CHAPTER 4 Speciation, Species Boundaries and Phylogeography of Amphibians Miguel Vences and David B. Wake 1. Introduction A. Species Concepts, Theories on IV. Genetic Criteria for Species Recognition Speciation and their Application in A. Allozyme Threshold Values Amphibians B. Mitochondrial Threshold Values B. Characters Used to Define Amphibian C. Discordance among Datasets Species D. DNA Barcoding and the Concepts of C. The Concept of Phylogeography MOTUs and Candidate Species II. Vicariant Species Formation E. Estimates of Amphibian Species Diversity A. Dichopatric versus Peripatric V. Correlates, Radiation Trajectories and B. Gene Flow, Dispersal and Amphibians on Hypothesis Testing Islands A. Correlates of Species Diversity C. Examples of Vicariant Species Formation B. Correlates of Genetic Diversity III. Adaptive Species Formation C. Trajectories of Amphibian Radiations AA.. EEccoollooggiiccaall DDiivveerrssiiffiiccaattiioonn aalloonngg CClliinneess VI. Acknowledgements aanndd TTrraannsseeccttss VII. References BB.. SSeexxuuaall SSeelleeccttiioonn:: CCoolloouurr,, CCrreessttss aanndd Calls C. Temporal Processes: Allochrony and Heterochrony D. Chromosome Rearrangements and Polyploidy I. INTRODUCTION A. Species Concepts, Theories on Speciation and their Application in Amphibians A RESULT of intense interest in species concepts during the past decade is the recognition that controversy has been focused more upon criteria for determining what species to recognize rather than upon what species, in general, are. The intellectual framework presented by de Queiroz (1998, 1999) suggested that despite the appearance of disagreement, there is fundamental agreement among the diverse definitions employed for species. De Queiroz argued that nearly all contemporary biologists accept the idea that species are segments of population-level evolutionary lineages. There remains the large question of what the defining properties of the taxonomic category "species" might be, but the primary issue is to discover when such lineages diverge and when lineages are finally split. The many species definitions — May den (1997) lists more than 20 — become 2614 AMPHIBIAN BIOLOGY criteria for species recognition under this perspective. The issue of delimiting species remains, but there can be agreement on the goal — an understanding of when evolutionary lineages have irretrievably diverged. As de Queiroz (1998) noted, there are still unresolved issues relating to whether successive species can exist in unbranched lineages, and whether asexual organisms form species, but an intellectual advance has been achieved. Delimiting species has been identified as a Renaissance issue in systematic biology by Sites and Marshall (2003), because of a number of new methods that result largely from the new kinds of data that have been introduced in recent years. This started about 30 years ago when use of allozyme data became common. Sites and Marshall (2004) dealt with twelve methods for delimiting species; there are more and no doubt new methods will arise as data generation and analysis become more sophisticated. It can be agreed, however, that there is something to be determined, establishment of permanently diverged lineages, and then use whatever evidence is available to make decisions in individual cases. A major issue is determining whether different researchers studying the same larger lineage, in the present case the clade Lissamphibia, have recognized similar entities as species. The useful reviews by Sites and Marshall (2003, 2004) contrasted population-based (their "non-tree") versus tree-based methods of determining species' status. This designation parallels what has emerged as a dichotomy concerning patterns or modes of species formation, that for the purposes of ther present review are identified as vicariant species formation and adaptive species formation. Vicariant species formation refers to the physical, usually geographic, isolation of fragments of a lineage, by whatever means, and their subsequent history of divergence. Included are the categories termed dichopatric (Bush 1994) and peripatric (Mayr 1954). Both are forms of allopatric species formation, the former referring to the splitting of a lineage into two fragments by some historical disruption of a formerly continuous geographic range, and the latter to what is sometimes called founder species formation, when a new geographic range is invaded. These two kinds of separation can lead to divergence in isolation, and tree-based methods typically are employed to determine whether such fragments should be recognized as species. Adaptive species formation refers to the situation in which spatially adjacent or overlapping populations have diverged while maintaining the possibility of genetic interaction. Included are parapatric and sympatric modes, and also hybrid species formation. Typically, population- based methods are used to determine whether the diverging units are separate populations or species; criteria include estimation of the degree of genetic exchange taking place. There is reason to think that vicariant