parsimony/biology

[Thomas DiBenedetto]

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

Kirk, I will respond to two of your messages (one re. things and events, the
other re. parsimony etc.) in this one message.
To begin with, let me dispose of a misunderstanding. When I said that in
Owen's time the "justification for recognizing meaningful sameness was an
unresolved question", I was not claiming that Owen was unclear in the manner
in which he distinguished homologues and homology. I merely meant that the
explanation of homology, be it descent, archetypes etc. was a matter in
dispute in the scientific culture of the time.
What I dont understand about your reading of Owen is that you seem to claim
that the term "homologues" refers only to similarities, and that only after
calling some property observed in different objects "homologues", do we then
investigate the "why" question. To me this begs the question of why
similarities are not simply called similarities, until the "why" question is
answered. To refer to them as "homologues" seems to me to imply that one has
accepted homology as the answer to the "why" question. You claim that we do
not name things *as a result of* our inference to causation, but surely we
_rename_ things (with the same name) once we accept homology as the causal
factor (as we do in renaming bones for instance).

As to the issue of individuals and classes. I havent read all of the
references you supplied, but for now my views of these matters are as
follows. I recognize that taxa can be seen as classes, from one perspective,
but also as individuals from another perspective. The two perspectives are
meaningful, because they reflect different aspects of our analysis. To
observe a strand of protein emerging from the skin of a specimen, and to
then group that specimen with another that also has such a strand, is to
erect a class of hair-bearing specimens. In this sense, taxon, as the
inferred larger group of hairy organisms is a class concept. But we do more
in phylognetics. We attempt to explain our perceptions of pattern by linking
them to abstract concepts which are deduced from our underlying assumptions.
We assume the historical existence of branching lineages through which
organisms transmit heritable traits to their descendants. These lineage
systems can be meaningfully analogized to individuals, as can the traits
themselves.
This brings us to a consideration of things and events. You identify an
organism as a thing; you are a thing. But to recognize you as an individual,
and also as a thing requries an inference. I could have seen you as an
infant, then again as a juvenile, and now as an adult. I could have
percieved three different things, distributed in time, all called Kirk. To
see these three things as an individual is to infer continuity through time;
or to recognize that things change over time, and that I have observed three
manifestations of the same thing. You claim that it is "events" which denote
the changes in things over time - I accept that you, as a thing, have
undergone changes over time, but does that make you less an individual, and
simply an instance of a class of events?

