News reporting about climate change often emphasizes people's emotions, including fear, hope, guilt, compassion, and even nostalgia (Höijer, 2010). Similarly, thinking about climate change evokes an array of emotional responses in people (Smith and Leiserowitz, 2014), even though, as a scientific topic, climate change “possesses few features that generate rapid, emotional, visceral reactions” (Weber, 2006; Markowitz and Shariff, 2012).

In this article, we are particularly interested in finding out how frequently used types of climate information influence an individual's emotional reactions (similar to Nabi et al., 2018 “classic” climate messages). A core of four messages are repeatedly used by media outlets, advocates, or scientists and/or suggested as useful for climate communication: consensus messaging that emphasizes scientific agreement about the causes and consequences of climate change (“consensus message”; Van der Linden et al., 2015, 2019; Cook et al., 2016); causal scientific process messaging that explains the mechanisms by which climate change occurs (“causes message,” Denning, 2020); impacts messaging that discusses the numerous and disparate impacts of the changing climate (“impacts message,” Hart and Feldman, 2014; Nabi et al., 2018); and solutions messaging that demonstrates methods of progress and highlights the action currently being taken (“solutions message,” similar to efficacy messages; Feldman and Hart, 2016, 2018; Nabi et al., 2018).

Although there has been some research that links these types of climate information with various specific emotional reactions, in this study, we expanded on this research and investigated how each of these four messages is linked to five different discrete emotions that are important in the climate context—guilt, anger, hope, sadness, and fear—and, in turn, how each of these emotions is linked to climate policy support. We defined the emotions as follows: Guilt is the experience of distress associated with believing oneself (or one's ingroup) to have contributed negatively to an outcome (Ferguson and Branscombe, 2010); anger is emotional arousal associated with a goal or plan being thwarted (Bandura, 1973); hope is a future-oriented positive assessment of an uncertain outcome (Lazarus, 1991); sadness arises when an irreparable loss is thought to have occurred (Bandura, 1973; Nabi, 1999); and fear arises when a threat is perceived (Lazarus, 1991). In the following sections, we examined each of the four types of climate information (consensus, causes, impacts, and solutions; see the Methods section for the operationalization in this study) and assessed the evidence—or lack thereof—of the types of emotional responses to those messages.

While there is a large body of research on cognitive reactions in response to consensus messaging (i.e., the Gateway Belief Model, Van der Linden et al., 2015, 2019), there has been relatively less research on emotional mechanisms. However, people report experiencing anger when exposed to information about the political and corporate barriers to U.S. climate action, a common feature of consensus messaging (Bieniek-Tobasco et al., 2019), indicating that consensus messaging may generate an important emotional response. Furthermore, for those concerned about climate change, learning about the scientific consensus on the issue may evoke either fear or hope. Fear may arise through realizing the inevitability of the consequences of continuing to burn fossil fuels. Conversely, hope may arise by fostering optimism that the scientific consensus will move attention toward implementing solutions, and away from debating the scientific validity of climate science claims.

A second commonly used message strategy is to provide scientific causal process information. To our knowledge, no study has examined emotional reactions to climate messages focusing on climate change's causes. However, Weber (2006) argues that scientifically focused messages about climate change per se are unlikely to generate emotions, so it may be that a message focused on understanding the causal mechanisms of climate change does not result in an emotional response. It is plausible, however, that learning how one's own actions are causing climate change could elicit emotions such as guilt and sadness.

A third common climate message is to provide information about climate change impacts. In some cases, climate impacts messaging has been shown to produce fear; for example, after viewing the images of mass extinction and drought stories in the documentary Years of Living Dangerously, participants reported feeling fearful of the consequences of climate change (Bieniek-Tobasco et al., 2019). Similarly, loss-framed climate information (communication that emphasizes what is lost through inaction rather than what is gained through action) has also been shown to increase fear (Nabi et al., 2018), indicating that an impacts message that focuses on what might be lost as a result of climate change may produce fear.

