Longitudinal spatial language intervention studies have shown that greater exposure to spatial language improves children's performance on spatial tasks. Can short naturalistic, spatial language interactions also evoke improved spatial performance? In this study, parents were asked to interact with their child at a block wall exhibit in a children's museum. Some parents were instructed to emphasize formal shape terms, others to emphasize spatial goals, and some were not provided scripts. Children were presented with a series of spatial reasoning tasks before and after this parental interaction, and the amount and type of spatial language during the training session was coded for parents and children. We found that (a) parents significantly increased their spatial language use when prompted, (b) children and parents used different types of spatial language in each of the scripted conditions, and
(c) children's spatial language during the interaction, and not parents', predicts children's subsequent improved puzzle performance.