Bunomys andrewsi ( Allen, 1911 )

Musser, Guy G., 2014, A Systematic Review Of Sulawesi Bunomys (Muridae, Murinae) With The Description Of Two New Species, Bulletin of the American Museum of Natural History 2014 (392), pp. 1-313 : 169-208

Bunomys andrewsi ( Allen, 1911)

Mus andrewsi Allen, 1911: 336 .

Rattus adspersus Miller and Hollister, 1921a: 71 .

Rattus penitus inferior Tate and Archbold, 1935a: 6 . Rattus penitus heinrichi Tate and Archbold, 1935a: 6 .

HOLOTYPE: USNM 175899 , the skin and skull of an old adult male (original

number 25) collected December 13, 1909, by R.C. Andrews. In the original description, Allen (1911: 336) referred to the holotype as 32193. That is a misprint for 31293, which is the number given to the specimen by staff at the American Museum of Natural History, where it was initially deposited and cataloged. Standard external measurements, oth- er relevant data, and measurements of the skull and dentition are listed in table 40. The skin is overstuffed but intact, the cranium and mandible are complete except for missing hamular processes in the pterygoid region, all incisors and molars are present (fig. 60).

TYPE LOCALITY: Pulau Buton (05 ° 00 9 S, 122 ° 55 9 E; locality 23 in the gazetteer and on the map in fig. 50) southeastern region of Sulawesi, Propinsi Sulawesi Tenggara, Indonesia .

EMENDED DIAGNOSIS: A member of the B. fratrorum group resembling the other species in that assemblage in body size ( LHB 5 137–195 mm, WT 5 95–222 g, ONL 5 37.1–45.5 mm) and further characterized by the following combination of traits: (1) a relatively long muzzle and broad face, and stocky body; (2) dorsal fur dense, lustrous, dark brown speckled with buff and black, ventral fur grayish white to buffy gray or ochraceous gray, digits white, dorsal surfaces of carpal and metacarpal regions typically white, range from grayish white to grayish buff in some samples; (3) claws on front feet short but not as short and delicate as in B. penitus , scanty ungual tufts at bases of front and hind claws; (4) tail shorter than length of head plus body ( LT / LHB 5 75 % – 92 %), dark brown on dorsal surface, white through mottled brown to solid brown on ventral surface (5) white tail tip uncommon (characterizes 20 % of 127 specimens) and when present is short relative to tail length (mean 5 7.7 %, range 5 1 % –13 %); (6) testes moderately small relative to body size (8 % – 15 %); (7) shape of sperm head similar to that of B. chrysocomus but shorter and slightly wider; (8) robust skull with a moderately long and narrow rostrum, wide upright zygomatic plate, long incisive foramina (actually and relative to length of skull), and moderately large ectotympanic bulla relative to skull size; (9) molars moderately large relative to size of skull and mandible; (10) cusp t3 occurs frequently on second upper molar (78 %) in sample from southwestern peninsula but infrequently (17 %) elsewhere; (11) anterolabial cusp characteristically present on second lower molar (94 % of sample) but occurs infrequently on third lower molar (18 %); (12) anterior labial cusplets typically absent from first lower molars, posterior labial cusplets typically present on first and second molars;

and (13) karyotype, 2N 5 42, FNa 5 56, FNt 5 58.

GEOGRAPHIC AND ELEVATIONAL DISTRI- BUTIONS: Labuan Sore (00 ° 37 9 S, 120 ° 03 9 E) at the southern terminus of the northern peninsula is the northernmost record of B. andrewsi . South of there voucher specimens indicate the species occurs in lowlands and

mountains of Sulawesi’s core, on the eastern peninsula, the two southern peninsulas, and on Pulau Buton off the coast of the southern margin of the southeastern peninsula (see the map in fig. 50).

Reliable elevations associated with each specimen document 30 m to 1600 m as the range in which most collections have been made. The upper limit is anchored by animals recently collected by Kevin Rowe (Museum Victoria) and Anang Achmadi (Museum Zoologicum Bogoriense) from the Mamasa area on Gunung Gandangdewata (see the gazetteer and table 6). Most of the specimens reported from there are from tropical lowland evergreen rain forest, but some were taken in the transition between lowland evergreen and montane forest habitats.

Collections of small mammals from the northern peninsula north of Labuan Sore (at the base of the peninsula) do not contain examples of Bunomys andrewsi . The narrow stretch of coastal lowlands and mountains between Labuan Sore and the northwest curve of the peninsula has not been adequately (if at all) surveyed for small mammals; whether B. andrewsi occurs there is unknown. The remainder of the northern peninsula, however, has been surveyed periodically by different collectors dating from the late 19th century and into the 20th century (the Sarasins, G. Heinrich, H.C. Raven, and J.J. Menden, for example), but no collections made then, some of which are large in number of specimens and diversity of taxa, include B. andrewsi . The northeastern section of the northern peninsula, that stretches east of the Gorontalo region, has been the focus of relatively intense collecting efforts by Dutch, English, German, Indonesian, and North American naturalists, and only B. fratrorum and B. chrysocomus have been documented from this region (see the gazetteers and distribution maps for these species).

The peninsular neck between Palu and Parigi, where Labuan Sore is located, may approximate the natural northern range boundary for B. andrewsi . That region is the southern limit for the geographic distribution of the tree squirrel, Prosciurillus leucomus (at Bumbarujaba, 00 ° 43 9 S, 120 ° 04 9 E; Musser et al., 2010) and confluence of the ranges for the northern peninsular Macaca hecki and southern Sulawesi Macaca tonkeana ( Fooden, 1969; Watanabe et al., 1991; Bynum et al., 1997; Groves, 2001). Results of future surveys for small mammals in this southern region of the northern peninsula would be most revealing.

SYMPATRY WITH OTHER BUNOMYS : Except for B. fratrorum , which is restricted to the northeastern end of the northern peninsula and is sympatric only with B. chrysocomus (see that account), the ranges of all other species of Bunomys fall within the broad geographic distribution for B. andrewsi (table 6). But little or no overlap (syntopy) characterizes the ranges of B. andrewsi and the other species. The usual pattern has populations of B. andrewsi occurring in lowlands and those of other species existing at higher elevations (elevationally parapatric). Three pairs of distributions form examples. At the western margin of the eastern peninsula, B. andrewsi has been collected in the lowlands (Sungai Ranu, 50– 100 m) and B. chrysocomus and B. prolatus only much higher on the adjacent Gunung Tambusisi (1372–1830 m). On the western side of the southeastern peninsula, B. andrewsi is documented from Wawo and Masembo in lowlands (50–550 m) adjacent to Pegunungan Mekongga, and B. penitus and B. chrysocomus are found in montane forest on that mountain (1500–2000 m). Bunomys andrewsi and B. coelestis are found on Gunung Lompobatang at the southern end of the southwestern peninsula where B. andrewsi was collected on the lower flanks up to 1100 m and B. coelestis higher on the volcano between 1830 and 2500 m.

Certain collections of B. andrewsi and B. chrysocomus provide the only reliable syntopic distributional records between the former and any other species of Bunomys (table 20). Both species were collected in the northern part of the west-central mountain block at Bakubakulu, 600 m, in the Puro Valley. Along my transect between 290 and 675 m I collected examples of each in the same trapline. I did not encounter B. andrewsi at higher elevations where B. chrysocomus was common. Subfossil remains of both species have been excavated from Ulu Leang I, a cave on the coastal plain at the southern end of the southwestern peninsula (see the account of B. chrysocomus for additional information covering distributional sympatry or overlap between that species and B. andrewsi ).

Bunomys andrewsi and B. penitus are regionally sympatric in the west-central mountain block but generally occur at contrasting elevations and forest formations. Most samples of B. andrewsi were obtained through an elevation range extending from about 100 to 1000 m and were caught in habitats associated with tropical lowland evergreen rain forest. Bunomys penitus is tied to montane forests: 1285–2287 m brackets the elevations at which samples were collected. There are no reliable records that document collection of the two species at the same place (syntopic).

Bunomys andrewsi is locally sympatric with B. penitus and B. torajae on Gunung Gandangdewata in the Quarles Range in the southern part of the west-central mountain block. Bunomys penitus and B. torajae were encountered in montane habitats between 2000 m and 2600 m, B. andrewsi was collected at 1600 m in tropical lowland rain forest.

DESCRIPTION: The original description of Mus andrewsi , obviously named after the collector, is based on two specimens, an old adult and a young adult. The older one is the holotype and Allen (1911: 336) described it as follows:

Fur long and soft, without spines. General color of the whole dorsal region cinnamon rufous finely grizzled with black-tipped and whitetipped hairs, the former prevailing, the hairs being grayish plumbeous at base darkening to blackish and broadly ringed subapically with pale rufous, and for the most part tipped narrowly with black, mixed with many wholly black-tipped hairs and a few light-tipped hairs; sides of the body similar in color to the back but much lighter; ventral surface whitish gray, the hairs darker but still light gray basally; top of head to nose like the back or slightly lighter; limbs like the sides of the body; feet fleshcolored, scantily clothed with short glistening white hairs; ears large, naked, much longer than broad, dark brown above and on sides, flesh color below and all around at the extreme tip.

Allen then provided some measurements, noted a few features of the younger specimen and remarked that

This species has many points of resemblance to Mus luzonensis Thomas , not only in coloration and proportions but in cranial characters. The type shows a tendency to albinism through the presence of scattered white hairs and one or two small bunches of white hairs on the back, and an irregular submedian line of wholly white hairs on the ventral surface.

Adult Bunomys andrewsi has a broad head, unpigmented rhinarium and lips, a stocky body, and in physical size (LHB 5 137– 195 mm, LT 5 110–167 mm, LHF 5 35– 44 mm, LE 5 22–30 mm, W 5 95–222 g, ONL 5 37.1–45.5 mm) resembles the other members of the B. fratrorum group (see the portrait of B. andrewsi in fig. 6). The dorsal coat is soft and moderately long (12–15 mm long on specimens taken in lowlands, up to 20 mm on rats from higher elevations). Guard hairs barely project beyond the overhair layer so the surface of the coat appears smooth; the texture is similar to that in B. chrysocomus . Overall color is a rich dark brown or brownish gray speckled with buff and black (underfur is gray and tipped with pale buff, overhairs are gray basally and tipped with buff, guard hairs provide the black peppering). Sides of the body are slightly paler but show more rusty and buffy highlights. Forearms are the same color as sides of body, except for a grayish brown to dark brown area 8–10 mm behind the wrist that contrasts with the metacarpal surfaces and grayish underparts.

The demarcation between coloration of dorsal and ventral coats is subtle in some animals where the underparts are dark buffy or ochraceous gray, but conspicuous in others where the ventral coat is primarily grayish white. Coloration of underparts is individually variable, but a similar chromatic range can be found in all large samples. At one end of the range is the darkest, dark grayish white (hairs are dark gray for most of their lengths, unpigmented at their tips) or dark grayish buff (hairs have buffy tips), which occurs most frequently in highland samples. The other extreme is pale grayish white or pale grayish-buff underparts (hairs are pale gray with either unpigmented or buffy tips), and predominates in lowland samples. In any large lowland series, the venter coloration is broken by white or buff strips on the chest, along the midventral region in some specimens, at the inguinal area in others. Two specimens from Lombasang are brighter than any of the total specimens I have studied, one has bright buffy gray underparts, the other bright ochraceous-gray ventral fur. On many specimens in samples from low elevations (Pinedapa, Kuala Nausu, Sungai Ranu, Masembo, and Wawo) there are dense orange or rusty patches or strips on the throat and chest, or the inguinal region, or scattered over the abdomen. One adult from Kuala Navusu has a cream venter slightly suffused with pale gray and is rusty around the chin and throat and pale rusty over the neck and chest.

