The specificity of social feedback processing:
An EEG study on the MFN and P300 event-related potentials
Laurens T. Kemp, Laura M. S. Dekkers1, Maurits W. van der Molen1
1Department of Developmental Psychology, University of Amsterdam, The Netherlands
Corresponding author:
Laurens T. Kemp, Rode Kruislaan 1119D, Diemen, The Netherlands, E: ltk1@live.nl, T: +31 6 50 81 35 90
Word counts: Abstract: 200
0. Abstract
We investigated the Medial Frontal Negativity (MFN) and the P300 event-related potentials, to determine whether they were sensitive to socially-evaluative effects, feedback congruency and/or valence. Six participants performed on a Social Judgment (SJ) task and a novel control task, coined the Incentive Judgment (IJ) task. In the SJ task, participants judged photographs of unknown others and decided whether they believed that the other person liked them or not. The IJ task, using the same format as the SJ task, asked whether a person would share money with them. Participants received (fictitious) positive or negative feedback upon their decisions. This control task allowed to control for rejection feedback that is not explicitly social while keeping the feedback pattern of congruence and valence identical to the SJ task. We found a larger MFN amplitude at Fz for incongruent versus congruent feedback in the SJ task, and also at Fpz for negative versus positive feedback in both tasks. Contrary to previous results, we did not find that the P300 was sensitive to socially-evaluative effects. These preliminary data are interesting in that they show an unexpected sensitivity to feedback valence, but interpretation of this pattern is not warranted given the small number of participants.
Keywords:
1. Introduction
Social rejection evokes a strong aversive response in most mammals, including humans (Vogt & Gabriel, 1993). The salience of social rejection is hypothesised to be due to social separation in mammals being similarly threatening to survival as physical injury (Eisenberger, 2012a; Panksepp, 1998). Over the past decades, researchers have begun to explore the neural correlates underlying the processing of social rejection. To pursue this goal, a number of functional Magnetic Resonance Imaging (fMRI) studies employ the Cyberball game, which is a ball-catching game between the participant and two other, virtual players. In this game, the ball is initially passed between all three players, but eventually the two other players will only pass the ball between each other, making the participant feel excluded (Williams & Jarvis, 2006). Multiple studies show that the anterior cingulate cortex (ACC) and the anterior insula show increased activity when a person is being ostracised in this game (Eisenberger, 2012b). Since these areas are also involved the processing of physical pain, recruitment of these regions during Cyberball points to a certain amount of overlap between the processing of physical pain and social rejection.
Additional studies using the Cyberball paradigm have been done while measuring Electro-encephalography (EEG), in order to investigate these patterns of activation on a more precise timescale. These studies indicate a frontal-central positive slow wave in response to social rejection (Crowley et al., 2009; Gutz, Küpper, Renneberg & Niedeggen, 2011). This is possibly reflective of the response of the ACC, as shown in the fMRI studies employing Cyberball (Eisenberger, 2012b).
The Cyberball paradigm is a good way to induce social stress in a person, but it is not precise in conveying social acceptance or rejection on a trial-by-trial basis, nor is it able to test for the influence of expectations of the participant. A paradigm that addresses these caveats has been developed by Somerville, Heatherton, and Kelly (2006), and is called the Social
Judgment (SJ) task. In their experiment, participants were led to believe that their photo had been judged by a panel of peers. During an fMRI session, they were shown photographs of panel members and were asked “Would this person like you?” Participants first indicated their expectation (‘Yes’/’No’) and were then presented (fictitious) feedback on their answer (‘Yes’/’No’). The possible combinations of feedback in this task are thus ‘Yes-Yes’ (expected acceptance), ‘No-No’ (expected rejection), ‘Yes-No’ (unexpected rejection) and ‘No-Yes’ (unexpected acceptance).
Together, these conditions allow for comparison of neural activity to social acceptance or rejection feedback, as well as feedback that is congruent or incongruent with prior expectations. Using this paradigm, Somerville et al. (2006) showed that the dorsal ACC responded primarily to expectancy violation, with increased dACC activity to incongruent as opposed to congruent feedback. The ventral ACC, on the other hand, was more sensitive to the social valence of the feedback, with more vACC activity in response to social acceptance as compared to rejection feedback. These findings were replicated and extended by Gunther Moor, Leijenhorst, Rombouts, Crone & Van der Molen (2010), who found increased activity in a number of areas, including the ventral medial prefrontal cortex (which includes the vACC). In ‘Yes’ expectation trials, acceptance, but not rejection, resulted in significantly more activation, whereas in ‘No’ expectation trials the reverse was true: rejection, but not acceptance, resulted in significantly more activation. This extends the results of Somerville et al. (2006), and shows that the effects of congruence and valence may be caused by differences between specific feedback conditions (Yes-No, etc.) rather than by congruency and/or valence in general.
