Is Natural Selection a Population-Level Causal Process? - Mechanism and Causality in Biology and Economics - page 168

decreases. In such a case, although most biologists would predict that drift tends to
dominate natural selection, Millstein would decline to make such a prediction.
These cases represent a divergence of Millstein's account from a broad consen-
sus among biologists. Such a divergence arises ultimately from Millstein's view
that when natural selection and drift operate together, there must be two separate
sampling processes. As a result, it is possible for the two separate sampling
processes either to increase their strengths at the same time or to decrease their
strengths at the same time. When such a possible scenario happens, it would be hard
to tell whether natural selection or drift tends to predominate. In contrast, if natural
selection and drift, when they operate together, are one single sampling process and
if their strengths always counteract each other, then neither is it possible for the two
strengths to go up together nor is it possible for them to diminish at the same time.
And an increase in the crucial quantity 4Ns means definitely two things: both an
increase in the strength of natural selection and a decrease in the strength of drift,
which in turn means that natural selection tends to dominate drift, and vice versa for
a decrease in the crucial quantity. It is unclear how, on Millstein's account, two
separate sampling processes would be so linked together that there definitely is a
trade-off between each other's strengths. Given that the causal relevance affects
only the strength of natural selection, such a trade-off would seem impossible.
Now, consider the following statement ( T ):
( T ) Whenever natural selection and drift operate together, a change in the strength of
natural selection implies an inverse change in the strength of drift, and vice versa.

I would say that any account of natural selection and drift would be dismissed as
inadequate if it fails to show that ( T ) is the case. Thus, one problem with Millstein's
account is clearly this: it fails to meet the statement ( T ). It turns out once again that
the problem with Millstein's account arises ultimately from her view that natural
selection and drift are two separate sampling processes.
Admittedly, Millstein's account, most of the time , leads to no divergence from
biologists' judgment as to whether natural selection or drift tends to predominate.
That is because when natural selection and drift operate together among a real
population of organisms, it seems reasonable to suppose that the degree of causal
relevance, which is essential to natural selection, remains constant , namely, a fixed
nonzero value. Accordingly, both the strength of natural selection and the strength
of drift depend exclusively on one single factor, i.e., the effective population size.
So although natural selection and drift are two separate sampling processes, there
could be a trade-off between their strengths. Still, theoretically speaking ,

Millstein's account is not consistent with ( T ).
Now let's return to Brandon's account. Recall that he characterizes drift as any
deviation from expected result due to sampling error. Brandon's characterization
has an immediate and significant advantage: drift as unrepresentative sampling
process is inherently related to the effective population size. The smaller the
effective population size is, the more likely it is that drift would be unrepresenta-
tive, and accordingly, the larger its strength would be. Although a sampling process
is drift if and only if it is unrepresentative, not every representative sampling