Saturday, January 24, 2015

BANDit cladogram evaluated - James and Pourtless 2009

Besides Hou et al. (1996), the only BANDit (Birds Are Not Dinosaurs) cladistic analysis has been that of James and Pourtless (2009).  Little has been said about this paper, with most of the commentary noting the backwards philosophy the authors had in coding characters with supposedly controversial homology as unknown instead of letting phylogenetic analysis establish homology.  Well I finally looked at it in a little detail, and found a far larger issue.

First, let's go over the basic tests and conclusions of James and Pourtless' study.  After making the almost correct observation that no published cladistic analysis has tested whether birds are dinosaurs (Senter 2004 did, but while the authors cite it they don't comment on it in that regard), James and Pourtless aim to correct this by constructing their own dataset with alternative proposed bird ancestors included.  With basal archosauriforms as the outgroup, the authors also include crocodylomorphs,  Longisquama, a few other pseudosuchians and numerous theropods.  Running their dataset of 242 characters and 79 taxa, they recover a large number of MPTs.  These are only ever illustrated as majority rule bootstrap trees after a posteriori pruning of taxa, and each alternative tree is basically a well supported Archosauria with a massive polytomy between crocodylomorphs, numerous theropod lineages and birds.  Longisquama weakly (55% bootstrap) groups with birds if 21 "controversially homologous" characters (which are only coded for non-dinosaurs) are excluded and non-bird maniraptorans are pruned away.  Statistical tests were done, seemingly showing the 'basal archosaur' (= Longisquama), crocodylomorph and dinosaur hypotheses of bird origins aren't any worse than each other.

Figure 12 of James and Pourtless (2009), showing majority rule bootstrap tree with numerous taxa including all maniraptorans pruned away a posteriori.  *Gasp* It looks like birds are closer to non-dinosaur Longisquama and that even crocs could be as close to birds as dinosaurs.  Were BANDits right all along? (Spoiler- NO)
How to explain this?  First, I should note that when you run their data, the resulting MPTs are pretty standard.  As shown in my typed cladogram below, while there are plenty of weird details going on in Theropoda, the basic structure of birds within Pennaraptora within Maniraptora within Coelurosauria within Theropoda within Saurischia within Archosauria is present.  The only odd relevant aspects are Effigia is an ornithomimosaur and Longisquama is deeply embedded in Pennaraptora.  Note there's no way of knowing Effigia groups with ornithomimosaurs by reading the paper, as its exact results aren't mentioned and it's always pruned out of the figured trees.  I find this suspicious.  In any case, Longisquama being sister to ornithines is discussed since this is the current favored alternative of BANDits.  Besides the paltry literature, the authors depended on photos at the KU for their coding of the genus, and present a cranial reconstruction.  This reconstruction looks as if it was done assuming a theropodan morphology for the taxon.  For example, there's a T-shaped structure at the anterodorsal corner of the orbit, interpreted as a lacrimal by the authors.  But it would also compare well to a prefrontal in a basal diapsid like Coelurosauravus, where the lacrimal is reduced.  Similarly, the strip under the supposed supraorbital ridge (yellow) would match the posterodorsal strip of the lacrimal/prefrontal in Coelurosauavus.  Or the wedge just behind this, while unidentified by the authors, would match the postfrontal of Coelurosauravus.  The supposed antorbital fossa (red) matches the external naris of Coelurosauravus well, and even the big posterodorsal hole ("?" in their figure) would match Coelurosauravus' if it's a supratemporal fenestra as identified by most previous authors.  This isn't to say that Longisquama is a close relative of Coelurosauravus (as Senter 2004 recovered), merely that the lens you view a fossil like Longisquama through based on photos can affect your identifications.  The controversial aspects of Longisquama's anatomy are similarly all coded as theropod-like- thecodont teeth, antorbital fenestra with maxillary fenestra, furcula, etc..

