Assessing the Role of Self-Efficacy in Reducing Psychological Reactance to Guilt Appeals Promoting Sustainable Behaviors

Abstract

Guilt is a widely used emotional appeal in environmental sustainability and other advocacy messages. However, unintended negative effects such as psychological reactance, where individuals resist or reject the message, have been identified, and scholars have recently suggested strategies to mitigate those effects. Using an experimental design, the study examined the role of self-efficacy cues (lower vs. higher magnitude behavioral recommendations) in the message recommendation component of guilt appeals that are designed to elicit more guilt than shame. Adult-aged participants (N = 341) were exposed to messages that described the negative effects of climate change on wild animals. Results from a serial mediation test indicated that viewing guilt messages with lower (vs. higher) magnitude behavioral recommendation was associated with greater self-efficacy beliefs and less perceived threat to freedom, which predicted more anticipated guilt and, finally, greater intention to take pro-environmental actions. These findings contribute to the theoretical development and practical applications regarding how self-efficacy-enhancing content in guilt appeals can reduce defensive responses and foster a stronger commitment to sustainable practices.

1. Introduction

Climate change is one of the most salient issues and substantial challenges impacting the United States and the rest of the world today [1]. Climate change’s cascading negative impacts, such as increasing temperatures, rising sea levels, frequent storms, and floods. have posed significant threats to human societies and natural ecosystems. Although levels of greenhouse emissions temporarily decreased due to COVID-19 lockdowns in 2020 as millions of people were quarantined at home to prevent the spread of the virus [2], the limited amount of reduction has a negligible impact on the impending climate crisis [3]. This global environmental crisis necessitates immediate and effective action to reduce the negative consequences of climate change.

Communication can play a role in shaping individuals’ attitudes, beliefs, and risk perceptions regarding climate change and can influence individuals’ mitigation behaviors [4,5,6]. However, studies have found that merely providing individuals information about the current or future negative consequences of climate change, such as animals and plant extinction, lack of clean water sources, and extreme weather events, is often not sufficient to change their behaviors [7,8]. To more effectively encourage public pro-environmental behaviors in helping mitigate climate change, it is important that climate-change risk messages be “emotionally rewarding and easy to follow” [9] (p. 1). To this end, scholars have suggested the crucial role of emotions in shaping the public’s mitigation behaviors and in effective communication about climate change [10,11].

Among various emotions, feelings of guilt and/or anticipated guilt have been found to be positively associated with the practice of sustainable behaviors—specifically, pro-environmental behaviors, or PEBs [12,13,14]—ostensibly due to both the moral underpinnings of the emotion of guilt and the moral aspects and drivers of PEBs [15,16]. Given the ability of guilt to motivate PEBs, messages promoting those behaviors may be influential to the extent that they elicit message recipients’ feelings of guilt for past behavior or anticipated guilt related to future behaviors. Perhaps not surprisingly, guilt appeals—persuasive messages that attempt to create a mental conflict between one’s behaviors and moral standards to encourage behavioral change—have been identified as a widely used emotional appeal in pro-environmental messages [17,18,19]. Well-designed guilt appeals are proposed to motivate behavior change because those who recognize a failure to live up to their own standards typically feel compelled to make amends or take reparative action in order to reduce feelings of guilt [15].

Despite extensive research on guilt appeals in promoting pro-environmental behaviors, significant gaps remain in understanding how to optimize these messages for effectiveness while minimizing unintended negative responses such as psychological reactance. This study addresses a critical gap in the literature by exploring the role of self-efficacy cues in the behavioral recommendations of guilt appeals, an area that has been overlooked. The study’s findings offer new insights into how guilt appeals can be structured to enhance their persuasive impact while reducing the likelihood of resistance and boomerang effects, thereby contributing additional knowledge in the area of communication regarding sustainable behaviors.

By examining the role of self-efficacy and reactance in shaping responses to guilt appeals through the lenses of the appraisal theory, social cognitive theory, extended parallel process model, and psychological reactance theory, this study aims to refine and enhance strategies for messaging about sustainable behaviors. Different from previous studies that have predominantly focused on the content of guilt arousal, this study provides a novel focus on the behavioral recommendations that accompany these appeals. The insights gained could inform the design of more effective communication strategies that promote sustainable behaviors while minimizing reactance and other counterproductive effects. The current study also advances understanding of message effectiveness by investigating the nuanced mechanisms between guilt, self-efficacy, perceived freedom-threat, psychological reactance, and pro-environmental behavioral intentions. Overall, the findings of this study aim to contribute to the theoretical understanding of guilt appeals and to provide practical implications for developing more effective climate-change communication strategies that can drive meaningful public engagement and sustainable actions.

2. Literature Review and Hypotheses

2.1. Dual Components and Conflicting Findings of Guilt Appeals

Guilt appeals typically have two components, including the external stimulus that arouses guilt and the appeal’s recommended action [15,20,21]. These types of messages aim to create a discrepancy between personal or societal standards and the message recipients’ current or previous behaviors [15,22]. Guilt appeals may focus on a past transgression, thereby leading to guilt arousal. Conversely, guilt appeals may focus on a behavior that has not yet occurred, which leads to anticipated guilt [20,21]. Although guilt appeals can motivate intended attitudinal and behavioral responses, they can also lead to unintended, negative emotional responses and even reactance-like or “backlash” effects (e.g., anger arousal and behavioral boomerang effects), which can ultimately result in failed or counter-productive persuasive attempts (i.e., boomerang effects) [23,24,25,26].

Attempts to understand such unintended effects have led to conflicting findings regarding the relationship between guilt intensity in messages and message responses: Some studies identified a curvilinear relationship between guilt-appeal intensity and post-message attitudinal and behavioral intention changes, whereby too much guilt can lead to message rejection or acting in opposition to the recommended behavior (i.e., boomerang effects) [23,27]. However, a recent meta-analysis found that although more intense guilt appeals were associated with greater levels of felt guilt among message recipients, those feelings were neither linearly nor curvilinearly related to the effect guilt appeal intensity on persuasion [28]. Scholars have suggested that such inconsistency reflects challenges related to the manipulation of guilt in persuasive messages as well as unintended emotional responses [28,29].

