PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON 119(4):477-515. 2006. The status of Ototnys or estes dollmani Heller, 1912 (: Otomyinae), a described from the Mathews Range, central

Michael D. Carleton* and Ellen Schaefer Byrne (MDC) Department of Vertebrate Zoology, National Museum of Natural History, Smithsonian Institution, Washington, D.C. 20560-0108, U.S.A., e-mail: [email protected]; (ESB) 4209 Lawn Avenue, Western Springs, Illinois 60558, U.S.A.

Abstract.•Morphological and morphometric comparisons of orestes dollmani, known only from Mount Gargues in the Mathews Range, central Kenya, indicate that the taxon is a valid distinct from other African forms under which it has been previously synonymized (O. irroratus, O. orestes, O. tropicalis). Based also on these comparisons, the morphological recognition and distribution of O. orestes (including thomasi Osgood) are further clarified in relation to O. tropicalis (including elgonis Wroughton) in Kenyan and O. typus proper in ; another taxon relegated to junior synonymy within O. typus, O. uzungwensis Lawrence and Loveridge from south-central , is resurrected to species. Certain traditional characters used in Otomys , in particular molar lamination, demonstrate conservative patterns of variation that complement spatial structure derived from morphometric analyses of craniometric data and that vindicate their continued utility in delimiting species. We argue that uncritical emphasis of polytypic species, applied following the biological species concept during the latter 1900s, has led to chronic underestimation of species diversity of Otomys confined to the Afromontane Biotic Region in eastern , in particular those populations that inhabit afroalpine environments.

Vlei or laminate-toothed rats of the Honacki et al. 1982, Corbet & Hill 1980, genus Otomys (: Muridae) con- 1986, 1991; Musser & Carleton 1993). stitute a distinctive morphological and Musser & Carleton (2005) recently ac- phylogenetic radiation indigenous to the knowledged 23 species. In the most savannas and highlands within Subsa- considered revisionary study attempted haran Africa (Bohmann 1952, Carleton & to date, Bohmann (1952) recognized only Musser 1984, Watts & Baverstock 1995, 11 species, most of them containing from Ducroz et al. 2001, Taylor et al. 2004a). 3 to 5 well-marked subspecies and the Since the circumscription of the group as exceptionally polymorphic O. irroratus a suprageneric taxon (Otomyinae Thom- embracing 23 geographic races. Much of as, 1897), the number of Otomys {sensu the indecision over specific diversity has lato) considered to be valid species has involved populations distributed across vacillated greatly, ranging from 30 (Eller- eastern African mountains north of the man 1941) to as few as 8 (Petter 1982) but Zambezi River•from the Nyika Plateau usually around 12-14 (Misonne 1974, and Eastern Arc Mountains, through the ranges and volcanoes fronting the West- ern and Eastern Valleys, to the * Corresponding author. Ethiopian Highlands. Within this region, 478 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON named forms have been variously and especially the last emerging M3. This inconsistently treated as subspecies of O. junction marks the border of the glossy irroratus, O. tropicalis, or O. typus. white enamel of the crown and dull white The present study focuses on one such cementum of the root and is visible as geographically localized population, Ot- a faint longitudinal line along the base of omys or es tes dollmani Heller (1912), de- the tooth (in lateral view). Those individ- scribed from Mount Gargues in the uals whose molars were incompletely Mathews Range, a small and relatively erupted, such that the line of the crown- low-elevation range lying to the north of root junction was visible on only the M1 in central Kenya (Fig. 1). or Ml•2, were categorized as juveniles. Following its description, dollmani has Adult specimens were identified by the been sequentially referred to O. orestes visibility on all molars of the crown-root (Dollman 1915, Allen 1939, Ellerman line, which initiaUy appears approximate- 1941), O. irroratus (Bohmann 1952), or ly even with the maxillary alveoli (young O. tropicalis (Misonne 1974, Musser & adults); by this growth stage, the occlusal Carleton 1993). Early on, however. Hoh- surface of the molar row approximates lster (1919:148) had pointedly raised a horizontal plane. With advancing wear, dollmani to species until "actual inter- the crown-root line is progressively dis- grading specimens are found," a pragmat- placed above the plane of the maxillary ic viewpoint endorsed by Musser & bone, crown height is correspondingly Carleton (2005). Here, we consolidate diminished, and dentinal cores of the morphological and morphometric evi- individual laminae progressively expand dence that decidedly supports the last in width and length (full and old adults). interpretation. Our juvenile tooth-wear stage essentially concurs with age classes 1 and 2 as Materials and Methods defined by Taylor & Kumirai (2001), and the young, full, and old adult wear Specimens reported herein consist prin- stages to their age classes 3, 4, and 5, cipally of skins with their associated respectively. skulls and are contained in the following Specimens of Otomys, especially those North American museum collections: originating from older historically impor- Carnegie Museum of Natural History, tant surveys, complicate study of their Pittsburgh (CM); Field Museum of Nat- craniometric variation because a substan- ural History, Chicago (FMNH); Museum tial percentage of skulls encountered in of Comparative Zoology, Harvard Uni- museum collections are damaged. Com- versity (MCZ); and the National Museum monly, nasal bones are dislodged, zygo- of Natural History, Smithsonian Institu- matic arches broken, or the braincase tion, Washington D.C. (USNM, formerly fractured. Hence, we have taken a robust U.S. National Museum). approach to our definition of operational Relative age was crudely assessed for taxonomic units (OTUs), identifying an- every specimen examined based on tooth- alytical samples that correspond to moun- wear stages, employing a combination of tains or ranges instead of single molar eruption and degree of molar localities and precise altitudes. Further, abrasion to identify four categories (juve- rather than eliminate those variables niles, young adult, full adult, old adult). taken on frequently broken cranial parts In view of the hypsodont molars charac- in order to amplify specimen numbers for teristic of Otomys, we used the appear- multivariate analyses, we have elected to ance of the basal crown-root junction to accept smaller sample sizes to capture gauge the degree of molar eruption. more fully size and shape differences. VOLUME 119, NUMBER 4 479

Cherangani jacksotíl^ ' íES (1891) • ' 1 ^ Ó ^ % ^ dollmani <%, • (1912)

^^x L'nubilus Mt Kenya \ '1-3' (1915)

malleus f thomasi ^ * (1915) .¿^t (1910) troplcalis percivali (1902) tk (1915) •^ '--'"^ •^ • squalus ^•^ (1915)

> 2500 m 2000-2500 m j 1500-200Öm

Fig. 1. Mountainous region of eastern Africa inhabited by populations of Otoinys addressed herein, indicating type localities of relevant taxa and principal geographic features.

Only length of rostrum, a measurement nomial species-group taxa of O. tropicalis that involves the nasal tips, was conven- are intended only to convey current tionally omitted in morphometric com- synonymy, not to denote any accepted parisons. subspecific classification. Full provenance Twelve OTUs, their two-letter abbrevia- and museum registration numbers are tions, and sample sizes of crania measured provided either in the Taxonomic Sum- were identified as follows. Provisional mary or in Appendix 1. Abbreviations are species taxonomy observes the arrange- used here and throughout the text for ment of Musser & Carleton (2005); tri- mount or mountains (Mt, Mts). 480 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

Otomys dollmani: Kenya, Mt Cargues tween the lateral edge of the upper lobe of (MG, n = 6, including the holotype the exoccipital condyles; length of ros- USNM 181790). trum (LR), an oblique line measured from Otomys irroratus: South Africa, East- the posterior bevel of the right zygomatic ern Cape Province (EC, n = 21); Kwa- notch to the end of the nasal bones at Zulu-Natal Province (KZ, n = 17); their midsagittal junction; breadth of Western Cape Province (WC, n = 32). rostrum (BR), caliper points on the nasal Otomys orestes: Kenya, Aberdare Mts process of the premaxillaries just inside and Mt Kenya (OR, n =10). the dorsal zygomatic notch; postpalatal Otomys tropicalis elgonis: Kenya, Aber- length (PPL), from the anteriormost bevel dare Mts (AM, n = 24); Uasin Gishu of the mesopterygoid fossa to the mid- Plateau (UG, n = 7); Kaimosi and notch of the basioccipital; length of bony Kakamega (KA, n = 21); Kenya and palate (LBP), from the anteriormost bevel , Mt Elgon (ME, n = 19). of the mesopterygoid fossa to the poste- Otomys tropicalis tropicalis: Kenya, Mt rior end of the left incisive foramen; post Kenya (MK, n = 28). palatal breadth (PPB), distance across the Otomys typus: Ethiopia, all localities maxillary bones just behind the third (ET, n = 14). molars; length of incisive foramen (LIE), Otomys uzungwensis: Tanzania, Ud- greatest anterior-posterior expanse mea- zungwa Mts (UZ, n = 14). sured on the left side; length of diastema Sixteen cranial and three dental vari- (LD), from the inside curvature of the left ables were measured to 0.01 mm, using upper incisor to the crown-root junction hand-held digital calipers while viewing (about the gum line) on anterior face of crania under a stereomicroscope. These the left Ml; breadth of zygomatic plate measurements, and their abbreviations as (BZP), from the rear edge of the right used in text and tables, are: condylo- zygomatic plate to its anterior edge, incisive length of skull (CIL), from the usually near the middle of the plate where occipital condyles to the medial projec- it is vertical; length of auditory bulla tion of bone (gnathix process) between (LAB), from the posteriormost curve of the incisors (an overall length of skull left ectotympanic bulla to the notch of the necessitated by the high incidence of eustacian tube (at the seam of the opaque Otomys encountered with missing or bone of the eustacian tube and translucent dislodged nasal bones); greatest zygomat- bone of the ectotympanic); depth of ic breadth (ZB), expanse across the auditory bulla (DAB), an oblique dimen- zygomatic arches, usually taken at the sion measured from the dorsal rim of the junction of the jugal bone and zygomatic auditory meatus to the ventralmost curve process of the squamosal; breadth of of the ectotympanic bulla; coronal length braincase (BBC), measured on the par- of maxillary toothrow (CLM), from the ietals, with the caliper points placed posterior edge of M3 to the crown-root midway between the glenoid fossa and junction on the anterior face of Ml, lambdoidal ridge and just above the usually at the posterior deflection of the squamosal flanges behind the zygomatic crown; width of the upper first molar arches; depth of skull at the middle orbit (WMl), distance across the edges of the (DBO), vertical distance between the middle lamina (t4-t6); breadth across lateral ledge of the maxillary above the upper incisors (Bis), outside caliper points anterior lamina of M3 and the dorsal on lateral surface of incisors near their surpaorbital crest of the frontal; least tips. In addition, the number of laminae interorbital breadth (lOB); breadth across composing the upper third molar (M3) occipital condyles (BOC), distance be- was scored for each cranium measured. VOLUME 119, NUMBER 4 481

Four external dimensions (to nearest dividually reference the upper (maxillary) whole mm) were transcribed from skin and lower (dentary) molars, respectively. tags as given by the collector: total length (TOTL), head and body length (HBL), Results and Comparisons tail length (TL), and hindfoot length (HFL). Most of the specimens we exam- In a sample of O. tropicalis from Mt ined originated from the early expeditions Kenya (n = 27), the contribution of to British East Africa (e.g., see Hollister secondary sexual dimorphism to within- 1919), and collectors of those specimens sample variation is statistically negligible, (E. Heller, J. A. Loring, E. A. Mearns) at least given the unbalanced nature of recorded a head and body length, instead analyzable material usually consohdated of total length, and measured the hind from museum collections. None of the 19 foot without claw. Thus, TOTL as craniodental measurements yielded ap- reported in Table 2 was usually obtained preciable differences according to sex by adding HBL and TL. Means and (Table 1). Taylor et al. (1993) reported ranges of external variables are provided that male O. irroratus average shghtly as a general index of size and bodily larger than females in most cranial proportions but were not subjected to dimensions, but that differences were morphometric comparisons. generally not significant. Accordingly, Standard descriptive statistics (mean, we combined the sexes in composing range, and standard deviation) were de- OTU samples for multi-group discrimi- rived for adult specimens (young, full, nant function analysis and in executing and old age classes combined) of the 12 principal component analyses. OTUs. Discriminant function classifica- Size variation within Otomys samples, tion, and principal component scores on the other hand, is readily apparent were computed using only the 19 cranio- from superficial inspection of crania and dental variables, all of which were first their impressive range in gross size, form, transformed to natural logarithms follow- rugosity, and definition of sutures. While ing the rationale developed by Bookstein some of this size range may be attribut- et al. (1985:23-26), Marcus (1990:80), and able to individual variation, much of it Bookstein (1991:101-102). Principal com- certainly reflects age-related size increase ponents (PC) were extracted from the (post-weaning growth) and correlates variance-covariance matrix, and variable strongly with the three adult age classes loadings are expressed as Pearson prod- based on tooth-wear. Many measure- uct-moment correlation coefficients of the ments thus display linear, incremental derived components or canonical variâtes increases in average size across the three (CV) with the original cranial measure- age classes, especially extremal dimen- ments. One-way analyses of variance sions (CIL, ZB) and those measured on (ANOVAs), apphed to sex or age as the facial region (LR, BR, LIE, LD); group effect, utilized the raw, non-trans- correspondingly, age-class as categorical formed measurements, and Bonferroni's effect in analyses of variance contributes post-hoc multiple comparison test, based significantly to non-geographic variation on Student's t statistic, was employed to for the majority of variables (Table 1). In isolate between-group significant differ- nearly all of those age-sensitive variables, ences. All analytical procedures were the significant differences reside between implemented using statistical packages the young adult and full adult or old adult contained in Systat (Version 10.2.01, age classes, but not between the full adult 2002). Throughout the text, we use the and old adult age classes (according to abbreviations Ml•3 or ml-m3 to in- Bonferroni post-hoc multiple comparison 482 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

Table 1.•Arithmetic means of craniodental variables and results of one-way ANOVAs for sex and age cohorts in a sample of adult Otomys tropicalLs from Mt Kenya (n = 27). See Materials and Methods for variable abbreviations.

