From Hate to Collective Violence: Research and Practical Implications

This paper deals with the question of how hate leads to collective physical violence: Why is it that people who are filled with hateful emotions sometimes use violent action but mostly do not? By themselves, underlying emotions such as hate are not sufficient conditions for physical violence. However, emotions are not irrelevant for the emergence of violence. My main argument is that situational interaction sequences create emotional dynamics that make collective actors overcome their inhibition threshold and act violently. Insights into the micro-timing of interaction sequences prior to violence are therefore crucial. As social movement demonstrations have recently become reconstructable in great detail, they are especially promising for analyzing the connection of micro-timings to collective violence. If we are able to identify sequences of micro-interactions and emotional dynamics leading to violence, we might be able to avoid violence by interrupting decisive sequences.


I. INTRODUCTION
Even if conflicts occur broadly over a variety of situations, cultures and countries, violent action-certainly one of the most radical ways to express hate-is rare. There are many occasions on which people feel hate toward others and are therefore motivated to use violence, but normally they do not (Collins, 2008;Klusemann, 2009). While violence is an available resource for any person at any time (Eckert & Willems, 2002, p. 1475, the choice to use violence depends on a wide range of alternative options of action. In this paper, I claim that the emotion of hate does not necessarily result in violent action. However, emotions are crucial in explaining the emergence of collective violence: I suggest that certain micro-interactional sequences produce specific emotional dynamics that can lead to collective violence. As I will show in this paper, interactional sequences are needed for translating hate into violent action, since they produce emotional dynamics that allow for actors to overcome their inhibition threshold for violence. Following Randall Collins' (2008) micro-sociology of violent confrontations, I state that these specific micro-situational sequences are required, in addition to structural and motivational factors, in order for hate to lead to violent action. A large body of studies suggests that certain situational trigger conditions can bring even peacefully motivated actors to use violent means (Collins, 2008;Gilcher-Holtey, 1995;Klusemann, 2009;Marx, 1998;Stott & Reicher, 1998). I claim that the sequence of microinteractions producing these emotional dynamics should be put at the center of attention when analyzing the emergence of violence. Several event courses in contentious politics (Tarrow & Tilly, 2007) and several event sequences and escalation courses (Eckert & Willems, 2002) recur repeatedly. We can find these sequences in connection with civil wars, police arrests, and massacres (Collins, 2008;Klusemann, 2009), and even in generally peaceful social movement demonstrations (Nassauer, 2010). Despite this central role of interaction sequences, emotional dynamics produced by interactions are key in understanding the emergence of violent action (Collins, 2008;Klusemann, 2009); they determine whether violence emerges and in which ways it does (Collins, 2008, p. 20). In addition to studying emotions in specific situations of violence, I suggest that we can complement Collins' approach by focusing on cognitive and interactional aspects.
I aim to show that sociological, psychological and political research on various topics-on violence, hate, social movements, contentious politics, and collective behavior-can be combined to study the connection of hate and violence. As situational interaction sequences can translate hate into physical violence, their avoidance can have practical implications for preventing the emergence of several forms of collective violence.
In this paper, I will first conceptualize hate and discuss the connection of emotions to rational decision-making. Subsequently, a working concept of violence will be elucidated: Violence is a physical action aiming to injure or kill another person. In the second part, empirical examples will illustrate micro-interaction sequences that are recurrent in violently ending events. These sequences and their connection to violence will be explained by a combination of three theoretical approaches : Collins' (2008: Collins' ( , 2009) microsociology of violent confrontations, Tilly's (2003) contentious politics approach, and Blumer's (1986) symbolical interactionism. Lastly, I will come back to the emotion of hate to draw conclusions on the connection between hate and physical violence. 2011] FROM HATE TO COLLECTIVE VIOLENCE 201 II. CONCEPTUALIZING HATE AND VIOLENCE

A. Hate-Concepts of Emotions and Rational Decision-making
In the following sections I will define the concept of emotions by distinguishing them from moods, affects, and feelings. I will position hate within the field of emotions and discuss the rationality of emotionally motivated actions.

Positioning Hate in the Field of Emotions
To conceptualize hate, first of all it is necessary to distinguish among several concepts commonly labeled as emotions-affects, emotions, moods, and feelings. These concepts are often used interchangeably and their definitions are still controversial. While affects are generally a shorter and more intense emotional state, a mood is a longer-term, less intense emotional state that is not related to a specific object. Emotions in the proper psychological definition are shorter and more intense than moods and are related to an object. Feelings, on the other hand, are often defined as the consciousness of emotions (Batson, Shaw & Oleson, 1992).