In order to investigate the cognitive processes involved in social feedback processing at more precise timescales, EEG studies that employed the SJ paradigm have been carried out. Van der Veen, Van der Molen, Sahibdin, and Franken (in press) used the SJ task to
investigate the P300 (also known as the P3), while Dekkers, Gunther Moor, Van der Veen, & Van der Molen (2013) measured the Medial Frontal Negativity (MFN, also known as the Feedback Related Negativity, FRN) and P300 event-related brain potentials (ERPs). These components will now be further elaborated.
The MFN is an ERP component that is time-locked to the onset of feedback presentation and peaks around 250 ms post-feedback at fronto-central regions.The MFN is believed to originate from the ACC (Ridderinkhof, Ullsperger, Crone & Nieuwenhuis, 2004). According to some recent studies, the MFN is strongest when a person receives feedback signalling that an outcome is different from what was expected, or that a behaviour was incorrect (for a review, see Van Noordt & Segalowitz, 2012). Dekkers et al. (2013) used the SJ task to further investigate the MFN in a social context, and found that it is specifically sensitive to feedback congruency (expected vs. unexpected) rather than valence (positive vs. negative). However, the exact properties that the MFN is sensitive to are still not completely clear.
The P300 is a positive slow-wave potential that is usually determined as the maximal peak between 200 and 600 ms post-feedback. It is typically strongest around central parietal regions (Wu & Zhou, 2009). The P300 has been associated with evaluating gains and losses (Wu & Zhou, 2009), probability estimation, and is modulated by motivationally significant tasks (Nieuwenhuis, Aston-Jones, & Cohen, 2005). These properties have also been demonstrated for the P300 in studies using the SJ task (Dekkers et al., 2013; Van der Veen et al., in press). More specifically, these studies found that the P300 is sensitive to socially-evaluative effects. Together, these ERPs may represent aspects of the encoding of social feedback.
In order to control for the common influence of cognitive performance feedback, as well as for the influence of self-referential cognition, Dekkers et al. (2013) used an
age-matching task (also see, Gunther Moor, Crone, & Van der Molen, 2010) that was designed to be as self-referential as the SJ task, but did not involve social judgment. That is, in the SJ task the participant was asked “Would this person like you?” while the control task asked “Is this person as old as you are?” This control task allowed social feedback to be compared to affectively neutral, non-social feedback.
Using this control task, Dekkers et al. (2013) showed that the P300 had a larger amplitude in the social task as compared to the non-social age-matching task, as well as after unexpected as compared to the expected feedback. The social vs. non-social effect was replicated by Van der Veen et al. (in press), who also found that there was a stronger P300 for expected social acceptance. This could result from this type of feedback having higher motivational significance.
A few issues remain unresolved in the previous studies. Namely, the Age Judgment tasks have not been able to create equivalent conditions with the SJ task, for two reasons. Firstly, the conditions representing negative feedback differ between tasks so that conditions of both tasks cannot be compared directly. That is, the negative feedback conditions for the AJ task (i.e. where the participant incorrectly judged a person to be younger or older) are the ‘Yes-No’ and ‘No-Yes’ conditions, whereas for the SJ task (where the participant is rejected) these are the ‘Yes-No’ and ‘No-No’ conditions. Secondly, and related to the first issue, the AJ task, but not the SJ task, lacks the ability to find valence effects (positive vs. negative feedback) that are separate from congruency effects (expected vs. unexpected). More specifically, in the AJ task, the ‘Yes-No’ and ‘No-Yes’ conditions both represent incongruent and negative feedback, and there is no condition that represents incongruent positive feedback, or congruent negative feedback.
In addition, the effect of varying degrees of social involvement has not been investigated, which still leaves the possibility that the effect of a stronger P300 in the SJ task
as compared to the AJ task is purely due to social cognition in general, rather than specifically due to the experience of being socially accepted (or rejected). This can be achieved by using money as the object of rejection, as is explained in the following section.
In this paper, we describe the preliminary findings on the MFN and P300 in two different judgment tasks: the Social Judgment (SJ) task (Somerville et al., 2006), and a newly developed Incentive Judgment (IJ) task. In the IJ task, the participant is asked “Would this person share money with you?” This question does create equivalent conditions with the SJ task, since all four feedback conditions (‘Yes-Yes’, ‘Yes-No’, ‘No-Yes’ and ‘No-No’) match up according to the combinations of positive/negative and expected/unexpected outcomes in the SJ task. Thus, the conditions in both tasks can be directly compared to one another. Additionally, the IJ task enables us to dissociate the effects of feedback congruency and valence, since a participant’s correct expectation (i.e. congruent feedback) that the other person does not want to share money with them still presents a negative outcome. Finally, feedback in this task does recruit aspects of social cognition. However, social feedback in this task is not as explicit in communicating acceptance or rejection as in the SJ task, because if a participant receives negative feedback on the IJ task, this can be explained by assuming the person on the photo is simply selfish, and that the negative feedback does not necessarily reflect on the participant personally. This difference in explicitness means we can expect the SJ task to evoke a stronger aversive response than the IJ task, which should be reflected in the ERP.
