Acheloma cumminsi, COPE, 1882

Gee, Bryan M, 2020, Size matters: the effects of ontogenetic disparity on the phylogeny of Trematopidae (Amphibia: Temnospondyli), Zoological Journal of the Linnean Society 190 (1), pp. 79-113 : 105-109

ACHELOMA CUMMINSI COPE, 1882

Synonyms: Trematops milleri Williston, 1909 ; Trematops thomasi Mehl, 1926 ; Trematops willistoni Olson, 1941 ; Trematops stonei Olson, 1970 ; Ach. dunni Polley & Reisz, 2011 .

Holotype: AMNH FARB 4205, partial skull with articulated mandibles and associated postcranial skeleton.

Type locality: Arroyo Formation, Clear Fork Group. Specimen collected from Coffee Creek in Baylor County, TX, USA.

Revised diagnosis: Trematopid diagnosed by the following autapomorphy: (1) presence of a dentulous ridge on the vomer extending anteroposteriorly. The following autapomorphy is reliably found only in adults: lateral exposures of the palatine (LEP) and of the ectopterygoid (LEE) excluded from the orbit by a lacrimal–jugal contact. Differentiated from E. cutlerensis , F. striegeli and Ma. laticeps by: presence of internarial fontanelle; elongation of the snout, with a naris that is substantially longer than the orbit; otic notch that is dorsoventrally constricted and with reduced lateral exposure of the supratympanic flange; choana prominently expanded anteromedially to form a ‘Y’-shaped contour.

Remarks: The diagnosis differs from that of previous workers (e.g. Polley & Reisz, 2011; Schoch & Milner, 2014) in being more conservative in light of ontogenetic considerations. The differential diagnosis is formed only by comparison with other large-bodied trematopids that may be interpreted as relatively mature (‘adults’). The motivations and philosophy for this are guided by discussion in the next section of this paper. The following features may be diagnostic or differential (and, in most instances, have previously been suggested as such), but cannot be determined definitively to be autapomorphies, because they occur only in adults of Acheloma and because of an absence of specimens of comparable maturity in the closely related Phonerpeton : (1) choana prominently expanded anteromedially (convergent with some highly nested amphibamiforms) to form a ‘Y’-shaped contour; (2) otic notch with a nearly horizontal ventral margin and a prominent dorsoventral constriction by the quadratojugal that largely obscures the posteroventral portion of the supratympanic flange in lateral profile; (3) preorbital region much longer than postorbital region; (4) elongate naris, much longer than orbit; and (5) posterior skull table transversely expanded (shared with F. striegeli ). The lateral exposures of the palatal bones are regarded as a more reliable feature, because there is no evidence for an ontogenetic transformation from exposures along the ventral orbital margin (e.g. Phonerpeton , many dissorophids) to those seen in Ach. cumminsi , in which they are either absent early in ontogeny or separated from the orbit in large individuals ( Gee et al., 2019a). The adopted approach of specifying that it is an autapomorphy possibly only expressed in more mature individuals distinguishes it from the vomerine ridge, which is known across the ontogenetic range of specimens.

The validity of both species of Phonerpeton is less controversial (or at least has not been stated as such in the literature), but it can be questioned along slightly different lines that were discussed briefly by Gee et al. (2019a). Phonerpeton was named by Dilkes (1990) for material from three different formations (Archer City, Nocona and Petrolia) in Texas, which represent a stratigraphic succession (listed from oldest to youngest). These formations are older than the Clear Fork Group, from which material of Ach. cumminsi is known and from which only one specimen of cf. Phonerpeton sp. (MCZ 2475) has been described ( Dilkes, 1993). Dilkes described only one species, P. pricei , which included ‘ Ach. pricei ’ [from the Archer City Formation (formerly Putnam)] and ‘ Ach. whitei ’ [from the Petrolia Formation (formerly Wichita)] of Olson (1941). Schoch & Milner (2014) split Phonerpeton into two species, P. pricei and P. whitei , thus following Olson’s original species-level distinctions and Dilkes’ genus-level distinction from Acheloma . As with Olson (1941), P. pricei is currently restricted to material from the Archer City Formation, whereas P. whitei is restricted to material from the Petrolia Formation. Dilkes (1990) also referred one specimen (MCZ 1485) from the Nocona Formation to P. pricei . This specimen was incompletely described and figured by Bolt (1974b) and was most recently suggested to be a new species of Anconastes by Schoch & Milner (2014) following Berman et al. (1987). Its scoring in this analysis (based on incomplete description by Bolt, 1974b, c) is nearly redundant with that of the most complete specimen of P. pricei (AMNH FARB 7150) and differs only in a few non-overlapping missing data entries. As such, the specimens cluster in the total morphological evidence permutation (83% of MPTs) and in the taxonomic reduction permutation (100% of MPTs) in the midst of an overall lack of resolution within Trematopidae ( Fig. 3; Supporting Information, Fig. S1). The rationale for taxonomic differentiation that was provided by Berman et al. (1987), such as the absence of an internarial fontanelle and the absence of lateral exposures of the palatal bones, conflicts with the only description and figuring of the specimen by Bolt (1974c), in which these features are present. However, having not examined the specimen personally and in the absence of any figures of the specimen beyond those of Bolt (1974c), it is not possible to reconcile these discrepancies here.

