Figure 8.

Summary figure. In the left column, a sample trial details that a fast response was chosen. Faster responses had higher probability distribution variance than slow responses; thus, there was more relative uncertainty for the chosen option, which was defined as an exploratory trial. Greater relative uncertainty correlated with peri-response theta power in mid-frontal and right frontopolar areas. Response characteristic slopes were taken from unthresholded TF-ROIs in Figure 7; conditions where participants selected the response with lower relative uncertainty (exploitation) were inverted from Figure 7c to demonstrate the continuous nature of the relationship between relative uncertainty of the selected response and theta power. This plot shows how individuals who used uncertainty to explore (positive ϵ) had significantly larger slopes only when exploring. The right column shows a sample feedback. Given the current estimation of action value for fast responses (25), this feedback (35) reflected a better-than-expected outcome (positive prediction error: +10). Prediction errors correlated with mid-frontal theta power. Response characteristic slopes were taken from unthresholded TF-ROIs from Figure 5; valenced prediction errors had similar slopes, yet negative prediction errors were characterized by a larger initial offset (greater power overall).