Abstract
We examined how leaders’ expressions of emotion and emotion-related appraisals in their speeches were associated with subsequent political aggression by their groups. We obtained records of speeches anchored to identified acts of aggression and selected for analysis those speeches that were available at three points in time prior to those acts. We then coded the speeches for their expressions of emotion and emotion-related appraisals and tested the differences in that content separately for groups that committed acts of aggression and those that did not, which we labeled acts of resistance. Leaders’ expressions of contempt and disgust and the appraisals related to them differentiated the two groups. We discuss the unique potential contributions of expressions of contempt and disgust to aggression and violence.
One way leaders motivate their followers to action is by strategically expressing emotions and emotion-related concepts in their language. The communication of emotion by leaders is important because emotions are rapid information-processing systems that aid individuals in making decisions and engaging in action with minimal conscious awareness (Tooby & Cosmides, 2008). They are immediate, unconscious, involuntary, and transient reactions that occur as a result of an appraisal of an event that has implications for the welfare of the organism and require immediate response (Ellsworth & Scherer, 2003; Lazarus, 1991). They are a major source of motivation (Frijda, Kuipers, & ter Schure, 1989; Tomkins, 1962, 1963) because they prime behaviors by initiating unique physiological signatures and mental structures (Levenson, 1999, 2003). Understanding the roots of group decisions and actions, therefore, requires an understanding of how emotions are communicated to groups by their leaders.
Emotion communication can occur directly by carefully using words to express emotions toward the out-group and indirectly by expressing various appraisal dimensions related to emotion in order to interpret out-group actions. Appraisals are the themes by which events are evaluated to elicit emotion (Ellsworth & Scherer, 2003; Lazarus, 1991). Anger, for instance, is elicited by appraisals related to goal obstruction, injustice, or norm violations (Matsumoto & Hwang, 2013). The direct expression of emotions can occur through the choice of words (e.g., angry, happy, afraid) as well as nonverbally through faces, voices, gestures, and body language. Emotions can also be expressed through emotion-laden attitudes, values, beliefs, opinions, and metaphors. Referring to “the evil axis,” for example, conveys anger and contempt (along with appraisals of injustice and violations of social hierarchies), while references to “infidel dogs” conveys disgust (as some animals are perceived as unclean, contaminated objects).
By expressing emotions and their appraisals, leaders set the tone for groups to interpret events in ways that facilitate certain decisions and behaviors. Leaders do this by creating narratives based on their appraisals (or reappraisals) of critical events and situations (see Sternberg, 2003, for a discussion on the role of stories in inculcating hatred) and by communicating emotion-related appraisals and expressing emotions in those narratives. Consider, for example, this excerpt of a speech by Mao Tse Dong on July 21, 1966, about the Chinese Cultural Revolution (Mao, 1966):
I say to you all: youth is the great army of the Great Cultural Revolution! It must be mobilized to the full. After my return to Peking I felt very unhappy and desolate. Some colleges even had their gates shut. There were even some which suppressed the student movement. Who is it who suppressed the student movement? Only the Pei-yang Warlords. It is anti-Marxist for communists to fear the student movement. Some people talk daily about the mass line and serving the people, but instead they follow the bourgeois line and serve the bourgeoisie. The Central Committee of the Youth League should stand on the side of the student movement. But instead it stands on the side of suppression of the student movement. Who opposes the great Cultural Revolution? The American imperialists, the Soviet revisionists, the Japanese revisionists and the reactionaries. (Italics added)
The emotional language used involves direct references to words such as “unhappy,” “desolate,” and “fear” and indirect references to words such as “suppression,” Warlords,” “bourgeoisie,” “imperialists,” “revisionists,” and “reactionaries.” This language facilitates certain ways of thinking and feeling in listeners that allow for emotion sharing to occur, priming potential future action.
Specific emotions facilitate specific thoughts and actions because each emotion has its own action tendency (Frijda et al., 1989) that rearranges the body’s physiological and cognitive priorities to prepare for actions (Levenson, 1999). In the case of aggression or violence, previous research has suggested that two emotions—anger and fear—may be key. Anger is considered a dangerous emotion (Averill, 1983; Berkowitz, 1990; Halperin & Gross, 2010), and the state of anger and offensive aggression activate the same brain areas (Harmon-Jones & Sigelman, 2001). Anger motivates violent responses to goal blockage, truncating ongoing transgressions by others and deterring additional ones (Fessler, 2010). Fear is also dangerous because it can be transformed into aggression in both humans and animals (Lorenz, 2002; Moyer, 1968), especially when escape from the fearful stimulus is not possible or desirable.