Considering previous results, we hypothesize that the MFN reflects the processing of a prediction error in the ACC (Van Noordt & Segalowitz, 2012), and so it is sensitive to feedback congruency (Dekkers et al., 2013). Our second hypothesis is that the processing of social feedback is reflected in the P300, which is particularly sensitive to motivationally significant feedback (Nieuwenhuis et al., 2005; Dekkers et al., 2013). Therefore, we expect
larger amplitudes for the SJ task compared to the IJ task, just as Dekkers et al. (2013) found for the SJ task as compared to the AJ task. Overall, we expect there to be a dissociation in social feedback processing for the MFN and the P300, where the former is more sensitive to congruency and the latter to context.
Furthermore, within the experimental conditions of the SJ task we may be able to find that expected positive feedback results in a larger P300-amplitude compared to other feedback conditions, consistent with the results of Van Der Veen et al. (2013). It is also possible that we find a higher P300 amplitude for unexpected versus expected outcomes across both tasks, as some studies have reported (e.g. Dekkers et al., 2013; Hajcak, Holroyd, Moser & Simons, 2005).
In order to test these hypotheses, we conducted an experiment using the SJ and IJ tasks while measuring EEG, electrocardiography and galvanic skin response. Behavioral measures were obtained by using three questionnaires. This paper will focus on the EEG results, as the analysis of heart rate and skin conductance was outside the scope of this paper. Only the EEG results of the first six participants are included due to time constraints. The results of the other subjects and physiological and behavioral measures are reported elsewhere.
2. Methods 2.1 Participants
Twenty-six students (17 female; ages 18-27; µage = 21; σage = 2.4), recruited via the University
of Amsterdam’s research recruitment website, participated in the study for either course credit or financial compensation. Inclusion criteria were: age between 18 and 27; right-handedness, proficient in the Dutch language, no current or history of neurological or psychiatric illness, no use of prescribed psychotropic medication, and no chronic use of illicit drugs. Written consent was obtained prior to participation. The study was approved by the local Ethics Committee of the University of Amsterdam. Following testing, two participants were
excluded, one who did not fulfil the inclusion criteria, and one who did not believe the cover story. For this preliminary study, only the first six participants (4 female; ages 18-23; µage =
21; σage = 1.8) were included in the analysis.
2.2 Stimulus Materials, Task Description, and Experimental Design
Students who expressed interest in participating in the experiment via e-mail or online sign-up were contacted by e-mail to inform them that they would be involved in a study that aimed at investigating first impressions. To this end, they were asked to send their portrait photograph to the researchers and were led to believe that peers from other universities would form impressions about their photographs before they would perform in the experimental test session. Participants were informed that photographs of these peers would be presented to them during the experimental test session and that they would have to perform two tasks in which they had to estimate how their peers had judged their photos. Unbeknownst to the participants, the photos that they submitted were not judged by anyone and the photos that were presented to them during the test session were of 330 volunteers.
The photos that were used in the tasks were of full-facing faces with a neutral expression. The persons photographed were men and women of ages 18 to 30 (Males: µage =
22.1; σage = 2.6; females: µage = 21.4; σage = 2.6) against a black background, and were divided
into two sets of 160 photos, one for each task. Ten additional photos (all female) were used for the practice set. Photos were randomly ordered, and no individual face was presented more than once. Two hundred and four photos were gathered from previous experiments (Gunther Moor et al., 2010; Dekkers et al., 2013);144 additional photos were taken at the Hogeschool van Amsterdam, Vrije Universiteit (Amsterdam), and Universiteit Utrecht. Of these photos, a selection was made to create equal distribution of men and women. All photographed persons provided written consent for the use of their photo for scientific purposes.
Faces (224 x 280 pixels) were presented using Presentation software (version 16.01; Neurobehavioral Systems, Albany, CA) in colour against a black background in the centre of a 22-inch computer monitor (refresh rate = 120 Hz, resolution = 1600 x 900 pixels).
Participants performed on two tasks: the Social Judgment (SJ) and the Incentive Judgment (IJ) task. In the SJ task, participants were asked to decide whether they expected to be liked or disliked by the person on the photograph (“Would this person like you?”). In the IJ task, participants were asked to decide whether they expected the person on the photograph to be willing to share money with them (“Would this person share money with you?”).
