Holotype

Field Museum of Natural History number 175934, an adult male prepared as skin, skull, and partial skeleton, collected 10 December 2002 by Steven M. Goodman original number SMG 13,425.

The round skin and skull are in fine condition. External data recorded on the skin tag include: TOTL, 335 mm; HBL, 150 mm; TL, 179 mm; HFL, 30 mm; EL, 26 mm; and WT, 91 gm. The animal was noted as having scrotal testes 12 × 5 mm with convoluted epididymides; tissues were preserved in EDTA ethylene diamine tetraacetic acid.

Type Locality

Madagascar, Fianarantsoa Province, Parc National de l'Isalo, 28 km SE Berenty-Betsileo, along Sahanafa River near foot of Bevato Mountain, 650 m; geographic coordinates: 22°19.0′S, 45°17.6′E.

Diagnosis

A species of Eliurus characterized in combination by a gray dorsal pelage, contrasting grayish-white venter, and a distinctive bright-white tail tuft fig. 1. Cranial size moderately large ONL ca. 37–41 mm; figs. 2, 3, molar rows absolutely and relatively long LM1–3 ca. 6.0–6.5 mm, incisive foramina relatively long and wide LIF/LD ca. 56%, and ectotympanic bullae large for the genus. Alveolus of lower incisor short, not forming a capsular process.

Referred Specimens

Three additional specimens are known, all from the PN de l'Isalo: FMNH 175933, a young adult male also collected by Goodman SMG 13,387 in the same month 5 Dec from 24 km W Ranohira bas, Andranohavo Canyon des Rats, 700 m, 22°28.9′S, 45°22.9′E; UADBA 10483, a young adult female obtained 2 April 1995, A. F. A. Hawkins FH 3, from a site within 300 m of where the holotype was obtained Georges Rabeony, personal commun.; and UADBA 46805 obtained between 2 and 9 February 1995 Hawkins, 1995, by A. F. A. Hawkins FH 23 in the Canyon des Rats.

Distribution

To date known only from the PN de l'Isalo, south-central Madagascar.

Description

Fur texture soft and fine. Cover hairs 9–10 mm in length over middle rump. Basal four-fifths of dorsal cover hairs bright plumbeous gray, distal one-fifth medium buff, some hairs tipped with a third band of faint dusky brown. Guard hairs dark brown to blackish, only slightly longer than the cover hairs except over the rump. General effect of upperparts a pale brownish gray; forehead, face, and lower legs a purer medium gray. Boundary of dorsal and ventral pelage clearly demarcated along sides. Ventrum variable in coverage of white hairs: entirely bright buff white UADBA 10483; or dull buff white over the chin, throat, and chest with some basal gray thinly showing over the lower abdomen and inguinum FMNH 175934; or wholly white hairs confined to the chin and throat, cover hairs over the chest and abdomen bicolored basal half gray, distal half white, and the underparts overall light grayish white FMNH 175933. Medial dusky brown band extends over the tarsus and proximal metatarsus; remainder of metatarsus and toes, as well as tops of forefeet, covered with bright white hairs; ungual tufts at base of claws well developed.

Caudal tuft prominent over distal half of tail length, tuft hairs becoming gradually longer toward the tip and measuring about 12–15 mm long over the terminal portion fig. 1. Terminal one-third of tail length covered with pure white hairs, forming a brilliant white tuft accentuated by the intensely black hairs of the proximal tuft distal white segment = 35 mm in UADBA 10483, 45 mm in FMNH 175934, and 75 mm in FMNH 175933. Proximal half of tail clothed with short black hairs that do not mask the caudal scales; scalation moderately textured; epidermis pale gray to whitish toward the base of the tail, lighter underneath, but paleness visually obscured by the black hairs.

Hind feet short and broad. Six plantar pads arranged as characteristic of the genus; pads not so bulbous and crowded as in E. antsingy. Pinnae medium-sized, more rounded, not so elongate as E. tanala Major 1896; color pale dusky brown externally, somewhat lighter internally, and sparsely clothed with fine brown hairs inside and outside.

