the identity of the Schomburgk brothers’ collections of S. guianensis ([Rob.] Schomburgk, 2^nd ser. 571 but likely mistranscribed for 574, P; 574, F, G, NY, P; Rich. Schomburgk 842, F image ex B, US) from Guyana, as they collected in or near the Pakaraima Mountains, and the lectotype, Schomburgk 574 (Cuatrecasas 1961), is from “Roraima” at the drainage divide of the Mazaruni watershed. All of these specimens have long-pedunculate inflorescences and cannot be confused with S. perryi. The Schomburgk expedition routes, including the approach to Mt. Roraima, skirt around the Mazaruni headwaters and were not within the presently known range of S. perryi (Roth 1923, van Dam 2002). A sterile collection (Guppy 308, NY) from southern Guyana, attributed by Cuatrecasas (1961) to S. amazonica, appears based on leaf similarities to be S. guianensis. The Humiriaceae of Guyana now totals six genera and 13 species (Funk et al. 2007).

The significance of inflorescence structure variation in Sacoglottis has not been previously emphasized, although it is implicit in details of the descriptions and key of Cuatrecasas (1961). His first species-key couplet relates to deciduous versus persistent bract condition, which appears correlated with other aspects of inflorescence architecture. Taxa with persistent bracts have long-peduncled, often large inflorescences, while those with deciduous bracts have short to non-existent (giving the appearance of multiple inflorescences per leaf axil) peduncles and generally smaller inflorescences. The deciduous bracts are rapidly lost, and bract morphology can be hard to document without very young inflorescences. An exception to this inflorescence structure dichotomy is Sacoglottis gabonensis, which has deciduous bracts and a long pedunculate (1–3 cm) inflorescence. Sacoglottis maguirei, a tepui endemic known only from the type collection (Maguire et al. 30693), has been mischaracterized as deciduous (Cuatrecasas 1961), but possesses previously overlooked very short, persistent bracts that are consistent with its long-peduncled inflorescence architecture. The bract condition for the other genera of Humiriaceae is persistent (Endopleura, Humiria, Hylocarpa), deciduous (Vantanea), or mixed (Duckesia, Humiriastrum, Schistostemon).

Schistostemon, presently containing nine species, was formerly a subgenus or section of Sacoglottis and elevated to a genus principally based on differences in androecial structure with a doubling of stamen number to 20, an increase to 30 anthers, and three stamen types instead of two (Cuatrecasas 1961). Of particular note in differentiating Schistostemon are the antesepalous longest stamens that are trifurcate at their apex to support three anthers per stamen (Fig. 2G, 3E). An overlooked difference between the genera is in pistil morphology, which is thin and as long or longer than the stamens in Sacoglottis, but in Schistostemon is stout and shorter than the stamens. Fruit and vegetative details are otherwise strikingly similar between the genera. The renewed suggestion by Herrera et al. (2010) of returning to a broader, pre-Cuatrecasas circumscription of the Sacoglottis-Schistostemon group (i.e., Sacoglottis s.l.) needs further study within the framework of a well-sampled molecular phylogenetic analysis that includes taxa with both types of inflorescence structure. Limited molecular phylogenetic evidence has indicated that Schistostemon is not monophyletic and embedded within a paraphyletic Sacoglottis (Bardon 2015).

Humiriaceae share stamens partly connate, filaments subulate and complanate, and anthers that are dorsifixed, versatile, caducous, lacking stomata, and possessing connective protrusions. Most of these shared androecial features are broadly distributed, even in combination, across the rosids (Endress and Stumpf 1991). The eight genera presently can be clearly defined based on details of androecial structure including number and morphology of both stamens and sporangia, although ovule number and endocarp structure are also important distinguishing generic characters. Our observations on gross androecial structure largely agree with and complement prior floral anatomy studies by Narayana and Rao (1969–1977). Variations in anther sporangium morphology are especially interesting from an evolutionary perspective in Humiriaceae due to transitions among the three distinct forms which include, (1) tetrasporangiate-dithecal with two lateral disporangiate thecae, a typical rosid anther configuration, in Vantanea, (2) tetrasporangiate-tetrathecal with four monosporangiate thecae arranged as two superposed pairs in Duckesia and Endopleura, and (3) disporangiate-dithecal with two monosporangiate thecae arranged as one pair in the remaining five genera. Disporangiate anthers characterize some families of rosids (e.g., Cucurbitaceae, Malvaceae), but are otherwise very sparse in Malpighiales (e.g., cleistogamous flowers of Viola, Violaceae; Endress and Stumpf 1991) except for Humiriaceae.

