Randomised prior feedback modulates neural signals of outcome monitoring

- FRN is sensitive to prior events but does this persist when learning is impossible?
- Four FRN theories were examined in a task with fully randomised feedback.
- Previous trial outcome modulated FRN during positive but not negative outcomes.
- Results consistent with a positive affective modulation interpretation of the FRN.
- FRN is determined by factors beyond prediction error (e.g. affective/motivational).

Substantial evidence indicates that decision outcomes are typically evaluated relative to expectations learned from relatively long sequences of previous outcomes. This mechanism is thought to play a key role in general learning and adaptation processes but relatively little is known about the determinants of outcome evaluation when the capacity to learn from series of prior events is difficult or impossible. To investigate this issue, we examined how the feedback-related negativity (FRN) is modulated by information briefly presented before outcome evaluation. The FRN is a brain potential time-locked to the delivery of decision feedback and it is widely thought to be sensitive to prior expectations. We conducted a multi-trial gambling task in which outcomes at each trial were fully randomised to minimise the capacity to learn from long sequences of prior outcomes. Event-related potentials for outcomes (Win/Loss) in the current trial (Outcomet) were separated according to the type of outcomes that occurred in the preceding two trials (Outcomet-1 and Outcomet-2). We found that FRN voltage was more positive during the processing of win feedback when it was preceded by wins at Outcomet-1 compared to win feedback preceded by losses at Outcomet-1. However, no influence of preceding outcomes was found on FRN activity relative to the processing of loss feedback. We also found no effects of Outcomet-2 on FRN amplitude relative to current feedback. Additional analyses indicated that this effect was largest for trials in which participants selected a decision different to the gamble chosen in the previous trial. These findings are inconsistent with models that solely relate the FRN to prediction error computation. Instead, our results suggest that if stable predictions about future events are weak or non-existent, then outcome processing can be determined by affective systems. More specifically, our results indicate that the FRN is likely to reflect the activity of positive affective systems in these contexts. Importantly, our findings indicate that a multifactorial explanation of the nature of the FRN is necessary and such an account must incorporate affective and motivational factors in outcome processing.