IUBio Biosequences .. Software .. Molbio soft .. Network News .. FTP

Rouse, Fauchald, Eibye-Jacobsen, Nielsen

Kirk Fitzhugh fitzhugh at almaak.usc.edu
Wed Dec 3 21:34:19 EST 1997


I'd like to take a moment to respond to Danny Eibye-Jacobsen's posting
here, where he presents a rebuttal to Greg Rouse's rebuttal to
Eibye-Jacobsen & Nielsen's critique of Rouse & Fauchald. I'll use the name
abbreviations employed by these individuals to keep things as short as

When E-J&N's paper originally appeared, I actively avoided getting into
this debate for two simple reasons: 1) the paraphyly issue as it pertains
to R&F's cladistic analysis using "Annelida", "Pogonophora", and
"Vestimentifera" was blown out of proportion by E-J&N, and rested on
specious reasoning; 2) E-J&N's particular advocacy of successive
approximations weighting, and the criticism of R&F for not using such a
method are unfounded.  Quite frankly, I found E-J&N's diatribe to go well
beyond what was necessary, and I think it unfortunate that Rouse had to
expend so much energy writing a lengthy rebuttal.

I have no problem with salient discussions of character interpretations,
but to bounce around the question of annelid monophyly relative to what
taxa are included at the level of phylum or class simply neglects the real
issue: that our understanding of cladistic relationships among annelid
groups at just about any hierarchical level is abysmal. E-J&N state that
R&F's analysis using pogonophorans and vestimentiferans as taxa equivalent
to the Annelida runs the risk of making the Annelida paraphyletic, thus
making it a "requirement" that at least an additional analysis is needed
wherein the former groups are excluded. Do I think the Annelida could be
paraphyletic because there have been vestimentiferan and pogonophoran
phyla? Sure. Is that reason enough to claim that R&F needed to take this
into consideration? DEFINITELY NOT. The whole point of R&F's paper was to
set the stage for looking at just these issues of
pogonophoran/vestimentiferan inclusion in the Annelida.  Personally, I
thought this was eminently clear in their paper - but maybe my thought
processes are not so well tuned to the obvious. That R&F included the
Pogonophora and Vestimentifera is of no detriment to what they attempted to
accomplish.  E-J&N raise the same concerns with regard to R&F's use of the
Echiura. The simple fact is that E-J&N's supposed solution to R&F's
perceived deficiencies is no solution at all, and such an alternate
analysis provides no greater illumination. If we want to address matters of
paraphyly, then take the time to look at relationships within the groups in
question! That is the only way to settle this matter, and again, my reading
of R&F's position was that that was just their intention in their
forthcoming papers.  What we have seen are needless arguments over
semantics, not solutions to real (or imagined) problems.

For some peculiar reason there is then debate on the use of consensus
trees. E-J states that "Four out of five cladists will agree that a
consensus tree is not worth much and that the information content is higher
in any one of the trees used to construct the consensus tree." This is a
straw man. Four out of five cladists will also tell you that they know the
difference between consensus trees and original trees. Using one or the
other resides in what it is one intends to get across to their audience.
Strict consensus trees are useful in showing relative branch movements,
i.e., topological phenomena. For that they do have some "worth".
"Information content" is meaningless in this context. That E-J would use
information content of a single tree as evidence against a consensus tree
is nonsensical. The "information content" of a tree relates to those
character data that are used to construct the tree. Any cladist (or at
least 4 out of 5) will tell you that character distributions on consensus
trees are not to be considered as the consensus tree is a compilation of
other topologies, not a summary of data per se.

E-J&N argue that R&F should have used successive approximations weighting
(SAW), under the premise that "weak" characters could be down weighted in
lieu of eliminating ambiguity and providing "more conclusive results."
E-J&N have stepped on a philosophical land mine.  First, the issue of
character weighting in any incarnation lacks any sort of empirical
justification, and without such justification, there's no basis for using
it as a critique of R&F's results. Now, to say that some characters are
"weak", or, as E-J says, that cladistics "rests on the sound [sic!] logic
that all characters are not created equal (i.e., we are not infallible in
determining homologies)" is irrelevant to the matter of weighting. Further,
the "equality" of characters is a nonissue. We make observations of shared
similarity among taxa; as systematists we are driven to account for such
order in the world. Where does equality of characters fit into all this?
Nowhere, and not least of all in cladistics. If we're going to start
ranking observations then it's got to be at the level of bare observations,
not at the point of making cladograms. I know of no one that wants to
tackle that monster.

