An Assault on the Large Reptile Tree (dino section)!

Note added after publication: For those interested the comments section sheds new light. 

Mickey Mortimer, blogger of the Theropod database, put a lot of work into rescoring the dinosaur portion of the large reptile tree. With those changes, here is the recovered tree (taxa abbreviated but those in the know will know).

|–Turfano
`–+–Gracili
`–+–Arizona
`–+–+–Loto
|  `–+–Popo
|     `–+–Effi
|        `–Shuvo
`–+–+–Pseudolago
|  `–Sile
`–+–Mara
`–+–Pisano
`–+–+–Scelido
|  `–+–Heterodonto
|     |–Hexinlu
|     `–+–Agili
|        `–+–Lesotho
|           `–Scutello
`–+–+–Masso
|  `–+–Theco
|     `–Saturn
`–+–Daemono
`–+–+–Panphag
|  `–Pampa
`–+–Herrera
`–+–Trial
|–SMNS
|–Tawa
`–Coelo

You’ll note that Saurischia is recovered with [Sauropodomorpha + Theropoda]. Panphagia + Pampadromaeus nest basal to Theropoda, and the [poposaurids + Arizonasaurus] nest close to the basal archosaur/crocodylomorph, Gracilisuchus. Ornithischians are in the middle.

A segment of the large reptile tree, focused on the Dinosauria.

Figure 1. A segment of the large reptile tree, focused on the Dinosauria.

These are different nestings from the large reptile tree (Fig. 1). Basically the Dinosauria has been flipped top to bottom.

Well, then, let’s test the Mortimer tree
You’ll recall when I removed taxa, clades, or large numbers of clades from the large reptile tree, the rest of the topology did not change. When I tested only skulls or only post-crania, the tree did not change. These are signs of stability and strength.

However,
when I removed the [poposaurs + Arizonasaurus] from the Mortimer tree, the topology reverted to that of the large reptile tree. That’s a big change. And there’s more:

Note
In the Mortimer tree theropods nest as the most derived clade of dinos, preceded by several clades of herbivores in this pattern:

Quoting from my reply to MM: “Turfanosuchus and Gracilisuchus (carnivores) at the base giving rise to Poposaurids (herbivores), then Silesaurids (herbivores), then Marasuchus (carnivore), then Pisanosaurus (herbivore), then a split between Ornithischians (herbivores) and Saurischians led by Sauropodomorphs on one branch (herbivores) and Daemonosaurus (?-vore) at the base of Panphagia + Pampadromaeus (herbivores) and theropods (carnivores). As you can see this is a varied mix of herbies and carnies, which is not the case in the Large Reptile Tree”

In the LRT the theropods nested as basal dinos, giving rise via Panphagia, Pampadromaeus and Daemonosaurus to herbivorous sauropodomorphs, poposaurids and ornithischia. Basically the same tree, just flipped top to bottom.

Not sure yet what ordering, scoring, what-have-you turned the Mortimer tree upside-down, but having theropods as derived and widely separating Gracilisuchus from Herrerasaurus raises red flags. It just ain’t right. Isn’t it more tenable that certain theropods gave rise to herbivores, first with saurischian pelves and later with ornithischian pelves?

The unstable topology is also bothersome. Removing taxa should not upset the topology. Hopefully we’ll come to a resolution on this. The devil is probably in the details many or all of which can be seen here at M. Mortimer’s blog post.

Arizonasaurus does not belong here. It nests with rauisuchians and shares few to  no synapomorphic traits with Gracilisuchus, a basal crocodylomorph.

I appreciate the work by MM and, perhaps, some of the rescoring is justified. I don’t know. In any case, the results do not appear to be tenable.

As always, I encourage readers to see specimens, make observations and come to your own conclusions. Test. Test. And test again.

Evidence and support in the form of nexus, pdf and jpeg files will be sent to all who request additional data.

8 thoughts on “An Assault on the Large Reptile Tree (dino section)!

  1. Excerpted/expanded from my reply to you…

    Note “my” tree is still mostly yours- I haven’t changed any postcranial characters or codings, and only corrected two taxa (Turfanosuchus and Marasuchus) for all cranial characters when I sent you the file. So the matrix you’re complaining is unstable is really yours with half the character formulations fixed, 17 characters added and 1.4% of the entries changed.

    Also Poposaurus and Daemonosaurus were carnivores (recurved, compressed teeth with small serrations), so your tree has to deal with that homoplasy too.

