Reward Prediction Errors, Not Expectations Or Outcomes, Drive Emotional Valence

What role do expectations, outcomes, and their mismatches play in generating emotions? The current work re-examines this question to test a key premise of the value-updating theory of emotions, a new theory of emotions introduced here. Specifically, we test the idea that emotions are triggered by signals of the need for value-updating, which typically include reward prediction errors (RPEs) – outcomes being better or worse than expected – but not expectations or outcomes per se. While some past work has supported this idea, there are also conflicting results. For example, a recent study showed that affect in a perceptual decision-making task was influenced by both expectations and outcomes, but not by RPEs: i.e., positive affect was driven by expecting trials to be easier, and by trials actually being easier, but not by trials being easier than expected. The current work builds off this study and re-evaluates its conclusion by modifying the task and analyses to account for multiple forms of RPEs (trial-level and block-level) and their temporal dynamics. Results indicated that only RPEs drove affect in this task, with a negligible effect of expectations or outcomes per se, contrary to previous conclusions. This work supports the central role of RPEs in driving emotions, a core prediction of the value-updating theory of emotions. Moreover, linking emotions to value-updating has broad theoretical implications, such as highlighting how value-updating may be key to resolving emotions and how difficulties in this process may underlie many emotion-related disorders.