[Fwd: Re: Probabilities on Phylogenetic Trees]

[James Francis Lyons-Weiler]

On Wed, 17 Sep 1997, Tom DiBenedetto wrote:

>  James Francis Lyons-Weiler wrote:
>
> >        My point remains because the only type of hypotheses of
> >        homology that are considered to begin with are those
> >        that have already passed a series of tests to distinguish
> >        homologies from non-homologies.
>
> This is a PROBLEM?
>
> >        Explicitly, the degree of
> >        corrorboration afforded to the sets of congruent hypotheses
> >        is lower than it could be because previously screened
> >        hypotheses of homology have a a HIGH, not a LOW,
> >        probability, as so the presence of congruence has a HIGH, not
> >        a LOW logical probability, so boldness is low.
>
> Here is a hint as to why your logic is faulty. In Poppers view (and
> you are using his concepts so I imagine you are making reference to
> him), a hypothesis of low boldness is indeed one with high
> probability. It is also a hypothesis which departs minimally from
> "adhocness". An ad hoc hypothesis is one which explains the details
> of particular phenomena, with minimal recourse to general notions. It
> is highly probable because it it says almost nothing beyond
> describing the immediate phenomenon, and so it has low boldness. In a
> parsimony analysis, the *least* parsimonious tree would be one which
> maximizes ad hoc statements. Those 4000 instances of hair are not
> one, they are 4000 separate instances. Describe each one
> separately,,,no generalization, no boldness, very hgh probability.
> Now think hard about what the opposite situation (*most* parsimonious
> tree) would be,,,,,
>
        But, Tom, you are switching opportunistically between
        hypotheses of homology and trees to suit your argument.
        I am NOT talking about the probability that the
        shortest tree is true.  I am talking about whether
        or not the parsimony criterion affords any amount
        of corroboration... I didn't say that the longest tree would be
        afforded a higher dose of corroboration, because that would imply
        that the parsimony criterion is a test.  The fact
        that the boldest hypothesis is, in fact, the longest
        tree is irrelevant to the suitability of parsimony
        to the task of being a critical test.

        JLW.