This paper describes two lines of research exploring a hypothetical function of imagery in the context of imitative actions: the mediation between perceptual and motor processes. Both experimental approaches, a sequence learning task and a timing imitation task, demonstrate that engagement into imagery as a temporally distinct activity between observation and performance is not required for accurate imitation. Moreover, evidence is provided that generative processes can take place during event observation itself, thus making a separate recoding stage redundant. Nevertheless, in the absence of a visual display, imagery of a movement sequence exerted similar learning effects as physical and observational practice, and visual and motor imagery were found to be equally effective rehearsal strategies for maintenance of temporal information in short-term memory.