Materials and Methods

Phylogenetic Assumptions and Taxon Sampling

Our phylogenetic analyses are based on the assumption of didelphid monophyly, which is strongly supported by a variety of nuclear-gene sequence datasets (e.g., Jansa and Voss, 2000; Amrine-Madsen et al., 2003; Meredith et al., 2008). Our ingroup terminals (table 2) consist of 44 species representing every currently recognized didelphid genus and subgenus, including all of the species previously analyzed by Jansa and Voss (2000, 2005), Voss and Jansa (2003), Voss et al. (2004a, 2005), Jansa et al. (2006), and Gruber et al. (2007). Altogether, this is by far the most taxon-dense and character-rich didelphid dataset ever analyzed.

Table 2

Marsupial Terminal Taxa Scored for Phylogenetic Analysis^a

The outgroups for this study include representatives of every extant ordinal-level clade that has ever been recovered within one or two nodes of Didelphidae in previous analyses of marsupial relationships, including caenolestids (Paucituberculata), dasyurids (Dasyuromorphia), Dromiciops (Microbiotheria), and peramelids (Peramelemorphia). We did not include Notoryctes (Notoryctemorphia) or diprotodontians because these taxa have never been considered close to didelphids, and because their highly derived morphologies raise many unresolved issues of homology that are beyond the scope of this study.

Comparative Morphology

Our morphological comparisons are based on skins, skulls, and fluid-preserved specimens (appendix 1), which we examined for discretely varying traits that could be scored for most of the terminal taxa in our analysis. Additionally, we summarize information about other phenotypic characters that, although not suitable for phylogenetic analysis due to continuous variation (e.g., body size), provide taxonomically useful descriptors. Information about morphological character variation among nondidelphid marsupials is largely based on our scoring of outgroup terminal taxa, but we also include remarks on the morphology of other Old World marsupials where these seem relevant or interesting.

To provide a maximally user-friendly reference, our morphological observations are summarized in two different formats. First, we describe taxonomic comparisons using normal (nontelegraphic) prose in the body of the text. These accounts, organized organ-system-by-organ system, are accompanied as necessary by illustrations of representative specimens. Second, we summarize morphological descriptors taxon-by-taxon using telegraphic prose in the generic accounts of our classification. Although largely redundant with our data matrix (appendix 4), these generic descriptions are easier to use for taxonomic identifications, and they provide an opportunity to summarize relevant observations that were not encoded as characters.

Formal character descriptions (in appendix 3) are accompanied by technical details about coding procedures and ordering criteria. Our criteria for character choice were described at length by Voss and Jansa (2003), so they are not repeated here. The only noteworthy methodological change in this study concerns our coding of polymorphisms. As in our previous study, we ignored rare variants (on the assumption that all characters are polymorphic given sufficiently large samples) and coded only those polymorphisms represented by nearly equal frequencies of alternative states (e.g., when presence and absence were both commonly observed for a given taxon). However, whereas we formerly (Voss and Jansa, 2003) treated such polymorphism as a separate state intermediate to the two fixed conditions, with transformations to and from the polymorphic state weighted as half-steps (the “scaled” option discussed by Wiens, 2000), we here conform with the prevailing custom of coding polymorphisms as taxonomic ambiguities (e.g., as “0/1” for a binary character). This coding change has minimal impact on our results (only weakly supported relationships are affected, even in separate analyses of morphology) but it facilitates likelihood modeling of morphological character evolution (Lewis, 2001) in Bayesian analyses of our combined datasets (see below).

Molecular Sequencing and Homology Comparisons

The laboratory procedures we used for DNA amplification and sequencing (including the names and locations of primers used in PCR reactions) have already been described for three of the protein-coding nuclear loci analyzed in this report: IRBP (Jansa and Voss, 2000), Dentin Matrix Protein 1 (DMP1; Jansa et al., 2006), and Recombination Activating 1 Gene (RAG1; Gruber et al., 2007). In addition, we sequenced two other protein-coding nuclear genes as described below.

We amplified 2.1 kb of BRCA1 exon 11 from genomic DNA in two fragments using the primers listed in table 3. The first fragment, comprising the upstream 1.2 kb of the exon, was amplified using primers F1 paired with R1218 or F47 paired with R1343. The second, downstream fragment (ca. 1 kb) was amplified with primers F1163 or F1163a paired with R2078 or R2151. These amplification products were then used in a second round of PCR to generate smaller pieces of suitable size for sequencing. For the upstream fragment, either F1 or F47 was paired with R743, and F593 was paired with either R1218 or R1343. For the downstream fragment, F1163 or F1163a was paired with R1780, and F1697 was paired with either R2078 or R2151. A single fragment of ca. 1 kb was amplified from vWF exon 28 using either F104 or F120 paired with R1141. This product was then used in a second round of PCR in which F1 or F47 was paired with either R665 or R742, and F557 was paired with R1141.

Table 3

Primers Used to Amplify and Sequence BRCA1 and vWF

Initial amplifications using genomic DNA as template were performed as 20 µl reactions using Ampli-Taq Gold polymerase (Perkin-Elmer Corp.) and recommended concentrations of primers, nucleotides, buffer, and MgCl₂. These genomic amplifications were performed using a four-stage touchdown protocol as described in Jansa and Voss (2000). Reamplification reactions were performed using Taq DNA polymerase (Promega Corp.) in 25 µl reactions for 30 PCR cycles. The resulting PCR products were sequenced in both directions using amplification primers and dye-terminator chemistry on an AB 3700 automated sequencer.

We searched the draft Monodelphis domestica genome using the BLAT (modified BLAST; Kent, 2002) algorithm to determine the copy number and chromosomal location of query sequences from M. brevicaudata. As reported on the Ensemble database ( www.ensemble.org/index.html), the available reference genome (assembly MonDom5) was released on October 2006 and has a base coverage of approximately 7.3×.

Alignment and Phylogenetic Analysis

We aligned DNA sequences with reference to translated amino-acid sequences using MacClade 4.08 ( http://macclade.org). Aligned sequences were then analyzed phylogenetically using maximum parsimony (MP) as implemented by PAUP* ver. 4.0b10 (Swofford, 1998), maximum likelihood (ML) as implemented by GARLI ver. 0.95 (Zwickl, 2006), and Bayesian inference as implemented by MrBayes ver. 3.1.1 (Ronquist and Huelsenbeck, 2003). For MP analyses, all molecular characters were treated as unordered and equally weighted, and all tree searches were heuristic with at least 10 replicates of random stepwise taxon addition followed by tree bisection-reconnection (TBR) branch swapping. To choose the best models of nucleotide substitution for ML and Bayesian analyses, we examined the fit of various models separately for each of our five gene partition (IRBP, vWF, BRCA1, DMP1, and first and second codon positions of RAG1) based on neighbor-joining trees of Jukes-Cantor-corrected distances using both hierarchical likelihood-ratio tests (hLRTs) and the Akaike Information Criterion (AIC) as implemented in ModelTest 3.7 (Posada and Crandall, 1998). Where the two approaches disagreed on model choice, we used the model selected by the AIC for reasons outlined by Posada and Buckley (2004). For both ML and Bayesian searches, the best-fit model was specified, but model parameter values were not fixed. For ML analyses, we conducted three independent runs of genetic-algorithm searches in GARLI, with random starting topologies and automatic termination after 10,000 generations with no improvement in log-likelihood scores. For Bayesian analysis of each gene partition, we conducted two independent runs of Metropolis-coupled Markov-chain Monte Carlo (MCMCMC), each with one cold and three incrementally heated chains. For each run, we assumed uniform-interval priors for all parameters, except base composition, which assumed a Dirichlet prior. Runs were allowed to proceed for 5 × 10⁶ generations, and trees were sampled every 100 generations. We evaluated the burn-in for each run, and pooled the post-burn-in trees to calculate estimated parameter distributions and posterior probabilities for each node. We assessed nodal support from MP and ML analyses using nonparametric bootstrapping (Felsenstein, 1985). Bootstrap values for the parsimony analyses (MPBS) were calculated in PAUP* from 1000 pseudoreplicated datasets, each of which was analyzed heuristically with 10 random-addition replicates with TBR branch swapping. Bootstrap values for the likelihood analysis (MLBS) were calculated in GARLI using genetic-algorithm searches of 1000 pseudoreplicated datasets, allowing model parameters to be estimated for each pseudoreplicate.

We analyzed the combined-gene dataset using ML as implemented in RAxML-VI-HPC (ver. 2.2.3; Stamatakis, 2006). We specified the GTRMIX model, which performs initial tree inference using a GTRCAT approximation, with final topology evaluation performed under a GTRGAMMA model. We allowed parameters to be estimated independently across the five gene partitions. To evaluate nodal support for this combined-gene, mixed-model analysis, we performed 1000 bootstrap replicates, again allowing model parameters to be estimated independently across the five genes. We also performed a Bayesian analysis of the combined-gene dataset, using the same MCMCMC settings given above. For this analysis, we specified the best-fit model for each gene, decoupled estimation of substitution parameters across the partitions, and allowed each gene to assume a separate rate.

We analyzed the nonmolecular (morphological + karyotypic) data alone and in combination with the molecular data using MP and Bayesian approaches. Due to the large number of suboptimal trees recovered from parsimony analysis of the nonmolecular dataset, we first performed 1000 random-taxon-addition replicates with TBR branch swapping, but saved only 10 trees per replicate. We then used this pool of 10,000 trees as the starting point for an unbounded heuristic search. Parsimony analysis of the combined (molecular + nonmolecular) dataset did not exhibit this problem; therefore, we analyzed the combined dataset using unbounded heuristic searches with 1000 replicates of random-taxon addition and TBR branch swapping. For Bayesian analysis of the morphological dataset alone, we specified the Mkv model (Lewis, 2001) with a Γ-distributed rate parameter and ascertainment bias corrected for omission of constant characters (lset coding  =  variable). For Bayesian analysis of the combined (molecular + nonmolecular) dataset, we specified this same model for the morphological data and applied the best-fit model to each of the five gene partitions. As above, we allowed parameters to be estimated independently across all partitions and used the same MCMCMC settings. To assess the impact of the large amount of missing data from Chacodelphys, we performed all analyses that included nonmolecular characters with and without this taxon.

Online Data Archives

All of the new molecular sequences produced for this study have been deposited in GenBank with accession numbers FJ159278–FJ159314 and FJ 159316–FJ159370 (for a complete list of GenBank accession numbers of all analyzed sequences, old and new, see table 9). All of our datasets, selected ML and Bayesian analyses, and associated trees have been deposited on TreeBase ( http://www.treebase.org) with accession numbers S2164, M4107, and M4108. Our nonmolecular data matrix has also been deposited on MorphoBank ( http://morphobank.geongrid.org) with accession number X600.

Size and External Features

Most opossums are externally unremarkable mammals with pointed muzzles, large rhinaria, well-developed vibrissae, prominent eyes, membranous ears, nonspinous pelage, subequal limbs, pentadactyl feet, and naked tails. In many of these respects they resemble other plesiomorphic marsupials (e.g., Dromiciops and dasyurids) as well as certain unspecialized placentals (e.g., solenodontids, rice tenrecs, gymnures, and tree shrews). Closer inspection, however, reveals numerous distinctive and phylogenetically informative details of didelphid external morphology.

Size

Didelphids are small to medium-sized mammals. The smallest Recent species is probably Chacodelphys formosa, the young adult holotype of which had a head-and-body length of 68 mm and probably weighed about 10 g (Voss et al., 2004a). By contrast, the largest living opossum, Didelphis virginiana, can measure almost 500 mm in head-and-body length and weigh more than 3000 g (Hamilton, 1958). Most didelphids, however, range in head-and-body length from about 100 to 300 mm and weigh between about 20 and 500 g (table 4). Although some authors have recognized “large”, “medium”, and “small” opossums, taxonomic assignments to discrete size categories are often arbitrary and sometimes misleading. For example, Reig et al. (1987: character 25) scored Metachirus as “large” and Caluromys as “medium” in their phylogenetic analysis, but linear measurements and weights that we compiled suggest that these taxa are indistinguishable in size. More taxonomically comprehensive compilations of morphometric data will probably document a continuum of didelphid size distributions.

Table 4

External Measurements (mm) and Weights (g) of Exemplar Didelphid Species^a

By comparison, Old World marsupials include some species that are smaller and others that are much larger than any didelphid. The smallest living Australian species, for example, is said to be Planigale ingrami, with an adult weight of only about 4 g, and the largest is Macropus rufus, males of which are said to weigh as much as 85 kg (Dawson et al., 1989). The extinct Pleistocene species Diprotodon optatum, however, may have weighed almost 2800 kg (Wroe et al., 2004).

Rhinarium and Mouth

The rhinarium, a prominent pad of naked glandular skin surrounding the nostrils, is divided by a median crease or sulcus (sulcus medianus; Ade, 1999) that extends from between the nares to the upper lip in all examined didelphids. The part of the rhinarium that borders the upper lip (pars supralabialis) is broad, and its ventral margin is notched by one or two distinct grooves on each side of the median sulcus (fig. 1). Two ventrolateral grooves are present on each side in most examined taxa, but only a single groove is present in Chironectes, Didelphis, Lestodelphys, Lutreolina, Metachirus, Monodelphis, Philander, and Thylamys pallidior. Dasyurids and Dromiciops have rhinaria that are essentially similar in gross morphology to those of didelphids, with a pars supralabialis that makes broad contact with the upper lip; however, only a single ventrolateral groove is present in these taxa (Pocock, 1926: figs. 26–28). By contrast, the supralabial part of the rhinarium is reduced to a narrow philtrum in caenolestids and peramelids, such that the groove-bearing ventral rhinarial margin of other marsupials is effectively absent.

Figure 1

Ventral view of rhinarium in Thylamys pallidior (A, UMMZ 156349) and Marmosa robinsoni (B, UMMZ 117236). Both species have a broad pars supralabialis (psl) that contacts the upper lip. Only a single groove (g) is present along the ventral margin of the pars supralabialis on each side of the median sulcus (ms) in T. pallidior, whereas two ventrolateral grooves are present in M. robinsoni. Scale bars  =  2 mm.

Figure 2

Sexually dimorphic wrist morphology of Marmosops pinheiroi. Adult males (left, AMNH 267346) possess an externally obvious lateral carpal tubercle that is supported internally by an enlarged pisiform bone (arrows); females (right, AMNH 267342) do not exhibit this trait.

The mouth is large in all didelphids, with a posterior angle (angulis oris; Brown, 1971) that extends posteriorly to a point below the eye; the upper and lower oral margins are smooth or irregularly wrinkled and anatomically featureless. Most other marsupials have anatomically featureless oral margins like those of didelphids, but caenolestids are uniquely provided with reciprocating fleshy lappets of unknown function on the upper and lower lips (Osgood, 1921: pl. 2; Bublitz, 1987: fig. 4; Patterson and Gallardo, 1987: fig. 3).

Facial Vibrissae and Markings

Didelphid facial vibrissae are grouped into discrete tracts that are easily homologized with those described and illustrated by Pocock (1914), Lyne (1959), and Brown (1971), whose terminology is followed here. All of the taxa we examined (including representative species from every genus) exhibit well-developed mystacial, submental, interramal, superciliary (supraorbital), and genal vibrissae. Labeled illustrations of didelphid facial vibrissae are provided by Pocock (1914: fig. 1A) and Lyne (1959: fig. 2). Most other plesiomorphic marsupials closely resemble didelphids in cranial vibrissal traits, but caenolestids lack interramal vibrissae (Lyne, 1959).

Didelphids exhibit conspicuous taxonomic variation in facial markings (frontispiece; Voss and Jansa, 2003: fig. 3). In Caluromys, a median streak of dark fur, unconnected with any other dark marking, extends from the rostrum to the frontal region. 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. No nondidelphid marsupial that we examined has a dark midrostral stripe.

The fur surrounding the eye is not distinctively colored in Caluromysiops, Lutreolina, or Monodelphis, but most other didelphids have masklike markings. 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 Glironia, Lestodelphys, and “marmosines” (taxa formerly included in “Marmosa” sensu Tate, 1933). 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. A blackish mask is also present in Metachirus, Chironectes, and Philander, but the dark circumocular fur in these taxa is usually continuous with dark fur on the crown of the head. Among nondidelphid marsupials, dark circumocular masks are present in Dromiciops, some dasyurids (e.g., Sminthopsis crassicaudata), a few peramelemorphians (e.g., Echymipera kalubu), and some small arboreal diprotodontians (e.g., Petaurus breviceps); only one examined outgroup taxon, Echymipera kalubu, has a mask that is more or less continuous with dark coronal fur.

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. 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 rather than a homologue of the condition seen in Metachirus and Philander.) We have not examined any nondidelphid marsupial with pale supraocular spots.

Gular Gland

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.” External signs of glandular activity tend to be maximally developed in fully mature males (loc. cit.).

No unambiguously glandular throat patch was observed in any examined specimens of Caluromys, Caluromysiops, Chironectes, Hyladelphys, Lutreolina, Philander, or Tlacuatzin. By contrast, fully adult male specimens of Cryptonanus, Gracilinanus, Lestodelphys, and Thylamys usually exhibit well-developed gular glands. A gular gland is also present on the young adult male holotype (and only known skin) of Chacodelphys formosa. Other didelphid genera (Marmosa, Marmosops, and Monodelphis) include some species that consistently develop adult male gular glands and others that just as consistently show no trace of such organs (Voss and Jansa, 2003). Although adult male Didelphis often have discolored gular fur, no glandular skin is macroscopically distinguishable. Because we were not able to examine any fully adult male specimens of Glironia, the occurrence of gular glands in this taxon is unknown.

Gular (or sternal) glands that are macroscopically and histologically similar to those of didelphids are present in most dasyurids (Cooper et al., 2005), but other plesiomorphic outgroup taxa (e.g., caenolestids, peramelids, Dromiciops) seem to lack all external traces of glandular activity on the throat or chest.

Body Pelage

All didelphids have one or more tracts of postcranial vibrissae (Brown and Yalden, 1973), including ulnar-carpal vibrissae (at the wrist), medial antebrachial vibrissae (at or near the middle of the forearm), anconeal vibrissae (at the elbow), and/or calcaneal vibrissae (on the ankle). Lyne (1959) reported the occurrence of postcranial vibrissae in several didelphid species based on his examination of pouch young, whose sensory hair follicles are easily seen because they are not obscured by coat hairs. Unfortunately, postcranial vibrissae are much harder to observe on fully furred adult specimens, the only material commonly available for most species. We found ulnar-carpal and medial antebrachial vibrissae on most examined didelphids, whereas anconeal and calcaneal vibrissae were often inapparent.

All didelphids have soft (nonspinous) fur consisting of two or more distinct types of hairs whose density and morphology determine the appearance and texture of the coat. Some taxa (e.g., Caluromys) have somewhat woolly fur that does not lie flat or exhibit the glossy highlights typically seen in the pelts of many other taxa, but textural differences are hard to define by objective criteria that can be used for character-state definitions or taxonomic diagnoses. The only structural (nonpigmental) feature of didelphid body pelage that seems useful in this context is the presence of uniquely long, coarse guard hairs that project conspicuously from the underfur in species of Didelphis.

The dorsal body pelage of most didelphids is uniformly colored and unpatterned, usually some shade of brownish or grayish, but some taxa are distinctively marked (see illustrations in Eisenberg, 1989; Redford and Eisenberg, 1992; Pérez-Hernández et al., 1994; Reid, 1997; Eisenberg and Redford, 1999). Blackish transverse bars connected middorsally on a pale-grayish background, for example, characterize Chironectes; dark scapular stripes are unique to Caluromysiops; three longitudinal stripes are present in several species of Monodelphis (e.g., M. theresa); a grayish middorsum contrasting with reddish flanks is exhibited by other species in that genus (e.g., M. brevicaudata); and a grayish midbody contrasting with reddish head and rump is seen in others (e.g., M. emiliae). The subtle but consistently diagnostic “tricolor” shading of Thylamys and Lestodelphys was described by Tate (1933: 209):

Instead of the usual bicolor system composed of a dorsal color, paling a little on the sides, which is replaced at a generally well-marked transition line by a distinct ventral color, the elegans group [ =  Thylamys] displays three distinct shades, separated from each other along each side by two lines of transition. The additional lines are subdorsal, running from a point at the center of the frons [forehead] past the inner edge of each ear (not including it), and straight backward through scapulae and hips, where they again approach the median line of the body and merge with the tail. This pair of lines encloses the major part of the dorsal area of head and body, the color of the area being very dark brownish-gray or grayish fuscous. The fuscous area is pointed at front, projecting forward between the ears, and narrows again to a point as it merges with the dark color of the upper surface of the tail. The second [lateral] area, light gray in color, frequently tinged with buffy or yellowish, extends [on each side] between the dark dorsal region and the edge of the belly color at the normal transition line. Ventral color either buffy, grayish, or snowy white.

The hair bases of the dorsal fur are heavily pigmented, usually dark gray (or grayish), in most didelphids, but species of Didelphis uniquely exhibit white dorsal underfur.

The ventral pelage also exhibits noteworthy taxonomic variation among didelphids. In some species the ventral fur is “gray-based,” a descriptor that applies when the individual hairs are grayish basally and abruptly paler (usually whitish, yellowish, or brownish) distally; the overall color of the ventral fur then depends on the degree to which the dark basal pigmentation shows through the paler superficial hue. In other species, the ventral fur is partially or entirely “self-colored,” a descriptor that applies when the individual hairs have the same pigmentation (usually whitish) from root to tip. Because marked differences in the patterning of gray-based versus self-colored ventral pelage can occur among closely related (congeneric) species, such variation is often described and illustrated in the revisionary taxonomic literature (e.g., Patton et al., 2000: fig. 41). However, the existence of numerous intermediate conditions spanning the entire range of taxonomic variation in ventral color patterns (e.g., from entirely gray-based to completely self-white ventral fur) precludes meaningful phylogenetic scoring of this character.

Whereas most of the pelage colors (and color patterns) described above are preserved on museum skins that have been protected in dark cabinets from the bleaching effects of light, other colors that can be striking in life fade quickly after death. The ventral pelage of live Monodelphis emiliae, for example, has been described as “bright glowing violaceous pink” (Emmons, 1997: 34), a lurid hue that is not retained in any examined museum specimens. What little is known about such fugitive pigments (some of which fluoresce under ultraviolet light) was summarized by Pine et al. (1985).

Most other marsupials have unremarkable body pelage that essentially resembles the common didelphid condition as described above, but postcranial vibrissae are reduced or absent in a few clades (Lyne, 1959) and peramelemorphians have stiff, dorsally grooved guard hairs (illustrated in cross section by Lyne and McMahon, 1951: figs. 51, 60) that impart a distinctively harsh, spinous texture to their fur. Most other marsupials also resemble didelphids in having unpatterned dorsal fur, although there are noteworthy examples of apparently convergent markings (e.g., those of Dromiciops somewhat resemble Chironectes'; Marshall, 1978b) and some strikingly divergent ones (e.g., the transverse sacral barring seen in Perameles gunnii; Lyne, 1951: pl. 1B). All examined outgroup taxa have dark dorsal underfur.

Wrist

The wrists of males and females are morphologically similar and externally featureless in most didelphids, but striking sexual dimorphism is present in certain small arboreal and scansorial forms (Lunde and Schutt, 1999). Grossly enlarged glabrous tubercles supported internally by carpal ossifications are exhibited by large adult males of Cryptonanus, Gracilinanus, Marmosops, Tlacuatzin, 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 of it can be attributed to ontogeny: tubercles are consistently present in the largest adult male specimens of species in which such structures occur, whereas they may be lacking in some smaller (presumably younger) conspecific males. Lunde and Schutt (1999) plausibly suggest that these structures function as clasping devices during copulation.

Lateral carpal tubercles (fig. 2) are more widespread than medial carpal tubercles, occurring in all taxa that exhibit any externally obvious sexual dimorphism in the wrist. We found these structures to be consistently present in large adult male specimens of Cryptonanus, Gracilinanus, Marmosops, Tlacuatzin, and most species of Marmosa. In addition, we observed medial carpal tubercles (Lunde and Schutt, 1999: fig. 3) in Marmosa demerarae, M. mexicana, M. paraguayana, M. regina, M. robinsoni, and M. rubra. Other didelphids appear to lack sexually dimorphic carpal tubercles,2 but we were not able to determine whether or not such structures are present in Chacodelphys and Glironia because no fully adult male specimens of either genus are currently available for study.

Figure 3

Dorsal views of right forefeet of Monodelphis brevicaudata (A, AMNH 140466), Marmosops incanus (B, MVZ 197629), Marmosa robinsoni (C, AMNH 259983), and Caluromys philander (D, AMNH 7433) illustrating generic differences in claw size and digital proportions. Monodelphis and Marmosops both have mesaxonic forefeet in which the third digit (dIII) is longest, whereas Marmosa has a paraxonic forefoot in which dIII and dIV are subequal. In Caluromys, dIV is the longest manual digit.

The water opossum Chironectes has carpal tubercles that, uniquely, are neither sexually dimorphic nor ontogenetically variable. In this taxon, juveniles and adults of both sexes possess a large, fleshy process on the outside of the wrist, resembling a sixth finger, that is supported internally by the pisiform (Augustiny, 1942: fig. 16; Mondolfi and Medina, 1957: fig. 14; Oliver, 1976: fig. 1b).

No examined nondidelphid marsupial has carpal tubercles of any kind.

Manus

The didelphid hand is provided with five well-developed clawed digits (fig. 3). When the digits are flexed, their tips converge toward the center of the palm, and it seems likely that at least the arboreal and scansorial forms—which tend to have relatively longer fingers than terrestrial taxa (Lemelin, 1999; Kirk et al., 2008)—are capable of manual prehension. In most didelphids, the manual digits are more or less evenly spaced, but in some arboreal taxa (e.g., Caluromys, Caluromysiops, Marmosa) the gap between dII and dIII is somewhat larger than the gaps separating other pairs of adjacent fingers, and it is possible that these taxa are incipiently schizodactylous (sensu Haines, 1958). The pollex (dI) of Chironectes is set off from the other manual digits by a wide gap and appears to be pseudo-opposable (sensu Napier, 1961); in fluid-preserved specimens it is often folded across the palm (Augustiny, 1942: fig. 16).

Figure 4

Plantar view of right hind foot of Gracilinanus marica (redrawn from Boas, 1918: pl. 1, fig. 2). Pedal characters that distinguish didelphids from most other marsupials include eleuthrodactyly, a large opposable hallux (digit I), a grooming claw (Putzkralle) on digit II, and prominent dermatoglyph-bearing plantar pads.

According to Tate (1947), the third and fourth digits of the didelphid manus are subequal and longer than the other fingers (fig. 3C), 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 exhibit this condition include Chacodelphys, Chironectes, Didelphis, Lestodelphys, Lutreolina, Marmosops (fig. 3B), Metachirus, Monodelphis (fig. 3A), Philander, and Thylamys. Yet another alternative condition is seen in Caluromys and Caluromysiops, in which dIV is slightly but distinctly longer than dIII (fig. 3D).

Many didelphids have small, weakly recurved manual claws that do not extend much (if at all) beyond the fleshy apical pad of each digit (fig. 3B, C), but some arboreal taxa such as Caluromys (fig. 3D) and others that are strictly terrestrial such as Monodelphis (fig. 3A) have large manual claws that extend well beyond the apical pads. The large manual claws of Glironia (not illustrated) are unusually deep, laterally compressed, and strongly recurved. Although illustrated taxonomic differences in claw length are striking, intermediate morphologies observed among other forms (e.g., Hyladelphys) make it difficult to recognize discrete character states for taxonomic analysis.

Like most other plantigrade mammals (Whipple, 1904; Brown and Yalden, 1973), didelphids usually have two well-developed metacarpal pads (thenar, hypothenar) and four interdigital pads on the hairless ventral (plantar or volar) surface of the manus. These six pads encircle a central palmar region that is either smooth or sparsely tubercular (as in most didelphids; see Hershkovitz, 1997: fig. 6A) or densely covered with small convex tubercles (as in Chacodelphys, Lestodelphys, and most species of Thylamys; see Carmignotto and Monfort, 2006: fig. 3C). In almost all species, the epidermis that covers the plantar pads is sharply differentiated from that of the surrounding plantar surface because it is provided with dermatoglyphs (friction or papillary ridges; Hamrick, 2001) resembling those on human fingertips. Plantar pads tend to be larger and to have more pronounced dermatoglyphs in arboreal opossums (e.g., Caluromys) than in terrestrial forms (e.g., Monodelphis), but a few didelphids exhibit qualitatively different morphologies.

The plantar pads of two Brazilian species of Thylamys (illustrated by Carmignotto and Monfort, 2006: figs. 3a, 3b) are fused together and fill the entire palmar region rather than encircling a central palmar surface as in other didelphids. Additionally, only the apices of the plantar pads are provided with dermatoglyphs in T. velutinus, whereas the pads of T. karimii lack dermatoglyphs completely. Instead, the plantar pads are mostly (T. velutinus) or entirely (T. karimii) covered with small convex tubercles like those that occur on the center of the palm in other species of Thylamys.

The ventral surface of the manus in the water opossum uniquely lacks any trace of plantar pads. Instead, the palm of Chironectes is essentially flat and densely covered with microscopically dentate tubercles whose fine structure was described and illustrated by Brinkmann (1911) and Hamrick (2001). Numerous smooth, hemisperical papillae (not mentioned by Hamrick, 2001), however, are scattered at regular intervals among the dentate tubercles and presumably have a different function (Brinkmann, 1911).

Among other plesiomorphic marsupials, dasyurids and Dromiciops most closely resemble didelphids in manual morphology, both taxa having five well-developed clawed digits. However, whereas dIII and dIV are subequal in Dromiciops, dIII is distinctly longer than dIV in dasyurids. By contrast, dI and dV are conspicuously reduced and nail bearing in caenolestids (Osgood, 1921: pl. 2, fig. 4), and the same digits are vestigial—lacking claws or nails—in peramelemorphians (Lyne, 1951: fig. 11); of the remaining manual digits, dIII is distinctly the longest in both of these groups. Manual plantar pads are indistinct or absent in all examined peramelemorphians but they are distinct in dasyurids, caenolestids, and Dromiciops. Whereas the plantar pads of Dromiciops and at least some dasyurids (e.g., Murexia) have dermatoglyphs, the plantar pads of other dasyurids (e.g., Sminthopsis crassicaudata) are tubercular, and those of caenolestids are smooth. The central palmar surface is essentially smooth (or irregularly creased) in caenolestids and Dromiciops, but it is densely tubercular in examined dasyurids and peramelids.

Pes

Most didelphids have an eleuthrodactylous hind foot with five well-developed digits that are all separate and freely movable (fig. 4). The only exception is Chironectes, whose pedal digits are bound together by webs of skin to form a paddlelike swimming organ (illustrated by Augustiny, 1942: Mondolfi and Medina, 1957; Oliver, 1976; Hershkovitz, 1997). Although Bensley (1903), Tate (1933), and Kirsch (1977b) suggested that some didelphids exhibit incipient fusion of dII and dIII, we have not observed any specimen with a hind foot that even remotely resembles the syndactylous condition seen in peramelemorphians and diprotodontians.3

Figure 5

Mammary morphology of adult female specimens of Hyladelphys kalinowskii (A, AMNH 267339) and Marmosops parvidens (B, AMNH 267344). Only two pairs of inguinal-abdominal teats are present in H. kalinowskii, which is unusual among didelphids in lacking an unpaired median teat; its mammary formula is 2–0–2  =  4. By contrast, the illustrated specimen of M. parvidens has four pairs of inguinal-abdominal teats plus an unpaired median teat (4–1–4  =  9).

The didelphid hallux (dI) is well developed, set off at a wide angle from the other pedal digits, and fully opposable. As in other marsupials, the hallux is claw- and nailless in opossums. Although this digit tends to be maximally developed in arboreal forms, it is always long enough to contact the tips of the other digits when they are flexed, even in such strictly terrestrial taxa as Lestodelphys, Lutreolina, and Monodelphis.

The remaining pedal digits bear well-developed claws that are laterally compressed and dorsoventrally recurved like those of most other plantigrade mammals (Brown and Yalden, 1973). The claws on dIV and dV are more or less symmetrically crescentic, with an unguis that is equally extensive on the axial and abaxial surfaces, and a subunguis that is only exposed ventrally. Digit II, however, bears a strikingly asymmetrical grooming claw (Putzkralle; Boas, 1918) with an abaxially exposed subunguis and a rounded, spoon-shaped apex. The subunguis is sometimes also exposed abaxially on the claw of dIII, but the asymmetry is consistently less than that of the second digital claw, and the third claw never has a rounded tip.

In the strictly terrestrial opossums 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). By contrast, dII, dIII, and dIV 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. It is noteworthy that the relative lengths of pedal digits do not strictly covary with those of the manual digits, as they might be expected to do if transformations of the pes and manus were functionally or developmentally determined by the same factors. For example, whereas dIV is the longest pedal digit in Marmosops, dIII is the longest manual digit in that genus.

In most didelphids (e.g., Philander; Hershkovitz, 1997: fig. 6B) the entire ventral surface of the hind foot is macroscopically naked from the apex of the heel to the tips of the toes. Under high magnification, a fine plantar pelage of very short hairs partially or completely covers the heel in some taxa (e.g., Marmosops), but intermediate conditions in other forms and variation among conspecific individuals suggest that such microscopic details are not a reliable basis for either phylogenetic analysis or taxonomic diagnosis. By contrast, a morphologically distinctive condition occurs in Lestodelphys and Thylamys (see Carmignotto and Monfort, 2006: fig. 3), whose heels are entirely covered by coarse (macroscopically visible) fur.

Most didelphids have six separate plantar pads like many other plantigrade mammals (Brown and Yalden, 1973), but several patterns of taxonomic variation are noteworthy: (1) The position normally occupied by the hypothenar (lateral tarsal) and fourth interdigital pads in most didelphids is occupied by a single elongate pad in Caluromys, Caluromysiops, and Glironia; although this large structure is plausibly interpreted as the result of fusion (hypothenar + interdigital 4), this scenario is complicated by the occasional presence (e.g., in AMNH 273038) of a small proximal metatarsal pad that might be a vestigial hypothenar. (2) The hypothenar is absent or variably present but clearly vestigial in Monodelphis, Lutreolina, Philander, and Didelphis. (3) The thenar (medial tarsal) and first interdigital pads are often in contact or indistinguishably fused among arboreal taxa (e.g., Caluromys and Marmosa), but a continuum of intermediate conditions precludes any unambiguous distinction between these morphologies and the separate conditions seen in terrestrial forms (e.g., Monodelphis and Thylamys). (4) Interdigital 2 is much larger than interdigital 3 in most arboreal/scansorial forms (e.g., Marmosa, Marmosops), but they are subequal in some terrestrial taxa (e.g., Monodelphis and Lutreolina).

The pedal plantar pads of most didelphids are provided with dermatoglyphs like those of the manual plantar pads. However, the dermatoglyph-bearing epithelium is much reduced in Thylamys velutinus, and in T. karimii the pedal plantar pads are entirely tubercular (lacking dermatoglyphs completely; Carmignotto and Monfort, 2006: fig. 3). As in so many other aspects of its appendicular morphology, Chironectes is unique among didelphids in lacking any trace of pedal plantar pads.

The eleuthrodactylous hind foot of Dromiciops resembles the common didelphid morphotype in having a large opposable hallux, a grooming claw on dII, dIV > dIII, and dermatoglyph-bearing plantar pads (Hershkovitz, 1999: fig. 21). By contrast, the hind foot of peramelids is syndactylous (with fused dII and dIII; Hall, 1987; Weisbecker and Nilsson, 2008); the hallux is small and effectively nonopposable in caenolestids (Osgood, 1921: pl. 2, fig. 3), dasyurids (Thomas, 1888: pl. 23, fig. 8), and peramelids (Lyne, 1951: fig. 12); dIII and dIV are subequal in examined caenolestids and dasyurids, both of which groups also lack a grooming claw on dII; peramelids do not have distinct plantar pads; the plantar pads of caenolestids are smooth; and the plantar pads of some dasyurids are tubercular. The plantar epithelium of the heel is macroscopically naked in most of these outgroup taxa, but the underside of the heel is coarsely furred in some dasyurids (e.g., Sminthopsis crassicaudata).

Pouch and Mammae

Based on our first-hand examination of parous adult female specimens, pouchlike enclosures for nursing young are unequivocally present or absent among didelphids. Although our sample sizes were always small, we observed no intraspecific variation in this aspect of female reproductive morphology, nor did we observe any intermediate condition between absence and presence of a pouch. Despite the fact that several distinctly different pouch configurations can be recognized among opossums, we provisionally recognize all pouches 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 Cryptonanus, Glironia, Gracilinanus, Hyladelphys, Lestodelphys, Marmosa, Marmosops, Metachirus, Monodelphis, Thylamys, and Tlacuatzin. By contrast, well-developed pouches were consistently found to be present in suitably prepared parous adult females of Caluromys, Chironectes, Didelphis, and Philander. Although Lutreolina was described as pouchless by Thomas (1888), Cabrera (1919), and Marshall (1978a), two fluid-preserved parous females that we examined (UMMZ 166634, USNM 536827) had pouches exactly resembling the morphology illustrated and described by Krieg (1924: fig. 11). Caluromysiops is said to have a pouch (Izor and Pine, 1987; Reig et al., 1987), but no explicit description or illustration of the female reproductive anatomy of this genus is available, nor were we able to examine suitably preserved parous female specimens. The presence or absence of a pouch remains undocumented for many opossums, notably Chacodelphys. Contradictory literature accounts of a pouch as present or absent in Metachirus were discussed by Voss and Jansa (2003).