Cranium moderately large ONL = 37–41 mm; figs. 2, 3, agreeing with largest geographic examples of E. majori e.g., from PN de Marojejy, smaller than E. antsingy and large-bodied E. tanala e.g., from RS d'Anjanaharibe-Sud. Dorsal profile of skull arched fig. 3, sloping gradually toward the rostrum and more steeply toward the occiput, not so flat as in E. tanala of similar size. Rostrum proportionally long LR/ONL = 36.2% for the genus, comparable to relative length in the E. tanala complex but longer than E. myoxinus or E. majori. Braincase rounded and smoothly contoured; subsquamosal fenestra patent and medium-sized, exposing the interior braincase and dorsal periotic; hamular process of the squamosal stout but well delineated. Zygomatic arches almost parallel-sided in the three specimens; zygomatic plate medium in width for the genus, forming a shallow but distinct dorsal notch, more incised than in E. penicillatus and E. majori. Incisive foramina bluntly pointed on their posterior ends, more acute at the anterior; wide and relatively long LIF/LD = 56% for the genus, in proportion resembling E. antsingy and the E. majori complex fig. 2. Posterior palatine foramina oval, situated within the maxillary-palatine suture at the level of the M1-M2; supernumerary palatal foramina occurring in the palatine bones, but minute in size and irregularly formed; slitlike palatal vacuities absent. Posterior margin of the bony palate terminates just short of the ends of the third molars; anterior expanse of mesopterygoid fossa broad and horseshoe-shaped, its sides gently incurved. Alisphenoid strut present on both sides in all three individuals examined. Auditory bullae relatively large for the genus, the ectotympanic covering nearly all of the periotic in ventral view, even more so than in E. antsingy and E. myoxinus. Mandible stocky in form, ascending ramus deep.

Enamel face of upper and lower incisors dull yellowish-orange in color. Alveolus of lower incisor relatively short, terminating posteriorly at the level of the coronoid process and well below the sigmoid notch; tip of incisor root indicated by slight mound on mandibular ramus but not forming a projecting capsular process. Toothrows long LM1-3 = 6.0–6.5 mm relative to cranial size, in proportion like those of E. majori and E. penicillatus; molars robust with laminate occlusal configuration as per the genus, higher crowned than most Eliurus but again approximating E. majori and E. penicillatus in this detail; upper and lower third molars about equal in length to second molars or only slightly shorter.

Other external, cranial, and dental features as described for the genus Carleton, 1994; Carleton and Goodman, 1998.

Morphological and Morphometric Comparisons

The cardinal traits of E. danieli implicate the eastern species E. majori and E. penicillatus or western E. antsingy as its closest relatives. All four are relatively large species of Eliurus and possess long incisive foramina that span over half the diastema fig. 2 and short lower incisors that lack a prominent capsular process. In combination, these qualitative features serve to distinguish the four from all other species of Eliurus see Carleton, 1994, Carleton and Goodman, 1998, and Carleton et al., 2001, for discussion of interspecific variation in these characters.

Other character differences associate E. danieli more closely to E. majori and E. penicillatus instead of E. antsingy. Foremost, specimens of E. danieli have long, robust molar rows in proportion to cranial size LM1-3 ca. 16–17% of ONL, a configuration that agrees with E. majori and E. penicillatus Carleton, 1994. In contrast, examples of E. antsingy have shorter toothrows in relation to cranial size LM1-3 ca. 13–14% of ONL, resembling in this regard all other species of Eliurus Carleton et al., 2001. A simple bivariate plot underscores the segregation of the two groups by dental proportion fig. 4; only one stray individual an E. danieli overlaps the otherwise clearly delineated data-point constellations. The morphogenetic determination of these dental shape contrasts is such that molar variables LM1-3, WM1 always contribute most heavily to group separation on the first factor extracted in all ordinations of our cranial measurements, whether principal component not illustrated or canonical variate fig. 5, table 2. Eliurus danieli, together with E. majori and E. penicillatus, also differ from E. antsingy in fur texture not so fine and soft, plantar pads not as large, bulbous, and close-set, size of upper and lower third molars not smaller than second molars, and molar hypsodonty not as low-crowned. These impressionistic differences are not so easily quantifiable as is the disparity in toothrow proportions.

