Figure 1. Experimental design and temporal characterization of separate outcome value systems.

(a) Schematic representation of the experimental paradigm. On each trial, two abstract symbols (selected from a larger set of three symbols) were presented for a maximum of 1.25 s. During this time, subjects had to select, by pressing one of two buttons, the symbol that was most likely to lead to a reward. Once a decision was made, a random delay was presented before the outcome was revealed. A tick and a cross were used to inform the participants of a positive (constant reward) and a negative (non-rewarding) outcome, respectively. Participants (n=20) performed 2 blocks of 170 trials each. (b) Multivariate discriminator performance (A_z) during positive versus negative outcome discrimination of outcome-locked EEG responses, averaged across subjects. Shaded error bars represent s.e. across subjects. The dotted line represents the average A_z value leading to a significance level of P=0.01, estimated using a bootstrap test. Two outcome value components (Early and Late) were revealed, with spatially distinct scalp topographies as estimated at time of maximum discrimination. (c) Single-trial discriminant component maps, for a representative subject. The four panels represent the discriminator amplitudes for the Early and Late components for positive and negative outcome trials using the training windows shown by the vertical white bars (solid: Early, dashed: Late). (d) Hypothetical value-related BOLD effects showing either greater overall BOLD signal for positive than negative outcomes and vice versa (red and blue curves, respectively). Three different BOLD predictors were used to model these effects: a conventional categorical regressor for positive versus negative outcomes and two parametric regressors modulated by the single-trial variability (STV) in the discriminator amplitudes of positive and negative outcomes in each of the Early and Late EEG components (extracted from subject-specific windows corresponding to the two components—solid and dashed windows, as seen in c).