When guilt works: a comprehensive meta-analysis of guilt appeals

Peng, Wei; Huang, Qian; Mao, Bingjing; Lun, Di; Malova, Ekaterina; Simmons, Jazmyne V.; Carcioppolo, Nick

doi:10.3389/fpsyg.2023.1201631

SYSTEMATIC REVIEW article

Front. Psychol., 28 September 2023

Sec. Personality and Social Psychology

Volume 14 - 2023 | https://doi.org/10.3389/fpsyg.2023.1201631

When guilt works: a comprehensive meta-analysis of guilt appeals

$\r\nWei Peng$ Wei Peng¹^*

Qian Huang²^*

Bingjing Mao³

Di Lun⁴

Ekaterina Malova⁵

Jazmyne V. Simmons⁶

Nick Carcioppolo⁷

¹Edward R. Murrow College of Communication, Washington State University, Pullman, WA, United States
²Department of Communication, University of North Dakota, Grand Forks, ND, United States
³TSET Health Promotion and Research Center, Stephenson Cancer Center, University of Oklahoma Health Sciences Center, Oklahoma, OK, United States
⁴Department of Health Disparities Research, University of Texas MD Anderson Cancer Center, Houston, TX, United States
⁵Simon Business School, University of Rochester, Rochester, NY, United States
⁶Division of Health Science, School of Allied Health Sciences, Florida Agricultural and Mechanical University, Tallahassee, FL, United States
⁷School of Communication, University of Miami, Coral Gables, FL, United States

Introduction: Guilt appeals are widely used as a persuasive approach in various areas of practice. However, the strength and direction of the persuasive effects of guilt appeals are mixed, which could be influenced by theoretical and methodological factors.

Method: The present study is a comprehensive meta-analysis of 26 studies using a random-effects model to assess the persuasive effects of guilt appeals. In total, 127 effect sizes from seven types of persuasive outcomes (i.e., guilt, attitude, behavior, behavioral intention, non-guilt emotions, motivation, and cognition) were calculated based on 7,512 participants.

Results: The analysis showed a small effect size of guilt appeals [g = 0.19, 95% CI (0.10, 0.28)]. The effect of guilt appeals was moderated by the theoretical factors related to appraisal and coping of guilt arousal, including attributed responsibility, controllability and stability of the causal factors, the proximity of perceiver-victim relationship, recommendation of reparative behaviors, and different outcome types. The effect was also associated with methods used in different studies.

Discussion: Overall, the findings demonstrated the persuasive effects of guilt appeals, but theoretical and methodological factors should be considered in the design and testing of guilt appeals. We also discussed the practical implications of the findings.

Introduction

Guilt appeals are a type of persuasive communication that evokes feelings of guilt to influence persuasive outcomes (O'Keefe, 2000). In essence, guilt appeals arouse guilt by activating a self-evaluative process through which the perceiver finds their behavior deviating from their moral and social standards (Tangney et al., 2007; Greenbaum et al., 2020). Aroused guilt elicits reparative behaviors toward the victim, such as apologies, compensation, and cooperative behaviors, to undo the harm and reduce conflicts (de Hooge et al., 2011; Chrdileli and Kasser, 2018). For this reason, guilt has been used to promote compliance in the forms of advertisements, health or educational materials, electronic information, and interpersonal conversations (Antonetti et al., 2018; Xu and Guo, 2018).

Empirical research testing the effects of guilt appeals has shown mixed results (Shehu et al., 2013; Noble et al., 2014; Antonetti et al., 2018). The inconsistent findings draw attention to several prominent issues in guilt appeal research. First, guilt induction involves multiple processes, meaning that the antecedents of guilt arousal may vary across different studies (Tangney et al., 2007; Greenbaum et al., 2020). One factor is the attribution of the cause of other's suffering (Tracy and Robins, 2006). Previous studies elicited guilt by focusing on the victim's suffering only (Chang, 2012; Lee, 2013) or by additionally attributing the responsibility to the perceiver (Antonetti et al., 2018). In addition, the relational distance between the perceiver and the victim could affect guilty feelings (Ghorbani et al., 2013). Second, guilt appeals seek to motivate reparative behaviors, but including such information often makes the perceiver feel coerced, resulting in weaker effectiveness (Graton et al., 2016). Third, a survey of studies testing guilt appeals has found a mixture of research methods and interests that vary in topic contexts, induction approaches, and methods of measurement.

Overall, distinct theoretical and methodological factors from past studies increase the uncertainty of guilt appeal research, calling for meta-analytical studies. Previous meta-analyses have demonstrated the effects of guilt appeals (O'Keefe, 2000; Boster et al., 2016; Xu and Guo, 2018; Turner and Rains, 2021) but had several limitations. Not all meta-analyses systematically accounted for various processes involved in guilt appeals under a consistent theoretical framework. Abbreviated investigations of non-theoretical moderators have been conducted so far. As a result, the variations of guilt induction and coping were not examined. Moreover, heterogeneity in methodological approaches, including but not limited to contexts, induction methods, and measurement in extant studies on guilt appeals, had not been fully evaluated.