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).

In all marsupials that possess a pouch the mammae are contained within it, but the mammae of pouchless taxa are variously distributed (Tate, 1933: fig. 3). In most pouchless didelphids (e.g., Glironia, Marmosa, Metachirus) the mammae are confined to a more or less circular inguinal/abdominal array that occupies the same anatomical position as the pouch in taxa that possess a marsupium. However, other pouchless opossums (e.g., Cryptonanus guahybae, Marmosops incanus) have bilaterally paired mammae that 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” or “thoracic” mammae (e.g., by Tate, 1933; Reig et al., 1987).

Most didelphids have, in addition to bilaterally paired mammae, an unpaired median teat that occupies the ventral midline approximately in the center of the circular abdominal-inguinal array (fig. 5B). Mammary counts for didelphids are therefore usually odd-numbered, an allegedly diagnostic attribute previously noted by Bresslau (1920), Osgood (1921), and Tate (1933, 1947, 1948a). The only exceptions that we have encountered are Caluromys lanatus, Glironia venusta,4 and Hyladelphys kalinowskii (fig. 5A), examined females of which have four mammae in two abdominal-inguinal pairs with no trace of a median teat. By convention, didelphid mammary complements are summarized by formulae representing the right-side (R), median (M), left-side (L), and total (T) teat counts in the format R–M–L  =  T. Thus, the mammary complement of the illustrated specimen of Marmosops parvidens could be written as 4–1–4  =  9, whereas that of Hyladelphys kalinowskii would be 2–0–2  =  4. Occasionally, lateral teats are unpaired, resulting in even-numbered totals for taxa that possess an unpaired median teat, but such anomalies are easily distinguished by formulae (e.g., 3–1–4  =  8 if the anteriormost right teat of AMNH 267344 were really missing in fig. 5B).

Figure 6

Dorsal and ventral cranial views of Marmosa murina showing principal osteological features mentioned in the text. Abbreviations: als, alisphenoid; atp, alisphenoid tympanic process; bo, basioccipital; bs, basisphenoid; cc, carotid canal; ect, ectotympanic; fm, foramen magnum; fpj, frontal process of jugal; fro, frontal; gf, glenoid fossa; gpa, glenoid process of alisphenoid; ip, interparietal; jug, jugal; lac, lacrimal; max, maxillary; mp, mastoid process (of petrosal); nas, nasal; occ, occipital condyle (of exoccipital); of, orbital fossa; pal, palatine; par, parietal; pcf, paracanine fossa; pogp, postglenoid process (of squamosal); pop, postorbital process; pre, premaxillary; prgp, preglenoid process (of jugal); pro, promontorium (of petrosal); ps, presphenoid; pt, pterygoid; rtp, rostral tympanic process (of petrosal); sq, squamosal; tcf, transverse canal foramen; tf, temporal fossa; za, zygomatic arch.

Among other plesiomorphic marsupials that we examined, a marsupium is absent only in caenolestids and some dasyurids (e.g., Murexia). The pouch of Dromiciops consists of lateral skin folds opening medially (Hershkovitz, 1999: fig. 14), whereas the pouches of other dasyurids are variously configured (Woolley, 1974), and those of peramelids open posteriorly. Most nondidelphid marsupials lack an unpaired median teat, and therefore normally have even-numbered mammary counts, but an unpaired median teat is said to be present in Rhyncholestes (e.g., by Osgood, 1924; Patterson and Gallardo, 1987).

Cloaca and Male Genitalia

In most didelphids the openings of the urogenital and rectal ducts are inguinal, 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 inguinal skin (Nogueira et al., 2004: fig. 1). Conflicting descriptions of didelphid cloacal morphology by Hershkovitz (1992a, 1992b) were discussed and refuted by Voss and Jansa (2003). All examined nondidelphid marsupials have an inguinal cloaca resembling the common didelphid condition, with the conspicuous exception of Dromiciops, in which the cloaca is basicaudal (Hershkovitz, 1999: fig. 12).

Descriptions of didelphid male genitalia (e.g., by Ribeiro and Nogueira, 1990; Martinelli and Nogueira, 1997; Nogueira et al., 1999a, 1999b; Nogueira and Castro, 2003) were recently reviewed and summarized by Nogueira et al. (2004), who tabulated morphological comparisons among 18 species in 12 genera. All examined opossums have a bifid penis (contra Hershkovitz, 1992b), but didelphid male genitalia exhibit conspicuous taxonomic variation in length, shape, disposition of the urethral grooves, presence/absence of diverticula, and other details. Unfortunately, the genitalic characters of many species remain unstudied, and we were unable to extend Nogueira et al.'s (2004) observations for this report.

Tail

Most didelphids have a tail that is substantially longer than the combined length of the head and body, but some taxa are much shorter tailed (table 5). The known range of relative tail length in the family is bracketed on the one hand by Gracilinanus emiliae, an arboreal species with a tail that may be almost twice as long as the combined length of its head and body (Voss et al., 2005: table 2), and on the other by some terrestrial forms (e.g., Monodelphis spp.) with a tail that may be less than half as long as the head and body. However, arboreal taxa are not always longer tailed than terrestrial forms. Tree-dwelling Glironia venusta, for example, has a tail that is subequal in length to its head and body, whereas ground-dwelling Metachirus nudicaudatus is much longer tailed.

Table 5

Relative Tail Length of Exemplar Species from 18 Didelphid Genera

Body pelage (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 Glironia (see da Silva and Langguth, 1989: fig. 1), Caluromysiops, and some species of Caluromys and Monodelphis (see Voss et al., 2001: fig. 29A). 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, Philander, and some species of Marmosa (e.g., M. paraguayana). In most other didelphids, however, body fur does not extend more than a short distance, if at all, onto the tail base.

The exposed skin of didelphid tails is variously pigmented. Some species have uniformly pigmented (concolored) tails that are usually some shade of grayish or brownish, but pale markings are not uncommon. Many species, for example, have parti-colored tails that are paler distally than basally. In most taxa with pale tail tips, the basal color is grayish or brownish and the transition is either gradual (the basal color fading to whitish distally; e.g., in Marmosops ocellatus) or irregularly mottled (e.g., as in Caluromys derbianus; frontispiece). However, in most specimens of Chironectes (frontispiece), Didelphis, Lutreolina, and Philander, the tail is blackish basally with an abruptly white tip. Dorsoventrally bicolored tails (which are grayish or brownish above and pale on the underside) represent another widespread pattern of didelphid caudal marking that is indistinct in some species but sharply defined in others. Different patterns of caudal markings often distinguish closely related species and can be used in combination with other external traits to identify congeneric taxa in the field (Voss et al., 2004b: table 4).

The epidermal scales that cover the unfurred nonprehensile surfaces of didelphid tails differ taxonomically in shape and arrangement (Tate, 1933: fig. 2). Square or rectangular caudal scales in unambiguously annular series occur in Chacodelphys, Cryptonanus, Thylamys, Tlacuatzin, and most species of Gracilinanus. Although individual cutaneous scales are indistinct in Lestodelphys and Monodelphis, the annular arrangement of caudal bristles in both taxa suggest that they also conform to the Thylamys pattern. By contrast, rhomboidal (diamond-shaped) or hexagonal scales in spiral series occur in Caluromys, Caluromysiops, Chironectes, Didelphis, Lutreolina, Marmosops, Philander, and most species of Marmosa. A few taxa, however, have caudal scales that appear to be morphologically intermediate between the annular and spiral morphotypes. In these taxa (e.g., Hyladelphys, Metachirus, Marmosa mexicana) the caudal scales are arrayed in annular series over the vertebral articulations and in more or less spiral series elsewhere. Glironia cannot be assigned to any of these caudal-scale patterns because the dorsal and lateral surfaces of its tail are covered by body fur and the entire ventral surface is modified for prehension (da Silva and Langguth, 1989: fig. 1).

Three stiff, bristlelike hairs 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 hard to distinguish in Monodelphis, the caudal hairs of species that we examined usually emerge from the skin in triplets, so this genus appears to conform to the widespread didelphid condition. The number of hairs normally associated with each caudal scale is impossible to assess in Glironia (whose dorsal caudal surface is furry and whose ventral caudal surface is modified for prehension; see above and below) and Lestodelphys (on which no discrete clustering within rows of caudal hairs could be distinguished with confidence). 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 most species of Marmosops 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).

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 Chacodelphys, Chironectes, Lestodelphys, Lutreolina, Metachirus, and Monodelphis are covered with unmodified scales from base to tip (for illustrations of Chironectes and Lutreolina, see Hershkovitz, 1997: fig. 7). In taxa conforming to this morphology, the tail tip is sometimes 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 (for illustrations of Didelphis and Philander, see Hershkovitz, 1997: fig. 7). Some taxa additionally exhibit caudal modifications for basal prehension. Whereas 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), the scales on the underside of the base of the tail are heavily cornified, forming hard raised tubercles in Glironia (see da Silva and Langguth, 1989: fig 1a), Caluromysiops, and some species of Caluromys.

The tail is a slender, muscular organ in most didelphids, but Thylamys and Lestodelphys have incrassate tails (Morton, 1980) in which fat is seasonally deposited. Incrassate tails can be recognized superficially by their characteristically swollen outline and soft texture in fresh and fluid-preserved material, and by their flattened, grease-stained appearance in most skins. When the external aspect of the tail leaves some room for doubt, the presence or absence of subcutaneous adipose tissue is easily determined by dissection. For example, a midventral incision that we made near the base of the tail of MZUSP 32097 provided unambiguous confirmation of Carmignotto and Monfort's (2006) statement that the tail of Thylamys macrurus is incrassate (contra Palma, 1997; 6162Creighton and Gardner, 2008).

Among other marsupials, representative caenolestids, dasyurids, and peramelemorphians that we examined have nonprehensile tails that are uniformly pigmented, annularly scaled, and lack basal extensions of body pelage; most are slender, muscular organs, but those of Rhyncholestes and some dasyurids (e.g., Sminthopsis crassicaudata) are incrassate. The tail of Dromiciops is incrassate, annularly scaled, and distally prehensile, with a conspicuous extension of body pelage onto the basal 1/4 to 1/3 of the caudal dorsum.

Cranium and Mandible

The best general description of didelphid cranial morphology is Wible's (2003) account of the skull of Monodelphis brevicaudata, which includes a review of the literature on didelphid cranial ontogeny, myology, and other relevant topics. Whereas Wible's description is mostly limited to external cranial features, Rowe et al. (2005) provided a detailed description of internal structures of the nasal skeleton of a closely related species, M. domestica. Together, these publications provide detailed and abundantly illustrated accounts of the didelphid head skeleton as exemplified by two species in a single genus that is increasingly popular among biomedical researchers.

Other didelphids, however, differ from Monodelphis in numerous aspects of cranial morphology (Voss and Jansa, 2003), including some characters that remain undescribed in the literature. Additionally, the literature contains no comprehensive account of the cranial features by which didelphids as a group differ from other marsupials, although important comparative observations were made by Osgood (1921), Tate (1947, 1948b), Archer (1976a), Reig et al. (1987), Marshall et al. (1995), Wroe (1997), Muizon (1998), and Sánchez-Villagra and Wible (2002), among others. Following a brief consideration of overall cranial shape, the following accounts summarize relevant cranial comparisons in a roughly anterior to posterior sequence along the dorsolateral contours of the skull, starting with the rostrum and nasal cavity and proceeding to the orbital region and braincase. A second sequence describes ventral cranial features beginning with the palate and proceeding to the basicranium and occiput. The principal osteological features of the didelphid skull and mandible are illustrated in figures 6 and 7.

Figure 7

Left lateral cranial and mandibular views of Marmosa murina showing principal osteological features mentioned in the text. Abbreviations: als, alisphenoid; ap, angular process; atp, alisphenoid tympanic process; conp, condylar process; corp, coronoid process; ect, ectotympanic; exo, exoccipital; fpj, frontal process of jugal; fr, foramen rotundum; fro, frontal; hpp, hamular process of pterygoid; iof, infraorbital foramen; ip, interparietal; jug, jugal; lac, lacrimal; lc, lambdoid crest; lf, lacrimal foramina; maf, masseteric fossa; max, maxillary; mef, mental foramina; nas, nasal; pal, palatine; par, parietal; pc, pars cochlearis (of petrosal); pcf, paracanine fossa; pm, pars mastoideus (of petrosal); pop, postorbital process; pre, premaxillary; rmf, retromolar fossa; sq, squamosal; ssf, subsquamosal foramen; sup, supraoccipital; zps, zygomatic process of squamosal.

Figure 8

Ventral view of rostrum in Marmosa rubra (A, MVZ 153280) and Monodelphis brevicaudata (B, AMNH 257203) illustrating taxonomic differences in premaxillary morphology. A broad, shelflike rostral process (rp) extends the suture between right and left bones well anterior to the incisors in M. rubra, but the premaxillae form only narrow alveolar rims anterior to I1 in M. brevicaudata, where the left and right bones are separated by a small tissue-filled gap (not a distinct suture).

Shape Variation

Didelphids exhibit conspicuous variation in overall cranial shape that is illustrated in the taxonomic accounts below (figs. 37–Figure 38Figure 39Figure 40Figure 41Figure 42Figure 43Figure 44Figure 45Figure 46Figure 47Figure 48Figure 49Figure 50Figure 51Figure 52Figure 5354). Although some shape variation is obviously allometric (larger opossums tending to have relatively larger rostra and temporal fossae but relatively smaller orbits and braincases than smaller opossums), size-independent shape contrasts are also obvious. Thus, the rostrum is relatively short and broad in both Hyladelphys and Lutreolina, whereas the rostrum is relatively long and narrow in Marmosops and Metachirus.5 A continuous taxonomic series of intermediate proportions makes any qualitative coding of such variation an arbitrary exercise, but shape differences are visually conspicuous and are sometimes useful for diagnosing clades discovered by analyzing other data. The nascent literature on didelphid cranial morphometrics is not large, but it includes studies of intraspecific ontogenetic variation (Abdala et al., 2001; Flores et al., 2003), multivariate comparisons of conspecific species (e.g., Ventura et al., 1998, 2002; Cerqueira and Lemos, 2000), analyses of evolutionary rates (Lemos et al., 2001), geometric modeling of taxonomic variation (Astúa de Moraes et al., 2000), and preliminary attempts to explain shape differences in functional terms (e.g., Medellín, 1991).

Rostrum

The premaxillae of many didelphids (Caluromysiops, Chacodelphys, Chironectes, Cryptonanus, Didelphis, Glironia, Hyladelphys, Lestodelphys, Lutreolina, Metachirus, Monodelphis, Philander, Thylamys, Tlacuatzin) are short, terminating abruptly in front of the incisors, and often lack a definitive suture between the left and right elements (fig. 8B). By contrast, in other didelphids the premaxillae are produced anteriorly as a more or less acutely pointed shelflike process that extends the bony rostrum beyond I1 and contains a distinct suture between the right and left bones (fig. 8A). Taxa that exhibit a well-developed premaxillary rostral process include Caluromys, Gracilinanus, Marmosa, and most examined species of Marmosops. Aspects of rostral process development that impact phylogenetic character scoring, including some discrepancies and ambiguities in the published literature on didelphid premaxillary morphology, were previously discussed by Voss and Jansa (2003: 22–23).

Figure 9

Detail of the anterior orbital region in Marmosops parvidens (A, AMNH 267359) and M. pinheiroi (B, AMNH 267345). In M. parvidens the lacrimal foramina (lf) are concealed from lateral view inside the orbit, whereas the lacimal foramina are laterally exposed anterior to the orbit in M. pinheiroi. Other abbreviations: fro, frontal; jug, jugal; lac, lacrimal; max, maxillary; nas, nasal; pal, palatine.

Didelphid nasal bones always extend anteriorly beyond the vertically oriented facial processes of the premaxillae to form a well-defined median apex, and they always extend posteriorly between the lacrimals. Despite such consistency, taxonomic differences in nasal length can be defined with respect to other anatomical landmarks. The tips of the nasal bones are produced anteriorly beyond I1 in most opossums, and the nasal orifice is consequently not visible from a dorsal perspective (figs. 37–Figure 38Figure 39Figure 40Figure 41Figure 42Figure 4344, 49–Figure 50Figure 51Figure 52Figure 5354). In Chironectes, Didelphis, Lutreolina, and Philander, however, the tips of the nasals do not extend so far anteriorly, and the nasal orifice is dorsally exposed (figs. 45–Figure 46Figure 4748).

The nasals are wider posteriorly than anteriorly in most didelphids, but they are not so wide posteriorly as to contact the lacrimals (except as rare, often unilateral, variants). The transition from the narrow anterior part of each nasal to the broader posterior part may be gradual or abrupt and usually occurs at or near the maxillary-frontal suture. By contrast with this widespread condition, a few taxa (Chacodelphys, Thylamys, Marmosops incanus) have uniformly narrow nasals with subparallel lateral margins (e.g., fig. 54). Due to the usual absence of nasolacrimal contact, the maxillae and frontals are normally in contact on both sides of the rostrum.

Most of the lateral surface of the didelphid rostrum is formed by the maxilla, which is prominently perforated above P2 or P3 by the infraorbital foramen. The foramen is notably larger in Chironectes than in other didelphids, presumably to accommodate the hypertrophied infraorbital nerve that conducts trigeminal sensory fibers from the well-developed mystacial vibrissae of this semiaquatic taxon (Sánchez-Villagra and Asher, 2002). Aside from small nutrient foramina that may be variably present or absent in some taxa, there are no other lateral rostral openings in didelphids; in particular, no fenestra is ever present at or near the conjunction of the maxillary, nasal, and frontal bones.

Other marsupials exhibit a wide range of rostral morphologies. In all examined dasyuromorphians, for example, the nasals are very short and do not extend anteriorly beyond the facial processes of the premaxillae; rather than forming an acute apex, they are notched medially, and the incisive foramina (in the floor of the nasal orifice) are exposed to dorsal view (e.g., in Dasyurus; Flores et al., 2006: fig. 4B). In Recent peramelemorphians, the nasals do not extend posteriorly between the lacrimals (Freedman, 1967: fig. 2), whereas Recent caenolestids have a prominent fenestra on each side of the rostrum at or near the conjunction of the nasal, frontal, and maxillary bones (Thomas, 1895; Osgood, 1921: pl. 20, fig. 2). In a few Old World marsupials (e.g., Lasiorhinus, Tarsipes) the nasals contact the lacrimals (separating the maxillae from the frontals), whereas the premaxillae contact the frontals (separating the nasals from both the lacrimals and the maxillae) in dactylopsiline petaurids (Tate, 1948a). By contrast, many other marsupials (e.g., Dromiciops) essentially resemble didelphids in rostral morphology.

Nasal Cavity

The complex morphology of the bony turbinals inside the didelphid nasal cavity is difficult to visualize without X-ray computed tomography, which has only been used to describe the endonasal skeleton of Monodelphis domestica (see Rowe et al., 2005). However, some turbinal structures are accessible to direct inspection. The maxilloturbinals, for example, 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. In most didelphids, the greater curvature of each maxilloturbinal scroll throws off secondary lamellae, which branch to form tertiary lamellae, some of which branch again to form an elaborate dendritic mass that fills much of the ventrolateral part of the nasal lumen. By contrast, the maxilloturbinals are simple, slender, unbranched (or sparsely ornamented) scrolls in all examined species of Monodelphis.

The morphology of the nasal cavity remains to be surveyed among other marsupials. However, some peramelemorphians (e.g., Echymipera, Perameles) appear to have simple (unbranched) maxilloturbinals, whereas the maxilloturbinals of examined dasyurids (e.g., Murexia, Sminthopsis) and Dromiciops closely resemble the elaborately dendritic condition seen in most didelphids. A monographic study of marsupial endonasal features based on high-resolution X-ray computed tomography (Macrini and Voss, in preparation) will doubtless provide additional points of comparison among didelphids and other marsupial clades.

Zygomatic Arch

The bones comprising the zygomatic arch exhibit few variable features among didelphids. The maxillary-jugal suture is always more or less straight or irregularly crescentic, a frontal process of the jugal is invariably present, and the jugal-squamosal suture is always deeply inflected (typically <- or ⊂- shaped). Likewise, a faceted preglenoid process of the jugal and a well-developed postglenoid process of the squamosal are always present, and a distinct glenoid (entoglenoid) process of the alisphenoid consistently forms part of the posterior zygomatic root. Although the zygomatic arch tends to be more gracile, to have a more pronounced suborbital deflection, and to have a more distinct frontal process in small opossums (with relatively large eyes and weakly developed masticatory muscles; e.g., Hyladelphys; fig. 40) than in large opossums (with relatively smaller eyes but massive masticatory muscles; e.g., Lutreolina; fig. 47), most didelphids exhibit intermediate zygomatic morphologies.

Contrasting features of the zygomatic region among other marsupial groups include: (1) the deeply inflected maxillary-jugal suture of peramelemorphians, which divides the jugal into distinct anterodorsal and anteroventral processes flanking a well-developed nasolabial fossa (Filan, 1990); (2) the absence of a frontal process of the jugal in caenolestids and some peramelemorphians (Osgood, 1921); (3) the absence of a faceted preglenoid process of the jugal in several Old World clades (e.g., Thylacinus and macropodoids); (4) the absence of a distinct postglenoid process of the squamosal in Hypsiprymnodon, Tarsipes, and vombatids; and (5) the absence of a glenoid process of the alisphenoid in Myrmecobius and many diprotodontians.

Orbital Mosaic

The anteriormost part of the didelphid orbit is formed by the lacrimal, which is always prominently exposed in lateral view. The lacrimal is perforated by one or more lacrimal foramina that are sometimes concealed within the orbit (fig. 9A) but usually open laterally on or near the orbital margin (fig. 9B). Most didelphids normally have two lacrimal foramina on each side, but Chironectes, Hyladelphys, and some populations of Didelphis virginiana usually have just one lacrimal foramen, and many other didelphids that normally have two lacrimal foramina occasionally have a single foramen on one or both sides of the skull. The orbital margin formed by the lacrimal is smoothly rounded in all didelphids, none of which exhibits lacrimal tubercles (e.g., like those seen in macropodids; Wells and Tedford, 1995: fig. 9) or distinct crests (as in Myrmecobius and some peramelemorphians). Unlike the condition seen in some Old World marsupials with maxillary-frontal contact on the medial wall of the orbit (Flannery et al., 1987: fig. 3), the didelphid lacrimal is always in posteroventral contact with the palatine.6

Figure 10

Oblique dorsolateral view of the left orbital floor in Thylamys venustus (A, AMNH 263562) and Monodelphis peruviana (B, AMNH 272695) illustrating taxonomic differences in sutural patterns. In Thylamys (and most other didelphids), the maxillary (max) and alisphenoid (als) bones are separated by the palatine (pal), but the alisphenoid extends anteriorly across the palatine to contact the maxillary in Monodelphis. Other osteological abbreviations: fro, frontal; hp, hamular process of pterygoid; jug, jugal; lac, lacrimal; par, parietal; sq, squamosal.

The medial wall of the didelphid orbit is perforated by several openings, including the sphenopalatine foramen (always in the palatine bone), the ethmoid foramen (in the suture between the orbitosphenoid and frontal), the sphenorbital fissure (between the palatine, orbitosphenoid, alisphenoid, presphenoid, and sometimes the pterygoid), and the foramen rotundum (in the alisphenoid). The configuration of these orbital perforations in all examined taxa is essentially similar to that illustrated for Monodelphis by Wible (2003: fig. 4). In particular, the foramen rotundum is always exposed to lateral view behind the sphenorbital fissure, from which it is invariably separated by a bony partition.7 Many other marsupials essentially resemble didelphids in these features, but some exhibit noteworthy differences. In Dromiciops, for example, the foramen rotundum is concealed from lateral view within a common vestibule that it shares with the sphenorbital fissure in the rear of the orbit (Giannini et al., 2004).

The maxillary and alisphenoid bones are usually separated by the palatine on the orbital floor of most didelphids (fig. 10A). By contrast, the maxillary and the alisphenoid are consistently in contact on the orbital floor of Lutreolina and Monodelphis (fig. 10B). The maxillary and alisphenoid are separated by the palatine in most nondidelphid marsupials, but the maxillary and alisphenoid broadly contact one another in all examined macropodoids (e.g., Macropus; Wells and Tedford, 1995: fig. 9A).

Figure 11

Dorsal views of the interorbital region in Marmosa murina (A, AMNH 267368), Marmosops noctivagus (B, AMNH 262402), and Gracilinanus agilis (C, MVZ 197439) illustrating taxonomic differences in the development of processes and beads. Distinct postorbital processes (pop) are normally present in fully adult specimens of most Marmosa species, but they are usually absent in Marmosops and Gracilinanus. Supraorbital beads (b), consisting of upturned dorsally grooved ridges, are present in some but not all species of Marmosops. Other abbreviations: ioc, interorbital constriction; poc, postorbital constriction. Scale bars  =  5 mm.

Interorbital Region

The morphology of the interorbital region is highly variable among didelphids. The frontals produce distinct postorbital processes in several genera, most of which are rather large opossums (e.g., Caluromys, Philander), but some of which are small (e.g., Tlacuatzin). In Caluromys, Caluromysiops, Glironia, Tlacuatzin, and most species of Marmosa (fig. 11A) the postorbital processes are flattened and more or less triangular, but the postorbital processes of Chironectes, Didelphis, Lutreolina, and Philander are often bluntly pyramidal or hornlike projections. Because frontal outgrowths tend to develop late in postweaning ontogeny (Abdala et al., 2001), postorbital processes are often absent in juvenile and subadult specimens of taxa that normally exhibit them when fully grown. The postorbital processes of Glironia uniquely consist of both frontal and parietal moieties (fig. 37).

Figure 12

Lateral view of posterior braincase in Marmosa murina (A, AMNH 272816) and Marmosops impavidus (B, AMNH 272709), illustrating the presence of a fenestra (fen) that exposes the petrosal (pet) between the parietal (par) and squamosal (sq) in Marmosops. The petrosal is not laterally exposed by a fenestra in the parietal-squamosal suture of Marmosa. Other abbreviations: fc, fenestra cochleae; ssf, subsquamosal foramen.

Among didelphids that normally lack distinct postorbital processes (Chacodelphys, Cryptonanus, Gracilinanus, Lestodelphys, Marmosops, Metachirus, Monodelphis, Thylamys), the supraorbital margins of the frontals may be beaded (with a dorsally upturned edge) or more or less smooth and featureless. Taxa with supraorbital beads include some species of Gracilinanus (e.g., G. emiliae), Hyladelphys, some species of Marmosops (e.g., M. noctivagus; fig. 11B), Marmosa rubra, and Metachirus. By contrast, the supraorbital margins are consistently smooth and essentially featureless in most specimens of Cryptonanus, Lestodelphys, Thylamys, and other species of Marmosops and Gracilinanus (e.g., G. agilis; fig. 11C).

Distinct postorbital processes are consistently absent in many nondidelphid marsupial groups (e.g., caenolestids, peramelemorphians, macropodoids), but they are present in some dasyuromorphians (e.g., Thylacinus, Sarcophilus) and in a few diprotodontians (e.g., Petaurus breviceps, thylacoleonids). Other aspects of interorbital morphology are too variable among nondidelphid marsupial clades to describe succinctly here, but none appear to provide an unambiguous basis for phylogenetic inference or taxonomic diagnosis.

Dorsolateral Braincase

The right and left frontals and parietals are separated by ontogenetically persistent median sutures in most didelphids (fig. 6), but the midfrontal suture is incomplete or absent in most juveniles and in all examined subadult and adult specimens of Chironectes (fig. 45), Didelphis (fig. 46), Lutreolina (fig. 47), and Philander (fig. 48). All examined juveniles and one young adult specimen (FMNH 84426) of Caluromysiops have complete midfrontal and midparietal sutures, but the right and left frontals and parietals are co-ossified in all of the remaining (fully adult) specimens of Caluromysiops that we examined. Most nondidelphid marsupials have ontogenetically persistent median sutures separating the braincase roofing bones, but the left and right frontals and parietals are co-ossified in all examined subadult and adult caenolestids (Osgood, 1924: figs. 1–3; Patterson and Gallardo, 1987: fig. 1). The midparietal suture is also fused in most examined postjuvenile peramelemorphians.

The scars that mark the dorsalmost origin of the temporalis muscle on each side of the braincase are widely separated, and no sagittal crest is developed in most didelphids. 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. By contrast, much larger sagittal crests that extend anteriorly onto the frontals—an unambiguously different condition—are consistently developed in adult specimens of Caluromysiops, Chironectes, Didelphis, Lutreolina, and Philander. Sagittal crests are altogether absent in caenolestids, Dromiciops, and in most small-bodied Old World marsupials, but well-developed sagittal crests are present in adult specimens of some dasyuromorphians (Dasyurus, Sarcophilus, Thylacinus), some peramelemorphians (Macrotis), and some diprotodontians (e.g., phalangerids, †Wakaleo, †Ekaltadeta).

The taxonomic distribution of parietal-alisphenoid versus frontal-squamosal contact on the lateral braincase of metatherian mammals has often been discussed by authors (e.g., Archer, 1976a; Marshall et al., 1995; Muizon, 1998; Wroe et al., 1998), not all of whom have correctly reported the distribution of these alternative conditions among didelphids (Voss and Jansa, 2003). In fact, parietal-alisphenoid contact is exhibited by all didelphids with the unique exception of Metachirus (fig. 44). Among other marsupials, parietal-alisphenoid contact is also found in caenolestids (Osgood, 1921: pl. 20, fig. 2), most dasyuromorphians (e.g., Dasycercus; Jones, 1949: fig. 8), some fossil peramelemorphians (e.g., Yarala; Muirhead, 2000), and many diprotodontians, whereas squamosal-frontal contact is present in some dasyuromorphians (e.g., Sminthopsis; Archer, 1981: fig. 4), all Recent peramelemorphians, and a few diprotodontians (Wroe et al., 1998). Dromiciops, the only microbiotherian for which the lateral braincase morphology is known, also exhibits squamosal-frontal contact.8

In most didelphids (Chironectes, Didelphis, Glironia, Lutreolina, Marmosa, Metachirus, Monodelphis, Philander, and Tlacuatzin) the petrosal is only exposed on the occiput (behind the lambdoid crest) and on the ventral surface of the skull (between the exoccipital, basioccipital, and alisphenoid; see Wible, 1990: fig. 1). In taxa conforming to this morphology, there are no gaps among the bones that normally form the lateral surface of the braincase (squamosal, parietal, and interparietal; fig. 12A). By contrast, the petrosal capsule that encloses the paraflocculus and the semicircular canals ( =  pars mastoideus or pars canalicularis) is consistently exposed through a fenestra in the suture between the parietal and squamosal in Chacodelphys, Cryptonanus, Gracilinanus, Thylamys, and some species of Marmosops (fig. 12B). Both conditions—presence and absence of a fenestra in the squamosal-parietal suture—occur as balanced polymorphisms (neither state clearly predominating) in Lestodelphys, Marmosops incanus, and M. noctivagus. No nondidelphid marsupial that we examined has a fenestrated squamosal-parietal suture.

Figure 13

Posterior and lateral views of the occipital region in Lestodelphys halli (A, B, UWZM 22422) and Metachirus nudicaudatus (C, D, AMNH 267009) illustrating taxonomic differences in several cranial characters discussed in the text, including alternative patterns of contact among the interparietal (ip), squamosal (sq), parietal (par), and mastoid ( =  pars mastoideus of petrosal, mas). Additionally, the incisura occipitalis (io), a distinct notch in the dorsal margin of the foramen magnum (fm), separates the right and left exoccipitals in Lestodelphys. By contrast, the incisura occipitalis is absent because medial processes of the right and left exoccipitals are in contact, excluding the supraoccipital from the dorsal margin of the foramen magnum in Metachirus. Another conspicuous taxonomic difference illustrated in these views concerns the paroccipital process (pp), which is a small, inconspicuous bony mass adnate to the mastoid in Lestodelphys. The paroccipital process of Metachirus is much larger and projects almost straight ventrally.

Interparietal

Although some authors (e.g., Novacek, 1993) have stated that marsupials lack an interparietal bone, a large interparietal is unequivocally present in Dromiciops (see Giannini et al., 2004: fig. 3), many diprotodontians (e.g., Tarsipes; Parker, 1890: fig. 2), and some dasyuromorphians (e.g., Myrmecobius). In these taxa, the sutures that separate the interparietal from adjacent bones (supraoccipital and parietals) are clearly visible and ontogenetically persistent. In all of the taxa we examined with such distinctly sutured interparietals, the interparietal-supraoccipital boundary coincides closely with the transverse (lambdoid) crest that marks the dorsalmost insertion of the neck extensor musculature.

An interparietal is also unambiguously present in didelphids, all of which exhibit a large, unpaired, wedge-shaped or oblong element between the left and right parietals in the same position (anterior to the lambdoid crest) as the sutured interparietals of other marsupials. The didelphid interparietal, however, is fused with the supraoccipital in all of the skulls we examined (including postweaning juveniles assignable to Gardner's [1973] age class 1), few of which show any trace of a suture.9 Fortunately, developmental studies of didelphid pouch young are available to prove the existence of a distinct interparietal center of ossification. In Didelphis the interparietal first appears around day 8 postpartum (p) and fuses with the supraorbital by day 28p (Nesslinger, 1956), whereas in Monodelphis these events occur on days 3p and 8p, respectively (Clark and Smith, 1993). By contrast, the didelphid interparietal never appears to fuse with the parietals, the sutures between them persisting even in the largest adult specimens we examined.

The didelphid condition (a large interparietal fused to the supraoccipital but not to the parietals) appears to be unique among marsupials. In young dasyurids, for example, the interparietal is small and separated by open sutures from neighboring bones, or it is absent; no examined dasyurid has a large median element fused to the supraoccipital that is suturally distinct from the parietals and wedged between their posterior borders. The interparietal appears to be completely missing in peramelemorphians, none of which show any trace of a bony element wedged between the parietals anterior to the lambdoid crest, even as juveniles. The condition in caenolestids is uninterpretable because no sutures persist on the posterodorsal braincase of any examined specimen.

The didelphid interparietal participates in two alternative patterns of contact among bones of the posterior braincase and the occiput. In most opossums the interparietal does not contact the squamosal because the parietal is in contact with the mastoid (fig. 13A, B). However, in Chironectes, Didelphis, Lutreolina, Philander, and in one examined specimen of Metachirus (fig. 13C, D) the interparietal contacts the squamosal and prevents the parietal from contacting the mastoid. Two species of Monodelphis (M. brevicaudata and M. emiliae) exhibit both conditions with approximately equal frequency among the specimens we examined. Most examined nondidelphid marsupials exhibit parietal-mastoid contact, but acrobatids and Tarsipes exhibit broad interparietal-squamosal contact (Parker, 1890: fig. 2), and phalangerids uniquely exhibit squamosal-exoccipital contact (Flannery et al., 1987: fig. 4C, D).

Figure 14

Palatal morphology of Thylamys venustus (AMNH 261254) illustrating nomenclature for fenestrae, foramina, and other features described in the text. Dental loci (I1–M4) provide convenient landmarks for defining the size and position of palatal structures. Abbreviations: if, incisive foramen; m, maxillary fenestra; mp, maxillopalatine fenestra; p, palatine fenestra; plpf, posterolateral palatal foramen.

Palate

The bony palate is variously perforated by foramina and fenestrae that exhibit considerable variation in occurrence, size, and position among marsupials. Unfortunately, inconsistent terminology has long inhibited clear communication about taxonomic differences. The nomenclature for palatal perforations adopted herein is illustrated in figure 14 and compared with other terminology in table 6.

Figure 15

Ventral midcranial view of Caluromys philander (A, AMNH 267002) and Marmosa demerarae (B, AMNH 266428) illustrating taxonomic differences in palatal morphology. In caluromyines the posterior palate slopes gently ventrally, and the palatal margin is arched (concave posteriorly) without strongly projecting lateral corners; the internal nares (or choanae, in) are very broad. In didelphines, however, the posterior palate is abruptly inflected ventrally and the palatal margin is more or less straight with projecting lateral corners (arrows in lower panel); the internal nares are narrow. Other abbreviations: als, alisphenoid; bs, basisphenoid; max, maxillary; pal, palatine; ps, presphenoid; pt, pterygoid.

Table 6

Terminology for Marsupial Palatal Foramina and Fenestrae

The incisive foramina are prominent slots in the anterior palate that transmit the nasopalatine ducts (Sánchez-Villagra, 2001a) together with nerves and blood vessels that remain to be identified in marsupials (Wible, 2003). In didelphids, these openings extend from the upper incisor arcade posteriorly to or between the canines. Although didelphid incisive foramina are always bordered posteriorly and posterolaterally by the maxillary bones, their dividing septum is formed primarily by the premaxillae. By contrast, the enormously elongated incisive foramina of caenolestids extend posteriorly between the premolar rows (opposite P1 or P2), and the posterior part of the dividing septum is formed by the maxillae (Osgood, 1921: pl. 20). The long incisive foramina of peramelemorphians do not extend behind C1, but the posterior part of their dividing septum is likewise formed by the maxillae. Most other marsupials (e.g., dasyuromorphians and Dromiciops) have incisive foramina that essentially resemble those of didelphids, with the conspicuous exception of some diprotodontians (e.g., Macropus; Wells and Tedford, 1995: fig. 9E) in which these openings are completely contained by the premaxillae.

