Re: Pneumaticity in Triassic pterosaurs

["David Marjanovic" <david.marjanovic@gmx.at>]

----- Original Message ----- 
From: "David Peters" <davidpeters@att.net>
Sent: Sunday, May 17, 2009 3:36 PM
Subject: Re: Pneumaticity in Triassic pterosaurs


> First of all, if Dr. Laurin was a referee, then you just blew his 
> anonymity. I hope that's okay with your mentor.

He almost always signs his reviews... are you sure he didn't this time? 
Also, I don't know for which journal he reviewed your manuscript; didn't you 
send it to several?

> Second, specifically, how, in your view, does a phylogenetic data matrix 
> differ from a phenetic one when dealing with fossil organisms? Are not 
> both concerned with morphology, characters and ratios?

We have _been_ through this! Right after Steve Brusatte and colleagues 
published their science paper on the speed of morphological evolution of 
various archosaur clades in the Triassic, you noticed that the dataset in 
the supp. inf. didn't give the tree that also was in the supp. inf., and 
Steve came in and explained why! Here! Onlist! And before he did, several 
other people discussed with you about this!

<headshake> Brain like a noodle sieve.

Short answer: It is very easy to put characters into a matrix that are 
correlated with each other, or that lack phylogenetic signal because they 
evolve too fast for the problem in question (imagine feather color in an 
analysis of Neornithes -- this is a very real problem in molecular 
phylogenetics, where it's called "saturation"); in a phenetic analysis, this 
doesn't matter (or may even be desired, as it was in that of Brusatte et 
al., who optimized their carefully designed phenetic matrix onto a 
phylogenetic tree to see which changes had happened along each internode), 
but it can seriously screw up a phylogenetic analysis. If two correlated 
characters are in the same matrix, that's like having a single character 
with double weight in it, and there's no reason for that double weight. It's 
arbitrary, and will give a wrong result. If a character without phylogenetic 
signal is in the matrix, that means there's random noise in the matrix which 
will destabilize the tree and may support a slightly less parsimonious one 
than the one that would be found without it.

And phylogenetic analyses are often very sensitive to "garbage in, garbage 
out". Fig. 10 of the paper by Anne Warren in the December 2007 issue of JVP 
(on the Early Carboniferous tetrapod *Ossinodus*) shows two totally 
different trees that were derived from fairly large matrices that differed 
in a single cell. Check it out and be amazed. And here 
http://dpc.uba.uva.nl/ctz/vol77/nr03/art02 is a whole paper on that 
phenomenon.

Long answer: Look it up. Steve's explanation was excellent as far as I 
remember.

> Third, if a given matrix and its scores deliver a single MPT in which all 
> sister taxa are similar in size, shape, niche and chronology, isn't that a 
> good result?

A result can't be better than the matrix it's calculated from... :-)

> > > I can show you the path of most parsimony if
> > you're interested.
> >
> > Erm, PAUP* will do that... :-)
>
> By that I meant, if you are accidentally excluding any pertinent taxa, you 
> should be told.

Oh, sure. I'll send you the taxon list once I'm done with making it. But I 
don't think you'll find any omissions; I'm totally megalomaniac on taxon 
inclusion. :-)

> And then on to the M. Habib replies:

(More precisely, my reply to his reply.)

> > Probably the idea is that if taxa are too distant
> > phenetically -- that is, by a subjective, intuitive estimate
> > of phenetic distance, not even by any attempt at
> > quantification --, they can't be sister-groups.
>
> No, simply that a larger test found better sister taxa here excluded.

I was talking about your use of "family resemblance".

> > - For the probably twentieth time, the vertebrate fossil
> > record -- let alone our knowledge of it!!! -- simply
> > isn't complete enough that we could expect to do what
> > the Unnameable Ones ask us to do (to present them a complete
> > series of transitional fossils documenting each and every
> > speciation-or-whatever between two arbitrary endpoints).
>
> Respectfully, that's an opinion, David.

Show me. :-)

Let me advance an argument: As of today, there are 6,487 known extant 
species of lissamphibians (http://www.amphibiaweb.org/, bottom). Let's 
ignore all those whose ecological niches didn't exist in the Early Permian. 
I suppose this means a few thousand species of rainforest frogs. We might be 
left with... let's err on the side of caution and say only 1,000 species.