and adaptive species formation might both be functioning in a given lineage experiencing continuing episodes of isolation and recontact. This kind of mixed perspective appears to cause the greatest controversy among systematists. In the end, the most direct evidence evinced in support of species-status is when populations of what are determined as lineages occur in sympatry. When there is hybridization in such circumstances controversy can and does arise. Hybridization is well known and much studied in amphibians, and often leads to differences in how species are determined. Litdejohn and Watson (1993) reported that hybrid zones are known in one-third of southeastern Australia's frog genera. In contrast to mammals (but not to other animal groups, e.g., birds), hybridizable species of amphibians can be relatively old, averaging an estimated 21 million years (Wilson et al. 1974; Prager and Wilson 1975). Accordingly, the observation of a natural hybridization zone between two species is not necessarily indicative of a recent speciation event. An explicit analysis of the frequency of modes of speciation in vertebrates was made by Lynch (1989). He presented a method for determining which of three modes of species formation had occurred: vicariant, peripheral isolate (these two he recognized as ends of a continuum and both fall into the present vicariant category) and sympatric (largely corresponding to the present adaptive category). Detailed information on geographic distri bution and robust phylogenetic hypotheses are necessary elements of his method. Lynch VENCES and WAKE: SPECIATION, SPECIES BOUNDARIES, PHYLOGEOGRAPHY OF AMPHIBIANS 2615 identified his assumptions and presented case examples that are considered later in this review. His pioneering effort anticipated the explosion of phylogeographic information resulting from techniques that readily generate large amounts of data on DNA sequences. B. Characters Used to Define Amphibian Species In the past, morphology was often the sole criterion used to determine species' status, and it remains the most general criterion. When two distinctive morphs, identifiable by at least two unrelated characers, occur in a single population with age and sex taken into account, they are considered species. Bioacoustics long has been employed as a criterion for frogs, and ever more sophisticated means of discrimination are employed (Schneider and Sinsch Chapter 8 this volume). The idea is that if two distinctive call types are found in a single population and when size of caller and such environmental variables as temperature can be excluded as explanations for the differences, two species are assumed to be involved, even if no morphological differences are detected. In most cases of cryptic species detected by bioacoustics in the tropics, the differences in call structure are distinct, easily recognizable by a trained observer, and leave little doubt that two reproductively isolated species are present. Starting about 1970, allozymes were employed to determine whether a population included two or more genetic subpopulations. Allozymes have been used to determine the status of species, e.g., Plethodon dorsalis was found to be sympatric with a greatly diverged form that resembled it closely in morphology, and that subsequently was named Plethodon websteri (Highton 1979). Allozymes have also been used to show the invalidity of taxa previously recognized as separate species on morphological grounds, e.g., Plethodon gordoni was shown not to differ allozymically from sympatric Plethodon dunni, from which it differed in coloration (Feder et al. 1978). Populational (non-tree based) methods (Sites and Marshall 2003, 2004) depend largely on genetic distances inferred from allozymic or microsatellite differentiation among populations. For many taxa and regions where populational data are unavailable because of low sample sizes, however, pairwise distances between individuals based on DNA sequences of mitochondrial genes (mtDNA, apparently strictly maternally inherited in amphibians) increasingly are used for species discrimination (section IV-B). Such data are primarily used, however, to construct phylogenetic hypotheses of the relationships of species and segments of species and to assess species' boundaries. In "total evidence" approaches (Wiens and Penkrot 2002) tree-based methods include other kinds of data as well. MtDNA haplotype clades recovered using established algorithms show relatively rapid coalescence and thus provide much useful data that can be interpreted historically (Avise 2004). Haplotype trees are often interpreted as if they are decisive in finding clades that when phylogeographically distinct are recognized as species. Single populations, however, can share two or more haplotypes, and presumptive species can be composites of paraphyletic or even polyphyletic haplotype lineages (Funk and Omland 2003). Sites and Marshall (2004) concluded that all methods will sometimes fail to delimit species' boundaries properly, and virtually all methods will require researchers to make qualitative judgements. In a