Abstract
When competing for scarce resources, groups can behave aggressively toward one another. Realistic conflict theory suggests that intergroup hostility internally ties groups together, thus improving intragroup functioning. In contrast, conflict spillover theory suggests that aggressive behaviors between groups can permeate to the intragroup level and thus worsen intragroup functioning. We reconcile these two opposite perspectives by introducing the relative group size as a moderator that determines when and how targeted verbal aggression from one group harms or improves intragroup functioning in the targeted group. We tested our hypotheses using a sample of in-situ observations of transcribed plenary discussions in the German national parliament and compared intergroup targeted verbal aggression by distinguishing targeted verbal aggression from two social groups (i.e., a new populist smaller party vs. a larger group of veteran parliament members). We measured targeted verbal aggression as a form of hostile intergroup behavior from each social group using computerized text analyses. We analyzed intragroup functioning using a measure of verbal mimicry. Our results show support for our hypotheses. We discuss theoretical and practical implications for the verbal aggression and intergroup relations literature.
Keywords
“Yes, (Stefan Keuter, Alternative for Germany, AfD) “Maybe then there won’t be so much (Stephan Protschka, Alternative for Germany, AfD)
Verbal aggression—defined as verbal communications of anger that violate social norms (Grandey, Dickter, & Sin, 2004)—is particularly prevalent in political workplaces (Kalmoe, 2014; McLaughlin, 2020; Seiter & Gass, 2010). These workplaces are typically characterized by intergroup tensions and hostility between political groups that compete for power: for example, former US president Trump’s calling his opponent “crooked Hillary” or the German populist party leader Alexander Gauland from the Alternative für Deutschland (AfD), who suggested “dumping” one his political rival “in Anatolia”. In politics, the goal of verbal attacks is to weaken members of an opposing group. But is verbal aggression actually “effective” in harming others? A better understanding of targeted verbal aggression is important for both theoretical and practical reasons (Glomb, 2002). First, we know from individual-level research and meta-analyses that verbal aggression at work has typically adverse personal consequences, including reduced task performance, impaired working memory (Rafaeli et al., 2012; Walker, Van Jaarsveld, & Skarlicki, 2017), negative impact on psychological and physical health as well as work-related withdrawal behaviors (Han, Harold, Oh, Kim, & Agolli, 2022). Second, verbal aggression is considered to be ubiquitous, that is, one of the most frequently experienced forms of workplace aggression (Glomb, 2002; Grandey et al., 2004; Neuman & Baron, 1998), which is why it is practically relevant to understand the various consequences of this behavior.
However, one problem is that verbal aggression has been predominantly studied at the individual or dyadic level, although aggressive acts rarely take place in a social vacuum. Thus, the broader impact of verbal aggression in group contexts is still severely underdeveloped (Miner et al., 2018; Van Knippenberg, 2003). Not only do social contexts influence whether aggression is exhibited, but they are also likely to impact the effects that aggression has on its targets (Hershcovis & Reich, 2013). This is a surprising blind spot since most scholars consider context to affect organizational behavior in general (Johns, 2006) and workplace aggression in particular (Hershcovis & Reich, 2013; Hershcovis, Reich, Parker, & Bozeman, 2012). For example, when verbal aggression is exhibited between different social groups, difference in group sizes could play a key role in whether verbal attacks from one group harm intragroup functioning within the targeted group. In other words, social context factors may determine the recognition, interpretation, and actual impact of aggressive behaviors (Hershcovis et al., 2012; Hershcovis, Cortina, & Robinson, 2020; Hershcovis & Reich, 2013).
Our research addresses these current shortcomings. Our key research question is: To what extent do verbally aggressive intergroup behaviors harm or benefit intragroup functioning? Two contrasting perspectives informed answers to this question. On the one hand, conflict spillover theory (Jehn et al., 2013) suggests that intergroup verbal aggression can infiltrate the intragroup level, implying that targeted verbal aggression should impair intragroup functioning. On the other hand, realistic conflict theory (Jackson, 1993) suggests that hostility between groups strengthens intragroup functioning, implying the opposite effect. We reconcile these two opposing lines of thought by taking stock of recent theorizing that considers intergroup behaviors crucially tied to relative size differences between the involved groups (Carton & Cummings, 2012). Specifically, we propose that relative group size is a key moderator that determines whether targeted verbal aggression (i.e., aggressive language that specifies a target by using second-person pronouns like “you”/“yours”) harms or improves intragroup functioning. We argue that targeted verbal aggression from small groups is less expected and will likely trigger more pronounced negative consequences (Walker, Van Jaarsveld, & Skarlicki, 2014), thus potentially harming the intragroup functioning of the targeted group.
Overall, our research offers three contributions to the literature. First, we expand knowledge about targeted verbal aggression in an intergroup context. That is, we go beyond extant research considering the detrimental consequences of individual or dyadic verbal aggression at work by offering a perspective from the intergroup relations literature. In doing so, we address the call for research on the broader impact of verbal aggression on group processes (Miner et al., 2018; Van Knippenberg, 2003) and add nuance to the aggression literature by differentiating between targeted versus nontargeted verbal aggression (the latter being aggressive language without a specific target). Second, we explore how group size differences modulate the effect of targeted verbal aggression on intragroup functioning. The consideration of this moderator speaks to Hershcovis and Reich (2013), who argued that we need to better understand how the social context can shape the negative effects of aggressive behavior to delineate meaningful (practical) implications. Third, by using a novel measurement approach, we expand existing measures of verbal aggression in real-world contexts. Specifically, we used a text-analytical approach to measure targeted verbal aggression in an intergroup context and its impact on intragroup functioning, which we captured via verbal mimicry in each group. Verbal mimicry is a language-based measure, which we argue to be a key aspect of intragroup functioning (Chartrand & Lakin, 2013; Lakin, Chartrand, & Arkin, 2008; Maddux, Mullen, & Galinsky, 2008; Van Baaren, Holland, Steenaert, & van Knippenberg, 2003; Van Baaren, Holland, Kawakami, & Van Knippenberg, 2004), and we provide further evidence for this. Our approach offers avenues to overcoming some limitations of many existing workplace aggression measures (Hershcovis, 2011; Hershcovis & Reich, 2013). In what follows, we delineate the theoretical background for our hypotheses (see Figure 1). Research model.
Theoretical Background
In many organizational settings, groups interact or even compete with other groups, which gives rise to the study of intergroup behaviors (Kramer, 1991; Van Knippenberg, 2003). Because groups compete for scarce resources (e.g., power, money, personnel, and help from management), intergroup behaviors are predisposed to hostility and aggression (Baldridge, 1971; Kramer, 1991). In other words, substantive tensions and hostile, aggressive behaviors can arise between different groups (Baldridge, 1971; Kramer, 1991). While intergroup behaviors occur between different groups, intragroup functioning deals with the processes within a group (Van Knippenberg, 2003). Research has indicated that conflicts and hostile behaviors between groups cannot only affect the relationships between the involved groups (Jackson, 1993; Park & DeShon, 2018; Van Knippenberg, 2003) but also their respective intragroup functioning, such as group coordination and relationships within the group (see Sherif et al., 1961; Sherif & Sherif, 1966; Van Bunderen, Greer, & Van Knippenberg, 2018).
