The electrocortical response to rewarding and aversive feedback: The reward positivity does not reflect salience in simple gambling tasks
Introduction
For the past 20 years, ERP researchers have increasingly focused on the differentiation between positive and negative feedback to understand reward processing and learning (Miltner et al., 1997, Krigolson, n.d). Across time estimation (Miltner et al., 1997, Becker et al., 2014), reinforcement learning (Baker and Holroyd, 2008, Holroyd et al., 2011), and simple gambling tasks (Gehring and Willoughby, 2002, Holroyd et al., 2004, Holroyd et al., 2006, Proudfit, 2015), studies have consistently observed a relative negativity that peaks approximately 300 ms following feedback indicating bad compared to good outcomes. This relative negativity has been referred to as the feedback error-related negativity (Miltner et al., 1997, Holroyd and Coles, 2002, Holroyd et al., 2006, Nieuwenhuis et al., 2004), feedback negativity (Yeung and Sanfey, 2004), feedback related negativity (Cohen et al., 2007, Hajcak et al., 2006, Liu et al., 2014), and the medial frontal negativity (Gehring and Willoughby, 2002). More recent accounts suggest that this negativity may be a N200 to unexpected events that require increased need for cognitive control (Holroyd, 2004, Holroyd et al., 2008), and that this N200 is suppressed by a reward-sensitive positivity on reward trials (Holroyd et al., 2008). When conceptualized as a relative positivity following reward, several authors have suggested naming the ERP accordingly, either as the feedback correct-related positivity or the reward positivity (RewP; Holroyd et al., 2008, Proudfit, 2015).
Several lines of evidence suggest rewards drive the ERP difference between positive and negative feedback, including experimental manipulations (Holroyd et al., 2006, Holroyd et al., 2008, Kujawa et al., 2013), principal components analysis (PCA) of the ERP waveform (Foti et al., 2011, Weinberg et al., 2014, Liu et al., 2014, Carlson et al., 2011), and correspondence of the RewP to both reward-related behavioral (Bress and Hajcak, 2013) and neural measures derived from fMRI (Carlson et al., 2011, Becker et al., 2014, Foti et al., 2014). Collectively, these data suggest a positive potentiation in the ERP following rewards that is reduced or absent on non-reward trials.
Functionally, the RewP is thought to reflect a reward prediction error signal, which codes whether outcomes are better or worse than expected (Holroyd and Coles, 2002, Holroyd et al., 2008, Walsh and Anderson, 2012, Sambrook and Goslin, 2015). Consistent with this view, the RewP is larger when rewards are unexpected (Holroyd et al., 2011) and larger in magnitude (Sambrook and Goslin, 2015). While there is much evidence to suggest that the RewP is a reward-related modulation of the ERP, recent studies have provided evidence for the possibility that the RewP instead reflects a salience prediction error (SPE) signal. That is, the RewP may instead differentiate high- from low-salience events, regardless of valence. In this view, rewards might elicit a RewP because reward is more salient than non-reward.
In particular, two studies have found more positive ERP responses to feedback indicating aversive outcomes relative to feedback signaling the omission of aversive outcomes (Soder and Potts, n.d, Talmi et al., 2013). In terms of their experimental design, both studies presented participants with an initial cue that induced expectations regarding the likelihood of the outcome on each trial; following this cue (S1), participants were presented with feedback (i.e., the S2) that indicated expected or unexpected reward, or with feedback that indicated an expected or unexpected punishment (i.e., electric shocks in Talmi et al., 2013; noise blasts in Soder & Potts, this issue). Both Talmi and colleagues, as well as Soder and Potts, found that the S2 indicating unexpected reward elicited a positivity in the waveform relative to unexpected non-reward; however, both studies also found that the S2 signaling unexpected punishment also elicited a positivity relative to unexpected punishment omission (Talmi et al., 2013, Soder and Potts, n.d). The notion that unexpected punishment would elicit a RewP is inconsistent with reward-related accounts and suggests instead that a RewP may be elicited by salient outcomes.
