Abstract
Abstract
Meta-analytic procedures were used to estimate the effect of experienced guilt on compliance. Examination of 47 effect sizes indicated that inducing guilt is an effective means by which to increase compliance, ρ = .26. Moreover, despite coding for numerous substantive and methodological moderators, there was no evidence of moderation in these data. Instead, correcting for measurement error in the independent variable and restriction in range in the dependent explained all variance in effect sizes, yielding a corrected effect size of ρ′ = .35.
Keywords:
Disciplines ranging from anthropology (Benedict, 1946) to communication (Boster et al., 1999), economics (e.g., Smith, 1759/2002), philosophy (e.g., Kant, 1785/1959), psychology (e.g., Freud, 1930/2005), and theology (e.g., Niebuhr, 1932) have addressed the etiology, essence, and effects of guilt. Consequently, treatments of the topic differ considerably.
Narrowing the issue to the study of the effect of guilt on compliance renders the topic more tractable by focusing the discussion of guilt’s causes, character, and consequences. Even within this more limited focus, however, there is almost a 50-year history of research that has addressed the effect of guilt on compliance. Moreover, although investigations presently continue to explore this relationship, there is no summary of the available literature. This paper addresses that lacuna by presenting a meta-analytic review of this relationship.
Like shame, fear, and anger, guilt can be characterized as an unpleasant emotional state. Distinguishing it from these other emotions, however, are guilt’s unique antecedents and consequents. Regarding the former, guilt is believed to result from accidental transgressions, so that harming another or allowing another to come to harm produces guilt (Baumeister, Stillwell, & Heatherton, 1994; Tangney & Dearing, 2002).1,2 Pertinent to the latter, because guilt is unpleasant, people either seek to anticipate it and then act in ways that avoid inducing it or try to undo harm when it has occurred.
According to the negative state relief model (NSRM, Cialdini, Darby, & Vincent, 1973), induced guilt will also result in a higher probability, or magnitude, of compliance (see Figure 1). As this diagram indicates, guilt inductions are expected to affect experienced guilt, such that those targets in a guilt or transgression condition experience more guilt than those in no guilt or no transgression control condition. In turn, higher levels of experienced guilt lead to a higher likelihood (or greater magnitude) of compliance with others’ requests.3 Because opportunities to comply with such requests allow guilty parties a means of providing restitution, they also provide a means of reducing or eliminating the negative affect associated with committing the transgression.
Published online:
18 February 2016![]({"type":"image","src":"/na101/home/literatum/publisher/tandf/journals/content/psif20/2016/psif20.v011.i01/15534510.2016.1142892/20170209/images/medium/psif_a_1142892_f0001_b.gif"}){kind=small}
Figure 1. The negative state relief model (based on *Cialdini et al., 1973).
So, for example, Cryder, Springer, and Morewedge (2012) conducted an experiment in which they asked subjects to choose between eating either red apple-flavored or vomit-flavored jellybeans, with almost all of them choosing the red apple flavor. Subjects were also given a series of instructions about the task before making their choice, but, because the instructions were in small print and were overly detailed, most subjects did not read them carefully. Subsequently, those in the guilt condition learned that the instructions had dictated that a co-participant be required to eat the remaining flavor—the vomit-flavored jellybeans. Those in the no guilt control condition, on the other hand, learned that instructions dictated that the co-participant be required to eat the same flavor—the red apple-flavored jellybeans. In other words, subjects’ failure to read the instructions carefully in the guilt condition resulted in harm to the other (by forcing him or her to eat vomit-flavored jellybeans), thus producing guilt. No harm was caused in the control condition, so no guilt would be expected to result. Indeed, when subjects engaged in a dictator game with the co-participant, guilty subjects gave more money to a partner than did control subjects. In this study, higher levels of guilt thus led to a higher magnitude of compliance.
The meta-analysis that follows examined the available data on the effect of induced guilt on compliance and estimated the magnitude of the effect. It also sought to ascertain whether or not the effect was moderated, and, if so, to identify those moderator variables.
Method
The procedures described by Schmidt and Hunter (2015) were employed in this meta-analysis. Specifically, after identifying all studies that met the inclusion criteria, the Pearson product–moment correlation coefficient between guilt as induced and compliance was calculated for all studies for which sufficient information to perform the calculation was available. Concurrently, study characteristics, both methodological artifacts and potential substantive moderators, were coded for each study for which information was available.
After information from each study was obtained, the analysis proceeded in three stages. First, a barebones meta-analysis (i.e., correcting for sampling error only) was conducted, in which weighted mean effect sizes and weighted variances were calculated. If the variance in effect sizes could not be attributed to sampling error, corrections were then made for other statistical artifacts. Finally, if additional variance in effect sizes still remained following these corrections, the possibility of substantive moderator effects was entertained.
Search criteria
PsycInfo, PsycArticles, Communication and Mass Media Complete, Communication Abstracts, JSTOR, Wiley Online, Web of Science, ProQuest Dissertations and Theses, and Sage Journals, were examined for relevant articles. Keywords included “guilt,” “transgress*,” “harm,” “compliance,” “comply,” “help*,” and “prosocial.” The reference sections of identified manuscripts were also examined for additional articles.
Inclusion criteria
Studies were retained if they met three conditions. First, they had to vary guilt experimentally without intentionally attempting to induce other affective states. Minimally, this requirement entails a guilt treatment, a no guilt control condition, and an independent group design in which participants are assigned (preferably randomly) to either the guilt or no guilt condition. So, for example, Darlington and Macker (1966) varied guilt by inducing subjects to perform poorly on a task (or not), thus depriving (or not depriving) a confederate of the opportunity to earn course credit. Thus, results from this experiment were included. Alternatively, Basil, Ridgway, and Basil (2006, Experiment 2) exposed subjects to advertisements designed to evoke either guilt, empathy, or both guilt and empathy. Because a no guilt comparison group that did not induce other emotions was unavailable, results from this experiment were not included. Studies of anticipatory guilt were also excluded from this meta-analysis because the dynamics of the two processes may diverge. Thus, separating their effects promotes conceptual and empirical clarity.
