OUP user menu

The Subfossil Occurrence and Paleoecological Significance of Small Mammals at Ankilitelo Cave, Southwestern Madagascar

Kathleen M. Muldoon, Donald D. de Blieux, Elwyn L. Simons, Prithijit S. Chatrath
DOI: http://dx.doi.org/10.1644/08-MAMM-A-242.1 1111-1131 First published online: 15 October 2009

Abstract

Small mammals are rarely reported from subfossil sites in Madagascar despite their importance for paleoenvironmental reconstruction, especially as it relates to recent ecological changes on the island. We describe the uniquely rich subfossil small mammal fauna from Ankilitelo Cave, southwestern Madagascar. The Ankilitelo fauna is dated to the late Holocene (∼500 years ago), documenting the youngest appearances of the extinct giant lemur taxa Palaeopropithecus, Megaladapis, and Archaeolemur, in association with abundant remains of small vertebrates, including bats, tenrecs, carnivorans, rodents, and primates. The Ankilitelo fauna is composed of 34 mammalian species, making it one of the most diverse Holocene assemblages in Madagascar. The fauna comprises the 1st report of the short-tailed shrew tenrec (Microgale brevicaudata) and the ring-tailed mongoose (Galidia elegans) in southwestern Madagascar. Further, Ankilitelo documents the presence of southwestern species that are rare or that have greatly restricted ranges today, such as Nasolo's shrew tenrec (M. nasoloi), Grandidier's mongoose (Galidictis grandidieri), the narrow-striped mongoose (Mungotictis decemlineata), and the giant jumping rat (Hypogeomys antimena). A simple cause for the unusual small mammal occurrences at Ankilitelo is not obvious. Synergistic interactions between climate change, recent fragmentation and human-initiated degradation of forested habitats, and community-level processes, such as predation, most likely explain the disjunct distributions of the small mammals documented at Ankilitelo.

Key words
  • biogeography
  • Holocene
  • Madagascar
  • range contraction
  • small mammals
  • subfossil
  • taxonomy

Subfossil mammals have been known from Madagascar for more than 100 years (Godfrey and Jungers 2002). By far the best-studied components of the subfossil fauna are the extinct giant lemurs (e.g., Jungers et al. 2002). In contrast, the associated remains of small mammals have received less attention. Although nonprimate mammalian specimens have been collected from paleontological deposits in Madagascar, they are typically not reported in the literature. Given that small-bodied taxa are the only direct means of comparing past and present mammal communities in Madagascar, subfossil small mammal remains provide an opportunity to decipher recent ecological changes that have taken place on the island. Such changes include shifts in species richness (Godfrey et al. 1999), geographic range contractions (Burney et al. 2008; Godfrey et al. 2004; Goodman and Rakotondravony 1996; Goodman et al. 2006; Ranivo and Goodman 2007a, 2007b), and species extinctions (Burney et al. 2008; Goodman et al. 2004, 2007). Comparative studies are necessary to assess the impact of recent ecological changes on surviving mammal communities in Madagascar. However, research has been hindered by a lack of appropriate comparative samples for the identification of small mammal subfossils.

Recently, skeletal remains representing a broad taxonomic spectrum of Malagasy small mammals have been collected as part of extant faunal surveys across the island (e.g., Goodman and Jenkins 2000; Goodman and Soarimalala 2004,2005). In light of these new comparative collections, the osteological basis for species identifications of small mammals can now be discussed. Whereas detailed differential diagnoses based on skeletal characters are not typically reported by zoologists, fragmentary skeletal remains are all that is available from paleontological sites. In particular, the identification of craniodental characters useful for species recognition is necessary to advance studies of subfossil small mammals in Madagascar.

Since 1994, paleontological work at Ankilitelo Cave, southwestern Madagascar, has unearthed an abundance of subfossil remains of small mammals in association with specimens of giant lemurs. In this paper, we present a formal description of the small mammal fauna from Ankilitelo, provide the morphological basis for species identification, and discuss the paleoecological implications of this assemblage.

Ankilitelo is located on the southernmost portion of the karst landscape of the Mikoboka Plateau (Fig. 1). Unlike the better-known cave systems in Madagascar, Ankilitelo consists of a narrow vertical shaft approximately 10 m in diameter and 145 m in depth, accessible only by a single entrance (Fig. 2). At 145 m, the shaft opens up into a large spacious cavern. A large talus slope leads from the shaft to a deeper part of the cave approximately 30 m below. Positioned directly under the shaft on the floor of the cavern is a pile of mud and debris, containing a very high density of bone in the surface layer to a depth of approximately 1 m. Subfossils are not found below this layer. Most of the subfossils are recovered from the area located directly below the shaft, but also are scattered down an immense slope below this point. The lower reaches of the talus slope do not contain subfossils. Subfossils were collected, recorded, and curated according to a series of areal zones on the cone surface so that associations are preserved.

Fig. 1

Map showing the general position of Ankilitelo Cave relative to modern forests mentioned in text. Ankilitelo is located 55 km northeast of Toliara in southwestern Madagascar (22°54.819′S, 43°52.610′E, Universal Transverse Mercator coordinates 38 K 0384818, datum WGS 84).

Fig. 2

Diagram of Ankilitelo Cave in profile and plan view. Cave length equals 635 m. Cave depth equals ′230 m. Surveyed June 1997. Prepared by Charles Savvas and Nancy Pistole.

The stratigraphy of the surface bone layer reflects dual processes of accumulation of material coming down the shaft at the apex of the pile, and sloughing of material down onto the talus slope. The shaft and cavern of Ankilitelo were formed underground as the limestone of the Mikoboka Plateau dissolved through the action of water trickling in from ground level. As the cave grew, the roof of the shaft became instable and collapsed. Subfossils could not have accumulated before the shaft broke through at the surface. Ankilitelo documents the youngest record of Malagasy subfossil lemurs (Burney et al. 2004). Specimens from Ankilitelo provide a date for Megaladapis of 630 ± 50 years ago, and for Palaeopropithecus of 510 ± 80 years ago (Simons et al. 1995), indicating that the extinct fauna of Madagascar overlapped temporally with humans for at least 1,800 years in the southwestern region (Burney et al. 2004). Radiocarbon dating of the small mammal samples from Ankilitelo confirms that the giant lemurs and small mammal fauna were deposited contemporaneously. A partial femur of Cryptoprocta collected on the slope of the cone was dated to 560 ± 40 years ago (Beta-201844). Another submitted specimen was not large enough to produce an accurate radiocarbon date. It is unlikely that the fauna represents a time-averaged assemblage. The narrow window of radiocarbon dates combined with the geology and stratigraphy of the cave suggests that there is no significant temporal signal to the small mammal remains. Instead, the subfossil assemblage appears to represent a relatively limited snapshot of time.

The rich accumulation of subfossils collected from the area directly below the opening of the cave indicates that Ankilitelo served as a natural trap for larger vertebrates that chanced to fall in. Natural traps are caves with vertical entrances that allow animals to enter, but exit is prohibited via the same entrance usually because of steep or inverted cave walls (Wolverton 2001, 2006). Animals often die from falling into such caves, or of starvation after surviving such a fall (Oliver 1989). Because of the depth of Ankilitelo (230 m), it is unlikely that any animals could have survived falling down the shaft. It is possible that the megafauna were attracted to moisture, food, or shelter, or a combination of these factors, in the forest at the cave opening, and thus more susceptible to accidental death. Similar scenarios have been proposed for the accumulation of cave bears in deep shaft caves in Missouri and Wyoming, and Atapuerca, Spain, during the Pleistocene (Andrews and Fernandez Jalvo 1997; Martin and Gilbert 1978; Wang and Martin 1993; Wolverton, 2001, 2006). The condition and size range of the small mammal remains at Ankilitelo strongly imply that they were brought into the cave by the action of predators. In addition to the abundant remains of giant lemurs, there are thousands of small mammal specimens, indicating that the sample was not taphonomically biased by size sorting.

Materials and Methods

More than 5,000 vertebrate specimens, representing mammals, birds, and herpetofauna, have been collected from Ankilitelo Cave under the direction of ELS. Skeletal remains in the mammalian assemblage demonstrate a high quality of preservation and include largely complete crania, jaws, isolated teeth, and postcranial elements. All subfossil material recovered from Ankilitelo has been collected as part of a collaborative effort between the Duke University Lemur Center, Division of Fossil Primates (Durham, North Carolina) and the Département de Paléontologie et Anthropologie Biologique, Université d'Antananarivo (Antananarivo, Madagascar).

Thirty-four species of mammals have been identified as subfossils from Ankilitelo Cave. The Ankilitelo fauna is best known as an extraordinarily rich repository for 5 species of extinct giant lemurs: Palaeopropithecus ingens, Megaladapis madagascariensis, Archaeolemur majori, Daubentonia robusta, and Pachylemur (Godfrey et al. 1999; Hamrick et al. 2000; Jungers et al. 1997, 2005; Shapiro et al. 2005; Simons et al. 2004; Wunderlich et al. 1996). In contrast, few dental remains of the extinct pygmy hippo (Hippopotamus) have been recovered from the cave. Among still-extant species, the giant jumping rat (order Rodentia: Hypogeomys antimena) is present in the Ankilitelo fauna based on an isolated right femur (UA 10440), but is not represented craniodentally. Introduced mammals are represented by black rats (Rattus rattus) and house mice (Mus musculus).