Therefore, the present meta-analysis aimed to achieve three goals. First, this study sampled a larger body of guilt appeal studies to investigate the overall effect size. Second, the study centered on the emotional mechanism of guilt appeals to evaluate relevant theoretical moderators that had not been examined before. Third, it investigated the moderating effects of methodological differences. The following section is organized into two parts. The first part examines the theoretical framework and mechanism of guilt appeals, which guides the review of theoretical moderators related to the processes of guilt induction and coping. The second part is a review of key methodological moderators—contexts, induction methods, narratives, modalities, and time points of measurement. A conceptual model of the present meta-analysis is available in Figure 1.

FIGURE 1

Figure 1. The conceptual model of the meta-analysis of guilt appeals with theoretical and methodological moderators.

Theoretical framework and moderators of guilt appeals

Guilt appeals arouse guilt as a discrete emotion to persuade (O'Keefe, 2000). There is substantial agreement that human emotions are adaptive responses of an organism to stimuli in the external environment (Mulligan and Scherer, 2012; Roseman, 2013). Each discrete emotion is aroused by a unique pattern of appraisals that evaluates the external stimulus with regard to different dimensions, such as motivational relevance, goal congruence, expectedness, control, and agency (Smith and Ellsworth, 1985; Moors et al., 2013; Moors, 2020). The perceiver also appraises the coping ability to reduce the demands from emotions arising from this event (Roseman, 2013; Scherer and Moors, 2019). This coping appraisal prompts mental and behavioral strategies to meet the needs of the personal goal and regulate the aroused emotions (Roseman, 2013; Scherer and Moors, 2019). Because the adaptive responses of emotions can change motivational priorities to address the causal factors, theorists reason that emotions can cause biased estimation of the possibility and desirability of an object, leading to changes in decision-making processes (Nabi, 2003; DeSteno et al., 2004; So et al., 2015).

Based on the appraisal theories, guilt appeals should be designed with two pivotal components: guilt induction and coping (Nabi, 1999; O'Keefe, 2002). In the first step, guilt appeals induce guilt in the perceiver by guiding their cognitive appraisals to evaluate specific dimensions of the stimulus event (O'Keefe, 2002). Once feeling guilty, a perceiver evaluates and adopts possible strategies to control the emotion (Dearing et al., 2005; de Hooge et al., 2011; Behrendt and Ben-Ari, 2012). Thus, in the second step, guilt appeals can recommend specific actions of the perceiver to regulate guilty feelings (O'Keefe, 2002). Based on the framework of discrete emotions, variations in theoretical factors related to guilt induction and coping can influence the effects of guilt appeals.

Guilt induction as theoretical moderators

Guilt induction is the first component of guilt appeals. Theoretical models conceptualize guilt as a self-conscious emotion elicited by the inconsistency between one's self-representation and moral or social standards (Tracy and Robins, 2006; Tangney et al., 2007). Guilt is associated with complex concepts of self and socialization needs, such as maintaining a friendship, as opposed to survival instincts (de Hooge et al., 2007; Cryder et al., 2012). We review several key factors of guilt induction, including attribution dimensions associated with responsibility, controllability, stability, and the proximity of the perceiver–victim relationship (Tracy and Robins, 2006; Tangney et al., 2007; Ghorbani et al., 2013).

Attribution

Attribution is an essential part of the appraisal in guilt elicitation (Weiner et al., 1982). In the attribution process, the perceiver evaluates the cause of the victim's suffering presented in guilt appeals regarding its responsibility, controllability, and stability (Tracy and Robins, 2004). From the perspective of the persuader, how the cause is presented and perceived concerning these three attribution dimensions is critical to eliciting guilt. The first dimension of attribution is responsibility. Emotion research shows that guilt arises from the sense of responsibility for inflicting harm on others (Tangney et al., 2007; Roberts et al., 2014; Miceli and Castelfranchi, 2018). Responsibility plays an essential role in the social and moral function of guilt as it motivates people to acknowledge their mistakes and take reparative actions in the aftermath of transgressions (Tangney et al., 2007). To activate the appraisal of responsibility, guilt appeals blame the perceiver as the cause of others' suffering in a specific social or moral context (Harth et al., 2013; Antonetti et al., 2018).

However, responsibility is not always a necessary condition of guilt arousal. Existential guilt is associated with a person's broader perception of their meaning and expectations in a human society (Binder, 2022). This type of guilt is often evoked by moral responsibility or social obligation, which does not result from an actual transgression one commits but from one's perceived failure to fulfill their meaning and duties in society (Zimmermann et al., 2011). Thus, existential guilt extends beyond mere compensation and punishment as described by Tangney et al. (2007) and could motivate people to engage in broader issues. Others contend that existential guilt is possibly driven by empathy, which links one with others in a social relation, to reduce others' suffering (Baumeister et al., 1994; Roberts et al., 2014). Thus, people would feel guilty about inequality impacting others, even though they did not cause its occurrence (O'Connor et al., 2000; Harth et al., 2013; Dull et al., 2021). For example, guilty feelings are experienced by the survivors of natural disasters (Hugelius et al., 2017; Dang et al., 2022), workplace layoffs (Karfakis and Kokkinidis, 2018; Mujtaba and Senathip, 2020), or cancer (Perloff et al., 2019). For this reason, without assigning responsibility, guilt appeals can still be persuasive by evoking the perceivers' role in a social system and their behaviors to address broad social issues (e.g., climate change, equality, and poverty) (Chang, 2011).

