Echiothrix centrosa Miller and Hollister, 1921

Musser, Guy G. & Durden, Lance A., 2014, Morphological And Geographic Definitions Of The Sulawesian Shrew Rats Echiothrix Leucura And E. Centrosa (Muridae, Murinae), And Description Of A New Species Of Sucking Louse (Phthiraptera: Anoplura), Bulletin of the American Museum of Natural History 2014 (391), pp. 1-87 : 34-47

Echiothrix centrosa Miller and Hollister, 1921

Echiothrix brevicula Miller and Hollister, 1921:67 .

HOLOTYPE: USNM 218706, the skin and skull of an adult male (original number 3077) collected January 9, 1917, by H.C. Raven. The skin was prepared as a conventional stuffed museum preparation; the skull and mandible are complete; all teeth are present. External, cranial, and dental measurements along with other data are listed in table 2.

TYPE LOCALITY: Winatu (01 ° 34 9 S, 119 ° 59 9 E; locality 7 in gazetteer and on the map in fig. 2 ), 762 m, in the core of the island, Propinsi Sulawesi Tengah, Indonesia .

EMENDED DIAGNOSIS: Echiothrix centrosa is similar to E. leucura in aspects of physical size as well as coloration and texture of fur but differs in other traits: (1) a shorter body and hind foot, but a longer tail in relation to length of head and body; (2) a smaller skull (as indexed by the lesser mean values for occipitonasal and rostral lengths; interorbital breadth; height and breadth of braincase; and lengths of diastema, bony palate, and postpalatal region), but longer incisive foramina, narrower rostrum, and wider bony palate and mesopterygoid fossa; (3) smaller molars with less complex cusp patterns: (4) cusp t3 occurs at a lower frequency on the first upper molar, a posterior cingulum is rarely present on the first and second upper molars: (5) an anterolabial cusp is absent from the second and third lower molars, posterior labial cusplets occur at low frequencies on the first and second lower molars but are absent from the third lower molar, and neither the first or second lower molar bears a posterior cingulum.

SPECIMENS EXAMINED: Total 39 (see Gazetteer and Specimens).

ETYMOLOGY: Miller and Hollister used centrosa to reference the provenances of their samples (Gimpu, Winatu, Besoa, and Tuare) from the central region of Sulawesi.

GEOGRAPHIC AND ELEVATIONAL DISTRIBUTIONS: Voucher specimens are from two localities on the northern peninsula and several places in the core of the island (see gazetteer and the map in fig. 2 ). Although collection sites are spotty, and expansive landscapes of Sulawesi are without records, the specimens at hand indicate that E. centrosa ranges throughout the northern peninsula west of the Gorontalo region (00 ° 31 9 N, 123 ° 03 9 E) and into the core of the island. All specimens come from the mainland between coastal lowlands and approximately 1000 m (see gazetteer), an interval that would be covered by tropical lowland evergreen rain forest.

The actual distribution of E. centrosa south of the northern peninsula is unknown. There are no records from the eastern, southeastern, or southwestern peninsulas. However, there are no reasons to think the species does not occur in tropical lowland evergreen rain forest on at least the eastern and southeastern peninsulas. Whether it also occurred on the southwestern peninsula before most of the lowland forest was removed and the landscape converted to agriculture (see Fraser and Henson, 1996; Whitten et al., 1987) or is present in remnant tracts of lowland forest (as described by Froehlich and Supriatna, 1996) is not known. Samples of subfossil murines excavated from caves in the southern part of the southwestern peninsula include species of Bunomys , Maxomys , Paruromys , Taeromys , Lenomys , and Rattus , but no Echiothrix ( Musser, 1984, also unpublished MSS.).

DESCRIPTION: Physical size and external features (color and texture of fur; lengths of head and body, hind foot, and ear; length of tail and its pigmentation pattern; and number of teats) of E. centrosa closely resemble those attributes shown by E. leucura (see description of that species). Descriptions of the skull, mandible, incisors, and molars for E. leucura also describe the basic shapes in these structures for E. centrosa except for the differences in cranial measurements and proportions along with size differences in molars and contrasts in their occlusal topographies, which are described below in the section covering comparisons.

Here we provide descriptions of a few characteristics of E. centrosa , primarily derived from material Musser collected, which are not currently available for E. leucura .

Testes and spermatozoal morphology: The testes are large relative to body size (16 % – 23 % of head and body length; table 14), the scrotal sac is densely haired, and the hairs behind the penis are stained by a brownishyellow exudate. There is no gross external sign of a midventral gland.

Gross morphology of the spermatozoa of Echiothrix centrosa has been described by Breed and Musser (1991; reported as E. leucura ). Basically, the asymmetrical sperm head is long and sickle shaped, the apical hook is short, ventral hooks are not present, and the tail is long (see the table of measurements and micrograph of the spermatozoa in Breed and Musser, 1991: 4 and 8, respectively).