Two samples from about 1000–1500 m deserve additional attention. Except for two specimens, underparts in the sample from Lombasang on the southwestern peninsula show the range in coloration typically seen in series from other places and lower elevations, but just darker: dark grayish white to grayish buff. Two rats have brighter coats, one is buffy gray everywhere, the other ochraceous gray. No specimen in the sample has rusty patches. Specimens in the sample from the Mamasa region (locality 19 in the gazetteer and on the map in fig. 50) have a thicker dorsal coat (15–20 mm) that is darker than lowland samples (the overhairs and their gray segments are longer, imparting a darker tone to the pelage) and darker than the Lombasang series. Underparts are also dark, ranging from dark grayish white to dark grayish buff, and two rats exhibit rusty patches on the neck and chest, similar to the pattern in lowland samples; otherwise, the only difference between this series and those from lower altitudes is the longer, darker pelage.

Ears (external pinnae) are moderately large and appear naked but are covered in short, fine, and unpigmented hairs. In life, the ears have a rubbery texture and are pigmented with gray and brown hues—the range extends from dark gray through brownish gray to dark brownish gray. The dried ears of stuffed museum skins have lost the rubbery texture of the live animal and dried to dark brown with no hint of the actual range in hue and tone.

A tail shorter than the combined length of head and body (LT/LHB 5 75 % –92 %) is typical for B. andrewsi , and the tail is relatively shorter than in samples of the other three species in the B. fratrorum group (table 41). Grayish brown to very dark brown is the range over the dorsal surface of the tail. The ventral surface exhibits the range from all white (from base to tip, and the tail appears conspicuously dorsoventrally bicolored) to monocolored brown or brownish gray. Intermediate variations range from tails that are gray or tan along the ventral surface to surfaces that are mottled pale grayish brown (all of each scale with brown pigment) or speckled with brown (pigment in center of scale). In most samples, the brown on the dorsal surface extends to the tail tip and individuals showing a white tip are uncommon (20 % of 133 specimens), and when present the white tip is short relative to length of tail (mean 5 7.7 %, range 5 1 % – 13 %; see table 8). The gray cast to the dorsal surface of the tail is evident in freshly caught rats, but is lost in the dry study skins where the dorsal surfaces appear brown.

Front and hind feet are long and slender. Front and hind digits are white, dorsal carpal and metacarpal surfaces are white, white lightly speckled with brown (hairs are brown at tips), or white speckled with brownish orange. The naked palmar and plantar surfaces are either unpigmented or show gray or pale brown tones. Claws are unpigmented, those on the front digits are, relative to lengths of the digits, about the same size as in B. fratrorum and larger than the delicate claws of B. penitus ; a sparse ungual tuft springs from the base of each front and hind claw.

Females have four teats, arranged in two inguinal pairs, the number of teats common to all species of Bunomys . The scrotal sac of males is gray and sparsely haired (appears naked), and the testes are small relative to body size (8 % –15 %; see table 9). Description of gross spermatozoal morphology is provid- ed by Breed and Musser (1991).

The dorsal coat of juveniles is shorter (up to 10 mm long) than that of adults, the hairs are finer, and the overall color is darker, without the buffy highlights of the adult coat. Underparts are grayish white. The range in coloration of the feet and tail matches that seen in samples of adults.

Bunomys andrewsi typically has a large skull with a long and narrow rostrum and a wide zygomatic plate (figs. 52–54, 60; tables 42, 51). The interorbit and braincase is moderately wide, the incisive foramina very long (but their posterior rims remain anterior to front surfaces of the first molars), as is the wide bony palate. Each dentary is robust and similar in shape to those of the other species in the B. fratrorum group.

Moderately large molars with simple occlusal patterns are typical of B. andrewsi (fig. 61). The frequencies of cusp t3 are geographically variable. It is present on the second upper molar in only 17 % of 70 specimens from the central core and rarely occurs on the third molar in that sample (4 %). By contrast, cusp t3 is present on the second molar in 78 % of 18 specimens from Lombasang on the southeastern peninsula and exists on the third molar in 22 % of that sample (table 10). As in most other species of Bunomys , an anterior labial cusplet is absent from the first lower molar, but a posterior labial cusplet typically forms part of the chewing surfaces of both the first and second lower molars (fig. 57; table 11). An anterolabial cusp is present on the second lower molar in most specimens (94 % of 71 individuals) but is less prevalent on the third molar (18 % of 66 specimens).

KARYOTYPE: 2N 5 42, FNa 5 56 and FNt 5 58, comprised of seven pairs of metacentric chromosomes, one pair of subtelocentrics, and 12 pairs of acrocentrics; the sex chromosomes are acrocentrics (table 12).

COMPARISONS: Bunomys andrewsi resembles the northeastern peninsular B. fratrorum in color of body fur, but the dissimilarities in the two species, as described in the account of B. fratrorum , are otherwise trenchant and no data suggests that samples of B. andrewsi are simply southern Sulawesi variants of B. fratrorum . In the Mamasa region, B. andrewsi and B. torajae are sympatric but elevationally separated, and morphologies of the two were contrasted in the account of B. torajae .

Outside of B. fratrorum and B. torajae , B. andrewsi requires comparison with three members of the B. chrysocomus group, and with B. penitus and B. karokophilus , n. sp., in the B. fratrorum group. Contrasts between B. andrewsi and B. karokophilus , n. sp., will take place in the account of the new species. Here I compare B. andrewsi first with B. penitus , then with B.chrysocomus , B. coelestis , and B. prolatus .

Bunomys andrewsi and B. penitus: The distribution of B. penitus takes in Pegunungan Mekongga on the southeastern peninsula and the montane forests in the west-central mountain block in Sulawesi’s core (see the map in fig. 51). Samples of B. andrewsi describe a more expansive range: lowlands on the southeastern peninsula and Pulau Buton, lowlands at the western end of the eastern peninsula, low and middle elevations in the mountains of Sulawesi’s core, and middle elevations at the southern end of the southwestern peninsula (see the map in fig. 50). I compare all these samples with those of B. penitus . Of the four taxa associated with B. andrewsi , two were described as subspecies of ‘‘ Rattus penitus . ’’ On the southeastern peninsula, the samples of B. andrewsi from Wawo and Masembo in lowlands immediately adjacent to Pegunungan Mekongga are geographically closest to the Mekongga population of B. penitus . This sample of B. andrewsi was described as ‘‘ Rattus penitus inferior ’’ by Tate and Archbold (1935a), ‘‘ inferior ’’ referring to what they viewed to be a lowland population of the montane B. penitus . ‘‘ Rattus penitus heinrichi ’’ was applied by Tate and Archbold (1935a) to the sample of B. andrewsi from the flanks of Gunung Lompobatang at the southern end of the southwestern peninsula. Bunomys penitus does not occur in montane habitats on that arm of Sulawesi; presumably, Tate and Archbold thought of ‘‘ heinrichi ’’ as a geographic form of the central Sulawesi penitus . Miller and Hollister’s (1921a) ‘‘ Rattus adspersus ’’ is based on a sample of B. andrewsi from the northern lowlands of Sulawesi’s core; B. penitus is common in montane habitats of that region. The two specimens on which ‘‘ Mus andrewsi ’’ ( Allen, 1911) is based were collected on Pulau Buton where B. penitus does not occur.

Bunomys penitus is physically similar to B. andrewsi , averaging slightly longer in lengths of head and body and hind foot, but broadly overlapping in length of ear and mass (table 41). The greatest dimensional contrast between the two species involves tail length, which is absolutely longer in B. penitus (M 5 159.8–171.8 mm) compared with B. andrewsi (mean 5 111.5–156.8 mm) and longer relative to length of head and body (LT/ LHB 5 88 % –102 % for B. penitus , 75 % –92 % for B. andrewsi ). The montane species clashes with the lowland form in its soft, long (up to 25 mm) and silky brownish-gray to grayishbrown dorsal pelage (dorsal coat up to 12 mm thick in most examples of B. andrewsi , soft but not silky, dark brownish gray tinged with bright buff); grayish white ventral coat (underparts in B. andrewsi are dark grayish white, grayish buff, or gray washed with brighter ochraceous hues); its tail that is grayish brown to brownish gray on the dorsal surface behind a white tip, and pure white over the ventral surface (brownish gray on top, ranges from white to brown below in B. andrewsi ); and a white tail segment present in 98 % of the sample that is long relative to tail length (mean 5 21 %, range 5 3 % –68 %; a white tip is present in only 20 % of the combined samples of B. andrewsi and when present is short relative to length of tail, mean 5 7.7 %, range 5 1 % –13 %; table 8).

Size of testes relative to head and body (table 9) and gross spermatozoa morphology ( Breed and Musser, 1991) differ between the two species. Bunomys andrewsi averages larger testes relative to body size (8 % –15 % for four samples) than does B. penitus (9 %; the sample is from the west-central mountain block; I have not seen fluid-preserved material from Pegunungan Mekongga). Both have an asymmetrical sperm head falciform in outline and a tail; the head is slightly longer, the apical hook much shorter, and the tail shorter in B. penitus (specimens from Gunung Kanino) than in B. andrewsi (an example from Kuala Navusu).

Both species have a 2N of 42, but different fundamental numbers (FNa 5 56 and FNt 5 58 for B. andrewsi , FNa 5 58 and FNt 5 60– 61 for B. penitus ; table 12).

The skull averages longer in samples of Bunomys penitus compared with population samples of B. andrewsi , and univariate means for all but six cranial and dental dimensions are also greater in B. penitus (table 42). By contrast, breadth across the zygomatic arches is less in B. penitus than in B. andrewsi , and its zygomatic plate is appreciably narrower. Postpalatal length and length and breadth of the incisive foramina are the same in the two species. See the skull illustrations for visual appreciation of these dimensional similarities and differences (figs. 52–54).

Specimen scores projected on first and second principal components provides a graphic quantitative summary of differences among the cranial and dental variables (fig. 62, upper graph). Covariation in nearly all variables, as expressed by their high and positive correlation coefficients (r 5 0.42– 0.79; table 47), is responsible for the presence of two clusters with very narrow overlap situated along the first axis, scores for the larger skulls of B. penitus to the right, scores for the smaller B. andrewsi to the left. Particularly influential values reflect the appreciably longer and wider rostrum of B. penitus , its longer bony palate, wider mesopterygoid fossa, larger bullae, and heavier molars. Negative loadings for breadths of zygomatic plate and bony palate indicate how much narrower, absolutely and relative to skull size, are these dimensions in B. penitus compared with B. andrewsi .

The distribution of scores in the ordination affirms identities of the holotypes (and by extension the sample of which each is a part) associated with each species. Scores for holotypes of ‘‘ Rattus penitus inferior ’’ and ‘‘ Rattus penitus heinrichi ’’ fall within the constellation representing all population samples of Bunomys andrewsi , as are the points for holotypes of ‘‘ Rattus adspersus ’’ and ‘‘ Mus andrewsi . ’’ Holotypes of ‘‘ Rattus penitus ’’ and ‘‘ Rattus sericatus ’’ are indicated by two scores in the center of the cloud of points defining B. penitus .

There are three qualitative dental distinctions between B. andrewsi and B. penitus . The first is the absence of a large labial cusplet adjacent to cusp t6 on the first upper molar in all examples of B. andrewsi I examined; such a cusp is found in 30 % of the sample of B. penitus (fig. 75). Second is the frequency of occurrence of cusp t3 on the second upper molar (table 10), which is rare in most samples of B. andrewsi (17 %; fig. 61) but relatively common in the sample of B. penitus (62 %; fig. 75). Third is frequency of the anterolabial cusp on the second and third lower molars (table 11). That cusp forms part of the anterolabial margin of the second molar in most examples of B. andrewsi (94 %) but is pesent in only about half the sample of B. penitus (55 %); it is found infrequently on the third molar in the sample of B. andrewsi (18 %) but is absent from all specimens surveyed of B. penitus .