Second, studies had to include a measure of behavioral compliance or helping behavior. The measure may have taken the form of overt action, such as signing a petition (Brock & Becker, 1966); or a pledge to act, such as agreeing to have friends participate in a subsequent experiment (Boster et al., 1999).
Third, manuscripts had to provide sufficient information to estimate an effect size for the guilt induction–compliance relationship. This information may have been reported directly, or the manuscript may have included sufficient descriptive statistics to allow an effect size calculation to be made.
Coding
Each identified article was screened for eligibility by at least one author. Articles that met the inclusion criteria were then read by a pair of authors, with both authors coding relevant study characteristics and effect sizes. Cases of disagreement were resolved by discussion among the authors. Coded characteristics included both methodological and substantive variables.
Methodological characteristics
Methodological characteristics included aspects of the design, inductions, measures, and the request. Coded design elements included: year of publication, the number of subjects, the sex composition of the sample, the study context (field, laboratory, online, or unclear), whether or not experimenters and confederates were present, and whether or not experimenters and confederates were blind to conditions. Other coded methodological characteristics included: the nature of the guilt induction (active, e.g. breaking an apparatus; passive, e.g. indirectly harming another as a result of inaction; and scenario, e.g. reading a brief guilt-inducing story), when and if guilt was measured (not measured, measured in a pilot study, measured before the compliance measure, measured after the compliance measure, measured both before and after the compliance measure, or unclear), the number of items measuring guilt, and the response scale for any guilt items.
Target compliance or helping behavior characteristics were also coded. First it was determined whether the study included an explicit compliance-seeking request or an unsolicited helping opportunity. If studies included an explicit request for compliance, coders determined the method by which the request was delivered (no request, face-to-face, written message, both written and face-to-face, or unclear), who was the requestor (an experimenter, a confederate, an unknown other, or unclear), and whether or not the requestor was blind to conditions (no, yes, or unclear). The number of indicators used to measure compliance or helping and the number of response options for each indicator were also coded.
Substantive characteristics
Substantive study characteristics included characteristics of the requestor and of the desired outcome that could impact the valence or magnitude of the relationship between guilt and compliance. Specifically, coders identified the person harmed in the guilt induction (an experimenter, a confederate, an unknown person, or a group or team member), the beneficiary of the request (an experimenter, a confederate, an organization or cause, an unknown other, a team or group member, or some other beneficiary), whether or not either the beneficiary or requestor was the person harmed in the guilt induction (no, yes, or unclear), and whether or not compliance resulted in personal gain or reward for the subject. Studies were also coded for whether the focal outcome was behavioral intent or behavior, in addition to the behavior (actual or intended) of those willing to comply (donating money, donating time, picking up objects, making phone calls, donating blood, or some other behavior).
Results
A total of 47 effect sizes from 34 different papers were estimable from this literature (see Table ). The distribution of correlations was symmetrical but leptokurtic, with the positive kurtosis stemming from the fact that 14 (29.8%) of the effect sizes fell between .20 and .29.
Table 1. List of studies included in meta-analysis.
The total sample size for these 47 studies was N = 2,800. The distribution of sample size was skewed positively and mesokurtic, with a mean sample size of N = 59.57 (SD = 41.07, Mdn = 46, Mode = 20).
The first guilt experiments were published in 1966, and this line of research continued until 2012, albeit not continuously. Most of these papers were published during the 1970s, and approximately three-quarters were published before 1980 (76.6%).
There was no evidence that effect sizes varied across time (r = .01, 95% CI [−.28, .30], r = −.11, 95% CI [−.35, .13] weighted by N), although there was a tendency for sample sizes to increase over time (r = .26, 95% CI [−.01, .53]). Moreover, although the association was modest, studies conducted on larger samples also tended to find smaller effects (r = −.23, 95% CI [−.50, .04]).
Barebones meta-analysis
The meta-analysis was conducted using the Hunter–Schmidt Meta-Analysis Programs (Schmidt & Le, 2014). The estimated weighted mean effect size was ρ = .26 (95% CI [.21, .31]) with an observed standard deviation of .17.
Notably, in a study with a sample size of 59 subjects (the approximate mean sample for this set of studies), an effect of r = .26 would be statistically significant (albeit barely) at p ≤ .05 (two-tailed test). Hence, the power to detect an effect of r = .26 in such an experiment would be approximately .50. Put differently, approximately one half of the studies in this literature would be expected to yield null effects, whereas the other one half would be expected to yield statistically significant effects. Examining the distribution of correlations in Table indicates that 27 of the 47 effects (57.4%) were statistically significant at p ≤ .05 (two-tailed test). This figure deviates little from the expected 50%, indicating that there is only modest evidence of this manifestation of publication bias.
Finally, results indicated that 49.9% of the variance in observed correlations could be attributed to sampling error. Hunter and Schmidt (2004; Schmidt & Hunter, 2015, p. 64) have suggested that when at least 75% of the variance in observed correlations can be accounted for by known and correctable artifacts, the remaining 25% is likely to be due to unobservable artifacts. Because the variance due to sampling error fell short of this cutoff, an attempt was made to correct for other artifacts.
Correcting for other artifacts
Enough information could be obtained from the 47 studies to correct for two artifacts (see Table ). The first correction addressed the fact that inductions varied in strength as assessed by available guilt induction-experienced guilt correlations (considered error of measurement in the independent variable by Schmidt & Hunter, 2015, pp. 261–266). There were eight studies in which enough information was reported to calculate the relationship between the guilt induction and experienced guilt.4
The second correction addressed the fact that as the amount of compliance in the control group increases it restricts the size of the guilt induction–compliance correlation (treated as restriction in range in the dependent variable by Schmidt & Hunter, 2015, pp. 123–ff). In studies with dichotomous dependent variables, the proportion of control group compliance was compared to the proportion averaged across all studies. These ratios were then used to compute the correction factor. In studies using continuous-dependent variables, corrections were calculated by comparing the standard deviation of compliance in the control group to a reference standard deviation. Because studies varied widely in the scales used to measure compliance, the reference standard deviation was set as the value that would be obtained if the coefficient of variation was .50. Both corrections were made using artifact distributions, rather than correcting individual effect sizes (see Schmidt & Hunter, 2015, Chapter 4).
