An MEG signature corresponding to an axiomatic model of reward prediction error

Beyond peaks and troughs: Multiplexed performance monitoring signals in the EEG

With the discovery of event-related potentials elicited by errors more than 30 years ago, a new avenue of research on performance monitoring, cognitive control, and decision making emerged. Since then, the field has developed and expanded fulminantly. After a brief overview on the EEG correlates of performance monitoring, this article reviews recent advancements based on single-trial analyses using independent component analysis, multiple regression, and multivariate pattern classification. Given the close interconnection between performance monitoring and reinforcement learning, computational modeling and model-based EEG analyses have made a particularly strong impact. The reviewed findings demonstrate that error- and feedback-related EEG dynamics represent variables reflecting how performance-monitoring signals are weighted and transformed into an adaptation signal that guides future decisions and actions. The model-based single-trial analysis approach goes far beyond conventional peak-and-trough analyses of event-related potentials and enables testing mechanistic theories of performance monitoring, cognitive control, and decision making.

The effects of trauma on feedback processing: an MEG study

The cognitive impact of psychological trauma can manifest as a range of post-traumatic stress symptoms that are often attributed to impairments in learning from positive and negative outcomes, aka reinforcement learning. Research on the impact of trauma on reinforcement learning has mainly been inconclusive. This study aimed to circumscribe the impact of psychological trauma on reinforcement learning in the context of neural response in time and frequency domains. Two groups of participants were tested - those who had experienced psychological trauma and a control group who had not - while they performed a probabilistic classification task that dissociates learning from positive and negative feedback during a magnetoencephalography (MEG) examination. While the exposure to trauma did not exhibit any effects on learning accuracy or response time for positive or negative feedback, MEG cortical activity was modulated in response to positive feedback. In particular, the medial and lateral orbitofrontal cortices (mOFC and lOFC) exhibited increased activity, while the insular and supramarginal cortices showed decreased activity during positive feedback presentation. Furthermore, when receiving negative feedback, the trauma group displayed higher activity in the medial portion of the superior frontal cortex. The timing of these activity changes occurred between 160 and 600 ms post feedback presentation. Analysis of the time-frequency domain revealed heightened activity in theta and alpha frequency bands (4-10 Hz) in the lOFC in the trauma group. Moreover, dividing the two groups according to their learning performance, the activity for the non-learner subgroup was found to be lower in lOFC and higher in the supramarginal cortex. These differences were found in the trauma group only. The results highlight the localization and neural dynamics of feedback processing that could be affected by exposure to psychological trauma. This approach and associated findings provide a novel framework for understanding the cognitive correlates of psychological trauma in relation to neural dynamics in the space, time, and frequency domains. Subsequent work will focus on the stratification of cognitive and neural correlates as a function of various symptoms of psychological trauma. Clinically, the study findings and approach open the possibility for neuromodulation interventions that synchronize cognitive and psychological constructs for individualized treatment.

Non-invasive stimulation reveals ventromedial prefrontal cortex function in reward prediction and reward processing

Introduction

Studies suggest an involvement of the ventromedial prefrontal cortex (vmPFC) in reward prediction and processing, with reward-based learning relying on neural activity in response to unpredicted rewards or non-rewards (reward prediction error, RPE). Here, we investigated the causal role of the vmPFC in reward prediction, processing, and RPE signaling by transiently modulating vmPFC excitability using transcranial Direct Current Stimulation (tDCS).

Methods

Participants received excitatory or inhibitory tDCS of the vmPFC before completing a gambling task, in which cues signaled varying reward probabilities and symbols provided feedback on monetary gain or loss. We collected self-reported and evaluative data on reward prediction and processing. In addition, cue-locked and feedback-locked neural activity via magnetoencephalography (MEG) and pupil diameter using eye-tracking were recorded.

Results

Regarding reward prediction (cue-locked analysis), vmPFC excitation (versus inhibition) resulted in increased prefrontal activation preceding loss predictions, increased pupil dilations, and tentatively more optimistic reward predictions. Regarding reward processing (feedback-locked analysis), vmPFC excitation (versus inhibition) resulted in increased pleasantness, increased vmPFC activation, especially for unpredicted gains (i.e., gain RPEs), decreased perseveration in choice behavior after negative feedback, and increased pupil dilations.

