AbstractThe riot that occurred on January 6th, 2021, at the Federal Capitol buildings represented a form of political violence rarely witnessed in the history of the United States. Thousands of people attended the riot, and many were criminally charged with violent and nonviolent offenses. Criminal case files provided by the Federal Bureau of Investigation contained social media posts made in anticipation of the riot, thus creating an unprecedented opportunity to assess linguistic indicators of future political violence. These data permitted the rare empirical exploration of why some individuals became violent when acting under the same pretext as their nonviolent peers. In the present study, individuals were divided into two groups: those who were charged with violent (n = 25) and nonviolent crimes (n = 55). Natural language analyses of their social media posts in the 3 months prior to January 6 identified that violent rioters used rhetoric which displayed weaker group affiliation. The violent rioters also used less analytical and more questioning language than their nonviolent counterparts. The findings indicate that violence during the U.S. Capitol Breach on January 6th was not an extension of zealous group identification, and that these rioters may have been unique in their motivations to engage in violent political action. This investigation included a small sample, and whether the language of those who were fortunate enough to evade penalty, or posted on encrypted social media networks, would adhere to the patterns in the current findings is unknown.
IntroductionIn the months following the 2020 presidential election, the Trump administration claimed the results of the recently held election were fraudulent1. Trump supporters met his call to march to the Capitol building on January 6th, 2021, to overturn the results. The gathering erupted into violence when his supporters stormed the Capitol resulting in death and destruction2.Research over the last decade suggests that few politicized individuals make the transition to violent political action, and the path is not necessarily linear nor gradual3. Political violence may be the ultimate manifestation of strong group loyalty coupled with emotional lability and aggressive tendencies. How this transition unfolds in the minds of those who eventually commit violence is unclear but some have suggested that group-identity processes4 may play a central role. If so, the path toward violence would be paved by the progression of indoctrination. Alternatively, group affiliation may be unimportant to violent political action. Their motivations may be more oriented toward the ideological goals5 of creating political upheaval or gaining societal significance than seeking to affiliate with the group. Group affiliation may also be unimportant to those who may be prone to aggression, as a basic personality trait.In some cases, terrorists have published manifestos in anticipation of their violence, and the language they use prior to the event distinguishes them from other political extremists who do not aggress. One example is that violent extremists (e.g., al Qa’ida) have used different language patterns that were less analytical than their non-violent Arabic-speaking counterparts6. These written accounts have been valuable in gaining an understanding of the factors that underlie violence of this manner. Many of those who participated in the insurrection on January 6th used social media accounts to announce their displeasure with the election results and declare intentions to gather at the Capitol. The Trump supporters at the Capitol followed a common social-media pattern in which homogenous networks coalesce around shared grievances which furthers groupthink and normalizes radical ideology7. Those who engage in political violence may declare their intentions prior to the event, and this has been referred to as “leakage”8. The ubiquity of social-media use provides the opportunity to identify leakage of impending violence in posts leading up to the Capitol Breach. Instances of leakage would not be expected to be as simple as an honest and direct pronouncement of imminent violence, but evidenced in ways that follow linguistic patterns that have been established in other research6,9.The data for the current analyses came from case files provided by the Federal Bureau of Investigation prior to the mass pardons of those who participated in the riot by the newly re-elected Trump in January, 2025. The FBI case files contained social media posts from November 1, 2020, through January 5th 2021, concerning the January 6th riot. Data were analyzed using the text analysis program LIWC-2210. We compared those who were charged and convicted with violent offenses to others who were charged and convicted but for non-violent actions. The availability of social media data provides a unique opportunity to witness the thought processes, via linguistic markers, of those who were planning their participation in a political insurrection, and two distinct hypotheses were tested. According