Recent work has identified strategies for manipulating guilt in a manner more likely to lead to intended message outcomes and to a better understanding of the relationship between experienced guilt and message-recommended behavior (or intentions). Effectively eliciting guilt without concurrently eliciting other unintended responses (e.g., shame, anger, psychological reactance) plays an important role in influencing message recipients’ attitudinal and behavioral responses, although doing so can also be challenging [22,30,31]. Therefore, this study tested the effectiveness of messages designed to elicit anticipated guilt regarding climate change risks and recommended PEBs without concurrent, unintended effects (i.e., psychological reactance and shame). Specifically, the study varied the behavioral recommendations component of the messages to examine the role of task magnitude (i.e., difficulty) and subsequent self-efficacy beliefs on psychological reactance, the extent of felt guilt, and intentions to practice the message-recommended PEBs.

An unintended and counter-productive audience response to guilt appeals is audience feelings of anger and/or of being manipulated by the message, or what some have termed reactance-like effects, especially in response to more intense guilt appeals [25,28,30]. Extant research has identified a variety of strategies that can effectively elicit feelings of guilt while mitigating such defensive responses to guilt appeals such as focusing on the problematic behavior rather than the person enacting the behavior (i.e., eliciting guilt but not shame [22,32,33]; invoking interpersonal connections in the message [34]; focusing on anticipated guilt for future actions (or lack thereof) rather than guilt related to one’s past behaviors (reactive guilt) [35,36,37]; and eliciting empathy for individuals described in the message [38,39,40].

These strategies have generally been related to message content designed to describe the problem/problematic behavior and to elicit guilt among message receivers, which is the first component of a typical guilt appeal. However, guilt appeals typically also include a second component: content designed to inform recipients about how to alleviate or avoid those feelings of guilt—behavioral recommendations [41]. Few studies have attempted to identify how this second component of the message might contribute to the success (or failure) of guilt appeals. For example, Yang and Kruschke (2024) found that messages recommending behaviors toward which participants felt lower obligation to perform, compared to those with higher obligation to perform, were generally associated with higher levels of reactance [29]. Of particular interest in the current study is the inclusion and nature of efficacy information in the behavioral recommendation component of guilt appeals.

2.2. Self-Efficacy and Responses to Guilt Appeals

Lindsay (2005) proposed that for people to respond favorably to a guilt appeal, they must believe that the behavioral recommendations are efficacious—in short, that recipients should perceive self-efficacy to take necessary action and that such action will help ameliorate the problem described (i.e., perceive response-efficacy) as well as reduce their feelings of guilt [42]. However, efficacy information was held constant in that study, with strong cues of both self- and response-efficacy in all messages. Brennan and Binney (2010) similarly proposed and found that guilt appeals can be motivating only when recipients believe that individual action is necessary and likely to affect important social change, and that message-prescribed actions are within their individual capacities (i.e., have positive self-efficacy beliefs) [24].

The extended parallel processing model (EPPM), designed to explain rejection of threat (fear-eliciting) messages in the health domain, posits that reactance is one possible negative response to threatening messages with no or weak self-efficacy information [43]. Based on propositions of the EPPM but applying the model to guilt appeals, Basil et al. (2008) proposed that self-efficacy cues and beliefs would lead to less maladaptive responses to such appeals [44]. In line with this proposition, they found that exposure to messages that included a less difficult behavioral recommendation (i.e., donate $2 to a charity) resulted in greater perceived self-efficacy, less reactance responses, higher levels of guilt, and greater intentions to donate than exposure to messages that included a more difficult behavioral recommendation (i.e., donate $200 to a charity). Of note, these results were found to be due to differences in perceived efficacy associated with the two behavioral recommendations rather than perceptions of the large donation recommendation as being unreasonable. Other scholars have recommended the inclusion of strong efficacy information in guilt messages because efficacy cues or beliefs may reduce feelings of shame experienced during guilt appeal exposure, which can in turn reduce the likelihood of eliciting psychological reactance [22]. Collectively, these works suggest that message recommendations that specify easier, rather than more difficult behaviors be practiced in order to mitigate the given problem could result in greater perceptions of efficacy, less defensive message responses (i.e., less reactance), and greater practice of recommended actions.

The magnitude of a behavior (i.e., difficulty of a task) is an important factor that influences individuals’ self-efficacy beliefs [45,46]. According to Bandura’s social cognitive theory, self-efficacy plays a crucial role in determining whether individuals adopt specific behaviors. The theory proposes that individuals are more likely to practice a behavior when they perceive greater self-efficacy—the belief in their own ability to successfully complete the task [45]. This relationship suggests that the perceived difficulty of a task can significantly impact an individual’s willingness to take action. Therefore, exploring whether the difficulty of the recommended behavior in guilt messages about a pro-environmental behavior would influence audiences’ self-efficacy is of theoretical and practical importance, given that people can adopt many different behaviors to protect the environment. For example, encouraging people to reduce plastic straw use (easier behavior/lower magnitude) or stop using plastic products (e.g., plastic bags or food containers/more difficult behavior and therefore higher magnitude) might both contribute to mitigating plastic pollution, but the difficulty level of these behaviors varies. Although more difficult behaviors are often needed to affect meaningful social change or environmental protection [7], recommending more difficult behaviors in advocacy messages might lead to defensive responses [47], at least among those who have not previously taken any similar actions. Therefore, we tested the proposition that a first step to mitigating rejection of guilt appeals could be starting with a recommendation of a less difficult (i.e., lower magnitude) behavior that recipients feel capable of performing.

Hypothesis 1 (H1). 

Message condition (1 = lower magnitude behavioral recommendation, 0 = higher magnitude behavioral recommendation) will be positively related to self-efficacy beliefs.

