Systematics of the Osteocephalus buckleyi species complex (Anura, Hylidae) from Ecuador and Peru

Abstract We present a new phylogeny, based on DNA sequences of mitochondrial and nuclear genes, for frogs of the genus Osteocephalus with emphasis in the Osteocephalus buckleyi species complex. Genetic, morphologic, and advertisement call data are combined to define species boundaries and describe new species. The phylogeny shows strong support for: (1) a basal position of Osteocephalus taurinus + Osteocephalus oophagus, (2) a clade containing phytotelmata breeding species, and (3) a clade that corresponds to the Osteocephalus buckleyi species complex. Our results document a large proportion of hidden diversity within a set of populations that were previously treated as a single, widely distributed species, Osteocephalus buckleyi. Individuals assignable to Osteocephalus buckleyi formed a paraphyletic group relative to Osteocephalus verruciger and Osteocephalus cabrerai and contained four species, one of which is Osteocephalus buckleyi sensu stricto and three are new. Two of the new species are shared between Ecuador and Peru (Osteocephalus vilmae sp. n. and Osteocephalus cannatellai sp. n.) and one is distributed in the Amazon region of southern Peru (Osteocephalus germani sp. n.) We discuss the difficulties of using morphological characters to define species boundaries and propose a hypothesis to explain them.


introduction
The Upper Amazon region has the highest alpha diversity of amphibians in the World with several sites exceeding 100 species in less than 10 km 2 (Bass et al. 2010). Remarkably, these figures may vastly underestimate the total diversity as shown by the discovery of large numbers of cryptic species with the use of genetic markers (e.g., Fouquet et al. 2007;Funk et al. 2011;Ron et al. 2006). These preliminary efforts suggest that the use of genetic characters is crucial to attain a complete understanding of the diversity and evolutionary history of Amazonian amphibians. This necessity is particularly pressing in widespread taxa with pervasive taxonomic problems.
One such group is Osteocephalus, a genus of hylid frogs widely distributed in the Amazon Basin, Guianas and upper drainages of the Magdalena and Orinoco rivers (Frost 2010). Osteocephalus are arboreal and nocturnal frogs with reproduction modes varying from deposition of eggs in lentic water and exotrophic tadpoles to deposition of eggs in bromeliads and oophagus tadpoles and biparental care (Crump 1974;Jungfer and Weygoldt 1999). There are 24 described species and reports of undescribed species are frequent (e.g., Jungfer 2010; Moravec et al. 2009;Ron et al. 2010). There is only one formally defined species group within Osteocephalus, the O. buckleyi complex. It was first proposed by Cochran and Goin (1970) to allocate O. buckleyi (Boulenger 1882), O. pearsoni (Gaige 1929), and O. cabrerai (Cochran and Goin 1970). Its first large scale review was carried out by Trueb and Duellman (1971) who examined the morphology of specimens from seven countries and concluded that the O. buckleyi complex (excluding O. verruciger Werner 1901) consisted of a single, morphologi-cally variable and widely distributed species. They synonymized O. cabrerai, O. carri (Cochran and Goin 1970), and O. festae (Peracca 1904) under O. buckleyi. The three species have been subsequently resurrected (Duellman and Mendelson 1995;Jungfer 2010;Lynch 2006). Recent reviews (Jungfer 2010;2011;Moravec et al. 2009;Ron et al. 2010) imply that the O. buckleyi species complex consists of nine species: O. buckleyi, O. cabrerai, O. carri, O. duellmani Jungfer 2011, O. festae, O. inframaculatus (Boulenger 1882), O. mutabor Jungfer and Hödl 2002, O. verruciger and an undescribed species sister to O. verruciger. A phylogeny based on mitochondrial DNA revealed strong support for the O. buckleyi complex as well as paraphyly in O. verruciger and O. buckleyi (Ron et al. 2010).
Despite recent contributions to the taxonomy of the group (e.g., Jungfer 2010; 2011) the O. buckleyi species complex still contains undescribed species as well as alpha taxonomic problems (Jungfer 2010;Ron et al. 2010) which attest the difficulties of correctly identifying species boundaries on the basis of morphological evidence alone. Herein we integrate genetic, morphological and advertisement call data to assess the phylogenetic relationships and species boundaries among populations of the O. buckleyi complex from Ecuador and Peru. The results demonstrate the existence of three new species, which are formally described here.

Methods
For ease of comparison, we generally follow the format of Trueb and Duellman (1971) for diagnosis and description. Morphological terminology and abbreviations follow Lynch and Duellman (1997). Notation for hand and foot webbing is based on Myers and Duellman (1982). Sex was determined by the texture of dorsal skin, the presence of nuptial pads or vocal sac folds, and by gonadal inspection. Specimens were fixed in 10% formalin and preserved in 70% ethanol. Snout-vent length is abbreviated as SVL. Examined specimens (listed in the type-series and Appendix I) are housed at the collection of the División de Herpetología, Centro de Ornitología y Biodiversidad (CORBIDI), Herpetology Collection at Escuela Politécnica Nacional (EPN-H), Museo de Historia Natural at Universidad San Marcos (MUSM), Museo de Zoología at Pontificia Universidad Católica del Ecuador (QCAZ), and Natural History Museum (BMNH). The pencil drawing of the holotype of O. cannatella sp. n. was made using a Wild Heerbrugg M3B 10×/21 stereo microscope equipped with a camera lucida.
Principal Components Analysis (PCA) and Discriminant Function Analysis (DFA) were used to assess the degree of morphometric differentiation between species. Only well preserved specimens (Simmons 2002) were measured for the following eight morphological variables, following Duellman (1970): (1) SVL; (2) head length; (3) head width; (4) tympanum diameter; (5) femur length; (6) tibia length; (7) foot length; and (8) eye diameter. All variables were log-transformed. To remove the effect of co-variation with SVL, the PCA was applied to the residuals from the linear regressions between the seven measured variables and SVL. We applied a multivariate analysis of variance (MANOVA) to tests for morphometric differences between sexes. Because we found significant differences in O. buckleyi, the PCA and DFA were applied on each sex separately. For the PCA, only components with eigenvalues > 1 were retained. The DFA was applied to the measured variables without size correction because we wanted to assess discriminability among species based on all the variables, including SVL. Sample sizes are: O. buckleyi 24 males, 3 females; O. cabrerai 7 males; O. cannatellai sp. n. 33 males, 3 females; O. festae 7 males, 18 females; O. germani sp. n. 2 males, 5 females; O. verruciger 22 males, 5 females; and O. vilmae sp. n. 4 males. Both PCA and DFA were conducted in JMP® 8.01 (SAS Institute 2008). Measurements were made using digital calipers (to the nearest 0.01 mm).
