Dissecting the Role of Dominance in Robberies: An Analysis and Implications for Microsociology of Violence.

The influential microsociological theory of violence advanced by Randall Collins suggests that emotional dominance preconditions physical violence. Here, we examine robbery incidents as counterevidence of this proposition. Using 50 video clips of real-life commercial robberies recorded by surveillance cameras, we observed, coded, and analyzed the interpersonal behaviors of offenders and victims in microdetail. We found no support for Collins's hypothesized link between dominance and violence, but evidence against it instead. It is the absence, not the presence, of emotional offender dominance that promotes offender violence. We consider these results in the light of criminological research on robbery violence and suggest that Collins's strong situational stance would benefit from a greater appreciation of instrumental motivation and cold-headed premeditation.

Introduction

In the emerging field of microsociology of violence, evidence is accumulating on how microinteractions shape violent conflicts. Much of this research has been based on Collins’s (2008) microsociological theory of violence, which has been applied to, for example, street violence (Weenink, 2014), protest and police violence (Nassauer, 2016), and mass atrocities (Klusemann, 2012). The present study examines the relevance of Collins’s theory for explaining why some commercial robberies result in violence and others do not. Although Collins (2008, pp. 174-186) considers his theory relevant for “muggings and holdups” and “armed robberies,” and despite scholarly calls for microsociological analysis of robbery violence (Balemba, Beauregard, & Mieczkowski, 2012; Lindegaard, Bernasco, & Jacques, 2015), Collins’s theory has so far not been systematically applied to the study of robbery violence.

The objective of this article is to investigate the main hypothesis of Collins’s (2008) microsociological theory of violence, asserting that “emotional dominance” of the perpetrator is a key condition for conflict to turn into violence. While this proposition may be plausible for other types of violence, we suggest that Collins’s discussion of robberies conveys a somewhat inconsistent picture of the situational mechanisms leading to violence in robberies. Moreover, these insistencies are mirrored in the existing empirical research of robberies in which victim- and offer-based research offers conflicting accounts of how and why some robberies turn violent.

The structure of this article is as follows. Initially, we outline Collins’s, partially inconsistent, account of how dominance and violence are associated. We then present our data and methodological strategy to evaluate Collins’s theory on robberies. Next, we present our empirical results, and finally, we discuss the need of the microsociology to advance toward a multilevel theory of violence, integrating situational (e.g., emotional dominance) and subjective (e.g., instrumental motives) aspects of causation.

Sequential Relationships Between Dominance and Violence

Collins’s (2008) microsociology of violence may be characterized as an attempt to generalize Goffman’s (1967) classic situational analysis of face-to-face interaction to violent encounters. In Collins’s view, situations offer a fruitful basis for studying violence, because it is situational processes that render it possible for humans to circumvent our sociobiologically based aversion for inflicting harm to others (see Heinskou & Liebst, 2016)—what he describes as “confrontational tension/fear,” which functions as a barrier against acting violently. The key pathway to circumvent the confrontational tension/fear is through the establishment of emotional dominance. In most violent-threatening encounters, robberies included, this occurs either by attacking a weak victim, or via encouragement by an audience. Specifically, emotional dominance implies that a perpetrator becomes focused on the emotional support of the audience or on controlling the weak victim—instead of getting caught up in a symmetrical interaction with the victim that produces solidarity and thus, in turn, confrontational tension/fear.

What first appears as a simple explanatory link between dominance and violence becomes increasingly complicated, however, when Collins considers the relevance of his theory for robbery situations. Across his discussions of robberies, we find claims that resonate with his general view of the dominance–violence link, but also considerations which diverge from this framework.

Dominance precedes violence

Closely aligned with his general argument, Collins initially suggests that situations with more than one robber have an increased likelihood of severe violent outcomes. This violent effect of emotional dominance is “very much like the audience effect . . .: bigger fighting groups create their own emotional zone, pump each other up with enthusiasm, and produce more serious violence that one-to-one fights” (Collins, 2008, p. 182). This proposition is in line with victim-based studies of robberies, which report that a dominant offender tends to use violence. If the victim influences offender violence, challenging the domination of the offender is as likely to prevent violence as to stimulate it (Block, 1981; Kleck & DeLone, 1993; Tark & Kleck, 2004). These findings suggesting that victim behavior is insignificant for offender violence have, however, been contested by offender-based studies (Bernasco, Lindegaard, & Jacques, 2013; Jacobs, 2013; Katz, 1988; Wright & Decker, 1997a, 1997b), which thus offer support for Collins’s alternative account on the role of dominance in robbery violence—that dominance precedes nonviolence.