Figure 1 depicts a schematic of the experimental design (A) and trial sequence (B). Each trial started with a central fixation cross presented with a 600-1600 ms jittered duration. The fixation cross was followed by a facial cue that remained on the screen for the rest of the trial. For each trial, participants were given 3000 ms to indicate their expectation. They did this by pressing a button with their right index or middle finger for either ‘Yes-’ or ‘No’-expectations, respectively. Immediately after their choice, their answer was displayed on the left part of the screen. After a delay of 1000 ms, positive (‘Yes’) or negative (‘No’) feedback was shown on the right part of the screen, indicating either that their answer was correct (‘Yes-Yes’ for positive feedback or ‘No-No’ for negative feedback) or that their answer was incorrect (‘Yes-No’ for negative feedback or ‘No-Yes’ for positive feedback), for a duration of 2000 ms. When there was no response within the 3000 ms response window, the message “TOO LATE!” appeared, followed by a new trial.
The order of both tasks was counterbalanced across participants. All participants performed a total of 170 trials for each task, consisting of 10 practice trials and four successive blocks of 40 test trials. On half of these trials they received ‘Yes’-feedback and on the other half they received ‘No’-feedback, which was distributed evenly across male and
female faces. Unbeknownst to the participants, feedback for all trials was generated pseudo-randomly by the computer.
Insert Figure 1 here 2.3 Procedure
At the start of the test session, participants read the information brochure and signed the informed consent form. Hereafter EEG, ECG, and GSR equipment was attached and participants were reminded of the purpose of the study by a rehearsal of the cover story. All participants were tested in a sound and electrical shielded EEG chamber, while sitting in a comfortable chair at a distance of approximately 75 cm from the computer monitor. Prior to starting each task, participants received verbal, followed by written instructions on the task at hand. Baseline EEG, ECG, and GSR measures were recorded during a two-minute period prior to each task. Participants were permitted a few minutes rest between tasks if they so desired. After the recording session, three questionnaires were administered (for references, see Dekkers et al., 2013). To test for the integrity of the cover story, participants were asked to write down their experiences during and thoughts about the study at the end of the test session. Participants were debriefed by e-mail after all participants were tested.
2.4 EEG Data Collection and Reduction
EEG was recorded with a 64-channel ActiveTwo system (BioSemi, Amsterdam, The Netherlands) using 64 Ag-AgCl electrodes mounted in an electrode cap (10/20 system), at a sampling frequency of 1024 Hz. On average, electrode offsets were kept below 20 µV. The BioSemi Common Mode Sense (CMS) active electrode and Driven Right Leg (DRL) passive electrode were used as grounds. To monitor eye blinks and movements, horizontal and vertical electro-oculography (EOG) were measured with two Ag-AgCl electrodes placed on the left and right cantus and above and below the right eye, respectively. The left and right mastoids were used as offline reference sites.
Offline analysis of the EEG time series was performed using BrainVision Analyzer (BVA, version 2.0.1.5528, Brain Products GmbH, 1998-2008). EEG time series were down-sampled to 512 Hz, re-referenced to the right and left mastoids, and band-bass filtered between 0.1-40 Hz (24dB/oct) and notch filtered at 50 Hz. EOG artefacts and EOG-related muscle contractions were removed from the data by using an ocular independent component analysis (ICA), as implemented in BVA. Bad channels were interpolated with neighbouring channels. Subsequently, 5600 ms epochs were created time-locked to the onset of the face stimulus, including a 100 ms pre-stimulus interval. Since gross disturbances at the time of stimulus presentation and response as well as feedback presentation could have influenced processing of feedback, these epochs, encompassing the entire trial, were used to inspect the EEG time series for artefacts, in order to ensure that only artefact-free trials were included in the analysis. Additionally, epochs where the response time of the participant was shorter than 100 ms were excluded. Artefact-free trials were then segmented in 1200 ms epochs time-locked to the onset of the feedback, including a 200 ms pre-feedback baseline. Automatic artefact rejection was then used to reject A) epochs with a signal exceeding a maximal voltage step of 50 µV/ms, B) epochs in which the difference between the maximum and minimum voltage in a time window of 100 ms did not exceed 0.50 µV, and C) epochs in which a difference of 200 µV occurred within an interval of 200 ms. Baseline correction was done using the 200 ms pre-feedback baseline, and epochs were then averaged per condition.
MFN was determined at Fpz, Fz, FCz, and Cz, and P300 peak amplitude at Cz, CPz, Pz, POz, Oz, and Iz, by using the local peak method. Peak markers of the P2, MFN and P300 were visually inspected and were corrected where necessary. Given that Dekkers et al. (2013) found maximal ERP amplitudes at midline channels, lateral electrodes were not included in the analysis. MFN peak amplitude was calculated in three steps. First, the P2 component was identified by determining the most positive value in the 150-250 ms post-feedback window.
Peak latency of the P2 component was then used as the onset of the negativity. Second, the most negative value was identified from the onset of the negativity to 350 ms after the onset of the feedback (i.e., MFN time-window). Finally, the difference between the P2 and most negative value was taken as MFN peak amplitude. The MFN was scored at 0 µV when no negativity could be identified within the MFN time-window. The latency at which peak negativity was found was taken as MFN peak latency.
