Abstract
We investigated the scope of the effect of disgust on moral judgments. In two field experiments (Experiment 1, N = 142, Experiment 2, N = 248), we manipulated whether participants were exposed to a disgusting odor. Participants then rated the permissibility of actions in two kinds of moral problems: dilemmas and transgressions. In both experiments, disgust did not affect moral judgments when we compared across exposure levels. However, self-reported disgust did predict moral judgments in the following cases: In Experiment 1, it was linked with decreased acceptability for dilemmas and transgressions alike; in Experiment 2, it was linked with decreased acceptability for dilemmas only. Findings also differed across the experiments when we regressed feelings of disgust onto participants’ utilitarian and deontological inclinations. Overall, the findings suggest that subjective feelings of disgust may provide a more sensitive measure of the effect of disgust on moral judgment than basing analysis on the presence of disgust elicitors.
Feeling disgust may cause people’s moral evaluations to be more severe. For example, when people smell disgusting odors, they give more negative moral evaluations to actions like eating one’s own dog (Schnall et al., 2008) or sacrificing a person to save several others (Ugazio et al., 2012). People have also been found to give more severe moral evaluations after recalling disgusting experiences (Schnall et al., 2008), seeing disgusting movie scenes (Harlé & Sanfey, 2010; Ugazio et al., 2012), and seeing subliminally presented faces expressing disgust (Cameron et al., 2013). Moreover, preventing people from feeling disgust (i.e., by having them consume ginger) reduces their moral condemnations of some moral violations (Tracy et al., 2019).
At the same time, effects of disgust on moral judgment have been called into question. Some papers have failed to replicate the impact of inducing disgust on moral judgments (e.g., David & Olatunji, 2011; Johnson et al., 2016). Also, a meta-analysis of studies on inducing disgust suggested its effect on moral judgment is marginally strong and disappears when controlling for publication bias (Landy & Goodwin, 2015; but see Schnall et al., 2015, for a response). Hence, the effect of disgust on moral judgment is subject to some controversy (for recent reviews of research relevant to this debate, see Giner-Sorolla et al., 2018; Rhee et al., 2019).
In this article, we attempt to better understand how disgust affects moral judgments. The most prominent explanation is the amplification hypothesis (Landy & Goodwin, 2015; Pizarro et al., 2011), which proposes that disgust makes bad actions seem more wrong than they otherwise would. We outline and test between two competing accounts of how amplification operates. However, as we will detail, our findings may also be informative about the relative merits of different approaches to investigating the effects of disgust on moral judgment.
Two Accounts of How Disgust Amplifies Moral Judgments
Disgust might affect moral judgment by specifically amplifying aversion to immoral actions. Consider a moral dilemma in which torturing a suspected terrorist may prevent an attack that will kill many innocent people. Eliciting disgust could specifically amplify aversion to the act of torture, without affecting feelings about the deaths that will result if the terrorist is not tortured. This specific or “narrow” form of amplification might arise because in considering moral dilemmas, people primarily focus on the action that may be taken and not on the consequences of inaction (Cushman, 2013; Trémolière & De Neys, 2013).
This narrow form of amplification is broadly consistent with dual-process accounts, which claim that evaluations of immoral actions, like torture, stem from fast and nonreflective emotional reactions, whereas considering the potential consequences of such actions (e.g., saving lives) depends on higher order cognition (Greene et al., 2008; Greene et al., 2001). That is, disgust might contribute more to emotional aspects of moral judgments than to deliberative aspects (see Conway & Gawronski, 2013, for related discussion).
Alternatively, disgust could operate more broadly. It might increase aversion to immoral actions, while also increasing aversion to the negative consequences of inaction. In the example above, it might make people view torturing the suspect more negatively, while also amplifying their negative judgments of not acting and allowing the victims of the bombing to be killed. It is plausible that inducing disgust increases condemnation of inaction because people are often described as feeling disgusted and outraged when they learn about inactions and omissions that permit harm (Booth, 2001; Dressler, 2000). For example, people expressed moral disgust toward David Cash Jr., who failed to intervene when his friend raped and murdered a 7-year-old in a casino bathroom (Lambeth, 2000). This said, emotional responses to inaction could be most potent when the actions needed to avoid harm are not themselves immoral.
