Abstract
We tested the prediction, derived from the hubris hypothesis, that bragging might serve as a verbal provocation and thus enhance aggression. Experiments 1 and 2 were vignette studies where participants could express hypothetical aggression; Experiment 3 was an actual decision task where participants could make aggressive and/or prosocial choices. Observers disliked an explicit braggart (who claimed to be “better than others”) or a competence braggart as compared with an implicit braggart (who claimed to be “good”) or a warmth braggart, respectively. Showing that explicit and competence bragging function as verbal provocations, observers responded more aggressively to the explicit and competence braggart than to the implicit and warmth braggart, respectively. They did so because they inferred that an explicit and a competence braggart viewed other people and them negatively, and therefore disliked the braggart. Rather than praising the self, braggarts are sometimes viewed as insulting others.
Provocation often elicits aggression (Berkowitz, 1989; Bettencourt, Talley, Benjamin, & Valentine, 2006). It may do so by activating hostile thoughts or feelings and by increasing physiological arousal (cf. the General Aggression Model, Anderson & Bushman, 2002). Although provocation may take many forms, ranging from minor frustrations to physical attacks, it is often verbal. Verbal provocation may in turn be overt, as in blatant personal insults, or can be veiled, as in disparagement humor (e.g., Caprara, Passerini, Pastorelli, Renzi, & Zelli, 1986; Janes & Olson, 2000; Zillmann, 1983). Yet, all hitherto studied forms of verbal provocation involve a communicator delivering an unpleasant message to and about others (“observers”). Packed as a serious statement or as a joke, the communicator insults or ridicules one or more observers.
We examined the effect of an even more subtle category of provocations, namely, verbal messages that superficially are not even about the observers. More specifically, we tested the prediction that certain types of bragging (i.e., flattering self-descriptions) provoke aggression. Narcissists are particularly notorious for bragging (e.g., Palmer, Ramsey, Morey, & Gentzler, 2016; Vangelisti, Knapp, & Daly, 1990). Still, they are not the only ones to do so. Many people are likely to brag from time to time, if only because most people believe that they are in many respects better than others (e.g., Hoorens, 1993). If supported, therefore, our prediction uncovers an important and thus far neglected cause of aggression.
The prediction that bragging may function as a verbal provocation rests upon the phenomenon that responses to bragging may depend on how observers believe the braggart views them. Before describing our research, therefore, we briefly review the relevant findings.
The Hubris Hypothesis: Why Self-Superiority Claims May Entail Provocation
Suggestive evidence that observers experience certain types of bragging as provocations comes from research on the hubris hypothesis (Hoorens, Pandelaere, Oldersma, & Sedikides, 2012). This hypothesis was developed to explain the finding that observers respond negatively to braggarts who explicitly compare themselves to others (by claiming to be “better”) whereas they respond positively to braggarts who implicitly compare themselves to others (by claiming to be “good”). This finding is remarkable for two reasons.
First, observers do not interpret explicitly comparative bragging as suggesting a particularly inflated self-view (Hoorens et al., 2012; Hoorens, Van Damme, Helweg-Larsen, & Sedikides, 2017; Van Damme, Hoorens, & Sedikides, 2016). Second, any self-judgment arguably rests on social comparison (e.g., Biernat, 2005; Corcoran & Mussweiler, 2009, 2010). For instance, saying “I am a good leader” is uninformative without the underlying idea “I am a better leader than (some) others.” To reflect the comparative nature of self-judgments, we consider any instance of bragging a “self-superiority claim.” To distinguish between claims that do versus do not emphasize the underlying comparison, we label instances of noncomparatively framed bragging “implicit self-superiority claims” or “implicit bragging” and of comparatively framed bragging “explicit self-superiority claims” or “explicit bragging” (Hoorens & Van Damme, 2012; Van Damme et al., 2016). Accordingly, we distinguish between implicit and explicit braggarts. The hubris hypothesis states that observers’ responses to implicit and explicit braggarts depend on how they believe the braggarts view others and, above all, them (i.e., the observers). More specifically, observers are hypothesized to relatively dislike explicit braggarts because they infer that these braggarts hold a negative view of others and hence of them. In previous studies, mediation analyses confirmed this reasoning (Hoorens et al., 2017; Van Damme, Hoorens, & Sedikides, 2016).
Besides the comparison being hidden or emphasized, one characteristic that determines observers’ reactions is whether braggarts flaunt warmth- or competence-related qualities. Competence (or agency) refers to one’s ability to attain personal goals. Warmth (or communion) refers to one’s ability to form connections with others (Abele & Wojciszke, 2014; Cuddy, Fiske, & Glick, 2008). In individualistic cultures, people’s self-esteem seems to depend on competence-related rather than on warmth-related feedback (Abele & Wojciszke, 2014; Gebauer, Wagner, Sedikides, & Neberich, 2013). Being rejected for one’s alleged incompetence therefore elicits more anger than being rejected for one’s alleged coldness (Çelik, Lammers, van Beest, Bekker, & Vonk, 2013). People also react more aggressively toward individuals calling them incompetent than to individuals calling them cold (DeBono & Muraven, 2014; Experiments 2-4). Consistent with this pattern, observers dislike competence braggarts more than warmth braggarts because they infer that the former view them as lacking in competence (Van Damme & Hoorens, 2016a).
If explicit and competence-related bragging feel like veiled messages about the inferiority of the observers, the latter should feel provoked by the bragging. We therefore predicted that observers would react more aggressively to explicit than to implicit braggarts and more aggressively toward competence braggarts than toward warmth braggarts.
The Present Research
Participants in three experiments read an implicit or an explicit self-superiority claim in the domain of warmth or competence. They judged the braggart and indicated how they thought the braggart would view others and them. Most importantly, all studies included a measure of aggression. Experiments 1 and 2 were vignette studies measuring hypothetical aggression whereas Experiment 3 measured actual aggression toward the braggart.
