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

Re: PDF-request: Original description of Lagosuchus and/or Marasuchus

 There's no "tone" to this other than mild frustration at the recent
 glut of cladograms with generally fewer than twenty taxa, many
 supraspecific and all selected a priori with no overriding study to
 support these choices.

Well, you sometimes give me the impression that you're frustrated at the fact that nobody sits down and does a really big study. That isn't done because it's just too much work, and you give the impression of not knowing how much work a big morphological phylogenetic analysis is.

 Case in point: Why were caseids lumped in with eupelycosaurs when
 there are plenty of other taxa out there that have more in common
 with caseids?

Like what?

I'm asking because I know a relevant, small, unpublished matrix (in press in JVP, so I suppose it'll come out sometime next year).

 Bill Parker wrote:

> Remember not only is the literature written from >first hand
> observation< (and this itself is a major assumption), but also from
> the authors personal interpretation of that observation.  This
> interpretation can and has often been influenced by factors such as
> poor preservation, bias based on preconceived ideas, subjective
> judgement of homology, misidentification, and sadly in some cases
> deliberate deception.

 Not sure why you feel the need to preach to the choir here. This is
 what I base all my papers on.

Well, no. You don't observe the specimens in person, you rely on the literature, even though the literature often lacks descriptions or illustrations of important body parts (which often only turned out to be interesting decades after publication), even though the illustrations that are there are often misleading about 3D relationships (this holds even for good photos [pers. obs.], never mind line drawings!), and so on.

 Bill Parker wrote: No.  First hand observation is crucial.

 Yes and no. See Bill's notes about preconceived ideas, etc. above,
 which support my POV. No, testing is crucial. Do the observations
 create weird autapomorphies? If so,  time to rethink.

No -- time to look at the specimens again. If the autapomorphies really are weird, as demonstrated by the specimens, _then_ it's time to rethink.

 Bottom line, we both know
 what we're doing -- most of the time.

Now that I've explained to you why potentially continuous characters should be ordered, and what the difference between a matrix for phenetic analysis and a matrix for phylogenetic analysis is, this is true (...I hope). 5 years ago it wasn't.

 Bill Parker wrote: There is absolutely no alternative for seeing the
 actual specimen.

 Yes and no. We've all seen poor and half-baked tracings of specimens
 that clearly could have been handled with more care and precision.
 Such tracings reflect the thinking of the observer, sometimes in a
 rush to get things out and published.

But you can't _tell_ whether a tracing is wrong, unless you've seen the specimen!

I've seen both published photos of the holotype skull of *Tseajaia* in dorsal view (Moss 1972, Berman et al. 1992). It took me quite some time to grasp the fact that they're both of the same skull; slight differences in angle and lighting make them look utterly different, including the positions of sutures.

(...And then there's the disgusting habit of tracing sutures in ink _on the specimen_ instead of on the photo. The skull in question has such ink on it, and maybe the amount of ink changed in those 20 years.)

 Also, let's think about Vancleava again, and it's misnesting.

You keep telling us as a matter of fact that it's misplaced, when you haven't even published your analysis.

Look, we can wait for a year. There's no problem.

 Bill Parker wrote: Sure the algorithm can handle the large dataset
 and spits out a tree, but phylogentic analysis is much more involved
 than data entry based on simple observation.

 Again, Bill, you're trying to make it sound like I don't know this.
 My cladograms end up with a single tree and some are built upon
 40,000+ data points. Subsets are also single trees. Any test you want
 to use on these trees you can use.

I don't see how the second paragraph addresses the first. I don't see any connection.

 Bill Parker wrote: …adding taxa is also something that needs to be
 done carefully given problems with 'missing data', utilizing species
 vs. supra-specific terminals, etc..

 We all know these things, Bill. I've been preaching to the list about
 the dangers of supra-specific taxa for years. Missing data? That's a
 red-herring. You already know you can add a skull, a pelvis or a
 foot to most cladograms and get pretty close to a correct nesting.

There are other problems.

One is when people score an entire presumed clade as having a character state, even when the character is unknown in some of the OTUs in question. Happens a lot, as you can see in my 2008 paper in Cont. Zool..

Another is when people throw out informative characters when they can only be scored for a few OTUs. Mickey Mortimer can give you a long rant about that, I'm sure.

A third is when a cell is scored as unknown because the literature doesn't describe it -- but makes clear that the body part in question is preserved! There are cases where I had to do this myself because I wasn't able to see the specimen.

 Bill Parker:

> Rejection of manuscripts should never be based on results but rather
> on flaws in methodology. If your methodology is sound and repeatable
> then it should be published no matter if the results are unorthodox.

 So, these are the brass tacks. You've been railing against my
 methodology. And I've been railing against yours. You say "You must
 see the fossil or be rejected!" I say, "You must include all
 pertinent taxa in your tree or you're going to get 'by default'

You're both right. That is the tragedy.

 If you have an opportunity to referee my paper, you will
 probably reject it based on your paradigm. I would do the same to
 yours if I saw systematic problems. My protest of your methodology is
 easily handled by simply adding the necessary taxa as a test to see
 if your topology remains the same.

This is only "easy" and "simple" if the literature describes and illustrates the necessary taxa unusually well or if you have the specimens at hand.

 I have done the work.

In a very, very, very sloppy fashion.

 And I am in a position to raise red flags.

Based on rather subjective reasons.

 My submissions on Amniote phylogeny have been
 rejected because, to your point, I have not personally visited all of
 the specimens, even though that has never been a requirement in
 other phylogenetic work of this size (and magnitudes smaller) from
 other workers and no one has ever tested the assumption that personal
 observation would be so superior to pulling data from the literature
 that it would change tree topologies.

It's because you disagree with the descriptive literature so much. You reinterpret fossils based on your own tracings of grainy photos -- I wouldn't take issue with a single suture without having seen the specimen.

In my abovementioned paper, I reinterpreted the homologies of a few skull roof bones of the lysorophian "lepospondyl" *Brachydectes* without having seen the specimen. But all I did was to exchange the labels on those bones; I didn't change anything about the course of sutures, presence of foramina, or whatever.

 To my point, the Amniote
 phylogency submission simply provides the overriding study that has
 been missing from Amniote studies.

The version I saw (several years ago, I hasten to add) was very sloppy. It was chock full of correlated characters, all were unordered, the limits between the states weren't explicitly justified and therefore indistinguishable from arbitrary...

 There's no glory in subjecting myself to your adjectives about my
 character, motivations and emotions. I do this because I'm only after
 one thing: what we're all after.

You're _not_ doing it. You're trying to get results without doing enough work.

 Getting back to evidence: Trialestes is indeed not well known in the
 literature. But enough has been said about it to add a few characters
 to a matrix and see (test) where it comes out. Lots of missing data,
 true, but not enough to have a loss of resolution, interestingly

Doesn't surprise me. How much resolution there is depends on the amount of character conflict, and that depends on the absolute number of scored cells, not the relative number of unknown ones.

 The work is already done and has
 been done for several years.

No. Not only hasn't it been published, it's so sloppy it wasn't worth publishing -- you didn't know what you were doing, you didn't know how phylogene.