Theorizing and Measuring Emotions in Conflict

The (New) Passion for Emotion in IR and Conflict Studies

Even though emotions such as resentment or disgust have consistently been evoked by classical figures of IR scholarship, the amplitude of their role, the exact route of their impact, the precise mechanisms of their influence have remained undertheorized until the 1970s. Considered to be an episodic individual factor, emotion was simply “off the radar-screen of international relations theorists” (Crawford 2000, 122). In the 1970s, the importance of psychological determinants in international relations and conflicts started to be more fully acknowledged, but almost only in the narrow sense of their negative impact on decision making. IR theorists conceived emotion in Descartes’s lineage as something that distracts rational judgment (e.g., Jervis 1976). According to this view, emotion short-circuits rational decision making because it proceeds through a supposedly distinct process, as an external factor affecting the usual conduct of reason (Mercer 2005, 92).

Findings in psychology have demonstrated, however, that the rational/irrational (emotional) dichotomy is a nonsensical one; the negative role attributed to emotion appears to be an unsound prejudgment. Advances in neuropsychology and cognitive sciences in the 1990s (e.g., Damasio 1994) indeed made clear that rationality is never empty of emotional input, thereby completing a long-standing “undermining of the image of a unified and rational self” (Kahler 1998, 919). These studies showed that emotion cannot be considered simply as an affect that every now and then hampers otherwise rational judgment but rather as a permanent state holding an enduring and systemic role in every cognitive process. Cognition needs emotion, which does not necessarily play a negative role. Consequently, the mechanistic logic of investigating a direct (and negative) impact of a particular emotion on a specific political decision makes less sense or at least gets highly complicated to investigate.

The study of emotions in international conflicts thereby redeveloped on new bases, thanks to scholars bridging psychology and political science. In 1996, Mercer’s (2005) programmatic International Studies Association (ISA) presentation “Approaching Emotions in International Politics” made clear that the role of emotion in IR was at the same time far from negligible and ill theorized. Crawford later provided a detailed portray of the dynamics and impacts of emotions in IR. Reflecting on this fresh, less simplistic understanding of emotion, she proclaimed, “the time is ripe for addressing emotion in world politics” (Crawford 2000, 130). The 2000s have consequently witnessed a vibrant discussion on the role of emotion in conflict, delivering increasingly sophisticated theoretical models and notable empirical insights. Within this research agenda, an insisting idea is to investigate emotion at a collective level to locate its influence in societal processes rather than solely in individual decision-making dynamics; this crucial idea, however, has been variously interpreted and attempts to test it empirically do not yet rest on sufficiently diverse and direct methods.

Holistic versus Communicational Conceptions of Collective Emotions

Cutting-edge studies on the role of emotion in conflicts converge toward the belief that emotion has a collective dimension, whose role in intergroup conflicts cannot be neglected. Scholars such as McDermott, Bar-Tal, or Halperin have convincingly argued that emotions not only are experienced by individuals but might also appear at the group level. Beyond the consensual opinion that “emotion, while most often experienced at the individual level, is inherently social and relational” (Fattah and Fierke 2009, 70), however, any further agreement is hidden behind a conceptual diversity and fuzziness that prevents the scholar from clearly identifying what a collective emotion is. Terms as various as “affective identity” (Ross 2006), “attitude” (Scherer 2005), collective emotion (Bar-Tal, Halperin, and de Rivera 2007), “collective emotional orientation” (Jarymowicz and Bar-Tal 2006), or “affective composition” (Kelly and Barsade 2001) are used to qualify the collective dimension of emotion. The very concept of “emotion” has lost specificity, as Scherer (2005) laments, “the number of scientific definitions proposed has grown to the point where counting seems hopeless” (p. 696). In this diversity, we nonetheless identify two main conceptions of collective emotion.

First, a holistic view suggests that the society itself might somehow be the subject of an emotion. Insisting on this perceptible collective reality, De Rivera (1992, 197) defines an “emotional climate” as “an objective group phenomenon that can be palpably sensed.” Similarly, Ross argues that “affects” “cut across individual subjects” (2006, 199). For these scholars, a collective emotion is something else than the simple addition of similar individual emotions: it is experienced in a real way by the group. Slightly different accounts suggest that groups are characterized not by their real experience of an emotion but by their propensity to trigger, among their members, particular emotions when a significant event occurs. Scherer (2005, 705) suggests that members of a group possess an identical “affect disposition,” that is, a common propensity to “to react with certain types of emotions even upon slight provocation”; sticking to this holistic perspective, Bar-Tal, Halperin, and de Rivera (2007, 443) define the similar concept of “collective emotional orientation” as “the tendency of a society to express a particular emotion.”

A second viewpoint stresses that the collective aspect of emotion only relates to its communicative dimension, to the process of its dissemination within a group; in a society, it may happen that some people individually experience the same emotion and communicate it in the same way, thereby contributing to its “collectivization.” As Fattah and Fierke (2009) explain, the experience of emotion is individual but is always expressed to others and hence tends to disseminate among group members. While Bleiker and Hutchison (2008) suggest that the external manifestations of emotion experienced at the individual level provoke the collectivization of emotion within a group, which then impacts on political decisions, Kelly and Barsade (2001) highlight leaders’ roles in intersubjective emotional contagion through their display of specific emotional repertoires. In this view, a collective emotion is not strictly speaking a reality but rather is a heuristic concept used to account for the communication and dissemination of emotion.

Both lines of thought have their merit and problem. The second interpretation rightly locates the experience of emotion at the sole individual level and stresses the importance of the communicative aspect of emotion within a group, but it does not fully acknowledge the importance of truly collective components in the triggering of precise emotions among the members of specific societies. Conversely, the first interpretation operates too blurry a boundary between the two distinct spheres of individual emotions and societal dynamics, but by doing so rightly emphasizes the connection between how individuals react emotionally to significant events and how the society to which they belong is organized by particular norms.