We predicted that observers would attribute a more negative view of others to an explicit (vs. implicit) braggart and to a competence (vs. warmth) braggart. We also predicted that observers from this inference would deduct that the braggart viewed them (the observers) negatively, and therefore would dislike the braggart (Van Damme et al., 2016). The novel prediction was that this dislike would in turn lead to greater aggression toward explicit than implicit braggarts and toward competence than warmth braggarts (Figure 1).
Causal path explaining aggression as a reaction to an explicit versus implicit self-superiority claim, as proposed by the hubris hypothesis.
In all experiments (a) participants were randomly assigned to the conditions of between-subjects manipulations, (b) there were many more women than men, so that we excluded participants’ gender from the analyses, and (c) we report all measures, manipulations, conditions, exclusions, and significant effects.
Experiment 1
Experiment 1 was a first test of the prediction that explicit (rather than implicit) and competence (rather than warmth) bragging would provoke aggression. Participants, serving as observers, read an explicit or an implicit self-superiority claim reflecting a braggart’s qualities as a friend (a warmth-related role) or as a student (a competence-related role). They indicated how much they liked the braggart and how they believed the braggart viewed others and them. Finally, participants went through a procedure measuring hypothetical aggression.
Among the most commonly used aggression measures is the Hot Sauce Paradigm (Lieberman, Solomon, Greenberg, & McGregor, 1999). Participants prepare samples of hot sauce to be eaten by an “other participant” who hates spicy tastes, with the size of the sample counting as a measure of aggression. In one adaptation of this paradigm (DeBono & Muraven, 2014; Experiment 2-4), participants determine how long another individual has to do math work, knowing that this individual hates math. These and other procedures have been criticized, however, for insufficiently allowing participants to react nonaggressively or even prosocially (e.g., Saleem, Anderson, & Barlett, 2015).
To avoid this limitation, we had participants imagine that they were in a position to assign coursework to the braggart, and listed three courses that the braggart should work on. They could express aggressive choices by selecting among the braggart’s nonpreferred courses. Although this approach has the disadvantage of measuring hypothetical rather than actual behavior, it has many advantages: it includes prosocial options, is easy to administer, and requires neither deception nor many materials.
Method
Participants and design
We tested 450 participants (363 women; Mage = 18.67, SDage = 2.00) in a 2 (claim type: implicit, explicit) × 2 (claim domain: friendship, studentship) between-subjects design. Participants were first year psychology students at a large Belgian university, who fulfilled a course option. An a priori power analysis using G*Power (Faul, Erdfelder, Lang, & Buchner, 2007) had shown that 210 participants should suffice to detect main effects with a medium effect size (i.e., f = .25; Cohen, 1969) at an alpha level of .05 in a 2 × 2 design. We had access to a group of more than double this size, and decided to include them all so that the study could serve as an experienced example in a class on methods.
Materials and procedure
Participants took part in groups of 24 to 29. Each received a booklet containing all instructions and materials. The instructions prompted participants to imagine someone, also a student of psychology, making a certain claim. They then read a self-superiority claim that was implicit or explicit and described the student’s qualities as a friend or a student. We used verbatim the self-superiority claims from previous research on the hubris hypothesis (Van Damme & Hoorens, 2016a; Studies 2-3). For instance, the claim about studentship read as follows, with the differences between the explicit and implicit claim appearing in square brackets and the explicit claim being first in order:
You know, I am a [better/good] student [than others are] . . . I prepare classes and workshops [more] thoroughly. I also [more] often succeed in bringing assignments to a good end. I have original ideas and spend a lot of time looking up relevant information, even if the subject doesn’t really appeal to me. I find it [easier/easy] to process large amounts of study material. If I [compare myself to others/look at myself], I may well say that I am [more] organized, diligent and inventive and that I have [more/a lot of] insight.
Participants responded to affiliation questions adapted from Helweg-Larsen, Sadeghian, and Webb (2002; Hoorens et al., 2017). Two of these here served as an unobtrusive measure of liking, namely, if the braggart was pleasant to be around (1 = not at all, 7 = definitely pleasant) and if participants would like to have the braggart as a friend (1 = not at all, 7 = very much). The remaining three served as filler items. Next, participants rated how they believed the braggart viewed others (braggart view of others) and them (braggart view of observer) on the seven warmth-related traits (e.g., helpful) and seven competence-related traits (e.g., decisive) from Van Damme, Hoorens, and Sedikides (2016; Study 2). Participants responded on 7-point rating scales ranging from 1 (not at all) to 7 (very much). 1
Finally, participants completed the measure of hypothetical aggression. They learned that the braggart was a female psychology student who had completed most courses of the first bachelor year. The text included several details to enliven her image, including that she had commuted but was now living in a dorm. Participants then learned that a group of students, including the braggart, had ranked the courses that they had taken. Participants saw the braggart’s ranking of 10 courses, with those at the top being the ones that she liked the most and that those at the bottom being the ones she hated. Participants imagined that the braggart had to write a paper on the topic of one course and that they were in the position to decide which one. They were requested to suggest three courses in their order of preference.
Results
Three participants (two from the explicit/friendship condition and one from the implicit/studentship condition) did not complete all ratings and were excluded from analyses. The means associated with the main effects appear in Table 1.
Aggression Toward, Liking and Respect for, and View of Others and the Observer Attributed to a Braggart as a Function of the Content (Studentship vs. Friendship) and Comparative Nature (Explicit vs. Implicit) of the Bragging.
Note. All explicit–implicit and student–friend comparisons are significant except where the means are in italics.
Experiments 1-2: Hypothetical aggression scores based on the participants’ first choice. Experiment 3: Actual scores of aggression (hurt) and prosociality (help).
Hypothetical aggression
We examined the rank number of participants’ first choices, which could vary from 1 (the braggart’s favorite course, a maximally prosocial choice) to 10 (the braggart’s least favorite course, a maximally aggressive choice). 2 We subjected these scores to a 2 × 2 analysis of variance (ANOVA) with claim type (implicit, explicit) and claim domain (friendship, studentship) as between-subjects variables. We found main effects of claim type, F(1, 443) = 13.39, p < .001, ηp2 = .029, and claim domain, F(1, 443) = 10.40, p = .001, ηp2 = .023. Participants reported that they would assign less pleasant courses to an explicit than to an implicit braggart, and to a studentship than to a friendship braggart.