In sociology, we can find different distinctions of forms of emotions. Jasper (1998), for example, distinguishes among emotions that are quicker to appear, shorter to last, and reactive, like urges and reflex emotions; longer lasting forms of emotions, which he calls "affective emotions"; and moods and others emotions in between these two groups.
In this paper, I will use a more basic distinction between reflex emotions, which are more immediate and object-related, and lasting emotions, such as moods and moral emotions, which are longer lasting and less object-related. Reflex emotions include the often so-called "basic emotions" (for a discussion on basic emotions, see Ekman, 1999), such as fear and anger; lasting emotions include trust and respect, love and hate.
Hate is an emotion usually understood to be one of the strongest forms of aversion a person can have. It can be defined as the willingness to endure costs in order to harm and express a violent dislike toward others (Cameron, 2009, p. 7). It is a longer lasting and less object-related form of anger-one of the main human emotions, next to fear, sadness, surprise, disgust, and happiness (Ekman, 2003;Klusemann, 2009). The emotion of hate can produce aggression, which refers to the potential to use violent action (Imbusch, 2002).
Hate is frequently assumed to be the underlying emotion that causes violence. In fact, since the late 1970s, the U.S. legislative system has even 202 JOURNAL OF HATE STUDIES [Vol. 9:199 recognized the existence of so-called hate crimes, a subset of criminal behavior defined with regard to the perpetrator's motivation (Jenness, 2003). Hate crimes are, for example, attacks on persons motivated by the victim's race, gender, or sexuality. Apart from such lasting emotions, reflex emotions are of crucial importance for the emergence of violence in face-to-face settings (Jasper, 1998). As violence requires at least two actors, face-to-face interactions precede almost every violent situation. Lasting emotions, such as hate, might be underlying causes of violence, but as we will see below, they might be less decisive in terms of causing violence in direct interactions.

Emotions and Rational Decision-making
In order to discuss the connection between hate and violent action, it is crucial to understand the extent to which emotionally motivated actions are rational. For centuries, scholars, politicians, and the public have discussed the role of emotions in our daily choices and actions: Do humans act rationally or irrationally? Do negative emotions cause people to do irrational things that they would never have done in a rational state of mind? When reflecting on emotions, the common argument suggests that people are generally good, acting based on rational calculations. Every now and then, however, they make irrational affect-guided choices that lead to deviant behavior, such as physical violence. In this regard, all emotions, not only hate, have frequently been identified with irrationality (Goodwin & Jasper, 2006). Moreover, they have been associated with causing morally negative and deviant behavior.
This identification of emotions with irrational action can be found in a number of approaches in the social sciences, particularly in relation to studies of collective behavior. Whether Le Bon characterized collective action as irrational and affect-guided, or whether resource mobilization theory and political process theory characterized collective action as unemotional and rational-either way, emotions were assumed to be irrational (Goodwin & Jasper, 2006). This assumption has recently been disputed by social science researchers. The clear-cut line between rational cognition and irrational emotion became increasingly fuzzy, as researchers assume that cognitions and feelings are strongly connected (Melucci, 1995, p. 45).
Emotions are increasingly viewed as being "a fundamental part of rational action, not a diversion from it" (Jasper, 2006, p. 28). Therefore, emotions and cognitions can hardly be divided in the way scholars did it for decades; much less can they be contrasted. In psychological research emotions have three common functions (Schneider & Dittrich, 1990): the release of actions, the regulation of the endurance and intensity of actions, 2011] FROM HATE TO COLLECTIVE VIOLENCE 203 and social communication (for a discussion of the four psychological traditions reflecting on the emergence and function of emotions, see Allen, 2005). Apart from these functions, emotions are not only part of our reactions to situations; they also shape our goals (Jasper, 1998, p. 398). These goals and the resulting actions are not automatically irrational simply because emotions are involved in the decision-making process. Since emotions and cognition lead people to make choices, both can lead to seemingly irrational actions people might later evaluate as mistakes (Jasper, 1998). However, the choices themselves may have been made rationally. Hence, when emotions are involved in decision making, resulting action, like violence, is not necessarily irrational; nor can it be labeled as purely affectguided. As we will see below, cognitive interpretations do play a role in leading from hate to violence.

The Concept of Violence
Certainly, violent action is one of the most radical actions to express hate. No society, religion, or culture is free of violence (Imbusch, 2002). Yet, no conclusive answer as to the causes of the multiple forms of violence exists. We are dealing with one of the most elusive and at the same time most complex social phenomena in social sciences (Hagan & Heitmeyer, 2002, p. 15).
