Terrestrialization, the evolution of non-aquatic oviposition, and miniaturization, the evolution of tiny adult body size, are recurring trends in amphibian evolution, but the relationships among the traits that characterize these phenomena are not well understood. Furthermore, these traits have been identified as possible “key innovations” that are predicted to increase rates of speciation in those lineages in which they evolve. We examine terrestrialization and miniaturization in sub-Saharan puddle frogs (Phrynobatrachidae) in a phylogenetic context to investigate the relationship between adaptation and diversification through time. We use relative dating techniques to ascertain if character trait shifts are associated with increased diversification rates, and we evaluate the likelihood that a single temporal event can explain the evolution of those traits. Results indicate alternate reproductive modes evolved independently in Phrynobatrachus at least seven times, including terrestrial deposition of eggs and terrestrial, non-feeding larvae. These shifts towards alternate reproductive modes are not linked to a common temporal event. Contrary to the “key innovations” hypothesis, clades that exhibit alternate reproductive modes have lower diversification rates than those that deposit eggs aquatically. Adult habitat, pedal webbing and body size have no effect on diversification rates. Though these traits putatively identified as key innovations for Phrynobatrachus do not seem to be associated with increased speciation rates, they may still provide opportunities to extend into new niches, thus increasing overall diversity.
Citation: Zimkus BM, Lawson L, Loader SP, Hanken J (2012) Terrestrialization, Miniaturization and Rates of Diversification in African Puddle Frogs (Anura: Phrynobatrachidae). PLoS ONE 7(4): e35118. https://doi.org/10.1371/journal.pone.0035118
Editor: Nathan G. Swenson, Michigan State University, United States of America
Received: August 10, 2011; Accepted: March 13, 2012; Published: April 10, 2012
Copyright: © 2012 Zimkus et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This study was supported by the Department of Organismic and Evolutionary Biology and the Museum of Comparative Zoology, Harvard University, and by National Science Foundation grant EF-0334939 (AmphibiaTree) to J. Hanken. Putnam expedition grants from the Museum of Comparative Zoology supported fieldwork by B. Zimkus in Ethiopia and Tanzania. Fieldwork in Tanzania was additionally supported by a Society for Integrative and Comparative Biology Fellowship for Graduate Student Travel. S. Loader is supported by a Swiss National Science Foundation grant (31003A-133067). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Trait shifts that enable extension into and colonization of new environments, or ecospace, and subsequent diversification in ways that were not previously possible are commonly cited as “key innovations” , . These characteristics are hypothesized to influence diversification rates, and include traits associated with life history, body size and ecology –. Few studies, however, have analyzed diversification rates within extensively sampled radiations to determine whether specific traits promote speciation due to the difficulty of assembling such comprehensive data sets –. The concept of key innovations is further complicated by the fact that successful adaptation to a new ecological niche must be attributable to a specific innovation. To identify potential key innovations and understand current patterns of species diversity, both intrinsic and extrinsic factors must be investigated simultaneously. Intrinsic characteristics provide the adaptive potential for radiation, while extrinsic factors, such as climate or habitat change, provide the ecological opportunity for diversification.
One possible key innovation in amphibians is the shift to terrestrial reproduction, which allows eggs to be laid away from standing water and facilitates the transition from aquatic to terrestrial existence. Amphibians are unique among tetrapod vertebrates in the tremendous diversity of ways in which they reproduce . The generalized reproductive mode in anurans comprises a biphasic pattern of development, which includes large clutches of small aquatic eggs and free-living, aquatic larvae that undergo metamorphosis to terrestrial adults. There are, however, nearly 40 alternate reproductive modes in anurans, many of which are characterized by small clutches of large terrestrial or arboreal eggs . In some species eggs hatch into non-feeding tadpoles and undergo metamorphosis after utilizing endogenous yolk supplies, whereas others employ direct development and lack the free-living larval stage entirely. This shift to a non-aquatic lifestyle is a recurring theme in amphibian evolution that is manifest as multiple independent occurrences of terrestrialization and direct development , –. The shift to an alternate mode of reproduction may enable a lineage to occupy a new ecological niche or habitat, in the same way that loss of the free-living, feeding larval stage may enable populations to persist in areas without permanent bodies of water , .
Another widespread phenomenon among animals, including amphibians, is miniaturization, the evolution of extremely decreased adult body size. Although many biologists have noted that body size and degree of terrestriality appear to be correlated within certain amphibian lineages, the relationship between these two trends has not been adequately investigated. Body size is undisputedly one of the most significant organismal traits because it influences many biological attributes, including behavior, ecology, life history and physiology. The selective advantages of miniaturization include predator avoidance , , exploitation of alternate food sources , utilization of physically smaller niches , ,  and attainment of reproductive maturity at an earlier age , . Miniaturization likely played an important role in the origin of the Lissamphibia, which includes all living amphibians, and is a recurring trend within all three extant orders. Miniaturization has been documented within anurans, caecilians and salamanders and has occurred independently many times in each clade , , –. Although miniaturization is an important factor in amphibian evolution, its relation to species diversification and its correlation with other proposed key innovations are poorly understood; recent studies have yielded conflicting results. In the global radiation of toads, the most speciose clades were characterized by large body sizes and large ranges . In contrast, the most speciose clades of Malagasy mantellid frogs were characterized by small body size and small ranges, which was attributed to the poor dispersal ability of small-bodied species . Finally, the diversity of Amazonian hylid treefrogs is not explained by body size evolution, but rather by the timing of colonization and long-term sympatry of multiple clades .
Puddle frogs of the genus Phrynobatrachus (Anura: Phrynobatrachidae) comprise approximately 80 species distributed across sub-Saharan Africa, where they are adapted to aquatic, semi-aquatic, and terrestrial habitats –. They comprise the second largest generic radiation on the continent, which makes them ideal for investigating broader evolutionary studies of continental speciation. Adult snout-vent lengths (SVL) vary greatly, from as little as 12 mm in some miniaturized species to greater than 50 mm in the largest species. The lineage also displays diverse reproductive modes. Although most species deposit large clutches of aquatic eggs, some species are semi-independent from permanent water and lay eggs in very small amounts of water where the tadpoles develop and feed –. Other species are endotrophic, with terrestrial eggs and non-feeding larvae that obtain developmental nutrients entirely from maternally provisioned yolk.