well- studied example of closely related species of fruit flies (Drosophila), Machado et al. (2002) and Machado and Hey (2003) concluded that a limited amount of gene flow can continue even after completion of speciation, and that simple bifurcating trees may in some cases be unable to reflect the complex history of species formation. Due to the many combinations of deterministic and stochastic processes associated with speciation, such fuzzy boundaries are unavoidable (Sites and Marshall 2004). This circumstance warns against oversimplified approaches to delimitation of species. C. The Concept of Phylogeography Since its formalization as a distinct field (Avise et al. 1987), phylogeography has been closely related to systematics. The key element in phylogeography is the development of 2616 AMPHIBIAN BIOLOGY intraspecific phylogenies, based mainly on mitochondrial DNA. Taken together with conceptual advances in population genetics (Avise 2000, 2004), modern phylogeography has greatly enhanced the abilities of systematists to function by facilitating the identification of historial population-level units within what have been recognized as species. Phylogeographic patterns are important elements of systematic analyses and increasingly there is a focus on the search for concordant patterns of distribution of allozymes, morphology and mitochondrial DNA haplotypes. Very large databases now exist that permit comparisons of different kinds of data with respect to patterns of geographic variation. When morphology, allozymes and mtDNA haplotypes show congruent patterns (e.g., Jockusch et at. 2001), determination of species seems straight-forward. Such patterns are central to understanding vicariant elements of species formation. When, however, incongruence is observed, as is becoming increasingly common (Funk and Omland 2003), the old problems of making determinations of species' status for geographically isolated populations return. While mtDNA is often treated as if it were a neutral marker, Ballard and Whitlock (2004) argued that adaptation often influences the course of its evolution, and they caution against many of the assumptions made by systematists. II. VICARIANT SPECIES FORMATION A. Dichopatric versus Peripatric A subtle shift in emphasis with respect to vicariant species formation has taken place in recent years. Earlier, range subdivision, now frequently termed dichopatric species formation, was emphasized, but founder effect, or peripatric, species formation has gained much attention. The former is based on the idea that a once continuous range is disrupted and some barrier to dispersal forms. Latent in the framework of this conception is the idea that species are to some degree maintained as entities by gene flow. When gene flow is disrupted, now isolated populational segments are free to diverge under the influence of natural selection as well as by neutral factors that require either small populational sizes, much time, or both. Eventually, when the two groups are reunited, a sympatry test occurs, and either the units are now fully independent or they demonstrate some degree of reproductive compatibility, ranging from production of sterile hybrids grading to situations in which essentially free interbreeding leads to merger of the once distinct units. Of course, in many instances the units have diverged to some degree and no sympatry test has happened. In such situations different criteria of independence of lineages are used, and controversy is inevitable. Wiens (2004a) argued that geographically isolated segments of species become species when populations are unable to penetrate marginal, peripheral environments. He also argued that the failure of populations to adapt to ecological conditions at the edge of the species' range is the fundamental issue in geographic isolation and thus sets up lineage-splitting, which is tantamount to species formation when accompanied by some degree of differentiation (how much and what kind remains controversial). In other words, dichopatric species formation places an emphasis on isolation where parapatric species formation focuses on adaptive divergence. Peripatric species formation contrasts with dichopatric in that previously unoccupied regions are invaded by small numbers of migrants and small populations are established that start with reduced variability and then, because they are small, experience ongoing genetic drift. Without going into detail concerning this somewhat more controversial mode of vicariant species formation, one can simply ask which pattern is most frequently encountered in amphibians. Using molecular markers, Avise (2004) has given criteria for determining which mode of speciation has occurred. Available data strongly suggests that most species formation in amphibians is vicariant, principally dichopatric. The most common evidence is the parapatric or allopatric distribution patterns of extant sister species (e.g., Roberts and Maxson 1985; Watson and Littlejohn 1985; Lynch 1989). Lynch (1989) listed the assumptions that he made for vicariant species formation as (1) an ancestral