Gossiping About Deviance: Evidence That Deviance Spurs the Gossip That Builds Bonds

Abstract

We propose that the gossip that is triggered when people witness behaviors that deviate from social norms builds social bonds. To test this possibility, we showed dyads of unacquainted students a short video of everyday campus life that either did or did not include an incident of negative or positive deviance (dropping or cleaning up litter). Study 1 showed that participants in the deviance conditions reported having a greater understanding of campus social norms than those in the control condition; they also expressed a greater desire to gossip about the video. Study 2 found that when given the opportunity, participants did gossip about the deviance, and this gossip was associated with increased norm clarification and (indirectly) social cohesion. These findings suggest that gossip may be a mechanism through which deviance can have positive downstream social consequences.

Keywords

gossip deviance social norms norm clarification social bonds open data open materials

Gossip—broadly defined as communication about the behavior of others (e.g., Peters & Kashima, 2015; E. R. Smith, 2014)—is what people generally do when they are together.¹ It may also be a reason that people come together in the first place. For instance, there is evidence that people who see behaviors that deviate positively or negatively from social norms (i.e., admirable or disgusting behaviors) are highly motivated to discuss these behaviors with others (Feinberg, Willer, Stellar, & Keltner, 2012; Heath, Bell, & Sternberg, 2001; Peters, Kashima, & Clark, 2009). One person’s deviance, then, seems to be the catalyst for other people’s social interactions. To the extent this holds true, gossip may be a mechanism through which deviance has greater social implications than is typically recognized. While the existing literature has thoroughly explored the implications of deviance for the person performing the deviant behavior (e.g., Kam & Bond, 2009), it has given less consideration to the downstream social implications of the deviant act itself. We explore these social implications by examining participants’ desire to—and actual sharing of—gossip about an instance of positive or negative deviance that was witnessed in the laboratory.

We expected that gossiping about deviance would confer benefits on gossipers by giving them a clearer understanding of the prevailing social norms and an increased sense of cohesion. Following this idea, theorists have on occasion suggested that the consequences of deviance may not be limited to derogation of a person performing a negative deviant act (Jetten & Hornsey, 2014) or celebration of a person performing a positive deviant act (although here, derogation is also possible; Heckert & Heckert, 2015). Indeed, Durkheim (1895/1964) argued that people who display deviant behavior serve important social functions by drawing observers’ attention to social norms. He suggested that this should, in turn, increase observers’ sense of unity and shared perspective. In the gossip literature, too, a number of authors have suggested that gossip has the capacity to clarify social norms and increase cohesion (Baumeister, Zhang, & Vohs, 2004; Ben-Ze’ev, 1994; Foster, 2004; Peters & Kashima, 2007; Rosnow, 2001; Wert & Salovey, 2004). However, there has been no consideration that gossip about deviance may be particularly important in this regard, either theoretically or empirically.

In sum, then, following Durkheim’s (1895/1964) classic work, we examined support for three assertions about deviance and gossip. First, people who observe another person acting in a way that deviates positively or negatively from a social norm will have a greater desire to gossip about this act than about an act that is not deviant. Second, to the extent that they actually engage in this gossip, observers should develop a clearer understanding of the relevant social norm. And third, this clarity should provide the basis for cohesion in terms of gossipers’ social bonds and sense of shared perspective. We put these expectations to their first test with two studies. In Study 1, we exposed participants to deviance in the lab and then measured their desire to gossip as well as their perceptions of norm clarification, social bonding, and shared reality. In Study 2, we allowed participants to actually exchange gossip before measuring these same social consequences.

Study 1

Method

Participants

Participants were 114 unacquainted university students who took part in the study in exchange for course credit. Participants were an average of 20.92 years old (SD = 5.60). Most were female (n = 85) and Australian (n = 82). We aimed to exceed a sample of 50 dyads, as above this point, multilevel models successfully converge (Maas & Hox, 2005). We fell just short of 20 dyads in each condition because of a high number of no shows that coincided with the end of the university’s research participation period. This study was approved by the authors’ institutional human research ethics board on March 14, 2016 (No. 2014000387).

Procedure

Participants were recruited to take part in an experiment that purported to examine the way in which people communicate after exposure to different kinds of media. Respondent dyads were seated side by side in front of a computer screen and asked to refrain from talking to one another. They were told that they would watch a short video together and that after this they might also be asked to spend 5 min discussing it. Each dyad was randomly shown one of three 2-min videos, depending on condition: negative deviance (n = 20), positive deviance (n = 18), or control (n = 19). The videos were shot from a single perspective and captured students going about their daily lives in a recognizable and well-frequented campus courtyard.

