Back in 2010, I wrote
a post about how surprisingly theropod-like Heterodontosaurus was, and ended by stating "Not that I'm saying Saurischia is wrong, I just think we can get so
comfortable with an idea that we stop looking at alternatives." This week I was surprised to learn of a new paper by Baron, Norman and Barrett (2017) which uses a phylogenetic analysis to recover that heterodox ornithichian+theropod topology. In particular, they recover herrerasaurids as basal sauropodomorphs, while ornithischians and theropods are joined together in the resurrected clade Ornithoscelida*. Vindication?
* Baron et al. redefine some clades (see below), but here I'm using the traditional concepts since I dislike some of their definitions.
 |
"Reduced strict consensus tree of [Baron et al.'s] main
analysis ..., after the exclusion of Saltopus elginensis, Agnosphitys
cromhallensis, Eucoelophysis baldwini and Diodorus scytobrachion." (after Baron et al., 2017) |
Well, let's see how good the study is first. I should preface this by noting that while traditional Saurischia has been recovered in every single previously published numerical cladistic analysis, most of these were crap by today's standards. The only exceptions are really Nesbitt et al.'s (2009)
Tawa analysis, which used 30 dinosauromorphs and 315 characters, Ezcurra's (2006)
Eucoelophysis analysis that used 26 dinosauromorphs and 287 characters, and Cabreira et al's
Ixalerpeton/
Buriolestes analysis that used 42 dinosauromorphs and 256 characters Yates' is large but focuses mainly on sauropodomorphs, and others are small (11 taxa, 98 characters for Langer and Benton, 2006; 18 taxa, 139 characters for Martinez et al., 2011; etc.). Baron et al. include 72 dinosauromorphs and 456 characters, so by that measure is much better.
It also uses only species-level OTUs and properly orders characters, so that's good. And it has excellent taxon sampling, including basically every basal dinosauromorph except
Ixalerpeton,
Buriolestes and
Isaberrysaura (which were presumably published too late) and a couple fragmentary taxa (
Teyuwasu,
Alwalkeria)
[Edit: I did not realize at the time it doesn't have Daemonosaurus, Scutellosaurus or rhodesiensis, and less importantly the fragmentary Chromogisaurus. The absence of Daemonosaurus is particularly odd given it's somewhat heterodontosaury features...]. In particular, ornithischians are well represented unlike any prior analysis (14 taxa vs. 5 in Nesbitt et al., 3 in Ezcurra, and 5 in Cabreira et al.). I'd say sauropodomorphs are almost over-represented, except that with this heterodox topology, it might have been a good thing to include even the basal gravisaurs they did just in case those ended up being the earliest diverging branches. Sauropodomorpha has the standard topology in their trees btw, except
Eoraptor is recovered as a theropod.
The next thing to check is if this result is robust, and at a whopping 20 extra steps needed to recover Saurischia, is it ever*! For comparison, the original TWG matrix needs 22 more steps to recover Bullatosauria and 23 steps to get Russell and Dong's (1994) crazy topology. They tested this in an IMO un-ideal way, by constraining all 42 saurischians to be in an exclusive clade. Would have been better to just use a backbone of e.g. (
Euparkeria,
Lesothosaurus (
Plateosaurus,
Coelophysis), in case e.g. herrerasaurids aren't dinosaurs. Baron et al. also deleted various sets of taxa but found "the recovery of Ornithoscelida as a monophyletic clade, as well as Sauropodomorpha+Herrerasauridae (Saurischia, new definition), occurred in all of these analyses, regardless of which taxa were included or excluded." So that's impressive. They also tested
Dimorphodon as an alternate outgroup (the defaults are
Euparkeria and
Postosuchus which is wrongly called an ornithosuchid; an ornithosuchid would have been better ironically, since they're the first branching pseudosuchian group if parasuchians aren't), and found the same topology.