2.3. Two-Step Model of Psychological Reactance with Self-Efficacy Beliefs

Previous research has found that messages recommending more (vs. less) difficult behavior can lead not only to lower perceptions of self-efficacy but also, in turn, to defensive message responses: Those who we feel unable to take the recommended risk-reducing actions might be motivated to deny the risk via message rejection [48]. The current study focuses on psychological reactance as a particularly relevant defensive response that has been proposed to result from exposure to persuasive messages with a high magnitude of behavioral request [47,49]. As an aversive defensive motivational state [50,51], psychological reactance is comprised of both anger (i.e., affective component) and negative cognitions (i.e., cognitive component) [52,53,54] that occur following a threat to freedom. Hence, psychological reactance theory [50,55] hinges on a perceived freedom-threat (step 1), which predicts state psychological reactance (i.e., a combination of anger and negative cognitions) (step 2) [52,53,56]. Of particular interest in our study is the relationship between self-efficacy beliefs and freedom-threat perceptions, specifically within the context of guilt appeal messages. We posit that guilt appeals that recommend more (vs. less/lower) difficult/higher-magnitude behaviors will lead to greater freedom-threat perceptions, especially when message recipients believe they are not capable of practicing the recommended behavior (i.e., have low self-efficacy beliefs). As noted by Rains and Turner (2007), requests for more substantial behaviors (e.g., abstaining from drinking alcohol vs. reducing alcohol consumption) can lead to perceived freedom threat because such behaviors may be perceived by message receivers as too great of an imposition on their “time, energy, or finances” [47] (p. 247) or as beyond their personal capabilities. The green nudge literature provides additional insight into how the magnitude of behavioral requests can be designed to minimize psychological reactance. Some evidence indicates that green nudging (using defaults or policy changes to encourage sustainable behavior) is more acceptable and effective when the nudged behavior is less effortful and when individuals believe they have the freedom to choose whether or not to engage in the behavior [57].

Therefore, when individuals believe that they lack sufficient resources or that the recommended behavior is too difficult to accomplish (e.g., never using any plastic products or containers again), they may perceive that their freedom to complete other, more-attainable behaviors (e.g., increasing plastic recycling) is being threatened (i.e., their choices are being limited, a key elicitor of psychological reactance). This perceived restriction can trigger reactance, as individuals feel their choices are being limited. In short, less perceived ability (self-efficacy) to perform a recommended behavior should be associated with greater perceived threat-to-freedom and subsequent psychological reactance [58] and with greater perceived need to restore freedom [59]. Thus, an inverse relationship is expected between self-efficacy and freedom-threat perceptions, such that greater self-efficacy beliefs will be associated with lower freedom-threat perceptions.

Hypothesis 2 (H2). 

Self-efficacy beliefs will be negatively associated with freedom-threat.

As described herein, research has demonstrated that psychological reactance is best modeled as a two-step process wherein freedom-threat predicts psychological reactance [51,52,53,54,60,61]. Thus, in our study, we have predicted that message condition (0 = higher magnitude/difficulty of behavioral recommendation, 1 = lower magnitude/difficulty) will lead to greater self-efficacy beliefs (H1), which will be associated with lower freedom-threat (see H2). Next, following the two-step model of psychological reactance [52,53,54,60,61], we predict that freedom-threat perceptions will be associated with greater psychological reactance (i.e., a combination of anger and counterarguments) [61]. Therefore,

Hypothesis (H3). 

Freedom-threat will be positively associated with psychological reactance.

2.4. Modeling Anticipated Guilt in the Two-Step Model of Psychological Reactance

In addition to considering the role of self-efficacy, we further advance the psychological reactance literature by revisiting the role of guilt as a consequence of a freedom-threat perception and the subsequent experience of psychological reactance. Indeed, the role of guilt, as an emotional response and its relation to psychological reactance has been examined in prior work, with each study using different variations to model guilt [30,44,62,63]. However, in a few of these studies, the authors did not measure freedom-threat, and therefore, freedom-threat could not be modeled as an antecedent of psychological reactance [30,44,62], which limits the generalizability of the results to psychological reactance theory. In contrast, in their study, Quick et al. (2015) measured freedom-threat as a predictor of psychological reactance (i.e., two-step model) with guilt modeled as an antecedent to freedom-threat [63]. They argued that guilt can generate greater freedom-threat and arouse subsequent psychological reactance because guilt appeals can evoke anger [31,41], which is the affective component of psychological reactance [52].

However, the current study deviates from Quick et al. (2013) given more recent findings pertaining to the defensive nature of psychological reactance theory [54]. In their studies, Clayton (2022) and Clayton and colleagues (2023) exposed participants to a series of either freedom-threatening or freedom-enhancing language in persuasive health messages [51,64]. They collected psychophysiological data as indices of cognitive and emotional processing during message exposure while also measuring self-report and memory measures post-message exposure. Both studies found that participants in the freedom-threatening-language message condition self-reported greater freedom-threat and psychological reactance (also modeled as a combination of anger and counterarguments) than participants in the freedom-enhancing message condition. Interestingly, participants in the freedom-threatening condition allocated significantly less attention to, and had worse memory of, message content relative to participants in the freedom-enhancing message condition. Clayton and colleagues concluded that perceived freedom-threat and psychological reactance are associated with an aversive motivational state and a cascade of defensive responses, specifically less cognitive allocation (i.e., attention) to the messages [20,51,64,65]. That is, if participants disengage with the message via defensive processing (e.g., cognitive avoidance), then they are less likely to attend to and subsequently have worse recognition memory for message content intended to elicit guilt [51,64]. Therefore,

Hypothesis 4 (H4). 

Psychological reactance will be negatively associated with anticipated guilt.

Appraisal theory suggests that emotions such as guilt arise when individuals evaluate a situation and realize that their actions or inactions have caused harm or violated their moral standards [66]. This theory provides a framework for understanding how guilt, as a moral emotion, can be evoked and utilized in promoting sustainable behaviors. For instance, messages that elicit recipients’ feelings of guilt often do so by making them aware of a prior or potential future behavior that is under their control and violates their personal or moral standards [15]. The emotional response to guilt often leads to coping strategies aimed at managing the emotional experience or altering the situation that caused it. In the context of sustainability-related guilt appeals, this response might result in reparative actions, such as adopting more sustainable behaviors to alleviate the emotional discomfort [14,16,67]. For instance, two meta-analyses have found a motivating influence of guilt on PEBs, with medium effect sizes identified for effects of guilt on behavior and medium-to-large effects for pro-environmental intentions [14,68]. Of note, scholars have suggested and found that messages that elicit reactive guilt (experienced for past transgressive behaviors) can lead to more defensive processing than messages that elicit anticipated guilt (feelings we imagine having if we do not take recommended action/behave in accordance with our standards in the future), because it can be easier to avoid a future transgression than to correct a past one [37]. Anticipatory guilt can also motivate PEBs because people tend to overestimate how strongly they will feel guilt in the future [13]. Based on those findings, we designed the guilt appeals in the current study to elicit anticipatory guilt and proposed that successfully doing so would lead to greater intention to take recommended actions.

Hypothesis 5 (H5). 

Anticipated guilt will be positively associated with pro-environmental behavioral intentions.

