According to a widely shared interpretation, research employing spontaneous-response false belief tasks demonstrates that infants as young as 15 months attribute (false) beliefs. In contrast with this conclusion, I advance an alternative reading of the empirical data. I argue that infants constantly form and update their expectations about others' behaviour and that this ability extends in the course of development to reflect an appreciation of what others can and cannot see. These basic capacities account for infants' performance in spontaneous-response false belief tasks without the need to assume the existence of a cognitive module specific for mental state attribution. My proposal suggests a plausible explanatory strategy for the problem of the representational format of the information processed in spontaneous-response false belief tasks.