The key difference between the videos involved a female confederate who walked from the right foreground toward a set of rubbish bins located 150 m away in the left background (screen shots are provided in Fig. 1). In the negative-deviance video, the confederate casually dropped an empty drink can partway through her journey. In the control video, the confederate dropped nothing but simply walked past the same drink can. In the positive-deviance video, the confederate stopped to pick up the drink can, which was already lying on the ground, and deposited it in the rubbish bins. In the two deviance conditions, the deviance behavior occurred approximately 35 s into the video.

Fig. 1.

Screenshots from the videos shown in the (a) negative-deviance, (b) control, and (c) positive-deviance conditions of Studies 1 and 2. These frames show the key moments at which the deviant behavior did or did not occur; circles highlight the empty drink can and the confederate.

After watching the video, dyads were told that they would not be required to discuss it. They were then asked to answer a series of questions measuring perceptions of the video² and of their partner on 7-point Likert scales (all of the following items were rated from 1, strongly disagree, to 7, strongly agree). The desire to discuss the deviance was measured with three items (α = .86): “I had a strong desire to share my feelings and opinions about the video that I watched,” “I would have liked my partner to share her / his feelings and opinions about the video that we watched,” and “I would have liked to spend time with my partner discussing our respective feelings and opinions on the topics we saw in the video.” Norm clarification was measured with six items (α = .84): “This video informed me about the ways in which people generally behave,” “From this video, I have a clearer sense of the ways in which people should behave,” “As a result of this video, I have a better sense of the appropriateness of certain behaviors,” “This video gave me a clearer idea of what it means to be a student [at this university],” “As a result of this video, I have learned about how I should behave,” and “This video motivated me to change the way that I behave.”

We also assessed the social cohesion of participants who had by this point spent about 10 min in close proximity. Social bonding was measured with three items (Peters & Kashima, 2007; α = .81): “I have a social bond with my partner,” “I connect with my partner,” and “I trust my partner.” Shared reality was measured with five items (Stukas, Bratanova, Peters, Kashima, & Beatson, 2010; excluding the two reverse-scored items resulted in a reliable scale: α = .72): “I would not rely on my partner’s judgments of other people” (reverse scored), “My partner is correct in the way in which he / she looks at the world,” “My partner and I have a similar impression of things,” “My partner and I are on the same wavelength,” and “My partner and I have different perspectives on the world” (reverse scored).

After this, participants were told that the litter-related behaviors were staged. As a manipulation and confound check, participants were asked to rate the salience and normativity of several behaviors captured by each video. These included the litter-related behaviors as well as two other behaviors that appeared in all three videos: (a) a group sitting and conversing on the lawn and (b) people taking a shortcut across the lawn by hopping over a chain fence (see the Supplemental Material available online for videos of each condition). Participants rated the salience of the three behaviors captured by their version of the video (i.e., one litter-related and two other) and the normativity of all five behaviors (i.e., three litter-related and two other). To elicit the ratings and minimize the effects of prior exposure, we presented participants with a screenshot for each behavior and asked them to rate its salience or normativity. Behavior salience was measured with two items (rs = .28–.79, all ps < .003): “I clearly remember seeing [behavior] when I watched the video” and “I spent some time thinking about [behavior].” Perceptions of the descriptive and injunctive normativity of the behaviors were each measured with three items (J. R. Smith et al., 2012; αs = .64–.89): “[Behavior] is typical of this university’s students,” “The majority of this university’s students [behavior] on a regular basis,” “[Behavior] regularly is important to the average student,” “Typical students of this university approve of those who [behavior] on a regular basis,” “The majority of students at this university approve of [behavior] on a regular basis,” and “The average student at this university supports [behavior] on a regular basis.”

Results

Deviance-manipulation check

A repeated measures analysis of variance (ANOVA) on the ratings of the descriptive normativity of the behaviors showed that participants perceived the negative (M = 2.27, SD = 0.92) and positive (M = 3.68, SD = 1.23) deviant behaviors as less typical of university students than the control behavior (M = 4.26, SD = 1.01) or either of the two other behaviors (sitting: M = 5.47, SD = 0.94; hopping the chain: M = 4.39, SD = 1.26), F(3.15, 355.89) = 142.19, p < .001, ε² = .56. Repeating this analysis for ratings of injunctive normativity showed that the negative deviance behavior was seen as attracting less approval (M = 1.89, SD = 0.99) and the positive deviance behavior as attracting more approval (M = 5.66, SD = 1.02) than the control behavior (M = 3.40, SD = 1.26) or chain hopping (M = 4.58, SD = 1.12), F(3.07, 346.50) = 316.10, p < .001, ε² = .74. Sitting on the lawn received the highest approval (M = 6.13, SD = 0.81). On average, therefore, the deviant litter-related behaviors were indeed perceived to deviate from social norms more than the control or other behaviors.

