T-U-R-T-L-E Power! Part 2: The weird and wacky origin of turtles.

Facebooktwitterredditpinterestlinkedintumblrmailby feather
A Galapagos tortoise struts about, secure in the knowledge that no one will ever know where the hell it came from. Photo from petcaregt.com

This is a long overdue follow up to my original Turtle Power article back in…yeah never mind the date.

As established previously, turtles are a strange, and highly diverse group of animals, but how did they come to be this way?

The turtle bauplan has been a phylogenetic double edged sword. On the one hand, the unique shell design, and the necessary body contortions associated with it, make chelonians a very easy group to classify. However, it is these same peculiarities that keep us from finding the ancestor to turtles. To date, there are no “half-turtles.” No good transitionals between one reptile group to that  of turtles. As such, the list of turtle ancestors runs all over Reptilia. Some paleontologists believe the origin lies at the base with reptiles like procolophonoids, and pareiasaurs. Others believe turtles are a bit more closely related to extant reptiles, and belong in, or alongside the sauropterygians (plesiosaurs, nothosaurs, and placodonts). There is even some evidence to suggest turtles are actually in the same reptile group as dinosaurs and crocodylians (Archosauria).

Curse that shell! - Photo by tompain

How can the list be this extensive? Read on to find out.

Since the turtle bauplan is so contorted from that of “standard” reptiles, researchers are forced to look at the least changed body parts (e.g. aspects of the skull), or to look for fossils that show signs of a turtle-like development (overlapping ribs, extensive osteoderms, etc). This limited selection of traits increases the frustration of chelonian phylogenetics as the character traits that have been retained, may be found in too many other taxa to help narrow down the possible relatives.

Parsing out life by holes in the head

A common method for distinguishing which amniote group was related to which, was to use the presence, or absence of holes (fenestrae) found on the skull. This method gave us two of the most commonly uttered taxonomic terms in vertebrate paleontology: synapsids and diapsids.

The trick to figuring out who goes in which camp was to look behind the orbits at the temporal region. Here a distinct set of fenestrae usually developed. They provided room for muscle bellies to bulge, allowing for the evolution of greater jaw closing power. The location of temporal fenestrae actually serve as a splendid example of how evolution can go about solving the same problem in multiple different ways.

The plesiomorphic, or archaic / primitive condition was to have a solidly roofed skull with no fenestrae whatsoever. This condition was deemed: anapsid, or “without arches.” From here at least three different conditions arose. One condition saw the evolution of a single temporal fenestra low and to the side of the skull — a condition referred to as synapsid (“one arch”). Synapsids are the group that includes mammals.  Another condition evolved that resulted in two temporal fenestrae. One up high on the skull, and another lower down and to the side. This group was the diapsids (two arches). This is the group that the majority (all?) of extant reptiles belong to. Lastly another condition developed that saw the evolution of a single temporal fenestra again, but this time it was up high, rather than low and on the side of the skull. This group was called the euryapsids (wide arch).

The classic diapsid condition as modeled by a tuatara (_Sphenodon_).
_Dimetrodon_ models the synapsid condition.

Though still taught today, the arch system of classification has fallen out of favour as the arches themselves have proven to be more variable than we once thought. For instance, the euryapsids are now believed to have been modified diapsids that just lost their lower arch. Further, in most extant squamates, there is usually only one arch present (and in many, no arch at all).

Nonetheless, this is where the initial placement of turtles as very early reptiles, came about. The turtle skull is considered to be solidly roofed over, thus making turtles anapsids, and placing them among the pareiasaurs, procolophonoids and other obscure animals.

The Usual Suspects: Anapsids

Loggerhead sea turtle (_Caretta caretta_) showing off its anapsid (or is it?) skull.

Even with the eschewing of the arch system of classification, there were other traits that linked turtles to early parareptiles. The presence of broad, flat ribs in Eunotosaurus africanus were originally viewed as a shared character with extant turtles. The broad ribs formed a sort of carapace that would have limited vertebral bending, and forced E.africanus to move in a tortoise like fashion (Sumida & Modesto 2001).  The loss of the tabular and ectopterygoid bones in turtles has been used as evidence for placing them close to the captorhinids (Carroll 1969). These were small, lizard-like reptiles that roamed the Permian of Pangea.

