Striking a balance, weighting and Cladistics
Thomas DiBenedetto
TDibenedetto at DCCMC.ORG
Thu Feb 15 11:02:34 CST 2001
-----Original Message-----
From: Thomas Lammers [mailto:lammers at VAXA.CIS.UWOSH.EDU]
10:40 PM 2/14/01 -0800, you wrote:
[with re.]
>>...why is there a desire to weight characters a
>>priori, unless to influence the outcome of an analysis away from results
>>driven by parsimony?
>Well, I suppose this IS the reason. But it is motivated by the knowledge
>that there is absolutely no reason to assume that evolution has been
>parsimonious.
If this is in fact the sole source of the motivation to weight, then I think
it is an unfortunate error in reasoning. Because in fact, the knowledge that
there is no reason to assume that evolution has been parsimonious, is a
knowledge shared equally by those who are strongly against weighting. Equal
weighting or non-weighting parsimony has never been based on an assumption
that evolution has been parsimonious. Parsimony is a criterion applied to
evidence, not an assumption about the underlying reality. It is an
epistemological tool, not an ontological assumption (if you want to get
fancy about it). Everyone understands and accepts that similar characers can
arise through convergent evolution, and that character states can be lost in
descendants. Were this not so, or were this to be merely a trivial aspect of
evolution, then phylogenetic systematics would be an unproblematical
enterprise.
It may be of some benefit to consider what one means by a character being
more or less reliable. True homologies are true homologies; there is no such
thing as a fractional homology. The problem arises in our ability to
percieve when a similarity is homologous or not. All characters in a matrix
are the result of long and careful study (hopefully). The systematist should
not propose the character unless she is confident, to the best of her
analytical abilities, that it is in fact a homolgy. From this perspective,
the only meaning to a weighting scheme is that it represents a metric of
one's confidence in the character hypothesis.
According to the methodological principles accepted by most anti-weighting
cladists, one must test ones proto-character-hypothesis rigorously before
adding it to the matrix, till one arrives at a threshold of confidence that
it is a valid homology. This is the locus of the equivalency between
characters; that they all are accepted as equally likely to be homologies
according to the experience and judgement of the systematist. One often sees
characters weighted at this point, because of the perception that similar
characters in other taxa have been shown to be problematical in other
studies, that the character has been shown to arise several times
convergently in other groups, thus being "unreliable". But such a
probabilistic argument for unreliability assumes that there is some common
process that drives independent characters towards similar appearances, with
some predictable regularity. I dont see any reason to believe that.
Another phrase one often hears, in discussions of weighting, is the
differential "importance" of a character. This is something I do not
understand at all. Characters are important in phylogenetic systematics if
they are homologies. If they are not homologies, they are not important. If
a homology is a something "small" like a particular nucleotide or something
"big", like a feather, does not make any difference that I can see.
-tom
More information about the Taxacom
mailing list