Open Access
How to translate text using browser tools
1 March 2006 THE AMPHIBIAN TREE OF LIFE
DARREL R. FROST, TARAN GRANT, JULIÁN FAIVOVICH, RAOUL H. BAIN, ALEXANDER HAAS, CÉLIO F.B. HADDAD, RAFAEL O. DE SÁ, ALAN CHANNING, MARK WILKINSON, STEPHEN C. DONNELLAN, CHRISTOPHER J. RAXWORTHY, JONATHAN A. CAMPBELL, BORIS L. BLOTTO, PAUL MOLER, ROBERT C. DREWES, RONALD A. NUSSBAUM, JOHN D. LYNCH, DAVID M. GREEN, WARD C. WHEELER
Author Affiliations +
Abstract

The evidentiary basis of the currently accepted classification of living amphibians is discussed and shown not to warrant the degree of authority conferred on it by use and tradition. A new taxonomy of living amphibians is proposed to correct the deficiencies of the old one. This new taxonomy is based on the largest phylogenetic analysis of living Amphibia so far accomplished. We combined the comparative anatomical character evidence of Haas (2003) with DNA sequences from the mitochondrial transcription unit H1 (12S and 16S ribosomal RNA and tRNAValine genes, ≈ 2,400 bp of mitochondrial sequences) and the nuclear genes histone H3, rhodopsin, tyrosinase, and seven in absentia, and the large ribosomal subunit 28S (≈ 2,300 bp of nuclear sequences; ca. 1.8 million base pairs; x̄ = 3.7 kb/terminal). The dataset includes 532 terminals sampled from 522 species representative of the global diversity of amphibians as well as seven of the closest living relatives of amphibians for outgroup comparisons.

The primary purpose of our taxon sampling strategy was to provide strong tests of the monophyly of all “family-group” taxa. All currently recognized nominal families and subfamilies were sampled, with the exception of Protohynobiinae (Hynobiidae). Many of the currently recognized genera were also sampled. Although we discuss the monophyly of genera, and provide remedies for nonmonophyly where possible, we also make recommendations for future research.

A parsimony analysis was performed under Direct Optimization, which simultaneously optimizes nucleotide homology (alignment) and tree costs, using the same set of assumptions throughout the analysis. Multiple search algorithms were run in the program POY over a period of seven months of computing time on the AMNH Parallel Computing Cluster.

Results demonstrate that the following major taxonomic groups, as currently recognized, are nonmonophyletic: Ichthyophiidae (paraphyletic with respect to Uraeotyphlidae), Caeciliidae (paraphyletic with respect to Typhlonectidae and Scolecomorphidae), Salamandroidea (paraphyletic with respect to Sirenidae), Leiopelmatanura (paraphyletic with respect to Ascaphidae), Discoglossanura (paraphyletic with respect to Bombinatoridae), Mesobatrachia (paraphyletic with respect to Neobatrachia), Pipanura (paraphyletic with respect to Bombinatoridae and Discoglossidae/Alytidae), Hyloidea (in the sense of containing Heleophrynidae; paraphyletic with respect to Ranoidea), Leptodactylidae (polyphyletic, with Batrachophrynidae forming the sister taxon of Myobatrachidae Limnodynastidae, and broadly paraphyletic with respect to Hemiphractinae, Rhinodermatidae, Hylidae, Allophrynidae, Centrolenidae, Brachycephalidae, Dendrobatidae, and Bufonidae), Microhylidae (polyphyletic, with Brevicipitinae being the sister taxon of Hemisotidae), Microhylinae (poly/paraphyletic with respect to the remaining non-brevicipitine microhylids), Hyperoliidae (para/polyphyletic, with Leptopelinae forming the sister taxon of Arthroleptidae Astylosternidae), Astylosternidae (paraphyletic with respect to Arthroleptinae), Ranidae (paraphyletic with respect to Rhacophoridae and Mantellidae). In addition, many subsidiary taxa are demonstrated to be nonmonophyletic, such as (1) Eleutherodactylus with respect to Brachycephalus; (2) Rana (sensu Dubois, 1992), which is polyphyletic, with various elements falling far from each other on the tree; and (3) Bufo, with respect to several nominal bufonid genera.

A new taxonomy of living amphibians is proposed, and the evidence for this is presented to promote further investigation and data

INTRODUCTION

Amphibians (caecilians, frogs, and salamanders) are a conspicuous component of the world's vertebrate fauna. They currently include 5948 recognized species with representatives found in virtually all terrestrial and freshwater habitats, in all but the coldest and driest regions or the most remote oceanic islands. The number of recognized species of amphibians has grown enormously in recent years, about a 48.2% increase since 1985 (Frost, 1985, 2004, unpubl. data). This growth reflects the increasing ease of collecting in remote locations and a significant growth of active scientific communities in a few megadiverse countries. Unfortunately, the rapid increase in knowledge of amphibian species diversity is coincident with a massive and global decline in amphibian populations (Alford and Richards, 1999; Houlahan et al., 2000; Young et al., 2001; S.N. Stuart et al., 2004) due to a diversity of factors, including habitat loss and fragmentation (Green, 2005; Halliday, 2005) but also possibly due to global environmental changes (Donnelly and Crump, 1998; Blaustein and Kiesecker, 2002; Heyer, 2003; Licht, 2003) and such proximate causes as emerging infectious diseases (Collins and Storfer, 2003).

Understanding of amphibian evolutionary history has not kept pace with knowledge of amphibian species diversity. For all but a few groups, there is only a rudimentary evolutionary framework upon which to cast the theories of cause, predict which lineages are most likely to go extinct, or even comprehend the amount of genetic diversity being lost (Lips et al., 2005). Indeed, it is arguable whether our general understanding of frog phylogenetics has progressed substantially beyond the seminal works of the late 1960s to early 1980s (Inger, 1967; Kluge and Farris, 1969; J.D. Lynch, 1971, 1973; Farris et al., 1982a). The major advances in frog taxonomy in the 1980s and 1990s were dominated by nomenclatural and largely literature-based phenotypic sorting (e.g., Dubois, 1980, 1981, 1984b; Laurent, 1986; Dubois, 1987 “1986”, 1992) that provided other workers with digestible “chunks” to discuss and evaluate phylogenetically. This has begun to change in the 2000s with the infusion of significant amounts of molecular evidence into the discussion of large-scale amphibian diversification. But, although recent molecular studies have been very illuminating (e.g., Biju and Bossuyt, 2003; Darst and Cannatella, 2004; Faivovich et al., 2005; Roelants and Bossuyt, 2005; San Mauro et al., 2005), so far they have not provided the general roadmap for future research that a larger and more detailed study could provide.

Among the three major taxonomic components of amphibian diversity, caecilians appear to have been the focus of the most significant study of large-scale evolutionary history (Gower et al., 2002; Gower and Wilkinson, 2002; M. Wilkinson et al., 2002; M. Wilkinson et al., 2003; San Mauro et al., 2004; M.H. Wake et al., 2005), although this may be an artifact of the relatively small size of the group (173 species currently recognized) and the few, mostly coordinated, workers. Salamanders are the best-known group at the species level, but salamander phylogenetic work has largely focused on the generic and infrageneric levels of investigation (e.g., Zhao, 1994; Titus and Larson, 1996; Highton, 1997, 1998, 1999; García-París and Wake, 2000; Highton and Peabody, 2000; Jockusch et al., 2001; Parra-Olea and Wake, 2001; Jockusch and Wake, 2002; Parra-Olea et al., 2002; Steinfartz et al., 2002; Parra-Olea et al., 2004; Sites et al., 2004), although there have been several important efforts at an overall synthesis of morphological and molecular evidence (Larson and Dimmick, 1993; Larson et al., 2003; Wiens et al., 2005).

Research on frog phylogenetics has also focused primarily on generic and infrageneric studies (e.g., Graybeal, 1997; Cannatella et al., 1998; Mendelson et al., 2000; Sheil et al., 2001; Channing et al., 2002a; Dawood et al., 2002; Faivovich, 2002; Glaw and Vences, 2002; Pramuk, 2002; Cunningham and Cherry, 2004; Drewes and Wilkinson, 2004; B.J. Evans et al., 2004; Pauly et al., 2004; Crawford and Smith, 2005; Matsui et al., 2005), and broader discussions of frog phylogenetics have been predominantly narrative rather than quantitative (e.g., Cannatella and Hillis, 1993; Ford and Cannatella, 1993; Cannatella and Hillis, 2004). Illuminating large-scale studies have appeared recently (Biju and Bossuyt, 2003; Haas, 2003; Darst and Cannatella, 2004; Roelants and Bossuyt, 2005; Van der Meijden et al., 2005; Faivovich et al., 2005). Nevertheless, a study of a broad sampling of amphibians, based on a large number of terminals, has not been attempted to date.

A serious impediment in amphibian biology, and systematics generally, with respect to advancing historically consistent taxonomies, is the social conservatism resulting in the willingness of many taxonomists to embrace, if only tacitly, paraphyletic groupings, even when the evidence exists to correct them. The reason for this is obvious. Recognizing paraphyletic groups is a way of describing trees in a linear way for the purpose of telling great stories and providing favored characters a starring role. Because we think that storytelling reflects a very deep element of human communication, many systematists, as normal storytelling humans, are unwilling to discard paraphyly. Unfortunately, the great stories of science, those popular with the general public and some funding agencies, almost never evidence careful analysis of data and precise reasoning or language. And, for much of its history, systematics focused on great narrative stories about “adaptive radiations” and “primitive”, “transitional”, and “advanced” groups rather than the details of phylogeny. These stories were almost always about favored characters (e.g., pectoral girdle anatomy, reproductive modes) within a sequence of paraphyletic groupings to the detriment of a full and detailed understanding of evolutionary history.

When one deconstructs the existing taxonomy of frogs, for example, one is struck by the number of groups delimited by very small suites of characters and the special pleading for particular characters that underlies so much of the taxonomic reasoning. Factoring in the systematic philosophy at the time many of these groups were named, both the origin of the problems and the illogic of perpetuating the status quo become apparent.

Our goal in this study is to provide remedies for the problems noted above, by way of performing a large phylogenetic analysis across all living amphibians and providing a taxonomy consistent with phylogeny that will serve as a general road map for further research. That such a diverse group of biologists (see list of authors) would be willing to set aside their legitimate philosophical differences to produce this work demonstrates the seriousness of the need. We hope that by providing considerable new data and new hypotheses of relationship that we will engender efforts to test our phylogenetic hypotheses and generate new ones. Regardless, the days are over of construing broad conclusions from small analyses of small numbers of taxa using small amounts of molecular or morphological data. We also think that the time is past for authoritarian classifications, rich in special pleading and weak on evidence (e.g., Dubois, 1992; Delorme et al., 2005; Dubois, 2005). In short, we hope that this publication will help change the nature of the conversation among scientists regarding amphibian systematics, moving it away from the sociologically conservative to the scientifically conservative. As noted by Cannatella and Hillis (2004: 444), the need for “scaling up” the rate of data collection is certainly evident (e.g., compare the evidentiary content of Cannatella and Hillis, 1993, with Cannatella and Hillis, 2004).

Nevertheless, even if we are successful in providing a roadmap for future work, this will not assure the health of amphibian systematics. Clearly, the task of understanding the evolution and ecological, morphological, and taxonomic diversity of amphibians is massive, yet funding remains insufficient to maintain a healthy amphibian systematics commmunity. Further, the institutional, interinstitutional, national and international infrastructure needed to promote the systematics research program needs to be greatly enhanced with respect to state-of-the-art collection facilities, digital libraries of all relevant systematic literature, interoperable collection databases, and associated GIS and mapping-related capacity, supercomputers and the improved analytical software to drive them, remotely accessible visualization instrumentation and specimen images, and enhanced data-aquisition technology, including massive through-put DNA sequencing, in addition to already-identified personnel, training, and financial needs related to exploring life on this planet and maintaining large research collections (Q.D. Wheeler et al., 2004; Page et al., 2005). There has been the salutary development of additional support in the training of systematists (e.g., Rodman and Cody, 2003) and important successes in increasing systematics capacity in a few megadiverse countries (e.g., Brazil; see de Carvalho et al., 2005), but it is also clear that increased research support is needed to assure another generation of evolutionary biologists capable of the detailed anatomical work to document how organisms have changed and diversified through time. But, especially in this time of increasing optimization of university hiring and retention policies on the ability of faculty to garner extramural funding, additional funding is needed to make sure that jobs exist for the systematists that are being trained.

About the Collaboration:

This collaboration was undertaken with the knowledge that everyone involved would have to compromise on deeply held convictions regarding the nature of evidence, methods of analysis, and what constitutes a reasonable assumption, as well as the nature of taxonomic nomenclature. Nevertheless, all data are provided either through GenBank or from  http://research.amnh.org/herpetology/downloads.html, and we expect several of the coauthors to deal in greater detail with the problems and taxonomic hypotheses noted in this paper, on the basis of even greater amounts of data with various taxonomic units within Amphibia and from their own points of view. We are unanimous in thinking that the capacity for systematic work needs to be expanded, and given existing university hiring and retention practices, this expansion can only take place through enhanced funding.

MATERIALS AND METHODS

Conventions and Abbreviations

Commands used in computer programs are italicized. Tissues are referenced in appendix 1 with the permanent collection number for the voucher specimen or, if that is unavailable, the tissue-collection number or field-voucher number. (See appendix 1 for acronyms.)

General Analytical Approach: Theoretical Considerations

Choice of Phylogenetic Method:

All phylogenetic methods minimize the number of character transformations required to explain the observed variation. Unweighted (equally weighted) parsimony analysis minimizes hypothesized transformations globally, whereas the assumptions (expressed as differential probabilities or costs) about the evolutionary process or perceived importance of different classes of transformations employed in statistical (maximum-likelihood, Bayesian analysis) and weighted parsimony methods minimize certain classes of transformations at the expense of others. Operational considerations aside (e.g., tree-space searching capabilities), disagreements between the results of unweighted parsimony analysis and the other methods are due to the increased patristic distance required to accommodate the additional assumptions. For this study, we chose to analyze the data under the minimal assumptions of unweighted parsimony. Given the size and complexity of our dataset, an important advantage of parsimony algorithms (whether weighted or unweighted) is that thorough analysis could be achieved in reasonable times given currently available hardware and software.

Nucleotide Homology and the Treatment of Insertions/Deletions (Indels):

The method of inferring nucleotide homology (i.e., alignment) and insertions/deletions (indels) and the treatment of indels in evaluating phylogenetic hypotheses are critically important in empirical studies. A given dataset aligned according to different criteria or under different indel treatments may strongly support contradictory solutions (e.g., W.C. Wheeler, 1995; Morrison and Ellis, 1997). Many workers infer indels as part of their procedure to discover nucleotide homology but then either treat the inferred indels as nucleotides of unknown identity by converting gaps into missing data or eliminate gap-containing column vectors altogether, because they are believed to be unreliable or because the method of phylogenetic analysis does not allow them (Swofford et al., 1996). Others argue that indels provide evidence of phylogeny but believe, we think incorrectly, that sequence alignment and tree evaluation are logically independent and must be performed separately (e.g., Simmons and Ochoterena, 2000; Simmons, 2004).

We treat indels as evidentially equivalent to any other kind of inferred transformation and as a deductively inferred component of the explanation of DNA sequence diversity observed among the sampled terminals. Furthermore, because nucleotides lack the structural and/or developmental complexity necessary to test their homology separately, hypotheses of nucleotide homology can be evaluated only in reference to a topology (Grant and Kluge, 2004; see also Frost et al., 2001). In recognition of these considerations, we assessed nucleotide homology dynamically by optimizing observed sequences directly onto competing topologies (Sankoff, 1975; Sankoff et al., 1976), thereby heuristically evaluating competing hypotheses by simultaneous searching of tree space. This is achieved using Direct Optimization (W.C. Wheeler, 1994, 1996, 1998, 1999; Phillips et al., 2000; W.C. Wheeler, 2000, 2001, 2002, 2003a, 2003b, 2003c) as implemented in the computer program POY (W.C. Wheeler et al., 1996–2003).

Determination of nucleotide homology is treated as an optimization problem in which the optimal scheme of nucleotide homologies for a given topology is that which requires the fewest transformations overall—that is, that which minimizes patristic distance, thus providing the most parsimonious explanation of the observed diversity. Determining the optimal alignment for a given topology is NP-complete1 (Wang and Jiang, 1994). For even a minuscule number of sequences, the number of possible alignments is staggeringly large (Slowinski, 1998), making exact solutions impossible for any contemporary dataset, and heuristic algorithms are required to render this problem tractable.

Phylogenetic analysis under Direct Optimization, therefore, addresses two nested NP-complete problems. POY searches simultaneously for the optimal homology/topology combination, and search strategies must take into consideration the extent of the heuristic shortcuts applied at both levels. The details of our analyses are discussed below under Heuristic Homology Assessment and Heuristic Tree Searching, with the general approach being to increase the rigor at both levels as the overall search progresses. In any heuristic analysis, a balance is sought whereby the algorithmic shortcuts speed up analysis enough to permit a sufficiently large and diverse sample of trees and alignments to discover the global optimum during final refinement, but not so severe that the sampling is so sparse or misdirected that the global optimum is not within reach during final refinement. Ideally, indicators of search adequacy (e.g., multiple independent minimum-length hits, stable consensus; see Goloboff, 1999; Goloboff and Farris, 2001; Goloboff et al., 2003) should be employed to judge the adequacy of analysis, as is now reasonable in parsimony analysis of large prealigned datasets (e.g., as performed by the software package TNT; Goloboff et al., 2003). However, current hardware and software limitations make those indicators unreachable in reasonable amounts of time for our dataset analyzed under Direct Optimization. The adequacy of our analysis may only be judged intuitively in light of the computational effort and strategic use of multiple algorithms designed for large datasets.

Taxon Sampling

The 532 terminals (reflecting 7 outgroup species, 522 ingroup species [with three redundancies]) included in our analysis are given in appendix 1. Because this study is predominantly molecular, outgroup sampling was restricted to the closest living relatives of living amphibians and did not include fossil taxa. These included two mammals, two turtles, one crocodylian, one squamate, and a coelacanth as the root. Our study was not designed to identify the sister taxon of tetrapods, and our use of a coelacanth instead of a lungfish was due to expediency and not a decided preference for any particular hypothesis of tetrapod relationship.

The remaining 525 terminals were sampled from the three orders of living amphibians. Our general criteria were (1) availability of tissues and/or sequences on GenBank, and (2) representation of taxonomic diversity. Although taxonomic rank per se is meaningless, our taxon sampling was guided to a large degree by generic diversity. Experience suggested that this “genus-level” sampling would thoroughly sample the diversity of living amphibians. The median number of species per genus for living taxa is only three, something that we think has to do with human perception of similarity and difference, not evolutionary processes. Some genera (e.g., Eleutherodactylus, ca. 605 species) are so large and/or diverse that directed subsampling of species groups was required to evaluate likely paraphyly (e.g., with respect to Phrynopus).

Summarizing, our sample constituted about 8.8% of all species of Recent amphibians currently recognized, with approximately the same proportion of species diversity sampled from each order. Of the ca. 467 Recent amphibian genera2, 326 (69.8%) are represented in our sample. We targeted 17 species of caecilians, representing 16 genera of all 6 family groups. Among salamanders we sampled 51 species from 42 genera of all 10 families. The bulk of our ingroup sample focused on frogs, with 437 terminals targeted. The remaining 457 terminals represent 454 anuran species from ca. 269 genera and 32 anuran families. A more extensive discussion of the terminals and the rationale behind their choice is presented under “Review of Current Taxonomy”.

Character Sampling

Morphology:

The 152 transformation series of morphology were incorporated directly from Haas (2003). Of his original 156 transformations, the gap-weighted morphometric transformations 12 (relative larval dermis thickness), 83 (cornua trabeculae proportions), 116 (ratio of anterior ceratohyal processes), and 117 (relative depth of anterior ceratohyla emargination) were excluded from our analysis because POY is unable to address noninteger transformations. We did include Haas' transformation 102 (presence/ absence of larval ribs) which he excluded from analysis because of difficulty in scoring absences; its inclusion did not alter his final topology and provided us the opportunity to incorporate known occurence of larval ribs in our final hypothesis.

Of the 81 frog and 4 salamander species in Haas' (2003) study, our study overlaps in 41 anurans and 2 caudates. We did not combine into one virtual taxon morphology from one species and DNA sequences from another, even if putatively closely related. Although that would have allowed us to incorporate more (and potentially all) morphological data, and in some cases it probably would not have affected our results detrimentally, because of our general skepticism regarding the current understanding of amphibian relationships we were unwilling to assume the monophyly of any group prior to the analysis.

DNA Sequences:

In light of the differing levels of diversity included in this study, we sought to sample loci of differing degrees of variability (i.e., rates). From the mitochondrial genome, we targeted the mitochondrial H-strand transcription unit 1 (H1), which includes the 12S ribosomal, tRNAValine, and 16S ribosomal sequences, yielding approximately 2,400 base pairs (bp) generated in 5– 7 overlapping fragments. We also targeted the nuclear protein coding genes histone H3 (328 bp), rhodopsin (316 bp), tyrosinase (532 bp), seven in absentia (397 bp), and the nuclear 28S ribosomal gene (ca. 700 bp), giving a total of approximately 2,300 bp of nuclear DNA. Primers used in PCR amplification and cycle-sequencing reactions (and respective citations) are given in table 1. When possible, terminals for which we were unable to generate all fragments were augmented with sequences from GenBank (see appendices 1, 2) under the assumption that the tissues were actually conspecific. The amount of sequence/terminal varied (fig. 1) with a range from 490 bp (Limnonectes limborgi) to 4,790 (Eleutherodactylus pluviacanorus), and the mean being 3,554 bp (see appendix 1).

Laboratory Protocols

Whole cellular DNA was extracted from frozen and ethanol-preserved tissues (liver or muscle) using either phenol-chloroform extraction methods or the Qiagen DNeasy kit following manufacturer's guidelines. PCR amplification was carried out in 25 μl reactions using Amersham Biosciences puRe Taq Ready-To-Go Beads. The standard PCR program consisted of an initial denaturing step of 3 minutes at 94°C, 35–40 cycles of 1 minute at 94°C, 1 minute at 45–62°C, and 1–1.5 minutes at 72°C, followed by a final extension step of 6 minutes at 72°C. PCR-amplified products were cleaned using the ARRAYIT kit (TeleChem International) on a Beckman Coulter Biomek 2000 robot. Cycle-sequencing using BigDye Terminators v. 3.0 (Applied Biosystems) were run in 8 μl reactions, and this was followed by isopropanol-ethanol precipitation and sequencing on either an ABI 3700 or ABI 3730XL automated DNA sequencer. Sequences were edited in Sequencher (Gene Codes).

Given the magnitude and complexity of this project (over 8,500 sequences were generated), the potential for errors to accumulate from a variety of sources (e.g., mislabeled vials, contamination, mispipetting, incorrect naming of files) was a serious concern. We took several measures to avoid errors. Tissues, stock solutions (including DNA extracts), and diluted working solutions were stored separately. Extractions were done at different times in batches of no more than 30 samples. Filtered tips were used to manipulate stock DNA extracts. Multichannel pipettes were used whenever possible, and all PCR cleaning was done using a Beckman Coulter Biomek 2000 robot. We extracted 100 tissues twice independently and sequenced at least one locus of each to confirm sequence identity, and we distributed multiple specimens of 10 species among different batches and generated all sequences for each to confirm species identifications and sequence identities and detect errors.

Sequence Preanalysis: Heuristic Error Checking

Numerous steps were taken to detect errors in DNA sequences. As is standard practice, we generated sequences in forward and reverse directions. The ca. 2400 bp of H1 were generated in 5–7 overlapping fragments, which allowed further sequence confirmation. We also compared the sequences generated for multiple extractions of the same tissues, as well as multiple specimens of the same species. Using Sequencher (Gene Codes) we selected all edited sequences for a given locus and used the “assemble interactively” option to establish the threshold at which a given sequence would align with any other sequence, which allowed identical and nearly identical sequences to be isolated for inspection. We compared questionable sequences with those of confirmed identity and sequences in GenBank.

The sequences that passed these tests were then aligned using ClustalX (Thompson et al., 1997). The resulting alignments and neighbor-joining trees for each partition were examined to detect aberrant sequences and formatting errors (e.g., reverse-complements). Finally, preliminary phylogenetic analyses were performed, and the resulting topologies were used to identify terminals that required further scrutiny. Extreme variance from expected position suggested the possibility of error and caused us to perform experiments to confirm sequence identities. We clarify that no sequence was eliminated solely because it did not fit our prior notions of relationships. Rather, the topologies were used heuristically to single out terminals/sequences for reexamination.

Once sequence identities were confirmed, sequences derived from the independent DNA extractions were merged. With a few exceptions noted later, those from conspecific specimens were merged into chimeras (with polymorphisms coded as ambiguities) to reduce the number of terminals in the analysis, but all sequences are deposited separately in GenBank (appendix 1).

Molecular Sequence Formatting

To allow integration of incomplete sequence fragments (particularly those obtained from GenBank; see Taxon Sampling Strategy and Character Sampling Strategy, above), accelerate cladogram diagnosis, and reduce memory requirements under Iterative Pass Optimization, we broke complete sequences into contiguous fragments. (This also improves the performance of POY's implementation of the parsimony ratchet; see Heuristic Tree Searching, below.) We did so sparingly, however, as these breaks constrain homology assessment by prohibiting nucleotide comparisons across fragments, that is, it is assumed that no nucleotides from fragment X are homologous with any nucleotides from fragment Y. As the number of breaks increases, so too does the risk of overly constraining the analysis and failing to discover the globally optimal solution(s).

We, therefore, inserted as few breaks as were necessary to maximize the amount of sequence data included, minimize the insertion of terminal N's, and attain maximum-length fragments of about 500 bases (table 2). Breaks were placed exclusively in highly conserved regions (many of which correspond to commonly used PCR primers), as recovery of such highly invariable regions is largely alignment-method independent and the inserted breaks do not prevent discovery of global optima. These highly conserved regions were identified via preliminary ClustalX (Thompson et al., 1997) alignments under default parameters. Except for their usefulness in placing fragments derived from different PCR primers and detecting errors (see Sequence Preanalysis, above), these preliminary alignments were used solely for the purpose of identifying conserved regions; they did not otherwise inform or constrain our phylogenetic analysis. Once appropriate conserved regions were identified, fragments were separated by inserting ampersands (&). Thus, the multiple fragments of the mtDNA cluster remain in the same file and order. The resulting POY-formatted files can be obtained from  http://research.amnh.org/herpetology/downloads.html or from the authors.

Analytical Strategy

We analyzed all data simultaneously using the program POY (W.C. Wheeler et al., 1996–2003) v. 3.0.11a (released May 20, 2003) run on the AMNH Parallel Computing Cluster. We visualized results using Winclada (Nixon, 1999–2002) and performed additional searches of implied alignments by spawning NONA (Goloboff, 1993–1999) from Winclada (see below).

Heuristic Homology Assessment:

Numerous algorithms of varying degrees of exhaustiveness have been proposed to optimize unaligned data on a given topology. Our search strategy employed three Direct Optimization algorithms. In order of increasing exhaustiveness and execution time, these were Fixed States Optimization (W.C. Wheeler, 1999), Optimization Alignment (W.C. Wheeler, 1996), and Iterative Pass Optimization (W.C. Wheeler, 2003a). As an indication of the magnitude of the problem of analyzing this 532-terminal dataset, execution time for a single random-addition sequence Wagner build (RAS), without swapping, on a 1.7 GHz Pentium 4 Dell Inspiron 2650 running WindowsXP was 2.69 hours under Fixed States and 3.26 hours under Optimization Alignment.

Although Fixed States Optimization was proposed as a novel means of conceptualizing DNA-sequence homology (W.C. Wheeler, 1999), we employed it here simply as a heuristic shortcut. Because Fixed States is so much faster than the Optimization Alignment algorithm, it allowed us to sample more thoroughly the universe of trees. (The speed-up for multiple replicates is actually much greater than noted earlier for a random-addition sequence Wagner build, as generating the initial state set is the slowest step in Fixed States analysis.) The trees obtained in Fixed States analyses were then used as starting points for further analysis under Optimization Alignment. The potential exists for the globally optimal tree (or trees that would lead to the global optimum when swapped under a more exhaustive optimization algorithm) to be rejected from the pool of candidates under the heuristic. To minimize this risk, we also generated a smaller pool of candidate trees under Optimization Alignment. The resulting 10 optimal and near-optimal candidate trees were then submitted to final evaluation and refinement under Iterative Pass optimization using iterativelowmem to reduce memory requirements. (For details on tree-searching algorithms see Heuristic Tree Searching, below.)

We did not employ exact during most searches, although we did use that command in the final stages of analysis. To verify lengths reported in POY, we output the implied alignment (W.C. Wheeler, 2003b) and binary version of the optimal topology in Hennig86 format with phastwincladfile and opened the resulting file in Winclada (Nixon, 1999–2002), following the procedure of Frost et al. (2001). Because each topology may imply a different optimal alignment, when multiple optimal topologies were obtained we examined them separately by inputting each as a separate file using topofile. Examination of the implied alignments, whether formatted as Hennig files or as standard alignments (impliedalignment), grants another opportunity to detect errors in formatting or sequencing (e.g., reverse complements; see Sequence Preanalysis, above).

Heuristic Tree Searching:

Efficient search strategies for large datasets are to a certain degree dataset-dependent (Goloboff, 1999), and, as discussed above, common indicators of sufficiency are unrealistic given current technological limitations. Therefore, rather than apply a simple, predefined search strategy (e.g., 100 random-addition sequence Wagner builds + TBR branch swapping), we employed a variety of tree-searching algorithms in our analysis, spending more computing time on those that proved most fruitful. Tree fusing (Goloboff, 1999) and TBR swapping were performed at various points throughout the analysis, and optimal trees from different searches were pooled for final tree fusing and TBR swapping, all of which was refined by submitting optimal topologies to swapping and ratcheting (see below) under Iterative Pass Optimization (W.C. Wheeler, 2003a).

See table 3 for a summary of general searching techniques. Initial runs used the approxbuild heuristic to speed up building of starting trees, but the resulting trees required much more subsequent refinement, nullifying the initial speed-up. Remaining analyses were therefore run without approxbuild. We conducted searches without slop or checkslop, both of which increase the pool of trees examined by swapping suboptimal trees found during the search. Although these steps can be highly effective, initial trials showed they were too time-consuming for the dataset (especially under Iterative Pass, where they would also be most relevant).

A variant of Goloboff's (1999) tree drifting was also used to escape local optima. Although it is based loosely on Goloboff's algorithm, the implementation in POY differs significantly in the way it accepts candidate trees during the search (see Goloboff, 1999, for his accept/reject calculation). In POY, the probability of accepting a candidate tree that is equal to or worse than the current optimum (better trees are always accepted) is given by 1/(n + c − b), where c is the length of the candidate topology, b is the length of the current optimum (best), and n is a user-specified factor that decreases the probability of accepting a suboptimal tree, effectively allowing the user to control the ease with which the search will drift away from the current optimum (we used the default of 2).

The parsimony ratchet (Nixon, 1999) was proposed for analysis of fixed matrices. Given that there are no prespecified column vectors to be reweighted under dynamic homology, the original approach had to be modified. In the current version of POY, the ratchet is programmed to reweight randomly selected DNA fragments. Our data were divided into 41 fragments (see table 2), so ratchetpercent 15 randomly reweighted 7 fragments, regardless of their length or relative position. In our analyses we reweighted 15–35% of the fragments and applied weights of 2–8×.

As a complementary approach, we also performed quick searches (few random-addition sequence Wagner builds + SPR) under indel, transversion, and transition costs of 3: 1:1, 1:3:1, and 1:1:3 and included the resulting topologies in the pool of trees submitted to tree-fusing and refinement under equal weights, following the general procedure of d'Haese (2003). Reweighting in this method is not done stochastically and therefore differs from both Nixon's (1999) original and POY's implementation of the ratchet. However, because it weights sets of transformations drawn from throughout the entire dataset, it is likely to capture different patterns in the data and may be a closer approximation to the original ratchet than POY's implementation. Both approaches attempt to escape local optima.

We also performed constrained searches by using Winclada to calculate the strict consensus of trees within an arbitrary number of steps of the present optimal, saving the topology as a treefile, constructing the group-inclusion matrix (Farris, 1973) in the program Jack2Hen (W.C.Wheeler, unpublished; available at  http://research.amnh.org/scicomp/projects/poy.php), and then employing constraint in the subsequent searches. To calculate the consensus we included trees within 100–150 steps of the current optimum, the goal being to collapse enough branches for swapping to be effective, but only enough branches to make for significant speed-ups of RAS + swapping, while still allowing discovery of optimal arrangements within the polytomous groups (see Goloboff, 1999: 420). This is effectively a manual approximation of Goloboff's (1999) consensus-based sectorial search procedure, the main difference being that we collapsed branches based only on tree length and not relative fit difference (Goloboff, 1999; Goloboff and Farris, 2001).

Using constraint files generated in the same way, we also input the current optimum as a starting point for ratcheting. This strategy avoids spending time on RAS builds of the unconstrained parts of the tree (which tend to be highly suboptimal) and seeks to escape local optima in the same way as unconstrained ratcheting, discussed earlier. However, there is a tradeoff in that the arrangements may be less diverse (and therefore unable to find global optima) but are likely to be, on average, closer to the optimum score than those examined through RAS.

As a further manual approximation of sectorial searches, we analyzed subsets of taxa separately by defining reduced datasets with terminals files that listed only the targeted terminals. More rigorous searches (at least 100 RAS + TBR for each of the reduced datasets) of these reduced datasets were then performed, and the results were used to specify starting topologies for additional searching of the complete dataset.

As a final attempt to discover more parsimonious solutions in POY, we also rearranged branches of current optima manually. As a general search strategy this would obviously be highly problematic, if for no other reason than that it would bias results. However, we performed this step primarily to ensure that the “received wisdom” was evaluated explicitly in our analysis. Our procedure was to open the current optimum in Winclada, target taxa whose placement was strongly incongruent with current taxonomy, and move them to their expected positions (or place them in polytomies, depending on the precision of the expectations). The resulting topology was saved as a treefile that was read into POY as a starting topology for diagnosis and refinement (e.g., swapping, tree-fusing). In this way we were sure that the more heterodox aspects of our results were not due simply to failing to evaluate the orthodox alternatives in our searches.

We analyzed the final implied alignment obtained in the final searches under Iterative Pass Optimization (i.e., the optimal solution found through all searching in POY) by carrying out 10 independent ratchet runs of 200 iterations each, using the default reweightings (Nixon, 1999). This ensured that heuristic shortcuts employed in POY to speed up optimization did not prevent discovery of global optima. It also ensures that users of other programs will be able to duplicate our results given our alignment.

Parallel Computing:

All POY runs were parallelized across 95 or 64 processors of the AMNH 256-processor Pentium 4 Xeon 2.8 GHz Parallel Computing Cluster. Initial analyses divided replicates among 5 sets of 19 processors using controllers, that is, 5 replicates were run simultaneously, each parallelized across 19 processors. Although that strategy may lead to a more efficient parallel implementation of POY (Janies and Wheeler, 2001), a shortcoming is that catchslaveoutput, which saves all intermediate results to the standard error file, is disabled when controllers is in use. Consequently, crashes (e.g., due to HVAC failures and overheating) or maintenance reboots result in the irrecoverable loss of days or weeks of analysis. To avoid this problem in subsequent runs, we parallelized each replicate across all processors and ran replicates serially, which allowed recovery from interrupted runs by inputting the intermediate results as starting points.

Support Measures:

We calculated support using the implied alignment of the optimal hypothesis. That is, the values reported reflect the degree of support by the hypothesized transformation series and not by the data per se. It is preferable to evaluate support based on the unaligned data, as that provides a more direct assessment of evidential ambiguity. (That is, it is possible for a clade to appear strongly supported given a particular alignment, but for support to dissolve when an alternative alignment is considered, meaning that the support by the data themselves is ambiguous.) We based support measures on the implied alignment because (1) it is much less time-consuming than support calculation under dynamic homology, and we preferred to concentrate computational resources on searches for the optimal solution; and (2) these values are directly comparable to those reported in the majority of phylogenetic studies, which derive support values from a single, fixed alignment.

To estimate Bremer values (Bremer, 1994), we output the implied alignment and optimal trees in Hennig86 format using phastwincladfile, converted it to NEXUS format in Winclada, and then generated a NEXUS inverse-constraints batch file in PRAP (K. Müller, 2004), which was analyzed in PAUP* 4.0 (Swofford, 2002). Given time constraints, tree searches for the Bremer analysis were superficial, consisting of only 2 RAS + TBR per group. Jackknife frequencies were calculated from 1000 replicates of 1 RAS per replicate without TBR swapping; jackknife analysis was performed by spawning NONA from Winclada.

REVIEW OF CURRENT TAXONOMY, THE QUESTIONS, AND TAXON SAMPLING

In this section we review the existing taxonomy of living amphibians and explain which species we sampled and what the justifications were for this sampling3. We also examine the evidentiary basis of the current taxonomy in an attempt to evaluate which parts provide a scientific template on which to interpret evolutionary patterns and trends, and which parts form an arbitrary and misleading structure that are merely anointed by time and familiarity or, worse, by authority. The canonical issue is obviously monophyly, so the question becomes: Does our taxonomy reflect evolutionary (i.e., monophyletic) groups? And, regardless of that answer, what is the evidentiary basis of the claims that have been made about amphibian relationships? Can we sample taxa in such a way as to test those claims? In this section we have, where practical, provided specific evidence from the published record as it bears on these questions. The reader should bear in mind that much of the current taxonomy rests on subjective notions of overall similarity and the relative importance of certain characters to specific Linnaean ranks. Even where knowledge claims derive from phylogenetic analysis, the evidence can be highly contingent on a specific phylogenetic context. We have not attempted to provide comparable characters among the taxa because such a description has yet to be accomplished in a detailed way (but see J.D. Lynch, 1973, and Laurent, 1986, for general attempts) and is outside the scope of this study. A general study would obviously change both the delimitation of the characters and the levels of generality.

Comparability of Systematic Studies

Throughout the review of current taxonomy that follows, we make only passing reference to the various analytical techniques used by various authors. There are two reasons for this. Not only is a deep review of techniques of phylogenetic inference beyond the scope of this paper, but it probably would be impossible for us to put together a quorum of authors to support any view beyond that it is monophyletic taxa that we are attempting to apprehend.

Our main concerns regard the repeatability of systematic analyses and that readers understand that many, if not most, of the analyses cited in this section are not rigorously comparable. In morphological studies it is common practice to report on individual transformation series and the logic behind treating these transformations as additive or nonadditive or whether these transformations can be polarized individually or not. This makes these analyses repeatable because workers can duplicate data as well as analytical conditions.

DNA sequence studies, however, have tended not to provide the information necessary for independent workers to repeat analyses, regardless of the accessibility of the original sequence data. In most cases, authors align their sequences manually (which is necessarily idiosyncratic and nonrepeatable, even if one uses models of secondary structure to help). In cases where alignment is done under algorithmic control, it is common to not cite the indel, transversion, and transition costs that went into the alignment, rendering these alignments unrepeatable. Also, many authors “correct” alignments by eye without explaining what this means or what these corrections were, further removing alignment from the sphere of repeatability. (This “correction” almost always means that the trees become longer.)

Of concern, at least for the AMNH authors, is that the assumptions of alignment may not be consistent with the assumptions of analysis. For instance, an author may have aligned sequences using one transversion: transition cost ratio but subsequently analyzed those data under an evolutionary model that makes entirely different assumptions about these relative costs. If the alignment method is not adequately specified, as is common in published works (e.g., Pauly et al., 2004), one is at a loss to know what component of the ultimate tree structure is due to the assumptions of alignment or to the assumptions of analysis. To illuminate the underlying incomparability of many molecular studies, we have provided in the relevant figure legends, and where this information can be gleaned from the publication, the alignment costs and whether the sequence was excluded for being “unalignable” (generally meaning that the authors did not like the number of gaps required to align the sequences), the amount of sequence and from what genes, and the kind of analysis (parsimony, Bayesian, or maximum-likelihood), and, if some general model of nucleotide evolution was assumed, what that model was. Because we are alarmed by the lack of explicitness in the literature regarding underlying assumptions, we urge editors to require that these pieces of information to be included in any works that pass over their desks. Having provided this preface to our review of current taxonomy as a caveat for readers, we now embark on a peregrination through the evidentiary basis of current amphibian taxonomy.

Amphibia

For the purposes of this paper, we are concerned with amphibians not as the fictional “transitional” group from fishes to amniotes, but as the taxon enclosing the extant crown clades Gymnophiona (caecilians), Caudata (salamanders), and Anura (frogs), together forming Lissamphibia of Gadow (1901) and most recent authors (e.g., Milner, 1988, 1993, 1994; Ruta et al., 2003; Schoch and Milner, 2004) or Amphibia in the restricted sense of being the smallest taxon enclosing the living crown groups (cf. de Blainville, 1816; Gray, 1825; de Queiroz and Gauthier, 1992; Cannatella and Hillis, 1993, 2004). We concur with authors who restrict application of the name Amphibia to the living crown groups, so for this study we use the terms “Amphibia” and “Lissamphibia” interchangeably.

Testing lissamphibian monophyly and the relationships among the three crown groups of amphibians was and continues to be daunting because morphologically the groups are mutually very divergent and temporally distant from each other and from nonamphibian relatives. Furthermore, testing lissamphibian monophyly may be outside the ability of this study to address inasmuch as the major controversy has to do with the phylogenetic structure of various fossil groups. Most authors regard Lissamphibia as a taxon imbedded in Temnospondyli (e.g., Estes, 1965; Trueb and Cloutier, 1991; Lombard and Sumida, 1992) whereas others regard frogs to be temnospondyls and salamanders and caecilians to be lepospondyls (Carroll and Currie, 1975; Carroll et al., 1999; Carroll, 2000a; J.S. Anderson, 2001). Laurin (1998a, 1998b, 1998c) regarded Lissamphibia to be completely within Lepospondyli, but more recent work (e.g., Ruta et al., 2003) returned a monophyletic Lissamphibia to the temnospondyls. (See Lebedkina, 2004, and Schoch and Milner, 2004, for extensive reviews of the alternative views of phylogeny of modern amphibian groups.) Because none of these paleontological studies adequately addressed living diversity, we hope that future work will integrate data presented here with fossil taxa as part of the resolution of the problem.

Regardless of the consideration of fossil taxa, the choice of Recent outgroups for analysis is clearly based on knowledge of the relationships of major tetrapod groups. A coelacanth (Latimeria) represents a near-relative of tetrapods, and among tetrapods, several amniotes (Mammalia: Didelphis and Gazella; Testudines: Pelomedusa and Chelydra; Diapsida: Iguana and Alligator) represent the nearest living relatives of amphibians. Although our choice of outgroups is made specifically to root the ingroup tree, our choice of terminals will allow weak tests of the various hypotheses of amniote relationships. The alternative relationships suggested by various authors is large, and an extensive discussion of these alternatives is outside the scope of this paper. Nevertheless, we show four topologies in figure 2. The most popular tree of amniote groups among paleontologists is shown in figure 2A and reflects the preferred topology of Gauthier et al. (1988a, 1988b), although some authors suggested, also on the basis of morphological evidence, that turtles are the sister taxon of lepidosaurs, with archosaurs and mammals successively more distantly related (Rieppel and de Braga, 1996; fig. 2B). This position, however, was disputed by M. Wilkinson et al. (1997). Also relevant to our study, some recent DNA sequence studies have found turtles to form the sister taxon of archosaurs (Zardoya and Meyer, 1998; Iwabe et al., 2005; fig. 2C), and others found turtles to be the sister taxon of archosaurs to the exclusion of lepidosaurs, with mammals outside this group (Hedges and Poling, 1999; Mannen and Li, 1999; fig. 2D). Our data will provide a weak test of these alternatives.

Assuming lissamphibian monophyly, the relationships among the three major groups of living lissamphibians remain controversial. On the basis of a parsimony analysis of morphological data, Laurin (1998a, 1998b, 1998c) suggested that salamanders are paraphyletic with respect to caecilians (although Laurin himself considered this conclusion implausible). Previously published molecular data placed salamanders as the sister taxon of either caecilians (Larson, 1991; Feller and Hedges, 1998) or frogs (Iordansky, 1996; Zardoya and Meyer, 2000, 2001; San Mauro et al., 2004; Roelants and Bossuyt, 2005; San Mauro et al., 2005). The latter arrangement is most favored by morphologists (e.g., Trueb and Cloutier, 1991). Additional tests using morphological data of the relative placement of the living lissamphibians will require evaluation of fossils, such as Albanerpetontidae (McGowan and Evans, 1995; Milner, 2000; Gardner, 2001, 2002) and the putative Mesozoic and Tertiary caecilians, salamanders, and frogs (Estes, 1981; Jenkins and Walsh, 1993; Shubin and Jenkins, 1995; Sanchíz, 1998; Carroll, 2000a; Gao and Shubin, 2001, 2003).

Gymnophiona

Caecilians (6 families, 33 genera, 173 species) are found almost worldwide in tropical terrestrial, semiaquatic, and aquatic habitats. A reasonably well-corroborated cladogram exists for at least the major groups of caecilians (Nussbaum and Wilkinson, 1989; Hedges and Maxson, 1993; M. Wilkinson and Nussbaum, 1996, 1999; Gower et al., 2002; M. Wilkinson et al., 2002; fig. 3). Taxon sampling has not been dense and taxonomic assignments are almost certain to change with the addition of new taxa and evidence. Nevertheless, it appears that most caecilian taxa are monophyletic, with the exception of “Ichthyophiidae” with respect to Uraeotyphlidae (Gower et al., 2002) and “Caeciliidae”, which includes most of the diversity (93 species; 54% of all caecilians) and which is paraphyletic with respect to Typhlonectidae (M.H. Wake, 1977; M. Wilkinson, 1991) and possibly with respect to Scolecomorphidae (M.H. Wake, 1993; M. Wilkinson et al., 2003).

The following taxa were sampled:

Rhinatrematidae (2 genera, 9 species):

A South American group, Rhinatrematidae is hypothesized to be the sister taxon of remaining caecilians and is clearly composed of the most generally plesiomorphic living caecilians (Nussbaum, 1977, 1979; Duellman and Trueb, 1986; San Mauro et al., 2004). They retain a tail (a plesiomorphy) but share the putatively derived characters of high numbers of secondary annuli, having an os basale, and lacking the fourth ceratobranchial. We sampled one species each of the two nominal genera (Rhinatrema bivittatum and Epicrionops sp.) to optimize characters for the family appropriately and to test the monophyly of this group.

Ichthyophiidae (2 genera, 39 species) and Uraeotyphlidae (1 genus, 5 species):

Tropical Asian Ichthyophiidae was hypothesized to form the sister taxon of Uraeotyphlidae (M. Wilkinson and Nussbaum, 1996; San Mauro et al., 2004), or to include Uraeotyphlidae (cf. Gower et al., 2002), or, currently less corroborated, to be the sister taxon of Uraeotyphlidae plus stegokrotaphic caecilians (i.e., “Caeciliidae” + Scolecomorphidae + Typhlonectidae; Nussbaum, 1979; Duellman and Trueb, 1986). The morphological diagnosis of Ichthyophiidae is contingent on whether Uraeotyphlus is within or outside of Ichthyophiidae, but the presence of angulate annuli anteriorly in ichthyophiids remains an apomorphy among these phylogenetic hypotheses. We have sampled one striped Ichthyophis (Ichthyophis sp.) that is not suspected to be close to Uraeotyphlus and one unstriped Ichthyophis (I. cf. peninsularis), which we suspect (M. Wilkinson and D.J. Gower, unpubl. data) to be phylogenetically close to Uraeotyphlus. Monophyly of the endemic and monotypic Indian group Uraeotyphlidae is supported by the morphological character of the tentacle being positioned below the naris. Our sole sample of this taxon is Uraeotyphlus narayani.

Scolecomorphidae (2 genera, 6 species):

The East African Scolecomorphidae was suggested to form the sister taxon of “Caeciliidae” + Typhlonectidae (Nussbaum, 1979), but because this suggestion was based on one of the early phylogenetic studies of caecilians, the sampling over which this generalization was made was small. Subsequent studies from mtDNA (M. Wilkinson et al., 2003) and morphology (M.H. Wake, 1993; M. Wilkinson, 1997) suggested that Scolecomorphidae, like Typhlonectidae, is imbedded within “Caeciliidae”. The monophyly of Scolecomorphidae is well-corroborated by morphology (Nussbaum, 1979; M. Wilkinson, 1997). Nevertheless, we sampled members of each of the two nominal genera (Crotaphatrema tchabalmbaboensis and Scolecomorphus vittatus), both as a test of scolecomorphid monophyly and to help optimize molecular characters for the family to the appropriate branch4.