Intergroup Behaviors: Targeted Verbal Aggression Between Groups
To better understand how hostility in intergroup behaviors shapes intragroup functioning, we focus on verbal aggression, defined as verbal communications of anger that violate social norms (including insults and swearing) (Glomb, 2002; Grandey et al., 2004). This definition is broad enough that it includes persistent and more severe forms of mistreatment (e.g., bullying; Einarsen & Skogstad, 1996) and lower intensity and less persistent forms of mistreatment (e.g., incivility; Andersson & Pearson, 1999). While the broader concept of workplace aggression also covers more extreme behaviors like physical assaults and violence (Neuman & Baron, 1998), verbal aggression excludes these behaviors and is predominantly focused on language or symbolic expressions of hostility against others (Neuman & Baron, 1998). This is also why verbal aggression is considered to be one of the most frequently experienced forms of aggression in the workplace (Glomb, 2002; Grandey et al., 2004; Neuman & Baron, 1998).
Overview of Aggression Concepts in the Academic Literature.
Note.* = Overlap with item from another construct, -- = Item unique to this construct within the five concepts listed.
As verbal aggression describes the form that aggression can take, rather than constituting a distinct type of workplace aggression, it is reflected in almost all types of workplace aggression constructs that have been differentiated to date. Constructs such as bullying, incivility, and abusive supervision all include verbally aggressive behaviors (see Table 1). Accordingly, rather than adding to the plethora of constructs in the workplace aggression literature, verbal aggression is a label for the way aggression can be expressed—irrespective of whether it can be classified as bullying, incivility, or any other type of workplace aggression. To move forward, we thus only distinguish between targeted and nontargeted verbal aggression as two forms of aggressive intergroup behavior. For example, a member from one group can be verbally aggressive without specifically targeting another group (e.g., “The East was good enough for the garbage. This is a mess!”). Alternatively, group members can also use verbal aggression, including words that directly target members from the other group (e.g., “Tell them that you give a shit about them, because with this regulation you are destroying the small piglet breeders and the small farmers.”). This distinction is important because targeted verbal aggression is particularly relevant in an intergroup context in which members from one group can direct their hostility toward a target as a form of expressing their animosity toward the other group.
Intragroup Functioning: Verbal Mimicry Within Groups
To understand how targeted verbal aggression affects intragroup functioning, we focus on verbal mimicry (operationalized as language style matching, LSM 1 ) as a key indicator of intragroup functioning. Verbal mimicry has been argued to constitute the “social glue” between members of a group (Lakin, Jefferis, Cheng, & Chartrand, 2003; Shi, Zhang, & Hoskisson, 2019), that is, a form of convergence that facilitates relational functioning and affiliative bonds (e.g., Rains, 2016; Van Swol & Kane, 2019). Conceptually, verbal mimicry taps into convergence between group members with respect to their language styles (i.e., the extent to which individuals use pronouns and function words when they speak; Pennebaker, 2011). When a group exhibits low levels of verbal mimicry, members use widely different language styles; that is, there is a stronger divergence between group members—indicating that they do not speak the “same language” and thus cannot understand each other properly (Gonzales, Hancock, & Pennebaker, 2010; Ireland, 2011).
Research has established that behavioral mimicry between social interaction partners is a strong measure of relational functioning, indicated by its positive associations with rapport (Bernieri, Reznick, & Rosenthal, 1988), liking (Lakin et al., 2008; Maddux et al., 2008), feelings of empathy (Stel & Vonk, 2010), helping behaviors (Van Baaren et al., 2003; 2004), and pro-social behaviors (e.g., Stel & Vonk, 2010; van Baaren et al., 2003). Furthermore, the perspective that verbal mimicry is indicative of positive relational outcomes is supported by field research in various contexts: For example, Meinecke and Kauffeld (2019) showed that verbal mimicry in dyads (leaders and their followers) was associated with the followers’ perception of the leaders’ likeability. On the intragroup level, research found that verbal mimicry of language styles is significantly associated with self-report measures of team cohesion (Castro-Hernandez, Swigger, & Ponce-Flores, 2014; Gonzales et al., 2010; Manson, Bryant, Gervais, & Kline, 2013; Muir, Joinson, Collins, Cotterill, & Dewdney, 2020; Scissors, Gill, & Gergle, 2008; Tausczik & Pennebaker, 2013). In political and negotiation contexts, that is, on the intergroup level, verbal mimicry of language styles during international diplomatic negotiations has been associated with stronger agreement (Bayram & Ta, 2019) and with improved rapport during online negotiations (Muir et al., 2020).
Verbal Aggression Between Groups and Intragroup Functioning
When does targeted verbal aggression between groups negatively affect intragroup functioning (i.e., verbal mimicry)? In our theoretical reasoning, we rely on two opposing perspectives, one being realistic group conflict theory (Jackson, 1993) and the other being a conflict spillover perspective (Jehn, Rispens, Jonsen, & Greer, 2013).
Realistic group conflict theory assumes that intergroup tensions arise when groups have incompatible goals or compete for the same resources. More crucially, this theory suggests that intergroup aggression is not contained at the intergroup level (Sherif & Sherif, 1966; Sherif, Harvey, White, Hood, & Sherif, 1961; Sherif, 1966) but may also permeate group boundaries, thus influencing intragroup functioning (Mead & Maner, 2012; Van Bunderen et al., 2018).
Based on realistic conflict theory, we would expect that targeted verbal aggression between groups (i.e., at the intergroup level) impacts intragroup functioning in such a way that the groups start sticking more closely together and pool their resources to defend themselves against external threats (e.g., Brewer, 2001; Sherif, 1966; Simmel, 1955; Stein, 1976; Tajfel, 1982). For instance, conflicts at the intergroup level have been shown to improve intragroup cooperation (Benard, 2012; Bornstein, 2003; Bornstein, Budescu, & Zamir, 1997; Dion, 1979; Sherif et al., 1961) and contribute to team resource pooling (Erev, Bornstein, & Galili, 1993; Gunnthorsdottir & Rapoport, 2006; Halevy, Weisel, & Bornstein, 2012). From this perspective, verbal aggression between groups should improve intragroup functioning; thus, we would expect targeted verbal aggression to be positively related to intragroup verbal mimicry.