Strengthening the Appraisal-based Framework: Circularity, Emotional Worldviews, and Emotional Configurations

Our theoretical model combines the strengths of both approaches and simultaneously avoids their respective weaknesses. In other words, we “attempt to understand how an individual micro-level instinct, an emotion, contributes to collective macro-level processes and outcomes” (Berezin 2002, 33) and the other way round. Important work has already been done in this perspective, which we seek to complete. “Aware of no comprehensive framework for understanding how emotions influence individual and collective beliefs, attitudes and behaviors regarding war and peace” (Halperin, Sharvit, and Gross 2011, 87), Halperin, Sharvit, and Gross recently put forward such an integrated model with the ambition to articulate the social and individual levels. Their appraisal-based framework suggests that individuals within a given group may experience similar specific emotions when a political event occurs, if they appraise this event in the same way. Appraisal could indeed be common as it is determined by three group-based factors, namely “long-term emotional sentiments,” “nonaffective factors,” and “framing.” However, they recognize that their model is in need of further sophistication, recommending that “further research…must address the unique interaction between long-term and short-term emotions and their confluence on conflict related to political process” (Halperin, Sharvit, and Gross 2011, 98).

Our aim is precisely to address this need by strengthening Halperin, Sharvit, and Gross’s original framework with three key improvements. First, we highlight the existence of group-specific “emotional worldviews” and stress their crucial impact. Second, we hypothesize the presence of complex “emotional configurations” determined by the emotional worldviews. Third, we theorize the circularity of group-based emotional dynamics. Thereby strengthened (see Figure 1), the appraisal-based framework becomes a comprehensive model of the interplay between the individual and collective levels and between the short-term and long-term dimensions of emotion in conflict.

OPEN IN VIEWER

Figure 1.

Interplay between the individual and collective levels of emotion in conflict.

The starting point to apprehend the model is the occurrence of a significant political event or the belief or recollection that an event has occurred. However, an event can be appraised in many ways and does not convey, in itself, a specific emotional charge. Group-based dynamics determine which emotions are triggered.

We first suggest that members of any given group are characterized by a common particular way of apprehending and understanding the social world through a particular overarching logic, which we call emotional worldview. ² An emotional worldview can be defined as an all-encompassing and coherent logic of understanding the social world that is based on the group’s collective memory itself made of socially constructed axioms (statements that are taken for granted, that cannot be contested) and myths (real traumatic past events that are perpetually reinterpreted) that are reproduced in language. Worldviews are not especially targeted toward another group but are constructed throughout the group’s history and therefore always contains intergroup components such as the group’s supposed hierarchical status or social role. We suggest that the axioms and myths of the worldview, through which members of a group interpret the social world, determine the cognitive appraisal of the event. First, they possess an “emotional disposition,” that is, an intrinsic propensity to trigger specific emotions. Second, and—later in the cyclic process—they impact the cognitive appraisal because they shape framing and feed long-term sentiments. Individuals experience the specific emotions subsumed by the emotional disposition of their group’s emotional worldview precisely because they are members of this society, so in any given society individuals are prone to experience the same emotions, with small variations due to personality traits. We argue that adding worldview to the original appraisal-based framework highlights the deep interconnections that bind together long-term sentiments, framing, and (seemingly) nonaffective factors. In their framework, Halperin, Sharvit, and Gross define a long-term sentiment as a “temporally stable emotional disposition toward a person, group, or symbol.” However, the source of this enduring predisposition remained to be identified. While nonaffective factors could be this source, the concept remains fuzzy and more importantly Halperin, Sharvit, and Gross do not map any interaction between the two elements. Bar-Tal et al.’s (2012, 41) concept of “ethos of conflict,” understood as a “configuration of central, shared societal beliefs,” is very close to what we describe here, but our use of the term emotional worldview instead aims at connecting with the “worldview problem” in cognitive science, and in particular with the theoretical space unfolded by Lakoff (2002), who showed that individual cognition is dependent of group-specific frames (worldviews), which inherently contain an emotional dimension. ³ In other words, we posit that in every group a specific worldview is shared that determines which long-term sentiments and framing will influence the cognitive appraisal of events. By emphasizing the importance of worldviews as prime determinants of emotions, we claim that seemingly nonaffective factors such as ideologies (which are encompassed in worldviews) actually carry heavy emotional charges—hence the intrinsic “emotional” dimension of worldviews. This conception helps us to better understand why emotion is linked with group affiliation (in line with Mackie, Devos, and Smith 2000).

Second, we suggest that events trigger not only one emotion but rather a complex set of specific emotions as well, which are experienced with defined intensities; we call this the emotional configuration. We indeed believe that one has to consider the occurrence, at the individual level, of an emotional configuration rather than a single emotion felt in a binary way. An emotional worldview may, for example, push individuals to react to an event with high levels of sadness and resentment or with low levels of rage and fear. Emotional configurations are dependent on the emotional worldview because some axioms and myths are prone to produce resentment (as it could be the case with nationalist ideologies that develop in formerly colonized countries), while others might tend to trigger anger (as it might happen with a worldview developed in a country whose power is constantly frustrated by the out-group’s constraints).

Third, we stress the circular dynamics of emotions in conflict, highlighting how the collectivization of emotions within a group influences how those who do not have a direct knowledge of the event appraise it and hence further communicate their emotions (not only through facial expressions and behaviors but also, as Lakoff (2002) showed, through specific language uses). In this process, political leaders and traditional mass (social) media play a key role (Halperin, Sharvit, and Gross 2011, 87). We identify four dynamics that make the collectivization of emotions a cyclic process. First, the worldview and emotional configuration that characterize a group shape the way direct witnesses, leaders and mass (social) media propagate the event, thereby creating a particular framing. The shared character of the worldview and the ensuing homogeneity of emotional configuration ensure the coherence of this framing and its acceptance within group. Second, the dissemination of the “framed” event allows individuals without previous knowledge of the event to also appraise it. The emotional configuration that they will consequently experience and propagate is expected to be similar to that of leaders, direct witnesses, and the media, both because they share a similar worldview and because they appraise the event according to their framing. Third, the dissemination of emotions within the group—this time not only by direct witnesses, leaders, and media but also by the entire population—reinforces the worldview and its emotional disposition. Indeed the recurrence of collectively experienced emotions in response to framed events acts as a powerful mechanism in the confirmation of worldviews (confirming its myths and axioms) and enduring long-term sentiments toward the out-group. Fourth, the collectivization of emotions increases the likelihood of a new conflict-related event.

