TY - JOUR
AU - Schoch, Rainer R.
AB - Abstract Among the diverse clade of Paleozoic dissorophoid amphibians, the small, terrestrial amphibamids and the neotenic branchiosaurids have frequently been suggested as possible antecedents of either all or some of the modern amphibian clades. Classically, amphibamids and branchiosaurids have been considered to represent distinct, but closely related clades within dissorophoids, but despite their importance for the controversial lissamphibian origins, a comprehensive phylogenetic analysis of small dissorophoids has thus far not been attempted. On the basis of an integrated data set, the relationships of amphibamids and branchiosaurids were analyzed using parsimony and Bayesian approaches. Both groups represent miniaturized forms and it was tested whether similar developmental pathways, associated with miniaturization, lead to an artificial close relationship of branchiosaurids and amphibamids. Moreover, the fit of the resulting tree topologies to the distribution of fossil taxa in the stratigraphic rock record was assessed as an additional source of information. The results show that characters associated with a miniaturized morphology are not responsible for the close clustering of branchiosaurids and amphibamids. Instead, all analyses invariably demonstrate a monophyletic clade of branchiosaurids highly nested within derived amphibamids, indicating that branchiosaurids represent a group of secondarily neotenic amphibamid dissorophoids. This understanding of the phylogenetic relationships of small dissorophoid amphibians provides a new framework for the discussion of their evolutionary history and the evolution of characters shared by branchiosaurids and/or amphibamids with modern amphibian taxa. amphibamids, Bayesian analysis, branchiosaurids, dissorophoids, fossil taxa, Lissamphibia, miniaturization, parsimony analysis, Paleozoic, stratigraphic consistency The small-bodied branchiosaurids and amphibamids are among the best known temnospondyl amphibians of the Paleozoic and are of particular interest because both groups have frequently been suggested as close relatives of some or all of the modern amphibian clades. The origins and relationships of frogs, salamanders, and the limbless caecilians of the neotropics are very controversial and represent one of the most debated problems in vertebrate systematics (Milner 1988, 1993; Trueb and Cloutier 1991; Laurin and Reisz 1997; Ruta et al. 2003; Schoch and Milner 2004; Vallin and Laurin 2004; Lee and Anderson 2006; Anderson 2007; Carroll 2007; Anderson, Reisz, et al. 2008). Three hypotheses are currently discussed in the literature (Fig. 1a-c). The first one is that of a polyphyletic origin with amphibamids as the closest relatives of frogs, branchiosaurids as the closest relatives of salamanders, and caecilians nested within lepospondyls and therein most closely related to the microsaur Rhynchonkos (Carroll 2007). The other two hypotheses both propose a monophyletic Lissamphibia, which is either the sister taxon of lysorophian lepospondyls (Laurin and Reisz 1997; Vallin and Laurin 2004) or variably of branchiosaurids or amphibamids within temnospondyls (Bolt 1969, 1991; Milner 1988, 1993; Trueb and Cloutier 1991; Ruta et al. 2003; Schoch and Milner 2004; Anderson 2007; Anderson, Reisz, et al. 2008). All recent molecular analyses support a monophyly of the three modern groups. Therein the intrarelationships of lissamphibians vary between the Procera hypothesis with salamanders + caecilians as sister taxon to frogs (Hedges and Maxson 1993; Hay et al. 1995; Feller and Hedges 1998) and the Batrachian hypothesis with salamanders + frogs as the sister taxon to caecilians (Zardoya and Meyer 2001; San Mauro et al. 2005; Zhang et al. 2005) (Fig. 1d,e). FIGURE 1. View largeDownload slide Competing phylogenetic hypotheses for the relationships of lower tetrapods and the origin of lissamphibians (a, b, c) and the intrarelationships of the three modern amphibian forms (d, e). a) Polyphyly hypothesis; b) Temnospondyl hypothesis; c) Lepospondyl hypothesis; d) Batrachia hypothesis; e) Procera hypothesis. FIGURE 1. View largeDownload slide Competing phylogenetic hypotheses for the relationships of lower tetrapods and the origin of lissamphibians (a, b, c) and the intrarelationships of the three modern amphibian forms (d, e). a) Polyphyly hypothesis; b) Temnospondyl hypothesis; c) Lepospondyl hypothesis; d) Batrachia hypothesis; e) Procera hypothesis. Despite their importance for the discussion of the origins of the modern amphibian groups from within the diverse Paleozoic taxa, phylogenetic assessments of amphibamids on the one hand and branchiosaurids on the other hand have only very recently been attempted (Schoch and Rubidge 2005; Schoch and Milner 2008). Amphibamids are known from Carboniferous and Permian paleoequatorial deposits from North America and Europe and represent small, terrestrial forms (Fig. 2f). Schoch and Rubidge (2005) published the first phylogenetic analysis of amphibamids and for the first time formally established the monophyly of the group. On the basis of their phylogenetic hypothesis, they furthermore discussed several aspects of the evolutionary history of the clade. Subsequently, a number of new amphibamid taxa were discovered and included in this data set with some modifications or extensions (Huttenlocker et al. 2007; Anderson, Henrici, et al. 2008; Anderson, Reisz, et al. 2008; Fröbisch and Reisz 2008). FIGURE 2. View largeDownload slide Relationships of higher temnospondyls (Euskelia, Yates and Warren, 2000) depicting the morphology and relative size of taxa. a) the eryopid Onchiodon; b) the zatracheid Acanthostomatops; c) the basal dissorophoid Micromelerpeton; d) the dissorophid Broiliellus (redrawn from Carroll, 1964); e) the branchiosaurid Apateon pedestris; f) the amphibamid Doleserpeton (redrawn from Bolt 1969). Scale bars equal 1 cm. FIGURE 2. View largeDownload slide Relationships of higher temnospondyls (Euskelia, Yates and Warren, 2000) depicting the morphology and relative size of taxa. a) the eryopid Onchiodon; b) the zatracheid Acanthostomatops; c) the basal dissorophoid Micromelerpeton; d) the dissorophid Broiliellus (redrawn from Carroll, 1964); e) the branchiosaurid Apateon pedestris; f) the amphibamid Doleserpeton (redrawn from Bolt 1969). Scale bars equal 1 cm. Branchiosaurids are small, gill-bearing amphibians known from hundreds of specimens and a large number of ontogenetic stages from the Permo-Carboniferous lake deposits of central Europe (Fig. 2e). Despite their excellent fossil record, a phylogenetic analysis of the group was not attempted until very recently because of the prevailing neotenic nature of almost all representatives of Branchiosauridae (Schoch and Milner 2008). Neoteny, the retention of larval somatic features in adult animals, is known to be associated with a large number of homoplasies in modern urodeles and therefore has strong effects on the assessment of phylogenetic relationships (Wake 1996; Wiens et al. 2005; Schoch and Milner 2008; Schoch 2009). An understanding of the relationships of small dissorophoids, that is, amphibamids and branchiosaurids, is essential for our understanding of the evolution of these groups and ultimately for the continuing discussion on the origins of frogs, salamanders, and caecilians. The specific relationships of amphibamids and branchiosaurids have thus far not been addressed in detail because individual phylogenetic analyses for each of the two groups were lacking. Branchiosaurids and amphibamids have previously been incorporated into large data set that focused on lower tetrapod relationships on a broad scale but were therein only represented by few taxa or composite codings (Milner 1988, 1993; Trueb and Cloutier 1991; Laurin and Reisz 1997; Ruta et al. 2003; Schoch and Milner 2004; Vallin and Laurin 2004; Anderson 2007; Anderson, Reisz, et al. 2008). Classically, branchiosaurids and amphibamids were considered to form two distinct, but closely related clades within dissorophoid amphibians. However, it has been tentatively suggested that branchiosaurids and amphibamids may not represent two distinct clades, but instead branchiosaurids may be neotenic amphibamids (Milner 1982; Anderson 2007; Anderson, Henrici, et al. 2008; Schoch and Milner 2008), a suggestion that if proven correct would have significant consequences for the discussion of character evolution within the group and in the evolution of the modern amphibian taxa. A combined phylogenetic analysis of amphibamids and branchiosaurids applying parsimony and Bayesian analytical approaches is the focus of this study, which represents the first comprehensive analysis of the relationships of small dissorophoid amphibians and provides a new framework for a discussion of the evolution of crucial morphological characters and developmental pathways shared with modern amphibian groups. MINIATURIZATION An important aspect that complicates the discussion of the relationships of branchiosaurids and amphibamids is that both groups represent miniaturized forms in comparison with their close relatives within dissorophoids and on a broader phylogenetic scale (Fig. 2). Miniaturization is widely distributed among metazoans (Hanken and Wake 1993) and is a well-studied phenomenon in many vertebrate clades (Miller 1996). With respect to amphibians, it is of particular significance, because miniaturization is thought to have played a crucial role in the origins of the modern amphibian groups. However, the impact of miniaturization on assessments of relationships has thus far not been addressed, but the morphological changes associated with a decrease in body size can artificially lead to a close grouping of miniaturized taxa. Generally, miniaturization is understood as the evolution of extremely small body size to a degree where the physiology, ecology, and behavior of the