Anger is associated with reward-related electrocortical activity: Evidence from the reward positivity

Transcription

1 Psychophysiology, 52 (2015), Wiley Periodicals, Inc. Printed in the USA. Copyright VC 2015 Society for Psychophysiological Research DOI: /psyp Anger is associated with reward-related electrocortical activity: Evidence from the reward positivity DOUGLAS J. ANGUS, a,b KEVIN KEMKES, b,c DENNIS J. L. G. SCHUTTER, d AND EDDIE HARMON-JONES b a School of Psychology, University of Sydney, Sydney, Australia b School of Psychology, University of New South Wales, Sydney, Australia c Graduate School of Life Sciences, Utrecht University, Utrecht, The Netherlands d Donders Institute for Brain, Cognition and Behaviour, Radboud University, Nijmegen, The Netherlands Abstract Previous research indicates that the reward positivity (RewP), an electrophysiological correlate of sensitivity and biases towards rewarding stimuli, is modulated by affective and motivational variables. Studies have provided evidence that states and traits associated with negative affect and reduced approach motivation are correlated with smaller RewP amplitudes. However, the possible confound of affective valence and motivational direction was not addressed in these studies. In the present study, we examined if anger, an emotion associated with negative affect and increased approach motivation, would affect RewP amplitude. We also investigated if RewP amplitude was related to the motivational properties of reward stimuli. One hundred male participants completed two emotion inductions intended to elicit feelings of either neutrality or anger. Each was followed by a simple gambling task, in which correct choices were followed by images of women in lingerie or swimwear. Although the RewP was elicited following each induction, there was no difference in amplitude between the neutral and anger induction. However, RewP amplitude was positively correlated with how much participants liked the reward stimuli, and this correlation was statistically larger following the anger induction. These results support a motivational interpretation for the differences in RewP amplitude reported in previous studies, suggesting that motivational direction and intensity, rather than affective valence, underlie differences in RewP amplitude. Moreover, the RewP appears to be affected by interactions between motivational state and the motivational value of reward stimuli. Descriptors: Motivation, ERPs, Affect, Emotion DJA was supported by an Australian Postgraduate Award. Portions of this work were funded by a grant from the Australian Research Council (DP ). Address correspondence to: Douglas Jozef Angus, Top South Badham 243, School of Psychology, University of Sydney, Sydney, NSW 2006, Australia. dang9080@uni.sydney.edu.au The receipt of rewards increases a variety of neural responses, and past research has revealed that positive affective traits and states that are directly associated with approach motivation further increase these responses. The present research was designed to test the counterintuitive possibility that anger, a negative affective state that is directly associated with approach motivation, would also further increase a neural response associated with the receipt of reward. Such a result would assist in understanding how different components of emotion influence reward-related neural responses. Many scientists view emotion as a complex multicomponent psychophysiological process (e.g., Izard, 2010; Lang, 1995). Many theories of emotion recognize motivational direction and affective valence as two important components of emotion. Motivational direction can be defined as the urge to go toward (approach) or away from (avoid) a stimulus (Harmon-Jones, Harmon-Jones, & Price, 2013), whereas affective valence (positive to negative) can be defined as the subjective evaluation of the feeling state (Harmon-Jones, Schmeichel, Mennitt, & Harmon-Jones, 2011). Often, affective valence and motivational direction are naturally confounded: most positive affective states (e.g., joy, interest) are associated with approach motivation, and most negative affective states (e.g., fear, disgust) are associated with avoidance motivation. Anger has been found to often violate this natural confound as anger is often considered as a negative feeling state that is associated with approach motivation Although anger is primarily associated with approach motivation, previous research has identified certain situations where the experience, expectation, and perception of anger coincides with neural activity and behaviors related to withdrawal motivation (Aarts et al., 2010; Zimmer-Gembeck & Nesdale, 2013; Zinner, Brodish, Devine, & Harmon-Jones, 2008). In these situations, feelings of anxiety are also experienced or expected alongside anger. For instance, greater relative right frontal asymmetry has been observed when participants are required to inhibit aggressive approach behaviors due to social pressures; in this situation, individuals report feeling angry and anxious (Zinner et al., 2008). In another study, although endorsement of withdrawal behaviors as a response to hypothetical social rejection was correlated with how angry participants reported they would feel, this relationship ceases to be significant when statistically controlling for participants expected anxiety (Zimmer-Gembeck & Nesdale, 2013). Research using photographs of facial expressions of anger suggests that these expressions may evoke approach, avoidance, or no motivational responses. The perception of facial expressions of anger is not equivalent to the experience of anger, and various trait and state variables influence emotive responses to these facial expressions as has been identified in past research (Aarts et al., 2010; van Honk et al., 2010). 1271