Sex Age M F Y A o Variable (14) (13) F(sex) (9) (14) (2) fXnge)

CIL 36.9 36.8 0.2 35.2 37.7 38.6 12.5*** ZB 19.1 19.1 0.4 18.8 19.4 • 2.1 BBC 15.6 15.2 1.8 14.9 15.7 15.4 5.2* lOB 4.3 4.2 2.0 4.3 4.2 4.1 1.9 BOC 8.3 8.3 2.3 8.2 8.3 8.6 1.6 DSO 11.9 11.9 0.0 11.3 12.2 12.7 ^ 1 y*** LR 13.2 13.3 0.4 12.7 13.5 14.0 7.3** BR 7.3 7.3 0.1 7.1 7.4 7.6 3.8* PPL 13.7 13.7 0.0 13.1 14.0 14.7 10.2** LBP 9.7 9.8 1.1 9.4 10.0 9.9 8.6** LIF 7.0 6.8 1.5 6.5 7.1 7.1 6.2** LD 9.7 9.5 0.1 8.7 10.2 10.3 18.8*** BZP 6.2 6.2 1.6 5.9 6.3 6.4 6.4** PPB 6.9 7.0 0.3 6.8 6.9 7.5 3.3 LAB 7.5 7.4 0.5 7.3 7.5 7.8 2.4 DAB 7.7 7.7 1.5 7.5 7.8 8.2 4.9* CLM 9.0 9.0 0.4 9.1 9.0 8.7 1.7 WMl 2.5 2.5 2.5 2.5 2.5 2.5 0.6 Bis 4.3 4.4 0.7 4.1 4.5 4.8 11.0**

* = P < 0.05; ** = ^ P s 0.01; *** = P s 0.001. tests). Such a pattern is predicated by taxonomically informative (also see Dis- sample sizes available for the three age cussion and accompanying Table 9). Fur- groups and corresponding mean differ- thermore, three of the dimensions that ences. Presumably, the under-representa- demonstrated no age variation (lOB, tion of the old adult age-class in this OTU CLM, WMl) proved to contribute fre- diminished the number of dimensions quently and strongly to among-sample that potentially exhibit significant age differentiation in nearly all ordinations effects, as well as lowered the magnitude performed. of F values. Dimensions of the molars In view of the many taxa addressed, (CLM, WMl), which decrease in crown three sets of morphometric comparisons height with occlusal abrasion after erup- were conducted, emphasizing both the tion, do not grow in length and width, past taxonomic assignments of dollmani unhke osseus components of the skull; in and different geographic scopes: (1) doll- contrast, the incisors increase in girth mani in relation to O. irroratus and O. with age (Table 1). Other than procedural tropicalis sensu lato, the taxa in which elimination of juvenile specimens, we did dollmani has been commonly synony- not adjust for size in the following mized; (2) dollmani vis-à-vis other taxa morphometric analyses. Although varia- found north of the Zambezi River (elgo- tion introduced by post-weaning growth nis, or est es, tropicalis, typus), in particular may be substantial within samples of highlighting its divergence with respect to Otomys, it is typically negligible relative orestes, the taxon under which dollmani to those extracted factors that emerged as was originally named; and (3) three VOLUME 119, NUMBER 4 483

Table 2.•External and cranial statistics (mean, ±1 SD, and observed range) for samples of Otomys doUmani (Kenya, Mt Cargues, n = 6), O. orestes (Kenya, both Mt Kenya, « = 3, and Aberdare Mts, n = 7), O. tropicalis (Kenya, Mt Kenya, n = 29), and O. irroratus (South Africa, Western Cape Province, « = 32). See Materials and Methods for variable abbreviations.

Variable O. dollmani O. orestes O. tropicalis O. irroratus

TOTL 236.7 ±11.7 232.7 ± 13.0 253.1 ± 16.1 275.5 ± 20.2 225-255 212-252 217-284 232-307 HBL 144.7 ± 5.9 159.6 ± 10.6 169.2 ± 12.3 171.0 ± 12.5 138-150 135-175 137-192 150-192 TL 92.0 ± 7.2 73.1 ± 10.2 83.9 ± 5.8 104.5 ± 10.2 85-105 61-93 70-96 81-120 HFL 25.1 ± 1.3 28.5 ± 1.6 31.9 ± 1.1 32.5 ± 1.4 23-27 25-30 29-35 30-35 CIL 33.3 ± 1.5 35.5 ± 1.5 36.8 ± 1.7 38.4 ± 2.1 31.1-35.1 33.1-38.3 34.1-39.9 34.7^3.4 ZB 17.3 ± 0.8 19.2 ± 1.0 19.2 ± 0.6 19.8 ± 0.9 16.2-17.9 16.8-20.5 18.2-20.1 18.0-22.3 BBC 13.6 ± 0.3 14.6 ± 0.5 15.4 ± 0.6 15.0 ± 0.3 13.4-14.1 14.0-15.5 14.5-16.4 14.3-15.5 lOB 4.1 ± 0.1 3.9 ± 0.3 4.3 ± 0.2 4.5 ± 0.2 4.0-4.3 3.6-1.4 3.9^.6 4.1^.9 BOC 7.5 ± 0.3 7.7 ± 0.2 8.3 ± 0.3 8.3 ± 0.3 7.2-7.9 7.4-8.0 7.7-9.1 7.8-9.3 DSO 10.5 ± 0.5 12.4 ± 0.7 11.9 ± 0.7 12.6 ± 0.8 9.8-10.9 11.4-13.6 10.7-13.5 11.2-14.5 LR 12.1 ± 0.7 13.3 ± 0.7 13.3 ± 0.6 14.2 ± 0.9 11.2-13.1 12.5-14.7 12.4-14.6 12.7-16.4 BR 6.9 ± 0.3 6.8 ± 0.3 7.3 ± 0.3 7.5 ± 0.4 6.6-7.3 6.3-7.3 6.9-7.9 6.4-8.5 PPL 12.5 ± 0.7 13.5 ± 0.7 13.7 ± 0.8 14.9 ± 0.9 11.5-13.2 12.3-14.6 12.4-15.0 13.5-16.9 LBP 8.6 ± 0.3 9.2 ± 0.5 9.7 ± 0.4 9.9 ± 0.5 8.2-9.0 8.6-10.3 8.6-10.5 9.0-11.5 LIF 6.1 ± 0.5 6.9 ± 0.4 6.9 ± 0.4 7.4 ± 0.5 5.4-6.7 6.3-7.5 6.2-7.8 6.5-8.2 LD 8.2 ± 0.7 8.8 ± 0.6 9.6 ± 0.9 9.6 ± 0.9 7.1-8.9 8.3-9.6 7.9-11.6 7.7-11.5 BZP 5.3 ± 0.3 5.4 ± 0.5 6.2 ± 0.4 6.1 ± 0.5 4.9-5.5 4.6-6.0 5.4-6.8 5.1-7.0 PPB 6.5 ± 0.2 7.0 ± 0.2 6.9 ± 0.3 7.3 ± 0.3 6.3-6.9 6.6-7.4 6.4-7.7 6.8-7.8 LAB 6.3 ± 0.3 7.8 ± 0.3 7.4 ± 0.3 7.2 ± 0.4 5.9-6.9 7.3-8.5 6.8-8.1 6.5-8.0 DAB 6.9 ± 0.1 8.2 ± 0.3 7.7 ± 0.3 7.9 ± 0.2 6.7-7.0 7.9-8.7 7.2-8.3 7.3-8.3 CLM 8.07 ± 0.17 8.68 ± 0.21 9.00 ± 0.26 9.31 ± 0.24 7.86-8.34 8.38-8.97 8.46-9.50 8.76-9.81 WMl 2.09 ± 0.05 2.32 ± 0.11 2.52 ± 0.10 2.35 ± 0.11 2.03-2.18 2.21-2.55 2.37-2.67 2.10-2.57 Bis 3.98 ± 0.20 3.70 ± 0.30 4.32 ± 0.29 4.15 ± 0.36 3.72^.29 3.32^.25 3.68^.90 3.30-5.01 disjunct highland taxa {orestes, typus, considered valid species (e.g., Taylor & uzungwensis) conventionally grouped Kumirai 2001, Musser & Carleton 2005); within O. typus. Otomys irroratus, O. hence, their inclusion in deliberate com- tropicalis, and O. typus are currently binations in the multivariate analyses 484 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON provided a comparative yardstick against equally markedly divergent from those of which to assess the craniometric differen- irroratus and tropicalis-elgonis, more so tiation of dollmani. Further, we routinely than are the latter taxa from one another employed both discriminant function and (Fig. 2, bottom). principal component analyses to explore Principal component analysis limited to patterns of craniodental differentation for specimens of irroratus and those of each set. The latter ordination method tropicalis-elgonis further corroborates summarizes covariation patterns based fundamental shape differences between on the log-transformed specimen data these taxa (Fig. 3). All variables load itself without prior assumptions about positively on the first component and group assignment. Together, the two most correlations are moderate to strong procedures lend different perspectives to (r = 0.55-0.97; Table 4). Fewer variables variable covariation and complement one load heavily on PC II, more or less another in supporting taxonomic conclu- orthogonal to the major axes of the taxon sions. samples, and these emphasize consistent Morphometric variation.•In discrimi- contrasts in shape: the longer (CLM) but nant function analysis of the 9 OTUs narrower (WMl) molar rows in irroratus, representing dollmani and irroratus-tropi- its relatively smaller ectotympanic bullae calis (including elgonis), well-defined phe- (LAB, DAB), and broader interorbital netic structure is retained for most of the region (lOB) compared with examples of species-group epithets represented (Fig. 2, tropicalis-elgonis (Table 4). The propor- top). The exceptions are tropicalis from tionally smaller bullae and wider inter- Mt Kenya and OTUs of elgonis, which orbit of O. irroratus reinforce the mor- overlap substantially in multivariate phometric findings obtained by Taylor & space and among which a posteriori Kumirai (2001). Several of these variables misclassifícations are commonplace (12- are the same as those highlighted on CV 1 47%). Similarly, the three samples of derived from the above 9-group discrim- irroratus lack unique identity, their CV inant function analysis (CLM, WMl, scores broadly overlapping (29-36% mis- lOB), but the lack of prior group defini- classifications among them), but collec- tions and removal of dollmani specimens tively they are set well apart from divulge other non age-related shape dif- examples of elgonis and tropicalis on ferences, in this instance the relatively CVl. Specimens of irroratus average greater inflation of the bullae in tropicalis- larger than those of tropicalis-elgonis in elgonis (LAB, DAB). Although scores of most variables (Table 2), an overall size irroratus and tropicalis-elgonis marginally disparity that accounts for some of the overlap on the first two components separation, but the most heavily weighted extracted, the major axes of their elliptical variables on CVl involve the longer and spreads are phenetically distinct (Fig. 3), broader palate in irroratus (LBP, PPB), their Y-intercepts being significantly dif- its longer basicranial axis (PPL) and ferent {F = 22.7, P < 0.0001) while their deeper skull (DSO), longer but narrower slopes are statistically inseparable {F = molar rows (CLM, WMl), and broader 0.23, P = 0.629). Juveniles were omitted interorbit (lOB) (Table 3). The conspicu- from our analyses, but age-related size ous isolation of dollmani issues from its increases (post-weaning growth) still ac- uniformly smaller size, as indicated by count for much of the elongate scatter variable correlations with CV2, especially observed within the species samples and its more delicate molars (CLM, WMl) for the oblique orientation of the group and smaller auditory bullae (LAB, DAB) constellations on the first and second (Tables 2, 3). The sample of dollmani is principal components. Accordingly, age- VOLUME 119, NUMBER 4 485

dollmani

elgonis irroratus > O

tropicalis

CV1 (50.1%)

- MG dollmani

- ME elgonis

- KA elgonis

- UG elgonis

- AM elgonis

- MK tropicalis

- WC irroratus

- EC irroratus

- KZ irroratus

70 MahalanobisD Fig. 2. Results of discriminant function analysis performed on 18 log-transformed craniodental variables as measured on 93 intact specimens representing 9 OTUs of the Otomys irroratus-tropicalis complex. Top, scatter plot of specimen scores on first two canonical variâtes (CV) extracted; irregular polygons enclose maximal dispersion of specimen scores around an OTU centroid (lettered symbols) and species-group taxa are indicated next to the bounding polygons (to avoid visual congestion, polygons surround maximal dispersion for all specimens of the taxa elgonis and irroratus and only OTU centroids are indicated). Bottom, cluster diagram (UPGMA) based on average Mahalanobis distances between OTU centroids; species-group epithets are indicated for each terminal OTU. See Table 3 for variable correlations and percent variance explained. 486 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

Table 3.•Results of 9-group discriminant func- north of the Zambezi River {elgonis, tion analysis performed on intact specimens (n = 93) orestes, tropicalis, typus) again reveals representing dollmani and taxa of the irroratus- clearly defined, non-overlapping structure tropicalis complex and using 18 log-transformed among most taxa (Fig. 4, top). OTUs of craniodental variables (excluding LR). See Fig. 2 and Materials and Methods for variable abbrevia- tropicalis and elgonis are extensively tions. congruent on the first two canonical variâtes extracted (cumulative variation Correlations explained = 76.6 %) and are jointly Variable CVl CV2 differentiated from samples of dollmani, OIL 0.35* -0.51*** orestes, and typus. Morphological diver- ZB 0.34* -0.51*** gence between tropicalis and examples of BBC -0.12 -0.62*** elgonis, as depicted in both the CVA lOB 0.46*** -0.25 BOC 0.10 -0.74*** scatter plot and UPGMA phenogram, is DSO 0.57*** -0.44** unremarkable and questions taxonomic BR 0.36* -0.34* segregation of the two. Instead, the PPL 0.38** -0.49*** linkage pattern (Fig. 4, bottom), accord- LBP 0.47*** -0.45*** ing to inter-centroid Mahalanobis dis- LIF 0.33 -0.45*** LD 0.22 -0.30 tances among OTUs, preserves some BZP 0.36* -0.33 geographic structure, insofar as those PPB 0.43** -0.48*** OTUs drawn from mountains to the east LAB -0.29 -0.62*** of the Rift Valley (AM, MK) cluster DAB 0.03 -0.60*** apart from those to the west of the Rift CLM 0.46*** -0.71*** WMl -0.53*** -0.57*** Valley (KA, ME, UG). Although skulls Bis 0.13 -0.11 of orestes and typus approximate those of Canonical correlations 0.94 0.86 tropicalis-elgonis in many cranial dimen- Eigenvalues 7.01 2.95 sions (Table 2), certain pronounced shape Percent variance 50.1 21.1 differences contribute to their separation * = P < 0.05; on CV 1, namely the interorbital con- ** = P s 0.01; striction (lOB), inflation of the auditory *** = P s 0.001. bullae (LAB, DAB), and depth of the cranial vault (DSO) compared with OTUs class as a categorical effect in one-way of tropicalis-elgonis (Table 5). Such broad analyses of variance contributes signifi- similarities in shape are reflected in the cantly to the spread of specimen scores last clustering division, in which orestes- along PC I (i^ = 34.3, P < 0.0001) but not typus form a pair-group apart from PC II (F = 0.18, P = 0.911); whereas, dollmani and tropicalis-elgonis. Systema- species as effect contributes substantially tists such as Bohmann (1952) and Mis- to dispersion on both PC I (F = 13.9, P < onne (1974) have arranged orestes as 0.001) and especially on PC II (F = 114.1, a subspecies of typus, but the morpho- P < 0.0001). The elongate constellations metric footprints of the two taxa are of specimen scores thus revealed and their distinctive, orestes being an overall smal- obhque orientation to PC I and II ler than typus. According to conform with the now familiar results a posteriori probabilities of group mem- obtained in multivariate comparisons of bership, the holotype (FMNH 16698) of other closely related (congeneric) muroid thomasi Osgood (1910) was classified with species (Carleton & Martinez 1991, Voss the OTU of orestes (P = 0.999). Once & Marcus 1992). more, small size overridingly influences Discriminant function analysis of doll- the morphometric segregation of dollmani mani and other forms that occur to the along CV 2 (Table 5), particularly as VOLUME 119, NUMBER 4 487