Stomach morphology: Echiothrix centrosa has a unilocular-pouched stomach, that is, it combines aspects of what Carleton (1973) regards as the two basic morphological designs within muroid rodents—unilocular-hemiglandular or bilocular-discoglandular. Carleton (1973: 10) described a unilocularhemiglandular stomach as:

single-chambered with a shallow incisura angularis that scarcely extends beyond the esophageal opening … the corpus is spacious, with a broad fornix ventricularis. Distribution of cornified and glandular linings coincides closely with the basic stomach divisions: cornified epithelium is found in the corpus while glandular epithelium is limited mainly to the antrum. The bordering fold crosses the lesser curvature at the apex of the incisura angularis and the greater curvature at a locus opposite the incisura angularis.

This single-chambered hemiglandular morphology, in which the glandular zones are separated by a smooth bordering fold and the incisura angularis is shallow, forms the gastric conformation that Carleton (1973, 1980) suggested represents the primitive evolutionary state among muroid rodents. The general unilocular-hemiglandular design is common to Sulawesian species in which the diets are composed of insects and fruit ( Margaretamys ); fruit, small seeds ( Haeromys ); fruit, vegetative and flowering plant parts, insects ( Lenomys and Eropeplus ); small vertebrates, arthropods, snails, earthworms (oligochaetes), fruit, fungi ( Bunomys ); primarily fruit, some insects ( Taeromys ); and mostly fruit ( Rattus ). In these species, the extent of glandular epithelium lining the antrum relative to the area of cornified epithelium of the corpus resembles the pattern exemplified by Rattus hoffmanni as illustrated in figure 11 , with a range of variation in which the glandular portion does not extend beyond the level of the esophageal orifice to a configuration where the glandular lining penetrates the corpus well past the esophageal opening.

In Carleton’s view (1973: 10), a biloculardiscoglandular stomach has:

a deep incisura angularis that projects well past the esophageal opening, thereby imparting a more strongly defined bipartite condition …. The fornix ventricularis arches further beyond the esophageal orifice than in a unilocular stomach, and recurves slightly toward the esophagus. The size of the corpus, especially the fornix ventricularis section, varies greatly between individuals, the degree of distension corresponding to the amount of food contained. The walls of the antrum are conspicuously muscular, particularly near the pyloric orifice. The zone of glandular epithelium, surrounded by a distinct bordering fold, is restricted to a small disc-shaped area on the greater curvature. Cornified epithelium covers the remainder of the antrum and corpus.

A bilocular-discoglandular pattern is the stomach morphology common to Maxomys hellwaldii ( fig. 11 ) and other Sulawesian species in that genus. A spacious fornix ventricularis inclines toward the esophagus and projects craniad beyond the esophageal opening, the incisura angularis is deep, and the walls of the antrum are muscular. The corpus and most of the antrum is lined with muscular cornified squamous epithelium. Glandular epithelium is restricted to a small patch in the antrum on the ascending portion of the greater curvature; the glandular zone is bounded by a bordering fold. Fruit, arthropods, snails, earthworms, and small vertebrates are consumed by M. hellwaldii .

The stomach of Echiothrix centrosa ( fig. 11 ) is unilocular and pouched. ‘‘Unilocular’’ because the corpus and antrum are not separated by a bordering fold, which transforms the bulk of the stomach into a single chamber, and the incisura angularis is shallow (either even with the esophageal orifice or barely extending beyond it). ‘‘Pouched’’ because the entire corpus and antrum is lined with cornified squamous epithelium, and glandular epithelium occurs only in a pouch situated on the greater curvature within the lumen of the stomach opposite the esophageal orifice. This pouch is lined with thick glandular epithelium and connects with the main chamber of the stomach through a small aperture at the end of a funnel-shaped or tubular neck. (We are indebted to Michael Carleton [in litt., 2013] for identifying Echiothrix ’s stomach conformation as unilocular-pouched.)

The possible correlation of stomach design and diet will be explored below in the section covering natural history aspects of E. centrosa .

Karyotype: The chromosomal composition has been recorded for a male E. centrosa (AMNH 225682) collected at Kuala Navusu: 2N 5 40, FNa 5 72 (total of autosomal arms) and FNt 5 75 (total number of arms, including XY). Autosomes comprise 14 pairs of submetacentrics with the first pair the largest and the others ranging in size from moderately large to small, two pairs of small acrocentrics, and three pairs of small metacentrics; the presumed X chromosome is a small submetacentric, the Y a small acrocentric ( Musser, 1990: fig. 10 ; documented as E. leucura ).

Juveniles: All the information provided in the preceding descriptions of E. centrosa and E. leucura is derived from adults. We have seen only two juveniles, both E. centrosa : a skin and skull from Pinedapa (USNM 219737) and skull only from Tuare (USNM 219714). All molars are erupted and exhibit a degree of wear similar to the juvenile molars of USNM 219714 illustrated in figure 10 . The dorsal pelage is brown and the ventral coat off-white in the juvenile from Pinedapa; although we have not seen live or freshly caught juveniles, we suspect the skin is discolored and colors in the live rat were likely grayish brown and pure white. The feet and most of the tail are white (with slight greasy discoloration). Fur covering upperparts from shoulders to rump is spinous but softer, without the rigid texture of the adult coat. The spines are long, and about half the width of adult spines (but still flat, flexible, and grooved), so the curly and soft underfur layer forms more of the coat than it does in adults, which is why the fur feels less bristly to the touch. Legs and sides of the body are covered with softer fur, composed mainly of soft and curly underhairs through which are scattered flattened, long spines. Mask on the face (around each eye) and pigmentation pattern of the tail is like that of adults. Ears are dry and brown, but were probably rubbery and dark gray in life. The ventral fur is soft to the touch and composed of both thin hairs and thicker flat hairs; the latter are much narrower than those in the adult coat.

