Re: polyphyletic Alvarezsauria assemblage

[Nick Longrich <longrich@alumni.princeton.edu>]

	Alvarezsaurids show a number of bizarre characters which 
appear to unite them. One of them is the caudally expanded, 
anteriorly tapered iliac blade, seen in Alvarezsaurus and the 
Mongolian material, nothing else looks like it in the Theropoda. 
Another is the strongly procoelous, ball-and-socket caudals which are 
anything but common in theropods, and the digital proportions are 
just bizarre: the second and fourth digits are more robust than the 
third which is almost unique, other than having evolved convergently 
in jerboas and Borogovia (I checked out the possibility that 
Borogovia was an alvarezsaur, however the laterally compressed second 
metatarsal and the well-developed medial condyle of the astragalus 
indicate that it is in fact troodontid).
	Its also worth noting that fusion of the carpals is 
notoriously variable. Protarchaeopteryx for example has distal 
carpals 1 and 2, unfused, capping metacarpals 1 and 2. The holotype 
of _Oviraptor philoceratops_ preserves a full fusion of the carpals 
and metacarpals, forming a carpometacarpus, which has not been 
reported in other oviraptorids. So should we now decide that 
Oviraptoridae is really polyphyletic and Oviraptor goes with 
Mononykus, Avimimus, and pygostylians, other "oviraptoridae" are more 
primitive, and Protarchaeopteryx is a basal form related to 
therizinosaurs, all based on this "key character" (all the other 
evidence being homoplastic noise)?


>      I've been looking at the characters that supposedly unite 
> Alvarezsauria, and all I can see is a bunch of homoplasic "noise".
>   The first two (small cervical epipophyses and neural spines) seem 
> to be very subject to convergence and found in other groups. 
> Procoelous caudals and sacrals pop up here and there in all kinds of 
> archosauromorphs (incl. the bird Patagopteryx), and they grade into 
> slightly procoelous-to-amphiplatyan forms in a variety of 
> neotheropods.  I find nothing convincing in this list of so-called 
> synapomorphies.  The similarities appear to be pretty superficial 
> and homoplasic.

	I'm not clear, how is this process different from the work of 
those who would say "I've been looking at the characters which 
supposedly unite birds and dinosaurs, and all I can see is a bunch of 
homoplasic 'noise'," and then argue that each supporting character is 
highly homoplastic or easy to evolve convergently, so it can be 
conveniently dismissed without an analysis?
	The problem with manually picking and choosing characters is 
we can get absolutely anything we want if we have the will to ignore 
the counterevidence. Take a set of animals, find characteristics that 
unite them, and call those  "key characters" or "good characters", 
and everything else can be dismissed as homoplasy. And lo and behold: 
all the evidence supports my hypothesis, except forthat evidence 
which doesn't.