Brassomys albidens, Voss & Meng & Prendini & Voss & Whiteley & Knight & Lunde & And & Melomys & Bulletin, 2009

Voss, Musser Scientific Publications Of The American Museum Of Natural History American Museum Novitates Bulletin Of The American Museum Of Natural History Anthropological Papers Of The American Museum Of Natural History Publications Committee Robert S., Meng, Chair Board Of Editors Jin, Prendini, Paleontology Lorenzo, Voss, Invertebrate Zoology Robert S., Whiteley, Vertebrate Zoology Peter M., Knight, Anthropology Managing Editor Mary, Lunde, And, New Guinea Coccymys, Melomys & Bulletin, Albidens Amnh, 2009, Systematic Reviews Of New Guinea Coccymys And ‘‘ Melomys’ ’ Albidens (Muridae, Murinae) With Descriptions Of New Taxa, Bulletin of the American Museum of Natural History 2009 (329), pp. 1-139 : 1-139

Brassomys albidens , new combination

HOLOTYPE AND TYPE LOCALITY: The holotype is an adult male (AMNH 150821), collected by W.B. Richardson (original number 4698) on August 14, 1938. It consists of a stuffed study skin (fig. 8) accompanied by a skull (fig. 10), all in good condition. Occlusal surfaces of the molars are worn, but cusp patterns are still evident. External, cranial, and dental measurements are listed in tables 7 and 20.

The type locality is about 15 miles (24 km) north of Mt. Wilhelmina (Gunung Trikora), at 3225 m near Lake Habbema (04 ° 499S, 138 ° 419E; locality 2 in fig. 3), which is nestled between the ridges forming northern slopes of the Snow Mountains (‘‘Sneeuw Gebergte’’ on older Dutch maps; Pegunungan Maoke is the Indonesian designation on contemporary maps and in modern gazetteers) in the Central Cordillera of western New Guinea (Papua Province, Indonesia).

REFERRED MATERIAL: There are five other modern specimens, all from the Snow Mountains, captured during October 1938, in the Bele River valley, 9 km northeast of Lake Habbema at 2800 m (04 ° 059S, 138 ° 509E; locality 3 in fig. 3). Date of collection, sex, age, and measurements for each example are listed in table 20. The skulls of three specimens are intact, but those of two others are damaged. Tate had identified only two of the five as albidens (AMNH 150531 and 150541), one as Pogonomys sylvestris (AMNH 150607), and two as Pogonomelomys ruemmleri (AMNH 150518 and 150923).

In addition to the holotype and five other modern specimens, there are three dentary fragments (AMF 134074, 134096, 134126) that were extracted from Late Pleistocene sediments excavated in Kelangurr Cave (04 ° 019S, 138 ° 089E; locality 1 in fig. 3), 2950 m. The cave is on the northern slopes of the Snow Mountains about 60 airline km west of Lake Habbema, in a valley confluent with the valley of the West Baliem River (see

TABLE 20

Measurements (mm) of the Six Known Modern Specimens of Brassomys albidens map and description in Flannery, 1999). Two of the fossils had been identified as Coccymys ruemmleri , and the third as ‘‘ C. sp.’’

GEOGRAPHIC DISTRIBUTION: Brassomys albidens is still known by only the original six skins and skulls obtained from 3225 and 2800 m, and the three Late Pleistocene fragments collected in Kelangurr Cave on the northern slopes of the Snow Mountains, but probably occurs in montane forest formations at comparable altitudes elsewhere in the Central Cordillera of western New Guinea. If the range of B. albidens is concordant with that of other montane mammals that are endemic to western New Guinea (see Helgen, 2005b, 2007a, 2007c), it should be found in suitable forest habitats along the length of the Snow Mountains and possibly in the adjacent Star Mountains. This mountain backbone is part of Flannery’s (1995: 37) ‘‘Western Sub-Province’’ of the ‘‘Tumbunan Biogeographic Province’’ defined by Schodde and Calaby (1972); the subprovince extends from the Wissel Lakes region in the west (03 ° 559S, 136 ° 159E) to the deep valley of the Strickland River in the east (06 ° 309S, 142 ° 049E). With very few exceptions, mammalogical surveys throughout the western Central Cordillera formed of the Snow and Star mountains have been sparse (see Geographic Distribution in the account of C. ruemmleri ).

DESCRIPTION AND COMPARISONS: Here we describe the external form, pelage, and characteristics of the skull and dentition of Brassomys albidens based on five adults and one juvenile. We note the contrasting condition in the samples of Coccymys ruemmleri (the type species of Coccymys ) from the Snow Mountains at the appropriate places in the account, discuss absolute and proportional differences in cranial and dental variables, and summarize the primary trenchant distinctions between the two genera in table 21.

Adults of both B. albidens and C. ruemmleri share a short masked face; long facial vibrissae, small body clothed in luxuriant, thick, and dark brown fur; and a thin, brownish gray tail that is longer than head and body (figs. 6, 8). Upperparts of B. albidens are typically brighter than in C. ruemmleri : dorsal pelage over most of the upperparts is dense (overhairs are up to 12 mm long), soft to the touch, and woolly in appearance, and a rich buffy brown from head to rump and along back and sides (overhairs are dark gray for most of their lengths and have rich buffy brown tips). Guard hairs project slightly beyond the overfur layer but not enough to alter the even contour of the fur covering or its color. The bright buffy brown extends onto the thighs, upper arms, and cheeks. A dark brown ring encircles each eye ( C. ruemmleri has darker, more somber upperparts, the coat is thicker, 11–14 mm, and lax instead of woolly, and the head is gray, patterned only by a blackish area around each eye that extends forward over each side of the muzzle to the bases of the mystacial vibrissae). The very long mystacial vibrissae are mostly black with silvery tips, the longest reaching 70 mm and extending well beyond the pinnae when laid against the head. The longest of the superciliary vibrissae extend to slightly beyond the pinnae, and the usual murine array of facial submental and interramal sensory hairs and ulnar and tarsal vibrissae are apparent (all except possibly the genal vibrissae, which we could not find—either they are absent or indistinguishable from overhairs in the dorsal coat). The ears are large, dark brown, and covered inside and out with fine, short hairs. Brassomys albidens has much larger and more expansive external pinnae than does C. ruemmleri (means and standard deviations are: 20.0 mm ± 1.58, range 18–22 mm, for B. albidens , N 5 5; 16.9 mm ± 1.09, range 15–19 mm for C. ruemmleri from Lake Habbema, N 5 16; see table 3 for univariate summaries from other samples of C. ruemmleri ). We compared length of ear between adults of B. albidens and a sample of adult C. ruemmleri from Lake Habbema, which were all likely measured by one collector, W.B. Richardson, so limits of the dimension would be consistent whether taken from base to crown or from notch to crown. Unfortunately, length of ear as it was measured does not convey the striking difference in size when skins are compared side-by-side.

The ventral coat of B. albidens is also soft and dense, and up to 10 mm thick. The entire underparts, from chin to base of tail, are pigmented and among the five adults ranges from whitish gray (AMNH 150923) through dark grayish white (AMNH 150541, 150607) and pale buffy gray (AMNH 150531) to intense and bright ochraceous gray in the holotype. The darker and buffy tones are produced by hairs that are dark gray for most of their lengths and have short unpigmented or buffy tips; relatively longer unpigmented tips result in paler expressions. Underparts of 36 of the 40 C. ruemmleri from the Snow Mountains that we examined are either whitish gray or a darker grayish white (the predominant tone), and only four have buffy gray underparts .

The tail is slender and longer than the head and body (LT/LHB ranges from 121% to 146% in five adults), but relatively shorter than in C. ruemmleri (tables 3, 7). It is covered in annuli of small, slightly swollen, and brown scales (19–24 scale rings per cm); these rings of scales abut one another near the base of the tail but appear to overlap slightly toward the tip in some specimens but not others. Three dark brown hairs, as long as three or four scales, emerge from beneath each scale, which imparts a hirsute texture to the entire tail, but no tuft at the tip (scale rows abut against each other in C. ruemmleri , the tail hairs are shorter relative to scale length, and while the texture is somewhat hirsute, it is slightly less hairy than in B. albidens ). All surfaces of the tail are covered by scales and hairs, without any sign of a long and naked gripping pad (thickened epidermis devoid of either scales or hair) over the dorsal surface of the distal fifth of the tail, indicating that the tail is not used as a dorsally prehensile organ (such a dorsal strip is prominent in C. ruemmleri , and the tail is used for dorsal prehensile grasping). The entire tail surface is brownish gray in all six specimens; AMNH 150607 is the only one with a short white tip (7 mm long, 5% of the tail length).

Dorsal surfaces of front feet and digits are unpigmented and densely covered with silvery hairs; metatarsal surfaces of the hind feet are pale brown, and densely covered with silvery and pale brown hairs, the digits and their covering of fine hairs are unpigmented. A tuft of silvery hairs springs from the base of each claw to cover about the basal half of each. The claws on both front and hind digits are unpigmented, large, long, and scytheshaped (metacarpal, metatarsal, and digital surfaces are sparsely haired in C. ruemmleri , the ungual tufts extend from bases of the digits to the tips, and the claws are smaller). Lengths of the digits relative to one another on front and hind feet match the proportions described for C. ruemmleri (the two middle digits of each front foot are the longest, the lateral digits slightly shorter; the three central digits of the hind foot are subequal in length and longer than the much shorter hallux, which bears a claw). The front foot is large; its palmar surface is naked, pale brown, and most of the area is formed by three large fleshy interdigital pads and posterior thenar and hypothenar mounds set close together (the large pads are evident even in their desiccated state on each dry study skin), resembling the palmar surface in such arboreal murines as the species of Margaretamys from Sulawesi ( Musser, 1981a: 278). ( C. ruemmleri has smaller front feet, which reflects its smaller body size, but the pads are less expansive relative to palmar surface.) Each hind foot is moderately wide across the carpal region and short relative to length of head and body (absolute length of hind foot in C. ruemmleri and B. albidens is about the same, so the former has a longer foot relative to body size, and it is also narrower; see tables 3, 7, and 20). Plantar surfaces are naked, pale brown, and bear four fleshy interdigital pads, a moderately large thenar, and much smaller hypothenar; the configuration is similar in C. ruemmleri . In both species, the pads are large but they do not occupy as much of the plantar surface as is seen in very specialized arboreal murines such as the species of Sulawesian Margaretamys ( Musser, 1981a: 279) .

No weights were recorded for the six specimens of B. albidens so we have no estimate of body mass. Also missing is a mammary count. All but 150618 were labeled as males in the field, do not exhibit any indication of teats, and some retain a dry scrotal sac. AMNH 150618 is in juvenile pelage and labeled as female, but there is no sign of teats on the dry skin (we cannot exclude the possibility that the specimen was incorrectly sexed in the field, which is easy to do with juveniles). A clear contrast exists between juvenile and adult coats. Juvenile pelage is slightly shorter (up to 10 mm) than that of adults and the hairs are finer, guard hairs project a bit beyond the overhairs, the texture is soft and more woolly, the upperparts are darker—a dark grayish brown—and the underparts are whitish gray. Juvenile C. ruemmleri (and C. shawmayeri ; we lack juveniles of C. kirrhos ) have a shorter coat than adults (up to 10 mm; 11–14 mm in adults) and the texture is finer and softer, but color of upperparts and underparts can hardly be distinguished from adult coverings—the underparts are whitish gray and the fur over head and body is warm brown.

The skull is small, as is that of C. ruemmleri (figs. 9, 10; table 4), but appears sturdy with a somewhat chunky rostrum, parallel zygomatic arches, and large, round braincase (‘‘full braincase’’ as Tate, 1951: 286, described it). The rostrum is moderately long and wide, and slightly tapered as viewed from a dorsal perspective, its lateral outlines barely interrupted by the low bump of each nasolacrimal canal. Smooth dorsolateral boundaries of the interorbital and postorbital regions define an hourglass-shaped interorbit in dorsal view. The back of the postorbital area is not defined by a vertical ridge where the frontal and squamosal bones meet, the conformation exhibited by C. ruemmleri , so there is no angular projection (in dorsal view) breaking the curved dorsolateral outlines of

TABLE 21 Summary of Notable Chromatic and Morphological Contrasts Between Adult Brassomys albidens and Coccymys ruemmleri (Differences are also reflected in the univariate summaries [table 4], ratio diagram [fig. 42], and illustrations of skins, skulls, and molars [figs. 6, 8–12, 40, 41].)

