[Taxacom] Why character-tracking doesn't happen?

Bob Mesibov mesibov at southcom.com.au
Sat Sep 13 05:39:00 CDT 2008


John Grehan wrote:

"I thought homoplasy was just a term to describe character similarities
that do not match the preferred tree, and as such they are not an
explanation of anything. The homoplasies themselves are then 'explained'
as convergence, parallelism etc."

I think of it this way: the null hypothesis is that sharing of a
particular character state reflects common ancestry. This is a good null
hypothesis because it is evidently so often true. The alternative
hypothesis is that the shared similarity does not reflect common
ancestry. Instead, it arises because there are selection pressures
driving lineages towards a common phenotype by mechanisms other than
simple inheritance of character states.

[Notice that I'm not saying the similarity couldn't arise in different
lineages by the same developmental mechanism, turned on independently.
I'm just saying that the character state similarity wasn't handed down,
node to node, by direct descent.]

One goal of phylogeny inference should be to minimise Type I error, i.e.
rejecting the common-ancestry explanation and accepting
convergence/parallelism when in fact common ancestry is the true cause
of the character-state sharing. This is a statistical interpretation of
'minimising homoplasy'.

When you come up with an optimised tree, you find by inspection that
there are character-state similarities in separate lineages. So far as I
know, there are no tests within any phylogenetic inference program which
enable you to directly estimate the likelihood of Type I error for
particular shared similarities in a given tree. Instead, the tests are
at one or two removes and look at likelihoods of particular branching
arrangements, which contain a range of both shared and non-shared
character states.

In this view of things you are correct in saying, as I did earlier, that
homoplasy is something supposed after the fact of tree-building. It is
an explanation for those pesky non-common-ancestry cases. (IMHO even
after a Platnick pruning of mistaken homologies there will still be such
cases.)

What I would like character-tracking to do is consider those cases more
carefully, rather than dismiss them as evolutionary noise or sampling or
analytical error. If there really wasn't a Type I error, then we can
learn something about evolution. If there was a Type I error, then the
tree needs re-examining. Either way we benefit. The current custom of
ignoring the details of homoplasy seems to me to be throwing away
information.
-- 
Dr Robert Mesibov
Honorary Research Associate
Queen Victoria Museum and Art Gallery and
School of Zoology, University of Tasmania
Home contact: PO Box 101, Penguin, Tasmania, Australia 7316
(03) 64371195; 61 3 64371195
http://www.qvmag.tas.gov.au/mesibov.html





More information about the Taxacom mailing list