Systematics of Oreobates and the Eleutherodactylus discoidalis species group (Amphibia, Anura), based on two mitochondrial DNA genes and external morphology

We present morphological and molecular (mitochondrial DNA, mtDNA) evidence supporting the validity and monophyly of the genus Oreobates. This genus also includes members of the former Eleutherodactylus discoidalis species group plus Eleutherodactylus heterodactylus. The presence of prominent conical subarticular tubercles and prominent supernumerary tubercles associated with the axis of fingers and toes, the presence of glandular axillary pads, and the absence of vocal sacs are proposed as morphological synapomorphies. Species of this taxon form a well-supported crown clade in a phylogeny including members of the genera Craugastor and Eleutherodactylus s.l. The sister taxon to Oreobates is the Eleutherodactylus martinicensis series; Oreobates does not appear to be closely related to the Eleutherodactylus binotatus series or to members of the Eleutherodactylus dolops and Eleutherodactylus nigrovittatus species groups. The taxonomic status of all species of Oreobates is reassessed. Hylodes philippi and Hylodes verrucosus are removed from the synonymy of Oreobates quixensis. We redescribe Oreobates cruralis on the basis of the holotype and new material from Bolivia and Peru, and restrict its distribution to the humid forests of the lowlands and adjacent foothills of the Andes, from southern Peru to central Bolivia. Oreobates granulosus is rediscovered, redescribed, and resurrected, on the basis of the examination of the holotype and additional material from Peru. Phylogenetic analyses of partial 16S mtDNA are used to test the independence of lineages (species). The 14 species of Oreobates are distributed from southern Ecuador to northern Argentina. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 152, 737–773.


INTRODUCTION
Research on systematics of Neotropical amphibians has experienced a boost in recent decades (Glaw & Köhler, 1998). Nevertheless, many taxonomic problems remain to be solved. New species are being described each year and many phylogenetic hypotheses are under review (Padial & De la Riva, 2006). Moreover, the degree of exploration is still incomplete for lineages, areas (Kress et al., 1998), and DNA sequences (Vences & Köhler, 2006). Among brachycephalids, 'Eleutherodactylus' are famous for their incomplete and complex taxonomy, together with the difficulties in their identification. This is probably, at least in part, because of the rarity of many species, which leads to low sampling for species descriptions. In order to facilitate taxonomic work, this large genus has been subdivided into several subgenera, series, and species groups (Lynch & Duellman, 1997). More recently, it has been split into several genera (Crawford & Smith, 2005;Frost et al., 2006). However, the taxonomic status and monophyly of most groups and taxa now included in the family Brachycephalidae have not yet been assessed. Examples of this include the Eleutherodactylus discoidalis species group and the genus Oreobates.
The E. discoidalis group was proposed by Lynch (1976), who recognized five species: Eleutherodactylus cruralis (Boulenger, 1902), Eleutherodactylus discoidalis (Peracca, 1895), Eleutherodactylus elassodiscus Lynch, 1973, Eleutherodactylus granulosus (Boulenger, 1903, and Eleutherodactylus nigrovittatus (Anderson, 1945). He defined this group on the basis of a unique combination of traits and treated it as monophyletic. Additionally, he noted some heterogeneity within the group, as evidenced by the presence of pointed digital tips in E. elassodiscus and E. nigrovittatus, in contrast to the rounded tips in the three southern species (E. cruralis, E. discoidalis, and E. granulosus). He also pointed out that the southern species resembled the Brazilian species Eleutherodactylus octavioi of the Eleutherodactylus binotatus species group in having large outer metatarsal tubercles. Later, Lynch (1989) considered E. granulosus a junior synonym of E. cruralis, and split the former E. discoidalis group, leaving it with only the two southern species: E. discoidalis and E. cruralis. Ischnocnema was then considered phylogenetically more primitive, or ancestral to Eleutherodactylus, and those species placed in the E. binotatus group were proposed as the species most closely related to Ischnocnema, and ancestral to the E. discoidalis group (Lynch, 1989). He also proposed that a complex of species, the Eleutherodactylus fitzingeri group of Lynch (1976) and Lynch & Myers (1983) (later divided into the genus Craugastor and the Eleutherodactylus conspicillatus group; Lynch, 1986;Crawford & Smith, 2005) would be advanced in relation to (and/or derived from) the frogs identified as members of the E. discoidalis group. Furthermore, Lynch (1989) also rejected Savage's (1987) hypothesis of relationships that considered Ischnocnema and the Mexican genus Tomodactylus as sister groups.
More recently, molecular phylogenetic analyses placed I. quixensis among Eleutherodactylus and as sister group of the E. binotatus species group ; however, because of the limited taxon sampling, the only supported conclusion was the nonmonophyly of Eleutherodactylus (see also Darst & Cannatella, 2004). Hence, the taxonomic status and phylogenetic relationships of Ischnocnema could not be assessed, other than saying that I. quixensis was close to what we call Eleutherodactylus (a position that was already held by Jiménez de la Espada 1872Espada , 1875. With the resurrection of Oreobates and its restriction to the Andean and Amazonian species, the hypotheses of relationships are even more uncertain. If I. verrucosa is an 'Eleutherodactylus' and Oreobates a valid taxon, which one among the several 'Eleutherodactylus' species groups is the sister clade of Oreobates? Moreover, is Oreobates monophyletic? Furthermore, if we assume no relationships of Oreobates with south-eastern Brazilian Eleutherodactylus (the E. binotatus series), a position held by Caramaschi & Canedo (2006), then we have to look for putatively related groups in the Andes or the Amazon. The E. discoidalis species group seems to be the best candidate. It shares with Oreobates many external morphological features, among them, those proposed by Lynch (1989) as synapomorphies for this group. Moreover, Padial et al. (2005a) already pointed out the difficulties of assigning some specimens from the Andes either to what was then considered Ischnocnema or to the E. discoidalis group. Furthermore, some characteristics of the advertisement call of Oreobates sanctaecrucis were considered similar, but putatively primitive, to those calls characteristic of the E. discoidalis group (Padial et al., 2008). Finally, a species of the Brazilian Shield recently rediscovered, Eleutherodactylus heterodactylus, shows morphological characters of both the E. binotatus and the E. discoidalis groups (Padial & De la Riva, 2005), and its phylogenetic relationships deserve a detailed study (Padial et al., 2008).
None of the previous hypotheses have been tested to date. Hence, the goal of our study is to answer the following questions. 3. Do former members of the E. discoidalis group belong to this group? 4. What is the current species diversity of those taxa? 5. To which group does E. heterodactylus belong?
To answer these questions we apply an integrative taxonomic approach, sensu Dayrat (2005), Will, Mishler & Wheeler (2005), and Padial & De la Riva (2006). We include molecular phylogenetics (using mitochondrial DNA, mtDNA) and classical taxonomic analyses. We also consider Padial et al.'s (2008) analyses on advertisement calls as additional evidence to test our hypothesis. Finally, we provide a taxonomic account, with remarks and redescriptions for several species the taxonomic status of which is unclear or still poorly known.

EXTERNAL MORPHOLOGY
We followed Lynch (1989) and Lynch & Duellman (1997) for morphological and colour characteristics used in the diagnosis and description. A single person (JMP) took measurements with a digital calliper to the nearest 0.01 mm, but to avoid pseudoprecision (Hayek, Heyer & Gascon, 2001), we rounded all measurements to only one decimal point. Abbreviations are as follows: EE, eye-eye distance; EL, eye length (measured horizontally); EN, eye-nostril distance; FA, arm length (from posterior margin of thenar tubercle to distal point of elbow); FL, foot length (from posterior border of inner metatarsal tubercle to tip of fourth toe); HL, head length (from posterior margin of lower jaw to tip of snout); HW, head width (measured at level of rictus); IND, internarial distance; SVL, snout-vent length; TH, thigh length (from vent to knee); TL, tibia length; TYH, tympanic membrane height; TYL, tympanic membrane length. We do not include values of interorbital distance (IOD) and upper eyelid width (EW). Our experience indicates that these parameters are usually of little utility because the preservation condition of specimens highly influences the measurements, and makes it difficult to obtain precise and comparable values. Colour characteristics were noted in life and in alcohol. We determined the age and sexual condition by dissection or observation of external secondary sexual characters. The condition of the trigeminal nerve (see Lynch, 1986) was determined through dissection of the skin above the tympanic area, and through a horizontal cut of the mandibular joint. Museum abbreviations other than cited by Leviton et al. (1985) are: Centro de Biodiversidad y Genética, Universidad Mayor de San Simón, Cochabamba, Bolivia (CBG); Colección Boliviana de Fauna, La Paz, Bolivia (CBF); Museo de Historia Natural Noel Kempff Mercado, Santa Cruz de la Sierra, Bolivia, Amphibian Collection (MNKA, formerly NKA); Museo de Historia Natural de la Universidad Mayor de San Marcos, Lima, Peru (MHNSM), Museo de Historia Natural, Universidad Nacional de San Antonio Abad del Cusco, Peru (MHNC). We include a list of specimens examined in the Appendix.

MOLECULAR PROCEDURES AND PHYLOGENETIC ANALYSES
The origin of sequences, collection numbers, and localities for each terminal and corresponding amplified fragment are detailed in Table 1. Following the recent hypotheses of hyloid relationships (Darst & Cannatella, 2004;Frost et al., 2006), we selected Cryptobranchus sp., Rhinella amboroensis (see Chaparro, Pramuk & Gluesenkamp, 2007), Leptodactylus griseigularis, and Leptodactylus rhodonotus as the outgroup taxa. Sequences not amplified by us were downloaded from the NCBI database (http://www.ncbi.nlm.nih.gov) or were provided by Matt Heinicke and Blair Hedges (Eleutherodactylus dolops and E. elassodiscus). We used the standard phenol-chloroform extraction protocol (Sambrook, Fritsch & Maniatis, 1989), with minor changes, to isolate genomic DNA. Fragments of approximately 591 bp from the mitochondrial gene 16S and approximately 350 bp of cytochrome b (cytb) were amplified after previously described PCR conditions (Hillis, Moritz & Mable, 1996), using the universal primers 16Sar-5′ and 16Sbr-3′, and cytbZ15-5′ and cytbB2-3′, respectively (Hillis et al., 1996;Goebel, Donnelly & Atz, 1999). Amplification PCR products were purified and sequenced by SecuGen SA in an ABI-PRISM 3700 instrument. Sequences were edited in Sequencher 4.6 (Gene Codes Corporation). Alignments were performed using the program CLUSTAL X 1.83.1 (Thompson et al., 1997) under default parameters. Ambiguously aligned regions were removed from the analysis (see below). For Bayesian phylogenetic analyses (Rannala & Yang, 1996) we used MrBayes version 3.2.1 (Huelsenbeck & Ronquist, 2001). The majority rule consensus tree was produced from four Metropolis-coupled Monte Carlo Markov chains (MCMC; Yang & Rannala, 1997); each run used one cold chain (the head chain) and three heated chains (scout chains). The analysis was run simultaneously for 10 million generations. Chain swapping and parameters update rates were monitored at the beginning and the end of the analyses to ensure that tree searches were being improved. Trees were sampled every 1000 generations. Burn-in was evaluated by examination of the standard deviation of split frequencies (> 0.01), to determine at what point the values had reached stationarity. Stationarity was reached after the first 1 000 000 generations, and hence the first SYSTEMATICS OF OREOBATES 739 1000 from 10 001 trees were discarded. Following Crawford & Smith (2005), default priors and conditions were used in all cases. Maximum parsimony (MP) analyses were performed in PAUP* 4.0b10 (Swofford, 1998) using heuristic searches under parsimony (with all characters weighted equally), with tree bisectionreconnection (TBR) branch swapping and 1000 random addition sequence replicates, with no limit imposed on the number of trees searched. In order to obtain estimates of clade support, nonparametric bootstrapping was performed with heuristic searches of 1000 replicate datasets, and 100 random addition sequences per dataset, with no limit imposed on the number of trees searched. Gaps were treated as fifth base in both the MP and the Bayesian analyses (Crawford & Smith, 2005). Phylogenetic analyses are presented in two ways. First, the MP and Bayesian analyses of a 16S plus cytb, with a data-partitioned dataset, included 756 equal-weight characters (427 constant, 85 parsimony uninformative, and 224 parsimony informative, with gaps considered as 'fifth-base') of 14 taxa including outgroups. Leptodactylus griseigularis and L. rhodonotus were used as outgroups for this analysis. We removed 38 and 12 base pairs corresponding to ambiguously aligned regions of the 16S dataset. Second, the MP and Bayesian analyses of the single 16S dataset, with larger taxon sampling, included 479 equal-weight characters (222 constant, 55 parsimony uninformative, and 192 parsimony informative, with gaps considered as 'fifth-base') of 32 taxa (four outgroups -Cryptobatrachus sp., R. amboroensis, L. griseigularis, and L. rhodonotus). Uncorrected pairwise distances for members of Oreobates were calculated in PAUP* 4.0b10 from this later dataset after removing ambiguous regions. This larger analysis, and the uncorrected pairwise distances analysis, aimed to test the independence and divergence of species included in Oreobates and the E. discoidalis group. However, given the limitations of only using mtDNA for phylogenetic inference at the species level (Funk & Omland, 2003), nuclear markers need to be incorporated to complement and/or test our phylogenetic analyses.

