Mimicry has become a subject of great interest because of its ability to signal affiliation and rapport, unconscious nature, and seemingly strategic application. Studies of this phenomenon have focused on dyad members and inspired several explanations, but none has gained broad acceptance. This dissertation explores a neglected line of research, that of third parties’ impressions of mimics, and uses the evidence gained to reflect on mimicry’s true nature. The first chapter challenges the common assumption that mimicry necessarily signals positive characteristics, such as prosociality or trustworthiness. Rather, when mimicry is directed towards condescending models, it can reduce third party onlookers’ impressions of the mimic’s competence. The second chapter shows that observers do not take an act of mimicry as an indication of a general similarity between a mimic and his model, but make contextualized interpretations, dependent on the mimic’s knowledge of his model’s reputation and beliefs. Third party onlookers do, however, tend to conclude that a mimic shares his model’s opinion -- but only when the mimic could have known this opinion. This leads to arguments in chapter four that the empirical facts relating to mimicry can be explained parsimoniously by assuming that mimicry emerges as part of an implicit learning process. Before this, the third chapter investigates the basis of the complicated perceptions of mimicry shown in earlier chapters. Experiments here test the involvement of the mirror system in detecting rapport and mimicry, showing that nonverbal judgments are remarkably robust to both traditional cognitive load (counting backwards) and interference methods targeting the motor system. The latter methods are shown to place general demands on attention, so that interference with non-verbal judgments cannot be claimed to due to motoric demands, specifically.