The attitrion of taxonomic expertise

James Lyons-Weiler weiler at ERS.UNR.EDU
Fri Oct 18 12:19:51 CDT 1996


On Fri, 18 Oct 1996, Tom DiBenedetto wrote:

> On Fri, 18 Oct 1996 09:19:36 -0700, James Lyons-Weiler wrote:
>
> re. synapomorphies and circularity
> >The "best fit" part is circular; the data that are used to choose the tree
> >that provides the best fit are used to test homology (the data).
>
> Sorry, but I think the circularity is in your head, not in the method
> :).  The hypotheses are combined into the pattern with the least
> incongruence. That is what we call a "best fit". Hypotheses which
are
 No need to apologize; the circularity is most certainly in my head, which
I recognize the problem.  The method itself (cladistic parsimony) does not
proceed in any such manner; parsimony algorithms first calcluate a
pairwise distance (purely a phenetic), and then go about finding the
shortest route possible between terminal taxa.  The problem is the same as
finding a Steiner graph.  Character transformations are then reported as
ticks on the tree.  The parsimony tree is the tree with the shortes
possible internodes.  That is what amounts to "best fit".  The tree
itself is derived from the data, and then the tree is used to infer
something about the data.  Sounds circular to me.


> incongruent are made to fit the best pattern by dissecting them into
> as many congruent parts as necessary to fit the pattern, All of this
> is done within the expectation that true homologies are congruent.
> The data are not used to find a best fit and then that best fit used
> to test the data. There is only one step; the fitting process IS the
> test of the homologies under the expectation of congruence.


If you expect to find homology, you'll find it.  But you won't know if
other hypotheses of homology have been rejected because you're accepting
erroneous synapomorphies when the tree is wrong.  It's not a good test,
but it is a rather nice way to summarize the data.

James Lyons-Weiler




More information about the Taxacom mailing list