538 Tim Lewens
cannot say this, because it equates having a function with being able to perform
that function; hence, it rules out malfunction altogether. This move would entail
that a mutant light coloured moth that appears in the population after the dark
wings go to fixation is not malfunctioning.
So what should the SE account say about light wings while the dark ones are
going to fixation? Are they malfunctioning or not? The account could say ‘no’
of many of these moths, perhaps on the grounds that a trait token only has a
function F if token homologous traits in its ancestors contributed to fitness by F-
ing. Ancestors of many of the light-winged moths did not (we can suppose) have
wings that contributed to fitness in this way. Alternatively, the account could
say ‘yes’, perhaps on the grounds that a trait token has a function F in case it
is homologous with other tokens that have contributed to fitness in the past by
F-ing. The light wings and the dark wings are homologous, and dark wings have
contributed to fitness by providing camouflage. This second option might seem
strange, but if the basic gist of the SE account is that the function of a trait is
whatever effect explains the selective success of traitsof that type, we might say
that we should treat dark moth wings, light moth wings after the dark ones have
gone to fixation, and light wings while the dark ones are going to fixation, all as
‘traits of the same type’.
Now let me come clean: the primary point of this excursion is not to suggest
that it is impossible for the SE account to manufacture a set of hard and fast
conditions, which are phrased in purely biological terms, that will tell us when
some trait is to count as malfunctioning. There may be all kinds of ways to
achieve this job, and I have sketched some above. My point, rather, is that there
is no particular way of doing this suggested by the practice of evolutionary biology
itself, nor does that practice make such a project needful. The biologist who
wishes to understand changes in the frequencies of traits in a population has no
need to decide which should count as malfunctioning, and which should count as
functioning: it is enough to construct a model that can explain trait fluctuations.
One might reply by pointing out that the concept of malfunction does play
an important role in the context of assessments of health and disease: diseased
traits are those that have ‘gone wrong’. Suppose the environmental change in
the example above, the one that results in the failure of a very high proportion
of traits in some population to augment fitness in the usual way, is in fact the
sudden presence in the human population of a virus that affects the development
of the kidneys so that they do not filter blood efficiently. Here humans may well
want to assert that most kidneys are now diseased. I tried to suggest above that
the study of evolution does not, in itself, need to make assessments of health and
disease in cases like this. This does not show that the health/disease distinction
should not be understood in biological terms (see [Boorse, 1997], for arguments
that it should be). Perhaps it is possible to equate disease with malfunction, and
to define malfunction in biological terms. Whatever the result of that debate, we
do not find such a malfunction concept playing a role in evolutionary biological
inquiry; rather, we construct that concept from evolutionary biology’s conceptual