Phylogenetic systematics of the Rana signata

Transcription

1 Biological Journal of the Linnean Society, 2002, 76, With 13 figures Phylogenetic systematics of the Rana signata complex of Philippine and Bornean stream frogs: reconsideration of Huxley s modification of Wallace s Line at the Oriental Australian faunal zone interface RAFE M. BROWN* and SHELDON I. GUTTMAN Zoology Department, Miami University, Oxford, Ohio USA Received 3 August 2001; accepted for publication 22 February the regional boundaries drawn in the area between Sundaland and Sahulland vary greatly according to the taxonomic group studied and... no real realistic single boundary for the faunas as a whole has been or probably can be established. In this connection, it is noteworthy that most of the studies and the boundary proposals have been based on flying animals, usually birds, bats, and insects... Simpson (1977, p. 116). It is worthy of notice that its staunchest defenders were those naturalists who actually studied and collected animal life on both sides of the line, like Dickerson and his associates in the Philippines... Mayr (1944, p. 4). The complexity, level, and width of the transition of the Oriental and Australian vertebrates vary with the different classes probable powers of crossing salt water barriers. Darlington (1957, p. 469). We addressed the evolutionary relationships and biogeographical patterns of a model organism of low relative dispersal ability by electrophoretically assaying the products of 42 presumptive gene loci in Philippine and Bornean members of the Rana signata complex of SE Asian stream frogs. Utilizing three distantly related species of ranid frogs to deeply root trees consisting of five more closely-related species and six in-group species of the Rana signata complex, we conducted phylogenetic analyses that produced concordant topologies, regardless of the data coding strategy employed. All analyses support the hypothesis of monophyly for the Rana signata complex on the whole, but none provides support for the monophyly of its Philippine members. Our analyses of morphometric and allozyme data (along with biogeographical information) indicate that (1) most previously-recognized Philippine and Bornean subspecies of the Rana signata complex should be recognized as full species in appreciation of their status as independent evolutionary lineages; (2) Rana picturata Boulenger (until very recently included in the synonymy of Rana signata signata) is deserving of specific rank; (3) the Mindoro Isl. (Philippine) population, previously confused with Rana signata similis of Luzon Isl. is a new species; (4) two major clades (((R. signata (R. grandocula + R. similis)) + (R. picturata (R. mangyanum + R. moellendorffi))) of Bornean + Philippine lineages are recognized, corresponding to two separate faunal exchanges between the Philippines and the edge of the Sunda Shelf; (5) invasions of the oceanic portions of the Philippine islands from the Sunda Shelf have occurred along both the eastern (Sulus Mindanao Samar Leyte Luzon) arc and the western (Palawan Busuanga Mindoro) island arcs; (6) northern reaches of Wallace s Line (as modified by Huxley) include exceptions to an otherwise discrete faunal separation. These results suggest the need for revision of this biogeographical barrier, increased recognition of temporal patterns of island connectedness and geographical proximity, and/or a greater appreciation of dispersal abilities of ranid frogs.. ADDITIONAL KEYWORDS: Chalcorana dispersal evolutionary radiations Hylarana islands new species Pulchrana Ranidae Sunda Shelf vicariance. *Corresponding author. Present address: Section of Integrative Biology (C0930) and Texas Memorial Museum, University of Texas, Austin, TX , USA. rafe@mail.utexas.edu 393

2 394 R. M. BROWN and S. I. GUTTMAN INTRODUCTION Situated between the Oriental and Australian continents, the Indo-Australian archipelago (Fig. 1) is known for being the most geologically complex group of islands in the world (Willis, 1937; Allen, 1962; Morley & Flenley, 1987; Hall, 1996, 1998; Hamilton, 1979; Audley-Charles, 1981). This region also supports the highest degree of biological endemism ever recorded (Wallace, 1896; Barbour, 1912; Mayr, 1944; Darlington, 1957; Carlquist, 1965; Whitmore, 1975, 1987; Brooks et al., 1997). Wallace s Line, the celebrated zoogeographical boundary that bisects the archipelago (Fig. 1; Wallace, 1860; Huxley, 1868; Darlington, 1957; Calaby, 1972; Quammen, 1996), is universally recognized as the sharpest faunal demarcation on the planet and the western edge of the interface between the Oriental and Australian faunal zones (see papers in Whitmore, 1987; Keast & Miller, 1996; Hall & Holloway, 1998). Treatments of this region have stressed that the adherence of various faunal groups to Wallace s Line is based on the ecological attributes and relative dispersal abilities of each group (Carlquist, 1965; Holloway & Jardine, 1968; Thorne, 1972; Keast, 1983; Heaney, 1986; Bremer, 1987; Dransfield, 1987; summaries in Whitmore, 1987). Virtually all empirical con km Sumatra Mainland Asia 100? 47 Java 110 Huxley Borneo siderations of Wallace s line and Huxley s line (Fig. 1) have utilized organisms with moderate to high dispersal abilities (e.g. birds, lizards, snakes, mammals, flying insects, coconut palms). For example, in many taxa, patterns of relatedness (Peterson & Heaney, 1993; Heaney & Ruedi, 1994), divergence and limitations to gene flow (Schmidt, Kitchener & How, 1995) can often be explained by insular geography and deep water channels such as those that give rise to Wallace s Line (see also Holloway & Jardine, 1968; Whitmore, 1975; Dransfield, 1981, 1987; Ollier, 1985). Still, vicariance hypotheses of phylogenetic discontinuity that evoke limits of dispersal afforded by the Sunda Shelf should include organisms of low relative dispersal abilities as a measure of this barrier s true biological importance. Surprisingly few such studies exist. The northern portion of Wallace s Line has been the subject of debate. The original boundary as perceived by Wallace (1860) has been altered in a manner that more accurately reflects the edge of the Sunda Shelf. Wallace originally identified a sharp zoogeographical boundary between Bali and Lombok in the south, that extended northward, between Borneo and Sulawesi (Fig. 1). Huxley (1868), asserting that Palawan megapodes and pheasants were taxonomically more similar to those from Borneo than to species from the Sulawesi Wallace New Guinea 10 Australia Figure 1. The Indo-Australian region, bisected by Wallace s Line (and Huxley s modification), which divides the northern Palawan land-bridge island chains (Calamian and Cuyo groups) from the rest of the oceanic portions of the Philippines. Numbered localities refer to collection sites listed in Appendix 1; see Fig. 4 for collection localities in northern Borneo and within the Philippines (box).

3 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 395 rest of the Philippines, modified Wallace s Line such that the greater portion of the Philippines were excluded from the edge of the Sunda Shelf, and Palawan Island and the Calamian group (Busuanga, Coron, and Cuyo Islands, Philippines) were associated with mainland Asia (see also: Everett, 1889; Mocquard, 1890; Mayr, 1944; Darlington, 1957; Simpson, 1977; Widmann, 1998). Huxley s modification of Wallace s line thus split the Philippines into two faunal subregions, with the Palawan satellite islands to the west and Mindoro immediately to the east (Fig. 1). Today, Huxley s Line (= Merrill- Dickerson Line of Widmann, 1998) distinguishes continental or land bridge islands from oceanic or volcanic islands that arose de novo from the ocean floor and have not been connected to the Sunda Shelf during Pleistocene sea-level reductions (Fig. 2; Hall, 1996, 1998). In order to generate hypotheses of relationships among island endemics, Pleistocene aggregate island formation has been visualized (Kloss, 1929; Feliciano & Pelaez, 1940; Rutland & Walter, 1974; Hamilton, 1979; Heaney, 1985a, b, 1986; Voris, 2000) by tracing underwater bathymetric contours around the Sunda Palawan Less than 120 m submarine contour 200 km 15 Cavite Passage Mindoro Mindoro Straits 10 Busuanga Borneo Balabac Sulu Archipelago Luzon Panay Guimaras Negros Cebu Jolo Tawitawi Batanes islands Babuyan islands Polillo Marinduque Catanduanes San Bernardino Straits Basilan Mindanao Samar Leyte Bohol Camiguin Figure 2. The Philippine Islands with Pleistocene aggregate island complexes indicated by tracing of the 120-m submarine bathymetric contour (following Heaney, 1985a, b, 1986). Note channel names discussed in the text. Shelf and adjacent oceanic islands. This approach has revealed which islands may have been connected by land-positive aerial exposure during Pleistocene sealevel retreats. Islands separated by less than 120 m water depth are assumed to have been connected at the end of the Pleistocene, and so it is hypothesized that these assemblages should support fauna more closely related to one another than to distant oceanic landmasses not conjoined at the end of the last glacial maxima (Heaney, 1985b). Five to seven major geological and faunal subregions have been recognized (Fig. 2) within the Philippines, each corresponding to late Pleistocene aggregate islands composed of several smaller islands of today (Taylor, 1928; Leviton, 1963; Brown & Alcala, 1970, 1994; Heaney, 1985a, b, 1986). These approaches have also elucidated potential entryways into and among the oceanic portions of the Philippines by land bridges, or substantial narrowing of channels between islands (Dupree, 1954; Myers, 1960, 1962; Heaney, 1985b, 1986; Auffenberg, 1988; Brown & Alcala, 1994; R. Brown, 1997; W. Brown, 1997). Systematists and biogeographers focusing on questions of botanical (Merrill, 1923), avian (Dickinson, 1991), mammalian (Heaney, 1986, 1991; Heaney & Rickart, 1990; Peterson & Heaney, 1993; Groves, 1994; Heaney & Ruedi, 1994), and anuran distribution patterns in the Philippines (Boulenger, 1894; Inger, 1954; Brown & Alcala, 1970) have continued to treat Palawan as a faunal extension of northern Borneo, noting biotic similarities between the two islands and their apparent dry land connection during the mid-pleistocene (c yr before present; Everett, 1889; Heaney & Rickart, 1990; Heaney, 1991; Peterson & Heaney, 1993). In contrast, Mindoro has been regarded as an extension of southern Luzon because of faunal similarity between the two islands (Taylor, 1928; Leviton, 1963; Brown & Alcala, 1970; Dickinson, 1991; Kennedy et al., 2000). Over the past 40 years the prevailing taxonomy of Philippine Amphibia (Inger, 1954) has received little critical attention in zoogeographical summaries (e.g. Cranbrook, 1988; Dickinson, 1991; Inger, 1999). Until very recently (Brown & Alcala, 1994; Brown, Brown & Alacala, 1997; Brown, Alcala & Diesmos, 1997; Alcala & Brown, 1998; Brown et al., 2000b; Brown, McGuire & Diesmos, 2000; McGuire & Alcala, 2000; see Brown & Diesmos, in press, for discussion), no systematic revisions of Inger s (1954) polytypic taxonomy have been undertaken. Systematic treatment of any of the problematic Sunda Shelf elements on Mindoro has never been attempted, though their presence has been repeatedly noted (Inger, 1954; Brown & Alcala, 1955; Myers, 1960, 1962; Leviton, 1963; Brown & Alcala, 1970, 1978; R. Brown, 1997; W. Brown, 1997; Alcala & Brown, 1998). We chose members of the R. signata complex (Inger,

