James Francis Lyons-Weiler
weiler at ERS.UNR.EDU
Wed Feb 26 09:40:18 CST 1997
On Wed, 26 Feb 1997, Tom DiBenedetto wrote:
> On Wed, 26 Feb 1997 09:59:11 -0400, Alan Harvey wrote:
> >What I find curious is the notion that character polarity (and ordering,
> >for that matter) should be done _prior_ to the phylogenetic analysis. This
> >is tantamount to saying that we don't know the evolutionary history of the
> >taxa (else why would we be doing the phylogenetic analysis?), but we _do_
> >know the evolutionary history of the characters that constitute the taxa
> >(else how could we provide a priori polarities and ordering?).
> Interesting point,,,but, perhaps it is better seen like this,,,by
> using outgroups we assert, or assume, that we know both the character
> and the taxic history at some particular (relatively high) level of
> generality,,,but we do not necessarily know either within the
> Character polarities imposed through the outgroup criterea rest on
> the assertions that certain taxa are close relatives, and certain
> states are ancestral to the ingroup,,,,even though ingroup polarities
> are thus determined, one really only is making an assumption about
> the condition outside of the ingroup.
The situation is a tad more complex than this, by my estimation. First is
the issue of "careful choice" of outgroup taxa. This is largely
unresolved, although some rules of thumb have ben proposed in the
literature (e.g., the sister taxon criterion). However, there is an
influence of evolutionary rates on the plesiomorphy content of particular
outgroup taxa (nad combinations of taxa), and the effect of the rate of
character evolution is complicated but _which_ characters have evolved and
how they have evolved in the ingroup AND outgroup respectively. Imagine a
scenario where a sister group to an ingroup has undergone an increased
rate of character evolution, and (by chance) most of the characters that
have changed in the ingroup (and have become synapomorphies distributed
variously among ingroup taxa) have also changed in the sister taxon. The
sister taxon may be less informative than another, more distantly related
lineage within which the "right" character have not evolved (i.e.,
have remained plesiomorphic with respect to the ingroup taxa).
The plesiomorphy content of the taxon farther out may be higher than that
of the sister taxon, even when rates are equal, depending upon WHICH
characters have evolved in the respective ingroup taxa and outgroup taxa.
Obviously, if rates in one outgroup lineages happen to have increased,
even less plesiomorphy will have been retained. Therefore, oneis making
assumptions about the character evolution within the outgroup AND ingroup.
Tree-based methods of inference make it difficult to accept or reject a
particular outgroup taxon (= a particular set that contains 1 or more
taxa). For the careful choice of outgroup taxa, one can constrain
outgroup states during the calculations of character-based measures of
signal (shameless plug o' the week).
Although PAUP does not use outgroup comparisons to polarize characters
first, there are good reasons for it; one has already been referred to
(knowledge of character evolution independent of organismal evolution has
been difficult (replace knowledge with "estimation" if you like). The
other is that FIG/FOG methods proposed to make use of them have not been
well accepted. Other criteria (common equals primitive) have also been
trounced. Perhaps given an adequately justified criterion for the
comparison of the plesiomorphy content of outgroup taxa, constrained
analyses might be found to be tenable?
Another issue is how information on plesiomorphy is used. I
think a constrained analyses that restrict distance measures, or steps on
a tree, to those inferred character changes that did not return to the
supposed plesiomorphic states. Such an approach would not assume that
evolution is irreversible, just that reversals should be ignored during
the selection is made difficult. [insert speculation caveat here].
More information about the Taxacom