Dominance precedes nonviolence

This interpretation is offered with Collins’s observation that armed robbers often use a situational technique of attacking the weak to gain emotional dominance: “The armed robber aims to acquire dominance by means of threat, but the threat may escalate to violence and murder if there is resistance from the victim” (Collins, 2008, p. 179). As such, and inconsistently with his general claim, Collins is suggesting that successfully established emotional dominance inhibits violence. The threat of violence substitutes the need for violence, until the point where victim resistance necessitates a violent reaction from the perpetrator. Interestingly, Collins hereby accommodates his critics, for example, Felson (2009), who questions the idea that big and strong individuals are more frequent users of violence. Rather, such emotionally dominant individuals

may not find it necessary to engage in an overt attack since they are threatening enough to gain compliance and deference without it. For example, robbers with guns are less likely to engage in an overt attack than unarmed robbers. (Felson, 2009, p. 583)

This view that dominance by threat inhibits rather than stimulates perpetrator violence is also in line with accounts from offender interviews in which it is typically described that they avoid using violence by means of establishing emotional dominance (Lindegaard et al., 2015; Luckenbill, 1981). Robbers apply a variety of techniques to ensure the relative weakness of the victim, particularly in the opening phase of the robbery. Luckenbill (1981) found that robbers generate compliance by announcing the robbery with speed, stealth, and disguise techniques. These techniques were also found in other offender-based studies of robberies (Mosselman, Weenink, & Lindegaard, 2018; Wright & Decker, 1997a). For example, Bernasco et al. (2013) report that robbers prefer to enter the robbery as if they were a “SWAT team” to ensure full domination and thereby avoid using violence. Entering the scene with force, in a group, and visibly carrying weapons ensures that even the most dangerous victims become relatively weak and compliant. The use of guns in robberies increases the ability to establish emotional dominance, resulting in a lower likelihood of victim resistance and consequently a lower likelihood of offender violence (Mosselman et al., 2018).

Nondominance precedes violence

Finally, Collins (2008, p. 179) suggests, as cited above, that the attempt to establish dominance by means of threat may escalate to violence if there is victim resistance. This argument is also at odds with Collins’s general theory: By associating violence with victim resistance—that is, essentially a process of attacking the strong rather than the weak—Collins assumed that it is the robbers’ nondominance rather than their dominance that stimulates the use of violence. As such, Collins is again accommodating his critics, including Felson (2009) and Kemper (2011), who suggest that Collins’s view on dominance, and the process of attacking the weak, cannot account for the fact that “offenders are more likely to kill their adversaries during an assault when they consider them more dangerous and threatening . . . Better to ‘finish them off’ if they might retaliate later on” (Felson, 2009, p. 583). In addition to these arguments in favor of Collins’s competing suggestion that nondominance underpins violence, this proposition is also in line with empirical evidence: Offender-based studies find that when offenders realize they are unable to establish dominance because of victim resistance, they tend to resort to violence as a technique for generating compliance (Bernasco et al., 2013; Katz, 1988; Wright & Decker, 1997a).

Taken together, the above discussion suggests that empirical evidence is needed to clarify the tensions in Collins’s theory of violence and to settle the dispute between the victim- and offender-based robbery studies that support conflicting propositions in Collins’s theory. Overall, victim-based studies find that dominance stimulates violence, while offender-based research shows that dominance leads to nonviolence, and that nondominance leads to violence. Specifically, there is a need for systematic observational studies on the actual microsequences of robberies—by contrast, the offender and victim-based studies, on which Collins relies, draw conclusions from self-reported and retrospective accounts. This call for research in the actual behavioral sequence of robberies is summarized by Balemba et al. (2012, p. 607):

Only after uncovering the complex offender–victim interactions that lead to varying levels of offender violence and coercion throughout the offending sequence will researchers be able to begin to advise potential victims as to the most protective course of action.