4 396 R. M. BROWN and S. I. GUTTMAN 1954) for several reasons. First, R. signata group frogs are an ideal choice for a study of Philippine biogeography because they occur on both Sunda Shelf land bridge islands (e.g. Palawan and its satellites), as well as the oceanic portions (Mindoro, Mindanao, Luzon and associated satellite islands) of the archipelago. Second, The Rana signata complex (Fig. 3) was selected because of the low relative dispersal ability assumed for these stream frogs (Inger, 1954; Myers, 1962; Carlquist, 1965). Finally, although higher-level relationships of ranids have not yet been resolved satisfactorily (see Ford & Cannatella, 1993), the available works do imply that the R. signata complex has a history outside of the Philippines (Inger, 1954, 1966; Taylor, 1962; Dubois, 1987, 1992; Inger & Tan, 1996a). That is, the putative taxa Hylarana, Chalcorana and Pulchrana (at times considered subgenera, genera or merely unsubstantiated hypotheses of affinity; Dubois, 1992; Inger, 1996) contain several morphologically cohesive groups of species that might prove to be monophyletic with pending phylogenetic analyses (Inger, 1966, 1984, 1996; Dubois, 1992; Matsui, 1994). The majority of these clades are African, mainland Asian, Indonesian and Bornean groups, some of which have invaded oceanic island archipelagos (Inger, 1966, 1984, 1999). The R. signata complex appears to be an example of this type of invasion. Its presumed closest relatives are either mainland Asian or Sunda Shelf species (Günther, 1872a, b; Boulenger, 1882; Inger, 1954, 1966; Dubois, 1987, 1992). Specifically, we ask: (1) Are the Philippine members of the R. signata complex a monophyletic assemblage or are they derived from multiple faunal exchanges with the edge of the Sunda Shelf (Borneo)? (2) Do the phylogenetic relationships in Philippine and Bornean stream frogs suggest order of island arc colonization or routes of dispersal into the Philippines? (3) Is there justification for maintaining the taxonomy of Inger (1954) under a lineage-based species concept? (4) And, finally, do relatively poor dispersers (for which these ranid stream frogs represent a suitable model) adhere to Huxley s modification of Wallace s Line and the largely unchallenged faunal division between northern Palawan and southern Mindoro Islands? TAXONOMY OF THE RANA SIGNATA COMPLEX Frogs of the Rana signata complex (Rana signata signata, R. s. similis, R. s. grandocula, R. s. moellendorffi, R. picturata, R. siberu; see Inger, 1954, 1966; Diagnosis of the Rana signata complex section, below) favour riparian habitats in montane and submontane regions throughout much of Borneo, Indonesia, peninsular Malaysia, Thailand and the Philippines (Günther, 1872a; Inger, 1954, 1966; Taylor, 1962; Inger & Stuebing, 1989; Inger & Tan, 1996b; Manthey & Grossmann, 1997). Rana signata (Fig. 3) was described by Günther (1872b) from Matang, Sarawak (Borneo). In the same year, he described R. similis from Laguna del Bay, Luzon Island, Philippines (Günther, 1872a), differentiating the two by the white labial stripe and smooth skin in R. similis. Boulenger (1882) listed R. similis in his catalogue but remarked that he considered it invalid as a species. Later, Boulenger (1920) placed R. similis and Mocquard s (1890) R. obsoleta (type locality: Kinabalu, Sabah) in the synonymy of R. signata. Boulenger s synonymy of R. obsoleta has remained unchallenged (see note added in proof, p. 443); however, Taylor (1920) regarded R. similis as a distinct Luzon form. The next member of the R. signata complex to be described was Rana moellendorffi Boettger 1893 from Palawan, Philippines. This species was placed in the synonymy of R. signata by Van Kampen (1923), who also submerged the Bornean species R. picturata (Fig. 3; Boulenger, 1920). Smith (1930, 1935) and Inger (1954, 1966) concurred with Van Kampen s action regarding R. picturata Boulenger 1920, but Inger (1954) retained the names moellendorffi and similis as island races of R. signata (see below) and restricted R. similis to Luzon and Leyte (the latter apparently in error; see Leviton, 1955; Gaulke, 1990). Prior to Inger s (1954) revision, Taylor (1920) considered the Sunda Shelf R. signata and the Philippine R. similis and R. moellendorffi all distinct species (see also Taylor, 1922a, 1962) and he named four more Philippine species: R. grandocula (Mindanao Island), R. yakani (Basilan Island), R. philippinensis (Mindanao) and R. melanomenta (Papahaug Island, Sulu Archipelago). Although the holotype of Taylor s R. melanomenta was destroyed by the firebombing of Manila during World War II (see Myers, 1960; W. Brown & Alcala, 1978; R. Brown, Ferner & Diesmos, 1997), the rest of Taylor s types were available for examination by Inger (1954), who concluded that R. yakani and R. philippinensis were synonyms of R. grandocula and that this, and all other Philippine forms, were indistinguishable from R. signata at the specific level. Inger s (1954, 1966) conclusions were based on nuptial pad shape, coloration and pattern, and body proportions; his assignment of these forms to the major island subgroups of the Philippines has stood uncontested since the publication of his 1954 monograph, save for Dubois (1992; see Duellman, 1993) listing of Inger s (1954) subspecies as full species without accompanying data or justification. Similarly, no data have yet been presented to substantiate the recent resurrection of R. picturata by Inger & Tan (1996b). Thus, despite Dubois (1992) and Duellman s (1993) listing of Inger s (1954) subspecies of R. signata as full species, many authors (e.g. Frost, 1985; Alviola et al., 1998) have considered Philippine populations to

5 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 397 Figure 3. Colour pattern variation in Philippine members of the Rana signata complex, including: (A) female R. signata from the state of Sabah, Borneo Isl., Malaysia; (photo: R. F. Inger); (B) male R. picturata from Ranchan Pool, the state of Sarawak, Borneo Isl., Malaysia (photo: L. L. Grismer); (C) female R. moellendorffi from the Municipality of Nara, Palawan Isl., Philippines; (D) female R. grandocula from the Municipality of Bilar, Bohol Isl., Philippines; (E) female R. mangyanum from the Municipality of Puerto Galera, Mindoro Isl., Philippines; (F) male R. mangyanum (holotype) from the Municipality of Puerto Galera, Mindoro Isl., Philippines; (G) dark phase male R. similis from the Municipality of Maria Aurora, Luzon Isl.; (H) light phase male R. similis from the Municipality of Irosin, Luzon Isl., Philippines (C H, photos: R. M. Brown).

6 398 R. M. BROWN and S. I. GUTTMAN be R. signata or continued to follow Inger (1954) and referred these populations to subspecies (i.e. Alcala, 1986; Dring, McCarthy & Whitten, 1989; R. Brown et al., 1996; Alcala et al., 1997; Alcala & W. Brown, 1998). Although Taylor (1920) compared his R. melanomenta to other members of the R. signata complex, Inger (1954) emphasized characters that suggest a relationship to R. glandulosa from Borneo. Until new material of R. melanomenta is available, we are unable to asses this taxon s status or phylogenetic affinities. We cannot comment further on this species beyond our tendency to tentatively consider it a valid species pending the collection of data to the contrary (see also Inger, 1954, 1999). More recently, Dring, McCarthy & Whitten (1989) described R. siberu, a remarkable new form presumably related to the R. signata complex and endemic to the Mentawai Islands, Indonesia. Recently, herpetologists conducting biodiversity surveys in the mountains of Sumatra have identified stream frogs presumably related to R. siberu (D. Iskandar, pers. comm.), although these identifications require confirmation. High degrees of colour pattern and body size variation in certain Bornean and peninsular Malaysian populations suggest that additional undescribed species may require recognition once additional data become available. MATERIAL AND METHODS SPECIES CONCEPT AND TAXONOMIC DECISIONS We agree with Wiens (1993) that taxonomic decisions regarding population status should be based on an explicit species concept (for discussion of recent species concepts, see symposia in: Otte & Endeler, 1989; Ereshefsky, 1992; Howard & Berlocher, 1998; Wilson, 1999). For this purpose, we adopt the General Lineage Concept (de Queiroz, 1998, 1999; see also Simpson, 1961; Wiley; 1978; Hull, 1980; Frost & Hillis, 1990) and consider a species a lineage segment of ancestor-descendant populations with a unique evolutionary history and predictable evolutionary future (fate). Recently, de Queiroz (1998, 1999) demonstrated that all modern species concepts are consistent with a simple unified principle of species as population-level lineage segments. de Queiroz (1998, 1999) suggested that species be described by the properties they acquire during the process of speciation (i.e. not by rigid, formulaic, predetermined criteria). We concur with this general conceptual framework and find it particularly appropriate in the context of the insular nature of populations in the Philippines (see discussions in Brown, McGuire & Diesmos, 2000; McGuire & Alcala, 2000; McGuire & Kiew, 2001; Brown & Diesmos, in press). Accordingly, taxa recognized herein possess fixed, diagnostic phenotypes of morphology, behaviour, and alleles and either exist sympatrically (with no evidence of intergradation) or allopatrically on separate Pleistocene aggregate island complexes, each separated by deep water (= 120 m channel depth) and with a known history of isolation. For the purposes of this analysis, we consider as distinct lineages populations that are (1) geographically isolated as insular endemics and morphologically and genetically distinct and (2) sympatric, reliably diagnosable populations for which the hypothesis of conspecificity confidently can be rejected by analyses of morphological, behavioural and genetic data (Frost & Hillis, 1990; Wiens, 1993). FIELD WORK Field work was conducted on the Philippine islands of Luzon, Polillo, Samar, Palawan, Mindanao, Bohol, Mindoro and Panay between 1992 and 1995 (Fig. 4 and Appendix 1). Frogs were collected from riparian habitats along elevational transects (Ruedas, Demboski & Sison, 1994; as modified by Brown, Ferner & Sison, 1995; Brown et al., 1996, 2000; Brown, McGuire & Diesmos, 2000), over-anaesthetized in chlorobutanol (1,1,1-trichloro-2-methyl-2-propanol) and dissected in the field for liver, skeletal muscle (quadriceps and tongue) and heart samples. Tissues were placed together in a single cryogenic vial and flash frozen by immersion in liquid nitrogen within 5 min of dissection. Once transported to the laboratory at Miami University (Oxford, Ohio, USA), tissues were frozen at -80 C until biochemical studies ( 2 years later) were conducted. Voucher specimens were photographed (Kodak Kodachrome 64; colours recorded in notes by R.M.B. at the time of capture), fixed in buffered 10% formalin, and transferred to 70% ethanol after being returned to the US 2 months later (Simmons, 1987). Voucher specimens (Appendix 1; fluid-preserved carcasses) used in electrophoretic analyses are deposited in collections at the National Museum of the Philippines (PNM), The Cincinnati Museum of Natural History (CMNH), the Texas Memorial Museum (TNHC), and the United States National Museum of Natural History (USNM; all museum acronyms, with the exception of CMNH, follow Leviton et al., 1985). MORPHOLOGICAL CHARACTERS AND MULTIVARIATE ANALYSES We examined 1067 fluid-preserved specimens of frogs of the R. signata complex from the Field Museum of Natural History (FMNH), The California Academy of Sciences (CAS), The Natural History Museum, London (BM), PNM, CMNH, TNHC and USNM for characters of diagnostic morphology (colour pattern, nuptial

7 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 399 Rana similis 120 Rana moellendorffi Rana grandocula Rana mangyanum Rana signata + Rana picturata no R. signata frogs Luzon Republic of the Philippines 250 km Polilio Catanduanes Mindoro Samar Busuanga Panay Cebu Leyte Palawan 16 36,37 Negros Bohol Camiguin 3 5 Balabac Borneo Basilan Jolo Tawi-tawi Mindanao Sulu Archipelago Figure 4. Geographical distribution of the Philippine members of the Rana signata complex. Numbers correspond to locality data presented in Appendix 1. Some points encompass two or more localities if these were in close proximity to one another or were located on small islands (i.e. Basilan). Other important Bornean (i.e. Bruneian and Malaysian localities in the state of Sarawak) and Indonesia (Siberut Isl.) localities are shown in Fig. 1 and listed in Appendix 1. pads, humeral glands, digital characters and body proportions) and morphometric characters were recorded for multivariate analyses. The following 32 mensural and meristic characters (Matsui, 1984; Heyer et al., 1990; Brown, McGuire & Diesmos, 2000) were recorded from fluid preserved specimens from all Philippine and Bornean populations used in electrophoretic analyses and from other localities where sample sizes were sufficient ( 30 specimens) for multivariate analyses: snout-to-vent length (SVL), head length (HL), snout length (SNL), interorbital distance (IOD), internarial distance (IND), eye diameter (ED), tympanic annulus diameter (TAD), head width (HW), upper arm length (UAL), forearm length (FAL), femur length (FL), tibia length (TBL), tarsus length (TSL), pes length (PL), manus N length (ML), fourth toe length (Toe4L), first toe length (Toe1L), first finger length (Fin1L), third finger length (Fin3L), nuptial pad length (NPL), nuptial pad width (NPW), width of terminal disc on fourth toe (Toe4DW), width of terminal disc on first finger (Fin1DW), width of terminal disc on third finger (Fin3DW), dorsolateral stripe width (if present; width recorded above tympanum; DLSW), metatarsal tubercle length (MTTL), humeral gland length in males (HG), eye-totympanum distance in females (ETD), number of transverse dark bars on femoral segment of hindlimb (FB), number of transverse dark bars on tibial segment of hindlimb (TB), the number of middorsal spots between dorsolateral light lines (MDSPOTS), and the number of light labial spots (LBSPOTS). All measurements were taken to the nearest 0.1 mm (with digital calipers and a microscope when necessary); only data scored by R.M.B. were used in an effort to reduce intermeasurer inconsistencies (Lee, 1990) and data were collected from specimens judged to be sexually mature adults on the basis of gonadal inspection (when possible) and/or development of secondary sexual characteristics. Symmetrical characters were scored on the specimen s right side. We qualitatively confirmed the assumptions of normality and homoscedasticity by examining frequency distributions of each variable (Sokal & Rohlf, 1981) and then performed multivariate statistics using software by Statistica TM. (StatSoft, 1994) and Statview TM. (Abacus Concepts, 1992). The data initially were explored and errors removed by examining graphs of each variable plotted against SVL (not shown). Although we tentatively identified groups on the basis of character differences, we performed Principal Component Analyses (PCA) to determine whether continuous morphological variation could form the basis of detectable structure in the data (group separation without a priori taxonomic assignment) and to examine for correlations of various variables with particular principal components. We also performed Canonical Variates Analyses (CVA) to separate the taxa more effectively. We examined standardized canonical coefficients to determine the relative contributions of individual variables to intergroup separation. We also examined factor structure matrices to examine for correlations of individual variables with particular canonical axes. In all multivariate analyses, we successively extracted components (in the case of the PCA), canonical coefficients and structural factors (CVA) until eigenvalues dropped below 1.0 and percentage contributions to the total cumulative variation dropped below 2%. All analyses were conducted separately for the sexes and were performed on logtransformed data. Initially, we used multivariate analyses to quantitatively describe morphological variation in the R.