These four dynamics allow us to understand why the different groups involved in a conflict may have a different understanding of a same event and thereby a different collective emotional configuration, as each group is characterized by a specific worldview which is continually reinforced through the collectivization of emotions. Moreover, these four dynamics increase the coherence of emotion configurations within a group, as the group-specific worldviews shape the appraisal of the event both directly and indirectly via the collectivization of emotions, the reinforcement of long-term sentiments and the framing of the event during its mediatization. In our opinion then, acknowledging the reality of collective emotion in conflict is to document the statistically attested presence in a society of a specific emotional worldview and the ensuing simultaneous experience, by individuals within a given society, of the same emotions when confronted to the same politically salient event.

Like the original appraisal-based framework, it further strengthens; this model is not testable in its entirety with a single empirical study. Our empirical analysis of the Israeli–Palestinian conflict aims at illustrating two aspects that we have added to the appraisal-based framework, namely emotional worldviews and emotional configurations; by doing so and by focusing on political leaders and social media, it also offers a first look at the circular dynamics of emotions collectivization.

Emotion in Politics: Insufficiencies in Method

IR scholars willing to integrate emotion into the analysis of conflicts have not yet accompanied their assertions with methods for empirical testing. The situation is paradoxical: while all major IR contributors have lamented on the lack of clear methodological procedures to study emotion in politics, they have not provided clear-cut, feasible operational guidelines. The absence of precise methodological suggestions in McDermott’s (2004) chapter on emotion in her Political Psychology in International Relations, or Ross’s confession that “methodological problems demand more attention that I am able to give them here” (2006, 212), is particularly telling. As Bleiker and Hutchison state (without providing a solution), “strangely there are still only very few systematic inquiries into emotions and even fewer related discussions on method” (2008, 115).

Conflict scholars with stronger ties with psychology have been more willing to design and implement adequate methods. Psychology offers several methods and techniques for the study of emotion, whose respective pertinence depends on what “component” (Scherer 2005) is scrutinized. However, the implementation of many of these tools, most famously medicotechnical devices such as electroencephalographs (EEGs) and magnetic resonance imagings (MRIs; ⁴ responsible for most of the “‘emotional revolution’ in psychology,” to use the words of Crawford [2000, 119]) is close to impossible in real political contexts and conflict settings. Computer-based facial expression capture and identification software developed in line with Ekman’s work (e.g., Ekman 1993) are interesting alternatives and provide noteworthy results, but their use is highly technical and is limited to the analysis of high-quality video images. Because of the difficulty to implement these techniques in real-life contexts of conflict, the majority of findings related to the role of emotion in such contexts have been provided by questionnaires and minimal-group experiments (e.g., Bizman and Hoffmann 1993; De Steno et al. 2004; Halperin 2008; Halperin et al. 2011). Although this latter effort has been pursued with rigor and did bring about the most robust insights available to date, it remains somehow artificial: data collected through minimal-group experiments cannot perfectly be generalized to real-life settings, and data obtained through questionnaires come from the individuals’ own accounts and not from a direct measurement of the actual experience of emotion that an actual event can trigger. Findings obtained through these methods would gain in being triangulated with data acquired from a direct measuring of emotion.

Analyzing Language, Directly Measuring Emotion: Using the LIWC

Among the three ways of directly examining emotions identified by Ohman (2006, 34), ⁵ “the uniquely human ability of language provides a means for people to make their feelings known to the outside world” and therefore becomes a privileged target for the scientific analysis of collective emotion. We follow Fattah and Fierke (2009, 70) who showed that “emotion finds expression only in a language and a culture.” In addition to this advantage in terms of practical feasibility, the study of emotional expressions in language has another advantage over the study of emotional expressions in behavior. ⁶ Verbal cues not only crystallize the presence of specific affects but also simultaneously provide insight on the prescriptive beliefs, the evaluative judgments, and the worldviews of authors and the group they belong to (Lakoff 2002). Hence, a language-based emotion analysis is apt at combining two dimensions, what is said and how it is said, without neglecting the former, as do behavior- or physiology-based studies. It is therefore perfectly suited to a study of interdependent emotional worldviews and emotional configurations.

To implement a language-based emotion analysis, we selected what is arguably the most robust software to detect emotion in discourses, the LIWC, and adapted it to the specificities of our theoretical propositions. As regards its theoretical principles, the LIWC is based on the idea that language use reflects the psychological state of the writer/speaker, in other words that “the words we use in daily life reflect what we are paying attention to, what we are thinking about, what we are trying to avoid, how we are feeling, and how we are organizing and analyzing our worlds” (Tausczik and Pennebaker 2010, 30). As regards its functioning logic, the LIWC “links daily word use to broader social and psychological processes” (Tausczik and Pennebaker 2010, 38), calculating the frequency, in any given text, of precise words and kinds of words that research in psychology has shown to be indicators of specific emotions; the higher the frequency of certain words and kinds of words, the most salient the associated emotion. A particularly high/low use of personal pronouns, of certain adjectives and nouns, or of words that have four syllables or more are the kinds of cues that the LIWC calculates when trying to assess and map the emotional tone of a text (including political speeches, parliamentary minutes, or political blog contributions).

This technical device, which is not new as such but has never been integrated and adapted in conflict studies, has the crucial advantage of studying emotion directly in political context, through the prism of language. It allows the analysis to rest on nonartificial, direct, large, and comparable data samples and to take into account a variety of different actors, some of them not available for questionnaires or testing. This approach is thereby particularly well suited to study the collective dimension of emotion. The collectivization of emotions inside a particular group (or across various groups) can be attested by examining with the LIWC multiple documents emanating from its members. For these reasons, we import the LIWC from its initial domain to study emotion in political conflict contexts. The LIWC, however, has a shortcoming, that is, it is built on a narrow definition of emotion, due to its original use in counseling. We therefore had to alter the initial functioning of the software by making it able to detect not only emotions but also emotional worldviews.