The endemic subfossil small mammals are described below using comparisons with extant material housed in the Division of Mammals, Smithsonian Institution (Washington, D.C.); Department of Mammalogy, American Museum of Natural History (New York); Mammal Department, Museum of Comparative Zoology, Harvard University (Boston); and the Division of Mammals, Field Museum of Natural History (Chicago). All measurements were taken using Mitutoyo Digimatic Calipers (Paleo-Tech Concepts, Crystal Lake, Illinois) calibrated to an accuracy of 0.01 mm. Craniodental variables were recorded in millimeters for each specimen examined following anatomical landmarks defined and illustrated in Giannini et al. (2006) and Czaplewski et al. (2008) for bats; Jenkins and Goodman (1999), MacPhee (1987), and Swindler (2002) for tenrecs; Carleton (1994) and Musser and Carleton (1993) for rodents; Coetzee (1977) and Wozencraft (2005) for euplerids; and Rasoloarison et al. (2000) and Swindler (2002) for primates. We generally follow the higher-level taxonomy of Wilson and Reeder (2005). The primates of Madagascar are currently undergoing extensive taxonomic revision based on molecular analyses (e.g., Andriaholinirina et al. 2006; Craul et al. 2007, 2008; Louis et al. 2006a, 2006b). Although significant species-level diversity in most lemur genera is apparent, diagnoses based exclusively on genetic distance must be treated with caution until supporting differential data are provided (Tattersall 2007). We therefore follow the higher-level taxonomy of Mittermeier et al. (2006) for Primates, with the exception that our treatment of Cheirogaleus follows Groeneveld (2008). The number of identifiable specimens (NISP) and minimum number of individuals (MNI) are presented for each species based on dental remains in order to describe relative taxonomic abundance in the faunal assemblage. Institutional abbreviations are DPC—Duke University Primate Center (now Duke University Lemur Center), Durham, North Carolina, and UA—Université d'Antananarivo, Antananarivo, Madagascar. Anatomical abbreviations are L.—left, R.—right, dent.—dentary, and max.—maxillary; lowercase refers to lower teeth and uppercase refers to upper teeth.

Systematic Paleontology

Because most of the taxa present at Ankilitelo are known extant species, detailed descriptions are unnecessary. Only diagnostic bony features are discussed, providing a skeletal key for the identification of small mammals from subfossil deposits.

Order Chiroptera Blumenbach, 1779

Suborder Microchiroptera Dobson, 1875

Superfamily Vespertilionoidea (Gray, 1821)

Family Miniopteridae Miller-Butterworth, 2007

Miniopterus Bonaparte, 1837

Miniopterus gleni Peterson et al., 1995

Glen's Long-fingered Bat Fig. 3A

Fig. 3

Chiroptera and Tenrecinae from Ankilitelo. A) Miniopterus gleni, skull DPC 18761b in lateral and inferior view; B) Otomops madagascariensis, skull DPC 18788a in lateral and inferior view; C) Mormopterus jugularis, left dentary DPC 18770d in lateral view; D) Setifer setosus, left dentary DPC 17293a in lateral view; E) Echinops telfairi, right dentary DPC 13764c in lateral view. Scale bar equals 5 mm.

Referred specimen.—DPC 18761b (partial cranium, L. max. M2–M3). NISP = 1; MM = 1.

Remarks.—Miniopterus gleni is represented in the Ankilitelo fauna by a partial cranium. Miniopterus can be distinguished from other Malagasy genera by the distinctive shape of its skull. When viewed laterally, the anterior braincase is domed and rises steeply from the low, flat rostrum, giving the face a distinctive “dished” shape.

Although the rostrum is broken in DPC 18761b, the orientation of the palate relative to the clivus forms a large inferior angle, which is diagnostic of this dorsally flexed morphology. The squared and gracile shape of the zygomatic arches, relative flatness of the auditory bullae, and distally projecting palatine spine also are characteristic of the genus.

Family Molossidae Gervais, 1855

Subfamily Molossinae Gervais, 1855

Otomops Thomas, 1913

Otomops madagascariensis Dorst, 1953

Malagasy Giant Mastiff Bat Fig. 3B

Referred specimens.—DPC 18788a (partial cranium, R. max. II, C, P4–M3, L. max. C, P4–M3); UA Unno (partial cranium, R. max. P4–M3, L. max. Il, C–M3; associated mandible, R. dent. i2–m3, L. dent. p3–m3); 10187 (partial cranium, R. max. P4–M3, L. max. P4–M3). NISP = 3; MNI = 3.

Remarks.—The skull of Otomops can be distinguished from that of other Malagasy genera by several features, including its comparatively large size. The rostrum is long, deep, and tapers anteriorly. The low braincase rises at a broad angle from the rostrum. The palate is U-shaped, broad, and squared posteriorly, in contrast to the parabolic contour in Miniopterus. The posterior palatine spine does not project distally. The interorbit is parallel-sided and narrow. The zygomatic arches are distinguished by a prominent superior projection at the midpoint of the arch.

Mormopterus Peters, 1865

Mormopterus jugularis (Peters, 1865)

Peter's Wrinkle-lipped Bat Fig. 3C.

Referred specimen.—DPC 18770d (L. dent. ml–m3). NISP = 1; MNI = 1.

Remarks.—Mormopterus jugularis is represented in the Ankilitelo fauna by a single dentary, which preserves several features diagnostic at the generic level. The mandible is distinctively robust for its small size. The angle of the mandible is broad and shallow. The angular process is displaced superiorly on the ramus, but is large and projects distolaterally. The angular process is deep superoinferiorly but narrow mesiodistally. The angular notch is broad and U-shaped. The mandibular condyle and coronoid process are in approximately the same plane, and separated by a broad and shallow sigmoid notch.

Order Afrosoricida Stanhope et al, 1998 Suborder Tenrecomorpha Butler, 1972

Family Tenrecidae Gray, 1821 Subfamily Tenrecinae Gray, 1821

Tenrec de Lacépède, 1799

Tenrec ecaudatus (Schreber, 1778)

Common Tenrec

Referred specimen.—DPC 13760k (juvenile, R. dent. dp3, p4, m2 in crypt). NISP = 1; MNI = 1.

Remarks.—Tenrec ecaudatus is rare in the Ankilitelo assemblage. A single incomplete dentary of a juvenile is referable to this species based on its large size and distinctively pronounced diastemata. The molariform p4, with distinctively squared pre- and postcingulids, also is characteristic of the monospecific genus.

Genus Setifer Froriep, 1806

Setifer setosus (Schreber, 1778)

Greater Hedgehog Tenrec Fig. 3D

Referred specimens.—DPC 13699a (juvenile, R. dent. dp3 or dp4); 13699b (juvenile, R. dent, edentulous); 13706a (L. dent. p2–ml, m3); 13725a (juvenile, L. dent. p4); 13725b (L. dent, c, p2); 13760L (juvenile, R. dent. p2, m3 in crypt); 17160a (juvenile, L. dent, edentulous); 17200n (R. dent. p2); 17200p (L. dent, edentulous); 17200q (juvenile, L. dent m3 in crypt); 17200r (juvenile, L. dent, edentulous); 17200t (R. p3); 17200w (R. max. P3–P4); 17209d (R. dent. p2–m2); 17209e (L. dent. p2–p3, ml); 17253h (L. dent, edentulous); 17253i (L. i2); 17320d (L. dent, edentulous); 17253J (L. p2); 17253k (L. p3); 172531 (L. ml); 17253m (L. c); 17253n (L. p3); 17259g (L. dent. c–m2); 17259h (L. m2 or m3); 17278f (R. dent, edentulous); 17278g (L. max. C, P3); 17293a (L. dent. i2–m3); 17293b (R. dent. i2–p3); 17320e (L. dent, edentulous); 17335b (R. dent. i2–p2); 17335c (L. dent p3); 18067a (R. dent, edentulous); 18067b (L. max. P3–P4); 18125a (R. max. C, P3– P4); 18125b (L. p4); 18125f (R. dent. p2–m3); 18125g (L. dent, edentulous); 18125h (juvenile, R. dent. i2, c, m3 in crypt); 18125i (L. dent, edentulous); 18125k (juvenile, R. dent. m3 in crypt); 18125x (R. dent, edentulous); 18785b (L. dent. p2–4); 18785c (R. dent. p2–p3); 18785d (R. dent, c–p3); 18785d (L. Il); 18785f (L. P3); 18785h (L. 12); 18785i (L. 12); 18785k (R. dent. p3–p4); 18787a O'uvenile, L. dent. p4–ml, m2 in crypt); 18877a (L. dent, edentulous); 18877b (L. p2); 18877c (R. dent. m3); 18953 (L. dent. c–m2); 18974a (L. dent. p3–p4); 18974b (L. dent, edentulous); 18974d (L. p4); 18899b (R. dent, edentulous); 18899c (L. dent, edentulous); 18911 (R. dent. i2, p2–m2); UA 9403 (L. dent, edentulous); 9886 (partial cranium, R. max. C, P2–P4, M3, L. max. C, P2–P4); 10192 (partial cranium, edentulous); 10195 (L. dent, edentulous); UnnoA (juvenile, R. dent. dp3–dp4, m2 in crypt); UnnoB (R. dent, edentulous); UnnoC (L. dent, edentulous). NISP = 67; MM = 22.

Remarks.—Setifer setosus is the most common tenrec found at Ankilitelo. Setifer can be distinguished from other tenrecines, Tenrec ecaudatus and Echinops telfairi, by its dental formula (i 2/2, c 1/1, p 3/3, m 3/3, total 36, in contrast to i 2/3, c 1/1, p 3/3, m 3/3, total 38 in Tenrec, and i 2/2, c 1/1, p 3/3, m 2/2, total 32 in Echinops). Setifer further differs from Tenrec by the lack of enlarged diastemata, and its much smaller body size. Setifer is larger than Echinops in most craniodental dimensions (Table 1). Other traits that separate Setifer from Echinops include its broader rostrum, and more laterally flared and prominent superior articular facets of the zygoma. The upper canine is taller in crown height than the central incisor, and the upper molars are not as mesiodistally compressed and demonstrate less-developed stylar regions. Juvenile specimens of Setifer can be particularly difficult to distinguish from adult Echinops. The most useful character for discerning juvenile from adult specimens is the morphology of dP4/dp4. In juveniles, dP4 is highly molariform, whereas the adult tooth is much less molariform and has a completely different anterobuccal portion. The deciduous p4 is smaller and overall more gracile than the adult p4. Features of the mandible, such as the roundness of the coronoid process and mandibular condyle, also are useful in determining the relative age of the individual.