Peak amplitude of the P300 was defined base-to-peak as the most positive value of the ERP within the 250-500 ms post-feedback window relative to the 200 ms pre-feedback baseline corrected base. The time at which the P300 reached peak amplitude was taken as P300 peak latency.
2.5 Statistical Analyses
Results were analysed using repeated measures Analyses of Variance (ANOVAs), carried out with IBM SPSS Statistics 20 (IBM Corporation, 1989-2011). Statistical analyses were evaluated against an alpha of .05, and alphas between 0.1 and .05 were taken as marginally significant.
To test our hypotheses, repeated measures ANOVAs were performed using a 2 (Task) by 2 (Valence) by 2 (Congruency) factorial design. In case of three-way interaction effects, we performed two separate 2 (Valence) by 2 (Congruency) ANOVAs for each Task, separately. We do not have hypotheses regarding the ERP latencies; these tests are exploratory.
3. Results
In Figure 2 and 3, the grand average ERP waveforms associated with the feedback stimulus at Fpz, Fz, Cz and Pz, are presented for each of the feedback conditions of both tasks. Figure 2 plots the waveforms associated with electrodes that were also analysed by Dekkers et al.
(2013) (MFN at Fz and P300 at Pz), while Figure 3 plots the waveforms associated with the electrodes showing maximal amplitudes for the MFN (Fpz) and P300 (Cz).
3.1 Preliminary Analyses
To determine at which electrode the MFN and P300 amplitudes were maximal, two repeated measures Analyses of Variance (ANOVAs) were performed on the mean amplitudes of both tasks. The MFN at Fpz, AFz, Fz, FCz, and Cz was compared in a 5 (Site) by 2 (Task) factorial design. The P300 at Cz, CPz, Pz, POz, Oz, and Iz was compared in a 6 (Site) by 2 (Task) factorial design. The analyses indicated that for the MFN, there were no main effects or interaction effects, all ps > .05. For the P300, within-subjects tests revealed a significant main effect of Site, F (5, 25) = 30.520; p < .001. Other main and interaction effects were not significant, all ps > .05. Because there were no main effects of Task or Task by Site interaction effects, we used the sites with the highest average amplitude for further analysis. The MFN amplitude was maximal at Fpz (SJ task: µmax = -3.73 µV; σ = 2.14 µV; IJ task: µmax
= -3.46 µV; σ = 1.49 µV) and the P300 amplitude was maximal at Cz (SJ task: µmax = 17.34
µV; σ = 5.47 µV; IJ task: µmax = 15.77 µV; σ = 6.80 µV). Given than Dekkers et al. (2013)
performed their analysis on electrode sites Fz (for the MFN) and Pz (for the P300), we also ran tests on these sites in order to improve comparability. Thus, further analysis was carried out on both Fpz and Fz for the MFN, and both Cz and Pz for the P300.
Insert Figures 2 and 3 here 3.2.1 MFN amplitude.
We predicted that the MFN amplitude would be stronger for incongruent compared to congruent feedback, and that feedback valence would have no effect. Contrary to these predictions, at Fpz, where MFN amplitudes were maximal, a significant main effect of Valence was shown, F (1, 5) = 7.597; p = .040; η2
p = .603, where negative feedback (Yes-No,
However, the results at Fz – the electrode site that was also tested by Dekkers et al. (2013) – did confirm our predictions. Here we found a marginally significant effect of Congruency, F (1, 5) = 6.518; p = .051; η2
p = .566, where incongruent feedback (Yes-No,
No-Yes) corresponded with a higher amplitude than congruent feedback (Yes-Yes, No-No). Furthermore, there was a marginally significant Task by Valence by Congruency three-way interaction on MFN amplitude at Fz, F (1, 5) = 5.447; p= .067; η2
p = .521. To further explore
this effect, the MFN amplitude at Fz is depicted in Figure 4. It suggests that the amplitudes in the congruent conditions are roughly similar for both tasks, but in the incongruent conditions the tasks seem to diverge, with the SJ task resulting in a much stronger MFN than the IJ task. We investigated this further by performing two repeated measures ANOVAs on the SJ task and the IJ task separately with a 2 (Valence) by 2 (Congruency) factorial design. This revealed that there was a significant main effect of Congruency for the SJ task, F (1, 5) = 14.504; p = .013; η2
p = .744, but not the IJ task, all ps > .05. For the SJ task, incongruent
feedback (Yes-No, No-Yes) corresponded to a higher amplitude than congruent feedback (Yes-Yes, No-No).
Insert Figure 4 here
3.2.2 MFN latency.
The MFN latency at Fpz showed a main effect of Congruency, F (1, 5) = 7.998; p = .037; η2 p= .615, where incongruent feedback corresponded with a later peak amplitude than congruent
feedback. A similar main effect of Congruency on MFN latency was observed at Fz, F (1, 5) = 9.017; p = .030; η2
p = .643. Other effects were absent, all ps > .05.
