Inhibitory processing in the false belief task: Two conjectures

Although it is well established that four-year-olds outperform three-year-olds on predicting behavior from false beliefs, this is only true when the false belief is coupled with a positive desire. Four-year-olds perform poorly in an otherwise standard false belief task when the protagonist’s desire is to avoid rather than to approach a target. We account for this by assuming that the attribution of a false belief involves inhibitory processing. We present two versions of an inhibition model of successful belief-desire reasoning. © Blackwell Publishers Ltd. 1998, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA. Developmental Science 1:2 pp 247–253 Address for correspondence: Professor Alan M. Leslie, Center for Cognitive Science, Psych. Bldg., Busch Campus, Rutgers University, Piscataway, NJ 08855, USA. E-mail: [email protected] Copyright © 2000. All rights reserved. task with two possible answers, A or B. A standard task demands a prediction of behavior from a belief-desire pair in which the belief is false and the desire positive. By hypothesis, recovering a false belief content intrinsically requires inhibition of a default. Whereas beliefs ought to be true, desires are not by default positive – the negative desire, not to burn one’s fingers, is a perfectly ordinary desire. Still, a need for inhibitory processing in order to identify the target of desire can be created extrinsically. Suppose that the agent’s desire is for whichever of targets A or B does not have property x. Under some circumstances, in order to identify which of A or B is the target of the agent’s negative desire, one might first have to identify the target which does have x, say, A, in order then to identify the NOT(x) target of the desire, i.e., B. Having attended to target A, the brain must subsequently disengage from A and shift attention to B. Our models assume that the disengagement and shifting requires inhibition of attention to A. For our models, it is ‘target shifting’ rather than negativity which is critical. In fact, though a false belief can be entirely positive, our models say that attributing a false belief requires target shifting. Suppose now that the subject is required to predict behavior from a false belief together with a ‘target shift’ desire. If we are correct, such a task will demand double inhibition. However, the two inhibitions cannot simply be summed to produce a stronger inhibition of the same target, because this will give the wrong answer (see below). Instead, the two inhibitions must interact so as to cancel each other out. We expect that inhibiting an inhibition will be hard, even though the fouryear­old subject can comfortably marshal a single inhibition.€ Testing the inhibition hypothesis Cassidy (1995) modified a standard false belief task (with positive desire) to one in which the agent negatively desires the object. In this task, the agent wants to look in whichever container the object is not. Cassidy’s four-year-olds all passed a standard false belief task. However, in the false belief with negative desire task, only 38% passed, a result usually associated with threeyear-olds. For Cassidy, this result was entirely unexpected but is predictable from the double inhibition hypothesis outlined above. We therefore needed to see if this result would replicate. A critical feature of our models is that in a false belief + negative desire task the two inhibitions will interact and not simply sum. Therefore, we reasoned that the difficulty of passing a true belief + negative desire task would not simply sum with the difficulty of passing a false belief + positive desire (i.e., standard) task to yield the difficulty of a false belief + negative desire task. Instead, there should be an interaction such that the difficulty of a task requiring double inhibition will be far greater than the sum of two tasks each requiring a single inhibition. Our study can be thought of as a 2 × 2 design with factors belief (true, false) × desire (positive, negative), though we did not actually test with a true belief + positive desire condition because we assumed this would be trivially simple for four-year-olds. We made passing the standard false belief + positive desire condition an inclusion criterion to ensure that all subjects could marshal the required single (false belief ) inhibition. We also asked subjects in the false belief + negative desire condition a standard Think question. It is important to notice that double inhibition is only required by a Prediction question. Only in predicting behavior are belief and desire considered together : only then will the two inhibitions interact. We predicted that subjects’ performance on the Think question would be significantly better than on the Prediction question. Second, we wanted to see if a second inhibition could be introduced other than by way of a negative desire. For this purpose, we included a Mixed-Up-Man scenario. This involves a character who has positive desires but who always acts in a way ‘opposite’ to his desire. The only way to predict what the MixedUp-Man will do is first to identify what a normal man would do, inhibit that outcome, and choose the alternative.