Archboldomys (Muridae: Murinae) Reconsidered: A New Genus and Three New Species of Shrew Mice from Luzon Island, Philippines

ABSTRACT Shrew mice of the genus Archboldomys are poorly known members of an endemic clade of vermivorous/insectivorous murid rodents confined to Luzon Island, Philippines. Three species of these small, ground-living, diurnal mice were previously known, all from a handful of specimens from a few localities. The pattern of morphological and genetic differentiation among additional specimens of shrew mice from our recent field surveys in the Central Cordillera and Sierra Madre mountains of Luzon document the presence of two distinct species groups within Archboldomys as previously defined, as well as three new species. Gene-sequence data from the mitochondrial cytochrome b and nuclear IRBP genes confirm the existence of six distinct species, but also show that Archboldomys, as previously defined, is composed of two clades that are not sister taxa. Reevaluation of the presumed morphological synapomorphies among these shrew mice, together with analyses of karyological and gene-sequence data, support the following: (1) erection of Soricomys, new genus; (2) transfer of A. kalinga and A. musseri to Soricomys; and (3) recognition of Archboldomys maximus, n. sp., Soricomys leonardocoi, n. sp., and Soricomys montanus, n. sp. The new genus and species are described, and their phylogenetic relationships, biogeography, and conservation are discussed.


INTRODUCTION
Shrew mice of the genus Archboldomys are part of one of the oldest known radiations of murid rodents in the Philippines, and provide important insights into the dynamics of mam¬ malian diversification on isolated oceanic islands (Jansa et al., 2006). Together with the shrew rats Chrotomys and Rhynchomys, their closest relatives, they form a clade of terrestrial murid rodents adapted to feeding on earthworms and soft-bodied invertebrates on the cool, damp forest floor of montane and mossy forest on the mountaintops of Luzon Island in the northern Philippines Duya et al., 2011;Jansa et al., 2006;Rickart et al., 1991Rickart et al., , 2011a. The sister taxon to these three genera is the genus Apomys, forest mice that also prefer soft-bodied inverte¬ brates, but additionally feed on seeds and small fruits. These four genera form a clade that is endemic to the Philippines (referred to informally as the Chrotomys Division by Musser and Carleton, 2005, or as members of the Hydromyini within the Murinae by Aplin and Helgen, 2010; see also Jansa and Weksler, 2004;Rowe et al., 2008;Heaney et al., 2009), exhibiting spectacular ecological and morphological diversity, from small arboreal mice to ground-hopping, long¬ snouted, shrewlike rats. At least 45 species are currently known (Heaney et al., 2010. Some members of this clade (e.g., Chrotomys and Rhynchomys) had previously confounded biologists, who speculated on the possible phylogenetic alliance(s) of this morphologically distinctive group. Based purely on morphological grounds, these Philippine endemics have been allied to Australo-Papuan murines in various subfamilial categories in the past, including Rhynchomyinae and Hydromyinae (Thomas, 1898;Ellerman, 1941;Simpson, 1945;Misonne, 1969). Among the members of this clade, the then monotypic genus Archboldomys, proved to be the most enigmatic. Since its description (Musser, 1982a) it was thought to be closely allied to Crunomys, another genus of shrewlike mice from the Philippines and Sulawesi (Musser, 1982a;Musser and Heaney, 1992). In 1998, the discovery of what was then considered another member of Archboldomys (A. musseri from Mt. Cetaceo, northeastern Luzon) and an additional new species of Crunomys (C. suncoides from Mindanao), together with karyotypic data from species of both genera, enabled a tentative reassessment of the inferred phylogenetic alliance between Archboldomys and Crunomys . The results confirmed the phylo¬ genetic link of Archboldomys with the Chrotomys Division, but found evidence that a close relationship with Crunomys was ambiguous, with possible convergence of the two on a shrew¬ like (i.e., Soricidae-like) morphology and trophic adaptation. This shrewlike morphology is associated with their heavy use of earthworms and other soft-bodied invertebrates captured in leaf litter Heaney et al., 1999;.
New characters and character states first evident in A. musseri and confirmed later in another new species, A. kalinga, were discussed within the context of broadening the morpho¬ logical range of Archboldomys Balete et al., 2006). Subsequent DNA sequence data analyses supported the placement of A. luzonensis within the endemic Chroto-mys Division of Philippine rodents, including Apomys, Chrotomys, and Rhynchomys; Crunomys, in contrast, was recovered as a sister taxon to Maxomys, within an entirely different clade of murines from the Philippines and the Sunda Shelf (Jansa et al., 2006).
Our additional fieldwork during 2006 to 2009 in the Central Cordillera and Mingan Mountains, both on Luzon Island ( fig. 1), resulted in additional specimens of shrew mice from several locations . Our analyses of their external, cranial, and dental features, together with additional karyological data from A. kalinga, and molecular sequence data from representatives of the Chrotomys Division, document the presence of three additional species of shrew mice described below. Unexpectedly, molecular sequence data has failed to support the monophyly of Archboldomys; although the two "species groups" are strongly supported as members of the Chrotomys Division, we find that Archboldomys, as currently defined, is paraphyletic and includes two divergent genera: Archboldomys and Soricomys, new genus, described below.