Could _Owenetta_ be the ancestral turtle? - Photo from Wikipedia
_Bradysaurus_ a pareiasaur...and an ancestral turtle? - Photo from Wikipedia

A reappraisal of the procolophonoid: Owenetta, lead the authors Reisz and Laurin (1991) to determine that this little critter was actually an ancestral turtle. The authors noted 10 diagnostic characters that Owenetta shared with testudines. Most notable was the formation of the tympanic notch.  The ear of chelonians is rather unique in how it is formed. Turtles lack a tympanum (the thin membrane that allows for the transmission of airborne vibrations) and have, instead, evolved a tight layer of skin over their tympanic opening that performs much the same task. This rearrangement seems to be related to the evolutionary movement of the jaw muscles through time. Because of this rearrangement, the otic capsule in turtles is formed by a unique compilation of bones (the squamosal and an enlarged quadratojugal). This makes turtle ears similar to mammal ears in terms of their diagnostic potential.  That Owenetta also possessed a similar ear, while no other Permian reptile group, or diapsid group did, became strong evidence for a procolophonoid ancestry for turtles.

However, other researchers were not so quick to agree. Michael Lee of the University of Queensland, did a phylogenetic study of turtles in relation to parareptiles and came to the rather fascinating conclusion that turtles were actually dwarf pareiasaurs (Lee 1993, 1997). Pareiasaurs were large terrestrial reptiles. Some grew to 3 meters (9.8ft) in length and weighed half a tonne. They had stocky bodies, erect hind limbs, tiny tails and sculpted heads.  From a glance they almost did seem a bit like turtles without the shell. Lee found that pareiasaurs, much like turtles, were a group united by a suite of unique traits, but that these traits could also be found in other animals. In particular, Lee cited the thickened braincase floor, acromian process, dermal armour, 4 presacral ribs, among others. Though turtles probably did not descend from the larger pareiasaurs, Lee does think that they might have been a separate offshoot of dwarf pareiasaurs “in which  adolescence and maturity came on long before the  normal  height  had  been  reached  but not  before great strength,  especially in the  limbs,  had  been  developed. (Lee 1997: pulled from Gregory 1946).

The Contenders: Diapsids

The apparent synapomorphies found between chelonians and the parareptilian anapsids did not go unchallenged. Olivier Reippel and Michael deBraga conducted a phylogenetic study of a wide array of reptilian taxa, and came to the conclusion that turtles were not anapsids at all, but rather a group of modified diapsids that closed off their temporal openings (Reippel and deBraga 1996).

In their initial test of turtle origins, Reippel and deBraga rightly pointed out that previous phylogenetic studies had been too narrowly focused. The authors questioned the assumption that turtles were anapsids. This assumption coloured the testing of turtle phylogenies by focusing on parareptiles only. When Reippel and deBraga ran their analysis they discovered, much to the shock of many, that turtles nested within the diapsids.  The authors found similarities in the jaw adductor musculature, temporal bone ossification, jugal shape and the presence of a centrale among the carpal bones instead of a radiale.

The modified diapsid (formerly euryapsid) skull of a plesiosaur. Was this a close relative to turtles?

Further work by Reippel and Reisz (1999) found that turtles nested particularly close, and even within the sauropterygia. In particular, turtles seemed to place rather closely around the placodonts; a group that has been affectionately referred to as reptilian walruses (Orenstein 2001). Given that the sauropterygians were the dominant sea critters of the Mesozoic, and that sea turtles made up a substantial portion of the sea life during the Cretaceous, as well as the common description for elasmosaurid plesiosaurs as “snakes threaded through the bodies of turtles,” the placement of Testudines among the sauropterygians almost seems apt. Furthermore, it might help explain the particularly perplexing issue of the evolution of the plastron.

As discussed in other parts of the site, the turtle shell is composed of both dorsal, and ventral armour; the carapace and plastron respectively. The issue with evolving a plastron is why bother unless one is worried about being attacked from below?  Barring super tall animals like giraffes and sauropods, attacks from below are pretty rare among terrestrial animals. Most other terrestrial vertebrates that have evolved armour (armadillos, ankylosaurs, various crurotarsans, certain lizard species) all have armour on their back and sides, but none on the belly region. At the very least it seems unlikely that a terrestrial animal would evolve ventral body armour.