Discussion

Our results support the pivotal role of the vmPFC in reward prediction and processing. Furthermore, they suggest that transient vmPFC excitation via tDCS induces a positive bias into the reward system that leads to enhanced anticipation and appraisal of positive outcomes and improves reward-based learning, as indicated by greater behavioral flexibility after losses and unpredicted outcomes, which can be seen as an improved reaction to the received feedback.

Magnetoencephalographic Correlates of Mood and Reward Dynamics in Human Adolescents

Despite its omnipresence in everyday interactions and its importance for mental health, mood and its neuronal underpinnings are poorly understood. Computational models can help identify parameters affecting self-reported mood during mood induction tasks. Here, we test if computationally modeled dynamics of self-reported mood during monetary gambling can be used to identify trial-by-trial variations in neuronal activity. To this end, we shifted mood in healthy (N = 24) and depressed (N = 30) adolescents by delivering individually tailored reward prediction errors while recording magnetoencephalography (MEG) data. Following a pre-registered analysis, we hypothesize that the expectation component of mood would be predictive of beta-gamma oscillatory power (25–40 Hz). We also hypothesize that trial variations in the source localized responses to reward feedback would be predicted by mood and by its reward prediction error component. Through our multilevel statistical analysis, we found confirmatory evidence that beta-gamma power is positively related to reward expectation during mood shifts, with localized sources in the posterior cingulate cortex. We also confirmed reward prediction error to be predictive of trial-level variations in the response of the paracentral lobule. To our knowledge, this is the first study to harness computational models of mood to relate mood fluctuations to variations in neural oscillations with MEG.

Feedback-Related Negativity and Frontal Midline Theta Reflect Dissociable Processing of Reinforcement

Prediction errors (PEs) encode representations of rewarding and aversive experiences and are critical to reinforcement processing. The feedback-related negativity (FRN), a component of the event-related potential (ERP) that is sensitive to valenced feedback, is believed to reflect PE signals. Reinforcement is also studied using frontal midline theta (FMΘ) activity, which peaks around the same time as the FRN and increases in response to unexpected events compared to expected events. We recorded EEG while participants completed a monetary incentive delay (MID) task that included positive reinforcement and negative reinforcement conditions with multiple levels of the outcome, as well as control conditions that had no reinforcement value. Despite the overlap of FRN and FMΘ, these measures indexed dissociable cognitive processing. The FRN was sensitive to errors in both positive and negative reinforcement but not in control conditions, while frontal theta instead was sensitive to outcomes in positive reinforcement and control conditions, but not in negative reinforcement conditions. The FRN was sensitive to the point level of feedback in both positive and negative reinforcement, while FMΘ was not influenced by the feedback point level. Results are consistent with recent results indicating that the FRN is influenced by unsigned PEs (i.e., a salience signal). In contrast, we suggest that our findings for frontal theta are consistent with hypotheses suggesting that the neural generators of FMΘ are sensitive to both negative cues and the need for control.

Cooperative update of beliefs and state-transition functions in human reinforcement learning

It is widely known that reinforcement learning systems in the brain contribute to learning via interactions with the environment. These systems are capable of solving multidimensional problems, in which some dimensions are relevant to a reward, while others are not. To solve these problems, computational models use Bayesian learning, a strategy supported by behavioral and neural evidence in human. Bayesian learning takes into account beliefs, which represent a learner’s confidence in a particular dimension being relevant to the reward. Beliefs are given as a posterior probability of the state-transition (reward) function that maps the optimal actions to the states in each dimension. However, when it comes to implementing this learning strategy, the order in which beliefs and state-transition functions update remains unclear. The present study investigates this update order using a trial-by-trial analysis of human behavior and electroencephalography signals during a task in which learners have to identify the reward-relevant dimension. Our behavioral and neural results reveal a cooperative update—within 300 ms after the outcome feedback, the state-transition functions are updated, followed by the beliefs for each dimension.