Field and Preparation Procedures
The specimens examined in this study were all collected by the authors and their associates, except the holotype of Archboldomys luzonensis, which was collected by D.S. Rabor in 1961 (Musser, 1982a;Heaney et al., 1999;Rickart et al., 1991Rickart et al., , 1998Balete et al., 2006). These speci¬ mens are deposited at the Field Museum of Natural History (FMNH), the United States National Museum of Natural History, Smithsonian Institution (USNM), and the Philippine National Museum (PNM). The capture and handling of animals in the field followed all rele¬ vant laws and regulations of the Philippines.
Tissue samples were taken from the thigh muscle of fresh specimens and preserved in either 95% ethanol or DMSO buffer. Most of the specimens were injected with saturated for¬ malin solution in the field, stored temporarily in 10% formalin, and subsequently transferred to 70% ethanol. Some skulls were then removed, cleaned with dermestid beetles, and briefly soaked in a weak ammonia solution. Some specimens were skeletonized in the field; these were cleaned in the same fashion as the skulls.
Reproductive data were taken in the field from skeletonized specimens or in the laboratory from autopsied specimens stored in 70% ethanol, including, for males: testes descent (scrotal or abdominal), testes size (length x width, in mm), and convolution of the epididymis; for females: number and position of mammae (inguinal, abdominal, axial), size and condition of mammae (small, large, lactating), presence or absence of vaginal perforation, number and size (crown to rump length, in mm) of embryos, and number of placental scars in the uterus.
Age determination was done in the field condition of freshly caught specimens, and subsequently validated based on molar wear and fusion of cranial sutures, following the age categories defined by Musser and Heaney (1992). We followed the terminology of Brown (1971) and Brown and Yalden (1973) for defining the external features of the head and limbs. Terminology for cusps on molar teeth ( fig. 2) and cranial foramina ( fig. 3) was adapted from Musser and Heaney (1992: figs. 2, 8). FIG. 2. Molar teeth of Archboldomys maximus, n. sp. (FMNH 193531, holotype), illustrating terminology used to describe cuspidation. Upper molars (left): cusps are numbered and referred to in the text with the prefix "t"; see Musser and Heaney, 1992. Lower molars (right): ala, anterolabial cusp; ali, anterolingual cusp; ed, entoconid; hd, hypoconid; md, metaconid; pc, posterior cingulum; pd, protoconid based on relative body size and reproductive FIG. 3. Lateral and ventral views of the cranium of Archboldomys maximus, n. sp. (FMNH 193531, holo¬ type), showing morphological features discussed in the text. Abbreviations: ab, auditory bulla; al, alisphenoid; bo, basioccipital; bs, basisphenoid; cc, carotid canal; et, eustachian tube; fo, foramen ovale; ms, mastoid; pet, petrosal portion of the petromastoid complex; pf, pterygoid fossa; pgf, postglenoid fossa; oc, occipital bone; pt, periotic part of the petromastoid complex; ptb, pterygoid bridge; ptr, pter¬ ygoid ridge; sq, squamosal; srza, squamosal root of zygomatic arch; stf, stapedial foramen; th, tympanic hook; trc, transverse canal. AMERICAN MUSEUM NOVITATES NO. 3754 Morphological Methods External measurements (in millimeters) were taken in the field from fresh specimens, includ¬ ing total length (TOTAL), length of tail (TAIL), length of hind foot including claws (LHL), length of ear from notch (EAR), and weight in grams (WT); field catalogs of the collectors with these measurements and additional notes have been deposited at EMNH or USNM. Additional mea¬ surements of length of overfur (LOE, measured in the middorsal region) and number of tail scale rings per centimeter (TSR, counted at a point on the tail one-third of the total length from the base) were taken from fluid specimens preserved in 70% ethyl alcohol. The length of head and body (HBL) is the difference of the length of tail subtracted from total length. Scanning electron micrographs of crania, mandibles, and teeth were made from uncoated specimens with an AMRAY 1810 scanning electron microscope.
Eighteen cranial and dental measurements of 44 adult specimens were taken with a digital caliper and recorded to the nearest 0.01 mm by Heaney, following the terminology and limits of these measurements defined in  see also Musser and Heaney, 1992) Descriptive statistics (mean, standard deviation, and range) of cranial and dental measure¬ ments were calculated from sample groups. We assessed quantitative phenetic variation through principal components analysis (PCA), using the correlation matrix of log10-transformed mea¬ surements of adult specimens, using SYSTAT 10 for Windows (SPSS, Inc., 2000).

Karyology Methods
We prepared karyotypes from bone marrow cells following in vivo methodology (Patton, 1967, as modified by Rickart et al., 1989). Freshly killed animals captured in snap traps were processed using a modified in vitro technique . Standard giemsa-stained karyotypes were prepared in the laboratory from fixed-cell suspensions. A minimum of 10 chromosome spreads was examined from each preparation. Chromosome terminology follows Rickart and Musser (1993).
Because sex chromosomes could not be identified with certainty in many karyotypes, fundamental number (FN) refers to the total number of chromosome arms (including those of the sex chromo¬ somes). Microscope slides and photomicrographs cross-referenced to voucher specimens at FMNH and USNM are housed at the Utah Museum of Natural History, University of Utah, Salt Lake City.

Molecular Genetic Methods
We used DNA sequences from the mitochondrial cytochrome b and the nuclear IRBP gene to infer phylogenetic relationships for members of the "vermivore clade" of Philippine endemic rodents (Clade D of Jansa et al., 2006). Our cytochrome b data set comprised 58 specimens, including at least one representative of all recognized species of Chrotomys, Rhynchomys, and Archboldomys, as well as several new specimens whose phylogenetic position in this clade was uncertain. In addition, we included sequences from seven species of Apomys. We also used sequences from Batomys and Phloeomys-members of the cloud rat clade (Clade E)-as out¬ groups to root our phylogenetic analyses (Jansa et al., 2006;Rowe et al., 2008). Our IRBP data set comprised 19 specimens that were a subset of the larger cytochrome b data set. These 19 specimens were chosen as representative individuals from the set of newly collected specimens, and included sequences from exemplar species of Rhynchomys and Apomys, as well as from all recognized species of Archboldomys and Chrotomys (see Jansa and Weksler, 2004).
DNA was extracted using a Qiagen DNA Minikit (Qiagen, Inc.) from tissue preserved in 95% ethanol. We amplified and sequenced the complete cytochrome b gene using primers and PCR conditions, as described in Jansa et al. (2006). DNA sequences were aligned using MUSCLE (Edgar, 2004) with default settings as imple¬ mented in Geneious v. 5.4.5 (Drummond et al., 2006). Where necessary, alignments were adjusted with reference to translated amino acid sequences for each gene. We analyzed each gene data set using maximum parsimony (MP) and maximum likelihood (ML) methods of phylogenetic inference. Analyses using MP were performed in PAUP* v4.0bl0 (Swofford, 2002) using heuristic searches with 1000 replicates of random taxon addition and TBR branch swap¬ ping. For the ML analyses, we determined the best-fitting model of sequence substitution among the models available in MrModelTest ver. 2.3 (Nylander, 2004) using the Akaike Infor¬ mation Criterion (AIC). We specified this model in an ML search in GARLI ver. 2.0 (Zwickl, 2006). Nodal support values were estimated using 1000 bootstrap replicates for the MP analyses and 300 replicates for the ML analyses. We also performed model-based (ML and Bayesian) phylogenetic analysis on the com¬ bined IRBP + cytochrome b gene data set. To do so, we evaluated the relative fit of four datapartitioning schemes: (1) a single partition including both genes; (2) two partitions, one for each gene; (3) three partitions, one for each codon position; and (4) six partitions, one for each codon position for each gene. We first determined the best-fitting model of sequence substitu¬ tion for each possible partition separately, using the AIC in MrModelTest. To test the relative fit of the four broad partitioning schemes to our combined-gene data set, we calculated the log-likelihood of each using Bayesian inference in MrBayes 3.1.2 (Ronquist and Huelsenbeck, 2003). To do so, we specified the resulting best-fit sequence substitution model for each data partition as appropriate, and we unlinked all substitution parameters across partitions. We estimated branch lengths relative to each partition and allowed the substitution rate to vary among the specified partitions. For each partitioning scheme, we ran two chains of Metropoliscoupled Markov chain Monte Carlo (MC3) for 3 x 106 generations, sampling every 100th generation. We discarded the first 10% of each run as burn-in and calculated the harmonic mean log-likelihood (HML) for each partitioning scheme using Tracer 1.4 (Rambaut and Drummond, 2007). The resulting HML values were used to calculate the AIC and Bayesian information criterion (BIC) scores for each partition (McGuire et al., 2007); the partitioning scheme with the lowest AIC and BIC scores was then specified in a second Bayesian run with 10 x 106 generations of MC3.
We also used this optimal partitioning scheme in a partitioned ML analysis as imple¬ mented in GARLI ver. 2.0 (Zwickl, 2006). We specified the best-fit model for each partition and allowed the model parameters to be estimated from the data. We performed five indepen¬ dent search replicates, but did not otherwise alter the default conditions for the tree search algorithm; nodal support values were obtained based on 100 bootstrap replicates.