Typhlonectidae (5 genera, 14 species):

The South American Typhlonectidae is a morphologically well-corroborated taxon of secondarily aquatic caecilians (M.H. Wake, 1977; Nussbaum, 1979; M. Wilkinson, 1991), clearly derived out of “Caeciliidae”. Although there are several nominal genera of typhlonectids, because of the highly apomorphic nature and highly corroborated monophyly of the taxon we sampled only Typhlonectes natans.

“Caeciliidae” (21 genera, 100 species):

This nominal taxon can be diagnosed only in the sense of being coextensive with the “higher” caecilians (Stegokrotaphia of Cannatella and Hillis, 1993) in lacking a tail, having a stegokrotaphic skull, and not being either a scolecomorphid or typhlonectid. We chose taxa from within the pantropical “Caeciliidae” that on the basis of previously published results (M.H. Wake, 1993; M. Wilkinson et al., 2003) we predicted would illuminate the paraphyly of “Caeciliidae” with respect to the presumptively derivative groups Typhlonectidae and Scolecomorphidae. We sampled: Boulengerula uluguruensis (Africa), Caecilia tentaculata (South America), Dermophis oaxacae (Mexico), Gegeneophis ramaswanii (India), Geotrypetes seraphini (Africa), Herpele squalostoma (Africa), Hypogeophis rostratus (Seychelles), Schistometopum gregorii (Africa), and Siphonops hardyi (South America).

Caudata

Salamanders (10 families, 62 genera, 548 species) are largely Holarctic and Neotropical and are the best known amphibian group, even though their phylogeny is notoriously problematic because of the confounding effects of paedomorphy on interpreting their morphology by (Larson et al., 2003; Wiens et al., 2005). Apparently independent paedomorphic lineages include Cryptobranchidae, Proteidae, and Sirenidae, as well as various lineages within Ambystomatidae and Salamandridae. Larson et al. (2003) provided an extensive discussion of salamander systematics, offering detailed discussion of the existing issues, although much of the supporting evidence was not disclosed. Until recently, Larson and Dimmick (1993) provided the received wisdom on salamander relationships based on a combined analysis of morphology (29 transformation series) and molecules (177 informative sites from rRNA sequences; fig. 4). The branch associated with internal fertilization in their tree (all salamanders excluding Sirenidae, Cryptobranchidae, and Hynobiidae) is corroborated primarily by a number of morphological characters that are functionally related to the secretion of a spermatophore (Sever, 1990; Sever et al., 1990; Sever, 1991a, 1991b, 1992, 1994).

Gao and Shubin (2001) provided a parsimony analysis of DNA sequences and morphology (including relevant fossils) suggesting that Sirenidae is not the sister taxon of the remaining salamander families, but the sister taxon of Proteidae (fig. 5). Otherwise, their results were largely congruent with those of Larson and Dimmick (1993). The exemplars used for their family-group tree were not provided nor were the distribution of morphological characters sufficiently detailed to allow us to include their data. Further, Larson et al. (2003), on the basis of molecular data alone (the data themselves not presented or adequately described beyond noting that they are from nuclear rRNA and mtDNA sequences), suggested the tree shown in figure 6. Larson et al. (2003) also noted that phylogenetic analysis of most morphological characters, other than those associated with spermatophore production, do not support the monophyly of their Salamandroidea (sensu Duellman and Trueb, 1986; all salamander families other than Sirenidae, Hynobiidae, and Cryptobranchidae). Although we address salamander phylogeny through the application of a large amount of molecular data, we did not address the morphological data set presented by Larson and Dimmick (1993) and Gao and Shubin (2001, 2003) because of the lack of correspondence between our exemplars and theirs and because this would have required reconciliation of these data with the frog morphology data we did include, an undertaking that is outside the scope of this study.

Most recently, Wiens et al. (2005) provided an analysis that included additional characters of morphology and the addition of data from RAG-1 DNA sequences (fig. 7). These authors presented results from different analytical approaches (e.g., maximum-likelihood, Bayesian, parsimony). We illustrate only the parsimony analysis of morphology + molecules, which most closely approximates our own assumption set. A paper by San Mauro et al. (2005) provided substantially similar results using the RAG-1 gene also used by Wiens et al. (2005).

Sirenidae (2 genera, 4 species):

Sirenidae is a North American, pervasively paedomorphic taxon, whose members are obligately aquatic and possess large external gills and lack pelvic girdles and hind limbs as well as eyelids. Only two genera (Siren and Pseudobranchus) are recognized. Sirenidae has been considered the sister taxon of the remaining salamanders by most authors because of its lack of internal fertilization (this is assumed on the basis of its lacking spermatophore-producing glands and not on any observation regarding its reproductive behavior) and its primitive jaw closure mechanism (Larson and Dimmick, 1993). Other morphological similarities (such as external gills and reduced maxillae) shared with other obligate paedomorphs have been more-or-less universally considered by authors to be convergent. Nevertheless, Gao and Shubin (2001), on the basis of an analysis of living and fossil taxa, concluded that sirenids are the sister taxon of proteids (fig. 5). Wiens et al. (2005) suggested, on the basis of a parsimony analysis of DNA sequences and morphology, that sirenids are the sister taxon of all other salamanders (fig. 7), although their Bayesian analysis placed Sirenidae as the sister taxon of Salamandroidea, with Cryptobranchoidea outside the inclusive group. We selected representatives of each nominal genus: Siren lacertina, S. intermedia, and Pseudobranchus striatus.

Hynobiidae (7 genera, 46 species):

The Asian Hynobiidae and Asian and North American Cryptobranchidae are usually considered each others' closest relatives because they share the putatively plesiomorphic condition of external fertilization and have the m. pubotibialis and m. puboischiotibialis fused to each other (Noble, 1931; Larson et al., 2003; Wiens et al., 2005). Hynobiids have aquatic larvae and transformed adults, and they retain angular bones in the lower jaw (presumed plesiomorphies). Morphological evidence in support of monophyly of this group are septomaxilla absent (also absent in plethodontids and ambystomatids), first hypobranchial and first ceratobranchial fused (also in Amphiuma), second ceratobranchial in two elements, and palatal dentition replaced from the posterior of the vomer (also in ambystomatids; Larson and Dimmick, 1993). Our selection of hynobiid taxa was restricted to Ranodon sibiricus and Batrachuperus pinchoni. Larson et al. (2003) suggested that Onychodactylus, especially, and several genera that we could not obtain (e.g., Hynobius), are not bounded phylogenetically by these taxa, so our analysis will not provide a rigorous test of hynobiid monophyly.

Cryptobranchidae (2 genera, 3 species):

Cryptobranchids are a Holarctic group represented by only three species in two genera, Cryptobranchus (eastern North America) and Andrias (eastern temperate Asia). We included all three species, Cryptobranchus alleganiensis, Andrias davidianus, and A. japonicus, to test the monophyly of Andrias and optimize “family” evidence to the appropriate branch. The monophyly of Cryptobranchidae is not seriously in doubt as these giant, obligately paedomorphic salamanders are highly apomorphic in many ways, such as in lacking gills or functional lungs, and instead respiring across the extensive skin surface. Like Hynobiidae and Sirenidae (presumably), cryptobranchids lack internal fertilization.

Proteidae (2 genera, 6 species):

Proteidae is another obligate paedomorphic perennibranch clade considered to be monophyletic because of its loss of the maxillae (also very reduced in sirenids, apparently independently) and the basilaris complex of inner ear (also lost in sirenids, plethodontids, and some salamandrids), and because it has other characters associated with paedomorphy, such as lacking a m. rectus abdominis (Noble, 1931). Unlike sirenids, cryptobranchids, and hynobiids, but like other salamander families, proteids employ internal fertilization through use of a spermatophore (Noble, 1931). In our analysis, we included two species of Necturus (of North America), N. cf. beyeri and N. maculosus, but were unsuccessful in amplifying DNA of the only other genus, Proteus (which is found only in the western Balkans). Nevertheless, Trontelj and Goricki (2003) did study Proteus and provided molecular evidence consistent with the monophyly of Proteidae, and Wiens et al. (2005), also reporting on both Necturus and Proteus, subsequently provided strong evidence in favor of its monophyly.

Rhyacotritonidae (1 genus, 4 species):

Western North American Rhyacotriton was originally placed in its own subfamily within Ambystomatidae (Tihen, 1958) but was shown to be distantly related to ambystomatines by Edwards (1976), Sever (1992), and Larson and Dimmick (1993), who considered it to be a family distinct from Ambystomatidae. Wiens et al. (2005) considered, on the basis of their parsimony analysis, that Rhyacotritonidae is the sister taxon of Amphiumidae + Plethodontidae. Good and Wake (1992) provided the most recent revision. Rhyacotritonidae retains a reduced ypsiloid cartilage and has at least one apomorphy associated with the glandular structure of the cloaca (Sever, 1992). Inasmuch as the four species are seemingly very closely related and morphologically very similar, we sampled only Rhyacotriton cascadae, although this leaves the taxon's monophyly untested.

Amphiumidae (1 genus, 3 species):

The amphiumas of eastern North America have reduced limbs and are obligate aquatic paedomorphs. They have internal fertilization and a suite of morphological features that are associated with spermatophore formation and internal fertilization. Some authors have associated Amphiumidae with Plethodontidae (sharing fused maxillae and reproductive behavior patterns; e.g., Salthe, 1967; Larson and Dimmick, 1993) and recent molecular studies place them here as well (Wiens et al., 2005). The three species are very similar and share many apomorphies, so we restricted our sampling to Amphiuma tridactylum.

Plethodontidae (4 subfamilies, 27 genera, 374 species):

Plethodontidae includes the large majority of salamander species, with most being in North America, Central America, and South America, with Speleomantes found in Mediterranean Europe and Karsenia found in the Korean Peninsula (Min et al., 2005). The monophyly of the group is not seriously questioned, as its members share a number of morphological synapomorphies such as nasolabial grooves in transformed adults and the absence of lungs (found in other groups as well; Larson and Dimmick, 1993). Starting in 2004, and while this project was in progress, understanding of the evolution of Plethodontidae moved into a dynamic state of flux with the publication of a series of important studies addressing substantial amounts of DNA sequence data and morphology (Chippindale et al., 2004; Mueller et al., 2004; Macey, 2005; Wiens et al., 2005). Before 2004, plethodontid phylogeny appeared to be reasonably well understood (D.B. Wake, 1966; D.B. Wake and Lynch, 1976; J.F. Lynch and Wake, 1978; D.B. Wake et al., 1978; Maxson et al., 1979; Larson et al., 1981; Maxson and Wake, 1981; Hanken and Wake, 1982; J.F. Lynch et al., 1983; D.B. Wake and Elias, 1983; Lombard and Wake, 1986; D.B. Wake, 1993; Jackman et al., 1997; García-París and Wake, 2000; Parra-Olea et al., 2004) with the group putatively composed of two monophyletic subfamilies (fig. 8), Desmognathinae and Plethodontinae, although the morphological evidence for any suprageneric group other than Desmognathinae and Bolitoglossini (a tribe in Plethodontinae as then defined) was equivocal.

Desmognathines (2 genera, 20 species; Desmognathus + Phaeognathus) as traditionally understood share nine morphological characters suggested to be synapomorphies (Schwenk and Wake, 1993; Larson et al., 2003), although at least some of them may be manifestations of a single transformation having to do with the unique method of jaw closure: (1) heavily ossified and strongly articulated skull and mandible; (2) dorsoventrally flattened, wedge-like head; (3) modified anterior trunk vertebrae; (4) enlarged dorsal spinal muscles; (5) hindlimbs larger than forelimbs; (6) stalked occipital condyles; (7) enlarged quadratopectoralis muscles; (8) modified atlas; and (9) presence of atlantomandibular ligaments. Most species have a biphasic life history, but at least some species have either nonfeedling larvae or direct development (Tilley and Bernardo, 1993). Plethodontinae in the pre-2004 sense (fig. 8) did not have strong morphological evidence in support of its monophyly, although Lombard and Wake (1986) suggested that possessing three embryonic or larval epibranchials is synapomorphic. Within Plethodontinae were included three nominal tribes: Hemidactyliini, Plethodontini, and Bolitoglossini.

Hemidactyliini (5 genera, 33 species) was the only putative plethodontine group with free-living larvae and transformation into adults (although this is shared with most desmognathines). Lombard and Wake (1986) suggested that Hemidactylium is the sister taxon of Stereochilus + (Eurycea, Gyrinophilus, and Pseudotriton) but provided only a single morphological character (parietal with a distinct ventrolateral shelf) in support of the monophyly of this group.

Plethodontini (3 genera, 62 species), as traditionally understood, was a heterogeneous assemblage composed of Plethodon, Aneides, and the more distant Ensatina (1 nominal species, but likely containing many species under any meaningful definition of that term; see Highton, 1998). Lombard and Wake (1986) suggested two morphological characters in support of the monophyly of this group (radii expanded and fused to basibranchial, and presence of a posterior maxillary facial lobe).

As traditionally viewed (before 2004), Bolitoglossini (15 genera, 222 species) represented a highly-speciose group in the New World tropics and west-coastal North America, with isolated representation in Mediterranean Europe. The group was characterized by having a projectile tongue, although this also appears in other plethodontids.

Lombard and Wake (1986) proposed a (nonparsimonious) scenario in which they suggested 10 synapomorphies of Bolitoglossini, all associated with the structure and function of the tongue. They regarded the supergenus Hydromantes (Hydromantes + Speleomantes) to be the sister taxon of the supergenus Bolitoglossa + supergenus Batrachoseps (containing solely Batrachoseps) based on two synapomorphies. Elias and Wake (1983) discussed phylogeny within Bolitoglossini and suggested the topology Hydromantes [including Speleomantes] + (Batrachoseps (Nyctanolis + other bolitoglossine genera)). Synapomorphies given by Elias and Wake (1983) for Bolitoglossini are (1) urohyal lost; (2) radii fused to the basibranchial; (3) long epibranchials relative to the ceratobranchials; (4) second ceratobranchial modified for force transmission; (5) presence of a cylindrical muscle complex around the tongue; (6) juvenile otic capsule configuration. The synapomophry for Batrachoseps + Nyctanolis + other bolitoglossine genera was reduction in number of caudosacral vertebrae to two. For the supergenus Bolitoglossa (Nyctanolis + other genera of bolitoglossines, excluding Batrachoseps and supergenus Hydromantes), they suggested that having the tail base with complex of breakage specializations was synapomorphic and for the supergenus Bolitoglossa excluding Nyctanolis they suggested that fused maxillae was a synapomorphy.

As noted above, in 2004–2005 three studies appeared that transformed our understanding of plethodontid relationships (Mueller et al., 2004; Chippindale et al., 2004; Macey, 2005). Although there are three relevant analyses, there are only two data sets. Mueller et al. (2004; fig. 9) presented a Bayesian analysis of complete mtDNA genomes; this data set was reanalyzed by parsimony and extensively discussed by Macey (2005; fig. 10). Another data set and analysis of combined morphology and DNA sequence evidence was provided by Chippindale et al. (2004; fig. 11).

All three studies suggested strongly not only that Plethodontinae (as traditionally understood) is paraphyletic with respect to Desmognathinae, but that the traditional view of plethodontid relationships was largely mistaken, presumably due in part to the special pleading for particular characters that underpinned the older system of subfamilies and tribes. Mueller et al. (2004) found that all three of the traditionally recognized plethodontine tribes, Bolitoglossini, Hemidactyliini, and Plethodontini, are polyphyletic. Chippindale et al. (2004) found Hemidactyliini and Plethodontini to be polyphyletic, with Bolitoglossini insufficiently sampled to test its monophyly rigorously. Macey (2005; fig. 10) also rejected the monophyly of Bolitoglossini and Hemidactyliini, in his reanalysis of the data of Mueller et al. (2004). Mueller et al. (2004; fig. 9) placed Hemidactylium as the sister taxon of Batrachoseps (a bolitoglossine) and the remaining hemidactyliines as the sister of a group of bolitoglossines (excluding Hydromantes and Speleomantes). Chippindale et al. (2004; fig. 11) considered Hemidactylium to be the sister taxon of all other bolitoglossines and hemidactyliines, and the remaining hemidactyliines to form the sister taxon of Hemidactylium + bolitoglossines (Hydromantes and Speleomantes not analyzed).

Chippindale et al (2004; fig. 11) provided a new taxonomy, recognizing a newly formulated Plethodontinae (including Plethodontini and Desmognathinae of the older taxonomy). The sister taxon of Plethodontinae was not named in their taxonomy, the component parts being named Hemidactyliinae (for Hemidactylium alone), Spelerpinae (for the remainder of the old Hemidactyliini), and Bolitoglossinae (identical in content to the old Bolitoglossini, these authors not having studied Hydromantes sensu lato). Mueller et al. (2004), followed by Macey (2005), showed that Hydromantes (in the sense of including Speleomantes) is not imbedded in Bolitoglossini, as previously supposed, but is imbedded in Plethodontinae. Macey (2005) arrived at the same taxonomy as Chippindale et al. (2004), although Macey (2005) placed Hemidactylium (Hemidactyliinae) as the sister taxon of the remaining plethodontids.

Clearly, the analyses of mtDNA-sequence data by Mueller et al. (2004) and Macey (2005) and of nuDNA, mtDNA, and morphology by Chippindale et al. (2004) 5 are strongly discordant with previous (and more limited) morphological and molecular results. Because of the timing of the appearance of these papers, our selection of taxa was chosen to address the older, more traditional view but may provide a weak test of the new view of plethodontid phylogeny and taxonomy.

We included in our analysis Hemidactylium scutatum (Hemidactyliinae) as well as the more “typical” hemidactyliines (Spelerpinae of Chippindale et al., 2004, and Macey, 2005): Eurycea wilderae and Gyrinophilus porphyriticus.

Of the new Plethodontinae (composed of former Desmognathinae, Plethodontini, and supergenus Hydromantes of Bolitoglossini) we sampled broadly. We included one species of western Plethodon, P. dunni, and one species of eastern Plethodon, P. jordani. We also included Aneides hardii and Ensatina eschscholtzii. Mueller et al. (2004), based on analysis of mtDNA, rejected the monophyly of Plethodontini, placing Ensatina as the sister taxon of desmognathines. (In a parsimony analysis of the same data, Macey, 2005, placed Ensatina as the sister taxon of Hydromantes.) The monophyly of Plethodon, in particular, is controversial, with some authors (e.g., Larson et al., 1981; Mahoney, 2001) finding the western species to be closer to Aneides to the exclusion of eastern species, and others (e.g., Chippindale et al., 2004; Mueller et al., 2004; Macey, 2005) finding Plethodon and Aneides to be rather distantly related. We bracketed the diversity (Titus and Larson, 1996) of desmognathines (the pre-2004 Desmognathinae) by sampling Phaeognathus hubrichti, Desmognathus quadramaculatus, and D. wrighti. Of the supergenus Hydromantes, formerly in Bolitoglossini, we sampled Hydromantes platycephalus and Speleomantes italicus.

Of Bolitoglossinae we sampled 11 of the 14 nominal genera: supergenus Batrachoseps (B. attenuatus and B. wrightorum), and supergenus Bolitoglossa (Bolitoglossa rufescens, Cryptotriton alvarezdeltoroi, Dendrotriton rabbi, Ixalotriton niger, Lineatriton lineolus, Nototriton abscondens, Oedipina uniformis, Parvimolge townsendi, Pseudoeurycea conanti, and Thorius sp.).

Salamandridae (18 genera, 73 species):

Salamandridae is found more-or-less throughout the Holarctic, with the bulk of its phylogenetic and species diversity in temperate Eurasia. Salamandrids are characterized by strongly keratinized skin in adults (except for the strongly aquatic Pachytriton), in addition to two cranial characters (presence of a frontosquamosal arch and fusion of the premaxillaries [reversed in Pleurodeles + Tylototriton, and Chioglossa]).

Titus and Larson (1995) provided a phylogenetic tree on the basis of a study of mt rRNA and morphology data (fig. 12). Scholz (1995; fig. 13) obtained similar results on the basis of morphology and courtship behavior. Zacj and Arntzen (1999) also reported on phylogenetics of Triturus, showing (as did Titus and Larson, 1995) that it is composed of two groups: (1) Triturus vulgaris + Triturus marmoratus species groups; (2) and Triturus cristatus group, but not addressing its polyphyly. Steinfartz et al. (2002) reported on salamandrid phylogeny and substantiated the polyphyly of Triturus and of Mertensiella. Subsequently (and appearing after this analysis was completed), García-París et al. (2004b) partitioned the polyphyletic “Triturus” into three genera (Triturus, Lissotriton, and Mesotriton), based on the suggestions that (1) Triturus, sensu stricto (Triturus cristatus + T. marmoratus species groups) is most closely related to Euproctus; (2) Mesotriton (Triturus alpestris) is the sister taxon of a group composed of Cynops, Paramesotriton, and Pachytriton); and (3) Lissotriton (Triturus vulgaris species group) is of uncertain relationship to the other components, but does not form a monophyletic group with either Mesotriton or Triturus. García-París et al. (2004a: 602) also suggested that ongoing molecular work (evidence undisclosed), will show Euproctus to be paraphyletic and that Triturus vittatus will not be included within Triturus, the oldest available name for this taxon being Ommatotriton Gray, 1850.

We could not address these final issues, these appearing well after the manuscript was written, but we chose taxa that should allow the basic structure of salamandrid phylogeny to be elucidated. To bracket this suggested topology with appropriate taxonomic samples we chose Euproctus asper, Neurergus crocatus, Notophthalmus viridescens, Pachytriton brevipes, Paramesotriton sp., Pleurodeles waltl, Salamandra salamandra, Taricha sp., Triturus cristatus, and Tylototriton shanjing.

Dicamptodontidae (1 genus, 4 species):

The North American Dicamptodon is related to Ambystomatidae (Larson and Dimmick, 1993; fig. 4) and, like them, some populations are neotenic (Nussbaum, 1976). Like other salamandroid salamanders they have internal fertilization and a suite of morphological features associated with forming and collecting spermatophores. Dicamptodon differs from Ambystomatidae in glandular features of the cloaca and in attaining a large size, but is considered by most workers as the sister taxon of Ambystomatidae (e.g., Larson et al., 2003fig. 6; Wiens et al., 2005fig. 7). We sampled both Dicamptodon aterrimus and D. tenebrosus.

Ambystomatidae (1 genus, 31 species):

North American Ambystomatidae is a morphologically compact family having internal fertilization via a spermatophore and the suite of morphological characters that support this attribute. Some populations exhibit neotenic aquatic adults.

The last summary of phylogeny within the group based on explicit evidence was presented by Shaffer et al. (1991; see also Larson et al., 2003), who provided a cladogram based on 32 morphological transformation series and 26 allozymic transformation series. The basal dichotomy in this tree is between Ambystoma gracile + A. maculatum + A. talpoideum on one hand, and all other species of Ambystoma, on the other. We were unable to obtain any of these three species, but we did sample Ambystoma cingulatum, A. mexicanum and A. tigrinum. Ambystoma mexicanum and A. tigrinum are very closely related, and A. cingulatum is distantly related to them. This is a weaker test of monophyly than we would have liked because it does not include A. gracile, A. maculatum, or A. talpoideum. Further, Larson et al. (2003) suggested that, in addition to A. gracile, A. maculatum, and A. talpoideum, A. jeffersonianum, A. laterale, A. macrodactylum, and A. opacum were likely to be outside of the taxa bracketed by our species, although the evidence for this was not presented.

Anura

Frogs (32 families, ca. 372 genera, 5227 species) constitute the vast majority (88%) of living species of amphibians and the bulk of their genetic, physiological, ecological, and morphological diversity. Despite numerous studies that point towards its deficiencies (e.g. Kluge and Farris, 1969; Lynch, 1973; Sokol, 1975, 1977; Duellman and Trueb, 1986; Ruvinsky and Maxson, 1996; Maglia, 1998; Emerson et al., 2000; Maglia et al., 2001; Scheltinga et al., 2002; Haas, 2003; Roelants and Bossuyt, 2005; San Mauro et al., 2005; Van der Meijden et al., 2005), the current classification continues in many of its parts to reflect sociological conservatism and the traditional preoccupation with groupings by subjective impressions of overall similarity; special pleading for characters considered to be of transcendent importance; and notions of “primitive”, “transitional”, and “advanced” groups instead of evolutionary propinquity. Understanding of frog relationships remains largely a tapestry of conflicting opinion, isolated lines of evidence, unsubstantiated assertion, and unresolved paraphyly and polyphyly. Indeed, the current taxonomy of frogs is based on a relatively small sampling of species and in many cases the putative morphological characteristics of major clades within Anura are overly-generalized, overly-interpreted, and reified through generations of literature reviews (e.g., Ford and Cannatella, 1993), of which this review is presumably guilty as well. This general lack of detailed understanding of anuran relationships has been exacerbated by the explosive discovery of new species in the past 20 years.

Currently, the most widely cited review of frog phylogeny is Ford and Cannatella (1993; fig. 14), which provided a narrative discussion of the evidence for a novel view of frog phylogeny without providing all of the underlying data from which this discussion was largely derived. The result was that the extent of character conflict within their data set was never adequately exposed. More recently, Haas (2003; fig. 15) provided a discussion of frog evolution, based primarily on new larval characters. Haas did, however, exclude several of the adult characters included by Ford and Cannatella (1993) as insufficiently characterized or assayed. More recently, important discussions of phylogeny have been made in the context of DNA sequence studies (Roelants and Bossuyt, 2005fig. 16; San Mauro et al., 2005fig. 17) that will be cited throughout our review.

The monophyly of frogs (Anura) relative to other living amphibians has not been generally questioned6 (although the universality of this taxon with respect to some fossil antecedent taxa has (e.g., Griffiths, 1963; Roĉek, 1989, 1990), and the number of morphological characters corroborating this monophyly is large—e.g., (1) reduction of vertebrae to 9 or fewer; (2) atlas with a single centrum; (3) hindlimbs significantly longer than forelimbs, including elongation of ankle bones; (4) fusions of radius and ulna and tibia and fibula; (5) fusion of caudal vertebral segments into a urostyle; (6) fusion of hyobranchial elements into a hyoid plate; (7) presence of keratinous jaw sheaths and keratodonts on larval mouthparts; (8) a single median spiracle in the larva, a characteristic of the Type III tadpole (consideration of this as a synapomorphy being highly contingent on the preferred overall cladogram); (9) skin with large subcutaneous lymph spaces; and (10) two m. protractor lentis attached to lens, based on very narrow taxon sampling (Saint-Aubain, 1981; Ford and Cannatella, 1993).

Haas (2003) suggested (fig. 15) an additional 20 synapomorphies from larval morphology: (1) paired venae caudalis lateralis short; (2) operculum fused to abdominal wall; (3) m. geniohyoideus origin from ceratobranchials I–II; (4) m. interhyoideus posterior absent; (5) larval jaw depressors originate from palatoquadrate; (6) ramus maxillaris (cranial nerve V2) medial to the m. levator manidbulae longus; (7) ramus mandibularis (cranial nerve V3) anterior (dorsal) to the m. levator mandibulae longus; (8) ramus mandibularis (cranial nerve V3) anterior (dorsal) to the externus group; (9) cartilago labialis superior (suprarostral cartilage) present; (10) two perilymphatic foramina; (11) hypobrachial skeletal parts as planum hypobranchiale; (12) processus urobranchialis short, not reaching beyond the hypobranchial plates; (13) commisura proximalis I present; (14) commisura proximalis II present; (15) commisura proximalis III present; (16) ceratohyal with diarthrotic articulation present, medial part broad; (17) cleft between hyal arch and branchial arch I closed; (18) ligamentum cornuquadratum present; (19) ventral valvular velum present; (20) branchial food traps present. Haas also suggested that the following were synapomorphies not mentioned as such by Ford and Cannatella (1993): (1) amplexus inguinal; (2) vertical pupil shape; (3) clavicle overlapping scapula anteriorly; and (4) cricoid cartilage as a closed ring.

“Primitive” Frogs

We first address the groups that are sometimes referred to collectively as Archaeobatrachia (Duellman, 1975) and traditionally are considered “primitive”, even though the component taxa have their own apomorphies and the preponderance of evidence suggests strongly that they do not form a monophyletic group (Roelants and Bossuyt, 2005; San Mauro et al., 2005).

Ascaphidae (1 genus, 2 species):

Ford and Cannatella (1993) considered North American Ascaphus (Ascaphidae) to be the sister taxon of all other frogs (fig. 14), although on the basis of allozyme study by Green et al. (1989) and, more recently, Roelants et al. (2005; fig. 16) and San Mauro et al. (2005; fig. 17), on the basis of evidence from DNA sequences, suggested that Ascaphidae + Leiopelmatidae forms a monophyletic group. Báez and Basso (1996) presented a phylogenetic analysis designed to explore the relationships of the fossil anurans Vieraella and Notobatrachus with the extant taxa Ascaphus, Leiopelma, Bombina, Alytes, and Discoglossus. Despite their restricted taxon sampling, their results also support the monophyly of Ascaphus + Leiopelma, although the authors considered their evidence weak for reasons of difficulty in evaluating characters.

Green and Cannatella (1993) did not find a monophyletic Ascaphus + Leiopelma. Ascaphus and Leiopelma share the presence of a m. caudalipuboischiotibialis and nine presacral vertebrae (Ford and Cannatella, 1993), both considered plesiomorphic within Anura7. Ascaphus has an intromittant organ (apomorphic) in males and a highly modified torrent-dwelling tadpole. The vertebrae are amphicoelous and ectochordal (Nicholls, 1916; Laurent, 1986), presumably plesiomorphic at this level of generality. Our sampled species for this taxon is Ascaphus truei, one of the two closely-related species.

Leiopelmatidae (1 genus, 4 species):

Isolated in New Zealand, Leiopelmatidae, like Ascaphidae, is a generally very plesiomorphic group of frogs. Nevertheless, it possesses apomorphies, such as ventral inscription ribs, found nowhere else among frogs (Noble, 1931; Laurent, 1986; Ford and Cannatella, 1993). Unlike Ascaphus, Leiopelma does not have feeding larvae (Archey, 1922; Altig and McDiarmid, 1999; Bell and Wassersug, 2003). As in Ascaphidae, the vertebrae are amphicoelous and ectochordal with a persistent notochord (Noble, 1924; Ritland, 1955) and both vocal sacs and vocalization are absent (Noble and Putnam, 1931).

Ford and Cannatella (1993) suggested that Leiopelmatidae is the nearest relative of all other frogs (excluding Ascaphidae) and listed five synapomorphies in support of this grouping (their Leiopelmatanura): (1) elongate arms on the sternum; (2) loss of the ascending process of the palatoquadrate; (3) sphenethmoid ossifying in the anterior position; (4) exit of the root of the facial nerve from the braincase through the facial foramen, anterior to the auditory capsule, rather than via the anterior acoustic foramen into the auditory capsule; (5) palatoquadrate articulates with the braincase via a pseudobasal process rather than a basal process.

Characters 4 (facial nerve exit) and 5 (palatoquadrate articulation) are polarized with respect to salamanders; the other three characters were likely polarized on the assumption that Ascaphus is plesiomorphic and the sister taxon of remaining frogs, thereby presupposing the results, although this was not stated. With respect to character 1 (the triradiate sternum), the parsimony cost of this transformation on the overall tree is identical if Ascaphidae and Leiopelmatidae are sister taxa and Bombinatoridae and Discoglossidae are sister taxa. The remaining characters, 2 and 3, were not discussed with respect to outgroups or reversals in the remainder of Ford and Cannatella's tree, implying that they are unreversed and unique.

With Ascaphus, Leiopelma shares the apomorphy of columella not present (N.G. Stephenson, 1951). Haas (2003) did not include Leiopelma in his analysis of exotrophic larval morphology because of their endotrophy. We included in our analysis Leiopelma archeyi and L. hochstetteri, which bracket the phylogenetic diversity of Leiopelmatidae (E.M. Stephenson et al., 1974), although it is not sufficient to test hypotheses of the evolution of direct development (exoviviparity in this case; Thibaudeau and Altig, 1999) within Leiopelma.

Discoglossidae8 (2 genera, 12 species) and Bombinatoridae (2 genera, 10 species):

Ford and Cannatella (1993; fig. 14) suggested that Bombina + Barbourula forms the sister taxon of all other frogs, exclusive of Leiopelmatidae and Ascaphidae, although recent molecular evidence (Roelants and Bossuyt, 2005; fig. 16) placed Bombinatoridae and Discoglossidae in the familiar position of sister taxa.

Ford and Cannatella's (1993) arrangement (fig. 14; i.e., paraphyly of Bombinatoridae + Discoglossidae) required a partition of the traditionally recognized Discoglossidae (sensu lato) to place Bombina and Barbourula in their own family, Bombinatoridae. In their system, Bombinatoridae + its sister taxon (all frogs excluding Leiopelmatidae and Ascaphidae) was named Bombianura. Bombianura is corroborated by four synapomorphies: (1) fusion of the halves of the sphenethmoid; (2) reduction to eight presacral vertebrae; (3) loss of the m. epipubicus (regained in Xenopus); and (4) loss of the m. caudalipuboischiotibialis. In addition, Abourachid and Green (1999) noted that although Leiopelma and Ascaphus do hop, they swim with alternating sweeps of the hind legs (the presumably plesiomorphic condition), unlike those in Bombianura, which swim with coordinated thrusts of the hind limbs, a likely synapomorphy.

Bombinatoridae was considered (Ford and Cannatella, 1993) to have as synapomorphies (1) expanded flange of the quadratojugal, and (2) presence of endochondral ossifications in the hyoid plate (both unreversed). We sampled four species of Bombina: B. bombina, B. microdeladigitora, B. orientalis, and B. variegata. The genus may be monophyletic, but no rigorous phylogenetic study has been performed so far, and paraphyly of Bombina with respect to Barbourula remains an open question. We could not obtain tissues of Barbourula so its phylogenetic position will remain questionable. Bombina has aquatic feeding tadpoles, but larvae of Barbourula are unknown and are suspected to be endotrophic (Altig and McDiarmid, 1999). Discoglossidae (sensu stricto) also has free-living aquatic tadpoles (Boulenger, 1892 “1891”; Altig and McDiarmid, 1999).

Ford and Cannatella (1993; fig. 14) also posited a taxon, Discoglossanura, composed of Discoglossidae (sensu stricto) and the remaining frogs (exclusive of Ascaphidae, Leiopelmatidae, and Bombinatoridae) which they suggested to be monophyletic on the basis of two synapomorphies: (1) bicondylar sacrococcygeal articulation; and (2) episternum present. Monophyly of Discoglossidae (sensu stricto) was supported by their possession of (1) V-shaped parahyoid bones (also in Pelodytes) and (2) a narrow epipubic cartilage plate.

Haas (2003; fig. 15) presented a cladogram that is both deeply at variance with the relationships suggested by Ford and Cannatella (1993) and, at least with respect to this part of their cladogram, consistent with the molecular evidence presented by Roelants and Bossuyt (2005; fig. 16). Haas (2003) presented six morphological synapomorphies of Discoglossidae + Bombinatoridae (as Discoglossidae, sensu lato) and rejected Discoglossidae (sensu Ford and Cannatella) as paraphyletic, placing Alytes as the sister taxon of the remaining members of Discoglossidae + Bombinatoridae. Synapomorphies of Haas' Discoglossidae are: (1) origin of m. intermandibularis restricted to the medial face of the cartilago meckelii; (2) larval m. levator mandibulae externus present as two bundles (profundus and superficialis); (3) posterior processes of pars alaris double; (4) cartilaginous roofing of the cavum cranii present only as taenia traversalis; (5) vertebral centra formation epichordal; and (6) processus urobranchialis absent. Synapomorphies suggested by Haas (2003; fig. 15) for Discoglossidae, excluding Alytes are (1) epidermal melanocytes forming an orthogonal pattern; (2) advertisement call inspiratory; and (3) pupil an inverted drop-shape (triangular). Of Discoglossidae (sensu stricto), we sampled one species of Alytes (A. obstetricans) and two species of Discoglossus (D. galganoi and D. pictus). Discoglossidae and Bombinatoridae show opisthocoelous and epichordal vertebrae according to Mookerjee (1931), Griffiths (1963), and Haas (2003). Kluge and Farris (1969: 23) suggested that vertebral development in Discoglossus pictus is perichordal, although Haas (2003) reported it as epichordal.

Roelants and Bossuyt (2005; fig. 16) and, with denser taxon sampling, San Mauro et al. (2005; fig. 17) provided substantial amounts of DNA evidence suggesting strongly that Bombinatoridae + Discoglossidae forms a monophyletic group, thereby rejecting Discoglossanura, Leiopelmatanura, and Bombianura of Ford and Cannatella (1993).

“Transitional” Frogs

The following few groups traditionally have been considered “transitional” from the primitive to advanced frogs, even though one component taxon in particular, Pipidae, is highly apomorphic in several ways. The monophyly of this collection of families was supported by some authors (e.g., Ford and Cannatella, 1993; García-París et al., 2003), but recent morphological (e.g., Haas, 2003; Pugener et al., 2003) and DNA sequence evidence (Roelants and Bossuyt, 2005; San Mauro et al., 2005) does not support its monophyly.

Ford and Cannatella (1993; fig. 14) suggested this group, Mesobatrachia, to be monophyletic and composed of Pipoidea (Pipidae + Rhinophrynidae) and Pelobatoidea (Pelobatidae [including Scaphiopodidae] + Megophryidae + Pelodytidae). They provided four synapomorphies for their Mesobatrachia: (1) closure of the frontoparietal fontanelle by juxtaposition of the frontoparietal bones (not in Pelodytes or Spea); (2) partial closure of the hyoglossal sinus by the ceratohyals; (3) absence of the taenia tecti medialis; and (4) absence of the taenia tecti transversum.

Pugener et al. (2003) rejected Mesobatrachia and suggested three synapomorphies for a clade composed of all frogs excluding pipoids. (This statement is based on Pugener et al.'s, 2003, figure 12; they provided no comprehensive list of synapomorphies.)

Haas (2003; fig. 15), in contrast, suggested a number of characters that placed Pipoidea as the sister taxon of all frogs except Ascaphidae (although he did not study Leiopelma). This is consistent with the molecular studies of San Mauro et al. (2005; fig. 17). Haas' characters also placed Pelobatoidea (as represented by his exemplars) as a paraphyletic series of Spea, (Pelodytes, Heleophryne), and Pelobates + Megophrys + Leptobrachium, “between” Discoglossidae (sensu lato) and Limnodynastes on a pectinate tree. This is inconsistent with the results of Roelants and Bossuyt (2005). Larval characters suggested by Haas (2003) to support the group of all frogs exclusive of Ascaphidae and Pipoidea are (1) m. mandibulolabialis present; (2) upper jaw cartilages powered by jaw muscles; (3) larval m. levator mandibulae externus main portion inserts in upper jaw cartilages; (4) insertion of the larval m. levator mandibulae internus in relation to jaw articulation lateral; (5) m. levator mandibulae longus superficialis and profundus in two bundles; (6) processus anterolateralis of crista parotica present; (7) processus muscularis present; (8) distal end of cartilago meckeli with stout dorsal and ventral processes forming a shallow articular fossa; and (9) ligamentum mandibulosuprarostrale present.

García-París et al. (2004b; fig. 18) presented mtDNA sequence evidence for the monophyly of Mesobatrachia although their outgroup sampling (which was limited to Ascaphus truei, A. montanus, Discoglossus galganoi, and Rana iberica) provided only a minimal test of this proposition. Even more recently, on the basis of more DNA sequence evidence and better sampling, Roelants and Bossuyt (2005; fig. 16) and San Mauro et al. (2005; fig. 17) found “Mesobatrachia” to have its elements in a paraphyletic series with respect to Neobatrachia. Roelants and Bossuyt (2005) found (Ascaphidae + Leiopelmatidae) + (Discoglossoidea + (Pipoidea + (Pelobatoidea + Neobatrachia))) and San Mauro et al. (2005) found Ascaphidae + Leiopelmatidae as the sister taxon of Pipoidea + (Discoglossoidea + (Pelobatidae + Neobatrachia)). In other words, their substantial difference is in Discoglossoidea (= Bombinatoridae + Discoglossidae) and Pipoidea changing places, with San Mauro et al.'s (2005) placement of Pipoidea agreeing with that of Haas (2003).

Pipoidea:

Pipoidea (Pipidae + Rhinophrynidae) is clearly well corroborated as monophyletic but not clearly resolved with respect to its rather dense fossil record. Ford and Cannatella (1993; fig. 14) considered Pipoidea to be supported by five morphological synapomorphies: (1) lack of mentomeckelian bones; (2) absence of lateral alae of the parasphenoid; (3) fusion of the frontoparietals into an azygous element; (4) greatly enlarged otic capsule; and (5) tadpole with paired spiracles and lacking keratinized jaw sheaths and keratodonts (Type I tadpole). Haas (2003) added a substantial number of larval characters: (1) eye position lateral; (2) opercular canal and spiracles paired; (3) insertion of m. levator arcuum branchialium reaching medially and extending on proximal parts of ceratobranchial IV; (4) m. constrictor branchialis I absent; (5) m. levator mandibulae internus shifted anteriorly; (6) m. levator mandibulae longus originates exclusively from arcus subocularis; (7) posterolateral projections of the crista parotica with expansive flat chondrifications; (8) arcus subocularis with a distinct processus lateralis posterior projecting laterally from the posterior palatoquadrate; (9) articulation of cartilago labialis superior with cornu trabeculae fused into rostral plate; and (10) forelimb erupts out of limb pouch, outside of peribranchial space. In addition, recent DNA sequence data (Roelants and Bossuyt, 2005; fig. 16) strongly support a monophyletic group of Rhinophrynidae + Pipidae.

Rhinophrynidae (1 genus, 1 species):

Tropical North American and Central American Rhinophrynus dorsalis is a burrowing frog with a number of apomorphies with respect to its nearest living relative, Pipidae: (1) division of the distal condyle of the femur into lateral and medial condyles; (2) modification of the prehallux and distal phalanx of the first digit into a spade for digging; (3) tibiale and fibulare short and stocky, with distal ends fused; and (4) an elongate atlantal neural arch. In addition to the previous characters provided by Ford and Cannatella (1993; fig. 14), Haas (2003; fig. 15) provided (1) larval m. geniohyoideus absent; (2) larval m. levator mandibulae externus present in two bundles (profundus and superficialis); (3) ramus mandibularis (cranial nerve V3) posterior (ventral) to m. levator mandibulae externus group; (4) endolymphatic spaces extend into more than half of the vertebral canal (presacral vertebrae 4 or beyond); (5) branchial food traps divided crescentrically; (6) cricoid ring with dorsal gap; and (7) urobranchial process very long. Available DNA sequence data (e.g., Roelants and Bossuyt, 2005) also suggest strongly that Rhinophrynus is the sister taxon of Pipidae. We sampled the single species in this taxon, Rhinophrynus dorsalis. Báez and Trueb (1997) noted that Rhinophrynus also has amphicoelous ectochordal vertebrae, as in Ascaphidae and Leiopelmatidae, which may be a synapomorphy of Rhinophrynidae at this level of generality.

Pipidae (5 genera, 30 species):

South American and African Pipidae is a highly apomorphic group of bizarre, highly aquatic species. Ford and Cannatella (1993) provided 11 characters in support of its monophyly: (1) lack of a quadratojugal; (2) presence of an epipubic cartilage; (3) unpaired epipubic muscle; (4) free ribs in larvae; (5) fused articulation between the coccyx and the sacrum; (6) short, stocky scapula; (7) elongate septomaxillary bones; (8) ossified pubis; (9) a single median palatal opening of the eustachian tube; (10) lateral line organs in the adults; and (11) loss of tongue. Báez and Trueb (1997) added to this list (fossil taxa pruned by us for purposes of this discussion): (1) the possession of an optic foramen with a complete bony margin formed by the sphenethmoid; (2) anterior ramus of the pterygoid arises near the anteromedial corner of the otic capsule; (3) parasphenoid fused at least partially with the overlying braincase; (4) vomer without an anterior process if the bone is present; (5) mandible bears a broad-based, bladelike coronoid process along its posteromedial margin; (6) sternal end of the coracoid not widely expanded; (7) anterior ramus of pterygoid dorsal with respect to the maxilla; and (8) premaxillary alary processes expanded dorsolaterally. Haas (2003) provided 11 additional larval characters: (1) origin of the m. subarcualis rectus II–IV placed far laterally; (2) anterior insertion of m. subarcualis rectus II–IV on ceratohyal III; (3) commissurae craniobranchiales present; (4) arcus subocularis round in cross section; (5) one perilymphatic foramen; (6) vertebral centra formation epichordal; (7) processus urobranchialis absent; and (8) ventral valvular velum absent, as well as these additional characters of the adult: (9) advertisement call without airflow; (10) pupil shape round; and (11) pectoral girdle pseudofirmisternal.

On the basis of morphology, Cannatella and Trueb (1988; fig. 19A) considered the generic relationships to be Xenopus + (Silurana + ((Hymenochirus + Pseudhymenochirus) + Pipa)). Subsequently, de Sá and Hillis (1990; fig. 19B), on the basis of a combined analysis of morphology and mtDNA, proposed the arrangement Hymenochirus (Xenopus + Silurana), and this was further corroborated by Báez and Trueb (1997) and Báez and Pugener (2003; who found [Hymenochirus + Pipa] + [Xenopus + Silurana]; fig. 19C), and suggested the following synapomorphies for Dactylethrinae (Xenopus + Silurana; fossil taxa pruned for this discussion): (1) scapula extremely reduced; (2) margins of olfactory foramina cartilaginous; (3) articular surfaces of the vertebral pre- and postzygopophyses bear sulci and ridges, with the prezygopophyses covering the lateral margin of the postzygopophysis; and (4) anterior process of the pterygoid laminae. They also suggested the following synapomorphies for Pipinae (Pipa + Hymenochirus) (fossil taxa pruned for purposes of this discussion): (1) wedge-shaped skull; (2) vertebrae with parasagittal spinous processes; (3) anterior position of the posterior margin of the parasphenoid; (4) possession of short coracoids broadly expanded at their sternal ends. In addition, they noted other characters of more ambiguous placement that optimize on this stem in this topology. Recent DNA sequence data (Roelants and Bossuyt, 2005; figs. 16, 19D), however, suggest a topology of Pipa + (Hymenochirus + (Xenopus + Silurana)).

We sampled three species of Dactylethrinae (Africa): Silurana tropicalis, Xenopus laevis, and X. gilli. From Pipinae (South America and Africa) we sampled Hymenochirus boettgeri, Pipa pipa, and P. carvalhoi. According to the cladogram provided by Trueb and Cannatella (1986), inclusion of either Pipa parva or P. myersi would have bracketed the phylogenetic diversity of Pipa somewhat better, although our sampling was adequate to test pipine (weakly), dactylethrine, and pipid monophyly, and the placement of Pipidae among other frogs.

Pelobatoidea:

Pelobatoidea (Megophryidae, Pelobatidae, Pelodytidae, and Scaphiopodidae) has also been the source of considerable controversy. Haas (2003; fig. 15) did not recover the group as monophyletic (see the earlier discussion under Mesobatrachia), although Ford and Cannatella (1993) suggested that synapomorphies include the presence of a palatine process of the maxilla and ossification of the sternum into a bony style. Gao and Wang (2001) found Pelobatoidea to be more closely related to Discoglossidae on the basis of a limited analysis of fossil taxa. But, García-París et al. (2003; fig. 18) suggested that Pelobatoidea is the sister taxon of Pipoidea on the basis of a maximum-likelihood analysis of mtDNA evidence, although their outgroup structure was insufficient to provide a strong test of this proposition. (This position was effectively rejected by recent molecular evidence [Roelants and Bossuyt, 2005; San Mauro et al., 2005; figs. 16, 17].)

Maglia (1998) also provided an analysis of Pelobatoidea, but because she constrained the monophyly of this group we are not sure how to interpret the distribution of her morphological evidence. Pugener et al. (2003) provided a cladogram based on morphology in which Pelobatoidea was recovered as monophyletic (and imbedded within Neobatrachia), but the underlying data were not provided. Roelants and Bossuyt (2005; fig. 16) suggested on the basis of DNA evidence that Pelobatoidea is the sister taxon of Neobatrachia, a result that is consistent with the older view of Savage (1973; cf. Noble, 1931). Dubois (2005) most recently treated all pelobatoids as a single family composed of four subfamilies, but this was merely a change in Linnaean rank without a concomitant change in understanding phylogenetic history.

