Mood dynamics are associated with learning and not choice
Updating predictions about which stimuli are associated with reward is an important aspect of adaptive behaviour believed to relate to prediction errors, the difference between experienced and predicted outcomes. Behavioural sensitivity to prediction errors flexibly adapts to environmental statistics. Prediction errors also influence affective states during risky choice tasks that do not require learning, but the relationship between emotions and adaptive behaviour is unknown. Here, using computational modelling we found that mood dynamics, like behaviour, are sensitive to learning-relevant model variables (i.e., probability prediction error). Unlike behaviour, mood dynamics are not sensitive to model variables that influence choice (i.e., expected value), and increasing volatility does not reduce how many trials influence affective state. Finally, depressive symptoms reduce overall mood more in volatile than stable environments. Our findings suggest that mood dynamics are selective for variables relevant to adaptive behaviour and suggest a greater role for mood in learning than choice.