    But more importantly, the reason your tree is so stable is because you miscode taxa so that PAUP thinks they are strongly supported as related. For instance, codings supporting your unconventional Theropoda with Marasuchus, Trialestes and SMNS 12352 include-

    1. Marasuchus having a skull width <1.2 times its height. How did you code that when all we have are lateral views of the braincase, squamosal, partial postorbital and a medial view of the maxilla?

    2. Trialestes having a skull width <1.2 times its height. How did you code this when all we have is a lateral view of the ventral skull?

    3 Trialestes having a maxillary palatal process. How could you code this when we only have a lateral view of the maxilla, which is sutured to the premaxilla fragment?

    4. Pantydraco ("Thecodontosaurus") and Pisanosaurus lacking a maxillary palatal process. How could you code these given the anterior maxilla for each is missing, and that's where the process would be?

    5. Herrerasaurus having an anteriorly pinched interchoanal rod. You used Sereno and Novas' (1993) dotted hypothetical vomer for this. How could you code it when they specifically say the anterior vomer isn't visible?

    6. Trialestes having an anteriorly pinched interchoanal rod. How can you code this when we know nothing of the palate of Trialestes?

    7. Marasuchus having a nasal with parallel sides. How can you code this when Marasuchus doesn't preserve nasals, and only has a medial view of the maxilla?

    8. SMNS 12352 having a naris angled less than 30 degrees downwards. How can you code this when the specimen doesn't preserve any part of the external naris?

    That's just a small sampling (Turfanosuchus has 33 miscodings, Gracilisuchus has 32, Arizonasaurus has 53, Trialestes has 26, Marasuchus has 24, SMNS 12352 has 17. Each out of 115), but that's why you get such stable results. Once Trialestes, Marasuchus and SMNS 12352 are corrected, they all go out of Dinosauria like in traditional phylogenies. If you want to defend your results, start by defending these nine codings of yours.

  2. Okay, first inning results. Some of the questionable scorings are due to restorations: following the rostral profile to terminate a premaxilla, for instance in SMNS 12352. Some are due to codes that added my mistake and are now gone, thank you. Turfanosuchus, not that it matters, does not have a flat skull table because the upper temporal bars are lower than the midline of the parietal. Turfanosuchus does not have an occiput far posterior to the quadrate (I’m gonna guess that was a typo on your part, and it doesn’t matter anyway because that character is not shared here). In Marasuchus not all, but some of the maxillary teeth are twice as deep as long, the smaller ones generally. Those are some of your miscodings. So neither of us is perfect.

    Adding Arizonasaurus, a rauisuchid, to “our” tree is not appropriate as it is a rauisuchian according to the LRT and was not in the original test set of dinos.

    Of the skull traits it looks like you used all of mine (115) added 113 more, all unlabeled, or are those my 113 postcranial traits for a total of 228? That would make sense. And 36 added labeled traits, from Nesbitt, I suppose.

    Altogether your 10 MPtrees had 453 steps, but if you rearrange the topology, flipping the theropods and poposaurids to match the LRT it takes only 9 more steps (462 without re-scoring your re-scorings). Delete the poposaurids and your data produces my tree, not yours.

    So I know and suspect there are some miscodings on both sides. But I have to tell you your results, nesting Theropods as more derived than other dinosaurs, should have told you there’s something wrong.

    My hat is off to you for your efforts, though not perfect. Neither were mine.

    • Well, this might be the most useful exchange we’ve had regarding your amniote analysis.

      1-7. I’m assuming you’re admitting these were miscodings based on “Some are due to codes that added my mistake and are now gone, thank you.”

      8. You say in SMNS 12352 you followed “the rostral profile to terminate a premaxilla”, and your reconstruction has a naris angled 32 degrees to the snout’s ventral margin. So if true, your own reconstruction would be contra your coding it having an angle of less than 30 degrees. But first, you extend the external naris posteriorly over soid surfaces of the nasal, when there is actually none of it preserved. Second, without knowing the shape of the premaxilla, you can’t even estimate this. It could have a basically straight dorsal snout margin and parallel-sided internarial bar like you illustrate, giving a low angle, or a convex margin like Herrerasaurus and/or internarial bar which is deeper at its base, which would give you a high angle. Third, even if your estimation were based on real data, being only 2 degrees from your cutoff point between states, surely it would be better to leave it coded unknown.

      Regarding Turfanosuchus’ skull table, I misinterpreted your character as meaning the frontoparietal only, not the supratemporal bar too. That makes two less of your codings wrong (it and Gracilisuchus for that same character).

