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

Fw: DML

Forwarded on request, with my answers included.

----- Original Message -----
From: David Hone
Sent: Wednesday, February 01, 2006 10:08 AM

Just a couple of extra points to add to David's latest notes. Again, apologies for the editing. I hope its readable.

I have to repeat this -- the original was in HTML, so I had to insert almost all ">" myself.

> >(although the discovery of new, small tyrannosaurs may have 'broken' > >that one too).
>
>You mean if *Nanotyrannus* is real?

Actually we don't need it! Eotyrannus and Dilong are both small and early taxa

But unfortunately I wasn't able to include either of them in my database, because both are only known from subadults. -- But I thought you had said the discovery of new, small tyrannosaurs may have "broken" the strong trend to increasing size found among tyrannosaurs, so I tried to find a late, small tyrannosaur as an example.

> >This is still a large assumption, but it is fair to allow that > > >sister-taxa are usually very similar.
>
>If the branch lengths involved are short enough.

Well, we have to make some assumptions about ancestors somewhere.

Only one: that the speed of evolution does not vary totally chaotically. Using this assumption Mesquite infers the character states of ancestors -- for continuous characters.

By assuring a large period of time between the two comparison taxa we can at least be confident that a hypothetical ancestor had diverged much less for the older taxa than the younger. 

I think something is missing in this sentence, but I can't find out what...

>Squared-change parsimony is the method the program uses. It _tells_ you >how big an ancestor was. This is better than any other proxy! If you do a >little extra work, it gives you confidence intervals on the size of >selected ancestors.
>
>Squared-change parsimony takes branch length into account. >Yates (2004) >used it, but McClade, which he used, regards all >branch lengths as >identical...

I'm not sure i had that available in 2001. Looks promising however.

Oh! MacClade was certainly available, but I don't know if the versions of that time had squared-change parsimony inbuilt.

>My point is that it was not the best available data at that or any time. >For example its sources include precladistic speculations, it is biased >toward those characters that the >most analyses use, and it lacks any >bias towards bigger or newer phylogenetic hypotheses, treating all the >same.

Perhaps, but then if you want a large, inclusive tree (which we did) then there were few large-scale phylogeneis available (apart from Sereno 1999 I'm struggling to think of any quickly).

Did the time not suffice for inserting small-scale phylogenies into this large-scale one? (Erm... apparently not, see below...)

I;m not sure thats necessarily a fair criticsism of the results.

I'm sure it's fair, but may hardly change the results. Except for details, the supertree was pretty much up to date. No drastic changes in the commonly accepted phylogeny had happened in the 10 years before it, and the first that came after it was the beginning of the total rebuilding of Ornithischia late last year.

Generally they were highly conservative, and limits were placed on derived data sets (ie those that fundamentally repeat old data).

That was not my impression. The supertree includes at least two of the almost yearly versions of the ever-growing theropod matrix by Holtz (1994 onwards), for example.

> >So if they could shrink, grow or stay the same and they
> >grew then this is positive evidence for a trend towards large size.
>
>Or for diversification with subsequent extinction of the small ones.

But *that* would be Cope's Rule sensu stricto (a right leaning Jablonski polygon). Large taxa getting larger and small taxa getting laerger (ie. both the minimum and maximum body sizes increse).

Large taxa diversifying evenly in both directions and small taxa dying out -- before they had a chance to evolve different body sizes -- would mean that the left border of the Jablonski polygon would become meaningless.

>I think you could have taken The Dinosauria I and looked up >how much >material is known for each genus... of course it's >too old to include >*Unenlagia*... Besides, how much time did >you have for your M. Sc. >thesis? Half a year?

Much less. We were working on dinosaurian supertree than got in touch with Davide Pisani via Mark Wilkinson and found out about the one then in progress, so half-way through I switched projects. Hence we needed data and fast. 

Oh. I see. :-(

> > >- The size of *Dilophosaurus* is not known. The only known
> > >specimen is subadult. Thus *D.* was most probably not slightly
> > >smaller but quite a bit bigger than *Liliensternus*.
> >
> >See above.
>
>Well, this is a case where both may have been somewhere around >6 or 7 >meters, instead of one being 50 to 150 % longer than >the other.

Yes, but we took an average of species within a genus so this should correct for this at least a little.

Not in this case. *Dilophosaurus* is monotypic (as mentioned, one specimen is known), and the single referred species of *Liliensternus* (*L. airelensis*... which may not belong there) is so fragmentary that I don't think it's included in your estimates.

>_Because_ the supertree has such flaws you should IMHO not have used it.

But was that known then? There was still a great deal of contention about much of the dinosauria back then (e.g. Upchurch and Wilson on sauropods, hence the large polytomy there). This was a comprehensive attempt at producing a decent consensus, using only trees since 1980 and derived from a datas matrix.

Many were not derived from a data matrix, and almost all cladograms from the 1980s are so tiny as to be close to useless.

I wouldn't say there "was still a great deal of contention" at that time.
Even Upchurch and Wilson agree (1995, 1998, and so on) with each other and disagree with precladistic conceptions on important issues like having the titanosaurs next to the brachiosaurs rather than the diplodocoids. The only arguably big differences between these two were the the position of Nemegtosauridae (diplodocoids according to Upchurch; titanosaurs according to Wilson) and the monophyly of Euhelopodidae (Upchurch finds them as a clade at the base of Eusauropoda; Wilson finds most of them there, but *Euhelopus* next to Titanosauria).
Everyone except maybe Sereno agreed on the paraphyly of "Hypsilophodontia", but no alternative analysis was published. It was therefore inevitable that the supertree showed a monophyletic Hypsilophodontia in spite of the fact that this was outdated.
The great rebuilding of the large-scale phylogeny of Ornithischia only begun late last year. From at least the middle 1990s onward everyone agreed on one topology.

You cannot a priori eliminate trees from a study such as this because you 'don't like' or 'don't agree with them so everything went in. 

Indeed not.

Instead, newer cladistic analyses test older cladistic analyses that include the same taxa. Often they are explicitely designed to test older ones. For this purpose, they contain more taxa, more characters and hopefully fewer coding mistakes than the older ones -- and _must_ therefore be better. They repla

New studies and new specimens might change our opinions but at the time it was hailed as pretty important piece of work.

I wouldn't generalize like this... have a look at the discussion that started here http://dml.cmnh.org/2002May/msg00419.html... oops, the "Next by Thread" links won't take you far enough, so start here http://dml.cmnh.org/2002May/msg00493.html.

>and it was composed of a total of 126 published phylogenies.
>
>This is actually a disadvantage -- see above.

I would disagree.

I mean that probably most of those were demonstrably outdated and/or precladistic guesses.

>And indeed, a VERY preliminary one-second analysis of my data points to >just the same consistent size increase across Dinosauria. But, like your >analyses, it does not distinguish the effects of time and phylogeny. This >is where we could (!) get a surprise.

I'd be interested to see, obviously.

Dave Hone

Having now got under some time pressure myself, I'll try to do the analyses as fast as possible, and after that, I'll rush to publish...

Scattergrams (for Jablonski polygons) for all five characters (with reconstructed ancestors, not terminal taxa, as data points) _subjectively_ look like there was no increase. There seems to be a left wall that only the birds and a few other coelurosaurs were able to break. The first dinosaurs are closer to this supposed left wall than to the right fringe, but not much (perhaps because *Herrerasaurus* was so big and *Silesaurus*, which I used as the only outgroup, wasn't all small either).

I can send those scattergrams (screenshots, as .png files, each a few tens of kB) to interested parties. But most of the nodes are unnamed, and I won't have the time to explain what the numbered nodes are.