This thesis project focused on the evaluation of a memory overload hypothesis in analogical reasoning tasks. It is said that individuals with moderate intellectual disability (mental age: 4-7) are confronted to a memory overload when they have to treat several relations at the same time in analogical matrices. In order to test this assumption, we created a computerized test using a touch screen, the Revised - Construction of Analogical Matrices Test (CAM-R), composed of a construction version, supposed to unload the memory, and a control version, supposed to overload the memory. As expected, memory abilities played a crucial role in the classic version and less in the construction version. Our results strongly indicated that, with the support of external memories, participants with moderate intellectual disability obtained better performances than without such support, and were able to reach a similar level of performance as typically developing children of the same mental age.