Among Recent didelphids, only Caluromys and Caluromysiops consistently lack well-formed maxillopalatine fenestrae (figs. 38, 39). Although small perforations in the maxillary-palatine sutures occur in most examined specimens of both genera, these are obviously vascular openings (perhaps homologous with the major palatine foramina of placental mammals) that do not resemble the nonvascular fenestrae of other didelphids. Glironia is sometimes said to lack palatal fenestrae (e.g., by Archer, 1982; Reig et al., 1987), but distinct maxillopalatine openings are present in most adult specimens that we examined (e.g., INPA 2570; fig. 37). When present, the left and right maxillopalatine fenestrae of didelphids are invariably separated by a broad median septum, and they never extend posteriorly behind the molar rows. Maxillopalatine fenestrae are very widely distributed among other marsupials, but they are absent in a few taxa (e.g., Myrmecobius, Dactylopsila, Phascolarctos). Among nondidelphid marsupials that possess these openings, the right and left fenestrae are often confluent because no median septum is developed (as in Rhyncholestes and Perameles gunnii), or they may extend posteriorly well behind the molar rows (as in acrobatids and some burramyids).

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 Cryptonanus, Didelphis, Gracilinanus, Lutreolina, Philander, Thylamys, some species of Marmosa, and some species of Marmosops (e.g., M. creightoni; fig. 52). Presence is also the modal condition for other taxa (e.g., Lestodelphys) in which individuals without palatine perforations are uncommon variants (Martin, 2005). By contrast, palatine fenestrae are consistently absent in Caluromys, Caluromysiops, Chironectes, Glironia, Hyladelphys, Metachirus, Monodelphis, some species of Marmosops (e.g., M. pinheiroi; fig. 53), and most species of Marmosa. The single intact skull of Chacodelphys has palatine fenestrae, but these are incompletely separated from the maxillopalatine openings; the normal morphology for this taxon remains to be determined. Palatine vacuities are less widely distributed than maxillopalatine openings among other marsupials, but they occur in some dasyurids (e.g., Sminthopsis; Archer, 1981: fig. 5A), some peramelemorphians (e.g., Isoodon; Lyne and Mort, 1981: fig. 9), and a few diprotodontians (e.g., Phascolarctos, Vombatus).

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 normally present in Chacodelphys, Gracilinanus, many species of Thylamys (e.g., T. pusillus), and in Tlacuatzin canescens. Most other didelphids lack maxillary vacuities except as rare (usually unilateral) variants. These palatal openings are not known to occur in nondidelphid marsupials.

The posterolateral palatal foramen perforates the maxillary-palatine suture behind M4 or posterolingual to M4 on each side of the skull in most metatherians. According to Archer (1976a), this foramen transmits the minor palatine artery from the maxillary artery to the ventral surface of the palate. In didelphids, caenolestids, Dromiciops, peramelemorphians, and stem metatherians, this foramen is completely surrounded by bone, but the foramen is incomplete or absent in many dasyurids (e.g., Murexia and Sminthopsis) because the maxillary and palatine processes that form the posterior border of this opening in other taxa fail to ossify (Wroe, 1997). In most didelphids the posterolateral palatal foramina are small and located behind the fourth upper molar, but these openings are conspicuously larger and extend lingual to the protocone of M4 on each side of the skull in Thylamys (fig. 54) and most specimens of Lestodelphys (fig. 51).

Two alternative morphologies of the posterior palate can be recognized among didelphids. In Caluromys, Caluromysiops, and 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 choanae are not strongly constricted (fig.15A). By contrast, the posterior palate of most other didelphids 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. 15B). Most nondidelphid marsupials have posterior palates that do not conform strictly to either of these morphotypes, but they more closely resemble those of Caluromys in lacking well-developed lateral corners.

Figure 16

Ventral view of left ear region in Marmosa murina (A, AMNH 267368), Marmosops pinheiroi (B, AMNH 267346), and Monodelphis theresa (C, MVZ 182775) illustrating taxonomic differences in secondary foramen ovale formation. In Marmosa, the mandibular branch of the trigeminal nerve (V³, reconstructed course shown by heavy arrow) emerges from the endocranial lumen via the foramen ovale (fo), which is bordered by the alisphenoid (als) and the petrosal (pet); the extracranial course of the nerve is unenclosed in this taxon. In Marmosops, however, the extracranial course of V³ is partially enclosed by a bony strut (st) that extends from the anteromedial surface of the alisphenoid tympanic process (atp) across the transverse canal foramen (tcf); the nerve then emerges from a so-called secondary foramen ovale. Another kind of secondary enclosure is seen in Monodelphis, where the nerve emerges from a secondary foramen ovale formed by a medial lamina (lam) of the alisphenoid tympanic process. Other abbreviations: bs, basisphenoid; bo, basioccipital; cc, carotid canal; ect, ectotympanic. Scale bars  =  5 mm.

Nasopharyngeal and Mesopterygoid Region

The roof of the nasopharyngeal passageway (or meatus; Rowe et al., 2005) is formed by the vomer, the palatines, and the presphenoid, of which only the palatines and the presphenoid are usually exposed to ventral view. Usually concealed inside the nasopharyngeal passageway, the vomer is deeply divided by the presphenoid into paired lateral processes that only occasionally extend posteriorly into the mesopterygoid fossa to contact the pterygoids (e.g., in Caluromysiops; fig. 39). Most other marsupials, however, exhibit different configurations of the nasopharyngeal roofing bones. In particular, the vomer is undivided and extends caudally to underlie the presphenoid (concealing most or all of it from ventral view) in caenolestids, dasyurids, peramelemorphians, and Dromiciops.

The left and right pterygoid bones are widely separated by the presphenoid and the basisphenoid in didelphids. In other marsupials, the vomer may also extend between the left and right pterygoids, but in Dromiciops the pterygoids are in midline contact, underlying the presphenoid and concealing its suture with the basisphenoid from ventral view; the conjoined pterygoids of Dromiciops form a strong sagittal keel, which is continuous with that formed anteriorly by the vomer and posteriorly by the basisphenoid.10

BASICRANIAL FORAMINA: A foramen that transmits the venous transverse canal perforates the alisphenoid anterolateral to the carotid canal in most didelphids (Sánchez-Villagra, 2001b; Sánchez-Villagra and Wible, 2002). Caluromys and Caluromysiops, however, consistently lack a transverse canal foramen (Archer, 1976a; Sánchez-Villagra and Wible, 2002; Voss and Jansa, 2003). According to Kirsch and Archer (1982: character 24), the “transverse canal” is absent in Metachirus nudicaudatus and Philander opossum, but most of the specimens we examined of both taxa had a transverse canal foramen on each side of the skull. The transverse canal foramen is normally present in most other marsupials (e.g., caenolestids, dasyurids, peramelemorphians, and Dromiciops), although its reduction or absence in a few taxa (e.g., Planigale, Tarsipes) has been noted by authors (Archer, 1976a; Aplin, 1990; Sánchez-Villagra and Wible, 2002).

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. 16A), but the foramen is sometimes contained entirely within the alisphenoid, and both conditions can be seen on opposite sides of the same skull (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 process provide a more accessible basis for taxonomic comparisons. Two different (apparently nonhomologous) conditions of secondary nerve enclosure occur among didelphids.

Figure 17

Oblique ventrolateral view of left ear region in Marmosops impavidus (A, MUSM 13284) and Philander mcilhennyi (B, MUSM 13299) illustrating taxonomic differences in ectotympanic suspension. Whereas the ectotympanic (ect) is suspended from the skull by attachments both to the petrosal (pet) and to the malleus (mal) in Marmosops, the ectotympanic of Philander is suspended only from the malleus (there is no attachment to the petrosal). Other abbreviations: atp, alisphenoid tympanic process; pro, promontorium (of petrosal); rtp, rostral tympanic process (of petrosal); sq, squamosal.

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. 16B) 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.

An alternative pattern of secondary foramen and canal formation is seen in Caluromysiops, Chironectes, Didelphis, Lutreolina, Philander, and some species of Monodelphis (e.g., M. theresa). In these taxa, V³ is broadly enclosed by an alisphenoid lamina (fig. 16C) 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 (Caluromys, Chacodelphys, Cryptonanuns, Glironia, Hyladelphys, Marmosa, Tlacuatzin, and most species of Monodelphis) lack secondary enclosures of the mandibular nerve except as rare (usually unilateral) variants.

Although many nondidelphid marsupials (e.g., caenolestids, Dromiciops, most dasyuromorphians) have only a primary foramen ovale, secondary foramina ovales are also widely distributed. Unfortunately, these traits remain to be widely surveyed, and inconsistent terminology (discussed by Gaudin et al., 1996) makes it difficult to interpret some published descriptions of marsupial basicrania. Although other developmental mechanisms have obviously been responsible for secondary enclosures of the mandibular nerve in some taxa (e.g., Phascolarctos), putative homologues of both didelphid patterns of secondary foramen formation can be recognized among certain Old World marsupials. At least some specimens of Thylacinus, for example, have a secondary foramen formed by an anteromedial bullar strut that spans the transverse canal foramen, whereas the secondary foramen ovale of Recent peramelemorphians (e.g., Echymipera, Perameles) is formed by a broad medial bullar lamina.

EAR REGION: All didelphids have an osteologically well-defined middle ear cavity or hypotympanic sinus (sensu van der Klaauw, 1931: 19). A cup-shaped tympanic process (or “wing”) of the alisphenoid invariably forms the anterior part of the floor of the middle ear, but this structure exhibits significant taxonomic variation in size. The alisphenoid tympanic process is large and extends far enough posteriorly to closely approximate or contact the rostral tympanic process of the petrosal in Caluromys and Caluromysiops (see Reig et al., 1987: fig. 44B, D). In other opossums, a distinct gap separates the alisphenoid tympanic process from the rostral tympanic process of the petrosal, such that at least part of the floor of the middle ear is membranous (e.g., in Marmosa and Monodelphis; Reig et al., 1987: fig. 41B, D). In Lestodelphys and Thylamys a rather indistinct medial process of the ectotympanic makes a small contribution to the floor of the middle ear, partially filling in the gap between the tympanic processes of the alisphenoid and petrosal (Reig et al., 1987: fig. 42B, D). The roof of the hypotympanic sinus in all Recent didelphids is almost exclusively formed by the alisphenoid, with the petrosal making only a small, often negligible, contribution. The didelphid squamosal does not participate in forming any part of the floor or roof of the middle ear cavity.

Caenolestids, dasyurids, peramelemorphians, and Dromiciops resemble didelphids in that the alisphenoid forms most of the anterior part of the floor of the middle ear. However, whereas caenolestids and some peramelemorphians (e.g., Echymipera) have a small alisphenoid tympanic process that does not contact the rostral tympanic process of the petrosal, the alisphenoid tympanic process is large and broadly contacts the petrosal in dasyurids, Dromiciops, and other peramelemorphians (e.g., Perameles). The hypotympanic sinus roof of dasyurids and Dromiciops is almost exclusively formed by the alisphenoid (as in didelphids), but a broad petrosal shelf forms part of the sinus roof in caenolestids and some peramelemorphians. Diprotodontians exhibit a very wide range of bullar morphologies (reviewed by Aplin, 1990), many of which include substantial squamosal participation.

The epitympanic recess of didelphids is a shallow concavity in the ventral surface of the petrosal pars canalicularis that encloses the mallear-incudal articulation (Wible, 2003: fig. 7D).11 Although this morphology is widespread among other marsupials, a few (e.g., Macrotis, Perameles) have an enormously inflated epitympanic recess that vaults high above the ossicles, leaving the mallear-incudal joint effectively unenclosed by bone.

The promontorium of the petrosal pars cochlearis is not marked by any vascular grooves in didelphids. Instead, the most conspicuous feature of this otherwise smooth, bulbous structure is the rostral tympanic process (sensu MacPhee, 1981; Wible, 1990). The rostral tympanic process is small, more or less conical, and perhaps serves only to anchor the posterior (ventral) limb of the ectotympanic annulus in some opossums (e.g., Glironia), but it is anteroposteriorly expanded, dorsally concave, and substantially contributes to the ventral enclosure of the middle ear in others (e.g., Marmosa; Reig et al., 1987: figs. 38–42). In Caluromys and Caluromysiops, the rostral tympanic process extends along the entire length of the promontorium from the fenestra cochleae to the anterior pole (Sánchez-Villagra and Wible, 2002). The rostral tympanic process of caenolestids and most peramelemorphians is small and more or less conical, but dasyurids, some peramelemorphians (e.g., Macrotis), and Dromiciops have an anteroposteriorly extensive and dorsally concave process that forms a substantial part of the bullar floor in these taxa.

In most didelphids, the fenestra cochleae is exposed in lateral or ventrolateral view as illustrated for Didelphis by Wible (1990: fig. 1). In others (e.g., Caluromys, Caluromysiops, Lestodelphys, Marmosa rubra, Monodelphis emiliae, and Thylamys), the fenestra cochleae is concealed within a chamber or sinus formed by a laminar outgrowth of the pars canalicularis that approximates or contacts the 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. The resulting auditory chamber—corresponding to the “periotic hypotympanic sinus” of Archer (1976a) and the “post-promontorial tympanic sinus” of Wible (1990)—is always small and uninflated in opossums.

The fenestra cochleae is also exposed in caenolestids and some peramelemorphians (e.g., Echymipera), in which tympanic processes of the petrosal are either small or absent. In other peramelemorphians (e.g., Perameles) and in dasyurids, however, the fenestra cochleae is concealed within a postpromontorial sinus formed by the seamless fusion of the rostral and caudal tympanic processes; in these forms (as in Caluromys and other didelphids with similar petrosal features), the sinus is small and uninflated. The fenestra cochleae of Dromiciops is enclosed in an inflated (bubblelike) postpromontorial sinus that, although strikingly unlike the smaller sinuses of other plesiomorphic marsupials, is also assumed to have resulted from fusion of caudal and rostral tympanic processes (Sánchez-Villagra and Wible, 2002: fig. 11).

No other bone (besides the alisphenoid, ectotympanic, and petrosal) participates in auditory sinus formation among Recent didelphids, which have only one or two hypotympanic cavities as described above. Caenolestids and Dromiciops are similar to didelphids in this respect, but Recent dasyuromorphians and some peramelemorphians (e.g., Macrotis and Perameles) have a well-developed epitympanic sinus formed by the squamosal. This structure is a cup-shaped cavity that, in all skulls retaining vestiges of auditory soft tissues, appears to be covered by the membrana Shrapnelli (pars flaccida of the tympanic membrane). Archer (1976a: 304) remarked that, “[i]n some didelphids (e.g., Metachirus) there is a depression in the squamosal which is clearly the homologue of this sinus,” but nothing resembling the squamosal epitympanic sinus of dasyurids occurs in any living opossum.

Ectotympanic and Ossicles

Although didelphids were described by van der Klaauw (1931: 26) as having a completely free ectotympanic, two distinct patterns of ectotympanic attachment can be recognized in the family. In most didelphids the anterior limb (or crus) of the ectotympanic 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 seems to be to the petrosal (fig. 17A). In six genera (Caluromys, Caluromysiops, Chironectes, Didelphis, Lutreolina, and Philander), however, the anterior limb of the ectotympanic is not directly attached to the skull, and the suspension is indirect, via the anterior (tympanic) process of the malleus (fig. 17B). In all didelphids with an indirect dorsal connection between the ectotympanic and the skull, the tympanic annulus is more or less ringlike because the posterior (ventral) limb is not expanded to form part of the floor of the middle ear cavity. By contrast, in some taxa with direct ectotympanic suspension, the posterior limb tends to be dorsoventrally flattened and laterally expanded, forming part of the floor of the external ear canal to a greater or lesser extent.12 The anterior and posterior limbs of the didelphid ectotympanic are always widely separated by a broad posterodorsal gap, the incisura tympanica.

Figure 18

Unworn premaxillary dentitions of Marmosops pinheiroi (A, AMNH 267341) and Lutreolina crassicaudata (B, AMNH 210422) illustrating taxonomic differences in the shape of the incisor crowns. In Marmosops and many other small didelphines, the crowns of I2–I4 are symmetrically rhomboidal, with subequal anterior (mesial) and posterior (distal) cutting edges that converge to form a sharp central apex; a distinct posterior corner (distostyle) is always present on I5. By contrast, in Lutreolina (and certain other taxa), the crowns of I2–I4 are conspicuously asymmetrical, with longer anterior than posterior cutting edges; a distinct distostyle is usually absent on I5. Scale bars  =  1 mm.

Caenolestids, dasyurids, most peramelemorphians, and Dromiciops essentially resemble didelphids in most aspects of ectotympanic morphology. The anterior limb is suspended loosely but directly from the basicranium, and it is separated from the posterior limb by a broad posterodorsal incisura tympanica. However, whereas the ectotympanic annulus in didelphids, caenolestids, dasyurids, and peramelemorphians is prominently exposed on the lateral aspect of the bulla, the ectotympanic is uniquely concealed from lateral view within the auditory bulla of Dromiciops (see Segall, 1969a: fig. 3, 1969b: fig. 4). The ectotympanics of Recent diprotodontians differ strikingly from those of didelphids and other marsupials, principally by forming tubelike extensions of the ear canal, by ankylosis or fusion with neighboring bones, and by dorsal migration and narrowing or closure of the incisura tympanica (Aplin, 1990).

The outermost auditory ossicles of didelphids conform to a pattern that is widespread among other plesiomorphic marsupials. In particular, the malleus invariably has a long, sharply inflected neck, a small orbicular apophysis, a well-developed lamina, and a manubrium that is approximately parallel to the anterior (tympanic or gracile) process; the body of the incus is distinctly differentiated from the long and short processes of that bone (Segall, 1969b: figs. 7, 9; Henson, 1974: fig. 20A). Although taxonomic differences in certain details of didelphid mallear and incudal structure were noted by Sánchez-Villagra et al. (2002) and Schmelzle et al. (2005), we did not attempt the requisite dissections to assess the phylogenetic distribution of the features they described. The didelphid stapes, however, exhibits striking morphological variation that can be determined without dissection.

Despite some intraspecific variation noted by Gaudin et al. (1996), most didelphids normally exhibit one or the other of two different stapedial morphotypes defined by Novacek and Wyss (1986). Most didelphids (e.g., Philander; Novacek and Wyss, 1986: fig. 5F) have a more or less triangular or stirrup-shaped, bicrurate stapes that is perforated by a large stapedial (obturator or intercrural) foramen. Other didelphids have a columelliform (or columnar) stapes that is imperforate (or microperforate: with a foramen whose maximum diameter is less than the width of a surrounding crus; Gaudin et al., 1996). Taxa with columelliform stapes include Caluromysiops, Lestodelphys, and several species of Monodelphis (e.g., M. peruviana, M. theresa).

The ossicular morphology of nondidelphid marsupials exhibits significant taxonomic variation. Whereas the mallei of caenolestids, dasyurids, and peramelemorphians essentially resemble those of didelphids in the features described above, the malleus of Dromiciops has a short, uninflected neck, no lamina, and a manubrium that forms an open angle with the anterior process (Segall, 1969a, 1969b). The stapes of Dromiciops is bicrurate (with a large foramen, like didelphids), but the stapes of caenolestids, dasyurids, and peramelemorphians are columelliform and imperforate (Novacek and Wyss, 1986; Schmelzle et al., 2005). Notoryctes and phalangeriform diprotodontians have ossicles that are strikingly unlike those of didelphids and other plesiomorphic marsupials (Segall, 1970; Sánchez-Villagra and Nummela, 2001).

Occiput

Fusion of the supraoccipital with the interparietal is common to all didelphids and occurs very early in postnatal life as has already been discussed, so the dorsal part of the lambdoid crest is not associated with any visible sutures, nor does it develop sesamoids for the insertion of nuchal muscles (as in peramelemorphians, see below). The crest itself is only weakly developed in juveniles and in adult specimens of the smaller forms, but it is prominent in fully adult specimens of all of the larger opossums. The mastoid (pars mastoideus of the petrosal) is always prominently exposed on the lateral occiput, where it is bordered by the supraoccipital, exoccipital, squamosal, and sometimes the parietal. All didelphids have a well-developed mastoid process at the ventrolateral corner of the occiput that is sometimes provided with a sesamoid.

The dorsal margin of the foramen magnum is formed by the exoccipitals and the supraoccipital in most didelphids (fig. 13A). In taxa that conform to this morphology, the foramen has a distinct middorsal emargination, the incisura occipitalis, that separates the left and right exoccipitals. By contrast, 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 the foramen and occluding the incisura (fig. 13C)—in adult specimens of Didelphis, Lutreolina, Metachirus, and Philander. Juvenile (and some subadult) specimens of these genera, however, resemble other didelphids in their occipital morphology, such that both conditions (presence and absence of a supraoccipital margin) can be observed in ontogenetic series of conspecific skulls (Abdala et al., 2001: fig. 3). The medial processes of the exoccipitals are closely approximated in Chironectes, but in all of the specimens we examined the supraoccipital still forms part of the dorsal margin of the foramen magnum (contra Sánchez-Villagra and Wible, 2002), and the incisura occipitalis persists as a narrow middorsal notch.

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. 13A, 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. 13C, D).

Among other plesiomorphic marsupials, dasyurids are essentially similar to didelphids in most occipital features, but mastoid processes are indistinct in caenolestids, and Dromiciops lacks paroccipital processes. Peramelemorphians differ from all other marsupials in having unique osseous structures that are sutured between the parietals and the supraoccipital on each side of the skull. Commonly illustrated but seldom labeled or described by authors, these are the lambdoid sesamoids of Filan (1990), who hypothesized that they function as force multipliers for M. rectus capitis dorsalis and other head extensors. Another occipital feature that might be unique to peramelemorphians is a sesamoid that is probably associated with the origin of the posterior belly of the digastric; this small bone occupies the apex of the paroccipital process in some peramelemorphians (e.g., Perameles), but it attaches to the otherwise smooth ventrolateral corner of the exoccipital in taxa that lack a paroccipital process (e.g., Microperoryctes).

Mandible

The didelphid mandible consists of an anteroposteriorly elongate horizontal ramus that contains the dental alveoli, an ascending ramus with well-developed coronoid and condylar processes, and a posteroventral angular process. The mandibular symphysis is never fused, and the retromolar space (between m4 and the base of the coronoid process) is either imperforate or pierced by tiny nutrient foramina of inconstant number and position. The masseteric fossa (a prominent concavity for the insertion of the eponymous muscle on the posterolateral surface of the jaw) is always imperforate, and its posteroventral border is defined by a distinct shelf. The articular condyle is transversely elongate and more or less semicylindrical. Although substantial variation in mandibular shape is apparent from the drawings that accompany our systematic accounts (figs. 37–Figure 38Figure 39Figure 40Figure 41Figure 42Figure 43Figure 44Figure 45Figure 46Figure 47Figure 48Figure 49Figure 50Figure 51Figure 52Figure 5354), other features are more readily characterized for taxonomic diagnoses and phylogenetic inference.

Most didelphids have two mental foramina that perforate the lateral surface of the horizontal ramus, 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. Chironectes, however, has only a single mental foramen, which is very large and occupies the same position as the anterior foramen in those taxa with two lateral mandibular perforations.

Alternative states of the form and orientation of the marsupial angular process were defined by Sánchez-Villagra and Smith (1997), but we are unable to consistently recognize the distinction between the “rod-like” and “intermediate” conditions they scored for didelphid exemplars. Instead, the common didelphid condition seems more appropriately characterized as “acute and strongly inflected” (Voss and Jansa, 2003: 34). Caluromys and Caluromysiops are the only didelphids with bluntly rounded and weakly inflected mandibular angles.

Most other plesiomorphic marsupials have mandibles that grossly resemble the common didelphid morphotype, but Dromiciops has only one mental foramen and some peramelemorphians (e.g., Perameles gunnii) often have three mental foramina. Additionally, Caenolestes has a small foramen in the retromolar space that Sánchez-Villagra et al. (2000) called the retrodental canal (after Hoffstetter and Villaroel, 1974), and it has a small masseteric foramen near the ventral margin of the masseteric fossa (Osgood, 1921), the lumen of which communicates with that of the much larger mandibular foramen on the medial surface of the jaw.

Dentition

All didelphids normally have 50 teeth that most authors now recognize as including I1–5, C1, P1–3, and M1–4 in the upper dentition and i1–4, c1, p1–3, and m1–4 in the lower dentition. As in other metatherians, only the third upper and lower premolars have deciduous precursors (Flower, 1867; Luckett, 1993; Luo et al., 2004; van Nievelt and Smith, 2005a, 2005b). Despite this prevailing consensus, marsupial tooth homologies have long been controversial, with the result that several alternative systems have been used to identify didelphid dental loci (table 7).

Table 7

Alternative Hypotheses of Marsupial Tooth Homologies

As far as we have been able to determine without dissections or X-rays (i.e., from external inspection of teeth in situ, and from the alveoli of specimens with loose teeth), all didelphids have single-rooted incisors and canines, double-rooted premolars and lower molars, and three-rooted upper molars; apparently, dP3 always has three roots and dp3 has two. Most of the following remarks therefore concern crown morphologies, which afford greater scope for taxonomic comparisons.

Upper Incisors

The didelphid first upper incisor (I1) is a more or less styliform, somewhat procumbent, hypsodont tooth that is conspicuously unlike I2–I5, from which it is always separated by a small diastema. Although taxonomic variation among opossums in I1 hypsodonty and diastema formation has been suggested by authors (e.g., Takahashi, 1974; Archer, 1976b; Creighton, 1984; Wroe et al., 2000), diagnostically useful differences are more easily recognized among the posterior teeth of this series.

The crowns of I2–I5 in most didelphids are approximately symmetrical, labiolingually compressed rhomboids 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. 18A). This is the morphology that Takahashi (1974: 414) termed “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 length (anteroposterior or mesiodistal dimension) from front to back, such that I2 appears visibly smaller than I5 in labial view. This tendency is very pronounced in some taxa with premolariform incisors (e.g., Marmosops, Metachirus) but seems to grade imperceptibly (via intermediate morphologies) to the essentially uniform toothrows of other taxa (e.g., Lestodelphys, Tlacuatzin) in which I2 and I5 are subequal in crown length.

Figure 19

Lateral views of left C1–P3 in Hyladelphys kalinowskii (A, AMNH 267338), Gracilinanus agilis (B, AMNH 134005), and Thylamys pusillus (C, AMNH 275445) illustrating taxonomic differences in the relative heights of P2 and P3 (toothrows are not drawn to the same scale).

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, Hyladelphys, 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. 18B). This is the morphology that Takahashi (1974) labeled “incisiform,” which she recorded for Chironectes in addition to Caluromys and other taxa in her study. However, the unworn juvenile dentitions of Chironectes that we examined 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 crown length from front to back, such that I2 is sometimes visibly larger than I5 in labial view. Aspects of didelphid upper incisor morphology reported by authors whose observations we have not been able to replicate were discussed by Voss and Jansa (2003: 35).

Most peramelemorphians and Dromiciops resemble didelphids in having five upper incisors, but I5 is missing in dasyurids, caenolestids, Echymipera, and Rhynchomeles.13 The first upper incisor is styliform in most of these taxa, but I1 is a laterally compressed, distinctly chisellike tooth in caenolestids, and it is a labiolingually compressed, flat-crowned tooth in peramelemorphians. The unworn posterior incisors have approximately rhomboidal crowns with subequal anterior and posterior cutting edges that converge to a sharp apex in most dasyurids and in Dromiciops, but I2–4 in caenolestids have strikingly asymmetrical crowns (appropriately described as “hatchet-shaped” by Osgood, 1921: 117), the posterior margins of which are deeply notched in Rhyncholestes (see Osgood, 1924: figs. 3b, c; Bublitz, 1987: fig. 13). In peramelemorphians, I2–4 are flat crowned (like I1), but I5 (when present) is caniniform. All of the upper incisors are single rooted in dasyurids, caenolestids, Dromiciops, and many peramelemorphians, but I5 is double rooted in Perameles gunnii.

Upper Canine

The didelphid upper canine (C1) is invariably separated from the posteriormost upper incisor by a wide diastema that may or may not contain a distinct paracanine fossa for the reception of the lower canine. In most didelphids, C1 occupies the premaxillary-maxillary suture, despite the fact that the maxilla always extends anterolaterally beyond this tooth, sometimes reaching the alveolus of I5; therefore, the premaxilla only contacts the C1 alveolus medially (fig. 14). In Caluromys and Caluromysiops, however, the upper canine alveolus is entirely contained within the maxilla (figs. 38, 39).

The upper canine is a simple unicuspid tooth in most didelphids, but some species (e.g., Gracilinanus emiliae; Voss et al., 2001: fig. 12A) have a small posterior accessory cusp, and others (e.g., Marmosops pinheiroi; fig. 53) have both anterior and posterior accessory cusps. Some care is needed in determining canine crown morphology, however, because accessory cusps are sometimes obliterated by wear (as inferred from their absence in old adults of species that uniformly exhibit such structures as juveniles and subadults). Alternatively, a false posterior accessory cusp is sometimes formed when the trailing edge of C1 is notched by occlusion with p1. Unlike other teeth whose size does not increase ontogenetically after the crown is fully erupted, didelphid canines often appear much larger in older adults than in younger animals because the root is continuously extruded from the alveolus throughout life.

Most other marsupials have single-rooted unicuspid upper canines like those of didelphids, but accessory cusps are present in some peramelids (e.g., Echymipera) and dasyurids (e.g., Antechinomys; Archer, 1976b). Additionally, the canine is double rooted and provided with accessory cusps in female Caenolestes and Rhyncholestes and in both sexes of Lestoros, whereas male Caenolestes and Rhyncholestes have single-rooted unicuspid teeth (Osgood, 1924; Bublitz, 1987). The position of C1 with respect to the maxillary-premaxillary suture has not been widely surveyed in Marsupialia, but the upper canine alveolus is entirely contained by the maxilla in Dromiciops, some peramelemorphians (e.g., Echymipera, Perameles), and some dasyuromorphians (Murexia, Sminthopsis). In several diprotodontian clades, C1 is absent.

Upper Premolars

The didelphid first upper premolar (P1) is usually in contact with or very close to C1, but a diastema is often present between P1 and P2. The gap between P1 and P2 tends to be individually and ontogenetically variable, however, and a continuous range of intermediates exist between taxa in which it is usually absent (e.g., Lestodelphys; fig. 51) and those in which it is normally present (e.g., Philander; fig. 48). By contrast, no diastema is ever present between P2 and P3, the crowns of which are in contact or closely approximated in all examined opossum dentitions.

Although consistently smaller than the posterior premolars, P1 in most didelphids is at least half the height or width of P2, which it essentially resembles in having a large, laterally compressed, more or less triangular central cusp that is usually flanked by smaller anterior and posterior accessory cusps. By contrast, P1 is a much smaller (apparently vestigial) tooth that lacks distinct occlusal features in Caluromys (fig. 38) and Caluromysiops (fig. 39). In a few specimens of the latter genus (e.g., AMNH 244364), P1 is missing.

The second and third upper premolars are always large and well-developed teeth. Each is surrounded by a broad basal cingulum in Caluromys and Caluromysiops, but the basal cingulum of P2 and/or P3 is often incomplete or indistinct in other didelphids. The second upper premolar is distinctly taller than P3 in Caluromys, Caluromysiops, and Hyladelphys (fig. 19A), but P3 is distinctly taller than P2 in Chironectes, Cryptonanus, Didelphis, Lestodelphys, Lutreolina, Monodelphis, Philander, and Thylamys (fig. 19C). By contrast, P2 and P3 are about the same height in Chacodelphys,14 Glironia, Gracilinanus (fig. 19B), Marmosa, Marmosops, and Metachirus.

Figure 20

Occlusal views of left upper and right lower didelphid molars illustrating features of crown morphology discussed in the text. Abbreviations: acid, anterior cingulid; alci, anterolabial cingulum; ecf, ectoflexus; encd, entoconid; encrd, entocristid; hycd, hypoconid; hycld, hypoconulid; hycld.n., hypoconulid notch; mec, metacone; mecd, metaconid; mecrd, metacristid; pac, paracone; pacd, paraconid; pacrd, paracristid; pomecr, postmetacrista; popacr, postparacrista; poprcr, postprotocrista; prc, protocone; prcd, protoconid; prmecr, premetacrista; prpacr, preparacrista; prprcr, preprotocrista; stB, stylar cusp B; stD, stylar cusp D.

When unworn, P3 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, Glironia, and Hyladelphys. In all other didelphids, 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 with single-bladed P3s.

Most other plesiomorphic marsupials have three upper premolars, but some dasyurids (e.g., Dasyurus) have only two. In addition to the usual diastema between P1 and P2, some adult dasyurids (e.g., Murexia) and all adult peramelemorphians have a diastema between P2 and P3. The third upper premolar (P3) is taller than P2 in caenolestids, Dromiciops, and some dasyurids (e.g., Murexia, Sminthopsis), but P2 is taller than P3 in other dasyurids (e.g., Myoictis). An anterior cutting edge is apparently absent from P3 in all nondidelphid marsupials.

Upper Milk Premolar

Although the deciduous upper premolar (dP3) of didelphids has consistently been described as large and molariform (Flower, 1867; Thomas, 1888; Bensley, 1903; Tate, 1948b; Archer, 1976b), there is noteworthy taxonomic variation in the morphology of this tooth. In most opossums dP3 is, indeed, a sizable tooth: its crown area ranges from about 42% to 96% of the crown area of M1, the tooth immediately behind it (Voss et al., 2001: table 5). An obviously functional element of the upper toothrow, dP3 occludes with both the deciduous lower premolar (dp3) and with m1. Hyladelphys, however, has a very small upper milk premolar (<10% of the crown area of M1; op. cit.) that appears to be functionally vestigial because it does not occlude with any lower tooth.

Although molariform, the didelphid upper milk premolar does not always match the teeth behind it in all occlusal details. In most specimens, the paracone of dP3 is located on the labial margin of the crown, and the stylar shelf is correspondingly incomplete, whereas the paracone is lingual to a continuous stylar shelf on all didelphid molars (see below). Among those didelphids whose milk premolars we examined, only Caluromys has a dP3 in which the paracone is usually lingual to a continuous stylar shelf.

The morphology of milk premolars remains to be widely surveyed in Marsupialia, but of those taxa that we examined, only Dromiciops has a large, molariform dP3 resembling the common didelphid condition (Marshall, 1982b: fig. 17; Hershkovitz, 1999: fig. 32). By contrast, dP3 is much smaller, more or less vestigial, and structurally simplified in caenolestids, dasyurids, and peramelemorphians (Tate, 1948a; Archer, 1976b; Luckett and Hong, 2000).

Upper Molars

Didelphid upper molars conform to the basic tribosphenic bauplan (Simpson, 1936) in having three principal cusps—paracone, protocone, and metacone—connected by the usual crests in a more or less triangular array (fig. 20). A broad stylar shelf and an anterolabial cingulum are invariably present; the centrocrista (postparacrista + premetacrista) and the ectoloph (preparacrista + centrocrista + postmetacrista) are uninterupted by gaps; the para- and metaconules are indistinct or absent;15 and there is no posterolingual talon. In addition, most didelphids have several (usually five or six) small cusps on the stylar shelf, for which most authors employ alphabetical labels (after Bensley, 1906; Simpson, 1929; Clemens, 1966). Of these, stylar cusp B (labial to the paracone) is more consistently recognizable than the others, but a stylar cusp in the D position (labial to the metacone) is often subequal to it in size.

Figure 21

Occlusal views of left upper molar rows illustrating taxonomic differences in relative widths of teeth (not drawn to the same scale). A, Caluromysiops irrupta (FMNH 84426); B, Caluromys philander (AMNH 267334); C, Glironia venusta (AMNH 71395); D, Marmosa regina (MVZ 190333); E, Marmosops impavidus (MUSM 13284); F, Lestodelphys halli (MMNH 15708).

Didelphids differ conspicuously in the relative width (transverse or labial-lingual dimension) of successive molars within toothrows. In some taxa, the anterior molars tend to be wide in proportion to more posterior teeth, but in others the posterior molars are relatively wider (fig. 21). The ratio obtained by dividing the width of M4 by the width of M1 (M4/M1; table 8) conveniently indexes this size-independent shape variation and ranges from a minimal value of 0.83 (in Caluromysiops) to a maximal value of 1.76 (in Lutreolina). Morphometric comparisons of dental measurements (Voss and Jansa, 2003: fig. 12) and visual inspections of toothrows suggest that this index is correlated with a pattern of molar occlusal transformation that has 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). These tendencies are redundantly described by numerous dental ratios that have been coded as independent characters in previous phylogenetic studies (for an extended discussion, see Voss and Jansa, 2003).

Figure 22

Lingual views of lower incisors. Top, Metachirus nudicaudatus (AMNH 266452) with distinct posterior accessory cusps on i1–i4. Bottom, Lutreolina crassicaudata (AMNH 210424) without distinct posterior accessory cusps on i1–i4.

Table 8

Upper Molar Measurements and Proportions^a

In effect, didelphids with relatively wide anterior molars have dentitions that can be described as not or only weakly carnassialized, whereas those with relatively wide posterior molars have strongly carnassialized teeth. Although the endpoints of this morphocline are strikingly different, the essentially continuous distribution of intermediate morphologies renders any attempt to make qualitative distinctions based on molar proportions an arbitrary exercise. We therefore employ “not carnassialized,” “weakly carnassialized,” “moderately carnassialized,” and “strongly carnassialized” as heuristic descriptors of a visually compelling but phylogenetically intractable aspect of didelphid dental variation.16

The ectoflexus is a V-shaped labial indentation of the stylar shelf that is either present or absent on tribosphenic marsupial molars. Among didelphids, Caluromys and Caluromysiops are the only taxa that lack any trace of an ectoflexus (e.g., figs. 21A, B). In most other opossums, a distinct ectoflexus is present on M3, on M2 and M3, or (rarely) on M1–M3. When ectoflexi are present on multiple teeth, they invariably increase in depth from anterior to posterior.

