Reptile relationships turn turtle⃛

Abstract

Turtles are so anatomically bizarre that their affinities with other reptiles remain contentious. Wilkinson et al.¹ showed how even the extensive morphological information amassed by Rieppel and deBraga^2,3 provides only weak support for their view that turtles are advanced diapsid reptiles, rather than descendants of primitive anapsid reptiles, as conventionally thought. But there seem to be errors in Rieppel and deBraga's data matrix, many involving turtles or their putative ‘anapsid’ relatives^4,5. I have corrected and reanalysed the data (see ‘Incorrectly coded characters’, overleaf), and find that Rieppel and deBraga's data actually support, rather than challenge, the traditional view.

Main

The modified data yield a tree where diapsid monophyly is restored and turtles are related to anapsid pareiasaurs (Fig. 1a). This result is consistent with other recent analyses^4,5. In particular, the tree is almost identical to that proposed in another detailed phylogenetic analysis of the entire Reptilia⁶ (Fig. 1b). The large impact of these apparently minor corrections to Rieppel and deBraga's data set is not surprising.

Figure 1: Alternative reptilian phylogenies.

a, The single most parsimonious tree (length 763 steps, consistency index 0.511, retention index 0.695), based on an analysis of the data matrix of Rieppel and deBraga2,3, after corrections detailed in footnote. Turtles (Testudines) are nested within ‘anapsid’ parareptiles, as the nearest relatives of pareiasaurs (Anthodon, Scutosaurus, Bradysaurus), and diapsid monophyly is restored. b, Phylogeny obtained from an independent study of higher-level reptile phylogeny6. Note the close similarity to a. Clades in common are indicated by horizontal bars. In both trees, diapsids are monophyletic and turtles are nested within parareptiles.

In their phylogeny (Fig. 1 in ref. 2) almost all of the characters interpreted as supporting turtle-diapsid affinities (those diagnosing clades 3 to 7) also occur in all or many anapsid ‘parareptiles’, or are absent (presumed reversed) in turtles. Thus, only a slight modification to the data caused turtles to shift from deep within diapsids (as lepidosaur relatives) to deep within parareptiles (as pareiasaur relatives).

My results, together with those of Wilkinson et al.¹, emphasize the importance of evaluating just how strongly a preferredtree is supported over alternatives. Rieppel and deBraga have identified a surprisingly strong phylogenetic signal linking turtles and advanced diapsids. However, the conventional views regarding the anapsid derivation of turtles and the monophyly of diapsid reptiles need not yet be abandoned.

Incorrectly coded characters. I have provisionally accepted codings for other disputed characters (for example the acromion process^2,3,4,5) where there is disagreement over the interpretation of homologies but room for debate. The data set was analysed using the same methods and outgroup rooting as refs 2 and 3. Characters are numbered as in ref. 3. 44-46, Stapes unknown in Bradysaurus and Anthodon. 65, Basal tubera on basioccipital and basisphenoid present in all pareiasaurs⁵. 70, Rieppel and deBraga interpret the anterior braincase ossification in the primitive turtle Proganochelys as unequivocally a sphenethmoid, but then code turtles as lacking the element, instead of as polymorphic. 73, Interpterygoid vacuity closed in Bradysaurus and slit-like in Anthodon⁷. 77, 82, Transverse flange of pterygoid oriented anterolaterally, and jaw joint located anterior to occiput, in turtles (primitively)^8,9. 87, Lateral shelf of surangular present in all pareiasaurs (for example, Figs 5-viii, 6 and Plate 33 of ref. 10). 103, Cervical centra keeled in all pareiasaurs (for example, Bradysaurus BMNH R1971, Scutosaurus PIN 2005/1532-8, Anthodon BPI 1/548). 120, 121, Two coracoids present in synapsids¹¹. Rieppel and deBraga suggest that the acromion in turtles might represent a modified anterior coracoid. If so, the coracoid foramen in turtles lies totally within the coracoid homologues. Under their interpretation, turtles must be coded as equivocal for both coracoid number and coracoid foramen position. 124, 126, 127, Humeral shaft short, supinator process absent, and ectepicondylar foramen present (primitively) in turtles⁸. 140, Fourth trochanter absent in all pareiasaurs⁵. 150, Well-defined articulation between astragalus and fourth distal tarsal present in Scutosaurus (for example PIN 2005/1877), condition unknown in Bradysaurus and Anthodon. 152, First distal tarsal present (primitively) in turtles^8,12. 160, Metapodials overlapping proximally in Bradysaurus (for example, plate 37 of ref. 13) and Scutosaurus (for example PIN 2005/1532), condition unknown in Anthodon. 164, Long unguals primitively present in turtles^8,9.