Morphological Comparisons
Substantial variation exists among adults of six putative species of shrew mice from Luzon in external, cranial, and dental measurements (tables 1-4). Males tend to average only slightly larger than females, so we pooled the measurements of both sexes for multivariate analyses.
A principal components analysis, conducted on 18 cranial and dental measurements (log10transformed) of 44 adult specimens of shrew mice having intact (or nearly intact) crania included Archboldomys luzonensis from Carmarines Sur Province (N = 7), A. maximus, n. sp., from Mt.
Pulag, Benguet Province, and Mt. Amuyao, Mountain Province (N = 10), and Soricomys leonardocoi, n. gen. and n. sp., from the Mingan Mountains, Aurora Province (N = 8). The first four components accounted for 87.6 % of the total variance (table 5). Most variables had positive loadings of high magnitude on the first component (accounting for 63.8% of the variance), indi¬ cating that much of the variation was related to size. The exceptions, with low loadings, were diastema length, postpalatal length, and breadth of incisors (which had a slight negative loading).
The second component (accounting for 10.1% of the variation) had high-magnitude positive loadings for basioccipital length, diastema length, postpalatal length, and breadth of the upper incisors, separating individuals with large skulls, relatively elongate diastemas and postpalatal regions, and broad upper incisors from individuals having the converse. The third component (8.9% of the variance) had high-magnitude positive loadings for postpalatal length, mastoid breadth, and basioccipital length, but high-magnitude negative loadings for diastema length, length of incisive foramina, and incisor width, which distinguishes individuals with the combina¬ tion of elongate skulls and postpalatal regions, broad mastoid regions, short diastemas, short incisive foramina, and narrow upper incisors from individuals having the converse. The fourth and following components had eigenvalues below 1.0, and so are not interpretable.
A bivariate plot of specimen scores on the first and second components ( fig. 4) shows clear separation of the six putative taxa into four groupings. On the first component, expressing mainly size variation, specimens of A. luzonensis together with A. maximus, n. sp., from Mt.
A PC A of the four putative species referred below to Soricomys, n. gen. (table 6, fig. 5), using the same set of cranial and dental measurements, showed many of the same features as those evident in the PC A just described (   Archboldomys luzonensis: As reported previously (Rickart and Musser 1993;, the karyotype has 2N = 26 and FN = 43 (table 7). The autosomal complement includes eight pairs of small to large metacentric or submetacentric chromosomes and four pairs of very small to medium-sized telocentric chromosomes. A large submetacentric X chro¬ mosome is present in both sexes. It is paired with a medium-sized telocentric Y chromosome in males, and with a large telocentric element in the single female available ( fig. 6A).
Soricomys kalinga, n. gen.: The karyotype of this species was reported previously (as Archboldomys musseri) by , based on a poor quality in vitro preparation.
The karyotype has 2N = 44, consisting of small to large telocentric or subtelocentric chromo¬ somes and some small submetacentric chromosomes. The X chromosome is tentatively identi¬ fied as the largest element, and is either telocentric or subtelocentric. The Y chromosome is tentatively identified as small and telocentric. The FN has not been determined ( fig. 6B).

COMPONENT^)
COMPONENT (3) FIG. 5. Projection of specimen scores onto the first two axes (A) and the third and fourth axes (B) of a prin¬ cipal components analysis of 18 cranial variables (log10-transformed) from 32 specimens representing four putative species of Soricomys, n. gen. Each species is shown enclosed in a maximally inclusive polygon.

Molecular Phylogenetic Analysis
The cytochrome b data set yielded 394 parsimony-informative characters for the 56 ingroup and two outgroup taxa (table 8). Parsimony analysis of this data set resulted in seven mini¬ mum-length trees with length of 1610 steps, Cl = 0.386, and RI = 0.817. The strict consensus of these seven trees differs topologically from the tree recovered from a maximum-likelihood analysis of the same data, but conflict is limited to poorly supported (bootstrap < 75%) nodes.
We therefore present the maximum-likelihood tree annotated with support values from the parsimony analysis ( fig. 7).
Among the available substitution models, a GTR + I + T4 model of sequence evolution pro¬ vided the best fit to the cytochrome b data (-InL = 8261.366). Parameter estimates indicate that cytochrome b is markedly guanosine poor in these taxa and that transitions greatly outnumber transversions (   specimens assigned to A. kalinga (assigned below to Soricomys montanus, n. sp.) and A. musseri are not closely related to the type species A. luzonensis, but are part of a distinct clade. In addi¬ tion, these data also recover the sister-taxon relationship between A. luzonensis and the large shrew mouse from Mt. Amuyao. The IRBP data do not recover a monophyletic Chrotomys, but monophyly is not inconsistent with these data: C. silaceus and C. sibuyanensis form an unre¬ solved polytomy with a clade comprising the remaining three species of Chrotomys. There is strong support for a sister-taxon relationship between the new genus and species of Chrotomys; otherwise, IRBP provides little additional support for intergeneric relationships.
A model that partitions the data set six ways (by gene and by codon) provided the best fit to the combined-gene data set (table 10) trees based on the individual gene data sets, and recovers the principle relationships discussed above. In addition, this tree provides additional intergeneric resolution: the split between Apomys and a clade comprising the remaining four genera receives some statistical support in this analysis (ML-bootstrap = 75%; BPP=0.98); however, relationships among Archboldomys, Rhynchomys, and the clade comprising Chrotomys and the new genus remain uncertain.

Taxonomy
Analyses of the combined morphological, chromosomal, and molecular data analyses of all representatives of the shrew mice of the Philippines strongly support the following taxo¬ nomic decisions: (1) redefinition of Archboldomys, (2) establishment of Soricomys, n. gen., described below, and (3) recognition of one new species of Archboldomys and two new species of Soricomys, n. gen. We emend the diagnoses of all previously recognized species of Archboldo¬ mys (A. luzonensis, "A." kalinga, and "A." musseri), and transfer two into the new genus Sorico¬ mys (S. kalinga and S. musseri).
Comments: Archboldomys and Crunomys: Musser (1982a) noted the similarity of Arch¬ boldomys to Crunomys, specifically C. melanius and C. celebensis, and hinted at a possibly close phylogenetic relationship, based in particular on the shape, simplified cuspidation, and occlusal patterns of their molars. He proposed the hypothesis that "the old native rodents of the Philippine Islands may represent an adap¬ tive radiation in which all members are more closely related to each other than to rats and mice on the Asian mainland to the west, Sulawesi and the Lesser Sundas to the south, or to Australia and New Guinea area to the east." In spite of including one species from Sulawesi (C. celebensis), Crunomys was grouped with Archboldomys, Chrotomys, Celaenomys (now part of Chrotomys;Rickart et al., 2005), and Rhynchomys. This was further reinforced in the first assessment of phylogenetic alliances of the native Philippine murid fauna (Musser and Heaney, 1992), in which the Crunomys Group was erected within the Old Endemic Division (Division I), consisting of A. luzonensis, Crunomys fallax, C. melanius, and C. rabori (now part of C. melanius; see Rickart et ah, 1998 Heaney, 2002). Our phylogenetic analyses of molecular data from representatives of these spe¬ cies and other members of the Chrotomys Division (Apomys, Chrotomys, and Rhynchomys) not only supported the above morphological and karyological differentiation between the two groups, it also failed to recover the monophyly of Archboldomys. Instead, it showed that the small-bodied members (Soricomys, n. gen.) are the sister taxon to Chrotomys, and the large¬ bodied members (Archboldomys sensu stricto) are sister to the clade that includes Soricomys, Chrotomys,and Rhynchomys (figs. 7,8). This evidence clearly supports the establishment of a new genus, Soricomys, that accommodates two shrew mice formerly attributed to Archboldomys (S. kalinga and S. musseri), and two additional new species, as described below.

