Impact of anger, compassion, and hope on climate action taking Authors: Dominika Zaremba, Michalina Marczak, Jarosław M Michałowski, Christian Andreas Klöckner, Małgorzata Wierzba, Artur Marchewka Published: 2026 ABSTRACT AbstractClimate change is a serious threat to people and ecosystems worldwide, making it increasingly important to understand what drives pro-environmental behavior (PEB). Emotions are recognized as key drivers of PEB, with underlying appraisals that seem to shape behavioral outcomes. We conducted a preregistered study on the influence of anger, compassion, and hope on PEB. Participants (N = 1029 in Poland, N = 1010 in Norway) read emotional stories designed to elicit a specific set of climate-related appraisals or neutral stories. Subsequently, they completed two tasks designed to measure different types of PEB: an effort task (Work for Environmental Protection Task; WEPT) and a monetary donation task. We hypothesized that reading emotional (as compared to neutral) stories would increase PEB and that distinct appraisal patterns would promote different behavioral intentions. We also predicted that demographic and personal factors would moderate the relationship between emotions and PEB. Contrary to expectations, results showed no effects of emotion manipulation on either measure. Older age, however, consistently predicted higher effort and donations. We discuss potential areas for further research, including the use of validated naturalistic and immersive stimuli, prolonged exposure to interventions, momentary assessment of behaviors, and matching behavioral tasks to emotional manipulation.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-56796-9.Subject terms: Environmental social sciences, Psychology, Psychology Introduction Despite the consensus on the need for systemic solutions to address climate change1,2, changes in individual lifestyles are also essential. In high-income countries, household consumption accounts for approximately 60% of greenhouse gas emissions3,4, so it is important to understand the drivers of individual pro-environmental behavior (PEB). For quite some time, models of PEB focused on cognitive factors, such as attitudes, social norms, values, judgments, incentives, and situational factors. The most prominent examples of such models are the Theory of Planned Behavior5,6, the Norm-Activation Model7,8, the Value-Belief-Norm Model9,10, or the Comprehensive Action Determination Model11. In comparison, the role of emotions in motivating PEB has been relatively underexplored. Recent evidence suggests that emotions associated with climate change (hereafter: climate emotions) can significantly influence judgments and behaviors concerning this issue12–14. Most research on climate emotions has focused on fear or anxiety, emotions linked to the experience of risk and uncertainty15–18. This research has yielded mixed findings, which some authors attribute to psychological reactance when threat-focused messaging is not paired with efficacy- or solution-oriented information19. Such communication may be most effective when it hits a “Goldilocks” balance, evoking enough anxiety to motivate engagement without becoming overwhelming20. On the other hand, some scholars argue that the empirical evidence for such backfire effects remains limited21. Other emotions, such as anger, hope, and compassion, have received relatively less attention. For example, anger often emerges as moral outrage toward those responsible for climate harms and can motivate punitive or corrective action22,23. Hope, particularly “constructive hope,” can inspire confidence in collective solutions and future-oriented engagement24–26. Compassion, rooted in empathy for victims of climate change, fosters a desire to alleviate suffering and reduce harm27,28. Emotions as drivers of climate action The Lazarus’ Cognitive-Motivational-Relational Theory of Emotion29 can be useful in explaining how appraisals shape distinct climate emotions. According to this framework, emotions arise from primary and secondary appraisals. Primary appraisal assesses whether an event holds goal relevance and goal congruence, corresponding to emotional arousal and valence, respectively30. Additionally, Lazarus29 argues that distinct emotions arise from differences in goal content (e.g., punitive goals in anger, affiliative goals in compassion). Furthermore, Lazarus’ theory distinguishes secondary appraisals29, such as attribution of responsibility (blame or credit), coping potential (perceived ability to influence the situation), and future expectations (anticipated outcomes), all of which seem to be relevant in the context of climate change. Indeed, these factors were previously discussed as critical in shaping climate engagement35,40. Appraisal theories provide a broader integrative framework for linking emotions with the cognitive determinants of PEB. Based on these theories29,31,32, one can postulate that climate emotions emerge from evaluations of climate-related events (primary appraisals) and the evaluations of one’s own capacity to respond to these events (secondary appraisals). These appraisal dimensions closely overlap with constructs that are central in prominent environmental-behavior models, such as perceived threat and efficacy beliefs, responsibility and norms, and objective costs and constraints6,8,11,33. Framed in this way, “cognitive predictors of PEB” are not separate from emotions, but instead they are components in the process of their elicitation. Appealing to emotions may improve engagement and persuasive outcomes of climate communication, although effects can vary by message features and audience context12,34,35. In terms of Lazarus’ theory, such communication should evoke both goal-relevant and goal-congruent appraisals36. However, to understand how distinct climate emotions influence PEB, it is also essential to consider the goal content. Different emotions align with various adaptive responses to climate change: holding perpetrators accountable, addressing climate-induced suffering, and developing long-term solutions. These responses can be encouraged by framing climate change using relatable narratives that emphasize proximity to personally valued aspects of life, such as safety, justice, and well-being37–39. Research has identified specific emotions that effectively drive PEB, including anger22,41, compassion42,43, and hope25,44,45, and secondary appraisals are key to understanding their effect on action taking. Table 1 outlines the particular appraisal structure of these emotions. Table 1General appraisal patterns used as a reference for stimulus construction, together with the expected action tendencies.EmotionEmotion objectPrimary AppraisalSecondary AppraisalGoals threatened by climate change: physical, social, and economic safety, health and well-being, procreation, identity, living in a just world, and good self-imageAngerPerpetrators of the climate crisis (e.g., polluters, policymakers) who cause the harm.Goal relevance: high (Perpetrators worsen climate change and endanger my safety.)Goal congruence: low(Their actions directly obstruct my goal of living in a safe world.)Goal content or motivational tendency: punitive action, protest, holding perpetrators accountableAttribution: blame (Their behavior is environmentally harmful; they are responsible for the crisis.)Coping potential: high (We can enforce rules and hold perpetrators accountable.)Future expectations: positive (The justice will be restored.)CompassionInnocent victims of climate change (e.g., elderly, children, animals) who suffer the consequences.Goal relevance: high (The suffering of the victims of climate change shows the world is unjust and unsafe.)Goal congruence: low(Harm to others contradicts my moral values and affiliative goals.)Goal content or motivational tendency: caring, helping, advocacy for vulnerable othersAttribution: lack of blame (The victims are innocent and do not deserve their suffering.)Coping potential: high(Despite personal costs, we hold a moral responsibility to help.)Future expectations: positive (The suffering will be reduced.)HopeA safe future can be secured through collective, coordinated, proactive efforts.Goal relevance: high (The problem