After correcting for these artifacts the effect size for these studies was ρ′ = .35 (95% CI [.28, .42]). Moreover, results indicated that 100% of the variance in observed effect sizes could be explained by the corrected artifacts.
Coded potential moderators
Because all variance in observed effect sizes could be explained by corrected artifacts, these results indicate that none of the coded potential moderators moderate the effect of guilt on compliance. Notably, there were two coded variables that appeared to be moderators before the correction for error of measurement and restriction in range were performed. First, it appeared that passive (e.g., verbal transgression, indirect harm via failed action) or active inductions (e.g., breaking important objects, harming another person) produced weaker effect sizes (overall mean r = .29, 95% CI [.25, .33]) than scenario inductions (mean r = .38, 95% CI [.30, .46]), z = 1.94, p = .05. Second, it appeared that studies using a larger proportion of female subjects tended to find smaller effects (r = −.43). Despite the suggestion of moderation, the finding that all observed variance can be explained by artifacts suggests that these effects do not differ for substantive reasons. Instead, chance associations between these variables and study artifacts create the illusion of moderation.
Primary study moderators
The fact that no coded moderators emerged in the meta-analytic calculations does not, however, warrant drawing the conclusion that the experienced guilt–compliance relationship is unmoderated. For example, it is possible that message characteristics, features of the experimental context, or personality characteristics could moderate the relationship. The reasons that they would not emerge as moderators in the meta-analysis are that they did not vary in many, if any, studies (e.g. most compliance messages were direct requests, most guilt studies did not require that the target lie) or that the values they assumed are assigned randomly across conditions (e.g. personality traits).
Nevertheless, some of these factors can be found by examining multi-factor primary studies and observing instances of non-additivity. When they emerge in primary studies, non-additive findings can provide a valuable guide for future research and subsequent theoretical development. Several of these non-additive effects were reported in the primary literature.
Message characteristics were identified as moderators in a number of studies. For example, Boster et al. (1999) reported that the effect of guilt on compliance was stronger when a positive self-feeling message (see Marwell & Schmitt, 1967) was used than when a direct request was employed. Cryder et al. (2012, Experiment 3) also report that the effect of guilt is stronger when the guilt party is able to fix the transgression that produced the guilt (integral condition, r = .43) than when the guilt party is not able to do so, (e.g. only ameliorate private feelings of guilt or improve reputation, incidental condition, r = .09). Furthermore, Cunningham, Steinberg, and Grev (1980, Experiment 2) found that the effect of guilt on compliance was more substantial in a negative message condition (r = .37) than in a positive message condition (r = −.08). Notably, what is termed a positive message by Cunningham et al. is a direct request, and what is termed a negative message is a debt strategy (Marwell & Schmitt, 1967).
Konoske, Staple, and Graf (1979) also find a non-additive effect of the nature of the compliance-gaining request. Specifically, in this experiment the effect of guilt on compliance was stronger in a non-deception condition (r = .51) than in a deception condition (r = −.05). In this case what is termed deception involves targets complying by making telephone calls that require them to lie; in the non-deception conditions they comply by making telephone calls in which they are not asked to lie. Consequently, the induction might be characterized more generally as making pro-social vs. anti-social requests.
Finally, two studies find non-additive effects of mood. McMillen (1971) and McMillen and Austin (1971) reported that induced guilt combined non-additively with a self-esteem induction to affect compliance. The effect of guilt on compliance was more modest for those in a positive self-esteem condition (r = −.09 and −.01) than for those in a no self-esteem condition (r = .51 and .44). Because the self-esteem induction involved receiving bogus feedback (none vs. positive) concerning one’s personality traits, it may be characterized as a mood induction (i.e., neutral mood vs. positive mood) so that guilt has a smaller effect on compliance when targets are in a very positive mood than when in a neutral mood. Alternatively, it may be matter of preserving one’s self-esteem, particularly the notion that one is a good person who typically behaves in a normative fashion (Steele, 1988). Thus, guilt may have a weaker effect on compliance when subjects can restore their self-esteem in other ways than complying with a request (e.g., by thinking about their positive qualities).5
Discussion
These data demonstrate the presence of a reasonably strong effect of guilt on compliance, ρ = .26 and ρ′ = .35. This effect is comparable to, or larger than, reported meta-analytic estimates of the effects of other compliance gaining techniques, including the foot-in-the-door (Beaman, Cole, Preston, Klentz, & Steblay, 1983, r = .09 for studies sampled and r = .21 for groups sampled; Dillard, Hunter, & Burgoon, 1984, r = .17), the door-in-the-face (Dillard et al., 1984; r = .15; Feeley, Anker, & Aloe, 2012, r = .13 verbal and r = .05 behavioral; O’Keefe & Hale, 2001; OR = 1.46), the legitimization of paltry favors (Andrews, Carpenter, Shaw, & Boster, 2008, r = .18; Bolkan & Rains, 2015; r = .22), the “but you are free” (Carpenter, 2013; r = .13), the disrupt-then-reframe (Carpenter & Boster, 2009; r = .21), and the low-ball (Burger & Caputo, 2015, r = .21).
There was also minimal evidence of some of the biases common to meta-analyses of various literatures. For example, the modest negative association between effect size and sample size indicates that publication bias is relatively minor in this literature.6
Because the meta-analysis indicated no evidence of substantive moderation, the weighted mean effect size provides the best available estimate of the strength of relationship between induced guilt and compliance. And, indeed, the effect may not be moderated. The finding of a homogeneous effect suggests that continued exploration of the moderators examined in this meta-analysis would not be useful. Instead, for future research in this area to be productive, it must move in different directions.
