We describe a partial skeleton of a new species of Bubalus (Bubalus) from soft karst near Balamban, Cebu Island, Philippines. The specimen is likely Pleistocene or Holocene in age and includes left and right humeri, a left metatarsal, 2 posterior thoracic vertebrae, 2 left lower molars, and a pair of ungual phalanges. Bubalus sp. nov. differs from all previously described Bubalus in both the size and proportions of the skeleton and in possessing a unique combination of discrete character states. Possible autapomorphies for Bubalus sp. nov. evident in the metatarsal include a very broad dorsal longitudinal sulcus; a broad, triangular anterior cubonavicular facet; and a sulcus that bisects a small tuberosity on the proximolateral surface. Limb elements of Bubalus sp. nov. are less than two-thirds the length of corresponding elements of the Asiatic water buffalo, B. (Bubalus) bubalis, and are about 80% the length of those of the tamaraw, B. (Bubalus) mindorensis; they are similar in length to limb bones of the lowland anoa, B. (Anoa) depressicornis, but are more robust. Mass estimates based on regression equations for modern bovids suggests a mass of 150–165 kg for Bubalus sp. nov.; this is approximately 25% smaller than B. mindorensis (180–220 kg) and at least 15% larger than B. depressicornis (approximately 135 kg). The small size of Bubalus sp. nov. relative to other B. (Bubalus) is likely attributable to island dwarfing; this is supported by a consistent relationship between body size and island size in Bubalus sp. nov., B. mindorensis, and B. bubalis, and by the relatively larger dentition of B. sp. nov. relative to body size. Bubalus sp. nov. is the 1st fossil mammal to be reported from Cebu Island and is the only nonproboscidean documented from the Negros–Panay Philippine Faunal Region. In conjunction with the presence of Bubalus on Mindoro Island (and potentially Luzon), this specimen suggests that Bubalus may once have ranged throughout the Philippines.
The Philippines comprise more than 7,000 islands situated between Borneo and Taiwan in the northeastern corner of the Malay Archipelago. The vast majority of these islands are tiny (only 1–2 km2) but the 2 largest (Luzon and Mindanao) are each approximately 100,000 km2 in area (Fig. 1). Although a few of the islands (the Palawan group) are continental, most are oceanic in origin, and their current configuration is the result of complex tectonic interactions among the Eurasian continental plate, the Philippine oceanic plate, and various microcontinental blocks (e.g., Hall 1998, 2002; Yumul et al. 2000; Zamoras and Matsuoka 2004).
The fractured geography of the Philippine islands and their variable degrees and timing of connection and separation have given rise to a diverse fauna of terrestrial vertebrates. At least 172 species of native mammals have been recorded, of which 111 (64%) are endemic (Heaney et al. 1998). This degree of endemism is exceeded only by Madagascar, an island of nearly twice the total area; on a per-area basis, the Philippines may harbor the greatest number of endemic mammals in the world (Heaney 1993). Because of this endemism, the Philippines has been recognized both as one of the most “megadiverse” countries and as one of the “hottest” biological hotspots, making it a top priority for conservation efforts (Ricketts et al. 2005; Shi et al. 2005). Newly discovered mammals continue to be described from the region (e.g., Rickart et al. 2002, 2003, 2005), indicating that the true species richness of the country is still unknown.
The large number of Philippine islands and their diverse mammal faunas make the archipelago particularly amenable to studies of island biogeography and the relative effects of colonization, extinction, and speciation (Heaney 1986, 2000). Such investigations have been aided by the recognition that the present configuration of islands is a geologically recent event. During the last glacial maximum of the late Pleistocene (approximately 20,000 years ago), the great quantity of water trapped in polar and continental ice sheets lowered sea levels worldwide, exposing areas of land that had previously been submerged (e.g., Bird et al. 2005; Meijaard 2003). In the Philippines, these exposed areas connected many previously isolated islands, resulting in 6 major Philippine faunal regions (Heaney 1991; Heaney et al. 1998; Fig. 1). Although these islands once again became isolated after the Pleistocene, disparities in faunal resemblance between intraregional and interregional island pairs are still dramatic (Heaney 1986; Heaney and Regalado 1998; Heaney et al. 1998) and may differ from those that existed before that event.
Unfortunately, the terrestrial fossil record of the Philippines is poor and has provided few insights into the development of the country's unique fauna. Thus far, only a few Pleistocene and Holocene ungulates have been reported (Bautista 1991; Beyer 1957; Koenigswald 1956; Table 1), although increased efforts to collect and study microvertebrate fossils could considerably expand knowledge of the recent biotic history of the islands (Reis and Garong 2001). The present report describes fossil remains of a new species of bovid that lived on Cebu Island, perhaps as recently as a few thousand years ago. The specimen was discovered in 1958 by Michael Armas, a mining engineer, during exploration for phosphate (Fig. 2). In 1995, it was brought to the attention of Dr. Hamilcar Intengan, who subsequently brought the bones to The Field Museum for initial identification by our late colleague, Steven McCarroll, and JJF. Armas and LRH visited the site in April 1999; it is an area of soft karst, formed from poorly consolidated coral reef. The fossil was found at the end of an approximately 10- to 11-m horizontal tunnel that had been dug into the side of a ridge for the purpose of phosphate mining. The sediments in which the fossil was found likely represent fissure-fill (i.e., sediments that had accumulated within a crack or crevice in the limestone). All elements of the specimen were found on a single day, in close proximity to each other, within loose matrix. No other fossils were found at this site or in other similar mining tunnels dug in the vicinity. This is the 1st fossil mammal reported from Cebu Island, and the only nonproboscidean fossil described from the entire Negros–Panay Faunal Region of the Philippines.