to a dominant theory of political radicalization3, there is an important distinction between support for political violence versus actively participating in it, as there is between violent and nonviolent (but still illegal) action.Strong group identification hypothesisRadicalization has been associated with heightened group identification11. For instance, identity-fusion theory11 asserts that the blurring of the boundary between personal and group identities fosters a willingness to aggress and sacrifice physical safety for the group12. Seeking stronger ties to a group is one manner in which people may resolve feelings of insecurity13. In the case of the Capitol Breach, the political insecurity that was fueled by assertions of the former president that the election results were fraudulent may have contributed to the need to affiliate14. The strong identification with a group may lead some toward radical action who would otherwise not engage in violence3. In this manner, those who became violent were anticipated to more strongly affiliate with the group and more fervently express their support for the insurrection than their nonviolent peers. Strong group identification should be evidenced in natural language processing as two facets of “unquestioning” and “affiliation”4. Unquestioning, in this case, would be expressed through confident language without evidence of ruminating or questioning their actions. Previous studies have captured this state through the low usage of “thinking through” words, such as think, try, because, should4, which are words contained the in standard LIWC output of “cognitive processes.” Those strongly identifying with the group tend to accept group norms, goals, and actions with limited thoughtfulness or absence of questioning. Affiliation is evidenced in greater use of “we” pronouns and prosocial words (e.g., “help”). Also, since negative emotions have been associated with strong group membership and support for political violence, we predicted that strong group identification would be displayed as passionate emotional declarations in social media which would include much anger9,15.Individual quest to gain significance hypothesisAnger at witnessing the political defeat of a popular leader may have contributed to violence, as the frustration-aggression hypothesis would suggest16. The political defeat would also be associated with perceptions of a loss in societal standing. The individual’s need to restore their social significance, in theory, justifies violence and enhances impulses to aggress17. For these reasons, we expect those seeking to restore significance through violence would use words containing more negative emotion9,18. Therefore, any evidence that the violent group engaged in more negative emotional rhetoric than the nonviolent group would fail to distinguish our hypotheses. Although justifications for violence may be reinforced by group dynamics19, strong group affiliation may not be a way to relieve frustration nor be necessary for one to gain significance. For instance, lone-actor terrorists may gain societal or historical significance despite being socially isolated20. Conspiratorial beliefs, which are common among lone actors21, are ideological narratives which are constructed and promoted to establish significance5,22.Those most prone to political violence may have strategized and openly considered their involvement in the January 6th demonstration in the weeks leading up to it. Their social-media posts may have displayed uncertainty about the actions needed to restore their significance. They may have also entertained complex conspiratorial narratives. So, in contrast to the group-identification hypothesis, those seeking to establish significance may show more evidence of an unsettled mind -- one prone to questioning, preoccupation, and uncertainty. Those seeking to gain significance may also self-characterize as having a unique capacity to act and feel that they alone are shouldering the burden of the necessity and urgency to act3. If this were the case, mounting anticipation and uncertainty would interfere with logical-reasoning skills6. Therefore, the language of those who became violent should show evidence of low analytical thinking and greater use of cognitive processing words6,18 which demonstrate more questioning and uncertainty.The two hypotheses, strong group identification and individual quest to gain significance, were tested by using natural language analysis (LIWC). The analysis was performed on social media posts of those who participated in the U.S. Capitol Breach in anticipation of the riot. Both hypotheses suggest that violent rioters should use more emotional language than nonviolent rioters; but the hypotheses differ regarding the degree of questioning language displayed by those who would behave violently. Violent rioters may use more questioning and less analytical language in their social media posts, according to the quest to gain significance hypothesis. The strong group identification hypothesis places emphasis on group affiliation