Holotype: MZB 41456 (= LSUMZ 36945), an adult male collected 8 March 2013 by J.L. Patton. The skin, skull, skeleton, and a tissue are held at MZB, while an additional frozen tissue is curated at LSUMZ. External measurements and weight recorded from the holotype are 182 mm × 98 mm × 19 mm × 9 mm = 12 g.

Type locality: Indonesia, Sulawesi Tengah, Toli Toli, Malangga Selatan, Mt. Dako; 1.10642° N, 120.9106° E, 512 m elevation.

Geographic distribution: Widespread on Sulawesi across most areas of endemism, with records from the west-central (Mts. Rorekatimbo and Torompupu, Central Sulawesi Province; Mt. Gandang Dewata, West Sulawesi Province; Mt. Latimojong, South Sulawesi Province), northwest (Mt. Dako, Central Sulawesi Province; Mt. Buliohuto, Gorontalo Province), north-east (Mt. Ambang, North Sulawesi Province), south-west (Mt. Bawakaraeng, South Sulawesi Province), east-central (Mt. Katopasa, Central Sulawesi Province), and south-east (Mt. Mekongga, Southeast Sulawesi Province) areas of endemism. The lack of records from the north-central area of endemism is almost certainly due to the lack of collecting effort in this region. Occurs over a wide elevational range from approximately 500 m on Mt. Buliohuto to at least 2600 m on Mt. Gandang Dewata (fig. 13; table 3).

Abbreviated diagnosis: Esselstyn et al. (2019) made detailed comparisons to other Sulawesi species found on Mt. Dako, of the north-west area of endemism. Briefly, Crocidura caudipilosa is a slender grayish to brownish shrew with a hairy tail (fig. 14A) that is moderately longer than the head and body (table 2). The feet are brown in dorsal view, with paler digits. The skull is gracile, with rounded features and a weakly developed dentition (fig. 15). Summary statistics for measurements of material we refer to this species are provided in tables 2 and 5.

Comparisons: This species is readily distinguished from all other Crocidura on Sulawesi by the density (combined number and length) of applied hairs on the tail (short hairs that lie close to the tail, not the longer bristles that typically project away from the tail on many species of Crocidura), but it is also distinguishable by various combinations of its body size, tail length relative to head-and-body length (fig. 9), gray to brown pelage, slight bicoloration (darker dorsum, paler venter), slender skull, and relatively small teeth. Its tail is shorter relative to head- and-body length than in any of the Elongata Subgroup members, but longer (absolutely and relatively) than in any other congeneric species on Sulawesi (fig. 9). Body size is considerably smaller than in C. elongata and C. quasielongata, but only slightly smaller than in C. microelongata (fig. 9). The skull and hind feet are not particularly elongate, as they are in members of the Elongata Subgroup. The ratio of braincase breadth to interorbital width is greater, on average, than in any other shrew species on Sulawesi, and considerably so compared to any of the Elongata Subgroup members (fig. 10).

Comments: Individuals of Crocidura caudipilosa have been caught (NMV Z56802) and observed (NMV Z62413) climbing trees (Esselstyn et al., 2019). The extent of time this species spends in trees is unknown, but the possibility that some shrews on Sulawesi exploit arboreal resources is a potentially promising explanation for how so many species coexist on the island. Musser (1982: 81) reported collecting an undescribed species of Crocidura “in moss growing 8 feet above ground around a tree trunk.” However, he never described this species, and it is unclear whether he was referencing a member of the Long-Tailed Group or some other species.

Crocidura caudipilosa is unusually widespread across Sulawesi, but Jukes-Cantor distances calculated from mtDNA sequences were <0.024 (fig. 4; supplementary data S3), suggesting recent, widespread movement of at least maternally inherited markers. Nuclear loci sampled from across the island show the same pattern, with variation detected at only five of 549 nucleotides among 56 sequences of apolipoprotein b, for instance.

This species is sister to Crocidura levicula in our UCE- (figs. 7, 8) and nuclear exon-based phylogenies, although statistical support from the latter estimate was absent (supplementary data S6). Our mitochondrial estimates placed C. caudipilosa in an unresolved clade of several small, darkly colored species of shrew, among which C. levicula is included (figs. 4, 5). This clade, which was well supported by our UCE analyses, contained no other member species of the Long-Tailed Group (figs. 7, 8).

Specimens examined: Mt. Ambang (LSUMZ 39243–39247), Mt. Bawakaraeng (MVZ 237625–237627; NMV Z57152, Z57013, Z57043, Z57044), Mt. Buliohuto (LSUMZ 38288–38291, 38296, 38588, 38656; NMV C37746, C37815), Mt. Dako (LSUMZ 36940, 36942, 36945; NMV C37267, C37304, C37305, C37330, C37351), Mt Gandang Dewata (FMNH 218759–218762, 218768, 218983–218986; MZB 34746, 34747, 34951, 34952), Mt. Katopasa (LSUMZ 39355–39357; MZB 39840, 39871, 39922; NMV C40186, C40219, C40220, Z61815, Z62413, Z62415), Mt. Latimojong (FMNH 213020–213028, 213427; MVZ 237612–237615; MZB 41649; NMV C38591), Mt. Mekongga (MWFB 8158, 13512), Mt. Rorekatimbo (FMNH 213246–213248, 213441), Mt. Torompupu (LSUMZ 39358, 39359).

Holotype: USNM 217534, an adult male obtained by H.C. Raven on 1 August 1916. The specimen comprises a skin and skull. External measurements recorded from the type are 214 mm × 120 mm × 22 mm; no ear length or weight was recorded.

Type locality: “Temboan (southwest from Tondano Lake), northeastern Celebes” (Miller and Hollister, 1921: 101; fig. 1). We estimate that the type locality is at 0.979° N, 124.605° E, 650 m elevation, which differs from other interpretations (e.g., Musser, 2014). See the gazetteer for a full explanation (appendix).

Geographic distribution: Apparently restricted to the northern peninsula of Sulawesi, where clear records are known from the northeast (the type locality, Temboan and Mt. Ambang, North Sulawesi Province) and north-west (Mt. Dako, Central Sulawesi Province, and Mt. Buliohuto, Gorontalo Province) areas of endemism (fig. 16). The absence of records from the north-central area of endemism are almost certainly due to the general lack of mammal collecting from this region. Miller and Hollister's (1921) paratypes from “Pinedapa, eastern Middle Celebes” and the two specimens from Mt. Rorekatimbo reported by Ruedi (1995) almost certainly represent Crocidura microelongata (see below). Crocidura elongata is known from a moderately broad range of elevations, with the low elevation records between 500 and 600 m from Mts. Buliohuto and Dako and high-elevation records reaching 1600 m on Mt. Dako (table 3). The type locality at Temboan is around 650 m elevation.

Emended diagnosis: A long-bodied, somewhat heavily built, moderately bicolored shrew with a very long tail and unusually long, pale hind feet (figs. 9, 14C, 17; tables 2, 5). The dorsal pelage is gray-brown overall; individual dorsal hairs are gray at the base and brown at the tip. The ventral pelage is paler, with individual hairs gray-based like those on the dorsum, except that the hair tips are silvery. The silvery appearance of the venter is most pronounced on the throat and chest area. In some specimens, however, this area has a reddish tint, due to variation in the color of the tips of some hairs. The mystacial vibrissae are dark proximally but white distally. The ears are prominent and pale. Dorsally, the feet are pale, transitioning from light brown near the wrist and ankle to nearly white on the digits. Ventrally, the feet show the same transition, but pigment is concentrated around the base of the pads (fig. 14C). The claws are translucent. The hind feet are long, even relative to head-and-body length (fig. 17). The long tail is subtly bicolored, with a brown dorsum and pale brown venter. Tiny applied hairs are present along the entire length of the tail, but they are barely visible to the naked eye (fig. 14C). However, these hairs are slightly longer and white near the tip of the tail, creating a very small white distal tuft. The long bristle hairs that are common at the base of the tail of many Crocidura are nearly absent in this species. The skull is long and slender, with a notably narrow braincase, interorbital region, and palate (figs. 10, 18A). Although the interorbital region is narrow relative to skull length, it is less constricted than the braincase. The long skull of C. elongata is attributable primarily to elongation of the postpalatal region; the rostral length is short relative to skull length (fig. 10). The lambdoidal ridge is prominent for a shrew of this size. The interorbital region is relatively straight (i.e., not strongly tapered) when viewed from the dorsal aspect. The dentition is prominent relative to the palatal breadth (fig. 18A).

Comparisons: Crocidura elongata is readily distinguished from most shrew species on Sulawesi by its long body and even longer tail, relatively pale color, and long and narrow hind feet and skull (tables 2, 5). Its head-and-body length is considerably longer than all species except C. rhoditis and C. quasielongata. Absolute hind-foot and ear lengths are on average greater in C. elongata than in any other shrew on Sulawesi (fig. 9; table 2). However, the other members of the Elongata Subgroup, first described below, can be difficult to distinguish (fig. 11; table 4). Compared to C. elongata, C. microelongata is smaller bodied, with a darker pelage and shorter average tail length. The thenar pad on the hind foot of C. elongata (fig. 14C) is considerably longer than in either C. microelongata (fig. 14D) or C. quasielongata (fig. 14B). The skull of C. microelongata is also shorter but nearly as wide at the braincase and interorbital region, hence its relative width (BB/CIL and IOW/CIL) is greater than that of C. elongata (figs. 10, 12). Compared to C. elongata, C. quasielongata has a similar head-and-body length, but, on average, a shorter tail and paler pelage (fig. 12, table 2). The skull of C. elongata is very similar in length to that of C. quasielongata (figs. 12, 18). However, in C. elongata the rostrum makes up a smaller proportion and the postpalatal region makes up a greater proportion of skull length than in C. quasielongata (figs. 10, 12). In addition, relative braincase breadth (BB/ CIL and BB/IOW) is, on average, slightly less in C. elongata than in C. quasielongata. A principal components analysis of 12 cranial measurements shows that these two species have broad, though not complete, overlap in multivariate morphometric space (fig. 11).

Comments: A nearly complete cytochrome b sequence (1109 bp) from a topotypical paratype (USNM 217535) is nearly identical to those from specimens we collected on Mt. Ambang (fig. 4), which is approximately 32 km to the southwest. Our inference based on mitogenomes also place USNM 217535 as a close relative to specimens from Mt. Ambang (fig. 5). These animals' mitochondrial sequences are also closely related to a series from Mt. Buliohuto and to specimens from Mt. Dako referred to as “dark elongata” by Esselstyn et al. (2019). Maximum Jukes-Cantor distances between these sample localities is 0.05 (fig. 4; supplementary data S5). Samples of this species from Temboan and Mts. Ambang, Buliohuto, and Dako formed a well-supported clade in our analyses of UCEs (figs. 7, 8) and concatenated nuclear exons (supplementary data S6).

The phylogeographic study of Eldridge et al. (2018), which examined the correspondence of genetic diversity in Elongata Subgroup shrews to the area of endemism paradigm expectations, conflated the three species of this subgroup with Crocidura elongata. More recently, specimens of C. elongata from Mt. Dako were referred to as “dark elongata” by Esselstyn et al. (2019) because the authors were unable to determine if either of two long-tailed species on Mt. Dako represented true C. elongata. Mt. Dako is the only area where we found C. elongata occurring in syntopy with another member of the Elongata Subgroup (C. quasielongata). Despite the confusing history of specimens in this subgroup, none of our analyses suggested a sister relationship between any two of these species. In our UCE species-tree inference, C. elongata was moderately supported as the sister to C. rhoditis and C. pseudorhoditis (fig. 7), but in our mitogenome estimate it was placed as sister to C. lea, although without statistical support (fig. 5).

Ruedi (1995) suggested a scansorial lifestyle for this species based on its long, naked-appearing tail and long hind feet. While this is certainly possible, direct evidence for a scansorial lifestyle is lacking, and these traits could be linked alternatively to a saltatorial locomotory style (Brosset, 1988). Some very limited evidence indicates that Crocidura caudipilosa, which has a long, but less extreme tail combined with a more typical hind-foot length, is a skilled climber (Esselstyn et al., 2019).

For coalescent species delimitation results, see the Crocidura quasielongata account below.

Specimens examined: Mt. Ambang (LSUMZ 39008–39013, 39015–39018, 39057, 39058, 39061, 39248–39251, 39257–39264, 39318; NMV C38009, C38032), Mt. Buliohuto (LSUMZ 38238, 38240, 38243–38247, 38251–38253; NMV C37742, C37752, C37760), Mt. Dako (LSUMZ 36905–36907, 36909, 36916, 36919, 36921, 36923, 36924, 36932; NMV C37248, C37249, C37303), Temboan (USNM 217534, 217535).

Holotype: MZB 43000 (= FMNH 213426), an adult male, collected on 1 March 2011 by J.A. Esselstyn. The specimen comprises a study skin, cleaned skull and skeleton, and frozen tissue samples. External measurements from the holotype are 206 mm × 111 mm × 20 mm × 10 mm = 12.5 g. The voucher specimen and a tissue sample will be permanently curated at MZB, with another tissue sample retained at FMNH.

Type locality: Indonesia, Sulawesi Selatan, Enrekang, Buntu Bato, Latimojong Village, Karangan, Mt. Latimojong, Bantanase; 3.40755° S, 120.0078° E, 2050 m.

Etymology: We combine “micro” with “elongata” because this species looks like a small version of C. elongata.

Geographic distribution: This species is broadly distributed across western portions of the west-central area of endemism of Sulawesi. We identified populations from Mt. Latimojong, South Sulawesi Province; Mt. Gandang Dewata, West Sulawesi Province; and Mts. Torompupu and Rorekatimbo, Central Sulawesi Province (fig. 16). Recorded from approximately 700 m on Mt. Latimojong to 2600 m on Mt. Gandang Dewata. Most specimens are from areas >1500 m (fig. 13; table 3).

