Re: trichotomy

Robert Scott (rscott@mail.utexas.edu)
13 Nov 1994 03:19:01 GMT

In article <sarimaCz52AG.M8H@netcom.com> Stanley Friesen,
sarima@netcom.com writes:
>Now, in the case in point, the two branches are *very* close
>in time, which is one of the "bad" combinations. [Also, the
>branches are *relatively* deep, but not really deep enough
>for the full random linkage effect to show up].
>
>So, this tends to support a skeptical attitude towards the
>molecular phylogenies in this case.

I tend to share your skepticism here.

I think there is an important distinction between phenetic and cladistic
approaches that is often overlooked by molecular phylogenies. Many
molecular phylogenies (like thos based on DNA-DNA hybridization) are
essentially phenetic - they assume that overall similarity implies
relatedness. This will only be true when overall similarity can be
unambiguously related to the time since two tax diverged - when molecular
evolution is purely clocklike. This assumption appears to be at best
tenuous.

Rather than overall similarity, phylogenies should be based on patterns
of characters that reflect evolutionary branching events. Shared-derived
characters would seem to reflect actual evolutionary branching events
better than overall similarity. Indeed, in the morphological arena,
cladistics appears to have triumphed over numerical taxonomy. Weighted
analysis of characters is prefered over measuring overall similarity.
Why then do many molecular phylogenies continue to rely on overall
similarity?

Cladistic analysis is applied to some kinds of genetic data like DNA
sequence data. However, this seems inappropriate. Such an analysis
essentially makes the base pair the operational unit of phylogenetic
analysis. Given that all base pairs only have 4 potential character
states, A,C,G and T, this seems foolish. Homoplasy is likely to be common
due simply to random chance - any to base pairs will automatically have a
12.5% chanc of being identical.

Thus, I would have more confidence in phylogenies based on the weighted
analysis of higher level characters. It is worth noting that the only
genetic evidence that relies on a weighted analysis of higher level
characters - a cladistic analysis of chromosomal karyotypes - is in
agreement with morphological evidence that puts chimps and gorillas in
their own clade. If these same characters are used to group humans,
chimps, and gorillas according to overall similarity, chimps and humans
form a cluster!