RESULTS
Inferred phylogenies and nodal support of Bayesian partitioned analyses of partial 16S and cytb support the monophyly of Oreobates (bayesian posterior probability, bpp = 100), and a group including the E. conspicillatus and Eleutherodactylus martinicensis series (sensu Lynch & Duellman, 1997) (bpp = 100; Fig. 1). Maximum parsimony partitioned analysis of partial 16S and cytb resulted in two equally mostparsimonious trees, also with statistical support for both clades (bootstrap support, bss = 98 and 85 for Oreobates and Eleutherodactylus, respectively; bates granulosus (Fig. 2). Oreobates quixensis and Oreobates lehri are the sister group of these two subclades in the Bayesian and MP analyses (Figs 1, 2). The sister group of Oreobates is the subgenus Eleutherodactylus (including the E. conspicillatus and E. martinicensis series), and the genus Eleutherodactylus turns out to be nonmonophyletic (Fig. 2). The clade that includes species assigned to the Central American genus Craugastor (Crawford & Smith, 2005;Frost et al., 2006) has the E. binotatus series as sister group, with low boostrap support (bss = 88%) and no Bayesian support. The E. dolops and E. nigro-vittatus species groups are not closely related to Oreobates. However, it must be stressed that our taxon and character sampling for all these groups, except our focal groups (Oreobates and the E. discoidalis group), are incomplete. What is intended herein is to answer the five questions posed in the Introduction. The clade Oreobates includes species assigned to the E. discoidalis group by Lynch (1989) and by us (Padial, Gonzáles & De la Riva, 2005b;Padial, Chaparro & De la Riva, 2007;Padial et al., 2008), and members of the genus Oreobates (sensu Caramaschi & Canedo, 2006). It also includes E. heterodactylus Figure 2. Majority rule consensus tree based on Bayesian phylogenetic analyses of partial 16S (c. 590 bp) mitochondrial DNA of some members of the genera Oreobates, Eleutherodactylus, and Craugastor. The numbers above branches are Bayesian posterior probabilities, followed by boostrap support for maximum parsimony topology. Values lower than 0.90 Bayesian posterior probability, or lower than 60 for boostrap, are not depicted. (Miranda-Ribeiro, 1937), which had already been suggested as belonging to the E. discoidalis group (Padial & De la Riva, 2005;Padial et al., 2008). Hence, members of the E. discoidalis species group are now included in the genus Oreobates. Following the Phylocode (Cantino & de Queiroz, 2004), the genus Oreobates can be node-based defined as the least inclusive crown clade including Oreobates discoidalis, O. lehri, O. quixensis, and O. sanctaecrucis. Additionally, differences in the 560-bp fragment of the 16S mtDNA also support the specific status of the species assigned to the former E. discoidalis group and to Oreobates (Table 2), although the value for the pair O. granulosus-O. sanderi is moderate to low (2.8%).
Regarding external morphology, we propose the presence of enlarged conical plantar supernumerary and subarticular tubercles, together with a smooth texture of the skin covering plantar surfaces, as a synapomorphy for Oreobates. Two other putative synapomorphies are the presence of axillary glandular pads and the absence of a vocal sac in males. Harvey & Sheehy (2005)  The genus Oreobates includes small to mediumsized frogs (SVL of males, 20-44 mm; of females, 25-63 mm) with the following characters: snout short; sexual dimorphism in size; cranial crests absent; body robust; limbs moderately long; skin of venter smooth; skin on dorsal surfaces from smooth to tuberculate; skin of plantar surfaces smooth (not considering plantar tubercles); axillary and/or inguinal glandular pads present; discoidal fold conspicuous; dorsolateral folds rudimentary or absent; males with faint or absent nuptial pads, vocal slits present, and vocal sac absent; tympanic membrane and annulus conspicuous; finger I longer or equal to finger II; finger tips usually rounded with reduced, or absent, disc struc-ture, when present only on fingers III and IV, and always with incomplete circumferential grooves and a poorly defined ungual flap; supernumerary and subarticular tubercles present, prominent, subconical to conical, on smooth plantar surface; toe V equal or slightly shorter than toe III, not reaching distal subarticular tubercle of toe IV (condition B sensu Lynch & Duellman, 1997); toes lacking discs; webbing absent; no tubercles on heel or tarsus; subarticular tubercles prominent and conical, supernumerary tubercles either absent or few, and round to prominent and conical; dorsal coloration overall brown, with an occipital W-shaped dark mark, an x-shaped mid-dorsal dark mark, and a broad and oblique dark band at the anterior margin of the flanks; dentigerous process of the vomers short, prominent, almost at the level of choanae or between them; mandibular ramus of the trigeminal nerve passing lateral to the m. adductor mandibulae externus (S condition sensu Lynch, 1986); reproductive mode by terrestrial eggs with direct development (mode 17 of Duellman & Trueb, 1986); advertisement call consisting of pulsed notes (4-32 pulses), and with low dominant frequency (2000-3800 Hz) (Padial et al., 2008).

OREOBATES
Oreobates choristolemma is most similar to O. granulosus, O. lehri, O. quixensis, O. sanctaecrucis, O. sanderi, and O. saxatilis, from which it can be distinguished by the combination of enlarged finger and toe tips with notched ungual flaps, dorsolateral folds, and iris metallic green. Additionally, it can be distinguished (character of other species in parentheses) from O. granulosus by the larger size of females (SVL, (Table 3); from O. lehri by shorter feet, FL/SVL = 50% (60-70%); from O. sanctaecrucis by dorsal coloration without scarlet flecks; from O. simmonsi by lacking ulnar tubercles (abundant, round, small, pungent ulnar tubercles), and head wider than long; from O. zongoensis by having dorsal skin irregularly covered with enlarged warts and granules (densely and homogeneously granular, round, pungent keratinized granules).