animal is affected (Hanken and Wake 1993). Miniaturization is, however, difficult to quantify and change from a “regular” size reduction to miniaturization may by a continuous transition (Hanken 1993; Hanken and Wake 1993). Size reduction can have profound effects on the organism, especially in cases where a minimum possible body size is approached. When dealing with fossil taxa, morphological changes in the skeleton are the only ones that can potentially be observed. Those are also the most conspicuous ones in extant forms and are frequently described in detail accompanied by analyses of how these morphologies are a result of, and themselves affect, changes of the associated soft tissues (Marshall and Corruccini 1978; Hanken 1983, 1984, 1993; Linke et al. 1986; Roth et al. 1992; Rieppel 1996). Paedomorphosis is usually implicated as the evolutionary pattern correlated with the reduction in size and some of the associated morphological changes. However, the actual heterochronic mechanisms leading to the size reduction can vary among taxa and have to be investigated in a phylogenetic framework for each considered taxon separately (Hanken 1993). This is especially important considering the frequency with which miniaturization can occur within a clade and the large degree of homoplasy associated with it. Wake (1992), for example, showed that a miniaturized morphology evolved at least 10 times within Plethodontidae, the lungless salamanders. Miniaturization was also shown to have evolved recurrently within rasporine fish, a clade containing Paedocypris, the smallest known vertebrate (Rüber et al. 2007). Different categories of morphological change are associated with miniaturization, with a combination of any number of them potentially occurring in different miniaturized taxa (Hanken 1982, 1983, 1984, 1985, 1993; Hanken and Wake 1993). These include reduction or structural simplification accompanying the decrease in size; hyperossification usually compensating for a weakening of the skeleton caused by structural simplification or reduction; morphological novelty as a consequence of skeletal rearrangements, which are a direct cause of size reduction; and increased variability, in particular, in the elements formed late in ontogeny and affected variably by the precocious truncation of the developmental trajectory. In the discussion of miniaturization, it is important to not only consider physical size but also the so-called “biological size” (Roth et al. 1992; Hanken and Wake 1993). The latter takes cell size into account, which is a direct measure of the size of the genome, that is, the amount of DNA. Genome size can vary widely between different taxa, even if they are closely related. Thereby, they may have the same physical size, but the cell size may be very different. The plethodontid salamander Thorius not only has an extremely small physical size but in addition also possesses very large cells. It thereby shows an even stronger size decrease with respect to the developmental mechanisms and principles of self-organization (Hanken and Wake 1993). PALEOZOIC DISSOROPHOIDS AND MINIATURIZATION While assessments of cell size in fossil taxa is largely impossible, identifying miniaturization in terms of a physical size decrease is certainly possible in the fossil record (e.g., Marshall and Corruccini 1978; Carroll 1990; Schoch and Rubidge 2005). The identification of the heterochronic mechanisms that lead to the miniaturized morphology and a potential recurrent evolution of miniaturization within a clade are more difficult to assess because detailed knowledge of the ontogeny and absolute age data as well as phylogenetic assessments derived independently of morphological data are usually unavailable for fossil forms. This problem is particularly compelling with regard to amphibians because miniaturization was repeatedly suggested to have played an important role in the evolution of major taxa, including the origins of the modern amphibian forms (Bolt 1977, 1979; Milner 1988, 1993; Carroll 1990). Milner (1993) discussed that a suite of characters supporting the temnospondyl-lissamphibian hypothesis are characters associated with miniaturization. Although he did not explicitly identify these characters, he listed the enlarged interpterygoid vacuities, the absence of the jugal and two coronoids (with one coronoid remaining), the absence of the ectopterygoid, and the straight, abbreviated ribs, all likely associated with small body size. As outlined above, branchiosaurids and amphibamids both have frequently been suggested as potential close relatives to either all modern amphibians or batrachians (salamanders+frogs) (Bolt 1969, 1977, 1991; Milner 1988, 1993; Trueb and Cloutier 1991; Ruta et al. 2003; Schoch and Milner 2004; Anderson 2007; Carroll 2007; Anderson, Reisz, et al. 2008), and they were found to be the only plausible antecedents among the known temnospondyls (Milner 1993). Both groups are undoubtedly very small compared with other representa- tives within Dissorophoidea (e.g., micromelerpetontids, trematopids and Ecolsonia) and with other temnos- pondyls in a broader phylogenetic scale such as the eryopids (Fig. 2). Amphibamids have previously been referred to as miniaturized dissorophoids (Milner 1988; Schoch 1995), and miniaturization is thought to have been a driving force in the evolution of this clade (Schoch 2002b; Schoch and Rubidge 2005; Schoch and Milner 2008). Characters that were mentioned as being associated with miniaturization in amphibamids by Schoch and Rubidge (2005) are the rudimentarily developed circumorbital elements, the poor dermal ornamentation, overall thin dermal bones, oversized orbits and otic notches, and the very extended interpterygoid vacuities of the palate. Notably though, the vertebral elements, neurocranium, and quadrates are well ossified (Carroll 1964; Bolt 1968; Clack and Milner 1993; Daly 1994; Schoch and Rubidge 2005; Huttenlocker et al. 2007; Fröbisch and Reisz 2008). In all these correlates of miniaturization, amphibamids resemble modern salamanders and frogs more closely than any other early tetrapod clade (Milner 1988). Amphibamids are classically interpreted as terrestrial forms and the latter features are associated with late ontogenetic stages and terrestrially (e.g. Schoch and Fröbisch 2006), which lead Schoch and Rubidge (2005) to hypothesize that derived amphibamids likely had a condensed larval development. This is supported by small-sample growth series of Platyrhinops lyelli (Schoch 2002b), an amphibamid whose larvae apparently shared the apomorphic branchial dentition with branchiosaurids, which was then lost during metamorphosis (Clack and Milner 2007). It is further indicated by the tiny body size of some amphibamid specimens from the South Grandfield locality in Texas, some of which have a skull length of no more than 15 mm, but show well-ossified braincase elements and jaw articulation (NBF, personal observation). The very small size is particularly conspicuous in the highly nested amphibamids Amphibamus, which also represents the smallest terrestrial dissorophoid, and Doleserpeton. This is contrasted by branchiosaurids, which share small body size within a similar size range with amphibamids, but neoteny is a pronounced feature, present in all but one member of this clade (Schoch and Fröbisch 2006; Fröbisch and Schoch 2009). Similar to amphibamids, branchiosaurids also show trends toward further miniaturization within the clade, indicated by very small forms nested higher in the phylogeny such as Apateon gracilis and Apateon dracyiensis. Apateon dracyiensis shows pronounced juvenile features in adult specimens, such as the short jugal and an open cheek (Boy 1987; Werneburg 2001, 2002; Schoch and Milner 2008), indicating a further precocious truncation of the ancestral developmental trajectory. The large sample size of some branchiosaurids also permits to observe a very high frequency of individual variation (Schoch and Milner 2008), which is a further characteristic of miniaturized species, although the degree of variability in other nonminiaturized temnospondyls is not well understood. On the one handthe small size of these two groups may lend support for the hypothesis that they represent the closest Paleozoic relatives of some or all modern amphibian taxa, considering that miniaturization is thought to have played an important role in the evolution of the extant amphibians, the vast majority of which is still rather small. On the other hand, it complicates the discussion of the relationships of amphibamids and branchiosaurids, character evolution within the clade, and its role in the evolution of the modern forms. The question arisesas to what impact miniaturization has on phylogenetic assessments of dissorophoid relationships. Miniaturization is known to produce a suite of homoplasies that can potentially have a strong effect on phylogenetic assessments. Indeed, Schoch and Rubidge (2005) state that amphibamids and branchiosaurids were likely exposed to similar constraints imposed by size reduction, although they clearly differ in characteristics not thought to be related to small body size. However, this hypothesis remained a general statement and was not tested further because a phylogenetic analysis of branchiosaurids and consequently no combined analysis of branchiosaurids and amphibamids had been attempted. In a study on the effects of neoteny (in the sense of the retention of larval somatic feature into sexual maturity) on the assessment of the relationships of modern salamander families, Wiens et al. (2005) found that developmental mechanisms could have profound effects on phylogenetic assessments based on adult morphological data, when similar large-scale developmental mechanisms are acting in distantly related clades. Although Wiens et al. (2005) were looking at two well-known developmental pathways in salamander ontogeny, neoteny (paedomorphosis), and metamorphosis, similar phenomena as observed by them for extant salamanders could be expected for this study on the effects of miniaturization on phylogenetic assessments. Paedomorphosis is also thought to be the primary mechanism leading to a miniaturized morphology and therefore a close clustering of miniaturized taxa as observed for neotenic taxa in salamanders can be expected. In their study, Wiens et al. (2005) suggested three approaches to assess the influence of neotenic characters on phylogenies: First, the exclusion of characters associated with neoteny; second, the exclusion of the neotenic taxa; and third coding the adult morphology of neotenic taxa as unknown, assuming that adults of neotenic taxa are not a comparable ontogenetic stage to adults of metamorphosed taxa. When these approaches are conveyed to the investigation of the effects of miniaturization on the relationships of small dissorophoids, the exclusion of characters associated with a miniaturized morphology is the only possible approach. The exclusion of the taxa would not be helpful because their relationships relative to each other and with other taxa would not be resolved (and for the same reason, this approach was also not chosen by Wiens et al. 2005). The third approach is also not feasible because there is no clear-cut delineation in the case of miniaturization as is between neotenic and metamorphosed morphologies that are correlated with distinct ontogenetic stages. Moreover, molecular characters that would provide additional and independent data for resolution after the deletion of a large amount of information are unavailable for studies on fossil taxa. In this study, characters associated with a miniaturized morphology were excluded from the data set in a second set of analyses to assess the impact of miniaturization on the phylogenetic relationships on small dissorophoid amphibians, investigating whether shared developmental pathways associated with miniaturization lead to an artificial intricate relationship between amphibamids and branchiosaurids. Moreover, the stratigraphic distribution of taxa relative to various tree topologies is investigated as an additional source of information for the comparison of tree topologies derived from the same data set (the complete data set and the data set without miniaturization characters), but different analytical approaches. The results of the analyses and their implications for the evolutionary history of small dissorophoids and the origins and intrarelationships of frogs, salamanders, and caecilians are discussed. MATERIAL AND METHODS Phylogenetic Analyses The following analyses are based on an integrated data matrix representing for the first time a comprehensive data set for the study of amphibamid and branchiosaurid relationships. The amphibamid data set is taken from Fröbisch and Reisz (2008), which in turn represents an extension of the data set by Schoch and Rubidge (2005) and includes all currently recognized amphibamid taxa. Furthermore, the recently assembled data set for branchiosaurid relationships by Schoch and Milner (2008) forms the second part of the integrated data matrix. Duplicate characters were deleted from the data set and the analysis presented here includes 31 taxa and 80 cranial and postcranial characters (Supplementary Appendices 1 and 2, available from http://sysbio.oxfordjournals.org/; TreeBASE submission number SN4543). The data analysis was divided into two components, which each utilized three analytical approaches. The first component represented the analysis of the complete data sets, the second one the analysis of a reduced data set, where characters thought to be associated with a miniaturized morphology were eliminated from the data matrix (see Supplementary Appendix 3). Both data sets were analyzed using a parsimony approach with all characters unordered, a parsimony approach with selected characters ordered (see discussion of individual analysis) and a Bayesian approach. Therefore, the two different data set and analytical treatments yielded six (consensus) tree topologies altogether. Additionally, a partition homogeneity test (Farris et al. 1995) was performed in PAUP (Swofford 2001) for the parsimony analyses of the complete data set with all characters ordered and selected characters unordered, respectively, to quantify a potential conflict between characters associated with miniaturization and the remaining characters of the matrix. The basal temnospondyls Dendrerpeton and Balanerpeton served as outgroups for all analyses. Parsimony analyses were conducted using the heuristic search mode of PAUP 4.0b10 (random-stepwise addition, 10 replicates) (Swofford 2001) and character evolution and optimization was investigated using MacClade 4.08 (Maddison and Maddison 1992). Node support was assessed by calculating bootstrap values with 1000 replicates and Bremer decay values. Bayesian analyses were conducted using Mr. Bayes 3.1.2. (Huelsenbeck and Ronquist 2005). The application of Bayesian approaches to morphological data sets represents a relatively recent approach providing an alternative to the traditional parsimony analyses in particular for analyses of fossil taxa that often show little robustness of trees or particular nodes due to the limited and often fragmentary data available in the fossil record (Lewis 2001; Snively et al. 2004; Müller and Reisz 2006). The Mk model was applied to the data set, which is currently the only available model for the analysis of morphological data, because it allows for a variable number of character states. It follows the assumption that character states can change with equal probability, in any direction, and at any time. Four chains with two independent runs were computed with a tree sampling at every one hundred generations. The first 5000 trees (the so called “burn-in”) were disregarded for the final evaluation. Log likelihood values were plotted against numbers of generations to test whether the Markov chain Monte Carlo chain converged on a stationary stage before tree sampling was initiated. This showed no obvious increase or decrease, indicating that the stationary stage was reached in the analyses prior to the start of tree sampling. All Bayesian analyses were run twice on the same data set, that is, without and with gamma shape distribution, the latter allowing for different rates of character change between different characters. The two runs per data set were compared by calculating the Bayes factor, which is based on the difference between the harmonic means of the log-likelihoods of each respective run multiplied by 2. It provides a measure that helps to assess whether variation in the rates of character change improves the fit of the data to the model. Generally a Bayes factor > 10 is considered to reflect a strong support for the gamma shape distribution (Kass and Raftery 1995; Müller and Reisz 2006). Stratigraphic Indices The fit of the tree topologies to the stratigraphic distribution of taxa in the rock record provides an additional source of information for the comparison of results yielded by different analytical treatments. Three commonly used indices for the fit of the stratigraphic data with the tree topology are the stratigraphic congruence index (SCI), stratigraphic congruence index (RCI), and the gap excess ratio (GER) all of which reflect slightly different aspects of the fit of stratigraphic data to a cladogram. The SCI (Huelsenbeck 1994) describes the congruence of a tree with the stratigraphic distribution of the taxa as the proportion of stratigraphically consistent nodes within the tree and the overall number of nodes. Therein, a node is considered stratigraphically consistent if the oldest first occurrence above a given node has the same age or is younger than the first occurrence date of the sister node. An SCI of 1.00 would therefore reflect a perfect fit of the tree topology to the stratigraphic distribution of the taxa. The two other measures, the RCI (Benton and Storrs 1994) and the GER (Wills 1999) incorporate time in their measures in form of minimum implied gaps (ghost lineages). The RCI incorporates a ratio of the minimum implied gaps and the actually observed stratigraphic ranges of the considered taxa and thereby provides a measure for the relative completeness of the fossil record (Benton and Storrs 1994; Hitchin and Benton 1997; Wills 1999). RCI values can range from large negative values, when the minimum implied gaps are longer than the actually observed stratigraphic ranges, to a theoretical maximum of 100% (Benton and Storrs 1994; Hitchin and Benton 1997; Siddall 1998; Wills 1999). In contrast, the GER does not include the actually observed stratigraphic range of the considered taxa and thereby does not conflate congruence of the tree topology and stratigraphic data with the completeness of the fossil record (Wills 1999). The GER calculates the excess ghost range based on the cladogram and the stratigraphic range data of the considered taxa over the minimal ghost range possible for the given stratigraphic data and is then expressed as a fraction of the total range of ghost values possible for the same stratigraphic data on any cladogram (Wills 1999). The highest value for the GER (1.00) can be obtained, even if only a point distribution of fossils is known, whereas in the latter case, the RCI values tend to become large and negative. Previously, measures of stratigraphic fit have been used as an independent optimality criterion to choose between several most parsimonious trees yielded by a single parsimony analysis, but it has also been argued that the indices reflect more the completeness of the data in general (Siddall 1998). Unlike in previous studies, the three indices are here used in a novel approach to compare trees that are derived from the same data set, but from different analytical methods, that is, the two