Character 1:

Rhinarium with two ventrolateral grooves flanking the median sulcus on each side (0); or with a single ventrolateral groove on each side (1). The didelphid rhinarium is divided by a median crease or sulcus (sulcus medianus; Ade, 1999) that extends from between the nares to the margin of the upper lip. On either side of this dividing fold, the ventral margin of the rhinarium is notched by one or two distinct grooves (fig. 2). Two ventrolateral grooves are present on each side in all examined caluromyines and most “marmosines”, but only a single groove was observed in Chironectes, Didelphis, Lestodelphys (P. Jenkins, personal commun.), Lutreolina, Metachirus, Monodelphis, Philander, and Thylamys pallidior. We scored Marmosa rubra as missing (“?”) in the absence of suitably preserved material. First noted by Thomas (1888), this character appears to have been ignored by subsequent students of didelphid systematics.2

Character 2:

Dark midrostral stripe absent (0); or present (1). A median streak of dark fur, unconnected with any other dark marking, extends from the rostrum to the frontal region (between the eyes) in both examined species of Caluromys (fig. 3, top right). In other didelphids, a chevron of dark coronal fur sometimes extends anteriorly between the ears and down the rostral midline, but the condition seen in Caluromys is distinctive and apparently nonhomologous.

Character 3:

Fur surrounding eye not distinctively colored (0); or eye surrounded by mask of dark fur (1). The circumocular fur is not distinctively colored in Caluromysiops, Lutreolina, or Monodelphis, but most other didelphids have mask-like markings (fig. 3). A circumocular mask or ring of dark (usually blackish) fur that contrasts sharply with the paler (usually brownish, whitish, or grayish) color of the crown and cheeks is present in “marmosines”, Glironia, and Lestodelphys. Species of Caluromys have essentially similar reddish-brown eye rings that contrast with grayish cheeks and crowns. A blackish mask is likewise present in all examined species of Didelphis, but this marking is inconspicuous in D. marsupialis and D. virginiana. We also scored a mask as present (state 1) in Metachirus, Chironectes, and Philander, taxa whose dark circumocular fur is continuous with dark fur on the crown of the head.

Character 4:

Circumocular mask contrasts with coloration of cheeks and crown (0); or polymorphic (1); or dark fur around eye continuous with dark coronal fur (2). When present, the didelphid circumocular mask is either discrete (contrasting abruptly in coloration with the surrounding genal and coronal fur; fig. 3, top left), or it is continuous with dark fur that covers the crown (between the ears; fig. 3, bottom right). In Philander frenata, however, both conditions occur with equal frequency in the material we examined. This character was scored as inapplicable (“-”) for taxa that lack a mask of distinctively colored circumocular fur (see character 3, above).

Character 5:

Conspicuous pale spot above each eye absent (0); or present (1). A distinct whitish supraocular spot is consistently present in species of Metachirus and Philander, resulting in the “four-eyed” marking by which these animals are commonly known (fig. 3, bottom right). Most other didelphids lack pale supraocular markings. An indistinct pale bar above each eye in Chironectes appears to be part of the unique transverse banding pattern in that taxon (see Character 7), rather than a homolog of the condition seen in Metachirus and Philander.

Character 6:

Gular gland absent (0); or present (1). As described by Tate (1933: 30), many didelphids have a cutaneous gular (throat) gland, the presence of which is indicated on dried skins and fluid-preserved specimens by a bare median patch of skin; often, but not invariably, the surrounding fur is discolored. According to Barnes (1977: 390), this secretory region contains “hypertrophied apocrine sudoriferous glands and sebaceous glands, both confined to the thickened dermis”. Because external signs of glandular activity tend to be maximally developed in fully mature males (op. cit.), we only scored this character if suitable adult male material (skins or fluid-preserved specimens) was examined.

No unambiguously glandular throat patch was observed in any examined specimens of Caluromys, Caluromysiops, Chironectes, Didelphis virginiana, Lutreolina, Marmosa canescens, M. lepida, M. murina, M. rubra, Marmosops impavidus, M. parvidens, M. pinheiroi, Micoureus, Monodelphis theresa, or Philander. Although most examined adult males of Didelphis albiventris and D. marsupialis had discolored gular fur, no glandular skin was macroscopically distinguishable in these taxa, for which we scored throat glands as absent. Because no adult male specimens of Glironia were examined, we scored this taxon as missing (“?”) in our matrix. Adult males of all other taxa exhibited well-developed gular glands. Tate (1933: 30) reported that throat glands are absent in Marmosops incanus, but every adult male skin and fluid-preserved specimen of M. incanus that we examined (e.g., MVZ 182768, 182769) had conspicuous gular patches of obviously glandular skin.

Character 7:

Dorsal body pelage more-or-less uniformly colored, unpatterned (0); or marked by dark transverse bars (1); or marked by paired scapular stripes (2); or marked by one median and two lateral stripes (3); or grayish middorsum contrasting with reddish or yellowish flanks (4); or grayish midbody (including flanks) contrasting with reddish head and rump (5); or “tricolored” (6). The dorsal body pelage of most didelphids is uniformly colored and essentially unpatterned (state 0), although some taxa assigned to this state are indistinctly darker middorsally than along their paler flanks. States 1–5 are autapomorphies for this study, but we scored them for descriptive completeness and to accommodate denser taxon sampling in future datasets. Among taxa scored in our current matix, dark transverse bars (state 1) are unique to Chironectes; dark scapular stripes (state 2) are unique to Caluromysiops; three longitudinal stripes (state 3) are unique to Monodelphis theresa; a grayish middorsum contrasting with reddish flanks (state 4) is unique to Monodelphis brevicaudata; and a grayish midbody contrasting with reddish head and rump (state 5) is unique to Monodelphis emiliae. The subtle but consistently diagnostic “tricolor” shading of Thylamys and Lestodelphys (state 6) was described by Tate (1933: 209) as follows:

Color illustrations of most of these dorsal pelage patterns are in Eisenberg (1989), Redford and Eisenberg (1992), Pérez-Hernández et al. (1994), Reid (1997), and Eisenberg and Redford (1999). In the absence of any obvious sequence of transformations among alternative dorsal fur markings, this character was treated as unordered in all of our analyses.

Character 8:

Dorsal underfur dark (0); or white (1). The hair bases of the dorsal fur are pigmented, usually dark gray (or grayish), in most didelphids. Species of Didelphis, however, uniquely exhibit white underfur.

Character 9:

Dorsal pelage hairs not grossly dissimilar in length or coarseness (0); or dorsal pelage with guard hairs conspicuously longer and coarser than underfur (1). Very long coarse guard hairs that project conspicuously from the underfur are uniquely exhibited by species of Didelphis. Patton and da Silva (1997) characterized the guard hairs of Philander mcilhennyi as diagnostically longer than those of other congeners, but specimens of P. mcilhennyi that we examined do not approach the condition coded above as state 1.

Character 10:

Manual digit III longer than other manual digits (0); or manual digits III and IV subequal and longer than other manual digits (1); or manual digit IV longer than other manual digits (2). According to Tate (1947: 108), the third and fourth digits of the didelphid manus are subequal and longer than the other manual digits, proportions that correspond to the paraxonic morphotype defined by Brown and Yalden (1973). Not all didelphids have paraxonic forefeet, however. Instead, many have a mesaxonic manus in which dIII is distinctly longer than the other fingers; taxa that we scored as exhibiting this condition include Chironectes, Didelphis, Lestodelphys, Lutreolina, Marmosops, Metachirus, Monodelphis, Philander, and Thylamys. A third alternative condition is seen in Caluromys and Caluromysiops, in which manual digit IV is distinctly longer than dIII. On the hypothesis that these three conditions represent a sequence of transformations reflecting the dominance of either III or IV as the main propulsive digit of the didelphid manus (with the subequal condition intermediate to the other two states), we treated this character as ordered (0 ⟷ 1 ⟷ 2) in all of our analyses.

Because the hardened digits of dried skins are often twisted, folded, or otherwise distorted, this character (and character 14, below) is more confidently scored from fluid- preserved material with elastic phalangeal joints that can be manipulated to align adjacent digits for relative length comparisons.

Character 11:

Central palmar surface of manus smooth, or sparsely provided with large flattened tubercles (0); or densely covered with small convex tubercles (1); or entire plantar surface of manus sandpapery, covered with microscopically dentate tubercles (2). The central palmar epithelium (the region of bare skin encircled by the plantar pads) of the manus is smooth in most didelphids, but the palms are very densely set with small convex tubercles in Thylamys and Lestodelphys. A few didelphids (e.g., Marmosa canescens, Monodelphis theresa) have manual palms that are sparsely provided with large flattened or weakly convex tubercles, a condition that more nearly resembles the common didelphid condition than that seen in Thylamys and Lestodelphys. Sandpapery plantar epithelium is an autapomorphy of Chironectes (see Hamrick, 2001) that cannot be confidently associated with either of the other states of this unordered character. Creighton (1984: characters 2, 6) scored the palmar texture of manus and pes as separate characters, but the taxonomic distribution of the states he recorded suggests that they do not result from independent transformations.

Character 12:

Externally conspicuous lateral carpal tubercles absent in both sexes (0); or large adult males with a prominent lateral carpal tubercle supported internally by the pisiform (1). In most didelphids the wrists of males and females are morphologically similar, but striking sexual dimorphism is present in some “marmosines” (Lunde and Schutt, 1999). Grossly enlarged glabrous tubercles supported internally by carpal ossifications are exhibited by large adult males of Gracilinanus, Marmosops, Micoureus, and some species of Marmosa. Two distinct kinds of tubercles can be distinguished, consisting of lateral (“ulnar”) tubercles supported internally by the pisiform, and medial (“radial”) tubercles supported by the prepollex (op. cit.). Although some intraspecific variation in the development of carpal tubercles has been documented, most examples can be attributed to ontogeny: tubercles are consistently present in the largest adult male specimens of species in which such structures occur, even though they may be lacking in some smaller (presumably younger) conspecific males.

Lateral carpal tubercles supported internally by the pisiform (Lunde and Schutt, 1999: fig. 2) are more widespread than medial carpal tubercles, occurring in all examined taxa that exhibit any conspicuous sexual dimorphism in the wrist. We scored these structures as present in adult males of Marmosa canescens, M. lepida, M. mexicana, M. robinsoni, M. rubra, Marmosops, and Micoureus. We scored Glironia and Gracilinanus microtarsus as missing (“?”) for this character in the absence of suitably preserved material. All other examined didelphid taxa appear to lack lateral carpal tubercles. Our observations differ from those of Lunde and Schutt (1999: table 1), who scored lateral (“ulnar”) tubercles as absent in Marmosa mexicana, M. robinsoni, M. rubra, and Micoureus demerarae. In fact, the hypertrophied pisiform seems to be more closely applied to the center of the wrist (and therefore somewhat less prominent) in Marmosa and Micoureus than it is in Marmosops, but we defer scoring such positional distinctions pending examination of relevant osteological preparations from all of the taxa in our study.