Figure 4

Scatter plot of length of the upper molar row LM1-3 versus greatest length of skull ONL for intact specimens of E. antsingy, E. danieli, E. majori, and E. penicillatus. Note that examples of the latter three species possess longer toothrows in relation to cranial size compared with those of E. antsingy.

Figure 5

Results of discriminant function analysis performed on 18 log-transformed craniodental variables, as measured on 130 specimens representing 13 OTUs of Eliurus see Materials and Methods. Top, projection of individual scores onto the first and second canonical variates extracted. Bottom, projection of individual scores onto the first and third canonical variates extracted. The maximally inclusive polygons enclose specimen scores around an OTU's centroid; in view of extensive overlap and visual confusion, the sample polygons for E. antsingy 4 OTUs and E. majori 7 OTUs were collapsed to enclose maximal dispersion around the species' grand centroid. See table 2 for variable correlations and percent variance explained.

Table 2

OTUs represented include four of E. antsingy, N = 24; one of E. danieli, N = 3; seven of E. majori, N = 90; and one of E. penicillatus, N = 13. See Materials and Methods and fig. 4

Eliurus danieli can be distinguished from E. majori and E. penicillatus by means of the following traits. Although similar in size to E. majori, specimens of E. danieli possess a conspicuous white caudal tuft. This trait is unknown in nearly all geographic samples of E. majori, in which the tuft is wholly dark brown or black to the tip Carleton, 1994; Goodman and Carleton, 1996. In the PN de Marojejy, however, Carleton and Goodman 2000 reported that nine of the 20 E. majori specimens collected possess a white caudal tip of variable extent. The white tail tip in those particular E. majori, however, is not accentuated by a contiguous band of dense black hairs as found in E. danieli. Furthermore, in other features and in morphometric analyses, the Marojejy sample unambiguously accords with typical E. majori Carleton and Goodman, 2000; also see next, not E. danieli. Although possessing a white caudal tuft as diagnostic of E. penicillatus, specimens of E. danieli are conspicuously larger in nearly all variables measured figs. 2, 3; table 1; in view of that great size disparity, the two species are strongly circumscribed in morphometric space fig. 5. Finally, in contrast to E. majori and E. penicillatus, E. danieli has gray dorsal pelage, bright in tone in both eastern species, dorsum blackish gray to blackish brown and somber colored; its venter is bright buff white or grayish white and clearly delineated from the dorsum in both eastern species, venter paler but similarly colored to the dorsum and dorsal-ventral pelage contrast less well marked; its ectotympanic bulla is absolutely larger table 1, obscuring nearly all of the periotic in ventral view in both eastern species, ectotympanic less inflated, exposing posteromedial wedge of the periotic; and its zygomatic plate projects forward to form a distinct dorsal notch in both eastern species, zygomatic plate narrower and dorsal notch weakly incised.

In projections of the first three canonical variates cumulative variation explained = 85.5% extracted from 13-group discriminant function analysis, the three specimens from PN de l'Isalo are wholly isolated in morphometric space from all other OTUs and species fig. 5. As commented above, the two molar variables LM1-3, WM1 most strongly influence separation of groups along the first canonical variate table 2. Other dimensions that load moderately along this factor include those that reflect palatal breadth BM1s, PPW and interorbital width IOB; compared with E. antsingy, the palate is relatively broader and interorbit narrower in E. danieli, E. majori, and E. penicillatus table 1. Size as a general factor is expressed along the second canonical variate, as indicated by the positive and moderately strong correlations Pearson's R = 0.6–0.8, P ≤ .001 of most variables with this factor table 2. In general size, specimens of E. danieli match the larger samples of E. majori and E. antsingy and conspicuously exceed the diminutive E. penicillatus in most dimensions quantified table 1. Between-group variation summarized by the third canonical variate conveys the morphometric association of E. danieli and E. penicillatus fig. 5. The few cranial variables BIF, LD, LBP, DAB, BZP that account for this association constitute proportional subtleties that are not readily apparent in side-by-side skull comparisons of the two species due to their gross size disparity; notable among these are the relatively spacious incisive foramina BIF and short diastema LD. Although visually unimpressive, such proportional features nevertheless influence the phenetic pairing of the E. danieli and E. penicillatus samples among all OTUs represented, as indexed by Mahalanobis distances between group centroids fig. 6. The multivariate level of divergence between the two Mahalanobis D² = 52.3 surpasses that derived for the most strongly differentiated samples within either E. majori or E. antsingy.