Recent derivation of a molecular phylogeny of sub-Saharan puddle frogs enables us to simultaneously investigate terrestrialization and miniaturization for the first time . Here we use this phylogeny, reanalyzed with improved species-tree methods, to test the following hypotheses: (1) terrestrialization is characterized by alternate modes of reproduction and low water dependence due to occupation of habitats with high humidity; (2) alternate reproductive modes evolved once within puddle frogs; (3) miniaturization is correlated with decreased pedal webbing; and (4) miniaturization and decreased pedal webbing are correlated with terrestrialization.We also use relative dating methods and diversification analysis to test two additional hypotheses: (5) character traits (e.g., reproductive mode, adult habitat, pedal webbing and body size) are significantly correlated with change in diversification rates within Phrynobatrachus, and (6) a single temporal or geographic events correspond to shifts in character trait changes across the Phrynobatrachus phylogeny, thereby implying a common extrinsic cause. Testing these hypotheses allows us to determine if morphological (intrinsic) and ecological (extrinsic) traits that characterize terrestrialization impact diversification patterns within puddle frogs, and possibly all amphibians, thereby allowing them to take advantage of resources not previously utilized by aquatic clades.
Taxon sampling and specimen preparation
All field collections made for this study were obtained with appropriate permits (Ethiopia: Wildlife Conservation Department collection permit 13/37/55/521, export permit 13/37/57/568; Tanzania: COSTECH research permit 2007-45-NA-2007-3, Trophy Export Certificate No. 50086-89). External morphology was assessed by examination of alcohol-preserved specimens (Table S1). A total of 279 Phrynobatrachus specimens were examined, comprising 54 species, as well as nine outgroup genera. Only adults were included; mature specimens were identified by the presence of secondary sexual characteristics and/or ossified carpal bones (using radiography). Radiographs were obtained with a Thermo Kevex digital x-ray (Model PXS10) in combination with a PaxScan amorphous silicon sensor array (Model 4030R) and ViVa version 2.0 (Varian Medical Systems, Inc.); specimens were x-rayed at 40 kV.
Phylogenetic relationships and relative dating
Previous hypotheses of Phrynobatrachus phylogeny were obtained by using maximum parsimony (MP), maximum likelihood (ML) and Bayesian (BI) methods for three datasets: mitochondrial DNA (mtDNA; approximately 2.5 kilobases, including 12S rRNA, valine-tRNA and 16S rRNA), nuclear gene RAG-1 (920 base pairs; bp) and combined partitions (mtDNA+RAG-1) . For this study, the combined-dataset phylogeny of Zimkus et al. (2010)  was reanalyzed with improved species-tree methods implemented using the program *BEAST (starBEAST; BEAST package version 1.6.1) . The consensus nuclear allele (after all SNPs within individuals were coded as missing information) for RAG-1 was matched to the mtDNA alignment for each specimen to create a 1∶1 mtDNA∶nuDNA dataset. The RAG-1 locus was analyzed with the codon-based SRD06 nucleotide substitution model , whereas mtDNA was run using the model selected as most closely fitting the data through AIC, AICc and BIC scores with jModelTest 0.1.1 , . jModelTest performs poorly when there is a large amount of missing data. The dataset therefore was trimmed to exclude sites with a high proportion of gaps, which yielded a reduced dataset of 2318 bp, and then further reduced to include only individuals with greater than 90% completeness of the chosen loci, which yielded 165 individuals. Likelihood scores were produced with the full set of possible models (88) using an estimated BIONJ tree topology. The model with the highest likelihood supported by BEAST was GTR+I+G. Substitution-model parameters were unlinked across data partitions, with scaled mutation rates (1∶15 nuDNA∶mtDNA) and strict molecular clocks. Strict clocks were used as relaxed clocks did not significantly improve the analysis. This scaled mutation rate reflects estimates from Crawford (2005)  and Lawson (2010) , who found a 10–20-fold increase in mutation rate for mtDNA. The chain was 25 million generations, logging every thousandth. The first 5.5 million trees were discarded as burn-in in accordance with a significant and sustained improvement in likelihood after this point (as assessed in TRACER) . Quality of the run was assessed in TRACER; each run was deemed suitable if the posterior had an ESS score above 200 with good mixing in the trace. The tree was run first with an unconstrained topology to determine clades in an unrooted system, and then was run constrained with the basal relationship of Phrynobatrachus clade A  to root the *BEAST maximum credibility tree. See supplementary materials for *BEAST XML file (Appendix S1).
Specimens identified as Phrynobatrachus mababiensis A, B and C are each a distinct species ; specimens of P. mababiensis B were unavailable for morphological examination, so this species was excluded from character analyses. Provisional identification of three specimens as Phrynobatrachus mababiensis A is based solely on geography; these specimens were collected closer to the locality of the specimen used to define this clade (MV 10.8; Namibia: Rundu)  than to the localities of P. mababiensis B and C. Phrynobatrachus natalensis was treated as a single species, although it likely comprises a complex of several species in a monophyletic group , . Species identifications follow Zimkus et al. (2010)  for the following putative new species: P. aff. latifrons; P. calcaratus A and B; P. werneri A and B; P. cf. hylaois.
Increased terrestriality is associated with species that exhibit alternate reproductive modes, including terrestrial deposition of eggs and/or tadpoles; see (c) Character coding and ancestral state reconstruction. To evaluate hypotheses of temporal congruence of terrestriality, we tested the null hypothesis that contemporaneous divergences explain the origins of species with alternate reproductive modes. Molecular dating techniques require calibration points but, as with many amphibian lineages, fossils are not available for the Phrynobatrachidae or any related African ranoids. In addition, geological vicariance-based events are not ideal calibration points because their use assumes that a specific event led to vicariant speciation , . Therefore, we inferred relative rather than absolute time scales from molecular data, which is a method recently used by Loader et al. (2007)  to determine if disjunct distributions of African caecilians are temporally congruent. We fixed the root node of the *BEAST consensus tree to an arbitrary value (10) and inferred the relative divergence times of those species that exhibit alternate reproductive modes. Confidence intervals (95%) for estimated divergence dates were used to accept or reject the hypothesis that species with increased reproductive terrestriality are temporally congruent. Regional temporal congruence was also evaluated by coding species according to distribution within five geographic zones based on biotic regions , which were used by Zimkus et al. (2010)  to reconstruct the historical biogeography of Phrynobatrachidae. These zones are (1) western Africa (west of, but not including, Nigeria; includes the Upper Guinean Forest Zone); (2) central Africa (including the Lower Guinean Forest Zone, Cameroon Volcanic Line, and lowland forests of the Congo River Basin); (3) Albertine Rift Mountains (including eastern Democratic Republic of Congo, Burundi, Rwanda and Uganda); (4) eastern Africa (including the Eastern Arc Mountains); (5) southern Africa (countries south of Malawi and excluding northern Mozambique). Species that are distributed in more than one geographic zone are coded according to the ancestral state of its most recent common ancestor .