Consistent with the “broad” view, a recent study found that people who are more prone to feelings of disgust show stronger tendencies against both actions that violate moral norms and decisions that omit acting in the interests of the greater good (Robinson et al., 2019, see their supplemental material). However, we know of no work examining whether inducing disgust affects assessments of inactions that allow harm to occur.
Testing Between the Accounts
We test between the broad and narrow accounts in two ways. First, we compare the effects of disgust on two kinds of moral problems: transgressions and dilemmas. In moral transgressions, an agent performs an immoral act without any moral justification and possibly only for self-benefit. For example, they steal money to buy a fancy smartphone or just for the thrill of it. The broad and narrow versions of the amplification theory both predict that disgust should amplify condemnation for these items.
In moral dilemmas, the agent performs an immoral act to secure other moral goals. For example, the agent kills one person to save the lives of many others. Because dilemmas include an immoral act, the narrow account again predicts that disgust should amplify condemnation (just as in transgressions). However, the broad account offers a different prediction: Because the immoral act serves a larger moral goal, disgust could also amplify aversion to the negative outcomes that would arise from not taking this action (i.e., if participants consider this counterfactual). Given this, the net effect of disgust could dampen for moral dilemmas—disgust might lead to less condemnation in dilemmas than in transgressions.
We also test between the broad and narrow accounts by comparing two inclinations in people’s moral judgments: deontology and utilitarianism. People show the deontological inclination by considering abstract rules regarding the acceptability of actions and show the utilitarian inclination by considering whether the outcome of an action serves the greater good. Crucially, both inclinations occur in the very same judgment (Białek & De Neys, 2017; Conway & Gawronski, 2013). For example, a person might consider torture unacceptable (deontology), yet advocate using it to save lives (utilitarianism). Again, the accounts make different predictions about how disgust affects these inclinations. The narrow account predicts that disgust should only affect the deontological inclination (i.e., because this account claims the effects of disgust are specific to actions that violate moral rules). In contrast, the broad account predicts that disgust should increase both inclinations. For example, disgust should make us view torture more negatively, while also making us view the potentially avoidable deaths of the bomb victims more negatively too.
We tested between these accounts in two experiments. As in Schnall et al. (2008), participants were tested in a natural setting outside, and disgust was induced by baiting a nearby trash can. We felt that conducting the experiment outside of a lab setting would reduce the chance of participants understanding that the experimenters were the source of their disgust, as awareness of the source could reduce its effect (e.g., Kugler et al., 2020; see Schnall et al., 2015, for discussion). However, in contrast with previous work on the effects of induced disgust on moral judgments, our participants responded to moral problems used to derive utilitarian and deontological tendencies—the battery of incongruent and congruent items devised by Conway and Gawronski (2013).
Experiment 1
Data and materials from both experiments are available at Open Science Framework (OSF) at this link. 1
Participants
We tested 142 participants (M age = 22 years, range = 19–35 years, 57 females). In both experiments, participants were tested at the academic campus of the Maria Curie-Skłodowska University (Lublin, Poland) over several days, during daytime hours (10 a.m.–6 p.m.), and were remunerated for their effort. This sample is larger than in Schnall et al.’s Experiment 1 (n = 127) and greater than the samples from 30 of the 31 published studies in Landy and Goodwin’s (2015) meta-analysis.
Procedure
Two experimenters sat at a table outdoors on a central path through the campus. They asked people passing by whether they were willing to participate in an academic survey. People who agreed completed a questionnaire (described below), either in the presence of a strong or moderate disgusting smell or with no smell present. Participants were then given a separate page asking whether they smelled anything (yes/no), whether the smell had distracted them while they completed the questionnaire (yes/no), and the extent to which they felt disgusted (7-point Likert-type scale). Finally, participants were thanked, debriefed, and remunerated in cash.
To create the disgusting smell, doses of Hukinol wildlife repellent were put into a nearby trash container (about 10 m from the table). A double dose (about 40 ml) was used for the “high-odor” group, a single dose (about 20 ml) was used for the “low-odor” group, and no dose was used for the “no-odor” group. These groups were tested on separate days. The smell lasted for about 45 min, which was long enough to test a participant. After 45 min, when the intensity of the smell decreased and almost disappeared, a new dose was put into the trash container. The experimenters started recruiting participants approximately 5 min after each baiting.