Figure 6

Phenogram produced from clustering UPGMA of Mahalanobis' distances derived from discriminant function analysis of 13 OTUs representing four species of Eliurus see fig. 5.

Ecological Notes

The PN de l'Isalo is located on the Isalo Massif, a vast area of badlands composed of ruiniform Jurassic sandstone. Much of the park includes rocky habitat with sparse short scrub or completely lacking any plants. Many deep canyons found within this formation contain permanent rivers that are in part responsible for the steeply cut valley sides, and the precipitous facies have a unique rupicolous plant community. Gallery forest can be found along river margins at the base of these canyons and at their mouths. An evergreen sclerophyllous forest formation dominated by the genus Uapaca Euphorbiaceae covers more gently sloping hillsides. Although the region receives around 850 mm of rain per year and has a pronounced dry season Nicoll and Langrand, 1989, the water-fed canyons retain moist conditions and support vegetation whose floral affinity is transitional between humid and deciduous formations Gautier and Goodman, 2003. Within these moist canyons occur numerous animals that represent taxa or species groups otherwise known from the eastern humid forests.

Microhabitats where specimens of E. danieli were captured are noted on skin tags or in original specimen catalogs as follows: “On rocky cliff about one-half way up Bevato Mt. in zone with little woody vegetation. Trapped on ground in small hole leading into rock fall” FMNH 175934, holotype; “on the rocky mountain of Sahanafa” [translated from the French] UADBA 10483; “In canyon with vegetation at base of rock face and along river banks—sandy soil. Trap on ground under opening of large boulder at river edge and surrounded by ferns and dense rupicolous vegetation” FMNH 175933; and “in the interior of the Canyon des Rats, a site of exposed rock and at a level more than 30 m above the water [= river] level” [translated from the French] UADBA 46805. This information suggests that the species is at least in part terrestrial and lives outside forest formations.

On the basis of considerable field experience with members of Eliurus, Goodman has never captured an individual in an ecological setting similar to that of the E. danieli holotype. The site at the foot of Mt. Bevato is a broad expanse of exposed pseudosteppe, on a steep rocky slope, and approximately 500 m from the nearest forest. According to M. George Rabeony, an employee of the park service who has accompanied numerous researchers in the PN de l'Isalo, this is also the locale where UADBA 10483 was trapped in April 1995. In late December 2002, Goodman returned to this site and placed thirty traps 10 Nationals and 20 Shermans on the ground under large rock outcrops between the forest edge and the middle portion of Mt. Bevato. The individual FMNH 175934 was captured in one of the highest-elevation traps set along the Bevato cliff face. Other mammals captured in this trap line include the introduced Rattus rattus and the large tenrecid Tenrec ecaudatus. A few individuals of E. myoxinus were trapped within the Sahanafa Forest, about 1.5 km from where the holotype of E. danieli was obtained. In the Andranohavo Forest Canyon des Rats, E. myoxinus and R. rattus were common and captured within a short distance of E. danieli.