Advertisement calls recordings were made with a Sennheiser™ ME-67 directional microphone with digital recorder Olympus™ LS10. Calls were analyzed using software Raven 1.2.1 (Charif et al. 2004) at a sampling frequency of 22.1 kHz and a frequency resolution of 21.5 Hz. Calls consist of two components, the first is a rattle note and the second is a quack note. Measured call variables are: (1) call rate: number of calls per second, (2) dominant frequency: frequency with the most energy, measured along all the call, (3) duration of first component note: time from the beginning to the end of note, (4) duration of second component: time from beginning of first quack to the end of the last, (5) first component interval: time from the end of last note of the first component to the beginning of the first note of the second component, (6) number of pulses: number of pulses in a first component note, (7) pulse rate: number of pulses/ duration of first component note, (8) duration of second component note: duration from beginning to end of a single quack, (9) quack rate: number of quacks/duration of second component. If available, several calls or notes were analyzed per individual to calculate an individual average. Original recordings are deposited in the audio archive of the QCAZ and are available through the AmphibiaWebEcuador website (http://zoologia.puce.edu.ec/vertebrados/anfibios/).

DNA extraction, amplification, and sequencing
Total DNA was extracted from muscle or liver tissue preserved in 95% ethanol or tissue storage buffer using standard phenol-chloroform extraction protocols (Sambrook et al. 1989). Polymerase chain reaction (PCR) was used to amplify the mitochondrial genes 12S rRNA ,16S rRNA, ND1 (with flanking tRNA genes), CO1, and control region. We amplified one DNA fragment for 12S, CO1, and the control region and one or two overlapping fragments for the last ~320 bp of 16S and the adjacent ND1 using primers listed in Goebel et al. (1999) and Moen and Wiens (2009). We also amplified the nuclear gene POMC as a single fragment using primers listed by Wiens et al. (2005). PCR amplification was carried under standard protocols. Amplified products were sequenced by the Macrogen Sequencing Team (Macrogen Inc., Seoul, Korea).

Phylogenetic analyses
We estimated phylogenetic relations between species of Osteocephalus based on newly generated sequence data for five mitochondrial (12S RNA, CO1, 16S, ND1, control region) and one nuclear gene (POMC) for a total of up to 4170 bp. To expand the species sampling, we also included sequences from GenBank. All samples are listed in Table 1. For the outgroup, we included one sample of Trachycephalus jordani and one of T. typhonius (based on Faivovich et al. 2005 andWiens et al. 2010). The completeness of the sequences varied considerably among individuals (specially for samples from GenBank which typically lacked three or more loci). Nevertheless, we included samples with missing data because analyses of both empirical and simulated matrices have shown that taxa with missing sequences can be accurately placed in model-based phylogenetic analyses if the number of characters is large, as in our matrix (for a review see Wiens and Morrill 2011).
Preliminary sequence alignment was done with MAFFT 6.814b software with the L-INS-i algorithm (Katoh et al. 2002). The sequence matrix was imported to Mesquite (version 2.72; Maddison and Maddison 2009) and the ambiguously aligned regions were adjusted manually to produce a parsimonious alignment (i.e., informative sites minimized). In protein coding loci, DNA sequences were translated to amino acids with Mesquite to aid the manual alignment. Phylogenetic trees were obtained using Bayesian inference.
Because our dataset includes several loci, it is unlikely that it fits a single model of nucleotide substitution. Thus, we partitioned the data to analyze each partition under a separate model. The best model for each partition was chosen with JModelTest version 0.1.1 (Posada 2008) using the Akaike Information Criterion with sample size correction as optimality measure. We also evaluated three different partition strategies: (i) a single partition, (ii) six partitions (one per loci), and (iii) twelve partitions (one for each codon position in protein coding loci plus one for each non protein coding loci). The best partition strategy was chosen by estimating Bayes factors using a threshold of 10 as evidence in favor of the more complex partition (Brandley et al. 2005).
Each Bayesian analysis consisted of two parallel runs of the Metropolis coupled Monte Carlo Markov chain for 5 × 10 6 generations. Each run had four chains with a temperature of 0.05. The prior for the rate matrix was a uniform dirichlet and all topologies were equally probable a priori. Convergence into a stationary distribution was determined by reaching average standard deviation split frequencies < 0.05 between runs. We also used software Tracer ver. 1.5 (Rambaut and Drummond 2007) to visually inspect convergence and stationarity of the runs. The first 50% of the sampled generations were discarded as burn-in and the remaining were used to estimate the Bayesian tree, posterior probabilities and other model parameters. Phylogenetic analyses were carried out in MrBayes 3.2.1 (Ronquist et al. 2012).
Because the only nuclear gene analyzed had low variability and few informative sites, it was concatenated to the mitochondrial genes into a single matrix. We recognize the advantages of species-tree methods (e.g., Edwards et al. 2007) but could not use them given the insufficient number of nuclear genes sampled. We encourage the application of those methodologies in future phylogenetic inferences in Osteocephalus.

Phylogenetic analyses
Throughout this section, genetic distances are uncorrected p-distances for gene 12S.    Fig. 1).
Populations of O. mutabor segregate latitudinally: the most divergent population (Puerto Bolívar) is the only north of the Napo and Aguarico rivers; the remaining populations are separated in one central and one southern clade, both with strong support. Pairwise genetic distances between populations are below 2% in all comparisons.
The phylogeny recovers a monophyletic O. verruciger (in contrast to Ron et al. 2010) divided in two clades with an unexpected geographic pattern. Loreto and Pacto Sumaco are at a distance of 20-50 km from Cosanga, Río Salado and other nearby localities in central Ecuador (Fig. 2). Yet, in they phylogeny the two samples are sister to samples from Cordillera Kampankis in Peru, at a distance of 370 km. Cordillera Kampankis is an isolated mountain range separated from the rest of the Andes by areas below 500 m above sea level. The records from Cordillera Kampankis are the first confirmed occurrences of O. verruciger in Peru. Genetic distances among O. verruciger samples range between 0 and 1.5%.
