Striking a balance, weighting and Cladistics

[Kirk Fitzhugh]

Tom,

Thanks for your comments. Having read Owen's works on homologue and
homology line-by-line, I'd like to disagree with your statement that in
Owen's time (or for Owen in particular?) "the justification for recognizing
meaningful sameness was an unresolved question." This is not true. Owen was
actually very clear in how he distinguished the terms homologue and
homology. The former was/is named similarities, the latter a causal
hypothesis to account for those similarities. Owen was careful to make a
distinction between what one observes as effects and what one conveys as an
explanation of those effects. This is consistent as well with current
epistemology that we do not name similar properties among objects *as a
result of* our inference of some causal hypothesis. In order to explain
something, one must first indicate the effects in need of explanation. To
say we observe "x" among objects A, B, and C is the basis for then asking
why A, B, and C have that property "x" as opposed to some other property.
Applying names to the qualities of things takes place as a matter of
routine prior to explaining why we perceive the same properties.

You mentioned congruence as a test of homology hypotheses. Unfortunately,
there is nothing within the context of congruence, or the expectation
thereof, that functions as a legitimate test. If one can establish this to
be so by way of showing that congruence is a deductive consequence, then
that would be great. The problem, however, is that what one wants to test
of a homology hypothesis is the claim that some common ancestor underwent
speciation. What must be tested is the claim that some causal event
occurred in the past by way of stipulating the deductive consequences of
that claim. If a cladogram is the melding of a group of homology
hypotheses, then congruence is irrelevant to the matter. But then, given
that incongruence can result from such melding shows that a cladogram as an
explanation is not explaining homology, but shared similarities.

On the subject of "testing," you mention that "I don't 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." Ok, well, in that case, a cladogram is no longer a causal
explanation. But then, what are those "specific hypotheses?" If homology
hypotheses are separate explanations, then they have been couched in proper
historical terms. A cladogram is not evaluating those hypotheses, it can't.
In other words, we can't make the simultaneous claim that a cladogram is a
causal account for some set of effects, but then say that the cladogram is
also some kind of test of other causal accounts.

I agree with you that a cladogram is a phylogenetic hypothesis. I don't see
how anything I've said thus far could possibly be construed otherwise. I've
been very careful in my choice of words to make distinctions between what
are the effects to which a cladogram is an explanation. Similarly, I've
been very careful to correctly apply the terms homologue and homology to
indicate that the latter is the explanation of the former. So yes, we do
"draw" cladograms for individual properties. Your example using Osteichthys
and hair is correct insofar as we can infer individual causal hypotheses.
But, there is no inferential basis for then saying all we need do is
combine these individual explanations into a grand explanation that is a
cladogram. The requirement of total evidence stipulates that all "relevant"
effects be placed in the premises from which one infers a cladogram. In
other words, holding to the view that a cladogram is an explanation, the
premises from which a cladogram is inferred are not homology hypotheses,
but rather the observed effects in need of explanation plus some causal
theory. What are therefore combined are not individual cladograms that
represent homology hypotheses, but instead the effects which per the
requirement of total evidence must be accounted for together.

On the matter of parsimony, the act of using such a criterion always falls
within the realm of nondeductive inference. Parsimony is irrelevant to
deduction, which anyone that is interested can find this out from any
elementary logic text. To say one has made a most parsimonious inference is
to say that the major premise has been applied as closely as possible to
the minor premises, such that the conclusion represents as best as possible
the conjunction of those premises. With cladograms recognized as
explanations, the application of parsimony is that of applying the premise
of common ancestry as much as possible to the properties to which we have
given the same name. The cladogram-as-explanation is "most parsimonious" in
that it has maintained the integrity of our perceptions of similarity as
much as possible in the act of explaining those similarities. That is the
essence of parsimony in cladistics. There are no hypotheses within the
premises that have been accepted or rejected. As is the overriding goal of
science, we have asked the questions of why there are regularities in the
world and we come up with possible explanations for those regularities.

Gotta stop there for tonight. Thanks again for all your thought-provoking
ideas!

Kirk