As the term violence can be imprecise, it is difficult to compare different actions that are commonly referred to as violence, for example a brawl, psychological violence, and state terror. How can we define and categorize violence if the term is applied so broadly?
First, as Imbusch (2002) emphasizes, violence is not to be confused with concepts of aggression or power. Power can be exercised by violence, but this does not have to be the case. Aggression is a psychological term, referring only to the potential to use violence (Imbusch, 2002). In this regard, Rucht (2002, p. 461) notes that the understanding of violence depends strongly on the respective culture. We can assume that the historical context is equally important.
According to Imbusch (2002), various types and dimensions of violence can be categorized. Types of violence are individual, collective, and state violence. Dimensions of violence range from metaphorical violence, to direct physical violence, to cultural-symbolical violence, to structural, institutional, or ritualized violence (Imbusch, 2002).
While some researchers highlight aspects of individual violence 204 JOURNAL OF HATE STUDIES [Vol. 9:199 (Gould, 2003), others focus on symbolical violence (Eder, 1998), institutional violence (Grimm, 2002), or violence in the civilization process (Elias, 1976;Sofsky, 1996). Depending on the concept of violence, explanations range from structural factors (Gurr, 1972;Graham & Gurr, 1969), to narratives (Rydgen, 2007;Lammont & Molnár, 2002), emotional dynamics (Collins, 2008;Klusemann, 2009), or struggles over social ranks (Gould, 2003). In this paper, I will propose a quite narrow definition of violence, which refers only to physical violence between persons; violence is a physical action aiming to injure or kill another person. I will refer only to collective physical violence in this paper, either by state or by non-state actors. Such a definition, which is close to a quotidian understanding by referring to physical and visible violence, can be useful, as it allows us to avoid blurry concepts and explanations (Rucht, 2002).
Since most societies provide different normative disaffirmations regarding different types of violence, the inhibition threshold for using physical violence against persons is usually higher than the inhibition threshold for using symbolical violence or violence against objects. Hence, I claim that the emergence of physical violence is even more unlikely and therefore particularly interesting.

The Unlikeliness of Violence
Although conflicts occur broadly in a variety of situations, cultures, and countries, violent incidents are rare, as "violence breaks with ordinary reality" (Collins, 2008, p. 130). Even though everybody could use violence at any time (Eckert & Willems, 2002, p. 1475, actors have so many alternative options of action that they usually choose another action. Thus, violence is used by only a few people and only in rare situations (Collins, 2008, p. 14). This finding is interesting for further research, as violence is not a common reaction; rather, people have to overcome an inhibition threshold first to be capable of violence. One example underlining the high inhibition threshold for violence is the study of S. L. A. Marshall on the performance of U.S. troops in World War II (Collins, 2008). The study revealed that only 15% of frontline troops actually fired their guns in combat. If soldiers did fire, most of their gunshots were ineffective. People are generally more likely to shoot the greater the physical distance from the enemy, or if a superior is standing next to them. Although his results are doubted by some researchers, Collins showed that the numbers of the study were supported by follow-up studies. This suggests that even in the military, actors try to avoid violence in most situations. On one hand, we could argue that the inhibition threshold for using violence in combat might be higher than it is in other forms of violence, as actors know that they are 2011] FROM HATE TO COLLECTIVE VIOLENCE 205 likely to kill someone. On the other hand, we could argue that the inhibition threshold might be lower, as soldiers are physically trained, mentally prepared, and, by the nature of their profession, requested to use violence, whereas civilians, for example protesters, are not. Generally, people tend to avoid violence even in situations where they threaten to use it. They also try to use violence near people who might settle the dispute, limiting the duration and thereby possible consequences right away (Collins, 2008). Therefore, violent interactions are often very short and many brawls end with the first punch.
Even if violence is not necessarily used when conflicts are present, hate has frequently been discussed in terms of favoring its emergence. In this regard, several violence studies highlight, for example, the importance of hate narratives in ethnic conflict (Eder, Giesen, Schmidtke, & Tambini, 2002;Kaufmann, 2006;Lammont & Molnár, 2002;Rydgren, 2007). By comprehensively integrating macro-as well as micro-structural influences on situational interactions, hate narrative explanations give much insight on the escalation of ethnic conflict (Eder, Giesen, Schmidtke, & Tambini, 2002). However, I claim that more immediate micro-situational trigger effects are needed to translate these hate narratives, and hate in general, into direct physical violence. Not every actor experiencing hateful feelings toward someone or something will use violent means. Violence is not a common reaction when emotions of hate are present. Actors might be motivated and armed-meaning motives as well as resources to use violence are given. Still, violence is not necessarily used (Klusemann, 2009, p. 8), if an inhibition threshold of certain emotions in a confrontational situation (see Collins' (2008) concept of confrontational tension and fear) is not overcome by additional conditions.