3

A o. îropicalis Y = 0,36X+ 0.91 F = 18.5, P< 0.001

CO ö

O. irroratus Y = 0.29X - 0.83 F = 7.5, P < 0,01

I

PC I (56.0%) Fig. 3. Scatter plot of specimen scores on the first two principal components (PC) extracted from ordination of 18 log-transformed craniodental variables as measured on 89 intact specimens representing samples of Otoniys irroratus and tropicalis-elgonis. See Table 4 for variable correlations and percent variance explained. expressed in its more delicate molar rows first component extracted (most variable (CLM, WMl). As discerned by Hollister correlations positive and large•Table 6; (1919), the skull oí dollmani in fact recalls ANOVA (age) highly significant•F = a small version of tropicalis-elgonis and 29.9, P < 0.0001). Variable correlations differs from orestes and typus in the same with the second component convey shape proportional traits as listed above. The contrasts in select anatomical regions: the phenetic association of dollmani with the narrower interorbital constriction (lOB) tropicalis-elgonis group, albeit at a large in orestes, its slimmer molars and incisors Mahalanobis distance (Fig. 4, bottom), is (WMl, Bis), narrower zygomatic plate consistent with Hollister's observation (BZP), and relatively larger bullae (LAB, and contradicts Heller's (1912) original DAB), compared with individuals of arrangement of the taxon as a subspecies tropicalis-elgonis (Table 6). A visual land- of orestes. mark approximately concords with the Confined only to the large-bodied univariate difference in expanse of the Kenyan samples, i.e., omitting dollmani zygomatic plate: the anterodorsal border and typus, principal component analysis of the plate in orestes is more or less reaffirms the phenetic intermingling of coincident with the premaxillary-maxil- specimens of îropicalis and elgonis and lary suture, but in tropicalis-elgonis, the suggests their joint discrimination from lead edge of the plate notably overlaps examples of orestes (Fig. 5). Again, post- and obscures the middle portion of the weaning size increases are captured on the suture. Taxon as effect is significantly 488 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

Table 4.•Results of principal component analy- function ordinations. In principal compo- sis performed on 89 intact specimens representing nents analysis. The Udzungwa Mts form examples of Otomys irroratus and O. tropicalis emerges as overall smaller than the (including elgonis) and using 18 log-transformed northern taxa but with a relatively craniodental variables (excluding LR). See Fig. 3 and Materials and Methods for variable abbrevia- broader rostrum and incisors and smaller tions. auditory buUae (Fig. 6, top; Table 7). Lawrence and Loveridge (1953:61) re- Correlations marked on the small skull and relatively Variable PCI PCll heavy rostrum of uzungwensis in their OIL 0.97*** -0.00 description. Specimens of orestes are ZB 0.93*** -0.05 wholly segregated from those of typus BBC 0.55*** 0.23 and uzungwensis on PC III (not figured) lOB 0.13 -0.70*** BOC 0.62*** 0.05 based on their relatively and absolutely DSO 0.94*** -0.21 smaller molars (Table 7). Given the un- BR 0.87*** -0.15 ambiguous deUneation of the three taxa in PPL 0.90*** -0.04 scatter-plots of principal components, 3- LBP 0.84*** -0.24 group discriminant function analysis LIP 0.89*** -0.07 LD 0.93*** 0.13 yields even clearer segregation, grading BZP 0.83*** -0.18 incrementally in craniodental size along PPB 0.66*** -0.16 CV 1 (Fig 6, bottom). Narrower incisors LAB 0.54*** 0.63*** (Bis) and smaller auditory bullae (DAB) DAB 0.72*** 0.43** largely account for divergence of exam- CLM 0.35* -0.51*** WMl 0.01 0.49*** ples of orestes on CV 2 (Table 7). Bis 0.84*** 0.17 Molar lamination and incisor sulca- Eigenvalues 10.1 1.9 tion.•Counts of laminae on the upper Percent variance 56.0 10.3 third molar (Table 8) vary intelligibly by * = P < 0.05; taxon and generally concur with patterns ** = P s 0.01; already documented (Bohmann 1952, *** = P s 0.001. Dollman 1915, Taylor & Kumirai 2001). The modal number of M3 laminae ranges reflected in the dispersion of PC II scores from 6 in dollmani and irroratus (in- {F = 68.3, P < 0.0001), but age-class is frequently 7); to 7 in elgonis, tropicalis, not (F = 0.5, P = 0.708). Scatter of the and uzungwensis (infrequently 6 or 8 in PC scores, patterns of variable loadings, each taxon); to 8 in typus (infrequently 7 and ANOVA effects are reminiscent of or 9). Within an OTU or species-group the PCA results obtained with the larger taxon, the modal number is pronounced samples of irroratus and tropicalis-elgonis and the range of variation typically spans (see Fig. 3), but the sparse representation only two laminae, exceptionally three of orestes erodes so clear a definition of (e.g., the elgonis sample from Mt Elgon). dual species constellations. The holotype A notable exception to laminar stability of thomasi falls nearer the margin of the involves the series of orestes from Mt orestes distribution (Fig. 5). Kenya, the area of its type locality. Specimens oi orestes, typus, and uzung- Thomas (1900) diagnosed orestes based, wensis, three disjunct highland taxa con- in part, on the possession of an M3 with ventionally viewed as races of a single only 6 laminae (citing 4 specimens, which species, O. typus (Bohmann 1952, Mis- we have not seen), in contrast to a 7- onne 1974), are easily distinguishable in laminated M3 that characterizes tropicalis multivariate space, whether derived from from the middle-upper slopes of Mt principal component or discriminant Kenya (Thomas 1902). In reporting the VOLUME 119, NUMBER 4 489

8

CM > O

o 2 CV1 (52.1%)

ET typus

OR Orestes

MG dollmani

AM elgonis

MK tropicalis

UG elgonis

ME elgonis

KA elgonis

100 60 40 20 Mahalanobis D Fig. 4. Results of discriminant function analysis performed on 18 log-transformed craniodental variables as measured on 64 intact specimens representing 8 OTUs of Otomys distributed north of the Zambezi River (dollmani, elgonis, orestes, tropicalis, typus). Top, scatter plot of specimen scores on first two canonical variâtes (CV) extracted; irregular polygons enclose maximal dispersion of specimen scores around an OTU centroid (lettered symbols), and species-group taxa are indicated next to the bounding polygons (to avoid visual congestion, the polygon surrounds maximal dispersion for all specimens of elgonis and only OTU centroids are indicated). Bottom, cluster diagram (UPGMA) based on average Mahalanobis distances between OTU centroids; species-group epithets are indicated for each terminal OTU. See Table 5 for variable correlations and percent variance explained. 490 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

Table 5.•Results of 8-group discriminant func- M3 laminae (FMNH 43444, 43446; tion analysis performed on intact specimens (n = 64) USNM 164308, 164333, 164334, 164357, representing dollmani and other taxa distributed 165334) and 6 have 7 laminae (USNM north of the Zambezi River {elgonis, orestes, 164293, 164295, 164297, 164304, 164329, tropicalis, typus) and using 18 log-transformed craniodental variables (excluding LR). See Fig. 4 164332). We cannot fully explain the and Materials and Methods for variable abbrevia- discrepancy between HoUister's and our tions. tallies of M3 laminae. We note that 8 of the 13 specimens are juveniles with un- Correlations worn M3s, and in many of these, a tiny Variable CVl CV2 seventh lamina is set below the occlusal GIL 0.30 -0.43* surface. Although visually obscure, it ZB 0.39 -0.37 seems evident that initiation of wear in BBC -0.15 -0.42* lOB -0.62*** -0.30 these individuals would have produced BOG 0.03 -0.75*** a 7-laminated M3, but perhaps Hollister DSO 0.68*** -0.22 scored their M3s as having 6 laminae. BR 0.09 -0.43* Regardless of the explanation, the notion PPL 0.25 -0.30 of 6 M3 laminae as narrowly diagnostic LBP 0.37 -0.55*** LIF 0.38 -0.32 of orestes became erroneously entrenched LD 0.22 -0.33 in the literature and influenced the BZP 0.04 -0.57*** taxonomic arrangements of subsequent PPB 0.43* -0.25 authors (Dollman 1915, Bohmann 1952, LAB 0.61*** -0.40 Setzer 1953), including the description of DAB 0.59*** -0.15 CLM 0.45** -0.79*** dollmani (which possesses only 6 laminae WMl 0.22 -0.89*** in all 7 specimens examined) as a sub- Bis -0.17 -0.36 species of O. orestes (Heller 1912). Canonical correlations 0.96 0.93 Next to molar lamination, the presence Eigenvalues 12.7 6.0 and pronouncement of longitudinal Percent variance 52.1 24.5 grooves (sulci) on the upper and lower * = P < 0.05; incisors have figured prominently in the ** = P s 0.01; taxonomy of Otomys (Wroughton 1906, *** = P s 0.001. Dollman 1915, Thomas 1918, Bohmann 1952). Unlike the variation in upper large series of Otomys recovered by the incisor grooving observed among south- Smithsonian African Expedition from Mt ern African forms, some of which wholly Kenya, Hollister (1919:147) reinforced lack sulci (e.g., littledalei), the notion that a 6-laminated M3 is eastern African populations considered diagnostic of O. orestes•"The last upper here are uniform in possessing a deep molar has six laminae in all specimens lateral groove and shallow medial groove. except one (No. 164329) in which there Development of grooves on the lower are distinctly seven, the last very small." incisor, however, exhibits appreciable Hollister referred 16 USNM specimens to variation among the species-group taxa orestes, leading Bohmann (1952) to as- surveyed. sume that he had meant "1 of 16 speci- As reported by many authors, two mens," but Hollister did not specify how longitudinal grooves mark the anterior many skulls were available for examina- face of the lower incisor in O. typus sensu tion. We have managed to reassemble 13 stricto (Dollman 1915, Bohman 1952, specimens with skulls from Mt Kenya Taylor & Kumirai 2001). The sulci are that were probably studied by Hollister. about equal in depth and notch three Among these 13, 7 individuals possess 6 points (cusps), about equal in height and VOLUME 119, NUMBER 4 491

A etgonis A tropicalis O Orestes thomasi 1 1

A A A A ^A^ A A A \ " A A *^ A A \i/ A A A A A 0 A A CM A A A A tN A A = O Ü A a. -1 ^'.

0 O O O 8 O 0

1 1

PC I (65.4%) Fig. 5. Scatter plot of specimen scores on the first two principal components (PC) extracted from ordination of 18 log-transformed craniodental measurements as measured on 52 intact specimens representing Kenyan samples of etgonis, orestes, and tropicalis, along with the type specimen of O. thomasi Osgood, 1910 (= ht, FMNH 16698 from Molo). See Table 6 for variable correlations and percent variance explained. width, at the cutting edge of the incisor. the secondary medial groove seems to be The working edge of the lower incisor in more strongly expressed in orestes, com- O. typus, as well as that in uzungwensis, pared with specimens of dollmani, elgonis, thus appears trifúrcate or trifíd. Among and tropicalis, and is usually absent in other forms examined here, the tip of the samples of irroratus. This last observation lower incisor is bifurcate or bifid, a deep conflicts with the interpretation of most lateral sulcus clearly inscribing a central authors, who have characterized O. irror- point that is wider and set higher than the atus as possessing one deep lateral and lateral point. In dollmani, elgonis, orestes, one faint medial groove (e.g., Misonne and tropicalis, a second, much shallower, 1974, Taylor & Kumirai 2001). To us, the longitudinal groove occurs medial to the configuration of grooves in O. irroratus deep lateral groove. Presumably, this more closely resembles O. denti, which secondary groove is homologous to the (Dollman 1915) described as having only deeper medial groove observed in exam- one deep lateral groove and separated it ples of O. typus and O. uzungwensis, but from other eastern African forms with the indentation is so very shallow that the a deep lateral and faint medial groove central incisor point macroscopically ap- (i.e., dollmani, elgonis, orestes, thomasi, pears undivided; that is, the tip of the tropicalis). Expression of incisor sulca- incisor is functionally bifid. In general. tion among otomyine populations is 492 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

Table 6.•Results of principal component analy- O. irroratus or O. tropicalis, the species sis performed on 52 intact specimens representing with which it is usually associated (Boh- Kenyan examples of elgonis, orestes, and tropicalis mann 1952, Misonne 1974). The taxa and using 18 log-transformed craniodental variables orestes (Kenyan highlands), typus (Ethio- (excluding LR). See Fig. 5 and Materials and Methods for variable abbreviations. pian plateau), and uzungwensis (Ud- zungwa Mts, Tanzania) comprise three Correlations distinctive phena, each of which is found Variable PCI PCII in widely separated mountain systems. GIL 0.98*** -0.01 Their union as a single morphogenetic ZB 0.88*** -0.08 entity, O. typus as conventionally ar- BBC 0.54** 0.35 ranged (Bohmann 1952, Misonne 1974, lOB -0.20 0.75*** Musser & Carleton 1993), should rest on BOG 0.55** 0.37 DSO 0.87*** -0.38 empirical data, not vague preconceptions BR 0.84*** 0.34 of intergradation following a polytypic PPL 0.89*** -0.05 species concept. Clear-cut morphometric LBP 0.83*** 0.12 delineation of southern African O. irror- LIF 0.89*** -0.20 LD 0.96*** 0.03 atus, with predominantly 6-laminated BZP 0.76*** 0.40* M3s, from eastern African O. tropicalis, PPB 0.57*** -0.23 with predominantly 7-laminated M3s, LAB 0.65*** -0.49** mirrors the multivariate results previously DAB 0.75*** -0.56*** CLM 0.15 0.21 reported by Taylor & Kumirai (2001) and WMl 0.22 0.54** solidifies the judgments of earlier authors, Bis 0.81*** 0.48* especially Dollman (1915), regarding the Eigenvalues 0.049 0.010 specific distinctiveness of the two. Percent variance 64.3 12.7 Synonymies of the Kenyan taxa central * = P < 0.05; to this study are presented below and ** = P s 0.01; trace name availability and subsequent *** = P s 0.001. name combinations as used in major systematic works. As discussed in the respective accounts, full allocation of extremely subtle, and categorization of junior synonyms to O. orestes and O. specific traits is complicated by the very tropicalis requires much additional study, small size scale. Ultimately, careful scan- including examination of primary types ning electron microscopy must be applied and original descriptive series, as well as to substantiate our impressions and the incorporation of new specimens and data incisor descriptions of past authors and to sources. Specimens examined include all develop a consistent comparative termi- individuals personally seen and identified nology. by us. Otomys dollmani Heller Taxonomic Summary

The morphological and morphometric Otomys orestes dollmani Heller, 1912:5 evidence presented above sustains Hollis- (type locality-Kenya, Mathews Range, ter's (1919) judgment in elevating doU- Mount Gargues [Warges, Uaraguess], mani to a valid species. Our results 7000 ft [2134 m]; holotype-USNM provide no convincing justification for 181790).•Dollman, 1915:157 (listed continuing to treat dollmani either as as valid subspecies).•Allen, 1939:346 a form of O. orestes, as originally de- (listed as valid subspecies).•Ellerman, scribed (Heller 1912), or as a subspecies of 1941:323 (listed as valid subspecies).• VOLUME 119, NUMBER 4 493

A Orestes A typus o uzungwensis 1 1 1 1 1

A

A A A - U A' A A A A Á O A Á CL ' O A A^ OO O 0 G O ©o o M O

1 1 1 1 -2-10 1 PC I (47.7%)

O -

> ü -2 -

-4 -

CV 1 (72.3%) Fig. 6. Ordinations performed on 34 intact specimens representing the disjunct highland taxa orestes, typus, and uzungwensis. Top, projection of specimen scores onto the first two principal components (PC) extracted from factor analysis; the score for the holotype (MCZ 26645) of O. uzungwensis Lawrence & Loveridge (1953) is indicated (= ht). Bottom, projection of specimen scores onto the first two canonical variâtes (CV) extracted from discriminant function analysis; average between-centroid divergences (Mahalanobis D^) among the three samples are indicated along the two-way arrows. See Table 7 for variable correlations and percent variance explained. 494 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

Table 7.•Results of principal component and discriminant function analyses performed on 34 intact specimens representing the disjunct highland taxa orestes, typus, and uzungwensis and using 17 log- transformed craniodental variables (excluding LR, DSO). See Fig. 6 and Materials and Methods for variable abbreviations.

