Friday, August 24, 2012

Turner et al. 2012- great review, sloppy analysis

Back in October 2011, Turner et al. redescribed Mahakala and included a phylogenetic analysis which basically combined the TWG matrix with Clarke's bird matrix.  This analysis of 102 taxa and 472 characters was not backed up with a matrix or character list in the paper.  However, Turner et al. did state "The character list, data matrix, and supplemental analysis data are available at http://www.anat.stonybrook.edu/people/facultypage/turner and at http://research.amnh.org/paleontology/staff/mark-norell."  Following links from the first page leads to http://www.anat.stonybrook.edu/aturner/lab/Datasets.html , which didn't and still doesn't have the dataset.  The second page doesn't seem to have links to any datasets.  Emailing Turner resulted in no reply.  I'm very grateful for Norell's help and hospitality during my AMNH visits, but it's concerning to have not only an unavailable dataset, but also false statements in a paper about a dataset's availibility.

Regardless, Turner et al. (2011) stated their dataset was "developed by Turner et al. (in press).", a paper on dromaeosaurid systematics and paravian phylogeny, and theropod workers have been waiting ever since.  Eleven months later, the paper has been published, and this post will contain my thoughts on it.  Having a matrix of 111 taxa and 474 characters, it's actually different than the Mahakala analysis, meaning the latter has STILL not been published.

First of all, it's useful to know the paper is an updated version of parts of Turner's 2008 thesis.  Besides this, the thesis contains the Mahakala redescription noted above, a redescription of Jinfengopteryx's skull, and a description of "Ichabodcraniosaurus" IGM 100/980 (though Turner doesn't name it).  Unfortunately, the thesis has several major issues and these were not altered for the published version.

But before my usual scathing critique, let's look at the good.  The paper provides updated diagnoses for many dromaeosaurids (notable exceptions are Pamparaptor, Itemirus, Luanchuanraptor, Ornithodesmus and Dromaeosauroides among taxa generally thought to belong to the family).  Even better, there are numerous useful photos, including Adasaurus(!!!!) and new material of Utahraptor.  Though why they felt the need to waste 4 of 7 Utahraptor photos on the previously well described and illustrated premaxilla is a mystery.  Now that Adasaurus holotype photos have finally been published, I'm free to distribute mine, so those will be strewn throughout this review.  One juicy fact is "The only description of [Achillobator] is nefarious in that it was published without the knowledge of any of the junior authors based on a preliminary draft left in Mongolia in 1997."  Like Senter (2011), Turner et al. synonymize Tsaagan and Linheraptor, though they do this properly by evaluating the latter's supposed diagnosis.  Overall, this section is excellent.

Skull of Adasaurus mongoliensis IGM 100/20 in posterolateral view.

The phylogenetic analysis is where things start to fall apart.  The good news is it's better than most TWG analyses in ordering most of the characters that need it and having explicit state definitions for the 0 state.  Also, he's improved on the Turner et al. (2007) analysis that added Daspletosaurus in that he now codes its postcranium.  Another good thing is that the crown birds seem genuinely coded and not just assumed to be the same in regard to states usually diagnosing deeper branches.  In general, combining Clarke's bird matrix into the TWG matrix is a great way to quickly and easily give a better look at Mesozoic paravian phylogeny.  But here, it seems to have been done hastily and inconsistently. 

Take the taxon sampling for instance.  Numerous taxa are excluded supposedly because they could not be examined firsthand (Borogovia, Urbacodon, Tochisaurus, Geminiraptor, Sinusonasus, Gansus, Ambiortus, Archaeorhynchus).  Yet other taxa are admitted to be coded based only on literature (Xiaotingia, Xixiasaurus, Liaoningornis, Pengornis, Cathayornis, Concornis, Songlingornis, Baptornis, Hesperornis, Iaceornis, Ichthyornis; 35 included taxa if you trust Appendix 1).  Yet surely a basalmost "euornithine" like Archaeorhynchus is more important to include than a fragmentary possible crown bird like Iaceornis.  Similarly, if the goal is to test paravian relationships, why add taxa like Albertosaurus, Proceratosaurus and Albinykus, but not Yandangornis, Zhongjianornis, Protarchaeopteryx or Falcarius?  The latter is explained in Turner's thesis, where he says since Zanno is examining therizinosaurs, he didn't want to scoop her, but Zanno's papers are all out now and Falcarius was already coded for TWG matrices like Li et al. (2009). 

I mentioned Appendix 1 earlier, which is a list of which references and specimen numbers the authors used and examined firsthand.  But it's clearly wrong.  Daspletosaurus is coded completely, but the two provided references don't allow coding for most of it.  Harpymimus and Garudimimus both have Kobayashi and Barsbold (2005a, b) listed, but those were not used for most of the characters since the TWG codings were taken directly from Norell et al. (2001).  Many taxa have listings which are contradicted in the main text.  I note "Neuquenraptor sp." (near certainly Pamparaptor) is listed here too and was included in the thesis analysis, but not the published analysis.  Oops.  More distressing is that Citipati osmolskae is said to be based on IGM 100/978 and 100/979, but actually the TWG portion of its codings are directly from the IGM 100/42 OTU of Norell et al. (2001).  So that OTU is actually a chimaera.

Cervical vertebrae of Adasaurus mongoliensis IGM 100/20 in right lateral view.

