[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

Re: Benton 1985, etc.

Okay, David, sure more taxa is better than fewer taxa. That goes
without saying.
Still, if all you want to do is find out if pteros are closer to one
taxon or another, you can find out with a dozen of each. Or five if
you want to get down and dirty.

This is wrong, and I'm speaking from experience. Using just 7 "microsaurs" and 6 dissorophoid temnospondyls plus an outgroup or two to find out the position of Lissamphibia in the tree will not get you a robust or defensible result (even when the albanerpetontids are included, *Eocaecilia* is coded after its 2007 redescription, and *Rubricacaecilia* is considered). This does not change even if you add the only well-known lysorophian (*Brachydectes*). For example, plenty of character states come out as autapomorphies of the ingroup, even though e. g. other temnospondyls that are not in the matrix retain the plesiomorphy. To find out the position of Lissamphibia, or arguably even its monophyly, you must, I repeat, must make a full-grown analysis of all limbed tetrapods. Any less than that doesn't work. (I even have unpublished reasons to think that the latest such analysis, with 102 taxa and 341 characters, still didn't resolve the question, though that may have more to do with the way of coding than the size of the matrix.)

For an outgroup, be safe. Try Ichthyostega.

Most certainly not. *I.* is so far away that you are inviting long-branch attraction. Plus, it has its own bizarre autapomorphies (like that middle ear phenomenon) which would increase that long-branch attraction even further. Also, using a single outgroup is a bad idea: several are necessary to determine the character state distribution at the base of the ingroup. People who only use *Allosaurus* or *Allosaurus* and *Sinraptor* as outgroups for a coelurosaur analysis don't even find tyrannosauroid monophyly (see the *Mahakala* paper vs Rauhut 2002) because *Tyrannosaurus* is much more similar to the carnosaurs than *Dilong* is. The only suitable outgroup (if finding out the position of the pterosaurs is the only goal) is a lot of stem-diapsids (plus perhaps a few other sauropsids as well) because you must make an analysis of the entire diapsid crown.

I've also checked matrices for miscodings: Irmis et al. 2007 and Hill
2005. It's daunting, but it can be done. You asked about time put in?
10 years with time out for meals, sleeping, etc. It's fun, so the
work goes fast.


10 years just for your amniote analysis, and not counting your day job?

re: pteros next to the Archosauromorpha. Well, lots of interest there. Benton 1985 nests pterosaurs between Lepidosauria and Trilophosaurus + Rhynchosaurs + Prolacertiformes + Archosauriformes. With Trilophosaurus and Rhynchosaurs actually closer to Lepidosauriformes (they are Rhynchocephalians with unfused ankles) than Archosauriformes, that's a very good nesting. 

In hindsight.

Also, why do you get *T.* and the rhynchosaurs as rhynchocephalians? Surely you know that the "beaks" of rhynchocephalians and rhynchosaurs are not primary homologues (the latter, like that of *T.*, being real beaks, consisting of the pmx with presumably a horny covering, the former being acrodont fused teeth)?

The only other nesting that is not
reflected in the comprehensive cladogram

(which is, again, not a cladogram, because it's not the outcome of a cladistic analysis)

is Prolacerta with Macrocnemus. In the absence of other taxa, evidently
convergence won out here.

See? Throwing too few taxa at a complicated problem with lots of convergence does not work.

In 23 years since Benton 1985, no one has published a matrix that includes any lepidosauriformes with pterosaurs. If you can find one, please send it to me. That's a powerful a priori paradigm at work. 

That's true. I'll probably be the first to do it.

You wrote: "No, I can't imagine that finger IV grew secondarily, so
it cannot have done that"? I'm not saying you're wrong; I'm saying
that if you're right, you're right for the wrong reasons.

Parsimony is the reason I use.

Then you don't need to mention a single character. Parsimony works across the whole matrix. It is what tells us that toe V did indeed grow in sauropodomorph evolution.

You wrote: *_WRONG_*. Benton had Younginiformes (as a whole!!!) and "Lepidosauriformes" derived from a common ancestor, and he had that common ancestor and Archosauromorpha derived from another common ancestor. This common ancestor is *_NOT_* "Youngina and kin". Why do you keep making such confusions? An ancestor on a cladogram is a node, not a terminal taxon!

It's shorthand, David. Youngina and kin means Younginiforms + more
primitive diapsids.


That's an outright misleading shorthand. Please abandon it.

You wrote: Why can't it happen that, at least with our imperfect knowledge of the imperfect fossil record, two large groups can be sister-groups of each other?

That's a world view that is the current paradigm. 

And that makes it wrong? :-)

That world view
permits pterosaurs, ichthyosaurs and diadectids, among others, to
remain mysterious and poorly connected from other amniotes.

Yeah, and? What makes you think we must already have found all of their side branches?

