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

Re: Resending

Félix  Landry wrote-

But the
fact is that, as far as I know, not one of the recent discoveries (let's say the
fuzzy ones from Yixian) has been correctly described, and we're already
building heaps of trees including these taxa.

Sinosauropteryx (Currie and Chen, 2001)
Huaxiagnathus (Hwang et al., 2004)
Sinornithosaurus' skull and pes (Xu and Wang, 2000; Xu and Wu, 2001)
Microraptor (Hwang et al., 2002)
Confuciusornis (Chiappe et al., 1999)
Sapeornis (Zhou and Zhang, 2003)
In addition, Caudipteryx (Zhou et al., 2000), Changchengornis (Chiappe et al., 1999), Eocathayornis (Zhou, 2002) and Vescornis (Zhang et al., 2004) have fairly detailed descriptions, though not monograph quality.

Of course new discoveries should be
reported as quickly as possible, but am I really wrong when I say that short
and superficial descriptions are more and more frequent. Well, maybe it's
just me, or maybe it's because we find more new species and have too little
time to do the job really well.

I'm not sure. Let's look at pre-90's Morrison theropods as a comparison to the Yixian theropods you used as an example above.
Ceratosaurus- described in 1884, first monographed in 1920
Torvosaurus- described in 1979, monographed in 1991
Marshosaurus- partially preserved so easily described in detail in its short original paper
Allosaurus- described in 1877, first monographed in 1920
Saurophaganax (="Saurophagus)- barely described in 1941, monograph will appear in Chure's upcoming publication
Coelurus- described in 1879, first monographed in 2005
Stokesosaurus- partially preserved so easily described in detail in its short original paper
Ornitholestes- described in 1903, postcrania monographed in 2005, skull will be monographed in Norell et al.'s upcoming publication
I don't think we're doing too badly with Yixian theropods. We have six monographed within five years of their original descriptions.

And while the rate of species description has gone up, it also seems to me the proportion of new species based on fairly complete skeletons has risen too. The pre-90's Morrison workers had eight fairly complete theropods to deal with, we have over 40 Yixian theropods.

Yes, or maybe more, paleobiology and ecology being largely based on
speculation as soon as they intend to be precise (what I have in mind is, for
example, the debate about T. rex as a scavenger or not). Phylogeny definitely is
science. Crappy analyses aren't, just like bad paleobiology isn't.

Ironically, I think the crappy analyses are the non-quantified ones. Like any attempt by the BAND/ABSRD/MANIAC crowd, for instance, is pretty unscientific. And even good qualitative analyses, like Welles' (1984) monograph on Dilophosaurus, really don't end up being as useful as good qualitative ones like Rauhut's (2003) on ceratosaurs sensu lato.

Agreed. But out of ten  analyses, how many do really contribute to a better
knowledge of these patterns?  I feel that having a tree at the end of your
paper is not really enough to make  a valuable contribution to phylogenetic

If people use the analyses the right way, I think most contribute. First of all, they're a condensed description of taxa. Sure, it's easier to make an important typo in a matrix than a paragraph, but it's nonetheless equivalent information. And no, a 1 or 0 doesn't substitute for a good illustration, but a sentence doesn't either. Second, they provide new characters for other analyses to use. Too many authors don't bother to actually test prior hypotheses by including all the characters used in them, but ideally they should.

>published morphological cladistic analyses are crap.  Too few  characters
>taxa, too subjectively defined states.

Yes, that's exactly what I was saying. I love cladistic analyses, when
they're good. The time used to make the crappy ones could be used to better know
the anatomy of the critters they're trying to classify, no?

Well, I think of this time as the beginning of the cladistic era, when we're all learning how to improve the method. I think we need to suffer through this stage to get everyone doing things right. And I see some encouraging steps being taken. When a new coelurosaur is described now, for instance, it's usually placed in the Theropod Working Group's matrix, along with some new characters. Then later papers describing new taxa and characters include the updated matrix, and thus the database is built up instead of starting from scratch every time. Also, the online databanks of characters and codings being developed by Sereno are going to be a huge step in the right direction. Much like GenBank is for molecular analyses.