8 400 R. M. BROWN and S. I. GUTTMAN signata complex, reduce variables to a manageable set of axes and explore multivariate space for contributions of various variables to intergroup dispersion. However, in final analyses, we specifically asked whether the species that historically have been confused (R. signata vs. R. picturata, and R. similis vs. R. mangyanum, new species, described below) could be discriminated on the basis of continuously varying mensural characters? ADVERTISEMENT CALLS Advertisement calls of frogs were recorded (by R.M.B.) in the field using a Sony WM DC6 Professional Walkman and a Sennheiser ME80 condenser microphone equipped with a K3U power module. Calls were recorded between June and August (the beginning of the rainy season). Recordings were collected between 19:00 and 22:00 h and ambient temperatures ranged from 24 to 25.5 C (no temperature corrections were needed due to the narrow range of temperature variation). All recordings were vouchered and specimens deposited in US museums (above and Appendix 1). Calls were recorded at distances ranging from 0.5 to 2.0 m and were digitized and analysed using Soundedit (Macromedia, 1995) and Canary (Charif, Mitchell & Clark, 1996) software on a Macintosh computer. We examined oscillograms (waveforms), audiospectrograms (sonograms) and results of the Fast Fourier Transformation (frequency spectra) for a variety of qualitative and quantitative characters (Table 1). Call rate ((total number of calls -1)/time from beginning of first call to beginning of the last) was determined during 3.5-min continuous calling intervals. Data were recorded from 3 to 5 individuals per species (10 15 calls per individual). PROTEIN ELECTROPHORESIS Heart, liver and skeletal muscle were combined (in a volume ratio of 1 : 2 : 2), homogenized in equal portions of 0.25 M sucrose: 2% 2-phenoxyethanol grinding solution, centrifuged for 3 min at r.p.m., and electrophoresed immediately. Supernatant was refrozen and stored at -80 C so that electrophoresis could be repeated as necessary to confirm allelic designations; a maximum of three thaw cycles were used and allozymes retained activity over these cycles. We employed standard horizontal starch gel electrophoretic procedures (Selander et al., 1971; Guttman, 1985; Richardson, Braverstock & Adams, 1986) and histochemical staining techniques (Murphy et al., 1996; Menchenko, 1994; Table 2); all gels were 12% (w/v) Sigma starch, run at 5 7 C. Eight buffer systems consistently resolved 42 loci (Table 1); interpretation of banding pattern, subunit structure and identification of polymorphic character states followed Ferguson (1980). Electromorphs were evaluated by running individuals of uncertain allelic assignment in adjacent lanes on comparison gels. In some cases (Table 1: ADH, LDH-1, GPI-2 and SOD) it was necessary to employ sequential electrophoresis to fully resolve cryptic heterogeneity (Highton & Hedges, 1995). With the assumption that electromorphs were homologous if they demonstrated identical mobilities, we assigned alphabetical designations to banding patterns (the most anodal migrant = A, the next = B, and so on) and subjected data to the following analyses. DATA CODING AND ANALYSES Estimates of genetic variability, heterozygosities, allele frequencies and chi-square tests for deviations from Hardy Weinberg expectation (Bonferonni-type alpha level correction employed for multiple comparisons) were obtained with BIOSYS-1 software (Swofford & Selander, 1981) analysed on a mainframe computer. We used allele frequencies to construct matrices of pairwise genetic distances (Rogers, 1972; Wright, 1978) and constructed UPGMA and Wagner Table 1. List and definitions of qualitative and quantitative variables scored in the call analyses. Variables were scored from the waveforms and frequency spectra representative of the following species: Rana similis, R. grandocula, R. moellendorffi and R. mangyanum Call length Call rate Pulse number Fundamental frequency Tonal introduction Supplementary pulses Dominant frequency identity Pulse redundancy Tonal attachments Frequency modulation Time (in s) from beginning to end of entire call (total number of calls 1)/time from beginning of first pulse to beginning of last pulse The total number of pulses per call Lowest frequency component of the call (not necessarily dominant) Presence/absence of tonal introduction at beginning of call Presence/absence of separate, distinct subpulses following main call pulses Emphasized frequency = fundamental or first harmonic Pulse structure similar throughout call or varied Noisy first portion of call reduces to tonal quality by end of pulse Presence/absence of frequency modulation at any point during call

9 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 401 Table 2. Electrophoretic protocols: Proteins visualized, number of loci, and counter-ion systems for the products of 42 presumptive loci (IUBNC, 1984) Enzyme locus Locus IUBNC No. loci Buffer(s) Arginine Phosphokinase APK-ARGK Adenosine Deaminase ADA Adenylate Kinase AK-ADK Alcohol Dehydrogenase ADH Aspartate Aminotransferase GOT Creatine Kinase CK Esterase EST L-Fucose Dehydrogenase FUCDH Fumarate Hydratase/Fumarase FUM Fructose-1, 6- Diphosphate Dehydrogenase FDP Glycerate Dehydrogenase G-2-DH Glucose Dehydrogenase GDH Glucosephosphate Isomerase GPI Hexokinase HK Lactate Dehydrogenase LDH Isocitrate Dehydrogenase IDH Malate Dehydrogenase MDH Malic Enzyme ME Phosphoglucomutase PGM l-leucyl-leucyl peptidase PEP Superoxidase Dimutase SOD Triose-phosphate Isomerase TPI Non-specific protein TP 2 1 Total = 42 loci 1: TC 8.0: Tris-citrate, tray = ph 8.0, gel = ph 8.0 ; 2: HC 5.7, Histidine Citrate, tray = ph 5.7, gel 1:6 dilution (dd H20): 3: HC 8.0, Histidine Citrate, tray = ph 8.0, gel = 1:29 dilution; 4: CT 8.2, Citric Acid-Tris, tray = ph 8.2; gel = 1:19 dilution; 5: TME, Tris EDTA Maleate, tray = ph 7.4, gel = 1:9 dilution; 6: RW, Ridgeway, tray = ph 8.5, gel = 1:100 dilution; 7: CT 6.0, Citric Acid-Tris, tray = ph 6.1, gel = ph 6.0; 8: THCL, Tris HCl, tray = Poulik, ph 8.2, gel = Tris HCl, ph 8.0. networks on two matrices. Owing to the fact that four of the intermediate outgroup species (R. chalconota, R. hosii, R. luctuosa and R. baramica; see below for justification of outgroup designations) were represented by older (8 15 year) tissue samples in which, presumably, some enzyme activity had been lost, we were unable to resolve electromorphs in one or more of these species for 14 of the 42 loci surveyed. Similarly, we were unable to resolve four of these same loci for R. picturata and R. signata. These loci were excluded (because BIOSYS requires at least partial data for each taxon for each locus) and an all 19 OTUs (reduced number of loci (n = 28)) matrix was utilized in a separate analysis with populations as terminals. Alternatively, we excluded the four taxa with incomplete allozyme data and constructed an all 40 loci (reduced number of OTUs (n = 15)) matrix with species (populations pooled) as terminal taxa. Phylogenetic analyses were performed after coding each locus as a character and various alleles (or allelic arrays; see below) as states (Murphy, 1993; Swofford et al., 1996). Although there is a consensus that allozyme data should be coded in this manner (vs. scoring the presence or absence of each allele) for phylogenetic analyses (Buth, 1984; Guttman et al., 1990; Murphy, 1993), there has been recent debate concerning how polymorphic allozyme data should be treated due to the usual abundance of intraspecific variation (Crother, 1990; Mabee & Humphries, 1993; Wiens, 1995, 1998, 2000; McGuire, 1996; Mink & Sites, 1996; Wiens & Servidio, 1997; Murphy & Doyle, 1998). We utilized the Manhattan distance/frequency parsimony coding approach (Swofford & Berlocher, 1987; Wiens, 1995; Berlocher & Swofford, 1997) in which a matrix of Manhattan distances, calculated between all pairs of taxa, was constructed from raw allele frequencies for each polymorphic locus. Manhattan distances were then entered into character step matrices to weight transitions between character states in subsequent parsimony analyses (see Berlocher & Swofford, 1997, for further explanation). This method is akin to the recommendations of Mabee & Humphries (1993) but

10 402 R. M. BROWN and S. I. GUTTMAN differs through the use of raw frequencies to incorporate frequency information while allowing multistate characters (McGuire, 1996; Mendoza-Quijano, Flores- Villela & Sites, 1998). In a recent review of methods of analysing allozyme data, Wiens (2000) demonstrated that methods that make use of allele frequency data perform well in a variety of conditions and generally out-perform methods that do not make use of this kind of information. In our analysis, 24 of the 40 potentially informative loci were intraspecifically variable and were input with character step matrices; all characters received equal weight in all phylogenetic analyses. In order to determine the effects of differing coding strategies on tree topologies, we also scored intraspecifically variable loci as unspecified polymorphisms by evoking PAUP s interpret multiple states as uncertainty option. Additionally, we employed a 90% cutoff option in which rare (usually autapomorphic) alleles were discarded (Guttman et al., 1990; Green & Borkin, 1993) and the remaining (with intraspecifically varying allele frequencies of 11 89%) terminals were assigned multiple character states and PAUP s interpret multiple states as uncertainty option was again invoked. These analyses were performed on unordered multistate characters and we rooted the tree with Rana vittigera, R. cancrivora, and Limnonectes magnus. We performed parsimony analyses using PAUP* (Swofford et al., 1996; Swofford, 1998). Owing to a relatively small number of taxa, we were able to use the Branch-and-Bound algorithm of Hendy & Penny (1982), guaranteeing that the shortest trees were recovered. Because the sister group to the R. signata complex had not been identified, we rooted trees by utilizing multiple outgroups including several potential sister group taxa. These included two species that have, at times, been allied in the same subgenus (or genus; Dubois, 1987, 1992) as the R. signata complex (R. baramica, R. luctuosa (= subgenus Pulchrana)), representatives of two other presumably related subgenera (R. chalconota, R. hosii (= Chalcorana), and R. erythraea (= Hylarana)), and three distantly related groups (Limnonectes magnus (= genus Limnonectes), R. cancrivora, and R. vittigera (= subgenus Fejervarya)). Tree stability and information content were investigated using 500 non-parametric bootstrap iterations (Felsenstein, 1985; Hillis & Bull, 1993). Phylogenetic signal was inferred from degree of hierarchical structure contained in the data as elucidated by univariate tree length distribution skewness statistics (g 1 statistic; Hillis, 1991; Huelsenbeck, 1991; Hillis & Huelsenbeck, 1992). For critical g 1 confidence values, we used the more conservative values for binary characters provided by Hillis & Huelsenbeck (1992). RESULTS MULTIVARIATE ANALYSES OF MORPHOLOGICAL VARIATION Standard univariate statistics for the morphometric variables are presented in Tables 3 and 4. Results of the multivariate analyses are presented separately for each sex below. Only data for 24 mensural characters could be reliably scored for use in multivariate analyses. The remaining meristic characters (FB, TB, MDSPOTS, LBSPOTS) and certain mensural characters that were difficult to consistently record (and had inflated degress of intrapopulational variation; e.g. Toe1L in both sexes and ETD, NPL and NPW in males) were excluded from multivariate analyses. These characters were nevertheless summarized (Tables 3 and 4) to aid in the diagnoses of species. Males Although 10 principal components accounted for 90% of the total variation, we extracted and discuss only the first three (comprising 64.5% of the total variance; Table 5) because eigenvalues for components IV XI were <1.0 and because, thereafter, each individual component accounted for <2% of the total variance. Furthermore, components IV XI did not form the basis of any detectable structure in the data (plots not shown). The loadings for PC I were all positive and generally large in magnitude (with the exception of IND, DLSW and HG), indicating that PC I primarily described body size variation. In plots containing all species (not shown) size-based PC I does not provide the basis for identifying as distinct any species pairs, although the species at the size extremes (R. similis and R. signata, the smallest species, and R. grandocula, the largest) exhibited reduction of overlap (not shown). Although some taxon-based structure in the data was detected between R. picturata and R. signata, overlap in size-based PC I was apparent (Fig. 5); in the case of R. similis and R. mangyanum, overlap was even more extreme. The PC II axis loaded most heavily and positively on DLSW, HG, MTTL, Toe4DW Fin1DW and TAD, indicating high correlations of these variables with this component. This shape-based axis formed the basis of some distinction between both the R. similis R. mangyanum and R. picturata R. signata comparisons, but some overlap remained between groups (Fig. 5). PC III loaded heavily and positively on IND and IOD and heavily and negatively on Fin1L, MTTL, Fin3L and HG. In the case of PC III, R. picturata was partially distinct from the remaining taxa on the basis of this axis (not shown) but both the R. similis R. mangyanum and R. picturata R. signata comparisons demonstrated considerable overlap (Fig. 5). In summary, the principal components analysis demonstrated that (in males)

11 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 403 Table 3. Morphometric variation (in mm) in adult members of the Rana signata complex from Borneo and the eastern island arc (Mindanao Luzon) of the Philippines. Table entries include mean ± 1 standard deviation (range below) and sample size (n). See text for character abbreviations. For ease of comparison, data from multiple islands are pooled by species for R. similis (Catanduanes, Polillo, Luzon) and R. grandocula (Bohol, Mindanao, Camiguin). Specimens previously referred to R. yakani Taylor (Basilan Isl. specimens) are summarized separately R. similis R. grandocula R. grandocula Luzon, Polillo, and Mindanao, Bohol, Basilan Isl. R. signata Catanduanes Islands Camiguin Islands (R. yakani types) Borneo male female male female male male female n = 128 n = 80 n = 172 n = 99 n = 36 n = 118 n = 27 SVL 38.8 ± ± ± ± ± ± ± HL 15.0 ± ± ± ± ± ± ± SNL 6.3 ± ± ± ± ± ± ± IOD 4.1 ± ± ± ± ± ± ± IND 4.0 ± ± ± ± ± ± ± ED 5.6 ± ± ± ± ± ± ± TAD 3.2 ± ± ± ± ± ± ± HW 12.5 ± ± ± ± ± ± ± UAL 7.4 ± ± ± ± ± ± ± FAL 8.9 ± ± ± ± ± ± ± FL 19.1 ± ± ± ± ± ± ± TBL 20.1 ± ± ± ± ± ± ± TSL 11.5 ± ± ± ± ± ± ± PL 19.1 ± ± ± ± ± ± ± ML 11.0 ± ± ± ± ± ± ± Toe4L 16.7 ± ± ± ± ± ± ± Toe1L 4.6 ± ± ± ± ± ± ± Fin1L 6.4 ± ± ± ± ± ± ±