2 1272 D.J. Angus et al. The Reward Positivity The attainment of rewards is facilitated by monitoring the environment for features signaling the success or failure of behaviors engaged in its pursuit. In humans, an evaluative step of this process has been examined using the reward positivity (RewP), an ERP that occurs in response to positive feedback (Hajcak, Moser, Holroyd, & Simons, 2006) and is suppressed by negative feedback (Proudfit, 2014). That is, the RewP is largest when feedback indicates that an action resulted in reward, and appears to be actively suppressed and thus smallest when feedback indicates that an action resulted in loss (Hajcak et al., 2006) or absence of reward (Bellebaum, Polezzi, & Daum, 2010). While a growing literature links intra- and individual differences in the RewP to affective states and traits, previous research has focused on variables that are either associated with positive affect and increased approach motivation or with negative affect and decreased approach. This has left the precise effects of affect and motivation unclear. In the present study, we use anger an emotion associated with negative affect and increased approach motivation to separate the potential influence of affective valence and motivational direction on the RewP. The RewP is a positive-going deflection that begins approximately 200 ms and peaks 300 ms after feedback onset and is sensitive to binary information about reward. Thus, it has been proposed as a marker of sensitivity to the motivating and reinforcing properties of reward stimuli (Bress & Hajcak, 2013; Proudfit, 2014). The RewP is likely generated by multiple neuronal populations. Some research has localized the RewP to the anterior cingulate cortex (ACC) in EEG (Gehring & Willoughby, 2002) and simultaneous EEG/fMRI research (Hauser et al., 2014), while other research has localized the RewP to the ventral striatum and areas of the medial prefrontal cortex (mpfc; Becker, Nitsch, Miltner, & Straube, 2014; Carlson, Foti, Mujica-Parodi, Harmon-Jones, & Hajcak, 2011; Carlson, Foti, Harmon-Jones, & Proudfit, 2014). The RewP has been frequently quantified and referred to as the feedback-related negativity (FRN), feedback error-related negativity (fern), and medial-frontal negativity (MFN). However, Proudfit (2014) recently argued that the neural response to feedback as indexed by the RewP/FRN/fERN/MFN is best conceptualized as a positive deflection elicited by reward feedback against a baseline negative deflection elicited by nonreward/punishment feedback, rather than the inverse. For example, recent studies show that trials in which feedback signals losses or nonreward elicit relative negative deflections, while feedback signaling reward elicits the characteristic positive deflection (e.g., Holroyd, Hajcak, & Larsen, 2006; Kujawa, Smith, Luhmann, & Hajcak, 2013). Similarly, recent EEG/fMRI research shows that, while greater BOLD (blood-oxygen-level-dependent) activation in the ventral striatum, ACC, and mpfc are associated with larger overall RewP amplitudes, feedback-locked hemodynamic responses occurred in response to reward feedback and not loss feedback (Becker et al., 2014). Moreover, conceptualizing the RewP as intrinsically related to feedback about the probable receipt of reward (compared to punishment or nonreward) provides a more parsimonious account of its relationship to the individual differences in reward processing discussed below. In light of the evidence presented by Proudfit (2014), we refer to the response as the RewP. Reward Positivity and Affect/Motivation The RewP is influenced by affective and motivational variables. In particular, states and traits associated with negative withdrawal-related affect are correlated with smaller RewP responses. For example, inverse relationships between RewP amplitude and depressive symptomology have been found, with smaller RewP amplitudes observed in healthy populations at risk of developing depression (Foti & Hajcak, 2009; Foti, Kotov, Klein, & Hajcak, 2011). Similar associations have been found for trait anxiety when feedback is unambiguous (Gu, Ge, Jiang, & Luo, 2010; Gu, Huang, & Luo, 2010). An inverse relationship between RewP and scores on the negative affect schedule has also been found (Santesso et al., 2012). Conversely, traits associated with positive affect and approach motivation correlate with larger RewPs in gambling tasks. For instance, studies have found that extraversion, a construct linked to positive affect and approach motivation, is positively correlated with RewP amplitude (Cooper, Duke, Pickering, & Smillie, 2014; Smillie, Cooper, & Pickering, 2011). Larger RewP amplitudes have also been observed in individuals who score higher on Carver and White s (1994) Behavioral Activation System (BAS) scale (Lange, Leue, & Beauducel, 2012) and scales related to it (Bress & Hajcak, 2013). These results suggest that variables related to affect and motivation may be related to modulated RewP amplitudes. That is, elevated negative affect, decreased reward seeking, and/or decreased approach-related behavior relates to smaller RewPs, while positive affect, reward seeking, and approach motivational behaviors are related to larger RewPs. Furthermore, the correspondence between RewP amplitude and affect has been shown to be partially state dependent. In one study, Foti and Hajcak (2010) had adult participants undergo neutral or sadness mood inductions, followed by an equiprobable gambling task. Although RewP amplitude did not differ as a function of mood induction, participants self-reported state sadness immediately following the induction predicted RewP amplitude in the subsequent task even when trait sadness and depressive symptomology were statistically controlled. Results showed that a greater propensity to react with sadness during the experiment was associated with a smaller RewP (Foti & Hajcak., 2010). Subsequent work replicated this effect in adolescents at high risk of developing depression (Foti et al., 2011). These findings collectively suggest that the RewP is modulated by negative affect, and have been interpreted as reflecting a potential mechanism for lowered reward responsiveness in various psychopathologies. However, it is unknown if the effects described above involve dampening of reward sensitivity by negative affect, or state alterations of and individual differences in approach/reward motivation. The study of affective and motivational influences on the RewP has largely been concerned with states and/or traits in which positive and negative affect are confounded with approach and withdrawal motivation, respectively. Although these studies commonly find that negative affect and withdrawal motivation are associated with smaller RewP amplitudes, the coupling of affective valence and motivational direction has precluded the disentangling of these potentially separate affective and motivational influences. Thus, it is unclear if variation in RewP amplitude is caused by changes and/or differences in affective valence or motivational direction. Anger and Motivation In contrast to sadness and anxiety, the negative affect of anger is consistently associated with elevated approach and reward-related motivation in behavioral, subjective, and neurophysiological