Since its resurrection, P. whitei has seldom been discussed. Schoch & Milner (2014) provided only a diagnosis of the taxon, without redescription or figures. This is unfortunate, because the only previous description and illustration is that of Olson (1941), which is similarly brief and nondescript. Sutural patterns, if preserved, are not apparent. Schoch & Milner (2014) diagnosed P. whitei by two features: (1) external naris shorter and more rounded; and (2) otic notch not slit-like. This diagnosis is suspect, because both features are influenced by ontogeny. Dilkes (1993) demonstrated that narial elongation and subdivision were ontogenetically influenced features in long-snouted trematopids. This can be corroborated by examination of specimens of various sizes from the Richards Spur locality attributed to ‘ Ach. dunni ’ ( Polley & Reisz, 2011). The formation of a more constricted slit-like otic notch has been suggested to be a derived feature that is unique to Acheloma ( Berman et al., 2011) . A similar condition, but without the greatly restricted lateral exposure of the notch, is also seen in the large referred specimen of Ma. laticeps and in the referred specimens of E. cutlerensis ( Berman et al., 1985; Dilkes & Reisz, 1987; Polley & Reisz, 2011; Milner, 2018). Along with the various types of most Acheloma and ‘ Trematops ’, all of these specimens exceed 12 cm in midline skull length. Small trematopid specimens <8 cm in length, such as those attributed to either species of Phonerpeton , the holotype of R. gothae and small specimens of Ach. cumminsi , have less constricted notches, with a distinct posterolateral angling and a large lateral exposure, suggestive of an ontogenetic influence ( Dilkes, 1990; Berman et al., 2011; Gee et al., 2019a).

The stratigraphic separation of P. pricei (Archer City Formation) and P. whitei (Petrolia Formation) is thus the most compelling support for their taxonomic separation and, as discussed above, stratigraphic occurrence is not a diagnostic attribute. Given that both species occur in Texas, it might be considered that P. whitei is an anagenetic derivative of P. pricei , but this is difficult to validate in the fossil record, because no morphological standard exists for establishing taxonomic differentiation. Furthermore, P. whitei is more primitive in features, such as the absence of a subdivided naris, and would thus represent a less parsimonious (although not impossible) reversal to a more primitive morphology based on its higher stratigraphic occurrence. I have not examined material of P. whitei personally and thus do not feel qualified to make a conclusive determination of the validity of the species. However, the morphological distinctions are both ontogenetically influenced attributes, stratigraphic occurrence is not diagnostic, and this analysis does not provide compelling evidence for their phylogenetic distinction. The separation of these taxa should be regarded sceptically until P. whitei can be redescribed in greater detail.

CONSIDERATIONS FOR TAXONOMIC DIAGNOSES

The discussion of P. whitei encapsulates the broader challenges associated with diagnosing small-bodied taxa that possess features suggestive of ontogenetic immaturity rather than body size diminution (dwarfism) or early diverging phylogenetic position. A similar point on problems of ontogeny and taxonomy was previously discussed by Gee et al. (2019a) regarding how to properly refer juvenile specimens to taxa that are represented and diagnosed by larger specimens. A point that I explore in this section is the utility and validity of diagnoses made by comparing ontogenetically disparate taxa. Outside Acheloma and Phonerpeton , only Ma. laticeps is represented by both small- and large-bodied specimens. The taxonomic separation of most trematopids is not in question, although Milner (2019) queried whether An. vesperus and E. cutlerensis , both from the Early Permian Cutler Formation of New Mexico, are synonymous. As noted above, this analysis does not recover tree topologies that support this synonymy.

