parsimony/biology

[Thomas DiBenedetto]

-----Original Message-----
From: Kirk Fitzhugh [mailto:kfitzhug at NHM.ORG]
re. Owen
>it can be argued that his understanding
>of the progression from perception to naming (homologue) to causal
>hypothesis (homology) is correct
Kirk,
Thank you very much for the extensive scholarly presentation. There are some
aspects of your explanation that think I have a bit of a different take on.
Let me explain, keeping in mind that I am not assuming that you disagree
with everything that I lay out.
Scientific explanations exist at many different levels, with much of the day
to day work of science (including systematic analyses) taking place within
the context of assumed answers to larger questions. The notion of hypotheses
within paradigms is relevant here. The question that we seem to be dealing
with here is "what are the causes of character similarities"? I get the
sense from several of your postings that you view the cladogram construction
process as yielding an answer to this question - that the results we achieve
apportion various explanations to different instances of similarities. But I
think it better to emphasize that systematics takes place within an
explanatory paradigm, and the results of our analysis merely stipulates the
extent to which this explanation can be applied to the data at hand.
Cladistic parsimony is a procedure that is embedded in the "descent with
modification" paradigm, and only yields explanations (cladograms) that fully
implement the consequences of this paradigm.

Owen was working in a different paradigm, one which answered the big "why"
question with the notion of an archtype. Much of the day to day work of an
Owenian taxonomist might be very similar to the work of a modern cladist,
the careful examination of specimens, the perception of similarities, and,
after much careful biological study, the naming of homologues. At this
point, I sense that Owen would go on to further develop the explanation by
fleshing out the nature of the specific archetype. In this sense, I
understand your claim that the explanation follows the naming. But at a
different level, I dont think one can deny that the notion of the archetype
was present before the analysis began; that the real "why" question is not
asked in the course of a specific analysis, but only less general questions
are asked, such as "are these particular organs really homologues", or "what
is the nature of the archetype to which these organs relate"?
In the course of naming homologues, the cladist is also asking less general
questions, for the paradigmatic answer to the larger "why" question,
"descent with modification", is taken as an assumption. Less general
questions that are relevant are "are these particular organs really
homologues" (same as for Owenians), and "what is the full extent of the
descent relationships of this homology"? These less general questions are
posed as hypotheses - "these organs are homologs", "this homology extends to
all these taxa" - and then the hypotheses are tested. The first of these
question/hypotheses is answered by both Owenians and cladists through the
procedures of careful comparative anatomy and related biological
investigations (the hypotheses are tested against the expectations of our
biological knowledge). But after naming homologues, after answering the
first question, the Owenian can simply proceed to flesh out the archetype,
but the cladist must do an additional test. Because of the implications of
the "descent" paradigm, one cannot fully accept similar organs as homologues
unless they also prove to be congruent with other sets of homologues. Thus
the need for the test of congruence - a final, non-biological test of
whether these similar organs really are homologues.

The more I ponder this, the more I think that I really must disagree with
your "naming before explanation" claim. As is the case in most of science,
cladistic phylogenetics represents the attempt to apply an explanation to
real world data to the greatest extent possible. The explanation is "descent
with modification", and the parsimony algorithm is simply one of many tests,
coming late in the procedure, that a specific descent hypothesis must
survive if it is to be accepted as a legitimate examplar of the general
explanation. The results of a cladstic analysis do not yield different
explanations for different sets of similarities - in fact the results do not
even offer a simple yes or no answer to the descent explanation. Characters
in a matrix have long since been accepted as being present in our specimens
because of descent with modification. The real answer that the cladogram
reveals is simply the extent, or the particular form of the descent
relationship explanation; not its correctness. For example, if you have an
analysis of 10 taxa, named A through J, and a character that unites A, B,
and J - i.e. a hypothesis that A,B,and J share descent from an exclusive
common ancestor- but your final tree is fully bifurcated with A and B on one
side and J on the opposite side, then you have not refuted the larger
descent explanation, you simply have broken up the singular descent
hypothesis into two; A and B are descended from a common ancestor that had
the trait, and J is descended from an independent ancestor that had a
similar trait. As a commonly heard quip in cladistics puts it "homoplasy is
simply homology at a lower level".

In summary, I would say that explanations are formulated before one
begins,(as Darwin remarked "How odd it is that anyone should not see that
all observations must be for or against some view if it is to be of any
service"), and the actual analysis merely attemtps to apply the explanation
to percieved similarities, to the greatest extent possible. This is true if
the explanation is archetypes or descent. Naming homologues represents the
arrival at the point at which one is willing to accept that the explanation
is tenable in a given situation. For Owen, this point is reached at the end
of the biological investigations. For cladists, one must wait for the
results of an additional test, the test of congruence, for that test is
demanded by the particular nature of our paradigm. Once the point is
reached, in either paradigm, one can articulate with confidence the general
explanation in a particular case. This may appear to represent an
explanation following a naming, but the analysis has not _chosen_ the
explanation, it has merely knocked down the obstacles (passed the tests)
that we (as good Popperian scientists trying to falsify our specific
hypotheses) have erected to our acceptance of the explanation.
-tom