More on the 'cladistics' of sequences

[Jim Whitfield]

It seems to me that in these cladistics vs. phenetics arguments about
molecular data, two related but distinct issues are being confused
some of the time.

The first is whether the data being employed are actual character
data (discrete or continuous) or are converted to distances in some
way.

The second is whether taxa are being clustered into a classification
only by shared similarity (=the phenetic philosophy) or not

For instance, neighbor joining typically converts DNA sequences to
corrected distances but is NOT fundamentally a phenetic method (it is
also not strictly cladistic).  It optimizes the distances to a tree,
and the tree IS intended to represent plausible phylogeny rather than
a classification by overall similarity.  Many people seem to
erroneously equate distance analysis with phenetics.  There ARE
reasons to prefer direct character data over distances for estimating
phylogeny (a lot of information is lost in the conversion to
distance), but distances have continued to be useful since they can
be "corrected" for saturation and other biases, and thereby provide
better estimates of divergence times.  Probably these are now much
better dealt with in a likelihood/Bayesian context, where the
hierarchical character data relationships are more directly preserved
than in distances.


--
James B. Whitfield
Associate Professor
Entomology, 320 Morrill Hall
505 S. Goodwin Ave
University of Illinois
Urbana, IL 61801
tel. (217)333-2567 (office)
(217) 265-8123 (lab)
FAX (217) 244-3499
email jwhitfie at life.uiuc.edu
lab web pages:  http://www.life.uiuc.edu/whitfield/