In the diagram the ‘pretend’-belief/desire generator feeds inputs corres-
ponding to the beliefs and desires of the target of the simulation into the
mind-reader’s own decision-making system. From then on this system
operates just as it would have if the inputs had been the mind-reader’s own
beliefs and desires, rather than simulated beliefs and desires. The diVerence
which makes this a case of oV-line running comes in the output process.
Whereas processing of the mind-reader’s own beliefs and desires would
eventuate in the formation of intentions and thereafter instructions to the
motor-control systems leading to action, output of the processing resulting
from ‘pretend’ inputs is side-tracked into a behaviour-predicting/ex-
plaining system. So instead of action, the result is expectations as to how
others will behave, or attributions of intentions and other mental states to
them.
Note from this how simulationism is liable to collapse into a hybrid view
which also invokes theoretical knowledge. For in this scheme of processing
both the‘pretend’-belief/desire generatorand thebehaviour-predicting/ex-
plaining systemclearly have important functions to discharge. The more
those two parts of the processing system are given to do, the more it will
seem that they must employ some body of tacit knowledge – an implicit
theory, in fact.
2 Problems for simulationism
In section 4 we will be reviewing the evidence derived from developmental
studies. We maintain that theory-theory does a better job of explaining this
evidence than simulationism. But it should be acknowledged that current
developmental evidence does more in the way of testifying to the modular
and innately speciWed character of mind-reading than it does to discrimi-
nate between theory-theory and simulationism. ReXection on mature folk-
psychological capacities reveals, however, a number of reasons why
theory-theory should be preferred to simulationism, as we shall now see.
2.1 Simulation and explanationSimulation is a feed-forward process. By feeding in ‘pretend’ inputs one
can get an output or result. This seems quite acceptable if you want to
predictwhat people will infer, or decide, or do, or how they will react. But
it is quite unclear how such a process could generate interpretations or
explanations of why someone has done something. Consider the analogy
with a wind-tunnel, often used by simulationists themselves. By putting a
model of a structure in a wind-tunnel you canWnd out how it will react to
various forces. But it is no good putting in a model of a damaged structure
Problems for simulationism 83