cladism's greatest weakness

[Richard Zander]

Tom DiBenedetto wrote:

> Richard Zander wrote:
> >I guess I'm a skeptic. Reminds me of another switch common in
systematics: saying at
> >the beginning of a paper that a reconstruction is being generated, then
> >presenting a classification at the end of the paper.
>
> What is the problem with that? Cladists believe in classifying in a
> manner that mirrors the reconstruction.

The problem is there is commonly no probabilistic reconstruction. By
probabilistic, I mean that no really acceptable basis is offered for an
expectation that at least more often than not you would be correct if you
acted upon the hypothesis.

> >What evidence have I of this? The presentation of one cladogram per
result.
> >Optimization, yes, is a valuable tool, and is the best guide to action -
> >given no loss upon failure. Is being wrong a loss to science? We have
> >hundreds of published cladograms...how many of them are somewhat wrong?
Very
> >wrong? How do you tell the difference?
>
> By examining the evidence, just like in all science. We do the best we
> can with the evidence that we have. Then we try to find more evidence.
> What more can you expect?

Well, choosing not to present a single answer when only a "best" answer is
available out of many different answers that are also supported by the data
and which are also not evolutionarily unreasonable.

>
> >A cladogram, in the presentation of
> >poorly supported detail, is not a reconstruction, which requires
> >considerable support in the form of reassurance that there are no
reasonable
> >alternatives. This is the realist burden and cladism's greatest weakness.
>
> But poorly supported cladograms are easily seen to be just that. So long
> as one uses all available evidence, even a poorly supported cladogram
> is state-of-the-art. And something to be improved upon by the next
> worker. As in all sciences, we can only start with what has already
> been learned, try to learn more, and train our students to do even more
> after us.

Tom, a poorly supported cladogram is an optimization, which is good.
Grouping similar taxa by any optimization maximizes phylogenetic info. That
makes it fine for a classification, sure enough. The difference is that
these optimized classifications are passed off as reconstructions -- not
well supported reconstructions but none-the-less very detailed poor
reconstructions. The devil is in the detail, and no attempt is made at
exorcism.

>
> >I believe there is good in modern systematics, but
> >except for publications using Bayesian Markov chain Monte Carlo analysis,
>
> Gee, I always thought the Bayesians had it all bass-ackwards, given
> that the probability calculus was designed for a consideration of the
> probabilities of *events*, not hypotheses. Probabilities of hypotheses
> are not consistent with the calculus of probability.

Bayesians deal with probability as psychological expectation.

> See both Edward's
> book on Likelihood, and Popper's works on the scientific method for
> critiques along these lines.

I have puzzled over both of these authors. Their books add to my ongoing
compilation of ingenious ways to explain away doubt and generate funding.
Likelihood does have uses, but can be easily misused. Edwards explains
support for a single hypothesis as the difference in likelihoods between
that off the hypotheses with the highest likelihood and and the  value
assigned to the hypothesis with the second highest value. No consideration
is given to other alternatives: third highest etc. values. Clearly a normal
distribution of likelihood values must be assumed for this to work. Does
this obtain in whatever it is we are talking about?Edwards is an enthusiast
for yet another quick way to be reassure oneself about essentially dubious
work. Because history cannot be directly tested, it is a fertile field for
statistical gimmicks and computerized revelation.
    Popper uses propositional logic like waving a 4th of July sparkler in
the dark to create pretty patterns of light, which persist for a while as
blind spots. His Logic of Scientific Discovery barely and limply mentions
simplicity and parsimony (p. 145), though he grossly simplifies expectation
with these ideas in Conjectures and Refutations. On the other hand, his
Poverty of Historicism is definitely against methodological essentialism (p.
136) of the type I ascribe to cladistic reconstruction, which is so
insidious and tendentious "that we are liable to feel that we see it [the
model], either within or behind the changing observable events, as a kind of
permanent ghost or essence."

    Sick of philosophy? Check out the philosophy humor pages at:
http://www.math.uni-frankfurt.de/~eik/fun/math/p.html


>
> >present forms of phylogenetic analysis do not provide acceptable
> >consideration of contrary hypotheses.
>
> Do you think all scientists should abandon their preference for the
> hypotheses that are most in accord with empirical data in favor of a
> non-judgemental consideration of all alternatives?

Again you plunk for a "best" answer and ignore my question in doing so. A
best explanation is fine for classification but insufficient for a
reconstruction.




Richard H. Zander, Curator of Botany
Buffalo Museum of Science
1020 Humboldt Pkwy
Buffalo, NY 14211 USA
email: rzander at sciencebuff.org
voice: 716-895-5200 x 351