Is it possible to perceive directly others’ mental states? Are mental states such as intentions visible in others’ behavior? Whereas abstract theoretical efforts have failed to resolve this debate, my aim in this discussion is to propose an empirically anchored approach to the problem. I will first briefly review the nature and strength of the empirical evidence for and against the observability of intentions and identify some of the pitfalls that might account for apparent contradictory findings. Then, drawing on recent and decisive studies, I will delineate an experimental strategy to overcome these pitfalls and provide a quantitative characterization of the ability to perceive intentions. I shall conclude by proposing an alternative conception of direct social perception founded on a definable and measurable relationship between the specific features of an observed behavior and the ability to read out mental states.