Osteocephalus festae samples were collected on both sides of the dry valley of the Marañón River. This valley, with elevations as low as 600 m, is part of the Huancabamba depression, a well-known biogeographic barrier in the Andes. Nevertheless, populations on both sides do not form reciprocally monophyletic groups. In some  Locality data from the literature (Duellman and Mendelson 1995;Jungfer 2010;Peracca 1904;Ron et al. 2010) and specimens deposited at Museo de Zoología of Pontificia Universidad Católica del Ecuador, the Herpetology Collection, Escuela Politécnica Nacional, and Centro de Ornitología y Biodiversidad CORBIDI.  Osteocephalus buckleyi-like individuals are grouped in four clades (A, B, C, and D in Fig. 1). Each clade has unique morphological features (see species descriptions) indicating that each represents a species. The external morphology of the lectotype of O. buckleyi (BMNH 1947.2.13.44, an adult male with nuptial excrescences, Figs 3-4) shows that it belongs to clade A because: (1) its body size (37.90 mm; Fig. 5) is within the range for adult males of Clade A (37.32-45.25 mm, n = 24) but below de range of clades B (48.23-51.85, n = 4) and C (38.47-57.21 mm, n = 24), (2) its relative tympanum size (tympanum diameter/SVL = 0.093; Fig. 5) falls outside the range of Clade C (0.056-0.084, n = 24 males) but within the range of clade A (0.072-0.095, n = 24 males), (3) it has conspicuous tarsal tubercles (absent in clade D), and (4) clade D have a geographic range that, according to the available specimens, does not overlap with the type locality (Canelos, Provincia de Pastaza, Ecuador, 650 m; Figs 2 and 6). Thus, we attach the binomial O. buckleyi to clade A. Clades B, C, and D cannot be assigned to any described species of Samples of O. buckleyi sensu stricto (clade A) have low genetic differentiation (uncorrected p from 0 to 0.7%) despite including localities separated by up to 450 km. As in O. mutabor, the most divergent populations in the phylogeny were those north of the Napo and Aguarico rivers (Cuyabeno and Tarapoa).
Osteocephalus cannatellai sp. nov comprises eight populations with genetic distances ranging from 0 to 1.7%. Populations group latitudinally forming a central and a southern clade. However, one of three samples (CORBIDI 9394) from the southern locality Pongo de Chinim (Kampankis) groups with the central localities.
Clade D comprises five samples from four populations. For two individuals (from Brazil and French Guyana) only GenBank sequences were available and thus we cannot determine if they belong to O. germani sp. n. The three remaining samples (from Peru) are assigned to O. germani.    (Duellman and Mendelson 1995;Jungfer 2010;Peracca 1904;Ron et al. 2010) and specimens deposited at Museo de Zoología of Pontificia Universidad Católica del Ecuador, the Herpetology Collection, Escuela Politécnica Nacional, and Centro de Ornitología y Biodiversidad CORBIDI.
Osteocephalus cannatellai differs from O. vilmae in having a narrower head (relative to SVL, mean male HW/SVL = 0.323, SD = 0.034, n = 33; O. vilmae mean male HW/SVL = 0.355, SD = 0.012, n = 5; differences are significant: t = 2.06, P = 0.046) and a smaller tympanum (relative to SVL, mean male TD/SVL = 0.069, SD = 0.007, n = 33; O. vilmae mean male TD/SVL = 0.087, SD = 0.006, n = 5; differences are significant: t = 5.17, P < 0.001). According to the phylogeny, O. cannatellai and O. vilmae are not sister species (Fig. 1) Head narrower than body, slightly longer than wide; snout truncate in lateral and dorsal views; distance from nostril to eye longer than diameter of eye; canthus rostralis distinct and rounded; loreal region concave; internarial area depressed; nostrils moderately protuberant, directed laterally; interorbital area flat, lateral margins of the frontoparietals inconspicuous through skin; eye large, strongly protuberant; tympanic membrane clearly evident, large, slightly wider than high, about two thirds of eye diameter, separated from eye by ca. 85% of its diameter; tympanic annulus distinct except dorsally where it is covered by supratympanic fold; posterior end of supratympanic fold reaches arm insertion. Arm slender, axillary membrane present, reaching one third of arm length; four small low tubercles present along ventrolateral edge of forearm; relative length of fingers I<II<IV<III; fingers bearing large, oval discs, that of third finger about three fourths of tympanum diameter; subarticular tubercles prominent, round to ovoid, single; supernumerary tubercles present; palmar tubercle small, elongated; prepollical tubercle large, flat, elliptical; prepollex enlarged; large dark keratinous nuptial excrescences covering inner surface of prepollex up to half the distance between subarticular tubercle and proximal border of disk of thumb; webbing formula of fingers I basal II1 2 / 3 -2 2 / 3 III2½-2 + IV. Medium sized to small tubercles on tibiotarsal articulation; scattered tubercles on tarsus, more abundant on outer edge; small tubercles scattered along ventrolateral edge of foot; toes bearing discs slightly wider than long, smaller than those of fingers; relative length of toes I<II<V<III<IV; outer metatarsal tubercle ill defined, small, round; inner metatarsal tubercle large, ovoid; subarticular tubercles single, round, protuberant; supernumerary tubercles restricted to the soles; webbing formula of toes I1-2II1-2 + III1 + -2 + IV2 --1V. Skin on dorsum, head, and dorsal surfaces of limbs smooth, with scattered tubercles; skin on flanks areolate; skin on venter coarsely granular; skin on ventral surfaces of head and thighs granular, those of shanks smooth. Cloacal opening directed posteriorly at upper level of thighs; short simple cloacal sheath covering cloacal opening; round tubercles below vent; two conspicuous white tubercles ventrolateral to vent. Tongue cordiform, widely attached to floor of mouth; dentigerous processes of the vomer angular, adjacent medially, posteromedial to choanae, bearing 12 and 9 (left/right) vomerine teeth; choanae trapezoidal, oblique; vocal sac barely distinct above the arm and below the ear.
Color of holotype in preservative. Dorsum brown with light gray to cream peripheral marks; dark brown, ill defined, transversal bar between orbits (Fig. 8); cream middorsal line from tip of snout to end of sacrum; dorsal surfaces of forearms brown with light gray and dark gray marks, dorsal surfaces of thighs light gray with dark gray transversal bands, dorsal surfaces of shanks and feet brown with dark gray marks. Venter brown with light cream yellowish spots, more abundant on posterior half of the body (Fig.  9); ventral surfaces of hindlimbs and forelimbs brown with dark brown marks and conspicuous white tubercles on forearms; outer half of ventral surfaces of forearms dark brown; sides of head brown with oblique white bar from posteroventral border of orbit to border of jaw, below tympanum (Fig. 10); vertical dark brown bar below eye, anterior to white bar; area behind white bar and eye dark brown except for brown  periphery of tympanum; iris light gray with black reticulations; flanks light gray anteriorly, cream posteriorly, areolate region with gray reticulation.