Alternatively, particularly when unexpected, verbal aggression between groups could also lead to negative cross-level “spillover” effects. As conflict spillover perspectives would suggest, conflicts and tensions between groups potentially propagate conflicts at the intragroup level (see Jehn et al., 2013; Keenan & Carnevale, 1989; Van Bunderen et al., 2018). Research supporting this perspective has shown that conflicts can be transmitted from the between-group level to the intragroup level in the form of intragroup power struggles (Van Bunderen et al., 2018). Therefore, rather than groups acting in unity, verbal aggression between groups might harm intragroup functioning by spurring fights over internal resource control. In sum, different theoretical perspectives exist as to how targeted verbal aggression could impact intragroup functioning.
A key to resolving these different perspectives could lie in the context surrounding these groups. In particular, prior research on workgroups suggests that difference in group size constitutes an important contextual variable (Carton & Cummings, 2012). With regard to the effects of group size on intragroup functioning, research has shown that numerically smaller groups are more prone to favoring members of their own group, suggesting better intragroup functioning in comparison to larger groups (Hewstone, Rubin, & Willis, 2002; Mullen, 1991; Simon, Aufderheide, & Kampmeier, 2001). Conversely, in larger groups, positively valued intragroup functioning (e.g., cooperation) typically decreases (e.g., Brewer & Kramer, 1986; Hamburger, Guyer, & Fox, 1975; for a review, see Pruitt, 1998; Yamagishi, 1992). The reason for this effect is that members of larger groups feel less efficacious and may have the perception that their contributions are dispensable (Kerr & Bruun, 1983), as well as feeling less responsibility to pursue the group’s welfare (Kerr, 1989; Liebrand, 1984).
In sum, research suggests that members of larger groups may be less prone to act toward collective interests. Accordingly, we expect that targeted verbal aggression between groups will most likely damage intragroup functioning within the larger of two hostile social groups. Thus, we hypothesize:
Targeted verbal aggression of a smaller group is negatively related to the larger group’s intragroup functioning (i.e., verbal mimicry).
The Role of Relative Group Size in Intergroup Encounters
Group size appears to be an important determinant of how intergroup aggression affects intragroup functioning. When groups compete for resources, differences in group sizes are an important context cue that helps competing groups to build expectations on whether to anticipate verbal aggression from their rivals (Lucchesi et al., 2020; McComb, Packer, & Pusey, 1994; Roth & Cords, 2016; Wilson, Kahlenberg, Wells, & Wrangham, 2012). For example, research in the field of behavioral ecology has for a long time explored the role of relative group size in intergroup encounters and provided consistent evidence that a numerical assessment of the opponent group is widespread among different species (Wilson et al., 2012). This type of research suggests that groups (from a variety of animal species) assess their opponent’s strength in numbers when facing a competitor group that has a relatively smaller number (McComb et al. 1994). Overall, this research highlights that asymmetries in group sizes between opposing groups play a critical role in the intensity of intergroup competition (Roth & Cords, 2016). Based on this rationale, we argue that the effect of targeted verbal aggression on group functioning is dependent on the relative group size of the aggressor group. In particular, with increasing group size of an aggressor group, the effects of targeted verbal aggression are reduced as the opponent group is more likely to make a numerical assessment and prepare for an attack. In contrast, verbal aggressions may not be expected from an opponent group that is relatively small (compared to the other group). When a targeted group is not prepared for aggressive behaviors, these behaviors are more surprising and can cause more harm. In line with this argument, research on uncivil encounters between employees and customers has shown that rude customer behavior can trigger more pronounced reactions when these behaviors were unexpected (Walker et al., 2014). Thus, when targeted verbal aggression comes from a relatively small group, it comes with an element of surprise for the larger group. Accordingly, the relative group size of an aggressor could be an important context cue that likely impacts the expectation of aggressive behavior and thus has a harmful impact.
When aggressor group size grows larger, the opponent group is more likely to expect aggression from their competitors. In line with realistic conflict theory, where groups (typically of equal size) expect hostile intergroup behaviors, targeted verbal aggression between groups is more likely to be associated with improved intragroup functioning (i.e., members sticking together, binding their resources, and acting more cohesively to defend the impending threat posed by the other group; Brewer, 1999; Sherif, 1966; Simmel, 1955; Stein, 1976; Tajfel, 1982).
To summarize, we expect that relative group size moderates the effect of targeted verbal aggression on intragroup functioning (i.e., verbal mimicry). As relative group size increases, targeted verbal aggression will elicit less negative effects on the targets’ intragroup functioning. That is, with increasing relative group size, targeted verbal aggression becomes less effective and may even have the effect of bringing members of the targeted group closer together.
Relative group size moderates the effect of the smaller group’s targeted verbal aggression on the targets’ intragroup functioning. With increases in relative group size, a smaller group’s targeted verbal aggression is less harmful for the target group’s intragroup functioning (i.e., verbal mimicry).
Methods
Research Context: Plenary Sessions of the 19th German Legislation Period
To better understand how targeted verbal aggression between two social groups relates to intragroup functioning of verbal mimicry, our study relied on recorded interactions, specifically stenotyped transcripts of plenary interactions between political representatives in the German parliament from the 19th governmental period. 2 We selected these transcribed plenary sessions because they allowed us to study verbal aggression in a socially embedded context (Benthin, 2019, November 13). These plenary discussions are public-sector meetings at one of the highest levels of a political institution. In those plenary sessions, the deputies compete for scarce resources in particular regarding attention that is paid toward proposals or drafts for novel bills; they can make counter-offers regarding these policy decisions. The debates usually follow government statements and deputies can deal with central political issues in particular drafts for new laws. In other words, these sessions serve as magnifying class in which groups compete for power. That is, these sessions serve the elected parties to debate about policies (of international and national scope), assess and amend the government’s legislative program, and discuss policy implications. The outcomes of these sessions have consequences on a national policy level.
Distinction of Groups
In our analyses, we distinguish between two social groups: members of the party Alternative for Germany (AfD, a newcomer party who comprise the smaller group) and non-AfD members (i.e., veterans who comprise the larger group). In our dataset, the Alternative für Deutschland (AfD) entered the 19th German parliament for the first time on the 24th of October 2017 (the starting point of our dataset), where they articulated the goal “to hunt the government.” 3 An analysis from two German media channels showed that mockery and scoffing comments (i.e., scoffing laughs) tripled in comparison with the 18th legislation period 4 (Benthin, 2019, November 13), concluding that the extent of verbal aggression has changed dramatically in the 19th legislation period. These changes have been associated with the entry of this new populist party (AfD), which obtained seats for the first time in the 19th legislation period. The AfD has received strong media attention due to various forms of verbal aggression (e.g., provocation and aggressive remarks) that occurred both inside (i.e., during the plenary session) and outside (i.e., in Twitter comments and interviews) of the parliamentary context (DPA, 2019, November 9; Benthin, 2019, November 13).