Even when the separation and validity of taxa is well supported, caution should still be exercised with respect to diagnosing these taxa. It is self-evident that many differences, some of which relate to different states of phylogenetic characters, can be identified between juveniles and adults of the same taxon, and this disparity can be expanded when comparing probable juveniles of one taxon with probable adults of another.This is particularly salient when using species-level OTUs (the traditional permutation) for which intraspecific ontogenetic variation may be masked by coding only the ‘adult’ condition and for which interspecific ontogenetic disparity is not explicitly taken into account. It is unsurprising, for example, that the highest nested long-snouted trematopid is also the largest: Ach. cumminsi . Although this might be the true evolutionary position of this taxon, it is undoubtedly strengthened by being compared with the distinctly smaller and distinctly immature Phonerpeton and R. gothae . This urges caution when evaluating phylogenies that use species-level OTUs for which the number of in-group taxa is relatively limited (either biologically or methodologically), because both resolution and nodal support might be artificially exaggerated by exclusion of data.

Artefacts associated with ontogeny can be exacerbated further when assessments of ontogenetic maturity rely on assumptions of conserved ontogenetic trajectories such that one taxon may be used as a guide for another. However, this assumption may not always be true (e.g. the long-snouted Ach. cumminsi is probably not a good model for development of the short-snouted Act. peabodyi ), and such assumptions should be both tested more directly and explicitly stated when used. These considerations are broadly applicable to palaeontology and are salient when features are proposed to be autapomorphic or when features are known to be influenced ontogenetically in at least some closely related taxa.

Some purported autapomorphies might appear to be autapomorphies only because the closest related taxa are not known from comparably mature specimens or because of heterochrony. For example, an anteromedial expansion of the choana is a purported autapomorphy of Acheloma ( Schoch & Milner, 2014) , but small specimens of Acheloma and all specimens of Phonerpeton possess the primitively oval choana ( Olson, 1941; Dilkes, 1990; Gee et al., 2019a). Given the similarities between Acheloma and Phonerpeton in overall skull shape and anatomical features at similar skull lengths, it might also be predicted that expansion of the choana could occur in Phonerpeton and simply be unknown, because no skulls> 7.5 cm (and apparently no ‘adults’) are known in this taxon. Milner (2019: 217) briefly discussed FMNH UC 481, an unpublished skull measuring 15 cm from the Archer City Formation, which might represent a potential adult of Phonerpeton , but this specimen has never been figured or described and requires further first-hand examination to contextualize properly. A previous study on the Late Triassic metoposaurid temnospondyl Apachesaurus gregorii Hunt, 1993 , which also appears to be represented only by juvenile specimens ( Gee & Parker, 2018), recommended critical examination of purported diagnostic features for ontogenetic influence and that such features should generally be avoided in diagnoses unless comparably mature specimens are known. I followed this recommendation in the above revised diagnosis for Ach. cumminsi .

I reiterate here that I am not proposing a departure from making diagnoses based on mature specimens (or the most mature specimen available), but these considerations are particularly important if an intended holotype is not mature. For example, it is not possible to diagnose Ach. cumminsi confidently with respect to Mo. calliprepes , which is represented only by a holotype skull measuring <4 cm in length. Workers intending to erect new taxa based on immature individuals should form the diagnosis based on comparably mature individuals. For example, some of the diagnostic features of Phonerpeton sensu Schoch & Milner (2014) are not diagnostic when compared with individuals of Acheloma of a similar size, such as a large, posteriorly open otic notch. It should also be examined critically whether a distinctly immature individual can be differentiated sufficiently in the context of the incomplete fossil record to warrant novel taxonomic distinction.

Diagnoses must account fully for the current understanding of ontogenetic transformations within a clade, in particular for taxa represented only by small-bodied and immature taxa. Questionable taxonomy that results from a failure to account for ontogeny (among other variables, such as non-ontogenetic intraspecific variation) has consequences beyond the validity of taxa. Poorly supported taxonomy can have wide-ranging influences on other palaeobiological studies. Assessments of broader macroevolutionary and ecological trends, such as biodiversity and rates of character evolution, can be confounded by improper interpretations of morphological diversity and disparity. These considerations are hardly exclusive to trematopids or to temnospondyls and should be considered carefully as part of future taxonomic and phylogenetic work. Where ontogeny simply cannot be characterized owing to gaps in the fossil record, more conservative approaches are recommended.