The questionnaire included a battery of 20 moral problems (Conway & Gawronski, 2013), including 10 moral dilemmas (incongruent items) and 10 moral transgressions (congruent items). The dilemmas involved situations where committing an immoral action will benefit society but not committing it will allow some societal harm to occur; the transgressions involve situations where committing the action provides a negligible societal benefit and not committing it carries no cost. For each item, participants first judged whether the action was morally permissible (yes/no) and rated their confidence in this decision (1 = a little, 2 = moderately, 3 = very much). As we detail below, these responses were recoded onto a single scale for some analyses.
The questionnaire also included 4 control items, as in Schnall et al. (2008, Experiment 2), in which participants rated the extent to which they were willing to engage in two novel activities (trying a meal they had never tasted, trying a risky sport) and two familiar ones (eating their favorite meal, watching their favorite movie). Participants responded using a 10-point Likert-type scale. 2
Results and Discussion
Manipulation Check
Participant’s ratings of disgust (1 = not disgusted at all; 7 = extremely disgusted), their awareness of a bad smell, and their reports of having been affected by the smell varied across testing groups. The effect of the odor manipulation on feelings of disgust was driven entirely by participants in the high-odor group. Disgust was greater for these participants (M = 3.48) than those exposed to low levels of the disgusting odor (M = 2.52), t(92) = −2.51, p = .039, and those not exposed to the odor (M = 2.52), t(94) = −2.36, p = .036 (p values Bonferroni-corrected), but did not differ between those groups, t(92) = −0.00, p = .998. As in Schnall et al. (2008), the odor manipulation did not affect nonmoral decisions like receptivity to familiar activities (M high = 7.80, M low = 7.76, M none = 8.09), F(2, 138) = 0.71, p = .496, or novel activities (M high = 6.91, M low = 7.17, M none = 6.80), F(2, 138) = 0.55, p = .580 (see Table 1).
Manipulation Checks of the Odor Manipulation in Both Experiments.
Disgust and Moral Dilemmas and Transgressions
We examined the effect of disgust on moral judgment in two ways. First, following most other studies, we examined whether moral ratings varied depending on the presence of disgust elicitors. To obtain continuous moral ratings, we recoded participants’ yes/no responses and three-level confidence ratings onto a 6-point scale, ranging from 1 = very confident no responses to 6 = very confident yes responses. This approach of combining the two measures is useful because it provides binary decisions needed for process dissociation analyses (see below) but also makes it possible to generate continuous scores for our initial analyses (for examples of other papers that recode responses in similar ways, see Shaw & Olson, 2014; Starmans & Friedman, 2012; Tenney et al., 2007; Trippas et al., 2013).
We then entered these moral ratings into a 3 (odor: high, low, none; between-subject) × 2 (type-of-problem: moral dilemma, moral transgression; within-subject) mixed-model analysis of variance (ANOVA; Figure 1). Transgressions were viewed as less permissible (M = 2.76) than dilemmas (M = 3.67), F(1, 139) = 185.54, p < .001,
Violin plots showing permissibility judgments by condition in Experiment 1 (left plots) and in Experiment 2 (right plots). Note. The width of the “violins” shows the proportion of responses at each level, and the horizontal bar shows the mean values. Points show responses from individual participants and are jittered to avoid overlaps.
By itself, this analysis suggests that inducing disgust did not affect moral judgment. However, a potential drawback with this analytic approach is that it overlooks variability in the extent to which participants actually feel disgust. For example, it misses out on the possibility that some participants in the high-odor group were greatly disgusted, while some others barely felt disgust at all. As such, this approach loses information, increasing the chance of null results. Hence, we also analyzed the data in another way that avoids this pitfall.