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 ∧-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 (e.g., Didelphis; see Reig et al., 1987 [character 1]; Wroe et al., 2000 [character 10]; Wible et al., 2001 [character 31]). Our observations agree with Johanson's (1996), 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 didelphids, 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 ∧-shaped—in Chacodelphys, Cryptonanus, Gracilinanus, Lestodelphys, Marmosa, Marmosops, Metachirus, Monodelphis, Thylamys, and Tlacuatzin. 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 seven remaining genera (Caluromys, Chironectes, Didelphis, Glironia, Hyladelphys, Lutreolina, Philander) in which the centrocrista has a weak labial inflection with a slightly elevated apex; these taxa can appropriately be described as weakly dilambdodont.

In many didelphids—Caluromys, Caluromysiops, Glironia, Gracilinanus, Hyladelphys, Marmosa, Tlacuatzin, and some species of Cryptonanus and Marmosops—the preprotocrista passes labially around the base of the paracone to join with the anterolabial cingulum. This results in the formation of a continuous shelf along the anterior margin of the tooth (fig. 21A–D), and taxa possessing this feature are sometimes said to have a “complete” anterior cingulum (Archer, 1976b: 3) or to exhibit “double rank prevallum-postvallid shearing” (Cifelli, 1993: 213). In the alternative morphology—exhibited by Chacodelphys, Chironectes, Didelphis, Lestodelphys, Lutreolina, Metachirus, Monodelphis, Philander, Thylamys, and other species of Cryptonanus and Marmosops—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. 21E, F). Voss and Jansa (2003: 39) discussed several inconsistencies in the published literature concerning the distribution of these traits among didelphids.

The postprotocrista also exhibits noteworthy variation among didelphids. In most genera, this crest decreases in height and width as it passes posterolabially around the base of the metacone before merging with the posterior surface of the tooth crown, leaving a distinct groove or gap in the posterior wall of the trigon basin. By contrast, the postprotocrista connects directly with the base of the metacone, such that the posterior wall of the trigon basin has no gap or groove, in Chironectes, Didelphis, Lutreolina, and Philander; instead, the unworn postprotocrista of these taxa exhibits a distinct carnassial notch near the base of the metacone.

Dasyurids and Dromiciops have tribosphenic upper molars that differ from those of didelphids in only minor details. Indeed, some dasyurids and didelphids are indistinguishable in published matrices of upper molar characters (e.g., Archer, 1976b: table 1), although there is a tendency for stylar cusp D to be larger than stylar cusp B when both structures are present in the former group (Wroe, 1997), and a few dasyurids (e.g., Sarcophilus) are conspicuously divergent in other upper molar traits. The upper molars of Dromiciops differ from those of most didelphids by having a linear centrocrista, indistinct stylar cusps, and a stylar shelf that is conspicuously reduced in width labial to the paracone on M1.17

The upper molars of Recent peramelemorphians differ from those of didelphids in several respects, most notably by having a discontinuous centrocrista: instead of forming a ∧-shaped (labially inflected) crest with an apex that is lingual to the stylar shelf, the postparacrista and the premetacrista of modern peramelemorphians terminate on the labial margin of the tooth, where they are separated by a small gap (Muirhead and Filan [1995] referred to this trait as the result of the centrocrista “breaching” the ectoloph). Additionally, all modern peramelemophians have a large posterolingual cusp on M1–M3; in most genera (e.g., Echymipera, Perameles) this is a hypertrophied metaconule, but in Macrotis it is the metacone (Archer, 1976b). Another peramelemorphian trait is that the preparacrista of M1 is a tall crest that passes posterolabially to a large stylar cusp at or near the C position (Archer, 1976b). The anterolabial cingulum is present in some peramelemorphians (e.g., Echymipera) but not in others (e.g., Perameles).

On caenolestid molars the principal labial cusps are hypertrophied stylar elements in the B and D positions (Osgood, 1921; Marshall, 1987; Goin and Candela, 2004; Goin et al., 2007). In both Caenolestes and Rhyncholestes, the paracone is absent (or indistinguishably fused with stylar cusp B). The metacone is also absent in Rhyncholestes but, although fused to the lingual aspect of stylar cusp D, the metacone is still clearly recognizable in Caenolestes. Given this interpretation of caenolestid cusp homologies, it follows that a low crest along the labial base of the tooth crowns in both genera is a neomorphic cingulum rather than a vestigial stylar shelf. The first two caenolestid upper molars also have a well-developed posterolingual cusp that has been interpreted either as a hypertrophied metaconule (Marshall, 1987; Goin and Candela, 2004; Goin et al., 2007) or as a neomorphic outgrowth of the postprotocingulum (Hunter and Jernvall, 1995). Another distinctive trait of the upper molars of Recent caenolestids is the complete absence of an anterolabial cingulum.

Upper Dental Eruption Sequences

As described by Tribe (1990), P3 is the last upper tooth to erupt in most didelphids. By contrast, P3 erupts before M4, which is the last upper tooth to emerge in Chironectes, Didelphis, Lutreolina, and Philander. In a few didelphids (e.g., Metachirus nudicaudatus, Monodelphis brevicaudata), M4 and P3 appear to erupt simultaneously. Molar eruption sequences have not been widely surveyed among other marsupials, but P3 seems to be the last upper tooth to erupt in most nondidelphid clades, with the exception of Dromiciops and caenolestids, in which M4 is usually last (Hershkovitz, 1999; Luckett and Hong, 2000).

Lower Incisors

The four lower incisors of didelphids are subequal in size and essentially similar in coronal shape, but the alveolus of i2 is always “staggered” (displaced lingually and dorsally relative to the alveoli of i1, i3, and i4), and the root of this tooth is correspondingly supported by a prominent labial buttress of alveolar bone (Hershkovitz, 1982, 1995). The unworn lower incisor crowns of most didelphids have a distinct lingual cusp or heel (fig. 22, top) that is indistinct or absent in Chironectes, Didelphis, Lutreolina (fig. 22, 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. No didelphid has a lobed or indented cutting edge on any of the lower incisor crowns.

Figure 23

Labial views of unworn anterior mandibular dentitions illustrating taxonomic differences in lower canine (c1) morphology. A, Lestodelphys halli (MMNH 17171) with erect, acutely pointed, and simple c1. B, Marmosa mexicana (AMNH 17136) with semiprocumbent and flat-crowned but simple c1. C, Marmosops pinheiroi (AMNH 267342) with fully premolariform c1 (procumbent and flat crowned with posterior accessory cusp).

Among other marsupials, Dromiciops resembles didelphids in having four lower incisors, whereas dasyurids and peramelemorphians have only three. The missing tooth in dasyurids and peramelemorphians is presumably i4, because there is only one tooth behind staggered-and-buttressed i2, and because it seems more likely that teeth should be lost from the end of the incisor series than from intermediate loci. In all of these polyprotodont groups, the lower incisors are short-crowned nonprocumbent teeth that differ from those of didelphids in only minor details.

The unworn lower incisors of most dasyurids (e.g., Murexia, Sminthopsis) have a distinct lingual cusp or heel, but this structure is lacking in at least some of the larger carnivorous forms (e.g., Dasyurus). A distinctive feature of unworn i3 in dasyurids is the consistent presence of a small posterior lobe on the cutting edge of the tooth.18 As implied by Bensley (1903: 113), this structure appears to be homologous with the much larger posterior lobe of i3 seen in peramelemorphians. All peramelemorphian lower incisors lack a lingual cusp or heel, but i1 and i2 are otherwise unremarkable in this group. The lower incisor dentition of Dromiciops is similar to that of didelphids except for the fact that i2 is neither staggered nor buttressed in microbiotherians (Hershkovitz, 1999: figs. 5A, 29E).

The anteriormost lower incisor of caenolestids is a long-crowned procumbent (gliriform) tooth that most authors (e.g., Ride, 1962; Marshall, 1980; Sánchez-Villagra, 2001b) assume to be either i1 or i2. Whichever it is, at least two of the four single-rooted unicuspids that normally occur between the gliriform tooth and the anteriormost undisputed premolar must also be incisors.19 None of the lower incisors of Caenolestes or Rhyncholestes has a staggered alveolus.

Lower Canine

The lower canine is an erect unicuspid tooth that is conspicuously differentiated from the incisors and premolars in all of the larger didelphids (Caluromys, Caluromysiops, Chironectes, Didelphis, Lutreolina, Philander), and in some of the smaller forms as well (Glironia, Hyladelphys, Lestodelphys, Monodelphis, Tlacuatzin). In all of these taxa, the unworn crown of c1 tapers to a sharp apical point (fig. 23A). By contrast, the lower canine is procumbent and tends to become premolariform in other small didelphids. Premolarization of c1 includes the development of a flattened anterior blade (rather than a sharp tip) and a posterior cingulum that sometimes produces a distinct accessory cusp. Whereas these tendencies are minimally indicated on the unworn teeth of Chacodelphys and Metachirus (which are flattened along their anterior margin but have no posterior cingulum), they are increasingly apparent in Cryptonanus, Gracilinanus, Marmosa (fig. 23B), and Thylamys. Premolarization of the lower canine is maximally developed in the smaller species of Marmosops (e.g., M. parvidens and M. pinheiroi), where c1 and p1–3 form a series of four strikingly similar teeth (fig. 23C). Generally speaking, taxa with well-developed accessory cusps on the upper canine also have a well-developed posterior accessory cusp on c1, so variation in these occluding teeth is at least partially correlated.

Figure 24

Occlusal views of right dp3 and m1 illustrating taxonomic differences in trigonid morphology of the deciduous tooth. Top, Marmosops impavidus (MUSM 13286) with a complete (tricuspid) dp3 trigonid. Bottom, Marmosa murina (MUSM 15297) with an incomplete (bicuspid) dp3 trigonid.

Large dasyurids (e.g., Dasyurus) have erect, unicuspid lower canines, but many small dasyurids (e.g., Sminthopsis) have procumbent, premolariform teeth. The lower canines are small and blunt in peramelemorphians. In Dromiciops, c1 is procumbent and more or less premolariform (Marshall, 1982b: fig. 14).

Lower Premolars

Didelphid lower premolars are unremarkable teeth, each having a single dominant cusp and a well-developed distal cingulum that often produces a posterior accessory cusp. Small diastemata are sometimes present between p1 and c1, between p1 and p2, and (rarely) between p2 and p3. The lower first premolar (p1) is vestigial and occlusally featureless in Caluromys and Caluromysiops, but it is well developed in all of the other genera.

The second lower premolar is distinctly taller than p3 in most didelphids, but p2 and p3 are subequal in Marmosops incanus and most examined species of Monodelphis and Thylamys. In Lestodelphys and in certain other species of Monodelphis (e.g., M. emiliae), however, p3 is distinctly taller than p2.

Most other plesiomorphic marsupials have three lower premolars that are essentially similar to those of didelphids, although some dasyurids (e.g., Dasyurus) have only two. Among the taxa we examined, p3 is taller than p2 in caenolestids, Dromiciops (see Marshall, 1982b: fig. 14), some dasyurids (e.g., Murexia), and some peramelemorphians (e.g., Echymipera), but p2 and p3 are subequal in height or p2 is the taller tooth in other dasyurids (e.g., Sminthopsis) and peramelemorphians (e.g., Macrotis).

Lower Milk Premolar

Taxonomically significant variation among didelphids in the size and shape of dp3 was reported by Voss et al. (2001: table 5). Many didelphids (Caluromys lanatus, Caluromysiops, Chironectes, Cryptonanus chacoensis, 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. 24, 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 Cryptonanus unduaviensis, Gracilinanus agilis, Lestodelphys, Marmosa, and Thylamys, which have incomplete, blade-like trigonids bearing only one or two distinct cusps (fig. 24, bottom). Although just the protoconid of dp3 is usually distinct in our material of Caluromys philander, the trigonid of this taxon is triangular in occlusal outline and more closely resembles the fully molariform condition than the blade-like uni- or bicuspid alternative. Among other plesiomorphic marsupials that we examined, only Dromiciops has a large and fully molariform dp3, whereas dasyurids and peramelemorphians have small, structurally simplified and obviously vestigial lower milk premolars.

Figure 25

Lingual views of right m2 and m3 illustrating taxonomic differences in entoconid size. Top, Thylamys pallidior (AMNH 262405) with a large entoconid (arrow). Bottom, Monodelphis peruviana (AMNH 272781) with an indistinct entoconid (arrow).

Lower Molars

Didelphid lower molars conform to the usual tribosphenic pattern and are readily described using standard nomenclature (fig. 20). The trigonid moiety of each tooth invariably consists of three distinct cusps (paraconid, protoconid, and metaconid) that are connected by two tall and deeply notched crests (paracristid and metacristid); three additional cusps (hypoconid, entoconid, and hypoconulid) surround the talonid basin. A well-developed anterior cingulid (precingulid) is always present, but a posterior cingulid (postcingulid) is just as consistently absent.20 The anterior cingulid never extends all the way to the base of the paraconid because it abuts on the hypoconulid of the preceding tooth (dp3 in the case of m1), resulting in a prominent concavity in the anterolingual occlusal outline (the “hypoconulid notch” of Archer, 1976b). The entocristid is a short crest that approximately parallels the lingual margin of the tooth.

Didelphid lower molars exhibit taxonomic shape variation that is correlated with carnassialization as previously described for the upper teeth. Thus, taxa with uncarnassialized molars have relatively small trigonids and large talonids, whereas opossums with highly carnassialized molars have relatively large trigonids and small talonids. Trigonid enlargement is invariably accompanied by an increase in occlusal relief, the protoconid and metacristid in particular tending to become very tall in the most carnassialized forms (e.g., Lestodelphys and Lutreolina). The width gradient previously noted for the upper molars is also seen in the lowers, m2 being the widest tooth in uncarnassialized dentitions whereas m3 or m4 is widest in progressively more carnassialized forms. As for the upper teeth, the highly divergent endpoints of this conspicuous gradient of dental shape variation are connected by an almost continuous taxonomic series of morphological intermediates.

Only a few lower molar traits that vary among didelphids seem useful for the purposes of taxon diagnoses and phylogenetic analysis. Among these, the labial prominence of the hypoconid tends to vary along the toothrow, so we standardized comparisons at the third molar locus. In most opossums, the hypoconid is labially salient (projecting beyond the protoconid or level with it) on m3, but the hypoconid is lingual to the protoconid in Chacodelphys, Lestodelphys, Lutreolina, and Monodelphis.

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. 25, top). In Chacodelphys and Monodelphis, however, the entoconid is very small, never more than subequal to the hypoconulid, and often smaller than that cusp (fig. 25, bottom); it becomes indistinct or is obliterated by wear in most adult specimens. Discrepant observations in the literature about didelphid entoconid size variation were discussed by Voss and Jansa (2003).

Figure 26

Scatterplot of percent GC content among didelphid sequences at first and second codon positions versus percent GC content at third codon positions for the five genes included in this study.

In most didelphids, the hypoconulid is much closer to the entoconid than it is to the hypoconid, and is said to be “twinned” with the former cusp. Although twinning of the hypoconulid and entoconid is often stated (or implied) to be universally present among tribosphenic marsupials (e.g., by Clemens, 1979; Kielan-Jaworowska et al., 2004), it is not. The hypoconulid and entoconid are not closely approximated in Caluromysiops, and in some specimens (e.g., FMNH 121522) the hypoconulid is at the center of the posterior talonid margin, equidistant to the hypoconid and the entoconid.

Most other plesiomorphic marsupials have lower molars that resemble those of didelphids in common tribosphenic features, but several differences are noteworthy. Dasyurids, for example, have a well-developed posterior cingulid on m1–m3, the anterior cingulid of peramelemorphian lower molars lacks a hypoconulid notch, the anterior cingulid is vestigial in Dromiciops (which lacks a hypoconulid notch on m1), and an entocristid is absent in some dasyurids (e.g., Sminthopsis) and peramelemorphians (e.g., Perameles).

As they are in so many other character complexes, caenolestids are the most divergent basal marsupials in lower molar morphology. In both Caenolestes and Rhyncholestes the paracristid is unnotched and decreases gradually in height from the apex of the protoconid to the anterolingual corner of the tooth; the paraconid is therefore indistinct. Whereas the paracristid runs in an uninterrupted arc from the protoconid to the anterolingual corner of the tooth on m1, the paracristid on m2 and m3 abuts the hypoconulid of the preceding tooth (m1 and m2, respectively), at which point it is abruptly deflected posterolingually. The only distinct trigonid cusp on m4 is the metaconid, from which a diminutive crest (comprising a vestigial metacristid and part of the paracristid) curves anterolabially to abut the hypoconulid of m3. The entocristid on m1–m3 is a long crest that is deflected labially, away from the lingual margin of the tooth, and terminates deep inside the talonid basin.

Summary

The morphological observations described above provide evidence of taxonomic differences in many phenotypic attributes suitable for phylogenetic analysis, including unambiguously distinguishable aspects of pelage color and structure; sexually dimorphic skin glands and carpal tubercles; manual and pedal digital proportions; the presence, absence, and morphology of cutaneous shelters for nursing young; caudal modifications for prehensility and fat storage; the sizes and shapes of cranial bones and bony processes; the presence, absence, and ontogenetic persistence of cranial sutures; the number and occurrence of cranial foramina and fenestrae; auditory morphology; and the sizes, shapes and eruption sequences of teeth. Because the relevant literature on mammalian functional morphology is too extensive to review effectively in this report, we can only assert that such taxonomic differences are likely to reflect adaptations for such diverse functions as concealment, locomotion, reproduction, sense perception, and food reduction. Anatomical trait variation (together with published information about karyotypic differences that are not reviewed above) is formally described as qualitative characters in appendix 3, where we also explain the criteria used to score individual specimens and discuss the justification for ordering multistate transformation series. The resulting data are summarized in appendix 4.

Breast Cancer Activating 1 Gene

The human BRCA1 gene consists of 22 coding exons distributed over ca. 100 kb of genomic DNA on the long arm of chromosome 17 (Miki et al., 1994). The gene is expressed in numerous tissues, but it is especially active in epithelial cells undergoing high levels of proliferation and differentiation (Miki et al., 1994; Casey, 1997). BRCA1 is a tumor suppressor gene whose unmutated product is a large (1863 amino acids) nuclear-cytoplasmic-shuttling phosphoprotein of unknown structure but with apparently essential roles in transcriptional regulation and DNA repair, among other intracellular functions (Rodríguez and Henderson, 2000; Dimitrov et al., 2001; Venkitaraman, 2001).

Nucleotide sequence data from BRCA1 exon 11 have been analyzed in several mammalian phylogenetic studies, usually in concatenated datasets with other genes (e.g., Adkins et al., 2001; Madsen et al., 2001; Delsuc et al., 2002; Douady et al., 2002; Raterman et al., 2006). Homology comparisons of the 2163 bp fragment that we amplified and sequenced for this study are consistent with the presence of a single full-length copy in Monodelphis domestica, where the gene appears to be located on chromosome 2. BRCA1 sequence data are available from 48 terminals in our analyses, including 41 didelphids (Caluromysiops and Lestodelphys were not sequenced) and all 7 nondidelphid outgroups. All of the marsupial sequences that we obtained and others that we downloaded from GenBank translate to open reading frame. Twenty independent insertion-deletion events (indels) ranging in length from 3 to 39 bp were required to align didelphid BRCA1 sequences; of these, 9 were unique to single taxa and 11 were shared by two or more taxa.

Dentin Matrix Protein 1

The human DMP1 gene (originally known as AG1; George et al., 1993) consists of six exons spanning about 14 kb of genomic DNA on the long arm of chromosome 4 (Hirst et al., 1997). DMP1 is an acidic calcium-binding secreted phosphoprotein of about 513 amino acids, most of which are encoded by exon 6 (Hirst et al., 1997). The protein is assumed to have some role in tissue mineralization based on its chemical properties and presence in the extracellular matrix of dentine and bone (Butler and Ritchie, 1995; Feng et al., 2003). However, mRNA transcripts and the protein itself have also been detected in a wide range of soft tissues, where the gene presumably has other functions (Terasawa et al., 2004).

Nucleotide sequence data from DMP1 exon 6 have been analyzed in several previous studies of mammalian intraordinal relationships (e.g., Van Den Bussche et al., 2003; Reeder and Bradley, 2004; Jansa et al., 2006). Homology comparisons of the 1176 bp fragment that we amplified and sequenced for this study are consistent with the presence of a single full-length copy in Monodelphis domestica, where the gene appears to be located on chromosome 5. We obtained DMP1 sequences from 42 didelphids (none is available from Caluromysiops). Fourteen indels ranging in length from 3 to 18 bp are required to align didelphid DMP1 sequences, including 6 that are unique to single taxa and 8 that are shared by two or more taxa. Unfortunately, didelphid DMP1 sequences cannot be aligned unambiguously with sequences from nondidelphid marsupials, so no outgroups are available to root phylogenetic analyses of this gene (Jansa and Voss, 2005; Jansa et al., 2006).

Interphotoreceptor Retinoid Binding Protein

The human IRBP gene (also known as the Retinol Binding Protein 3 gene, RBP3) consists of four exons spanning about 9.5 kb of genomic DNA on the long arm of chromosome 10 (Liou et al., 1989; Fong et al., 1990; Pepperberg et al., 1993). The gene product is a large (1230 amino acids) secreted glycolipoprotein that is primarily found in the extracellular matrix between the retinal pigment epithelium (RPE) and the neural retina of the eye; the protein apparently functions as a two-way carrier of retinoid between the rod photoreceptors and the RPE, and therefore plays a critical role in visual pigment regeneration (Pepperberg et al., 1993).

Nucleotide sequences from IRBP exon 1 were among the first molecular character data to provide convincing support for mammalian superordinal relationships (Stanhope et al., 1992, 1996; Springer et al., 1997b), and they have also been analyzed in several subsequent studies focused on lower-level relationships (e.g., Jansa and Voss, 2000; Mercer and Roth, 2003; Weksler, 2003; Jansa and Weksler, 2004). The 1158 bp fragment that we amplified and sequenced from didelphids is homologous with a single full-length copy in Monodelphis domestica, where the gene is apparently located on chromosome 1. IRBP sequences are available from 43 didelphid terminal taxa (including Caluromysiops and Lestodelphys) and 7 nondidelphid outgroups. No indels are required to align didelphid IRBP sequences, all of which translate to open reading frame (Jansa and Voss, 2000, 2005; Voss and Jansa, 2003).

Recombination Activating 1 Gene

The human RAG1 gene contains a single uninterrupted exon of 3.1 kb located on the short arm of chromosome 11 adjacent to the RAG2 locus; the gene product is a large DNA-binding protein of 1043 amino acids (Schatz et al., 1989; Oettinger et al., 1990, 1992). Recombination activating genes are expressed only in the nucleus of developing B and T lymphocytes, where the RAG1 and RAG2 proteins act synergistically to make double-stranded breaks at specific recognition sequences, initiating recombination of variable (V), diversity (D), and joining (J) gene segments; completion of the V(D)J rearrangement process yields functional immunoglobulin and T-cell receptor proteins, an essential step in the development of a mature immune system (Gellert, 2002).

Nucleotide sequence data from RAG1 have previously been analyzed in several vertebrate phylogenetic studies (e.g., Groth and Barrowclough, 1999; Barker et al., 2002; Baker et al., 2004; Gruber et al., 2007). The 2790 bp fragment that we amplified and sequenced for this study is homologous with a single full-length copy in Monodelphis domestica, where the gene is apparently located on chromosome 5. Although RAG1 occupies the same chromosome as DMP1, the considerable distance that separates these loci (>2 × 10⁸ bp) suggests that they belong to different linkage groups. We obtained full-length RAG1 sequences from 41 didelphids (excluding Caluromysiops and Lestodelphys), and we downloaded shorter (543 bp) outgroup sequences from Genbank. Ingroup and outgroup sequences all translate to open reading frame. Only two indels (of 6 bp each) were required to align didelphid RAG1 sequences; one indel is unique to Metachirus nudicaudatus, whereas the other is shared by Marmosops parvidens and M. pinheiroi.

von Willebrand Factor

The human vWF gene consists of 52 exons spanning 178 kb of genomic DNA on the short arm of chromosome 12 (Mancuso et al., 1989). The gene is normally active only in vascular endothelium and in bone-marrow megakaryocytes (Ruggieri and Zimmerman, 1987). The mature (post-translationally processed) gene product is a large (2050 amino acids) glycoprotein that is assembled into multimers of various sizes; these multimers are either secreted into the plasma or into the subendothelial matrix (by endothelial cells), or they are packaged into the µ granules of platelets (by megakaryocytes). Multimeric vWF mediates adhesion of platelets to exposed subendothelium following vascular injury, and it serves as the plasma chaperone of blood factor VIII, an important regulatory protein in the coagulatory cascade (Ruggieri and Zimmerman, 1987; Ginsburg and Bowie, 1992; Sadler, 1998).

Nucleotide sequence data from exon 28 of the von Willebrand factor gene have been analyzed in many previous mammalian phylogenetic studies (e.g., Porter et al., 1996; Huchon et al., 1999; Neumann et al., 2006). The 963 bp fragment that we amplified and sequenced for this study is homologous with a single full-length copy in Monodelphis domestica, where the gene is apparently located on chromosome 8. We obtained vWF sequences from 41 didelphids (excluding Caluromysiops and Lestodelphys) and seven outgroup taxa, all of which translate to open reading frame. Only three indels (two of 3 bp and one of 6 bp) are required to align didelphid vWF sequences; all three indels are unique to single taxa.

Summary, Sequence Characteristics, and Model Fitting

Available information suggests that the five nuclear genes sequenced for this study exist as single full-length copies in the only didelphid genome currently available for inspection, that they are unlinked (on separate chromosomes or very widely separated), that they are active in different tissues, and that their translated protein products have widely differing functions (table 10). Therefore, it is reasonable to expect that taxonomic patterns of nucleotide substitution at these loci provide substantially independent evidence of phylogenetic relationships. Additionally, the coding sequences we obtained appear to be fully functional (all translate to open reading frame), and none exhibits alignment ambiguities among our ingroup taxa. Other sequence characteristics that might affect the utility of these genes for phylogenetic analysis are considered below.

Table 10

Biological Properties of Genetic Loci Sequenced for This Study^a

Base-compositional heterogeneity, as indicated by significant taxonomic variation in GC (versus AT) content, is now widely recognized as a potential problem for phylogenetic inference (see literature reviewed by Gruber et al., 2007). Of the five genes analyzed herein, BRCA1 and DMP1 exhibit the lowest overall GC content (fig. 26). For BRCA1, GC content at all codon positions is remarkably consistent across taxa. However, GC content at third codon positions (GC₃) of DMP1 is substantially higher in two taxa (Hyladelphys and Tlacuatzin) than it is in other didelphids (Jansa and Voss, 2005). As reported elsewhere, third codon positions of RAG1 are so highly variable in GC content among didelphids that plausible phylogenetic results cannot be recovered from datasets that include these positions (Gruber et al., 2007); therefore, only first and second codon positions of RAG1 are analyzed below. The GC content of vWF is most similar to that of RAG1, but vWF does not exhibit the exceptionally high taxonomic variation in GC₃ seen in the latter gene. Compared to the other loci, IRBP has relatively high GC content at all codon positions, with one taxon (Cryptonanus chacoensis) exhibiting elevated GC content at first and second codon positions.

Figure 27

The strict consensus of 1068 minimum-length trees resulting from parsimony analysis of the nonmolecular (morphological + karyotypic) dataset with Chacodelphys included. Divided circles at each resolved node indicate support from parsimony bootstrap and Bayesian analyses. For parsimony bootstrap analysis (MP), white indicates bootstrap frequencies ≤50%, grey indicates bootstrap frequencies between 50% and 75%, and black indicates bootstrap frequencies ≥75%. For the Bayesian analysis (BPP), white indicates posterior probabilities <0.95, whereas black indicates posterior probabilities ≥0.95.

Additional sequence characteristics can be effectively summarized in the context of model fitting (Posada and Crandall, 1998). The best-fitting model of nucleotide substitution for each gene is either a general-time-reversible (GTR) model or a modification thereof (TVMef) that assumes equal base frequencies and equal transition rates (table 11). In addition, the best-fitting model for each gene includes among-site rate heterogeneity as approximated by a gamma distribution (Γ), and a proportion of invariant sites (p_inv) applies to three genes. Fitted base-compositional parameters (πA, πC, πG, πT) for GTR-modeled loci are relatively even for IRBP, slightly skewed towards A for the RAG1 partition, strongly skewed towards A for BRCA1, and strongly biased towards A and G for DMP1. All five genes exhibit higher rates of transitions than transversions; in addition, first and second positions of RAG1 exhibit a relatively high rate of A-to-C transversions. The degree of among-site rate heterogeneity was lowest for BRCA1 (α  = 1.48), and highest for the RAG1 partition (α  =  0.41).

Table 11

Results of Maximum-likelihood Model Fitting with Single-gene Datasets

Analyses of Nonmolecular Data

Despite the large amount of missing data from Chacodelphys (table 2), including this taxon in a parsimony analysis of our nonmolecular (morphological + karyotypic) characters resulted in fewer equally parsimonious trees (with a correspondingly better-resolved strict consensus) than we obtained from a parsimony analysis with Chacodelphys excluded (table 12). Among other noteworthy features of this topology (fig. 27), didelphid monophyly is only weakly supported, as are most patterns of intergeneric relationships. In fact, only six genera and intergeneric clades are strongly supported (with bootstrap values >75%) by parsimony analysis of these data: (1) a group containing Lestodelphys and Thylamys; (2) a group that includes Caluromys and Caluromysiops; (3) the genus Didelphis; (4) a group that includes Lutreolina, Philander, and Didelphis; (5) a group that includes the latter three genera plus Chironectes; and (6) the genus Monodelphis. However, several other groups (e.g., the genus Thylamys, the genus Marmosops, the genus Caluromys, the genus Philander, and Philander + Didelphis) receive moderate bootstrap support.

Figure 28

The tree resulting from maximum-likelihood analysis of the IRBP dataset under its best-fitting model (GTR + I + Γ, ln-likelihood  =  −6817.82; table 11). Support statistics from a parsimony bootstrap analysis, a maximum-likelihood bootstrap analysis, and a Bayesian analysis are indicated at each resolved node. For the parsimony and maximum-likelihood analyses (MP and ML, respectively), white indicates bootstrap frequencies ≤50%, grey indicates bootstrap frequencies between 50% and 75%, and black indicates bootstrap frequencies ≥75%. For the Bayesian analysis (BPP), white indicates posterior probabilities <0.95, whereas black indicates posterior probabilities ≥0.95.

Table 12

Results of Parsimony Analyses of Datasets with Nonmolecular Characters

Bayesian analysis of these nonmolecular data (implementing the Mkv model of Lewis, 2001) provides strong support (posterior probabilities >95%) for most of the same clades that are strongly supported by parsimony, with the notable exception of (4) among those listed above. In addition, there is strong Bayesian support for the monophyly of Marmosops (only moderately supported by parsimony). However, Bayesian support is weak for several nodes with moderate parsimony bootstrap support, including the genus Thylamys, the genus Caluromys, and the genus Philander.

Most of the obvious differences between these results and our previously published analyses of didelphid morphological character data are attributable to rooting. Whereas we previously (Voss and Jansa, 2003; Voss et al., 2004a; Jansa and Voss, 2005) rooted our estimates of didelphine relationships using Caluromys, Caluromysiops, and Glironia as outgroups, these analyses are rooted using nondidelphid marsupial outgroups. Considered as an undirected network, the ingroup topology of figure 27 is almost identical to that obtained by Jansa and Voss (2005: fig. 1C) for the same set of didelphid terminal taxa.

Single-gene Analyses

Maximum-likelihood topologies with superimposed summaries of nodal statistics from three analyses (MP, ML, and Bayesian) of each single-gene dataset (figs. 28–Figure 29Figure 30Figure 3132) illustrate similarities and differences in clade support among the five loci we sequenced. Because our analytic results for IRBP and DMP1 have already been published (Jansa and Voss, 2005), only summary comments are warranted here. Briefly, the monophyly of all genera represented by two or more terminal taxa is strongly supported by both of these genes (figs. 28, 29) with the exception of Gracilinanus (weakly to moderately supported by IRBP, not recovered as monophyletic by DMP1), Thylamys (only weakly supported by Bayesian analysis of DMP1), and Didelphis (inconsistently recovered as monophyletic by analyses of IRBP). On the assumption that the DMP1 tree (lacking outgroup terminals) is appropriately rooted within a node or two of Glironia, several patterns of intergeneric relationships are also supported by both genes, including the nested clades (Metachirus (Chironectes (Lutreolina (Didelphis + Philander)))); a group that includes Lestodelphys and Thylamys; a group that includes Cryptonanus, Gracilinanus, and Lestodelphys + Thylamys; a group that includes the latter four genera plus Marmosops; and a group that includes Marmosa, Monodelphis, and Tlacuatzin. As reported elsewhere (Jansa and Voss, 2005), there are no examples of hard incongruence to be found in comparing analytic results between IRBP and DMP1 because all conflicting nodes (e.g., those resolving the position of Hyladelphys) have weak support from one or both genes.

Figure 29

The tree resulting from maximum-likelihood analysis of the DMP1 dataset under its best-fitting model (GTR + Γ, ln-likelihood  =  –5943.66; table 11). Four of the five deletion events that optimize as synapomorphies on this topology are indicated with vertical gray bars; the fifth (not shown) is shared by Marmosa regina, M. paraguayana, M. demerarae, M. murina, and M. lepida, but this clade is subtended by a zero-length branch. Because alignable DMP1 sequences are not available from outgroup taxa (see text), this tree is rooted with Glironia. Conventions for indicating nodal support are described in the caption to figure 28. The broken arrow indicates the alternative position of Tlacuatzin recovered by ML bootstrap analysis (the position for this taxon recovered by ML analysis of the original data was not supported by bootstrapping).

Figure 30

The tree resulting from maximum-likelihood analysis of first and second codon positions of RAG1 under its best-fitting model (GTR + I + Γ, ln-likelihood  =  −6343.58; table 11). The single synapomorphic deletion event in this gene is indicated with a vertical gray bar. Conventions for indicating nodal support are described in the caption to figure 28. Asterisks indicate nodes resolved by ML analysis of the original data that lack bootstrap support.

Figure 31

The tree resulting from maximum-likelihood analysis of the BRCA1 dataset under its best-fitting model (GTR + Γ, ln-likelihood  =  −15078.42; table 11). Vertical bars indicate unique and unreversed synapomorphic insertions (black) or deletions (gray) in this gene. Conventions for indicating nodal support are described in the caption to figure 28. The broken arrow indicates the alternative position of Glironia recovered by parsimony analysis.

Figure 32

The tree resulting from maximum-likelihood analysis of the vWF dataset under its best-fitting model (TVMef + I + Γ, ln-likelihood  =  −6648.62; table 11). Conventions for indicating nodal support are described in the caption to figure 28. The broken arrow indicates the alternative position of Marmosa rubra supported by Bayesian analysis.

Figure 33

The tree resulting from a mixed-model maximum-likelihood analysis of the concatenated (five-gene) dataset. Conventions for indicating nodal support are described in the caption to figure 28. The broken arrow indicates the alternative position of Glironia recovered by parsimony analysis.

Analyses of first and second codon positions of RAG1 (fig. 30) support the monophyly of most polytypic didelphid genera (except Caluromys, Didelphis, and Philander), but some genera recovered as monophyletic groups (e.g., Thylamys, Marmosops, and Monodelphis) do not receive consistently strong support. Many of the same intergeneric relationships supported by IRBP and DMP1 were also recovered, notably including the branching patterns among Metachirus, Chironectes, Lutreolina, and Philander + Didelphis. Additionally, Marmosops was recovered as the sister group to a clade that includes Cryptonanus, Gracilinanus, and Thylamys. Unfortunately, the uncertain position of the root (probably an artifact of incomplete outgroup sequences; see above) is reflected in the lack of strong support for any basal relationships in this topology.

Analyses of sequence data from BRCA1 produced the most compelling single-gene estimate of didelphid relationships recovered to date, with strikingly consistent high nodal support values and an almost completely resolved consensus topology (fig. 31). Among other salient features, every polytypic opossum genus with the exception of Didelphis was recovered with strong support, as were almost all of the higher-level clades recovered with strong support by any analysis of other single-gene datasets (the only exceptions are groups that include Lestodelphys and Caluromysiops, taxa from which BRCA1 sequences are not available). Three of these merit particular attention. One is “clade H” of Jansa and Voss (2000), which unites the Thylamys cluster with Metachirus and the large opossums with 22 chromosomes (Chironectes, Lutreolina, Philander, and Didelphis). Another is the subfamily Didelphinae of traditional usage (excluding Hyladelphys, Caluromys, and Glironia), and the third is Hyladelphys + Didelphinae. A novel result (uniquely recovered with strong support by BRCA1 but not strongly contradicted by any other gene) is the sister-group relationship between Marmosa and Tlacuatzin. In effect, the only relevant intergeneric nodes not convincingly resolved by BRCA1 are the one that clusters Cryptonanus with Gracilinanus (recovered with only weak to moderate support by MP, ML, and Bayesian analyses) and the position of the ingroup root. The latter was placed between Glironia and Caluromys with moderate support from MP, but it was placed between Caluromys + Glironia and other didelphids with weak support in ML and Bayesian analyses.