CONSIDERATIONS FOR TEMNOSPONDYL PHYLOGENY

This study is the first to examine the phylogeny of a temnospondyl clade using specimen-level OTUs and is one of the first to test explicitly for the influence of ontogeny on phylogenetic reconstruction. Depending on perspective, temnospondyl phylogeny may be regarded as an ideal case study for exploring phylogenetic inference or an area badly in need of revision with respect to the effects of ontogenetic disparity on phylogeny (these are not mutually exclusive). Temnospondyls are a morphologically diverse clade, with a long temporal duration (Carboniferous to Cretaceous) that would extend to the present day if their putative affinities to Lissamphibia are accepted. The clade exhibits a remarkable diversity of lifestyles associated with a similarly broad range of ontogenetic trajectories (e.g. Schoch, 2009). Body sizes were also extremely diverse, ranging from amphibamiforms of a comparable size to extant caudates to capitosaurs of crocodilian proportions. Unsurprisingly, temnospondyl phylogeny remains weakly resolved in many areas and at various taxonomic levels, with disparity between studies produced by different working groups (e.g. Yates & Warren, 2000; Pawley, 2006; Ruta, 2009; McHugh, 2012; Schoch, 2013; Pardo et al., 2017). This additional disparity poses problems that remain to be addressed directly, because it further limits the reliability of assumptions when comparing taxa of disparate life histories and for reconciling morphological disparity; a small neotenic taxon almost certainly does not reach the same morphological landmarks or progress along the same developmental timeline as a large terrestrial taxon.

In the context of the present study, ecological and life-history disparity are less substantial. Mattauschia laticeps , here recovered as an early diverging trematopid (at least when some measure of in-group resolution is achieved), measures 12 cm in skull length. This is larger than virtually all non-trematopid dissorophoids and the third largest trematopid after Ach. cumminsi and E. cutlerensis . This suggests that relatively large body size appeared early in trematopid evolution and that the markedly small size of many Carboniferous-age holotypes in comparison to the larger Permian forms is not necessarily a hallmark of early trematopids. Furthermore, there is no evidence for a diversity of ecologies among trematopids that would hint at marked heterochrony. Generally speaking, body size and ecology are conserved within comparable temnospondyl clades (e.g. Angielczyk & Ruta, 2012). However, a counterpoint to this can be observed in the sister clade to Trematopidae , the Dissorophidae . The smallest dissorophids are represented by type specimens ~ 5 cm in skull length (e.g. Carroll, 1964), but the largest members achieved skull lengths> 30 cm (e.g. Gubin, 1980; Liu, 2018). The specific case of dissorophids is probably distinct from that of trematopids, because the largest dissorophids are all Middle Permian taxa (suggesting a trend of increasing body size into the later part of the Permian), but this remains to be tested in a fashion similar to the present study. Dissorophoidea is even more diverse, with a great diversity of small-bodied aquatic micromelerpetids and branchiosaurids in addition to dissorophids and trematopids, and is itself a microcosm of Temnospondyli in its entirety. Assessing the effects of both real (heterochronic) and artificial (gaps in the fossil record) disparity will be important for further resolving the intrarelationships of temnospondyls. To date, few studies have attempted such in-depth analyses, and those that do have focused on the diminutive amphibamiform dissorophoids as the putative closest relatives of lissamphibians (e.g. Fröbisch & Schoch, 2009; Pérez-Ben et al., 2018). Examining the effects of secondarily reduced body size (e.g. lapillopsids, ‘latiscopids’) and secondarily acquired paedomorphosis remains a largely unexplored area and might have major implications for both the evolutionary history of temnospondyls and the origins of the lissamphibian clades.

FUTURE DIRECTIONS FOR TREMATOPID PHYLOGENY

T h e p r e s e n t s t u d y h a s d o n e m o r e t o i d e n t i f y limitations and shortcomings of existing methods and materials than to provide robust, discrete information regarding the evolution of trematopids. However, what prospects exist for future research into this clade? From a methodological perspective, the use of continuous characters and data merits consideration. This approach has essentially never been applied to any temnospondyl clade (but for continuous data binned into discrete characters in an analysis of Metoposauridae , see Chakravorti & Sengupta, 2019). Likelihood approaches remain relatively underused for both dissorophoids and temnospondyls at large and should be explored further. From a perspective of the availability of material, re-examination of poorly described historical specimens and specimens of disputed taxonomy might clarify some trematopid relationships and morphological patterns. ‘Discovery’ and examination of unpublished specimens, such as the possible adult of Phonerpeton mentioned by Milner (2019) and undoubtedly others not mentioned in the literature, could shed additional light on ontogenetic patterns or presently unrecognized morphological diversity. Unfortunately, even if these approaches are undertaken, the current state of trematopid systematics is greatly hindered by the absence of specimens beyond the holotype for many taxa and the apparent immaturity of some of those holotypes. It may well be that without additional collection of material, for which there is no assurance, the phylogeny and evolutionary history of trematopids cannot be explored further.

