[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]
Re: Phylogenies, science, tea-time and more...
>I object to the use of "Carnosauria" because that name comes with a concept
>(big theropod... pictures like Burian's semi-sprawling *Tarbosaurus*). At
>least it's anchored to *Allosaurus* :-)
Well, carnosaurs have been held to be very active and unsprawling predators
before their actual systematic content was redesigned by cladistics (think
Bakker's works, or the first edition of The Dinosauria...). And as far as I
know they still are all "big theropods", very concordant with the picture we
have when we think "carnosaur". The term Dinosauria should not be wiped out
because for most lay people it means "any big prehistoric beast". I tend to
a subjective preference for saving old names even if their meaning has to
change (isn't there a Brontosauria somewhere???), but there is also a
reason for it: if we have to invent new names every time cladistics show a
new clade, we'll soon run out of names... Both approaches can generate a fair
amount of confusion, but I think it's better to see an already known name and
ask "what's it current definition(s)?" than to see a new name without a clue
of what it means. People not expert on a particular phylogenetic group should
still be able to read works about it.
There is another advantage of keeping using old names under cladistic
standards: it makes anybody see very clearly what the cladistic point of view
us discover. For example, it looks like most tetrapod phylogenies have wiped
the term Reptilia out, not to have it inlude birds or mammals, because it
wouldn't fit with the popular picture. But birds definitely are reptiles in
tree-thinking, and calling them sauropsids or anything else just won't change
anything about that. It will just prevent people from understanding that,
indeed, birds are derived reptiles. A few days ago I was explaining to my best
friend that she was an amniote. As I had previously described what the "Bob
basal amniote" looked like, she just couldn't see what the term meant. She
wouldn't catch why a very salamander-like creature AND her would be amniotes,
while salamanders and frogs wouldn't. That's because she was absolutely not
tree-thinking. I feel that using old names in trees just makes clearer what a
>> Ja! But the problem is that in a large part of current dino research
>> is no more than a tool for building some matrix, that, stuffed into some
>> computer program, will release the new version of a phylogeny, just to be
>> destroyed two months later. I'm not saying cladistics are coming from
>> nowhere and
>> pointless, just that they seem to be a little bit too overriding these
>There are two reasons. Firstly, the number of characters used per study
>currently grows faster than linearly. Some people add tens of characters to
>their matrices and publish the updates on average once per year. The other
>is the addition of taxa -- by discovery. *Eotyrannus*, *Dilong* and perhaps
>*Tanycolagreus* and the redescribed *Coelurus* can tell us more about the
>phylogenetic position of Tyrannosauroidea than we could have found out
That's why we need more detailed anatomical (re)descriptions and less
hastily done computer analyses. Huge amounts of horizontal study are necessary
produce a little amount of vertical one, if you see what I mean. And it looks
like many people want to go straight to the vertical knowledge
(reconstructing evolutionary patterns) before the hard, dry and boring
anatomical) work is done.
>A deep knowledge of anatomy can contribute a lot to finding previously
>overlooked characters and to coding them correctly (like, coding
>non-independent characters as one character rather than 2 or 10,
>them as size- and/or ontogeny-related...). And it should. :-)
Yes, and explicitely.
>> After all, most of the important dinosaur clades were
>> identified well before cladistics, on a morphological basis.
>Their interrelationships, however, weren't. Remember when some people
>thought Dinosauria was triphyletic simply because nobody had tried to show
Isn't that also because they were not really interested in finding that out?
With the fossils we now have at our disposal, if we try to draw a picture of
how the early evolution of dinosaurs, their monophyly can be very strongly
supported (by diagnosis, and by building an anatomically compelling
evolutionary scenario) without using cladistics. If they confirm that, all the
If they don't, we have to take another look, but not necessarily to find them
to be right.
>What would that comparative study do in terms of phylogenetics? Would it be
>computerless cladistics? Or would it produce a scenario-based phylogeny,
>scenario being based on "cladistics with 3 characters"?
Yes and no... Take early dino evolution again: it looks quite likely that
changes in limb biomechanics were a very important aspect of dinosaur early
success. So it's not pointless to try to draw a scenario of how dino limbs
evolved from "thecodont" limbs (all those ankle reversals and
counter-reversals...). I don't say this should replace cladistics, but I think
it could have a
bigger place along it. Or after you have done an analysis, even if it looks
quite well supported parsimony-wise, if it's in complete contradiction with
the fossil record and paleogeography seem to indicate, then this should also
be taken into consideration. This is just extending parsimony, after all,
though on a qualitative basis. If the phylogenetic picture that comes from a
tree is not that neat, that is if you can't easily transform it into a
story of how evolution happened, then alternative scenarios could be at
least mentioned, even if they're a little less parsimonious from a strictly
quantitative and anatomical point of view. Some works include this, but most
>It may not be reliable*, but it's scientific. I mean, the alternative would
>in effect be to say that A is more closely related to C _because_ these two
>only shares 3 apomorphies! See above.
