Altered neural processing of social signals such as angry facial expressions has been associated with increased aggressive behavior, but evidence for this relationship in healthy persons using ecologically valid experimental designs is lacking. We presented socially relevant videos of facial expressions in a functional magnetic resonance imaging (fMRI) version of the well-established Taylor Aggression Paradigm and investigated 41 healthy male participants, of whom 32 were included in the analysis. In each round of this competitive reaction time task, participants observed their opponent while he selected a punishment level for him, bearing either a neutral or angry facial expression. Afterward, participants in turn selected a punishment level for their opponent. Across participants, reactivity of the medial orbitofrontal cortex (OFC) to angry facial expressions was negatively related to aggressive behavior. Within participants and across trials, activity in the anterior cingulate cortex (ACC) was positively related to aggressive behavior specifically in response to angry expressions. Moreover, we found an effect of angry expressions on neural activity patterns during later stages of the task, demonstrating that the effect of angry expressions on neural reactivity is more than just a short-lived, stimulus-driven response. Our results underscore the importance of OFC and ACC for the shaping of socially adaptive responses to provocation.

Introduction

Angry facial expressions serve important communicatory functions in human social interactions (Blair 2003) and their neural processing is closely connected to social adjustment, especially the regulation of aggressive impulses. The medial orbitofrontal cortex (OFC), including ventral parts of the anterior cingulate cortex (ACC), is important for emotional decision making, especially in social interactive contexts (review in Rudebeck et al. 2008) and, indeed, the processing of angry faces is associated with increased activity in right lateral OFC and ACC (Blair et al. 1999). Moreover, orbitofrontal lesions in patients are associated with deficits in the recognition of angry expressions (Blair and Cipolotti 2000), deficits in response reversal (Fellows and Farah 2003) and increases in reactive aggression (Grafman et al. 1996; Anderson et al. 1999; Blair 2012). Compared with healthy controls, patients with intermittent explosive disorder, who are prone to impulsive aggression, show reduced reactivity of the medial OFC to angry faces, increased reactivity of the amygdala, and reduced negative OFC–amygdala connectivity (Coccaro et al. 2007).

Less is known about the role of the OFC in aggressive behavior in nonpathological groups. Decreased activity of the medial OFC has been observed during the imagination of aggressive behavior in healthy subjects (Pietrini et al. 2000). Beaver et al. (2008) found a negative relationship between medial OFC reactivity to angry faces and scores on the subscale “drive” of the behavioral inhibition/activation system scales (BIS/BAS; Carver and White 1994), which has been associated with increased aggressive tendencies (Harmon-Jones 2003; Carver 2004). Moreover, negative connectivity between medial OFC and amygdala in response to angry faces was found to be reduced in high BAS-driven individuals (Passamonti et al. 2008). However, the BIS/BAS system scales are not a measure of aggression and direct evidence for a relationship between the OFC response to angry faces and aggression in healthy populations is to date lacking.

This is in part due to the fact that reactivity to angry expressions has so far not been investigated in the context of an actual social interaction. Recently, there has been a call for more ecologically valid paradigms in neuroimaging studies of social cognition and empathy, with a special emphasis on the link between neural effects and behavior (Zaki and Ochsner 2012).

We therefore investigated the relationship between reactive aggressive behavior and neural reactivity to angry facial expressions in the context of a realistic social interaction in healthy participants. To this end, we adapted an established method for eliciting and measuring reactive aggressive behavior in the laboratory, the Taylor aggression paradigm (TAP; Taylor 1967), such that it was suitable for use in combination with functional magnetic resonance imaging (fMRI). The TAP is set up as a competitive reaction time task, during which the loser of each round gets punished by an aversive stimulus, the intensity of which is determined by the winner.

On a within-subject basis, it has been shown that aggressive responding to provocation in the TAP is associated with increased activity in the dorsal striatum and the dorsal ACC (Krämer et al. 2007). In another fMRI study using the TAP, Lotze et al. (2007) implemented video feedback of the opponent at the end of each trial, when the punishment was administered. They found increased activity in the medial OFC while watching the other person's suffering and a positive correlation between the selected punishment level and activity in the dorsomedial prefrontal cortex.

In our version of the TAP, we showed a video of the opponent at the beginning rather than the end of a trial, showing the moment of his selection of the punishment level for the participant, with either a neutral or angry facial expression. We expected that, in comparison to neutral expressions, angry faces would elicit increased activity in OFC, ACC, and amygdala. Furthermore, we hypothesized that across subjects, reactivity of the medial OFC would be negatively associated with aggressive behavior, while expecting a reverse effect for the amygdala.

As also mentioned above, the processing of angry facial expressions has been associated with activity in the lateral OFC (Murphy et al. 2003). We also tested for a relationship between reactivity of this region and aggression, in order to determine the specificity of the proposed relationship between medial OFC reactivity and aggression.