Braggart liking
We subjected the liking scores (α = .90) to a 2 × 2 ANOVA with claim type (implicit, explicit) and claim domain (friendship, studentship) as between-subjects variables. We found a main effect of claim type, F(1, 443) = 89.18, p < .001, ηp2 = .168. As predicted, participants disliked an explicit braggart (M = 2.79, SD = 1.26) more than an implicit braggart (M = 3.92, SD = 1.42). We also found a main effect of claim domain, F(1, 443) = 68.74, p < .001, ηp2 = .134. Participants disliked a studentship braggart (M = 2.86, SD = 1.15) more than a friendship braggart (M = 3.85, SD = 1.56). The Claim Type × Claim Domain interaction was not significant, F(1, 443) = 0.97, p = .326.
Braggart view of others
We subjected mean ratings for the braggart’s view of others’ warmth (α = .90) and competence (α = .92) to a 2 × 2 × 2 ANOVA with claim type (implicit, explicit) and claim domain (friendship, studentship) as between-subjects variables and rating dimension (warmth, competence) as a within-subjects variable. We found main effects of claim type, F(1, 443) = 68.50, p < .001, ηp2 = .134, and claim domain, F(1, 443) = 58.05, p < .001, ηp2 = .116. Participants attributed a more negative view of others to an explicit braggart than to an implicit braggart and to a studentship braggart than to a friendship braggart.
The main effects were qualified by two interactions: Claim Type × Claim Domain, F(1, 443) = 5.67, p = .018, ηp2 = .013, and Claim Type × Claim Domain × Rating Dimension, F(1, 443) = 27.17, p < .001, ηp2 = .058. Participants attributed a more negative view of others’ warmth and competence to explicit than to implicit braggarts. If the claim was about friendship, the difference was larger for warmth (Mexplicit = 3.89, SD = 1.20; Mimplicit = 5.06, SD = 0.80), t(222) = 8.61, p < .001, than for competence (Mexplicit = 3.70, SD = 1.25; Mimplicit = 4.29, SD = 1.05), t(222) = 3.85, p < .001. If the claim was about studentship, the difference was larger for competence (Mexplicit = 2.90, SD = 1.09; Mimplicit = 3.64, SD = 1.05), t(221) = 5.16, p < .001, than for warmth (Mexplicit = 3.82, SD = 0.90; Mimplicit = 4.06, SD = 0.85), t(221) = 2.02, p = .044. The rating dimension main effect was also significant, F(1, 443) = 156.14, p < .001, ηp2 = .323.
Braggart view of observer
We subjected the ratings for the braggart’s view of the observer’s warmth (α = .88) and competence (α = .86) to a similar analysis. We found main effects of claim type, F(1, 443) = 10.18, p = .002, ηp2 = .022, and claim domain, F(1, 443) = 26.41, p < .001, ηp2 = .056. Participants attributed a more negative view of them to an explicit than an implicit braggart and to a studentship than a friendship braggart.
These main effects were qualified by a Claim Type × Claim Domain × Rating Dimension interaction, F(1, 443) = 4.14, p = .042, ηp2 = .009. If the claim was about friendship, the explicit–implicit difference was significant for warmth (Mexplicit = 4.93, SD = 0.92; Mimplicit = 5.36, SD = 0.88), t(222) = 3.60, p < .001, but not for competence (Mexplicit = 4.12, SD = 1.10; Mimplicit = 4.30, SD = 0.96), t(222) = 1.34, p = .183. If the claim was about studentship, an almost significant explicit–implicit difference occurred for competence (Mexplicit = 3.61, SD = 1.08; Mimplicit = 3.86, SDi = 0.93), t(221) = 1.89, p = .061, but not for warmth (Mexplicit = 4.77, SD = 0.82; Mimplicit = 4.88, SD = 0.96), t(221) = 0.91, p = .362. The rating dimension main effect was also significant, F(1, 443) = 156.14, p < .001, ηp2 = .323.
Mediational analyses
We tested a serial multiple mediation model to test the causal path that according to the hubris hypothesis leads from claim type to aggression (Figure 1). We applied the bootstrapping method using Hayes’ (2013) PROCESS macro for SPSS (version 2.15; model 6). We entered hypothetical aggression as the dependent variable, claim type as a dummy-coded independent variable (0 = implicit, 1 = explicit), and braggart view of others, braggart view of participant, and liking as serial mediators (in that order). This analysis also tested all indirect effects including only one or two of the proposed mediators (in the order in which they were entered). Bootstrap estimates are based on 5,000 resamples and all reported tests are two-tailed t tests.
The total direct effect of claim type on hypothetical aggression was significant, B = 0.93, t(445) = 3.68, p < .001. The predicted indirect effect representing the full causal path was also significant, point estimate 0.05, 95% CI [0.02, 0.10]. We furthermore found an indirect effect of claim type via braggart view of others and braggart liking, point estimate 0.12, 95% CI [0.06, 0.23], and an indirect effect of claim type via braggart liking, point estimate 0.30, 95% CI [0.15, 0.49]. When the mediators were controlled for, the direct effect of claim type on aggression was no longer significant, B = 0.34, t(445) = 3.68, p = .221.
We also conducted a mediation analysis on the effect of claim domain on aggression. We entered hypothetical aggression as the dependent variable, claim domain as a dummy-coded independent variable (0 = friendship, 1 = studentship), and braggart view of others’ competence, braggart view of the participant’s competence, and braggart liking as serial mediators (in that order). The total direct effect of claim domain on aggression was significant, B = 0.82, t(445) = 3.24, p = .001, as was the predicted indirect effect representing the full causal path, point estimate 0.02, 95% CI [0.01, 0.05]. The indirect effect of claim domain via braggart view of the participant’s competence and braggart liking was also significant, point estimate 0.02, 95% CI [0.01, 0.06]. We furthermore found an indirect effect of claim domain via braggart view of others’ competence and braggart liking, point estimate 0.12, 95% CI [0.06, 0.22], and an indirect effect of claim domain via braggart liking, point estimate 0.27, 95% CI [0.15, 0.45]. When the mediators were controlled for, the direct effect of claim domain on aggression was no longer significant, B = 0.34, t(445) = 1.27, p = .203.