To accommodate this view, we present the results of an investigation of robberies based on the analysis of surveillance camera recordings. These data offer—as increasingly recognized by scholars of interpersonal violence (Philpot, Liebst, Møller, Lindegaard, & Levine, 2019)—a more objective basis to consider what lessons real-life interactional sequences of robberies might teach microsociology of violence. Given the above discussion, we have reason to expect that Collins should adjust his general theory to his robbery-specific theorizations, and hence, we hypothesize that in the context of commercial robberies, the offenders’ emotional dominance (as indicated by the absence of victim resistance) is negatively associated with their use of physical violence.

Data and Method

A major limitation of most systematic empirical studies of violence is that they rely on personal accounts of the participants, either directly through personal interviews with the researchers or indirectly through transcripts of recorded interviews with police officers in criminal investigations. For example, it has been demonstrated that due to cognitive constraints, conscious misrepresentation, and poor investigative interviewing techniques, the reliability of the accounts of offenders, victims, and witnesses of crime is poor (Vrij, Hope, & Fisher, 2014). The present study aimed to overcome this limitation by direct observation of interpersonal behavior in commercial robberies, captured by surveillance cameras. Using video footage to directly observe behavior is consistent with Collins’s empirical strategy and also addresses the plea for conducting naturalistic observations in social psychology, criminology, and sociology (Lindegaard & Bernasco, 2018; Mortensen & Cialdini, 2010; Reiss, 1991). Advantages of using footage of behavior is that the researchers themselves can directly observe and classify behavior, rather than rely on interpretations of participants; that footage does not need to be observed in real time but can be played back in slow motion as many times as needed; and that interrater reliability is much easier to verify (see Lindegaard & Copes, 2017).

Our sample consists of 50 CCTV clips of real-life robberies recorded in Amsterdam and Rotterdam, the Netherlands, during the years 2011–2014, and part of these data have previously been reported in Lindegaard, Bernasco, and De Vries (2016) and Mosselman et al. (2018). Amsterdam and Rotterdam are the two largest cities in the Netherlands and together have 40% of all robberies in the Netherlands. To save data collection resources, it was decided to restrict the sample to Amsterdam and Rotterdam. While the footage in the Rotterdam area is stored centrally, no central data warehouse for robberies exists in Amsterdam. From Amsterdam, data were obtained from the pool of clips stored for the purpose of media use and search warrants. As such, these robberies may overrepresent unsolved high-impact robberies.

Only clips that conformed to specific criteria were included in the sample: the clips contain a commercial robbery; the clips include some degree of face-to-face interaction between victims and offenders; the clips contain video material rather than just still images; and the clips have a technical and visual quality that render a systematic behavioral coding possible. Regarding the latter criteria, we should note that quality of the footage (e.g., resolution and number of frames per second) was generally good, so that postures and movements were clearly visible. Facial expressions were difficult to observe, mainly because the camera angles rarely allowed observations of the participants’ faces.

A total number of 184 video clips were obtained from the police, of which 47 did not contain video material (only still images) or could not be played due to technical reasons (most cameras installed in businesses have their own video format and proprietary video player and there are many different standards). Of the remaining 137 clips, 75 did not include any offender–victim interaction. For example, they only showed the offenders entering through the front door. Four clips did not depict a commercial robbery but another event, such as an assault or a theft. Finally, of the 58 remaining clips, 8 were excluded from the analysis because an essential variable could not be determined.

The robberies in our data took place in a variety of commercial land use settings (e.g., supermarkets, bars/restaurants, jewelers) and proceeded typically less than 3 min. The majority of the robberies were committed by one or two offenders. There were typically three or fewer victims, and in about half of the robberies, several bystanders were present. Large numbers of bystanders were typical for supermarket robberies. In most cases, the offenders had weapons that were visible in the footage (e.g., a firearm, sharp, or blunt object). In a minority of the cases, a victim had access to a weapon, often an improvised weapon such as a bottle, a chair, or another object present in the shop. More details are provided in Table 1.

Table 1.

Key Features of the 50 Robberies.

