This study uses conversation analysis (CA) to investigate the coupling of specific linguistic items with specific gestures in second language (L2) learning over time. In particular, we are interested in how gestures accompany learning of new vocabulary. CA-informed studies of gesture have previously shown the importance of embodiment in L2 use and situated learning, substantiating the idea that gestures are ingrained in the ongoing work to achieve and maintain intersubjectivity. We investigate how the relation between talk and related embodied activities develops over time and indicates learning processes. Specifically, we show how gesture-talk combinations that the co-participants deploy at a certain point in time to display understanding are reused at later occasions. Examples investigated here are the prepositions "under" and "across", and our data indicate that at least these specific linguistic constructions are deeply embodied in ways that change slowly over time around an iconic core, as the linguistic items are being learned, suggesting a strong link between L2 vocabulary learning and gestures.