* To those who noted the 66% bootstrap for Ornithoscelida (Cerny on the DML, *cough cough*), realize that with very incomplete taxa like most fossils, bootstrap percentages mean very little. It's easy for a fragmentary taxon like
Nyasasaurus to fall inside Ornithoscelida, for example. And every time it does, that lowers the bootstrap score.
 |
| The Lori analysis if limited to the taxonomic scope of Barron et al. (2017). The topology changed from my last post after the inclusion of non-dinosauromorph outgroups, but is still unlike either the traditional topology or Baron et al.'s. Instead, it features non-dinosaurian herrerasaurids and a Phytodinosauria with paraphyletic sauropodomorphs. |
Of course, I just posted about recovering a heterodox dinosaur topology using the Lori matrix, which has more characters than even Baron et al.'s (481 are parsimony informative for basal dinosauromorphs) and 71 dinosauromorphs within Baron et al.'s scope (more coelophysoids, less plateosaurs). Yet I don't trust that part of the Lori tree. That's because the Lori matrix wasn't designed to sort basal dinosauromorphs, so lacks many of the proposed characters that support traditional clades. Is Baron et al.'s similar? To determine that, we need to check and see if the characters traditionally used to support Saurischia are included. Of the 28 characters recovered as supporting the clade in Nesbitt et al.'s study, 27 are included by Baron et al., and one is even (correctly) divided into two characters since it was a composite. The one excluded character ("Femur, surface between the lateral condyle and crista tibiofibularis on the distal surface: smooth (0); deep groove (1)") doesn't actually support Saurischia anyway. Cabreira et al.'s analysis was published after Baron et al. had done their analysis, so the latter can't be faulted for not including all of their characters. Ironically, they also found 28 characters, but 16 of these are different than Nesbitt et al.'s. Five of these sixteen are not in Baron et al.'s matrix, which considering the 20 steps needed to get Saurischia there, aren't a help even assuming coding accuracy. Finally, Ezcurra had 39 characters of which 23 aren't used in the prior two, and 9 of these aren't in Baron et al.'s. Even adding these nine to the preceding five, it doesn't match the twenty. So I'd say Baron et al. did a decent job testing saurischian monophyly character-wise, and that adding published suggested saurischian characters aren't going to overturn Ornithoscelida. For comparison, only 5 of Nesbitt et al.'s characters are in the Lori matrix and only 5 of Cabreira et al.'s new ones and 5 of Ezcurra's new ones, though the Lori topology eliminates Saurischia by recovering Phytodinosauria instead.
So far it's looking very good for Ornithoscelida. However- here are some caveats. First, Baron et al.'s matrix has some correlated characters. Examples are "Third cervical vertebra, centrum length: 0, subequal to the axis centrum; 1, longer than the axis centrum (Gauthier, 1986; Langer and Benton, 2006; Nesbitt, 2011)" compared to "Elongation of cervical centrum (cervicals 3–5): 0, less than 3.0 times the centrum height, 1, 3.0-4.0 times the centrum height, 2, >4.0 times the centrum height (Upchurch, 1998; Pol et al., 2011b). ORDERED"; "Digit I with metacarpal: 0, longer than the ungual; 1, subequal or shorter than the ungual (Sereno, 1999; Langer and Benton, 2006; Nesbitt, 2011)" compared to "Manual digit I, first phalanx: 0, is not the longest non-ungual phalanx of the manus; 1, is the longest non-ungual phalanx of the manus (Gauthier, 1986; Langer and Benton, 2006; Nesbitt, 2011)", and "Deltopectoral crest: 0, less than 30% the length of the humerus; 1, more than 30% the length of the humerus (Bakker and Galton, 1974; Benton, 1990; Juul, 1994; Novas, 1996; Benton, 1999)" compared to "Humerus, apex of deltopectoral crest situated at a point corresponding to: 0, less than 30% down the length of the humerus; 1, more than or equal to 30% down the length of the humerus but less than 50% down the length of the humerus; 2, more than 50% down the length of the humerus (modified from Bakker and Galton, 1974; Benton, 1990; Juul, 1994; Novas, 1996; Benton, 1999, Nesbitt, 2011)." There are also no polymorphies in the matrix, which is impossible in reality with taxa like Coelophysis and Plateosaurus known from tons of variable individuals. So it's not an excellent character list.
Second, despite the great taxon sample,
Buriolestes would seem to be a most relevant taxon to include, being a supposed basalmost sauropodomorph that is very theropod-like. Maybe it'd fall out as a theropod in Baron et al.'s trees, since
Eoraptor does, but who knows until it's added. Another issue is that no neotheropods are included, unlike Nesbitt et al.'s and Ezcurra's matrices. If coelophysoid-grade taxa are paraphyletic, that's fine. But if
Dilophosaurus,
Cryolophosaurus,
Dracovenator and (?)