Materials and Methods
The descriptions below are based on direct observations of the original material (left humerus, vertebrae, teeth, and unguals) or casts of the original material (right humerus and left metatarsal) that are now housed at the National Museum of the Philippines (PNM). Morphological comparisons were made with osteological specimens from the Recent mammal collections of the Division of Mammalogy at The Field Museum of Natural History (FMNH) and the University of Michigan Museum of Zoology (UMMZ); a list of these specimens is provided in Appendix I.
All measurements were made to the nearest 0.5 mm, using a digital caliper. Estimated measurements (e.g., for articulated specimens) are indicated by parentheses. Statistical analyses were performed using SPSS (SPSS Inc., Chicago, Illinois) on an Apple G4 computer (Capertino, California). The dagger symbol (†) is used to designate extinct species.
Mammalia Linnaeus, 1758
Artiodactyla Owen, 1848
Bovidae Gray, 1821
Bovinae Gray, 1821
Bovini Gray, 1821
Bubalina Pilgrim, 1939
Bubalus Hamilton-Smith, 1827
Bubalus (Bubalis) bubalis, †B. brevicornis, B. (Anoa) depressicornis, †B. guzhensis, †B. mephistopheles, B. (Bubalus) mindorensis, †B. murrensis, †B. palaeindicus, †B. palaeokerabau, †B. platyceros, B. (Anoa) quarlesi, †B. sivalensis, †B. teilhardi, †B. tingi, †B. triangulatus, †B. wansijocki, †B. youngi.
As is evident from the list above, in addition to the extant species, a large array of fossil species have been referred to Bubalus. The majority of these species are incompletely known, however, and are differentiated primarily by horn core characters (e.g., Geraads 1992; Xue and Li 2000; Young 1936). The new species described herein is represented by 2 teeth and various postcranial elements and therefore cannot be directly compared to most fossil taxa. Of necessity, therefore, the diagnosis below focuses on character states that distinguish the new Cebu Island species from extant species of Bubalus, the only taxa for which sufficient postcranial specimens are available for diagnostic differentiation. The new species can only be stated with certainty to differ from all fossil species of Bubalus by its dramatically smaller size.
Four extant species of Bubalus are currently recognized (Grubb 2005; Nowak 1999). The most widespread of these is Bubalus bubalis, the Asiatic water buffalo; domestic B. bubalis is found virtually throughout the world (Kierstein et al. 2004), but wild populations have declined and are considered endangered (Nowak 1999). Water buffaloes are the largest Bubalus, with wild males weighing more than 1,000 kg (Popenoe 1983). Some authorities distinguish between the wild and domestic forms, using the name B. arnee for the former and reserving B. bubalis for the latter (e.g., Geraads 1992; Groves 1969). We do not make this distinction in the present report because the 2 species cannot be readily distinguished from one another by anatomical differences. Bubalus mindorensis, the tamaraw or tamarau, is endemic to the Philippine island of Mindoro (Custodio et al. 1996). It is much smaller than B. bubalis (see below) and is highly endangered because of a variety of factors, including hunting and habitat destruction (Heaney and Utzurrum 1991; Oliver 1993). B. bubalis and B. mindorensis are generally grouped together in the subgenus Bubalus (Grubb 2005). The subgenus Anoa includes 2 very similar species of dwarf buffalo endemic to Sulawesi, Indonesia: B. depressicornis, the lowland anoa, and B. quarlesi, the mountain anoa (Burton et al. 2005; Dolan 1965; Groves 1969). They are the smallest Bovini, but the largest endemic mammals of Sulawesi, and are also endangered (Burton et al. 2005).
PNM 2006-A, an associated partial skeleton including left and right humeri, left metatarsal, 2 thoracic vertebrae (?T9 and ?T11), 2 unguals, and left m1–2. More of the skeleton (including several ribs and additional teeth and vertebrae) was present when the specimen was discovered (M. Armas, pers. comm.), but these were given away or lost subsequently; the aforementioned elements were the only ones available to us for study. The individual elements are partially permineralized (i.e., “fossilized”) but the degree of permineralization varies both within and between elements. The original specimen will be deposited at the National Museum of the Philippines; casts of the bones will be retained in the geology collections of The Field Museum (FMNH PM 61097) and a duplicate set of casts will retained in the Recent mammal collections.
The specimen was collected from the end of a horizontal tunnel in soft karst at approximately 50 m elevation in K-Hill near Balamban, Cebu Island, Philippines (approximately 10°51′N, 123°72′E; Fig. 2). No stratigraphic data were recorded when the specimen was collected.