Hylodes cruralis Boulenger (1902)
Hylodes cruralis Boulenger, 1902: 396. Holotype: BM 1947.2.15.70 (formerly 1901. Type locality: 'La Paz, Bolivia, 4000 m' (Fig. 4). Lynch (1989) redescribed this species after examining the holotype. He also studied numerous specimens from Bolivia and Peru deposited in several collections. He confronted two problems. One, the type locality of H. cruralis was La Paz, Bolivia, 4000 m a.s.l. (type collected by P. O. Simmons). Lynch considered this locality to be in error, which was subsequently supported by De la Riva (1990), De la Riva (1993), and De la Riva et al. (2000). The second problem was the great variability in the size of adults. Most of these specimens were in very poor preservation condition, and therefore many subtle morphological characters were difficult or impossible to observe or identify with confidence. Moreover, the specimens studied by Lynch came from very different altitudes and habitats. For example, the largest sample available to him was the series AMNH 6060-73, and the locality of procedence 'Juliaca' was also in error, as it lies in the dry altiplano of southern Peru.
De la Riva et al. (2000) previously stated that what was considered E. cruralis could be in reality a composite of more than one species. Some species described recently were similar in external appearance to E. cruralis (O. ibischi and O. madidi), and differences in advertisement calls were pivotal for assessing their distinctness (Reichle, Lötters, & De la Riva, 2001;Padial et al., 2005b;Padial et al., 2008). We studied 181 specimens of E. cruralis from Bolivia and Peru, collected by us as well as those deposited in various collections, including the holotype and almost all specimens studied by Lynch (1989). After O. madidi, O. sanderi, and O. lehri, O. cruralis can be defined with confidence. The redescription of E. cruralis by Lynch (1989) is not only based on the holotype, but on a series of specimens. Furthermore, in our point of view, the series examined by him contain three different species (O. cruralis, O. granulosus, and O. sanderi). Hence, in order to avoid confusion we provide a thorough diagnosis of O. cruralis and a redescription based exclusively on the holotype. The study of the intraspecific variation is based on additional specimens listed in the Appendix.
Diagnosis: A small Oreobates (SVL of adults, 20.3-33.6 mm) characterized as follows: (1) skin on dorsum coarsely shagreened without keratinized granules, texture composed of small, round, low, flat warts, regular in size, only some of them slightly enlarged; warts on flanks slightly larger than those of dorsum; a pair of incomplete dorsolateral folds composed by enlarged warts; venter smooth; posterior surfaces of limbs smooth; discoidal fold present; postrictal glands present; (2) tympanic membrane and annulus distinct, about half the eye length; supratympanic fold weak and short; (3) head large, slightly longer than wide; snout round in dorsal and lateral views; canthus rostralis sinuous in dorsal view, round in profile; (4) cranial crests absent; upper eyelid covered by small granules; (5) dentigerous process of vomers large, situated posteromedial to choanae; (6) males with vocal slits and no nuptial pads; (7) hands with long and slender fingers, first finger longer than second; subarticular tubercles large, prominent,  conical; supernumerary tubercles large, prominent, round to conical, smaller than subarticular tubercles; tips of fingers III and IV truncate, slightly enlarged, lacking circumferential grooves and ungual flaps; lateral fringes and keels on fingers from moderate to absent; (8) ulnar tubercles absent; (9) no tubercles on heel and tarsus; (10) inner metatarsal tubercle ovate to round, prominent; outer metatarsal tubercle smaller, round, prominent; supernumerary tubercles conical, prominent; (11) toes long and slender (foot length 50% of SVL), lateral fringes weak or absent, webbing absent; fifth and third toes reaching midpoint of second subarticular tubercle of toe IV; tips of toes moderately enlarged, rounded, ungual flap not indented; (12) axillary gland present; (13) dorsal coloration pale brown to dark brown or greyish-brown, with W-shaped occipital and mid-dorsal X-shaped dark marks, or a pair of cream dorsolateral stripes and a short longitudinal sacral stripe; throat and chest light grey to dark brown; belly cream with brown mottling or reticulations on anterior margin.
Oreobates cruralis can be distinguished from other Oreobates (characters of other species in parentheses) as follows: from O. choristolemma by lacking keratinized granules on dorsum, smaller size (SVL of the single known adult female, 46.4 mm) (Table 3), and lacking indented ungual flap on finger discs; from O. discoidalis by having warty dorsal skin (finely shagreened with few enlarged warts in some specimens); moderately enlarged and truncate tips on fingers III and IV (enlarged and ovate); dentigerous process of vomers posteromedial to choanae (between choanae); canthus rostralis sinuous in dorsal view and round in lateral profile (straight and sharp); differences in advertisement call and habitat (Padial et al., 2008). From O. granulosus by having slightly enlarged and truncate tips on fingers III and IV (rounded, not enlarged); dorsal skin coarsely shagreened, composed of round low warts, without keratinized granules on dorsum (dorsal skin with low, round, nonpungent keratinized granules and warts); smaller SVL of adult females, 24.9-33.6 mm (SVL, 34.4-39.5 mm), and head longer than wide (equal) ( Table 3); by having numerous, conical, and prominent supernumerary tubercles on feet (supernumerary tubercles low, few, round). From O. heterodactylus by having coarsely shagreened dorsal skin with enlarged warts (smooth); slightly enlarged and truncate tips of fingers III and IV (very enlarged and ovate); numerous, conical, and prominent supernumerary tubercles on feet (supernumerary tubercles low, few, round); advertisement call, and habitat (Padial et al., 2008). From O. ibischi by having tympanum length half or less than half of eye length (more than half of eye length); coarsely shagreened dorsal skin with enlarged warts (smooth to finely shagreened with some enlarged warts); slightly enlarged and truncate tips on fingers III and IV (enlarged and ovate, finger tips two times wider than the digit); numerous, conical, and prominent supernumerary tubercles on feet (supernumerary tubercles low, scarce, round); head longer than wide (wider than long); differences in advertisement call and habitat (Padial et al., 2008). From O. lehri by having slightly enlarged and truncate tips on fingers III and IV (finger tips not expanded); first finger longer than second (finger I equal to finger II); smaller size of adult males and females (SVL, 31.0-39.9 mm) ( Table 3); shorter feet, FL/SVL = 50% (60%); numerous conical and prominent supernumerary tubercles on feet (supernumerary tubercles low, few, round). From O. madidi by having dorsal skin with small, round, uniform warts, with sparse enlarged warts (homogeneously warty, larger warts), slightly enlarged and truncate tips on fingers III and IV (rounded finger tips), and advertisement call (Padial et al., 2005b(Padial et al., , 2008. From O. quixensis by smaller size, mean SVL of males and females, 24.8 and 29.3 mm, respectively (mean SVL of adult males, 39.0 mm, N = 9; mean SVL of adult females, 50.4 mm, N = 14) ( Table 3); coarsely shagreened dorsal skin with enlarged warts (dorsal skin coarsely tuberculate, with enlarged and prominent warts and granules, some of them keratinized); slightly enlarged and truncate tips on fingers III and IV (rounded finger tips). From O. sanctaecrucis by smaller size, mean SVL of males and females, 24.8 and 29.3 mm, respectively (mean SVL of adult males, 35.2 mm, N = 3; mean SVL of adult females, 46.0 mm, N = 3) ( Table 3); coarsely shagreened dorsal skin with enlarged warts (dorsal skin coarsely tuberculate, with enlarged and prominent warts and granules, some of them keratinized); slightly enlarged and truncate disc on fingers III and IV (rounded finger tips). From O. sanderi it differs by smaller size (mean SVL of adult males, 29.2 mm, N = 6; mean SVL of adult females, 36.5 mm, N = 4) ( Table 3); head longer than wide (wider than long); coarsely shagreened dorsal skin with enlarged warts (dorsal skin covered by sparse keratinized granules and some warts); slightly enlarged and truncate tips on fingers III and IV (rounded finger tips). From O. saxatilis by smaller size, mean SVL of females, 29.3 mm (49.0 and 43.7 mm, holotype and paratype, respectively) ( Table 3); coarsely shagreened dorsal skin with enlarged warts (dorsal skin coarsely tuberculate, with enlarged and prominent warts and granules, some of them keratinized); slightly enlarged and truncate disc on fingers III and IV (rounded finger tips). From O. simmonsi by having coarsely shagreened dorsal skin with enlarged warts (densely granular with round, pungent, keratinized granules); slightly enlarged and truncate tips on fingers III and IV (rounded finger tips); ulnar tubercles absent (abundant, round, small). From E. zongoensis SYSTEMATICS OF OREOBATES 751 by having coarsely shagreened dorsal skin with enlarged warts (densely granular with round, pungent, keratinized granules); slightly enlarged and truncate tips on fingers III and IV (rounded finger tips); head longer than wide (wider than long).
Description of the holotype: An adult female (small ovarian eggs) with head slightly longer than wide; snout round in dorsal view, and round to subacuminate in lateral profile; nostrils slightly protuberant, orientated dorsolaterally; canthus rostralis straight in dorsal view, round in frontal profile; loreal region slightly concave, sloping gradually to the lips; lips not flared; upper eyelid without tubercles, but covered by small warts; no cranial crests. Supratympanic fold distinct, thin, short; tympanic membrane and its annulus, distinct; tympanic membrane slightly ovate vertically, its length about two-fifths of eye length; two postrictal glands. Choanae not concealed by palatal shelf of the maxillary arch when roof of mouth is viewed from below; choanae large, round, lateral, separated by distance equal to four times the diameter of a choana; the specimen has one discernible vomerine odontophore, situated posteromedial to and far from the choanae. Skin texture of dorsal surfaces and posterior parts of hind limbs composed of small, low, round warts; ventral surfaces smooth; a W-shaped occipital fold; no dorsolateral folds; discoidal fold evident.
Ulnar tubercles absent; palmar tubercle bifid, flat, prominent; thenar tubercle ovate, prominent, twothirds of the size of the palmar tubercle; supernumerary tubercles large, round, prominent, smaller than subarticular tubercles; subarticular tubercles large, prominent, subconical; tips of fingers I and II round, not enlarged, and those of fingers III and IV moderately enlarged, truncate; basal lateral fringes on fingers I-III; relative length of fingers: II < I Յ IV < III.
Toes long and slender (foot length 50% of SVL); heel and tarsus lacking tubercles or folds; inner metatarsal tubercle round, prominent, slightly larger than outer metatarsal tubercle; outer metatarsal tubercle round, conical; six supernumerary tubercles, small, round to elongate; subarticular tubercles prominent, subconical to conical; toes with basal lateral fringes; toes I and II with tips rounded, not expanded, toes III-V with slightly enlarged and truncate tips; ungual flap not indented; relative length of toes, I < II < V < III < IV; toes III and V reaching penultimate subarticular tubercle of toe IV.
Dorsal surfaces reddish brown, with darker marks including a W-shaped occipital dark brown mark, an interocular bar, two subocular dark brown stripes, and a supratympanic stripe. Arms and hindlimbs light brown with transverse brown stripes; concealed sur-faces of hindlimbs brown; flanks beige with dark brown spots that merge in an oblique wide band posterior to the arm insertion; ventrally cream with fine grey mottling on throat, chest, and anterior margin of belly.  O. granulosus), the variation in size of this species is still considerably high compared with other members of this genus. However, this is the species for which the largest data set has been analysed, and hence such observed variability may be influenced by sample size. Adult males of O. cruralis range from 20.3 to 30.2 mm, and females from 24.9 to 33.6 mm. The sizes of males and females overlap broadly. Another very relevant character is skin texture, but this character is easily diagnosable under a stereomicroscope. The dorsal skin texture of this species varies from scarcely warty (almost smooth) to densely and homogeneously warty with few or abundant enlarged warts. We consider the skin texture to be warty, although it has also been considered as rugose (Boulenger, 1902), shagreened, or smooth to finely areolate (Padial et al., 2005a, b). The relevant distinction is that the dorsal skin of this species lacks granules (i.e. hard, keratinized structures, usually rounded or conical) or tubercles (enlarged, prominent, conical warts). All dorsal warts are always flat, low, and constitute soft structures. As in O. granulosus and O. sanderi, there are two colour morphs. The most commom morph consists of a pale to dark brown or reddish brown dorsum, with pale and dark irregular flecks, a W-shaped mark on occipital region, and a x-shaped mark on the mid-dorsum, one or two broad dark oblique bands on flanks, and dark interocular, labial, and transversal bars on the extremities. This pattern occurs in 60 of 91 (66%) specimens for which the colour pattern was noted. Another 26 specimens (29%) show the following colour pattern: a pale to dark brown or reddish brown dorsum outlined by a pair of pale dorsolateral bands, a sacral stripe, and two wide dark brown stripes on flanks, one anterior to the groin and one posterior to the arm insertion, with the transverse bars on the extremities less evident. The intensity and tonalities of both patterns varies. A third, rare colour pattern, shared only with E. madidi, is the presence of a thin, white (in alcohol, yellow in life) mid-dorsal stripe from snout to vent. This pattern was observed in five specimens (one female and four males). In all three morphs, the ventral pattern is similar: overall cream with fine mottling on throat and chest, mottling varying in density, intensity, and colour, from grey to dark brown. Other variable characters are as follows. Although breeding males of O. cruralis do not have nuptial pads, a nonswollen white region can be observed on the dorsal surface of the thumbs of some individuals. Axillary glands can be present or absent, or present on only one side; finger fringes can be present (weak) or absent; a faint thin mid-dorsal fold is shown in some specimens; the number of supernumerary tubercles on a single toe varies from two to four, and from weak to prominent; the degree of enlargement of finger tips of fingers III and IV varies, although this variation seems to be related to fixation. The most divergent specimens studied by us are those from La Hoyada (MNK A 5577 and ZFMK 72644), which are overall dark greyish-brown, and have conspicuous finger and toe fringes. Specimens from the lowlands of Peru (KU207749, 215461-2) seem to have slightly indented ungual flaps on the fingers. Specimens from higher altitudes are more affected by parasitic subdermal mites (see Wohltmann, Köhler & Martin, 2006), which may lead one to think they have unusual skin structures.  Cei (1987) includes some observations that do not correspond with characters of O. cruralis. The broad range of adult size and intraspecific genetic distances, the broad altitudinal gradient occupied by this species (including several life zones), and the broad latitudinal distribution, together with some differences in advertisement calls (Padial et al., 2008), suggest that the name O. cruralis is perhaps still being applied to more than one species. The best candidates to be recognized as new species are those populations from humid montane forests and cloud forests in the departments of Cochabamba and Santa Cruz, Bolivia, but our data are still inconclusive.

