Striking a balance, weighting and Cladistics

[Thomas DiBenedetto]

-----Original Message-----
From: Kirk Fitzhugh [mailto:kfitzhug at NHM.ORG]

Kirk, Thanks for the thoughtful and insightful response. I have a slightly
different take on some of your points, and am not convinced that my view of
a cladogram as a summary of homology hypotheses is illogical. I will briefly
outline my perspective on this.
You begin by responding to my statement that characters are both hypotheses
of homology and simple phylogenetic hypotheses by pointing out that these
are really the same thing. I have no objection to that perspective. You then
go on to discuss the relationship between naming and explanation, stating:
"the naming of properties simply denotes our
perceptual beliefs that what we observe are identical in some sense. It is
only after the application of such names that we can ask the question as to
why those properties exist among some set of individuals".
You then ground your statement in the distinction that Owen made between
"homologues" (structures given the same names), and "homology" (the
explanation or justification for applying the same name).
My view on this is as follows. We have moved beyond the time of Owen, when
the justification for recognizing meaningful sameness was an unresolved
question. We do operate in a context in which the historical evolution of
life is accepted as a basic assumption, and thus the justification for
same-namedness, for homology, is rooted in the notion of shared descent from
a common ancestor. Given this fact, and my general approach to science, I
would reformulate your statement (quoted above) in this way: "the naming of
properties (in systematics)is always a hypothesis, denoting our perceptual
belief that the properties we observe are identical in a historical sense.
After the application of such names, we must then test these hypotheses
against the expectation, derived from theory, of congruence with other
homologies".
In other words, I dont think that the process of gathering up characters
into a matrix and then applying some protocol to the matrix represents an
inquiry into the nature of why those similarities exist. Rather, this
procedure is a test of whether the specific hypotheses are tenable within
the assumed context of historical descent.
Which then brings us to the consideration of cladograms. I have no objection
to seeing the cladogram as a causal explanation in itself. But you seem to
view the cladogram drawn as a result of the parsimony analysis (or any other
protocol) as being something more than a cladogram drawn from a single
character. To me, a cladogram is merely a graphical representation of a
phylogenetic hypothesis. We seem to agree that a single character is itself
a simple phylogenetic hypothesis. Can we draw a cladogram from a single
character? Yes, why not? To flesh out the example we have been referring to;
if we are doing a phylogenetic analysis of Osteichthys, and use a single
character - hair - we can draw a cladogram which will have a basal polytomy,
with one branch for each species in the study, except that one of the
branches will have a distal node which itself is a polytomy giving rise to
three branches - the platypus, the kangaroo and the human. Not a very
informative cladogram, but nonetheless, it contains all of the formal
elements that are present in a more resolved cladogram. Surely we can
interpret this simple cladogram in a manner that recognizes the branch
element below the distal node as representing the ancestral mammal, and the
distal node itself as a crude approximation of the speciation event(s) which
gave rise to the three mammalian species. We can then draw a similar, simple
cladogram for another character - say tetrapod limbs - and have another
near-bush, with one of the basal branches having a distal node that gives
rise to all of the tetrapod species. We can then combine the information in
these two simple cladograms in a logical/mathematical procedure and yield a
combined hypothesis, drawn in such a manner than there are two nodes besides
the basal node, one more distal (less inclusive) than the other,
representing Tetrapoda and Mammalia. We can continue to do this, adding
sufficient characters to eventually yield a fully resolved cladogram, and,
if we were very lucky, there might be no homoplasy whatsoever. Would this
fully resolved cladogram lack any of the formal or theoretical elements that
any other cladogram does? I dont think so. And please note, at this point we
have not made recourse to our parsimony algorithm.
The use of parsimony enters into the picture only when we encounter
characters (and their associated simple cladograms) that contradict other
characters, and at that point the parsimony principle merely guides us in
preferring one summary cladogram over another, under the criterion of
minimizing the rejection of any of the simple hypotheses in whole or in
part. I continue not to see any formal or theoretical factors that are
introduced by applying this principle.

You stated at the end of your first objection that, under my formulation,
one would not need cladograms, only homology hypotheses. I respond that
cladograms are simply representations of phylogenetic hypotheses, which you
agree are the same thing as homology hypotheses, so I fail to really see the
distinction. What we really would like to have of couse, are cladograms of
many characters, not just simple one-character cladograms, which brings us
to your second objection - the need for total evidence.

I agree with you that total evidence is a fundamental principle. You claim
that my formulation does not contain an imperative to use total evidence,
for that would only arise in a context in which various effects are seen to
necessarily be causally related. I am not sure that I really object to that
formulation, but I think it is present in my approach at a basic level. One
of the principle assumptions we all share is that the heritable traits of
organisms are potential evidence of historical relatedness. Given our
further assumption regarding the evolution of life, and the descent of
traits in lineages, it seems to me that we have a fundamental appreciation
of the fact that the totality of the heritable traits of organisms is the
universe of our concern. I think that it is also a fundamental principle of
science that explanations are to be required to account for all relevant
evidence. Once we define the universe of relevancy (all heritable traits),
the imperative for total evidence flows directly. I dont see why my
formulation would cause us to escape this imperative. To repeat a point made
above, the "complete" cladogram is not the answer to a big "why" question,
it is the selection of one out of a finite set of "which" questions - which
hypothesis best accounts for the evidence, which character hypotheses are
tenable under the theoretical expectation of congruence. "Congruence" itself
is a notion which seems to demand total evidence.

I agree with your view that (from Farris) we can see homoplasy as a
non-ancestry "explanation" for similarities, and that that "explanation"
derives from the cladogram. I recognize the explanatory power of the
cladogram. But that explanatory power does not have its origin in the
application of the parsimony algorithm; the real explanatory power resides
in the individual homology hypotheses themselves. The parsimony algorithm
merely adjudicates, logically, conflicts amongst a set of explanations -
rejecting the ancestry explanation of some characters and tossing them into
the void of "some other explanation needed", for they fail to survive the
test of expected congruence. In other words, "homoplasy" is not a real
explanation, it is a general term for "proposed explanation rejected", and
that is the only added "explanation" that the parsimony algorithm delivers.

The parsimony algorithm rejects some explanations, but does not add any of
its own. It merely implements a test, the test of congruence, and through
this, allows some previously defined explanations to persist, and not
others. It yields the most simple, efficient summary explanation of the
evidence in light of the theoretical expectation of congruence.

Or so it seems to me,,,,Sorry that I havent been brief,,
-tom