TABLE 21 (Continued)

the interorbital and postorbital regions. Beyond the postorbital boundaries, the deep braincase is round in outline, and smooth in texture: the parietal joins the squamosal smoothly without any roughened places or bead marking the attachment sites for temporal muscles, an inconspicuous linear beading defines each lamboidal ridge, and each mastoid is moderately inflated. The interparietal is deep (anterior-posterior dimension) and wide. Moderately thin zygomatic arches are parallel or flare slightly toward the front, and each arch projects only slightly outward from the side of the skull (arches flare out more in C. ruemmleri , and are not parallel but taper toward the rostrum); anterior margins of the dorsal maxillary roots of the zygoma form right angles to the rostrum and are not indented— no zygomatic notch. (A shallow convex dorsal margin between rostrum and top of the zygomatic plate represents the zygomatic notch in C. ruemmleri .) The jugal component of each zygoma is short.

When the skull is viewed from a lateral perspective, the dorsal outline rises straight and even along the rostrum and interorbit to arch over the braincase down to the occiput, which barely overhangs the occipital condyles (slightly greater cranial flexion in C. ruemmleri , which has a relatively deeper occiput). The rostrum is rectangular in lateral view, about as deep near the incisors as at the dorsal and ventral zygomatic roots (tapered in C. ruemmleri , higher near the zygomatic plates than at the incisors); margins of the nasals and premaxillaries project slightly beyond the incisor faces, and the nasolacrimal capsules are large but only slightly inflated. The zygomatic plate is very narrow, and its straight, vertical, leading edge does not project anterior to the dorsal maxillary root of the zygomatic arch, so there is no zygomatic notch; the posterior margin of the ventral maxillary root of the plate is level with the anterior third of the first molar (there is a much wider zygomatic plate in C. ruemmleri , both absolutely and relative to cranial dimensions [see figs. 41 and 43, table 4], with a slight zygomatic notch, the posterior edge lies in front of the first molar). The tendon for the superficial masseter muscle attaches to a robust knoblike tubercle projecting ventrolaterally from the ventral maxillary root of the zygomatic plate (the origin of the superficial masseter is simply a slightly roughened spot or low bump at the base of the plate in C. ruemmleri ; fig. 41). Behind the orbit, the squamosal root of the zygomatic arch originates midway on the side of the braincase. Posterior to the squamosal zygomatic root and dorsad of the auditory bulla, the squamosal is intact except near the squamosal-exoccipital suture where it is perforated by a large subsquamosal foramen through which the periotic is visible. The dorsolateral margin of the braincase is mostly smooth—a slight bevel marks the temporal ridge—and formed by the union of parietal and squamosal. A small portion of the parietal drops either just below the dorsolateral margin of the braincase or about halfway between the margin and top of the zygomatic root; this short projection and the squamosal form the wall of the braincase (slight temporal beading is evident along dorsolateral margins of the braincase in C. ruemmleri , and a much larger portion of the parietal projects farther ventrally to form part of the lateral braincase wall). The junction of the orbitosphenoid, alisphenoid, and frontal bones forms a solid section of the braincase wall, unbroken by a sphenofrontal foramen. The inner walls of the braincase are smooth, without squamosal-alisphenoid grooves. A wide bony alisphenoid strut (present in all six skulls) separates the foramen ovale accessorius from the combined buccinator-masticatory foramen. The auditory bulla is more tightly attached to the squamosal in B. albidens than it is in C. ruemmleri : the postglenoid foramen is a slit and the ventral postalar fissure is narrow (these openings are much wider in C. ruemmleri ); the periotic exposed along the dorsolateral margin of the bullar capsule projects forward as a thin tegmen tympani but does not overlap the squamosal. The occiput is not as deep as it is in C. ruemmleri — its extension beyond the occipitals is not as great, probably a reflection of the lesser degree of cranial flexion in B. albidens .

The wide incisive foramina are conspicuous when the skull is viewed from a ventral perspective: they occupy about 70% of the diastema, which is similar to the proportion in C. ruemmleri , but unlike that species, the posterior edges are anterior to front faces of the first molars (figs. 9, 10). The wide bony palate is about as long as the parallel molar rows, and its posterior margin is level with the front half of each third molar (even with backs of the third molars or projects slightly beyond them in C. ruemmleri ). Its surface is flat and smooth except where textured by a shallow pair of palatal grooves and large posterior palatine foramina set even with the anterior third of each second molar (surface of bony palate is longitudinally indented in C. ruemmleri , and the palatal grooves are deeper). The very wide mesopterygoid fossa is slightly narrower than the bony palate and its dorsolateral margins are intact or breached by elongate or short sphenopalatine vacuities. The adjacent pterygoid plates (5 parapterygoid plates) are triangular in ventral view and mostly intact except where the middle section is perforated by small sphenopterygoid vacuities, and at the back of the plate where it is pierced by the large ventral opening of the foramen ovale. Between the foramen ovale and middle lacerate foramen, a shallow groove scores the posterolateral area of the plate’s ventral surface. Each pterygoid fossa is moderately deep. The posterolateral margin of the pterygoid consists of either a threadlike bony ridge lateral to the foramen ovale (as in the holotype, fig. 10) or a slightly wider but still narrow and delicate ridge (the comparable ridge in C. ruemmleri is wider, more robust). A very short bony eustachian tube projects from each ectotympanic (auditory) bulla, and that capsule is very large and somewhat inflated relative to size of the skull—length of bulla is 18% of occipitonasal length. The bullar capsule has the same basic form as does the relatively smaller capsule in Coccymys , but it and the underlying periotic are just larger, and the capsule itself is deeper, not relatively wider because it shields about as much of the periotic surface as does the capsule in Coccymys , leaving exposed in ventral view a posteromedial wedge of that element between capsule and basioccipital and a thinner segment extending forward; the length of the medial edge of the capsule does not touch the lateral margin of the basioccipital. The dorsal wall of the carotid canal is formed by the periotic and its medial wall by the basioccipital, a configuration similar to that diagrammed for Oligoryzomys by Carleton and Musser (1989: 33). (the bulla is much smaller in C. ruemmleri , about 14% of the skull length, and not inflated, the ectotympanic does not conceal most of the periotic in ventral view, which is exposed as a posteromedial wedge and tapered flange extending forward separating the ectotympanic from the basioccipital margin). All of the specimens we examined possessed a large stapedial foramen penetrating the crevice (petromastoid fissure) between the bullar capsule and posteromedial wedge of the periotic. A slitlike or small, oval middle lacerate foramen separates the bullar capsule from the posterior margin of the pterygoid plate, a reflection of the inflated bulla (wide opening between capsule and pterygoid in C. ruemmleri , a result of its much smaller ectotympanic bulla).

All specimens of Brassomys albidens and Coccymys ruemmleri we studied share a carotid arterial pattern that is derived for muroid rodents in general but primitive for members of subfamily Murinae (character state 2 of Carleton, 1980; pattern 2 described by Voss, 1988; conformation diagrammed for Oligoryzomys by Carleton and Musser, 1989). Our sample of B. albidens is represented only by skulls, but they possess certain cranial foramina and bony landmarks, along with dried vessels, also apparent in skulls of C. ruemmleri , that signal the derived plan already described in the account of C. ruemmleri .

Each dentary is elongate, with a slim and long tubular ramus between incisors and molar row that gently curves upward at about 30 °, a low ascending ramus, delicate coronoid process, and elongate condyloid projection (figs. 10, 42). The posterior margin between condyloid and angular processes is deeply concave, and a wide concave sigmoid notch separates the small coronoid from the projecting condyloid process. The labial surface of the ascending ramus is smooth, lacking any external indication of the incisor capsule, which is concealed within the dentary and terminates at or near the base of the coronoid process. The dentary of C. ruemmleri is stocky by comparison with that of B. albidens , the ramus between molar row and incisor is shorter and deeper and curves upward at about 45 °, the ascending ramus is higher, the coronoid process more robust, and the condyloid projection wider and stocky (figs. 9, 42); the ascending ramus is smooth in some specimens of C. ruemmleri , but in others is sculptured by a low mound containing the incisor capsule that runs diagonally along the ascending ramus; in both textures, the capsule ends as a large, bony, labial knobby projection ending slightly below, even, or above the top of the sigmoid notch between coronoid and condyloid processes.

Enamel layers of the upper incisors are white or cream, and enamel covering the lower incisors is white (dense orange on uppers and paler orange on lowers in C. ruemmleri ). The incisor faces are smooth, lacking either grooves or shallow sulci. The incisors emerge from the rostrum at a right angle to the occlusal plane of the molars (orthodont configuration, as defined by Thomas, 1919).

The upper incisors of B. albidens have a distinctive shape and extent of enamel cover that are unusual among most murines, that conspicuously diverge from the incisor form in Coccymys ruemmleri (figs. 9, 10). The incisors of B. albidens project straight down from the rostrum (orthodont) rather than curve down and back (opisthodont) as they do in C. ruemmleri , and they are shorter and thinner (in the anterior-posterior plane). Measurements from an old adult B. albidens (AMNH 150541) and an old adult C. ruemmleri (AMNH 151277) are typical: the distance from the anterior edge of the incisor alveolus down to the incisor tip is 4.8 mm in B. albidens and 5.8 mm in C. ruemmleri , and the anterior-posterior thickness (measured near the alveolus well above the wear facet) is 1.2 mm for B. albidens and 1.8 mm for C. ruemmleri . Brassomys albidens has a larger skull than C. ruemmleri (tables 4, 7), so its incisor is both absolutely smaller and smaller relative to skull size. In side view, the white enamel covers one-third to one-half or more of the labial surface of the incisor in B. albidens , even though the incisor is no thicker than the orange enamel layer in C. ruemmleri , where it covers only one-fourth to one-third of the labial surface. Finally, the incisor tips form a gently convex cutting edge in B. albidens , but a straight edge in C. ruemmleri .

Lower incisors of B. albidens are slightly shorter than those in C. ruemmleri (5.9 mm and 6.1 mm, respectively; measured from the ventral alveolar margin to the incisor tip on the same two adults identified previously) and narrower (1.1 mm and 1.5 mm, respectively, measured in the same plane as the uppers). White enamel covers half of the labial incisor surface in B. albidens , but in C. ruemmleri the pale orange enamel covers only a third of the labial surface. The incisor tips form a convex cutting surface in B. albidens , but a straight edge in C. ruemmleri .

Numbers of roots for the first and second upper (maxillary) molars in B. albidens deviate from the pattern considered primitive for murines ( Musser and Newcomb, 1983). Instead of three roots beneath each molar, the primitive pattern in C. ruemmleri , each each first and second upper molar is anchored by four roots: one large anterior, a smaller lingual, and two posterior, each about the size of the lingual. The increase from three roots to four results from a divided posterior root. In many other murines with multiple roots, especially Rattus and its generic allies (in the Rattus Division of Musser and Carleton, 2005), the posterior root is intact and it is the lingual root that becomes divided. Each third upper molar in B. albidens has three roots, and each lower (mandibular) molar has two holdfasts, all representing the generalized murine character state.

The upper and lower molars are brachydont, narrow relative to skull breadth, and abut against one another in each row with slight overlap (figs. 11, 12). The first upper molar forms nearly half of the maxillary toothrow, the second molar slightly more than a third, and the third molar relatively much smaller, about 16% of the row. Proportions are similar for the lower molars. Relative to length of molar rows, size of the first and second molars in C. ruemmleri are similar to those in B. albidens , but the third lower in each jaw is relatively much larger, 20% of the respective molar row.