Bubalus cebuensis differs from all previously described Bubalus in the size and proportions of the humerus and metatarsal (and presumably the rest of the postcranial skeleton). Based on linear dimensions, B. cebuensis is less than two-thirds the size of modern B. bubalis and is about 80% the size of modern B. mindorensis. B. cebuensis is similar in size to members of the subgenus Anoa (within approx. 10% for limb element lengths), but its skeleton is much more robust. The humerus, metatarsal, and vertebrae differ noticeably from those of B. (Anoa) and display a mixture of characters shared with B. bubalis, B. mindorensis, or both (e.g., roughly spherical humeral head, large greater tuberosity on humerus, low and broad deltopectoral crest, elongate metatarsal relative to humerus, straight metatarsal in medial view, small posterior cubonavicular facet on articular surface of metatarsal, T9–11 spinous processes straight in lateral view, and mammillary processes absent on T9–11; Table 2). The relative width of the dorsal longitudinal sulcus on the metatarsal is greater than that exhibited by any modern Bubalus and may be an autapomorphy for the new species. The unusual configuration of the proximal end of the metatarsal (i.e., with a broad, triangular anterior cubonavicular facet on the articular surface and a sulcus that bisects a small tuberosity on the proximolateral surface) may also be autapomorphic for B. cebuensis.
Age and distribution
The specimen is likely Pleistocene or Holocene in age, based on the geology of the region, but we were unable to obtain an absolute date because of a lack of preserved collagen (i.e., carbon dating was unsuccessful; University of Arizona, Tucson, Accelerator Mass Spectrometry Laboratory sample AA57785). The species is known only from the type locality.
After Cebu Island, the type locality and only known locality for the species. The specific epithet is analogous to that of the tamaraw, B. mindorensis, named in reference to Mindoro Island, Philippines, and reinforces the insular nature of this endemic species.
The humerus (Fig. 3) is very similar in length to that of B. depressicornis but is much more robust, as evidenced by all other humeral measurements (Table 3). The head of the humerus is roughly spherical in shape, similar to the condition in B. mindorensis and B. bubalis; in B. depressicornis the proximal surface of the humeral head is flattened, and the transition between the proximal and caudal surfaces is more abrupt, approximating a right angle. As in other Bubalus, the greater tubercle in B. cebuensis projects far above the humeral head and arches medially over the intertubercular groove, creating a tunnel. The greater tubercle is much larger and more developed in B. cebuensis, B. mindorensis, and B. bubalis compared to B. depressicornis. A well-defined pit is present in the tubercular fossa (i.e., the area anterior to the humeral head and medial to the base of the greater tubercle); this pit is absent in B. depressicornis and is present but less defined in B. mindorensis and B. bubalis. The greater tubercle extends inferiorly beyond the articular surface of the humeral head in B. cebuensis and B. bubalis; this contrasts with the condition in B. mindorensis, in which the greater tubercle ends at approximately the same level as the articular surface. A rugose deltopectoral crest is present in B. cebuensis, but the deltoid tuberosity is low and broad, strongly differing from the conditions observed in B. depressicornis (in which it is a thin, platelike process) and B. mindorensis (in which it is a high ridge). It resembles the condition in B. bubalis, but is less developed. The scar marking the insertion of the teres minor muscle is large and broad in B. cebuensis, again more closely resembling the condition in B. bubalis than in either B. depressicornis or B. mindorensis (in which it is narrow). B. cebuensis differs from both B. bubalis and B. mindorensis in having a smaller lesser tubercle and in having a less-pronounced medial ridge on the lesser tubercle for insertion of the subscapularis muscle.
To investigate size and shape variation of the humerus in Bubalus, a principal components analysis was performed using the raw data for the 9 measurements presented in Table 3. The factor loadings for the first 2 principal components (PCs) are presented in Table 4 and the specimens are plotted against these 2 factors in Fig. 4. As expected, PC1 clearly represents size (all variables have high positive loadings), and B. cebuensis plots closer to B. depressicornis on this axis than to any other species. PC2 represents humeral robustness; midshaft and distal anteroposterior diameters have high positive loadings, and length has a high negative loading. This axis clearly distinguishes the robust humeri of B. cebuensis and B. bubalis from the more gracile humerus of B. mindorensis, whereas the humeral morphology of B. depressicornis is intermediate between these 2 groups.
In contrast to the humerus, the metatarsal of B. cebuensis is more similar in length to that of B. mindorensis than B. depressicornis (Fig. 5; Table 3). As in both B. mindorensis and B. bubalis, it is straight in medial view; in B. depressicornis, this bone is slightly bowed (concave dorsally) and is much more gracile. The most conspicuous feature of the metatarsal in dorsal view, the dorsal longitudinal sulcus (i.e., the sulcus for the extensor digitorum longus tendon), is quite broad in B. cebuensis. It is absolutely wider than that of the larger B. mindorensis and B. depressicornis, and is equal in width to that of the much larger B. bubalis (and thus proportionally is much wider). The sulcus also differs from that of B. mindorensis and B. depressicornis in lacking overarching sides, although this may be an artifact of preservation.
