Thesis defense – Carmen Guerrero Marquez

Sensorimotor learning (SML) is essential for acquiring and mastering the wide variety of motor skills we use in daily life. Riding a bike, playing an instrument, singing and dancing are paradigmatic examples: such skills are never innate, but are refined through repetition and practice. This process depends on the precise coordination of movements by dedicated neural circuits, among which the basal ganglia–thalamo–cortical (BGT) loop plays a central role in motor learning across vertebrates. Reinforcement learning (RL) is thought to underlie this process, including the acquisition of human speech.

Songbirds are a powerful model for studying skill learning. Juveniles learn their song by trial-and-error imitation of an adult tutor, using a specialized song system composed of a BGT loop and a cortical motor pathway. Within this system, the lateral magnocellular nucleus of the anterior nidopallium (LMAN), the cortical output of the BGT loop, injects variability into song production through its projections to the motor pathway. RL models propose that LMAN also sends an efference copy of these variations to Area X, the basal ganglia nucleus of the song system, which is homologous to the mammalian striatum and pallidum. However, the functional organization of LMAN and Area X, and how they encode vocal variability, remain poorly understood.

To address this, we combined chronic and acute large-scale in vivo electrophysiology in zebra finches to examine functional connectivity and neural dynamics in LMAN and Area X during singing. Our results suggest that song variability is not encoded by single neurons, but emerges from coordinated population activity in both nuclei. We also identify topographically organized functional channels linking LMAN and Area X that may convey motor variability signals. Together, these findings shed light on how BGT circuits represent vocal features.