Since its first developments, the software has been perfected and its scientific pertinence and validity secured (Kahn et al. 2007; Tausczik and Pennebaker 2010). The LIWC has already been used to document the presence and impact of emotion in politics broadly speaking but only rarely and within narrow research designs. It has nonetheless already provided promising insights. For example, Taylor and Thomas (2008) have demonstrated the importance of matching stakeholders’ “linguistic styles” to secure successful negotiations. Pennebaker and Lay (2002) have highlighted the chronological variations in Rudolph Giuliani’s emotions throughout the press conferences of his mandate, Pennebaker, Slatcher, and Chung (2005) have contrasted the emotions expressed in John Kerry’s, John Edward’s and Al Gore’s interviews, and Baele and Sterck (2015) have studied the security rationale of EU immigration policies. Considering the political dimension of social media, Cohn, Mehl, and Pennebaker (2004) have assessed the emotional response of bloggers after 9/11. Given all this, it does not seem exaggerated to claim that we should be able to measure the emotional tone of political texts and contexts in international politics through the use of the LIWC software.

A Two-step Method

Coherently with our model, our method has to start with a significant political event; it then proceeds in two steps, in whichever order. These two steps are presented here in a programmatic fashion in order to allow further research on emotional worldviews and emotional configurations.

First the LIWC is launched to measure emotions, strictly speaking, in selected texts (speeches, parliamentary minutes, weblogs, interviews, etc.), ⁷ in order to establish the emotional configuration. The LIWC provides two kinds of results. On one hand, it measures the general emotional tone of the texts under scrutiny, with two raw scores: one reflecting the saliency of positive emotion in the text and the other one reflecting negative emotion. On the other hand, it measures the saliency of several simple emotions (the LIWC initial working scheme currently allows the analyst to get scores for three emotions, namely, anxiety, anger, and sadness) and one complex emotion (resentment, which has to be added to the LIWC initial working scheme). ⁸ Because pronouns are robust indicators of the general emotional tone of political discourses in the sense that they reveal the magnitude of intergroup tensions (Pennebaker 2011), another set of measurement is then carried out, aimed at detecting unusually high frequencies of specific pronouns, namely, “I” words (I, me, my, etc.), “we” words, and “they” words. More precisely, it is expected that “extreme groups use ‘I-words’ at extremely low rates, ‘we-words’ and ‘they-words’ at very high rates. Their members are making distinctions between their ingroups and those who are different, which is the basis of their emotional tone.” ⁹ For each emotion or pronoun, the software provides a raw quantitative score, which reflects the percentages of words and kinds of words that are particularly associated with the expression of a specific emotion, out of the total amount of words in the text under scrutiny. Scores do not therefore possess any essential meaning; indeed, these numbers have to be compared through the adequate statistical tools with other results—ideally with already established norms of various types of texts, such as blogs, emotion-laden reports, controlled writings, and the like (these norms are provided with the LIWC). Comparisons allow the analyst to conclude that political texts under scrutiny are statistically similar/different or have abnormal scores on specific emotions. This mapping of the respective intensities of specific emotions reveals emotional configurations. If multiple texts are considered within a single sample, their coherence can be assessed by examining the coefficient of variation of the sample and comparing it with that of standard LIWC texts.

Second, this initial search is expanded by measuring several elements susceptible of partaking to an emotional worldview; they both convey a clear emotional charge and are of particular interest, given the case under scrutiny. For example, an analyst who studies emotions in nationalist leaders’ discourses may look for linguistic clues evoking particular traumas and rituals specifically linked with the nationalist movement at stake. The LIWC has the ability to compute variables close to such elements, but the analyst has to complement this program by creating new LIWC variables in accordance with the specificities of his or her case. ¹⁰ Again, the resulting data have to be compared, via adequate statistical tools, with that provided by standard texts.

Setting the Stage: The Statehood Bid in Speeches and Blogs

On September 23, 2011, Mahmud Abbas handed over an application letter to Ban Ki-Moon for full UN membership for Palestine. Shortly after, Abbas made a speech supporting this bid for statehood at the UN General Assembly, followed on the tribune later that day by Benyamin Netanyahu, speaking on the same issue. The bid has then been withheld due to diplomatic tensions, only to be rehandled with more success but in a less ambitious form a year after. On October 31, 2011, Palestine was nonetheless admitted as a full member to the UNESCO during the organization’s general conference.

We examine the emotional worldview and emotional configuration that permeate through two key speeches, on one hand, and an extensive series of blogs, on the other hand. We analyze Abbas’s and Netanyahu’s speeches made at the UN General Assembly on the day the bid was presented because they are supposed to express the official views of, respectively, the Palestinian Authority and Israel on the statehood bid in particular and the state of the conflict in general. It is worth noting that leaders usually receive the assistance of speechwriters. However, given the importance of speeches at the United Nations, it is very likely that leaders contribute for the largest part to these interventions and fully endorse them. Furthermore, speechwriters are expected to share their leader’s worldview and emotions. Besides, leaders are capable of hiding their emotions, which should encourage us to interpret our results as underestimations. Our data show, however, that the LIWC is able to detect emotion in speeches, putting forward already remarkable results.