Pelobatidae (1 genus, 4 species) and Scaphiopodidae (2 genera, 7 species):

Ford and Cannatella (1993; fig. 14) diagnosed Pelobatidae (including Scaphiopodidae in their sense) on the basis of (1) fusion of the joint between the sacrum and urostyle; (2) exostosed frontoparietals; and (3) presence of a metatarsal spade supported by a well-ossified prehallux. As noted earlier, Haas (2003; fig. 15) did not recover Pelobatidae (sensu lato) as monophyletic, instead placing Spea phylogenetically far from Pelobatidae, more distant than Heleophryne. More recently, García-París et al. (2003; fig. 18) provided molecular data suggesting that Pelobatidae and Scaphiopodidae are not each other's closest relatives. These results were augmented by the DNA sequence studies of Roelants and Bossuyt (2005) and San Mauro et al. (2005), both of which supported Scaphiopodidae as the sister taxon of Pelodytidae + (Pelobatidae + Megophryidae) (figs. 16, 17). All species have typical exotrophic aquatic larvae (Altig and McDiarmid, 1999). We sampled Spea hammondii, Scaphiopus couchii, and S. holbrooki from Scaphiopodidae, and Pelobates fuscus and P. cultripes from Pelobatidae.

Pelodytidae (1 genus, 3 species):

Ford and Cannatella (1993; fig. 14) diagnosed Pelodytidae as having a fused astragalus and calcaneum (also found in some Centrolenidae; Sanchíz and de la Riva, 1993) and placed them in their Pelobatoidea as did García-París et al. (2003; fig. 18). Haas (2003), however, recovered Pelodytes in a polytomy with Heleophryne, Neobatrachia and Megophrys + Pelobates + Leptobrachium. We sampled Pelodytes punctatus as our exemplar of Pelodytidae. Larvae in pelodytids are also typical free-living exotrophs (Altig and McDiarmid, 1999).

Megophryidae (11 genera, 129 species):

Ford and Cannatella (1993; fig. 14) diagnosed Megophryidae as having (1) a complete or nearly complete absence of ceratohyals in adults; (2) intervertebral cartilages with an ossified center; and (3) paddle-shaped tongue. Haas (2003; fig. 15) recovered a group consisting of the megophryids (Leptobrachium and Megophrys being his exemplars) and Pelobates but did not resolve the megophryids sensu stricto. Evidence for this megophryid + Pelobates clade is: (1) distal anterior labial ridge and keratodont-bearing row very short and median; (2) vena caudalis dorsalis present; (3) anterior insertion of the m. subarcualis rectus II–IV on ceratobranchial III; (4) m. mandibulolabialis superior present; (5) adrostral cartilage very large and elongate; and (6) cricoid ring with a dorsal gap.

Dubois (1980) and Dubois and Ohler (1998) suggested that megophryids form two subfamilies based on whether the larvae have funnel-shaped oral discs (Megophryinae), an apomorphy, or nonmodified oral discs (Leptobrachiinae), a plesiomorphy. Megophryinae includes Atympanophrys, Brachytarsophrys, Megophrys, Ophryophryne, and Xenophrys. Their Leptobrachiinae includes Leptobrachella, Leptolalax, Leptobrachium, Oreolalax, Scutiger, and Vibrissaphora. Delorme and Dubois (2001) presented a consensus tree (fig. 20) based on 54 transformation series of morphology (not including Vibrissaphora). This tree suggests that Megophryinae (Megophrys montana being their exemplar) is deeply imbedded within a paraphyletic Leptobrachiinae (the remaining megophryid exemplars being of this nominal subfamily); that Scutiger is composed of a paraphyletic subgenus Scutiger and a monophyletic subgenus Aelurophryne), and that Oreolalax is composed of a paraphyletic subgenus Oreolalax and a monotypic subgenus Aelurolalax.

Within megophryines, Xie and Wang (2000) noted conflict between isozyme and karyological data regarding the monophyly of Brachytarsophrys, and also noted that Atympanophrys is only dubiously diagnosable from Megophrys or Xenophrys. They also suggested that Xenophrys may not be diagnosable from Megophrys.

Lathrop (1997) suggested that, among nominal leptobrachiines, Leptolalax has no identified apomorphies. Xie and Wang (2000) noted that Oreolalax is diagnosable from Scutiger on the basis of unique maxillary teeth and that Vibrissaphora has apomorphies (e.g., keratinized spines along the lips of adults), although the effect of recognizing Oreolalax and Vibrissaphora on the monophyly of Scutiger has not been evaluated. Similarly, the monophyly of Leptobrachium is undocumented.

The species sampled for DNA sequences were Brachytarsophrys feae, Leptobrachium chapaense, L. hasselti, Leptolalax bourreti, Megophrys nasuta, Ophryophryne hansi, O. microstoma, Xenophrys major (formerly X. lateralis). We were unable to obtain samples of Atympanophrys, Leptobrachella, Oreolalax, Scutiger, and Vibrissaphora, so, although we are confident that our sampling will allow phylogenetic generalizations to be made regarding the family, most of the problems within the group (e.g., the questionable monophyly of Leptobrachium, Leptolalax, Megophrys, Scutiger, and Xenophrys) will remain unanswered.

“Advanced” Frogs—Neobatrachia

Neobatrachia9 includes about 96% of extant frogs and is a poorly understood array of apparently likely paraphyletic groups with apomorphic satellites. So, at this juncture in our discussion the quantity of evidence suggested by authors to support major groups, and the quality of published taxonomic reasoning drops significantly to the realm of grouping by overall similarity and special pleading for particularly favored characters. Like the larger-scale Archeobatrachia (primitive frogs), Mesobatrachia (transitional frogs), and Neobatrachia (advanced frogs) of prephylogenetic systematics, Neobatrachia also has within it its own nominally “primitive” groups aggregated on plesiomorphy (e.g., Leptodactylidae), as well as its own nominally “transitional” and “advanced” groups (e.g., Ranidae and Rhacophoridae, Arthroleptidae and Hyperoliidae). Further, the unwillingness of the systematics community to change taxonomies in the face of evidence is best illustrated here. For example, Brachycephalidae was shown to be imbedded within the leptodactylid taxon Eleutherodactylinae, but the synonymy was not made by Darst and Canntella (2004), and Leptopelinae was shown to be more closely related to Astylosternidae than to hyperoliine hyperoliids by Vences et al. (2003c), but was retained by those authors in an explicitly paraphyletic Hyperoliidae.

Ford and Cannatella (1993; fig. 14) suggested five characters in support of the monophyly of Neobatrachia: (1) (neo)palatine bone present; (2) fusion of the third distal carpal to other carpals; (3) complete separation of the m. sartorius from the m. semitendinosus; (4) presence of an accessory head of the m. adductor longus; and (5) absence of the parahyoid bone. In addition, Haas (2003; fig. 15) presented the following larval characters (but see Heleophrynidae): (1) upper lip papillation with broad diastema; (2) cartilage of the cavum cranii forms tectum parietale; (3) secretory ridges present; and (4) pupil horizontally elliptical. The character of central importance historically to the recognition of this taxon is the (neo)palatine bone, a character not without its own controversy.

“Hyloidea”:

The worldwide Hyloidea, for which no morphological synapomorphy has been suggested, was long aggregated on the basis of its being “primitive” with respect to the “more advanced” Ranoidea, although molecular evidence under certain analytical methods and assumptions supports its monophyly (Ruvinsky and Maxson, 1996; Feller and Hedges, 1998). Hyloidea is defined by the plesiomorphic (at least within Neobatrachia) possession of arciferal pectoral girdles (coracoids not fused) and simple procoelous vertebrae, although descriptions of both characters have been highly reified through repetition and idealization. More recently, Biju and Bossuyt (2001: fig. 21) suggested on the basis of a DNA sequence analysis that Hyloidea, as traditionally viewed, is paraphyletic with respect to Ranoidea, but within “Hyloidea” is a monophyletic group largely coextensive with “Hyloidea”, but excluding Heleophrynidae, Limnodynastidae, Myobatrachidae, Nasikabatrachidae, Sooglossidae, and, presumably Rheobatrachidae as well. Darst and Cannatella (2004; fig. 22) redelimited Hyloidea as the descendants of the most recent common ancestor of Eleutherodactylini, Bufonidae, Centrolenidae, Hylinae, Phyllomedusinae, Pelodryadinae, and Ceratophryinae, thereby excluding Heleophrynidae, Limnodynastidae, Myobatrachidae, Rheobatrachidae, and Sooglossidae (and by implication, presumably Nasikabatrachidae) from Hyloidea10. For this discussion, we retain the older, more familiar definition of Hyloidea as all neobatrachians excluding the ranoids.

Heleophrynidae (1 genus, 6 species):

South African Heleophryne was considered by Ford and Cannatella (1993; fig. 14) to be a member of Neobatrachia and Hyloidea. The synapomorphy of Heleophrynidae suggested by these authors includes only absence of keratinous jaw sheaths in exotrophic free-living larvae. Haas (2003; fig. 15), in contrast, placed Heleophrynidae outside Neobatrachia in a pectinate relationship among “pelobatoids” or as the sister taxon of Pelobates, Leptobrachium, and Megophrys. Heleophryne is included at this level in Haas' analysis by having (1) m. tympanopharyngeus present; (2) m. interhyoideus posterior present; (3) m. diaphragmatopraecordialis present; (4) m. constrictor branchialis I present; and (5) interbranchial septum IV musculature with the lateral fibers of the m. subarcualis rectus II–IV invading the septum, and lacking the characters listed by Haas for Neobatrachia. In addition, the vertical pupil and ectochordal vertebrae tie heleophrynids to myobatrachines, and non-neobatrachians. Recent DNA sequence evidence (San Mauro et al., 2005; fig. 17) strongly supports Heleophrynidae as the sister taxon of all other neobatrachians (although Biju and Bossuyt, 2003, also on the basis of molecular evidence as well had suggested that Heleophrynidae is the sister taxon of Limnodynastidae + Myobatrachidae).

We sampled Heleophryne purcelli and H. regis. These species are likely close relatives (Boycott, 1982) so broader sampling (to have included H. rosei, whose isolation on Table Mountain near Cape Town suggests a likely distant relationship to the other species) would have been preferable.

Sooglossidae (2 genera, 4 species) and Nasikabatrachidae (1 genus, 2 species):

Sooglossidae is a putative Gondwanan relict (Savage, 1973) on the Seychelles, possibly related to myobatrachids as evidenced by sharing with that taxon the plesiomorphy of ectochordal vertebrae (J.D. Lynch, 1973), although Bogart and Tandy (1981) suggested a relationship with the arthroleptines (a ranoid group). In fact, the group is plesiomorphic in many characters, being arciferal (although having a bony sternum; see Kaplan, 2004, for discussion of the various meanings of “arcifery”) and all statements as to its relationships, based on morphology, have been highly conjectural. Biju and Bossuyt (2003; fig. 21) suggested on the basis of DNA sequence evidence that Sooglossidae is the sister taxon of the recently discovered Nasikabatrachus, found in the Western Ghats of South India. Nasikabatrachus has so far had little of its morphology documented. They also found Sooglossidae + Nasikabatrachidae to form the sister taxon of all other neobatrachians.

We sampled one species each of the nominal sooglossid genera (Nesomantis thomasseti and Sooglossus sechellensis). Of Nasikabatrachidae we sampled Nasikabatrachus sahyadrensis as well as sequences attributed by Dutta et al. (2004) only to an unnamed species of Nasikabatachidae, also from the Western Ghats. Although Dutta et al. did not name their species as new, they explicitly treated it as distinct from N. sahyadrensis (Dutta et al., 2004: 214), and we therefore follow their usage. (Our statement that Nasikabatrachidae contains two species rests on this assertion, although any clear diagnosis of the second has yet to be cogently provided.) All species of Sooglossidae that are known are endotrophic according to Thibaudeau and Altig (1999). Sooglossus sechellensis has free tadpoles that are carried on the back of the mother. The tadpoles are likely endotrophic, but this is not definitely known (R.A. Nussbaum, personal obs.). Dutta et al. (2004) reported exotrophic tadpoles occurring in fast-flowing streams for their unnamed species of Nasikabatrachidae.

Limnodynastidae (8 genera, 50 species), Myobatrachidae (11 genera, 71 species), and Rheobatrachidae (1 genus, 2 species):

Different authors consider this taxonomic cluster to be one family (Myobatrachidae, sensu lato) with two or three subfamilies (Heyer and Liem, 1976); to be two families, Limnodynastidae and Myobatrachidae (Zug et al., 2001; Davies, 2003a, 2003b); or to be three families, Limnodynastidae, Myobatrachidae, and Rheobatrachidae (Laurent, 1986). Because Rheobatrachidae (Rheobatrachus; Laurent, 1986) was only tentatively associated with Myobatrachidae by Ford and Cannatella (1993), we retain its familial status for clarity of discussion.

Limnodynastidae, Myobatrachidae, and Rheobatrachidae are primarily united on the basis of their geographic propinquity on Australia and New Guinea (Tyler, 1979; Ford and Cannatella, 1993). And, only one line of evidence, that of spermatozoal morphology, has ever suggested that these taxa taken together are monophyletic (Kwon and Lee, 1995). Heyer and Liem (1976) provided a character analysis that assumed familial and generic monophyly, but this was criticized methodologically (Farris et al., 1982a). Rheobatrachinae (Rheobatrachus) is of uncertain position, although Farris et al. (1982a) in their reanalysis of Heyer and Liem's (1976) data, considered it to be part of Limnodynastinae. Ford and Cannatella (1993) subsequently argued that Rheobatrachinae is more closely related to Myobatrachidae than to Limnodynastidae, although this suggestion, like the first, rests on highly contingent phylogenetic evidence. Moreover, Myobatrachidae may be related to Sooglossidae (J.D. Lynch, 1973) and Limnodynastinae to Heleophrynidae (J.D. Lynch, 1973; Ruvinsky and Maxson, 1996), although these views are largely conjectural inasmuch as the character evidence of J.D. Lynch (1973) was presented in scenario form.

Ford and Cannatella (1993) suggested, on the basis of discussion of characters presented by Heyer and Liem (1976), that Myobatrachidae (Myobatrachinae in their sense and presumably including Rheobatrachus) has four morphological synapomorphies: (1) presence of notochordal (ectochordal) vertebrae with intervertebral discs; (2) m. petrohyoideus anterior inserting on the ventral face of the hyoid; and, possibly, (3) reduction of the vomers and concomitant reduction of vomerine teeth (J.D. Lynch, 1971).

Ford and Cannatella (1993) suggested several synapomorphies of Myobatrachidae and Sooglossidae to the exclusion of Limnodynastidae: (1) incomplete cricoid cartilage ring; (2) semitendinosus tendon inserting dorsal to the m. gracilis (in myobatrachines excluding Taudactylus and Rheobatrachus, which have a ventral trajectory of the tendon; (3) horizontal pupil (except in Uperoleia; (also vertical in Rheobatrachus; limnodynastines primitively have a vertical pupil according to Heyer and Liem, 1976, although several have horizontal ones); (4) broad alary process (Griffiths, 1959a), which they found in Myobatrachidae and Rheobatrachus (as well as in Adenomera, Physalaemus [in the sense of including Engystomops and Eupemphix], and Pseudopaludicola); and (5) divided sphenethmoid.

Ford and Cannatella (1993) reported at least one synapomorphy for Limnodynastidae: a connection between the m. intermandibularis and m. submentalis (also found in Leptodactylinae and Eleutherodactylinae according to Burton, 1998b). Rheobatrachus was diagnosed by having gastric brooding of larvae—an unusual reproductive mode, to say the least. It is tragic that the two species are likely now extinct (Couper, 1992).

Read et al. (2001) provided a phylogenetic study of myobatrachine frogs (fig. 23) based on mtDNA sequence data that assumed monophyly of the group and used only Limnodynastes to root the myobatrachine tree. The evolutionary propinquity of Limnodynastes (Limnodynastidae) and Myobatrachus (Myobatrachidae) was supported on the basis of DNA sequence evidence by Biju and Bossuyt (2003).

We were able to sample at least one species for most of the genera of the three nominal families. For Limnodynastidae we sampled at least one species for all nominal genera: Adelotus brevis, Heleioporus australiacus, Lechriodus fletcheri, Limnodynastes depressus, L. dumerilii, L. lignarius, L. ornatus, L. peronii, L. salmini, Mixophyes carbinensis, Neobatrachus sudelli, N. pictus, Notaden melanoscaphus, Philoria sphagnicola. Recent authors (e.g., Cogger et al., 1983) have considered Kyarranus to be a synonym of Philoria, and we follow this. J.D. Lynch (1971) provided morphological characters that are evidence of monophyly of Kyarranus + Philoria (e.g., presence of stubby fingers and concealed tympana as well as direct development—Littlejohn, 1963; De Bavay, 1993; Thibaudeau and Altig, 1999).

For Rheobatrachidae, we obtained Rheobatrachus silus. And, for Myobatrachidae, we obtained at least single representatives of all nominal genera: Arenophryne rotunda, Assa darlingtoni, Crinia nimbus, C. signifera, Geocrinia victoriana, Metacrinia nichollsi, Myobatrachus gouldii, Paracrinia haswelli, Pseudophryne bibroni, P. coriacea, Spicospina flammocaerulea, Taudactylus acutirostris, and Uperoleia laevigata. With exceptions, this taxon selection will not allow us to comment on generic monophyly, but it will identify major monophyletic groups and questions that will guide future research. All rheobatrachids and most myobatrachids have endotrophic larvae and various degrees of direct development (Thibaudeau and Altig, 1999).

“Leptodactylidae” (57 genera, 1243 species):

“Leptodactylidae” holds the same position in the Americas as Myobatrachidae (sensu lato, as containing Limnodynastidae and Rheobatrachidae) does in Australia—a likely nonmonophyletic hodgepodge “primitive” holochordal or rarely stegochordal, arciferal, and procoelous neobatrachian group united by geography and not synapomorphy. “Leptodactylidae” is currently divided into five subfamilies, some of which are not clearly monophyletic (or consistently diagnosable) and some of which may be polyphyletic (Ruvinsky and Maxson, 1996; Haas, 2003; Darst and Cannatella, 2004; Faivovich et al., 2005; San Mauro et al., 2005; figs. 17, 22, 24).

J.D. Lynch (1971, 1973) considered leptodactylids to be divided into four subfamilies, on the basis of both synapomorphy and symplesiomorphy: (1) Ceratophryinae (for Ceratophrys and Lepidobatrachus); (2) Elosiinae (= Hylodinae of other authors; for Crossodactylus, Hylodes, and Megaelosia); (3) Leptodactylinae (for Barycholos, Edalorhina, Hydrolaetare, Leptodactylus [including Adenomera], Limnomedusa, Lithodytes, Paratelmatobius, Physalaemus [including Engystomops and Eupemphix], Pleurodema, and Pseudopaludicola); and (4) Telmatobiinae, aggregated on the basis of plesiomorphy. Within his Telmatobiinae Lynch defined five tribes, each aggregated on a variable basis of synapomorphy and symplesiomorphy: Telmatobiini (Batrachophrynus, Caudiverbera, Telmatobius, and Telmatobufo); Alsodini (Batrachyla, Eupsophus [including Alsodes], Hylorina, and Thoropa); Odontophrynini (Macrogenioglottus, Odontophrynus, and Proceratophrys); Grypiscini (Crossodactylodes, Cycloramphus, and Zachaenus); and Eleutherodactylini (Eleutherodactylus, Euparkerella, Holoaden, and Ischnocnema, as well as several other genera subsequently placed in the synonymy of Eleutherodactylus), with Scythrophrys being left incertae sedis. Subsequently, Heyer (1975) provided a preliminary clustering (based on the nonphylogenetic monothetic subset method of Sharrock and Felsenstein, 1975) of the nominal genera within the family that assumed monophyly of both the family and the constituent genera (see Farris et al., 1982a, for criticism of the approach) in which Heyer identified, but did not recognize formally, five units that were recognized subsequently (Laurent, 1986) as Ceratophryinae, Eleutherodactylinae, Cycloramphinae, Leptodactylinae, and Telmatobiinae. J.D. Lynch (1978b) revised the genera of Telmatobiinae, where he recognized three tribes: Telmatobiini (Alsodes, Atelognathus, Batrachophrynus, Eupsophus, Hylorina, Insuetophrynus, Limnomedusa, Somuncuria, and Telmatobius), Calyptocephalellini (Caudiverbera and Telmatobufo), and Batrachylini (Batrachyla and Thoropa). The justification for this arrangement was partially based on character argumentation, although plausibility of results was based on subjective notions of overall similarity and relative character importance. A cursory glance at figure 24 (Faivovich et al., 2005) shows that several of these groups are nonmonophyletic.

Burton (1998a) suggested on the basis of hand muscles (although his character polarity was not well supported) that the leptodactylid tribe Calyptocephalellini is more closely related to the South African Heleophrynidae than to other South American leptodactylids. San Mauro et al. (2005; fig. 17) suggested on the basis of DNA sequence data that Caudiverbera (Calyptocephalellini) is more closely related to at least some component of Limnodynastidae (Lechriodus) than to other South American “leptodactylids”. Another leptodactylid satellite is Brachycephalidae, a small monophyletic taxon, likely the sister taxon of Euparkerella (Leptodactylidae: Eleutherodactylinae) based on digit reduction (Izecksohn, 1988; Giaretta and Sawaya, 1998). Similarly, Rhinodermatidae (Rhinoderma) is a small group that is likely also a telmatobiine leptodactylid (Barrio and Rinaldi de Chieri, 1971; Lavilla and Cei, 2001), differing from them in having partial or complete larval development within the male vocal sac and, except for Eupsophus, in having endotrophic larvae (Formas et al., 1975; Altig and McDiarmid, 1999).

Laurent (1986) provided the subfamilial taxonomy we employ for discussion (his arrangement being the formalization of the groupings tentatively recommended by Heyer, 1975). He recognized Ceratophryinae (in the larger sense of including J.D. Lynch's Odontophrynini, transferred from Telmatobiinae), Telmatobiinae (including calyptocephallelines and excluding J.D. Lynch's Eleutherodactylini), Cycloramphinae (as Grypiscinae, including Grypscini and Elosiinae of J.D. Lynch), Eleutherodactylinae, and Leptodactylinae.

“Ceratophryinae” (6 genera, 41 species):

Reig (1972) and Estes and Reig (1973) suggested that the leptodactylid subfamily Ceratophryinae was “ancestral”, in some sense, to Bufonidae, although others rejected this (e.g., J.D. Lynch, 1971, 1973). Laurent (1986), following Heyer (1975), transferred Macrogenioglottus, Odontophrynus, and Proceratophrys (J.D. Lynch's tribe Odontophrynini) into this nominal subfamily, with Ceratophrys, Chacophrys, and Lepidobatrachus being placed in Ceratophryini. Haas (2003; fig. 15) presented morphological evidence that Ceratophryini and Odontophrynini are not each other's closest relatives (following J.D. Lynch, 1971), with Odontophrynus most closely related to Leptodactylus, and the clade Ceratophryini (Lepidobatrachus + Ceratophrys) most closely related to hylids, exluding hemiphractines. Duellman (2003) treated the two groups as subfamilies, Odontophryninae and Ceratophryinae, presumably following the results of Haas (2003), and this was followed by Dubois (2005). Faivovich et al. (2005; fig. 24) also found Ceratophryinae to be polyphyletic. We sampled exemplars from all nominal ceratophryid genera except Macrogenioglottus, which is similar to Odontophrynus (J.D. Lynch, 1971) and karyologically similar to Proceratophrys (Silva et al., 2003; Odontophrynus not examined in that study) that we doubt that this will be an important problem. Ceratophryini does have synapomorphies, for example: (1) transverse processes of anterior presacral vertebrae widely expanded; (2) cranial bones dermosed; and (3) teeth fanglike, nonpedicellate (J.D. Lynch, 1971, 1982b), although nominal Odontophrynini does not have unambiguously synapomorphies, and the group is united on overall similarity. All ceratophryids have free-living exotrophic larvae (Altig and McDiarmid, 1999). We sampled three species of Ceratophryini (Ceratophrys cranwelli, Chacophrys pierotti, and Lepidobatrachus laevis) and three species of Odontophrynini (Odontophrynus achalensis, O. americanus, and Proceratophrys avelinoi). Our sampling of Proceratophrys should have been denser, but this proved a practical impossibility.

“Cycloramphinae” (10 genera, 79 species):

Haas (2003) suggested that this group may be closely related to Dendrobatidae, in part supporting the earlier position of Noble (1926) and J.D. Lynch (1973) that the hylodine part of this nominal subfamily (Crossodactylus, Hylodes, and Megaelosia) is paraphyletic with respect to Dendrobatidae. Faivovich et al. (2005; fig. 24) recovered Crossodactylus (their exemplar of this group) as the sister taxon of Dendrobatidae. Laurent (1986) recognized this subfamily, thus unifying J.D. Lynch's (1971, 1973) Grypiscini and Elosiinae (= Hylodinae), although the evidentiary basis for uniting these was based on Heyer's (1975) results based on monothetic subsets, not parsimony. (Note that J.D. Lynch, 1971, had considered his Grypiscini to be close to Eleutherodactylini on the basis of overall similarity.) Grypiscines and hylodines differ in (1) the shape of the transverse processes of the posterior presacral vertebrae, being short in hylodines and not short in grypiscines; (2) the shape of the facial lobe of the maxillae (deep in grypiscines, shallow in hylodines); (3) the shape of the nasals (large and in median contact in grypiscines, small and widely separated in hylodines); and (4) whether the nasal contacts the frontoparietal (contact in grypiscines, no contact in hylodines). We were unable to obtain samples of Crossodactylodes, Rupirana, or Zachaenus, but we did obtain at least one species of every other nominal genus in the group: Crossodactylus schmidti, Cycloramphus boraceiensis, Hylodes phyllodes, Megaelosia goeldii, Paratelmatobius sp., Scythrophrys sawayae, and Thoropa miliaris. Denser sampling of this particular taxon would have been preferable, but what we obtained will test cycloramphine monophyly and its putative relationship to Dendrobatidae and will provide an explicit hypothesis of its internal phylogenetic structure as the basis of future studies.

Duellman (2003) did not accept Laurent's (1986) unification of J.D. Lynch's Hylodinae and Grypiscini and recognized Hylodinae (Crossodactylus, Hylodes, and Megaelosia) as a different subfamily from Cycloramphinae. Duellman distinguished Hylodinae and Cycloramphinae by T-shaped terminal phalanges in Hylodinae and knoblike terminal phalanges in Cycloramphinae; and glandular pads on the dorsal surface of the digits, absent in Hylodinae and present in Cycloramphinae. However, neither the particulars of distribution of these characters in the taxa nor the levels of universality of their application as evidence was discussed. Duellman (2003) also suggested that Hylodinae and Cycloramphinae differ in chromosome numbers, with 13 pairs in Cycloramphinae and 3 pairs in Hylodinae. However, Kuramoto (1990) noted that hylodines in Duellman's sense have 11–13 pairs of chromosomes, and cycloramphines in Duellman's sense also have 11–13 pairs, so Duellman's statement is taken to be an error.

Eleutherodactylinae (13 genera, 782 species):

The only suggested synapomorphy of this taxon is direct terrestrial development of large eggs deposited in small clutches (J.D. Lynch, 1971). The universality of direct development in this group is based on extrapolation from the few species for which direct development has been observed; the occurrence of large, unpigmented eggs, and because free-living larvae are unknown (see cautionary remarks in Thibaudeau and Altig, 1999). Inasmuch as this taxon contains the largest vertebrate genus, Eleutherodactylus (ca. 600 species) of which the vast majority are not represented by genetic samples, this taxon will remain inadequately sampled for some time. There has never been any comprehensive phylogenetic study of the relationships within the group and the likelihood of many (or even most) of the non-Eleutherodactylus genera being components of Eleutherodactylus is high. Indeed, Ardila-Robayo (1979) suggested strongly that for the taxon currently referred to as Eleutherodactylus (sensu lato) to be rendered monophyletic it would need to include Barycholos, Geobatrachus, Ischnocnema, and Phrynopus (and likely Adelophryne, Phyllonastes, Phyzelaphryne, Holoaden and Euparkerella, and Brachycephalidae [Izecksohn, 1971; Giaretta and Sawaya, 1998; Darst and Cannatella, 2004; Faivovich et al., 2005]11). Regardless, many of the nominal eleutherodactyline genera represent rare and extremely difficult animals to obtain (e.g., Atopophrynus, Dischidodactylus), so our sampling of this particular taxon is clearly inadequate to address most systematic problems. We could not obtain samples of Adelophryne, Atopophrynus, Dischidodactylus, Euparkerella (even though it was suggested to be closely related to Brachycephalidae), Geobatrachus, Holoaden, Phyllonastes, or Phyzelaphryne. We hope that work in the near future can rectify this with the recognition of major monophyletic groups from within Eleutherodactylus. What we could sample of the non-Eleutherodactylus eleutherodactyline taxa were Barycholos ternetzi, Ischnocnema quixensis, and two species of Phrynopus. Of Eleutherodactylus (sensu lato) we sampled two species of the North American subgenus Syrrhophus (Eleutherodactylus marnocki of the E. marnocki group of J.D. Lynch and Duellman, 1997, and E. nitidus of the E. nitidus group of J.D. Lynch and Duellman, 1997); one species of the Antillean subgenus Euhyas (Eleutherodactylus planirostris of the E. ricordii group of J.D. Lynch and Duellman, 1997); two species of the South American subgenus Eleutherodactylus (E. binotatus and E. juipoca, both of the E. binotatus group of J.D. Lynch, 1978a; see also J.D. Lynch and Duellman, 1997); and six species of the Middle American subgenus Craugastor12 (E. bufoniformis of the E. bufoniformis group of J.D. Lynch, 2000, E. alfredi of the E. alfredi group of J.D. Lynch, 2000, E. augusti of the E. augusti group of J.D. Lynch, 2000, E. pluvicanorus of the E. fraudator group of Köhler, 2000, E. punctariolus and E. cf. ranoides13 of the E. rugulosus group of J.D. Lynch, 2000) and E. rhodopis of the E. rhodopis group of J.D. Lynch, 2000). (For expediency, all of these are noted in “Results” in combination with their subgeneric names; e.g., Eleutherodactylus (Syrrhophus) marnockii is treated as Syrrhophus marnockii.) As noted earlier, we expect that Eleutherodactylus will be found to be paraphyletic with respect to a number of other eleutherodactyline taxa (e.g., Barycholos, Phrynopus, and Ischnocnema) and hope that this selection will allow some illumination of this. Nevertheless, we are aware that this tiny fraction of the species diversity of Eleutherodactylus is insufficient to fully resolve the phylogeny of this massive taxon and that the value of the results will be in highlighting outstanding problems and providing a basis for future, more densely sampled studies.

Leptodactylinae (12 genera, 159 species):

Monophyly of this group is supported by the possession of foam nests (except in Limnomedusa [Langone, 1994] and Pseudopaludicola [Barrio, 1954], and in some species of Pleurodema [Duellman and Veloso M., 1977]) and the presence of a bony sternum (rather than the cartilaginous sternum of other leptodactylids; J.D. Lynch, 1971). However, Haas (2003; fig. 15) sampled three species of Leptodactylinae (Physalaemus biligonigerus, Leptodactylus latinasus, and Pleurodema kriegi) for mostly larval morphology and found the group to be para- or polyphyletic with respect to Odontophrynus, and with Physalaemus14 and Pleurodema forming, respectively, more exclusive outgroups of Haas' hylodines and dendrobatids. In Darst and Cannnatella's (2004) phylogenetic analysis of mtDNA (fig. 22), their leptodactyline exemplars are monophyletic in the maximum-likelihood analysis of mtDNA, but polyphyletic in the parsimony analysis. In Faivovich et al.'s (2005; fig. 24) parsimony analysis of multiple mtDNA and nuDNA loci, exemplars of most genera of Leptodactylinae obtained as monophyletic, with the exception of Limnomedusa. Therefore, the monophyly of Leptodactylinae is an open question. We could not obtain samples of Hydrolaetare (or the recently resurrected Eupemphix and Engystomops), but we sampled at least one species of each of the other nominal leptodactyline genera: Adenomera hylaedactyla, Edalorhina perezi, Leptodactylus fuscus, L. ocellatus, Limnomedusa macroglossa, Lithodytes lineatus, Physalaemus gracilis, Pleurodema brachyops, Pseudopaludicola falcipes, and Vanzolinius discodactylus). Our sampling of Leptodactylus is not dense enough to evaluate well the likely paraphyly of this taxon with respect to others, such as Adenomera (Heyer, 1998), being restricted to only two of the five nominal species groups. Leptodactylines vary from having endotrophic larvae, facultatively endotrophic larvae (Adenomera) to having exotrophic, free-living larvae (Edalorhina, Engystomops, Eupemphix, Leptodactylus, Lithodytes, Physalaemus, Pleurodema, Pseudopaludicola, Vanzolinius; Altig and McDiarmid, 1999).

“Telmatobiinae” (11 genera, 98 species):

Telmatobiinae is a similarity grouping of mostly austral South American frogs. As currently employed, contents of this subfamily stem from Laurent's (1986) formalization of Heyer's (1975) informal grouping. Telmatobiines are currently arranged in three tribes (J.D. Lynch, 1971, 1978b; Burton, 1998a): Telmatobiini (Alsodes, Atelognathus, Eupsophus, Hylorina, Insuetophrynus, Somuncuria, and Telmatobius); Batrachylini (Batrachyla and Thoropa); and Calyptocephalellini (Batrachophrynus, Caudiverbera, and Telmatobufo). All telmatobiines have aquatic, exotrophic larvae except Eupsophus, which has endotrophic larvae (Altig and McDiarmid, 1999), and Thoropa, which is semiterrestrial (Bokermann, 1965; Wassersug and Heyer, 1983; Haddad and Prado, 2005).

Batrachylini (in J.D. Lynch's sense of including Thoropa) is diagnosed by having terrestrial eggs and aquatic to semiterrestrial larvae and T-shaped terminal phalanges. Laurent (1986) did not (apparently) accept J.D. Lynch's (1978b) transferral of Thoropa into Batrachylini, and retained Thoropa in Cycloramphinae following Heyer (1975).

Calyptocephalellini was most recently discussed and diagnosed by Burton (1998a) on the basis of hand musculature, but the character argumentation was essentially that of overall similarity, not synapomorphy. Formas and Espinoza (1975) provided karyological evidence for the monophyly of Calyptocephallelini (although they did not address Batrachophrynus). Cei (1970) suggested on the basis of immunology that Calyptocephalellini is phylogenetically distant from leptodactylids, being closer to Heleophrynidae than to any South American leptodactylid group. J.D. Lynch (1978b) suggested the following to be synapomorphies of Calyptocephalellini (composed of solely Caudiverbera and Telmatobufo): (1) occipital artery enclosed in a bony canal; and (2) very broad pterygoid process of the premaxilla. In addition, (1) a very long cultriform process of the parasphenoid; and (2) presence of a medial process on the pars palatina of the premaxilla are osteological characters suggested by J.D. Lynch possibly to unite Batrachophrynus with Caudiverbera and Telmatobufo.

We sampled representatives of two of the genera of Calyptocephallelini (Caudiverbera caudiverbera and Telmatobufo venustus). We could not sample Batrachophrynus, which was considered a calyptocephalelline by Burton (1998a), and in some of the cladograms presented by J.D. Lynch (1978b) Batrachyophrynus was considered to form the sister taxon of his Calyptocephalellini, so its absence from our analysis is unfortunate.

“Telmatobiini” of J.D. Lynch (1978b) is explicitly paraphyletic with respect to Batrachylini and as such has no diagnosis other than that of the inclusive clade “Telmatobiini” + Batrachylini: (1) presence of an outer metatarsal tubercle (dubiously synapomorphic), and (2) reduction of imbrication on the neural arches of the vertebrae. Among species of “Telmatobiini” we sampled Alsodes gargola, Atelognathus patagonicus, Eupsophus calcaratus, Hylorina sylvatica, Telmatobius jahuira, T. cf. simonsi, and T. sp. Of Batrachylini, we sampled Batrachyla leptopus. On this basis we provide a weak test of telmatobiine relationships with regard to Batrachylini. We were unable to sample any member of Insuetophrynus or Somuncuria.

“Hemiphractinae” (5 genera, 84 species):

Mendelson et al. (2000) provided a cladogram of Hemiphractinae but assumed its monophyly and its hylid affinities, as had all authors since Duellman and Gray (1983) and Duellman and Hoogmoed (1984). Haas (2003) suggested (fig. 15), on the basis of morphological data, that his examplar of Hemiphractinae, Gastrotheca, was far from other hylids and imbedded within a hetereogeneous group of leptodactylids and ranoids. Darst and Cannatella (2004), who examined one exemplar species each of Gastrotheca and Cryptobatrachus, suggested on the basis of mtDNA evidence that Hemiphractinae is polyphyletic, with Cryptobatrachus closest to direct-developing eleutherodactylines, and Gastrotheca imbedded in another group of leptodactylids. Similarly, in the analysis by Faivovich et al. (2005; fig. 24) of multiple mtDNA and nuDNA loci, hemiphractines do not appear as monophyletic. They recovered one clade composed of Gastrotheca and Flectonotus, one clade composed of Stefania and Cryptobatrachus, and they found Hemiphractus to form a clade with the few included exemplars of Eleutherodactylinae and Brachycephalidae. Further, inasmuch as the sole noncontingent synapomorphy of nominal Hemiphractinae, bell-shaped larval gills, has not been surveyed widely in direct-developing leptodactylids, we consider the morphological evidence for the monophyly of the hemiphractines to be questionable.

Faivovich et al. (2005; fig. 24) transferred “Hemiphractinae” out of a reformulated Hylidae and into “Leptodactylidae” on the bases that continued inclusion in Hylidae would render Hylidae polyphyletic; its nominal inclusion in “Leptodactylidae” did no violence to a taxon already united solely by plesiomorphy and geography; and placing it incertae sedis within Hyloidea was to suggest its possible placement outside of the “leptodactylid” region of the overall tree, which it is not. “Hemiphractinae” is a grouping of South American frogs united by (1) brooding of eggs on the female's back, generally within a dorsal depression or well-developed pouch; (2) possession in the developing larvae of bell-shaped gills (Noble, 1927); and (3) presence of a broad m. abductor brevis plantae hallucis (Burton, 2004). Larvae may be exotrophic and endotrophic among species of Gastrotheca and Flectonotus, and endotrophic alone in Cryptobatrachus, Hemiphractus, and Stefania. Based on Faivovich et al.'s (2005) topology (fig. 24), claw-shaped terminal phalanges and presence of intercalary cartilages between the ultimate and penultimate phalanges must be considered either convergent with those found in Hylidae or plesiomorphically retained in hylids (and lost in intervening lineages), while the proximal head of metacarpal II not between prepollex and distal prepollex, and the larval spiracle sinistral and ventrolateral (Duellman, 2001) are convergent with those in the Phyllomedusinae. Our sampling of Gastrotheca is not dense enough to allow for the detection of the paraphyly suggested by Mendelson et al. (2000). Our sampling precludes evaluation of paraphyly of any of the nominal genera. Nevertheless, we did sample at least one species per genus, which allows us to test the monophyly of the hemiphractines based on more extensive outgroup sampling. Our sampled taxa are Cryptobatrachus sp., Flectonotus sp., Gastrotheca fissipes, G. cf. marsupiata, Hemiphractus helioi, and Stefania evansi.

Brachycephalidae (1 genus, 8 species):

This tiny group of diminutive south- to southeastern Brazilian species are united by (1) the absence through fusion of a distinguishable sternum; (2) digital reduction (possibly homologous with that in Euparkerella and Phyllonastes in Eleutherodactylinae); and (3) complete ossification of the epicoracoid cartilages with coracoids and clavicles (Ford and Cannatella, 1993; Kaplan, 2002). Brachycephalidae was suggested to be imbedded within Eleutherodacylinae (Izecksohn, 1971; Giaretta and Sawaya, 1998), which also shows direct development. Further, Darst and Cannatella (2004; fig. 22) provided molecular data to link this taxon to Eleutherodactylinae, but continued its recognition despite the demonstrable paraphyly that its recognition requires. Although there are several named and unnamed species in the genus, the monophyly of the group is not in question (Kaplan, 2002), and we sampled the type species, Brachycephalus ephippium, for this study.

Rhinodermatidae (1 genus, 2 species):

As noted earlier, the Chilean Rhinodermatidae is a likely satellite of a paraphyletic “Leptodactylidae”; it is like them in having procoelous and holochordal vertebrae. J.D. Lynch (1973) conjectured that Rhinodermatidae is the sister taxon of Bufonidae, whereas Lavilla and Cei (2001) suggested that Rhinoderma is within the poorly-defined “Telmatobiinae” (“Leptodactylidae”). The only notable synapomorphy of Rhinodermatidae is the rearing of tadpoles within the vocal sacs of the male, although Manzano and Lavilla (1995) also discussed myological characters that are possible synapomorphies. Two species are currently recognized, Rhinoderma darwinii and R. rufum. We sampled R. darwinii.

Dendrobatidae (ca. 11 genera, 241 species):

The monophyly of Dendrobatidae has been upheld consistently (e.g., Myers and Ford, 1986; Ford, 1993; Haas, 1995; Clough and Summers, 2000; Vences et al., 2000b), but different datasets place Dendrobatidae at various extremes within the neobatrachian clade. It is either nested deeply within hyloids and arguably related to cycloramphine leptodactylids (Noble, 1926, 1931; J.D. Lynch, 1971, 1973; Burton, 1998a; Haas, 2003; Faivovich et al., 2005); the sister group of Telmatobius (Vences et al., 2003b); or closely related to Hylinae (Darst and Cannatella, 2004). Alternatively, they have been suggested to be deeply imbedded within ranoids, usually considered close to arthroleptids or petropedetids (Griffiths, 1959b, 1963; Duellman and Trueb, 1986; Ford, 1993; Ford and Cannatella, 1993; Grant et al., 1997). Rigorous evaluation of the support for these contradictory hypotheses is required.

Ford and Cannatella (1993; fig. 14) provided the following as synapomorphies of Dendrobatidae: (1) retroarticular process present on the mandible; (2) conformation of the superficial slip of the m. depressor mandibulae; and (3) cephalic amplexus. They mentioned other features suggested by other authors but that were suspect for one reason or another. Haas (2003; fig. 15) considered the following to be synapomorphies that nest Dendrobatidae within hylodine leptodactylids: (1) guiding behavior observed during courtship; and (2) T- or Y-shaped terminal phalanges. Like most other frogs, most dendrobatids have aquatic free-living tadpoles (with some endotrophy in Colostethus), although the parental-care behavior of carrying tadpoles to water on the back of one of the parents appears to be synapomorphic (Altig and McDiarmid, 1999), although among New World anurans it also occurs in Cycloramphus stejnegeri (Heyer and Crombie, 1979).

Taxon sampling was designed to provide the maximal “spread” of phylogenetic variation with a minimum number of species: Allobates femoralis, Ameerega boulengeri, Colostethus undulatus, Dendrobates auratus, Mannophryne trinitatis, Minyobates claudiae, Phobobates silverstonei, and Phyllobates lugubris. We did not sample any representative of Aromobates, Cryptophyllobates, Nephelobates, nor did we sample either of two generally-not-recognized genera Oophaga or Ranitomeya. On the basis of ongoing work by T. Grant, we think that all of these are imbedded within our sampled genera and their absence does not hamper our ability to test dendrobatid monophyly and place the family in the larger phylogenetic scheme.

Allophrynidae (1 genus, 1 species):

South American Allophryne has been (1) very provisionally associated with Hylidae (J.D. Lynch and Freeman, 1966); (2) asserted to be in Bufonidae on the basis of morphology (Laurent, 1980 “1979”; Dubois, 1983; Laurent, 1986), the evidence for this latter position not actually presented until much later by Fabrezi and Langone (2000); (3) imbedded within Centrolenidae, on the basis of morphology (Noble, 1931); or (4) placed as the sister taxon of Centrolenidae on the basis of mtDNA sequence studies (Austin et al., 2002; Faivovich et al., 2005). Cognoscenti of frogs will marvel at the vastness separating these various hypotheses. Ford and Cannatella (1993) noted that Allophryne lacks the interacalary cartilages of hylids and centrolenids and suggested that placement in any taxon other than Neobatrachia is misleading. Haas (2003; fig. 15) did not examine Allophryne. We sampled the single species, Allophryne ruthveni. Larvae are unknown (Altig and McDiarmid, 1999).

Centrolenidae (3 genera, 139 species):

Centrolenidae has long been thought to be close to, or the sister taxon of, Hylidae (J.D. Lynch, 1973; Ford and Cannatella, 1993; Duellman, 2001) because of the occurrence of interacalary cartilages between the ultimate and penultimate phalanges. On the basis of mostly-larval morphology, Haas (2003) recovered (weakly) Centrolenidae as the sister taxon of all Neobatrachia except for Limnodynastes (Limnodynastidae), because it lacked all characters that Haas' analysis suggested were synapomorphies of Neobatrachia. The analysis of Faivovich et al. (2005; fig. 24) of multiple mtDNA and nuDNA loci recovered an Allophryne + Centrolenidae clade nested within a grade of “Leptodactylidae”. Clearly, the diversity of opinions on the placement of Centrolenidae is great. For our analysis we selected species of the three nominal genera: Centrolene geckoideum, C. prosoblepon, Cochranella bejaranoi, and Hyalinobatrachium fleischmanni. Larvae of centrolenids are aquatic or burrowing exotrophs (Altig and McDiarmid, 1999).

Hylidae (48 genera, 806 species):

Hylidae, as traditionally recognized, was recently shown to be polyphyletic (Ruvinsky and Maxson, 1996; Haas, 2003; Darst and Cannatella, 2004; Faivovich et al., 2005). As an interim measure to resolve this problem Faivovich et al. (2005) transferred “Hemiphractinae” into “Leptodactylidae”, thereby restricting Hylidae to the Holarctic and Neotropical Hylinae, tropical American Phyllomedusinae, and Australo-Papuan Pelodryadinae (and thereby formalizing the implication of Darst and Cannatella, 2004).

Our notions of hylid relationships extend from the recent revision by Faivovich et al. (2005; fig. 24), who provided a phylogenetic analysis of multiple mtDNA and nuDNA loci. Their study addressed 220 hylid exemplar terminals as well as 48 outgroup taxa. For our study, we considered including all terminals from the Faivovich et al. (2005) study, which would have allowed a more rigorous test, but the increased computational burden was judged too great for the expected payoff of increased precision within Hylinae. Our sampling strategy aimed to be sufficiently dense to test the position of hylids among other frogs and the monophyly of the major clades without unduly exacerbating computational problems.

Hylinae (38 genera, 586 species):

Our sampling structure of Hylinae was guided by the results of Faivovich et al. (2005). Beyond their genetic evidence, monophyly of this subfamily is corrobated by at least one morphological synapomorphy: tendo superficialis digiti V (manus) with an additional tendon that arises ventrally from the m. palmaris longus (Da Silva In Duellman, 2001). All hylines for which it is known have free-living exotrophic larvae (Altig and McDiarmid, 1999). Faivovich et al. (2005) recognized four monophyletic tribes within Hylinae: Cophomantini (Aplastodiscus, Bokermannohyla, Hyloscirtus, Hypsiboas, and Myersiohyla); Hylini (Acris, Anotheca, Bromeliohyla, Charadrahyla, Duellmanohyla, Ecnomiohyla, Exerodonta, Hyla, Isthmohyla, Megastomatohyla, Pseudacris, Plectrohyla, Ptychohyla, Smilisca [including former Pternohyla], Tlalocohyla, and Triprion); Dendropsophini (Dendropsophus, Lysapsus, Pseudis, Scarthyla, Scinax, Sphaenorhynchus, and Xenohyla); and Lophiohylini (Aparasphenodon, Argenteohyla, Corythomantis, Itapotihyla, Nyctimantis, Osteocephalus, Osteopilus, Phyllodytes, Tepuihyla, and Trachycephalus).

In this study we included representatives of these four tribes: Cophomantini (Aplastodiscus perviridis, Hyloscirtus armatus, H. palmeri, Hypsiboas albomarginatus, H. boans, H. cinerascens (formerly Hypsiboas granosus; see Barrio-Amorós, 2004: 13), H. multifasciatus); Dendropsophini (Dendropsophus marmoratus, D. minutus, D. nanus, D. parviceps, Lysapsus laevis, Pseudis paradoxa, Scarthyla goinorum, Scinax garbei, S. ruber, Sphaenorhynchus lacteus); Hylini (Acris crepitans, Anotheca spinosa, Charadrahyla nephila, Duellmanohyla rufioculis, Ecnomiohyla miliaria, Exerodonta chimalapa, Hyla arbrorea, H. cinerea, Isthmohyla rivularis, Pseudacris crucifer, P. ocularis, Ptychohyla leonhardschultzei, Smilisca phaeota, Tlalocohyla picta, and Triprion petasatus); and Lophiohylini (Argenteohyla siemersi, Osteocephalus taurinus, Osteopilus septentrionalis, Phyllodytes luteolus, Trachycephalus jordani, and T. venulosus).