Variable	N	%
Robbery duration
Up to 1 min	16	32
2–3 min	20	40
More than 3 min	14	28
Location
Supermarket	13	26
Bar/restaurant	5	10
Hotel	5	10
Jewelry store	4	8
Liquor store	4	8
Convenience store	3	6
Household item store	3	6
Drugstore	2	4
Gas station	2	4
Other (each 1)	9	18
Number of offenders
1	16	32
2 or more	34	68
Number of victims
1	29	58
2 or more	21	42
Number of bystanders
None	18	36
1	6	12
2 or more	26	52
Offender weapon
No weapon	8	16
Some weapons	42	84
Victim weapon
No weapon	43	86
Some weapons	7	14
Total	50	100

The coding was conducted in accordance with a detailed behavioral codebook (or “ethogram”) compiled from initial qualitative observations of data and existing robbery behavioral definitions applied in the literature (Lindegaard et al., 2015). For example, “offender displays firearm” is an example of an offender-specific behavior. It is described as an offender carrying a gun that is visible to victims and any bystanders, but not pointing the gun at anyone. An example of a victim-specific behavior is “victim prevents value transfer,” which is defined as any behavior by which a victim directly blocks the offender from obtaining valuable goods.

Next, the behavioral codes were assigned to the higher level categories of offender violence and victim resistance. Specifically, “offender uses (any) weapon,” “offender kicks victim,” “offender punches victim,” “offender grabs victim aggressively,” and “offender pushes victim aggressively” were assigned the offender violence label. Victim resistance included “victim actively prevents transfer of values,” “victim moves away (escape),” “victim attempts to negotiate (verbally),” “victim hides,” “victim runs after offender,” “victim threats offender with a weapon,” “victim punches offender,” “victim kicks offender,” “victim grabs or pushes offender aggressively,” and “victim uses weapon.” Reflecting our research interest in the robbery sequences, as well as a concern for avoiding simultaneity bias in our statistical analysis, these behaviors were only included as victim resistance if they occurred before any perpetrator violence had taken place. Any victim resistance starting after perpetrator violence can obviously not be a cause of that violence and was therefore ignored in the coding procedures. Furthermore, note that the behaviors in incidents with multiple offenders or victims were coded on a situational, aggregate level. For example, “offender kicks victim” was coded if any of the offenders kicked any of the victims and “victim punches offender” was coded if any of the victims punched any of the offenders.

Moreover, given that Collins describes emotional dominance in terms of being in control of the situation, it was decided to define emotional dominance as the reverse of victim resistance, so that victim resistance indicates the lack of emotional dominance, and the absence of resistance indicates emotional dominance. This definition reflects the relational nature of emotional dominance, which is not an attribute of the offender but a situational property of the offender-victim interaction. Collins’s definition of emotional dominance also includes elements of the victims’ emotions, such as feeling helpless, frozen, and suffocated, and of their behaviors, in particular their passivity and lack of resistance (Collins, 2008).

All clips were double-coded independently by two research assistants who held a Master’s and a Bachelor’s degree, respectively, in social science and who had both received an extensive customized coding training for this research. It should, however, be noted that the practice of coding the video data has taken place as an iterative rather than linear process. Although the literature offers some coding frameworks for analyzing video-recorded violent crimes (e.g., Klusemann, 2012; Nassauer, 2016), none were directly applicable to commercial robberies. Therefore, we had to incrementally evolve our methodological practice, especially with regard to what level of granularity the behaviors of interest should be coded. A fine-grained coding (e.g., separate body movements, like a person raising his or her arm) proved very hard to code reliably and was therefore discarded in favor of coding more coarse-grained behavioral patterns (e.g., “offender displays firearm”), which proved more reliable to code. Behavioral codes with this level of behavioral granularity were very reliable according the α ≥ .8 criterion suggested by Krippendorff (2004).