Character coding and ancestral state reconstruction
We categorically coded and reconstructed the evolution of four traits: two associated with terrestrialization (reproductive mode and adult habitat) and two associated with miniaturization (pedal webbing and body size). All discrete and multi-state characters were reconstructed using maximum likelihood (ML) in Mesquite on a set of 27,000 post burn-in trees obtained from *BEAST analyses. These analyses were completed under a single-rate Mk likelihood model (mk1) with the likelihood decision threshold set to 1.0 (default). Adult body size and extent of pedal webbing were coded as continuous characters and reconstructed on a pruned consensus tree of the combined dataset using weighted squared-change parsimony in Mesquite 2.01 . Those species for which body size data are not available were excluded from all character evolution analyses and ancestral state reconstructions. We also conducted tests of phylogenetic signal using Pagel's λ  calculated in R v. 2.11.1 ; a λ value of 0 implies evolution independent of the phylogenetic tree, whereas a value of 1 indicates that the probability of shared inheritance between species is proportional to their relatedness. If a trait exhibits phylogenetic signal, we expect that the phylogeny would be useful in explaining the distribution of character values, but if the phylogenetic signal is low, our ability to infer patterns of character evolution are limited.
Data regarding reproductive mode were gathered from both published literature and the IUCN Global Amphibian Assessment (Table S2) . Reproductive mode was coded as a binary character for diversification rates analyses: (0) aquatic eggs with feeding, aquatic tadpoles, or (1) alternate reproductive mode. Reproductive mode was also coded as a multi-state character for character evolution analyses: (0) aquatic eggs with feeding, aquatic tadpoles; (1) terrestrial eggs with aquatic tadpoles; (2) terrestrial eggs with terrestrial tadpoles; or (3) direct development, metamorphosed froglets hatch directly from terrestrial eggs. Species without documented reproductive mode data were assumed to have aquatic eggs and larvae. Females of those species were dissected to determine egg size whenever possible. Presence of large eggs was considered indicative of terrestrial clutches, but this condition was not found in any specimen examined.
Adult habitat: independence from water bodies.
Adult habitat was divided into two categories by reorganizing the IUCN habitats classification scheme . Species were coded as (0) dependent on temporary or permanent water bodies, or (1) semi-independent or independent from temporary or permanent water bodies. We assumed low water dependence if the preferred habitat included only areas of high humidity or damp substrates in one or both of these categories: subtropical/tropical moist lowland forest (IUCN code: 1.6) or moist montane forest (IUCN code: 1.9).
Extent of pedal webbing is variable among Phrynobatrachus species; it ranges from (0) absent or rudimentary with 3.0–4.0 phalanges free on toe IV, to (1) moderate to extensive with 0–2.9 phalanges free on toe IV. In addition to this binary classification, we created both continuous and categorical datasets of pedal webbing. The continuous measure was calculated by quantifying the number of phalanges free on the fourth toe (Table S1); the maximum value quantified was used as an heuristic measure for the species. If the extent of webbing attached to one side of the fourth toe differed from that on the other side, the foot was classified with more extensive webbing and fewer phalanges free. Evolution of pedal webbing was also reconstructed as a discrete character by allocating data among five categories: (1) webbing absent, 4 phalanges free on toe IV; (2) rudimentary webbing, 3–3.9 phalanges free on toe IV; (3) moderate webbing, 2–2.9 phalanges free on toe IV; (4) moderate webbing, 1–1.9 phalanges free on toe IV; and (5) extensive webbing, 0–0.9 phalanges free on toe IV.
Evolution of body size was investigated by using maximum values for each species. Snout-vent length (SVL) was determined from direct measurements of alcohol-preserved specimens using digital calipers (±0.1 mm; Table S1); all measurements were taken by one person (BZ). Puddle frog females are often larger than males, but there are many exceptions. In addition, species either lack external morphological characters that reliably distinguish between sexes, or male secondary sexual characteristics may be exhibited only during the breeding season . Therefore, maximum body size for each ingroup species was estimated by using the largest SVL measured, regardless of sex. Maximum SVL approximates the upper bounds for each species; its use enables the inclusion of specimens of both sexes. These values serve as an heuristic measure of maximum body size because of error associated with different sampling intensities among species . Data from the literature were utilized to determine maximum body size for a particular species if few or no specimens were available for examination. Maximum SVL for outgroup genera was estimated by using data from the literature (Table S1). Previous studies of anuran body size evolution utilize binary characters to examine reduction in size and suggest that species smaller than 20 or 25 mm SVL should be considered miniature , . This method is not optimal because it designates an arbitrary threshold to distinguish miniaturized versus non-miniaturized taxa. We therefore coded and reconstructed adult body size as binary, multistate and continuous characters to maximize both refinement and power in analyses. We used results of ancestral reconstructions of the continuous characters to investigate possible thresholds for multistate and binary analyses. As a binary character, body size was coded as (0) small, SVL 10–25 mm, or (1) large, SVL≥25.01 mm; as a multistate character, body size was coded as (1) 10–17.5 mm; (2) 17.51–25 mm; (3) 25.01–40 mm; (4) 40.01–55 mm; and (5) ≥55.01 mm.
We used a continuous-time Markov model implemented in the Discrete module of BayesTraits – to investigate relationships between specific pairs of binary traits: (1) reproductive mode and habitat; (2) reproductive mode and maximum webbing; (3) reproductive mode and snout-vent length; (4) maximum pedal webbing and habitat; (5) maximum pedal webbing and snout-vent length; (6) snout-vent length and habitat. BayesTraits takes into account phylogenetic relationships among species for the across-species analysis of comparative data. To incorporate phylogenetic uncertainty, we used 27,000 post burn-in trees obtained generated from *BEAST analyses. Evidence for correlated evolution is obtained when the dependent model (H0) fits the data better than the independent model (H1). The likelihood ratio (LR) statistic compares the goodness-of-fit and is defined as follows: LR = 2 log [log-likelihood (H0)−log-likelihood (H1)]. The LR is then compared to χ2 with d.f. = 1 (χ2 = 3.84 with a confidence threshold of 0.05).