Pelodryadinae (3 genera, 168 species):

Knowledge of phylogenetic relationships among Australo-Papuan hylids is poorly understood, beyond the pervasive paraphyly of nominal “Litoria” with respect to the other two genera, Nyctimystes and Cyclorana (Tyler and Davies, 1978; King et al., 1979; Tyler, 1979; Maxson et al., 1985; Hutchinson and Maxson, 1987; Haas and Richards, 1998; Haas, 2003; Faivovich et al., 2005). Faivovich (2005) noted one morphological synapomorphy of Phyllomedusinae + Pelodryadinae, the presence of a tendon of the m. flexor ossis metatarsi II arising only from distal tarsi 2–3. Evidence for the monophyly of Pelodryadinae remains unsettled. Haas (2003), on the basis of six exemplars, recovered the subfamily as paraphyletic with respect to hylines and phyllomedusines. Tyler (1971c) noted the presence of supplementary elements of the m. intermandibularis in both Pelodryadinae (apical) and Phyllomedusinae (posterolateral). These characters were interpreted by Duellman (2001) as nonhomologous and therefore independent apomorphies of their respective groups. If these conditions are homologues as suggested by Faivovich et al. (2005) on the basis of their preferred cladogram, the polarity between the two characters is ambiguous because either the pelodryadine or the phyllomedusinae condition might be ancestral for Phyllomedusinae + Pelodryadinae. Because our study aims to provide a general phylogenetic structure for amphibians, not to resolve all systematic problems, and in light of ongoing research by S. Donnellan, we have not sampled “Litoria” densely enough to provide a detailed resolution of relationships within Pelodryadinae. Nevertheless, we sampled densely enough to provide additional evidence regarding the paraphyly of “Litoria” with respect to Cyclorana or Nyctimystes. Species sampled in this group are Cyclorana australis, “Litoriaaurea, “L.” freycineti, “L.” genimaculata, “L.” inermis, “L.” lesueurii, “L.” meiriana, “L.” nannotis, Nyctimystes dayi, and N. pulcher. All pelodryadines appear to have free-living exotrophic larvae (Tyler, 1985; Altig and McDiarmid, 1999).

Phyllomedusinae (7 genera, 52 species):

There is abundant morphological and molecular evidence for the monophyly and phylogenetic structure of this subfamily of bizarre frogs. Synapomorphies of the group include: (1) vertical pupil; (2) ventrolateral position of the spiracle; (3) arcus subocularis of larval chondrocranium with distinct lateral processes; (4) ultralow suspensorium; (5) secondary fenestrae parietales; and (6) absence of a passage between the ceratohyal and the ceratobranchial I (Haas, 2003). Faivovich et al. (2005) discussed several other characters likely to be synapomorphies of Phyllomedusinae. Faivovich et al. (2005) demonstrated on the basis of molecular data that Cruziohyla is the sister taxon of the remaining genera, which are further divided in two clades, one containing Phasmahyla and Phyllomedusa, and the other containing the remaining genera (Agalychnis, Hylomantis, Pachymedusa, and Phrynomedusa). Our taxon sampling reflects this understanding: Agalychnis callidryas, Cruziohyla calcarifer, Phasmahyla guttata, and Phyllomedusa vaillanti.

Bufonidae (35 genera, 485 species):

Bufonidae is a worldwide hyloid clade of noncontroversial monophyly, although the 35 genera for the most part are weakly diagnosed (e.g., Andinophryne, Bufo, Crepidophryne, Pelophryne, and Rhamphophryne). Ford and Canntella (1993) suggested the following synapomorphies for Bufonidae: (1) presence of a Bidder's organ (although absent in Melanophryniscus [Echeverria, 1998] and Truebella [Graybeal and Cannatella, 1995]); (2) unique pattern of insertion of the m. hyoglossus; (3) absence of the m. constrictor posterior (Trewevas, 1933); (4) teeth absent (also in some basal telmatobiines, Allophryne, some dendrobatids, and some rhacophorids); (5) origin of the m. depressor mandibulae solely from the squamosal and associated angle or orientation of the squamosal (Griffiths, 1954; also in several other anurans—see Manzano et al., 2003); (6) presence of an “otic element”, an independent ossification in the temporal region that fuses to the otic ramus of the squamosal (Griffiths, 1954; also known in two genera of Ceratophryini, Ceratophrys and Chacophrys, but unknown in LepidobatrachusWild, 1997, 1999). Ford and Cannatella (1993) considered characters 2–6 to be insufficiently surveyed but likely synapomorphic. Da Silva and Mendelson (1999) also noted the possibility that the possession of inguinal fat bodies and having a xiphisternum free from the underlying m. rectus abdominis are synapomorphies of Bufonidae, or some subtaxon of that group.

Dubois (1983, 1987 “1985”) recognized five nominal subfamilies, not predicated on any phylogenetic hypothesis or, seemingly, any concern for monophyly (Graybeal and Cannatella, 1995; Graybeal, 1997).

Graybeal and Cannatella (1995) provided a discussion of the monophyly of most of the genera within Bufonidae that is extremely useful. They noted that many bufonid genera are monotypic and therefore not eligible for tests of monophyly: Altiphrynoides Dubois, 1987 “1986”; Atelophryniscus McCranie, Wilson, and Williams, 1989; Bufoides Pillai and Yazdani, 1973; Crepidophryne Cope, 1889; Didynamipus Andersson, 1903, Frostius Cannatella, 1986; Laurentophryne Tihen, 1960; Mertensophryne Tihen, 1960; Metaphryniscus Señaris, Ayarzagüena, and Gorzula, 1994; Pseudobufo Tschudi, 1838; Schismaderma Smith, 1849; and Spinophrynoides Dubois, 1987 “1986”.

Graybeal and Cannatella (1995) noted that many genera lack evidence of monophyly: Adenomus Cope, 1861 “1860”; Andinophryne Hoogmoed, 1985; Bufo Laurenti, 1768; Nectophrynoides Noble, 1926; Pedostibes Günther, 1876 “1875”; Pelophryne Barbour, 1938; Peltophryne Fitzinger, 1843; Rhamphophryne Trueb, 1971; Stephopaedes Channing, 1979 “1978”; and Wolterstorffina Mertens, 1939. Graybeal and Cannatella (1995) noted the following genera to show evidence of monophyly: Ansonia Stoliczka, 1870; Atelopus Duméril and Bibron, 1841; Capensibufo Grandison, 1980; Dendrophryniscus Jiménez de la Espada, 1871 “1870”; Leptophryne Fitzinger, 1843; Melanophryniscus Gallardo, 1961; Nectophryne Buchholz and Peters, 1875; Nimbaphrynoides Dubois, 1987 “1986”; Oreophrynella Boulenger, 1895; Osornophryne Ruiz-Carranza and Hernández-Camacho, 1976; Truebella Graybeal and Cannatella, 1995; and Werneria Poche, 1903.

Graybeal (1997) provided the latest estimate of phylogeny within the entire Bufonidae. Unfortunately, although the morphological results were presented, the morphological data matrix and morphological transformation series were not, though they presumably are available in her unpublished dissertation (Graybeal, 1995). Her DNA sequence data and analytical methods are available, however. There have been serious reservations published about the quality of Graybeal's 16S sequence data (Harris, 2001; Cunningham and Cherry, 2004)15 and the paper was largely a narrative largely focused on comparing parsimony, maximum-likelihood, and neighbor-joining techniques. For our discussion we present two of her trees that rest on analytical assumptions similar to our own: (1) a strict consensus of 82 equally parsimonious trees based on the unweighted molecular data alone (fig. 25A); and (2) her combined morphology + molecular tree (fig. 25B). Her molecular results suggest that, of the exemplars treated in that particular tree (fig. 25A), Melanophryniscus is the sister taxon of all other bufonids, and Atelopus + Osornophryne form the sister taxon of the remaining bufonids, excluding Melanophryniscus. (This would suggest that presence of a Bidder's organ is not a synapomorphy of Bufonidae, but of a smaller component of that taxon.) She also suggested that Peltophryne (the Bufo peltocephalus group) is far from other New World bufonids, that Bufo gargarizans is far from her other exemplar of the B. bufo group (B. bufo), and that the two members of the B. viridis group (sensu Inger, 1973), B. calamita and B. viridis, are isolated phylogenetically from each other. Nevertheless, resolution was not strongly corroborated. The combined morphology + molecular analysis provides less resolution at the base of the tree and placed Bufo viridis and B. calamita far apart, but it did resolve the Bufo bufo group as monophyletic (B. bufo and B. gargarizans being her exemplars). Beyond that, her results do not offer a great deal of resolution. Although Graybeal (1997; fig. 25) and, more recently Pauly et al. (2004; see “Taxonomy of Living Amphibians”) provided estimates of bufonid phylogeny and started to delineate the paraphyly of “Bufo” within Bufonidae, taxonomy within “Bufo” remains largely parsed among similarity-based species groups (Blair, 1972b; Cei, 1972; Inger, 1972; R.F. Martin, 1972). These species groups have enjoyed considerable popularity and longevity of use, but, with exceptions, it is not clear whether their recognition continues to be helpful in promoting scientific progress, inasmuch as no attempt so far has been made to formulate these groups in phylogenetic terms.

Grandison (1981) provided a phylogenetic data set for African bufonids that she assumed were closely related to Didynamipus. Her data were reanalyzed and her tree was corrected by Graybeal and Cannatella (1995), and this tree is presented herein (fig. 26). On the basis of Grandison's (1981) evidence, Dubois (1987 “1985”) partitioned former Nectophrynoides into four nominal genera: Spinophrynoides (with aquatic larvae), Altiphrynoides (with terrestrial larvae), Nectophrynoides (ovoviviparous), and Nimbaphrynoides (viviparous). Graybeal and Cannatella's (1995; fig. 26) reanalysis suggests, at least on the basis of Grandison's (1981) evidence, that “Nectophrynoides” (sensu stricto) remains paraphyletic. Nectophryne and Wolterstorffina also appear paraphyletic in this tree, although Graybeal and Cannatella (1995) suggested additional characters in support of the monophyly of Nectophryne. This topology may be deeply flawed, however, because Graybeal's (1997) tree of morphology and molecules (fig. 25) show that among the exemplars shared with the study of Grandison (1981), Altiphrynoides, Didynamipus, Nectophrynoides, and Werneria are not necessarily particularly closely related. Didynamipus, in particular, is more closely related to Asian Pelophryne and South American Oreophrynella than to the others in the group addressed by Grandison (1981).

Cunningham and Cherry (2004) provided a DNA sequence study of putatively monophyletic African 20-chromosome Bufo (fig. 27). They suggested that the 20-chromosome toads form a monophyletic group with a reversal to 22-chromosomes in the Bufo pardalis group (their exemplars being B. pardalis and B. pantherinus). They also suggested that Stephopaedes and Bufo lindneri (a member of the B. taitanus group, long associated with Mertensophryne and Stephopaedes) form a monophyletic group, that on the basis of larval morphology also includes Mertensophryne. The sister taxon of this Mertensophryne group they suggested is the Bufo angusticeps group, with more distant relatives being the Bufo vertebralis group and Capensibufo.

Because of this lack of a corroborated global phylogeny of Bufonidae16, we attempted to sample as widely as possible. The nominal bufonid taxa we, unfortunately, were unable to sample are Adenomus, Altiphrynoides, Andinophryne, Atelophryniscus, several of the species groups of nominal Bufo, Bufoides, Churamiti, Crepidophryne, Frostius, Laurentophryne, Leptophryne, Mertensophryne, Metaphryniscus, Nimbaphrynoides, Oreophrynella, Parapelophryne, Pseudobufo, and Truebella. Several of these (e.g., Andinophryne, Bufoides, and Pseudobufo) are likely imbedded within sampled genera.

At least some of the bufonids are descriptively firmisternal, such as Atelopus, Dendrophryniscus, Melanophryniscus, Oreophrynella, and Osornophryne. Others (Leptophryne) approach this condition (Laurent, 1986; but see Kaplan, 2004, for discussion of the various meanings of “firmisterny”). Some bufonids exhibit various kinds of endotrophy: Altiphrynoides (nidicolous; M.H. Wake, 1980), Didynamipus (direct development; Grandison, 1981), Laurentophryne (direct development; Grandison, 1981), Nectophryne (nidicolous; Scheel, 1970), Nectophrynoides (oviductal-ovoviviparous; Orton, 1949), Nimbaphrynoides (viviparous; Lamotte and Xavier, 1972), Oreophrynella (direct development; McDiarmid and Gorzula, 1989), and Pelophryne (nidicolous; Alcala and Brown, 1982). Others are also suspected to have endotrophic larvae or direct development: Crepidophryne, Dendrophryniscus, Frostius, Metaphryniscus, Osornophryne, Rhamphophryne, Truebella, and Wolterstorffina (Peixoto, 1995; Thibaudeau and Altig, 1999). Unfortunately, our inability to sample any of these taxa other than Didynamipus, Nectophryne, Nectophrynoides, Osornophryne, Pelophryne, Rhamphophryne, and Wolterstorffina prevents us from elucidating the details of the evolution of life history in this group or the considerable morphological variation in bufonid larvae, including such things as fleshy accessory respiratory structures on the head (e.g., on Stephopaedes, Mertensophryne, Bufo taitanus) and flaps on the head (Schismaderma).

Regardless of the taxa we could not include, we were able to sample a worldwide selection of 62 bufonid species: Ansonia longidigitata, A. muelleri, Atelopus flavescens, A. spumarius, A. zeteki, Bufo alvarius, B. amboroensis, B. andrewsi, B. angusticeps, B. arenarum, B. cf. arunco, B. asper, Bufo aspinia, B. biporcatus, B. boreas, B. brauni, B. bufo, B. camerunensis, B. celebensis, B. cognatus, B. coniferus, B. divergens, B. galeatus, B. granulosus, B. guttatus, B. gutturalis, B. haematiticus, B. latifrons, B. lemur, B. maculatus, B. margaritifer, B. marinus, B. mazatlanensis, B. melanostictus, B. nebulifer, B. punctatus, B. quercicus, B. regularis, B. schneideri, B. spinulosus, B. terrestris, B. tuberosus, B. viridis, B. woodhousii, Capensibufo rosei, C. tradouwi, Dendrophryniscus minutus, Didynamipus sjostedti, Melanophryniscus klappenbachi, Nectophryne afra, N. batesi, Nectophrynoides tornieri, Osornophryne guacamayo, Pedostibes hosei, Pelophryne brevipes, Rhamphophryne festae, Schismaderma carens, Stephopaedes anotis, Werneria mertensi, and Wolterstorffina parvipalmata. This sampling, while not dense overall given the size of Bufonidae, allows a rigorous test of the monophyly and placement of Bufonidae among anurans, as well as a minimal test of the monophyly and relationships of many groups. Most important, the results, together with the obvious deficiencies in taxon sampling, will provide an explicit reference point for future, more thorough studies of the internal phylogenetic structure of Bufonidae.

Ranoidea:

Ranoidea is an enormous group of frogs, arguably monophyletic, grouped largely on the basis of one complex morphological character of the pectoral girdle (i.e., firmisterny, the fusion of the epicoracoid cartilages), except where considered to be nonhomologous (possibly Dendrobatidae, some bufonids and pipids; see Kaplan, 1994, 1995, 2000, 2001, 2004, for discussion). In addition, most ranoids are reported as diplasiocoelous, although the definitions of amphicoely, anomocoely, procoely, diplasiocoely, as well as ectochordy (= perichordy), epichordy, holochordy, and stegochordy (= epichordy) in frogs remains controversial17. Ranoidea contains noncontroversially Arthroleptidae, Astylosternidae, Hemisotidae, Hyperoliidae, Mantellidae, Microhylidae, Petropedetidae, Ranidae, and Rhacophoridae. More controversially included (see above) is Dendrobatidae, which is placed by various authors within Hyloidea. Haas (2003; fig. 15) did not recover Ranoidea as monophyletic in his analysis of larval characteristics, instead finding major ranoid groups (e.g., ranids, rhacophorids, hemisotids + hyperoliids + microhylids) interspersed among various hyloid groups (e.g., Physalaemus, Pleurodema, Odontophrynus + Leptodactylus, and bufonids). Discussing the evidence that supports the monophyly of the various ranoid groups is extremely difficult, partly because of the highly contingent nature of the evidence and, more commonly because historically the groups were assembled on the basis of overall similarity or special pleading for specific characteristics.

As understood by most workers, the questions regarding Ranoidea fall into two categories: (1) What are the phylogenetic relationships within Microhylidae?; and (2) What are the phylogenetic relationships within “Ranidae” (sensu lato as including all other ranoid subfamilial and familial taxa). The possibility of paraphyly of “Ranidae” (sensu lato) with respect to Microhylidae does not seem to have been considered seriously. We know of no definitive evidence that would reject this hypothesis, although microhylids predominantly have broadly dilated sacral diapophyses, a presumed plesiomorphy, and ranoids predominantly have round sacral diapophyses (Noble, 1931; J.D. Lynch, 1973), although in the absence of an explicit cladogram the optimization of this transformation and the number of convergences is questionable. Nevertheless, we will restrict our comments to the ranoids, excluding microhylids, while noting that any study of ranoid phylogenetics must address the position of microhylids within the ranoid framework.

Within the nonmicrohylid ranoid group, modern progress in our understanding must be dated from the publication of Dubois (1981), in which he presented a discussion of ranoid nomenclature with reference to the attendant published morphological diversity of Ranidae as then understood. Although nonphylogenetic in outlook, subsequent papers by Dubois (1983, 1984b, 1987 “1985”, 1992) provided workers with phenotypic groupings and a working taxonomy that in earlier manifestations, at least, were useful as rough approximations of phylogenetic groups. This approach was criticized for its lack of a phylogenetic rationale and overgeneralization of characters (Inger, 1996). But because there was little else with which to work, the taxonomies of Dubois have been influential. The most substantive differences between Dubois' classifications (e.g., Dubois, 1992, 2005) and those of other authors (e.g., Vences and Glaw, 2001) revolve around category-rank differences, particularly with respect to the rank and content of Rhacophoridae (variably including Mantellidae as a subfamily, or as Rhacophorinae placed as a subfamily within Ranidae or with Mantellidae and Rhacophoridae as distinct families), with the status of the various components of “Ranidae” left as an open question. With the exception of the recent papers by Marmayou et al. (2000) and Roelants et al. (2004), which dealt only with Asian taxa, and Van der Meijden et al. (2005), which focused on an African clade, no comprehensive attempt has been made to address the phylogenetics of the entire Ranoidea.

Arthroleptidae, Astylosternidae, and Hyperoliidae:

Arthroleptidae, Astylosternidae, and Hyperoliidae are poorly understood African families that have been joined and separated by various authors (Dubois, 1981; Laurent, 1984b; Dubois, 1987 “1985”, 1992) and even suggested to be related to at least two microhylid subgroups, Scaphiophryninae (Laurent, 1951) and Brevicipitinae (Van der Meijden et al., 2004). Ford and Cannatella (1993) regarded Arthroleptidae (sensu Dubois, 1981; including Astylosternidae) as a metataxon (Donoghue, 1985; Estes et al., 1988; Archibald, 1994), even though no evidence was suggested to support the monophyly of a group composed of Arthroleptidae and Astylosternidae and as originally proposed was considered to be paraphyletic (Laurent, 1951) with respect to Hyperoliidae (Hyperoliinae in Laurent's usage). Laurent (1986) suggested that the group composed of Arthroleptidae, Astylosternidae, and Hyperoliidae is distinguished from Ranidae (sensu lato) by having: (1) a cartilaginous metasternum without a bony style (presumably plesiomorphic at this level of generality); (2) second carpal free; (3) third distal tarsal free; (4) terminal phalanges generally hooked; (5) pupil usually vertical (usually horizontal in Hyperoliinae, although vertical in some—e.g., Afrixalus, Heterixalus, Kassina, Phlyctimantis; vertical in Leptopelinae); and (7) metatarsal tubercle absent or poorly developed. None of these characters is demonstrably synapomorphic.

Arthroleptidae (sensu dubois, 1992; 3 genera, 49 species):

Laurent and Fabrezi (1986 “1985”) provided a discussion of the phylogeny of genera within this African taxon and suggested a relationship of (Arthroleptis + Coracodichus) + (Cardioglossa + Schoutedenella), although the evidence for this scenario is unclear. Like astylosternines and hyperoliids, arthroleptids possess a cartilaginous sternum, a vertical pupil (horizontal in most hyperoliines), and a free second distal carpal, all of which are questionable as to level of universality and polarity. The monophyly of this taxon has never been rigorously tested by phylogenetic analysis within a well-sampled larger group although Biju and Bossuyt (2003; fig. 21), on the basis of a relatively small sampling of frogs found Hyperoliidae to be polyphyletic, and Vences et al.'s (2003c; figs. 28, 29) analysis of mtDNA suggested that Arthroleptis, Schoutedenella, and Cardioglossa form a clade, either as the sister taxon of Astylosternidae + Leptopelinae, or as the sister taxon of Hyperoliinae. We sampled: Arthroleptis tanneri, A. variabilis, Cardioglossa gratiosa, C. leucomystax, Schoutedenella schubotzi, S. sylvatica, S. taeniata, and S. xenodactyloides. We were unable to sample a member of Coracodichus (if recognized as distinct from Arthroleptis). Within Arthroleptidae, Arthroleptis, Schoutedenella, and Coracodichus have direct development (Laurent, 1973), but Cardioglossa have free-living, feeding larva (Lamotte, 1961; Amiet, 1989; Altig and McDiarmid, 1999; Thibaudeau and Altig, 1999).

Astylosternidae (5 genera, 29 species):

The African Astylosternidae traditionally has been allied with Arthroleptidae and Hyperoliidae (see above), although the evidentiary justification for this appears to be overall similarity rather than synapomorphy. Like arthroleptines and hyperoliids, astylosternids have a cartilaginous sternum, a vertical pupil (except in Leptodactylodon), and a free second distal carpal, all of which are questionable as to level of universality. For our analysis we sampled one species of each nominal genus: Astylosternus schioetzi, the presumably closely related Trichobatrachus robustus, and Leptodactylodon bicolor, Nyctibates corrugatus, and Scotobleps gabonicus. Vences et al. (2003c; figs. 28, 29) suggested on the basis of mtDNA evidence that Leptopelinae (Hyperoliidae) is either imbedded within a paraphyletic Astylosternidae or a paraphyletic Arthroleptidae, but they did not express this in the taxonomy. Astylosternus and Trichobatrachus have exotrophic aquatic larvae; in Scotobleps, the larva is unknown; and in Leptodactylodon the exotrophic aquatic larva has an upturned mouth presumably to feed on the surface film (Amiet, 1970).

Hyperoliidae (18 genera, 2 subfamilies, 250 species):

The African treefrogs of the family Hyperoliidae are currently divided into two subfamilies: Hyperoliinae, which is united by the presence of a gular gland (Drewes, 1984), and Leptopelinae, which was found by Vences et al. (2003c18; figs. 28, 29) to be more closely related to Astylosternidae than to Hyperoliinae. Vences et al. (2003c) further discussed some of the characters that Drewes (1984) used in his analysis of the family. Like Hyperoliinae, Leptopelinae lacks fusion of the second tarsal element and fusion of the second distal carpal (Drewes, 1984; fig. 30). Channing (1989) reanalyzed the morphological data provided by Drewes (1970; fig. 30) and provided different cladistic interpretations of these data; this reanalysis and the underlying characters were discussed in detail by J.A. Wilkinson and Drewes (2000). All hyperoliids for which it is known have free-living exotrophic larvae (Altig and McDiarmid, 1999). With the exception of a partial revision of Hyperolius (Wieczorek et al., 1998; Wieczorek et al., 2000, 2001), only the intergeneric relationships within Hyperoliidae have been addressed phylogenetically (Drewes, 1984; Channing, 1989; Richards and Moore, 1998) and paraphyly of Hyperolius and Kassina remain strong possibilities. Our genetic sampling included four species of the sole genus in Leptopelinae (Leptopelis argenteus, L. bocagei, L. sp., and L. vermiculatus). Of Hyperoliinae we were less complete, as we were not able to sample any member of Callixalus, Chlorolius, Chrysobatrachus, Kassinula, Phlyctimantis, or Semnodactylus. Nevertheless, we were able to obtain genetic samples of all remaining genera: Acanthixalus spinosus, Afrixalus fornasinii, A. pygmaeus, Alexteroon obstetricans, Heterixalus sp., H. tricolor, Hyperolius alticola, H. puncticulatus, H. tuberilinguis, Kassina senegalensis, Nesionixalus thomensis (transferred back into Hyperolius during the course of this study by Drewes and Wilkinson, 2004), Opisthothylax immaculatus, Phlyctimantis leonardi, and Tachycnemis seychellensis.

Hemisotidae (1 genus, 9 species):

Relationships of the African taxon Hemisotidae are also unclear (Channing, 1995). Like Rhinophrynus and Brachycephalus, Hemisus lacks a distinguishable sternum. Haas' (2003; fig. 15) study of larval morphology found it to be the sister taxon of Hyperoliidae among his exemplars. Blommers-Schlösser (1993) suggested on the basis of one morphological synapomorphy (median thyroid gland) that hemisotines should be united with brevicipitine microhylids. That hemisotines and brevicipitines are quite dissimilar cannot be disputed (Channing, 1995; Van Dijk, 2001), and the putative phylogenetic relationship between the two taxa was corroborated via molecular data only recently (Van der Meijden et al., 2004; fig. 31), although Loader et al. (2004) could not place with confidence Hemisus with brevicipitines on the basis of mtDNA sequence evidence. Emerson et al. (2000b) on the basis of mtDNA and a small amount of morphology also allied hemisotids with microhylids, although hemisotids retain a Type IV tadpole unlike the Type II tadpoles of microhylids (or direct development as in brevicipitines). We sampled only Hemisus marmoratus, one species of the single genus, of this morphologically compact taxon.

On the basis of the tree of Van der Meijden et al. (2004), Dubois (2005) recognized an enlarged family, Brevicipitidae, composed of six subfamilies: Astylosterninae, Arthroleptinae, Brevicipitinae, Hemisotinae, Hyperoliinae, and Leptopelinae. For our discussion we maintain the older, more familiar taxonomy.

Microhylidae (69 genera, 432 species):

Microhylidae is a worldwide, arguably well-corroborated taxon (J.D. Lynch, 1973; Starrett, 1973; Blommers-Schlösser, 1975; Sokol, 1975, 1977; Wassersug, 1984; Haas, 2003; but see Van der Meijden et al., 2004 (fig. 31), who suggested that the taxon is paraphyletic with respect to the hemisotines), although the internal relationships of Microhylidae are certainly problematic (Burton, 1986; Zweifel, 1986, 2000). The subfamilial taxonomy or taxonomic differentia have not changed materially since the revision by Parker (1934), with the exception of the treatment of Phrynomeridae as a subfamily of Microhylidae by J.D. Lynch (1973), the inclusion of the Scaphiophryninae by Blommers-Schlösser (1975), and the demonstration of the evolutionary propinquity of Hemisotidae and Brevicipitinae by Van der Meijden et al. (2004; fig. 31). Beyond the isolation of brevicipitines from other microhylids, the allozyme data of Sumida et al. (2000a) suggest that the subfamilial definitions and generic assignments within nominal Genyophryninae and Asterophryninae may require change. Indeed, Savage (1973) had synonymized the two subfamilies on the basis of their geographical and morphological similarity.

Savage (1973) suggested that Dyscophinae is polyphyletic, with the Asian Calluella more closely related to asterophryines than to the Madagascan Dyscophus. Blommers-Schlösser (1976) reviewed the controversy and retained Dyscophus and Calluella in Dyscophinae. Our taxon sampling allows us to test whether Dyscophinae is monophyletic or diphyletic.

Ford and Cannatella (1993) identified five larval synapomorphies for Microhylidae (although these cannot be documented in lineages with direct development such as in brevicipitines, asterophryines, and genyophrynines, so the level of universality of these characters is questionable): (1) absence of keratodonts in tadpoles; (2) ventral velum divided medially; (3) glottis fully exposed on buccal floor; (4) nares not perforate; and (5) secretory ridges of branchial food traps with only a single row of secretory cell apices. In addition, adults are characterized as having 2–3 palatal folds (palatal folds also being found in Hemisus). Van de Meijden et al. (2004) suggested on the basis of molecular evidence that Hemisotidae + Brevicipitinae is more closely related to Hyperoliidae, Arthroleptidae, and Astylosternidae than to an otherwise monophyletic group of microhylids (fig. 31). Therefore, the only articulated questions so far regarding the monophyly of Microhylidae are whether Hemisotidae is imbedded within it (see above) or, with Brevicipitinae, more closely related to non-microhylid groups, although the definition, historical reality, and content of the various subfamilies are controversial.

Scaphiophryninae (2 genera, 11 species):

The Madagascan microhylid subfamily Scaphiophryninae has no demonstrable synapomorphies in support of its monophyly, but if its monophyly is assumed it is widely considered to be the sister taxon of the remaining Microhylidae. At least some authors (e.g., Dubois, 1992) regard it as a distinct family. Ford and Cannatella (1993) suggested that larval synapomorphies that place it in association with the remaining Microhylidae (at least for those that have larvae) are (1) the possession of a median spiracle in the larvae; (2) gill filaments poorly developed or absent; (3) modifications of buccal pumping mechanisms (short lever arm on ceratohyal, small buccal floor area); (4) absence of m. suspensoriohyoideus; and (5) lack of separation of the mm. quadrato-, hyo-, and suspensorioangularis. Parker (1934) reported the taxon as diplasiocoelus like most other ranoids, although he noted Hoplophryninae (Parhoplophryne + Hoplophryne), Asterophryinae, and some members of his Microhylinae (e.g., Melanobatrachus, Metaphrynella, Myersiella, Phrynella) as procoelous. Parker (1934) noted that Scaphiophryne retains a complete sphenethmoid, thereby excluding it from Microhylidae, which, as he applied the name, included only those taxa where the sphenethmoid is either in two parts, or, more rarely, not ossified at all. Haas (2003) suggested on the basis of larval morphology that Scaphiophryninae is polyphyletic, with Scaphiophryne forming the sister taxon of the remaining microhylids, and Paradoxophyla more closely related to Phrynomerinae. Clearly the monophyly of this taxon is controversial, but we, unfortunately, were unable to sample Paradoxophyla and so could not test the monophyly of Scaphiophryninae. We were able to obtain only a representative of the other genus, Scaphiophryne marmorata. Our results regarding the Scaphiophryninae will therefore remain incomplete.

Asterophryinae (8 genera, 64 species) and Genyophryninae (11 genera, 142 species):

Zweifel (1972) and Burton (1986) last reported on phylogenetics of the Australo-Papuan Asterophryinae (fig. 32). Genyophryninae is also Australo-Papuan but extends into the Philippines and Lesser Sundas. No major revision or broad-scale phylogenetic study has appeared since Parker (1934), although Burton (1986) did suggest evidence that it is paraphyletic with respect to Asterophryinae. Sumida et al. (2000a) noted that some allozyme evidence suggested that Asterophryinae is imbedded within a paraphyletic Genyophryninae. Savage (1973) considered Genyophryninae to be part of Asterophryinae based on the dubious nature of the procoely–diplasiocoely distinction; that they share direct-development; and, in part, that they are both biogeographically centered in New Guinea.

Zweifel (1971) summarized the distinction between the subfamilies as (1) maxillae often overlapping the premaxillae and usually in contact anteriorly (Asterophryinae; this presumably is apomorphic), maxillae not overlapping the premaxillae (Genyophryninae); (2) vertebral column diplasiocoelous (rarely procoelous; Asterophryinae), procoelous (Genyophryninae); and 3) tongue subcircular, entirely adherent, often with a median furrow and posterior pouch (Asterophryinae), tongue oval, half-free behind, no trace of a median furrow or pouch (Genyophryninae; shared with Cophylinae). Genyophryninae and Asterophryinae share direct development (Zweifel, 1972; Thibaudeau and Altig, 1999). Our sampling will allow us to test the hypotheses of relationship so far published and elucidate the possible paraphyly of Genyophryninae. Unfortunately, we could sample only one species of Asterophryinae, Callulops slateri, which will not allow us to test its monophyly. The effect of excluding representatives of Asterophrys, Barygenys, Hylophorbus, Mantophryne, Pherohapsis, Xenobatrachus, and Xenorhina is unknown.

Of Genyophryninae, we were able to sample Aphantophryne pansa, Choerophryne sp., Cophixalus sphagnicola, Copiula sp., Genyophryne thomsomi, Liophryne rhododactyla, Oreophryne brachypus, and Sphenophryne sp. We were unable to sample Albericus, Austrochaperina, Oreophryne, or Oxydactyla.

Brevicipitinae (5 genera, 22 species):

Like the Australo-Papuan Asterophryninae and Genyophryninae, the African Brevicipitinae has direct development (Parker, 1934; Thibaudeau and Altig, 1999). Parker (1934) considered the subfamily to be distantly related to all other microhylid taxa and characterized by the retention of a medially expanded vomer. Parker (1934) reported the taxon as diplasiocoelus like most other ranoids. The species within the subfamily are closely similar and unlike all other microhylids in general habitus, although the monophyly of the group has never been tested rigorously.

Blommers-Schlösser (1993) suggested (the presence of a median thyroid gland being the sole synapomorphy) that brevicipitines should be united with hemisotids (but see Channing, 1995, who considered this change premature at the time because of the otherwise trenchant differences between them). Van der Meijden et al. (2004; fig. 31) and Loader et al. (2004) provided molecular data in support of a hemisotid–brevicipitine relationship. Of the nominal genera we were unable to sample the monotypic Balebreviceps and Spelaeophryne. The effect of excluding these taxa is unknown, although Loader et al. (2004) recovered Spelaeophryne in a clade with Probreviceps and Callulina, to the exclusion of Breviceps. We were able to sample at least one species of the remaining nominal genera: Breviceps mossambicus, Callulina kisiwamitsu, C. kreffti, and Probreviceps macrodactylus. Because there are 13 species of Breviceps and 3 species of Probreviceps, we were unable to test rigorously the monophyly of these taxa.

Cophylinae (7 genera, 41 species):

The Madagascan Cophylinae is similar to Dyscophinae and Genyophryninae in retaining maxillary and vomerine teeth (except in Stumpffia) but differs from Dyscophinae in having procoelous vertebrae and from Dyscophinae and Genyophryninae in having a divided vomer (Parker, 1934); none of the characters is demonstrably synapomorphic. Blommers-Schlösser and Blanc (1993) provided a cladogram (fig. 33A) of the genera based on nine morphological characters, in which they suggested that Plethodontohyla was paraphyletic and that Platypelis did not have apomorphies to assure its monophyly. Andreone et al. (2004 “2005”) recently provided a maximum likelihood analysis of 1173 bp of mtDNA (fig. 33B), in which he documented Plethodontohyla paraphyly. Because these sequences became available after our analyses were completed, we did not sample Cophyla, Madecassophryne, or Rhombophryne. The effect of this on the placement of the subfamily will remain unknown, although we did sample four species of the four remaining genera: Anodonthyla montana, Platypelis grandis, Plethodontohyla sp., and Stumpffia psologlossa. Unfortunately, because of our limited sampling we will not be able to test rigorously the results of either Blommers-Schlösser and Blanc (1993) or Andreone et al. (2004 “2005”). Species of Cophylinae have nidicolous larvae (Blommers-Schlösser and Blanc, 1991; Glaw and Vences, 1994).

Dyscophinae (2 genera, 10 species):

The Madagascan Dyscophinae is distinguished from most other microhylid subfamilies by retaining maxillary and vomerine teeth, otherwise known only in Cophylinae and Genyophryninae, both of which are procoelous rather than diplasiocoelous (Parker, 1934) as in Dyscophinae. Savage (1973) had regarded Calluella as associated with the direct-developing Asterophryinae and any similarities with Dyscophus as reflecting plesiomorphy. We sampled one species each of the two nominal genera: Calluella guttulata and Dyscophus guineti.

Melanobatrachinae (3 genera, 4 species):

On the basis of geography alone (East Africa [2 genera] and southern India [1 genus]), one would suspect that this is not a monophyletic taxon. Nevertheless, the three genera share an incomplete auditory apparatus (convergent in Balebreviceps [Brevicipitinae]; Largen and Drewes, 1989) and fusion of the sphenethmoid with the parasphenoid (Parker, 1934). Savage (1973), followed by Laurent (1986) and Dubois (2005), placed Melanobatrachus in Microhylinae and retained Hoplobatrachus and Parhoplophryne in Hoplophryninae, but did so only by discarding absence of the auditory apparatus and fusion of the sphenethmoid to the parasphenoid, as convergences, without offering specific characters that conflicted with these as synapomorphies. Although we are suspicious of the monophyly of this taxon, we stick with the most parsimonious hypothesis (monophyly of Melanobatrachinae, sensu lato) until alternative evidence emerges.

Apparently based on information provided for Hoplophryne by Barbour and Loveridge (1928) and Noble (1929), Parker (1934) generalized that all members of his Melanobatrachinae lack a free-swimming tadpole, the larvae with “metamorphosis taking place on land, but not in an egg”. No reproductive or developmental data on Parahoplophryne or Melanobatrachus have been published (Daltry and Martin, 1997). Thibaudeau and Altig (1999) listed Melanobatrachus and Parhoplophryne as having endotrophic larvae, presumably because of the earlier statement by Parker (1934). McDiarmid and Altig (1999: 13), however, listed Hoplophryne as exotrophic, because Barbour and Loveridge (1928: 256) reported vegetable matter in the guts of larvae and because R. Altig examined AMNH larvae of Hoplophryne and inferred that they could feed (R.W. McDiarmid, personal commun.). Laurent (1986) reported the taxon (Parhoplophryne and Hoplophryne in his Hoplophrynnae; Melanobatrachus in his Microhylinae) as procoelous, unlike most other ranoids so this may also be synapomorphic. Unfortunately, we were able to sample only Hoplophryne rogersi and so will not be able to comment on the monophyly of Melanobatrachinae.

Microhylinae (30 genera, 133 species):

The American and tropical Asian Microhylinae have free-swimming tadpoles (except for a few species, such as Myersiella microps, that have direct development; Izecksohn et al., 1971). Although microhylines can be morphologically characterized, they have no known synapomorphies, and their monophyly is deeply suspect. According to Parker (1934), maxillary and vomerine teeth are absent (as in several other extra-Madagascar subfamilies); the vomer is much reduced and usually divided; the sphenethmoid is divided or absent; and the vertebrae are diplasiocoelous (or rarely procoelous). Wild (1995) provided a cladogram of New World genera (fig. 34), but this assumed that the New World group is monophyletic and was unclear about the outgroup(s) used to polarize the transformations. Laurent (1986) treated the Old World and New World components separately, implying some kind of taxonomic division. This was followed, without discussion, by Dubois (2005), who recognized Gastrophryninae for the New World component and Microhylinae for the Old World component. We are not aware of any evidence in support of this arrangement so we retain the old taxonomy. Of the 30 nominal genera we were able to sample representatives of 14: Chaperina fusca, Ctenophryne geayei, Dasypops schirchi, Dermatonotus muelleri, Elachistocleis ovalis, Gastrophryne elegans, G. olivacea, Hamptophryne boliviana, Kalophrynus pleurostigma, Kaloula pulchra, Microhyla heymonsi, Microhyla sp., Micryletta inornata, Nelsonophryne aequatorialis, Ramanella obscura, and Synapturanus mirandaribeiroi). We were not able to sample Adelastes, Altigius, Arcovomer, Chiasmocleis, Gastrophrynoides, Glyphoglossus, Hyophryne, Hypopachus, Metaphrynella, Myersiella, Otophryne, Phrynella, Relictovomer, Stereocyclops, Syncope, and Uperodon. Most of these appear to be clustered with sampled taxa. The exclusion of Otophryne and Uperodon, however, is particularly regrettable. Our sampling will not allow detailed elucidation of the evolution of life-history strategies. Adelastes, Altigius, Gastrophrynoides, Hyophryne, Kalophrynus (nidicolous), Myersiella (direct development), Phrynella, Synapturanus (nidicolous), and Syncope (nidicolous) have endotrophic larvae that exhibit (or are suspected to exhibit) various degrees of truncation of larval development (Thibaudeau and Altig, 1999). That we lack representatives of about half of these is lamentable, but our results will provide an explicit starting point for future, more detailed studies. The remaining genera have exotrophic larvae of typical microhylid morphology (Altig and McDiarmid, 1999).

Phrynomerinae (1 genus, 5 species):

The African Phrynomerinae is diagnosable from Microhylinae solely by possessing intercalary cartilages between the ultimate and penultimate phalanges (Parker, 1934). Like most other ranoids it is diplasiocoelous. Of this small taxon we sampled Phrynomantis bifasciatus. Phrynomantis typically has aquatic, exotrophic microhylid larvae (Altig and McDiarmid, 1999).

“Ranidae” (ca. 54 genera, 772 species):

Ranidae is a large ranoid taxon, that is likely paraphyletic with respect to Mantellidae and Rhacophoridae—at least on the basis of molecular evidence (Vences and Glaw, 2001; Roelants et al., 2004; Van der Meijden et al., 2005). Ford and Cannatella (1993; fig. 14) suggested that the group is paraphyletic, or, at least, that it does not have recognized synapomorphies. Nevertheless, Haas (2003; fig. 15) suggested the following to be synapomorphies for Ranidae, excluding other ranoids: (1) cartilaginous roofing of the cavum cranii present as taenia transversalis and medialis; (2) free basihyal present; and (3) firmisterny (convergent elsewhere in Haas' tree).

Laurent (1986) included the mantellines and rhacophorids in his Ranidae, a content that allows at least two other characters (distinctly notched tongue and bony sternal style) to be considered as possible synapomorphies (Ford and Cannatella, 1993). (These are, however, incongruent with characters suggested by Haas, 2003).

Dubois and coauthors (Dubois, 1992; Dubois and Ohler, 2001; Dubois et al., 2001; Dubois, 2005) suggested a taxonomy of 11– 14 subfamilies of uncertain monophyly or relationship with respect to each other. For discussion, we recognize Dubois' subfamilies, except as noted. As discussed by Inger (1996), the diagnostic features supporting Dubois' (1992) classification at the time of that writing frequently reflected overgeneralized and postfacto approximations for clusters that were aggregated with overall similarity, not synapomorphy, as the organizing principle. The relationships suggested by this taxonomy (and Dubois, 2005, as well) can be at variance with evidence of monophyly, notably evidence from DNA sequences (Emerson and Berrigan, 1993; Bossuyt and Milinkovitch, 2000; Emerson et al., 2000b; Marmayou et al., 2000; Biju and Bossuyt, 2003; Roelants et al., 2004), so this taxonomy requires careful evaluation.

Ceratobatrachinae (6 genera, 81 species):

Ceratobatrachinae is composed of direct-developing species found from western China (i.e., Ingerana) to the Indo-Australian archipelago (Batrachylodes, Discodeles, Palmatorappia, Platymantis, and the monotypic Ceratobatrachus). Ceratobatrachinae represents the direct-developing part of Cornuferinae sensu Noble (1931) and Platymantinae of later authors (e.g., Savage, 1973; Laurent, 1986). Those taxa formerly included in Cornuferinae or Platymantinae that exhibit unforked omosterna and/or free-living tadpoles (what are now Amolops, Huia, Meristogenys, Staurois, Hylarana [sensu lato], and Micrixalus) are now placed in Raninae or Micrixalinae. Batrachylodes is inferred to have direct development (Noble, 1931; Brown, 1952; Duellman and Trueb, 1986; Thibaudeau and Altig, 1999), but unlike other members of Ceratobatrachinae, Batrachylodes has an entire omosternum (rather than being forked). Noble (1931) regarded Batrachylodes as derived from his Cornufer (= Platymantis) and, by inference, exhibiting direct development. Because of the character conflict of omosternum shape and life-history, Brown (1952) regarded Batrachylodes as related either to “Hylarana” (exotrophic, entire omosternum) or to the Ceratobatrachus group (direct-developing, forked omosternum). Laurent (1986) treated Batrachylodes as a member of Raninae, although Boulenger (1920) had noted the intraspecific plasticity of omosternum shape, the only evidence supporting placement of Batrachylodes in Raninae. This arrangement was accepted by Dubois (1987 “1985”), although subsequently, Dubois (2005) transferred Batrachylodes out of Raninae and into Ceratobatrachinae, presumably on the basis of the direct development. Our analysis should provide more evidence on the placement of this taxon.

Dubois (1992) recognized Ceratobatrachini within his Dicroglossinae, but later (Dubois et al., 2001) considered it to be a subfamily, of unclear relationship to Dicroglossinae. Even later, Dubois (2003) stated, on the basis of unpublished molecular data, that Ceratobatrachini is a tribe within Dicroglossinae. Van der Meijden et al. (2005) presented DNA sequence evidence that Ceratobatrachus is outside of Dicroglossinae, and on that basis (Dubois, 2005) once again embraced the subfamilial rank Ceratobatrachinae. Roelants et al. (2004; fig. 35), in a study of predominantly Indian taxa, provided molecular evidence that suggest that Ingerana is in Occidozyginae, rather than in Ceratobatrachinae, although Dubois (2005), without discussion, did not accept this.

Of this group we sampled Batrachylodes vertebralis, Discodeles guppyi, Ceratobatachus guentheri, Platymantis pelewensis, P. weberi, and Ingerana baluensis. Thus, we only lack Palmatorappia from this group19. Although we obviously cannot test the monophyly of these individual genera (except Platymantis), our taxon sampling is adequate to test the monophyly of the inclusive group.

Conrauinae (1 genus, 6 species):

Until the recent publication by Dubois (2005), this genus (Conraua) had been placed on the basis of overall similarity in a monotypic tribe, Conrauini, in Dicroglossinae (Dubois, 1992). Conrauini was proposed (Dubois, 1992) for the West African genus Conraua, the diagnostic characters being the retention of a free-living tadpole stage (plesiomorphic), with a larval keratodont formula of 7–8/6– 11 (see Dubois, 1995, for the definition of keratodont formula) and lateral line not retained into adulthood (plesiomorphic). Van der Meijden et al. (2005; fig. 36), on the basis of DNA sequence data, showed that Conraua is not close to Dicroglossinae but the sister taxon to a taxonomically heterogeneous group of southern African ranoids, including Afrana, Cacosternum, Natalobatrachus, Petropedetes, Pyxicephalus, Strongylopus, and Tomopterna. Kosuch et al. (2001; figs. 38, 39), on a relatively small amount of evidence, had previously placed Conraua alternatively as either the sister taxon of Limnonectes (based on 16S alone) or as the sister taxon of Tomopterna + Cacosternum (based on combined 12S and 16S). The latter result was suggestive of the more complete results of Van der Meijden et al. (2005). Although characters have not been suggested that are clearly synapomorphic, the group is morphologically compact and monophyly is likely. Of the six species we sampled two: Conraua robusta and C. goliath.

Dicroglossinae (12 genera, 152 species):

Recounting the taxonomic history of Dicroglossinae is difficult inasmuch as it was originally formed on the basis of overall similarity, and the content has varied widely, even by the same authors. Only recently has its concept begun to be massaged by phylogenetic evidence. Dubois (1987 “1985”, 1992) diagnosed Dicroglossinae (in the sense of including Conrauinae and excluding Paini) as having the omosternum moderately or strongly bifurcate at the base and the nasals usually large and in contact with each other and with the frontoparietal, although none of these characters is demonstrably synapomorphic. The most recent taxonomy of Dicroglossinae (Dubois, 2005) recognized four tribes: Dicroglossini (for Euphlyctis, Fejervarya, Hoplobatrachus, Minervarya, Nannophrys, and Sphaerotheca), Limnonectini (for Limnonectes, as well as some taxa considered by most authors to be synonyms of Limnonectes), Occidozygini (for Occidozyga and Phrynoglossus), and Paini (for Chaparana, Nanorana, and Quasipaa).

Dicroglossini was diagnosed by Dubois (1992; in the sense of including Occidozyginae) as retaining a free-living tadpole (plesiomorphic) and having a lateral line system that usually is retained into adulthood (presumably apomorphic, but not present in Occidozyga, sensu stricto). As conceived by Dubois (1992), the taxon contained Euphlyctis, Occidozyga, and Phrynoglossus. Fei et al. (1991 “1990”) and, subsequently, Dubois et al. (2001) on the basis of published and unpublished molecular evidence (Marmayou et al., 2000fig. 37; Kosuch et al., 2001figs. 38; Delorme et al., 2004fig. 40) placed Occidozyga and Phrynoglossus in the subfamily Occidozyginae, and transferred without discussion into Dicroglossini Fejervarya and Hoplobatrachus (from Limnonectini) and Sphaerotheca (from Tomopterninae), and Nannophrys (from Ranixalinae).

Grosjean et al. (2004), building on the earlier work of Kosuch et al. (2001) suggested on the basis of several mtDNA and nuDNA loci that Euphlyctis is the sister taxon of Hoplobatrachus with Fejervarya, Sphaerotheca, Nannophrys, and Limnonectes forming more distant relations, a result that is consistent with the tree of Roelants et al. (2004; fig. 35).