View this table:
Table 1

Selected tooth measurements (in mm) for Afrosoricida from Ankilitelo. Measurements are in mm, given as means ± standard deviations.

p4Lp4Wm1LmlWm2Lm2Wm3Lm3WP4LP4WM1LM1WM2LM2WM3LM3W
Setifer setosus1.0 ± 0,1 n = 101.0 ± 0.1 n = 102.5 ± 0.3 n = 83.0 ± 0.2 n = 82.2 ± 0.3 n = 72.6 ± 0.1 n = 62.0 ± 0.2 n = 42.0 ± 0.1 n = 42.4 ± 0.2 n = 44 3 + 0.2 n = 4
Echinops telfairi0.7 n = 20.9–1.0 n = 2
Geogale aurita1.1 n = 10.9 n = 11.1 n = 11.1 n = 11.0 n = 11.1 n = 11.2 n = 10.8 n = 11.3 ± 0.1 n = 31.3 ± 0.06 n = 31.2 ± 0.06 n = 31.5 ± 0.1 n = 31.0 ± 0.1 n.= 31.8 ± 0.1 n = 30.4 ± 0.06 n = 32.0 ± 0.2 n = 3
Microgale brevicaudata0.8 n = 10.7 ai = 10.9 n =10.8 n = 10.9 n = 10.9 n = 11.3 n = 10.8 n = 11.4 ± 0.0 n = 31.3 ± 0.2 n = 31.2 ± 0.1 n = 31.5 ± 0.1 n = 31.1 ± 0.1 n = 31.7 ± 0.1 n = 30.6–0.7 n = 21.7–1.8 n = 2
Microgale nasoloi Microgale cf. M. majori1.0 ± 0.1 n = 70.7 ± 0.1 n = 71.0 ± 0.1 n = 80.9 ± 0.1 n = 81.2 ± 0.1 n=51.0 ± 0.1 n = 51.5–1.7 n = 20.8–0.9 n = 21.5–1.6 n=21.3–1.4 n = 21.3–1.4 n = 2 0.9 n=11.7 n = 2 1.2 n = 11.3 n = 2 0.9 n = 11.8 n = 2 1.3 n = 10.9–1.0 n = 22.1–2.2 n=2

Echinops Martin, 1838

Echinops telfairi Martin, 1838

Lesser Hedgehog Tenrec Fig. 3

Referred specimens.—DPC 13706c (juvenile, R. dent, edentulous, m2 in crypt); 13706d (L. dent, edentulous); 13761a (L. dent, edentulous); 13764c (R. dent. i2–p4); 18090a (L. dent, edentulous); 18125c (L. p3); 18125L (juvenile, L. dent, edentulous, m2 in crypt); 18125p (juvenile, L. dent, edentulous, m2 in crypt); 18125q (L. dent, edentulous); 18125r (L. dent, edentulous); 18125s (L. dent, edentulous); 18125t (L. dent, edentulous); 18785e (R. dent. i2–p4); 18785J (R. m2); 18967a (juvenile, L. dent. i2–p3, m2 in crypt); 18967b (juvenile, R. dent. il–p3); UA 10193 (R. dent, edentulous); 10194 (L. dent, edentulous); UnnoD (R. dent. i2). NISP = 19; MNI =11.

Remarks.—These specimens are referable to Echinops telfairi based on the combination of size (Table 1), and several dental characters, such as the distinct incisiform lower canine and mesiodistally compressed lower molars. Echinops has the most reduced dentition within the Tenrecidae, with only 2 molars in each jaw. More-detailed craniodental differences between Echinops and Setifer were described above.

Subfamily Geogalinae Trouessart, 1881

Geogale Milne-Edwards and A. Grandidier, 1872

Geogale aurita Milne-Edwards and A. Grandidier, 1872

Large-eared Tenrec Fig. 4A

Fig. 4

Geogalinae and Oryzorictinae from Ankilitelo. A) Geogale aurita, skull DPC 22049a in occlusal view; B) Microgale brevicaudata, skull DPC 18770a in occlusal view; C) M. nasoloi, skull 18125e in occlusal view; D) M. cf. M. majori, right maxilla DPC 22054 in occlusal view. Scale bar equals 1 mm.

Referred specimens.—DPC 18761c (partial cranium, R. P3–M3, L. P4–M2); 22049a (partial cranium, R. M1–M3, L. P4–M3); 22049b (partial cranium, R. P4–M3, L. P4–M2); 22051 (R. dent. p4–m3); 22096b (L. dent, edentulous); 22096c (L. dent, edentulous). NISP = 6; MNI = 3.

Remarks.—Geogale aurita is identified at the generic level by its unique dental formula (i 2/2, c 1/1, p 3/2, m 3/3, total 34) and small size. Its cranium is elongated, with a flat and straight dorsal profile. The zygomatic plate is broad with prominent maxillary projections. The interorbit is narrow and elongated. The rostrum is broad, especially when compared with specimens of Microgale of similar or smaller size. The most Geogale-like species of Microgale is M. brevicaudata, which also is found at Ankilitelo. Geogale can be distinguished from M. brevicaudata by several craniodental features. The dental formula of Microgale is i 3/3, c 1/1, p 3/3, m 3/3, total 40, whereas Geogale has a reduced dentition. The skull of Geogale is flatter in dorsal profile with a smaller braincase, and the lambdoid crest is more distinct. The posterior palatine spine is compressed and projects more prominently inferoposteriorly. The anterior dentition of both the maxilla and mandible is more robust in Geogale, and P2 and P3 are reduced relative to P4. Additionally, m3 differs from the condition in M. brevicaudata in that it lacks an expanded talonid.

Subfamily Oryzorictinae Dobson, 1882

Microgale Thomas, 1882

Microgale brevicaudata G. Grandidier, 1899

Short-tailed Shrew Tenrec Fig. 4B

Referred specimens.—DPC 13780c (partial cranium, R. max. 12, C, P3–M3; L. max. C, P3–M3); 17200u (R. dent. p3–m3); 18770a (partial cranium, R. max. C–M3; L. max. C–M3); 22077 (partial cranium, R. max. P3–M3, alveoli for I1–P2; L. max. P3–M1). NISP = 3; MNI = 3.

Remarks.—Species of Microgale share several primitive dental characters that distinguish them from other genera of Malagasy tenrecs. These include a dental formula of i 3/3, c 1/ 1, p 3/3, m 3/3, total 40, upper incisors and canines with distostyles, zalambdodont upper molars, and lower molars dominated by a large tapering protoconid. Microgale brevicaudata can be distinguished from its congeners by a robust skull and short rostrum (MacPhee 1987). In profile, the nasal region has a swollen, slightly convex outline and appears to ascend more sharply toward the top of the skull than in other species. M3 is reduced relative to the anterior molars, and has a small, squared talon that is not elongated buccolingually, in contrast to the condition in M. nasoloi. The talonid of m3 is well developed, with a distinct hypoconid, hypoconulid, and pronounced entoconid ridge. M. brevicaudata can be further differentiated from the other species of Microgale that are found at Ankilitelo based on size (Table 1). A new species, MacPhee's shrew tenrec (M. macpheei) was recently described from subfossil deposits at Andrahomana Cave (Goodman et al. 2007). M. brevicaudata is notably smaller than published measurements for M. macpheei.

Micro gale nasoloi Jenkins and Goodman, 1999

Nasolo's Shrew Tenrec Fig. 4C

Referred specimens.—DPC 13760c (L. dent. p4–m2); 13780d (R. dent. p4–m3); 17200s (L. dent. p3–2); 17200v (R. dent., ml, m3); 18125d (R. dent. p3–p4); 18125e (partial cranium, R. max. C–M3; L. max. P2–M2); 18770b (R. dent. p4–m3); 18786g (R. dent. p4–m3); 18786h (L. dent. p4–m2); 22047 (R. dent. p4, m2–m3); 22070 (partial cranium, R. max. P3, M1–M3; L. max. C–M3); 22096a (R. dent. p4–ml). NISP = 2; MNI = 2.

Remarks.—Microgale nasoloi was described based on 2 distinctive specimens collected in southwestern Madagascar (Jenkins and Goodman 1999). M. nasoloi can be easily distinguished from other species of Microgale present at Ankilitelo by its larger size, upper molars with distinctive buccolingually elongated talons, especially prominent on M3, and m3 with a developed talonid, although the cusps are less defined.

Microgale cf. M. majori (Thomas, 1918)

Major's Shrew Tenrec Fig. 4D

Referred specimen.—DPC 22054 (R. max. M1–M2). NISP = 1; MNI = 1.

Remarks.—DPC 22054 is a right maxillary fragment preserving only M1 and M2. The molars are of the zalambdodont form typical of Microgale, which is not diagnostic among species. The mesiostyle, anterior ectostyle, posterior ecostyle, and distostyle are present and well defined. The centrobuccal cleft, which separates the anterior from the posterior ectostyle, is more prominent in M2 than M1.

There are 5 species of Microgale that have been reported to occur in southwestern Madagascar: M. brevicaudata, M. nasoloi, M. jenkinsae (Jenkins’ shrew tenrec), M. longicaudata (lesser long-tailed shrew tenrec), and M. majori (Major's long-tailed shrew tenrec), although a 6th species has recently been described, but is not included here (Olson et al. 2009). DPC 22054 is clearly well below the possible size range for either M. brevicaudata or M. nasoloi, the other shrew tenrecs present in the Ankilitelo sample (Fig. 5). M. jenkinsae was recently described from the Forêt des Mikea, a forest in close proximity to Ankilitelo Cave, on the basis of 2 subadult specimens (Goodman and Soarimalala 2004). M. jenkinsae is a small member of the genus that can be differentiated from other shrew tenrecs by size and the presence of a single-rooted p2. In the absence of associated mandibular material, comparisons of M1 area reveal that DPC 22054 falls below the documented size range for M. jenkinsae. It is important to note that although it is possible that DPC 22054 represents a juvenile individual, it is still much smaller than subadult M. jenkinsae. When comparing M1 area, DPC 22054 falls within the approximate size range of M. pusilla, but is significantly smaller than M. longicaudata (t = 3.696, d.f. = 4, P < 0.05) and M. majori (t = 4.980, d.f. = 10, P < 0.001; Fig. 5). Microgale pusilla is not known to occur in the drier portions of Madagascar today, although it was reported from disintegrated owl pellets of unknown age from the subfossil sites of Lelia and Anjohimpaty in southwestern Madagascar (MacPhee 1986, 1987). These specimens are more likely referable to M. jenkinsae (Goodman and Soarimalala 2004).