Correlations Variable PCI PCII PCIII cvi CV2

GIL 0.97*** 0.03 0.14 0.56* 0.07 ZB 0.78*** 0.20 0.45 0.43 -0.27 BBC 0.18 0.59** -0.01 0.43 -0.32 lOB 0.19 -0.27 -0.07 -0.08 0.24 BOC 0.58** 0.26 -0.36 0.70*** 0.16 BR 0.54 -0.72*** 0.09 -0.24 0.43 PPL 0.78*** -0.13 0.39 0.24 -0.08 LBP 0.87*** 0.17 -0.17 0.66** 0.22 LIF 0.89*** 0.10 0.20 0.59** -0.02 LD 0.93*** 0.00 0.19 0.53* 0.07 BZP 0.74*** -0.34 -0.26 0.27 0.53 PPB 0.57* 0.68*** 0.01 0.79*** -0.27 LAB 0.48 0.68** 0.21 0.63** -0.51 DAB 0.43 0.67** 0.29 0.50 -0.57* CLM 0.59* 0.50 -0.57* 0.91*** 0.25 WMl 0.63** 0.21 -0.59* 0.70*** 0.45 Bis 0.42 -0.87*** -0.03 -0.33 0.62** Eigenvalues 0.034 0.017 0.006 14.7 5.6 Percent variance 47.7 23.2 8.1 72.3 27.7

* = P < 0.05; ** = P s 0.01; *** = P s 0.001.

Setzer, 1953:334 (retained as valid sub- skull profile relatively flat as per O. species). tropicalis (noticeably arched in O. ores- Otomys dollmani: Hollister, 1919:147 (el- tes); lower incisor with bifid tip and faint evation to species).•Musser & Carle- medial sulcus, M3 with 6 laminae. ton, 2005:1526 (retention as valid spe- Remarks.•A key feature of dollmani cies). that influenced Heller (1912) to describe it Otomys irroratus dollmani: Bohmann, as a subspecies of O. orestes is the 1952:34 (new name combination, re- presence of 6 laminae in the M3, a number tained as valid subspecies). considered typical of the latter as di- Otomys tropicalis [dollmani]: Misonne, agnosed by Thomas (1900). The occur- 1974:34 (new name combination, listed rence of only 6 M3 laminae is actually less in synonymy without indication of typical of O. orestes than appreciated by rank).•Musser & Carleton, 1993:681 Thomas; 6 or 7 M3 laminae occur in (listed in synonymy without indication approximately equal frequency in the of rank). small sample of orestes proper from Mt Distribution.•Known only from the Kenya available to us (Table 8). Other- type locality in central Kenya (Fig. 9). wise the two forms are strikingly dissim- Emended diagnosis.•A diminutive spe- ilar (Figs. 7, 8), as aptly summarized by cies of Otomys, much smaller than O. Holhster (1919:148), who elevated doll- orestes and O. tropicalis in most cranio- mani to species: "It is much darker in dental and external dimensions quanti- color, has a considerably longer tail, and fied; tail relatively and absolutely long; the skull lacks almost entirely any ap- VOLUME 119, NUMBER 4 495

Table 8.•Counts of laminae in the upper third molars and lower incisor configuration in selected taxa representing the Otomys irroratus-tropicalis complex. M3 laminae are tabulated by general locality and incisor morphology is characterized at the level of taxon.

M3 laminae Lower incisor Taxon/Locality n 6 7 8 9 tip medial groove dollmani bifid faint Mt Gargues 7 7 • • • elgonis bifid faint Aberdare Mts 32 • 30 2 Uasin Gishu Plateau 9 • 7 2 Kaimosi 15 1 14 • Kakamega 14 • 14 • Mt Elgon 23 3 19 1 • irroratus bifid absent Eastern Cape 21 21 • • KwaZulu-Natal 17 15 2 • Western Cape 32 29 3 • • Orestes bifid shallow Aberdare Mts 10 • 10 • Mt Kenya 13 7 6 • • tropicalis bifid faint Mt Kenya 59 1 55 3 • typus trifid deep Ethiopia 56 • 2 42 12 uzungwensis trifid deep Tanzania 20 1 18 1 •

pearance of the highly arched interorbital (type locality = Jomberi Range, northeast region or wide spreading zygomata of of Mt Kenya), both of which he treated as orestes. The skull is in fact almost exactly subspecies of O. irroratus, and empha- like skulls of the tropicalis group except sized the nearness of their type localities for small size and the lamina formula." (see Fig. 1) in suggesting their synonymy Tail length of O. dollmani is absolutely (he prudently refrained from formally (Table 2) and relatively longer (TL/HBL doing so). However, traits of nubilus as « 63%) than either O. orestes (TL/HBL related in the original description•its « 46%) or O. tropicalis (TL/HBL = dark brown pelage, shorter tail, larger 50%); in proportional length, it more cranial size, and M3 with 7 laminae•to closely resembles O. irroratus (TL/HBL = us convey closer resemblance to popula- 61%). Hollister's assessment of cranial tions of elgonis, as appreciated by the size and shape differences is borne out by describer Dollman (1915:160): "A very our multivariate comparisons (Figs. 2, 4), dark race, related to elgonis and tropica- in which the sample of O. dollmani is lis." strongly differentiated from examples of Otomys dollmani is so far documented O. irroratus, O. orestes, and O. tropicalis only from Mt Gargues, the type locahty. (including elgonis) (also see Figs. 7, 8). Dollman (1915:157) mentioned additional Bohmann (1952) drew attention to specimens of dollmani from Mt Gargues similarities between dollmani and nubilus in the British Museum of Natural Histo- 496 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

Fig. 7. Dorsal (top) and ventral (bottom) views (ca. 1.75X) of adult crania of three species of Otomys in Kenya: left, O. dollmani (USNM 184046; GIL = 32.7 mm), a female from Mount Cargues; middle, O. tropicalis (elgonis) (USNM 184003; GIL = 36.5 mm), a male from Kaimosa Forest; right, O. orestes (USNM 164332, GIL = 35.7 mm), a male from Mount Kenya. Heller (1912) described dollmani as a subspecies of O. orestes, and Misonne (1974) placed it as a synonym of O. tropicalis. ry, collected by Percival, but considered discrepancy exists about the altitude of them "too young to be of any systematic the type locality. Heller (1912) indicated use." We have not examined these. New 7000 ft in his description, but as Hollister biological surveys should be undertaken (1919) observed, all skin tags read 6000 ft. to determine whether the species occurs in Heller's field notes suggest that either or higher sectors of the Mathews Range, just both altitudes are possible (see next to the north of Mt Cargues. Minor paragraph). Because Heller offered no VOLUME 119, NUMBER 4 497

"Heavy forest of juniper, Podocarpus, ^ > olive, crotón palms and Drocaenus ..." Within the Forest Zone, he further discerned lower and upper sectors, the lower one forest proper and the upper consisting of "Giant lobelia, date palms, bramble, lupine, lichen and moss." Al- though Otomys was recorded as an in- habitant pecuhar to the Forest Zone of Mt Gargues, Heller did not specify whether they were found in the lower or upper sector, or in both. Specimens examined.•Kenya, Math- ews Range, Mount Gargues, 6000 ft (USNM 181790, 184041-184046). Otomys orestes Thomas Otomys irroratus orestes Thomas, 1900:175 (type locality-Kenya, Mount Kenya, Teleki Valley, 13,000 ft [3962 m]).• Wroughton, 1906:275. Otomys [orestes] orestes: Dollman, 1915: (elevation to species, de facto arrange- ment as nominate subspecies).•Hollis- ter, 1919:147 (retention as monotypic species).•Allen, 1939:346 (listed as Fig. 8. Lateral views (ca. 1.75X) of adult crania of three species of Otomys in Kenya (same speci- species and vahd subspecies).•Eller- mens as illustrated in Fig. 7): top, O. dollmani; man, 1941:323 (hsted as species and middle, O. tropicalis (elgonis); bottom, O. orestes. valid subspecies).•Setzer, 1953 (re- tained as valid species and subspe- reason for usage of 7000 ft, we followed cies).•Musser & Carleton, 2005:1528 Hollister and reported the figure as found (retention as species, indication of on original specimen labels. synonyms). In his description of O. orestes doll- Otomys typus orestes: Bohmann, 1952:47 mani. Heller (1912) noted that five speci- (new name combination as vahd sub- mens were secured in the forest glades at species).•Misonne, 1974:33 (listed in the extreme summit of Mt Cargues. He synonymy without indication of was shghtly more expansive about habitat rank).•Musser & Carleton, 1993:682 in his field notes (27 Aug-1 Sep 1911; (listed in synonymy without indication Division archives, USNM): "A of rank). few Otomys irroratus [preliminary field Otomys thomasi Osgood, 1910: (type identification] taken at 6300 ft on weedy locality-Kenya, Molo; type specimen- hillsides. Rarer at 7000 feet." In his field FMNH 16698). journal. Heller identified three "floral and Otomys [thomasi\ thomasi: Dollman, faunal zones" along the slopes of Mt 1915:153 (listed as species, de facto Gargues. Specimens of O. dollmani orig- arrangement as nominate subspe- inated from the uppermost, which he cies).•Holhster, 1919:147 (hsted as called the Forest Zone, 5000-7100 ft valid species and subspecies).•Allen, (summit 8000 ft), and characterized as 1939:346 (listed as valid species and 498 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

• O. dollmani • O. orestes • O. tropicalis

Fig. 9. Distribution of Otomys dollmani, O. orestes, and O. tropicalis in Kenya based on specimens examined by us.