Since “collective emotions take place in the public sphere, including the mass media and public speeches” (Bar-Tal, Halperin, and de Rivera 2007, 447), and given the importance of mass media in the diffusion of social norms of intergroup perception and behavior in conflict settings (Paluck 2009), we also put a series of blogs under scrutiny, in order to unveil the emotional worldview and emotional configuration of a particular interpretive community or “online public sphere” (Hindman 2008) constituted by the numerous online political commentators of the conflict. The aim is not to treat blogs and speeches indifferently, as they play different roles, but rather to assess the similarity of worldviews and emotions in the two texts in order to unveil a hypothetical in-group “profile” shared by both leaders and the media, which are two the prominent actors in the collectivization of emotions within a group. While speeches are far from unknown for the political analyst, Internet-based texts are still neglected for various reasons. They constitute, however, not only a valid but also a highly valuable material for pragmatic and methodological reasons. On the pragmatic side, and as the recent uprisings in the Maghreb and Mashrek illustrate, the Internet in general and blogs and social networks in particular are no more a negligible part of conflicts. It was estimated that the spectacular growth of blogging brought about approximately 30 million blogs worldwide in 2008, a significant amount of which being politically oriented (Hindman 2008). Influential opinions and powerful emotions are developed through this channel outside traditional political spaces, making the political blogosphere evolve “from an interesting curiosity to an enduring feature of the political landscape” (Karpf 2008, 35) that should not be neglected. Methodologically speaking, it makes sense to work with blogs if one wants to get insight on the truly social, collective dimension of emotion—an aspect which is not necessarily attained by investigating political leaders’ speeches. Blogs are intrinsically communicative media, creating networks and discussions. As Vatrapu, Robertson, and Dissanayake (2008) showed, political blogs emanate from a desire to display and communicate emotions related to a political issue. Through this “digital circulation of affects” (Kuntsman 2010), participative Internet such as blogs and Twitter both at once foster in-group cohesion across territorial boundaries by strengthening transnational “unbounded citizenships” (Cammaerts and Van Audenhove 2005) and accentuate intergroup separation and differences (Cobb and Kuklinski 1997). Aouragh (2011) and Kuntsman (2010) have illustrated these elements in the case of the Israeli–Palestinian conflict, while Zeitzoff (2011) has used the case study of the 2008–2009 Gaza conflict to demonstrate the usefulness of quantitatively scrutinizing the social media to determine the intensity of a conflict.

To ground the analysis on a coherent basis, we selected blogs according to three criteria. First, blogs had to be exclusively dedicated to commenting the Israeli–Palestinian issue and to be either pro-Palestinian or pro-Israeli, as this is necessary for the comparative logic of our analysis, which aims to contrast different families of political texts from different groups with standard samples. By pro-Israeli and pro-Palestinian blogs, we understand blogs that are unambiguously in support either of Israel or Palestine, blogs whose posts are clear-cut expressions of support of one side and condemnation of the other side (most political blogs dedicated to the conflict actually present themselves in such way). Second, we did not include blogs that reproduce without any form of comment or presentation journal articles or contributions from other blogs. Third, blogs had to be written by native English speakers from outside Israel and the Palestinian territories. Of course, the analysis should ideally have taken into account all pro-Palestinian and pro-Israeli blogs regardless of their language or geographical location. However, the technical requirements of the LIWC (which necessitates near-native English texts) would have led us to include local blogs written in English and not local blogs written in Hebrew or Arabic, thereby introducing a selection bias. We opted for the second-best, imperfect yet most methodologically secure option of selecting only blogs from the English-speaking world, acknowledging the unbounded character of political groups, which expand across geographical boundaries. In his qualitative analysis of pro-Israeli and pro-Palestinian diasporic cyberspaces, Kuntsman (2010) indeed particularly stressed the “elastic” character of the geographical limits of these groups. Previous findings on diasporas and transnational group belonging, which emphasize the strength of membership across borders and the high intensity of in-group feelings among expatriate members (e.g., Connor 1994; Diekhoff and Jaffrelot 2004), tend to suggest that our choice should not significantly alter ideal results. Our statistical analysis is made on the basis of the posts from the selected blogs that discuss the bid for the official international recognition of Palestine as a state, both at the United Nations as such and at the UNESCO. Selected blog entries cover a period of two months before and after the bid, ranging from July 23 until the December 31 of the same year. ¹¹

First Step: Measuring Emotional Configurations

Starting with blogs, we aggregated all posts from each blog within a single text corresponding to a blog and classified these texts in either the pro-Palestinian or the pro-Israeli category. We subsequently operated all these single blog texts with the LIWC ¹² and looked at the scores for the LIWC lexicons named “posemo” (measuring positive emotions), “negemo” (measuring negative emotions), “anxiety,” “anger,” “sadness,” and “resentment.” We also computed scores for pronouns. For each outcome, we computed the mean and the standard error of the two series of blogs. To measure the emotional tone of the blogs, we provide two statistics; namely, the first allows us to compare the various samples, and the second tests the internal coherence within each sample.

In order to compare pro-Israeli and pro-Palestinian blogs, we used a one-tailed Welch’s t-test to detect significant differences between the means of the two samples of blogs. The same method was used to assess whether or not these samples are statistically different from the following three LIWC reference texts: “LIWC blogs” is a large random sample of blogs, ¹³ “LIWC control writing,” and “LIWC emotion writing” are aggregations of thousands of texts expressing, respectively, remarkably low and particularly high emotionality. ¹⁴ This first set of comparisons allows us to detect whether or not the two families of blogs show a high presence of emotion, both generally speaking through the positive/negative emotion and pronouns results and then more specifically through anger, anxiety, sadness, and resentment scores. As Table 1 shows, pro-Israeli and pro-Palestinian blogs show a strikingly similar (indeed almost identical) emotional profile. There is indeed no statistically significant difference between the two groups of blogs regarding any of the specific emotions under scrutiny (at a 5 percent threshold). The ideas are expressed in both samples of blogs with a language that reflects an analogous, high emotional profile. Pro-Palestinian and pro-Israeli blogs actually present a tamed emotional tone in comparison with blogs in general; however, it appears that blogs in general are highly emotional texts—their norms are similar to those of the “LIWC emotional text.” Under this light, we can say that the political blogs under scrutiny here are indeed clearly emotional texts. One result is of particular interest, that is, anger. While blogs on the Israeli–Palestinian conflict do not demonstrate a higher level of anger than do other blogs in general, both groups of blogs under scrutiny here score higher than emotional texts; pro-Israeli texts even surpass them to a statistically significant extent, with a rate of anger that is no less than six times more important than that of the LIWC control writing. Given the importance given to anger by scholars working on emotion in conflicts and its role in the Israeli–Palestinian conflict (e.g., Halperin and Gross 2011), this result is noteworthy. When considering blogs, we can therefore conclude that emotion—especially anger—noticeably permeates from the radical narratives. As regards pronouns, pro-Palestinian and pro-Israeli blogs present again a very similar profile; the mean number for “they” words is even the same one. More precisely, they score remarkably low on “I” words and high on “they” words, as expected by theory on emotion, language, and intergroup conflict. As previously noted, this reflects the fact that contributors share both strong in-group feelings and out-group rejection.

