The Feedback-Related Negativity Signals Salience Prediction Errors, Not Reward Prediction Errors

Abstract

Modulations of the feedback-related negativity (FRN) event-related potential (ERP) have been suggested as a potential biomarker in psychopathology.A dominant theory about this signal contends that it reflects the operation of the neural system underlying reinforcement learning in humans. The theory suggests that this frontocentral negative deflection in the ERP 230 -270 ms after the delivery of a probabilistic reward expresses a prediction error signal derived from midbrain dopaminergic projections to the anterior cingulate cortex. We tested this theory by investigating whether FRN will also be observed for an inherently aversive outcome: physical pain. In another session, the outcome was monetary reward instead of pain. As predicted, unexpected reward omissions (a negative reward prediction error) yielded a more negative deflection relative to unexpected reward delivery. Surprisingly, unexpected pain omission (a positive reward prediction error) also yielded a negative deflection relative to unexpected pain delivery. Our data challenge the theory by showing that the FRN expresses aversive prediction errors with the same sign as reward prediction errors. Both FRNs were spatiotemporally and functionally equivalent. We suggest that FRN expresses salience prediction errors rather than reward prediction errors.

Principal components analysis of reward prediction errors in a reinforcement learning task

“…The importance of the second axiom should be clear when one considers that the activity of such components will be the same for large -RPEs and small +RPEs, rendering them unsuitable as RPE encoders. These salience components had latencies, durations and strengths strongly resembling Foti et al's (2011) PCA study, and also resembled the behaviour of ERP components shown in other recent studies and reviews (San Martin, 2012;Talmi et al, 2013).…”

“…The summed effect of these components is a waveform which appears sensitive only to 5 +RPE size. There is now mounting evidence of such salience encoding components in the same temporal interval as the FRN (Hauser et al, 2014;Talmi, Atkinson, & El-Deredy, 2013) and such components therefore stand to account entirely for the apparent preferential sensitivity of the FRN to +RPEs shown in Walsh and Anderson's review.…”

Spatial filters yield stable features for error-related potentials across conditions

“…Noticeably these signals, linked to cognitive monitoring processes, have also been reported to be rather stable across different recording days [11], and feedback characteristics [12], [9]. As a matter of fact, they are not strongly modulated by the stimulus presentation rate [13], although they may vary depending on factors such as the predictability of the stimulus [14] These studies typically compare the signal across different conditions without assessing the classification performance across different experimental conditions (i.e., generalisation across days or feedback presentation speeds). In these studies, both the pre-processing steps and the classifier parameters are specifically suited for a given experimental condition.…”

Evidence for a general performance‐monitoring system in the human brain

“…One possibility to reconcile this conflicting evidence follows from the suggestion that FRN may comprise two distinct components (Heydari & Holroyd, 2016; Holroyd, Pakzad‐Vaezi, & Krigolson, 2008); the expectancy component (N200 response) is thought to track surprising or unexpected task‐related information (Talmi, Atkinson, & El‐Deredy, 2013) while the valence component (reward positivity) is considered to index processing of reward information (Holroyd et al, 2008). In line with this model, one study reported two distinct spatiotemporal principal components contributing to the FRN.…”