Longisquama skull (counterslab of PIN 2584/4) and interpretation of James and Pourtless (2009) on top.  Coelurosauravus skull reconstruction at bottom after Evans and Haubold (1987), modified so that the entire spiked temporal element is the squamosal after Schaumberg et al. (2007).  I see the supposed antorbital fossa (afo in red), but don't think its obviously a fossa, as opposed to a fenestra (e.g. the naris) or bone process itself.  Similarly, the entire ventral edge of the supposed antorbital fenestra (aof) looks like a crack to me, including the supposed dorsal process of the maxilla (dpm), supposedly there to support an interfenestral bar.  I don't see any posterior edge of the antorbital fenestra either, so that that whole area could be a solid maxilla.  The ventral process (vp) is apparent, but if anything seems separated from the nasal by a suture (shown by the green space).
As for the matrix, there are of course lots of miscodings (e.g. character 5 is for dermal armor, which is coded unknown for almost all theropods and Effigia, plus coded absent in Ceratosaurus, the one theropod that actually has it).  Some of the characters are formed terribly while others are partitioned in a way that weights them (e.g. each dentigerous element has its own character for tooth serrations, each a three state ordered character of the form '0- all serrated; 1- some serrated; 2- none serrated'; thus any taxon leading from serrated to unserrated teeth needs six steps to do so).  The 21 "controversially homologous" characters include every manual character due to the I-II-III vs. II-III-IV issue, though oddly these aren't even coded for dinosaurs with five digits like Herrerasaurus.  There's also a huge imbalance of taxa, with a whopping 45 non-avian theropods plus 16 birds, only 9 pseudosuchians, 3 non-archosaurs, Eoraptor, Marasuchus and Scleromochlus.  These are all issues, but not the most important one.

|--Proterosuchus
`--+--Euparkeria
...`--+--Erythrosuchus
......`--+--Pseudosuchia(Ornithosuchus,Postosuchus,Crocodylotarsi)
.........|--Scleromochlus
.........|--Marasuchus
.........`--+--Eoraptor
............`--+--Herrerasaurus
...............|--Carnosauria("Syntarsus",Dilopho,Cerato,Sinraptor,Allo,Tyranno)
...............`--+--Juravenator
..................|--Sinosauropteryx
..................|--Compsognathus+Huaxiagnathus
..................|--Guanlong+Dilong
..................|--Effigia+Ornithomimosauria
..................|--Microvenator
..................`--+--Coelurus
.....................|--Ornitholestes
.....................|--Falcarius
.....................`--+--Therizinosauroidea
........................`--+--Oviraptorosauria+Dromaeosauridae
...........................`--+--Troodontidae
..............................`--+--Pelecanimimus+Avimimus+Parvicursorinae
.................................`--+--Caudipteryx
....................................`--+--Longisquama
.......................................`--Ornithes


Actual strict consensus of James and Pourtless' data, simplified so that genera which form monophyletic clades are represented by their clade names.  Note no dinosaurs were coded for manual characters, Effigia is an ornithomimosaur, and Longisquama is deeply nested in Theropoda.

No, the biggest problem with James and Pourtless' analysis is that its matrix consists mostly of characters designed to diagnose coelurosaur clades (e.g. 96 from Clark et al.'s 2002 TWG analysis; 26 from Chiappe's 2002 bird analysis) plus those suggested by BANDits to group Longisquama and crocs with birds.  So what happens when you analyze Longisquama in a coelurosaur/bird matrix after reconstructing its skull with theropod presumptions and pretending no dinosaur preserves hands?  Given it doesn't preserve sacrum, pelvis, hindlimbs or tail, and that the vertebrae are basically uncodable, it emerges as a coelurosaur.  When chatting with Nick Gardner about this I joked "I bet if you coded e.g. Coelurosauravus' front half into the matrix, it would be coelurosaurian too."  So I coded Coelurosauravus' front half, the same parts preserved for Longisquama, and lo...

