
Fiigure 1. The turtle mimic Eunotosaurus from the Middle Permian was actually closer to Acleistorhinus.
Before yesterday,
the best images of the skull of Eunotosaurus were simple drawings (Fig. 1). Now, thanks to Bever et al. (2015) we have CT scans of several skulls in several views (Fig. 2). For those we are all most grateful!

Figure 2. Click to enlarge. Eunotosaurus from CT scans. Left column, DGS tracings with restoration to in vivo positions. Middle column, CT scans from Bever et al. Right column, top, reconstructions by Bever et al. Bottom, juvenile skull with incorrect scale bar and correct scale bar.
Bever et al. reports, “If turtles are closely related to the other living reptiles then we would expect the fossil record to produce early turtle relatives with diapsid skulls. That expectation remained unfulfilled for a long time, but we can now draw the well-supported and satisfying conclusion that Eunotosaurus is the diapsid turtle. The skull of Eunotosaurus grows in such a way that its diapsid nature is obvious in juveniles but almost completely obscured in adults.”
The important word
in that paragraph is “IF.” Turtles are related to other living reptiles, but they are not either kind of diapsid (lepidosaur or archosaur). And, as you’ll see, the skull of Eunotosaurus does not grow in such a way that its diapsid nature is obvious. Bever et al. made several mistakes in their observations and analyses, as will be shown.
Bever et al. note that “Eunotosaurus africanus, a 260-million-year-old fossil reptile from the Karoo Basin of South Africa, whose distinctive postcranial skeleton shares many unique features with the shelled body plan of turtles.
What Bever et al. fail to note
is that such an armored postcrania has evolved several times independently within the Reptilia. In addition to
Eunotosaurus and
turtles there are
Sinosaurosphargis, Henodus and
Cyamodus, not to mention ankylosaurs, aetosaurs, armadillos and pangolins.
Bever et al. note: “Scepticism regarding the status of Eunotosaurus as the earliest stem turtle arises from the possibility that these shell-related features are the products of evolutionary convergence.”
True.
The
large reptile tree indicates these shell-related features are convergent in the above-named taxa.

Figure 3. Sister taxa according to Bever et al. Eunotosaurus purportedly nests between the diapsid, Ascerosodontosaurus, and the turtles. The large reptile tree, on the other hand, finds that only the turtles are related to each other. Click to enlarge.
Bever et al. note: “Our phylogenetic analyses indicate strong cranial support for Eunotosaurus as a critical transitional form in turtle evolution. Furthermore, we find unexpected evidence that Eunotosaurus is a diapsid reptile in the process of becoming secondarily anapsid.”
That’s big news.
Worthy of a Nature paper, if valid. At it happens, secondary anapsids do occur in
mesosaurs and
ankylosaurs, but not turtles. The large reptile tree indicates turtle ancestors were always anapsids. And that
Eunotosaurus and kin were all synapsid in skull morphology, but were not related to the real synapsids: pelycosaurs + therpasids + mammals (pan-mammals). Bever et al. did not test enough taxa or pertinent taxa. Most importantly, even if
Eunotosaurus juveniles did have an upper temporal fenestra (Figs. 4, 5), they did so alone. All their closest sisters did not and adult
Eunotosaurus did not have the upper temporal fenestra, but instead had a bone in that area.

Figure 4. Juvenile Eunotosaurus showing purported left upper temporal fenestra left by an absent supratemporal.
Bever et al. note: “This is important because categorizing the skull based on the number of openings in the complex of dermal bone covering the adductor chamber has long held sway in amniote systematics, and still represents a common organizational scheme for teaching the evolutionary history of the group.”
Held sway?? What Bever et al. fail to note
is that several of these skull morphologies arose more than once by convergence. This fact is not commonly known, but recovered in the large reptile tree. Remind yourself not to follow anyone who holds “held sway” up as evidence.

Figure 5. Eunotosaurus juvenile right side showing purported absent supratemporal. At right a possible displaced supratemporal is identified as bump on the otherwise humpless parietal that pretty well fits the opening left by its taphonomic shift.
Bever et al. note: “Our results suggest that Eunotosaurus represents a crucially important link in a chain that will eventually lead to consilience in reptile systematics, paving the way for synthetic studies of amniote evolution and development.”
Actually
Eunotosaurus currently represents a dead end in the large reptile tree.
The Bever et al figure 1
misidentifies several bones on the skull of
Proganochelys. They mistakenly follow the original bone identifications, where were changed
here.
- The postorbital + postfrontal was mistakenly labeled ‘postorbital.’
- The tabular was mislabeled ‘supratemporal.’
- The supratemporal was mislabeled ‘squamosal.’
- The squamosal was mislabeled ‘quadratojugal.’
- The quadratojugal was not separated from the jugal.