COMPARISONS: Miller and Hollister (1921: 67) diagnosed Echiothrix centrosa as:

Like Echiothrix leucura Gray , of North Celebes; but more grayish, less buffy, in color; with more cream-buff, less yellowish, under-parts and inner surfaces of limbs. Ears smaller; teeth smaller, the length of entire upper tooth row about equal to that of first and second molars of leucura .

Farther along in the description, Miller and Hollister remarked that the new species

is like Echiothrix leucura of North Celebes in size and proportions of the skin and skull; except that it has smaller ears and smaller teeth. In color it is conspicuously different from a

TABLE 6

Descriptive Statistics for Cranial and Dental Measurements (mm) Derived from Population Samples for Echiothrix leucura and E. centrosa

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

E. centrosa

All samples Northcentral Core

E. leucura Molinggapoto Bumbulan Kuala Navusu–Pinedapa Sadaunta–Besoa P value Variable N 5 16 N 5 30 N 5 3 N 5 1 N 5 19‘‘ brevicula ’’ N 5 7‘‘ centrosa ’’ (KN–P versus S–B)

ONL 55.0 ± 1.56 52.3 ± 1.36 52.3 ± 1.27 53.0 51.8 ± 1.26 53.4 ± 1.16 **

(51.7–57.8) (50.1–54.8) (51.4–53.7) (50.1–54.2) (51.5–54.8) ZB 23.5 ± 1.14 23.2 ± 0.80 22.6 ± 0.31 23.8 23.2 ± 0.78 23.5 ± 0.95 —

(21.5–25.2) (21.4–25.1) (22.3–22.9) (21.8–25.1) (21.4–24.3) IB 7.3 ± 0.26 6.8 ± 0.38 7.4 ± 0.17 7.3 6.6 ± 0.22 7.1 ± 0.33 **

(6.9–7.9) (6.2–7.7) (7.3–7.6) (6.2–7.0) (6.7–7.7) LR 22.4 ± 1.10 20.6 ± 0.87 20.9 ± 0.81 20.7 20.4 ± 0.76 21.2 ± 1.04 —

(20.0–24.0) (19.2–22.3) (20.3–21.8) (19.2–22.3) (19.9–22.3) BR 7.3 ± 0.46 7.5 ± 0.33 7.0 ± 0.00 7.1 7.5 ± 0.31 7.6 ± 0.28 —

(6.7–8.3) (6.8–8.1) (6.8–8.1) (7.2–8.0) BBC 19.4 ± 0.40 18.8 ± 0.37 18.4 ± 0.50 18.8 18.7 ± 0.31 19.1 ± 0.28 **

(18.5–20.0) (17.9–19.6) (17.9–18.9) (18.0–19.2) (18.8–19.6) HBC 14.8 ± 0.53 14.2 ± 0.36 13.9 ± 0.31 14.6 14.1 ± 0.30 14.5 ± 0.34 *

(13.5–15.5) (13.6–15.0) (13.2–14.2) (13.6–14.7) (14.1–15.0) BZP 3.5 ± 0.24 3.5 ± 0.27 3.6 ± 0.10 3.8 3.4 ± 0.22 3.6 ± 0.35 —

(3.1–4.0) (3.1–4.0) (3.5–3.7) (3.1–3.8) (3.1–4.0) LD 19.0 ± 0.99 18.5 ± 0.78 18.2 ± 0.85 18.5 18.2 ± 0.74 19.2 ± 0.73 *

(16.7–20.5) (17.3–20.1) (17.3–19.0) (17.3–19.9) (18.0–20.1) PPL 18.4 ± 0.95 18.1 ± 0.46 17.7 ± 0.25 18.3 18.1 ± 0.43 18.3 ± 0.59 —

(16.9–20.3) (17.0–19.2) (17.5–18.0) (17.0–19.0) (17.5–19.2) LBP 12.1 ± 0.69 10.1 ± 0.55 11.1 ± 0.15 10.4 9.9 ± 0.41 10.5 ± 0.36 **

(10.9–13.1) (9.0–11.2) (10.9–11.2) (9.0–10.8) (10.0–11.1) BBPM1 4.6 ± 0.53 4.9 ± 0.32 4.6 ± 0.58 5.0 4.9 ± 0.24 4.9 ± 0.43 —

