<p>Five experiments investigated evidence for a dual-process account of mindreading (Apperly, 2010). This account is motivated by two puzzles: First, why is it that three-year-olds fail standard false-belief tests when looking patterns infer that infants are sensitive to others’ false beliefs? Secondly, why is adult mindreading sometimes slow and effortful, and at other times fast and effortless? The seemingly contradictory observations may be explained by drawing upon two relatively distinct mindreading abilities: ‘Efficient’ processing supports precocious infant performances in non-verbal tasks and fast-paced social interaction in adults, while the later developing ‘flexible’ processing permits full blown understanding of beliefs and facilitates correct verbal responding in standard false-belief tests. Evidence for this theory can be sought by exploiting the idea that there are ‘signature limits’ to the type of information that can be efficiently processed. One conjecture is that representations underpinning efficient belief-tracking relate agents to objects, leading to the prediction that efficient processing cannot handle false-beliefs involving identity. Experiments 1 and 2 used a novel action-prediction paradigm to determine if adults’ reaction-time patterns differed between two false-belief tasks, one involving a standard change-of-location scenario, and one which also incorporated an identity component. The findings revealed equivalent flexible processing across both tasks. However, there were distinct reaction-time profiles between the tasks such that efficient belief-tracking was only observed in the change-of-location task. The absence of efficient processing in the task incorporating an identity component supports the conjecture that efficient belief-tracking is limited to relational, rather than propositional attitudes. A second conjecture is that representations underpinning efficient belief-tracking either do not specify agents’ locations or do not specify objects’ orientations. This leads to the prediction that efficient belief-tracking alone will not yield expectations about agents’ perspectives. In a novel object-detection paradigm, Experiments 3 to 5 tested the extent to which adults efficiently tracked the belief of a passive bystander in two closely-matched but conceptually distinct tasks. In a task involving homogenous objects, reaction times were involuntarily influenced by the presence of the bystander. By contrast, in a second task in which the object could be differently perceived depending on where the agent was located in relation to that object, the presence of the agent did not influence adults’ response times, supporting the second conjecture.</p>