We argue that imitation is a learning response to unintelligible actions, especially to social conventions. Various strands of evidence are converging on this conclusion, but further progress has been hampered by an outdated theory of perceptual experience.

Comparative psychology continues to be premised on the doctrine that humans and non-human primates only perceive others’ physical “surface behavior,” while mental states are perceptually inaccessible.

However, a growing consensus in social cognition research accepts the direct perception hypothesis: primarily we see what others aim to do; we do not infer it from their motions. Indeed, physical details are overlooked – unless the action is unintelligible. On this basis we hypothesize that apes’ propensity to copy the goal of an action, rather than its precise means, is largely dependent on its perceived intelligibility.

Conversely, children copy means more often than adults and apes because, uniquely, much adult human behavior is completely unintelligible to unenculturated observers due to the pervasiveness of arbitrary social conventions, as exemplified by customs, rituals, and languages.

We expect the propensity to imitate to be inversely correlated with the familiarity of cultural practices, as indexed by age and/or socio-cultural competence. The direct perception hypothesis thereby helps to parsimoniously explain the most important findings of imitation research, including children’s over-imitation and other species-typical and age-related variations.