(3.0–5.2) (4.2–5.3) (4.3–5.3) (4.4–5.3) (4.3–5.3) BMF 2.3 ± 0.18 2.5 ± 0.19 2.1 ± 0.12 2.5 2.6 ± 0.10 2.4 ± 0.12 **

(2.0–2.6) (2.0–2.8) (2.0–2.2) (2.4–2.8) (2.2–2.6) LIF 9.3 ± 0.65 9.7 ± 0.54 9.1 ± 0.71 9.4 9.7 ± 0.49 10.0 ± 0.51 —

(8.2–10.3) (8.5–10.7) (8.5–9.9) (8.8–10.7) (9.1–10.7) BIF 3.3 ± 0.33 3.3 ± 0.23 3.3 ± 0.15 3.4 3.2 ± 0.21 3.6 ± 0.17 **

(2.7–3.9) (2.8–3.7) (3.1–3.4) (2.8–3.5) (3.3–3.7)

)

S–B

value versus — — —

P KN–P

(

Core Sadaunta–Besoa

’’ N 7 centrosa 5 ‘‘ 0.11 ± 6.1 (–) 5.9 6.2 ± 0.17 6.4 (6.1 6.6 –) ± 0.12 2.1 2.3 – () 1.9

centrosa

Navusu–Pinedapa 5 ’’ brevicula

N

‘‘ 19 6.1 ± 0.15) – 6.3 (5.8 ± 6.4 0.22 () 5.9 – 6.7 ± 0.06 2.1 1.9 2.1 () –

E

.

Kuala

) TABLE 6 Continued (Bumbulan 1 N 5 5.9 6.2 2.1

Northcentral

Molinggapoto 5 3 N 5.8 ±

0.15 5.7 (6.0 –) ± 6.4 0.25 6.6) 6.3 – (± 0.00 2.1.

significant

samples 30 5 0.15 ± 6.3 –) ± 0.20 6.7) – ± 0.08 –) 2.3 5 not

All

N 6.0 (5.7 6.4 (5.9 2.1 (1.9 —;

0.05

#

leucura 16 5 ±

0.18 6.5) – ± 0.39 – 8.0) ± 0.10 2.5) – P

*

5

E

. N

6.1 (

5.8 7.3 (6.6 2.3 (2.2; 0.01

#

3 P

Variable LB ALM 1

– BM 1 ** 5 series of leucura from Temboan, North Celebes; the yellowish-buff tints of leucura are replaced by vinaceous-gray, especially noticeable on the flanks; and the belly is a whitish cream-buff rather than deep yellowish-buff.

In the following contrast between E. leucura and E. centrosa , the taxon brevicula , also described by Miller and Hollister and also from the central part of Sulawesi, is included as a synonym of E. centrosa . Reasons for doing so are set forth below in the section explaining allocation of the synonym.

In the samples at hand of E. centrosa and E. leucura , which includes the material examined by Miller and Hollister, and contrary to their observations, we see differences in size and proportions of external measurements. Mean values for length of head and body (206.0 mm for E. centrosa , 215.8 mm for E. leucura ; P 5 0.013) and length of hind foot (51.5 mm and 53.5; P 5 0.003) are significantly different— E. centrosa averages smaller. Means for length of tail (240.9 mm for E. centrosa , 240.2 mm for E. leucura ; P 5 0.877), however, are statistically identical—therefore, E. centrosa has on average a shorter body and hind foot but a longer tail relative to length of head and body (sample sizes are N 5 30 for E. centrosa and N 5 13 for E. leucura for all three variables).

Other external differences between E. centrosa and E. leucura that were described by Miller and Hollister are not apparent to us. Both species have dark gray to bluishgray dorsal coats; both have underparts that are either all white (in live animals) or white with patches or streaks of buff, orange, or pale rust (the rust may be a stain because the hairs in these colored places have pigmented debris adhering to them while hairs in the white region are clean); both show a similar color pattern on the tail, including length of the dorsal white portion (means of dorsal white length/tail length are 59.1 % for E. centrosa and 58.5 % for E. leucura ; table 5); and both possess two pairs of inguinal teats.

To Miller and Hollister, the smaller ears of centrosa —as well as brevicula , the other taxon they described from central Sula-wesi—were diagnostic in relation to E. leucura . But their descriptions of the new

TABLE 7 Results of Principal-Components and Discriminant-Function Analyses Comparing Samples of Echiothrix leucura and E. centrosa Correlations (loadings) of 16 cranial and 2 dental log-transformed variables are based on 16 E. leucura and 30 E. centrosa . See figure 12 .

taxa and comparisons were based on material obtained by H.C. Raven, who did not measure length of ear on freshly caught rats. Miller and Hollister simply measured ears (from the notch) on the dry museum skins; because such ears are shriveled and their lengths contracted, they recorded smaller values than would be obtained from freshly caught animals. They reported, for example, 29.4 mm for the holotype of centrosa and 28.1 mm for the holotype of brevicula . Musser also measured the ear (from notch to crown of pinna) on specimens he collected within the geographic ranges of centrosa and brevicula and did not encounter any ear length less than 32 mm, a value within the range of measurements obtained by G. Heinrich for a sample of E. leucura from Rurukan (table 4).

Except for its smaller ears and teeth, E. centrosa was considered by Miller and Hollister to be similar to E. leucura in cranial size and proportions. In addition to the contrast in molar size, (see below), there are—contrary to the perceptions of Miller and Hollister—impressive distinctions between the two species in certain cranial dimensions and proportions. Univariate mean differences in cranial variables show that, compared with E. leucura , the central Sulawesian E. centrosa has a shorter skull and narrower interorbit, a shorter but wider rostrum, smaller braincase, shorter diastema and postpalatal region, shorter but wider bony palate, wider mesopterygoid fossa, and longer incisive foramina (table 6).