A triangular process projects cranially from the posterior edge of the tarsal articular surface of the metatarsal in B. cebuensis. Although it is incompletely preserved, it is longer than that of B. depressicornis and is more similar in form to that of B. mindorensis and B. bubalis. In extant Bubalus, this process supports various facets for articulation with the posterior portions of the cubonavicular (= centroquartal) and fused tarsals II and III. In B. depressicornis and B. bubalis, the posterior cubonavicular facet is mediolaterally elongate; in B. mindorensis, it is small and circular. The size and position of this facet cannot be discerned precisely in the fossil cast, but it appears it would have more closely resembled the condition in B. mindorensis than in B. bubalis or B. depressicornis. The remaining tarsal facets are much broader in B. bubalis, B. mindorensis, and B. cebuensis than in B. depressicornis. In B. cebuensis, the anterior cubonavicular facet tapers slightly anteriorly; in both B. mindorensis and B. bubalis it is more constant in breadth. The facet for tarsals II and III is relatively larger in B. bubalis than in B. mindorensis; the condition in B. cebuensis more closely resembles that in B. mindorensis.
The tuberosity on the proximomedial end of the planter surface of the metatarsal is proportionately broader in B. cebuensis than in B. depressicornis, B. mindorensis, and B. bubalis. This tuberosity is proximodistally elongate in both B. mindorensis and B. bubalis, but extends further proximally in the former. In B. cebuensis, the tuberosity is broad proximally, resembling a triangle with its apex pointed distally. In both B. mindorensis and B. bubalis, a sulcus delimits the medial edge of the tuberosity from the plantar surface; no sulcus is present in B. depressicornis. In B. cebuensis, the feature that appears to be the homologous sulcus is positioned more laterally, essentially bisecting the tuberosity; this condition strongly contrasts with that present in all extant species of Bubalus.
In B. cebuensis, a slight crest extends distally from the proximomedial tuberosity of the metatarsal along its plantar surface; it is similarly developed in B. mindorensis. This crest is more strongly developed in B. bubalis, and is essentially absent in B. depressicornis. The heads of the metatarsal of B. cebuensis are reminiscent of those of modern Bubalus.
A principal component analysis was performed to explore size and shape variation in the metatarsal among different Bubalus. Seven of the measurements presented in Table 3 were used in this analysis; breadth of the dorsal longitudinal groove was excluded because the condition in B. cebuensis differs so dramatically from that of other Bubalus. The factor loadings for PC1 and PC2 are presented in Table 4 and the specimens are plotted against these 2 factors in Fig. 4. Again, PC1 clearly represents size (all variables have high positive loadings), but in contrast to the humeral results, B. cebuensis is positioned between B. depressicornis and B. mindorensis on this axis, reflecting the relatively larger size of the metatarsal in the new species compared to the humerus. PC2 represents metatarsal robustness to some degree—midshaft mediolateral diameter has a high positive loading and length has a high negative loading—but proximal and midshaft anteroposterior diameters also exhibit somewhat higher negative loadings. There is less discrimination along this axis than in the same axis in the principal component analysis of humeral measurements, reflecting a greater amount of individual or ontogenetic shape variation, or both, in the Bubalus metatarsal than in the humerus. Both B. mindorensis and B. bubalis vary greatly in their scores on PC2, but show surprisingly little overlap; B. cebuensis and B. depressicornis are within the range of variation of B. bubalis, distinct from B. mindorensis.
Two thoracic vertebrae of B. cebuensis are preserved but they differ in several respects from thoracic vertebrae of modern Bubalus (Fig. 6); because of this, their precise positions within the thoracic series are uncertain. Both preserved vertebrae exhibit relatively small costal articular facets (both on the vertebral body and transverse process), which suggests they pertain to the caudal half of the thoracic region (i.e., T7–13). They differ from vertebrae in this region of B. depressicornis in being larger, more robust, and lacking well-developed mammillary processes on the dorsal surfaces of the transverse processes. Both vertebrae exhibit lateral vertebral foramina, a feature present in most Bubalus vertebra.
The more complete of the 2 vertebrae (Fig. 6A) is likely the more anterior; its spinous process is less vertical (positioned at an angle of just under 45° to the horizontal), its transverse costal facets are slightly larger and directed more cranially, and its vertebral body is longer and narrower. It compares reasonably well with T9 of B. mindorensis in having a spinous process that is long and straight and mediolaterally expanded at its distal end (although the epiphysis is not preserved in the fossil vertebra). The caudal articular facets are positioned at approximately the same angle as the spinous process (45°) and are elliptical. The vertebra differs from the T9 of B. mindorensis, however, in having the transverse costal facets oriented more laterally (as opposed to cranially). In this respect, it is more similar to T10 of B. mindorensis. The vertebra differs dramatically from both T9 and T10 of B. depressicornis; in those vertebrae, the spinous processes arch caudally (i.e., they are concave caudally) and the caudal costal facet is positioned more dorsally on the vertebral body than the cranial costal facet (they are at the same level in B. cebuensis). In anterior view, the neural arch more closely resembles that of B. bubalis than B. mindorensis; the transverse processes do not extend as far superiorly as in B. mindorensis. The vertebra is intermediate in length between that of B. depressicornis and B. mindorensis and the vertebral body is wider than tall, more closely resembling the condition in B. mindorensis than in either B. depressicornis or B. bubalis (Table 5).