Besides, it's not true. Ichthyosaurs and thalattosaurs, for example, are quite similar, and diadectids and *Tseajaia* don't differ in much more than their teeth while being very similar to *Limnoscelis* and a bit less similar to Amniota as a whole.

This results from using suprageneric taxa and just a few dozen taxa at
most (i.e. your suprageneric "two large groups").


It is not at all necessary that those large groups in the example be OTUs.

A comprehensive cladogram, on the other hand, using species and a few hundred taxa demonstrates that such enigmas do indeed have close generic sister taxa somewhere on the tree. And you'll be able to chart and see the spectral blend of morphologies across the tree. 

This is only a good argument if that blend is the most parsimonious tree.

Try not to hang your hopes on just one character. I'll just hazard a guess, but I'll bet this unpublished pterosaur has a prepubis, an elongated pedal digit V, hyper-elongated manual digit IV, multiple cusp teeth, chevrons parallel to the centra, an elongated coracoid, a sternal complex (interclavicle + clavicle + sternum), a non terminal naris, an antorbital fenestra without a fossa and other characters no dinosauromorph has.

To be fair, ao fenestrae without fossae are seen just outside the archosaur crown-group, two nodes away from Dinosauromorpha, or three if the pterosaurs are included.

re: Huehuecuetzpalli: you wrote: What is so pterosaurian about it? Well, its actually closer to Cosesaurus, which leads to pteros, but here goes:

1. non terminal naris
2. ascending process of premaxilla extends beyond naris
3. low distinct coronoid process (a bump)
4. posteriorly sharp quadratojugal making poor contact with quadrate


It doesn't have a quadratojugal at all.

5. short descending process of squamosal
6. clavicles, interclavicle
7. posterior lean to quadrate

All of this, the lack of a qj included, is normal for squamates. What do you mean by "clavicles, interclavicle" -- surely not their presence, which is a plesiomorphy?

8. eight cervicals (pre-scapula verts)

Please, please, please get another definition. This one depends on the size and the life position of the scapula -- the former is variable and the latter not trivial to reconstruct as well as very unstable post mortem. I recommend the transition between cervical ribs (short) and dorsal ribs (long).

9. low scapula

You mean "poorly ossified scapula". Squamate scapulae -- in fact tetrapod scapulae in general -- ossify from the glenoid upward, and *H.* is a baby. You must code all immature specimens as unknown for this character.

10. poorly ossified carpals (phylogenetic, not ontogenetic character -- this is how the centralia migrate to become preaxial carpals and pteroids)

Dude, this is an ontogenetic character. Denying it doesn't make it otherwise. Don't act as if you knew nothing about tetrapod ontogeny in general and squamate ontogeny in particular. You are making yourself ridiculous.

This doesn't mean that there's zero phylogenetic signal in this character. The signal lies in the "point" in ontogeny at which the carpals ossify. In many aquatic tetrapods they never ossify, so you can clearly use "carpals ossified IN ADULTS: no (0), poorly (1), fully (2)" as a character (if you quantify it). But simply saying that this ossification is "not ontogenetic" is nothing short of ridiculous. Never mind the, as far as I can see, complete lack of aquatic features in *H.*, and the obvious fact that it's a baby (all epiphyses are still separate, as mentioned, even those of the suproccipital).

11. manual digit IV is the longest
12. metacarpals III and IV subequal
13. manual digit V reduced but retaining three phalanges
14. fibula less than half diameter of tibia
15. unfused tarsals
16. metatarsals III and IV subequal and longer than I and II
17. pedal digit IV the longest
18. Metatarsal V short and torsioned
19. Pedal 5.1 elongated.

Sounds all normal for squamates or even larger groups -- 18 for example is a famous autapomorphy of the diapsid crown-group, and 11 and 17 are the normal state of affairs for the whole clade of pentadactyl vertebrates (plus probably *Tulerpeton*). In other words, you are citing plesiomorphies as if they were apomorphies. You're even aware of that: you say "retaining" in 13.

What does 15 mean? Unfused astragalocalcaneum? I bet *H.* doesn't have that, because the astragalocalcaneum fuses in squamates before it even chondrifies. I'll check.

20. Tail chevrons are preserved parallel to centra. Was this their orientation in life? 

I'll try to find out.

21. Posterior tail attenuated. 

Can you quantify this?

22. and it's the right size 

It isn't any size. It's a baby.

Besides, when you sent me your matrix 2 1/2 years ago, size was a binary character: bigger or smaller than a house cat. This is not how to deal with a continuous character. Why not make a lot more states out of it? (This would allow you to give *H.* a minimum size: however large the adult may have been, it can't have been smaller than the preserved specimen, so the proper coding would be partial uncertainty rather than just a question mark.) And how did you determine the limit between the states? Surely you didn't just pick it arbitrarily, but did some kind of math?

Having size as a character in the matrix is also dangerous because it's difficult to exclude that you have any size-related characters in the matrix.