But expression of the combination of anger, contempt, and disgust may play an important and heretofore unrecognized role in understanding aggression and violence because of their unique sociomoral and sociocultural functions and their purported roles as components of hatred. These emotions are elicited by violations of moral codes, including individual rights, autonomy, and self-relevance (for anger); communal codes and hierarchy, and other’s incompetence or lack of intelligence (for contempt); and purity, sanctity, and moral untrustworthiness (for disgust) (Hutcherson & Gross, 2011; Rozin, Lowery, Imada, & Haidt, 1999). Sternberg (2003) proposed that anger, contempt, and disgust comprised the three components of hate, which in turn contribute to aggression and violence. We propose that expressions of appraisals related to and the emotions of anger, contempt, and disgust are associated with acts of physical aggression against others because these emotions comprise the three components of hatred, and followers who share their leader’s anger, contempt, and disgust are more easily motivated toward the devaluation and destruction of others. Through the careful use of language (and nonverbal behaviors) associated with these emotions, leaders are in a position to motivate, escalate, or defuse situations, and incite or suppress action, through emotion.
We tested these ideas in a study examining the emotions expressed in the words used by world leaders and leaders of ideologically motivated groups talking about their archrival out-groups in their speeches. We obtained records of such speeches, anchored to an identified act of aggression, and selected for analysis speeches that were available at three points in time prior to those acts to test the hypothesis that verbal expressions of anger, contempt, and disgust and their related appraisal dimensions are associated with violence and hostility against the out-group. We also included for comparison a small group of acts and speeches of ideologically motivated groups that had despised opponent out-groups but did not result in violence. We analyzed the speeches for their emotional content and tested the differences in that content separately for groups that committed an act of aggression (AoA) and those that did not, which we labeled as acts of resistance (AoR), and tested the following hypotheses:
Hypothesis 1: Speeches associated with AoAs will have significantly greater expressions of appraisals related to anger, contempt, and disgust compared with AoRs.
Hypothesis 2: Speeches associated with AoAs will have significantly greater expressions of anger, contempt, and disgust compared with AoRs.
Method
Source Acquisition
We first identified AoAs committed by ideologically motivated groups using the following criteria: (a) the act was motivated by ideological motives, including racial and political; (b) the act was not an immediate retaliation to an act of aggression by the other party; (c) the act was a violent action against a defined out-group, with the intention of causing physical harm, reduced quality of life, and/or denial of basic human rights; and (d) there was a clear leader of the group who made speeches across multiple points in time. For comparison purposes, we also identified nonviolent acts of resistance by ideologically motivated groups using the following criteria: (a) the act was motivated by ideological motives, including racial and political; and (b) the act was a nonviolent action against a defined out-group without the intention to cause physical harm, reduced quality of life, and/or denial of basic human rights of others. To find these acts we consulted historical subject matter experts; accessed published resources with lists of historical and contemporary acts of aggression and resistance; accessed web-based resources of governmental agencies such as the CIA and FBI, as well as nongovernmental websites such as Globalconflict.org; and contacted authors of books or papers on related subjects to seek guidance both about which subjects to consider and to learn of sources for textual data. We also used news of current events from U.S. and international media sources.
When potential acts were identified, we then searched for texts of speeches at three different points in time: 3, 6, and 12 months before the event. These time periods were chosen as we considered 1 year as an adequate range of time to see possible changes in expressed emotions. For this study, we included only those acts and groups for which at least one speech text was found for all points in time. When sources were not originally in English, we used available English translations. This resulted in the acquisition of source material from 20 AoAs and 5 AoRs (Table 1). The total across all speeches and events included 7,800 sentences and 191,763 words (1,682 sentences and 45,061 words came from AoRs).
Listing of Acts of Aggression (AoAs) and Acts of Resistance (AoRs) Used in this Study.