parsimony analyses (characters unordered vs. selected characters ordered) and the Bayesian analysis. Measures of stratigraphic fit for tree topologies derived from the complete and the reduced data set have to be treated separately because they are based on a different number of characters. SCI, RCI, and GER were calculated using the program Ghosts version 2.3 (Wills 1999). Stratigraphic data were amended from the default settings of the program to reflect the stages and according to absolute ages as given in the most recent geologic timescale (Gradstein et al. 2004). One thousand permutations of the range data over terminal taxa yielded a randomized distribution of values to assess whether the measures yielded for the tree topologies vary significantly from a random distribution. As suggested by Wills (1999), the data of the actual ranges were considered significantly better than expected by chance, if it was greater than all but a small tail of randomized values (5%). The SCI and the RCI have been criticized, in particular, because they were shown to be theoretically influenced by tree topology, that is, pectinate versus balanced trees (Siddall 1997, 1998), and the same holds true for the GER (Wills 1999). However, although this theoretical relationship between these values and tree balance exists, Hitchin and Benton (1997) showed on the basis of a large number of cladograms derived from empirical data that SCI and RCI values of balanced trees were on average as high as for unbalanced trees, not reflecting the expected inverse relationship. For the purpose of this study, the imbalance value I(s) (Shao and Sokal 1990) was assessed for all six trees to test for the possible correlation between tree balance and values for stratigraphic fit. None of the three measures (SCI, RCI, and GER) were found to be significantly correlated with tree balance (Tables 1 and 2). TABLE 1. Table of measures of consistency of tree topology with stratigraphic distribution   SCI  RCI  GER  Imbalance value I(s)  Complete data set          Parsimony analysis all characters unordered  0.52  − 0.87%  0.812  201  Parsimony analysis selected characters ordered  0.52  − 0.87%  0.812  240  Bayesian analysis  0.59  6.67%  0.831  311  Reduced data set          Parsimony analysis all characters unordered  0.52  − 11.88%  0.784  244  Parsimony analysis selected characters ordered  0.59  1.74%  0.819  265  Bayesian analysis  0.59  3.77%  0.824  320    SCI  RCI  GER  Imbalance value I(s)  Complete data set          Parsimony analysis all characters unordered  0.52  − 0.87%  0.812  201  Parsimony analysis selected characters ordered  0.52  − 0.87%  0.812  240  Bayesian analysis  0.59  6.67%  0.831  311  Reduced data set          Parsimony analysis all characters unordered  0.52  − 11.88%  0.784  244  Parsimony analysis selected characters ordered  0.59  1.74%  0.819  265  Bayesian analysis  0.59  3.77%  0.824  320  Note: Imbalance value I(s) from Shao & Sokal (1990). View Large TABLE 2. Spearman's rank and P values for the correlation analysis of the imbalance value I(s) with measures of stratigraphic fit   Spearman's rs  P value  SCI  0.878  0.103  RCI  0.754  0.133  GER  0.754  0.104    Spearman's rs  P value  SCI  0.878  0.103  RCI  0.754  0.133  GER  0.754  0.104  View Large RESULTS AND DISCUSSION Complete Data Set Parsimony analysis (characters unordered).— The data set was analyzed in a parsimony analysis via a heuristic search with all characters equally weighted and unordered. The search resulted in 13 most parsimonious trees, the strict consensus of which as well as Bootstrap values and Bremer support values are shown in Figure 3a. The tree length is 270 steps with a consistency index of 0.524 and a rescaled consistency index of 0.406. FIGURE 3. View largeDownload slide Strict consensustrees of the parsimony analyses of the complete data set with all character unordered (a) and selected characters ordered (b). Bold numbers indicate bootstrap values, italic numbers indicate Bremer support. A and B in the cladogram refer to the nodes of amphibamids and branchiosaurids, respectively. FIGURE 3. View largeDownload slide Strict consensustrees of the parsimony analyses of the complete data set with all character unordered (a) and selected characters ordered (b). Bold numbers indicate bootstrap values, italic numbers indicate Bremer support. A and B in the cladogram refer to the nodes of amphibamids and branchiosaurids, respectively. The analysis results in a polytomy at the base of the amphibamid clade, which however clearly includes branchiosaurids. Therein, the monophyly of branchiosaurids is well supported and the amphibamid Georgenthalia clusters as their closest relative. Although much of the relationships within amphibamids remain unresolved, a clade including the basal taxa ([Micropholis, Pasawioops] Tersomius) is recovered, which has also been proposed by Fröbisch and Reisz (2008). The branchiosaurid relationships differ in some respects from those proposed by Schoch and Milner (2008). They recognized two distinct clades within branchiosaurids, the Apateon-clade, and the Melanerpeton-clade, and the oldest branchiosaurid taxon Branchiosaurus was positioned at the base of both of these clades. In this analysis, the relationships within the Apateon-clade are consistent with Schoch and Milner's (2008) results, but Branchiosaurus is positioned in a polytomy with the Apateon-clade and Melanerpeton-clade. A distinct Melanerpeton-clade is also retrieved in this analysis with Melanerpeton humbergese as its the basalmost member, which differs from Schoch and Milner (2008) result, where M. eisfeldi is positioned at the base of this clade. The relationships of the remaining members of Melanerpeton-clade are comparable with Schoch and Milner's results with Leptorophus tener and Schoenfelderpeton prescheri in a well-supported sister taxon relationship. Micromelerpeton forms the basalmost member of Dissorophoidea. The relationships of dissorophids, trematopids, and Ecolsonia remain unresolved in this analysis, but all four taxa are positioned in a more derived position relative to Micromelerpeton as yielded by the individual amphibamid analysis (Fröbisch and Reisz 2008) and branchiosaurid analysis (Schoch and Milner 2008). Parsimony analysis (selected characters ordered).— In the analysis of branchiosaurid relationships by Schoch and Milner (2008), all characters were treated as ordered. Their decision to order particularly multistate characters was based on the well-documented ontogenetic character transformations in many taxa included in their analysis, implying that evolutionary and ontogenetic character transformations are correlated. This explicitly referred to evolutionary modifications of ontogenies and was not meant to imply recapitulation in its historical sense see Schoch and Milner (2008) for further details on their argumentation]. Considering the good knowledge of the ontogenetic series of many taxa in which distinct patterns such as paedomorphosis have been recognized without changes to the ancestral sequence, this is a valid justification for ordering characters. However, some characters of Schoch and Milner's (2008) original study were amended for this analysis, for example, by adding additional characters states, to accommodate the wider range of taxa and morphologies considered here (see Supplementary Appendix 1). Therefore, the different character states do not necessarily reflect an ontogenetic or evolutionary transformation series anymore and thus cannot be treated as ordered under the justification outlined above. Only six characters were treated as ordered in this modified version of the parsimony analysis (characters #35, 38, 62, 63, 68, and 79). The parsimony analysis resulted in 14 most parsimonious trees. The strict consensus tree as well as the Bootstrap and Bremer decay values are shown in Figure 3b. The tree length for this tree is 272 steps with a consistency index of 0.520 and a rescaled consistency index of 0.405. Ordering of the characters results in a more resolved topology compared with the analysis with all characters unordered. Branchiosaurids remain a well-defined clade, but again with a polytomy at its base, including Branchiosaurus and the monophyletic Apateon-clade and Melanerpeton-clade. The relationships within the Apateon-clade and Melanerpeton-clade are the same as recovered for the parsimony analysis with all characters unordered. The nesting of amphibamids is well resolved in this analysis and clearly supports a paraphyly of amphibamids with respect to branchiosaurids. A basal split can be recognized in the Amphibamidae (incl. branchiosaurids), which coincides with the split into two distinct clades of amphibamids retrieved by Fröbisch and Reisz (2008). One clade includes the primitive amphibamids [(Micropholis, Pasawioops) Tersomius] and a second clade contains the derived amphibamids as well as the monophyletic Branchiosauridae. Therein, as in the parsimony analysis with all characters unordered, Georgenthalia forms the sister taxon to branchiosaurids. Even more highly nested is a clade including the remaining derived amphibamid taxa (((((Amphibamus, Platyrhinops) Doleserpeton) Gerobatrachus) Plemmyradytes) Eoscopus). As in the previous analysis, the monophyly of Dissorophoidea is well supported with Micromelerpeton as the basalmost representative of this clade. Dissorophines and Cacopines form a robust clade, but their relationships with Trematopidae and Ecolsonia are unresolved. Bayesian analysis.