Comparisons: See Description and Comparisons section of A. maximus, below.
Distribution: Known only from Mt. Isarog, southeastern Luzon Island Balete and Heaney, 1997;Heaney et al., 1999). Ecology: Archboldomys luzonensis is a montane and mossy forest specialist, occurring at 1350 m to 1750 m; it was not recorded below 1350 m Balete and Heaney, 1997;Heaney et al., 1999). Laurels, myrtles, oaks, and podocarps were the common canopy and emergent trees in these forests, often reaching a canopy height of ca. 12-20 m in montane forest, and 5-12 m in mossy forest. Moss cover ranged from moderate in montane forest, to covering most surfaces of trees and ground in mossy forest; leaf litter in both habitats was extensive, and the humus layer was thick. Vegetation of these habitats was described in more detail in Heaney et al. (1999).
The morphological adaptations of A. luzonensis (e.g., small, robust body with short tail, short limbs, and small ears, short dense pelage, and long, sharp claws) suggest a semifossorial habit. It is active during the day, possibly extending from dawn to dusk; in 1988, seven of eight individuals captured were taken during the day and the eighth near dawn .
They were captured mostly along runways under root tangles, under fallen logs, or beside ground vegetation. In 1993-1994, two individuals were observed foraging in leaf litter in mossy forest during the day (Balete and Heaney, 1997). Stomach contents, consisting of amphipods, larval and adult arthropods, and earthworms, suggest a vermivorous/insectivorous diet . Earthworms were abundant where this species was captured .

Females have two pairs of inguinal mammae. Two pregnant females, recorded in late March
and late April, each had a single embryo . The species appeared to be moderately common in mossy forest on Mt. Isarog, with an estimated density of ca. 4.5 + 1.63 individuals/ha (Balete and Heaney, 1997).
Other nonvolant small mammals recorded along with Archboldomys luzonensis in montane and mossy forest included Crocidura grayi, Apomys microdon, A. musculus, Batomys cf. grand, Chrotomys gonzalesi, Phloeomys cumingi, Rattus everetti, and Rhynchomys isarogensis (Balete and Heaney, 1997;Heaney et al., 1999;Rickart et al., 1991). Of these, five are members of the Chrotomys Division: A. microdon, A. musculus, C. gonzalesi, and R. isarogensis (table 12).  1885 m (193531 [holotype], 193944). NO. 3754 Etymology: From the superlative of magnus (Latin: "great, large"), to highlight its being the largest of the seven species of shrew mice (Archboldomys, Crunomys, and Soricomys) on Luzon. We recommend as the English common name either "large Cordillera shrew mouse" or "Cordillera archboldomys" Diagnosis: Archboldomys maximus is a shrew mouse similar to A. luzonensis, but readily defined by the following combination of external, cranial, and dental characters and propor¬ tions (tables 1, 2, figs. 9-14): (1) long, robust body; (2) dark chestnut pelage; (3) tail nearly as long as head and body, with fine scale rows; (4) relatively long hind feet; (5) long, broadly ovate skull with tapered rostrum and more strongly inflated cranium; (6)  The pelage of A. maximus is uniformly grayish chestnut, longer and thicker dorsally than ventrally. The rostrum is tapered and the face is broad (fig. 12A). The pale grayish-brown vibrissae extend slightly beyond the ears. The lips and rhinarium are pigmented pale grayish brown. The eyelids are finely edged with dark gray, surrounded by a narrow paler band covered with short dark grayish-brown fur. The ears are small, round, medium grayish brown, and covered with short black hairs.