Procedure: Computer-Aided Text Analyses (CATA)
We used stenotyped and fully transcribed plenary sessions obtained from the archival repository of the German Bundestag (National Parliament of the Federal Republic of Germany). 5 We analyzed 113 fully transcribed sessions between 24th October 2017 and 13th September 2019. This sample contained approximately 6.5 million words (roughly equivalent to 19,000 A4 pages of transcribed text); each session contained about M = 57,476 words. We combined all sessions into one large dataset containing N = 71,500 speaker turns for our analyses.
To obtain measures of targeted verbal aggression (intergroup behaviors) and verbal mimicry (intragroup functioning), we conducted computer-aided text analysis (CATA, Short et al., 2018) using the transcribed sessions. CATA describes a class of different methods to analyze textual data in terms of their content and psycholinguistic properties (Short et al., 2018). CATA can be used with various text data (e.g., shareholder letters, online content, transcripts, speeches, calls, and group interactions). In this study, we applied it to verbatim transcriptions of these parliamentary discussions. We obtained summary text files from the AfD party and parliamentary veterans (non-AfD). Using CATA, we calculated the proportion of words in a text file that matched established lexical dictionaries, generating objective, replicable, and comparable measures for our constructs of interest (Short, McKenny, & Reid, 2018).
One advantage of the CATA method is that it explicitly captures verbal behavior, which taps into a more proximal and unobtrusive representation of actual behaviors rather than directly asking participants about their intentions using more traditional survey measurements (Klonek, Gerpott, Lehmann-Willenbrock, & Parker, 2019; Short et al., 2018). Two broad classes of CATA methods can be distinguished that fall into a spectrum of inductive to deductive methods (Short et al., 2018): The first (inductive) approach looks at patterns of words to derive higher-order themes that characterize a corpus of text. The second approach (which we adopted here) is deductive and analyzes the frequency of words occurring in a text corpus and matches them to one (or more) corresponding predetermined lexical validated dictionaries (that measure the underlying constructs of interest). Researchers can use CATA to explore how frequently words related to a construct occur (Mathieu et al., 2021). Extant research has identified and developed numerous validated lexical categories—including, for example, verbal aggression (Walker et al., 2017) or positive affect (Bantum & Owen, 2009)—that scholars have used for construct measurement (Gelfand et al., 2015; Mathieu et al., 2021). To operationalize our focal study measures, we used two software programs that allowed us to code the transcripts with validated lexical dictionaries that were relevant to our research questions. These software programs were the linguistic word count and inquiry (LIWC 2015; Pennebaker, Booth, & Francis, 2007) and the “Basic Unit-Transposable Text Experimentation Resource” (Boyd, 2021). We will explain the operationalization of our measures in more detail in the subsequent section.
Targeted and Nontargeted Verbal Aggression
To capture verbal aggression using CATA, we used five different dictionaries/lexica 6 (anger, swear, threats, hate, and aggression) that were all developed to capture the concept of verbal aggression and that showed evidence of construct validity in published research. These conceptually similar lexica were highly intercorrelated (indicated by high internal consistencies, i.e., Cronbach’s α).
In the following, we will describe each of the five dictionaries.
For the anger and swear dictionaries, we used two lexica incorporated in the LIWC 2015 software (Pennebaker et al., 2007). The anger lexicon contains words to assess the occurrence of anger-related words in natural language (e.g., “abuse,” “destroy,” “fight,” and “rage,” Meier et al., 2018). In terms of validity, multiple studies have demonstrated that the LIWC lexicon for anger correlates significantly with human-coded anger ratings (r = .25) and provides concurrent and discriminant validity with respect to other constructs (Bantum & Owen, 2009; Liess et al., 2008). The linguistic inquiry and word count [LIWC] swear lexicon screens for 244 swear words (e.g., “bullshit,” “stupid,” and “idiot”; Meier et al., 2018).
Second, we used the hate and threat lexica from a psycholinguistic dictionary that was developed to understand language use in the context of grievance-fueled violence threat assessments (Van der Vegt, Mozes, Kleinberg, & Gill, 2021). This dictionary has shown validity in distinguishing texts from violent and non-violent individuals (Van der Vegt et al., 2021). The dictionary specifically measures threat-related constructs and can be used for a wide variety of violence and extremism fueled by grievance. We selected the hate lexicon (containing 177 words, e.g., annoy, bitter, destruct, disgust, and evil) and the threat lexicon (containing 156 words, e.g., warn, provoke, fight, and revenge).
Third, we used the aggression lexicon from a dictionary developed by Gelfand et al. (2015), 7 which captures words related to aggression and wrongdoing.
We calculated percentage scores for each of the five lexica (number of words associated with the lexicon divided by the total number of words spoken) in each session (for AfD, i.e., smaller group, vs. non-AfD members, i.e., larger group). Using percentage score (%) accounts for differences in speaker lengths (i.e., overall word count). That is, we counted the number of words that belong to a respective lexicon (e.g., “anger”) and divided this count by the number of total words (aggressive and non-aggression words). For each session, we computed the mean across these five lexica (anger, swearing, threat, hate, and aggression) into an overall measure of verbal aggression separately for the smaller (AfD) group and the larger (non-AfD) group.
In a final step (explained below), we distinguished whether verbal aggression included a target (or did not include a target). This resulted in four measures of verbal aggression: (1) Targeted verbal aggression from smaller group (AfD), (2) Targeted verbal aggression from larger group (non-AfD), (3) Nontargeted verbal aggression from smaller group (AfD), (4) Nontargeted verbal aggression from larger group (non-AfD). Supplemental Appendix A gives examples for CATA coded statements from these four categories. Supplement Appendix B provides the full list of the words included in each of the five lexica. Supplement Appendix C compares the extent of targeted and nontargeted verbal aggression between AfD and non-AfD at the session-level.
Targeted Verbal Aggression From Smaller Group
To measure targeted verbal aggression from the smaller group, we examined whether AfD speakers used verbal aggression in combination with pronouns like you and yours (cf., Walker et al., 2017). In all sessions, speakers address other party members directly using the formal you (German: “Sie”). Thus, we identified speaker turns in which speakers communicated aggressive words in combination with the German “Sie”
8
(e.g., “yes,
For each session, we used a percentage measure to capture targeted verbal aggression (count of targeted verbal aggression from AfD divided by the total number of words spoken by AfD members). This measure of targeted verbal aggression showed an internal consistency of αtargeted aggression, AfD = .76 across the five lexica (i.e., anger, swearing, harm, aggression, threat, and hate).