Segment Identification
Once source material was acquired, it was necessary to identify the specific segments of each speech that were related to the out-group because speeches were generally about various issues, much having nothing to do with the out-group. To identify speech segments that contained references to the out-group, coders were trained in the background of the events and how to identify instances when the speaker was referring to the out-group. We captured not only direct nominal references to the out-group—such as Osama bin Laden using the words United States, America, or Zionists—but also more subtle, categorical references such as “infidel,” “imperialist,” or “enemies of freedom.” Coders were also trained in identifying oblique references that more sophisticated politicians might make when referring to an out-group, including references to a group according to a problem they create for the in-group (e.g., when a Russian prime minister refers to “threats to the safety and well-being of former citizens of the Soviet Union in the Caucasus” when referring to Chechen rebels, or when a Chinese leader talks about the “territorial integrity of China,” which may refer to dissent in one of a number of regions, such as Tibet and Taiwan).
Two coders independently read each obtained document, annotating the start and end points of text passages in which the out-group was mentioned. The coders then compared annotations and produced an arbitrated listing of them. Texts in which both coders agreed on the out-group identification were selected for further analysis. These text extractions ranged from several sentences to more than 10 pages in length.
Emotion Annotation
The speeches included many indirect emotion and appraisal references that were not analyzable according to current technologies such as the Linguistic Inquiry and Word Count (LIWC; Pennebaker, Francis, & Booth, 2001) or Leximancer (Smith & Humphreys, 2006). Thus, we created a coding scheme specifically for this study that consisted of two different aspects of emotions and their expressions. The first involved the coding of emotion-related appraisal dimensions, which were defined as the underlying, evaluated theme that elicits a particular emotion. The appraisal dimensions coded were extracted from those associated with anger, contempt, disgust, fear, and positive emotions as described by emotion appraisal researchers (e.g., Ellsworth & Scherer, 2003; Lazarus, 1991). Particular attention was paid to the appraisals associated with anger, contempt, and disgust posited by Rozin, Lowery, et al. (1999) and included Obstruction (for anger, with subcategories Divisive, Inaction, Deception, and Liabilities), Injustice (for anger, with subcategories Unfairness, Crime, and Violence), Superiority (for contempt, with subcategories Strength, Beneficial Influence, and Virtues), Inferiority (for contempt, with subcategories Weakness, Corrupting Influence, Vices, and Name Calling), and Intolerability (for disgust). The dimension Threat was coded for fear. The following appraisal dimensions were coded for positive emotions: Advancement (with subcategories Harmonious, Action, and Assets), Peace-Seeking, Justice-Seeking (with subcategories Defending/Avenging, Effective Justice, and Capacity), and Humility. A Not Applicable category was also included (used in less than 1% of the coding). All subcategories were summed within categories for analyses.
The second type of coding concerned the specific emotion expressed in the language. Emotion classifications were inspired by Ortony, Clore, and Collins’s (1990) model, which assigned emotions according to whether they pertained to an outcome, action, or object/person, and denoted the categories of pleased versus displeased and approval versus disapproval. Based on those distinctions, emotions expressed in the text were coded as Joy, Hope, Pride, Admiration, Gratitude, Compassion, Distress, Fear, Anger (Reproach in Ortony et al.’s, 1990, classification scheme), Contempt, or Disgust. Coding of Anger, Contempt, and Disgust allowed us to test directly our hypotheses concerning these emotions; coding of the other emotions allowed us to examine whether they functioned differently than anger, contempt, and disgust. A Not Applicable category was also included (used in less than 1% of the coding).
A single coder annotated each sentence in the sample for both the appraisal dimensions and emotion expressed. Interrater reliability was established with a second coder who coded a randomly selected sample of 20% of the speech texts. Reliability was adequate for both appraisals and emotions (κ = .85 and .88, respectively). Both coders were blind to the identification of the time frame, speaker, and source of the text (i.e., AoA vs. AoR) coded, and were instructed to score the emotions from the writer’s perspective, not their own.
To avoid violating assumptions of independence in the data, the appraisals and emotions were tallied and then averaged across sentences and speeches within each event and time frame to generate a set of codes for each point in time for each event. Thus, the final codes used in the analyses were the codes for each appraisal and emotion averaged across speeches for each event separately for each of the three time periods.