The other two attributional dimensions are stability and controllability (Tangney et al., 2007). The stability dimension is crucial to distinguish guilt from other moral emotions, such as shame (Tracy and Robins, 2004). Compared with shame, guilt arises from one's attention to the unstable aspect of the self, such as behaviors, as opposed to one's personality, character, and other stable aspects (Greenbaum et al., 2020). To influence the appraisal of stability, guilt appeals provide evaluation information showing that the nature of a causal event is changeable (Xu and Guo, 2018). The attribution to unstable causes is often achieved by focusing on a specific incident or behavior of the perceiver as the cause of the victim's suffering (Roberts et al., 2014).

Moreover, to influence the appraisal of controllability, guilt appeals attribute the victim's suffering to a cause over which the perceiver has control (Antonetti and Maklan, 2014). As a result, the perceiver can develop a stronger sense of agency over the process and consequences of the causal event (Roberts et al., 2014). A theoretical review explains that the action taken by the self is an integral part of morality, which is linked to guilt as a moral emotion (Jennings et al., 2015), that is, in addition to the person's identity and environment, the actions a person takes to influence society and self-shape the characteristics of their morality (Jennings et al., 2015). On the other hand, controllability should be distinguished from stability attribution in the appraisal of guilt (Tracy and Robins, 2004). While a cause is attributed to be unstable, a person might not feel guilty if the locus of control is external (Antonetti and Maklan, 2014). Thus, the effects of guilt appeals might depend on whether the guilt appeals attribute a negative event to a cause that is controlled by the perceiver.

The proximity of perceiver–victim relationships

The elicitation of guilt can vary by the proximity of the relationships between the perceiver and the victim (Baumeister et al., 1994). Transgressors usually have a stronger feeling of guilt when they are interpersonally close to the victims, such as friends or family (Ghorbani et al., 2013). Because guilt involves one's self-representation in a moral or social context, people can more easily appraise the effects of their misdeeds on others and empathetically feel their suffering if there is a close connection (Tangney et al., 2007; Ruckstaetter et al., 2017). In addition, transgressors can be motivated to reconcile their transgressions due to the concern about alienating those in close relationships (Ghorbani et al., 2013).

Additionally, a personal relationship is not always required for guilt induction. Guilt can be elicited by the wrongdoings committed by in-group members onto others (Iyer et al., 2007; Zimmermann et al., 2011). For example, people felt guilty about their country's actions toward disadvantaged groups in the Dutch colonization of Indonesia (Doosje et al., 1998) and the Israeli–Palestinian conflicts (Roccas et al., 2006). In response to the finding, Doosje et al. (1998) explained that transgressions committed by in-group members might threaten the perceiver's social identity and arouse collective guilt. In addition, individuals may develop a sense of community toward workplaces or other institutions (Nowell and Boyd, 2014). Guilty feelings might result from the perception that one had not responded to the misfortune of their community (Zimmermann et al., 2011).

Thus, we seek to understand whether the proximity of the relationship between a perceiver and a depicted victim moderates the effectiveness of guilt appeals. When appraising a negative event, a perceiver may react differently after assessing whether the victim is in a close relationship. Hence, the proximity of the perceiver–victim relationship will be tested as a theoretical moderator.

Guilt coping as a theoretical moderator

There is a clear consensus among emotion theories that negative human emotions must be regulated by changing the causal event directly or releasing the mental state from its influence (Lazarus, 1991; Ellsworth and Scherer, 2003; Roseman, 2013). Once feeling guilty, the perceiver seeks and evaluates possible strategies to cope with the emotion (Tracy and Robins, 2006; Tangney et al., 2007). In this process, following the guilt induction process, guilt appeals provide the perceiver with a certain action or counterfactual thinking to remedy guilt-arousing causes (O'Keefe, 2002; Antonetti et al., 2015). Thus, the meta-analysis will test the following two moderators to understand how the effects of guilt appeals are associated with the regulation of aroused guilt.

Recommendation of reparative behaviors

Guilt appeals in various contexts provide immediately accessible information to help the perceiver take measures directly against the source of the negative feelings (Agrawal and Duhachek, 2010; Antonetti et al., 2015). Extensive research has shown that guilt motivates reparative behaviors to amend wrongdoings and undo the harm to others (Ketelaar and Tung Au, 2003; Tangney et al., 2007; Shen, 2018; Vaish, 2018). The reparation process helps transgressors control their guilty feelings (Tangney et al., 2007). Confession, apology, cooperative, and prosocial behaviors are among the most common reparative behaviors in social interactions (Tangney et al., 2007; de Hooge et al., 2011; Behrendt and Ben-Ari, 2012). For example, Ilies et al. (2013) found that in the workplace, employees who felt guilty about their counterproductive work behaviors were more likely to offer helping behaviors. In interpersonal conflicts, guilty feelings prompt people to engage in cooperative approaches (Behrendt and Ben-Ari, 2012). Other studies suggest that apologies were related to guilt, but not to shame (Ruckstaetter et al., 2017; Chrdileli and Kasser, 2018).