Analyses of sequence data from vWF (fig. 32) strongly support the monophyly of all polytypic didelphid genera except Didelphis and Philander. In addition, they strongly support some of the same intergeneric clades common to our other single-gene results, including (Chironectes (Lutreolina (Philander + Didelphis))); a group that includes Thylamys, Gracilinanus, and Cryptonanus; and another that includes Marmosa, Monodelphis, and Tlacuatzin. A novel result—not strongly supported by any other gene and strongly contradicted by our RAG1 results—is a sister-group relationship between Gracilinanus and Thylamys. No other intergeneric clades are consistently strongly supported by parsimony, likelihood, or Bayesian analyses of vWF sequence data; indeed, most are weakly supported by two or more analyses.

In summary, most didelphid genera were recovered as monophyletic groups with strong nodal support values from most genes (table 14). Support statistics suggest that the most problematic genera are Didelphis (recovered as monophyletic with strong support only by analyses of DMP1), Philander (inconsistently recovered as monophyletic by analyses of RAG1 and vWF), and Gracilinanus (not recovered as monophyletic by analyses of DMP1). Although Caluromys was not recovered as monophyletic by analyses of RAG1 sequences, this result is plausibly attributable to the unstable position of the ingroup root.Table 13

Table 13

Results of Parsimony Analyses of Six Molecular Datasets

Table 14

Support for Generic Monophyly from Single-gene Datasets^a

Several patterns of intergeneric relationships were also consistently recovered, usually with strong support. These include the nested clades (Chironectes (Lutreolina (Didelphis + Philander))) and a group comprised of Thylamys, Gracilinanus, and Cryptonanus. In addition, both loci (IRBP, DMP1) from which sequence data were available for Lestodelphys support a sister-group relationship between that genus and Thylamys. Lastly, all of the recovered single-gene trees are consistent with our assumption of didelphid monophyly.

With a single exception, no pattern of intergeneric relationships strongly supported by any analysis of these genes was strongly contradicted by another. Examples of relationships strongly supported by some analyses and not strongly contradicted by any other include the sister-group relationship of Marmosops to Thylamys + Gracilinanus + Cryptonanus; the sister-group relationship of Metachirus to the large opossums with 22 chromosomes; a group that includes Marmosa, Monodelphis, and Tlacuatzin; the group Marmosa + Tlacuatzin; the subfamily Didelphinae; and the clade Hyladelphys + Didelphinae. In effect, the only example of hard incongruence among patterns of intergeneric relationships supported individually by these loci concerns the relationships of Cryptonanus. This obviously problematic taxon was recovered with strong nodal support either as the sister group of Gracilinanus emiliae (in all analyses of DMP1), or as the sister group of Gracilinanus (in all analyses of RAG1), or as the sister group of Thylamys + Gracilinanus (in all analyses of vWF).

Analyses of Concatenated Genes

Maximum parsimony, maximum likelihood, and Bayesian analyses of concatenated sequence data from all five genes (7320 bp) resulted in well-resolved and strikingly similar trees with high support values at most nodes (fig. 33). Among other analytic similarities, all polytypic genera were recovered as monophyletic groups, as were all of the higher-level clades common to two or more of the single-gene analyses discussed above. However, three intergeneric nodes remain weakly or inconsistently supported.

Figure 34

Alternative resolutions for three patterns of intergeneric relationships that exhibit nodal posterior probabilities <0.95 in the Bayesian analysis of the concatenated five-gene dataset (see fig. 33). A, B, C, alternative relationships among Glironia, Hyladelphys + Didelphinae, and Caluromys + Caluromysiops; D, E, F, alternative relationships among Tlacuatzin, Marmosa, and Monodelphis; G, H, I, alternative relationships among Cryptonanus, Thylamys + Lestodelphys, and Gracilinanus. Black and grey circles at nodes indicate posterior probabilities ≥0.95 or <0.95, respectively. Branch lengths are shown at the same scale across all trees.

The first of these concerns the position of the ingroup root. Whereas maximum parsimony weakly supports placing the root between Glironia and other didelphids (which were recovered as a monophyletic group with a bootstrap frequency of 74%), maximum likelihood places the root between (Glironia (Caluromys + Caluromysiops)) and Hyladelphys + Didelphinae; however, likelihood bootstrap support for the former group is weak (<50%). Bayesian posterior probabilities for three alternative placements of the ingroup root (fig. 34A, B, C) suggest that the parsimony and likelihood solutions are almost equiprobable but slightly favor the latter.

Figure 35

The 50% majority-rule consensus of post-burnin trees resulting from a mixed-model Bayesian analysis of the combined (nonmolecular + molecular) dataset without Chacodelphys. Conventions for indicating nodal support are described in the caption to figure 28. The broken arrow indicates the alternative position of Glironia recovered by parsimony analysis.

The second equivocal node concerns the relationships of Marmosa, Tlacuatzin, and Monodelphis. Although Marmosa and Tlacuatzin were recovered as sister taxa, support for this clade was uniformly weak. Bayesian posterior probabilities for all possible resolutions of this node (fig. 34D, E, F), however, suggest a somewhat clearer ranking than was obtained for alternative placements of the ingroup root.

The third problematic issue concerns relationships among Gracilinanus, Cryptonanus, and Thylamys + Lestodelphys. Here, parsimony provides moderate bootstrap support (58%) for grouping Gracilinanus with Cryptonanus as does likelihood (70%). Although Bayesian support for Gracilinanus + Cryptonanus is weak, posterior probabilities clearly favor this clade over other phylogenetic alternatives (fig. 34G, H, I).

Analyses of Combined Datasets

Maximum parsimony and Bayesian analyses of the combined (nonmolecular + molecular) data without Chacodelphys recover the same higher-level topologies that were obtained from analyses of the concatenated-gene data, with only a few noteworthy changes in nodal support (fig. 35). For example, parsimony bootstrap support for the monophyly of Didelphis increases slightly when nonmolecular character data are added (from 80% to 93%), as does the Bayesian posterior probability for placing the ingroup root between (Glironia (Caluromys + Caluromysiops)) and other didelphids (from 0.51 to 0.87). Similarly, adding nonmolecular characters marginally improves the Bayesian posterior probability for Marmosa + Tlacuatzin (from 0.63 to 0.83).

Figure 36

The 50% majority-rule consensus of post-burnin trees resulting from a mixed-model Bayesian analysis of the combined (nonmolecular + molecular) dataset including Chacodelphys. Conventions for indicating nodal support are described in the caption to figure 28. The broken arrow indicates the alternative position of Glironia recovered by parsimony analysis.

By contrast, including Chacodelphys in a combined-data parsimony analysis dramatically erodes bootstrap support for five intergeneric nodes (fig. 36). In fact, the strict consensus of 12 minimum-length trees (not shown) places Chacodelphys as the sister group of Monodelphis with weak (MPBS  =  49%) support. Bayesian analysis, however, strongly supports the membership of Chacodelphys in a clade that also includes Gracilinanus, Cryptonanus, and Thylamys + Lestodelphys, and posterior probabilities for other nodes are substantially unaffected by taxon addition.

Figure 37

Glironia venusta (based primarily on INPA 2570, a subadult male from the Rio Urucú, Amazonas, Brazil). Because p3 is incompletely erupted on INPA 2570, the morphology and adult position of this tooth was reconstructed from other specimens. The postorbital processes become more pronounced with age.

Discussion

Analyses of our combined dataset (excluding Chacodelphys; fig. 35) effectively summarize all of the shared phylogenetic signal that we have found to date in morphology, karyotypes, and five unlinked protein-coding nuclear genes. Additional support is provided by 16 parsimony-informative indels (figs. 29, 30, 31) that were not coded as phylogenetic characters but which optimize as synapomorphies on this topology. Lastly, this tree is congruent with the results of independent phylogenetic analyses based on sequence data from intron 1 of the nuclear transthyretin (TTR) gene and the mitochondrial 12S rDNA gene (Steiner et al., 2005). To the best of our knowledge, no strongly conflicting valid results have been obtained in any other published phylogenetic analysis of molecular or morphological character data.22 Although this tree is fully resolved with high support values at most nodes, several outstanding problems merit comment.

As we have previously noted elsewhere, Chacodelphys exhibits conflicting patterns of derived morphological similarities with Thylamys + Lestodelphys on the one hand and with Monodelphis on the other (Voss et al., 2004a). Weak parsimony support for nodes along the phylogenetic path between Thylamys + Lestodelphys and Monodelphis in the combined-data analysis that includes Chacodelphys (fig. 36) presumably reflects such character conflict, although it is noteworthy that Bayesian support for the same nodes is unaffected by taxon addition. Although we are convinced that Chacodelphys is closely related to Lestodelphys and Thylamys by a host of phenotypic resemblances too indefinite to code as characters but too numerous to ignore, a compelling analytic solution to this vexing problem is unlikely to be forthcoming until at least some of the missing molecular data for Chacodelphys can be obtained from fresh material.

However, even substantial amounts of sequence data are clearly not enough to resolve other phylogenetic uncertainties. Indeed, it is not a little frustrating that, with >7000 bp of protein-coding nuclear sequence in hand, the relationships of Glironia, Cryptonanus, and Tlacuatzin should still be problematic. Glironia is of special concern, because the two plausible phylogenetic resolutions for this genus (fig. 34A, B) determine the root of the didelphid radiation. The alternative phylogenetic resolutions of Cryptonanus and Tlacuatzin within their respective groups do not seem like comparably weighty issues, but each could affect ecobehavioral character optimizations of significant evolutionary interest. Phenotypically, Cryptonanus is certainly more similar to Gracilinanus than it is to Thylamys or Lestodelphys, whereas Tlacuatzin is undeniably more similar to Marmosa than it is to Monodelphis. The positions of these genera in our combined-data trees are therefore plausible despite the absence of compelling parsimony or Bayesian support.

Although it might seem best to wait until these few remaining issues are convincingly resolved before proposing a formal classification, there is no guarantee that additional data will soon be forthcoming or useful. In the meantime, other relationships that are strongly supported by our results merit nomenclatural recognition, and the contents of some taxa currently recognized as valid need to be revised on the basis of our analytic results. As documented below, no previous classification is consistent with what is now confidently known about didelphid phylogenetic relationships.

A Revised Phylogenetic System

As reviewed above, all previous classifications of Recent opossums contain nonmonophyletic groups. Examples include “Marmosini” sensu Reig et al. (1985), “Marmosidae” sensu Hershkovitz (1992b); “Marmosopsini” sensu Kirsh and Palma (1995), and “Monodelphini” sensu Gardner (2005, 2008). Therefore, a revised phylogenetic system is needed.

The classification suggested below is based on the following principles: (1) taxa should be demonstrably monophyletic; (2) binomial nomenclature should remain as stable as possible consistent with the requirement of monophyly; (3) it is not necessary to name every node in a cladogram; (4) if future discoveries can be reasonably anticipated, taxa should be named so that these can be incorporated with minimal disruption; and (5) usage should be consistent with the International Code of Zoological Nomenclature. Points (1) and (2) should be familiar to most readers and, we trust, are not controversial. Point (3) might be disputed by those who feel that cladograms should be completely described by classifications, but the mnemonic value of names is lost if they proliferate beyond reasonable need, the hierarchical ordering implied by the consistent use of familiar suffixes (e.g., -idae, -inae, -ini) is compromised if many clade names with unfamiliar endings are interpolated, and naming weakly supported clades that might disappear when new character data or taxa are added to future analyses is not in the interest of nomenclatural stability. Point (4) refers to long branches, which can be interpreted as artifacts of extinction or incomplete taxon sampling (Horovitz, 1999); this notion provides some justification for naming monotypic higher taxa to accommodate future discoveries of new forms. Point (5) acknowledges the necessity for widely accepted rules to govern the use of technical names.

Because detailed generic synonymies have recently been published by Gardner (2008), we do not provide them here. However, a chronological list of available names with information about type species and current status (table 15) contains most of the information needed to understand current binomial usage. Under the indented heading “Contents” we list all of the valid lower-level taxa included by each higher-level taxon; for example, the genera referred to a tribe, or the species referred to a genus. In lists of congeneric species, those currently regarded as valid are in boldface, followed by an alphabetic list of available junior synonyms in parentheses; however, names of doubtful application (e.g., brasiliensis Liais, 1872, which might be a junior synonym of more than one currently recognized species of Didelphis; see Cerqueira and Tribe, 2008) are not included.

Under “Diagnosis” we list unique combinations of traits that distinguish members of suprageneric taxa from other marsupials, whether or not such traits are unambiguously assignable by parsimony as relevant synapomorphies (listed in appendix 5). Under “Morphological description” (for genera) we list all phenotypic descriptors that apply to included species; traits of particular importance for identification are indicated by italics. Ranges of metrical traits (e.g., 170–210 mm) are approximations that match or slightly exceed known minimal and maximal adult values that we believe to be reliable based on published and unpublished sources. Except as noted otherwise, all descriptions of cranial traits are based on adult specimens, and all dental descriptions are based on unworn teeth.

Under “Remarks” we discuss evidence for monophyly and other issues that relate to taxon recognition and rank. Although ranks are biologically arbitrary, they affect spelling and information retrieval in the prevailing Linnaean system, and they are regulated by widely accepted rules of biological nomenclature. The lack of species-level revisionary studies for most genera is also discussed under this heading.

Family Didelphidae Gray, 1821

CONTENTS: Glironiinae, Caluromyinae, Hyladelphinae, and Didelphinae.

DIAGNOSIS: Didelphids can be distinguished unambiguously from other marsupials by their soft (nonspinous) body pelage; five subequal claw-bearing manual digits; five separate pedal digits (loosely connected by webbing in Chironectes), of which the hallux (dI) is large and opposable and the second (dII) bears an asymmetrical grooming claw; inguinal cloaca; prehensile tail (secondarily lacking external evidence of prehensility in some taxa); long nasal bones (extending anteriorly beyond the facial processes of the premaxillae and posteriorly between the lacrimals); unfenestrated rostrum; uninflected maxillary-jugal suture; distinct foramen rotundum (not confluent or sharing a common vestibule with the sphenorbital fissure); alisphenoid-parietal contact (except in Metachirus); large interparietal fused to the supraoccipital; complete posterolateral palatal foramina; deeply divided vomer that does not conceal the presphenoid or extend posteriorly into the mesopterygoid fossa; lack of midline contact between the left and right pterygoids; unkeeled basisphenoid; well-developed tympanic process of the alisphenoid; well-developed rostral tympanic process of the petrosal; lack of a squamosal epitympanic sinus; laterally exposed ectotympanic; unspecialized malleus (with long sharply inflected neck, small orbicular apophysis, and well-developed lamina); distinct mastoid and paroccipital processes; unreduced dental complement of 50 teeth; nongliriform incisors; large milk premolars (except in Hyladelphys); tribosphenic upper molars, each with a continuous stylar shelf, uninterrupted centrocrista, and reduced or absent para- and metaconules; and tribosphenic lower molars, each with a well-developed anterior cingulid (notched for the hypoconulid of the preceding tooth), notched paracristids, and without any trace of a posterior cingulid.

Didelphids have often been compared with dasyurids, but the literature contains no adequate statement of the many characters that distinguish these superficially similar yet highly divergent clades. Among the external and craniodental features treated in this report, dasyurids consistently differ from didelphids by their small, nonopposable hallux; lack of a grooming claw on pedal digit II; a nonprehensile tail (provided with a terminal tuft of hairs that is never present in opossums); anteriorly truncated and medially notched nasals; a small interparietal that, when present in juvenile skulls, is suturally distinct from the supraoccipital; incomplete posterolateral palatal foramina; an undivided vomer that extends posteriorly to underlie the presphenoid within the mesopterygoid fossa; well-developed squamosal epitympanic sinuses; four upper and three lower incisors; a distinct posterior lobe on the unworn cutting edge of i3; vestigial milk premolars; and a distinct posterior cingulid on the lower molars.

Originally described as a didelphid by Thomas (1894) and long maintained in that family by subsequent authors (e.g., Simpson, 1945), Dromiciops differs from opossums by its basicaudal cloaca; a foramen rotundum recessed in a common vestibule with the sphenorbital fissure; frontal-squamosal contact; a large interparietal that is suturally distinct from the supraoccipital; an undivided vomer that underlies the presphenoid within the mesopterygoid fossa; left and right pterygoid bones that contact one another in the midline; keeled basisphenoid; a concealed ectotympanic; a specialized malleus (with a short uninflected neck, no orbicular apophysis, and no lamina); lack of a paroccipital process; a discontinuous stylar shelf on M1; an unstaggered i2 alveolus; and a vestigial anterior cingulid on the lower molars (of which m1 lacks a hypoconulid notch).

Caenolestids, peramelids, and other marsupials are sufficiently distinct from opossums that explicit comparisons would be pointless here. Some stem metatherians (e.g., herpetotheriids) that are strikingly similar to didelphids in most respects (including ear morphology; Gabbert, 1998) differ from didelphids by having a distinct posterior cingulid on the lower molars in addition to the diagnostic endocranial and postcranial features described by Sánchez-Villagra et al. (2007).23

REMARKS: A wide range of fossil taxa have at one time or another been regarded as didelphids (e.g., by Simpson, 1935, 1945; Clemens, 1979; Marshall, 1981; McKenna and Bell, 1997), but phylogenetic analyses suggest that most of the extinct forms once thought to be closely related to Recent opossums (e.g., †Alphadon, †Andinodelphys, †Glasbius, †Herpetotherium, †Jaskhadelphys, †Pediomys) are stem metatherians and not members of the crown group Marsupialia (Rougier et al., 1998; Wible et al., 2001; Luo et al., 2003; Sánchez-Villagra et al., 2007). Herein we explicitly restrict Didelphidae to living didelphimorphians, their most recent common ancestor, and all of its descendants.

Even more restrictive concepts of Didelphidae have been proposed, but none is widely accepted. Hershkovitz (1992b), for example, used Didelphidae to include just the large opossums with 2n  =  22 chromosomes (Chironectes, Didelphis, Lutreolina, Philander), whereas Kirsch and Palma (1995) excluded Glironia, Caluromys, and Caluromysiops from the family. Because Didelphidae in any of these applications (ours, Hershkovitz's, or Kirsch and Palma's) is monophyletic, the choice among them must be justified by other criteria. In our view, the name Didelphidae and its colloquial equivalent (“didelphids”) are so deeply entrenched in the literature as referring to all Recent opossums that more restrictive applications would serve no adequate compensatory purpose.

Although didelphid monophyly is impressively supported by nucleotide sequence data (e.g., Jansa and Voss, 2000; Amrine-Madsen et al., 2003; Meredith et al., 2008), unambiguous morphological synapomorphies of the family remain to be confidently identified. Because our analyses did not include any nonmarsupial outgroup, the position of the marsupial root node was not determined, and the phylogenetic interpretation of character-state transformations on the branch separating didelphids from nondidelphid marsupials is correspondingly equivocal. If, as most recent analyses suggest, didelphids are the basalmost branch of Marsupialia (Nilsson et al., 2004; Amrine-Madsen et al., 2003; Horovitz and Sánchez-Villagra, 2003; Asher et al., 2004; Sánchez-Villagra et al., 2007; Meredith et al., 2008; Beck, 2008), then such transformations might be didelphid synapomorphies, or they could be synapomorphies of the unnamed clade that includes caenolestids and Australidelphia.

Comparisons with stem metatherians that are believed to be close outgroups to Marsupialia, including †Mayulestes (see Muizon, 1998), †Pucadelphys (see Marshall et al., 1995), and †Herpetotherium (see Gabbert, 1998; Sánchez-Villagra et al., 2007) suggest that most of the craniodental traits by which didelphids differ from other marsupials are plesiomorphic, but one exception merits comment. Bone homologies on the posterodorsal braincase have received little attention in the literature, but they appear to be phylogenetically informative at many taxonomic levels within Metatheria. In particular, the presence of a large undivided interparietal bone that is wedged between the parietals anteriorly and fused to the supraoccipital posteriorly may be a didelphid synapomorphy. According to Muizon (1998: 38, fig. 6), paired interparietal ossifications are fused to the parietals in †Mayulestes, but the basis for this interpretation (which has no analog among living metatherians) is not explained, and it seems equally possible that the interparietal(s) is(are) absent in this taxon. In †Pucadelphys, paired interparietals (“postparietals”) are suturally distinct from the parietals and from the supraoccipital (Marshall et al., 1995: 50, fig. 12). The occiput of †Herpetotherium has not previously been described, but in a well-preserved specimen that we examined (127684 in the AMNH vertebrate paleontology catalog) there is no bone wedged between the parietals and the supraoccipital, which share a suture just behind the lambdoid crest. Therefore, the available evidence suggests that the didelphid condition is unique.

Other alleged morphological synapomorphies of didelphids include features of the spermatozoa (Temple-Smith, 1987), postcranial skeleton (Horovitz and Sánchez-Villagra, 2003), and petrosal (Ladevèze, 2007). However, the phylogenetic interpretation of such traits is compromised by sparse ingroup sampling. Spermatozooa, for example, have not been studied from many genera (e.g., Caluromysiops, Glironia, Hyladelphys, Lutreolina, Marmosops, Thylamys), and published phylogenetic analyses of postcranial and petrosal characters have not included basal didelphids (analyzed didelphid terminal taxa in both of the osteological studies cited above belong to the subfamily Didelphinae). Hopefully, future studies will help fill in many of these taxonomic gaps and contribute to a better assessment of anatomical character support for didelphid monophyly.

Only Recent didelphids are formally classified and described below, but several South American Neogene genera represented by well-preserved cranial material (†Hyperdidelphys, †Thylatheridium, †Thylophorops) are clearly members of the didelphimorph crown clade. Pending a phylogenetic analysis of their relationships with living forms, we follow current paleontological judgments of taxonomic affinity in suggesting where these fossils belong. By contrast, most other fossil didelphids are only represented by dental fragments from which few characters can be scored, and their relationships to extant taxa are correspondingly ambiguous.

Subfamily Glironiinae, new

CONTENTS: Glironia.

DIAGNOSIS: Members of this clade differ from other didelphids by the extension of soft fur along the dorsal surface of the tail from base to tip (the dorsal surface of the tail is macroscopically naked distally in other opossums), by strongly recurved and laterally compressed manual claws (manual claws are much less strongly recurved and laterally compressed in other opossums), and by postorbital processes that are formed by the frontals and parietals (postorbital processes are absent or are formed only by the frontals in other opossums).

REMARKS: The name Glironiidae as used by Hershkovitz (1992a, 1992b, 1999) was a nomen nudum because it was not accompanied by a statement of diagnostic characters (ICZN, 1999: Article 13). To our knowledge, no family-group name based on Glironia is technically available from any other publication.

Glironia Thomas, 1912

Figure 37

Figure 38

Caluromys philander (based on AMNH 266409, an adult male from Paracou, French Guiana).

CONTENTS: venusta Thomas, 1912 (including aequatorialis Anthony, 1926; and criniger Anthony, 1926).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body probably ca. 170–210 mm; adult weight probably ca. 100–200 g (measurements from the only two adult specimens known to have been measured by the American method and accompanied by weight data are in table 4). Rhinarium with two ventrolateral grooves on each side of median sulcus; dark circumocular mask present; pale supraocular spot absent; dark midrostral stripe absent; throat gland unknown (no suitably preserved adult male specimens have been examined for this feature). Dorsal body pelage unpatterned, brownish; dorsal underfur gray; dorsal guard hairs short and inconspicuous; ventral fur grayish or gray-based whitish, with or without self-whitish pectoral markings. Manus paraxonic (dIII  =  dIV); manual claws strongly recurved, laterally compressed, and much longer than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium smooth; carpal tubercles absent. Pes unwebbed; dIV longer than other pedal digits; plantar surface of heel naked. Pouch absent; mammae 2–0–2  =  4, all abdominal/inguinal; cloaca present. Tail about as long as combined length of head and body or a little longer, slender and muscular (not incrassate), and densely furred from base to tip except along ventral midline; naked ventral caudal surface modified for prehension with raised tubercles near base and apical pad bearing dermatoglyphs.

Premaxillary rostral process absent. Nasals long, extending anteriorly beyond I1 (concealing nasal orifice from dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. Lacrimal foramina concealed within anterior orbital margin or exposed laterally, usually two on each side. Orbits very large; supraorbital crests well developed; flattened triangular postorbital processes present, formed by both frontal and parietal bones. Left and right frontals and parietals separated by persistent median sutures. Parietal contacts alisphenoid on lateral braincase (no frontal-squamosal contact). Sagittal crest absent. Petrosal not laterally exposed through fenestra in parietal-squamosal suture (fenestra absent). Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine fenestrae usually present, but small; palatine fenestrae absent; maxillary fenestrae absent; posterolateral palatal foramina small, not extending anteriorly between M4 protocones; posterior palatal morphology Caluromys-like (without prominent lateral corners, the choanae not constricted behind). Maxillary and alisphenoid not in contact on floor of orbit (separated by palatine). Transverse canal foramen present. Alisphenoid tympanic process smoothly globular, without anteromedial process or posteromedial lamina enclosing extracranial course of mandibular nerve (secondary foramen ovale absent), and not in contact with rostral tympanic process of petrosal. Anterior limb of ectotympanic suspended directly from basicranium. Stapes triangular with large obturator foramen. Fenestra cochleae exposed, not concealed by rostral and caudal tympanic processes of petrosal. Paroccipital process small, adnate to petrosal. Dorsal margin of foramen magnum bordered by exoccipitals and supraoccipital, incisura occipitalis present.

Two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 asymmetrical (“incisiform”), with longer anterior than posterior cutting edges; I5 separated from I4 by small diastema in some specimens (e.g., INPA 2570) but not others (e.g., FMNH 41440). Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps. First upper premolar (P1) present, smaller than more posterior premolars but well formed and not vestigial; second and third upper premolars (P2 and P3) subequal in height; P3 with both anterior and posterior cutting edges. Molars weakly carnassialized (postmetacristae longer than postprotocristae); relative widths M1 < M2 < M3 > M4; centrocrista weakly inflected labially on M1–M3; ectoflexus absent or indistinct on M1 and M2, distinct but usually shallow on M3; anterolabial cingulum continuous with preprotocrista (complete anterior cingulum present) on M3; postprotocrista without carnassial notch. Last upper tooth to erupt is P3.

Lower incisors (i1–i4) with distinct lingual cusps. Lower canine (c1) erect, acutely pointed, and simple (without a posterior accessory cusp). Second lower premolar (p2) taller than p3; morphology of lower milk premolar (dp3) unknown (no juvenile specimens examined). Hypoconid labially salient on m3; hypoconulids twinned with entoconids on m1–m3; entoconids much taller than hypoconulids on m1–m3.

DISTRIBUTION: Glironia is currently known from fewer than two dozen specimens collected at widely scattered localities between 300 and 1000 m above sea level in Brazil (Amazonas, Mato Grosso, Pará, Rodônia), eastern Ecuador, eastern Peru, and eastern Bolivia (Díaz and Willig, 2004; Santos-Filho et al., 2007; Barkley, 2008). Although the genus probably also occurs in southeastern Colombia, no specimens are known to have been collected there (Díaz and Willig, 2004). Most records accompanied by definite habitat information are from primary or secondary Amazonian rainforest, but a single specimen was recently reported from the upper Paraguay Basin (Santos-Filho et al., 2007), and several Bolivian records are from dry forest (Tarifa and Anderson, 1997; Emmons, 1998).

REMARKS: Only the type species of Glironia is currently recognized as valid, but it seems probable that additional taxa are represented among the material now preserved in museum collections.

Subfamily Caluromyinae Reig et al., 1987

CONTENTS: Caluromys and Caluromysiops.

DIAGNOSIS: Caluromyines can be distinguished from other confamilial taxa by their long fourth manual digit (dIII is the longest manual digit, or dIII and dIV are subequal in all other didelphids); a completely ossified palate (maxillopalatine and sometimes additional palatal fenestrae are consistently present in most other didelphids); lack of a transverse canal foramen (transverse canal foramina are almost invariably present in all other didelphids); an alisphenoid tympanic process that contacts or closely approximates the rostral tympanic process of the petrosal (the alisphenoid tympanic process and the rostral tympanic process of the petrosal are widely separated in other didelphids); an indirectly suspended anterior limb of the ectotympanic (the anterior limb of the ectotympanic is directly suspended from the skull in most other didelphids); a fenestra cochleae that is concealed in a sinus formed by the rostral and caudal tympanic processes of the petrosal (the fenestra cochleae is exposed in most other didelphids); a blunt and weakly inflected angular process (the mandibular angle is acute and strongly inflected in all other didelphids); an upper canine alveolus that is completely contained by the maxilla (C1 occupies the premaxillary-maxillary suture in all other didelphids); a vestigial or absent first upper premolar (P1 is present and nonvestigial in all other didelphids); a tall second upper premolar that much exceeds P3 in height (P2 is either subequal to or smaller than P3 in most other didelphids); and lack of an ectoflexus on all of the upper molars (a distinct ectoflexus is present, at least on M3, in all other didelphids).

REMARKS: Caluromyine monophyly is strongly supported by morphological characters (appendix 5) and IRBP gene sequences (fig. 28), and this clade is recovered with strong support in analyses of concatenated genes (fig. 33) and total evidence (nonmolecular + molecular characters; figs. 35, 36) despite the overall incompleteness of our data from Caluromysiops (table 2). Two deletions at the DMP1 locus that currently optimize as synapomorphies of Caluromys (fig. 29) might prove to be caluromyine synapomorphies when this gene is eventually sequenced for Caluromysiops. Although most recent authors (e.g., Reig et al., 1985, 1987; Kirsch and Palma, 1995; McKenna and Bell, 1997; Gardner, 2005, 2008) have included Glironia in the Caluromyinae (or Caluromyidae), that genus appears to represent an ancient lineage of basal didelphids that may not be closely related to Caluromys + Caluromysiops (see Jansa and Voss, 2000; this report, above). The authorship of family-group names based on Caluromys is often attributed to Kirsch (1977b), but no such name in that work fulfills the technical criteria for nomenclatural availability (ICZN, 1999: Article 13). Apparently, the first explicit statement of characters purported to differentiate this taxon from other opossums appeared in Reig et al. (1987: 72).

Caluromys J.A. Allen, 1900

Figure 38

Figure 39

Caluromysiops irrupta (a composite illustration based on several imperfect adult specimens, mostly from zoos: AMNH 208101, 244364; USNM 397626; FMNH 84426).

CONTENTS: derbianus Waterhouse, 1841 (including antioquiae Matschie, 1917; aztecus Thomas, 1913; canus Matschie, 1917; centralis Hollister, 1914; fervidus Thomas, 1913; guayanus Thomas, 1899; nauticus Thomas, 1913; pallidus Thomas, 1899; pictus Thomas, 1913; pulcher Matschie, 1917; pyrrhus Thomas, 1901; and senex Thomas, 1913); lanatus Olfers, 1818 (including bartletti Matschie, 1917; cahyensis Matschie, 1917; cicur Bangs, 1898; hemiurus Miranda-Ribeiro, 1936; jivaro Thomas, 1913; juninensis Matschie, 1917; lanigera Desmarest, 1820; meridensis Matschie, 1917; modesta Miranda-Ribeiro, 1936; nattereri Matschie, 1917; ochropus Wagner, 1842; ornatus Tschudi, 1845; and vitalinus Miranda-Ribeiro, 1936); and philander Linnaeus, 1758 (including affinis Wagner, 1842; cajopolin Müller, 1776; cayopollin Schreber, 1777; cayopollin Kerr, 1792; dichurus Wagner, 1842; flavescens Brongniart, 1792; leucurus Thomas, 1904; trinitatis Thomas, 1894; and venezuelae Thomas, 1903).

These taxa are currently allocated to two subgenera, Caluromys J.A. Allen, 1900 (containing only C. philander), and Mallodelphys Thomas, 1920 (containing C. derbianus and C. lanatus), that can be distinguished unambiguously by the traits noted below.

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 210–300 mm; adult weight ca. 190–500 g. Rhinarium with two ventrolateral grooves on each side of median sulcus; dark circumocular mask present; pale supraocular spot absent; dark midrostral stripe present; throat gland absent. Dorsal pelage unpatterned, grayish or reddish brown, or indistinctly mottled with same colors; dorsal underfur grayish; dorsal guard hairs short and inconspicuous; ventral fur gray based or self-colored. Manual digit IV longer than other manual digits; manual claws longer than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium smooth or sparsely tuberculate; carpal tubercles absent. Pedal digits unwebbed; pedal digit IV longer than other pedal digits; plantar surface of heel naked. Pouch present, consisting of separate lateral skin folds opening medially (subgenus Caluromys) or a deep pocket opening anteriorly (subgenus Mallodelphys); known mammary formulae apparently 2–0–2  =  4 (in C. lanatus) and 3–1–3  =  7 (in C. philander); cloaca present. Tail much longer than combined length of head and body, slender and muscular (not incrassate); body fur extends onto tail base to the same extent dorsally as ventrally (subgenus Caluromys) or much further dorsally than ventrally (subgenus Mallodelphys); naked caudal integument dark (grayish or brownish) basally and whitish distally with mottled transition; caudal scales in spiral series, each scale with three subequal bristlelike hairs emerging from distal margin (the central hair of each caudal-scale triplet longer than the lateral hairs but not grossly swollen or petiolate); naked ventral caudal surface modified for prehension, with raised tubercles near base (Mallodelphys only) and apical pad bearing dermatoglyphs.

Premaxillary rostral process present. Nasals long, extending anteriorly above I1 (concealing nasal orifice from dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. Lacrimal foramina exposed to lateral view at or near anterior orbital margin, usually two on each side. Large, flattened, triangular postorbital processes of frontals present. Left and right frontals and parietals separated by persistent median sutures. Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest weakly developed on interparietal and/or along midparietal suture (not extending anteriorly to frontals) in some large adult specimens. Petrosal not laterally exposed through fenestra in parietal-squamosal suture (fenestra absent). Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine fenestrae usually absent (small fenestrae, never extending anteriorly beyond M1 or posteriorly beyond M2, are uni- or bilaterally present in occasional specimens); palatine and maxillary fenestrae absent; posterolateral palatal foramina small, not extending anteriorly between M4 protocones; posterior palate gently sloping ventrally, usually with arched caudal margin and without prominent posterolateral corners (the choanae unconstricted behind). Maxillary and alisphenoid not in contact on floor of orbit (separated by palatine). Transverse canal foramen absent. Alisphenoid tympanic process more or less conical (acutely pointed ventrally), usually without posteromedial lamina enclosing extracranial course of mandibular nerve (secondary foramen ovale absent),24 and contacting or closely approximating rostral tympanic process of petrosal. Anterior limb of ectotympanic indirectly suspended from basicranium (by malleus). Stapes triangular with large obturator foramen (in subgenus Caluromys) or subtriangular with small but patent foramen (in subgenus Mallodelphys). Fenestra cochleae concealed in sinus formed by rostral and caudal tympanic processes of petrosal. Paroccipital process small, rounded and subtriangular, adnate to petrosal. Dorsal margin of foramen magnum bordered by supraoccipital and exoccipitals, incisura occipitalis present.

Two mental foramina usually present on lateral surface of each hemimandible; angular process blunt and weakly inflected.

Unworn crowns of I2–I5 asymmetrical (“incisiform”), with much longer anterior than posterior cutting edges. Upper canine (C1) alveolus contained entirely in maxillary bone; C1 simple, without accessory cusps. First upper premolar (P1) minute, vestigial, and lacking any consistent occlusal features; second upper premolar (P2) much taller than P3; P3 with both anterior and posterior cutting edges; upper milk premolar (dP3) large and molariform. Molars not carnassialized (postprotocristae and postmetacristae are subequal in subgenus Mallodelphys) or weakly carnassialized (postmetacristae are slightly longer than postprotocristae in subgenus Caluromys); relative widths usually M1 < M2 > M3 > M4 (subgenus Mallodelphys) or M1 < M2 < M3 > M4 (subgenus Caluromys); centrocrista weakly inflected labially on M1–M3; ectoflexus consistently absent on all upper molars; anterolabial cingulum continuous with preprotocrista (complete anterior cingulum present) on M3; postprotocrista without carnassial notch. Last upper tooth to erupt is P3.

Lower incisors (i1–i4) with distinct lingual cusps. Lower canine (c1) erect, acutely pointed, and simple (without a posterior accessory cusp). Second upper premolar (p2) much taller than p3; lower milk premolar (dp3) trigonid with or without paraconid (when present, the paraconid is small and often indistinct). Hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid much taller than hypoconulid on m1–m3.

DISTRIBUTION: Caluromys occurs from the Mexican state of Veracruz southward throughout most of the forested regions of Central and South America (including Trinidad) to eastern Bolivia, eastern Paraguay, and northeastern Argentina (Hall, 1981; Gardner, 2008). Most specimens of C. lanatus and C. philander (e.g., those reported by Handley, 1976; Anderson, 1997) are from lowland or lower montane rainforest; however, C. philander has also been observed in dry forest (Emmons, 1998), and C. derbianus ranges from sea level to over 2000 m and occurs in both rainforest and dry forest (Handley, 1966; Bucher and Hoffmann, 1980; Sánchez et al., 2004).

REMARKS: Generic monophyly (vis-à-vis Caluromysiops) is supported by IRBP sequence data (fig. 28), by the presence of a distinct midrostral stripe, and by the presence of a small but distinct rostral process of the premaxillae (the latter two traits optimize as unambiguous generic synapomorphies; appendix 5).

The genus Caluromys has never been revised, and there is substantial divergence in morphology and/or cytochrome b sequences among some allegedly conspecific populations (Voss et al., 2001; Patton and Costa, 2003). Indeed, no revisionary study has documented the conspecific status of the many subjective synonyms listed above for either C. philander or C. lanatus. It is an open question as to whether the subgenera Caluromys and Mallodelphys will prove to be reciprocally monophyletic or not in future phylogenetic analyses based on denser taxon sampling.