Targeted Verbal Aggression From Larger Group
Using the same procedure as outlined above, we also coded transcripts for targeted verbal aggression from the larger group (non-AfD, e.g., “
Nontargeted Verbal Aggression From Smaller Group
We measured nontargeted verbal aggression by counting aggressive words that occurred in speaker turns without second-person pronouns and dividing it by the total number of words spoken by the smaller group (AfD) in the session (e.g., “Maybe then there won’t be so much
Nontargeted Verbal Aggression From Larger Group
Nontargeted verbal aggression from the larger group (non-AfD) was also coded (e.g., “I’m just talking myself into a
Relative Group Size
To assess the relative group size of the smaller group (for each session), we divided the number of AfD speakers by the number of all speakers (AfD and non-AfD) from the same plenary session (Mullen, 1991). This can be formalized as: N (AfD)/[N (AfD)+N(Non-AfD)], M = 15.41%, SD = 2.68). This measure is equivalent to and therefore perfectly correlated with the relative group size (of the non-AfD speaker group (N (non-AfD)/[N (AfD)+N(Non-AfD)]). 9
When this measure has a value of 50%, it means that the sizes between the two groups are balanced. When this measure is smaller (vs. larger) than 50%, the AfD group size is smaller (vs. larger) than the non-AfD group size. On average, the AfD group was smaller (15.41%) relative to the non-AfD group of parliamentary veterans. However, there were variations in relative group size (for the smaller group) across the sessions, which meant that the relative group size of the smaller group varied between sessions.
Intragroup Verbal Mimicry
To assess the extent of intragroup verbal mimicry processes for the respective groups, we calculated LSM scores in each plenary session across all members from each group (i.e., for the smaller AfD group and the larger non-AfD group, respectively). For the calculation of LSM, we followed the procedural analytical steps outlined by Gonzales et al. (2010).
First, we calculated each speaker’s language style (i.e., the extent of a speaker’s use of function words). That is, we calculated the extent to which each speaker (in a session) used nine types of function words: auxiliary verbs (e.g., to be, to have), articles (e.g., an, the), common adverbs (e.g., hardly, often), personal pronouns (e.g., I, they, we), indefinite pronouns (e.g., it, those), prepositions (e.g., for, after, with), negations (e.g., not, never), conjunctions (e.g., and, but), and quantifiers (e.g., many, few). For personal pronouns, for example, the percentage use might be 6.4% for Speaker 1, 5.7% for Speaker 2, etc. We based the selection of nine function words on the procedure suggested by Gonzales et al. (2010) and by Ireland et al. (2011), who developed the LSM measure, which includes an exhaustive list of function word dictionaries developed by Tausczik and Pennebaker (2013). From a psychological perspective, it has been argued that function words reflect how people are communicating, whereas content words convey what they are saying. In other words, function words arguably reflect unconscious processes (as opposed to content-specific words that were used for the verbal aggression measures). Second, we calculated LSM scores by comparing the individual language styles with the overall percentage of the remaining speakers in the same session (see Gonzales et al., 2010). Using this strategy, we calculated separate LSM scores for each speaker within a session. One LSMi score is calculated for each individual (of one session), and one LSMG-i score is calculated for the remaining speakers (of that same session).
To illustrate the LSM score for one lexical category (i.e., personal pronouns), LSM for speaker i is calculated in the following way:
LSMi =1‒ [(|prepi ‒ prepG|) / (prepi +prepG + 0.001)]*
Where prepG is the LSM score for the remaining speakers in the same session.
**In the denominator, 0.001 is added to prevent empty sets (see also Shi et al., 2019).
We carried out these calculations for all nine function words (i.e., auxiliary verbs, articles, common adverbs, and personal pronouns) and for all individual speakers (who contributed to each session, i.e., LSMi, LSMj, …. LSMn). The group-level LSMG scores were then derived by calculating the average across for LSMi to LSMn for all speakers from the same session. Finally, we calculated a composite LSM score by averaging across all nine function categories. Appendix D provides further details on validity analyses regarding the intragroup verbal mimicry measures.
Intragroup Verbal Mimicry (in the Larger Group)
To assess the extent of intragroup verbal mimicry processes for the larger group, we calculated LSM scores in each plenary session across all members (excluding AfD members). The internal consistency for the LSM score across the nine lexical categories was α = .94. LSM scores range from 0 to 1, where 1 reflects perfect LSM (high levels of verbal mimicry) between all speakers, and zero means no LSM (i.e., no verbal mimicry) between speakers (M = 0.89, SD = 0.05).
Intragroup Verbal Mimicry (in the Smaller Group)
Intragroup verbal mimicry for the smaller group was measured by calculating LSM scores (same procedure as described above) for all AfD speakers (i.e., excluding non-AfD speakers) (α = .79, M = 0.88, SD = 0.04).
Control Variables
Group Familiarity
We controlled for group familiarity because prior research has shown that group familiarity (i.e., the degree of shared experience that group members amass over time; Espinosa, Slaughter, Kraut, & Herbsleb, 2007) can affect group processes, such as coordination (Grijalva, Maynes, Badura, & Whiting, 2020) and information elaboration (Maynard, Mathieu, Gilson, Sanchez, & Dean, 2018). Group familiarity is often operationalized as the number of episodes a group has spent working together. Each plenary session constitutes a distinguishable episode. Following the approach of Grijalva et al. (2020) for archival data, we thus ordered the plenary session chronologically. We used this operationalization to capture the increasing level of familiarity arising from working together throughout the legislative period.
Session Size
Session size was measured by the number of all plenary speakers who actively contributed (i.e., speaking a minimum of 10 words) to a session.
Verbal Positive Affect
The literature suggests that positive affect is linked with positive relational group outcomes, such as social integration (Knight & Eisenkraft, 2015). Accordingly, we controlled for the level of verbal positive affect using the LIWC positive affect lexicon (containing 2239 words, such as “lovely,” “lucky,” “passion,” “perfect,” “pleasant,” and “awesome”). We calculated verbal positive affect for the smaller group (AfD: M = 2.59, SD = 0.36) and the larger group separately (non-AfD: M = 3.02, SD = 0.26).
Verbal Aggression (from a Previous Session)
To control for carry-over effects from verbal aggression from a previous session, we also captured the extent to which the smaller (AfD) group (and larger non-AfD group) exhibited targeted and nontargeted verbal aggression in a previous session.
Analytical Strategy
We conducted all analyses on the plenary session-level. To test H1 and H2, we ran ordinary least square stepwise regressions, using verbal mimicry (of the larger group) as the dependent variable. We used the relative group size of the smaller group, the level of targeted verbal aggression (from smaller group), and their multiplicative interaction (relative group size × targeted verbal aggression from smaller group).
Based on theoretical considerations (Bernerth & Aguinis, 2016), our analyses included multiple control variables [(group familiarity, verbal positive affect (for the smaller and larger group), targeted verbal aggression (for the smaller and larger group), nontargeted verbal aggression (for the smaller and larger group), and both targeted and nontargeted verbal aggression from previous sessions (for the smaller and larger group)].
Following best practice recommendations by Bernerth and Aguinis (2016), we ran our models once with the inclusion of these controls (see Supplement Appendix G and I) and once without the inclusion of controls. We compared the model fit to determine if the inclusion of controls is warranted. We only kept control variables in the final analyses when adding a control variable to the model resulted in a significant increase in model fit. Our results (particularly the resulting inference from hypothesis tests) did not change when we included these control variables.
