Fig. 7.

Predicted reward during the delay period and responses to the go cue and to the reward delivery. (A) In trials with stimulus-present choices during the delay period the predicted reward increased with the stimulus amplitude in a graded manner. In FA trials (red line) its temporal profile was similar to that observed when a low-amplitude stimulus is perceived. (B) The predicted reward during the delay period was higher in trials where the Bayesian module detected a stimulus. It was higher in miss than in CR trials due to the detection of the stimulus when the amplitude was high. At the go cue, because of the reset mechanism, the reward predictions in the four trial types collapsed in approximately the same value and immediately after they separated into two values corresponding to the possible decisions. (C) The RPEs at the go cue were lower in stimulus-present decisions (hit and FA trials) than in stimulus-absent choices (miss and CR trials). According to the model this gradation was determined by the modulation of the reward prediction described in A and B and by the reset mechanism.