Diagnosis: Crocidura microelongata is a somewhat large Crocidura with a long tail and long, slender hind feet and skull (tables 2, 5). The dorsal pelage is gray-brown overall, with individual hairs having a gray-brown base and brown tip (fig. 14D). The ventral pelage is more silver, comprising individual hairs with a dark gray base and silver tip. The mystacial vibrissae are dark proximally for a third of their length but white distally. The hind feet are long in absolute terms and relative to head-and-body length (figs. 9, 17). Dorsally, the feet are brown, abruptly transitioning to pinkish white near the base of the phalanges (in some specimens, it is a gradual transition). Ventrally, the hind feet are nearly white, but brown pigment is present around the lateral, posterior margin of the hind foot and around the base of the thenar and hypothenar pads (fig. 14D). In some specimens, the 1st and 4th interdigital pads are also pigmented around the base. The palmar surface is entirely pinkish white. The claws are translucent. The tail is subtly bicolored, with a brown dorsum and pale brown venter. Tiny applied hairs are present along the entire length of the tail, but they are barely visible to the naked eye along most of the length of the tail (fig. 14D). These hairs are slightly longer and white near the tip of the tail, creating a very small distal white tuft. In a minority of specimens, the integument is also white for the distal ≤20 mm of the tail. The tail bristles that are common at the base of the tail of many Crocidura are nearly absent in this species (fig. 14D). The skull is somewhat long and slender, with a tapering interorbital region, and moderately robust dentition (fig. 18B). The braincase is dorsoventrally inflated and bulges outward in the parietal region as compared with the more posterior inter-parietal. Relative to skull length the rostrum is quite short (fig. 10). The braincase is somewhat narrow relative to skull length, but the interorbital region is wide (fig. 10).

Comparisons: This species is readily distinguished by its ratio of tail length to head-and-body length (fig. 9; table 2) from all Crocidura species on Sulawesi except other members of the Long-Tailed Group. Within the Long-Tailed Group, C. caudipilosa is smaller and has a much hairier tail, shorter hind foot, paler pelage, and lower relative interorbital width (IOW/CIL) but greater relative braincase breadth (BB/CIL) (figs. 9, 10). Within the Elongata Subgroup, C. microelongata has a smaller body size, shorter tail, shorter and slightly darker hind feet, and shorter skull than C. elongata and C. quasielongata (fig. 12; table 5). The thenar pad on the hind foot is shorter than in C. elongata, but comparable to that of C. quasielongta (fig. 14). Although the skull is shorter, its breadth is similar to that of the other two species, whether measured at the braincase, interorbital area, or rostral region (figs. 10, 12). The interorbital region, however, is more tapered than in either of the two most similar species (fig. 18). The dentition of C. microelongata is slightly smaller in proportion to the skull than in either C. elongata or C. quasielongata (fig. 18).

Comments: Although we have not examined the specimens Ruedi (1995) identified as Crocidura elongata from Mt. Rorekatimbo (IZEA 4365 and 4396), a published cytochrome b sequence from IZEA 4396 is indistinguishable from mitochondrial sequences we obtained from Mt. Rorekatimbo samples of C. microelongata. Our mitochondrial gene trees placed C. microelongata as either sister to the other 19 species that make up Sulawesi's endemic radiation (all species except C. nigripes) or to all members of the endemic radiation except C. musseri. However, statistical support for these hypotheses was nonexistent (fig. 4) or modest (fig. 5). Our analyses of nuclear DNA placed C. microelongata as part of the basal comb with no clear sister relationship (figs. 7, 8; supplementary data S6). See the next account for results of nuclear DNA species delimitation analyses.

Specimens examined: Mt. Gandang Dewata (MZB 34736–34741, 34743–34745, 34748, 34749, 34751, 34753, 34755–34757, 38463, 38472, 38473; FMNH 218544–218547, 218584, 218593–218603, 218969, 218972; NMV Z21764), Mt. Latimojong (MZB 40935, 40937, 40938, 43000; FMNH 212990–213005; MVZ 237567, 237569–237572, 237594, 238121; NMV C38534), Mt. Rorekatimbo (FMNH 213146–213162, 213164–213173, 213435, 213436), Mt. Torompupu (MZB 43013, 43014; NMV C40139).

Holotype: MZB 43001 (= LSUMZ 36939), an adult male, collected on 15 March 2013 by J.L. Patton. The specimen was prepared as a study skin, cleaned skull and skeleton, and frozen tissues. External measurements from the holotype are 215 mm × 126 mm × 22 mm × 11 mm = 16 g. The voucher specimen and a tissue sample will be permanently curated at MZB, with another tissue sample retained at LSUMZ.

Type locality: Indonesia, Sulawesi Tengah, Toli Toli, Mt. Dako; 1.10998° N, 120.90339° E, 410 m.

Etymology: We combine “quasi” with “elongata” because this species resembles C. elongata.

Geographic distribution: Widespread on Sulawesi and recorded from all areas of endemism except the north-central and north-east (fig. 16). Populations identified here are from the west-central (Wasponda and Mts. Balease and Torompupu, Central Sulawesi Province; Salu Tiwo of Mt. Gandang Dewata, West Sulawesi Province; Mt. Latimojong, South Sulawesi Province), south-west (Mt. Bawakaraeng, South Sulawesi Province), south-east (Mt. Mekongga, Southeast Sulawesi Province), east-central (Mts. Katopasa and Tompotika, Central Sulawesi Province), and northwest (Mt. Dako, Central Sulawesi Province) areas of endemism (fig. 16). We found this species from approximately 200 m at Salu Tiwo (low elevation Mt. Gandang Dewata site) to 1700 m on Mts. Mekongga and Katopasa and 1800 m on Mt. Bawakaraeng (fig. 13; table 3).

Diagnosis: A long-bodied, somewhat heavily built, moderately bicolored shrew with a long tail and long, pale hind feet (table 2). The dorsal pelage ranges from gray-tan to gray-brown overall, with individual dorsal hairs having a gray base and tan to brown tip. Some specimens also have a narrow, tan band between the gray proximal section and brown tip of each hair. The ventral pelage has silver highlights, the visual effect of individual hairs each with a pale tip. In many specimens, the chest and throat area has a slight red tint due to variation in the color of the hair tips. The ears are large and pale. The mystacial vibrissae are dark proximally, but white distally, with the dark portion occupying 20%–80% of vibrissa length. The hind feet are long absolutely (table 2) and relative to head-and-body length (fig. 17). Dorsally, the feet range from entirely white, to brown at the ankle and wrist, slowly transitioning toward the white digits. Ventrally, the feet show the same transition (except in specimens where they are entirely white), but pigment is concentrated around the base of each foot pad. The claws are translucent (fig. 14B). The tail is subtly bicolored, with a brown dorsum and pale brown venter. Tiny applied hairs are present along the entire length of the tail, but they are barely visible to the naked eye along most of the tail. However, these hairs are slightly longer and white near the tip of the tail, creating a very small, sometimes white, distal tuft. Tail bristles are nearly absent (fig. 14B). The skull is long, primarily due to elongation of the postpalatal region, but not the rostral region (figs. 10, 18C). Relative to skull length, the braincase is especially narrow (BB/CIL) and the interorbital region somewhat narrow (IOW/CIL; fig. 10) and relatively untapered. The lambdoidal crest is prominent. The dentition is robust (fig. 18C).

Comparisons: This species is easily distinguished by the ratio of tail length to head-and-body length (most specimens >110%; fig. 9) from all Crocidura species on Sulawesi except other members of the Long-Tailed Group. Within this group, C. quasielongata is considerably larger, much stockier, has a much narrower relative braincase breadth (BB/CIL), and has a much less hairy tail than C. caudipilosa. Within the Elongata Subgroup, C. quasielongata is, on average, intermediate in size, tail length, and hind-foot length between the smaller C. microelongata and the slightly larger C. elongata (fig. 9; tables 2, 5). Crocidura quasielongata has long ears, second in length only to C. elongata (table 2). Although these measurements overlap between the species, the averages differ. On the hind foot, the thenar pad is shorter than in C. elongata (fig. 14), despite the similar hind-foot lengths shared by these species. Also, on average, C. quasielongata is paler than either C. elongata or C. microelongata, but there is substantial color variation across the range of the species. The pelage varies in overall color from tan to dark brown. The palest specimens of C. quasielongata are from Mt. Tompotika and Salu Tiwo of Mt. Gandang Dewata, while animals from Mt. Dako are slightly darker, and specimens from Mts. Balease, Katopasa, and Torompupu are darker still. These color differences may reflect a tendency for low-elevation animals to be paler than those sampled at higher elevations, perhaps an elevational version of Gloger's Rule (Gloger, 1833). The palest specimens have a middle color band that is light gray on individual hairs. The skull of C. quasielongata is long, with a very narrow braincase and interorbital region. The interorbital region is also rather straight, barely tapering anteriorly. In this regard, the skull is very similar to, but slightly shorter than that of C. elongata, and it is much more elongate than in C. microelongata (fig. 12). In proportion to skull length, rostral length is greater, but postpalatal length is lesser in C. quasielongata than in C. elongata and C. microelongata (fig. 10). The dentition is slightly more robust than in C. microelongata but comparable to that of C. elongata (fig. 18).

Comments: Although Crocidura microelongata is mostly separated in our PCA (fig. 11) and in univariate measures (fig. 12), C. elongata and C. quasielongata are much more difficult to distinguish, with average differences apparent only from large series of specimens identified with molecular data. Based on current sampling, geography can be a reliable predictor except that we found both species occurring on Mt. Dako and there remains a large sampling gap between Mt. Dako and the west-central area of endemism (i.e., the base of the northern peninsula; fig. 16). On Mt. Dako, we trapped C. elongata around 512 and 1600 m, whereas we captured C. quasielongata only around 410 m. We found C. quasielongata at higher elevations in other parts of the island and C. elongata at lower elevations at other localities (fig. 13; table 3). Thus, these two species may have a parapatric distribution partitioned by elevation on the one mountain where we found them together. Because these two species are so different morphologically from all the other species on the island, it would not be surprising if they fill similar functional niches and thus one excludes the other wherever they interact.

Despite the morphological similarities of Elongata Subgroup members, genetic evidence is clear that these three species are distinct from each other and do not form a clade. None of our phylogenetic analyses (mtDNA, nuclear exons, or UCEs) even hinted at a sister relationship between any of the three species. Our mitochondrial inferences placed Crocidura quasielongata as sister to C. caudicrassa (figs. 4, 5), but this seems unlikely because C. caudicrassa is sister to the phenotypically similar C. brevicauda in UCE trees (figs. 7, 8). In our UCE species tree, C. quasielongata is sister to a mix of Ordinary Group and Rhoditis Group species (fig. 7). Despite these mixed signals, the lack of basal resolution among species in our phylogenetic estimates leaves the door open to possible sister relationships among Elongata Subgroup members. The similarities in cranial proportions (e.g., relative widths; fig. 10) between C. elongata and C. quasielongata suggest either inherited similarity, or remarkable convergence. If any of these three species ever show a sister relationship in future analyses, we suspect it will be this pair.

Our BPP analyses compared the three members of the Elongata Subgroup, even though we did not infer them to form a clade in any of our phylogenetic analyses. The samples of each species were chosen to represent the full extent of our geographic sampling for the range of each species. Samples included comprise Crocidura elongata from three mountains (N = 17), C. microelongata from four mountains (N = 24), and C. quasielongata from eight mountains (N = 28). The matrix is 92% complete and all analyses supported all three species with 1.0 posterior probability.

Specimens examined: Mt. Balease (FMNH 210542–210559, 210596–210601), Mt. Bawakaraeng (NMV Z57377, Z57061), Mt. Dako (MZB 43001; LSUMZ 36935–36938), Mt. Katopasa (LSUMZ 39483, 39486–39490, 39492; NMV C40177, C40201, C40202, Z62311, Z61792), Mt. Latimojong (MZB 40936, MVZ 237573), Mt. Mekongga (MWFB 8091, 8094, 8113, 8114, 8122, 8130, 13507, 13509, 13511), Salu Tiwo (FMNH 218548–218557, 218559–218561, 218563–218579; MZB 38448, 38452), Mt. Tompotika (FMNH 213339–213342), Mt. Torompupu (LSUMZ 39418–39422, 39426–39430; MVZ 238107, 238108, NMV C40240, C40241, C40243–C40245, C40247, C40254, C40259–C40262, C40281, C40282, C40288, C40293), Wasponda (FMNH 210560–210562).

Holotype: USNM 217550, an adult male collected 3 August 1916 by H.C. Raven. Prepared as a skin and skull. External measurements from the type are 153 mm × 70 mm × 17 mm; ear length and weight were not recorded.

Type locality: “Temboan, northeastern Celebes” (Miller and Hollister, 1921: 102; fig. 20). Temboan is located at 0.979° N, 124.605° E, 650 m elevation in the Southeast Minahasa Regency, North Sulawesi Province, 6 km south of Kalait. See the gazetteer (appendix) for a detailed explanation of our interpretation of Raven's type locality.

Geographic distribution: Known only from the north-east area of endemism (Temboan and Mt. Ambang, North Sulawesi Province; fig. 20). Ruedi (1995) reported specimens from Mt. Rorekatimbo, but mitochondrially (Ruedi et al., 1998), these match the sister taxon, Crocidura pseudorhoditis. Musser (1987) reported C. rhoditis from southwestern and central Sulawesi, but those records likely represent other taxa. We recorded this species between 1400 and 1700 m on Mt. Ambang (fig. 13; table 3). Specimens from Temboan are from low elevation habitat, probably around 650 m.

Emended diagnosis: Crocidura rhoditis is one of the larger shrews on Sulawesi (tables 2, 7), with a stocky build (fig. 17). Its overall color is paler than most species on the island, with a medium gray dorsal pelage and slightly paler venter, often with cinnamon highlights. The feet are pale, especially the forefeet. The digits of the hand are nearly white or pale pink colored, but those of the hind feet are off-white. The tail is slightly shorter than head-and-body length, thick at the base, and similar in color to the pelage. Tail bristles are present, but sparse along the proximal half of the tail (fig. 21D). The lips are paler than the surrounding pelage and the rhinarium is also generally pale, but the upper margins of each nostril have a dark rim. The skull of C. rhoditis is large, with the rostral length making up an unusually large proportion of skull length (RL/CIL; fig. 10). The braincase is wider than expected for a shrew of this size, but the interorbital region is narrow relative to the braincase breadth (figs. 10, 22A). The lambdoidal ridge is fairly prominent and the squamosal-parietal suture also forms a ridge that extends horizontally from above the internal ear to above the glenoid fossa. This ridge is visible from a dorsal aspect and highlights the tapering shape of the anterior portion of the braincase margin. The angle at which the braincase narrows anteriorly is similar to the tapered shape of the interorbital region. The dentition is robust, occupying a large portion of the palate (fig. 22A).