To assess the effects of reproductive mode, adult habitat preference, pedal webbing and body size on diversification of puddle frogs, we completed diversification analyses using a species tree generated from *BEAST. Differential diversification was tested using BiSSE  as implemented through the Diversitree package in R v. 2.11.1 ,  using discrete binary characters (previously outlined). We estimated the parameters λ0, λ1, μ0, μ1, q01 and q10 for the species-tree, allowing each binary character state to have unique diversification, extinction and transitions. BiSSE analyses were run with the correction for incomplete sampling unbiased for character state following FitzJohn et al. (2009) . Of the 82 recognized species of Phrynobatrachus , 58 were used in these analyses. We then ran a Likelihood Ratio Test (LRT) that compared this tree to a tree constrained to have equal diversification rates regardless of character state to determine if the character-dependent speciation model was a significant improvement over a character-independent model (for additional details regarding these analyses, see tutorial at http://www.zoology.ubc.ca/prog/diversitree/).
Phylogenetic species-tree relationships
Phylogenetic reconstructions using *BEAST yielded the same three major clades identified by Zimkus et al. (2010; Fig. 1, A–C) . All relationships within clade A are both well supported and concurrent with recent phylogenetic reconstructions . Similarly, most nodes are well supported within clade C, which differs only slightly from Zimkus et al. (2010) . Clade B shows somewhat different relationships from Zimkus et al. (2010)  with the ancestor of the group likely originating in West Africa instead of East Africa (Fig. 1). Several nodes within clade B have lower bootstrap support values due to missing data in the corresponding taxa. Additional sequence data may be required to resolve relationships within this group.
Species are color-coded according to their distribution within five geographic zones (see map inset): (1) western Africa, green; (2) central Africa, blue; (3) Albertine Rift, orange; (4) eastern Africa, red; and (5) southern Africa, purple. Each species marked with an asterisk (*) is present in two biogeographic zones and was color-coded according to the ancestral state of its most recent common ancestor . Species or genera coded in black are present in more than one zone and cannot be assigned unambiguously to a zone by ancestral state. Species that deposit terrestrial eggs or have terrestrial tadpoles are indicated by the egg and tadpole images on the right. Numbers above branches are posterior probabilities; asterisks (*) indicate values greater than 0.95. The three major clades of puddle frogs are indicated (A–C).
Evolution of reproductive mode and independence from water
Most Phrynobatrachus species breed in small, lotic bodies of water and have aquatic eggs with free-living, feeding tadpoles. Yet, reproductive modes that provide autonomy from permanent water bodies have evolved independently at least seven times (Figs. 1, 2). Both medium-sized and miniaturized taxa appear to exhibit alternate reproductive modes (Table S2). Only seven species of Phrynobatrachus are semi-independent or independent from temporary or permanent water bodies; all inhabit subtropical/tropical moist lowland forest and/or moist montane forest. Each species with low water dependence also exhibits an alternate reproductive mode (Figs. 1, 2). Reproductive mode and adult habitat did not show a strong phylogenetic signal (both: p = 1, λ<0.001) using Pagel's λ , thus ancestral character states were not reconstructed.
Ancestral state reconstructions for body size (maximum adult snout-vent length) and webbing (minimum number of phalanges free on toe IV for a given species) are mapped to each node using squared-change parsimony. Continuous snout-vent length values have been allocated among five bins: (1) 10.01–17.5 mm (red); (2) 17.51–25 mm (orange); (3) 25.01–40 mm (yellow); (4) 40.01–55 mm (green); (5) >55.01 mm (blue). Numbers of phalanges free of webbing on toe IV have been allocated among five bins: (1) webbing absent, 4.0 phalanges free (red); (2) webbing rudimentary, 3–3.9 phalanges free (orange); (3) webbing moderate, 2–2.9 phalanges free (yellow); (4) webbing moderate, 1–1.9 phalanges free (green); (5) webbing extensive, 0–0.9 phalanges free (blue). Gray circles denote either missing data or those internal nodes for which reconstructions are not possible. Species with alternate reproductive modes are shown in red. The three major clades of puddle frogs (A–C) are indicated.
Evolution of pedal webbing and body size
Pedal webbing and body size both showed a strong phylogenetic signal (webbing: p = 0.0005, λ = 1; body size: p<0.0001, λ = 1) using Pagel's λ ; therefore, ancestral states could be inferred for these two traits. Ancestral state reconstructions reveal a trend towards reduction of pedal webbing within each major clade of puddle frogs. The basal-most clade (Fig. 2, A; Fig. S1) includes species with moderate-to-extensive webbing (1–3.9 phalanges free of webbing on toe IV). The second major clade (B) includes species with absent, rudimentary or moderate webbing (2–4 phalanges free). No species within this clade exhibits extensively or completely webbed toes, and only one species, P. uzungwensis, exhibits moderate webbing (1–1.9 phalanges free). The third major clade (C) exhibits extensive variation among species, including absent, rudimentary, moderate and extensive webbing. The most recent common ancestor (mrca) of the puddle frog lineage is reconstructed as having moderate webbing (2–2.9 phalanges free).
Ancestral state reconstructions of adult body size reveal patterns similar to those for pedal webbing within each of the three major clades. The basal-most clade (Fig. 2, A; Fig. S2) includes most species with mean SVL greater than 25 mm and is the only clade of puddle frogs with species that exceed 40 mm. Evolution of extremely small body size (mean SVL less than 17.5 mm) has occurred ten times within the two remaining major clades (B, C). All species within the second major clade (B) have mean SVL less than 25 mm, and several of these species are smaller than 17.5 mm. The last major clade (C) also exhibits repeated reduction of body size, with seven independent instances of reduction from an ancestral SVL between 25 and 39.9 mm to less than 25 mm. The reconstructed SVL of the mrca of the phrynobatrachid lineage—58 mm—is significantly larger than most extant species.