— The Bayesian analysis was performed in two consecutive runs, one with and one without gamma shape distribution. The harmonic mean without gamma shape distribution is − 911.82, and − 906.10 with gamma shape distribution, resulting in a Bayes factor of 11.44. On the basis of this value, allowing rate variation between different characters appears to be a better choice for this data set. The result of the Bayesian analysis also shows a paraphyly of amphibamids with respect to branchiosaurids (Fig. 4), therefore supporting the suggestion that branchiosaurids are a subgroup of amphibamids (Milner 1982; Anderson 2007; Anderson, Henrici, et al. 2008; Schoch and Milner 2008). The basal split of Amphibamidae can also be recognized in this analysis, with one clade containing the basal amphibamids Micropholis, Pasawioops, and Tersomius and a second clade comprising the derived amphibamids as well as all branchiosaurids. Plemmyradytes is positioned at the base of this second clade and the next successively higher nested taxa are Eoscopus and Georgenthalia. Branchiosaurids constitute a monophyletic clade that forms the sister taxon to the derived amphibamids {[(Amphibamus, Gerobatrachus) Doleserpeton] Platyrhinops}, suggesting a slightly higher nesting of branchiosaurids than retrieved by the parsimony analysis with selected characters ordered. FIGURE 4. View largeDownload slide Tree topology yielded by the Bayesian analysis of the complete data set. Numbers indicate posterior probability values. A and B in the cladogram refer to the nodes of amphibamids and branchiosaurids, respectively. FIGURE 4. View largeDownload slide Tree topology yielded by the Bayesian analysis of the complete data set. Numbers indicate posterior probability values. A and B in the cladogram refer to the nodes of amphibamids and branchiosaurids, respectively. Within branchiosaurids, Branchiosaurus clusters at the base of the Apateon-clade and the relationships within the latter are essentially similar to those yielded by the parsimony analysis with all characters unordered. The Bayesian analyses also resulted in a distinct Melanerpeton-clade, with Melanerpeton eisfeldi as its basalmost member. The relationships of amphibamids retrieved by the Bayesian analyses strongly resemble those that were retrieved in the analysis of Fröbisch and Reisz (2008). Moreover, branchiosaurids are positioned as a distinct clade between the highly nested and the three more basal members of the derived amphibamids. At the base of the Dissorophoidea, the dissorophids, Ecolsonia, and trematopids form a distinct clade and as in the parsimony analyses, Micromelerpteon forms the basalmost member of the dissorophoids. Reduced Data Set To investigate the possibility that miniaturized body size is affecting the phylogenetic assessment of branchiosaurid and amphibamid relationships, 13 characters that are associated with a miniaturized morphology, as previously suggested in the literature, were excluded from the data set. The deleted characters and their effect on the morphology are discussed in Supplementary Appendix 3. Parsimony analysis (all characters unordered).— The exclusion of characters associated with miniaturization reduces the overall resolution and results in 30 most parsimonious trees (Fig. 5a). The analysis with the reduced data set results in branchiosaurids being clearly nested within the derived amphibamids. As in the previous analyses, distinct Melanerpeton- and Apateon-clades within branchiosaurids are recovered, but therein Branchiosaurus forms the sister taxon to the Melanerpeton-clade rather than being more closely associated with the Apateon-clade as in the Bayesian analysis of the complete data set. The recently described Gerobatrachus (Anderson, Reisz, et al. 2008) forms the sister taxon to the monophyletic Branchiosauridae with Amphibamus, Doleserpeton, Platyrhinops, and Eoscopus as successively more basal taxa. Therefore, as in the trees retrieved from the analyses of the complete data set, the Branchiosauridae are nested within derived amphibamids but are even higher nested within that clade. The relationships of the remaining amphibamids are not resolved, but the monophyly of the clade amphibamids+branchiosaurids is supported. The relationships of the dissorophids, trematopids, and Ecolsonia are the same as retrieved by the Bayesian analysis of the complete data set and Micromelerpeton represents the basalmost member of the monophyletic Dissorophoidea. FIGURE 5. View largeDownload slide Strict consensus trees of the parsimony analyses of the reduced data set with all character unordered (a) and selected characters ordered (b). Bold numbers indicate bootstrap values, italic numbers indicate Bremer support. A and B in the cladogram refer to the nodes of amphibamids and branchiosaurids, respectively. FIGURE 5. View largeDownload slide Strict consensus trees of the parsimony analyses of the reduced data set with all character unordered (a) and selected characters ordered (b). Bold numbers indicate bootstrap values, italic numbers indicate Bremer support. A and B in the cladogram refer to the nodes of amphibamids and branchiosaurids, respectively. Parsimony analysis (selected characters ordered).— Only five multistate characters were treated as ordered in the reduced data set compared with six characters in the complete data set because character 63 (the contact between maxilla and quadratojugal) was excluded as a character associated with miniaturization. The ordering of characters increased the resolution compared with the analysis of the reduced data set with all characters unordered and resulted in six most parsimonious trees. The strict consensus tree (Fig. 5b) shows that the monophyletic clade of branchiosaurids is again nested within the derived amphibamids, and therein Georgenthalia forms their immediate sister taxon as also retrieved by the parsimony analysis of the complete data set with selected characters ordered. Highly nested within the clade of derived amphibamids+branchiosaurids is a further clade comprising {[(Amphibamus, Platyrhinops) Gerobatrachus] Doleserpeton}. Contrary to the previous analyses, in this study Plemmyradytes clusters within the clade of basal amphibamids in a sister taxon relationship with Micropholis with Pasawioops and Tersomius as successively more basal taxa. A close relationship of Plemmyradytes with Micropholis was also suggested by Huttenlocker et al. (2007) in the initial description of Plemmyradytes, but this taxon was found to be more closely related to members of the clade of derived amphibamids in a subsequent, more inclusive study of amphibamid relationships (Fröbisch and Reisz 2008). Eoscopus represents the basalmost member of the clade amphibamids+branchiosaurids. Within branchiosaurids, Branchiosaurus forms the basalmost member of this monophylum. This is consistent with the results of Schoch and Milner's (2008) analysis of branchiosaurid relationships, but differs from the other analyses in this study, where the position of Branchiosaurus remained either unresolved or Branchiosaurus variably fell either with the Apateon-clade or the Melanerpeton-clade. The two clades within branchiosaurids are also reflected in the results of this analysis, whereas the relationships within the Apateon-clade are essentially similar to those of the previous analyses. Apart from the sister taxon relationship of Leptorophus tener and Schoenfelderpeton prescheri, the relationships within the Melanerpeton-clade are unresolved in a polytomy at its base in the strict consensus tree. Bayesian analysis.— Bayesian analyses with and without gamma shape distribution were performed on the basis of the reduced data set. The Bayes factor was calculated to assess whether the implementation of the gamma shape parameter increases the fit of the model to the data. The harmonic mean without gamma shape parameter was − 779.05 and with gamma shape parameter − 772.86 resulting in a Bayes factor of 6.19. Therefore, there is no strong support for running the analyses allowing for variable rates of character change. Branchiosaurids form a monophyletic group nested highly within derived amphibamids (Fig. 6). Gerobatrachus and Amphibamus are in a sister taxon relationship and are the most derived members of this clade also forming the sister taxon to branchiosaurids, whereas Doleserpeton, Platyrhinops, and Eoscopus are positioned successively more basal. A second clade comprises {[(Micropholis, Pasawioops) Plemmyradytes] Tersomius}, showing that Plemmyradytes is nested within basal amphibamids as previously indicated in the parsimony analysis of the reduced data set with selected characters ordered. Georgenthalia is positioned at the very base of the clade amphibamids+branchiosaurids. The clade comprising dissorophids, Ecolsonia, and trematopids forms the sister clade to amphibamids+branchiosaurids and as in all previous analyses Micromelerpeton is the basalmost representative of the monophyletic Dissorophoidea. FIGURE 6. View largeDownload slide Tree topology yielded by the Bayesian analysis of the reduced data set. Numbers indicate posterior probability values. A and B in the cladogram refer to the nodes of amphibamids and branchiosaurids, respectively. FIGURE 6. View largeDownload slide Tree topology yielded by the Bayesian analysis of the reduced data set. Numbers indicate posterior probability values. A and B in the cladogram refer to the nodes of amphibamids and branchiosaurids, respectively. Stratigraphic Consistency An independent data set for the assessment of phylogenetic relationships in the form of molecular data is unavailable for fossil forms that would allow for a broader investigation of the effects of character exclusion from the data set. However, another source of data is available for fossil taxa that can be considered for an additional, albeit not an alternative, source of information for the comparison of tree topologies, which is the distribution of taxa in the stratigraphic rock record. The relative appearance of fossil taxa in the fossil record was suggested as an independent source of information, in particular, when stratigraphic information was not included in the initial data set used for tree building (Huelsenbeck 1994). Three measures for the fit of tree topology to the stratigraphic record