Finally, solutions messaging has been shown to produce hope in some research (Chadwick, 2015; Feldman and Hart, 2018; Nabi et al., 2018). Similarly, information about the human health benefits of addressing climate change has been shown to produce feelings of hope (Myers et al., 2012), as have depictions of “young” and “everyday” people (vs. elites) participating in addressing climate change (Bieniek-Tobasco et al., 2019). Additionally, solutions messaging is posited to decrease fear (Extended Parallel Process Model, Witte, 1992) and has been shown to do so (Hornsey and Fielding, 2016). However, in other research, solutions and/or efficacy messaging have not produced an emotional response (Feldman and Hart, 2016). Solutions messages have also been linked to both increased (Valentino et al., 2009) and decreased anger (Tausch et al., 2011), indicating an unclear expectation of how a solutions message might be related to anger.

Given the lack of research on the emotional impacts of some messages and mixed findings for others, we asked the following question:

Emotions surrounding climate change have been theorized and found to be among the most important predictors of such climate change responses as policy support and individual action (Roeser, 2012; Smith and Leiserowitz, 2014; Stevenson et al., 2015; Brosch, 2021). As individual action has a limited capacity to generate the wholesale societal changes necessary to effectively confront climate change (Gillard et al., 2016), policy support is a particularly important consideration in this context. In this study, we focused on five discrete emotions: guilt, anger, hope, sadness, and fear. In the following section, we reviewed the evidence that each of these emotions may be linked to climate policy support.

Guilt has been shown to lead to a behavior that seeks to repair the harm caused (Lewis, 1971; Dearing et al., 2005). In the environmental context, guilt has been positively associated with attitudes, intentions, and behaviors (Mallett, 2012; Harrison and Mallett, 2013; Rees et al., 2015; Moore and Yang, 2020), including willingness to engage in personally costly practices such as conserving energy and paying green taxes (Ferguson and Branscombe, 2010). Therefore, we tested whether guilt is positively associated with climate change policy support.

Anger arising from perceiving injustice that resulted in unfair outcomes has been associated with participation in collective action (Van Zomeren et al., 2008) and a desire for retribution (Nabi, 2003). Anger is often centered on a concrete and identifiable culprit as the cause of a negative situation (Smith and Ellsworth, 1985; Lazarus, 1991). In the context of climate policy options, we anticipated that anger would be associated with a preference to regulate emissions from carbon producers to address the perceived wrongs that they have perpetuated (Roseman et al., 1994; Harth et al., 2013). Therefore, we tested whether anger is positively associated with climate change policy support.

Hope has been theorized to be a motivator for action (Snyder, 2002; Chadwick, 2015). In the context of climate change, hope has been highlighted as a missing component in the climate action chain (Hulme, 2009; Roser-Renouf and Maibach, 2010; Ojala, 2012; Nabi et al., 2018), as it has been linked to a movement toward a goal outcome (Lazarus, 1991). Hope has been found to be a strong predictor of climate policy support (Marlon et al., 2019), and positive emotions generally are “linked…to people's engagement with climate change, largely in a productive manner” (Schneider et al., 2021). Therefore, we tested whether hope is positively related to climate change policy support.

Sadness can lead to an effort to reestablish what has been lost (Nabi, 1999). As most Americans perceive climate losses as being experienced by others who are far away in time and space rather than themselves (Leiserowitz, 2006; Weber, 2016) and as certain communities are especially vulnerable to climate impacts (Harrington et al., 2018), we tested whether sadness is positively related to climate change policy support.

Fear often leads to a desire for self-preservation (Lazarus, 1991; Nabi, 2003). However, the relationship between fear and action is unclear (Bieniek-Tobasco et al., 2019), with some studies finding a link between fear and action outcomes (Witte and Allen, 2000; Meijnders et al., 2001a,b; Swim and Bloodhart, 2015), while others do not, perhaps due to differences in self-or response efficacy or the level of fear experienced (O'Neill and Nicholson-Cole, 2009). The relationship between fear and policy support is also murky, as the desire for self-protection can lead to support for more strict policy regulation (Comartin et al., 2009) and/or for new policy options (Goodall et al., 2013). Therefore, we tested whether fear is positively related to climate change policy support.