The univariate mensural cranial and dental differences are also summarized by results from multivariate analyses ( fig. 12 ). There the top scatterplot contains the distribution of specimen scores for all population samples of E. centrosa and E.

TABLE 8

Descriptive Statistics for Measurements (mm) of Maxillary and Mandibular Molar and Alveolar Rows Derived from Samples of Echiothrix leucura and E. centrosa

Mean ± 1 SD, observed range in parentheses, and size of sample are listed.

leucura projected on the first and second principal components. Scores for specimens in population samples from regions west of Gorontalo in the northern peninsula (Molinggapoto, Bumbulan) and from the core of the island (Sadaunta-Besoa, and Kuala Navusu–Pinedapa) with small skulls and molars are scattered in the left half of the plot: these are samples of E. centrosa . Specimen scores for samples from the northeastern portion of the northern peninsula east of the Gorontalo region, animals with larger skulls and molars, fall in the right half of the ordination: these scores identify E. leucura . Covariation among most variables indicates size to be the primary influence for dispersion of the scores along the first component, as indicated by the many positive and high correlations on that axis (r 5 0.57–0.94). Four variables yielded moderate to high negative correlations (r 5 ‾ 0.19 to ‾ 0.72; table 7), which point to the broader rostrum, bony palatal bridge, and mesopterygoid fossa along with a longer incisive foramina relative to skull size (indexed by occipitonasal length) in E. centrosa compared with E. leucura . The patterns of significant loadings of cranial and dental measurements are generally mirrored by the univariate mean contrasts (table 6).

Results of discriminant-function analysis (which identifies the variables most strongly separating the centroids of predefined population samples) provide a sharper resolution of morphometric affinity among samples, but the pattern is generally similar to that illuminated by results of the principal-components analysis. Individual specimen scores projected on first and second canonical variates form two nonoverlapping constellations ( fig. 12 , bottom): scores for E. leucura , with the larger skulls and molars, fall in the left half; points representing E. centrosa , with smaller skulls and molars, are contained in the right half. Size is the force separating scores along the first variate as indicated by covariation among most of the same variables that proved significant in the principalcomponents analysis (table 7). Here, the analysis again shows that E. centrosa has a shorter skull and rostrum, narrower interorbit, smaller braincase, shorter bony palate, and smaller molars (as measured by alveolar length of maxillary molar row and breadth of first upper molar) (r 5 ‾ 0.51 to ‾ 0.91), but a wider rostrum, bony palate, and mesopterygoid fossa, coupled with a longer incisive foramina (r 5 0.28–0.62) compared with E. leucura .

Miller and Hollister noted the contrast in molar size between E. centrosa and E. leucura , and the difference is striking as shown by the univariate means for lengths of maxillary molar rows: 6.4 mm for E. centrosa and 7.0 mm for E. leucura (table 8). Furthermore, there is no overlap in the observed range of values. Mean differences in lengths of the mandibular molar rows is comparable (6.4 mm for E. centrosa and 7.1 mm for E. leucura ), although there is slight overlap in the observed range of measurements. The shorter maxillary and mandibular molar rows in E. centrosa are related to its smaller molars compared with those of E. leucura , as estimated by means for breadth of first upper molars (table 8).

Contrast in molar size is not the only dental distinction between the two species. The molars of E. centrosa also have less complex occlusal surfaces, as indicated by the

TABLE 9 Presence or Absence of Particular Cusps and Cusplets on Maxillary and Mandibular Molars in Samples from Echiothrix leucura and E. centrosa a

TABLE 9

(Continued)

E. leucura : northeastern arm of the northern peninsula (Rurukan, Tondano, and Temboan).

E. centrosa : northern peninsula (Molinggapoto and Bumbulan) and Core (Kuala Navusu, Pinedapa, Sungai Sadaunta, Winatu, Gimpu, Besoa, and Tuare).

None of the specimens examined showed an anterocentral cusp as part of the anteroconid, or anterior labial cusplets on any of the three mandibular molars.

difference between the two species in frequency of occurrence of particular cusps and cusplets (table 9, fig. 10 ). In the maxillary molars, cusp t3 occurs at a lower frequency on the second molar in E. centrosa (25 %) compared with E. leucura (45 %), and a posterior cingulum is present on the first and second molars in only 1 % of the sample of E. centrosa but present in 90 % of the sample of E. leucura . In the mandibular molars, an anterolabial cusp is not present on either the second or third molars in E. centrosa (occurs in 30 % of the sample of E. leucura ); none of the specimens of E. centrosa surveyed showed a posterior cingulum on the first and second molars (frequency is 92 % on the first molar and 70 % on the second molar in E. leucura ); and a posterior labial cusplet is present on the first molar in 46 % of the sample, on the second molar in 19 %, and not present on the third molar in the sample of E. centrosa (92 % on the first molar, 40 % on the second, and 20 % on the third molar are the frequencies for E. leucura ).