Given this result, we further decided that it was justified to introduce new and revised codes after the initial interrater reliability testing phase had been finalized, in so far that these codes had at least the same level of behavioral granularity as the codes that were assessed to have a high reliability. Specifically, we made two revisions after the interrater reliability tests had been conducted, namely, by introducing two distinctions between (a) “offender guides victim with a push” and “offender pushes victim aggressively” and between (b) “offender guides victim by grabbing” and “offender grabs victim aggressively.” In the original coding, we had lumped there categories together, but our subsequent qualitative assessment of data suggested that this was invalidly crude (e.g., guiding and aggressive pushing serves different goal-directed purposes). Because the reliability of all other behavioral measures was good, and because we had no reason to assume that the results would be different for these four new measures, we did not perform double coding and reliability calculations of the new measures. Essentially, here we decided to weigh the value of the validity of the coding over an exact assessment of its reliability. Finally, to secure a consistent application of these revised codes to data, all cases were recorded by an additional student assistant trained for the purpose.

Results

Table 2 presents our data in a 2 × 2 contingency table displaying the frequency distribution of victim resistance and offender violence. From Table 2, it can be established that most robberies—approximately four out of five cases—unfold without the robbers using physical violence. This limited prevalence of violence across our data suggests that robbers typically accomplish their deed by means of threat of physical violence instead of actual physical violence. This finding provides initial support for our hypothesis, assuming that violence typically develops because of threats having failed to establish emotional dominance in the face of victim resistance.

Table 2.

Frequency Distribution of the Level of Victim Resistance and Offender Violence.

	Dominance (No Victim Resistance)	No Dominance (Victim Resistance a )	Total
Offender violence a	3	6	9
No offender violence	36	5	41
Total	39	11	50

Note. Pearson’s χ2(1) = 12.76 (p < .001). Fisher’s exact test p = .002. Odds ratio = 15.2.

See Appendix for detailed behavior categories.

Next, we examine the extent to which these relatively uncommon cases of violence are dependent on varying levels of emotional dominance. Both Pearson’s chi-square test and Fisher’s exact test indicate a statistically significant association between these two measures. Combined with the relatively large magnitude of the odds ratio of 15.2, this result provides evidence for our hypothesis that failed attempts to establish emotional dominance are associated with violent outcomes. Keeping in mind that we have measured nondominance as active victim resistance, our data suggest that the use of violence in robberies typically unfolds as a sequential process in which the robber’s threats are most likely to escalate to violence if there is resistance from the victim—that is, when the victim claims the role of the strong rather than the weak.

Discussion

Collins’s microsociological theory of violence proposes that in situations of conflict, emotional dominance of a perpetrator over a victim is a key condition for violence. Based on a sample of 50 CCTV recordings of real-life robberies, we evaluated the role of emotional dominance in violent-threatening robbery incidents. The findings do not support Collins’s proposition, suggested in his general theory of violence. To the contrary, we find that violence is less common in robberies where perpetrators achieve emotional dominance than in situations in which they do not. Our finding is in line with offender-based research on robbery (Bernasco et al., 2013; Jacobs, 2013; Katz, 1988; Wright & Decker, 1997a, 1997b) and with issues raised by scholarly critics of Collins’s general thesis on the dominance–violence link. Taken together, the cumulated evidence and theoretical issues suggest that the explanatory span of Collins’s theory should be broadened, so as to be able to distinguish between situations where emotional dominance increases the likelihood of violence and situations where it functions as a shield against violence.

One way forward could be to nullify the inconsistencies facing Collins by reassessing emotional dominance as an explanatory condition that is indeterminate with regard to what action tendencies are triggered (Elster, 2011). Emotional dominance may thus be similar to the other emotional states, such as the feeling of fear from which it is effectively impossible to predict the action outcome of fight, flight, freeze, or faint. Elster (2011) highlights that such indeterminacy of emotional mechanisms should be taken as an analytical starting point for studying emotions. With the progress of knowledge, however, it might be possible to identify the causal pathways linking emotions and actions. As such, the question is whether we have the knowledge to determine under what conditions dominance stimulates violence and under what conditions it inhibits.

In considering this, we may return to Collins’s scholarly critics, several of whom are suggesting that his situational perspective needs to be balanced against some notion of the subjective motivation in violence (e.g., Felson, 2009; Smith, 2015; Wieviorka, 2014). Here, predatory crimes like robberies are a case in point, given that such deviance would simply not occur if it were not for the perpetrator’s purposeful action to attain some goal by illicit means (Clarke & Cornish, 1985; Felson & Tedeschi, 1993, though see Feeney, 1986). This emphasis on instrumentality of predatory crimes provides a framework that allows us to distinguish between situations where emotional dominance triggers violence and where it counteracts this outcome (Felson, 2006; Lindegaard et al., 2015).