and unquestioning loyalty to a cause as a condition of violence.MethodsSampleThe U.S. Department of Justice (DoJ) published FBI case reports submitted to the federal court, which included documentation (i.e., screenshots and/or transcriptions) of defendants’ social media posts related to their participation in the January 6th Capitol Breach. These case files were once available at the United States Department of Justice, Capitol Breach Cases website23. This publicly available information was used to gather data in January 2024, and the exact charges filed as well as social-media data were aggregated into a dataset.For inclusion of social media data into our dataset, the language content needed to be posted or shared openly to all the individual’s followers, thus text sent privately through direct messaging (e.g., Facebook Messenger) was excluded. The language used in each social media post was required to be original content; in other words, if the individual posted language written by another person (i.e., a “repost” of another person’s text, or an image that contains text such as an event flyer), it would be excluded from the dataset. The language was required to be written text, thus, spoken words captured on video and posted to social media were also excluded from the dataset. A total of 234 defendants’ case reports included social media data that met the aforementioned criteria. Of these, 108 had uploaded posts between the time of the election (November 3rd, 2020) and the Capitol Breach on January 6th, 2021. Social media posts made on January 6th were excluded because these would not be made in anticipation of the event. Many individuals made several posts to social media accounts (mean number of posts = 2.48, standard deviation = 2.01), and the total accumulation of text was used to perform LIWC analysis. The overall amount of text varied, and we excluded those individuals from the analysis who failed to create enough text to justify linguistic analysis. Of those 108 individuals criminally charged, those who used less than 20 words were excluded, yielding a final data sample of 81 individuals. Among those 81, only one case had been dismissed and was excluded from the analysis. Of the remaining 80, all had been sentenced. Table 1 contains demographic information on the sample used in the analysis.Table 1 Demographic variables of violent and nonviolent rioters for this analysis relative to the larger sample of the mob27, including means, percentages and standard deviationsFull size tableThere was a total of 26 unique charges filed against those individuals (Supplementary Table 1). The mob of rioters used in the sample was comprised of mostly white men from nearly every state in the U.S., many of whom were affiliated with militia groups. On average, those included in this study were charged with 4.6 crimes. Each unique charge presented across all cases was listed as a variable in the master dataset (e.g., the variable labeled “Charge 1” was Act of Physical Violence in a Capitol Grounds or Building, the variable labeled “Charge 2” was Aiding and Abetting, etc.). These charges were then bifurcated between those which were violent in nature (e.g., contained some element of assault), versus those which did not. This yielded 55 individuals in the nonviolent group and 25 in the violent group. All study procedures described herein were approved by an Institutional Review Board (IRB00021, Federal Wide Assurance #6316).Data AnalysisLanguage analyses were performed using LIWC-22 software. Although the program analyzes 118 language dimensions, the current study focused on dimensions that are theoretically relevant to the project. For our analyses, we were interested in (I) unquestioning affiliation, (II) analytical thinking, and (III) emotional expression. Unquestioning affiliation was calculated using the equation from4 which uses summed z scores obtained from the LIWC categories of cognitive processes and affiliation. The cognitive processes category included the sum of certitude, tentative, differentiation, and discrepancy subcategories. Analytical thinking, formerly known as the “categorical dynamic index” is a combination of function words, including pronouns. The major category of emotion was used, which contains all LIWC subcategories of anger, sadness, anxiety, and positive and negative emotions. A full list of LIWC categories and subcategories is published elsewhere10.The study of terrorism and LIWC has employed simple between-group parametric analyses9. Following that precedent, we used planned independent-samples t-tests to compare violent to non-violent actors among four LIWC output variables. For all analyses, equal variances could not be assumed to be uniform across both groups, and Levene’s test was used to determine which t-test values to use for statistical interpretation (equal or unequal variances). In the event of a failure to reject the null hypothesis, Bayesian factors were calculated for independent samples using priors of unequal variances. D’Agnostino and Person’s test was used to