cosmopolitan distribution, (2) geographic ranges of sister lineages VENCES and WAKE: SPECIATION, SPECIES BOUNDARIES, PHYLOGEOGRAPHY OF AMPHIBIANS 2617 juxtaposed and not overlapping, and (3) sister species with ranges of equivalent sizes. Distributional gaps between sister taxa result from extinction, and sympatry requires dispersal. His peripheral isolate mode (peripatric) assumed distributions of sister lineages as well to be are juxtaposed and not overlapping, but ancestral distributions are not cosmopolitan, and the sister lineages have geographic ranges that differ markedly in size. The parental species has a large distribution. Distributional gaps are expected because the peripheral isolate resulted from dispersal. Sympatry accordingly requires more dispersal. Lynch's sympatric model assumed ancestral distributions that were not cosmopolitan and geographic ranges of sister taxa that overlap, but there were no expectations about sizes of geographic ranges, distributional gaps, or dispersal. For 66 cases analysed (three genera of frogs — Eleutherodactylus, Ceratophrys, and Rana — three of cyprinodontiform fishes, and one of passerine birds) his vicariant mode explained 71% of the variation, the peripheral isolate mode 15%, and the sympatric mode 6%. For the Rana dataset (from a resolved cladogram of 23 species in the leopard-frog complex [Hillis et al. 1983]), Lynch estimated that geographic-subdivision species formation accounted for 72.7% of the 22 speciation events, peripheral isolates for 4.9-9.2%, and sympatric for 4.6-18.2%. As Lynch noted, these results are surprising because of the recent emphasis on peripheral isolates and sympatric modes. The Rana pipiens complex has long been recognized to be especially difficult taxonomically; whether the case is unusual is difficult to determine because few amphibian clades have been studied so thoroughly (the green-frog complex of Europe is another group of Rana that is well studied, but it differs in its unique sexual system, see below). For Ceratophrys (phylogenetic hypothesis from Lynch [1982]), five of the six speciation events were attributed to vicariance, and one to peripheral isolates. For Eleutherodactylus (phylogenetic analysis from Miyamoto [1983]), five of eight events (62.5%) were attributed to vicariance, one to peripheral isolates, one to sympatric differentiation, and one was indeterminate. Lynch, concerned with the then-prevalent biogeographic assumption of nondispersal, concluded that dispersal was unnecessary (e.g., because of habitat alterations during the Pleistocene) and not a logical necessity. While the analyses of Lynch are informative, they are strongly dependent on threshold values in the proportion of overlap between sister species, or on relative range sizes (Fig. 1) (Chesser and Zink 1994), and thereby on phylogenetic and distributional knowledge. Phylogenetic hypotheses are being generated at rapid rates but many of the new species being described from tropical regions are known from only a single or narrowly restricted locality. Examination of extensive phylogeographic, phylogenetic and distributional databases that currently exist suggests that the general conclusions of Lynch are valid and that the vast majority of amphibian speciational events probably were vicariant by means of geographic subdivision (see also Wiens 2004a). Considering the wealth of phylogenetic data on amphibians now available and the improved knowledge of the distribution of many species, Lynch's methodology can be applied to further datasets, and its combination with other approaches such as modelling of environmental niches (Graham et al. 2004) seems promising. An important perspective is also the inclusion of genetic distances, which can be used to infer the ages of nodes and thereby to use plots of range overlap through evolutionary age of nodes (Fig. 2) to make inferences about the geographic mode of species formation. B. Gene Flow, Dispersal and Amphibians on Islands In the typical dichopatric mode of species formation the critical issue is cessation of gene flow, which is related to dispersal capability. Amphibian species typically show high levels of geographic variability in molecular markers (see also below, section V). Very few amphibian species give evidence of species-wide gene flow. Within direct-developing species in particular, differentiation in situ most probably proceeds through a combination of neutral and selective factors operating over long spans of time. This may explain why there are more than 25 species of Thorius in a small part of eastern Mexico, more than 140 species of Eleutherodactylus in Ecuador, and more than 60 species of direct-developing frogs in 2618 AMPHIBIAN BIOLOGY b ed #^ ^1^ ^s ^ 2 13 3 12 1 2 3 1 V \/ 1 \ \/ \ \/ N \ V "***N XX Xir Sympatnc speciation X X JXr Dichopatric speciation ||&*<afjE- X / Peripatric speciation ] } >^ Dichopatric speciation ^ Sympatric speciation *r Sympatric speciation '-otSr^'. \S"* f (30% criterion) g (90% criterion) h (5% criterion) 12 3 12 3 1 2 3 \ \/ \ \/ N V\ \ V / \^ f Sympatric speciation X^ ^Dichopatric speciation