198 Marzia Soavi
The third element is the Object Structure, which must include a specifi cation of all the
components that are relevant from the functional point of view, a specifi cation of materials
and dimensions.
A criticism that can be leveled against my proposal is that a functional characterization
of a kind is functional to the extent that it does not take into account the details of the
physical realization but typically is of the input-output form, or even a simple output form.
When we say that the function of a heart is to pump blood or that the function of a knife
is to cut, we are adopting an output style. In the present proposal, the functional charac-
terization includes instead the physical description of the structure of the object function-
ally characterized. I do not want to stick to the idea that artifact kinds must be purely
functional in the sense here intended; what has been proposed here can be regarded as a
functional-structural characterization.
The notion of “function” mentioned in the attribution criterion proposed here includes
a specifi cation of the object’s structural type, and for this reason it largely avoids the
problems related to multiple realizability. What I suggest is that artifact kinds are not in
fact specifi ed through those rough functional descriptions traditionally related to the
meaning of ordinary artifact terms but through the specifi c functional-structural descrip-
tions related to the design.
The mere input-output function is the one adopted in the characterization of kinds such
as chairs, watches, cars, and so forth, typically mentioned in antirealist arguments. Both
Millikan (1984) and Elder (1989) have already observed that kinds such as car or chair
are simply too generic to be considered as real kinds and that more specifi c kinds might
be better candidates for real artifact kinds. Elder proposed the example of the Eames desk
chair 1957, Millikan the example of the 1969 Plymouth Valiant 100. The criticism could
be further developed by saying that no matter how specifi c we are in the description of
the kind, as long as we adopt a functional characterization, the possibility of multiple
realizations remains, and this is precisely the difference with natural kinds. We can identify
the essence of water with the molecular structure H 2 O and perhaps the essence of a species
with a certain genetic content, but for artifacts we either specify a functional kind, thus
having the problem of multiple realizations, or we specify a structural type, thus abandon-
ing the functional characterization of artifact kinds.
I reply as follows. First, it is worth observing that there is no straightforward way of
concluding that multiple realizability holds for every functional kind. It is possible that a
certain function might be realized in only one way, but of course this cannot be a conclu-
sive argument.
Second, I contest the idea that signifi cant multiple realizability is really possible for
artifact kinds, because what happens during the design phase of artifact production is that
a certain function is related to a certain physical structure. The type of artifact defi ned by
a certain design is the result of such a connection. It seems to be rather arbitrary to decide
to stop the characterization of artifact kinds at a general functional level, as the nature of