"Parsimony" program for clade studies

[Doug Yanega]

Norbert Kondla wrote:

>i am surprised that this note did not generate some debate on the discussion
>group.

I think perhaps the problem was that Jim was asking questions from a
position of (forgivable) near-total ignorance as to what systematics is
about, so even a friendly attempt to correct his misconceptions could take
a few pages worth of text. It's tough to bring someone completely up to
speed on what parsimony and cladistics are in a paragraph or two. I'll make
a few comments below.

> others swear on the value of genitalia and the list of
>potential characters is pretty long. an interesting observation on genitalia
>is that this character suite seems to be based on the very old lock and key
>hypothesis which i do not think has ever been proven.

Try reading Bill Eberhard's book on this topic. It's pretty well proven
that - at the very least - species-specific recognition cues are important
in the evolution of animal genitalia, and even if it's not a matter of
"fit" in the physical sense, it *is* a matter of fit in the *sensory* sense
(if that makes sense). ;-)

>> I read the other day about a program to determine the degree to which
>> species, genera, etc., are related to each other, and the tool in use was
>> a
>> program using the principle of parsimony.

Cladistics does not determine "degree of relatedness" in the sense you
imply - it is used to form hypotheses about the *pattern* of relatedness,
which is a different thing from "distance" between taxa. Parsimony is a
process by which you take multiple independent hypotheses of character
homology (basically, "whenever character X looks the same in two or more
groups it is because they share common ancestry") and analyze them together
simultaneously so as to group the taxa in a way that leads to the rejection
of the smallest total number of these hypotheses.

> The name suggests a comparison
>> of
>> some sort of characteristics of the critters under study with those
>> sharing
>> the most being the most closely related.

Nope. It does not *necessarily* matter how *many* characters are shared,
but rather what the patterns of shared characters may be. You can have two
taxa that share many features but which are more closely related to *other*
taxa with which they share fewer features. Some features are more
informative than others.

> Anyone have an opinion on the
>> veracity
>> of the results compared to, say, a DNA analysis?

You can analyze DNA data using the exact same procedures, and many people
do, but there are problems with doing so, mostly stemming from a
fundamental difference in how one makes hypotheses of character homology
using DNA sequences. In general, then, a parsimony analysis of DNA data is
likely to be less trustworthy than one of morphological data.
        Many folks who do DNA work *only* use a different technique, called
"maximum likelihood", to work out phylogenies, but this technique requires
that you start with a model of genetic evolution FIRST and then the results
of your analysis are dependent on how good your model is. You can't run
this sort of analysis with morphological features because there are no
expectations as to how morphology evolves. This is a main reason why
there's so much debate and argument among systematists these days. People
are using two very different techniques to do the same thing, and there's
no way to know who's right.

>How would one go about
>> choosing the most meaningful characteristics for comparison? How would one
>> weigh the characteristics so chosen?

This is where the human factor comes in. Total objectivity would dictate
that we use every characteristic we can measure, and that we perform no
weighting. The other extreme is that we use only those characters that we
feel subjectively are important, and that we weigh them according to our
opinions. Virtually everyone operates somewhere between these two extremes
these days. A good taxonomist can make the difference between a good
analysis and a bad one because of the ability to reject characters at the
outset that will create problems (like using color patterns of wings in
Heliconiine butterflies), and determine some cases of false homologies
(convergence) *before* running an analysis. To some degree, though, it's
all trial and error.

Does that help?

Peace,


Doug Yanega       Dept. of Entomology           Entomology Research Museum
Univ. of California - Riverside, Riverside, CA 92521
phone: (909) 787-4315
                  http://www.icb.ufmg.br/~dyanega/
  "There are some enterprises in which a careful disorderliness
        is the true method" - Herman Melville, Moby Dick, Chap. 82