Sinosaurus are coelophysoids, we're missing a whole side of the avepod tree. And that could be important since heterodontosaurids seem particularly similar to coelophysoids among avepods. So add
Ceratosaurus,
Elaphrosaurus,
Piatnitzkysaurus and
Monolophosaurus?
Also, how accurate is the coding? There's no premade TNT file and the matrix is presented in a way that without copy-pasting it's basically impossible to check codings. But the last character is an obvious one to code- "Singular and unbranched filamentous integumentary structures covering, or partially covering, the outer body: 0, absent; 1, present. NEW". Depressingly, the following taxa are coded as lacking filamentous integument-
Saltopus, Silesaurus, Lewisuchus, Pisanosaurus, Fruitadens, Abrictosaurus, Heterodontosaurus, Eocursor, Lesothosaurus, Scelidosaurus, Emausaurus, Agilisaurus, Hexinlusaurus, Jeholosaurus, Dracoraptor, Massospondylus carinatus, Guaibasaurus, Unaysaurus, Plateosaurus, Efraasia, Pantydraco, Herrerasaurus, Eodromaeus, Coelophysis, Eoraptor, Procompsognathus, Euparkeria, Dilophosaurus, Sinosaurus, Pampadromeus and
Dimorphodon. Obviously none of these taxa preserve integument (does
Dimorphodon preserve pterofuzz?), excluding osteoderms in a few. So this is a massive and worthless miscoding (only
Tianyulong preserves filaments among the OTUs and none is known to lack them). I can only hope this misrepresents the coding accuracy of the matrix.
So where does this leave us? Honestly, I'm rather strongly in favor of Ornithoscelida. For all my caveats, I could say the same for at least Nesbitt et al.'s matrix. I haven't examined Ezcurra's or Cabreira et al.'s in depth. Also, Baron et al. did include a lot of new ornithischian info, both taxon- and coding-wise. Given the apparent robusticity of the result, the size of the matrix, and the coverage of proposed saurischian characters, any analysis that aims to disprove Ornithoscelida needs to use this study as its base.
 |
| Holotype and referred specimen of Nyasasaurus, Middle Triassic massopod? (after Nesbitt et al., 2013) |
One other unexpected result of this analysis is the position of
Nyasasaurus. In its unofficial pre-cladistic days, considered a prestosuchid or thecodontosaurid (heh, what are those? ;) ), Nesbitt et al. (2013) recovered it as closer to Dinosauria than silesaurids, but outside examined
ornithischians, sauropodomorphs and
Tawa+avepods. Well, Baron et al. recovered it as... a massopod sauropodomorph outside Anchisauria. In the Middle Triassic. Which would be weird. Oddly enough,
Nyasasaurus emerges sister to
Plateosaurus in the Lori result. Hmm...
FINALLY (long post), Baron et al. redefine several clades given their new topology. This is where I have some serious issues.
First, Dinosauria becomes "The least inclusive clade that includes
Passer domesticus,
Triceratops horridus and
Diplodocus carnegii." I'm generally of the Kischlat school of phylogenetic definitions, where only taxa originally considered part of the clade (or not) are used. But sauropodomorphs weren't included when Dinosauria was coined, so I'll make an exception here. Since Sauropodomorpha is based on Sauropoda, and Sauropoda was based on
Atlantosaurus,
Apatosaurus,
Morosaurus and
Diplodocus, I'm fine with
Diplodocus being used (better than the standard
Saltasaurus). But of course I'm a defender of
D. longus remaining the type species (and submitted a Comment to the BZN over it), so would use it instead of
D. carnegii, which was unknown to Marsh. Similarly, I don't like making birds dinosaurs by fiat, so would rather use
Allosaurus fragilis for the theropod specifier since it was what Marsh based Theropoda on in 1881. For ornithischians, I follow Kischlat's arguments that
Stegosaurus armatus is ideal since
Triceratops was unknown to Seeley in 1887. So I'd suggest-
The least inclusive clade that includes
Allosaurus fragilis,
Stegosaurus armatus and
Diplodocus longus.
Ornithoscelida is defined as "The least inclusive clade that includes
Passer domesticus and
Triceratops horridus." It's not a great name for the clade, because Huxley included sauropodomorphs in the group-
Thecodontosaurus as a scelidosaurid and
Cetiosaurus as an iguanodontid. Huxley's point was to use it to relate all dinosaurs with
Compsognathus, which was not considered dinosaurian at the time. But since there are honestly no preexisting names for this concept, and this is a landmark paper, we might as well accept it. Just as in the above example though, birds were never considered ornithoscelidans, and ceratopsians were still unknown back in 1870. As Huxley's dinosaurian families were Megalosauridae, Scelidosauridae and Iguanodontidae, I'd suggest the following definition-
The least inclusive clade that includes
Megalosaurus bucklandii and
Iguanodon bernissartensis.