CONCLUSIONS

It is a widely held assumption that OTUs represented by juvenile individuals will be recovered in earlier diverging positions within a phylogeny founded on morphological characters. However, this assumption is rarely tested explicitly among extinct tetrapods. It is particularly important to examine this directly in temnospondyls, which exhibit a diverse set of ecologies, body sizes and life histories over> 200 Myr of evolution (and possibly further, given the proposed ties to Lissamphibia). Even if other studies address the same question in other systems and validate such assumptions, the specific nuances in which these mechanisms operate will vary with the study system and its associated developmental processes. Although the absence or underdevelopment of features is typically associated with immaturity and a more primitive morphology, there are certainly exceptions that are often attributed to heterochrony (e.g. paedomorphosis) and that are nevertheless coded into phylogenetic matrices. An example is the absence of the sphenethmoid (a relatively late-ossifying element) in immature specimens of terrestrial taxa, such as trematopids. Conversely, the persistent absence of a sphenethmoid throughout ontogeny in many aquatic taxa is regarded as a derived condition in many secondarily aquatic, and often paedomorphic, Mesozoic clades, which illustrates one example of a character for which the derived condition in one group may be the primitive and juvenile condition in another. Heterochrony has been documented to exert discernible influences on lissamphibian phylogeny (e.g. Wiens et al., 2005), and similar considerations should be explored further within temnospondyls, given a distinctly greater range of body size and the diversity of life histories.

In this study, I have provided a direct test of the effects of ontogenetic disparity on the phylogenetic reconstruction of the Trematopidae using a specimen-level analysis. This study is the first either to sample a temnospondyl clade using specimen-level OTUs or to assess the influence of ontogenetic disparity on the phylogenetic reconstruction directly. Various permutations that included the traditional species-level analysis produced differing topologies with various degrees of resolution, most of which lacked support for in-group relationships. Given the relatively complete character sampling and the complete taxon sampling, these findings suggest that various factors may be affecting the topology. It is naturally more difficult to either prove definitively or rule out causal mechanisms or to assess the relative degree of influence of multiple mechanisms in deep time, because there is no alternative source of data to reconcile conflicting or unresolved topologies. Although this study did not recover evidence for a clear directional bias created by ontogenetic disparity, I propose that it does provide a degree of support for the hypothesis that ontogenetic disparity exerts an observable influence on trematopid phylogeny (e.g. correlation of size with recovered phylogenetic position among long-snouted taxa). Ontogenetic disparity among the highly limited specimens of short-snouted trematopids might also contribute to the lack of resolution of their intrarelationships, although this requires additional specimens of differing maturity to the currently known ones (e.g. adults of R. gothae ) to examine further. Previous trematopid analyses have achieved resolution but with a distinctly limited taxonomic sample. Therefore, it should be considered that the present sample of trematopid specimens is insufficient for resolving the phylogeny of the clade confidently.

This study emphasizes the importance of exploring potentially confounding variables in analyses of other early diverging tetrapod clades through closer, explicit examination of the sensitivity of phylogenetic methods, such as OTU sampling (e.g. Anderson, 2001; Kearney & Clark, 2003) and character construction ( Harris et al., 2003; Simões et al., 2017), and independent testing of the structure and the support of tree topologies by multiple, independent working groups (e.g. Marjanović & Laurin, 2019). Trematopids are not unique among Palaeozoic temnospondyls or other tetrapod clades in being represented by a paucity of specimens that capture only limited fragments of a more substantial ontogenetic range. However, they are probably not an ideal candidate for more philosophical tests of phylogenetic methodologies, such as the applicability of parsimony vs. likelihood approaches for morphological datasets, given the findings of poor resolution in this study amid ontogenetic interference. Future work must carefully examine and directly address such considerations to ensure a high fidelity of phylogenetic inferences and taxonomic diagnoses.