Comparisons: Crocidura rhoditis is larger than all Sulawesi Crocidura except C. elongata, C. quasielongata, C. nigripes, and the two Thick-Tailed Group members, C. brevicauda and C. caudicrassa. Its mass-to-HBL ratio is greater than in all other species except C. caudicrassa (fig. 17) and its relative rostral length (RL/CIL) is greater than in all other species (fig. 10). The tail of C. rhoditis is much shorter absolutely and relative to body length than in members of the Long-Tailed Group, but longer absolutely than in all other species except C. pseudorhoditis (fig. 9). The dorsal pelage of C. rhoditis is similar in color to C. elongata, darker than C. quasielongata, and paler than C. nigripes. The feet of C. rhoditis are paler than those of most other species, with the exceptions of C. lea (much smaller), C. pallida (smaller), C. elongata (longer tail), and C. quasielongata (longer tail). Within the Rhoditis Group, C. rhoditis is easily confused with C. pseudorhoditis, its apparent sister species, but it is larger in all dimensions than the other two members of the Rhoditis Group, C. australis and C. pallida. Both C. rhoditis and C. pseudorhoditis have a pale integument, but C. rhoditis is larger, slightly paler, and has a narrower thenar pad on the hind foot (fig. 21). Although differences in body length are modest (fig. 23), C. rhoditis is stockier than C. pseudorhoditis (fig. 17) and size differences are apparent in condyloincisive length and braincase breadth (fig. 23; table 7). In addition to the greater relative rostral length mentioned above, C. rhoditis also has a lesser relative interorbital width (IOW/CIL) than all other members of the Rhoditis Group (fig. 10). The narrowness of the interorbital region is also reflected in that it is lesser relative to braincase breadth (IOW/BB) in C. rhoditis compared to C. pseudorhoditis, whereas braincase breadth relative to skull length (BB/CIL) does not differ appreciably between the two species (figs. 10, 22). The skull of C. rhoditis has a prominent ridge at the suture of the squamosal and parietal bones. This ridge is present, but less conspicuous in C. pseudorhoditis. A PCA of cranial measurements largely separates these two taxa along the first axis, which represents size (fig. 19; table 6). The differences in size measurements between C. rhoditis and its sister species are somewhat more pronounced when only syntopically sampled specimens from Mt. Ambang are considered (fig. 23), perhaps indicating a role for character displacement in these species.

Comments: Mitochondrial sequences from the type series (USNM 217550, 217552, and 217554, and FMNH 43858) are all very similar to cytochrome b sequences from our Mt. Ambang specimens, with maximum intraspecific Jukes-Cantor distances ≤0.011 (fig. 4; supplementary data S5). Parapatrically distributed, phenotypically similar specimens with distinct mitochondrial sequences (mean Jukes-Cantor distance = 0.08) and smaller skulls are described below as Crocidura pseudorhoditis. Although the morphological differences between these two species are slight, the fairly large mitochondrial divergence, corroborating inferences from nuclear exons and UCEs (figs. 7, 8; supplementary data S6), and their syntopy on Mt. Ambang (fig. 20) strongly support their recognition as distinct taxa. See the account of C. pseudorhoditis for results of BPP analyses.

Specimens examined: Mt. Ambang (LSUMZ 39032, 39033, 39037, 39038, 39050, 39053–39055, 39062, 39064, 39065, 39068, 39284, 39296, 39319, 39321; NMV C38022, C38031), Temboan (USNM 217550, 217552, 217554, FMNH 43858).

Holotype: MZB 43002 (= LSUMZ 39310), an adult female, collected on 22 February 2016 by H. Handika. The specimen was prepared as a cleaned skull and skeleton, study skin, stomach, intestine, and frozen tissues. External measurements from the type are 148 mm × 70 mm × 16 mm × 11 mm = 11 g. The voucher specimen and a tissue sample will be permanently curated at MZB, with another tissue sample retained at LSUMZ.

Type locality: Indonesia; Sulawesi Utara; Bolaang Mondgondow, Passi Timur; Insil; Mt. Ambang, near Lake Aliyah; 0.76385° N, 124.41188° E, 1481 m elevation.

Etymology: We combine “pseudo” with “rhoditis” because this species looks very similar to C. rhoditis.

Geographic distribution: We recorded this species from the northwestern portion of the west-central area of endemism (Mts. Torompupu and Rorekatimbo, Central Sulawesi Province), and the north-west and north-east areas of endemism of the northern peninsula (fig. 20; Mt. Dako, Central Sulawesi Province; Mt. Buliohuto, Gorontalo Province; and Mt. Ambang, North Sulawesi Province). The absence of records from the north-central area of endemism is almost certainly due to the lack of general mammal surveys and specimens from this region. Both specimens we report from Mt. Rorekatimbo were trapped by Ruedi (1995); we did not collect any Crocidura pseudorhoditis on Mt. Rorekatimbo, despite working at similar elevations. We found this species at a range of elevations (fig. 13), the low on Mt. Buliohuto (500 m) and the high on Mt. Rorekatimbo (2200 m; table 3).

Diagnosis: Crocidura pseudorhoditis is a moderately large shrew (tables 2, 7) with a somewhat stocky build, medium gray, moderately bicolored pelage, a tail that is shorter than head- and-body length, and dorsal foot surfaces that range from light gray-brown to pinkish white. The forefeet are paler than the hind feet and the claws are unpigmented and accompanied by a small tuft of white hairs on the hands, but on the hind feet, some of these hairs are dark. The tail is barely dorsoventrally bicolored proximally, but this transitions to a uniformly colored tail for the distal half of tail length. The tail has only sparse bristles over the proximal half of its length (fig. 21C). The lips are gray-brown, but the nose is distinctly paler. The palmar surface ranges from gray-brown to white, being palest on the digits. The color of the plantar surface also transitions from darker to paler colors toward the digits (fig. 21C), but all aspects are darker than on the forefeet. The skull is long (table 7) and broader at the braincase and interorbital region than expected given its length (figs. 10, 22B). The rostrum is long relative to the postpalatal portion of skull length, reflected in the RL/CIL (fig. 10). The suture between the squamosal and parietal bones is often open and in the shape of a sickle blade just below the opening of the sinus canal (fig. 22B). The maxillary bridge is wide.

Comparisons: Crocidura pseudorhoditis is smaller than C. rhoditis, C. elongata, and C. quasielongata, but larger than C. caudipilosa, Rhoditis Group members C. australis and C. pallida, all members of the Small-Bodied Group, and all members of the Ordinary Group except C. nigripes, which is similarly sized (fig. 19; tables 2, 7). The body of C. pseudorhoditis is more robust than those of C. microelongata, C. australis, and C. pallida, but less stocky than in C. rhoditis and C. caudicrassa (fig. 17). Crocidura pseudorhoditis is slightly darker (both pelage and feet) than C. rhoditis and tail bristles, though not abundant, are still more prevalent than in C. rhoditis (fig. 21). The thenar and hypothenar pads on the hind foot are both more rounded than the oblong pads of C. rhoditis (fig. 21). Crocidura pseudorhoditis is similar in pelage color to C. australis and C. pallida. Braincase breadth relative to skull length is greater than in many of the species with somewhat similar body sizes, including C. rhoditis, and all Elongata Subgroup and Thick-Tailed Group species. However, C. pallida is comparable in this regard, and C. australis has an even greater relative skull breadth no matter where it is measured (fig. 10). The braincase is a bit more angular in C. pseudorhoditis than in C. rhoditis, with the widest part of the cranium in C. pseudorhoditis forming an obtuse point (fig. 22). However, the angularity of the braincase does not reach the degree seen in C. nigripes. The ratio of rostral length to skull length is large in C. pseudorhoditis and exceeded only by that of C. rhoditis; the other Rhoditis Group members have comparatively short relative rostral lengths (RL/CIL) and outside the Rhoditis Group, C. nigripes and the two Thick-Tailed species (defined below) are comparable (fig. 10). Crocidura pseudorhoditis is larger in head-and-body length, foot length, and all cranial dimensions than C. australis and C. pallida (fig. 19; tables 2, 7). The slight difference in size between C. pseudorhoditis and C. rhoditis appears to be enhanced where the two species cooccur (Mt. Ambang; fig. 20), especially condyloincisive length (fig. 23). A principal components analysis of 12 cranial measurements that only included these two species readily separated them on the first axis, which represents size (fig. 19; table 6).

Comments: While Crocidura pseudorhoditis is on average smaller than C. rhoditis, the two species are easily confused phenotypically and they occur in sympatry in the north-east area of endemism (fig. 20; table 3). The large mitochondrial distances between C. rhoditis and C. pseudorhoditis, along with the slight but consistent differences in size, which appear to be enhanced where they are syntopic (fig. 23), strongly suggest these are independently evolving populations. Our BPP analyses tested species limits with an alignment that contains 15 C. rhoditis and 58 C. pseudorhoditis and is 96% complete. The analyses supported these species' distinction with 1.0 posterior probability with all prior combinations in all runs.

One of the few cases of clarity from our phylogenetic analyses is the consistent and well-supported sister relationship between Crocidura rhoditis and C. pseudorhoditis (figs. 4, 5, 7, 8; supplementary data S6). Given their phenotypic similarity, this makes sense from a morphological perspective. In our UCE species tree, we found these two species sister to C. elongata (fig. 7).

Specimens examined: Mt. Ambang (LSUMZ 39030, 39031, 39034–39036, 39039–39043, 39045, 39048, 39049, 39051, 39052, 39056, 39059, 39060, 39063, 39066, 39067, 39278, 39280–39283, 39293, 39297, 39298, 39300–39309, 39311–39314, 39316, 39317, 39322; MZB 43002; NMV C37985, C37992, C37996, C38017), Mt. Buliohuto (LSUMZ 38275–38278, 38284; NMV C37793), Mt. Dako (LSUMZ 36973, 36978, 36979, 36986, 36990, 36991, 36993, 36995, 36998, 37000, 37002, 37005, 37007, 37008, 37011–37016, 37033, 37034; MZB 38564–38566; NMV C37252, C37254, C37264, C37268, C37292, C37297, 37308, 37315, C37328, 37329, C37363, C37367), Mt. Torompupu (NMV C40308), Mt. Rorekatimbo (RMNH 38409 (= IZEA 4406), IZEA 4407).

Holotype: MZB 43003 (= MVZ 237610), an adult of unknown sex collected by K.C. Rowe on 30 October 2016. The specimen was preserved as a study skin, cleaned skull (fig. 24) and skeleton, and frozen tissues. External measurements from the holotype are: 141 mm × 65 mm × 16 mm × 11 mm = 9.8 g. The voucher specimen and a tissue sample will be permanently curated at MZB, with another tissue sample retained at MVZ.

Type locality: Indonesia, Sulawesi Selatan, Sinjai, Sinjai Barat, Gunung Perak Village, Mt. Bawakaraeng; 5.308463° S, 119.948661° E, 2390–2550 m elevation.

Etymology: We use the Latin for “southern,” as this species is the most southern member of the Rhoditis Group.

Geographic distribution: This species is known only from the south-west area of endemism and was only collected from Mt. Bawakaraeng, South Sulawesi Province (fig. 20) on traplines placed around 1660–2040 and 2390–2550 m elevation (fig. 13; table 3).

Diagnosis: Crocidura australis is a medium-sized shrew (tables 2, 7) with a thick, medium-brown dorsal pelage and slightly paler venter. Hairs of the middorsum are 6–7 mm long. The color of the tail matches that of the dorsal pelage, but the feet are paler than the surrounding fur, particularly on the digits (fig. 21A). The tail is slightly shorter than the head-and-body length and not distinctly bicolored. Applied hairs on the tail are inconspicuous and bristles are sparsely distributed along the proximal half of tail length. The claws are translucent and surrounded by small tufts of white hairs. On the hind foot, the claws are long and the tufts more prominent than on the forefoot. The foot pads are more darkly pigmented than the surrounding plantar and palmar surfaces, but the difference is greater on the hind foot than on the forefoot. The external ears, though not small, are indistinct from the surrounding fur, due to the matching color, length, and density of the pelage. The mystacial vibrissae are relatively short, and mostly unpigmented. A few pigmented vibrissae are found posterior to the shorter, unpigmented vibrissae. Those with pigment are pigmented only proximally, typically for no more than half of their length. The braincase is wide relative to skull length (fig. 10), high, and somewhat angular, with a lateral point in the mastoid region and relatively prominent lambdoidal ridges (fig. 24A). The ridge formed by the parietal-squamosal suture is indistinct. The interorbital region is also wide relative to skull length (fig. 10), but it is strongly tapered. Despite the strongly tapered interorbital region, the maxillary process is not prominent when viewed from the dorsal aspect. The dentition is somewhat prominent relative to palatal width (fig. 24A).

Comparisons: Crocidura australis is substantially larger in body size than all members of the Small-Bodied Group and somewhat larger than C. musseri, C. ordinaria, and C. solita of the Ordinary Group (fig. 9). It is smaller than members of the Thick-Tailed Group and Elongata Subgroup. Relative to members of the Rhoditis Group, it is smaller than C. rhoditis and C. pseudorhoditis in all dimensions except ear length, but similar in size to C. pallida (figs. 19, 23; tables 2, 7). In color, C. australis is darker than all other members of the Rhoditis Group, particularly on the feet. Crocidura ordinaria and C. solita of the Ordinary Group are similar in color and only a little smaller, but their braincases are more rounded than in C. australis. The somewhat angular shape of the braincase, however, is nevertheless more rounded than in C. pseudorhoditis. The great relative breadth of the braincase (BB/CIL) in C. australis distinguishes it from all other species except C. baletei (much smaller and darker), C. levicula (much smaller and darker), C. musseri (smaller and darker), and C. ordinaria (smaller). In comparison to other Rhoditis Group species, C. australis is substantially smaller in condyloincisive length and braincase breadth than the much larger skull of C. rhoditis and the somewhat larger skull of C. pseudorhoditis (fig. 19). Condyloincisive length is somewhat smaller than in C. pallida, although these two species are quite similar in head-and-body length. The interorbital region and labial breadth at M2 are also narrower than noted in C. rhoditis or C. pseudorhoditis (table 7). Relative rostral length (RL/ CIL) in C. australis is less than in any other member of the Rhoditis Group, though only slightly so compared to C. pallida.

Comments: Published references to C. rhoditis from the southwestern peninsula (Musser, 1987; Ruedi, 1995) probably refer to this species. Crocidura australis was consistently inferred as sister to the clade containing C. ordinaria and C. solita of the Ordinary Group with varying degrees of support (figs. 4, 5, 7, 8; supplementary data 6).

Specimens examined: Mt. Bawakaraeng (MZB 40991, 41027, 43003; NMV Z56801, NMV Z57200, NMV Z57223).

Holotype: MZB 43004 (= FMNH 210607), an adult female collected by J.A. Esselstyn on 18 October 2010. The specimen consists of a dried skin, cleaned skull (fig. 24B) and skeleton, and tissue sample. It carried one embryo measuring 18 mm in crown-rump length. External measurements from the holotype are: 140 mm × 62 mm × 15 mm × 8 mm = 11 g. The voucher specimen and a tissue sample will be permanently curated at MZB, with another tissue sample retained at FMNH.