In this analysis, we examined whether moral judgments varied with participants’ self-reported levels of disgust (for a similar analytic approach, see Van Dillen et al., 2012, Study 2). A generalized estimating equations (GEE) model (linear data, independent correlation matrix) with type-of-problem (moral dilemmas, moral transgression) and self-reported level of disgust as predictors found a significant main effect of disgust, Wald χ2(1) = 6.81, p = .009, and type-of-problem, Wald χ2(1) = 71.78, p < .001, but no interaction, Wald χ2(1) = 0.79, p = .375; see Figure 2. Contrary to the initial analysis above, the main effect of disgust suggests that inducing disgust does amplify the severity of moral judgments. Further, the main effect of problem-type again shows that actions in transgressions are viewed as less acceptable than those in dilemmas. Finally, as predicted by the “narrow” account of amplification, there was no difference between the effects of disgust for dilemmas and transgressions. 3
Permissibility judgments by self-reported disgust in Experiments 1 (left plot) and in Experiment 2 (right plot). Note. Points are jittered to avoid overlaps; bands in plot on right show ±1 standard error.
Disgust and Deontological and Utilitarian Inclinations
To test whether disgust differentially affected deontological and utilitarian inclinations, we used the process dissociation approach (Conway & Gawronski, 2013). In this approach, the probability of accepting harm in moral dilemmas and rejecting harm in moral transgressions is equated with utilitarian moral inclination; the probability of rejecting harm in both moral dilemmas and moral transgressions is equated with deontological moral inclination; accepting harm in moral transgressions is equated with neither inclination. Because the deontological and utilitarian inclinations depend only on whether harm is accepted or not, they are calculated using participants’ binary yes/no judgments.
We used regressions to examine the effect of self-reported disgust on each inclination. Disgust predicted deontological inclination, R 2 = .04, β = .19, t(141) = 2.31, p = .023, B 10 = 2.50, but not utilitarian inclination, R 2 = .02, β = −.13, t(141) = −1.59, p = .115, B 01 = 1.76 (Figure 3). This finding is consistent with the narrow version for the amplification hypothesis, but not with the broad version.
Deontological and utilitarian moral inclinations as a function of disgust in Experiments 1 (left plot) and in Experiment 2 (right plot). Note. Points are jittered to avoid overlaps; bands show ±1 standard error.
Finally, we confirmed that our data replicated findings that deontological and utilitarian scores are not correlated, r(142) = .07, p = .418, B
01 = 6.88 (Conway & Gawronski, 2013; Muda et al., 2018) and that females are more deontological than males (M
female = 0.27, M
male = −0.18), F(1, 140) = 7.33, p = .008,
Experiment 2
This experiment used a similar design to Experiment 1 but with a few changes. First, we dropped the low-odor condition, as it did not produce more self-reported disgust than the no-odor condition. We also increased the power by increasing the sample to 248 participants. This allows us to detect (with .8 power and α of .05) correlations as small as r = .18 (two-tailed) and differences between groups of d = .31 (one-tailed). This experiment was preregistered at https://aspredicted.org/blind.php?x=g9ya3p.
Participants
We tested 248 participants (M age = 21 years, range = 18–63 years, 131 females), 2 fewer participants than specified in the preregistrations. An additional five participants were preliminarily counted as having completed the survey but were excluded (as specified in the preregistration) because they did not respond to all of the moral problems.
Procedure
Two experimenters sat at a table outdoors on a central path through the campus and asked people passing by whether they were willing to participate in an academic survey. Participants were tested in one of two between-subject conditions: no-odor and high-odor. These conditions were run on separate days. On days where the high-odor condition was run, doses of Hukinol animal repellant were periodically put in trash containers 10–15 m from the experimenter’s table. Participants first completed the battery of moral problems, then responded to the Body Consciousness Scale, and finally answered the manipulation check items and provided their demographics. In this experiment, participants responded to each moral problem on a 6-point Likert-type scale ranging from 1 = no, I consider it to be completely impermissible to 6 = yes, I consider it to be completely permissible. The manipulation checks (awareness of the odor, being affected by odor, and level of experienced disgust) were rated on 7-point Likert-type scales ranging from 1 = not at all to 7 = very much so.
Results and Discussion
Manipulation Check
Participant’s ratings of disgust, their awareness of a bad smell, and their reports of having been affected by the smell each varied across the high-odor and no-odor conditions. Disgust was greater for participants exposed to the disgusting odor (M = 2.88) compared to those not exposed to it (M = 2.23), t(242) = −3.00, p = .004. As in Schnall et al. (2008), odor did not affect nonmoral decisions like receptivity to familiar activities (M odor = 8.18; M none = 8.24), t(244) = 0.37, p = .715, or novel activities (M odor = 6.33, M none = 6.78), t(244) = 1.77, p = .078 (also see Table 1). Again, although disgust ratings remaining relatively low, our manipulation was effective in inducing disgust.
