Is it possible to perceive others' mental states? Are mental states visible in others' behavior? In contrast to the traditional view that mental states are hidden and not directly accessible to perception, in recent years a phenomenologically-motivated account of social cognition has emerged: direct social perception. However, despite numerous published articles that both defend and critique direct perception, researchers have made little progress in articulating the conditions under which direct perception of others' mental states is possible. This paper proposes an empirically anchored approach to the observability of others' mentality – not just in the weak sense of discussing relevant empirical evidence for and against the phenomenon of interest, but also, and more specifically, in the stronger sense of identifying an experimental strategy for measuring the observability of mental states and articulating the conditions under which mental states are observable. We conclude this article by reframing the problem of direct perception in terms of establishing a definable and measurable relationship between movement features and perceived mental states.