Type locality: Indonesia, Sulawesi Selatan, Luwu Utara, Sukamaju, Mt. Balease; 2.4995° S, 120.4874° E, 900 m elevation.

Etymology: We use the Latin pallida to highlight the pale color of the feet of this species.

Geographic distribution: Recorded from the west-central (Mt. Gandang Dewata, West Sulawesi Province; Mts.Torompupu and Balease, Central Sulawesi Province; Mt. Latimojong and Rindingallo, South Sulawesi Province), east-central (Mts. Katopasa and Tompotika, Central Sulawesi Province), and south-east areas of endemism (Mt. Mekongga, Southeast Sulawesi Province; fig. 20). Across these areas, we found this species over a broad elevational range, from approximately 100 to over 2500 m (fig. 13; table 3).

Diagnosis: Crocidura pallida is a moderately sized shrew (figs. 9, 23; tables 2, 7) with very pale feet (fig. 21B) and a somewhat pale ventral side of the tail. The tail is shorter than the head-and-body length (fig. 9; table 2). The dorsal pelage is gray to gray-brown while the ventral pelage is pale gray. The pinna and the dorsal side of the tail match the dorsal pelage, but the dorsal side of the feet are distinctly paler. The dorsal surface of the hand is nearly white, but some pigment is present near the wrist. The hind feet show a similar pattern but are modestly darker. Ventrally, the feet are also pale, especially on the digits, which are usually white. The darkest parts of the hind foot are usually the thenar and hypothenar pads (fig. 21B). The pads of the forefeet are rarely pigmented. Tail bristles are sparse to nearly absent (fig. 21B), extending along no more than the proximal third of the tail length. The tail varies from uniformly colored to moderately bicolored with a paler ventral side. The skull is typical of Crocidura of this size. The braincase is generally rounded, but a subtle lateral point is evident at the mastoid region when viewed from a dorsal aspect (fig. 24B). The braincase is somewhat inflated dorsoventrally, and it is wide relative to skull length (fig. 24B), as are the interorbital region and rostrum (fig. 10; table 7). The wide interorbital region gives the maxillary process a relatively weak appearance. In some individuals, the nasal passage is particularly inflated and laterally bulging, further obscuring the maxillary process. The rostrum is somewhat long, relative to skull length (fig. 10). The maxillary bridge is usually narrow, with an anteriorly placed lacrimal foramen. The posterior portion of the hard palate is narrow, sandwiched between broad molars.

Comparisons: Crocidura pallida is smaller than C. rhoditis, C. pseudorhoditis, C. nigripes, and members of the Elongata Subgroup and Thick-Tailed Group. It is considerably larger than all members of the Small-Bodied Group and somewhat larger than all members the Ordinary Group except C. nigripes. In absolute terms and relative to head-and-body length, the tail length is comparable to members of the Ordinary, Thick-Tailed, and Rhoditis groups, but substantially shorter than in all members of the Long-Tailed Group and considerably longer than in any member of the Small-Bodied Group (fig. 9). Outside of the Rhoditis Group, most species have much darker feet than C. pallida. These include C. caudipilosa and C. microelongata of the Long-Tailed Group, C. normalis, C. musseri, and C. nigripes of the Ordinary Group, both Thick-Tailed Group members, and all members of the Small-Bodied Group except C. lea. Within the Rhoditis Group, C. pallida is smaller in absolute measurements (fig. 9) and more delicately built than C. rhoditis and C. pseudorhoditis (fig. 17); it is paler with a relatively narrower braincase and interorbital region than the otherwise similarly proportioned C. australis. Rostral length makes up a smaller proportion of skull length (RL/CIL) in C. pallida than in either C. rhoditis or C. pseudorhoditis, but this trait is comparable in C. australis (fig. 10).

Comments: Substantial mitochondrial genetic divergence is evident between populations from Mt. Katopasa, Mt. Tompotika, and the remaining sites (up to 0.089 Jukes-Cantor distance; fig. 4; supplementary data S3). However, these populations form a cohesive set of morphological specimens, form a clade in our phylogenetic analyses of nuclear genes (figs. 7, 8; supplementary data S6), and all of the mitochondrial variation is partitioned geographically (i.e., no sympatry between divergent mitochondrial clades). We therefore did not divide them further.

We identified a single specimen of Crocidura pallida from Mt. Latimojong (MVZ 237618), a locality where C. solita is abundant. Although C. pallida is slightly larger, these two can be difficult to distinguish. As such, it is possible that this specimen is a slightly large individual of C. solita with mtDNA introgressed from C. pallida. Unfortunately, we did not obtain nuclear loci (exons or UCEs) from this specimen and therefore cannot test for introgression. Contamination of this cytochrome b sequence is unlikely because it is unique in our alignment. We favor the hypothesis that MVZ 237618 is C. pallida because its cytochrome b sequence differs slightly from C. pallida sequences from nearby localities (i.e., Mts. Gandang Dewata and Balease) and phenotypically, it is nearer the averages for C. pallida than C. solita.

Phylogenetic estimates were not consistent regarding the relationships of Crocidura pallida. Our mitochondrial gene trees put it as sister to a clade of Small-Bodied species, C. caudipilosa, and C. normalis (figs. 4, 5). However, our nuclear-based inferences placed C. pallida as part of the large basal polytomy (figs. 7, 8; supplementary data S6).

Specimens examined: Mt. Balease (FMNH 210580–210592, 210608, 210609, MZB 43004), Mt. Gandang Dewata (FMNH 218687–218702, 218989; MZB 34872, 34886, 34888, 34889), Mt. Katopasa (LSUMZ 39527, 39529–39538; MVZ 238115–238118; NMV C40187, C40192, C40199, C40206, C40214, C40217, C40307, Z56723, Z62366, Z61754, Z62414), Mt. Latimojong (MVZ 237618), Mt. Mekongga (MWFB 8059, 8115, 8125, 8139, 8150, 8161, 8162, 8195, 8196, 8438, 8439), Rindingallo, Tana Toraja (MSB 93256), Mt. Tompotika (FMNH 213366–213369), Mt. Torompupu (NMV C40307).

Holotype: USNM 217553, an adult male collected by H.C. Raven on 3 August 1916. Prepared as a skin and skull. External measurements from the holotype are 111 mm × 51 mm × 14 mm; ear length and weight were not recorded.

Type locality: “Temboan, northeastern Celebes” (Miller and Hollister, 1921: 102; fig. 25). Temboan is located at 0.979° N, 124.605° E, 650 m elevation in the Southeast Minahasa Regency, North Sulawesi Province, 6 km south of Kalait. See the gazetteer (appendix) for details of our interpretation of Raven's field notes.

Geographic distribution: Probably endemic to the northern peninsula. We identified specimens of this species from the northwest area of endemism (Mt. Dako, Central Sulawesi Province and Mt. Buliohuto, Gorontalo Province) and the type locality is in the north-east area of endemism (Temboan, North Sulawesi Province). We did not find this species on Mt. Ambang (north-east area of endemism), where its ecological position is apparently held by another tiny, but not closely related shrew, Crocidura tenebrosa (fig. 25). We recorded C. lea at both high and low elevations on both Mts. Dako (512 and 1600 m) and Buliohuto (400–600 and 1200 m; table 3). We did not collect any specimens of this species at our 410 m site on Mt. Dako. The type locality (Temboan) is at roughly 650 m.

Emended diagnosis: Crocidura lea is a very small shrew (tables 2, 8) with a brown dorsal pelage, gray-brown ventral pelage, and pale feet. The pinnae are prominent and slightly paler than the surrounding fur, especially at the lobe. The mystacial vibrissae are generally unpigmented, but a few, typically the longer ones, are brown or black proximally. Dorsally, foot color transitions from gray-brown at the wrist and ankle to pale pinkish white on the digits (fig. 27B). The claws are small, unpigmented, and sometimes lack the surrounding tuft of white hairs present in many other species. Ventrally, the forefeet mostly lack pigment, but the hind feet are darker. In the hind feet, color transitions from brown posteriorly from the heel, and to a lesser extent laterally, to white at the digits. However, pigment is concentrated in the pads, particularly around the base of the thenar, hypothenar, and in some cases the 4th (lateralmost) interdigital pad (fig. 27B). The tail is shorter than head-and-body length, sometimes dorsoventrally bicolored, and thinly covered with very small, dark brown applied hairs. A few sparse bristles are present along the proximal three-fourths of the tail (fig. 27B). The skull is short, narrow, and somewhat rounded in overall shape (fig. 28B). Relative to body size, the skull is slightly longer than expected, and relative to skull length, the braincase (BB/CIL) is especially narrow, but the interorbital region is not unusual in its relative width (IOW/CIL; fig. 10). The maxillary process is weak (fig. 28B). The I3 (U2) is smaller than the C (U3). The parastyle of P4 is modest.

Comparisons: Crocidura lea is much smaller than all Sulawesi Crocidura outside of the Small-Bodied Group (fig. 9). Among the Small-Bodied species, only C. baletei and C. tenebrosa are known from the northern peninsula. Crocidura levicula, C. parva, and C. mediocris are apparently restricted to the central core and eastern and southwestern peninsulas (fig. 25). We therefore focus our comparisons on the cooccurring, northern peninsula species. However, first we note that the allopatric C. levicula has a broader skull (figs. 10, 26; tables 8, 9), shorter tail, shorter hind feet, and more tail bristles than does C. lea (figs. 27, 29; table 2). Among the Small-Bodied shrews of the northern peninsula, C. lea has a slightly paler and more strongly dorsoventrally bicolored pelage than C. baletei or C. tenebrosa. Both of the latter species are relatively uniform in color, C. baletei being medium brown and C. tenebrosa dark brown. Crocidura lea also has paler skin wherever it is exposed than either C. baletei (somewhat darker) or C. tenebrosa (much darker). Dorsally, the feet of both C. baletei and C. tenebrosa are uniformly dark, whereas the foot color transitions posteriorly from brown at the heel to white or pink at the digits in C. lea (fig. 27). Similarly, the tails of both C. baletei and C. tenebrosa are not obviously bicolored, and both are darker. Tail bristles are present on approximately three-fourths of the tail length in all three species, but they are much more abundant and longer on the tails of C. baletei and C. tenebrosa than in C. lea (fig. 27). Applied hairs on the tail are also much more visible in the two new species than in C. lea. Tail length in C. lea is greater, on average, than in C. baletei, and substantially so compared to C. tenebrosa (fig. 9; table 2). The skulls of C. lea are comparable in length, but significantly narrower at the braincase, interorbital region, maxillary process, and to a lesser extent at the rostrum than those of C. baletei and C. tenebrosa (figs. 10, 30; table 8). Crocidura lea occupies distinct morphometric space from the other northern peninsula endemics in bivariate plots of skull length and width and the first two principal components from an analysis of 12 cranial measurements (fig. 26). The C (U3) is larger than I3 (U2) in C. lea, but they are usually subequal in occlusal surface area in C. baletei and C. tenebrosa. The parastyle of P4 is more prominent in C. tenebrosa than in C. lea or C. baletei.

Comments: An 1111 bp cytochrome b sequence from the holotype (USNM 217553) provides strong evidence that the tiny shrews we collected on Mts. Buliohuto and Dako (samples from Mt. Dako were referred to as “Crocidura pale lea” by Esselstyn et al., 2019) represent this species. Interestingly, we did not find C. lea on Mt. Ambang, our sample site nearest the type locality. Rather, the tiny shrews we collected from Mt. Ambang represent an undescribed species (C. tenebrosa) that does not appear to be a close relative of C. lea. We suspect that C. lea is widespread at relatively low elevations on the northern peninsula, but is displaced by smaller-range, darker-colored endemics in some higher-elevation areas (fig. 13; table 3). Our only surveys on Mt. Ambang were centered on 1500 m elevation (fig. 3). For BPP species delimitation results, see the C. tenebrosa account below.

Ruedi (1995) reported Crocidura lea from Mt. Rorekatimbo, but we refer those specimens to C. normalis (see below) based on a cytochrome b sequence from Ruedi et al. (1998).

None of our phylogenetic estimates found strong support for the closest relative(s) of Crocidura lea, but some consistencies are nevertheless apparent. Our mitochondrial analyses placed C. lea as the sister to C. rhoditis and C. pseudorhoditis (fig. 4) or C. elongata (fig. 5), while our UCE species tree placed C. lea as sister to the clade containing C. elongata, C. rhoditis, and C. pseudorhoditis (fig. 7). Other nuclear-based inferences lacked support (fig. 8; supplementary data S6).

Specimens examined: Mt. Buliohuto (LSUMZ 38254–38262, 38271–38274; NMV C37692, C37715, C37716, C37720, C37721, C37735, C37744, C37829), Mt. Dako (LSUMZ 36947–36950, 36952, 36954–36956, 36958, 36960, 36964, 36968–36972), Temboan (USNM 217553).

Crocidura “dark lea” Esselstyn et al., 2019: 1715. Informal name.

Holotype: MZB 43005 (= LSUMZ 36959), an adult male, collected by J.A. Esselstyn on 10 March 2013. Specimen consists of a cleaned skull (fig. 28A), a fluid-preserved body, and frozen tissues. External measurements from the holotype are: 119 mm × 50 mm × 12 mm × 8 mm = 4.9 g. The voucher and a tissue sample will be permanently curated at MZB, with another tissue sample retained by LSUMZ.

Type locality: Indonesia, Sulawesi Tengah, Toli Toli, Galang, Malangga Selatan, Mt. Dako; 1.10607° N, 120.93853° E, 1600 m.

Etymology: Named in honor of Danilo S. Balete, an accomplished mammal systematist who passed away much too early in 2017. Danny, as he is known, was a tireless collector, whose years of field research in the Philippines built the kind of museum collections needed for comprehensive catalogs of biodiversity. Without the focused, long-term efforts of people like Danny, the type of comprehensive taxonomic research we are attempting here would not be possible.

Geographic distribution: Apparently endemic to montane forests of the north-west area of endemism. We recorded this species only on Mts. Dako and Buliohuto (fig. 25) at elevations around 1600 and 1400 m, respectively (fig. 13). Our sampling efforts at lower elevations (400–600 m; fig. 3) on both mountains did not find this species (table 3). Surveys in the mountains west and south of Mt. Dako (i.e., the base of the northern peninsula) and east of Mt. Buliohuto are needed to draw firmer conclusions regarding the geographic distribution of Crocidura baletei.

