Abstract
How social contact is perceived as rewarding and subsequently modifies interactions is unclear. Dopamine (DA) from the ventral tegmental area (VTA) regulates sociality, but the ongoing, unstructured nature of free behavior makes it difficult to ascertain how. Here, we tracked the emergence of a repetitive stereotyped parental retrieval behavior and conclude that VTA DA neurons incrementally refine it by reinforcement learning (RL). Trial-by-trial performance was correlated with the history of DA neuron activity, but DA signals were inconsistent with VTA directly influencing the current trial. We manipulated the subject's expectation of imminent pup contact and show that DA signals convey reward prediction error, a fundamental component of RL. Finally, closed-loop optogenetic inactivation of DA neurons at the onset of pup contact dramatically slowed emergence of parental care. We conclude that this component of maternal behavior is shaped by an RL mechanism in which social contact itself is the primary reward.