Bunomys andrewsi and B. chrysocomus: As documented by modern samples, both species are regionally sympatric at the southern end of the northern peninsula, the western margin of the eastern peninsula, the southeastern peninsula and Pulau Buton, and Sulawesi’s core (see tables 6 and 20). Each has been taken in the same trapline in three places: the PuroValley at Bakubakulu, 600 m, and the valley of the Sungai Miu along Sungai Oha Kecil at 290 m and Sungai Sadaunta at 675 (see the account of B. chrysocomus ). Both species are also recorded from the southern end of the southwestern peninsula where the comparative material consists of subfossils and modern specimens representing B. andrewsi and subfossils only for B. chrysocomus ; no modern examples of B. chrysocomus have been collected anywhere on the south- western peninsula south of the Tempe Depression (identified in fig. 1). Identity of the subfossils and their comparisons with B. andrewsi and subfossil examples of B. chrysocomus are discussed in a later section (see Subfossils).

Bunomys chrysocomus is physically smaller than B. andrewsi , averaging less in lengths of head and body, hind foot, ear, and mass (tables 19, 41). The tail of B. chrysocomus is typically shorter than that of B. andrewsi (mean range of population samples 5 120.0–

TABLE 47 Results of Principal-Components and Analyses Comparing All Population Samples of Bunomys andrewsi with Those of B. penitus and B. chrysocomus Correlations (loadings) of 16 cranial and two dental log-transformed variables are based on 98 B. andrewsi , 185 B. penitus , and 232 B. chrysocomus ; see figure 62.

142.7 mm for B. chrysocomus , mean range 5 111.5–156.8 mm for B. andrewsi ) but at the same time averages longer in relation to length of head and body (LT/LHB 5 84 % – 98 % for B. chryscocomus , 75 % –92 % for B. andrewsi ).

Bunomys chrysocomus and B. andrewsi are not easily separated by color of fur and appendages, particularly in those places where examples of each were taken in the same trapline. Texture and length of the fur covering upperparts of head and body is similar in the two species. The coat is dark brownish gray in most examples of B. chrysocomus and only lightly speckled (buffy bands of the hairs are dull and short) while the pelage of B. andrewsi is brighter with more intense buffy and yellowish speckling (buffy bands are bright and wide); ears and tops of the front and hind feet are similarly pigmented in both species. Some examples of B. chrysocomus —from Gunung Balease and Gunung Nokilalaki, for example—have very dark, brownish-black dorsal fur with muted speckling and are unlike nearly all examples of the brighter B. andrewsi (exceptions are the specimens from the Mamasa area that were trapped at 1500–1600 m and have much darker upperparts than is usual in most samples of B. andrewsi ). Variation in color pattern of the tail is similar in the two species, a low percentage of specimens in samples of each have a white-tipped tail, and in both species that unpigmented segment is short relative to length of tail (table 8). Bunomys chrysocomus has longer, more gracile claws on the front feet.

While preserved specimens of each species can be difficult to distinguish, the living animals are easier: B. chrysocomus has a longer muzzle with an upturned snout in contrast to the broad face of B. andrewsi without an upturned rhinarium.

Sexually mature males can be separated by size of testes (table 9), which in B. chrysocomus are large relative to body size (mean length of testes/length of head and body 5 22 %), but in B. andrewsi are relatively smaller (8 % –15 %). This relative size disparity is strikingly evident in freshly caught animals where sexually mature examples of each species are taken at the same place.

Configuration of spermatozoa is similar in the two species, but the sperm head is shorter and wider in B. andrewsi ( Breed and Musser, 1991) .

Bunomys chrysocomus has a small skull and weak molars compared with B. andrewsi . Univariate mean values for all cranial and dental variables in combined population samples of B. andrewsi exceed univariate means for those dimensions in all the combined samples of B. chrysocomus (table 42). Similar metric distinctions exist in samples of each species obtained from most places where the two are regionally sympatric (table 48). The dimensional contrasts can be appreciated in the images of skulls of B. chrysocomus arranged in figures 16 and 99– 101 with those of B. andrewsi portrayed in figures 52–54 and 99–101. Samples from the northern part of the west-central mountain

TABLE 48

Descriptive Statistics for Cranial and Dental Measurements (mm) Contrasting Particular Samples of Bunomys

chrysocomus with Samples of B. andrewsi

Mean ±1 SD, and observed range (in parentheses) are listed.

Pulau Buton Eastern peninsula

Gunung Balease

B. chrysocomus B. andrewsi B. chrysocomus

Variable B. andrewsi RMNH 21256 Sungai Ranu Gunung Tambusisi B. andrewsi B. chrysocomus

N 2 1 5 6 8 3

ONL 40.4 ± 2.62 36.9 44.0 ± 1.27 38.3 ± 1.34 42.8 ± 1.33 40.2 ± 0.23

(38.5–42.2) (41.9–45.0) (35.8–39.5) (40.3–44.8) (40.1–40.5) ZB 19.5 ± 0.64 — 20.9 ± 0.16 17.8 ± 0.61 21.1 ± 0.74 18.8 ± 0.17

(19.0–19.9) (20.7–21.1) (17.1–18.4) (19.7–21.9) (18.7–19.0) IB 6.6 ± 0.07 6.0 6.5 ± 0.13 6.5 ± 0.25 7.0 ± 0.19 6.9 ± 0.27

(6.5–6.6) (6.3–6.6) (6.1–6.8) (6.7–7.3) (6.7–7.2)

LR 13.8 ± 1.06 — 15.7 ± 1.05 13.3 ± 0.70 15.4 ± 0.60 14.2 ± 0.45

(13.0–14.5) (14.3–17.1) (12.4–14.1) (14.6–16.1) (13.7–14.6) BR 7.6 ± 0.64 6.1 8.0 ± 0.46 6.5 ± 0.48 7.9 ± 0.46 7.0 ± 0.20

(7.1–8.0) (7.4–8.7) (5.6–6.9) (7.5–8.9) (6.8–7.2) BBC 15.8 ± 0.21 15.0 16.5 ± 0.34 15.6 ± 0.42 16.6 ± 0.39 16.2 ± 0.21

(15.6–15.9) (16.1–16.8) (15.1–16.2) (15.8–17.1) (16.0–16.4) HBC 11.0 ± 0.28 11.0 11.8 ± 0.29 10.14 ± 0.26 11.7 ± 0.48 11.5 ± 0.55

(10.8–11.2) (11.5–12.2) (10.3–11.0) (11.1–12.2) (11.0–12.1) BZP 3.5 ± 0.42 2.7 4.1 ± 0.29 2.9 ± 0.08 4.2 ± 0.34 3.0 ± 0.30

(3.2–3.8) (3.9–4.6) (2.8–3.0) (3.4–4.5) (2.7–3.3) LD 10.6 ± 1.06 9.5 12.0 ± 0.46 10.5 ± 0.55 11.4 ± 0.67 10.8 ± 0.20

(9.8–11.3) (11.5–12.5) (9.6–11.1) (10.1–12.2) (10.6–11.0) PPL 14.3 ± 0.99 13.5 15.2 ± 0.60 13.2 ± 0.49 15.0 ± 0.79 13.9 ± 0.27

(13.6–15.0) (14.4–16.0) (12.3–13.6) (13.7–16.0) (13.7–14.2) LBP 8.1 ± 0.71 7.1 8.9 ± 0.78 7.6 ± 0.48 8.1 ± 0.34 8.2 ± 0.06

(7.6–8.6) (7.8–9.9) (6.7–8.1) (7.7–8.7) (8.1–8.2) BBP 3.8 ± 0.35 3.1 3.9 ± 0.35 3.5 ± 0.22 4.1 ± 0.22 3.9 ± 0.21

(3.5–4.0) (3.6–4.1) (3.2–3.8) (3.8–4.5) (3.7–4.1) BMF 2.9 ± 0.21 2.6 3.5 ± 0.28 2.7 ± 0.21 3.1 ± 0.18 3.0 ± 0.21

(2.7–3.0) (3.2–3.8) (2.4–2.9) (2.9–3.4) (2.8–3.2) LIF 7.3 ± 0.00 6.3 8.5 ± 0.66 6.7 ± 0.26 8.2 ± 0.41 6.0 ± 0.35

(7.8–9.5) (6.5–7.2) (7.6–8.6) (5.7–6.4) BIF 2.8 ± 0.28 2.4 3.1 ± 0.15 2.3 ± 0.12 3.0 ± 0.11 2.4 ± 0.00

(2.6–3.0) (2.9–3.3) (2.2–2.5) (2.8–3.1)

LB 7.3 ± 0.21 6.9 6.6 ± 0.23 6.5 ± 0.19 6.6 ± 0.24 6.6 ± 0.27

(7.1–7.4) (6.3–6.9) (6.2–6.7) (6.3–6.9) (6.4–6.9) CLM1–3 7.2 ± 0.28 5.9 7.4 ± 0.22 6.3 ± 0.21 7.6 ± 0.24 6.2 ± 0.10

(7.0–7.4) (7.1–7.6) (5.9–6.5) (7.3–8.0) (6.1–6.3) BM1 2.2 ± 0.21 2.1 2.3 ± 0.06 2.0 ± 0.12 2.4 ± 0.08 2.1 ± 0.00

(2.0–2.3) (2.2–2.3) (1.8–2.1) (2.3–2.5)

block are an exception because the B. andrewsi from there are smaller in body size and the range of variation in some of the cranial variables overlaps in samples of the two species; this contrast will be amplified in a later section.

Morphometric distinctions between B. chrysocomus and samples of B. andrewsi are summarized by a scatter plot containing specimen scores projected onto first and second principal components where they form two oblique elliptical clouds, one representing B. chrysocomus , the other B. andrewsi (fig. 62, lower graph). Major axes of the elliptical spreads of scores are parallel, phenetically distinct: their Y-intercepts are significantly different between the two species (+0.873 versus ‾ 1.984; F 5 117.57, P 5 0.000), but their slopes are comparable (1.140 versus 1.075; F 5 0.20, P 5 0.652). Moderate to high positive loadings for nearly all variables (r 5 0.42–0.91; table 47) on the first component reflect size, scores representing B. andrewsi with the larger skulls and molars to the right, those for B. chrysocomus with the smaller skulls and molars to the left with overlap in the center of the scatter plot. Loadings for skull size (occipitonasal length and zygomatic breadth), length and breadth of the rostrum, size of braincase, breadth of zygomatic plate, lengths of diastema and the postpalatal region, expanse of the incisive foramina, and size of molars are especially influential in dispersing scores along the first axis. All these dimensions are less in B. chrysocomus (table 42), especially overall size of the skull, width of the zygomatic plate, size of the rostrum along with extent of the incisive foramina, and molar robustness, which form salient distinguishing landmarks readily appreciated in the images of skulls portrayed in the figures as well as in side-by-side comparisons of actual skulls.

Frequencies of a cusp and cusplet on lower molars differ between the two species (table 10). An anterior labial cusplet occurs on the first molar in half of the B. chrysocomus examined but is absent from all the specimens of B. andrewsi surveyed. The third molar bears an anterolabial cusp in 65 % of the sample of B. chrysocomus but is present at a much lower frequency in the specimens of B. andrewsi (18 % of the sample).

Bunomys andrewsi and B. coelestis : Bunomys coelestis inhabits montane forests on Gunung Lompobatang where samples were collected between 1800 and 2500 m; B. andrewsi is found in tropical lowland evergreen rainforest at lower elevations on the volcano with the only large sample coming from Lombasang at 1100 m; the species is also documented by subfossils excavated from deposits in caves on the western coastal plain. Bunomys andrewsi and B. coelestis are contrasted here because individuals in the peninsular population of B. andrewsi on average are smaller in body size than those in other geographic samples of the species, and many cranial dimensions, compared to samples of B. andrewsi from elsewhere on the island are more similar to B. coelestis (tables 49, 51).

