Darwin (was: Phylogenetic evidence)

[pierre deleporte]

A 12:28 26/07/2002 +0200, Zdenek Skala wrote :
>Perhaps I was not clear enough, so:
>
>Pierre:
> >In A, it takes the form of a derived "0", which is a plain historical
> avatar of "1".
> >The class DCBA is supported by the shared equivalent trait "1" = "derived
> >0" in A.
>I agree completely with the only reservation that class theory knows
>nothing about history but only on the objects and characters, so the
>argument that 'historically 0=1' is invalid in this context (and hence:
>the clades are not classes).

This is the point. My class theory (Mahner & Bunge 1997) is apparently
different from yours. My class theory accepts any kind of "equivalent
properties of individals" for classifying into classes of equivalence (by
difference with similarity clustering). A "character of an object" may thus
be historically defined (why not ?).

Formally, do you suggest that this would imply a complete recoding of the
data matrix, suppressing all ambiguities (homoplasies) by transforming them
in different apomorphies ?
Well, I think it's exactly so. And would we not have classes, based on this
recoded data matrix ?
No problem for me, because this operation has not the ennoying implication
you suggested (that there would be "no more homoplasy").
We should never calculate homoplasy ratio on this ad-hoc-transformed data
matrix (will effectively be zero homoplasy all the times), but on the
initial one. Of course.

If it seems shocking to "recode" incompatible data this way, we should keep
in mind that this is the mere logical consequence of accepting the
cladogram as "the" phylogeny (= the provisionally least dissatisfying
phylogeny !).

Refusing to "resolve" homoplasies into independent apomorphies means
refusing the cladogram as a reliable phylogeny. You believe the cladogram
or not. No other way.

>  (...) Homoplasy only can be defined as a consequence of character state
> equivalence appearing in distant lines  - e.g. when -1 is always "the
> same" though appearing several times (this is the meaning of the words
> "character conflict"). Without character conflicts you will always get a
> fully resolved cladogram.

You mean character state equivalence in the data matrix, and for
calculating homoplasy ratio... But charater state equivalence after
accepting the cladogram = phylogeny is different, because conflicts are
resolved. "Distant lines similarities" are reinterpreted in the form of
local, distinct homologies by descent, and this is the clear basis for
clades = synapomorphy-based classes of equivalence in phytogenetic systematics.

>2. Or one accepts character-state equivalence (i.e. "-1" is always the
>same) resulting in the homoplastic character states / character conflicts.
>This, however, in turn means that these apomorphies/homoplasies are also
>equivalent to the plesiomorphic state and then ... (clades are not classes).

Such groups are effectively not classes, but they are not the clades of
cladists (see above). In this case, you simply don't believe that the
cladogram is a phylogeny, or you reject the criterion of historical
continuity by descent (= contiguity on the cladogram) as the definition of
homology, i.e. you reject the cladistic notion of homology (but then, don't
be surprised not to get classes).

Accepting a phylogeny for classification requires accepting the resolution
of conflicts according to this phylogeny. Choosing the best fitting
cladogram means accepting the best
character-conflict-resolving-phylogeny (at the lower cost in terms of
remaining homoplasy = ad hoc resolution of conflict).

So: I accept to recode my data matrix according to ad hoc "homoplastic
character conflict" resolution, with reference to the accepted most
parsimonious cladogram = more likely phylogeny, and I base my classes of
equivalence of taxa on this fully consistent nested set of reinterpreted
characters(-states).
This is the basis for phylogenetic classification, not the basis for
initial phylogeny inference (which has some residual homoplasy ratio, which
I never loose from sight).
Agreement yet ?

Best,
Pierre