3.3.1 P300 amplitude.
We predicted higher P300 amplitude for the SJ task compared to the IJ task. Contrary to this expectation, at neither Cz nor Pz, there were significant effects of P300 amplitude across tasks and/or conditions, all ps > .05.
3.3.2 P300 latency.
P300 latency at Cz showed a significant main effect of Congruency, F (1, 5) = 17.973; p = . 008; η2
p = .782, where incongruent feedback (Yes-No, No-Yes) corresponded to a later peak
amplitude than congruent feedback (Yes-Yes, No-No). P300 latency at Pz showed no main or interaction effects (all ps > .05).
4. Discussion
In this paper, we described the preliminary results of an EEG experiment investigating the effect of social acceptance and rejection feedback on the MFN and P300 ERPs. We used the Social Judgment task, a task where participants indicated their expectation on whether a person on a photograph would like them or not, followed by feedback on whether or not they were actually liked or disliked. We used the Incentive Judgment task as a control task, which has equivalent feedback conditions on valence and congruency to the Social Judgment task, and additionally involves social cognition without explicit social judgment. This task required participants to indicate whether they expected the person on the photograph to share money with them, after which they received feedback indicating whether the person in question indeed would or would not be willing to do this.
Our results indicate that the MFN at Fpz, in both tasks, had a higher amplitude for negative versus positive feedback, an effect that is contrary to our predictions. In addition, the MFN at Fpz peaked later for incongruent versus congruent feedback. In order to allow a comparison with the results of Dekkers et al. (2013), the MFN at electrode site Fz was also tested. The MFN at Fz had a higher amplitude, for incongruent versus congruent feedback, which is consistent with our predictions regarding amplitude. MFN peak amplitude also occurred later for incongruent versus congruent feedback. We tested the P300 at electrode sites Cz and Pz. At Cz, it only showed a longer latency for incongruent versus congruent feedback, while there
were no effects at Pz. This also did not match our prediction regarding a task-specific effect for P300 amplitude.
4.1 Medial Frontal Negativity
For electrode site Fz we found a marginally significant effect of congruency on MFN amplitude, which was included in a marginally significant Task by Valence by Congruency interaction effect. From the additional tests, we found that the SJ task alone showed a significant effect of congruency. This partially matches previous results on the MFN and the SJ task (Dekkers et al., 2013; Hajcak et al., 2006). However, we found no such effect on the IJ task, which was used in this experiment for the first time. This is unexpected given that Dekkers et al. (2013) found a congruency effect across tasks when using the Age Judgment task as a control for the SJ task. This points to the possibility that the IJ task is less suitable as a control task when it comes to measuring the MFN. However, our limited sample size makes the effects in our sample difficult to detect.
Inspecting the MFN amplitudes displayed in Figure 4 lets us see that negative feedback tends to evoke a (non-significantly) larger MFN than positive feedback, with the exception of unexpected negative feedback on the IJ task. Furthermore, the amplitudes for negative feedback on the IJ task appear nearly identical. This implies that participants do not react as strongly to negative feedback in the IJ task (regardless of whether it is expected or unexpected) compared to the SJ task. This may be a result of the different motivational significance of the two tasks. There is some evidence that the MFN is also sensitive to the motivational significance of the feedback (Yeung, Holroyd & Cohen, 2005), so this effect will have to be considered if the IJ task is used in future experiments.
The higher MFN amplitude for negative feedback at Fpz contradicts previous research on the Social Judgment task, which indicated that the MFN was stronger for incongruent feedback, but not sensitive to feedback valence. It is possible that in this task, the MFN is
indeed sensitive to feedback valence, but this is contrary to our expectations. One possible explanation for the valence effect is that we analysed the MFN at electrode site Fpz, whereas previous studies have focused on Fz. Measuring the MFN at Fpz may have introduced additional influences on the ERP that are not commonly associated with the MFN. In other words, we may be measuring a brain potential other than the MFN, which happens to share the same time window, in addition to the signal that presumably originates in the dACC (Somerville et al., 2006). It is possible that this finding is related to an earlier fMRI study on the SJ task, which found that the ventral ACC did respond to valence (Somerville et al., 2006). We cannot confidently say whether this effect is specific to electrode site Fpz, since we did not perform source modelling on our data. It should be noted, however, that the ERP on Fpz suffers from a large amount of noise, as can be seen in Figure 3. Combined with the small sample size, this means that this finding is unreliable.
We discovered that at electrode sites Fpz and Fz, incongruent feedback resulted in later peak amplitude than congruent feedback for the MFN. We did not include this in our hypotheses, but it is possible that incongruent feedback is more complex than congruent feedback, and takes longer to process. Nevertheless, it is evident that this result is consistent with the findings of Dekkers et al. (2013). In contrast, their results also indicated that the MFN latency was larger for the AJ task compared to the SJ task, which we were not able to find here. The absence of an effect on MFN latency may be because the IJ task, which we used as a control task, is more similar to the SJ task than the AJ task. Although this points to the validity of the IJ task as a control task, we need to be careful in interpreting a lack of an effect as evidence for this.
