Résumé : The 'learning and performance' conundrum has for a long time puzzled the field of cognitive neuroscience. Deciphering the genuine functional neuroanatomy of motor sequence learning, among that of other skills, has thereby been hampered. The main caveat is that changes in neural activity that inherently accompany task practice may not only reflect the learning process per se, but also the basic motor implementation of improved performance. Previous research has attempted to control for a performance confound in brain activity by adopting methodologies that prevent changes in performance. However, blocking the expression of performance is likely to distort the very nature of the motor sequence learning process, and may thus represent a major confound in itself. In the present study, we postulated that both learning-dependent plasticity mechanisms and learning-independent implementation processes are nested within the relationship that exists between performance and brain activity. Functional magnetic resonance imaging (fMRI) was used to map brain responses in healthy volunteers while they either (a) learned a novel sequence, (b) produced a highly automatized sequence or (c) executed non-sequential movements matched for speed frequency. In order to dissociate between qualitatively distinct, but intertwined, relationships between performance and neural activity, our analyses focused on correlations between variations in performance and brain activity, and how this relationship differs or shares commonalities between conditions. Results revealed that activity in the putamen and contralateral lobule VI of the cerebellum most strongly correlated with performance during learning per se, suggesting their key role in this process. By contrast, activity in a parallel cerebellar network, as well as in motor and premotor cortical areas, was modulated by performance during learning and during one or both control condition(s), suggesting the primary contribution of these areas in motor implementation, either as a function or not of the sequential content of movements. Our findings thus highlight the multifaceted nature of the link between performance and brain activity, and suggest that different components of the striato-cortical and cerebello-cortical motor loops play distinct, but complementary, roles during early motor sequence learning.