subspecies).•EUerman, 1941:323 (listed Emended diagnosis.•Compared with as valid species and subspecies). O. tropicalis, post-auricular patches pres- Otomys typus thomasi: Bohmann, 1952:41 ent, tail relatively and absolutely shorter, (new name combination as valid sub- hindfoot smaller. Skull strongly arched species).•Misonne, 1974:33 (listed in (flatter profile in O. tropicalis); rostrum synonymy without indication of rank). narrower, nasal tips not so abruptly •Musser and Carleton, 1993:682 (listed flared toward tips; interorbit more nar- in synonymy without indication of rank). rowly constricted; zygomatic arches ex- Otomys orestes [thomasi]: Musser & Car- panded anteriorly, somewhat boxy leton, 2005:1529 {thomasi allocated as (bowed and widest in middle and shghtly synonym without indication of rank). convergent anteriorly in O. tropicalis), maxillary process of zygoma deeper and Distribution.•Discontinuous in alpine zygomatic plate shorter; parapterygoid settings, ca. 2700^200 m, from western fossa deeper, more cavernous, anterior and central Kenya (Fig. 9), presumably to portion slightly undercutting hard palate; northeastern Tanzania (Grimshaw et al. bullae relatively and absolutely larger. 1995, reported as O. typus zinki). Molar rows and incisors narrower; M3 VOLUME 119, NUMBER 4 499 with 6-7 laminae (8-9 in O. typus); lower tency of these contrasts and their taxo- incisor tip bifid (trifid in O. typus), with nomic value. medial groove consistently expressed al- Although some systematists have beit shallow (not deep as per O. typus; viewed the status of orestes as a distinct wider, shallower, more facet-like in O. species, following the taxonomic action of tropicalis). Dollman (1915; also Hollister 1919, Set- Remarks.•Thomas (1900) differential- zer 1953, Musser & Carleton 2005), ly compared O. irroratus orestes from the others have treated it as a subjective alpine zone of Mt Kenya (13,000 ft) to synonym of O. typus, observing the pre- a then unnamed form from lower slopes cedent introduced by Bohmann (1952; (8000-10,000 ft) on the same mountain, also Misonne 1974, Musser & Carleton not to O. irroratus sensu stricto from 1993). This synonymy is unsupported by southern African. Shortly thereafter, he our analyses, which unambiguously re- (1902) named the other Mt Kenyan form veal orestes to be as distinct from O. typus tropicalis, which he also described as in Ethiopia as it is from contiguous a subspecies of O. irroratus. Our mor- populations of O. tropicalis in Kenya. phometric results support Thomas' dis- Coupled with the singular morphometric cernment of two forms along the upper footprint of orestes, its historical inclu- slopes of Mt Kenya, one present in alpine sion as a subspecies of O. typus is settings {orestes) and the other {tropicalis) questioned by qualitative differences in in the montane forest-bamboo zone, and M 3 lamination (6-7 laminae in O. orestes add weight to the subtle cranial features versus 8-9 in O. typus) and lower incisor that he visually appreciated in separating sulcation (1 deep and 1 shallow groove in them (Figs. 7, 8). The more strongly O. orestes versus 2 deep creases in O. arched skull (DSO) of O. orestes, its typus) and by the pronounced disjunction narrower interorbit (lOB), and narrower of alpine habitats in the Ethiopian High- molars (WMl) are dimensions that all lands and those on high mountains that contribute prominently to segregating border the Eastern Rift Valley (see orestes from OTUs of tropicalis in mor- Discussion). phometric space (see Figs. 4, 5; Tables 5, Which other species-group taxa belong 6); to these size and shape contrasts as junior synonyms of O. orestes"? Musser may be added the narrower incisors & Carleton (2005) listed five {malleus (Bis) of O. orestes, its relatively and Dollman 1915, percivali Dollman 1915, absolutely larger ectotympanic bullae squalus Dollman 1915, thomasi Osgood (LAB, DAB), and shorter zygomatic 1910, zinki Bohmann 1943) based on plate (BZP), which also contribute signif- original descriptions, comparisons con- icantly to morphometric segregation but ducted by early workers (Dollman 1915, which were not mentioned by Thomas Hollister 1919, Bohmann 1952), and bio- (1900). The smaller hindfoot in O. orestes geographic plausibility. Of those five, we and its noticeably shorter tail, as quanti- have examined only the holotype fied in Table 2, are also consistent with (FMNH 16698) and referred specimens the diagnostic contrasts mentioned by (FMNH 16694, 16699) of O. thomasi Thomas. None of our measured variables from Molo, Kenya. Curiously, Osgood reflects the differences in depth of the (1910) contrasted his new species only zygomatic arch (see Fig. 8) and area of with O. irroratus tropicalis, but the qual- the parapterygoid fossa, apparent to us itative features he noted for thomasi• and mentioned in the emended diagnosis; narrow nasals, conspicuously arched presumably a geometric morphometric skull, and creamy buff post-auricular approach could better analyze the consis- patches•recall those of O. orestes, as 500 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON later stressed by Dollman (1915), Boh- tion based on its isolation from Afroal- mann (1952), and Lawrence & Loveridge pine zones in Eastern Rift mountains of (1953) in their taxonomic comparisons. central Kenya (Fig. 1) and on the distinc- The type is a large individual (Bohmann tive pelage colors recounted by Bohmann [1952] considered it very old, but we (1952). graded it as full adult), but the other two The concept of O. orestes advanced specimens (skulls incomplete) fit within here essentially circumscribes Dollman's the range of size variation of the series of (1915) Otomys Division B, Section I, to O. orestes from the Aberdare Mts. En- which he assigned the species O. thomasi tered as an unknown in 8-group discrim- (including malleus and squalus) and O. inant function analysis, the type was orestes (including dollmani). He consid- indisputably assigned as a member of the ered the two to be morphologically highly orestes OTU {P = 0.999), in contrast to similar but emphasized the difference in the 7 other OTU options available among M3 laminae, 7 in thomasi and 6 in orestes, dollmani, elgonis, tropicalis, and typus; in in maintaining them as separate species. our principal component results, the score As detailed in the Results and Compar- of the specimen fell at the margin of PC I isons, 6 or 7 laminae occur in equal variation for orestes, consistent with frequency within our small sample of Bohmann's interpretation of its advanced orestes proper from Mt Kenya; 7 laminae age, and outside the sphere of dispersion uniformly characterize the M3s in sam- of elgonis and tropicalis specimens ples from the Aberdare Mts and Uasin (Fig. 5). Although these results support Gishu Plateau, including the type and allocation of thomasi Osgood to the topotypes of thomasi. The constancy of synonymy of O. orestes, given the data laminae counts, however, must be viewed at hand, we stress that sample sizes are cautiously given our small locality sam- abysmally small. The specific union of the ples, an insufficiency that once again taxon {thomasi) named from highlands to prescribes new material. We are im- the west of the Rift Valley in Kenya with pressed that Dollman's (1915) and Boh- those {prestes, malleus, percivali, squalus) mann's (1952) taxonomic comparisons described from mountains to the east must regularly featured orestes vis-à-vis tho- be tested with fresh material and other masi, squalus, and malleus, and were it not information bases. for the mistaken acceptance of laminar Our specimens from the summit of the difference, we suspect that they would Aberdare Mts may essentially represent have considered those taxa even more topotypes of Dollman's (1915) O. thomasi closely related, if not synonymous. All squalus (type locality-Mt Kinangop, such populations of O. orestes inhabit Aberdare Mts, 12,000 ft), as the series open habitats above treeline and exhibit was reported by Hollister (1919), but we a characteristic morphology. At lower have not studied the type specimen itself. altitudes within the Aberdare Mts and on Examination of that type, along with Mt Kenya, examples of O. orestes have those of malleus and percivali, original been collected in sympatry with O. type series, and, as importantly, fresh tropicalis (see respective Specimens exam- topotypic material, is required to certainly ined). Whether such instances reflect resolve the synonymy of these epithets occupation of the same biotope or wheth- described from the Aberdare Mts. Of the er alpine and montane forest elements are synonyms attributed by Musser & Carle- intermixed at those elevations cannot be ton (2005), O. typus zinki Bohmann deduced from the collectors' field jour- (1943), described from Mt Kilimanjaro nals. Long term ecological studies and at 3800 m, especially merits reconsidera- elevational surveys, such as those con- VOLUME 119, NUMBER 4 501 ducted by Clausnitzer (2001) and Claus- to 3200 m, from southern Sudan and nitzer & Kityo (2001) on Mt Elgon, are Ethiopia, through eastern D. R. Congo, needed. Uganda, and western Kenya, to Specimens examined.•Kenya: Aber- and northeastern Tanzania. dare Mts, summit, 11,000 ft (USNM Remarks.•Together with O. irroratus 184033-184035); Aberdare National Park, orestes, O. i. tropicalis was the second Fishing Lodge, 8760, 9000, and 9500 ft kind of vlei rat that Thomas (1902) recog- (USNM 589995-589997); Uasin Gishu nized as dwelling on the upper western Plateau, 20 mi N Eldama Ravine, 9000 ft slopes of Mt Kenya. He (1902:314) pithily (USNM 164290); Molo (FMNH 16694, characterized tropicalis as "The northern 16698, 16699; type and original topo types strong-coloured form of O. irroratus, with of O. thomasi Osgood, 1910); Mt Kenya, seven laminae to the last upper molar," 10,700 ft (FMNH 43444; USNM 164308, and considered it to be the widespread 164346, 164357, 164358), 13,500 ft (US- Otomys in "Eastern Tropical Africa, from NM 164293-164295, 164297, 164329, British East Africa [Uganda, Kenya] to 164332-164334), 13,700 ft (FMNH 43445, Nyasaland [Malawi]." Wroughton (1906) 43446; USNM 164300-164305, 164344, considered the Zambezi River to be the 164345). geographic divide that separated northern Otomys tropicalis Thomas (tropicalis) and southern {irroratus) forms of the O. irroratus complex, but Dollman Otomys irroratus tropicalis Thomas, (1915:149) accentuated its importance 1902:314 (type locality-Kenya, western when he raised tropicalis to species: "... slope of Mt Kenya, 10,000 ft).• tropicalis taking the place of irroratus, Wrought on, 1906:274 (listed as vaHd which species, on account of its cranial subspecies).•Bohmann, 1952:33 (re- structure and lamina formula, is not tained as vaUd subspecies); Delany, considered a near enough relative for 1975:53 (listed as synonym without the name to be used in connection with indication of rank).•Honacki et al., the East-African forms; in this manner, 1982:445 (given as synonym without we confine irroratus and its subspecies to indication of rank). the country south of the Zambesi." In Otomys [tropicalis] tropicalis: Dollman, spite of a workable diagnosis and persua- 1915:157 (elevation to species, de facto sive biogeographic context, mammalo- arrangement as nominate subspecies).• gists thereafter dithered over treatment HoUister, 1919:148 (Hsted as valid spe- of tropicalis as a race of O. irroratus cies and subspecies).•^Allen, 1939:347 (Bohmann 1952, Delany 1972) or as (listed as valid species and subspecies).• a distinct species (Allen 1939, Hatt 1940, EUerman, 1941:322 (Hsted as valid spe- Misonne 1974, Musser & Carleton 1993). cies and subspecies).•Misonne, 1974:34 The recent study of Taylor & Kumirai (listed as species, indication of syno- (2001) brings decisive illumination to this nyms).•Corbet & Hill, 1980:159 (Hsted issue in demonstrating that clear cranio- as species).•Corbet & Hill, 1986:178 metric differentiation accompanies the (Hsted as species).•Corbet & Hill, qualitative dental and pelage contrasts 1991:167 (Hsted as species).•Musser & between O. irroratus and O. tropicalis long Carleton, 1993:681 (Hsted as species, ago appreciated by Thomas, Wroughton, indication of synonyms).•Musser & and Dollman. Our results reinforce the Carleton, 2005:1529 (listed as species, conclusion of Taylor & Kumirai (2001) indication of synonyms). apropos the specific distinction of O. Distribution.•As currently understood, tropicalis from O. irroratus in southern highlands in eastern Africa, mostly 1800 Africa and additionally provide evidence 502 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON for segregation of O. tropicalis from O. The core distribution of O. tropicalis dollmani, O. orestes, and O. typus sensu occupies the highlands of western and stricto in eastern Africa. central Kenya (Fig. 9), but the limits of its As with O. orestes, the legitimacy of range, to the north, south and west, are synonyms allocated to O. tropicalis and poorly documented. We have examined refinement of its geographic distribution the type and referred series of Otomys invite future studies that pose specific orestes giloensis Setzer, 1953 (type local- taxonomic hypotheses within critical re- ity•Sudan, Gilo, 6500 ft), known from gional contexts, that amplify field surveys the Imatong Mts in southern Sudan and sample sizes, and that incorporate (FMNH 73901, 79456-79463, 108207, new data bases. Musser & Carleton 108208). Setzer contrasted his new Suda- (2005) assigned eight species-group taxa nese subspecies with the Kenyan taxa to O. tropicalis•e/goKW Wroughton 1910 orestes dollmani (sensu Heller 1912) and o. (Elgonyi, Mt Elgon, Kenya), faradjius orestes (sensu Dollman 1915), a narrow Hatt 1934 (Faradje, D. R. Congo), ghigii scope of comparison shaped by his in- de Beaux 1924 (Bugala, Uganda), giloen- terpretation of 6 laminae on the M3 of his sis Setzer 1953 (Gilo, Imatong Mts, type series. Our examination of those Sudan), nubilus Dollman 1915 (Jomberi specimens uniformly revealed a 7-lami- Range, Kenya), rubeculus Dollman 1915 nated M3, a discrepancy that must reflect (Kagambah, Uganda), vivax Dollman different criteria for scoring laminae. In 1915 (Mt Nyiro, Kenya), vulcanis Lönn- view of the affinity and rank supposed by berg and Gyldenstolpe 1925 (Mt Sabinio, its describer, along with Bohmann's D. R. Congo)•based on the comparative (1952) union oí orestes with typus, giloen- context given in many of the original sis was afterwards swept under the descriptions, which recurringly referenced synonymy of O. typus sensu lato (Misonne tropicalis or elgonis or both. They further 1974, Musser & Carleton 1993). Never- identified three morphological assem- theless, in 8-group discriminant function blages among these epithets that may analysis including samples of elgonis, merit specific status: tropicalis proper on orestes, tropicalis, and typus, the holotype Mt Kenya, elgonis (provisionally includ- oí giloensis (FMNH 73901) was associat- ing faradjius, ghigii, giloensis, nubilus, ed with two of the smaller-bodied OTUs vivax), and rubeculus (provisionally in- of elgonis, at comparable levels of a pos- cluding vulcanis). Wroughton (1910) de- teriori probabihties (KA, P = 0.466; ME, scribed elgonis as being most similar to P = 0.398), not with the samples of O. tropicalis, except for smaller size and orestes from Kenya or O. typus from darker pelage color, a resemblance ro- Ethiopia. This analytical consignment bustly sustained by results of our mor- concurs with our estimation of its affinity phometric analyses (Figs. 2, 4, 5). The derived from inspection of skins and usual overlap in multivariate space among skulls: giloensis is a darkly colored examples of tropicalis and elgonis contra- animal, slightly smaller and more uni- dicts the speculation of Musser and formly chocolate brown than typical Carleton and justifies their contin- elgonis, but it possesses the cranial shape, ued synonymy. Although the type speci- lower incisor grooving, and 7-laminated mens of tropicalis and elgonis were not M3 that otherwise recall the elgonis consulted, we are confident that our morph. The presence of O. tropicalis in samples from the slopes of Mt Kenya southwestern Ethiopia was reported by and Mt Elgon comprehensively represent, Musser & Carleton (2005), who noted it respectively, the morphologies intended (FMNH 28165) as altitudinally parapa- by Thomas (1902) and Wroughton (1910). tric with O. "typus" on Mount Albasso. VOLUME 119, NUMBER 4 503

The kinds, detailed distribution, and or whether it is a valid species endemic to ecology of Otomys in Ethiopia invite the Eastern Arc Mountains and related to wholesale revision (also see Lavrench- the O. tropicalis complex. At whichever enko et al. 1997). rank, such a phyletic affinity is biogeo- Towards the south, the material re- graphically as persuasive as its present ported from Mts Meru and Kilimanjaro classification under O. denti, otherwise in northeastern Tanzania plausibly repre- known from mountains along the upper sents O. tropicalis (Demeter & Hutterer Western Rift Valley. 1986, Grimshaw et al. 1995), based on the To the west, the synonymy of rubeculus authors' descriptions of morphology and- Dollman (1915) and vulcanis Lönnberg & or elevational range and habitats. A small Gyldenstolpe (1925), from southwestern series from the West Usambara Mts, Uganda and northernmost Rwanda, re- a range forming the eastern portion of spectively, under O. tropicalis or with one the Eastern Arc Mts, also fits with O. another should be objectively demon- tropicalis (Shume, 6000-7000 ft [USNM strated. One can only suspect, for in- 381484-381487]; Sunga [USNM 340800]). stance, that much of the confusing Sunga is the type locality of O. denti variation that Dieterlen (1968) encoun- sungae, which Bohmann (1943) named as tered within O. "irroratus" from moun- an outlier subspecies to the main distri- tains around Lake Kivu indicated the bution of O. denti in Western Rift presence of two or more species, one of mountains (type locahty = Uganda, Mt which should bear the regional epithet Ruwenzori, 6000 ft). Compared with O. vulcanis. Furthermore, the morphological denti proper, however, these specimens and genetic homogeneity of populations exhibit more spatulate-shaped nasals, represented by these epithets, described a broader zygomatic plate that partially from mountains rising astride the West- covers the premaxillary-maxillary suture ern Rift Valley, with O. tropicalis and (thus a deeper zygomatic notch), larger others named from Afromontane high- otic capsules, broader upper incisors, lands bordering the Eastern Rift Valley, lower incisors with a faint medial groove, must be viewed skeptically based alone on and M3 with 7 laminae [Bohmann (1943) geographic occurrence and on the pro- reported 6 laminae based on the type and nounced biotic discontinuity established 1 other specimen in the BMNH]. All of between the two mountain systems (e.g., these traits are indicative of O. tropicalis, see Moreau 1966, White 1978, Dowsett in particular the elgonis morphology. The 1986, Carleton et al. 2006). rich blackish-brown pelage in these 5 Specimens examined.•Kenya: Aber- specimens favorably compares to Boh- dare Mts, 10,500 ft (USNM 164279, mann's (1943, 1952) description of O. 164280), summit, 11,000 ft (USNM denti sungae, but such a dark tone is 184025-184032, 184037-184340); Aber- approached by other taxa currently asso- dare Mts, Changongorra, 7000 ft (USNM ciated with O. tropicalis•e.g., more 184019-184023); Aberdare National Park, strongly saturated examples of elgonis Fishing Lodge, 8760, 9000, and 9500 ft and the type series oí giloensis and nubilus (USNM 589998, 589999, 590002, 590003, (Dollman 1915, Setzer 1953), taxa also 590012-590015); 12 mi N Eldama Ravine, described from lower mountains near the 8800 ft (USNM 164281); Uasin Gishu distributional periphery of O. tropicalis. If Plateau, 30 mi N Ravine, 8600 ft (USNM the USNM specimens prove to be repre- 164286-164289); Uasin Gishu Plateau, sentative of sungae proper, future in- Nzoia River, 6000 ft (USNM 164282- vestigation should determine whether the 164285); Kaimosi Forest (USNM taxon is a junior synonym of O. tropicalis 184003-184017); Kakamega (MCZ 504 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