Remarks
Distribution: This species inhabits the lowland rainforests, humid forests, and cloud forests of the Andean foothills from 200 to 2000 m a.s.l., from Department Cusco in southern Peru to Department Santa Cruz in central Bolivia (Figs 4,5). This species also reaches the semideciduous forests of the inter-Andean valleys of central Bolivia. Cei (1987) tentatively cited O. cruralis for Argentina without locality, but Lavilla & Cei (2001) discarded this possibility. Köhler (2000), Reichle et al. (2001), and Padial et al. (2008) described its advertisement call. Some data on its biology and ecology can be found in Köhler (2000), Doan & Arizábal (2002), and Duellman (2005).
Hylodes discoidalis Peracca, 1895 Hylodes discoidalis Peracca, 1895: 24. Lectotype: MZUT An 427.1. Type locality: 'Tucumán (Argentina)' (Fig. 4), coordinates approx. 26°50′S, 65°13′W. Lynch (1989) redescribed this species based on type material, but neither Peracca (1895) nor Lynch designated one of the syntypes as a lectotype. Lynch's (1989) redescription is based on the syntypes BM 1947.2.15.63-65 and several syntypes deposited in the MZUT, Torino, Italy. The lectotype was designated by Gavetti & Andreone (1993) from the MZUT series. Although they mention that the lectotype was in a good state of preservation, the plate provided evidence that the specimen was not very well preserved. Moreover, the description of the lectotype is rather incomplete (for example it does not include data on skin texture, tubercles on plantar surfaces, or the degree of development of finger and toe tips). The BM series that we studied is better preserved and, in order to avoid future misidentification, we herein describe the best-preserved specimen of this series (BM 1947.2.15.63) (Fig. 6E, F), which is in general very similar to the lectotype.
Oreobates discoidalis can be distinguished from O. choristolemma, O. cruralis, O. granulosus, O. lehri, O. madidi, O. quixensis, O. sanctaecrucis, O. sanderi, O. saxatilis, O. simmonsi, and O. zongoensis by having dorsal skin smooth to finely shagreened, and enlarged and ovate finger tips on fingers III and IV. These two characters are shared with O. heterodactylus and O. ibischi. Nevertheless, O. heterodactylus exhibits a pattern of dorsal coloration consisting of arrow-shaped and X-shaped dark dorsal marks, and it occurs on the Brazilian Shield; O. ibischi has moderately enlarged and truncate finger tips, low warts on dorsal skin, and head wider than long. These three species can be further distinguished by differences in advertisement calls (see Padial & De la Riva, 2005;Padial et al., 2008).
Description of BM 1947.2.15.63: An adult female (SVL, 39.9 mm), with head slightly longer than wide; snout subacuminate in dorsal view and round in lateral profile; nostrils slightly protuberant, orientated dorsolaterally; canthus rostralis straight in dorsal view, sharp in frontal profile; loreal region flat; lips not flared; upper eyelid without tubercles or warts; no cranial crests. Supratympanic fold prominent; tympanic membrane and annulus distinct; tympanic membrane large, round, longer than half of the eye length; two postrictal glands on each side of the head. Choanae not concealed by palatal shelf of the maxillary arch when roof of mouth is viewed from below; choanae large, round, lateral, separated by distance equal to seven times the diameter of a choana; dentigerous process of vomers ovate, medial to choanae, slightly separated from each other, but far from choanae. Skin texture of dorsal surfaces smooth to finely shagreened; ventral surfaces smooth; no occipital or dorsolateral folds; discoidal fold conspicuous.
Ulnar tubercles absent; palmar tubercle round, flat, prominent; thenar tubercle ovate, prominent, twothirds of the size of palmar tubercle; supernumerary tubercles large, conical, prominent, smaller than subarticular tubercles; subarticular tubercles large, prominent, conical; tips of fingers I and II round, not enlarged, and those of fingers III and IV very large and truncate; lateral fringes absent; relative length of fingers: II < I Յ IV < III.
Toes moderately long (foot Յ 50% of SVL); heel and tarsus lacking tubercles or folds; inner metatarsal tubercle ovate, prominent, slightly larger than outer metatarsal tubercle; outer metatarsal tubercle large, round, conical; three inconspicuous supernumerary tubercles, small, round; subarticular tubercles prominent, conical; toes with faint basal membrane between toes II and III; toes I, II and V with rounded, not expanded, tips; toes III and IV with slightly enlarged and truncate tips; ungual flap not indented; relative length of toes I < II < V < III < IV; toes III and V reaching penultimate subarticular tubercle of toe IV.
Colour in preservative: Dorsal surfaces reddish brown, with many irregular dark and light (cream, grey, and brown) marks. Tympanic fold almost black; subocular bars dark brown; flanks as dorsum, but with a dark brown, irregular, oblique wide band posterior to the arm insertion; ventrally cream, with intense reddish brown mottling on throat, chest, anterior margin of belly, and proximate surfaces of flanks. Remarks: In the label of the jar containing O. discoidalis specimens catalogued as BM 98. 7.7.1920 there is a hand-written note with the word 'Types'. Nevertheless, these specimens are not types. They were collected by Dr Borelli at San Lorenzo, Province Jujuy, Argentina, and were probably sent to the BM by Peracca with the types [Peracca (1895, 1897) studied Borelli's collections]. San Lorenzo is also the locality of the two syntypes catalogued as NMW 16510 (Häupl, Tiedeman & Grillitsch, 1994). The locality 'Tucumán' of the lectotype designated by Gavetti & Andreone (1993) is quite ambiguous, as it may refer to Province Tucumán or to San Miguel de Tucumán (popularly known as Tucumán). We suspect that these specimens may originate from near the town of San Miguel de Tucumán, probably from the Sierra de San Javier, as do the paralectotypes BM 1947.2.15.63-65. The specimen KU 182814 (see Harvey & Keck 1995: fig. 3) that was reported as O. cruralis is in fact an adult female of O. discoidalis. Cei (1980) provided a basic but accurate description of this species and an inaccurate illustration of the plantar surfaces. He also provides an illustration of the pectoral girdle. Padial et al. (2008) (1992) provided information about the lingual structures in relation to feeding habits. Lavilla & Cei (2001) consider this species as a poorly known taxon.

