Positivism vs Realism

[James Francis Lyons-Weiler]

On Fri, 12 Dec 1997, Thomas Pape wrote:

> At 10.08 1997-12-11 -0800, James Francis Lyons-Weiler wrote:
>
> >        Parsimony is a criterion, but the belief that
> >        the criterion will lead to meaningful estimates of the past
> >        requires a bolus of faith I'd rather not choke on.
>
> I cannot quite understand this scepticism towards parsimony. Why should you
> EVER choose an estimate of the 'truth' that requires a LESS SIMPLE
> explanation? How can an estimate which requires MORE AD-HOC explanations
> ever be superior?

        When there is evidence that the explanation needed is more
        complex than parsimony would allow.

> >        Congruence is expected sometimes, but not always -
> >        descent with modification only partly explains the distribution
> >        of states among taxa...
>
> Did I get this wrong or do you mean that evolution - at its very heart - can
> have other causes than "descent with modification"? Even a homoplastic state
> has to be, at some level, considered descent with modification. If descent
> with modification is our starting point, expecting congruence seems
> reasonable. That we - practically always - do find a certain amount of
> incongruence is NOT at odds with EXPECTING congruence in general. When
> should we ever expect incongruence on a priori grounds for any particular
state?

        Descent with modification has always stressed the "like
        produces like" component, and the degree of modification
        can influence our ability to estimate past processes.
        My comment should have been more obviously directed at
        inheritance at the lineage level - where the diconnect
        between geneaology and cladogenesis exists.  So, yes,
        lineages can and are expected to aquire new states that
        are independent of their ancestors, ruling pure descent
        explanations - and whether these new states then become
        informative depends both on the degree to whichg other
        states sampled also changed on that internode, as well
        as whether the new states change again on the next
        internode.  Parsimony clearly assumes model of
        lots of descent with little modification, whether it
        is used as a summary, a criterion, or test (the last
        bit there for completeness).

>
> The existence of homoplasy as such, and the amount of homoplasy in
> particular, does not change the rationale behind using parsimony as the tool
> for providing the best estimate of the past.

        You are quite right here technically speaking.  But our ideas
        on the expectedprevalence of homoplasy, its independence and
        its distribution have everything to do with whether we can be (I
        mean as individuals) confident about the use of the term "best".
        There is either duplicity or myopia in the position that allows
        that in some instances, parsimony will mislead in providing
        and estimate, but that we can still can that estimate
        "best".  Why not seek other evidence from within the
        characters themselves; covariation, implied nestedess -
        and do so independent of parsimony trees?  There are numerous
        types of analyses that take precedence over parsimony
        because they can predict, for a specific case, that parsimony
        is likely to fail.  A blanket recommendation for parsimony
        is not justified, nor is one needed.


        James Lyons-Weiler