31375, 31416-31420, 31432-31437, 31439, altogether too slight a character on which 31440; USNM 184018); Molo (FMNH to base species ...." In this particular 16693, 16695); Mt Elgon National Park, instance, quantitative comparisons of the 15 km from Chorlim Gate towards Koi- degree of nasal flaring have proven help- tobos Peak (USNM 590004-590006, ful for discrimination of closely related 590011); Mt Kenya, west slope, 7000- species of Otomys (Taylor & Kumirai 13,700 ft (CM 2836, 2837; FMNH 43819- 2001, Taylor et al. 2005). For Ellerman 43822; USNM 164277, 164291-164293, and other taxonomists of the middle 164296, 164298, 164299, 164307, 164309- 1900s, however, certain classes of char- 164312, 164314-164324, 164326, 164330, acters, especially subtle size and shape 164331, 164337-164343, 164347-164350, distinctions, were a priori viewed as 164353, 164354, 164356, 164359-164363, indicative of geographic variation and 164365-164373, 164375); Mt Kenya, Naro deemed useful only for recognition of Mom, 10,000 ft (USNM 590000, 590001); subspecies. Mt Kenya, Naro Moru Gate, Meteoro- Bohmann's (1952) study appeared at logical Station, 10,000 ft (USNM 590007- the crest of this new "population think- 590010); Naivasha Plains, Nalsitichu Riv- ing" and constituted the watershed pub- er (USNM 184036); 10 mi E Nyeri lication in applying this research para- (USNM 164278). Uganda: Mt Elgon, Sipi digm to patterns of inter- and (MCZ 31269-31272, 31373, 31374, 31377- intraspecific variation among named 31381, 31426-31432). forms of Otomys (He recognized the single genus Otomys and no subgenera; Discussion most authors have retained Parotomys as generically distinct.). Eleven species, The biological species narrative and nearly all polytypic and some of these current underestimation of diversity morphologically highly heterogeneous, within Otomyinae.•By the close of the were recognized by Bohmann. His con- fertile era of taxonomic discovery of struct of an exceptionally polymorphic O. African Rodentia (ca. 1880-1930), 28 or irroratus, distributed over much of Sub- 30 species of otomyines were recognized saharan Africa, was successively broad- (Allen 1939, Ellerman 1941). Emergence ened to subsume three forms that he had of the New Systematics (Huxley 1940) maintained as species•anchietae, lamina- and biological species concept (Mayr tus, and typus•further decreasing the 1942) in the middle 1900s profoundly number of species to only 8 (Dieterlen influenced how mammalogists treated 1968, Petter 1982). Unfortunately, the and nomenclaturally conveyed interpop- recommendations in the latter contribu- ulation variation. Specific taxa based on tions lacked the synoptic regional and supposedly minor features of size and taxonomic perspective developed by Boh- pelage during the early descriptive phase mann and compounded his underestima- were routinely reconsidered as geographic tion of otomyine diversity. By the late variants or subspecies, often with little or 1900s, when taxonomic listings of mam- no presentation of supportive data. Ex- malian species became popular, the num- amination of the pivotal reasoning for ber of valid species had stabilized at 12 or 13 some of these synonymies is instructive. (Misonne 1974, Corbet & Hill 1980, 1986, For example, Ellerman et al. (1953:307), 1991; Honacki et al. 1982, Musser & aligned angoniensis as a subspecies of O. Carleton 1993•The departure from Boh- irroratus, contra Roberts (1951) who mann's recognition of 11 reflected the maintained the two as distinct based on tendency in these works to maintain eastern distal nasal shape, because "this seems African tropicalis as distinct.). Classifíca- VOLUME 119, NUMBER 4 505 tory consensus over the past 30 yr should alpha-level systematics of otomyine ro- not be confused with systematic under- dents in the latter 1900s: in aggregating so standing: repetition without recourse to many more or less distinctive forms as new data or new methods of evaluating old supposedly intergrading subspecies, as data does little to advance it. was impelled by the persuasive orthodoxy Given such systematic quiescence, the of the biological species concept, taxono- notion has passively entered the recent mists could afterwards only lament the literature that differentiating traits rehed immense plasticity of morphological traits upon in the pioneering studies of oto- and their insufficiency as diagnostic myines are too highly variable to be characters at the species level. taxonomically useful at the species level. That perception may have stemmed Lawrence & Loveridge (1953:60), for from the words of the illustrious taxono- example, remarked that "the grooving mist Oldfield Thomas (1918:204), who on the lower incisors and the lamina considered "the grooves on the incisors, formula of M3/ in Otomys are highly and the numbers of the molar laminae, plastic." More recently, Taylor and Ku- used so effectively by Wroughton and mirai (2001:162) prefaced their cranio- Dollman for the sorting of species, are so metric study by observing that "the plastic, and show so wide a range of taxonomy of the group is highly unstable variation, that, however useful for specific due to variability in diagnostic characters, distinction, they have to be used with usually involving the number of laminae great caution when generic divisions are in M3 and Ml, grooving of the inci- in question." Careful rereading of his sors...." That these characters are in- comments, however, indicates that he ordinately variable and extremely plastic readily acknowledged the utility of these is implicit in the disparate synonymies features for species-level systematics• and heterogeneous morphologies that "however useful for specific distinc- Bohmann (1952) submerged under the tion"•but not for genus-group arrange- "species" O. irroratus and O. typus, but ments. Thomas' statement was meant to his study did not constitute an empirical introduce his emphasis of other cranial demonstration of such variabihty. A pro- features (bullar size, nasal expansion, ponent of Rensch's (1929) rassenkreis interorbital constriction), in a paper ex- concept, the intellectual predecessor to pressly designed to revise the classifica- Mayr's (1942) polytypic species, and tion of Otomyinae and to define new intent on identifying geographically replac- genera and subgenera (e.g., Anchotomys, ing subspecies, Bohmann's revision re- Myotomys, Parotomys), not to reevaluate mains an insightful narrative of evolu- specific limits. His observations on the tionary and biogeographic diversification morphological plasticity and extraordi- within Otomys, but it less critically nary variation of otomyine dentitions addressed among-population variation have been taken out of context as they and character intergradation as bases for bear on species-level questions. delimiting species. In consequence of the Notwithstanding Thomas' (1918) de- immense amount of inter-population var- preciation of molar lamination and in- iation embraced by his permissive defini- cisor grooves for delineating major tions of species limits, it was a small step lineages among otomyine species (genera for Dieterlen (1968) and Petter (1982) to and subgenera), it is tellingly ironic that rationalize the morphological boundaries he was nonetheless obliged to rely on of O. irroratus as also embracing anchie- these very same characters in his taxo- tae, laminatus, and typus. In hindsight, nomic key and genus-group diagnoses. a curious circularity has pervaded the Indeed, we find it noteworthy that traits 506 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON of molar lamination and incisor sulcation exceeded those of taxon or OTU as a post have regularly served in identification hoc grouping variable (especially in keys to otomyine species, whether em- DFAs wherein OTU was procedurally ployed in taxonomic or faunal works inherent to the computation), but the (Wroughton 1906, Dollman 1915, Eller- significance level attained by M 3 laminae man et al. 1953, Misonne 1974, Delany often matched those broader group 1975, Smithers 1983, Meester et al. 1986, effects. Furthermore, all three of those Taylor 1998). The practical application of group effects conventionally dwarfed the these characters issues from their pro- contribution of sex and age cohorts to nounced modal development within pop- dispersion of factor scores. A notable, ulation samples distributed over broad and predictable, exception to this gener- geographic regions, which is the actual alization involved the influence of age- pattern of variation so far documented class upon first principal component for otomyines (Table 8; Dollman 1915, scores extracted in two of the three factor Bohmann 1952, Taylor & Kumirai 2001). analyses (see explanation under Results); Numbers of M3 laminae do not range in these instances, taxon, OTU, and-or from 4 to 10 within populations and M3 laminae emerged as important cate- modal counts do not vary randomly gorical effects on the second component among geographic samples. Such useful- derived (Table 9). In summary, differ- ness in keys, in and by itself, controverts ences in number of M3 laminae covary the notions of inordinate plasticity and intelligibly with unambiguous discrimina- uninformative variation, particularly in tion of otomyine populations and species a rodent superfamily wherein morpholog- in morphometric space, and these pat- ical differences among closely related terns intimate a significant degree of species usually rest upon minimally per- morphogenetic integration and ontoge- ceptible features of size, shape, and pelage netic canalization. Applied in concert texture or color. A combination of with such continuous variables and other continuous and quahtative features is discrete traits, molar lamination is a tax- typically necessary for workable defini- onomically informative and highly useful tion and recognition of taxa at the species character system for distinguishing oto- level in Muroidea. Considered against myine species. this experience, readily grasped meristic Of the taxa considered herein, Boh- characters such as molar laminae are mann (1952) had allocated some to O. a boon not to be casually dismissed. irroratus (23 subspecies, including doU- And in point of fact, variation in M3 mani, elgonis, and tropicalis) and some to laminae regularly complements spatial O. typus (11 subspecies, including orestes structure evident in the multivariate and thomasi). Taylor & Kumirai (2001) summaries of craniodental covariation. have previously cemented the morpho- Although number of laminae was scored metric argument for separation of south- for each specimen used in our various ern African O. irroratus from eastern morphometric comparisons, we empha- African O. tropicalis sensu lato and from size that it was not entered as an O. typus sensu lato (i.e., according to analytical variable. Still, post hoc tests Bohmann 1952, Misonne 1974). We of M 3 number as categorical effect in regard the retention of O. dollmani, O. analyses of variance typically disclosed orestes (including thomasi), O. tropicalis significant mean differences among the (including elgonis), and O. uzungwensis as extracted factors, whether principal com- vahd species, distinct from O. irroratus ponent or canonical varíate (Table 9). and O. typus, as a rough first-glimpse to The magnitude of F values seldom the species diversity yet to be uncovered VOLUME 119, NUMBER 4 507

Table 9.•One-way ANOVAs generated for various group effects on results of discriminant function (DFA) and principal component (PCA) analyses of Otomys: 9-group DFA (see Fig. 2); 8-group DFA (see Fig. 4); 3-group DFA (see Fig. 6); PCA of samples of O. irroratus and O. tropicalis (including elgonis; see Fig. 3); PCA of samples of O. orestes and O. tropicalis (including elgonis; see Fig. 5); and PCA of samples of O. orestes, O. typus, and O. uzungwensis (see Fig. 6).

Variable F(taxon) F(OTU) F(sex) F(age) F(M3 laminae)

DFA: 9 OTUs CV 1 scores 170.1*** 73.7*** 0.3 3.2* 80.1*** CV 2 scores 31.3*** 30.9*** 3.5 0.1 1.3 DFA: 8 OTUs CV 1 scores 134.9*** 101.6*** 1.9 1.3 6.7** CV 2 scores 75.6*** 47.8*** 0.8 0.1 19.3*** DFA: 3 Highland OTUs CV 1 scores 228.1*** • 3.1 2.7 14.2*** CV 2 scores 87.3*** • 0.0 0.8 4.1* PCA: O. irroratus-O. tropicalis PC I scores 13.9*** 5.7*** 0.5 34.3*** 4.5* PC II scores 114.1*** 24 4*** 0.3 0.2 30.8*** PCA: O. orestes-O. tropicalis PC I scores 2.1 3.5** 1.8 29.9*** 1.6 PC II scores 68.3*** 44.6*** 0.0 0.5 3.3* PCA: O. orestes-O. typus-O. uzungwen '.sis PC I scores 9.0** • 1.9 1.7 4 9** PC II scores 38.1*** • 3.7* 7.8** 1.5