Discussion
The novel finding of Experiment 1 was that observers show more hypothetical aggression toward explicit than implicit braggarts and more hypothetical aggression toward competence than warmth braggarts. This greater aggression is the outcome of a chain of judgments connecting claim type or domain, inferred view of others, inferred view of observer, and claimant liking. Observers’ inferences from explicit (more than implicit) and competence (more than warmth) self-superiority claims that the braggart views other people, and hence them negatively, and observers’ dislike for the braggart replicated previous research on the hubris hypothesis.
One limitation of Experiment 1 was that we measured hypothetical aggression after participants had judged the braggart and the braggart’s alleged view of others and them. It is possible, therefore, that the effect of bragging type and domain on hypothetical aggression occurred because participants’ inferences about the braggart had been primed. Another limitation was that the choice options in the measure of hypothetical aggression were competence-related (courses). Assigning a nonfavorite course to a braggart could serve two goals: giving a not-so-pleasant task or making the braggart underperform. As participants could curtail the braggart’s competence-related performance but not the braggart’s warmth-related “performance,” the design may have favored aggression against competence braggarts. We addressed these limitations in Experiment 2.
Experiment 2
Experiment 2 tested the predictions using a refined design. All participants saw the braggart’s ranking of activities before completing the measurements in a counterbalanced order. The measure of hypothetical aggression came either first or last and was adapted to involve activities that were unrelated to the contents of the bragging.
Method
Participants and design
We tested 337 first year social sciences students at a large Belgian university (210 women; Mage = 19.22, SDage = 1.70) in a 2 (claim type: implicit, explicit) × 2 (claim domain: friendship, studentship) × 2 (measure order: aggression first, aggression last) between-subjects design.
Materials and procedure
In mass testing sessions of 75 to 85 participants, each participant received a booklet containing all stimulus materials. Except for the measure of hypothetical aggression, all stimuli and measures were identical to those in Experiment 1. In the introduction, participants read that they would learn about another student (of unspecified gender, unlike Experiment 1) at their university. They read a quote where a student described life at the dorm and explained that residents took turns on a number of chores. The quote continued: “This is my ranking of the chores according to how much I like them. The chores at the top are the ones I like the most; the ones at the bottom are the ones I hate.” Participants then saw the student’s ranking of 10 chores (e.g., “Organizing a party at the dorm,” “Cleaning the bathrooms”). Half of the participants started with the measure of hypothetical aggression. The other half first rated the braggart and how they believed the braggart viewed other people and them.
Results
One participant (implicit/friendship/aggression first) did not complete all ratings and was excluded from analyses. The means associated with main effects appear in Table 1.
Hypothetical aggression
We analyzed the aggression data as in Experiment 1. The ANOVA on participants’ first choices with claim type (implicit, explicit), claim domain (friendship, studentship), and measure order (aggression first, aggression last) as between-subjects variables yielded main effects of claim type, F(1, 328) = 23.64, p < .001, ηp2 = .067, and claim domain, F(1, 328) = 11.69, p = .001, ηp2 = .034. Participants indicated that they would assign less pleasant chores to an explicit than to an implicit braggart, and to a studentship than to a friendship braggart.
Neither claim domain nor claim type was involved in interactions with measure order, all Fs < .2. Still, we tested if their effects were significant within the measure order conditions. When aggression was measured first, the main effect of claim type was even stronger, F(1, 165) = 15.02, p < .001, ηp2 = .083, than when it was measured last, F(1, 163) = 9.32, p = .003, ηp2 = .054. The main effect of claim domain was also stronger when aggression was measured first, F(1, 165) = 6.43, p = .012, ηp2 = .038, than when it was measured last, F(1, 163) = 5.36, p = .022, ηp2 = .032.
Braggart liking
We subjected observers’ liking for the braggart (α = .81) to a 2 × 2 × 2 ANOVA with claim type (implicit, explicit), claim domain (friendship, studentship), and measure order (aggression first, aggression last) as between-subjects variables. The ANOVA yielded a main effect of claim type, F(1, 328) = 37.69, p < .001, ηp2 = .103. Participants disliked an explicit braggart (M = 4.02, SD = 1.22) more than an implicit braggart (M = 4.71, SD = 0.93). The claim domain main effect was also significant, F(1, 328) = 34.13, p < .001, ηp2 = .094. Participants disliked a studentship braggart (M = 4.03, SD = 1.11) more than a friendship braggart (M = 4.69, SD = 1.07). No other effects, including those involving measure order, were significant, Fs < 3.29, ps > .071.
Braggart view of others
We subjected mean ratings for the braggart’s view of others’ warmth (α = .88) and competence (α = .86) to a 2 × 2 × 2 × 2 ANOVA with claim type (implicit, explicit), claim domain (friendship, studentship), and measure order (aggression first, aggression last) as between-subjects variables, and rating dimension (warmth, competence) as a within-subjects variable. We found main effects of claim type, F(1, 328) = 66.96, p < .001, ηp2 = .170, and claim domain, F(1, 328) = 15.14, p < .001, ηp2 = .044. Participants attributed a more negative view of others to an explicit than to an implicit braggart, and to a studentship than to a friendship braggart.
