Striking a balance, weighting and Cladistics

[Thomas DiBenedetto]

-----Original Message-----
From: Ken Kinman [mailto:kinman at HOTMAIL.COM]
>    I don't how many nucleotide changes were involved in the evolution of
>feathers, but it was a large number.

Ken, I dont know how many nucleotide changes were involved in the
modification of an archosaurian scale into a feather either, but, as an
aside, I admit that I wouldn't be shocked if it were not a large number. If
one were to isolate the responsible nucleotide change(s), use them as
characters, and then also code "feather", then clearly there would be a
non-independence problem. But the thrust of your argument seems to be that,
so long as the number of nucleotide changes is greater than one, then the
use of a character "feather", weighted as 1, represents a problematical
under-weighting. I do not necessarily agree.
 I do not think that a weighting scheme should serve as a scorecard for the
number of process steps that one imagines has underlain a particular
transformation. We all admit that we dont have a clue as to what went into
the tranformation to feather. It could have been one nucleotide change, or
100 non-independent changes that "should be weighted" as one, or 100
independent changes. We cant code what we dont know. What we do know, is
that "feather" is an identifiable character-state difference relative to the
condition found in other archosaurs. If that is all that we know, in this
comparative context, then that is all we can code. The matrix is simply a
compendium of the comparative hypotheses that we have about a group of
species, concerning their heritable traits. If further study of the genetics
and developmental biology of your critters leads you to redefine and expand
the number of your characters, that is fine. It will represent progress. I
dont have any reason to suspect though, that this will lead to wholesale
changes in our understanding of relationships.

>Even individual nucleotides are of differential importance and will
>someday be weighted routinely in some fashion.

I dont doubt that this will happen, but I dont think it will be useful. Once
again, the matrix is not a numerically coded textbook of developmental
biology, it is a compendium of hypotheses of homology - "these two states
are the same historically and thus are evidence of exclusive common
ancestry".

>...Granted there is great difficulty in determining how some
>characters may be related to one another, ... but we can't just throw up
>our hands and give up on differential weighting just because it is a very
>difficult task.

Once again, the issue for a phylogenetic analysis is not how characters are
related to each other except to the extent that one determines whether they
are independent of each other or not. A hypothesis of homology based on a
character "feather" is as much a unit of evidence for exclusive common
ancestry as a hypothesis of homology based on the 23rd nucleotide in a
particular gene, so long as they are independent. It is at this level, as
equally valid hypotheses of homology, that the characters must be treated
equally in a matrix.
Whether my perspective is or is not correct, it is based on methodological
principles, and not on a "throwing up of my hands" in the face of a
difficult task.

-tom