Reaction time profiles of adults' action prediction reveal two mindreading systems.

Human beings are able to quickly step into others' shoes to predict peoples' actions. There is little consensus over how this cognitive feat might be accomplished. We tested the hypotheses that an efficient, but inflexible, mindreading system gives rise to appropriate reaction time facilitation in a standard unexpected transfer task, but not in a task involving an identity component. We created a new behavioural paradigm where adults had to quickly select whether an actor would reach, or not reach, for an object based on the actor's false belief about the object's location. By manipulating the type of object we compared participants' responding behaviour when they did and did not have to take the actor's perspective into account. While the overall accuracy reflected a high level of flexible belief reasoning across both tasks, the pattern of response times across conditions revealed a limit in the processing scope of an efficient mindreading system. Thus, we show, for the first time, that there are indeed different profiles of reaction times for object-location scenarios and for object-identity scenarios. The results elevate growing evidence that adult humans have not one, but two mindreading systems for dealing with mental states that underlie action.