Results
Predicting Intragroup Verbal Mimicry in the Larger Group
Descriptive Data for and Correlations for Variables at the Session Level.
Note. N =110–113, *p < .05, **p < .01, AfD = “Alternative für Deutschland.”
Regression Model Predicting Intragroup Verbal Mimicry (in the Larger Group, non-AfD).
Note. †p .<10, *p <.05, **p <.01 (two-sided), smaller group = AfD (“Alternative für Deutschland”).
Table 3 displays the moderated regression analyses predicting intragroup verbal mimicry in the larger group. H1 predicted that targeted verbal aggression of a smaller group is negatively related to the larger group’s intragroup functioning (i.e., verbal mimicry). In line with H1, we found a negative main effect for targeted verbal aggression (from the smaller group) on verbal mimicry (the larger group) (B = −.11, p = .003; see Table 3). Next, we included the interaction term (Model 2, Table 3). The estimate for the interaction term predicting verbal mimicry (in the larger group) was also significant (B = 3.10, p < .001), thus offering support for H2.
To better interpret the significant moderation effect (H2), we used the Johnson–Neyman technique to highlight regions of significance (Hayes, 2017). This method is preferable to the more traditionally used simple slope methods because the choice of the conditional values for the moderator variable in these traditional methods is ultimately arbitrary (“pick-a-point” approach in which researchers typically select the following three values for the moderator: −1SD, mean, +1SD). The Johnson–Neyman technique essentially works backward and identifies the full range of the moderator for which the interaction is significant; this technique identifies the values of the moderators (here: relative group size of the smaller group) for which the relationship between targeted verbal aggression (from smaller group) and intragroup verbal mimicry (in the larger group) is significant. The upper line in the resulting plots indicates the upper region boundaries of significance (the higher 2.5%), and the lower line indicates the lower region boundaries of significance (the lower 2.5%). The middle line indicates the direction (i.e., positive or negative) of the relationship. The upper and lower confidence bands indicate the regions of significance for the moderator. When the upper and lower band are both below zero (or when they are both above zero on the x-axis), the moderator has a significant interaction effect on the x-to-y relationship (Hayes, 2017).
Figure 2 shows the plotted confidence bands for the moderating role of relative group size (AfD) on the relationship between targeted verbal aggression from smaller group (x) and intragroup verbal mimicry (larger group) (y). As shown in Figure 2, when the AfD relative group size was below 17%, their targeted verbal aggression had a significantly negative association with intragroup verbal mimicry of the larger group. For sessions with an AfD relative group size above 17%, targeted verbal aggression from the AfD group no longer had a significant association with intragroup verbal mimicry (of the larger group). Furthermore, when the relative group size (AfD) was higher than 25% in a session, targeted verbal aggression from the AfD group was positively associated with intragroup verbal mimicry in the larger group. This flipped direction of the effect supports the interaction hypothesis (H2), which argues that the relative group size is a key social context factor that leads to higher expectations of verbally aggressive behavior (when their group size is relatively large) by members of the other group. These expectations of aggressive behavior due to relative group size appear not only to buffer the negative impact of targeted verbal aggression but are even associated with improved intragroup functioning in the other group. Regions of significance for the moderator (Johnson–Neyman technique) for predicting verbal mimicry in the larger group. Note. Johnson–Neyman plot for the moderating role of AfD relative group size on the relationship between targeted verbal aggression (of the smaller group) and intragroup verbal mimicry (in the larger group). −1SD corresponds to low levels in AfD relative group size, M corresponds to mean levels of AfD relative group size, and +1SD corresponds to high levels of AfD relative group size. The bounds of significance are indicated by the two vertical dotted lines: In the Johnson–Neyman plot, the relationship between targeted verbal aggression and intragroup verbal mimicry (larger group) is significant for relative group size levels below 17%, non-significant for AfD relative group size levels ≥17% or ≤25%, and significant for AfD relative group size levels higher than 25%.
Finally, we conducted additional analyses to rule out alternative explanations for the interaction effect. For example, the possibility that there is a small group of very aggressive, but active, AfD members and the possibility that effect is diffused as more (but less aggressive) AfD members participate. Our additional analyses suggest that the interaction effect cannot be explained by this alternative hypothesis (see online Supplement Appendix H for details).
Supplemental Analyses: Predicting Intragroup Verbal Mimicry in AfD
Regression Model Predicting Intragroup Verbal Mimicry (in the Smaller Group, AfD).
Note. †p .<10, *p <.05, *p <.01 (two-sided), smaller group = AfD (“Alternative für Deutschland”).
We tested the effect of targeted verbal aggression from the larger group on intragroup verbal mimicry in the smaller (AfD) group and the moderation of relative group size (H2, which predicted that relative group sizes moderate the effect of targeted verbal aggression on intragroup verbal mimicry).
Specifically, we evaluated the effect of the control variables on intragroup verbal mimicry (in the smaller group) in multiple steps [Step 1: familiarity, session size; Step 2: targeted/nontargeted verbal aggression (both from the larger and smaller group) from a previous session; Step 3: positive affect (smaller group) and positive affect (larger group); Step 4: Targeted/nontargeted verbal aggression (larger group) and nontargeted aggression (smaller group) from the current session].
The inclusion of session size resulted in a significant model fit (Step 1: ∆F(2, 105) = 24.44, ∆R 2 = .32, p < .001), while including verbal aggression from the previous session did not improve model fit (Step 2: ∆F(4, 101) = 1.82, ∆R 2 = .05, p = .132). Furthermore, including control variables from Step 3 (∆F(2, 99) = 3.07, ∆R 2 = .04, p = .051) and Step 4 (∆F(3, 96) = 2.63, ∆R 2 = .05, p = .054) did not improve model fit significantly (for details see Supplement Appendix I). Although changes in model fit with controls for Step 3 and Step 4 were non-significant, we noticed that p-values were often borderlining at the 0.05-threshold, hence, we inspected which variables contributed to this. This inspection revealed that positive affect (larger group) (B = 2.75, p = .015) and nontargeted verbal aggression (smaller group) (B = −.15, p = .012) had significant effects on intragroup verbal mimicry (of the smaller group). Thus, we decided to include these control variables in our focal analyses.
Table 4 displays the results of the moderated regression analyses predicting intragroup verbal mimicry (in the smaller group). There was no significant effect for targeted verbal aggression (from the larger group) (B = −.04, p= .665; see Table 4) on intragroup verbal mimicry (of the smaller group). However, our results showed a significant interaction effect between targeted verbal aggression from the larger group with relative group size on intragroup verbal mimicry (in the smaller group) (B = 4.63, p = .046; see Table 4), which provides further support for H2.
