Probabilities on Phylogenetic Trees

Richard Zander bryo at COMMTECH.NET
Tue Sep 9 08:08:30 CDT 1997

Daniel Barker wrote:
> Could someone please explain what probabilities of phylogenetic trees
> actually mean?

Well, my understanding is that two state changes are half as probable
than one, therefore a tree that has one extra state change is half as
probable as one 1 step shorter. This is arguable, since surely different
characters have different mutation rates, and the rates in different
directions may be different, and the regularity assumptions of Bayesian
analysis (all characters have equal probability of occurrence, and they
are distributed equally, and don't affect other characters i.e. no
introgression, etc.) are all confounding. On the other hand, Keynes
figured that such regularity assumptions are okay if you don't know more
information: there is a nice, lengthy explanation of probability in:

Pap, A. 1962. An Introduction to the Philosophy of Science. Macmillan
Co., N.Y.

My take on the belief-oriented and frequentist controversy is that
scientists use Bayesian-type analyses (formal or informally) when first
approaching a phenomenon, then when more information is available and
the subject is better delineated, frequency analyses are performed.
Thus, there is no controversy. With the results of phylogenetic
analysis, however, you are stuck with the clustering and there is no way
to do frequency-type analyses. So your Bayesian-type analysis has to be
very firm, very probabilistic, before you have a good hypothesis on
which to base a classification (not just a low-probability "best"
hypothesis). Thus all my discussion about probability.


Richard H. Zander, Buffalo Museum of Science
1020 Humboldt Pkwy, Buffalo, NY 14211 USA bryo at

More information about the Taxacom mailing list