is real, but effective solutions exist.)Goal congruence: high (These solutions help me reach my goal of living in a safe world.)Goal content or motivational tendency: constructive problem-solving, grassroots organizingAttribution: credit (People acting for the climate deserve appreciation.)Coping potential: high (People can innovate and organize successfully.)Future expectations: positive (The world will be sustainable.) General appraisal patterns used as a reference for stimulus construction, together with the expected action tendencies. Goal relevance: high (Perpetrators worsen climate change and endanger my safety.) Goal congruence: low (Their actions directly obstruct my goal of living in a safe world.) Goal content or motivational tendency: punitive action, protest, holding perpetrators accountable Attribution: blame (Their behavior is environmentally harmful; they are responsible for the crisis.) Coping potential: high (We can enforce rules and hold perpetrators accountable.) Future expectations: positive (The justice will be restored.) Goal relevance: high (The suffering of the victims of climate change shows the world is unjust and unsafe.) Goal congruence: low (Harm to others contradicts my moral values and affiliative goals.) Goal content or motivational tendency: caring, helping, advocacy for vulnerable others Attribution: lack of blame (The victims are innocent and do not deserve their suffering.) Coping potential: high (Despite personal costs, we hold a moral responsibility to help.) Future expectations: positive (The suffering will be reduced.) Goal relevance: high (The problem is real, but effective solutions exist.) Goal congruence: high (These solutions help me reach my goal of living in a safe world.) Goal content or motivational tendency: constructive problem-solving, grassroots organizing Attribution: credit (People acting for the climate deserve appreciation.) Coping potential: high (People can innovate and organize successfully.) Future expectations: positive (The world will be sustainable.) Anger Climate anger manifests as moral outrage toward those who knowingly contribute to climate change46–48. It stems from perceived injustice and is often directed at individuals or institutions responsible for the harm, with prosocial outcomes more likely when the harm affects others49. In the context of climate injustice, anger may arise when people recognize that those least responsible for emissions suffer the greatest consequences23. In general, this emotion motivates confrontational behaviors - such as activism, protests, and advocacy - to hold perpetrators accountable40,41,50. However, when anger lacks a specific target and is directed at abstract entities (e.g., humanity in general), it tends to lead to inaction and fatalistic beliefs about human selfishness or the impossibility of change48,51. On the contrary, anger directed toward specific entities and institutions has been shown to increase public support for climate mitigation52. Compassion Compassion, focused on the innocent victims of climate change (e.g., the elderly, children, animals), stems from the perception of their suffering. It reflects goal incongruence, particularly the violation of affiliative or moral goals, and leads to a tendency to approach, help, and support23,27. The effects of compassion can be contradictory to those of anger - compassion refocuses the attention and energy to helping the victims, and the tendency to punish the perpetrators decreases53. Some of the mobilizing effects of compassion can be explained by a reduced psychological distance to climate change and an understanding of its impact on living beings54,55. Compassion is associated with greater support for climate mitigation efforts42, self-reported donations to environmental organizations43, and engagement in climate change activism28. However, compassion does not always lead to action. Repeated exposure to suffering can cause compassion fatigue, psychic numbing, or burnout56. When victims remain anonymous, or their numbers increase, compassion tends to “fade,” reducing mobilization57,58. Hope Hope relates to the future of the climate crisis - it arises from appraisals of goal congruence, possibility, uncertainty of a favorable outcome, and positive future expectations59. It can be elicited by emphasizing attainable goals and feasible pathways for action24,60. Research on hope related to climate action provided mixed results, with researchers distinguishing between “false hope” or denial-based hope, which involves misguided beliefs about climate change, and “constructive hope,” which is based on the belief that collective, coordinated efforts can still halt climate change25,44. When hope is based on vague optimism rather than grounded in realistic pathways for change, it may foster complacency rather than engagement61–63. Constructive hope has been positively linked to self-reported PEB, support for specific policies, and political engagement24,34, whereas false hope was found to lead to inaction64. Moreover, if hope is not paired with a sense of coping efficacy and concrete behavioral options, it may feel naïve or unrealistic, especially in the face of escalating climate threats24. Research suggests that effective hope appeals must balance optimism with realism and emphasize efficacy to maintain engagement65,66. Measuring pro-environmental behavior Historically, many studies have depended on PEB measures that, while cost-effective and convenient, often lack validity. For example, approximately 80% of studies on environmental behaviors published in the Journal of Environmental Psychology in 2015 and 2016 rely solely on self-reports of behavior, intentions, and attitudes67. However, declarations and decisions made in hypothetical scenarios are typically poor models of real-life situations68–70. For instance, real donations tend to be 27% lower than hypothetical ones71. Therefore, it is crucial to use measures of PEB that ensure more ecological validity than self-reports or hypothetical, inconsequential scenarios72,73. Charitable donation is an example of a behavior with real-world consequences and is easy to use in experimental settings. Previous studies found that emotions play a crucial role in motivating donations to climate advocacy groups74 and are more effective than social norms alone75. However, for some demographics (e.g., economically privileged people), the decision to donate may not be substantially effortful. This is why it is also important to measure people’s willingness to sacrifice other valuable resources, such as their time or cognitive effort. Such effort can be measured with paradigms such as the Pro-environmental Behavior Task67 or the Work for Environmental Protection Task76. Moreover, the goals associated with these tasks can be linked to specific emotions. According to appraisal theories, anger may lead to a preference for actions that punish corporations and policymakers driving climate change, compassion may motivate efforts to help those affected by climate change, and hope may foster investments in mitigation solutions. Current study In the current study, we aimed to investigate the causal relationship between distinct climate emotions and PEB. Building on Lazarus’29 Cognitive-Motivational-Relational Theory of Emotion, we argue that cognitive appraisals play a key role in shaping specific climate actions. While arousal, as an essential aspect of emotional response, may account for overall motivation to act, the direction of effort likely depends on a more nuanced interpretation of the situation that triggered the emotions. We hypothesized that exposure to stimuli designed to evoke specific climate-change-related appraisals and emotions would increase PEB (H1). We investigated whether engagement in PEB would differ across emotion conditions, that is, anger, compassion, or hope (H2). These emotions were selected as an initial focus because prior research has linked each of them to approach-oriented tendencies relevant to climate action23,25,43,77,78. We also hypothesized that experiencing different emotions would increase distinct action tendencies (H3). In particular, anger was expected to relate to the tendency to punish perpetrators, compassion