Agent-specific learning signals for self-other distinction during mentalising

Humans have a remarkable ability to simulate the minds of others. How the brain distinguishes between mental states attributed to self and mental states attributed to someone else is unknown. Here, we investigated how fundamental neural learning signals are selectively attributed to different agents. Specifically, we asked whether learning signals are encoded in agent-specific neural patterns or whether a self–other distinction depends on encoding agent identity separately from this learning signal. To examine this, we tasked subjects to learn continuously 2 models of the same environment, such that one was selectively attributed to self and the other was selectively attributed to another agent. Combining computational modelling with magnetoencephalography (MEG) enabled us to track neural representations of prediction errors (PEs) and beliefs attributed to self, and of simulated PEs and beliefs attributed to another agent. We found that the representational pattern of a PE reliably predicts the identity of the agent to whom the signal is attributed, consistent with a neural self–other distinction implemented via agent-specific learning signals. Strikingly, subjects exhibiting a weaker neural self–other distinction also had a reduced behavioural capacity for self–other distinction and displayed more marked subclinical psychopathological traits. The neural self–other distinction was also modulated by social context, evidenced in a significantly reduced decoding of agent identity in a nonsocial control task. Thus, we show that self–other distinction is realised through an encoding of agent identity intrinsic to fundamental learning signals. The observation that the fidelity of this encoding predicts psychopathological traits is of interest as a potential neurocomputational psychiatric biomarker.

In order for people to have meaningful social interactions, they need to infer each other’s beliefs. Converging evidence from humans and nonhuman primates suggests that a person’s brain can represent a second person’s beliefs by simulating that second person’s brain activity. However, it is not known how the outputs of those simulations are identified as ‘yours and not mine’. This ability to distinguish self from other is required for social cognition, and it may be impaired in mental health disorders with social cognitive deficits. We investigated self–other distinction in healthy adults learning about an environment both from their own point of view and the point of view of another person. We used computationally identified learning variables and then detected how these variables are represented by measuring magnetic fields in the brain. We found that the human brain can distinguish self from other by expressing these signals in dissociable activity patterns. Subjects who showed the largest difference between self signals and other signals were better at distinguishing self from other in the task and also showed fewer traits of mental health disorders.

Correlation of cue-locked FRN and feedback-locked FRN in the auditory monetary incentive delay task

Reflecting the discrepancy between received and predicted outcomes, the reward prediction error (RPE) plays an important role in learning in a dynamic environment. A number of studies suggested that the feedback-related negativity (FRN) component of an event-related potential, known to be associated with unexpected outcomes, encodes RPEs. While FRN was clearly shown to be sensitive to the probability of outcomes, the effect of outcome magnitude on FRN remains to be further clarified. In studies on the neural underpinnings of reward anticipation and outcome evaluation, a monetary incentive delay (MID) task proved to be particularly useful. We investigated whether feedback-locked FRN and cue-locked dN200 responses recorded during an auditory MID task were sensitive to the probability and magnitude of outcomes. The cue-locked dN200 is associated with the update of information about the magnitude of prospective outcomes. Overall, we showed that feedback-locked FRN was modulated by both the magnitude and the probability of outcomes during an auditory version of MID task, whereas no such effect was found for cue-locked dN200. Furthermore, the cue-locked dN200, which is associated with the update of information about the magnitude of prospective outcomes, correlated with the standard feedback-locked FRN, which is associated with a negative RPE. These results further expand our knowledge on the interplay between the processing of predictive cues that forecast future outcomes and the subsequent revision of these predictions during outcome delivery.

Local context influences memory for emotional stimuli but not electrophysiological markers of emotion-dependent attention

Emotional enhancement of free recall can be context dependent. It is readily observed when emotional and neutral scenes are encoded and recalled together in a “mixed” list, but diminishes when these scenes are encoded separately in “pure” lists. We examined the hypothesis that this effect is due to differences in allocation of attention to neutral stimuli according to whether they are presented in mixed or pure lists, especially when encoding is intentional. Using picture stimuli that were controlled for semantic relatedness, our results contradicted this hypothesis. The amplitude of well‐known electrophysiological markers of emotion‐related attention—the early posterior negativity (EPN), the late positive potential (LPP), and the slow wave (SW)—was higher for emotional stimuli. Crucially, the emotional modulation of these ERPs was insensitive to list context, observed equally in pure and mixed lists. Although list context did not modulate neural markers of emotion‐related attention, list context did modulate the effect of emotion on free recall. The apparent decoupling of the emotional effects on attention and memory, challenges existing hypotheses accounting for the emotional enhancement of memory. We close by discussing whether findings are more compatible with an alternative hypothesis, where the magnitude of emotional memory enhancement is, at least in part, a consequence of retrieval dynamics.