ISMII1I11P> . X 7 Sympatric speciation v. 3 / Dichopatric speciation ^r Dichopatric speciation '' Sympatric speciation /. Inferring the mechanism of speciation using interspecific phylogenies and current distributions. Partly after Lynch (1989) and Losos and Glor (2003). Given the distributional patterns in (a), the first row shows the influence of phylogeny on the assumed speciation mode, (b) Sympatric speciation is assumed for the origin of 1 and 3 because they are sister taxa and their distributions overlap. Dichopatric speciation is assumed for the split between 2 and 1-3 because the distributional area of 2 does not overlap with the merger of the distributional areas of 1 and 3. (c) Dichopatric speciation is assumed for the origin of 1 and 2 because they are sister taxa and their distributions are of similar size and do not overlap. Sympatric speciation is assumed for the split between 3 and 1-2 because the distributional area of 3 overlaps with merged areas of 1 and 2. (d) Peripatric speciation is assumed for the origin of 2 and 3 because they are sister species, their ranges do not overlap and the range of one is much smaller than the range of the other. Sympatric speciation is assumed for the split between 1 and 2-3 because the distributional area of 1 overlaps with merged areas of 2 and 3. Note, however, that there might be a contradiction in this simplistic reasoning because the overlap between 1 and 2-3 results from a presumable dispersal of 3 subsequent to the origin of 1. Given the distributional patterns in (e), the second row shows the influence of the definition of range overlap on the inferences of mode of speciation. (f) If an overlap of 30% of a species' distribution with that of a second species is defined as the cutoff, then taxa 2 and 3 are considered to have overlapping ranges and to have originated in sympatry, but the small overlap between taxon 1 and lineage 2-3 is not considered and dichopatric speciation is assumed, (g) If a very strict criterion is employed and overlap is only accepted if the ranges of two taxa coincide by 90%, then dichopatric speciation is inferred for both splits, (h) If an hypothetical relaxed criterion of 5% is employed, then even the small overlap in range between taxon 1 and lineage 2-3 would be considered and sympatric speciation inferred for both splits. Sri Lanka (Manamandra-Arachi and Pethiyagoda 2005; Meegaskumbura et al. 2002; Meegaskumbura and Manamandra-Arachi 2005). What is important is spatial heterogeneity and long spans of time, because most of these species have narrowly restricted distributions and appear to be old (based on molecular divergence levels, e.g., Crawford 2003a). European populations of Salamandra salamandra range over most of the continent, from Spain to Greece, and probably recolonized Central Europe after the last glaciation. The occurrence of specific east-European and west-European haplotypes as well as allozymic alleles suggests that this recolonization originated from at least two source populations, probably the Iberian peninsula and the Balkans (Steinfartz et al. 2000). Two divergent populations in northern Spain (S. s. bernardezi) and southern Italy (S. s. gigliolii) surprisingly were sister to each other. These probable remnants from a previous colonization event seem to have maintained their separate genetic identity even though they are not separated by geographic barriers from closely related neighbouring populations (Steinfartz et al. 2000). This example illustrates how gene flow and admixture, at least of mitochondrial haplotypes, can be restricted even among neighbouring and closely related amphibian populations. Vicariant or peripatric speciation is obvious for amphibian species endemic to islands. Amphibians long have been thought to be unable to disperse over the sea but now there is compelling evidence, at least for frogs (Hedges et al. 1992; Vences et al. 2003b). Amphibians show near absence on true oceanic islands, with Mayotte, Sao Tome and Principe (with endemic frogs, and Sao Tome with a probable endemic caecilian; Measey VENCES and WAKE: SPECIATION, SPECIES BOUNDARIES, PHYLOGEOGRAPHY OF AMPHIBIANS 2619 a Allopatric — infrequent shifts b Sympatric — infrequent shifts Q. Q_ 1° i_ CD CD > > o O <D CD CO CO C CO <xcRxxo co JCX> -O- evolutionary age evolutionary age Allopatric — frequent shifts d Sympatric — frequent shifts Q. JP. % C>D CD > O O CD CD CD CO 8 c c CO CO to- r\ LV-W- evolutionary age evolutionary age Fig. 2. Inferring the mechanism of speciation using range overlap, phylogenetic relationships and evolutionary age (as roughly equivalent to genetic differentiation). The graphs show hypothetical examples obtained by modelling under the assumption that range movements occur by large-scale shifts of entire species' ranges. Schematically redrawn from Barraclough and Vogler (2000). Note that when frequent range shifts occur the patterns produced by the two speciation modes cannot be reliably distinguished. et al. 2006) being rare exceptions (Schatti and Loumont 1992; Vences et al. 2003). Molecular data (e.g., Hedges et al. 1992; Vences et