Taxonomic and Biogeographic Remarks

The phenetic union of E. danieli and E. penicillatus and their joint possession of a white tail tuft, largely to the exclusion of E. majori, offer a provisional hypothesis of sister-group relationship. Nonetheless, we are not persuaded that this cognate relationship will be necessarily sustained by other data sources, such as gene-sequence studies. First, with only four specimens at hand, the sample size of E. danieli is too small to adequately represent variational bounds in the species population. Second, as enumerated above, E. majori and E. penicillatus do share a number of morphological similarities not found in E. danieli. In view of their distribution in eastern humid forest, the possibility that E. majori and E. penicillatus form a clade vicariantly related to E. danieli in west-central Madagascar remains as viable an interpretation and see below.

Compared with other Eliurus that have a white caudal tuft E. grandidieri, E. penicillatus, E. tanala, that possessed by E. danieli is the most visually striking. As we currently understand interspecific affinities see Discussion and assuming that an all-dark tuft represents the primitive state, a terminal white tuft has evolved at least three times within the genus. The significance of these contradistinctive caudal flags to social behavior of the species or in interspecific encounters deserves investigation.

Although the Isalo Massif today receives only moderate amounts of rainfall per year and experiences a pronounced dry season, organisms identified from regional paleontological sites supply ample biotic evidence that local climatic conditions were distinctly more mesic during the Late Pleistocene and Holocene Burney, 1997; Goodman and Rakotozafy, 1997; Godfrey et al., 1999. Relicts from this more mesic phase of late Quaternary climate persist in the modern flora of the Isalo area, most notably in the deep canyons where many plants show phytogeographic affinities to the eastern humid forests. The same relictual biogeographic pattern has also emerged from the study of certain birds, reptiles, and amphibians that inhabit the Isalo Massif e.g., Raxworthy and Nussbaum, 1996, 1997; Fjeldsa et al., 2000. Two kinds of taxonomic evidence have been identified among those vertebrate groups: 1 the same species is found in common between Isalo and sites in the eastern humid forests; or 2 closely related taxa occur in these two areas and are presumed to have originated through east-west vicariance. The presence of E. danieli in the Isalo region and its proposed sister-group relationship to E. majori and/or E. penicillatus exemplify the second line of evidence for a former east-west continuity of humid forest.

Etymology

We name this new species to honor Professor Daniel Rakotondravony, Département de Biologie Animale, Université d'Antananarivo, especially for his contributions to our knowledge on Malagasy rodents and for helping to foster zoological research on Madagascar for numerous national and foreign scientists.

Eliurus majori Group E. majori Thomas, 1895; E. penicillatus Thomas, 1908; E. danieli Carleton and Goodman, 2006

As reviewed above, these three species constitute a strongly circumscribed morphologic subset within the genus. They are moderately large in size and the hind feet comparatively short. The tail is well furred, but the caudal hairs gradually lengthen toward the tip rather than abruptly elongate as an expansive terminal tuft. Two of the three species bear a white caudal tip E. danieli, E. penicillatus, but certain northern populations of the third E. majori vary in the incidence of white tipping PN de Marojejy–Carleton and Goodman, 2000. Their molar rows are exceptionally robust with the third molars subequal to the second, the incisive foramina are wide and long > 50% of diastemal length, and the lower incisor is short and does not form a protruding capsular process see Carleton, 1994: fig. 20. Carleton 1994 thought that E. majori and E. penicillatus retained several plesiomorphic traits and his cladistic interpretation represented the two as a basal clade within Eliurus; however, the molecular results of Jansa et al. 1999 indicate that, although E. majori and E. danieli, reported as “sp. A” subtended the E. tanala and E. myoxinus groups, E. grandidieri-Voalavo composed the most basal cladistic position among eliurine Nesomyinae.

