Arachnids and phylogeny

[Kirk Fitzhugh]

At 08:59 AM 4/11/01 -0500, you wrote:
>Hayashi and Wheeler worked on a molecular phylogeny for the arachnid
>orders, but unfortunately their analysis gave nonsensical results.
>They combined their molecular data with the morphological data of
>Schulz for a total evidence analysis, and not surprisingly came up
>with Schulz's tree (with one or two minor differences in the
>Pulmonata).  However, because the molecular data detracted from
>rather than contributed to the robustness of the tree, this cannot be
>considered corroboration.

Can one really discern the "robustness" of a hypothesis by way of the data
that hypothesis is supposed to explain? This would seem to assume the a
priori existence of explanatory truth, from which we evaluate our
hypotheses, or that one has the power to judge when one set of data impinge
on the truthful explanation of other data. Otherwise, the correct means of
assessing the robustness of any hypothesis is to seek "corroborating
instances," which of course cannot be found within the very data used to
infer the original hypothesis. Just because we don't like an inferred
hypothesis does not mean that it's the fault of the data used to infer the
hypothesis. The relevance of data to hypothesis inference is only useful
insofar as the ability of the hypothesis to account for those data. One
cannot use that same hypothesis to decide which data from which the
hypothesis was inferred to critique those data.

Kirk

"I find that my mind is so fixed by the inductive
method that I cannot appreciate deductive reasoning:
I must begin with a good body of facts and not from
principle (in which I always suspect some fallacy)
and then as much deduction as you please."

C. Darwin, 1874

------------------------------------------
Kirk Fitzhugh, Ph.D.
Associate Curator of Polychaetes
Research & Collections Branch
Los Angeles County Museum of Natural History
900 Exposition Blvd
Los Angeles CA 90007
Phone:   213-763-3233
FAX:     213-746-2999
e-mail:  kfitzhug at nhm.org
------------------------------------------