OPEN IN VIEWER

Table 1.

Emotional Configuration and Pronouns in Blogs.

	Positive emotions	Negative emotions	Anxiety	Anger	Sadness	Resentment	I	We	You	They
Mean of LIWC outputs (SD in parenthesis)
Pro-Israeli blogs	2.63 (0.53)	1.83 (0.31)	0.36 (0.15)	0.91 (0.28)	0.32 (0.18)	2.66 (0.63)	0.26 (0.22)	0.70 (0.28)	0.18 (0.13)	0.97 (0.28)
Pro-Palestinian blogs	2.75 (0.48)	1.97 (0.42)	0.34 (0.07)	0.78 (0.18)	0.45 (0.31)	3.07 (0.53)	0.81 (0.81)	1.19 (1.05)	0.29 (0.27)	0.97 (0.31)
LIWC blogs	3.55 (2.20)	2.34 (1.96)	0.33 (0.62)	0.98 (1.34)	0.44 (0.77)	2.24 (1.54)	6.59 (3.27)	0.62 (0.99)	1.05 (1.66)	0.51 (0.75)
LIWC control  writing	1.90 (1.40)	0.73 (0.87)	0.21 (0.41)	0.16 (0.36)	0.14 (0.33)	1.44 (0.92)	8.52 (4.40)	0.94 (1.36)	0.20 (0.72)	0.41 (0.62)
LIWC emotional  writing	3.40 (1.64)	2.69 (1.49)	0.68 (0.66)	0.69 (0.77)	0.63 (0.64)	2.93 (1.27)	10.38 (2.94)	0.73 (1.03)	0.38 (0.76)	0.71 (0.82)
Comparison using Welch’s t-test: p values
Pro-Israeli vs. Pro-  palestinian blogs	0.30	0.21	0.36	0.09*	0.13	0.05*	0.04**	0.10	0.13	0.48
Pro-Israeli vs. LIWC  reference’s blogs	0.00***	0.00***	0.26	0.17	0.01**	0.01**	0.00***	0.15	0.00***	0.00***
Pro-Palestinian vs.  LIWC reference’s  blogs	0.00***	0.03**	0.74	0.01**	0.87	0.00***	0.00***	0.14	0.00***	0.00***
Pro-Israeli vs. LIWC  control writing	0.00***	0.00***	0.00***	0.00***	0.00***	0.00***	0.00***	0.00***	0.30	0.00***
Pro-Palestinian vs.  LIWC control  writing	0.00***	0.00***	0.00***	0.00***	0.01***	0.00***	0.00***	0.25	0.17	0.00***
Pro-Israeli vs. LIWC  emotional writing	0.00***	0.00***	0.00***	0.00***	0.00***	0.06*	0.00***	0.33	0.00***	0.00***
Pro-Palestinian vs.  LIWC emotional  writing	0.00***	0.00***	0.00***	0.09*	0.06*	0.23	0.00***	0.11	0.16	0.02**

Note: LIWC = linguistic inquiry and word count.

***1 percent. **5 percent. *10 percent significance.

In order to be sure that each family of blogs is coherent as regards emotions, we also conducted an F-test for comparing the coefficients of variation of the blogs and the three LIWC reference texts. ¹⁵ The lower the coefficient of variation in a sample, the higher is the coherence of the sample. What clearly appears from this test is the remarkable internal cohesion of each family of blogs, as the emotional expression is uniform within each family (Table 2). All coefficients of variation in emotions in pro-Israeli and pro-Palestinian blogs are significantly lower than those in LIWC standard blogs. Coefficients of variation in pronouns also lower, except for “I” words (which is explained by their low frequency in our blog samples compared to the LIWC blogs). This difference is sharp; that is, for emotions, all but one coefficients of variation in our samples are lower than one-third of those in LIWC blog samples. In other words, emotional configuration within each group of blogs is highly homogeneous.

OPEN IN VIEWER

Table 2.

Internal Cohesion of Emotions in Blogs.

	Positive emotions	Negative emotions	Anxiety	Anger	Sadness	Resentment	I	We	You	They
Coefficient of variation of LIWC output (M/SD)
Pro-Israeli blogs	0.20	0.17	0.44	0.31	0.56	0.24	0.85	0.41	0.71	0.29
Pro-Palestinian blogs	0.18	0.22	0.22	0.24	0.71	0.18	1.04	0.91	0.95	0.33
LIWC blogs	0.62	0.84	1.87	1.37	1.77	0.69	0.50	1.59	1.58	1.48
LIWC control writing	0.74	1.20	1.96	2.20	2.33	0.64	0.52	1.45	3.63	1.52
LIWC emotional writing	0.48	0.55	0.96	1.12	1.01	0.43	0.28	1.41	1.97	1.16
F-test for common coefficients of variation (Forkman 2009)
Pro-Israeli vs. LIWC reference’s blogs	0.00***	0.00***	0.00***	0.00***	0.00***	0.00***	0.01**	0.00***	0.02**	0.00***
Pro-Israeli vs. LIWC control writing	0.00***	0.00***	0.00***	0.00***	0.00***	0.00***	0.02**	0.00***	0.00***	0.00***
Pro-Israeli vs. LIWC emotional writing	0.00***	0.00***	0.00***	0.00***	0.02**	0.00***	0.00***	0.00***	0.01***	0.00***
Pro-Palestinian vs. LIWC reference’s blogs	0.00***	0.00***	0.00***	0.00***	0.05*	0.00***	0.01***	0.30	0.36	0.00***
Pro-Palestinian vs. LIWC control writing	0.00***	0.00***	0.00***	0.00***	0.03**	0.00**	0.02**	0.36	0.12	0.00***
Pro-Palestinian vs. LIWC emotional writing	0.00***	0.00***	0.00***	0.00***	0.35	0.00	0.00***	0.38	0.24	0.00***

Note: LIWC = linguistic inquiry and word count.

