For several decades, scientists have held that newborn human infants can imitate adult behaviors like mouth opening, tongue protruding, and finger movements. This has been difficult to explain. In particular, it is not clear how newborn infants could know enough about their own bodies and movements, and how these map onto the bodies and movements of others, to be able to match their own felt movements to the seen movements of others. Prominent theories posit the existence of inborn mechanisms that automatically link visual input from the movements of others to the infant's already existing motor programs to produce the same movements. Such mechanisms would be instances of the inheritance of specific knowledge. In this review, we explain how current data argue against the existence of an inborn mechanism for linking seen actions with programs for performing those actions. We then sketch an alternative, multicausal explanation for the gradual development of imitation across infancy and beyond. WIREs Cogn Sci 2017, 8:e1410. doi: 10.1002/wcs.1410 For further resources related to this article, please visit the WIREs website.