Color of holotype in life. Based on digital photographs. Dorsum brown with green peripheral marks; dark brown, ill defined, transversal bar between orbits (Fig. 7E); dorsal surfaces of forearms brown with green and dark brown marks, dorsal surfaces of thighs dark green with dark brown transversal bands, dorsal surfaces of shanks and feet brown with dark brown marks. Sides of head brown with oblique lime bar from posteroventral border of orbit to border of jaw, below tympanum; vertical dark brown bar below eye, anterior to lime bar; area posterior to lime bar and eye dark brown except for brown periphery of tympanum; flanks light green, areolate region with dark reticulation.
Etymology. The specific name cannatellai is a noun in the genitive case and is a patronym for David C. Cannatella, who with his research has enriched the understanding of the evolution of Neotropical amphibians. He has also contributed to amphibian studies in Ecuador by providing funding and training to local scientists.
Variation. Variation in dorsal and ventral coloration of preserved specimens is shown in Figures 8 and 9. Dorsal background coloration varies from cream to light gray or brown; irregular dark brown or dark gray marks are always present (Fig. 8). Some specimens have a cream middorsal line from the tip of the snout to the mid sacrum (QCAZ 49570) or the vent (QCAZ 39633). In females, the dorsum always lacks tubercles while in males it varies between lacking tubercles (QCAZ 32508) and having scant and ill-defined non-keratinized tubercles (e.g., QCAZ 48814 and 49569). The prominence of the tubercles seems to decrease in preserved specimens: when collected, QCAZ 48744 had large conspicuous dorsal tubercles (Fig. 7), in preservative tubercles are barely noticeable.   Ventral surfaces of preserved specimens (Fig. 9) vary from light gray (QCAZ 40909) to brown (QCAZ 31031). In most specimens, there are dark brown or dark gray spots, more abundant posteriorly (e.g., QCAZ 49439); QCAZ 39633 has brown blotches on the chest and venter; QCAZ 48804 has similar marks that also reach the gular region. In two Peruvian specimens ventral surfaces are light gray with few brown spots posteriorly (CORBIDI 09553) or with light brown spots, slightly visible, on gular region and belly (MUSM 28050). The gular regions in some Peruvian specimens are brown (e.g., CORBIDI 09507, 10532). Ventrally, limbs vary from light gray or light brown to dark brown; in QCAZ 33256 and 39587 black dots are present on limbs; scant cream tubercles can be present in the external edge of the forearm (e.g., QCAZ 32512). The skin of the anterior and posterior surfaces of thighs and the concealed surfaces of shanks are pale in the Peruvian specimens. The vent region is light gray to dark brown with dark brown dots. Flanks are cream to light gray, areolate in the anterior two-thirds and smooth posteriorly. In specimens from Peru the flanks are completely areolate. The areolate portion has a dark brown reticulation.
Head shape is truncate in dorsal and lateral view (e.g., QCAZ 39579). Lateral head coloration varies between dark brown (QCAZ 49569) to cream (QCAZ 32506). There is a cream subocular mark. The tympanic annulus is concealed dorsally and has lighter color than the background. The distal subarticular tubercle on Finger IV is single (e.g., QCAZ 40909) or bifid (e.g., QCAZ 45272).
Morphometric data pertain to adults and are summarized in Table 3. In the examined series, the largest male has a SVL of 57.21 mm and the largest female 72.77 mm; mean male SVL = 46.84 mm (n = 33, SD = 4.31), mean female SVL = 66.55 mm (n = 3, SD = 5.44). Females are significantly larger than males (t = 7.66, df = 33, P < 0.001). A MANOVA on the residuals of the regressions between SVL and the other measured variables indicates lack of significant differences between sexes in size-free morphometry (F = 0.239, df = 17, P = 0.060).
Color in life. Based on digital photograph of adult male QCAZ 48744 (Fig. 7 C-D): dorsum green with irregular light and dark brown marks; canthal region green with cream subocular mark and olive green diffuse band along the posterior half of upper lip; tympanum light brown; flanks light green with dark brown reticulation anteriorly and irregular dark brown blotches posteriorly; dorsal surfaces of thighs, shanks and forelimbs green with transversal brown bands; venter brown with irregular dark brown and cream marks; iris bronze with diffuse brown mid-horizontal line and black reticulations.
Based on digital photograph of juvenile QCAZ 40859 (Fig. 7 F): dorsum green with dark brown marks; upper lip cream with transversal brown bars; flanks light green with brown marks; dorsal surfaces of arms, thighs and shanks green with brown transversal bars; external edge of tarsus with white tubercles; iris bronze with black reticulations and diffuse mid-horizontal dark band between the pupil and posterior border of iris.
In life the Peruvian specimens have extensive blue coloration in the groins, concealed surfaces of thighs and tibia, dorsal surfaces of tarsus, armpits and posterior surfaces of arms (e.g., CORBIDI 10534; Fig. 7 G-H). The iris is highly variable from light cream to brownish cream and dark brown (CORBIDI 09394); there are always black reticulations and a diffuse mid-horizontal dark band. Some individuals have a diffuse vertical dark band below the pupil.
Green coloration in life changes to cream in preserved specimens.
Call. Males call from vegetation next to rivers or streams. Acoustic parameters of the advertisement call of O. cannatellai are shown in Table 4. The call consists of two  components. The first is obligatory and consists of one to five rattle-like notes. The second component is facultative and consists of one to three quacks. The first component is pulsed and lacks harmonic structure; the second component has visible harmonics and reaches higher amplitude than the first component (Fig. 11). The advertisement calls of O. cannatellai differ markedly from those of O. buckleyi. Calls of O. buckleyi (Fig. 11) consist of a pulsed rattle-like note repeated at irregular intervals and without a second component. Those calls have a shorter duration, higher repetition rate, and fewer pulses than calls of O. cannatellai.
Distribution and ecology. Osteocephalus cannatellai has been recorded at twelve localities, all of them south of the Napo river, in the Ecuadorian and Peruvian Amazon basin (Provincias Morona Santiago, Napo, Orellana, Pastaza, Zamora-Chinchipe, and Datem del Marañón; Fig. 2 Most specimens were collected at Río Pucayacu, a river surrounded by a mixture of primary and secondary forest. Frogs were found next to the river, perching over broad leaves or on tree branches 50 to 230 cm above the ground. At the collection site, the river has an average width of approximately 10 m, fast running water, and a rocky bottom. Males were calling next to the river between June 26 and July 3 2010. Several adults and a juvenile were found on a small stream, tributary of Río Rivadeneira, surrounded by secondary forest, near Río Pucayacu, in March 2008.