For this reason, guilt appeals might be more persuasive if recommending reparative actions can immediately satisfy the perceiver's need to make amends (Graton and Mailliez, 2019). In addition, guilt appeals with reparation suggestions convey the repairable nature of the events for which people feel guilty (Tangney et al., 2007), that is, the negative consequences of someone's wrongdoings can be counteracted by the recommended actions in guilt appeals (Graton et al., 2016). If no actions or remedies are provided in guilt appeals, the perceiver might feel that the negative consequences cannot be offset, and their recovery from guilt is inhibited (LeBlanc et al., 2020). In turn, people would seek to deal with the negative feelings directly without changing the cause (Antonetti and Baines, 2015). As a result, guilt appeals may be avoided or rejected (Graton et al., 2016).

However, explicit information about reparative behaviors does not necessarily strengthen the effectiveness of guilt appeals. In a series of experiments, Graton and Ric (2017) found that guilty feelings have increased attention to and positive attitudes toward implicit cues of making amends. Additional demands for reparative behaviors, such as asking for charitable donations, could make people feel coerced. These findings bring uncertainty into whether recommending reparative behaviors in guilt appeals can influence persuasive effectiveness.

Outcomes as a theoretical moderator

The process of guilt appeals may vary outcomes, resulting in variations in the effect size of guilt appeals. The first outcome is guilt, which often serves as a validity indicator of guilt appeals (O'Keefe, 2000). Guilt is a mediating process of guilt appeals before more distal persuasion outcomes. Through the elicitation of guilt, guilt appeals engender persuasion resulting from the perceiver's motivation to control the guilty feelings (Tangney et al., 2007; Allard and White, 2015). Guilt elicited by guilt appeals is thus expected to be consistent and strong. However, if the stimulus event framed in a guilt appeal is relatively more important to the perceiver's goals other than maintaining the social or moral dimension of self-identity, more ambiguous or other emotions may arise, reducing the intensity of guilty feelings (Tracy and Robins, 2006; Tangney et al., 2007).

Attitude is another essential outcome of guilt appeals. Attitude is generated from an appraisal process in which the perceiver evaluates the attributes of the persuasion object (Fishbein and Ajzen, 1975). While early theoretical work primarily focused on cognitive antecedents of attitudinal assessments, more recent studies show that emotional information is an essential dimension of the evaluations for attitude (Fishbein and Middlestadt, 1995; Zanna and Rempel, 2008). Guilt appeals influence attitudes by providing information that can cause biased estimation of the value and probability of specific attributes associated with a persuasion object (DeSteno et al., 2004). Variations of attitude depend on whether certain attributes of the object are appraised as potentially helpful to cope with the aroused guilty feelings and attain one's goal related to social and moral issues (Allen et al., 2005; Antonetti and Baines, 2015). The desirability of the persuasion object can also be lower if the guilt appeals elicit other emotions or conflict with incidental emotions (DeSteno et al., 2004).

Behaviors and behavioral intentions are considered focal outcomes of guilt appeals. Behavior is at the center of guilt experiences. Guilt is elicited by specific wrongdoings and the regulation process is often achieved by reparative behaviors (Tangney et al., 2007; Roberts et al., 2014). As an emotion regulation mechanism, behavioral actions are taken by individuals to address the problems that cause emotions (Scherer and Moors, 2019). In this sense, guilt appeals suggest specific behavioral actions that can undo the harm to the victim and reduce guilty feelings (Tangney et al., 2007; Antonetti and Baines, 2015). Thus, guilt appeals are often used in persuasion to gain behavioral compliance (Agrawal and Duhachek, 2010; Chang, 2012). If behavioral outcomes provide little help to control negative feelings, maladaptive guilt experiences can result in rejection or avoidance of the entire message and recommended behaviors (Kim et al., 2011; Antonetti and Baines, 2015).

Lastly, guilt is expected to change motivation and cognition (Bradley et al., 2001; Pessoa, 2009). Aroused guilt disrupts thoughts and directs cognitive resources to fix eliciting events and control aversive feelings (Tangney et al., 2007; Scherer and Moors, 2019). Motivation is an intermediate process between emotional arousal and cognitive or behavioral responses. Emotion is aroused when a person's motivation-related goal is satisfied or at risk (Roseman, 2013). In response, motivation prioritizes specific cognitive resources (e.g., attention) and executive functions (e.g., behaviors) to act toward the goal (Pessoa, 2009). Thus, guilt can motivate a person to engage in reparation behaviors that help maintain the goal related to one's self-representation or interpersonal relationships (Tangney et al., 2007; Greenbaum et al., 2020).