was expected to relate to a tendency to help victims, and hope was expected to relate to supporting the development of promising solutions. Finally, we explored whether relationships between emotions and behavior would vary as a function of several commonly studied predictors of PEB: level of valence and arousal79, belief in climate change, concern about climate change13, perceived climate action efficacy80, psychological distance80, willingness to sacrifice for the environment81, and demographic characteristics, such as age82 and socio-economic status83 (H4). The moderator variables were selected because they have been widely examined as predictors of PEB in previous research84–86, but also because of how they map conceptually onto appraisal-related dimensions. Specifically, belief in climate change, climate concern, and psychological distance index primary appraisals related to relevance and urgency, whereas perceived personal and collective efficacy and willingness to sacrifice map onto secondary appraisals related to coping potential and feasibility of action. Demographic and socio-economic indicators were included as markers of resources and constraints that can shape the availability of behavioral options. Although hypotheses H2 and H4 were preregistered, we treat them here as exploratory. Existing literature on the determinants of PEB is still characterized by substantial heterogeneity in operationalizations, as well as variation in effect sizes and even effect directions25,87,88. Therefore, rather than testing highly specific directional predictions, we used these analyses to examine whether distinct emotions and individual predispositions might shape climate-related decisions in different ways. The hypotheses for this study were preregistered before data collection: https://osf.io/cdym2. The study was conducted within a formal Polish-Norwegian research collaboration, which enabled pooling of expertise and resources in both countries. Although the study was not designed as a cross-cultural comparison, including both samples allowed us to examine whether the preregistered effects were robust across contexts that differ in climate-policy history, energy systems, and social conditions. We therefore treat the national context as a theoretically motivated robustness check. Both Poland and Norway are economies heavily dependent on fossil fuels, but with different policies regarding climate change46,47. Both societies exhibit high climate awareness, with climate change widely recognized as a serious threat. Climate emotions in both contexts appear to be morally and socially embedded rather than purely individual experiences, and may function as motivators for climate action46–48. Poland faces substantial structural barriers to decarbonization while being the most coal-dependent European economy, accounting for 10.7% of the EU’s net greenhouse gas emissions as of 202389. Although climate change and environmental pollution were consistently identified as major threats by Poles between 2019 and 2022, concern has declined notably since 2020, suggesting a shift from mobilization to normalization of the climate crisis90. Many renewable technologies still evoke skepticism, and energy transition in Poland is framed as necessarily gradual, conditioned on equity and just transition provisions for coal-mining regions that retain considerable political influence91,92. At the same time, Poland’s electricity system has changed rapidly in recent years. Solar photovoltaics has been one of the fastest-growing markets in the EU, driven in part by bottom-up adoption by individual prosumers, despite persistent grid and regulatory constraints93. In contrast, Norway’s context is characterized by high social trust, state capacity, and a longer history of climate policy implementation46. Despite this, Norway faces a distinct set of tensions as both a climate policy leader and the world’s fifth-largest oil exporter in a situation scholars have characterized as the “Norwegian climate paradox”94. Norway’s petroleum-dependent economy, which finances the comprehensive welfare state, may introduce specific psychological dynamics around climate responsibility, including potential for organized denial or avoidance95. Finally, resistance to certain climate policies and renewable technologies persists, including protests over onshore wind expansion96. Materials and methods Participants The study was conducted in Poland (N = 1029) and Norway (N = 1010) among adult native Polish speakers and native Norwegian speakers. The sample size was based on the preregistration. We included a diverse group of participants, with a demographic profile (gender, age, and education level) broadly reflecting the Polish and Norwegian populations (see Appendix A, Table A.1 and A.2). Importantly, we aimed to recruit a similar number of participants with low, moderate, and high levels of climate change concern. Participants received remuneration equivalent to 20 PLN (~ 4.5 EUR) and 200 NOK (~ 17 EUR), respectively. Importantly, they were offered the opportunity to donate some or all of the money they earned to a climate change organization (World Wide Fund for Nature, WWF). The study was conducted online, and volunteers were recruited by an external company via mass mailing, targeting a broad range of respondents. Each participant was dynamically assigned to the group with the fewest observations at the time. Procedure A purpose-built, secure web application was used to collect the data. The participants first read the description of the study’s aims and provided their informed consent. The experiment used a between-subjects design with an emotion manipulation, in which participants were randomly assigned to one of the emotion conditions. Next, participants completed several questionnaires measuring relevant variables that could affect pro-environmental intentions and behaviors. We measured the level of climate change belief and concern, personal and collective action efficacy80, psychological distance to climate change97, willingness to sacrifice for the environment81, as well as demographic variables (gender, age, place of residence, education level, socio-economic status, parenthood status). The exact content of items can be found in Appendix A Fig. 1. Overview of the experimental procedure. Emotion manipulation To elicit emotions, we used stimuli from the Emotional Climate Change Stories (ECCS) dataset98; see Appendix A. ECCS stimuli were developed to be ecologically plausible while instantiating appraisal-relevant features associated with several distinct climate emotions (anger, anxiety, compassion, hope, and guilt), repeatedly linked to climate engagement and pro-environmental behavior in prior research12,21,88,99. It was previously shown in the ECCS dataset validation study that some stories elicited mixed emotional responses, possibly because of the recruitment of overlapping appraisals98. Therefore, a classification algorithm was used to select the stories that elicited the specific target emotion to the greatest extent100. Story content was additionally checked against appraisal-theoretic criteria to ensure that the selected stories broadly related to appraisal patterns summarized in Table 1. Individual stories were selected to approximate these patterns, but some that deviated from them could elicit a slightly different emotional experience. Participants were allocated to experimental groups (ANG = anger induction; COM = compassion induction; HOP = hope induction) or to the control group (NEU = emotionally neutral state) and were asked to read 9 stories at their own pace. The mean length of stories in Poland was 294.6 characters (SD = 54.39), while in Norway, it was 281.3 (SD = 59.88). To assess the effectiveness of the manipulation, we then asked the participants to rate their emotional state (valence, arousal, intensity of anger, compassion, and hope). Outcome measures Participants were asked about their donation intention (‘Would you make a donation to a climate change organization?’; answer provided on a 4-point scale from “Definitely no” to “Definitely yes”) and donation aim preference (‘If you were to make a donation to a climate change advocacy organization, which