Heydari and Holroyd (2016) have reported competing findings from a study in which participants navigated a virtual T maze and received feedback in rewarding and aversive conditions. Feedback indicated absence or presence of monetary reward in the rewarding condition, and absence or presence of small shock in the aversive condition. They found the RewP to be more positive to feedback indicating receipt of monetary reward as compared to its omission, and to feedback indicating omission of shock relative to impending shock. Thus, this study utilized a similar paradigm to those from Talmi et al. (2013) and Soder & Potts (current issue) by employing rewarding and aversive conditions, however, their results demonstrated the RewP tracked feedback valence rather than salience.
In the studies from Talmi et al. (2013) and Soder and Potts (current issue), the S1–S2 design was used to induce expectations regarding outcomes. However, participants never made choices—there was no response requirement in either the Talmi et al., or Soder and Potts experiments. This is particularly relevant given the fact that experimental results suggest that the RewP is maximized by feedback that follows volitional choice (Walsh and Anderson, 2012, Yeung and Sanfey, 2004). Moreover, many studies that have examined the RewP do so in the context of simple guessing tasks in which reward and loss are equiprobable on each trial (Gehring and Willoughby, 2002, Holroyd et al., 2004, Holroyd et al., 2006, Proudfit, 2015).
The current study employed a simple guessing task and within-subject design to examine whether feedback that signaled impending shock or safety would elicit a RewP. Subjects were administered two identical versions of a guessing task: a monetary version in which choices led to either monetary gain or loss—and an aversive version in which choices led to either safety from shock or punishment with shock. In this way, we employed identical features as Talmi and colleagues and Soder and Potts, however, feedback followed participant choices and were equiprobable on each trial. Traditional and principal component analysis (PCA)-derived factors were analyzed to assess the impact of outcome on ERPs. If the more rewarding outcomes (i.e., monetary gain and safety from shock) elicit a relative positivity compared to non-rewarding outcomes (i.e., monetary loss and punishment), the data would support the role of the RewP in reward-related process. If more salient outcomes (i.e., monetary gain and punishment) elicit a positivity relative to less salient outcomes (i.e., monetary loss and safety from shock), the data would support the SPE model and sensitivity of the RewP to salient outcomes.
Section snippets
Participants
Forty-one undergraduates from the introduction to psychology subject pool at Stony Brook University participated for course credit. The sample was college-aged (M = 20 years, SD = 3.70), 65.8% female, and ethnically diverse, including 38.1% Caucasian, 33.3% Asian, 14.3% Black, and 4.8% Latino. Demographic information was obtained through an initial screening e-mail. Informed consent was obtained prior to participation and the research protocol was approved by the Institutional Review Board at Stony
RewP
A 2 (outcome: best outcome [gain/safety], worst outcome [loss/punishment]) × 2 (task: money, shock) repeated measures ANOVA on mean activity from 250 to 350 ms following feedback at Cz confirmed that the ERP was more positive following desirable outcomes (i.e., gain and safety feedback; M = 17.71, SD = 10.11) than undesirable outcomes (i.e., loss and punishment feedback; M = 13.19, SD = 11.13; F(1, 40) = 19.12, p < 0.001, ηp2 = 0.33). The main effect of outcome is depicted in the ERP waveforms in Fig. 1. There
Discussion
The current study examined traditional ERP and PCA-based scores in the time window of the RewP to feedback indicating monetary gains and losses, as well as to feedback indicating safety and punishment, to determine whether more rewarding or more salient outcomes elicit the RewP. Consistent with previous studies on the RewP, when examining both the traditional ERP- and PCA-based scores, monetary gains compared to losses were associated with a relative positivity that peaked around 300 ms at
References (39)
- et al.
Ventral striatal and medial prefrontal BOLD activation is correlated with reward-related electrocortical activity: a combined ERP and fMRI study
NeuroImage
(2011) - et al.
Reward expectation modulates feedback-related negativity and EEG spectra
NeuroImage
(2007) The ERP PCA toolkit: an open source program for advanced statistical analysis of event-related potential data
J. Neurosci. Methods
(2010)- et al.
Optimizing principal components analysis of event-related potentials: matrix type, factor loading weighting, extraction, and rotations
Clin. Neurophysiol.
(2005) - et al.
Reward dysfunction in major depression: multimodal neuroimaging evidence for refining the melancholic phenotype
NeuroImage
(2014) - et al.
A new method for off-line removal of ocular artifact
Electroencephalogr. Clin. Neurophysiol.
(1983) - et al.
The feedback-related negativity reflects the binary evaluation of good versus bad outcomes
Biol. Psychol.
