[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

Re: Popper and Palaeontology (was: Re: "running" elephants - locomotary analoges)

> 2. Evolution is most parsimonious (respectively, what we call "maximum
> likelihood" is really that), or at least it's parsimonious enough that the
> few extra reversals and convergences would not change the cladogram,
> except if we had lots more data and would use those to make the cladogram.

    I have never been happy with this characterization of parsimony, and I
think a number of others share my displeasure. The assertion that parsimony
analysis requires that "evolution is parsimonious" is like saying that using
newtonian mechanics presupposes that the universe works in a strictly
newtonian fashion. I do not believe any scientist who uses newtonian
mechanics would, if pressed, agree to that statement. Newtonian mechanics is
an approximation, a model; it can give us reasonable approximations of the
true values under a given set of assumptions without having a 1:1
correspondance with the phenomena it models.

    In the same way, parsimony, likelihood, or any other optimality
criterion, is never expected to accurately reconstruct the way evolution
"works" (if such a thing CAN be accurately reconstructed). It is (depending
upon your point of view) either a model intended to give the closest
approximation of evolution that we can get on theoretical grounds, or an
imperfect model that represents the only reasonable approach to solving the
problem currently available (but see my post on likelihood and Bayesian
methods), or it is a logical procedure for reconciling conflicting character
data that has no pretensions to being an overt model of evolution. None of
these applications requires that "evolution [be] parsimonious."

    To paraphrase Paul Sereno (Pete, you may wish to note the date and
time): parimony does not imply that homoplaisy is minimal, it merely
minimizes assumed homoplaisies.

    And another thing: exactly HOW could evolution be parsimonious? What
does that mean? No one has ever answered this to my satisfaction. The
possible interpretations I have arrived at almost always lead to some pretty
wild assumptions (e.g., evolution is an active process, morphological
characters are objectively real, etc.). I'm open to a good explanation, but
I've never heard one. Yet people seem really comfortable bandying this
statement around as if we all KNOW what it means. I'm waiting...

ALSO: I'm not sure I understand your comment about maximum likelihood. I can
say that several folks have constructed likelihood models that produce the
same tree as parsimony, and have used these models to address the properties
of parsimony. If you buy that line, then parsimony actually makes more
assumptions than likelihood, and has a demonstrable statistical problems
(e.g., we actually would NOT expect adding more data to produce a correct
answer under some circumstances).

    I don't miss the Clade Wars. :)