Electrophysiological correlates reflect the integration of model-based and model-free decision information

In this study, we investigated the interplay of habitual (model-free) and goal-directed (model-based) decision processes by using a two-stage Markov decision task in combination with event-related potentials (ERPs) and computational modeling. To manipulate the demands on model-based decision making, we applied two experimental conditions with different probabilities of transitioning from the first to the second stage of the task. As we expected, when the stage transitions were more predictable, participants showed greater model-based (planning) behavior. Consistent with this result, we found that stimulus-evoked parietal (P300) activity at the second stage of the task increased with the predictability of the state transitions. However, the parietal activity also reflected model-free information about the expected values of the stimuli, indicating that at this stage of the task both types of information are integrated to guide decision making. Outcome-related ERP components only reflected reward-related processes: Specifically, a medial prefrontal ERP component (the feedback-related negativity) was sensitive to negative outcomes, whereas a component that is elicited by reward (the feedback-related positivity) increased as a function of positive prediction errors. Taken together, our data indicate that stimulus-locked parietal activity reflects the integration of model-based and model-free information during decision making, whereas feedback-related medial prefrontal signals primarily reflect reward-related decision processes.

Temporal dissociation of salience and prediction error responses to appetitive and aversive taste

The feedback-related negativity (FRN), a frontocentral ERP occurring 200-350 ms after emotionally valued outcomes, has been posited as the neural correlate of reward prediction error, a key component of associative learning. Recent evidence challenged this interpretation and has led to the suggestion that this ERP expresses salience instead. Here, we distinguish between utility prediction error and salience by delivering or withholding hedonistically matched appetitive and aversive tastes, and measure ERPs to cues signaling each taste. We observed a typical FRN (computed as the loss-minus-gain difference wave) to appetitive taste, but a reverse FRN to aversive taste. When tested axiomatically, frontocentral ERPs showed a salience response across tastes, with a particularly early response to outcome delivery, supporting recent propositions of a fast, unsigned, and unspecific response to salient stimuli. ERPs also expressed aversive prediction error peaking at 285 ms, which conformed to the logic of an axiomatic model of prediction error. With stimuli that most resemble those used in animal models, we did not detect any frontocentral ERP signal for utility prediction error, in contrast with dominant views of the functional role of the FRN ERP. We link the animal and human literature and present a challenge for current perspectives on associative learning research using ERPs.

MEG Signatures of a Perceived Match or Mismatch between Individual and Group Opinions

Humans often adjust their opinions to the perceived opinions of others. Neural responses to a perceived match or mismatch between individual and group opinions have been investigated previously, but some findings are inconsistent. In this study, we used magnetoencephalographic source imaging to investigate further neural responses to the perceived opinions of others. We found that group opinions mismatching with individual opinions evoked responses in the anterior and posterior medial prefrontal cortices, as well as in the temporoparietal junction and ventromedial prefrontal cortex in the 220–320 and 380–530 ms time windows. Evoked responses were accompanied by an increase in the power of theta oscillations (4–8 Hz) over a number of frontal cortical sites. Group opinions matching with individual opinions evoked an increase in amplitude of beta oscillations (13–30 Hz) in the anterior cingulate and ventral medial prefrontal cortices. Based on these results, we argue that distinct valuation and performance-monitoring neural circuits in the medial cortices of the brain may monitor compliance of individual behavior to the perceived group norms.

The Feedback-related Negativity Codes Components of Abstract Inference during Reward-based Decision-making

