3TA (was Human and ape phylogeny)

[pierre deleporte]

Hi René, David, and taxacomers

I think this thread has no more connexion with the Human-apes thread, hence 
my re-naming.

I will not comment all the points made by René (although I disagree with 
all of them in the slightest details). I just try to point at the core of 
the question on two points: semantics, and models.

- semantics :

At 11:56 05/04/2003 +0200, René Zaragüeta Bagils wrote :
>(...)
>it is logically inconsistent to optimize "reversals" because they cannot 
>be "apomorphic", by definition (except in phenetics...)
>(...)
>Pierre Deleporte wrote:
>"But if you really like phenetics, use three-taxon analysis in its last 
>versions [...] This makes the method non-evolutionary, i.e. fitting no 
>conceivable evolutionary process, and thus non-phylogenetic at all."
>Re:
>What do you mean by "last versions"? I do not agree with "modified 
>three-taxon analysis" because it is as phenetic as standard parsimony (it 
>optimizes absences). It seems to me that it is you who like phenetics 
>(e.g. when you say that "apomorphic reversal to 'loss of tail' can support 
>a clade")

Yes I mean "modified 3TA", which has been demonstrated as being pure 
phenetics (i.e. clustering on the basis of overall similarity, and not 
optimizing homology by descent).

A basic misconception in the three-taxon-analysis school is to consider 
that "phenetics" means: "grouping (sometimes) on the basis of the absence 
of a feature". This is ignoring the current definition of phenetics (= 
grouping on the basis of overall similarity, as repeatedly stated on this 
list), and also ignoring that phenetic methods can implement an illimited 
variety of similarity or distance measures, including methods which group 
on the basis of the overall co-occurence of presence of features, to the 
exclusion of the co-occurrence of absence. You then get a phenetic 
clustering on the basis of overal similarity of presence of features only 
(would then 3TA proponents call it "non-phenetic"?).
This very peculiar acception of "phenetics" in 3TA jargon makes their 
comments on the topic very difficult to understand for the naive outsider. 
A complete dialogue of the deaf in fact.

René also wrote:
>Optimizing instances and non-instances of characters at the same time is
>simply phenetic (just so).

No, phenetics is not better defined as "just so". Sorry if my last post 
could suggest this. Applied to molecular data under the character evolution 
model of "molecular clock", a phenetic clustering is clearly phylogenetic, 
the evolutionary model justifying the approach is explicit (whether the 
model is biologically realistic in this or that case is debatable of 
course, and molecularists have developed methods to test for this in the 
structure of the data matrix).

But applied to morphological data, phenetics is "just so" (because there is 
no model of "morphological clock" justifying the use of the method for 
phylogeny inference on such data). Notably, the phenetic school of 
systematics in the seventies presented itself as biologically "just so" 
(they did not pretend to infer phylogenies). Its explicit goal was stable 
classification/nomenclature, but not biologically meaningful systematics.

See also recent posts, like David Orlovich, or B.J. Tindall :
" It is difficult to see how a data set can be phenetic. It may be handled 
in a phenetic way (overall similarity)."
and Peter Hovenkamp for his comment on the notion of phylogenetic 
(transforming) characters.


- evolutionary models

René wrote:
>However, how can absences evolve (=transform, I guess you should say) in
>your model? What do you mean by the evolution of the "lack of a tail"? How 
>can, for instance, natural selection work on the absence of something?

Very simply: evolution acts on absence by selecting against corresponding 
presence. Selection acts in selecting in a population between the 
descendants of individuals with no tail (ancestral state) and of mutants 
with a tail (new feature), on the basis of differential fitness 
(differential number of reproductive decendants from the two types of 
individuals with/without a tail according to their differential adaptation 
to the environment), or possibly on the basis of mere genetic drift (= 
chance alone) in small populations.
"Evolutionary transformation" is not magics.  Biology is meaningful. 
Considering population biology, population genetics, genetics of the 
development, can give useful insights into how evolution can proceed. 
Artificial selection too (Darwin's pigeons to start with). Provided people 
take an interest in such matters, of course.

