"Motor skills are acquired through sensorimotor learning and can be executed without awareness. Experimental evidence suggests that the acquisition and execution of motor skills rely on multiple neural networks. On one hand, the basal ganglia (BG) are necessary for the acquisition of a motor skill through sensorimotor learning, but not for its execution later on. On the other hand, premotor networks, such as cortical motor areas, are required for both the learning and execution of motor skills. The respective role of the BG and premotor networks in sensorimotor learning however remains unclear. We aim to deepen our knowledge of the neural mechanisms underlying the acquisition of motor skills and to understand the respective contribution of BG and premotor networks during this process. Our general hypothesis is that DA-dependent learning in the GB drive plastic vocal production during early phases of vocal learning, while extended practice allows cortical premotor networks to be “trained” by the BG output, and to become capable of driving vocal production independently of the BG. To test this hypothesis, we will set up a comprehensive theoretical model of the song system relying heavily on available physiological, anatomical and behavioral data. In this model, we will investigate possible mechanisms allowing the optimization of the vocal output through learning in the BG circuit followed by a transfer of the motor control to the premotor networks. Experimentally testable predictions of the model will be formulated in terms of activity changes in the BG and premotor networks. We will use innovative experimental techniques to test these predictions and collect further data about neural activity in the song system. In particular, we will record song-related neural activity and investigate changes in functional connectivity in the circuit in vivo in a paradigm of adult song learning."