Salience-confound check

To check whether participants in the two deviance conditions were the only ones exposed to behaviors that were sufficiently attention grabbing to allow for later discussion, we compared the ratings of behavior salience in all three conditions using one-way ANOVAs. Means and confidence intervals are provided in Table 1. These analyses revealed that the litter-related and chain-hopping behaviors were more salient in the deviance conditions than in the control condition, F(2, 108) = 68.25, p < .001, ε² = .56, and F(2, 111) = 15.76, p < .001, ε² = .22, respectively. Notably, though, the seated group attracted equally high salience ratings in the three conditions, F(2, 111) = 0.29, p > .250, ε² = .01. Our finding that participants in the control condition were exposed to at least one highly salient behavior means that all participants, regardless of condition, had some basis for later discussion. This provides reassurance that the deviance manipulation was not confounded with the salience of potential discussion topics.

Table 1.

Study 1: Mean Ratings of Behavior Salience and Social Cohesion

Rating type and variable	Negative-deviance condition	Control condition	Positive-deviance condition
Behavior salience
Littering	6.41_a [6.08, 6.75]	3.22_b [2.56, 3.89]	6.51_a [6.22, 6.81]
Chain hopping	5.28_a [4.71, 5.80]	4.18_b [3.59, 4.89]	6.25_c [5.94, 6.56]
Sitting	6.28 [6.05, 6.52]	6.17 [5.82, 6.43]	6.31 [6.03, 6.58]
Social cohesion
Discussion desire	4.44_a [4.02, 4.86]	3.72_b [3.29, 4.15]	4.27_ab [3.83, 4.71]
Norm clarification	3.73_a [3.40, 4.06]	2.69_b [2.35, 3.03]	3.82_a [3.47, 4.17]
Social bonding	2.75 [2.38, 3.12]	2.70 [2.32, 3.08]	2.86 [2.47, 3.25]
Shared reality	3.78 [3.59, 3.97]	3.94 [3.74, 4.13]	3.97 [3.77, 4.17]

Note: Dyad N = 57, participant N = 114. Values in brackets are 95% confidence intervals. Within a row, means with different subscripts are significantly different (p < .05); social-cohesion standard errors were clustered within dyads.

The social consequences of mere exposure to deviance

Intraclass correlation coefficients (ICCs) suggested that while the dyad level did not account for variance in ratings of discussion desire (ICC = −.09), it did account for variance in ratings of shared reality (ICC = .17), social bonding (ICC = .20), and norm clarification (ICC = .21). We therefore accounted for the multilevel structure of our data for the latter three variables. Raw means and confidence intervals are provided in Table 1.

To assess the consequences of exposure to deviance on discussion desire, we used the ordinary-least-squares (OLS) method to regress participants’ ratings onto two condition dummy variables (one representing the negative-deviance condition with a value of 1, otherwise 0; another representing the positive-deviance condition with a value of 1, otherwise 0). The condition dummy variables accounted for 5% of the variance in discussion desire, F(2, 111) = 3.05, p = .051. As expected, participants in the negative-deviance condition expressed a stronger desire to discuss the video with their partner than did control participants—b = 0.72, t(111) = 2.38, p = .019; this difference was marginal in the positive-deviance condition—b = 0.55, t(111) = 1.76, p = .081. Participants’ ratings in the negative- and positive-deviance conditions did not differ from one another, F(1, 111) = 0.32, p = .575. This supports our claim that exposure to deviance may mobilize subsequent interactions among observers.

To assess the consequences of exposure to deviance on the remaining variables, we fitted two-level random-effects maximum-likelihood regression models to participants’ ratings (Rabe-Hesketh & Skrondal, 2012). This approach allowed us to contend with the potential loss of independence that was associated with the nesting of participants (Level 1) within dyads (Level 2). We allowed intercepts to vary in order to model the variance in ratings that could be attributed to differences among dyads. To our surprise, there was evidence that simply exposing participants to deviance affected their ratings of norm clarification, likelihood-ratio (LR) χ²(2, N = 114) = 22.08, p < .001, with participants in the deviance conditions reporting significantly higher norm clarification than participants in the control condition—negative-deviance condition: b = 1.04, z = 4.36, p < .001; positive-deviance condition: b = 1.13, z = 4.62, p < .001. Participants’ ratings in the negative- and positive-deviance conditions did not differ from one another, χ²(1, N = 114) = 0.14, p = .708. Thus, it appears that observers do not need to gossip about a deviant act to gain a clearer understanding of local norms.