>How can you distinguish between the 'transformation' of the absence of a 
>tail to the presence of a tail and the 'transformation' of the absence of 
>a leg to the presence of a
>tail?

Humm... I'm no specialist, but in my humble opinion it seems that they do 
not fit exactly in the same place on otherwise similar squeletons. This 
remark is as old as the "principe des connections" of Geoffroy 
Saint-Hilaire, and basic to all homology assessment in morphology, with an 
obvious equivalent in molecular studies (sequence alignment and secondary 
structure). Centuries of comparative anatomy make sense, and modern 
molecular biology too. Provided people take an interest in such matters, of 
course.

OK you question my model, fine, but may I question yours ? I fear not.

In my view, the basic problem with 3TA is the absence of an explicit model 
of character evolution. All biological evolutionary consideration is 
absent, as if "replaced" by general notions of "hierarchy" and 
"relationship", which are certainly not biological in themselves.

Thanks also to David Williams for his comments, which help illustrating the 
point.
David wrote:

"One wonders quite what Deleporte imagines are "conceivable evolutionary 
processes", one wonders quite where Deleporte's special knowledge comes 
from, quite how he knows what is and is not a "conceivable evolutionary 
process"."

OK David, so much for me, I certainly made an infortunate tactical error in 
stating "no conceivable", I should have written instead "no explicit 
evolutionary model in 3TA" .

So, once a proponent of 3TA, just one, anyone (but at least one !), will 
have explicitly stated what model of character evolution he is 
implementing, the question of this model being "conceivable" or not for a 
contemporaneous biologist will only be raised. I infortunately jumped over 
this obligatory stage of the discussion, and anticipated the conclusions 
according to my own inability, despite my sincere efforts, of conceiving a 
biological model of character evolution fitting 3TA procedures. Sorry David.

The problem is that this discussion is impossible at the present time, 
because not a single proponent of 3TA, and symptomatically neither David 
nor René in their last posts on this list, ever stated what notion of 
biological evolution, of character transformation, was implemented in this 
approach.
Norman Platnick was the only one I know who tempted something this way 
(trying to justify 3TA as an improved implementation of evolutionary 
homology optimization, and whose attempts were rebutted in the nineties, by 
myself among others). Platnick's papers were notably separated from 
Nelson's ones at this time of the debate, Nelson never arguing optimization 
of evolutionary character homology (but only "more precise use of 
parsimony" and "relationships" instead, which is not biological in itself).

So, I can just reiterate the question: dear David or René, what conception 
of the evolutionary process, or what biological conception of the 
similarities and differences of characters in taxa are you implementing, 
when you perform 3TA ? Or more simply (like I already asked David five 
years ago): "Why do you do that rather than something else", as 
phylogeneticians ?

Note that I consider that you are, as I do, trying to infer a phylogeny 
from a data matrix of terminal taxa and characters. If not, forget all what 
I said, my comments do not concern non-phylogenetic approaches.

But please, don't just answer "relationship", or "hierarchy", not even 
"precise statement of relationships" or "cladistic idea of relationships" 
(like in last post by David),  I think I really know this by heart since 
the time, and I'd rather like to read something, anything biologically 
meaningful instead.
Similar or different traits are present in different taxa: how do you 
figure out this may the h...ll come from, biologically speaking ? People 
who have no answer to this question cannot logically infer a phylogeny from 
a data matrix. Inferring the history of evolution (phylogeny) requires the 
explicit implementation of some notion of evolution as a process. If not, 
then "anything goes"... not for me, thanks.

best and cheers (David, I pay the beers !   :-)
Pierre


Pierre Deleporte
CNRS UMR 6552 - Station Biologique de Paimpont
F-35380 Paimpont   FRANCE
Téléphone : 02 99 61 81 66
Télécopie : 02 99 61 81 88