Functions 537anomalies — rare instances that conform to no pattern of theoretical interest —
rather than as malfunctioning traits. They are discounted for the same reason
that any science might discount an extremely improbable accident — because
they are too unusual to have general impact, not because they are examples of
something that has isn’t doing what it is supposed to. Some traits that are present
at a high frequency may appear to be detrimental to the fitness of their bearers.
The biologist might try to explain their presence in terms of drift, or in terms
of pleiotropic linkage to some other trait that is under selection. But neither
explanation describes the trait as malfunctioning. Suppose, however, that the
trait is present at high frequency because it is descended from a trait that was
selected for, but whose developmental pathway has been perturbed by some large-
scale environmental change, with the result that what is now present is a new
trait which reduces fitness. If the perturbation is short-lived — one generation in
duration — we might write the whole thing off as an anomaly in the context of
the overall dynamics of the population, rather than a malfunction. If, on the other
hand, the perturbation is long-lived, then we will need to readjust the terms of
our population analysis to see the new trait as one of lasting evolutionary interest.
In neither case is it pressing to judge that the new trait is malfunctioning. All
of these observations are borne out by the fact that while evolutionary biologists
regularly hypothesise functions for traits, they are very rarely in the business of
offering claims about trait malfunctions.
What does the SE account say about malfunction? The basic gist of the SE
account is that a trait is malfunctioning just in case it fails to have the effect that
explains its presence. More formally, a token trait is malfunctioning if and only
if it does not have the effect (or is less proficient in its performance of the effect)
that accounts for the prevalence of traits of that type. This sounds sensible. Just
as an artefact malfunctions if it cannot do what it is intended to do, so a trait
malfunctions if it cannot do what it is selected for. But there are cases where it
is not so clear what we should say. Imagine a population of peppered moths that
contains both dark and light forms in equal proportions. Suppose there has been no
predation from birds. Now suppose that birds arrive in the moths’ habitat, and the
dark moths begin to increase their frequency because they are better camouflaged
against the dark trees. It seems clear that once the dark moths are at fixation,
in the event that a light moth appears in the population through mutation we
can say that these light wings are malfunctioning. But what if the light moth
appears in the population through migration from a population in which the trees
are light in colour? Presumably the SE account will say that this moth’s wings
are malfunctioning too, although this is not clear, especially if there is no pedigree
of providing camouflage in this moth’s ancestors. What about the wings of the
light coloured moths a few generations after selection for camouflage begins, and
before the dark variety goes to fixation? Here is what the SE account should not
say: these light coloured wings are not malfunctioning, because camouflage does
not explain the presence in the population of light coloured wings, hence moths
with light coloured wings do not have camouflage as a function. The SE account