III. FROM HATE TO VIOLENCE
After having conceptualized hate and violence, in the following section I will elaborate upon my argument that emotional dynamics, produced by micro-interactional sequences, lead to violence. It is therefore useful to discuss empirical examples of lethal violence and my own preliminary findings on non-lethal violence. While the context and consequences of both forms of violence are very different, nevertheless, we find similar patterns leading to violence. Subsequently, these examples can be tied back to theory to come to a conclusion on the connection between hate and violence.

A. Empirical Examples -Pathways to Violence
A large number of empirical studies (Collins, 2008;Gilcher-Holtey, 206 JOURNAL OF HATE STUDIES [Vol. 9:199 1995; Klusemann, 2009;Marx, 1998) support the assumption that underlying emotions alone do not automatically lead to violence. At the same time, background variables-such as gender, class, ethnicity, or culture-cannot explain the emergence of violence (Collins, 2009). Moreover, the classical social science approaches, explaining violence by resource mobilization, political opportunity structure, or relative deprivation, cannot conclusively account for the emergence of forms of unplanned collective violence (Nassauer, 2010).
In this regard, research on protest policing-the way in which the police handle protest (Della Porta & Reiter, 1998, p. 1)-gives vital insights on the emergence and impact of different police strategies (Della Prota & Reiter, 1998;Earl, 2006;Gillham & Noakes, 2007;McCarthy & McPhail, 2005). Protest policing strategies are important for the police conduct at demonstrations and therefore also influence interactions and the possible emergence of violence. However, violence emerges at some protests where the police use a certain strategy (negotiated management, escalated force, or, recently, strategic incapacitation [see Gilham & Noakes, 2007]), but not at all of them. Nor does violence emerge only when one of these strategies is used. Thus, I assume that the protest policing strategy is not the only explanatory factor for violence in demonstrations.
At the same time, we can find certain pathways in several event courses leading to violence. I argue that these interaction sequences lead to emotional dynamics that can allow actors to overcome their inhibition threshold and use violence.

Patterns in Lethal Violence
In his study on the 1995 Srebrenica massacre in Bosnia-and-Herzegovina, Stefan Klusemann (2009) shows how the different phases of conflict escalation and the immediate interaction between General Mladic, commander of the Bosnian-Serb Army troops, and Lieutenant-Colonel Karremans, commander of the Dutch peacekeeping forces in Srebrenica, led to a specific local emotional dynamic that caused violence. In this conflict, hate narratives and ethnic motivations to kill were long present. However, confrontational tension and fear needed to be overcome and emotional dominance established before one side actually used physical violence.
With several examples, Collins (2008) illustrates how the establishment of emotional dominance by one side prior to violent action gives way to physical dominance. Several empirical examples of soldiers' actions in war situations show how, after the building up of tension and frustration on one side, a sudden move to action takes place; the enemy is caught and becomes emotionally weak. The side that has built up tension moves to a 2011] FROM HATE TO COLLECTIVE VIOLENCE 207 frenzied rush of violence (Collins, 2008, p. 88). Examples in point are the massacres perpetrated by U.S. troops in Vietnam, particularly in the villages of Ha Na and My Lai (Collins, 2008). The troops entered the villages believing them to be an enemy base. During the approach to the village, tension and fear rose. Once the troops arrived in the village, no enemy troops could be found-all soldiers had already left and only civilians, mostly women and children, remained. Confronted with the obviously inferior enemy, the troops got into an emotional rush, killing all civilians they found-in the village of My Lai 300 to 400 persons were killed-and ultimately destroying the whole village. Collins also states that the reports suggest that at this point the troops felt panic-they all described strong psychological symptoms of fear and panic. Retrospectively, they could not believe that it was they who had carried out such violence. But due to the immediate micro-situation, they felt total domination over a weak victim and entered into a violent rush. Once taken by this rush, they got even more charged up while the victim became even more passive and helpless. Both moods reinforced each other in this specific pathway to violence (Collins, 2008, p. 102). The situational dynamic in this example, which leads into a tunnel of frenzied violence, is what Collins (2008) calls forward panic: Two sides confront each other over a long period in which tension builds up. At a certain turning point, the tense confrontation changes to the overwhelming advantage of one side and the weakness of the other side. This leads to a "hot rush," "piling on" (a one-sided attack of many against few), and an "overkill" (repeated beating) (Collins, 2008, p. 89). Forward panics can be found in violent crime, police arrests, and war situations, as in the example from Vietnam.

