Probably the best critic of Dawkins is Elliott Sober. I recommend his
book, "The nature of selection : evolutionary theory in philosophical
focus."
I'd also recommend Griffiths' and Sterelny's "Sex and Death: An
Introduction to Philosophy of Biology," because Sterelny is one of the
philosophers of biology most sympathetic to Dawkins, but the book is
still critical of his ideas.
I'll just copy some old posts I made on this subject.
Perhaps the most straightforward argument against the gene being the
unit of selection is offered by Sober and Lewontin in "Artifact, Cause,
and Genic Selection," reprinted in Sober, ed., Conceptual Issues in
Evolutionary Biology.
Consider heterozygote superiority. If you specify the selection
coefficients of individual genes, their values will vary as the
population changes in gene frequency. In contrast, the selection
coefficients of individual genotypes remain constant.
We therefore lose the ability to think of selection as a force, as the
cause of differences of fitness. Models of selfish genes can produce the
same results as models of selection acting at a higher level, but you
lose the concept of selection as a force and as a cause of evolution,
rather than just a measure of a propensity to change.
As an illustration, let p be the frequency of dominant allele A and q
be the frequency of recessive allele a, where p + q = 1. Let w1 be the
fitness of AA, w2 the fitness of Aa, and w3 the fitness of aa. Before
selection, the population will contain AA, Aa, and aa in the proportion,
p^2:2pq:q^2. The average fitness of the population will be p^2w1 +
2pgw2 + q^2w3. The population will move toward a stable equilibrium
frequency p' = (w3 - w2) / ( (w1 - w2) + (w3 - w2)).
Now, let's try this with selfish genes. If we want to determine the
fitness of the individual allele A, W.A, we calculate that
W.A * frequency of A before selection = frequency of A after selection *
average fitness. Frequency of a before selection is p, after selection
it is w3p^2 + w2pq / W. Hence, W.A = w1p + w2q. Similarly, W of the
recessive allele a = w3q + w2p.
Hence, the fitness of single genes is the just the average of
the fitness values of the genotype, weighted by the frequency of their
occurrence in the genotype.
The fitness of the each of the three genotypes, AA, Aa, and aa, are
constants. They do not change as the population reaches equilibrium.
But the fitness of individuals genes does change constantly as their
frequency changes. Whereas the fitness of the genotypes has a real
relationship to the viability of the organism, the fitness of the
allele does not. You can calculate the fitness of alleles if you
know the fitness of the genotype, but you cannot in general calculate
the fitness of the genotype from the fitness values of the alleles.
The selfish genes model leads to a lost of information.
For example, assume the homozygotes are lethal. The equilibrium
frequency of each allele will be 0.5. Before selection, the three
genotypes will be in the proportion, 1/4, 1/2, 1/4. After selection, 0,
1, 0. When they reproduce, the population will return to 1/4, 1/2, 1/4,
then selection will occur and return the population to 0, 1, 0. And so
on and so on. But according to the selfish gene model, at equilibrium
the fitness values of both genes are 1, and their selection coefficients
are therefore 0. Therefore, the selfish gene model provides no means for
explaining why the zig-zag is 0. According to the selfish gene model,
selection is not happening at equilibrium. We have lost the concept of
selection as something external to fitness. Selfish genes also provide
no model of heterozygote superiority, since it measures only the fitness
of individual alleles, and can't tell us that the diploid genotype is
superior.
Sober and Lewontin go on to provide examples of where selfish genes -
genic selection - might occur: they mention that chromosomes in the
house mouse that contain the t-allele have an increased chance of being
represented in the sperm pool of heterozygous mice. The t-allele might
be thought of as a selfish gene at this level. Also, dominant alleles
that are always lethal, and phenotypic traits that are controlled by
a single locus where the heterozygote is intermediate in fitness between
the two homozygotes, could be modeled as selfish genes. Genes that have
no effect on the phenotype, as with most selfish DNA, also could be
modelled using genic selection.
But whenever the fitness of a gene depends on what's present at some
other locus, the selection coefficient of individual alleles will be
the average over all genetic contexts, and - most importantly -
they will change as the population evolves, and the idea of selection as
the external cause of fitness differences is lost.
Sober and Lewontin argue that genic selection will be rare, since
the fitness of alleles often depends on the presence or absence of other
alleles. They discuss chromosome inversions in the
grass-hopper Moraba scura, and in general point out that many models of
stabilizing selection depend on the fitness of individual loci being
determined by the presence of other genes.
---
One point [Griffiths and Sterelny]
make is that gene interaction isn't the only problem for genic
selection. Consider that two alleles could have the same phenotypic
effect: if A and B are distinct alleles and they both can make
the difference X in the organism, isn't more sensible to say X is
selected, rather than A and B are selected?
The "gene selectionism" Dawkins popularized depends on the
concept of "genes for" traits: if individual genes are to be
the targets of selection, they must be "visible": they have
to be real entities with similar phenotypic effects in a
particular context. Kim Sterelny and Paul Griffiths observe
that the phrase "gene for" can indicate a molecular sequence
that makes a difference in the phenotype of a particular
individual, or it can indicate those sequences that consistently
make a particular difference in the phenotype in a given
context. In the former case, there is no guarantee that the
gene will have the same phenotypic effects in other
individuals and in other contexts; while in the latter case,
although the "genes" have a constant phenotypic effect, there
is no guarantee that the DNA sequences underlying these
effects are identical or part of the same lineage. They argue
that "gene selectionist" ideas depend on an "empirical bet":
unless "genes lineages ... have some form of underlying
molecular unity and some form of similar phenotypic effect ...
we can make no sense of the idea that the fate of phenotypes
affects evolution only through its effect on gene lineages."
Sterelny and Griffiths go on to note that the use of the
phrase "gene for" in medical genetics has obscured some of the
issues involved:
The fact that a serious empirical commitment is being
made here is obscured by the fact that the "gene for"
locution has its primary home in medical genetics, where
in many cases the complexities of the gene/phenotype relation
can be ignored. Medical genetics discovers "genes for" disease
phenotypes. These involve some major defect in an evolved gene
that, in its normal role, interacts with many other genes. Such
disease genes include those for albinism, melanism, and other
pigment changes caused by defect pigment-making genes, and
those for dwarfism caused by hormonal defects. They also
include the most famous case of all: the genes responsible for
phenylketonuria and its variants, caused by the absence
of an enzyme or failure of its function. These pathological
genes impair normal development to such an extent that they
dominate variance in the phenotypic traits they affect
under almost any background conditions. If we were interested
only in the evolution of disruptions of existing phenotypes, we
could perhaps assume that the difference-maker genes for
such traits were real entities at both the molecular and the
phenotypic level. But biology is at least as interested in
the evolution of new complex phenotypes as in such disruptions.
_______________________________________________
Marxism-Thaxis mailing list
Marxism-Thaxis@lists.econ.utah.edu
To change your options or unsubscribe go to:
http://lists.econ.utah.edu/mailman/listinfo/marxism-thaxis