      Regarding its occiput vs. quadrate, you have it coded as “Occiput close to quadrate”. I found this to be wrong as it had quadrate condyles far posterior to its occiput (see http://postimage.org/image/o8f8eahm3/ with Erythrosuchus on the right to show occiput position compared to parietals), as I wrote in my notes. But I did make a typo in the matrix, as you note.

      In Marasuchus, one of the nine preserved maxillary teeth (the fourth one back) is taller than twice its width. The others are shorter. You coded it as maxillary teeth taller, I coded it as maxillary teeth shorter. It should be coded as having both (0+1).

      The same more traditional tree results when these four codings are fixed.

      I’m not surprised that it takes few steps to get back to your topology, or that taxon deletion causes that, since I reported this the moment my changes to your matrix resulted in a more standard arrangement. So it’s only on the cusp of getting the standard one. We’ll see what happens once more codings are corrected. If you catch more errors by me, do report them.

      The unlabeled 113 characters are your mandibular and postcranial ones. The ones after that are 27 from Nesbitt, and 5 spun off of your cranial traits when you had single characters coding for more than one thing (e.g. teeth with constricted roots and teeth with large serrations).

  3. You wrote:
    Herrerasaurus having an anteriorly pinched interchoanal rod. You used Sereno and Novas’ (1993) dotted hypothetical vomer for this. How could you code it when they specifically say the anterior vomer isn’t visible?

    The interchoanal rod is the vomer. It may or may not be pinched, that’s not the character trait. The rostrum is pinched, forcing choanae that are lateral to the maxilla to a medial position. That was what I was shooting for. So it doesn’t matter that the vomer is visible or not in this case. I went by the reconstruction of Sereno and Novas.

    You wrote:
    Trialestes having an anteriorly pinched interchoanal rod. How can you code this when we know nothing of the palate of Trialestes?

    Again, I don’t describe the “rod.” The rostral shape in palatal view is what I am describing.

    You wrote about the naris angle of SMNS 12352.

    It nests with Marasuchus (no naris preserved) between Herrasaurus (> 30) and Tawa + Coelophysis (<30), so it doesn't matter in the end.

    You mentioned there were 33 rescorings for Turfanosuchus,
    yet my Turfanosuchus nests close to aetosaurs (and the similarly long-necked Yarasuchus and Ticinosuchus, which is where Ezcurra et al. (2010) and you nested it.
    So what does that mean, assuming your were correct? Or I was correct? Apparently the differences in your data and mine were of little importance ultimately.

    Again, you seem to be glossing over the errant topology of your tree, nesting theropods as derived from several clades of plant-eating grades, and separated from otherwise morphologically close relatives as Gracilisuchus and Turfanosuchus. I realize egos are involved here. I have often reported on data changes in the large reptile tree and will continue to do so. So that's not an issue. What I do object to is the lack of self-criticism when your results produce such "odd bedfellows." Unfortunately, that trait follows a long tradition of paleontologists who likewise see nothing wrong when their trees nest taxa together that obviously do not resemble one another and do not make sense in an evolutionary topology that should demonstrate gradual accumulations of traits in derived taxa, but often do not.

    • 1-4. I assume you accept these as miscodings on your part, as before.

      5. For Herrerasaurus, the character is “Internal nares position: (0) lateral; (1) medial vomers narrow; (2) medial rostral pinch; (3) medial posterodorsal to vomers; (4) medial, vomers wide.” I suppose we can chalk that up to an unclear state definition. Except then how do you explain e.g. Silesaurus, which has a much more pinched snout and thus pinched choanae (both singly and when viewed together), yet you coded it as state 1 “medial vomers narrow”. Ditto for Lesothosaurus. Mind expliticly clarifying the meaning of each state, so that I can code it properly?

      6. For Trialestes, how do you know the “rostral shape in palatal view” when we only have a lateral drawing of the skull?

      7. Another assumed accepted miscoding.

      8. SMNS 12352 “nests with Marasuchus (no naris preserved) between Herrasaurus (> 30) and Tawa + Coelophysis (<30), so it doesn't matter in the end." But it does matter because being coded with a low angle helps it group with Tawa and Coelophysis, while the high angle would help it go almost anywhere else in Dinosauria, or even with Saltoposuchus, which most experts think it is close to.

      "You mentioned there were 33 rescorings for Turfanosuchus,
      yet my Turfanosuchus nests close to aetosaurs (and the similarly long-necked Yarasuchus and Ticinosuchus, which is where Ezcurra et al. (2010) and you nested it.
      So what does that mean, assuming your were correct? Or I was correct? Apparently the differences in your data and mine were of little importance ultimately."