aim would you prefer to support?’). Participants were presented with the following choices: fighting against corporations and policies that exacerbate climate change, aiding those affected by climate change, investing in solutions to mitigate climate change, or none of the above. Pro-environmental behaviors were measured with two tasks: the Work for Environmental Protection Task (WEPT76) and a donation task. In WEPT, participants could voluntarily engage in 15 short number identification tasks. For each task completed by the participant, we transferred a small donation to WWF (0.5 PLN in Poland or 4 NOK in Norway; a maximum of 7.5 PLN or 60 NOK). To ensure participants put in the effort, we informed them that donations would depend on task accuracy (the task was considered completed with at least 90% of numbers identified correctly). In the donation task, participants decided whether they wanted to donate part of their remuneration to WWF. While a composite pro-environmental behavior score (cPEB; a weighted average of completed WEPT tasks and donation amount) was calculated as per preregistration, it did not provide additional insights beyond the individual tasks and is therefore not reported in the main text to maintain clarity. Results for cPEB can be found in the Supplementary Materials. The entire procedure (administration of questionnaires, emotion manipulation, and post-manipulation measures of pro-environmental intentions and behaviors) took approx. 20 min. Data analysis Data analysis was conducted using R statistical software101. Data from participants who did not complete the study were discarded. In addition, data quality was assessed by comparing participants’ responses regarding sex, age, and level of concern about climate change with their answers to the same questions in the screening survey. Participants were excluded if their responses were inconsistent across the two surveys. Based on these criteria, we excluded 36 Polish participants (6 due to inconsistent reporting of sex, 22 due to inconsistent age, and 8 due to inconsistent level of concern about climate change) and 61 Norwegian participants (28 due to inconsistent reporting of sex, 13 due to inconsistent age, and 27 due to inconsistent level of concern about climate change). Some participants were excluded for more than one reason. The proposed research questions were tested with the following methods: H1. Compared with the neutral condition, participants exposed to any emotional stories (pooled together) will show higher PEB, operationalized as (H1a) greater WEPT effort and (H1b) larger donations. These hypotheses were tested with Welch two-sample t-tests within each country sample. H2. Compared with the neutral condition, participants exposed to different emotional stories will show a different level of PEB: ANG vs. NEU, COM vs. NEU, and HOP vs. NEU. In addition, PEB will differ between the three emotion conditions (ANG, COM, HOP; planned set of pairwise contrasts). These effects were investigated separately for (H2a) WEPT and (H2b) donations within each country sample with a one-way analysis of variance (ANOVA). H3. Donation aim preferences will depend on the induced emotion condition: (H3a) anger will increase preference for punitive/corrective aims, (H3b) compassion will increase preference for helping victims, and (H3c) hope will increase preference for solution-focused aims. These hypotheses were tested using chi-square tests of independence (aim × condition) within each country sample. H4. Beyond experimental conditions, individual differences (climate concern, efficacy, psychological distance, willingness to sacrifice, and demographics) will moderate PEB outcomes (WEPT, donations), investigated using multiple regression models within each country sample. To characterize the sensitivity of the final samples, we conducted sensitivity analyses for the main preregistered tests, assuming two-sided α = 0.05 and 80% power. The study had 80% power to detect small differences between the neutral and pooled emotional conditions, corresponding to Cohen’s d = 0.201 in Poland and d = 0.206 in Norway. For the four-group ANOVAs, the minimum detectable effects were Cohen’s f = 0.103 in Poland and f = 0.104 in Norway, corresponding to η² ≈ 0.011 in both countries. For the chi-square tests of donation aim preference, the minimum detectable associations were Cohen’s w = 0.129 in Poland and w = 0.134 in Norway, corresponding to Cramer’s V = 0.074 and V = 0.077, respectively. All datasets and data analysis scripts are available on https://github.com/nencki-lobi/PEB. Results After data quality control, the final samples comprised n = 1029 participants from Poland and n = 1010 from Norway. For clarity, details of all reported analyses (including confidence intervals and effect sizes) are presented in the Appendices B-E. On average, participants in Poland spent more time reading stories (M = 15.11, Mdn = 11.88, SD = 35.41 s) compared to participants in Norway (M = 10.18, Mdn = 8.54, SD = 12.47 s). Reading times in Poland exhibited greater variability, as indicated by a higher standard deviation. We assessed the effectiveness of emotion manipulation by comparing the story ratings within and between groups; descriptive statistics are shown in Table 2. As expected, participants in the HOP group rated their emotional state as more positive than those in the NEU, ANG, and COM groups. Also, the reported level of arousal was higher for participants exposed to emotional stories (ANG, COM, HOP) than for those exposed to neutral stories (NEU). Furthermore, each group of participants gave the highest ratings on the scale corresponding to the intended emotion. Specifically, the ANG group gave the highest ratings on the anger scale, the COM group on the compassion scale, and the HOP group on the hope scale. NEU stories were rated slightly positive. COM stories were also characterized by elevated anger ratings. This pattern was observed in both Poland and Norway, indicating consistent effects across cultures. Inferential manipulation-check results are reported in Table 3. Table 2Means and standard deviations of story ratings across experimental conditions in Poland and Norway. Ratings were measured on a continuous scale for valence, arousal, anger, compassion, and hope. Bolded text shows ratings on matched emotion scales.Rating scalePoland: M (SD)Norway: M (SD)NEUN = 260ANGN = 257COMN = 261HOPN = 251NEUN = 244ANGN = 256COMN = 255HOPN = 255Valence70.11 (20.53)20.58 (23.13)20.54 (22.60)81.66 (19.74)62.96 (20.85)31.02 (26.50)32.38 (26.97)73.50 (21.17)Arousal38.81 (25.18)63.17 (23.98)69.74 (22.75)60.78 (24.93)44.81 (26.03)53.40 (24.76)63.85 (24.14)58.13 (24.87)Anger11.52 (20.45) 70.86 (25.35) 63.86 (27.60)14.47 (22.53)23.62 (26.65) 62.64 (24.15) 54.76 (26.70)24.89 (26.81)Compassion25.06 (25.68)40.17 (29.45) 82.10 (21.91) 29.12 (28.52)48.53 (26.99)38.25 (28.24) 72.15 (21.99) 57.79 (27.11)Hope48.77 (31.19)21.12 (25.00)31.99 (27.00) 76.38 (24.25) 48.07 (26.45)32.25 (25.18)36.20 (24.84) 68.51 (24.19) Means and standard deviations of story ratings across experimental conditions in Poland and Norway. Ratings were measured on a continuous scale for valence, arousal, anger, compassion, and hope. Bolded text shows ratings on matched emotion scales. Table 3Manipulation checks: inferential statistics. Contrasts correspond to the following planned comparisons (valence: HOP vs. NEU; ANG vs. NEU; COM vs. NEU; arousal: EMO vs. NEU; emotion specificity: anger, compassion, and hope ratings compared in the target group vs. the remaining groups). EMO = pooled emotional conditions (ANG, COM, HOP). Reported omnibus F test with degrees of freedom, p value, and partial η², for each planned contrast, estimated mean difference (Δ), t statistic, p value (unadjusted, planned comparisons), and standardized effect size (d).Rating scaleF, pPartial η²Planned contrastΔt p d Poland ValenceF(3, 1025) = 574.90, p <.00010.63HOP vs. NEU11.56.053< 0.00010.536ANG vs. NEU−49.5−26.121< 0.0001−2.298COM vs. NEU−49.6−26.247< 0.0001−2.300ArousalF(3, 1025) = 79.82, p <.00010.19EMO vs. NEU25.814.819< 0.00011.063AngerF(3, 1025) = 438.90, p <.00010.56ANG vs. others40.923.525< 0.00011.694CompassionF(3, 1025) = 251.80, p <.00010.42COM