Educated thinking and even intuition also are scientific, or else we would
still be living in prehistory. They should not replace quantitative methods,
just highlight both their strenghts and weaknesses.
>I like reweighting, though. But then, it is _dishonest_ to _only_ publish
>reweighted trees, as the "Basal Ornithopoda" chapter of D2 does.
>* This is something we can try to _quantify_. Some people prefer to draw
>stippled lines -- I prefer to read "bootstrap value = 79 %". Yet another
>advantage of cladistics. (It is, however, possible to produce clades that
>are both wrong and strongly supported...)
I do like stippled lines; they sometimes are scientifically very
parsimonious. :-) If our knowledge of some aspect of evolution is poorly
known, it's no
dishonor to admit it while waiting for new papers and discoveries. As I
understand it, bootstrap value is something like an indicator of reliability
tree (which I admit is just as good or even better), but what is it exactly?
>Every once in a while a very strongly supported clade emerges, and people
>think "why didn't I think of that earlier". Take the molecular analyses
>again and again and again make whales and hippos sistergroups.
>used to think it must be wrong. But early this year and late last year
>morphological analyses found the same. The mystery of artiodactyl phylogeny
>is no longer the position of the whales but that of the ruminants!
Well, as there was very little tree-thinking before cladistics, people
probably were not looking for the right things. I just say that tree-thinking
should not stop to be thinking because it's got a very handy calculation tool.
>> [...] if some new species is described from just one incomplete
>> specimen, then a discussion on its phylogenetic position, based
>> on careful examination of what is known
>> (I mean, an actual TEXT about that), should work just as well,
>> or better, than a computer calculation.
>I think this depends on how detailed the computerized analysis is. If it's
>as detailed as the text, it should be at least as reliable.
But you need a lot of text to show that said analysis is really good. That
was just the text I was actually talking about. It keeps being absolutely
necessary to make good science, even if you get a lot of help from computer
>I think it just needs to be done carefully and explicitely. If time permits
>(and as computers keep getting faster, it normally does), simply repeat the
>analysis with different coding assumptions. Has been done in the literature.
This is fair science, but doesn't it show how thin the cladistic support for
a particular view of phylogeny is? I admit without any calculation it would
be worse. But the right way to code characters, if there is one, can only be
found with more and more thorough research, etc... You've already got my
>> The thing I can't grasp about parsimony is why one femur
>> character and another cranial one should have more weight
>> than one single vertebral one, for example.
>The reason is simple. There is no objective method of determining the true
>weight of a character. So we must resort to parsimony.
It can be reasonably objective if it is discussed and everybody agrees (more
or less :-) in light of ontogeny, previous evolution, biomechanics, etc...
>Model-based approaches that take branch length into account, however, are
>coming. There is one paper that uses Bayesian methods for morphology
>doesn't explain how, at least not yet).
Great. There certainly are methods to improve cladistics. However, this
looks like more and more calculations and less and less calculation-free
thinking. I keep thinking it might be dangerous.
>> I feel cladistics are sometimes a bit like asking "are American
>> democrats or republicans?", and answering that according to the last
>> elections it's most parsimonious to conclude they're republicans. I know
>> that this is a very rough caricature, but anybody working in cladistics is
>> confronted to just that kind of problem.
>> By the way, what are "Bayesian methods"?
>They try to resolve precisely this problem! :-) It would help if I had
>actually understood them... anyway... this approach has been used in many
>recent molecular analyses. You start with... I don't know how you start. I
>only remember the two most important features: The results come with
>"posterior probabilities", a support value similar to bootstrap &
>except more reliable, because it's based on a kind of statistics called
>(Markov chain Monte Carlo); and you can apply complex weighting schemes
>(GTR...) by not giving everything the same "prior probability". The main
>advantage is that it produces probabilites rather than yes-or-no answers.
Looks very interesting, at it seems to imply discussion both before and
after the actual calculation, which is ideal. Anybody here has more precise
(website, pdf, whatever) on how this works?
>>As far as I know, knowing that
>>tyrannosaurs are closer to birds than to allosaurs did not improve in any
>>our understanding of how tyrannosaurs lived.
>It has. To start with, phylogenetic bracketing teaches us that the default
>assumption should be that *Tyrannosaurus rex* had feathers, while the
>evidence is at best equivocal for *Allosaurus*. Then we can go on about...
>perhaps its respiratory system, which may well have been more birdlike than
>that of *Allosaurus*, which in turn could imply different metabolic rates,
>different running speeds, whatever. Even how we interpret the biomechanics
>of its legs, hips and tail is influenced by our knowledge on its
>phylogenetic position. And so on. Hey, its brain!
>Really. "Nothing makes sense in evolution, except in the light of a good
>phylogeny". I'll try to look up the exact wording and the authors. :-)
Right. And, may I add, nothing makes sense in phylogeny except in the light
of, well, a good many things many analyses unfortunately don't have. :-)