On Tue, 20 Jun 1995, Dennis Paulson wrote:

> David wrote: "Simply observing that because Baltimore and Bullock's are
> more like each other than either is like Altamira or Streak-backed (i.e.
> gross phenotypic similarity) and concluding that they are more closely
> related to one another is not an appeal to parsimony."
>=20
> Well, "gross phenotypic similarity" *where there is no evidence clearly
> pointing toward convergence* (all italicized) has counted for a lot in th=
e
> history of both taxonomy and phylogeny, and I don't think it matters
> whether you adhere to a biological-species concept or not. It's *only*
> when molecular data have been contrary to this that people have questione=
d
> earlier conclusions. I agree entirely that hybridization is not the poin=
t
> here, but I would surely be stimulated by its frequent occurrence to look
> closely at the relationship between two species!
>=20
> Gross phenotypic similarity in this pair includes many similarities in
> plumage patterns, including characters not really shared by any other
> orioles, for example whitish-bellied immature and orange-edged tail in
> males; size, bill size and shape (identical); vocalizations, both calls a=
nd
> songs (the NGS guide makes them sound more different than they are);
> habitat choice (not that there are *that* many other oriole habitats from
> which to choose, but the Orchard, for example, did so); nest construction
> (somewhat different but, again, more alike than the putative relatives) a=
nd
> egg pattern (tentative conclusion, small sample of Baltimore and no
> Streak-backed, but Baltimore like Bullock's rather than like Altamira).
> Any of these similarities are possible by convergence, but there seem too
> many to me--thus I'll go back to parsimony.

This has turned out to be quite a can of worms.

Convergence vs. sister-group relationship of Baltimore=20
and Bullock'=92s are not the only possibilities here. Bullock's and=20
Baltimore could have inherited the characters in question from=20
a common ancestor and retained them unchanged while the=20
characters changed in other descendants of the same common=20
ancestor. For example, in Scott Freeman'=92s tree the=20
relationships of Streak-backed and Bullock'=92s to the Baltimore=20
+ Altamira clade are ambiguous. That is, we don'=92t know if=20
there was a 3-way split between Streak, Bull, and the=20
Balt+Alta clade, if Streak and Bull are sister groups (i.e.=20
Bull+Streak is a clade that is the sister group of the=20
Balt+Alta clade), if Streak is the sister group of Balt+Alta,=20
or is Bull is the sister group of Balt+Alta. This last=20
possibility would make the sharing of Dennis'=92s characters by=20
Bull and Balt, followed by a change in these characters in=20
ancestor of Altamira, just as parsimonious as convergence=20
between Baltimore and Bullock'=92s. =20

But it would still be more parsimonious to have Bullock'=92s and=20
Baltimore as sister groups to one another *if* the characters=20
Dennis cited are *novelties* relative to their homologs in=20
other orioles. But we have no evidence that that is the case. =20
And the characters Dennis cited are contradicted by similar=20
kinds of characters -- for example, Baltimore shares with=20
Orchard and Scott'=92s the trait of having a full black hood and=20
back, wheras Bullock'=92s has a black "bib" that is more like=20
that of Streak-backed and Altamira than the full hood of=20
Baltimore. =20

How can we tell which of these resemblances indicate=20
especially close relationship, which represent convergence,=20
and which simply represent lack of change in ancestral=20
characters? On the basis of of gross similarity the orioles=20
having hoods are each other'=92s closest relatives, and th=20
orioles having bibs are each other'=92s closest relatives. But=20
under the cladistic approach, only evolutionarily novel=20
versions of characters are evidence of close relationship. =20
For example, consider the bib vs. hood distribution of black=20
in these orioles, and assume for the sake of discussion that=20
the orioles in quesion share a common ancestor that had either=20
a hood or a bib. If the ancestor had a bib, then the=20
occurence of a hood would be a novel character, and would be=20
evidence that Baltimore, Orchard and Scott'=92s share a common=20
ancestor that is not shared by Bullock'=92s and the others. But=20
if the common ancestor of the entire group had a hood, than=20
the bib would be the novel character, and it would indicate=20
that Streak-backed, Altamira, and Bullock's=92 share a comon=20
ancestor not shared by Baltimore and the others. Note that in=20
either case, the taxa lacking the new character are *not*=20
automatically assumed to be each other's=92 closest relatives.
For example, in the case of the hood being the ancestral=20
character, Baltimore, which has a hood, could well be more=20
closely related to Bullock'=92s and the other bibbed orioles that=20
it is to the other hooded orioles (we would have to recognize this=20
on the basis of another character). =20

Note there there will be contradictions between the=20
above characters and those Dennis listed. Parsimony is=20
crucial here, as we search for an arrangement of taxa that=20
minimizes the number of evolutionary changes in each=20
character, when all of the characters are considered together. =20
In practice, there are computer algorithms that do this for=20
you. Dennis, if you want to get more plumage characters,=20
etc., from a larger set of orioles, we can plug them into PAUP=20
and see if your phenotypic evidence, when treated=20
cladistically, really disagrees with Scott Freeman'=92s tree. =20

So, in response to Gene'=92ssuggestion that the differences between=20
the methods under discussion are not that great, gross=20
similarity ("phenetics") and cladistics really do produce=20
different trees for these taxa on the basis of this character. =20
I don'=92t think it is an exaggeration at all to call these=20
approaches (and attendant mindsets) different *paradigms*. =20
Anyone who seriously questions the magnitude of the=20
differences is advised to peruse issues of Systematic Zoology=20
from the 1970s when these issues were hotly and extensively debated. =20
The distinctions are not subtle, and the two approaches quite=20
often produce very different *arrangements* of taxa using the=20
same datasets. The phenetic paradigm -- degree of similarity=20
equals degree of relationship -- actually antedates evolution=20
as a concept. Cladistics came on the scene as a radical replacement=20
in the 60s, but was not really appreciated utnil the 70s. DNA-DNA=20
hybridization uses the phenetic paradigm: degree of similarity of=20
DNA is assumed to equal degree of relationship. Generally speaking,=20
any method that expresses degree of relationship in terms of "distance"=20
is phenetic. Cladistic analysis can be used with molecular=20
characters -- for example, parsimony analysis of DNA sequences --=20
and with morphological, or behavioral characters. The=20
resulting tree, a cladogram, is as much a map of character=20
change (i.e., evolutionary transformations) as it is a tree of=20
phylognetic relationshisps. This is of great utility to=20
people who want to study how evolutionary change *actually=20
happens* (e.g, plumage evolution). Check the pages of recent=20
issues of Evolution and similar journals for example of such=20
applications. =20

David Wright
dwright at u.washington.edu