Brain Signatures of Human Skill Learning: From Single Movements to Movement Sequences
Abstract
Sequences of finger movements, such as making a cup of coffee or playing the piano, have a key role in our lives. An important neuroscientific question is how such movement sequences are represented in the brain. The central goal of this thesis was to investigate how different brain regions represent individual movements, and how these representations change when learning sequences of movements. To that end, we used 1) high-field functional magnetic resonance imaging (fMRI) to measure brain activation in humans while they produced finger movements on a keyboard-like device, and 2) advanced multivariate analyses to characterize the brain representations underlying the acquisition and control of finger movements. First, we examined the functional architecture of individual finger movements (Chapter 2). To dissociate sensory processing from movement, we designed an experiment including active finger movements and passive finger stimulation. We found that while the contralateral hemisphere represented individual fingers equally well during active movement and passive stimulation, the ipsilateral hemisphere represented fingers more clearly during active movement. Next, we assessed how brain representations for sequences of finger movements develop with learning (Chapters 3 and 4). Healthy volunteers were trained to execute a set of finger presses over five weeks and underwent repeated fMRI sessions. The results revealed widespread learning-related changes in premotor and parietal regions, including overall reduction in activation and a reorganization of how individual sequences are represented (Chapter 3). Contrary to previous research, none of these changes were observed in the primary motor cortex (M1). This distinction in learning between M1 and association regions was further supplemented by utilizing repetition suppression analysis and multivariate pattern analysis (Chapter 4). We demonstrate that M1 primarily represents the starting finger of the sequence, an effect which diminishes upon repeated sequence execution, but does not further represent sequence-specific features. Conversely, association regions reflect sequence identity and remain more stable across repetitions. Altogether, these studies revealed that M1 and association regions are differentially involved in execution and learning of movement sequences. The broad implication of this research is that premotor and parietal regions are likely fundamental to learning sequential skills, extending beyond finger movements.