The species complex is principally confined to montane and sclerophyllous montane forest settings in the eastern humid forest Carleton, 2003. Size within E. majori is fairly uniform over much of its geographic range, from the southern Anosyenne Mountains in the south to the Northern Highlands, but the northernmost population on the isolated peak Montagne d'Ambre is diminutive. With skulls arrayed side-by-side, the Montagne d'Ambre animals “look” like a separate species, but shape features and genetic data clearly affiliate this montane outlier with E. majori proper Carleton and Goodman, 1998; Jansa et al., 1999. Extralimital populations have been also uncovered in remnant montane forests of the Central High Plateau Jansa et al., 1999; Carleton et al., 2001. That from the PN de l'Isalo is herein diagnosed as a new species based on recently collected material. Still other western populations of uncertain status possess majori-like features, namely those from the RS d'Ambohijanahary and RS d'Ambohitantely appendix 1; these specimens have an all-dark tail tuft and approximate the Montagne d'Ambre sample of E. majori in their smaller size.

Carleton 1994 doubtfully retained E. penicillatus as distinct from E. majori, but the accumulation of larger samples of the latter species, now available over a broad north-south geographic swath, continues to underscore the clear-cut morphometric separation of the two Carleton and Goodman, 2000; this study. Jansa et al.'s 1999 cytochrome-b genetic evidence divulged close relationship of the Isalo species, E. danieli called “sp. A”, and E. majori, whereas our morphometric evaluation identified E. penicillatus as an alternative possibility of cognate affinity. Jansa et al. 1999 lacked tissues for E. penicillatus, which until recently has been known only from the type locality, Ampitambe, and all specimens collected by Forsyth Major in 1895–1896 see Jenkins and Carleton, 2005. A lone specimen UADBA 16209 collected in 2000 from Ankerana, about 35 km NE of Fandriana, represents only the second locality that we consider definitely referable to E. penicillatus, but it supplies evidence that the species still survives. Ampitambe is about 45 km SE Fandriana see Carleton and Schmidt, 1990, for discussion. Perhaps the distribution of E. penicillatus, like E. petteri, is notably localized along the eastern flank of the Central High Plateau. Renewed search for the species and preservation of fresh material to investigate its status and relationship with regard to E. majori and E. danieli are warranted.

Eliurus antsingy Group E. antsingy Carleton, Goodman, and Rakotondravony, 2001

Eliurus antsingy is the most recent of the morphological themes within Eliurus to be formally recognized. A large species with a dark tail tuft and extremely bulbous plantar pads, E. antsingy possesses some characteristics that intimate relationship with the E. majori group: namely, small and rounded pinnae; a short lower incisor that terminates in an indistinct incisive capsule; and long and wide incisive foramina. However, unlike species of the E. majori group, its molars are decidedly less robust in relation to cranial size fig. 4, and the upper and lower third molars are smaller than the second; according to these features, E. antsingy resembles other species of Eliurus. Until evaluated with other data sources, we isolate E. antsingy in its own group.

The distribution of E. antsingy is highly disjunct, so far known from three widely separate regions fig. 7. As conveyed by the specific epithet, the first examples identified, from the PN de Bemaraha and PN de Namoroka, appeared to be restricted to limestone tsingy physiography and associated forest habitat. Recent collecting has extended the range northward, including other sites in tsingy in the RS d'Ankarana but also near massive metamorphic rock outcrops at the edge of transitional dry-humid forest in the Forêt d'Ambilondambo, near Daraina see appendix 1 for detailed localities. Collecting sites with known elevation are all relatively low, 50–550 m. The animals from the PN de Namoroka, as noted by Carleton et al. 2001, are slightly smaller and possess all-white venters compared with those from the PN de Bemaraha. In ventral pelage color, those from Ankarana and Daraina recall the series from PN de Namoroka but are again smaller by an increment comparable to that between the Bemaraha and Namoroka series. The overall amount of craniodental differentiation among the three regions approximates that derived among samples of the broadly distributed E. majori fig. 6, as the latter is currently understood, and they marginally overlap in morphometric space not illustrated. We continue to recognize these regional samples as the one species E. antsingy, but stress the need for collections from intermediate localities, wherever potential habitat exists, and for the application of genetic data to assess levels of divergence and phylogeographic structure.

Figure 7

Geographic occurrence of the new species E. danieli, known only from PN de l'Isalo, and its eastern relatives E. majori and E. penicillatus. Also illustrated is the currently documented range of E. antsingy based on new records reported herein.