Comparative analyses and diversification rates
Reproductive mode is highly correlated with adult habitat (Fig. 3A), but there is no correlation with pedal webbing. Although likelihood-ratio test statistics for reproductive mode and snout-vent length are not significant, some correlation is apparent (Fig. 3B). Likelihood-ratio tests indicate that body size is correlated with pedal webbing and adult habitat (Fig. 3C, D), yet adult habitat is not correlated with pedal webbing. In BiSSE analyses corrected for random incomplete sampling, only reproductive mode is associated with a change in diversification rate (LRT, p = 0.0487; Table 1). Though independence from temporary or permanent water bodies, pedal webbing and body size were non-significant in relation to changes in diversification rates (p = 0.0592, p = 0.0741, and p = 0.2160, respectively; Table 1), there is a clear relationship between characteristics associated with terrestriality and decreased diversification rate. Extinction rates were consistently not impacted by character state.
Frequency distributions of likelihood-ratio values were estimated from 27,000 trees (post burn-in trees obtained from *BEAST). Dashed lines indicate significant p-values at the 5% threshold. (a) Reproductive mode and adult habitat; (b) reproductive mode and snout-vent length; (c) maximum pedal webbing and snout-vent length; (d) adult habitat and snout-vent length.
We reject the null hypothesis of temporal congruence with a single, common biogeographic event that explains the terrestrialization of individual Phrynobatrachus species, using the confidence intervals to discriminate between hypotheses (Fig. 1). Regional temporal congruence also is rejected: many species within the same or adjacent geographic zones that exhibit alternate reproductive modes did not diversify contemporaneously (Fig. 1). Species within adjacent geographic zones that exhibit terrestrialization also do not exhibit temporal congruence; P. dendrobates (Albertine) and P. krefftii (East) have non-overlapping confidence intervals, and P. sandersoni (Central) and the four species from West Africa that exhibit alternate reproductive modes also have non-overlapping confidence intervals. Confidence intervals of four of the six West African pairs do overlap (P. guineensis and P. tokba; P. phyllophilus and P. tokba; P. guineensis and P. villersi; P. tokba and P. villersi).
Key innovations can confer the adaptive potential required to utilize a resource in a way not previously possible, and such traits may be correlated with subsequent species diversification if radiations into new environmental niches are successful. The transition from water to land is one of the major evolutionary milestones in the history of life, which typically requires a number of significant changes to both morphology and ecology. We thus hypothesized that traits characterizing terrestrialization in puddle frogs might represent key innovations, which enable colonization of environments that are uninhabited by aquatic species. Indeed, puddle frogs have repeatedly become independent from aquatic habitats, with alternate reproductive modes and low water dependence evolving multiple times independently. Yet, and contrary to our hypothesis, puddle frog clades with alternate reproductive modes and low water dependence actually have lower rates of diversification than clades that exhibit biphasic life histories with aquatic egg deposition and high water dependence. Though the traits that characterize terrestrialization enabled species to survive without permanent sources of water, complete terrestriality did not lead to increased diversification. Therefore, we conclude that even though an intrinsic change may have provided the adaptive potential for radiation, without an increase in moist terrestrial climates to facilitate this lifestyle, terrestriality did not increase the ecological opportunity for diversification in the Phrynobatrachus.
Most puddle frog species exhibit aquatic egg development, but terrestrial reproductive modes appears to have evolved in Phrynobatrachus at least seven times. This recurring trend towards terrestrialization is intriguing: although it may have allowed colonization of new niches, it has neither maintained nor elevated diversification rates. Instead, diversification rates decrease significantly in clades that evolve traits associated with terrestrialization. While this is a significant trend in Phrynobatrachus, a correlation between terrestrial development and decreased diversification remains to be investigated in anurans more generally. In addition, plasticity of reproductive mode is understood only superficially in anurans . In addition to confirming the reproductive mode(s) of species for which we hypothesized aquatic reproduction, future investigations of terrestrialization in Phrynobatrachus should also consider abiotic factors, such as light level, temperature, and humidity, to determine the degree of plasticity of reproductive mode (i.e., aquatic and/or terrestrial egg deposition) within individual species. Although reproductive plasticity, if found in Phrynobatrachus, might obscure the adaptive significance of terrestrialized or aquatic strategies, it would provide a more complete understanding of the shift from aquatic to terrestrial development.
We hypothesized that miniaturization is correlated with terrestrialization, but our analysis reveals that these two trends generally are decoupled, even though each has evolved multiple times. Miniaturized body size and reduction of pedal webbing recur within each major clade of puddle frogs. Clade A, which includes large species with moderate-to-extensive webbing, displays only slight reduction in either body size or webbing relative to its most recent common ancestor. The second major clade (B) includes only small species in which webbing generally is rudimentary or absent. The smallest species within clade B are neither one another's closest relatives nor do they occupy the same geographic region. Instead, miniaturized body size evolved at least eight times independently within this group. Clade C also exhibits recurrent body size reduction. However, body sizes in this clade are generally larger than those in clade B, and there are rare instances of body size increase. Webbing is highly variable within clade C, ranging from absent or rudimentary in some species to extensive in others, but such differences are highly correlated with body size. Neither pedal webbing nor body size was correlated with diversification rate.
Reduction of pedal webbing in puddle frogs may be a direct consequence of miniaturization. It is well known that miniaturized amphibians reduce or even lose components of their limbs, including bones of the hands and feet and even whole digits , –. The correlation between body size and pedal webbing in puddle frogs also is not surprising given that most small species inhabit leaf-litter as adults (the largest species are semi-aquatic). The link between miniaturization and exploitation of leaf-litter and moss habitats has been discussed in relation to many other frog groups –. Similar correlations between body size and ecology are evident in other amphibians. For example, the largest salamander species are fully aquatic, whereas the smallest species are terrestrial or arboreal . It is somewhat surprising, however, that pedal webbing is not associated with either reproductive mode or water dependence. Although we hypothesized that reduction of pedal webbing would be advantageous to species that are independent of aquatic environments, it is possible that webbing confers an advantage, or at least does not confer a disadvantage, in terrestrial environments. Foot webbing can help maximize forward thrust during swimming, but it also may greatly increase the surface area of the foot that comes in contact with terrestrial surfaces, such as rocks or leaves, thereby conferring increased stability. Future studies should assess the degree of terrestriality of species within this lineage by quantifying time within aquatic environments to more fully resolve the correlation between ecology and pedal webbing.