Caluromysiops Sanborn, 1951

Figure 39

Figure 40

Hyladelphys kalinowskii (based on AMNH 267338, an adult male from Paracou, French Guiana).

CONTENTS: irrupta Sanborn, 1951.

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body probably ca. 250–300 mm (Izor and Pine, 1987); adult weight probably ca. 300–500 g. Rhinarium with two ventrolateral grooves on each side of median sulcus; fur of head uniformly pale, without any marking (dark circumocular mask, supraocular spots, and midrostral stripes absent); throat gland absent. Dorsal pelage pale overall but prominently marked by paired blackish stripes extending from forelimbs to shoulders and thence posteriorly as narrowing parallel bands to rump; dorsal underfur dark (apparently grayish but discolored with age in all examined specimens); dorsal guard hairs short; ventral fur variously colored (gray-based buffy in some specimens, self-buffy in others). Manual digit IV longer than other manual digits; manual claws longer than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium smooth; carpal tubercles absent. Pedal digits unwebbed; dIV longer than other pedal digits; plantar surface of heel usually naked. Pouch said to be present (Izor and Pine, 1987) but morphological details unknown; mammary formula unknown; cloaca present. Tail longer than combined length of head and body, slender and muscular (not incrassate); densely covered with body fur from base almost to tip except along ventral midline and posterolaterally; posterolateral caudal scales in spiral series, each scale with three subequal bristlelike hairs emerging from distal margin; naked ventral surface modified for prehension with raised tubercles near base, and apical pad bearing dermatoglyphs.

Premaxillary rostral process absent. Nasals long, extending anteriorly above I1 (concealing nasal orifice from dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. One or two lacrimal foramina exposed on each side on or near anterior orbital margin. Large flattened, triangular postorbital processes of frontals present, continuous with supraorbital crests. Left and right frontals and parietals co-ossified in most adult specimens (median sutures incomplete or obliterated). Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest present, prominently developed on parietals and extending onto frontals in most adult specimens. Petrosal not laterally exposed through fenestra in parietal-squamosal suture (fenestra absent). Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine, palatine, and maxillary fenestrae absent (palate completely ossified); posterolateral palatal foramina small, not extending anteriorly between M4 protocones; posterior palatal morphology conforms to Caluromys morphotype (without strongly developed lateral corners, the choanae unconstricted behind). Maxillary and alisphenoid widely separated by palatine on floor of orbit. Transverse canal foramen absent. Alisphenoid tympanic process smoothly globular, apparently always with posteromedial lamina enclosing extracranial course of mandibular nerve (secondary foramen ovale present), and extending posteriorly to contact or closely approximate rostral tympanic process of petrosal. Anterior limb of ectotympanic indirectly suspended from basicranium (by malleus). Stapes usually columellar and imperforate (rarely microperforate). Fenestra cochleae usually concealed in sinus formed by rostral and caudal tympanic processes of petrosal (but not in USNM 397626, possibly due to pathology or postmortem damage). Paroccipital process small, adnate to petrosal. Dorsal margin of foramen magnum bordered by supraoccipital and exoccipitals, incisura occipitalis present.

Two mental foramina usually present on lateral surface of each hemimandible; angular process obtuse, weakly inflected.

Unworn crowns of I2–I5 asymmetrical (“incisiform”), with longer anterior than posterior cutting edges. Upper canine (C1) alveolus contained entirely within maxillary bone; C1 simple (without accessory cusps), and very large in adult specimens. First upper premolar (P1) absent or minute and vestigial; second upper premolar (P2) much taller than P3; P3 with both anterior and posterior cutting edges; upper milk premolar (dP3) large and molariform. Molars not carnassialized (postmetacristae subequal to postprotocristae); relative widths M1 < M2 > M3 > M4; centrocrista straight (uninflected) on M1–M3; ectoflexus absent on all upper molars; anterolabial cingulum continuous with preprotocrista (complete anterior cingulum present) on M3; postprotocrista without carnassial notch. Last upper tooth to erupt is P3.

Lower incisors (i1–i4) with distinct lingual cusps. Lower canine (c1) erect, acutely pointed, and simple (without a posterior accessory cusp). Second lower premolar (p2) much taller than p3; lower milk premolar (dp3) large and nonvestigial, but paraconid not well developed (trigonid incomplete). Hypoconid labially salient on m3; hypoconulid not twinned with entoconid on any lower molar; entoconid much taller than hypoconulid on m1–m3.

DISTRIBUTION: Caluromysiops is currently known from just seven localities in the rainforested Amazonian lowlands of southeastern Colombia, eastern Peru and western Brazil (Emmons, 2008).

REMARKS: Various authors (e.g., Cabrera, 1958; Simpson, 1972; Izor and Pine, 1987) have questioned whether it is useful to separate Caluromysiops from Caluromys, but none explicitly considered the suite of trenchant morphological characters that distinguish these taxa. Among other features, Caluromysiops differs from Caluromys by its lack of a circumocular mask and dark midrostral stripe, presence of dark scapular stripes, shorter tail (table 5), absence of a premaxillary rostral process, co-ossified frontals, presence of a well-developed sagittal crest, presence of a secondary foramen ovale, globular (versus conical) alisphenoid bullae, columelliform stapes, linear centrocrista, and untwinned hypoconulid. Although there is no cladistic problem with combining these taxa under a single generic name, maintaining current binomial usage is likewise unobjectionable and serves to separate the monophyletic cluster of species currently known as Caluromys from its long-branched sister taxon.

Subfamily Hyladelphinae, new

CONTENTS: Hyladelphys.

DIAGNOSIS: Members of this clade uniquely differ from other didelphids by their vestigial milk dentition (dP3/dp3 are large, more or less molariform teeth in other opossums; Voss et al., 2001: table 5). Additional contrasts among hyladelphines and members of other didelphid subfamilies were tabulated by Jansa and Voss (2005: table 2).

REMARKS: Given the phylogenetic results at hand, a monotypic suprageneric taxon for Hyladelphys could either be ranked as a subfamily (as it is here) or as a tribe (if the genus were referred to the Didelphinae). Although the issue of rank is not biologically meaningful, the former option serves to emphasize the intermediate position of this odd little opossum between basal didelphids (Glironia and caluromyines) and the speciose radiation of lineages that are more closely related to Didelphis. The branch leading to Hyladelphys is very long in most molecular reconstructions of didelphid anagenesis (e.g., fig. 33), suggesting an ancient history of independent evolution accompanied by extinction of transitional forms; the same conclusion is suggested by its uniquely reduced milk dentition and by characters of the postcranial skeleton (Flores, 2009). In effect, excluding Hyladelphys from Didelphinae simplifies the diagnosis of the latter clade and provides a new higher taxon to accommodate fossil relatives of the former, should any be discovered.

Hyladelphys Voss et al., 2001

Figure 40

Figure 41

Marmosa murina (based on AMNH 267368, an adult of unknown sex from Paracou, French Guiana). The inset shows the morphology of the unworn lower canine (with flattened apical blade; based on AMNH 266416, a young adult female from the same locality).

CONTENTS: kalinowskii 137Hershkovitz, 1992.

MORPHOLOGICAL DESCRIPTION: Combined adult length of head and body ca. 75–95 mm; adult weight ca. 10–20 g. Rhinarium with two ventrolateral grooves on each side of median sulcus; dark circumocular mask present, extending posteriorly from mystacial region to base of ear on each side of face; pale supraocular spot absent; dark midrostral stripe absent; throat gland absent. Dorsal pelage unpatterned reddish-brown with dark-gray hair bases; dorsal guard hairs short and inconspicuous; ventral fur self-white. Manus paraxonic (dIII  =  dIV); manual claws relatively large, strongly recurved, and slightly longer than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium smooth or sparsely tubercular; carpal tubercles absent. Pedal digits unwebbed; dIV longer than other pedal digits; plantar surface of heel naked. Pouch absent; mammae 2–0–2  =  4, all abdominal-inguinal; cloaca present. Tail much longer than combined length of head and body, slender and muscular (not incrassate), and apparently naked (without a conspicuously furred base); caudal scales in both annular and spiral series, each scale with three subequal bristle-like hairs emerging from distal margin; ventral caudal surface modified for prehension distally, with apical pad bearing dermatoglyphs.

Premaxillary rostral process absent. Nasals long, extending anteriorly above I1 (concealing most of nasal orifice from dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. Lacrimal foramina exposed laterally on or near anterior orbital margin, one or two on each side. Orbits very large; interorbital region strongly convergent anteriorly, with beaded dorsolateral margins; postorbital processes absent. Left and right frontals and parietals separated by persistent median sutures. Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest absent. Petrosal not exposed laterally through fenestra in parietal-squamosal suture (fenestra absent). Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine fenestrae present; palatine and maxillary fenestrae absent; posterolateral palatal foramina small, not extending anteriorly between M4 protocones. Posterior palatal morphology more Didelphis-like than Caluromys-like (with moderately well-developed posterolateral corners, the internal choanae distinctly constricted behind). Maxillary and alisphenoid bones not in contact on floor of orbit (widely separated by palatine). Transverse canal foramen present. Alisphenoid tympanic process smoothly globular, without anteromedial process or posteromedial lamina (secondary foramen ovale absent), and not in contact with rostral tympanic process of petrosal. Anterior limb of ectotympanic suspended directly from basicranium. Stapes triangular with large obturator foramen. Fenestra cochleae exposed, not concealed by rostral and caudal tympanic processes of petrosal. Paroccipital process small and adnate to petrosal. Dorsal margin of foramen magnum bordered by supraoccipital and exoccipitals, incisura occipitalis present.

One or two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 asymmetrical (“incisiform”), with much longer anterior than posterior cutting edges. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps. First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; second upper premolar (P2) much taller than P3; P3 with both anterior and posterior cutting edges; milk premolar (dP3) very small, vestigial, and lacking distinct occlusal features. Upper molars not strongly carnassialized (postmetacristae only slightly longer than postprotocristae); relative widths M1 < M2 < M3 > M4; centrocrista weakly inflected labially on M1–M3; ectoflexus indistinct on M1, shallow on M2, distinct on M3; anterolabial cingulum continuous with preprotocrista (complete anterior cingulum present) on M3; postprotocrista without carnassial notch. Last upper tooth to erupt is P3.

Lower incisors (i1–i4) with distinct lingual cusps. Lower canine (c1) erect, acutely pointed, and simple (without a posterior accessory cusp). Second lower premolar (p2) taller than p3; lower milk premolar (dp3) small, vestigial, and lacking distinct occlusal features. Hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid taller than hypoconulid on m1–m3.

DISTRIBUTION: Currently known from fewer than a dozen specimens, Hyladelphys has been reported from just nine localities in the rainforested lowlands of eastern Peru, central Amazonian Brazil, southern Guyana, and French Guiana (Astúa, 2007). Extrapolating from these scanty data is obviously problematic, but it would not be surprising to find this elusive taxon anywhere in Amazonia.

REMARKS: High levels of molecular divergence between sequenced specimens of Hyladelphys from French Guiana and Peru, together with geographic variation in morphological characters, suggest that additional species remain to be described in this genus (Jansa and Voss, 2005).

Subfamily Didelphinae Gray, 1821

CONTENTS: Marmosini, Metachirini, Didelphini, and Thylamyini.

DIAGNOSIS: Members of this clade uniquely differ from other didelphids by lacking an anterior cutting edge on P3 (the third upper premolar is double bladed in all basal didelphids). In addition, Didelphinae differs from Glironiinae by having a caudal dorsum that is macroscopically naked, at least near the tip; an unpaired median teat; well-developed posterolateral palatal corners; and postorbital processes that (when present) are formed only by the frontals. Didelphinae additionally differs from Caluromyinae by having a shorter fourth manual digit (dIII is subequal to or longer than dIV), an incompletely ossified palate (maxillopalatine and sometimes other fenestrae are invariably present), transverse canal foramina, a wide gap between the alisphenoid tympanic process and the rostral tympanic process of the petrosal, an acute and strongly inflected mandibular angle, an upper canine alveolus that lies within the premaxillary-maxillary suture, a nonvestigial P1, a larger P3 (subequal to or taller than P2), and a distinct ectoflexus on M3. Didelphinae additionally differs from Hyladelphinae by having an unpaired median teat, nonvestigial milk teeth, and a larger P3 (subequal to or taller than P2).

REMARKS: The monophyly of Didelphinae is strongly supported by all analyses of BRCA1 sequences (fig. 31), concatenated sequence data from five genes (fig. 33), and combined (nonmolecular + molecular) data when Chacodelphys is excluded (fig. 35). Subfamilial monophyly is also supported by a uniquely shared insertion (not coded for phylogenetic analysis) at the BRCA1 locus (fig. 31). However, only two morphological traits optimize as unambiguous synapomorphies of Didelphinae: an unpaired median teat and a strongly labially inflected centrocrista (subsequently reversed in Didelphini; appendix 5). In addition to genera referred to the tribes named above, the genus †Hyperdidelphys (here unassigned to tribe) clearly belongs to this subfamily based on characters described and illustrated by Goin and Pardiñas (1996) and others subsequently scored for phylogenetic analysis by D.A. Flores and the senior author.

Tribe Marmosini 137Hershkovitz, 1992

CONTENTS: Marmosa, Monodelphis, †Thylatheridium, and Tlacuatzin.

DIAGNOSIS: Marmosines can be distinguished from other members of the subfamily Didelphinae by lacking a pouch (present in all Didelphini), by lacking an anteromedial bullar process spanning the transverse canal foramen (present in Metachirini and most Thylamyini), by lacking a fenestra in the squamosal-parietal suture (present in most Thylamyini), and by having the supraoccipital form part of the dorsal margin of the foramen magnum (absent in adult Metachirini and most adult Didelphini).

REMARKS: The monophyly of Marmosini as construed herein is strongly supported by all analyses of BRCA1 (fig. 31), vWF (fig. 32), concatenated sequence data from five genes (fig. 33), and our combined (nonmolecular + molecular) dataset excluding Chacodelphys (fig. 35). However, no morphological trait optimizes as an unambiguous synapomorphy of this molecularly robust clade. Instead, all of the diagnostic features listed above appear to be plesiomorphic within the subfamily Didelphinae.

Although the name Marmosini was used by Reig et al. (1985, 1987) it was not accompanied by an explicit statement of taxonomically differentiating characters. To the best of our knowledge, the first family-group name based on Marmosa is technically available from Hershkovitz (1992b). An alternative name for this clade is Monodelphini, which McKenna and Bell (1997) attributed to Talice et al. (1961). However, Talice et al. did not mention any characters purported to differentiate Monodelphini from other didelphids, so their name is unavailable (ICZN, 1999: Article 13). To the best of our knowledge, the first family-group name based on Monodelphis is also available from Hershkovitz (1992b). As first revisors in the sense of the Code (ICZN, 1999: Article 24), we select Marmosini 137Hershkovitz, 1992, to have precedence over Monodelphini 137Hershkovitz, 1992.

Our reluctance to recognize any subtribal distinction between Marmosa and Monodelphis is based on the uncertain position of Tlacuatzin, which is not convincingly resolved despite the large amount of data at hand (>2000 parsimony-informative characters; table 12). In the event that a sister-group relationship between Marmosa and Tlacuatzin (weakly indicated by supermatrix analyses; figs. 33–Figure 34Figure 3536) were strongly supported by some future dataset, it would then make sense to recognize one subtribe (e.g., Marmosina) for those taxa and another (Monodelphina) for Monodelphis and †Thylatheridium. Although it seems indisputable that Monodelphis and †Thylatheridium are closely related (Goin and Rey, 1997), we note that this hypothesis has yet to be tested analytically, nor is it certainly known whether or not these taxa are reciprocally monophyletic.25

Marmosa Gray, 1821

Figure 41

Figure 42

Monodelphis brevicaudata (based on AMNH 257203, an adult female from San Ignacio de Yuruaní, Bolívar, Venezuela).

CONTENTS: We recognize two subgenera, Marmosa Gray, 1821, and Micoureus Lesson, 1842 (see Remarks, below).

The subgenus Marmosa contains andersoni Pine, 1972; lepida Thomas, 1888 (including grandis Tate, 1931); mexicana Merriam, 1897 (including mayensis Osgood, 1913; savannarum Goldman, 1917; and zeledoni Goldman, 1917); murina Linnaeus, 1758 (including bombascarae Anthony, 1922; chloe Thomas, 1907; dorsigera Linnaeus, 1758; duidae Tate, 1931; guianensis Kerr, 1792; klagesi J.A. Allen, 1900; macrotarsus Wagner, 1842; madeirensis Cabrera,. 1913; maranii Thomas, 1924; meridionalis Miranda-Ribeiro, 1936; moreirae Miranda-Ribeiro, 1936; muscula Cabanis, 1848; parata Thomas, 1911; roraimae Tate, 1931; tobagi Thomas, 1911; and waterhousei Tomes, 1860); quichua Thomas, 1899 (including musicola Osgood, 1913); robinsoni Bangs, 1898 (including casta Thomas, 1911; chapmani J.A. Allen, 1900; fulviventer Bangs, 1901; grenadae Thomas, 1911; isthmica Goldman, 1912; luridavolta Goodwin, 1961; mimetra Thomas, 1921; mitis Bangs, 1898; nesaea Thomas, 1911; pallidiventris Osgood, 1912; ruatanica Goldman, 1911; and simonsi Thomas, 1899); rubra Tate, 1931; tyleriana Tate, 1931 (including phelpsi Tate, 1939); and xerophila Handley and Gordon, 1979.

The subgenus Micoureus contains alstoni J.A. Allen, 1900 (including nicaraguae Thomas, 1905); constantiae Thomas, 1904 (including budini Thomas, 1920); demerarae Thomas, 1905 (including areniticola Tate, 1931; domina Thomas, 1920; esmeraldae Tate, 1931; limae Thomas, 1920; and meridae Tate, 1931); paraguayana Tate, 1931 (including cinerea Temminck, 1824 [preoccupied]); phaea Thomas, 1899 (including perplexa Anthony, 1922); and regina Thomas, 1898 (including germana Thomas, 1904; mapiriensis Tate, 1931; parda Tate, 1931; rapposa Thomas, 1899; and rutteri Thomas, 1924).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 100–210 mm; adult weight ca. 20–170 g. Rhinarium with two ventrolateral grooves on each side of median sulcus; dark circumocular mask present; pale supraocular spot absent; dark midrostral stripe absent; throat gland absent in some species (e.g., M. murina) but present in adult males of other species (e.g., M. mexicana). Dorsal pelage unpatterned, superficially brownish, reddish, or grayish, but dorsal hair bases always dark gray; dorsal guard hairs short and inconspicuous; ventral fur superficially whitish, yellowish, or orange, wholly or partly gray based (apparently never completely self-colored). Manus paraxonic (dIII  =  dIV); manual claws about as long as or slightly longer than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium smooth or sparsely covered with flattened tubercles (never densely tuberculate); carpal tubercles completely absent in some species (e.g., M. murina), or only lateral carpal tubercles present in adult males (e.g., in M. lepida), or both medial and lateral carpal tubercles present in adult males (e.g., in M. robinsoni). Pedal digits unwebbed; dIV longer than other pedal digits; plantar surface of heel macroscopically naked, but all or part of heel covered with microscopic hairs in some species (e.g., M. rubra). Pouch absent; mammae 3–1–3  =  7 (e.g., in M. lepida) to 7–1–7  =  15 (in M. mexicana), all abdominal-inguinal; cloaca present. Tail always substantially longer than combined length of head and body; slender and muscular (not incrassate); without a conspicuously furred base in some species (e.g., M. murina) or tail base conspicuously furred to about the same extent dorsally as ventrally (e.g., in M. paraguayana); naked caudal integument unicolored (all dark) in most species but mottled distally with white spots and/or white tipped in others (e.g., M. paraguayana); caudal scales in spiral series (e.g., in M. murina) or in both spiral and annular series (e.g., in M. mexicana), each scale usually with three subequal bristlelike hairs emerging from distal margin; ventral caudal surface always modified for prehension distally, with apical pad bearing dermatoglyphs.

Premaxillary rostral process present in most species (absent in M. xerophila). Nasals long, extending anteriorly beyond I1 (concealing nasal orifice in dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. Lacrimal foramina usually two on each side, exposed to lateral view on or near anterior orbital margin. Supraorbital margins with distinct beads or prominent crests; flattened, triangular postorbital processes usually present in fully mature adults but substantially larger in some species than in others (absent or indistinct in M. rubra). Left and right frontals and parietals separated by persistent median sutures. Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest never developed. Petrosal usually not exposed laterally through fenestra in parietal-squamosal suture (fenestra absent).26 Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine fenestrae present; palatine fenestrae absent in most species (but consistently present in some; e.g., M. mexicana); maxillary fenestrae absent; posterolateral palatal foramina small, never extending anteriorly between M4 protocones; posterior palatal morphology conforms to Didelphis morphotype (with well-developed lateral corners, the choanae constricted behind). Maxillary and alisphenoid not in contact on floor of orbit (separated by palatine). Transverse canal foramen present. Alisphenoid tympanic process smoothly globular, without anteromedial process or posteromedial lamina enclosing the maxillary nerve (secondary foramen ovale usually absent),27 and not in contact with rostral tympanic process of petrosal. Anterior limb of ectotympanic suspended directly from basicranium. Stapes usually triangular, with large obturator foramen (microperforate and imperforate stapes occur as rare variants in several species). Fenestra cochleae exposed in most species, but fenestra concealed in sinus formed by caudal and rostral tympanic processes of petrosal in M. rubra. Paroccipital process small, rounded, adnate to petrosal. Dorsal margin of foramen magnum bordered by supraoccipital and exoccipitals, incisura occipitalis present.

Two mental foramina usually present on lateral surface of each hemimandible (one foramen or three foramina occur as rare, usually unilateral variants in some species); angular process acute and strongly inflected.

Unworn crowns of I2–I5 symmetrically rhomboidal (“premolariform”), with subequal anterior and posterior cutting edges, and increasing in length (mesiodistal dimension) from I2 to I5. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple (without accessory cusps in most species) or with posterior accessory cusp only (in M. lepida). First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; second and third upper premolars (P2 and P3) subequal in height; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Molars moderately carnassialized (postmetacristae are visibly longer than postprotocristae); relative widths usually M1 < M2 < M3 < M4; centrocrista strongly inflected labially on M1–M3; ectoflexus indistinct or absent on M1, shallow but usually distinct on M2, and consistently deep on M3; anterolabial cingulum continuous with preprotocrista (complete anterior cingulum present) on M3. Last upper tooth to erupt is P3.

Lower incisors (i1–i4) with distinct lingual cusps. Unworn lower canine (c1) usually semiprocumbent, with flattened bladelike apex, with or without distinct posterior accessory cusp. Second lower premolar (p2) taller than p3; lower milk premolar (dp3) trigonid incomplete (bicuspid). Hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid much taller than hypoconulid on m1–m3.

DISTRIBUTION: Species of Marmosa collectively range from the Mexican state of Tamaulipas southward throughout most of Central and tropical South America to Bolivia, Paraguay, and northern Argentina (Hall, 1981; Creighton and Gardner, 2008b; Gardner and Creighton, 2008b). Although most species inhabit lowland rainforests, a few are restricted to dry habitats (e.g., M. xerophila; Handley and Gordon, 1979) and some species of the subgenus Micoureus occur in montane rainforest at elevations in excess of 2000 m (e.g., M. mapiriensis sensu Tate, 1933: 76).

REMARKS: The monophyly of Marmosa (including Micoureus) is supported by sequence data from five genes analyzed separately (figs. 28–Figure 29Figure 30Figure 3132), in tandem (fig. 33), and in combination with nonmolecular characters (figs. 35, 36); generic monophyly is also supported by a uniquely shared insertion at the BRCA1 locus (fig. 31). Only a single morphological character, the possession of a rostral process of the premaxillae, optimizes as an unambiguous generic synapomorphy (appendix 5), but even this weak phenotypic evidence is compromised by the absence of a rostral process in M. xerophila, a species that we did not score for this study.

No published phylogenetic analysis supports the reciprocal monophyly of Marmosa and Micoureus, both of which have been treated as valid genera by recent authors (e.g., Gardner, 2005; Creighton and Gardner, 2008b; Gardner and Creighton, 2008b). Instead, species of Micoureus have consistently been recovered as nested within a paraphyletic group of Marmosa species (Kirsch and Palma, 1995; Voss and Jansa, 2003; Jansa and Voss, 2005; Steiner et al., 2005; Jansa et al., 2006; Gruber et al., 2007). Obviously, there are several alternative taxonomic solutions to this problem.

One solution would be to treat Micoureus as a junior synonym of Marmosa without recognizing any subgenera of the latter. Another would be to recognize Micoureus as a subgenus of Marmosa and to name new subgenera for other monophyletic clusters of Marmosa species. A third would be to recognize Micoureus as a genus and to describe new genera as needed to make Marmosa monophyletic. Unfortunately, the first option would result in the loss of a useful name (Micoureus, see below), whereas the second and third options are not currently workable because many species of Marmosa have not been included in any phylogenetic analysis, and their relationships are correspondingly obscure. Our interim solution is to move the currently intractable problem of paraphyly from the generic to the subgeneric level. Although taxonomic rank is biologically arbitrary, it affects binomial usage, which should be conformable with phylogenetic relationships insofar as these are known. In effect, because the use of generic names is obligatory under the current Linnaean system, it is crucial that genera be monophyletic.

The monophyly of Micoureus has been supported in every sequencing study to date that has included two or more exemplar species (e.g., Patton et al., 1996; Voss and Jansa, 2003; this study), and it may often be appropriate to indicate this fact in contradistinction to the paraphyly of the subgenus Marmosa. Where appropriate, this can be achieved using double quotes for the latter, as for the species Marmosa (“Marmosa”) lepida as contrasted with Marmosa (Micoureus) demerarae. Predictably, the subgeneric classification of Marmosa will be refined as future studies based on denser taxon sampling yield increasingly resolved estimates of relationships within this speciose group.

Most of the species herein referred to Marmosa have not been revised since Tate (1933), and some currently recognized synonymies are the result of uncritical lumping by subsequent authors (e.g., Hershkovitz, 1951; Cabrera, 1958). Recent analyses of both mtDNA sequence data (e.g., by Patton et al., 2000; Patton and Costa, 2003) and morphological characters (Rossi, 2005) suggest that several nominal taxa currently listed as synonyms (e.g., of M. demerarae, M. murina, and M. robinsoni) are probably valid species. Therefore, significant changes to the species-level taxonomy of Marmosa should be expected soon.

Monodelphis Burnett, 1830

Figure 42

Figure 43

Tlacuatzin canescens (based on primarily on USNM 511261, an adult female from Rancho Sapotito, Nayarit, Mexico; some dental details were reconstructed from USNM 125659, an adult male from Los Reyes, Michoacan, Mexico).

CONTENTS: adusta Thomas, 1897 (including melanops Goldman, 1912); americana Müller, 1776 (including brasiliensis Erxleben, 1777; brasiliensis Daudin, 1802; trilineata Lund, 1840; and tristriata Illiger, 1815); brevicaudata Erxleben, 1777 (including brachyuros Schreber, 1777; dorsalis J.A. Allen, 1904; hunteri Waterhouse, 1841; orinoci Thomas, 1899; sebae Gray, 1827; surinamensis Zimmermann, 1780; touan Bechstein, 1800; touan Shaw, 1800; touan Daudin, 1802; and tricolor E. Geoffroy, 1803); dimidiata Wagner, 1847 (including fosteri Thomas, 1924); domestica Wagner, 1842 (including concolor Gervais, 1856); emiliae Thomas, 1912; glirina Wagner, 1842; handleyi Solari, 2007; iheringi Thomas, 1888; kunsi Pine, 1975; maraxina Thomas, 1923; osgoodi Doutt, 1938; palliolata Osgood, 1914; peruviana Osgood, 1913; reigi Lew and Pérez-Hernández, 2004; ronaldi Solari, 2004; rubida Thomas, 1899; scalops Thomas, 1888; sorex Hensel, 1872 (including henseli Thomas, 1888; itatiayae Miranda-Ribeiro, 1936; lundi Matschie, 1916; and paulensis Vieira, 1950); theresa Thomas, 1921; umbristriatus Miranda-Ribeiro, 1936; and unistriatus Wagner, 1842.

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 70–200 mm; adult weight ca. 15–150 g. Rhinarium with one ventrolateral groove on each side of median sulcus; dark circumocular mask absent; pale supraocular spot absent; dark midrostral stripe absent; throat gland present in adult males of most species but possibly absent in some (e.g., M. theresa). Dorsal pelage coloration highly variable, but dorsal hair bases always dark gray; dorsal guard hairs short and inconspicuous; ventral fur self-colored or gray based, highly variable in surface pigmentation. Manus mesaxonic (dIII > dIV); manual claws very long, extending well beyond fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present, but pads small and dermatoglyphs sometimes indistinct; central palmar epithelium smooth or sparsely tuberculate; carpal tubercles absent in both sexes. Pedal digits unwebbed; pedal digit III longer than digit IV; plantar surface of heel naked. Pouch absent; mammae 4–1–4  =  9 (all abdominal-inguinal; e.g., in M. brevicaudata) to 13–1–13  =  27 (including pectoral teats; e.g., in M. sorex); cloaca present. Tail much shorter than combined length of head and body; thick but muscular, not incrassate; tail conspicuously furred at base to about the same extent dorsally as ventrally (e.g., in M. emiliae), or caudal fur extends farther dorsally than ventrally (e.g., M. brevicaudata), or tail base unfurred (e.g., M. peruviana); unfurred caudal surfaces covered with macroscopic bristlelike hairs, not naked-appearing; caudal scales often inapparent but always in annular series; relationship between caudal scales and hairs usually obscure, but subequal hairs usually arranged in triplets; ventral caudal surface not modified for prehension.

Premaxillary rostral process absent. Nasals long, extending anteriorly beyond I1 (concealing nasal orifice from dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture. Maxillary turbinals (viewed through the nasal orifice) simple or sparsely ornamented scrolls, not elaborately branched. Lacrimal foramina (usually two on each side) prominently exposed on orbital margin or on face anterior to orbit. Orbits small, interorbital region more or less parallel sided (usually without conspicuous constrictions); supraorbital margins smoothly rounded, without beads or distinct postorbital processes (but blunt, indistinct processes occasionally developed in large specimens of some species; e.g., M. emiliae). Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest absent (e.g., in M. theresa) or small (usually not extending to frontals; e.g., in M. brevicaudata). Petrosal not exposed laterally through fenestra in parietal-squamosal suture (fenestra absent). Parietal-mastoid contact usually present (interparietal seldom contacts squamosal).

Maxillopalatine fenestrae present; palatine fenestrae usually absent; maxillary fenestrae absent; posterolateral palatal foramina small, not extending anteriorly between M4 protocones; posterior palatal morphology conforms to Didelphis morphotype (with moderately well-developed lateral corners, the choanae somewhat constricted behind). Maxillary and alisphenoid in contact on floor of orbit (not separated by palatine). Transverse canal foramen present. Alisphenoid tympanic process smoothly globular; posteromedial lamina forming secondary foramen ovale present in some species (e.g., M. theresa) or lamina and secondary foramen ovale absent (e.g., in M. brevicaudata). Anterior limb of ectotympanic suspended directly from basicranium. Stapes triangular with large obturator foramen (e.g., in M. brevicaudata), or columellar and microperforate or imperforate (e.g., in M. peruviana). Fenestra cochleae exposed in most species, but fenestra concealed in sinus formed by rostral and caudal tympanic processes of petrosal in M. emiliae. Paroccipital process small, rounded, and adnate to petrosal. Dorsal margin of foramen magnum bordered by supraoccipital and exoccipitals, incisura occipitalis present.

Two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 symmetrically rhomboidal (“premolariform”), with subequal anterior and posterior cutting edges, and usually increasing in length (mesiodistal dimension) from I2 to I5. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 usually simple (without accessory cusps), but small posterior accessory cusp sometimes present (e.g., in M. peruviana). First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; third upper premolar (P3) taller than P2; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Molars highly carnassialized (postmetacristae much longer than postprotocristae); relative widths consistently M1 < M2 < M3 < M4; centrocrista strongly inflected labially on M1–M3; ectoflexus shallow on M1, deeper on M2, and consistently deep on M3; anterolabial cingulum and preprotocrista discontinuous (anterior cingulum incomplete) on M3. Last upper tooth to erupt is P3 in some species (e.g., M. peruviana), or P3 and M4 erupt simultaneously (e.g., in M. brevicaudata).

Lower incisors (i1–i4) with distinct lingual cusps. Second lower premolar (p2) subequal in height to p3 (e.g., in M. brevicaudata), or p3 taller than p2 (e.g., in M. emiliae); lower milk premolar (dp3) trigonid usually complete (tricuspid). Hypoconid lingual to protoconid (not labially salient) on m3; hypoconulid twinned with entoconid on m1–m3; entoconid subequal to or smaller than hypoconulid on m1–m3.

DISTRIBUTION: Species of Monodelphis occur in lowland and montane rain forests and dry forests from eastern Panama throughout most of tropical and subtropical South America to about 37°S in eastern Argentina (see range maps in Pine and Handley, 2008); recorded elevations range from sea level to at least 2500 m (on the eastern slopes of the tropical Andes). The apparent absence of Monodelphis from the trans-Andean lowlands of western South America is noteworthy, but this is possibly an artifact of inadequate collecting; a representative of the M. adusta complex is rumored to occur along the Pacific littoral of Colombia and northern Ecuador (Solari, 2007), but no specimen-based records have yet been published from that region.

REMARKS: The monophyly of Monodelphis vis-à-vis other Recent didelphids is convincingly supported by 10 nonmolecular characters (appendix 5); by a uniquely shared and unreversed deletion at the BRCA1 locus (fig. 31); and by sequence data from five genes analyzed separately (figs. 28–Figure 29Figure 30Figure 3132), together (fig. 33), and in combination with morphology and karyotypes (figs. 35, 36).

Tlacuatzin Voss and Jansa, 2003

Figure 43

Figure 44

Metachirus nudicaudatus (based on AMNH 267009, an adult male from Paracou, French Guiana).

CONTENTS: canescens J.A. Allen, 1893 (including gaumeri Osgood, 1913; insularis Merriam, 1908; oaxacae Merriam, 1897; and sinaloae J.A. Allen, 1898).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 110–160 mm; adult weight ca. 30–70 g. Rhinarium with two ventrolateral grooves on each side of median sulcus; dark circumocular mask present; pale supraocular spot absent; dark midrostral stripe absent; throat gland absent. Dorsal pelage unpatterned, usually pale gray or grayish brown; dorsal underfur gray; dorsal guard hairs short and inconspicuous; ventral fur self-whitish, -cream, or -buffy. Manus paraxonic (dIII  =  dIV); manual claws about as long as fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium sparsely tuberculate; adult males with well-developed lateral carpal tubercles but not medial carpal tubercles. Pedal digits unwebbed; dIV longer than other pedal digits; plantar surface of heel naked. Pouch absent; mammae 4–1–4  =  9 or 5–1–5  =  11, all abdominal-inguinal; cloaca present. Tail about as long as combined length of head and body, slender and muscular (not incrassate); body pelage present on basal 1/10 or less of tail; naked caudal integument uniformly grayish or faintly bicolored (darker dorsally than ventrally), occasionally parti-colored (with a whitish tip); caudal scales in annular series, each scale with three subequal bristlelike hairs emerging from distal margin; ventral caudal surface modified for prehension distally, with apical pad bearing dermatoglyphs.

Premaxillary rostral process absent. Nasals long, extending anteriorly beyond I1 (concealing nasal orifice from dorsal view), and usually widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. Lacrimal foramina visible in lateral view at or near anterior orbital margin, usually two on each side. Flattened, triangular postorbital processes well developed in fully adult specimens. Left and right frontals and parietals separated by persistent median sutures. Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact. Sagittal crest absent. Petrosal not exposed laterally through fenestra in parietal-squamosal suture (fenestra absent). Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine fenestrae present; palatine fenestrae absent; maxillary fenestrae present, usually opposite M2, sometimes partially confluent with maxillopalatine openings; posterolateral palatal foramina small, not extending anteriorly between M4 protocones; posterior palatal morphology conforms to Didelphis morphotype (with well-developed lateral corners, the choanae constricted behind). Maxillary and alisphenoid not in contact on floor of orbit (separated by palatine). Transverse canal foramen present. Alisphenoid tympanic process smoothly globular, without anteromedial process or posteromedial lamina enclosing extracranial course of mandibular nerve (secondary foramen ovale absent), and not in contact with rostral tympanic process of squamosal. Anterior limb of ectotympanic directly suspended from basicranium. Stapes usually subtriangular and microperforate or imperforate (rarely triangular with a large obturator foramen). Fenestra cochleae exposed, not concealed by rostral and caudal tympanic processes of petrosal. Paroccipital process of exoccipital small, rounded, adnate to petrosal. Dorsal margin of foramen magnum bordered by supraoccipital and exoccipitals, incisura occipitalis present.

Two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 symmetrically rhomboidal (“premolariform”), with subequal anterior and posterior cutting edges, and subequal in length (mesiodistal dimension) from I2 to I5. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps. First upper premolar (P1) present, smaller than posterior premolars but well formed and not vestigial; second and third upper premolars (P2 and P3) subequal in height; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Molars moderately carnassialized (postmetacristae longer than postprotocristae); relative widths usually M1 < M2 < M3 < M4; centrocrista strongly inflected labially on M1–M3; ectoflexus usually absent on M1, shallow on M2, deep and distinct on M3; anterolabial cingulum continuous with preprotocrista (complete anterior cingulum present) on M3; postprotocrista without carnassial notch. Last upper tooth to erupt is P3.