Bunomys coelestis is physically larger than animals in the peninsular sample of B. andrewsi , a difference reflected in its greater univariate means for lengths of head and body, tail, hind foot, and ear (compare the values for B. coelestis listed in table 19 with those for the sample of B. andrewsi from Lombasang in table 41). Bunomys coelestis has darker upperparts, a longer muzzle, and appreciably longer claws at the ends of the front digits.

Compared with B. andrewsi from Lombasang, B. coelestis typically has a longer but narrower skull (indexed by occipitonasal length and zygomatic breadth); wider interorbit; longer but narrower rostrum (elongate compared with the relatively stocky rostrum of B. andrewsi ); wider zygomatic plate; longer diastema, longer and wider bony palate, and longer basicranial region (measured by postpalatal length); less spacious incisive foramina (shorter and narrower than the expanded foramina in B. andrewsi ); smaller auditory bulla; and smaller molars (table 49). Size of the braincase (breadth and height) is comparable in the two species.

These contrasts in cranial and dental dimensions indicated by univariate mean values are summarized in a scatter plot where individual specimen scores are projected onto first and second principal components (fig. 63, upper graph). Two widely separated groups of scores represent B. coelestis in the right half of the ordination and B. andrewsi from Lombasang in the left, positions influenced by the longer skull of B. coelestis along with its longer rostrum and diastema, and longer and wider bony palate (r 5 0.42–0.76; table 50). Moderate to high negative correlations (r 5 ‾ 0.40 to ‾ 0.88) point to the wider skull of B. andrewsi (indexed by zygomatic breadth) along with its broader rostrum and mesopterygoid fossa, longer and broader incisive foramina, larger bullae, and heavier molars as compared with B. coelestis , contrasts between the two species that are mirrored by univariate means (table 49).

Bunomys andrewsi and B. prolatus: Samples of three species of Bunomys come from Gunung Tambusisi and adjacent lowlands in the western region of the eastern peninsula of Sulawesi. Bunomys chrysocomus and B.

TABLE 49

Descriptive Statistics for Cranial and Dental Measurements (mm) Contrasting Particular Samples of Bunomys andrewsi with Samples of B. coelestis and B. prolatus

Mean ± 1 SD and observed range (in parentheses) are listed.

prolatus were collected by C.H.S. Watts from slopes of the mountain (1372–1829 m), and he obtained B. andrewsi from adjacent lowland forest in the Sungai Ranu area (50 m). The large-bodied Bunomys prolatus could possibly be confused with B. andrewsi , which is also characterized by large body size in the Tambusisi region.

While B. prolatus and B. andrewsi are comparable in body size (estimated by length of head and body), B. prolatus has a much shorter tail (absolutely and relative to length of head and body), shorter hind feet, but relatively larger ears (compare the measurements of B. prolatus listed in table 18 with those of B. andrewsi from Kuala Navusu, Pinedapa, and Wawo + Masembo listed in table 41; I lack external measurements for the sample from Sungai Ranu). As is typical of the contrast between montane and lowland populations, the lush dorsal coat of B. prolatus is soft and dense, 20–25 mm long; that of most B. andrewsi , and especially the rats from Sungai Ranu, are harsher and 12– 15 mm long. The robust and very long front claws of B. prolatus (fig. 36) are unlike the weaker and short front claws, a conformation typical of specimens in all population samples of B. andrewsi .

The skull of B. prolatus has a gracile aspect to it compared with the stocky conformation that is usual for B. andrewsi (compare the images of B. prolatus in figs. 37–39 with those for B. andrewsi from Kuala Navusu portrayed in figs. 52–54). The more delicate cranial conformation in B. prolatus , so visually apparent in the cranial illustrations, is reinforced qualitatively by univariate means for cranial and dental variables where most dimensions average less than those for the sample of B. andrewsi from Sungai Ranu; exceptions are interorbital breadth and length of bullar capsule, which average greater in B. prolatus (table 49).

Contrasts among the cranial and dental variables are summarized in multivariate space by the scatter plot of specimen scores projected onto first and second principal components (fig. 63, lower graph). Aggregations of scores identifying the specimens of B. prolatus and those for B. andrewsi from Sungai Ranu are widely separated along the first axis, reflecting the general disparity in size as indicated by the large and positive loadings (r 5 0.45–0.99; table 50) for all but two variables, which mirrors the contrasts in univariate means (table 49). The two high negative loadings are for interorbital breadth (‾ 0.86) and length of bulla (‾ 0.51), both absolutely and relatively greater in B. prolatus (table 49).

The long fur of B. prolatus , its robust and long front claws, elongate skull, wide interorbit, narrow zygomatic plate and mesopterygoid fossa, short and narrow incisive foramina, large bulla, and link to montane forest are strongly diagnostic compared with the larger-bodied B. andrewsi from the nearby lowlands.

GEOGRAPHIC VARIATION: There is conspicuous variation in body size among geographic samples of B. andrewsi . During 1974, I was trapping between 290 and 675 meters along the Sungai Oha Kecil, Sungai Miu, and Sungai Sadaunta in the northern part of the west-central region of Sulawesi and encountering mostly B. chrysocomus . Here and there another kind of rat showed up in the traps that in body size and fur color resembled B. chrysocomus but with a shorter muzzle. It was also similar to B. andrewsi but physically smaller than rats in samples of that species with which I was familiar. Later I worked in lowlands east of the west-central mountain block but not too distant from Pinedapa, the place where H.C. Raven collected a series of rats in 1918 that would subsequently be described by Miller and Hollister (1921a) as Rattus adspersus . Near my camp on Kuala Navusu, I caught what clearly looked to be adspersus . The rats resembled animals taken at the Sungai Oha Kecil, Sungai Miu, and Sungai Sadaunta (means for LHB 5 163.1 mm and weight 5 113.7 g for the Puro-Sungai Miu sample) but were larger in body size (LHB 5 177.4 mm and weight 5 154.6 for the sample from Kuala Navusu).

The difference registered by my two collections generally reflects the magnitude of variation in adult body size among all the samples of B. andrewsi I have studied (see table 41). Generally, large physical size characterizes specimens in the population samples from northeast–central core (Labuan Sore, Kuala Navusu, and Pinedapa), south–central core (Gunung Balease, Malili area, and Sukamaju), and the eastern + southeast peninsulas (Sungai Ranu, Wawo, and Masembo). Animals averaging smaller in body size form the population samples from lowlands and highlands in the west-central mountain block—northern west-central highlands (Puro-Sungai Miu, Tamalanti, and Tuare) and southern west-central highlands (Mamasa Area)—and the southwest peninsula (Lombasang).

To determine the geographic trends in body size, I relied on my measurements of cranial and dental dimensions (measurements for head and body and appendages were made by different collectors who may or may not have utilized the same endpoints). Morphometric relationships among the population samples are summarized in the univariate descriptive statistics listed in table 51, results of principal-components and discriminant-function analyses (figs. 64, 65), cluster diagram (fig. 66), and descriptive exposition in the text.

Intersample variation in cranial and dental measurements is apparent in the ordination of specimen scores for all seven population samples of B. andrewsi projected on first and second principal components (fig. 64); embedded in the scatter plot are results concordant with geographic differences in body size

TABLE 50 Results of Principal-Components Analyses Comparing Two Samples of Bunomys andrewsi with Samples of B. coelestis and B. prolatus Correlations (loadings) of 16 cranial and two dental log-transformed variables are based on 12 B. andrewsi , 19 B. coelestis , and eight B. prolatus . Upper graph: B. andrewsi from lower slopes of Gunung Lompobatang is contrasted with B. coelestis from higher on the volcano. Lower graph: The lowland B. andrewsi from Sungai Ranu is contrasted with the montane B. prolatus from Gunung Tambusisi. See figure 63.

I encountered in my trapping. Moderate to high positive loadings (r 5 0.46–0.95; table 52) among nearly all variables influence the dispersal of scores along the first axis; variation in size is stressed here with scores for the smallest skulls sprinkled to the left in the scatter plot, those for larger skulls to the right. Reflected in the spread of points along the first axis is individual variation and that attached to adults of different ages (young to old) but there is also a geographic pattern. Scores for specimens with the largest skulls are from eastern + southeast peninsulas (empty triangle; Sungai Ranu, Masembo, and Wawo) and cluster along the far right of the entire scatter. Northern west-central region (inverted filled triangle; Puro-Sungai Miu, Tamalanti, and Tuare) and southwest peninsula (empty diamond; Lombasang) are represented by scores filling the left half of the ordination, and these samples contain the smallest skulls (table 51). Straddling these outlying clumps are scores for specimens in northeast–central region (filled circle; Labuan Sore, Kuala Navusu, and Pinedapa), south–central region (empty circle; Gunung Balease, Malili area, and Sukamaju), and southern west-central region (asterisk; Mamasa Area). Of the two scores representing Pulau Buton, the very young adult aligns with the smaller specimens, the other—the old adult holotype —nests in the center of the cloud of scores for northeast–central region. Although there is a geographic trend in size of skulls, clusters representing the population samples greatly overlap and not one of them is isolated within the ordination.

Shape information can be extracted from the spread of scores along the second component (table 52). Most specimens from northeast-central region (filled circle in fig. 64) have a relatively narrower zygomatic plate, wider mesopterygoid fossa, larger bullae, and shorter molar row compared with animals represented by the points in the top half of the scatter plot, especially those from the west-central mountain block (northern west-central region and southern west-central region population samples).

A sharper picture of geographic trends in intersample variation of cranial and dental variables emerged from results of discriminant-function analysis. Among the first, second, and fourth canonical variates extracted, the spread of scores along the first variate (fig. 65, upper graph) is a response to the positive moderate to high correlations (or loadings; r 5 0.32–0.68; table 53) for the variabes. Size is reflected along this axis: the right half of the ordination contains samples from the west-central mountain block (northern west-central region and southern westcentral region) and southwest peninsula, which are generally characterized by small skulls; the left half holds scores for largerbodied animals from eastern + southeast peninsulas and northeast–central region (east of the west-central mountain block); scores

TABLE 51

Descriptive Statistics for Cranial and Dental Measurements (mm) Derived from Population Samples of

Bunomys andrewsi

Mean ± 1 SD and observed range (in parentheses) are listed.

for samples from south–central region overlap the groups of scores in the left and right halves of the scatter plot. Points representing the two specimens from Pulau Buton align with those representing animals with larger skulls. The high positive loadings for overall size of skull (occipitonasal length and zygomatic breadth), breadth of braincase and zygomatic plate, length of diastema, postpalatal length, breadth of bony palate, expanse of incisive foramina, and length of molar row indicate that these dimensions are not as great in samples from the west-central mountain block and Lombasang on the southwestern peninsula compared with series from elsewhere in Sulawesi, mainly those landscapes east of the west-central mountain block.

Shape differences are indicated by the spread of scores along the second variate in the upper graph in figure 65. High positive loadings for nearly all variables (r 5 0.27– 0.69; table 53), along with isolation of the scores for eastern + southeast peninsulas (empty triangle) on the right margin of the ordination, and southwest peninsula (empty diamond) at the left extreme, point to the relatively overall larger cranial and dental dimensions in the former and relatively smaller dimensions in the latter compared to the other four samples clumped together in the center of the ordination, which includes the holotype of andrewsi (star).

The lower scatter plot in figure 65 reveals a pattern of covariation in cranial and dental measurements among the population samples similar to that in the upper scatter plot of figure 65. The large and positive loadings for breadth of rostrum (0.45) and length of bulla (0.69) on the fourth variate (see table 53) isolate scores for the two specimens from Pulau Buton from those points signifying specimens in all the other population samples. Compared with the six population samples, whether they contain small, large, or skulls of intermediate size, animals in the sample from Pulau Buton have a relatively wider rostrum and much larger tympanic bulla, particularly the holotype of andrewsi . A large bulla is characteristic of the small sample from Pulau Buton: 7.2 mm is the mean length of bulla for Pulau Buton, and 6.3–6.8 mm is the range of means for the other six population samples.