were considered in this study, the SCI (Huelsenbeck 1994), RCI (Benton and Storrs 1994), and GER (Wills 1999). For the complete data set, all three values are highest for the tree yielded by the Bayesian analysis (Table 1). The highest RCI value of 6.67% indicates that shorter ghost ranges relative to the observed ranges had to be implied for this tree topology than for the other two trees. The values for the trees yielded by the two parsimony analyses are the same with an SCI value of 0.52 and a negative RCI value of − 0.87%. GER values, although ranging between 0 and 1, generally rank trees in the same order of goodness of fit of the topology with the stratigraphic distribution as the RCI. Therefore, the Bayesian tree has the highest GER value (0.831) and the parsimony trees both have values of 0.812. Tree topologies produced by the analyses of the reduced data set also show quite similar GER values, whereas the Bayesian tree again has the highest value of 0.824 (Table 1). The RCI varies from a large and negative value of − 11.88% for the parsimony tree with all characters unordered, to 1.74% for the tree of the parsimony analysis with selected characters ordered, to 3.77% for the Bayesian tree, indicating that very different length of ghost ranges relative to the actual stratigraphic ranges have to be inferred for the three different topologies. The SCI is lowest for the parsimony analysis with all characters unordered, but in general, the difference between the SCI values of the three trees is quite small. For both data sets, the measures of fit of the tree topology to stratigraphic distribution indicate the best fit for the Bayesian trees. If consistency of the tree topology with the stratigraphic distribution of taxa in the rock record is used as an independent optimality criterion to chose between different tree topologies, these trees therefore were to be given preference. CONCLUSIONS Relationships of Small Dissorophoid Amphibians The results of the comprehensive phylogenetic analyses of small dissorophoid amphibians indicate a close relationship of branchiosaurids and amphibamids. Although the resulting tree topologies vary in some respects between the different subsets and analytical approaches, consistent patterns can be discerned. Micromelerpeton forms the basalmost member of the monophyletic Dissorophidea in all analyses. In the analyses that resolved the relationships of dissorophids, Ecolsonia and trematopids, Dissorophinae and Cacopinae form a well-supported sister clade with Ecolsonia and Trematopidae as successively more inclusive taxa. Branchiosaurids form a monophyletic group in all of the yielded trees and represent a robust clade as reflected in particularly high support values in the parsimony and Bayesian analyses. In all analyses, amphibamids are invariably paraphyletic with respect to branchiosaurids, whereas the latter are consistently nested within the clade of derived amphibamids. Support values for the clade including amphibamids and branchiosaurids vary between the analyses. Bayesian analyses of both the complete and reduced data set show strong support for this clade with posterior probability values of 0.98 and0.97, respectively. The parsimony analyses show rather weak support for this clade, except for the analysis of the complete data set with selected characters ordered with a bootstrap value of 86% for this node. Amphibamus, Doleserpeton, Gerobatrachus, and Platyrhinops are always the most highly nested taxa within the clade of derived amphibamids with only little variation in their relationships to each other. The close relationship of the amphibamids Pasawioops, Micropholis, and Tersomius to the exclusion of the derived amphibamid taxa was previously recognized in the analysis of amphibamid relationships (Fröbisch and Reisz 2008). This clade is recovered in all analyses, where amphibamid relationships are resolved. Plemmyradytes variably falls within this clade of basal amphibamids or clusters more closely with the derived amphibamids. The somewhat variable placement of this taxon has previously been recognized and is likely a result of the fragmentary nature of the fossil material (Fröbisch and Reisz 2008). Likewise, the somewhat variable placement of Georgenthalia was previously reflected in analyses of amphibamid relationships (excluding branchiosaurids; Anderson, Henrici, et al. 2008; Fröbisch and Reisz 2008) and may be the result of a juvenile ontogenetic stage of the only known specimen. The exclusion of characters that are thought to be associated with miniaturization did not result in a placement of amphibamids and branchiosaurids as more distantly related taxa as was expected if the common pathways associated with miniaturization were responsible for their shared similarities. In contrast, all analyses performed with the reduced data set invariably resulted in the paraphyly of amphibamids with respect to branchiosaurids, therefore supporting an intricate relationship of the amphibamid taxa with branchiosaurids. Partition homogeneity tests performed in PAUP demonstrate that the two data sets consisting of the 13 characters associated with miniaturization and the remaining characters of the matrix, respectively, were congruent in the complete data set with all characters unordered (P = 0.61) as well as in the analysis with selected characters ordered (P = 0.49). This further corroborates that miniaturization is not responsible for the close relationship of branchiosaurids and amphibamids reflected in the phylogenetic analyses. A possible paraphyly of amphibamids with respect to branchiosaurids has previously been suggested tentatively (Milner 1982; Anderson 2007; Anderson, Henrici, et al. 2008; Schoch and Milner 2008), but was thus far not tested in a comprehensive phylogenetic framework. The paraphyly of amphibamids with respect to branchiosaurids is clearly supported by all six analyses of the complete data set. If the additional information from the fossil record, that is, stratigraphic consistency of the retrieved tree topologies is taken into consideration, the support is highest for the trees yielded by the Bayesian analyses. Both of these trees support a high nesting of branchiosaurids within the clade of derived amphibamids with either Amphibamus and Gerobatrachus, or {[(Amphibamus, Gerobatrachus) Doleserpeton] Platyrhinops} as closest relatives to branchiosaurids. Evolutionary Implications Life-history strategies.— What does this tell us about the evolutionary history of small dissorophoid amphibians? Amphibamids and branchiosaurids differ profoundly in their life-history strategies, which in amphibamids are reflected in characters associated with an effective terrestrial locomotion and feeding, and in branchiosaurids in characters associated with the almost universal presence of neoteny. However, in his important study on the branchiosaurids from the Rotliegend of Germany, Boy (1972) discussed that despite the neotenic nature of branchiosaurids, they show a mosaic of characters associated with terrestriality (e.g., shortening of the trunk region, flattening of the skull, and an enlarged otic notch) and obvious characters of aquatic animals (e.g., external gills with specialized branchial denticles for filter feeding, laterally compressed tails with tail fins, and weakly ossified and undifferentiated limb skeletons). On the basis of these observations, Boy (1972) proposed an evolutionary scenario, suggesting that branchiosaurids were a primarily terrestrial group, possibly with representatives living in an upland environment or in the upstream parts of rivers, where preservational conditions are poor and therefore these terrestrial counterparts remain unknown. The excellent fossil record of branchiosaurids therein reflects only the secondarily aquatic forms, which much like modern salamanders acquired most of their aquatic adaptations via paedomorphosis. Although Boy's scenario was mainly hypothetical at the time, it becomes more conceivable in the light of this and other recent studies. The results of the phylogenetic analyses show branchiosaurids highly nested within amphibamids, which supports a terrestrial origin of the clade comprising amphibamids and branchiosaurids as depicted in Boy's (1972) scenario, with branchiosaurids forming a clade of secondarily aquatic, neotenic amphibamids. Therein, many of the characteristic synapomorphies of branchiosaurids, such as the apomorphic filter feeding apparatus, evolved as adaptations to the aquatic environment in this group, as has been previously suggested (Schoch and Milner 2008). The recently described amphibamid Georgenthalia was found in the Lower Permian Bromacker quarry in Germany (Anderson, Henrici, et al. 2008), which has been interpreted as preserving an upland fauna (Eberth et al. 2000), which at least indicates that terrestrial amphibamids were present in the upland environments at a similar time. Moreover, Platyrhinops co-occurs with Branchiosaurus, the earliest known member of the branchiosaurid clade, in the Upper Carboniferous (Moscovian) coal deposits of Nyrany in the Czech Republic. It is likely, however, that branchiosaurids already experienced a period of separate evolutionary history as a distinct group of neotenic amphibamids predating the Moscovian, given the specialization visible in the branchial apparatus and palatal anatomy seen already in Branchiosaurus as the earliest known member of the clade. The phylogeny of branchiosaurids indicates that the occurrence of metamorphosed members of the species A. gracilis probably represents a reversal to the plesiomorphic life-history pattern of metamorphosis (Schoch and Fröbisch 2006; Schoch and Milner 2008). The life history of modern amphibians is characterized by a strong condensation of events associated with metamorphosis and the first evidence for this condensation in the fossil record was reported in branchiosaurids (Schoch and Fröbisch 2006). Although detailed ontogenetic series of amphibamids are not available at this time, the data at hand suggest a condensed larval phase in amphibamids as well (Schoch 2002b; Schoch and Rubidge 2005), which in the framework of the phylogenetic relationships yielded in this study would be considered a plesiomorphic state for the clade comprising amphibamids and branchiosaurids. Pedicellate dentition.