Lower incisors (i1–i4) with distinct lingual cusps. Lower canine (c1) erect, acutely pointed, and simple (without a posterior accessory cusp). Second lower premolar (p2) taller than p3; lower milk premolar (dp3) with incomplete (usually bicuspid) trigonid. Hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid taller than hypoconulid on m1–m2, usually subequal in height to hypoconulid on m3.

DISTRIBUTION: Tlacuatzin is apparently endemic to Mexico, where it occurs in tropical dry 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; Zarza et al., 2003).

Tribe Metachirini 137Hershkovitz, 1992

CONTENTS: Metachirus.

DIAGNOSIS: Metachirini is uniquely distinguished from other members of the subfamily Didelphinae (indeed, from all other didelphids) by contact between the frontal and squamosal on the lateral surface of the braincase.

REMARKS: Although Metachirus could logically be referred to the Didelphini, a new name would then be needed for the clade now awkwardly referred to in the literature as the “large 2n  =  22 opossums” (Chironectes, Didelphis, Lutreolina, Philander). Because many morphological characters unambiguously diagnose the latter group, whereas only a few diagnose the group that includes the large 2n  =  22 opossums plus Metachirus (see appendix 5), the present arrangement seems preferable. Additionally, because the branch leading to Metachirus is a long one (e.g., in fig. 33), we expect that fossils will eventually be found to occupy it, and that such discoveries will minimally disrupt the suprageneric nomenclature if a tribe is already available to accommodate them.

The name Metachirini was credited by Gardner and Dagosto (2008) to Reig et al. (1985), but no family-group name based on Metachirus fulfills the technical criteria for availability (ICZN, 1999: Article 13) prior to Hershkovitz (1992b), who effectively diagnosed Metachirinae in a key. By the Principle of Coordination (ICZN, 1999: Article 36), Metachirini is available with the same authorship and date.

Metachirus Burmeister, 1854

Figure 44

Figure 45

Chironectes minimus (based on AMNH 96759, an adult male from Ilha do Taiuna, Pará, Brazil).

CONTENTS: nudicaudatus E. Geoffroy, 1803 (including antioquiae J.A. Allen, 1916; colombianus J.A. Allen, 1900; bolivianus J.A. Allen, 1901; dentaneus Goldman, 1912; imbutus Thomas, 1923; infuscus Thomas, 1923; modestus Thomas, 1923; myosuros Temminck, 1824; personatus Miranda-Ribeiro, 1936; phaeurus Thomas, 1901; and tschudii J.A. Allen, 1900).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 250–300 mm; adult weight ca. 250–500 g. Rhinarium with one ventrolateral groove on each side of median sulcus; dark circumocular mask present, continuous with dark coronal fur; pale supraocular spot present; dark midrostral stripe absent; throat gland present in adult males. Dorsal pelage unpatterned, usually some shade of grayish or reddish brown; dorsal underfur gray; dorsal guard hairs short and inconspicuous; ventral fur self-whitish, -cream, or -buffy. Manus mesaxonic (dIII > dIV); manual claws shorter than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium smooth or sparsely tuberculate; carpal tubercles absent. Pedal digits unwebbed; dIV longer than other pedal digits; plantar surface of heel naked. Pouch absent; mammae 4–1–4  =  9, all abdominal-inguinal; cloaca present. Tail longer than combined length of head and body, slender and muscular (not incrassate), unfurred except at base; naked caudal integument bicolored basally (brownish or grayish dorsally, paler ventrally), gradually becoming all pale distally; caudal scales in both annular and spiral series (neither pattern predominating), each scale with three subequal bristlelike hairs emerging from distal margin; ventral caudal surface not externally modified for prehension.

Premaxillary rostral process absent. Nasals long, extending anteriorly above or beyond I1 (concealing nasal orifice from dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. Two lacrimal foramina laterally exposed on each side on or just anterior to orbital margin. Interorbital region very broad, with distinctly beaded supraorbital margins; postorbital processes absent. Left and right frontals and parietals separated by persistent median sutures. Frontal and squamosal in contact on lateral braincase (no parietal-alisphenoid contact). Sagittal crest absent or weakly developed just anterior to occiput. Petrosal not exposed laterally through fenestra in parietal-squamosal suture (fenestra absent). Parietal-mastoid contact normally present (interparietal usually does not contact squamosal).

Maxillopalatine fenestrae present; palatine and maxillary fenestrae absent; posterolateral palatal foramina small, not extending anteriorly between M4 protocones; posterior palatal morphology conforms to Didelphis morphotype (with well-developed lateral corners, the choanae constricted behind). Maxillary and alisphenoid usually not in contact on floor of orbit (separated by palatine in most examined specimens). Transverse canal foramen present. Alisphenoid tympanic process laterally compressed (not globular), with anteromedial process enclosing extracranial course of mandibular nerve (secondary foramen ovale present), and not in contact with rostral tympanic process of petrosal. Anterior limb of ectotympanic suspended directly from basicranium. Stapes triangular with large obturator foramen. Fenestra cochleae usually exposed, not or incompletely concealed by rostral and caudal tympanic processes of petrosal. Paroccipital process large, erect, projecting ventrally (not adnate to petrosal). Dorsal margin of foramen magnum bordered by exoccipitals only, incisura occipitalis absent.

Two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 symmetrically rhomboidal (“premolariform”), with subequal anterior and posterior cutting edges, and increasing in length (mesiodistal dimension) from I2 to I5. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps. First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; second and third upper premolars (P2 and P3) subequal in height; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Molars highly carnassialized (postmetacristae much longer than postprotocristae); relative widths M1 < M2 < M3 > M4 or M1 < M2 < M3 < M4; centrocrista strongly inflected labially on M1–M3; ectoflexus usually shallow on M1 and M2 but distinct on M3; anterolabial cingulum and preprotocrista usually discontinuous (anterior cingulum usually incomplete)28 on M3; postprotocrista without carnassial notch. Third upper premolar (P3) and M4 erupt simultaneously.

Lower incisors (i1–i4) with distinct lingual cusps. Lower canine (c1) procumbent, with flattened bladelike anterior margin, but usually without a distinct posterior accessory cusp. Second lower premolar (p2) taller than p3; lower milk premolar (dp3) trigonid complete (tricuspid). Hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid much taller than hypoconulid on m1–m3.

DISTRIBUTION: Metachirus is widely distributed in humid lowland and lower montane forests (usually below 2000 m) from Chiapas, Mexico (Medellín et al., 1992) to northern Argentina (Flores et al., 2007). Published distribution maps (e.g., Reid, 1997; Gardner and Dagosto, 2008) indicate several range disjunctions, but some of these may be artifacts of inadequate collecting effort in suitable habitats.

REMARKS: Although only a single species of Metachirus is currently regarded as valid, recent mtDNA sequencing studies have revealed deep molecular divergence (>10% at the mitochondrial cytochrome b locus; Patton et al., 2000; Patton and Costa, 2003) among supposedly conspecific geographic populations, and it seems likely that several of the names currently listed as junior synonyms of M. nudicaudatus will be resurrected as valid species by future revisionary studies.

Tribe Didelphini Gray, 1821

CONTENTS: Chironectes, Didelphis, Lutreolina, Philander, and †Thylophorops.

DIAGNOSIS: Members of the tribe Didelphini differ from all other members of the subfamily Didelphinae by their possession of a well-developed pouch, black-and-white marked tail, nasal tips that do not extend above I1, co-ossified frontals, well-developed sagittal crest, ectotympanic suspension via the malleus only, delayed eruption of M4, and lower incisors without a distinct lingual cusp.

REMARKS: The monophyly of Didelphini is strongly supported by most phylogenetic analyses published to date (e.g., Kirsch and Palma, 1995; Kirsch et al., 1997; Jansa and Voss, 2000, 2005; Voss and Jansa, 2003, this report; Steiner et al., 2005; Jansa et al., 2006; Gruber et al., 2007). The only exceptions appear to be primitive parsimony (“Wagner”) trees, unaccompanied by support statistics, that were based largely on dental characters (e.g., Reig et al., 1987: figs. 64–67). All of the diagnostic morphological features listed above, together with several other craniodental and karyotypic traits, optimize as unambiguous tribal synapomorphies (appendix 5). The monophyly of Didelphini is additionally supported by a uniquely shared deletion at the DMP1 locus (fig. 29), and by two uniquely shared deletions at the BRCA1 locus (fig. 31).

Chironectes Illiger, 1811

Figure 45

Figure 46

Didelphis marsupialis (based on AMNH 266459, an adult male from Paracou, French Guiana).

CONTENTS: minimus Zimmermann, 1780 (including argyrodytes Dickey, 1928; bresslaui Pohle, 1927; cayennensis Turton, 1800; guianensis Kerr, 1792; gujanensis Link, 1795; langsdorffi Boitard, 1845; palmata Daudin, 1802; panamensis Goldman, 1914; paraguensis Kerr, 1792; sarcovienna Shaw, 1800; variegatus Olfers, 1818; and yapock Desmarest, 1820).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 250–320 mm; adult weight ca. 510–790 g. Rhinarium with one ventrolateral groove on each side of median sulcus; dark circumocular mask present, continuous with dark coronal fur (the whole top of the head except for an indistinct pale bar over each eye is blackish); pale supraocular spot absent; dark midrostral stripe absent; throat gland absent. Dorsal pelage pale silvery gray with blackish transverse bars connected by blackish middorsal stripe (when fresh; old museum skins become brownish or reddish with age); dorsal underfur pale gray; dorsal guard hairs short; ventral fur self-white. Manus mesaxonic (dIII >dIV); manual claws shorter than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads absent (plantar epithelium uniformly and very densely covered with minutely denticulate tubercles among which are scattered at regular intervals many smooth, hemispherical papillae); large, fleshy carpal tubercle supported internally by pisiform present in both sexes. Pedal digits webbed from base to terminal phalanges; dIV longer than other pedal digits; plantar surface of heel naked. Pouch present, opening posteriorly; mammae 2–1–2  =  5; cloaca absent (urogenital and rectal openings widely separated by furred skin). Tail longer than combined length of head and body, thick and muscular (not incrassate); body fur extends onto basal one-sixth or less of tail, to about the same extent dorsally as ventrally; naked caudal integument (beyond furry base) blackish with abruptly whitish tip (rarely all blackish); caudal scales in spiral series, each dorsal scale usually with three subequal bristlelike hairs emerging from distal margin (ventral scales often have 4–5 hairs each); ventral caudal surface not externally modified for prehension.

Premaxillary rostral process absent. Nasals short, not extending anteriorly above I1 (exposing nasal orifice in dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. One lacrimal foramen usually present on each side just outside anterior orbital margin. Blunt, hornlike postorbital frontal processes present in adults. Right and left frontals co-ossified, midfrontal suture incomplete or absent; left and right parietals separated by persistent midparietal suture. Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest present, well developed on parietals, and extending anteriorly onto frontals. Petrosal not laterally exposed through fenestra in parietal-squamosal suture (fenestra absent). Parietal-mastoid contact absent (interparietal narrowly contacts squamosal).

Maxillopalatine fenestrae present; palatine and maxillary fenestrae absent; posterolateral palatal foramina not extending anteriorly between M4 protocones; posterior palatal morphology conforms to Didelphis morphotype (with well-developed lateral corners, the choanae constricted behind). Maxillary and alisphenoid usually not in contact on floor of orbit (uni- or bilateral contacts were observed as rare variants). Transverse canal foramen present. Alisphenoid tympanic process small and uninflated, usually with medial lamina enclosing extracranial course of mandibular nerve (secondary foramen ovale usually present), and not in contact with rostral tympanic process of petrosal. Anterior limb of ectotympanic suspended indirectly from basicranium (by malleus). Stapes triangular, with large obturator foramen. Fenestra cochleae exposed, not concealed by rostral and caudal tympanic processes of petrosal. Paroccipital process large and erect, not adnate to petrosal. Dorsal margin of foramen magnum bordered by supraoccipital and exoccipitals, incisura occipitalis present but narrow.

One mental foramen present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 symmetrically rhomboidal (“premolariform”), with subequal anterior and posterior cutting edges, and subequal in length (mesiodistal dimension) from I2 to I5. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps. First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; third upper premolar (P3) taller than P2; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Upper molars highly carnassialized (postmetacristae much longer than postprotocristae); relative widths M1 < M2 < M3 > M4 or M1 < M2 < M3 < M4; centrocrista only weakly inflected labially on M1–M3; ectoflexus shallow but distinct on M1 and M2, deepest on M3; anterolabial cingulum and preprotocrista discontinuous (anterior cingulum incomplete) on M3; postprotocrista with carnassial notch. Last upper tooth to erupt is M4.

Lower incisors (i1–i4) without distinct lingual cusps. Lower canine (c1) erect, acutely pointed, and simple (without a posterior accessory cusp). Second lower premolar (p2) taller than p3; lower milk premolar (dp3) large and molariform with a complete (tricuspid) trigonid; hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid much taller than hypoconulid on m1–m3.

DISTRIBUTION: Chironectes occurs along streams in moist lowland and lower montane forests (the highest elevational record appears to be about 1900 m; Handley, 1976) from the Mexican state of Oaxaca throughout most of Central America and tropical South America to northeastern Argentina (Misiones) and Uruguay (González and Fregueiro, 1998). As mapped by Marshall (1978d), the geographic range of this genus is strikingly disjunct, but collecting localities subsequently reported by Anderson (1997), Linares (1998), and Brown (2004) have closed many distributional gaps. As mapped by Stein and Patton (2008a), Chironectes appears to be absent from central Amazonia, but an unvouchered record from the lower Tapajos (George et al., 1988) suggests that this hiatus may be an artifact of inadequate collecting. Although water opossums are known to inhabit gallery-forested streams in the Cerrado (Mares et al., 1989), there are no published records from the Caatinga and Chaco.

REMARKS: Given the very broad geographic distribution of this unrevised genus, it seems probable that at least some of the many putative synonyms of Chironectes minimus will be found to represent valid taxa in future analyses of morphological and/or molecular data.

Didelphis Linnaeus, 1758

Figure 46

Figure 47

Lutreolina crassicaudata (based AMNH 254513, an adult female from Punta Lara, Buenos Aires, Argentina).

CONTENTS: albiventris Lund, 1840 (including bonariensis Marelli, 1930; dennleri Marelli, 1930; lechei Ihering, 1892; leucotis Wagner, 1847; paraguayensis J.A. Allen, 1902; poecilotis Wagner, 1842; and poecilonota Schinz, 1844); aurita Wied-Neuweid, 1826 (including koseritzi Ihering, 1892; longipilis Miranda-Ribeiro, 1935; and melanoidis Miranda-Ribeiro, 1935); imperfecta Mondolfi and Pérez-Hernández, 1984; marsupialis Linnaeus, 1758 (including battyi Thomas, 1902; cancrivora Gmelin, 1788; caucae J.A. Allen, 1900; colombica J.A. Allen, 1900; etensis J.A. Allen, 1902; insularis J.A. Allen, 1902; karkinophaga Zimmermann, 1780; particeps Goldman, 1917; richmondi J.A. Allen, 1901; and tabascensis J.A. Allen, 1901); pernigra J.A. Allen, 1900 (including andina J.A. Allen, 1902; and meridensis J.A. Allen, 1902); and virginiana Kerr, 1792 (including boreoamericana J.A. Allen, 1902; breviceps Bennett, 1833; californica Bennett, 1833; cozumelae Merriam, 1901; illinensium Link, 1795; pigra Bangs, 1898; pilosissima Link, 1795; pruinosa Wagner, 1843; texensis J.A. Allen, 1901; woapink Barton, 1806; and yucatanensis J.A. Allen, 1901).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body 310–495 mm; adult weight 600–5100 g. Rhinarium with one ventrolateral groove on each side of median sulcus; dark circumocular mask indistinct in some species (D. aurita, D. marsupialis, D. virginiana) but distinct in others (D. albiventris, D. imperfecta, D. pernigra); pale supraocular spot absent; dark midrostral stripe absent; throat gland absent. Dorsal pelage unpatterned; dorsal underfur white; dorsal guard hairs long, coarse and conspicuous, giving the pelage a distinctively shaggy appearance; ventral fur pale (usually whitish) tipped with black. Manus mesaxonic (dIII > dIV); manual claws usually longer than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium smooth or sparsely tuberculate; carpal tubercles absent. Pedal digits unwebbed; dIV slightly longer than other pedal digits in some species (e.g., D. marsupialis) or dII, dIII, and dIV subequal (e.g., in D. albiventris and D. virginiana); plantar surface of heel naked. Pouch well developed, opening anteriorly; mammae normally 3–1–3  =  7 to 6–1–6  =  13 or more29; cloaca present. Tail slightly longer than combined length of head and body, slender and muscular (not incrassate); basal 1/6 to 1/4 densely furred (covered with body pelage); naked caudal integument blackish proximally and abruptly whitish distally; caudal scales in spiral series, each scale usually with three subequal bristlelike hairs emerging from distal margin; ventral caudal surface variably modified for prehension distally (prehensile modifications are more strongly developed in D. auritus and D. marsupialis than in the other species), but dermatoglyph-bearing apical pad consistently present.

Premaxillary rostral process absent. Nasals short, not produced anteriorly above I1 (exposing nasal orifice in dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. Lacrimal foramina one or two on each side, exposed laterally on orbital margin or on face just anterior to orbit. Postorbital processes bluntly pyramidal, maximally developed in old adults. Left and right frontals co-ossified (midfrontal suture incomplete or absent), but left and right parietals separated by persistent midparietal suture. Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest large, well developed, and extending onto frontals. Petrosal not exposed laterally through fenestra in squamosal-parietal suture (fenestra absent). Parietal-mastoid contact absent (interparietal narrowly contacts squamosal).

Maxillopalatine and palatine fenestrae present; maxillary fenestrae absent; posterolateral palatal foramina not extending anteriorly lingual to M4 protocones; posterior palate with prominent lateral corners, the choanae abruptly constricted behind. Maxillary and alisphenoid not in contact (separated by palatine) on floor of orbit. Transverse canal foramen present. Alisphenoid tympanic process small and uninflated, usually with posteromedial lamina enclosing extracranial course of mandibular nerve (secondary foramen ovale usually present), and not in contact with rostral tympanic process of petrosal. Anterior limb of ectotympanic suspended indirectly from basicranium (by malleus). Stapes usually triangular with large obturator foramen but sometimes columellar and imperforate (varies within species). Fenestra cochleae exposed, not concealed by rostral and caudal tympanic processes of petrosal. Paroccipital process large, erect (not adnate to petrosal) and projecting ventrally. Dorsal margin of foramen magnum bordered by exoccipitals only, incisura occipitalis absent.

Two mental foramina usually present on lateral surface of each hemimandible (one foramen is present unilaterally in a single examined specimen of D. albiventris); mandibular angular process acute and strongly inflected medially.

Unworn crowns of I2–I5 asymmetrical (“incisiform”), with much longer anterior than posterior cutting edges. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps. First upper premolar (P1) smaller than posterior premolars, but well-formed and not vestigial; third upper premolar (P3) taller than second (P2); P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Molars highly carnassialized (postmetacristae conspicuously longer than postprotocristae); relative widths M1 < M2 < M3 > M4 or M3 < M4; centrocrista only weakly inflected labially on M1–M3; ectoflexus shallow, indistinct, or absent on M1 and M2, but consistently deep and distinct on M3; anterolabial cingulum and preprotocrista discontinuous (anterior cingulum incomplete) on M3; postprotocrista with carnassial notch. Last upper tooth to erupt is M4.

Lower incisors (i1–i4) without distinct lingual cusps. Second lower premolar (p2) much taller than p3; lower milk premolar (dp3) with complete (tricuspid) trigonid. Hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid much taller than hypoconulid on m1–m3.

DISTRIBUTION: Didelphis is a quintessentially eurytopic genus that ranges from southern Canada throughout most of North America (except the Rockies, the northern Great Plains, the Great Basin, the arid southwest, north-central Mexico, and Baja California; Hall, 1981), all of Central America, and most of South America (to about 40°S latitude in central Argentina; Flores et al., 2007). South American collection records (mapped by Cerqueira and Tribe, 2008) represent almost every non-desert tropical and subtropical biome on the continent.

REMARKS: Despite contradictory or ambiguous phylogenetic results from most sequenced loci analyzed separately (table 14), the monophyly of Didelphis is strongly supported by parsimony and Bayesian analyses of morphology (fig. 27), a concatenated five-gene dataset (fig. 33), and combined (nonmolecular + molecular) supermatrices (figs. 35, 36). Revisionary studies of the genus Didelphis in its modern sense were initiated by Allen (1901, 1902) and continued by Krumbiegel (1941), Gardner (1973), Cerqueira (1985), and Lemos and Cerqueira (2002). Useful summaries of morphological characters that distinguish Didelphis species in local faunas are provided by Mondolfi and Pérez-Hernández (1984), Catzeflis et al. (1997), Cerqueira and Lemos (2000), Flores and Abdala (2001), and Ventura et al. (2002). Analyses of mitochondrial DNA sequence variation within and among South American species were reported by Patton et al. (2000) and Patton and Costa (2003).

Lutreolina Thomas, 1910

Figure 47

Figure 48

Philander opossum (based primarily on AMNH 266387, an adult female from Paracou, French Guiana; minor details were reconstructed from AMNH 266386 and 266994).

CONTENTS: crassicaudata Desmarest, 1804 (including crassicaudis Olfers, 1818; bonaria Thomas, 1923; ferruginea Larrañaga, 1923; lutrilla Thomas, 1923; macroura Desmoulins, 1824; paranalis Thomas, 1923; travassosi Mirando-Ribeiro, 1936; and turneri Günther, 1879).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 240–345 mm; adult weight ca. 300–800 g). Rhinarium with one ventrolateral groove on each side of median sulcus; head entirely pale and unmarked (dark circumocular mask, pale supraocular spots, and dark midrostral stripe absent); throat gland absent. Dorsal pelage unpatterned, usually some shade of yellowish brown with grayish hair bases; dorsal guard hairs short and inconspicuous; ventral fur gray based or self-buffy. Manus mesaxonic (dIII > dIV); manual claws much longer than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium more or less smooth; carpal tubercles absent. Pedal digits unwebbed; dIII longer than other pedal digits; plantar surface of heel naked. Pouch present, opening posteriorly; mammae 4–1–4  =  9 to 5–1–5  =  11; cloaca present. Tail usually slightly shorter than combined length of head and body, thicker than in other large opossums but muscular, not incrassate; furred to about the same extent dorsally and ventrally for basal one-third to one-half; naked caudal integument blackish proximally but abruptly whitish near tail tip; caudal scales in spiral series, each scale usually with four or five subequal bristlelike hairs emerging from distal margin; ventral caudal surface not externally modified for prehension.

Premaxillary rostral process absent. Nasals short, not extending anteriorly above I1 (exposing nasal orifice in dorsal view), and widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. Lacrimal foramina two on each side, exposed laterally on orbital margin or on face just anterior to orbit. Short, blunt, hornlike postorbital processes usually present in large adult specimens. Left and right frontals co-ossified (midfrontal suture incomplete or absent), but left and right parietals separated by persistent midparietal suture. Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest present, well developed on parietals and extending anteriorly onto frontals. Petrosal not laterally exposed through fenestra in squamosal-parietal suture (fenestra absent). Parietal-mastoid contact absent (interparietal narrowly contacts squamosal).

Maxillopalatine and palatine fenestrae present; maxillary fenestrae absent; posterolateral palatal foramina not extending anteriorly between M4 protocones; posterior palatal morphology conforming to Didelphis morphotype (with well-developed lateral corners, the choanae abruptly constricted behind). Maxillary and alisphenoid in contact (overlying palatine) on floor of orbit. Transverse canal foramen present. Alisphenoid tympanic process small and usually rounded (somewhat more inflated than in Chironectes or Didelphis), with broad lamina enclosing extracranial course of mandibular nerve (secondary foramen ovale present), and not in contact with rostral tympanic process of petrosal. Anterior limb of ectotympanic suspended indirectly from braincase (by malleus). Stapes triangular with large obturator foramen; fenestra cochleae exposed (not enclosed by bony laminae). Paroccipital process large, erect, directed posteroventrally. Dorsal margin of foramen magnum bordered by exoccipitals only, incisura occipitalis absent.

Two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 asymmetrical (“incisiform”), with much longer anterior than posterior cutting edges. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps. First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; third upper premolar (P3) taller than P2; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Upper molars highly carnassialized (postmetacristae conspicuously longer than postprotocristae); relative widths M1 < M2 < M3 < M4; centrocrista only weakly inflected labially on M1–M3; ectoflexus usually distinct only on M3; anterolabial cingulum and preprotocrista discontinuous (anterior cingulum incomplete) on M3; postprotocrista with carnassial notch. Last upper tooth to erupt is M4.

Lower incisors (i1–i4) without distinct lingual cusps. Lower canine (c1) erect, acutely pointed, and simple (without a posterior accessory cusp). Second lower premolar (p2) taller than p3; lower milk premolar (dp3) large and molariform with complete (tricuspid) trigonid. Hypoconid lingual to protoconid (not labially salient) on m3; hypoconulid twinned with entoconid on m1–m3; entoconid taller than hypoconulid on m1 and m2, but sometimes subequal in height to hypoconulid on m3.

DISTRIBUTION: The known distribution of Lutreolina consists of two disjunct regions of nonforest vegetation in South America (Stein and Patton, 2008b). Specimens traditionally referred to the subspecies L. crassicaudata turneri are from widely scattered localities in the savanna lowlands (below 900 m) in eastern Colombia and central Venezuela, and from the isolated but adjacent savannas of southern Venezuela and northern Guyana (Voss, 1991: 91); although specimens of this form are currently unknown from Surinam and French Guiana, they could be expected to occur in the as yet poorly surveyed coastal savannas of either (or both) countries and in the adjacent Brazilian state of Amapá. Southern South American specimens (representing several nominal taxa in addition to the nominotypical subspecies L. c. crassicaudata) are much more commonly collected in a wide range of tropical, subtropical, and temperate habitats in Paraguay, Uruguay, eastern Bolivia, southern Brazil, and northern Argentina (Flores et al., 2007; Stein and Patton, 2008b).

REMARKS: Although several morphologically diagnosed subspecies of Lutreolina have been treated as valid by authors (e.g., Thomas, 1923; Ximénez, 1967; Graipel et al., 1996; Stein and Patton, 2008b), no published analysis of molecular variation has tested the implicit assumption that only a single species occurs across the vast range of South American landscapes from which specimens have been collected.

Philander Brisson, 1762

Figure 48

Figure 49

Cryptonanus unduaviensis (based on AMNH 209152, an adult female from the Río Baures, Beni, Bolivia).

CONTENTS: andersoni Osgood, 1913 (including nigratus Thomas, 1923); deltae Lew et al., 2006; frenatus Olfers, 1818 (including azaricus Thomas, 1923; quica Temminck, 1824; and superciliaris Olfers, 1818); mcilhennyi Gardner and Patton, 1972; mondolfii Lew et al., 2006; olrogi Flores et al., 2008; and opossum Linnaeus, 1758 (including canus Osgood, 1913; crucialis Thomas, 1923; fuscogriseus J.A. Allen, 1900; grisescens J.A. Allen, 1901; melantho Thomas, 1923; melanurus Thomas, 1899; pallidus J.A. Allen, 1901; and virginianus Tiedemann, 1808).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 250–350 mm; adult weight ca. 280–700 g. Rhinarium with one ventrolateral groove on each side of median sulcus; dark circumocular mask present, usually continuous with dark coronal fur (the coronal fur is not dark in some examined specimens of P. frenatus); pale supraocular spot present; dark midrostral stripe absent; throat gland absent. Dorsal pelage unpatterned grayish or blackish, or with grayish flanks and black middorsal stripe; dorsal underfur gray; dorsal guard hairs usually short (longer middorsally than along flanks in P. mcilhennyi); ventral fur variously pigmented, usually gray-based buffy or cream, or entirely grayish (variable within and among examined taxa). Manus mesaxonic (dIII > dIV); manual claws about as long as fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium smooth or sparsely tuberculate; carpal tubercles absent. Pedal digits unwebbed; dIV longer than other pedal digits; plantar surface of heel naked. Pouch present, opening anteriorly; mammae usually 2–1–2  =  5 or 3–1–3  =  7; cloaca present. Tail longer than combined length of head and body, slender and muscular (not incrassate); furred dorsally and ventrally to about the same extent for basal one-fifth; naked caudal integument blackish proximally and abruptly whitish distally; caudal scales in spiral series, each scale with 4–6 bristlelike hairs emerging from distal margin; ventral caudal surface modified for prehension distally, with apical pad bearing dermatoglyphs.

Premaxillary rostral process absent. Nasals short, not extending anteriorly above I1 (exposing nasal orifice in dorsal view), and widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. Lacrimal foramina usually two on each side, exposed laterally on orbital margin or on face just anterior to orbit. Interorbital region smoothly rounded, without supraorbital beads or crests; short, blunt, hornlike postorbital processes usually present in large adult specimens. Left and right frontals co-ossified (midfrontal suture incomplete or absent), but left and right parietals separated by persistent midparietal suture. Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest present, well developed on parietals and extending anteriorly onto frontals. Petrosal not exposed laterally through fenestra in squamosal-parietal suture (fenestra absent). Parietal-mastoid contact absent (interparietal narrowly contacts squamosal).

Maxillopalatine and palatine fenestrae present; maxillary fenestrae absent; posterolateral palatal foramina small, not extending anteriorly between M4 protocones; posterior palatal morphology conforming to Didelphis morphotype (with prominent lateral corners, the choanae constricted behind). Maxillary and alisphenoid usually separated by palatine on floor of orbit (maxillary-alisphenoid contact occurs unilaterally or bilaterally in a few specimens). Transverse canal foramen usually present. Alisphenoid tympanic process small and uninflated, usually with broad lamina enclosing extracranial course of mandibular nerve (secondary foramen ovale present), and not contacting rostral tympanic process of petrosal. Anterior limb of ectotympanic indirectly suspended from basicranium (by malleus). Stapes usually triangular with large obturator foramen. Fenestra cochleae exposed (not concealed by rostral and caudal tympanic processes of petrosal). Paroccipital process large, erect, directed posteroventrally. Dorsal margin of foramen magnum bordered by exoccipitals only, incisura occipitalis absent.

Two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 asymmetrical (“incisiform”), with much longer anterior than posterior cutting edges. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps. First upper premolar (P1) smaller than posterior premolars but well-formed and not vestigial; third upper premolar (P3) taller than P2; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Molars highly carnassialized (postmetacristae conspicuously longer than postprotocristae; relative widths M1 < M2 < M3 < M4; centrocrista only weakly inflected labially on M1–M3; ectoflexus usually distinct only on M3; anterolabial cingulum and preprotocrista discontinuous (anterior cingulum incomplete) on M3; postprotocrista with carnassial notch. Last upper tooth to erupt is M4.

Lower incisors (i1–i4) without distinct lingual cusps. Lower canine (c1) erect, acutely pointed, and simple (without a posterior accessory cusp). Second lower premolar (p2) much taller than p3; lower milk premolar (dp3) large and molariform with complete (tricuspid) trigonid. Hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid much taller than hypoconulid on m1–m3.

DISTRIBUTION: Species of Philander occur in lowland and lower montane moist forests (to 1600 m; Reid, 1997) from the Mexican state of Tamaulipas southward throughout most of Central America (Hall, 1981) and tropical South America (Patton and da Silva, 2008) to northern Argentina (Chaco, Formosa, and Misiones; Flores et al., 2007). Collecting localities mapped by Anderson (1997), Linares (1998), and Brown (2004) suggest that Philander is largely absent from intervening arid, semiarid, or treeless landscapes (e.g., the Chaco, Cerrado, Caatinga, and Llanos), with a few exceptions that may have been taken in gallery forests.

REMARKS: Although inconsistently supported by our separate nuclear-gene analyses (table 14) and by Patton et al.'s (1996) analyses of sequence data from the mitochondrial cytochrome b locus, the monophyly of Philander is moderately well supported by nonmolecular characters (fig. 27) and it is strongly supported by our analyses of concatenated genes and combined datasets (figs. 33, 35, 36). The current species-level classification given above largely follows Patton and da Silva (1997), who analyzed mtDNA haplotype diversity among 42 Central and South American specimens that they interpreted as representing six valid taxa (andersoni, canus, frenatus, fuscogriseus, mcilhennyi, opossum); three unsequenced taxa (azarica, melanurus, pallidus) were also recognized as valid on the basis of morphological traits. Whereas andersoni, frenatus, and mcilhennyi were recovered as monophyletic haplotype groups that Patton and da Silva recognized as monotypic species, their concept of opossum included several subspecies (azarica, canus, fuscogriseus, melanurus, pallida) and was not recovered as a clade. Based on their phylogenetic analysis (op. cit.: fig. 2), it would be logical to recognize fuscogriseus and canus as valid species, but the authors reasonably cautioned against facile taxonomic decisions based on the results of analyzing a single maternally inherited gene. Indeed, the genus clearly requires a more comprehensive revisionary treatment, with particular attention devoted to obtaining analyzable character data from hitherto unrepresented taxa (notably azarica, crucialis, grisescens, melanurus, and pallidus). Our placement of azarica in the synonymy of P. frenatus uncritically follows Gardner (2005), who also allocated nominal taxa not treated by Patton and da Silva (1997) to the synonymy of P. opossum. Although an alternative species-level classification of Philander proposed by Hershkovitz (1997) was not based on any explicit analysis of specimen data, it raised several issues not adequately treated by other authors and drew attention to the possible existence of undescribed taxa. New species described by Lew et al. (2006) and Flores et al. (2008) contribute additional taxonomic complexities and underscore the need for critical revisionary work on this difficult genus.

The tiresome nomenclatural controversy regarding the use of Philander Tiedemann, 1808, versus Metachirops Matschie, 1916, as the correct generic name for the gray four-eyed opossums (Pine, 1973; Hershkovitz, 1976; Gardner, 1981) was made moot by a recent ruling of the International Commission on Zoological Nomenclature (ICZN, 1998) that definitively established the availability of Philander from Brisson (1762).

Tribe Thylamyini 137Hershkovitz, 1992

CONTENTS: Chacodelphys, Cryptonanus, Gracilinanus, Lestodelphys, Marmosops, and Thylamys.

DIAGNOSIS: Members of the tribe Thylamyini differ from all other didelphids by having a fenestra in the parietal-squamosal suture through which the petrosal is visible on the lateral surface of the braincase. Additionally, most thylamyines (Chacodelphys and Cryptonanus are exceptions) differ from otherwise similar marmosines by having a secondary foramen ovale formed by an anteromedial bullar process that spans the transverse canal foramen.

REMARKS: The monophyly of Thylamyini (without Chacodelphys, from which sequence data are unavailable) is strongly supported by parsimony, likelihood, and Bayesian analyses of IRBP (fig. 28) and BRCA1 (fig. 31); by likelihood and Bayesian analyses of DMP1 (fig. 29); and by Bayesian analysis of RAG1 (fig. 30). It is also strongly supported by all analyses of a concatenated five-gene dataset (fig. 33), and by Bayesian analysis of a combined (nonmolecular + molecular) dataset that includes Chacodelphys. The presence of a fenestra in the parietal-squamosal suture (appendix 5) and a uniquely derived insertion at the BRCA1 locus (fig. 31) both optimize as thylamyine synapomorphies.

Apparently, the earliest family-group name based on Thylamys is technically available from Hershkovitz (1992b), who erroneously attributed authorship of Thylamyinae to Reig et al. (1987).

Chacodelphys Voss et al., 2004

CONTENTS: formosa Shamel, 1930 (including muscula Shamel, 1930).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body 68 mm (external measurements are only available from the young adult holotype); adult weight unknown but probably ca. 10 g. Rhinarial morphology unknown (no fluid-preserved specimens have been examined); dark circumocular mask present but narrow and inconspicuous; pale supraocular spot absent; dark midrostral stripe absent; gular gland present. Dorsal fur unpatterned, brownish, somewhat darker middorsally than along flanks, but pelage not distinctly tricolored; dorsal underfur gray based; dorsal guard hairs short and inconspicuous; ventral fur gray-based buffy. Manus mesaxonic (dIII > dIV); manual claws shorter than fleshy apical pads of digits; manual plantar pads present, but presence/absence of dermatoglyphs unknown; central palmar surface of manus densely covered with small convex tubercles; occurrence of carpal tubercles unknown. Pedal digits unwebbed; dIV slightly longer than other pedal digits; plantar surface of heel naked. Mammary formula, morphology of pouch (if any), and presence/absence of cloaca unknown. Tail shorter than combined length of head and body, slender and muscular (not incrassate); body pelage not extending onto tail base; tail densely covered with short hairs and distinctly bicolored (dark above, pale below); caudal scales arranged in annular series, each scale with three subequal bristlelike hairs emerging from distal margin; ventral caudal surface not externally modified for prehension.