The less-complete vertebra (Fig. 6B) appears to represent T11. It possesses caudal articular facets that are directed inferiorly from the base of the spinous process, precluding the possibility of its referral to T12 or T13 (both of which have facets that are oriented more vertically and laterally, as in the lumbar vertebrae). The vertebra also exhibits a more vertical spinous process than typical T12 and T13 of modern Bubalus. It cannot represent T10, however, if the identification of the more complete vertebra described above is correct; when placed in approximate articulation, the 2 vertebrae do not appear to represent consecutive positions. The vertebra is more similar to T11 of B. mindorensis than of B. depressicornis, although the transverse processes are less well developed than in B. mindorensis. The vertebral body is very robust; it is intermediate in length between that of B. depressicornis and B. mindorensis, but is similar in breadth to the latter, making it relatively broader than in any modern material of Bubalus examined (Table 5). The within-species variability in this character is not known. The relative height of the vertebral body approximates that of B. bubalis.
The morphology of the unguals of B. cebuensis is similar to that of other Bubalus (Fig. 5D). The most conspicuous feature is a large, anteroposteriorly elongate depression on the posterolateral surface, an area that normally exhibits 1 or more large nutrient foramina in extant taxa. In B. cebuensis, it appears that several of these nutrient foramina might have coalesced. Whether this is a constant, discriminating feature of the taxon is unknown. This more closely resembles the condition seen in some individuals of B. bubalis than of B. mindorensis (no comparative specimens of B. depressicornis were available).
It is unclear whether these elements represent manual or pedal unguals. In B. mindorensis, the manual unguals appear to circumscribe a larger arc than the pedal ones. Additionally, the medial sides of the plantar surfaces are flatter in manual unguals; they are slightly upturned in the pedal unguals. The B. cebuensis unguals display an intermediate morphology between the anterior and posterior, precluding a confident identification of homology and suggesting a slightly different relationship between the ungual and pedal phalanges in this extinct taxon. The left ungual measures 50 mm (anteroposterior) × 25 mm (mediolateral); the right measures 49.5 mm (anteroposterior) × 23 mm (mediolateral).
The only dental elements preserved are 2 isolated lower molars, left m1–2 (Fig. 7; Table 6). In overall form, they closely resemble the corresponding teeth of modern Bubalus. In size, they are most similar to B. mindorensis; they are slightly smaller mesiodistally but are comparable buccolingually (Fig. 8). Both teeth exhibit moderate wear, suggestive of a mature individual; only the talonid fossette is present in m1, whereas in both trigonid and talonid fossettes are present in m2. They closely resemble the state of wear exhibited by UMMZ 157862 (B. bubalis).
The enamel is thick relative to tooth size in both m1 and m2, but this may be a function of wear stage. In m1, the crown height is 14.8 mm labially and 17.2 mm lingually. The corresponding values for m2 are 12.0 mm and 12.8 mm. A well-developed labial projection is present between the trigonid and talonid in both teeth. The presence or absence of this feature appears to vary individually and with wear in extant Bubalus. The enamel islands approximate a figure eight in B. cebuensis; enamel island shape also appears to vary with wear state in extant Bubalus, and a larger sample size of both B. cebuensis and other Bubalus would be required to test the taxonomic significance of this character.
One other long bone (lacking epiphyses) originally was found with the holotype, but we were unable to examine it firsthand. Based on study of a plaster cast preserving little detail, it most likely represents a ?right radius, potentially referable to B. cebuensis. It certainly represents an individual distinct from that of the holotype, however, in which all appendicular epiphyses are solidly fused (i.e., any sutures have been obliterated). The element measures 136 mm in length; the proximal end measures 47.5 × 24.5 mm and the distal end measures 34.5 × 28 mm.
Only a single cladistic analysis has examined relationships within Bovini. Geraads (1992) performed a phylogenetic analysis of 32 fossil and extant taxa based on 57 morphological characters. The monophyly of Asiatic buffaloes (the traditionally recognized Bubalina clade) was not supported in the consensus tree, but Geraads (1992) favored a slightly longer tree that included a monophyletic Bubalina (including Bubalus, †Hemibos, Anoa, and †Proamphibos). All but one of the characters used in this analysis were craniodental, however, precluding testing of the phylogenetic position of B. cebuensis using this data set. Xue and Li (2000) examined relationships among Chinese fossil Bubalus, but that analysis also was based exclusively on craniodental (especially horn core) characters. Rautian et al. (2000) examined morphological and molecular differentiation within the Bovini (including a few fossil species), but did not perform a phylogenetic analysis nor include any extinct genera. Among extant taxa, Rautian et al. (2000) advocated a common ancestry of Bubalus and Anoa exclusive of other taxa. The same association was favored by Geraads (1992) among extant Bovini.