4.2 P300
The P300 is associated with motivationally significant feedback, and has been shown to be stronger for the SJ task compared to an Age Judgment task (Dekkers et al., 2013; Van Der
Veen et al., in press). Unfortunately, we were unable to detect any effect of task on the P300 amplitude, on neither the Pz nor Cz electrode sites. This is contrary to our prediction of a stronger P300 for the SJ task, and also to our prediction of a possible effect of expected positive feedback. For this experiment, this suggests that the motivational significance of the SJ and IJ tasks are not experienced differently, and thus the P300 amplitudes do not differ. In order to confirm this, more data is required, since our limited sample size makes this result potentially unreliable.
The lack of significant differences between tasks could be attributed to the difference between the control tasks used here and by Dekkers et al. (2013) and Gunther Moor et al. (2010). Specifically, the Incentive Judgment task involves aspects of social cognition, namely a judgment of personality, which was not involved in the Age Judgment task. Therefore, the control task was a closer approximation of the cognitive process that we investigated in the SJ task – that is, judging how likeable someone finds you – and so we could expect the ERPs across tasks to be similar. There are two possible explanations for this result:
1) The IJ task did not sufficiently distinguish the type of feedback from the SJ task, and the participant still experienced social rejection in spite of being told that they were being judged on whether the other person would share money with them, instead of whether they would be liked based on their appearance. Judging from the answers the participants provided on their experience of both tasks, this seems unlikely, since it appears that they did not indicate the IJ task to be as distressing as the SJ task.
2) The IJ task is sufficiently distinct from the SJ task, but the difference in P300 amplitude is solely attributable to social cognition rather than the processing of social acceptance or rejection. This explanation is consistent with our result that the P300 did not differ significantly between tasks. However, we should be careful with interpreting a lack of an effect when our sample size is low.
The lack of an effect for expected positive feedback is not readily explained. This effect may become detectable with more data, but given that Dekkers et al. (2013) did not find this effect either, it may only be weakly present. It is possible that in the experiments using the SJ task and the related control tasks, the response of the P300 to social feedback may be modulated by social context alone. However, the literature regarding the P300 provides some evidence that it responds to additional factors, such as positive or unexpected feedback (Bellebaum & Daum, 2008), so the possibility of finding valence and congruency-specific
effects in experiments using the SJ task remains open.
Finally, the P300 at Cz also showed a longer latency for incongruent compared to congruent feedback. At Pz, on the other hand, P300 latency did not show any significant differences, which is consistent with Dekkers et al. (2013). The latency of the P300 has been associated with stimulus evaluation time (Kutas, McCarthy & Donchin, 1977), which makes sense in the context of this experiment: an unexpected result is likely to cause participants to evaluate the stimulus more, i.e. think about why they were wrong.
4.3 Conclusions
Among our most important findings is that the MFN at Fz was larger for incongruent compared to congruent feedback when the Social Judgment task was used, indicating that the MFN is primarily sensitive to congruency. This is consistent with the theory that the MFN represents the evaluation of outcomes independent of their context or valence (Alexander & Brown, 2011). In addition, we observed that the MFN was sensitive to valence when measured at frontal polar electrode sites (Fpz). The valence effect at the frontal polar electrodes warrants further investigation.
Our data on the P300 proved to be inconclusive, and we were unable to find any of the predicted effects. Although we cannot confidently say these results are reliable, they have nevertheless given us some information about the Incentive Judgment task. The lack of
task-specific effects may be due to the P300’s sensitivity to social cognition rather than social feedback. If this is indeed the case, the IJ task requires additional research to determine how exactly it relates to the SJ task.
Future studies could investigate the validity of the IJ task further by including a task that determines in what degree self-referential context is a factor in social feedback processing. An impersonal variant of the IJ task would ask “Would this person share money with someone?” In this case, the participants would be told that they have not been judged, and that they are simply to make a personality assessment based on someone’s photo. This would allow us to determine whether the P300 is sensitive to feedback that is social in nature (namely, the feedback concerns personality traits), or is dependent on self-referential processing. Such an experiment would serve to further elucidate the role of the P300 in social feedback processing.
5. References
Alexander, W. H., & Brown, J. W. (2011). Medial prefrontal cortex as an action-outcome predictor. Nature neuroscience, 14(10), 1338-1344.
Bellebaum, C., & Daum, I. (2008). Learning‐related changes in reward expectancy are
reflected in the feedback‐related negativity. European Journal of Neuroscience, 27(7), 1823-1835.