al. 2003b, 2004b) have revealed that overseas dispersal in frogs may be more common than previously thought, and this explanation probably applies to many of the occurrences of endemic species on Caribbean and Pacific islands (Fiji and Palau, and islands in the Sunda Region and Philippines) (see also Kaiser et al. 1994). Despite this change of paradigm there is little doubt that amphibians belong to the lower end of the spectrum of relative dispersal ability (Inger and Voris 2001; Brown and Guttman 2002). Salamanders on islands are usually of the same species as on the adjacent mainland. A rare exception is Oedipina maritima from Isla Escudo de Veraguas (Garcia-Paris and Wake 2000), an island in the Caribbean Sea off the coast of Panama that is also the home of an endemic species of sloth. Another exception is Batrachoseps pacificus, endemic to the northern Channel Islands in the Pacific Ocean off southern California (Jockusch et al. 2001). On the small Atlantic islands of Cies and Ons, off the Galician coast in northwestern Spain, separated from the mainland only for about 9 000-7 000 years, populations of Salamandra salamandra occur that are morphologically divergent from the mainland populations by being slightly smaller and with a reduced yellow pattern on Ons (Galan Regalado 2004). These salamanders, however, appear to be genetically rather similar to those from the 2620 AMPHIBIAN BIOLOGY mainland (S. Steinfartz, pers. comm.) and therefore provide a good example of how rapidly different morphologies can evolve under fully allopatric conditions. A contrast to these examples is the presence of nonendemic salamanders and caecilians on Isla Gorgona, an island in the Pacific, west of Buenaventura, Colombia (Brame and Wake 1963). The generalization that amphibians are usually poor dispersers and highly philopatric (Blaustein et al. 1994), with a limited osmotic tolerance (Balinsky 1981), is correct, and these conditions affect both pattern and process of species formation prevalent in this group. Their populations in most cases show a strong phylogeographic structuring (Avise 2000); this has been found in studies such as those of Alexandrino et al. (2000, 2002), Austin et al. (2002, 2004), Babik et al. (2004), Barber (1999a, 1999b), Bos and Sites (2001), Burns et al. (2004), Chiari et al. (2005, 2006), Church et al. (2003), Crawford and Smith (2005), Donnellan et al. (1999), Garcia-Paris and Jockusch (1999), Goldberg et al. (2004), Green et al. (1996), Hoffman and Blouin (2004), Jaeger et al. (2005), Johansson et al. (2006), Masta et al. (2003), Macey et al. (2001), McGuigan et al. (1998), Mulcahy et al. (2000), Nielson et al. (2001), Noonan and Gaucher (2005), Riberon et al. (2001), Rowe et al. (1998), Rowe et al. (2005), Schneider et al. (1998), Shaffer et al. (2000, 2004), Shaffer et al. (2000), Steele et al. (2005), Symula et al. (2003), Vieites et al. (2006), Zeisset and Beebee (2001), and many others reviewed in more detail in the following sections. Slade and Moritz (1998) observed that introduced populations of Bufo marinus were uniform in their mitochondrial sequences, validating interpretations of phylogeographic structure as indicative of autochtonous status (Vences et al. 2004b). Lougheed et al. (1999) found that gene flow in an Amazonian poison frog, Epipedobates femoralis, was not relevantly hindered by major rivers but the phylogeographic structure instead reflected ancient ridges, no longer evident on the landscape. Funk et al. (2005) observed in the Columbian spotted frog Rana luteiventris, based on microsatellite data, that mountain ridges had the effect of strongly reducing gene flow among populations situated on either site of the ridge. In addition, they also found reduced gene flow among low-elevation and high-elevation populations, and a reduced genetic variation at high elevations. Their "valley-mountain" model of population structure suggests two ways in which adaptive and non-adaptive influences can interact in affecting genetic structure and differentiation at small spatial scales: (1) Differentiation (that is presumably non-adaptive) may occur following a restricted dispersal across physical barriers (ridges). (2) Adaptive differentiation may occur in frogs dispersing across an elevational gradient; their survival rates after dispersal may be lower because of missing adaptations to local ecological conditions or they may have lower rates of reproduction because of elevational differences in breeding phenology. Newman and Squire (2001), on the basis of a literature survey on fine-scale population structures in amphibians, and their own microsatellite data on wood frog populations (Rana sylvatica), concluded that in general amphibian populations exhibit a high degree of spatial structure, particularly when inter-populational distances exceed several kilometres. At the finest scales (<l-2 km) populational differences were not predictable unless some barrier to dispersal were present. At this scale, the authors found similar allele frequencies suggestive of high gene flow; they emphasized the possible importance of