Participants were recruited through academic newsletters (e.g., faculties of psychology, general studies), social media, and SurveyCircle. Climate activists from Germany and Austria were recruited via e-mail by various activist groups (e.g., Last Generation, Students for Future, Fridays for Future, Parents for Future, Artists for Future, and Scientists for Future). The inclusion criterion was, Age ≥ 16 years. We excluded datasets with missing values (listwise deletion) and individuals who stated that they did not believe in man-made climate change. Participants who completed the experiment could take part in a voucher raffle (20 x 10 EUR or a donation to a non-profit environmental organization), and psychology students of the University of Mainz received extra course credit for participation.

From the 461 individuals who started the survey, 178 participants (38.6%) were excluded due to missing data, and nine (1.9%) for not believing in man-made climate change. A flowchart per group can be found in Supplement A (see Hennemann et al., 2026). Finally, 274 participants were included in the analyses (75 activists and 199 non-activists). In the total sample, the average age was 33.2 years (SD = 14.8, range 16–73), 210 individuals (76.6%) indicated female gender, 63 (23%) were parents, and the majority were highly educated (at least A-level graduation: 252, 92%). Two-thirds of participants (n = 181, 66%) rated themselves as “Strongly” or “Very strongly” affected by climate change (5-point scale from 1 “Not at all” to 5 “Very strongly”). Activists were significantly older (d = 0.65 [0.37, 0.93]), more strongly affected by climate change (76% vs. 62% responding “Very strongly” or “Strongly”), more often parents (54% vs. 46%), reported less intolerance of uncertainty (d = -0.41 [-0.68, -0.14]), higher negative climate affectivity (d = 0.67 [0.39, 0.94]) and more pro-environmental behavior (d = 1.70 [1.38, 2.01]) than non-activists, see Table 1.

The study was conducted online, and the data collection was carried out anonymously via soscisurvey.de (Leiner, 2019) from July 27 to September 8, 2023. The study was preregistered (see Schröder, 2023). Other than pre-registered, the psychometric analyses of the self-constructed Climate Emotions Questionnaire (CEQ) are not reported in full detail here (see Supplement C, Hennemann et al., 2026, for more details). Furthermore, we formulated Hypothesis (a) regarding differences between the two ER strategies as a specific hypothesis rather than exploratively (corresponding to additional hypotheses under “secondary analyses” in the pre-registration), based on the latest evidence (Wenzel et al., 2023), and extended the explorative analyses (c) on the association of ER effectiveness and frequency with PEB.

Participants were first informed about the background and procedure of the study and provided informed consent. Next, participants completed a set of questionnaires regarding demographics, climate emotions, emotion regulation, intolerance of uncertainty, PEB, and current mental distress. Afterward, participants completed the ER task, which followed a 3 (trial condition: reappraisal vs. suppression vs. control) x 2 (group: activists vs. non-activists) x 2 (time: pre- vs. post-regulation) mixed factorial design. Trial condition and time were within-subject factors, group was a between-subject factor, and affective valence and arousal served as dependent variables. First, participants were introduced to the two ER strategies, reappraisal and suppression, and the control condition via 3-minute information videos (text version see Supplement B, Hennemann et al., 2026). This was followed by a practice trial for each of the ER strategies and the control condition. After ensuring an understanding of the ER task and strategies, participants completed the main ER task with 24 trials (eight trials per condition). The trial procedure is shown in Figure 1. Emotion-eliciting images were presented in random order for 4 seconds, followed by a first rating of affective valence and arousal. Then, also in random order, one of the three trial conditions was prompted in the center of the screen (“reappraisal”, “suppression”, or “view”) for 3 seconds and the same image was presented for 7 seconds, during which the respective strategy should be applied. Subsequently, affective valence and arousal were rated again, as well as the effort of implementing the respective strategy. The sequence, stimulus presentation times, and rating procedures were designed in accordance with established ER paradigms and have demonstrated validity and reliability in prior studies (Ochsner et al., 2002; Schnabel et al., 2022; Waugh et al., 2022).