3 The reward positivity and anger 1273 domains (Carver & Harmon-Jones, 2009; Harmon-Jones, 2007; Harmon-Jones, Harmon-Jones, Abramson, & Peterson, 2009; van Honk, Harmon-Jones, Morgan, & Schutter, 2010). The association between anger and approach motivation is bidirectional. In one direction, the evocation of anger typically but not exclusively occurs when the pursuit or acquisition of rewards are hindered, goal-directed actions are thwarted, or what ought to occur does not (Carver & Harmon-Jones, 2009; Frijda, 1986). In the other direction, the subjective feeling of anger typically increases individuals propensity to engage in behaviors that involve approaching and engaging with the source of the anger, or to seek out rewarding and appetitive stimuli unrelated to the source of the anger (see below, Ford et al., 2010). Thus, anger may permit a way of examining whether negative affect or approach motivation is associated with RewP amplitudes. First, trait and state anger are positively correlated with scores on the BAS scale (Carver, 2004; Harmon-Jones, 2003; Harmon- Jones & Peterson, 2008), a self-report measure of individual differences in the functioning of a reward-related approach motivation system (Carver & White, 1994). Although anger was not assessed in past RewP studies examining BAS, their findings hint at a possible relationship between approach-related motivation that which characterizes anger and RewP amplitude during gambling tasks. Second, evidence for a potential relation between anger and the RewP comes from studies examining the allocation of visual attention to reward stimuli. Ford et al. (2010) had participants undergo one of four emotion induction protocols, designed to elicit a neutral, fearful, excited, or angry feeling. Subsequent to the emotion induction, participants completed a simple visual attention task in which images rated as high in positive valence (reward) were presented alongside those rated as being high in negative valence (threat) or neutral in valence. Participants who underwent the anger or excitement emotion inductions spent relatively longer fixating on reward than threat images compared to those in the neutral condition. More specifically, the elicitation of anger or excitement was associated with greater gaze fixation to reward stimuli than threat, over and above that observed for participants who underwent a neutral induction. This pattern was reversed for participants who underwent the fear induction. More recently, Ford, Tamir, Gagnon, Taylor, and Brunye (2012) found that higher trait anger predicted longer fixation times to reward than to control or threat-related images. Present Study In the present study, we examined the influence of anger and reward motivation on the RewP. In particular, we focused on the effect of state anger, elicited by an extended version of the mood induction protocol used in a previous anger study (Ford et al., 2010). Participants completed neutral and anger emotion inductions, the order of which was counterbalanced across the sample. Following each mood induction, participants then completed a simple gambling task modeled on that used by Foti and Hajcak (2010). In the present version of this task, reward stimuli were images of women in lingerie or swimwear and nonreward stimuli were images of rocks. This methodological decision was based on much past research that has found that men respond with reward-related responses to images of women (e.g., Kampe, Frith, Dolan, & Frith, 2001; Kranz & Ishai, 2006; Poli, Sarlo, Bortoletto, Buodo, & Palomba, 2007; Sabatinelli, Bradley, Lang, Costa, & Versace, 2007; Walter et al., 2008). At the conclusion of the study, participants completed a questionnaire assessing their subjective liking of the reward and nonreward stimuli. We had three primary hypotheses regarding the relation between the RewP, anger, and reward motivation. First, if the RewP is largely sensitive to motivational factors, such as increased or decreased approach motivation, then its amplitude should be greater following an anger emotion induction than following a neutral emotion induction. On the other hand, if the RewP is sensitive to negative affect, then RewP amplitude should be smaller following the anger induction than following the neutral induction. Second, we predicted that the more likable that individuals evaluate reward stimuli, the larger the amplitude of the RewP. This prediction follows from the idea that likable stimuli are wanted more and result in greater RewP amplitudes (e.g., Jia, Zhang, Li, Feng, & Li, 2013). Finally, if anger increases the RewP, then it should increase the degree to which likable stimuli are wanted and thus increase the correlation between RewP amplitudes and liking. Participants Method One hundred and five male volunteers participated in exchange for course credit or monetary compensation. All participants were either first-year undergraduate students in psychology at the University of New South Wales (UNSW) (N 5 56) or recruited via a UNSW paid-participation website (N 5 49). When enrolling in the study, the men indicated that they were willing to view pictures of women in lingerie/swimwear. All participants received written information and provided informed written consent. The study was approved by the UNSW Human Research Ethics Advisory Panel for Psychology. Data from five participants were lost due to technical problems, leaving a sample of 100. Procedure On arrival, participants were provided with a brief overview of the study, as well as a consent form. Then, EEG electrodes were attached to the participant. Stereo headphones were placed over the ears. All materials were presented on a 23-inch full HD computer screen with a refresh rate of 100 Hz, and administered using Presentation software (Neurobehavioral Systems, Inc.). The experimenter explained the gambling task to the participant by guiding him through two practice trials. Once the participant understood the gambling task, the experimenter left the room and turned off the light. Participants first completed the Buss and Perry (1992) Aggression Questionnaire, followed by either the neutral or anger emotion induction, the order of which was counterbalanced across the sample. Once the first induction had finished, the first block of the gambling task began, followed by the remaining emotion induction, and the second block of the gambling task. At the conclusion of the second block of the gambling task, participants completed the post-task questions. After all questions were completed, EEG sensors were removed, and the participant was debriefed. Materials and Design Emotion inductions. Participants underwent two emotion induction procedures modeled on those used by Ford et al. (2010). Each induction began with an autobiographical recall task. Participants were asked to try and relive an event in their mind s eye and write about this event for 8 min.

