Abstract Mimetic schemas, unlike the popular cognitive linguistic notion of image schemas, have been characterized in earlier work as explicitly representational, bodily structures arising from imitation of culture-specific practical actions (Zlatev 2005, 2007a, 2007b). We performed an analysis of the gestures of three Swedish and three Thai children at the age of 18, 22 and 26 months in episodes of natural interaction with caregivers and siblings in order to analyze the hypothesis that iconic gestures emerge as mimetic schemas. In accordance with this hypothesis, we predicted that the children's first iconic gestures would be (a) intermediately specific, (b) culture-typical, (c) falling in a set of recurrent types, (d) predominantly enacted from a first-person perspective (1pp) rather than performed from a third-person perspective (3pp), with (e) 3pp gestures being more dependent on direct imitation than 1pp gestures and (f) more often co-occurring with speech. All specific predictions but the last were confirmed, and differences were found between the children's iconic gestures on the one side and their deictic and emblematic gestures on the other. Thus, the study both confirms earlier conjectures that mimetic schemas “ground” both gesture and speech and implies the need to qualify these proposals, limiting the link between mimetic schemas and gestures to the iconic category.