Saurischia is defined as "The most inclusive clade that contains
Diplodocus carnegii but not
Triceratops horridus", which is easily the worst definition of the bunch. A clade of just herrerasaurids and 1990's+ standard sauropodomorphs is NOT Saurischia. Seeley clearly constructed Saurischia to combine theropods and sauropods based on e.g. pelvic structure and axial pneumaticity. All proposed definitions either self destruct in this topology or ironically include only theropods. So I say if Ornithoscelida is a thing, then Saurischia just goes away.
Theropoda is defined as "The most inclusive clade that contains
Passer domesticus but not
Diplodocus carnegii or
Triceratops horridus." Again, birds shouldn't be theropods by definition, and neither Mesozoic taxon was known when Marsh named Theropoda. Using the logic from the above examples, I'd suggest- The most inclusive clade that contains
Allosaurus fragilis but not
Diplodocus longus or
Stegosaurus armatus.
Ornithischia is defined as "The most inclusive clade that contains
Triceratops horridus but not
Passer domesticus or
Diplodocus carnegii." So again, I would suggest- The most inclusive clade that contains
Stegosaurus armatus but not
Allosaurus fragilis or
Diplodocus longus.
Sauropodomorpha is defined as "The most inclusive clade that contains
Diplodocus carnegii but not
Triceratops horridus, Passer domesticus or
Herrerasaurus ischigualastensis." This is the second most troubling definition because there's no reason to bring herrerasaurids in to it. Indeed,
Herrerasaurus was regarded AS a sauropodomorph by basically every author after its initial description and before the mid 80's cladistic revolution.
I even state on the Database "While these historical comparisons of
Herrerasaurus
to sauropodomorphs were generally vague, unpublished analysis suggests herrerasaurids
can be moved to a position as basalmost sauropodomorphs without adding many
extra steps" based on my experiments with a partially corrected Nesbitt et al. dataset. Just regard herrerasaurids as basal sauropodomorphs in this topology and use the definition- The most inclusive clade that contains
Diplodocus longus but not
Stegosaurus armatus or
Allosaurus fragilis.
The take away- I provisionally accept Ornithoscelida as better supported than Saurischia, but just get rid of the latter in this topology and let herrerasaurids be sauropodomorphs.
As a last question- what does this say about
Chilesaurus? It wasn't included, but is the prime example of an ornithischian-theropod cross after all....
References- Russell and Dong, 1994. The affinities of a new theropod
from the Alxa Desert, Inner Mongolia, People’s Republic of China. Canadian
Journal of Earth Sciences. 30(10), 2107-2127.
Ezcurra, 2006. A review of the systematic position of the dinosauriform archosaur
Eucoelophysis baldwini Sullivan & Lucas, 1999 from the Upper Triassic of New Mexico, USA. Geodiversitas. 28(4), 649-684.
Langer and Benton, 2006. Early dinosaurs: A phylogenetic study. Journal of Systematic
Palaeontology. 4(4), 309-358.
Nesbitt, Smith, Irmis, Turner, Downs and Norell, 2009. A complete skeleton of
a Late Triassic saurischian and the early evolution of dinosaurs. Science. 326,
1530-1533.
Martinez, Sereno, Alcober, Columbi, Renne, Montanez and Currie, 2011. A basal
dinosaur from the dawn of the dinosaur era in Southwestern Pangaea. Science.
331, 206-210.
Nesbitt, Barrett, Werning, Sidor and Charig, 2013. The oldest dinosaur? A Middle
Triassic dinosauriform from Tanzania. Biology Letters. 9(1), 20120949.
Cabreira, Kellner, Dias-da-Silva, da Silva, Bronzati, Marsola, Muller,
Bittencourt, Batista, Raugust, Carrilho, Brodt and Langer, 2016. A
unique Late Triassic dinosauromorph assemblage reveals dinosaur
ancestral anatomy and diet. Current Biology. 26(22), 3090-3095.
Baron, Norman and Barrett, 2017. A new hypothesis of dinosaur relationships and early dinosaur evolution. Nature. 543, 5601-506.
