Dopamine (DA) loss in Parkinson’s disease (PD) causes debilitating motor deficits. However, dopamine is also widely linked to reward prediction and learning, and the contribution of dopamine-dependent learning to movements that are impaired in PD—which often do not lead to explicit rewards—is unclear. Here, we used two distinct motor tasks to dissociate dopamine’s acute motoric effects vs. its long-lasting, learning-mediated effects. In dopamine-depleted mice, motor task performance gradually worsened with task exposure. Task experience was critical, as mice that remained in the home cage during the same period were relatively unimpaired when subsequently probed on the task. Repeated dopamine replacement treatments acutely rescued deficits and gradually induced long-term rescue that persisted despite treatment withdrawal. Surprisingly, both long-term rescue and parkinsonian performance decline were task specific, implicating dopamine-dependent learning. D1R activation potently induced acute rescue that gradually consolidated into long-term rescue. Conversely, reduced D2R activation potently induced parkinsonian decline. In dopamine-depleted mice, either D1R activation or D2R activation prevented parkinsonian decline, and both restored balanced activation of direct vs. indirect striatal pathways. These findings suggest that reinforcement and maintenance of movements—even movements not leading to explicit rewards—are fundamental functions of dopamine and provide potential mechanisms for the hitherto unexplained “long-duration response” by dopaminergic therapies in PD.

Learning critically drives parkinsonian motor deficits through imbalanced striatal pathway recruitment