Patterns in Non-lethal Violence
Which exact interactions bring actors to establish this sort of emotional dominance prior to violence? Which conditions are needed to do so, given that violence is such a rare phenomenon and so difficult to carry out? To analyze conditions leading from hate to violence in greater detail, I propose to look at event courses preceding collective physical violence at social movement demonstrations.
Social movements are generally the peaceful form of contentious claim-making (Tilly & Tarrow, 2007). In contrast to lethal violence, violence is not integral to this form of contentious action; thus it happens unplanned and very rarely. However, two arguments speak for comparing social movements with the above-mentioned studies and for analyzing patterns leading to violence in demonstrations. 208 JOURNAL OF HATE STUDIES [Vol. 9:199 First, it is remarkable that violence does occur even at demonstrations of generally peaceful social movements, as several additional obstacles to exist. Violence is much more irrational and more unlikely to take place in demonstrations than in other circumstances of contentious politics. From an empirical point of view, there are far fewer violent protests than peaceful ones; in France in the 1990s, for example, only 5% of all demonstrations ended violently (Fillieule & Jobard, 1998, p. 70), and in the U.S. from 1970 to 2000, only 2% of the protests show injury or property damage (McAdam, Sampson, Weffer-Elizondo, & MacIndoe, 2005, p. 9). Graham and Gurr (1969, p. 789) also comment that the organizational form of groups allows for a better control over actions at modern protests and thereby allows an exhibition of force without using physical violence. Protesters have alternatives, of which violence is usually the last and most desperate one. It is also the option of action that leads to the fewest cases of success for the group (Graham & Gurr, 1969). The fact that violence is even more unlikely to take place in this setting, and that the inhibition threshold might therefore be even higher, makes social movement demonstrations particularly interesting.
Secondly, we are capable of analyzing the micro-timing prior to violence in greater detail than in other forms of violence, for example warrelated events. Demonstrations are social phenomena where visual footage is more easily and extensively available, due to high media coverage and copious reports by nongovernmental organizations (NGOs) and the police. The recent video revolution makes it possible to focus more closely on micro-interactions and to identify specific sequences.
For my analysis of event courses, I capitalize on this new source of data. I rely on visual data, reports from governments, NGOs, the police and observers, police radio traffic protocols, and reports of participants. I reconstructed demonstration courses in western European countries and the U.S. to analyze their connection to violence. In the following section I present preliminary findings from this analysis.
When analyzing social movements, scholars found that sequences of whole protest cycles are visible (Minkoff, 1997;Tarrow, 1994). At the same time, as I will show, specific event sequences can also be found within a protest. I claim that these micro-timings are crucial in producing violence in a demonstration with generally peacefully motivated protesters and police forces.
When we look at several demonstrations (such as the Night of the Barricades, Paris, 1968;the WTO protests in Seattle, 1999;andthe G8 protests in Genoa, 2001 andRostock, 2007)  lence. An extract of an interview with an Italian police officer on duty at the 2001 G8 protests in Genoa illustrates the expectancy of violence, the spread of rumors, and the almost panic-like fear prior to the police operation: The tension among us was sky high: for the whole foregoing week we had been told that the demonstrators would have pistols, and would be throwing infected blood and ball bearings covered in acid at us. On the Friday evening after that lad's death [protester Carlo Giuliani, who was shot by an Italian police officer] they told us, that a carabiniere had died too. (Della Porta & Reiter, 2006, p. 26) Several of these demonstrations include extraordinary stress for the participants. For instance, police forces have to cope with very long shifts, insufficient sleep and food, insufficient knowledge of the area or of the course of action of the operation, and being clearly outnumbered (GdP, 2007;ARC, 2010).
Additionally, we find strikingly similar micro-interaction sequences within several demonstration courses (Nassauer, 2010). For example, police see how protesters destroy objects or build barricades and do not understand why no order is given to make arrests. They constantly hear threatening news from their colleagues over the police radio; in the example of November 30, 1999, at the WTO protests in Seattle, bad news for the police units followed in quick succession over the police radio traffic. Announcements gave information about protesters preparing for civil disobedience, about platoons being threatened, outnumbered, attacked, or announcing their withdrawal, for example "about 200 protesters shut down intersection. Taking fencing down. No police presence. Delegates in area. Unknown status," or a "large number of protesters on bus, do not have personnel to hold line" (ACR, 2010). Thereupon these units are, for example, sent into other parts of the city with little knowledge of their task or the general police strategy, and without any knowledge of the area, as police union reports on the operations in Rostock state (GdP, 2007). One of the few things they believe to know for certain is that they are in a threatening situation. Unable to communicate with other units because the police radio traffic is interrupted (Della Porta & Reiter, 2006;GdP, 2007), they are suddenly confronted with protesters of the so-called black block (see McCarthy & McPhail, 2005, p. 14) and cannot get any assessment of the general situation from police coordination (ARC, 2010;Della Porta & Reiter, 2006, p. 179). The visual footage I analyzed shows that these units are very likely to use violence when the police-protesters line breaks up-especially on easy targets and when chasing protesters.