One set of possible directions emerges from examining the evidence of moderation reported in primary experiments. Particularly, message variations, such as positive self-feeling and debt, may produce different effects of guilt on compliance, as might the pro- vs. antisocial character of the request. A positive self-feeling message, in contrast to a direct request, may focus the target more clearly on what might be done to eliminate the negative affect generated by the guilt induction. Alternatively, in contrast with a direct request, a debt message may focus targets on the harm they have done, enhancing the strength of the guilt induction. Finally, compared with performing a prosocial act, requesting that someone reduce their obligation by performing an antisocial act may be considered too great a psychological cost for relieving the negative affect brought about by the guilt induction.
Additionally, mood, or self-esteem, holds the promise of moderating the effect, with larger effects expected under conditions of neutral mood and moderate self-esteem than positive mood or high self-esteem. Using mood as an example, those in a very positive mood may be able to discount psychologically the negative affect produced by the guilt induction. Or, the positive mood induction may counteract the negative affect produced by the guilt induction so that little relief is perceived as needed, rendering compliance unnecessary.
Each of these effects provides a fruitful avenue for future research. Certainly additional primary research of this nature is necessary before these hypotheses can receive a rigorous test.
There are other important limitations characteristic of the manner in which this literature has developed that also require future research to address. The simplest is the quantity of studies available to be summarized. Certainly, 47 effect sizes allow some relatively firm conclusions to be drawn. Nevertheless, additional, targeted experimental results would allow moderator hypotheses to be examined with greater rigor and would reduce second-order sampling error (Schmidt & Hunter, 2015, pp. 391–393). The limited mean sample size that is characteristic of guilt and compliance experiments also contributes to sampling error, further emphasizing the importance of additional experiments. Given that scenario inductions produced effects comparable in magnitude to direct and indirect inductions, online data collection may be employed more frequently. Consequently, larger samples can be obtained more efficiently in future research.
A conceptual challenge in studies of guilt and compliance involves distinguishing guilt and shame, and this challenge is particularly crucial because of its important empirical implications. Both guilt and shame arise from transgressions (Tangney & Dearing, 2002, p. 17), so they cannot be distinguished by their antecedents.7 Instead, a common way of distinguishing them is by their consequences. Guilt is thought to result in cognition focused on the person harmed, leading to regret, remorse, and thoughts about ways to repair relationships. In contrast, shame is thought to result in cognition focused on the self, feelings of a diminished self, and a desire to avoid interaction (Leary, 2004, pp. 97–98).
It is because of the differential consequences of guilt and shame that they are expected to be associated differently with compliance. Because guilty targets may view compliance with a request as an act of restitution, they are more likely to comply as a means of reducing their negative effect. Thus, there is likely to be more compliance among guilty subjects than among control subjects. And, there is likely to be a positive correlation between experienced guilt and the dependent variable, compliance.
In contrast, the self-focus of shamed targets disrupts their ability to empathize with the harmed party (Leary, 2004, pp. 97–98). This disruption results in a lack of feelings of remorse and regret, and leads to a lower probability (or magnitude) of compliance with subsequent requests. Therefore, there is expected to be less compliance among shamed subjects than among control subjects, producing a negative correlation between experienced shame and the dependent variable, compliance.
If transgressions bring about both guilt and shame in the transgressing party, and if guilt is associated positively with compliance but shame is associated negatively with compliance, then the interesting possibility depicted in Figure 2 arises. Because the transgression induction is likely associated positively with both experienced guilt and experienced shame, the predicted correlation between guilt and shame is also expected to be positive (they are driven by a common cause). But, if the correlation between experienced guilt and compliance is positive, and the correlation between experienced shame and compliance is negative, then the direct causal impact of guilt on compliance (the path coefficient, or standardized regression coefficient) will be suppressed unless shame is controlled. Thus, there are two reasons to expect that guilt and compliance experiments underestimate the experienced guilt–compliance relationship. First, guilt and compliance experiments estimate the impact of the guilt induction on compliance, an indirect effect, and do not focus on the theoretical relationship of interest, the experienced guilt–compliance effect, that is, notably, unaffected by the strength of the guilt induction (see Figure 1).8 Second, the effect of shame is not controlled in these experiments, so that the experienced guilt–compliance relationship is likely suppressed.
Published online:
18 February 2016![]({"type":"image","src":"/na101/home/literatum/publisher/tandf/journals/content/psif20/2016/psif20.v011.i01/15534510.2016.1142892/20170209/images/medium/psif_a_1142892_f0002_b.gif"})
Figure 2. The impact of experienced shame on the guilt–compliance relationship.
Additionally, because transgressions, or scenarios requiring targets to imagine transgressions, are likely to induce both guilt and shame, primary research would be informed by experimental work that seeks to develop inductions that either separate the two emotions, if possible, or measure the extent to which each emotion is induced. A strategy outlined by Spencer, Zanna, and Fong (2005) has the potential to be useful for this purpose. Spencer et al. (2005) raise the possibility of multiple data collections. In an initial experiment the effect of the guilt induction on experienced guilt and experienced shame (see Tangney & Dearing, 2002, for measures of these constructs) can be assessed; in a second experiment the impact of the induction on compliance can be measured. Of note is that fact that this approach avoids combining the potentially reactive measures of experienced guilt and experienced shame with the measure of compliance.
Two additional limitations, one methodological and the other substantive, have promise for informing future research. First, the vast majority of these data were collected with US residents, typically university students, serving as subjects. Given the important intercultural differences that have been observed in cognition (Nisbett, 2003), it would not be surprising to find substantial intercultural differences in emotion as well. Cross-national research would also provide the opportunity to examine any differences in the experienced guilt–compliance relationship between guilt cultures and shame cultures (Benedict, 1946). For cultures that rely more on shame for behavioral sanctions, for example, guilt might have a weaker impact on compliance.