— Pedicely is an important character that has frequently been discussed in connection with the origins of the three modern amphibian groups, which all display a pedicellate dentition. Pedicely is definitively known from the three most derived amphibamid taxa, Amphibamus, Gerobatrachus, and Doleserpeton (Bolt 1969, 1977, 1979; Anderson, Reisz, et al. 2008) and has also been reported for Tersomius, but this taxon needs taxonomic revision before a more confident discussion of its characters in a phylogenetic framework is possible. Pedicellate teeth have also been reported for a single specimen of the branchiosaurid Apateon caducus (Schoch and Carroll 2003), indicating the possible presence of pedicely within the branchiosaurid clade. Therefore, the evolution of pedicellate dentition may have occurred at the base of the clade including derived amphibamids and branchiosaurids. Ontogenetic data and phylogeny.— The excellent fossil record of branchiosaurids allowed for a detailed study of their ontogeny, in particular their ossification sequence, and many similarities between branchiosaurids and extant amphibians, in particular salamanders have been described (Schoch 1992, 2002a, 2004; Schoch and Carroll 2003; Schoch and Fröbisch 2006; Carroll 2007; Fröbisch et al. 2007). The uniqueness of the fossil record of branchiosaurids at the same time holds the problem that in many cases it is difficult to assess when certain development pathways evolved in the fossil record and on what level they may represent true synapomorphies or are parallelisms (Schoch 2006; Anderson 2007; Fröbisch et al. 2007). The phylogenetic position of branchiosaurids nested within derived amphibamids could indicate that many of the ontogenetic features observed in branchiosaurids may represent symplesiomorphies for amphibamids including branchiosaurids (or an even more inclusive group), which however, will only be possible to test when more data on the ontogeny of amphibamids become available. It becomes clear that a mosaic of features associated with miniaturization on the one hand and the evolution of different life-history strategies on the other hand characterize the evolution of small dissorophoid amphibians of the Paleozoic. An understanding of the relationships of these groups is essential for the ongoing discussion of lissamphibian origins. The reduction of the palatal elements, the pedicellate dentition, and several aspects of their ontogeny, such as metamorphosis and neoteny, are features that are shared by amphibamids and/or branchiosaurids and the modern taxa. A better understating of the relationships of amphibamids and branchiosaurids with each other provides a new framework for the discussion of the evolution of these characters and a better assessment of their explanatory power for lissamphibians origins. SUPPLEMENTARY MATERIAL Supplementary material can be found at http://sysbio.oxfordjournals.org/. FUNDING This work was funded by a graduate award through a National Science and Engineering Research Council of Canada discovery grant to R. L. Carroll and McGill Graduate Student Fellowships. We would like to thank R.L. Carroll and H.C.E. Larsson for fruitful discussions and advise. N.B.F. is grateful to Johannes Müller for an introduction to Bayesian analyses and helpful discussions on this topic. J. Fröbisch read earlier versions of this manuscript and with many helpful comments and discussions greatly helped to improve it. We would very much like to thank J.A. Boy, A.R. Milner, and J.S. Anderson for providing unpublished manuscripts and many helpful discussions. The constructive comments by A.R. Milner and an anonymous reviewer greatly helped to improve this paper. References Anderson JS.  Anderson JS,  Sues H-D.  Incorporating ontogeny into the matrix: a phylogenetic evaluation of developmental evidence for the origin of modern amphibians,  Major transitions in vertebrate evolution ,  2007 Bloomington (IN) Indiana University Press(pg.  182- 227) Anderson JS,  Henrici A,  Sumida SS,  Martens T,  Berman DS.  Georgenthalia clavinasica, a new genus and species of dissorophoid temnospondyl from the early Permian of Germany, and the relationships of the family Amphibamidae,  J. Vertebr. Paleontol ,  2008, vol.  28 (pg.  61- 75) Google Scholar CrossRef Search ADS   Anderson JS,  Reisz RR,  Fröbisch NB,  Scott D,  Sumida SS.  A stem batrachian from the Early Permian of Texas and the origin of frogs and salamanders,  Nature ,  2008, vol.  453 (pg.  515- 518) Google Scholar CrossRef Search ADS PubMed  Benton MJ,  Storrs GW.  Testing the quality of the fossil record: paleontological knowledge is improving,  Geology ,  1994, vol.  22 (pg.  111- 114) Google Scholar CrossRef Search ADS   Bolt JR. ,  The osteology and relationships of Doleserpeton annectens, a new rhachitomous amphibian from the Lower Permian of Oklahoma [PhD thesis] ,  1968 [Chicago (IL)] University of Chicago(pg.  1- 176) Bolt JR.  Lissamphibian origins: possible protolissamphibian from the Lower Permian of Oklahoma,  Science ,  1969, vol.  166 (pg.  888- 891) Google Scholar CrossRef Search ADS PubMed  Bolt JR.  Dissorophoid relationships and ontogeny, and the origin of the Lissamphibia,  J. Paleontol ,  1977, vol.  51 (pg.  235- 249) Bolt JR.  Nitecki MH.  Amphibamus grandiceps as a juvenile dissorophid,  Mazon creek fossils ,  1979 New York Academic Press(pg.  529- 564)  evidence and implications Bolt JR.  Schultze H-P,  Trueb L.  Lissamphibian origins,  Origins of the higher groups of tetrapods: Controversy and consensus ,  1991 Ithaca (NY) Comstock Publishing Associates(pg.  194- 221) Boy JA.  Die Branchiosaurier (Amphibia) des saarpfaelzischen Rotliegenden (Perm, SW-Deutschland). Abh. hess,  Landesamt Bodenf ,  1972, vol.  65 (pg.  6- 137) Boy JA.  Studien über die Branchiosauridae (Amphibia: Temnospondyli; Ober-Karbon - Unter-Perm. Systematische Übersicht,  N.Jb. Geol. Paläont. Abh ,  1987, vol.  174 (pg.  75- 104) Carroll RL.  Early evolution of the dissorophoid amphibians. Bull. Mus. Comp. Zool,  Harv ,  1964, vol.  131 (pg.  161- 250) Carroll RL.  A tiny microsaur from the lower Permian of Texas: size constraints in Paleozoic tetrapods,  Palaeontology ,  1990, vol.  33 (pg.  893- 909) Carroll RL.  The Palaeozoic ancestry of salamanders, frogs, and caecilians,  Zool. J. Linn. Soc. ,  2007, vol.  150  suppl. 1(pg.  1- 140) Google Scholar CrossRef Search ADS   Clack J,  Milner A.  The amphibamid Platyrhinops, morphology and metamorphosis,  J. Vertebr. Paleontol ,  2007, vol.  27 pg.  59A  Google Scholar CrossRef Search ADS   Clack JA,  Milner AR.  Heidtke U.  Platyrhinops from the Upper Carboniferous of Linton and Nyrany and the family Amphibamidae (Amphibia: Temnospondyli),  New research on Permo-Carboniferous faunas ,  1993(pg.  185- 191)  Bad Dürkheim (Germany): Pollichia Daly E.  The Amphibamidae (Amphibia: Temnospondyli), with a description of a new genus from the Upper Pennsylvanian of Kansas,  Univ. Kansas Misc. Publ. ,  1994, vol.  85 (pg.  1- 59) Eberth DA,  Berman DS,  Sumida SS,  Hopf H.  Lower Permian terrestrial paleoenvironments and vertebrate paleoecology of the Tambach Basin (Thuringia, Central Germany): the upland Holy Grail,  Palaios ,  2000, vol.  15 (pg.  293- 313) Google Scholar CrossRef Search ADS   Farris JS,  Kallersjo M,  Kluge AG,  Bult C.  Constructing a significance test for incongruence,  Syst. Biol. ,  1995, vol.  44 (pg.  570- 572) Google Scholar CrossRef Search ADS   Feller AE,  Hedges SB.  Molecular evidence for the early history of living amphibians,  Mol. Phylogenet. Evol. ,  1998, vol.  9 (pg.  509- 516) Google Scholar CrossRef Search ADS PubMed  Fröbisch NB,  Carroll RL,  Schoch RR.  Limb ossification in the Paleozoic branchiosaurid Apateon (Temnospondyli) and the early evolution of preaxial dominance in tetrapod limb development,  Evol. Dev. ,  2007, vol.  9 (pg.  69- 75) Google Scholar CrossRef Search ADS PubMed  Fröbisch NB,  Reisz RR.  A new Lower Permian amphibamid (Dissorophoidea, Temnospondyli) from the fissure fill deposits near Richards Spur, Oklahoma,  J. Vertebr. Paleontol ,  2008, vol.  28 (pg.  1015- 1030) Google Scholar CrossRef Search ADS   Fröbisch NB,  Schoch RR.  The largest specimen of Apateon and the life history pathway of neoteny in the Paleozoic temnospondyl family Branchiosauridae,  Fossil Rec ,  2009, vol.  12 (pg.  83- 90) Google Scholar CrossRef Search ADS   Gradstein FM,  Ogg JG,  Smith AG. ,  A geologic timescale ,  2004 Cambridge (MA) Cambridge University Press Hanken J.  Appendicular skeletal morphology in minute salamanders, genus Thorius (Amphibia: Plethodontidae): growth regulation adult size determination, and natural variation,  J. Morphol ,  1982, vol.  174 (pg.  57- 77) Google Scholar CrossRef Search ADS   Hanken J.  Miniaturization and its effects on cranial morphology in plethodontid salamanders, genus Thorius (Amphibia, Plethodontidae): II. The fate of the brain and sense organs and their role in skull morphogenesis and evolution,  J. Morphol ,  1983, vol.  177 (pg.  255- 268) Google Scholar CrossRef Search ADS PubMed  Hanken J.  Miniaturization and its effects on cranial morphology in plethodontid salamanders, genus Thorius (Amphibia: Plethodontidae). I. Osteological variation,  Biol. J. Linn. Soc. ,  1984, vol.  23 (pg.  55- 75) Google Scholar CrossRef Search ADS   Hanken J.  Morphological novelty in the limb skeleton accompanies miniaturization in salamanders,  Science ,  1985, vol.  229 (pg.  871- 874) Google Scholar CrossRef Search ADS PubMed  Hanken J.  