Results
Hypothesis 1: Speeches Associated With AoAs Will Have Significantly Greater Expressions of Appraisals Related to Anger, Contempt, and Disgust Compared With AoRs
We computed Time (3) by Group (2: AoA vs. AoR) mixed ANOVAs separately on Obstruction, Injustice, Superiority, Inferiority, and Intolerability, as these were the appraisal dimensions related to anger, contempt, and disgust (see Table 2 for description). No effects involving Group were significant for Obstruction, Injustice, or Superiority. The main effects of Group on Inferiority and Intolerability, however, were significant—F(1, 24) = 5.88, p < .05,
Means (and SD) for Appraisal Dimensions and Expressed Emotions Related to Anger, Contempt, and Disgust at Three Time Periods.
Because of the small sample sizes, we retested the Group marginal means of the five target appraisal dimensions using bootstrapped t tests. These were significant for Obstruction, Inferiority, and Intolerability (all ps < .05), and marginally significant for Superiority (p < .06). AoRs had significantly higher scores than AoAs on Obstruction, but AoAs had higher scores than AoRs on Superiority, Inferiority, and Intolerability. Table 3 presents 95% confidence intervals (CIs), separately for AoAs and AoRs. Note that the CIs for Inferiority and Intolerability (the appraisal dimensions associated with contempt and disgust, respectively) did not overlap between the two groups.
95% Confidence Intervals of the Marginal Means for the Target Variables, Separately for AoAs and AoRs.
Note. AoA = act of aggression; AoR = act of resistance.
We also computed Mann–Whitney U tests comparing the two groups on the marginal means. The Us were significant for Obstruction and Inferiority, and marginally significant for Intolerability in the same directions reported above. Combined with the bootstrap procedures, these findings provided some mitigation against the small sample sizes.
Because the target appraisal dimensions were likely intercorrelated, we also computed logistic regressions using Group (i.e., AoAs vs. AoRs) as the dependent variable and the marginal means of the five target appraisal dimensions as predictors (simultaneous entry) with bootstrapping. The overall model was significant, χ2(5, 25) = 19.47, p < .01, and classification statistics indicated correct classification for all but one case (96%). Bootstrapped tests of the regression coefficients indicated that the coefficients for Obstruction, Injustice, Inferiority, and Intolerability were significant (all ps < .05), and marginally significant for Superiority (p < .08).
For comparison purposes, we computed the same mixed ANOVAs reported above on the nontarget appraisal dimensions —Threat, Advancement, Peace-Seeking, Justice-Seeking, and Humility. None produced any significant main effects of Group (
Thus consistent with prediction, AoAs had higher scores on contempt- and disgust-related appraisals than AoRs, but contrary to prediction, AoRs had higher scores on anger-related appraisals than did AoAs. These findings provided partial support for Hypothesis 1.
Hypothesis 2: Speeches Associated With AoAs Will Have Significantly Greater Expressions of Anger, Contempt, and Disgust Compared With AoRs
We computed Time (3) by Group (2: AoA vs. AoR) mixed ANOVAs on each of the expressed emotions as well. There was a marginally significant main effect of Group on Anger, indicating that AoRs tended to have higher scores on Anger than AoAs, F(1, 22) = 3.15, p < .10,
As above we retested the Group marginal means of the target emotions using bootstrapped t tests. These were significant for Anger, Contempt, and Disgust (all ps < .05) in the same directions reported above. Note in Table 3 that none of the CIs for these three emotions overlapped between the two groups.
We also computed Mann–Whitney U tests comparing the two groups on the marginal means. The Us were significant for Anger and Contempt, and marginally significant for Disgust in the same directions as reported above.
We computed logistic regressions using Group (i.e., AoAs vs. AoRs) as the dependent variable and the marginal means of the three target emotions as predictors (simultaneous entry) with bootstrapping. The overall model was significant, χ2(3, 25) = 14.80, p < .01, and correctly classified 84% of the cases. Bootstrapped tests of the regression coefficients indicated that the coefficients for Anger and Contempt were significant (ps < .05), and marginally significant for Disgust (p < .06).