Hylodes granulosus
Diagnosis: A medium-sized Oreobates (SVL of adults, 22.6-36.9 mm) characterized as follows: (1) skin of dorsum granular, with round keratinized granules and small, sparse, low, flat warts, only some of them slightly enlarged; warts on flanks slightly larger than those of dorsum; occipital W-shaped fold and/or x-shaped fold on mid-dorsum; venter smooth; posterior surfaces of limbs smooth, groin areolate; discoidal fold present; no dorsolateral folds; large postrictal glands; (2) tympanic membrane and annulus distinct, both with length about half of the eye length; supratympanic fold weak and short; (3) head large, as wide as long; snout short, round in dorsal and lateral views; canthus rostralis slightly convex or sinuous in dorsal view, round in profile; (4) cranial crests absent; upper eyelid covered by small warts; (5) dentigerous process of vomers large, oblique, almost in contact, situated posteromedial to choanae; (6) males with vocal slits and nuptial pads; (7) hands with long and slender fingers, first finger equal or slightly shorter than second; subarticular tubercles large, prominent, round to conical; supernumerary tubercles large or small, round to conical, smaller than subarticular tubercles; tips of fingers III and IV round, slightly enlarged, lacking circumferential grooves and ungual flap; lateral fringes and keels on fingers absent; (8) ulnar tubercles absent; (9) no tubercles on heel or tarsus; (10) inner metatarsal tubercle ovate, prominent; outer metatarsal tubercle smaller, round, prominent; supernumerary tubercles inconspicuous, small, or absent; (11) toes long and slender (foot length 50-60% of SVL), lateral fringes weak or absent, webbing absent; toe V reaching the distal margin of the second subarticular tubercle of toe IV, and toe III reaching the proximal margin; tips of toes moderately enlarged, rounded, with ungual flap not indented; (12) axillary, sacral, and inguinal glands present; (13) dorsal coloration pale brown to dark brown with cream flecks, sometimes with W-shaped occipital and/or mid-dorsal X-shaped dark marks, or a pair of cream dorsolateral stripes and a short longitudinal sacral stripe; throat and chest cream with fine brown mottling; belly cream with brown mottling or reticulations on anterior margin.
Oreobates granulosus can be distinguished from other Oreobates (characters of other species in parentheses) as follows: from O. choristolemma by smaller size of females [SVL of the only known adult females, 46.4 mm (holotype) and 41.0 mm (ZFMK 72569)] (Table 3), and lacking indented ungual flap on discs. From O. discoidalis by having warty and granular dorsal skin (finely shagreened with few enlarged warts in some specimens); tips of fingers III and IV rounded (enlarged and ovate); dentigerous process of vomers posteromedial to choanae (between choanae); canthus rostralis convex or sinuous in dorsal view, and round in lateral profile (straight and sharp). From O. cruralis by having slightly rounded, not enlarged, finger tips (enlarged and truncate tips on fingers III and IV); dorsal skin with low, round, nonpungent keratinized granules and warts (dorsal skin coarsely shagreened, SYSTEMATICS OF OREOBATES 755 composed of round low warts, without keratinized granules on dorsum); larger size, mean SVL of adult females, 36.9 mm (mean SVL of adult females, 29.3 mm) and head as long as wide (head longer than wide) ( Table 3); inconspicuous supernumerary tubercles, low, few, round (numerous, conical, and prominent supernumerary tubercles on feet). From O. heterodactylus by having dorsal skin with low, round, nonpungent keratinized granules and warts (smooth); rounded, slightly developed tips of fingers III and IV (enlarged and ovate). From O. ibischi by having tympanum length half or less than half of the eye length (more than half of the eye length); granular and warty dorsal skin (smooth to finely shagreened with some enlarged warts); rounded, slightly developed tips of fingers III and IV (enlarged and ovate discs two times wider than the digit); head as long as wide (wider than long). From O. lehri by having dorsal skin with low, round, nonpungent keratinized granules and warts (homogeneously warty); by smaller size, mean SVL of adult males, 26.6 mm (mean SVL of adult males, 32.1 mm) ( Table 3); shorter feet, FL/SVL = 50-60% (60-70%). From O. madidi by having dorsal skin with low, round, nonpungent keratinized granules and warts (homogeneously warty); head as wide as long (longer than wide); shorter feet, FL/SVL = 50-60% (50%). From O. quixensis by smaller size, mean SVL of males and females, 26.6 and 36.9 mm, respectively [mean SVL of adult males, 39.0 mm, N = 9; mean SVL of adult females, 50.4, N = 14) ( Table 3); dorsal skin with low, round, nonpungent keratinized granules and warts (dorsal skin coarsely tuberculate, with enlarged and prominent warts and granules, some of them keratinized); inconspicuous supernumerary tubercles on feet, low, scarce, round (numerous, conical and prominent supernumerary tubercles on feet). From O. sanctaecrucis by the smaller size of adult females, 36.9 mm (mean SVL of adult females, 46.0, N = 3) ( Table 3); dorsal coloration without scarlet flecks; supernumerary tubercles inconspicuous, scarce (numerous supernumerary tubercles, prominent); longer feet, FL/SVL 50-60% (50%). From O. sanderi by having supernumerary tubercles inconspicuous, scarce (numerous supernumerary tubercles, prominent); longer feet, FL/SVL = 50-60% (FL/SVL = 50%); head as long as wide (wider than long). From O. saxatilis by its smaller size, mean SVL of females 36.9 (adult females, 49.0 and 43.7 mm, holotype and paratype, respectively) ( Table 3); dorsal skin with low, round, nonpungent keratinized granules and warts (dorsal skin coarsely tuberculate, with enlarged and prominent warts and granules, some of them keratinized); head as long as wide (wider than long). From O. simmonsi by having dorsal skin with low, round, nonpungent keratinized granules and warts (densely granular, round, pungent, keratinized granules); ulnar tubercles absent (abundant, round, small, pungent); head as long as wide (longer than wide). From O. zongoensis by having dorsal skin with low, round, nonpungent keratinized granules and warts (densely granular, round, pungent keratinized granules); slightly enlarged and truncate discs on fingers III and IV (rounded finger tips); head as long as wide (wider than long).
Description of the holotype: An adult male with vocal slits and single faint white, glandular, nonspinous nuptial pad on each thumb; head as long as wide; snout round in dorsal view and lateral profile; nostrils small, orientated dorsolaterally; canthus rostralis slightly convex in dorsal view, round in frontal profile; loreal region slightly concave, sloping gradually to the lips; lips not flared; upper eyelid without tubercles but covered by small granules; no cranial crests. Supratympanic fold indistinct; tympanic membrane and annulus distinct; tympanic membrane nearly round, its length about half of the eye length; two or three postrictal glands. Choanae not concealed by palatal shelf of the maxillary arch when roof of mouth is viewed from below; choanae large, round, anterolateral, separated by distance equal to five times the diameter of a choana; dentigerous process of the vomers larger than choanae, oblique, large, prominent, almost in contact, situated posteromedial to choanae, bearing vomerine teeth; vocal slits short, placed posterolaterally. Skin of dorsal surfaces and posterior parts of hind limbs granular, with round, low, keratinized granules and sparse, low, irregular warts; ventral surfaces smooth; W-shaped occipital fold; no dorsolateral folds; a thin mid-dorsal fold; discoidal fold evident, almost reaching the groin.
Ulnar tubercles absent; palmar tubercle round, large, flat; thenar tubercle ovate, large, prominent, its size around two-thirds of the size of the palmar tubercle; supernumerary tubercles low, round, small; subarticular tubercles round, larger than supernumerary tubercles; finger tips round, slightly enlarged, but almost the same width as corresponding digits; fingers lacking lateral fringes and keels; relative length of fingers: I < IV < II < III.
Colour: In preservative, the dorsal coloration is pale brown to dark brown with cream flecks, a W-shaped occipital dark mark, a pair of faint cream dorsolateral stripes from posterior margin of eyes to occipital region, and a white spot on the middle of the occipital region; head brown with cream flecks, two dark subocular bars; throat and chest cream with fine brown mottling; belly cream with brown mottling or reticulations on anterior margin; flanks cream with fine brown mottling, coalescing in an oblique, broad band posterior to arm insertion; extremities pale brown with diffuse reddish brown transverse bars. Variation: There is scarce variation in the qualitative characters other than colour pattern. Juveniles are almost identical to adults. There is marked sexual dimorphism in size (females are larger than males), and males have longer feet and larger EL in relation to HW and EN (see Table 3). There are two basic colour morphs. One is the colour pattern of the holotype, consisting of a pale to dark brown dorsum, with pale and dark irregular flecks, and a W-shaped dark brown or cream mark on the occipital region, and a x-shaped dark or pale mark on mid-dorsum, a pale sacral stripe, and dark transverse bars on extremities (Fig. 6G). The alternative colour pattern (present on five of 26 specimens) consists of a dark brown dorsum delimited by a pair of pale dorsolateral bands, a sacral stripe and two wide dark brown stripes on flanks, one anterior to groin and one posterior to arm insertion, and transverse bars on extremities that are less evident than in the first colour pattern. Both colour patterns are shared by O. cruralis, O. lehri, and O. sanderi. The intensity of brown mottling on the throat and chest also varies. Glandular pads are present on axillary, groin, and sacral regions, but are sometimes completely absent or absent from only one side. The adult breeding males USNM 299011-12 present glandular pads on axillary, groin, sacral regions, and thumb. It is likely that the development of these glandular pads occurs only during the breeding season. In life, the dorsal texture and folds are also similar to preserved specimens. For example, the specimen MHNC 5328 shows incomplete dorsolateral folds composed of warts, an occipital W-shaped fold a x-mid-dorsal fold, and three enlarged round warts on the dorsal surface of the snout.
Description of colour patterns in life are based on the specimens MHNC 5328 and MHNC 5335 (Fig. 3E, F). The adult female MHNC 5328 showed the follow-ing coloration. Head greenish brown, two subocular and one labial irregular dark brown bars, canthus and loreal region with the same coloration; tympanic fold dark brown; lower lip barred with broad dark brown bars separated by light yellow stripes; postrictal glands orange; tympanic membrane purplish brown, annulus greenish brown; an interocular dark brown bar. Dorsum reddish brown with greenish brown tonalities; some scattered red and orange warts on sacral and occipital regions; dark brown W-shaped occipital mark interiorly outlined by orange tones; mid-dorsum with an irregular X-shaped dark brown mark; sacral region with a transverse dark brown mark that continues towards the groin, forming two oblique bands in the posterior margin of the flanks; flanks light orange to pinkish cream, with some dark brown flecks; a broad irregular, oblique band on the anterior and posterior margins of each flank; groin orange. Arms greenish brown with irregular, transverse bands; arm insertion orange; limbs greenish brown with well-defined transverse dark brown bands and orange granules; plantar surfaces dark brown with grey tubercles. Throat purple, with irregular yellow spots; chest purple; belly yellowish; ventral and posterior surfaces of limbs orange. Iris coloration composed of metallic green, yellow, and orange tones, and black reticulations. Pupil horizontal, black, with a vertical black stripe ventrally. The adult male MHNC 5335 showed the following coloration. Head brown with two subocular and one labial irregular dark brown bars, canthus and loreal region with the same coloration; tympanic fold black; lower lip barred with broad dark brown bars separated by light yellow stripes; postrictal glands orange; tympanic membrane purplish brown, annulus brown; an interocular dark brown bar outlined anteriorly with two cream spots. Dorsum dark reddish brown with some scattered orange warts on sacral and occipital regions; dark brown W-shaped occipital mark interiorly outlined by orange; a pair of bold black spots in the dorsolateral region; sacral region with a thin transverse black stripe that continues toward the groin, forming two oblique bands on the posterior margin of the flanks; flanks brown with some light flecks; groin orange. Arms and limbs brown with diffuse transverse bars and orange warts. Throat brown, with white spots; belly yellow with white spots on dark brown groin anterolaterally; ventral and posterior surfaces of limbs fleshy orange. Iris bronze with black reticulations. Pupil horizontal, black, with a vertical black stripe ventrally.
Remarks: The original description by Boulenger (1903) is short but accurate. Although he overlooked the difference in relation to the degree of development of supernumerary tubercles between O. granulosus SYSTEMATICS OF OREOBATES 757 and O. cruralis, he did note the pivotal difference in skin texture. He mentioned 'skin slightly rugose' (Boulenger, 1902: 580) for O. cruralis, and 'granulate with small warts' for O. granulosus (Boulenger, 1903: 553). This species, which was known from the holotype only, is now known from 25 additional specimens (see Appendix). BM 1905.5.31.21 corresponds to seven specimens (two small juveniles, two young females, a subadult female, and two adult males) collected at Limbani, Carabaya, Department Puno, Peru, by Ockenden. Boulenger identified them as O. granulosus, but he did not include them as types. BM 1905.5.31.14.20 includes four specimens (two juveniles, a subadult male, and a subadult female) from Agualani, Carabaya, Department Puno, Peru. BM 1907.5.7.17-18 correspond to an adult female and a subadult female, respectively. Both specimens were collected by Ockenden very close to the type locality.
Diagnosis: A small Oreobates (SVL of adults, 24.6-30.8 mm) characterized as follows: (1) skin on dorsum finely shagreened to smooth; venter smooth; posterior surfaces of limbs smooth; discoidal fold present; postrictal glands present; (2) tympanic membrane and annulus distinct, its length about two-thirds of the eye length; supratympanic fold well developed; (3) head longer than wide; snout slightly pointed in dorsal view, round in lateral profile; canthus rostralis sharp, slightly concave; (4) cranial crests absent; upper eyelid tubercles absent; (5) dentigerous process of the vomers medial to choanae; (6) males with vocal slits and a single faint nuptial pad; (7) hands with long and slender fingers, first finger longer than second; subarticular tubercles enlarged and subconical; supernumerary tubercles small, rounded; terminal discs of fingers I and II rounded, not enlarged, those of fingers III and IV markedly enlarged, ovate; lateral fringes and keels on fingers absent; (8) between two and four small ulnar tubercles; (9) no tubercles on heel and tarsus; (10) inner metatarsal tubercle small, high, ovoid; outer more rounded, subequal; supernumerary tubercles round to conical; (11) lateral fringes absent, webbing absent; fifth toe shorter than third; tips of toes moderately enlarged, rounded, ungual flap not indented; (12) axillary gland present; (13) dorsal coloration light reddish brown with dark brown to black spots and marks, outlined by cream; in preservative, dorsal regions brownish grey; snout dark grey with darker marks; throat cream with dense and fine brown mottling; venter immaculate.
Oreobates heterodactylus can be distinguished from O. choristolemma, O. cruralis, O. granulosus, O. lehri, O. madidi, O. quixensis, O. sanctaecrucis, O. sanderi, O. saxatilis, O. simmonsi, and O. zongoensis by having dorsal skin smooth to finely shagreened, and enlarged and ovate finger tips on fingers III and IV. These two characters are shared with O. discoidalis. Nevertheless, O. heterodactylus exhibits a pattern of dorsal coloration consisting of arrow-shaped and X-shaped dark dorsal marks, and it occurs on the Brazilian Shield. Oreobates ibischi has moderately enlarged and truncate finger tips, low warts on dorsal skin, and head wider than long. These three species can be further distinguished by differences in advertisement calls (see Padial & De la Riva, 2005;Padial et al., 2008). (2005) and Padial et al. (2008), based on the similarity of external morphology and advertisement calls, already noted that this species could belong to the E. discoidalis group.