Eliurus myoxinus Group E. myoxinus Milne Edwards, 1885; E. minor Major, 1896

Affiliation of these two species in a single species group at first seems incongruous given that each was conventionally accepted as one of the only two valid species of Eliurus Ellerman, 1949; Petter 1972, 1975; Honacki et al., 1982; Corbet and Hill, 1991. In light of current research, the polytypic construct of E. myoxinus that so long masked the distinctive morphologies and genetic divergence of forms such as E. majori and E. tanala almost defies comprehension. Unencumbered by identification with these species and others, the morphological similarity of E. myoxinus and E. minor is more readily appreciable: size small E. minor to medium E. myoxinus; skull stockily built with a relatively broad interorbit, short rostrum, and incisive foramina medium in length; dorsal pelage color dominated by brown tones; caudal pilosity dense and tuft long, covering one-half to two-thirds of the tail length. Although no special affinity was appreciated by Carleton 1994, Jansa et al. 1999 presented molecular data that lend confidence in the sister-group relationship of E. myoxinus and E. minor.

Morphological resemblance of the two species is more convergent among their northernmost populations. Carleton 1994 drew attention to the large size of E. minor that inhabit lowlands around Antongil Bay; robust proportions also characterize those obtained from the RS de Manongarivo. Size variation within E. myoxinus, on the other hand, trends from larger in the south Petriky Forest to smaller in the northwest PN de l'Ankarafantsika, as the species range was then known to Carleton et al. 2001. Recent surveys have now documented E. myoxinus even farther to the north, from the RS de Manongarivo, Daraina Forest, and the northwestern sector of the PN de Marojejy in Antsiranana Province Goodman and Soarimalala, 2002; Soarimalala and Goodman, 2003; appendix 1. These northern samples continue the pattern of size decline, representing the smallest specimens of the species we have so far examined, but they generally conform to the cranial proportions and external morphology that typify E. myoxinus populations to the south unpublished data and analyses. Sample sizes and geographic representation of both forms have vastly improved since Carleton's 1994 revision and should enable a critical and much-needed evaluation of the complex. Careful scrutiny should be devoted to northern Madagascar where the ranges of E. myoxinus and E. minor appear to interdigitate in a complex manner, unlike the southern facies of their distributions where biotic transition is dramatically abrupt and their morphological differentiation is just as sharply marked Goodman et al., 1999.

Eliurus petteri Group E. petteri Carleton, 1994; E. grandidieri Carleton and Goodman, 1998

These two species are medium in size, possess soft-textured and short pelage, and have small auditory bullae, delicately built upper and lower incisors, and an extremely short lower incisor without a capsular projection. Their tails are relatively long and the tail tuft very short and weakly developed in both ca. distal one-fourth of tail length, one with a white-tip E. grandidieri and the other dark E. petteri. Some of these conditions are the extreme expression of the character as so far observed within the genus see Carleton and Goodman, 1998: 173. Genetic investigation to date is limited to E. grandidieri, which surprisingly is revealed as most closely allied to Voalavo gymnocaudus according to these data Jansa et al., 1999. These results puzzle us in view of the morphological traits that characterize Voalavo and unambiguously distinguish it from all species of Eliurus Carleton and Goodman, 1998: 189–194. Further, Voalavo possesses a relatively large midventral gland, unknown in any species of Eliurus, that produces a distinctly musky odor. Although the two genus-group taxa are undoubtedly close kin Carleton and Goodman, 1998; Jansa and Weksler, 2004, the possible paraphyly of Eliurus as indicated earlier by Jansa et al. 1999 requires further corroboration and its generic diagnosis and contents amended if appropriate. Inclusion of tissues from E. petteri, along with additional geographic samples of E. grandidieri, may provide a fresh taxonomic scope to do so.

