Historically, research in the field of human neuroscience has focused on the cerebral cortex. In contrast, our understanding of the subcortex lags behind. The human subcortex comprises hundreds of unique structures, which together encompass approximately 25% of the brain’s volume. Studying these structures is hindered by technical difficulties: The subcortex lies deep in the brain, limiting the quality of neuroimaging data. In this thesis, ultra-high field 7 Tesla magnetic resonance imaging (MRI) and computational modelling approaches are used to study the structure and function of the human subcortex. Chapter 2 describes biophysical properties of 17 subcortical structures across the adult lifespan. Chapters 3 and 4 focus on the combination of reinforcement learning and evidence accumulation models to adequately describe behavior in instrumental learning tasks. Chapter 5 focuses on optimizing 7 Tesla fMRI for studying the human subcortex. Chapter 6 studies the role of 12 subcortical gray matter structures in decision making and reinforcement learning. Chapter 7 tests whether there are three functionally relevant subdivisions in the subthalamic nucleus in a decision-making task. Chapter 8 tests for evidence for three subdivisions in the human subthalamic nucleus using post-mortem specimens.