Vegetation types (according to the classification of Sierra et al. 1999 ) are: (1) Amazonian Mountain Range Evergreen Foothill Forest, characterized by a mixture of  Amazonian and Andean vegetation with a canopy of 30 m (Río Pucayacu, Bobonaza, and Yawi), (2) Amazonian Lowland Evergreen Forests, characterized by high plant alpha-diversity and a canopy of 30 m with emergent trees that reach 40 m (Huino, Río Maderoyacu, Reserva Yachana), and (3) Amazonian Lower Montane Evergreen Forest, with an elevational range of 1300 to 2000 m above sea level, its canopy can reach 25 to 30 m (Nuevo Israel; Sierra et al. 1999. Specimens from Peru were collected in Cordillera de Kampankis within an elevational range of 300 to 365 m above sea level in tall, closed-canopy forest on low hills with well-drained soils at the base of the mountains. The soils have variable proportions of silt, clay and sand, but there are some small patches of sandy soil and limestone outcrops. The forest canopy is about 30 m tall, with emergent trees reaching 45 m. All individuals were collected in riparian vegetation of low-velocity and low-volume streams with rounded slate rocks lining the stream bed. Some individual were found on leafs of dense populations of rheophytic plants or shrubby Pitcairnia aphelandriflora (Bromeliaceae). Other individuals were found on branches of bushes between 50 and 200 cm above the ground. Other arboreal frogs at the site were O. mutabor, Hypsiboas cinerascens, and Gastrotheca longipes.
Paratypes. (Fig. 15  Diagnosis. Throughout this section, coloration refers to preserved specimens unless otherwise noted. Osteocephalus germani is a medium-sized species of Osteocephalus having the following combination of characters: (1) size sexually dimorphic; maximum SVL in males 41.45 mm (n = 2), in females 50.76 (n = 2); (2) skin on dorsum bearing tubercles in males, smooth in females; (3) skin on flanks areolate; (4) hand webbing formula varying from I basal II2 --3 -III2½-2IV to I basal II2-3III3 --3 -IV; foot webbing formula varying from I1-1½II1 --2III1-2IV1½-1 -V to I1 + -2II1 + -2III1 + -2 + IV2-1V; (5) dorsum varying from brown with dark brown marks to light gray with dark brown marks; (6) venter light cream with or without dark brown flecks; (7) cream suborbital mark present, clear labial stripe absent; (8) flanks cream to brownish cream with dark brown blotches and flecks; (9) dermal roofing bones of the skull weakly exostosed; (10) bones green in life; (11) in life, iris golden to reddish golden with fine dark reticulation; (12) paired vocal sacs, located laterally, behind jaw articulation, (13) juvenile coloration unknown; (14)   Description of holotype. Adult male, 41.26 mm SVL, head length 12.79, head width 14.23, eye diameter 5.23, tympanum diameter 3.79, femur length 22.3, tibia length 23.1, foot length 17.97. Head narrower than body, slightly wider than long; snout rounded in dorsal view and truncate in lateral view; distance from nostril to eye longer than diameter of eye; canthus rostralis distinct and straight; loreal region concave; internarial area depressed; nostrils moderately protuberant, directed laterally; interorbital area with tiny keratinized conical tubercles, lateral margins of frontoparietals inconspicuous through skin; eye large, strongly protuberant; tympanic membrane clearly evident, large, slightly wider than high, about two thirds of eye diameter, separated from eye by ca. 85% of its diameter; tympanic annulus distinct except dorsally where it is covered by supratympanic fold; posterior end of supratympanic fold reaches arm insertion. Arm slender, axillary membrane present, reaching less than one third of arm length; four small low tubercles present along ventrolateral edge of forearm; relative length of fingers I<II<IV<III; fingers bearing large, oval discs, that of third finger about three fourths of tympanum diameter; subarticular tubercles prominent, round to ovoid except for bifid distal subarticular tubercle of Finger IV; supernumerary tubercles present; palmar tubercle small, elongated; prepollical tubercle large, flat, elliptical; prepollex enlarged; large dark keratinous nuptial excrescences covering inner surface of prepollex up to two thirds the distance between subarticular tubercle and proximal border of disk of thumb; webbing absent between fingers I and II; webbing formula of fingers II2 --3III2½ -3 -IV. Small tubercles on tibiotarsal articulation; dorsal surface of tarsus covered by tiny keratinized conical tubercles, more abundant on outer edge; minute tubercles scattered along ventrolateral edge of foot; toes bearing discs slightly wider than long, smaller than those of fingers; relative length of toes I<II<V<III<IV; outer metatarsal tubercle ill defined, small, round; inner metatarsal tubercle low, ovoid; subarticular tubercles single, round, protuberant; supernumerary tubercles restricted to the soles; webbing formula of toes I1-2II1-2III1 + -2IV2 --1V. Skin on dorsum, head, and dorsal surfaces of limbs shagreen covered by conical tubercles with keratinized tips, tiny on head and limbs; skin on flanks areolate; skin on venter coarsely granular; skin on ventral surfaces of head and thighs granular, those of shanks smooth. Cloacal opening directed posteriorly at upper level of thighs; short simple cloacal sheath covering cloacal opening; round tubercles below vent; two conspicuous white tubercles ventrolateral to vent at lower level of thighs. Tongue cordiform, widely attached to floor of mouth; dentigerous processes of the vomers angular, adjacent medially, posteromedial to choanae, bearing 5 and 6 (left/right) vomerine teeth; choanae trapezoidal, oblique; vocal slits moderately long, extending diagonally from lateral end of tongue toward to the angle of snout; vocal sac indistinct above the arm and below the ear.
Color of holotype in preservative. Dorsum light brown with dark brown peripheral marks; dark brown transversal bar between orbits with fine pale borders; dorsal surfaces of forearms grayish brown with dark brown marks, dorsal surfaces of thighs, shanks, and feet grayish brown with diffuse brown transversal bars. Venter light cream with dark brown flecks on the throat and thoracic region and absent on posterior half of the body; ventral surfaces of hindlimbs and forelimbs dirty cream with dark brown flecks on the lateral borders of shanks; outer half of ventral surfaces of forearms dirty cream; sides of head brown with white subocular band extending, below tympanum, two little brown blotches below the eye; tympanic membrane dark brown and area in the periphery of tympanum light brown dorsally and grayish brown behind the tympanum; flanks grayish white, areolate region with dark brown reticulation and flecks. Iris silver with dark brown mid-horizontal line and thin black reticulations.
Color in life. Dorsum brown with irregular dark brown marks; flanks brownish cream with dark brown spots and flecks; dorsal surfaces of thighs, shanks, and forelimbs brown with transversal dark brown bands. Venter whitish cream with brown flecks in throat; ventral surfaces of thighs tan. Iris bronze with thin black reticulations (G. Chávez field notes April 2010).