In contrast to the maxillary cusp rows in Coccymys , which are formed by discrete cusps that have not merged in young animals or only narrowly coalesced in older individuals with more worn coronal surfaces, most of the cusp rows in B. albidens are smooth, without clear indication of their cuspidate origin because the cusps merge early in ontogeny and even in juveniles are no longer discrete entities (a juvenile is shown in fig. 11). Also, the rows of cusps in Coccymys are set farther apart and not as tight as in B. albidens . The occlusal surface of each first upper molar of B. albidens is formed by three broad rows of cusps and a posterior cingulum (see figs 11, 12, where occlusal views of Brassomys are contrasted with those of Coccymys ). The first two rows each conform to a chevron in occlusal view, especially in worn molars, resembling the conformation in Melomys (see the occlusal views of Melomys rufescens in Musser, 1982a: 39). Cusps t1, t2, and t3 that make up the first row are so completely coalesced that their boundaries are obliterated (such cusps are evident in the anterior lamina of C. ruemmleri , and the elongate central cusp t2 is broadly merged with the smaller labial cusp t3 to form a slightly bowed configuration, not a chevron). A posteriorly oriented lingual cusp t4 merged with a large triangular central cusp t5 and elongate diagonally oriented labial cusp t6 forms the second arched row; these cusps in the second row retain a bit of their definition in juveniles but lose their boundaries in older individuals (the second lamina in C. ruemmleri is entirely cuspidate, cusp t5 is narrow, wide, and nearly straight, and the row is gently arcuate, not chevron-shaped). A very large triangular cusp t8 joined to a labially directed much smaller cusp t9 comprises the third row (cusp t8 is compressed, from front to back, in C. ruemmleri , not triangular). In adults, the three cusp rows are connected at their lingual margins where the posterior edge of cusp t1 fuses with a short and high spur projecting from the anterolingual surface of cusp t4, and the posterior edge of that cusp merges with the anterolabial edge of cusp t8; there is no lingual cusp t7 or connecting ridge between cusps t4 and t8. At the labial margin, a short ridge projects forward from the anterior surface of cusp t6 to meet the posterolabial edge of cusp t3 near its base, and the posterolabial edge of cusp t6 broadly fuses with a high enamel spur from the end of cusp t9. Between these lingual and labial connections, each cusp row forms a chevron-shaped basin bounded by enamel ridges, especially high along the lingual margins. The connections are absent or not as pronounced on molars with little wear (in Coccymys , the first cusp row is unattached to the second along the lingual margin; on the labial side a low ridge connects cusp t3 with the front margin of cusp t6 near the cingulum in some specimens, and a low cingular ridge extends forward from the tip of cusp t9 to fuse with the posterolabial margin of cusp t6; the first and second cusp rows do not become basined with wear). The posterior cingulum is small relative to overall occlusal surface, but remains recognizable even after considerable wear.

A large cusp t1 defines the anterolingual border of the second molar, and a much smaller cusp t3 sits on the anterolabial margin; the latter varies from a small recognizable cusp to a nubbin on the cingular margin. Occlusal configurations of the first complete row (formed of cusps t4, t5, and t6) and the posterior row (consisting of cusps t8 and t9) are similar to those exhibited by the second and third cusp rows on the first molar. Also, as in the first molar, a lingual enamel ridge connects the isolated cusp t1 with the anterior margin of cusp t4, and the posterior edge of that cusp is merged with the anterolingual margin of cusp t8; there is no cusp t7. On the labial margin, the anterior face of cusp t6 bears a vertical ridge that meets the small cusp t3 near its base if that cusp is large enough; a comparable ridge projects forward from the anterolabial edge of cusp t9 to the posterolabial margin of cusp t6 (may be absent from one molar but present on its opposite). The posterior cingulum is present but sometimes smaller than its counterpart in the first molar.

A large cusp t1, arcuate anterior row formed by completely fused cusps t4, t5, and t6, and small, chunky posterior lamina (mostly cusp t8, possibly including a small cusp t9, but difficult to identify) form the chewing surface of each third molar. The two cusp rows are evident in the juvenile (fig. 11) but coalese after wear into dentine surround- ed by an enamel ring (the anterior row is larger in C. ruemmleri , and the posterior lamina much larger and wider relative to occlusal surface, reflections of the relatively much larger third molar). A posterior cingulum is absent, and cusp t3 is either absent or undetectable because it has so completely merged with the anterolabial surface of the molar.

The first lower molar in each mandibular row is relatively short and narrow, and its coronal surface is formed by three rows of cusps (fig. 12), their conformation similar to that characterizing Melomys (see illustration of M. rufescens in Musser, 1982a: 39). The front lamina, the anteroconid, is solid, small relative to surface of the first molar, with an obtuse anterior outline, and formed from the complete fusion of a large anterolingual and much smaller anterolabial cusp (no evidence of an anterocentral cusp); the two cusps are completely fused even in the juvenile example of B. albidens (anteroconid in C. ruemmleri is relatively larger and rectangularlike in occlusal outline, consisting of a chunky anterolingual cusp and only slightly smaller anterolabial cusp that are separated along most of their medial borders in unworn and moderately worn teeth but broadly coalesced in older adults where the medial borders are oblitered; fig. 12). A triangular and large fused metaconid and protoconid create the second row of cusps that is separated from the anteroconid (fusing with it only after much wear and usually only narrowly at the labial and lingual margins). A large and oblong entoconid and hypoconid form most of the posterior one-third of the molar, and behind that row of cusps is a wide and triangular posterior cingulum. Cusps forming these two rows behind the anteroconid are angled posteriorly (broad chevron shape in occlusal view) and each pair has completely coalesced along the midline even in juveniles showing little wear (cusps forming comparable laminae in Coccymys display a more linear orientation and fuse only at a later stage of wear than in Brassomys ). The occlusal surface of the somewhat rectangular second lower molar consists of two rows of cusps, each a spread chevron in occlusal view, and a large and elongate posterior cingulum (here also the chewing surfaces in Coccymys , especially that of the posterior lamina, are straighter and the cusps only narrowly merged). The anterolabial margin of the molar supports a small anterolabial cusp that fuses with the protoconid after only little wear; its opposite, the anterolingual cusp, is absent. The chunky third molar has a simple occlusal surface created by two laminarlike cusp rows, similar in configuration to the occlusal topography in Coccymys . An anterolabial cusp is absent from some specimens but present on others, although often inconspicuous because it has nearly melted into the hypoconid. The posterior margin of the tooth is without a posterior cingulum, the usual state in most murines.

Anterior and posterior labial cusplets are absent from the juvenile but can be dectected on a few of the adults where they are indicated by short projections from the anterolabial corners of the cusp rows (a relatively large and discrete posterior labial cusplet is present on each first and second molar in C. ruemmleri ).

Differences between our samples of Brassomys albidens and Coccymys ruemmleri in absolute size of measured external, cranial, and dental variables are evident from the univariate summaries listed in tables 3 and 4. Adult B. albidens have larger bodies but relatively shorter tails than do the samples of C. ruemmleri (table 3) and exceed the latter in dental measurements and all cranial dimensions except for zygomatic breadth, breadth and height of braincase, breadth of zygomatic plate, length of diastema, and length and breadth of incisive foramina in which measurements are about the same or less (table 4).

Proportional differences in cranial and dental measurements are portrayed in a ratio diagram (fig. 43), which was constructed from the measurements taken from the holotype and 14 other adults of Coccymys ruemmleri from Lake Habbema (table 4), and three adults of Brassomys albidens : the holotype from Lake Habbema and two from Bele River, the only adults with intact skulls (table 4). Compared with the sample of Coccymys ruemmleri , the smaller series of Brassomys albidens is proportionally significantly narrower across the zygomatic arches and has a smaller braincase relative to occipitonasal length (length of skull). Brassomys albidens also has a significantly wider rostrum relative to its length, a much narrower zygomatic plate relative to most other cranial dimensions, shorter diastema relative to occipitonasal length and lengths of the palatal bridge and postpalatal region, shorter incisive foramina relative to occipitonasal length, narrower bony palate relative to its length, appreciably wider mesopterygoid fossa relative to breadth of bony palate, larger bullae relative to postpalatal and occipitonasal lengths, and larger molars (as indexed by length of maxillary molar row and breadth of first upper molar) relative to occipitonasal length. Some of these proportional contrasts are easily visualized in figures 9 and 10 where skulls of the two genera are compared. There the heavier rostrum of Brassomys is apparent, and its narrower zygomatic plate, larger bullae, wider mesopterygoid fossa, and heavier molars are clearly evident.

HABITAT: Brassomys albidens is still represented only by six modern specimens, each consisting of a stuffed museum skin and skull. All were collected by members of the 1938–1939 Archbold Expedition to New Guinea during August to November 1938, along the northern ramparts of the Snow Mountains in western New Guinea. The holotype is recorded from the Lake Habbema camp at 3225 m, and the other five are labeled as originating from the Bele River camp at 2800 m (9 km northeast of Lake Habbema; see the gazetteer for Coccymys and map in fig. 3). The landscape in the environs of Lake Habbema is a mosaic of subalpine forest, alpine grassland, and marsh. Descriptions of the environments there were summarized in our habitat account for Coccymys ruemmleri and described in lucid detail by Archbold et al. (1942) and Brass (1941). Here we summarize accounts of the environment at 2800 m and nearby where all but one of the six examples of Brassomys albidens were obtained.

Archbold and Brass (1941:265) described how the entire Bele River valley was insulated from the trade winds by high spurs: ‘‘There was therefore little air movement, and although mists frequently enveloped the more prominent local ridges and sometimes filled the little valleys, there was no very regular massing of clouds in the immediate camp locality.’’ The camp was occupied ‘‘during a period of variable weather, with pleasant sunny mornings and showery afternoons, successions of overcast rainy days, and one dry spell in which no rain fell for four days.’’ Between October 15 and November 9, the mean minimum ambient temperature was 7.0 ° C (range, 4.5 ° –10.0 °), the mean maximum 17.5 ° C (range, 13.5 ° – 21.5 °).

This camp at 2800 m was erected, as described by Brass (1941: 312),

in the bottom of the rather steep-falling little valley of one of the headwater streams of the Bele River. Sharp spur ridges, considerably lower than those that hemmed in the river basin, rose 50 to 100 m. from the bed of the stream on both sides. The country was entirely forested except for landslips along the streams, a few small clearings planted with Pandanus and containing native huts, and occasional wet shrubby openings which had probably been enlarged by the natives who used them for resting places when hunting and traveling in the mountains [fig. 44]. Although well up in the cloud belt, weather conditions were variable, and mists not very frequent or regular in occurrence in the immediate neighbourhood.

‘‘Beech-forests continuous with those of the Bele Camp area [2200 m] and equally tall in the valleys clothed the slopes up to about 3,100 meters, where they met in a narrow ecotone subalpine coniferous forests dominated by Podocarpus papuanus ’’ ( Archbold et al., 1942: 263).

Brass (1941: 313) was impressed with the forests and their zonal components, particularly the magnificent beech ( Nothofagus ) zones:

The forest had many faces. In the usual splendid development, that clothed ridges and valleys alike, the dominants [mostly Nothofagus ] grew to 35–40 m high and up to 1–1.5 m. in diameter above their spurred base, the tallest forest being found on the slopes [fig. 45]. On occasional flat seepage-wet benches in the valleys, the rather stunted forests carried a thick layer of green bryophytes on the ground and lower tree-trunks and held few of the usual ferns and orchids. On narrow ridges, especially above 2850 m., the stand degenerated into a typical open ‘‘mossyforest’’ 15 m. or less in height, with a deep ground cover of brownish hepatics and matted surface roots, the trees themselves blanketed and cushioned with mosses and hepatics and hung with pendulant ferns and orchids to 5– 8 m. of their length [figs 46, 47]. … In general, ground, trees and undergrowth were abundantly mossed down to 2600 m. or lower on the summits of spurs, and about 2700 m. in the valley bottoms.

All six specimens of Brassomys albidens lack any details indicating microhabitat, whether the rats were caught on the ground or above in the forest understory (‘‘Brought in by Natives’’ is notated on the skin tag of AMNH 150923 from 2800 m). The five individuals from 2800 m were certainly caught in forest because beech and ‘‘mossy’’ formations dominated the landscape at the 2800-m camp. We suspect the holotype also came from the forested ridges around Lake Habbema, but no formal record corroborates that assumption.

BIOLOGY: No reproductive data are associated with any of the six specimens representing B. albidens , and direct evidence about diet, which could be gleaned from feeding captive animals or examining stomach contents, is unavailable. While reproductive information is impossible to extract from the dry, stuffed skins (five of the six specimens are male, the sixth is a juvenile female without any sign of teats on the skin— we don’t know if the rat was correctly sexed in the field), we can infer diet drawing from morphology of the skull and dentition. We suspect that B. albidens is an invertebrate predator, based on architecture of the particular cranial, mandibular, and dental traits described below.