GEOGRAPHIC VARIATION: Except for the 19 specimens in the Kuala Navusu–Pinedapa sample, the other population samples are small. Three from Molinggapoto and one from Bumbulan are the only specimens we have examined from the northern peninsula west of the Gorontalo area, and the seven in Sadaunta-Besoa are the only specimens from middle elevations in central Sulawesi west of the lowlands in which Kuala Navusu and Pinedapa are located. Without larger samples from more places, especially from transects that would connect the current collection localities, we cannot provide a definitive pattern of geographic variation using external traits and measurements from cranial and dental variables. For now only a few comments covering the different population samples is possible.

The specimens from Molinggapoto and Bumbulan have white underparts without patches of pigmentation. Both all-white underparts and white ventral coats with patches or streaks of buff or rust occur in the samples from the core of Sulawesi. More specimens from the northern peninsula both east and west of Gorontalo are needed to determine the frequency of occurrence of allwhite versus partially pigmented underparts in the population there.

The pattern of covariation among cranial and dental variables as revealed by principalcomponents and discriminant-function analyses places the specimen from Bumbulan within or close to the cluster of specimen scores representing the Sadaunta-Besoa sample ( fig. 12 ). Its physical size and fur color are indistinguishable from specimens of comparable age in the Sadaunta-Besoa sample, at least those individuals with all-white underparts.

In the ordinations generated by multivariate analyses, scores for the Sadaunta-Besoa sample cluster close to those representing the Kuala Navusu–Pinedapa sample, both in principal components and canonical variate scatter plots ( fig. 12 ). There are some perceptible differences between the two samples. Animals from the lowlands at Kuala Navusu and Pinedapa have slightly shorter ventral coats and slightly smaller skulls than do those in the Sadaunta-Besoa sample, differences that will be elaborated upon in the section below where the reasons are given for treating brevicula as a synonym of E. centrosa . Again, the significance of these observations within the context of revealing geographic variation in external and cranial traits is elusive without samples from intervening regions between the two population samples and without samples from elsewhere in the core of the island and the various peninsulas.

The specimens from Molinggapoto are geographically closest to the range of E. leucura , but their cranial and dental morphometric attributes, along with molar occlusal cusp patterns, link them with the specimen from Bumbulan and those collected in the core of Sulawesi. However, there are a few morphometric differences, which are graphically summarized in the canonical variate scatterplot in figure 12 . There, along the first axis, the specimen scores for Molinggapoto are aligned in the right half of the ordination along with scores for the samples from Bumbulan and central Sulawesi, all of which we identify as E. centrosa ; scores for E. leucura , with larger skulls and molars, fall into the left half of the graph. But along the second axis, the points for Molinggapoto are separated from those representing the other examples of E. centrosa . Compared with that material, the Molinggapoto sample has a relatively wider interorbital region and zygomatic plate, narrower mesopterygoid fossa, and smaller bullae, as indicated by the high positive and negative correlations for these variables (table 7).

Future attempts to recover any patterns of geographic variation within what is identified as E. centrosa here will depend on study of additional material collected from a greater expanse of Sulawesi and analyses of molecular as well as morphometric data.

ALLOCATION OF SYNONYM: Echiothrix brevicula Miller and Hollister, 1921: 67 . The holotype is USNM 219744 , the skin and skull (measurements are listed in table 3) of an adult male collected January 29, 1918, by H.C. Raven. The type locality is Central Sulawesi, Pinedapa (01 ° 25 9 S, 120 ° 35 9 E; locality 4 in gazetteer and on the map in fig. 2 ), Propinsi Sulawesi Tengah, Indonesia .

The Latin brevis means ‘‘short’’ (also connotes ‘‘small’’ and ‘‘narrow’’), and Miller and Hollister used brevicula to reflect what they perceived to be the smaller size of the specimens in the sample from Pinedapa as compared with their sample of E. leucura from Temboan in the northeastern end of the northern peninsula and E. centrosa from the west-central highlands. Miller and Hollister (1921: 68) diagnosed Echiothrix brevicula as follows:

Differs from Echiothrix leucura and E. centrosa in smaller size; smaller hind feet; much more vinaceous, less buffy or yellowish, coloration; darker underparts, buff or reddish-buff rather than yellowish or whitish; and smaller, less narrowed skull. Ears and teeth small, as in centrosa .

The authors remarked that the new species ‘‘is easily separated from E. leucura and E. centrosa by its small size, peculiar coloration, and the less narrowed skull. It has small ears and small teeth as in E. centrosa .’’

Our sample of brevicula includes the type series Raven obtained at Pinedapa and the sample Musser collected from Kuala Navusu; both localities are in the coastal lowland region east of the west-central highlands (see gazetteer and the map in fig. 2 ). Specimens from each place are inseparable in physical size, color of fur, morphometric traits associated with the skull and molars, and molar occlusal patterns formed by cusps and cusplets. We subjected cranial and dental measurements from all the samples identified here as E. centrosa to principal-components analysis and on the resulting scatterplot (not illustrated here) the scores representing specimens from Pinedapa and Kuala Navusu intermingle to form a tight cluster. The two geographic samples were combined to form a single population sample (Kuala Navusu–Pinedapa) used to generate the multivariate analytical results illustrated in figure 12 .