In sum, the few studies that have examined relationships among fossil Bovini have relied almost exclusively on craniodental characters and thus cannot be used to directly test the phylogenetic affinities of B. cebuensis. The lack of postcranial data in those analyses is due to the paucity of associated postcrania for fossil taxa. Although a comprehensive revision of fossil bovine postcrania would likely help identify postcranial characters useful for phylogenetic analysis, such an undertaking is well beyond the scope of the present report of this single new endemic island taxon.
Among extant Bubalus for which postcranial data are available, B. cebuensis much more closely resembles B. (Bubalus) than B. (Anoa); B. cebuensis shares a variety of discrete morphological character states with both B. bubalis and B. mindorensis, but shares none with B. depressicornis (Table 2). Given the greater similarity in size between B. cebuensis and B. (Anoa) than between the former and B. (Bubalus), the character states common to B. cebuensis and B. (Bubalus) cannot be attributed to allometry and are more likely indicative of phylogenetic relationship. Based on these resemblances, B. cebuensis can reasonably be referred to B. (Bubalus). Within B. (Bubalus), B. cebuensis shares more character states with B. bubalis than with B. mindorensis. The latter 2 species share several traits not observed in B. cebuensis, however, which may indicate a closer relationship between them than between either and B. cebuensis. The relationships among these 3 species are probably best considered unresolved until additional material of B. cebuensis is discovered and a thorough cladistic analysis (incorporating postcranial features, as well as living and fossil taxa) can be performed.
Long-bone lengths for B. cebuensis suggest an animal roughly similar in stature (i.e., limb length) to B. depressicornis; in contrast, the breadths of these bones suggest a more massive animal. To test this assertion, we estimated the body mass of B. cebuensis using postcranial regression equations for modern bovids published by Scott (1983). Eleven humeral and metatarsal measurements examined by Scott (1983) were preserved in the available material of B. cebuensis and 10 of these were used to estimate body mass; metatarsal length was excluded because it is poorly correlated with body mass in modern bovids (Scott 1983). The remaining 10 variables produced mass estimates of approximately 115–215 kg with X̄ = 157.2 kg and SD = 36.6 kg (Table 7). Humeral mass estimates exhibited a much greater range than metatarsal mass estimates, but the average values were quite similar (154.0 kg and 162.0 kg, respectively). A reasonable mass estimate for this particular specimen of B. cebuensis is roughly 150–165 kg. Given the variability in estimates derived from individual measurements in living taxa, this estimate is probably within 15% of the true mass of this fossil animal (Scott 1983).
As discussed by Scott (1983), accurate individual or species body masses are difficult to obtain for extant large herbivores; this type of information is rarely recorded in the field, and body masses listed for these mammals frequently include only trophy or zoo animals (which are not representative of the species in general). In the case of rare ungulates such as B. mindorensis and B. (Anoa), even poor mass estimates are comparably scarce. Only 2 independent body masses have been published for B. mindorensis: Talbot and Talbot (1966) estimated the mass of a female zoo animal at 600 lb (= 275 kg) and Roth and Montemayor-Taca (1971) estimated the mass of a female zoo animal at 180–220 kg. Although sexual dimorphism is evident in the skull of B. mindorensis, body mass and limb proportions do not vary with sex (Custodio et al. 1996), and so these estimates also should apply to male B. mindorensis. Using the regression equations of Scott (1983), an average body mass of approximately 210 kg was obtained for the 2 male specimens examined in the present study; this accords well with the estimate of Roth and Montemayor-Taca (1971) and highlights the lack of sexual dimorphism in this species.
Groves (1969) reported a body mass of 56 kg for an adult B. quarlesi and indicated it was the only body mass for B. (Anoa) that he was aware of; this same mass was used by Scott (1983) for B. depressicornis, presumably because no better data were available. Burton et al. (2005) estimated the body mass of B. quarlesi as <150 kg, but provided no data for individual specimens; they estimated the body mass of B. depressicornis as <300 kg, but noted that no specimen has ever been recorded at more than 150 kg. Based on skull length (Groves 1969), there does not appear to be dramatic sexual size dimorphism in B. (Anoa). Using the regression equations of Scott (1983), a body mass of approximately 135 kg was obtained for the single female specimen of B. (A.) depressicornis examined in the present study. Therefore, this suggests that a mass estimate of <150 kg for B. depressicornis may be more accurate than an estimate of <300 kg.
The data considered above indicate that B. cebuensis was intermediate in body mass between B. depressicornis and B. mindorensis, being about 15% larger than the former and 25% smaller than the latter. This contrasts with long-bone lengths, which suggest a stature more similar to that of B. depressicornis than B. mindorensis. The prominent muscle scars on the humerus and the breadth of the dorsal longitudinal sulcus of the metatarsal support the interpretation of relatively large appendicular muscles in B. cebuensis, indicative of its larger body mass (relative to B. depressicornis). Together, these observations paint a picture of a short, heavy-bodied Bubalus, perhaps similar in stature to an anoa, but certainly of greater mass; the structure of the skull is unknown, but the animal may have resembled a small tamaraw.