Dubois (1992) also recognized a tribe Limnonectini diagnosed nearly identically with Conrauini (Conrauinae of this review), differing only in the larval keratodont formula of 1–5/2–5, which is arguably plesiomorphic. Nominal genera contained in this group occur from tropical Africa to tropical Asia with most taxonomic diversity being in Asia: Hoplobatrachus, Limnonectes, and Fejervarya (which was considered a subgenus of Limnonectes at the time). In addition Marmayou et al. (2000; fig. 37) and Delorme et al. (2004; fig. 40) suggested on the basis of mtDNA evidence that Sphaerotheca (formerly in Tomopterninae; Dubois, 1987 “1985”) and Taylorana (now a synonym of Limnonectes; originally considered to be a member of Limnonectini [Dubois, 1987 “1985”] but subsequently transferred to Ceratobatrachinae by Dubois, 1992) are in Limnonectini.

Sphaerotheca, therefore, is likely not to be closely related to Tomopterna, as one would have expected given that the species of Sphaerotheca were long placed in Tomopterna (Pyxicephalinae). Roelants et al. (2004; fig. 35) also placed Nannophrys in Dicroglossinae (by implication) on the basis of mtDNA and nuDNA evidence, substantiating the earlier assessment by Kosuch et al. (2001; figs. 38) which was made on less evidence. It was previously assigned to Ranixalini by Dubois (1987 “1985”) and to Dicroglossini by Dubois et al. (2001). Dubois et al. (2001: 55) implied on the basis of various published and unpublished mtDNA data that Euphlyctis (formerly in his Dicroglossini), Fejervarya, Hoplobatrachus, Minervarya, Nannophrys, and Sphaerotheca (formerly in his Limnonectini) should be included in a reconstituted Dicroglossini.

Delorme et al. (2004; fig. 40) demonstrated—as had Roelants et al. (2004; fig. 35)— that Lankanectes is phylogenetically distant from Limnonectes.

Of these taxa we sampled rather broadly: Euphlyctis cyanophlyctis; Fejervarya cancrivorus, F. kirtisinghei, F. limnocharis, and F. syhadrensis; Hoplobatrachus occipitalis and H. rugulosus; Limnonectes acanthi, L. grunniens, L. heinrichi, L. kuhlii, L. limborgi (formerly Taylorana limborgi), L. poilani, and L. visayanus; Nannophrys ceylonensis; Sphaerotheca breviceps and S. pluvialis. On the basis of this sampling we should be able to evaluate the reality of this taxon and, at least to some degree, the monophyly of the contained genera.

Occidozygini is a tropical Asian group of arguable position. Marmayou et al. (2000; fig. 40) presented mtDNA evidence that Occidozyga and Phrynoglossus are not within Dicroglossinae but are outside of a clade composed of Rhacophoridae and other members of a paraphyletic Ranidae. Fei et al. (1991 “1990”) had already transferred Occidozyga (sensu lato) out of Dicroglossinae and into its own subfamily on the basis of larval characters and this evidence supported the view that Dicroglossinae, as previously conceived, is polyphyletic. Roelants et al.'s (2004) greater sampling of Asian ranoids suggested that Ingerana (nominally in Ceratobatrachinae) is in this clade and together form the sister taxon of a reformulated Dicroglossinae (fig. 35), which together are the sister taxon of a clade composed of Mantellidae, Rhacophoridae, and Raninae. No African taxa were examined by Marmayou et al. (2000; fig. 37), Roelants et al. (2004; fig. 35), or Delorme et al. (2004; fig. 40), so the relative position and monophyly of Occidozyginae and Dicroglossinae needed to be further elucidated. This issue was addressed by Van der Meijden et al. (2005; fig. 36), who did analyze Asian and African taxa simultaneously and found Occidozygya lima to sit within their Dicroglossinae. Dubois (2005), on the strength of the evidence produced by Van der Meijden et al. (2005), returned Occidozyginae to Dicroglossinae as a tribe. We sampled Phrynoglossus baluensis, P. borealis, P. martensii, and Occidozyga lima.

Paini is a montane Asian tribe diagnosed among ranids by having an unforked omosternum (and was therefore formerly included in Raninae by Dubois, 1987 “1985”, 1992) and males having black, keratinous ventral spines (presumably a synapomorphy with Nanorana; Jiang et al., 2005: 357). Paini according to Dubois (1992) was composed of two genera, each with four subgenera: genus Chaparana with subgenera Annandia, Chaparana, Feirana, and Ombrana; genus Paa with subgenera Eripaa, Gynandropaa, Paa, and Quasipaa. Dubois et al. (2001), citing unpublished DNA sequence, suggested that Paini be transferred from Raninae to Dicoglossinae. Jiang and Zhou (2001, 2005; fig. 41), Jiang et al. (2005; fig. 42), Roelants et al. (2004; fig. 35), and Van der Meijden (2005; fig. 36) on the basis of published DNA sequence evidence, suggested that Dicroglossinae, with a forked omosternum, is paraphyletic with respect to Paini, with an unforked omosternum. For this reason Roelants et al. (2004) and Jiang et al. (2005) transferred Paini out of Raninae and into Dicroglossinae. Larvae in the group are exotrophic and aquatic (Altig and McDiarmid, 1999).

Jiang et al. (2005) recently provided a phylogenetic study (fig. 42) of Paini on the basis of 12S and 16S rRNA fragments. Unfortunately, that study appeared too late to guide our choice of terminals, but their results are important in helping us interpret our own results. They found Paa to be paraphyletic with respect to Chaparana and Nanorana; Chaparana to be polyphyletic with the parts imbedded within “Paa”; and Nanorana to be deeply imbedded within “Paa”. Within Paini they recognized two groups: (1) Group 1, composed of “Chaparana”, several species of “Paa”, and Nanorana, characterized by spines forming two patches on the chest (save C. quadranus, the type of subgenus Feirana, which does not have spines on the chest); and (2) Group 2, composed of “Paa” species associated previously with the subgenera Quasipaa Dubois, 1992 (P. robertingeri), and one species nominal of the genus Chaparana, subgenus Feirana Dubois, 1992 (Paa yei). The second group is characterized by having spines as a single group, more or less over the entire venter, but this characteristic is sufficiently variable among subgroups as not to be diagnostic practically except in the not-Nanorana group sense. These authors recommended that the generic name Quasipaa be applied to Group 2, but for unstated reasons hesitated to resolve taxonomically the nonmonophyly of Chaparana and Paa in their Group 1. Nanorana Günther, 1896, is the oldest available name for their first group.

Three nominal genera are definitely included in Paini: “Chaparana” (polyphyletic; see above); Nanorana; and “Paa” (paraphyletic with respect to “Chaparana” and Nanorana20). We sampled Nanorana pleskei, Quasipaa exilispinosa and Q. verrucospinosa but did not sample “Chaparana” or “Paa” (sensu stricto).

Jiang et al. (2005) did not mention or address three supraspecific taxa usually associated with Paini. The first is Eripaa Dubois, 1992, whose type and only species is Rana fasciculispina Inger, 1970. Eripaa Dubois, 1992, was named and is currently treated as a subgenus of Paa. Although Eripaa exhibits spines on the entire chest and throat, such as in group 2 of Jiang et al. (2005), they are uniquely distinct from all other “Paa”, “Chaparana”, and Nanorana species in that these spines are clustered in groups of 5–10 on circular whitish tubercles. We cannot hazard a guess as to how Eripaa is related to the rest of Paini. The second is Annandia Dubois, 1992, whose type and only species is Rana delacouri Angel, 1928. Annandia was originally named as a subgenus of Chaparana Bourret, 1939, but recently, Dubois (2005), without discussion of evidence, treated Annandia as a genus in Limnonectini. Perhaps this was done because this species bears a smooth venter, with spinules only clustering around the anus (Dubois, 1987 “1986”). Regardless, this is a large taxonomic change (from Paini to Limnonectini) and because no evidence was produced or discussed to justify this change, we must consider the status of this taxon questionable. The third is Ombrana Dubois, 1992, whose type and only species is Rana sikimensis Jerdon, 1870). Ombrana Dubois, 1992, was originally proposed as a subgenus of Chaparana. This species also posseses spinules only around the anus, prompting Dubois (1987, “1986”) to consider it evidence of a unique reproductive mode, and thus a close relative of Annandia delacouri. Unfortunately, we did not sample any of these three taxa, so their status will remain questionable.

Lankanectinae (1 genus, 1 species):

This subfamily was named for Lankanectes corrugatus of Sri Lanka by Dubois and Ohler (2001). Its distinguishing features are (1) forked omosternum (plesiomorphy); (2) vomerine teeth present (presumed plesiomorphy); (3) median lingual process absent (likely plesiomorphy); (4) femoral glands absent (likely plesiomorphy); (5) toe tips not enlarged (arguable polarity); (6) tarsal fold present (likely plesiomorphy at this level); and (7) lateral line system present in adults (also in Phrynoglossus and Euphlyctis, but presumably apomorphic). Roelants et al. (2004; fig. 35) and Delorme et al. (2004; fig. 40) subsequently suggested on the basis of mtDNA and nuDNA evidence that Lankanectes is far from Limnonectes, where it had been placed by Dubois (1992). Roelants et al. (2004) placed it as the sister taxon of Nyctibatrachinae, and Delorme et al. (2004) placed it as the sister taxon of Nyctibatrachinae + Raninae. We sampled the sole species, Lankanectes corrugatus.

Micrixalinae (1 genus, 11 species):

Tropical Asian Micrixalus (11 species) is the sole member of this taxon, diagnosed by Dubois (2001) as differing from Dicroglossinae in lacking a forked omosternum (possibly apomorphic), lacking vomerine teeth, having digital discs (present in some limnonectines and otherwise widespread in Ranoidea) and having a larval keratodont formula in its aquatic tadpoles of 1/0 (likely apomorphic) (Dubois et al., 2001). On the basis of mtDNA and nuDNA evidence, Roelants et al. (2004; fig. 35) considered Micrixalinae to be the sister taxon of Ranixalinae. We were able to sample Micrixalus fuscus and M. kottigeharensis. Although this provides only a minimal test of the monophyly of Micrixalus, it allows us to place the taxon phylogenetically.

Nyctibatrachinae (1 genus, 12 species):

Nyctibatrachinae contains the Indian taxon Nyctibatrachus and is characterized by having a forked omosternum (likely plesiomorphic), vomerine teeth present, digital discs present, femoral glands present (shared with Ranixalinae and some Dicroglossinae) and an aquatic tadpole with a keratodont formula of 0/0 (likely apomorphic; Dubois et al., 2001). Of this taxon we sampled Nyctibatrachus cf. aliciae and N. major.

Petropedetinae (2 genera, 10 species); Phrynobatrachinae (4 genera, 72 species) and Pyxicephalinae (13 genera, 57 species):

Until recently, members of Petropedetinae and Phrynobatrachinae, as well as several genera now assigned to Pyxicephalinae (e.g., Anhydrophryne, Arthroleptella, Cacosternum, Microbatrachella, Natalobatrachus, Nothophryne, and Poyntonia) were considered members of “Petropedetidae” (sensu lato), aggregated on the basis of overall similarity, with no evidence for its monophyly ever suggested. Noble (1931) recognized his Petropedetinae (Arthroleptides and Petropedetes), as united by the possession of dermal scutes on the upper surface of each digit and otherwise corresponding osteologically and morphologically with Raninae. Noble (1931) also recognized Cacosterninae for Cacosternum and Anhydrophryne, united by lacking a clavicle and having palatal ridges. He related the cacosternines to brevicipitines, and the remainder of the genera then named he allocated to Raninae.

Laurent (1941 “1940”) addressed the confusion between Arthroleptis and Phrynobatrachus and transferred Petropedetes, Anhydrophryne, Phrynobatrachus (including Natalobatrachus), Dimorphognathus, and Arthroleptella into his Phrynobatrachinae. Laurent (1941) subsequently provided an anatomical characterization of the group.

Laurent (1951) transferred Cacosterninae into Ranidae and moved Microbatrachella into Cacosterninae. Poynton (1964a) suggested that Phrynobatrachus is deeply paraphyletic with respect to Cacosterninae and therefore considered Laurent's Phrynobatrachinae (= Petropedetinae) and Cacosterninae to be synonyms. Subsequent authors (e.g., Dubois, 1981; Frost, 1985) uncritically followed this unsupported suggestion, although there have been significant instances of workers continuing to recognize Cacosterninae and Petropedetinae as distinct (e.g., Liem, 1970; J.D. Lynch, 1973).

Another morphologically compact African group was Pyxicephalinae (Dubois, 1992), composed of Pyxicephalus (2 species) and Aubria (2 species). The taxon was diagnosed by at least four synapomorphies (Clarke, 1981): (1) cranial exostosis; (2) occipital canal present in the frontoparietal; (3) zygomatic ramus being much shorter than otic ramus; and (4) sternal style a long bony element tapering markedly from anterior to posterior. Dubois' (1992) reasoning for excluding this taxon from Dicroglossinae is not clear, but presumably had to do with the distinctive appearances of Pyxicephalus and Aubria.

Dubois (1992) also recognized a subfamily Tomopterninae, for Tomopterna (sensu lato, at the time including Sphaerotheca, now in Dicroglossinae, Limnonectini). The diagnosis provided by Clarke (1981) presumably applies inasmuch as he examined only African species (Tomopterna, sensu stricto), even though the optimization of these characters on his cladogram may well be contingent on being compared only with other African ranids: (1) zygomatic ramus much shorter than otic ramus; (2) outline of anterior end of cultriform process pointed, with lateral borders tapering to a point; (3) distal end of the anterior pterygoid ramus overlapping the dorsal surface of the posterior lateral border of the palatine; (4) no overlap of the anterior border of the parasphenoid ala by the medial ramus of the pterygoid in the anterior–posterior plane; (5) sternal style short, tapering posteriorly; (6) dorsal protuberance of the ilium not or only slightly differentiated from the spikelike dorsal prominence; and (7) terminal phalanges of the fingers and toes reduced, almost conelike.

In 2003 this untidy, but familiar arrangement began to unravel. Dubois (2003), removed Cacosterninae from “Petropedetidae” without discussion, apparently anticipating evidence to be published elsewhere, although Kosuch et al. (2001; fig. 38) had suggested earlier that Cacosternum was more closely related to Tomopterna and Strongylopus than it was to Petropedetes. The content of this taxon was stated to be Anhydrophryne, Arthroleptella, Cacosternum, Microbatrachella, Nothophryne, Poyntonia (from Petropedetidae), and, possibly Strongylopus and Tomopterna (from Ranidae).

Van der Meijden et al. (2005; fig. 36) suggested Phrynobatrachus to be the sister taxon of Ptychadena. On this basis Dubois (2005) recognized a ranid subfamily Phrynobatrachinae, containing Phrynobatrachus, but also allocated to this subfamily, without discussion, Dimorphognathus, Ericabatrachus, and Phrynodon. Petropedetes and Conraua formed successively more distant outgroups of the southern African clade of Van der Meijden et al. (2005), so Dubois (2005) removed Conrauini (Conraua) from Dicoglossinae and placed it in its own subfamily, Conrauinae, and recognized Petropedetinae for Petropedetes, as well as the presumably closely allied Arthroleptides. The southern African clade of Van der Meijden et al. (2005; fig. 36) was composed of Cacosternum (formerly of Petropedetidae), Afrana and Strongylopus (formerly of Raninae), Natalobatrachus (formerly of Petropedetidae), Tomopterna (Tomopterninae), and Pyxicephalus (Pyxicephalinae), a group that Dubois (2005) allocated to an enlarged Pyxicephalinae. Aubria was asserted by Dubois (2005) to be in this group because it was grouped by morphological evidence with Pyxicephalus. Amietia he transferred into the group without discussion, but presumably because they appeared to him to be related to Strongylopus and Afrana. He transferred Arthroleptella, Microbatrachella, Nothophryne, and Poyntonia into Pyxicephalinae, presumably because he thought that they were more likely to be here than close to either Petropedetinae or Phrynobatrachinae.

Of Dubois' (2005) Petropedetinae (which presumably is diagnosed as by Noble, 1931) we were able to sample both genera: Arthroleptides sp. and Petropedetes cameronensis, P. newtoni, P. palmipes, and P. parkeri.

Of the newly constituted Phrynobatrachinae, we were also able to sample species from three of four genera: Dimorphognathus africanus, Phrynobatrachus auritus, P. calcaratus, P. dendrobates, P. dispar, P. mababiensis, P. natalensis, and Phrynodon sandersoni. We did not sample Ericabatrachus, an unfortunate omission, inasmuch as we are unaware of the evidence for Dubois' (2005) association of Ericabatrachus with Phrynobatrachinae, other than the statement that it is “Phrynobatrachus-like” (Largen, 1991). Phrynobatrachus, at least for the species which it is known, have exotrophic larvae. Larvae are unknown in Dimorphognathus and Ericabatrachus, and Phrynodon is endotrophic (Amiet, 1981; Altig and McDiarmid, 1999).

Of the reformulated Pyxicephalinae we were able to sample Aubria (Aubria subsigillata [2 samples21]) and Pyxicephalus (Pyxicephalus edulis) as well as several of the taxa recently transferred into this taxon including Anhydrophryne rattrayi, Arthroleptella bicolor, Cacosternum platys, and Natalobatrachus bonebergi. We also sampled members of Afrana (A. angolensis and A. fuscigula), Tomopterna (T. delalandii), Strongylopus (S. grayii), and Amietia (A. vertebralis), but for reasons having to do with the evidentiary basis and history of taxonomy in Raninae, considerable discussion of these genera is presented there. We did not sample Microbatrachella, Nothophryne, or Poyntonia. Pyxicephalines have exotrophic larvae, with the exception of Anhydrophryne and Arthroleptella, which are endotrophic; unknown in Nothophryne (Hewitt, 1919; Procter, 1925; DeVilliers, 1929; Altig and McDiarmid, 1999). This selection should allow us to test the phylogenetic results of Van der Meijden et al. (2005).

Ptychadeninae (3 genera, 51 species):

Ptychadeninae is a morphologically compact group of sub-Saharan ranids diagnosed (Clarke, 1981; Dubois, 1987 “1985”, 1992) by having: (1) an otic plate of the squamosal covering the crista parotica in dorsal view and extending mesially to overlap the otoccipital; (2) palatines absent; (3) clavicles reduced; (4) sternal style a short compact element tapering anteriorly to posteriorly; (5) eighth presacral vertebra fused with sacral vertebra; and (6) the dorsal protuberance of ilium smooth-surfaced and not prominent. The three nominal genera in the taxon are Ptychadena (47 species), Hildebrandtia (3 species), and Lanzarana (1 species) of which we sampled only Ptychadena anchietae, P. cooperi, and P. mascareniensis. Because we did not sample Hildebrandtia and Lanzarana, we did not adequately test the monophyly of this group. Nevertheless, assuming the group to be monophyletic, our three species of Ptychadena allow us to test the placement of Ptychadeninae within Ranoidea. For his analysis Clarke (1981) assumed that Ptychadeninae is imbedded within other African ranids, although a lack of comparison with Asian members of the group makes this assumption questionable. Van der Meijden et al. (2005; fig. 36) suggested that Ptychadena is the sister taxon of Phrynobatrachus among his exemplars, thereby implying that Ptychadeninae is the sister taxon of Phrynobatrachinae.

“Raninae” (ca. 8 genera, 309 species):

“Raninae” is a catch-all largely Holarctic and tropical Asian taxon united because the members do not fit into the remaining subfamilies and have unforked omosterna. Until recently, “Raninae” included two tribes: Paini and Ranini (Dubois, 1992). However, Paini and Nanorana of Ranini were transferred to Dicroglossinae on the basis of mtDNA and nuDNA evidence (Roelants et al., 2004fig. 35; Jiang et al., 2005fig. 42), so Raninae, as we use it, is coextensive with Ranini of Dubois (1992), itself dubiously monophyletic22.

“Raninae” is distributed on the planet coextensively with the family and is united by the lack of putative apomorphies, either in the adult or in the larvae. There does not appear to be any reason to suggest that this nominal taxon is monophyletic.

The starting point of any discussion of Ranini must be Dubois (1992), who provided an extensive, and controversial, taxonomy. Because the distinction between ranks (section, subsection, genus, and subgenus) in Dubois' system appears to rest primarily on subjective perceptions of similarity and difference, the evidentiary basis of this taxonomy is unclear, even though we accepted his system as a set of bold phylogenetic hypotheses. Nevertheless, most of these taxa are imperfectly or incompletely diagnosed and to lay the foundation for our results and concomitant taxonomic remedies, we discuss this taxonomy in greater depth than we do most of the remainder of current amphibian taxonomy. Suffice it to say that we think that we sampled “Rana” diversity sufficiently to provide at least a rudimentary phylogenetic understanding of the taxon as a starting point for future, more densely sampled studies.

Within his Ranini, Dubois (1992) recognized six genera: Amolops, Batrachylodes, Nanorana, Micrixalus, Rana, and Staurois (table 4). Of these, two continue to be placed in this taxon (Amolops and Rana [sensu lato]) (Dubois, 2005). Staurois, Nanorana and Micrixalus have subsequently been transferred out of Ranini, Staurois to a new tribe, Stauroini (Dubois, 2005), Nanorana to Dicroglossidae (Roelants et al., 2004; fig. 35), and Micrixalus to a distant Micrixalinae (Dubois et al., 2001). Batrachylodes was provisionally transferred, without substantial discussion, by Dubois (2005) to Ceratobatrachinae.

Within both Amolops and Rana, Dubois recognized several subgenera, that other authors (e.g., Yang, 1991b) considered to be genera, as we do, although we arrange the discussions by Dubois' genera and subgenera. Dubois (2003) arranged Raninae into two tribes (Amolopini for the taxa with cascade-adapted tadpoles, i.e., Amo, Amolops, Huia, Meristogenys, Chalcorana, Eburana, Odorrana) and Ranini (for everything else). This system represents typical nonevolutionary A and not-A groupings, although Amolopini in this form is testable. Dubois (2005) subsequently did not embrace Amolopini, because it was too poorly understood, but he did erect Stauroini for Staurois, because Roelants et al. (2004) placed Staurois as the putative sister taxon of other ranines.

Amolops, Amo, Huia, and Meristogenys: Amolops has been recognized in some form since Inger (1966) noted the distinctive tadpole morphology (presence of a raised, sharply defined abdominal sucker). Like other cascade-dwelling taxa, larvae of Amolops (sensu lato) all share high numbers of keratodont rows. Subsequently, Yang (1991b) recognized two other genera from within Amolops: Meristogenys and Huia. Amolops (sensu stricto) has one possible synapomorphy (short first metacarpal, also found in Huia), and three synapomorphies joining Huia and Meristogenys to the exclusion of Amolops (lateral glands present in larvae; four or more uninterrupted lower labial keratodont rows; and longer legs).

Subsequently, Dubois (1992) treated Meristogenys and Huia as subgenera of Amolops, and added a fourth subgenus, Amo (including only Amolops larutensis). Amo was diagnosed (Boulenger, 1918) as having a digital disc structure similar to species of Staurois (i.e., having a transverse groove or ridge on the posteroventral side of the disc continuous with a circummarginal groove to define a hemisphere; Boulenger, 1918) and as having axillary glands (after Yang, 1991b) that are otherwise unknown in Amolops.

Although Dubois (1992) considered Amolops (sensu stricto), Amo, Huia, and Meristogenys to be subgeneric parts of a monophyletic genus Amolops, other authors (e.g., Yang, 1991b) considered at least Amolops, Huia, and Meristogenys as genera. For consistency we treat as genera Amo, Amolops, Huia, and Meristogenys. Our samples were Amolops (A. chapaensis, A. hongkongensis), Huia (H. nasica), and Meristogenys (M. orphocnemis). We were unable to sample Amo larutensis.

Staurois: The definition of Staurois (digital discs broader than long; T-shaped terminal phalanges in which the horizontal part of the T is longer than the longitudinal part; outer metatarsals separated to base but joined by webbing; small nasals separated from each other and frontoparietal; omosternal style not forked [Boulenger, 1918]) has also been used to define Hylarana (Boulenger, 1920; see below). Although some larval characters are shared among species of Staurois (deep, cup-like oral disc in the tadpole, no glands or abdominal disc in tadpole; Inger, 1966), the diagnostic value of these characters is unknown due to the large number of ranid species whose adults are morphologically similar to those of Staurois, but whose larvae remain undescribed. Our single exemplar of Staurois, S. tuberilinguis, is not sufficient to test the monophyly of the genus. Although no one has suggested that Staurois is polyphyletic, or that it is paraphyletic with respect to any other group, both of these remain untested possibilities. Roelants et al. (2004; fig. 35) provided evidence that Staurois is the sister taxon of remaining ranines.

Rana (sensu Dubois, 1992)23: Rana of Dubois (1992) is diagnostically coextensive with his Ranini (our “Raninae”), and no features provided in his paper exclude “Rana” from being paraphyletic with respect to Staurois, Amolops (sensu Dubois, 1992), or Batrachylodes. So, as we discuss the internal taxonomy of “Rana” as provided by Dubois, readers should bear in mind that Amolops (sensu lato), Batrachylodes, and Staurois, as discussed by Dubois (1992), must be regarded as potential members of all infrageneric taxa that do not have characters that specifically exclude them. (And, at least with respect to Dubois', 1992, Rana subgenera, Strongylopus and Afrana, DNA sequence data have been published that suggest that they have little relationship with other ranines [Van der Meijden et al., 2005; fig. 36].) With respect to “Rana” specifically, Dubois (1992) provided a system of sections, subsections, and subgenera that has posed serious challenges for us: Rather than a synapomorphy scheme, or even a system of carefully-evaluated characteristics, the various taxa appear to represent postfacto character justifications of decidedly nonphylogenetic and subjectively arrived-at groups. We found Dubois' (1987 “1985”, 1992) arrangement to be inconsistent with the preponderance of evidence in certain instances (see the discussion of inclusion of Aquarana in his section Pelophylax, below) and the underlying diagnostic basis of the system to contain overly-generalized statements from the literature (Inger, 1996) that are not based on any comprehensive comparative study of either internal or external morphology. For instance, larvae may have dorsal dermal glands, lateral dermal glands, or ventral dermal glands in various combinations (e.g., Yang, 1991b). These characters have become larval dermal glands present or absent in Dubois' (1992) diagnoses, thereby conflating the positional homology of these features. Although we address deficiencies here and in the Taxonomy section, for other critiques see Emerson and Berrigan (1993), Matsui (1994), Matsui et al. (1995), Inger (1996), Bain et al. (2003), and Matsui et al. (2005).

As noted earlier, several, if not most taxa recognized by Dubois within his “Rana” are effectively undiagnosed in a utilitarian sense (i.e., they are diagnosed sufficiently only to make the names available under the International Code; ICZN, 1999). In addition, several are demonstrably nonmonophyletic (Matsui, 1994; Matsui et al., 1995; Inger, 1996; Tanaka-Ueno et al., 1998a; Emerson et al., 2000a; Marmayou et al., 2000; Vences et al., 2000a; B.J. Evans et al., 2003; Roelants et al., 2004; Jiang and Zhou, 2005). Unlike the superficially similar situation in Eleutherodatylus (sensu lato) where it is straightforward to get specific information on individual species and where the nominal subgenera and most related genera, even if they do not rise to the level of synapomorphy schemes, have been diagnosed largely comparatively, the subgeneric (and generic, in part) diagnoses of ranids are not comparable, and the purported differentiating characters frequently do not bear up to specimen examination (e.g., Tschudi, 1838; Boulenger, 1920; Yang, 1991b; Fei et al., 1991 “1990”; Dubois, 1992).

Historically, taxonomists approached Rana (sensu lato) as being composed of two very poorly defined similarity groupings: (1) those that have expanded toe tips (likely plesiomorphic) that at one time or another have been covered by the name Hylarana; and (2) those that lack expanded toe tips, and that have more-or-less always been associated with the generic name Rana. Most authors since Boulenger (1920) recognized the lack of definitive “breaks” between the two groups, and Dubois was the first to attempt to summarize the relevant taxonomic literature and to divide Rana (sensu lato) into enough groups to allow some illumination of the problem. Our issue with his system is that it is impossible to tell from the relevant publication (Dubois, 1992) which species have actually been evaluated for characters and which have merely been aggregated on the basis of overall similarity or erected on the basis of specially-favored characters.

Dubois' primary division of Rana was into eight sections of arguable phylogenetic propinquity to each other or to other ranine genera (see table 4). We discuss these with reference to his diagnoses and other literature relevant to their recognition:

(1) Section Amerana. Dubois (1992) erected his subgenera Amerana and Aurorana for parts of the Rana boylii group of Zweifel (1955), which he placed in their own section, Amerana. Most previous work (e.g., Case, 1978; Farris et al., 1979; Post and Uzzell, 1981; Farris et al., 1982b; Uzzell and Post, 1986) had placed these frogs from western North American close to, or within, the Eurasian Rana temporaria group. Nevertheless, section Amerana was recognized by Dubois (1992) on the basis of a combination of characters, none unique but corresponding to the Rana boylii group identified by ribosomal data by Hillis and Davis (1986; fig. 43). This group had been suggested by Hillis and Davis (1986) to be in a polytomy with what Dubois regarded as his section Rana (R. temporaria and R. sylvatica were the exemplar species in their analysis), a group composed of a part of Dubois' section Pelophyax (Aquarana), and his sections Lithobates and Pantherana. Moreover, Hillis and Davis' (1986; fig. 43) results suggested that neither of the groups subsequently identified by Dubois (1992) as the subgenera Aurorana and Amerana are monophyletic. Subsequent work (Hillis and Wilcox, 2005; fig. 44) has provided substantial amounts of evidence in support of the nominal subgenus Aurorana being polyphyletic, and the subgenus Amerana being paraphyletic. Hillis and Wilcox (2005) used the section Amerana + Rana temporaria to root the remainder of their tree, so their overall tree cannot be taken as additional evidence of evolutionary propinquity of the section Amerana being in a monophyletic group with Rana temporaria, to the exclusion of all other North American Rana, inasmuch as this was an assumption of their analysis, based on earlier work (e.g., Case, 1978).

Dubois (1992) provided no unique morphological features to diagnose section Amerana, and because of his use of present-or-absent as a characteristic, the characters provided in his table 1 fail to rigorously distinguish section Amerana from sections Hylarana, Lithobates, Pelophylax, Rana, or Strongylopus (now in Pyxicephalinae on the basis of DNA sequence evidence—Dubois, 2005; Van der Meijden et al., 2005). Within Amerana, Dubois recognized two subgenera, Amerana and Aurorana, differing in the expansion of toe tips (mildly expanded in Amerana; not expanded in Aurorana), rows of larval keratodonts (4–7/4–6 in Amerana; 2–3/3–4 in Aurorana) karyotype (derived in Amerana; primitive in Aurorana). This subgeneric distinction is not phylogenetically consistent with the results of Hillis and Davis (1986; fig. 43), who presented evidence suggesting that Dubois' Aurorana is paraphyletic with respect to his Amerana (making one wonder what the purpose was in naming two subgenera). Macey et al. (2001) subsequently provided additional molecular evidence for paraphyly of Aurorana with respect to Amerana. Examples of this section in our analysis are Amerana muscosa and Aurorana aurora (see table 4).

(2) Section Amietia (including a single subgenus, Amietia, for two species in the Lesotho Highlands of southern Africa). The sole synapomorphy of Amietia is the umbraculum over the eye in the larva. The diagnosis of section Amietia is otherwise phylogenetically indistinguishable on the basis of the table of characters provided by Dubois (1992), from Amerana, Hylarana, Lithobates, Rana, or Strongylopus. We sampled Amietia vertebralis. Amietia was transferred into Pyxicephalinae by Dubois (2005) on the apparent but undiscussed assumption that it is closely related to Strongylopus, which was placed by Van der Meijden et al. (2005) in that group on the basis of DNA sequence evidence.

(3) Section Babina (for the Rana holsti and Rana adenopleura groups). The unique synapomorphy for this group is a large “suprabrachial” gland (sensu Dubois, 1992) on the sides of reproductive males (which can be difficult to assess in nonreproductive animals). The diagnosis of section Babina does not otherwise allow it to be practically separated from the sections Amerana, Hylarana, Lithobates, Pelophylax, Rana, or Strongylopus. Within section Babina, Dubois recognized two subgenera, Babina (with a large fingerlike prepollical spine, an apomorphy) and Nidirana (members of the Babina section lacking the apomorphy of the subgenus Babina). Fei et al. (2005) considered Nidirana to be a subgenus of their Hylarana, but their taxonomy was presented for only the Chinese fauna, so the wider implication of this action is not known. Of this section we sampled no member of the subgenus Babina, although we did sample Nidirana adenopleura and N. chapaensis. Babina and Nidirana have also been associated with “Hylarana” (see below), so Dubois' (1992) reason for recognizing this as a section distinct from section Hylarana is unclear.

(4) Section Lithobates. This section is not rigorously diagnosable by the features presented by Dubois' (1992: his table 1) from sections Amerana, Hylarana, Rana, or Strongylopus. However, Lithobates is consistent with the phylogenetic tree of American Rana provided by Hillis and Davis (1986; fig. 43), presumably the source of the concept of this section. Hillis and Davis placed this taxon, on the basis of DNA substitutions, as the sister taxon of part of Dubois' section Pelophylax, the subgenus Pantherana. Within section Lithobates, Dubois recognized four subgenera: Lithobates (Rana palmipes group), Sierrana (Rana maculata group), Trypheropsis (Rana warszewitschii group), and Zweifelia (Rana tarahumarae group). All of them are consistent with the tree provided by Hillis and Davis (1986). Dubois (1992) offered the following morphological characters which may be synapomorphies: Lithobates differs from other members of the section by having tympanum diameter larger or equal to the diameter of the eye; Sierrana without diagnostic characters that differentiate it from the section diagnosis; Trypheropsis by having an outer metatarsal tubercle (unusual in American ranids); and Zweifelia with sacrum not fused with presacral vertebrae. Hillis and Wilcox (2005; fig. 44) presented evidence that suggests that section Lithobates of Dubois (1992) is paraphyletic, with part of Dubois' subgenera Sierrana (R. maculata), and all of his subgenera Trypheropsis, and Lithobates falling within one monophyletic group, but Zweifelia (the Rana tarahumarae group) and another part of Sierrana (R. sierramadrensis) forming the sister taxon of Dubois' subgenus Pantherana, the Rana pipiens group of Hillis and Wilcox (2001).

Our exemplars of this section are Lithobates palmipes, Sierrana maculata, and Trypheropsis warszewitschii. We did not sample Zweifelia.

(5) Section Pelophylax. The characters provided by Dubois for his section Pelophylax will not rigorously diagnose it from Amerana, Hylarana, Rana, or Strongylopus. Further, the association of his subgenera Aquarana (former Rana catesbeiana group), Pantherana (former Rana pipiens group), Pelophylax (former Ranaesculenta” group), and Rugosa (Rana rugosa group) is curious inasmuch as we are unaware that anyone had previously suggested such a relationship. All published evidence that was available to Dubois at the time of his writing (e.g., Case, 1978; Post and Uzzell, 1981; Hillis and Davis, 1986; Pytel, 1986; Uzzell and Post, 1986) suggested that this section is polyphyletic, with Dubois' subgenus Pantherana (of his section Pelophylax) more closely related to his section Lithobates, than to any other member of section Pelophylax. Indeed, the subgenera Aquarana and Pantherana of Pelophylax are both more closely related to both the sections Lithobates, Rana, and Amerana, than they are to the Old World members of section Pelophylax according to the evidentiary literature (i.e, Case, 1978; Post and Uzzell, 1981; Hillis and Davis, 1986; Pytel, 1986; Uzzell and Post, 1986). There never was any evidence for the monophyly of section Pelophylax sensu Dubois, while there was considerable evidence against it. Recently, Hillis and Wilcox (2005; fig. 44) have provided molecular evidence that Aquarana (their Rana catesbeiana group) is the sister taxon of Rana sylvatica, and together the sister taxon of all other American Rana, with the exception of the section Amerana (their Rana boylii group).

The subgenera recognized by Dubois within section Pelophylax have more justification for their monophyly. Aquarana is distinct on the basis of its large snout–vent length and its tympanum diameter, which is greater than eye diameter in males. Rugosa is separated by its “small” adult snout–vent length. Pantherana and Pelophylax are separated from Aquarana and Rugosa by their “medium” size and spots on the dorsum, but are otherwise undiagnosable from each other by features presented by Dubois (1992). Fei et al. (1991 “1990”, 2005) consistently considered Pelophylax and Rugosa to be a distinct genera, but these authors generalized solely over the Chinese fauna rather than attempting to draw global distinctions. From Aquarana (Rana catesbeiana group) we sampled Aquarana catesbeiana, A. clamitans, A. grylio, and A. heckscheri. Of Pantherana (Rana pipiens group) we sampled Pantherana berlandieri, P. capito, P. chiricahuensis, P. forreri, P. pipiens, and P. yavapaiensis. Of Pelophylax we sampled R. nigromaculata and P. ridibunda. We did not sample Rugosa.

(6) Section Pseudorana. This section cannot be rigorously diagnosed on the basis of information given by Dubois (1992) from section Hylarana. Pseudorana was named by Fei et al. 1991 “1990”) as a distinct genus for Rana sauteri, R. sangzhiensis, and R. weiningensis. Subsequently, Fei et al. (2000) coined Pseudoamolops for Rana sauteri, suggesting, on the basis of its having a large ventral sucker on the tadpole, that it is more closely related to Amolops (sensu lato) than to Pseudorana. Although the ventral sucker found in Pseudoamolops is associated with the oral disc of the tadpole, in Amolops the ventral sucker sits posterior to the oral disc. Fei et al. (2000) suggested that Pseudoamolops is the sister taxon of the remainder of their Amolopinae (Amo, Amolops, Huia, and Meristogenys) and derived with respect to a paraphyletic Hylarana, although Tanaka-Ueno et al. (1998a) had previous suggested on the basis of DNA sequence analysis that Pseudorana sauteri is imbedded within the brown frog clade (Rana temporaria group), although that analysis had addressed no member of nominal Amolopinae. We were able to sample Pseudoamolops sauteri and Pseudorana johnsi to test the placement of these species.

(7) Section Rana. This section cannot be diagnosed rigorously from sections Amerana, Hylarana, Lithobates, Pelophylax, or Strongylopus on the basis of characters presented by Dubois (1992). The association of Rana sylvatica with the Rana temporaria group has been controversial, with Hillis and Davis (1986) providing weak evidence for its placement with Rana temporaria, and Case (1978) suggesting that Rana sylvatica is phylogenetically within other North American Rana (sensu lato). Hillis and Wilcox (2005; fig. 44) recently provided molecular evidence in support of Rana sylvatica being the sister taxon of the Rana catesbeiana group (Aquarana of Dubois, 1992). In addition to noncontroversial members of the Rana temporaria group (Rana japonica and R. temporaria) we sampled Rana sylvatica to test whether it was a member of the Rana temporaria group or, as suggested previously, imbedded within a North American clade.

(8) Section Strongylopus. This section also is not phylogenetically diagnosable on the basis of Dubois' (1992) suggested evidence from sections Amerana, Hylarana, Lithobates, Pelophylax, or Rana. If the autapomorphies of Babina and Amietia are not considered, there also is nothing in the diagnosis of section Strongylopus that would prevent it from being paraphyletic with respect to Babina or Amietia. Nevertheless, DNA sequence evidence of Van der Meijden et al. (2005; fig. 36) places Strongylopus in Pyxicephalinae, and Dubois (2005) presumed that Afrana and Amietia also should be so allocated. Section Strongylopus is seemingly a geographically determined unit, not a phylogenetically determined one. Within section Strongylopus, Dubois recognized two subgenera that differ in size and color of larvae (long and dorsally black in Afrana; modest length and entirely black in Strongylopus), foot length (short in Afrana; long in Strongylopus), and webbing (less webbing in Afrana than in Strongylopus).

Van der Meijden (2005; fig. 36) provided a phylogenetic tree, based on mtDNA and nuDNA sequence data, that placed Strongylopus and Afrana in a heterogeneous clade (which they termed the “southern African ranid clade”, and which Dubois, 2005, considered as an expanded Pyxicephalinae), along with Tomopterna (Tomopterninae), Cacosternum and Natalobatrachus (“Petropedetidae”), and Pyxicephalus (Pyxicephalinae). Because the evidence of Van der Meijden et al. (2005; fig. 36) is the first phylogenetic evidence that bears on this issue, we follow that taxonomy, but note that nothing in morphology so far supports this arrangement.

We sampled Afrana angolensis, A. fuscigula, and Strongylopus grayii.

(9) Section Hylarana. We have left section Hylarana to the end of this discussion because it represents the heart of the problem of “Rana” systematics. The name Hylarana has had an historically unstable application, alternatively being considered synonymous with Rana, or treated as a distinct subgenus or genus with an ill-defined content, and diagnosed in several different, even contradictory ways (e.g., Tschudi, 1838; Günther, 1859 “1858”; Boulenger, 1882, 1920; Perret, 1977; Poynton and Broadley, 1985; Laurent, 1986; Fei et al., 1991 “1990”; Dubois, 1992), although it is almost always associated with frogs that exhibit expanded toe tips. The original diagnostic character of the genus Hylarana Tschudi, 1838 (type species: Rana erythraea Schlegel, 1827) is the presence of a dilated disc on the tips of the toes (a character that can now be seen to encompass many of the species of Ranidae and its immediate outgroups). Günther (1859 “1858”) revised the diagnosis to include “males with an internal subgular vocal sac” (i.e., lacking gular pouches) as a character, and increased the composition to five Asian and African species (including Hylarana albolabris and H. chalconota).

Because of the ambiguity of the diagnostic character of dilated toe disc, Boulenger (1882, 1920) believed Hylarana to be a “group of polyphyletic origin”, but suggested that it was a subgenus of Rana, removing vocal sac condition as a diagnostic character and expanding its definition: dilated digital discs with circummarginal grooves, T-shaped terminal phalanges, and an unforked omosternal style (Boulenger, 1920: 123; as Hylorana). All of his putatively diagnostic characters have greater levels of generality than “Hylarana”. He listed 62 species from Australasia, including Rana curtipes, R. guentheri, and R. taipehensis (the latter implicit, as he synonomized it with R. erythraea; Boulenger, 1920: 152–155).

Perret (1977: 842) listed ten African species of the genus Hylarana (including H. galamensis), revising the diagnosis as follows: precoracoids ossified, transverse, approaching each other medially; metasternum ossified, elongated; males with or without gular pouches; males with brachial (humeral) glands. Poynton and Broadley (1985: 139) revised the diagnosis in their account of African Hylarana: only some species with expanded digital discs; broad brown to golden band from head to urostyle; upper lip white; males with single or paired baggy gular pouches. Laurent (1986: 761) further revised the diagnosis of Hylarana: without transverse grooves on finger discs.

Fei et al. (1991 “1990”) moved some species from Hylarana into a new genus Odorrana. They diagnosed their new genus Odorrana by having: omosternum extremely small, colorless spines present on chest of male in breeding condition. Despite the etymology of the generic name, Fei et al. (1991 “1990”), did not include odoriferous secretions as one of the characters uniting the genus. In addition, they included six species (O. anlungensis, O. kwangwuensis, O. swinhoana, O. tiannanensis, O. versabilis, and O. wuchuanensis) known not to have colorless spinules on the chest of the male. Subsequently, Ye and Fei (2001; fig. 45), on the basis of a phylogenetic study of Chinese Odorrana (including Eburana in their sense), suggested that only the Odorrana andersoni group (O. andersoni, O. grahami, O. hainanensis, and O. margaretae) have large chest spines, with small spines otherwise only in O. schmackeri. Chest spines were reported as absent in all other species of Odorrana that they studied: O. anlungensis, O. exiliversabilis, O. hejiangensis, O. kuangwuensis, O. livida, O. lungshengensis, O. nasuta, O. swinhoana, O. tiannanensis, O. versabilis, and O. wuchuanensis.

Fei et al. (1991 “1990”: 138–139) further divided Hylarana into two subgenera, Hylarana and Tenuirana based on the following characters (Tenuirana in parentheses): anterior process of hyoid long, curved outwards (long, straight); tips of digits with or without a horizontal groove (always present on toes); feet almost fully webbed (half webbed); body not long or slender (long, slender); snout blunt and rounded (long, pointed); limbs moderate (long, slender); dorsolateral folds distinct to extremely broad (narrow); humeral gland or shoulder gland present in males (absent); gular pouches present in male (absent); and tadpole vent tube dextral (medial). As part of the Chinese fauna, they included R. nigrovittata and R. guentheri (under the subgenus Hylarana) and R. taipehensis (the type species of the subgenus Tenuirana) in Hylarana. Although they did not discuss R. erythraea (the type species of Hylarana), its inclusion in the subgenus Hylarana was implied.

As noted earlier, Dubois (1992) partitioned species formerly associated with one or more of the historical manifestations of Hylarana into several sections, subsections, and subgenera (see table 4) of which the sections Babina (subgenera Babina and Nidirana) and Hylarana (subsections Hydrophylax and Hylarana) are particularly relevant to this discussion of “Hylarana”-like frogs (although the section Hylarana, in Dubois' system was not precluded by any evidence from being paraphyletic to any or all of the other sections defined by him). Sections Babina and Hylarana are distinguishable in Dubois' system solely by the possession of a suprabrachial gland (apomorphy) in section Babina. This gland is not found in section Hylarana which at least as portrayed by Dubois (1992) and noted above, has no apomorphies. All other characters overlap or are identical between the two sections.

Dubois placed the collection of subgenera that he aggregated under section Hylarana into two subsections: a humeral gland-bearing group (subsection Hydrophylax) and a group characterized by having indistinct or absent humeral glands (subsection Hylarana). The presence of a humeral gland is an apomorphy, so at least prior to analysis we considered this single character as evidence of monophyly of Dubois' subsection Hydrophylax, leaving the condition “humeral glands indistinct or absent” as plesiomorphic (although we would have liked to know the distribution of “indistinct” humeral glands within the groups where Dubois reported them as indistinct or absent). During analysis, however, Matsui et al. (2005; fig. 46) provided DNA sequence evidence suggesting that that the subsection Hydrophylax is paraphyletic at least with respect to Chalcorana chalconota and (subgenus) Hylarana (subsection Hylarana) and that subsection Hylarana is polyphyletic with Hylarana (subgenus) and Chalcorana chalconota being independently derived of the main group of subsection Hylarana, which included all of their exemplars of subgenera Eburana and Odorrana, as well as Chalcorana hosii.

Within the apomorphic subsection Hydrophylax (well-developed humeral gland-bearing group) Dubois (1992) recognized several weakly or undiagnosed (except in the nomenclatural sense) subgenera: Amnirana, Humerana, Hydrophylax, Papurana, Pulchrana, and Sylvirana. According to Dubois (1992; his table II), Humerana is distinguished from other members of the subsection by the absence of an outer metatarsal tubercle; Amnirana and Pulchrana are not rigorously diagnosable from each other; Papurana and Pulchrana are not rigorously diagnosable from each other; and Hydrophylax can be diagnosed from Sylvirana only on the basis of the absence of an expanded disc and lateral groove on finger III and toe IV. Marmayou et al. (2000; fig. 37) presented DNA sequence evidence that Sylvirana (a humeral gland-bearing taxon) is paraphyletic with respect to Hylarana (subgenus) and Pelophylax, both of which lack humeral glands, suggesting that his subsection Hydrophylax (of section Hylarana) is paraphyletic. We sampled Amnirana albilabris, Hydrophylax galamensis, Papurana daemeli, Sylvirana guentheri, S. maosonensis, S. nigrovittata, and S. temporalis. We were unable to sample any member of Pulchrana, although Matsui et al. (2005; fig. 46) provided evidence that it is related to a group of subsection Hydrophylax, including Sylvirana, as well as an imbedded piece of subsection Hylarana, Chalcorana chalconota.

The “indistinct or absent” humeral-gland group (subsection Hylarana) is not rigorously diagnosable on the basis of apomorphies from any of the other sections of Rana (except for Amietia [now in Pyxicephalinae] and Babina) or from other genera of Ranidae. We, therefore, must assume that it is a mixture of groups with no necessary phylogenetic propinquity or to the exclusion of other ranid groups. The subgenera coined and aggregated under subsection Hylarana by Dubois (1992) are variably diagnosable. Marmayou et al. (2000; fig. 37) provided DNA sequence evidence for the polyphyly of subsection Hylarana (as well as for the paraphyly of the other subsection, Hydrophylax; see above), by placing Hylarana (subgenus) and Chalcorana very distant from each other evolutionarily.