>OR I could reverse your argument. Once we get the basics of paleobiology
>down, let's move on to determining low level relationships. We already
>Tyrannosaurus is a terrestrial carnivore, I can't see the sheer utility in
>all these papers trying to determine its speed, cranial strength, binocular
>overlap, etc.. You change some assumptions in Hutchinson's running model,
>and come up with completely different results. ;)

Point taken. However, even if it is very tricky and largely based on
speculation, if I had to chose, I would better learn how T. rex behaved as a living
animal in a living environment than whether Daspletosaurus is more closely
related to it or to Gorgosaurus. I don't expect everybody to share this
aesthetic, as you say, but I do expect not to be the only one...

Fair enough, and no doubt true. I think you have as much right to be happy with the current trends in paleontology as I do though, for along with quantitative phylogenetics comes quantitative biomechanics. And the latter seem far more reliable than the traditional "robust jaws means a strong bite" or "big cnemial crest means a fast runner" statements that dictated paleobiology before the 90's.

This is fine for me, but... I get the impression than most published
analyses are somewhat hiding behind the fact they were calculated, by a computer. If
you include a thorough description of characters used, coding methods, etc...
and seriously discuss the results (concordance with fossil record, and the
like), then using PAUP is perfect. Just stuffing some unjustified matrix into
it and publish the result is not.

I agree the things you mention make a MUCH better paper. Rauhut's (2003) is a perfect example (at least the version in his thesis was, I don't know if the resulting paper included all the character discussion and illustration). But like I said, even crappy analyses contribute in some ways, and we're passing through a learning stage for cladistics now.

>PAUP only does work we could do ourselves if we weren't so slow. We don't
>use parsimony because it's easy, we use it because it's objective. That's
>what makes cladistics more scientific than non-quantitative methodologies.

Well character definition is not that objective. Or it would be if we had an
absolutely complete picture of the organism including ontogeny, individual
variation and the like. But since we don't... Basically, as I understand it,
using parsimony is saying "there 51% chances that tree A is the right one, and
49% that it's tree B, so we'll accept tree A". Evolution acts on the whole
organism, not on separate parts of it; I always feel like determining a NUMBER
of shared characters involves a reasonable amount of cheating with logics.

Character definition is very subjective now, but I'm trying to reduce that in my analysis, and I'm sure others have their own ideas too. For example, instead of having states defined as "structure a more than x% of structure b", where x is an arbitrary value chosen because it supports some preconceived clade, I look at the entire range of variation in my sample and define each state as a tenth of that variation space. The character is ordered, of course. Yes, the choice of 1/10 is still subjective, but it's the highest PAUP allows in various ways, and getting much more detailed than that starts potentially involving individual variation and taphonomic distortion. Another example- when a growth range is known for a taxon, I note how and if a character coding differs between ontogenetic stages. Then, I can use this information to code other taxa known only from juveniles or subadults in ways that will limit the effects of ontogeny on phylogeny. This will all be detailed in my publication, and is one of the many reasons it's taking so long. :)
As for your dispute with parsimony, perhaps you can suggest an alternative objective method for morphological phylogenetics?

David Marjanovic wrote-

To be fair, Chure's has been upcoming for years. How old is his thesis?


I have seen computer-free cladistics, however.

Yup. Like Paul's (1984) paper on segnosaur relationships. 50 characters are used, and it takes 20 more steps to make segnosaurs theropods than it does to make them ornithischian relatives. I should try to add oviraptorosaurs to the matrix and run it in PAUP. :)

_One_ page?

Yeah. The paper's not much more than an osteology. Half a page of comparison to Megalosaurus, half a page to Acrocanthosaurus, and a couple sentences about Chilantaisaurus.

Mickey Mortimer