vs. others50.726.655< 0.00011.910HopeF(3, 1025) = 201.40, p <.00010.37HOP vs. others42.421.625< 0.00011.570 Norway ValenceF(3, 1006) = 203.00, p <.00010.38HOP vs. NEU10.54.888< 0.00010.438ANG vs. NEU−31.9−14.824< 0.0001−1.326COM vs. NEU−30.6−14.183< 0.0001−1.270ArousalF(3, 1006) = 25.95, p <.00010.07EMO vs. NEU13.77.444< 0.00010.547AngerF(3, 1006) = 150.60, p <.00010.31ANG vs. others28.214.950< 0.00011.081CompassionF(3, 1006) = 77.00, p <.00010.19COM vs. others24.012.630< 0.00010.915HopeF(3, 1006) = 106.90, p <.00010.24HOP vs. others29.716.276< 0.00011.179 Manipulation checks: inferential statistics. Contrasts correspond to the following planned comparisons (valence: HOP vs. NEU; ANG vs. NEU; COM vs. NEU; arousal: EMO vs. NEU; emotion specificity: anger, compassion, and hope ratings compared in the target group vs. the remaining groups). EMO = pooled emotional conditions (ANG, COM, HOP). Reported omnibus F test with degrees of freedom, p value, and partial η², for each planned contrast, estimated mean difference (Δ), t statistic, p value (unadjusted, planned comparisons), and standardized effect size (d). Participants correctly completed a total of 8394 WEPT tasks in Poland and 5891 tasks in Norway (54% and 38% of all possible tasks, respectively). 240 WEPT tasks in Poland and 300 in Norway were completed incorrectly (number identification task accuracy < 90%). Participants in Poland completed an average of 8.16 WEPT tasks (SD = 5.72, Mdn = 7) and donated 4.08 PLN (~ 0.95 EUR), corresponding to 54.4% of the total possible donation (7.5 PLN). Participants in Norway completed on average 5.83 WEPT tasks (SD = 5.41, Mdn = 4) and donated 23.32 NOK (~ 2.04 EUR), corresponding to 38.9% of the total possible donation (60 NOK). In the donation task, participants could donate a portion of their remuneration (up to 20 PLN in Poland, ~ 4.5 EUR; 200 NOK in Norway, ~ 17 EUR). On average, participants in Poland donated 3.60 PLN (~ 0.84 EUR; SD = 4.99, Mdn = 1), corresponding to 18% of their remuneration. Participants in Norway donated 63.57 NOK (~ 5.55 EUR; SD = 79.01, Mdn = 10), corresponding to 31.8% of their remuneration. The correlation between the completed WEPT tasks and donations was small (r =.14 in Poland, r =.16 in Norway). Registered analyses Our first hypothesis posited that experiencing any emotion related to climate change would increase PEB. The results from Welch two-sample t-tests revealed no significant differences between groups exposed to emotional and neutral stories. Specifically, in Poland, the t-tests yielded non-significant results for WEPT performance (t(452.58) = −1.53, p =.127, d = −0.14, 95% CI [−0.33, 0.04]) and donation size (t(449.99) = 0.91, p =.363, d = 0.09, 95% CI [−0.10, 0.27]). Similarly, in Norway, the t-tests showed non-significant results for both WEPT performance (t(394.65) = 1.29, p =.199, d = 0.13, 95% CI [−0.07, 0.33]) and donation size (t(433.35) = −0.68, p =.497, d = −0.07, 95% CI [−0.25, 0.12]). Effect sizes were very small in both countries. We also proposed that the level of engagement in PEB would vary between ANG, COM, HOP, and NEU groups. However, ANOVA results did not support this hypothesis. There were no significant effects for WEPT performance (F(3, 1025) = 0.87, p =.455, η² = 0.0026) and donation size (F(3, 1025) = 0.92, p =.429, η² = 0.0027) in Poland. Similarly, in Norway, no significant effect of group was found for WEPT performance (F(3, 1006) = 1.09, p =.353, η² = 0.0032) and donation size (F(3, 1006) = 1.31, p =.270, η² = 0.0039). Effect sizes were very small in both countries. We also tested whether our manipulation increased action tendencies associated with distinct emotions. Specifically, anger was expected to relate to a tendency to punish perpetrators, compassion to a tendency to help victims, and hope to a tendency to support promising solutions. However, chi-squared tests examining the preferences for donation aim did not yield significant differences in either the Polish or Norwegian samples (in Poland: χ² = 16.581, df = 9, p =.056, Cramer’s V = 0.05; in Norway: χ² = 9.559, df = 9, p =.387, Cramer’s V = 0.01). A regression analysis was conducted to investigate whether other variables considered in prior literature were associated with the PEB outcome measures. For each country, we fitted linear models including experimental group, participants’ post-manipulation valence and arousal ratings, belief in climate change (BCC), climate change concern (CCC), personal and collective action efficacy (PCAE), psychological distance (PD), willingness to sacrifice for the environment (WTS), gender, age, place of residence (RES), level of education (EDU), parenthood status (Parenthood) and socio-economic status (SES). Multicollinearity was low in the preregistered main effect regression models (GVIF-adjusted VIFs; maximum GVIF^(1/(2*Df)) = 1.69), indicating that the inclusion of multiple predictors was unlikely to have substantially inflated standard errors. WEPT (Poland) For WEPT task completion, the model explained a small proportion of variance (R² = 0.063, adj. R² = 0.038; F(27, 997) = 2.49, p <.001). The experimental condition was not a significant predictor (all condition contrasts ns). Age was a positive predictor of WEPT performance (b = 0.0492, p <.001). Donations (Poland) For donation amounts, the model explained a small proportion of variance (R² = 0.071, adj. R² = 0.046; F(27, 997) = 2.84, p <.001). The experimental condition was again not significant (all condition contrasts ns). Higher WTS, age, and education predicted larger donations (WTS: b = 0.0699, p =.018; age: b = 0.0539, p <.001; education: relative to primary education, secondary, vocational, and university/college categories ps ≤ 0.037). WEPT (Norway) For WEPT task completion, the model explained a moderate proportion of variance (R² = 0.178, adj. R² = 0.155; F(28, 969) = 7.51, p <.001). The experimental condition was not a consistent predictor of WEPT. Relative to the NEU group, ANG group showed a small negative association with WEPT that was at the threshold of significance (b = −1.0068, p =.050), whereas the COM and HOP were not significant (ps ≥ 0.19). Age was again a strong positive predictor of WEPT (b = 0.1174, p <.001). Compared with the reference BCC category, endorsing ‘rather not happening’ and higher PD was associated with lower WEPT (BCC: b = −2.2058, p =.002, PD: b = −0.1778, p =.029). Reporting finding it difficult on present income predicted higher WEPT (b = 1.2869, p =.005). Donations (Norway) For donation amounts, the model explained a moderate proportion of variance (R² = 0.149, adj. R² = 0.124; F(28, 969) = 6.04, p <.001). The experimental condition effect was not significant. Age was again a strong positive predictor (b = 0.9498, p <.001). Higher CCC (levels 4 and 5), higher WTS, male sex, and lower SES categories were associated with larger donations (CCC: ps < 0.012; WTS: b = 1.2316, p =.002; sex: b = −13.1672, p =.007; SES: ps < 0.021). In summary, across both countries, experimental condition was not a robust predictor of PEB outcomes in the regression models, whereas age emerged as the most consistent positive predictor. Consistent with this pattern, observed means for WEPT and donations were very similar across ANG, COM, HOP, and NEU groups in both countries. Full regression results are presented in Table 4, and condition-level descriptive statistics are presented in Table 5. Table 4Estimate values of predictors for the outcome measures of PEB. Legend: BCC = belief in climate change, CCC = climate change concern, PCAE = personal and collective action efficacy, PD = psychological distance to climate change, WTS = willingness to sacrifice for the environment, RES = place of residence, EDU = level of education, Parenthood = parenthood status, SES = socio-economic status, p = p-value, ns = non-significant. P-values reported for individual regression coefficients are uncorrected and are provided for exploratory interpretation.Outcome measure/predictorPoland (n = 1029)Norway (n = 1010)WEPTDonationsWEPTDonationsIntercept0.94, ns−2.344, ns−1.254, ns−9.148, nsCategory: ANG0.278, ns−0.7, ns−1.007, p =.05*8.789, nsCategory: COM0.431, ns−0.449, ns−0.436, ns3.254, nsCategory: HOP0.505, ns−0.566, ns−0.611, ns0.289, nsValence−0.003, ns−0.008, ns−0.005, ns−0.091, nsArousal0.012, ns−0.001, ns0, ns−0.053, nsBCC: Rather happening0.184, ns−0.295, ns−0.334, ns−5.394, nsBCC: Rather not happening0.709, ns0.031, ns−2.206, p =.002**−16.935, nsCCC: 21.88, ns−0.663, ns0.68, ns21.713, nsCCC: 30.978, ns−0.352, ns1.097, ns19.021, nsCCC: 41.372, ns0.309, ns0.997, ns33.221, p =.012*CCC: 50.462, ns0.561, ns1.312, ns43.457, p =.003**PCAE0.085, ns−0.028, ns0.035, ns0.192, nsPD−0.084, ns0.005, ns−0.178, p =.03*−1.507, nsWTS0.061, ns0.07, p =.02*0.051, ns1.232, p =.002**Gender: Male−0.277, ns0.211, ns−0.581, ns−13.167, p =.01**Gender: Other8.321, ns4.633, ns3.048, ns−2.026, nsAge0.049, p<.001***0.054, p<.001***0.117, p<.001***0.95, p<.001***RES: Suburbs of a city−1.319, ns−0.994, ns0.211, ns3.284, nsRES: Small city0.721, ns0.262, ns0.164, ns10.823, nsRES: Country village−0.303, ns0.38, ns0.601, ns11.813, nsEDU: Secondary0.177, ns2.476, p =.01**0.733, ns−4.508, nsEDU: Vocational−0.118, ns1.904, p =.04*0.843, ns−4.353, nsEDU: University/College0.667, ns2.713, p<.003**0.828, ns−2.128, nsEDU: othernot applicablenot applicable−0.522, ns40.213, nsParenthood (no)−0.031, ns0.497, ns0.047, ns2.502, nsSES: Coping on present income0.868, ns0.345, ns0.428, ns−13.321, p =.02*SES: Finding it difficult on present income1.357, p =.05*−0.403, ns1.287, p =.004**−26.584, p<.001***SES: Finding it very difficult on present income0, ns−2.202, ns0.38, ns−36.675, p<.001*** Estimate values of predictors for the outcome measures of PEB. Legend: BCC = belief in climate change, CCC = climate change concern, PCAE = personal and collective action efficacy, PD = psychological distance to climate change, WTS = willingness to sacrifice for the environment, RES = place of residence, EDU = level of education, Parenthood = parenthood status, SES = socio-economic status, p = p-value, ns = non-significant. P-values reported for individual regression coefficients are uncorrected and are provided for exploratory interpretation. Table 5Descriptive statistics for key PEB outcomes by experimental condition in Poland and Norway. Values are observed means with standard deviations in parentheses. WEPT = Work for Environmental Protection Task performance.GroupPoland: M (SD)Norway: M (SD)WEPTDonationsWEPTDonations NEU 7.69 (5.65)3.85 (4.96)6.23 (5.59)60.68 (75.15) ANG 8.27 (5.71)3.62 (5.21)5.38 (5.19)70.01 (82.87) COM 8.47 (5.71)3.76 (5.46)5.80 (5.27)66.19 (79.23) HOP 8.20 (5.81)3.17 (4.22)5.95 (5.58)57.28 (78.31) Descriptive statistics for key PEB outcomes by experimental condition in Poland and Norway. Values are observed means with standard deviations in parentheses. WEPT = Work for Environmental Protection Task performance. Adding preregistered condition by moderator interactions did not improve model fit, F(73, 896) = 1.11, p =.256. Improvements were not observed for either outcome in Poland (WEPT: p =.441; donations: p =.435) nor in Norway (WEPT: p =.777; donations: p =.256). Because these main-effects models already provided little evidence for condition effects (and all p-values are uncorrected), subsequent condition-by-predictor interaction findings were not treated as a basis for substantive conclusions about moderation and are not reported here. The specific analyses can be found in Appendix A (Table A.3). Unregistered analyses Considering the mostly insignificant findings from our preregistered analyses, we investigated other factors that might have influenced the results. Even though the experimental manipulation was on average successful (as demonstrated in Table 2), individual differences in experienced emotions are to be expected. Therefore, we decided to run analyses on a filtered dataset comprising participants for whom the manipulation (elicitation of emotion) was successful. We operationalized successful manipulation using participants’ emotional state ratings. In the following analyses, we included: (1) participants from ANG, COM, and HOP groups who rated their current intensity of anger, compassion, or hope, respectively, as higher than 50% of the available scale; (2) participants from the NEU condition who rated their arousal as lower than 50% of the available scale. Applying these criteria, we obtained a filtered dataset of 827 participants from Poland (202 excluded) and 718 participants from Norway (292 excluded). We compared participants with successful versus unsuccessful manipulation to assess whether manipulation success is associated with participant characteristics. In Table 6, we summarize the most important differences. Because the two groups - further referred to as the successful and the failed group - differ in both overall sample size and the number of participants assigned to each experimental condition, we report descriptive statistics only. In the Polish sample, the failed group reported lower baseline climate engagement, particularly under emotional conditions. In the failed group, participants in ANG/COM/HOP reported substantially lower WTS than in the successful group. The failed group was also less likely to report a strong BCC and high CCC. The Norwegian sample showed the same general pattern of results. In ANG/COM/HOP, the failed group again reported lower WTS, lower BCC, and lower CCC. Table 6Descriptive differences in climate engagement between participants with successful versus failed manipulation, shown separately for each experimental condition and for Poland and Norway. Conditions (ANG, HOP, COM, NEU) are presented in columns. Willingness to sacrifice for the environment (WTS; sum score range 0–40) is reported as n, mean (SD); ΔM = mean_successful - mean_failed. High belief is defined as the percentage endorsing “Definitely happening” on the Belief in Climate Change (BCC) item; high concern is defined as the percentage endorsing “5 (Extremely concerned)” on the Climate Change Concern (CCC) item; Δpp = percentage points (successful - failed).CountryMeasureStatisticANGHOPCOMNEU Poland WTS (0–40) Successful: n, M (SD)215, 29.01 (7.43)214, 28.43 (7.82)233, 27.70 (8.38)165, 25.54 (8.92)Failed:n, M (SD)42, 22.83 (10.49)37, 17.22 (10.06)28, 17.75 (11.20)95, 29.47 (7.90) ΔM 6.18 11.22 9.95 −3.93 High BCC (5) Successful:n (%)142 (66.0)135 (63.1)145 (62.2)107 (64.8)Failed:n (%)21 (50.0)12 (32.4)10 (35.7)54 (56.8) Δpp 16.0 30.7 26.5 8.0 High CCC (5) Successful:n (%)82 (38.1)82 (38.3)87 (37.3)39 (24.1)Failed: n (%)7 (16.7)0 (0.0)2 (7.4)33 (34.7) Δpp 21.4 38.3 29.9 −10.6 Norway WTS (0–40) Successful: n, M (SD)183, 26.32 (7.81)200, 26.64 (7.39)212, 26.42 (7.31)123, 23.40 (9.61)Failed:n, M (SD)73, 20.82 (9.31)55, 17.04 (8.87)43, 18.63 (9.89)121, 27.08 (7.49) ΔM 5.50 9.60 7.79 −3.68 High BCC (5) Successful:n (%)104 (56.8)112 (56.0)115 (54.2)62 (50.4)Failed: n (%)18 (24.7)13 (23.6)16 (37.2)56 (46.3) Δpp 32.1 32.4 17.0 4.1 High CCC (5) Successful:n (%)48 (26.2)48 (24.0)64 (30.2)30 (25.0)Failed: n (%)9 (12.7)3 (6.0)6 (14.6)27 (22.3) Δpp 13.5 18.0 15.6 2.7 Descriptive differences in climate engagement between participants with successful versus failed manipulation, shown separately for each experimental condition and for Poland and Norway. Conditions (ANG, HOP, COM, NEU) are presented in columns. Willingness to sacrifice for the environment (WTS; sum score range 0–40) is reported as n, mean (SD); ΔM = mean_successful - mean_failed. High belief is defined as the percentage endorsing “Definitely happening” on the Belief in Climate Change (BCC) item; high concern is defined as the percentage endorsing “5 (Extremely concerned)” on the Climate Change Concern (CCC) item; Δpp = percentage points (successful - failed). Failed: n, M (SD) Successful: n (%) Failed: n (%) Successful: n (%) Failed: n, M (SD) Successful: n (%) Successful: n (%) After filtering the data, we again investigated differences between groups exposed to emotional versus neutral stories in terms of WEPT and donations using Welch t-tests. In Poland, the emotional condition showed higher WEPT performance than the neutral condition (M = 8.73 vs. 7.23), t(255.80) = −3.09, p =.002, and the effect size was small (Cohen’s d = −0.39, 95% CI [−0.63, −0.14]). Donation amount did not differ between conditions (M = 3.75 vs. 3.85, t(256.84) = 0.22, p =.824, d = 0.03, 95% CI [−0.22, 0.27]). In Norway, no significant differences