extinction- recolonization founder events, driven, for example, by periodic drying of wetlands, in contributing to the development of genetic subdivision by increasing the rate of stochastic fluctuation in allele frequencies. Palo et al. (2004b) studied the geographic subdivision of populations of the widespread frog, Rana temporaria, using microsatellite markers. They observed a high level of substructuring even in northern Fennoscandia, a region that was presumably colonized less than 10 000 years ago following the last glaciation. They suggested that processes other than restricted dispersal capacity needed to be explored to explain the high degree of populational subdivision in amphibians. Nevertheless, compared to other areas, the genetic differentiation among these populations is rather low (as is evident from their near-identical mitochondrial haplotypes; Palo et al. [2004b]), and this species provides one out of many examples that amphibians are capable of very rapidly moving into newly available space. In fact, many species essentially followed the VENCES and WAKE: SPECIATION, SPECIES BOUNDARIES, PHYLOGEOGRAPHY OF AMPHIBIANS 2621 glacial fronts northward at the end of the Pleistocene, establishing new populations of low genetic differentiation (e.g., Larson et al. 1984), although the mitochondrial and nuclear genetic signatures of these processes are not always fully concordant in space (see section IV below). Despite some complications (see Marsh and Trenham 2001), amphibian spatial dynamics resemble classical metapopulation models, with subpopulations in breeding waters blinking in and out of existence, and extinction-recolonization regularly taking place through stochastic or deterministic processes. Case studies of introduced frog populations have shown that low numbers of individuals can found viable populations in which nuclear genetic variability is not significantly reduced and bottleneck effects and subsequent demographic explosions are sometimes not, or only weakly, detectable, although genetic differentiation in terms of microsatellite allele frequencies is apparent (e.g., Rowe et al. 1998; Estoup et al. 2001; Zeisset and Beebee 2003). Several other population genetic studies have shown that genetic differentiation proceeds with increasing physical barriers to gene flow (e.g., Hitchings and Beebee 1997, 1998; Gibbs 1998). To understand the driving forces of speciation under fully vicariant (mostly peripatric) conditions in amphibians, the comparison of recent insular endemics with their closest relatives, once these have been identified, appears to be a fruitful field of study. Gene flow from insular populations, whether they originated by dispersal or vicariance, to their mainland relatives would almost certainly be interrupted, and insular populations of organisms, in general, tend to have reduced genetic diversity (Frankham 1997). They may also be subject to faster rates of molecular evolution due to their smaller population size (Johnson and Seger 2001). C. Examples of Vicariant Species Formation In species with less radical geographic isolation than true of insular species, the patterns of variation are usually more complex. Determining the species-status of geographic segments of clades is a major challenge. Geographic variation and allopatry pose difficulties for all the characters typically used to determine species-status. Comparative studies of modes and patterns of species formation necessarily assume that all species are equivalent, so critera for species recognition are important. Yet, even for well studied groups there continue to be difficulties. Important components of the debate over recognition and delineation of species are the patterns of geographic variation and their interpretation. Following are a few examples showing how different authors have dealt with this problem. The findings from many phylogeographic studies of amphibians show that the distributions of haplotype clades are almost entirely exclusive, with relatively little geographic overlap even within a species (but see below). The studies of Mayr (1963) and many others, which led to the view that vicariance species formation is the dominant mode for terrestrial vertebrates were based on the existence of geographic variation frequently interpreted as subspecific. While the use of subspecies has been largely abandoned (but see below), many of the populational units originally so designated subsequently have been raised to the specific level, following an argument that geographic units are also genealogical units (as is the case with phylogeographic units) and are to some degree incipient species. Cracraft (1997) argued that such basal diagnosable units are effective functional equivalents of evolutionarily significant units and that they should be formally named as species. Certainly such units would have many advantages over traditionally recognized species, but the main issue is what constitutes a population-level unit. Discordances between datasets show that there is no absolute criterion for what constitutes a species. Nevertheless, the current trend is