Regarding the current findings, we suggest that the absence of violence in the majority of robbery cases may be attributed a subject–situation interplay by which the perpetrator uses the situational property of emotional dominance instrumentally to accomplish the robbery nonviolently. As such, if emotional dominance is a sufficient means to attain the robber’s goal, it may be considered rational to choose this strategy—given that direct use of violence is potentially more costly in terms of formal and informal penalties, as well as experienced confrontational tension/fear (Jacobs, 2013; Lindegaard, Bernasco, Jacques, & Zevenbergen, 2014). Furthermore, when emotional dominance fails because of victim resistance, the robbers use violence as an instrumental means to overcome this situational constraint preventing the acquisition of the desired goal.

Collins (2008, p. 21) is reluctant to adapt such instrumentalist arguments into his framework, given that his “preferred strategy is to push as far as possible with a situational approach”—only when this approach has been emptied out, scholars should in his view engage in the development of a “multilevel theory” of violence that combines situational–person interactions. However, we believe that the case of robberies offers counterevidence to Collins’s main theory that precipitates the advancement of such multilevel theoretical framework. This argument, it should be added, does not exclude the possibility that Collins’s general understanding of the dominance–violence link is plausible in other types of violent crime, less underpinned by predatory motivations. As such, the most compelling evidence in favor of Collins’s theory tends to be found in the context of hot-headed encounters, such as street violence (Weenink, 2014). Here, the interpersonal aggression is typically linked to Collins’s (2008, pp. 85, 132) term “forward panic,” describing conflict interactions in which a stepwise buildup of tension/fear is released in an “emotional rush” in the moment the victim shows weakness. Such emotional–motivational climate, however, is clearly different from the robbery situations in which the emotional tone is set by the robber’s cold-headed instrumentality (Contreras, 2013; Katz, 1991).

Adding to these considerations, we would like to add that our goal-directed account of why emotional dominance inhibits violence should not be restricted to considering the “proximate” causes (i.e., cognitive reasoning) that leads robbers to abstain from using violence. This argument may be extended to consider the “ultimate” cause (i.e., evolutionary function) underlying this nonviolence pattern (Mayr, 1961; Smith & Parker, 1976). We believe that Collins’s concept of emotional dominance is at odds with an evolutionary theory of emotional expressions. Here, it is suggested that humans have a biologically innate capacity to express and recognize shame (i.e., submission) and pride (i.e., dominance) in bodily postures (Darwin, 1872; Tracy & Matsumoto, 2008). The adaptive value of these dominance–submission displays is most likely that they inhibit violence: “reliably recognizing shame displays in others can be vitally beneficial for conflict avoidance. Just as appeasing reduces conflict for those who display shame, it also reduces conflict for those who observe it” (Martens, Tracy, & Shariff, 2012, p. 401). If the innate shame/submission expression increases, rather than decrease, the likelihood of physical aggression, evolution would simply select against this trait and eventually wipe it out from the human gene pool. The innate nature of dominance/submission emotions indicates, in other words, that they serve an evolutionary–adaptive function in inhibiting violence.

This study has limitations that warrant consideration. One issue relates to how violence is defined in our analysis. The validity of our Collins’s evaluation hinges ultimately on the agreement between our operationalization and his concept of violence. The problem is, however, that Collins does not offer an unequivocal definition of violence—as he puts it, it

is not useful to insist that violence must fit an exact preconceived definition . . . Where do we draw the boundaries? Are threats a form of violence? Clearly they are close enough to it that we have to put them into the model of situational dynamics. (Collins, 2008, p. 24)

Although perhaps sensible, this argument effectively immunizes Collins’s theory against being tested directly. In so far that the conceptual boundary between “violence” and “nonviolence” remains blurred, our empirical result (i.e., dominance prompts nonviolence) can be transposed to the opposite result, simply by changing the ad hoc definition of violence to include more nonphysically aggressive behaviors (e.g., threat). Our definition of physical violence did not include threat of physical violence alone, without actual violence being used. This definition allows us to empirically demonstrate not only that the offenders use physical violence in only 18% (9 out of 50) of the commercial robberies but also that this is largely a consequence of victims refraining from resistance.