Rostrum: Brassomys albidens has a rectangular rostrum that is as deep near the incisors as it is at the level of the nasolacrimal canals, and wide relative to its length, especially compared with the more usual murine configuration, such as that expressed by Coccymys , for example (figs. 9, 10). The relative size of this rectangular bony enclosure may reflect an increase in volume of the nasal cavity and greater expanse of nasal epithelium leading to enhanced olfactory acuity, an advantage in a wet, cool, and mossy microenvironment where the air is saturated with scents of wet moss and decaying vegetation. Most species of murine rodents that live in tropical montane forest and prey upon insects and earthworms (annelid worms) typically exhibit increased volume in the nasal cavity. Either the rostrum projects appreciably forward as a tube beyond the incisors, as in the Sulawesian Sommeromys ( Musser and Durden, 2002) ; or the rostrum is elongate, whether a long and narrow tube between incisors and molar rows, the configuration in the Sulawesian species of Melasmothrix and Tateomys ( Musser, 1982b) , and Philippine Rhynchomys ( Musser and Heaney, 1992; Balete et al., 2007); or the rostrum is less protracted than in those genera, but still relatively long and with the nasals and premaxillaries projecting moderately beyond the incisors, the pattern in the Philippine Archboldomys , for example ( Musser, 1982b; Balete et al., 2006). Modifying the rostrum from the generalized slightly tapered shape seen in Coccymys to a rectangular box may be another solution for increasing nasal volume.

Zygomatic plate: Brassomys albidens has a very narrow zygomatic plate (both in absolute width and relative to size of the skull; see table 4 and fig. 43), its posterior border is directly over the first molar (the anterior third of that tooth), and the ventral zygomatic root bears a robust masseteric tubercle that projects ventrolaterally from the anteroventral edge of the plate (figs. 10, 41), the origin of the superficial masseter muscle. This combination is generally characteristic of tropical murines and Neotropical sigmodontines that feed primarily on invertebrates (some species also include small vertebrates in their diet). Among murines with a similar zygomatic architecture and knobby masseteric projection are the New Guinea hydromyins, species of Hydromys , Parahydromys , Baiyankamys , Crossomys , Leptomys , Paraleptomys , and Pseudohydromys (see cranial illustrations in Musser and Heaney, 1992: 88; Flannery, 1995: 523–526; Helgen, 2005: 7, 2007b; Flannery, 1995, summarizes diets, also see Musser et al., 2008, who provide dietary information for Leptomys ); the Philippine Archboldomys ( Musser, 1982b: 21; Musser and Heaney, 1992: 72; Rickart et al., 1998: 19; Balete et al., 2006; diet is documented by Rickart et al., 1991, Heaney et al., 1999, and Balete et al., 2006); Melasmothrix naso and species of Tateomys , endemics from Sulawesi ( Musser, 1969: 16; Musser and Durden, 2002: 27; see Musser, 1982b, and Musser and Durden, 2007, for diet compositions); and the African Colomys ( Dieterlen, 1983: 85–87) , which preys on limnetic invertebrates and fish ( Dieterlen and Statzner, 1981). Prime examples among Neotropical sigmodontines are the ichthyomyines, species of Ichthyomys, Rheomys, Neusticomys, Anotomys , and Chibchanomys ( Voss, 1988, monographed the group and elaborated on diets). A narrow zygomatic plate and robust, projecting masseteric tubercle are common to all these muroids, but position of the posterior border of the plate relative to the first upper molar varies among the genera, from just anterior to the molar, through the condition in Brassomys albidens , to the end of the upper molar.

Two genera of murine vermivores, the Philippine Rhynchomys and Sulawesian Echiothrix , both predators primarily of annelid worms (earthworms), are an exception to the genera summarized above. The species in each genus possess a relatively narrow zygomatic plate, but only a low, raised area or scoring on the bone identifies the anchor for the superficial masseter muscle, not a large bony projection. But little or no mastication is required by these rats to process their primary food. Echiothrix possesses a full complement of molars in each quadrant of the jaw, but they are very small relative to the large and robust skull ( Musser, 1990). Rhynchomys lacks third molars and the remaining teeth are tiny—nearly vestigial—relative to the large, elongate skull ( Musser and Heaney, 1992; Balete et al., 2007). Using only its incisors, Echiothrix cuts earthworms into sections, which are swallowed intact ( Musser, 1990, and field observations); similar pieces of earthworms have been found in stomachs of Rhynchomys ( Heaney et al., 1999; Balete et al., 2007).

Tropical Indo-Australian murines that are either primarily herbivorous or include fruit and insects in their diets exhibit a wider zygomatic plate relative to skull size and only a rugose spot or small bump (no large, knobby projecting tubercle) at the base of the ventral zygomatic root for insertion of the superficial masseter muscle (our observations). Descriptions of the zygomatic conformation in the three species of Coccymys are examples. Other New Guinea examples can be appreciated by viewing illustrations of skulls in Flannery (1995) of Hyomys (see also the figure in Musser, 1981b: 151), Uromys , Mallomys (see Musser, 1981b: 79), Xenuromys , some Chiruromys and Pogonomys , Melomys , Paramelomys , Pogonomelomys , Abeomelomys , and some Rattus ; species in these genera, at least those with dietary information, are entirely or primarily herbivorous (see summaries in Flannery, 1995). Species of Tarsomys , Limnomys , Batomys , Crateromys , Phloeomys , and some Apomys are Philippine endemics that are entirely herbivorous or consume fruit and invertebrates (see cranial illustrations in Musser and Heaney, 1992; dietary information is record- ed by Rickart et al., 1991, 2003; Gonzales and Kennedy, 1996; Heaney et al., 1999, 2006). Endemic Sulawesian herbivores are Lenomys meyeri and Eropeplus canus (see cranial illustrations in Musser, 1981b); fruit and seeds are preferred by Haeromys minahasae ( Musser, 1990, figured the skull); and a mixed diet of fruit and invertebrates is common to the three species of Margaretamys (skulls are illustrated in Musser, 1982a), along with Bunomys chrysocomus and Maxomys hellwaldii (see cranial renditions in Musser, 1991); dietary information comes from Musser’s feeding in the field and survey of stomach contents. This array of species, in addition to possessing a less specialized zygomatic configuration, also lacks the specializations of the rostrum, incisors, cusp patterns of molars, and dentary configuration seen in the specialized murines described just above and in the paragraphs below.

Incisors: The conformation of the incisors in B. albidens suggests an invertebrate diet. The uppers are short and thin (relative to skull size) and emerge from the rostrum at a right angle. The enamel is unpigmented and wraps around the side to cover up to half of the labial surface of each incisor, and their tips form a gently convex cutting edge. Lower incisors are also relatively short and thin, with unpigmented enamel that covers a comparable extent of labial surface, and the tips are awl-shaped. They emerge from the dentary at a low angle. The positions of the upper and lower incisors relative to the rostrum and dentaries in B. albidens , along with unpigmented or pale incisor enamel (white to cream and pale yellow is the usual range seen in other species) that extends laterally to cover at least half of the labial surface, and a convex incisor cutting edge (whether gently convex as in B. albidens , convex, or markedly so as seen in Melasmothrix and Chrotomys ) is common to the Sulawesian shrew rats Melasmothrix and Tateomys . Melasmothrix naso preys on earthworms and dipteran larvae; the two species of Tateomys are earthworm predators ( Musser, 1982b). The Philippine Chrotomys displays the extreme configuration in which the enamel covers nearly all the labial surface of the upper and lower incisors ( Musser and Heaney, 1992). Chrotomys has an elongate rostrum and upper incisors that are procumbent with a marked convex cutting edge; invertebrates comprise the diet of those species of Chrotomys for which this information is available ( Rickart et al., 1991; Heaney et al., 1999).

Molars: Configuration of the occlusal surfaces on the upper molars may provide supporting evidence for an invertebrate diet. The labial and lingual arms of the chevrons forming the chewing surfaces are connected at their outer margins either by their edges fusing or by a high enamel ridge. The arms of the chevrons in the first and second rows on the first molar, and the first complete row on the second tooth are transformed through wear into chevron-shaped basins bounded by labial and lingual walls; even cusp t1 on the second molar is basined. This wear pattern is not as evident in the juvenile specimen, but is already developed in very young adults (fig. 11). In our experience, basined chewing surfaces, no matter what shape the cusp row takes, are associated with diets composed primarily of invertebrates and small vertebrates. Examples are found in New Guinea hydromyins ( Jackson and Woolley, 1993; Flannery, 1995; Musser et al., 2008), Sulawesi Melasmothrix and Tateomys ( Musser, 1982b) , Philippine Archboldomys and Chrotomys ( Rickart et al., 1991; Heaney et al., 1999; Balete et al., 2006), African Colomys ( Dieterlen, 1983; Dieterlen and Statzner, 1981), and Neotropical ichthyomyines ( Voss, 1988).

Dentary: The shape of the dentary in B. albidens is another indicator of a specialized diet (figs. 10, 42). The ramus between molar row and the base of the incisor is tubular, and the curve made by the ramus and incisor is gently convex, not more than about 30 ° from the horizontal plane. The rest of the dentary is low and elongate. The overall form of the dentary and incisor resembles, but is not as extremely elongated as, the shapes of dentaries in the Sulawesian shrew rats ( Musser, 1982b) and the Philippine invertebrate predators ( Musser, 1982b; Musser and Heaney, 1992; Balete et al., 2006, 2007).

General conclusions: We look at Brassomys albidens as a nocturnal, likely arboreal, invertebrate predator. Length of the tail, which is appreciably longer than the head and body (tables 3, 7, 20), and configurations of the digits and hind feet in B. albidens , along with sharp and recurved claws, are consistent with arboreal activity. A wide and short hind foot relative to body size is characteristic of rats that nest and forage in the forest canopy, or nest on or below the ground but forage in trees. For example, the ratio of hind foot to head and body lengths ranges from 19% to 23% in samples of the endemic New Guinea genera of Chiruromys and Pogonomelomys , from 20% to 24% in the three arboreal species of Sulawesian Margaretamys , and from 17% to 18% in species of the New Guinea genus Pogonomys (percentages were computed from mean values of adults listed in Musser, 1981a and Flannery, 1995). The proportional range in B. albidens is 21% to 24% (computed from values listed in table 20). By contrast, the combined range in our population samples of C. ruemmleri , C. shawmayeri , and C. kirrhos , which have narrow and long hind feet relative to body size and are likely scansorial rather than arboreal, is 24% to 30% (computed from means in table 3). The range is 29% to 30% in the scansorial Lorentzimys nouhuysi , which nests above ground in forest understory and on the ground among rocks (Flannery, 1995). We view a scansorial species as one that forages over the ground (partly terrestrial) and climbs into shrubs, tangles of woody vines, small trees, Pandanus crowns, and lower branches of larger trees within the forest understory (partly arboreal); nests may be anywhere in that vertical range, as is apparently characteristic of L. nouhuysi . The tail is longer than the head and body in all the species in these genera, whether arboreal or scansorial ( Musser, 1981a; Flannery, 1995).

Brassomys albidens certainly seems capable of moving about in the forest canopy, but also in the understory and on the ground. Whether it is wholly arboreal, nesting and foraging in the canopy (our perception of complete arboreality), which describes the behavior of Chiruromys and Pogonomelomys (Flannery, 1995) as well as the species of Margaretamys (Musser’s field observations), or nests in ground burrows but forages in the canopy, which is typical of species of Pogonomys (Flannery, 1995) and members of the Sulawesian Rattus xanthurus group (Musser’s field observations), is of course unknown.

We lack data regarding foraging activities for B. albidens , but again can speculate drawn from its external traits and a cranial character. The large external pinnae and inflated ectotympanic bullae possessed by B. albidens likely provide sharp auditory acuity, which would certainly aid in detecting nocturnal predators, especially owls, but might also be used to listen for invertebrates moving through wet moss, beneath cracked bark on tree limbs, trunks, and woody vines, and in decaying leaf litter, whether on the ground, in the tops of Pandanus , or in epiphytes and ferns in the crowns of trees. Such foraging activities would also be facilitated by sharp olfaction, possibly a function of the expansive nasal cavity in the rectangular rostrum.

We are certain that B. albidens is nocturnal. In our experience, its pelage coloration, body form, and arboreality (if we are correct) are associated with rats that are active during the night rather than during daylight hours, at least in the Asian tropics. The species of Sulawesian Melasmothrix and Philippine Archboldomys are good examples of tropical diurnal rats. These are small-bodied murines in which the tail is shorter than length of head and body or coequal, and upperparts are dark chestnut, and underparts are chromatically similar or slightly paler; all are primarly diurnal ( Musser, 1982b; Heaney et al., 1999). Within the Indo-Australian tropical murine fauna, we don’t know of any arboreal or scansorial murine that is not nocturnal.