      Ezcurra et al. didn't even include Yarasuchus and Ticinisuchus, and found Turfanosuchus to be in a different position than you did, away from aetosaurs to boot. Their tree is-
      (Sclero+Dinoiformes)(Turfano(Qiano(Para(Stagono(Sauro+crocs)))))
      Yours is-
      (Para(Sauro((Qiano+Stagono)(Turfano(Dinoiformes(Sclero+crocs))))))
      In both (and Nesbitt's), it's close to the base of Dinosauriformes, so works as an outgroup for our test, but besides that the trees share almost nothing in common among included taxa.

      As for the "errant" topology of my tree, I must reiterate "my" tree here is really your matrix slightly modified and does not fully represent my views.
      In Nesbitt's (2011) tree, which I'll accept as mine until I see a better analysis-
      The base of Archosauria is carnivorous, Revueltosaurus+aetosaurs evolve herbivory, shuvosaurs evolve herbivory, derived silesaurs evolve herbivory, ornithischians do, and sauropodomorphs do. So that's five steps.
      In your tree, Revueltosaurus does, aetosaurs do, phytodinosaurs do, Poposaurus reverses to be carnivorous and Daemonosaurus reverses to be carnivorous. That's five steps too. The trees are equally parsimonious as far as diet goes.

      But even ignoring that, you think of similarity largely in terms of gross ecotype, but I could easily speak of your tree having "odd bedfellows". Why are sauropodomorphs with ornithischians when they share hyposphenes, long manual phalanx I-1, a C-shaped primordial sacral rib, a large lateral condyle of metacarpal I, etc. with theropods? What odd bedfellows sauropodomorphs and ornithischians are in your tree. Where's your self-criticism about that? The answer of course is that every analysis will have some characters that don't fit well, causing odd bedfellows if we concentrate on those. That's what homoplasy is. Nesbitt's analysis has characters related to herbivory, they're just outnumbered by others. In the end, I would have to say most biologists don't think outwardly dissimilar taxa are obviously unrelated and make no evolutionary sense. We're all fine with pangolins being related to carnivores for instance, while you have them inside Xenarthra by anteaters. How likely is it all the other evolutionary biologists are wrong about extensive convergence and reversal in gross ecotype being the norm, while your lone intuitions are correct?

  4. Thank you for your comments, Mickey. I don’t plan my trees out. They just grow the way they grow. I just have a larger field in which taxa have a wider gamut of opportunities to nest in. To your points about “odd bedfellows, imagine a world in which it was widely accepted that pterosaurs evolved from small, bipedal lizards with long lateral toes and various extradermal membranes. Now, along comes that heretic David Peters saying, “Not so fast! My analysis indicates that flat-headed, tiny handed, no lateral toe, osteoderm-layered Scleromochlus, the basal croc is the closest ancestor. And if not that, Parasuchians make great ancestors. I say so because I have tested them and this is how they nest. ” Now you’ve got good reason to gripe.

    Yes, herbivory evolves lots of times. But it’s time for you to admit the suit of traits that basal dinos like Herrerasaurus, share with Gracilisuchus, Trialestes and Turfanosuchus. Who says poposaurs and daemonosaurus were carnivores?

    • No one is saying Scleromochlus or parasuchians are ancestors of pterosaurs. It’s like Feduccia always claiming we think dromaeosaurids are birds’ ancestors. Sister group does not equal ancestor, and both dinosauromorphs and pseudosuchians are closer to pterosaurs than parasuchians are in Nesbitt’s tree anyway. Also, Scleromochlus may have had phalanges on mtV (the two known examples are missing many pedal elements and don’t taper to points) and the ‘osteoderms’ may just be scales (they’re preserved as sandstone impressions, so we don’t know if they contained bone). Now I think you’re probably correct pterosaurs are closer to Longisquama and such, but it’s still good to represent the other side fairly.

      As for carnivory, Poposaurus (Parker and Nesbitt, 2013) and Daemonosaurus both have recurved, transversely compressed, sharp lateral teeth with small serrations, like Megalosaurus, rauisuchuians, etc.. The anterior teeth of Daemonosaurus are rounded in section, but this is true of e.g. Coelophysis, Duriavenator, Compsognathus as well, so is common in carnivores. Why would they not be carnivores besides the circular logic that you place them in a clade of herbivores?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.