Second, as de Hooge, Nelissen, Bruegelmans, and Zeelenberg (2011) point out, there can be a dark side to atoning for one’s guilt. In a series of experiments employing the dictator game they found that inducing guilt resulted in compensating behavior toward victims, but not at the expense of the guilt party. Instead, resources were transferred from non-harmed others to the victim. Such experiments pose a challenge to the view of guilt as a moral emotion. Whether or not there are conditions in which guilty parties compensate victims (with money, time, or effort) at their own expense also remains a potentially fruitful avenue for future research.
Finally, neither the theoretical development in this literature, or the available data, allowed tests of alternative theories of the induced guilt–compliance relationship. Nevertheless, one interesting omission is the literature is noteworthy. The NSRM predicts that compliance reduces, or eliminates, the guilt that arises from committing an unintentional transgression. Notably, that hypothesis remains unexamined. The results of a test of that hypothesis, particularly if the data are inconsistent with it, have the potential to spur additional theorizing in this area.
Conclusions
This review adds to the growing set of quantitative reviews that estimate the strength of the effects of techniques designed to gain behavioral compliance by summarizing the impact of induced guilt on compliance. The findings show that induced guilt has a moderately strong impact on compliance, although, because of the indirect nature of the effect and the possibility that it is suppressed by confounding of guilt inductions (shame), the effect of experienced guilt on compliance may be substantially stronger. Importantly, there was no meta-analytic evidence of moderation in these data after correcting for methodological artifacts, even though two specious moderators were identified before corrections were made. This finding reinforces the Schmidt and Hunter (2015) dictum that a barebones meta-analysis (examining the effect of sampling error only) is an incomplete meta-analysis.
Disclosure statement
No potential conflict of interest was reported by the authors.
| Authors | Year | Study number | N | r | rxx | u |
|---|---|---|---|---|---|---|
| Boster et al. | 1999 | 1 | 60 | .32 | .33 | .53 |
| Brock and Becker | 1966 | 1 | 72 | .60 | – | – |
| Carlsmith and Gross | 1969 | 1 | 40 | .50 | – | 1.34 |
| Carlsmith and Gross | 1969 | 2 | 20 | .63 | – | 1.61 |
| Cialdini, Darby, and Vincent | 1973 | 1 | 15 | .22 | – | .67 |
| Cryder, Springer, and Morewedge | 2012 | 1 | 122 | .46 | .81 | .41 |
| Cryder, Springer, and Morewedge | 2012 | 2 | 22 | .45 | .32 | 1.09 |
| Cryder, Springer, and Morewedge | 2012 | 3 | 139 | .26 | – | .45 |
| Cunningham, Steinberg, and Grev | 1980 | 1 | 45 | .35 | – | .77 |
| Cunningham, Steinberg, and Grev | 1980 | 2 | 80 | .18 | – | 2.23 |
| Darlington and Macker | 1966 | 1 | 26 | .31 | – | .71 |
| de Hooge, Nelissen, Bruegelmans, and Zeelenberg | 2011 | 1 | 74 | .39 | .78 | .26 |
| de Hooge, Nelissen, Bruegelmans, and Zeelenberg | 2011 | 2 | 143 | .32 | .93 | .33 |
| de Hooge, Nelissen, Bruegelmans, and Zeelenberg | 2011 | 3 | 44 | .37 | .80 | .90 |
| Ellis | 1978 | 1 | 27 | −.11 | – | 1.05 |
| Fisher | 1971 | 1 | 160 | .13 | – | – |
| Freedman, Wallington, and Bless | 1967 | 1 | 62 | .29 | – | .94 |
| Freedman, Wallington, and Bless | 1967 | 2 | 62 | .35 | – | .88 |
| Freedman, Wallington, and Bless | 1967 | 3 | 66 | .30 | – | 1.26 |
| Harris and Samerotte | 1976 | 1 | 100 | .38 | – | 1.06 |
| Kahler | 1978 | 1 | 20 | .26 | – | .28 |
| Katz, Glass, Lucido, and Farber | 1979 | 1 | 99 | .09 | – | 1.20 |
| Katz, Glass, Lucido, and Farber | 1979 | 2 | 82 | .10 | – | .95 |
| Katz, Glass, Lucido, and Farber | 1979 | 3 | 57 | .26 | – | .72 |
| Katzev, Edelsack, Steinmetz, Walker, and Wright | 1978 | 1 | 41 | .22 | – | .96 |
| Katzev, Edelsack, Steinmetz, Walker, and Wright | 1978 | 2 | 36 | .20 | – | 1.15 |
| Kelln and Ellard | 1999 | 1 | 28 | .20 | .17 | .58 |
| Konecni | 1972 | 1 | 62 | .28 | – | 2.08 |
| Konoske, Staple, and Graf | 1979 | 1 | 30 | .54 | – | .59 |
| McMillen | 1970 | 1 | 16 | .52 | – | .89 |
| McMillen | 1971 | 1 | 22 | .55 | – | 1.23 |
| McMillen and Austin | 1971 | 1 | 20 | .84 | – | .64 |
| McMillen, Jackson, and Austin | 1974 | 1 | 50 | .43 | – | .48 |
| Neumann | 1974 | 1 | 160 | .01 | – | .70 |
| Neumann | 1974 | 2 | 142 | −.07 | – | .64 |
| Neumann | 1974 | 3 | 46 | .16 | – | .67 |
| Noel | 1971 | 2 | 18 | −.30 | – | 3.55 |
| O’Malley and Andrews | 1983 | 1 | 60 | .29 | .30 | 2.00 |
| Peters | 1974 | 1 | 32 | .23 | – | 2.68 |
| Polman and Ruttan | 2012 | 2 | 64 | .23 | – | .58 |
| Rawlings | 1968 | 1 | 20 | .28 | – | .82 |
| Regan | 1971 | 1 | 27 | .28 | – | .67 |
| Regan, Williams, and Sparling | 1972 | 1 | 40 | .42 | – | 2.23 |
| Riordan, Dunaway, Haa, James, and Kruge | 1984 | 1 | 20 | .05 | – | .48 |
| Schwartz and David | 1976 | 1 | 72 | .24 | – | – |
| Wallace and Sadalla | 1966 | 1 | 26 | .55 | – | 2.17 |
| Yinon, Bizman, Cohen, and Segec | 1976 | 1 | 120 | .43 | – | .34 |
Notes. Values in the rxx column are guilt induction-experienced guilt correlations (considered error of measurement in the independent variable). Values in the u column are correction factors for restriction in range in the dependent variable.