To test the relationships (H1–H5) within a four-mediator sequential mediation chain following the two-step model of psychological reactance, we proposed the following hypothesis:

Hypothesis 6 (H6). 

Message condition (1 = lower magnitude/difficulty of behavioral recommendation, 0 = higher) will be associated with increased self-efficacy (see H1), which will be negatively related to freedom-threat (see H2), which will be positively related to psychological reactance (see H3), and psychological reactance will be negatively associated with anticipated guilt (see H4), which will be positively associated with intentions to follow message recommendations (behavioral intentions) (H5).

3. Method

3.1. Design and Stimuli

An online experiment used a 2 × 4 between-subjects design that manipulated magnitude/difficulty of the behavioral recommendation (lower vs. higher) and a message replication factor varying types of animals described as suffering from effects of climate change (moose vs. bears vs. beetles vs. salamanders). Portraying animals that are similar vs. dissimilar to humans can influence audiences’ emotional responses toward suffering animals with more empathy and concern typically shown for animals with greater phylogenic similarity to humans [69,70,71]. Therefore, the phylogenic similarity of animals suffering from climate change described in the stimuli were varied to control for possible differences in guilt arising from the type of animal depicted. Furthermore, to enhance the credibility and ecological validity of the messages, the selected animals were listed as vulnerable or endangered species affected by climate change by reliable scientific organizations such as National Geographic and World Wildlife Fund. Additionally, these animals/insects were selected because they are less symbolic than some other animals (e.g., butterflies symbolize transformation and freedom, bees are commonly associated with industriousness and productivity, and polar bears have already come to symbolize climate change threats in general). The choice of less symbolic animals/insects was intended to reduce the risk of emotional or cultural biases affecting participants’ responses.

Guilt messages featured short paragraphs accompanied by photographic images informing participants of the negative effects of climate change on moose/bears/beetles/salamanders and attributed the cause to (non-specific) human behavior that harms the environment. To elicit guilt, messages stated that human activities have significantly contributed to climate change and have led to animal endangerment and extinction. The animals and insects were described as having difficulty finding food and suffering from increasing heat and damage to their habitats (e.g., wildfires). Based on research indicating that guilt and shame manifest via different cognitive appraisals and have different action tendencies, with shame being less likely to lead to reparative behaviors [72] and greater likelihood of message rejection [22], messages were designed to elicit guilt but not shame. The messages referred to “humans’ negligent” and “reckless behaviors” included as guilt-eliciting terms but not including the word “you” or putting the focus on the individual’s specific behaviors (e.g., [22,33]). The message concluded with four behavioral recommendations (described below) to mitigate climate change and effects on animals.

Message-recommended behaviors were based on results of a pilot test (N = 229) of undergraduate students who rated their perceived difficulty of performing 17 pro-environmental behaviors (1 = extremely difficult to 7 = extremely easy). Behaviors chosen were rated as moderately difficult to perform, with mean perceived difficulty scores ranging from 3.5–4.7. The message-recommended frequency of performing each of the pro-environmental behavior was adjusted to manipulate lower vs. higher magnitude/difficulty of behavioral recommendation. Messages with lower-magnitude/less difficult recommendations encouraged participants to “try to use less plastic by avoiding using plastic straws and avoiding products that are packaged in plastic (such as beverages, fresh fruits and vegetables, and cleaning or beauty products)”, “try to eat or purchase vegetables which are organic or locally-grown once a month”, “try to recycle plastics at home, school, or work once a month”, and “try to use less air conditioning in summer or heat in winter”. Messages with higher-magnitude/more difficult behavioral recommendations encouraged participants to do the same behaviors “always” (always avoid plastic, always avoid using heating and air conditioning, always recycle, and always eat organic or locally grown vegetables).

Mean rating scores for the behaviors examined in the pilot-test and experimental stimuli can be found in the supplemental materials document located on our Open Science Framework (OSF site) (see pilot study folder and STIM folder): https://osf.io/24r9a/?view_only=516c4f0acf58431c92069df71202eaf0.

3.2. Experiment Procedure

This study was approved by the university’s Institutional Review Board (IRB) prior to data collection. Participants were recruited from Prime Panels, an online marketing research panel. Participants were invited via the survey platform to join the study. Inclusion criteria ensured that participants were at least 18 years old and provided informed consent before participating. Participants also received small monetary incentives for their participation.

The data collection process involved multiple stages to capture pretest and posttest responses. Participants first completed a pretest questionnaire online via the survey platform. The pretest questionnaire measured biospheric values, outcome efficacy, and demographics. After two days, participants received an online link directing them to the posttest stimulus material. This delay was designed to minimize carryover effects from the pretest. Participants were randomly assigned to one of the eight message conditions. They were then directed to complete the post-message questionnaire. The post-message questionnaire measured emotions, self-efficacy, threat-to-freedom, reactance (i.e., anger and counterarguing), anticipated guilt, and pro-environmental behavioral intentions.

3.3. Participants

Three-hundred and eighty-one participants recruited from Prime Panels completed the pretest, but three failed an attention check. Among the 378 qualified pretest participants, 361 completed the full experiment, but one failed an attention check in the posttest. The two attention checks were designed to identify and exclude participants who were not paying sufficient attention to the survey items. In particular, we instructed participants to “Please select Strongly Disagree for this item” in the pretest and “Please select Feeling this way very strongly as your response to this question” in the posttest. Those who did not comply were excluded from the analysis. Additionally, nineteen outliers identified as being more than 1.5 times the interquartile range (IQR) above the third quartile or below the first quartile on the boxplot of each dependent variable and state shame were removed, resulting in a final sample of 341 participants. Participants’ ages ranged from 18 to 78 (M = 39.43, Mdn = 38.00, SD =11.18). In terms of race and/or ethnicity, most participants self-identified as White (74.5%), 13.8% identified as Black or African American, and 7.3% identified as Asian. In terms of ethnicity, 8.2% of the participants identified as Hispanic, Latino, or as having Spanish origins such as Mexican, Puerto Rican, or Cuban. The sample was fairly gender-balanced with 51.3% of participants identifying as male and 48.7% as female.

The ideal sample size for the study was calculated using Aberson’s “pwr2ppl” R package (Version 0.4.0). The Pearson correlation between each pair of variables was set as Pearson’s r = 0.25. Results indicated an approximate sample size of N = 350 was sufficient to yield a power (1 – β) of 0.8 for the serial mediation tests. Additionally, a post-hoc power analysis indicated that the sample size N = 341 achieved a power of 0.98.