assess normality. Cohen’s d was used to assess effect size. All statistics and data visualizations were performed in SPSS (v. 29.0) and Prism Graphpad (v. 10.0). This study was not preregistered.ResultsIndependent-samples t-tests (two-tailed) were performed and significant differences between groups were revealed for unquestioning affiliation with a strong effect; t(78) = 2.41, p = 0.02, d = 0.58, CI:0.10-1.06. The violent group used fewer analytic words; t(78) = 2.06, p < 0.04, d = 0.50, CI: 0.20-0.97. All data in Fig. 1 passed tests of normality. Differences in emotional words were not apparent between groups (t(78) = 0.95, p = 0.34, d = 0.23, CI:-0.25-0.70), and there was moderate evidence to accept the null hypothesis (BF = 3.61). Differences in rhetorical negative affective tone were not apparent between groups; t(78) = 0.96, p = 0.34, d = 0.23, CI:-0.24-0.71, BF = 3.59. The number of words (LIWC-22 “word count”) did not differ between groups; t(78) = −0.06, p = 0.95, d = -0.02, CI:-0.49-0.46, BF = 5.46). All LIWC-22 output variables and between-group comparisons are listed in Supplementary Table 2.Fig. 1: Linguistic differences between violent and nonviolent rioters at the Capitol Breach in their social media posts prior to January 6th, 2021.Linguistic inquiry word count (LIWC-22) was performed on social media posts of 26 violent and 55 nonviolent rioters charged for participation at the U.S. Capitol Breach. The figure compares both groups on two different indices from LIWC, these being unquestioning affiliation and analytic language. Means and 95% confidence intervals are depicted in red.Full size imageDiscussionNatural language analysis (LIWC-22)10 identified significant and robust differences between nonviolent and violent offenders among the measures for unquestioning affiliation and analytical thinking (Fig. 1). The most surprising discovery was that the violent rioters were much lower in unquestioning affiliation4 than their non-violent peers. The results run counter to the commonsense notion that violent behavior reflects extreme levels of group identification. The lesser group affiliation expressed by violent rioters is noteworthy because it suggests that they exist on the periphery of the group of political rioters. Furthermore, the degree of questioning expressed in language of violent actors was evident in their text, for example: “Seeing this country tear itself apart breaks my heart and weighs heavy on my chest, nobody wants to do what we all know needs to be done, but it’s our duty to make sure this never happens to any future generations ever again. I’m willing to give up everything I have and my entire lineage to make that happen.” This finding, coupled with the lower use of analytical words, supports the quest for significance hypothesis. One caveat is that there was no greater use of emotional words nor negative rhetorical tone, which ran counter to both of our hypotheses.The unsettled, questioning mind of violent rioters was apparent in their tendency to adopt the rhetoric of conspiracy. As one violent rioter said: “When you’re prevented from watching as a half million votes are counted, it’s hard to have evidence. There probably isn’t much evidence that Stalin cheated to win elections either… Simple question: When a party flagrantly ignores election laws, is there supposed to be zero consequences for that? Are we supposed to just say: Sorry, it’s too late, all the illegal votes have already been counted? If the Democrats get away with their crimes, there’s nothing to stop them from cheating in future elections. Why shouldn’t amoral operatives cheat if there are no consequences?” The adoption of these misconceptions about the validity of the election and the origins of the fraud indicate that those who became violent strongly identified with the need24 to correct perceived injustice and establish their societal significance by doing so. These findings are not unique; others have found that conspiratorial narratives are emblematic of violent political action, in particular lone-actor terrorists9. Although their methods of violence differ, lone actor terrorists and violent insurrectionists may share the essential need to convey narratives15 which support their unique perspectives.Embedded within unquestioning affiliation are LIWC-22 subcategories which include differentiation and tentative words (e.g., “if” or “or”, Supplementary Table 2). These words were used by violent actors to create logistical conditions under which violent action would be triggered, such as: “If you accept the reality that the election was stolen, then you cannot accept Biden as the new president & neither can I… I won’t sit idle while the nation is stolen.” And also, “If people do not fight now -- go out and protest the vote and election fraud -- then your voting days are over.” Although only speculative, perhaps the cognitive tendency to develop conditional triggers served as psychological thresholds beyond which violence would be deemed contextually acceptable.The transition to a new leadership in January