In the several demonstration courses we likewise observe a long period 210 JOURNAL OF HATE STUDIES [Vol. 9:199 in which protesters seem to dominate the situation, and the police suddenly become active and respond with a massive use of force. As in a forward panic, the police are suddenly moved to action and when using violence especially target isolated groups, as a large body of video recordings and scientific reconstructions show (Bobbi & Busse, 2002;Della Porta & Reiter, 2006, p. 26;ARC, 2000a): They beat up individuals who have fallen down and attack bystanders, such as shop owners or the press. Just as in Collins's (2008) examples, we see that crowd violence actually happens when the crowd breaks up into smaller groups. When one individual falls down, he or she is commonly attacked by three to six opponents, becomes passive, and stops resisting (Collins, 2009). This is the point at which most violence-and the most harmful violence-takes place. This chain is one possible micro-interactional sequence that is likely to produce police violence in demonstrations. However, according to my observations, we find this situation to be strikingly similar in several demonstrations that end violently. We find similar patterns for violence by protesters.
Two aspects are particularly interesting in these examples. First, we see that interactions are unintended and chained. We see that generally peaceful actors make ad-hoc decisions in what are, for them, extraordinary situations. I showed that the course of events was often neither structurally determined nor individually or collectively intended, for example by actors motivated by hate or other hostile attitudes. In these event sequences, chained uncoordinated decisions and reactions collide. Situational decisions of the movement and situational police reactions create unplanned situations of interaction that, as I argue here, lead to violence. In several examples, the police did not act upon clearly defined rules; rather, decisions were made due to situational interactions (Della Porta & Reiter, 2006, p. 22;Gilcher-Holtey, 1995;ARC, 2000a). The same accounts for the actions of protesters. Therefore, I suggest that uncoordinated reactions lead to recurrent patterns of interaction sequences that cause violence. The patterns we find in social movement demonstrations are similar to the above-described forms of lethal collective violence. In both cases, micro-timing is crucial (Collins, 2009, p. 575).
Second, it is remarkable that we can observe many demonstrations where a large number of protesters from the black block attend, who sometimes have already collected stones to engage in violence against objects and who might look as they were motivated by hate. Simultaneously, the police use repressive tactics that seem to heat up the situation even more. Still, these demonstrations do not necessarily lead to any form of physical violence.
At the same time, we can see that those demonstrations that do end 2011] FROM HATE TO COLLECTIVE VIOLENCE 211 violently show strikingly similar sequences of interactions between the police and protesters within the event course and prior to violence. We can assume that these immediate face-to-face interactions change the actors' attitudes and emotions toward the situation and toward each other.

Toward a Theoretical Framework on Violence
I claim that we can connect three theoretical approaches to explain these events: The importance of interactions (Tarrow & Tilly, 2007) can be linked with the relevance of situations and their interpretations (Blumer, 1968) to show how situational interaction dynamics can lead to violence, by allowing actors to overcome confrontational tension and fear (Collins, 2008). I will first briefly explain the basic arguments of these theories to subsequently combine them and discuss the benefits of this theoretical approach.
Based on the above-stated insights, I join Collins (2008) in assuming that background conditions-and here I include long-lasting emotionsmight be necessary but certainly not sufficient conditions for a violent outcome. In contrast, situational factors are always necessary and sometimes even sufficient conditions for the emergence of violence. We see that micro-situational factors can keep violently motivated people from using violence and bring peacefully motivated people to use violence and even to kill (Klusemann, 2009). Hence, background conditions have to pass situational turning points in order to lead to violent action. Collins (1993, pp. 204 ff) assumes that people generally act due to bounded rationality-calculations of loss and gain that are limited by the actors' "cognitive capacity in the face of complexity." Their choices are facilitated by deciding to take actions that might lead to the highest level of positive emotions. Violence requires specific emotional conditions, which emerge in interaction: In these interactions one side becomes passive and emotionally weak. The other side becomes emotionally stronger in this specific local situation and turns to emotional dominance (Collins, 2008;Klusemann, 2009). This side overcomes confrontational emotions and uses violence.
