Striatal dopamine signals reflect perceived cue-action-outcome associations in mice. This study investigates how dopamine signals in the striatum of mice change as they learn cue-action-outcome associations based on implicit and changing task rules. Using fiber photometry, the researchers measured dopamine levels in the ventral striatum (VS), dorsomedial striatum (DMS), and dorsolateral striatum (DLS) while mice learned these associations. Reinforcement learning models of the behavioral and dopamine data showed that rule changes lead to adjustments of learned associations. After rule changes, mice discarded learned associations and reset outcome expectations, resulting in dissociations of cue- and outcome-triggered dopamine signals that depended on the adopted behavioral strategy. As the animals rediscovered the impact of their own actions, reward predictions for the different trial events were recoupled, indicating an increased understanding of the current task. The study found that dopamine signals triggered by the cue and outcome evolve differently during learning. In the VS, all events triggered large dopamine transients. In the DMS, responses to the instruction cue dominated responses to the outcome, whereas the opposite pattern was observed in the DLS. The results suggest that dopaminergic reward prediction errors reflect an agent's perceived locus of control. The study also found that negative dopamine RPEs were not inversely scaled with task performance but tended to diminish across learning. This is in contrast to parallel offsets between positive and negative dopamine RPEs, which reflect externally driven outcome expectations. The findings indicate that dopamine signals do not simply passively track reward rates when they are under the control of the animals' actions, in agreement with evolving belief states regarding the nature of the task. The study also showed that temporal difference learning models captured the main behavioral and dopaminergic signatures of the present task, suggesting that active learning could be well approximated by state-action value learning. The results highlight the role of dopamine in the acquisition of complex associations and the importance of internal representations in learning.Striatal dopamine signals reflect perceived cue-action-outcome associations in mice. This study investigates how dopamine signals in the striatum of mice change as they learn cue-action-outcome associations based on implicit and changing task rules. Using fiber photometry, the researchers measured dopamine levels in the ventral striatum (VS), dorsomedial striatum (DMS), and dorsolateral striatum (DLS) while mice learned these associations. Reinforcement learning models of the behavioral and dopamine data showed that rule changes lead to adjustments of learned associations. After rule changes, mice discarded learned associations and reset outcome expectations, resulting in dissociations of cue- and outcome-triggered dopamine signals that depended on the adopted behavioral strategy. As the animals rediscovered the impact of their own actions, reward predictions for the different trial events were recoupled, indicating an increased understanding of the current task. The study found that dopamine signals triggered by the cue and outcome evolve differently during learning. In the VS, all events triggered large dopamine transients. In the DMS, responses to the instruction cue dominated responses to the outcome, whereas the opposite pattern was observed in the DLS. The results suggest that dopaminergic reward prediction errors reflect an agent's perceived locus of control. The study also found that negative dopamine RPEs were not inversely scaled with task performance but tended to diminish across learning. This is in contrast to parallel offsets between positive and negative dopamine RPEs, which reflect externally driven outcome expectations. The findings indicate that dopamine signals do not simply passively track reward rates when they are under the control of the animals' actions, in agreement with evolving belief states regarding the nature of the task. The study also showed that temporal difference learning models captured the main behavioral and dopaminergic signatures of the present task, suggesting that active learning could be well approximated by state-action value learning. The results highlight the role of dopamine in the acquisition of complex associations and the importance of internal representations in learning.