Small body size is correlated with increased species diversity in numerous lineages , , . Although body size is not significantly correlated with diversification rate, it is highly probable that size reduction was a key factor in the successful radiation of puddle frogs across sub-Saharan Africa. Size decrease may have enabled smaller species to populate new microhabitats or niches, allowing them to coexist with larger puddle frog species. Whereas the overall trend of decreasing body size may have facilitated the successful radiation of this group, extreme miniaturization does not appear to lead to increased diversification. Interestingly, both miniaturization and direct development are hypothesized to be key traits that allowed the successful range expansion of African squeaker frogs (Arthroleptis) from Central Africa into adjacent biogeographic regions . However, the appearance of these characteristics at the base of Arthroleptis phylogeny and the predominant trend of increasing body size in that genus contrasts with the patterns of repeated terrestrialization and decreasing body size observed in Phrynobatrachus.
Temporal and spatial patterns of terrestrialization
The extensive geographic range, substantial intrapopulational variation and tiny adult body size of many species of Phrynobatrachus are some of the reasons that the systematics of this genus has until recently been poorly understood. However, our understanding of puddle frog diversity, like that of many other tiny vertebrates, has been greatly improved with the advent of molecular approaches. Recent identification of a large number of undescribed and often cryptic species, many of which are extremely small, and the construction of a robust molecular phylogeny allow us to begin to unravel the complex patterns of speciation within this lineage , , –. Though the current dataset and recently published phylogeny  provide much more clarity than ever before, greater geographical sampling and the inclusion of additional loci in future phylogenetic analyses would strengthen our understanding of relationships within a lineage that exhibits extreme diversity in both size and degree of terrestriality.
Terrestrial breeding modes have evolved repeatedly within Phrynobatrachus and are not linked to a single biogeographic event or recent time period. This complex evolutionary history contrasts with that of most of Africa's fauna and flora, which are significantly impacted by frequent, recent, continent- and regional-scale drying events , . For amphibians with restrictive habitat requirements (e.g., generally moist habitats), these fluctuations would have led species to adapt to dry habitats at continental or regional scales. This pattern was recently documented for the global radiation of toads, but speciation was scaled over a longer period and was not confined to recent Pleistocene changes . Puddle frogs exhibit a complex pattern characterized by multiple, asynchronous events that explain the origins of terrestrialized breeding strategies. Although a single temporal event cannot explain the Africa-wide terrestrialization of puddle frogs, temporal congruence of speciation of some West African species exhibiting alternate reproductive modes suggests that they may have become terrestrialized in response to a common local cause. Given the decrease in diversification rate that accompanies changes in reproductive mode, we conclude that evolution of terrestrialization is not a precursor to radiation in puddle frogs. It does, however, confer the adaptive potential for some species to successfully inhabit and diversify in areas without permanent water.
Evolution of pedal webbing. Ancestral state reconstructions of the extent of webbing (minimum number of phalanges free on toe IV) are mapped on each node using using maximum likelihood (ML) in Mesquite on a set of 27,000 post burn-in trees. Number of phalanges free of webbing on toe IV are allocated among five bins: (1) webbing absent, 4 phalanges free (white); (2) webbing rudimentary, 3–3.9 phalanges free (blue); (3) webbing moderate, 2–2.9 phalanges free (green); (4) webbing moderate, 1–1.9 phalanges free (yellow); (5) webbing extensive, 0–0.9 phalanges free (black). Red indicates the fraction of trees for which that node is not present; nodes with equivocal reconstructions are indicated in grey. The three major clades of puddle frogs (A–C) are indicated.
Ancestral state reconstructions for body size (maximum adult snout-vent length) mapped on each node using maximum likelihood (ML) in Mesquite on a set of 27,000 post burn-in trees. The color scheme corresponds to five discrete states used in reconstruction: (1) 10.01–17.5 mm (white); (2) 17.51–25 mm (blue); (3) 25.01–40 mm (green); (4) 40.01–55 mm (yellow); (5) >55.01 mm (black). Red indicates the fraction of trees for which that node is not present; nodes with equivocal reconstructions are indicated in grey. The three major clades of puddle frogs (A–C) are indicated.
Specimens examined, including snout-vent length (SVL) and pedal webbing.
We thank J. Campbell and C. Roelke (UTA), A. Channing (University of the Western Cape, South Africa), R. Drewes (CAS), M. Menegon (Museo Tridentino), M.-O. Rödel (Humboldt University Berlin), J. Rosado (MCZ) and D. Wake (MVZ) who made available specimens under their care. B. Zimkus extends sincere thanks to R. Kerney, L. Mahler, M. Munisi and D. Pawlos for assistance in the field. Many thanks to Mr. T. Hailu (Ethiopian Wildlife Conservation Department), Mr. M. Mushi (Tanzanian Commission on Science and Technology) and Dr. J. Keyyu (Tanzanian Wildlife Research Institute) for assistance with collection and export permits. We thank Y. Wu, C. Linnen and H. Müller for useful discussion.
Conceived and designed the experiments: BMZ. Performed the experiments: BMZ. Analyzed the data: BMZ LL. Wrote the paper: BMZ LL SL JH.
- 1. Galis F (2001) Key innovations and radiations. In: Wagner GP, editor. The character concept in evolutionary biology. San Diego: Academic Press. pp. 581–605.
- 2. Sahney S, Benton MJ, Ferry PA (2010) Links between global taxonomic diversity, ecological diversity, and the expansion of vertebrates on land. Biol Lett 6: 544–547.
- 3. Dial KP, Marzluff JM (1988) Are the smallest organisms the most diverse? Ecology 69: 1620–1624.
- 4. Marzluff JM, Dial KP (1991) Life-history correlates of taxonomic diversity. Ecology 72: 428–439.
- 5. Barraclough TG, Hogan JE, Vogler AP (1999) Testing whether ecological factors promote cladogenesis in a group of tiger beetles (Coleoptera: Cicindelidae). Proc R Soc B 266: 1061–1067.
- 6. Gittleman JL, Purvis A (1998) Body size and species-richness in carnivores and primates. Proc R Soc B 265: 113–119.
- 7. Owens IPF, Bennett PM, Harvey PH (1999) Species richness among birds: body size, life history, sexual selection or ecology? Proc R Soc B 266: 933–939.
- 8. Hodges S (1997) Floral nectar spurs and diversification. Int J Plant Sci 158: S81–S88.
- 9. Farrell BD, Sequeira AS (2004) Evolutionary rates in the adaptive radiation of beetles on plants. Evolution 58: 1984–2001.
- 10. Ree R (2005) Detecting the historical signature of key innovations using stochastic models of character evolution and cladogenesis. Evolution 59: 257–265.