Character 13:

Externally conspicuous medial carpal tubercles absent in both sexes (0); or large adult males with a prominent medial tubercle supported internally by the prepollex (1). We observed medial carpal tubercles (Lunde and Schutt, 1999: fig. 3) in large adult male specimens of Marmosa mexicana, M. robinsoni, M. rubra, and Micoureus. Because no suitably preserved material of Glironia and Gracilinanus microtarsus was available for examination, we scored these taxa as missing (“?”). Medial carpal tubercles appear to be consistently absent in all of the other didelphid taxa examined.

Character 14:

Pedal digit III distinctly longer than adjacent digits II and IV (0); or pedal digits II–IV subequal (1); or pedal digits II–IV progressively increasing in length, with IV distinctly longer than III (2). In Lestodelphys, Lutreolina, and Monodelphis the third pedal digit is longer than the adjacent second and fourth digits, and the hind foot is therefore mesaxonic (sensu Brown and Yalden, 1973) despite the opposability of digit I. By contrast, pedal digits II, III, and IV are subequal (none distinctly longer than the others) in Didelphis albiventris and D. virginiana. Among all other examined didelphid taxa (including Didelphis marsupialis), the second, third, and fourth pedal digits progressively increase in length, such that dIV is the longest toe opposing the hallux (see Boas, 1918: figs. 2, 3). Although taxonomic variation in pedal digit lengths was previously scored for phylogenetic analysis by Creighton (1984: character 9), his observations for Lestodelphys (said to have a long fourth digit) and Lutreolina (said to have subequal third and fourth digits) differ inexplicably from ours. Note that the states of this character do not covary with those of character 10, as might be expected if transformations of manual and pedal morphology were functionally or developmentally determined by the same factors. For example, whereas dIII is the longest manual digit in Marmosops, dIV is the longest pedal digit in that genus.

Character 15:

Pedal digits free, without extensive webbing (0); or all pedal digits bound together by extensive fleshy webs (1). Webbed hind feet (illustrated by Augustiny, 1942: fig. 17; Mondolfi and Medina, 1957: fig. 3; Oliver, 1976: fig 1C; and Hershkovitz, 1997: fig. 3) are a conspicuous autapomorphy of Chironectes.

Character 16:

Plantar epithelium of tarsus entirely or partly glabrous (0); or densely covered with coarse fur (1). In most didelphids the plantar surface of the hindfoot appears to be glabrous (hairless) from toe-tips to heel. In others (e.g., Marmosops) the distal tarsus is partially covered by a velvet-like pelage of microscopic hairs, but there is still a glabrous patch of skin at the apex of the heel. Because a range of intermediates blur the distinction between these two conditions (as when a narrow band of glabrous epithelium connects the naked apex of the heel to the broader expanse of naked skin over the metatarsus), we combine them as a single state (0). By contrast, the tarsus is entirely covered by coarse (macroscopic) hairs in Lestodelphys and Thylamys, an unambiguously different morphology that we code accordingly (as state 1).

Character 17:

Pouch absent (0); or present (1). Based on our examination of parous adult female specimens, pouch-like enclosures for nursing young are unequivocally present or absent among the didelphid taxa included in this study. Although our sample sizes were always small, we observed no intraspecific variation in this character, nor did we observe any intermediate condition between absence and presence of a pouch. Despite the fact that distinctly different pouch configurations (scored as a separate character, below) can be recognized within the family, we provisionally recognize all such enclosures as homologous in the absence of a priori evidence to the contrary.

We found no trace of a pouch in suitable material (fluid-preserved specimens and carefully prepared skins) of parous adult female Glironia, Gracilinanus, Lestodelphys, Marmosa, Marmosops, Metachirus, Micoureus, and Monodelphis. Well-developed pouches were consistently found to be present in suitably prepared parous adult female Caluromys, Chironectes, Didelphis, and Philander. Although Lutreolina was described as pouchless by Thomas (1888), Cabrera (1919), and Marshall (1978a), two fluid-preserved specimens that we examined (UMMZ 166634, USNM 536827) had pouches exactly resembling the morphology illustrated and described by Krieg (1924: fig. 11).3 We scored Caluromysiops as having a pouch based on statements to that effect by Izor and Pine (1987) and Reig et al. (1987), but the absence of compelling documentation (explicit descriptions or illustrations of examined specimens) is noteworthy, and our scoring should be considered provisional. Lacking suitable specimens or reliable literature descriptions of parous adult female specimens, we scored Monodelphis theresa, Thylamys pallidior, and T. venustus as missing (“?”) for this character.

Some noteworthy confusion exists concerning the presence or absence of a pouch in Metachirus. Despite the fact that this genus was clearly described and illustrated as having a well-developed pouch by Enders (1937), all previous and subsequent authors have described it as pouchless, and none of the parous adult females we examined (AMNH 255815; USNM 461138, 577756) had any trace of a marsupium. Enders himself (in a personal communication cited by Pine, 1973) believed that he was mistaken in his taxonomic identification of pouched specimens that he earlier reported as Metachirus, which were unfortunately not preserved as vouchers. Given these reasons to doubt the reliability of Enders' (1937) observations, we scored Metachirus as pouchless rather than polymorphic.

Character 18:

Pouch consists of separate lateral skin folds opening medially (0); or lateral folds connected posteriorly, pouch opening anteriorly (1); or lateral folds connected anteriorly, pouch opening posteriorly (2). The marsupium of Caluromys philander uniquely consists of deep lateral skin folds that enclose the nursing young and open in the midline (resembling the morphology that Tyndale-Biscoe and Renfree [1987: fig. 2.8] incorrectly attributed to didelphids in general). In Caluromys lanatus, Didelphis, and Philander, however, the lateral pockets are joined posteriorly, forming a more extensive enclosure that opens anteriorly (Enders, 1937: fig. 19). Yet another condition is exhibited by Chironectes and Lutreolina, in which the lateral pockets are connected anteriorly, forming a marsupium that opens posteriorly (Krieg, 1924: fig. 11A; Oliver, 1976: fig. 1B). We scored this character as inapplicable (“-”) for all pouchless species, but we scored Caluromysiops (whose pouch morphology has not been described) and the few taxa for which it is not certain whether a pouch is present or absent as missing (“?”). In the absence of any clear indication of intermediacy among the alternative conditions of the didelphid marsupium, this character was treated as unordered in all of our analyses.

Character 19:

Mammae all abdominal/inguinal, more-or-less confined to pouch region (0); or extending anteriorly beyond pouch region to thoracic region (1). In all marsupials that possess a pouch the mammae are contained within it, but the mammae of pouchless taxa are variously distributed (see Tate, 1933: fig. 3). In most pouchless didelphids (e.g., Glironia, Marmosa, Micoureus, Metachirus, some species of Marmosops, and some species of Monodelphis), the mammae are confined to a more-or-less circular inguinal/abdominal array that occupies the same anatomical position as the pouch does in taxa that possess a marsupium. In other pouchless didelphids (e.g., Lestodelphys and Marmosops incanus), however, bilaterally paired mammae extend anteriorly well beyond the pouch region. Although most of these anterior teats are not actually located on the upper chest, they are usually referred to as “pectoral” mammae by authors (e.g., Tate, 1933; Reig et al., 1987: character 42). Because the presence or absence of pectoral mammae cannot be confidently inferred from published teat counts (e.g., those tabulated by Bresslau, 1920), and because unused mammae are inconspicuous, examination of well-preserved parous adult females is essential for accurate scoring of this character. We were unable to examine suitable material of Gracilinanus microtarsus ourselves, but we scored pectoral mammae as present in this taxon based on Tate's (1933: 193) description and illustration (op. cit.: fig. 3) of a fluid specimen formerly preserved in the British Museum of Natural History (BMNH 82.9.30.42).4 We scored Caluromysiops irrupta as having only inguinal/abdominal mammae based on the reported presence of a pouch in that taxon (see character 17, above). Tate (1933) and Reig et al. (1987) reported pectoral mammae as present in Monodelphis, but the Monodelphis specimens we examined had only abdominal/inguinal teats. Lacking suitable material, we scored Micoureus paraguayanus, Monodelphis theresa, Thylamys pallidior, and T. venustus as missing (“?”) for this character.

Character 20:

Urogenital and rectal openings closely juxtaposed and sharing a common mucosa (0); or urogenital and rectal openings widely separated by furred skin (1). In most didelphids the openings of the urogenital and rectal ducts are closely juxtaposed and share a continuous mucosa. In life, both openings are normally recessed in a common sinus (cloaca), but this feature is obscured in male specimens with everted genitalia. Chironectes uniquely lacks a cloaca because the urogenital and rectal openings are widely separated by furred skin (Oliver [1976: fig. 1] correctly illustrated this morphology in the female, but his illustration of the male omits the urogenital orifice). Scoring this character requires fluid-preserved specimens, which were unavailable for Marmosa rubra (scored as missing [“?”] in our matrix). In a few cases, relevant published information was used to score taxa that we were unable to examine (e.g., Nogueira et al.'s [1999a] description of Glironia), but the literature also contains misleading information about didelphid cloacal anatomy.

Hershkovitz (1992a: 203–205; 1992b: 11–13) reported a diversity of cloacal types among marsupials, including: (1) consistent lack of separation between the urogenital and rectal openings (cloaca present in both sexes); (2) male urogenital and rectal openings separated by a “perineal membrane” or “perineum” (cloaca absent in males, present in females); and (3) consistent separation of urogenital and rectal openings (cloaca absent in both sexes). He explicitly attributed the first condition to Caluromys, Gracilinanus, and some species of Marmosa; the second to Caluromysiops, Marmosops, Metachirus, and Micoureus; and the third to the large 2n = 22 opossums (“didelphids” in his usage).