of 2021 was associated with large-scale societal instability which accompanied a sitting President’s refusal to accept political defeat. Devoted members of the Republican Party may have experienced a form of identity-uncertainty related to the government’s instability, and both identity uncertainty and personal uncertainty have been linked violence25. However, this crisis of uncertainty should be resolved by ardent group identification; theoretically, it is the need to affiliate with a group which leads to violent action25. We found the opposite; violent actors were less motivated by social affiliation, and they may have been driven by an ambition to maximize their significance by promoting ideology5,17,22. Populist support for violence in the United States is associated with fears that the majority will be marginalized by a growing racial and ethnic demographic26. The rioters in the sample were nearly entirely white men, and their perceptions of a loss of significance17 may be an important motivating factor fueling the growth of far right-wing affiliation in the U.S. and support for political violence.In a similar pattern to what Pennebaker6 found when comparing al Qa’ida to nonviolent Arabic-language speaking groups, violent actors at the Capitol Breach were less analytical relative to their nonviolent counterparts. Theoretically, the mounting anticipation of a major event (in this case on January 6th) produces uncertainty which is incompatible with analytical thinking (i.e., logical-formal reasoning). We expected that the anticipatory distress would be evidenced in anger or negative affect, but the between group comparisons for emotion and negative affective tone did not differ, statistically.LimitationsOur study involved a small sample and may have been underpowered to detect subtle distinctions in emotionality between those who were charged with violent crimes (e.g., assault) and nonviolent crimes (e.g., disorderly conduct) at an event which attracted passionate far-right wing political rioters. Another factor which may limit our conclusions is that we aggregated text across multiple social-media postings between November 2020 and January 6th, 2021. As others have found6, rhetoric can change in the months leading up to a violent event, and we may have missed opportunities to identify those important linguistic shifts. Our conclusions are also limited to the social media data that was available to the criminal prosecutors of those rioters who were arrested for their participation. Prosecutorial discretion may have also influenced whether a rioter was charged and convicted of a violent crime, irrespective of the degree of actual violence that occurred on January 6th. Whether the language of those who were fortunate enough to evade penalty, or posted on encrypted social media networks, would adhere to the patterns in the current findings is unknown.ConclusionWe assessed whether natural language processing can be harnessed to identify harbingers of political violence. The evidence reported herein demonstrates that those charged with violent crimes were more cognitively preoccupied and less analytical relative to their criminally charged nonviolent counterparts. Importantly, violent actors also displayed less evidence of unquestioning group affiliation, which suggests that these individuals may exist as fringe members of the group or are unconcerned with their standing within the group. This was a surprising discovery which supports the quest to gain significance hypothesis. Those who support Trump may be distinct in their motivations and unique in the world of political activists; but, the linguistic characteristics that we have identified can transcend demographics, socioeconomics, and politics. Therefore, they may be generalizable to racial-majority violent political activists beyond the current political climate in the United States, and this remains to be tested in the context of the rise of racial-majority violence elsewhere in the world.The availability of social media data linked to verified overt behavior presented an unprecedented glance into the motivations of violent political activists. We found that the language used by violent offenders sets them apart from their nonviolent peers and reveals that they are less socially integrated. A major strength of our analysis is that there was a suitable nonviolent group with whom to compare linguistic evidence of leakage, and the between group differences were statistically robust. In this report, we have been able to use social media texts as an indicator of violent political action, and this furthers the evidence that natural language analysis can be used to advance public safety by isolating language characteristics associated with approaching acts of violence.
Data availability
LIWC output Data for this manuscript can be located at the Center for Open Science https://osf.io/dvqkb/. Anonymized data which includes all social media posts can be made available upon request.
Code availability
The code for SPSS can be found at https://osf.io/dvqkb/.