Cluster analysis based on squared Mahalanobis distances results in the pattern of phenetic relationships among the seven population samples portrayed in figure 66. Two primary groups of samples are apparent. The largest unites all samples from mainland Sulawesi except those at the most southern margins of the island, Pulau Buton off the south coast of the southeastern peninsula and Lombasang at the southern end of the southwestern peninsula, which form a group discrete from that containing the other five population samples.

Within the larger assemblage, the phenetic relationships among samples generally mirror geography. Specimens from lowlands at Labuan Sore, Kuala Navusu, and Pinedapa (northeast-central region) link with animals from Gunung Balease, the Malili area, and Sukamaju (south-central region), all localities east or at the margin of the west-central mountain block. Samples from the northern and southern parts of the west-central mountain block are united in the diagram and are linked (between 30 and 40 distance units) to the previous two samples from east of that mountainous region. These two sets join the sample of animals from the west end of the eastern peninsula at Sungai Ranu and Wawo and Masembo in the lowlands adjacent to Pengunungan Mekongga on the southeastern peninsula. All these samples describe a distribution covering the core of Sulawesi and the southeastern peninsula. The trend in skull size (and generally body size) extends from the west-central mountain block where populations are characterized by small skulls on average to east of the mountain block where larger-bodied animals predominate, with the largest skulls in the samples from Sungai Ranu and the southeastern peninsula. The names ‘‘ Rattus adspersus ’’ ( Miller and Hollister, 1921a), based on the series collected at Pinedapa, and ‘‘ Rattus penitus inferior ’’ ( Tate and Archbold, 1935a), applied to specimens from Masembo and Wawo, are associated with this large geographic block of samples.

Other traits I examined on specimens in the five population samples do not show patterns of intersample variation that accords with the geographic trend in size associated with cranial and dental variables. For example, I did not detect any geographic variation in qualitative traits of the skull or teeth. And with few exceptions, all the adults have a moderately soft and long dorsal coat that is dark brown or brownish gray speckled with buff and black; in any large sample, the ventral coat ranges from pale grayish white or pale grayish buff to dark grayish white, dark grayish buff, buffy gray, or ochraceous-gray (some specimens show rusty patches); ears are brown or brownish gray; dorsal surfaces of feet are white, gray, or buffy; the tail is shorter than combined length of head and body (LT/ LHB 5 85 % –92 %), brownish over the dorsal surface, white, mottled, or brown on the ventral surface, and either lacks or shows a relatively short white tip (see table 8).

There is variation in thickness and color of fur covering head and body that corresponds to differences in elevation, which is illustrated by samples from two regions. Specimens in the Mamasa area (the southern part of the west-central mountain block) come from higher elevations (900–1600 m) than do most examples of B. andrewsi and have a thicker dorsal coat (15–20 mm) that is much darker than lowland samples. Underparts also average darker, ranging from dark grayish white to dark grayish buff.

The population sample of the south-central region contains specimens collected between 830 and 925 m on Gunung Balease and those obtained at 450 m in the Malili area, approximately 80 km southeast of Gunung Balease, that cannot be distinguished by cytochrome- b sequences (molecular trees passed to me by J.L. Patton and K.C. Rowe, in. litt., 2011), body size, cranial and dental traits, or pelage thickness. They do show, however, slight differences in coat color that is related to elevation. Specimens from Gunung Balease have darker upperparts (dark brown to brownish black speckled with buff) than do the rats in the sample from the Malili region, which in pelage color is closely similar to the specimens from Pinedapa and Kuala Navusu and others from 400 m and lower (rich brown speckled with buff over the upperparts). Underparts of both the Balease and Malili specimens range from grayish white to grayish buff (the buff forms either a pale or intense wash over the underparts), but the mountain animals are slightly darker.

The two examples from Pulau Buton, which comprise the type series of ‘‘ Mus

TABLE 52 Results of Principal-Components Analysis Performed on Population Samples of Bunomys andrewsi Correlations (loadings) of 16 cranial and two dental log-transformed variables are based on 98 B. andrewsi ; see figure 64.

andrewsi ’’ ( Allen, 1911) are distinguished from all other samples primarily by their relatively large ectotympanic bullae, slightly shorter but wider rostrum, short incisive foramina, and slightly average smaller skull. They also have very short tails in relation to head and body (LT/LHB 5 75 %; 85 % –92 % is the range for the other geographic samples). Color of the fur and appendages closely resembles samples from farther north in the lowlands, Kuala Navusu, Pinedapa, and the Malili area, for example. It is their average smaller skull that may be most responsible for linking the two specimens with the sample from Lombasang at the southern end of the southwestern peninsula, the type series of ‘‘ Rattus penitus henrichi ’’ ( Tate and Archbold, 1935a); otherwise the animals in each sample are quite different. The Lombasang rats were collected at 1000 m and have a thicker coat (15–20 mm thick as opposed to 12–15 mm for the Buton specimens from the lowlands) that is a richer and brighter brown, a relatively longer rostrum, longer incisive foramina (absolutely and relative to skull length), and smaller bullae (relative size is similar to specimens in samples from the center of the island and southeastern peninsula. Also, a small cusp t3 occurs on the second upper molar in 78 % of the Lombasang sample; the young adult from Pulau Buton lacks a comparable cusp (molars are too worn to determine presence or absence of cusp t3 on the other Buton specimen).

To summarize, I provisionally regard the variation in coat color, body size, and the cranial and dental measurements described above to be contained within the phenetic boundary of a single species. Here are highlights of this hypothesis and suggestions for future inquiry:

TABLE 53 Results of Discriminant-Function Analysis Performed on Population Samples of Bunomys andrewsi Correlations (loadings) of 16 cranial and two dental log-transformed variables are based on 98 B. andrewsi ; see figure 65.

I can report that the sample from Lombasang broadly overlaps those from the westcentral mountain block (northern west-central and southern west-central regions) in magnitude of many cranial and dental variables, as reflected in the scatter plots of specimen scores derived from principal-components and canonical-variate analyses (figs. 64, 65). Simply based on morphometric similarities, the southwest peninsular and west-central mountain block samples appear to be phenetically similar.

Results of cluster analysis based on squared Mahalanobis distances, however, do not link the sample from Lombasang with those from the west-central mountain block (fig. 66). In addition to the diagram in figure 66, I generated several other cluster diagrams (none are illustrated here) using different combinations of all samples, reducing the larger population samples into smaller entities for some analyses, and eliminating the two specimens from Pulau Buton for others. In each result, the configuration of population samples formed by the linked northeast-central region and southcentral region connected to the linked northern west-central region and southern west-central region (west-central mountain block) retained its integrity (the island’s core unit). In different iterations, the sample from Lombasang either linked with Pulau Buton or in the absence of that sample connected to the core assemblage, or to a cluster consisting of the core samples and the eastern + southeast seninsulas sample, but never directly with either of the samples from the west-central mountain block.

Morphological and genetic information derived from new material collected in the southern part of Sulawesi’s core just north of the Tempe Depression as well as along the peninsula south of the depression would be most welcome and likely would resolve the present ambiguous relationship between the sample from Lombasang and those from the core of the island. Of all the samples I examined for this review, the one from Lombasang might represent a separate species, a supposition I entertained earlier (Musser, 1991; Musser and Carleton, 1993). (5) The images of skulls portrayed in figure 67 visually summarize the variation in size of skulls among population samples. On the left is the holotype of andrewsi from Pulau Buton—its wider rostrum (relative to its length), shorter incisive foramina, and larger bullae are evident. On the right is the holotype of heinrichi from Lombasang on the southwestern peninsula. The middle skull is from Kuala Navusu and represents the typical size and conformation of many skulls in samples from Kuala Navusu, Pinedapa, and other localities on the margin of the west-central mountain block and places in the lowlands east of there and on the southeastern peninsula, and is closely similar in size and shape to the holotypes of adspersus and inferior.

NATURAL HISTORY: Included here is a summary of information covering habitat and diet, along with my few observations of nest construction.

Habitat: Bunomys andrewsi has been collected in primary forest, secondary growth, and village gardens. Along the Sungai Oha Kecil and Sungai Sadaunta in the northern part of the west-central region, and in the Kuala Navusu area, my helpers and I trapped B. andrewsi in primary forest along streams and on hillsides—all sites were shaded where the ground remained either damp or wet (descriptions of trap sites are summarized in table 54). Mean ambient air temperature ranged from 66.9 ° to 78.8 ° F along Sungai Oha Kecil and Sungai Sadaunta, and 73.6 ° to 80.9 ° F at Kuala Navusu, with relative humidities reaching 100 % in all three places (table 3). Individuals were caught in traps placed on decaying tree trunks and limbs spanning streams and ravines; in runways beneath trunks and limbs that are remnants of treefalls lying on the forest floor, on wet stream terraces covered by shrubs, small trees and canopied by taller streamside forest; beneath moss-covered rocks on wet hillsides, and beneath roots of living trees. Decaying tree-falls were especially productive places to set traps because they provided excellent cover for runs beneath the trunks and limbs and are usually shaded by a dense cover of shrubs, ferns, rattan rosettes, and saplings that offered a constant damp or wet microenvironment. Examples of the typical places where we caught B. andrewsi are presented in figures 68 and 69. We did not encounter the rats in forest high on steep hillsides or ridgetops, areas that dry out faster than protected hillsides above streams and stream banks and terraces, which remain wet.

TABLE 54

Summary of Microhabitats at Trapping Sites, Stomach Contents, and other Relevant Information for Specimens of Bunomys andrewsi Collected in Central Sulawesi, 1974–1975

Descriptions of the trapping sites and contents of stomachs are summarized from my field journals (in

Mammalogy Archives at AMNH). All collection localities are in old-growth tropical lowland evergreen rain forest; all rats were caught during the night. See habitats in figures 68 and 69.

Locality, AMNH and Elevation

(ASE-field) numbers (m) Date Trap site and other information

Sungai Oha Kecil

224630 (1649) 290 July 1974 In wide runway beneath roots of trees growing along margin of ravine between the small Sungai Oha Kecil and the larger Sungai Miu. Stomach: nearly empty, remains of rhinotermitid termites mixed with small chunks of fruit.

224631 (1838) 457 Aug. 1974 On rotting, smooth trunk bridging Sungai Oha Kecil in streamside

224632 (1845) forest. Stomach: 224631 had a few pieces of fig (seeds still attached to

224633 (1858) inner wall of the rind); 224632 was full of large brown seeds (same as found in 224107 from Sungai Sadaunta), some complete (dark brown seed coat, seed roughly triangular to oblong), others partly digested, intact seeds and pieces of seed coat held in a whitish gelatinous substance; 224633 full of same fruit as found in 224632.

Sungai Miu

224145 (1541) 350 Apr. 1974 On smooth, wet, and rotting trunk lying over narrow tributary stream; wet streamside, shrubby understory at either end of trunk, surrounded by tall river terrace forest. The rats Taeromys celebensis , Rattus hoffmanni , and Maxomys muschenbroekii were caught at same spot on different nights; trunk clearly used by rodents to cross stream. Stomach: some bait; packed with rhinotermitid workers and soldiers (mostly head capsules, body tissue, legs); fruit pulp and a few seeds; a piece that resembles snail (tissue, not shell).

Sungai Sadaunta

224107 (1344) 675 Feb. 1974 In runway beneath rotting, wet trunk lying on wet ground in understory of streamside forest. Stomach: partly full of fig remains (tiny seeds, some still adhering to inside of skin, pieces of fig wall); also a few large, hard brown seeds from a different fruit; small earthworm cut into segments.

224116 (1363) 675 Feb. 1974 On damp ground beneath boulder in forest understory adjacent to Sungai Sadaunta. Stomach: distended with bait.