Subgenus Chalcorana (Chalcorana chalconota being our exemplar, and the type of the taxon) is a morphologically very poorly diagnosed subgenus within the subsection Hylarana, with dermal glands present or not in the larvae, outer metatarsal tubercle present or not, male with paired subgular vocal pouches present or not, animal pole of egg pigmented or not, and the only likely synapomorphy is the relative size of the fingers (I < II; Dubois, 1992). Matsui et al. (2005; fig. 46) provided evidence that Chalcorana is broadly polyphyletic, with Chalcorana chalconota close to subsection Hydrophylax and C. hosii close to members of Eburana. Matsui et al. (2005) suggested that this was not surprising as Chalcorana chalconota lays pigmented eggs and has a larval keratodont formula of 4–5/3 (Inger, 1966), whereas Chalcorana hosii has pigmentless eggs and larvae with a keratodont formula of 5–6/4. Matsui et al. (2005) transferred Chalcorana hosii into Odorrana (sensu lato, as including Eburana), with the status of the remaining species of nominal Chalcorana left questionable.

Clinotarsus is a monotypic taxon (Clinotarsus curtipes) that is also poorly diagnosed, with larvae attaining a large size and having a somewhat high (but not exclusively) larval keratodont formula of 8/6–8 (Chari, 1962; Dubois, 1992), both characteristics found in Nasirana as well. We sampled the single species, Clinotarsus curtipes.

Subgenera Eburana and Odorrana (sensu Dubois, 1992) are putatively distinguished from each other by Eburana having (1) discs with a circumlateral groove on finger III and toe IV (present or absent in Odorrana); (2) external metatarsal tubercle present or absent (absent in Odorrana); (3) gular pouches (variable, including the Eburana condition, in Odorrana); (4) no unpigmented spines on the chest in males (putatively present in Odorrana, according to Dubois, 1992, but absent in most species, being present in Odorrana only in the Odorrana andersoni group [see above] and two species of the Odorrana schmackeri group [O. schmackeri and O. lungshuengensis]; see C.-C. Liu and Hu, 1962; Hu et al., 1966, 1973; Yang and Li, 1980; L. Wu et al., 1983; Fei, 1999; Fei and Ye, 2001, Ye and Fei, 2001; see also Bain et al., 2003; Bain and Nguyen, 2004); (5) animal pole of egg unpigmented (pigmented in Odorrana, except O. anlungensis, O. exiliversabilis, O. hejiangensis, O. kwangwuensis, O. lungshengensis, O. nasuta, O. tiannanensis, O. versabilis [C.-C. Liu and Hu, 1962; Hu et al., 1966; Yang and Li, 1980; Fei, 1999; Fei and Ye, 2001; Fei et al., 2001; Ye and Fei, 2001; see also Bain et al., 2003; Bain and Nguyen, 2004]).

Ye and Fei (2001; fig 45) on the basis of morphology, and Jiang and Zhou (2005; fig. 41), on the basis of DNA sequence evidence have demonstrated that recognition of Eburana renders Odorrana paraphyletic. With a different sampling of species of Eburana and Odorrana, Matsui et al. (2005; fig. 46) provided DNA sequence evidence that nominal Eburana is paraphyletic with respect to at least one member of Odorrana (O. schmackeri) and one species of Chalcorana (C. hosii). On this basis Matsui et al. (2005) considered Eburana to be part of Odorrana (along with Chalcorana hosii).

As noted above, a number of characters suggested by Dubois (1992) to diagnose various taxa have taxonomic distributions to suggest more widespread occurrence. Colorless chest spinules (a putative character of Odorrana) are also present in Huia nasica (B.L. Stuart and Chan-ard, 2005), Nidirana adenopleura, and the holotype of N. caldwelli (R. Bain, personal obs.). The one putative apomorphy of Eburana is character 5 (lacking a pigmented animal pole on the egg) which is known from at least three other genera: Odorrana (see above), Amolops (e.g., A. chunganensis), and Chalcorana (e.g. C. hosii) (Bain et al., 2003; Bain and Nguyen, 2004).

Bain et al. (2003) transferred Rana chloronota (which they thought Dubois, 1992, had in hand as his exemplar of “Rana livida”) from Eburana to Odorrana on the following bases: it has odoriferous skin secretions (implied to be characteristic of Odorrana by way of the formulation of the name by Fei et al., 1991 “1990”); its chromosomes have submetacentric pairs and positions of secondary constrictions more similar (in some cases almost identical) to other species of Odorrana than to other species of Eburana (Li and Wang, 1985; Wei et al., 1993; Matsui et al., 1995); and molecular data (Murphy and Chen, unpublished), although it has unpigmented eggs and lacks pectoral spinules. The implication is that (1) odoriferous skin secretions may be unreported for other Eburana species, or (2) odoriferousness, presence of spinules, and egg color may be homoplastic. We sampled Eburana chloronota and Odorrana grahami. Although this will not allow us to test the monophyly of Eburana or Odorrana, it will help illuminate the extent of the problem.

Fei et al. (2005; fig. 45) have since divided Odorrana (sensu Fei et al., 1991 “1990”) into two subgenera: Bamburana and Odorrana. Bamburana was distinguished from subgenus Odorrana (sensu Fei et al., 2005) by the following characters: dorsolateral folds present (absent in Odorrana), upper lip with sawtooth spinules (absent in Odorrana); xiphisternum without notch (deeply notched in Odorrana); sternum widened posteriorly (sternum not widened posteriorly in Odorrana). Odorrana (Bamburana) versabilis (the type species) and O. (Bamburana) nasuta do not have white spines on the chest of the male, but the other species, O. (Bamburana) exiliversabilis does. According to this diagnosis, Bamburana should also include O. trankieni (Orlov et al., 2003). Nevertheless, Ye and Fei (2001; fig. 45) provided a cladogram based on 29 character transformations of morphology that suggest strongly that Bamburana renders the subgenus Odorrana as paraphyletic. We did not sample any species of nominal Bamburana, but on the basis of the study of Ye and Fei (2001) we can reject its recognition.

Glandirana was coined by Fei et al. (1991 “1990”) as a genus, a position they have maintained consistently (Fei et al., 2005). Nevertheless, Glandirana was placed by Dubois (1992) within subsection Hylarana, where it was diagnosed by Dubois as lacking digital and toe pads, although it retains a lateral groove on the toe tips as found in other groups that do have enlarged digital pads. With the exception of the lateral toe grooves in Glandirana, we are unaware of any morphological character that would prevent assignment of Glandirana to sections Amerana, Pelophylax, or Rana. Jiang and Zhou (2005), on the basis of DNA sequence evidence, placed Glandirana as the sister taxon of Rugosa and together as the sister taxon of a group composed of Amolops, Nidirana, Pelophylax, and Rana (fig. 41). We sampled Glandirana minima.

Subgenus Hylarana is also weakly diagnosed by comparative characters, with the only morphological apomorphies suggested by Dubois (1992) being the low number of rows of labial keratodonts in larvae (shared with Glandirana and sections Amerana, Pelophylax, and Rana; tadpoles unknown in Pterorana and Tylerana). We sampled Hylarana erythraea and H. taipehensis. Matsui et al. (2005; fig. 46) suggested, on the basis of DNA sequence evidence that Hylarana (a member of Dubois', 1992, subsection Hylarana) is imbedded within his subsection Hydrophylax.

Subgenus Tylerana is diagnosed from the remaining Hylarana-like taxa by having a large oval gland on the inner side of the arm in males (Boulenger, 1920; Dubois, 1992). We sampled Tylerana arfaki.

Subgenera Sanguirana, Pterorana, and Nasirana, which we did not study, were reported by Dubois (1992) to have dermal glands on the larvae (unknown in Pterorana), well-developed digital discs, and outer metatarsal tubercles (unknown in Pterorana). Two of the three subgenera, Nasirana and Pterorana, contain single species that have distinctive autapomorphies. Nasirana alticola can be distinguished from other Hylarana-like frogs by the large size of its larvae (shared with Clinotarsus), the ocellated color pattern on the larval tail (larvae of Pterorana and Tylerana unknown), the fleshy prominence on the nose of the adult, and the relatively high 7–9/8–9 keratodont formula (Dubois, 1992), which may suggest that it is a member of one of the cascade-dwelling clades. Similarly, Pterorana khare is distinguished from other ranid frogs by the fleshy folds on the flanks of the adult. Matsui et al. (2005) did not study Sanguirana or Pterorana, but suggested that Nasirana is the sister taxon of a group composed of subsection Hydrophylax and Chalcorana chalconota (nominally part of subsection Hylarana).

Ranixalinae (1 genus, 10 species):

Ranixalinae is another Indian endemic. It contains only Indirana, and is characterized by terrestrial tadpoles with a keratodont formula of 3–5/3–4. Otherwise, it is diagnostically identical to Nyctibatrachinae (Dubois et al., 2001). Dubois (1999a: 89) doubted that Nyctibatrachinae was distinguishable from Ranixalinae and suggested that Blommers-Schlösser's (1993) distinction between Ranixalinae (as Indiraninae), Nyctibatrachinae, and Nannophrys (which Blommers-Schlösser placed in the otherwise African Cacosterninae and Dubois placed in Ranixalinae) might be substantiated by additional evidence.

Van der Meijden (2005; fig. 36), recently placed, weakly, Indirana as the sister taxon of Dicroglossinae on the basis of mtDNA and nuDNA sequence data.

We sampled two species of Indirana (Indirana sp. 1 and Indirana sp. 2).

Rhacophoridae (10 genera, 267 species) and Mantellidae (5 genera, 157 species):

Some authors consider Afro-Asian Rhacophoridae and Madagascan Mantellidae to be families (e.g., Vences and Glaw, 2001; Van der Meijden et al., 2005). Others consider them subfamilies of Ranidae (e.g., J.D. Lynch, 1973; Dubois, 1987 “1985”, 1992; Roelants et al., 2004) or subfamilies of a larger Rhacophoridae (e.g., J.A. Wilkinson and Drewes, 2000; J.A. Wilkinson et al., 2002). Regardless, their taxonomic histories are deeply entwined and we treat them in our discussion as families.

Liem (1970) provided the first character-analysis-based study of phylogeny of the group (including the mantellids in his sense) in which the mantellids were considered basal to the remaining rhacophorids (fig. 47A). Channing (1989) followed with a more rigorous analysis of Old World treefrogs and proposed that Buergeria is the sister taxon of the remaining rhacophorids (including the mantellines; fig. 47B), which he called Buergeriinae and Rhacophorinae, respectively. In his arrangement the mantellids were included as basal members of Rhacophorinae. Ford and Cannatella (1993) noted at least four synapomorphies that distinguish Rhacophoridae + Mantellidae from other ranoids: (1) presence of intercalary elements (presuming that hyperoliids are not the sister taxon); (2) one slip of the m. extensor digitorum communis longus inserts on the distal portion of the fourth metatarsal; (3) outermost slip of the m. palmaris longus inserts on the proximolateral rim of the aponeurosis palmaris; and (4) possession of a bifurcate terminal phalanx. J.A. Wilkinson and Drewes (2000) discussed the analyses by Liem (1970) and reanalysis of these data by Channing (1989) and suggested further analytical refinements but noted considerable instability in the morphological evidence (fig. 47C).

More recent work has suggested that mantellids are the sister taxon of rhacophorids (e.g., Emerson et al., 2000b; Richards et al., 2000; Roelants et al., 2004; Delorme et al., 2005), with this group imbedded within Ranidae. Vences and Glaw (2001) suggested that Mantellidae is composed of three subfamilies: Boophinae (Boophis), Laliostominae (Aglyptodactylus and Laliostoma), and Mantellinae (Mantella and “Mantidactylus”). Vences et al. (2003d) arranged these subfamilies as Boophinae + (Laliostominae + Mantellinae), with “Mantidactylus” deeply paraphyletic with respect to Mantella, and several of the subgenera of “Mantidactylus” paraphyletic or polyphyletic.

J.A. Wilkinson et al. (2002; fig. 48) proposed a phylogeny of rhacophorines, based on mtDNA sequence data. They found mantellines to be the sister taxon of rhacophorines, and that within rhacophorines, that Buergeria is the sister taxon of all others. They also found Chirixalus to be polyphyletic, a problem that was addressed, in part, by the recognition of Kurixalus by Ye, Fei, and Dubois (In Fei, 1999), for “Chirixaluseiffingeri. Some other taxonomic problems were left open by J.A. Wilkinson et al. (2002): the recognition of “Chirixaluspalbebralis, which is isolated phylogenetically from the majority of rhacophorids; the monophyletic grouping of the type species of Chirixalus (Chirixalus doriae) with that of Chiromantis (Chiromantis xerampelina); and the weakly supported sister clade of Chirixalus-Chiromantis of Chirixalus vittatus, with the type species of Polypedates, P. leucomystax.

Delorme et al. (2005) have since proposed a taxonomy of Philautini (Rhacophoridae; fig. 49). Although a tree was provided, the evidence (molecular or morphological) that provided the tree structure was not provided, and inasmuch as phylogenetic propinquity was not the organizing principle of their proposed taxonomy, their taxonomy is not consistent with the phylogeny they proposed. Although reported to be based largely on the same data set as the rhacophorid study of J.A. Wilkinson et al. (2002; 12S and 16S rRNA), the tree proposed by Delorme et al. (2005) also included data from rhodopsin and from morphology (number and content of transformations undisclosed), but Delorme et al. (2005) did not include the tRNAValine gene included by J.A. Wilkinson et al. (2002). Because none of the underlying data were formally provided, methods of alignment and analysis were also not provided. Substantially less resolution is evident in the Delorme et al. (2005) tree (fig. 49) than in the J.A. Wilkinson et al. (2002) tree (fig. 48), although they agree that (1) mantellines are the sister taxon of rhacophorines; (2) Buergeria is the sister taxon of all remaining rhacophorids; (3) Theloderma and Nyctixalus are sister taxa; (4) Chirixalus is paraphyletic with respect to Chiromantis and likely polyphyletic (see points 6 and 7); (5) Rhacophorus may be paraphyletic with respect to a possibly nonmonophyletic Polypedates; (6) a monophyletic unit exists that is composed of Kurixalus eiffingeri and Aquixalus idiootocus and A. verrucosus (the latter two were transferred, respectively, by Delorme et al., 2005, from “Chirixalus” and “Rhacophorus” into an explicitly paraphyletic or polyphyletic Aquixalus, without disclosure of phylogenetic evidence; see comment below); (7) “Chirixaluspalpebralis is demonstrably not in a monophyletic group with remaining Chirixalus.

Delorme et al. (2005) recognized a paraphyletic/polyphyletic Aquixalus containing two nominal subgenera: (1) Aquixalus (paraphyletic/polyphyletic if Aquixalus idiootocus and A. verrucosus are included; if they are excluded from Aquixalus the monophyly of the remaining subgenus Aquixalus remains arguable); (2) Gracixalus (type species: Philautus gracilipes Bourret, 1937) for the “Chirixalusgracilipes group, which they treated as phylogenetically distant from “C.” palpebralis, thereby suggesting that the palpebralis group of Fei (2001), composed, in Fei's usage, of Philautus palpebralis, P. gracilipes, P. medogensis, P. ocellatus, and P. romeri, is nonmonophyletic. Nevertheless, because J.A. Wilkinson et al. (2002) and Delorme et al. (2005) presumably had so much underlying evidence in common, the fact of their substantial topological differences between their results is surprising, although many of the internal branches of the J.A. Wilkinson et al. (2002) tree are weakly supported and possibly could be modified by the undisclosed rhodopsin and morphology data of Delorme (2005). Nevertheless, a tree without associated evidence (that of Delorme et al., 2005) cannot test a tree that has evidence attached to it (the tree of J.A. Wilkinson et al., 2002).

Because Delorme et al. (2005; fig. 49) do not accept (apparently) phylogenetic propinquity as the organizing principle in taxonomy, they (1) created a new paraphyletic genus, Aquixalus (including Chirixalus idiootocus and Rhacophorus verrucosus, which they simultaneously figured to be closer evolutionarily to Kurixalus eiffingeri than to other members of their Aquixalus), (2) retained a nonmonophyletic Chirixalus (with respect to Chiromantis and “Chirixaluspalpebralis), and (3) recognized Philautini (Philautus + Theloderma + Nyctixalus + “Aquixalus”), for which the predominance of their own evidence, as demonstrated by their tree, does not reject paraphyly. In particular, it is not clear why these authors transferred Chirixalus idiootocus into a paraphyletic “Aquixalus”, so for our overall discussion, we will not follow the transfer of “Chirixalusidiootocus into a paraphyletic/polyphyletic “Aquixalus”, because this taxonomic change disagrees with the phylogenetic tree (albeit, data free) proposed in the same publication.

In our analysis we sampled Boophinae (Boophis albilabris, B. tephraeomystax); Laliostominae (Aglyptodactylus madagascariensis, Laliostoma labrosum); Mantellinae (Mantella aurantiaca, M. nigricans, Mantidactylus cf. femoralis, M. peraccae); Buergeriinae (Buergeria japonica); Rhacophorinae (“Aquixalus” (Gracixalus) gracilipes [formerly in Chirixalus or Philautus], “Chirixalusidiootocus, Chirixalus doriae, C. vittatus, Chiromantis xerampelina, Kurixalus eiffingeri, Nyctixalus pictus, N. spinosus, Philautus rhododiscus, Polypedates cruciger, P. leucomystax, Rhacophorus annamensis, R. bipunctatus, R. calcaneus, R. orlovi, and Theloderma corticale).

RESULTS

Sequence Length Variation and Notes on Analysis

Length variation among the four nuclear protein coding genes was minimal. Following trimming of primers, all histone H3-complete products were 328 bp, and all SIA-complete products were 397 bp. All but one of the rhodopsin-complete products were 316 bp; the sequence for Alytes obstetricans was 315 bp, as was the sequence of this species deposited previously on GenBank (AY364385). Most tyrosinase products were 532 bp, exceptions being Xenophrys major and Ophryophryne hansi, which were 538 bp. Tyrosinase was by far the most difficult fragment to amplify (tyrosinase sequences were sampled for only 38% of the terminals), and this difficulty impedes understanding of the significance of this length variation. The “closest” taxa for which we were able to obtain sequences for this locus were Xenopus laevis (from GenBank AY341764) and Hemisus marmoratus (both of which are 532 bp), so it is unclear whether the greater length of this tyrosinase fragment is characteristic of some megophryids or a more inclusive clade. The homologous tyrosinase sequence for Petropedetes parkeri downloaded from GenBank (AY341757) was 535 bp. As with the megophryids, the generality of this length is unclear. However, the length of Arthroleptides sp. is 532, so it is likely that the increased length is restricted to some or all species of Petropedetes.

Length variation was much more extensive and taxonomically widespread in the ribosomal loci. Among complete H1 sequences, the shortest length of 2269 bp was found in Afrana fuscigula. The longest sequence was that of the outgroup terminal Latimeria chalumnae (2530 bp), followed by Ptychadena mascareniensis (2494 bp) and Silurana tropicalis (2477 bp). Length variation was too extensive for clear phylogenetic patterns to emerge. However, although extensive variation in the length of the 28S sequences occurred even among closely related species (e.g., 744 bp in Schoutedenella schubotzi and 762 bp in S. xenodactyloides), numerous clades may be characterized by their 28S length. For example, of the 20 salamander 28S fragments with no missing data, all had a length of 694 bp, except Pseudoeurycea conanti and Desmognathus quadramaculatus, which were 695 bp. The only other species of 694 bp in this study were the two turtles (Pelomedusa subrufa and Chelydra serpentina) and the pelodryadine frog, Nyctimystes dayi. Length variation in 28S is greater among caecilians (683–727 bp), but it is still more restricted than in anurans (685–830 bp).

Among the sampled anurans, this 28S fragment is > 700 bp in all but six species (appendix 3). Mantella nigricans and M. aurantiaca differ from all other taxa in that their 28S sequence is 685 bp (28S sequences were not generated for Mantidactylus, but they were for Laliostoma, Aglyptodactylus, and numerous rhacophorids, which have 28S sequences of 709–712 bp). As mentioned earlier, the 28S sequence of Nyctimystes dayi is 694 bp, and that of the related Litoria genimaculata is 690. The remaining outliers are Bufo punctatus (700 bp) and Microhyla sp. (698 bp) which differ from close relatives by > 50 bp and > 25 bp, respectively. Ascaphus truei, Leiopelma archeyi, and L. hochstetteri are all 703 bp, as are the included species of Pelodytes and Spea. Similarly, Alytes and Discoglossus are the only sampled species with a 28S fragment of 706 bp.

Although these variations in length do not provide evidence of phylogeny independent of the underlying indel and nucleotide transformation events, their phylogenetic conservativeness makes them useful diagnostic tools, and we therefore note 28S sequence length, where relevant, in the taxonomic sections that follow.

Parsimony analysis by POY of the combined data set resulted in a single most parsimonious solution of 127019 steps. Although optimizing the implied alignment on the topology found in POY verified the length reported in POY, ratcheting of the implied alignment in NONA spawned from Winclada resulted in four most parsimonious trees of length 127,017 steps, and these are our preferred hypotheses. The only differences between the POY and NONA solutions involve the placement of (1) Glandirana and (2) Brachytarsophrys feae. This conflict is also seen among the four 127017-step trees, resulting in the polytomies seen in the strict consensus (fig. 50 [provided as a multipage insert]).

Topological Results and Discussion

A consensus of the four equally most parsimonious trees is shown in figure 50 (insert). Most clades are highly corroborated by molecular evidence (and in some places by morphological evidence). Although only an imperfect surrogate for a measure of support (something that so far eludes us), the Bremer (= decay index) and jackknife values all speak to a highly corroborated tree. (See appendix 4 for branch length, Bremer support, and jackknife values.) Because this study rests on the largest amount of data ever applied to the problem of the relationships among amphibians, we think that the obtained tree is a step forward in the understanding of the evolutionary history of amphibians. We do, of course, have reservations about parts of the overall tree. But, upon reflection, we realized that most of the parts of the tree that concerned us were those that (1) we considered insufficiently sampled relative to known species and morphological diversity (e.g., Bufonidae); or (2) are groups for which no other evidence-based suggestions of phylogeny had ever been provided (e.g., parts of traditionally recognized Ranoidea). Nevertheless, familiarity has much to do with notions of plausibility, the root of the problem of social conservatism in amphibian systematics.

We discuss results under two headings and with reference to several different figures. The primary focus in this first section, “Results”, is to address issues of relationship among, and monophyly of, major groups (nominal families and subfamilies and nomenclaturally unregulated taxa). We also make general taxonomic recommendations in this section. Under the second heading, “Taxonomy”, we discuss further results and various taxonomic issues under the appropriate taxonomic category. Bremer and jackknife values are reported for each branch in figure 50 (insert; as well as in other figures, where relevant) but are otherwise only occasionally mentioned in text.

The general tree shown in figure 50 (insert), with 532 terminals, is obviously too complex and detailed for easy discussion, so we will refer to subtrees in different figures. Relevant taxa (branches) have the molecular data summarized by name and/or number in appendix 4. We first discuss the results relative to the Review of Current Taxonomy at or above the nominal family-group level, with reference to families that appear to be monophyletic and those that are paraphyletic and polyphyletic. In the case of paraphyly and polyphyly we offer remedies in this section that are paralleled in more detail in the Taxonomy section, where we propose a monophyletic taxonomy for all but a few problematic amphibian groups and discuss aspects of our results that are relevant to the systematics of that particular group, such as monophyly of nominal genera and various taxonomic remedies to problems that our results highlighted.

Outgroup Relationships

In our results, Latimeria is outside of the tetrapod clade, and amniotes form the sister taxon of amphibians. This topology was conventional, at least for paleontologists and morphologists (e.g., Gauthier et al., 1988a, 1988b; fig. 2A). Within Amniota, we found turtles to be the sister taxon of diapsids (archosaurs + lepidosaurs) and this inclusive group to be the sister taxon of mammals. Our molecular data do not support the suggestion by Rieppel and de Braga (1996), based on morphology, that turtles are more closely related to lepidosaurs than to archosaurs. Our molecular results disagree with the results of Mannen and Li (1999), Hedges and Poling (1999), and Iwabe et al. (2005), in which turtles were found to be closely related to archosaurs, with lepidosaurs, and mammals as successively more distant relations. An analysis of why our molecular results are congruent with the conventional tree of morphology (fig. 2A) and not with previous molecular results is largely outside the scope of this paper. Nevertheless, our analysis was a parsimony analysis, as were the studies of Gauthier et al. (1988a; 1988b). The molecular study of Hedges and Poling (1999) rested on a large amount of DNA evidence (ca. 5.2kb), but their alignment was made under a different set of evolutionary assumptions from that used in their phylogenetic analysis. A stronger test of amniote relationships will be made by combining morphology and all available DNA evidence and analyzing these data under a common set of assumptions.

Amphibia (Lissamphibia) and Batrachia

Our results (figs. 50 [insert], 51) corroborate the monophyly of amphibians (Lissamphibia of Parsons and Williams, 1963; Amphibia of Cannatella and Hillis, 1993) with reference to other living taxa, although our data obviously cannot shed any light on the placement of the lissamphibians among fossil groups. We also found the three groups of lissamphibians to be strongly supported (fig. 50 [insert], branches 7, 24, 74). Furthermore, our DNA sequence data indicate that the caecilians are the sister taxon of the clade composed of frogs plus salamanders (Batrachia; fig. 50 [insert], branch 23), the topology preferred by Trueb and Cloutier (1991). Our data reject (1) that living amphibians are paraphyletic with respect to Amniota (Carroll and Currie, 1975; J.S. Anderson, 2001); (2) that salamanders are paraphyletic with respect to caecilians (Laurin, 1998a, 1998b, 1998c); and (3) the hypothesis, based on smaller amounts of evidence, that caecilians and salamanders are closest relatives (Feller and Hedges, 1998). Our data suggest strongly that the arrangement favored by morphologists (e.g., Trueb and Cloutier, 1991; Iordansky, 1996; Zardoya and Meyer, 2000, 2001; Schoch and Milner, 2004) is also the arrangement favored by the preponderance of the molecular evidence (e.g., San Mauro et al., 2005), that living amphibians form a monophyletic group with respect to Amniota, and that frogs and salamanders are more closely related to each other than either is to the caecilians (contra Feller and Hedges, 1998). The effect of including fossils and a much more complete morphological data set are not known, but we note that our molecular data are consistent with the preponderance of morphological data so far published.

Salamanders (Caudata) and frogs (Anura) are each also monophyletic, a result that will surprise no one, even though the morphological evidence for monophyly of the salamanders, in particular, is weak (Larson and Dimmick, 1993).

Gymnophiona

In general form our cladogram (fig. 50 [insert], fig. 52) agrees with the conventional view of caecilian relationships (fig. 3). Like Nussbaum (1977, 1979) and later authors (e.g., Duellman and Trueb, 1986; San Mauro et al., 2004; San Mauro et al., 2005) we find that Rhinatrematidae is the monophyletic sister taxon of the remaining caecilians. This placement appears well-corroborated on both morphological and molecular grounds.

Ichthyophiidae is paraphyletic with respect to Uraeotyphlidae (this being highly corroborated by our molecular data), and can be restated as Ichthyophis is paraphyletic with respect to Uraeotyphlus. This outcome was arrived at previously by Gower et al. (2002). There is a single morphological character, angulate annuli anteriorly, that supports the monophyly of the ichthyophiids (sensu stricto, excluding Uraeotyphlus), but the amount of molecular evidence in support of Uraeotyphlus being nested within Ichthyophis indicates that this character was either reversed in Uraeotyphlus or independently derived in different lineages of “Ichthyophis”. Under these circumstances, Uraeotyphlus must be transferred to Ichthyophiidae, and although treatment of “Ichthyophis” is beyond the scope of this study, we expect subsequent work (denser sampling of ichthyophiids and addition of new data) to delimit the nature of this paraphyly and reformulate infrafamilial taxonomy. The effect of this change is minimal, because Uraeotyphlidae contains a single genus, and no hierarchical information is lost by placing Uraeotyphlidae in the synonymy of Ichthyophiidae.

As expected from previously published DNA sequence (M. Wilkinson et al., 2003) and morphological evidence (M.H. Wake, 1993; M. Wilkinson, 1997), we found Scolecomorphidae to be imbedded within Caeciliidae. The evidence for this is strong (appendix 4, branches 12, 14, 16), and we therefore consider Scolecomorphidae to be a subsidiary taxon (Scolecomorphinae) within Caeciliidae. Similarly, Typhlonectidae is deeply imbedded within Caeciliidae, a result previously noted (M.H. Wake, 1977; Nussbaum, 1979; M. Wilkinson, 1991; Hedges et al., 1993). Typhlonectidae is here regarded as a subsidiary taxon (as Typhlonectinae) within a monophyletic Caeciliidae, although the genera of the former “Caeciliinae” remain incertae sedis within the Caeciliidae.

Our results differ slightly from those presented by M. Wilkinson et al. (2003), which were based on a smaller amount of sequence data (mt rRNA only). Like us, M. Wilkinson et al. (2003) found Scolecomorphidae and Typhlonectidae to be imbedded within “Caeciliidae”, although in a different and less strongly corroborated placement. Our placement of Siphonops (South America) as the sister taxon of Hypogeophis (Seychelles) and together the sister taxon of Gegeneophis (India), is the only unanticipated result. In light of the strong support it received in our analysis, this conclusion deserves to be evaluated carefully.

Caudata

Among previously published cladograms our results (fig. 53) most resemble the tree of salamander families suggested by Gao and Shubin (2001; fig. 5) and diverge slightly from the results presented by Larson and Dimmick (1993; fig. 4) and Wiens et al. (2005; fig. 7) in placing sirenids (which lack spermatophore-producing organs) as the sister taxon of Proteidae (which, like other salamandroid salamanders has spermatophore-producing organs), rather than placing the sirenids as the sister taxon of all other salamander families. (The Bayesian analysis of Wiens et al., 2005, however, placed cryptobranchoids as the sister taxon of remaining salamanders, suggesting that there is internal conflict within their data set.) Other recent results found, on the basis of RAG-1 DNA sequence evidence (Roelants and Bossuyt, 2005; San Mauro et al., 2005), and on the basis of RAG-1, nuRNA, and morphology (Wiens et al., 2005), Sirenidae to be the sister taxon of remaining salamanders, the traditional arrangement. Because our molecular evidence did not overlap with theirs, and with the arguable example of Wiens et al. (2005), their amount of evidence is smaller than ours, these results require additional testing. Our results do not reject the monophyly of any of the nominal families of salamanders, a result that is consistent with previous studies. Except as noted later, the remaining results are conventional.

Hynobiidae and Cryptobranchidae:

Unlike the results of Larson and Dimmick (1993; fig. 4), San Mauro et al. (2005; fig. 17), Roelants and Bossuyt (2005; fig. 16), and Wiens et al. (2005; fig. 7) our results place these taxa as the sister taxon of all other salamanders, and not as the sister taxon of all salamanders excluding sirenids (the relationship recovered by Larson and Dimmick, 1993, San Mauro et al., 2005, and Roelants and Bossuyt, 2005). The monophyly of hynobiids plus cryptobranchids is not controversial, nor is that of Cryptobranchidae. In the case of Hynobiidae, as noted in the taxonomic review, our sampling is insufficient to address any of the generic controversies (summarized by Larson et al., 2003: 43–45) and is only a minimal test of the monophyly of Hynobiidae.

Sirenidae and Proteidae:

Unlike Larson and Dimmick (1993) and more recent morphological and molecular studies (Roelants and Bossuyt, 2005; San Mauro et al., 2005; Wiens et al., 2005), but like Gao and Shubin (2001; fig. 5), we recovered Sirenidae not as the sister taxon of all other salamanders but as the sister taxon of Proteidae. Our highly corroborated results and the results of Gao and Shubin (2001) suggest that the perennibranch characteristics of Proteidae and Sirenidae are homologous. On this topology the cloacal apparatus for spermatophore formation is a synapomorphy at the level of all salamanders, excluding Cryptobranchidae and Hynobiidae, with a loss in Sirenidae. Alternatively, it is a convergent development in Proteidae and in the ancestor of Salamandridae, Rhyacotritonidae, Dicamptodontidae, Plethodontidae, Amphiumidae, and Ambystomatidae. The effect of combining the morphological data presented by Wiens et al. (2005) with all of their and our molecular data remains an open question, although we note that their morphological-only data set produced a result in which Sirenidae + Proteidae form a monophyletic group. Thus, it is not clear that this is a simple morphology-versus-molecules issue. Rather than oversimplify and misrepresent that paper, we leave the question open as to what the result will be when all molecular and morphological data are combined.

As noted earlier, our results reject a monophyletic Salamandroidea (all salamanders, excluding Cryptobranchidae, Hynobiidae, and Sirenidae). This taxon was diagnosed by internal fertilization through the production of spermatophores (produced by a complex system of cloacal glands) and having angular and prearticular bones fused (also found in Sirenidae). The hypothesis that sirenids and proteids form a taxonomic group is quite old: It was first suggested by Rafinesque (1815; as Meantia; see the discussion in appendix 6).

Rhyacotritonidae and Amphiumidae:

We resolved the polytomy found in the tree of Gao and Shubin (2001) of Plethodontidae, Rhyacotritonidae, and Amphiumidae into Rhyacotritonidae + (Amphiumidae + Plethodontidae), a conclusion also of Wiens et al (2005). Although we did not test the monophyly of either Rhyacotriton or Amphiuma, in neither case is this seriously in question. As noted earlier, the position of Amphiuma with respect to plethodontids is conventional (Larson, 1991; Larson and Dimmick, 1993).

Plethodontidae:

Our tree differs trenchantly from those of authors prior to 2004 (e.g., D.B. Wake, 1966; Lombard and Wake, 1986), but is similar in general form to those of Mueller et al. (2004) on the basis of complete mtDNA genomes, Macey's (2005) reanalysis of those data, and the tree of Chippindale et al. (2004), based on 123 characters of morphology and about 2.9 kb of mtDNA and nuDNA. In those studies and in ours Amphiumidae and Rhyacotritonidae were obtained as successively more distant outgroups of Plethodontidae. In the three previous studies (Chippindale et al., 2004; Mueller et al., 2004; Macey, 2005) as well as in ours, the desmognathines are in a clade with the plethodontines (Ensatina, and Plethodon). Our data (as well as those of Mueller et al., 2004, and Macey, 2005) also found Hydromantes and Speleomantes to be in this plethodontine clade, not with “other” bolitoglossines.

In our results, as well as those of Mueller et al. (2004) and Chippindale et al. (2004), all other plethodontids (the old Hemidactyliinae and Bolitoglossini) are placed in a group that forms the sister taxon of the first group. The evidence for these groupings is strong (appendix 4; fig. 53). The placement of Hydromantes and Speleomantes in the first group by our data is strongly corroborated, being placed within the desmognathines (a result that runs counter to the morphological evidence as presented by Schwenk and Wake, 1993). Mueller et al. (2004) obtained Hydromantes (including Speleomantes) in the same general group as we did, but placed as the sister taxon of Aneides. In the details of placement of Batrachoseps, Hemidactylium, and our few overlapping bolitoglossine genera, we differ mildly. Our differences from the tree of Macey (2005) are difficult to explain. The amount of evidence marshalled by Macey (the same aligned data set as Mueller et al., 2004), is on the order of 14kb of aligned mtDNA sequence. Our mtDNA set is a subset of that, but analyzed differently, particularly with respect to alignment. Alignment of the data set of Mueller et al. (2004) was done with different transformation costs than used in analysis, and this alignment was accepted for reanalysis by Macey (2005). Further, a number of our exemplars (i.e., Plethodon dunni, P. jordani, Desmognathus quadramaculatus, Phaeognathus, Hydromantes platycephalus, Eurycea wilderae, Gyrinophilus porphyriticus, Thorius sp., Bolitoglossa rufescens, and Pseudoeurycea conanti) are represented in our analysis by sequences that are not part of the mtDNA genome. Although we provisionally accept the results of Macey (2005; fig. 10) as based on a much larger amount of data than our results, it may be that the single biggest cause of different results between our analysis and his is the method of alignment. One will know only when that data set is analyzed using direct optimization.

Chippindale et al. (2004; fig. 11) suggested a taxonomy, consistent with their tree, for Plethodontidae. Plethodontinae in their sense corresponds to the group composed of the former Desmognathinae and former Plethodontini. Within the second group composed of hemidactyliines and bolitoglossines they recognized Hemidactyliinae (Hemidactylium), Spelerpinae Cope, 1859 (Eurycea [sensu lato], Gyrinophilus, Stereochilus, and Pseudotriton), and Bolitoglossinae (for all of the bolitoglossine genera studied). Macey (2005) came to the same taxonomy, but placed Hemidactyliinae as the sister taxon of remaining plethodontids, the relative position of the other groups remaining the same. He also placed Hydromantes (including Speleomantes) in Plethodontinae. These two genera had previously been associated with Bolitoglossini (D.B. Wake, 1966; Elias and Wake, 1983).

Our results regarding placement of Hydromantes and Speleomantes imply either that the morphological synapomorphies of the Desmognathinae, mostly manifestations of the bizarre method of jaw opening in which the lower jaw is held in a fixed position by ligaments extending to the atlas–axis complex, are reversed in the hydromantine clade or that this peculiar morphology is convergent in Desmognathus and Phaeognathus.

Previous to the study of Mueller et al. (2004), who found Plethodon to be monophyletic on the basis of analysis of mtDNA sequence data, all published evidence pointed to paraphyly of Plethodon with respect to Aneides (e.g., Larson et al., 1981; Mahoney, 2001). Our analysis of a variety of DNA sequence data suggests also that the eastern and western components of Plethodon do not have a close relationship, being united solely by symplesiomorphy. Had it not been for the appearance of the recent paper by Chippindale et al. (2004), we would have erected a new generic name for western Plethodon (for which no name is currently available). But, the denser sampling of plethodons and different selection of genes in the Chippindale et al. (2004) paper suggests that a study including all of the available data and a denser sampling is required before making any taxonomic novelties.

We recovered former Bolitoglossini as polyphyletic, with the traditional three main components (supergenera Batrachoseps, Hydromantes, and Bolitoglossa; D.B. Wake, 1966) being found to have little in common with each other. Our tree of bolitoglossines (sensu stricto) is not strongly corroborated. Nevertheless, that the three groups of bolitoglossines should be recovered as polyphyletic is not shocking inasmuch as the amount of evidence that traditionally held them together was small.

Salamandridae:

Our results largely correspond to those of Titus and Larson (1995) and especially with those presented by Larson et al. (2003). Our tree differs from the topology suggested by Larson et al. (2003), which was based on more extensive taxon sampling but less DNA evidence, in that we get additional resolution of the group Neurergus + (Triturus + Euproctus), where in the tree provided by Larson et al. (2003) these taxa are in a polytomy below the level of Paramesotriton + Pachytriton.

Dicamptodontidae and Ambystomatidae:

Dicamptodon is recovered as the sister taxon of Ambystomatidae, the same phylogenetic arrangement found by previous authors (Sever, 1992; Larson and Dimmick, 1993; Wiens et al., 2005). The monophyly of Dicamptodon was only minimally tested, although Dicamptodon monophyly is not seriously in doubt (Good and Wake, 1992). Inasmuch as Dicamptodontidae was recognized on the basis of its hypothesized phylogenetic distance from Ambystomatidae (Edwards, 1976), a hypothesis now rejected, we propose the synonymy of Dicamptodontidae with Ambystomatidae, which removes the redundancy of having two family-group names, each containing a single genus. The reformulated Ambystomatidae contains two sister genera, Dicamptodon and Ambystoma.

Ambystomatidae was found to be monophyletic, at least with reference to our exemplar taxa, and the sister taxon of former Dicamptodontidae. Although we have not severely tested the monophyly of Ambystoma, others have done so (e.g., Shaffer et al., 1991; Larson et al., 2003), and its monophyly is well corroborated.

Anura

As mentioned earlier and in the taxonomic review, the amount of morphological and DNA sequence evidence supporting the monophyly of Anura is overwhelming. We think that our data make a strong case for a new understanding of frog phylogeny. Even though most of our results are conventional with respect to understanding of frog phylogenetics, our purpose is not to conceal this understanding, but to bring the taxonomy of frogs into line with their phylogenetic relationships. For discussion we adopt the Ford and Cannatella (1993) tree (fig. 14) as the traditional view of phylogeny (although not of nomenclature). We first discuss the non-neobatrachian frogs (fig. 54).

Ascaphidae and Leiopelmatidae:

Ascaphidae and Leiopelmatidae are recovered in our analysis as parts of a monophyletic group, mirroring the results of Green et al. (1989), Báez and Basso (1996), and more recent authors (Roelants and Bossuyt, 2005; San Mauro et al., 2005). The paraphyly of this grouping, as suggested by Ford and Cannatella (1993), is rejected. If our results are accurate, the five morphological synapomorphies suggested by Ford and Cannatella (1993) of Leiopelma plus all frogs excluding Ascaphus must be convergences or synapomorphies of all living frogs that were lost in Ascaphus. Nevertheless, the hypothesis of Ford and Cannatella (1993) was based largely on the unpublished dissertation of Cannatella (1985; cited by Ford and Cannatella, 1993), who rooted his analysis of primitive frogs on Ascaphus on the basis of two plesiomorphic characters found among frogs uniquely in Ascaphus: (1) facial nerve passes through the anterior acoustic foramen and into the auditory capsule while still fused to the auditory nerve; (2) salamander-type jaw articulation in which there is a true basal articulation. All other characters placing Leiopelma as more closely related to all non-Ascaphus frogs were optimized by this assumption, requiring their polarity to be verified. Furthermore, the support for the Ascaphus + Leiopelma branch is very high (Bremer = 41, jackknife = 100%), so it is unlikely that five morphological characters (of which three have not been rigorously polarized) can reverse this. Placing Ascaphus and Leiopelma as sister taxa allows some characters to be explained more efficiently. Thus, the absence of the columella in these two taxa can be seen to be a synapomorphic loss. Ritland's (1955) suggestion that the m. caudalipuboischiotibialis in Leiopelma and Ascaphus may not be homologous with the tail-wagging muscles of salamanders, and the more traditional view of homology with these muscles are both consistent with our results. To remove the redundancy of the family-group names with the two genera (Ascaphus and Leiopelma), we assign Ascaphus to Leiopelmatidae (as did San Mauro et al., 2005). Roelants and Bossuyt (2005) retained Ascaphidae and Leiopelmatidae as separate families and resurrected the name Amphicoela Noble, 1931, for this taxon. Amphicoela is redundant with Leiopelmatidae (sensu lato) when Ascaphidae and Leiopelmatidae are regarded as synonymous, as we do.

Pipidae and Rhinophrynidae:

We found, as did Haas (2003) and San Mauro et al. (2005), and as was suggested even earlier by Orton (1953, 1957), Sokol (1975), and Maglia et al. (2001) that Rhinophrynidae + Pipidae is the sister taxon of all non-leiopelmatid frogs. This result is strongly supported by our evidence (fig. 54; appendix 4, branches 77, 78, 84). Recent suggestions had alternatively placed Pipoidea as the sister taxon of Pelobatoidea (Ford and Cannatella, 1993; their Mesobatrachia) or as the sister taxon of all other frogs (Maglia et al., 2001; Pugener et al., 2003). All three of these arrangements are supported by morphological characters, although Haas' arrangement is more highly corroborated. Haas (2003) suggested nine apomorphies that exclude Pipoidea and Ascaphidae from a clade composed of all other frogs. Pugener et al. (2003) suggested three synapomorphies for all frogs excluding pipoids. (This statement is based on examination of their figure 12; they provided no comprehensive list of synapomorphies.) Ford and Cannatella (1993) suggested that four characters support Mesobatrachia: (1) closure of the frontoparietal fontanelle by juxtaposition of the frontoparietal bones (not in Pelodytes or Spea); (2) partial closure of the hyoglossal sinus by the ceratohyals; (3) absence of the taenia tecti medialis; and (4) absence of the taenia tecti transversum. However, on the basis of Haas' (2003) morphological data alone, these characters are rejected as synapomorphies. However, the mtDNA molecular results presented by García-París et al. (2003) support the recognition of Mesobatrachia (Pelobatoidea + Pipoidea). Nevertheless, these authors included only three non-pipoid, non-pelobatoid genera (Ascaphus, Discoglossus, and Rana) as outgroups, which did not provide a strong test of mesobatrachian monophyly. Placement of Pipoidea as the sister taxon of all other non-leiopelmatid frogs requires rejection of Discoglossanura, Bombinatanura, and Mesobatrachia of Ford and Cannatella (1993), a rejection that is strongly supported by our study.

In our analysis, as well as in all recent ones (Ford and Cannatella, 1993; Báez and Pugener, 2003; Haas, 2003), Pipoidea (Rhinophrynidae + Pipidae) is monophyletic, as are the component families. A novel arrangement in our tree is Hymenochirus being placed as the sister taxon of Pipa + (Silurana + Xenopus). This result differs from the cladograms of Cannatella and Trueb (1988), de Sá and Hillis (1990), Báez and Pugener (2003), and Roelants and Bossuyt (2005; fig. 16). Although our results are highly corroborated by our data, a more complete test would involve the simultaneous analysis of all of the sequence data with the morphological data of all relevant living and fossil taxa. As noted in figure 55, the rooting point of the pipid network appears to be more important to the estimates of phylogeny than differences among networks.

The placement of Pipidae + Rhinophrynidae as the sister taxon of all frogs, save Leiopelmatidae + Ascaphidae, suggests strongly that the fusion of the facial and trigeminal ganglia (Sokol, 1977) found in pelobatoids, pipoids, and neobatrachians, but not in Discoglossidae and Bombinatoridae is homoplastic. Similarly, the absence of free ribs in the adults of pelobatoids, neobatrachians, and pipoids, but their presence in Leiopelma, Ascaphus, and Discoglossidae, requires either independent losses in pipoids and pelobatoids + neobatrachians, or an independent gain in discoglossids + bombinatorids. Roelants and Bossuyt (2005) noted fossil evidence that would support the independent loss in pipoids and Acosmanura (Pelobatoidea + Neobatrachia).

Discoglossidae and Bombinatoridae:

Ford and Cannatella (1993) partitioned the former Discoglossidae (sensu lato) into Discoglossidae (sensu stricto) and Bombinatoridae because their evidence suggested that former Discoglossidae was paraphyletic, with Bombinatoridae and Discoglossidae forming a graded series between the Ascaphidae and Leiopelmatidae on one hand, and all other frogs on the other hand. As noted in the taxonomic review, this partition was based on two characters shared by discoglossines and all higher frogs and absent in the bombinatorines. Haas (2003) rejected this topology with six character transformations supporting the monophyly of Bombinatoridae and Discoglossidae. In addition to Haas' characters, we have strong molecular evidence in support of the monophyly of this taxon (Discoglossidae + Bombinatoridae), as well as the subsidiary families.

Unlike Haas (2003), but like recent molecular studies (Roelants and Bossuyt, 2005; San Mauro et al., 2005), we did not recover Alytes as the sister taxon of the remaining discoglossines and bombinatorines. We included Haas' six characters supporting that topology in our analysis, and the taxon sampling for this part of the tree is nearly identical in the two studies, so it appears that molecular evidence in support of a topology of Alytes + Discoglossus is decisive. The only rationale for considering Discoglossidae and Bombinatoridae as separate families rested on the assertion of paraphyly of the group (Ford and Cannatella, 1993), a position now rejected. Nevertheless, we retain the two-family arrangement because this reflects the state of the literature and is consistent with recovered phylogeny.

Pelobatoidea:

Haas (2003) did not recover Pelobatoidea (Megophryidae, Pelobatidae, Pelodytidae, Scaphiopodidae) as monophyletic. Although we included his morphological data in our analysis, we find Pelobatoidea to be highly corroborated, which suggests very interesting convergences in tadpole morphology. García-París et al. (2003; fig. 18) also found Pelobatoidea to be monophyletic, on the basis of their DNA evidence, and suggested a topology of Scaphiopodidae + (Pelodytidae + (Megophryidae + Pelobatidae)), with relatively low Bremer values on the branch tying Scaphiopodidae to the remaining taxa. In our results we recover all of these family-group units as monophyletic and highly supported. But, our data show strongly a relationship of (Pelodytidae + Scaphiopodidae) + (Pelobatidae + Megophryidae) (fig. 54).

Neobatrachia:

As in all previous studies, we found Neobatrachia to be highly corroborated by many transformations (figs. 50 [insert], 56, 58, 59, 60). What is particularly notable in the broad structure of Neobatrachia is the dismemberment of Leptodactylidae and Hylidae as traditionally formulated, as well as the placement of Heleophrynidae outside of the two major monophyletic components, for our purposes referred to here as (1) Hyloidea, excluding Heleophrynidae and (2) Ranoidea.

Heleophrynidae:

Haas (2003) suggested that Heleophryne may be related to Pelobatoidea, a suggestion that is not borne out by our simultaneous analysis of Haas' data and our molecular data. Earlier authors (e.g., J.D. Lynch, 1973) addressed the phylogenetic position of Heleophryne and associated it with Limnodynastidae on the basis of overall similarity, or with Limnodynastidae + Myobatrachidae on the basis of DNA sequence data (Biju and Bossuyt, 2003). But recently San Mauro et al. (2005) suggested, on the basis of DNA sequence evidence, that Heleophrynidae is the sister taxon of remaining Neobatrachia. We obtained the same placement of Heleophrynidae as did San Mauro et al. (2005).