Methodological moderators of guilt appeals

Methodological differences should be assessed to understand the varying effects of guilt appeals. The effects of guilt appeals can be influenced by different contexts (Massi Lindsey, 2005; Chang, 2012), induction methods (e.g., Allard and White, 2015; La Ferle et al., 2019), or outcome variables (Noble et al., 2014). Below, we review methodological factors related to the effectiveness of guilt appeals.

Contexts of persuasion objects

Different contexts of persuasion objects can account for variations in effects in that each context may result in different appraisal patterns (e.g., causal attribution, reparative behaviors, and personal relevance). Commonly seen in advertising and marketing, guilt appeals are used to promote a product or service by provoking the perceiver's examination of self-representation against a person's goal and claiming how the persuasion object (e.g., a product) helps them reach the goal (Antonetti and Baines, 2015). Guilt appeals can also change self-efficacy and behaviors in health and medical contexts (Xu and Guo, 2018). Guilt appeals in a health-related context seek to portray certain health behaviors as wrongdoing that violates socially acceptable rules and generates negative impacts on people in social relationships, such as the harmful effects of human papillomavirus infection on a sexual partner (Babin, 2009). Another important context is social issues with broad impacts, such as environmental protection (Baek and Yoon, 2017; Schneider et al., 2017) and charity giving (Chang, 2011; Xu, 2022). Guilt feelings arise when people consider themselves not fulfilling their moral and social responsibility and these causes (Zimmermann et al., 2011; Urbonavicius et al., 2019). Guilt appeals in this context motivate prosocial and altruistic behaviors to alleviate guilty feelings (de Hooge et al., 2011; Urbonavicius et al., 2019).

Induction methods

Two methods of guilt induction commonly used in previous research could influence the effects of guilt appeals. One method is based on persuasive messages, which provide a brief piece of information that matches the core appraisal theme of guilt to arouse guilty feelings in the perceiver (Nabi, 1999, 2003). For instance, a guilt appeal message portrays the victim's suffering and frames the perceiver's controllable behavior as the cause. The other method uses recall and imagination, commonly through verbal instructions in interpersonal communications (Tracy and Robins, 2006). The perceiver is asked by the persuader to recall a salient incident in their lives that once aroused guilt and then to re-experience it in the current setting (Allard and White, 2015; Graton et al., 2016).

The influence of induction methods on guilt appeals might center on the relevance of information. According to the self-reference effect, information about oneself is clearly organized in memory and easily accessible (Block, 2005). Despite its popularity, message-based guilt appeals can be perceived as less relevant to the perceiver when describing the suffering of an unknown party (Ghorbani et al., 2013). As a result, the effects of guilt appeals may be hampered (Chang, 2012; Ghorbani et al., 2013). In this sense, recall might be advantageous to elicit self-conscious emotions like guilt (Tracy and Robins, 2006; Siedlecka and Denson, 2019). Through recall, the perceiver can reproduce the emotional experience and examine whether certain aspects of themselves have caused negative impacts on others (Tracy and Robins, 2006). Thus, this study will test whether induction methods moderate the effects of guilt appeals.

Narrative

A growing body of literature points to emotional experiences as the foundation of narrative persuasion (Green and Brock, 2000; Nabi and Green, 2015; Liu and Yang, 2020). Using narrative in guilt appeals has also been tested in the literature. Past studies have narratives to evoke feelings of guilt by providing the victim's stories or by engaging the perceiver in summoning personal experiences (Graton et al., 2016; Antonetti et al., 2018, study 1b). In contrast, non-narrative guilt appeals typically use reasoning, logical arguments, and statistical information (Chang, 2012; Antonetti et al., 2018, study 1a). However, the overall effects of narrative vs. non-narrative persuasion are mixed (Shen et al., 2015; Freling et al., 2020). We seek to test whether using narratives in guilt appeals is associated with stronger persuasiveness.

Modalities

Guilt appeals can benefit from different modalities (e.g., text, audio, video, or mixed modes), but the effects may vary. According to multimedia learning theory (Mayer and Moreno, 2003; Mayer, 2014), information in multiple modalities increases attention to essential content, adds visual cues to exemplify the central theme, and reduces the cognitive burden through redundant information. Visual content may build mental experiences through which a person can experience the if-then scenario in guilt arousal (Chien and Chang, 2012; Antonetti et al., 2015). Thus, it is reasonable to argue that these benefits of multimedia information could support the processing of guilt appeals to achieve the intended effects. However, multimedia information with a poor design might also distract a perceiver from scrutinizing the claims in guilt appeals (Mayer, 2014). For example, irrelevant visual information might cause incidental processing and unintended effects (Sung and Mayer, 2012). Given limited evidence to inform the most effective modality for guilt appeal, the present meta-analysis tests whether the effectiveness of guilt appeals varies by different modalities.