Vantanea has long been considered the most ancestral genus of Humiriaceae based on androecial structure (Cuatrecasas 1961, Narayana and Rao 1969, 1977, Herrera 2010, Kubitzki 2014), although molecular phylogenetic evidence suggests a nested placement (Bardon 2015). A complex interpretation by Narayana and Rao (1976a, b, 1977) on the origin of the disporangiate anthers based on the position of the sporangia relative to the enlarged connective in bud transverse sections posits that in Humiria and Schistostemon the dorsal (outer) sporangia were lost, while in Humiriastrum and Sacoglottis the ventral (inner) sporangia were lost. We observed the anthers of Sacoglottis and Schistostemon to be nearly indistinguishable when finely comparing anthers of the same type and did not find any transverse shifts in sporangial position to suggest differing origins. Moreover, anther similarities in regard to sporangial position and dehiscence lines extend across all the disporangiate genera (in part for Humiriastrum), although differences exist in connective details. Humiriastrum is now more homogeneous with our generic transfers (i.e., Duckesia liesneri and Vantanea spiritu-sancti); however, the remaining species have two different thecal orientations, which could correspond to different origins (i.e., from dorsal or ventral sporangia). While H. dentatum and H. glaziovii in particular differ from the rest in regard to orientation, by other morphological bases they firmly belong to Humiriastrum and it is premature to reconsider generic affiliation. Whether the disporangiate bauplan has one or more origins needs further testing with floral development and phylogenetic studies. The superposed pairs of sporangia (tetrasporangiate-tetrathecal) in Duckesia and Endopleura have been thought to represent an intermediate state in the transition from tetrasporangiate-dithecal to disporangiate-dithecal anthers (Cuatrecasas 1961). Moreover, they establish homologies and clearly show that the ventral sporangia can assume a basal position with distal-proximal dehiscence. Some phylogenetic evidence has indicated those genera may be embedded within a disporangiate clade and may not be sister groups (see Herrera et al. 2010, Bardon 2015), although we found no major morphological differences to contradict a single origin of the superposed-sporangia anther type. The disporangiate and superposed-sporangia anthers of Humiriaceae, sometimes described as “dehiscing by detachment” (Cuatrecasas 1961, Kubitzki 2014), have markedly eccentric stomia that fit modern definitions of valvate and have similarities in dehiscence lines to established valvate taxa such as Hamamelis L. and Grubbia P.J. Bergius (Hufford and Endress 1989, Endress and Stumpf 1990). Their fundamentally valvate nature has not been noted in prior broad comparative anther surveys and apparently does not occur in any other rosid (see Endress and Stumpf 1990, 1991).

Disporangiate-dithecal anthers are often correlated with specialized floral biology, which can be enabled by restrictive anther openings such as valves (Endress and Stumpf 1990). The pollination biology of Humiriaceae has been only carefully studied for Humiria balsamifera (Holanda et al. 2015), and indicated bee pollination of its nectariferous flowers. For the valvate taxa (i.e., all except Vantanea), further study is needed to see if the anther valves constrain pollen release. The greatest floral variation (i.e., perianth size, stamen number and length, color) occurs in Vantanea with its relatively unspecialized anthers. Vantanea guianensis in particular with its showy red flowers in large clusters is likely hummingbird pollinated and recorded as visited by hummingbirds by Gentry & Stein 46932 (US). However, it differs from typical flowers with that pollination syndrome in not being especially tubular in nature with spreading 30–40 mm long petals, which are shed early before the androecium, and stamen connation for only 5–7 mm of their 25–35 mm length. Variable morphological features that likely affect pollination biology and need further investigation in an evolutionary framework include flower color (white to greenish or red to pink), disc morphology (cupular to variously dissected), filament ornamentation (papillose to smooth), stamen length (long and thin to short and robust), stamen number (10–200+), staminodes (presence or absence), stamen connation (nearly free to half connate), anther ornamentation (hairs or connective elaborations), and mode of dehiscence (longitudinal slits or valvate). Most taxa have elaborated and likely glandular connectives and discs. Stamen connation can affect pollinator nectar access by forming a tube around the disc that varies in height among genera and species (e.g., Sacoglottis spp., Table 2), although nectar robbing by piercing the tube is known for Humiria (Holanda et al. 2015). Anthesis is likely generally of short duration based on the morphology patterns of versatile, caducous anthers and delicate stigmas, although this has been directly little studied (Holanda et al. 2015).

We documented EFNs in all species of Humiriaceae, and found distributional and morphological differences. Some of these differences appear informative at the generic level, including the marginal glands of Humiria, entire eglandular margins of Vantanea, and the exceptionally large laminar glands of Hylocarpa. Our EFN survey, although mostly qualitative in nature, also reveals that their species-level taxonomic value is limited due to high variation in abundance and position, and low intra-specific morphological variation. The EFN ubiquity indicates that they are phylogenetically conserved (i.e., plesiomorphic) for the family, and is a conclusion that differs greatly from character reconstructions that could be hypothesized under the prior understanding of their taxonomic distribution in only three species. This conclusion is also important when considering EFN evolution in the broader context of Malpighiales where they have been documented in 13/39 families (Weber and Keeler 2013), and when considering correlations between EFNs and higher lineage diversification rates (Marazzi and Sanderson 2010, Weber and Agrawal 2014). Humiriaceae contains relatively low extant species-richness despite EFN richness. The gross morphology and variation of Humiriaceae EFNs are consistent with reports from other families (e.g., Tilney and van Wyk 2004, Marazzi et al. 2012, Gonçalves-Souza et al. 2016), although the glands appear unremarkable with no specialized pores or unusual locations on the plants. The shallow teeth forming the crenations along the leaf margins of Sacoglottis gabonensis have been studied in detail, and on the young leaves were found to secrete nectar that attracts ants (Belin-Depoux 1993). Belin-Depoux (1993) did not comment on the basilaminar glands, which we found to be relatively easy to overlook and confuse with the marginal setae scars (see Results). The glands along the leaf margins of Humiria differ from those elsewhere in the family in their position as an often dense submarginal row. They are not modified teeth or setae, which are lacking, but appear to be typical laminar glands developmentally shifted to the margin. Humiria spp. are largely savanna dwellers unlike most of the other Humiriaceae taxa in forests, and habitat shifts may have influenced their EFN evolution. Such habitat shifts have been implicated in leaf defense adaptive evolution, including decrease in EFN abundance and increase in glandular trichomes in neotropical Bignoniaceae (Nogueira et al. 2012). In addition to the extrinsic defenses conferred with the glands, Humiriaceae leaves appear rich in intrinsic defenses due to their thickened and tannin-rich nature.