The main effects were qualified by a Claim Type × Claim Domain × Rating Dimension interaction F(1, 328) = 11.44, p = .001, ηp2 = .034. Participants attributed a more negative view of others’ warmth and competence to an explicit than to an implicit braggart. If the claim was about friendship, the difference was larger for warmth (Mexplicit = 4.08, SD = 0.93; Mimplicit = 4.95, SD = 0.69), t(166) = 6.91, p < .001, than for competence (Mexplicit = 3.85, SD = 0.82; Mimplicit = 4.38, SD = 0.78), t(166) = 4.34, p < .001. If the claim was about studentship, the difference was larger for competence (Mexplicit = 3.44, SD = 0.82; Mimplicit = 4.11, SD = 0.81), t(166) = 5.29, p < .001, than for warmth (Mexplicit = 4.05, SD = 0.83; Mimplicit = 4.47, SD = 0.69), t(166) = 3.56, p < .001. Except for the main effect of rating dimension, F(1, 328) = 104.70, p < .001, ηp2 = .242, no other effects were significant, Fs < 1.83, ps > .178.
Braggart view of observer
The analysis on the ratings for the braggart’s view of the observer’s warmth (α = .86) and competence (α = .83) yielded a main effect of claim type, F(1, 328) = 33.30, p < .001, ηp2 = .092. Participants attributed a more negative view of them to an explicit than to an implicit braggart. The main effect of rating dimension was also significant, F(1, 328) = 156.14, p < .001, ηp2 = .323, with participants inferring that the braggarts had a more negative view of their competence (M = 4.27, SD = 0.90) than of their warmth (M = 4.94, SD = 0.81). No other effects were significant, Fs < 2.64, ps > .105.
Mediation analyses
We tested the serial mediation model from claim type to aggression as in Experiment 1. The results were highly similar. The total direct effect of claim type on aggression was significant, B = 1.36, t(334) = 4.77, p < .001. The predicted indirect effect including all mediators was also significant, point estimate 0.04, 95% CI [0.01, 0.11]. As in Experiment 1, we also found an indirect effect of claim type via braggart view of others and braggart liking, point estimate 0.22, 95% CI [0.09, 0.42]; an indirect effect of claim type via braggart view of others, point estimate 0.56, 95% CI [0.23, 0.96]; and an indirect effect of claim type via braggart liking, point estimate 0.11, 95% CI [0.01, 0.30]. None of the other indirect effects was significant, 95% CIs ⊂ [−0.32, 0.10]. When the mediators were controlled for, the direct effect of claim type on aggression was marginally significant, B = 0.54, t(334) = 1.85, p = .066.
We also tested the same serial mediation model from claim domain to aggression as in Experiment 1. The total direct effect of claim domain on aggression was significant, B = 0.96, t(334) = 3.30, p = .001. The predicted indirect effect representing the full causal path failed to reach significance, point estimate 0.01, 95% CI [−0.01, 0.04]. The indirect effect of claim domain via braggart view of others’ competence and braggart liking was significant, point estimate 0.11, 95% CI [0.04, 0.24]. We also found an indirect effect of claim domain via braggart view of others’ competence, point estimate 0.17, 95% CI [0.03, 0.42], and an indirect effect of claim domain via braggart liking, point estimate 0.30, 95% CI [0.14, 0.57]. No other indirect effects were significant, 95% CIs ⊂ [−0.07, 0.07]. When the mediators were controlled for, the direct effect of claim domain on aggression was no longer significant, B = 0.36, t(334) = 1.27, p = .207.
Discussion
Experiment 2 replicated the main findings of Experiment 1. Observers expressed greater hypothetical aggression toward the explicit (vs. the implicit) braggart and toward the competence (vs. the warmth) braggart. They did so because the negative view of others and of them that they attributed to the explicit and competence braggart made them dislike these braggarts.
Importantly, Experiment 2 disambiguated the findings of Experiment 1 in two manners. One is that it showed that the effects were not limited to situations where observers were encouraged to think about the braggart before expressing hypothetical aggression. The second was that it also showed that these effects were not limited to situations where the aggression measure was related to the domain of a claim.
Still, Experiment 1 and Experiment 2 had an important limitation, namely, that the measure of aggression was hypothetical. Our findings may thus reflect how people think they would respond rather than their actual responses. Experiment 3 addressed this limitation.
Experiment 3
Experiment 3 took our test of the extension of the hubris hypothesis a step further by examining actual behaviors. More specifically, we used the Tangram Help/Hurt task (Saleem et al., 2015; Saleem, Barlett, Anderson, & Hawkins, 2017), which allows researchers to calculate scores of prosocial behavior (through a helping score) and aggressive behavior (through a hurting score).
Tangram puzzles consist of seven geometric pieces which can be used to form a specified shape. The more pieces required, the more difficult and time-consuming a tangram puzzle is. In the Tangram Help/Hurt task, participants assign 11 out of 30 puzzles of varying difficulty to another participant, knowing that the latter will be eligible to win a prize if he/she completes 10 puzzles in 10 min. The participant can help by selecting easy tangrams, or hurt by selecting difficult ones. Showing that the Tangram Help/Hurt task is sufficiently sensitive to reveal effects of verbal provocation on aggression, participants in several studies made more hurtful choices in response to negative feedback (Saleem et al., 2015, Studies 5-6).
Method
Participants and design
We tested 266 first year psychology students fulfilling a course option at a large Belgian university. All were subjected to a 2 (claim type: implicit, explicit) × 2 (claim domain: friendship, studentship) × 2 (measure order: aggression first, aggression last) between-subjects design. In 15 cases, participants’ answers to the debriefing questions revealed that they were suspicious about whether someone would truly have to solve the tangram puzzles they assigned. The data of suspicious participants and one participant who reported not having understood the task were excluded from the analyses. Of the remaining 250 participants, 209 were female and two did not indicate their gender (17-29 years, Mage = 18.38, SD = 1.15).
Materials and procedure
Participants took part in an online testing session consisting of this and an unrelated study. The order in which participants did the two studies was counterbalanced. We presented all materials using LimeSurvey (www.limesurvey.org).
Participants first went through the informed consent procedure. They read that another student (of unspecified gender) at their university had during an experimental session written about himself/herself. Participants would have to judge that other participant and assign tasks for him or her to do in another experimental session. They then read the self-superiority claim that the other participant had allegedly made. The claims were identical to those used in Experiments 1 and 2. Just as in Experiments 1 and 2, participants could revisit the claim during the measures.
