Darwin-L Message Log 3:73 (November 1993)

Academic Discussion on the History and Theory of the Historical Sciences

This is one message from the Archives of Darwin-L (1993–1997), a professional discussion group on the history and theory of the historical sciences.

Note: Additional publications on evolution and the historical sciences by the Darwin-L list owner are available on SSRN.

<3:73>From mayerg@cs.uwp.edu  Tue Nov 16 14:39:55 1993

Date: Tue, 16 Nov 1993 13:49:34 -0600 (CST)
From: Gregory Mayer <mayerg@cs.uwp.edu>
Subject: Re: phenetics vs cladistics vs evol. class.
To: darwin-l@ukanaix.cc.ukans.edu

	Bob O'Hara's first installment of his reply to Lynn Hanninen's
query is a clear exposition of many of the issues involved, and one with
with which I find myself in almost total agreement, especially on the
importance of distinguishing phylogenetic inference (=historical
reconstruction) from classification (=a set of names and their
definitions).  It is with some trepidation, therefore, that I venture
to suggest an amendment to his characterization of phenetics as a method
of phylogenetic inference.  (In fact, it might be better for those
encountering these issues for the first time to read O'Hara's full
exposition first before returning to my suggestion here.)
	My suggestion is that phenetics is best characterized as _not_
being a method of phylogenetic inference at all.  It is quite right to say
that phenetics is concerned with overall similarity.  But it is not
concerned with phylogenetic inference.  Let me illustrate this with two
lines from Sneath & Sokal's _Numerical Taxonomy_ (Freeman, San Francisco,
1973; the classic text of phenetics):
	Taxonomic relationships are evaluated purely on the basis of	resemblances existing _now_ in the material at hand. [p. 9,	emphasis in original]
	The separation of overall similarity (phenetics) from evolutionary	branching sequences (cladistics) is an important advance in	taxonomic thinking. [p. 10]
	The distinction between phenetics and phylogenetics is clearly
acknowledged, and phylogenetic intent is disavowed.  Part of my
trepidation in raising this issue is that phenetics has a complex history,
and its aims have shifted over time; I do not wish to rehash these
history-of-the-discipline matters.  And if this were my only quibble with
O'Hara, i.e. as to what it was the pheneticists actually wanted to do, I
would not bother to mention it.
	What I think _is_ worth mentioning is that phylogenetics is an
intention, not a technique.  What distinguishes a phylogenetic method from
a phenetic method is that a phylogenetic method produces an estimate of
history.  As O'Hara notes, history happened only one way, and there is
thus a "parametric" history to be inferred.  Some of our estimates of
this history will be good, and some will be bad, and one reason that an
estimate could be bad is that the method of inference is not very good.  I
think this is what O'Hara meant when he contrasted overall similarity with
an analysis based on the distinction between derived and ancestral
character states: he meant that, in general, overall similarity is not as
good a method of inference.  Methods of historical inference are justified
on the basis of their ability to reconstruct history.  Depending on the
historical process, and the data available, different methods may be best.
Given a particular historical process, someone might argue that, say,
single-linkage clustering of Jaccard coefficients of similarity gives the
best estimate of the history.  That single-linkage clustering and Jaccard
coefficients are used by pheneticists does not make the method phenetic;
it is phylogenetic because it is a method of inferring history.  Now it
might be a bad method, but that depends on the nature of the historical
	A statistical analogy may make the point clearer.  One way to
estimate the mean of a distribution is to rank all the observations and
choose the value of the central observation as the estimate of the mean.
This is a method of inference.  If we have good reason for believing that
the distribution is symmetrical, then this is not a bad method (not the
best, but not bad).  If the distribution is not symmetrical, then this is
a bad method.  The method is better or worse depending on the nature of
the distribution.
	Let me conclude with an example where many biologists have argued
that overall similarity _is_ a good estimate of phylogeny.  Some molecular
evolutionists have concluded that, at the molecular level, evolution
proceeds at a constant rate at a given locus.  If this is true, and if
evolution is divergent, then the overall similarity at a locus _is_
proportional to the time since common ancestry, and an average-linkage
tree of the overall similarities will be a good estimate of the phylogeny.
The premises of the justification may be false, in which case the
technique will be bad as a method of historical inference.  Studies have
been done of exactly how untrue the premises must be before the method
fails.  But if the premises are true, then it will be a good method.
	Thus I endorse the clear distinction of phylogenetic inference and
classification.  But phenetics and evolutionary classification are both
about classification.  Within phylogenetics, many techniques are or have
been advocated; what makes them phylogenetic is that their proponents
intend them to be estimates of phylogenetic history.  We may judge a
technique as more or less successful, but they are phylogenetic nonetheless.

Gregory C. Mayer

Your Amazon purchases help support this website. Thank you!

© RJO 1995–2022