Our side-by-side comparisons of specimens of taxa representing each of Hershkovitz's morphotypes revealed no corresponding differences in their cloacal anatomy with the unique exception of Chironectes noted above. For example, adult male specimens of Caluromys lanatus (supposedly representing cloacal type 1; e.g., AMNH 273038) and Caluromysiops irrupta (type 2; e.g., FMNH 60398) exhibit identical arrangements of the urogenital and rectal openings. Likewise, Nogueira et al.'s (1999b) detailed study of male reproductive anatomy did not reveal any differences in the cloacal anatomy of Caluromys (allegedly type 1), Metachirus (type 2), and Didelphis (type 3). The only plausible explanation for Hershkovitz's unrepeatable and clearly problematic observations is that they resulted from careless interpretation of preservational artifacts.5

Character 21:

Body pelage extends onto tail conspicuously farther dorsally than ventrally (0); or body pelage extends onto dorsal and ventral surfaces of tail to about the same extent (1); or body pelage does not extend appreciably onto tail (2). Body pelage, here defined as soft fur composed of ordinary coat hairs that are not associated with epidermal scales, extends to a variable extent onto didelphid tails. Body fur extends onto the tail much farther dorsally than ventrally in Caluromys lanatus, Caluromysiops, Glironia, Monodelphis brevicauda, and M. emiliae. By contrast, body fur extends onto the tail dorsally and ventrally to about the same extent (from about one-sixth to about one-third the length of that organ) in Caluromys philander, Chironectes, Didelphis, Lutreolina, Micoureus paraguayanus, and Philander. In the remaining taxa examined for this study (Gracilinanus, Lestodelphys, Marmosa, Marmosops, Metachirus, Micoureus demerarae, M. regina, Monodelphis adusta, M. theresa, Thylamys) body fur does not extend more than a short distance onto the tail base (less than or equal to about one-eighth of caudal length). In the absence of any obvious intermediates among these alternative conditions, we treated this character as unordered in all of our analyses.

Character 22:

Caudal integument uniformly pigmented or indistinctly or irregularly marked (0); or blackish basally and abruptly whitish distally (1). Pale tail tips are not uncommon among didelphids, but the observed range of taxonomic variation exhibits few discontinuities for character-state definition. The exposed skin of most didelphid tails is grayish or brownish and, in most taxa with pale tail-tips, the color transition is either gradual (the basal color fading to white) or irregular (mottled). Dorsoventrally bicolored tails (which are grayish or brownish above and pale on the underside) represent another caudal marking pattern that is difficult to score as a discrete character state because it is indistinct in some taxa.

By contrast, the tails of Chironectes, Didelphis, Lutreolina, and Philander are blackish basally with abruptly white tips. Hershkovitz (1997: 34) confusingly described the tail of Philander as “entirely brown or mottled to particolored from tip to as much as distal three-fifths of tail more or less unpigmented”, but all of the specimens of Philander that we examined (with the exception of obviously discolored material) have clearly black-and-white tails. This marking is variable in Lutreolina (a few specimens having all-black tails), but most (75%) of the skins we examined have white-tipped tails, and we scored this modal condition rather than create an intermediate state for intraspecific polymorphism. In Chironectes, 38 of 40 examined specimens (95%) had black tails with white tips. Although significant geographic variation has been reported in the extent of the white caudal marking within some Didelphis species (Gardner, 1973: 10–11), the pattern itself seems remarkably persistent—no population having been described with all-black or all-white tails—and we did not observe any intraspecific polymorphism in our samples of this genus.

Character 23:

Tail scales in predominantly annular series (0); or in annular and spiral series (1); or in predominantly spiral series (2). The epidermal scales that cover the unfurred nonprehensile surfaces of didelphid tails differ taxonomically in shape and arrangement. Square or rectangular caudal scales in unambiguously annular series (state 0) occur in Gracilinanus microtarsus, Marmosa canescens, and Thylamys. Although individual cutaneous scales are indistinct in Lestodelphys and Monodelphis, we also scored those taxa with state 0 based on the annular arrangement of their caudal-hair triplets. By contrast, rhomboidal (diamond-shaped) or hexagonal scales in spiral series occur in Caluromys, Caluromysiops, Chironectes, Didelphis, Lutreolina, Marmosa lepida, M. murina, M. rubra, Marmosops, Micoureus, and Philander. A few taxa, however, have caudal scales that appear to be morphologically intermediate between the annular and spiral morphotypes. In these taxa (Metachirus, Marmosa mexicana, M. robinsoni) the caudal scales appear to be in annular series over the vertebral articulations and in more-or-less spiral series elsewhere. In Glironia, the dorsal and lateral caudal surfaces are covered by body fur, and the entire ventral surface is modified for prehension (see da Silva and Langguth, 1989: fig. 1); therefore, no unmodified caudal scales are present in this taxon, for which we score the character as inapplicable (“-”) in our matrix.

The shape and arrangement of didelphid caudal scales were coded as separate characters by Creighton (1984), but these are clearly not independent features. Rhomboidal scales cannot be efficiently packed around a cylinder in annular series, for example, nor can square scales be packed in spiral series (for illustrations, see Tate, 1933: fig. 2). Because caudal scale shape and packing geometry necessarily covary, and because scale arrangements are more easily characterized—despite the ambiguities recognized above—than scale shapes, we neither coded scale shape as a separate character nor included it in our state definitions. Since the three conditions defined above appear to represent a sequential series of transformations with state 1 as the obvious intermediate, we treated this character as ordered (0 ⟷ 1 ⟷ 2) in all of our analyses.

Character 24:

Unfurred ventral surface of tail base covered with smooth scales (0); or provided with raised tubercles (1). The unfurred ventral surface of the tail base in most didelphids is covered by smooth flat scales, each of which is soft and flexible (yielding easily to the tip of a probe on fluid-preserved material). By contrast, the scales on the underside of the base of the tail are heavily cornified, forming hard raised tubercles in Caluromys lanatus, Caluromysiops irrupta, and Glironia (see da Silva and Langguth, 1989: fig 1a).

Character 25:

Ventral surface of tail tip covered by unmodified hair-bearing scales (0); or ventral surface of tail tip modified for prehension (1). Although all didelphids are perhaps capable of caudal prehension to some extent, external morphological features associated with this behavior are variably developed in the family. The unfurred caudal surfaces of Chironectes, Lestodelphys, Lutreolina, Metachirus, and Monodelphis are covered with unmodified scales (each bearing three or more bristles) from base to tip. In taxa conforming to this morphology, the tail-tip may be provided with a smooth terminal button, but never with a ventrally expanded apical pad bearing dermatoglyphs. The tails of all other didelphids are provided with a distal prehensile surface that may be smooth or covered by modified scales (conspicuously unlike those of the caudal dorsum) but is always transversely creased and glabrous; in taxa conforming to this morphology, the tail-tip is invariably provided with a ventrally expanded pad bearing dermatoglyphs (da Silva and Langguth, 1989: fig 1b; Hershkovitz, 1997: fig. 7).

Creighton's (1984: character 19) coding of caudal prehensile modifications was based in part on the presence or absence of a median groove or sulcus, a distinction that we were unable to recognize consistently in our material. Additionally, his scoring of Metachirus and Lutreolina as indistinguishable from Philander and Didelphis in morphological traits associated with caudal prehension is inconsistent with our observations.

Character 26:

Caudal scales bearing three hairs each (0); or more than three hairs each (1). Three hairs usually emerge from the posterior margin of each caudal scale in most didelphids, but four or more hairs usually emerge from each caudal scale in Lutreolina and Philander. Although individual cutaneous scales are often hard to distinguish in Monodelphis, the caudal hairs of species that we examined usually emerge from the skin in triplets, a condition that we interpreted as corresponding to state 0. Because the entire caudal dorsum is covered with body fur and the entire caudal ventrum is more-or-less glabrous and modified for prehension in Glironia, we were unable to code it for this character. We were also unable to confidently determine the state of this character in Lestodelphys based on the dried skins at hand. Both Glironia and Lestodelphys are therefore scored as missing (“?”) in our matrix.

Creighton (1984: character 20) scored Philander as having three hairs per caudal scale, but our material of P. frenata, P. mcilhennyi, and P. opossum consistently exhibits four or more hairs per scale, the same condition described and illustrated for this genus by Hershkovitz (1997: 14, fig. 8). According to Creighton's text, only Chironectes and Lutreolina have more than three hairs per caudal scale, but his matrix (op. cit.: table 5) records this condition for Lutreolina and Metachirus. Four hairs emerge from beneath some ventral scales near the base of the tail in our material of Chironectes, but the modal count dorsally and distally is three in that genus, as it is in Metachirus.

Character 27:

Tail hairs emerging from each scale not grossly differentiated, varying in length but subequal in thickness (0); or central hair much thicker than lateral hairs (1). The hairs that emerge from the posterior margin of each caudal scale are not grossly differentiated, varying in length but subequal in thickness, in most didelphids. However, in some species of Marmosops (including all of those sampled in this study), the middle hair of each caudal-scale triplet is conspicuously thicker than the lateral hairs, and it is often more darkly pigmented (Gardner and Creighton, 1989). Glironia could not be scored for this character for the reasons previously explained (see character 26).

Character 28:

Tail not incrassate (0); or incrassate (1). The tail is a slender, muscular organ in most didelphids, but Thylamys and Lestodelphys have incrassate tails in which fat is seasonally deposited (Morton, 1980). Incrassate tails can be recognized superficially by their characteristically swollen, carrot-shaped outline and soft texture in fresh and fluid-preserved material, and by their flattened, grease-stained appearance in most skins.

Character 29:

Premaxillae not produced anteriorly beyond I1 (0); or forming a distinct, shelf-like rostral process (1). The premaxillae of many didelphids (Caluromysiops, Chironectes, Didelphis, Glironia, Lestodelphys, Lutreolina, Marmosa canescens, Metachirus, Monodelphis, Philander, and Thylamys) are short, terminating abruptly in front of the incisors, often without a definitive suture between the left and right elements (fig. 4B). In other didelphids, however, the premaxillae are produced anteriorly as a more-or-less acutely pointed shelf-like process that extends the bony rostrum beyond I1 and contains a distinct suture between the right and left bones (fig. 4A). Taxa that we scored as possessing a rostral process of the premaxillae include Caluromys, Gracilinanus, Marmosa (except M. canescens), Micoureus, and most examined species of Marmosops. The specimens of Marmosops incanus that we examined had oddly truncated premaxillae that might be preparation artifacts (the rostral process is sometimes damaged on carelessly prepared specimens, presumably when the nasal cartilages are cut in the last step of skin removal) or an intermediate condition between presence and absence, an uncertainty that we coded as a missing datum (“?”).

Taxonomic variation in the shape of the premaxillae among didelphids has been noted by authors, notably Creighton (1984: character 33), who scored these bones as long (>1 mm) and “pointed” anterior to I1 only in the murina group of Marmosa (= Marmosa sensu stricto); other didelphid terminals in his analysis were scored as having short, rounded premaxillae. However, we did not observe that the premaxillae are consistently longer—or differently shaped—in Marmosa than in the other taxa scored as having a rostral process (state 1) in our matrix. Goin (1993) cited the possession of “shortened premaxillae” as synapomorphic for Caluromyidae, but undamaged specimens of both Caluromys species scored for this analysis have small but distinct rostral processes.

