108 Maarten Franssen
It is the/a proper function of an item x of an organism o to do that which items of x’s type did to
contribute to the inclusive fi tness of o’s ancestors, and which caused the genotype, of which x is the
phenotypic expression, to be selected by natural selection.
This defi nition coincides with the reduction to the direct proper function of biological
items of Millikan’s much more general account of function. What caused the genotype of
which a token x is the phenotypic expression to be selected by natural selection is what
items of x’s type contribute on the average to the average inclusive fi tness of o’s ancestors.
The causal explanation of the selection does not require the positive contribution of every
historical possessor of tokens of the type corresponding to x to the selection of the
genotype.
Davies might choose to save his argument by claiming that an organism with a mal-
functioning token of T is not the expression of the genotype that was selected, since that
genotype codes for functioning tokens of T. This would be a relevant defense if all defor-
mation resulting in malfunctioning had its origin in genotypic defects. Indeed, in his book,
Davies occasionally writes as if he believes this to be the case. However, deformation can
be the effect of many different causes. Even congenital deformations are not necessarily
the effect of genotypic faults but can also be the result of disturbances in the biochemical
circumstances in which the fetus developed. It is therefore perfectly possible for a pheno-
type with a malfunctioning token to be the expression of the “faultless” genotype that was
selected.
As already noted, the type of which x is a token cannot be the functional type, that is,
the type that is defi ned in terms of x’s function. For biological items, however, such type
defi nitions are hardly, if ever, a problem: they are furnished by morphological or devel-
opmental considerations or by homology, as Davies is well aware (2001, p. 199, n. 6).^7
The PF-type theories, therefore, are able to attribute malfunctions. This does not mean,
however, that only these theories are able to describe the situation where a particular item
fails to do or is incapable of doing what would be necessary for performing a particular
function. The CR-type theories describe it in precisely the terms that I have been using so
far: in terms of type and token behavior. If, on the PF-type or etiological theories, an item
malfunctions, then on any theory, it does not show the behavior that tokens of its type
normally or typically show, or it does not now show—or is currently unable to show—the
behavior that it used to show and that counted as performing a particular function. Propo-
nents of the CR-type theories consider this sort of description to be adequate enough.
So there are items of which the PF-type theories say that their function is to F, but they
malfunction, whereas the CR-type theories do not attribute to them the function to do F.
In deciding whether one of these positions is the “correct” one, it is important to recognize
which intuitions, exactly, need to be saved here. Surely we would like to remain justifi ed,
at least within certain limits, in holding that a malfunctioning (i.e., broken) pump is still
a pump, and similarly a malfunctioning (i.e., deformed, diseased) heart is still a heart.^8