In support of this,
the long
Eunotosaurus supratemporal is actually homologous to the mislabeled ‘squamosal’ (actually the supratemporal) of
Proganochelys (same shape, same placement), despite their separation on the large reptile tree. Noting that would have added evidence to the Bever et al claims, but they let that one pass and stuck to the old paradigm. Basal turtles, like
Elginia and
Meiolania, and pre-tuirtles like
Sclerosaurus and Arganaceras, had supratemporal horns, not squamosal horns.
The Bever et al cladogram (their figure 1) has several errors.
- Eothyris is not a member of the Pan-mammals (Synapsida), but is in the Milleretta and Casea clade.
- The term ‘Parareptilia’ does not represent a monophyletic clade.
- Youngina does not belong on this cladogram of several Lepidosauromorpha, but belongs on another branch, the Archosauromorpha.
- Elginia is a more basal and earlier stem turtle than Proganochelys, but is omitted in the cladogram.
- Some of the drawings used are inaccurate, but that may matter only a wee bit.
Bever et al. report, “The skull of Eunotosaurus is relatively short and wide. Its compact snout bears approximately 23 robust, subthecodont marginal teeth and nasals that are longer than the frontals.” Actually the nasals are subequal to the frontals. You can measure them yourself.
Bever et al. report a “moderately sized suborbital fenestra,” but I did not see any space between the ectopteryoid and palatine, only between the palatine and anterior pterygoid. In any case, the palate of Eunotosaurus does not resemble the palate of basal turtles.
Bever et al. report,
“A lower temporal arcade is not present, leaving the cheek open ventrally.” Actually the adult jugal has a very tiny quadratojugal process. In the juvenile (Figs. 4, 5) the jugal has a large process that invades the lateral temporal fenestra, as in
Acleistorhinus and
Deloryhnchus, the verified sisters of
Eunotosaurus. We just looked at that new process
here.
Bever et al report, “The roof of the adductor chamber in a juvenile skull is marked by a distinct upper temporal fenestra (on both sides.”
Actually there are several new fenestra in the juvenile skull (Fig. 4). And where is the supratemporal? On the left, the supratemporal appears to be crushed down into, rather than arching over the other temporal bones. On the right (Fig. 5) the supratemporal appears to have shifted medially. Now it looks like an unintended bump. So these purported ‘upper temporal fenestra’ were not really present in vivo. Furthermore, there are no genuine diapsids that nest close to Eunotosaurus in the large reptile tree.
In the Bever et al. phylogenetic studies
(Fig. 3) Eunotosaurus nests between basal Sauropterygia + Acerosodontosaurus and the turtles, Proganochelys + Odontochelys. Many suprageneric taxa were used, so there were many opportunites to cherry pick traits to fit a preconceived idea. That does not happen in the large reptile tree where every taxon is represented by a specimen or species. The large reptile tree is still fully resolved, and breaks down into fully resolved subsets.
Bever et al. report, “The adductor chamber (in Eunotosaurus) is closed posteriorly by what appears to be an expanded tabular, though this bony plate may represent a posterolateral flange of the parietal. The putative tabular fills a space corresponding to what would otherwise be a moderately sized post-temporal fenestra.” That is a tabular.
Bever et al write, “The presence of a lower temporal fenestra in Eunotosaurus supports the hypothesis that the characteristically closed cheek of modern turtles is a secondary condition that evolved through expansion of the jugal, quadratojugal, and squamosal.”
Actually
the lower (lateral) temporal fenestra in
Eunotosaurus is homologous with that of
Acleistorhinus and kin. Turtle ancestors never had a lateral temporal fenestra according to the large reptile tree, but came closest with
bolosaurs.
Removal
of
Acleistorhinus and
Delorhynchus from the large reptile tree nests
Eunotosaurus with its other sisters,
Feeserpeton and
Australothyris. One look, or deep study, and you’ll see why. Removal of those taxa nests
Eunotosaurus between cousins
Milleretta and the casesauria. Removal of those taxa nests
Eunotosaurus with second cousins,
Limnoscelis,
Orobates and kin. Still no change in tree topology. Removal of these taxa finally nests
Eunotosaurus between
Stephanospondylus and
Bashkyroleter, each at the base of their own clades with a slight change in tree topology. Still
Eunotosaurus does not nest with turtles, Ascerosodontosaurus or basal sauropterygians.
At least some of the problem
with the Bever et al. report rests with their use of prior phylogenetic analyses, and/or accepting the results of those analyses without their own critical analysis of recovered sister taxa. For instance, the swimming
Claudiosaurus and they gliding
Kuehneosauridae appear as sister taxa, while
Claudiosaurus and another similar swimmer,
Hovasaurus, do not. Those are problems. Nesting
Diadectomorpha outside the Amniota is a problem.. Nesting bulky
Rhynchosaurs as relatives of gliding Kuehneosaurs is also a problem. None of these problems exist in the large reptile tree, where all sister taxa look alike and display gradual accumulations of derived traits. Perhaps Bever et al. should have created their own matrix from scratch, as in the large reptile tree.
