Elsevier

Neuroscience Letters

Volume 526, Issue 1, 20 September 2012, Pages 5-9
Neuroscience Letters

Feedback-related negativity encodes outcome uncertainty in the gain domain but not in the loss domain

https://doi.org/10.1016/j.neulet.2012.08.017Get rights and content

Abstract

Outcome uncertainty plays an important role in decision making. The present study used event-related potentials (ERP) combined with a simple gambling task to investigate the time course of outcome uncertainty processing in the gain and the loss domains. Behaviorally, participants showed a preference for the uncertain option in the gain domain, whereas they showed no such preference in the loss domain. In ERP analyses, the feedback-related negativity (FRN) amplitudes mirrored the behavioral results: for gains, cue-elicited FRN amplitudes were more negative for uncertain than for certain cues, whereas for losses, cue-elicited FRN amplitudes for certain cues and uncertain cues were not different. In contrast, the P300 amplitude was sensitive to the valence of cues (gains or losses) but not to the uncertainty of the outcome. These results suggest that the outcome uncertainty is rapidly and differentially processed in the gain and loss domains.

Highlights

► We investigated the outcome uncertainty processing in the gain and loss domains. ► The cue-elicited FRN and the cue-elicited P300 were measured in a gambling task. ► The FRN was sensitive to the outcome uncertainty only in the gain domain. ► The P300 was sensitive to the valence of cues but not to the uncertainty of outcomes.

Introduction

Humans quickly evaluate external information and optimize their behavior in response to the constantly changing environment. Using event-related potentials (ERP), previous studies showed that a negative ERP component, known as the feedback error-related negativity (feedback-ERN) or feedback-related negativity (FRN), is more pronounced for negative feedback than for positive feedback [5], [8], [11], [17], [20], [31]. Evidence from source localization suggests that the FRN is generated in the anterior cingulate cortex (ACC)/medial prefrontal cortex (mPFC) [5], [11].

Holroyd and Coles [11] proposed a reinforcement learning theory of FRN (RL-FRN theory) and suggested that the FRN may reflect the reward prediction error signal. This signal may be encoded by the midbrain dopaminergic system and projected to the ACC, where it elicits the FRN and where it is used to reinforce the appropriate behavior. According to this theory, the FRN reflects the function of the evaluation system located in the ACC and should be sensitive to deviations from the expected event outcome. Consistent with this view, previous studies have observed that the amplitude of the FRN increases when the outcome is less predictable or more unexpected [3], [9], [16], [23], [28], and when the active choice is made [15].

Although most studies on FRN have focused on stimuli in the outcome phases, recent studies have shown that predictive cues in the anticipation phase can also elicit the FRN [4], [14], [33]. Predictive cues carry important information about potential outcomes, such as valence and reward probability. These cues have been extensively used in gambling tasks [9], [10], [11], [30], but the time course of cue processing is incompletely understood. An extension of the RL-FRN theory [4] would predict that the FRN is elicited by the first indication of reward prediction errors, regardless of whether this information is available in the anticipatory cue phase or the outcome phase. Using a gambling task in which the participants were presented with a gain cue indicating a monetary gain of 1 yuan or 0 yuan, each with 50% probability, or a loss cue representing a monetary loss of 0.5 yuan or 0 yuan, each with 50% probability, Yu and Zhou [33] reported that the loss cue elicited a more negative cue-elicited FRN (cue-FRN) than the gain cue. In another recent study, Dunning and Hajcak [4] found that cues indicating a 100% chance of losing elicited stronger cue-FRN than cues indicating a 100% chance of winning. More recently, Liao et al. [14] observed a FRN-like negative deflection for the difference wave between negative expectation cues (20% winning) and positive expectation cues (80% winning). Taken together, these findings suggest that the valence of the upcoming outcome is encoded by the cue-FRN.

However, the uncertainty of the outcome was not investigated in these previous studies. In contrast to reward probability, uncertainty refers to the variance of potential outcomes [24] regardless of valence. In previous studies, the uncertainty of the potential reward was the same in the gain domain and the loss domain. In the study by Yu and Zhou [33], positive cues and negative cues were both associated with the highest uncertainty because the outcomes were the most unpredictable (50% each). In a study by Dunning and Hajack [4], both cues were certain because the outcomes were 100% predictable. In a study by Liao et al. [14], both cues were associated with the same uncertainty (the outcome can be predicted with 80% accuracy), although the valence of the cues was different (one positive and the other negative).

The main purpose of the current study was to investigate the cue-FRN modulating effects of outcome uncertainty during the performance of a gambling task. A recent fMRI study found that ACC activity is modulated by the uncertainty of the reward environment during feedback monitoring [1], which suggests that the ACC may track the reward uncertainty. Given that the ACC is the most likely generator of the FRN, we predicted that the cue-FRN amplitude would also be modulated by the outcome uncertainty. In addition, some evidence suggests that the activity in human reward-sensitive brain regions is modulated by the context [12], [13], [19], [22]. A recent ERP study showed that the feedback-elicited FRN (feedback-FRN) amplitude is modulated by the reward probability in the gain but not the loss domain [3]. Thus, we predicted that the cue-FRN amplitude is also sensitive to outcome uncertainty in the gain but not in the loss domain.

Studies have suggested that the P300 amplitude is sensitive to the outcome magnitude [26], [32], the valence of reward [7], [31], and both outcome valence and magnitude [28]. Based on these findings, we predicted that the P300 amplitude would be more positive for cues in the gain domain than for those in the loss domain regardless of uncertainty.