***1 percent. **5 percent. *10 percent significance.

Let us now turn to speeches. Because in our case speeches are single texts, the adequate statistical analysis has to be different than that carried out for blogs. Rather than comparing sample means and coefficients of variation, we propose to examine whether or not Abbas’s and Netanyahu’s speeches at the General Assembly could statistically be considered similar to standard emotion-laden texts, on one hand, and emotion-controlled texts, on the other hand. Practically, for each speech and each outcome, we calculate the probability (one-tailed p value) that the speech could be considered as being part of the population of emotion-laden texts or emotion-controlled texts. Considering the overarching categories of positive and negative emotions (Table 3), it is striking to remark that both Abbas’s and Netanyahu’s speeches similarly reveal a positive emotional tone. A look at the p values teaches us that both Netanyahu’s and Abbas’s speeches cannot be considered to be controlled texts and that one cannot conversely reject their categorization as emotional texts. In sum, both speeches share a remarkably similar emotional profile, showing a positive tone. Their negative emotional tone is moderate, scoring between controlled and emotional texts. Regarding specific emotions, the picture is similar, that is, no speech unambiguously shows a significantly high level of a specific emotion, except from anger in Netanyahu’s speech. While Abbas’s speech scores rather high on this variable, Netanyahu’s speech presents a significantly higher level of anger. Regarding pronouns, both speeches present analogous patterns compared to those of blogs, that is, a very low use of “I” words and a high use of “they” words are typical of intergroup conflict narratives. This time we find a high frequency of “we” words, which reinforce this pattern. It is also interesting to note Netanyahu’s high use of “you” words, which probably reflects his attempt to connect with the audience and create a bond of emotional empathy.

OPEN IN VIEWER

Table 3.

Emotional Configuration and Pronouns in Speeches.

	Positive emotions	Negative emotions	Anxiety	Anger	Sadness	Resentment	I	We	You	They
Mean of LIWC reference texts (SD in parenthesis)
Control writing	1.90 (1.40)	0.73 (0.87)	0.21 (0.41)	0.16 (0.36)	0.14 (0.33)	1.44 (0.92)	8.52 (4.40)	0.94 (1.36)	0.20 (0.72)	0.41 (0.62)
Emotional writing	3.40 (1.64)	2.69 (1.49)	0.68 (0.66)	0.69 (0.77)	0.63 (0.64)	2.93 (1.27)	10.38 (2.94)	0.73 (1.03)	0.38 (0.76)	0.71 (0.82)
Mean and comparison of Netanyahu’s speech with reference text (p values)
Netanyahu speech	3.63	1.93	0.20	1.03	0.28	3.04	2.01	2.38	1.18	0.90
H0: speech =  control writing	0.11	0.08*	0.49	0.01***	0.34	0.04**	0.07*	0.14	0.09*	0.22
H0: speech =  emotional writing	0.44	0.30	0.23	0.33	0.29	0.47	0.00***	0.06*	0.15	0.41
Mean and comparison of Abbas’s speech with reference text (p values)
Abbas speech	3.78	1.69	0.32	0.63	0.55	2.55	0.66	2.56	0.11	1.08
H0: speech =  control writing	0.09*	0.14	0.39	0.10*	0.11	0.12	0.04**	0.12	0.45	0.14
H0: speech =  emotional writing	0.41	0.25	0.29	0.47	0.45	0.38	0.00***	0.04**	0.36	0.33

***1 percent. **5 percent. *10 percent significance.

In sum, and considering both blogs and speeches, the overall emotional tone of the texts under analysis clearly proves to be present and in a remarkable way. First and foremost, each group possesses a specific, coherent emotional configuration, defined by the intensity of its expression of anxiety, anger, sadness, and resentment. It is remarkable that political speeches and blogs, which are supposed to be very different sorts of expression, are highly similar in this respect. Second and no less clearly, both groups appear to share a similar emotional configuration. To be sure, the pro-Israeli emotional configuration is characterized by a more pronounced saliency of anger than the pro-Palestinian emotional configuration, but both are far above normal standards. ¹⁶

Second Step: Measuring Emotional Worldviews

We followed the same steps for the measurement of emotional worldview. We computed four lexicons, each reflecting societal beliefs that are susceptible to participate in a pro-Israeli or pro-Palestinian emotional worldview. Given the literature on the emotional dimension of the Israeli–Palestinian conflict, we computed the amount of religious references; preoccupations with death; suggestions that misinformation, lies, and biases about the conflict are ubiquitous in the media and the political sphere; and allusions to the Nazi regime and the genocide it perpetrated. One can indeed feel and understand (and eventually act upon) an emotional episode—here, the Statehood bid—by referring to socially shared religious truths (labeled “religious” in the tables), in direct reference to the traumatic Holocaust past (labeled “Nazi” in the tables), as a matter of existential struggle between life and death (labeled “Death” in the tables), or as yet another proof that perceptions on the ingoing conflict are massively biased in a prejudicial way (labeled “Truth” in the tables). All these elements clearly convey emotion and have been evoked in literature on the conflict. ¹⁷

It has to be noted that neither allusions to Nazi crimes nor the issue of biased perception can be measured by the LIWC in its original version, contrarily to references to death and religion; just like we measured resentment, we had to create two specific lexicons and to run them with the reference texts in order to establish proper statistical norms. For the Nazi element, we created a new LIWC lexicon with twenty-five highly representative key words of this issue, carefully avoiding including any word that could possess another meaning than its Nazi reference (the word camp, for example, does not only refer to concentration camps). We proceeded in an analogous fashion for the making of the truth lexicon. ¹⁸

Table 4 shows the comparison of LIWC scores for worldview elements in blogs. Whereas the emotional tones of the two sets of blogs were hardly distinguishable, measuring emotional worldview elements let as hypothesized appear sharp differences between them and in comparison with reference texts. The religious element is remarkably salient on both sides, but pro-Israeli blogs obtain a considerably higher score than pro-Palestinian blogs; references to religion are no less than six times more present than in the LIWC emotion text and approximately two and a half times more present than in pro-Palestinian blogs. References to the Holocaust and the NAZI regime are clearly recurrent in pro-Israeli posts, and far less frequent in pro-Palestinian blogs; the Statehood bid is regularly commented, by pro-Israeli actors, with direct references to the Holocaust. References to death are also multiple and more prominent in pro-Israeli blogs, even if pro-Palestinian blogs already score high in comparison with emotional writings, that is, pro-Israeli blogs have twice as much references to death than the LIWC emotion text (and ten times more than the LIWC control text). These observations about death and the Holocaust are consonant with previous (qualitative) studies on the presence of an “existential fear” in Israeli society (e.g., Shalit 1994) that takes its roots in collective memory. The picture is similar for references to supposedly biased opinions and information on the conflict. Whereas this preoccupation is not absent from pro-Palestinian narratives, it permeates much more heavily (almost 60 percent more) from pro-Israeli blogs, which makes it a particularly salient feature of the pro-Israeli emotional worldview.