Bever et al. write: “Although it is the case that neither our parsimony nor Bayesian analyses recover the dominant molecular solution of a unique turtle–archosaur relationship, both approaches do agree that turtles arose somewhere within the greater diapsid radiation.”
No one knows
why turtle DNA is close to archosaur DNA. It is not supported by morphology.
Bever et al. write: “Singular expression of a lower temporal fenestra once unambiguously diagnosed mammals and their stem lineage, but is now recognized in a number of phylogenetically disparate ‘anapsid’ parareptiles, and may represent the ancestral
condition for the amniote crown. That is a false statement. The ancestral condition in
basal amniotes was anapsid. Such a statement by Bever et al. is one based on hope, not observation.
Bever et al. write: “This observed pattern of concentrated homoplasy near the evolutionary origin of a character state is congruent with the concept of a protracted zone of variability that may commonly confound attempts to resolve the early history of clades and character systems” \This is observed only in bad cladograms. There is no such thing as a “protracted zone of variability.” At least I haven’t seen one. How would you measure it?
Bever et al. write,
“The amniote upper temporal fenestra has enjoyed a famously uncomplicated history, being a nearly consistent fixture of the diapsid body plan since its first appearance in the Carboniferous.” This is another false statement, according to the large reptile tree. The so-called Diapsida had
two origins. Bever et al are working with outmoded data and small taxon sets.
Bever et al. write,
“The combined morphologies expressed in the juvenile and adult specimens of Eunotosaurus provide not only the earliest direct evidence of an upper
temporal fenestra in a putative stem turtle, but the first evidence for how that fenestra may have closed before the origin of the turtle crown clade.” Except the juvenile does not have an upper temporal fenestra (careful examination reveals otherwise) and so it cannot close at this point. Bever et al did not examine the specimen carefully enough to find the taphonomically displaced supratemporals. Turtle embryos do not recapitulate this change during ontogeny. No sister taxa have an upper temporal fenestra. This is just a unique fossil that has successfully tempted a set of paleontologists into accepting its illusion, just like the pterosaur,
Sordes and the thalattosaur,
Vancleavea.
Bever et al write: “The role of the supratemporal is an aspect of the Eunotosaurus model of temporal closure that would not have been predicted based on earlier studies.” No shit, Sherlock. “An anteriorly expanded supratemporal that develops late in postnatal development to cover the upper temporal fenestra must currently be considered an autapomorphy of Eunotosaurus.” Be wary of autapomorphies. Evolution does not create many of those.
Morevover, what is wrong with the logic here?
According to Bever et al. the giant supratemporal closed off the upper temporal fenestra in Eunotosaurus, but the supratemporal is a tiny bone at the back of the skull in Proganochelys. The several misidentifications in Proganochelys are noted above.
Bever et al. answer this logic conundrum with,
“but it is important to stress that an expanded supratemporal is not a necessary component of an evolutionary model of fenestral closure in turtles that has Eunotosaurus as a central figure. For example, an analogous secondary expansion of the supratemporal partially or completely covers the upper temporal fenestra of the marine thalattosaurs.” This is false. The upper temporal fenestra in thalattosaurs is not reduced by expansion of the supratemporal, which remains small and not involved. Rather the upper temporal fenestra in thalattosaurs is reduced to a slit or less by the straightening (loss of curvature) of the parietal and postorbital rims.
Bever et al. further speculate, “Moreover, the construction of the modern turtle skull played out over a time span of tens of millions of years, and it is well attested that dermal bones in
major vertebrate lineages can shift back and forth considerably relative to the underlying tissues on these timescales.” As you have seen so far, Bever et al. have no idea how the turtle skull evolved, based on evidence from the large reptile tree. They have not included the correct ancestral taxa, but have pulled similar taxa from other clades. They are working in a hopeful vacuum of outmoded analyses, small taxon lists and incorrect bone identities.
Bever et al. conclude, “It is thus evident that the turtle skull, like the turtle postcranium,
underwent profound modifications during its history that similarly obscured anatomical evidence for phylogenetic affinities by the time the crown-group condition was reached.” This is, as you know by now, a conclusion based on mistake after mistake.
Finally,
let us remind ourselves that turtles developed their own method for creating space for larger temporal adductors. It seems illogical that they would first lose this space by early closure of the upper temporal fenestra only to reacquire additional space by anterior expansion of the posttemporal fenestra and subsequent skull emargination.
But logic is not a factor here.
This is evolution. Phylogenetic analysis, as recovered in the large reptile tree, provides the story of how the turtle got its skull… and shell… and all of its other accessories.
And if the juvenile Eunotosaurus did have missing supratemporals creating quasi and temporary upper temporal fenestrae, like the fontanelle in the baby human skull, then the diapsid status of turtles is still not verified. Turtles have no diapsid relatives when enough pertinent taxa are included in analysis.
References
Bever GS, Lyson TR, Field DJ and Bhular B-A S 2015. Evolutionary origin of the turtle skull. Nature published online Sept 02. 2015.