New Lucy: morphology vs. molecular data

Steve Manning sdmanning at ASUB.EDU
Mon Mar 29 12:50:31 CST 2004

At 11:46 AM 3/29/2004 -0600, Richard.Zander at MOBOT.ORG wrote:
>Charles Fort (Book of the Damned) pointed out that scientists often protect
>their more cherished theories by building a wall that keeps out contrary
>data. Both morphologists and DNAers at least sometimes build such walls
>(those who don't please raise your hands), and the current thread reflects
>A way to bust through the wall between morphologically and molecularly based
>theorizing is with the Bayes equation, using the probability of results of
>one or the other method as a prior. If the arrangements agree, then the
>posterior probability should go up; if they disagree, then it should go
>For instance, suppose two arrangments (one morphological and one molecular)
>disagree. To calculate the probablity that the morphologically based branch
>arrangement is correct, use 1/2 times (1 minus the probabilty of the
>molecularly determined arrangment) as a prior, which should be approximately
>the probability of the morphological arrangment as determined molecularly.
>Use the method in for calculating
>or estimating both molecularly and morphpogically based branch arrangement
>Naturally all traits (morphological and molecular) must be independent, as
>Tindall's recent post emphasizes. Differential lineage sorting and other
>problems must be addressed, and if significant must reduce the posterior

Is independence of traits not the default assumption that most make in the
absence of data to the contrary?

Might it be that the assumption should rather be that two traits are not
independent unless there are data to the contrary?  I sort of wonder
whether our grounding in Mendelian genetics leads us to behave as if the
one gene - one morphological trait model is the norm and gene interactions
are the exception in analyzing morphology.

To get around relying on either of the above assumptions, what do you think
is the best way of evaluating (calculating) the probability that any two or
more traits are "independent"?  (A different probability from the one
mentioned by Richard in the preceding paragraph).  Are there enough data
out there to do this?

After all, for example, there apparently are less human genes than earlier
estimated, indicating that (1) more genes than previously expected have
pleiotropic effects or that (2) gene interactions including timing of when
they are active and when not are so complex that a relatively few genetic
changes can skew the morphology (either toward lesser or more phenotypic
variation) more than expected from just a one-independent-gene-at-a-time

I hope the above makes sense.  Before this post I did think to myself that
maybe the greater morphological congruence with orangutans could be because
the particular divergences between humans and chimps happened to produce
greater phenotypic effects, on average, than the particular divergences
between humans and orangutans.  This would not necessarily have to reflect
natural selection as the  main driving force.

>I expect that branch arrangements that are already "accepted" will be
>amenable to this technique, and most problematic arrangments will remain
>problematic. "Rigor" in phylogenetics does not necessarily provide a
>reliable solution. A branch arrangment that is intuitively ((AB)C,D) may be
>rigorously analyzed and AB may be supported by 5 advanced states, AC by 3
>and BC by 2. The arrangement ((AB)C,D) remains equivocal, rigorously
>Richard H. Zander
>Bryology Group
>Missouri Botanical Garden
>PO Box 299
>St. Louis, MO 63166-0299
>richard.zander at <mailto:richard.zander at>
>Voice: 314-577-5180
>Fax: 314-577-9595
>Bryophyte Volumes of Flora of North America:
>Res Botanica:
>Shipping address for UPS, etc.:
>Missouri Botanical Garden
>4344 Shaw Blvd.
>St. Louis, MO 63110

More information about the Taxacom mailing list