However, this is not to be understood as a strict behavioral approach in the sense of Le Bon (2006), explaining violence by affects (here defined in the common sense understanding, as shorter and more intense emotional states) and reactions. I suggest that, drawing on Blumer (1968), we can understand the micro-sociology of violent confrontations from a symbolical interactionist perspective, meaning that the cognitive interpretation of the situation plays a crucial role. Thereby we can suppose that sometimes seemingly irrational actors carry out their actions due to bounded rationality (compare Stark, 1999), by chained interactions with other actors and their spontaneous interpretations of specific micro-situations. This means that people reflect due to bounded rationality in a circle of action, communication (also non-verbal), and interpretation. Based on these interpretations, micro-interactions can sometimes lead to unforeseeable results (Blumer, 1968), including violence.
These chained interactions are visible in the findings of my behavioral observations of visual data and of evaluation reports on police operations at demonstrations. Reports frequently mention a chain of successive actions, reactions, and interpretations within specific situations. They report not deliberate consequences of actions, as actors did not fully understand the significance of "complex, at times chaotic series of incidences" (ARC, 2000b, p. 4). Reports often suggest that due to complex interactions, a certain unintended dynamic of its own was inherent in the events (ARC, 2000a, p. 14;SPD, 1999, p. 1).
This theoretical combination of the micro-sociology of violent confrontations, the contentious politics approach, and symbolical interactionism implies several benefits. First, we can bridge the gap between the affectguided assumption of the sociology of masses and the rational approach advocated by most social movement research. When using violence, people do not rationally plan or calculate their actions in advance. However, neither do they just react in an affective way. They act due to bounded rationality and their assumption about what might bring them a higher level of positive emotions. We can therefore assume that situational interaction is crucial to lead to violence. Emotions, which emerge in these interaction sequences, are nevertheless crucial in producing violence.
Furthermore, I claim that this approach can explain that violence emerges without being caused solely by structural factors, discontent, rational calculations, or available resources. Neither is violence caused solely by motivations and lasting emotions, such as hate. Rather, it is a product of emotional dynamics, produced through specific sequences of micro-interactions.
Also, this theoretical framework can explain why violence emerges at some events where hate-filled individuals are present, but not at all of them, even if, for example in social movements, organizing groups (and therewith mostly also motivations and resources) and structural conditions remain the same. Moreover, the approach can account for the fact that the sequences of micro-interactions that take place prior to different forms of collective violence are so much alike. Lastly, it can explain why people generally try to 2011] FROM HATE TO COLLECTIVE VIOLENCE 213 avoid violence, but make use of it even in situations in which there seems to be no rational gain.

Hate and Other Emotions in the Emergence of Violence
By having conceptualized emotions, hate, and violence and having shown empirical examples and possible theoretical explanations, we saw that emotional dynamics produced by interaction sequences can lead to violence. Hence, we see that emotions do play a role in leading to violence. Coming back to our initial discussion, we can now debate: Which emotions are produced by situational interactions? Which role does hate play in comparison with other emotions in the emergence of collective violence?
Typically, we would assume that, for example, soldiers need a certain motivation, mood, or emotion to be able to kill, considering the high inhibition threshold. We would assume reflex emotions as anger, or lasting emotions as hate toward the enemy troops, to be main motivations. However, Collins' (2008) analysis of reflex emotions of soldiers, based on their faces and body postures on pictures, shows that surprisingly, only 6% of the soldiers show the emotion of anger. At the same time, one-third of the soldiers show strong or mild fear, and one-third show tension and concentration. Anger is very rare and is shown either by prisoners, or where muscular force is at work (e.g., to hinder a group of protesters from entering a site). "There is probably more anger in civilian life than in actual combat," Collins (2008, p. 70) concludes. Tension and fear are the primary emotions to be found in pictures of combat situations. Therefore, Collins (2009, p. 2) claims it to be a common cliché that anger is directly linked to violent behavior. Expressing anger is even far from actual fighting. In ethnic conflict, for example, hate narratives are often operating in the background, causing an antagonistic relationship between two ethnic groups. Yet violence occurs only if actions follow a certain pattern (see Collins, 2008, p. 115). The victims of ethnic violence are not picked due to their higher economic success, or other jealousy, but because group A perceives group B as strong and threatening. However, group A might attack because in an immediate local situation, it assumes that group B can safely be attacked (Collins, 2008, p. 121). Pictures of police arrests, as well as interviews with soldiers in combat, show the same: The dominant emotion experienced and expressed by body and face is fear (Collins, 2008;. With regard to the example of hate crimes, Collins (2009, p. 572) assumes even racists need a micro-situational advantage that allows the release of violent