Archboldomys maximus, new species
The front feet of A. maximus are small, with relatively short, robust digits bearing long, opaque claws with decurved tips, except the pollex, which is short and bears a nail. The dorsal and palmar surfaces are uniformly pigmented dark grayish brown. The palmar pads consist of three interdigitals, and a thenar and hypothenar, which are larger than the interdigitals. The hind feet are long and slender, with relatively short digits; the longest middle digits including claws are less than one-third the length of the hind foot. The claws are opaque, shorter than on the forefeet and with short decurved tips. Both the dorsal and plantar surfaces, including digits and plantar pads, are pigmented dark grayish brown. The plantar pads are small relative to plantar surface and consist of four interdigitals, a large thenar, and smaller, round hypothe¬ nar; the metatarsal is small and rounded; the rest are ovate and larger.
The tail of the holotype of A. maximus is 13% shorter than the combined length of head and body; the paratypes have from 20% shorter to slightly longer tail than the combined length of head and body (table 1). There are ca. 20 tail scale rows/cm (TSR) in the holotype, 20-24 TSR in paratypes; each scale bears three short hairs. Tail is uniformly pigmented dark grayish brown, with its dorsal surface covered with short, similarly pigmented hairs and its ventral surface covered with a mix of dark and pale to unpigmented hairs.
The skull of A. maximus is smooth, with gracile, tapered rostrum and dorsolaterally swol¬ len cranium (figs. 3,9,13). In lateral view, its dorsal profile is nearly straight from the top of the skull to about the anterior base of the premaxilla, from which the profile assumes a short and smoothly shallow anteriad convexity brought about by the upturned nasal tips. The nasal tip projects beyond the anterior edge of the premaxillae. The slightly raised bony capsule of the upper incisor root terminates medially near the suture with the maxilla, opposite the dorsal opening of the small and narrow lachrymal canal. The opening of this canal slants caudad, and its nearly flat outer wall barely projects beyond the lateral outline of the capsular swelling of the upper incisor root. The zygomatic plate slants dorsad, relative to the upper molar tooth row. The zygomatic process of the squamosal is low on the cranium, anchored less than a mil¬ limeter above the dorsal edge of the postglenoid foramen.
The tympanic hook of the squamosal is slender and short relative to cranial size. The squamoso-mastoid foramen is small relative to the postglenoid vacuity and is either obscured entirely by a thin membrane, as in the holotype, or partially obscured ventrally by the wide, posterior extension of the periotic part of the petromastoid. The mastoid of adults has no fenestra (e.g., FMNH 193526, 193943, 193944). In younger adults, a narrow slit in various stages of closure is present in the mastoid (as in the left mastoid of the holotype [the right one is completely ossified], FMNH 193531; fig. 13). A small mastoid foramen is situated along the medial occipital suture.
The incisive foramina of A. maximus are long and narrow, round edged posteriad and smoothly tapered anteriad ( fig. 13, table 2). A small interpremaxillary foramen is present, lying visibly posterior to a line connecting the posterior edges of the incisors ( fig. 9). The alisphenoid canal is small and circular, bridged by the thin alisphenoid strut, both well concealed under the wide pterygoid ridge. The pterygoid ridge itself is also long, extending posteriad and becoming the pterygoid bridge over the foramen ovale. The postglenoid vacuity is spacious and domed, and the widest part appears nearly round due to the smooth concavity of the posterior periotic part of the petromastoid opposite the domed dorsal edge; in the holotype, the left vacuity is sealed by a membranous covering. The auditory bulla is small relative to cranial area, and anteroventrally inflated (viewed right side up), obscuring the anterior half of the petrosal in ventral view.
Dental features are strikingly similar between A. maximus and A. luzonensis, both sharing slightly (A. luzonensis) or strongly (A. maximus) ophistodont upper incisor procumbency, upper incisors in ventral view that are sharply angled on the posterior edge ( fig. 9), and squarish first upper molars. But several dental and mandibular features and measurements easily distinguish A. maximus from A. luzonensis (figs. 2, 3, 9-14, table 2;Musser, 1982a: figs. 13, 32, 33; fig. 12), including longer molariform tooth row (shorter in A. luzonensis), larger molars with complete cuspidation on the first upper molar ( fig. 2; smaller, with absent or missing cusp t3 and cusp t9 on Ml), longer and more gracile mandible (shorter and more robust), narrow angular process (wider), and shorter and narrower coronoid process (longer and wider).
The maxillary molars of A. maximus (figs. 2, 14) are marked by their large size. The first upper molar appears squarish due to the conspicuous broadening of the lingual cusps tl and t4 that together contribute to the width of Ml, which is about three-fourths of its length. The cuspidation of the three rows of Ml appears complete ( fig. 2), though coalesced in the broad laminar outlines of the deeply basined occlusal surface, cusp t3 on the first row is barely visible at the edge of the worn, broad, and downsloping labial lamina, and cusps t7 and t9 have either merged with t8 or are absent. The second molars are smaller than the first, but the anterolingual convexity of cusps tl and t4 makes them wider than long, and trapezoidal in occlusal outline.
Cusps t2 and t3 are missing on the first row. The second and third rows appear to have com¬ plete cuspidation, coalesced in the laminar outlines of their occlusal surfaces. The third molar has less than half the surface area of the second molar, and has a cordate occlusal outline brought about by the broad lingual cusp tl and what appears to be the anterolabially oriented second row, with presumed cusps t4, t5, and t6 coalesced into the narrow laminar outline of its occlusal surface. The slanting orientation of this second row results in posteriad placement of cusp t4 (figs. 2, 14). No evidence of a third row is discernible.
The dentary of A. maximus is slightly longer and more gracile than in A. luzonensis (figs. 10, 13;Musser, 1982a: fig. 13). The labial surface of the mandible is smooth, and the capsular process is barely discernible. The placement of the mandibular and mental foramina is similar to A. luzonensis. The mandibular molar row ( fig. 14B) is slender. The lower incisors are robust, sharply pointed, and the anterior surface covered with pale yellow enamel ( fig. 13). The incisor is broadly procumbent, forming a smooth, wide curve from the anterior edge of the first man¬ dibular molar alveolus. The coronoid process is short, slender, and backswept, extending slightly more than halfway from its base to the base of the adjacent condyle. The condyloid process is short and robust. The angular process is slender with a narrowly angular and slightly upturned tip, forming a deep sigmoidal notch with the posterior edge of the condyle. Ecology: Archboldomys maximus has been captured in old-growth montane and mossy forest, from 1885 m to 2690 m . There were slightly more diurnal (8) than nocturnal/crepuscular (5) captures. Success with traps baited with live earthworms was higher than with fried coconut, but the effect was not statistically significant. Stomach analy¬ sis suggests a predominantly vermivorous feeding habit; stomach contents of six individuals all included earthworms, in addition to centipedes (order Geophilomorpha) in three indi¬ viduals, and a few fragments of insect exoskeletons (body and integument of adults and larvae) in four individuals.
Females have two pairs of mammae, both inguinal. Reproductive activity early in the year was apparent. Five of six males caught in March had scrotal testes, one of which had a testes size of 10 x 7 mm with convoluted epididimys; during the same period, two of seven females had large mammae, one of which was pregnant with a single embryo and had one placental scar (Balete, Heaney, and Rickart field notes in FMNH).