Hall BK. ,  Adaptation of bone growth to miniaturization of body size ,  1993 Boca Raton (FL) CRC Press(pg.  79- 104)  Bone Hanken J,  Wake DB.  Miniaturization of body size: organismal consequences and evolutionary significance. Ann. Rev. Ecol,  Syst ,  1993, vol.  24 (pg.  501- 519) Hay JM,  Ruvinsky I,  Hedges SB,  Maxson LR.  Phylogenetic relationships of amphibian families inferred from DNA sequences of mitochondrial 12S and 16S ribosomal RNA genes,  Mol. Biol. Evol. ,  1995, vol.  12 (pg.  928- 937) Google Scholar PubMed  Hedges SB,  Maxson LR.  A molecular perspective on lissamphibian phylogeny,  Herpetological Monographs ,  1993, vol.  7 (pg.  27- 42) Google Scholar CrossRef Search ADS   Hitchin R,  Benton MJ.  Congruence between parsimony and stratigraphy: comparison of three indices,  Paleobiology ,  1997, vol.  23 (pg.  20- 32) Google Scholar CrossRef Search ADS   Huelsenbeck JP.  Comparing the stratigraphic record to estimates of phylogeny,  Paleobiology ,  1994, vol.  20 (pg.  470- 483) Google Scholar CrossRef Search ADS   Huelsenbeck JP,  Ronquist F.  MrBayes (Bayesian analysis of Phylogeny). Version 3.1.1,  2005  Available from: http://mrbayes. csit.fsu.edu/download.php Huttenlocker AK,  Pardo JD,  Small BJ.  Plemmyradytes shintoni, gen. et sp. nov., an Early Permian amphibamid (Temnospondyli: Dissorophoidea) from the Eskridge Formation, Nebraska,  J. Vertebr. Paleontol ,  2007, vol.  27 (pg.  316- 328) Google Scholar CrossRef Search ADS   Kass RE,  Raftery AE.  Bayes factors,  J. Am. Stat. Assoc. ,  1995, vol.  90 (pg.  773- 795) Google Scholar CrossRef Search ADS   Laurin M,  Reisz RR.  Sumida SS,  Martin KL.  A new perspective on tetrapod phylogeny,  Amniote origins ,  1997 New York Academic Press(pg.  9- 59) Lee MSY,  Anderson JS.  Molecular clocks and the origin(s) of modern amphibians,  Mol. Phylogenet. Evol. ,  2006, vol.  40 (pg.  635- 639) Google Scholar CrossRef Search ADS PubMed  Lewis PO.  A likelihood approach to estimating phylogeny from discrete morphological character data,  Syst. Biol. ,  2001, vol.  50 (pg.  913- 925) Google Scholar CrossRef Search ADS PubMed  Linke R,  Roth G,  Rottluff B.  Comparative studies on the eye morphology of lungless salamanders, family Plethodontidae, and the effect of miniaturization,  J. Morphol ,  1986, vol.  189 (pg.  131- 143) Google Scholar CrossRef Search ADS   Maddison WP,  Maddison DR.  MacClade: analysis of phylogeny and character evolution,  1992 Sunderland (MA) Sinauer Associates Marshall LG,  Corruccini RS.  Variability, evolutionary rates, and allometry in dwarfing lineages,  Paleobiology ,  1978, vol.  4 (pg.  101- 119) Google Scholar CrossRef Search ADS   Miller PJ.  Miniature vertebrates: the implications of small body size,  1996 Oxford Clarendon Press Milner AR.  Small temnospondyl amphibians from the middle Pennsylvanian of Illinois,  Palaeontology ,  1982, vol.  25 (pg.  635- 664) Milner AR.  Benton MJ.  The relationships and origin of living amphibians,  The phylogeny and classification of the tetrapods ,  1988 Oxford Clarendon Press(pg.  59- 102) Milner AR.  The Paleozoic relatives of lissamphibians,  Herpetological Monographs ,  1993, vol.  7 (pg.  8- 27) Google Scholar CrossRef Search ADS   Müller J,  Reisz RR.  The phylogeny of early eureptiles: comparing parsimony and Bayesian approaches in the investigation of a basal fossil clade,  Syst. Biol. ,  2006, vol.  55 (pg.  503- 511) Google Scholar CrossRef Search ADS PubMed  Rieppel O.  Miller PJ.  Miniaturization in tetrapods: consequences for skull morphology,  Miniature vertebrates: the implications of small body size ,  1996 Oxford Claredon Press(pg.  47- 61) Roth G,  Dicke U,  Nishikawa K.  How do ontogeny, morphology, and physiology of sensory systems constrain and direct the evolution of amphibians?,  Am. Nat ,  1992, vol.  139 (pg.  105- 124) Google Scholar CrossRef Search ADS   Rüber L,  Kottelat M,  Tan H,  Ng P,  Britz R.  Evolution of miniaturization and the phylogenetic position of Paedocypris, comprising the world's smallest vertebrate,  BMC Evol. Biol. ,  2007, vol.  7 pg.  38  Google Scholar CrossRef Search ADS PubMed  Ruta M,  Coates MI,  Quicke DLJ.  Early tetrapod relationships revisited,  Biol. Rev. ,  2003, vol.  78 (pg.  251- 345) Google Scholar CrossRef Search ADS   San Mauro D,  Vences M,  Alcobendas M,  Zardoya R,  Meyer A.  Initial diversification of living amphibians predated the breakup of,  Pangaea. Am. Nat ,  2005, vol.  165 (pg.  590- 599) Google Scholar CrossRef Search ADS   Schoch RR.  Comparative ontogeny of early Permian branchiosaurid amphibians from southwestern Germany,  1992, vol.  222 (pg.  43- 83)  Palaeontographica, Abt. A Schoch RR.  McNamara KJ.  Heterochrony in the development of the amphibian head,  Evolutionary change and heterochrony ,  1995 Chichester (UK) John Wiley & Sons(pg.  107- 124) Schoch RR.  The early formation of the skull in extant and Paleozoic amphibians,  Paleobiology ,  2002, vol.  28 (pg.  278- 296) Google Scholar CrossRef Search ADS   Schoch RR.  The evolution of metamorphosis in temnospondyls,  Lethaia ,  2002, vol.  35 (pg.  309- 327) Google Scholar CrossRef Search ADS   Schoch RR.  Skeleton formation in the Branchiosauridae: a case study in comparing ontogenetic trajectories,  J. Vertebr. Paleontol ,  2004, vol.  24 (pg.  309- 319) Google Scholar CrossRef Search ADS   Schoch RR.  Skull ontogeny: developmental patterns of fishes conserved across major tetrapod clades,  Evol. Dev. ,  2006, vol.  8 (pg.  524- 536) Google Scholar CrossRef Search ADS PubMed  Schoch RR.  Evolution of life cycles in early amphibians,  Ann. Rev. Earth Planet. Sci. ,  2009, vol.  37 (pg.  15.1- 15.28) Google Scholar CrossRef Search ADS   Schoch RR,  Carroll RL.  Ontogenetic evidence for the Paleozoic ancestry of salamanders,  Evol. Dev. ,  2003, vol.  5 (pg.  314- 324) Google Scholar CrossRef Search ADS PubMed  Schoch RR,  Fröbisch NB.  Metamorphosis and neoteny: alternative pathways in an extinct amphibian clade,  Evolution ,  2006, vol.  60 (pg.  1467- 1475) Google Scholar CrossRef Search ADS PubMed  Schoch RR,  Milner AR.  Arratia G,  Wilson MVH.  Structure and implications of theories on the origins of lissamphibians,  Recent advances in the origin and early radiation of vertebrates ,  2004 Munich Verlag Dr. Friedrich Pfeil(pg.  345- 377) Schoch RR,  Milner AR.  The intrarelationships and evolutionary history of the temnospondyl family Branchiosauridae,  J. Syst. Palaeontol ,  2008, vol.  6 (pg.  409- 431) Google Scholar CrossRef Search ADS   Schoch RR,  Rubidge BS.  The amphibamid Micropholis from the Lystrosaurus Assemblage Zone of South Africa,  J. Vertebr. Paleontol ,  2005, vol.  25 (pg.  502- 522) Google Scholar CrossRef Search ADS   Shao K-T,  Sokal RR.  Tree balance,  Syst. Zool ,  1990, vol.  39 (pg.  266- 276) Google Scholar CrossRef Search ADS   Siddall ME.  Stratigraphic indices and tree balance: a reply to Hitchin and Benton,  Syst. Biol. ,  1997, vol.  46 (pg.  569- 573) Google Scholar CrossRef Search ADS   Siddall ME.  Stratigraphic fit to phylogenies: a proposed solution,  Cladistics ,  1998, vol.  14 (pg.  201- 208) Snively E,  Russel AP,  Powell GL.  Evolutionary morphology of the coelurosaurian arctometatarsus: descriptive, morphometric and phylogenetic approaches,  Zool. J. Linn. Soc. ,  2004, vol.  142 (pg.  525- 553) Google Scholar CrossRef Search ADS   Swofford D.  PAUP*: phylogenetic analysis using parsimony (* and other methods),  Version 4 0b10 ,  2001 Sunderland (MA) Sinauer Associates Trueb L,  Cloutier R.  Schultze H-P,  Trueb L.  A phylogenetic investigation of the inter- and intrarelationships of the Lissamphibia (Amphibia: Temnospondyli),  Origins of the higher groups of tetrapods: controversy and consensus ,  1991 Ithaca (NY) Comstock Publisher Associates(pg.  223- 313) Vallin G,  Laurin M.  Cranial morphology and affinities of Microbrachis, and a reappraisal of the phylogeny and lifestyle of the first amphibians,  J. Vertebr. Paleontol ,  2004, vol.  24 (pg.  56- 72) Google Scholar CrossRef Search ADS   Wake DB.  Adler K.  An integrated approach to evolutionary studies of salamanders,  Herpetology: current research on the biology of amphibians and reptiles. Proceedings of the 1st World Congress of Herpetology ,  1992 Oxford (OH) Society for the Study of Reptiles and Amphibians(pg.  163- 177) Wake DB.  Sanderson MJ,  Hufford L.  Introduction,  Homoplasy ,  1996 San Diego (CA) Academic Press(pg.  17- 25) Werneburg R.  Apateon dracyiensis - eine frühe Pionierform der Branchiosaurier aus dem europäischen Rotliegend, Teil1: Morphologie. Veröff. Naturhist. Mus,  Schleusingen ,  2001, vol.  16 (pg.  17- 36) Werneburg R.  Apateon dracyiensis - eine frühe Pionierform aus dem europäischen Rotliegend, Teil 2: Paläoökologie. Veröff. Naturhist. Mus,  Schleusingen ,  2002, vol.  17 (pg.  17- 32) Wiens JJ,  Bonett RM,  Chippindale PT.  Ontogeny discombobulates phylogeny,  Syst. Biol. ,  2005, vol.  54 (pg.  91- 110) Google Scholar CrossRef Search ADS PubMed  Wills MA.  Congruence between phylogeny and stratigraphy: randomizationn tests and the gap excess ratio,  Syst. Biol. ,  1999, vol.  48 (pg.  559- 580) Google Scholar CrossRef Search ADS   Zardoya R,  Meyer A.  On the origin of and phylogenetic relationships among living amphibians,  PNAS ,  2001, vol.  98 (pg.  7380- 7383) Google Scholar CrossRef Search ADS PubMed  Zhang P,  Zhou H,  Chen Y-Q,  Liu Y-F,  Qu L-H.  Mitogenomic perspectives on the origin and phylogeny of living amphibians,  Syst. Biol. ,  2005, vol.  54 (pg.  391- 400) Google Scholar CrossRef Search ADS PubMed  © Society of Systematic Biologists
TI - Testing the Impact of Miniaturization on Phylogeny: Paleozoic Dissorophoid Amphibians
JF - Systematic Biology
DO - 10.1093/sysbio/syp029
DA - 2009-07-03
UR - https://www.deepdyve.com/lp/oxford-university-press/testing-the-impact-of-miniaturization-on-phylogeny-paleozoic-knaXyzfcmv
SP - 312
EP - 327
VL - 58
IS - 3
DP - DeepDyve
ER -