4 1274 D.J. Angus et al. In the anger condition, participants were asked to recall and write about an emotional event from their past in which they felt very angry, while in the neutral condition they were asked to recall and write about an event in which they felt little of any emotion. After writing for 8 min, participants were asked to maintain body postures and facial expressions associated with their feelings for 1 min. In both conditions, participants were instructed to recreate the [angry, neutral] feeling to the best of your ability. At the same time, make your face and body express this [angry, neutral] feeling and hold it while you feel [angry, neutral] for 1 min. Lastly, participants listened to instrumental music intended to evoke feelings consistent with their writing condition. For the anger condition, this was Refuse/Resist by Apocalyptica (Tuppinen, 1998). For the neutral condition, this was Indecision by Meyer et al. (2000). Song durations were 3 min 12 s and 3 min 16 s, respectively. These autobiographical recall and music tasks have worked well in selectively manipulating target emotions in past published experiments investigating the effects of anger (Ford et al., 2010). In general, embodiment, recall, and music manipulations are effective methods of manipulating emotional state (Lench, Flores, & Bench., 2011). Due to time constraints and the fact that these manipulations have worked well in past published experiments (Ford et al., 2010, 2012; Lench, Flores, & Bench, 2011) as well as in pilot tests in our lab, we did not include self-report manipulation checks. However, it seems unlikely that the manipulation would be ineffective given past results using the same manipulation. Participants completed both mood inductions, and induction order was counterbalanced across the sample. Gambling task. Participants completed a modified version of a gambling task used in previous studies (Carlson et al., 2011; Foti & Hajcak, 2010). In this task, participants were instructed to choose which one of two doors had a reward behind it, pressing the left or right key on a Cedrus RB530 response pad (Cedrus Corporation, San Pedro, CA). Door stimuli remained on screen until a response was made, at which point they were replaced with a fixation cross presented for 500 ms. If the choice was correct, the fixation cross was replaced with a green upward arrow, serving as positive feedback stimuli and lasting for 2,000 ms. On incorrect trials, a red downward arrow was presented for 2,000 ms. Another fixation cross was then presented for 500 ms, followed by reward or nonreward stimuli. Reward stimuli were 60 unique color photos of attractive women wearing lingerie or swimsuits, while nonreward stimuli were 60 unique color photos of rocks or minerals (from Gable & Harmon-Jones, 2010a). All photos had identical dimensions and were each presented once during the course of the study. Each block of the gambling task consisted of 60 trials, with positive feedback and reward stimuli given on 50% of these, and feedback was completely predictive of the (non)reward stimulus. Feedback and reward order were pseudorandomized with no more than three repetitions of reward or nonreward trials presented during each block, and were identical for all participants. Although participants were informed that their choice would determine the outcome of each trial, feedback (reward or nonreward) was predetermined and independent of their response. Regardless of the preceding emotion induction, feedback and reward/nonreward image order was identical within each block. A total of two blocks were completed during the study, one after each emotion induction. Self-report questions. Prior to the emotion inductions and gambling task, participants completed the Buss Perry Aggression Questionnaire (Buss & Perry, 1992). The reliability of each the four subscales of the questionnaire ranged from acceptable to good (physical aggression, a 5.73; verbal aggression, a 5.60; anger, a 5.75; hostility, a 5.81). At the end of the experiment after both emotion inductions and gambling tasks had been completed, participants answered several questions. These consisted of a series of questions evaluating their liking of the reward and nonreward stimuli, as well as questions gauging belief in the validity of the gambling task. All questions were answered on a scale from 1 (notatall)to5(extremely). The reward-related questions were: How attractive did you find the pictures of women? (M , SD 5.72), How interested are you in viewing pictures of women in lingerie? (M , SD 5.82), How arousing did you find the pictures of women? (M , SD 5.81), and How attractive is this woman? The latter question was asked for each of the 60 reward pictures. The average score of these 60 questions was calculated and included as a single item into further analysis (M , SD 5.55). Two questions were asked about how attracted the participants were to pictures of rocks. These questions were: How attractive did you find the pictures of rocks? (M , SD 5.67) and How arousing did you find the pictures of rocks? (M , SD 5.67). Participants were also asked, "How much did you think your door choices affected whether you got to see pictures of women or rocks? (M , SD ) as a measure of perceived agency. A liking score for the reward stimuli was calculated by summing participants responses to the questions about the pictures of women (M , SD ), and a liking score for the nonreward stimuli was calculated by summing responses to the questions about the pictures of rocks (M , SD ). Reliability for each composite score was acceptable (reward: 4 items, a 5.761; nonreward: 2 items, a 5.669). To isolate liking of reward stimuli, the neutral composite score was subtracted from the reward composite score for each participant, producing a reward liking score (M , SD ). Physiological Recording and Data Reduction EEG was recorded at 64 locations according to the International EEG system, using Ag\AgCl-tipped electrodes. Data were amplified and sampled at 4000 Hz by a BioSemi Active Two system (BioSemi, Amsterdam, The Netherlands) and referenced to the common mode sense (CMS). The ground consisted of the CMS active and passive driven right leg (DRL) electrode that forms a feedback loop driving the participant s average potential as close as possible to the analog-to-digital converter (i.e., the amplifier zero ) reference voltage in the A/D box ( Electrodes were placed on the left and right earlobes (A1, A2) for offline rereferencing. The electrooculogram (EOG) was recorded from electrodes placed 1 cm lateral to the outer canthi of each eye, and from the sub- and supraorbital regions of the right eye. Electrophysiological data were recorded using Actiview (version 6, BioSemi). Raw data were processed offline in Brain Vision Analyzer software (version 2.0, Brain Vision, Gilching, Germany). EEG data were downsampled to 500 Hz to facilitate subsequent processing steps, rereferenced to the average of A1 and A2, and filtered (bandpass filter Hz, 50 Hz notch filter, 24 db/oct roll-off). Eye blinks with aberrant morphology (e.g., double blinks) and gross motor artifacts identified by visual inspection of raw data were removed to improve blink correction and artifact rejection procedures. Data were segmented for each trial, beginning 200 ms prior