Section snippets

Participants

Twenty-four participants (8 male) between the ages of 20 and 25 years (M = 22.92, SD = 1.52) participated in the paid study. All participants were right-handed and were screened for psychiatric and neurological problems. The study was approved by the Psychology Research Ethics Committee of South China Normal University. All participants were given written, informed consent and were informed of their right to withdraw from the study at any time.

Task and materials

As shown in Fig. 1, the outcome uncertainty (certain

Behavioral performance

All of the participants had more than 40 artifact-free trials in each cue condition, and these data were used for the rest of the computations. The performance-based bonus ranged from −8 to 18 yuan, and the participants received rewards of 22–44 yuan.

In the gain domain, the uncertain option was chosen at a rate of 60.59% (SD = 19.85%), which is significantly higher than what would be expected by chance [t(23) = 2.61, p = 0.016]. In the loss domain, there was no significant difference between the

Discussion

In the present study, the cue-FRN began at about 200 ms and peaked at about 300 ms after the presentation of cues, sharing similar onset and peak latencies as the classic feedback-FRN. However, the former had a broader scalp distribution than the latter. It is possible that the cue processing and the feedback processing involve distinct neural circuitry [4]. Further studies are needed to compare the generators of the cue-FRN and the feedback-FRN using source localization techniques.

The main aim

Conclusion

In summary, the present data indicate that the cue-FRN elicited by predictive cues preceding a decision is sensitive to the outcome uncertainty in the gain domain but not in the loss domain. Our findings suggest that the cue-FRN serves as a negative signal to indicate the aversiveness of uncertainty in the gain domain.

Acknowledgments

This study was supported by “Guangdong Province Universities and Colleges Pearl River Scholar Funded Scheme (GDUPS 2011)” and grant from the Guangdong Key Laboratory of Mental Health and Cognitive Science, South China Normal University. We thank Jie Li and two anonymous reviewers for their helpful comments and constructive critique.

References (34)

  • E. Snyder et al.

    Long-latency evoked potentials to irrelevant, deviant stimuli

    Behavioral Biology

    (1976)
  • Y. Wu et al.

    The P300 and reward valence, magnitude, and expectancy in outcome evaluation

    Brain Research

    (2009)
  • T.E. Behrens et al.

    Learning the value of information in an uncertain world

    Nature Neuroscience

    (2007)
  • C. Bellebaum et al.

    It was not MY fault: event-related brain potentials in active and observational learning from feedback

    Cerebral Cortex

    (2010)
  • J.P. Dunning et al.

    Error-related negativities elicited by monetary loss and cues that predict loss

    Neuroreport

    (2007)
  • W.J. Gehring et al.

    The medial frontal cortex and the rapid processing of monetary gains and losses

    Science

    (2002)
  • G. Hajcak et al.

    Brain potentials associated with expected and unexpected good and bad outcomes

    Psychophysiology

    (2005)
  • Cited by (9)

    • Medial frontal cortex response to unexpected motivationally salient outcomes

      2018, International Journal of Psychophysiology
      Citation Excerpt :

      First, the MFC may receive distinct reward prediction error and punishment prediction error signals from differential neural systems (e.g., different catecholamine systems; Daw et al., 2002). Consistent with previous studies that reported that probability/magnitude are only coded in the gain domain, thus reflecting RewP activity (Cohen et al., 2007; Deng et al., 2012; San Martin et al., 2010), it is possible the MFC receives a signed RPE signal from the dopamine reward system for feedback employing appetitive outcomes, while feedback employing aversive outcomes produces an oppositely signed punishment prediction error from a different neural system. The MFC has connections with a number of subcortical regions, including the spinothalamic system (Dum et al., 2009) and the lateral basal nucleus of the amygdala (Morecraft et al., 2007; Roy et al., 2009).

    • Randomised prior feedback modulates neural signals of outcome monitoring

      2016, NeuroImage
      Citation Excerpt :

      The Outcomet-1 × Outcomet-2 interaction approached significance (F [1, 26] = 3.61 p = .069, η2p = .122), but decomposing this revealed no statistically reliable effects (p values > .12). These findings are consistent with previous research, as the majority of existing studies report a larger P3 peak for positive compared to negative feedback (Deng et al., 2012; Zhou et al., 2010), although this effect can reverse in environments in which losses are more infrequent than gains (Bland and Schaefer, 2011; Mushtaq et al., 2013). Whilst the pattern of the FRN did not interact with the type of choice selected (i.e. Risk vs. Safe; see Mushtaq et al. Data in Brief, Section 6), reward expectancy might have been contingent on response outcome combinations of previous trials,2 and in particular if participants on any given trial chose the same type of gamble as in the previous trial (“Stay” response), or a different type of gamble (“Switch” response).

    • Larger N2 and smaller early contingent negative variation during the processing of uncertainty about future emotional events

      2014, International Journal of Psychophysiology
      Citation Excerpt :

      Using the aforementioned paradigm, some studies have emphasized the perception period; that is, they have explored the effects of the uncertain cues on the processing of the emotional events. In gambling, the outcome evaluations of winning or losing a wager elicited stronger brain activity (e.g., P2, FRN (feedback-related negativity) and P3) when preceded by uncertain cues than when preceded by certain cues (Deng et al., 2012; Xu et al., 2011). When viewing negative pictures preceded by uncertain cues rather than certain cues, greater unpleasant feelings were experienced, larger skin conductance responses was elicited (Grupe and Nitschke, 2011) and stronger brain activity in the insula and the amygdala was evoked (Sarinopoulos et al., 2010).

    View all citing articles on Scopus
    View full text