12 404 R. M. BROWN and S. I. GUTTMAN Table 3. Continued R. similis R. grandocula R. grandocula Luzon, Polillo, and Mindanao, Bohol, Basilan Isl. R. signata Catanduanes Islands Camiguin Islands (R. yakani types) Borneo male female male female male male female n = 128 n = 80 n = 172 n = 99 n = 36 n = 118 n = 27 Fin3L 7.7 ± ± ± ± ± ± ± NPL 4.6 ± ± ± ± NPW 1.6 ± ± ± ± Toe4DW 1.2 ± ± ± ± ± ± ± Fin1DW 1.1 ± ± ± ± ± ± ± Fin3DW 1.1 ± ± ± ± ± ± ± DLSW 1.3 ± ± ± ± ± ± ± MTTL 1.7 ± ± ± ± ± ± ± HGL 1.8 ± ± ± ± ETD 1.4 ± ± ± ± ± ± ± FB 3.7 ± ± ± ± ± ± ± TB 3.5 ± ± ± ± ± ± ± MDSPOTS ± ± LBSPOTS ± ± taxon-based data structure in multivariate space was qualitatively detectable but in no cases were species markedly distinguished on the basis of this analysis. As expected, the CVA more effectively discriminated between males of each species than did the PCA. CV I forms the basis of near complete separation between both species pairs (Fig. 6). CV coefficients weighing heavily on this axis include HG and DLSW (both positive) and to a lesser extent HL and TBL (both negative; Table 5). CV II does not discriminate between either of the species pairs. CV III forms the basis of partial discrimination between R. signata and R. picturata but not R. similis and R. mangyanum. This axis weights heavily on TBL, SVL (both positive) DLSW and Toe4L (both negative). The factor structure matrix indicates the presence of a moderate positive correlation between CV I and the variables DLSW and MTTL (Table 5). Rana picturata syntypes fall at one end of the range of CV III. In particular, one specimen (BM ) falls as an extreme outlier in ordinations of CV I vs. III (Fig. 6), suggesting that it does not fall

13 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 405 Table 4. Morphometric variation (in mm) in adult members of the Rana signata complex from Borneo and the western island arc (Palawan Mindoro) of the Philippines. Table entries include mean ± 1 standard deviation (range below) and sample size (n). See text for character abbreviations R. moellendorffi R. mangyanum R. picturata male n = 102 female n = 47 male n = 66 female n = 40 male n = 148 female n = 36 SVL 38.5 ± ± ± ± ± ± HL 15.2 ± ± ± ± ± ± SNL 6.4 ± ± ± ± ± ± IOD 4.4 ± ± ± ± ± ± IND 4.3 ± ± ± ± ± ± ED 5.7 ± ± ± ± ± ± TAD 3.5 ± ± ± ± ± ± HW 12.8 ± ± ± ± ± ± UAL 7.4 ± ± ± ± ± ± FAL 8.9 ± ± ± ± ± ± FL 19.5 ± ± ± ± ± ± TBL 20.3 ± ± ± ± ± ± TSL 11.6 ± ± ± ± ± ± PL 19.6 ± ± ± ± ± ± ML 11.1 ± ± ± ± ± ± Toe4L 17.5 ± ± ± ± ± ± Toe1L 4.6 ± ± ± ± ± ± Fin1L 6.5 ± ± ± ± ± ± Fin3L 7.7 ± ± ± ± ± ± NPL 4.7 ± ± ±

14 406 R. M. BROWN and S. I. GUTTMAN Table 4. Continued R. moellendorffi R. mangyanum R. picturata male n = 102 female n = 47 male n = 66 female n = 40 male n = 148 female n = 36 NPW 1.7 ± ± ± Toe4DW 1.2 ± ± ± ± ± ± Fin1DW 1.2 ± ± ± ± ± ± Fin3DW 1.2 ± ± ± ± ± ± DLSW 1.5 ± ± ± ± ± ± MTTL 1.9 ± ± ± ± ± ± HGL 2.9 ± ± ± ETD 1.5 ± ± ± ± ± ± FB 3.2 ± ± ± ± ± ± TB 3.2 ± ± ± ± ± ± MDSPOTS 17.6 ± ± ± ± LBSPOTS 2.6 ± ± ± ± into a morphologically cohesive group with the other specimens of Rana picturata examined here (see comments below and in the R. picturata taxonomic account). The classification analysis correctly classified % of specimens with their correct a priori species grouping (R. grandocula: 81.8% correct; R. similis: 90.3%; R. mangyanum: 77.0%; R. moellendorffi: 85.3%; R. picturata; 93.6%; R. signata; 93.7%; mean = 88.3%). The syntypes (n = 4) of R. picturata were 75% correctly classified, with one of the specimens misclassified as R. moellendorffi. In fact, this individual may in fact be a specimen of R. moellendorffi (see comments under species account for R. picturata) indicating that the type series may contain more than one species. The holotype of R. signata was correctly classified as were the holotype and paratype of R. siberu. Females Because the results of multivariate analyses of data from females generally did not qualitatively differ from those of males (but see exceptions, below), we present here only an abbreviated account of these analyses and omit duplication of the figures. For females, we extracted eight principal components before accounting for 90% of the total variance. The first three principal components accounted for 76.5% of the variation. With the exception of ED, DLSW and ETD all loadings in PC I were large and positive (Table 6), indicating that PC I generally can be interpreted as a size variable in females. Although we observed no complete taxon-based group structure on the PC I axis, the largest species, R. grandocula, was nearly separated from R. signata, R. similis and R. moellendorffi (not shown). Body size variation in R. mangyanum (vs. R. similis) and

15 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 407 Table 5. Multivariate analyses of 24 mensural characters (male specimens). Principal components PCs I III, extracted from the correlation matrix, and the canonical coefficients for canonical axes CVs I III (standardized by within-group pooled standard deviations for each variable). The final column is the factor structure components (FSC) for axis I. All data were log transformed Variable PC I PC II PC III CV I CV II CV III FSC I SVL HL SNL IOD IND ED TAD HW UA FAL FL TBL TSL PL ML Toe4L Fin1L Fin3L Toe4DW Fin1DW Fin3DW DLSW MTTL HG Eigenvalue Cum. % Var R. picturata (vs. R. signata) was apparent in PC I, but species overlap persisted. PC II loaded heavily and positively on DLSW, Toe4DW, Fin1DW and Fin3DW, and formed the basis of near-complete separation between R. mangyanum and R. signata. Rana picturata and R. signata were nearly distinguishable on the basis of this axis (as are R. similis and R. mangyanum), but some group overlap remained. PC III loaded most heavily and positively on ED and Fin3DW, and negatively on ETD and DLSW (Table 6). In a plot of all species considered (not shown), no species are distinguishable on the basis of this axis. With respect to the principal components analysis, females exhibited limited taxon-based structure, with some components nearly forming the basis of taxonbased structure in the data. In females, the CVA more effectively discriminated between groups than did the PCA, as expected. CV I formed the basis of separation between both species pairs, with minimal overlap between each group (not shown). CV coefficients that weighed heavily in this axis included TBL (positive), DLSW (negative), and to a lesser extent TSL, IOD, FL (positive) and UAL (negative; Table 6). CV II partially discriminated between R. similis and R. mangyanum but not between R. signata and R. picturata. PL and TBL weighed heavily and positively in CV II, and HL (positive), and SVL (negative) contributed to a lesser extent in this axis. CV III formed the basis of near complete discrimination between R. signata and R. picturata but not R. similis and R. mangyanum. This axis weighed heavily on MTTL, TBL, TAD (positive), IND and HL (negative). The factor structure matrix indicated the presence of a strong negative correlation between CV I and DLSW and weakly positive correlations between this axis and TBL, TSL FinIL, IOD and PL (Table 6). The classification analysis correctly placed between 80.8 and 92.3% of specimens within their correct a priori species grouping (R. grandocula: 89.8% correct; R. similis: 90.1%; R. mangyanum: 85.9%; R. moellendorffi: 80.8%; R. picturata; 92.1%; R. signata; 92.3%; mean = 88.3%).

16 408 R. M. BROWN and S. I. GUTTMAN 3 A 2 C PC I PC I 0 PC I B PC II PC III PATTERNS OF ADVERTISEMENT CALL VARIATION The distribution of qualitative call characters (Table 7) both supports the taxonomic identity of each species (i.e. each species produces a distinguishable call with fixed differences between it and other members of the Rana signata complex; Fig. 7) and supports a pattern of two closely related Philippine species pairs (each species pair shares characters to the exclusion of the other species pair). Depending on character states in the unsampled lineages for which call recordings were unavailable (R. signata and R. picturata), some of these characters (e.g. pulse redundancy, and identity PC I D PC II PC III Figure 5. Bivariate orientation of the first two principal components plotted separately for male specimens of R. similis and R. mangyanum (A,B) and for males of R. picturata and R. signata (C,D). The position of one R. picturata syntype (BM ) that may represent a specimen of R. moellendorffi (see text) is indicated with an open triangle (arrow). Open circles = R. similis, shaded circles = R. mangyanum, open squares = R. picturata, and shaded squares = R. signata. of the emphasized frequency) may represent synapomorphies for their respective clades. Pulse number per call (Fig. 8A) exhibited a pattern in which members of a species pair did not differ significantly from one another but each species in a pair differed significantly from the species in the other pair (Sheffe s test; intraclade comparisons, P values >0.05; interclade comparisons, P values <0.0001); nevertheless, there was a significant main species effect (ANOVA: F = 46.9; d.f. = 3,4; p < ). Call rate (Fig. 8B) differed significantly between all species (1 way ANOVA, followed by Sheffe s post hoc comparisons: P values = 0.001) and an analysis of variance (ANOVA)

17 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE A 4 C CV I CV I -2 CV I B CV II detected a significant main species effect (F = 50.07; d.f. = 2, 12; P = ). Fundamental frequencies (Fig. 8C) of the calls were very similar between R. similis and R. grandocula (Sheffe s P > 0.05) and each of these species differed significantly from R. moellendorffi (Sheffe s P < 0.001) but not from R. mangyanum (Sheffe s P > 0.05). Finally, call length (Fig. 8D) exhibited a significant main species effect (ANOVA: F = 46.35; d.f. = 3, 4; P = ) and all species comparisons were significant (Sheffe s P values <0.001) -1.0 CV III Figure 6. Bivariate orientation of the first two canonical variate scores for males of R. similis and R. mangyanum (A,B) and R. picturata and R. signata (C,D). Open circles = R. similis, shaded circles = R. mangyanum, open squares = R. picturata (exclusive of syntypes), and shaded squares = R. signata (exclusive of holotype), open triangles = R. picturata syntypes, shaded triangle = R. signata holotype. Arrow denotes one R. picturata syntype (BM ) that may represent a specimen of R. moellendorffi (see text). CV I D CV II CV III except for the comparison between R. similis and R. mangyanum (P > 0.05). ALLELE FREQUENCY VARIATION AND GENE DIVERSITY Of the 42 gene loci resolved, two (GDH-2 and PEP) were monomorphic across all taxa and 16 were polymorphic among, but monomorphic within, species (Table 8). Eight more loci exhibited limited intraspecific variation in one to three species (one allele preva

18 410 R. M. BROWN and S. I. GUTTMAN Table 6. Multivariate analyses of 24 mensural characters (female specimens). Principal components PCs I III, extracted from the correlation matrix, and the canonical coefficients for canonical axes CVs I III (standardized by within-group pooled standard deviations for each variable). The final column is the factor structure components (FSC) for axis I. All data were log transformed Variable PC I PC II PC III CV I CV II CV III FSC I SVL HL SNL IOD IND ED TAD HW UAL FAL FL TBL TSL PL ML Toe4L Fin1L Fin3L Toe4DW Fin1DW Fin3DW DLSW MTTL ETD Eigenvalue Cum. % Var Table 7. The distribution of qualitative call characters in Philippine members of the Rana signata stream frog complex Species R. similis R. grandocula R. moellendorffi R. mangyanum Tonal introduction + Supplementary pulses + Dominant frequency = fundamental + + Tonal elements before/after pulse + + Pulse redundancy + + Frequency modulation + lent at a frequency of 90%), and the remainder (n = 16) possessed intermediate allele frequencies of 50 89% for one to three species. In the case of these latter two patterns, the remaining loci (other than the one to three species that varied at these loci) were characterized by alleles that did not vary intraspecifically and were polymorphic interspecifically. Of the 14 loci not resolved in R. baramica, R. luctuosa, R. chalconota and R. hosii, four were also unresolvable in R. signata and R. picturata (Table 8). Pairwise genetic distances of the all-otus (reduced-loci) and reduced-otus (all-loci) matrices (Appendices 2 5) are slightly different, depending on the number of loci included. Four loci (CK-1, CK-2, LDH-1, ADA-1) could not be resolved in R. picturata or R. signata. For the calculations of genetic distances, allele frequency

19 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 411 Figure 7. Audiospectrograms (frequency vs. time) and waveforms (amplitude vs. time) from advertisement calls of (A) R. mangyanum, (B) R. moellendorffi, (C) R. similis and (D) R. grandocula. summary statistics and phenetic analyses, these loci will not be considered further; they are included only in cladistic analyses (see below). Conspecific populations of the R. signata complex were consistently more similar to each other than to heterospecifics (Appendices 2 and 3) and the degree of divergence between most taxa (Appendices 4 and 5) depended primarily on the number of fixed allelic differences between species. Conspecific populations invariably clustered together in phenetic analyses (UPGMA and Wagner phenograms; not shown) and the topology of clustering networks consistently demonstrated the presence of two major groups within the R. signata complex. One of these was composed of R. signata, R. similis and R. grandocula and another consisted of R. picturata, R. moellendorffi and R. mangyanum. No fixed allelic differences were detected between any populations on the same island except for those from Borneo assigned to R. signata and R. picturata. In subsequent analyses we pooled data for populations within Philippine Pleistocene aggregate islands and considered these to be distinct species (see