Diagnosis: Crocidura baletei is a very small shrew (tables 2, 8) with a brown pelage and skin color, where exposed. Overall, this species is quite uniform in color, with the feet, pinnae, and tail matching the surrounding fur. The venter and dorsum are similarly colored. The pinnae are, however, paler around their lower margin. The mystacial vibrissae are pigmented for about the proximal half of their length, but unpigmented distally. The color of the digits is only slightly paler than that of the rest of the foot (fig. 27A). Tail length is shorter than head-and-body length (fig. 9; table 2) and bristles are present for approximately three-fourths of tail length. Small, applied hairs are present on the tail, but difficult to see with the naked eye. The skull is short, particularly the rostral portion, relative to the overall body size of the species (fig. 28). Relative to skull length, the skull is broad overall, especially at the braincase (fig. 10). On the upper dentition, I3 (U2) and C (U3) are subequal and the parastyle of P4 is modest (fig. 28A).

Comparisons: This shrew is much smaller than all other Sulawesi Crocidura outside of the Small-Bodied Group (fig. 9). Within the Small-Bodied Group, C. lea and C. tenebrosa are the only other taxa known from the northern peninsula. Among the tiny shrews of the northern peninsula, C. baletei and C. tenebrosa are more uniform in color than C. lea, with C. baletei being paler (medium brown) than C. tenebrosa (dark brown). Crocidura lea has a paler integument, wherever it is exposed, and it is more strongly bicolored than either C. baletei (somewhat darker) or C. tenebrosa (much darker). Dorsally, the feet of both C. baletei and C. tenebrosa are uniformly dark, whereas the foot color transitions posteriorly from brown at the heel to white or pink digits in C. lea. Similarly, the tails of both C. baletei and C. tenebrosa are not obviously bicolored, and both are darker. Tail bristles are present on approximately three-fourths of the tail length in all three species, but they are much more abundant and longer on the tails of C. baletei and C. tenebrosa than in C. lea (fig. 27). Applied hairs on the tail are also more visible in the two new species than in C. lea. Tail length in C. baletei is slightly shorter than in C. lea, but substantially longer than in C. tenebrosa (fig. 9). On the hind foot, the thenar pad is wider and more prominent than in C. lea, but comparable to that of C. tenebrosa (fig. 27). The skulls of all three species are similar in length, but that of C. baletei is considerably wider than that of C. lea (figs. 10, 26, 30). The braincase breadths relative to skull lengths (BB/CIL) are similar in C. baletei and C. tenebrosa, but the interorbital region is relatively wider (IOW/CIL) in C. tenebrosa (fig. 10). Relative rostral length (RL/CIL) is slightly less in C. baletei than in C. tenebrosa (fig. 10). The parastyle of P4 is less prominent in C. baletei than in C. tenebrosa.

Comments: See comments for Crocidura tenebrosa, below.

Specimens examined: Mt. Buliohuto (LSUMZ 38264, 38265, 38268, 38270; NMV C37777, C37794, C37800), Mt. Dako (LSUMZ 36946, 36951, 36953, 36961–36963, 36965, 36966; MZB 43005; NMV C37306).

Holotype: MZB 43006 (= LSUMZ 39272), an adult male collected by H. Handika on 21 February 2016. The specimen comprises a study skin, cleaned skull and skeleton, and frozen tissues. External measurements from the holotype are 100 mm × 37 mm × 11 mm × 8 mm = 6.28 g. The voucher specimen and a tissue sample will be permanently curated at MZB, with additional tissues retained by LSUMZ.

Type locality: Indonesia, Sulawesi Utara, Bolaang Mondgondow, Passi Timur, Insil, Mt. Ambang, near Lake Aliyah; 0.76385° N, 124.41188° E, 1481 m elevation.

Etymology: Tenebrosa is Latin for “dark,” describing the overall color of this animal.

Geographic distribution: Recorded only on Mt. Ambang in the north-east area of endemism (North Sulawesi Province) between 1400 and 1500 m elevation (fig. 25; table 3). Future surveys of other mountains in the north-central area of endemism and at the eastern extreme of the north-east area of endemism may extend the known range of this species.

Diagnosis: Crocidura tenebrosa is a very small (fig. 9; tables 2, 8), very dark shrew. The pelage and all areas of exposed skin (feet, tail, pinna, lips, and nose) are uniformly dark brown (fig. 27C). The overall dark pelage of this animal appears to be primarily from a darker proximal base of individual hairs. The mystacial vibrissae are black for most of their length, but unpigmented at the tips. The soles of the feet are mostly dark brown, but the areas that tend to accumulate pigment (base of the foot pads) are black (fig. 27C). The claws are translucent. The tail is short, with a notable density of applied hairs and a comparative abundance of bristles extending for three-fourths of the tail length. The skull is short, particularly the postpalatal portion, relative to its width (figs. 10, 28C) and to overall body size. Posteriorly, the skull is rounded, with a smoothly inflated and broad braincase (fig. 28C). The lambdoidal ridge is thin, but relatively high for such a delicate skull. The interorbital region is smoothly tapered, and the maxillary process is relatively inconspicuous, a consequence of the broad interorbital region. The palate is narrow, with a somewhat prominent dentition. The C (U3) and I3 (U2) are subequal in occlusal surface area. The parastyle of P4 is prominent.

Comparisons: Crocidura tenebrosa is smaller than all Sulawesi shrews except the other members of the Small-Bodied Group, which are all comparable in body size. Among these, only C. lea and C. baletei are found on the northern peninsula, near the range of C. tenebrosa. Crocidura tenebrosa is readily distinguished from C. lea by its more dark and uniform color, shorter tail (fig. 9), and wider skull (figs. 10, 26). Crocidura tenebrosa is distinguished from its sister species, C. baletei, by its shorter tail (fig. 30; table 2), darker integument, shorter relative hind-foot length (fig. 17), and slightly narrower braincase, but wider interorbital region and rostrum (fig. 30; table 8). The ratio of the braincase breadth to interorbital width is slightly lower for C. tenebrosa than C. baletei (fig. 10). For additional details, see comparisons sections of C. lea and C. baletei, above.

Comments: We tested species limits with BPP between Crocidura baletei and C. tenebrosa, among these two species and C. lea, and among these three species and C. levicula. The alignments we analyzed contain: 6 and 13; 6, 13, and 12; and 6, 13, 12, and 34 individuals per species, respectively. Alignments are 91%, 87%, and 93% complete. All analyses supported all species with a posterior probability of 1.

Crocidura baletei and C. tenebrosa are probable sister species (figs. 7, 8) and phenotypically similar. We define them as distinct species because they have modest differences in color and skull shape, they are divergent in their mitochondrial DNA (mean 0.076 Jukes-Cantor distance; supplementary data S4), they are reciprocally monophyletic in our UCE, mitochondrial, and nuclear exon trees (figs. 4, 5, 7, 8; supplementary data S6), and finally since both species have been recorded only above 1000 m elevation (fig. 13), they are likely isolated from each other by the low, dry habitat of the Gorontalo Divide. These two small-bodied highland endemics may be replaced across the northern peninsula lowlands by the similarly sized but paler colored and distantly related C. lea.

Specimens examined: Mt. Ambang (LSUMZ 39019–39023, 39025, 39267–39271, 39273, 39274; MZB 43006).

Holotype: USNM 219450, an adult female collected by H.C. Raven on 13 February 1918. Specimen prepared as a skin and skull. External measurements from the type are 108 mm × 44 mm × 11 mm; ear length and weight were not recorded.

Type locality: “Pinedapa, Middle Celebes” (fig. 25). Pinedapa is located near the north coast of Central Sulawesi Province, in Poso Regency, just west of the town of Mapane. Musser (2014) placed the locality at 1.4167° S, 120.5833° E, 31 m elevation.

Geographic distribution: The type locality is near sea level in the northwestern part of Sulawesi's west-central area of endemism. We found this species to occur in the northern part of the west-central area of endemism (Mts. Rorekatimbo and Torompupu, Central Sulawesi Province), in two lowland areas near the boundary between the west-central and south-east areas of endemism (Wasponda and Tolala, Central Sulawesi Province), and in the east-central area of endemism (Mts. Katopasa and Tompotika, Central Sulawesi Province; fig. 25). These records span a broad elevational range, from approximately 400 m on Mt. Tompotika to 2000 m on Mt. Rorekatimbo (fig. 13; table 3).

Emended diagnosis: A very small (tables 2, 8), medium- to dark-brown shrew, with almost uniformly colored pelage, feet, pinna, and tail. The ventral pelage is only slightly paler than the dorsal pelage. The digits are similarly pigmented as the other foot parts. The tail is unusually short (table 2), with an abundance of bristles (some of which are pigmented) extending along the proximal two-thirds of tail length (fig. 29A). The hind foot is short, with darkly pigmented pads. The claws are pale, perhaps lightly pigmented, and surrounded by a small tuft of brown hairs (fig. 29A). The skull is small but broad, more so at the braincase than at the interorbital region (figs. 10, 31A). Relative to body size and skull breadth, the skull is quite short, particularly the rostrum (fig. 10). The C (U3) is equal to or larger in occlusal area than the I3 (U2).

Comparisons: Crocidura levicula is one of the smallest shrew species on Sulawesi, only being comparable to other members of the Small-Bodied Group (fig. 9). It has a shorter relative tail length than any other shrew species on Sulawesi (fig. 9). Its hind-foot length is shorter than in all Sulawesi shrews except one other member of the Small-Bodied Group, C. parva (fig. 9; table 2). Crocidura levicula is also darker in color than both C. lea and C. baletei, but not C. tenebrosa. Absolute tail length is slightly less in C. levicula than in C. parva. Tail bristles are more abundant on specimens of C. levicula than in C. lea, C. tenebrosa, and C. parva. The skull of C. levicula is small, delicate, and broad at the braincase and to a lesser degree at the interorbital region (fig. 10). Its skull width relative to length is greater than in C. lea, C. parva, and C. mediocris, and to a lesser degree, greater compared to C. tenebrosa (fig. 26). Crocidura levicula has a lesser skull length and breadth than C. baletei (fig. 26), but these two species have similar relative skull widths (fig. 10). Samples of Crocidura levicula are nearly nonoverlapping with other members of the Small-Bodied Group in morphometric space in bivariate plots of skull length versus width and along the first two axes from a PCA of 12 cranial measurements (fig. 26). In C. levicula, the C (U3) is equal to or greater in occlusal surface area than is I3 (U2) (fig. 31), which is similar to the conformation in C. lea.

Comments: Miller and Hollister (1921) named this species based solely on the holotype, and unfortunately, we failed to obtain DNA from this specimen. We followed the Tsai et al. (2020) phenol-chloroform extraction protocol and used two separate extractions. The addition of resalting the ethanol supernatants did not yield a quantity of DNA that could be detected with a Qubit 2.0 fluorometer using the dsDNA High Sensitivity Assay Kit (Thermo Fisher Scientific, Waltham, MA). As such, our identification of Crocidura levicula is based on the morphological similarity of the holotype to specimens we collected in areas nearest the type locality.

One specimen (NMV Z63390) from Mt. Torompupu appears to have an introgressed mitochondrion from syntopic Crocidura ordinaria, a member of the Ordinary Group detailed below. Our identification of this specimen is based on morphology alone; we did not sequence any nuclear DNA from this specimen and therefore cannot test our introgression hypothesis.

In our phylogenetic estimates, UCE data confidently placed Crocidura levicula as sister to C. caudipilosa (figs. 7, 8). However, our nuclear-exon and mitochondrial inferences placed this species as a member of a clade of mostly Small-Bodied species that also includes C. caudipilosa (fig. 5; supplementary data S6).

Crocidura levicula was delimited by all BPP analyses. See the C. tenebrosa account above for details.

Specimens examined: Mt. Katopasa (LSUMZ 39512–39516; NMV C40180, C40181, C40200, C40203, C40207, Z55557), Pinedapa (USNM 219450), Mt. Rorekatimbo (FMNH 213192, 213193, 213271), Tolala (FMNH 210576, 210577), Mt. Tompotika (FMNH 213343–213365), Mt. Torompupu (LSUMZ 39446–39449; NMV C40292, C40294, C40306, C40311, Z63365), Wasponda (FMNH 210578, 210579).

Holotype: MZB 43007 (= FMNH 210603), an adult female collected by J.A. Esselstyn on 22 October 2010. The specimen consists of a dried skin, cleaned skull and skeleton, and tissue samples. External measurements from the holotype are: 102 mm × 41 mm × 11 mm × 7 mm = 3.7 g. The voucher specimen and a tissue sample will be permanently curated at MZB, with an additional tissue sample retained by FMNH.

Type locality: Indonesia, Sulawesi Selatan, Luwu Utara, Sukamaju, Mt. Balease; 2.50002° S, 120.48726° E, 862 m elevation.

Etymology: Mediocris is Latin for “average,” applied in recognition that this is a species of shrew with no striking phenotypic traits worthy of hanging a descriptive name on.

Geographic distribution: This species is found across the southern and western portions of the west-central area of endemism (Mts. Torompupu and Balease, Central Sulawesi Province; Salu Tiwo, West Sulawesi Province; and Mt. Latimojong, South Sulawesi Province) and in the south-east area of endemism (Mt. Mekongga, Southeast Sulawesi Province; fig. 25). We found this species from approximately 200 to 1900 m elevation (fig. 13). Although the species was present at Salu Tiwo (ca. 200 m), we did not find it at higher elevations on Mt. Gandang Dewata (table 3).

Diagnosis: Crocidura mediocris is another small shrew (tables 2, 8), with a delicate build and dark gray to medium brown pelage that is only slightly paler on the venter. The tail is shorter than head-and-body length and moderately dorsoventrally bicolored. The feet are somewhat paler than the pelage, more so around the digits (fig. 29B). The pelage covers the body completely, but individual hairs are short (2–3 mm at middorsum). Mystacial vibrissae are short relative to body size, and unpigmented along most of their length. The tail is covered in moderately dense bristles along the proximal two-thirds of its length and a relatively high density of short, dark, applied hairs along its entire length (fig. 29B). The dorsal surfaces of the digits and, in some specimens, the feet are light brown. The plantar and palmar surfaces are generally dark, but the ventral sides of the digits are white (fig. 29B). As is typical, more pigment is present on the posterior and lateral portions of the palmar and plantar surfaces. The skull is small and delicate, with a relatively robust dentition (fig. 31B). The braincase is somewhat narrow and the interobital region is narrow relative to skull length (fig. 10). The maxillary process is modest and the maxillary bridge is narrow (fig. 31B). The lambdoidal ridge is prominent for such a small species.

