A priori homology (was positivism)

James Francis Lyons-Weiler weiler at ERS.UNR.EDU
Sat Dec 6 16:58:52 CST 1997

On Sat, 6 Dec 1997, MAGarland wrote:

> Tom:
> >I'm not sure what he is referring to here. The traditions and
> >methodologies of comparative anatomy have never struck me as
> >particularly naive.
> Doug:
> >Did Sneath propose an alternative "non-naive" way of determining homologies
> and >states?
> Sneath was referring to the general problem of determining "what should be
> compared with what."  He admitted this was a problem for phenetics as well as
> for any scientific comparisons.  His view appears to be that Hennigian
> cladistics naively skips over this problem.  Two more quotes:  "to assume that
> one could solve evolutionary homology without even considering general
> homology and, furthermore, to determine ancestral and descendant characters
> states a priori appeared to me impossible, indeed quixotic."  "One cannot
> reconstruct phylogeny from synapomorphies if one must first know the phylogeny
> to recognize correctly the synapomorphies."
> Unfortunately, Sneath didn't propose a "non-naive" way to determine homology
> in this paper.  And I think you can guess that he'd favor a straight phenetic
> approach.  For phylogenetic reconstruction, he seemed to favor the dreaded
> statistical phylogeneticists like Felsenstein.

        I have consistently, I hope, avoided jumping to push
        the techniques I have and am developing to examine and
        measure the types of structure in discrete matrices
        after they have been assembled, but here I can't refuse.

        Both sides of the Great Debate of the '70s have claimed
        victory, (i.e., cladists vs. pheneticists).  It is with
        some irony that a large number of analytical techniques
        have emerged that actually uses both phenetic and
        cladistic types of information.  How well a phenetic measure
        matches a cladistic measures happens to be a decent
        indicator of how well the matrix meets many of the assumptions
        of cladistic parsimony; i.e., how much the matrix looks like
        one that gives a decent tree when cladistic parsimony is
        the criterion of choice.  Kim has shown the utility of
        finding common trees using different criteria, and I think
        the message is not independent: we can better understand
        what we're in for with our data if we try to determine the
        types of structure present in a data set, and then set
        about to quantify the amount of each type present.  If a
        pheneticist examines a phenogram, or a cladist examines a
        cladogram, they are both looking at only a portion of the
        information present in the matrix - and rather often
        have little idea about which components of matrix
        structure are driving their favorite analysis.  The
        independent types of analyses I find so useful (not all
        of which are my own) are attempts to make explicit
        and viewable the ongoings and information and noise
        present in data matrices.  My ear to the rail tells me
        that others have been developing tree-independent measures
        of signal. Spectral analysis, for instance, is useful for
        looking at the parsimony statements present in a matrix -
        and is one step closer to revealing incongruence without
        looking at noise as signal.  If a data set that would
        result in a tree reveals "congruence" where these other
        tests demonstrate sufficient incongruence that places the
        tree-based summary of congruence (i..e, the "reconstruction")
        in serious doubt, then it seems clear which is the more
        severe test.

        James Lyons-Weiler

More information about the Taxacom mailing list