P < 0.05; P < 0.01; = P s 0,001. in eastern Africa. As remarked in the cies, coupled with continued investigation above taxonomic summaries, the homo- of Otomys found south of the Zambezi geneity of O. orestes and O. tropicalis as River (e.g., see Taylor et al. 2004b), will arranged herein are highly questionable. surely restore the diversity within Oto- Otomys typus, even with removal of O. myinae at least to the level of 28-30 orestes and O. uzungwensis and as re- species, although doubtfully the same stricted to Ethiopian plateaus, remains ones enumerated by Allen (1939) and a composite (see Lavrenchenko et al. Ellerman (1941). 1997, Musser & Carleton 2005). The In light of resurgent assessments of stature of O. dartmouthi and O. jacksoni, species limits within Otomyinae (Dieter- viewed as species by earlier authors and len & Van der Straeten 1992, Taylor & Musser & Carleton (2005) or classified Kumirai 2001, this study), Guy Doll- under O. typus by Bohmann (1952) and man's (1915) classification of Otomys in others, still requires validation. Synonyms eastern Africa earns a retrospective salute attributed to O. denti (kempi and sungae) for its modern outlook. His study elevat- deserve réévaluation. New species will ed to species many forms treated as likely emerge from the upper elevations subspecies of O. irroratus by Thomas of other mountains in eastern Africa that (1900, 1902) and Wroughton (1906)• have yet to receive thorough field survey. e.g., O. angoniensis, O. denti, O. orestes, And so on. Redressing these inadequa- and O. tropicalis•and following the 508 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON retrograde interlude in the middle 1900s groups (Moreau 1966, Dowsett 1986, when the stature of most was again buried White 1978, 1981), most of which are under O. irroratus, their validity as species congruent but which differ between the is gaining renewed appreciation. Dollman disciplines in exact limits and in name. further underscored the biogeographic Otomys dollmani and O. orestes, as un- significance of the Zambezi River in derstood here, fall within the Kenya separating eastern and southern species montane group (according to the orni- complexes of Otomys. Of the several thological designation). Otomys tropicalis, patronyms associated with the genus, O. as currently arranged, is found in the dollmani Heller (1912) is among the most Kenya and East Congo montane groups appropriate and deserved. but may extend to the periphery of the Montane biotopes in eastern Africa and Ethiopian Plateau and Tanganyika- the diversification o/Otomys.•The Ken- Nyasa Ranges. The endemism levels and yan species of Otomys central to this localized distributional patterns exhibited study are restricted to moist montane by other plants and that inhabit environments in eastern Africa (Fig. 9), these same mountain systems expose the the majority of localities falling within an probable composite nature of this pro- elevational belt covering 1800 to 4000 m. visional arrangement of O. tropicalis Two species are found in middle to upper (White 1978, Dowsett 1986, Stuart et al. montane zones, one narrowly (O. doll- 1993, Bowie et al. 2004, 2006; Carleton et mani) and one more broadly distributed al. 2006,). (O. tropicalis), and one occurs higher in In our view, the conceptual approach ericaceous vegetation and tropical alpine to understanding specific radiation of habitats (O. orestes). Based on such Otomys, especially in eastern African ecological occurrences and a naturally highlands, has been misdirected, hindered discontinuous highland setting, these spe- by the historical preference for large, cies comprise elements endemic to the broadly distributed polytypic species and Afromontane Biotic Region. The region implicit reliance on a model of geographic is sometimes used in a broad sense to (allopatric) speciation. Bohmann's (1952) include both wet montane forests and definition of O. typus, as expanded by open habitats above continuous tree Misonne (1974) and currently accepted in canopy (i.e., following White 1978, most systematic works, exemplifies our 1981); others consider the capstone point. The species' distribution was de- afroalpine vegetation to comprise a sepa- cidedly patchy, extending from the Ethi- rate biotic region (Hedberg 1986, Claus- opian plateaus (nominate O. typus and nitzer 2001). According to either defini- the subspecies fortior, helleri, and mal- tional scope, the Afromontane Region of kensis), across the highest mountain tops Subsahara Africa is highly disjunct and in eastern African•Ruwenzori Mts (dart- archipelago-like in its occurrence, consist- mouthi), Mt Elgon (jacksoni), Mau Es- ing of several mountain ranges and carpment (thomasi), northern and south- highland blocks that are diverse in geo- ern sectors of the Aberdare Mts (malleus logical age and orogenic formation; these and squalus), Mt Kenya (orestes), and Mt nonetheless share a characteristic and Kilimanjaro (zinki)•to as far south as largely stenotopic high-altitude flora and Mt Rungwe and the Udzungwa Mts in fauna. Within the Afromontane Region western Tanzania (uzungwensis). Observ- (sensu lato), both botanical and ornitho- ing the tenets of a polytypic species, logical biogeographers have distilled dis- Bohmann was forced to assume that the tributional patterns and foci of endemism isolated populations indicated the past to denote seven major mountainous general distribution of O. typus in VOLUME 119, NUMBER 4 509 eastern Africa and former interconnec- afroalpine morphotype may be sketched tions among afroalpine biotopes, pre- as having a deeper and more luxuriant sumably during pluvial maxima of the pelage with pale post-auricular patches, Pleistocene. a relatively short tail, a more conspicu- Botanical evidence for such past con- ously arched cranium and constricted nections of afroalpine zones has not been interorbit, narrower rostrum with less forthcoming. Hedberg (1986:457) re- expansive nasal tips, relatively larger marked that "... intermountain contacts auditory bullae, M3 with more laminae, may have been established for the mon- and lower incisors with two grooves. Such tane forest belt, but such contacts are an idealized morphology involves tenden- unlikely to have been possible for the cies of character development, but when ericaceous and afroalpine belts," and that examined in detail within geographically "... there is no evidence of any large-scale cohesive regions, not all of these qualita- Pleistocene intermountain dispersal of tive characteristics are equally expressed afroalpine vascular plants." Working in the taxa that Bohmann gathered under with giant senecios {Dendrosenecio, Aster- typus. That is to say, development of aceae), a spectacular genus iconographie incisor grooving, numbers of M3 laminae, of afroalpine habitats, Knox & Palmer and presence-absence of post-auricular (1995:10350) observed that "The current patches occur in different and unique geographic distribution is insular and combinations (this study, Bohmann 1952, would have remained almost entirely so, also see Musser & Carleton 2005). More- even given the response of the vegetation over, pronounced morphometric diver- to the estimated extremes of past climates gence in size and shape demonstrably ...." Instead, these authors have invoked accompanies such qualitative contrasts, at long distance seed dispersal by wind and least in typus, orestes, and uzungwensis, by birds (in feathers or mud on feet) to and together, these discontinuities more partly account for colonization of afroal- convincingly argue specific boundaries, pine zones and chorological similarities in not subspecific intergradation. The same their plant communities. Still, endemism pattern is apparent for certain taxa levels within the afroalpine flora are confined to the Afromontane zone prop- exceedingly high in the mountains of er, i.e., O. dollmani and O. tropicalis, and eastern Africa and Ethiopia, and many additional study of the latter will likely plant endemics are restricted to a single divulge other instances of specific differ- mountain (Hedberg 1986). Furthermore, entiation within Afromontane environ- the geologically young age of some ments coincident with major mountain eastern African volcanoes (ca. 1 million - systems. years old•e.g., Mts Kilimanjaro and Rather than a classical model of Meru; Griffiths 1993) indicates that allopatric speciation, the possibility of heights sufficient to support an alpine multiple independent originations (speci- zone have been attained only recently, ation) of otomyines in afroalpine envir- a spatiotemporal consideration that also onments on eastern African mountain- disputes notions of the former connectiv- tops deserves to be tested as an alternative ity of their uppermost plant zones with hypothesis to explain their diversification. those in more northern and southern Phylogenetic results recently obtained for mountains. other organisms confined to moist mon- Bohmann (1952), of course, perceived tane environments within tropical lati- real morphological similarities that per- tudes offer instructive analogies. Giant suaded him to group these taxa under senecios {Dendrosenecio), for example, are a widely distributed Otomys typus. This prominent members of afroalpine floras 510 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON in eastern Africa where populations of type may be anticipated. In addition to Otomys co-occur. Knox & Palmer forms once synonymized under typus, (1995:10349) summarized that the diver- denser sampling of populations and taxa sification of giant senecios "... primarily currently placed within O. dent i and involved repeated altitudinal radiation, especially O. tropicalis will prove crucial both up and down the mountains, leading to exploring these possibihties. to morphological paralleHsm in both A paleontológica! coda.•Otomyine ro- directions." In a genus of Neotropical dents are thought to have evolved from an butterfly, Hall (2005:5) concluded that arvicanthine-like stock sometime in the "The overwhelming pattern discernable in late Miocene, a hypothesis robustly sup- Ithomiola is of vertical speciation, ..., one ported by morphological, paleontological, elevational band at a time, following or and molecular studies (Pocock 1976, 1987; concomitant with orogenic cycles of Carleton & Musser 1984, Chevret et al. uplift." Recent molecular studies of Afri- 1993, Senegas & Avery 1998, Ducroz et al. can birds have disclosed complex vicari- 2001, Senegas 2001). Furthermore, south- ance patterns, bursts of speciation, and ern Africa is believed to encompass the the concentration of evolutionarily young area of origin of the subfamily, as in- species in montane regions of eastern dicated by the presence of the oldest Africa, concomitant with pronounced known fossils (Pocock 1976, 1987; Denys vegetational, climatic, and geological 1989, Senegas & Avery 1998, Senegas changes during the Quaternary (Fjeldsâ 2001), including the annectant fossil genus & Lovett 1997, Roy et al. 1997, 1998; Euryotomys (Mio-Pliocene, South Africa), Bowie et al. 2006). In light of the dynamic and by the geographic distribution of biogeography and prolific phylogenetic morphological apomorphies (Bohmann diversification that have attended the 1952, Denys 1989). Earliest fossil species Pleistocene history of other montane of Otomys proper date from the middle to organisms, the apparent evolutionary late Pliocene (2-3.5 million years ago) in stasis of a terrestrial rodent of low southern Africa (Senegas & Avery 1998, vagility, as refiected in the conventional Senegas 2001) and from the early Pleisto- taxonomies and wide distributions of O. cene (1-2 million years ago) in eastern tropicalis and O. typus, is difficult to Africa (Denys 1989). As with the later reconcile if not patently absurd. appearance of fossils, the distribution of The above examples represent only apomorphies within Otomys•notably the circumstantial evidence and do not spe- greater incidence of taxa with a 7-, 8-, or 9- cifically improve understanding of the laminated M3s {O. laminatus of south- phyletic radiation of otomyine ern Africa is a conspicuous exception), in eastern African mountains. The strik- lower incisors with dual sulci, and prom- ing morphological differentiation evident inently arched cranial vaults•also suggest among taxa {O. dollmani, O. orestes, O. that more highly derivative lineages are tropicalis, O. uzungwensis) once consid- localized within eastern Africa highlands. ered to be subspecies of widely ranging Otomys petteri, an extinct species named species, namely O. irroratus or O. typus, is from Olduvai Bed I, early Pleistocene of not circumstantial and provides a frame- Tanzania (Denys 1989), warrants com- work for future phylogenetic investiga- ment in this regard and apropos the tion. A pattern of regionally localized evolutionary significance of O. dollmani. cladogenesis, situated within the major Key diagnostic characters of O. pet- Afromontane physiographic groups iden- teri• its relatively small size, 6-laminated tified for other organisms, and indepen- M3, and lower incisors with a deep lateral dent evolution of the afroalpine morpho- groove and shallow medial groove are the VOLUME 119, NUMBER 4 511 same traits that recall O. dollmani. In can forms with 7 or 8 M3 laminae, i.e., O. contrasting the fossil O. petteri with living tropicalis (including elgonis), O. orestes, Otomys, Denys (1989:714) mistakenly and O. typus. Phylogenetic studies apply- attributed 7-8 laminae as characteristic ing gene-sequence analyses should ex- of the M3 in the "tropicaiis-irroratus plore this possibility. group." The possession of 6 M3 laminae by southern African O. irroratus was long Acknowledgments ago appreciated by Thomas (1902), who emphasized this contrast in his definition For permission to examine specimens, of tropicalis (as a subspecies of O. either through loans or collection visits, irroratus). The characteristic modal con- we thank S. McLaren (CM); L. R. trast in M3 laminae between southern Heaney, J. D. Phelps III, and W. T. irroratus and eastern tropicalis was also Stanley (FMNH); and M. Lawrence and acknowledged by other authors in early J. Chupasko (MCZ). The elegant skull classifications of Otomys, whether as photographs were provided by D. F. separate subspecies (Wroughton 1906) Schmidt (Mammal Division, USNM), or species (Dollman 1915). Along with and D. Cole (Office of Information O. irroratus proper, O. dollmani emerges Technology, NMNH) offered assistance as another member of the "tropicaiis- and advice on GIS mapping. Data irroratus group," as currently understood, collection for this project was initiated whose M3 possesses only 6 laminae. by E. Schaefer while she was a student in Réévaluation of the Olduvai sample is the NMNH Research Training Program. required to determine whether O. petteri Finally, we appreciate the comments and more closely resembles O. irroratus or O. suggestions of the reviewers G. G. Mus- dollmani, instead of O. angoniensis as ser, P. J. Taylor, and W. T. Stanley. Of highlighted by Denys in her comparisons course, any errors of commission or with recent Otomys. Owing to our differ- omission remain entirely our own. ent measuring protocols•one focused on crania, the other on individual molars• Literature Cited we cannot critically contrast size of O. dollmani and O. petteri. Still, certain Allen, G. M. 1939. A checklist of African mam- mals.•Bulletin of the Museum of Compara- characteristics of O. petteri noted by tive Zoology, Harvard 83:1-763. Denys•the low crown height and in- , & A. Loveridge. 1933. Reports on the complete fusion of cusps in the posterior scientific results of an expedition to the M3 laminae•do not suggest examples of southwestern highlands of Tanganyika Terri- O. dollmani. In our view, O. dollmani tory. II. .•Bulletin of the Museum of Comparative Zoology at Harvard College represents an in situ differentiation un- 75:47-140. ique to the Mathews Range, Kenya, and Bohmann, L., von. 1943. Zwei neue O/omj'.ç-Rassen is certainly a species distinct from O. aus Ostafrika.•^Zoologischer Anzeiger 143: petteri as known from the early Pleisto- 153-155. cene of northern Tanzania (If demon- . 1952. Die afrikanische Nagergattung Ot- omys F. Cuvier.•Zeitschrift für Säugetier- strated to be synonymous, dollmani has kunde 18:1-80. priority over petteri). The retention of 6 Bookstein, F. L. 1991. Morphometric Tools for M3 laminae in O. dollmani intimates that Landmark Data. Cambridge University it is a relict surviving from the early Press, Cambridge, xvii + 435 pp. cladogenesis of Otomys in eastern Africa. , B. Chernoff, R. L. Eider, J. M. Humphries, Jr., G. R. Smith, & R. E. Strauss. 1985. According to this interpretation, O. doll- Morphometrics in Evolutionary Biology.• mani is predicted to be a cladistically The Academy of Natural Sciences of Phila- primitive outlier relative to eastern Afri- delphia, Special Publication 15:xvii + 277 pp. 512 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