OPEN IN VIEWER

Table 4.

Emotional Worldview Elements in Blogs.

	Religious	Death	Nazi	Truth
Mean of LIWC outputs (SD in parenthesis)
Pro-Israeli blogs	1.29 (0.61)	0.30 (0.19)	0.09 (0.08)	0.21 (0.10)
Pro-Palestinian blogs	0.52 (0.21)	0.19 (0.09)	0.03 (0.02)	0.14 (0.10)
LIWC blogs	0.32 (0.71)	0.16 (0.48)	0.01 (0.04)	0.12 (0.12)
LIWC control writing	0.18 (0.46)	0.03 (0.12)	0.01 (0.06)	0.07 (0.19)
LIWC emotional writing	0.22 (0.56)	0.18 (0.43)	0.01 (0.06)	0.09 (0.22)
Comparison using Welch’s t-test: p values
Pro-Israeli vs. Pro-palestinian blogs	0.00***	0.03**	0.01***	0.07*
Pro-Israeli vs. LIWC reference’s blogs	0.00***	0.01***	0.00***	0.00***
Pro-Palestinian vs. LIWC reference’s  blogs	0.02**	0.32	0.07*	0.62
Pro-Israeli vs. LIWC control writing	0.00***	0.00***	0.00***	0.00***
Pro-Palestinian vs. LIWC control  writing	0.00***	0.00***	0.01**	0.03**
Pro-Israeli vs. LIWC emotional writing	0.00***	0.01**	0.00***	0.00***
Pro-Palestinian vs. LIWC emotional  writing	0.00***	0.37	0.01***	0.07*

Note: LIWC = linguistic inquiry and word count.

***1 percent. **5 percent. *10 percent significance.

As for emotional configurations, the analysis of the coefficients of variation documents the remarkable coherence of each blog sample with respect to worldviews (Table 5). All coefficients in our samples are lower than those of reference texts.

OPEN IN VIEWER

Table 5.

Internal Cohesion of Emotional Worldview Elements in Blogs.

	Religious	Death	Nazi	Truth
Coefficient of variation of LIWC output (M/SD)
Pro-Israeli blogs	0.48	0.65	0.91	0.51
Pro-Palestinian blogs	0.42	0.47	0.84	0.70
LIWC blogs	2.21	2.99	3.15	0.94
LIWC control writing	2.56	4.52	10.62	2.81
LIWC emotional writing	2.54	2.38	12.40	2.50
F-test for common coefficients of variation (Forkman 2009)
Pro-Israeli vs. LIWC reference’s blogs	0.00***	0.00***	0.04**	0.01***
Pro-Israeli vs. LIWC control writing	0.00***	0.00***	0.02**	0.00***
Pro-Israeli vs. LIWC emotional writing	0.00***	0.00***	0.02**	0.00***
Pro-Palestinian vs. LIWC reference’s blogs	0.00***	0.00***	0.07*	0.39
Pro-Palestinian vs. LIWC control writing	0.00***	0.00***	0.04**	0.02**
Pro-Palestinian vs. LIWC emotional writing	0.00***	0.00***	0.04**	0.02**

Note: LIWC = linguistic inquiry and word count.

***1 percent. **5 percent. *10 percent significance.

The analysis of emotional worldview elements in Abbas’s and Netanyahu’s speeches brings about a clear pattern of results in great coherence with blogs: once again, emotional worldviews are coherently shared within both sides—but this time not really between both sides. A second observation is that on all aspects, Netanyahu’s speech scores higher than does Abbas’s. The Holocaust worldview is salient in Netanyahu’s speech but not in Abbas’s; the religious worldview appears important on both sides but with a remarkably high score in Netanyahu’s speech; the truth worldview is no less than approximately five times more salient in Netanyahu’s speech than in Abbas’s one; finally, references to death are also omnipresent in both texts but in greater proportions in Netanyahu’s account (Table 6).

OPEN IN VIEWER

Table 6.

Emotional Worldview Elements in Speeches.

	Religious	Death	Nazi	Truth
Mean of LIWC reference texts (SD in parenthesis)
Control writing	0.18 (0.46)	0.03 (0.12)	0.01 (0.06)	0.07 (0.19)
Emotional writing	0.22 (0.56)	0.18 (0.43)	0.01 (0.06)	0.09 (0.22)
Mean and comparison of Netanyahu’s speech with reference text (p values)
Netanyahu speech	1.25	0.28	0.10	0.38
H0: speech = control writing	0.01**	0.01**	0.07*	0.05*
H0: speech = emotional writing	0.03**	0.41	0.07*	0.09*
Mean and comparison of Abbas’ speech with reference text (p values)
Abbas speech	0.50	0.24	0.00	0.08
H0: speech = control writing	0.25	0.03**	0.46	0.48
H0: speech = emotional writing	0.31	0.44	0.47	0.49

***1 percent. **5 percent. *10 percent significance.

In sum, each group is characterized by a specific emotional worldview made of an ensemble of framing elements that are present with particular intensities both in blogs and in speeches. Contrarily to emotional configurations, pro-Israeli and pro-Palestinian emotional worldviews are neatly distinguishable. The pro-Israeli emotional worldview appears to be particularly unique, not only in content but more strikingly in intensity. ¹⁹