We also investigated the neural basis of within-subject variability in aggressive behavior and its dependence on the facial expression shown before.

Materials and Methods

Participants

Forty-one healthy male participants were recruited from the local university for participation in this study. We did not include female participants, because men and women might differ in their response to a male opponent. All participants were free of neurological and psychiatric disorders (self-report) and all but 5 were right-handed. All participants gave written informed consent and received 8 Euro per hour as compensation for their participation. The study was approved by the Ethics Committee of the University of Lübeck and performed according to the Declaration of Helsinki.

Video Stimuli

We recorded 60 2-s video sequences, which showed the participants' opponent during his punishment selections for the TAP. The same video sequences were used for all participants, all including the same opponent. At the beginning of each sequence, the opponent looked downward, ostensibly selecting a punishment level via his keyboard. This was followed by a gaze shift into the camera, such that each sequence ended with the opponent looking directly at the participant. During 40 video clips, the opponent's facial expression was neutral. During 20 clips, he showed an angry expression. The opponent's expression was stable throughout each 2-s sequence.

Taylor Aggression Paradigm

Before entering the scanner, participants were introduced to their opponent for the TAP, who was the same for all participants and a confederate of the experimenters. Together with the opponent, participants received written instructions for the TAP and were informed that, at the beginning of each trial, they would be able to observe their opponent via webcam, while he made his punishment selection.

The TAP consisted of 3 runs with 20 trials each. The time-course of a single trial of the TAP is shown in Figure 1. At the beginning of each trial, participants saw a 2-s video sequence of their opponent selecting a punishment level for them. This was followed by the participant's punishment selection (decision phase), the reaction time task and the outcome phase. During the reaction time task, participants had to press a button as quickly as possible in response to a visual cue. During the outcome phase, they were informed whether they won or lost and which punishment level their opponent had selected. In lost trials, this was followed by the punishment stimulus.

Figure 1.

Time-course for one trial of the Taylor aggression paradigm: A video of the opponent is presented, before the participant selects the punishment level. After the reaction time task, the participant learns the decision of the opponent and whether he has won or lost.

Figure 1.

Time-course for one trial of the Taylor aggression paradigm: A video of the opponent is presented, before the participant selects the punishment level. After the reaction time task, the participant learns the decision of the opponent and whether he has won or lost.

We implemented punishment in the TAP as an aversive noise which could be adjusted in terms of loudness on a scale from 1 to 8. Outcome (win or lose) and the opponent's punishment selection were preprogrammed. In order to keep the social interaction plausible, angry trials were distributed unequally across the 3 runs. Three angry trials were presented in the first run, 7 in the second, and 10 in the third run. Within each run, the sequence of angry and neutral trials was randomized. Angry trials were more likely to follow trials in which the participants won (range 12–16 across participants) than trials in which participants lost (range 4–8). During angry trials, the opponent selected punishment levels from 5 to 8 (mean = 6.4) and, during neutral trials, the opponent selected punishment levels from 3 to 6 (mean = 4).

Postexperimental Assessment and Personality Questionnaires

To ensure the videos were perceived as intended, participants were shown 5 videos of each category again outside of the scanner and rated each video on a 19-point Likert scale from −9 to 9 according to how sad, angry, concentrated, and scary they perceived the person in the video to be. Valence ratings for angry and neutral videos were compared using paired t-tests.

Afterward, participants filled out a questionnaire assessing any suspicions they might have had about the true aims of the study and additional questionnaires that were not used in the current analyses.

MRI Data Acquisition

Structural and functional MRI images were recorded on a Philips Achieva 3-T scanner (Philips Healthcare, the Netherlands) with a standard 8-channel head coil. Functional images (N = 624) were acquired using a single-shot gradient-echo echo-planar imaging (EPI) sequence sensitive to blood oxygen level–dependent (BOLD) contrast [repetition time = 2500 ms, echo time = 25 ms, flip angle 80°, in-plane resolution 2.5 × 2.5 mm2, image matrix 80 × 80, field of view (FOV) 200 mm, slice thickness 2.5 mm, 47 transversal slices, SENSE factor 2.0]. High-resolution structural images were obtained applying a T1-weighted 3D turbo gradient-echo sequence with SENSE (FOV = 240 mm; matrix = 240 × 240; 180 sagittal slices of 1 mm thickness).

Data Analysis

One participant was excluded from data analysis due to excessive movement (>4 mm) during functional scanning and 2 due to pathological findings in their anatomical images. Five participants were excluded, because they reported suspicion about their opponent's participation in the TAP and 1 participant was excluded, because he did not select a punishment level in more than one-third of trials. Thus, 32 participants (mean age = 23.3 years, ±2.7 years) fulfilled predefined selection criteria and were included in the analyses.

