The attitrion of taxonomic expertise

James Lyons-Weiler weiler at ERS.UNR.EDU
Fri Oct 18 08:47:24 CDT 1996

On Fri, 18 Oct 1996, Doug Yanega wrote:

> Actually, I'm quite sure it goes beyond that, as James suggested - it
> is not that morphologists *choose* not to use statistical tests, but
> that many of the statistical tests used to evaluate molecular data can
> NOT be used on morphological data; what is the *probability* that a
> bristle on a fly head will appear? Disappear? Become thicker, or longer,
> or flattened, or plumose, etc.?? Molecular statistical analyses
> frequently invoke character transition probabilities, for which there are
> available estimates. And yes, I have seen folks who claim this makes for
> inherent superiority. I notice, too, that despite claims otherwise at the
> outset, this *has* become a bit of a molecules vs. morphology debate
> though this is perhaps only natural, given that the reduction in the
> proportion of morphologists in the systematic community does indeed seem
> to be directly related to the growing molecular ranks. Whether or not any
> of us put stock in the philosophical aspects of the debate, it does seem
> to be translating into a genuine battle as far as employment and funding,
> and I for one find it disheartening to see traditional systematics
> getting the short end of the stick so consistently, when everyone DOES
> realize how truly important it is.

With the new kinds of statistical analyses, morphology are given equal
footing.  I've seen cases where the morphological data blows available
molecular data out of the water! I'm not for or against morphological or
molecular data; it seems to me that the objective of a given study should
determine that.  What bothers me most is that molecular sequence data
results in gene trees, and not always organismal (species) trees, and then
people study morphology on the gene trees.  Sometime the molecular trees
are wrong (e.g., long branch attraction, mutation saturation); in fact,
estimates of the parameters for the complex models o nucleotide evolution
are sometimes wrong, but we can't test them anymore than we can directly
measure the probability of hairs growing in places on bodies in the past.
Take for instance, maximum likelihood.  It does not include parameters for
mRNA editing, and those sequences don't even make a phenotype (protein)
without being changed after transcription!  When the models don't include
parameters that we don't know about (i.e., increased rates of evolution
resulting in long branches), they are underspecified.  It takes only a
slight error in the parameter estimation for the result to be wrong.

The statistics I describe are simple, don't require specified models of
character evolution, have clear statistical bases... I could go on with my
propaganda, but I'm making myself ill.  I hope they help.


  \  /   /    \  /           JAMES LYONS-WEILER           ______________
   \/   /      \/                                        |..............|
    \  /       /                                         |..............|
     \/       /              DOCTORAL PROGRAM IN         |..............|
      \      /               ECOLOGY, EVOLUTION, AND     |...***........|
       \    /                CONSERVATION BIOLOGY        |..*****.......|
        \  /                                             |.******.......|
         \/                  1000 VALLEY ROAD/186        |********......|
    ______________           THE UNIVERSITY OF            --------------
   | will perform |          NEVADA, RENO
   |  statistical |          RENO, NEVADA 89512-0013
   | phylogenetic |
   | analyses for |         "(Biology) is not religion; if it were, we'd
   |    food      |          have a much easier time raising money."
    --------------                       -Leon Lederman

More information about the Taxacom mailing list