The phenomenon of insular “dwarfing” has been observed in many groups of large mammals, including proboscideans (Hooijer 1970; Roth 1990; Vartanyan et al. 1993), hippopotamids (Burney et al. 2004; Simmons 1988), cervids (Lister 1989), and possibly hominids (Brown et al. 2004; Morwood et al. 2005). Together with its converse—insular gigantism—insular dwarfing has been the subject of many studies over the past 40 years (e.g., Anderson and Handley 2002; Case 1978; Foster 1964; Heaney 1978; Lawlor 1982; Lomolino 1985, 2005; Meiri et al. 2004; Melton 1982; Michaux et al. 2002; Sondaar 1977; Van Valen 1973). Although the causes (and patterns) of insular body-size change are still being debated, certain cases continue to be exemplars of how a major life-history trait can evolve rapidly, over a geologically brief period of time, after isolation (e.g., Lister 1989).
Given the much larger size (in terms of both stature and mass) of the closest relatives of B. cebuensis, both fossil and extant, it is clear that this Cebu Island form represents another case of insular dwarfing. In fact, the long fossil record of Bubalus in Asia suggests that both B. mindorensis and B. cebuensis may be dwarf forms of B. bubalis that arose by dispersal to and within the Philippines. Given the rarity of fossils in the Philippines and the lack of a secure phylogeny for the species of B. (Bubalus), details of such a scenario must remain provisional. However, certain independent aspects of the distribution and morphology of these taxa do support this general interpretation.
Heaney (1978) examined body-size variation in the tricolored squirrel, Callosciurus prevostii, and found a significant correlation between body size and island size. Based on this correlation and patterns of variation in other mammals, he constructed a model predicting that the effects of food limitation, predation, and interspecific competition on body size would vary depending on the body size of the species in question and the area of the island. In some cases, these factors would be expected to produce insular dwarfs; in others, giants would result. For large mammals, food limitation was proposed as the most significant factor affecting insular body size, and a direct correlation between body size and island size was the expected result. In support of this model, Heaney (1978) noted such a correlation in Pleistocene populations of Elephas falconeri, where increasingly smaller-bodied taxa are found on progressively smaller Mediterranean islands (Maglio 1973). Although this pattern may not apply to carnivorans (Heaney 1978; Meiri et al. 2004) or tree sloths (Anderson and Handley 2002) it does seem to be generally applicable to large ungulates, including Bubalus.
Based on body mass, B. cebuensis is some 25% smaller than B. mindorensis; B. mindorensis is itself less than half the size of wild (mainland) B. bubalis. Such a pattern in body size would be predicted based simply on the relative sizes of Cebu Island (5,088 km2), Mindoro Island (10,243 km2), and mainland Asia, if island size were a major factor in body-size evolution. This pattern does not hold if one substitutes the entire area of the late Pleistocene Negros–Panay Faunal Region for that of Cebu Island—a region probably 10 times larger—but this is a moot point if B. cebuensis evolved during the Holocene; given the numerous other examples of post-Pleistocene dwarfing in large mammals, B. cebuensis certainly could have undergone the observed changes in body size subsequent to Cebu's isolation. No well-preserved Bubalus have been discovered on other islands that might permit testing of this scenario; although isolated teeth from Luzon have been referred to B. mindorensis (Beyer 1957), they have not been critically examined (Custodio et al. 1996), and probably are not useful for estimating body size in bovids except in very general terms (Scott 1983; see also below).
Size reduction in insular dwarfs is not generally isometric across all parts of the body; allometric relationships have been noted both for the dentition (Fortelius 1985; Lister 1989) and distal limb elements (Köhler and Moyà-Solà 2001; Sondaar 1977), among other structures. In dentition, island dwarfs tend to have relatively larger teeth, an attribute apparently exhibited by B. cebuensis; although much less massive than B. mindorensis, the teeth of these 2 species are similar in size (Fig. 8). Based on this pattern of positive dental allometry, it has been suggested that paedomorphosis is the mechanism of insular dwarfing in some lineages (Fortelius 1985).
In contrast to the dentition, the metatarsal of B. cebuensis does not exhibit the pattern of size reduction typical of island dwarfs; it is 71.5% the length of the humerus in B. cebuensis, a value comparable to that of B. bubalis examined (68–72.5%). However, the metatarsal does appear to be reduced in B. mindorensis, being only 59–60% the length of the humerus. Whether such reductions in distal limb elements are attributable to paedomorphosis or simply are locomotor adaptations (e.g., Sondaar 1977) may depend on the mosaic pattern of insular evolution in the particular species in question; the metatarsal of B. mindorensis more closely resembles an osteologically mature B. bubalis than a slightly less mature one (Fig. 4), but testing for such an ontogenetically driven trend would certainly require examining much younger individuals of B. bubalis.
