Polytomies

Doug Yanega dyanega at MONO.ICB.UFMG.BR
Tue Jun 17 11:12:27 CDT 1997


Kip Will wrote:
>
>>Well, suppose we are dealing with, say, 19 different characters, of which
>>10 support ((AB)C) and 9 support (A(BC)). Or perhaps I should say there are
>>branch lengths of 9 in one case and 10 in the other. In any case, ((AB)C)
>>appears to be a (slightly) more parsimonious tree. Is this really enough to
>>regard the potential polytomy as being "resolved"?  I think that to publish
>>only the "shorter" tree in this case is actually misleading. [snip]
>
>I don't see it that way. If your only evidence for showing resolution is
>character "number 19" (might be any one of the ten), then that *is* the
>evidence. If you suspect that character is misleading (from some outside non
>character evidence, say biogeography) then you should discuss that fact. [snip]
>I think that even the most stubborn supporter of a polytomy can't ignore
>character evidence that groups the taxa. You mentioned earlier that this
>begs the question of how many. *One* good (subjectivly considered so by the
>expert) synapomorphy unites two terminals versus a third terminal. Both
>dichotomy and hierarchy is established. If one is not going to accept the
>resolution on the *one* piece of character evidence then you must justify
>your choice.  Justification means disregarding one or more pieces of
>evidence. We should be careful not to "throw out the baby with the bath water".

While I agree with your conclusion - that, objectively, we have to use the
evidence we have and discuss it honestly - it is nonetheless distressing to
think how *terribly* unlikely we are to ever be able to recognize a natural
polytomy, regardless of what method of phylogenetic reconstruction we use,
or what character sets. Just sit down for a minute, summon up some quick,
small hypothetical character sets for a "species + peripheral isolates"
scenario, and see what happens - unless each peripheral isolate has only
purely apomorphic modifications relative to the parent, it is VERY hard to
generate a data set that cannot be fully resolved (as long as one accepts
"homoplasies" in any resulting shortest tree).
        If, for example, we use Harvey Ballard's example and include "grows
in rocky soil" as a character, or imagine that isolates at higher altitude
evolve more pubescent foliage, then what will happen is that these
homoplasies will appear to be the only synapomorphies in the analysis! The
species sharing such a homoplasy will form a well-supported group, leaving
other relationships unresolved (unless there are other homoplasies which
form other groups). Even if the various homoplasies suggest conflicting
grouping topologies (which might not always happen, if for example there
are ecological pressures working simultaneously on several features), it is
*still* quite likely that there will be a shortest tree. This does not mean
I would ever ignore the results of such an analysis, or that I am proposing
some alternative! - simply that we need to admit that natural polytomies
are something that may be trivially easy to "resolve" in a purely objective
analysis, rather than something which we have a decent chance of detecting,
and we may be throwing out quite a few babies in our bath water already.
This is something we will have to live with.
        I might also note that while many of us may be content with the
idea of a good character "subjectively considered so by the expert", there
are folks out there who like to slam dunk people who think subjectivity has
any place in modern systematics (or, for example, if your character set is
100% sequence data, then there can be no subjectively superior or inferior
characters, anyway, as there can be when ecological or biogeographical
observations come into play). I think the present example is one which
offers a good reason why subjectivity DOES have a place, myself. A
systematist can publish the shortest tree in the results section, yet still
go into a discussion of why he/she doesn't believe it should be accepted at
face value. That seems like honest science, to me.

Sincerely,

Doug Yanega    Depto. de Biologia Geral, Instituto de Ciencias Biologicas,
Univ. Fed. de Minas Gerais, Cx.P. 486, 30.161-970 Belo Horizonte, MG   BRAZIL
phone: 031-448-1223, fax: 031-44-5481  (from U.S., prefix 011-55)
                  http://www.icb.ufmg.br/~dyanega/
  "There are some enterprises in which a careful disorderliness
        is the true method" - Herman Melville, Moby Dick, Chap. 82




More information about the Taxacom mailing list