20 412 R. M. BROWN and S. I. GUTTMAN Number of pulses per call Fundamental frequency (khz) A sim grand moel mang C sim grand moel mang Call rate (calls s -1 ) Mean call length (s) sim grand moel mang sim grand moel mang Figure 8. Variation in four call characters in Philippine members of the Rana signata complex: (A) number of pulses per call; (B) call rate; (C) fundamental frequency; and call duration (D); sim = R. similis; grand = R. grandocula; moel = R. moellendorffi; mang = R. mangyanum. B D species concept section above, and taxonomic conclusions, below). Mean number of electromorphs per locus (E l ) ranged from 1.0 to 1.2. Percent polymorphic loci (% p) ranged from 0 to 15%, and mean heterozygosities (by direct count and as predicted by Hardy Weinberg) were extremely low, ranging from 0.0 to 0.02 (Table 9). The following species had one or more loci with genotype frequencies significantly deviating from the predictions of Hardy Weinberg expectations (sequential Bonferroni alpha level protection adjusted for the number of comparisons; loci included in parentheses): R. vittigera (HK-1, ME-2), R. erythraea (ME-2), R. grandocula (ME-1, ADH-2), R. picturata (ME-1, FDP), R. signata (MDH-1, MDH-2, GOT-1, ME-2), R. similis (AK-1). Data on individual genotypes by locality and species are available on request from R.M.B. In summary, estimates of genetic variability, allele frequencies, heterozygosities, low number of deviations from Hardy Weinberg expectations, and clustering topologies of distance networks all verified the presence of relatively low intraspecific genetic variation. Higher degrees of genetic variation were structured on an interisland scale and were detected in the form of fixed or nearly fixed differences between allopatric populations on separate Pleisticene aggregate island complexes (Tables 8 and 10; Fig. 2). Lineages of Philippine stream frogs on separate island bank platforms had 2 22 (mean ± SD = 14.8 ± 8.9, n = 6) fixed allelic differences between species and Bornean species in sympatry were separated by 14 fixed allelic differences (Table 10). SUMMARY OF TAXONOMIC CONCLUSIONS Fixed allelic differences (and numerous nearly fixed differences; Table 9) were detected between all pairwise comparisons of the four allopatric populations in the Philippines and two sympatric populations in Borneo. In accordance with a lineage-based species

21 Table 8. Electromorph frequencies for 19 populations of 14 SE Asian ranid species examined electrophoretically. Population numbers (No.) correspond to species/localities included in Fig. 4 and Appendix 1 (44ba = R. baramica, 41ca = Rana cancrivora, 43ch = R. chalconota, 41er = R. erythraea, 9gr, 10gr = R. grandocula, 45ho = R. hosii, 44lu = R. luctuosa, 9ma = L. magnus, 38 mo, 39mo = R. moellendorffi, 34mg = R.mangyanum, 44pi = R. picturata, 44si = R. signata, 23sm 26sm = R. similis, 41vi = R. vittigera); n = sample size; dash ( ) designates a locus that was unresolved for the species/population indicated. GDH-2 and PEP were fixed (monomorphic) across all populations and are not included in this table Population number No. 44ba 41ca 43ch 41er 9gr 10gr 45ho 44lu 9ma 38mo 39mo 34mg 44pi 44si 23sm 25sm 21sm 26sm 41vi (n) Locus PGM-1 A B C D E PGM-2 A B C 0.91 D E 0.02 F G IDH-1 A B C D IDH-2 A B C D E 1.00 SOD A B C D 1.00 nloaded from by guest on 06 October 2018 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 413

22 Table 8. Continued Population number No. 44ba 41ca 43ch 41er 9gr 10gr 45ho 44lu 9ma 38mo 39mo 34mg 44pi 44si 23sm 25sm 21sm 26sm 41vi (n) LDH-1 A B 1.00 C D E F G H LDH-2 A B C D E ADH-1 A B C D E F 1.00 ADH-2 A B C D G2DH-1 A B G2DH-2 A B C D E nloaded from by guest on 06 October R. M. BROWN and S. I. GUTTMAN

23 GDH-1 A B AK-1 A B C AK-2 A B 1.00 C ARGK A 1.00 B C D E F EST-1 A B C D E 1.00 EST-2 A B C ADA-1 A * B C 1.00 D 1.00 ADA-2 A B C nloaded from by guest on 06 October 2018 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 415

24 Table 8. Continued Population number No. 44ba 41ca 43ch 41er 9gr 10gr 45ho 44lu 9ma 38mo 39mo 34mg 44pi 44si 23sm 25sm 21sm 26sm 41vi (n) HK-1 A B C D 1.00 E 0.09 HK-2 A B C TPI-1 A B C 1.00 D 1.00 E 1.00 TPI-2 A B C 1.00 FDP A B C FUCDH A 1.00 B C D E CK-1 A B C 1.00 nloaded from by guest on 06 October R. M. BROWN and S. I. GUTTMAN

25 CK-2 A B CK-3 A B C D 1.00 CK-4 A B C 1.00 D TP-2 A B C 0.22 TP-3 A B 1.00 C D FUM A B C 1.00 MDH-1 A 1.00 B 1.00 C D 1.00 E F MDH-2 A 1.00 B C D E nloaded from by guest on 06 October 2018 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 417

26 Table 8. Continued Population number No. 44ba 41ca 43ch 41er 9gr 10gr 45ho 44lu 9ma 38mo 39mo 34mg 44pi 44si 23sm 25sm 21sm 26sm 41vi (n) GPI-1 A B C D 1.00 E F GPI-2 A 0.64 B C D E F G 1.00 GOT-1 A B C D E GOT-2 A B 1.00 C D 1.00 E F 1.00 ME-1 A B C D 0.03 ME-2 A 0.01 B C D *For Rana signata, only six individuals could be resolved for ADA-1 (all scored as aa ). nloaded from by guest on 06 October R. M. BROWN and S. I. GUTTMAN

27 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 419 Table 9. Summary statistics: mean sample size (n) per population, mean number of electromorphs per locus (E l), per cent polymorphic loci (% p) and direct count (DC) vs. expected (E hw; predicted by Hardy Weinberg equilibrium) mean heterozygosity (SE in parentheses). The subtable on the left summarizes the all OTUs (reduced number of loci) matrix; on the right are similar summaries of results for the all-loci (reduced number of OTUs) matrix Reduced-loci, all OTUs included Reduced OTUs, all loci included Mean heterozygosity Mean heterozygosity n E l % p DC E hw E l % p DC E hw R. baramica ( ) ( ) ( ) R. cancrivora (0.1) (0.01) (0.01) (0.1) (0.01) (0.01) R. chalconota ( ) ( ) ( ) R. erythraea (0.1) (0.02) (0.02) (0.1) (0.013) (0.016) R. grandocula (0.1) (0.01) (0.01) (0.1) (0.004) (0.007) R. grandocula (0.1) (0.0) (0.01) (0.1) (0.011) (0.010) R. hosii ( ) ( ) ( ) R. luctuosa ( ) ( ) ( ) L. magnus (0.0) (0.01) (0.01) (0.0) (0.001) (0.001) R. moellendorffi (0.1) (0.001) (0.001) (0.1) (0.002) (0.002) R. moellendorffi (0.0) (0.002) (0.002) (0.0) (0.001) (0.001) R. mangyanum (0.1) (0.02) (0.03) (0.1) (0.01) (0.017) R. picturata (0.1) (0.02) (0.01) (0.1) (0.01) (0.011) R. signata (0.0) ( ) ( ) (0.1) (0.01) (0.03) R. similis (0.1) (0.01) (0.02) (0.0) (0.005) (0.01) R. similis (0.2) (0.0) (0.0) (0.0) (0.01) (0.013) R. similis (0.0) (0.0) (0.0) ( ) ( ) ( ) R. similis (0.1) (0.004) (0.004) (0.1) (0.003) (0.004) R. vittigera (0.1) (0.03) (0.02) (0.1) (0.02) (0.01)

28 420 R. M. BROWN and S. I. GUTTMAN Table 10. Summary of interspecifically variable biochemical characters: numbers of fixed allelic differences (and 90% frequency differences in parentheses) between species R. moellendorffi R. grandocula R. similis R. signata R. picturata R. mangyanum 5 (1) 20 (4) 22 (2) 19 (0) 16 (2) R. moellendorffi 18 (4) 22 (1) 17 (2) 11 (5) R. grandocula 2 (4) 4 (1) 12 (4) R. similis 2 (0) 10 (1) R. signata 14 (3) A concept and in recognition of evidence suggesting unique evolutionary histories (cleary diagnosable morphological, behavioural and biochemical character states) and unique evolutionary fates (allopatry on different Pleistocene aggregate island complexes separated by 120 m channel depth) of lineage segments, we recognize each of Inger s subspecies (signata, similis, grandocula, moellendorffi) as distinct evolutionary lineages (species). Furthermore, as we have demonstrated the presence of 14 fixed allelic differences between two independent lineages in sympatry in Borneo and confirmed their identities by examination of the types and morphometric analyses of large samples, we recognize R. signata and R. picturata (rediagnosed below) as distinct species. Lastly, we detected 22 fixed allelic differences (Table 10) between the Mindoro population previously referred to R. s. similis and true R. similis from the near type locality (Laguna, Luzon), rendering the current taxonomic arrangement untenable. Thus, the Mindoro population should be reognized as a new species; we diagnose and describe this species below Polymorphic Coding Rana mangyanum R. moellendorffi R. picturata R. similis R. grandocula R. signata R. hosii R. chalconota R. baramica R. luctuosa R. erythraea R. cancrivora R. vittigera L. magnus B Rana mangyanum R. moellendorffi R. picturata R. similis R. grandocula R. signata R. hosii R. chalconota R. baramica R. luctuosa R. erythraea R. cancrivora R. vittigera L. magnus Majority Coding (90% Cutoffs) Figure 9. The 50% majority-rule consensus tree for (A) polymorphic and (B) majority (90% Cutoff) coding approaches (see text for explanation and tree statistics). Values above branches indicate bootstrap support if 50%. PHYLOGENY ESTIMATION Forty informative loci were available for the initial phylogenetic analyses. Analysis of the data coded by the treating intraspecific variation as unspecified polymorphisms (Polymorphic alternative) resulted in a single most parsimonious tree (Fig. 9A) of a length 200, a consistency index of 0.890, a retention index (RI) of and a g 1 tree length distribution skewness statistic of We take the strong left skew to the tree length frequency distribution (P 0.01; Hillis & Huelsenbeck, 1992) as evidence that data contain more hierarchical structure than randomized data, a finding consistent with the hypothesis of phylogenetic signal (Huelsenbeck, 1991; Hillis & Huelsenbeck, 1992). The ingroup topology consists of two major clades within the Rana signata complex, each composed of two Philippines species and a single Bornean species (((R. picturata (R. moellendorffi + R. mangyanum)) + (R. signata (R. similis + R. grandocula)), respectively). Bootstrap support for relationships within the ingroup topology (Fig. 9A) is

29 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 421 moderate to strong, approaching or exceeding 70 (corresponding, in one study, to 95% probability that the clade was real; Hillis & Bull, 1993) at some nodes. Basal portions of the tree were poorly supported and were not upheld in the majority rule consensus topology from the bootstrap analysis. The analysis of the majority (90%) cutoff coding approach generated a single tree of 169 steps (Fig. 9B), a CI of (0.824 rescaled) and a g 1 statistic of (P 0.01). The topology of the ingroup was maintained in this second analysis, with two major clades, each containing two Philippine species and a basal Bornean representative. Although bootstrap support for basal nodes generally was poor (<70), distal portions of the R. signata complex were moderately to well supported in some cases. The Manhattan distance/frequency parsimony coding approach, in which step matrices (Appendix 6) were employed to weight character state transitions between taxa, generated a single tree (Fig. 10A) of steps (effects of weighting scheme removed), a CI of 0.820, RI of and a g 1 of (P 0.01). The ingroup topology of the single shortest tree resembled earlier analyses and the bootstrap tree for this analysis is presented in Fig. 10(B). Two clades were recovered, each with two Philippine species and a single Bornean representative. As before, bootstrap support for portions of the R. signata complex was A E B D A C mangyanum moellendorffi picturata signata similis grandocula baramica luctuosa hosii chalconota erythraea vittigera cancrivora moderate but more basal nodes were relatively unstable. Although R. similis was found to be the sister taxon to R. grandocula in the bootstrap analysis, support for this relationship was poor (<50%); evidence of monophyly for the R. signata + R. similis + R. grandocula clade was, nonetheless, strong. See Appendix 7 for character support. Distance methods employed (UPGMA and Wagner network construction; not shown) resulted in networks with ingroup topologies similar to those presented above. In both analyses, two major clusters were recovered (one with R. picturata, R. moellendorffi and R. mangyanum and the other with R. signata, R. similis and R. grandocula). However, as in the case of the frequency coding method used in parsimony analyses, the position of R. signata varied. In UPGMA networks, R. similis and R. grandocula were most similar to one another while, in the Wagner network, R. signata and R. similis clustered together (Brown, 1997). DIAGNOSIS OF THE RANA SIGNATA COMPLEX In this section we provide a general diagnosis of the Rana signata complex. Members of the clade can be distinguished from other SE Asian ranid frogs by the combination of the following morphological characters: (1) moderate adult body size (30 70 mm); (2) first B 100 magnus Figure 10. The single tree generated by (A) the Manhattan distance/frequency parsimony approach and (B) the 50% majority rule consensus tree from the bootstrap analysis. Branch support between labelled nodes (in A) is presented in Appendix 7.