Character 30:

Premaxillae extend to upper canine alveoli (0); or upper canine alveoli contained entirely in maxillary bones (1). A distinct palatal process of the premaxilla extends posteriorly to the upper canine alveolus on each side in most didelphids, but the palatal process is absent and C1 is contained entirely within the maxillary in Caluromys and Caluromysiops.

Character 31:

Maxilloturbinals large and elaborately branched (0); or small and unbranched (1). The maxilloturbinals are thin, dorsally scrolling sheets of bone that occupy the lower part of the didelphid nasal fossa, where they arise from the inner surface of the maxilla on each side (for a general introduction to turbinal anatomy, see Moore, 1981). In most didelphids, the greater curvature of each maxilloturbinal scroll throws off secondary lamellae, which branch to form tertiary lamellae, which branch again to form an elaborate dendritic mass that presumably supports an extensive investing mucosa. By contrast, the maxilloturbinals are simple, slender, unbranched (or only sparsely ornamented) scrolls in all examined species of Monodelphis.

Character 32:

Nasal tips produced anteriorly above or beyond I1, obscuring nasal orifice in dorsal view (0); or posterior to I1, exposing nasal orifice dorsally (1). The tips of the nasal bones are produced anteriorly to a point above or beyond I1 in most didelphids, and the nasal orifice is consequently not visible from a dorsal perspective. In Chironectes, Didelphis, Lutreolina, and Philander, however, the tips of the nasals do not extend so far anteriorly, and the orifice is dorsally exposed.

Character 33:

Nasals conspicuously wider posteriorly than anteriorly (0); or nasals uniformly narrow, their lateral margins subparallel (1). In most marsupials (and extinct metatherians) the nasals are much broader posteriorly (at or near the maxillary/premaxillary suture) than anteriorly. This phylogenetically widespread condition characterizes all didelphids scored for this analysis with the exception of Marmosops incanus (see Mustrangi and Patton, 1997: fig. 14) and Thylamys (see Flores et al., 2000: fig. 7), which usually have much narrower nasals with subparallel lateral margins.

Character 34:

Postorbital processes of frontals absent or indistinct (0); or present as flattened, wing-like extensions of supraorbital crests (1); or present but not flattened or associated with supraorbital crests (2). Postorbital processes of the frontal bones are absent or indistinct in all examined adult specimens of Gracilinanus, Lestodelphys, Marmosa rubra, Marmosops, Metachirus, Monodelphis, and Thylamys, Although postorbital processes are commonly reported as present (without any qualifying descriptor) among other didelphid taxa, we distinguish two distinct morphologies of postorbital frontal outgrowths. In Caluromys, Caluromysiops, Glironia, Marmosa (except M. rubra), and Micoureus, the postorbital processes are flattened and more-or-less triangular projections of the supraorbital crests that shield the dorsal margin of the eye cavity. By contrast, the postorbital processes in Chironectes, Didelphis, Lutreolina, and Philander are bluntly pyramidal or horn-like projections that are not associated with supraorbital crests. Because frontal outgrowths develop late in postweaning ontogeny (Abdala et al., 2001), it is important that this character be scored from examination of large adult skulls. As none of the observed states of frontal morphology appear to be obviously intermediate to the others, this character was treated as unordered in our analysis.

Character 35:

Scars of M. temporalis origin on braincase not fused middorsally to form sagittal crest, or sagittal crest small, not extending to frontals (0); or sagittal crest large and extending to frontals (1). In most didelphids the right and left scars of origin of M. temporalis are widely separated on the braincase, and no sagittal crest is developed. In large adult specimens of Caluromys, Lestodelphys, and Monodelphis, however, a small sagittal crest is sometimes developed over the interparietal or along the midparietal suture. Because such small crests are often indistinct, we include the entire range of variation (from unambiguous absence to small) as a single state. By contrast, much larger sagittal crests that extend anteriorly along the midfrontal suture—an unambiguously different condition—are consistently developed in adult specimens of Caluromysiops, Chironectes, Didelphis, Lutreolina, and Philander.

Character 36:

Parietal and alisphenoid in contact on lateral aspect of braincase (0); or no parietal-alisphenoid contact (1). Among marsupials and extinct metatherians, either the parietal and alisphenoid are in contact on the lateral surface of the braincase, or those bones are separated by contact between the frontal and squamosal (Archer, 1976a; Marshall and Muizon, 1995; Muizon, 1998). Parietal-alisphenoid contact is exhibited by all of the didelphids we examined with the unique exception of Metachirus, in which the frontal and squamosal are in contact as previously reported by Wroe et al. (2000: character 64). Kirsch and Archer (1982: character 28) incorrectly scored Metachirus as resembling other didelphids in this trait.

Character 37:

Petrosal not exposed on lateral aspect of braincase (0); or polymorphic (1); or consistently exposed laterally by fenestra between squamosal and parietal bones (2). In most didelphids, the petrosal is only exposed on the occiput (where it is bounded by the squamosal, exoccipital, and supraoccipital) and ventrally (between the exoccipital, basioccipital, and alisphenoid; see Wible, 1990: fig. 1). In taxa conforming to this morphology (Caluromys, Caluromysiops, Chironectes, Didelphis, Glironia, Lutreolina, Marmosa, Metachirus, Micoureus, Monodelphis, Philander), there is no petrosal exposure through openings on the lateral surface of the braincase. By contrast, the petrosal capsule that encloses the paraflocculus and semicircular canals (= pars canalicularis or pars mastoideus) is consistently exposed through a fenestra in the suture between the squamosal and parietal in Gracilinanus, Thylamys, and some species of Marmosops. Both conditions occur as polymorphisms (neither state clearly predominating) in Lestodelphys, Marmosops incanus, and M. noctivagus. For Marmosa lepida we coded the unambiguously modal condition (no lateral petrosal exposure) despite a single specimen (MNHN 1998–306) with a distinct fenestra in the squamosal-parietal suture on both sides of the skull. As for other transformation series involving polymorphic states, we treated this character as ordered (0 ⟷ 1 ⟷ 2) in all of our analyses with transformations between adjacent states assigned a parsimony cost of 0.5.

Character 38:

Maxillopalatine fenestrae consistently absent or indistinct (0); or polymorphic (1); or large and distinct (2). Marsupial palatal fenestrations have often been discussed in a phylogenetic context, but homoplasy and variability rather than homology and taxonomic consistency have been emphasized by most authors (e.g., Marshall, 1979; Reig et al., 1987). Reservations about coding fenestral characters for phylogenetic analysis are certainly appropriate if taxonomic comparisons are not ontogenetically standardized,6 or if palatal openings are simply scored as present or absent without reference to anatomical location (as by Kirsch and Archer, 1982: character 48; Reig et al., 1987: character 27; Wroe et al., 2000: character 47). Although distinct fenestral loci have long been recognized by taxonomists (e.g., Tate, 1933; Archer, 1981; Hershkovitz, 1992b), inconsistent terminology makes it difficult to recognize homologies among the conditions described by different authors. The fenestral nomenclature adopted in this report is illustrated in figure 5.

Among didelphids, only Caluromysiops and Caluromys lack consistently well-formed maxillopalatine fenestrae. Small perforations in the maxillary-palatine sutures occur in most examined specimens of both genera (see Izor and Pine, 1987: fig. 1), but these appear to be vascular openings and do not resemble the larger, obviously nonvascular fenestrae of other didelphids. Archer (1982) and Reig et al. (1987) reported that Glironia lacks palatal fenestrations, but distinct maxillopalatine openings (illustrated by Marshall, 1978b: fig. 2) are present in some adult specimens of this rare taxon, which we coded as polymorphic. Maxillopalatine fenestrae are unambiguously present in all of the remaining didelphids that we examined. As for other transformation series involving polymorphisms, we treated this character as ordered (0 ⟷ 1 ⟷ 2) in all of our analyses with transformations between adjacent states assigned a parsimony cost of 0.5.

Character 39:

Palatine fenestrae absent (0); or present (1). Many didelphids have, in addition to maxillopalatine fenestrae, separate openings in the posterior palate that are entirely contained within the palatine bones. Palatine fenestrae are consistently present in adult specimens of Didelphis, Gracilinanus, Lutreolina, Marmosa mexicana, Philander, Thylamys, and some species of Marmosops (e.g., M. impavidus, M. incanus). Presence is also the modal condition for Lestodelphys and Marmosops noctivagus, in which individuals without palatine perforations appear to be uncommon variants. Palatine fenestrae are consistently absent in Caluromys, Caluromysiops, Chironectes, Glironia, Metachirus, Monodelphis, some species of Marmosops (e.g., M. parvidens, M. pinheiroi), and most species of Marmosa and Micoureus. Absence is the modal condition for Marmosops pinheiroi and Micoureus paraguayanus, in which a single specimen each exhibited small unilateral palatine perforations.

Character 40:

Maxillary fenestrae absent (0); or present (1). Most didelphids with fenestrated palates have only maxillopalatine or maxillopalatine and palatine openings. A few species, however, have additional fenestrae that are located in the maxillary bone between the maxillopalatine fenestra and the toothrow (at the level of M1 or M2) on each side of the palate. Maxillary fenestrae are consistently present in Gracilinanus microtarsus and Marmosa canescens, and presence is also the modal condition in our material of Thylamys venustus. All other didelphid taxa in our study consistently lack maxillary vacuities except as rare unilateral variants. Creighton (1984: character 26) referred to these openings as “mesolateral” or “mediolateral” fenestrae and noted that they are sometimes confluent with the maxillopalatine fenestrae in Marmosa canescens. Our observations are consistent with his remarks, but it does not seem necessary to create an additional state for the confluent morphology that is variably expressed in that species.

Character 41:

Posterolateral palatal foramina posterior to fourth molars (0); or extending lingual to M4 protocones (1). In all didelphids—and many other metatherians—a large foramen perforates the maxillary-palatine suture posterolingual to the molar row on each side. According to Archer (1976a), this foramen transmits the minor palatine artery from the maxillary artery to the ventral surface of the palate. In most didelphids the posterolateral palatal foramina are small and located behind the fourth molar (fig. 6), but these openings are conspicuously larger and extend anteriorly lingual to the protocone of M4 on each side in Lestodelphys and Thylamys (fig. 5). Creighton (1984: character 24) referred to these openings as “posterolateral fenestrae” and defined their size with respect to the width of M4, but the dimensions of that tooth exhibit independent taxonomic variation that we prefer to score as a separate character (below). Published illustrations of the skull of Lestodelphys (see Marshall, 1977: fig. 1; Reig et al., 1987: fig. 50) depict the posterolateral foramina as posterior to M4, but in all of the specimens we examined the foramina conform to state 1 as described above.