5 The reward positivity and anger 1275 to and ending 800 ms subsequent to feedback onset, with the former serving as the baseline. Segmented data were DC detrended (Hennighausen, Heil, & R osler, 1993), corrected for ocular artifact (Gratton, Coles, & Donchin, 1983; Miller, Gratton, & Yee, 1988, without average substraction), and baseline corrected. Segments with channels containing physiological artifact were identified using an automated procedure that removed data with (a) voltage shifts of more than 50 lv between samples, and (b) a maximum voltage difference of more than 300 lv within a trial (as recommended by Foti & Hajcak, 2010). Segments containing button presses were removed to prevent motor-related artifact from distorting the RewP. The data of five subjects were excluded from analysis as they had fewer than 20 artifact-free trials for any feedback outcome within either the neutral or anger conditions. The remaining 95 participants had a minimum of 20 trials for each feedback outcome within each condition (neutral nonreward, 20 30, M ; neutral reward, 24 30, M ; anger nonreward, 21 30, M ; anger nonreward, 24 30, M ). To quantify RewPs, trials were first averaged within feedback outcome (e.g., reward or nonreward) and within condition (e.g., neutral or anger). The ERP to the feedback in each induction condition was quantified as the mean amplitude over FCz between 250 and 350 ms after feedback onset for reward and nonreward trials; the difference waves between reward and nonreward trials were also calculated (Horan, Foti, Hajcak, Wynn, & Green, 2012). This method produced six mean amplitudes: neutral reward, neutral nonreward, neutral RewP, anger reward, anger nonreward, and anger RewP (the RewP is the difference wave). Analyses of self-report and EEG data were primarily conducted using SPSS (version 20, IBM, Armonk, NY). Grubbs test (Grubbs, 1969) was conducted in GraphPad Prism version 6.0 (GraphPad, San Diego, CA) to detect outliers. One participant s ERPs were identified as being a significant outlier and were not included in subsequent analyses. Self-report data from one participant were lost due a recording software failure. Pairwise comparisons were Bonferroni corrected for multiple comparisons. RewP Results Reward versus nonreward. A 2 (Reward vs. nonreward) 3 2 (Condition: neutral vs. anger) 3 2 (Order: neutral first, then anger vs. anger first, then neutral) repeated measures analysis of variance (ANOVA) was conducted to determine if induction order affected the reward and nonreward ERP waveforms in each induction condition. Because many of the within-subject analyses of separate reward and nonreward waveforms are redundant with analyses of the RewP difference wave (e.g., Condition 3 Reward vs. nonreward), we only report analyses involving induction order. There was no significant effect of induction order on overall amplitudes, F(1,92) <.01, p 5.978, g p 2 <.01. There was, however, an Induction Order 3 Emotion Induction Condition interaction averaging over reward and nonreward trials, F(1,92) , p <.001, g p Pairwise comparisons indicated that, regardless of induction order, amplitudes were more positive (to both types of feedback) following the first emotion induction than following the second (both ps <.001). The three-way interaction between reward, induction condition, and induction order was not significant, F(1,92) 5.10, p 5.756, g p 2 <.01). Difference wave RewP. One-sample t tests indicated that the RewP was significantly different from zero in both neutral, t(94) , p <.001, d , and anger, t(94) , p <.001, d , conditions (Figure 1). A 2 (Condition: neutral vs. anger) 3 2 (Order) repeated measures ANOVA was conducted to compare the RewP amplitudes in each induction condition. This revealed that there was no significant main effect of condition, F(1,92) , p 5.398, g p 2 <.01; induction order, F(1,92) , p , g p 2 <.00, and no significant interaction between the two, F(1,92) , p , g p 2 <.01. Thus, while the RewP was successfully elicited, its amplitude did not differ as a function of emotion induction, or the order in which inductions were delivered. Correlations Between ERP and Self-Report Variables First, the relations between the RewP amplitude in each condition and self-reports was examined using a series of Pearson correlations. 2 As shown in Table 1, within the neutral condition, reward liking scores were positively correlated with reward trial ERP amplitude, and perceived agency was positively correlated with the RewP difference wave amplitude. Within the neutral condition, reward liking was not significantly correlated with RewP, r 5.08, p In the anger condition, reward liking scores were positively correlated with the RewP difference wave amplitude, r 5.28, p (Figures 2 and 3). Steiger s Z (Steiger, 1980) was used to determine if the correlations between reward liking and RewP difference wave amplitude differed between conditions. The correlation between reward liking and the RewP within the anger condition was significantly larger than the same correlation within the neutral condition, Z , one-tailed p (Lee & Preacher, 2013). To be able to depict this correlational result in RewP difference wave amplitudes, we plotted the ERPs of participants in the upper and lower quartiles for reward liking score (Figure 2). As shown in Figure 2, RewP difference wave amplitudes did not differ significantly between upper and lower quartiles in the neutral condition, t(44) , p 5.306, d , but did in the anger condition, t(44) , p 5.033, d Thus, within the anger condition but not within the neutral condition, larger RewPs were significantly associated with greater subjective liking of the rewards. These results suggest that the anger induction increased differences in the neural response to reward signals between individuals who found those stimuli to be especially desirable, and those who found them to be less so. For angry 2. For each of the correlation and regression analyses, standardized residuals (Std. Res.) were used to identify outliers (threshold = 6 3) and Cook s distance was used to identify excessively influential data points. No consistent outliers were identified across all analyses. Specifically, a single data point had a Std. Res. in excess of the threshold, but this was only observed for the correlation between reward liking and the RewP following the neutral induction (standard resolution ), and the regression model that predicted RewP amplitudes following the neutral induction as a function of reward liking and anger RewP amplitudes (standard resolution ). The removal of this data point did not substantially affect the results of the correlation analysis (r 5.06, p 5.568). The regression model was also largely unaffected, F(2,91) , p <.001, adjusted R ), and reward liking still did not predict the neutral RewP amplitude (b , t(91) 5.65, p 5.516). Because this outlier did not appear to be the result of artifact, or to be driving or dampening key findings, we elected to retain it in the reported analysis. No data points were identified as being excessively influential for either of the correlations (Cook s distance <.087) or any of the regression models (Cook s distance <.18).