Specimens that can be assigned to what Miller and Hollister would have identified as centrosa consist of the holotype from Winatu and collections from Tuare, Besoa, Gimpu, Kulawi, and Sungai Sadaunta, all localities in the west-central highlands west of the eastern coastal lowlands containing Kuala Navusu and Pinedapa (see the maps in Musser et al., 2010: 16, 81).

The type series of brevicula from Pinedapa does differ in magnitude of two external measurements from the series collected by Raven at Temboan, which is the sample Miller and Hollister used as their representative of leucura . Means for length of head and body (197.0 mm for brevicula , 215.4 mm for leucura ; P 5 0.003) and length of hind foot (50.3 mm and 54.7 mm; P 5 0.0001) are significantly different, but means for length of tail are not (239.6 mm and 250.0 mm; P 5 0.082). (Sample sizes are N 5 12 for brevicula from Pinedapa and N 5 7 for N. leucura from Temboan.) The Temboan sample has a larger body and longer hind foot than the series from Pinedapa, and the tail is shorter relative to length of head and body; this is the same pattern of differences seen between all samples of E. leucura and all samples of E. centrosa (including brevicula ) that we presented previously (see comparisons in the account of E. centrosa ).

In contrast, the distinction between brevicula and centrosa as noted by Miller and Hollister cannot be corroborated by us. We compared external measurements from the type series of brevicula (Pinedapa; N 5 12) with the specimens Miller and Hollister assigned to centrosa (Gimpu, Winatu, and Besoa; N 5 3). Means for length of head and body (197.0 mm for brevicula , 206.7 mm for centrosa ; P 5 0.37), length of tail (239.6 mm and 233.3 mm; P 5 0.80) and length of hind foot (50.3 mm and 51.7 mm; P 5 0.25) are not significantly different.

Specimens (all adults) that Musser collected from Sadaunta (representing centrosa ; N 5 6) and those he obtained from Kuala Navusu (representing brevicula ; N 5 7) are closely similar in physical size as indexed by means for length of head and body (215.3 mm for brevicula , 213.8 mm for centrosa ; P 5 0.70), length of tail (249.9 mm and 241.2 mm; P 5 0.25), and length of hind foot (52.3 mm and 53.2 mm; P 5 0.29), but the Sadaunta sample has slightly larger ears (33.3 mm and 34.8 mm), although the difference is not statistically significant (P 5 0.23); see also table 4. The sample from Kuala Navusu has a larger mean weight (267.3 g) than that from Sadaunta (237.6 g), just the opposite of expectation if brevicula were physically smaller as Miller and Hollister claimed.

Coloration and texture of the fur is similar in the samples of centrosa and brevicula . Both contain specimens with dark gray or dark bluish-gray dorsal coats and white ventral coats, with some individuals showing buffy or rusty streaks on the chest and abdomen. One difference is the shorter ventral fur on the specimens from Kuala Navusu and Pinedapa; all were taken between 31 and 122 m, specimens representing centrosa , with slightly softer and longer fur, are from 400– 985 m. Another difference is that the dorsal coat is somewhat softer in the specimens comprising the Sadaunta-Besoa sample compared with the slightly harsher and stiffer fur typical of animals in the Kuala Navusu–Pinedapa series.

It is unclear what Miller and Hollister saw when they described brevicula as having a ‘‘less narrowed skull’’ than either leucura or centrosa —perhaps they meant a skull less narrow, with the implication that it was wider than those in the samples of leucura and centrosa . Specimens in the samples from Pinedapa and Kuala Navusu, as well as those from landscapes to the west at higher elevations (the Sadaunta-Besoa population sample), do have appreciably smaller skulls compared with the northeastern peninsular E. leucura , as does all the material at hand from places west of the Gorontalo region ( figs. 5 , 6 ; table 6).

Restricting the focus to the core of Sulawesi, there are significant differences in univariate means between the Kuala Na-vusu–Pinedapa ( brevicula ) and SadauntaBesoa ( centrosa ) samples for eight cranial dimensions. Most means for the Kuala Navusu–Pinedapa sample average smaller, including interorbital breadth as well as breadth and height of braincase, three variables that could be used conceivably to index degree of cranial narrowness—in this case brevicula actually has a more narrowed skull, not a ‘‘less narrowed skull,’’ as contrasted with centrosa . Significant differences between means for other variables highlight the average longer skull in centrosa , its longer diastema and bony palate, wider incisive foramina, but narrower mesopteryTABLE 10 Results of Principal-Components Analyses Comparing Samples of Echiothrix ‘‘ centrosa ’’ (Sadaunta–Besoa) and E. ‘‘ brevicula ’’ (Kuala Navusu–Pinedapa) Correlations (loadings) of 16 cranial and 2 dental log-transformed variables are based on 16 E. leucura and 30 E. centrosa . See figure 13 .

goid fossa (table 6). For the other cranial measurements, as well as the dental variables, there are no significant differences in mean values between the Kuala Navusu–Pinedapa and Sadaunta-Besoa samples. Keep in mind the dissimilar sample sizes involved (N 5 19 for Kuala Navusu–Pinedapa, N 5 7 for Sadaunta-Besoa), each containing a range in age from young-to-old adults.