The ER video instructions were adapted from Schnabel et al. (2022). Each video contained general (i.e., not climate-related) information about the respective strategy and provided examples, see Supplement B, Hennemann et al. (2026). As stimulus material, 24 negative climate images from the Affective Climate Images Database (Lehman et al., 2023) were selected, with the highest relevance and negative valence ratings within the database, so that the images showed as much variance as possible in terms of their content (e.g., ice floes, floods, animal suffering, etc.). The relevance to climate change of the selected images varied between 6.5 and 8.2 (M = 7.3, Scale 1–9), their valence from 1.4 to 3.3 (M = 2.6), and the arousal between 4.9 and 5.6 (M = 5.2) on a scale from 1 to 9, with low values indicating negative valence and low arousal. The list of images selected for this study can be found in Supplement C, Hennemann et al. (2026). The images were edited using GIMP, Version 2.10.34 (Natterer, 2023) for presentation in SoSci Survey and scaled to a width (or height for vertical images) of 600 pixels.

Valence and arousal were assessed with a 9-point self-assessment manikin scale (Bradley & Lang, 1994), ranging from 1 (“Unpleasant”) to 9 (“Pleasant”) for valence and 1 (“Excited”) to 9 (“Calm”) for arousal. For better understanding, both scales were recoded (1–“Pleasant to 9–“Unpleasant” and 1–“Calm” to 9–“Excited”) so that higher pre-post-regulation difference scores indicated more effective emotion regulation. The effort of ER implementation (“How difficult was it for you to implement the strategy?”) was rated on a 5-point scale from 1 (“Very difficult”) to 5 (“Very easy”) and was inversely recoded afterwards for better understanding (higher values indicating more effort).

Negative climate emotions (fear, sadness, anger, guilt, hopelessness) were assessed with a self-constructed questionnaire including 25 items that were answered on a 5-point scale from 1 (“Does not apply”) to 5 (“Does apply”). The questionnaire was intended as a “short version” based on a set of established questionnaires and findings suggested a 3-factorial structure with the factors ‘negative climate affectivity’ (McDonald’s ω = .93), ‘climate guilt’ (ω = .90) and ‘climate hopelessness’ (ω = .54), total scale ω = .92 (see Supplement D, Hennemann et al., 2026, for all items). Furthermore, we adapted the items of the Emotion Regulation Questionnaire (ERQ; Gross & John, 2003) to assess the preference for emotion regulation (reappraisal, suppression) in the context of the climate crisis. The ERQEco contained 8 items (see Supplement E, Hennemann et al., 2026), which are answered on a 7-point Likert scale (from 0–“Not true at all” to 6–“Completely true”). Internal reliability of the total scale was ω = .75. Individual and collective PEB over the past 12 months was assessed with 10 items based on the PEB measure by Stanley et al. (2021) with a 5-point answer scale (1–“Strongly disagree” to 5–“Strongly agree”), see Supplement F, Hennemann et al. (2026), for items. Internal reliability of the total scale was ω = .84. Intolerance of uncertainty was assessed with 12 items of the Intolerance of Uncertainty Scale short form (IUS-12; Carleton et al., 2007), which are rated on a 5-point answer scale ranging from 1 (Not at all characteristic of me) to 5 (Entirely characteristic of me). Internal reliability was ω = .89. Mental distress (i.e., depression, anxiety) in the past two weeks was assessed using the Patient Health Questionnaire 4 (PHQ-4; Kroenke et al., 2009) with two items each, which are rated on a 4-point scale from 0 (“Not at all”) to 3 (“Nearly every day”). Internal reliability was ω = .82.