Figure 3 shows the plotted confidence bands for the moderating role of relative group size on the relationship between targeted verbal aggression from the larger group (x) and intragroup verbal mimicry (of the smaller group) (y). As shown in Figure 2, when the relative group size of non-AfD speakers was below 76% (which is more than two standard deviations below the average relative size of this group, M = 84.4%), then their targeted verbal aggression had a significantly negative association on intragroup verbal mimicry of the smaller (AfD) group. In other words, only when the group of non-AfD speakers was substantially smaller relative to their average relative group size (M = 84.4%), then targeted verbal aggression from this group was negatively associated with the smaller group’s intragroup functioning. Regions of significance for the moderator (Johnson–Neyman technique) for predicting intragroup verbal mimicry in the smaller group. Note. Johnson–Neyman plot for the moderating role of larger group relative group size on the relationship between targeted verbal aggression (larger group) and intragroup verbal mimicry (of the smaller AfD group). −1SD corresponds to low levels in non-AfD relative group size, M corresponds to mean levels of non-AfD relative group size, and +1SD corresponds to high levels of non-AfD relative group size. The bounds of significance are indicated by the two vertical dotted lines: In the Johnson–Neyman plot, the relationship between targeted verbal aggression (of larger group) and intragroup verbal mimicry (in the smaller AfD group) is significant for non-AfD relative group size levels below 76% and non-significant for non-AfD relative group size levels ≥76%.
Discussion
Our study’s primary goal was to investigate how verbal aggression between groups might potentially harm intragroup functioning in the targeted group. In doing so, we specifically extended previous research by considering recent theorizing on group size differences (Carton & Cummings, 2012) and introduced relative group size as a moderator. To empirically test our hypotheses, we applied a text-analytic approach using plenary discussions between a newcomer party (with the explicit goal to disrupt) and veteran members of the parliament and coded targeted verbal aggression (from both groups) to investigate their impact on their respective intragroup functioning. We further argued that relative group size would moderate this relationship because it helps the other group to build expectations on whether they should anticipate verbal attacks. In other words, targeted verbal aggression coming from a relatively smaller group would be more unexpected, potentially causing detriments to the intragroup functioning of the target.
Conversely, with increasing group size, opponents might anticipate and prepare against targeted aggression, thus leading to less harmful or even positive effects of targeted verbal aggression on intragroup functioning. In line with our reasoning, our results showed that relative group size and targeted verbal aggression interactively affected the level of intragroup functioning (measured as verbal mimicry). When relative group size diminished, targeted verbal aggression harmed verbal mimicry in the other group, whereas when relative group size increased (relative to the other group), targeted verbal aggression was no longer associated with verbal mimicry in the other group; to the point that (when relative group size grew larger) targeted verbal aggression even showed a positive association with intragroup verbal mimicry in the other group. Our study provides important insights into the intergroup aggression literature, explaining when targeted verbal aggression between groups has the potential to harm (or to improve) intragroup functioning.
Theoretical Implications
Our study has at least three theoretical implications for the intergroup literature in general (e.g., Benard, 2012; Bornstein, 2003; LeVine & Campbell, 1972; Sherif, 1966) and for the organizational behavior literature with respect to verbal aggression more specifically (e.g., Hershcovis & Reich, 2013; Miner et al., 2018).
First, our research contributes to a better understanding of how verbal aggression is communicated by considering whether the verbal attack is personal (i.e., targeted) or not. In an intergroup setting, verbal aggression can be expressed without necessarily addressing a target or making it personal. Because the vast majority of verbal aggression research (including concepts like incivility) relies on survey methods that do not distinguish between targeted versus nontargeted verbal aggression, there has been little attention regarding how verbal aggression is enacted as a behavior (for exceptions, see Walker et al., 2017). This insight has broader ramifications for improving our understanding of the types of verbal aggression that can then affect either individual or intragroup functioning. Our findings entail that future research on workplace verbal aggression should move away from an attributional perspective that differentiates between instigators versus victims of aggression toward a perspective that focuses on verbal aggression as a behavior that can be targeted versus nontargeted. In fact, previous research has speculated that this may be important against the backdrop that “targets may become perpetrators, and vice versa” and that “there is mounting evidence that individuals often occupy both roles” (Hershcovis & Reich, 2013, p. S29). In line with this notion, our supplemental analysis shows that not only the focal group’s (i.e., AfD) targeted aggression impaired intragroup functioning when their relative group size was smaller, but that similarly, targeted verbal aggression from the larger group was also moderated by their relative group size with respect to how it impacted the intragroup functioning in the smaller group.
Second, our study contributes to a better understanding of when targeted verbal aggression is harmful by introducing relative group size as an important social context factor that moderates the relationship between targeted verbal aggression and intragroup functioning. In doing so, we not only address the call for a better understanding of (social context) moderators that change the focal relationships between verbal aggression and its outcomes (Hershcovis & Reich, 2013) but also manage to reconcile two seemingly opposing views on the impact of intergroup aggression on intragroup functioning. That is, our research helps us understand under which conditions intergroup hostility has the potential to divide (conflict spillover perspective; Jehn et al., 2013) or unite (realistic conflict perspective; Sherif & Sherif, 1966) a group in terms of its effect on verbal mimicry. Specifically, we were able to show that relative group size is an important social context cue, which influences how hostility between groups can either be detrimental or conducive to intragroup functioning. Social context cues such as group size can shape expectations of aggression, which may be the key to explaining the opposing effects of intergroup verbal aggression on intragroup functioning. With low or no expectation of aggression, intergroup verbal aggression can potentially tear groups apart; when expectations are present, however, they have the potential to prepare and unite groups against their opponents’ verbal attacks. The results of our study indicate that intergroup research is well-advised not to analyze intergroup behaviors irrespective of their social context (in the form of group sizes). In other words, considering the group size as an important social context factor may have far-reaching consequences for the intergroup literature more generally.
Third, our research also provides an important methodological extension with respect to studying verbal aggression “in the wild,” thereby addressing the recently growing interest in using innovative methods to study phenomena within real-world contexts (Klonek et al., 2019). So far, most studies of verbal aggression in the OB and management literature typically focus on perceptions of verbal aggression (either by surveying the targets about how much they experienced this behavior or by asking perpetrators whether they exhibited verbal aggression). In this study, we relied on objective and unobtrusive measures of verbal aggression by analyzing spoken language (through transcripts) in terms of the extent to which group members use an aggressive tone, thus using a measure of verbal aggression that does not rely on introspection. These measures have the advantage that they do not rely on introspection or participants’ accurate and full awareness of how or why they feel, think, react, or behave in a certain way (Hill et al., 2014). In other words, unobtrusive verbal aggression measures can be especially helpful to our understanding of verbal aggression because the impact of this behavior may span far beyond what targets or witnesses may introspectively realize or convey. For our purposes, we were able to study how targeted verbal aggression from one group impacted the intragroup functioning of the other group (which may not necessarily have been consciously aware of this effect). In doing so, we were also able to abstain from differentiating between “perpetrator” and “targets.” That is, through our focus on verbal behavior, we explicitly consider both groups as potential aggressors. In doing so, we speak to research that has questioned whether “labels such as ‘perpetrator’ and ‘target’ are practically or theoretically meaningful” (Hershcovis & Reich, 2013, p. S27). More generally, our research showcases that using (archival) text data can offer high levels of external validity and a potentially more objective approach to understanding verbal aggression and its adverse impact on intragroup outcomes. Relatedly, a focus on behavioral indicators could also help overcome the challenges of construct proliferation in the workplace mistreatment literature (Hershcovis, 2011; Hershcovis & Reich, 2013).