Additional species of native nonvolant mammals we documented on Mt. Amuyao included
Apomys abrae (at lower elevations), and the giant cloud rats Crateromys schadenbergi and Phloeomys pallidus .
Distribution: Luzon Island only; currently known only from the Central Cordillera, the Mingan Mountains, and the northern Sierra Madre ( fig. 1).
Comments: Comparisons to Archboldomys are given above. Soricomys kalinga  and S. musseri  were both origi¬ nally described as members of Archboldomys. Several cranial and dental characters highlighted during the original diagnosis of S. musseri, including straight nasal tips, short incisive foramina, small postglenoid vacuities, prominent mastoid fenestra, and smaller molars, are now among the diagnostic features of Soricomys (table 11). Also, as noted in the original description, the ventral pelage of S. musseri is slightly shorter and paler than the dorsal fur (with a gradual is similar to the divergence between the two species of Archboldomys (9.1%); within the range of divergence values among species of Chrotomys (2.8% to 11.7%); and much higher than the diver¬ gence values observed among recognized species of Rhynchomys (1.4% to 3.1%; fig. 7).
A PCA analysis of cranial and dental measurements ( fig. 4, table 5) that included all species of Archboldomys and Soricomys ( fig. 4, table 5) demonstrated clear separation of S. musseri from Soricomys leonardocoi, n. sp., on the second axis, with S. kalinga, S. montanus, n. sp., and S. musseri having a proportionately longer skull, more elongate diastema and postpalatal region, and broader upper incisors near their tips, and S. leonardocoi, n. sp., having the converse. Soricomys musseri differed from S. kalinga slightly on the first axis, an indication of its overall larger size. Soricomys musseri was most similar to S. montanus, n. sp., differing in scoring slightly higher on the first axis, indicating greater size. A PCA that included only the four species of Soricomys ( fig. 5, table 6) produced similar results, with S. musseri loading less heavily on the first component and more heavily on the second component than S. leonardocoi, n. sp., indicating that S. musseri averages smaller overall and has shorter basioccipital length, rostral depth, rostral length, diastema length, and breadth of incisors near their tips. Soricomys musseri differed from S. kalinga and S. montanus, n. sp., in this PCA on the first component, which generally indicates its larger size. Ecology: Soricomys musseri has been documented in montane and mossy forest on Mt. Ceta¬ ceo, from 1500 m to 1650 m Duya et al., 2007Duya et al., ,2011). Based on trapping success, S. musseri appeared to be uncommon, with only four captures in 2395 ground-trap nights; the majority (75%, N = 3) were caught with earthworm bait, possibly indicating a vermivorous diet . Males with scrotal testes were captured in May and June; the two females, cap¬ tured in June, were young adults with two pairs of small inguinal mammae. Also recorded in the montane and mossy forest habitat of S. musseri were two other members of the Chrotomys Division: Apomys musculus and A. sierrae; another member of the clade, A. microdon, was documented slightly lower on the mountain, at 1400 m (table 12). Other nonvolant small mammals recorded on the mountain included Crocidura grayi, Bullimus luzonicus, and Rattus everetti .
Comments: The attribution of the type locality in the original description to Callao Munic¬ ipality was erroneous, as Callao is actually a barangay (a smaller political unit within a munici¬ pality), instead of Penablanca, which is the correct name of the municipality. The first specimen, which later became the holotype of this species, was initially reported as Crunomys fallax (Danielsen et al., 1994).
Karyology: 2N = 44; FN unknown ( fig. 6B; . Comparisons: Distinguished from S. musseri, n. sp., by the following features (tables 3, 4; figs. 9-11; Rickart et al., 1998: figs. 10, 12, 13;Balete et al., 2006: figs. 2-6): (1)  Bali-it in Balbalan Municipality, Kalinga Province (figs. 1, 15). Inspection of figure 15 leads us to suggest that this species occurs north of the drainages of the Agno and Chico rivers (including the Sabangan river, a tributary of the Chico); these two drainages form a decliv¬ ity across the crest of the Central Cordillera, with their headwaters meeting at a narrow pass that reaches a maximum elevation of about 1800 m, a short distance to the northwest of Mt. Data. On this basis, we hypothesize that this species will be found to occur from approximately Mt. Tinangdanan (ca. 5 km SW of Sagada) along the chain of mountains stretching NNE to at least as far as Mt. Macopa, which lies along the boundary of Abra and Apayao provinces. The type locality lies slightly north of the center of this hypothesized distribution ( fig. 15).
Females have two pairs of inguinal mammae. Males with scrotal testes were captured in February, and pregnancy was recorded in April: two pregnant females each bore two embryos, and one also had one placental scar .  The incisive foramina of S. montanus are small, round-edged posteriad and smoothly tapered anteriad; nearly half as long as the diastema ( fig. 16, table 4). The alisphenoid canal is small and narrow, partly hidden under the pterygoid ridge (as seen from ventral view), but otherwise fully exposed in the holotype, due to the missing alisphenoid strut reminiscent of the Crunomys pat¬ tern, though only partially evident in C. suncoides (Musser, 1982a: fig. 23;Rickart et al., 1998: figs. 4, 7); in all the paratypes, however, the alisphenoid canal is positioned under the posterior edge of the thin pterygoid ridge, behind the short and narrow alisphenoid strut.
Because the pterygoid strut is absent in the holotype, the pterygoid ridge merges posteriad with the pterygoid bridge that extends over the foramen ovale to the anterior edge of the elon¬ gate middle lacerate foramen ( fig. 16); in all the paratypes, as in the congeners, the pterygoid ridge terminates posteriad at its juncture with the alisphenoid strut ( fig. 16; Rickart et al., 1998: fig. 12). The postglenoid vacuity is spacious and appears broadly triangular due to a blunt (and slightly arched) angle formed by the dorsal confluence of the backward-slanted anterior edge of the tympanic hook and forward-slanted posterior edge of the squamosal, with the straight ventral edge of the periotic part of the petromastoid as its base ( fig. 16). This shape of the postglenoid vacuity in S. montanus approximates that of S. kalinga, but in the latter the shorter tympanic hook results in a pronounced anteriad skew (Balete et al., 2006: figs. 4-5); in contrast, it is smoothly arched in S. musseri (Rickart et al., 1998: figs. 6, 13;Balete et al., 2006: figs. 4-5) and somewhat rectangular in S. leonardocoi, n. sp. (fig. 17). The auditory bulla is relatively small and ventrolaterally inflated, gradually tapering anteriad and smoothly merging with the pointed eustachian tube (fig. 16); in all congeners, this anterior bullar inflation is conspicuous and forms a pronounced constriction that clearly demarcates the boundary with the narrow eusta¬ chian tube (figs. 10, 17;Rickart et al., 1998: figs. 12-13;Balete et al., 2006: fig. 4).
The skull of S. montanus closely resembles those of its congeners, but is most similar to S. kalinga. However, it can be distinguished from S. kalinga by the following combination of traits: (1) longer skull with slightly dorsolaterally inflated cranium (shorter and dorsolaterally flattened in S. kalinga); (2) zygomatic breadth and zygomatic plate narrower (wider); (3) (FMNH 190972, holotype) in dorsal, ventral, and lateral views. laminar outline on their third row, which is largely comprised of the coalesced cusps t7 and t8. M3 is relatively large in relation to Ml and M2, with the broadly convex cusp tl account¬ ing for nearly half of its occlusal surface, and the rest formed by the coalesced cusps of the second row; the presence of the third row is not evident.
Soricomys montanus is distinguished from S. musseri by: (1) shorter maxillary molar tooth row (longer in S. musseri); (2) narrower Ml (broader); and (3)  Soricomys montanus has mandibles that closely resemble those of its congeners in general shape and placement of mental and mandibular foramina. They differ mainly in relative size and shape of the processes (fig. 16, table 4;Balete et al., 2006: fig. 4;Rickart et al., 1998: fig. 13). The coronoid process is relatively long, robust, and broadly backswept to about two-thirds the length of the condyloid process. In contrast, the coronoid process of S. kalinga is shorter and slender, though equally broadly backswept; it is longer, more delicate, and narrowly backswept in S. mus¬ seri; in S. leonardocoi, n. sp., it is about as long but more slender. The condyloid process of S. montanus is more robust and its angular process relatively shorter than in congeners.  (Heaney et al., 2006, unpubl. specimens and field notes at FMNH; Rickart et al., 2011b). It showed a high level of tolerance to forest disturbance on Mts. Amuyao, Data, and Pulag, but was absent in heavily disturbed habitats dominated by pine forest or by agriculture at the elevations where they occurred . We found S. montanus to be predominantly diurnal and to show a preference for earthworm bait (Balete, Heaney, and Rickart . 7).
Soricomys leonardocoi, new species Description and Comparisons: The pelage of S. leonardocoi is dark grayish chestnut dorsally, pale grayish brown and shorter ventrally, without distinct patterning. The lips and rhinarium are medium gray ( fig. 12C). Mystacial vibrissae are dark gray with pale tips that extend beyond the ears. The eyelids are dark gray surrounded by a narrow paler band covered with short, dark, grayish-brown fur. The dark gray ears are small, rounded, and covered with short, dark hairs. The front feet of S. leonardocoi are small, with long, slender digits bearing long, opaque claws with decurved tips, except the pollex, which is short and bears a nail. The dorsal surface is pigmented medium grayish brown, and the palmar surface and palmar pads, consisting of three small inter¬ digitals and two larger pads (thenar and hypothenar), are paler and uniformly pigmented. The hind feet are long and slender, with long, slender digits bearing long, opaque claws. They are dorsally pigmented medium grayish brown, darker on the digits, and covered with grayish-brown fur; dark gray to pale grayish-brown ungual tufts extend to about three-quarters of the length of claws. The plantar surface is uniformly pigmented dark grayish brown; six plantar pads, consisting of four interdigitals, a thenar, and a hypothenar, are small relative to the plantar surface, and uni¬ formly pigmented the same darkness as the plantar surface except slightly paler at the distal tips.
The skull of S. leonardocoi, as in congeners, is smooth and the braincase is longer than wide, appearing rectangular in general outline. The rostrum is short and tapered. Laterally, the dorsal profile is nearly straight from the top of the skull to the tip of the rostrum. The nasals of S. leon¬ ardocoi are straight, and terminate near the level of the anterior edges of the premaxillae. The upper incisor root is enclosed in a slightly raised bony capsule within the premaxilla and termi¬ nates medially near the suture with the maxilla, opposite the dorsal opening of the small and narrow lachrymal canal. The opening of this canal slants caudad and its swollen outer wall con¬ spicuously broadens the base of the rostrum. The zygomatic plate has a nearly straight, vertical anterior edge relative to the upper molar row. The zygomatic process of the squamosal is anchored posteriorly about 1 mm above the dorsal edge of the postglenoid foramen and attached higher laterally on the cranium. The tympanic hook of the squamosal is broadly tapered ventrad and slanted caudad. This broadening of the tympanic hook dorsally is coupled by the intrusion mediolaterally of a small bony mastoidal spur into the squamoso-mastoid foramen, resulting in a very narrow opening ( fig. 17). A relatively large mastoid fenestra is near the posterior margin of the mastoid, opposite the small mastoid foramen along the occipital suture ( fig. 17).
In ventral aspect of the skull, the incisive foramina of S. leonardocoi are small, round edged posteriad and smoothly tapered anteriad; they are slightly more than half the length of the diastema ( fig. 17, table 4). The alisphenoid canal is small and narrow. It is positioned under the posterior edge of the thin pterygoid ridge, behind the short and narrow alisphenoid strut. The pterygoid ridge itself terminates posteriad at the juncture with the alisphenoid strut, opposite which a relatively wide pterygoid bridge extends over the foramen ovale to the anterior edge of the elongate middle lacerate foramen. The postglenoid vacuity is relatively spacious and appears rectangular due to its straight dorsal edge (domed in congeners), and an almost straight ventral edge formed by the periotic part of the petromastoid (fig. 17). The auditory bulla is relatively small and ventrolaterally inflated, obscuring as much as the anterior half of the petrosal in ventral view. Soricomys kalinga and S. montanus have a similar pattern of bullar inflation; it is less pronounced in S. musseri.
The incisors of Soricomys leonardocoi are small and have smooth anterior surfaces. The upper incisors emerge from the rostrum at a right angle (orthodont) and have yellowish-orange enamel.
The ventral tips are nearly straight edged. In ventral view, the outline of each incisor is U-shaped deflected laterally, forming a deep, V-shaped gap along their posteromedial margin ( fig. 9). The lower incisors are long and thinly pointed and have pale yellow enamel. The upper molars are small and narrow, but with low and sloping labial cusps that do not have corresponding convexity.
The first upper molar is only slightly more than half as wide as it is long, and appears roughly rectangular in outline (especially in individuals with worn molars). The convex projection of cusps tl and t4 forms a distinct, shallowly bilobed and broadly tapered lingual outline on Ml.
Cusp t3 appears to be present in the shallow and sloping labial edge of the coalesced laminar outline of the first row of Ml, but is difficult to detect in most specimens with worn molars; t3 is undetectable on M2. Cusp t9 in both Ml and M2 is barely detectable on the broad laminar outline on the third row. M3 is relatively large in relation to Ml and M2, with the broadly convex cusp tl accounting for about a third of the occlusal surface of this molar.
The dentition of S. leonardocoi is closely similar to congeners, differing mainly in relative sizes and cuspidation pattern (figs. 9, 14, 17, table 4;Balete et al., 2006: fig. 6;Rickart et al., 1998: fig. 8). It is readily distinguishable from S. musseri by the following combination of dental features: (1) longer maxillary molar tooth row (shorter in S. musseri); (2) broader Ml (nar¬ rower); and (3) narrower incisors at their tips (broader), and from S. kalinga and S. montanus through targeted surveys of isolated mountains and mountain chains, using multiple sampling procedures to ensure detection of species that have varied food and foraging habits in different habitats along the elevational gradient (e.g., Balete et al., 2009;Rickart et al., , 2011b. In the case of Archboldomys and Soricomys, the use of live earthworms as bait, the targeting of montane and mossy forest at high elevations, and surveys of even small mountain chains, have been crucial in detecting the animals.

