Dopamine waves as a mechanism for spatiotemporal credit assignment
Significant evidence supports the view that dopamine shapes reward-learning by encoding prediction errors. However, it is unknown whether dopamine decision-signals are tailored to the functional specialization of target regions. Here, we report a novel set of wave-like spatiotemporal activity-patterns in dopamine axons across the dorsal striatum. These waves switch between different activational motifs and organize dopamine transients into localized clusters within functionally related striatal subregions. These specific motifs are associated with distinct task contexts: At reward delivery, dopamine signals rapidly resynchronize into propagating waves with opponent directions depending on instrumental task contingencies. Moreover, dopamine dynamics during reward pursuit signal the extent to which mice have instrumental control and interact with reward waves to predict future behavioral adjustments. Our results are consistent with a computational architecture in which striatal dopamine signals are sculpted by inference about instrumental controllability and provide evidence for a spatiotemporally \"vectorized\" role of dopamine in credit assignment.