Character 42:

Posterior palate gently sloping ventrally, usually with arched caudal margin and without prominent corners, choanae unconstricted behind (0); or abruptly inflected ventrally, with approximately straight caudal margin and prominent lateral corners, choanae constricted (1). Two alternative morphologies of the posterior palate and the internal choanae can be recognized among didelphids. In caluromyines (Caluromys, Caluromysiops, Glironia) the posterior palate slopes ventrally without any abrupt inflection, the caudal palatal margin is usually broadly arched, and prominent lateral corners are not developed; behind the palate, the internal choanae are not strongly constricted (fig. 6A). By contrast, the posterior palate of didelphines is abruptly inflected ventrally, and the caudal palatal margin is more-or-less straight with prominent lateral corners; behind the palate, the choanae are strongly constricted (fig. 6B). Rather than scoring each of these transformations (slope of the palatal roof, shape of the palatal margin, etc.) separately, we recognize a pattern of correlated changes that appears to involve a reorientation of the oral airway. Although some examined individuals of species scored with either condition differ in minor details (e.g., in some specimens otherwise conforming to state 1 the posterolateral palatal corners are less strongly produced than in others), the overall pattern of posterior palatal morphology is distinctly recognizable in all of the taxa examined for this analysis.

Character 43:

Maxillary and alisphenoid separate (0); or in contact on orbital floor (1). The maxillary and alisphenoid do not contact one another in most didelphids, although these bones are often closely approximated on the floor of the orbit (fig. 7A). By contrast, the maxillary and alisphenoid are consistently in contact on the orbital floor of Lutreolina and Monodelphis (fig. 7B). In occasional specimens of some taxa that do not otherwise exhibit maxillary-alisphenoid contact (e.g., Chironectes, Marmosa murina, Metachirus, Philander mcilhennyi) a thread-like process of the maxillary narrowly contacts the alisphenoid (often unilaterally), but we did not score such uncommon variants as polymorphisms.

Character 44:

Transverse canal foramen absent (0); or present (1). A foramen that transmits the venous transverse canal is present anterolateral to the carotid foramen in many marsupials and extinct metatherians (Sánchez-Villagra, 2001; Sánchez-Villagra and Wible, 2002). The typical didelphid condition (presence) is clearly depicted by Wible (1990: fig. 1), but the transverse canal foramen is sometimes incorrectly identified in other published illustrations of didelphid skulls (e.g., by Hershkovitz [1992b: fig. 18; 1997: fig. 11], who mislabeled it as the “foramen rotundum”).7 Caluromys and Caluromysiops, however, lack a transverse canal foramen (Archer, 1976a; Sánchez-Villagra and Wible, 2002). According to Kirsch and Archer (1982: character 24), the “transverse canal” is absent in Metachirus nudicaudatus and Philander opossum, but all of the specimens we examined of both taxa had a transverse canal foramen on each side of the skull. The transverse canal foramen is occasionally present unilaterally, and it is absent on both sides of the skull in rare examples of some species that otherwise consistently exhibit these openings. Rather than code such infrequent variation as polymorphisms, we scored the modal condition (presence or absence) for each didelphid terminal in our analysis. Sánchez-Villagra and Wible (2002) analyzed the presence or absence of an intramural connection between the left and right transverse canal foramina in their recent survey of didelphid basicranial characters, but confident scoring of this feature requires sectioned or suitably broken skulls, which were not available for the large majority of taxa included in our study.

Character 45:

Extracranial course of mandibular nerve not enclosed by bone (0); or consistently enclosed by anteromedial strut of alisphenoid bulla (1); or usually enclosed by posteromedial bullar lamina (2). The openings through which the mandibular division of the trigeminal nerve (V³) exits the skull have been variously named by systematists. Following Gaudin et al. (1996), we use the term “foramen ovale” for the primary orifice through which V³ exits the endocranial lumen of the adult skull. In most didelphids, the foramen ovale is bordered by the alisphenoid and the petrosal (fig. 8A), but the foramen is sometimes contained entirely within the alisphenoid, and both conditions can be seen on opposite sides of the same skull (see Gaudin et al., 1996: fig. 6). Because it is often difficult to determine the position of this opening with respect to relevant endocranial sutures, however, we did not score the position of the foramen ovale per se. Instead, secondary enclosures of V³ by outgrowths of the alisphenoid tympanic wing provide a more accessible basis for taxonomic comparisons.

Gaudin et al. (1996) distinguished three states of secondary enclosure of the mandibular nerve by alisphenoid outgrowths among marsupials: absence (no secondary enclosure), incomplete enclosure (presence of a secondary foramen but no canal), and complete enclosure (presence of a canal continuous with the primary foramen). Among didelphids, however, there are two different (apparently nonhomologous) conditions of secondary nerve enclosures that are not distinguished by this coding scheme.

In juveniles and adults of Gracilinanus, Lestodelphys, Marmosops, Metachirus, and Thylamys the extracranial course of V³ is enclosed by an alisphenoid process or strut that arises from the anteromedial surface of the bulla and extends anteriorly, medially, and dorsally to span the transverse canal foramen. In the smaller species with this condition (e.g., Marmosops pinheiroi; fig. 8B) the extracranial course of V³ remains unenclosed between this process and the primary foramen ovale, but in larger species (e.g., Marmosops noctivagus) a sheet of bone produced from the posterior edge of the process extends caudally to form a more-or-less complete canal late in postnatal life.

The alternative pattern of secondary foramen and canal formation (state 2) is seen in Caluromysiops, Chironectes, Didelphis, Lutreolina, Monodelphis theresa, and Philander. In these taxa, V³ is broadly enclosed by an alisphenoid lamina (fig. 8C) that extends along the posteromedial bullar surface, but there is no anteromedial process spanning the transverse canal foramen. Many juvenile specimens of some taxa scored with this condition (e.g., Didelphis) do not have a fully enclosed secondary foramen and canal, but they usually show some laminar development; subadults and young adults show progressively more complete enclosure; and old adults (large specimens with heavily worn teeth) usually have completely enclosed foramina and canals (Abdala et al., 2001). All other didelphid taxa examined for this analysis (Caluromys, Glironia, Marmosa, Micoureus, Monodelphis adusta, M. brevicaudata, M. emiliae) lack secondary enclosures of the mandibular nerve except as rare (usually unilateral) variants.

We treat this character as unordered, an analytic option consistent with our hypothesis that the two kinds of secondary enclosures described above are nonhomologous alternative conditions.

Character 46:

Anterior limb of ectotympanic directly attached to skull (0); or indirectly attached via malleus (1). Although didelphids were described by van der Klaauw (1931: 26) as having a completely free ectotympanic, two distinct patterns of attachment can be recognized in the family. In most didelphids (Glironia, Gracilinanus, Lestodelphys, Marmosa, Marmosops, Metachirus, Micoureus, Monodelphis, Thylamys) the anterior (dorsal) limb of the ectotympanic (tympanic annulus) is directly connected to the skull near the point where the squamosal, alisphenoid, and petrosal are juxtaposed behind the postglenoid process. Where the connection can be seen clearly (dried remnants of soft tissues frequently obscure this feature), the actual attachment usually seems to be to the petrosal (fig. 9A). Alternatively (in Caluromys, Caluromysiops, Chironectes, Didelphis, Lutreolina, and Philander), the anterior limb of the ectotympanic is not directly attached to the skull, and the suspension of the tympanic annulus is indirect, via the tympanic (anterior) process of the malleus (fig. 9B).

In all examined taxa with an indirect dorsal connection between the ectotympanic and the skull (state 1), the tympanic annulus is more-or-less ringlike because the posterior (ventral) limb is not expanded medially to form part of the floor of the middle ear cavity. By contrast, in taxa with direct ectotympanic suspension, the posterior limb tends to be dorsoventrally flattened and medially expanded, forming part of the bullar floor to a greater or lesser extent. This aspect of taxonomic variation in ectotympanic morphology, obviously correlated with the suspensory difference scored herein, was previously coded for phylogenetic analysis by Kirsch and Archer (1982: character 52), Creighton (1984: character 28), Reig et al. (1987: character 31), and Goin and Rey (1997: character 11), but their discrepant observations (e.g., regarding Metachirus and Monodelphis) reflect the difficulty of unambiguously distinguishing degrees of ectotympanic expansion, or (perhaps) differences in the species-level exemplars chosen to represent generic terminals in those analyses (see Jansa and Voss, 2000: 70–71).

Character 47:

Fenestra cochleae of petrosal exposed in lateral or ventrolateral view (0); or fenestra concealed within a sinus formed by the caudal and rostral tympanic processes (1). In most didelphid and extinct metatherian petrosals, the fenestra cochleae is completely exposed in lateral or ventrolateral view (as illustrated for Didelphis by Wible, 1990: fig. 1). In Caluromys, Caluromysiops, Lestodelphys, Marmosa rubra, Monodelphis emiliae, and Thylamys, however, the fenestra cochleae is concealed within a chamber or sinus formed by a laminar outgrowth of the pars canalicularis that approximates or contacts an expanded rostral tympanic process of the pars cochlearis. This lamina, which appears to be homologous with the caudal tympanic process of MacPhee (1981), is invariably fused posteroventrally with the paroccipital process (of the exoccipital), which also throws forward a small “tympanic” outgrowth. Wroe et al. (2000: characters 66, 67) coded what appears to be the same taxonomic variation as two characters (one each for the “paroccipital tympanic process” and the “mastoid tympanic process”), but only a single anatomical transformation (sinus formation) is apparent to us among the taxa included in this study. Sánchez-Villagra and Wible (2002) recorded contact between the caudal and rostral tympanic processes in Lestodelphys and Thylamys, but did not observe such contact in the other taxa listed above.

Character 48:

Paroccipital process small and rounded or subtriangular, its base broadly adnate to the petrosal and its apex directed posteroventrally (0); or a large erect process usually directed ventrally (1). The paroccipital process of the exoccipital is inconspicuous in most didelphids, consisting of a low, rounded or subtriangular bony mass for muscle attachment that is broadly adnate to the petrosal; the indistinct apex of the process in taxa that conform to this widespread condition is directed posteroventrally (fig. 10A, B). By contrast, in Chironectes, Didelphis, Lutreolina, Metachirus, and Philander the paroccipital process is much larger, more erect, and points almost straight ventrally in most examined specimens (fig. 10C, D).