For comparison purposes we also computed the same mixed ANOVAs on the other expressed emotions. The main effect of Group on Admiration was significant, F(1, 24) = 4.66, p < .05,
We also computed logistic regressions using Group (i.e., AoAs vs. AoRs) as the dependent variable and the marginal means of the nontarget emotions as predictors (simultaneous entry) with bootstrapping. The overall model was significant, χ2(8, 25) = 25.02, p < .01. Bootstrapped tests of the regression coefficients for the individual variables indicated that the regression coefficients for Joy, Admiration, and Fear were significant (all ps < .05), indicating that AoRs had higher amounts of these emotions than AoAs.
These findings provided partial support for Hypothesis 2 and indicated that, consistent with prediction, AoAs had greater expressions of contempt and disgust than AoRs but that contrary to prediction AoRs had greater expressions of anger than AoAs.
Post Hoc Analyses
Different aspects of the English language have changed across time (Aitcheson, 1991; Romaine, 1999), and given the fairly large time span of events in our sample, it may have been possible that language reflecting emotion may have differed across the events studied. We examined this by splitting our data set at a natural division point in our data set—pre– and post–World War II. This point in history produced geopolitical, population, and industrial changes around the world that coincided with changes in the spread and use of English (Baugh & Cable, 2002). It also produced an even distribution of events within our sample. Thus, we recomputed the ANOVA analyses reported above separately for events occurring before and after the end of World War II. Essentially the same findings were obtained.
To examine if any of the appraisals or emotions changed across time, we computed paired t tests and Wilcoxin signed ranks tests on each of the targetdependent variables, testing the difference between 6 and 3 months prior to the event separately for AoAs and AoRs. AoRs had a significant decrease in Obstruction from 6 to 3 months prior to the event on both the paired t-test, t(4) = 6.61, p < .01, d = 2.96, and the Wilcoxin, p < .05. AoRs also had a significant decrease in Anger from 6 to 3 months prior to the event on the Wilcoxin, p < .05; this effect was marginally significant on the paired t-test, t(4) = 2.10, p < .10, d = .94. No other effect was significant.
Discussion
The findings provided partial support for both hypotheses. Consistent with predictions, speeches associated with AoAs had significantly greater appraisals related to contempt and disgust (Inferiority and Intolerability), and significantly more expressions of contempt and disgust, than AoRs. (The means for Intolerability and Disgust were quite low and may be affected by floor effects; readers are cautioned to interpret these findings with this caveat.) These findings were obtained using standard parametric tests, bootstrapped parametrics, nonparametrics, bootstrapped logistic regressions, and when separate analyses of the events pre– and post–World War II were conducted.
These findings were not produced without limitation, perhaps the biggest of which concerned the selection of the events analyzed and the small sample sizes. Although there were many other events from which to potentially sample, our requirement of obtaining source materials at three points in time reduced the number of events that were usable for study. In fact, the entire corpus we dealt with included a considerable number of speeches and text, but in order to not violate assumptions of independence codes were averaged across sentences and speeches. Also we were limited by our coding systems to analyzing English language texts, which required us to find translations of some speeches. It is possible that the exact emotional content of the non-English speeches were not conveyed validly in the translations, which may have confounded the results. Future studies of non-English texts conducted in the target language without translation can address this important issue and would provide evidence for the possible pancultural universality of our findings.
Another limitation of our findings had to do with our choice of analyses. A better analytic strategy would have involved either a discriminant or logistic regression involving all emotion and appraisal variables at all three time frames and their interactions as predictors, or a causal model based on time series analysis. Factor analyses of the appraisal and emotion variables may have resulted in a reduced set of variables to test. We chose not to conduct these analyses because they required a much larger cases-to-predictors ratio than was available in our study to produce reliable results. (Recall that in order to avoid violations of independence, data were averaged across sentences and speeches within time frame and event, and events were cases, resulting in a total N of 25.) For this reason, and because we had specific hypotheses about specific appraisals and emotions, we chose to report univariate analyses so that readers could determine what occurred on the level of individual variables (although the use of many single variables in a small sample is likely to lead to a number of Type II errors). Because we were sensitive to the relatively low power afforded by the small sample size, we augmented the analyses with bootstrapping procedures and nonparametric statistics. And as a preliminary step to using more powerful parametric statistics we presented log regressions separated by target and nontarget appraisals and emotions, aggregated across time frames and supplemented by bootstrapping procedures, reckoning that these would not be as egregious a violation of the cases-to-predictor ratios. Nevertheless, readers are cautioned to interpret the findings from these analyses vis-à-vis the low cases-to-predictors ratio. Future research involving larger numbers of cases will be able to address the very important question concerning the causal relationships between expressed emotions and appraisals across time as they lead to violence.