Fig. 5

Scatterplot showing discrimination of maxillary specimens of Microgale from Ankilitelo based on size. M1 area is used to illustrate this separation based on the large number of preserved specimens in the subfossil sample. Note that the M1 area of DPC 22054 falls within the range of extant M. pusilla, but is significantly smaller than extant M. longicaudata, M. majori, and M. jenkinsae (known only from subadult individuals). Ankilitelo Micro gale are represented by solid symbols, extant Microgale by open symbols.

A recent molecular and morphological study of cryptic species boundaries in long-tailed shrew tenrecs suggests that multiple, distinct species clusters occur within M. long-icaudata (Olson et al. 2004). If these clusters represent phylogenetic groupings, the holotype of M. longicaudata corresponds to a large-bodied highland clade of long-tailed shrew tenrec, and a resurrected M. majori belongs to a small-bodied, geographically widespread clade. M. majori has been reported from Analavelona (Olson et al. 2004). M. majori is thus the only small-bodied, long-tailed shrew tenrec now believed to exist in western Madagascar. One of the morphological characters used to distinguish between M. longicaudata and M. majori is the width of M3, which unfortunately is not preserved in DPC 22054. We therefore provisionally refer DPC 22054 to M. majori based on geographic location.

Order Carnivora Bowdich, 1821

Suborder Feliformia Kretzoi, 1945

Family Eupleridae Chenu, 1850

Subfamily Euplerinae Chenu, 1850

Cryptoprocta Bennett, 1833

Cryptoprocta ferox Bennett, 1833

Fosa

Referred specimens.—DPC 17152 (partial cranium, edentulous); 17215g (R. dent. p2–ml); 17260 (L. dent. p3–ml). NISP = 3;MNI = 1.

Remarks.—Cryptoprocta is easily distinguished from other Malagasy carnivorans at the generic level by its considerably larger size. In addition, its craniodental anatomy reflects adaptations toward extreme carnivory. Such characters include a reduced dental formula (i 3/3, c 1/1, p 3/3, m 1/1, total 32), robust mandible with a long and reinforced symphysis, extremely large canines, and greatly enlarged, shearing carnassial dentition.

Subfossil remains have revealed that a larger species of this genus, C. spelea, has recently become extinct on Madagascar (Goodman et al. 2004; Grandidier 1902; Lamberton 1937, 1939; Petit 1935), although no radiocarbon date is available for subfossil specimens of Cryptoprocta. In general, C. spelea is significantly more robust, and has more massive teeth than the extant C. ferox. The presence of 2 size classes of Cryptoprocta at some subfossil localities suggests that both species of hypercarnivore may have lived in temporal sympatry in some parts of the island (Goodman et al. 2004).

Radiocarbon dates supporting this fact are needed. The Ankilitelo specimens fall slightly above or within the size range of modern C. ferox, and within the range of subfossil C. ferox for measurements of lower tooth lengths (Table 2).

View this table:
Table 2

Dental measurements of Ciyptoprocta ferox from Ankilitelo and extant C. ferox. Measurements for modern specimens are presented as mean ± standard deviation, minimum and maximum measurements.

Modern Cryptoproctaa (n = 18)Ankilitelo Ciyptoproctaa,b (n = 2)Subfossil Ciyptoproctac (n = 3)Subfossil Ciyptoprocta speleac,d
p3L7.8 ± 0.4 6.5–8.08.0–8.18.5 ± 0.8 7.7–9.29.2
p4L8.9 ± 0.6 7.9–10.010.0–10.110.3 ± 1.1 9.1–11.711.7
mlL11.6 ± 0.9 10.2–13.812.3–12.612.8 ± 0.5 12.4–13.413.4
  • a Comparative data collected during this study.

  • b The Ankilitelo sample consists of 2 mandibular specimens. Therefore, only the range is presented.

  • c Measurements from Goodman et al. (2004).

  • d Only means are presented for C. spelea in Goodman et al. (2004).

Subfamily Galidiinae Gray, 1865

Galidia I. Geoffroy Saint-Hilaire, 1837

Galidia elegans I. Geoffroy Saint-Hilaire, 1837

Ring-tailed Mongoose

Figs. 6A and 6B

Fig. 6

Euplerinae from Ankilitelo. Galidia elegans, A) partial maxilla DPC 17200k in occlusal view; B) left dentary DPC 17314f in lateral and occlusal view. Scale equals 5 mm. C) Galidictis grandidieri, partial cranium DPC 18104a in occlusal and dorsal view. Scale equals 1 cm. Mungotictis decemlineata, D) partial cranium DPC 13765f in occlusal and dorsal view; E) right dentary DPC 18107 in lateral view. Scale bar equals 1 cm.

Referred specimens.—DPC 17199m (R. dent. i2–ml); 17200k (max. L. I1–I2, R. I1–I3, P2–P3); 17253q (R. dent, ml); 17314f (partial mandible, L. c–m2, R. i2); 17333 (partial mandible, L. i2, c–m2, R. c–m2); 18769 (L. dent. i2–m1). NISP = 6; MNI = 4.

Remarks.—Several distinguishing characters separate Galidia from the other carnivorans in the subfossil sample at the genus level. In particular, the talon or buccodistal face of P3 is expanded, and M2/m2 is greatly reduced relative to M1/m1. Additionally, Galidia has a short and robust mandible coupled with relatively slender canines. The postcanine dentition is characterized by extremely trenchant, buccolingually compressed cusps, which are especially prominent on ml. The trigonid of ml is dominated by a large protoconid. The metaconid is greatly reduced, lingually offset, and widely separated from the paraconid. This arrangement creates the lingually open or splayed trigonid that is characteristic of the genus.

Galidictis I. Geoffroy Saint-Hilaire, 1839

Galidictis grandidieri Wozencraft, 1986

Grandidier's Mongoose Fig. 6C

Referred specimens.—DPC 17252a (partial cranium, edentulous); 17252b (partial cranium preserving frontal); 18104a (partial cranium, R. P2–M2, L. P3–P4); 18104b (L. dent. i2, c, p2, p4); 18782 (R. max. 12, P2–P4). NISP = 5; MNI = 4.

Remarks.—Galidictis is characterized by extremely large and sharply recurved canines, particularly in the lower dentition, and a correspondingly robust anterior maxilla and mandibular symphysis. Reliable cranial features for distinguishing Galidictis from other carnivorans at Ankilitelo include the broad rostrum, reduced supraorbital processes, greater postorbital constriction, and flared, rounded zygoma. In particular, the anatomy of the auditory region, with an inflated bullae, and drawn-out, tubular ectotympanic, is unique at the genus level.

Mungotictis Pocock, 1915

Mungotictis decemlineata (A. Grandidier, 1867)

Narrow-striped Mongoose

Figs. 6D and 6E

Referred specimens.—DPC 13765f (partial cranium, R. P2–M3, L. I1–I2, P2–M2); DPC 18107 (R. dent. p2–ml). NISP = 2; MNI = 1.

Remarks.—The skull of Mungotictis decemlineata is smoothly convex in dorsal profile, with a broad interorbit, and less prominent supraorbital processes than are present in Galidictis. Mungotictis retains the 1st upper premolar (for a dental formula of i 3/3, c 1/1, p 4/3–4, m 2/2, total 38–40). The dentition is characterized by bunodont but well-defined cusps, and well-developed stylar regions on the upper molars. M2 is not as greatly reduced relative to M1 as is the condition in Galidia, and the teeth are overall smaller than those of Galidictis. The lower carnassial, ml, has a closed trigonid and distinctive, basinlike, distally expanded talonid. Mungotictis is easily distinguished from the other Malagasy carnivorans at Ankilitelo on the basis of size.

Order Rodentia Bowdich, 1821

Suborder Myomorpha Brandt, 1855

Superfamily Muroidea Illiger, 1811

Family Nesomyidae Major, 1897

Subfamily Nesomyinae Major, 1897

Macrotarsomys Milne-Edwards and G. Grandidier, 1898

Macrotarsomys bastardi Milne-Edwards and G. Grandidier,

1898

Bastard's Big-footed Mouse Fig. 7A

Fig. 7

Nesomyinae from Ankilitelo. A) Macrotarsomys bastardi, right dentary DPC 17200cc in occlusal view; B) Macrotarsomys petteri, left dentary DPC 17253p in occlusal view; C) Eliurus myoxinus, right dentary DPC 18125v in occlusal view; D) Eliurus sp. indet., right dentary DPC 17213d in occlusal view. Scale bar equals 1 mm.

Referred specimens.—DPC 17200cc (R. dent. m1–m3); 17200dd (R. dent. ni2–m3); 17200ee (L. dent, edentulous). NISP = 2; MNI = 2.

Remarks.—Three mandibular specimens can be allocated to Macrotarsomys bastardi on the basis of the characteristically simple, low-crowned and cuspidate dentition, and the retention of the anteroconid on ml. M. bastardi is easily distinguished from other species of Macrotarsomys by its small size. M. bastardi can be differentiated from introduced Mus musculus, also found at Ankilitelo, because Mus is significantly smaller than M. bastardi. Further, in Macrotarsomys, the cusps are connected by low crests in an alternating pattern, such that tooth wear results in a zigzag pattern of dentine exposure, unlike the buccolingually transverse, or lophidate, arrangement in Mus. Further, m3 is not reduced relative to the anterior molars in Macrotarsomys.

Macrotarsomys petted Goodman and Soarimalala, 2005

Petter's Big-footed Mouse Fig. 7B

Referred specimens.—DPC 13760e (L. dent, ml); 17200x (L. dent. m3); 17244c (R. dent. m3); 17253p (L. dent. m2– m3); 17259J (L. dent. m3); 18070d (L. max. M1–M3); 18090d (R. dent. m3); 18125w (L. dent. m3); 18762i (R. dent. m3); 18786J (L. max. M1–M3); 18877f (R. dent. m3); 21996 (L. max. M1); 22021a (R. dent. ml–m3); 22022a (L. max. M1–M3); 22022b (R. max. M1, M3); UA 9772 (L. dent. m3); 9773 (L. dent, edentulous); 10188 (R. dent, edentulous); 10189 (R. dent, edentulous); 10318 (R. dent. m2). NISP = 19; MNI = 7.