were observed for either WEPT (M = 6.18 vs. 6.54), t(170.11) = 0.64, p =.524, d = 0.10, 95% CI [−0.20, 0.40], or donation amount (M = 72.16 vs. 59.07), t(186.88) = −1.73, p =.085, d = −0.25, 95% CI [−0.54, 0.04]). When comparing the group effect on PEB measures, the only statistically significant effect was observed for WEPT performance in the Polish sample (F(3, 823) = 3.28, p =.020, η² = 0.01). Subsequent post-hoc analyses revealed significant differences between groups ANG - NEU (p =.026, Cohen’s d = 0.29, 95% CI [0.02, 0.56]) and COM-NEU (p =.031, Cohen’s d = 0.28, 95% CI [0.02, 0.54]), with participants in emotional conditions exhibiting greater effort in WEPT. No other significant effect of the group on any other measure of PEB was identified in either the Polish or the Norwegian sample. In Poland, the chi-squared test for differences in donation aim preference was significant (χ² = 19.02, df = 9, p =.025, Cramer’s V = 0.07) when applied to the filtered dataset. Compared with the frequency of donation aim preferences in the NEU group, observed frequencies differed in the ANG condition (χ² = 15.989, df = 3, p =.001), with participants in the ANG condition showing a stronger preference for punitive donations (punishing the perpetrators) than those in NEU. In Norway, the chi-squared test for donation aim preference yielded a non-significant result (χ² = 15.95, df = 9, p =.068). Finally, when testing for potential moderation effects, we did not observe an improvement in explained variance when comparing direct effect regression models and moderation models for the filtered dataset (see Appendix A, Tables A.4 and A.5). Discussion This study aimed to investigate the causal relationship between emotional experiences and PEB in two underrepresented national contexts (in Poland and Norway). Drawing from appraisal theories of emotion, we expected that emotions would increase PEB and that distinct emotions would elicit different action tendencies, such as a desire to punish perpetrators (anger), help victims (compassion), or support solutions (hope)23,28,40. Contrary to these predictions, we did not observe any differences in PEB between participants exposed to emotional stories and those exposed to neutral stories. We also did not identify differences in action tendencies related to distinct emotions. Age was the only consistent predictor of PEB, with older participants donating more and spending more time on the WEPT task. These results were consistent across both Polish and Norwegian samples, and overall effect sizes were small. The manipulation checks showed that the stories elicited the intended emotions: participants in the ANG, COM, and HOP groups gave the highest ratings on the corresponding target scales in both countries (see Table 2). For planned comparisons, differences between ratings on relevant scales were statistically significant (see Table 3). However, on the descriptive level, the emotional reactions to stories were not entirely distinct, in the sense that other emotions were co-elicited to some extent. This was especially visible for the COM stories, where mean compassion ratings were highest, but mean anger ratings were also elevated. This suggests that COM stories may have activated concern for victims, as well as a perception of their harm resulting from moral violation. This partial overlap between emotions may have increased within-condition heterogeneity and attenuated average between-condition differences. Furthermore, even though stories belonged to one category, they could still differ in more subtle aspects of appraisal structure. The diversity of actors, events, and circumstances reflects the bottom-up development of ECCS from qualitative material47. For example, anger could be elicited not only by environmental harm itself but also by perceived hypocrisy, such as when people presenting themselves as environmental advocates act in ways that contradict their stated values. Although both reactions may be labeled as anger, they may reflect partly different appraisal patterns, centered respectively on harm or injustice versus norm violation and inconsistency. The impact of story content on appraisal structure could also play a role in the neutral condition. NEU stories were intended to function as non-emotional, non-climate-related control stimuli. However, some stories describing mundane social interactions contained prosocial content (e.g., a person informing a stranger about their untied shoelaces), which may have evoked a slight positive emotional reaction. This is supported by the observation that, descriptively, NEU stories were rated positively in valence, though not as positively as HOP stories. It is debatable whether incidental positive affect unrelated to climate change would impact PEB102. However, from the perspective of goal framing theory103, prosocial content may have shifted participants’ focus toward normative concerns, thereby increasing the likelihood of PEB. Additionally, because of non-climate-related content, NEU stories differed from emotional stories not only in emotional tone but also in topic. Future studies should consider using an additional passive control condition or baseline stimuli that are more affectively and normatively neutral. We argue, however, that the contamination of emotion categories is unlikely to fully account for the null effects. If the main mechanism were simply that the emotions elicited by different story categories were insufficiently specific, one might still expect emotional stories, as a broader category, to increase PEB relative to neutral stories. This was not observed (H1). Another consideration is that elicited emotions may not have been sufficiently intense, sustained, or behaviorally consequential to shift immediate behavioral choices in a brief, single-session experiment. The evidence for the effectiveness of brief interventions is mixed: some studies report that even 5-minute interventions change observable climate advocacy behaviors104,105, but most report modest average effects and substantial heterogeneity across intervention types, with outcomes depending on the study context and design25,106,107. Some authors argue that short emotional manipulation may not be sufficient to impact attitudes108 or substantially alter behavior in the long term109 and that prolonged interventions may allow for the accumulation of effects110. The potential mismatch between elicited action tendencies and available behavioral options is also to be considered. For example, anger-related punitive or norm-enforcing motivations may not map neatly onto donating money or investing time, which could dampen emotion-specific effects. Related work using ECCS in a different configuration reported effects on advocacy behaviors such as petition signing104, suggesting that emotions elicited by ECCS can be effective across different behavioral targets. To examine whether the null effects could be partly explained by unsuccessful emotion induction in some participants, we conducted additional exploratory analyses in a subset of participants for whom the manipulation appeared successful. Story ratings allowed us to define criteria for successful manipulation: participants in the emotional conditions were retained if they reported experiencing the target emotion to a substantial degree, whereas participants in the neutral condition were retained if they reported low arousal. These analyses should be interpreted cautiously because the manipulation-success subset did not appear to be a random subset of the full sample. Descriptively, and acknowledging imbalance in subgroup sizes, participants who responded successfully to emotional stories were more climate-engaged than those who did not. Across both countries, successful manipulation in the ANG, COM, and HOP conditions was associated with stronger belief that climate change is happening, higher climate concern, and greater willingness to sacrifice. In contrast, applying the successful-manipulation criterion in the NEU condition tended to exclude participants with these characteristics, because climate-engaged participants were more likely to respond emotionally even to neutral or low-arousal material. This pattern makes the subgroup