Dekkers, L., Gunther Moor, B., Van Der Veen, F., Van Der Molen, M. (2013). Cardiac and electro-cortical concomitants of social feedback processing. Manuscript submitted for publication.
Eisenberger, N. I., Lieberman, M. D., & Williams, K. D. (2003). Does rejection hurt? An fMRI study of social exclusion. Science, 302(5643), 290-292.
Eisenberger, N. I. (2012a). The neural bases of social pain: evidence for shared representations with physical pain. Psychosomatic Medicine, 74(2), 126-135.
Eisenberger, N. I. (2012b). The pain of social disconnection: examining the shared neural underpinnings of physical and social pain. Nature reviews neuroscience, 13(6), 421-434.
Gunther Moor, B. , Crone, E. A., & van der Molen, M. W. (2010). The Heartbrake of Social Rejection Heart Rate Deceleration in Response to Unexpected Peer Rejection. Psychological science, 21(9), 1326-1333.
Gunther Moor, B., van Leijenhorst, L., Rombouts, S. A., Crone, E. A., & Van der Molen, M. W. (2010). Do you like me? Neural correlates of social evaluation and developmental trajectories. Social neuroscience, 5(5-6), 461-482.
Gu, R., Lei, Z., Broster, L., Wu, T., Jiang, Y., & Luo, Y. J. (2011). Beyond valence and magnitude: A flexible evaluative coding system in the brain. Neuropsychologia, 49(14), 3891-3897.
Hajcak, G., Holroyd, C. B., Moser, J. S., & Simons, R. F. (2005). Brain potentials associated with expected and unexpected good and bad outcomes. Psychophysiology, 42(2), 161-170.
Holroyd, C. B., & Coles, M. G. (2002). The neural basis of human error processing:
reinforcement learning, dopamine, and the error-related negativity. Psychological review, 109(4), 679-709.
Kutas, M., McCarthy, G., & Donchin, E. (1977). Augmenting mental chronometry: The P300 as a measure of stimulus evaluation time. Science, 197(4305), 792-795.
Nieuwenhuis, S., Aston-Jones, G., & Cohen, J. D. (2005). Decision making, the P3, and the locus coeruleus--norepinephrine system. Psychological bulletin, 131(4), 510-532. Ridderinkhof, K. R., Ullsperger, M., Crone, E. A., & Nieuwenhuis, S. (2004). The role of the
medial frontal cortex in cognitive control. science, 306(5695), 443-447. Sumter, S. R., Bokhorst, C. L., Steinberg, L., & Westenberg, P. M. (2009). The
developmental pattern of resistance to peer influence in adolescence: Will the teenager ever be able to resist?. Journal of Adolescence, 32(4), 1009-1021.
Williams, K. D., & Jarvis, B. (2006). Cyberball: A program for use in research on
interpersonal ostracism and acceptance. Behavior Research Methods, 38(1), 174-180. Van der Veen, F.M., van der Molen, M.W., Sahibdin, P.P., & Franken, I.H.A. (in press) The Heart-Break of Social Rejection Versus the Brain Wave of Social Acceptance. Social Cognitive and Affective Neuroscience.
Van Noordt, S. J., & Segalowitz, S. J. (2012). Performance monitoring and the medial
prefrontal cortex: a review of individual differences and context effects as a window on self-regulation. Frontiers in human neuroscience, 6.
Wu, Y., & Zhou, X. (2009). The P300 and reward valence, magnitude, and expectancy in outcome evaluation. Brain research, 1286, 114-122.
Yeung, N., Holroyd, C. B., & Cohen, J. D. (2005). ERP correlates of feedback and reward processing in the presence and absence of response choice. Cerebral Cortex, 15(5), 535-544.
Figure captions
Figure 1: Schematic of the experimental design
A: Possible combinations of feedback received in the Social Judgment task and the Incentive Judgment task. B: Example trial. Adapted from Dekkers et al. (2013)
Figure 2: Grand average ERP waveforms for both tasks at electrodes Fz and Pz, which were also analysed by Dekkers et al. (2013).
Left panels: Social Judgment Task. Right panels: Incentive Judgment Task. Upper panels: Electrode Fz. Lower panels: Electrode Pz.
Figure 3: Grand average ERP waveforms for both tasks at electrodes Fpz and Cz, where the MFN and P300, respectively, amplitudes were maximal.
Left panels: Social Judgment Task. Right panels: Incentive Judgment Task. Upper panels: Electrode Fpz. Lower panels: Electrode Cz.
Figure 4: MFN amplitude at electrode Fz.
Error bars indicate the standard error of the mean (SEM). SJ = Social Judgment task, IJ = Incentive Judgment task.
Figure 2
Figure 3
Figure 4 Positive Negative -6 -5 -4 -3 -2 -1 0 SJ - Congruent SJ - Incongruent IJ - Congruent IJ - Incongruent Valence A m p lit u d e ( µ v)