Behavioral control is influenced not only by learning from the choices made and the rewards obtained but also by “what might have happened,” that is, inference about unchosen options and their fictive outcomes. Substantial progress has been made in understanding the neural signatures of direct learning from choices that are actually made and their associated rewards via reward prediction errors (RPEs). However, electrophysiological correlates of abstract inference in decision-making are less clear. One seminal theory suggests that the so-called feedback-related negativity (FRN), an ERP peaking 200–300 msec after a feedback stimulus at frontocentral sites of the scalp, codes RPEs. Hitherto, the FRN has been predominantly related to a so-called “model-free” RPE: The difference between the observed outcome and what had been expected. Here, by means of computational modeling of choice behavior, we show that individuals employ abstract, “double-update” inference on the task structure by concurrently tracking values of chosen stimuli (associated with observed outcomes) and unchosen stimuli (linked to fictive outcomes). In a parametric analysis, model-free RPEs as well as their modification because of abstract inference were regressed against single-trial FRN amplitudes. We demonstrate that components related to abstract inference uniquely explain variance in the FRN beyond model-free RPEs. These findings advance our understanding of the FRN and its role in behavioral adaptation. This might further the investigation of disturbed abstract inference, as proposed, for example, for psychiatric disorders, and its underlying neural correlates.

Reward positivity: Reward prediction error or salience prediction error?

The reward positivity is a component of the human ERP elicited by feedback stimuli in trial-and-error learning and guessing tasks. A prominent theory holds that the reward positivity reflects a reward prediction error signal that is sensitive to outcome valence, being larger for unexpected positive events relative to unexpected negative events (Holroyd & Coles, 2002). Although the theory has found substantial empirical support, most of these studies have utilized either monetary or performance feedback to test the hypothesis. However, in apparent contradiction to the theory, a recent study found that unexpected physical punishments also elicit the reward positivity (Talmi, Atkinson, & El-Deredy, 2013). The authors of this report argued that the reward positivity reflects a salience prediction error rather than a reward prediction error. To investigate this finding further, in the present study participants navigated a virtual T maze and received feedback on each trial under two conditions. In a reward condition, the feedback indicated that they would either receive a monetary reward or not and in a punishment condition the feedback indicated that they would receive a small shock or not. We found that the feedback stimuli elicited a typical reward positivity in the reward condition and an apparently delayed reward positivity in the punishment condition. Importantly, this signal was more positive to the stimuli that predicted the omission of a possible punishment relative to stimuli that predicted a forthcoming punishment, which is inconsistent with the salience hypothesis.

Reward-Related Neural Activity and Adolescent Antisocial Behavior in a Community Sample

Behavioral research has found evidence supporting reward dominance in adolescence with externalizing disorders, but findings from neuroimaging studies have been largely heterogeneous. We examined the Feedback-Related Negativity (FRN) and P3b in relation to self-reported externalizing behavior among 78 adolescents (11-18 yrs) during a monetary gambling task with concurrent high-density electroencephalogram. As expected, the P3b and the FRN demonstrated greater evoked activity to reward and punishment, respectively. Further, high externalizing behavior was associated with greater P3b difference and reduced FRN difference in response to reward and punishment, suggesting that externalizing behaviors may be associated with both reward dominance and reduced feedback-monitoring.

Does the processing of sensory and reward-prediction errors involve common neural resources? Evidence from a frontocentral negative potential modulated by movement execution errors

In humans, electrophysiological correlates of error processing have been extensively investigated in relation to decision-making theories. In particular, error-related ERPs have been most often studied using response selection tasks. In these tasks, involving very simple motor responses (e.g., button press), errors concern inappropriate action-selection only. However, EEG activity in relation to inaccurate movement-execution in more complex motor tasks has been much less examined. In the present study, we recorded EEG while volunteers performed reaching movements in a force-field created by a robotic device. Hand-path deviations were induced by interspersing catch trials in which the force condition was unpredictably altered. Our goal was twofold. First, we wanted to determine whether a frontocentral ERP was elicited by sensory-prediction errors, whose amplitude reflected the size of kinematic errors. Then, we explored whether common neural processes could be involved in the generation of this ERP and the feedback-related negativity (FRN), often assumed to reflect reward-prediction errors. We identified a frontocentral negativity whose amplitude was modulated by the size of the hand-path deviations induced by the unpredictable mechanical perturbations. This kinematic error-related ERP presented great similarities in terms of time course, topography, and potential source-location with the FRN recorded in the same experiment. These findings suggest that the processing of sensory-prediction errors and the processing of reward-prediction errors could involve a shared neural network.