An observed character, such as hair, is a thing. But our biological
knowledge and evolutionary assumptions lead us to infer that the strand of
protein on your head is merely one manifestation of an historical individual
- that the individual (lets call it the "heritable trait") has an ontogeny,
that the strand of protein is merely a manifestation of the trait which has
undergone epigenetic and genetic events, and events of parental transmission
as well. Your parents had this trait, and their parents as well. The hair on
your father's head is the _same_ hair as on your head, not because they look
similar, or have similar chemistry, but because they are seperate
manifestations, distributed in time and space of the _same_ trait, an
"individual" with continuity. This is a literal reading of Owen's defintion
of homology - "the _same_ organ..". Your hair is homologus to that of the
kangaroo for the same reason, not because of similarity, but because of an
unbroken continuity of form and matter undergoing the events of
transmission, copying, expression, and ontogeny. In brief, hair as the
observed strand of protein, is inferred (hypothesized) to be the
manifestation of an historical individual. This is what we mean by homology.
And these are the inferences we make as we convert a series of laboratory
observations into a hypothesis of homology. An additional inference, once
again a deduction from assumptions, is that the class of specimens assembled
are in fact part of a higher-level historical individual, are part of the
lineage system of phylogeny whose existence we assume, and whose particular
form we are attempting to determine.
In other words, we empirically define classes and hypothesize that they
conform to the historical "individuals" that we deduce from our basic
evolutionary assumptions.
Just as I do not see you as merely an instance of a series of causal events,
nor merely as a thing, but as an individual (my hypothesized inferences
regarding your individuality being consistently corroborated), so to I do
not see monophyletic taxa as mere instances of causal events. They are
empirical classes proposed as, and corroborated as historical individuals.
All of this of course is prior to the implementation of the parsimony
algorithm. In an earlier message I described to you a process by which we
define a character, infer its homology, and also infer/ hypothesize the
phylogenetic relationships of the species involved. And that we can then
draw this as a simple cladogram (and yes, I agree with your comments that
cladograms, whether simple of "grand", are representations of our
hypothesized explanations). You object that we cannot then proceed, as I say
we do, to combine these individual explanations into a grand cladogram. But
I fail to see any reason in your argument as to why we cannot do so. I am
sure that you are sufficiently aquainted with the history of our field to
recognize that the process I outlined is essentially the process known as
"Hennigian argumentation" - the procedure Hennig outlined to combine the
relationship-statements of individual characters into a fully resolved tree.
That is, in fact exactly what we do. Hennig failed to explicitly address the
problem of dealing with character conflict, but the algorithm that Farris
and Kluge developed did in fact address this issue, and they went out of
their way to demonstrate how the principles that underlie the algorithm are
firmly rooted in Hennigian principles. Perhaps it is useful to recognize
that cladistic parsimony was essetially a reform of the methods of
evolutionary systematics - that the general approach follows along similar
lines. Essentially, cladists demanded explicit character defintions, decried
the use of subjective criterea for weighting, emphasized the distinction
between apomorphy and plesiomorphy, and resisted the use of explicit models
of evolutionary dynamics. But both approaches are rooted in the careful
examination of specimens, the perception of patterns of similarites, and the
hypothesizing of these similarities as homologies indicative of relationship
between the species involved.
Parsimony, once again, enters the picture only when these homology
hypotheses are in conflict with each other (effectivly always, that is).
When one cannot use the Hennigian approach of merely combining congruent
phylognetic hypotheses (simple cladograms), then one makes recourse to a
procedure that resolves conflict with a preference for the scheme (tree)
that minimizes the refutation of the indiviual hypotheses.

Hennig outlined the procedure to combine congruent hypotheses, and it is
congruence which underlies the combining strategy of the parsimony
algorithm. Why congruence? It is, contrary to your view, a deductive
consequence of our assumptions, and serves as a test of the individual
hypotheses. I disagree with your claim that the relevant test for a homology
hypothesis is whether the common ancestor underwent speciation. That might
be an interesting test (although I dont readily see how to implement it),
but it simply is not what we are after in cladistics. Our assumptions
regarding descent with modification, and the existence of a branching
pattern of lineages encompassing all of life allow us to deduce a prediction
that all true homologies must be congruent with each other and with the
lineage branching pattern. We assemble a set of characters that we
hypothesize to be true homologies. If they are, then they must be congruent
with each other. That is the test. We select the "grand" explanation that
specifies the minimal set of charcter hypotheses that simply cannot be "fit"
into a congruent phylogenetic pattern.
I do not understand why you feel that, given this approach, the cladogram
can no longer be seen as a causal explanation. I assume that you agree that
the simple cladograms (character homologies) are explanations. The "grand
cladogram merely fits them together, under the constraint of congruence, to
form an additive "grand" explanation.

On the question of the logical form of the use of parsimony, I find that I
agree with the general form of your argument, but that you focus on
different factors than I do. You claim that the essence of the use of
parsimony in cladistics is to maintain the integrity of our perceptions of
similarity as much as possible in the act of explaining. I would claim that
parsimony maintains the integrity of our separate explanations as much as
possible in the act of combining those separate explanations into a "grand"
explanation, the separate explanations being recognized as arising from
independent sources,  relevant to the question at hand.
To put this in formal terms, we apply our major premise, congruence, as much
as possible to our set of minors, putative homologies, to yield the best
possible conjunction of those premises.

Well there is much left unsaid - sorry I couldnt weave a response to
everything into this. Thanks for all your time and effort
-tom