Notes
1. Baumeister (20052005; Baumeister, Stillwell, & Heatherton, 1994) emphasizes harm that is done to those with whom one has close relationships as the source of guilt. Whether Baumeister denies that guilt arises from harming others with whom one is not close, or, instead, believes that the magnitude of guilt arising from such transgressions is smaller, is unclear. Nevertheless, as the meta-analytic results will show, the outcomes of experiments assessing the effect of guilt on compliance demonstrate that guilt is induced in these experiments (albeit imperfectly, and possibly with confounded inductions), that the person harmed and the person making the request for compliance are generally unknown to the person in whom guilt is induced, and that induced guilt has an impact on compliance.
2. Generally, non-accidental, or intentional, transgressions would not be anticipated to generate guilt; they would rarely produce the remorse or regret that so commonly ensues after the transgression. Such events could, however, generate such guilt-related feelings after the passage of a substantial amount of time. The experiments reviewed in the meta-analysis that follows are cross-sectional with measures of compliance taken very soon after the guilt induction. Thus, the latter possibility remains unexamined in the observed literature. Additionally, the possibility that guilt could arise from something other than a transgression (a lack of action, such as failing to discharge a duty, being considered a type of transgression) can be entertained. For instance, persons may imagine that they committed a transgression, or through inaction committed a transgression, that they did not. The literature reviewed in this meta-analysis includes inductions that require subjects to imagine committing a transgression that they would not have actually committed. The effect of these inductions on compliance does not differ substantially from those in which actual transgressions were committed.
3. The research thread that examines the impact of guilt on compliance is marked by a lack of more general theory, as well as by a lack of theoretical controversy. Instead, disagreements exists as to the conditions under which induced guilt leads to experienced guilt (e.g., see Note 1), other consequents of induced guilt (e.g., see Carlsmith & Gross, 1969), and moderators of the relationship between experienced guilt and compliance (de Hooge, Nelissen, Bruegelmans, & Zeelenberg, 2011).
4. As would be expected, stronger inductions were associated with stronger effects of the guilt induction on compliance, r = .47.
5. We would like to thank an anonymous reviewer for raising this point.
6. In addition to the negative association between effect size and sample size, an examination of a funnel plot showed the usual pattern of greater variance in effect sizes as sample size decreased.
7. Some have argued that the type of situation is an antecedent that distinguishes guilt and shame (Benedict, 1946). As Tangney and Dearing (2002, p. 14) point out, pertinent data are inconsistent with this claim.
8. If the NSRM is correct, then the relationship between experienced guilt and compliance can be estimated. Specifically, from the covariance algebra of causal analysis it follows that the induced guilt–compliance correlation is the product of the induced guilt-experienced guilt and experienced guilt–compliance correlations. Thus, the experienced guilt–compliance correlation can be estimated as the induced guilt–compliance correlation divided by the induced guilt-experienced guilt correlation. Using the figures obtained in this meta-analysis the experienced guilt–compliance correlation is estimated to be .52 (.64 if the if the corrected induced guilt–compliance correlation is employed).
References
- Andrews, K. R., Carpenter, C. J., Shaw, A. S., & Boster, F. J. (2008). The legitimization of paltry favors effect: A review and meta-analysis. Communication Reports, 21, 59–69. doi:10.1080/08934210802305028 [Taylor & Francis Online], [Google Scholar]
- Basil, D. Z., Ridgway, N. M., & Basil, M. D. (2006). Guilt appeals: The mediating effect of responsibility. Psychology and Marketing, 23, 1035–1054. doi:10.1002/mar.20145 [Crossref], [Web of Science ®], [Google Scholar]
- Baumeister, R. F. (2005). The cultural animal: Human nature, meaning, and social life. Oxford, UK: Oxford University Press. [Crossref], [Google Scholar]
- Baumeister, R. F., Stillwell, A. M., & Heatherton, T. E. (1994). Guilt: An interpersonal approach. Psychological Bulletin, 115, 243–267. doi:10.1037/0033-2909.115.2.243 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- Beaman, A. L., Cole, C. M., Preston, M., Klentz, B., & Steblay, N. M. (1983). Fifteen years of foot-in-the door research: A meta-analysis. Personality and Social Psychology Bulletin, 9, 181–196. doi:10.1177/0146167283092002 [Crossref], [Web of Science ®], [Google Scholar]
- Benedict, R. (1946). The chrysanthemum and the sword: Patterns of Japanese culture. Boston, MA: Houghton Mifflin. [Google Scholar]
- Bolkan, S., & Rains, S.A. (2015). The legitimization of paltry contributions as a compliance-gaining technique: A meta-analysis testing three explanations. Communication Research. Advance online publication. doi:10.1177/0093650215602308 [Crossref], [Google Scholar]
- *Boster, F. J., Mitchell, M. M., Lapinski, M. K., Cooper, H., Orrego, V. O., & Reinke, R. (1999). The impact of guilt and type of compliance-gaining message on compliance. Communication Monographs, 66, 168–177. doi:10.1080/03637759909376470 [Taylor & Francis Online], [Web of Science ®], [Google Scholar]