We were able to get some understanding of what aspect of the social norms may have been clarified by examining how normativity ratings for the three litter-related behaviors varied as a function of experimental condition. A 3 (behavior; within participants) × 3 (condition; between participants) mixed-design ANOVA revealed that prior exposure conditioned the extent to which the behaviors were perceived to differ in terms of descriptive normativity, F(3.22, 178.69) = 2.98, p = .030, ε² = .03, but not injunctive normativity, F(3.55, 197.21) = 0.47, p > .250, ε² = .00. We used one-way ANOVAs to compare descriptive-normativity ratings across conditions for each behavior in turn. This revealed that participants in the positive-deviance condition rated picking up litter as marginally more typical (M = 4.01, SD = 1.08) than did participants in the negative-deviance condition (M = 3.38, SD = 1.18), F(2, 111) = 2.59, p = .079, ε² = .05. Participants in the negative-deviance condition rated walking past litter as more typical (M = 4.50, SD = 0.98) than did participants in the positive-deviance condition (M = 3.94, SD = 1.05), F(2, 111) = 3.14, p = .047, ε² = .05. Perceptions of the typicality of dropping litter did not vary, F(2, 111) = 0.98, p > .250, ε² = .02. Thus, it seems that exposure to positive deviance was associated with increased expectations that students would not ignore (and may pick up) litter, relative to exposure to negative deviance.

Unlike the findings for norm clarification, there was no evidence that exposure to deviance affected participants’ social bonding, LR χ²(2, N = 114) = 0.30, p > .250, or sense of shared reality, LR χ²(2, N = 114) = 1.89, p > .250. There was also no evidence that deviance affected social bonding and shared reality indirectly through norm clarification. Specifically, creating generalized multilevel structural equation models (Rabe-Hesketh, Skrondal, & Pickles, 2004) of the impact of the deviance dummies on cohesion through norm clarification revealed that all indirect effects were nonsignificant (parameter standard errors were computed with the delta method; Oehlert, 1992): negative deviance on social bonding, indirect effect = 0.18, z = 0.92, p > .250; positive deviance on social bonding, indirect effect = 0.20, z = 0.93, p > .250; negative deviance on shared reality, indirect effect = 0.07, z = 1.13, p > .250; positive deviance on shared reality, indirect effect = 0.07, z = 1.13, p > .250. In Study 2, we examined whether exposure to deviance has consequences for social cohesion when (and to the extent that) participants actually gossip about the deviant act.

Study 2

Method

Participants

Participants were 130 unacquainted university students who took part in the study in exchange for course credit. Participants were an average of 20.39 years old (SD = 4.94). Most were female (n = 103) and Australian (n = 86). Data collection continued until there was a minimum of 20 dyads in each condition. We slightly exceeded these numbers because more students than expected showed up. We additionally excluded one negative-deviance dyad as they had an undeclared preexisting relationship. Therefore, for analytic purposes, 128 students’ data were included. This study was approved by the authors’ institutional human research ethics board on April 16, 2014 (No. 2014000387).

Procedure

As in Study 1, participant dyads were recruited for an experiment on media and communication, and each was randomly assigned to be shown one of the three 2-min videos of campus life, depending on condition: negative deviance (n = 19), positive deviance (n = 22), or control (n = 23). Respondents were led to expect that they would be asked to discuss the video, and after watching the video, each dyad was left alone in the room for 5 min. They were told that they were free to talk about any aspect of the video they wished to and that their conversation would be recorded for later analysis. At the end of the 5 min, the experimenter returned and asked respondents to complete a questionnaire about their perceptions of their conversation and their partner. This included the Study 1 scales of social bonding (α = .76) and shared reality (we again excluded the two reverse-scored items to form a reliable scale: α = .79) and an amended version of the norm-clarification scale (this time, participants rated how their conversation had clarified their understanding of norms; α = .90).³

Results

Conversation coding

Two independent coders (the first author and a research assistant) rated each dyad’s verbal expressions of approval and disapproval of the deviant behavior on 5-point scales (0 = none, 4 = strong shared expressions). The ratings were reliable (approval: r = .80, p < .001; disapproval: r = .81, p < .001) and were averaged for each dyad. The coders also calculated the total length of time that participants spent discussing the following topics: litter and litter-related behavior, r = .94, p < .001; the seated group, r = .97, p < .001; people hopping over the chain, r = .99, p < .001; nonsocial topics, including the weather, buildings, trees, and wildlife, r = .88, p < .001; and personal topics, including interests, background, and plans, r = .91, p < .001. These times were also averaged.