Remarks.—Macrotarsomys petteri was named on the basis of a single specimen captured in the Forêt des Mikea (Goodman and Soarimalala 2005). M. petteri is the largest known species of Macrotarsomys. It shares with other members of the genus the characteristic brachydont and cuspidate dentition. Both M1 and ml retain the anterocone(id). The cusps are connected by low crests in an alternating pattern, such that tooth wear results in a zigzag pattern of dentine exposure. This characteristic wear pattern easily separates Macrotarsomys from Rattus, in which the cusps are connected by transverse loph(id)s. M. petteri is further distinguished from Rattus by the absence of a centrally placed, posterocingulid cuspid on the lower molars.

Eliurus Milne-Edwards, 1885

Eliurus myoxinus Milne-Edwards, 1885

Dormouse Tufted-tailed Rat Fig. 7C

Referred specimens.—DPC 13725d (R. ml or m2); 13761b (L. dent. m1–m3); 13780b (L. dent. m2–m3); 17199h (L. dent, m1); 17213e (L. dent. m1–m2); 17290e (L. dent, ml, m3); 18044b (R. dent. ml–m3); 18045a (L. dent. m1–m3); 18064a (L. dent. m3); 18070b (L. dent. m1–m2); 18125u (R. dent. m3); 18125V (R. dent. m1–m3); 18773c (L. dent. m3); 18789f (L. dent, ml, m3); 18773d (R. dent, ml, m3); 18773e (R. dent. ml–m3); 18786L (R. dent. ml–m3); 18877g (R. dent. m2–m3); 18928a (R. dent. m3); 21988 (R. dent. m3); 21989a (R. dent. ml–m3); 22019a (L. dent. ml–m3); 22019c (R. dent. m2–m3); 22019d (R. dent. ml–m2); 22020a (L. dent. m3); 22074a (R. dent. m2–m3); UA 10389 (L. dent. m2–m3). NISP = 27; MNI = 13.

Remarks.—In addition to several traits of the skull, key distinguishing features of Eliurus include moderately hypsodont molars with planar occlusal surfaces and undefined cusps. The molar crowns are configured as 3 laminae oriented nearly transverse to the longitudinal axis of the tooth, with individual lamina not connected by medial enamel connections (mures and murids), but uniting at their labial and lingual edges with wear. Eliurus myoxinus can be distinguished from other members of the genus by various subtle features of the occlusal morphology, such as the relative compression, angle, and completeness of the lamina; the degree of reduction of M3/m3 relative to the anterior molars; and size.

Eliurus is the most diverse rodent genus on Madagascar (Carleton 2003). Recent fieldwork has resulted in the discovery or resurrection of several new species in the northern and eastern rain forests, but E. myoxinus remains 1 of only 2 species thought to inhabit Madagascar's drier landscapes (Carleton and Goodman 2007). As would be expected in such a geographically widespread species, E. myoxinus demonstrates a tremendous amount of size variation, such that the northern samples from the Réserve Naturelle Intégrale d'Ankarafantsika average slightly but consistently smaller in most craniodental dimensions when compared with southern populations (Carleton et al. 2001; Carleton and Goodman 2007; Fig. 8). The mandibular specimens referred to E. myoxinus from Ankilitelo show characters of the dentition that are reminiscent of the western species, and fall within this range of size variation.

Fig. 8

Scatterplot showing separation of subfossil and modern specimens of Eliurus. Based on ml area, there are 2 statistically significant clusters within the referred subfossil E. myoxinus; however, this variation is encompassed within the broad range described for extant E. myoxinus. Eliurus sp. indet. is well outside of the range of E. myoxinus and is in fact larger than any known modern Eliurus species. The distinctive shape differences between E. myoxinus (buccolingually narrow molars) and Eliurus sp. indet. (more squared molars) also are captured. Ankilitelo Eliurus are represented by solid symbols, extant Eliurus by open symbols.

When comparing size dimensions of the subfossil sample of E. myoxinus, there are 2 significantly distinct groups on the basis of dental measures (ml area, t = 8.307, d.f. = 10, P < 0.001; Fig. 8). These 2 size groups also show subtle differences in occlusal morphology, such that the larger specimens have lamina that are buccolingually transverse and not mesially arcuate, and distal laminae that are reduced relative to mesial laminae. Future revisions based on field collections and genetic data may uncover a new species of Eliurus from southwestern Madagascar; however, these specimens of Eliurus are referred to E. myoxinus.

Eliurus sp. indet. Fig. 7D

Referred specimens.—DPC 17213d (R. dent. ml–m3); 17244b (R. dent. ml–m3); 17302d (R. dent. ml–m3); 18044a (R. dent. ml–m3); 18070a (L. dent. ml–m3); 18090c (R. dent. ml–m2); 18877e (L. max. M1–M3); 18898a (L. dent. m2); 18929a (R. dent. ml–m2); 22020b (L. dent. m2). NISP =10; MNI = 6.

Remarks.—There is subfossil material in the Ankilitelo sample that does not clearly fit within the hypodigm of Eliurus myoxinus. These specimens can be distinguished from E. myoxinus on the basis of their much larger overall dental dimensions, which fall above the known range for any described species of Eliurus (ml area, t = −7.787, d.f. = 23, P < 0.001; Fig. 8). Material referred to Eliurus sp. indet. can be readily distinguished by its much larger size, distinctively squared lower molars, particularly ml, and features of the occlusal morphology. Although patterns of morphological variation can be difficult to assess for this group (Carleton 2003), the Ankilitelo specimens may represent an unrecognized species. E. myoxinus is currently 1 of only 2 species of tufted-tailed rat thought to exist in western Madagascar (E. antsingy, the tsingy tufted-tailed rat, is found only in the dry deciduous forests of Bemaraha and Namoroka—Carleton et al. 2001). A number of studies have demonstrated the utility of DNA variation in illuminating potential species boundaries between morphologically similar, or cryptic taxa (Avise 2000; Olson et al. 2004; Yoder et al. 2000). Before the Ankilitelo specimens can be considered as representing an extinct form, the results of small mammal surveys of the Mikoboka Plateau and molecular analysis need to be considered in assessing species diversity in southwestern Eliurus.

Order Primates Linnaeus, 1758

Suborder Strepsirrhini É. Geoffroy Saint-Hilaire, 1812

Infraorder Lemuriformes Gray, 1821

Family Cheirogaleidae Gray, 1873

Subfamily Cheirogaleinae (Gray, 1873)

Microcebus É. Geoffroy Saint-Hilaire, 1834

Microcebus griseorufus Kollman, 1910

Gray-brown Mouse Lemur Fig. 9A

Fig. 9

Microcebus from Ankilitelo. A) M. griseorufus, partial skull DPC 18754a, right maxillary dentition in occlusal view; B) M. murinus, left maxilla DPC 18772g in occlusal view. Scale equals 1 mm.

Referred specimens.—DPC 17213c (R. max. P4–M2); 17242h (L. dent. p2–m3); 17259d (L. dent. p3–m3); 18069b (L. dent, edentulous); 18069c (R. dent. p4–m2); 18069d (R. dent. p4–m3); 18069e (L. dent. p2–p4); 18754a (partial cranium, R. P3–M3, L. P4–M2); 18772c (R. dent. p4–ml); 18772d (R. dent. ml–m2); 18772e (L. dent. p4–m2); 18784b (L. dent. p3–m3); 18784c (R. dent. p4); 18786b (R. dent, c–m3); 18786e (R. dent. m2–m3); 18801a (R. dent. p2–m3); 18842b (L. dent. m2); 18842h (L. max. P3–M3); 18902 (R. dent. p2–m3); 18926b (L. dent. p2, ml–m3); 21995a (R. dent. p4–ml); 21995b (R. dent. m2); 21995c (R. dent. p4–ml); 22052a (L. dent. p4–ml); 22053a (L. dent. m2); 22053b (R. dent. m2–m3); 22071c (R. dent. ml–m2); 22075 (R. max. P4–M3); 22078 (L. dent. p4–m3); 22093a (R. dent. p4–m2); 22093b (L. dent. p4–m3); 22093c (L. dent. p4); 22095d (R. dent. p4). NISP = 33; MNI = 16.

Remarks.—Two species of Microcebus are present in the Ankilitelo subfossil deposits. In light of current taxonomic revisions of the genus, this diagnosis was confirmed by examining the intraspecific variation in dental remains in detail (e.g., Cuozzo 2008; Cuozzo et al., in press). The degree of metric variation in 1st and 2nd lower molar lengths in the Ankilitelo sample of Microcebus exceeds that of a single geographic population of extant M. griseorufus from Amboasary (Table 3). In addition to its diminutive size, M. griseorufus can be distinguished from M. murinus (gray mouse lemur), also recovered at Ankilitelo, by several morphological characters. The P4 of M. griseorufus has a reduced protocone, giving the tooth a triangular occlusal shape. The M1 and M2 of M. griseorufus are distinguished by a slight indentation on the distal border of the tooth, coupled with a smaller hypocone on the distolingual shelf. The parastyle and metastyle appear to be slightly less developed than the condition in M. murinus. The lower 4th premolar has a prominent cingulid and distal heel, but these features are not as developed as in M. murinus. Additionally, M. griseorufus has distinctly squared m1–m2 instead of the mesiodistally elongated rectangular molars typical of other species of Microcebus, such as M. murinus and M. myoxinus (pygmy mouse lemur). The subfossil forms of M. griseorufus are significantly larger than their modern counterpart, and are not significantly different in size from either M. murinus or M. myoxinus (Muldoon and Simons 2007).

View this table:
Table 3

Metric variability compared between the subfossil sample of Microcebus from Ankilitelo and extant Microcebus griseorufus.

ml Lm2 L
nCVann
Subfossil Microcebusb217.69246.61
M. griseorufusc1083.411083.71
  • a All coefficients of variation (CVs) were corrected for sample size following Sokal and Rohlf (1995) and Plavcan and Cope (2001).

  • b Comparative data collected during this study.

  • c Data from Cuozzo (2008).