But what about the coding?  First, it's obvious the TWG part of the matrix was just copied from all of the earlier iterations going back to Norell et al. (2001) for most taxa.  So Harpymimus is still woefully incompletely scored despite being well described for seven years now, and Tsaagan is still only coded for the holotype despite Turner synonymizing it with the very complete Linheraptor in this very paper.  Sadly, this paper is the first real TWG example of my favorite pet peeve.  See if you can tell what I mean based on Turner's codings for Tarbosaurus-

???????????????????0???????????????????1???????????????????????
???????????????????????????????????????????????????????????????
?????????????????0?????????????????????????????????????????????
?????????????2????????????????????????????????0???????????????1
1111011112??2011000000000??00??????0000???000?????????????00???
???????????????????????????????????????????????????????????????
????????????????????????????????????????????????????????????000
000??02000??0001?0??00100000???????0

Now you don't even have to know what the characters are to see the problem.  Tarbosaurus is known from well described basically complete specimens, so nearly all of those question marks shouldn't be there.  Baptornis is uncoded for almost every non-Clarke-based character.  Patagopteryx is uncoded for many as well.  For the Clarke-based characters, Troodon, Jinfengopteryx, most of Saurornitholestes, the front half of his jinfengopterygine troodontids and the back half of Alxasaurus and Graciliraptor are uncoded.

Turner didn't recover Songlingornithidae, which seems odd since Clarke did.  Wanna know why?  He didn't notice Clarke et al. (2006) accidentally switched a few pectoral characters in their matrix, then partially recoded them himself.  It's just sloppy.

Dorsal vertebrae, sacrum, pelvis and femur of Adasaurus mongoliensis IGM 100/20.

When it comes to understanding Clarke's bird characters and applying them to non-birds, Turner makes some mistakes.  He codes all long-tailed theropods polymorphic for caudal transverse process length (some avians have very short processes on their free caudals, but instead of just comparing long-tailed theropod proximal caudals, I guess Turner coded all their free caudals, which of course start out with long processes that decrease to nothing distally).  Turner "rewords" the alvarezsaurid TWG character of a transversely convex distal ulna (visible in front/back view) with the ornithothoracine Clarke character of a posteriorly convex distal ulna (visible in side view), so that alvarezsaurids and ornithothoracines are both coded as having the same character.  Another example is how he codes the obturator tuber of birds as an obturator process, when birds lack an obturator process (see Hutchinson, 2001 for details).  But then he codes birds as lacking the obturator tuber, leaving it only known in some dromaeosaurids, when actually those dromaeosaurids share the obturator tuber with birds.

Turner's conclusions also seem inappropriate.  Try this one on for size- "On the other hand, only 1 additional step is required to place Troodontidae as the sister taxon to Avialae (fig. 73).  These only slightly less parsimonious topologies could be interpreted as an indication of weakness in deinonychosaurian monophyly. We are inclined to interpret (and think it is more readily borne out by the data) that this is instead a reflection of the overall morphological similarity of the basal members of each paravian clade (e.g., compare Mahakala to IGM 100/1126 or Archaeopteryx)."  But but... the refutation is too obvious to even state.  Interestingly, the quote is straight from his thesis (replacing I with we), so apparently adding Xiaotingia, Anchiornis and other taxa didn't change this number.  As those two genera are the basalmost troodontids in his topology and are very bird-like, it would be quite the coincidence for them to not affect this test, so I wonder if Turner truly did re-run the constraint analysis after the changes were made to his thesis' analysis.

Lower leg and pes of Adasaurus mongoliensis IGM 100/20.  Note the small pedal ungual II is incorrect (Kubota, pers. comm. to Senter, 2012).

Or "Constraining Rahonavis ostromi as a basal avialan requires seven additional steps beyond the most parsimonious topology, and constraining Rahonavis to a more derived placement within Avialae requires 11 additional steps (fig. 79). Taken together with the strong morphological support for the Unenlagiinae clade and strongly unparsimonious nature of an ‘‘avialan’’ Rahonavis, it has emerged that there is no reason to
consider Rahonavis as a problematic taxon."  So six steps is supposedly unambiguous enough to be non-controversial, but he finds Dilong to be closer to birds than tyrannosaurids and it takes twelve more steps to constrain as a tyrannosauroid (as stated in the thesis).  By that logic non-tyrannosauroid Dilong must be incredibly secure.  Ugh.

So my basic message is, the review section is excellent with tons of useful new information, but the phylogenetic analysis is more flawed than most TWG efforts.

References- Turner, 2008. Phylogenetic relationships of paravian Theropods. PhD Thesis. Columbia University. 666 pp.

Turner, Pol and Norell, 2011. Anatomy of Mahakala omnogovae (Theropoda: Dromaeosauridae), Togrogiin Shiree, Mongolia. American Museum Novitates. 3722, 1-66.

Turner, Makovicky and Norell, 2012. A review of dromaeosaurid systematics and paravian phylogeny. Bulletin of the American Museum of Natural History. 371, 1-206.

2 comments:

  1. Have you actually run the matrix? I tried to build it from the appendix, but it appears two taxa (Tianyuraptor and Microvenator) are missing 1 character each. The matrix is supposed to be on Morphobank, but the link (http://morphobank.org/permalink/?660) just leads to a page reading: "This project is not yet publicly available".

    Sigh.

    ReplyDelete
    Replies
    1. Hahaha. That Morphobank thing is priceless considering the 2011 matrix was never put online at the cited websites either. But no, I haven't tried running this matrix yet.

      Delete