Given these tested links between climate messages and emotions and between emotions and climate policy support, we also tested whether there are indirect effects of climate messages on policy support through emotions.

In an online study, participants were randomly assigned to one of the four video conditions (consensus, causes, impacts, and solutions) or a no-video control group.¹ Participants were first asked a set of five screening questions to determine their partisanship and issue segment. If the quota for their group had not been met, they were asked to read and sign the consent form, which included a pledge to watch the videos to their conclusion. They were then randomly assigned to watch one of the four videos (or the no-video control). Then, they completed a posttest questionnaire, which included measures of the emotions they felt while watching the video (or when considering climate change, if in the control group); the target of their anger, asked of those who said they felt angry; their climate change attitudes and policy preferences; and demographics. The protocol was approved by the George Mason University Institutional Research Board.

Quota sampling by the online sample vendor, Qualtrics, was implemented between August and October 2021 to recruit participants based on self-identified partisanship and their climate change issue public audience segment.² Recruitment of the smaller segments was challenging, and the balance of Republicans and Democrats within segments was skewed in the initial data collection, requiring additional sampling, which was conducted by Climate Nexus.³

For the five conditions included in this study, the final sample (N = 3,023) was 36% men and 59% women, with 5% missing or other. The median age was between 45 and 54 years, with an average level of education of “some college”. Concerning race, 84% of participants identified as white, 10% as Black, 3% as Asian or Pacific Islander, 2% as Native American, and 7% as Hispanic or Latino.

Four whiteboard videos focusing on the scientific consensus on climate change, the causal process of climate change, climate change impacts, and solutions to address climate change were created for this study (Table 1) and were modeled after NASA's Earth Minutes.⁴ The videos are of similar length (78–105 s), have a narrator with a similar voice, and were created by the same company, Truscribe. Most of the text and imagery for the science and impacts videos were taken directly from Earth Minutes. Content in the impacts video was also informed by the Fourth National Climate Assessment; the text of the consensus video was largely based on the message used by Van der Linden et al. (2015); and the text of the solutions video was based on content from Hawken (2017). The video scripts are included in the Supplementary material.⁵

To assess whether the emotions differed by a condition in comparison to the control, five regressions were run, predicting each emotion from the experimental conditions (with the control condition as the referent dummy-coded condition).

The conditions elicited different emotional reactions (Figure 1). In comparison to the control condition, the causes condition increased guilt (b_guilt.causes = 0.601, p < 0.05), the impacts condition increased sadness (b_{sadness.impacts} = 0.560, p < 0.05), and the solutions condition increased hope (b_{hope.solutions} = 0.540, p < 0.05). Conversely, the consensus, causes, and impacts conditions each decreased hope in comparison to the control condition (b_{hope.consensus} = −0.553, p < 0.05; b_hope.causes = −0.688, p < 0.01; b_hope.impacts = −0.880, p < 0.001); while the solutions condition decreased anger, sadness, and fear in comparison to the control condition (b_{anger.solutions} = −1.110, p < 0.001; b_{sadness.solutions} = −1.226, p < 0.001; b_{fear.solutions} = −1.116, p < 0.001).

Next, we tested whether each emotion was independently related to support for climate change policy. Using regression, we predicted climate change policy support from guilt, anger, hope, sadness, and fear, controlling for condition and political affiliation.⁶ Each emotion, except for guilt (b = 0.004, p = 0.775), was uniquely and positively related to climate policy support (b_{policysupport.anger} = 0.129, p < 0.001; b_{policysupport.hope} = 0.065, p < 0.001; b_{policysupport.sadness} = 0.049, p < 0.001; b_{policysupport.fear} = 0.058, p < 0.001).

Hayes' (2017) PROCESS model 4 was used to test whether the conditions had significant indirect effects on climate policy support through emotion (see Table 2 and Figures 2–5 for all indirect effects), controlling for political affiliation. The consensus and causes conditions both had significantly negative indirect effects on policy support through hope, while no other indirect effect was significant for either of these conditions [see Figures 2, 3; ab_{policysupport.hope.consensus} = −0.035, Bootstrapped CI = (−0.072, −0.002); ab_{policysupport.hope.causes} = −0.044, Bootstrapped CI = (−0.083, −0.011)].