Another aspect regarding population biology of B. albidens is its relative abundance in the wild. Whether the few documented specimens of this unusual murine actually represent its rarity in the montane forests of the Central Cordillera, or reflect an artifact of collecting techniques is unknown. We suspect the latter. Small-bodied nocturnal, arboreal rats that are invertebrate specialists are notoriously difficult to trap or snare, and may partly explain the few specimens encountered during the 1938–1939 Archbold Expedition.

If our ecological description of B. albidens is close to reality, it would be, within the endemic murine fauna of New Guinea, the only known arboreal invertebrate specialist. All the hydromyins for which diet has been determined—species of Hydromys , Parahydromys , Baiyankamys , Crossomys , Leptomys , Paraleptomys , and Pseudohydromys —are terrestrial or amphibious, and carnivorous or primarily invertebrate specialists (Flannery, 1995; Helgen, 2005b; Musser et al., 2008). Species in the other New Guinea endemic genera for which dietary data are available, whether terrestrial, arboreal, or scansorial, are either obligate herbivores (forbs, leaves, and grasses), primarily fruit and seed consumers with invertebrates constituting only a portion of their diets (see summaries in Flannery, 1995), or fungal consumers such as Protochromys fellowsi (see following section).

Hopefully, future mammalogical surveys in wet, mossy montane forests of the Central Cordillera in western New Guinea may encounter Brassomys and provide information either bolstering or modifying our hypothesis about arboreality and diet that we propose here based on the six known modern specimens represented by stuffed museum skins and accompanying skulls.

SYMPATRIC ASSOCIATIONS: The six examples of B. albidens were collected during the 1938–1939 Archbold Expedition to the Snow Mountains. Base camp was at Hollandia (today the city of Jayapura) on the coast. Although most of the time there was devoted to construction and organization of the camp, some mammals were collected in the vicinity of Hollandia, in the foothills of the adjacent Cyclops Mountains, and near Sentani Lake. The major thrust of the Expedition, however, as summarized by Archbold et al. (1942: 201–102), was to survey the north slope of the Snow Mountains. … Between the fairly well-known Weyland Mountains to the west and the mountains about the headwaters of the Sepik to the east, the north slope of the central range was practically unknown biologically and very little in any way. … It was the central part of this unknown area, between Mt. Wilhelmina and the Idenburg River, that the expedition studied. This included collections from camps from near sea level to near the upper limit of vegetation. Specialists in mammals, birds, insects and plants spent on the average about a month at each camp.

Between June 1938 and April 1939, 11 camps had been established between the Idenburg River (Bernard camp, 50 m) and the flanks of Mt. Wilhelmina (camp 11, 3800 m). Their locations and elevations are indicated in figure 48, which is a reproduction of the original map of the area published in the expedition summary ( Archbold et al., 1942).

Collection activities along the transect resulted in the most complete inventory of New Guinea mammals made at that time, and is significant to framing the faunal context for B. albidens . About 2484 mammals representing 79 indigenous species of echidna, bats, marsupials, and rodents were obtained between the Idenburg River and Mt. Wilhelmina. The number of specimens of each species collected at the different altitudes is summarized in figure 49.

Collections of mammals were also made in the vicinity of coastal Hollandia, adjacent foothills of the Cyclops Mountains, and nearby Lake Sentani. We excluded these specimens from the transect but listed them in a separate table (table 22). Hollandia and the other two northern coastal collection localities are separated from the wide lowland valley of the Idenburg River by ‘‘a broad tract of low mountain country, completely forested,’’ which attains ‘‘a fairly even elevation of perhaps 800 to 1,000 meters and lay in closely parallel ridges trending eastsoutheast and west-northwest’’ ( Archbold et al., 1942: 207). The northern foothills of the Snow Mountains actually begin on the southern margins of the Idenburg River valley where the mountains rise ‘‘abruptly from the plains’’ ( Archbold et al., 1942: 234). Most of the species collected near or at Hollandia, in the foothills of the Cyclops Mountains, and near Lake Sentani were also obtained in the vicinity of Bernhard camp and the 850-meter camp, and are members of the Membrano River basin lowland fauna in the northern lowlands of New Guinea ( Helgen, 2007c).

We obtained the numbers for each species by counting the specimens stored in cases in the Department of Mammalogy. We either identified or verified determinations by other researchers of all the specimens and generally followed current taxonomy—published or reviews being prepared for publication (see Helgen, 2007c and references cited there) except for the following taxa. We use the marsupial genus Murexia for the species habbema , melanurus , and naso following Krajewski et al. (2007) rather than the genera employed by Van Dyck (2002) for these species. Two rodents indicated as ‘‘sp.’’ in the chart are distinct species, one in Paraleptomys , the other in Pogonomys , and will be described in a coming revision of Paraleptomys (Helgen et al., in prep.) and a systematic review of small-bodied, montane species of Pogonomys (Musser and Lunde, MS) . Helgen identified the specimens of Uromys nero and U. ‘‘ caudimaculatus ’’ (usually lumped together as U. caudimaculatus ), which will be supported in his forthcoming systematic revision of the genus.

Future systematic revision of particular groups will change the views of certain species listed in figure 49. Helgen (2007c), for example, provisionally recognizes Murexia wilhelmina as a distinct species rather than just a population of M. melanurus (we treat the two as the same). Helgen (2007c) also notes the existence of an undescribed species of Microperoryctes in the Snow Mountains, and that particular taxon, plus others now listed as species in the literature— the marsupials, Phalanger gymnotis and Spilocuscus maculatus , the rodents, Paramelomys platyops , Paramelomys rubex , and Pogonomelomys mayeri , and the bat Nyctimene albiventer —is each likely a complex of species.

We omitted dogs and feral pigs, along with the introduced murines, Rattus exulans and R. rattus from our list of species. Specimens of R. exulans were obtained at Hollandia ( Taylor et al., 1982: 327) but not inland at camps along the transect. This species ‘‘is presumably native to southeast Asia, and there is no doubt that it was introduced into much of its present range along with human exploration in prehistoric times’’ ( Taylor et al., 1982: 276). One example of R. rattus comes from the Bele River camp at 2200 m ( Taylor et al. 1982: 330). ‘‘ Rattus rattus has followed human European settlement of New Guinea and adjacent islands and has established itself in almost every lowland Europe- an colonization’’ ( Taylor et al., 1982: 284).

The chart provides an altitudinal picture of the mammal fauna, however incomplete, of which Brassomys albidens is a part, along the northern slopes of the Snow Mountains. During the survey, one species of monotreme ( Zaglossus ), one species of bat, 12 species of marsupials, and 18 species of rodents are recorded from either 2800 m or 3225 m, or both—the altitudes from which B. albidens were collected. A few of these species do not occur in forest ( Rattus richardsoni , for example, inhabits alpine tussock grassland) but most are forest inhabitants. The lack of precise microhabitat information about where specimens were trapped, and the uncertainties tied to altitudes for the material brought in by natives (see habitat account for Coccymys ruemmleri ) likely contribute to inaccuracies. Results of the survey also underestimate the actual number of species occurring in the habitats between the lowlands and 4050 m, particulary for bats and for nonvolant species that are usually difficult to collect, such as those moss mice ( Pseudohydromys ).

Within a zoogeographical context, Brassomys albidens is part of a cluster of murines so far known only from the montane forests and alpine grasslands of the Snow Mountains that includes the following: the marsupials Murexia habbema , an undescribed species of Microperoryctes ( Helgen, 2007c) , and Dendrolagus mbaiso ; and the murine rodents Coccymys ruemmleri , Mallomys gunung , Baiyankamys habbema , Pseudohydromys occidentalis and an undescribed species in that genus ( Helgen, 2007a, 2007b), an undescribed species of Pogonomys (Musser and Lunde, MS) , Paraleptomys wilhelmina and an undescribed species of Paraleptomys (Helgen et al., in prep.), Melomys frigicola , Rattus richardsoni , and Rattus arrogans .

FOSSIL SAMPLES: Brassomys albidens is represented by two right dentary fragments and one left (from three individuals) extract- ed from primary matrix and lag deposits (mostly mud and fossil bone) collected in Kelangurr Cave, 2950 m, situated in a valley confluent with the valley of the West Baliem River and not far (8 km west) from the settlement of Kwiyawagi, which is 60 km or so west of Lake Habbema on the southern slopes of the Snow Mountains ( Flannery, 1999; Hope and Apline, 2007: 251, provide a photograph of the cave). The matrix and lag sediments were obtained from the walls and floor of the first chamber in the cave (chamber 1). The deposits yielded pieces of eight kinds of marsupials, one microchiropteran, and seven species of rodents ( Flannery, 1999: 342). Although a few of the species were represented by material thought to be Recent, the bulk of the specimens extracted from the deposit were heavily mineralized, and two dating techniques indicated that the

TABLE 22 The 432 Specimens Representing 35 Species of Native Mammals Obtained on the North Coast during the 1938–1939 Archbold Expedition to Western New Guinea a Mammals were collected in the vicinity of Hollandia (Jayapura; 02 ° 329S, 140 ° 439E), the nearby Sentani Lake region (02 ° 379S, 140 ° 319E), and foothills of the Cyclops Mountains ( Gunung Cycloop ; 02 ° 319S, 140 ° 389E) .

TABLE 23 Measurements of Lower (Mandibular) Molars and Alveolar Lengths in Samples from the Snow Mountains of Modern and Fossil (Late Pleistocene) Brassomys albidens a deposit in chamber 1 was accumulated between 25,000 and 20,000 B.P. (Late Pleistocene).

All three fossils consist of the anterior portion of the ramus beneath the toothrow. In one of these, the incisor is broken off and a relatively unworn first molar is present; the other two retain intact incisors, complete alveolar sections, but no molars (table 23). The pieces are densely mineralized and the enamel covering the incisor dentine is dark gray, a discoloration from the white typical of modern specimens. The shape of the ramus fragment, conformation of the incisor and extent of its enamel covering, alveolar length of the molar row, and size plus occlusal pattern of the first molar match these traits in dentaries from the modern material collected at Lake Habbema and the Bele River valley.

When Flannery visited Kelangurr Cave in 1994, the area was covered with ‘‘high upper montane forest,’’ but Flannery (1999: 349) noted that a ‘‘number of lines of evidence suggests that the fossil deposit preserved in Chamber 1 of Kelangurr Cave accumulated at a time [25,000 –20,000 B.P.] when the surrounding area was vegetated with tussock grassland and some alpine scrub. Forest was almost certainly absent, as the remains of obligate forest dwelling mammals are almost entirely absent from the deposit.’’ But the information we have presented in the preceding section strongly points to B. albidens as an obligate forest dweller. Possibly the landscape then was a mosaic of forest and tussock grassland, similar to the environment around Lake Habbema (3225 m). Another possibility is that the rats now represented by fossils were taken at lower altitudes in forest and carried by owls to the roost at the cave. Flannery purports that the accumulation of small mammal fossils in the cave sediments is the result of owl predation, whether by an extinct Pleistocene species or a Pleistocene population of the modern sooty owl ( Tyto tenebricosa ). The features of the three B. albidens and 63 Coccymys ruemmleri (see that account) from Kelangurr Cave we studied certainly resemble those of the dentary fragments extracted from modern owl pellets collected on Mt. Wilhelm that were regurgitated by a sooty owl (see the account of C. ruemmleri ).

The three fossils document the presence of Brassomys albidens in the Late Pleistocene, indicate it was likely one of the prey items sought by owls, and acknowledge its membership in a montane fauna mirroring the species composition of the modern assemblage of small mammals now living in landscapes of forest and tussock grassland on the northern slopes of the Snow Mountains (see fig. 49, and Flannery, 1999).

GENERIC COMPARISONS AND OTHER OBSERVATIONS

Our discussion here derives largely from comparing the specimens of Brassomys albidens and examples of Coccymys ruemmleri (our comparative example of the species) with the large collection of endemic New Guinea murines stored at the American Museum of Natural History, focusing on samples of Melomys , Paramelomys , Mammelomys , Protochromys , Abeomelomys , and Pogonomelomys . All comparisons made are among adults.