analyses difficult to interpret. Any effects observed after applying the manipulation-success criterion may reflect either the experimental manipulation, pre-existing climate engagement, or their combination. After repeating the preregistered analyses in this subset, most effects on PEB remained statistically nonsignificant, especially in the Norwegian sample. In the Polish sample, anger and compassion were associated with greater WEPT effort, but these effects were sparse and modest. Therefore, they should be treated as exploratory, particularly because the subgroup selection was not preregistered and may have been related to baseline climate engagement. The robustness of the null effects was also visible at the country level. In both Poland and Norway, the preregistered analyses converged on the same conclusion: despite clear effects on self-reported emotions, the manipulation did not reliably translate into higher WEPT effort, donations, or emotion-specific action tendencies. The results did not diverge, despite Poland and Norway differing on some macro-level factors that could plausibly affect the variables assessed in the current study. For instance, lower social trust in Poland might have led to more negative appraisals of the collective efficacy, lower willingness to sacrifice (captured by PCAE and WTS); a high-trust, high-capacity welfare state setting of Norway might have facilitated more positive appraisals of the collective efficacy; different geographic and climatic circumstances might have impacted the perceived psychological distance to climate change (PD). These country-level differences were not the focus of this study; however, future work can test these macro-level moderators explicitly within the present dataset. Summary and future directions In this study, we sought to address common challenges in the study of climate emotions by explicitly defining target emotions, appraisals, and hypothesized action tendencies. Despite the absence of preregistered effects, the present findings provide useful constraints for theory and intervention design. Preregistered null results help reduce publication bias and inflated effect-size estimates, improving cumulative inference in psychological science111,112. They also provide important boundary conditions for theories, prevent unnecessary duplication of effort, and offer insights into methodological refinements. We propose several recommendations for future research. Because the present stories elicited emotions but may not have produced sufficiently specific, intense or sustained emotional episodes, future work may benefit from intentionally constructed, immersive and naturalistic stimuli39,113,114. High-immersion media, including video, interactive simulations, and virtual reality, provide rich multisensory cues and may elicit strong emotional responses115,116. So far, developing such stimuli has been very resource-intensive, but the rise of generative AI makes it increasingly feasible to create large, naturalistic stimulus sets using precise prompts. At the same time, researchers in environmental psychology should not overlook the substantial social and environmental costs of using such tools117. Overwhelming evidence indicates that hypothetical scenarios and self-report measures of PEB do not accurately predict real behaviors118. Still, creating a behavioral paradigm that replicates the critical contingencies of real-life situations remains challenging119. In most daily behaviors, PEB demands an investment of various resources (e.g., money, time, cognitive or physical effort) that otherwise would be spent differently. Thus, we advocate for the intentional choice of tasks to measure pro-environmental behaviors in intervention studies. Framing of the behavior should be matched to the action tendency produced by manipulation (e.g., punishing perpetrators after induction of anger, repairing damage after induction of guilt)40. Furthermore, considering the types of costs (money, time, and effort) associated with the task is essential, especially since different demographic groups may value them differently. Finally, to address the issue of remote and uncertain environmental consequences of climate-friendly choices, researchers should use experimental tasks with measurable outcomes72,118. Another potential area for improvement is the length of the intervention. Previous research suggests that a single, short emotional manipulation may not be sufficient to affect attitudes108 or substantially alter behavior in the long term109. On the other hand, in a fixed-length experimental setting (where participants expect to spend a certain amount of time on the task), a negative correlation between intervention length and time spent on WEPT was observed105. Prolonged, bite-sized interventions may allow for the accumulation of behavioral effects110. Future research can take advantage of the ubiquity of mobile devices interwoven into participants’ lives - for example, by exposing people to numerous stimuli at different times of the day, over days or weeks. This consideration also leads to the question of the role of more stable predispositions, such as climate emotion disposition measures120,121. Some PEB may be more strongly expressed as an interaction between situational elicitation and individual differences in trait-like climate emotions, with baseline levels of these emotions measured before manipulation in future studies. In summary, our results do not imply that climate emotions are irrelevant. By providing a preregistered test using behavioral outcomes across two national contexts, this study contributes boundary conditions for emotion-based climate interventions and motivates future work that combines stronger or repeated inductions with behaviorally and psychologically matched outcome measures. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Author contributions D. Zaremba: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Validation, Visualisation, Writing-original draft; Writing-Review and editing; M. Marczak: Funding acquisition, Conceptualization, Methodology, Writing-Review and editing; J. M. Michałowski: Funding acquisition, Project administration, Writing-Review and editing; C. A. Klöckner: Conceptualization, Funding acquisition, Project administration, Supervision, Writing-Review and editing; A. Marchewka: Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing-Review and editing; M. Wierzba: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Project administration, Software, Supervision, Validation, Visualization, Writing-Review and editing. Funding The research conducted in this project is financed by the Norwegian Financial Mechanism for 2014–2020 Grant No. 2019/34/H/HS6/00677. Data availability The code used for the analyses can be accessed at: https://github.com/nencki-lobi/PEB. The preregistration is available at: https://osf.io/cdym2. Appendices contain detailed information about sample demographics, detailed analyses, as well as full study materials. Code availability The code used for the analyses can be accessed at: https://github.com/nencki-lobi/PEB. Declarations Competing interests The authors declare no competing interests. Ethics approval The research was carried out in compliance with the principles of the Declaration of Helsinki. The study protocol was approved by the SWPS University of Social Sciences and Humanities Research Ethics Committee in Poland (approval no. 2021-52-12). Consent to participate Informed consent was obtained from all individual participants included in the study. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Małgorzata Wierzba and Artur Marchewka share equal senior contribution. Contributor Information Dominika Zaremba, Email: d.zaremba@nencki.edu.pl. Artur Marchewka, Email: a.marchewka@nencki.edu.pl. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Data Availability Statement The code used for the analyses can be accessed at: https://github.com/nencki-lobi/PEB. The preregistration is available at: https://osf.io/cdym2. Appendices contain detailed information about sample demographics, detailed analyses, as well as full study materials. The code used for the analyses can be accessed at: https://github.com/nencki-lobi/PEB.