6 1276 D.J. Angus et al. Figure 1. Event-related potentials to reward and nonreward feedback stimuli, following neutral and anger emotion inductions. The reward positivity (RewP) is isolated by calculating the difference between reward and nonreward trials. individuals, the more they liked the reward stimuli, the more reward-related neural activity they showed when viewing information indicating that they would receive those rewards (as compared to not receiving them). Also, across both conditions, larger RewPs were associated with greater perceived agency. These results suggest that, regardless of emotion induction condition, this neural response to reward signals was greater when individuals believed they had control over obtaining the reward. Self-report measures of trait aggression were not significantly correlated with RewP amplitudes in the neutral or anger conditions, or with reward liking scores. Next, multiple regressions were conducted to examine the unique contribution of reward liking to difference wave RewP amplitude in each emotion induction condition, while first controlling for (a) RewP amplitude in the other induction condition (e.g., RewP in the neutral condition as a predictor of RewP in the anger condition), and then (b) other self-report variables and induction order. Adjusted R 2 is provided as a measure of variance explained by each model. Within the neutral condition, the RewP was significantly predicted by the first regression model, F(2,91) , p <.001. While the model accounted for 17% of variance, the only significant unique predictor was RewP amplitude in the anger condition (b 5.44, t(91) , p <.001). Reward liking did not predict RewP amplitudes (b , t(91) 5.44, p =.664). The second model including self-report variables and induction order also predicted RewP amplitude, F(8,85) , p 5.006; see Table 2. While this second model explained 14.1% of variance, the only significant unique predictor was still RewP amplitude in the anger condition, b 5.38, t(85) , p =.001. Because adjusted R 2 factors in the number of predictors were used in the regression model, the addition of predictors that were unrelated to RewP amplitudes in the neutral condition caused the adjusted R 2 to decrease, rather than increase. A different picture emerged for the RewP in the anger condition. The first model significantly predicted RewP amplitudes, F(2,91) , p <.001, with this model explaining 23% of variance. Both reward liking (b 5.25, t(91) , p =.008) and neutral RewP amplitude (b 5.41, t(91) , p <.001) were significant unique predictors. The second model explained 24.6% of variance, F(8,85) , p <.001; see Table 3. Importantly, reward liking remained a significant unique predictor of RewP amplitudes (b 5.25, t(85) , p =.009). Discussion The present study sought to clarify affective and motivational influences on the RewP, an ERP difference wave observed when Table 1. Bivariate Correlations Between RewP Amplitudes and Self-Report Variable Neutral R 2. Neutral NR.70*** 3. Neutral RewP.59*** Anger R.76***.57***.40*** 5. Anger NR.54***.57***.10.68*** 6. Anger RewP.44***.16.43***.62*** RL ** 8. Agency **.24* ** Physical Verbal Anger Hostility Note. For readability, and as their values are not relevant to the present study, the correlations within the aggression subscales are omitted, although all were positively correlated (rs >.30). Self-report variables are italicized for readability. Neutral R 5 neutral reward; Neutral NR 5 neutral nonreward; Neutral RewP 5 neutral reward positivity; Anger R 5 anger reward; Anger NR 5 anger nonreward; Anger RewP 5 anger reward positivity; RL 5 Reward Liking; Agency 5 perceived agency; Physical 5 physical aggression subscale; Verbal 5 verbal aggression subscale; Anger 5 anger subscale; Hostility 5 hostility subscale. *p <.05. **p <.01. ***p <.001. df 5 97.