Another concern relates to the operationalization of the other measure included in our analysis: emotional dominance. By definition, emotional dominance describes an intersubjective emotional process, yet we operationalize emotional dominance from a single subject’s behavior—that is, victim resistance. As such, we assume that it is plausible to make inferences from “behaviors” to “emotions” and from “subjective” to “intersubjective” states—yet, we lack the direct evidence for evaluating whether we have constructed a valid proxy for emotional dominance. This limitation taps into the broader issue that study of human emotions suffers from a deficit of methods, especially for measuring emotions as they unfold in interactional processes (Bernasco et al., 2013; Clay-Warner & Robinson, 2015; de Gelder, 2009). Given that one of the unique features of microsociology—compared to, say, criminology (Van Gelder, Elffers, Nagin, & Reynald, 2013)—is the sensitivity to the emotional processes of violence, future research should prioritize the development of validated measures of aggressive emotion processes, including emotional dominance.

Furthermore, the selectivity of data sources is a major limitation, in particular with respect to video data accessed via police sources, without control over the selection process (Lindegaard & Bernasco, 2018). Our empirical conclusions are based on a sample that is necessarily selective because it only includes robberies in businesses with a working surveillance camera and only robberies in Amsterdam and in Rotterdam. Due to how footage is stored by the police, the Amsterdam cases probably also involve more high-impact robberies with investigative priority than those in Rotterdam. Although a random sample would have been preferable, non-probability convenience samples may be less of a concern if, as in the current case, the research purpose is theory testing rather than descriptive inferences to a wider study population (Calder, Phillips, & Tybout, 1981; Collins, 1983). Although selectivity is to some extent unavoidable because crime recordings are usually not made for research purposes, we suggest that future researchers should attempt to gain some control over what is recorded, where it is recorded, how it is recorded, and which parts of it are saved and selected for the analysis.

In concluding this article, we highlight two suggestions for future research on the relation between emotional dominance, resistance, and physical violence. First, following up on the conclusion that the microsociological perspective should consider instrumental motivation and cold-headed premeditation in addition to situational properties and interactions, it would be useful to triangulate our conclusions on commercial robberies with findings on other types of events that typically involve physical violence or threats of physical violence, such as street robberies, sexual offenses, or domestic disputes. Second, we suggest that the field of violence research could benefit not only from methodological knowledge in animal ethology (Jones et al., 2016) but also from substantive findings on dominance, submission, and physical aggression in the animal kingdom. Although we acknowledge that animal behavior does not automatically generalize to humans, our conclusions about violence in robberies appear to correspond with the finding that submissive behavior inhibits aggression in many group-living species (Lindegaard & Copes, 2017; Preuschoft & Schaik, 2000).

Observed Frequencies of Behavior Categories Coded as Victim Resistance and Offender Violence (Source of Table 2).

Behavior Categories Coded “Victim Resistance” a	Frequency b
Victim moves away (escape)	1
Victim hides	1
Victim actively prevents transfer of values	8
Victim runs after offender	3
Victim threatens with weapon	2
Victim uses weapon	3
Victim punches/kicks offender	0
Victim grabs offender aggressively	0
Victim pushes offender aggressively	1
Victim uses other types of violence against offender	3
Total number of behaviors coded “victim resistance”	22

Behavior categories were only coded “victim resistance” when they occurred before the first instance of offender violence.

Note that the frequencies sum to more than the 11 robberies reported in Table 2 because in various robberies multiple victim resistance behaviors were observed.

Behavior Categories Coded “Offender Violence”	Frequency a
Offender uses firearm	0
Offender uses sharp object	0
Offender uses other weapon	2
Offender punches/kicks victim	1
Offender grabs victim aggressively	6
Offender pushes victim aggressively	1
Offender tears/hauls victim	1
Victim uses other types of violence against offender	2
Total number of behaviors coded “offender violence”	13

Note that the frequencies sum to more than the nine robberies reported in Table 2 because in various robberies multiple offender violence behaviors were observed.