Microcebus murinus (J. F. Miller, 1777)

Gray Mouse Lemur Fig. 9B

Referred specimens.—DPC 13760x (R. m2); 13760y (R. m2); 18066b (R. dent. m2); 18069a (L. dent, edentulous); 18772g (L. max. P3–M3); 18772h (R. m2); 18784d (L. dent, edentulous); 18784e (R. dent, edentulous); 18784f (R. dent, edentulous); 18784g (R. dent, edentulous); 18786a (L. dent. p4–m3); 18786c (L. dent. p4, m2); 18786d (R. dent, edentulous); 18786f (R. max. P3–M3); 18794a (R. dent, edentulous); 18806a (R. dent, edentulous); 18842c (L. dent, edentulous); 18880a (R. dent, edentulous); 18899a (R. dent, edentulous); 18926a (R. dent, edentulous); 18928b (L. dent, edentulous); 18974e (L. p2); 18974f (L. C); 22052b (L. dent. m2); 22071a (R. dent, edentulous); 22071b (R. dent, edentulous); 22073 (L. max. P4–M3); 22095a (L. dent, edentulous); 22095b (R. dent, edentulous); 22095c (L. dent, edentulous). NISP = 30; MNI = 13.

Remarks.—Microcebus murinus is less common in the Ankilitelo deposits than M. griseorufus. It is distinguished from M. griseorufus by its larger size. Additionally, P4 has 2 cusps (a large paracone and smaller protocone), a prominent lingual cingulum, and a parastyle and metastyle. The buccal stylar region is better formed on M1–M2. The upper molars have greater development of the hypocone, giving the lingual and distal borders of M1–M2 a squared shape. There are well-developed cingulids on all lower premolars. The lower molars have spacious trigonids, contributing to the typically rectangular shape of m1 and m2.

Cheirogaleus É. Geoffroy Saint-Hilaire, 1812

Cheirogaleus medius É. Geoffroy Saint-Hilaire, 1812

Western Fat-tailed Dwarf Lemur

Referred specimens.—DPC 13760h (R. dent. p2–ml); 13765d (R. dent, edentulous); 17199k (R. p4); 17199L (L. p4); 17200g (R. dent. ml–m3); 17200J (partial cranium, L. M3); 17278d (partial cranium, R. Il, C–P3, L. I1–P3); 17290d (L. dent. p2–ml); 18055a (L. dent. p2–m2); 18772a (R. dent, edentulous); 18772b (R. dent, edentulous); 18772f (R. m3); 18773a (R. m2); 18773b (R. p4); 18784h (L. dent, edentulous); 187784i (L. dent, edentulous): 181874j (L. dent, edentulous); 18784k (R. dent, edentulous); 18842d (R. dent. p4); 18908a (R. dent, edentulous); 21994a (R. dent, edentulous); 21994b (R. dent, edentulous). NISP = 21; MNI = 10.

Remarks.—A recent analysis of field and museum specimens, considering both the genetic and morphological variation, recognizes only 3 clades of CheirogaleusC. medius, C. major, and C. crossleyi (Groeneveld 2008). The Ankilitelo material is referred to the C. medius group based on the upper canine with a moderately developed lingual talon, presence of a diastema between the anterior and middle maxillary premolars, upper premolars that are oval in occlusal view, and more bunodont molar cusps (Groves 2000).

Family Lepilemuridae Gray, 1870

Subfamily Lepilemurinae (Gray, 1870)

Lepilemur I. Geoffroy Saint-Hilaire, 1851

Lepilemur leucopus (Major, 1894)

White-footed Sportive Lemur

Referred specimens.—BPC 13688 (R. dent. p2–m3); 13710 (R. dent. p2–m3); 13760i (R. dent. p3–m3); 17199f (R. dent. p3–m3); 17199g (R. dent. p4–m3); 17199j (L. p3); 17200b (R. max. P3–M3); 17213a (R. max. C); 17213b (L. max. C); 17253f (L. dent. m2–m3); 17278e (L. dent. p4–m3); 18053 (R. dent. p2–m3); 18066c (L. max. P3–M3); 18117a (R. dent. p4–m3); 18117b (R. max. P3–P4); 18117c (R. M1); 18117d (R. max. M2–M3); 18765 (L. dent. p3–m3); 18784a (R. dent. p4–m3); 18800a (R. p2); 18800b (L. p3); 18800c (L. dent, c, p3–m3); 18800d (R. dent. p4–m3); 18839 (complete mandible, L. il–m3, R. il–m3); 18842a (R. dent. ml–m2); 18896 (partial cranium, L. C–M3, R. C–M3); 18901a (R. dent. p4–m3); 18901b (R. dent. m2–m3); 18927a (L. dent. m2–m3); 18927b (R. dent. p4–m3); 18951a (L. dent. p2–m3); 22048 (R. max. M2–M3); 22094 (R. dent. p3–m3). NISP = 33; MNI = 15.

Remarks.—Lepilemur is the only extant genus of Malagasy primate that lacks permanent upper incisors. The dentition is distinctively trenchant, with well-developed shearing crests adapted for a folivorous diet. The upper canines are long, incisive, and curve distally, but are compressed buccolingually. A lingual pillar separates the lingual surface into mesial and lingual grooves. P2 is caniniform, whereas P3–P4 are bicuspid. The upper molars have a well-developed paracone, metacone, and protocone, but the hypocone is not present. The lower molars are cuspidate and cresty, appearing W-shaped in occlusal view because of the well-developed ectolophid. The lower 3rd molar has a large hypoconulid that is separated from the entoconid by a large notch. The Ankilitelo specimens of Lepilemur are intermediate in size between L. leucopus and L. ruficaudatus (red-tailed sportive lemur—Muldoon and Simons 2007). The Ankilitelo specimens can be allocated to L. leucopus by the more buccolingually compressed upper canine with a smaller distal heel, and lower premolar morphology. In L. leucopus, p3 is buccolingually compressed and has no metaconid, unlike the condition in L. ruficaudatus. The lower 4th premolar of L. leucopus differs from that of L. ruficaudatus in that it is not molariform, the mesial aspect of the tooth is not expanded, and there is no trigonid present.

Family Lemuridae Gray, 1821

Subfamily Lemurinae (Gray, 1821)

Lemur Linnaeus, 1758

Lemur catta Linnaeus, 1758

Ring-tailed Lemur Fig. 10A

Fig. 10

Lemurids from Ankilitelo. A) Lemur catta, partial skull DPC 18753b, right maxillary dentition in occlusal view; B) Eulemur rufus, left maxilla DPC 17199e in occlusal view. Scale equals 5 mm.

Referred specimens.—DPC 13689 (L. dent. p3–m3); 13701 (R. dent. p3–m3); 13702 (L. dent. p3–m3); 13706b (juvenile, L. dent, edentulous, m2 in crypt); 13709 (R. max. P2–M3); 13722a (R. p2); 13736c (L. C); 13760f (R. dent, ml); 13764a (R. dent. p3–m3); 13765a (L. dent. p2–m3); 13765c (partial cranium, braincase); 13765e (R. max. P2–P4); 13775a–2 (L. max. P4–M2); 17156 (R. max. P2, P4–M3); 17209a Guvenile, R. dent. dp3–ml, m2 in crypt); 17209c (L. C); 17234 (partial cranium, braincase); 17242a (R. dent. p2–m3); 17242c (R. dent. ml–m3); 17242e (L. max. P3–M3); 17251a (L. max. C, P4–M3); 17253d Guvenile, L. dent. dp4–ml); 17253e (L. dent. p3–m3); 17259a (R. dent. p3–m3); 17259b (R. dent. p3–m3); 17259c juvenile, L. dent. dp4–ml); 17259m (R. c); 17264a (R. max. P2–P3); 17278a-l (L. dent. p3–m3); 17278a–2 (L. dent, edentulous); 17278a–3 (R. max. M1–M3); 17290a (R. dent. p3–m3); 17290c (L. p2); 17290f (L. incisor); 17302b (L. dent. p3–ml); 17303a (R. dent. m2–m3); 17320a (L. dent. p2–m3); 17320b (L. dent. p3–m3); 17320c (L. dent. p3–m3); 17335a (L. dent. p2–m3); 17335d (R. dent, il, c, p2); 18052a (L. dent. p4–); 18052b (R. max. I2–P3); 18108 (L. max. P3–M1, M3); 18109a (R. dent. p3–m2); 18109b (R. dent. m3); 18110b (R. dent. ml–m3); 18110c (L. p4); 18121a (R. C); 18168a (R. M2); 18199b (R. dent. p4–m3); 18199c (R. dent. m2); 18199d (R. dent, il–ml); 18753a (L. dent. p4–m3); 18753b (partial cranium, L. P3–M3, R. P3–M3); 18764b (L. dent. p3–m3); 18776b (L. dent. m2); 18776c (R. dent. p4–m3); 18776d (R. dent, edentulous); 18794b (L. p3); 18842e (R. C); 18842f Guvenile, R. max. M1); 18900a (L. dent. m3); 18900b (R. dent. p3, ml–m2); 18925a (R. dent. ml–m3); 18794c (L. dent, edentulous); 22004 (L. dent. ml–m3); 22042a (R. dent. edentulous); 22081a (L. c); UA 10005 Guvenile, L. dent. dp3–m2); 10006 (R. dent. ml–m3); 10007 (R. dent, c, p3, ml–m2); 10075 (R. dent. p4–m3); 10076 (R. dent, ml); 10077 (L. dent. m2–m3); 10078 (R. dent. ml–m3); 10079 (R. max.); 10080 (L. max); 10089 (R. dent. m2–m3); 10090 (R. dent, il, p2– m3); 10091 Guvenile, R. dent. il–m2); 10092 (L. max. I2–C); 10093 (R. max); 10139 (L. dent. c–p4); 10140 (L. dent, ml– m3); 10141a (R. dent. p3–ml); 10141b (R. max. edentulous); 10141c (L. I); 10141d (L. c); 10141e (R. c); 10141f (L. p4); 10141g (L. m2); 10141h (L. m3); 10142 Guvenile, R. dent. m2 in crypt); 10143 Guvenile, L. dent, ml, m2 in crypt); 10147 (R. dent. ml–m3); 10148 (L. dent, c, p4–m3); 10168 (R. dent. p3–m3); 10169 (R. max. P4–M3); 10171a (R. max. P4–M3); 10171b (L. max. M2–M3); 10171c (L. M1); 10171d (fragmentary maxillary teeth); 10172 (L. max. C, P4–M1); 10173 (R. max. P3–M3); 10178 (R. dent. p3–m3); 10179 (L. dent. m2–m3); 10180 (L. dent. p4–m2); 10181a (L.P2); 10181b (R. P2); 10181c (fragmentary maxillary teeth); 10190 (L. P2); 10191 (R. P2); 10198 (R. max. P3–M3); 10199 (L. max. M1–M3); 10200 (R. max. P4–M1); 10239 (L. max. P3–M3); 10454 (partial cranium, edentulous); UA 9622 (partial cranium, R. I2–M3, L. P2–P4, M2–M3); 9763 (R. dent. p4–m3); 9764 (L. dent. p3–m3); 9765 (R. dent. m2–m3); 9766 (R. dent, edentulous); 9767 (L. dent. p3–m3); 9768 (R. P2); 9769 (L. max. P2–P3); 9770 (L. M1); 9771 (R. max. M1–M3); Unno (upper molar). NISP = 127; MNI = 38.

