mutual cognition, such as mutual pretence, mutual jealousy, or mutual
knowledge. Suppose that Peter and Pauline are mutually aware of the
candle on the table between them – each can see the candle and can see that
the other sees it; and this is mutually known to them both. This is surely a
situation we can understand, and can begin to make predictions about.
For example, we can predict that neither willWnd it necessary to comment
on the existence of the candle to the other; that each would be surprised if
the otherdidmake such a comment; that each would be surprised if the
other werenotsurprised if they themselves should make such a comment;
and so on. But how can I simultaneously simulate Peter simulating Pau-
line, who is in turn simulating Peter simulating Pauline, and so on? Here if
I try to do it by ‘dropping down a level’, I just simulate Peter and simulate
Pauline – but that will not do at all, since it is crucial that I should simulate
their reasoning about the mental states of the other, and their reasoning
about what the other will reason about their own mental states.
It looks as if simulationists will have to concede that mind-reading of
mutual cognition is handled by some sort of body of general knowledge, in
addition to simulation. But then they may try to claim that this knowledge
is learned through simulation in the course of normal development, hence
preserving the view that simulation is at the core of our mind-reading
abilities. But this looks pretty implausible when one reXects that children
engage in, and understand, complex forms of mutual pretence, at least by
the age of four (Jarroldet al., 1994a). Since this is the age by which children
acquire a properly representational conception of the mind (and hence can
understand that beliefs can be false – see section 4 below), and since you
cannot understand mutual pretence without such a conception, there is a
real problem here for simulationism – namely, to explain how both com-
petencies can emerge so close together in development.
2.4 Cognitive penetrability?These three points against simulationism appear to us to be quite decisive;
and to be a good deal more convincing than a rather technical line of
argument which has been pressed by Stich and Nichols (1992, 1995;
Nicholset al., 1996). They dub their argument against simulationism ‘the
cognitive penetrability argument’, but this is a rather confusing label
(liable to be confused with another sort of penetrability, the sort which
modules do not aVord). So we prefer to think of it as ‘the theoretical
fallibility argument’. The thinking behind this line of argument is that if
there is some area in which people regularly tend to behave irrationally in
ways which are surprising to the folk, but where we consistently fail to
predict this – wrongly expecting them to behave rationally – then we are
88 Mind-reading