Dissociable norepinephrine and dopamine representations during strategic decision-making and reward learning
Adaptive decision-making in competitive environments requires learning from immediate reward outcomes while tracking longer-timescale information to adjust behavioural strategies and remain unpredictable against others. Dopamine (DA) and norepinephrine (NE) have long been implicated in reward-guided behaviours, but their contributions to strategic choice remain unclear. We trained mice to play a zero-sum matching pennies game against a computer opponent that exploited regularities in their choice and reward history, such that optimal performance required stochastic choice decoupled from immediate reward. We combined fibre photometry recordings, behavioural and neural modelling, and a within-subject multi-task design to determine how NE and DA contribute to strategic choice in the matching pennies game versus reward-guided behaviour in a two-armed bandit (2AB) task. DA responses were dominated by trial-by-trial reward outcomes and consistently tracked model-derived reward prediction errors (RPE) across tasks and behavioural states. In contrast, NE activity reflected salient task events, predicted upcoming choice switches, and tracked longer-timescale variables including model-inferred behavioural strategy and average reward rate. In addition, NE responses were selectively enhanced following rewarded switches when these outcomes were informative for deciding whether to stay in or switch from the current strategy; and this enhancement could not be explained by RPE. Incorporating pre-choice NE activity into a reinforcement-learning model to modulate choice generation improved model prediction in matching pennies but not in the 2AB task. Together, these findings reveal complementary neuromodulatory correlates of competitive decision-making across timescales: DA primarily guides learning from immediate outcomes through RPE, whereas NE additionally represents contextual information and choice stochasticity relevant for shaping long-term strategies.