Kuala Navusu

225666 (3199) 55 Sept. 1975 Live trap on ground beneath decaying trunk (2 ft diameter) resting on

225667 (3247) slope above a stream. Trunk is 60–70 ft long and extends down the

225657 (3247) hill, over the stream and onto the opposite bank; this old tree-fall is lying in scrubby understory, and a large, long opening persists in the canopy of the hillside forest. A Maxomys hellwaldii was trapped in same place. Two Paruromys dominator and one Echiothrix centrosa were trapped on same trunk but on top of that portion bridging the stream. Ate small gecko during captivity (see text).

225665 (3238) 46 Sept. 1975 On ground beneath rotting trunk (2 ft diameter) lying in dense shrubbery on terrace of stream; part of a canopy tree that fell and broke into long sections that are now decaying. Good clear run beneath trunk and dense shrubby cover on both sides. Caught Maxomys hellwaldii in same spot (see fig. 65) and trapped Bunomys andrewsi 225648 nearby. Stomach: 225666 was empty except for a bit of bait; 225667 empty except for unidentifiable mash and a few fragments of insect sclerites; 225657 nearly empty, unidentifiable mash, few fragments of insect sclerites.

225659 (3360) 122 Oct.1975 On rocky, wet slope above a giant Dracontomelon dao ; here the rocks are covered with a thin layer of moss, soil and leaf litter lay among the rocks; area is shrubby; here and there vines and low understory trees grow over the rocks; clumps of tall rattan are scattered over the slope as are palms ( Licuala celebica , Pinanga caesia , and Livistona rotundifolia ); scattered understory trees form a partial canopy that shades the area; TABLE 54 (Continued)

Locality, AMNH and Elevation

(ASE-field) numbers (m) Date Trap site and other information

tall emergents, in addition to D. dao , grow upslope and downslope. Trap was placed on damp ground beneath several rocks. Stomach: full of fruit (not identified) and remains of small insects, mostly larvae.

225660 (3361) 122 Oct. 1975 On leaf-covered, wet ground alongside rocks on slope where 225659 was caught and about 5 ft from that rat; rocks partly covered by shrubs. A Maxomys hellwaldii was caught in same spot. Ate invertebrates and fruit during captivity (see text).

225661 (3371) 122 Oct. 1975 On wet ground beneath a shrub-covered group of rocks on same rocky slope where 225659 and 225660 were caught, only 3 ft from 225659. Stomach: crammed with mostly fruit mash from figs (large seeds, chunks of rind) and numerous rhinotermitid workers and soldiers.

225662 (3399) 122 Oct. 1975 In runway beneath rotting trunk lying on hillside above a rocky ravine; runway 6 in. wide extending length of limb (15 ft); parts of trunk covered with shrubs, leaf litter, and debris, so runway is partly concealed. Understory dense, young fan palms( Livistona rotundifolia ), Licuala celebica ., species of rattan, and a variety of shrubs; closed understory canopy with scattered tall trees nearby; ravine is shaded, cool, and moist, staying that way much longer than the exposed higher slopes away from the ravine. Caught a Maxomys hellwaldii in same spot. Stomach: distended with bait and unidentifiable fruit mash.

225663 (3416) 122 Oct. 1975 On ground in rocky area where 225659–225661 were caught; set trap beneath shrub and moss-covered rocks in front of burrow entrance that leads down into spaces among rocks. Stomach: full, mash of fig remains (seeds and skin of fruit) in which are mixed small beetle larvae.

225664 (3422) 122 Oct. 1975 On damp ground under moss-covered rocks beneath shrubs and litter; same rocky slope where 225659–225661, and 225663 were caught, a rocky area about 10 ft square in hillside forest. Stomach: full of fruit mixed with small insects (ant, beetle larvae)

225658 (3537) 61 Oct. 1975 On very rotten section of trunk lying across wet and deep, V-shaped, and rocky ravine that is a tributary of main stream; hillsides steep; forest uneven; canopy trees like ula ( Diospyros macrophylla ) and tea ( Artocarpus sp. ) predominate—the latter is common; same general area where 225649 was caught. Stomach: full of finely masticated figs with remains of adult and larval insects.

225649 (3549) 54 Nov. 1975 On top of decaying tree trunk (12 in. diameter) lying across deep rocky

225651 (3562) ravine (12 ft wide across top even with trunk, 12 ft deep from trunk to ravine bed) that is a tributary to the main stream and drains this section of the eastern-facing hillside. Generally good hillside forest, but it is open in this spot due to an old treefall that must have cleared a large area (about fourth of an acre) forming a large opening in the forest canopy; shrub and shrubby growth on slope above the ravine, and large sections of rotting trunks and limbs. The forested slopes are steep; on left side (facing up the ravine) slopes are densely covered with a field of tall (5 ft) ferns and nearby is an area of scrub consisting of palms, shrubs, saplings, euphorbs, elephant ears, and streamside figs. There are few rocks exposed on the slope along sides of the ravine. There is good cover under the rotten wood that lay in part of the scrub; a side ravine, rocky, shallow, and dry but mossy, courses down the slope and joins main ravine about where the rat was trapped; forest on right side of ravine is undisturbed; a high broken canopy surrounds the area; ula ( Diospyros macrophylla ), laru ( Myristica fasciculate ), and tea ( Artocarpus sp. ) are common; Madhuca sp. here also—the usual hillside-terrace flora; a huge (125 ft high, 4 ft diameter) pohon tea ( Artocarpus sericarpus )

towers in solitude over the scrubby area, apparently not felled by the TABLE 54 (Continued)

Locality, AMNH and Elevation

(ASE-field) numbers (m) Date Trap site and other information

old treefall; 225658 taken 25 ft above on trunk across main ravine; caught 225651 here at same spot. Stomach: 225649 full of insect remains (similar to 225648) mixed in brown fruit mash; 225651 empty.

225648 (3554) 46 Nov. 1975 On terrace of main stream, about 4 ft above stream in area of old treefall; long sections of trunks and large limbs lay rotting on hillside and portion of stream terrace; about a month ago caught a Bunomys andrewsi nearby beneath one of these sections; 225648 caught beneath rotting section of trunk (1.5 in. diameter, 25 ft long). This piece of trunk is covered by rotting branches, woody vines and other debris, as well as herbaceous vines and shrubs; the space beneath most of the trunk provides protected run for small animals; area stays damp even when there is little or no rain; thick debris and leaf litter cover forest floor; this part of terrace is not flooded as are washed parts nearby; understory is open, but area of treefall is surrounded by good forest; undercover is dense and shrubby as is typical of open areas that are regenerating after treefalls. Stomach: full of well-masticated and digested insects (legs, antennae, sclerites, larval abdomens; no termites) mixed with a little fruit, including a few hard oblong seeds.

225647 (3548) 46 Nov. 1975 In runway (6–8 in. wide) along rotting trunk (1.5 in. diameter, 20 ft long) on hillside above third tributary and main stream; good forest; trunk is from an old treefall that is shaded by two large clumps of rattan and a young fan palm ( Livistona rotundifolia ), so the ground is clear; also muddy, partly covered by leaf litter; canopy is closed overhead by mainly an emergent strangler Ficus ; ula ( Diospyros macrophylla ) is common, understory forest intact. Stomach: crammed full with worker and soldier rhinotermitid termites; nothing else.

225650 (3563) 54 Nov. 1975 On damp ground that is a protected runway beneath two limbs from an

225654 (3652) ula that lay on hillside above main stream; good hillside forest, but here in this spot several large trees have fallen, probably within last year, so the understory is scrubby and filled with rattan, shrubs, tall gingers, short young trees, young palms, and vines. The two limbs are covered with herbaceous and woody vines that fell with the tree; they, along with the undercover, form a matlike canopy over rotting trunks and limbs, shading the ground and providing good cover. The area is usually wet, whereas farther up the hillside, the ground dries quickly; entire area for about one-eighth acre is a jumble of scrub-enclosed rotting trunks and limbs interspersed between tall canopy trees; 225652 and 225654 also caught here at same spot. Stomach: 225650 partly full, cut chunks of earthworm, many insect sclerites, all in an unidentifiable brownish mass; 225654 filled with orange mash containing brown globs mixed with parts of tiny insects.

225652 (3579) 54 Nov. 1975 In same area of terrace slope high above stream where 225650 was caught. Trap set beneath rotten trunk (2.5 ft diameter) lying horizontally across hillside; it is surrounded by dense undergrowth of shrubs, seedlings and debris; understory is also dense; a huge rattan winds through and over everything (probably fell with the tree), the palm Licuala celebica , small Pandanus , and short trees abound; so space beneath trunk is enclosed on sides by good, dense cover and stays relatively damp. This spot is in an area of old treefalls that are interspersed among the remaining canopy and tall understory trees, and the treefalls form large, scrubby open areas in forest canopy. Stomach: ate variety of invertebrates and fruit during captivity (see text).

TABLE 54 (Continued)

Bunomys andrewsi has also been obtained in habitats where the original forest has been altered by human activity. The samples from Gunung Balease and the Malili area (Desa Lawaki Jaya) were trapped in secondary forest (J.L. Patton, personal commun., 2011). At Omu, a village at the lower end of my transect, a rat was ‘‘taken in a garden at edge of a swampy area (notation on the skin tag). In the Mamasa area, B. andrewsi was commonly encountered in second-growth forest and cacao gardens (K.C. Rowe, in litt., 2012).

We never trapped examples of B. andrewsi on woody vines above ground or in trees. It, like other species of Bunomys , is terrestrial and active during the night.

Diet: Bunomys andrewsi consumes a variety of invertebrates, vertebrates, and fruit, but not fungi; the categories and range of items in each are similar to those constituting the diet of B. chrysocomus (see that account and table 13). My dietary records derive from feeding three captive B. andrewsi collected at Kuala Navusu (two adults and a juvenile) and study of samples from stomachs of animals captured at Kuala Navusu, along my transect in the west-central region (Sungai, Oha Kecil, Sungai Miu, and Sungai Sadaunta), and specimens from Gunung Balease and the Malili area (obtained by J.L. Patton, K.C. Rowe, and J.A. Esselstyn); see tables 54 and 55.

All three captives aggressively attacked and consumed invertebrates. Each would instantly grab the offered earthworm or insect from my fingers, scamper to an opposite corner of the cage with the prey in its mouth, stop and watch me for a few seconds, then turn its back to me and begin eating. Even flies were chased about the cage.

TABLE 55

Summary of Stomach Contents for Specimens of Bunomys andrewsi collected in Central Sulawesi at Gunung Balease (830–925 m) and the Malili Region (Desa Lawaki Jaya, 450 m) during October and November, 2010 Collection locality and specimen numbers are included. The two collection localities are in second-growth tropical lowland evergreen rain forest (collectors: J.L. Patton, K.C. Rowe, and J.A. Esselstyn). Locality, MVZ and collector numbers Contents of stomach Gunung Balease 225684 Crammed ; some pieces of fruit pulp and skin, but rest is mostly conglomeration of rhinotermitid ( JLP 25238) termites (no soldiers, only workers and winged stages), at least one geophilomorph centipede ; legs and isolated sclerites from other kinds of small insects (probably small beetles).

225687 Full; mostly pulp and tiny seeds from figs; pulp and hard thin skin from one other kind of fruit; many (JAE 3474) fragments of rhinotermitid termites (highly fragmented, mostly workers); badly decomposed small macrolepidopteran larval skins; several segments from posterior tip of an earthworm (also some

tissue that looks like internal earthworm parts); legs and sclerites from small adult beetles.

225688 Full; small amount of fruit mash; few large hard, composite seeds that break into smaller chunks; (JAE 3492) large clumps of rhinotermitid termites (bodies of workers and soldiers, disintegrated head capsules); many larval remains from three different kinds of beetles; remains (legs, wings, pieces

of elytra) of many small adult beetles; tiny bits of debris that resembles soil; fragments of moss.

Clearly mostly full of insects.

