Biological Approaches to Mental Representation 561tion: Tomorrow is the day after today.^6 And so, as long as DAY, TODAY and
AFTER have instantiations with which we have interacted, it is not clear that
TOMORROW presents a deep problem.
Perhaps definitionism does not work for all complex concepts. Perhaps, it does
not work for UNICORN. But the philosophical and psychological literature on
concepts offers other ideas about how complex concepts could be structured and
it could be that different complex concepts have different kinds of structures.^7
(3) Functional indeterminacy [Dretske, 1986; Fodor, 1990a]. A further problem
arises for teleosemantics because there are different ways of describing the dispo-
sitions for which traits have been selected. One reason for this is that traits are
selected for a concertina of effects. They are not as a rule selected for simply
doingx, but for doingx^1 , which brings aboutx^2 , which brings aboutx^3 ,which
ultimately tends to contribute to, say, gene replication. Another reason is that
there are can be different but extensionally equivalent descriptions available for
the environmental features with which a cognitive system interacts. The much-
discussed case of the frog’s prey-capturing mechanism has been used to illustrate
both parts of this problem.
The frog’s perceptual system is adapted for responding to stimuli that are small,
dark, and moving past its retinas. In a normal frog, such stimuli produce a char-
acteristic neural event in its optic tectum (here,R) which in turn causes the frog
to turn to face its prey and try to catch it by approaching it, and snapping its
tongue or jaws. This in turn allows the frog to ingest food, digest it, and circulate
nutrients, which then promote the full range of bodily functions. Thus the frog’s
perceptual mechanism was selected for a complex causal role — for its disposition
to respond to stimuli that are small, dark and moving, but also for its disposition
to guide the frog’s snapping, help it to catch food, provide nutrients to cells, and
so on. On the simplifying assumption that, in the environment in which frogs
evolved, small, dark, moving dots were co-extensive with flies and frog food, we
can also see how different descriptions of co-extensive environmental features can
(^6) The fact that this definition does not unequivocally settle all possible questions of application
is not a reason to reject it. A good definition need only settle issues that are genuinely settled
by the concept being defined.
(^7) Perhaps teleosemantics would be in more serious trouble if it could be shown that some
representational simples have non existent intentional objects. Georges Rey [manuscript] argues
that this could be the case. He cites psychological theories (like those of [Marr, 1982], and
[Biederman, 1990]) that propose that perceptual processing employs “geons”, or volumetric shape
representations, e.g., of cones and cylinders. Suppose for example, that CIRCLE was a simple
of our visual system. Since there are no actual perfect circles, CIRCLE is an empty concept. It
is debatable whether we possess a visual simple of a circle that is as exacting as Rey’s argument
requires. But extant versions of teleosemantics imply, not only that do we not have such a simple,
but that we cannot have one.
I’m not sure how troubled the proponents of teleosemantics should be about this possibility. If
there really were a need to admit visual simples of Platonic ideals, perhaps accommodation could
be made. We would need to examine the details of the case, but neural components can, for
instance, be adapted for responding more or less enthusiastically to more or less good instances
of certain properties. Maybe there is room for actual selection history with actual non-ideal
instances to ground reference to ideal ones or approximations of ideal ones.