Neural evidence for description dependent reward processing in the framing effect

Human decision making can be influenced by emotionally valenced contexts, known as the framing effect. We used event-related brain potentials to investigate how framing influences the encoding of reward. We found that the feedback related negativity (FRN), which indexes the “worse than expected” negative prediction error in the anterior cingulate cortex (ACC), was more negative for the negative frame than for the positive frame in the win domain. Consistent with previous findings that the FRN is not sensitive to “better than expected” positive prediction error, the FRN did not differentiate the positive and negative frame in the loss domain. Our results provide neural evidence that the description invariance principle which states that reward representation and decision making are not influenced by how options are presented is violated in the framing effect.

Neurophysiology of performance monitoring and adaptive behavior

Successful goal-directed behavior requires not only correct action selection, planning, and execution but also the ability to flexibly adapt behavior when performance problems occur or the environment changes. A prerequisite for determining the necessity, type, and magnitude of adjustments is to continuously monitor the course and outcome of one's actions. Feedback-control loops correcting deviations from intended states constitute a basic functional principle of adaptation at all levels of the nervous system. Here, we review the neurophysiology of evaluating action course and outcome with respect to their valence, i.e., reward and punishment, and initiating short- and long-term adaptations, learning, and decisions. Based on studies in humans and other mammals, we outline the physiological principles of performance monitoring and subsequent cognitive, motivational, autonomic, and behavioral adaptation and link them to the underlying neuroanatomy, neurochemistry, psychological theories, and computational models. We provide an overview of invasive and noninvasive systemic measures, such as electrophysiological, neuroimaging, and lesion data. We describe how a wide network of brain areas encompassing frontal cortices, basal ganglia, thalamus, and monoaminergic brain stem nuclei detects and evaluates deviations of actual from predicted states indicating changed action costs or outcomes. This information is used to learn and update stimulus and action values, guide action selection, and recruit adaptive mechanisms that compensate errors and optimize goal achievement.

Developmental changes in performance monitoring: how electrophysiological data can enhance our understanding of error and feedback processing in childhood and adolescence

Performance monitoring includes learning from errors and feedback and depends on the functioning of the mediofrontal cortex, especially the anterior cingulate cortex (ACC) and the mesencephalic dopamine system. Error and feedback monitoring develop during childhood until early adulthood and are important in a lot of learning situations. The aims of this article are twofold: First, to review the present literature on the development of performance monitoring, and second, to highlight how electrophysiological data can contribute to the understanding of error and feedback processing in childhood and adolescence.

Neural dynamics of reward probability coding: a Magnetoencephalographic study in humans

Prediction of future rewards and discrepancy between actual and expected outcomes (prediction error) are crucial signals for adaptive behavior. In humans, a number of fMRI studies demonstrated that reward probability modulates these two signals in a large brain network. Yet, the spatio-temporal dynamics underlying the neural coding of reward probability remains unknown. Here, using magnetoencephalography, we investigated the neural dynamics of prediction and reward prediction error computations while subjects learned to associate cues of slot machines with monetary rewards with different probabilities. We showed that event-related magnetic fields (ERFs) arising from the visual cortex coded the expected reward value 155 ms after the cue, demonstrating that reward value signals emerge early in the visual stream. Moreover, a prediction error was reflected in ERF peaking 300 ms after the rewarded outcome and showing decreasing amplitude with higher reward probability. This prediction error signal was generated in a network including the anterior and posterior cingulate cortex. These findings pinpoint the spatio-temporal characteristics underlying reward probability coding. Together, our results provide insights into the neural dynamics underlying the ability to learn probabilistic stimuli-reward contingencies.

Negative reward expectations in Borderline Personality Disorder patients: neurophysiological evidence

Borderline Personality Disorder (BPD) patients present profound disturbances in affect regulation and impulse control which could reflect a dysfunction in reward-related processes. The current study investigated these processes in a sample of 18 BPD patients and 18 matched healthy controls, using an event-related brain potentials methodology. Results revealed a reduction in the amplitude of the Feedback-Related Negativity of BPD patients, which is a neurophysiological index of the impact of negative feedback in reward-related tasks. This reduction, in the effect of negative feedback in BPD patients, was accompanied by a different behavioral pattern of risk choice compared to healthy participants. These findings confirm a dysfunctional reward system in BDP patients, which might compromise their capacity to build positive expectations of future rewards and decision making.

The feedback-related negativity signals salience prediction errors, not reward prediction errors

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.