For behavioral data, we compared mean punishment selections between angry and neutral trials to investigate the effect of video condition on aggression. To investigate provocation effects for the overall greater number of angry trials and thus high selections of the opponent in the third block, we also compared mean punishment selections between the first and third block of the TAP, averaged across both neutral and angry trials. Finally, to investigate the effect of feedback concerning the outcome phase of one trial on the selection in the following trial, we performed a two-factor ANOVA with the within-subject factors outcome (win vs. lose) and opponent's selection (low [≤4] vs. high [>4]).

For the analysis of MRI data, we used SPM8, an open-source, Matlab-based toolbox (Wellcome Trust Centre for Neuroimaging, London, UK). Functional images were preprocessed using standard procedures: temporal adjustment for differences in slice time acquisition, motion correction, co-registration of EPI images with T1-weighted structural images, segmentation of structural images, normalization into MNI space and spatial smoothing with an 8-mm kernel.

For each participant, we set up a general linear model (GLM) on the first level, with 2 regressors for the decision phase (angry and neutral), 4 regressors for the outcome phase (angry and neutral for won and lost trials separately), regressors of no interest (the reaction time task and the punishment noise), as well as motion regressors. On the second level, we set up separate GLMs for the decision and outcome phases.

For all whole-brain analyses, we performed a cluster-wise control of family-wise error (α = 0.05 FWE). The cluster-defining threshold was P = 0.001, the critical cluster size was 213 voxels.

Regions of Interest

Since we had a priori hypotheses about effects in medial OFC and amygdala and these are regions susceptible to signal loss, we defined regions of interest (ROIs) for these areas using the following method: using ROI images from the automatic anatomic labeling (images MNI_Frontal_Med_Orb and MNI_Rectus) atlas (Tzourio-Mazoyer et al. 2002), we created masks for the left and right medial OFC including bordering areas of the gyrus rectus. We then applied these masks to the effects of interest contrast (contrasting neutral and angry trials against baseline) for the decision phase at the second level, with a threshold of P < 0.01 uncorrected. This ensures that only voxels which show any stimulus-related activity are included in the analyses. By this method, we created one ROI each for the right (MNI coordinates of ROI center: 15, 18, −14; cluster size 86 voxels) and left (−12, 20, −14; 95 voxels) OFC (Fig. 3A).

For the amygdala, Bzdok et al. (2012) show in a recent meta-analysis, that it can be divided into 3 distinct subregions that differ in their structure, function, and connectivity profile: a laterobasal (LB), a centromedial (CM), and a superficial (SF) nuclei group. The SF group, which includes the anterior amygdaloid area, the amygdalopyriform transition area, the amygdaloid-hippocampal area, and the ventral and posterior cortical nuclei (Amunts et al. 2005), is associated with social information processing (Bzdok et al. 2012). We used this parcellation as implemented in the Anatomy toolbox (Amunts et al. 2005) to create 3 sub-ROIs for the amygdala using the same method described for the OFC ROIs above. We created 6 amygdala ROIs (Supplementary Fig. 1): LB left/right (−26, −4 −22; 291 voxels /30, −3, −23; 291 voxels), CM left/right (−24, −9, −9; 58 voxels/27, −9, −9; 68 voxels) and SF left/right (−18, −4, −16; 188 voxels/21, −3, −15; 177 voxels).

Note that we only used the effects of interest contrast for the definition of ROIs. In all functional analyses, we used the angry >neutral contrast.

In order to test for effects in the lateral OFC, we defined 6-mm spheres around the coordinates reported in Murphy et al. (2003) for reactivity to anger signals (left OFC: −38, 27, −8; right OFC 40, 38, −12).

For ROI analyses, we used the SPM toolbox marsbar (Brett et al. 2002) and a significance threshold of P < 0.05.

Decision Phase

For the decision phase, we focused on a 5-s period beginning with the video onset, contrasting angry >neutral trials on the first level. On the second level, we defined a one-sample t-test for the angry >neutral contrast to test for main effects of video valence.

Using the same contrast, we regressed mean punishment selections against neural reactivity across participants. For our a priori ROIs in the amygdala and the OFC, we calculated Pearson correlation coefficients between the angry >neutral contrast values and mean punishment selections.

In order to test whether within-subject variability in aggressive behavior was related to variability in the neural reactivity to the videos, we performed a parametric analysis with punishment selections, z-standardized across all trials within each participant, as parametric modulator. We contrasted angry against neutral trials to test for brain regions that selectively respond to high selections in angry compared with neutral trials, or vice versa.

Three participants had to be excluded from the parametric analysis, because they failed to show any variance in punishment selection for at least one condition in at least 1 run, interfering with the model fit. Thus, data of 29 participants are reported for this analysis.