3.4. Control Measures and Manipulation Check

Biospheric values (α = 0.91) commonly predict pro-environmental concerns and behaviors and so were measured as a check on equivalence of research conditions in a pretest by four items from Schwartz (1992) [73]. Example items include “Protecting the environment: preserving nature”, “Unity with nature: fitting into nature”, and “Respecting the earth: harmony with other species”. Participants were asked to rate the importance of each value as a guiding principle in their lives [−1 (opposed to my values), 0 (not important) to 7 (extremely important)] (M = 4.51, SD = 1.25).

Trait Shame (α = 0.78). Although the stimulus messages were designed to elicit more guilt than shame (e.g., [22]), some individuals might have a tendency to experience higher levels of shame than others. Therefore, trait shame was measured in a pretest using three items that typically have relatively high factor loadings (0.78–0.86) from the Behavioral Shame Subscale of the Experience of Shame Scale (ESS) [74]. Responses were recorded using a 4-point Likert scale (1 = not at all; 4 = very much). Sample item: “Do you feel ashamed when you do something wrong?” (M = 2.62, SD = 0.82).

State shame (α = 0.90) was used as a manipulation check to ensure that messages elicited more guilt than shame. Feelings of shame were measured by four items adapted from Boudewyns et al. (2013) and Tangy (2002) [22,75]. Participants indicated the extent to which they experienced the following feelings (0 = not feeling this way at all to 7 = feeling this way very strongly) after viewing the message: “I want to sink into the floor and disappear”, “I feel ashamed”, “I feel like I am a useless person”, and “I feel embarrassed”. (M = 1.99, SD = 1.24).

3.5. Dependent Measures

Self-efficacy (α = 0.82) was measured with three items measuring capacity from Chang (2021) and two items measuring autonomy from Yzer (2012), both with 7-point Likert response categories (1 = strongly disagree to 7 = strongly agree) [76,77]. Sample items are: “I am confident in my ability to do the pro-environmental behaviors recommended in the message”, and “whether or not I will practice the pro-environmental behaviors recommended in the message is under my control”. (lower-magnitude recommendation: M = 5.45, SD = 0.97; higher-magnitude recommendation: M = 5.22, SD = 1.08).

Threat to freedom (α = 0.89) was measured by four items, “The message tried to make a decision for me”, “The message tried to manipulate me”, “The message threatened my freedom to choose”, and “The message tried to pressure me” on a 5-point Likert scale from 1 = strongly disagree, 5 = strongly agree (lower-magnitude recommendation: M = 2.43, SD = 1.19; higher-magnitude recommendation: M = 2.53, SD = 1.18) [52].

Psychological reactance was measured following the intertwined model of psychological reactance [52,53], which includes anger and counterarguments [61]. Anger (α = 0.89) was measured by four items: angry, irritated, annoyed, and aggravated on a semantic-differential scale from 1 = none of this feeling, 5 = a great deal of this feeling (lower-magnitude recommendation: M = 2.07, SD = 1.06; higher-magnitude recommendation: M = 2.02, SD = 0.98) [52]. Counterarguments (α = 0.90) were measured by three items from Silvia (2006): “Did you criticize the message you just saw while you were watching it?” and “While watching the message, were you skeptical of what was being said?” and “Did you think of points that went against what was being said while you were watching the message?” (1 = no, not at all to 5 = yes, very much) (lower magnitude-recommendation: M = 2.14, SD = 1.25; higher-magnitude recommendation: M = 2.30, SD = 1.24) [78]. Both variables were moderately and significantly correlated (Pearson’s r = 0.31, p < 0.01). Therefore, both were standardized and summed to create the psychological reactance variable (lower-magnitude recommendation: M = −0.10, SD = 1.69; higher-magnitude recommendation: M = −0.02, SD = 1.48) [61,65,79,80].

Anticipated guilt (α = 0.96) was adapted from a well-established scale in the environmental context [81]. Participants indicated the extent to which they would feel the listed emotions in a conditional form, “If during the next two weeks you will not act pro-environmentally to help mitigate climate change and help the animal/insect in the story to survive, how much do you think you would feel “guilty”, “remorse”, and “sorry” along with three distractors “disappointed”, “unsatisfied”, and “fearful” (1 = not feeling this way at all to 7 = feeling this way very strongly). (Lower-magnitude recommendation: M = 3.53, SD = 2.08; higher-magnitude recommendation: M = 3.68, SD = 2.00).

Pro-environmental behavioral intentions (α = 0.70) were measured by asking participants to indicate how likely they were to perform the following behaviors within the next two weeks: “reduce plastic use”, “purchase organic or locally grown vegetables in grocery stores”, and “reduce air conditioning/heat use” along with four distractor items (1 = extremely unlikely to 7 = extremely likely) (lower-magnitude recommendation: M = 4.62, SD = 1.58; higher-magnitude recommendation: M = 4.82, SD = 1.38).

4. Results

4.1. Preliminary Analyses

Prior to hypothesis testing, a one-way ANOVA was conducted to examine whether the replication factor of animal type (moose or bear vs. beetles or salamander) had equivalent effects on feelings of guilt and shame. Results indicated that differences by animal type were not significant within the psychological distance factor (similar to humans vs. dissimilar to humans) on reported guilt and shame (see

Table 1. ANOVA results of animal types within psychological distance factor.

Animal Type	Psychological Distance	Guilt				Shame
		F	df	p	η2	F	df	p	η2
Bear	Similar to Humans	3.59	1/172	0.06	0.02	0.03	1/172	0.86	0.0002
Moose
Salamander	Dissimilar to Humans	2.55	1/165	0.11	0.02	0.99	1/165	0.32	0.01
Beetle

for the ANOVA results). Therefore, the original eight experimental conditions were collapsed into two research conditions: higher (n = 166) vs. lower (n = 175) magnitude/difficulty of behavioral recommendation.