Now if the animal had its start in the sea that would be a completely different matter. Marine life is a true 3D affair. Attacks can come from above, below, and from the sides. It would seem that ventral body armour would be most likely develop in this medium. Adding further support for this is the presence of certain placodont species that looked awfully close to turtles. Remember it is the placodonts that are believed to be the sauropterygians closest to turtles.

Species like Placochelys placodonta and Henodus chelyops looked so similar to turtles that one would be hard pressed to tell them apart from a distance.

So there we go. That must be it right? The sauropterygian origin has a statistically better phylogeny behind it, more diagnostic characters, known species that look an awful lot like turtles, and a habitat that explains that problematic plastron.

Chinks in the Armour (the Problem of Convergence)

The argument from Reippel and others are honourable in that they force paleontologists to question long held assumption and avoid dogma for fear of how it can colour one’s findings. However, as elegant as the diapsid argument is, some glaring problems arise upon closer inspection.

While it is true that attacks from below are rare in terrestrial animals, it doesn’t mean its impossible. A well armoured animal may do well to hunker down and wait out an attack from a predator of smaller, or equal size. However, what’s to stop larger predators from flipping a small, but well protected animal on its belly? If turtles did descend from an animal like Eunotosaurus africanus, then perhaps the plastron developed as a way to counter the stiffened vertebral column, which would have prevented the animal from curling up like a modern day armadillo (or armadillo lizard).

Another thing the plastron can do is confer greater strength to the carapace by way of the bony connective bridge. This locks both sides of the carapace together, preventing any splaying and creating a much stronger structure (Orenstein 2001).  It remains possible that the plastron evolved on land after all.

_Henodus chelyops_ sure looked like a turtle...on the outside. - photo from Dinosaurs and other prehistoric animals Könemann

Then there is the issue of H.chelyops and P.placodonta. Both animals are extremely turtle like animals, but neither animal was a turtle. Both Henodus and Placochelys evolved their shell by incorporating plates of osteoderms — a dermally derived bone. As discussed in part I, the chelonian shell is a unique mixture of dermal and endochondral bone. Despite their extreme similarity, neither placodont incorporated endochondral bone into their shell. What we have here is an impressive case of convergence.

Convergence is commonplace in the history of life. Evolution might find multiple ways to solve the same problem, but it also has a tendency to reinvent the wheel a lot too. This reinvention can be a real problem for phylogenetic systematics. For a lot of phylogenetic programs, convergence is viewed as a last resort. While this tends to keep findings honest for small cases of homoplasy, cladistic trees can wind up producing false relationships when dealing with taxa that show extreme convergence.

Which brings us back to the diapsid argument.  If one removes sauropterygians from a phylogenetic analysis of turtle origins, turtles wind up returning to the parareptilia (Reippel and Reisz 1999). Could it be, then, that the only reason turtles are showing up in the diapsids at all is because of the extreme convergence seen in some members of the group?

Archosauria Comes in from Left Field

Sauropterygia wasn’t the only contender to the testudine throne. Phylogenetic work by Blair Hedges and Laura Poling (1999) found that turtles nested rather comfortably within…archosaurs?

The authors compared 24 nuclear genes and mitochondrial DNA of representatives from all the major extant reptile groups. Their results showed that — molecularly — turtles shared more in common with crocodiles than they did with squamates. This result forced chelonians into the archosaur branch of the reptilian family tree.

Other molecular phylogenies produced similar results (Kumazawa & Nishida 1999, Rest et al 2003).  Turtles would routinely show up as archosaurs using molecular systematics.

I believe I’ve made my thoughts clear on molecular phylogenetics before. The big problem with many molecular phylogenies is that they tend to lack morphological support. While the position of turtles had been unclear before, there was never any morphological support for an archosaurian affinity for these guys. Further, some of these molecular phylogenies (Hedges & Poling 1999) were also finding tuataras (Sphenodon) to nest with archosaurs too, and we have a pretty good idea of where tuataras fall on the reptile family tree.

