From: Dinogeorge@aol.com
Reply-To: Dinogeorge@aol.com
To: dinosaur@usc.edu
Subject: Re: Triassic Sauropods
Date: Fri, 6 Apr 2001 19:06:43 EDT
In a message dated 4/6/01 2:29:51 PM EST, twilliams_alpha@hotmail.com
writes:
<< Perhaps in overall form, but the devil's in the detail.
_Opisthocoelicaudia_ shows a *lot* of titanosaurian characters, and I
don't
think they can be discarded just be invoking the specter of
"convergence".
("Convergence is an insidious and treacherous trap, baited and waiting
for
the unsuspecting worker... etc etc) >>
Here's how long-branch convergence works. Suppose you and a friend play a
game of heads and tails. He tosses a coin, you toss a coin. If the coins
come
up alike (heads-heads or tails-tails), you win; if the coins come up
unalike
(heads-tails or tails-heads) he wins. After a couple of hundred tosses,
you'll both be about even, usually. His string of tosses will match your
string of losses about 50% of the time. Your tosses and his tosses are
random
and completely unrelated, yet they share about fifty "similarities."
Now replace each pair of coin tosses by a different bistate character in
two
related taxa being analyzed cladistically. Let the taxa evolve for a few
million years to give the characters some time to shake out (a "long
branch";
this amounts to playing the game with a few score coin tosses). What
happens?
Quite a few characters, just by chance, are going to shake out the same
way,
just like your coin tosses being alike 50% of the time. This is the "noise"
in cladistic analysis. How many characters have to shake out alike before
you
have a genuine phyletic relationship? Take a look at the character lists
for
oviraptorosaurs and segnosaurs listed in the Alxasaurus article (for
example). The similarities are all over the map, no system to them. Does
this
amount to relationship, or is this just long-branch noise? Likewise with
the
sauropods. Are the similarities between Opisthocoelicaudia and titanosaurs
noise, or are they indicative of some kind of close phyletic relationship?
Euhelopus has a similar problem in Sereno's analysis.
And this long-branch problem afflicts molecular analyses perhaps even worse
than it afflicts morphological analyses, since amino-acid sequences and DNA
sequences are more like discrete coin tosses than are morphological
characters.
I'd like to hear comments about how cladistic analyses might get around
this
problem.