During our daily life, we use a variety of gradually learned motor skills, such as riding a bike, playing guitar, or even speaking. Speech is a fundamental behavior acquired through sensorimotor learning during which a baby discovers the sets of action to execute for saying a new word through motor exploration and perception of her own actions. Previous research indicates that the basal ganglia (BG) are a likely candidate which could evaluate the discrepancy between expected and actual outcome and guide trial-and-error learning. While the BG drive behavioral adaptations that minimize errors early during learning, these adaptations are slowly incorporated in the downstream premotor cortical networks. It was proposed that the BG, through its subcortico-cortical loop, stabilize these adaptations by training premotor cortical areas. Here, we hypothesize that BG-cortical stabilization might occur during sleep through active processes, such as coordination of large neuronal assemblies between BG-cortical areas, which is a putative neurophysiological substrate allowing the long-term imprinting of recently acquired skills. To disentangle the BG-cortical dynamics during motor skill learning and the impact of sleep on these dynamics, we propose to use the songbird model and study the acquisition of their song learning behavior. Song learning is a natural form of sensorimotor learning akin to human speech acquisition during which a juvenile learns his song by imitating an adult bird (a tutor). Sleep in songbirds include mammalian-like features and is crucial for song learning. Moreover, songbirds have a set of interconnected brain nuclei dedicated to song, including a BG-thalamo-cortical loop, and that shares analogies with brain areas involved in human language. Songbirds thus represent an outstanding animal model for understanding the neurobiological bases of motor skill learning and the role of sleep. We hypothesize that sleep plays a critical role in the dynamical reorganization of the BG-cortical network and the consolidation of song. We expect specific events during the song learning stage (songs produced by the juveniles, exposure to the tutor song) to impact offline activity within the BG-cortical network. Offline activity, such as synchronized slow waves or replay bursts over the whole network could consolidate BG-driven adaptation in song in premotor nuclei. We thus expect important changes of coordinated activity within neuronal assemblies of the BG-cortical network during song learning. High-frequency oscillations, similar to mammalian ripples, occur during sleep in the songbird BG and neurons active during these oscillations are also active during singing. We thus hypothesize that offline oscillations contribute to song learning. We will perform longitudinal experiments, monitoring daily changes in the BG-cortical dynamics at the neuronal level and relate it with song acoustic features and song learning performance. We will perform large-scale simultaneous extracellular recordings across the BG and cortical nuclei. We will run experiments to establish cause-effect relationships between epochs of BG activity and song learning performance by chronically disturbing high-frequency oscillations in sleeping birds and assess their impact on birdsong learning We will thus reveal how synchronization between various neuronal populations involved in sensorimotor learning during sleep underlies consolidation in skill learning. The proposed research is at the interface between relevant basic biological issues for understanding sleep-related neurobiological processes at play during learning complex motor skills and pathological issues for assessing the impact both at the neuronal and behavioural level of the alterations of these processes.