7 The reward positivity and anger 1277 Figure 2. Top: Event-related potentials to reward and nonreward feedback stimuli for upper and lower reward liking quintiles, following emotion inductions. Middle: The reward positivity (RewP) for each condition. Bottom: Scalp distribution of the RewP (from left to right: neutral low liking, neutral high liking, anger low liking, anger high liking). feedback indicates that an action will result in a reward. Participants underwent anger and neutral emotion inductions, each followed by a simple gambling task, and rated the likeability of reward stimuli. While there was no overall difference in RewP amplitudes between induction conditions, consistent with previous research (Foti & Hajcak., 2010), the RewP was positively correlated with individual differences in reward liking (e.g., Jia et al., 2013). This correlation was only significant following the anger induction, and it differed from the same correlation within the neutral condition. Moreover, within the anger condition, reward liking remained a unique predictor of RewP amplitude even when statistically controlling for induction order, self-report measures of trait aggression, perceived

8 1278 D.J. Angus et al. Table 3. Multiple Regression Predicting RewP Amplitude in Anger Condition Variable B SE B b t RL ** Order Neutral RewP *** Agency ** Physical Verbal Anger Hostility Note. RL 5 reward liking; Order 5 induction order; Neutral RewP 5 neutral reward positivity; Agency 5 perceived agency; Physical 5 physical aggression subscale; Verbal 5 verbal aggression subscale; Anger 5 anger subscale; Hostility 5 hostility subscale. *p <.05. **p <.01. ***p <.001. df Figure 3. Scatter plots of reward positivity (RewP) amplitude over FCz in neutral (top) and anger (bottom) conditions as a function of selfreported reward liking score. Both plots include line of best fit. agency, and RewP amplitude following the neutral induction. Taken together, these results suggest that the RewP is sensitive to angerrelated approach motivation rather than negative affective valence when individuals desire the reward. While anger is an emotion with negative affect, it appears to strengthen the relationship between the motivational properties of reward stimuli and a neural marker of reward sensitivity (Proudfit, 2014). Research indicates that traits associated with positive affect and/or increased approach motivation are correlated with larger RewP amplitudes (e.g., Bress & Hajcak, 2013; Cooper et al., 2014; Lange et al., 2012; Smillie et al., 2011), while states and traits Table 2. Multiple Regression Predicting RewP Amplitude in Neutral Condition Variable B SE B b t RL Order Anger RewP *** Agency Physical Verbal Anger Hostility Note. RL5 reward liking; Order 5 induction order; Anger RewP 5 anger reward positivity; Agency 5 perceived agency; Physical 5 physical aggression subscale; Verbal 5 verbal aggression subscale; Anger 5 anger subscale; Hostility 5 hostility subscale. *p <.05. **p <.01. ***p <.001. df associated with negative affect and/or reduced approach motivation are correlated with smaller RewP amplitudes (e.g., Foti & Hajcak., 2010; Foti et al., 2011; Santesso et al., 2012). In the present study, we found evidence suggesting that the results obtained in past research are due to differences in motivational direction rather than affective valence. Two complementary findings support this claim. Firstly, we observed that the RewP was correlated with how much participants reported liking the reward stimuli. This measure serves as an estimate of the motivational and appetitive value that participants placed on the reward stimuli. Secondly, the strength of this correlation was increased following an established anger emotion induction. If the positive correlation between stimulus likability and the RewP in the present study was driven by the negative affect associated with anger, then this correlation should have been smaller following the anger induction. Instead, we found that the correlation was larger and robust to statistical correction for other unique predictors of RewP amplitude. It is also possible that the effects reported in the present study and previous (e.g., Foti & Hajcak., 2010) are closely related to the arousing properties of emotion inductions and their resulting motivational intensity. Anger and sadness have been conceptualized as possessing different levels of arousal (e.g., Russell, 1980). In contrast to low-arousal emotions such as sadness, anger typically provokes high levels of arousal and greater motivation to act in an approach-related fashion. However, based on the accumulated evidence to date, we believe that approach motivational intensity rather than arousal per se provides a more accurate explanation of the current results. That is, past research has found other arousing, negative states and traits associated with avoidance motivation to be inversely related to the RewP (Gu, Ge et al., 2010; Gu, Huang, & Luo, 2010; Santesso et al., 2012). An alternative interpretation of the RewP posits that the negative deflection observed in response to nonreward feedback reflects reward prediction error (e.g., the violation of an expectation to receive reward) caused by a phasic decrease in the activation of midbrain dopaminergic neurons that project to the ACC (Holroyd, Nieuwenhuis, Yeung, & Cohen, 2003). These reward prediction errors adjust to expectations of reward or punishment (or nonreward) with greater negative deflections occurring when there is a mismatch between the predicted outcome and the outcome as indicated by feedback. Interpreted within this theoretical framework, the results of the present study suggest that participants who particularly liked the reward stimuli tended to overpredict reward (rather than nonreward) following an increase in anger-related approach motivation. That is, participants who desired the reward stimuli

