Abstract: Habit learning involves gradual acquisition of associations between stimuli and responses, whereas declarative memory is concerned with episodes and facts. The experiments herein examine the core features that may distinguish the two types of knowledge and how habit learning develops and interacts with declarative memory. Three experiments examined the effects of performing a secondary task on habit learning and declarative knowledge. In experiment one, declarative knowledge, but not habit learning, was diminished after learning under dual task (DT) conditions. Learning under single task (ST) conditions allowed acquisition of more flexible information applicable outside the learning context, whereas learning under DT conditions did not transfer to a different context. Experiment two used functional magnetic resonance imaging (fMRI) to study the neural correlates of learning under ST and DT conditions. ST performance and the degree of declarative knowledge were tied to the MTL. DT conditions led to worse declarative knowledge and performance tied to the BG. Performance could be supported by distinct neural substrates and the characteristics of the knowledge supported by these substrates differed. Experiment three assessed retention of habit learning after ST and DT learning conditions. After one week performance of the task learned under ST conditions decreased more than one learned under DT conditions, suggesting that habit learning was less vulnerable to forgetting. The fourth experiment examined overtraining of a task using fMRI. Overtraining decreased neural responses to feedback information, but the behavioral response to changes in outcome contingencies remained flexible. A final experiment examined skill learning in patients with schizophrenia. Skills thought to depend on different parts of the BG were differentially affected in schizophrenia, which further supports the idea that the BG are functionally heterogeneous. Multiple types of knowledge may support performance on a task. Conditions of learning affect the neural substrates supporting performance and characteristics of the knowledge, such as whether it will be flexible and applicable outside the learning context or instead encapsulated in the learning context, but potentially more robust over time. Characterizing how different types of knowledge contribute to behavior in concert is essential to fully understanding learning and memory functions in humans.