Time points of measurement

The effects of guilt appeals may depend on different time points of measurement after persuasion occurs. Most existing studies measured the effects of guilt appeals immediately after exposure (Peloza et al., 2013; Graton et al., 2016) because emotional responses are often transient (Brosch, 2021; Ogunfowora et al., 2023). Nevertheless, recent research suggests that immediate measurement might not capture the desired effects of guilt appeals as people might delay their decisions to adopt the suggested behavior in persuasion (Antonetti et al., 2018; Brosch, 2021; Kemp, 2023). Thus, the present meta-analysis will understand how different measurement time points can cause variation in the observed persuasive effects of guilt appeals.

The present meta-analysis

The persuasive effects of guilt appeals have been synthesized by past meta-analyses, but several gaps remain to be addressed. Both O'Keefe (2000, p. 74) and Boster et al. (2016) included guilt appeal studies based on the transgression-compliance mechanism and found medium effect sizes (rs = 0.28 and 0.26, respectively). However, both studies presumed based on a priori knowledge that guilt would drive behavioral compliance of transgressors. Thus, their findings should not be interpretable as the effects of guilt appeal because transgressors' behavioral compliance is not necessarily explained by guilt (Cialdini et al., 1973; Baumeister et al., 1994). The other three meta-analyses studied guilt appeals that intentionally aroused guilt emotion for persuasive outcomes but suffered from multiple limitations. First, O'Keefe (2000, p. 80) compared five guilt appeal studies with varying explicitness and found that more explicit guilt appeals were associated with less persuasive effects (r = −0.26). However, this meta-analysis did not explicate the processes related to the intensity of guilt induction. Second, Xu and Guo (2018) demonstrated moderate effects of guilt-based persuasive messages but their findings were limited to health-related outcomes. In addition, several included studies manipulated persuasive constructs that were not intended to evoke guilt (e.g., norm perception, Lee and Paek, 2013). Furthermore, their analysis only examined a limited number of moderators, including message modality, self/other focus, and health behaviors. In a more recent meta-analysis, Turner and Rains (2021) found a small effect size of guilt appeals on attitudes or behavioral intentions and medium effects on guilt and anger. However, similar to previous meta-analyses, moderators associated with the appraisal dimensions of guilt arousal, coping process, and outcomes were not of primary interest. Their study did not account for heterogeneity in outcomes of guilt appeals.

The present study hence aimed to conduct a comprehensive synthesis of guilt appeals. There are four strengths of this study. First, we examined the studies that intentionally evoked guilt as an emotional mechanism of persuasion. Persuasion studies that manipulated the constructs not intended to evoke guilt were excluded. Thus, this meta-analysis provided a more robust and accurate estimate of the effectiveness of guilt appeals. Second, guided by appraisal theories of emotion, we systematically accounted for key factors associated with guilt induction and coping. The absence of such inquiries in previous studies could contribute to inconsistent results (O'Keefe, 2002; Tangney and Dearing, 2002). Thus, examining these factors would offer important insights into variations in guilt arousal and persuasive outcomes. Third, previous meta-analyses primarily assessed the effects on attitude and behaviors. In contrast, our study evaluated additional dependent variables to assess the overall effects of guilt appeals thoroughly. Fourth, we evaluated various methodological characteristics of past guilt appeal studies. As the body of guilt appeals expands, this analytic approach can critically address the limitation of previous meta-analyses in which methodological differences were not of focus.

The first goal of this meta-analysis is to understand the overall effectiveness of guilt appeals. To this end, we proposed the following research question.

RQ1: Compared with equivalent neutral true control conditions, do guilt appeals generate greater persuasive effects?

Also, as reviewed, theoretical factors involved in guilt induction and coping could influence the effectiveness of guilt appeals. The following hypotheses were proposed to test whether these factors moderated the effects of guilt appeals.

H1: The effects of guilt appeals are moderated by the presence of attribution dimensions related to (a) responsibility, (b) controllability, or (c) stability in guilt induction.

H2: The effects of guilt appeals are moderated by the proximity of the relationship between the perceiver and the victim in guilt induction.

H3: The effects of guilt appeals are moderated by the recommendation of reparative behaviors to cope with aroused guilt.

H4: The effects of guilt appeals are moderated by different types of outcomes.

Across a large body of past research, various method approaches to guilt appeals may influence the effectiveness of guilt appeals. The following research question is proposed to understand which methodological factors moderate the effects of guilt appeals.

RQ2: Are the effects of guilt appeals moderated by (a) context of persuasion objects, (b) induction methods, (c) narrative, (d) modalities, and (e) time point of measurement?

Methods

Search process

We used the Population, Intervention, Comparator, Outcomes, and Study (PICOS) design model in accordance with the PRISMA guideline (Moher et al., 2009) to develop the search strategy. The details of our PICOS criteria are provided in Table 1.

TABLE 1

Table 1. The PICOS (populations, interventions, comparators, outcomes, study design) inclusion criteria and initial article screening results.

