Is Natural Selection a Population-Level Causal Process? - Mechanism and Causality in Biology and Economics - page 162

Generally speaking, causal relevance, as is well known, implies statistical
relevance. That means in C -discriminate sampling processes, physical differences
among organisms are statistically relevant to differences in reproductive success.
So F -discriminate and C -discriminate sampling processes turn out to have some-
thing in common, namely, statistical relevance. It is statistical relevance as common
core that permits the term “discriminate” to have a meaning broad enough to cover
both selection-for and selection-of. Now, the term “discriminate,” in its broader
sense, means sampling processes in which physical differences among organisms
are statistically relevant to differences in reproductive success.
So a solution to the problem of equivocality is available, and it helps to secure
the characterization of selection-of as a kind of discriminate sampling process, in
contrast to Millstein's account. Despite the solution, I suspect that what is really at
issue concerns not whether I use the term “discriminate” equivocally; the real issue
concerns instead why natural selection, as Millstein proposes, should be confined to
C -discriminate sampling processes alone. Indeed, Millstein's account of natural
selection is causalist in its entirety, which definitely bars F -discriminate sampling
processes, in general, and selection-of, in particular, from counting as selection
processes. It is interesting to note that Millstein never characterizes natural selec-
tion in terms of fitness. Although Millstein (personal communication) does not
doubt that one can have a fitness-based account of natural selection, she avoids
using the term “fitness” because it is a controversial term, and she does not need it to
make the points she wants to make. In Millstein's view, a causalist account of
natural selection remains defensible without resort to the contentious term “fitness.”
I surely agree with her on this point. Nonetheless, it seems appropriate to separate
two questions apart: what is an adequate causalist account of natural selection? And
what is an adequate account of natural selection tout court? Indeed, though I agree
that Millstein succeeds in offering an adequate causalist account, I have been
arguing contra Millstein that the idea of natural selection should be broadened
enough to include F -discriminate sampling processes. Given that the argument I
offered above is based on the controversial term “fitness,” its strength accordingly
is limited if it is to be addressed to a causalist like Millstein. So let me try not to
argue against her in terms of fitness.
Consider again the case of selection-of. Selection-of is a sampling process that
biologists often name “hitchhiking.” Such a name suggests that selection-of cannot
occur all by itself: whenever there is selection-of, there is selection-for. Let's call
such a property of selection-of hitchhikliness . Needless to say, the property of
hitchhikliness is essential to selection-of. Recall that on Millstein's causalist
account of natural selection, there is no way to count selection-of as a discriminate
sampling process. Thus, whenever selection-of and selection-for are both present,
granting that selection-for is a discriminate sampling process, there must be two,
rather than one, sampling processes. In short, Millstein's causalist account implies
that it is impossible for selection-of to be numerically identical to selection-for.
However, such an implication is problematic, for it surely is possible that a
selection-of sampling process is numerically identical to a selection-for sampling
process. I would say the property of hitchhikliness, which is essential to selection-
of, provides evidence in support of this claim: selection-of and selection-for on