OREOBATES
Diagnosis: A small Oreobates (SVL of adults, 31.3-38.7 mm) characterized as follows: (1) skin on dorsum finely shagreened with some low, inconspicuous warts; venter smooth; posterior surfaces of limbs smooth; discoidal fold present; postrictal glands present; (2) tympanic membrane and annulus distinct, its length slightly more than two-thirds of the eye length; supratympanic fold weak; (3) head wider than long; snout round in dorsal view and lateral profile; canthus rostralis slightly convex, round; (4) cranial crests absent; upper eyelid tubercles absent; (5) dentigerous process of vomers medial and almost at the level of choanae; (6) males with vocal slits and a single faint nuptial pad; (7) hands with long and slender fingers, first finger longer than second; subarticular tubercles enlarged and subconical; supernumerary tubercles small, rounded; terminal discs of fingers I and II rounded, not enlarged, but with those of fingers III and IV markedly enlarged and ovate; lateral fringes and keels on fingers absent; (8) ulnar tubercles absent; (9) no tubercles on heel and tarsus; (10) inner metatarsal tubercle small, high, ovoid; outer more rounded, subequal; supernumerary tubercles round, small; (11) lateral fringes absent, webbing absent; fifth toe shorter than third; tips of toes scarcely enlarged, rounded, ungual flap not indented; (12) axillary gland present; (13) dorsal coloration cream to pale brown with dark brown to black spots and undefined marks, outlined by cream; in preservative, dorsal regions pinkish grey with undefined dark brown marks; venter immaculate.
Oreobates ibischi can be distinguished from O. choristolemma, O. cruralis, O. granulosus, O. lehri, O. madidi, O. quixensis, O. sanctaecrucis, O. sanderi, O. saxatilis, O. simmonsi, and O. zongoensis by having dorsal skin finely shagreened and enlarged, and ovate finger tips on fingers III and IV. O. ibischi can be easily distinguished from the most similar species (O. discoidalis, O. heterodactylus, and O. cruralis) (Reichle et al., 2001). Nevertheless, the authors mention that the dorsal skin is finely tuberculate, but following Lynch & Duellman's (1997) terminology the dorsal skin of this species is finely shagreened with some low, inconspicuous warts. Moreover, Reichle et al. (2001) mention the absence of nuptial pads in males of O. ibischi, but re-examination of the adult holotype revealed single, white, glandular nuptial pads on the dorsal surface of each thumb. They distinguished this species from O. discoidalis on the basis of enlarged tips of fingers III and IV, but this character is present in both species (O. discoidalis has more developed finger tips). In contrast, the presence of nuptial pads was mentioned for O. discoidalis, but we were not able to detect this structure in any adult male of O. discoidalis. Moreover, a diagnostic character provided by Reichle et al. (2001)  Distribution: Only known from the semideciduous forests of the type locality and a nearby locality in Central Bolivia (see Appendix and Figs 4,5). It is expected to occur in the semideciduous forests from Department Santa Cruz to Northern Argentina. (PADIAL ET AL., 2007) COMB. NOV. Padial et al., 2007 (Fig. 3I) Eleutherodactylus lehri Padial, Chaparro & De la Riva, 2007: 115. Holotype: USNM 537848. Type locality: 'Apurimac River Valley, Camisea Natural Gas Pipeline, Wayrapata Camp, 2445 m asl (12°50′10′′S, 73°29′43′′W), Department Cusco, Peru' (Fig. 4).

Eleutherodactylus lehri
Diagnosis: A robust Oreobates (SVL of adults, 28.5-39.9 mm) characterized as follows: (1) skin on dorsum coarsely shagreen with enlarged keratinized granules, granules regular in size, small, round, with only some of them slightly enlarged; granules on flanks slightly larger than those of dorsum; venter smooth; posterior surfaces of limbs smooth; discoidal fold present; no dorsolateral folds; postrictal glands weak SYSTEMATICS OF OREOBATES 759 or absent; (2) tympanic membrane and annulus distinct, their length about half of the eye length; supratympanic fold weak, short; (3) head large, slightly longer than wide or subequal; snout round in dorsal and lateral views; canthus rostralis sinuous in dorsal view, round in profile; (4) cranial crests absent; upper eyelid covered by small granules; (5) dentigerous process of vomers large, situated posteromedial to choanae; (6) males with vocal slits; (7) hands with long and slender fingers, first finger about the same length as second; subarticular tubercles round to conical, well developed; supernumerary tubercles low, not prominent, round to conical, smaller than subarticular tubercles; terminal discs of fingers truncate to round, not enlarged, lacking circumferential grooves and ungual flap; lateral fringes and keels on fingers absent; (8) ulnar tubercles absent; (9) no tubercles on heel and tarsus; (10) inner metatarsal tubercle ovate to round, prominent; outer smaller, round, prominent; supernumerary tubercles absent; (11) toes long and slender (foot length 60-70% of SVL), lacking lateral fringes or keels (or very weak at the base of toe III), webbing absent; fifth and third toes reaching midpoint of second subarticular tubercle of toe IV; tips of toes moderately enlarged, rounded, with ungual flap not indented; (12) axillary glands present; (13) dorsal coloration pale brown to dark brown or grayish brown, with darker marks and bold spots; a pair of cream dorsolateral stripes and a short longitudinal sacral stripe present in some specimens; snout dark brown with darker bars; throat and chest light grey to dark brown; belly cream with brown mottling or reticulations on anterior margin.
Oreobates lehri can be distinguished from all other species of the genus by the combination of long feet, fingers I, II, and IV of the same length, the absence or weakly developed supernumerary tubercles on the feet, round and not enlarged finger tips, and granular dorsal skin. In most species of Oreobates the foot length = 50% of the SVL, whereas in O. lehri the foot length Ն 60% of the SVL. For additional differences refer to Padial et al. (2007).
Distribution: Oreobates lehri is known from the type locality (Fig. 4) and from two localities in the Cosñipata Valley, Cusco, Peru. Both localities are situated in cloud forests (Fig. 5) and are separated by a distance of approximately 200 km. Nothing is known about its natural history.  (Fig. 4).
Diagnosis: A small Oreobates (SVL of adults, 27.9-33.0 mm) characterized as follows: (1) skin on dorsum heavily warty, warts regular in size, low, small, rounded, some of them enlarged on the posterior part of the head and scapular region; warts on flanks larger and more elongate than those of dorsum; venter smooth; posterior surfaces of limbs uniformly warty; discoidal fold present, barely visible; no dorsolateral folds; postrictal glands well developed; (2) tympanic membrane and annulus distinct, about half of the eye length; supratympanic fold weak, short; (3) head large, longer than wide; snout round to subacuminate in dorsal and lateral views; canthus rostralis rounded; (4) cranial crests absent; upper eyelid tubercles absent; (5) dentigerous process of vomers large, situated posteromedial to choanae; (6) males with vocal slits and a single faint nuptial pad on thumb; (7) first finger longer than second; subarticular tubercles enlarged and subconical to conical in profile; supernumerary tubercles small; terminal discs of fingers rounded, not enlarged; tips of fingers III and IV round; lateral fringes and keels on fingers absent; (8) ulnar tubercles absent; (9) no tubercles on heel and tarsus; (10) inner metatarsal tubercle ovoid, outer more rounded, subequal; (11) toes lacking lateral fringes or keels; webbing absent; fifth toe shorter than third; discs of toes moderately enlarged, rounded; supernumerary tubercles conical, prominent; (12) axillary gland present; (13) dorsal coloration dark brown to black, with diffuse dark marks; in preservative, dorsal regions brownish grey or purplish grey; snout dark grey with darker marks; throat cream with dense and fine brownish grey mottling; venter cream with scarce to dense brownish grey fine mottling.
Oreobates madidi is most similar to O. cruralis. It differs from O. cruralis by having a homogeneously warty dorsal surface, large warts (skin with small, round, uniform warts, smaller warts with sparse enlarged warts), rounded finger tips (slightly enlarged and truncate tips on fingers III and IV), and a different advertisement call (Padial et al., 2005b(Padial et al., , 2008. It can be distinguished from other small Oreobates (O. discoidalis, O. heterodactylus, and O. ibischi) by having rounder finger discs (truncated and enlarged on fingers III and IV); it can be distinguished from larger species of Oreobates by lacking keratinized granules on dorsum.
Distribution: Oreobates madidi inhabits the humid montane forests of the Andean slopes of northern Bolivia, at elevations of 900-1500 m a.s.l. (Figs 4, 5). The species is only known from La Cascada and Serranía Eslabón, two localities in the Department La Paz, separated by a distance of 150 km. Oreobates quixensis Jiménez de la Espada, 1872: 87. Types: not given in the original descriptions. Type locality: San José de Moti, Province Napo, Ecuador (Fig. 4).

OREOBATES
Diagnosis: A large Oreobates (SVL of adults, 35-63 mm) characterized as follows: (1) skin of dorsum granular, with round keratinized granules and small, sparse, prominent, and enlarged warts; occipital W-shaped fold and/or x-shaped fold on middorsum; venter smooth; posterior surfaces of limbs smooth, groin areolate; discoidal fold present; no dorsolateral folds; large postrictal glands; (2) tympanic membrane and annulus distinct, both with length about half of the eye length; supratympanic fold weak and short; (3) head large, wider than long; snout short, round in dorsal and lateral views; canthus rostralis slightly convex or sinuous in dorsal view, round in profile; (4) cranial crests absent; upper eyelid covered by prominent warts and granules; (5) dentigerous process of vomers large, oblique, almost in contact, situated posteromedial to choanae; (6) males with vocal slits, no nuptial pads; (7) hands with long and slender fingers, first finger longer than second; subarticular tubercles large, prominent, round to conical; supernumerary tubercles large, prominent, round to conical, smaller than subarticular tubercles; tips of fingers III and IV round, not enlarged, lacking circumferential grooves and ungual flap; lateral fringes and keels on fingers absent; (8) either one or two ulnar tubercles, round, low; (9) no tubercles on heel or tarsus; (10) inner metatarsal tubercle ovate, prominent; outer metatar-sal tubercle smaller, round, prominent; supernumerary tubercles large, prominent, conical, smaller than subarticular tubercles; (11) toes long and slender (foot length 50-60% of SVL), lateral fringes weak or absent, webbing absent; tips of toes round; (12) axillary glands present; (13) dorsal coloration pale brown to dark brown with purple tonalities and cream flecks, sometimes with W-shaped occipital and/or mid-dorsal X-shaped dark marks, and a short longitudinal sacral stripe; ventral surfaces cream with fine to dense brown mottling or reticulations.
Oreobates quixensis is the largest species of the genus. It differs from all small species of Oreobates by the combination of dorsum with enlarged warts and keratinized granules, head wider than long, and rounded finger discs. Among the medium-sized and large Oreobates, it is most similar to O. saxatilis, but O. quixensis has larger subconical tubercles on dorsum and a more heavily mottled venter.