The species albidens was originally described as a member of Melomys ( Tate, 1951) but the range in anatomical variation and geographic distributions once thought to characterize that genus ( Rümmler, 1938; Ellerman, 1949; Tate, 1951) has been drastically altered and is now partitioned among four monophyletic clusters, each of generic rank: Melomys , Paramelomys , Mammelomys , and Protochromys ( Menzies, 1996; Musser and Carleton, 2005). Furthermore, Mammelomys is not even closely related to Melomys and its allies, as affirmed by morphological ( Menzies, 1996; Musser and Carleton, 2005) and molecular ( Rowe et al., 2008) data. This new and more accurate view does not negate the exclusion of albidens from the altered definition of Melomys , or from any of the other three genera. The combination of slightly overlapping rings of small and flat tail scales, three hairs associated with each scale, no dorsal strip near the tail tip that is devoid of hair and scales; deep and smooth braincase reflecting some cranial flexion; rectangular-shaped rostrum; very narrow zygomatic plate in which the anterior margin does not project forward past the maxillary zygomatic root, the posterior border sitting over the first molar, and the anterolabial surface bearing a projecting knoblike masseteric tubercle; large and inflated ectotympanic (auditory) bulla; generalized (for murine rodents) carotid circulatory pattern; white incisor enamel and its extent over the labial surface; small size of incisors relative to size of skull and mandible; parallel molar rows, fusion of rows along labial and lingual margins on first and second upper molars; and an elongate dentary place albidens outside the range bracketed by the anatomical variation defining the generic boundaries of Melomys , Paramelomys , and Mammelomys .

Protochromys is based on fellowsi , which was named and described by Hinton (1943). Of medium body size and long-tailed, with dense, soft and woolly fur, and buffy gray underparts, P. fellowsi inhabits high montane forests at 2400–2500 m where it has been recorded from only a few places in the Central Cordillera of Papua New Guinea, primarily in the Hagen and Bismarck ranges (Flannery, 1995: 292, as Melomys fellowsi ). When he described Melomys albidens, Tate (1951: 286) had suspected a close relationship between it and Hinton’s Melomys fellowsi , because the latter was also characterized by white incisors and large bullae. Tate had borrowed two specimens of fellowsi (Archbold Expeditions would later acquire a large sample from Mt. Wilhelm) from the British Museum for comparison with his material. After summarizing morphological differences between the two species, which emphasized that the auditory bullae of albidens were even relatively larger than those in fellowsi, Tate wrote that ‘‘One can only conclude that these two species are not only thoroughly distinct from each other but also from any member of the three large groups of Melomys that form the principal part of the genus.’’ Hinton’s fellowsi has white or cream incisor enamel, a narrow zygomatic plate, and relatively large auditory bullae, which are among the traits used by Menzies (1996: 416) in diagnosing the genus Protochromys to embrace fellowsi and separate it from Melomys . Except for these features, however, external, cranial, and dental traits of P. fellowsi resemble those seen in most species of Melomys and Paramelomys (our observations derived from AMNH material; also compare Flannery’s, 1995, illustrations of the skulls of fellowsi , species of Melomys and Paramelomys , and albidens ).

In addition to physical size ( Protochromys fellowsi is a much larger rat; see measurements in Flannery, 1995), P. fellowsi and B. albidens diverge in many anatomical traits. Protochromys fellowsi has larger scales on the tail. There are 11–13 annuli per cm, all pressed into the epidermis; a single hair about the length of a scale springs from each scale; and the tail appears brown, smooth, and naked. The scales are smaller in B. albidens , 17–24 rings of scales per cm; proximal scale rings are pressed into the epidermis but the distal rings have their edges raised, giving the impression of overlap; there are three hairs per scale that are up to 3–4 scales long; and the tail appears somewhat hairy, definitely not naked. Protochromys has a large, stocky skull showing little cranial flexion so that the braincase is low and the dorsal profile (in lateral view) is nearly straight; conspicuous beading runs along dorsolateral margins of the postorbital region and onto the braincase, and the rostrum is slightly tapered in lateral view (marked cranial flexion and high braincase in Brassomys without postorbital or temporal beading or ridging, rostrum rectangular in lateral view). Both Protochromys and Brassomys have a narrow zygomatic plate, but in Protochromys the anterior border projects forward beyond the edge of the dorsal maxillary zygomatic root to form a distinct zygomatic notch, the posterior border of the plate is even with the anterior alveolar margin of the first molar, and the masseteric attachment is a small, rough circular area on the bone surface; in Brassomys , the plate does not project beyond the dorsal zygomatic root, there is no zygomatic notch, the posterior margin of the plate is even with the anterior half of the first molar, and there is a large, projecting knoblike masseteric tubercle—the contrasts are striking. The incisive foramina are shorter relative to diastemal length in Protochromys than in Brassomys . Large bullae relative to skull size are characteristic of Protochromys but only if compared with species in Paramelomys and Melomys of similar cranial dimensions. The bullae are larger in Brassomys , as Tate noted, not only in absolute size but strikingly larger relative to size of skull. An elongate dentary is shared by both genera, but the angular process does not project backward in Brassomys as far as it does in Protochromys , and Brassomys has a relatively longer coronoid process. Among the diagnostic traits Menzies (1996) described for Protochromys was a very narrow alisphenoid strut, but that is probably variable, for the strut is robust and wide relative to skull size—comparatively just as wide as the prominent struts in Coccymys and Brassomys in all 23 examples of Protochromys we surveyed in the AMNH collection.

Protochromys and Brassomys share similar incisor configurations. In both, the uppers are small relative to skull size, and project from the rostrum at a right angle, the enamel is white or cream and forms about half of the labial surface of each tooth, and the cutting edge is gently convex. Lower incisors in each genus have white enamel that covers about half the labial surface of each tooth. In both genera, occlusal outlines of the first two rows on the first molar and anterior row on the second each take the form of a chevron (the last chunky row in each molar is also somewhat chevron-shaped) in which the cusp outlines are completely obliterated, even in young individuals, a configuration similar to the patterns seen in Melomys and Paramelomys , but here the molar similarities end. The rows are not coalesced along either their lingual or labial margins in Protochromys , which is unlike the configuration we described in Brassomys . Occlusal patterns formed by cusp rows on the lower molars are similar in both genera. A trenchant difference between the two genera not mentioned by either Tate (1951) or Menzies (1996) is the number of roots beneath each first upper molar. Both Brassomys and Protochromys possess two roots beneath each lower molar, which is the primitive pattern for murines and muroid rodents in general ( Carleton, 1980; Musser and Newcomb, 1983). The first and second upper molars of Brassomys each has four roots, anterior, lingual, and divided posterior; the third upper molar has three roots, the primitive pattern. In Protochromys , however, each of the upper molars is anchored by four roots, but the third and fourth roots result from a divided lingual anchor, not a divided posterior holdfast; the pattern is specialized, but unlike the derived condition in Brassomys .

Set against this array of external, cranial, and dental contrasts between Brassomys and Protochromys , their shared incisor conformation and elongate dentary are likely convergent. Otherwise, traits associated with skins and skulls of Protochromys point to a close morphological (and likely phylogenetic) relationship with Melomys and Paramelomys , not with Brassomys .

We wondered if the general convergence in incisor anatomy and long and low dentary between Protochromys and Brassomys might indicate an invertebrate diet for the former. Because no dietary information has been recorded for P. fellowsi , we looked to fluid-preserved specimens stored in AMNH, five from Nondugl (183606, 183622, 183628, 183630, and 183632) and three from Mt. Wilhelm (192406, 192408, and 192411). Contents of all nine stomachs contained small rubbery chunks of caps and stalks from mushrooms only, not invertebrates. The pieces showed varying states of digestion. In most of the stomachs the chunks formed a solid mass difficult to break apart with forceps and severely distending the stomach. Some stomachs held only an undigested section of a bulbous stalk (the width of a little finger) and fragments of cap. Black flecks were a part of each mass. The same kinds of mushrooms, judged by texture and color, were found in each stomach, and are of the kinds that emerge from the ground as stalks and cap to produces spores. We sent samples to Tom May (mycologist at the Australian National University, Canberra) who verified the material as ‘‘definitely fungal’’ and noted that two different types were present, members of the phylum Basidiomycota (T. May, personal commun., 2008). Mycophagy is not uncommon among some species of Australian murids ( Claridge and May, 1994) but until now has not been reported in New Guinea rodents.

Judged from our sample (which is small), presence of white incisors, in the case of P. fellowsi , does not point to an invertebrate diet, at least during the period when the specimens were collected. The shape of the zygomatic plate in P. fellowsi —its anterior spine that projects forward beyond the dorsal maxillary root to form a conspicuous notch, and position of the posterior margin relative to the first molar—is not a specialized configuration and similar to the shapes and relative position found in such genera as Rattus and Melomys , for example. This type of plate is also characteristic of two species of Bunomys endemic to Sulawesi that include fungi in their diet, in that case primarily rubbery shelf fungi (Musser’s observations in the field). One of these Bunomys , an undescribed species, relies nearly exclusively on a jelly shelf fungus ( Auricularia sp. ) growing on rotten, wet tree trunks and limbs straddling shaded streams in primary forest.

Abeomelomys sevia is another New Guinea montane endemic murine bearing a superficial similarity to Brassomys . Two names are associated with the species: sevia was initially described by Tate and Archbold (1935) as a species of Melomys , and tatei was later proposed by Hinton (1943) as a species in Pogonomelomys . The name sevia was transferred to Pogonomelomys by Rümmler (1938). Later, Menzies (1990) separated sevia from both Melomys and Pogonomelomys by placing it in the new genus Abeomelomys and did not recognize tatei as a separate entity. Abeomelomys sevia has been recorded only from Papua New Guinea, where it is found between 1400 and 3100 m in the Central Cordillera and on the Huon Peninsula ( Menzies, 1990; Flannery, 1995). A recent phylogenetic analysis of sequences for autosomal nuclear loci and mitochondrial genes linked Abeomelomys to Mallomys and Mammelomys in a clade separate from other New Guinea ‘‘Old Endemics’’ ( Rowe et al., 2008).

As with Brassomys albidens , A. sevia has a thick and soft dorsal coat and ventral pelage composed of gray-based hairs, and a tail much longer than head and body (see the photograph in Flannery, 1995: 264, and description in Menzies, 1990). The upperparts of A. sevia , however, are reddish brown, not dark brown as in B. albidens , and the tail is covered in rows of large, flat, nonoverlapping rings of scales with three short hairs emerging from beneath each scale, and the dorsal surface of the terminal 5 mm at the tip of the tail is hairless and scaleless, indicating ability for dorsal prehensile grasping (see the drawing in Flannery, 1995: 263); the conformation of the tail scales is similar to that seen in species of Melomys , Paramelomys , and Pogonomelomys (rings of scales abut or slightly overlap in B. albidens , the scales are smaller, the scale hairs are relatively longer, and the dorsal surface near the tip remains covered in scales and hairs; the tail appears hairy while that of A. sevia is only slightly hirsute, appearing nearly naked). Abeomelomys sevia is physically larger than B. albidens (for example, range in length of head and body for five adult male A. sevia is 124–135 mm [Flannery, 1995: 264], that for five adult B. albidens is 111–122 mm [table 20], a size difference also reflected in the skull (see photographs in Flannery, 1995: 534; also compare the skull of B. albidens in fig. 10 with that of A. sevia in fig. 50).

The two species resemble each other in overall skull shape, relative length of incisive foramina, and other traits, but diverge sharply in other characters. Incisor enamel, for example, is orange in Abeomelomys sevia , not white as is typical of B. albidens , and the size of the upper and lower incisors relative to the skull and mandible, and extent of enamel covering over the labial incisor surface is like the generalized condition in Coccymys and species in Melomys , Paramelomys , Mammelomys , and Pogonomelomys ; the molar rows noticeably diverge posteriorly but are parallel in B. albidens ; the auditory bullae are moderately small relative to skull size, not large and somewhat inflated; the zygomatic arches converge anteriorly, instead of being either parallel or diverging slightly posteriorly as in B. albidens ; the zygomatic plate is broad with only a roughened site on the bone for attachment of the superficial masseter, and unlike the very narrow zygomatic plate in B. albidens with its knoblike masseteric tubercle; the interorbit is longer, and dorsolateral margins of interorbit and postorbital region are marked by low beading, dissimilar to the smooth borders in B. albidens with its hourglass-shaped interorbit (in dorsal view); and each dentary is stocky, not elongate, with a short ramus between molar row and incisor base, distinct from the slim and long ramus seen in B. albidens . Length of molar rows relative to skull length are equivalent in A. sevia and B. albidens , but occlusal cusp patterns on maxillary molars are somewhat less complex in A. sevia , more similar to those configurations in species of Melomys , Paramelomys , and Pogonomelomys where the rows of cusps take on a definite chevron shape, and its third molar is very large relative to others in the toothrow (figs. 51, 52). Occlusal surfaces of cusp rows in B. albidens are somewhat chevron-shaped, but the third upper molar is relatively much smaller. In addition, cusp rows are merged at their labial borders on first and second molars in B. albidens , and enamel ridges join cusp rows along their lingual margins; this configuration is absent in A. sevia ; cusp t9 projects from the central cusp t 8 in a straight line, rather than diagonally as in A. sevia , where cusp t9 and cusp t8 coalesce to form a large triangle in occlusal outline; and of the two laminae on the third molar of B. albidens , the front lamina is arched, and the back lamina is small, not extending to labial and lingual margins; the counterparts in A. sevia are straight and wide, and reach labial and lingual margins (fig. 51). Finally, the number of roots beneath upper molars differs in the two species. In both B. albidens and A. sevia , each lower molar has two roots, a primitive pattern. The first and second upper molars of B. albidens are each anchored by four roots, anterior, lingual, and divided posterior holdfasts, which is specialized; three roots anchor each third upper molar. Each upper molar of A. sevia has three roots, anterior, lingual, and posterior, the primitive state.