(2006) - et al.
The good, the bad and the neutral: electrophysiological responses to feedback stimuli
Brain Res.
(2006) - et al.
Cognitive load impacts error evaluation within medial-frontal cortex
Brain Res.
(2012) - et al.
Reinforcement-related brain potentials from medial frontal cortex: origins and functional significance
Neurosci. Biobehav. Rev.
(2004)
Learning from experience: event-related potential correlates of reward processing, neural adaptation, and behavioral choice
Neurosci. Biobehav. Rev.
Show me the money: the impact of actual rewards and losses on the feedback negativity
Brain Cogn.
Which way do I go? Neural activation in response to feedback and spatial processing in a virtual T-maze
Cereb. Cortex
A single-trial estimation of the feedback-related negativity and its relation to BOLD responses in a time-estimation task
J. Neurosci.
Self-report and behavioral measures of reward sensitivity predict the feedback negativity
Psychophysiology
Evaluating two-step PCA of ERP data with geomin, infomax, oblimin, promax, and varimax rotations
Psychophysiology
Differentiating neural responses to emotional pictures: evidence from temporal-spatial PCA
Psychophysiology
Event-related potential activity in the basal ganglia differentiates rewards from nonrewards: temporospatial principal components analysis and source localization of the feedback negativity
Hum. Brain Mapp.
The medial frontal cortex and the rapid processing of monetary gains and losses
Science
Cited by (31)
Do food images as action outcomes evoke a reward positivity?
2021, Brain and CognitionDissociating the effect of reward uncertainty and timing uncertainty on neural indices of reward prediction errors: A reward positivity (RewP) event-related potential (ERP) study
2021, Biological PsychologyCitation Excerpt :However, due to recent evidence showing that the underlying ERP component is driven primarily by rewards and an absence of reward that appears as a negativity (Proudfit, 2015), the terminology RewP will be used throughout this paper. RewP amplitude is thought to be partially driven by the activity of cortical dopamine neurons (Meadows, Gable, Lohse, & Miller, 2016), and, as such, serves as a neurobiological index of reward prediction errors (Mulligan & Hajcak, 2018; Sambrook & Goslin, 2015). Along with indexing reward prediction errors, amplitude of the RewP varies based on multiple characteristics of feedback.
The aversion positivity: Mediofrontal cortical potentials reflect parametric aversive prediction errors and drive behavioral modification following negative reinforcement
2021, CortexCitation Excerpt :Prior experiments examining positive and negative reinforcement are also confounded by failure to match outcome modality. Mulligan and Hajcak (2018) compared reward/loss to punishment/omission, and Heydari and Holroyd (2016) and Talmi et al. (2013) compared reward/omission with punishment/omission. However, in these tasks, conditions differed in outcome modality (money vs shock) and reinforcement timing (money is paid at the end of the study, but shocks are delivered immediately), making it impossible to draw direct comparisons between positive and negative reinforcement.
Reward Processing Abnormalities and Promising New Directions for Understanding Suicide Vulnerability
2021, Biological Psychiatry: Cognitive Neuroscience and NeuroimagingReward-Related Neural Predictors and Mechanisms of Symptom Change in Cognitive Behavioral Therapy for Depressed Adolescent Girls
2021, Biological Psychiatry: Cognitive Neuroscience and NeuroimagingCitation Excerpt :Specifically, the RewP has been linked to activity within the mesocorticolimbic reward circuit (e.g., ventral striatum and medial prefrontal cortex) (12,13) and dorsal anterior cingulate cortex (17); conversely, the LPP has been associated with a more distributed set of cortical and subcortical regions linked with visual, attentional, and emotion processing, including occipital, parietal, inferotemporal, and lateral prefrontal regions, as well as the amygdala and insula (47–51). In addition, in contrast to the RewP, which reflects initial reactivity to the receipt of rewards [but see studies linking the RewP/FRN to unexpected outcomes or feedback indicating safety, e.g., (52)], the LPP reflects more sustained attention toward and engagement with emotional or motivationally salient content (and not specific to only rewards). Although this is speculative, depressed adolescents exhibiting more sustained neural engagement to rewarding or motivationally salient feedback may be relatively more likely to successfully engage in and benefit from cognitive and behavioral activities prescribed in CBT.