Why are all these molecular datasets producing such a strange trees? One likely culprit behind the association of turtles with archosaurs is the phenomenon of long-branch attraction. This is the production of an erroneous relationship by a cladistic program due to the assignment of homology to similar traits found in two rapidly evolving lineages. Long branch attraction is most evident in molecular phylogenies due to the inherent limitation of DNA sequences (you only have 4 base pairs to work with).  In cases with two groups that are actually widely separated from one another, the random accumulation of identical base pairs is enough to draw the two branches together on a phylogenetic tree (Bergsten 2005).  It is likely that this, more than anything else, is what seems to keep drawing turtles and archosaurs together.

The other big problem with molecular phylogenetics, and one that will likely remain a big problem, is that it is limited to extant taxa only. DNA sequences rarely survive past 10,000 years. Turtle ancestry goes back at least 230 million years. Having to rely on sequence data from relatives that separated from each other over 230 million years ago is bound to create problems.

Current Consensus?

Interestingly, in the past two years we have had some surprising developments on the turtle ancestry front. New taxa have been discovered, and evolutionary development studies have even thrown their hat in the ring.

So what have we discovered?

Odontochelys Plays for Sauropterygia

The "new" old guy. _Odontochelys semitestacea_. - Photo by Marlene Donnelly

In 2008, Chun Li and others discovered the fossil of a small marine critter from the lower Carnian of China (~225 mya). They named the animal Odontochelys semitestacea : the “toothed turtle on the half shell” (Li et al 2008). This animal was special for two reasons. One, at the age it was found in, O.semitestacea was now the oldest known turtle, and from the diagnostic material collected it was definitely a turtle. O.semitestacea predated Proganochelys quenstedti (the former oldest known turtle) by 5 million years.

The other important feature about Odontochelys semitestacea was in its species epithet. The animal had only half a shell. After decades of searching, scientists had finally discovered the ever elusive “half turtle.” Perhaps now one could answer the all important question of who the ancestor was?

It turned out that the plastron was the first part of the shell to develop. This fossil evidence agreed with previous evo-devo studies that showed the same thing in developing turtle embryos (Reippel 1993). The carapace showed only rudimentary development with neural plates being present, but remaining separate from the neural spines. The ribs in O.semitestacea were expanded and resembled the ribs in embryonic snapping turtles (Li et al 2008).  This, along with the marine sediment Odontochelys was found in, suggests that the turtle bauplan did start in an aquatic environment.

When Li et al incorporated O.semitestacea into their phylogenetic analysis, they found turtles to root pretty nicely next to sauropterygians; well within Diapsida.

So turtles are diapsids.

Proganochelys Pinch Hits for Archosauria

The "old" old guy. _Proganochelys quenstedti_ - Photo from the Staatliches museum.

Not long after the O.semitestacea discovery, a new paper by Bhullar and Bever (2009) purported to have found the first ever morphological evidence for an archosaur relationship of turtles. The authors went back and looked at the former oldest turtle Proganochelys quenstedti to see if they could find anything particularly archosaurian about the animal. What they discovered was one bone in particular,that happened to be very archosaur like in shape: the laterosphenoid.

Doing a detailed comparative analysis of the laterosphenoid in P.quenstedti and archosaurs, the authors found that both bones looked nearly identical. Other ancient turtles like Kayentachelys aprix are also said to have possessed such a bone. Modern turtles have since obliterated the laterosphenoid, and it was not preserved in Odontochelys semitestacea, which makes this singular bone hard to use for diagnosis. Nonetheless, the authors did incorporate it into a phylogenetic test of turtle origins. In an unconstrained tree (where convergence was allowed), turtles were found to be a sister group to archosauromorphs, while in a constrained tree (laterosphenoid evolves only once), they became sister taxa to archosauriformes. This weak congruence found in these trees was not lost on the authors, as they suggested the need for much more work in this arena before one could say for certain one way, or the other.

So there you have it. An admittedly weak argument, but the first one to actually use morphological data to say that:

turtles are archosaurs.