Except for the measure of aggressive behavior, all measures were identical to those in Experiments 1 and 2. We used the online version of the original Tangram Help/Hurt task (Saleem et al., 2015). Participants read an explanation of tangram puzzles, including the fact that their difficulty level depends on the number of forms that are needed to solve them, and watched an instruction video showing someone solving several ones (borrowed from the supplementary materials of Saleem et al., 2015). To ensure that they watched the video, we announced that we would ask a question about it. After participants had done the Tangram Help/Hurt task, we asked how many human faces featured on it (the correct answer was zero). We also asked whether the participants had experienced problems with the sound and/or image of the video, and if they had understood the task (right after participants had watched the video). Eleven participants indicated that they had experienced problems with their video and/or audio player and 11 participants did not correctly answer the question about the video. All of these participants did indicate, however, that they understood the task.
Next, participants were informed that the participant whose self-description they had read could win a gift coupon of €25 if he/she would be able to solve 10 tangram puzzles within 10 min. Participants were asked to assign 11 tangrams to this other participant, who would never know who had assigned the puzzles, by clicking on them. The various tangrams were aligned according to their difficulty (easy–medium–hard) in a table similar to the one used by Saleem et al. (2015).
At the end of the session, participants responded to statements about their motives to select specific tangrams: “I wanted to give the participant a variety of tangrams,” “I wanted to give the participant tangrams that were as similar to each other as possible,” “I wanted to help the participant win the prize,” “I wanted to make it difficult for the participant to win the prize,” “I wanted to give the participants beautiful tangrams,” “I wanted to give the participants easy tangrams,” and “I wanted to give the participants difficult tangrams.” Participants answered on a scale ranging from 1 (strongly disagree) to a 5 (strongly agree). Finally, they answered open-ended questions designed to assess suspicion (based on the funneled debriefing procedure reported in Saleem et al., 2015): “Do you have any questions or remarks (of any sort) about this study?,” “Which prediction do you think do we want the test with this study?,” and “Was there something that you found strange, confusing, or uncomfortable regarding what you had to do?.” Participants were then thanked for their participation.
Results
The means associated with the main effects appear in Table 1
Aggression
We calculated a helping score (the number of easy tangrams selected minus 1), a hurting score (the number of difficult tangrams selected minus 1), and a difference score (the hurting score minus the helping score; Saleem et al., 2015). We subjected each score to a 2 × 2 × 2 ANOVA with claim type (implicit, explicit), claim domain (friendship, studentship), and measure order (aggression first, aggression last) as between-subjects variables.
On the hurting score, the claim type main effect was significant, F(1, 242) = 12.22, p = .001, ηp2 = .048, as was the claim domain main effect, F(1, 242) = 4.62, p = .033, ηp2 = .019. Participants selected more difficult tangrams for an explicit than for an implicit braggart, and more difficult tangrams for a studentship than for a friendship braggart.
On the helping score, we also found main effects of claim type, F(1, 242) = 9.22, p = .003, ηp2 = .037, and claim domain, F(1, 242) = 5.23, p = .023, ηp2 = .021. Participants selected fewer easy tangrams for an explicit than for an implicit braggart and fewer easy tangrams for a studentship than for a friendship braggart. Not surprisingly, therefore, the difference score also yielded main effects of claim type, F(1, 242) = 15.23, p < .001, ηp2 = .057, and claim domain, F(1, 242) = 8.85, p = .003, ηp2 = .033.
Liking
The 2 × 2 × 2 ANOVA on the liking scores (α = .89) with claim type (implicit, explicit), claim domain (friendship, studentship), and measure order (aggression first, aggression last) as between-subjects variables yielded significant main effects of claim type, F(1, 239) = 41.37, p < .001, ηp2 = .148, and claim domain, F(1, 239) = 12.53, p < .001, ηp2 = .050, which were qualified by an interaction of claim type, claim domain, and measure order, F(1, 239) = 5.11, p = .025, ηp2 = .021.
Participants liked an explicit braggart less than an implicit braggart and a studentship braggart less than a friendship braggart. They generally liked the claimant less if they had first assigned tangrams than if they had not. The difference was significant in the studentship–implicit condition (Maggression first = 4.09, SD = 1.14; Maggression last = 4.73, SD = 0.98), t(65) = 2.49, p = .015, but not in the friendship–implicit condition (Maggression first = 4.85, SD = 1.32; Maggression last = 5.19, SD = 1.00), t(61) = 1.11, p = .272, or in the friendship–explicit condition (Maggression first = 3.64, SD = 1.60; Maggression last = 4.21, SD = 1.73), t(57) = 1.30, p = .199. Only in the studentship–explicit condition did participants report that they liked the claimant more if they had first assigned tangrams than if they had not (Maggression first = 3.77, SD = 1.25; Maggression last = 3.03, SD = 1.25), t(56) = 1.95, p = .056.
Braggart view of others
We calculated mean ratings for the braggart’s view of others’ warmth (α = .94) and competence (α = .94). We subjected these scores to a 2 × 2 × 2 × 2 ANOVA with claim type (implicit, explicit), claim domain (friendship, studentship), and measure order (aggression first, aggression last) as between-subjects variables and rating dimension (warmth, competence) as a within-subjects variable. We found main effects of claim type, F(1, 239) = 43.69, p < .001, ηp2 = .15, and claim domain, F(1, 239) = 11.16, p = .001, ηp2 = .045. Participants attributed a more negative view to an explicit than to an implicit braggart and a less positive view to a studentship than to a friendship braggart.