Remarks:
The original description by Jiménez de la Espada (1872) is very succinct, but the redescription by Jiménez de la Espada (1875) is very accurate. Several authors have subsequently diagnosed this species (e.g. Lynch & Schwartz, 1971;Duellman, 1978;Rodríguez & Duellman, 1994). Lynch & Schwartz (1971) placed O. quixensis in the genus Ischnocnema, and considered Hylodes philippi Jiménez de la Espada, 1875 and Hylodes verrucosus Jiménez de la Espada, 1875 as junior synonyms of O. quixensis. These two Hylodes were not described and were only illustrated (Jiménez de la Espada, 1875). Based on the illustrations, Lynch & Schwartz (1971) Lynch & Schwartz (1971: 107) considered all of the types lost. This status prevailed until this paper (see Frost, 2006). The type locality of O. quixensis was considered 'la provincia de Quijos' by Lynch & Schwartz (1971), following the original description. The type locality of H. verrucosus was considered unknown. For H. philippi they provided the correct former holotype number (MNCN 202) and the corresponding locality of origin: San José de Motí, Provincia Napo, Ecuador. Although Jiménez de la Espada never mentioned catalogue numbers in his publications, the types are preserved in the amphibian collection of the Museo Nacional de Ciencias Naturales de Madrid [see González (2006) for a detailed catalogue of the Espada's collection at the MNCN]. Our examination of the types revealed significant differences between the two Hylodes taxa and O. quixensis. For example, both H. verrucosus and H. philippi lack supernumerary plantar tubercles, a character illustrated in detail by Jiménez de la Espada (1875) for O. quixensis. Moreover, none of the latter species belongs to the genus Oreobates, and both belong to the E. dolops species group (sensu Lynch, 1989). The taxonomic status of these two species will be discussed elsewhere.
Jiménez de la Espada (1875) mentioned 11 specimens of O. quixensis on which he based a more accurate description. In the MNCN there are 14 specimens collected and identified by Jiménez de la Espada. One comes from San José de Moti, Municipio San José de Mote (or Nuevo), Provincia Napo; the other 13 specimens come from Municipio Archidona de Quijos, Provincia Napo, Ecuador. The specimen MNCN 1708 (formerly 330), an adult female from San José de Moti, corresponds to the specimen measured by Jiménez de la Espada (1875: 104). On the reverse of the original catalogue card corresponding to this specimen there is the inscription 'Cotipo' (= cotype); at the back, there is the inscription 'genotipo' (= type of the genus). Moreover, the card states that this specimen was illustrated in Jiménez de la Espada (1875). The original catalogue number 1154 corresponds to 13 specimens (MNCN 1709-21) from Archidona de Quijos, Napo, Ecuador. Nevertheless, Jiménez de la Espada (1875) only mentioned 11 specimens for this locality. González (2006) mentioned that two of these specimens could correspond to H. verrucosus or H. philippi; however, we find them all assignable to O. quixensis.
We consider the specimen MNCN 1708 (formerly 330), an adult female from San José de Moti (from Municipio San José de Mote (or Nuevo), Provincia Napo, Ecuador, collected in June 1885 by Marcos Jiménez de la Espada, the lectotype of O. quixensis ( Fig. 6I-J), and the specimens MNCN 1709-21 (formerly 1154) from Archidona, Ecuador, collected by Marcos Jiménez de la Espada on 4 May 1865, the paralectotypes of O. quixensis. The lectotype corresponds with the specimen measured and illustrated by Jiménez de la Espada (1875).
Distribution: Upper Amazon Basin in Colombia, Ecuador, Peru, Bolivia (Department of Pando), and western Brazil, up to 1000 m a.s.l. (Frost, 2006). The advertisement call was described by Duellman (1978) and Heyer & Gascon (1995), both of which also provide natural history information.  (1) skin of dorsum granular, with round keratinized granules and small, sparse, low, flat warts, some of them slightly enlarged; warts on flanks slightly larger than those of dorsum; occipital W-shaped fold and/or x-shaped fold on mid-dorsum; venter smooth; posterior surfaces of limbs smooth, groin areolate; discoidal fold present; no dorsolateral folds; large postrictal glands;

OREOBATES
(2) tympanic membrane and annulus distinct, both with length about half of the eye length; supratympanic fold weak, short; (3) head large, wider than long; snout short, round in dorsal and lateral views; canthus rostralis slightly convex or sinuous in dorsal view, round in profile; (4) cranial crests absent; upper eyelid covered by small granules; (5) dentigerous process of vomers large, oblique, almost in contact, situated posteromedial to choanae; (6) males with vocal slits, no nuptial pads; (7) hands with long and slender fingers, first finger longer than second; subarticular tubercles large, prominent, round to conical; supernumerary tubercles large or small, round to conical, smaller than subarticular tubercles; tips of fingers III and IV round, not enlarged, lacking circumferential grooves and ungual flap; lateral fringes and keels on fingers absent or weak; (8) single ulnar tubercle, weak; (9) no tubercles on heel or tarsus; (10) inner metatarsal tubercle ovate, prominent; outer metatarsal tubercle smaller, round, prominent; supernumerary tubercles round, low; (11) toes long and slender (foot length 50% of SVL), lateral fringes weak or absent, webbing absent; tips of toes round, with ungual flap not indented; (12) axillary glands present; (13) dorsal coloration pale brown to dark brown with cream flecks, sometimes with W-shaped occipital and/or mid-dorsal X-shaped dark marks; a short longitudinal sacral stripe; throat and chest cream with brown mottling; belly cream with brown mottling; in life, conspicuous red to scarlet spots in axiles, dorsum, and shanks.
Oreobates sanctaecrucis can be distinguished from the rest of the species of Oreobates by the combination of dorsum with warts and keratinized granules, round finger and toe tips, and red to scarlet spots in axiles, dorsum, and shanks.

Distribution: Oreobates sanctaecrucis inhabits the humid montane forests and cloud forests of the Andean slopes of central Bolivia, in Departments
Cochabamba and Santa Cruz, at elevations of 1000-2100 m a.s.l. (Figs 4,5).
Diagnosis: A medium-sized Oreobates (SVL of adults, 28-38 mm) characterized as follows: (1) skin of dorsum granular, with round keratinized granules and small, sparse, low, flat warts, only some of them slightly enlarged; warts on flanks slightly larger than those of dorsum; occipital W-shaped fold and/or x-shaped fold on mid-dorsum; venter smooth; posterior surfaces of limbs smooth, groin areolate; discoidal fold present; dorsolateral folds weak; large postrictal glands; (2) tympanic membrane and annulus distinct, both with length of about half of the eye length; supratympanic fold weak, short; (3) head large, as wide as long; snout short, round in dorsal and lateral views; canthus rostralis slightly convex or sinuous in dorsal view, round in profile; (4) cranial crests absent; upper eyelid covered by small warts; (5) dentigerous process of vomers large, oblique, almost in contact, situated posteromedial to choanae; (6) males with vocal slits and nuptial pads; (7) hands with long and slender fingers, first finger longer than second; subarticular tubercles large, prominent, round to conical; supernumerary tubercles large or small, round to conical, smaller than subarticular tubercles; tips of fingers III and IV round, not enlarged, lacking circumferential grooves and ungual flap; lateral fringes and keels weak; (8) ulnar tubercles absent; (9) no tubercles on heel or tarsus; (10) inner metatarsal tubercle ovate, prominent; outer metatarsal tubercle smaller, round, prominent; supernumerary tubercles conspicuous, conical; (11) toes long and slender (foot length 50% of SVL), lateral fringes weak or absent, webbing absent; tips of toes round, with ungual flap not indented; (12) axillary glands present; (13) dorsal coloration pale brown to dark brown with cream flecks, sometimes with W-shaped occipital and/or mid-dorsal X-shaped dark marks, or a pair of cream dorsolateral stripes and a short longitudinal sacral stripe; throat and chest cream with fine brown mottling; belly cream with brown mottling or reticulations on anterior margin.
Among the medium-sized and large species of Oreobates, O. sanderi is most similar to O. granulosus, from which it can be distinguished by having conspicuous supernumerary tubercles (weak or absent), shorter feet, FL/SVL = 50% (50-60%), and incipient dorsolateral folds (absent). It can be distinguished from small species of Oreobates by the combination of keratinized granules on dorsum, and fingertips round and not enlarged.

Remarks:
The original description and diagnosis are accurate (Padial et al., 2005a). For distinguishing this species from other members of Oreobates see the operational key below. Its advertisement call is unknown.

OREOBATES SAXATILIS
Diagnosis: A large Oreobates (SVL of adults, 44.5-50.0 mm) characterized as follows: (1) skin of dorsum granular, with round keratinized granules and small, sparse, low, flat warts, only some of them slightly enlarged; warts on flanks slightly larger than those of dorsum; venter smooth; posterior surfaces of limbs smooth, groin areolate; discoidal fold present; no dorsolateral folds; large postrictal glands; (2) tympanic membrane and annulus distinct, both with length of about two thirds of the eye length; supratympanic fold weak, short; (3) head large, wider than long; snout short, round in dorsal and lateral views; canthus rostralis slightly convex or sinuous in dorsal view, round in profile; (4) cranial crests absent; upper eyelid covered by small warts and granules; (5) dentigerous process of vomers large, transverse, situated posteromedial to choanae; (6) vocal slits and nuptial pads condition unknown; (7) hands with long and slender fingers, first finger longer than second; subarticular tubercles large, prominent, round to conical; supernumerary tubercles large, conical, smaller than subarticular tubercles; tips of fingers III and IV round, not enlarged, lacking circumferential grooves and ungual flap; lateral fringes and keels on fingers absent; (8) single round, low, ulnar tubercle; (9) no tubercles on heel or tarsus; (10) inner metatarsal tubercle ovate, prominent; outer metatarsal tubercle smaller, round, prominent; supernumerary tubercles conspicuous, conical; (11) toes long and slender (foot length 50% of SVL), lateral fringes absent, webbing absent; tips of toes rounded, not enlarged, with ungual flap not indented; (12) condition of axillary, sacral, and inguinal glands unknown; (13) dorsal coloration greyish brown with darker brown markings; throat and chest cream with dense brown mottling; belly immaculate.
Among the large Oreobates, O. saxatilis is most similar to O. quixensis, but differs by being smaller and having more scattered warts, and by having an immaculate belly. From all small Oreobates it differs by its larger size and by having keratinized granules on dorsum and non-enlarged fingertips.

Remarks:
The original description and diagnosis are accurate (Duellman, 1990). However, Duellman (1990) mentioned the presence of narrow lateral fringes on fingers and toes, a character that we did not find in the specimens examined. Its advertisement call is unknown; nothing is known about the natural history of this species.
Oreobates simmonsi can be distinguished from all species of Oreobates by the combination of homogeneously granular dorsal surfaces and extremities, venter granular, and long feet with conspicuous supernumerary tubercles.

Remarks:
The original description and diagnosis are accurate. It is based on two specimens (Lynch, 1974). Its advertisement call is unknown. Nothing is known about the natural history of this species. It is only known from the types. The assignment to Oreobates is tentative.

OREOBATES
Oreobates zongoensis is most similar to O. simmonsi, from which it can be distinguished by a head wider than long, smooth skin on arms, and rounded and small supernumerary tubercles. It can be distinguished from other Oreobates species by the combination of homogeneously granular dorsum, head wider than long, supernumerary tubercles round and small, and dorsal and ventral surfaces homogeneously dark brown.

Remarks:
The original description and diagnosis are accurate, even though they are based on a single specimen (Reichle & Köhler, 1997). For distinguishing this species from other members of Oreobates see the operational key below. Its advertisement call is unknown; nothing is known about the natural history of this species. It is only known from the holotype. The assignment to Oreobates is tentative.
Distribution: Known only from the type locality (Fig. 4).