Remarks.—Lemur catta is the most abundant small mammal in the Ankilitelo sample. Maxillary specimens referred to Lemur can be differentiated from other lemurids because the upper molars lack a protostyle. This character is variable in species of Eulemur, in which the protostyle can be small, or very large and prominent. This variability presents difficulties in distinguishing large specimens of Lemur from Eulemur. Maxillary specimens of Lemur can be further distinguished from Eulemur based on the following characters: smaller overall size, upper canine is not as robust and is more buccolingually compressed; P2 is buccolingually compressed relative to the condition in Eulemur; P3 is less complex, with no parastyle, no lingual ridge, and is not as trenchant; and P4 has a much smaller protocone. Mandibular specimens of L. catta differ from those of Eulemur by their smaller overall size. Further, p3 is unicuspid, less complex, and is buccolingually compressed with a slight to absent talonid heel. The p4 has a metaconid that is not set as high on the flank of the protoconid and is less distinct than the condition seen in Eulemur. The largest lower molar in the toothrow is m2, such that ml < m2 > m3. The lower molars are smaller with cusps being more distinct from the marginal ridge. The lower molars have a distinctive but small metastylid distal to the metaconid, which is less pronounced down the toothrow. The entoconid is distinct from the marginal ridge, and the entoconid notch is broad and distinctly U-shaped. The distal border of the talonid is not as rounded as it is in Eulemur, but is distinctly squared. The Ankilitelo specimens of L. catta are larger than their modern counterpart (Muldoon and Simons 2007).

Eulemur Simons and Rumpler, 1988

Eulemur rufus (Audebert, 1799)

Red-fronted Brown Lemur Fig. 10B

Referred specimens.—DPC 13685a (partial cranium, R. C–M3, L. C–M3); 13685b (R. dent. c–m3); 13685c (L. dent, c–m3); 13690 (R. dent. p4–m2); 13691 (R. dent. p4–m2); 13725c (juvenile, R. dent, edentulous, ml in crypt); 13760g (L. M2); 13760J (R. c); 13764b (R. max. P3–P4, M2–M3); 13775a-l (L. dent. p4–m3); 17199a (R. dent. p4–m3); 17199b (L. dent. p4–m3); 17199c (L. dent. p4–m3); 17199d (R. max. P2–M1); 17199e (L. max. P2–M3); 17200a (juvenile, R. max. dP2–dP4, M1); 17200c Guvenile, L. dent. dp4); 17200d (juvenile, L. dent. dp4); 17200e (L. dent. m2); 17200h (L. max. M2–M3); 17200i (L. M1); 17200m (L. max. edentulous); 17202a (L. max. C, P4); 17202b (L. max. M2–M3); 17209b (R. max. P3–M3); 17242b-1 (R. dent. i2, p2, p4); 17242b-2 (R. dent. m2); 17242d (L. max. C, P2–P3); 17242f (R. m2); 17242g (R. m3); 17244a (R. C); 17253a (L. dent. p4–m2); 17253b (R. dent. p3–m2); 17253c (L. dent. p3–ml); 17253g (L. m3); 17259e Guvenile, L. max. C, dP2, dP3, dP4, M1); 17259f (R. max. C. dP2, dP3, dP4); 17278b-l (L. dent. m2– m3); 17278b-2 (R. dent, edentulous); 17290b (L. dent. p4– m2–m3); 13702a (juvenile, L. dent, ml); 17302c (R. max. P2–P3); 17314a (L. dent. p3–m3); 17314b (L. max. M2); 17314c (L. max. P3); 17314d (L. P4); 17314e (L. max. P3–M3); 18056a (R. max. 11–12, P2); 18110a (L. dent. p2, p4–m3); 18118 (partial cranium, edentulous); 18124a (L. dent. p3–p4); 18199a (juvenile, L. dent. dp4–ml); 18764a (R. dent. m2– m3); 18764c (R. max. I1–); 18776a (R. dent. p4–m2); 18785a (L. M2); 18795a (R. dent. p4–ml); 18796a (L. C); 18796b (R. C); 18842f (L. max. M1–M3); 18948 (R. dent. i2– c, p3–m3); 18949 (L. max. P3–M3); 22043 Guvenile, L. dent. dp4–ml); UA 9492 (R. max. P3–M3); 9623 (L. dent. p3–p4, m2); 9624 (L. dent. p4–ml, m3); 9625 (R. dent. m3); 9626 (R. dent. p3–ml); 9627 (R. dent, edentulous). NISP = 69; MNI = 19.

Remarks.—Eulemur rufus is abundant in the subfossil sample. On average, E. rufus is larger than Lemur catta. Maxillary specimens referred to E. rufus are differentiated from those of L. catta by the presence of a protostyle on the upper molars. In addition, P2 is larger and less buccolingually compressed with a distinct lingual cingulum. P3 has a large paracone and distinct lingual ridge extending from the crown tip to base, which terminates in a parastyle. P3 is taller in buccal view than P2. P4 has a paracone and a protocone, with a small protostyle and distinct lingual cingulum. M1 and M2 have a hypocone and protostyle, making the occlusal outline of the tooth very square. M3 is smaller relative to the anterior molars, with no hypocone or protostyle. In the lower dentition, p3 is more complex than the condition seen in Lemur catta, and sometimes has a small metaconid high on the lingual flank of the protoconid. The talonid heel of p3 is larger and more squared in appearance because of the presence of the metaconid. Lower molars in Eulemur have marginal ridges that incorporate the cusps, and rounded distal borders. The specimens of E. rufus from Ankilitelo are smaller than their modern counterpart (Muldoon and Simons 2007).

Family Indridae Burnett, 1828

Propithecus Bennett, 1832

Propithecus verreauxi A. Grandidier, 1867

Verreaux's Sifaka

Referred specimens.—DPC 18066a (R. dent. ml–m3). NISP = 1; MNI = 1.

Remarks.—Propithecus can be distinguished from other primates in the subfossil assemblage by its larger size and by the W-shaped ectolophid of lower molars when viewed occlusally. This pattern is shaped in ml–m2 by well-formed protoconids, metaconids, hypoconids, and entoconids connected by crests. The protoconid is connected to the parastylid by a paracristid, and the cristid obliqua interlocks the distal surface of metaconid to hypoconid. The lower 3rd molar has a distinct hypoconulid, which is widely separated from the entoconid. P. verreauxi of the extreme south can be differentiated from P. deckenii, a western species that occurs in the dry forest formations north of the Tsiribihina River, by its smaller size and less trenchant crests.

Discussion

The importance of pit caves is demonstrated by the excellent preservation of small vertebrates in association with abundant remains of extinct taxa. With a total of 34 different mammal species recognized, Ankilitelo is one of the most diverse subfossil faunas in Madagascar. All of the subfossil small mammals reported from Ankilitelo are attributable to modern taxa, and nearly all of these are endemic species.

Range contractions.—The occurrence of extant species in the subfossil deposits at Ankilitelo that do not live in the region of the cave today demonstrates that there have been significant geographic range contractions of several taxa, ranging from approximately 100 to 500 km, in the last 500 years (Fig. 11). In particular, the shrew tenrecs M. nasoloi, M. brevicaudata, and M. majori, the carnivoran G. elegans, and the rodent H. antimena have modern distributions in areas of Madagascar that are more humid than the region surrounding Ankilitelo Cave on the Mikoboka Plateau today. The presence of these taxa at Ankilitelo may suggest that the climate of the southwest was not always as dry as it is now.

Fig. 11

Map showing modern distributions of still-extant mammals found in the subfossil fauna at Ankilitelo. A) Microgale; B) carnivorans; C) Hypogeomys; and D) Lepilemur. Abbreviations: Ge, Galidia elegans; Gg, Galidictis grandidieri; Ha, Hypogeomys antimena; Ll, Lepilemur leucopus; Lr, Lepilemur ruficaudatus; Mb, Micro gale brevicaudata; Md, Mungotictis decemlineata; Mm, Microgale majori; and Mn, Microgale nasoloi. Ankilitelo is represented by a black star.

Shrew tenrecs (Microgale) are generally associated with the eastern rain forests, although specimens have recently been recorded from western and southwestern Madagascar (Jenkins and Goodman 1999; Olson et al. 2004). M. nasoloi was named on the basis of 2 specimens (Jenkins and Goodman 1999). Given the efficiency of recent small mammal surveys in this area, M. nasoloi is considered extremely rare. Today, M. nasoloi is restricted to 2 forests in a small portion of southwestern Madagascar (Vohibasia and Analavelona). These humid forests are considered floristically transitional between eastern humid and western dry deciduous forest, yet structurally they are similar to dry deciduous forest (Morat 1973). In contrast, M. brevicaudata has been recorded in the humid forests of eastern Madagascar, in the Sambirano (although it was not found in a more recent survey of the same area—Goodman 2003 a), and in western Madagascar (Goodman 2003a; Goodman et al. 1996; Goodman and Jenkins 2000; Raxworthy and Rakotondraparany 1988). Several intensive small mammal surveys have been conducted in southwestern Madagascar, and no evidence of this species has been found. It is likely that the Mangoky River is a large barrier to dispersal, especially for small-bodied nonvolant species. M. majori is the most widely distributed shrew tenrec. It occurs today in lowland and midaltitude eastern rain forest from Montagne d'Ambre in the extreme north as far south as the Andringitra Massif, as well as in Kirindy Forest (Centre de Formation Professionnelle Forestière), and Analavelona Massif (Olson et al. 2004). Ankilitelo is located outside the confirmed geographic range for all 3 of these shrew tenrec species. The presence of M. nasoloi, M. brevicaudata, and M. majori at Ankilitelo is evidence that these allopatric taxa once lived in sympatry, well outside of their current ranges (Fig. 11A ).

