This study investigates different modalities of social interaction and their effects on the progressiv mastery of spatial representations in children of age 5 to 6 years and 8 months. More precisely, children are confronted in four experimental situations with spatial transformations that are performed by an adult. These transformations are either 1) correct, 2) incorrect at a level superior to the subject’s initial one, 3) incorrect at an inferior level but still possessing a certain systematic structure, 4) completely absurd. In a control situation subjects work with the same material; however, this time there is no interaction with an adult. One finding is that models superior to the subject’s level can provoke progress. Another, perhaps even more important finding is that progress can also be provoked by inferior models to the subject’s initial level. This result was corroborated in a complementary experiment which further demonstrated that repetitiv interactions with an adult model proposing an inferior performance can stabilise the subject’s progress.