- *Brock, T. C., & Becker, L. A. (1966). “Debriefing” and susceptibility to subsequent experimental manipulations. Journal of Experimental Social Psychology, 2, 314–323. doi:10.1016/0022-1031(66)90087-4 [Crossref], [Web of Science ®], [Google Scholar]
- Burger, J. M., & Caputo, D. (2015). The low-ball compliance procedure: A meta-analysis. Social Influence, 10, 214–220. doi:10.1080/15534510.2015.1049203 [Taylor & Francis Online], [Web of Science ®], [Google Scholar]
- *Carlsmith, J. M., & Gross, A. E. (1969). Some effects of guilt on compliance. Journal of Personality and Social Psychology, 11, 232–239. doi:10.1037/h0027039 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- Carpenter, C. J. (2013). A meta-analysis of the effectiveness of the “but you are free” compliance-gaining technique. Communication Studies, 64, 6–17. doi:10.1080/10510974.2012.727941 [Taylor & Francis Online], [Google Scholar]
- Carpenter, C. J., & Boster, F. J. (2009). A meta-analysis of the effectiveness of the disrupt-then-reframe compliance gaining technique. Communication Reports, 22, 55–62. doi:10.1080/08934210903092590 [Taylor & Francis Online], [Google Scholar]
- *Cialdini, R. B., Darby, B. L., & Vincent, J. E. (1973). Transgression and altruism: A case for hedonism. Journal of Experimental Social Psychology, 9, 502–516. doi:10.1016/0022-1031(73)90031-0 [Crossref], [Web of Science ®], [Google Scholar]
- *Cryder, C. E., Springer, S., & Morewedge, C. K. (2012). Guilty feelings, targeted actions. Personality and Social Psychology Bulletin, 38, 607–618. doi:10.1177/0146167211435796 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- *Cunningham, M. R., Steinberg, J., & Grev, R. (1980). Wanting to and having to help: Separate motivations for positive mood and guilt-induced helping. Journal of Personality and Social Psychology, 38, 181–192. doi:10.1037/0022-3514.38.2.181 [Crossref], [Web of Science ®], [Google Scholar]
- *Darlington, R. B., & Macker, C. E. (1966). Displacement of guilt-produced altruistic behavior. Journal of Personality and Social Psychology, 4, 442–443. doi:10.1037/h0023743 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- *de Hooge, I. E., Nelissen, R. M. A., Bruegelmans, S. M., & Zeelenberg, M. (2011). What’s moral about guilt? Acting “prosocially” at the disadvantage of others. Journal of Personality and Social Psychology, 100, 462–473. doi:10.1037/a0021459 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- Dillard, J. P., Hunter, J. E., & Burgoon, M. (1984). Sequential-request persuasive strategies: Meta-analysis of foot-in-the-door and door-in-the-face. Human Communication Research, 10, 461–488. doi:10.1111/j.1468-2958.1984.tb00028.x [Crossref], [Web of Science ®], [Google Scholar]
- *Ellis, K. J. (1978). Personality and the transgression-altruism phenomenon (Unpublished doctoral dissertation). Rosemead Graduate School of Professional Psychology, La Mirada, CA. [Google Scholar]
- Feeley, T. H., Anker, A. E., & Aloe, A. M. (2012). The door-in-the-face persuasive message strategy: a meta-analysis of the first 35 years. Communication Monographs, 79, 316–343. doi:10.1080/03637751.2012.697631 [Taylor & Francis Online], [Web of Science ®], [Google Scholar]
- *Fisher, R. A. (1971). The effects of guilt and shame on public and private helping (Unpublished doctoral dissertation). University of Wisconsin – Madison, Madison, WI. [Google Scholar]
- Freedman, J. L., Wallington, S. A., & Bless, E. (1967). Compliance without pressure: The effect of guilt. Journal of Personality and Social Psychology, 7, 117–124. doi:10.1037/h0025009 [Crossref], [Web of Science ®], [Google Scholar]
- Freud, S. (2005). Civilization and its discontents. (J. Strachey, Trans.). New York, NY: Norton (Original work published in 1930). [Google Scholar]
- *Harris, M. B., & Samerotte, G. C. (1976). The effects of actual and attempted theft, need, and a previous favor on altruism. The Journal of Social Psychology, 99, 193–202. doi:10.1080/00224545.1976.9924772 [Taylor & Francis Online], [Web of Science ®], [Google Scholar]
- Hunter, J. E., & Schmidt, F. L. (2004). Methods of meta-analysis: Correcting error and bias in research findings (2nd ed.). Thousand Oaks, CA: Sage. [Crossref], [Google Scholar]
- *Kahler, B. J. (1978). Helping and self-punishment as alternative behaviors following accidental harm-doing (Unpublished doctoral dissertation). University of Kansas, Lawrence, KS. [Google Scholar]
- Kant, I. (1959). Foundations of the metaphysics of morals, and what is enlightenment? (L. W. Beck, Trans.). Indianapolis, IN: Bobbs-Merrill (Original work published in 1785). [Google Scholar]
- *Katz, I., Glass, D. C., Lucido, D., & Farber, J. (1979). Harm-doing and victim’s racial or orthopedic stigma as determinants of helping behavior. Journal of Personality, 47, 340–364. doi:10.1111/j.1467-6494.1979.tb00207.x [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- *Katzev, R., Edelsack, L., Steinmetz, G., Walker, T., & Wright, R. (1978). The effect of reprimanding transgressions on subsequent helping behavior: Two field experiments. Personality and Social Psychology Bulletin, 4, 326–329. doi:10.1177/014616727800400233 [Crossref], [Web of Science ®], [Google Scholar]
- *Kelln, B. R. C., & Ellard, J. H. (1999). An equity theory analysis of the impact of forgiveness and retribution on transgressor compliance. Personality and Social Psychology Bulletin, 25, 864–872. doi:10.1177/0146167299025007008 [Crossref], [Web of Science ®], [Google Scholar]
- *Konecni, V. J. (1972). Some effects of guilt on compliance: A field replication. Journal of Personality and Social Psychology, 23, 30–32. doi:10.1037/h0032875 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- *Konoske, P., Staple, S., & Graf, R. G. (1979). Compliant reactions to guilt: Self-esteem or self-punishment. The Journal of Social Psychology, 108, 207–211. doi:10.1080/00224545.1979.9711633 [Taylor & Francis Online], [Web of Science ®], [Google Scholar]
