cladistic

Joe Felsenstein joe at GENETICS.WASHINGTON.EDU
Tue Apr 20 15:48:38 EST 1993


Doug Eernisse wrote (in part):

> I agree with Joe's characterization of cladistic classification
> but in terms of inferring phylogenies, cladistic methods have
> generally been associated with the parsimony method and, while
...
> One of Hennig's primary contributions to systematics was his
> distinction between types of similarity. He advocated that it is
> inappropriate to use symplesiomorphic (shared primitive)
> similarities as supporting evidence for the monophyly (or
> holophyly) of a clade. Only synapomorphic (shared evolutionary
> novelties relative to one or more outgroups) features should
> be used. In this sense, distance methods are not cladistic
> because they confound plesiomorphic and apomorphic similarities.

Doug and I differ on this.  He quotes Hennig as saying that only
shared evolutionary novelties should be used as evidence for
monophyly.  He doesn't quite say that this is his view.  If he thinks
that parsimony should be called "cladistic" because it is the kind of
method Hennig favored, that is a weak way of defining "cladistic"
(Hennig also favored centers-of-origin biogeography but no one has
advocated calling that "cladistic").

I suspect Doug does agree with Hennig on inappropriateness of using
shared primitive states as evidence, and wants to call parsimony
"cladistic" because it adheres to this principle.  Surely distance
matrix methods do use both shared derived and shared ancestral states
without distinction in their calculations.  Under the statistical model
that they assume, they should do so: the fraction of shared ancestral
states is evidence and should be used appropriately.  If one agrees with
the use of such models (and I do) then there is no big distinction between
distance matrix and parsimony (and other kinds of) methods: they all make
use of some of the information present, in various different ways.  They
all can be valid methods from a statistical point of view.

It is very much like making a big distinction between medians and means
as measures of the center of a normal distribution, and calling one
"cladistic" and the other "phenetic" and arguing that the latter confounds
information with noise and this should never be used.



More information about the Mol-evol mailing list