Taxonomic hypothesis-testing

[James Lyons-Weiler]

On Sun, 31 Mar 1996, Joe Laferriere wrote:

[1/2 previous message snipped]
> > In this sense, taxonomy is a science:  in a well-done taxonomic study,
> > hypotheses are being put forth, criteria for rejection of those hypostheses
> > are expostulated, data are presented, and conclusions derived from the data
> > made on the basis of those criteria.
> >
> So, then, let me see if I get this straight. If a taxonomist is trying to
> decidee whether two species should be placed in the same genus or not,
> s/he sets up alternative hyoptheses:
>
> H1: The species are congeneric, vs.
>
> H0: The species are not congeneric.
>
> S/he then sets up an experiment to test whether s/he can demonstrate one
> hypothesis or the other with 95% confidence. Do I have this right? I have
> never seen any taxonomist do anything of the sort. Whether a classicalist
> examining dusty old herbarium specimens with a hand-lens, or a cladist
> producing computer-generated tree diagrams, it still always comes down to
> a question of personal judgment as whether to call a given monophyletic
> group one genus [familiy, order, etc.] or more than one. There is no
> universal consensus whatsoever on how to do this.
>
This is true.  However, the following questions can be asked on the basis
of whether a null can or cannot be rejected:

1) the data that are collected and placed in a character state matrix
appear to be informative with respect to phylogenetic history;

2) whether there is sufficient support to reject a star phylogeny (which
is a null phylogenetic hypothesis).

The taxonomy of organisms is a science in the same way that astronomy is a
science: both are highly empirical, and probability has little to do with
observations.  As a science, pure empiricism retards the process of
discovery, excepting, of course, serendipity.

Phylogenetics, on the other hand, is becoming more and more
statistical.  The artifical classes mentioned above may or may not be
reflected in a pattern of diversification, let alone monophyly;  evolution
has cared not a whit how humans will classify its products.

If there is a convention for erecting groups that is phylogenetically
based, I'd like to know more about it.  But I wonder if there should be.
There is a growing realization that most published phylogenies contain
grevious errors, and while the work is furiously being conducted on
providing measures of error, it is not yet clear how to measure support
for monophyletic groups (those thinking bootstrap hold back, it does not
measure support; however, see Zharkikh and Li, 1995. Mol. Phylo. Evol.
vol. 4).  Confidence in phylogenies is now based on how well different
methods of phylogeny estimation (no longer referred to as
'reconstruction') perform under bad conditions.  Hillis, Huelsenbeck,
Debry, Swofford, and others (e.g., Scho:niger, 1996) have, over the past
few years, conducted what they call "power analyses" using simulations
where rates vary (i.e., the Felsenstein zone).  Parsimony and
maxlikelihood models all have describable limitations (low efficiency;
i.e., they take more data to converge on the correct topology).  These
analyses are based on four taxon trees (obviously, extrapolation to larger
groups should be a hedged bet).

In one sense, the goal of natural classifications is to capture or reflect
geneaological information.  But I think the link between taxonomy and
[evolutionary] process is even weaker than the link between phylogeny and
process.  Both should be made stronger, and more statistical, IMHO.

Some cladists (NOT I) might assert that a classicalist using personal
judgement is doing cladistics after all, but s/he is simply weighting some
(or in some instances, one) character over others.  I think the original
point was that personal judgement applies in all of science, whether the
science is mostly empirical, or more probabilistic.  The trick is
to see a science as being continually ontogenetic; when concepts that we
find useful (personally and as a social unit) become so entrenched that we
start calling them "facts", or otherwise rely entirely on their truth
content for the outcome of our reasonings, we can't help but canalize
ourselves into comfort and habit.  When we stop learning, and start
imposing our comforts and habits on others, we set up road blocks to
scientific progress.  After all, what better can we do then to find the
situational specificities of our a priori assumptions, or paradigms?  When
we define such limits, we rightly deprive methodology, as well as
empirical experience, of undue authority.  So as a science proceeds, it
(inevitably?) changes, sometimes for the worse, sometimes for the better.

If a group erected as a monophyletic genus (note the combination of
phylogeny and taxonomy), the hypothesis of monophyly can now be critically
tested, but the question of sufficiency remains.  However, the issue of
what happens to the  as a result of the falsified (by whatever
convention) hypothesis of monophyly is, as correctly pointed out, a
matter of authority by convention.  This needs work if taxonomy is to
pull itself out of the empricist realm of science.  The question is
whether it needs to, and if so, how might this be accomplished?

JLW

_________________________________________________________________
JAMES LYONS-WEILER                PH:    (702) 359-6391
ECOLOGY, EVOLUTION AND            FAX:   (702) 784-4583
CONSERVATION BIOLOGY/186          eMAIL: weiler at grass.ers.unr.edu
1000 VALLEY ROAD
THE UNIVERSITY OF NEVADA, RENO
RENO, NEVADA 89512-0013
_________________________________________________________________