Disgust and Moral Dilemmas and Transgressions
As in the first experiment, we examined the effect of disgust on moral judgment in two ways. We started by examining whether moral ratings varied depending on the presence of disgust elicitors. A 2 (odor: high, none; between-subject) × 2 (type-of-problem: moral dilemma, moral transgression; within-subject) mixed-model ANOVA (Figure 1) revealed that transgressions were rated less permissible (M = 2.65) than dilemmas (M = 3.77), F(1, 238) = 614.11, p < .001,
In our second approach to examining the effect of disgust on moral judgment, we tested whether moral judgments varied with participants’ self-reported levels of disgust. A GEE model (linear data, independent correlation matrix) with the predictors’ type-of-problem (moral dilemmas, moral transgression) and self-reported level of disgust (entered as a covariate) found a significant main effect of type-of-problem, Wald χ 2 (1) = 348.19, p < .001, resulting because transgressions were viewed as less acceptable than dilemmas. There was no overall effect of disgust, Wald χ 2 (1) = 1.20, p = .273. However, disgust interacted with type-of-problem, Wald χ 2 (1) = 25.02, p < .001; see Figure 2. This interaction resulted because greater self-reported disgust was associated with lower acceptability ratings for moral dilemmas, Wald χ 2 (1) = 10.78, p < .001, but not for moral transgressions, Wald χ 2 (1) = 0.95, p = .331. These findings contrast with those from the first experiment (i.e., where the effects of self-reported disgust did not differ across the two types of moral problems). Moreover, these findings are consistent with neither the broad nor narrow account of amplification—neither account predicts stronger effects of disgust for dilemmas than transgressions.
Disgust and Deontological and Utilitarian Inclinations
We again used the process dissociation approach (Conway & Gawronski, 2013) to test whether disgust differentially affected deontological and utilitarian inclinations. Since this approach analysis requires dichotomized data, we recoded our permissibility ratings as responses of no (1–3) and yes (4–6), as outlined in our preregistration.
We used regressions to examine the effect of self-reported disgust on each inclination. Increased disgust was associated with a decrease in utilitarian inclination, R 2 = .065, β = −.25, t(234) = −4.02, p < .001, but it did not predict deontological inclination, R 2 = .004, β = .06, t(234) = 0.91, p = .364 (Figure 3). We followed the nonsignificant regression with Bayesian regression, with β binomial prior a = 3, b = 3. Evidence for the effect of self-reported disgust on utilitarian inclination was strong, B 10 = 242.40, whereas evidence against its effect on deontological inclination was weak, B 01 = 4.76. Although these findings sound different from those from our first experiment, Figure 3 suggests the findings were overall quite similar. Nonetheless, as with the analysis in the previous section, these findings were not predicted by the narrow account of amplification nor the broad account.
For consistency with Experiment 1, we conducted several additional analyses. First, both types of inclinations were uncorrelated, r(238) = .125, p = .054. Compared to males, females were more deontological (M female = 0.34, M male = −0.39), t(238) = 6.12, p < .001, but equally utilitarian (M female = −0.06, M male = 0.06), t(238) = −0.93, p = .354. Finally, in contrast with Experiment 1, self-reported disgust was weakly related to willingness to engage in both novel activities, R 2 = .027, β = −.16, t(241) = −2.57, p = .011, and familiar ones, R 2 = .016, β = −.13, t(241) = −2.00, p = .046.
General Discussion
We examined the effects of induced disgust on moral judgments in two experiments. Our findings were mixed, in regard to both the overall effect of inducing disgust on moral judgments and our test between the narrow and broad versions of the amplification hypothesis. We visit these issues in turn.
In both experiments, induced disgust had no effect on moral judgment when we compared participants across conditions varying in exposure to a noxious odor (i.e., no, low, and high exposures in Experiment 1; no and high exposures in Experiment 2). This null finding is consistent with other studies that have failed to find such effects (e.g., David & Olatunji, 2011; Ghelfi et al., 2018; Johnson et al., 2016) and with the conclusions of a meta-analysis that claimed that the effect of induced disgust on moral judgment is at best small (Landy & Goodwin, 2015). At the same time, it is possible that we did not see condition-level effects of disgust because our attempts to elicit disgust were only moderately successful (i.e., in both experiments, disgust ratings were relatively low).