Etymology.
The new species is dedicated to our colleague German Chávez (COR-BIDI), one of the best friends of PJV, for his contributions to Peruvian herpetology and collecting the type series and tissues of this new species.
Variation. Variation in dorsal and ventral coloration of preserved specimens is shown in Figure 14. Dorsal background coloration varies from light brown to light gray; irregular dark brown marks are always present. In females, the dorsum lacks tubercles while in males tubercles are present. The single male paratype (CORBIDI 06660) differs from the holotype in having non-keratinized tubercles.
Ventral surfaces of preserved specimens (Fig. 14) are whitish cream. All the specimens have scattered dark brown flecks on the anterior half of the venter. Ventrally, limbs vary from whitish cream to tan; scant white tubercles can be present in the external edge of the forearm of males (e.g., CORBIDI 06660). The vent region is light brown or dark. Flanks are whitish cream to light gray, areolate in the anterior half and nearly smooth posteriorly. The areolate portion is completely covered by dark brown reticulation and flecks.
Snout is truncate in lateral view except for a female with rounded snout (COR-BIDI 06633). Lateral head coloration varies from dull brown (CORBIDI 06633) to cream with dark brown blotches (CORBIDI 05505). The tympanic annulus is concealed dorsally and has lighter color than the background. The distal subarticular tubercle on Finger IV is bifid in all the specimens.
Adult morphometric data are summarized in Table 3. In the examined series, the largest male has a SVL of 41.45 mm and the largest female 50.76 mm; mean male SVL = 41.35 mm (n = 2, SD = 0.13), mean female SVL = 49.96 mm (n = 2, SD = 1.13).
Color in life. Based on digital photograph of adult male CORBIDI 06660: dorsum brown with irregular dark brown marks and some scattered light green blotches; canthal region greenish brown with greenish cream subocular mark and dark labial bars; tympanum light brown; flanks light green with dark brown reticulation and dark brown blotches posteriorly; dorsal surfaces of thighs, shanks, and forelimbs brown with transversal dark brown bands and scattered light green blotches. Iris bronze with brown horizontal midline and thin black reticulations.
Based on digital photograph of adult female CORBIDI 06633: dorsum brown with few scattered irregular dark brown marks; canthal region dark brown with greenish cream subocular mark speckled by three small dark brown blotches; tympanum light brown; flanks light brown with dark brown blotches; ventrolateral region cream with fine dark reticulation; dorsal surfaces of thighs, shanks, and forelimbs brown with transversal dark brown bands. Anterior half of venter whitish cream with fine brown reticulation in throat and chest; posterior half of venter and ventral surfaces of thighs tan; iris bronze with diffuse brown mid-horizontal line and thin black reticulations.
Based on digital photograph of adult female CORBIDI 05505 (Fig. 15): dorsum green with irregular dark brown marks; canthal region green with brown mottling and white subocular mark extending to the lips as a white labial stripe along posterior half of the jaw; tympanum light brown; flanks white with dark brown reticulation and small dark brown blotches posteriorly; dorsal surfaces of thighs, shanks, and forelimbs green with transversal dark brown bands and flecks. Venter white with scattered brown flecks on throat and chest. Iris reddish gold with diffuse brown mid-horizontal line and thin black reticulations. Based on digital photograph of adult female CORBIDI 08284 (Fig. 15): dorsum light brown with irregular dark brown marks; canthal region brown and greenish white subocular mark; tympanum light brown; flanks light brown with small dark brown blotches; dorsal surfaces of thighs, shanks, and forelimbs light brown with transversal dark brown bands. Venter dull cream. Iris bronze with diffuse brown mid-horizontal line and thin black reticulations.
Distribution and ecology. Osteocephalus germani is known from three localities in southern Peru (Fig. 6). Pongo de Manique and Comunidad Nativa de Poyentimari are in premontane forest on the Upper Urubamba River basin (vegetation types according to ONERN 1976) in the Amazonian foothills of the southern Peruvian Andes, at elevations of 670-725 m; Comunidad Nativa de Chokoriari is Terra Firme Amazonian lowland forests on the lower Urubamba River basin in the southern Peruvian Amazon lowlands, at elevation of 434 m. In Pongo de Mainique the new species was found close to rocky streams in low-hill primary forest with arboreal ferns and abundant epiphytes. At this locality, O. germani was sympatric with O. castaneicola and O. mimeticus. In Comunidad Nativa de Poyentimari, O. germani was found close to rocky streams in a step area of very wet high-hill primary forest with abundant ferns (including arboreal), epiphytes, lichens and mosses. At this locality the new species was sympatric with O. mimeticus. In Comunidad Nativa de Chokoriari, O. germani was found close to a black-water slow-running creek in a patch of secondary forest, surrounded by pastures for cattle and plantations. The forest was dominated by bamboo and Cecropia spp. and the creek had sandy soils covered by leaf litter. No other species of Osteocephalus were found in this locality.
All specimens were collected next to temporary pools, perching over broad leaves or on tree branches 100 to 200 cm above the ground. Many streams surround the collection sites.
