In adults, a wide range of results argue that sensory-motor systems are involved during conceptualprocessing. Following a developmental approach, the dissertation asks whether the developmentof concepts might be embodied and deals with the influence of action on conceptual knowledge.Classical models of development suggest that conceptual knowledge develops from one mechanism.Nelson’s position argues for a derivation of concepts from the interaction children have in theirenvironment, while Quinn and Eimas rather suggest that first categories are built from visual similaritydetection. However, children variability in taking account different information when categorizingchallenges these models. We adopt a global, differential and interactionist approach to consider thatvariability in categorization might be explained by a distinction across domain (living vs. artifacts) butalso by a distinction across manipulability of objects. We further consider that manipulability of objects isreflected in the organization of taxonomic knowledge to propose that children conceptual processingare grounded in the sensorial and motor interactions they have with objects. In adults, different studiesshow that actions influences conceptual processing; these studies give support to embodiedcognition theories. We conducted seven studies in 5 – to 9- year- old children to assess the hypothesesof an embodied development of conceptual knowledge. Two studies test the hypothesis of sensorimotorsimulations during conceptual processing. The five following studies directly assessed theinfluence of action on concepts was assessed in five other studies. We contrasted the influence of graspand use. Grasp training, either with full hand or pinch, allow children to take into account volumetricinformation that is subsequently used as cue to categorize new objects. Use training, either push and rollor press, results in a weaker influence that differs with ages. Finally analyses of eye movement patternduring target identification among distractors that could be either grasped or pushed allow us todisentangle the influence of perceived grasp affordances from the influence of training by itself. From adevelopmental point of view; grasp affordances seem to be automatically detected by the age offive, and whatever the training condition. Information computed during training seems to be graduallytaken into account from seven to nine, with use information influence occurring later than graspinformation. Moreover, children performances might be modulated by the concordance or the discordance between the perception of grasp affordances and information from action training.