The main effects were qualified by a Claim Type × Rating Dimension interaction, F(1, 239) = 5.13, p = .024, ηp2 = .021, and a Claim Type × Claim Domain × Rating Dimension interaction, F(1, 239) = 23.89, p < .001, ηp2 = .091. Participants generally believed that explicit braggarts viewed others’ warmth and competence more negatively than implicit braggarts. If the claim was about friendship, the difference was larger for warmth (Mexplicit = 4.04, SD = 1.16; Mimplicit = 4.99, SD = 0.96), t(120) = 4.95, p < .001, 95% CI [0.57, 1.33], than for competence (Mexplicit = 3.92, SD = 1.08; Mimplicit = 4.62, SD = 1.01), t(120) = 3.65, p < .001, 95% CI [0.32, 1.07]. If the claim was about studentship, the difference was larger for competence (Mexplicit = 3.24, SD = 1.23; Mimplicit = 4.35, SD = 0.93), t(123) = 5.76, p < .001, 95% CI [0.73, 1.49], than for warmth (Mexplicit = 4.00, SD = 0.91; Mimplicit = 4.40, SD = 0.83), t(123) = 2.61, p = .010, 95% CI [0.10, 0.72]. Except for a main effect of rating dimension, F(1, 240) = 42.60, p < .001, ηp2 = .149, no other effects were significant, Fs < 3.43, ps > .065.
Braggart view of observer
We subjected the ratings for the braggart’s view of the observer’s warmth (α = .94) and competence (α = .89) to a similar analysis as the ratings for the braggart’s view of others.
We found main effects of claim type, F(1, 239) = 32.70, p < .001, ηp2 = .120, and claim domain, F(1, 239) = 3.93, p = .049, ηp2 = .016. Participants attributed a more negative view of them to an explicit than to an implicit braggart and to a studentship than to a friendship braggart. The main effects were qualified by a marginally significant Claim Type × Claim Domain × Rating Dimension interaction, F(1, 239) = 3.71, p = .055, ηp2 = .015. If the claim was about friendship, the explicit–implicit difference was larger for warmth (Mexplicit = 4.54, SD = 1.19; Mimplicit = 5.37, SD = 0.80), t(120) = 4.56, p < .001, 95% CI [0.47, 1.20], than for competence (Mexplicit = 3.96, SD = 1.05; Mimplicit = 4.55, SD = 0.91), t(120) = 3.29, p = .001, 95% CI [0.23, 0.94]. If the claim was about studentship, it was larger for competence (Mexplicit = 3.58, SD = 0.99; Mimplicit = 4.25, SD = 0.91), t(123) = 3.98, p < .001, 95% CI [0.34, 1.01], than for warmth (Mexplicit = 4.62, SD = 1.01; Mimplicit = 5.05, SD = 0.96), t(123) = 2.44, p = .016, 95% CI [0.08, 0.78]. Except for a main effect of rating dimension, F(1, 239) = 162.89, p < .001, ηp2 = .405, no other effects were significant, Fs < 3.41, ps > .066.
Motives
Participants’ answers to the open question proved hard to unequivocally code as many were quite ambiguous. For instance, several participants indicated that they had chosen rather difficult tangrams to provide the individual with a pleasantly challenging task or that they had strived to give the individual an easy, but not too easy task [as to be boring]. We therefore did not analyze these data.
We subjected participants’ answers to the motives questions to ANOVAs with claim type (implicit, explicit) and claim domain (friendship, studentship) as between-subjects variables. We found main effects of claim type for the items “I wanted to help the participant win the prize,” “I wanted to make it difficult for the participant to win the prize,” “I wanted to give the participants easy tangrams,” and “I wanted to give the participants difficult tangrams”: Fs(1, 243) > 5.80, ps < .02; ηp2 > .023. On the latter two motives, we also found main effects of claim domain, Fs(1, 243) > 4.95, ps < .03; ηp2 > .020. As compared with the implicit braggart, participants reported they wanted to help the explicit braggart less (Mexplicit = 3.21, SD = 1.32; Mimplicit = 3.67, SD = 1.33) and wanted less to give this braggart easy tangrams (Mexplicit = 2.73, SD = 1.42; Mimplicit = 3.15, SD = 1.42). On the contrary, they reported they wanted to make it more difficult to win the prize (Mexplicit = 2.61, SD = 1.43; Mimplicit = 2.04, SD = 1.20) and wanted more to give this braggart difficult tangrams (Mexplicit = 2.79, SD = 1.39; Mimplicit = 2.22, SD = 1.45). As compared with the warmth braggart, participants reported that they wanted less to give the competence braggart easy tangrams (Mstudentship= 2.76, SD = 1.42; Mfriendship= 3.15, SD = 1.43) and more to give this braggart difficult tangrams (Mstudentship = 2.73, SD = 1.38; Mfriendship = 2.25, SD = 1.26). The other motives items did not yield any significant differences.
Mediation analysis
We tested the same serial mediation models as in the previous experiments, using the difference between the hurting and helping scores as the dependent variable. All path coefficients are shown in Figure 2.
Path coefficients in a serial multiple mediation model predicting aggression (difference scores) from claim type via braggart view of others, braggart view of observer, and braggart liking in Experiment 3.
The total direct effect of claim type on aggression was significant, B = 2.41, t(245) = 3.38, p = .008. Importantly, the predicted indirect effect including all mediators was also significant, point estimate 0.32, 95% CI [0.10, 0.70]. In addition, we found indirect effects via braggart view of others and liking to aggression, point estimate 0.47, 95% CI [0.18, 0.93]; via braggart view of observer and liking to aggression, point estimate 0.13, 95% CI [0.02, 0.33]; and via liking to aggression, point estimate 0.62, 95% CI [0.17, 1.31]. When the mediators were controlled for, the direct effect of claim type on aggression was not significant, B = 0.63, t(245) = 0.86, p = .390. The total direct effect of claim domain on aggression was significant, B = 1.53, t(245) = 2.12, p = .05. Importantly, the predicted indirect effect including all mediators was significant, point estimate 0.11, 95% CI [0.02, 0.31]. We also found indirect effects via braggart view of others’ competence and liking to aggression, point estimate 0.29, 95% CI [0.10, 0.70]. When controlling for the mediators, the direct effect of claim type on aggression was not significant, B = 0.55, t(242) = 0.82, p = .415.