Eunotosaurus Drags Turtles Back to Parareptilia

Just a couple of weeks ago a surprising turn of events occurred. A new paper by Tyler Lyson and others suggested that turtles might still be anapsids after all.

The granddaddy of all turtle ancestors: _Eunotosaurus africanus_.

Taking a nod from Li et al’s description of Odontochelys semitestacea, the authors set out to do a phylogenetic analysis that incorporated more stem, or “transitional” taxa. Rather than recode everything, the authors used the data from Li et al’s phylogeny (where turtles came out as sauropterygians), but they included Proganochelys quenstedti and Eunotosaurus africanus. The former was a definite early turtle, while the latter had once been considered an ancestral turtle relative. Their results were quite surprising.  By simply including transitional taxa into the Li et al matrix, turtles were pulled all the way back to parareptilia. Not only that, but they seemed to form a sister group relationship with E.africanus. In other words, by incorporating more data into their phylogeny, the authors were able to obtain a result that echoed the initial placement of turtles over 90 years ago.

In the process of doing this, the authors noted something else that was rather important. The ribs of E.africanus were not only expanded, but T-shaped in cross section. What does that mean? According to the authors, this suggests that expansion of the ribs in E.africanus occurred via a secondary ossification, and the incorporation of dermal components. Both are hallmarks of turtle shell development. It would appear that E.africanus may have been an ancestral turtle after all.

The big take-home message of Lyson et al’s paper can be summed up in this quote from their paper ([ ] = mine):

…although some morphological support exists for all three hypotheses [parareptiles, sauropterygians, archosaurs], the overriding signal in the major morphological datasets actually agree and converge on turtles as nested within ‘parareptiles’ and not diapsids

The marked topological shift in the Li et al (2009) analysis following the addition of two extinct species re?ects the importance of transitional fossils when long branches separate the origins of major crown clades

So there you have it, turtles are parareptiles.

Where do we go now?

So who are the ancestors of turtles? As it currently stands, turtles seem to be anapsid parareptiles. Whether that makes them procolophonoids like Owenetta, pareiasaurs, or a completely different group more closely related to Eunotosaurus remains to be answered.

Even then, that is just the current position. As Lyson et al mentioned, the inclusion of different taxa, can vastly change the topology of a phylogeny. Turtle phylogenetics serve as a sobering reminder that phylogenetic trees are just hypotheses on the relationships of different taxa. They are ultimately the products of the data given, and the assumptions used. Both of these things need to be kept in mind when viewing a phylogenetic tree. Even trees that show high bootstrap support, or Bayesian values, are not necessarily correct.  They just show that they are statistically sound for the data, and assumptions given to them.

So where does that leave us with turtles?

Well if you ask me, the answer to turtle ancestry is obvious.

They’re aliens. Why else would they be saucer shaped?  >: )

~ Jura

Next time: Not sure just yet, but T-U-R-T-L-E Power will continue. The next installments will mostly focused on case study taxa.