Outcome Phase

For the 4-s outcome phase, we defined 4 contrasts against baseline (e.g., angry trial, outcome win) on the first level. On the second level, we defined a 2 × 2 flexible factorial design with the factors condition (angry vs. neutral) and outcome (win vs. lose), with both factors and the interaction included in the design matrix. The factor subject was also defined but not included in the design matrix.

Results

Behavioral Data

Mean punishment level selection across all participants was 4.3 with a standard deviation of 0.9 (punishment levels ranged from 1 to 8). There was no difference in punishment selections between angry and neutral trials (P = 0.43, df = 31), but punishment selections increased across the experiment, with selections for the third block being significantly higher than for the first block (mean selections 4.49 and 3.93; standard error = 0.12; t = 4.54; P < 0.001). Furthermore, analyzing punishment selections in regard to the preceding outcome phase showed a significant effect for the factor opponent's selection (F31,1 = 35.4, P < 0.001). Participants selected higher punishments following high selections by the opponent (means 4.6 and 4.0; standard error = 0.1; t31 = 5.5). There was a trend toward higher punishment selections following lost trials than won trials (means 4.42 and 4.27; standard error = 0.08; P = 0.08). In angry trials, participants were slightly faster in making the punishment selection than in neutral trials (mean reaction times 1.22 and 1.31 s, standard error 0.03, P = 0.01). For the reaction time task, there was no difference in reaction times between angry and neutral trials (P = 0.24).

Postexperimental video ratings of perceived emotions confirmed that angry videos were perceived as angrier than neutral videos (t31 = 18.82, P < 0.001). Angry videos were also perceived as sadder (t31 = 5.96, P < 0.001), more concentrated (t31 = 3.63, P < 0.01), and scarier (t31 = 9.11, P < 0.001).

Functional Imaging Data

Decision Phase

To investigate the general effect of angry facial expressions on neural activation patterns, we contrasted angry against neutral trials for the decision phase of the TAP (5 s following video onset). Activity in the left superior to medial frontal gyrus, left superior temporal gyrus, bilateral inferior frontal gyrus, and right middle temporal gyrus was increased during angry trials (Fig. 2A; Table 1). Activation clusters for the inferior frontal gyri partially overlapped with the lateral OFC ROIs and the angry >neutral comparison was highly significant for these ROIs (P < 0.001).

Table 1

Neural effects for the decision phase with a cluster-wise control of P < 0.05 FWE-corrected

Region Hemisphere MNI coordinates n voxels Cluster-peak t-value 
Angry >neutral 
 Medial frontal gyrus −2, 44, 40 1121 7.24 
 Superior temporal gyrus −44,−56,22 562 5.79 
−48, −28, −2 272 4.75 
 Inferior frontal gyrus −56, 24, 10 872 5.72 
52, 20, 18 1481 5.59 
 Middle temporal gyrus 62, −54, 6 360 5.40 
Parametric modulation angry >neutral 
 Cingulate gyrus −2, 10, 34 213 4.37 
Region Hemisphere MNI coordinates n voxels Cluster-peak t-value 
Angry >neutral 
 Medial frontal gyrus −2, 44, 40 1121 7.24 
 Superior temporal gyrus −44,−56,22 562 5.79 
−48, −28, −2 272 4.75 
 Inferior frontal gyrus −56, 24, 10 872 5.72 
52, 20, 18 1481 5.59 
 Middle temporal gyrus 62, −54, 6 360 5.40 
Parametric modulation angry >neutral 
 Cingulate gyrus −2, 10, 34 213 4.37 
Figure 2.

(A) Effects for the angry >neutral contrast for the decision phase in medial frontal, bilateral inferior frontal, and superior temporal gyri (cluster-defining threshold P = 0.001, 0.05 FWE-corrected at the cluster level). (B) Cluster in the dorsal anterior cingulate cortex which showed a significant difference between angry and neutral trials for the parametric modulation of punishment selections for the decision phase (left column; cluster-defining threshold P = 0.001, 0.05 FWE- corrected at the cluster level; n = 29) and the corresponding mean parametric modulation coefficients for neutral and angry trials. (C)Activation patterns for the angry >neutral contrast for the outcome phase (P < 0.0001 uncorrected, cluster threshold 20 voxels).

Figure 2.

(A) Effects for the angry >neutral contrast for the decision phase in medial frontal, bilateral inferior frontal, and superior temporal gyri (cluster-defining threshold P = 0.001, 0.05 FWE-corrected at the cluster level). (B) Cluster in the dorsal anterior cingulate cortex which showed a significant difference between angry and neutral trials for the parametric modulation of punishment selections for the decision phase (left column; cluster-defining threshold P = 0.001, 0.05 FWE- corrected at the cluster level; n = 29) and the corresponding mean parametric modulation coefficients for neutral and angry trials. (C)Activation patterns for the angry >neutral contrast for the outcome phase (P < 0.0001 uncorrected, cluster threshold 20 voxels).

