Clades are classes

[pierre deleporte]

A 12:06 09/09/2002 +0200, vous avez écrit :
>Pierre, I do not try to offend you; I would only like to separate things 
>that have not much in common:

No problem, I never felt that you tried anything of this kind.

>-----Original Message-----
> >>I do not believe that you can place a new property (that was not
> >>observable before) to the set of taxa *after* the analysis to make it 
> class.
>
> >I don't believe that you can forbid anybody to base a classification of
> >living beings on the basis of the results of a previous phylogenetic
> >analysis. The process of phylogeny reconstruction should simply not be
> >confused with the process of classifying on that basis.
>
>Of course - so why are we discussing the class concept? Your taxa are 
>perfect clades (even monophyletic if you want); why to try to interpret 
>them within the class theory? To create distance-based taxa is nothing bad 
>nor unscientific.

Agreed : there is no philosophical problem involved there for me. We can 
"classify" a lot of ways. The important point for me is the meaning of our 
classification(s) (and good communication of this meaning). The "class 
theory" question is (for me) a side problem.
I simply understand that clades are classes, according to the class theory, 
once one recodes the initial data matrix into a completely consistent set 
of nested synapomorphies, which is what we do (at least implicitly) when we 
accept a cladogram as a reliable representation of the evolutionary history 
of clades (phylogeny).
A snake is a tetrapod. Not in the data matrix (it has no legs), but as 
interpreted once the phylogeny is accepted (belongin to the clade 
"tetrapods", it has "lost legs", which is historically different from 
having no legs as in outgroups and remote ancestors).

Of course clades are not classes if you stick to the initial data matrix. 
But cladists call it a matrix of putative synapomorphies, different from 
the a posteriori interpreted matrix of confirmed (or post-hoc interpreted) 
synapomorphies. "Lost legs" is a confirmed synapomorphy among tetrapods. 
The sequence of changes in a character in the history of a group is a 
relevant hierarchy of character states (including possible "misses").

> >I fear you have a rather naive notion of "observable" properties in biology
> >(and in natural sciences at large). There is no
> >"naturally-obvious-self-evident" properties....
>
>Indeed, but it does not mean that anything goes.

OK, but not anything goes in a scientific phylogenetic analysis. Concepts, 
methods, and even relevant"data" have to be argued.

>  What seems to be suspicious to me is to rearrange data matrix by the 
> results of analysis done on just this data matrix.

Data matrix plus an explanatory theory : you can't ignore that ! So there 
is no circling round, the explanatory theory is added to the data matrix in 
order to provide a new interpretation including the best fitting nested set 
of synapomorphies.

>  It is not interpretation, IMO, but illegitimate manipulation with the data.

In no way (no offense at all, your statement is simply wrong). Unless all 
scientific interpretation is "manipulation of the data", e.g. Earth 
revolving around the sun is "manipulation of the data" (it is not "the bare 
facts", you know...).

>  Imagine simply what are you doing:

I don't imagine, I just do it

>  cladogram is indicating that character 1 (say flower color) has changed 
> first from -1 to +1 (say, violet-magenta). Then (in taxon A) it changed 
> again (magenta-violet). Now, you are saying that the A has preserved the 
> magenta colour in some respect.

OK, I don't mean that it has preserved it, I mean that its ancestor likely 
had it, and thus that the magenta is the marker of the whole clade, even if 
modified in violet in some descendants. I pretend to have deciphered this 
history of the character "colour", supported by the congruence of all 
relevant characters supporting this phylogenetic reconstruction and thus 
this likely evolutionary scenario against other possible ones.

>  I wonder why? It does not make any sense, even within the phylogenetic 
> methodology - the character as such ("flower colour") is something that 
> is homologous through the taxa but the individual character states 
> (violet/magenta) are distributed in distinct regions of the 
> cladogram/groups of taxa. If the ch.state is transformed, its 
> plesiomorphic state is simply missing - this is the base for all 
> cladogram construction and I see no reason why to "recode" characters 
> after analysis.