An initial literature search was conducted in 2020 through the following databases: PsycINFO, ProQuest Dissertations & Theses Global, ERIC, MEDLINE, and Communication and Mass Media Complete via EBSCO. The search terms “Guilt^*” AND [“appeal” OR “campaign” OR “adverti^*” OR “message” OR “intervention”] were used. Publicly available doctoral and master's degree theses were also retrieved. A total of 1,240 citations were exported to a citation manager. The initial sample included studies published from 1929 to 2020. To avoid file drawer issues, authors and experts in the field were consulted using the listserve “Communication Research and Theory Network's (CRTNET)” to solicit unpublished studies in their personal libraries, and responses were accepted throughout 2020. This procedure yielded no additional articles. After removing duplicates (n = 72), 1,168 articles remained in the initial screening phase.

Inclusion and exclusion criteria

Figure 2 provides a PRISMA flowchart indicating the overall inclusion and exclusion procedure. Inclusion and exclusion criteria were developed and verified based on PICOS. The screening results are summarized in Table 1. In the first round, two coders screened the abstracts of 1,168 articles (n = 1,168) and excluded (a) five articles (n = 5) not written in English (n = 5); (b) the comparator criterion: 1,014 articles reported studies that compared the effects of guilt with other emotions, mixed emotions, or variants of guilt without a true control condition (n = 1,014); (c) the outcomes criterion: four articles reported studies that measured guilt as the only outcome variable (n = 4); and (d) the study design criterion: 92 studies reporting non-quantitative studies only (n = 92). A total of 53 articles remained after the first round of screening.

FIGURE 2

Figure 2. PRISMA flow chart.

In the second round of screening, the coders further assessed the full text of the remaining articles based on the PICOS criteria and removed the following articles: (1) the population criterion: no articles were excluded (n = 0) due to no criterion placed on demographic characteristics; (2) the interventions criterion: (a) four articles reported no studies that manipulated guilt appeals as an independent variable (n = 4) and (b) seven articles reported studies that manipulated different levels or subtypes of guilt (e.g., strong or naturalistic guilt) (n = 7); (3) the comparators criterion: (a) no articles were excluded due to not reporting at least one guilt appeal condition and one emotion-neutral control condition (n = 0); (4) the outcomes criterion: one article with studies that only reported the outcomes based on indirect rating or observation results by non-participant judges (n = 1); and (5) the study design criterion: (a) one article reported studies that did not use a between-subject design with a guilt appeal and a neutral control condition (n = 1).

In addition, the articles that reported no usable statistical results for effect size calculation were excluded, including (a) 20 articles with studies that provided no statistical results used to calculate effect sizes (i.e., mean and standard deviation, F-test values, t-test values, Cohen's d, and Pearson's r values) (n = 20), and (b) two articles (n = 2) with studies that only reported omnibus F-tests for the comparison of multiple conditions (e.g., F-tests of guilt, shame, hope, and control conditions) without reporting mean and standard deviation of guilt and neutral control conditions. Studies reported only unstandardized coefficients from multiple regression were excluded because effect sizes computed from these coefficients would be partial and under the influence of other predictors in the multiple regression (Hunter and Schmidt, 1990). Thus, coefficients did not reflect the meaningful effects unless corresponding correlation coefficients were obtained (Peterson and Brown, 2005).

A final sample of 18 articles yielding 26 studies was included in further analyses. The references of all included articles are listed in Appendix C.

Coding procedure

Appendix A lists the coded variables and their operational definitions. In the present meta-analysis, we also coded demographic and descriptive information as situational moderators of the included studies (i.e., publication type, sampling frame, funding source, study location, and year of publication). The studies were coded independently by two coders. The coders were graduate students who had completed courses and conducted relevant research projects related to persuasion and emotions. Each coder was responsible for half of the studies. Seven studies (27% of all selected studies) were sampled to compute the intercoder reliability. Intercoder reliability was computed with Krippendorff (2004) alpha (α) for categorical variables and intraclass correlation (ICC) for continuous variables (Shrout and Fleiss, 1979). Intercoder reliability of coded variables ranged from 0.74 to 1.00 (α bar = 0.95, ICC = 0.87) within the category of agreement (Landis and Koch, 1977). Intercoder reliability was also assessed during the coding of the full sample (Lacy and Riffe, 1996; Neuendorf, 2017). All rating disagreements were resolved through discussion until 100% agreement was achieved for each variable.

Calculation of effect size

Appendix B explains how effect sizes were calculated. Because studies of relatively small sample sizes were included, Hedges' g was applied to indicate effect size, correcting for small sample sizes (Hedges, 1981; Hedges and Olkin, 1985). A positive Hedges' g indicated an increase of the effect on study outcomes was elicited by guilt appeals compared to an equivalent non-guilt or true control condition; and a negative Hedges' g indicated a decrease of the effect was elicited by guilt appeals compared to an equivalent non-guilt or true control condition. For studies where the guilt appeal was linked to a reduction in outcome variables for which a greater score indicated negative effects, the computed sign of the effect size was reversed, so all positive differences reflected an increase in the dependent variables.

