eagleshould mostly be a false alarm, triggered by any large birdXying
overhead. For the costs to the squirrels of taking cover under a tree are
small, whereas the gains, on those occasions when thereisan eagle ap-
proaching, are very large. Better to hide unnecessarily many times, than to
risk not hiding when you should, and be eaten. So in terms ofinformation
carriedit looks like the alarm call would just mean ‘big bird’. But when we
consider thefunctionof the alarm call in the lives of the squirrels, we see
that it has been selected for in virtue of those occasions, and only those
occasions, when it carries the information that an eagle is above. And then
this, according to teleo-semantics (and in accordance with intuition), is
what it means.
3.2 Two distinctionsIn developing her version of teleo-semantics, Millikan (1984) draws an
important distinction between theproducersandconsumersof mental
representations. And she claims that it is the consumers which are primary
when it comes to determining intentional contents. In the case of a visual
percept, for example, the producer system will be the visual module which
constructs that representation out of the information striking the retina.
And the consumer system will be the various practical reasoning and
action-control systems whichuse(or can use) that percept in the course of
their normal functioning. (SeeWgure 3.3, for example. Note that what is a
consumer system for one type of mental state can be a producer system for
another. An inferential system which generates beliefs from percepts is a
consumer system relative to the latter, but is itself a producer system for
the practical reasoning faculty.) Now the function of any given mental
state lies in its evolved eVects (the eVects it issupposedto have) on the
consumers of that state. So it is to the latter that we need to look inWxing
the content of the state in question.
It seems to us that Millikan has an important point here, which is
independent of teleological approaches as such, and which embodies a
signiWcant criticism of informational (or ‘indicator’) semantics. The point
is that the meaning of a sign,for a system, can only be the meaning which it
has for the processes within that system which consume, or make use of,
that sign. It is no good a sign carrying information about some worldly
state of aVairs if, so to speak, the rest of the system doesn’t know that it
does! This now holds out some hope of making progress with the dis-
junction problem. If we ask whymousedoes not meanshrew-or-mouse–
given that I frequently misrecognise shrews as mice, and so given that
mouseoften carries information about shrews – the answer can be that it is
because the rest of the system only operates in ways appropriate to mice,
Teleo-semantics 169