The mensural differences noted here are reflected in a principal-components ordination where most of the scores for SadauntaBesoa congregate to the right along the first axis of the scatter plot but overlap those signifying the Kuala Navusu–Pinedapa sample to the left ( fig. 13 ), a pattern influenced by the moderate to large positive correlations for 13 of the 18 variables (r 5 0.39–0.83; table 10) a measure of the average larger means for some measurements in the Sadaunta-Besoa series. Whether these univariate mean distinctions seen in samples at hand measure real geographic variation in the two populations will have to be determined by study of larger samples equal in number of specimens and equal in age classes. Nevertheless, present evidence from phenetic traits associated with skins, skulls, and teeth persuades us that the sample from Kuala Navusu and Pinedapa, corresponding to what Miller and Hollister described as E. brevicula (originally based on the sample from Pinedapa), represents a coastal lowland population of E. centrosa with slightly shorter ventral fur, harsher dorsal coat, and some cranial dimensions that average smaller. This specific homogeneity is reinforced by sucking lice ( Anoplura ): hosts in the Kuala Navusu–Pinedapa sample as well as those in the west-central highlands (Sadaunta-Besoa sample) are parasitized by the same species of louse, Polyplax beaucournui , n. sp. (see the section describing ectoparasites of Echiothrix ).

Figures

Fig. 2. (above) The island of Sulawesi and environs, showing Sulawesi’s four peninsulas and its ‘‘core’’ or central portion. Boundaries separating regions of endemism discussed here are indicated by dashed lines: SM-D 5 lowland reaches of the Sungai Onggak Mongondaw and Sungai Onggak Dumoga; SB-DL 5 lowlands of the Sungai Bone and Danau Limboto; TD 5 Tempe Depression. (opposite page) Collection localities for samples of Echiothrix leucura and E. centrosa. Numbers key to localities described in the gazetteer. Note the range of E. leucura, which is east of SM-D.

Fig. 11. Ventral views of the unilocular-hemiglandular stomach (in midfrontal section) of Rattus hoffmanni (AMNH 226041) and the bilocular-discoglandular stomach of Maxomys hellwaldii (AMNH 224939) contrasted with the unilocular-pouched stomach of Echiothrix centrosa (AMNH 225043). Abbreviations: a, antrum; bf, bordering fold; c, corpus; ce, cornified squamous epithelium; d, anterior end of duodenum; e, posterior end of esophagus; fv, fornix ventricularis; ge, glandular epithelium; gp, glandular pouch; ia, incisura angularis; mc, muscular walls of the corpus; p, pylorus; ps, pyloric sphincter. Scale lines 5 10 mm.

Fig. 10. Occlusal views of left maxillary (top row) and mandibular (bottom row) molar rows in four specimens of Echiothrix (lingual margin of each row is on the left, labial on the right). A, young adult E. leucura (USNM 217807, Temboan; CLM1–3 and clm1–3 5 7.5 mm). The next three drawings portray wear stages in E. centrosa: B, juvenile (USNM 219719, Tuare; CLM1–3 5 6.4 mm, clm1–3 5 6.6 mm); C, young adult (AMNH 225680, Kuala Navusu; CLM1–3 and clm1–3 5 6.5 mm); D, adult (AMNH 225683, Kuala Navusu; CLM1–3 5 6.5 mm, clm1–3 5 6.4 mm).

Fig. 12. Scatterplots of specimen scores representing the sample of Echiothrix leucura (filled circle; N 5 16) and the four population samples of E. centrosa (N 5 30; Molinggapoto 5 asterisk, Bumbulan 5 cross, Kuala Navusu–Pinedapa 5 square, Sadaunta-Besoa 5 triangle) projected on first and second principal components extracted from principal-components analysis (upper), and on first and second canonical variates extracted from discriminant-function analysis (lower) of 16 cranial and two dental logtransformed variables. Scores for holotypes: Echiothrix centrosa (filled triangle, Winatu in Sadaunta-Besoa sample) and Echiothrix brevicula (filled square, Pinedapa in Kuala Navusu–Pinedapa sample). See table 7 for correlations (loadings) of variables with extracted components or canonical variates and for percent variance explained in both ordinations.

Fig. 5. Cranium and left dentary of Echiothrix leucura (AMNH 101246; ONL 5 56.6 mm, CLM1–3 5

Fig. 6. Cranium and left dentary of Echiothrix centrosa (AMNH 225043; ONL 5 52.5 mm, CLM1–3

Fig. 13. Scatterplot of specimen scores representing the sample of Echiothrix ‘‘centrosa’’ (Sadaunta- Besoa; asterisk; N 5 7) and the two population samples of E. ‘‘brevicula’’ (Kuala Navusu–Pinedapa; filled circle 5 Kuala Navusu, empty square 5 Pinedapa; N 5 19) projected on first and second principal components extracted from principal-components analysis of 16 cranial and two dental log-transformed variables. See table 10 for correlations (loadings) of variables with extracted components and for percent variance explained.