30 422 R. M. BROWN and S. I. GUTTMAN finger longer than second; (3) tips of digits distinctly expanded, about 1.5 times penultimate phalanges; (4) circummarginal grooves present on tips of digits; (5) two metatarsal tubercles present, distinct, medially contacting or slightly separated; (5) vocal sacs internal, paired, subgular; (6) humeral glands present; (7) two velvety nuptial pads joined and extending together from base of digit I to (but not beyond) first subarticular tubercle; (8) venter smooth; (9) coloration dark with dorsolateral spot rows or stripes in most specimens; (10) webbing of toes slightly variable, from 1/2 penultimate phalanges free of web on toes II, III and IV, to toes fully webbed to terminal discs. To facilitate identification of the R. signata complex members, Table 11 presents a summary of the distribution of variation in diagnostic morphological characters for each species. In the following section we present species accounts, summaries of literature, and diagnoses for each presently known species (except for the presumably valid R. melanomenta; see comments, above, and in Inger, 1954). We also describe the new species from Mindoro, and include a dichotomous key. SPECIES ACCOUNTS AND DIAGNOSES Rana grandocula Taylor 1920 Rana grandocula Taylor 1920 (Type locality = Bunawan, Agusan Province, Mindanao Isl., Philippines); Inger, Rana philippinensis Taylor 1920 (Type locality = Mindanao, Isl., Philippines). Rana yakani Taylor 1920 (Type locality = Abungabung, Basilan Isl., Philippines). Rana signata grandocula: Inger 1954; Rabor & Alcala 1959; Brown & Alcala 1967; Alcala 1986; Ross & Lazell 1991; Alcala & Brown Rana signata similis: Inger 1954 (part; Leyte Isl. populations). Rana signata: Frost 1985 (part). Rana (Pulchrana) grandocula: Dubois 1992; Duellman Diagnosis. A relatively large species of the R. signata complex, (x = 43.6 ± 3.6)mm SVL in males and (x = 53.8 ± 6.7)mm SVL in females; with a variable dorsal pattern: 63.5% of specimens with brown to dark brown dorsum, possessing darker indistinct flecks and blotches (remainder homogeneous brown); dorsolateral lines present (white to light grey; x = 2.1 ± 0.2, mm wide in males; x = 1.3 ± 0.3, mm in females) and complete (76% of specimens), broken (3.6%), or obscured by middorsal colouration and replaced by sharp stratified boundary between lighter middorsal and darker lateral regions (20.4%); dorsolateral lines or stratified dorsolateral boundary continues on snout (but lines do not continue on lateral edges of palpebra), distinguishing light dorsal portions of rostrum from darker loreal regions; skin smooth to finely granular; humeral glands of males small (x = 1.8 ± 0.4; mm), flat, unpigmented; nuptial pads joined in all but 2.8% of male specimens examined; 84.7% of specimens with 3 7 (males: x = 4.5 ± 0.9; females: x = 4.3 ± 0.9) dark brown femoral bars and 2 6 (males: x = 3.8 ± 0.6; females: x = 4.1 ± 0.6) transverse tibial bars (diffuse on some specimens but always complete); supralabial stripe yellow (36.1% of specimens), light brown (25.3%), white (24%), or light grey (14.6%), extending through the subtympanic region only (15.5%), to mid subocular region (49.5%), or to subnarial region (35.0%); loreal region black (62.4%), brown (29.7%), or light grey (7.9%); ventral coloration light grey (90.1%), grey blotched with brown (5.5%), or creamy yellow (4.4%); throat light grey (72.0%), or very dark grey to black (28.0%). Rana grandocula is distinguished from all other species in the R. signata complex by a larger body size and lighter brown (vs. black), more variable colour pattern; it is further distinguished from R. siberu, R. mangyanum, R. picturata and R. moellendorffi by its smaller, flat, unpigmented humeral gland (vs. longer, prominently raised, darkly pigmented), and from the latter two species by smooth to finely granular skin (vs. moderately to coarsely glandular) and by the presence of complete (often faint, occasionally broken) dorsolateral lines or stratified dorsolateral coloration (vs. spot rows or fused, irregular lines); R. grandocula can be distinguished from R. similis, R. siberu, R. moellendorffi, R. signata and R. mangyanum by the absence (vs. presence) of a distinct, bright line crossing lateral edges of palpebra (continuation of the canthal stripe) and further from P. similis and P. siberu by brown dorsal coloration (vs. black with two yellow or red dorsolateral stripes. It is easily distinguished from R. signata, R. picturata and R. moellendorffi by the absence (vs. presence) of distinct white or yellow middoral spots. Further differences in diagnostic alleles, advertisement calls, morphological characters and body proportions are presented in Tables 3, 7, 8, 10 and 11. Comments. Large samples of this species from several islands of the Mindanao Pleistocene Aggregate Island Complex make discussion of interisland variation possible. The population from Camiguin exhibits unusually prominent dorso-lateral folds (possibly an artifact of preservation) and are among the largest of species examined by us. Basilan Island specimens (Taylor s R. yakani paratypes) lack light dorsolateral lines and males are smaller than Mindanao males (Table 4).

31 Table 11. Distribution of diagnostic morphological characteristics in the Philippine and Bornean members of the Rana signata complex. DL = dorsolateral; n = sample size. See text for character definitions R. similis R. grandocula R. signata R. mangyanum R. picturata R. moellendorffi n = 175 n = 307 n = 145 n = 106 n = 184 n = 149 Nuptial pads 80.7% connected 97.2% connected 24.0% connected 92.5% connected 100% connected 97.1% connected 19.3% divided 2.8% divided 76.0% divided 7.5% divided 0% divided 2.9% divided DL stripes 95.2% complete; 76.0% complete; 64.8% complete; 98.1% complete; 0% complete; 0% complete; 4.8% broken stripes 3.6% broken stripes 35.2 % broken stripes 1.9% broken stripes 11.4% broken stripes; 5.4% broken stripes; 20.4% absent 77.7% spot rows; 57.1% spot rows; 10.9% no pattern 37.5% no pattern DL stripe straight, white straight, white straight, wavy, white irregular, yellow irregular, yellow shape, colour to yellow to grey yellow to yellow to orange to orange Dark bars 76.5% present, 84.7% present, 98.6% present, 93.4% present 100% present, 76.0% present, on limbs complete complete complete broken centrally complete complete 23.5% absent 15.3% absent 1.3% absent 6.6% absent 24.0 % absent Middorsal 20.1% grey with dark 63.5% with dark 100% distinct 61.1% with medial 100% distinct 97.1% distinct coloration flecks and blotches; flecks and blotches; round spots projections; round spots round spots; 79.9% homogeneous 36.5% homogeneous and bars 21.9% blotched; and bars 2.9% reticulate very dark grey to black 17.0% homogeneous brown to black Venter 74.6% light grey; 90.1% light grey; 42.5% brown with 57.0% light grey; 46.6% brown with 39.5% light grey; colour 8.6% black; 5.5% blotched brown; distinct white spots; 26.8% yellow; distinct white spots; 37.2% pale yellow; 6.8% pale or white; 4.4% creamy yellow 29.2% solid brown; 8.2% solid brown 39.2% light grey; 17.4% brown with 5.8% creamy yellow; 28.3% light grey 8.0% mottled grey 14.2% solid brown distinct white spots 4.2% blotched brown; and dark brown 5.9% black Throat 63.9% light grey; 72.0% light grey; 44.1% brown with 56.3% light grey; 46.6% brown with 40.1% grey colour 36.1% dark grey 28.0 % very dark distinct white spots; 23.1% yellow; distinct white spots; 30.2% pale yellow brown, or black grey, or black 26.6% solid brown; 20.6% dark brown 38.2% light grey; 9.0% brown 24.3% light grey; 13.1% brown; 20.7% black 5.0% black 2.1% black Labial stripe 62.0% yellow; 36.1% yellow; 30.8% yellow with spots; 93.3% white; 96.9% yellow spots; 98.1% yellow spots colour 36.0% white; 25.3% light brown; 25.0% white with spots; 5.9% yellow; 3.1% solid black 1.9% solid black 2% absent (black) 24.0% white; 15.3% yellow spots only; 0.8% light grey 14.6% light grey 12.4% absent (black); 9.4% white stripe only; 7.1% yellow stripe only Labial stripe 59.3% mid eye; 49.5% mid eye; 61.3% mid eye; 100% mid eye 100% absent 100% absent extends from 38.7% nares; 35.0% nares; 31.5% tympanum only (spots only) (spots only) tympanum to 2.0% absent 15.5% tympanum only 7.2% nares Loreal region 80.3% black; 62.4% black; 100% black 46.3% black; 94.5% black; 97% black colour 12.8% brown 29.7% brown; 46.3% brown; 5.5% brown 1.1% yellow-tan 6.9% light grey 7.9% light grey 7.4% light grey 1.9% brown Middorsal smooth to finely smooth to finely smooth to finely smooth to moderately moderately to moderately to skin texture granular granular granular granular coarsely granular coarsely granular nloaded from by guest on 06 October 2018 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 423

32 424 R. M. BROWN and S. I. GUTTMAN Range. Mindanao, Basilan, Biliran, Bohol, Samar, Leyte, Camiguin, Dinagat and possibly other landbridge islands of the Mindanao Pleistocene Aggregate Island Complex, Philippines (Fig. 4). The extent of this species south-western distribution into the Sulu archipelago is unknown. Rana similis (Günther, 1872) Polypedates similis Günther, 1872a (Type Locality = Laguna del Bay, Luzon Isl., Philippines ). Rana similis: Boulenger, 1882; Taylor, 1920, 1922b; Inger, Rana (Hylorana) signata: Boulenger, Rana signata similis: Inger, 1954; Alcala, 1986; Brown et al., 1996; Alcala & Brown, Rana signata: Frost, 1985 (part). Rana (Pulchrana) similis: Dubois, 1992; Duellman, Diagnosis. A relatively small member of the R. signata complex, SVL mm SVL in males (x = 38.8 ± 2.2) and (x = 51.9 ± 4.7) mm SVL in females; with a distinct and relatively invariant colour pattern: middorsum homogeneous very dark grey to black (79.9% of specimens) or grey with indistinct black flecks and blotches (20.1%); dorsolateral lines bright white to yellow, (x = 1.3 ± 0.2) mm wide in males, (x = 1.4 ± 0.2) mm in females, extending from sacral region, along the body, across lateral edge of palpepebra and canthus and to tip of rostrum; dorsolateral light lines complete in all but 4.8% of specimens examined; skin smooth to finely granular skin; humeral glands of males small, (x = 8 ± 0.4) mm long, flat, unpigmented; nuptial pads divided in 19.3% of specimens examined (remainder joined); 76.5% with 3 6 (males: x = 3.7 ± 0.7; females x = 3.9 ± 0.7) complete transverse black femoral bars and 3 6 (males: x = 3.5 ± 0.6, females: x = 3.9 ± 0.8) complete tibial bars; supralabial stripe yellow (62.0%), white (36.0%), or absent (2.0%), extending from the tympanic region to the mid subocular region (59.3%) or to the subnarial region (38.7%); loreal region black (80.3%), brown (12.9%), or light grey (6.9%). Rana similis is one of only two members of the R. signata complex (the other is R. siberu) that exhibits the generally invariant colour pattern of a full black body with two thin, straight dorsolateral stripes from rostrum to sacral region; R. mangyanum is similar superficially, but possesses thicker, wavy dorsolateral stripes with more variable middorsal coloration and with medial light colour projections extending from, or medially connecting dorsolateral lines in 61.1% of specimens examined; R. similis differs from the distantly allopatric R. siberu by the presence of pigmented (vs. unpigmented) eggs, yellow (vs. red) dorsolateral lines, and smaller humeral glands (vs. highly enlarged); this species is distinguished from R. grandocula by its smaller body size and black (vs. brown) general coloration; it differs from R. signata by the absence of distinct middorsal spots and a lower percentage of specimens with divided nuptial glands (19.3 vs. 76.0); it is further distinguished from R. mangyanum, R. picturata and R. moellendorffi by its smooth to finely granular skin (vs. moderately to coarsely granular), by its smaller, white or unpigmented low humeral gland (vs. larger, prominently raised, darkly pigmented) and from the latter two species by the presence in 95.2% of specimens with complete, thin dorsolateral lines and by the absence (vs. presence) of distinct white or yellow middoral spots. Further differences in diagnostic alleles, advertisement calls, morphological characters, and body proportions are presented in Tables 3, 7, 8, 10 and 11. Range. Luzon, Pollio, Catañduanes and possibly other land-bridge islands of the Luzon Pleistocene Aggregate Island Complex, Philippines (Fig. 4). Rana moellendorffi Boettger Rana moellendorffi Boettger, 1893 (Type locality = Culion Isl., Philippines); Boulenger, 1920; Taylor, 1920; Inger, Rana (Hylarana) signata: Van Kampen, Rana signata moellendorffi: Inger, 1954; Alcala, 1986; Alcala & Brown, Rana signata: Frost, 1985 (part). Rana (Pulchrana) moellendorffi: Dubois, 1992; Duellman, Diagnosis. A relatively small member of the R. signata complex, (x = 38.5 ± 2.5) mm SVL in males and (x = 52.7 ± 5.2) mm SVL in females; with a somewhat variable colour pattern: generally a dark species with yellow spots and irregular markings: 97.1% of specimens examined with body and middorsum black to dark brown with 8 48 (males: x = 17.6 ± 6.6; females: x = 17.7 ± 5.0) distinct, yellow or pale spots, short bars, and irregular markings, in no discernable pattern (37.5% of specimens), or with spots fused into incomplete dorsolateral lines (5.4%) or spot rows (57.1%), occasionally connecting into reticulate pattern (2.9%, black reduced dorsally); when present, canthal strip invariably connects to dorsolateral stripe via complete stripe on lateral edge of palpebra; skin moderately to coarsely granular; humeral glands of males very large, (x = 2.9 ± 0.6) mm long, raised, yellow, or darkly pigmented; nuptial pads joined in all but 2.9% of specimens examined; 76.0% with 2 4 (males: x = 3.2 ± 0.4, females: x = 3.5 ± 0.5) complete, black transverse femoral and 2 4 (males: x = 3.2 ± 0.3, females: x = 3.5 ± 0.5) tibial bars; 24.0% lack femoral and/or tibial bars); when present, trans-

