Affective valence does not reflect progress prediction errors in perceptual decisions.

Affective valence and intensity form the core of our emotional experiences. It has been proposed that affect reflects the prediction error between expected and actual states, such that better/worse-than-expected discrepancies result in positive/negative affect. However, whether the same principle applies to progress prediction errors remains unclear. We empirically and computationally evaluate the hypothesis that affect reflects the difference between expected and actual progress in forming a perceptual decision. We model affect within an evidence accumulation framework where actual progress is mapped onto the drift-rate parameter and expected progress onto an expected drift-rate parameter. Affect is computed as the difference between the expected and actual amount of accumulated evidence. We find that expected and actual progress both influence affect, but in an additive manner that does not align with a prediction error account. Our computational model reproduces both task behavior and affective ratings, suggesting that sequential sampling models provide a promising framework to model progress appraisals. These results show that although affect is sensitive to both expected and actual progress, it does not reflect the computation of a progress prediction error.

Do performance-monitoring related cortical potentials mediate fluency and difficulty effects on decision confidence?

It remains unclear to what extent the performance-monitoring (PM) processes involved in perceptual decisions are also involved in more complex reasoning decisions. To address this problem, we examined whether PM-related cortical potentials known to correlate with confidence ratings in the context of simple perceptual decisions also correlate with confidence ratings in a more complex reasoning task. In an EEG experiment, 49 participants had to quickly decide whether an equation (e.g. 9 * 7 = 65) was correct or incorrect and then report their confidence in that decision. Task difficulty and response fluency were varied to manipulate confidence. Pooling responses across correct as well as incorrect trials, we analyzed amplitudes of Error-Related Negativity (pooled ERN) and Error Positivity (pooled Pe). We found that pooled ERN but not pooled Pe had a positive significant correlation with decision confidence and that unpooled ERN correlated with confidence for correct, but not for error trials. Further analysis indicated that although pooled ERN correlated with task difficulty, neither pooled or unpooled potentials mediated the effects of difficulty or fluency on confidence. These results suggest that pooled ERN might reflect implicit performance monitoring and that confidence for simple perceptual and complex reasoning decisions may rely on different neural mechanisms.