action. Lasting emotions and motivations favor a hostile and antagonistic relationship, but they cannot replace the micro-situational trigger conditions that establish certain emotional dynamics through face-to-face interactions. Hence, 214 JOURNAL OF HATE STUDIES [Vol. 9:199 neither the lasting emotion of hate, nor the reflex emotion of anger, is sufficient to cause violence; tension and fear are much closer connected to the immediate emergence of violence. Coming back to preliminary findings of my analysis discussed above, we see that fear is also visible in the emergence of violence at demonstrations. Evaluation reports of police departments, government committees, or the police union on police operations at demonstrations, such as the WTO protests in Seattle in 1999 (see ARC, 2000a;2000b;SPD, 1999), or the G8 protests in Rostock in 2007 (see GdP, 2007), assume that perceived decisions of the operational police command and bad equipment and support led the officers in the field to feel let down. When looking at police reports and police radio protocols during demonstrations, we see how the hesitation of the operational command to stop protesters from, for example, destroying shop windows and the unavailability of backup in situations where a platoon was "losing it" (ACR, 2010) led officers to feel abandoned and thereby increased tension (GdP, 2007). As the operational command in Seattle became overburdened and was neither able to decide where to deploy new troops for backup, nor able to communicate sufficiently with all units, platoons frequently had to make operational decisions by themselves (SPD, 1999, p. 6) in assessment of the chaotic situation they confronted. Several reports use the word "panic" to describe the psychological state officers were in prior to using violence. In preparation for the Seattle WTO protests in 1999, police documents record an atmosphere of "concern, if not panic" (ARC, 2000a, p. 9) that increased as the summit got closer. Evaluation committees conclude that we know retrospectively that the various threats police feared during the operation did not materialize. However, at the time and on the spot, police had "legitimate reason to be seriously concerned for their personal safety" and to fear that they were in "serious danger" (ARC, 2000b, p. 4). This level of panic, which was illustrated by the statement of the Italian police officer above, also becomes evident in police radio communication and exaggerated crowd estimations by the police in Seattle (ARC, 2000a, p.12).
In conclusion, we would generally assume that violence emerges in demonstrations because protesters feel anger and hate toward the state and the police, or that the police feel anger and hate toward protestors engaging in protest, civil disobedience, or rioting. Certainly these emotions toward the other group might be present in demonstrations of social movements and favor the context for the emergence of physical violence. Nevertheless, empirical evidence leads us to assume that emotions such as fear and tension are much more closely linked to the use of violence than the reflex emotion of anger or the lasting emotion of hate. 2011] FROM HATE TO COLLECTIVE VIOLENCE 215 IV. CONCLUSION In this paper I discussed the role of hate in the emergence of collective violence. Using mainly preliminary findings of my analysis of violence at social movement demonstrations and relevant findings from other authors, I showed that neither an underlying social conflict nor hate and other lasting emotions by themselves are sufficient conditions for the emergence of physical violence. I emphasized that emotions do, however, play a role in leading to violence; situational interaction sequences create emotional dynamics that allow for collective actors to overcome their inhibition threshold and act violently.
My argument was buttressed by examples of ethnic violence, war incidents, and social movement demonstrations. With preliminary findings from my analysis of demonstration courses, I pointed out that decisions made by police and protesters resulted from their limited understanding of a complex and exceptional situation, leading to a chain of uncoordinated interactions. I also showed that hate is an important underlying emotion for the emergence of violence, as it creates an antagonistic relationship. Reflex emotions, however, play a more crucial role in face-to-face interactions that precede violent actions. While we would assume anger to be a main emotion in these face-to-face settings, tension and fear were shown to be much more important for leading to violence.
In conclusion, I showed that the timing of micro-situations and specific sequences of interactions that produce certain dynamics of reflex emotions are more relevant in producing violence than lasting emotions, such as hate. Nevertheless, we can assume that the context of hate can make tension, fear, and thus violence more likely in specific situations.
In combining aspects of the works of Collins (2008Collins ( , 2009), Tilly and Tarrow (2007), and Blumer (1968), I suggested that we can explain these recurring patterns of violence. People act due to bounded rationality and choose their actions according to their spontaneous interpretations of the situation and to their calculation as to which action might bring them more positive emotions. Even if people act in an emotional situation, they do not necessarily act irrationally.
Which practical guidelines can we draw from these insights? This micro-sociological approach to the analysis of violence might provide valuable new possibilities for preventing collective violence. The proposed analysis can determine certain visible and tangible conditions, combinations of conditions, and sequences, that lead to violence in the course of a demonstration. Once we acquired a more profound understanding, certain interactional sequences could be interrupted and thereby violence might be avoided.