For sample size estimation, a power analysis with G*Power, Version 3.1 (Faul et al., 2007) for a repeated measure (rm)ANOVA (time x trial condition x group) and two dependent variables (affective valence, arousal) yielded a necessary sample size of N = 152 to detect a small to moderate effect (η²p = 0.05) with a power of .80 and a .05 level of significance. To test group differences in demographics and self-report measures, we conducted independent t-tests for continuous and χ² tests for categorical data. To analyze ER effectiveness, repeated-measures analyses of variance (rmANOVAs) were conducted separately for valence and arousal, testing the effects of time (pre- vs. post-regulation), trial condition (reappraisal vs. suppression vs. control), and group (activists vs. non-activists). To investigate whether reappraisal would be more effective at decreasing negative affect and arousal than suppression and the control condition (Hypothesis A), we analyzed the main and interaction effects of time and trial condition and report effect sizes as partial η². As an indicator of potential ER differences between activists and non-activists (Hypothesis B), we explored time x trial condition x group interactions. For more in-depth analyses, we calculated difference scores for valence and arousal (pre minus post-regulation) as indicators of ER effectiveness, with higher scores indicating a stronger decrease in negative valence and arousal. With these ER effectiveness scores, we conducted independent t-tests to compare both groups for each ER strategy using a conservative level of significance of .005 to account for multiple testing (Benjamin et al., 2018). Differences in effort ratings were tested using dependent (within-trial conditions) and independent (between groups) t-tests. For the effort ratings, individual missing values occurred (n = 14 participants) and were addressed using item-wise mean imputation. We tested the relationship between ER effectiveness and PEB (Hypothesis C) by conducting Spearman correlations (corr_ρ) between the difference scores for valence and arousal ratings (pre minus post-regulation) and self-reported PEB. We also exploratively analyzed the correlation between reappraisal and suppression frequency in the ERQEco and PEB using the same correlation procedure. Data processing and statistical analyses were conducted with R, Version 4.2.1 (2022-06-23 ucrt) (R Core Team, 2023) and JASP, Version 0.17.3 (JASP Team, 2023) with an alpha of 0.05 (two-sided testing). All materials, data, and code are openly available through the Open Science Framework (see Hennemann et al., 2025).

Mixed ANOVAs revealed significant main effects of time (pre- vs. post-regulation) for valence (F(1,272) = 182.51, p < .001, η² = 0.40) and arousal (F(1,272) = 115.74, p < .001, η² = 0.30). Accordingly, participants reported a decrease in negative affect from before to after the regulation phase (see Table 2).

The analysis furthermore revealed a significant main effect of ER strategy on valence, F(2,544) = 5.52, p = .004, η² = 0.02, but not on arousal, F(2,544) = 2.13, p = .119, η² = 0.008, and a significant main effect of group on valence, F(1, 272) = 4.92, p = .027, η² = 0.02, but not on arousal, F(1, 272) = 1.49, p = .223, η² = 0.005. Average pre-regulation values for valence (ps ≥ .638) and arousal (ps ≥ .327) did not differ significantly between trial conditions. Activists showed higher pre-regulation valence values, t(272) = 2.32, p = .021, but not arousal values, t(272) = 1.38, p = .169, than non-activists.

The time x strategy interaction effects were significant for valence, F(2,544) = 32.68, p < .001, η² = 0.11, and arousal, F(2,544) = 18.33, p < .001, η² = 0.06. In line with Hypothesis (a), within-subjects post-hoc tests of pre-post-regulation difference values indicated that reappraisal was significantly more effective than suppression for both valence, t(273) = 7.42, p < .001, d = 0.45, and arousal, t(273) = 5.84, p < .001, d = 0.35. Participants were also significantly more effective in the reappraisal trials compared to the control trials at decreasing negative valence, t(273) = 5.48, p < .001, d = 0.33, and arousal, t(273) = 2.55, p = .011, d = 0.15. Suppression was significantly less effective compared to the control trials for both valence, t(273) = -3.05, p = .003, d = -0.18, and arousal, t(273) = -4.01, p < .001, d = -0.24; see also Figure 2.

Regarding group differences in the effectiveness of emotion regulation (b), the time x strategy x group interaction effects were significant for valence, F(2,544) = 4.57, p = .013, η² = 0.02, but not for arousal ratings, F(2,544) = 1.65, p = .194, η² = 0.006. Between-group posthoc tests of pre-post-regulation difference values revealed non-significant group differences for reappraisal and suppression for either valence or arousal (ps ≥ .105). However, activists, compared to non-activists, reported significantly higher decreases in negative affect, t(272) = 2.28, p = .023, d = 0.31, and marginally non-significant decreases in arousal, t(272) = 1.95, p = .053, d = 0.26, in the view trials; see also Figure 2.

Participants rated the ER strategies on average as moderately effortful (M = 2.69 to 2.99 on a 1 – 5 rating scale, see Supplement G, Hennemann et al., 2026). Reappraisal was rated significantly more effortful than view only, but not than suppression. The control condition was rated as the least effortful strategy. Activists rated reappraisal on average significantly more effortful to implement than non-activists (p = .033, d = 0.30), with no significant group differences for suppression or the control condition, see Table 3.

