MainRecent years have seen a global increase in the criminalization of peaceful protest1, leading the United Nations to call on governments to protect citizens’ rights to dissent2. Repression, defined as measures that ‘prevent, control, or constrain non-institutional collective action’3, is particularly notable in the case of climate activism4,5,6. Faced with insufficient government action to avert climate breakdown7,8,9,10, climate activists have escalated their use of disruptive protest tactics11,12,13 (for example, roadblocks). In response, the authorities have intensified restrictions on protest by selectively enforcing counterterrorism laws13 and through new legislation, such as the 2023 Public Order Act in England and Wales13,14. Consequently, climate activists in countries with historically open civil societies now face increased personal risks while protesting, including surveillance, arrest, fines and lengthy prison sentences13. We investigate the drivers of participation in collective climate action, despite the risk of such punitive consequences, among climate activists involved in the protest group Extinction Rebellion.Collective action involves individuals acting collectively to achieve common interests15,16. Conflicting evidence exists on whether and how repression (dis)inhibits collective action intentions1. Some studies find that repression deters engagement17,18, while others report evidence consistent with a galvanizing effect19,20,21. There has, moreover, been little differentiation between the psychological effects of anticipated repression, that is, the belief that activists may encounter repression, and actual lived experiences of repression in the form of arrests, fines, surveillance or jail time. In situ research on crowd dynamics indicates that harsh treatment of protesters by the police can intensify protest behaviour through the delegitimization of the authorities and collective empowerment of the crowd22. Research examining individuals’ action motivations has similarly found that direct experiences of police violence increased protest intentions in general23, willingness to make sacrifices for the movement23 and intentions to engage in disruptive collective actions specifically21. The latter finding was explained by reduced fear about protesting21. The potential of experienced repression to legitimize disruptive action is documented anecdotally in the climate activist context12,13, where the recent wave of protest criminalization has spurred calls to ‘double down’ on disruptive tactics and to the emergence of groups engaging in covert actions to directly disrupt oil and gas companies12,24. Similarly, recent qualitative evidence suggests that perceived police repression is used by activists to justify participation in law-breaking25,26.The social identity model of collective action (SIMCA)27 integrates three key psychological drivers of engagement. These are subjective grievances that give rise to a sense of injustice27,28,29, identification with an aggrieved or politicized group (that is, a group that is committed to effect social change)30,31,32,33,34 and a sense of efficacy27. Recent contributions have extended the efficacy concept to encompass a broader range of considerations—distinguishing political efficacy (the belief that one’s group can achieve its political aims)20 from identity consolidation efficacy35 the belief that protest strengthens the movement) and participative efficacy36 (the belief that one’s own participation makes a meaningful contribution to the group)35,36.While SIMCA aims to account for collective action intentions generally, two main forms of collective action have since been distinguished. Normative collective actions15,37 are conceptualized as actions that lie within the societal framework (for example, ratified protest marches or petition signing). Non-normative actions21,37,38,39 deliberately break the rules of the established system and violate societal norms (for example, roadblocks, public building occupations or superglue lock-ons)33. The psychological drivers of non-normative forms of collective action are much less researched27, particularly for climate activism26,33,40,41,42 (for exceptions, see refs. 25,43,44). As a result, our understanding remains limited on whether existing models of collective action apply to more disruptive, non-normative acts and how government repression might escalate different types of protest actions.Research also highlights the importance of justice-related emotions such as anger or outrage to enhance action intentions37,39,45 and fear, as an action-inhibiting emotion40,46, to reduce action intentions19,21,47,48. Although both normative and non-normative actions frequently co-occur and overlap in their psychological drivers49, evidence indicates that non-normative actions are more strongly predicted by feelings of contempt than anger37,50. Like anger, contempt is considered an ‘other-condemning’ emotion, but has distinct characteristics37,50 and promotes disengagement from its object50,51. When felt towards political elites, contempt can reduce the need to adhere to social norms and enable actions that challenge the established order, including non-lawful and even violent action (note that we do not include violent non-normative action tendencies in our analysis, as these are less relevant in the current climate activist context; see also refs. 37,38,52).Non-normative actions are more likely to occur as the perceived efficacy of conventional action drops37,53,54. The so-called nothing-to-lose hypothesis states that a sense of low political efficacy and despair about the prospect for change incites riskier, non-normative action (see refs. 37,53 for evidence, but also see ref. 21 for inconsistent findings). Limited research exists, however, on how participative efficacy and identity consolidation efficacy are linked to future non-normative collective action intentions, and whether low levels of these forms of efficacy might also create the conditions that promote non-normative action intentions.We investigate psychological predictors of intentions to engage in normative and non-normative collective climate action using data from a recent survey of Extinction Rebellion activists—a rarely accessed sample from a prominent climate protest movement in the UK. We invited 160,103 Extinction Rebellion UK mailing list subscribers to participate in a 10-minute survey on climate activism (Methods). In total, 1,662 subscribers entered the survey between 19 December 2024 and 25 January 2025. The final sample size (after exclusion checks) was 1,375 (Fig. 1).Fig. 1: Demographic overview of the sample.Detailed information on the sample (n = 1,375 participants) is presented in Supplementary Information. Map created with Flourish (https://flourish.studio). Credit: icons, UXWing (https://uxwing.com).Full size imagePredictors of normative and non-normative climate action intentionsWe used hierarchical multiple regression to examine the incremental predictive value of theoretically distinct groups of variables on collective climate action intentions. Each block was entered on the basis of its conceptual precedence, beginning with past event participation as a predictor of future intentions, and concluding with emotions as the most proximal drivers of action intentions. Model 1 regressed action intentions on past participation. In model 2, we entered the extended SIMCA27 variables (environmental concern, environmental movement identification, political efficacy, identity consolidation efficacy and participative efficacy). Model 3 added anticipated and experienced repression. Finally, in model 4, we added emotions in response to repression, specifically anger/outrage, contempt and fear (see Table 1 for hierarchical regression results; for an overview of all standardized coefficients in the final model, see Fig. 2).Table 1 Hierarchical multiple regression predicting normative and non-normative action intentionsFull size tableFig. 2: Predictors of normative and non-normative collective climate action intentions.Forest plot showing standardized regression coefficients (circles) and 95% CIs (error bars) from model 4 of a hierarchical multiple regression predicting normative (blue) and non-normative (yellow) collective climate action intentions. Model 4 includes all predictors entered simultaneously. Coefficients represent standardized effects. Sample sizes were n = 1,301 for normative action tendencies and n = 1,300 for non-normative action tendencies.Full size imageAlthough there were no significant direct predictive effects for anticipated repression over and above past participation and SIMCA variables in model 3, experienced repression positively predicted non-normative action intentions (B = 0.33, s.e. = 0.03, P