Comparisons: Crocidura mediocris is one of the smallest species on Sulawesi, easily distinguished on body size alone from all species outside the Small-Bodied Group (fig. 10). Crocidura normalis is the smallest member of the Ordinary Group and may be closely related to C. mediocris and C. parva, also of the Small-Bodied Group (figs. 5, 7). Compared to C. normalis, C. mediocris is smaller, paler colored, and has a shorter tail (fig. 10; table 2). Among the Small-Bodied species, C. mediocris has the shortest average head-and-body length. Its tail is also short, and relative tail length is comparable to C. baletei, greater than in C. levicula, C. tenebrosa, and C. parva, and only slightly less than in C. lea (fig. 9). In color, the feet of C. mediocris are darker than those of C. lea. Aside from the relative tail-length differences, C. mediocris is similar to C. levicula in overall proportions but differs in having a longer and slightly narrower skull (table 8), paler pelage and feet, more rounded hypothenar on the hind foot (fig. 29), and on average, fewer tail bristles. Crocidura mediocris is also paler than both C. tenebrosa (fig. 27C) and C. parva (fig. 29C). In cranial proportions, C. mediocris has a greater average condyloincisive length than C. levicula and C. parva, but is comparable in this regard to C. lea, C. baletei, and C. tenebrosa (table 8). The length of the rostrum, relative to skull length (RL/CIL), is greater on average than in C. lea, C. levicula, and C. parva (fig. 10). The relative breadth of the braincase (BB/ CIL) is greater than in C. lea and slightly more than in C. parva, but less than in the other Small-Bodied species. In morphometric space, a bivariate plot of condyloincisive length versus braincase breadth shows that C. mediocris occupies a width-to-length ratio between the narrow C. lea and C. parva and the wide C. levicula, C. baletei, and C. tenebrosa (fig. 26A). A PCA of cranial dimensions shows that specimens of C. mediocris occupy moderately distinct morphometric space, overlapping primarily with individuals of C. lea and C. parva (fig. 26A).

Comments: Crocidura mediocris is closely related to C. parva. See the C. parva account for a discussion of phylogenetic relationships between these two phenotypically similar species. For coalescent species delimitation results, see the C. normalis (Ordinary Group) and C. parva accounts below.

Specimens examined: Mt. Balease (FMNH 210563–210575, 210602, 210604–210606; MZB 43007), Mt. Gandang Dewata (FMNH 218617, 218619, 218620, 218622–218638, 218640–218642, 218644–218649), Mt. Latimojong (MVZ 237574), Mt. Mekongga (MWFB 8092, 8104, 8121, 8131–8136, 8140–8142, 8144, 8145, 8148, 8149, 8152, 8153, 13508, 13510, 13513, 13514), Salu Tiwo (FMNH 218614–218616, 218618, 218621, 218639, 218643, 218650, 218651), Mt. Torompupu (LSUMZ 39452–39457, 39555; MVZ 238109–238114; NMV C40246, C40251, C40253, C40274, C40278, Z62199).

Holotype: MZB 43008 (= MVZ 237577), an adult female collected on 18 October 2016 by H. Handika. The specimen is preserved as a study skin, cleaned skull (fig. 31C) and skeleton, and frozen tissues. External measurements recorded in the field from the type are: 101 mm × 41 mm × 2 mm × 7 mm = 3.9 g. Hind-foot length for the holotype was erroneously recorded as 2 mm; the true value is almost certainly 12 mm. The voucher specimen and a tissue sample will be permanently curated at MZB, while an additional tissue sample will be retained at MVZ.

Type locality: Indonesia, Sulawesi Selatan, Sinjai, Sinjai Barat, Gunung Perak Village, Mt. Bawakaraeng; 5.286815° S, 119.961406° E, 1713–1754 m elevation.

Etymology: Parva is Latin (feminine) for “tiny,” used in recognition of the diminutive body size of this species.

Geographic distribution: This animal was extremely abundant on Mt. Bawakaraeng of the south-west area of endemism (South Sulawesi Province; fig. 25). We did not encounter it anywhere else and it may be endemic to this region. All specimens were collected between 1700 and 2500 m elevation (table 3). Habitat destruction has eliminated low- and middle-elevation forests in South Sulawesi, which may have altered the natural elevational distribution of this species.

Diagnosis: Crocidura parva is among the smallest shrews on Sulawesi and it has the narrowest average braincase breadth of any species on the island (tables 2, 8). For a shrew of its head-and-body length, it has a rather delicate build (fig. 17). Crocidura parva has a gray-brown to dark brown pelage, with a slightly grayer venter. The fur is not particularly dense, and individual hairs on the middorsum are only about 3 mm in length, despite all specimens coming from mid to high elevation. Overall, the color of the skin on the feet, ears, lips, and tail match that of the surrounding pelage. The mystacial vibrissae are short (most <10 mm) and pigmented proximally for no more than half of their length. The tail length is short absolutely, and less than head-and-body length (fig. 9; table 2). Tail bristles are relatively abundant and are present along about two-thirds of the proximal tail length (fig. 29C). The skull is short, but also quite narrow at the braincase, giving it an elongate and dorsoventrally compressed overall appearance. The lambdoidal ridge is indistinct. The width of the interorbital region is middling relative to skull length (fig. 10) and the maxillary process is not prominent. Rostral length forms a relatively small proportion of skull length (fig. 10). The molar row is crowded on the narrow palate (fig. 31C). On the upper cheek teeth, the metacone is unusually prominent relative to the protocone, particularly on P4.

Comparisons: Crocidura parva is much smaller than all known Sulawesi shrews except other members of the Small-Bodied Group. Externally, C. parva is rather uniform in color, as are C. levicula, C. baletei, and C. tenebrosa, but other species have paler feet than bodies, including C. lea and to a lesser extent, C. mediocris. The overall color of C. parva is slightly darker than C. mediocris, and considerably darker than C. lea. The tail has more bristles than in C. mediocris, but fewer than C. levicula (fig. 29). The pelage is shorter than in other species, except that C. mediocris is similar. The skull of the tiny C. parva is slightly smaller overall than seen in any of the other very small taxa (table 8) and relative rostral length (RL/CIL) is less than in C. mediocris, C. baletei, and C. tenebrosa (fig. 10). Most distinctively, the braincases of the other species are more dorsoventrally inflated, particularly in the case of C. lea. The flat braincase of C. parva is narrower than in all other Small-Bodied species, but the range of values overlaps with those from C. lea, C. levicula, and C. mediocris (fig. 26). Bivariate plots of cranial length versus width and the first two axes from a PCA of 12 cranial dimensions show C. parva to occupy largely distinct morphometric space, but to overlap some with C. lea, C. levicula, and C. mediocris (fig. 26). Finally, in C. parva, on the upper molariform teeth, the protocone is relatively low, compared with those of the other Small-Bodied shrews (figs. 28, 31).

Comments: We found this species only on Mt. Bawakaraeng, but our nearest sample sites are on Mt. Latimojong (fig. 25), some 215 km away. Much of the intervening habitat between Mts. Bawakaraeng and Latimojong is low elevation and dry, with the lowest area (the Tempe Depression) considered a barrier to many vertebrate taxa. Nevertheless, sampling additional sites in the south-west area of endemism may extend the known range of Crocidura parva. We interpret this probable isolation as additional support for distinguishing C. parva from C. mediocris, which is phenotypically similar. These two species are sister taxa in our mitogenome phylogeny (fig. 5), but each is paraphyletic with respect to the other in our UCE (figs. 7, 8) and nuclear exon phylogenies (supplementary data S6). Because of the slow evolutionary rate of UCEs and exons, we do not consider this to be strong evidence against their distinction.

We tested species limits in BPP between Crocidura parva and C. mediocris and among C. parva, C. mediocris, and C. normalis. The former analysis employed 20 individuals of each species and the matrix is 95% complete. The second analysis used the same dataset, but with the addition of 36 specimens of C. normalis. This matrix is 94% complete. All analyses supported each species with a posterior probability of 1.

Specimens examined: Mt. Bawakaraeng (MVZ 237575, 237576, 237578–237581, 237583, 237584, 237600–237609; MZB 43008; NMV Z57156, Z56280, Z56302, Z56314, Z56319, Z56995, Z57008, Z57012, Z57045, Z57046, Z57047, Z57057, Z57064, Z57103, Z57129, Z57208).

Holotype: MZB 43009 (= MVZ 237632), an adult female collected on 12 August 2016 by H. Handika, and prepared as a cleaned skull, formalin-fixed carcass, and tissue samples. External measurements of the holotype are 130 mm × 56 mm × 15 mm × 9 mm = 12.0 g. The voucher specimen and a tissue sample will be permanently curated at MZB and an additional tissue will be retained at MVZ.

Type locality: Indonesia, South Sulawesi Province, Luwu Regency, Latimojong District, Ulusalu, Gamaru Village, Mt. Latimojong; 3.43103° S, 120.09457° E, 2518 m elevation.

Etymology: Brevicauda is Latin for “short tail,” indicating the relatively stubby tail that this species possesses.

Geographic distribution: Only known from Mt. Latimojong of the west-central area of endemism (fig. 1), between 2300 and 2500 m elevation (fig. 13). The surveys we conducted on Mt. Latimojong during 2011 between 2000 and 2300 m did not find this species (table 3).

Diagnosis: Crocidura brevicauda is a fairly large shrew (tables 2, 12) with a stocky build (fig. 17) and relatively short tail (fig. 9). The tail is somewhat thick and dark brown with many applied hairs. It has sparse, but long bristles along nearly its entire length (fig. 33A). The long dorsal fur (7–8 mm at middorsum) is gray-brown to dark brown, with a gray base and short brown tip. The fur coat is sufficiently thick and long that it somewhat obscures the ears. The hind feet are darker than the forefeet, the former being brown to dark brown dorsally, while the latter are pale brown. On all four feet, the thenar and hypothenar are less prominent than the four interdigital pads (fig. 33A). The skull width relative to length is midsized at the braincase and interorbital region (figs. 10, 34A). The rostrum is long relative to skull length (fig. 10). The lambdoidal ridge is prominent and the lateral margins of the braincase are somewhat angular, forming a dull point at the squamosal-parietal suture (fig. 34A).

Comparisons: Crocidura brevicauda is considerably larger than all members of the Small-Bodied Group, larger than C. australis and C. pallida of the Rhoditis Group, has a much shorter tail than members of the Long-Tailed Group, is slightly smaller and has a narrower braincase than C. rhoditis and C. pseudorhoditis of the Rhoditis Group, is darker in color (pelage and feet) than all members of the Rhoditis Group, and is larger than all members of the Ordinary Group (detailed below) except C. nigripes. Crocidura nigripes, however, is usually very darkly colored, particularly on its feet, whereas C. brevicauda has pale brown to dark brown feet (fig. 33). Crocidura brevicauda also has a somewhat stockier body (fig. 17), much wider interorbital region (fig. 10), more rounded braincase, and a less robust dentition (fig. 34A) than C. nigripes. Crocidura brevicauda has a more luxuriant pelage than all other species except C. musseri and C. caudicrassa, the other member of the Thick-Tailed Group. Crocidura caudicrassa is generally similar to C. brevicauda, but can be distinguished by its even thicker fur, heavier body (fig. 17), longer skull, wider rostrum and braincase, and thicker tail (table 12).

Comments: The extent of montane habitats on Mt. Latimojong is fairly limited and isolated from other such areas (fig. 1). As such, the geographic range of Crocidura brevicauda may be quite small. See the next account for details of phylogenetic relationships and coalescent species delimitation results.

Specimens examined: Mt. Latimojong (MVZ 237568, 237597, 237616, 237629–237631, 237633; MZB 43009).

Holotype: MZB 34795, an adult female collected on 28 October 2011 by K.C. Rowe and preserved as a cleaned skull (fig. 34B), fluid-preserved body (fig. 33B), and tissue sample (NMV Z21760). External measurements are 144 mm × 62 mm × 17 mm × 10 mm = 16.5g.

Type locality: Indonesia, West Sulawesi Province, Kabupaten Mamasa, Desa Tondok Bakaru, Kampung Rantepangko, Mt. Gandang Dewata, Post 3; 2.84534° S, 119.38216° E, 2580–2640 m elevation.

Etymology: Caudicrassa is Latin for “thick tail,” identifying this species' most distinctive trait.

Geographic distribution: Known only from the type locality around 2600 m on Mt. Gandang Dewata and from a single specimen collected by Luis Ruedas at 2120 m elevation from an area approximately 40 km ESE of Gandang Dewata in Rindingallo, of the west-central area of endemism (fig. 1; table 3).

Diagnosis: A large (tables 2, 12), stocky shrew (fig. 17) with chocolate dorsal pelage (more reddish brown in MSB 93104 from Rindingallo), slightly grayer ventral pelage, and an unusually thick tail (fig. 33B). The dorsal hairs are dark gray at the base, with the overall pelage color determined by the browner approximately 1 mm tip. The tips of ventral hairs are a paler brown. The tail is shorter than head-and-body length (fig. 9) and holds conspicuous bristles spread along nearly its entire length (fig. 33B). The bristles are pigmented proximally for roughly half their length in the series from Mt. Gandang Dewata, but only for approximately 1 mm on the specimen from Rindingallo (MSB 93104). The tail varies from pale brown to dark brown dorsally and is slightly paler on the ventral surface. The overall color of the tail is largely determined by the many applied hairs that cover the tail scales. The feet are similar in color to the tail, dorsally and ventrally and both the tail and feet are slightly paler than the dorsal pelage. The feet are small relative to body mass, but not body length (fig. 17). The interdigital pads are exceptionally prominent, but the thenar and hypothenar are not unusual (fig. 33B). The pelage is unusually thick, with hairs at the middorsum approximately 8–10 mm long. The skull is large and robust, with a broad interorbital region (figs. 10, 34B). The rostrum is long relative to skull length (fig. 10; table 12). From a dorsal view, the braincase appears narrow, largely from the effect of the broad interorbital region. In dorsal view, the anteriorly converging lines formed by the sinus canal, interorbital margin, and rostrum above the maxillary process are quite straight. In ventral view, the rostrum is narrow relative to the broad posterior palate. The molar row is robust (fig. 34B).