to use fine levels of diagnosability in recognizing amphibian species. Two plethodontid salamander genera that have been studied in detail, Batrachoseps and Plethodon, are examples of vicariant species formation. Batrachoseps contained two species as recently as 1954 (Hendrickson 1954), but subsequently the widespread Californian taxon, B. attenuatus, has been extensively subdivided into the following species: B. attenuatus, 2622 AMPHIBIAN BIOLOGY B. gavilanensis, B. luciae, B. incognitus, B. minor, B. major, B. pacificus, B. nigriventris, B. gregarius, B. diabolicus, B. relictus, and B. simatus (Jockusch and Wake 2002). In addition, new geographically isolated species have been found: B. campi, B. robustus, B. regius, B. kawia, B. stebbinsi and B. gabrieli. At least two species (B. gregarius and B. major) contain haplotypic lineages that are not sister, but in general the distribution both of species and haplotype clades are either non-overlapping or overlap only narrowly. Only B. nigriventris, which co- occurs with seven species, is broadly sympatric with others, and it is invariably the smallest or most attenuate member of sympatric pairs. Species of Plethodon once considered to be widespread have been broken into many species, usually with minimal or no overlap, and with frequent zones of hybridization where species interact in what are presumed to be zones of secondary contact (e.g., Highton and Peabody 2000). For example, species once diagnosed by morphological differences show geographically distinct patterns of genetic differentiation, as measured using allozymes. The species known as P. jordani, a polytypic species with many montane isolates, now is broken into seven geographically distinct species, and P. glutinosus now contains 16 species. Some of these species are joined by regions of genetic exchange considered by Highton (e.g., 1998, 2000) to be hybrid zones but by others (Petranka 1998; Wake and Schneider 1998) to be regions of genetic interchange because the so-called hybrid zones are wide in relation to dispersal distances of individuals and the two parental types are not sympatric. This complicated system involves many parapatric and sympatric occurrences, but sympatry is between populations formerly included in P. jordani and P. glutinosus. There is limited sympatry between members of the more traditional species. Plethodon glutinosus and P. aureolus co-occur at one locality (and were once confused). Plethodon hentucki and P. glutinosis are sympatric over most of the range of P kentucki, but some hybridization occurs. The Plethodon cinereus group is also complex, but in it there are a series of geographically distinct species, some of which have very narrow parapatric contacts, and one species, P. cinereus, that is more widely distributed and with some areas of sympatry with others. Plethodon cinereus may have competitively displaced other members of its group, leading to their geographic restriction. Alternative interpretations for both Batrachoseps and Plethodon would recognize some of the parapatric assemblages of species as portions of a geographically fragmented species- complex, as is the current case with Ensatina (see below). Petranka (1998) offered an alternative interpretation of Highton's taxonomy of the P. glutinosus complex. He considered only two of Highton's 15 newly recognized species to have reached the specific level: P. aureolus and P kentucki (sympatric with other members of the complex and with very limited or no hybridization). A third species, P. teyahalee (Petranka used an alternative and invalid name, P. oconaluftee) was tentatively recognized as a species because at one site at least it is sympatric with another form without hybridization; it does hybridize, however, with P. glutinosus {sensu lato) in many areas. Thus, Petranka considered 12 species to be subjective synonyms of P. glutinosus. This extreme view was ignored by Highton (2000), whose manuscript may have been prepared prior to the appearance of Petranka's (1998) book. Later, a committee of herpetologists (Crother et al. 2000) endorsed Highton's taxonomy. If Petranka's taxonomy were adopted, the extensive genetic interactions among members of the complex would be viewed as indications that geographic fragments of a widespread ancestral form were still capable of genetic interaction and had not crossed an irreversible genetic barrier. Species of Batrachoseps display allopatric or parapatric distributions. An example is the B. pacificus complex, which includes genetically distinct groups of populations that Yanev (1980) considered to be "semi-species" on the basis of allozymes, but which she did not name. When Jockusch et al. (1998, 2001) found mtDNA to have congruent distributions, Yanev's semi-species were recognized as full species and several more species recognized. Four of these species have closely contiguous, parapatric distributions in central coastal California: B. gavilanensis, B. luciae, B. incognitus and B. minor. These morphologically similar (but not identical) forms are distinct allozymically, and have mtDNA haplotypes that are