X-Message-Number: 11818
From: Daniel Ust <>
Subject: Testing evolutionary explanations
Date: Tue, 25 May 1999 09:38:59 -0400

This might seems a bit off topic, but on this list and many others,
evolutionary arguments have been invoked to defend or attack various
positions.  The typical form of these arguments has been to link some trait
with natural selection via a "just so" or story scenario, expressing a
qualitative and intuitive model of causation.  Many biologists have
criticized this method for being arbitrary.   (For instance, B. C. Goodwin's
"Evolution and the Generative Order" (in _Theoretical Biology: Epigenetic
and Evolutionary Order from Complex Systems_ (1992[1989]).)

Some might argue using a flawed method of explanation is better than using
none at all.  Science can work with faulty explanations.  Instances of such
pepper history.  Bohr's model of the atom, e.g., is now known to be wrong,
but it did focus attention on the important traits of atoms at that time,
and hacked a path through which other thinkers could travel toward better
theories.

It can also be argued that any explanation we have at any point will always
be flawed.  Human knowledge will never be perfect or complete.  I would also
agree with this and that a "just so" explanation can be a starting point for
further inquiry.

Still, in evolutionary biology there are better ways of evaluation than
making up a good scenario of how things came to be.  They are not perfect,
though they are almost all better than "just so" explanations in that they
offer a testing methodology.  Some of these methods have the benefit, so
important in science, of being independent of
particular causal theories of evolution -- such as the neoDarwinian
"Synthetic" Theory, the entropic "Unified Theory" (of Brooks and Wiley), and
the Neutral Theory (of Motoo Kimura).  These methods include paleontology,
population biology, phenetics, and cladisitics.
Here I will concentrate on cladistics.

Cladistics uses the traits of biology groups, such as populations, species
and genera, to reconstruct the lineage of the groups.  The method aims at a
transparent recovery of the history of cladogenic events-i.e., splits in the
ancestral group.  By "transparent" is meant a method that does not rely on
intuitive notions, one which should allow other workers to
quickly and repeatedly get the same results without bias.  This is akin to
addition or multiplication.  If someone adds one number to another to
generate a third, anyone should be able to repeat the process and clearly
see how the third number is generated.  (See Wiley et al. (1990)  _The
Compleat Cladist: A Primer of Phylogenetic Procedures_ .)

To give an example of this, Barbara A. Block et al. test the hypothesis that
endothermy evolved only once in fishes in "Evolution of Endothermy in Fish:
Mapping Physiological
Traits on a Molecular Phylogeny" (Science (260) 1993 April 9).  Endothermy
is, in lay terms, warm-bloodedness.  Put simply, the hypothesis is that
warm-bloodedness came about once and was passed along to all other
warm-blooded fishes.  This would mean
that all warm-blooded fishes should be more related to each other than they
are to non-warm-blooded fishes -- in the same way that I'm more related to
my brother than I am to his wife.

Block et al. examine the distribution and changes in DNA to determine that
endothermy in fact has evolved independently three times in fishes.  This is
the equivalent of saying
that not all warm-blooded fishes are more related to each other than they
are to non-warm-blooded fishes.  Thus, they tested the above hypothesis and
it failed.  A "just so"
explanation most likely would not have revealed this.

I don't have enough time here to over the whole method, but cladistic
methods work by comparing traits (which can be anything from a complex
behavior to DNA sequences)
between groups (e.g., species) and seeing in terms of these which groups are
more closely related.  This allows one to recapture the history of evolution
as the sequence of branchings between traits.  For instance, two species of
fish which are closely related in lots of traits might differ in endothermy,
giving evidence that endothermy evolved separately.

To put this a bit more clearly, imagine three sister species (species
sharing a common ancestor) which form a clade (meaning there are only these
species and no other that share this common ancestor): A, B, and C.
Cladistically, this set can give rise to four hyptheses. The first is that
they all evolved directly from their common ancestor, meaning there were no
intermediary splits in their line.  The evidence for this would be that none
of them shared any other traits besides the ones all of them shared.  Let's
call their ancestor M (for mother species:).  M has traits (0,0,0,0,0) --
each 0 representing a binary trait: you have it or you don't have it.  A has
traits (1,0,0,0,0); B (0,1,0,0,0), and C (0,0,1,0,0).  They all share the
last two traits (the last two 0s), but differ completely in the others.
Graphically this could be modeled by on point M with three lines leading to
points A, B, and C.

Imagine now instead that while A, B, and C share a common ancestor, A and B
share another common ancestor which C does not share.  In other words,
between M and A-B, there lines another species which is the mother of the
A-B sister pair.  We'll call this M'.  Evidence for this would be something
like M and C have the same traits as above, but A has traits (1,0,0,0,0) and
B has traits (1,1,0,0,0).  Note A and B share that first 1 with each, but
not with C; they also differ by the third 0 with C and share 0s four and
five with C.  Graphically, this could be pictured as a point M with one line
leading to C and another leading to M' AND two lines leading from M' to A
and B.  Thus M evolved into M' and C, and M' then evolved into A and B.

The two remaining possibilities here, which I won't cover in such detail,
are that B and C are more closely related or that A and C are more closely
related.

Aside from the order in which the branches happened, why would this be
important?  One might, say, have an evolutionary hypothesis like long billed
hummingbirds in South America evolved from shorter billed ones in response
to environmental pressure.  I bring up this example BECAUSE not only has
someone done the analysis but it seems almost obvious to be true.  In actual
cladistic analysis, it was found this was not the case.  The short billed
ones actually came later, evolving from their longer billed cousins.  If we
had stuck with "just so" stories of evolution, I doubt we'd have uncovered
this one.  (The article in which this was covered was in an issue of
BioScience about 5 years ago.  I can't find it right now, but believe the
whole issue was dedicated to hummingbirds.  I'm sure a title search of the
back issues on that keyword should arrive at it.)

This is one method of testing evolutionary hypotheses.  I won't cover
population biology here, except to say it relies on math and has been used
extensively to cover INTRAspecies trait changes -- stuff we can see happen
in the lab by running experiments.  Hence the name "population" instead of
"clade" or "species" -- a population is considered a subset of a species.

None of this means that "just so" arguments are always wrong, but a "just
so" argument by itself should NOT be taken as a valid scientific argument.
At best, it is a guess, and, without further evidence to back, must remain
so.  It can suggest areas for future research, but is by no means the end
goal.

I hope the above proves helpful for CryoNet list members.  We should aim at
the truth -- not presenting explanations merely to confirm us in our
prejudices.

Vivamus!

Daniel Ust

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=11818