The range of E. grandidieri is more extensive than that of E. petteri, the former known in the Northern Highlands and northern Central High Plateau, within montane and upper montane forest ca. 1200–1900 m, and the latter restricted to a small area in east-central Madagascar, at intermediate elevations ca. 400–1100 m within Toamasina Province Carleton, 2003. Recent recoveries of E. petteri—from Andriantantely, Saharay, and Sandranantitra appendix 1—originated from the same localized area and reaffirm the species' distinctive morphology.

Eliurus tanala Group E. tanala Major, 1896; E. webbi Ellerman, 1949; E. ellermani Carleton, 1994

These are large species, E. ellermani and some northern samples of E. tanala being fully as robust as E. antsingy, but in cranial and dental proportions, the three differ strikingly from that species and from those of the E. majori group. The molars are shorter and narrower relative to cranial size, with the third molars smaller than the second; upper and lower incisors are stout, and the lower is long and terminates as a moderately E. webbi to strongly E. tanala, E. ellermani pronounced capsular process on the upper mandibular ramus. The lunate incisive foramina are medium in length for the genus, supernumerary palatal foramina are commonplace, often coalesced as elongate vacuities, and the subsquamosal fenestra is small most E. webbi or absent E. tanala, E. ellermani. The pinnae are pliant and elongate, averaging the longest in the genus, and the caudal tuft is well developed but confined to the terminal 30–40% of tail length. Carleton 1994 speculated that E. webbi is sister species to the clade E. tanala–E. ellermani, an expectation not sustained by the few cladistic characters he mustered. Jansa et al. 1999, using cytochrome b sequences, demonstrated phylogenetic pairing of E. webbi and E. tanala among the eight species studied.

The E. tanala group presents many taxonomic problems that require attention. For one, the status of E. ellermani as a species distinct from E. tanala merits reexamination. Application of molecular techniques to fresh material from the amended type locality of E. ellermani Carleton and Goodman, 1998, along with examples of the large, creamy white–bellied E. tanala from the Northern Highlands, should provide the pivotal evidence.

As in the E. majori group, western outliers of the E. tanala assemblage have been recently documented Carleton et al., 2001; appendix 1. Compared with eastern E. tanala proper, those from the PN de Bemaraha have a distinctive color pattern—notably the pale brown dorsum, entirely creamy-white underparts, and the alternately banded tail tuft—but in cranial shape and their robust size, they suggest E. ellermani and northern E. tanala. That from the RS de Maningoza is similar in size and body color but possesses a monocolored tail tuft. Such chromatic contrasts, coupled with the wide geographic hiatus between the western tanala-like populations and the eastern range of E. tanala proper, cast doubt on their conspecific stature. The taxonomic status and distributional extent of the western morphs invite critical review.

Morphological variability within both E. tanala and E. webbi increases in the northern segments of their ranges and complicates their identification Carleton and Goodman, 1998, 2000. Variability is particularly evident in the invasiveness of all creamy-white hairs on the ventrum, the constancy of tail tuft coloration, closure of the subsquamosal fenestra, and overall size. Across the central and southern parts of their distributions, such features appear more stable in their population expression and meaningfully contribute to specific discrimination Carleton, 1994. The mixture of traits exhibited by certain northern samples, such as those from the Daraina region appendix 1 and perhaps the RS de Manongarivo Goodman and Soarimalala, 2002—listed as E. webbi, is so atypical that specific allocation is uncertain. These specimens may constitute a different species altogether or represent exceptional morphological convergence of E. webbi toward the E. tanala morphotype. Teasing apart these hypotheses will require denser transects within the troublesome geographic region in question, broad distributional sampling of typical E. tanala and E. webbi against which to gauge the differentiation of the northern samples, and application of multiple information sources.

In summary, we can be confident that Eliurus comprises many more than the two species so long recognized as valid by Ellerman 1949 and others. Nevertheless, the genus obviously requires further field, museum, and laboratory investigation to improve its alpha-level systematics. Some of the research challenges outlined above will undoubtedly reveal new species, but of greater systematic interest, the next phase of Eliurus investigation should aim to refine relationships among species and to explore patterns of their biogeographic origination.