9 The reward positivity and anger 1279 developed a stronger prediction of reward, which was violated on the presentation of nonreward feedback. This interpretation may suggest a relationship between anger-related approach motivation and increased sensitivity to reinforcement learning and punishment sensitivity (e.g., Massar, Rossi, Schutter, & Kenemans, 2012). Conclusion There is increasing evidence that the RewP is a useful biomarker for assessing reward seeking and responsiveness. However, research supporting this position has unavoidably conflated the potential influences of affective valence and reward motivation, leading to ambiguity over which of these constructs is responsible for modulated RewP amplitudes in clinical and nonclinical populations. In the present study, we showed for the first time that anger-related approach motivation increases the positive relationship between the desirability of reward stimuli and the RewP, suggesting that motivational direction and intensity rather than affective valence drives potentiation or attenuation of the RewP. References Aarts, H., Ruys, K. I., Veling, H., Renes, R. A., de Groot, J. H. B., van Nunen, A. M., & Geertjes, S. (2010). The art of anger: Reward context turns avoidance responses to anger-related objects into approach. Psychological Science, 21, doi: / Becker, M. P. I., Nitsch, A. M., Miltner, W. H. R., & Straube, T. (2014). A single-trial estimation of the feedback-related negativity and its relation to BOLD responses in a time-estimation task. Journal of Neuroscience, 34, doi: /JNEUROSCI Bellebaum, C., Kobza, S., Thiele, S., & Daum, I. (2010). It was not my fault: Event-related brain potentials in active and observational learning from feedback. Cerebral Cortex, 20, doi: /cercor/ bhq038 Bress, J. N., & Hajcak, G. (2013). Self-report and behavioral measures of reward sensitivity predict the feedback negativity. Psychophysiology, 50, doi: /psyp Buss, A. H., & Perry, M. (1992). The aggression questionnaire. Journal of Personality and Social Psychology, 63, doi: / Carlson, J. M., Foti, D., Harmon-Jones, E., & Proudfit, G. H. (2014). Midbrain volume predicts fmri and ERP measures of reward reactivity. Brain Structure and Function, 220, doi: /s Carlson, J. M., Foti, D., Mujica-Parodi, L. R., Harmon-Jones, E., & Hajcak, G. (2011). Ventral striatal and medial prefrontal BOLD activation is correlated with reward-related electrocortical activity: A combined ERP and fmri study. NeuroImage, 57, doi: /j.neuroimage Carver, C. S. (2004). Negative affects deriving from the behavioral approach system. Emotion, 4, doi: / Carver, C. S., & Harmon-Jones, E. (2009). Anger is an approach-related affect: Evidence and implications. Psychological Bulletin, 135, doi: /a Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: The BIS/BAS scales. Journal of Personality and Social Psychology, 67, doi: / Cooper, A. J., Duke, E., Pickering, A. D., & Smillie, L. D. (2014). Individual differences in reward prediction error: Contrasting relations between feedback-related negativity and trait measures of reward sensitivity, impulsivity and extraversion. Frontiers in Human Neuroscience, 8. doi: /fnhum Ford, B. Q., Tamir, M., Brunye, T. T., Shirer, W. R., Mahoney, C. R., & Taylor, H. A. (2010). Keeping your eyes on the prize: Anger and visual attention to threats and rewards. Psychological Science, 21, doi: / Ford, B. Q., Tamir, M., Gagnon, S. A., Taylor, H. A., & Brunye, T. T. (2012). The angry spotlight: Trait anger and selective visual attention to rewards. European Journal of Personality, 26, doi: / per.1840 Foti, D., & Hajcak, G. (2009). Depression and reduced sensitivity to nonrewards versus rewards: Evidence from event-related potentials. Biological Psychology, 81, 1 8. doi: /j.biopsycho Foti, D., & Hajcak, G. (2010). State sadness reduces neural sensitivity to nonrewards versus rewards. NeuroReport, 21, doi: / WNR.0b013e Foti, D., Kotov, R., Klein, D. N., & Hajcak, G. (2011). Abnormal neural sensitivity to monetary gains versus losses among adolescents at risk for depression. Journal of Abnormal Child Psychology, 39, doi: /s Frijda, N. H. (1986). The emotions. Cambridge, England: Cambridge University Press. Gable, P. A., & Harmon-Jones, E. (2010a). Late positive potential to appetitive stimuli and local attentional bias. Emotion, 10, doi: /a Gehring, W. J., & Willoughby, A. R. (2002). The medial frontal cortex and the rapid processing of monetary gains and losses. Science, 295, doi: /science Gratton, G., Coles, M. G. H., & Donchin, E. (1983). A new method for offline removal of ocular artifact. Electroencephalography and Clinical Neurophysiology, 55, doi: / (83) Grubbs, F. E. (1969). Procedures for detecting outlying observations in samples. Technometrics, 11, doi: / Gu, R., Ge, Y., Jiang, Y., & Luo, Y.-J. (2010). Anxiety and outcome evaluation: The good, the bad and the ambiguous. Biological Psychology, 85, doi: /j.biopsycho Gu, R., Huang, Y. X., & Luo, Y. J. (2010). Anxiety and feedback negativity. Psychophysiology, 47, doi: /j x Hajcak, G., Moser, J. S., Holroyd, C. B., & Simons, R. F. (2006). The feedback-related negativity reflects the binary evaluation of good versus bad outcomes. Biological Psychology, 71, doi: / j.biopsycho Harmon-Jones, C., Schmeichel, B. J., Mennitt, E., & Harmon-Jones, E. (2011). The expression of determination: Similarities between anger and approach-related positive affect. Journal of Personality and Social Psychology, 100, doi: /a Harmon-Jones, E. (2003). Anger and the behavioral approach system. Personality and Individual Differences, 35, doi: / S (02) Harmon-Jones, E. (2007). Trait anger predicts relative left frontal cortical activation to anger-inducing stimuli. International Journal of Psychophysiology, 66, doi: /j.ijpsycho Harmon-Jones, E., Harmon-Jones, C., Abramson, L. Y., & Peterson, C. K. (2009). PANAS positive activation is associated with anger. Emotion, 9, doi: /a Harmon-Jones, E., & Peterson, C. K. (2008). Effect of trait and state approach motivation on aggressive inclinations. Journal of Research in Personality, 42, doi: /j.jrp Harmon-Jones, E., Harmon-Jones, C., & Price, T. F. (2013). What is approach motivation?. Emotion Review, 5, doi: / Hauser, T. U., Iannaccone, R., St ampfli, P., Drechsler, R., Brandeis, D., Walitza, S., & Brem, S. (2014). The feedback-related negativity (FRN) revisited: New insights into the localization, meaning and network organization. NeuroImage, 84, doi: /j.neuroimage Hennighausen, E., Heil, M., & R osler, F. (1993). A correction method for DC drift artifacts. Electroencephalography and Clinical Neurophysiology, 86, doi: / (93)90008-J Holroyd, C. B., Hajcak, G., & Larsen, J. T. (2006). The good, the bad and the neutral: Electrophysiological responses to feedback stimuli. Brain Research, 1105, doi: /j.brainres Holroyd, C. B., Nieuwenhuis, S., Yeung, N., & Cohen, J. D. (2003). Errors in reward prediction are reflected in the event-related brain potential. Neuro- Report. 14, doi: /01.wnr a5 Horan, W. P., Foti, D., Hajcak, G., Wynn, J. K., & Green, M. F. (2012). Impaired neural response to internal but not external feedback in schizophrenia. Psychological Medicine, 42, doi: / S