Across-Subject Variability of Aggressive Behavior and its Neural Basis

In order to test if differences in the neural reactivity to angry videos were related to between-subject variability in aggressive behavior, we performed a regression analysis on the angry >neutral contrast with mean punishment selection as regressor. We found a significant negative correlation between punishment selection and neural reactivity for the left medial OFC ROI (r = −0.42, P = 0.02; Fig. 3B). Moreover, the most aggressive participants show negative contrast values, that is, reduced mOFC activity to angry relative to neutral faces, whereas the least aggressive participants show positive contrast values, that is, increased activity to angry relative to neutral faces. This explains why, averaged across the whole group, we did not find an effect of facial expression on mOFC activity (Fig. 2A). In the right mOFC, this correlation approached significance (r = −0.33; P = 0.06). We found no significant correlations with the amygdala ROIs (all P >0.2), the lateral OFC (P >0.3) or on the whole-brain level. Importantly, the correlation for the left mOFC was not driven by previous trial outcome, as sorting decision phases according to preceding trial outcomes showed no effect for either mOFC ROI.

Figure 3.

(A)Left and right medial orbitofrontal regions of interest. (B)Negative correlation between aggressive behavior and the angry >neutral contrast value for the decision phase averaged across all voxels in the left medial orbitofrontal region of interest.

Figure 3.

(A)Left and right medial orbitofrontal regions of interest. (B)Negative correlation between aggressive behavior and the angry >neutral contrast value for the decision phase averaged across all voxels in the left medial orbitofrontal region of interest.

Since the amygdala ROIs are rather large, reaching into neighboring tissue structures, we additionally performed a voxel-wise regression analysis limited to the bilateral amygdala ROIs, to ensure that we did not miss a true effect. This approach did not yield any significant findings, even at a liberal threshold of P < 0.01 uncorrected, supporting the results from the ROI analysis.

Within-Subject Variability of Aggressive Behavior and its Neural Basis

Since 3 subjects did not show sufficient variability in behavior to allow for a parametric analysis (see Materials and Methods section for details), data of 29 participants are reported. We found a positive effect for the dorsal anterior to middle cingulate gyrus (BA 24/32; Fig. 2B). To understand this effect, we extracted the contrast values for this cluster for angry and neutral trials. As can be seen in Figure 2B, activity in this region showed a positive correlation with punishment selections specifically in angry trials, but was unrelated to aggressive behavior in neutral trials.

Outcome Phase

For the outcome phase, we found significant effects for the factors condition (angry vs. neutral) and outcome (win vs. lose), but no significant interaction.

In angry compared with neutral trials, activity was increased in the left temporal pole, middle temporal gyrus, and precentral gyrus, the right inferior frontal gyrus, fusiform gyrus, superior parietal lobule, and thalamus (Fig. 2C; Table 2). At a significance threshold of 0.0001 uncorrected, cluster threshold 20 voxels, we also found increased activity in some other regions including the left amygdala (Fig. 2C).

Table 2

Neural effects for the outcome phase with a cluster-wise control of P < 0.05 FWE-corrected

Region Hemisphere MNI coordinates n voxels Cluster-peak t-value 
Angry >neutral 
 Temporal pole −36, 24 −22 2964 7.48 
 Inferior frontal gyrus 36, 30, −6 5037 7.10 
 Fusiform/lingual gyrus 38, −48, −20 17217 6.12 
 Precentral gyrus −42, −8, 56 441 5.56 
 Thalamus 6, −4, 6 1973 5.36 
 Superior parietal lobule 32, −54, 52 281 4.93 
 Middle temporal gyrus −48, −32, −6 646 4.48 
Region Hemisphere MNI coordinates n voxels Cluster-peak t-value 
Angry >neutral 
 Temporal pole −36, 24 −22 2964 7.48 
 Inferior frontal gyrus 36, 30, −6 5037 7.10 
 Fusiform/lingual gyrus 38, −48, −20 17217 6.12 
 Precentral gyrus −42, −8, 56 441 5.56 
 Thalamus 6, −4, 6 1973 5.36 
 Superior parietal lobule 32, −54, 52 281 4.93 
 Middle temporal gyrus −48, −32, −6 646 4.48 

For the win >lose comparison, we found increased activity in the right caudate and middle frontal gyrus, as well as the left hippocampus. The opposite contrast (lose >win) showed activation in the bilateral superior temporal gyrus, the right precentral gyrus, the right anterior insula, and the midbrain (Supplementary Table 2). These effects replicate findings from previous fMRI studies using the TAP (Krämer et al. 2007; Beyer et al. 2014) and as they are not relevant for our research question, we do not include them in our discussion.

Discussion

To investigate the relationship between neural responses to angry faces and aggression, we designed a paradigm in which the facial expression was behaviorally significant to the participant and the stimulus eliciting the neural response was also the target for the aggressive behavior. We found that OFC reactivity to angry expressions is negatively related to aggression across the normal range of aggressive behavior in healthy persons. Activity in the dorsal ACC was related to the trial-by-trial adaptation of behavior.