- 11. Alfaro ME, Santini F, Brock CD, Alamillo H, Dornburg A, et al. (2009) Nine exceptional radiations plus high turnover explain species diversity in jawed vertebrates. Proc Natl Acad Sci USA 106: 13410–13414.
- 12. Martin CH, Wainwright PC (2010) Trophic novelty is linked to exceptional rates of morphological diversification in two adaptive radiations of Cyprinodon pupfishes. Evolution 65: 2197–2212.
- 13. Duellman WE, Trueb L (1994) Biology of Amphibians. Baltimore: Johns Hopkins Univ. Press.
- 14. Haddad CFB, Prado CPA (2005) Reproductive modes in frogs and their unexpected diversity in the Atlantic forest of Brazil. Bioscience 55: 207–217.
- 15. Jameson DL (1955) Evolutionary trends in the courtship and mating behavior of Salientia. Syst Zool 4: 105–119.
- 16. Goin CJ (1960) Amphibians, pioneers of terrestrial breeding habits. Rep Smithson Inst 1959: 427–455.
- 17. Crump ML (1974) Reproductive strategies in a tropical anuran community. Misc Publ Mus Nat Hist, Univ Kansas 61: 1–68.
- 18. Salthe SN, Mecham JS (1974) Reproductive and courtship patterns. In: Lofts B, editor. Physiology of the Amphibia, vol. 2. New York: Academic Press. pp. 309–521.
- 19. Lamotte M, Lescure J (1977) Tendances adaptatives a l'affranchissement du milieu aquatique chez les amphibiens anoures. Terre Vie 31: 225–311.
- 20. Bogart JP (1981) How many times has terrestrial breeding evolved in anuran amphibian? Monitore Zool Ital Suppl XV: 29–40.
- 21. Duellman WE (1989) Alternative life-history styles in anuran amphibians: evolutionary and ecological implications. In: Bruton MN, editor. Alternative Life-history Styles of Animals. Dordrecht, The Netherlands: Kluwer Academic. pp. 102–126.
- 22. Salthe SN, Duellman WE (1973) Quantitative constraints associated with reproductive mode in anurans. In: Vial JS, editor. Evolutionary Biology of the Anurans: Contemporary Research on Major Problems. Columbia, MO: Univ. of Missouri Press. pp. 229–249.
- 23. Clarke BT (1996) Small size in amphibians: its ecological and evolutionary implications. Symp Zool Soc Lond 69: 201–224.
- 24. Blanckenhorn WU (2000) The evolution of body size: what keeps organisms small? Quart Rev Biol 75: 385–407.
- 25. Swedmark B (1964) The interstitial fauna of marine sand. Biol Rev 39: 1–42.
- 26. Higgins RP, Thiel H (1988) Introduction to the Study of Meiofauna. Washington, D.C.: Smithsonian Institution Press.
- 27. Gould SJ (1977) Ontogeny and Phylogeny. Cambridge, MA: Harvard University Press.
- 28. Nussbaum RA (1984) Amphibians of the Seychelles. In: Stoddart DR, editor. Biogeography and Ecology of the Seychelles Islands. The Hague, The Netherlands: W. Junk. pp. 379–415.
- 29. Trueb L, Alberch P (1985) Miniaturization and the anuran skull: a case study of heterochrony. In: Duncker HR, Fleischer G, editors. Functional Morphology of the Vertebrates. New York: Gustav Fischer Verlag. pp. 113–121.
- 30. Hanken J, Wake D (1993) Miniaturization of body size: organismal consequences and evolutionary significance. Ann Rev Ecol Syst 24: 501–519.
- 31. Wake MH (1986) The morphology of Idiocranium russeli (Amphibia: Gymnophiona), with comments on miniaturization through heterochrony. J Morphol 189: 1–16.
- 32. Yeh J (2002) The effect of miniaturized body size on skeletal morphology of frogs. Evolution 56: 628–641.
- 33. Blackburn DC (2008) Biogeography and evolution of body size and life history of African frogs: Phylogeny of squeakers (Arthroleptis) and long-fingered frogs (Cardioglossa) estimated from mitochondrial data. Mol Phylogenet Evol 49: 806–826.
- 34. Van Bocxlaer I, Loader SP, Roelants K, Biju SD, Menegon M, Bossuyt F (2010) Gradual adaptation towards a dispersal phenotype initiated the global radiation of toads. Science 327: 679–682.
- 35. Wollenberg KC, Vieites DR, Glaw F, Vences M (2011) Speciation in little: the role of range and body size in the diversification of Malagasy mantellid frogs. BMC Evol Biol 11: 217.
- 36. Wiens JJ, Pyron RA, Moen DS (2011) Phylogenetic origins of local-scale diversity patterns and the causes of Amazonian megadiversity. Ecol Lett 14: 643–652.
- 37. Frost DR (2011) Amphibian Species of the World: an online reference. Version 5.5 (31 January, 2011). Electronic Database available: http://research.amnh.org/vz/herpetology/amphibia/. American Museum of Natural History, New York, USA.
- 38. Zimkus BM, Blackburn DC (2008) Distinguishing features of the sub-Saharan frog genera Arthroleptis and Phrynobatrachus: a short guide for field and museum-based researchers. Breviora 513: 1–12.
- 39. Zimkus BM, Rödel M-O, Hillers A (2010) Complex patterns of continental speciation: molecular phylogenetics and biogeography of sub-Saharan puddle frogs (Phrynobatrachus). Mol Phylogenet Evol 55: 883–900.
- 40. Stewart MM (1967) Amphibians of Malawi. Albany, New York: State University of New York Press.
- 41. Rödel M-O (1998) A reproductive mode so far unknown in African ranids: Phrynobatrachus guineensis Guibé & Lamotte, 1961 breeds in tree holes (Anura: Ranidae). Herpetozoa 11: 19–26.
- 42. Altig R, McDiarmid RW (1999) Diversity: familial and generic characterizations. In: McDiarmid RW, Altig R, editors. Tadpoles: The Biology of Anuran Larvae. Chicago, IL: University of Chicago Press. pp. 295–337.
- 43. Rödel M-O (2000) Herpetofauna of West Africa Volume I. Amphibians of the West African Savanna. Frankfurt: Edition Chimaira.
- 44. Rödel M-O, Ernst R (2002) A new Phrynobatrachus from the Upper Guinean rain forest, West Africa, including a description of a new reproductive mode for the genus. J Herpetol 36: 561–571.