Rostral process of premaxillae absent. Nasals long, extending anteriorly beyond I1 (concealing nasal orifice from dorsal view), with subparallel lateral margins (not widened posteriorly). Maxillary turbinals large and elaborately branched. Two lacrimal foramina laterally exposed on each side just anterior to orbital margin. Supraorbital margins smoothly rounded, without beads or processes; strongly marked interorbital and postorbital constrictions present. Right and left frontals and parietals separated by persistent median sutures. Parietal and alisphenoid bones in contact (no squamosal-frontal contact). Sagittal crest absent. Petrosal exposed laterally though fenestra in parietal-squamosal suture. Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine fenestrae present and very large; palatine fenestrae present but incompletely separated from maxillopalatine openings; maxillary fenestrae very small but bilaterally present near M1/M2 commissure; posterolateral palatal foramina small, not extending lingual to M4 protocones; posterior palate conforms to Didelphis morphotype (with prominent lateral corners, the internal choanae abruptly constricted behind). Maxillary and alisphenoid not in contact on orbital floor (separated by palatine). Transverse canal foramen present. Alisphenoid tympanic process probably smoothly globular (broken in both examined specimens), without anteromedial process or posteromedial lamina enclosing extracranial course of mandibular nerve (secondary foramen ovale absent), and probably not contacting rostral tympanic process of petrosal. Anterior limb of ectotympanic directly suspended from basicranium. Stapes triangular, perforated by large obturator foramen. Fenestra cochleae exposed, not concealed by rostral and caudal tympanic processes of petrosal. Paroccipital process small, adnate to petrosal. Dorsal margin of foramen magnum formed by supraoccipital and exoccipitals, incisura occipitalis present.

Two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 symmetrically rhomboidal (“premolariform”), with subequal anterior and posterior cutting edges, and increasing in length (mesiodistal dimension) from I2 to I5. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps. First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; second and third upper premolars (P2 and P3) subequal in height (see footnote 14, above); P3 with posterior cutting edge only; upper milk premolar (dP3) morphology unknown (no juvenile specimens examined). Upper molars highly carnassialized (postmetacristae conspicuously longer than postprotocristae); relative widths M1 < M2 < M3 < M4; centrocrista strongly inflected labially on M1–M3; ectoflexus absent on M1, very shallow on M2, distinct only on M3; anterolabial cingulum and preprotocrista discontinuous (anterior cingulum incomplete) on M3; postprotocrista without carnassial notch. Last upper tooth to erupt is probably P3.

Lower incisors (i1–i4) with distinct lingual cusps. Lower canine (c1) semiprocumbent, with flattened bladelike apex, but simple (without a distinct posterior accessory cusp). Second lower premolar (p2) taller than p3; lower milk premolar (dp3) morphology unknown (no juvenile specimens examined). Hypoconid lingual to protoconid (not labially salient) on m3; hypoconulid twinned with entoconid on m1–m3; entoconid very small, subequal in height to hypoconulid on m1–m3.

DISTRIBUTION: Chacodelphys is currently known from just five localities in the Argentinian provinces of Chaco and Formosa (between 25–27° S latitude and 58–60° W longitude; Teta et al., 2006). However, because the humid savanna habitats that occur at these sites resemble those found elsewhere in northern Argentina, eastern Bolivia, western Paraguay, and southwestern Brazil, it would be reasonable to expect that the genus is more widely distributed.

REMARKS: In the absence of undamaged cranial material, we are unable to provide an illustration for this genus. However, photographs of two imperfect skulls were published by Voss et al. (2004a: fig. 2) and Teta et al. (2006: fig. 2).

The problematic relationships of this genus were discussed at length by Voss et al. (2004a) and seem unlikely to be resolved definitively in the absence of nuclear gene sequence data. As discussed above, the tribal membership of Chacodelphys is most compellingly supported by Bayesian analysis of a dataset that combines nonmolecular characters with nuclear-gene sequence data (fig. 36), but it is consistent with subjective assessments of similarity based on integumental and craniodental traits (e.g., by Tate, 1933). Although separate analyses of nonmolecular character data suggest that this genus may be more closely related to Thylamys and Lestodelphys than to other thylamyines (fig. 27), it seems premature to formalize such weakly supported results by subtribal nomenclature.

Cryptonanus Voss et al., 2005

Figure 49

Figure 50

Gracilinanus agilis (based on MVZ 197437, an adult male from Mato do Vasco, Minas Gerais, Brazil).

CONTENTS: agricolai Moojen, 1943; chacoensis Tate, 1931; guahybae Tate, 1931; ignitus Díaz et al., 2002; and unduaviensis Tate, 1931.

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 80–120 mm; adult weight ca. 15–40 g. Ventral margin of rhinarium with two shallow grooves on each side of median sulcus; dark circumocular mask present; pale supraocular spot absent; dark midrostral stripe absent; gular gland present in adult males. Dorsal body pelage unpatterned, usually grayish or reddish brown; dorsal fur gray based; dorsal guard hairs very short and inconspicuous; ventral fur gray based or self-colored (varying among species). Manus paraxonic (dIII  =  dIV); manual claws shorter than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar surface of manus sparsely tubercular (neither smooth nor densely covered with convex tubercles); lateral carpal tubercles present in adult males. Pedal digits unwebbed; dIV longer than other pedal digits; plantar epithelium of heel naked. Pouch absent; mammae 4–1–4  =  9 (all abdominal-inguinal) to 7–1–7  =  15 (with pectoral teats); cloaca present. Tail longer than combined length of head and body, slender and muscular (not incrassate); body pelage not extending more than a few mm onto tail base; unfurred caudal integument more or less bicolored (dark above, paler below) in most specimens; caudal scales in distinctly annular series, each scale with three subequal bristlelike hairs emerging from distal margin; ventral caudal surface modified for prehension distally, with apical pad bearing dermatoglyphs.

Rostral process of premaxillae absent. Nasals long, extending anteriorly beyond I1 (concealing nasal orifice from dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture. Maxillary turbinals large and elaborately branched. Two lacrimal foramina laterally exposed on each side on or just anterior to orbital margin. Supraorbital margins rounded, without beads or processes (a few old individuals have incipient postorbital processes); distinct interorbital and postorbital constrictions usually present in juveniles and young adults. Left and right frontals and parietals separated by persistent median sutures. Parietal and alisphenoid in contact on lateral braincase (no squamosal-frontal contact). Sagittal crest absent. Petrosal laterally exposed through fenestra in parietal-squamosal suture. Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine fenestrae large; palatine fenestrae present; maxillary fenestrae absent; posterolateral palatal foramina small, not extending lingual to M4 protocones; posterior palate conforms to Didelphis morphotype (with prominent lateral corners, the internal choanae abruptly constricted behind). Maxillary and alisphenoid not in contact on orbital floor (separated by palatine). Transverse canal foramen present. Alisphenoid tympanic process smoothly globular, without anteromedial process or posteromedial lamina enclosing extracranial course of mandibular nerve (secondary foramen ovale absent), and not contacting rostral tympanic process of petrosal. Anterior limb of ectotympanic directly suspended from basicranium. Stapes triangular, perforated by large obturator foramen. Fenestra cochleae exposed, not concealed by rostral and caudal tympanic processes of petrosal. Paroccipital process small, adnate to petrosal. Dorsal margin of foramen magnum formed by supraoccipital and exoccipitals, incisura occipitalis present.

Two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 symmetrically rhomboidal (“premolariform”), with subequal anterior and posterior cutting edges, slightly increasing in length (mesiodistal dimension) from I2 to I5. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 usually with one or two small accessory cusps (the posterior accessory cusp is more consistently distinct than the anterior cusp). First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; third upper premolar (P3) taller than P2; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Upper molars strongly carnassialized (postmetacristae conspicuously longer than postprotocristae); relative widths M1 < M2 < M3 > M4 or M1 < M2 < M3 < M4; centrocrista strongly inflected labially on M1–M3; ectoflexus shallow or absent on M1, deeper on M2, and consistently deep on M3; anterolabial cingulum continuous with preprotocrista (complete anterior cingulum present) on M3 in some species (e.g., C. unduaviensis) but anterolabial cingulum and preprotocrista discontinuous (anterior cingulum incomplete) on M3 in others (e.g., C. chacoensis); postprotocrista without carnassial notch. Last upper tooth to erupt is P3.

Lower incisors (i1–i4) with distinct lingual cusps. Lower canine (c1) procumbent, with flattened bladelike apex, usually with small posterior accessory cusp (but often absent on even moderately worn teeth). Second lower premolar (p2) taller than p3; lower milk premolar (dp3) with complete trigonid (tricuspid in in C. chacoensis) or with incomplete trigonid (bicuspid in C. unduaviensis). Hypoconid labially salient (level with labial apex of protoconid) on m3; hypoconulid twinned with entoconid on m1–m3; entoconid much taller than hypoconulid on m1–m3.

DISTRIBUTION: Cryptonanus is known from mostly unforested tropical and subtropical biomes south of the Amazon River and east of the Andes, including the Caatinga, Cerrado, Chaco, and northern Pampas; collection localities mapped by Voss et al. (2005) and D'Elía and Martínez (2006) are from Paraguay, Uruguay, eastern Bolivia, northern Argentina, and eastern Brazil.

REMARKS: Our species-level taxonomy of Cryptonanus follows Voss et al. (2005) who, however, cautioned that this arrangement is provisional and unsupported by rigorous tests of species limits. In particular, the genetic distinctness of C. agricolai, C. chacoensis, C. ignitus, and C. unduaviensis is not overwhelmingly indicated by available morphological data and merits careful evaluation. Molecular sequence comparisons would be an especially welcome supplement to the scant phenotypic evidence at hand.

Gracilinanus Gardner and Creighton, 1989

Figure 50

Figure 51

Lestodelphys halli (based on UWZM 22422, an adult male from Clemente Onelli, Río Negro, Argentina; and MVZ 173727, an adult male from Lihuel-Calel, La Pampa, Argentina).

CONTENTS: aceramarcae Tate, 1931; agilis Burmeister, 1854 (including beatrix Thomas, 1910; buenavistae Tate, 1931; and peruana Tate, 1931); dryas Thomas, 1898; emiliae Thomas, 1909 (including longicaudus 137Hershkovitz, 1992); marica Thomas, 1898 (including perijae 137Hershkovitz, 1992); and microtarsus Wagner (1842).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 75–130 mm; adult weight ca. 10–50 g. Rhinarium with two ventrolateral grooves on each side of median sulcus; dark circumocular mask present; pale supraocular spot absent; dark midrostral stripe absent; throat gland present in adult males. Dorsal pelage unpatterned reddish or grayish brown with dark-gray hair bases; dorsal guard hairs short and inconspicuous; ventral fur mostly gray-based buff or orange (e.g., in G. microtarsus), but self-white in some species (e.g., G. emiliae). Manus paraxonic (dIII  =  dIV); manual claws small, not extending beyond fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium smooth or sparsely tuberculate; lateral carpal tubercles present in fully adult males, but medial carpal tubercles absent in both sexes. Pedal digits unwebbed; dIV longer than other pedal digits; plantar surface of heel naked. Pouch absent; mammary formula (unknown for most species) 3–1–3  =  7 to 4–1–4  =  9; cloaca present. Tail much longer than combined length of head and body, slender and muscular (not incrassate), and macroscopically naked (furred only near base); naked caudal integument usually unicolored (dark) but sometimes indistinctly bicolored (dark above, paler below) at base; caudal scales in annular series (most species) or annular and spiral series (G. emiliae), each scale with three subequal bristlelike hairs emerging from distal margin; ventral caudal surface modified for prehension distally (with naked median groove and apical pad bearing dermatoglyphs).

Premaxillary rostral process present. Nasals long, extending anteriorly beyond I1 (concealing nasal orifice from dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. Two lacrimal foramina present on each side but usually inconspicuous in lateral view (partially concealed within anterior orbital margin). Interorbital region more or less parallel-sided, without well-developed constrictions; supraorbital margins smoothly rounded in most species (e.g., G. dryas) or with well-developed beads (in G. emiliae); postorbital processes usually absent except in large adult specimens of G. microtarsus. Left and right frontals and parietals separated by persistent median sutures. Parietal and alisphenoid in contact on lateral braincase (no squamosal-frontal contact). Sagittal crest absent. Petrosal laterally exposed through fenestra in parietal-squamosal suture. Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine fenestrae present; palatine fenestrae present; maxillary fenestrae present in most species (but very small and occasionally absent uni- or bilaterally in G. emiliae); posterolateral palatal foramina small, not extending lingual to M4 protocones; posterior palatal morphology conforms to Didelphis morphotype (with well-developed lateral corners, the choanae constricted behind). Maxillary and alisphenoid not in contact on floor of orbit (separated by palatine). Transverse canal foramen present. Alisphenoid tympanic process flask shaped or globular (not laterally compressed and without a well-developed ventral process), almost always with anteromedial strut enclosing extracranial course of mandibular nerve (secondary foramen ovale usually present), and not contacting rostral tympanic process of petrosal. Anterior limb of ectotympanic directly suspended from basicranium. Stapes triangular, with large obturator foramen. Fenestra cochleae usually exposed (partially concealed by caudal tympanic process of petrosal in some specimens of G. aceramarcae and G. microtarsus). Paroccipital process small, rounded, and adnate to petrosal. Dorsal margin of foramen magnum bordered by exoccipitals and supraoccipital, incisura occipitalis present.

Two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 symmetrically rhomboidal (“premolariform”), with subequal anterior and posterior cutting edges, increasing in length (mesiodistal dimension) from I2 to I5 (e.g., in G. microtarsus), or I2–I5 crowns subequal in length (e.g., in G. emiliae). Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps except in G. emiliae (which consistently has a posterior accessory cusp). First upper premolar (P1) smaller than posterior premolars but well-formed and not vestigial; second and third upper premolars (P2 and P3) subequal in height; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Molars carnassialized (postmetacristae longer than postprotocristae); relative widths M1 < M2 < M3 > M4 or M1 < M2 < M3 < M4; centrocrista strongly inflected labially on M1–M3; ectoflexus indistinct or absent on M1 and M2 but consistently distinct on M3; anterolabial cingulum continuous with preprotocrista (complete anterior cingulum present) on M3; postprotocrista without carnassial notch. Last upper tooth to erupt is P3.

Lower incisors (i1–i4) with distinct lingual cusps. Second lower premolar (p2) taller than p3; lower milk premolar (dp3) trigonid incomplete (unicuspid in the few specimens we were able to score for this trait). Hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid taller than hypoconulid on m1–m3.

DISTRIBUTION: Species of Gracilinanus occur in tropical and subtropical moist forests in lowland and montane landscapes (to at least 3350 m; Voss et al., 2004b: table 12) from northern Venezuela to Bolivia, Paraguay, southeastern Brazil, northeastern Argentina, and northern Uruguay (D'Elía and Martínez, 2006; Teta et al., 2007; Creighton and Gardner, 2008a). Apparently, only one Argentinian record and two Uruguayan records of Gracilinanus are valid, the remaining published localities for this genus from both countries having been based on specimens of Cryptonanus (see Voss et al., 2005; D'Elía and Martínez, 2006; Teta et al., 2007). Gracilinanus is said not to occur in central Amazonia (Patton and Costa, 2003), but the species to be expected there (G. emiliae) is difficult to collect and this distributional hiatus is perhaps an artifact of inadequate faunal sampling.30

REMARKS: The contents of this genus have changed significantly in the years since it was first described by Gardner and Creighton (1989), principally by removal of distantly related forms to new genera (see Voss et al., 2005: table 1). With the exception of Gracilinanus microtarsus and G. agilis, species of this genus are not common in museum collections, and geographic distributions remain poorly documented. In addition, current species concepts will doubtless change as more information becomes available. For example, molecular data (summarized by Costa et al., 2003) suggest the presence of an undescribed, morphologically cryptic species among the material currently identified as G. microtarsus; the geographic distribution of material currently identified as G. emiliae is improbably broad; and some specimens from northern Colombia and eastern Peru do not resemble any currently recognized taxa. Therefore, the species-level diversity of Gracilinanus seems certain to increase with future revisionary research.

The monophyly of Gracilinanus in its currently restricted sense is strongly supported by parsimony, likelihood, and Bayesian analyses of RAG1 (fig. 30), BRCA1 (fig. 31), vWF (fig. 32), concatenated sequence data from five genes (fig. 33), and combined datasets that include both nonmolecular and molecular characters (figs. 35, 36); it is also moderately supported by parsimony and likelihood analyses of IRBP (fig. 28). We have no plausible explanation as to why analyses of DMP1 strongly support the conflicting hypothesis that G. emiliae is the sister taxon of Cryptonanus (fig. 29). Although Gracilinanus is recovered as a clade by parsimony analysis of nonmolecular data (fig. 27), only two morphological traits optimize as unambiguous generic synapomorphies (appendix 5).

Lestodelphys Tate, 1934

Figure 51

Figure 52

Marmosops creightoni (based on CBF 6552 and UMMZ 155999, adult males from the valley of the Río Zongo, La Paz, Bolivia).

CONTENTS: halli Thomas, 1921.

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body probably ca. 120–160 mm; adult weight probably ca. 60–100 g (very few reliably measured specimens accompanied by weights are available). Rhinarium with one ventrolateral groove on each side of median sulcus; dark circumocular mask present; pale supraocular spot absent; dark midrostral stripe absent; throat gland present in adult males. Dorsal pelage tricolored (distinctly darker middorsally than on flanks); dorsal underfur dark gray; dorsal guard hairs short and inconspicuous; ventral fur self-white from chin to anus. Manus mesaxonic (dIII > dIV); manual claws much longer than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium densely tuberculate; carpal tubercles absent. Pedal digits unwebbed; dIII longer than other pedal digits; plantar surface of heel coarsely furred. Pouch absent; mammae 8–1–8  =  17 or 9–1–9  =  19, of which the anteriormost teats are “pectoral”; cloaca present. Tail shorter than combined length of head and body, incrassate, and unfurred except at base; caudal integument bicolored (dark above, distinctly paler below); tail scales difficult to distinguish but apparently in annular series; caudal hairs all subequal in length and thickness; ventral caudal surface not modified for prehension.

Premaxillary rostral process absent. Nasals long, extending anterior to I1 (concealing nasal orifice in dorsal view), and conspicuously widened posteriorly near premaxillary-maxillary suture. Maxillary turbinals elaborately branched. Two lacrimal foramina present on each side, exposed in lateral view anterior to orbital margin. Supraorbital margins smoothly rounded, without beads or crests; postorbital frontal processes usually absent (indistinct processes are occasionally developed in old adult males). Left and right frontals and parietals separated by persistent median sutures. Parietal and alisphenoid bones in contact on lateral surface of braincase (no frontal-squamosal contact). Sagittal crest usually absent but weakly developed along midparietal suture (not extending anteriorly to frontals) in some large specimens.31 Petrosal exposed through fenestra in parietal-squamosal suture in some individuals but not in others. Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine fenestrae present but sometimes quite small; palatine fenestrae variable but usually present (Martin, 2005: fig. 2); maxillary fenestrae absent; posterolateral palatal foramina very long, usually extending anteriorly lingual to M4 protocones; posterior palatal morphology conforms to Didelphis morphotype (with strongly produced posterolateral corners, the choanae constricted behind). Maxillary and alisphenoid not in contact (separated by palatine) on floor of orbit. Transverse canal foramen present. Alisphenoid tympanic process smoothly globular, with well-developed anteromedial process enclosing extracranial course of mandibular nerve (secondary foramen ovale present), and not contacting rostral tympanic process of petrosal. Anterior limb of ectotympanic directly suspended from basicranium. Stapes usually columelliform and microperforate or imperforate. Fenestra cochleae concealed in sinus formed by rostral and caudal tympanic processes of petrosal. Paroccipital process small, rounded, and adnate to petrosal. Dorsal margin of foramen magnum bordered by supraoccipital and exoccipitals, incisura occipitalis present.

Two mental foramina usually present on lateral surface of each hemimandible (three foramina are present unilaterally on two specimens examined); angular process acute and strongly inflected.

Unworn crowns of I2–I5 symmetrically rhomboidal (“premolariform”), with subequal anterior and posterior cutting edges, slightly increasing in length (mesiodistal dimension) from I2 to I5. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps. First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; third upper premolar (P3) taller than P2; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Molars strongly carnassialized (postmetacristae much longer than postprotocrista); relative widths M1 < M2 < M3 < M4; centrocrista strongly inflected labially on M1–M3; ectoflexus shallow or absent on M1, consistently present and distinct on M2, consistently deep on M3; anterolabial cingulum and preprotocrista discontinuous (anterior cingulum incomplete) on M3; postprotocrista without carnassial notch. Last upper tooth to erupt is P3.

Lower incisors (i1–i4) with distinct lingual cusps. Lower canine (c1) erect, acutely pointed, and simple (without a posterior accessory cusp). Third lower premolar (p3) taller than p2; lower milk premolar (dp3) large, but trigonid incomplete (uni- or bicuspid). Hypoconid lingual to protoconid (not labially salient) on m3; hypoconulid twinned with entoconid on m1–m3; entoconid taller than hypoconulid on m1–m3.

DISTRIBUTION: Most known specimens of Lestodelphys have been collected in semidesert shrubland and steppe habitats in Patagonian Argentina between 41° and 47°S latitude, but there are two outlying records from the Monte desert (between 32° and 38°S) that may represent a relictual population (Sauthier et al., 2007); all reported elevations associated with collected specimens are less than 1000 m above sea level (Birney et al., 1996; Martin, 2003). As noted by Flores et al. (2007), the distribution of Lestodelphys extends further south than that of any other living marsupial.

Marmosops Matschie, 1916

Figures 52, 53

Figure 53

Marmosops pinheiroi (based on AMNH 267345, a young adult male from Paracou, French Guiana), a small species that differs from M. creightoni (fig. 52) in several trenchant craniodental features including laterally exposed lacrimal foramina, absence of palatine vacuities, and presence of accessory cusps on C1.

Figure 54

Thylamys pusillus (based on AMNH 275442, an adult female, and 275445, an adult male, both from Estancia Bolívar, Tarija, Bolivia).

CONTENTS: bishopi Pine, 1981; cracens Handley and Gordon, 1979; creightoni Voss et al., 2004; fuscatus Thomas, 1896 (including carri J.A. Allen and Chapman, 1897; and perfuscus Thomas, 1924); handleyi Pine, 1981; impavidus Tschudi, 1845 (including caucae Thomas, 1900; celicae Anthony, 1922; madescens Osgood, 1913; oroensis Anthony, 1922; sobrinus Thomas, 1913; and ucayalensis Tate, 1931); incanus Lund, 1840 (including bahiensis Tate, 1931; and scapulatus Burmeister, 1856); invictus Goldman, 1912; juninensis Tate, 1931; neblina Gardner, 1990; noctivagus Tschudi, 1844 (including albiventris Tate, 1931; collega Thomas, 1920; dorothea Thomas, 1911; keaysi J.A. Allen, 1900; leucastrus Thomas, 1927; lugendus Thomas, 1927; neglectus Osgood, 1915; politus Cabrera, 1913; purui Miller, 1913; stollei Miranda-Ribeiro, 1936; and yungasensis Tate, 1931); ocellatus Tate, 1931; parvidens Tate, 1931; paulensis Tate, 1931; and pinheiroi Pine, 1981 (including woodalli Pine, 1981).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 90–195 mm; adult weight ca. 20–140 g. Rhinarium with two ventrolateral grooves on each side of median sulcus; dark circumocular mask present; pale supraocular spot absent; dark midrostral stripe absent; throat gland consistently present in adult males of some species (e.g., M. incanus, M. noctivagus) but apparently absent in others (e.g., M. impavidus, M. parvidens). Dorsal pelage unpatterned, usually some shade of dull grayish brown, but distinctly reddish or grayish in some species; dorsal hair bases always dark gray; dorsal guard hairs short and inconspicuous; ventral fur self-colored (whitish, buffy, or brown) or entirely or partially gray based. Manus mesaxonic (dIII > dIV); manual claws small, shorter than fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium smooth or sparsely tuberculate; lateral carpal tubercles externally conspicuous on wrists of large adult males; medial carpal tubercles absent in both sexes. Pedal digits unwebbed; dIV longer than other pedal digits; plantar surface of heel macroscopically naked (a microscopic pelage of very short hairs, not coarse fur, is usually present). Pouch absent; mammae varying among examined species from 3–1–3  =  7 to 7–1–7  =  15, anteriormost pairs “pectoral” when mammae ≥11; cloaca present. Tail longer than combined length of head and body, slender and muscular (not incrassate), and macroscopically naked (without a conspicuously furred base); naked caudal integument uniformly dark (usually grayish), bicolored (dark above, pale below), or parti-colored (dark distally, paler distally); caudal scales in spiral series, each scale with three bristlelike hairs emerging from caudal margin; median hair of each caudal-scale triplet usually much thicker and darker than lateral hairs; ventral caudal surface modified for prehension distally (with naked median groove and apical pad bearing dermatoglyphs).

Premaxillary rostral process long and well developed in some species (e.g., M. parvidens, M. pinheiroi) but short in others (e.g., M. noctivagus) and apparently absent in some (M. incanus). Nasals long, extending anteriorly beyond I1 (concealing nasal orifice in dorsal view), and conspicuously widened posteriorly near maxillary-frontal suture except in M. incanus (which has narrow, more or less parallel-sided nasals). Maxillary turbinals (viewed through nasal orifice) elaborately branched. Lacrimal foramina, usually two on each side, concealed from lateral view inside orbit (e.g., in M. parvidens) or exposed on orbital margin (e.g., in M. pinheiroi). Interorbital region without abrupt constrictions; supraorbital margins rounded in some species (e.g., M. parvidens), beaded in others (e.g., M. noctivagus); postorbital processes usually absent except in old males of some species (e.g., M. impavidus32). Left and right frontals and parietals separated by persistent median sutures. Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest absent. Petrosal consistently exposed laterally through fenestra in parietal-squamosal suture in most species (fenestra polymorphically absent in M. incanus and M. noctivagus). Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine fenestrae present; palatine fenestrae present in some species (e.g., M. creightoni; fig. 52), but absent in others (e.g., M. pinheiroi; fig. 53); maxillary fenestrae absent; posterolateral palatal foramina small, not extending anteriorly lingual to M4 protocones; posterior palatal morphology conforms to Didelphis morphotype (with prominent lateral corners, the choanae constricted behind). Maxillary and alisphenoid not in contact on floor of orbit (separated by palatine). Transverse canal foramen present. Alisphenoid tympanic process small, often bluntly conical or laterally compressed (never smoothly globular), always with a well-developed anteromedial process enclosing the extracranial course of mandibular nerve (secondary foramen ovale present), and not contacting rostral tympanic process of petrosal. Anterior limb of ectotympanic directly suspended from basicranium. Stapes triangular, with large obturator foramen. Fenestra cochleae exposed, not concealed by rostral and caudal tympanic processes of petrosal. Paroccipital process small, rounded, and adnate to petrosal. Dorsal margin of foramen magnum bordered by supraoccipital and exoccipitals, incisura occipitalis present.

Two mental foramina usually present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 symmetrically rhomboidal (“premolariform”), with subequal anterior and posterior cutting edges, increasing in length (mesiodistal dimension) from I2 to I5. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 without accessory cusps in some species (e.g., M. creightoni; fig. 52), or with posterior accessory cusp (e.g., M. juninensis), or with anterior and posterior accessory cusps (e.g., M. pinheiroi; fig 53). First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; second and third upper premolars (P2 and P3) subequal in height; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Molars strongly carnassialized (postmetacristae much longer than postprotocristae); relative widths usually M1 < M2 < M3 < M4; centrocrista strongly inflected labially on M1–M3; ectoflexus shallow or indistinct on M1, shallow but usually distinct on M2, and consistently deep on M3; anterolabial cingulum and preprotocrista discontinuous (anterior cingulum incomplete) on M3 in some species (e.g., M. noctivagus) but anterolabial cingulum continuous with preprotocrista (anterior cingulum complete) in others (e.g., M. parvidens); postprotocrista without carnassial notch. Last upper tooth to erupt is P3.

Lower incisors (i1–i4) with distinct lingual cusps. Second lower premolar (p2) subequal in height to p3 in some species (e.g., M. incanus) but distinctly taller than p3 in others (e.g., M. impavidus); lower milk premolar (dp3) trigonid complete. Hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid usually taller than hypoconulid on m1–m3.

DISTRIBUTION: Species of Marmosops occur in tropical lowland and montane moist forests from central Panama southward along the Andes and throughout Amazonia to eastern Bolivia and southeastern Brazil (see maps in Gardner and Creighton, 2008a). The genus is apparently unknown from Paraguay, Argentina, and Uruguay. See Mustrangi and Patton (1997) for verified specimen records based on modern revisionary research in southeastern Brazil, Patton et al. (2000) for the same in western Amazonia, Voss et al. (2001) for northeastern Amazonia, and Voss et al. (2004b) for Bolivia. Many other published geographic data purporting to represent records of Marmosops species (e.g., Anderson, 1997; Brown, 2004) are based on misidentified material (Voss et al., 2004b).

REMARKS: The monophyly of Marmosops is uniformly strongly supported by parsimony, likelihood, and Bayesian analyses of IRBP (fig. 28), DMP1 (fig. 29), BRCA1 (fig. 31), vWF (fig. 32), and concatenated sequence data from five genes (fig. 33). The genus is also recovered with strong support by parsimony and Bayesian analyses of combined datasets that include nonmolecular and molecular characters (figs. 35, 36), and by Bayesian analyses of nonmolecular characters (fig. 27) and RAG1 sequences (fig. 30). Spirally arranged caudal scales, the petiolate morphology of the central hair in each caudal-scale triplet, the presence of a rostral process of the premaxillae, and unique deletions at the DMP1 and BRCA1 loci optimize as unambiguous generic synapomorphies (appendix 5; figs. 29, 31). As noted elsewhere, Kirsch and Palma's (1995) suggestion that Marmosops is paraphyletic was based on misidentified voucher material (Voss and Jansa, 2003: 57).

Discrepancies between our species-level synonymies for Marmosops and those in Gardner (2005) are explained by Voss et al. (2004b). Few of the currently recognized species have received critical revisionary attention, and it seems likely that several widespread taxa (e.g., M. fuscatus, M. noctivagus, and M. impavidus) will prove to be composite. For regional revisions of Marmosops see Mustrangi and Patton (1997), Patton et al. (2000), and Voss et al. (2001, 2004b).

Thylamys Gray, 1843

Figure 54

CONTENTS: cinderella Thomas, 1902 (including sponsorius Thomas, 1921); elegans Waterhouse, 1839 (including coquimbensis Tate, 1931; and soricinus Philippi, 1894); karimii Petter, 1968; macrurus Olfers, 1818 (including griseus Desmarest, 1827); pallidior Thomas, 1902; pusillus Desmarest, 1804 (including bruchi Thomas, 1921; citellus Thomas, 1912; nanus Olfers, 1818; and verax Thomas, 1921); tatei Handley, 1957; velutinus Wagner, 1842 (including pimelurus Reinhardt, 1851); and venustus Thomas, 1902 (including janetta Thomas, 1926).

MORPHOLOGICAL DESCRIPTION: Combined length of adult head and body ca. 75–140 mm; adult weight ca. 10–65 g. Rhinarium with two ventrolateral grooves on each side of median sulcus in most examined species (but T. pallidior has only a single ventrolateral groove on each side); dark circumocular mask present; pale supraocular spot absent; dark midrostral stripe absent; throat gland present in fully adult males, and apparently also in adult females of some species (Carmignotto and Monfort, 2006). Dorsal pelage usually grayish and distinctly darker middorally than laterally (“tricolored”), but dorsal fur brownish and without conspicuous patterning in T. karimii and T. velutinus (Carmignotto and Monfort, 2006); dorsal hair bases dark gray; dorsal guard hairs short and inconspicuous; ventral fur self-white in some species (e.g., T. pusillus), wholly or partly gray-based whitish or yellowish in others (e.g., T. venustus). Manus mesaxonic (dIII > dIV); manual claws about as long as fleshy apical pads in most species (but claws extend well beyond apical pads in T. karimii and T. velutinus; Carmignotto and Monfort, 2006); dermatoglyph-bearing manual plantar pads present in most species; central palmar epithelium densely covered with small convex tubercles; carpal tubercles absent in both sexes. Pedal digits unwebbed; dIV longer than other pedal digits; plantar surface of heel coarsely furred. Pouch absent; mammae varying among species from 4–1–4  =  9 (in T. karimii; Carmignotto and Monfort, 2006) to 7–1–7  =  15 (in T. elegans; Tate, 1933), including “pectoral” teats (when the total number of mammae ≥ 11); cloaca present. Tail longer than combined length of head and body in most species (except in T. karimii and T. velutinus; Carmignotto and Monfort, 2006); apparently always incrassate (but quantity of stored fat may vary seasonally); furred only at base (to about the same extent dorsally as ventrally); unfurred caudal integument bicolored (dark above, pale below) in most species and parti-colored in some (T. macrurus has an all-white tail tip); tail scales in annular series, each scale with three subequal bristlelike hairs emerging from distal margin; ventral caudal surface modified for prehension distally (with naked median groove and apical pad bearing dermatoglyphs) in most species, but ventral prehensile surface reduced or absent in T. karimii and T. velutinus (Carmignotto and Monfort, 2006).

Premaxillary rostral process absent. Nasals long, extending anteriorly above or beyond I1 (concealing nasal orifice in dorsal view), and uniformly narrow (with subparallel lateral margins) in most specimens. Maxillary turbinals elaborately branched. Two lacrimal foramina present on each side, exposed to lateral view on anterior orbital margin in most species, but usually concealed inside orbit in others (e.g., T. tatei). Interorbital and postorbital constrictions often distinct; supraorbital margins rounded in some species (e.g., T. pallidior), squared or beaded in others; postorbital processes usually absent or indistinct but occasionally present in old adults of some species (e.g., T. macrurus). Left and right frontals and parietals separated by persistent median sutures. Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest absent. Petrosal almost always exposed laterally through fenestra in parietal-squamosal suture. Parietal-mastoid contact present (interparietal does not contact squamosal).

Maxillopalatine fenestrae present; palatine fenestrae present; maxillary fenestrae present in some species (e.g., T. pusillus) but absent in others (e.g., T. pallidior); posterolateral palatal foramina very large, extending anteriorly between M4 protocones; posterior palatal morphology conforms to Didelphis morphotype (with prominent lateral corners, the choanae constricted behind). Maxillary and alisphenoid not in contact on floor of orbit (separated by palatine). Transverse canal foramen present. Alisphenoid tympanic process smoothly globular, with anteromedial process enclosing extracranial course of mandibular nerve (secondary foramen ovale present), and not contacting rostral tympanic process of petrosal. Anterior limb of ectotympanic directly suspended from basicranium. Stapes triangular, with large obturator foramen. Fenestra cochleae concealed in sinus formed by rostral and caudal tympanic processes of petrosal. Paroccipital process small, rounded, adnate to petrosal. Dorsal margin of foramen magnum bordered by supraoccipital and exoccipitals, incisura occipitalis present.

Two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.

Unworn crowns of I2–I5 symmetrically rhomboidal (“premolariform”), with subequal anterior and posterior cutting edges, increasing in length (mesiodistal dimension) from I2 to I5. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 without accessory cusps in all examined species (but see Carmignotto and Monfort, 2006).33 First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; third upper premolar (P3) taller than P2; P3 with posterior cutting edge only; upper milk premolar (dP3) large, molariform, and nonvestigial. Molars strongly carnassialized (postmetacristae much longer than postprotocristae); relative widths usually M1 < M2 < M3 < M4; centrocrista strongly inflected labially on M1–M3; ectoflexus absent or indistinct on M1, usually present but shallow on M2, consistently deep and distinct on M3; anterolabial cingulum and preprotocrista discontinuous (anterior cingulum incomplete) on M3; postprotocrista without carnassial notch. Last upper tooth to erupt is P3.

Lower incisors (i1–i4) with distinct lingual cusps. Second and third upper premolars (p2 and p3) subequal in height; lower milk premolar (dp3) trigonid usually incomplete (bicuspid). Hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid much taller than hypoconulid on m1–m3.

DISTRIBUTION: Species of Thylamys collectively range from north-central Peru (Ancash) southward along the Andes and Pacific coastal lowlands to central Chile; in the unforested landscapes south of Amazonia, the genus extends eastward across Bolivia (Anderson, 1997), Paraguay, and Argentina (Flores et al., 2007) to Uruguay (González et al., 2000) and eastern Brazil (Bahia, Pernambuco, and Piauí; Carmignotto and Monfort, 2006). Recorded elevations range from near sea level to at least 3800 m (Solari, 2003).

REMARKS: The monophyly of Thylamys is strongly supported by parsimony, likelihood, and Bayesian analyses of IRBP (fig. 28), BRCA1 (fig. 31), vWF (fig. 32), and concatenated sequence data from five genes (fig. 33). Strong support for this clade is also provided by parsimony and likelihood analyses of DMP1 (fig. 29) and by parsimony and Bayesian analyses of RAG1 (fig. 30) and combined datasets that include both nonmolecular and molecular characters (figs. 35, 36). The monophyly of Thylamys is also recovered with moderate support by parsimony analysis of nonmolecular characters (fig. 27), but only one morphological trait (uniformly narrow nasals; see appendix 5) optimizes as an unambiguous generic synapomorphy.

We tentatively recognize bruchi as a synonym of Thylamys pusillus following Voss et al. (in press), who explain why the former name does not belong in the synonymy of T. pallidior (contra Creighton and Gardner, 2008c), and we refer sponsorius to the synonymy of T. cinderella following Braun et al., (2005). The names pulchellus Cabrera, 1934, and fenestrae Marelli, 1932, might be synonyms of T. pusillus and T. pallidior, respectively, but we have not seen the holotypes and published information about these nominal taxa is insufficient to support any definite conclusions about them. Many other species-level issues in this genus remain problematic despite much recent taxonomic work (e.g., Palma et al., 2002; Solari, 2003; Braun et al., 2005; Carmignotto and Monfort, 2006).