- Leary, M. R. (2004). The curse of the self: Self-awareness, egotism, and the quality of human life. Oxford: Oxford University Press.10.1093/acprof:oso/9780195172423.001.0001 [Crossref], [Google Scholar]
- Marwell, G., & Schmitt, D. R. (1967). Dimensions of compliance-gaining behavior: An empirical analysis. Sociometry, 30, 350–364. doi:10.2307/2786181 [Crossref], [Google Scholar]
- *McMillen, D. L. (1970). Transgression, fate control, and compliant behavior. Psychonomic Science, 21, 103–104.10.3758/BF03335787 [Crossref], [Google Scholar]
- *McMillen, D. L. (1971). Transgression, self-image, and compliant behavior. Journal of Personality and Social Psychology, 20, 176–179. doi:10.1037/h0031686 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- *McMillen, D. L., & Austin, J. B. (1971). Effect of positive feedback on compliance following transgression. Psychonomic Science, 24, 59–61.10.3758/BF03337892 [Crossref], [Google Scholar]
- *Mcmillen, D. L., Jackson, J. A., & Austin, J. B. (1974). Effects of positive and negative requests on compliance following transgression. Bulletin of the Psychonomic Society, 3, 80–82.10.3758/BF03333403 [Crossref], [Google Scholar]
- *Neumann, S. (1974). The effects of experimentally induced guilt and shame on helping behavior (Unpublished doctoral dissertation). University of Wisconsin – Madison, Madison, WI. [Google Scholar]
- Niebuhr, R. (1932). Moral man and immoral society: A study in ethics and politics. New York, NY: Scribner. [Google Scholar]
- Nisbett, R. E. (2003). The geography of thought: How Asians and Westerners think differently … and why. New York, NY: Free Press. [Google Scholar]
- *Noel, R. C. (1971). Helping behavior as a consequence of a prior transgression (Unpublished doctoral dissertation). University of Colorado, Boulder CO. [Google Scholar]
- O’Keefe, D. J., & Hale, S. L. (2001). An odds-ratio based meta-analysis of research on the door-in-the-face influence strategy. Communication Reports, 14, 31–38. doi:10.1080/08934210109367734 [Taylor & Francis Online], [Google Scholar]
- *O’Malley, M. N., & Andrews, L. (1983). The effect of mood and incentives on helping: Are there some things money can’t buy? Motivation and Emotion, 7, 179–189. doi:10.1007/BF00992901. [Crossref], [Web of Science ®], [Google Scholar]
- *Peters, J. E. (1974). The effects of locus-of-control and temporary affective states of the benefactor on altruism (Unpublished doctoral dissertation). University of Montana, Missoula, MT. [Google Scholar]
- *Polman, E., & Ruttan, R. L. (2012). Effects of anger, guilt, and envy on moral hypocrisy. Personality and Social Psychology Bulletin, 38, 129–139. doi:10.1177/0146167211422365 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- *Rawlings, E. I. (1968). Witnessing harm to other: A reassessment of the role of guilt in altruistic behavior. Journal of Personality and Social Psychology, 10, 377–380. doi:10.1037/h0026819 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- *Regan, J. W. (1971). Guilt, perceived injustice, and altruistic behavior. Journal of Personality and Social Psychology, 18, 124–132. doi:10.1037/h0030712 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- *Regan, D. T., Williams, M., & Sparling, S. (1972). Voluntary expiation of guilt: A field experiment. Journal of Personality and Social Psychology, 24, 42–45. doi:10.1037/h0033553 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- *Riordan, C. A., Dunaway, F. A., Haa, P., James, M. K., & Kruge, D. (1984). Prosocial behavior following transgression: Evidence for intrapsychic and interpersonal motives. The Journal of Social Psychology, 124, 51–55. doi:10.1080/00224545.1984.9924536 [Taylor & Francis Online], [Web of Science ®], [Google Scholar]
- Schmidt, F. L., & Hunter, J. E. (2015). Methods of meta-analysis: Correcting error and bias in research findings (3rd ed.). Thousand Oaks, CA: Sage. [Crossref], [Google Scholar]
- Schmidt, F. L., & Le, H. (2014). Hunter–Schmidt meta-analysis programs [Statistical software]. Iowa City, IA: University of Iowa. [Google Scholar]
- *Schwartz, S., & David, A. B. (1976). Responsibility and helping in an emergency: Effects of blame, ability and denial of responsibility. Sociometry, 39, 406–415. doi:10.2307/3033505 [Crossref], [PubMed], [Google Scholar]
- Smith, A. (2002). The theory of moral sentiments. Cambridge: Cambridge University Press (Original work published in 1759).10.1017/CBO9780511800153 [Crossref], [Google Scholar]
- Spencer, S. J., Zanna, M. P., & Fong, G. T. (2005). Establishing a causal chain: Why experiments are often more effective than meditational analyses in examining psychological processes. Journal of Personality and Social Psychology, 89, 845–851. doi:10.1037/0022-3514.89.6.845 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]
- Steele, C. M. (1988). The psychology of self-affirmation: Sustaining the integrity of the self. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 21, pp. 261–302). New York, NY: Academic Press. [Crossref], [Google Scholar]
- Tangney, J. P., & Dearing, R. L. (2002). Shame and guilt. New York, NY: Guilford Press.10.4135/9781412950664 [Crossref], [Google Scholar]
- *Wallace, J., & Sadalla, E. (1966). Behavioral consequences of transgression: I. The effects of social recognition. Journal of Experimental Research in Personality, 1, 187–194. [Google Scholar]
- *Yinon, Y., Bizman, A., Cohen, S., & Segec, A. (1976). Effects of guilt-arousal communications on volunteering to the civil guard: A field experiment. Bulletin of the Psychonomic Society, 7, 493–494.10.3758/BF03337260 [Crossref], [Google Scholar]