Moreover, participants’ self-reported gender identity did not differ by research condition, X² (1, N = 341) = 0.83, p = 0.36, and Cramer’s V = 0.05. Furthermore, biospheric values did not significantly differ between the two research conditions: lower magnitude/difficulty (M = 4.42, SD = 1.28) vs. higher (M = 4.61 SD = 1.20), t(339) = 1.45, p = 0.15, and Cohen’s d = 0.16. However, participants’ self-reported trait shame differed significantly between the higher (M = 2.51, SD = 0.82) and lower magnitude/difficulty conditions (M = 2.72, SD = 0.80), t(339) = −2.34, p = 0.02, and Cohen’s d = 0.25. Therefore, trait shame was used as a control variable for hypothesis testing. Additionally, a post-hoc MANCOVA was conducted to examine whether age had an unintended influence on self-efficacy, freedom threat, reactance, anticipated guilt, and pro-environmental behavioral intentions. The results showed that these dependent measures were not significantly influenced by the participants’ age across the two research conditions, Wilks’ Lambda = 0.98, F(5, 334) = 1.30, p = 0.26, and ηp² = 0.02.

Finally, a paired-samples t-test indicated that messages overall elicited significantly more guilt (M = 3.60, SD = 2.03) than shame (M = 1.99, SD = 1.24) among participants, t(340) = −16.23, p < 0.001, and Cohen’s d = 0.88; the same pattern occurred within the higher-magnitude/difficulty condition, t(165) = −12.53, p < 0.001, and Cohen’s d = 0.97, and within the lower-magnitude/difficulty condition, t(174) = −10.55, p < 0.001, and Cohen’s d = 0.80. An independent samples t-test also indicated that feelings of shame did not differ between message conditions (lower vs. higher magnitude/difficulty), t(339) = −0.48, p = 0.63, and Cohen’s d = 0.05. Therefore, messages were considered to have elicited significantly more guilt than shame, and shame was found to not differ significantly across the two main conditions.

4.2. Tests of Hypotheses

In order to test the hypotheses and investigate the sequential process by which message condition influenced PEBIs through potential mediating variables, a four-mediator sequential mediation analysis was conducted using PROCESS for SPSS v4.2 (Model 6), with 10,000 using bootstrap simulations. We used PROCESS because of its ability to directly test the significance of indirect effects of research conditions on the outcome variable using bootstrapped confidence intervals. This approach provides more accurate and reliable estimates compared to other multi-step regression methods [82]. Additionally, compared to other methods such as maximum likelihood estimation, PROCESS’s ordinary least squares (OLS) method enhances statistical power and reduces the Type II error rate [83,84]. In the model, condition was entered as the predictor, self-efficacy was entered as the first mediator, freedom threat was entered as the second mediator, psychological reactance was entered as the third mediator, anticipated guilt was entered as the fourth mediator, and pro-environmental intentions was entered as the outcome variable [84]. Trait shame was included as a covariate in the analysis and predicted reactance (B = 0.19, p < 0.05, CI [0.02, 0.36]) and anticipated guilt (B = 0.47, p < 0.001, CI [0.23, 0.72]). By controlling for trait shame in the model, we accounted for its influence on both reactance and anticipated guilt, thus isolating observed effects of research condition and mediators in the mediation analysis from the effects of individual differences in trait shame.

Both unstandardized (B) and standardized path coefficients (β) are reported for estimates of the association between the variables in hypothesis tests (unstandardized and standardized coefficients are reported in the results section; unstandardized coefficients are displayed in the serial mediation model (Figure 1)) [84]. See Figure 1 and

Table 2. Total, direct, and indirect effects among modeled variables.

Total, Direct, and Indirect Effects		Pro-Environmental Behavioral Intentions
Total and Direct Effects of X on Y	Unstandardized	Standardized	Boot SE	t	Boot LLCI	Boot ULCI
Total effect of X on Y	−0.1951	-	0.1623	−10.20	0.2304	−0.5144
Direct effect of X on Y	−0.2801	-	0.1153	−20.43	00.156	−0.5069
Indirect effects of X on Y	Unstandardized	Partially Standardized	Boot SE		Boot LLCI	Boot ULCI
Total indirect effect of X on Y	0.0851	-	0.1210	-	−0.1549	0.3194
Indirect effect 1: X→M1→Y	0.1223	0.0823	0.0558	-	0.0160	0.2356
Indirect effect 2: X→M2→Y	0.0026	0.0018	0.0326	-	−0.0638	0.0689
Indirect effect 3: X→M3→Y	0.0001	0.0001	0.0075	-	−0.0199	0.0133
Indirect effect 4: X→M4→Y	−0.1223	−0.0823	0.0623	-	−0.2508	−0.0048
Indirect effect 5: X→M1→M2→Y	−0.0288	0.0194	0.0160	-	0.0030	0.0650
Indirect effect 6: X→M1→M3→Y	0.0001	0.0001	0.0015	-	−0.0034	0.0032
Indirect effect 7: X→M1→M4→Y	0.0425	0.0286	0.0221	-	0.0048	0.0922
Indirect effect 8: X→M2→M3→Y	0.0000	0.0000	0.0048	-	−0.0106	0.0103
Indirect effect 9: X→M2→M4→Y	0.0011	0.0007	0.0138	-	−0.0282	0.0288
Indirect effect 10: X→M3→M4→Y	−0.0006	−0.0004	0.0044	-	−0.0103	0.0091
Indirect effect 11: X→M1→M2→M3→Y	0.0004	0.0003	0.0048	-	−0.0087	0.0111
Indirect effect 12: X→M1→M2→M4→Y	0.0120	0.0081	0.0074	-	0.0010	0.0297
Indirect effect 13: X→M1→M3→M4→Y	−0.0003	−0002	0.0009	-	−0.0027	0.0007
Indirect effect 14: X→M2→M3→M4→Y	−0.0002	−0.0001	0.0028	-	−0.0067	0.0056
Indirect effect 15: X→M1→M2→M3→M4→Y	−0.0016	−0.0011	0.0023		−0.0071	0.0022

Note: X = Condition, M1 = Self-efficacy, M2 = Freedom threat, M3 = Reactance, M4 = Anticipated guilt, Y = Pro-environmental behavioral intentions. Number of bootstrap samples for bias corrected bootstrap confidence intervals: 10,000. Level of confidence of all confidence intervals: 95.

for the serial mediation analysis results. The bivariate correlations are presented in Section 1 located in the results folder on our OSF site.

H1 predicted that participants who viewed a guilt-appeal that described an animal or insect harmed by climate change with a lower magnitude/difficulty of recommended behavior would report higher self-efficacy beliefs than the participants who viewed a guilt appeal with a higher magnitude/difficulty recommendation. Model results indicated support for this hypothesis—B = 0.25 (β = 0.25), SE = 0.11, t(338) = 2.27, p < 0.05, 95% CI [0.03, 0.47].