Regardless of these limitations, the findings provided interesting insights into the relationship between emotion and language and the contribution of expressions of anger, contempt, and disgust to aggression and violence. That AoRs had greater appraisals and expressions of anger than AoAs was unexpected given the previous literature suggesting a key role for anger in aggression (reviewed earlier). One explanation for this finding may be the fact that from 6 to 3 months prior to the event, appraisals related to Obstruction decreased for AoRs but not for AoAs, expressions of Anger decreased for AoRs but increased (nonsignificantly) for AoAs, and appraisals related to Injustice increased (nonsignificantly) for AoAs but remained the same for AoRs. Thus, it may have been the case that the role of anger as a precursor to violence diminished enough for AoRs but remained high enough for AoAs. The elevated levels of expressions of anger and anger-related appraisals for AoRs, combined with the relatively lower levels of contempt and disgust, may itself serve a purpose in political discourse and the motivation of others to some kind of action, just not aggression or violence.
Our findings suggest that contempt and disgust may be the relatively more important active ingredients in the fueling of aggression and violence than anger, or that the combination of anger with contempt and disgust is important. Anger, contempt, and disgust are commonly confused, often co-occur, and have not been differentiated well in the literature. They overlap semantically, and contempt and disgust may be subcategories of a larger semantic category of anger (Shaver, Schwartz, Kirson, & O’Connor, 1987). Observers judging facial expressions of these emotions are more likely to confuse them with each other than with other emotions (Matsumoto & Ekman, 2004). Researchers often group these emotions together; for instance, in the Stereotype Content Model (Cuddy, Fiske, & Glick, 2007), anger, shame, contempt, disgust, frustration, hate, resentment, and uneasy are all categorized into the single label “contempt.” Events eliciting them in combination often occur in real life, although they may be collectively labeled as “anger.” And laypersons confuse these terms in day-to-day usage, as when parents chastise a child by saying “I was disgusted by your behavior” when in fact they really mean they were angry but not physically nauseated.
But these emotions are quite different. As mentioned earlier, anger is triggered by goal obstruction, injustice, or norm violations; contempt, however, is about status and moral or ethical superiority, and disgust is the emotion of contamination. Anger is an emotion about what happened; contempt and disgust are emotions about the nature of the actors and what should be done about them. The function of anger is to remove obstacles, whereas the function of contempt is to make a statement about “inherent” moral superiority; the function of disgust is to eliminate or repulse contaminated objects. These emotions have different physiologies, invoke different mental state changes, and produce different nonverbal expressions (Ekman, 1999). Each is universal across human cultures in their own right, not only in its signal properties (Ekman, 1993; Matsumoto & Willingham, 2006, 2009; but see critiques of this area by Feldman Barrett, 2006), but also in terms of elicitors (Hutcherson & Gross, 2011; Rozin, Haidt, & McCauley, 1999; Rozin, Lowery, et al., 1999).
These differences have very different implications for action. These emotions allow groups and individuals to not only appraise the actions of others but also to make evaluations of the nature of the actors—especially their moral character—and what should be done about them. That is, anger is focused on the act, but contempt says the act or actor is beneath oneself, and disgust says that the act or actor should be eliminated. When people feel these emotions it is easier to make an evaluation that the target of their emotions is inherently bad or contaminated, and that there is no chance for rehabilitation, thus making a permanent assessment of the moral worthiness of the opponent group rather than a temporary judgment about an act committed by that group. While laypersons often do not distinguish among anger, contempt, and disgust, these distinctions are important and provide a road map as to how they can synergize each other to produce acts of violence. Although anger has received the most research attention, we propose that it is contempt and disgust added to anger, and not necessarily anger alone, that leads to aggression and violence.