Kirsch and Archer (1982: character 51) coded the paroccipital process as “long” in most of the didelphids included in their analysis (Caluromys, Chironectes, Didelphis, Glironia, Lutreolina, Marmosa, Metachirus, Monodelphis, Philander) versus “very small but not absent” in Lestodelphys. However, we were not able to distinguish any aspect of paroccipital morphology that corresponded to their pattern of taxonomic scoring in our material. Instead, the condition of the paroccipital process in Lestodelphys appeared to be indistinguishable from the widespread morphology among small didelphids, which includes a modest range of intergrading forms but not the hypertrophied state seen in Metachirus and the 2n = 22 opossums.

Rougier et al. (1998) and Wible et al. (2001) used the term “paracondylar” for the muscular process of the exoccipital (after Evans and Christensen, 1979),8 reserving “paroccipital” for a muscular process of the petrosal, but our usage follows Coues (1872), Flower (1885), and most other works of comparative mammalian osteology. That the two structures are homologous (as bony sites for muscle attachment) is suggested by the fact that no taxon scored by Rougier et al. (1998) and Wible et al. (2001) for the “paracondylar” character (number 135 in their matrix) and for the paroccipital character (number 131) was observed to have both processes. On this hypothesis, future analyses of basal therian relationships should include one character for presence/absence of a paroccipital process and another for its position (on the petrosal or exoccipital).

Character 49:

Dorsal margin of foramen magnum formed by exoccipitals and supraoccipital (0); or by exoccipitals only (1). The dorsal margin of the foramen magnum is formed by the exoccipitals and the supraoccipital in most didelphids (fig. 10A), but medial processes of the left and right exoccipitals are joined at the midline above the foramen magnum (excluding the supraoccipital from the dorsal margin of that opening; fig. 10C) in adult specimens of Didelphis, Lutreolina, Metachirus, and Philander. Juvenile (and some subadult) specimens of these four genera, however, resemble other didelphids in their occipital morphology, such that both states can be observed in ontogenetic series of conspecific skulls (see Abdala et al., 2001: fig. 3). Sánchez-Villagra and Wible (2002) coded Chironectes as exhibiting the same morphology as the four genera listed above, but the supraoccipital forms part of the dorsal margin of the foramen magnum in all of the ten specimens of Chironectes that we examined for this character (not all of the adult skulls listed in appendix 1 have visible occipital sutures). The unusual and obviously derived condition that we code as state 1 was apparently first described by Coues (1872), but seems to have remained unremarked by subsequent comparative anatomists until scored for phylogenetic analysis by Rougier et al. (1998).

Character 50:

Mandible with two mental foramina (0); or with only one (1). Most didelphids have two mental foramina that perforate the lateral surface of the mandible, of which the more anterior is usually larger and located below p1 or p2, and the more posterior is usually smaller and located below p3 or m1 (Tate, 1933: fig. 6). Chironectes, however, consistently exhibits only a single mental foramen, which is very large and occupies the same position as does the anterior foramen in those taxa with two lateral mandibular perforations.

Character 51:

Angular process acute (more-or-less pointed) and strongly inflected (0); or obtuse (bluntly rounded) and weakly inflected (1). Alternative states of the form and orientation of the marsupial angular process were defined by Sánchez-Villagra and Smith (1997), but we were unable to consistently recognize the distinction between the “rod-like” and “intermediate” conditions they scored for didelphid exemplars. Instead, both rod-like and intermediate angular processes are here combined as “acute and strongly inflected”, a condition common to most didelphids. Only Caluromys and Caluromysiops have bluntly rounded and weakly inflected mandibular angles. Sánchez-Villagra and Wible (2002) also treated this character as binary and recorded the same taxonomic variation that we recognize, but coded the medial inflection of the mandibular angle as “present” (state 0) or “absent” (state 1).

Character 52:

I2–I5 with approximately symmetrical, rhomboidal crowns (0); or with conspicuously asymmetrical crowns (1). The upper first incisor (I1) of all didelphids is a more-or-less styliform tooth, conspicuously unlike I2–I5, and usually set apart from them by a small diastema. Although authors have remarked taxonomic differences in the degree of hypsodonty of I1, such variation appears to be continuous rather than discrete among the numerous terminals included in our analysis. Instead, phylogenetically useful variation in incisor morphology is more easily scored for the more posterior teeth in this series.

The crowns of I2–I5 are approximately symmetrical rhomboids in most didelphids, with subequal anterior (mesial) and posterior (distal) cutting edges that converge to form a sharp central apex; usually, a distinct anterior angle (mesiostyle) and a posterior angle (distostyle) can be seen on unworn teeth (fig. 11A). This is the morphology that Takahashi (1974: 414) labeled “premolariform”, as exemplified by Monodelphis, Marmosa, and Metachirus in her study. In many taxa with posterior upper incisors of this type, the tooth crowns increase in breadth (antero-posterior or mesio-distal dimension) from front to back, such that I2 appears visibly smaller than I5 in labial view. This tendency (which was coded for phylogenetic analysis by Creighton, 1984: character 43) is very pronounced in some taxa with premolariform incisors (e.g., Micoureus, Marmosops, Metachirus) but seems to grade imperceptibly (via intermediate morphologies) to the essentially uniform toothrows of other taxa (e.g., Lestodelphys and Marmosa canescens, in which I2 and I5 are subequal) without obvious discontinuities to permit qualitative coding.

An alternative morphology in which the crowns of I2–I5 are conspicuously asymmetrical, with much longer anterior (mesial) than posterior (distal) cutting edges, occurs in Caluromys, Caluromysiops, Didelphis, Glironia, Lutreolina, and Philander. On unworn teeth of these taxa, the anterior angle (mesiostyle) is more frequently distinct than is the posterior angle (distostyle), which is often entirely lacking on I5 (fig. 11B). This is the morphology that Takahashi (1974) labeled “incisiform”, which she recorded for Chironectes in addition to Caluromys, Didelphis, Lutreolina, and Philander in her study. However, the unworn juvenile dentitions of Chironectes that we examined (e.g., AMNH 126979, 127563, 264573) have approximately symmetrical, rhomboidal crowns. There is a tendency (more marked in Glironia than in other genera of this category) for I2–I5 to decrease in breadth from front to back, such that I2 is sometimes visibly larger than I5 in labial view.

Some published descriptions of didelphid incisor morphology are impossible to reconcile with our observations. Archer (1976b: character 3), for example, coded the upper incisor crowns of Marmosa as “spatulate” versus the presumably nonspatulate condition that he recorded for other didelphids in his data matrix (e.g., Didelphis, Metachirus, and Philander). Subsequently, Kirsch and Archer (1982: character 2) coded the upper incisors of Caluromys, Didelphis, Lestodelphys, Lutreolina, Marmosa, and Philander as “round” versus the “spatulate” condition they observed in Chironectes, Glironia, Monodelphis, and Metachirus. Unfortunately, no definitions of these ambiguous and apparently contradictory descriptors were provided, and we have not been able to discover any aspect of incisor shape that corresponds to the pattern of taxonomic scoring in either study.

Character 53:

Upper canine simple, without distinct accessory cusps (0); or with distinct posterior accessory cusp (1); or with anterior and posterior accessory cusps (2). Although the upper canine (C1) is a simple unicuspid tooth in most didelphids, some species (e.g., Marmosa lepida) have a small but distinct posterior accessory cusp, and others (e.g., Marmosops parvidens and M. pinheiroi; see Voss et al. 2001: fig. 26) have both anterior and posterior accessory cusps. Some care is needed in scoring this character because posterior accessory cusps are sometimes obliterated by attrition (as inferred from their absence in old adults of species that uniformly exhibit such structures as juveniles and subadults). Alternatively, a false accessory cusp is sometimes formed when the posterior edge of C1 is notched by occlusion with p1. Therefore, confident scoring of this character should be based on unworn dentitions. Because no didelphid known to us has an anterior cusp without having a posterior cusp, state 1 appears to represent an intermediate condition in the premolarization of C1. We therefore treated this character as ordered (0 ⟷ 1 ⟷ 2) in all of our analyses.

Because taxa with distinct accessory cusps on C1 also have distinct accessory cusps on c1 (which is likewise premolarized), we did not score the latter as an independent character.

Character 54:

First upper premolar (P1) large, at least one-half the height or width of P2 (0); or vestigial, less than one-third the height or width of P2, or absent (1). Most didelphids have a relatively large first maxillary premolar (as operationally defined above) that essentially resembles P2 and P3 in having a well-defined central cusp flanked by anterior and posterior cutting edges that persist as recognizable occlusal features into adult life. In Caluromys and Caluromysiops, however, P1 is relatively much smaller and lacks prominent occlusal features, usually persisting into adulthood (if at all) as a simple peglike structure; P1 is bilaterally absent in some specimens of Caluromysiops (e.g., AMNH 244364). Glironia exhibits an apparently intermediate morphology, but more closely resembles the common didelphid condition (state 0), a similarity that we scored as such rather than create a unique code for this taxon. Although Reig et al. (1987: character 22) recorded the same taxonomic pattern of variation that we observed for this character, Kirsch and Archer (1982: character 44) did not. Unfortunately, the lack of explicit scoring criteria in previous phylogenetic analyses precludes any substantive evaluation of such discrepancies.

Character 55:

P2 distinctly taller than P3 (0); or P2 and P3 subequal in height (1); or P3 distinctly taller than P2 (2). Didelphid variation in the relative sizes of P3 and P2 was first scored for phylogenetic analysis by Kirsch and Archer (1982: character 4), who recognized only two states (“P3 greater than P2” versus “P3 less than P2”). Scoring by this criterion, however, is problematic for taxa in which these teeth are subequal, and size differences are hard to assess without explicit reference to dental dimensions. Reig et al. (1987: character 23) scored P3 as either “longer” or “shorter” than P2, but crown height is a more convenient criterion for visual comparisons (Wroe, 1997; Wroe et al., 2000).

Among the taxa included in this analysis, P2 is distinctly taller than P3 only in Caluromys and Caluromysiops. P3 is distinctly the taller tooth in Chironectes, Didelphis, Lestodelphys, Lutreolina, Monodelphis, Philander, and Thylamys. P2 and P3 are subequal in the remaining didelphids examined in this study. Because state 1 is clearly intermediate to the others, we treated this character as ordered (0 ⟷ 1 ⟷ 2) in all of our analyses.

Character 56:

P3 with well-developed anterior and posterior cutting edges (0); or only posterior cutting edge well developed (1). The unworn upper third premolar is provided with sharp anterior and posterior cutting edges, each of which extends from the cingulum to the apex of the tooth, in Caluromys, Caluromysiops, and Glironia. In all of the remaining didelphids we examined, however, the anterior margin of the tooth is rounded, and only the posterior cutting edge is well developed. Small anterior blades are variably present near the base of P3 in Chironectes and Philander, but the apex of the tooth is always rounded anteriorly as in the other taxa that we code with state 1.

