Abstract : This thesis research aims to test the potential of music as a mnemonic support for new learning in normal elderly and Alzheimer's disease (AD) participants. Several studies have highlighted the beneficial effects of music on cognition in aging and dementia. At the same time, in young adults, the idea that music could serve as a mnemonic support is highly debated. Yet, very few studies addressed this question in aging or dementia. In the present work, we conduct two studies in a cohort of 8 mild Alzheimer's disease and 7 matched control participants. The first study concerns verbal memory, and compares learning and different retention delays of lyrics (unknown texts) that are either spoken or sung. When lyrics are sung, different degrees of melody familiarity are contrasted. Moreover, as motor activity is strongly related to music, we compare two learning procedures that are either synchronized or not with the production of these lyrics during encoding: 1) participants sing in unison with the model or 2) participants hear the model without singing. Results of this study are presented and discussed in the first two articles of the experimental section. Globally, music does not show aid for learning measured in immediate recall; we even observed a harmful effect when lyrics are sung on a non-familiar melody. But music helps long-term retention of lyrics, particularly for AD participants. Nevertheless, music does not clearly interact with learning procedure involving unison singing. The second study of the thesis investigates the learning of gesture sequences. Similarly to the first study, we explore influence of music versus silence as background accompaniment, and synchronized production versus observation of gestures during encoding. Results (article 3) showed again no interaction between background accompaniment and learning procedure, but different effects of each variable on both groups. Learning gestures with synchronized production is beneficial for normal controls, but harmful for AD participants. On the other hand, musical accompaniment led to greater benefit for AD participants than for controls. In the general discussion, we present the implications of these results for fundamental and clinical neuropsychology. We propose some recommendations to maximize these effects and make them relevant for therapeutic care.