Comparisons: Crocidura caudicrassa is larger (as estimated from skull length) than most other species of shrew on Sulawesi. The exceptions are C. elongata and C. quasielongata of the Elongata Subgroup, C. nigripes of the Ordinary Group, and C. rhoditis of the Rhoditis Group. Among these relatively large animals, C. caudicrassa is much stockier than C. nigripes and all members of the Long-Tailed Group (fig. 17; table 2). Only C. rhoditis has a body form nearly as robust, but it is substantially paler in pelage and skin color, particularly on the feet, and has a longer hind foot and thinner tail than C. caudicrassa (fig. 9; table 2). Crocidura caudicrassa could be confused with C. nigripes, despite its stockier body. However, C. caudicrassa differs by having a wider interorbital region relative to both braincase breadth and skull length and a less robust dentition than C. nigripes. Crocidura caudicrassa could also be mistaken for C. brevicauda, the only other member of the Thick-Tailed Group, but the latter has a less stocky body form, less prominent interdigital pads, shorter skull length, narrower rostrum and interorbital region, lesser braincase height, and less elongate interorbital region (table 12). The most distinctive feature of C. caudicrassa is its thick tail, which has no equal among Sulawesi's Crocidura, but is approached by the somewhat thick tail of C. brevicauda (fig. 33). The pelage of C. caudicrassa is thick—only those of C. brevicauda and C. musseri, the latter of which is much smaller and does not have a particularly thick tail, are comparable.

Comments: We tested species limits between the two Thick-Tailed species using BPP on a small dataset containing 13 Crocidura caudicrassa and two C. brevicauda. Twelve of the C. caudicrassa specimens lack three of the five loci in this alignment, and thus it is only 54% complete. Nevertheless, BPP delimited these two species with posterior probability of 1.0 in all replicates.

The area around Mt. Gandang Dewata is an expansive region of montane habitat that is almost entirely unexplored by mammalogists. Although we describe Crocidura caudicrassa as a montane endemic, it may have a somewhat larger geographic range in this area. The single specimen from Rindingallo (∼40 km from Mt. Gandang Dewata) hints at this possibility.

We inferred a sister relationship between Crocidura caudicrassa and C. brevicauda in our analyses of UCEs (figs. 7, 8) and nuclear exons (supplementary data S6), but not in our analyses of mitochondrial DNA (figs. 4, 5). These two phenotypically similar species each appear to be montane endemics restricted to neighboring areas of high elevation, and thus we suspect that our UCE inferences reflect the correct relationship. Jukes-Cantor cytochrome b distances between these two species averaged 0.12, close to the mode of interspecific divergences among Sulawesi shrews and more than twice the median of intraspecific differences (fig. 6; supplementary data S4). Their morphological differences and presumed isolation by lowland habitats between Mts. Gandang Dewata and Latimojong further support their distinction.

Specimens examined: Mt Gandang Dewata (MZB 34792–34798, 34801–34805), Rindingallo, Tana Toraja (MSB 93104).

Holotype: USNM 217545, an adult male collected by H.C. Raven on 4 August 1916 and prepared as a skin and skull. External measurements from the holotype are 131 mm × 51 mm × 14 mm; ear length and weight were not recorded.

Type locality: “Temboan (southwest from Tondano Lake), northeastern Celebes” (Miller and Hollister, 1921: 101; fig. 1). We estimate Temboan is in the Southeast Minahasa Regency of North Sulawesi Province, 6 km south of Kalait, at 0.979° N, 124.605° E, 650 m. See the gazetteer (appendix) for a full justification of our interpretation of Raven's notes.

Geographic distribution: Crocidura nigripes is generally regarded as a Sulawesi endemic that occurs broadly across lowland areas of the island (fig. 1), but the species was tentatively reported from Obi Island by Fabre et al. (2018). This latter record is based solely on the similarity of cytochrome b sequences between C. nigripes and a single specimen from Obi; we have not examined the Obi specimen. On Sulawesi, we recorded this species from the north-east (Mt. Ambang, Temboan, and Lembeh, North Sulawesi Province), north-central (Toraut, North Sulawesi Province), north-west (Mt. Buliohuto, Gorontalo Province and Mt. Dako, Central Sulawesi Province), west-central (Mts. Balease, Rorekatimbo, and Torompupu, Toare, Tolai, and Sungai Miu, Central Sulawesi Province; Mt. Gandang Dewata (including our lowland sample area at Salu Tiwo), West Sulawesi Province), and east-central areas of endemism (Mt. Katopasa and Peleng Island, Central Sulawesi Province). Notably, we did not record the species from the south-west or southeast areas of endemism. Similarly, Musser (1987) reported the species only from the west-central area of endemism and northern peninsula. The full elevational extent of C. nigripes records across these areas extends from near sea level to just over 2000 m (fig. 13; table 3).

Emended diagnosis: A medium-sized shrew (tables 2, 14) with a medium build and uniformly dark pelage, feet, and tail (light chocolate to black; fig. 36). The tail is shorter than head-and-body length. The hind foot is somewhat short relative to head-and-body length (fig. 17), and usually dark brown to black. The skull is broad, and quite angular, with a prominent maxillary process, broad palate, and dentition with a large occlusal area relative to palatal area (fig. 37B). The interorbital region is unusually narrow, which enhances the prominence of the braincase breadth and maxillary process (fig. 10). Rostral length makes up a large proportion of skull length (fig. 10).

Comparisons: Crocidura nigripes is generally very dark in color with a short tail and moderate body size, but there is substantial variation in size and color. The feet are often nearly black, which will, in most cases, distinguish this species from all Sulawesi shrews except the darkest members of the Small-Bodied Group, which are much smaller. Typically, C. nigripes is smaller than C. elongata, C. quasielongata, C. caudicrassa, and C. rhoditis, comparable in size to C. microelongata, C. pseudorhoditis, C. brevicauda, and C. australis, and larger than the remaining species known from Sulawesi. It is the largest species in the Ordinary Group. These size differences can be observed using measures of skull length, head-and-body length, mass, or hind foot length (figs. 9, 17). Tail length is shorter than head-and-body length, which will distinguish C. nigripes from all similarly sized shrews (as judged by HBL, CIL, etc.) except the two Thick-Tailed species, which have greater relative hind-foot lengths (HF/HBL) and stockier bodies (figs. 9, 17). The skull of C. nigripes has a more angular braincase, more prominent maxillary process (enhanced by the narrow interorbital region), and more robust dentition (larger occlusal surface area) than all other Sulawesi shrews. The relative interorbital width (IOW/CIL) is substantially less than in all other species except C. elongata and C. quasielongata (fig. 10), two species with much longer skulls, tails, hind feet, and bodies.

Comments: Phylogenetically, Crocidura nigripes is more closely related to shrew species endemic to Sundaland than to any others that live on Sulawesi (Ruedi et al., 1998; Esselstyn et al., 2009, 2019; Hinckley et al. 2021). Our phylogenetic analyses conducted on far more species than previous studies reinforce this conclusion. We placed C. nigripes as sister to C. palawanensis, a species endemic to the Palawan Island group in the southwestern Philippines (figs. 7, 8). Hinckley et al. (2021) estimated C. nigripes as a close relative of C. foetida, a Bornean endemic that was not included in our analyses.

Miller and Hollister (1921) described two subspecies, Crocidura nigripes nigripes from Temboan and C. n. lipara from Gimpoe, in Central Sulawesi Province. Ruedi (1995) treated northern peninsula populations as C. n. nigripes and specimens from other parts of the island as C. n. lipara. He noted size differences with this geographic division, but also highlighted the limited sample of available specimens. We treat C. n. lipara as a synonym of C. nigripes because we have not identified any geographically partitioned morphological (fig. 38; tables 15, 16) or genetic diversity with our improved sampling. Jukes-Cantor cytochrome b distances between specimens sampled from across Sulawesi are <0.01 (fig. 4; supplementary data S3). As such, we see no reason to maintain the subspecific distinction between C. n. nigripes and C. n. lipara.

Specimens examined: Mt. Ambang (LSUMZ 39044, 39279, 39320), Mt. Balease (MZB 38491; FMNH 210593, 210595, 210611), Mt. Buliohuto (LSUMZ 38279–38281, 38283, 38286, 38287, 38292–38295, 38297; NMV C37778), Mt. Dako (LSUMZ 37017–37019, 37022, 37025–37029, 37032, 37035–37049; NMV C37257, C37265, C37307, C37318, C37368, C37369), Mt. Gandang Dewata (FMNH 218708), Gimpoe (=Gimpu) (USNM 219444, 219446), Lake Lindu (USNM 218668, 501223), Mt. Katopasa (LSUMZ 40925; NMV C40227, C40229, C40230), Koelawi (= Kulawi) (FMNH 47361), Kuala Navusu (AMNH 226105, 226106, 226108, 226109, 226110, 226112), Peleng Island (AMNH 109215), Pinedapa (USNM 219435), Mt. Rorekatimbo (FMNH 213190, 213191, 213252), Salu Tiwo (FMNH 218703, 218704, MZB 38488, 38489), Toraut (RMNH 38389 [= IZEA 4413], 38390 [= IZEA 4415]), Lembeh (FMNH 31844), Sungai Miu (AMNH 223989), Sungai Tolewonu, Tolai (AMNH 226537–226539), Temboan (FMNH 43859; USNM 217536, 217541–217545), Toare (= Tuare or Toware) (FMNH 47362, USNM 219434, 219437–219440, 219443, 219447), Mt. Torompupu (LSUMZ 39469–39474).

Holotype: MZB 16791 (= IZEA 4397), an adult female caught 18 August 1991. Measurements from the holotype are TC (head-and-body length): 66 mm × Q (tail length): 57 mm × PP (hind-foot length): 13.5 mm = Pds (weight): 6.5 g.

Type locality: Mt. Rorekatimbo, Central Sulawesi, at 1° 16′ S, 120° 15′ E, 2230 m elevation.

Geographic distribution: Apparently restricted to high elevation areas (>1800 m) in the northern portions of the west-central area of endemism east of the Palu-Koro Fault in Central Sulawesi Province (figs. 13, 39; table 3). Ruedi (1995) reported specimens of this species from only the type locality on Mt. Rorekatimbo. We collected specimens that match Ruedi's description and are mitochondrially identical (based on cytochrome b sequences from Ruedi et al., 1998) from the same mountain, at three nearby sites (1.3093° S, 120.3092 °E, 2020 m; 1.2926° S, 120.3064° E, 2180 m; 1.2884° S, 120.3104° E, 2250 m). Musser collected additional specimens from 1800–2300 m elevation on the neighboring Mt. Nokilalaki (ca. 15 km W of Mt. Rorekatimbo).

Emended diagnosis: A medium-sized shrew (tables 2, 14) with a thick, dark brown pelage that is only slightly paler on the venter. Foot, ear, tail, and rhinarium color closely match the pelage. Tail length is shorter than head-and-body length. The tail has inconspicuous applied hairs that are visible with the naked eye only on the distal third of tail length and sparse bristle hairs spread from the tail base to at least its midpoint (fig. 36A). The braincase is notably inflated, with great width, giving an overall wedge-shaped appearance when viewed from the dorsal aspect (fig. 37A). The squamosal-parietal suture is often open, leaving a tear-drop-shaped opening anteriorly, directly below the anterior opening of the sinus canal (sensu McDowell, 1958).

Comparisons: Crocidura musseri is smaller than members of the Elongata Subgroup and all Rhoditis Group members except C. pallida. It is larger than all members of the Small-Bodied Group. Within the Ordinary Group, C. musseri is smaller than C. nigripes, slightly larger than the sympatric C. normalis, and comparable to C. ordinaria and C. solita. Relative to head-and-body length, C. musseri has a shorter tail than all members of the Long-Tailed and Rhoditis groups, C. nigripes, and the sympatric C. solita. The pelage of C. musseri is thicker than that of any other shrew species we recognize from Sulawesi, except C. caudicrassa. Overall, the color of C. musseri is substantially darker than specimens from the Long-Tailed and Rhoditis groups, C. lea, and C. mediocris. Within the Ordinary Group, C. musseri is slightly darker than C. solita and C. normalis. Relative to skull length, skull width at both the braincase and interorbital region is greater than most other species, including all members of the Ordinary Group (figs. 10, 37). The unusual width of the skull (table 14) gives C. musseri a more wedge-shaped appearance than any other species when viewed from the dorsal aspect (fig. 37A). Crocidura nigripes has a more angular braincase, a more robust dentition, and a much narrower relative interorbital region (IOW/CIL). Relative rostral length (RL/CIL) in C. musseri is less than in C. nigripes, but greater than in C. normalis (fig. 10). Crocidura musseri is distinguished from all sympatric shrews by a combination of its short tail length, thick and dark pelage, and wide, wedge-shaped skull with small teeth (figs. 9, 36, 40).

Comments: Our description of Crocidura musseri is based primarily on the series of specimens we collected in 2011 very near the type locality. This series matches the holotype (MZB 16791) and the description of Ruedi (1995). Cytochrome b sequences from our series are identical to one from a paratype (IZEA 4403; Ruedi et al., 1998). None of our phylogenetic analyses supported a close relationship of C. musseri to any other species. Our mitogenome result placed C. musseri as sister to all other members of Sulawesi's endemic radiation with some support (fig. 5), while our UCE species tree placed it as sister to C. microelongata and C. pallida, but without statistical support (fig. 7).

Specimens examined: Mt. Nokilalaki (AMNH 223495, 225516, 225517, 225521, 225525–225527, 225530, 225531), Mt. Rorekatimbo (FMNH 213249–213251, 213253–213256, 213258–213269, 213442, 213443; MZB 16791).

Holotype: MZB 43010 (= FMNH 213437), an adult male, collected by J.A. Esselstyn on 27 March 2011. The specimen was prepared as a study skin, cleaned skull (fig. 41A) and skeleton, and frozen tissues. External measurements from the holotype are: 124 mm × 55 mm × 13 mm × 8 mm = 4.9 g. The voucher specimen and a tissue sample will be permanently curated at MZB, with an additional tissue sample retained at FMNH.

Type locality: Indonesia, Sulawesi Tengah, Poso, Huasa, Sedoa, Lore Lindu National Park, Mt. Rorekatimbo; 1.2884° S, 120.3104° E, 2250 m.

Etymology: Normalis is Latin for “normal,” used in recognition that this is yet another species of shrew with no striking phenotypic traits worthy of attaching a descriptive name.

Geographic distribution: We recorded this species at relatively high-elevation sites in Sulawesi's west-central (Mt. Gandang Dewata, West Sulawesi Province; Mts. Latimojong and Rorekatimbo, Central Sulawesi Province), east-central (Mt. Katopasa, Central Sulawesi Province) and south-east areas of endemism (Mt. Mekongga, Southeast Sulawesi Province; figs. 13, 39). We found this species at sites ranging from 1400 to 2600 m (fig. 13, table 3).

