This study investigates the role of striatal dopamine in mice learning to associate cues, actions, and outcomes in a task with implicit and changing rules. The researchers used fiber photometry to measure dopamine signals in the ventral striatum (VS), dorsomedial striatum (DMS), and dorsolateral striatum (DLS) as mice learned to integrate these associations. Key findings include: 1. **Learning and Behavioral Strategies**: Mice transitioned from a response bias strategy to one guided by the instruction cue, indicating they learned that outcomes were context-dependent and not externally determined. 2. **Dopamine Signals**: Cue-triggered dopamine signals did not increase with learning, while outcome-triggered signals decreased. This suggests that these signals were uncoupled when the animals had not yet formed a clear association between cues and outcomes. 3. **Rule Switches**: After rule changes, mice discarded learned associations and reset outcome expectations. Cue- and outcome-triggered dopamine signals became dissociated and dependent on the adopted behavioral strategy. 4. **Relearning**: As mice learned the new association, coupling between cue- and outcome-triggered dopamine signals and task performance re-emerged, indicating an increased understanding of the task. 5. **Temporal Difference Learning**: A temporal difference reinforcement learning (TDRL) model captured the behavioral and dopaminergic signatures, showing that rule switches led to adjustments in learned cue-action-outcome associations. 6. **Negative Prediction Errors**: Negative dopamine signals (deflections) in error trials did not scale with task performance, suggesting that behavioral errors had less impact on learning in expert animals. 7. **Subregional Differences**: The study found differences in dopamine signal dynamics and learning-related changes across striatal subregions, with the DLS showing more action value prediction errors and the VS showing more Pavlovian-like encoding. Overall, the results suggest that dopaminergic reward prediction errors reflect the animals' perceived locus of control, highlighting the importance of goal-directed actions in learning and cognitive processes.This study investigates the role of striatal dopamine in mice learning to associate cues, actions, and outcomes in a task with implicit and changing rules. The researchers used fiber photometry to measure dopamine signals in the ventral striatum (VS), dorsomedial striatum (DMS), and dorsolateral striatum (DLS) as mice learned to integrate these associations. Key findings include: 1. **Learning and Behavioral Strategies**: Mice transitioned from a response bias strategy to one guided by the instruction cue, indicating they learned that outcomes were context-dependent and not externally determined. 2. **Dopamine Signals**: Cue-triggered dopamine signals did not increase with learning, while outcome-triggered signals decreased. This suggests that these signals were uncoupled when the animals had not yet formed a clear association between cues and outcomes. 3. **Rule Switches**: After rule changes, mice discarded learned associations and reset outcome expectations. Cue- and outcome-triggered dopamine signals became dissociated and dependent on the adopted behavioral strategy. 4. **Relearning**: As mice learned the new association, coupling between cue- and outcome-triggered dopamine signals and task performance re-emerged, indicating an increased understanding of the task. 5. **Temporal Difference Learning**: A temporal difference reinforcement learning (TDRL) model captured the behavioral and dopaminergic signatures, showing that rule switches led to adjustments in learned cue-action-outcome associations. 6. **Negative Prediction Errors**: Negative dopamine signals (deflections) in error trials did not scale with task performance, suggesting that behavioral errors had less impact on learning in expert animals. 7. **Subregional Differences**: The study found differences in dopamine signal dynamics and learning-related changes across striatal subregions, with the DLS showing more action value prediction errors and the VS showing more Pavlovian-like encoding. Overall, the results suggest that dopaminergic reward prediction errors reflect the animals' perceived locus of control, highlighting the importance of goal-directed actions in learning and cognitive processes.