Indeed, the issue of homology is wholly separate from that of SAW. SAW was
originally intended to give greater weight of evidence to those characters
with higher cladistic reliability, i.e., with lowest levels of homoplasy.
This is an elegant argument but equally fallacious. How does one denote
"reliability"? If it is to be by way of getting results that accord with
some sort of independent measure of truth, then we have no means of
measuring such a feature. If we have independent measures of truth then
there's no need to do cladistics! This is exactly the logical nightmare
that ensnares proponents of maximum likelihood. Arguments in favor of
weighting have consistently failed because they rely on some notion of
"reliability". But then, proponents of weighting have failed to see
cladistics for what it really is: a process of developing explanatory
hypotheses to account for our observations of shared similarity among taxa.
SAW has the deficiency of denying the relevance of looking at homoplasy.
Seeing features as homologous is profoundly easy - accounting for why some
of our notions of homology do not hold up on cladograms, i.e., appear as
homoplasy, should be given the highest priority. Trying to down weight
homoplasy out of existence is simply a ludicrous endeavor. It is nothing
more than science at it's lowest. Cladists would be far better served if
the vacuous attempts at weighting were removed from consideration.

E-J says "Claus and I regard [SAW} as a necessary part of any complete
[sic] cladistic analysis." Why??? On what basis does one say that a set of
SAW trees have relevance over all minimum-length trees? I can guarantee you
that there is no basis for selecting any of the former over the latter. Let
me repeat myself - there are no empirically founded methods of character

The debate over "soft" and "hard" polytomies is about as irrelevant as how
much weight to give characters as an unfounded excuse to reduce the number
of trees. Rouse was quite correct to raise concerns as to how one could
possibly "know" when polytomies are soft or hard, which is not the same as
simply recognizing that polytomies exist as part of an explanatory
hypothesis. E-J simply evades the issue with equivocation: "does the
hypothesis that a polytomy at the phylum level can be hard...have ANY
credence? We don't believe it does." But, using this logic reduces all
polytomies to being soft. But, so what? But, is this a basis for removing
trees? But, of course not. All we've done is forsaken one level of
ignorance for another for no reason whatsoever. This is just another
example of poor science being used to make a method look flashy. When
issues in science focus more on methods than on the basis for use of the
methods, we should all be very worried. And, it's time to be worried.

So, have E-J&N shown fault with R&F? Not really. What they've shown is that
yes indeed the Annelida might be paraphyletic; but, we as well as R&F have
known that all along. Did E-J&N provide any substantive alternative to the
quandary of Annelida paraphyly? No. Was there then a need to have these
long-winded exchanges? No. Our time would have been better served talking
about the ignorance we all share as to much more fundamental issues.

Cladistics is seeing an unfortunate trend in practitioners wanting more
bells and whistles in computer programs, peculiar coding methods of
characters, and reading things out of cladograms that simply don't have a
basis in fact. I guess this makes us all feel that we're doing big-time
science. Feelings, however, are irrelevant. It's ironic that the proponents
of parsimony were right all along - cladograms represent our basic
observations in terms of common ancestry with ad hoc hypotheses of
homoplasy where needed. The simplicity and reality of these hypotheses has
been lost on many who try to read too much into cladograms, or have some
idea of how data such as nucleotide sequences evolve such that inane,
contrived weighting schemes and tossing out of data are deemed necessary.

Have a nice day...

Kirk Fitzhugh
<fitzhugh at almaak.usc.edu>

   Discuss  =  <annelida at net.bio.net> = talk to all members
   Server =  <biosci-server at net.bio.net> = un/subscribes
   Archives  = http://www.bio.net:80/hypermail/ANNELIDA/
   Resources = http://biodiversity.uno.edu/~worms/annelid.html

More information about the Annelida mailing list

Send comments to us at biosci-help [At] net.bio.net