Diagnosis: Crocidura normalis is a medium-sized shrew (tables 2, 14), with a light build (fig. 17) and dark brown pelage, feet, lips, pinna, and tail. Dorsoventrally, the color is relatively uniform. The mystacial vibrissae are short relative to body length and darkly pigmented proximally for about half of their length. The tail can be as long as head-and-body length but is usually shorter (fig. 9). It is relatively densely covered with bristles, many of which are darkly pigmented from the base of the tail along approximately two-thirds of its length (fig. 40A). The small, applied hairs that typically cover the tail of Crocidura are less distinctly noticeable than in many other species. The feet are brown to black, usually uniformly so, including the digits and the thenar and hypothenar are subequal in area (fig. 40A). The claws are translucent. The skull is relatively gracile, being narrow across its entire length for a Crocidura of this size (fig. 10; table 14). Rostral length makes up a comparatively small proportion of skull length (fig. 10). The maxillary process is not prominent, the palate is narrow, and the occlusal surface area of the dentition is small relative to skull size (fig. 41A). The maxillary bridge is narrow.

Comparisons: Crocidura normalis is smaller than all members of the Long-Tailed, Thick-Tailed, and Rhoditis groups, and larger than all members of the Small-Bodied Group. As the smallest member of the Ordinary Group, C. normalis is only slightly larger than C. mediocris (the largest member of the Small-Bodied Group). It differs from C. mediocris in having a darker, denser pelage (individual hairs approximately 5 mm at middorsum) and longer tail both absolutely and relative to HBL (fig. 9). Within the Ordinary Group, C. musseri has a wider braincase relative to skull length, more robust body (fig. 17), longer rostrum relative to skull length (fig. 10), and paler feet than C. normalis. Crocidura nigripes has a short tail, like that of C. normalis, but is substantially larger, usually has darker feet, and the skull of C. nigripes shows a more prominent dentition, longer rostrum, and narrower interorbital region relative to skull length (fig. 10). Crocidura ordinaria and C. solita are slightly larger than C. normalis, have longer tails, both absolutely and relative to HBL, have hypothenar pads larger than thenar pads, and have greater relative rostral lengths (RL/CIL), braincase breadths (BB/CIL), and interorbital widths (IOW/CIL) (fig. 10).

Comments: In addition to the coalescent species delimitation analyses described above in the Crocidura parva account, we used BPP to test species limits between C. normalis and C. mediocris. The dataset consisted of 20 specimens of C. mediocris and 36 individuals of C. normalis. The sequence alignment is 93% complete. All replicates supported these species with 1.0 posterior probability. Unfortunately, we do not have nuclear exon data from the two specimens of C. normalis from Mt. Mekongga, which formed a clade independent of C. normalis from other localities in the cytochrome b and mitogenome gene trees (figs. 4, 5). In our UCE inferences, the Mekongga specimens of C. normalis are represented and the species is monophyletic in both our species tree and concatenated estimates (figs. 7, 8). If specimens from additional localities from the south-east and west-central areas of endemism are collected, reexamining these populations would be worthwhile. In both UCE estimates, C. normalis is a member of a clade comprising several Small-Bodied Group members and C. caudipilosa (figs. 7, 8).

Ruedi (1995) referred three specimens (IZEA 4393, 4394, 4042) to Crocidura lea and later Ruedi et al. (1998) sequenced a fragment of cytochrome b in one of these individuals (IZEA 4394). That sequence is identical to our C. normalis cytochrome b sequences from Mt. Rorekatimbo. Although we have not examined these three specimens, we presume they all represent C. normalis.

Specimens examined: Mt. Gandang Dewata (MZB 34809, 34810, 34812, 34821; FMNH 218604–218613, 218652–218658, 218973–218975), Mt. Katopasa (NMV Z61813), Mt. Latimojong (FMNH 213006–213008, 213010, 213011, 213013, 213014, 213030, 213032; MVZ 237595, 237622–237624), Mt. Mekongga (MWFB 8151, 8154), Mt. Rorekatimbo (FMNH 213174–213189, 213257, 213438; MZB 43010).

Holotype: MZB 43011 (= FMNH 218726), an adult male, collected on 5 May 2012 by J.A. Esselstyn. The specimen comprises a skull (fig. 41B), formalin-fixed body, and frozen tissues. External measurements from the holotype are 142 mm × 65 mm × 16 mm × 10 mm = 11.0 g. The voucher specimen and a tissue sample will be permanently curated at MZB and an additional tissue sample will be retained at FMNH.

Type locality: Indonesia, Sulawesi Barat, Mamasa, Mamasa, Tondok Bakaru, Rantepangko, Mt. Gandang Dewata; 2.8181° S, 119.3823° E, 2200 m elevation.

Etymology: Ordinaria is Latin for “ordinary,” used in recognition that this is yet another species of shrew with no striking or unique phenotypic traits worthy of a descriptive name.

Geographic distribution: Crocidura ordinaria is found in the west-central area of endemism (Mt. Gandang Dewata, West Sulawesi Province and Mt. Torompupu, Central Sulawesi Province) (fig. 39). The species spans an unusually broad elevational range from approximately 200 to 2600 m (fig. 13; table 3).

Diagnosis: A moderately sized shrew (tables 2, 14) with a somewhat stocky build. The pelage is medium to dark gray-brown and thick, with hairs at the middorsum typically 6–7 mm long. The venter is paler, with hairs that are pale gray at the tip, but dark gray at the base. In some specimens, the tips of some ventral hairs are reddish brown, giving the belly and chest pale cinnamon highlights. The mystacial vibrissae are short and darkly pigmented for at least half their lengths. Dorsally, the feet are nearly as dark as the pelage, but some specimens have much paler digits (fig. 40B). The tail is shorter than head and body (fig. 9) and the abundance of tail bristles and applied hairs is variable. The skull of this species is typical in its length (relative to body size) for a Sulawesi shrew but is wide at the braincase and interorbital region relative to skull length (fig. 10; table 14). The braincase, though broad, is not especially inflated vertically. The dentition is more prominent than expected given the palatal width (fig. 41B), but, otherwise, the skull of Crocidura ordinaria is unremarkable.

Comparisons: Crocidura ordinaria is a moderately sized member of the Ordinary Group. The tail is shorter than head-and-body length (fig. 9; table 2). Members of the Long-Tailed Group are larger and have much longer tails. Rhoditis Group members C. rhoditis and C. pseudorhoditis are larger. Crocidura pallida and C. australis, also members of the Rhoditis Group, are only slightly larger than C. ordinaria in head-and-body length. Compared to C. ordinaria, C. pallida has paler feet dorsally, a narrower relative braincase breadth (BB/CIL) and a narrower relative interorbital width (IOW/CIL) whereas C. australis has a narrower relative interorbital width, but wider relative braincase breadth than C. ordinaria (fig. 10). All members of the Small-Bodied Group are much smaller than C. ordinaria. Within the Ordinary Group, C. nigripes has darker feet and a relatively much narrower interorbital region and braincase than C. ordinaria. Compared to C. ordinaria, C. musseri has a thicker, darker pelage, and darker feet. Crocidura normalis is darker in color, smaller in body size, has more bristles on its darkly colored tail, and its skull is narrower with a shorter relative rostral length (RL/CIL) than C. ordinaria. Crocidura solita, another member of the Ordinary Group, is difficult to distinguish morphologically from C. ordinaria. External measurements are nearly identical between these two species (fig. 9), but C. ordinaria has a higher mass-to-length ratio (fig. 17). The pelage and feet of C. ordinaria are darker and have a smaller hypothenar (fig. 40), on average, than in C. solita. Cranially, C. ordinaria has a wider skull, observable in the absolute and relative breadths at the rostrum, interorbital region, and braincase (figs. 10, 42; table 14). Principal components analyses of external and cranial dimensions show that these two species overlap broadly in multivariate morphometric space, more so with external measurements than with cranial measurements (fig. 43; tables 17, 18).

Comments: See extensive comments detailing our decision to distinguish this species from Crocidura solita in the next species account.

Specimens examined: Mt. Gandang Dewata (FMNH 218727–218743, 218763–218767, 218769, 218771, 218772; MZB 34806, 34808, 34816, 34820, 34827, 34833, 34856, 34870, 34871, 34876, 34877, 34883, 43011; NMV Z21910, Z21911, Z21938), Salu Tiwo (FMNH 218744), Mt. Torompupu (LSUMZ 39480; NMV C40252, C40273).

Holotype: MZB 43012 (= FMNH 213044), an adult male collected 1 March 2011 by J.A. Esselstyn. The specimen consists of a cleaned skull, fluid preserved body, and frozen tissue samples. External measurements from the holotype are 147 mm × 68 mm × 16 mm × 9 mm = 9.5 g. The voucher specimen with a tissue sample will be permanently curated at MZB, with an additional tissue sample retained at FMNH.

Type locality: Indonesia; Sulawesi Selatan; Enrekang; Buntu Bato; Latimojong Village; Karangan; Mt. Latimojong, Bantanase; 3.40755° S, 120.0078° E, 2050 m.

Etymology: Solita is Latin for “usual,” used in recognition that this is another species of shrew with no striking phenotypic traits worthy of attaching a descriptive name.

Geographic distribution: We recorded Crocidura solita on three mountains in the west-central area of endemism (Mts. Latimojong and Rorekatimbo, Central Sulawesi Province; and Mt. Gandang Dewata, West Sulawesi Province; fig. 39). Records of this species span a broad range of elevations, from 700 to 2600 m. On Mt. Gandang Dewata, C. solita occurred syntopically with its sister species, C. ordinaria, at middle and high elevations (around 1600 and 2600 m; fig. 13; table 3).

Diagnosis: A medium-sized shrew (tables 2, 14) with a medium gray dorsal pelage and lighter gray ventral pelage. Tail length is slightly shorter than head-and-body length (fig. 9). The tail is only slightly bicolored, but some individuals have at least a few white applied hairs, giving it a somewhat silvery appearance, particularly near the tip. Dorsally, the feet are paler than the surrounding pelage, typically transitioning from light brown posteriorly to nearly white on the digits (fig. 40C). Ventrally, the feet are darker around the thenar and hypothenar pads than on the surrounding plantar and palmar surfaces; the hind feet tend to be darker than the forefeet. The external ears are slightly paler than the surrounding pelage. The mystacial vibrissae are short and mostly unpigmented, but a few of the longer, more posterior vibrissae are pigmented at their base. The skull is average in length (relative to HBL) for a Sulawesi shrew and slightly wide at the braincase and interorbital region relative to skull length (fig. 10). The braincase is rounded, but with a somewhat prominent lambdoidal crest (fig. 41C). The suture between the squamosal and parietal bones is often open, leaving a long slit below the sinus canal. The interorbital region is well tapered. The maxillary bridge is thin.

Comparisons: For comparisons to most species, see the Crocidura ordinaria comparisons section above. Compared to C. ordinaria, the pelage is less dense (i.e., shorter hairs on the dorsum) and both the pelage and feet are paler on average and the body is less stocky (fig. 17). The skull of C. solita is slightly smaller, and it is also narrower both absolutely and relative to skull length (figs. 10, 42), with a narrower palate and less robust dentition (table 14). The suture between the squamosal and parietal bones is often open, leaving a long slit in C. solita, but this suture is usually closed in C. ordinaria.

Comments: We posit that Crocidura solita and C. ordinaria are phenotypically similar sister species. Morphologically, we note subtle differences of body size, relative skull widths, and coloration. In the absence of other evidence, these slight morphological differences would be too little to justify dividing these specimens into two species, and we would not have suspected multiple taxa were represented by these specimens if it were not for our use of mitochondrial sequences. Syntopic specimens of these two species from Mt. Gandang Dewata are separated by a mitochondrial Jukes-Cantor distance of 0.064 (0.087 when all specimens are used in the calculation) and specimens of each species from the zone of parapatry is more closely related in the mitochondrial gene tree to conspecifics from other parts of the island (figs. 4, 5; supplementary data S2–S5). The mitochondrial distance between syntopic individuals and closer relatedness to allopatric populations suggests that these phenotypically similar animals are most likely evolving independently of each other.

In our mitochondrial gene trees, Crocidura ordinaria and C. solita are well-supported sister taxa and reciprocally monophyletic (figs. 4, 5). Our inferences using UCEs again indicate a close relationship with the two species forming a clade, but each species is reciprocally paraphyletic in our concatenated and species-tree inferences, the latter of which treated individuals as “species” and therefore did not force monophyly (figs. 7, 8). Our phylogenetic estimate from concatenated nuclear exons also inferred reciprocal paraphyly for this species pair (supplementary data S6).

We tested species delimitation models with BPP between Crocidura solita and C. ordinaria using a dataset that included 64 C. solita and 22 C. ordinaria sampled from across their ranges. The alignment of five exons is 82% complete. All analyses of this dataset supported species delimitation with 1.0 posterior probability.

Overall, we interpret the combination of subtle morphological differences (weak evidence), deep mitochondrial divergence in the face of syntopy (strong evidence), and our BPP results (modest evidence) as supporting species distinction. We conclude that the paraphyly in our UCE inferences is most likely the result of incomplete lineage sorting and ancestral polymorphism. In coming to this conclusion, we are admittedly leaning heavily on mitochondrial data. However, the great differences between the mitochondrial sequences where the two species cooccur provides strong evidence for differentiation, at least between females. If these specimens represented a single species, we presume that any substantive gene flow would lead to one of the divergent mitochondrial types being favored, either because it is independently superior to the other, or because it is superior in its integrative functioning with the relevant parts of the nuclear genome (sensu Sloan et al., 2017; Hill, 2019). Mitochondria are foundational to energy metabolism; viewing them as a neutrally evolving locus would ignore their functional importance and extensive interactions with nuclear gene products (Nabholz et al., 2008; Lane, 2011; Sloan et al., 2017; Tobler et al., 2019). Although many cases of mitochondrial divergence within species have been documented, they rarely involve sympatry of the divergent mitochondrial types (but see Wayne et al., 1990; Morgan-Richards et al., 2017), and far more often than not, mitochondrial divergence is consistent with species limits (Hebert et al., 2003; Monaghan et al., 2005; Esselstyn et al., 2012a; Cao et al., 2016). Taking in all the evidence, we believe these two sets of samples represent independently evolving populations that should be recognized as species despite the limited evidence for phenotypic or nDNA differentiation.

Specimens examined: Mt. Gandang Dewata (FMNH 218770, 218773–218780, 218976; MZB 34807, 34811, 34813, 34814, 34823, 34826, 34828, 34829, 34831, 34881, 34885, 34887), Mt. Latimojong (MZB 41669–41677; FMNH 213009, 213012, 213015–213019, 213029, 213031, 213033–213037, 213039–213043, 213045–213071; MVZ 237585–237593, 237596, 237598, 237599, 237611, 237617, 237619–237621, 237628, 237634, 238122; MZB 43012; NMV C38479, C38493, C38516), Mt. Rorekatimbo (FMNH 213163, 213194–213245, 213270), Rindingallo, Tana Toraja (MSB 93125).