Human and ape phylogeny
pierre.deleporte at UNIV-RENNES1.FR
Fri Apr 4 18:58:10 CST 2003
A 17:39 04/04/2003 +0200, René Zaragüeta-Bagils wrote :
>I agree with John Grehan's arguments. But I should not focus only on this
>- Molecular characters are NEVER polarized (so they are phenetic, in
Phenetics can certainly not be defined as "no polarization".
And molecular characters are polarized by the outgroup criterion when
treated with parsimony.
>- The more general problem is that Farris' optimization (= cladistic
>computer programs) is flawed. Cladistics is supposed to be a way of testing
>a conjecture (initial hypothesis) of hierarchical relationships. However,
>Farris' optimization produces no hierarchies, but unrooted trees (something
>that some Americans call "networks"). There is only an isomorphism (~
>equivalence) between rooted trees and hierarchies. So an outgroup is needed,
>not because of cladistics, but because of Farris' implementation of
>cladistics used in computer programs.
>- Farris' optimization is flawed in (at least) another sense: if your
>hypothesis proposes a grouping of, say, animals having a tail, your computer
>program may tell you that there is a group characterized by... ABSENCE of
>tail. However, you never made this hypothesis. It is your computer program's
>There are other problems with Farris' optimization.
>A solution, nevertheless, exists: three-taxon analysis.
>Département Histoire de la Terre UMR 8569
>Equipe SICC (Systématique, Informatique, Cladistique et Chronologie)
>Muséum national d'Histoire naturelle
>8 rue Buffon
>75005 Paris - France
Farris optimization, or Wagner trees, or standard parsimony analysis, is
not "flawed", it just fits a particular model of evolution. You can allow
character reversals, and thus reversals can tend to "support" a clade, for
instance apomorphic reversal to "loss of tail" can support a clade. If this
agrees with your model of character evolution, it's OK. If not, use PAUP
"stepmatrix" tool and weight character step costs against reverals, and so
on for an illimited series of models with possible differential charcter
All depends on your model of character evolution.
But if you really like phenetics, use three-taxon analysis in its last
versions : its a true, and unexpected, rediscovery of phenetics, by people
initially dealing with "parsimony" in the "cladistic" school. 3TA
implements no implicit or explicit notion of character evolution. Even the
basic notion of homology by descent between similar character states is
broken down by the splitting of characters into separate "three taxon
statements" treated independently. Different states of a same character and
different occurences of a same character state are split into peaces with
This makes the method non-evolutionary, i.e. fitting no conceivable
evolutionary process, and thus non-phylogenetic at all.
Phenetics did not care with evolution either, it was "overall similarity
for overall similarity's sake", just so. Like 3TA is 3TA for 3TA's sake.
3TA has obviously no "flaw", and cannot have any flaw, merely because it
has no model to implement, and thus provides no handle to evaluate its
Reminds me of a least some version of "Panbiogeography"... :-)
CNRS UMR 6552 - Station Biologique de Paimpont
F-35380 Paimpont FRANCE
Téléphone : 02 99 61 81 66
Télécopie : 02 99 61 81 88
More information about the Taxacom