Phylogenentic Relationships
Our analyses of DNA sequence data from the mitochondrial cytochrome b ( fig. 7) and the nuclear IRBP ( fig. 8) genes provide robust evidence that the Chrotomys Division is a monophyletic unit, as first documented by Jansa et al. (2006). As discussed above, our earlier studies of morphology led us to view Archboldomys as a monophyletic unit containing two "species groups," but the genetic data presented here provide unambiguous evidence that these two "species groups" are not sister taxa, but rather are distantly related members of the Chrotomys Division. Many of the morphological similarities that we previously interpreted as homologies (e.g., dark reddish-brown pelage both dorsally and ventrally, tail of moderate length, reduced dentition) may represent convergent characters, perhaps associated with the largely diurnal habit of foraging for earthworms, insects, and other invertebrates in leaf litter.
The extent of ecological differences that allow syntopy of A. maximus with S. kalinga on Mt.
Amuyao (and possibly more widely within the Central Cordillera) may now be seen as espe¬ cially important in understanding the evolution of diversity within the Chrotomys Division, and deserve further investigation. Moreover, the sister-group relationship of Soricomys with Chrotomys now becomes one of special interest, with the implication that the burrowing habit of Chrotomys is a feature derived since divergence from the ancestor shared with the more generalized Soricomys. Detailed studies of the anatomy and character evolution in the Chrotomys Division, including changes in limbs and feet, might provide insights on the fac¬ tors facilitating coexistence.
The karyotype of Soricomys montanus is very similar to those of several members of the Chrotomys Division, including species in the genera Chrotomys, Rhynchomys, and the Apomys subgenus Megapomys that have karyotypes with 2N = 44 and FN = 52 to 58 with a majority of telocentric chromosomes (table 7; Rickart and Musser, 1993;. The 2N = 44 karyotype of S. kalinga, while imperfectly known, also is consistent with this group. The occurrence of this widespread 2N = 44 karyotype in taxa positioned throughout the clade strongly suggests that it represents a primitive arrangement for the Chrot¬ omys Division. In contrast, the karyotype of Archboldomys luzonensis is unique among Philip¬ pine murids examined to date, both in basic arrangement of the autosomes and in the apparently unique sex chromosome system (Rickart and Musser, 1993). The position of A. luzonensis within the Chrotomys Division ( fig. 7; see also Jansa et al., 2006: fig. 1) indicates that this karyo¬ type is a distinctive, highly derived condition. Unfortunately, the karyotypes of A. maximus, S. leonardocoi, and S. musseri are unknown, so that the utility of the unique features of the A. luzonensis karyotype in defining and contrasting the two genera remain unknown. m ( fig. 15). With the exception of the division between northern and southern Central Cordillera, the other members of the Chrotomys Division (Apomys, Chrotomys, and Rhynchomys) exhibit the same pattern of distribution, and further illustrate how isolation of mountain habitats provides another level of insularization that contribute to the observed richness of mammalian diversity in this isolated oceanic island Heaney, 2000;Rickart et al., 1998Rickart et al., , 2005Musser and Freeman, 1981;Steppan et al., 2003).
(7) The community composition of the nonvolant small mammals on each mountain or mountain range further illustrates the breadth and range of adaptive radiation in the Chrotomys Division (table 12; Balete et al., 2006Balete et al., , 2007Duya et al., 2011;Heaney et al., 1999Heaney et al., , 2005Jansa et al., 2006;Rickart et al., 1991Rickart et al., , 2011a. We have recorded six members of this clade occur¬ ring at a single locality on Mt. Bali-it , and seven at a locality on Mt. Amuyao Heaney et al., unpubl. data and specimens at FMNH). We currently have evidence of species of Archboldomys and Soricomys occurring sympatrically only in the Central Cordillera, on Mt. Amuyao, where both species were predominantly diurnal, were captured in leaf litter on the ground surface, showed preference for earthworm bait, and had remains of invertebrates in their stomachs (see species accounts above). We note that adult A. maximus average about 45 g and S. montanus about 27 g, which may indicate some difference in trophic level or foraging activities. Clearly, communities of murid rodents on Luzon are often comprised of many closely related species. All members of the Chrotomys Division are thought to be descended from a species that reached the Philippines ca. 12-15 Mya (Jansa et al., 2006).