The impacts condition also had a significant negative indirect effect through hope [ab_{policysupport.hope.impacts} = −0.056, Bootstrapped CI = (−0.096, −0.022)] but had a significantly positive indirect effect through sadness [Figure 4; no other indirect effects were significant for the impacts condition; ab_{policysupport.sadness.impacts} = 0.028, bootstrapped CI = (0.002, 0.063)].

Finally, the solution message produced significant and negative indirect effects through anger, sadness, and fear, along with a significant and positive indirect effect through hope [Figure 5; the indirect effect through guilt was not significant; ab_{policysupport.anger.solutions} = −0.141, Bootstrapped CI = (−0.218, −0.070); ab_{policysupport.sadness.solutions} = −0.059, Bootstrapped CI = (−0.106, −0.021); ab_{policysupport.fear.solutions} = −0.063, Bootstrapped CI = (−0.112, −0.025); ab_{policysupport.hope.solutions} = 0.036, Bootstrapped CI = (0.026, 0.074)].

Several limitations of this research should be noted and considered. Of primary interest, the messages were not designed to evoke an emotional response, and research has shown that emotionally charged messages result in a more significant behavioral change (and, presumably, policy support change) than informational messages (Dickerson et al., 1992; Aitken et al., 1994). While this is a limitation of this study, it is also a strength in that the current study offers an examination of emotional responses to the “baseline” climate messages and demonstrates that these informational messages, in fact, have effects on emotions (especially the suppression of some emotions). Future research should compare how emotionally loaded versions of these messages perform.

Furthermore, the experimental design assessed the effects of a single message at one time point with no counter or additional messages, an approach that is not strong in ecological validity. The design also did not manipulate the emotional mediators, making it impossible to claim causality in the indirect effects (although causal claims are warranted for the messages to emotions and messages to policy support links). Future research could address this limitation by manipulating both the independent variables and the mediators.

Additionally, perceptions of climate change are closely tied to an individual's political affiliation (Weber, 2010, 2016), and messages have been received differently across the political spectrum (e.g., Feldman and Hart, 2016). For the sake of parsimony, we did not investigate the moderating role of political affiliation; however, future research should examine how political affiliation affects the relationships between messages and emotions and between emotions and policy support.

Finally, there was an intercorrelation between some of the emotions (Supplementary Table 1), indicating that those who react emotionally to a message experience a range of emotions. This high intercorrelation does indicate that “a rising tide will lift all boats”—or that high emotionality about climate is likely associated with high support for the policy. Additionally, in an effort to reduce respondent burden while simultaneously assessing a range of emotions, emotions were measured with only a single self-reported measure. Future research would benefit from assessing emotions in a more robust way with multiple indicators. Additionally, future studies would benefit from including more positive emotions, such as pride, responsibility, care, compassion, and solidarity.

These results point to some notes of caution for climate messengers: specifically, the climate messages had negative indirect effects on policy support through all of the emotions simultaneously (the total indirect effect) due to the suppression (in comparison to the control condition) of hope for the consensus, science, and impacts conditions and the suppression of anger, sadness, and fear for the solutions condition. This finding indicates the power that emotions have; these overall negative indirect effects through emotion completely offset the direct positive effect of the messages on policy support.

We have demonstrated that emotions are important for understanding how people react to climate messages and whether those messages influence support for climate policies. Our results also suggest that these types of climate information are perhaps better utilized in combination rather than isolation, that is, by linking multiple types of message content together (consensus, impacts, and solutions, for example). Furthermore, practitioners and researchers should consider how to develop messages that engage people on multiple emotional fronts, accentuating positive emotions (e.g., hope, compassion) without sacrificing other types of emotional engagement (e.g., guilt, anger, sadness, or fear).

In sum, climate messages that are intended to activate policy support should (1) seek to engage the audience emotionally (because emotions about climate change are strongly related to climate policy support), (2) simultaneously elicit multiple emotions, and (3) avoid decreasing some emotions while increasing others (i.e., not trading off sadness and hope).