|--Proterosuchus
`--+--Euparkeria
...`--+--Erythrosuchus
......`--+--Pseudosuchia(Sclero,Ornithosu,Posto,Crocodylo)
.........`--+--Marasuchus
............`--+--Eoraptor
...............`--+--Herrerasaurus
..................`--+--Carnosauria("Syntarsus",Dilopho,Cerato,Sin,Allo,Tyranno)
.....................`--+--Compsognathidae
........................`--+--Guanlong+Dilong
...........................`--+--Microvenator
..............................`--+--Ornitholestes
.................................`--+--+--Coelurosauravus
....................................|..`--+--Effigia
....................................|.....`--Ornithomimo (inc. Pele,Longisquama)
....................................`--+--Coelurus
.......................................`--+--Falcarius
..........................................`--+--Therizinosauroidea
.............................................`--+--Ovirapt(inc.Caudi)+Dromaeo
................................................`--+--+--Troodontidae
...................................................|..`--Avimimus+Parvicursorinae
...................................................`--+--Sinovenator
......................................................`--Ornithes


I won my own bet.  Interestingly, this tree also has a lot more resolution and details which agree with the consensus (e.g. Pelecanimimus in Ornithomimosauria, Caudipteryx in Oviraptorosauria), and Longisquama is moved from Avialae to Ornithomimosauria.  Note Coelurosauravus is not thought by anyone to have anything to do with dinosaurs or birds, but still ends up as a maniraptoriform in James and Pourtless' matrix.  This fairly neatly proves my idea that Longisquama could be a far more basal diapsid and still emerge as a coelurosaur.  Note too that Longisquama is also an ornithomimosaur once the front half of Coelurosauravus is included, perhaps suggesting more signal between them than between Longisquama and birds.

So my conclusion is that very few characters were included that would support e.g. Tetanurae, Avepoda, Theropoda, Saurischia, Dinosauromorpha.  Just enough are included to get a basically consensus phylogeny even without manual characters, though not enough to properly place Effigia or the front half of Coelurosauravus.  The latter plus the theropodan assumptions in Longisquama's anatomy makes placement of thus genus particularly untrustworthy.  Certainly, not enough characters were included to survive bootstrap analysis, where random characters are deleted or repeated and the analysis is rerun.  It doesn't mean that "both the "early-archosaur" and "crocodylomorph" hypotheses are at least as well supported as the BMT [BAD] hypothesis", it means that James and Pourtless made a crappy analysis.

References-  Evans and Haubold, 1987. A review of the Upper Permian genera Coelurosauravus, Weigeltisaurus and Gracilisaurus (Reptilia: Diapsida). Zoological Journal of the Linnean Society. 90(3), 275-303.

Hou, Martin, Zhou and Feduccia, 1996. Early adaptive radiation of birds: Evidence from fossils from Northeastern China. Science. 274, 1164-1167.

Senter, 2004. Phylogeny of the Drepanosauridae (Reptilia: Diapsida). Journal of Systematic Palaeontology. 2, 257-268.

Schaumberg, Unwin and Brandt, 2007. New information on the anatomy of the Late Permian gliding reptile Coelurosauravus. Palaontologische Zeitschrift. 81(2), 160-173.

James and Pourtless, 2009. Cladistics and the origin of birds: A review and two new analyses. Ornithological Monographs. 66, 78 pp.

5 comments:

  1. Thus demonstrating the evils and irrelevance of cladistic analysis. Feduccia chortles from the shadows.

    ReplyDelete
  2. Haha! I didn't expect our chat to be the genesis of a new blog post. There's no surprise here to me. I'm also skeptical if the basal archosauromorphs or pseudosuchians were even scored correctly.

    ReplyDelete
  3. Wait. Ornithes? How is that defined, and where was it published?

    Note Coelurosauravus is not thought by anyone to have anything to do with dinosaurs or birds

    Not nowadays and not in a long time, but its name means "coelurosaur grandfather".

    ReplyDelete
  4. Ornithes Martyniuk, 2012
    Definition- (Archaeopteryx lithographica + Passer domesticus)
    Martyniuk, 2012. A Field Guide to Mesozoic Birds and Other Winged Dinosaurs. Vernon, New Jersey. Pan Aves. 189 pp.

    And yeah, Piveteau (1926) described Coelurosauravus as an ornithosuchid basal to coelurosaurs, but by 1930 it was recognized as an eosuchian or whatever they called that grade of reptiles back then.

    ReplyDelete
  5. Oh. That. Right. I almost remembered. :-] I don't like that definition much...

    ReplyDelete