‘pretend beliefs’ of another person into our own system(s) of inferential
processing, and then use the output of derived beliefs for attribution to the
other; and that we can similarly start from an assumed goal and work out
(on our own behalf, as it were) the steps necessary to achieve that goal,
again then attributing the various sub-goals to the other.
There are at least three powerful reasons why an advocate of theory-
theory should accept this much in the way of simulation:
(1) Anything likea comprehensive theory of thinkingwould be too large to
handle, especially as a sub-system within a theory-of-mind module
(this was Heal’s original point). Much of human inference isholisticin
character, at least to the extent that what inferences people will draw
from what will depend, very largely, upon their perceptions of what
other beliefs of theirs are relevant to the inferential steps in question.
(This was the point discussed in chapter 3, section 5.2. We return to the
issue of holism in chapter 7.) A comprehensive theory of thinking
would then have to be, at the same time, a theory ofrelevance. In order
to predict what someone will infer from some new belief of theirs,
using theory alone, I would have to know, not just what their other
beliefs are, but also which ones they will take to be relevant.
(2) We have to acknowledge, in any case, a capacity to process inferences
on the basis of suppositions, because that is what hypothetical or
counterfactual reasoning is. Much human reasoning can begin by
supposingsuch-and-such to be the case, and reasoning from there. It is
in this way that we can foresee the consequences of adopting a new
belief in advance of accepting it, or of a new plan of action in advance
of executing it.
(3) People’s own inferential capacities impose limits on the inferences they
can assign to others. Thus, in the case of belief-attribution we can
hardly allow a cognitive mismatch such that the agent cannot see what
to infer fromP, but can perfectly well handle an inference fromPon
someone else’s behalf. This is what we might call ‘the Watson con-
straint’: if Dr Watson cannot work out who the murderer must have
beenfor himself, then he cannot work it out on Holmes’ behalf, either.
But if our theory of mind really did embody a complete theory of
thinking, then it certainly ought to be possible, in principle, for people
topredictthoughts in others which they cannot arrive atin propria
personaby using their own theoretical or practical reasoning systems.
Theory-theory can quite happily make this much in the way of a conces-
sion to simulationism. What theory-theory shouldnotconcede is that
simulation is needed for anything other than inferential enrichment, going
from already attributed beliefs to further beliefs, or from already at-
90 Mind-reading