Among the carnivorans, there are 3 exceptional occurrences in the Ankilitelo subfossil fauna. Today, G. elegans is the most widely distributed of the native Malagasy Carnivora. Three subspecies occur in distinct geographic areas: the eastern humid forest belt (G. e. elegans); the extreme northwest, including the Sambirano region, Montagne d'Ambre, and drier forests around Ankarana (G. e. dambrensis); and the dry forest areas around the limestone karst formations at Bemaraha and Namoroka (G. e. occidentalis—Goodman 2003c; Fig. 11B). No part of the range of any of these 3 subspecies includes the any part of the Mikoboka Plateau. The presence of Galidia at Ankilitelo represents the 1st extant or subfossil report for this species from the southwest.

In contrast to the nearly ubiquitous distribution of G. elegans, M. decemlineata and G. grandidieri are among the rarest carnivorans in Madagascar. The modern range of Mungotictis was thought to be restricted to the succulent woodland ecoregion between the Tsiribihina River to the north and the Mangoky River to the south (Fig. 11B). In recent surveys of the Forêt de Mikea, 2 specimens were collected that confirm the existence of a southern subspecies in the vicinity of the Mikoboka Plateau (Goodman et al. 2005; Thomas and Kidney 2005). The Mungotictis at Ankilitelo represent the 3rd and 4th specimens of adults ever recorded, and confirm the existence of a subspecies that is locally endemic to the Southern Mikea region. Galidictis was described in 1986, and until 1989 was known only from 2 museum specimens (Wozencraft 1986, 1990). The only confirmed locality for this species is in southwestern Madagascar at Tsimanampetsotsa (Goodman 2003d; Mahazotahy et al. 2006). Subfossil remains of Galidictis also have been reported from Ankazoabo Cave, about 50 km south of Lac Tsimanampetsotsa (Goodman 1996). The presence of Galidictis in the subfossil deposits at Ankilitelo suggests that this species occurred well to the north of its current range in the recent past. Further survey work is needed in the remaining forested areas to determine the extent of its distribution. Based on the subfossil record at Ankilitelo, Galidia, Mungotictis, and Galidictis were once sympatric in this region of southwestern Madagascar.

The endemic rodent H. antimena currently survives as a single extant population in the Menabe region of southwestern Madagascar, near Morondava. Previous paleontological work has demonstrated that H. antimena once had a broad distribution in southwestern Madagascar (Fig. 11C; Goodman and Rakotondravony 1996). The record from Ankilitelo further substantiates that Hypogeomys has suffered a dramatic range contraction within the last 500 years, likely associated with habitat destruction (Goodman et al. 2006).

There is 1 unusual primate occurrence in the subfossil fauna at Ankilitelo, L. leucopus. It is currently thought that each of the recognized species of sportive lemur occupies a distinct range around Madagascar's periphery (Thalmann and Ganzhorn 2003; but see Craul et al. 2008). Today, the southwestern population of Lepilemur (considered by us to be L. leucopus, but see Louis et al. 2006b) is thought to be limited to the north by the Onilahy River, where it is replaced by L. ruficaudatus (Mittermeier et al. 2006; Fig. HD). The presence of L. leucopus at Ankilitelo is a northern extension of the confirmed range for this species. However, the actual distribution of sportive lemurs in the southwest may be more complex than currently thought (Andriaholinirina et al. 2006; Craul et al. 2008; Godfrey et al. 1999; Louis et al. 2006a; Seddon et al. 2000). Individuals sighted in 1998 in the region of Ankilitelo had gray coats with white ventra, as do L. leucopus, and seemed to lack the red highlights and reddish tails that distinguish L. ruficaudatus.

What factors lead to distributional change?—There are several possible explanations for the apparent change in the small mammal community at Ankilitelo. One possibility is that climate change altered the environment in southwestern Madagascar in ways that exceeded the physiological tolerances of species demonstrating range contractions. Evidence for climatic shifts in southwestern Madagascar is based on a pollen core from the subfossil site Andolonomby, which indicates a mesic period starting 5,000 years ago followed by climatic aridity between 3,500 and 2,500 years ago (Burney 1993). When humans arrived on the island approximately 2 millennia ago, this arid phase had shifted the vegetation from dry forest toward a palm savanna and spiny bushland. Furthermore, the existence of relict patches of eastern rain-forest flora and fauna in the west, in the region of Zombitse-Vohibasia and the mist oasis called Analavelona Massif, serves to underscore recent habitat changes in southwestern Madagascar (Jenkins and Goodman 1999; Nicoll 2003). Elements of the extant fauna and flora of these areas indicate that these isolated humid forests, and in particular Analavelona Massif, may be a refuge for biota that had much more extensive distributions in southwestern Madagascar in the recent past (Raxworthy and Nussbaum 1997). However, there is an absence of evidence for significant climate change in southwestern Madagascar in the last 500 years. As such, climate change seems unlikely to explain the unusual occurrences at Ankilitelo.

A 2nd hypothesis is that recent fragmentation and human-initiated degradation of forested habitats explains the disjunct distributions of the some of the Ankilitelo small mammals. The palynological data from Andolonomby indicates that anthropogenic changes began at 1,700 years ago with a sharp decline in the dung fungus Sporormiella, a proxy for megafaunal biomass, followed by a concomitant rise in charcoal values within a few centuries (Burney et al. 2003). This appears to indicate human hunting of naïve megafauna, followed by increased fire as grazing pressure declined (Burney et al. 2004). Subsequent changes documented by the pollen record include the decline of trees associated with closed formations, and an increase in grasses and other open-country, fire-tolerent species (Burney 1993). Examination of these data suggests a complex series of human-caused events, culminating in widespread landscape transformation in the southwest shortly after human arrival. Furthermore, analysis of aerial photographs and Landsat images indicates recent dramatic fragmentation of western dry forest, and southwestern spiny forest from 1953 to 2000 (Harper et al. 2007). The detrimental effects of forest fragmentation on small mammals in western and southern Madagascar have been documented (Ganzhorn et al. 2003; Scott et al. 2006), although factors constraining the current distribution of forest-limited species remain unclear. To further investigate this hypothesis, ecological habits of forest-limited species, and resource availability within forested and fragmented areas need to be assessed in detail, because current information is limited. For example, the disharmonious association of components of the fauna at Ankilitelo could be the result of individualistic ecological responses of species to habitat disturbance, as documented elsewhere for extinct and extant species (e.g., Alroy 1999; Barnosky 2004; Bellows et al. 2001; Brant and Orti 2003; Cannon 2004; Runck and Cook 2005).

Finally, it is possible that community-level processes, such as competition, predation, and hunting by humans could have resulted in distributional changes observed in the Ankilitelo fauna. Introduced species, such as Rattus and Mus, are of particular importance because evidence indicates they may be replacing endemic rodents through competition (Goodman 1995). Rattus (NISP = 28) and Mus (NISP = 4) are rare at Ankilitelo compared to endemic taxa, suggesting that competition was not a major factor structuring this community 500 years ago. Census data comparing the abundance of endemic and introduced rodents in southwestern forests today would shed light on this hypothesis. Likewise, increasing evidence supports the idea that endemic avian and mammalian predators, as well as human hunters, have profoundly impacted mammal populations in Madagascar (e.g., Brockman et al. 2008; Goodman 2003b; Karpanty 2006). For animals isolated in fragments, natural or invasive predators can be major causes of declines and extirpations (Crooks and Soulé 1999; Irwin, et al. 2009; Rasoloarison et al. 1995). Taphonomic evidence of predation at Ankilitelo supports this hypothesis, although there is no evidence of hunting by humans on the subfossil remains.

Much remains to be understood in emerging regional syntheses of Madagascar's paleoenvironments. Given the ecological variation in southwestern Madagascar during the Holocene and in recent times, a single coherent explanation for the unusual small mammal occurrences at Ankilitelo is not obvious. Synergistic interactions among the hypotheses out-lined above likely play a key role in interpreting the biogeography of Madagascar's extant fauna (Burney et al. 2003, 2004). Although the causes of the changes in the small mammal fauna are still far from being clear, the Ankilitelo fauna adds useful details regarding the past of the southwestern region of the island, and underscores the need for increased attention to this understudied portion of Madagascar's history.

Acknowledgments

We are grateful to the members of the Ankilitelo field crews, whose caving experience and bravery were invaluable, especially C. Hildreth, M. Minton, M. Oliphant, N. Pistole, T. Rasmussen, C. Savvas, V. F. H. Simons, G. Hermas Randriatahina, F. Ravoavy, L. Rakotondramanga, S. Rabenjarisoa, and F. Vololontsoa Randrianasolo. We especially acknowledge the support of the late Gisèle Randria, the late Madame Berthe Rakotosamimanana, and A. Rasoamiaramanana (UA Département de Paléontologie et Anthropologie Biologique) for facilitating our research at Ankilitelo. For access to skeletal collections in their care, we thank J. Randrianasy (UA), L. Gordon (Smithsonian Institution, Washington, D.C.), E. Westwig (American Museum of Natural History, New York, New York), J. Chupasko (Museum of Comparative Zoology, Harvard University, Cambridge, Massachusetts), and B. Stanley (The Field Museum, Chicago, Illinois). Scanning electron microscopy images were taken by C. Daghlian at the Electron Microscope Facility, Dartmouth Medical School. We thank G. Gunnell, S. Goodman, and T. Rasmussen for helpful comments on earlier drafts of this paper. S. Dobson assisted with the preparation of figures. We thank L. Godfrey and 1 anonymous reviewer for improving the manuscript. This research was supported by The Field Museum, Geological Society of America, Sigma Xi, National Science Foundation (BCS 0408732), American Philosophical Society, Claire Garber Goodman Fund, and Department of Anatomy Dartmouth Medical School (KMM), National Geographic Society (7692-04), and National Science Foundation (SBR 96-30350) (ELS). This is Duke Lemur Center publication 1139.

Footnotes

  • Associate Editor was Elizabeth R. Dumont.

Literature Cited

View Abstract