Overall analysis and outliers

Outliers are frequently present in meta-analyses and can distort pooled effect estimates of final effects; therefore, outliers were detected and removed from analyses (Hedges and Olkin, 1985; Schmidt et al., 1993; Viechtbauer and Cheung, 2010). We determined an effect size as an outlier if its confidence interval did not overlap with the confidence interval of the pooled effects. Such an effect size was less likely to be part of the population of pooled effects by the present meta-analysis (Viechtbauer and Cheung, 2010). Using the R package “altmeta” (Lin and Chu, 2016), we identified two outliers (i.e., 1.87 and 2.21), which were from Allard and White (2015; Studies 2 and 5).

The removal of these outliers yielded 127 effect sizes from 26 studies that examined the effects of guilt appeals with multiple measurements in the current meta-analysis. For most studies included, we extracted more than one effect size. Considering potential interdependency among effect sizes from the same study, we conducted all moderation analyses using the Robust Variance Estimation (RVE; Gleser and Olkin, 2009; Hedges et al., 2010; Pustejovsky and Tipton, 2022) for adjusting for effect size tendencies while retaining appropriate Type I error rates. All analyses were performed in R (version 4.2.1). We used the “robumeta” package (Fisher and Tipton, 2015) to perform the overall and moderator analyses, including meta-regression modeling. All test statistics (i.e., Q values) were conducted using the “metafor” package (Viechtbauer, 2010). Pairwise comparisons were performed using the “ClubSandwich” package (Pustejovsky and Tipton, 2018, 2022; Pustejovsky, 2020) and “multicomp” package (Bretz et al., 2010). Appendix D of the Supplemental material lists the acronyms in this article.

Results

Description of studies

A total of 127 effect sizes from 26 studies (N = 7,512) were included in the meta-analysis. Sample sizes ranged from 61 to 1,054 (M = 288.90, SD = 261.22). Participants' mean age ranged from 19.34 to 54 years (M = 28.68, SD = 11.65).

Publication bias

The rank correlation test (Kendall's tau) of Begg and Mazumdar (1994) suggested a non-significant relationship between standard error and effect size (Kendall's tau = 0.08, p = 0.17), indicating no publication biases. In addition, Egger's regression (Sutton, 2009) test showed a non-significant relationship between standard error and effect size (z = 0.24, p = 0.81), which indicated no evidence for publication bias. The trim and fill method of Duval and Tweedie (2000) suggested that zero studies were missing on the right side of the funnel plot (SE = 6.43). The graphical evaluation of the funnel plot (Figure 3) shows a symmetrical inverted shape, indicating that the effects of the included studies were variable and scattered to both sides of the point estimate of effect. Thus, the appearance of the funnel plot suggests no missing studies, either due to publication bias or sampling errors. Rosenthal's (1979) fail-safe N was 6,035 at the observed significance level of 0.0001 and the perceived significance level of 0.05. We thus concluded that it is unlikely that this meta-analysis missed a substantial number of studies. Given these results, we concluded that potential publication biases were absent in extant literature examining the effects of guilt appeals. A funnel plot with trim and fill and Egger's regression is provided in Figure 3.

FIGURE 3

Figure 3. Funnel plot with trim and fill and Egger's regression.

Overall analysis

Excluding outliers, the overall effect size of guilt appeal produced on all types of outcomes was significant, Q(126) = 902.31, t(24.2) = 3.11, I-squared = 91%, tau-square = 0.24 (SE = 0.04), with g = 0.19, SE = 0.05, rho = 0.8, 95% CI (0.10, 0.28), p < 0.001. The results indicate the heterogeneity in true effect sizes of guilt appeal (Higgins et al., 2003). Thus, we used the random-effects model in the analysis. Providing supportive evidence to address RQ1, this finding suggested a small effect size of guilt appeals. In contrast, the overall effect size of guilt appeal, including outliers, was 0.24, p < 0.001, suggesting that having outliers in subsequent analyses would inflate the overall effect size. The post-hoc statistical power of the overall analysis was 80%. Thus, without additional studies, the sample size was large enough to detect the true effect sufficiently.

Analysis of moderators

All moderator analyses were conducted for studies with the removal of outlier estimations. We reported the results of individual moderators. In addition, a meta-regression model with theoretical moderators as the predictors was also specified to explain the effects of guilt appeals (see details in Tables 1–3).

Analysis of theoretical moderators

Responsibility attribution

Responsibility attribution moderated the effects of guilt appeals. H1a was supported. Guilt appeals that attributed responsibilities to the perceiver produced smaller effects compared to those not attributing responsibilities (Table 2).

TABLE 2

Table 2. Effect sizes by theoretical moderators (s^a = 26, k^b = 127).

Controllability attribution

The meta-analysis supported H1b, indicating that controllability attribution moderated persuasive effects on outcomes. Guilt appeals depicting controllable causes generated smaller effects than those not depicting controllable causes.

Stability attribution

Stability attribution moderated the effects of guilt appeals, supporting H1c. Guilt appeals attributing the victim's suffering to an unstable cause produced greater effects on outcomes than did those not depicting such causes.

The proximity of perceiver–victim relationship

The proximity of the perceiver–victim relationship moderated the persuasive effects of guilt appeals, supporting H2. Guilt appeals featuring the victim in a personal relationship with the perceiver or being a non-human subject generated greater persuasive effects than those featuring victims in other relationships.