Conservation
Soricomys montanus is widespread in the southern portion of the Central Cordillera, where it maintains populations in both old-growth and secondary forest, and even occurs in small patches of highly degraded forest on Mt. Data . Archboldomys maximus and S. leonardocoi are members of a rich assemblage of endemic mammals associated with montane and mossy forest on Mt. Amuyao and Mt. Mingan, respectively. These new species appeared to be in moderate but stable populations in their respective habitats, as with the other members of the small mammal community there. Their habitats have not been subjected to the extensive logging that has drastically reduced the lowland dipterocarp forest of much of Luzon and the rest of the Philippines (Environmental Studies for Social Change, 1999). On both Mt. Amuyao and Mt. Mingan, high-elevation habitats are still in good condition. How¬ ever, we noted that on Mt. Amuyao, construction of telecommunication facilities on the peak, with accompanying forest clearing along the route from the town center to the peak, have increased forest loss in the area. Additional negative impacts, including recent forest fires from the increasing number of mountain climbers and hikers, were also evident along the trails. Mt.
Mingan, on the other hand, had much lesser signs of disturbance away from populated areas.
On both mountains, and over large portions of the Central Cordillera, the indigenous tribes continue to manage the forests following their traditional customs and are responsible for keeping the forest largely protected. Mt. Bali-it and the adjacent mountains in Kalinga are managed through traditional practices by the Banao tribe . In contrast, for instance, Mt. Data National Park, which is managed by the national government, has lost much of the high elevation forest to commercial vegetable farming . During our fieldwork there in 2006 we documented S. montanus along the grassy edge of a trail within the remnant mossy forest. However, we failed to find Chrotomys silaceus and Rhynchomys soricoides, first discovered there by John Whitehead in 1895, or Abditomys latidens, found on Mt. Data in 1946(Musser, 1982bThomas, 1895Thomas, , 1898Sanborn, 1952). At least four more species docu-mented on Mt. Data appeared to have become locally extinct by 1946: Carpomys melanurus, C. phaeurus, Crateromys schadenbergi, and Phloeomys pallidus (Thomas, 1898;Sanborn, 1952 Rickart et al., 2011b).
The beneficial effects of traditional management of the forest resources by the indigenous peoples are an important factor in the conservation of the forest and its rich biodiversity in Balbalasang, Kalinga Province . We caution against the marginalization of sustainable traditional practices of the indigenous peoples in the process of formalizing and centralizing management decisions in protected areas, as demonstrated on Mt. Data, which, ironically, is a national park. Management interventions aimed at encouraging and further strengthening the traditional forest resource use practices by indigenous peoples on Mt.
Amuyao and Mt. Mingan may be an important contribution to the conservation of the unique mammalian diversity in these areas, as an integral part of the creation and/or management of protected areas to ensure the long-term sustainability of biodiversity protection and conserva¬ tion of these mountains Rickart et al., 2011b).

ACKNOWLEDGMENTS
We offer our sincere thanks to everyone who offered support and assistance during this project, in the DENR and NCIP, and at the provincial, municipal, and barangay levels. We acknowledge the hard work and expert assistance of N. Antoque and J.B. Sarmiento, without whose help this project would not have been successfully completed. We also thank other members of our team who assisted us in our various activities in the field, including R.A.N. and figures were prepared by V. Simeonovski, L. Kanellos, and A. Niedzielski. A. Niedzielski also helped with preparation of the manuscript. Our use of the scanning electron microscope was skillfully facilitated by B. Strack. We thank J.A. Esselstyn, G.G. Musser, and R.S. Voss for their many helpful comments and suggestions for improving this manuscript.
We thank all of the good people of barangays Macalana, Fiantin, Latang, and Gawana, and the Dumagat tribe in Gabaldon, headed by the late L. Padua, who accompanied and helped us as porters, guides, or cooks, and who tolerated our seemingly odd activities in town and on the mountains. In Barlig, Jeana Away provided unending help in organizing our project, as well as arranging housing and working space for us, and the staff of the Seaworld Inn and Restau¬ rant, especially Manang Lourdes, showed endless patience and hospitality during our stay.
The Mingan Mountain project was funded in part by the Regional Natural Heritage Pro¬ gram of the Australian Government to the Conservation International-Philippines. Our special thanks to the Negaunee Foundation, and the Barbara Brown, Ellen Thorne Smith, and Marshall Field Funds of the Field Museum for their long-term support of our research, including field¬ work on Mt. Amuyao and Mt. Mingan, and for collection-based research at FMNH, without which our research and conservation activities in the Philippines would not be possible.