Aggressive Behavior and Neural Response to Angry Facial Expressions

Studies in patient populations with brain lesions in the OFC suggest that aggressive behavior should be negatively correlated with reactivity to angry expressions in orbitofrontal areas and positively with reactivity of the amygdala (Coccaro et al. 2007; Blair, 2012). We found a negative correlation between aggressive behavior and reactivity of the left medial OFC, supporting the hypothesis that stronger reactivity of the OFC to angry faces facilitates more controlled, nonaggressive behavior. This effect was specific for the medial OFC, notably. While the lateral inferior frontal cortex showed clear reactivity to anger signals, reactivity of this region did not vary as a function of aggression.

It has been suggested that the OFC is involved in the processing of complex social stimuli and that responsiveness of this area to negative expressions is important for the shaping of appropriate social behavior (Blair and Cipolotti 2000; Kringelbach and Rolls 2003). A comparison of patients with early (<16 months) and late (adult age) prefrontal cortical damage has shown increased antisocial behavior especially for the patients with early damage, underlining the importance of this brain area for the acquisition of social rules across the lifespan (Anderson et al. 1999). OFC reactivity in our study explained variability in aggression between subjects rather than on a trial-by-trial basis. Thus, low reactivity of this area to negative social signals appears to be a risk factor for aggression, although more studies are needed to investigate the stability of OFC reactivity across time and across different settings.

It should be noted that the correlation effect we found was in the moderate range only. This is not surprising, however, given that the explanatory power of neural activity in one particular brain region for such complex social behavior as aggression cannot be expected to be very high. The selection of the medial OFC and amygdala ROIs was based on clear a priori hypotheses, and the results demonstrate the relevance of the OFC for regulation of aggression. Future studies might also take a more exploratory approach examining the explanatory value of the complete data using multivariate regression analyses, for instance.

We did not find a significant correlation between aggression and amygdala reactivity. Reactivity of the amygdala to angry expressions is known to depend on a number of personality variables (Carre et al. 2012), and the study which found a positive correlation with aggression included a group of pathologically aggressive individuals (Coccaro et al. 2007). Our data suggest that amygdala reactivity might play a less decisive role for nonpathological aggression, at least in competitive reciprocal interactions as were implemented here. In an economic exchange game, Gospic et al. (2011) found a positive relationship between amygdala response to unfair offers and rejection of these offers. Yu et al. (2014) observed increased activity in the amygdala, among other regions, with increasing frustration in a monetary reward paradigm. These findings suggest that in situations involving monetary nonreward rather than aversive punishment, amygdala activity is related to the level of frustration and may also play a role in punishing behavior.

On a within-subject basis, activity of the dorsal anterior cingulate cortex (dACC) was positively correlated with punishment selection during angry trials, whereas this effect was absent during neutral trials. Activity in the dACC has been related to a wide range of different functions, often subsumed under the general term of cognitive control (Shenhav et al. 2013). Different theoretical models have been put forward associating the dACC with conflict monitoring (Carter and van Veen 2007), particularly monitoring of stimulus-response conflicts (Duncan and Owen 2000), encoding of action-reward associations (Rushworth et al. 2011) or allocating control based on its expected value (Shenhav et al. 2013). Our findings suggest that reacting aggressively to an opponent bearing an angry expression constitutes a situation demanding a higher level of cognitive control, possibly because reacting aggressively to a threatening opponent is a high-conflict response. Similarly, in a previous study with the TAP (Krämer et al. 2007), activity in the dACC was found to be increased during aggressive retaliation toward a highly provocative opponent. In the current study, the opponent on average selected higher punishment levels during angry trials, which participants may have learned during the experiment. That is, as in the previous study, we find dACC activity when participants chose to retaliate despite the risk of higher punishment.

Notably, participants did not generally select higher punishments in angry trials. Rather, participants selected higher punishments if the opponent had selected a high punishment in the previous trial. Punishment selections increased from the first to the third block, likely due to the increased number of angry trials with higher punishments in the third block. It would be interesting to investigate interactive effects of the preceding outcome phase and angry versus neutral facial expression on neural reactivity and behavior. However, such an analysis would require greater trial numbers than given in the current study.

Neural Reactivity to Angry Facial Expressions

Independently of aggressive behavior, we found increased activity in response to angry relative to neutral videos in inferior frontal and medial frontal cortex. These regions were close to the lateral OFC and dorsal ACC clusters reported for the corresponding contrast by Blair et al. (1999), supporting the reliability of this effect. Beyond this, we found increased activity in the superior and middle temporal gyrus. It has previously been shown that compared with static images, dynamic facial expressions elicit increased activity in superior temporal sulcus and amygdala, suggesting increased socioemotional processing (Arsalidou et al. 2011).