To test associations of ER effectiveness and PEB (see Research Question C), we investigated correlations between pre- vs. post-regulation difference scores for valence and arousal in the ER task and PEB total scores. Neither reappraisal nor suppression effectiveness scores for valence or arousal were significantly associated with PEB. Only the view-only strategy for decreasing arousal was significantly positively associated with PEB (corr_ρ = .22, p < .001), see also Table 4. Stratified by group, this correlation became significant in activists (corr_ρ = .23, p < .001) but not in non-activists (corr_ρ = .18, p = .118).

Exploratively, we analyzed correlations between self-reported habitual ER, assessed by the ERQEco, with PEB. Habitual reappraisal did not correlate significantly with PEB (corr_ρ = –.004, p = .539), and habitual expression suppression correlated negatively with PEB (corr_ρ = –.27, p < .001).

Despite the strengths of this study, several limitations should be acknowledged that may influence the interpretation of the findings. Firstly, although the sample comprises some demographic diversity and includes different levels of activism, it remains representative rather for a mostly young, female, well-educated, central European (“WEIRD”) population (Henrich et al., 2010), limiting the generalizability, especially to individuals in other countries with higher imminent exposure to environmental threats. As research points towards gender differences in emotion regulation for example that women tend to use a greater variety of emotion regulation strategies and demonstrate more flexibility in their implementation (Goubet & Chrysikou, 2019; McRae et al., 2008; Nolen-Hoeksema, 2012) and considering the high percentage of female participants in our sample, future research should further investigate how different gender identities influence the effectiveness and preference of emotion regulation strategies in the context of emotions in response to climate change threat. However, most participants in our sample indicated to be at least strongly affected by climate change. Future studies could investigate ER processes in populations that are exposed to different levels of environmental threat. Secondly, compared to a more controlled laboratory setting, conducting the ER task online may have compromised internal validity, primarily due to the inability to ensure participants' full attention and adherence. However, ensuring instruction adherence is a general challenge in ER tasks, regardless of the setting. For example, in an ER task by Opitz et al. (2015), at least half of the participants did not adhere to the instructed reappraisal strategy. Another limitation is that automatic and implicit regulation processes can interfere with instructed ER strategies. To mitigate this, we implemented a brief initial stimulus presentation (4 seconds). However, automatic regulation processes could not be fully controlled throughout the trial. In this context, previous research has highlighted the diverging effects of instructed versus habitual emotion regulation, for example, through studies employing regulatory choice paradigms (Mauersberger et al., 2018; Schnabel et al., 2022). Thirdly, the use of non-specific instructions for ER strategies concerning climate change could have led to heterogeneity in implementation tactics and should be adapted more closely to the subject. Additionally, the specificity of our findings to “climate emotions” remains uncertain, as no comparator stimuli were included (e.g., neutral images or negative images with non-climate-related content). Along those lines, we confined the experiment to two prominent ER strategies, i.e., reappraisal and (expression) suppression. Future research should include other frequently used ER strategies, e.g., distraction or acceptance (Lennarz et al., 2019), and use regulatory choice trials to provide a more holistic understanding of ER preferences and flexibility (Aldao et al., 2015). For example, problem-solving tactics and acceptance were used most frequently to reappraise everyday life stressors in an ecological momentary assessment study in college students (Marciniak et al., 2024). Fourthly, the number of trials was kept rather low to increase retention within the online experimental design and could be increased for more robust results. Finally, we only included a retrospective self-report measure of PEB in our study, which suffers from limited validity (for a systematic review, see Lange, 2023). Future experimental studies should consequently implement consequential behavioral paradigms to assess PEB, such as the Work for Environmental Protection Task (Lange & Dewitte, 2022).

In this experimental study, we corroborated cognitive reappraisal as an effective strategy to decrease negative emotions when confronted with climate change-related imagery. Additionally, both activists and non-activists were able to effectively implement reappraisal, although activists experienced more effort. As previous research identified potential drawbacks of reappraisal use on political action, we suggest a closer investigation of reappraisal tactics in future studies. Similar to previous research, expressive suppression was identified as an ineffective strategy for the reduction of negative emotions in response to climate change threat. As the frequent use of expressive suppression was additionally associated with less self-reported PEB, individuals and communities may rather profit from expressing emotions related to climate change threat. In sum, a constructive approach to climate change requires ER strategies that reconcile affective well-being with the motivation to act pro-environmentally. Although reappraisal can be effortful, it may achieve both goals.