Because ruemmleri was originally described as a species of Pogonomelomys , and Brassomys albidens has been allied with ruemmleri in the past, we briefly address here why albidens is not a member of Pogonomelomys . As currently documented in the published literature, Pogonomelomys contains two white-bellied, long-tailed arboreal species: P. mayeri , with records from hill forests scattered over northern New Guinea, from the Weyland Range in the west to the Huon Peninsula in the east; and the largerbodied P. bruijnii , which is rare in collections and recorded from a few lowland localities in western and southern New Guinea, including the Vogelkop Peninsula and the island of Salawati, the Fly River drainage, and Mt. Bosavi ( Menzies, 1990; Flannery, 1995; Aplin et al., 1999; Helgen, in litt., 2008; Helgen et al., in press). Helgen (2007c), however, notes that inclusion of only two species underestimates the actual species diversity in the genus, and more than these two will be recognized after a new taxonomic review of Pogonomelomys . Both currently accepted species of Pogonomelomys are physically larger than B. albidens (see measurements in Flannery, 1995), have pure white underparts (whitish gray to ochraceous-gray in B. albidens ), a mosaic pattern of tail scales (not abutting or overlapping annuli) with very short scale hairs so the tail appears naked (see drawing in Flannery, 1995: 63), and a short dorsal strip at the tip of the tail that is devoid of scales and hair, reflecting the ability for dorsal prehensility (abutting or slightly overlapping rows of small scales in B. albidens , with long scale hairs so tail appears somewhat hairy, no dorsal prehensile structure near tail tip). Tail anatomy typical of Pogonomelomys is similar to that characterizing Abeomelomys , which also resembles Pogonomelomys in other aspects of external anatomy, as well as many characteristics associated with the skull and dentition (figs. 50–52). Generally, most of the cranial differences that separate Brassomys from Abeomelomys are equivalent to the contrasts between Brassomys and Pogonomelomys . There are three clear exceptions: Pogonomelomys has a stockier skull, much shorter incisive foramina relative to diastemal length, and an appreciably wider zygomatic plate than in either Brassomys or Abeomelomys . Relative sizes of the maxillary molars and the coronal patterns of the cusp rows on these teeth are very similar in Pogonomelomys and Abeomelomys and diverge sharply from the configurations in Brassomys (compare figs. 11 and 12 with 51). Furthermore, the first and second upper molars in Brassomys have four roots; those in Pogonomelomys and Abeomelomys have three roots.

How closely related B. albidens might be to other New Guinea ‘‘Old Endemics’’ is unclear, but no anatomical traits support its membership in any one of the genera to which we have compared it, or to any other New Guinea murine.

COCCYMYS : We have described the taxonomic history of ruemmleri , the type species of Coccymys , in a previous section. It was initially described as a member of Pogonomelomys ( Tate and Archbold, 1941) , a placement always considered unsatisfactory ( Tate, 1951; Lidicker, 1968), and eventually Menzies (1990) designated ruemmleri as the type species of the new genus Coccymys . Menzies (1990: 133) noted that C. ruemmleri ‘‘differs from Pogonomelomys spp. in that the cranium is more rounded and less angular; the anterior palatine foramina are much longer; the zygomata are more tapering, less parallel sided; the molar crowns are more complex; the tail scales are overlapping, not mosaic, and the mammae form 3, not 2, pairs.’’ In addition, both species of Pogonomelomys have pure white venters, contrasting with the whitish gray or grayish white underparts in Coccymys . Rows of tail scales actually abut against one another in Coccymys but they do not form a mosaic pattern.

Some of Menzies’s cranial contrasts are evident in figures 9 and 50 where skulls of Coccymys and Pogonomelomys are presented. There are, however, other sharp cranial differences between the two genera. For example, compared with species of Pogonomelomys , Coccymys has a less robust skull with a higher and wider braincase and deeper occiput relative to overall skull size, reflecting appreciably greater cranial flexion (small braincase relative to interorbital and rostral regions of skull in Pogonomelomys , and shallower occiput); the interorbital and postorbital regions are short relative to the large braincase, their dorsolateral edges smooth, and the interorbit is hourglass-shaped in dorsal view (interorbital and postorbital form a long bridge, in dorsal view, connecting rostrum to braincase, its sides are either straight or diverge posteriorly, and the dorsolateral margins are marked by high shelflike ridges in Pogonomelomys ); the anterior margin of the zygomatic plate barely projects forward past the maxillary zygomatic root, so the zygomatic notch is slight or not present (deeper zygomatic notch in Pogonomelomys , reflecting greater extension of the leading edge); posterior edge of the bony palate is even with the backs of the third molars or extends slightly beyond them (margin even with anterior edges of the third molars in Pogonomelomys ).

The molars of Coccymys are wide relative to their lengths, but relatively narrower in Pogonomelomys . Occlusal surfaces of upper molars in all three species of Coccymys are ‘‘more complex’’ in the sense that each row of cusps maintains its cuspidate origin even after much wear compared to the coronal surfaces of the teeth in Pogonomelomys where the cusp patterns are evident on unworn and slightly worn molars, but tend to disappear with wear into chunky chevrons (compare fig. 11 showing molar of Coccymys with fig. 51 where the upper molars of Pogonomelomys are illustrated). Shape of the anterior lamina on the first molar is very similar in the two genera, a posteriorly oriented cusp t1 and nearly straight segment formed by coalesced cusps t2 and t3. The second lamina on that tooth in Coccymys is, except for cusp t4, essentially straight and the cusps remain discrete (narrowly merged) even with significant wear. The second row in Pogonomelomys , by contrast, has the form of a chevron, and even with little wear cusp t4 abuts the central cusp t5, which in turn is fully fused with the labial cusp t6 (as in the young example of P. bruijnii ); in older animals (the example of P. mayeri ), cusp t4 is completely merged with cusp t5. In the third lamina in Coccymys , cusp t8 resembles a somewhat distorted diamond shape in occlusal outline, and cusp t9 projects in a straight line to the labial margin of the molar. In Pogonomelomys , cusp t8 is more triangular in outline partly because cusp t9 has so completely merged with cusp t8 that its boundaries are obliterated, even in young rats (as shown in the young P. bruijnii pictured in fig. 51); that portion of the third lamina that represents cusp t9 has a posterolabial orientation, not one straight to the labial side. Connecting cusps t4 and t 8 in Coccymys is a high enamel ridge (discernable even in worn molars)—no such structure is present in the species of Pogonomelomys .

The shape of the anterior row of cusps on the second molar is similar in both Coccymys and Pogonomelomys in that cusp t4 is long and narrow, and cusp t5 is elongate, a conformation apparent in little worn teeth ( P. bruijnii in fig. 51) but less evident in worn molars ( P. mayeri in fig. 51). Cusp t6, however, is discrete and much larger relative to cusp t 5 in Coccymys and evident after much wear; cusp t6 is relatively smaller in Pogonomelomys and is already completely merged with cusp t 5 in young animals, and in older rats the row assumes the shape of a thin, widely spread chevron. The posterior cusp row in Pogonomelomys is similar in occlusal outline to its counterpart on the first molar and differs from the configuration on the molar of Coccymys in the same ways. A strong ridge connects cusps t4 and t8 on the second molar in Coccymys , but not in Pogonomelomys . There is a low cingular extension of cusp t8 that is oriented anterolingually, but it does not form a high ridge or cusp in Pogonomelomys . Enamel ridges connect labial margins of the cusp rows in Coccymys (figs. 11, 51), but comparable structures are absent in Pogonomelomys .

Length of the third upper molar relative to the others in the row is similar in Coccymys and Pogonomelomys , but this tooth is more square in the latter due to the wider laminae that form it. The occlusal cusp patterns are similar, except that the anterior and posterior laminae are parallel to one other or nearly so, at least in P. mayeri .

Except for the relatively much larger third molar possessed by Pogonomelomys , its occlusal patterns formed by cusp rows on the first and second upper molars resemble those simple (no lingual or labial connecting ridges) chevron-shaped patterns characteristic of Melomys (see the molar row of M. rufescens illustrated in Musser, 1982a: 39) and Abeomelomys . The more complicated, gently arcuate clearly cuspidate patterns in Coccymys are strikingly divergent.

Coronal laminar patterns of the lower molars are generally similar in the two genera but diverge in five traits (figs. 12, 52). As with the upper (maxillary molars), cusps forming the laminae in Coccymys are better defined and not as fully coalesced along the midline of each molar compared with the completely fused cusps of even young animals in Pogonomelomys , and the molars are wider relative to their lengths. Coccymys has a roughly rectangular anteroconid formed by a large anterolingual cusp and slightly smaller anterolabial cusp, which are separate in young rats but fuse completely in older individuals. The anteroconid in Pogonomelomys is smaller relative to the molar outline, and consists of a large anterolingual cusp fused with a very small anterolabial cusp; outlines of the two cusps can be seen in slightly worn molars (as in P. bruijnii in fig. 52) but are quickly lost with age. An anterolabial cusp forms the anterolabial border of the second molar in Coccymys , but is absent from the tooth in Pogonomelomys . Posterior labial cusplets are prominent along the labial margins of the first and second molars in Coccymys , but are not present in Pogonomelomys . Finally, the posterior lamina on the third molar of Coccymys is wider, extending from labial to lingual margins, while its counterpart in Pogonomelomys is narrower, its labial margin set in from the labial border.

Both genera exhibit the same number of molar roots: three beneath each upper, two anchoring each lower—the primitive configurations.

Except for overall size, the dentaries of Coccymys and Pogonomelomys are much alike (figs. 9, 50). Both are robust with a thick ramus between the molar row and incisor alveolus, expressing the same degree of curvature from the base of the ramus to the incisor tip. Both have a high ascending ramus, prominent coronoid process, and robust condyloid projection. There are two evident differences. First, the posterior concave margin of the dentary between the condyloid and angular processes is shallower in Coccymys . Second, the end of the incisor capsule is a large knob that projects outward from the labial surface of the ramus and its round top is about even with the margin of the sigmoid notch between coronoid and condyloid processes. The end of the incisor alveolus lies within the dentary in Pogonomelomys , and its terminus is indicated by a slight swelling at about the middle of the base of the coronoid process and well below the margin of the sigmoid notch.

Lidicker’s (1968) comparative analyses of phallic morphology among New Guinea murines is the only published comparative inquiry related to anatomical systems not associated with skins and skulls. He examined specimens of Pogonomelomys mayeri and ‘‘ Pogonomelomys ’’ ruemmleri and exclaimed that ‘‘The penis of P. ruemmleri is the most distinctive and unique of all the native forms examined’’ (p. 630) and did not consider P. mayeri and ruemmleri to be closely related, an estimate substantiated by comparisons of external, cranial, and dental traits.