The new taxon described above is the 1st fossil mammal of any age to be reported from Cebu Island and the only nonproboscidean to be documented from the Negros–Panay Faunal Region. In conjunction with the presence of Bubalus on Mindoro Island (and potentially Luzon), discovery of this specimen suggests that Bubalus may once have ranged throughout the Philippines—a hypothesis that hopefully will be tested by future paleontological sampling across the islands. Similarly, this discovery, combined with the exceptionally high endemism and diversity of the extant mammal fauna of the Philippines, suggest that many more new late Cenozoic to Holocene fossil species remain to be recovered from this region. B. cebuensis is clearly referable to the subgenus Bubalus, and is diagnostically differentiated from B. mindorensis and B. bubalis by metric and morphological characteristics. In life, B. cebuensis was probably similar in stature to the lowland anoa, B. depressicornis, but regression equations from modern bovids suggest that it was approximately 15% more massive. The overall small size of B. cebuensis relative to other B. (Bubalus) appears to be attributable to island dwarfing, an explanation that is supported by the consistent relationship between body size and island size in B. cebuensis, B. mindorensis, and B. bubalis. Although the relatively large dentition of B. cebuensis suggests paedomorphosis as a possible mechanism of body-size reduction in this lineage, this is not supported by the relatively large (i.e., normal for this clade) size of the metatarsal relative to the humerus. Additional material of B. cebuensis would facilitate testing of paedomorphosis as a mechanism for dwarfing in these bovids, as would a detailed study of the postcranial osteology of B. mindorensis. Although the exact age of B. cebuensis is unknown, the condition of the material and its small size suggest it is no older than Pleistocene, and possibly Holocene. Further study of B. cebuensis and other island dwarfs may provide insights into the evolution of small-bodied hominins such as Homo floresiensis, yet another reason to be interested in dwarf mammals on oceanic islands in Indo-Australia.
We thank M. Armas and H. Intengan for their generosity and sense of curiosity in recognizing the scientific value of the specimen. At The Field Museum of Natural History, S. McCarroll, W. Simpson, and W. Stanley gave valuable assistance with the preliminary identification and accessioning of the specimen; J. Weinstein and M. Widhalm skillfully executed the photographs; and A. Shinya graciously made the molds and casts. P. Myers and S. Hinshaw provided access to specimens at the University of Michigan Museum of Zoology. Carbon isotope analyses were conducted by the Accelerator Mass Spectrometry Laboratory at the University of Arizona. We thank 2 anonymous reviewers for providing constructive assessments of this manuscript.
List of specimens examined from the Recent mammal collections of the Division of Mammalogy at The Field Museum of Natural History (FMNH) and the University of Michigan Museum of Zoology (UMMZ).
Bubalus (Anoa) depressicornis
FMNH 98791, mature female, skull and skeleton, from Lincoln Park Zoo, Chicago, Illinois.
Bubalus (Bubalus) mindorensis
FMNH 18817, mature male, skull and skeleton, from Mindoro Island, Philippines; FMNH 43300, mature male, skull only, Mindoro Island, Philipines; FMNH 43301, young male, skull only; UMMZ 84106 (holotype of Anoa mindorensis Steere), mature male, skeleton only, from Catuiran River, Mindoro Island, Philippines.
Table 1.—Extinct and extant Philippine ungulates listed by faunal region (Bautista 1991; Heaney et al. 1998). Animals obviously brought to the Philippines by the Spaniards (e.g., Equus) are not included
Table 2.—Variation in selected morphological characters among species of Bubalus. Character states shared between B. cebuensis and other species are underlined; character states that are autapomorphic for B. cebuensis are italicized. T9–11 = 9th through 11th thoracic vertebrae
Table 3.—Humeral and metatarsal measurements (mm) for selected specimens of Bubalus. Measurements at proximal and distal ends are greatest diameters. Humeral trochlear diameter was measured in a proximodistal orientation. FMNH = The Field Museum of Natural History; PNM = National Museum of the Philippines; UMMZ = University of Michigan Museum of Zoology; AP = anteroposterior; ML = mediolateral
Table 4.—Factor loadings and proportion of variation for first 2 principal components (PCs) from principal component analyses of Bubalus humeral (H) and metatarsal (MT) measurements. PC1 accounts for 96.8% of the variation in the data; PC2 accounts for 2.3%. AP = anteroposterior; ML = mediolateral; NA = not available
Table 5.—Measurements of 9th (T9) and 11th (T11) thoracic vertebrae for selected specimens of Bubalus. FMNH = The Field Museum of Natural History; PNM = National Museum of the Philippines; UMMZ = University of Michigan Museum of Zoology; AP = anteroposterior
Table 6.—First (m1) and 2nd (m2) lower molar measurements for selected specimens of Bubalus. FMNH = The Field Museum of Natural History; PNM = National Museum of the Philippines; UMMZ = University of Michigan Museum of Zoology; L = length (measured mesiodistally); W = width (measured buccolingually)
Table 7.—Humeral (H) and metatarsal (MT) variables used to predict body mass of Bubalus cebuensis using measurements from the holotype and predictive equations from Scott (1983). Predictive equations are in the form: log(body mass) = b × log(variable) + a. The mean of the predicted body masses for B. cebuensis is 151.4 kg and SD = 40.0. For more detailed explanations of variables, see Scott (1983). AP = anteroposterior; ML = mediolateral