In support of H2, greater perceived self-efficacy was associated with less freedom threat after viewing the message—B = −0.44 (β = −0.38), SE = 0.06, t(337) = −7.49, p < 0.001, 95% CI [−0.55, −0.32].

In support of H3, higher levels of freedom threat were, in turn, associated with greater psychological reactance—B = 0.74 (β = 0.55), SE = 0.06, t(336) = 11.51, p < 0.001, 95% CI [0.61, 0.87].

H4 predicted that psychological reactance would be negatively associated with anticipated guilt. However, model results indicated that psychological reactance was not associated with anticipated guilt—B = 0.07 (β = 0.05), SE = 0.08, t(335) = 0.86, p = 0.39, 95% CI [−0.09, 0.22]. Therefore, H4 was not supported. [Of note, the bivariate correlation analysis showed that psychological reactance was significantly and positively associated with anticipated guilt [Pearson’s r = −0.12, p < 0.05]. Additionally, though not hypothesized, freedom-threat significantly and negatively predicted anticipated guilt.

Finally, in support of H5, greater anticipated guilt was associated with greater pro-environmental behavioral intentions—B = 0.30 (β = 0.41), SE = 0.03, t(334) = 9.73, p < 0.001, 95% CI [0.24, 0.36]. An examination of the P-P plot for these two variables indicated that the relationship was positive and linear. Additionally, a (not hypothesized) direct effect of magnitude of recommended behavior on intentions was identified, such that a lower magnitude predicted less intention to perform the recommended behaviors. Further, freedom-threat had a (not hypothesized) direct, negative effect on intentions, and, in line with existing research and theory (but not hypothesized), self-efficacy had a direct positive effect on intentions.

H6 predicted a four-mediator sequential mediation process by which participants who read a guilt appeal describing an animal or insect being harmed by climate change with lower magnitude/difficulty of recommended behavior would report greater intention to practice pro-environmental behavior via greater perceived self-efficacy (M1), then lower feelings of freedom threat (M2) and greater psychological reactance (M3), and lower anticipated guilt (M4). Results indicated the proposed serial indirect effect of message condition on pro-environmental behavioral intentions was not significant—B = −0.002 (β = −0.0011), BootSE = 0.002, 95% CI [−0.01, 0.002]—due to the non-significant individual path from reactance to anticipated guilt, which ultimately caused the overall mediation effect to be non-significant. Therefore, H6 was not supported. However, the analysis identified a significant and positive indirect effect of message condition (lower magnitude/difficulty of recommended behavior) on intentions via greater self-efficacy, lower freedom threat, and greater guilt—B = 0.012 (β = 0.008), BootSE = 0.001, 95% CI [0.001, 0.030].

5. Discussion

Prior research has found that guilt appeals, if not designed appropriately, might inadvertently trigger defensive responses in audiences and might not influence desired behaviors. This study sought to address this issue by investigating how advocacy messages promoting more sustainable behaviors could mitigate psychological reactance through behavioral recommendations aimed at enhancing self-efficacy. In contrast with previous studies that focused on the type of language (e.g., dogmatic, controlling) used in the message, this study examined how the nature of the behavioral recommendation itself can influence defensive responses in the form of psychological reactance [29,44,47]. Moreover, the study employed messages designed to elicit more guilt than shame in an effort to limit defensive responses and to further illuminate the relationship between post-message anticipated guilt and intentions to engage in message-recommended sustainable behaviors.

With respect to the possibility of limiting psychological reactance to the messages, it was proposed that a first step in mitigating reactance could be for messages to recommend lower magnitude/easier behaviors. We found that exposure to messages with the option to practice moderately difficult actions at a fairly low frequency (i.e., practice the behavior once a month or just “try” to practice the behavior: a lower-magnitude recommendation) rather than at a high frequency (i.e., “always”: a higher-magnitude recommendation) was associated with greater perceived self-efficacy to take recommended pro-environmental actions. This finding is consistent with propositions from Bandura (1977) and supports previous research indicating that recommendations of less (vs. more) difficult behavior in guilt appeals were associated with greater self-efficacy [44,45].

Scholars have recommended incorporating strong efficacy cues in guilt appeals, as increased feelings of self-efficacy may help reduce defensive responses to these messages [22,24,42,44,48]. The current study offers further evidence by identifying a relationship between self-efficacy and defensive responses but modeled the defensive response of psychological reactance as a two-step process, in line with current recommendations [52,53,56]. The results supported the hypothesis that greater self-efficacy would be negatively related to perceived threat to freedom (H2), which is the initial step in the two-step psychological reactance model [54]. When individuals perceive a recommended behavior as unattainable and beyond their capabilities (i.e., have low self-efficacy beliefs), they might experience stronger perceptions that their personal freedom to complete other, more-actionable behaviors is being threatened [47]. We suggest the greater self-efficacy resulting from the lower-magnitude behavioral recommendation in this study, led participants to feel less constricted and to experience greater autonomy, which in turn reduced their psychological reactance. This finding is consistent with the core tenets of psychological reactance theory (H3), which posits that a greater perceived threat to freedom leads to stronger psychological reactance (i.e., a combination of anger and counterarguments).

In summary, the findings of H1–H3 identified a potentially effective way to mitigate reactance to guilt appeals that recommend PEBs: “start small” by asking recipients to (at least initially) make less difficult changes to their behavior. Although promoting more effortful behaviors may be essential for substantial social or environmental change, such recommendations could reduce efficacy beliefs and trigger defensive message responses, resulting in failed persuasive attempts [7]. On the other hand, initiating an advocacy campaign with more modest recommendations could lead to significant change over time because those recommendations can lead to behavioral trials. Behavioral trials might facilitate more substantial long-term behavior change by fostering increased self-efficacy, mastery, and/or a shift in self-identity over the long run [85,86,87]. Additionally, the current approach could also mitigate negative feelings toward the message sponsor, a common issue with ineffective guilt appeals [31]. Hence, starting small to enhance self-efficacy and mitigate psychological reactance could provide additional benefits to those advocating sustainable behaviors in the long run. Furthermore, this finding also resonates with the concept of green nudges, which aim to make sustainable choices more accessible without creating unnecessary barriers [57].