Discussion
Experiment 3 replicated the findings of Experiments 1 and 2. Explicit and competence bragging do not only make observers dislike the braggart because they believe that the braggart views others and them negatively, but also provoke aggression. Experiment 3 showed that this aggression-provoking effect is not limited to hypothetical aggression but extends to actual behavior and goes hand in hand with a reduction of prosocial behavior.
General Discussion
Insulting or ridiculing messages provoke aggression. We extended the register of verbal provocations with an understudied category of provocative messages: self-superiority claims that are overtly comparative or flaunt the braggart’s competence. These types of bragging provoke aggressive behavior and inhibit prosocial behavior toward the braggart.
The Subtlest Provocation?
The ability of explicitly comparative or competence-related bragging to elicit aggression is fascinating because it ostensibly is not even about the addressed individuals. Braggarts may easily live in the conviction that they have not said anything about anyone but themselves. If the observer responds aggressively, the link with the bragging may completely escape the braggart.
The difference between reactions to explicit and implicit bragging and between reactions to competence and warmth bragging is the more intriguing as the braggart’s underlying self-evaluation arguably rests upon social comparison. Still, observers react to explicit (more than implicit) and competence (more than warmth) bragging with greater aggression. The reason is that they interpret these types of bragging as conveying a negative view of others and therefore of them, which makes them dislike the braggart. In other words, explicit and competence-related self-superiority claims are viewed as insulting the observer rather than as flattering the claimant.
The Hubris Hypothesis Extended
Our research supports and extends the hubris hypothesis. It supports it by replicating the causal path from an explicit versus an implicit self-superiority claim to dislike for the braggart via an inferred negative view of other people and them (i.e., the observer). It extends it by showing that observers’ dislike for explicit braggarts encourages aggression toward these braggarts. Hence, negative reactions to explicit self-superiority claims are not merely cognitive and affective, but translate into interpersonal behavior.
Our research furthermore shows that observers react more aggressively to braggarts if these flaunt their competence rather than their warmth. This novel finding is consistent with the finding that observers dislike competence more than warmth braggarts (Van Damme & Hoorens, 2016a), that people react more aggressively to individuals who view them as incompetent than to individuals who view them as cold (DeBono & Muraven, 2014), and that being rejected because of a lack of competence makes people angrier than being rejected because of a lack of warmth (Çelik et al., 2013).
Measuring Aggression Through Preferences and Decisions for Others’ Activities
Observers did not react very aggressively toward claimants overall. Modest levels of aggression are common in laboratory experiments. It is important to remember, however, that the crucial finding was that observers show more aggression to explicit than to implicit braggarts and to competence than to warmth braggarts.
Our paradigms are, first, adaptations of the Hot Sauce Paradigm (Experiments 1-2) and, second, the recently developed Tangram Help/Hurt task (Experiment 3). Besides not requiring deception or cover stories, the former are easy to use, require little preparation (for researchers), and are quick to complete (for participants). They can also easily be adapted to research circumstances and populations. Their disadvantages are that they measure hypothetical rather than actual aggression and that it is difficult to disentangle aggressive or prosocial choices. The Tangram Help/Hurt task addresses these issues by examining actual choices and yielding separate scores for aggressive and prosocial behaviors. Interestingly, the three experiments yielded highly consistent results. This observation is consistent with DeWall, Baumeister, Stillman, and Gailliot (2007), who also observed a great similarity between determinants of actual and hypothetical aggression.
Limitations and Directions for Further Research
Our samples predominantly included women. Men generally show more aggression than women (for a review, see Archer, 2004). Our research may thus have provided a rather low estimate of the aggression that some types of bragging elicit. It is also possible, however, that men find explicitly comparative competence bragging more acceptable than women do. If this is true, our estimate may have been rather high. As noted in the introduction, the composition of our participant pool did not allow solid tests of gender differences. To explore if these are likely in more balanced studies, however, we exploratorily reanalyzed the liking and aggression scores in Experiments 1 and 2 including gender as an independent variable (in Experiment 3, some cells included no more than two or three men). Claim type and claim content did not interact with gender.
The verbal provocation in our research occurred in a context without any other stressors. Research on the provocation–aggression link has shown that stressors (such as noise; Donnerstein & Wilson, 1976) enhance provocation effects. Therefore, one intriguing question is whether hearing people explicitly brag or brag about their competence at, say, a crowded reception might elicit even more aggression than in a quieter setting.
On the other hand, explicitly comparative and competence bragging do not always need to provoke aggression. There may be situations where they are considered appropriate or at least acceptable, such as during job interviews. One would expect that observers in those situations would respond less harshly to braggarts, but whether or not they do so remains a question. Previous research on the hubris hypothesis has furthermore shown that observers dislike explicit braggarts more when these belong to their ingroup than when these belong to an outgroup (Van Damme & Hoorens, 2016b). One might therefore expect that observers respond more aggressively toward ingroup braggarts than toward outgroup braggarts. It is unclear to which extent our participants viewed the braggart as an ingroup member. We presented the braggart as a psychology student, just like the participants, however, so it is likely that they did. In that case, less aggression might occur toward braggarts who clearly are outgroup members. Our future research will clarify these issues.
Conclusion
While people brag to reap benefits, their claims may insult others if they explicitly compare the self to others or brag about their competence. Observers may then take the bragging as communicating a negative view of others and them, which not only inspires dislike but even aggression. Braggarts who believe that people are unfriendly to them because they cannot stand the braggarts’ great qualities sometimes do good to realize that something else may be at play. Others may very well accept that braggarts are lucky with themselves. They simply cannot stand that braggarts demean them.
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
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research was supported by the Special Research Fund of KU Leuven–Leuven University under grant OT/12/38, awarded to the last author. The Special Research Fund had no involvement in the design of the experiments; the collection, analysis, or interpretation of the data; the writing of the article; or the decision to submit it.
Supplemental Material
The online supplemental material is available with the manuscript on the PSPB website.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