HYPOTHESES OF RELATIONHIPS
Phylogenetic analyses of mtDNA support the hypothesis of monophyly for the genus Oreobates. The monophyly is also supported by the presence of three putative morphological synapomorphies. The genus Oreobates now comprises all members of the former E. discoidalis group (sensu Padial et al., 2007) and Oreobates (sensu Caramaschi & Canedo, 2006). The structure of the advertisement calls in Oreobates also suggests close relationships (Padial et al., 2008). The sister group of Oreobates is a well-supported clade including the E. conspicillatus and E. martinicensis series of the subgenus Eleutherodactylus (Figs 1, 2). However, as E. binotatus, E. dolops, and E. elassodiscus (representing the E. binotatus, E. dolops, and E. nigrovittatus species groups of the subgenus Eleutherodactylus) do not cluster with the E. martinicensis series (which includes the type species of Eleutherodactylus), we consider Eleutherodactylus to be nonmonophyletic. This conclusion is also supported by Darst & Cannatella (2004) and Frost et al.'s (2006) molecular phylogenies. The genus Eleutherodactylus should probably be restricted to the E. martinicensis and E. conspicillatus series. Our data also support the hypothesis of monophyly of Craugastor (Crawford & Smith, 2005), but E. binotatus clusters with no statistical support with this taxon. Hence, either the E. binotatus series is the sister taxon of Craugastor or Craugastor includes members of this series. However, our limited dataset and scarce taxon sampling for these groups do not allow for definitive conclusions.
The most important synapomorphy suggested by us for Oreobates was already proposed by Lynch (1989) for two species (O. cruralis and O. discoidalis): presence of conical subarticular and supernumerary plantar tubercles. Jiménez de la Espada (1875) also described this character for O. quixensis. This character is also present in all other species assigned to Oreobates by Caramaschi & Canedo (2006). Lynch (1989) compared the skull morphology of several putatively related species, but he only selected this synapomorphy to split the former E. discoidalis species group (which by that time included species without conical subarticular tubercles, and without supernumerary tubercles). The two other remaining groups (E. dolops and E. nigrovittatus species groups) were considered very closely related to each other, but not to the E. discoidalis group. This hypothesis is corroborated by our mtDNA analyses (Fig. 2). Although the presence of conical subarticular and supernumerary plantar tubercles may be considered of doubtful phylogenetic utility because of its potential adaptive function, and hence being more likely to result from convergent evolution (Hennig, 1966), the degree of development does not SYSTEMATICS OF OREOBATES 765 seem to be directly related to habitat use. For example, it could be hypothesized that those species with more terrestrial habits or those that live on hard substrates may show the most enlarged tubercles. Nevertheless, species auch as O. cruralis, O. discoidalis, O. heterodactylus, and O. ibischi usually perch on bushes or are found on leaves, yet they present different degrees of development in plantar tubercles. In two terrestrial species, O. granulosus and O. lehri, the supernumerary tubercles of the feet are reduced in number and size. On the other hand, the degree of development of external finger discs seems to be associated to the arboreal life of these frogs. For example, terrestrial and rock-dwelling species such as O. saxatilis, O. quixensis, or O. granulosus have less developed finger discs.
As stated above, Lynch (1971) and Lynch & Schwartz (1971) considered Oreobates a junior synonym of Ischnocnema. More recently, Caramaschi & Canedo (2006) resurrected Oreobates and placed Ischnocnema in the synonymy of Eleutherodactylus. Both Lynch (1971) and Caramaschi & Canedo (2006) used some osteological differences to support their positions, but these characters were neither proposed nor tested as synapomorphies. The most significant difference mentioned by Caramaschi & Canedo (2006) to synonymize Ischnocnema was the presence of 'T-shaped terminal phalanges'. Jiménez de la Espada (1872, 1875) also described the osteology of O. quixensis and stated that the moderate bifurcation of the terminal phalanges was an intermediate character between Hylorhina sylvatica and Hylodes (= Eleutherodactylus). Nevertheless, members of what we consider Oreobates have very well bifurcated T-shaped terminal phalanges (e.g. O. discoidalis and O. heterodactylus).
The presence of conical supernumerary and subarticular tubercles probably influenced the synonymization of Oreobates with Ischnocnema, and led to the hypothesis of the relationship proposed by Lynch (1989), in which Ischnocnema was the ancestor of members of the E. discoidalis group. Indeed, E. verrucosus (formerly I. verrucosa) or E. octavioi (of the E. binotatus series), from south-eastern Brazil, also show prominent conical subarticular tubercles. Nevertheless, the development of plantar tubercles in E. verrucosus and E. octavioi seems to be associated with skin texture. These species have general granular skin texture that extends to the plantar surfaces, which contain some scattered enlarged granules similar to supernumerary tubercles. In the genus Oreobates the supernumerary tubercles are strictly associated with the axis of the phalanges, whereas in E. verrucosus the plantar tubercles are randomly distributed on the plantar surface. Furthermore, although in Oreobates each finger or toe bears between one and three supernumerary tubercles, in E. verrucosus there are only between one and three supernumerary tubercles for the whole plantar surface. Unfortunately, we were not able to test the relationships of Oreobates with E. verrucosus from south-eastern Brazil in our phylogeny. Nevertheless, no synapomorphy is shared by these species and Oreobates to suggest a close relationship. Indeed, the degree of development of the terminal phalanges seems to be associated with the degree of development of disc structures. The development of these structures is also variable within Oreobates, which suggests its limited utility for inferring phylogenetic relationships. Hence, until these hypotheses of relationships can be empirically tested, we agree with the biogeographical argument of Caramaschi & Canedo (2006) that considers the relationship of Oreobates (Andean and Amazonian species) with E. verrucosus (south-eastern Brazil) as doubtful.
Another hypothesis of relationship proposed by Lynch (1989) suggests that members of the E. binotatus series are intermediate between Ischnocnema and members of the E. discoidalis group, and that the latter would be annectant to the E. fitzingeri group (now Craugastor). However, although our molecular phylogeny is insufficient to address this problem, it does not show any close phylogenetic relationship of E. binotatus or Craugastor with the Oreobates clade. Examination of Lynch's (1971Lynch's ( , 1989 illustrations of the skulls of O. quixensis and O. discoidalis reveals a morphology of the vomers similar to that of members of the E. binotatus series and Euhyas. This is the only character that suggests close relationships, but, as Lynch (2001) stated, this character seems to be nonhomologous for both Euhyas and E. binotatus. We extend this assumption to Oreobates. Thus, there is no putative synapomorphy to support this hypothesis.

HYPOTHESES OF ORIGIN
Combining our information about distribution, habitat use, and phylogenetic relationships, we can hypothesize about the origin of the taxa under study (Graham et al., 2004). The highest diversity and endemicity of Oreobates is reached in the Eastern Cordillera of the Central Andes (sensu Gregory-Wodzicki, 2000). This diversity seems to be related to vicariant events as a result of Andean orogeny (Lynch & Duellman, 1997;Gregory-Wodzicki, 2000;Doan, 2003), and/or habitat fragmentation promoted by climatic changes (Weir, 2006;Hughes & Eastwood, 2006). Only three species are found in the lowlands (O. quixensis, O. cruralis, and O. heterodactylus). Oreobates quixensis occurs in the Upper Amazon Basin and invades the Andean foothills, where it reaches altitudes of 1000 m a.s.l. (Frost, 2006). Oreobates cruralis also inhabits the humid forests of the Andean foothills up to 2000 m a.s.l. Some members of the group, such as O. ibischi and O. discoidalis, occupy the semideciduous forests of the inter-Andean dry valleys, from Central Bolivia to northern Argentina, and O. cruralis reach the northern limit of this habitat (close to the Amazonian forests). Padial et al. (2008) found that differences in calls of some Oreobates were related to distribution in a certain habitat, and suggested that these differences could reflect consequences of isolation together with adaptation to a particular environment. Moreover, they found a geographical pattern in call variation along the latitudinal axis of the Andes, and proposed two possible scenarios for explaining these patterns. In the first, the colonization of the Andes by ancestors of the E. discoidalis group could have taken place in southern or central Bolivia, with subsequent expansion to the north (Peru) and to the south (northern Argentina). This hypothesis requires the E. discoidalis group to be related to the E. binotatus group (as hypothesized by Lynch, 1989). As the E. binotatus series is not closely related to Oreobates (Fig. 2), we reject this hypothesis.
In the second hypothesis, members of the E. discoidalis group would have emerged from a common ancestor of O. quixensis or a northern member of the E. binotatus series (Lynch, 1989). Our data discard the possibility of the E. binotatus series being related to Oreobates, but rather we suggest that O. quixensis could be the sister species to the rest of Oreobates (Figs 1, 2). In this case, the origin of Oreobates would be the Upper Amazon Basin or the Andean foothills. Indeed, members of Oreobates are direct-developing frogs that require moist conditions for reproduction, and the greatest species diversity in the genus occurs in humid forests of the Andean slopes from 1000 to 2000 m a.s.l. (Fig. 5). Those species inhabiting drier forests always seem to be restricted to humid microhabitats, such as for example the vegetation close to streams or caves. Moreover, there are two wellsupported clades within Oreobates (Fig. 2), one including former members of Oreobates (robust frogs with marked granular skin and very short, weakly pulsed calls) and another including former members of the E. discoidalis group (small and slender frogs, with warty or shagreened dorsal skin, and longer calls with higher numbers of pulses). These facts suggest two events of diversification, one involving more robust Oreobates species (i.e. O. choristolemma, O. granulosus, O. sanctaecrucis, O. sanderi, and O. saxatilis) in the Andean hills close to the Upper Amazon Basin, with a subsequent colonization to the south, and another diversification of smaller Oreobates species (O. discoidalis, O. cruralis, O. ibischi, and O. madidi) in the Andes from Bolivia to Argentina.
The most isolated species, from other members of Oreobates, is O. heterodactylus, which occurs in two isolated mountain ranges of the Precambrian Brazilian Shield (Heyer & Muñoz, 1999;Padial & De la Riva, 2005). This species inhabits relictual humid forests in small canyons and hills of isolated rocky outcrops. The distribution of this species is interesting because this area lacks any other brachycephalid species. This area lies in between the range of Andean and eastern Brazilian taxa. The occurrence of E. heterodactylus in this area probably made the correct assignment of this species to any species group difficult. For example, Lynch & Myers (1983) placed this species in the former E. fitzingeri group, whereas Lynch & Duellman (1997) included it in the E. binotatus group from the Atlantic Forests, something that Heyer & Muñoz (1999) considered as not making zoogeographical sense. Padial & De la Riva (2005) pointed out that O. heterodactylus resembles species of the E. binotatus (Lynch & Duellman, 1997) and E. discoidalis (sensu Lynch, 1989) species groups. Subsequently, Padial et al. (2008) suggested putative relationships with members of the E. discoidalis group based on similarities in advertisement calls. However, according to our analyses O. heterodactylus clearly belongs to Oreobates. Moreover, our phylogeny supports a close relationship of O. heterodactylus with the clade including putatively advanced Oreobates in the Andes of Bolivia and Argentina. Indeed, O. discoidalis, O. ibischi, and O. heterodactylus are very similar in morphology, and share a similar habitat (semideciduous forests). Therefore, the distribution pattern of these species indicates a broader distribution of Oreobates in the lowlands during more humid periods, and evidences a biogeographical connection between the Andes and the Precambrian Brazilian Shield (see also Pennington, Prado & Pendry, 2000).

CONCLUSIONS
Phylogenetic analyses of mtDNA, together with morphological synapomorphies and bioacoustical similarities, support the monophyly of Oreobates and the nonmonophyly of Eleutherodactylus. Oreobates contains 14 species, including former members of the E. discoidalis species group, E. heterodactylus and species assigned to Oreobates by Caramaschi & Canedo (2006). This genus is the sister group of the E. martinicensis and E. conspicillatus series of the subgenus Eleutherodactylus, and it is not closely related to members of the E. dolops or E. nigrovittatus groups, nor is it closely related to members of the E. binotatus series or E. verrucosus (formerly I. verrucosa). Members of Oreobates occur from Amazonian Colombia to northern Argentina, with an isolated vicariant species in the Precambrian Brazilian Shield SYSTEMATICS OF OREOBATES 767 (O. heterodactylus). The highest diversity of Oreobates occurs in humid montane forests of the Andean Cordillera Oriental (1000-2000 m a.s.l.).
projects REN/GLO 2001-1046 and CGL2005-03156 of the Spanish Ministry of Education and Science (I. De la Riva, Principal Investigator).

NOTE ADDED IN PROOF
While this paper was in press, two important publications appeared (Heinicke, Duellman & Hedges, 2007;Hedges, Duellman & Heinicke, 2008) in which former families Brachycephalidae and Eleutherodactylae are revised. The genus Oreobates is placed in the new family Strabomantidae, and presumably would include also the former species Phrynopus pereger.