a phenetic-cladistics saga (abridged) - was paraplaying (and bullies)

[pierre deleporte]

At 18:04 07/04/2004 -0400, Frederick Schueler wrote:
> >[...] a classification in which each taxon is
> >tagged, not with a category, but with a date, and a differentiation
> >measure. Estimating the date is a straight-forward hypothesis, and it's
> >often done. I don't know what the differentiation measure might be -
> >something related to the old phenetic goal of 'overall similarity,' and
> >maybe derived from a discriminant function between sister groups.

It's classically done under the "molecular clock" model, with fossils for 
minimal datation of the apparition of some morphological synapomorphy (= 
"calibration") and "phenetic-like" (= overall similarity-based) molecular 
distance as a measure.
But it does not correspond at all with the old phenetic goal of "overall 
similarity per se", it is evolutionary biology overtly (while early 
phenetics were agnostic): overall molecular similarity makes evolutionary 
sense under the character evolution model of regular divergence. Modern 
approaches cautiously perform prior evaluation of the plausibility of such 
a model for explaining the data matrix (distribution of distances) before 
attempting a molecular clock datation.

>I think it's too bad phenetics didn't attach itself to cladistics at an
>early era, but by the time it became apparent that cladistics was going
>to succeed, perhaps the two schools had called each other so many hard
>names that a reconciliation wasn't feasible.

I don't believe too much in personalizing the evolution of scientific 
thought. Bullying has strictly no chance against logics in the long run (I 
hope so), we certainly have better uses for adrenalin, and possible 
"reconciliation" should strictly be a matter of logics, not personal 
feelings. Science is neither a democraty, nor a family circle.

Possibly "attaching" phenetics to cladistics would have required an early 
and sound understanding of both approaches, their variants, and their 
possible connections or lack of. But clarification emerged painstakingly 
(and still does), except perhaps in a few - cryptic - genial brains... (not 
mine for sure).

I suggest that you distinguish a formal tool (measure of overal similarity, 
let's say a given computer program if you like) and the possible logics for 
using it in different contexts.

Early phenetics was not evolutionary. The overt logic was to class through 
"overall similarity", and quite arbitrarily so; the explicit goal was 
stabilizing the classification. It failed (more characters, more taxa, and 
different arbitrary distance measures... produced lasting instability). It 
also failed (imo) because nobody really needed this sort of classification 
(it had explicitly no biological meaning, by construction). In current 
practice it survives somehow, and rather informally, as a possible first 
attempt to class "difficult" groups when no phylogenetic signal obviously 
appears according to known characters.
Molecular clock approach is not a resurrection of the phenetic logics, it's 
a new use for an old tool, in a completely renewed logic (evolutionary 
biology).

Beyond early formal mixing with phenetics in "numerical taxonomy" (see 
Richard Jensen's last post), "Cladistics" splitted in two:
- agnostic cladistics (= strict "pattern" cladistics). Some proponents 
inadvertently rediscovered phenetics (= approx. Jaccard index, i.e. 
classing on overalll similarity for presence, not for absence, of character 
states considered separately). The common point with phenetic logic is 
biological agnosticism. Just class for classification sake... Frequently 
rooted in the myth of "The" unique and obvious pattern of subordination of 
characters, which you would just have to check out there (our basic 
disagreement, Frederick).
- evolutionary cladistics (= maximum parsimony phylogeny inference 
methods). Still frequently "brother-and-enemy" to maximum likelihood 
approaches in the mind of its proponents. Not for long I guess. Dig out the 
implicit model, and choose your optimization algorithm, cladistic or 
likelihood, and you'll get the same optimal topology anyway... this is 
matter for thought indeed... Evolutionary cladistics and ML are a 
complement to tentative molecular clock datations (see above). Conditions 
for approximate matching between cladogram and phenogram have been 
explained in other posts. The common point between modern evolutionary 
cladistics, likelihood and molecular clocks is explicit evolutionary 
explanation.

Look Frederick: arn't many "junctions" or "attachments" already done some way?
(And I really don't mean to be politically correct   ;-).

I talked systematics here (I mean organizing knowledge a consistent way). 
Of course, as for formal classification, you can use one or a several 
juxtaposed criteria (multi-criteria data bank and retrieving system). It is 
the attempt to "optimally combine" different criteria in a unique hierarchy 
which is open to arbitrary decisions, hence lasting quarrels of personal 
opinion (see Ken Kinman's lasting quest of an optimal clade-grade 
combination; but he also implicitly hints at a multi-criterion data bank, 
which does not impose artistic combination of criteria, just neutral 
juxtaposition of them).

>imho,
>fred.

in mine,
thanks for stimulating questions,
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