Hyloidea, excluding Heleophrynidae:

Hyloidea, as traditionally composed, consists of all arciferal groups of neobatrachians and was expected (on the basis of absence of morphological evidence) to be broadly paraphyletic with respect to Ranoidea, or firmisternal frogs (Microhylidae, Ranidae, and their satellites, Mantellidae, Rhacophoridae, Hyperoliidae, Arthroleptidae, Astylosternidae, and Hemisotidae), or monophyletic on the basis of molecular data (Ruvinsky and Maxson, 1996; Feller and Hedges, 1998; Faivovich et al., 2005; San Mauro et al., 2005). In our results, Hyloidea is only narrowly paraphyletic, with the bulk of the hyloids forming the sister taxon of ranoids and only Heleophrynidae outside of this large clade (a conclusion also reached by San Mauro et al., 2005). Within the restricted (non-heleophrynid) Hyloidea, a unit composed of Sooglossidae and the newly discovered Nasikabatrachidae forms the sister taxon of the remaining hyloids (cf. Biju and Bossuyt, 2003; San Mauro et al., 2005). For the most part, the traditional family-group units within Hyloidea were found to be monophyletic, the exceptions being predictable from preexisting literature: Leptodactylidae was found to be composed of several only distantly related groups, and Hylidae (in the sense of including Hemphractinae) was confirmed to be paraphyletic or polyphyletic (see below).

Sooglossidae and Nasikabatrachidae:

The South Indian Nasikabatrachus and the Seychellean sooglossids form an ancient taxon united by considerable amounts of molecular evidence (fig. 56). Biju and Bossuyt (2003) placed Nasikabatrachus as the sister taxon of the sooglossids and our results corroborate this. We are unaware of any historical (in the sense of history of systematics) or other reason to regard Nasikabatrachus as being in a family distinct from Sooglossidae, and on the basis of the molecular evidence we consider Nasikabatrachus to be the sole known mainland member of Sooglossidae. The antiquity of this united group is evident in its placement as the sister taxon of all other non-heleophrynid hyloids. Its phylogenetic position as well as its presence both in India and in the Seychelles suggests that the taxon existed before the final breakup of Pangaea in the late Mesozoic.

Myobatrachidae, Limnodynastidae, and Rheobatrachidae:

Because of the absence of morphological synapomorphies uniting the Australo-Papuan groups Myobatrachidae, Limnodynastidae, and Rheobatrachidae (in our usage), and because of the suggestion of a special relationship between Myobatrachidae and Sooglossidae and between Limnodynastidae and Heleophrynidae (J.D. Lynch, 1973), we were surprised that the preponderance of evidence corroborates a monophyletic Myobatrachidae + Limnodynastidae + Rheobatrachidae (fig. 56). Nevertheless, there is only one morphological character involved in these alternatives (condition of the cricoid ring: complete or incomplete), so, in retrospect, our surprise was unwarranted.

With respect to Myobatrachidae (sensu stricto; Myobatrachinae of other authors), our results are largely congruent with those of Read et al. (2001). The positions of Metacrinia and Myobatrachus are reversed in the two studies. The trenchant difference between our results is in the placement of Paracrinia. Our results placed it strongly as the sister taxon of Assa + Geocrinia, whereas Read et al. (2001) placed it as the sister taxon of the myobatrachids that they studied, with the exception of Taudactylus. Conclusive resolution of this problem will require all available evidence to be analyzed simultaneously.

We include Mixophyes (formerly in Limnodynastidae) and Rheobatrachus (sole member of former Rheobatrachidae) in Myobatrachidae (sensu stricto); Read et al. (2001) did not include those taxa in their study. We obtain a sister-taxon relationship between Mixophyes and Rheobatrachus (although this is only weakly corroborated) and association of Mixophyes (and Rheobatrachus) with Myobatrachinae, inasmuch as Mixophyes has traditionally been assigned to Limnodynastinae. Further discussion can be found in the Taxonomy section.

“Leptodactylidae”:

The paraphyly and polyphyly of “Leptodactylidae” is starkly exposed by this analysis, being paraphyletic with respect to all hyloid taxa except Heleophrynidae and Sooglossidae (fig. 57). Because of the extensiveness of the paraphyly and the complexity of the reassortment of the subsidiary groupings, the various units of a paraphyletic/polyphyletic “Leptodactylidae” must be dealt with before the remainder of Hyloidea can be addressed. Specifically the following nominal families are imbedded within “Leptodactylidae”: Allophrynidae, Brachycephalidae, Bufonidae, Centrolenidae, Dendrobatidae, Hylidae, Limnodynastidae, Myobatrachidae, and Rhinodermatidae. To provide the tools to allow us to discuss the remainder of the hyloid families, we here provide a new familial taxonomy with reference to the old taxonomy provided in figure 50 (insert). We start at the top of figure 56 and address the subfamilies of “Leptodactylidae” as we come to them.

“Telmatobiinae”:

“Telmatobiinae” is found to be polyphyletic (figs. 56, 57, 58, 59), with the austral South American Calyptocephalellini (Telmatobiinae-1: Telmatobufo + Caudiverbera) forming the sister taxon of the Australo-Papuan Myobatrachidae, Limnodynastidae, and Rheobatrachidae; Telmatobiinae-2 being paraphyletic with respect to Batrachyla (Telmatobiinae-3: Batrachylini); and Ceratophryini (Lepidobatrachus (Ceratophrys + Chacophrys)); and Telmatobiinae-4 (Hylorina, Alsodes, Eupsophus) being the sister taxon of a taxon composed of part of the polyphyletic Leptodactylinae (Limnomedusa) and Odontophrynini (Proceratophrys and Odontophrynus; part of nominal Ceratophryinae). As noted in the taxonomic review, Telmatobiinae was united by overall plesiomorphic similarity (e.g., exotrophic tadpoles, non-bony sternum). That the molecular data show Telmatobiinae to be polyphyletic is neither surprising nor unconventional.

The Chilean and Peruvian telmatobiine clade composed of Caudiverbera and Telmatobufo is monophyletic on both molecular and morphological grounds; is highly corroborated as the sister taxon of the Australo-Papuan Myobatrachidae + Limnodynastidae + Rheobatrachidae; and is phylogenetically distant from all other telmatobiine “leptodactylids” (see also San Mauro et al., 2005; fig. 17). (The inclusion of Batrachophrynus is discussed under Batrachophrynidae in the Taxonomy section.) This result is not unexpected as calyptocephallelines have long been suspected to be only distantly related to other telmatobiine leptodactylids (Cei, 1970; Burton, 1998a). Moreover, the region they inhabitat is also home to Dromiciops, a marsupial mammal most closely related to some groups of Australian marsupials and not to other South American marsupials (Aplin and Archer, 1987; Kirsch et al., 1991; Palma and Spotorno, 1999). The previous association of Calyptocephalellini with the South American Telmatobiinae was based on overall similarity with geographically nearby groups. As the sister taxon of the Australian Myobatrachidae + Limnodynastidae, it would be acceptable to place Calyptocephallelinae within some larger familial group, but to maintain familiar usage (and because we have resolved Limnodynastidae, Myobatrachidae, and Rheobatrachidae into redefined Limnodynastidae and Myobatrachidae) we consider it as the family Batrachophrynidae (the oldest available name for calyptocephallelines as currently understood; see “Taxonomy” and appendix 6 for discussion of application of this name).

As suggested by Lynch (1978b), one part of Telmatobiinae-2; (fig. 59), Telmatobiini, is paraphyletic with respect to Batrachylini (Batrachylus) as well as to Ceratophryinae-1 (Ceratophryini). The oldest name for the clade Telmatobiinae-2 (Telmatobius, Batrachyla, Atelognathus) + Ceratophryinae-1 (Ceratophrys, Chacophrys, and Lepidobatrachus) is Ceratophryidae. Within this family we recognize two subfamilies, Telmatobiinae (Telmatobius) and Ceratophryinae (for all remaining genera). Within Ceratophryinae we recognize two tribes: Batrachylini (Batrachyla + Atelognathus) and Ceratophryini (for Ceratophrys, Chacophrys, and Lepidobatrachus). (See the Taxonomy section for further discussion.)

As noted earlier, another former component of Telmatobiinae (Telmatobiinae-3; see figs. 57, 59) is recovered as the sister taxon of one piece of “Leptodactylinae” (Limnomedusa) plus Odontophrynini (Ceratophryinae-2, formerly part of Ceratophryinae). (The polyphyly of “Leptodactylinae” will be addressed under the discussion of that subfamilial taxon.) Because no documented morphological synapomorphies join the two groups of nominal Ceratophryinae (Odontophrynini and Ceratophrynini), and they had previously been shown to be distantly related (Haas, 2003), this result does not challenge credibility. (See further discussion in the Taxonomy section.)

“Hemiphractinae”:

“Hemiphractinae”, which was transferred out of Hylidae and into Leptodactylidae by Faivovich et al. (2005), is united by possessing bell-shaped gills in developing embryos and bearing eggs on the dorsum in shallow depressions to extensive cavities. The subfamily has not been found to be monophyletic by any recent author (Darst and Cannatella, 2004; Faivovich et al., 2005). In our results (figs. 57, 58) we found (1) Hemiphractus is the sister taxon of hyloids, excluding Batrachophrynidae, Myobatrachidae (including Rheobatrachidae), Limnodynastidae, Sooglossidae (including Nasikabatrachidae), and Heleophrynidae; (2) Flectonotus + Gastrotheca; and (3) Stefania + Cryptobatrachus are successively more distant from a clade [branch 371] bracketed by Hylidae and Bufonidae. The evidence for this polyphyly is quite strong, so we recognized three families to remedy this: Hemiphractidae (Hemiphractus), Cryptobatrachidae (Cryptobatrachus + Stefania), and Amphignathodontidae (Flectonotus + Gastrotheca).

Eleutherodactylinae and Brachycephalidae:

Eleutherodactylinae is paraphyletic with respect to Brachycephalidae (Brachycephalus) (fig. 57, 58). There is nothing about Brachycephalus being imbedded within Eleutherodactylus (sensu lato) that requires any significant change in our understanding of morphological evolution, except to note that this allows the large eggs and direct development of Brachycephalus to be homologous with those of eleutherodactylines. This result was suggested previously (Izecksohn, 1971; Giaretta and Sawaya, 1998; Darst and Cannatella, 2004), and no evidence is available suggesting that we should doubt it. Further, to impose a monophyletic taxonomy, we follow Dubois (2005: 4) in placing Eleutherodactylinae Lutz, 1954, into the synonymy of Brachycephalidae Günther, 1858. All “eleutherodactyline” genera are therefore assigned to Brachycephalidae. Previous authors (e.g., Heyer, 1975; J.D. Lynch and Duellman, 1997) have suggested that Eleutherodactylus (and eleutherodactylines) is an explosively radiating lineage. Our results, which places brachycephalids as the sister taxon of the majority of hyloid frogs refocuses this issue. The questions now become (as suggested by Crawford, 2003): (1) Why are the ancient brachycephalids morphologically and reproductively conservative as compared with their sister taxon (composed of Cryptobatrachidae, Amphignathodontidae, Hylidae, Centrolenidae, Dendrobatidae, and Bufonidae, as well as virtually all other “leptodactylid” species)? (2) Why are there so few species in the brachycephalid (eleutherodactyline) radiation relative to their sister group (the former composed of some 700 species, mostly in nominal Eleutherodactylus, and the latter consisting of more than twice as many species)? Additional comments on this taxon will be found under Brachycephalidae in the Taxonomy section.

“Leptodactylinae”:

Although “Leptodactylinae” has at least one line of evidence in support of its monophyly (bony sternum), the molecular data unambiguously expose its polyphyly, with its species falling into two units (fig. 57, 59). The first of these (Leptodactylinae 1–2), is paraphyletic with respect to the cycloramphine unit, called Cycloramphinae-1 in figures 57 and 59, Paratelmatobius and Scythrophrys (an arrangement partially consistent with the suggestion of J.D. Lynch, 1971, that at least Paratelmatobius belongs in Leptodactylinae). The second unit (Leptodactylinae-3, Limnomedusa) is the sister taxon of Odontophrynini (Ceratophryinae-2). Limnomedusa was previously united with other leptodactylines solely by its possession of a bony sternum, but it lacks the foam-nesting behavior found in most other leptodactylines (exceptions being Pseudopaludicola, Paratelmatobius, and some species of Pleurodema). Regardless, the association of Limnomedusa with Leptodactylinae has always been tentative (Heyer, 1975). So, our discovery (corroborating the results of Faivovich et al., 2005) that Limnomedusa is not part of Leptodactylinae is not unexpected; nor does it require extensive homoplasy in the morphological data that are available. We recognize this unit (Leptodactylinae-1 + Cycloramphinae-1 + Leptodactylinae-2; figs. 57, 59, branch 430) as Leptodactylidae (sensu stricto), a taxon that is much diminished compared with its previous namesake but that is consistent with evolutionary history. Further discussion is found under Leptodactylidae in the Taxonomy section.

“Ceratophryinae”:

“Ceratophryinae” (sensu lato) is polyphyletic, with its two constituent tribes, Odontophrynini (Ceratophryinae-2) and Ceratophryninae (Ceratophryinae-1) (sensu Laurent, 1986) being only distantly related (figs. 57, 59, branches 446, 458). As noted elsewhere in this section, there has never been any synapomorphic evidence to associate these two groups. Thus, their distant relationship is not surprising or even unconventional, inasmuch as Barrio (1963; 1968) and Lynch (1971) suggested that these two units are distantly related. Ceratophryini is imbedded in a taxon (figs. 57, 59: Telmatobiinae-2; branch 441) that is weakly corroborated, but is here recognized as a family Ceratophryidae. Odonotophrynini is resolved as the sister taxon of Limnomedusa (formerly in Leptodactylinae), together residing in a group composed largely of former cycloramphines.

“Cycloramphinae” and Rhinodermatidae:

“Cycloramphinae” (sensu Laurent, 1986) was also found to be polyphyletic (figs. 57, 59) in three distantly related groups. Our molecular data overcome the few morphological characters that might be considered synapomorphies of the relevant group. The first of these groups, labeled Cycloramphinae-1, is composed of Scythrophrys and Paratelmatobius and is imbedded within Leptodactylidae (sensu stricto; as part of Leptodactylinae, as discussed earlier.) The second unit, which is labelled Cycloramphinae-2, is Elosiinae (= Hylodinae) of Lynch (1971); although it is relatively weakly corroborated by molecular evidence, it is united by morphological evidence suggested by Lynch (1971, 1973). Cycloramphus (part of Cycloramphinae-2) is tightly linked to Rhinoderma (Rhinodermatidae), one of the points of paraphyly of former Leptodactylidae. Cycloramphinae-2 forms a paraphyletic group with respect to Rhinodermatidae, Telmatobiinae-2, Leptodactylinae-3, and Odontophrynini (Ceratophryinae-2). Because no morphological characteristics that we are aware of would reject this larger grouping, we place these five units into a single family, for which the oldest available name is Cycloramphidae. Within this, we recognize two subfamilies: Hylodinae (for Crossodactylus, Megaelosia, and Hylodes) and Cycloramphinae for the remainder of this nominal family-group taxon.

Our DNA sequence evidence places Thoropa (Cycloramphinae-3) as the sister taxon of the monophyletic Dendrobatidae (figs. 57, 60). We were surprised by this result, because none of the morphological characters that had been suggested to ally Hylodinae with Dendrobatidae are present in Thoropa (T. Grant, personal obs.), and Thoropa most recently has been associated with Batrachyla (J.D. Lynch, 1978b). Nevertheless, our molecular data support this arrangement, and Thoropa has never been more than tentatively associated with the grypiscines (= cycloramphines; Heyer, 1975). Furthermore, manual rearrangements of hylodines and Thoropa used as starting trees for further analysis inevitably led to less parsimonious solutions or returned to this solution as optimal (as implied by the Bremer values). Our first inclination was to place Thoropa into Dendrobatidae, so as not to erect a monotypic family. However, Dendrobatidae, as traditionally conceived, is monophyletic and has a large literature associated with it that addresses a certain content and diagnosis that remained largely unchanged for nearly 80 years. For this reason, we place Thoropa into a monotypic family, Thoropidae, to preserve the core diagnostic features of Dendrobatidae for the large number of workers that are familiar with the taxon.

Centrolenidae and Allophrynidae:

As suggested by Noble (1931), Austin et al. (2002), and Faivovich et al. (2005), Allophryne is closely related to Centrolenidae, together forming a monophyletic group that is the sister taxon of a group composed of most of the former Leptodactylinae (fig. 59; branch 426). Our data reject a close relationship of Centrolenidae to Hylidae, as well as the suggestion by Haas (2003), made on the basis of larval morphology, that Centrolenidae may not be a member of Neobatrachia. Allophryne shares with the centrolenids T-shaped terminal phalanges (J.D. Lynch and Freeman, 1966), which is synapomorphic at this level. We regard Allophryne as a part of Centrolenidae, the sister taxon of a taxon composed of Centrolene + Cochranella + Hyalinobatrachium (which has as a morphological synapomorphy intercalary phalangeal elements).

Brachycephalidae:

Our study found Brachycephalus to be imbedded within Eleutherodactyinae, indeed, within Eleutherodactylus (sensu lato; fig. 57, 58). Previous authors (e.g., Izecksohn, 1971; Giaretta and Sawaya, 1998) suggested that Brachycephalus is allied with Euparkerella (Eleutherodactylinae) on the basis of sharing the character of digital reduction. We did not sample Euparkerella, which could be imbedded within a paraphyletic Eleutherodactylus. This proposition remains to be tested. As noted earlier, Brachycephalidae and Eleutherodactylinae are synonyms, with Brachycephalidae being the older name.

Rhinodermatidae:

We found Rhinoderma to be imbedded within a clade composed largely of South American cycloramphine leptodactylids (figs. 57, 59), more specifically as the sister taxon of Cycloramphus. Because the only reason to recognize Rhinodermatidae has been its autapomorphic life history strategy of brooding larvae in the vocal sac, we place Rhinodermatidae into the synonymy of Cycloramphidae.

Dendrobatidae:

We found Dendrobatidae to be monophyletic and the sister taxon of Thoropa. The former statement is conventional, the latter, surprising. Nevertheless, the highly corroborated nature of this placement (cladistically in the same neighborhood as hylodines, with which it was considered closely allied by some authors, e.g., Noble, 1926, and Lynch, 1973) should close discussion of whether the firmisternal dendrobatids are derived from some austral South American arciferal group (here strongly supported; for dendrobatid girdle architecture see Noble, 1926; Kaplan, 1995) or related to some ranoid or ranid group, a conclusion suggested by some lines of morphological evidence (Blommers-Schlösser, 1993; Ford, 1993; Grant et al., 1997). Thoropa + Dendrobatidae form the sister taxon of Bufonidae. This phylogenetic arrangement is highly corroborated and suggests that Ameerega Bauer, 1986 (a senior synonym of Epipedobates Myers, 1987; see Walls, 1994) is polyphyletic, a result that is consistent with previous studies (e.g., Santos et al., 2003; Vences et al., 2003b). Taxon sampling was limited in all studies to date, however, and we leave it to more exhaustive analyses to assess the details of the relationships within Dendrobatidae.

Hylidae:

If hylids are considered to contain Hemiphractinae (see above), then Hylidae would be catastrophically paraphyletic with respect to leptodactylids (excluding the former calyptocephalellines [Batrachophrynidae]), dendrobatids, bufonids, Allophryne, and centrolenids (figs. 57, 58, 59). This arrangement suggests that the claw-shaped terminal phalanges and intercalary cartilages taken previously to be synapomorphies of Hylidae (sensu lato) are homoplastic and not synapomorphic for Hylidae. Because Hylidae (sensu lato) is broadly para- or polyphyletic, we adopt the concept of Hylidae adopted by Faivovich et al. (2005), that is Hylinae + Phyllomedusinae + Pelodryadinae.

Hylidae (sensu stricto, excluding “Hemiphractinae”) is monophyletic and highly corroborated. Our results are largely congruent with the results of Faivovich et al. (2005), which were based on more sequence evidence and denser sampling of hylids. Faivovich et al. (2005) should be referenced for the evidentiary aspects of hylid phylogenetics. The only significant difference between our results and theirs is that our exemplars of Hyla form a paraphyletic group with respect to Isthmohyla and Charadrahyla, and Hypsiboas is paraphyletic with respect to Aplastodiscus, and the tribe Dendropsophini is not monophyletic as delimited by Faivovich et al. (2005). However, because our density of sampling and evidence is less than in that study, our results do not constitute a test of those results, and we leave their taxonomy unchanged.

Hylinae has long been suspected of being paraphyletic, but our results and those of Faivovich et al. (2005) strongly corroborate the notion that Hylinae is monophyletic and the sister taxon of Pelodyradinae + Phyllomedusinae, both of which are also strongly corroborated as monophyletic.

The apparent polyphyly of Nyctimystes in our results may be real, although our paucity of sampling prevents us from delimiting the problem precisely. Similarly, the long-recognized (Tyler and Davies, 1978; King et al., 1979; Tyler, 1979; Maxson et al., 1985; Hutchinson and Maxson, 1987; Haas, 2003; Faivovich et al., 2005), pervasive paraphyly of Litoria in Pelodryadinae with respect to both Cyclorana and Nyctimystes has obviously been a major problem in understanding relationships among pelodryadines. Ongoing research by S. Donnellan and collaborators aims to rectify these issues in the near future.

Bufonidae:

That Bufonidae is a highly corroborated monophyletic group is not surprising; that we have a reasonably well-corroborated phylogenetic structure within Bufonidae is a surprise (figs. 50 [insert], 60). Like Graybeal (1997; fig. 25), we found Melanophryniscus (which lacks Bidder's organs) to form the sister taxon of the remaining bufonids (which, excluding Truebella, have Bidder's organs). Within this clade, Atelopus + Osornophryne forms the sister taxon of the remaining taxa.

The paraphyly of Bufo with respect to so many other bufonid genera had previously been detected (e.g., Graybeal, 1997; Cunningham and Cherry, 2004), but some associations are unconventional. The relationship of Bufo margaritifer with Rhamphophryne conforms with their morphological similarity, but the nesting of this clade within a group of Asian Bufo was unexpected. The association of Bufo lemur (a species of former Peltophryne in the Antilles) with Schismaderma (Africa) is novel, as is the placement of this group with Bufo viridis and Bufo melanostictus, although Graybeal (1997), at least in her parsimony analysis of molecular data, suggested that Peltophryne was associated with Bufo melanostictus, an Asian taxon.

Obviously, denser sampling will be required to resolve bufonid relationships, but the current topology provides an explicit hypothesis for further investigation. Clearly, Bufo must be partitioned into several genera to remedy its polyphyly/paraphyly with respect to several other nominal genera and to provide a reasonable starting place from which to make progress. For more discussion and the beginnings of this partition, see Bufonidae in the Taxonomy section.

Ranoidea:

Monophyly of Ranoidea (in the sense of excluding Dendrobatidae) was strongly corroborated in our analysis, as well as by other recent analyses (Roelants and Bossuyt, 2005; San Mauro et al., 2005). Ranoidea in our analysis is divided into two major groups (see figs. 50 [insert], 56, 61, 62, 63, 65), which correspond to (1) a group composed of a para- or polyphyletic Microhylidae, Hemisotidae, Hyperoliidae, paraphyletic Astylosternidae, and Arthroleptidae (figs. 61, 62); and (2) a giant paraphyletic “Ranidae” and its derivative satellites, Mantellidae and Rhacophoridae (fig. 63, 65). This is summarized on the general tree (fig. 50 [insert]).

Microhylidae and Hemisotidae:

Our results (figs. 50, 61, 62) do not support the traditional view of subfamilies and relationships suggested by Parker (1934) in the last revision of the family. The notion of polyphyletic Microhylidae falling into two monophyletic groups—(1) Brevicipitinae (as the sister taxon of Hemisotidae); and (2) the remaining microhylids—extends from the suggestion by Blommers-Schlösser (1993) that Hemisotidae and Brevicipitinae are closely related. Because the Type II tadpole that was considered a synapomorphy in microhylids (Starrett, 1973) is not present in brevicipitines (which have direct development) and hemisotids have a Type IV tadpole, there was never any particular evidence tying brevicipitines to the remaining microhylids. Moreover, only a single synapomorphy tied brevicipitines to hemisotines (Channing, 1995), so the evidence for paraphyly/polyphyly of microhylids also was not strong. As suggested by Van der Meijden et al. (2004; and consistent with the results of Biju and Bossuyt, 2003, and Loader et al., 2004, but contrary to the Scoptanura hypothesis of Ford and Cannatella, 1993), we find Brevicipitinae and Hemisotidae to form a monophyletic group, and this taxon to be more closely related to Arthroleptidae, Astylosternidae, and Hyperoliidae than to remaining Microhylidae. For this reason we regard brevicipitines as a distinct family, Brevicipitidae. (We find Dubois', 2005, proposal that Arthroleptidae, Astylosternidae, Brevicipitidae, Hemisotidae, and Hyperoliiidae be considered subfamilies of an enlarged Brevicipitidae, to be an unnecessary perturbation of familiar nomenclature.)

Within the larger group of “microhylids”, Microhylinae is broadly paraphyletic with respect to the remaining subfamilies, with Phrynomantis (Phrynomerinae) being situated near the base of our sampled microhylines, Hoplophryne (Melanobatrachinae) placed weakly next to Ramanella (Microhylinae), and Cophylinae (based on our exemplars of Anodonthyla, Platypelis, Plethodontohyla, and Stumpffia) being found to be monophyletic and placed as the sister taxon of Ramanella (Microhylinae) + Hoplophryne (Melanobatrachinae). Surprisingly, Scaphiophryne (Scaphiophryninae) is deeply imbedded among the microhylids and the sister taxon of part of “Microhylinae” (branch 130, subtending Kaloula, Chaperina, Calluella, and Microhyla). Ford and Cannatella (1993) and Haas (2003) had considered Scaphiophryne to form the sister taxon of the remaining microhylids on the basis of larval features, but because we included Haas' (2003) morphological data in our analysis, we can see that these features must be homoplastic.

Microhylinae is nonmonophyletic, with (1) some taxa clustered around the base of the Microhylidae and weakly placed (e.g., Kalophrynus, Synapturanus, Micryletta); (2) a group of Asian taxa (e.g., KaloulaMicrohyla) forming the sister taxon of Scaphiophryne; and (3) a New World clade (i.e., the group composed of Ctenophryne, Nelsonophryne, Dasypops, Hamptophryne, Elachistocleis, Dermatonotus, and Gastrophryne) placed as the sister taxon of Cophylinae + Melanobatrachinae + Ramanella.

Our picture of “Microhylinae” runs counter to the little phylogenetic work that has been done so far, especially with respect to the cladogram of New World taxa by Wild (1995). Wild's (1995; fig. 34) cladogram assumed New World monophyly, was rooted on a composite outgroup, and is strongly incongruent with our topology. Our solution is to (1) recognize Gastrophryninae for the New World taxa that do form a demonstrably monophyletic group (including Ctenophryne, Nelsonophryne, Dasypops, Hamptophryne, Elachistocleis, Dermatonotus and Gastrophryne); and (2) restrict Microhylinae to a monophyletic group including Calluella, Chaperina, Kaloula, and Microhyla. The genera that we have not assigned to either Gastrophryninae or Microhylinae (sensu stricto), or that are clearly outside of either group (e.g., Synapturanus or Kalophrynus), we treat as incertae sedis within Microhylidae. The arrangement asserted without evidence by Dubois (2005), of an Old World Microhylini and New World Gastrophrynini, within his Microhylinae, is specifically rejected by the basal position in our tree of Kalophrynus and Synapturanus, far from our Microhylinae and Gastrophryninae.

As suggested by Savage (1973), Dyscophinae is polyphyletic, with Calluella deeply imbedded within Asian microhylines and Dyscophus placed as the sister taxon of a group composed of members of Asterophryinae (Cophixalus, Choerophryne, Genyophryne, Sphenophryne, Copiula, Liophryne, Aphantophryne, Oreophryne) and Asterophryinae (Callulops). Genyophryninae is clearly paraphyletic with respect to Asterophryinae, as suggested by Savage (1973) and Sumida et al. (2000a). For this reason we regard Asterophryinae and Genyophryninae as synonyms, with Asterophryinae being the older name for this taxon. This allows the optimization of direct development as a synapomorphy for the combined taxon.

Arthroleptidae, Astylosternidae and Hyperoliidae:

We found an African group composed of Hyperoliidae, Astylosternidae, and Arthroleptidae to constitute a highly corroborated clade, the sister taxon of Hemisotidae + Brevicipitidae (fig. 62). This existence of this group was suggested previously but has not been substantiated by synapomorphies (Laurent, 1951; Dubois, 1981; Laurent, 1984b; Dubois, 1987 “1985”, 1992). Within this group we found Hyperoliidae (excluding Leptopelis) to form a monophyletic group.

Phylogenetic structure within Hyperoliidae has been contentious, with various arrangements suggested by different authors. Our results differ significantly from all previously published hyperoliid trees (Drewes, 1984; Channing, 1989; Vences et al., 2003c). Like Vences et al. (2003c), we found Leptopelis (Hyperoliidae) to form a monophyletic group that is separate from the remainder of Hyperoliidae and placed with a group composed of the Astylosternidae + Arthroleptidae. The consideration of Leptopelinae as a subfamily of Hyperoliidae cannot be continued because it renders Hyperoliidae (sensu lato) paraphyletic. We restrict the name Hyperoliidae to the former Hyperoliinae, which in addition to our molecular data, is supported by the synapomorphic presence of a gular gland (Drewes, 1984).

We found Astylosternidae to be paraphyletic with respect to Arthroleptidae, with Scotobleps (Astylosternidae) being the sister taxon of Arthroleptidae (fig. 62). No previous hypotheses of relationship within Astylosternidae or Arthroleptidae have been rigorously proposed (Vences et al., 2003c), so our results are the first to appeal to synapomorphy. Our finding that Schoutendenella is paraphyletic with respect to Arthroleptis is particularly noteworthy because recognition of Schoutedenella as distinct from Arthroleptis has been contentious (e.g., Laurent, 1954; Loveridge, 1957; Schmidt and Inger, 1959; Laurent, 1961; Poynton, 1964b; Laurent, 1973; Poynton, 1976; Poynton and Broadley, 1985; Poynton, 2003). Laurent and Fabrezi (1986 “1985”) suggested that Schoutedenella is more closely related to Cardioglossa than to Arthroleptis, an hypothesis rejected here.

Ranidae, Mantellidae, and Rhacophoridae:

Our results for this group are similar in some respects to those presented by Van der Meijden et al. (2005; fig. 36). Differences in results may be due to our denser taxon sampling, to their greater number of analytical assumptions, their inclusion of RAG-1 and RAG-2, which we did not include, or their lack of 28S, seven in absentia, histone H3, tyrosinase, and morphology, which we did include. Final resolution will require analysis of all of the data under a common assumption set.

We found a taxon composed of a broadly paraphyletic “Ranidae”, and monophyletic Mantellidae + Rhacophoridae to form the sister taxon of Microhylidae + Hemisotidae + Hyperoliidae + Arthroleptidae + Astylosternidae (fig. 50 [insert], 61, 63). The results are complex but are comparable to a group of smaller studies that dealt overwhelmingly with Asian taxa (Tanaka-Ueno et al., 1998a, 1998b; Bossuyt and Milinkovitch, 2000; Emerson et al., 2000a; Marmayou et al., 2000; Bossuyt and Milinkovitch, 2001; Kosuch et al., 2001; Grosjean et al., 2004; Roelants et al., 2004; Jiang and Zhou, 2005). This overall result varies widely from Bossuyt and Milinkovitch (2001), who found Mantellinae + Rhacophorinae as the sister taxon of Nyctibatrachinae + Raninae; this clade sister to Dicroglossinae + Micrixalinae, and Ranixalinae sister to them all.

We find Ptychadeninae (Ptychadena being our exemplar genus) to be the sister taxon of the remaining “Ranidae”, a highly corroborated result (fig. 63). The sister taxon of Ptychadeninae is composed of Ceratobatrachinae (Ingerana, Discodeles, Ceratobatrachus, Batrachylodes, and Platymantis) and the remaining “ranids”. Here we differ significantly from Roelants et al. (2004), inasmuch as they considered Ingerana to be an occidozygine, whereas we find Ingerana to be in Ceratobatrachinae, where it had originally been placed by Dubois (1987 “1985”).

We find a major African clade (fig. 63; branch 192), similar to the results of Van der Meijden et al. (2005). One clade (branch 193) is Phrynobatrachinae of Dubois (2005), composed of a paraphyletic Phrynobatrachus, within which Phrynodon and Dimorphognathus are imbedded. A second component (branch 200) is composed of Conrauinae (Conraua), Ranixalinae (Indirana), Petropedetinae, and Pyxicephalinae sensu Dubois (2005). Petropedetinae of Dubois (2005) (Petropedetes + Arthroleptides, subtended by branch 205), forms the sister taxon of Indirana (Ranixalinae of Dubois, 2005). Pyxicephalus + Aubria (branch 210) form the sister taxon of the Pyxicephalinae of Dubois (2005), the “southern African clade” of Van der Meijden et al. (2005): Tomopterna, Arthroleptella, Natalobatrachus, Afrana, Amietia, Strongylopus, Cacosternum, and Anhydrophryne. We place (1) Phrynobatrachus (and its satellites Phrynodon and Dimorphognathus) in Phrynobatrachidae; (2) Arthroleptides, Conraua, Indirana, and Petropedetes in Petropedetidae; (3) Afrana, Amietia, Anhydrophryne, Arthroleptella, Aubria, Cacosternum, Natalobatrachus, Pyxicephalus, Strongylopus, and Tomopterna in Pyxicephalidae, as had Dubois (2005). (See fig. 63 and further discussion of these groups in the Taxonomy section.)

Roelants et al. (2004), who did not include any African taxa in their study, proposed Indirana to be the sister taxon of Micrixalinae, although their evidence did not provide resolution beyond a polytomy with (1) the LankanectesNyctibatrachus clade; and (2) the ranine-rhacophorine-mantelline clade. However, we found Indirana to be deeply imbedded in an African clade otherwise composed of Conraua, Arthroleptides, and Petropedetes (a clade we consider a family, Petropedetidae). Dissimilarly, Van der Meijden et al. (2005) found, albeit weakly, Indirana as the sister taxon of Dicroglossinae. Nevertheless, our result is highly corroborated, although it is based on less overall evidence than that of Van der Meijden et al. (2005), although as noted previously, analyzed differently. Our sequence evidence for Indirana is the same 12S and 16S GenBank sequences produced/ used by Roelants et al. (2004), so contamination or misidentification is not an issue.

Like Roelants et al. (2004), we find occidozygines to form the sister taxon of Dicroglossinae, with the latter containing Paini (our exemplares being members of Nanorana and Quasipaa), which had been transferred from Raninae into Dicroglossinae by Roelants et al. (2004). Unlike their data, ours place Nanorana not within Paa, but as the sister taxon of a clade composed of Fejervarya (which we show to be paraphyletic), Sphaerotheca, Nannophrys, Euphlyctis, and Hoplobatrachus.

Our results are broadly consistent with several other studies showing that Hoplobatrachus (Limnonectini) is the sister taxon of Euphlyctis (Dicroglossini) (Bossuyt and Milinkovitch, 2001; Kosuch et al., 2001; Grosjean et al., 2004; Roelants et al., 2004). Limnonectini (sensu Dubois, 1992) is therefore rejected as nonmonophyletic. Limnonectes (including Taylorana Dubois, 1987 “1986”, as a synonym; a result congruent with Emerson et al., 2000a) forms the sister taxon of a clade formed by Paini (Quasipaa), Nanorana, Nannophrys, and the remaining members of “Limnonectini” (Fejervarya, Sphaerotheca, and Hoplobatrachus) and Dicroglossini (Euphlyctis), a result congruent with Grosjean et al. (2004). Marmayou et al. (2000) found Fejervarya + Sphaerotheca to form the sister taxon of a monophyletic Limnonectes + Hoplobatrachus, but they did not include Euphlyctis in their study. Roelants et al. (2004; fig. 35), and Jiang et al. (2005; fig. 42), and Jiang and Zhou (2005; fig. 41) found Paini to be imbedded within this group (Dicroglossinae), and our results confirm their result. This suggests that a character that has been treated as of particular importance to ranoid systematics, forked or entire omosternum, is considerably more variable than previously supposed (see Boulenger, 1920: 4), regardless of the weight placed on this character by some taxonomists (e.g., Dubois, 1992).

Our topology is not consistent with that of Roelants et al. (2004), Jiang et al. (2005), and Van der Meijden et al. (2005) in that we do not recover a monophyletic Paini, instead finding our exemplars (2 species of Quasipaa and 1 of Nanorana) to form a pectinate series leading to “Fejervarya” + Hoplobatrachus (Euphlyctis and Nannophrys were pruned for this discussion because they were not part of the study of Jiang et al., 2005; fig. 42). Although our topological differences from the results of Roelants et al. (2004) apparently reflect differences in evidence and sampling, we have more of both. The difference between our results and those of Jiang et al. (2005) seemingly do not reflect differences at the level of descriptive efficiency at the level of unrooted network. We do have a bit more resolution between their groups 1 and 2 as a paraphyletic grade, rather than as a polytomy. By treating Hoplobatrachus and Fejervarya as their outgroups on which to root a tree of Limnonectes + Paini, the study by Jiang et al. (2005) inadvertantly forced Paini to appear monophyletic. Examination of the trees and associated unrooted networks (fig. 64) support this view. That Euphlyctis, Hoplobatrachus, and Nannophrys lack spines on the forearms and belly as in Paini is incongruent evidence. Nevertheless, it does strengthen our view that Group 1 of Jiang et al. (2005) deserves generic recognition, and that Paini, as nonmonophyletic, must be placed into the synonymy of Dicroglossinae. (See the account of Dicroglossinae in the Taxonomy section.)

A trenchant difference between our results and those of Roelants et al. (2004; but the same as found by Van der Meijden et al., 2005) is in the placement of Lankanectes + Nyctibatrachus. Roelants et al. (2004) placed this taxon outside of most of “Ranidae” (excepting Micrixalinae and Indiraninae, which we also found to be placed elsewhere). We find Lankanectes + Nyctibatrachus to be the sister taxon of Raninae, excluding Amietia, Afrana, and Strongylopus (and Batrachylodes, transferred to Ceratobatrachidae, as discussed earlier).

Dubois' (1992) Amolops (containing the subgenera Amo [which we did not study], Amolops, Huia, and Meristogenys) is demonstrated to be polyphyletic (a result congruent with Roelants et al., 2004; who did not study Huia; fig. 65). At least with respect to our exemplars, the character of a ventral sucker on the larva is suggested by our results to be convergent in Amolops (in the sense of including Amo), Huia, and Meristogenys (as well as in Pseudoamolops).

As expected, the genus Rana (sensu Dubois, 1992) is shown to be wildly nonmonophyletic, with Dubois' sections Strongylopus (Afrana and Strongylopus) and Amietia (Amietia) being far from other “Rana” in our results. (This result is consistent with that of Van der Meijden et al., 2005, and was anticipated by Dubois, 2005.) In this position, Section Strongylopus is paraphyletic with respect to Cacosternum + Anhydrophyne (fig. 63). As noted earlier, we transfer Sections Strongylopus and Amietia out of Ranidae and into a newly recognized family, Pyxicephalidae, as was done by Dubois (2005). (See the Taxonomy section for further discussion.)

As noted in the Review of Current Taxonomy, understanding the phylogeny of Hylarana-like frogs (Dubois' sections Babina and Hylarana) is critical to understanding ranid systematics. Our results show that Boulenger (1920) was correct that “Hylarana” (sensu lato) is polyphyletic, or at least wildly paraphyletic. The plesiomorphic condition in Ranidae is to have expanded toe digits, as in Rhacophoridae + Mantellidae and farther outgroups, so this discovery merely illuminates that “Hylarana” was constructed on the basis of plesiomorphy. Dubois' (1992) Section Hylarana is paraphyletic with respect to Amolops, Meristogenys, and Huia, as well as most of sections Babina, Amerana, Rana, Pelophylax, and Lithobates. Further, the Hylarana subsection Hydrophylax (his humeral-gland group) is polyphyletic (as hinted at by the results of Matsui et al., 2005; fig. 46), in our results placing this group in two places: (1) Sylvirana guentheri forms the sister taxon of subgenus Hylarana (in the non-humeral-gland group) and (2) in a group containing Hydrophylax galamensis and Papurana daemeli. Our findings are largely congruent with the results of Roelants et al. (2004), who suggested “S.” guentheri as sister to H. erythraea, but who also suggested that this “erythraea clade” is sister to a clade containing Sylvirana nigrovittata. Marmayou et al. (2000; fig. 37) found strong support for H. erythraea and H. taipehensis as sisters, and weak support for “S.” guentheri to be part of that clade. They did show, weakly but consistently, that Sylvirana is polyphyletic with respect to Hylarana (Marmayou et al., 2000). Kosuch et al. (2001) found Amnirana to be the sister taxon of Hydrophylax galamensis + Sylvirana gracilis. Differences in data size and taxon sampling may account for differences in tree topology among these studies, but the substantial results are similar. Roelants et al. (2004) included exemplars of subgenus Hydrophylax sensu Dubois and subgenus Hylarana sensu Dubois, but not Amnirana as in our study. Kosuch et al. (2001) included exemplars of Hydrophylax and Amnirana, but not Hylarana, as was done for our study; but Roelants et al. (2004), Kosuch et al. (2001), and Marmayou et al. (2000) did not include species of Papurana or Tylerana.

The subsection Hylarana (the non-humeral-gland group) is polyphyletic as well. (This is not surprising, as subsection Hylarana never did have any suggested synapomorphies; again, this is consistent with the results of Matsui et al., 2005.) The component subgenus Hylarana is most closely related to Sylvirana guentheri (subsection Hydrophylax); subgenus Chalcorana (subsection Hylarana) is most closely related to Hydrophylax + Amnirana (subsection Hydrophylax); Tylerana (subsection Hylarana) is most closely related to Papurana (subsection Hydrophylax); and Clinotarsus (subsection Hylarana) forms the sister taxon of Meristogenys (subgenus of Amolops sensu Dubois, 1992). Glandirana (subsection Hylarana) is the sister taxon of Pelophylax (section Pelophylax). Eburana (subsection Hylarana) is the sister taxon of Huia (subgenus of Amolops sensu Dubois), and our exemplar of Odorrana (subsection Hylarana) is the sister taxon of “Amolopschapaensis, a result similar to those of Jiang and Zhou (2005; fig. 41), who found Eburana nested within Odorrana (see the Taxonomy section for further discussion).

As suggested by Hillis and Davis (1986) and confirmed by Hillis and Wilcox (2005), Dubois' (1992) Section Pelophylax is polyphyletic, with one part, Pelophylax (sensu stricto), being found most closely related to Glandirana (section Hylarana) and the part composed of Aquarana and Pantherana being paraphyletic with respect to Dubois' Section Lithobates, as well as one species in his Section Rana (R. sylvatica). Our results do not conflict with Roelants et al. (2004), who found Pelophylax (P. lessonae, P. nigromaculata) to be the sister taxon of Amolops cf. ricketti (A. ricketti and P. lessonae not included in our study). Roelants et al. (2004) also found that the AmolopsPelophylax clade is sister to a “Sylvirana”–HylaranaChalcoranaHydrophylaxPulchrana clade, which is largely consistent with our findings. (We did not study Pulchrana.) Jiang and Zhou (2005) had results that were only partly congruent with ours and with those of Roelants et al. (2004). Jiang and Zhou (2005; fig. 41) found Pelophylax to form a monophyletic group with Nidirana and Rana, and this group formed the sister taxon of Amolops. The next more inclusive group was found to include the RugosaGlandirana clade.

Dubois' (1992) Section Amerana is recovered as monophyletic and the sister taxon of Pseudorana + Rana + Pseudoamolops. Section Rana (our exemplars being Rana japonica, R. temporaria, and R. sylvatica) is recovered as polyphyletic, with one component (Rana japonica and R. temporaria) being paraphyletic with respect to Pseudoamolops, and another (R. sylvatica) forming the sister taxon of Pantherana (section Pelophylax) + Section Lithobates.

Excluding Dubois' (1992) section Amerana, we find American Rana (i.e., Aquarana, Lithobates, Trypheropsis, Sierrana, Pantherana, and Rana sylvatica) to form a monophyletic group, a conclusion reached previously by Hillis and Wilcox (2005; fig. 44). Section Amerana (subgenera Aurorana plus Amerana [former Rana aurora and R. boylii groups]) is most closely related to the Rana temporaria group (including Pseudorana and Pseudamolops), an arrangement that suggests the results of Case (1978) and Post and Uzzell (1981). Further discussion and generic realignments are provided in the Taxonomy section.

Mantellidae and Rhacophoridae:

We find Mantellidae and Rhacophoridae to be monophyletic sister taxa deeply imbedded within the traditional “Ranidae”, together placed as the sister taxon of Raninae + Nyctibatrachinae (fig. 65). The monophyly of the combined Mantellidae and Rhacophoridae is not controversial and was suggested by a number of authors on the basis of DNA sequence data (e.g., Emerson et al., 2000b; Richards et al., 2000; J.A. Wilkinson et al., 2002; Roelants et al., 2004; Roelants and Bossuyt, 2005; Van der Meijden et al., 2005) as well as the morphological data of Liem (1970).

For mantellids, the phylogenetic structure we obtained is identical to that obtained by Vences et al. (2003d): Boophis ((Aglyptodactylus + Laliostoma) + (Mantidactylus + Mantella)), but different from that of Van der Meijden (2005) ((Aglyptodactylus + Laliostoma) + (Boophis + (Mantella + Mantidactylus)). Although Vences et al. (2003d) demonstrated that Mantidactylus is deeply paraphyletic with respect to Mantella, our limited taxon sampling did not allow us to test that result rigorously.

The basal dichotomy of Rhacophoridae is as suggested by Channing (1989), with Buergeria forming the sister taxon of the remaining rhacophorids. But beyond that level, however, our results are quite different. This is not surprising, given the inherent conflict and lack of resolution in the morphological data gathered so far, as discussed by J.A. Wilkinson and Drewes (2000). We will not discuss in detail the minor differences between our results and those of J.A. Wilkinson et al. (2002) because, although our taxon sampling was somewhat different, we included all of the same genes used in that study, as well as our own.

Our tree suggests polyphyly of Chirixalus, a conclusion to which others had previously arrived (e.g., J.A. Wilkinson et al., 2002): (1) one relatively basal clade (our Kurixalus eiffingeri and “Chirixalusidiootocus) noted previously by J.A. Wilkinson et al.'s (2002) study for which the name Kurixalus Ye, Fei, and Dubois (In Fei, 1999) is available; (2) the group associated with the name Chirixalus (Chirixalus doriae and C. vittatus) forming a paraphyletic grade with respect to Chiromantis (also illustrated by Delorme et al., 2005; fig. 49); and (3) our “Chirixalusgracilipes, except for Buergeria, being the sister taxon of all rhacophorids. We, unfortunately, did not sample “Chirixaluspalpebralis, which J.A. Wilkinson et al. (2002; fig. 48) found in a similar, basal, position, although as shown by the dendrogram published by Delorme et al. (2005; fig. 49), “Chirixaluspalpebralis, which we did not study, will likely be found to be quite distant from Aquixalus (Gracixalus) gracilipes, once Aquixalus is adequately sampled for molecular analysis.

A TAXONOMY OF LIVING AMPHIBIANS

The taxonomy that we propose is consistent with the International Code of Zoological Nomenclature (ICZN, 1999). It will appear to some that we have adopted an unranked taxonomy. This is partially true, but only for above-family-group nomenclature unregulated by the Code. Regardless of widespread perception, the Code does not govern nomenclature above the family group. In fact, it barely mentions the existence of Linnaean nomenclature above the rank of the family group, and it does not specify particular ranks above that category. Our suggested taxonomy is predicated on the recognition that the community of taxonomists has largely discarded its concerns regarding ranks above the family-group level. For example, one no longer hears arguments regarding whether Aves is a class, coordinate with a Class Amphibia, or whether it is at most a family within Archosauria. The reason for this withering of concerns about ranks is that the concerns do not constitute an empirical issue. Notions of rank equivalency are always based on notions of levels of divergence, age, content, or size that are bound to fail for a number of theoretical or empirical reasons24. But, because nominal families and the ranks below them have been regulated by a more or less universally accepted rulebook for more than 160 years (Stoll, 1961), we are not inclined to easily throw out that rulebook or the universal communication that it has fostered. Even though several of the criticisms of Linnaean nomenclature are accurate, the alternatives so far suggested have their own drawbacks. The International Code can be changed, and we expect that changes will be made to meet the needs of modern-day problems.

