66 Françoise Longy
Notes
- This idea lies at the core of the traditional intentionalist theory of artifacts. For a quick historic survey of this
tradition see McLaughlin (2001: 42–62). - See, for instance, Millikan (1984: ch.1); Bigelow and Pargetter (1987: §III); (Griffi ths 1993: §8).
- Since it is debatable whether all the functions attributed are of a single sort and whether all such attributions
can be associated with etiological explanations and normative (or teleological) claims, it is better to start from
the hypothesis that there may be two sorts of functions, independent of each other, only one of which is related
to etiological explanations and normative claims. To avoid any ambiguity, some authors have called functions
of this latter sort “teleofunctions.” Since I am concerned here with teleofunctions only, I can keep using the usual
term function without fear of creating ambiguity. - For Hempel, functions do not correspond to a type of scientifi cally admissible property. For Cummins, they
do—they are physical dispositions—but they make up no single type since the difference between them and
ordinary physical dispositions, nonfunctional ones, is not ontological but pragmatic. See Longy (in press) for
clarifi cation. - This is what I expound on in Longy (in press).
- For the sake of discussion, we temporarily endorse the distinction between intended and selected functions.
Such a distinction is supposed to apply to the proper function of an item, that is, to the one attached to it as a
member of an artifact type but not to the possible occasional use functions it may get in some particular circum-
stances, such as when a pencil is used as a hairpin. - See, for instance, Neander (1991: 174) for a classical defi nition of SEL.
- Of course there is no objection to putting together heterogeneous things for pragmatic reasons. “Pets” is a
good example of what we may call a pragmatic category. It is not, however, a category that will have a place in
biology, contrary to the categories of “dogs” or “mammals.” An ontological category is a category that carves
nature at some of its joints and has for this reason a place in science. We cannot deal at length with this notion
here, but it is discussed in Longy (in press). - This conclusion, a clear negation of pluralism, is the only point on which I disagree with Perlman (this
volume). I agree totally with the agenda he sets for the further development of theories of teleofunction, with
his four DON’Ts, but unlike him I think one has to renounce pluralism to satisfy these DON’Ts, especially the
one of not drawing a hard line between natural functions and artifact functions (see note 8). - Sperber is not the only one to have pointed out the differences between natural and cultural mechanisms
and the diffi culties encountered when trying to transpose natural selection to sociocultural phenomena. In regard
to this question, the recent book by Tim Lewens (2004) is of particular interest since it focuses on functions.
The well-argued conclusion of Lewens is that SEL, defi ned as it currently is with an explicit reference to natural
selection, cannot be extended to artifact functions because, to put it briefl y, sociocultural selection differs largely
from natural selection (2004:140–157). I agree with him on this point, but I nonetheless dispute his fi nal conclu-
sion that there is a simple analogy between biological and artifactual functions. In line with what Sperber does
when he replaces selection by the broader notion of “propagation,” I defend instead that what we should aim at
is a more general and abstract notion of “teleofunction.” A step toward abstraction makes it, indeed, possible to
preserve both diversity and continuity. Different sorts of selective mechanisms are present in the sociocultural
realm. Some of these are quite similar to natural selection while others are not. However, with a more abstract
notion of “function,” we are no longer obliged to draw a line between the genuine and the watered-down mecha-
nisms of selection. - Sperber (2007: 128). As he specifi es, a propagation is a repeated reproduction that supposes neither a defi nite
copying mechanism nor a strong inheritance (new items do not necessarily have to “inherit all their relevant
properties from previous tokens of the type” [2007: 127]). - It depends of course on the limits placed upon the vagueness of the term artifact. Here it applies only to
material entities. - In fact things are more complex. SEL will account for the attribution of a peeling function if the selection
is somehow relative to the peeling capacity. The difference in the quality of the blade-handle connection is clearly
such a case. A less well connected blade will sooner loosen or make peeling diffi cult by wobbling. The colored
handle case is more problematic. If this variant won simply because it answered better the aesthetic taste of