Contrary to our hypotheses, we did not find increased amygdala activity as an immediate response to angry videos. However, recent studies point toward additional variables related to amygdala reactivity to angry expressions, such as trait anxiety (Carre et al. 2012, 2013), and compared with fearful expressions, angry expressions generally elicit amygdala activity less reliably (Blair 2003).

Interestingly, the facial expression presented at the beginning of a trial also strongly affected neural activity during the outcome phase. After the opponent looked angry during the decision phase, we found increased activity in a range of brain areas associated with mentalizing (temporal pole, IFG, fusiform gyrus, precentral gyrus; Hooker et al. 2008; Zaki and Ochsner 2012) and a trend toward increased activity in the left amygdala. Our results show that the neural reactivity to angry expressions is not just a short-lived stimulus-driven response. Rather, angry expressions serve as context information, shaping the neural response to subsequent socially relevant events.

Conclusions

Our results show that, in normal-range aggressive behavior, the orbitofrontal cortex plays a similar role as has been suggested for pathological aggression. Furthermore, they allow for a differentiation of neural processes related to inter- and intrapersonal variability in reactivity to negative social feedback. Beyond these findings, we replicated results from previous studies, demonstrating the feasibility of using complex and more natural interactive paradigms in social neuroscience research. As shown here, such paradigms can be useful for the direct relation of neural activation patterns to behavior.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/.

Funding

This study was supported by the German Research Foundation (KR3691/5-1) and through intramural funding by the University of Lübeck.

Notes

We thank Susanne Schellbach for help during data acquisition and Shejan von Fintel for his participation as actor. Conflict of Interest: None declared.