Bergsten, J. 2005. A Review of Long-Branch Attraction. Cladistics. vol.21:163-193.
Bhullar, B-A.S. and Bever, G.S. 2009. An Archosaur-Like Laterosphenoid in Early Turtles (Reptilia: Pantestudines). Breviora. vol.518:1-10.
Carroll, R.L. 1969. A Middle Pennsylvanian Captorhinomorph and the Interrelationships of Primitive Reptiles. J.  Paleo. vol.43:151-170.
Gregory WK.  1946.  Pareiasaurs Versus Placodonts as Near Ancestors to the Turtles. Bull.  AMNH. vol.86:276-323
Hedges, S.B. and Poling, L.L. 1999. A Molecular Phylogeny  of  Reptiles.  Science  283:998-1001
Kumazawa, Y. and Nishida, M. 1999. Complete Mitochondrial DNA Sequences of the Green Turtle and Blue-Tailed Mole Skink: Statistical Evidence for Archosaurian Affinity of Turtles. Mol. Biol. Evol. vol.16(6):784-792
Lee, M.S.Y. 1993.  The  Origin  of  the  Turtle  Body Plan: Bridging  a  Famous Morphological  Gap. Science. vol.261:  1716  1720.
Lee, M.S.Y. 1997. Pareiasaur Phylogeny and the Origin of Turtles. Zoo. J. Linn. Soc. vol.120:197-280
Lyson, T.R., Bever, G.S., Bhullar, B-A.S.,Joyce, G.J., GAuthier, J.A. 2010. Transitional Fossils and the Origin of Turtles. Biol. Lett. doi:10.1098/rsbl.2010.0371
Orenstein, R. 2001. Turtles, Tortoises & Terrapins: Survivors in Armor. Firefly Books. Buffalo, NY. 304 pgs. ISBN: 1-55209-605-x
Rieppel, O. 1993. Studies on Skeleton Formation in Reptiles. Patterns of Ossification in the Skeleton of Chelydra serpentina Linnaeus (Reptilia, Testudines).  J. Zool. vol.231: 487–509
Reippel, O. and deBraga, M. 1996. Turtles as Diapsid Reptiles. Nature. vol.384(5):453-455
Reippel, O. and Reisz, R.R. 1999. The Origin and Early Evolution of Turtles. Ann. Rev. Eco. Syst. vol.30:1-22
Reisz, R.R., and Laurin, M. 1991. Owenetta and the Origin of Turtles. Nature. vol.349:324-326.
Rest, J.S., Ast, J.C., Austin, C.C., Waddell, P.J., Tibbetts, E.A., Hay, J.M., Mindell, D.P. 2003. Molecular Systematics of Primary Reptilian Lineages and the Tuatara Mitochondrial Genome. Mol. Phyl. Ev. vol.29(2):289-297
Facebooktwitterredditpinterestlinkedintumblrmailby feather

6 Responses to T-U-R-T-L-E Power! Part 2: The weird and wacky origin of turtles.

  1. Avatar Apollonios
    Apollonios says:

    One is the answer. Turtles came after complex mutations brought by significant amounts of radiation were released into the Earth’s environment, when Proto-Greeks/Apollonians with their flying disks found the way to go back in time. There, with no humans existing yet, they could perform wild nuclear wars against the other dark side without damaging chthonic humans. As a byproduct of these imence nuclear war, enormous quantities of radiation was flooding periodically the Earth’s environment, causing strange mutations. Now, not only the origin of turtles can be found, but other obscure ancestries of other groups of organisms, such as flowering plants and the natural nuclear reactors present on earth two billions of years ago in Africa can be easily explained. The existence today of living fossils might be the result of transportal of life between the different ages of the earth by apollonian disks.

  2. Hey, it was great to see you up at the lab.

    Anyway, my 2 cents would be that existing phylogenetic analyses to tackle the problem of turtle origins suffer from taxon and character sampling biases. What I’ve found over the past few weeks while working on coding taxa into Mueller’s (ultimately derived from earlier Rieppel matrices) analysis is that the codings for Testudines don’t gel well with well known basal turtles such as Proganochelys and Kayentachelys. I’m currently working on coding a lot of different taxa related to turtle origins as part of a collaborative project… so we’ll see how things go. It’s worth noting that the archosauromorph contender paper didn’t find turtles as archosauromorphs except without using a constraint to force the relationship.


  3. aww man, this is what I get for writing at 3:30 in the morning on little sleep, I just noticed you already touched on the issue of constraints… so yeah, one of the salient points of the paper suggesting a close relationship based on the laterosphenoid isn’t a well supported statement within the paper itself. I can’t recall off the top of my head and my refs aren’t handy, but some parareptiles may also show some a laterosphenoid (pleurosphenoid) – like ossification (Lee 1995).

  4. It was nice to finally meet in person. I have to admit, when I heard you were coming down it actually reminded me that I needed to get the second part of this thing done already.

    The matrix you are working on sounds neat. It should be interesting to see where it turtles wind up landing this time.

  5. Your link about your thoughts on molecular phylogenetics leads to a post about komodo dragons, which, while interesting and worth reading, doesn’t address molecular phylogenies. Your tuatara link has some comments, but not what I think you were referring to.

    • Yeah, I’m not exactly certain how YARPP categorizes things. I think it just picks up keywords and assumes relations from there. It seems to work okay for most of the posts, but every once in awhile it seems to make what appears to be a random post selection.