Remarks. In the phylogeny (Fig. 1), two specimens from gen bank (EF376030 from French Guiana and AY843705 from Río Jurua, Brazil) are grouped with O. germani in a strongly supported clade (PP = 0.96) and are likely conspecific or represent one or two closely related species. The specimen from French Guiana was reported as "O. oophagus" by Salducci et al. (2002Salducci et al. ( , 2005; the specimen from Brazil was reported as "O. cabrerai" by Faivovich et al. (2005). Both individuals appear to be misidentified.  Description of holotype. Adult male, 51.85 mm SVL, head length 18.9, head width 19.0, eye diameter 6.8, tympanum diameter 4.9, femur length 28.0, tibia length 28.7, foot length 22.1. Head narrower than body, nearly as wide as long; snout truncate in lateral and dorsal views; distance from nostril to eye longer than diameter of eye; canthus rostralis distinct and straight; loreal region concave; internarial area depressed; nostrils moderately protuberant, directed laterally; interorbital area flat, lateral margins of frontoparietals distinct through skin; eye large, strongly protuberant; tympanic membrane clearly evident, slightly wider than high, about two thirds of eye length, separated from eye by ca. 85% of its diameter; tympanic annulus distinct except dorsally where it is covered by supratympanic fold; posterior end of supratympanic fold reaches mid arm insertion. Arm slender, axillary membrane present, reaching one third of arm length; three small low tubercles present along ventrolateral edge of forearm; relative length of fingers I<II<IV<III; fingers bearing large, oval discs, that of third finger about three fourths of tympanum diameter; subarticular tubercles prominent, round to ovoid, bifid in distal subarticular tubercle of Finger IV; supernumerary tubercles present; palmar tubercle small, elongated; prepollical tubercle large, flat, elliptical; prepollex enlarged; large dark keratinous nuptial excrescences covering inner surface of prepollex almost reaching the proximal border of disk of thumb; webbing basal between fingers I and II; webbing formula of fingers I basal II1½-2½III2+-2IV. Medium sized to small tubercles on tibiotarsal articulation; scattered low tubercles on tarsus, more abundant on outer edge; small tubercles scattered along ventrolateral edge of foot; toes bearing discs slightly wider than long, smaller than those of fingers; relative length of toes I<II<V<III<IV; outer metatarsal tubercle ill defined, small, round; inner metatarsal tubercle large, ovoid; subarticular tubercles single, round, protuberant; supernumerary tubercles restricted to the soles; webbing formula of toes I1-2 -II1-2III1-2IV2-1 -V. Skin on dorsum, head, and dorsal surfaces of limbs shagreen, with scattered tubercles; minute keratinized conical tubercles present on the eyelids and dorsal surface of head; skin on flanks areolate with big flattened warts; skin on venter coarsely granular; skin on ventral surfaces of head and thighs granular, that on shanks smooth. Cloacal opening directed posteriorly at upper level of thighs; short simple cloacal sheath covering cloacal opening; round tubercles below vent; two distinct white tubercles ventrolateral to vent. Tongue cordiform, widely attached to floor of mouth; dentigerous processes of the vomers angular, adjacent medially, posteromedial to choanae, bearing 9 and 6 (left/right) vomerine teeth; choanae trapezoidal, oblique; vocal slits short and curved posteroventral to the angle of snout at the base of tongue; vocal sac barely distinct above the arm and below the ear.
Color of holotype in preservative. Dorsum brown with a single diffuse interorbital mark; dorsal surfaces of forearms brown with diffuse brown bands; dorsal surfaces of hindlimbs brown with diffuse dark brown marks on shanks and feet. Venter dirty cream with light brown spots, more abundant on posterior half of the body; ventral surfaces of hindlimbs and forelimbs dirty cream without marks but with distinct white tubercles on forearms; outer half of ventral surfaces of forearms dark brown; sides of head light brown with oblique white bar from posteroventral border of orbit to border of jaw, below tympanum; vertical diffuse brown bar below eye, anterior to white bar; area behind white bar and eye dark brown including periphery of tympanum; flanks dirty cream, areolate region with brown reticulation. Iris silver with a brown mid-horizontal line and thin black reticulations.
Color of holotype in life. Based on digital photograph (Fig. 15). Dorsum pale brown without marks; canthal region pale brown with diffuse pale green subocular mark and dark stripe along the posterior half of upper lip; tympanum pink; flanks light green without marks; dorsal surfaces of thighs and tarsus pale brown with greenish brown transversal bands, forearms greenish brown; tibia pale brown without marks; anterior and posterior surfaces of thighs, concealed surfaces of tibia, and metatarsus pale blue. Venter dirty cream with light brown spots, more abundant on posterior half of the body; ventral surfaces of hindlimbs and forelimbs dirty cream. Iris dirty cream with brown transversal midline and black reticulations.
Etymology. The specific name is a patronym for Vilma Duran, in recognition of her continued work and efforts toward the improvement of the herpetological collection of CORBIDI and also for collecting the holotype and tissue of this new species. Figure 17. Dorsal background coloration varies from light brown to brown; irregular dark brown or dark gray marks are always present (Fig. 17). Flanks are always cream to grayish cream. Two specimens have a cream middorsal line from the tip of the snout to the vent (CORBIDI 06469, QCAZ 51205). The prominence of the tubercles can decrease in preserved specimens: when collected, CORBIDI 01086 had large conspicuous dorsal tubercle, in preservative tubercles are barely noticeable.

Variation. Dorsal and ventral coloration of preserved specimens is shown in
Ventral surfaces of preserved specimens (Fig. 17) vary from cream to vanilla. In most specimens, there are dark brown spots, more distinct posteriorly or in the throat (e.g., CORBIDI 06469); ventrally, limbs vary from dirty cream to light brown; all specimens have small white tubercles in the external edge of the forearm. The vent region is gray to brown with dark brown flecks or dots. Flanks are cream to gray, areolate, with dark brown reticulations, dots, and blotches along the entire flank or restricted to the posterior half (e.g. CORBIDI 05031).
Head shape is truncate in dorsal view and truncate in lateral view. Lateral head coloration varies from light brown with dark mottling (CORBIDI 01086) to grayish white with dark brown canthus rostralis and preocular stripes (CORBIDI 05031). All specimens have a white to cream subocular mark. The tympanic annulus is concealed dorsally and has lighter color than the background. The distal subarticular tubercle on Finger IV is bifid in all specimens.
Adult morphometric data are summarized in Table 3. In the examined series, the largest male has a SVL of 55.77 mm; mean male SVL = 50.74 mm (n = 6, SD = 3.17).
Color in life. Based on a digital photograph of adult male CORBIDI 01086: dorsum light brown with irregular dark brown and light green marks; canthal region greenish brown with white subocular mark and dark brown band along posterior half of upper lip; tympanum pink contrasting with dark brown tympanic annulus; flanks light green with dark brown reticulation anteriorly and few irregular dark brown blotches posteriorly; dorsal surfaces of thighs, shanks and forelimbs brown with dark brown transversal bands; posterior surfaces of thighs light green; venter white speckled with light brown blotches; iris light cream with brown mid-horizontal line and fine black reticulations.
Distribution and ecology. Osteocephalus vilmae is know from seven localities in the Peruvian and Ecuadorian Amazon basin (northern Loreto region), four at Río Corrientes (Jibarito, Nuevo Corrientes, Pampa Hermosa, and Shiviyacu), two near Rio Pastaza in the border Ecuador-Peru (Andoas and Capahuari Norte) and one at Provincia de Orellana, Pompeya-Iro road (Fig. 2). The elevations of these localities are between 150 to 270 m above sea level. Maximum airline distance between localities is 158 km. The Peruvian localities are dominated by Terra Firme forest. Specimens collected in Capahuari Norte were found in a stream surrounded by a mixture of primary and secondary forest. In Jibarito, Pampa Hermosa, and Shiviyacu the frogs were found in primary forest in a swamp close to a stream. All specimens were next to the streams, perching on tree branches 100 to 200 cm above the ground. Osteocephalus vilmae occurs sympatrically with O. buckleyi at km 80 Pompeya-Iro road. At the Peruvian localities it co-occurs with O. mutabor and O. planiceps.