The extant literature differentiating anger, contempt, and disgust support these ideas. Studies of emotions in interpersonal conflicts indicate that contempt and disgust, not anger, are associated with the breakup of relationships (Gottman & Levenson, 2002; Gottman, Levenson, & Woodin, 2001). Dehumanized targets activate brain regions associated with disgust (Harris & Fiske, 2006), lending credence to the notion that disgust plays an important role in the dehumanization process (Cortes, Demoulin, Rodriguez, Rodrigues, & Leyens, 2005; Demoulin et al., 2004), which may be necessary for violent acts.
The nonfindings on the nontarget appraisals suggested that appraisals related to anger, contempt, and disgust were the major, if not the only, emotion-related appraisal dimensions that differed between the groups. It may have been possible that appraisals related to emotions other than those assessed in this study, such as pride, guilt, shame, or embarrassment, may have differed between the groups, and future research will need to address this possibility. For example, humiliation has been suggested as important to suicide terrorism (Atran, 2003; Linder, 2001); and other emotions may be important in the emotional messages of ideologically motivated groups, but perhaps not directly to aggression and violence. Also there were no differences across time periods analyzed. Another study involving the same source material did in fact report group differences in anger, contempt, and disgust across time, with AoAs increasing in their expressions of these emotions nearer the event while AoRs decreased (Matsumoto, Hwang, & Frank, 2012). The coding scheme used in that study, however, was completely different from the one used here; thus, it was possible that differences between the studies were due to the specific coding scheme used (although the findings concerning elevated levels of contempt and disgust for AoAs were consistent across the studies). Finally, that AoRs had elevated levels of expressed Admiration and Joy compared with AoAs was interesting, and combined with their greater levels of anger but not contempt and disgust suggested that leaders of groups that committed AoRs moderated their anger with some positive emotions but were not very contemptuous or disgusted. But the relatively low levels of Admiration and Joy, and the fact that the appraisal dimensions associated with these emotions did not differ between the groups, may temper this interpretation. That the groups differed on both expressed emotions and appraisals for anger, contempt, and disgust suggested a saturation of the emotion message related to these emotions, providing a stronger basis for the association between emotion language and the possible suggested solutions for course of action (aggression or resistance).
Our findings may have practical application. Monitoring the expression of emotions by group leaders may provide not only early warning mechanisms of impending possible aggression but also a method to gauge the effects of one’s own group’s actions on other groups. For example, how groups interpret the political actions or decisions by another group may be assessed by monitoring emotional expressions about those actions and decisions and may be used to curtail future escalations of dangerous emotions that may lead to eventual violence. Developing systems to assess emotions among members of groups, and at different levels within the groups, may provide ways to gain insights about the degree to which emotion sharing may occur within groups, which may be important for political justification of leader decisions. Such systems may be akin to rumor-monitoring systems that are useful in assessing counterinsurgency operations in many areas of the world, where the battle concerning knowledge and information is as important as kinetic operations. Emotions expressed in social media, for instance, may have given glimpses of the Arab Spring, and of the escalation to mass action in the future.
Emotions expressed in words may only be part of the overall emotional message delivered. Nonverbal behaviors such as facial expressions and tone of voice that accompany the emotionally-laden language may amplify or deamplify the messages delivered by words. It is quite possible, therefore, that when emotionally laden language is imbedded within a rich repertoire of nonverbal behaviors that also portray emotions, the saturation index of the overall emotional message to listeners may be substantially more powerful than just the words alone. Future studies examining the combined emotional signals portrayed in verbal and nonverbal behaviors will address this important issue.
Finally, demonstrating that leaders of ideologically motivated groups express emotions in their speeches does not demonstrate that members of those groups hearing those speeches accurately perceive those emotions as intended. And even if they do, it is an open question as to whether those perceived emotions in turn spur them on to aggression or not. Action requires means, motive, and opportunity. Emotions may provide the motive, but groups require means and opportunity to aggress. Thus, it is clear that emotions by themselves may be necessary but not sufficient to explain aggression. There are many links to the puzzle of how emotions from leaders contribute to group action that need to be addressed in future research. Documenting the emotions expressed by the leaders of those groups is the first link in an emotional chain of events that may turn out to be a helpful predictor of imminent terrorist acts.
Footnotes
Acknowledgements
The authors would like to express their appreciation to the two anonymous reviewers and the editor for their helpful comments on an earlier draft of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This report was prepared with the support of Research Grant FA9550-09-1-0281 from the Air Force Office of Scientific Research.