225689 Full; packed with remains of rhinotermitid termites (workers) and white spherical eggs (likely (JAE 3493) termite eggs); a bit of tan fruit mash and rind; pieces of fig rind with large seeds embedded in the inner wall; pieces of exoskeleton of a large adult insect.

225693 Full; few macrolepidopteran larval skins; many remains (head capsules, legs, abdomen) of (JLP 25237) rhinotermitid termites; at least one large cursorial (1.5–2 in. long) beetle larva chewed into pieces; internal tissue and sclerites of legless smaller beetle larvae; few ants; a few very small

intact legless beetle larvae (usually found on fruit); unidentified pulp (blackish debris eroding

from blackish chunks).

225694 Full of reddish brown fruit mash; mixed in are a few fragments of a small adult insect (probably a (JLP 25238) beetle). 225707 Partially full; mostly brown, tan, and reddish fruit pulp; a few fragments of a small insects; one small (KCR 1166) macrolepidopteran larval skin. The insect parts are minor and fruit pulp is the main component. 225708 Partly full; mostly tan fruit cut into small pieces; moderate amount of rhinotermitid termites (head (KCR 1170) capsules most evident); intact geophilomorph centipede; few fragments of a small adult insect. 225710 Full, some peanut butter bait, rest fruit mash and one chewed large beetle larva (covered with (KCR 1175) setae). 225711 Full; fruit pulp and skin; numerous fragments from small adult beetles mixed in with the pulp. (KCR 1177) 225712 Partly full; mostly fruit pulp and skin, including coconut bait; fragments of legs and isolated (KCR 1179) sclerites of small insects. 225812 Full; tan, somewhat fibrous fruit mash with slivers of skin; mixed in are a few fragments of small (KCR 1168) insects. Desa Lawaki Jaya 225690 Partly full; contents as in 225715, remains of rhinotermitid termites and what appears to be termite (JAE 3539) nest debris; no sign of figs. 225691 Full; dark gray claylike debris containing small white inclusions (same material as in 1198); tannish (JAE 3540) orange pulp packed with white seedlike objects, one of the figs; fragments of rhinotermitid termites (bodies, fragmented head capsules).

225700 Full: dark brownish fruit mash in which are mixed fragments of small insects; a few rubbery (JLP 25308) blackish pieces that are probably snail. 225703 Partially full; fruit pulp mixed with unidentified debris; numerous rhinotermitid workers and (JLP 25312) winged stages. 225705 (25325) Partially full; chopped fruit pulp mixed with fragments of rhinotermitid termites. 225715 Packed; black and tan debris in which are mixed fragments of many rhinotermitid termites. The (KCR 1192) debris may be part of termite nest—seen it in one other stomach (25237); pieces of fig rind and a lot of endosperm containing tiny fig seeds.

TABLE 55

(Continued)

Locality, MVZ and collector numbers Contents of stomach

225716 Full; all remains of rhinotermitid termites mixed with debris that appears to be from termite nest. (KCR 1197) 225717 Full; a few intact adult ants, ant fragments and head capsules; many ant pupal cases; chunks of (KCR 1198) rubbery invertebrate tissue that is likely snail; tiny black ‘‘seeds’’ everywhere that could be fig, some larger black seeds; clay-like debris containing small elongate white particles packed

together; a few rhinotermitid termites. Looks as if rat dug into or beneath in rotting wood and

found a few termites and ant colony.

225719 Full, mostly fruit pulp containing small tan seeds; also chunks of invertebrate tissue that resembles (KCR 1202) well chewed earthworms, but cannot detect external indication of segments (could be snail tissue but did not find operculum).

225720 Full; same material as in 225715, remains of rhinotermitid termites and what appears to be termite (KCR 1203) nest debris; no sign of figs. 225721 Packed; unidentified debris and rhinotermitid termite remains; also what appears to be pieces of (KCR 1204) fruit.

If a rat was eating fruit and offered an insect, the fruit was instantly dropped and the insect yanked from my fingers.

Items constituting the diet of B. andrewsi are summarized below.

Earthworms —Captives grabbed earthworms from my fingers and consumed them in less than a minute, handling the worms in the same fashion as described for B. chrysocomus . Cut segments of earthworms were found in stomachs of other B. andrewsi (tables 54, 55).

Snails —When the three captives were offered snails (half-inch to an inch in diameter), they unhesitatingly began to process them. Each rat held the shell in its front paws, made an opening with its incisors and then extracted the snail’s body, all of which was consumed. The behavior was similar to that I described for B. chrysocomus (see that account).

Insects and other arthropods —Captives were offered moths, two kinds of large crickets (3 inches long), grasshoppers, praying mantises, several kinds of adult beetles, and large beetle larvae found in dry and rotting wood. All were quickly accepted and eaten, the behavior with these prey items was very similar to that described for B. chrysocomus . Contents of stomachs included rhinotermitid termites, adult ants and ant pupal cases, a variety of adult and larval beetles (including legless and cursorial larvae), macrolepidopteran larvae, and geophilo- morph centipedes (tables 54, 55). Rhinotermitid termites were the most common insect found in stomachs, 18 of 42 stomachs contained these insects (I omitted stomachs that were empty or contained only bait); the stomach of a rat collected at Kuala Navusu was crammed with worker and soldier rhinotermitid termites and contained no other invertebrates and no fruit.

The stomach contents of a rat collected on Gunung Balease illustrates the kinds of insects procured by B. andrewsi . The stomach was full, containing a few macrolepidopteran larval skins, many remains of rhinotermitid termites (head capsules, legs, abdomens), at least one large cursorial beetle larva (about 2 inches long) chewed into pieces, internal tissue and sclerites of smaller legless beetle larvae, a few very small intact legless beetle larvae, a few ants, and a bit of unidentified blackish debris.

Other invertebrates —I gave an adult male a small land crab, which he quickly accepted, used his front feet to manipulate the crab, bit the carapace in several places, but then rejected it, and showed no further interest.

Vertebrates —I did not find vertebrate remains in the stomachs examined, but two of the captives readily accepted and consumed geckos. The largest offered was about 7 inches long, 4 inches of which were head and body. I held it by the body and presented it to an adult male who instantly grabbed the gecko by the head and pulled it out of my fingers. He bit the head, chest, and abdomen several times and then manipulated the lizard until its head was between the front paws and then in his mouth. By this time the tail had come away from the body and the gecko was immobile. The rat began eating from the nose back toward the rump. Occasionally the lizard jerked, but the rat quickly placed a hind foot on the lizard’s pelvis to stabilize the gecko as it chewed on the other end. Once, the rat turned the gecko’s body around and ate from the rump for a bit, then returned to the head, finally consuming the entire body; the detached tail was also eaten. The process took just a few minutes and only a few pieces of breast bone remained on the cage floor. It was clear the rat recognized the gecko as prey and did not hesitate to attack and eat it. Another rat ate a smaller gecko in much the same fashion. I found the geckos inside decaying tree limbs lying on the forest floor, the same kinds of rotting wood that provides habitat for cerambycid beetle adults and larvae, rhinotermitid termites, and ant pupae, as I described in the account of B. chrysocomus .

Fruit —Different fruits offered the three captives were unambiguously accepted or rejected. Figs—whether from tall canopyforming strangler Ficus , other nonstrangler species that contributed to the canopy, or understory species—were accepted and readily consumed by all the captives. Figs are one of the most common of the identifiable fruit remains found in contents of stomachs (tables 54, 55). Also accepted and eatern were fruit from understory palms ( Pinanga sp. ); from the trees Sandoricum sp. , Sapium sp. , Palaquium obtusifolium , and another unidentified member of Sapotaceae ; large seeds from the tree Hydnocarpus sumatrana ; fruit from the vine Gnetum cuspidatum ; and fruit from Pandanus .

The captives ignored the dark green fruit from Matthaea sanata , a shrubby tree growing in understory; from the nutmegs, Knema sp. and Myristica sp. ; the citruslike fruit from the understory tree Dillenia serrata ; large, hard nuts from fruit of the understory tree, Pangium edule ; and the large, green fruits from the understory tree, Kjellbergiodendron celebicum .

Rarely did a stomach contain only fruit. For example, one rat from Gunung Balease had a stomach full of mostly pulp and tiny seeds from figs as well as pulp and hard thin skin from one other kind of fruit. Many fragments of rhinotermitid termites, small macrolepidopteran larval skins, several segments of an earthworm, and legs and sclerites from small adult beetles were mixed with the fruit remains.

Fungi —The ear fungus Auricularia delicata (called ‘‘ karoko ’’ by the local people) and a few kinds of jelly fungi were broken into pieces by all three of the captives, but the fragments were eventually dropped to the cage floor and ignored. I did not find macroscopic remains of fungi in any of the stomachs examined.

Overview —Like B. chrysocomus (see that account), B. andrewsi consumes fruit, mostly figs, disdains fungi, and is an aggressive predator of invertebrates and small vertebrates. Near camp on the Kuala Navusu, my helpers and I searched for places on the forest floor where B. andrewsi would find invertebrates. We first looked to earthworms. During the night after a rain we found earthworms scattered on the ground surface, beneath leaf litter, or in pools of water near the stream. During the day we located them in rotting sections of tree trunks and limbs or in the soil beneath. We also found burrows, each marked by a pile of processed soil, some as tall as 2 inches. Several small worms might occupy a single burrow and were found 2– 6 inches below ground. Burrows were always on well-drained slopes or stream terraces. Because the earthworms are spotty in their distribution, it is difficult to provide any meaningful estimate of worms available to B. andrewsi per unit area. We did find one place on a steep hillside above a wet ravine where a piece of wood had decayed into slivers. The soil was stable. We collected 21 earthworms (8 g total) in a square meter; all occurred throughout the soil from from near the surface to a depth of 6 inches.

The most common site for both large and small worms was within and beneath rotting trunks and limbs from old treefalls. At a certain stage of decomposition the wood becomes soft and is mixed at soil level with a gray or bluish gray dense claylike layer; the worms are in this layer, and in passages within the wet and rotting pulpy wood.

Decaying sections of old treefalls lying on the forest floor also provide habitat for a variety of invertebrates in addition to earthworms that are sought and eaten by B. andrewsi . I tore apart a long section of rotting trunk (15 ft long, 10–12 inches in diameter) lying on a terrace about 30 ft above a stream. The pulpy wood was saturated with water in places and the underlying claylike layer nearly liquefied. Pigs had torn into the wood in two places. I extracted 30 grams of invertebrates (excluding termites—the terrestrial rhinotermitid termites, I presume—that infested the entire trunk and were too numerous to collect and weigh): 21 earthworms (some large, up to 5 inches long, most smaller, 2–3 inches), 3 mole crickets, and 6 small adult and nymphal cockroaches. From different and drier decaying trunks, I extract- ed scorpions, two kinds of diplopods (millipedes), several kinds of chilopods (centipedes), large adult and larval beetles, a gecko, and a legless lizard.

Small frogs and snails seemed to be most prevalent on the banks and terraces bordering streams. I found snails on top of the leaf litter and just above ground on stems of shrubs and ferns.

Nests: I placed dry leaves in the cages of two adults and a juvenile. Each formed a pile of leaves, then burrowed into the side, and came out at the top of the pile. During the day the rat slept in the top of the pile nearly concealed on all sides by the leaves. After a few days the leaves became compacted and each rat would shift them around until they again formed a deep cup-shaped sleeping chamber.

I assume that B. andrewsi lives in underground burrows. Near Kuala Navusu one rat was caught in a trap placed at the mouth of a burrow partially concealed by rocks on a hillside.

ECTOPARASITES, PSEUDOSCORPIONS, AND ENDOPARASITES: Sucking lice, ticks, and mites are the ectoparasites recorded to date from Bunomys andrewsi (table 14). An undescribed species of Hoplopura (Anoplura) is unique to B. andrewsi (Durden and Musser, MS.) .