ReferencesEggers, A. C., Garro, H. & Grimmer, J. No evidence for systematic voter fraud: A guide to statistical claims about the 2020 election. Proc. Natl. Acad. Sci. 118, e2103619118 (2021).Article
PubMed
PubMed Central
Google Scholar
Li, Q., King, B. G. & Uzzi, B. Quantifying social media predictors of violence during the 6 January US Capitol insurrection using Granger causality. J. R. Soc. Interface 21, 20240314 (2024).Article
PubMed
PubMed Central
Google Scholar
McCauley, C. & Moskalenko, S. Understanding political radicalization: The two-pyramids model. Am. Psychol. 72, 205–216 (2017).Article
PubMed
Google Scholar
Ashokkumar, A. & Pennebaker, J. W. Tracking group identity through natural language within groups. PNAS Nexus 1, pgac022 (2022).Article
PubMed
PubMed Central
Google Scholar
Ellenberg, M. & Kruglanski, A. W. Power of the network and power from the network: Group processes and radicalization. Group Process. Intergroup Relat. 13684302241240704 https://doi.org/10.1177/13684302241240704 (2024).Pennebaker, J. W. Using computer analyses to identify language style and aggressive intent: The secret life of function words. Dyn. Asymmetric Confl. 4, 92–102 (2011).Article
Google Scholar
Van Swol, L., Lee, S. & Hutchins, R. The banality of extremism: The role of group dynamics and communication of norms in polarization on January 6. Group Dyn. Theory Res. Pract. 26, 239–251 (2022).Article
Google Scholar
Meloy, J. R. & O’Toole, M. E. The concept of leakage in threat assessment. Behav. Sci. Law 29, 513–527 (2011).Article
PubMed
Google Scholar
Baele, S. J. Lone-Actor Terrorists’ emotions and cognition: an evaluation beyond stereotypes. Polit. Psychol. 38, 449–468 (2017).Article
Google Scholar
Boyd, R., Ashokkumar, A., Seraj, S. & Pennebaker, J. The Development and Psychometric Properties of LIWC-22. https://doi.org/10.13140/RG.2.2.23890.43205 (2022).Swann, W., Klein, J. & Gómez, A. Comprehensive identity fusion theory (CIFT): New insights and a revised theory. https://doi.org/10.1016/bs.aesp.2024.03.003. (2024).Gómez, Á., Vázquez, A. & Atran, S. Transcultural pathways to the will to fight. Proc. Natl. Acad. Sci. USA. 120, e2303614120 (2023).Article
PubMed
PubMed Central
Google Scholar
Hogg, M. A. Uncertainty–Identity Theory. in Advances in Experimental Social Psychology vol. 39 69–126 (Academic Press, 2007).Hogg, M. A. & Adelman, J. Uncertainty–Identity Theory: Extreme Groups, Radical Behavior, and Authoritarian Leadership. J. Soc. Issues 69, 436–454 (2013).Article
Google Scholar
Baum, M. A., Druckman, J. N., Simonson, M. D., Lin, J. & Perlis, R. H. The political consequences of depression: how conspiracy beliefs, participatory inclinations, and depression affect support for political violence. Am. J. Polit. Sci. 68, 575–594 (2024).Article
Google Scholar
Berkowitz, L. Frustration-aggression hypothesis: examination and reformulation. Psychol. Bull. 106, 59–73 (1989).Article
PubMed
Google Scholar
Kruglanski, A. W. et al. Frustration–aggression hypothesis reconsidered: the role of significance quest. Aggress. Behav. 49, 445–468 (2023).Article
PubMed
Google Scholar
Phadke, S., Samory, M. & Mitra, T. Pathways through conspiracy: the evolution of conspiracy radicalization through engagement in online conspiracy discussions. Proc. Int. AAAI Conf. Web Soc. Media 16, 770–781 (2022).Article
Google Scholar
Kruglanski, A. W. et al. The psychology of radicalization and deradicalization: how significance quest impacts violent extremism. Polit. Psychol. 35, 69–93 (2014).Article
Google Scholar
Gill, P., Horgan, J. & Deckert, P. Bombing alone: tracing the motivations and antecedent behaviors of lone-actor terrorists. J. Forensic Sci. 59, 425–435 (2014).Article
PubMed
Google Scholar
Baele, S. J. Conspiratorial Narratives in Violent Political Actors’ Language. J. Lang. Soc. Psychol. 38, 706–734 (2019).Article
Google Scholar
Jasko, K., LaFree, G. & Kruglanski, A. Quest for Significance and Violent Extremism: The Case of Domestic Radicalization. Polit. Psychol. 38, 815–831 (2017).Article