- 45. Heled J, Drummond AJ (2010) Bayesian inference of species trees from multilocus data. Mol Phylogenet Evol 27: 570–580.
- 46. Shapiro B, Rambaut A, Drummond AJ (2006) Choosing appropriate substitution models for the phylogenetic analysis of protein-coding sequences. Mol Biol Evol 23: 7–9.
- 47. Guindon S, Gascuel O (2003) A simple, fast and accurate method to estimate large phylogenies by maximum-likelihood. Syst Biol 52: 696–704.
- 48. Posada D (2008) jModelTest: Phylogenetic Model Averaging. Mol Biol Evol 25: 1253–1256.
- 49. Crawford AJ (2003) Relative rates of nucleotide substitution in frogs. J Mol Evol 57: 636–641.
- 50. Lawson LP (2010) The discordance of diversification: evolution in the tropical-montane frogs of the Eastern Arc Mountains of Tanzania. Mol Ecol 19: 4046–4060.
- 51. Rambaut A, Drummond AJ (2007) (2007) Tracer v1.4, Available: http://beast.bio.ed.ac.uk/Tracer.
- 52. Zimkus BM, Schick S (2010) Light at the end of the tunnel: insights into the molecular systematics of East African puddle frogs (Anura: Phrynobatrachidae). Syst Biodiv 8: 39–47.
- 53. Upchurch P (2008) Gondwanan break-up: Legacies of a lost world? Trends Ecol Evol 23: 229–236.
- 54. Kodandaramaia U (2011) Tectonic calibrations in molecular dating. Curr Zool 57: 116–124.
- 55. Loader SP, Pisani D, Cotton JA, Gower DJ, Day JJ, Wilkinson M (2007) Relative timescales reveal multiple origins of parallel disjunct distributions of African caecilian amphibians. Biol Lett 3: 505–508.
- 56. Burgess N, Hales JD, Underwood E, Dinerstein E, Olson D, et al. (2004) Terrestrial Ecoregions of Africa and Madagascar. Washington, DC.: Island Press.
- 57. Maddison WP, Maddison DR (2010) Mesquite: a modular system for evolutionary analysis. Version 2.73. Available: http://mesquiteproject.org.
- 58. Pagel M (1999) The maximum likelihood approach to reconstructing ancestral character states of discrete characters on phylogenies. Syst Biol 48: 612–622.
- 59. R Development Core Team (2010) R: A language and environment for statistical computing, reference index version 2.11.1. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0. Available: http://www.R-project.org.
- 60. IUCN (2011) IUCN Red List of Threatened Species. Version 2011.1. http://www.iucnredlist.org.
- 61. Pagel M (1999) The maximum likelihood approach to reconstructing ancestral character states of discrete characters on phylogenies. Syst Biol 48: 612–622.
- 62. Pagel M, Meade A, Barker D (2004) Bayesian estimation of ancestral character states on phylogenies. Syst Biol 53: 673–684.
- 63. Pagel M, Meade A (2006) Bayesian analysis of correlated evolution of discrete characters by reversible-jump Markov chain Monte Carlo. Am Nat 167: 808–825.
- 64. Maddison WP, Midford PE, Otto SP (2007) Estimating a binary character's effect on speciation and extinction. Syst Biol 56: 701–710.
- 65. FitzJohn RG, Maddison WP, Otto SP (2009) Estimating trait-dependent speciation and extinction rates from incompletely resolved phylogenies. Syst Biol 58: 595–611.
- 66. Touchon JC, Warkentin KM (2008) Reproductive mode plasticity: aquatic and terrestrial oviposition in a treefrog. Proc Natl Acad Sci USA 105: 7495–7499.
- 67. Hanken J (1982) Appendicular skeletal morphology in minute salamanders, genus Thorius (Amphibia: Plethodontidae): growth regulation, adult size determination, and natural variation. J Morphol 174: 57–77.
- 68. Hanken J (1985) Morphological novelty in the limb skeleton accompanies miniaturization in salamanders. Science 299: 871–874.
- 69. Blackburn DC (2008) Evolution of diversity in African frogs (Arthroleptis and Cardioglossa). Unpubl. Ph.D. dissertation, Harvard University.
- 70. Lehr E, Coloma LA (2008) A minute new Ecuadorian Andean frog (Anura: Strabomantidae, Pristimantis). Herpetologica 64: 354–367.
- 71. Lehr E, Catenazzi A (2009) A new species of minute Noblella (Anura: Strabomantidae) from southern Peru: the smallest frog of the Andes. Copeia 2009: 148–156.
- 72. Kraus F (2010) New genus of diminutive microhylid frogs from Papua New Guinea. Zookeys 48: 39–59.
- 73. Kraus F (2011) At the lower size limit for tetrapods, two new species of the miniaturized frog genus Paedophryne (Anura, Microhylidae). Zookeys 154: 71–88.
- 74. Wells KD (2007) The ecology and behavior of amphibians. Chicago, IL: University of Chicago Press.
- 75. Stanley SM (1973) An explanation for Cope's rule. Evolution 2: 1–26.
- 76. May RM (1978) The dynamics and diversity of insect faunas. In: Mound LA, Waloff N, editors. Diversity of Insect Faunas. Oxford: Blackwell Scientific. pp. 188–204.
- 77. Zimkus B (2009) Biogeographic analysis of puddle frogs across Cameroon and description of a new species of Phrynobatrachus (Anura: Phrynobatrachidae) endemic to Mount Oku, Cameroon. Zool J Linn Soc 157: 795–813.
- 78. Blackburn DC (2010) A new puddle frog (Phrynobatrachidae: Phrynobatrachus) from the Mambilla Plateau in eastern Nigeria. Afr J Herp 59: 33–52.
- 79. Schick S, Zimkus BM, Channing A, Köhler J, Lötters S (2010) Systematics of ‘Little Brown Frogs’ from East Africa: Recognition of Phrynobatrachus scheffleri and description of a new species from the Kakamega Forest, Kenya (Anura: Phrynobatrachidae). Salamandra 46(1): 24–36.
- 80. Hamilton AC (1988) Guenon evolution and forest history. In: Gautier-Hion A, Bourlière F, Gautier JP, Kingdon J, editors. A Primate Radiation: Evolutionary Biology of the African Guenons. Cambridge, UK: Cambridge University Press. pp. 13–34.
- 81. Kingdon J (1989) Island Africa. Princeton, NJ: Princeton University Press.