Morphometric comparisons among species
Three components with eigenvalues > 1.0, accounting for 70.9% of the total variation, were extracted from the PCA for males (Table 5). The highest loadings were femur length and tibia length for PC I, eye diameter and tympanum diameter for PC II, and head length for PC III (    Table verruciger) are significant (all P values for t tests < 0.04); O. verruciger also shows significant differences with all the remaining species (all P values < 0.004). Three components with eigenvalues > 1.0 were extracted from the PCA for females ( Table 5). The three components accounted for 75.7% of the total variation. The highest loadings for the PCA for females were tibia length and femur length for PC I, head width for PC II, and eye diameter and head length for PC III (Table 5). As in the PCA for males, there is wide overlap in morphometric space among species (Fig. 18). The only exception is O. cannatellai, which segregates from the other species along PC II . However, larger sample sizes are required to confirm this differentiation.
In the DFA classification on males, all O. cabrerai, O. festae, O. germani, O. verruciger, and O. vilmae were correctly classified (n = 7, 7, 2, 22, and 5 respectively). In O. buckleyi, 17 out of 24 specimens were correctly classified (4 were misclassified as O. cabrerai, 2 as O. germani, and 1 as O. vilmae); in O. cannatellai, only 4 out of 33 specimens were incorrectly classified, 2 as O. buckleyi and 2 as O. vilmae. Overall, the DFA show morphometric differentiation among the analyzed species. The DFA on females shows even better discrimination because all individuals were correctly assigned to their own species.

Discussion
Similarly to previous studies on Amazonian amphibians (e.g., Elmer et al. 2007;Fouquet et al. 2007;Funk et al. 2011) our results document a large proportion (300% increase) of hidden diversity within a set of populations that were previously treated as a single widely distributed species. Moreover, because most of our sampling was restricted to Ecuador and Peru, it is likely that there are even more species than found in our study. These results highlight the need to carry out large-scale genetic surveys of Amazonian amphibians to achieve a more realistic understanding of their diversity and evolution.
Genetic evidence is a valuable taxonomic tool but, in most cases, is insufficient to define species boundaries without reference to other sets of characters like advertisement calls or external morphology. Taxonomic reviews of Amazonian amphibians suggest that morphological characters are too conservative to define species boundaries because closely related species share similar morphology (e.g., Elmer et al. 2007;Fouquet et al. 2012;Funk et al. 2011;Lougheed et al. 2006;). Our results, however, indicate that in some groups, like the O. buckleyi species complex, this is not necessarily the case. The three new species described here are diagnosable with morphological evidence alone and are distinctive from the other species of the complex. Morphological differences are also evident between O. buckleyi, O. cabrerai, O. carri, O. festae, and O. mutabor. Thus, none of the species of the complex are strictly cryptic (i.e., all of them can be identified using morphological characters) although their diagnosis based on morphology is challenging. Other groups of Amazonian amphibians on which phylogenetic analyses of DNA have led to the discovery of species that turned out to be morphologically distinct are the Hypsiboas fasciatus-calcaratus complex (Funk et al. 2011) and the Pristimantis "ockendeni" complex (Elmer and Cannatella 2008).
We suspect that the difficulty in defining species boundaries based on morphology arises from the high intraspecific polymorphism in coloration characteristic of most groups of dull-colored Amazonian amphibians like Osteocephalus and Pristimantis (see for example Fig. 2 in Elmer and Cannatella 2008 and Figs 3, 8, 14, 17 herein). If this is the case, understanding the evolutionary processes that generate and maintain polymorphism in coloration could help to predict which Amazonian taxa are more likely to contain "cryptic" diversity. One plausible process is frequency dependent predation which, occurs when the probability of predation is inversely correlated to the frequency of a given prey type in the population (for a review see Punzalan et al. 2005). Under this scenario, predators use search images to find preys and are better at detecting previously seen prey types because they have learned to find them. Although other processes could also explain polymorphisms (e.g., deferential selection associated with spatial variation in backgrounds), the available evidence suggests that some form of frequency dependent selection is the most likely explanation for color polymorphism in anurans (Milstead et al. 1974;Wells 2007, pp. 715).
Most Osteocephalus have a predominantly and highly polymorphic brown coloration and are cryptic against the background where they are found by day (Deichmann 2008;Deichmann and Williamson 2007;SRR pers. obs.) If polymorphisms are an adaptation to avoid falling into search categories of visually oriented predators, the difficulties of species delimitation based on morphological characters could be a byproduct of this selective pressure. This hypothesis needs to be tested empirically because if verified it could help to understand why several groups of Neotropical amphibians contain a large proportion of cryptic species.

Biogeography and speciation
Examination of the geographic ranges of sister species can provide insights into modes of speciation. Our phylogeny of the O. buckleyi species complex recovered four sister species pairs of which one is sympatric (O. buckleyi-O. vilmae) and three are allopatric. Among the allopatric pairs, two involve a lowland species sister to a highland species. Osteocephalus mutabor occurs at lower altitudes (range 230-1240 m) than its sister species, O. festae (860-2383 m). Similarly, O. cannatellai has a lower distribution (200-1290 m) than its sister species, O. verruciger (950-2120 m). Because most species of Osteocephalus are restricted to elevations below 1000 m, the distributions of O. festae and O. verruciger probably represent parallel and recent colonization events from the lowlands. This geographic pattern suggests that speciation has been a result of ecological mediated selection along an altitudinal gradient. Interestingly, both highland species resemble each other closely in external morphology (Figs 15 and 18) suggesting convergence as a byproduct of adaptation to similar environments. Speciation associated with ecological divergence along altitudinal gradients was also reported by Graham et al. (2004) in dendrobatid frogs and more recently by Salerno et al. (2012) between Tepuihyla and its sister lowland species, "Osteocephalus" exophthalmus.
At the intraspecific level, we found low genetic divergence with the only exception of Osteocephalus festae (up to 2.8% of uncorrected p distance in gene 12S). We also found a concordant geographic pattern of divergence in O. buckleyi and O. mutabor because in both the most divergent population was the most northern of them, in the Cuyabeno region. High divergence of samples from Cuyabeno relative to others to the south was also reported for Pristimantis kichwarum (Elmer and Cannatella 2008). Samples of O. mutabor, O. buckleyi (sensu stricto) and O. cannatellai show genetic structure generally congruent with geography (i.e., geographically close localities tend to be genetically similar). Overall, our intraspecific sampling reveals low levels of genetic differentiation and genetic variation geographically structured.