Farewell to: Farewell to Species - reticulation

[Richard Jensen]

On Thu, 3 Feb 2000, A Mitchell wrote:

> Well how many of us deal with "the ideal case" in our (real life) every
> day work?  If any of us did, we would soon be out of a job because nobody
> needs a Ph.D. and years of experience to do something as easy "the ideal
> case" (every character has a consistency index of 1 and phylogeny
> reconstruction is child's play).  So in the "practical, ordinary,
> conventional, usual, everyday case" cladistics will fail to accurately
> represent hybridization events.
>
> SIGH!!
>

Years ago (late '70s-early '80s), when the phenetocladistic wars were
being hotly contested, I  observed two things.  First, the
original examples of parsimony analysis (e.g., the first Farris/Kluge
paper), were done with very clean data sets (i.e., CI of approximately 1);
at a meeting I presented my attempts with oaks and Walter Fitch and others
were quite interested to see the numerous "most parsimonious trees" and
the high levels of homoplasy in my data set.  And, I was able to find
shorter trees by intuitive branch-swapping and simply shuffling the deck
(in the days when the data were on key-punched cards).

Second, I have always suspected that one reason cladistics caught on so
quickly, compared to classical NT, was the relative simplicity of doing a
cladistic analysis.  After all, it is quite easy to conduct a cladistic
analysis by hand and the mathematical approach (as presented in Farris'
early papers) was also quite simple. NT, on the other hand, based on
multivariate statistics, was much more esoteric.  While a cluster analysis
can also be done simply by hand (maybe with a calculator), ordinations are
far more demanding.

Admittedly, from the view of getting a solution, parsimony analysis is a
far more difficult problem than conducting a principal components
analysis.  But in those days, many were not aware of the fact that a
single data set could produce so many equally parsimonious trees.

Besides, cladistics was designed specifically to construct phylogenetic
hypotheses while phenetics is based on the criterion of overall
similarity, which may yield a very good approximation of phylogenetic
relationships, but is not intended to do so.  I believe many of the
questons related to the informaton-content and utility of the resultant
classifications are as yet unresolved.

So, I use phenetics quite a bit because of the ubiquity of hybrids in my
favorite organisms (oaks).  But, I do not claim that my results reflect
phylogenetic relationships.  At this point, my view is that a phenetic
classification of oaks is much more desirable than a cladistic
classification, especially with respect to utility.  That doesn't mean
that cladistics should not be applied, just that the cladistic results are
a hypothesis that is likely to change, perhaps significantly, as more
data are gathered and we develop a better understanding of exactly what
constitutes a species (however defined) in oaks.


Richard J. Jensen      |   E-MAIL: rjensen at saintmarys.edu
Dept. of Biology       |   TELEPHONE: 219-284-4674
Saint Mary's College   |   FAX: 219-284-4716
Notre Dame, IN  46556  |