Nonetheless, when we used an alternative analytic approach, we did find evidence that inducing disgust affects moral judgments, as participants’ self-reported levels of disgust predicted the severity of their moral judgments. One strength of this analytic approach is that it captures individual differences in the extent to which disgust elicitors affect moral judgment. A weakness, though, is that this approach may conflate other factors with induced disgust. For example, self-reported disgust may be affected by how prone participants are to feel disgust, and variations in this trait may also affect moral judgment (David & Olatunji, 2011; Inbar et al., 2009; Jones & Fitness, 2008; Wagemans et al., 2018).
Together these findings raise critical questions about which analytic approach is best for investigating effects of induced disgust on moral judgment. Self-reported disgust may provide a more direct test of the amplification hypothesis because the ultimate driver of changes in moral judgments is likely the sensation of disgust rather than the presence of a disgust elicitor. However, as acknowledged above, both approaches have their pros and cons. Related debates already exist in clinical research, where psychopathology can be conceptualized in terms of both discrete categories and continuous measures (e.g., Markon et al., 2011; also see Rucker et al., 2015). These findings also suggest it might be helpful to examine existing data sets using both analytic approaches where this is possible. Perhaps the best approach would be to ensure that investigations of the effects of disgust on moral judgment always use disgust elicitors that are near-universal in their effectiveness and lead to similar levels of disgust in all participants. To our knowledge, though, no studies have hit upon elicitors with this quality.
The Scope of Amplification
We also observed mixed findings when attempting to test between the narrow and broad versions of the amplification hypothesis. The findings from the first experiment favored the narrow view, both when examining responses to the dilemmas and transgressions and when looking at participants’ deontological and utilitarian inclinations. (Both analyses used self-reported disgust ratings.) But we did not replicate these findings in the second experiment, which was substantially better powered. Moreover, the findings from that experiment did not support either version of the amplification hypotheses that we initially outlined. For example, the narrow account predicted that increased disgust should be linked with decreased moral acceptability for dilemmas and transgressions alike, whereas the broad account predicted a stronger effect for transgressions. In contrast, in our second experiment, increased self-reported disgust predicted less approval for moral dilemmas only.
Why did some findings from the first experiment did not replicate in the second? One reason may be that the second experiment had a larger sample and was therefore more sensitive to subtleties in participants’ responses. It also used a more direct measure of moral judgment. In the first experiment, participants gave binary judgments of whether actions were morally permissible, followed by confidence ratings. In contrast, participants in the second experiment rated moral permissibility on a continuous scale. Although both approaches yielded responses on 6-point scales, the meaning of the responses may differ. For example, maximal confidence that an action is impermissible is not the same as rating it maximally impermissible. Perhaps the confidence and extremity of moral judgments have different relations to self-reported disgust.
Although the findings from the second experiment did not support either version of the amplification account, we believe they are explicable. For example, they could indicate that disgust affects how people resolve conflicts among moral principles (i.e., dilemmas involve a conflict between deontological and utilitarian principles, whereas moral transgressions do not). Broadly consistent with this, Cheng et al. (2013) found that disgust consistently increased condemnation of moral dilemmas (plane crash and trolley dilemma) but increased condemnation for only one of two of moral transgressions (lost wallet but not lying on a resume). This said, other explanations for the observed pattern are possible, and the findings need to be replicated and probed further before an explanation can be determined.
More generally, these findings echo other recent work in suggesting that care needs to be taken when investigating effects of disgust, in distinguishing between different kinds of moral items. Many papers have treated moral dilemmas as equivalent to other kinds of moral items (e.g., Cheng et al., 2013; Johnson et al., 2016; Schnall et al., 2008), as did the meta-analysis (Landy & Goodwin, 2015). But the effects of disgust could vary substantially across different kinds of moral items. If so, this could help explain why the effects of disgust on moral judgments have been unreliable—the effects may vary across different kinds of moral judgments.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
JF's contributions to this research were funded by a grant from the Natural Sciences and Engineering Research Council of Canada; OF's contributions to this research were funded by a grant from the Social Sciences and Humanities Research Council of Canada.