Google Scholar
District of Columbia | Capitol Breach Cases | United States Department of Justice. https://www.justice.gov/usao-dc/capitol-breach-cases Retrieved on 01.2024 (2021).Kruglanski, A. W., Molinario, E., Ellenberg, M. & Di Cicco, G. Terrorism and conspiracy theories: A view from the 3N model of radicalization. Curr. Opin. Psychol. 47, 101396 (2022).Article
PubMed
Google Scholar
Hogg, M. A., Kruglanski, A. & van den Bos, K. Uncertainty and the Roots of Extremism. J. Soc. Issues 69, 407–418 (2013).Article
Google Scholar
Piazza, J. A. Populism and Support for Political Violence in the United States: Assessing the Role of Grievances, Distrust of Political Institutions, Social Change Threat, and Political Illiberalism. Polit. Res. Q. 77, 152–166 (2024).Article
Google Scholar
Denbeaux, M., Crawley, D. & report, additional. The January 6 Insurrectionists: Who They Are and What They Did. (2023).Download referencesAcknowledgementsThis project was funded by the National Institute of Justice, grant award15PNIJ-21-GG-02726-DOMR. The authors wish to thank Neil Shortland, PhD, who, along with Sofis and Pennebaker, is a co-principal on the grant and helped to conceptualize research objectives and methods of the grant. The National Institute of Justice had no role in study design, data collection and analysis, the preparation of the manuscript, nor the decision to pursue publication.Author informationAuthors and AffiliationsAdvocates for Human Potential Department of Innovation 490-B Boston Post Road Sudbury, Massachusetts, 01976, USAAri P. Kirshenbaum, Gideon Cunningham, Lydia Mudd & Michael J. SofisUniversity of North Carolina at Charlotte Public Policy Program 9201 University City Blvd Charlotte, North Carolina, 28223, USAGideon CunninghamUniversity of Texas at Austin Department of Psychology 08 E. Dean Keeton Street, Mail Stop A8000, Austin, Texas, 78712, USAJames W. PennebakerAuthorsAri P. KirshenbaumView author publicationsYou can also search for this author inPubMed Google ScholarGideon CunninghamView author publicationsYou can also search for this author inPubMed Google ScholarLydia MuddView author publicationsYou can also search for this author inPubMed Google ScholarMichael J. SofisView author publicationsYou can also search for this author inPubMed Google ScholarJames W. PennebakerView author publicationsYou can also search for this author inPubMed Google ScholarContributionsKirshenbaum: primary writing, conceptualization, data analysis and presentation. Cunningham: writing, editing and conceptualization. Mudd: data aggregation and curation. Sofis: conceptualization and editing. Pennebaker: writing, editing, conceptualization, and data analysis.Corresponding authorCorrespondence to
Ari P. Kirshenbaum.Ethics declarations
Competing interests
The authors declare no conflicts of interest.
Peer review
Peer review information
Communications Psychology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editors: Jennifer Bellingtier and Marike Schiffer. A peer review file is available.
Additional informationPublisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Supplementary informationTransparent Peer Review fileSupplementary Materialsreporting-summaryRights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Reprints and permissionsAbout this articleCite this articleKirshenbaum, A.P., Cunningham, G., Mudd, L. et al. Online language of violent rioters displayed weak group affiliation preceding the U.S. Capitol Breach.
Commun Psychol 3, 46 (2025). https://doi.org/10.1038/s44271-025-00232-7Download citationReceived: 25 September 2024Accepted: 14 March 2025Published: 23 March 2025DOI: https://doi.org/10.1038/s44271-025-00232-7Share this articleAnyone you share the following link with will be able to read this content:Get shareable linkSorry, a shareable link is not currently available for this article.Copy to clipboard
Provided by the Springer Nature SharedIt content-sharing initiative