James Francis Lyons-Weiler
weiler at ERS.UNR.EDU
Sat Mar 8 07:29:39 CST 1997
On Fri, 7 Mar 1997, Tom DiBenedetto wrote:
> On Fri, 7 Mar 1997 15:39:59 -0800 (PST), James Francis Lyons-Weiler
> >Tom DiBenedetto wrote:
> >>the inference would be different than the historical reality. You
> >seem to equate "failure" with arriving at results that differ from
> >the expectations of generaliztions drawn from different and
> >independant circumstances.
> > Sorry for being blunt, but this response seems hackneyed; you
> > elsewhere offer a justification of cladistics based on first
> > principles of evolution; certainly Hennig
> > justified cladistics as following naturally
> > from an understanding of Darwinian (selectionist)
> > thinking. Perhaps these points about generalizations
> > are ubiquitous (it does take a leap of inference
> > to take a tree as an estimate of phylogeny.
> Strikes me that there is a big difference between developing methods
> structured around deductions from higher-level theories (what you
> assert Hennig and I did), and structuring tests which are made
> sensitive to inductive generalizations from previous tests (what I
> accuse y'all of doing).
> As to the phylogeny point,,,let me make this (possibly nit-picky)
> response,,,,(one never knows what might set the light-bulbs off)...:
> Although I understand the inferential leap that those of you who
> "estimate" phylogeny are making,,for us I think it is somewhat
> different. The analogous leap which we make is not so much from tree
> -> phylogeny, but rather from character -> homology. Each
> corroborated homology sitting in a matrix awaiting 'final"
> corroboration in the congruence test, is already a little
> phylogenetic hypothesis; the leap has already been made. The logical
> combination of homologies is also a logical combination of
> phylogenetic hypotheses, such that when one arrives at a tree, all
> the leaping has already been done.
Your distinction that Hennig was deductive is presumptiuous; it was not
clear then, not is it clear now, that enough is known about evolutionary
processes to safely state that one is making an inference from the general
to specific case (deduction). I accept that you do in fact make a leap,
and a happy to see you recognize it (character -> homology). That doesn't
make it less profound, or ambitious.
Nevertheless, there remains a difference in viewpoint. let me contrast
your test of homologies to my test for plesiomorphy content. You have a
criterion by which you think (as does or did Farris) that the mpt is the
tree with the least falsified set of hypotheses of homology. I have a
criterion by which the plesiomorphy content of particular putative
outgroup taxa (which may include more than a single lineage) is compared.
Your next inference is that inferred transformations among characters
between internodes in the mpt correspond to evolutionary changes, and that
whatever changes appear to be synapormophic then corroborate hypotheses of
homology. My next inference is about the entire set of character (states)
found in the putative outgroup taxa; thus far, all I've said is that as a
set they have high or low plesiomorphy content. A more critical test for
the use of mpts as test of apparent homologies would ask whether or not
the same inferences would be drawn about a particular hyothesis of
homology (hi) if hi were absent from the character state matrix use to
select the mpt. Two new pieces of information emerge here, and shouldn't
be conflated: 1) the degree to which the mpt relies on the hi, and 2) the
utility of the mpt as a test for hi.
For plesiomorphy content, any individual character state in the optimal
outgroup may NOT be plesiomorphic; what's important is that most or many
are. Therefore, there is no leap that therefore all the states in an
optimal outgroup are in fact plesiomorphic. Other criteria (e.g.,
parsimony) may be useful to that end if prior knowledge is available that
the set of hypotheses of plesiomorphy isn't misleading.
Any lightbulbs go off? :)
More information about the Taxacom