Then let's find out how many potential ecological analogs to those 
lissamphibians we know from the Early Permian fossil record. Maybe 20 
species of "branchiosaurs"; maybe 20 species, probably less, of (other?) 
amphibamids; a handful species of "microsaurs", not all of which count 
because some are more analogous to amphisbaenians or maybe other squamates; 
seymouriamorphs count because of their aquatic larvae, but for the entire 
Early Permian (Cisuralian) this adds at most, like, 10 species (half of 
those in places with very uncertain stratigraphy)...

I estimate about 50, distributed through thirty million years. Today, right 
now, we seem to have two hundred times that. If that isn't a red flag that 
shows us how incomplete the fossil record is...

> Also, when you say 'complete series' I hope you don't mean every mother's 
> son. What I'm saying is you'll be able to line them up like the famous 
> Australopithecus to Homo sapiens march.

Famous march?

You seem to have missed the last few decades on that one. It's a pretty 
bewildering bush. We don't even know where *Kenyanthropus* and *Homo* fit 
into the *Australopithecus* tree, or maybe outside of it...

> Or as I did from bacteria to humans in "From the Beginning" (1991)now 
> available only from Amazon.

The two reviews I find on Amazon are not detailed enough to explain what you 
mean...

Incidentally, the "from bacteria" part just got more interesting:

James A. Lake, Jacqueline A. Servin, Craig W. Herbold & Ryan G. Skophammer: 
Evidence for a new root of the tree of life, Systematic Biology 57(6), 
835 -- 843 (December 2008)

Abstract:
> >
Directed indels, insertions or deletions wihtin paralogous genes, have the 
potential to root the tree of life. There we apply the top-down rooting 
algorithm to indels found in PyrD (dihydroorotate dehydrogenase), a key 
enzyme involved in the de novo biosynthesis of pyrimidines, and HisA 
(P-ribosylformimino-AICAR-P-isomerase), an essential enzyme in the histidine 
biosynthesis pathway [these two are paralogous to each other, having arisen 
from a gene duplication before the origin of the crown-group of life]. 
Through the comparison of each indel with its two paralogous outgroups [the 
second being a gene called HisF], we exclude the root of the tree of life 
from the clade that encompasses the Actinobacteria, the double-membrane 
prokaryotes [gram-negative bacteria], and their last common ancestor. In 
combination with previous indel rooting studies excluding the root from a 
clade consisting of the Firmicutes [gram-positive bacteria], the Archaea, 
and their last common ancestor, this provides evidence for a unique 
eubacterial root for the tree of life located between the 
actinobacterial--double-membrane clade and the archaeal-firmicute clade. 
Mapping the distributions of genes involved in peptidoglycan and lipid 
synthesis onto this rooted tree parsimoniously implies that the cenancestral 
prokaryotic population consisted of organisms enclosed by a single, 
ester-linked lipid membrane, covered by a peptidoglycan layer.
<<

Another surprise, published earlier, is that Firmicutes is paraphyletic: the 
bacilli are more closely related to us than the clostridia. Sum (from fig. 
2):

--+--+--Actinobacteria
 |  `--gram-negative bacteria
 `--+--Clostridia
    `--+--Bacilli
       `--+--Archaea (imaginably paraphyletic)
          `--Eukaryota, except for the many genes which come from 
gram-negative bacteria (probably all from the mitochondrium)

> The whole idea is just to get a picture of what really happened.

But that picture comes from reading the tree, not the other way around. 
Phylogeny first, scenario second.

> > In fact, that's not limited to vertebrates. It does not
> > even hold for ammonites and conodonts; that's why
> > stratophenetics (using total resemblance and stratigraphy to
> > construct a tree and then interpret that tree as a
> > phylogenetic one) does not work.
>
> Okay. If you're referring to my submission,

I'm not.

> the stratigraphy was added after the phylogenetics, just to see where 
> things fell.

I didn't see that part, and there's nothing wrong with it. There's also 
stratocladistics, which tries to include stratigraphic information in its 
parsimony decisions. My point was that _even for taxa with a much better 
fossil record than vertebrates_ the fossil record is too incomplete to allow 
stratophenetics to work, in other words, too incomplete that we could expect 
to be able to see all the family resemblance you are talking about. That you 
didn't do stratophenetics is not relevant to that point.