Probabilities on phylogenetic trees

Richard Zander bryo at COMMTECH.NET
Thu Sep 18 09:01:59 CDT 1997

I'm beginning to think that all criticisms of various points of
interpretation are valid.
  I wonder if, for any particular data set, there is a pool of trees
that are similar to trees of maximum parsimony, maximum likelihood and
dendrograms of various clustering methods, all of which are not grossly
different from what we might expect given the theory of descent with
modification, and given the usual lack of information about ancestors,
are all equiprobable or almost equiprobable as being the same as the
true tree.
  If convergence is more common in closely related taxa, well mustn't it
be very common in the most closely related taxa? Imagine two closely
related species entering similar habitats, habitats that are different
from that their shared lineage; they both develop similar characters,
and this must be common in evolution. Why shouldn't three closely
related taxa with two of them sharing one or two advanced characters be
interpreted as a tree formed due to convergence than one formed by
shared ancestory? Either way is an arbitrary choice.
  From the above pool of reasonable hypotheses, we can choose to make a
classification artificially by selecting one through demanding that all
covariance (a la J. L-W.) be interpreted if possible as indicating
ancestral relationships. This is where the increase in tree detail given
by computerization goes way beyond what is justified by a theory of
  This way everybody is *wrong*, and we can finally all agree!


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

More information about the Taxacom mailing list