Bowie, R. C. K., J. Fjeldsâ, S. J. Hackett, & T. M. Denys, C. 1989. Phylogenetic affinities of the oldest Crowe. 2004. Molecular evolution in space East African Otomys (Rodentia, Mammalia) and through time: mtDNA phylogeography from Olduvai Bed I (Pleistocene, Tanza- of the Olive Sunbird (Nectarinia olivaceall nia).•Neues Jahrbuch für Geologie und obscura) throughout continental Africa.• Paléontologie, Monatshefte 44:705-725. Molecular Phylogenetics and Evolution Dieterlen, F. 1968. Zur Kenntnis der Gattung 33:56-74. Otomys (Otomyinae; Muridae; Rodentia). , , , J. M. Bates, & T. M. Beiträge zur Systematik, Ökologie und Bio- Crowe. 2006. Coalescent models reveal the logie zentralafrikanischer Formen.•Zeits- relative roles of ancestral polymorphism, chrift für Säugetierkunde 33:321-352. vicariance, and dispersal in shaping phylo- , & E. Van der Straeten. 1992. Species of the geographical structure of an African montane genus Otomys from Cameroon and Nigeria forest robin.•Molecular Phylogenetics and and their relationship to East African Evolution 38:171-188. forms.•Bonner Zoologische Beiträge 43: Carleton, M. D., & G. G. Musser. 1984. Muroid 383-392. rodents. Pp. 289-379. in S. Anderson and J. DoUman, G. 1915. On the swamp-rats (Otomys) of K. Jones, Jr., eds.. Orders and families of East Africa.•Annals and Magazine of Nat- Recent mammals of the world. John Wiley ural History, series 8, 15:149-170. and Sons, New York, 686 pp. Dowsett, R. J. 1986. Origins of the high-altitude , J. Kerbis Peterhans, & W. T. Stanley. 2006. avifaunas of tropical Africa. Pp. 557-585. in Review of the Hylomyscus denniae group F. Vuilleumier and M. Monasterio, eds.. High (Rodentia: Muridae) in eastern Africa, with altitude tropical biogeography. Oxford Uni- comments on the generic allocation of Epimys versity Press, New York-Oxford, x + 649 pp. endorohae Heller.•Proceedings of the Bi- Ducroz, J.-F., V. Volobouev, & L. Granjon. 2001. ological Society of Washington 119:293- An assessment of the systematics of arvi- 325. canthine rodents using mitochondrial DNA sequences: evolutionary and biogeographical Chevret, P., C. Denys, J. J. Jaeger, J. Michaux, & F. implications.•Journal of Mammalian Evo- Catzeflis. 1993. Molecular and paleontologi- lution 8:173-206. cal aspects of the tempos and mode of EUerman, J. R. 1941. The families and genera of evolution in Otomys (Otomyinae: Muridae: living rodents. Volume 2, Family Muridae. Mammalia).•Biochemical Systematics and London: British Museum (Natural History), Ecology 21:123-131. xii + 690 pp. Clausnitzer, V. 2001. Rodents of the afro-alpine , T. C. S. Morrison-Scott, & R. W. Hayman. zone of Mt. Elgon. Pp. 427^43. in C. Denys, 1953. Southern African mammals 1758 to L. Granjon, and A. Poulet, eds., African 1951: a reclassification. British Museum small mammals. IRD Editions, Collection (Natural History), London, 363 pp. colloques et séminaires, Paris, 570 pp. Fjeldsâ, J., & J. C. Lovett. 1997. Geographic , & R. Kityo. 2001. Altitudinal distribution of patterns of old and young species in African rodents (Muridae and Gliridae) on Mt. Elgon, forest biota: the significance of specific Uganda.•Tropical Zoology 14:95-118. montane areas as evolutionary centres.• Corbet, G. B., & J. E. Hill. 1980. A world hst of Biodiversity and Conservation 6:325-346. mammalian species. British Museum (Natural Griffiths, C. J. 1993. The geological evolution of History), London, viii + 226 pp. East Africa. Pp. 9-21. in J. C. Lovett and S. , & . 1986. A world list of mammalian K. Wasser, eds., Biogeography and Ecology species. Second ed. British Museum (Natural of the Rain Forests of Eastern Africa. Cam- History), London, 254 pp. bridge University Press, Cambridge, ix + 341 , & . 1991. A world hst of mam- pp. malian species. Third ed. British Museum Grimshaw, J., N. Cordeiro, & C. Foley. 1995. The (Natural History) Publications, London, viii mammals of Kilimanjaro.•Journal of East + 243 pp. African Natural History 84:105-139. de Graaff, G. 1981. The rodents of Southern Africa. Hall, J. P. 2005. Montane speciation patterns in Butterworths, Durban, 267 pp. Ithomiola butterflies (Lepidoptera: Riodini- Delany, M. J. 1975. The rodents of Uganda. dae): are they consistently moving up in the Trustees of the British Museum (Natural world?•Proceedings of the Royal Society B History), London, 165 pp. 272(1580):2457-2466. Demeter, A., & R. Hutterer. 1986. Small mammals Hatt, R. T. 1934. Fourteen hitherto unrecognized from Mt. Meru and its environs (Northern African rodents.•American Museum Novi- Tanzania).•Cimbebasia 8:199-207. tates 708:1-15. VOLUME 119, NUMBER 4 513

. 1940. Lagomorpha and Rodentia other Mayr, E. 1942. Systematics and the Origin of Species, than Sciuridae, Anomaluridae and Idiuridae, From the Viewpoint of a Zoologist. Columbia collected by the American Museirm Congo University Press, New York, xiv + 334 pp. Expedition.•Bulletin of the American Mu- Meester, J. A. J., I. L. Rautenbach, N. J. Dippenaar, seirm of Natural History 76:457-604. & C. M. Baker. 1986. Classification of Hedberg, O. 1986. Origins of the afroalpine flora. southern African mammals.•Transvaal Mu- Pp. 443^68. in F. Vuilleumier and M. seum Monograph 5:1-359. Monasterio, eds.. High altitude tropical bio- Misonne, X. 1974. Rodentia, main text. Part 6 in J. geography. Oxford University Press, New Meester and H. W. Setzer, eds.. The mam- York-Oxford, x + 649 pp. mals of Africa, an identification manual. Heller, E. 1912. New rodents from British East Smithsonian Institution Press, Washington, Africa.•Smithsonian Miscellaneous Collec- D.C., 39 pp (not continuously paginated). tions 59:1-20. Moreau, R. E. 1966. The bird faunas of Africa and HoUister, N. 1919. East African mammals in the its islands. Academic Press, New York, 424 United States National Museum. Part II. pp. Rodentia, Lagomorpha, and Tubuliden- Musser, G. G., & M. D. Carleton. 1993. Family tata.•Bulletin of the United States National Muridae. Pp. 501-755. in D. E. Wilson and Museum 99(2):1-184. D. M. Reeder, eds.. Mammal species of the Honacki, J. H., K. E. Kinman , & J. W. Koeppl World, a taxonomic and geographic refer- (eds.). 1982. Mammal species of the world, ence. Second Edition. Smithsonian Institution a taxonomic and geographic reference. Allen Press, Washington D.C., 1206 pp. Press, Inc., and The Association of Systematic , & . 2005. Superfamily Muroidea. Pp. 894-1531. in D. E. Wilson and D. M. Collections, Lawrence, Kansas, 694 pp. Reeder, eds.. Mammal species of the World, Huxley, J. S. (ed.). 1940. The New Systematics. a taxonomic and geographic reference. Third Clarendon Press, Oxford, viii + 583 pp. Edition. Johns Hopkins University Press, Kingdon, J. 1974. East African mammals: an Baltimore, Vol. 2: xviii + 745-2142 pp. atlas of evolution in Africa. (Hares and Osgood, W. H. 1910. Diagnoses of new East African Rodents). Academic Press, London, 2B:343- mammals, including a new genus of Mur- 704 + Ivii. idae.•Field Museum of Natural History, Knox, E. B., & J. P. Palmer. 1995. Chloroplast Zoological Series 10:5-13. DNA variation and the recent radiation of Petter, F. 1982. Les parentes des Otomys du Mont the giant senecios (Asteraceae) on the tall Oku (Cameroun) et des autres formes rap- mountains of eastern Africa.•Proceedings of portées a O. irroratus (Brants, 1827) (Roden- the National Academy of Sciences 92: tia, Muridae).•Bonner Zoologische Beiträge 10349-10353. 33:215-222. Lavrenchenko, L. A., A. N. MiUshnikov, V. M. Pocock, T. N. 1976. Pliocene mammalian micro- Aniskin, A. A. Warshavsky, & W. Gebrekidan. fauna from Langebaanweg: a new fossil 1997. The genetic diversity of small mammals genus linking the Otomyinae with the Mur- of the Bale Mountains, Ethiopia.•SINET: inae.•South African Journal of Science Ethiopian Journal of Science 20:213-233. 72:58-60. Lawrence, B., & A. Loveridge. 1953. Zoological . 1987. Plio-Pleistocene fossil mammalian results of a fifth expedition to East Africa. I. microfauna of southern Africa•a preliminary Mammals from Nyasaland and Tete. With report including description of two new fossil notes on the genus Otoniys.•Bulletin of the muroid genera (Mammalia: Rodentia).•Pa- Museum of Comparative Zoology at Harvard laeontologia Africana 26:69-91. College 110:1-80. Rensch, B. 1929. Das Prinzip geographischer Lovett, J. C. 1993. Climatic history and forest Rassenkreise und das Problem der Artbil- distribution in eastern Africa. Pp. 23-32. in J. dung. Bomtraeger, Berlin. C. Lovett and S. K. Wasser, eds., Biogeogra- Roberts, A. 1951. The mammals of South Africa. phy and Ecology of the Rain Forests of Trustees of "The mammals of South Africa" Eastern Africa. Cambridge University Press, book fund, Johannesburg, 700 pp. Cambridge, ix + 341 pp. Roy, M. S., P. Arctander, & J. Fjeldsâ. 1998. Marcus, L. F. 1990. Traditional morphometrics. Speciation and taxonomy of montane green- Pp. 77-122. in F. J. Rohlf and F. L. Book- buls of the genus Andropadus (Aves: Pycno- stein, eds.. Proceedings of the Michigan notidae).•Steenstrupia 24:51-66. Morphometrics Workshop. The University , J. M. Cardoso da Silva, P. Arctander, J. of Michigan Museiun of Zoology, Special Garcia-Moreno, & J. Fjeldsâ. 1997. The Publication Number 2, viii + 380 pp. speciation of South American and African 514 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON

birds in montane regions. Pp. 325-3434. in D. , J. Meester, & T. Kearney. 1993. The P. Mindell, ed., Avian Molecular Evolution taxonomic status of Saunder's vlei rat, and Systematics. Academic Press, New York, Otomys saundersiae Roberts (Rodentia: Mur- XX + 382 pp. idae: Otomyinae).•Journal of African Zool- Senegas, F. 2001. Interpretation of the dental ogy 107:1-26. pattern of the South African fossil Euryot- Thomas, O. 1897. On the genera of rodents: an omys (Rodentia, ) and origin of attempt to bring up to date the current otomyine dental morphology. Pp. 151-160. arrangement of the order.•Proceedings of in C. Denys, L. Granjon, and A. Poulet, eds., the Zoological Society of London [1896]: African small mammals. IRD Editions, Col- 1012-1028. lection Colloques et Séminaires, Paris, 570 pp. . 1900. List of mammals obtained by Mr. H. , & M. Avery. 1998. New evidence for the murine J. Mackinder during his recent expedition to origins of the Otomyinae (Mammalia, Rodentia) Mount Kenya, British East Africa.•Proceed- and the age of Bolt's Farm (South Africa).• ings of the Zoological Society of London South African Journal of Science 94:503-507. 1900:173-180. Setzer, H. W. 1953. Four new mammals from the . 1902. On some new forms of Otomys.• Anglo-Egyptian Sudan.•^Journal of the Wash- Annals and Magazine of Natural History, ser. ington Academy of Sciences 43(10):333-335. 7, 10:311-314. . 1956. Mammals of the Anglo-Egyptian . 1918. A revised classification of the Sudan.•Proceedings of the United States Otomyinae, with descriptions of new genera National Museum 106(3377):447-587. and species.•Annals and Magazine of Nat- Smithers, R. H. N. 1983. The mammals of the ural History, ser. 9, 2:203-211. Southern African Subregion. University of Watts, C. H. S., & P. R. Baverstock. 1995. Pretoria, Republic of South Africa, 736 pp. Evolution in some African Murinae (Roden- Stuart, S. N., F. P. Jensen, S. Brogger-Jensen, & R. tia) assessed by microcomplement fixation of I. Miller. 1993. The zoogeography of the albumin.•Journal of African Zoology 109: montane forest avifauna of eastern Tanzania. 423^33. Pp. 203-228. in J. C. Lovett and S. K. Wasser, White, F. 1978. The afromontane region. eds., Biogeography and ecology of the rain Pp. 463-513. in M. J. A. Werger, ed., forests of eastern Africa. Cambridge Univer- Biogeography and ecology of southern sity Press, Cambridge, ix + 341 pp. Africa. Junk, The Hague, xv + 1439 pp. SYSTAT for Windows. 2002. Systat Software Inc., . 1981. The history of the Afromontane Version 10.2.01. archipelago and the scientific need for its Taylor, P. J. 1998. The smaller mammals of conservation.•African Journal of Ecology KwaZulu-Natal. University of Natal Press, 19:33-54. Pietermaritzburg, 139 pp. Wroughton, R. C. 1906. Notes on the genus , & A. Kumirai. 2001. Craniometric relation- Otomys.•^Annals and Magazine of Natural ships between the Southern African vlei rat, History, ser. 7, 18:264-278. Otomys irroratus (Rodentia, Muridae, Oto- . 1910. Servals and an Otomys from East myinae) and allied species from north of the Africa.•Annals and Magazine of Natural Zambezi River. Pp. 161-181. in C. Denys, L. History, ser. 8, 5:205-207. Granjon, and A. Poulet, eds., African small mammals. IRD Editions, Collection Collo- Associate Editor: Gary R. Graves ques et Séminaires, Paris, 570 pp. , C. Denys, & M. Mukerjee. 2004a. Phytogeny of the African murid tribe (Roden- Appendix 1 tia), based on morphological and allozyme evidence.•Zoológica Scripta 33:389^02. Listed below are additional specimens of Ot- , A. Kumirai, & G. Contrafatto. 2004b. omys that formed the basis for the morphological Geometric morphometric analysis of adaptive comparisons, sample statistics, morphometric cranial evolution in southern African lami- analyses, and counts of M3 laminae presented nate-toothed rats (Family: Muridae, Tribe: herein. Otomyini).•Durban Museum Novitates 29:110-122. Otomys irroratus.•South Africa: Western Cape , , & . 2005. Species with fuzzy Province, 2 mi N Clanwilliam (USNM 343731- borders: the taxonomic status and species 343735); Goudveld, 10 mi N Knysna (USNM limits of Saunders' vlei rat, Otomys saunder- 343761-343776); Klein Brakrivier, 9 km N Mossel- siae Roberts, 1929 (Rodentia, Muridae, Oto- baai (USNM 343743-343761); Kluitjieskraal, 7 mi myini).•Mammalia 69:297-322. SW Swellendam (USNM 343741, 343742); Pakhuis VOLUME 119, NUMBER 4 515

Pass (USNM 343739, 343740); 6 mi WNW Rede- Otomys typus.•Ethiopia: Arussi Plateau, Albasso linghuys (USNM 469704^69707); 22 mi SW (USNM 259506); Gojam, N'jabara (USNM Worcester (USNM 469695^69703). South Africa: 259507); Bale Mts, Dinsho (MCZ 57319); Simien Eastern Cape Province, 19 mi S Aliwal North Mts, Mt Geech, 11,200 ft (MCZ 26992); Simien (USNM 344783-344788); Coloniesplaats, 28 mi N Mts, 10,000 ft (MCZ 34372); Simien Mts, Devark, Graaf Reinet (USNM 344795, 344796); 2 mi E 9200 ft (MCZ 34373); Arussi (FMNH 28125, 28126, Houtbosdorf (USNM 382365-382373); Kenkel 28129, 28132-28139, 28159-28164); Gojam (FMNH Bosch (USNM 221390-221399, 221400, 221401, 28141-28143, 28172-28177, 28179); Shou (FMNH 221415); Pirie, 12 mi NW King Williams Town 28140); Sidamo, (FMNH 28167); Simien Mts (USNM 344774-344782); 4 mi W Port St. Johns (FMNH 28146-28151, 28153-28158). (USNM 381220-381228); 5 mi W Somerset East Otomys uzungwensis.•Malawi: Nyika Plateau, (USNM 452460^52462). South Africa: KwaZulu- 7000 ft (MCZ 43947, 43950). Tanzania: Udzungwa Natal Province, 2 mi W Gardens Mts, Kigoro (MCZ 26637, 26638, 26640-26644); (USNM 381185-381187, 381207-381219); Kilgob- Udzungwa Mts, Dabaga (MCZ 26645 [holotype], bin, 8 mi W Dargle Station (USNM 381195- 26646, 26648-26653, 26655, 26656, 26664); Poroto 381197); Sani Pass (USNM 381198-381206). Mts, Mbeya, 9000 ft (MCZ 51147).