When Tate (1951: 316) expressed his unease with the generic allocation of ruemmleri to Pogonomelomys , he thought there were two subdivisions within the genus, the ‘‘ mayeri-bruijniii group,’’ which are the only two species that currently constitute Pogonomelomys ( Menzies, 1990; Musser and Carleton, 2005), and the ‘‘ sevia-rümmleri group,’’ which has been split into two genera, Abeomelomys and Coccymys , with the type species sevia and ruemmleri , respectively ( Menzies, 1990; Musser and Carleton, 2005). Tate’s grouping of ruemmleri and sevia compels us to compare Coccymys with Abeomelomys . Aside from ventral fur with gray bases and unpigmented tips clothing underparts of head and body, which is like Coccymys , Abeomelomys closely resembles Pogonomelomys in morphology of external traits, skull and mandible characters, and dentition, and the features associated with these different anatomies differ from comparable morphology in Coccymys in much the same way and degree reflected by the contrasts between Coccymys and Pogonomelomys (compare figs. 9, 11, and 12 with 50– 52). There are a few exceptions, and in these traits Abeomelomys resembles Coccymys . Compared with Pogonomelomys , the skull of Abeomelomys appears less robust overall; the interorbital and postorbital regions are shorter and their dorsolateral margins smooth, without the shelflike ridges seen in Pogonomelomys ; the anterior margin of the zygomatic plate is about even with the leading edge of the maxillary zygomatic notch, so the zygomatic notch is barely evident; and the incisive foramina are longer relative to diastemal length, their posterior margins almost reaching the anterior alveolar edges of the first molars.

Molars of Abeomelomys and Pogonomelomys are similar (figs. 51, 52). In both, the first and second upper molars are long relative to their respective widths, and the third upper molar is large relative to size of others in the row; occlusal patterns formed by laminae on the upper molars are basically like the patterns in Pogonomelomys , and neither has any sign of lingual or labial connecting ridges. Lower molars are also equivalent in occlusal patterns. The posterior lamina on the third molar is relatively wider in Abeomelomys (extending to both labial and lingual margins), but that is the basic configurational difference. Each upper molar has three roots and each lower is anchored by two roots in both genera. As with most skull traits, contrasts between Coccymys and Pogonomelomys are also reflected in the dental dissimilarities between Coccymys and Abeomelomys .

Coccymys and Abeomelomys differ somewhat in sperm morphology ( Breed and Aplin, 1994). The spermatozoal anatomy in Coccymys is ‘‘typified by a long and narrow, falciform sperm head and a long flagellum with indistinct boundary between mid- and principal pieces’’ (p. 24) and no accessory ventral processes, a configuration also found in Xenuromys barbatus , Lorentzimys nouhuysi , and both species of Mammelomys among the New Guinea endemics that were sampled as well as three species of endemic New Guinea Rattus . In contrast to these genera, Abeomelomys sevia ‘‘has a very distinctive sperm head, with a fairly broad lateral surface and a long apical hook’’ (p. 23). Sperm morphology of Abeomelomys is, according to Breed and Aplin (1994), somewhat intermediate between the conformation in Coccymys and that typical of Hyomys and Anisomys , both of which have a superficial ‘‘ Rattus -like’’ spermatozoal anatomy. Neither Coccymys nor any of the other genera mentioned above have a sperm head with two accessory ventral hooks, which is characteristic of nearly all native Australian murines (except species of Rattus ), Australian and New Guinea hydromyins, and the New Guinea Mallomys , Pogonomys , and Chiruomys ( Breed and Aplin, 1994; Breed, 1997).

There is no question that Coccymys ruemmleri , along with its vicariant relatives, C. shawmayeri and C. kirrhos , are members of a generic clade different from that identified by the labels, Pogonomelomys and Abeomelomys . Morphologies associated with skins, skulls, and dentition do not support a ‘‘ sevia-rümmleri group’’ ( Tate, 1951) or, in combination with phallic and spermatozoal traits, inclusion in Pogonomelomys .

COCCYMYS AND BRASSOMYS : Although albidens was first described as a member of Melomys ( Tate, 1951) , and subsequent assessments of its phylogenetic affinities suggested a tie to Coccymys ruemmleri (Flannery, 1990; Menzies, 1990; Musser and Carleton, 1993), our comparisons between the two enumerated in a previous section contradict such a close relationship. Some traits associated with skins and skulls are shared by Brassomys and Coccymys , but these are largely primitive (see Carleton, 1980; Musser and Newcomb, 1983; Carleton and Musser, 1989; Weksler, 2006): three hairs per tail scale and thin, slightly raised tail scales; lengths of digits relative to one another; full complement of palmar and plantar pads; skull with smooth interorbital and postorbital dorsolateral margins, with the interorbit hourglass-shaped in dorsal view; cranium smooth without temporal ridging or only slight salients, and inconspicuous lamboidal beading; narrow zygomatic plate with a vertically straight anterior margin that does not project forward past the leading edge of the dorsal maxillary root or barely does so; prominent alisphenoid struts; generalized murine carotid arterial pattern; posterior cingulum on first and second upper molars; and two roots beneath each lower molar. Brassomys and Coccymys also share small body size; thick and soft fur (lax in Coccymys , woolly in Brassomys ), brownish dorsal coats (in C. ruemmleri and C. shawmayeri , not C. kirrhos ), soft and dense ventral pelage with gray-based hairs everywhere; long facial vibrissae; and a brownish tail that is longer than head and body. Polarities of these traits, however, are difficult to assess without a comprehensive phylogenetic analysis of endemic New Guinea murines. Dark, thick, and soft fur, for example, is common to small-bodied species inhabiting cool and wet forests at high altitudes. A tail longer than head and body is possessed by many scansorial and arboreal New Guinea murines. Menzies (1990: 132) noted that Coccymys ruemmleri and the montane Pogonomys sylvestris (which also has soft and thick pelage, is small in body size, with a tail much longer than the head and body), are extraordinarily similar in external physical characteristics and fur coloration, yet cranial and dental features, along with phallic and spermatozoal traits, indicate a distant relationship. Relatively long incisive foramina and a large squamosal foramen are shared by Coccymys and Brassomys but their polarities are equivocal (see Carleton, 1980; Musser and Newcomb, 1983; Weksler, 2006). A conspicuous degree of cranial flexion characterizes skulls of both Brassomys and Coccymys , particulary compared with the flatter crania in Pogonomelomys , Protochromys , Melomys , Mammelomys , Uromys , and other New Guinea endemics (see skull illustrations in Flannery, 1995), which is a specialization ( Carleton and Musser, 1989). The flexion is somewhat more pronounced in Coccymys , however, and that in Brassomys resembles the skull outlines (in lateral view) of other small-bodied murines such as Pogonomys sylvestris ; so as an indicator of relationships this trait is ambiguous.

Occlusal surfaces of upper molars in Coccymys and Brassomys are derived compared with the patterns in early murines such as the late Miocene Chinese Linomys for example ( Storch and Ni, 2002), but in strikingly divergent ways (figs. 8–10), which are reflected in the comparisons described in a previous section. Two dental characteristics are shared by Coccymys and Brassomys . Occlusal outlines of the posterior cusp rows on the first and second upper molars are alike in both genera, formed from a large cusp t8 and much smaller cusp t9 that is broadly united with cusp t8 and projects labially in a straight line. This orientation of cusp t8 relative to cusp t9 is likely primitive; it characterizes some of the earliest murines, such as the Chinese late Miocene Linomys and Pakistan Progonomys of similar age, and even the middle Miocene Antemus chinjiensis , considered by most researchers to be the earliest murid ( Jacobs, 1978; Storch and Ni, 2002). The other shared trait is an enamel ridge connecting the labial margins of the second and third rows on the first upper molar and comparable rows on the second along the labial margin (fig. 11). This is a derived (apomorphic) configuration, but its significance in reconstructing the degree of phylogenetic affinity between Coccymys and Brassomys is difficult to assess without a comprehensive survey of character states in ‘‘Old Endemic’’ New Guinea murines that would be employed in a phylogenetic analysis.

Aside from the shared cranial and dental traits described above, most of which are primitive (the polarity of others is ambiguous and the significance of a few unknown), the postulated affinity between Brassomys and Coccymys evaporates in the face of their cranial and dental divergence. The specialized morphology of the incisors and mandible, and marked degree of ectotympanic bullar inflation (and large external pinnae relative to body size) indicate that Brassomys is morphologically unique among endemic New Guinea murines, and not clearly a part of any previously identified monophyletic group, at least at the level of divergence we identify as generic.

Within the diversity of endemic New Guinea murines, the phylogenetic affinity of Coccymys is puzzling, but how Brassomys fits within this fauna remains a mystery. Coccymys ‘‘ ruemmleri ’’ was sampled by Watts and Baverstock (1994; their sample was actually C. shawmayeri ) in their study of the interrelationships among species of New Guinea murines using microcomplement fixation of albumin to measure immunological distances among genera. Their results identified three groups of genera: the ‘‘ Hydromys Clade,’’ containing Hydromys , Parahydromys , Leggadina , Mesembriomys , Xeromys , Crossomys , Leptomys , Pseudohydromys (including Mayermys and Neohydromys ; see Musser and Carleton, 2005), Uromys , Solomys , and Melomys ; the ‘‘ Anisomys Clade,’’ which included Coccymys , Chiruromys , Pogonomys , Hyomys , Macruromys , and Mallomys ; and a ‘‘ Lorentzimys Clade’’ containing only Lorentzimys . According to their one-way measurements of immunological distances, ‘‘ Coccymys and Hyomys showed no clear affinities with any genus, but appear to be distinctive members of the Anisomys clade’’ (p. 298). They also noted that ‘‘ Coccymys ruemmleri was available only as an antigen and its placement can only be considered indicative. It proved difficult to place but, on balance, seems to be a member of the Anisomys clade. … However, given the immunological distances recorded within what are currently thought of as Uromys and Melomys … this placement must remain very tentative’’ (p. 301). Results generated by Watts and Baverstock are at odds with the clustering proposed by Lidicker (1968), based on phallic anatomy, in which ‘‘ ruemmleri ’’ was included in a ‘‘ Uromys group’’ that also contained species of Melomys , Uromys , Hyomys , and Pogonomelomys . Results from a comparative study of sperm morphology that included Coccymys (as ruemmleri , but the sample was C. shawmayeri ) were discordant with both Lidicker’s arrangement and the hypothesis presented by Watts and Baverstock, as well as being ambiguous in identifying the closest phylogenetic relative of Coccymys ( Breed and Aplin, 1994; Breed, 1997). Along with the New Guinea endemics, Abeomelomys , Anisomys , Chiruromys , Hyomys , Macruromys , Mallomys , Mammelomys , Pogonomelomys , Pogonomys , and Xenuromys , and the Timorese Coryphomys, Musser and Carleton (2005: 904) included Coccymys in a ‘‘ Pogonomys Division’’ of Murinae , which admittedly is built on the meager phylogenetic evidence available in the literature and personal observations, but at the time seemed a reasonable proposal to test. Recent phylogenetic analysis of nuclear and mitochondrial genes breaks the ‘‘ Pogonomys Division’’ into two separate clades, one containing Anisomys , Macruromys , Chiruromys , Hyomys , and Pogonomys (and also includes Lorentzimys , which Musser and Carleton, 2005, placed in its own division), the other composed of Mammelomys , Abeomelomys , and Mallomys ; Coccymys was not sampled ( Rowe et al., 2008). Currently, all data indicating that Coccymys is more closely related to Melomys and its relatives ( Uromys , Paramelomys , Protochromys , and Solomys ) than to at least some members in the ‘‘ Pogonomys Division’’ is ambiguous.

Where Brassomys would fit within a pattern of phylogenetic reconstruction of endemic New Guinea murines is unknown. Relative lengths of tail and hind feet suggest specializations for arboreal habits, and the long, dense, somewhat woolly fur reflects an adaptation to montane environments, but otherwise its external form does not exhibit any striking specializations. The tail, for example, lacks any anatomical indication that it can be used as a prehensile organ, and it is covered in annuli of small scales, each scale supporting three hairs, which is a primitive pattern for murines. The skull and teeth retain a constellation of primitive characteristics (see above), but the combination of a rectangular rostrum, relatively narrow zygomatic plate and its orientation relative to the first molars, knoblike masseteric tubercle, configurations of upper and lower incisors, inflated bullar capsule, and divided posterior root beneath each first and second molar is unique to B. albidens . Uncovering the phylogenetic affinities of Brassomys , as well as Coccymys , will require rigorous phylogenetic analyses utilizing data derived from anatomical surveys and molecular sources. Perhaps we can state that with the exception of hydromyins ( Hydromys , Baiyankamys , Microhydromys , Parahydromys , Crossomys , Leptomys , Paraleptomys , Pseudohydromys , and Xeromys ), all of which are terrestrial or amphibious, and invertebrate or small vertebrate predators, B. albidens may be one of the few arboreal invertebrate predators among the other ‘‘Old Endemic’’ New Guinea genera (which exclude endemic Rattus ). In New Guinea ‘‘Old Endemic’’ genera, the diet, at least for those species for which dietary information is available, consists primarily of fruits, forbs, leaves, and grasses, as well as mushrooms.