References

Amunts
K
Kedo
O
Kindler
M
Pieperhoff
P
Mohlberg
H
Shah
NJ
Habel
U
Schneider
F
Zilles
K
2005
.
Cytoarchitectonic mapping of the human amygdala, hippocampal region and entorhinal cortex: intersubject variability and probability maps
.
Anat Embryol (Berl)
 .
210
:
343
352
.
Anderson
SW
Bechara
A
Damasio
H
Tranel
D
Damasio
AR
1999
.
Impairment of social and moral behavior related to early damage in human prefrontal cortex
.
Nat Neurosci
 .
2
:
1032
1037
.
Arsalidou
M
Morris
D
Taylor
MJ
2011
.
Converging evidence for the advantage of dynamic facial expressions
.
Brain Topogr
 .
24
:
149
163
.
Beaver
JD
Lawrence
AD
Passamonti
L
Calder
AJ
2008
.
Appetitive motivation predicts the neural response to facial signals of aggression
.
J Neurosci
 .
28
:
2719
2725
.
Beyer
F
Münte
TF
Erdmann
C
Krämer
UM
2014
.
Emotional reactivity to threat modulates activity in mentalizing network during aggression
.
Soc Cogn Affect Neurosci
 . .
Blair
RJ
2003
.
Facial expressions, their communicatory functions and neuro-cognitive substrates
.
Philos Trans R Soc Lond B Biol Sci
 .
358
:
561
572
.
Blair
RJ
2012
.
Considering anger from a cognitive neuroscience perspective
.
Wiley Interdiscip Rev Cogn Sci
 .
3
:
65
74
.
Blair
RJ
Cipolotti
L
2000
.
Impaired social response reversal. A case of ‘acquired sociopathy
’.
Brain
 .
123
(Pt 6)
:
1122
1141
.
Blair
RJ
Morris
JS
Frith
CD
Perrett
DI
Dolan
RJ
1999
.
Dissociable neural responses to facial expressions of sadness and anger
.
Brain
 .
122
(Pt 5)
:
883
893
.
Brett
M
Anton
J-L
Valabregue
A
Poline
J-B
2002
.
Region of interest analysis using an SPM toolbox [abstract]
. In:
Presented at the 8th International Conference on Functional Mapping of the Human
Brain
;
2002 June 2–6; Sendai, Japan. Available on CD-ROM in NeuroImage. Vol. 16, No. 2.
Bzdok
D
Laird
AR
Zilles
K
Fox
PT
Eickhoff
SB
2012
.
An investigation of the structural, connectional, and functional subspecialization in the human amygdala
.
Hum Brain Mapp
 .
34
:
3247
3266
.
Carre
JM
Fisher
PM
Manuck
SB
Hariri
AR
2012
.
Interaction between trait anxiety and trait anger predict amygdala reactivity to angry facial expressions in men but not women
.
Soc Cogn Affect Neurosci
 .
7
:
213
221
.
Carre
JM
Murphy
KR
Hariri
AR
2013
.
What lies beneath the face of aggression?
Soc Cogn Affect Neurosci
 .
8
:
224
229
.
Carter
CS
van Veen
V
2007
.
Anterior cingulate cortex and conflict detection: an update of theory and data
.
Cogn Affect Behav Neurosci
 .
7
:
367
379
.
Carver
CS
2004
.
Negative affects deriving from the behavioral approach system
.
Emotion
 .
4
:
3
22
.
Carver
CS
White
T
1994
.
Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment - the BIS BAS scales
.
J Pers Soc Psychol
 .
67
:
319
333
.
Coccaro
EF
McCloskey
MS
Fitzgerald
DA
Phan
KL
2007
.
Amygdala and orbitofrontal reactivity to social threat in individuals with impulsive aggression
.
Biol Psychiatry
 .
62
:
168
178
.
Duncan
J
Owen
AM
2000
.
Common regions of the human frontal lobe recruited by diverse cognitive demands
.
Trends Neurosci
 .
23
:
475
483
.
Fellows
LK
Farah
MJ
2003
.
Ventromedial frontal cortex mediates affective shifting in humans: evidence from a reversal learning paradigm
.
Brain
 .
126
:
1830
1837
.
Gospic
K
Mohlin
E
Fransson
P
Petrovic
P
Johannesson
M
Ingvar
M
2011
.
Limbic justice—amygdala involvement in immediate rejection in the Ultimatum Game
.
PLoS Biol
 .
9
:
e1001054
.
Grafman
J
Schwab
K
Warden
D
Pridgen
A
Brown
HR
Salazar
AM
1996
.
Frontal lobe injuries, violence, and aggression: a report of the Vietnam Head Injury Study
.
Neurology
 .
46
:
1231
1238
.
Harmon-Jones
E
2003
.
Anger and the behavioral approach system
.
Pers Individ Dif
 .
35
:
995
1005
.
Hooker
CI
Verosky
SC
Germine
LT
Knight
RT
D'Esposito
M
2008
.
Mentalizing about emotion and its relationship to empathy
.
Soc Cogn Affect Neurosci
 .
3
:
204
217
.
Krämer
UM
Jansma
H
Tempelmann
C
Münte
TF
2007
.
Tit-for-tat: the neural basis of reactive aggression
.
Neuroimage
 .
38
:
203
211
.
Kringelbach
ML
Rolls
ET
2003
.
Neural correlates of rapid reversal learning in a simple model of human social interaction
.
Neuroimage
 .
20
:
1371
1383
.
Lotze
M
Veit
R
Anders
S
Birbaumer
N
2007
.
Evidence for a different role of the ventral and dorsal medial prefrontal cortex for social reactive aggression: An interactive fMRI study
.
Neuroimage
 .
34
:
470
478
.
Murphy
FC
Nimmo-Smith
I
Lawrence
AD
2003
.
Functional neuroanatomy of emotions: a meta-analysis
.
Cogn Affect Behav Neurosci
 .
3
:
207
233
.
Passamonti
L
Rowe
JB
Ewbank
M
Hampshire
A
Keane
J
Calder
AJ
2008
.
Connectivity from the ventral anterior cingulate to the amygdala is modulated by appetitive motivation in response to facial signals of aggression
.
Neuroimage
 .
43
:
562
570
.
Pietrini
P
Guazzelli
M
Basso
G
Jaffe
K
Grafman
J
2000
.
Neural correlates of imaginal aggressive behavior assessed by positron emission tomography in healthy subjects
.
Am J Psychiatry
 .
157
:
1772
1781
.
Rudebeck
PH
Bannerman
DM
Rushworth
MF
2008
.
The contribution of distinct subregions of the ventromedial frontal cortex to emotion, social behavior, and decision making
.
Cogn Affect Behav Neurosci
 .
8
:
485
497
.
Rushworth
MF
Noonan
MP
Boorman
ED
Walton
ME
Behrens
TE
2011
.
Frontal cortex and reward-guided learning and decision-making
.
Neuron
 .
70
:
1054
1069
.
Shenhav
A
Botvinick
MM
Cohen
JD
2013
.
The expected value of control: an integrative theory of anterior cingulate cortex function
.
Neuron
 .
79
:
217
240
.
Taylor
SP
1967
.
Aggressive behavior and physiological arousal as a function of provocation and the tendency to inhibit aggression
.
J Pers
 .
35
:
297
310
.
Tzourio-Mazoyer
N
Landeau
B
Papathanassiou
D
Crivello
F
Etard
O
Delcroix
N
Mazoyer
B
Joliot
M
2002
.
Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain
.
Neuroimage
 .
15
:
273
289
.
Yu
R
Mobbs
D
Seymour
B
Rowe
JB
Calder
AJ
2014
.
The neural signature of escalating frustration in humans
.
Cortex
 .
54
:
165
178
.
Zaki
J
Ochsner
K
2012
.
The neuroscience of empathy: progress, pitfalls and promise
.
Nat Neurosci
 .
15
:
675
680
.