Immatures of the ticks, Dermacentor sp. and Haemaphysalis sp. ( Acari , Ixodidae ), have also been collected from a variety of hosts in addition to B. andrewsi : shrews (the endemic Crocidura elongata , and the commensal Suncus murinus ), rusa ( Rusa timorensis , nonnative), three endemic squirrels ( Rubrisciurus rubriventer , Hyosciurus heinrichi and H. ileile ), 10 other species of endemic murid rodents ( Bunomys fratrorum and B. chrysocomus ; Echiothrix centrosa ; Maxomys hellwaldii , M. musschenbroekii , and M. wattsi ; Paruromys dominator ; Taeromys sp. ; Rattus hoffmanni and R. facetus [recorded as R. marmosurus ]), and three nonnative Rattus tanezumi (recorded as R. rattus ), R. argentiventer , and R. exulans ( Durden et al., 2008; L.A. Durden, personal commun.). Two species of mites in the genus Laelaps ( Acari , Laelapidae ) have been documented from Bunomys andrewsi ( Van Peenen et al., 1974) .

Chiridiochernes platypalpus is a pseudoscorpion described from B. andrewsi that was trapped at the base of Gunung Lompobatang on the southwestern peninsula ( Muchmore, 1972).

Two kinds of endoparasites are known infect B. andrewsi . Host specimens from the Mamasa region were found to be parasitized by the nematode Bunomystrongylus miyagii (Trichostrongylina, Heligmonellidae ), which is host-specific (Hasegawa and Mngali, 1996). Bunomys andrewsi is also a host for a species of liver fluke, Platynosomoides , which was also recovered from the nonnative Rattus tanezumi ( Van Peenen et al., 1974) .

SYNONYMS: Three scientific names prove to by synonyms of Bunomys andrewsi . Data attached to the holotype and type locality, description of the taxon as originally published, and reasons for its inclusion in B. andrewsi are presented below, ordered by publication date.

Rattus adspersus Miller and Hollister, 1921a: 71 . HOLOTYPE: USNM 219602 , an adult male (skin and skull; measurements are listed in table 40) collected January 22, 1918, by H.C. Raven (original number 3427). TYPE LOCALITY: Indonesia, Sulawesi, Propinsi Sulawesi Tengah, Pinedapa (01 ° 25 9 S, 120 ° 35 9 N), 100 ft (30 m; locality 3 in the gazetteer and on the map in fig. 50), northeastern margin of the central core.

Miller and Hollister (1921a:71) diagnosed Rattus adspersus as:

Related to Rattus chrysocomus (Hoffmann) of North Celebes, but general coloration darker; the characteristic agouti-like flecking more brownish, less yellowish; sides and underparts especially less yellowish. Tail shorter, almost unicolor, only very slightly lighter colored on underside near base, and without light colored tip. Skull smaller.

The specimens of adspersus , remarked Miller and Hollister (1921a: 72),

have been compared with about 150 skins and skulls of Rattus chrysocomus (including for the present Mus fratrorum Thomas ) from numerous localities in North Celebes. The specimens of chrysocomus vary considerably in size but average larger than examples of the new form from Middle Celebes, with much more distinctly yellowish flecking. The northern species always has the terminal portion and the whole underside of the tail whitish. The color of the upperparts in some specimens of adspersus approaches very closely to that of the type specimen of Rattus andrewsi (Allen) from Pulo Boeton, off the coast of southeastern Celebes, which proves to be a member of the chrysocomus group.

Miller and Hollister identified as adspersus 23 specimens from Pinedapa and two from Tuare. Three additional examples are from Labuan Sore (see locality 1 in the gazetteer), one had originally been determined to be Rattus rallus by Miller and Hollister, the other two as Rattus hoffmanni . Their adspersus is unquestionably phenetically distinct from the samples collected by Raven from the northeastern peninsula that Miller and Hollister thought were B. chrysocomus but are actually B. fratrorum . The holotype of andrewsi resides in USNM and was handy for comparison. While Miller and Hollister recognized a link between adspersus and andrewsi in coloration of the dorsal pelage, they presumbably did not think the resemblance indicated attributes of a single species, which is my hypothesis based on analyses of morphometric traits as well as chromatic aspects of the pelage.

Miller and Hollister’s reference of adspersus to a ‘‘ chrysocomus group’’ was reaffirmed strongly by Tate (1936: 554) when remarking on the holotype of adspersus he wrote that ‘‘The skull of this form unquestionably indicates its affinity to the chrysocomus group.’’ Between 1936 and 1969, adspersus retained its status as a species of Rattus in, first, a Rattus chrysocomus group ( Ellerman, 1941), then a Rattus coelestis group ( Ellerman, 1949), in subgenus Rattus ( Laurie and Hill, 1954) , and finally in subgenus Bullimus of Rattus ( Misonne, 1969) . Except for chrysocomus , fratrorum , colestis, koka, and andrewsi, Ellerman (1949) and Laurie and Hill (1954) relegated all the other taxa that had been associated with a Rattus chrysocomus group to subspecies of R. adspersus (see table 4).

Rattus penitus inferior Tate and Archbold, 1935a: 6 . HOLOTYPE: AMNH 101059 (skin and skull; measurements are listed in table 40), an adult male collected January 23, 1932, by G. Heinrich (original nuber 808). TYPE LOCALITY: Indonesia, Propinsi Sulawesi Tenggara (southeastern peninsula of Sulawesi), Wawo (03 ° 41 9 S / 121 ° 02 9 E), 50 m (locality 21 in gazetteer and on the map in fig. 50).

Rattus penitus inferior was characterized by Tate and Archbold (1935a: 6) as ‘‘A large member of the chrysocomus group with coarser pelage than either penitus penitus or penitus sericatus . Under parts irregularly suffused with hazel. Skull massive for the group, with long, broad palatal foramina.’’ They then provided a description of the holotype:

Pelage of type with guard hairs distinct, though not longer than wool hairs; color of tips of hairs dull fawn color, the gray bases showing through and dulling the total effect. Under parts with fur mostly gray-based, tips whitish on throat, a little on sides and inside of limbs, otherwise tips hazel. Line of demarcation not sharp. Tail white beneath, not white-tipped.

Skull of type with large muzzle and sloping zygomatic plate. Length of palatal foramina exceeding length of molar crowns (118 per cent), which in turn exceeds length of bulla (110 per cent). The molar series, though short, is made up of rather heavy individual teeth (width of m 1 is 62 per cent of its length).

Their description of inferior was based on the holotype and nine additional specimens from southeastern Sulawesi (see localities 16 and 17 in the gazetteer and on the map in fig. 50). Both Tate (1936) and Ellerman (1941) treated inferior as a subspecies of Rattus penitus, Sody (1941) associated inferi- or with penitus but arranged the latter as a species of Frateromys , and later Ellerman (1949) treated inferior as a subspecies of Rattus adspersus , a link accepted by Laurie and Hill (1954). By the early 1990s, inferior was subsumed under Bunomys andrewsi ( Corbet and Hill, 1992) where it has remained ( Musser and Carleton, 1993, 2005; see table 4). See the preceeding section on Comparisons where the identity of the holotype of inferior as a sample of B. andrewsi and not B. penitus is demonstrated.

Rattus penitus heinrichi Tate and Archbold, 1935a: 6 . HOLOTYPE: AMNH 101006 (skin and skull; measurements are listed in table 40), an adult male collected August 31, 1931, by G. Heinrich (original number 365). TYPE LOCALITY: Indonesia, Propinsi Sulawesi Tengah, Lombasang (05 ° 16 9 S, 119 ° 55 9 E), 1100 m (locality 25 in the gazetteer and on the map in fig. 50), in the foothills of Gunung Lompobatang , southwestern peninsula of Sulawesi.

Rattus penitus heinrichi was characterized by Tate and Archbold (1935a: 6) as ‘‘A medium-sized member of the chrysocomus group with dense though rather crisp pelage.’’ The holotype was described as follows:

Pelage of type coarser than that of R. p. penitus or R. p. sericatus , and much denser than R. p. inferior. Color above, rather strong sayal brown at tips of hairs, giving a general effect of natal brown due to dark-colored hair bases showing through from below. Under parts long-haired, the hairs between pinkish buff and cream buff, with gray bases. Hands and feet thinly clothed with whitish hairs. Tail whitish beneath, the scale-hairs about one and one-half scale lengths.

Skull of type more delicately formed than that of p. inferior, with narrower rostrum (? constant), smaller palatal foramina, shorter bullae, narrower mesopterygoid fossa.

Twenty-one specimens from Lombasang (locality 25 in gazetteer), in addition to the holotype, were identified as heinrichi .

After its description, heinrichi remained a subspecies of Rattus penitus in synoptic works published during the late 1930s and early 1940s ( Tate, 1936; Ellerman, 1941); was transferred to Frateromys , its identity intact as a subspecies of penitus , by Sody (1941); and later arranged as a subspecies of Rattus adspersus by Ellerman (1949) and Laurie and Hill (1954). During the 1990 s, Corbet and Hill (1992) and Musser and Carleton (1993) recognized heinrichi as a separate species, but by the next decade it lost its identity under Bunomys andrewsi ( Musser and Carleton, 2005) where it currently remains (table 4). Why heinrichi is a peninsular population of B. andrewsi and not a subspecies of B. penitus is discussed in the preceeding section covering Comparisons.

SUBFOSSILS: Two maxillary fragments, three pieces of dentaries, and an incisor fragment are the subfossils representing B. andrewsi that have been found in caves on the southwestern peninsula of Sulawesi.

Two maxillary fragments, two pieces of dentaries, and an isolated segment of incisor (figs. 70, 71; tables 56, 57) were excavated from a shallow deposit forming the floor of a small rock shelter, Batu Edjaja II, that has been described by Mulvaney and Soejono (1970; 167, plate V; also see locality 26 in the gazetter and on the map in fig. 50). A few artifacts were recovered (geometric microliths and pottery shards), but the deposit had been so disturbed that no reliable age could be attached to them or to the rodent remains. A sample of charcoal from near the bottom of the deposit was determined to be ‘‘modern’’ by C-14 dating. David Bulbeck (in litt., 1997) informed me that ‘‘Batu Ejaya 2 has abundant geometric microliths suggesting some occupation by at least 3000–2000 B.P., but the basal radiocarbon date is modern and the deposits are obviously disturbed comprehensively’’ (also see Bulbeck, 2004; Simons and Bulbeck, 2004).

One right dentary fragment lacking molars (fig. 71; tables 56, 57) was recovered from sediments excavated at Ulu Leang I, a cave in the Maros region of the southwestern peninsula (locality 24 in the gazetteer and on the map in fig. 50; also see the map of archaeological sites in Simons and Bulbeck, 2004: 168). Excavations in the cave were described by Glover (1976), who also provided me with 8785 ± 45 years B.P. as the date of the horizon (layer V) in which the piece was located (Glover, personal commun.); a different assessment by Bulbeck (2004: 132) gives around 7500 B.P. for layer V. Specimens of B. chrysocomus were found in sediments from the same cave (see account of that species).

I compared all these subfossil fragments with modern samples of B. chrysocomus from Sadaunta in the west-central region of Sulawesi (no modern samples of B. chryso-

TABLE 56

Subfossil Bunomys andrewsi from the Southwestern Peninsula of Sulawesi

Cave and specimen Age Description

BATU EJAJA II

AMNH 265015 adult Left maxillary fragment containing partial first molar and intact second molar (fig. 70).

AMNH 265015 adult Left maxillary fragment with intact first molar (fig. 70).

AMNH 265013 young Fragment of left ramus of dentary with intact molar row; masseteric ridges and mental adult foramen evident; no incisor (fig. 71).

AMNH 256014 adult Piece of anterior portion of right ramus of dentary containing intact first molar; no incisor (fig. 71).

AMNH 265020 adult Distal fragment of right lower incisor.

ULU LEANG I AMNH 266974 adult Anterior portion of right dentary with most of incisor; no molars (fig. 71).