The recoding aknowledges the acceptation of this complex scenari: there are 
two different "magenta", one is ancient, the other is "historically 
transformed from violet".

>  Even your argument that "-1 is a historical avatar of +1" does not 
> change anything - be it "avatar" or not, the +1 is still missing. This 
> argument seems to me to result from a poor distinguishing between the 
> *character* and *character state* (quite frequent mistake in such discussions).

You are right to insist on the distinction betwen character and character 
state.
Now, when you say the character state is "missing", I say it is 
"transformed", that is to say: present under another (and sometimes 
misleading) form. Present under another form is not missing.
Maybe the discussion of the case of derived character missing (not 
character state missing) is more relevant. Snakes are tetrapods, they have 
four (missing) legs. Earthworm don't have four (missing) legs. Phylogeny 
reconstruction tells us this, and this may be implemented in a phylogenetic 
classification of life, possibly with inferred nested synapomorphies 
defining classes of equivalence.

> >Not at all if you aknowledge the interpretative nature of "scientific
> >observations".
>
>Huh?? The (+1) is a perfect homology common to the (paraphyletic) group 
>BDE; what is bad with this interpretation?

Nothing, just that this is a symplesiomorphy of BDE relatively to the 
character state "reversed absence" in A, and thus a misleading similarity 
for a phylogentic classification ,unless you aknowledge the "presence of 
reversed loss of 1" in A.
If you have already accepted the phylogeny, of course, unless you will 
effectively stick to the basic data matrix and ignore the historical 
phylogenetic interpretation. It is just as you please.

>  On the other side, the reversal (-1) can be viewed as autapomorphy of A. 
> If there would be further taxa-splitting A1+A2 it will be even new 
> synapomorphy. You can imagine it as a transformation series (-1) - (+1) - 
> (++1), say "flowers violet - flowers magenta - flowers white". In such a 
> case and WITHIN THE CLASS THEORY the BDE is indeed a class - it has a 
> unique (and even homologous) character. That this is unacceptable from 
> the CLADISTIC point of view is completely another story.

Well, a given symplesiomorphy can be used to support a class. I accept this 
perfectly well. The problem will arise from compatibility with other 
characters if you take them all as "face value similarity".
On the other hand, the completely compatible and nested set of 
synapomorphies will define classes of equivalence a consistent way 
throughout the group at stake.
You say we don't "observe" them. I just say we can infer them from 
phylogenetic analysis.

>However, we were not discussing the phylogeny reconstruction but classes - 
>I only pointed that "to be a class" will not distinguish between cladistic 
>and phenetic (and eclectic) taxa.

OK. So my point is simply that synapomorphies can define "real" classes of 
equivalence, provided that homoplasies and reversals are inferred from 
phylogenetic analysis and coded as a nested set of synapomorphies, 
including "historical modification to absence".
Snakes are tetrapods = four legged vertebrates, even if you don't see the 
legs. They are not tetrapods before phylogenetic analysis, but after. If 
you stick to the data matrix, which is definitely not a phylogeny, this 
will not make sense of course. And the absence of legs in snakes will still 
make a similarity with earthworms, which idea the "cladists" dislike (those 
so strange people... ;-)     ).

I agree with you that the rough data matrix does not provide a nested set 
of synapomorphies supporting a true classification (classes of equivalence).
I argue that the post hoc reinterpretation of the data set as a nested set 
of synapomorphies provides such a basis for classification.
I disagree with you that this does not "make sense". It makes phylognetic 
sense, which is all right for a phylogenetic classification.
I agree with you that it would be no shame if this or that way of grouping 
taxa would not be a "true classification". But it merely happens that 
synapomorphies, by recoding convention, do support classes (which is no 
shame either).

But maybe we will agree on this:

"This is not to deny the legitimacy of different classifications of 
organisms. After all, classifications are conventions, hence neither true 
nor false. It is perfectly legitimate (for a home-maker) to classify 
organisms into, say, "Edibilia" and "Inedibilia"."
(Mahner and Bunge 1997, p. 252).

best,
Pierre