33 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 425 verse hindlimb bars always complete; supralabial region black or very dark brown with 0 6 (males: x = 2.6 ± 1.3; females: x = 2.8 ± 0.8) pale yellow spots or vertical bars (1.9% solid black); loreal region black (97% of specimens), yellowish tan (1.1%) or brown (2.9%); ventral coloration pale grey (39.5%), pale yellow (37.2%), brown with distinct white spots (17.4%) or black (5.9%). Rana moellendorffi differs from R. picturata by its generally more variable shape in dorsal light markings (spots mixed with short bars, other irregular markings, or a reticulate pattern vs. large subcircular spots in R. picturata) and by its tendency towards a lower average number of middorsal (Table 10) and labial spots; R. moellendorffi differs from R. mangyanum by the absence (vs. presence in all but 1.8% of specimens examined) of thick dorsolateral light stripes, by a slightly smaller mean body size (38.5 vs mm in males; 52.7 vs mm in females), by the presence of light labial spots or bars (vs. labial stripe present in 100% of R. mangyanum) and, when present, by transversely complete dark tibial crossbars (vs. broken by light brown central strip on dorsal surface of tibial region when present in R. mangyanum); the species further differs from R. siberu, R. grandocula, and R. similis by the presence (vs. absence) of labial spots or bars, by the presence (vs. absence) of distinct spots, bars, or irregular middorsal markings and, excepting R. siberu, by its larger, pigmented humeral glands; R. moellendorffi differs from R. signata by the absence of dorsolateral stripes, by its usually (97.1%) undivided nuptial pads (vs. 76.0% divided), by (when present) its irregular dorsolateral configuration of sometimes fused spot rows (vs. straight, narrow, seldomly broken dorsolateral stripes), by the absence (vs. presence in some form) of a labial stripe and by its generally more coarsely granular (vs. smooth or finely granular) skin. Further differences in diagnostic alleles, morphological characters, advertisement calls and body proportions are presented in Tables 4, 7, 8, 10 and 11. Range. Balabac, Palawan, Busuanga, Coron, Culion, Caluit and possibly other land-bridge islands of the Palawan Pleistocene Aggregate Island Complex, Philippines (Fig. 4). Rana signata (Günther, 1872) Polypedates signatus Günther 1872b (Type locality = Matang, Sarawak, Borneo, Malaysia). Rana signata: Boulenger, 1882; Frost, 1985 (part); Inger & Stuebing, 1989 (specimen illustrated is R. picturata); Inger & Tan, 1996a; Inger & Steubing, 1997; Chan-ard et al., 1999 (one specimen illustrated may be R. signata; remainder are R. picturata); Inger, Rana (Hylorana) signata: Boulenger, Rana (Hylarana) signata: Van Kampen, 1923 (part). Rana signata signata: Inger, 1954; Inger, 1966, Rana (Pulchrana) signata: Dubois, 1992; Duellman, Diagnosis. A relatively small member of the R. signata complex, (x = 36.5 ± 1.4) mm SVL in males and (x = 50.1 ± 2.6) mm SVL in females: with a generally invariant colour pattern: dorsum black or very dark brown with complete or near complete thin yellow dorsolateral lines from sacral region, along body, across lateral edge of palpebra, to tip of rostrum; middorsal region with 2 23 (males: x = 11 ± 4.5; females: x = 12 ± 3.4) distinct pale to bright yellow spots and short bars (consisting of one or more connected spots); dorsolateral stripes continuous with stripe along lateral edge of palpebra, the latter continuous with complete canthal stripes; skin smooth to finely granular; humeral glands small, (x = 1.5 ± 0.3) mm long, flat to slightly raised, unpigmented; nuptial pads joined in 76% of specimens, divided in 24%; 98.6% of specimens with 3 5 (males: x = 3.6 ± 0.6; females: x = 3.6 ± 0.6) femoral and 2 5 (males: x = 3.6 ± 0.7; females: x = 3.5 ± 0.6) complete, transverse tibial bars; supralabial region highly variable: labial stripe absent (labial region black: 12.4%; labial region spotted: 15.3%), or present (72.3% of specimens examined) and white (9.4%) or yellow (7.1%) or present and preceded anteriorly by 1 5 (males: x = 2.0 ± 1.3; females: x = 2.2 ± 0.9) supralabial spots (yellow in 30.8% of specimens; white in 25.0%); when present, supralabial stripe extends under the tympanum only (31.5% of specimens examined), to mid subocular region (61.3%), or forward to subnarial region (7.2%); loreal region invariably black; ventral coloration brown with distinct subcircular white spots (42.5% of specimens examined), homogeneous brown (29.2%), or light grey (28.4%); throat brown with distinct white spots (44.1% of specimens examined), homogeneous brown (26.6%), light grey (24.3%), or black (5.0%). Rana signata differs from R. picturata, the sympatric congener with which it has often been confused, by its smaller mean body size (36.5 vs mm SVL in males; 50.1 vs in females), the presence of complete or nearly complete thin dorsolateral lines (vs. irregular, thicker fused spot rows), the presence in 72.3% of specimens of a light labial stripe (vs. spots in all R. picturata) a smaller, non-pigmented humeral gland ( mm long in R. signata vs mm and darkly pigmented in R. picturata) and by fewer middorsal spots (2 23 in males and 6 19 in females of R. signata vs and in males and females of R. picturata); it differs from R. grandocula and R. similis by the presence (vs. absence) of

34 426 R. M. BROWN and S. I. GUTTMAN middorsal spots, by a greater percentage (76.0) of divided nuptial pads (vs. 2.8% in R. grandocula and 19.3% in R. similis); R. signata differs from R. siberu by the presence (vs. absence) of middorsal spots, by yellow (vs. red) dorsolateral lines, by the presence of dark transverse limb bars (limbs spotted in R. siberu); from R. moellendorffi and R. mangyanum it is distinguished by its smaller humeral glands ( mm vs mm in R. moellendorffi and mm in R. mangyanum), complete, thin, straight dorsolateral stripes (vs. spot absent, or present in the form of fused or unfused spot rows in R. moellendorffi and present but thick and wavy in R. mangyanum), distinct (vs. absent, diffuse or irregular) middorsal spots and, in the case of R. moellendorffi, by the variable presence (in 72.3% of specimens) of labial stripes (vs. labial spots or vertical bars); R. signata is further distinguished from R. mangyanum by the presence of complete (vs. centrally broken) transverse hindlimb bars when present. Further differences in diagnostic alleles, morphological characters and body proportions are presented in Tables 3, 8, 9, and 10. Range. Southern Thailand, peninsular Malaysia, Sabah and Sarawak (Malaysia); possibly other localities in Borneo and possibly Sumatra (D. Iskandar, pers. comm.; Figs 1 and 4). Rana picturata Boulenger Rana (Hylorana) picturata Boulenger, 1920 (Original localities = Bidi Caves, Sarawak, Kina Balu, Sabah, Borneo (Malaysia) and Barabas (Barabai), Kalimantan, Borneo (Indonesia)) Rana (Hylarana) signata: Van Kampen, 1923 (part). Rana signata signata: Inger, 1954 (part); Inger, 1966 (part); Inger, 1984 (part). Rana signata: Frost, 1985 (part); Manthey & Grossman 1997 (specimens illustrated are R. picturata); Inger & Stuebing, 1989 (specimen illustrated is R. picturata); Chan-ard et al., 1999 (two specimens illustrated are R. picturata; the other may be R. signata); Inger, Rana picturata: Inger & Tan, 1996a; Inger & Steubing, Diagnosis. A moderately sized member of the R. signata complex, (x = 39.2 ± 3.4) mm SVL for males and (x = 55.3 ± 4.7) mm SVL for females; dorsum black or very dark brown with yellow to orange dorsal spots throughout middorsum and forming two dorsolateral spot rows or near complete, thick dorsolateral lines composed of fused spot rows; when present, dorsolateral lines not fused in sacral region but becoming fused anteriorly on body and forming a more cohesive line across lateral edge of palpebra and canthus, to tip of rostrum; middorsal region with (males: x = 29.9 ± 7.2; females: x = 30.9 ± 7.8) distinct spots and short bars consisting of one or more connected spots, usually in combination with many diffuse or faint middorsal spots; skin moderately to coarsely granular; humeral glands of males large, (x = 3.0 ± 0.3) mm in length, raised, darkly pigmented; nuptial pads invariably joined; 3 5 (x = 3.6 ± 0.7) transverse black bars present on femoral segment of hindlimbs of all male specimens examined, 3 4 (x = 3.6 ± 0.5) in females; 3 4 (males: x = 3.2 ± 0.4; females: x = 3.5 ± 0.5) tibial bars present; supralabial region with immaculate black (3.1% of specimens) or with 1 6 (males: x = 4.2 ± 0.6; females: x = 4.4 ± 0.8) cream to yellow spots or vertical labial bars; spots connected in subtympanic and suboccular region to form thick subocular bar in 3.2% of specimens; ventral coloration brown with distinct round white spots (46.6% of specimens examined), light grey (39.2%), or homogeneous brown (14.2%); throat colour brown with white spots (46.6% of specimens examined), light grey (38.2%), homogeneous brown (13.1%), or black (2.1%). Rana picturata qualitatively differs from R. signata by the presence of spot rows or thick dorsolateral lines composed of fused spot rows (vs. distinct, thin, usually complete stripes), more labial spots, the complete absence (vs. 72.3% presence) of labial stripes in all examined specimens, a larger, raised and pigmented (vs. small, flat, unpigmented) humeral gland, and by the presence of (vs. 2 23) middorsal spots; R. picturata differs from R. moellendorffi by its more consistent round shape of middorsal spots (vs. irregular bars and markings), and greater number of middorsal and labial spots; R. picturata differs from R. mangyanum by the presence of distinct middorsal spots (at most, faint middorsal blotches present in R. mangyanum), by the presence of yellow to orange (sometimes fused) dorsolateral spot rows (vs. tan or yellow wavy dorsolateral stripes in R. mangyanum) and labial spots (vs. yellow or white labial stripe in R. mangyanum); R. picturata is distinguished from R. siberu by the presence of yellow or orange (vs. red) dorsolateral coloration, and by transversely barred ulnar, tibial, and femoral limb segments (spotted in R. siberu); R. picturata differs from R. similis and R. grandocula by the presence (vs. absence) of distinct middorsal spots, the presence of dorsolateral spot rows (vs. complete dorsolateral stripes), and by the presence of labial spots (vs. white or yellow labial stripes). Further differences in diagnostic alleles, morphological characters and body proportions are presented in Tables 4, 7, 8, 10 and 11. Comments. One of the syntypes of R. picturata in the Natural History Museum, London (BM )