All taxonomies are rough and ready in the sense that, except for the most general level of communication, they must be qualified implicitly or explicitly with respect to variation in taxon content according to various authors, controversies regarding diagnosis, or, more subtly, the taxon sampling regime (Delorme et al., 2004) and underlying data used to infer the existence of particular taxa. In other words, taxonomies are constructions for verbal and written communication that are inherently limited because they represent sets of theories of relationship and do not communicate information on underlying data or assumptions of analysis. Precision in communication is enhanced by background knowledge on the part of those using the system for communication or, even better, having the relevant tree(s) and data set(s) available from which the taxonomy was derived. For an example of how taxonomies always must be qualified, Ford and Cannatella (1993) explicitly defined Hylidae as the most recent common ancestral species of Hemiphractinae, Hylinae, Pseudinae, Pelodryadinae, and Phyllomedusinae and all of its descendants. This definition was implicitly changed by Darst and Cannatella (2004) to be the ancestor of Pelodryadinae, Phyllomedusinae, and Hylinae, and all of its descendants, because Hemiphractinae was discovered to be paraphyletic and phylogenetically distant from “other” hylids. A casual glance at our tree will show that an application of Ford and Cannatella's (1993) cladographic definition of Hylidae would render as hylids nearly all arciferal neobatrachians, with the exception of Batrachophrynidae, Heleophrynidae, Limnodynastidae, Myobatrachidae, and Sooglossidae—a far cry from any content familiar to any who have used these terms and certainly not promoting precision in the discussion of synapomorphies or even casual notions of similarity25. Furthermore, the molecular evidence that optimizes as synapomorphies for Hylidae (sensu stricto) in the study of Darst and Cannatella (2004) must differ from those proposed by Faivovich et al. (2005) simply because the ingroup and outgroup taxon sampling of the latter is so much denser than that of the former. As taxa are sampled more and more densely, more and more nonhomology will be detected, with concomitant improvements in estimates of phylogeny (W.C. Wheeler, 1992; Zwickl and Hillis, 2002). The controversy as it exists today, regardless of sloganeering, is about how to portray in words hypotheses of monophyly, and revolves not about precision of communicating tree structure or underlying data, but about how to maintain consistency of communication among authors and across studies with a minimum of qualification. All systems so far suggested have limitations; like all maps they must have limitations to be useful. Linnaean taxonomy does promote useless rank controversies, but, as noted above and discussed more fully below, rigid application of cladographic definitions of taxonomic names (such as the method proposed by de Queiroz and Gauthier, 1992) brings other kinds of nomenclatural instability as well.

It is beyond the scope of this work to discuss at length the theory and practice of taxonomy and nomenclature. The ranked and rankless alternatives to expressing phylogenetic relationships in words theoretically are endless but most recently and most clearly discussed by Kluge (2005). To oversimplify his paper, currently competing systems for expressing phylogenetic relationships in words are (1) Linnaean system (Linnaeus, 1758); (2) Annotated Linnaean system (Wiley, 1981); (3) what Kluge termed “Descent Classification” and proponents call “Phylogenetic Taxonomy” (de Queiroz and Gauthier, 1992); (4) the “Set Theory Classification” system of Papavero et al. (2001), as termed by Kluge; and (5) Kluge's (2005) “Phylogenetic System”.

We have taken a sixth approach, one that we think is based on common sense, especially with respect to how systematists use taxonomies and with respect to the state of the discussion, which is still very preliminary and reflecting a deep ambivalence on the part of taxonomists (for all sides of the controversy see: Wiley, 1981; de Queiroz, 1988; de Queiroz and Gauthier, 1994; Cantino et al., 1997; Cantino et al., 1999; Benton, 2000; Nixon and Carpenter, 2000; Withgott, 2000; Kress and DePriest, 2001; Niklas, 2001; Papavero et al., 2001; Pennisi, 2001; Brummitt, 2002; Carpenter, 2003; Keller et al., 2003; Kojima, 2003; Nixon et al., 2003; Schuh, 2003; Kluge, 2005; Pickett, 2005). What we do think is that the conversation will continue for some time and that changes will take place, all discussed fully and not driven by the overheated sloganeering that, unfortunately, characterizes so much of the rhetoric at this time—on all sides—inasmuch as this is a political, not a scientific controversy (see Pickett, 2005, for discussion). With respect to our approach to taxonomy, we, in effect, take the easy way out, we follow the International Code of Zoological Nomenclature (ICZN, 1999) for regulated taxa (family group and down) and apply an unranked taxonomy for unregulated taxa (above family group), the hypotheses for these taxa being derived from their included content and diagnostic synapomorphies.

We expect that regulated nomenclature will increasingly be pushed toward the terminal taxa and that unregulated taxa will increasingly be rankless. The reason for this is that there really is a practical limit to the number of ranks that workers are willing to use. Systematists seemingly are not enamored of new ranks such as grandorders, hyperfamilies, epifamilies, and infratribes (e.g., Lescure et al., 1986) or of the redundancies and controversies over rank that are part and parcel of ranked nomenclature (e.g., see Dubois, 2005). So, our observation is that sociological pressures will push workers towards ever smaller families, especially because there is no scientific or sociological pressure to construct larger families. Regardless, we think that this process will correspond with enormous progress in phylogenetic understanding.

We suggest that the content of an above-family taxon as originally formed by an author renders an implied hypothesis of descent, even if the concept of that taxon predates any particular theory of descent with modification. We spent considerable time determining the original intent of various taxonomic names. Unfortunately, an examination of the original content of the groups denoted by these taxonomic names obviated the need to use many of them because they deviated so widely from all but a few of our phylogenetic hypotheses (e.g., Salientia in the original sense of Laurenti, 1768, not only includes all frogs, but shares Proteus with his Gradientia, a novel phylogenetic hypothesis!).

In some cases (e.g., Caudata), we set aside the intent of the original author in favor of widespread current usage as suggested by subsequent authors. The wisdom of this kind of action is open for discussion (see Dubois, 2004b, 2005), but increasingly the International Commission of Zoological Nomenclature appears to be moving toward usage rather than priority as an important criterion to decide issues, so we take this to be the appropriate strategy.

As noted above, we are unconvinced that cladographic rules governing name assignment (sensu de Queiroz and Gauthier, 1992) necessarily engender enhanced stability or precision of discussion (except in the special case of the crown-group approach to delimitation). However, we do think that associating names of extant taxa with content-specific, ostensively derived concepts (cf. Patterson and Rosen, 1977) will go a long way toward reducing the “wobble” of diagnoses associated with extant taxa as membership changes. One need only look at the history of the use of “Amphibia” to see how the lack of an overarching concept of the taxon has resulted in considerable drift of content and diagnosis. As noted by Laurin (1998a: 10), until Huxley (1863), the term Amphibia applied only to Recent taxa. Haeckel (1866) and Cope (1880) rendered Amphibia paraphyletic by the addition of some fossil taxa, with other authors (e.g., Romer, 1933) continuing the trend until all fossil tetrapods that were not “reptiles” were considered to be members of “Amphibia”. Amphibia was returned to monophyly only by Gauthier et al. (1989) and subsequently restricted back to the groups of original intent by de Queiroz and Gauthier (1992).

Although the discussion is generating considerable self-examination by systematists, we think that cladographically assigned taxonomic names (de Queiroz and Gauthier, 1992) introduce a new kind of nomenclatural instability by tying names, not to content, types, or diagnoses but to tree topology26. Avoiding this instability requires great caution in the application of that naming convention. Nevertheless, in our judgment it is unlikely that a fourth “order” of living amphibians will be discovered, so application of the cladographic rules suggested by de Queiroz and Gauthier (1992) governing the application of the names Anura, Caudata, and Gymnophiona could be salutary for purposes of discussing fossil relatives of these crown groups.

Our strategy in designing a taxonomy for unregulated taxa is to preserve, as nearly as practical, the originally implied phylogenetic content of named above-family-group taxa. We also attempted to apply older names for above family-group taxa, but because of the constant redefinition of many of these taxa, we could solve these only on an ad hoc basis, depending on use, original intent, and recency of coining of the name(s).

In several cases, we changed the ranks of some regulated taxa from subfamilies to families to provide flexibility and help workers in the future with the problems inherent in ranked hierarchies. Because all names above the regulated family group are unaddressed by the International Code of Zoological Nomenclature (ICZN, 1999) we have regarded all of these names as unranked, but within the zone normally associated with class and order (whatever that might mean to the reader). We have not been constrained by recommendations regarding name formations and endings for ranks above the level of family group simply because we believe that these are unworkable and that they merely exacerbate the previously recognized problems of taxonomic ranks (de Queiroz and Gauthier, 1992).

Although we argue that taxonomy should reflect knowledge of phylogeny as closely as possible, by eliminating all paraphyly and recognizing all clades, we focused our attention primarily on the taxonomy of clades above the “genus level” for three reasons. First, for the most part our taxon sampling was inadequate to test prior hypotheses of intrageneric relationships for most genera. The practical implication of this inadequacy is that we lack evidence to refer the majority of species in a more refined generic taxonomy, which would require those species to be placed as incertae sedis, a cumbersome solution with little payoff. The other alternative—expanding the content of genera to enforce monophyly is equally unsatisfactory in these cases, as it overlooks the finer-level knowledge of phylogeny that exists but, for practical reasons was not brought to bear in this analysis. Secondly, the bulk of phylogenetic research since the mid-1970s has focused primarily on “genus-level” diversity, which means that a considerable amount of evidence, both molecular and morphological, has been generated for those groups, most of which was not included in the present study. Third, we see the value of the present contribution to be in framing finer level problems that are better addressed by regional specialists who can achieve more exhaustive taxon and character sampling.

Our consensus tree is shown in figure 50 (insert), which also displays the current and recommended family-group taxonomy. We modify the current generic taxonomy in places in this section, but those changes are not reflected in the figure for purposes of clarity in “Results”. With minor exceptions, all clades are highly corroborated by molecular evidence (and morphological evidence on many branches as well) as estimated by Bremer values and parsimony jackknife frequencies (see below and appendix 4 for these values by branch). Because this study rests on the largest amount of data applied to the problem of the relationships among living amphibians, we provide a new taxonomy that we think will provide a better reference for additional progress.

This taxonomy of living amphibians is based on a phylogenetic analysis of 532 terminals, on the basis of a total of 1.8 million bp of nuDNA and mtDNA sequence data (x̄ = 3.7 kb/terminal) in addition to the morphological data from predominantly larval morphology presented by Hass (2003), the only comparable data set across all frogs. Despite the fact that this is, so far, the most data-heavy analysis of amphibians, we expect to be criticized for presenting this taxonomy for four reasons:

(1) This taxonomy will be criticized both as premature and as not conservative. However, the underlying cladogram reflects the best overall estimate of phylogeny on the most thorough dataset applied to the issue. The alternative—to stick for sociological reasons to an old taxonomy that is clearly misleading and based on relatively little evidence—certainly will not efficiently promote additional research. Some will attempt to defend as conservative the old arrangements, especially favored paraphyletic groups, but mostly this will mean socially conservative, not scientifically conservative, something detrimental to scientific progress. As revealed in the “Review of the Current Taxonomy”, much of the existing taxonomy of amphibians stands on remarkably little evidence and has simply been made plausible through decades of repetition and reification.

A similar argument is that we should retain the status quo with respect to taxonomy until we are “more sure” of a number of weakly recovered relationships. This position ignores how little evidence underlies the existing classifications. Indeed, our taxonomy explains more of the evolution of amphibian characteristics than the existing classification(s) and has the distinction of attempting to be explicitly monophyletic over all of the evidence analyzed. We are surely mistaken in several places, but this is better than continuing to recognize taxonomic groups that are known to be inconsistent with evolutionary history, regardless of social convention. We do go beyond our data in several places (e.g., Brachycephalidae, Bufonidae) and recognize some groups whose monophyly we have not rigorously tested. The reason for this is to attempt to delimit new hypotheses and not sit idly by while major problems are concealed by convention. Critics may charge that this is no different from post facto “diagnosis” of subjective similarity groupings (e.g., Dubois, 1987 “1985”, 1992). However, in each case we think there is good reason to expect our taxa to obtain as monophyletic—and that leaving the taxonomy as it exists does nothing to promote improved understanding of evolutionary history.

(2) Some will be critical of the fact that we have not included all of the morphological data that have been presented by other authors. Early in the development of this work, we made an attempt to marshal the disparate but extensive number of characters presented by such authors as J.D. Lynch (1973), Estes (1981), Duellman and Trueb (1986), Milner (1988), Nussbaum and Wilkinson (1989), Trueb and Cloutier (1991), Ford and Cannatella (1993), Larson and Dimmick (1993), Milner (1993 (1994), McGowan and Evans (1995), Shubin and Jenkins (1995), M. Wilkinson and Nussbaum (1996), Laurin and Reisz (1997), Laurin (1998a), Maglia (1998), Carroll et al. (1999), M. Wilkinson and Nussbaum (1999), Carroll (2000a), Laurin et al. (2000), Milner (2000), J.S. Anderson (2001), Gardner (2001), Kaplan (2001), Zardoya and Meyer (2001), Gardner (2002), Gower and Wilkinson (2002), Laurin (2002), Scheltinga et al. (2002), and Báez and Pugener (2003). What we found, not surprisingly, is that different studies tended to generalize across different exemplars, even if they were working on the same groups, and that in some cases putative synapomorphies had been so reified through repetition in the literature that it was difficult, if not impossible, to ascertain which taxa (much less which specimens) had actually been evaluated for which characters. We also found that many of the new characters remain in unpublished dissertations (e.g., Cannatella, 1985; Ford, 1990; S.-H. Wu, 1994; Graybeal, 1995; da Silva, 1998; Scott, 2002), where ethics dictates they not be mined for information if they are new, and prudence dictates that the information in them not be taken at face value if they are old and still unpublished.

Further, most of the paleontological literature reflects such incomplete sampling of living taxa as to oversimplify living diversity. (One does not read evolution from the rocks, but the rocks certainly are an undersampled component of our study.) Reconciling all morphological descriptions of characters in comparable form, obviously, is the next big step, for someone else, and in combined analysis this will constitute a test of our results and taxonomy. This problem calls for careful evaluation of all morphological characters across all taxonomic groups concomitant with the evaluation of relevant fossil groups. This is a big task, but one worth doing well. Unfortunately, this kind of infrastructural science is not flashy and therefore will not attract funding from already oversubscribed and underfinanced granting agencies. (See Maienschein, 1994, for an essay on the dangers to science from the preoccupation by administrators and funding agencies with the “cutting edge”.)

(3) Some will criticize our analytical methods. We have been conservative with respect to analytical assumptions. Beyond attempting to maximize explanatory efficiency, some workers prefer to incorporate assumptions about the evolutionary process by the addition of particular evolutionary models. This is obviously a discussion that we think will continue for a long time because of the serious philosophical and evidentiary issues involved.

Some will be uncomfortable that such a large proportion of our data are molecular (even though most of our results are generally conventional). We believe that it is better to present a taxonomy that represents explicit, evidence-based hypotheses of relationships than to retain a taxonomy solely because we are used to it. Some will want to exclude all sequence data that require alignment. Unfortunately, this assumes that same-length sequences lack evidence of having had length variation, an assumption not supported by evidence (Grant, unpubl.). Others will want to “correct” alignments manually (although this is likely to increase the number of transformations required to explain sequence variation). Although such methodological choices are crucial and should continue to be debated (indeed, we urge authors and editors of empirical papers to be more explicit about both their methods of alignment and analysis and their reasons for employing them), the issue at hand is that it is time to move away from a taxonomy known to be fatally flawed and that promotes misunderstanding and into a scientific dialogue that will promote a much improved understanding of the evolution of amphibian taxic, life history, and morphological diversity.

(4) We will be trivially criticized for formulating new taxonomic names with 19 authors. Times change and collaborations on this scale are necessary to answer global questions. That a new name can have 19 authors may be cumbersome, but, authorship is not part of the scientific name. And, regardless of recommendations made in the Code (ICZN, 1999) this authorship reflects accurately the extensive effort in collecting samples, sequencing, data analysis, and writing that work on this scale requires.

Although our results will undoubtedly allow considerable progress to be made, by nearly doubling the number of amphibian species for which DNA sequences are available in GenBank, projects such as this one generate questions as well as answers. Our results therefore will provide a reasonably well-tested departure point for future studies by identifying outstanding problems that are especially worthy of investigation.

Taxonomic Accounts

Below we present ancillary information and discussion to accompany the taxonomy presented in figures 50 (insert) and 66 (a reduced tree of family-group taxonomy). (Table 5 provides names of taxa/branches on the interior of the tree shown in figure 66, and figure 67 provides the taxonomy of amphibians in condensed form.) Most morphological evidence is addressed in accounts, but molecular synapomorphies are provided where relevant in appendix 5, with branch numbers corresponding with those noted in the various figures. We are conservative in the scientific sense in that we stick close to the preponderance of evidence and not to tradition. Genera in bold listed under Content represent those from which one or more species were included in our analysis (as DNA sequences either generated or by us or others and available via GenBank). A justification is provided for inclusion of taxa that were not sampled. Synonymies provided in the family group and below conform to the International Code of Zoological Nomenclature (ICZN, 1999). We include citations only to original uses and not to emendations, rank changes, or incorrect subsequent spellings. More extensive discussion of specific nomenclatural issues are dealt with in appendix 6. A summary of generic name changes is presented in appendix 7. We do not address fossil taxa, although they can be placed within this framework with relatively little effort. Dubois (2005) recently provided a taxonomy of living amphibians and their fossil relatives (Neobatrachii in his sense). Because his taxonomy appeals to a taxonomic philosophy deeply steeped in the importance of ranks and personal authority and the unimportance of evidence and logical consistency with evolutionary history, we comment on it only where necessary.

For taxa above the family group, which are not regulated by the Code, homonymy remains an unresolved issue in amphibian nomenclature because, even if the original author intended one content (i.e., one hypothesis of relationship), subsequent authors saw (and may see) little problem in redefining these names to fit revised hypotheses of relationship. For these taxa we do not provide a synonymy because in the absence of any regulatory tradition of above-family-group nomenclature, we have tried to optimize on the hypothesis of relationship intended by the author (or redefiner) of that taxon. Although we do not provide a “synonymy” in the accounts of unregulated taxa, we variably note in appendix 6 (“Nomenclature”) synonyms, near-synonyms, and problematic nomenclatural issues.

The structure of the taxonomic accounts is straight-forward with several categories of information: (1) the name and author of the taxon (and where appropriate and to enhance navigation among records, bracketed numbers are associated that correspond to the numbered branches in our various figures and tables in “Results”); (2) a list of available names if application of the name is regulated by the International Code of Zoological Nomenclature; (3) an etymology if the name of a taxon is used for the first time; (4) the name and branch number of the immediately more inclusive taxon; (5) the name and branch number of the sister taxon; (6) a statement of the geographic distribution of the taxon; (7) the concept of the taxon in terms of content; and (8) a characterization and diagnosis, which is merely a general summary of the salient features of the animals that are included in the taxon under discussion, and characters (either synapomorphic or not) that differentiate this taxon from others. Where a character is thought to be a synapomorphy, this is stated. If the explicit statement is not made, then the character should be assumed to be of unknown polarity. Because we included Haas' (2003) characters in the analysis, for each group we list all unambiguously optimized synapmorphies for that data set, reported using Haas' original numbering scheme (e.g., Haas 34.1). Otherwise, we have not attempted to be exhaustive nor to make these differentia explicitly comparable for the simple reason that the challenge of sorting out the published record regarding the morphological characteristics of amphibians will be enormous and, clearly, is outside of the scope of this work27. Regardless, that next step is an important one in elucidating the morphological evolution of amphibians. The characterization and diagnosis is followed by (9) various systematic comments and discussion. Considerable taxonomic “sausage making” is evident in these sections, particularly with respect to the larger and more chaotic genera, which we have not been shy about partitioning because considerable redistribution of taxonomic names needs to happen if we are going to progress towards a taxonomy that reflects evolutionary history. In some places our changes have not been successful in producing a taxonomy that is entirely monophyletic. Our rationale for failing to propose a more precise taxonomy was given earlier, and we are confident that future work will correct this shortcoming in our proposal. To that end, we emphasize and discuss the specific problems and inadequacies for each of these cases. Some workers will not appreciate the loose-ends that remain untied and will prefer the old approach of concealing these questions. Our position, however, is that unless these problems are advertised, the sociological response of the scientific community will be to let sleeping dogs lie.

In a few places in the taxonomy, we do not render taxonomic changes suggested by our tree. In the cases of “Eleutherodactylus” and “Centrolene”, our sampling density is so low compared to the species diversity that our results could not be practically translated into an informative taxonomy. In two other cases, the reason is that we do not consider our results to constitute a sufficient test of a published cladogram, based on a data set that includes as a subset the data over which we generalized. The first of these is in Hylinae, where our data represent a subset of the data (and concomitant results) of Faivovich et al. (2005), meaning that our analysis does not constitute an adequate test of their results. The second is plethodontid salamanders, where the placement of certain taxa (i.e., Hemidactylium and Batrachoseps) in our tree is based on a subset of data in a published tree (Macey, 2005), which came to different conclusions regarding those critical taxa, based, at least with respect to those taxa, on a more inclusive data set (although the assumptions of analysis were subtly different). In these two cases we do not reject the conclusions of these authors, pending even more inclusive analyses.

[6] AMPHIBIA GRAY, 1825

  • Amphibia Gray, 1825: 213. (See appendix 6 for further nomenclatural discussion.)

  • Range:

    Worldwide on all continents except Antarctica and most oceanic islands, in cold-temperate to tropical habitats.

    Concept and content:

    Amphibia is a monophyletic taxon composed of [7] Gymnophiona J. Müller, 1832, and [23] Batrachia Latreille, 1800, constituting the crown group (i.e., living) amphibians (sensu Amphibia Gray, 1825; not Amphibia of Linnaeus, 1758; cf. de Queiroz and Gauthier, 1992).

    Characterization and diagnosis:

    Beyond our molecular data, Amphibia is diagnosed by many morphological characters. Amphibians, like mammals, retain plesiomorphically the glandular skin of ancestral tetrapods. They do not have the apomorphy of epidermal scales found in sauropsids (turtles and diapsids).

    Trueb and Cloutier (1991) and Ruta et al. (2003) provide extensive discussions of the synapomorphies of Amphibia (as Lissamphibia) in the context of fossil groups. Synapomorphies of Amphibia include (Trueb and Cloutier, 1991): (1) loss of the postparietal bones; (2) loss of the supratemporal bone; (3) loss of the tabular bone; (4) loss of the postorbital bones; (5) loss of the jugal bone; (6) loss of the interclavicle; (7) loss of the cleithrum; (8) papilla amphibiorum present in ear; (9) opercular element associated with the columella; (10) fat bodies present that originate from the germinal ridge associated with the gonads; and (11) pedicellate and bicuspid teeth that are replaced mediolaterally (reversed in some taxa).

    Systematic comments:

    Amphibia is highly corroborated as a taxon, but this only implies that all living amphibians are more closely related to each other than to any other living species and does not address the placement of amphibian groups within the larger structure of relevant fossil tetrapods. All work so far on the overall placement of amphibians (lissamphibians) among fossil groups has depended on inadequate sampling of living taxa and, with the exception of Gao and Shubin (2001), has ignored available molecular data. We hope that additional work on fossil groups, combined with the data presented here, and a better account of living diversity, will further elucidate those relationships.

    [7] GYMNOPHIONA J. MÜLLER, 1832

  • Gymnophiona J. Müller, 1832: 198. (See appendix 6 for nomenclatural discussion.)

  • Immediately more inclusive taxon:

    [6] Amphibia Gray, 1825.

    Sister taxon:

    [23] Batrachia Latreille, 1800.

    Range:

    Pantropical, except for Madagascar and southeast of Wallace's Line; not yet reported from central equatorial Africa.

    Concept and content:

    Gymnophiona is a monophyletic taxon containing the living caecilians (cf. J. Müller, 1832; Cannatella and Hillis, 1993): [8] Rhinatrematidae Nussbaum, 1977, and [9] Stegokrotaphia Cannatella and Hillis, 1993.

    Characterization and diagnosis:

    Caecilians are a bizarre group of legless amphibians, primitively oviparous with aquatic larvae (Rhinatrematidae, Ichthyophiidae), although some species are ovoviparous (with or without direct development) and burrowing, as reflected by considerable numbers of osteological modifications.

    Beyond our molecular data, the following morphological characters have been suggested to be synapomorphies (Nussbaum and Wilkinson, 1989; Trueb and Cloutier, 1991): (1) lacking limbs and girdles (except for one antecedent fossil taxon not included in the crown group; Carroll, 2000b); (2) presence of a dual jaw-closing mechanism; (3) presence of an eversible phallodeum in males formed by a portion of the cloacal wall; (4) annuli encircling the body; (5) paired sensory tentacles on the snout.

    Systematic comments:

    M. Wilkinson has an extensive morphological matrix of more than 180 character transformations (see also Nussbaum and Wilkinson, 1995; M. Wilkinson and Nussbaum, 1996, 1999), which will appear elsewhere, analyzed in conjunction with this and additional evidence.

    [8] FAMILY: RHINATREMATIDAE NUSSBAUM, 1977

  • Rhinatrematidae Nussbaum, 1977: 3. Type genus: Rhinatrema Duméril and Bibron, 1841.

  • Immediately more inclusive taxon:

    [7] Gymnophiona J. Müller, 1832.

    Sister taxon:

    [9] Stegokrotaphia Cannatella and Hillis, 1993.

    Range:

    Tropical northern South America from Amazonian Peru and Brazil, through eastern Ecuador, Colombia, Venezuela, and the Guianas.

    Content:

    Epicrionops Boulenger, 1883; Rhinatrema Duméril and Bibron, 1841.

    Characterization and diagnosis:

    Rhinatrematids are oviparous with aquatic larvae. They are strongly annulated with numerous secondary and tertiary grooves. Like ichthyophiids, rhinatrematids have a short tail and the eyes are visible, although they lie beneath the skin in bony sockets. The tentacle arises near the anterior edge of each eye, and the middle ear contains a stapes (Nussbaum, 1977).

    Beyond the molecular evidence, the following morphological characters have been suggested to be synapomorphies (Duellman and Trueb, 1986; M. Wilkinson and Nussbaum, 1996): (1) dorsolateral process of the os basale present; (2) loss or fusion of the prefrontal with the maxillopalatine; (3) secondary annulus/primary annulus greater than one; and (4) fourth ceratobranchial absent. In addition, the prefrontals are fused with the maxillopalatine as in caeciliids, but not in ichthyophiids and outgroups, rendering the optimization of this character arguable.

    [9] STEGOKROTAPHIA CANNATELLA AND HILLIS, 1993

  • Stegokrotaphia Cannatella and Hillis, 1993: 2.

  • Immediately more inclusive taxon:

    [7] Gymnophiona J. Müller, 1832.

    Sister taxon:

    [8] Rhinatrematidae Nussbaum 1977.

    Range:

    Tropics of southern North America, South America, equatorial East and West Africa, islands in the Gulf of Guinea, Seychelles, and India; Philippines and India to southern China, Thailand, Indochina and the Malayan archipelago.

    Concept and content:

    Stegokrotaphia is a monophyletic group containing [10] Ichthyophiidae Taylor, 1968, and [12] Caeciliidae Rafinesque, 1814 (cf. Cannatella and Hillis, 1993).

    Characterization and diagnosis:

    Stegokrotaphian caecilians show variation in reproductive mode (from aquatic larvae to ovoviviparity) and morphology, with some retaining tails (Ichthyophiidae) and others (Caeciliidae) having lost them (even though a pseudotail may be present). The eyes may be visible (e.g., Ichthyophis), completely hidden beneath bone (e.g., Scolecomorphus), or completely absent (Boulengerula). Unlike in rhinatrematids, the tentacle originates in front of the eye and may be nearly as far forward as the nostril. A stapes is generally present but is lost in some taxa (Nussbaum, 1977).

    Beyond the molecular evidence, the following morphological characters have been suggested to be synapomorphies (Duellman and Trueb, 1986; M. Wilkinson and Nussbaum, 1996): (1) mouth subterminal or recessed rather than terminal; (2) tentacular opening anterior to the anterior edge of the eye; (3) frontal and squamosal articulate; (4) stegokrotaphic skull; (5) vomers in contact throughout their entire length; (6) sides of the parasphenoid converge anteriorly; (7) quadrate and maxillopalatine lack articulation; (8) squamosal and frontal in contact; (9) pterygoid reduced; (10) basipterygoid present; (11) retroarticular process long and usually curved dorsally; (12) third and fourth ceratobranchial fused; (13) anterior fibers of the m. interhyoideus do not insert on ceratohyal; (14) m. interhyoideus posterior in two bundles; (15) orientation of m. interhyoideus posterior is longitudinal rather than oblique; and (16) m. depressor mandibulae longitudinally oriented rather than vertically oriented.

    [10] FAMILY: ICHTHYOPHIIDAE TAYLOR, 1968

  • Epicria Fitzinger, 1843: 34. Type genus: Epicrium Wagler, 1828. Suppressed for purposes of priority but not homonymy in favor of Ichthyophiidae by Opinion 1604 (Anonymous, 1990: 166).

  • Ichthyophiidae Taylor, 1968: 46. Type genus: Ichthyophis Fitzinger, 1826. Placed on Official List of Family-Group Names in Zoology by Opinion 1604 (Anonymous, 1990: 166–167).

  • Uraeotyphlinae Nussbaum, 1979: 14. Type genus: Uraeotyphlus Peters, 1880 “1879”.

  • Immediately more inclusive taxon:

    [9] Stegokrotaphia Cannatella and Hillis, 1993.

    Sister taxon:

    [12] Caeciliidae Rafinesque, 1814.

    Range:

    India to southern China, Thailand, and through the Malayan archipelago to the Greater Sunda Islands and Philippines.

    Content:

    Caudacaecilia Taylor, 1968; “IchthyophisFitzinger, 1826 (see Systematic Comments); Uraeotyphlus Peters, 1880 “1879”.

    Characterization and diagnosis:

    Ichthyophiids are oviparous with aquatic larvae, both features being plesiomorphies. Like rhinatrematids, ichthyophiids plesiomorphically retain a true tail. Eyes are externally visible beneath the skin and are in bony sockets. The tentacle arises between the nostril and the eye, generally closer to the eye in Ichthyophis and Caudacaecilia and anterior near the nostril in Uraeotyphlus. A stapes is present (Nussbaum, 1977).

    Beyond the molecular evidence supporting the monophyly of this group the following morphological characters have been suggested to be synapomorphies (M. Wilkinson and Nussbaum, 1996): (1) vomers in contact anteriorly (convergent in Siphonops, Scolecomorphus, and Gegeneophis); (2) atria divided externally; (3) anterior pericardial sac long and extensive; (4) posterior internal flexures in the m. rectus lateralis II; (5) tracheal lung present (also in Typhlonectes).

    Systematic comments:

    As noted in “Results”, the preponderance of evidence suggests that “Ichthyophis” is paraphyletic with respect to Uraeotyphlus. Unfortunately, the number of species currently assigned to “Ichthyophis” is large and mostly unsampled, and the relationships among them (and Caudacaecilia [unsampled by us] and Uraeotyphlus) are unclear. Nussbaum and Wilkinson (1989: 31) suggested that Caudacaecilia and Ichthyophis might both be polyphyletic inasmuch as they are diagnosed solely on single characters of known variability. We do not place Caudacaecilia and Uraeotyphlus into the synonymy of Ichthyophis, although to do so would certainly render a monophyletic taxonomy. Ongoing work by M. Wilkinson, Nussbaum, and collaborators should provide a monophyletic taxonomy without resorting to that minimally informative one. In the interim we place quotation marks around “Ichthyophis” (the only ichthyophiid genus for which we have evidence of paraphyly). In the face of strong evidence of paraphyly of “Ichthyophis”, maintaining a family-group name for Uraeotyphlus is unnecessary, and we therefore place Uraeotyphlinae in the synonymy of Ichthyophiidae. Other than assuming that the morphological synapomorphies are sufficient, stronger evidence of monophyly of Ichthyophiidae will require sampling of Caudacaecilia and more “Ichthyophis”. Nevertheless, we make the hypothesis that Ichthyophiidae is a monophyletic taxon and trust that others will elucidate this further.

    [12] FAMILY: CAECILIIDAE RAFINESQUE, 1814

  • Cecilinia Rafinesque, 1814: 104. Type genus: Caecilia Linnaeus, 1758. See Dubois (1985: 70). Authorship but not spelling to be conserved following Opinion 1830 (Anonymous, 1996: 68–69).

  • Caeciliadae Gray, 1825: 217. Type genus: Caecilia Linnaeus, 1758.

  • Siphonopina Bonaparte, 1850: 1 p. Type genus: Siphonops Wagler, 1828.

  • Typhlonectidae Taylor, 1968: xi, 231. Type genus: Typhlonectes Peters, 1880 “1879”.

  • Scolecomorphidae Taylor, 1969a: 297. Type genus: Scolecomorphus Boulenger, 1883.

  • Dermophiinae Taylor, 1969b: 610. Type genus: Dermophis Peters, 1880 “1879”.

  • Herpelinae Laurent, 1984a: 199–200. Type genus: Herpele Peters, 1875.

  • Geotrypetoidae Lescure et al., 1986: 162. Type genus: Geotrypetes Peters, 1880.

  • Grandisoniilae Lescure et al., 1986: 164. Type genus: Grandisonia Taylor, 1968.

  • Indotyphlini Lescure et al., 1986: 164. Type genus: Indotyphlus Taylor, 1960.

  • Afrocaeciliiti Lescure et al., 1986: 164. Type genus: Afrocaecilia Taylor, 1968.

  • Brasilotyphlili Lescure et al., 1986: 166. Type genus: Brasilotyphlus Taylor, 1968.

  • Pseudosiphonopiti Lescure et al., 1986: 166. Type genus: Pseudosiphonops Taylor, 1968.

  • Oscaecilioidae Lescure et al., 1986: 167. Type genus: Oscaecilia Taylor, 1968.

  • Gymnopiilae Lescure et al., 1986: 168. Type genus: Gymnopis Peters, 1874.

  • Potamotyphloidea Lescure et al., 1986: 169. Type genus: Potamotyphlus Taylor, 1968.

  • Pseudotyphlonectini Lescure et al., 1986: 170. Type genus: Pseudotyphlonectes Lescure, Renous, and Gasc, 1986.

  • Immediately more inclusive taxon:

    [9] Stegokrotaphia Cannatella and Hillis, 1993.

    Sister taxon:

    [10] Ichthyophiidae Taylor, 1968.

    Range:

    Tropics of Mexico, Central America, and South America; equatorial East and West Africa and islands in the Gulf of Guinea, Seychelles, and India.

    Content:

    Atretochoana Nussbaum and Wilkinson, 1995; Boulengerula Tornier, 1896; Brasilotyphlus Taylor, 1968; Caecilia Linnaeus, 1758; Chthonerpeton Peters, 1880; Crotaphatrema Nussbaum, 1985; Dermophis Peters, 1880; Gegeneophis Peters, 1880; Geotrypetes Peters, 1880; Grandisonia Taylor, 1968; Gymnopis Peters, 1874; Herpele Peters, 1880; Hypogeophis Peters, 1880; Idiocranium Parker, 1936; Indotyphlus Taylor, 1960; Luetkenotyphlus Taylor, 1968; Microcaecilia Taylor, 1968; Mimosiphonops Taylor, 1968; Nectocaecilia Taylor, 1968; Oscaecilia Taylor, 1968; Parvicaecilia Taylor, 1968; Potomotyphlus Taylor, 1968; Praslinia Boulenger, 1909; Schistometopum Parker, 1941; Scolecomorphus Boulenger, 1883; Siphonops Wagler, 1828; Sylvacaecilia Wake, 1987; Typhlonectes Peters, 1880.

    Characterization and diagnosis:

    Caeciliids represent the bulk of caecilian diversity and, not surprisingly, show considerable morphological and reproductive variation. Some taxa are oviparous with aquatic larvae (e.g., Praslinia), whereas others are oviparous with direct development in the egg (e.g., Hypogeophis, Idiocranium, and Boulengerula), and others are viviparous (e.g., Schistometopum, Dermophis, and typhlonectines). Unlike Ichthyophiidae and Rhinatrematidae, no caeciliid possesses a true tail, although some (e.g., typhlonectines) have a pseudotail. Most species are terrestrial and burrowing, although some (e.g., typhlonectines) are secondarily aquatic. At least one species (Atretochoana eiselti: Typhlonectinae) is totally lungless (Nussbaum and Wilkinson, 1995). Most taxa have stapes, but all scolecomorphines lack them (Nussbaum, 1977).

    Beyond the molecular evidence, the following morphological characters have been suggested to be synapomorphies of this group (M. Wilkinson and Nussbaum, 1996): (1) tail absent; (2) premaxillae and nasal bones fused; (3) septomaxillae reduced or absent (reversed in Scolecomorphus); (4) pterygoid absent; (5) basiptergygoid process large (small in Scolecomorphus); (6) fused third and fourth ceratobranchials greatly expanded; and (7) vent circular or transverse, not longitudinally oriented (reversed in Scolecomorphus).

    Systematic comments:

    Recognition of the nominal families Typhlonectidae (Atretochoana, Chthonerpeton, Nectocaecilia, Potomotyphlus, Typhlonectes) and Scolecomorphidae (Crotaphatrema and Scolecomorphus) renders Caeciliidae paraphyletic. Although we expect that ongoing work by M. Wilkinson, Nussbaum, and collaborators will provide a more refined taxonomy, these currently recognized taxa can be retained as subfamilies (Scolecomorphinae Taylor, 1969, and Typhlonectinae Taylor, 1968) with no paraphyly implied as long as the remaining caeciliids are not placed within a subfamily. (A Caeciliinae recognized as nomenclaturally coordinate with Scolecomorphinae and Typhlonectinae would merely push the paraphyly to the subfamily level, as was done by Hedges et al., 1993.) Although molecular evidence corroborates the monophyly of Scolecomorphinae, the following morphological characters also diagnose that taxon (Nussbaum and Wilkinson, 1989; M. Wilkinson and Nussbaum, 1996): (1) temporal fossa secondarily large (also in Typhlonectinae, though not homologously); (2) premaxillae separate; (3) septomaxilla present; (4) prefrontals present; (5) basipterygoid process small; and (6) no stapes. Similarly, for Typhlonectinae, the following apomorphic characters diagnose that taxon (Nussbaum and Wilkinson, 1989; M. Wilkinson and Nussbaum, 1996): (1) temporal fossa secondarily large (also in Scolecomorphinae, though not homologously); and (2) choanae large, with well-developed valves.

    [23] BATRACHIA LATREILLE, 1800

  • Batrachii Latreille, 1800: xxxvii. A Latinization of Batraciens Brongniart, 1800b, emended to Batrachia by Rafinesque, 1814: 103. (See appendix 6 for nomenclatural discussion.)

  • Immediately more inclusive taxon:

    [6] Amphibia Gray, 1825.

    Sister taxon:

    [7] Gymnophiona J. Müller, 1832.

    Range:

    Cosmopolitan in cold-temperate to tropical habitats, except for extreme northern latitudes, Antarctica, and most oceanic islands.

    Concept and content:

    Batrachia is a monophyletic taxon containing [24] Caudata Fischer von Waldheim, 1813, and [74] Anura Fischer von Waldheim, 1831 (cf. Cannatella and Hillis, 1993; cf. Latreille, 1800).

    Characterization and diagnosis:

    Batrachia is a taxon whose living members of the two component groups (salamanders and frogs) are so different (and mutually apomorphic) that their synapomorphies are not obviously reflected in external appearance. The annectant members of the taxon are all fossil and not well known. For practical purposes, Batrachia is composed of living amphibians that are not members of Gymnophiona.

    Beyond our molecular evidence, the following morphological characters have been suggested to be synapomorphies of this group (Trueb and Cloutier, 1991): (1) loss of a postfrontal bone; (2) loss of the surangular bone; (3) loss of splenial bone; (4) loss of dermal scales; (5) absence of an articulation of the anterior ptergygoid ramus with the palatine; (6) absence of an ectopterygoid; (7) absence of a stapedial foramen; (8) presence of a papilla neglecta; (9) presence of a carotid labyrinth; (10) choanal tube opens into the archenteron during development; and (11) pronephros modified for sperm transport.

    Systematic comments:

    Feller and Hedges (1998) coined the name Procera (for which Homomorpha Fitzinger [1835] is an available older name) for a clade composed of salamanders and caecilians that they believed to be monophyletic. Procera was supported by analysis of 2.7 kb of sequence from four mtDNA genes. We have not attempted to reanalyze the data of Feller and Hedges (1998), but we note that we also used 12S and 16S fragments of the mt rRNA genes and tRNAValine. They also used sequences from a portion of the tRNALeucine gene, which we did not. Unlike Feller and Hedges (1998), we included substantial evidence from nuDNA sequences (see “Materials”), with the result that we have employed almost half again as much sequence as they did and more than 43 times as many terminals. Our results strongly support the relationship corroborated by morphological evidence (Trueb and Cloutier, 1991), which is caecilians + (frogs + salamanders). This arrangement, in turn, is consistent with the recognition of Batrachia Latreille (1800) and as intended by Trueb and Cloutier (1991). Furthermore, for our data alternative topologies required considerably more steps: (1) frogs + (caecilians + salamanders) required 84 additional steps; and (2) salamanders + (caecilians + frogs) solution required an additional 85 steps.

    [24] CAUDATA FISCHER VON WALDHEIM, 1813

  • Caudati Fischer von Waldheim, 1813: 58, an apparent latinization and reranking of Caudati A.M.C. Duméril, 1806: 95 (which was coined as a family-group taxon and is therefore unavailable for above-family-group taxonomy). Emended here to conform to the traditional spelling, Caudata (see Stejneger, 1907). Not Caudata Scopoli (1777), as attributed incorrectly by Stejneger, 1907: 215. (See appendix 6 for nomenclatural discussion.)

  • Immediately more inclusive taxon:

    [23] Batrachia Latreille, 1800.

    Sister taxon:

    [74] Anura Fischer von Waldheim, 1831.

    Range:

    Temperate Eurasia, northwestern Africa, and North America, and in disjunct populations throughout tropical America.

    Concept and content:

    Caudata is a monophyletic group composed of all living salamanders (cf. Cannatella and Hillis, 1993), the subsidiary taxa being [25] Cryptobranchoidei Noble, 1931, and [29] Diadectosalamandroidei new taxon.

    Characterization and diagnosis:

    Salamanders are immediately recognizable because they are the only living amphibians to have both forelimbs and tails. Their primitive aspect is restricted only to general body plan. Salamanders show many osteological losses and morphological simplifications from their non-caudatan ancestors. Unlike the other two major clades of living amphibians, whole groups of salamanders are known for paedomorphic lineages with varying degrees of retention of larval characteristics in the aquatic adults (e.g., Cryptobranchidae, Sirenidae, Proteidae, and various members of the Ambystomatidae [e.g., Amybystoma dumerilii] and Plethodontidae [e.g., Eurycea tridentifera]). Most salamanders transfer sperm via the production of spermatophores, but like frogs and caecilians, salamanders primitively have external fertilization with free-living aquatic larvae.

    Beyond our molecular evidence, Caudata is diagnosed by the following morphological characters, judged to be synapomorphies (modified from Trueb and Cloutier, 1991; Larson and Dimmick, 1993; Larson et al., 2003): (1) incomplete maxillary arcade; (2) presence of a tuberculum interglenoideum; (3) scapulocoracoid and scapula fused (reversed in sirenids); (4) no operculum and columella detached (modified in some hynobiids, plethodontids, salamandrids, and ambystomatids); and (5) male anterior ventral glands present (reversed in sirenids). In addition, Trueb and Cloutier (1991) discussed a number of other features that may be synapomorphic but are highly contingent on cladogram topology.

    [25] CRYPTOBRANCHOIDEI NOBLE, 1931

  • Cryptobranchoidea Noble, 1931: 473. Explicit order emended to Cryptobranchoidei by Tamarunov, 1964b: 159. (See appendix 6 for nomenclatural note.)

  • Immediately more inclusive taxon:

    [24] Caudata Fischer von Waldheim, 1813.

    Sister taxon:

    [29] Diadectosalamandroidei new taxon.

    Range:

    Eastern United States and southeastern Canada in North America; in Eurasia from Kamchatka west through Siberia to eastern European Russia to Turkmenistan, Afghanistan, and Iran and eastward through central China to Korea and Japan.

    Concept and content:

    Cryptobranchoidei is a monophyletic taxon composed of [27] Cryptobranchidae Fitzinger, 1826, and [26] Hynobiidae Cope, 1859.

    Characterization and diagnosis:

    Cryptobranchoidei exhibits external fertilization (one genus showing a unique kind of spermatophore formation) and other features primitive for Caudata. Although one group (Cryptobranchidae) consists of paedomorphic giants with distinctive apomorphies such as lateral folds of skin, the bulk of species (Hynobiidae) are generalized forms that are similar in many ways to the ancestral salamander.

    Beyond the molecular evidence, the following morphological characters are likely synapomorphies (Noble, 1931; Larson and Dimmick, 1993; Larson et al., 2003): (1) fusion of the m. pubotibialis and m. puboischiotibialis; and (2) ribs unicapitate (also in Anura).

    [27] FAMILY: CRYPTOBRANCHIDAE FITZINGER, 1826

  • Cryptobranchoidea Fitzinger, 1826: 42. Type genus: Cryptobranchus Leuckart, 1821.

  • Menopomatidae Hogg, 1838: 152. Type genus: Menopoma Harlan, 1825.

  • Andriadini Bonaparte, 1839: 131. Type genus: Andrias Tschudi, 1837.

  • Protonopsina Bonaparte, 1840: 101 (p. 11 of offprint). Type genus: Protonopsis LeConte, 1824.

  • Salamandropes Fitzinger, 1843: 34. Type genus: Salamandrops Wagler, 1830.

  • Megalobatrachi Fitzinger, 1843: 34. Type genus: Megalobatrachus Tschudi, 1837.

  • Sieboldiidae Bonaparte, 1850: 1 p. Type genus: Sieboldia Gray, 1838.

  • Protonopsidae Gray, 1850a: 52. Type genus: “Protonopsis Barton, 1824” (= Protonopsis LeConte, 1824).

  • Immediately more inclusive taxon:

    [25] Cryptobranchoidei Noble, 1931.

    Sister taxon:

    [26] Hynobiidae Cope, 1859.

    Range:

    Central China; Japan; eastern temperate North America.

    Content:

    Andrias Tschudi, 1837; Cryptobranchus Leuckart, 1821.

    Characterization and diagnosis:

    Cryptobranchidae is a taxon composed of three species of giant, obligately aquatic paedomorphs. Like other cryptobranchoids, they lack internal fertilization and share a suite of internal characters primitive for Caudata. Adults lack gills and the lungs are nonfunctional, so nearly all respiration is across the extensively folded and wrinkled skin (Noble, 1931; Bishop, 1943).

    Beyond the molecular evidence, the following morphological characters have been suggested to be synapomorphies (Larson and Dimmick, 1993; Larson et al., 2003): (1) dorsoventrally flattened bodies; (2) presence of folds of skin forming flaps along the lateral margins of the body; and (3) septomaxilla absent (also in some salamandrids, Amphiumidae, and Perennibranchia).

    Systematic comment:

    The monophyly of Cryptobranchidae was never seriously in doubt, but our results (appendix 5) and those of Larson et al. (2003) demonstrate that Cryptobranchus is the sister taxon of Andrias, an arrangement suggested, but not substantiated, by Estes (1981).

    [26] FAMILY: HYNOBIIDAE COPE, 1859 (1856)

  • Ellipsoglossidae Hallowell, 1856: 11. Type genus: Ellipsoglossa Duméril, Bibron, and Duméril, 1854.

  • Hynobiidae Cope, 1859: 125. Type genus: Hynobius Tschudi, 1838.

  • Protohynobiinae Fei and Ye, 2000: 64. Type genus: Protohynobius Fei and Ye, 2000.

  • Immediately more inclusive taxon:

    [25] Cryptobranchoidei Noble, 1931.

    Sister taxon:

    [27] Cryptobranchidae Fitzinger, 1826.

    Range:

    Japan, Korea, and Kamchatka west through Siberia and China to eastern European Russia to Turkmenistan, Afghanistan, and Iran.

    Content:

    Batrachuperus Boulenger, 1878; Hynobius Tschudi, 1838; Onychodactylus Tschudi, 1838; Pachyhynobius Fei, Qu, and Wu, 1983; Protohynobius Fei and Ye, 2000; Ranodon Kessler, 1866; Salamandrella Dybowski, 1870.

    Characterization and diagnosis:

    Hynobiids are unremarkable salamanders, predominantly exhibiting a biphasic life history with external fertilization and females lacking spermathecae. Lungs are usually developed, except in Onychod