Other workers have remarked marsupial variation in P3, distinguishing “inflated” or “bulbous” teeth on the one hand from “laterally compressed”, “bladelike”, or “normal” teeth on the other, but we are unable to recognize these descriptors as distinct alternatives among the taxa included in our analysis. Contradictory coding for P3 morphology in different studies (e.g., Didelphis, scored as “bulbous” by Reig et al. [1987: character 24], but as “laterally compressed” by Wroe et al. [2000: character 6]) clearly indicates the ambiguity of such undefined contrasts. Well-developed anterior cutting edges occur on taxa with “bulbous” P3s (e.g., Caluromysiops) and with “normal” teeth (e.g., Glironia) according to Reig et al.'s (1987) scoring of premolar shape variation.

Character 57:

First upper molar (M1) wider than M4, the molar dentition as a whole not or weakly carnassialized (0); or M4 wider than M1, the molar dentition moderately to strongly carnassialized (1). Didelphid molar dentitions differ conspicuously in the relative width (transverse or labial-lingual dimension) of teeth within toothrows. In Caluromys and Caluromysiops the anterior molars are wide in proportion to more posterior teeth, but the posterior teeth are much wider in other didelphids. Our visual comparisons of toothrows suggest that these proportional differences are correlated with a complex pattern of molar transformation that has previously been described in the literature as “carnassialization” (Reig and Simpson, 1972: 534) or as “an emphasis on postvallum-prevalid shear” (Muizon and Lange-Badré, 1997).

Tribosphenic molar occlusion results in shearing forces created by the relative movement of opposing edges on the upper molar trigons and the lower molar trigonids, together with the crushing action of the protocone as it penetrates the talonid basin. Because adjacent marsupial molars are not separated from one another by diastemas, and because upper and lower teeth interlock tightly in occlusion, enlargement or reduction of one dental structure necessarily involves reciprocal changes in other parts. As described by numerous authors (e.g., Reig and Simpson, 1972; Reig et al., 1987; Muizon and Lange-Badré, 1997), the shearing (carnassial) action of marsupial molars is emphasized in taxa with large metacristae (on the posterior wall or vallum of the trigon, hence “postvallum”) that occlude with large paracristids (on the anterior wall of the trigonid, hence “prevallid”); the same taxa also tend to have relatively small paracones, protocones, metaconids, and talonids. These tendencies are redundantly described by numerous dental ratios that have been coded as independent characters in previous phylogenetic studies (see appendix 4).

We measured molar widths (from the stylar shelf at or near the “A” position to the lingual surface of the protocone) in order to document the proportional differences we observed involving the most anterior and posterior elements of the toothrow, and we measured the lengths of the metacrista and postprotocrista of M3 to obtain an index of carnassialization.10 A bivariate plot of these data (fig. 12) effectively illustrates the divergence of Caluromys and Caluromysiops from other didelphids in both respects, and the essentially continuous variation that precludes recognition of additional states (despite a wide range of morphologies) among members of the latter group.

Character 58:

Centrocrista of M1–M3 linear, its apex almost level with the floor of the trigon basin (0); or weakly V-shaped, its apex distinctly elevated above the trigon basin (1); or strongly V-shaped, its apex high above the trigon basin (2). The shape of the centrocrista (postparacrista + premetacrista) and the descriptive nomenclature associated with its alternative states have been a source of some confusion among marsupial researchers. Traditionally, taxa with a V-shaped (labially inflected) centrocrista have been described as “dilambdodont” because the ectoloph (preparacrista + centrocrista + postmetacrista) is then W-shaped (like two inverted lambdas), whereas taxa with a straight (uninflected) centrocrista have usually been called “predilambdodont” (e.g., by Reig et al., 1987). Unfortunately, neither descriptor applies unambiguously to some didelphids (Goin, 1997), and certain taxa have been coded with contradictory character states in different phylogenetic datasets. Didelphis, for example, was coded as having a V-shaped centrocrista by Reig et al. (1987: character 1) and Wible et al. (2001: character 31), but Wroe et al. (2000: character 10) scored it as having a linear centrocrista. Our coding reflects Johanson's (1996) observation that the shape (linearity versus labial inflection) of this crest is correlated with its occlusal relief (apical height above the trigon basin), and we recognize an intermediate condition for taxa that do not conform with either traditionally recognized morphotype.

Among the taxa included in our study, only Caluromysiops has a truly linear centrocrista on M1–M3. In this taxon, the apex of the centrocrista is essentially level with the floor of the trigon basin, clearly conforming to the predilambdodont condition defined by Johanson (1996). By contrast, the centrocrista is strongly inflected labially (buccally)—and therefore distinctly V-shaped—in Gracilinanus, Lestodelphys, Marmosa, Marmosops, Metachirus, Micoureus, Monodelphis, and Thylamys; the apex of the crest is elevated well above the trigon floor in these taxa, which are unambiguously dilambdodont sensu Johanson (1996). The intermediate condition occurs in the six remaining genera (Caluromys, Chironectes, Didelphis, Glironia, Lutreolina, Philander) in which the centrocrista has a weak labial inflection with a slightly elevated apex; these taxa could appropriately be described as weakly dilambdodont. Because these states appear to represent successive stages in a single complex transformation of occlusal architecture with the weakly dilambdodont condition intermediate to the others, we treated this character as ordered (0 ⟷ 1 ⟷ 2) in all of our analyses.

Character 59:

Upper molars without a distinct ectoflexus on any tooth (0); or distinct ectoflexus present on one or more teeth (1). Among the taxa examined for this analysis, only Caluromys and Caluromysiops lack any trace of an ectoflexus on the upper molars. A distinct ectoflexus is present on M3, on M2 and M3, or (rarely) on M1–M3 in all other didelphids. Rougier et al. (1998: character 19) scored the presence or absence of a “deep” ectoflexus on M2 and M3, but a complete range of intermediate conditions of the ectoflexus from shallow to deep among the taxa treated herein precludes unambiguous scoring of such distinctions.

Character 60:

Preprotocrista and anterolabial cingulum joined to form a continuous shelf along the anterior margin of M3 (0); or crista and cingulum separate, not forming a continuous shelf (1). In many didelphids (Caluromys, Caluromysiops, Glironia, Gracilinanus, Marmosa, Marmosops parvidens, Micoureus) the preprotocrista passes labially around the base of the paracone to join with the anterolabial cingulum, forming a continuous shelf along the anterior margin of the crown of each of the first three maxillary molars (fig. 13, left). In the alternative morphology exhibited by Chironectes, Didelphis, Lestodelphys, Lutreolina, Marmosops (except M. parvidens), Metachirus, Monodelphis, Philander, and Thylamys, the preprotocrista extends only to a point at or near the base of the paracone; the anterolabial cingulum does not converge toward the preprotocrista in these taxa, but passes obliquely dorsally such that the two crests are discontinuous on the anterior surface of the tooth crown (fig. 13, right). Because this occlusal feature sometimes varies along the toothrow, we standardized our taxonomic scoring by recording the condition observed on M3.

Archer (1976b: characters 31, 32), Creighton (1984: character 53), and Wroe et al. (2000: character 25) referred to state 0 as comprising a “complete” anterior cingulum, and state 1 as an “incomplete or absent” cingulum. However, whereas Creighton and Wroe et al. reported essentially the same taxonomic variation that we observed, Archer recorded the presence of a complete cingulum in Philander opossum, a species (and genus) that uniformly lacks this feature in our material. Similarly, Rougier et al. (1998) and Wible et al (2001: character 33) scored the preprotocrista of Didelphis as extending labially beyond the base of the paracone, an apparent lapsus; in all specimens of Didelphis that we examined, the preprotocrista terminates at the base of the paracone with no trace of a labial extension.

Character 61:

Fourth upper molar erupts before P3 (0); or M4 and P3 erupt simultaneously (1); or P3 erupts before M4 (1). As described by Tribe (1990), M4 is the penultimate upper tooth to erupt in most didelphids, followed by P3; by contrast, P3 erupts before M4 in Chironectes, Didelphis, Lutreolina, and Philander. We also coded an intermediate condition for taxa in which M4 and P3 erupt simultaneously (Metachirus nudicaudatus, Monodelphis brevicaudata) and treated this character as ordered (0 ⟷ 1 ⟷ 2) in all of our analyses. Suitable subadult material was not available to determine eruption sequences for several species (Micoureus paraguayanus, Monodelphis emiliae, M. theresa, Philander frenata), which are scored as missing (“?”) in our data matrix.

Character 62:

Distinct lingual cusp of lower incisors present (0); or absent (1). The unworn lower incisors of most didelphids have a distinct lingual (posterior) cusp or heel (fig. 14, top) that is entirely lacking in Chironectes, Didelphis, Lutreolina (fig. 14, bottom), and Philander. This character has the same taxonomic distribution among didelphids as Takahashi's (1974) distinction between the “subrectangular” teeth of Chironectes, Didelphis, Lutreolina, and Philander on the one hand, and the “suboval” lower incisors of Caluromys, Marmosa, Metachirus, and Monodelphis on the other, but we are unable to appreciate the basis for her shape descriptors.

Character 63:

Second lower premolar distinctly taller than p3 (0); or p2 and p3 subequal in height (1); or p3 distinctly taller than p2 (2). The second lower premolar is distinctly taller than p3 in most didelphids, but p2 and p3 are subequal in Thylamys and in most species of Monodelphis. In Lestodelphys and Monodelphis emiliae, however, p3 is distinctly taller than p2. Due to the obviously intermediate condition represented by state 1, this character was treated as ordered (0 ⟷ 1 ⟷ 2) in all of our analyses.

Character 64:

Deciduous lower third premolar (dp3) with distinctly tricuspid (complete) trigonid (0); or with unicuspid or bicuspid (incomplete) trigonid (1). Although most previous descriptions of didelphid milk teeth have described them as large and molariform (Flower, 1867; Thomas, 1888; Bensley, 1903; Tate, 1948; Archer, 1976b), taxonomically significant variation in the size and shape of dp3 was reported by Voss et al. (2001: table 5). Many didelphids (Caluromys lanatus, Caluromysiops, Chironectes, Didelphis, Lutreolina, Marmosops, Metachirus, Monodelphis emiliae, Philander) have a fully molariform dp3 in which the trigonid is represented by its normal complement of three cusps (paraconid, protoconid, metaconid) in a more-or-less triangular configuration (fig. 15, top). A fully molariform dp3 is also the modal condition for Monodelphis brevicaudata, in which seven of nine specimens examined with intact milk dentitions had complete trigonids. By contrast, dp3 is only partially molariform in Lestodelphys, Marmosa, Micoureus, and Thylamys, which have incomplete, blade-like trigonids bearing only one or two distinct cusps (fig. 15, bottom). Scoring Caluromys philander for this character was somewhat problematic because only the protoconid of dp3 is usually distinct in our material, but the trigonid is triangular in outline and more closely resembles the fully molariform condition than the blade-like bicuspid alternative. We did not examine juvenile specimens of several taxa (Glironia, Gracilinanus, Marmosops incanus, Micoureus paraguayanus, Monodelphis adusta, M. theresa, and Philander frenata) that are scored as missing (“?”) in our matrix.