This study also attempted to enhance understanding of the relationship between psychological reactance and guilt by testing whether reactance would lead to anticipated guilt regarding pro-environmental action. Contrary to our prediction (H4), greater psychological reactance was not significantly associated with less guilt. However, the lack of a relationship between these two variables in the mediation model could be explained by a suppression effect of freedom threat on guilt: freedom-threat was a highly significant predictor of guilt and was the strongest predictor of that variable in the model. Additional evidence for a suppression effect of threat can be found in the bivariate correlation analysis indicating a significant, positive association between reactance and anticipated guilt, Pearson’s r = −0.12, p < 0.05. Although not hypothesized, we found freedom-threat had a direct and negative relationship with pro-environmental intentions. Thus, it is probable that a message could threaten one’s freedom and lead to a boomerang effect without the experience of psychological reactance [63,88]. Indeed, prior research testing psychological reactance theory has shown that, counter to their predictions, freedom-threat was directly and negatively associated with attitudes [63] and negatively associated with behavioral intentions [88]. These findings suggest that psychological reactance is a sufficient but may not be a necessary condition for the elicitation of a boomerang effect, whereas freedom-threat appears to be both necessary and a sufficient condition in the prediction of a boomerang effect. Similarly, our model revealed that freedom-threat directly and negatively predicted anticipated guilt. Hence, it seems likely that freedom-threat could decrease audiences’ anticipated guilt without evoking state psychological reactance. It’s also possible that when individuals perceive themselves as having their freedom threatened by a message, they may feel constrained or coerced, leading to reduced feelings of personal responsibility, which is a key motivation of anticipated guilt [89,90]. Consequently, this might ultimately decrease recipients’ anticipated guilt.

The results of this study add to those from other work (e.g., [22]) indicating that guilt appeals that do not concurrently elicit strong feelings of shame can reduce defensive responses and facilitate a positive relationship between guilt and message-recommended behavior (or intentions, in this case) (H5). This finding is consistent with prior research identifying positive correlations between anticipated guilt and behavior [13,14,35] and seems to underscore the understanding of anticipated guilt as being tied to a sense of moral responsibility [91,92]. This moral dimension of anticipated guilt can serve as a motivator to encourage individuals to engage in sustainable behaviors such as PEBs, which are often driven by a sense of moral obligation [93,94]. However, more work is needed to assess the long-term effects of guilt appeals on recommended and spillover behaviors [95].

Finally, the four-mediator sequential mediation process predicted by H6 [guilt appeals with lower-magnitude recommended behaviors would lead to greater behavioral intentions via greater perceived self-efficacy (M1), lower feelings of freedom threat (M2), greater psychological reactance (M3), and lower anticipated guilt (M4)] was not supported. This is likely because the predicted relationship between psychological reactance and guilt was not identified, seemingly due to a suppression effect stemming from the influence of freedom threat on guilt. We note that the indirect effect of message condition (lower magnitude/difficulty of recommended behavior) on intentions via greater self-efficacy, lower freedom threat, and greater guilt was significant and positive. Additionally, a direct effect of magnitude of recommended behavior was identified in the full mediation model, suggesting that higher-magnitude recommendations were associated with greater intent to act when all of the mediating variables were controlled for in the model. These two findings seem to point back to the tension described above of whether to “start small” with behavioral recommendations. A higher-magnitude recommendation could lead to more impactful pro-environmental behavior, but, as our mediation analysis indicates, it can also lead to less self-efficacy and more defensive messages responses, which can lead to less behavioral intention or potentially to boomerang effects.

6. Conclusions and Limitations

This study adds to the understanding of the influence of guilt appeals on intention to practice message-recommended sustainable behaviors by further illuminating the mediating roles of self-efficacy and reactance. Particularly, the study offers several important theoretical and practical implications for communicating about sustainability and pro-environmental behaviors. In terms of theoretical contributions, the findings advance the understanding of psychological mechanisms, specifically how self-efficacy and psychological reactance mediate the relationship between guilt appeals and pro-environmental intentions. Our results suggest that recommending smaller magnitude/less difficult behaviors may result in greater intention to act by increasing self-efficacy and reducing defensive message responses, such as freedom threat, and by increasing anticipated guilt. By identifying these mediating factors, the study contributes to the broader literature on messaging about sustainable behaviors. Additionally, the study differentiates between the impacts of guilt and shame in environmental communication. The linear relationship between anticipated guilt and behavioral intention identified herein suggests that appeals that do not elicit significant amounts of shame or reactance can be effective in motivating pro-environmental behaviors. This distinction enriches theoretical discussions by providing a more nuanced understanding of how specific moral emotions might drive behavior changes.

Practically, the study offers valuable insights for designing more effective sustainability advocacy messages. Our findings suggest that utilizing guilt appeals with lower-magnitude behavioral recommendations can reduce psychological reactance and increase the likelihood of individuals engaging in sustainable practices. This provides important implications for public campaigns promoting sustainability, as it offers strategies for designing messages that are more persuasive because they are less likely to evoke audiences’ resistance. Furthermore, the identified linear relationship between anticipated guilt and behavioral intention suggests that crafting emotional appeals that focus on sustainable actions while avoiding eliciting audiences’ feelings of shame can better motivate them to engage in pro-environmental behaviors. Such findings provide evidence-based guidance for practitioners seeking to enhance public engagement in sustainability and environmental conservation efforts while minimizing individuals’ defensive responses. Ultimately, these findings can help practitioners promote more effective and impactful behavior changes, contributing to broader sustainability goals in societies.

The findings of this study should be considered in light of a few limitations. Firstly, we measured anticipated guilt rather than experienced or reactive guilt. Defensive message responses might have been stronger and behavioral intentions weaker if we had measured and included either of the latter two forms of guilt in our model [13]. Additionally, although we identified a positive relationship between anticipated guilt and behavioral intentions, mean reported feelings of anticipated guilt were at a moderate level. Therefore, our method and data might have precluded a test of the linear relationship across a broader range of anticipated guilt and across messages designed to elicit higher levels of guilt. Furthermore, findings should also be considered in light of the limited racial and ethnic diversity of the sample. Cultural variations in values and norms could potentially influence the significance of moral emotions in predicting pro-environmental intentions. Another limitation is that the findings are based on measures of just few environmental behaviors, and their applicability to other types of specific pro-environmental actions remains uncertain. These limitations highlight the need for further research to explore the role of self-efficacy and guilt across diverse contexts to better understand the broader applicability of the findings.