Cladobabble

[Josh Smith <smithjb@sas.upenn.edu>]

> Dinogeorge:
>
> >Is it just me, or do other people find this endless hunt for ever more
> >"characters" just plain CRAZY?

> Chris:
>
> What?  You find the constant search for evidence of relationships unwise
> (or even unintelligent)?  We're searching for more characters because we
> keep finding new fossils and turning new technologies (and new search
> patterns) toward older discoveries.
>
> Phylogenetics, from a morphological perspective, is a feedback loop.  Dr. X
> publishes a phylogenetic hypothesis.  It suggests something unorthodox
> about taxa A, B, and C.  So Dr. Y takes a closer look at A, B, and C
> (perhaps Dr. X only read about one of these from the literature, and was
> unable to study them first-hand).  Dr. Y sees something Dr. X missed, and
> recodes a few characters she feels were inadequately described.  She also
> comes up with a new hypothesis, which looks a whole lot like its
> predecessor from Dr. X, but has some nuances.  Then Dr. Z takes a look....

    Josh:

    Wow.  How many times have we been down this same road....?  At the risk of
sounding like a broken CD, more characters is certainly not a bad goal, but 
where
I feel the emphasis should be next placed is on better characters (whatever the
heck that really means...).  The two really big problems with phylogenetic
systematics as is it currently practiced is the inherent subjectivity of 
character
selection and the extremely qualitative means by which those characters are
described and thought of. We should expect that subsequent cladograms will
resemble their predecesors--we "test" a phylogenetic "hypothesis" by the exact
same means that we "come up with the hypothesis" in the first place.  If I want 
to
understand the relationship between  _Giganotosaurus_ and 
_Carcharodontosaurus_, I
will examine these taxa and select a number of what I (and those of my peers who
have done similar analyses before me) deem to be useful characters (there are a
number of criteria by which this is done, but the researcher is still in a
position of deciding which characters to use or discard from his or her
analysis).  Combined with some outgroup taxa (also chosen), an analysis is run 
and
a "phylogenetic hypothesis" is generated in the form of a cladogram.  The only 
way
that I know of to "test" this hypothesis is to get some more characters or cut
some out (or of course modify characters) and run another analysis by the same
methodology.  This process is circular.  As I understand it (and someone please
chime in if I have this wrong, although that doesn't seem to be a problem with
this list....), to go into the analysis already presuming some relationship
between _Giganotosaurus_ and _Carcharodontosaurus_ is considered bad form 
because
it generally causes the researchers to choose characters to support that idea (a
"bad" practice).  If I think a rock sample is a basalt (my hypothesis based on 
its
color and its grain size if all I have is the sample), I run a petrographic
analysis and determine its constituent minerals and percentages (which have been
arbitrarily, but concretely, defined for rock "taxa") and see if it falls within
basalt or not. If I additionally have outcrop context, I have information about
the genesis of the sample which helps to further constrain the diagnosis of 
basalt
or not basalt.  I chose this example because the hypothesis genesis and testing
means are largely the same as they are in cladistics.  Although both situations
are largely circular, we know what is important in igneous rock classification
(the actual boundaries between the rock types are completely arbitrary, but it
doesn't matter because everyone knows and agrees on them), bulk elemental
composition with respect to mineralogy, and genesis mechanism. We do not yet 
know
what is important in defining taxa.  Cladistics and clade definitions are not 
the
problem--the species concept is the problem.  The neontologists have enough
trouble dealing with this when they have the entire organism to play with.  We
have to be content with some isolated bits and pieces of the organism.  The
species concept is an amorphous and argued point in biology, we are certainly 
far
from having a good one in paleo.  As such, a move toward increasing numbers of
characters is one logical progression of trying to better understand these bits
and pieces...absolutely.  However, the part that is in serious need of revision
here is the determination of the utility of the characters themselves.  
Massaging
the data or the techniques by which we massage the data after the data are 
already
in the matrix is one thing, but what about the data themselves? We really need 
to
move beyond character states like "orbit keyhole shaped" and crap like that.  
This
is not helping us.  It is way too subjective.  Even using terms like gracile and
robust.  Are my and Dr. X's definitions of robust the same thing?  Unless they 
are
specifically defined, these terms don't mean anything.  If I, reading Dr. X's
analysis, cannot be sure of what robust means and exactly how the character was
obtained, how can I repeat the analysis?  If I cannot repeat the analysis, it
isn't science.  I love that we are working towards increasing numbers of
characters, although I don't foresee this trend continuing forever, but I really
think we need to spend some energy towards better understanding and improving 
the
way we use the data that we do collect.  If I cannot see what the character 
really
means on the physical specimen, then it really doesn't matter what elegant
transformations the algorithms do it.  I agree with Chris totally that the
abandonment of Linnean groups is a good thing--there are no defining
characteristics of a family--it doesn't mean anything.  How can we use a term 
that
is defined as "whatever a good taxonomist says it is?"  I also agree with George
in that "believing" in a phylogeny is not a good thing--it is like emotionally
supporting ones hypothesis (another "bad" practice).  Just because Dr. X's tree
says that _Carcharodontosaurus_ and _Giganotosaurus_ are sister taxa doesn't 
mean
that they are in the real world.  It simply means that Dr. X's analysis, based 
on
those characters and those descriptions from the taxa and specimens examined,
places them as sister taxa.  Nothing more.  We are putting these animals in
arbitrary boxes that do not necessarily have anything to do with objective
reality--but this doesn't matter if we can agree on the boxes.

There are people that love cladistics and there are people who hate it, but 
with a
very few exceptions, no one is working in improving it.  I have been known to
speak out against cladistics quite a bit, but what I am usually on about is the
characters selection and description problem and the religious zeal by which 
some
tote the methodology.  People are working on improving phylogenies, sure, but to
my understanding very few paleontologists are working on the character problem.
As far as I know, no one in igneous petrology sits around and debates on how to
classify rocks--these debates happened 150 years ago.  We are not there yet, but
we should be approaching all trees with some caution until we deal with the
fundamental issues that exist within the methodology by which we are generating
these trees.

Tracy:

It's getting way out of hand. Tom says his article in Gaia is going to torn 
apart
at the SVP. Give me a break!!!

Chris:

>
> Yes!  How dare Tom do science by attempting to falsify a hypothesis!  Tom,
> shame!

Josh:
    Yeah, I don't see how we can criticize Tom for doing exactly what the
generator of a hypothesis should be doing: trying his or her hardest to punch
holes in it.



--
Josh Smith
Department of Earth and Environmental Science
University of Pennsylvania
471 Hayden Hall
240 South 33rd Street
Philadelphia, PA  19104-6316
(215) 898-5630 (Office)
(215) 898-0964 (FAX)
http://www.sas.upenn.edu/~smithjb
http://www.sas.upenn.edu/earth/vertebratepaleo/