Character 65:

Lower third molar hypoconid labially salient (0); or m3 hypoconid lingual to salient protoconid (1). The hypoconid of m3 is labially salient (projecting beyond the protoconid, or level with it) in most didelphids, but the hypoconid is lingual to the protoconid on m3 in Lestodelphys and Monodelphis. Goin and Rey (1997: character 20) described the hypoconids of Monodelphis as “poco salientes labialmente” versus the alternative condition of “[h]ipocónidos salientes labialmente” attributed to the tribe Marmosini (including Lestodelphys), an observation that conflicts with our assessment, but for which we can offer no explanation.

Character 66:

Entoconid large and well developed on m1–m3 (0); or very small or indistinct (1). Most didelphids have a well-developed entoconid that is about as tall as the hypoconid and much exceeds the adjacent hypoconulid in height and breadth on m1–m3 (fig. 16, top). In Monodelphis, however, the entoconid is very small, never more than subequal to the hypoconulid, and often smaller than that cusp (fig. 16, bottom); it becomes indistinct or is obliterated by wear in most adult specimens. Our scoring of didelphid entoconid variation differs only in minor details from coding schemes previously suggested by Archer (1976b: character 43), Creighton (1984: character 58), Goin and Rey (1997: character 21), Rougier et al. (1998: character 54), and Wroe et al. (2001: character 41), chiefly by discounting cusp shape (as too subjective to categorize) and by not distinguishing cusp reduction from absence (an entoconid could be distinguished in all examined specimens with newly erupted teeth). Kirsch and Archer (1982: character 17) coded Lestodelphys as having a small entoconid like that of Monodelphis, but our material of Lestodelphys (especially specimens with newly erupted teeth; e.g., MMNH 17171) more closely resembles the entoconid morphology that we observed in other didelphids.

Character 67:

Hypoconulid at posterolingual margin of talonid, “twinned” with entoconid on m1–m3 (0); or at midline of tooth, approximately equidistant to hypoconid and entoconid, not “twinned” with the latter cusp (1). Among didelphids, only Caluromysiops has a hypoconulid that is not “twinned” with the entoconid, but this trait is widespread among placentals and basal therians (see data matrices in Cifelli, 1993; Wible et al., 2001). We scored this character on the first three lower molars because the entoconid is sometimes missing from m4.

Character 68:

Robertsonian equivalents {acr1 + acr5, met1} present as a single metacentric chromosome (0); or as two acrocentric chromosomes (1).

Character 69:

Robertsonian equivalents {acr2 + acr8, met2} present as a single metacentric chromosome (0); or as two acrocentric chromosomes (1).

Character 70:

Robertsonian equivalents {acr3 + acr10, met3} present as a single metacentric chromosome (0); or as two acrocentric chromosomes (1).

Character 71:

Robertsonian equivalents {acr6 + acr9, met4} present as a single metacentric chromosome (0); or as two acrocentric chromosomes (1).

Type Species:

Didelphis (Micoureus) canescens Allen, 1893, subsequently transferred to Marmosa by Allen (1897) and thereafter known as Marmosa canescens throughout the 20th century mammalogical literature (e.g., by Tate, 1933; Gardner, 1993).

Geographic Distribution:

Apparently endemic to Mexico, where it occurs in seasonally dry (deciduous) tropical forests from Sonora southward (principally along the Pacific littoral and adjacent slopes of the coastal cordilleras) to Oaxaca and Chiapas; isolated populations also occur in the northern part of the Yucatan Peninsula and on the Tres Marías Islands (Hall, 1981; Wilson, 1991; Reid, 1997). Armstrong and Jones (1971) implied that T. canescens occurs in Guatemala, but we have not examined specimens from that country, nor have we seen any published references to vouchered Guatemalan records.

Contents:

Only a single valid species is currently recognized following Tate (1933) and Gardner (1993), who treated gaumeri Osgood (1913), insularis Merriam (1898), oaxacae Merriam (1897), and sinaloae Allen (1898) as synonyms or subspecies of canescens. However, substantial geographic variation has been noted by authors (e.g., Wilson, 1991), and sufficient material (amounting to several hundred specimens in U.S. and Mexican museums) is now available for a long-overdue critical revision of these nominal taxa. Among them, the status of the Yucatecan population (gaumeri) merits particular attention due to its zoogeographically unusual disjunction from supposedly conspecific western Mexican forms (see Remarks, below).

Etymology: From the Nahuatl (Aztec) word for “opossum” (Karttunen, 1983), which is still in colloquial use as a singular masculine noun (e.g., by Villa-Ramírez, 1991).

Diagnosis and Comparisons:

Small didelphines that can be distinguished from all other members of the subfamily by nonmolecular characters analyzed in this report, among which the following are salient points of morphological comparison. Rhinarium with two ventrolateral grooves; dark ocular mask present; supraocular spots absent; gular gland absent; dorsal fur unpatterned (unicolored grayish or reddish-gray), with short, inconspicuous guard hairs; manual digits III and IV subequal; large adult males with well-developed lateral carpal tubercles but not medial carpal tubercles; plantar surface of hind foot naked from heel to toes; pedal digit IV longer than other pedal digits; marsupium absent; tail essentially naked (without a conspicuously furred base), covered by epidermal scales in annular series except for distal prehensile surface, not incrassate. Rostral process of premaxillae absent; nasals conspicuously wider posteriorly than anteriorly; very large, flattened, wing-like postorbital processes present in mature adults; sagittal crest absent; parietal and alisphenoid in contact on lateral braincase; no lateral petrosal exposure through fenestra between parietal and squamosal; maxillopalatine fenestrae long (usually extending from the level of P3 to M3); palatine fenestrae absent; maxillary fenestrae present, often confluent with maxillopalatine openings; posterolateral palatal foramina posterior to M4 protocones; posterior palate with prominent lateral corners, the internal choanae constricted behind; transverse canal foramen present; secondary foramen ovale absent; ectotympanic suspension direct; fenestra cochleae exposed; paroccipital process of exoccipital small, rounded, adnate to posterior aspect of petrosal; dorsal margin of foramen magnum formed by supraocciptal and exoccipitals. Crowns of I2–I5 symmetrically rhomboidal and subequal (not increasing in size from front to back); P2 and P3 subequal in height; P3 without an anterior cutting edge; upper molars strongly dilambdodont; distinct ectoflexus present on M2 and M3 (much deeper on the latter tooth); anterior cingulum complete on M3. Distinct lingual cusp present on i1–i4; c1 an erect, dorsally recurved tooth (not procumbent and premolariform); p2 much taller than p3; dp3 with incomplete (bicuspid) trigonid; entoconid a tall sharp cusp, much higher than hypoconulid on m1, m2, and sometimes m3.

Although Marmosa and Tlacuatzin are superficially similar, T. canescens differs from all other species currently referred to Marmosa (sensu Gardner, 1993) in several nonmolecular characters that we coded for phylogenetic analysis, including: caudal scales in unambiguously annular series (character 23), absence of a premaxillary rostral process (character 29), possession of maxillary palatal vacuities (character 40), and 2n = 22 chromosomes. Two additional comparisons also provide diagnostic criteria, but these were not coded for phylogenetic analysis due to morphological intermediates observed among other didelphine taxa. The first concerns the morphology of I2–I5, which have rhomboidal crowns that increase in breadth from front to back in Marmosa, such that I2 is visibly narrower than I5 in lateral view. The crowns of I2–I5 are also rhomboidal in Tlacuatzin, but in that taxon they do not increase in breadth from front to back, and the crowns of I2 and I5 appear subequal in lateral view. The second comparison involves the lower canine, which is a procumbent tooth with a flattened, blade-like apex and (occasionally) a small posterior accessory cusp in Marmosa. By contrast, c1 is an erect, simple, recurved-conical tooth in Tlacuatzin. Additional character-state differences (scored in appendix 5) distinguish Tlacuatzin canescens from Marmosa murina (the type species of Marmosa), but these are not consistently useful for generic diagnosis.

Tlacuatzin canescens resembles Marmosa andersoni (the type species of Stegomarmosa Pine, 1972) in having large postorbital processes (Pine, 1972: fig. 1), but these taxa are otherwise dissimilar. Based on our examination of the Peruvian type specimen (FMNH 84252), M. andersoni differs from T. canescens by having tail scales in spiral series; a long rostral process of the premaxillae; large palatine fenestrae; no maxillary fenestrae; upper incisor crowns that increase in breadth from I2 to I5; and a procumbent, apically flattened c1. Unfortunately, the karyotype of M. andersoni is unknown. Based on these and other character data, we see no evidence for a close relationship between T. canescens and M. andersoni, and we concur with the current treatment of Stegomarmosa as a synonym or subgenus of Marmosa.

Remarks:

Tlacuatzin canescens is part of a distinctive fauna that inhabits dry tropical forests in western Mexico. Recent syntheses of distributional data have documented the impressive diversity of western Mexican dry forests, to which at least 25 mammalian species are endemic (Ceballos, 1995; Ceballos and García, 1995; Ceballos et al., 1998). A different fauna inhabits the dry tropical forests of the Yucatan Peninsula, however, where T. canescens also occurs. The Yucatecan dry forest fauna is less easily defined than that of western Mexico because dry forest on the Yucatan Peninsula grades into more mesic vegetation formations, but most Yucatecan endemics are primarily dry-forest species (e.g., Heteromys gaumeri, Otonyctomys hatti, Mazama pandora; see Reid, 1997; Medellín et al., 1998).

Tlacuatzin is one of seven mammalian genera endemic to Mesoamerican dry forests, the other six consisting of a shrew (Megasorex), one bat (Musonycteris), and four rodents (Hodomys, Osgoodomys, Otonyctomys, Xenomys). Without exception, all of these genera are currently considered to be monotypic (Wilson and Reeder, 1993). Whereas five are endemic to western Mexico (Megasorex, Musonycteris, Hodomys, Osgoodomys, Xenomys) and one is endemic to the Yucatan Peninsula (Otonyctomys), only Tlacuatzin has a disjunct distribution in both dry-forest regions. In fact, Tlacuatzin appears to be one of only two mammalian taxa (of any rank) endemic to Middle American dry forests that occurs both in western Mexico and the Yucatan, a biogeographic anomaly that begs critical attention.13