35 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 427 bears the collectors tag labelled A. Everett on one side and Palawan on the other, suggesting that this specimen may actually be an example of R. moellendorffi from Palawan Island. In fact, this specimen conforms to the diagnosis of R. moellendorffi (above), and was classified with that species in the classification analysis. Cataloged locality data for BM states that the specimen originated on Mt. Kinabalu (along with BM male ), contrary to the Palawan locality data attached to the specimen itself. We are unable to determine if this inconsistency results from a collector s error, ambiguity entered into the British Museum s database during the recataloging of its specimens following World War II, or from some other error. In any case, our data suggest that the type series for R. picturata may contain more than one species and that designation of a lectotype would be advisable at this time. Designation of a lectotype of Rana picturata. Owing to the possibility of locality errors and uncertainty regarding species content of Boulenger s (1920) type series of Rana picturata, we designate a large male specimen, BMNH (collected on Mt. Kinabalu by A. Everett), as the lectotype for the species. This action has the added benefit of restricting the type locality for this species to a well-studied preserve (Mt. Kinabalu, Sabah) where populations of Rana picturata can be easily studied in the future. This specimen has the following measurements (mm): SVL 45.9, HL 17.0, SL 6.8, IOD 4.5, IND 4.5, ED 7.7, TAD 3.6, ET, 1.5, HW 15.2, UAL 8.4, FAL 10.1, FL 18.5, TBL 20.8, TSL 14.3, PL 23.4, ML 14.1, Toe4L 21.8, Toe1L 6.6, Fin1L 8.4, Fin3L 10.3, NPL 5.8, NPW 1.6, Toe4DW 1.1, Fin1DW 1.0, Fin3DW 1.1, DLSW 1.8, MTTL, 2.5, HG 4.4. The specimen exhibits a typical R. picturata colour pattern of a dark body with two yellow dorsolateral stripes composed of fused spot rows, broken on the left in five places and on the right in three places. The specimen is very dark brown with 20 distinct spots of varying sizes between the dorsolateral stripes; it has three bars on the femoral segments of the hindlimbs and four on the tibial segments. Forelimbs are tan with three dark brown bars each. The snout and lateral portions of the head are brown; the specimen has three light spots on the upper lip and the lower lip is spotted white on dark grey background. The ventral surfaces are dark grey with large white spots. The skin is finely rugose on dorsal surfaces of the body, and becomes coarsely rugose laterally. Range. Kalimantan (Indonesia), Brunei, Sabah and Sarawak (Malaysia), and possibly peninsular Malaysia (J. McGuire, L. Grismer, N. Das, and R. Inger, pers. comm.; Figs 1 and 4). Rana siberu Dring, McCarthy & Whitten Rana siberu Dring et al., 1989 (Type locality = Teitei Bulak, Sabeuleleu, Siberut Isl., Mentawai Archipelago, Indonesia); Inger, Rana (Pulchrana) siberu: Dubois, 1992; Duellman, Diagnosis. A relatively small species of the R. signata complex 39.5 mm SVL for the male holotype, 44.9 mm SVL for a female paratype; Dring et al. (1989) report a body size range of mm for three adult males; dorsum black with near complete pale (red in life) dorsolateral lines from tip of rostrum, across canthal region, lateral edge of palpebra, along body, and continuing to sacral region; middorsal spots absent; skin finely to coarsely granular; humeral glands of the holotype highly enlarged, raised, darkly pigmented, 4.5 mm in length (Dring, McCarthy & Whitten report humeral glands of mm); nuptial pads absent; limbs with distinct round spots (some fused into bilobed marking) on black background; lateral portions of body (flanks) with similar spots; labial region of holotype and one paratype black with four yellow spots; loreal region black; ventral colour brown with yellow markings on throat. Rana siberu differs from all members of the R. signata complex by the presence of unpigmented (vs. pigmented) eggs in females, by the absence (vs. presence) of nuptial pads in males, and by the presence of distinct pale spots on the limbs (vs. bars or indistinct blotches in all other species). This species also differs from all other R. signata complex members by the presence of red dorsolateral stripes (vs. white, yellow, pale orange or tan in other species), by extreme reduction in webbing of toes (one phalanx of third and fifth and two to two and a half phalanges of fourth toe are free of web; vs. web nearly complete in all other R. signata complex members) and by a distinct advertisement call (Dring, McCarthy & Whitten, 1989). Range. Siberut Isl., Mentawai island group (Fig. 1), and possibly montane forested regions of Sumatra (Bengkulu, Barisan Selatan, and Jambi; D. Iskandar, pers. comm.), Indonesia. Rana mangyanum, sp. nov. Rana signata similis: Inger, 1954 (part); Brown & Alcala, 1955; Alcala, 1986; Alcala & Brown, Rana signata: Frost, 1985 (part); Alviola et al., Diagnosis. A moderately sized species of the R. signata complex, (x = 40.9 ± 4.6) mm SVL for males and (x = 58.2 ± 4.4) mm SVL for females; dorsum black or very dark brown with complete, thickened pale tan to yellow irregular, wavy, or straight

36 428 R. M. BROWN and S. I. GUTTMAN dorsolateral lines from tip of rostrum, through canthal region, across lateral edge of palpebra, and continuing along body to sacral region; dorsolateral lines (x = 2.2 ± 0.4) in males and (x = 3.2 ± 0.4) in females; middorsum homogeneous dark brown to black (in 17.0% of specimens examined), dark grey or brown with black blotches (21.9%), or with dorsolateral lines with medial projections that transverse (Figs 3 and 11) middorsum in (61.1%); skin finely to moderately granular; humeral glands of males raised, pigmented, (3.4 ± 0.4) mm in length; nuptial pads joined in 92.% of specimens; 2 5 (males: x = 3.3 ± 0.7; females: x = 3.5 ± 0.7) transverse black bars present on femoral and 2 5 (males: x = 3.3 ± 0.6; females: x = 3.5 ± 0.9) bars present on tibial segments of hindlimbs; tibial segment of limbs with incomplete transverse bars (centre dorsal strip of tan or brown interrupting transverse bars along dorsal axis of limb); white or yellow supralabial stripe from subtympanic to subocular region of all specimens examined; ventral coloration light grey (57.0% of specimens examined), yellow (26.8%), homogeneous brown (8.2%), or mottled grey and dark brown (8.0%); throat colour light grey (56.3% of specimens), yellow (23.1%), or homogeneous dark brown (20.6%). Rana mangyanum is readily distinguished from all other members of the R. signata complex by the presence of the thick, irregular dorsolateral light lines (= 3 mm wide in females; wavy and of uneven thickness, often giving rise to transverse markings medially spanning the middorsal region; Fig. 3) and by the presence of incomplete dark bars on the tibial segment Figure 11. Palmar view of manus (ML = 11.5 mm), plantar view of pes (PL = 19.4) and lateral view of head (HL = 15.8) of female paratype TNHC Illustrations by C. Sheil.

37 SE ASIAN RANA SYSTEMATICS AND WALLACE S LINE 429 of the hindlimb (Fig. 3; vs. complete in other species possessing bars); this species differs from R. signata, R. similis and R. grandocula by its much larger, raised and darkly pigmented humeral glands (vs. humeral glands small, flat, unpigmented); it is distinguished from R. moellendorffi, R. signata and R. picturata by the absence (vs. presence) of yellow, orange or white middorsal spots; from R. siberu by virtue of its thick pale cream to yellow dorsolateral light lines (red and much thinner in R. siberu) and by the absence (vs. presence) of distinct spots on the limbs; from R. similis by thicker dorsolateral light lines and more variable middorsal coloration; from R. grandocula by the constant presence of thicker dorsolateral light lines (vs. thin dorsolateral lines, faint or obscured by middorsal coloration). Further differences in diagnostic alleles and morphological characters, advertisement calls and body proportions are presented in Tables 4, 8, 9, and 10. Holotype. PNM 6270 (Fig. 3). Philippines, Mindoro Island, Oriental Mindoro Province, Municipality of Puerto Galera (within 1 km of the border of the Municipality of San Teodoro), Barangay Villaflor (15 km from Puerto Gallera City on Puerto Gallera-Calapan Road), Tamaraw Falls (unnamed river), 150 m above sea level (masl) (Site 34, Figs 4 and 13). Collected 28 June 1995, 19:00 h (1.5 h after sunset) by Renato E. Fernandez. The holotype was alone and calling from a rock (100 cm diameter), in middle of a stream (4 m wide) when captured. Paratypes (numbers in parentheses correspond to localities in Fig. 4 and Appendix 1) (33) CMNH , one mature female, two mature males and one immature female Philippines, Oriental Mindoro Province, Municipality of Calapan, Barangay Lantuyan, base of Mt. Halcon, Dulangan River, collected approximately 19:00 h 8 June 1992 by R. M. Brown and J. F. Barcelona at 90 masl; CAS-SU 16887, , FMNH 95867, six mature males and one mature female, Mindoro Island, Oriental Mindoro Province, Mt. Halcon, Alcate River (15 km W of Victoria City), collected April May 1954 by A. C. Alcala; CAS-SU , , , , 22362, 23523, 22 mature males and three mature females, Mindoro Island, Oriental Mindoro Province, Municipality of Tarogin, 30 km SE of Mt. Halcon, collected 1 April 1963 by Q. Alcala and party, masl; CAS-SU 22175, 22255, mature male and female, same locality but collected 7 April, 1963 by A. Alcala, 30 masl; CAS-SU 22176, 22238, 22357, a mature male and two mature females, same locality but collected 31 March 1963 by A. Alcala below 30 masl; CAS-SU 22239, mature male, same locality but collected 1 March 1963 by A. C. Alcala: masl; CAS-SU 22586, mature male, Oriental Mindoro, Municipality of Tarogin, 30 km S of Calapan, bank of Tarogin river, collected 21 April 1963 by A. C. Alcala, between 30 and 130 masl; CAS-SU 22210, mature male, Mindoro Island, Oriental Mindoro Province, SE slope of Barawanan Peak, E side of Mt. Halcon, collected 11 April 1963 by A. C. Alcala and party, between 600 and 660 masl; (34) TNHC , five males and seven females, same locality as the holotype, collected January 1996 by J. A. McGuire and R. I. Crombie; USNM , nine mature males and one mature female, Oriental Mindoro, Municipality of Puerto Galera, Mt. Alinyalan (SW of Puerto Galera City), 500 masl, collected by C. A. Ross and party; USNM , one immature recently metamorphosed froglet of undetermied sex, five mature males, two immature females, and six mature females, same locality as holotype, collected 19:20 21:30 h, 8 March 1995 by R. I. Crombie and R. Santos; USNM , two mature males and eight mature females, same locality as holotype, collected 20:00 22:30 h, 10 March 1995, by R. I. Crombie and R. Santos; USNM , two mature males, one immature and seven mature females, same locality as holotype, collected 19:00 23:30 h, 15 January 1996, by R. I. Crombie and J. A. McGuire; TNHC , and PNM , one immature recently metamorphosed froglet of undetermined sex and 13 mature males, same locality as holotype, collected 19:00 20:00 h, June, 1995 by R. E. Fernandez. Etymology. The specific epithet is derived from the genitive plural formation of the term Mangyan that collectively refers to the six tribal groups of Mindoro island. Mangyans posses distinctive racial, linguistic and cultural identities that have become increasingly threatened as their forested ancestral homelands have been developed for mining, timber harvesting and agriculture (Lopez, 1976; Kikuchi, 1984; Schult, 1991). We hope that calling attention to Mangyan natural heritage will impart enhanced recognition of their cultural heritage and that both will be preserved for future generations. External description of holotype. The holotype (Fig. 11) is an undissected male in excellent condition: SVL 43.6, HL 18.0, SL 7.5, IOD 4.5, IND 4.5, ED 6.5, TD 4.4, HW 14.6, UAL 8.6, FAL 10.5, FL 20.8, TBL 22.7, TSL 12.6, FTL 21.9, ML 12.2, Toe4L 20.1, Toe1L 5.7, Fin1L 6.0, Fin3L 9.1, NPL 5.0, NPW 1.7, Toe4DW 1.4, Fin1DW 1.2, Fin3DW 1.3, DLSW 2.0, MTL 1.9, HGL 3.1, ETL 1.4; mass 7.1 g. Snout obtusely rounded (Fig. 12), moderately elongate, terminally flattened and caudo-ventrally angled, extending well beyond lower jaw in lateral view; SL/HL = 0.42; head slightly narrower than widest

38 430 R. M. BROWN and S. I. GUTTMAN Figure 12. Stream side habitat at the type locality: Tamaraw Falls, Barangay Villaflor, Municipality of Puerto Galera, Oriental Mindoro, Philippines. point on body; HW/HL = 0.81; HL/SVL = 0.41; canthus rostralis sharply angular; loreal region deeply concave; supralabial region barely visible in dorsal aspect; nares laterally protuberant, nearly terminal on snout; IOD/IND = 1.0; IOD/ED = 69.2; labial region thick and swollen; interorbital region flat, as wide as a single palpebrum; dorsal rostrum flat; eyes large and protuberant, laterally orientated beyond jaw when viewed in ventral aspect; cornea extending beyond lateral edge of palpebrum when viewed from above; pupil horizontally elliptical; tympanum distinct, immediately behind eye; tympanum slightly raised, smaller than eye; TAD/ED = 0.68; supratympanic ridge faintly evident, continuing caudally and ventrally towards the rictus of the jaw; ventral edge of tympanum ventrally in contact with elongate postsymphysial tubercle; post-symphysial tubercle discontinuous, composed of elongate, dorsally curved portion immediately caudal to articulation of jaw and in contact with more caudal rounded portion; latter two structures separated by 0.7 mm hiatus from another oblong tubercle dorsal to insertion of forearm. Four vomerine teeth atop the dentigerous process of vomer; premaxillary and maxillary teeth present; choanae large ( mm), round, widely separated, partially obscured by maxilla when viewed from below; vocal sac apertures round, 0.1 mm in diameter, just medial to the articulation of the jaw; infrasymphysial knob present, bordered laterally by two distinct swellings; suprasymphysial notch present; vocal sacs paired, internal subgular; tongue elongate, rhomboidal, stretching nearly to glottis, free for more than half its length, just covering vocal sac apertures at its widest point, with central terminal notch. Skin on occiput, dorsum and sacral region finely granular (texture very fine but distinct; homogeneous); caudal edge of palpebrum, dorsal postocular region, suboccular region and supralabial areas also finely granular; skin within post-tympanic and supratympanic ridge moderately granular than skin just above and behind ridge; skin in anal and ventral femoral regions coarsely granular; skin on forelimbs, hindlimbs, venter and dorsal surfaces of manus and pes smooth; at 50 skin on lateral and dorsal surfaces of the body with numerous white-tipped asperities, thickest in sacral region, absent on palpebra. Upper arm slender save for presence enlarged humeral glands; forearm musculature hypertrophied, robust; UA/FA = 0.82; FA/ML = 0.86; FA/SVL = 0.24; order of fingers from shortest to longest II I IV III (first and second finger nearly equal in length); FIL/FIIIL = 0.67; fingers without webs (Fig. 12); distal ends of fingers moderately dilated to discs not more than 1.5 times width of penultimate phalanges; FIDW/FIIIDW = 0.92; discs with circummarginal grooves separating dorsal from ventral surfaces; dorsal surface of articulation of ultimate and penultimate phalanges with transverse, rounded, inverted