Practical Implications
Intergroup hostile behavior frequently arises—not only between political parties but also between nations, companies, departments within companies, and in many more workplace contexts (Benthin, 2019, November 13; Kalmoe, 2014). Accordingly, the insights of our study can be used to derive practical implications that can be used in education about intergroup relations and how to better deal with verbal aggression between groups. More specifically, our study is relevant to understanding under which conditions intergroup hostility either threatens or fosters intragroup functioning, thereby providing important advice for groups to prepare or intervene toward acts of verbal aggression.
First, our differentiation between targeted versus nontargeted verbal aggression emphasizes that it is particularly important to intervene when verbal attacks become personal. For example, when group members notice that the verbal tone in a discussion is becoming more aggressive and that these attacks become targeted (e.g., noticing second-person pronouns like “you”/“yours”), they should promptly intervene to de-escalate the situation and prevent further harm.
Second, groups should be educated that relative size differences matter in intergroup contexts. Specifically, relative size affects how vulnerable groups are to targeted verbal aggression from other groups. Particularly, when the relative group size of aggressors is relatively small, a larger group is more vulnerable to targeted verbal aggression.
Third, as we elaborate in more detail in the future research section, our study also demonstrates that innovative methodological approaches (i.e., automatically tracking the level of verbal aggression in spoken words) could be of use for the detection of targeted verbal aggression. While these novel methods are only just starting to emerge as a function of recent technological developments (Buengeler, Klonek, Lehmann-Willenbrock, Morency, & Poppe, 2017; Klonek et al., 2019), there is potential for them to be used for real-time interventions in the (near) future. For example, group members could learn how to identify verbal aggression in a discussion (based on paying attention to aggressive words) and use this for determining when things are moving in the wrong direction. Offering communication training that helps group members understand how certain words are hurtful to others could be particularly important, as some members may express verbal aggression without being aware of it. This could allow co-workers to address the verbal aggression immediately and could help to repair broken social relationships. Increased awareness and calling out of verbally aggressive language could be an important first step in dealing with group members that display verbal aggression.
Limitations
Our research has the following limitations. First, although our study’s reliance on real plenary discussions constitutes a strength in terms of external validity, it is uncertain how generalizable these results would be to other populations (e.g., groups with a hierarchical structure such as workgroups with a manager). Second, our measure of verbal aggression was restricted to verbatim expressions, whereas it should be noted that anger and threats can also be communicated through gestures and other non-verbal channels. That is, our study falls short in measuring non-verbal aggression (rolling one’s eyes, facial expressions, pointing with a fist at someone, etc.). Third, our theoretical arguments (i.e., relative group size moderating the impact of targeted verbal aggression) relied on some unmeasured mechanisms (e.g., expectations of whether verbal aggression will occur). However, due to the nature of the archival dataset, it was not possible to obtain perceptual self-report measures. Fourth, our study used observational data; that is, we did not experimentally manipulate the predictor and moderator variable, which renders our study limited with respect to making causal conclusions. Fifth, some of the verbal aggression measures had internal consistencies below a recommended (and more conservative) cut-off value of α ≥ .80, which indicates some limitations regarding the reliability of these measures. It should be noted that reliability measures for text-based and lexical measures are typically much lower than what researchers would expect for survey-based measures (Tauscik & Pennebaker, 2013). That is, while it is considered standard practice to report reliability like Cronbach’s alpha, some researchers have questioned whether these analytical approaches, which were originally intended to evaluate multi-item self-report surveys, are appropriate methods to estimate the reliability of text-based dictionary measures (Boyd & Pennebaker, 2015; Meier et al., 2018).
Future research
Our study brings some exciting opportunities for future research. First, against the backdrop of the increasing use and ease of digital technologies that allow easy transcriptions of verbal behaviors, we hope to see more research using CATA-based measures of verbal aggression. Due to the COVID-19 pandemic, many organizations have shifted work practices toward virtual meetings using digital meeting platforms (like Slack or MS Teams) that often automatically transcribe verbatim what participants are saying during these meetings. Our study provides methodological ideas about how to use similar approaches to better understand verbal aggression (and other forms of mistreatment) during recorded workplace interactions. This also provides an opportunity for technology-based intervention studies that evaluate how technology-based feedback could be improved to reduce the occurrence of targeted verbal aggression or better ways to intervene when this happens. It might be promising to utilize technology for real-time warnings about targeted verbal aggression (Buengeler et al., 2017). In that regard, our study may give engineers, computer scientists, and social researchers some ideas for assisting organizations in developing tools that track the level of targeted verbal aggression and thinking about ways that help improve how group members interact with one another. Future research also would need to identify how this feedback should be delivered, whether it is accepted by the aggressors, and when it is most effective. To conclude, we are excited to see what the future of technology-supported text analysis will have to offer for a better understanding of intergroup relations, which could ultimately help reduce the risk of harming intragroup functioning.
Second, future research should further investigate social context moderators that shape the impact of workplace verbal (and physical) forms of aggression on the targets of aggression. This could include departmental affiliation, organizational or team hierarchy position, or organizational norms. We also hope to see more group-level studies and the use of experimental designs considering the role of social context, which can help organizations to better understand the role of mistreatment and aggression, when it happens, how it can be prevented, and under what conditions it has most adverse effects.
Supplemental Material
Supplemental Material - When Groups of Different Sizes Collide: Effects of Targeted Verbal Aggression on Intragroup Functioning
Supplemental Material for When Groups of Different Sizes Collide: Effects of Targeted Verbal Aggression on Intragroup Functioning by Florian Klonek, Fabiola, Gerpott, and Lisa Handke in Group & Organization Management.
Footnotes
Author’s contribution
The first author (F.K.) conceptualized the research ideas, collected the data, developed all measures, carried out the analyses, wrote the paper, and responded to the reviewer and editorial feedback. The second author (F.G.) assisted in responding to reviewer feedback. The third author (L.H.) helped in double-coding data for supplemental analyses.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplementary material for this article is available on the online.
Notes
Author Biographies
Associate Editor: Ivana Vranjes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