Conversation content following exposure to deviance

The means and confidence intervals of the conversation codes are provided in Table 2. To examine whether participants’ Study 1 discussion desire translated into actual gossip about litter-related behavior, we ran OLS regression with dummy variables representing each of the deviance conditions (the control condition was the reference; see the Study 1 Method for coding). The condition dummies predicted the length of time that dyads spent gossiping about deviance, F(2, 61) = 12.19, p < .001, and as expected, dyads in the two deviance conditions spent significantly longer talking about litter-related behavior than did dyads in the control condition—negative deviance vs. control: t(61) = 4.93, p < .001, ε² = .28; positive deviance vs. control: t(61) = 2.03, p = .046, ε² = .05. Dyads in the negative-deviance condition spent significantly longer talking about littering than did dyads in the positive-deviance condition, F(1, 61) = 8.65, p < .005.

Table 2.

Study 2: Mean Conversation Codes and Social-Cohesion Ratings

Rating type and variable	Negative-deviance condition	Control condition	Positive-deviance condition
Conversation code
Litter-related discussion	68.79_a [49.65, 87.93]	5.04_c [−12.35, 22.44]	30.36_b [12.57, 48.15]
Chain-hopping discussion	15.21 [8.05, 22.38]	6.65 [0.14, 13.17]	16.07 [9.41, 22.73]
Sitting-group discussion	35.05_ab [19.53, 50.57]	18.50_b [4.39, 32.61]	45.18_a [30.76, 59.61]
Nonsocial discussion	9.66_a [0.59, 18.73]	32.35_b [24.11, 40.59]	25.05_b [16.61, 33.47]
Personal discussion	57.11 [25.23, 88.98]	65.20 [36.22, 81.31]	51.68 [22.06, 81.31]
Approval of deviant behavior	0.05_a [−0.36, 0.47]	0.20_a [−0.18, 0.57]	1.30_b [0.91, 1.68]
Disapproval of deviant behavior	2.13_a [1.71, 2.55]	0.17_b [−0.21, 0.56]	0.25_b [−0.14, 0.64]
Social cohesion
Norm clarification	3.95_a [3.54, 4.36]	2.93_b [2.56, 3.30]	3.54_a [3.16, 3.92]
Social bonding	4.55 [4.22, 4.88]	4.55 [4.25, 4.85]	4.44 [4.13, 4.75]
Shared reality	4.99 [4.70, 5.28]	5.17 [4.90, 5.44]	4.84 [4.57, 5.12]

Note: Dyad N = 64, participant N = 128. Values in brackets are 95% confidence intervals. Approval and disapproval of deviant behavior were coded using 5-point response scales (0 = none, 4 = strong shared attitude); all other conversation codes were recorded in seconds. Within a row, means with different subscripts are significantly different (p < .05).

We repeated this analysis for each dyad’s expressed disapproval and approval of the deviant behavior (see Table 2). The condition dummies significantly predicted expressions of disapproval, F(2, 61) = 29.45, p < .001, with dyads in the negative-deviance condition expressing significantly more disapproval than dyads in the control condition, t(61) = 6.90, p < .001, ε² = .40, or positive-deviance condition, F(1, 61) = 43.09, p < .001, ε² = .40. The condition dummies also significantly predicted expressions of approval, F(2, 61) = 12.04, p < .001, ε² = .28, and dyads in the positive-deviance condition expressed significantly more approval than dyads in the control condition, t(61) = 4.07, p < .001, ε² = .19, or negative-deviance condition, F(1, 61) = 19.65, p < .001, ε² = .41. Therefore, when given the opportunity, participants do indeed gossip about deviant behaviors; they also take the opportunity to derogate negative, and celebrate positive, deviant behaviors.

As is apparent from the results shown in Table 2, these were not the only ways in which conversations differed across condition. Repeating the above analysis revealed differences in the time dyads spent discussing the seated group, R² = .11, F(2, 61) = 3.57, p = .034, people hopping over the chain, R² = .08, F(2, 61) = 2.47, p = .093, and nonsocial topics, R² = .19, F(2, 61) = 7.02, p = .002. There were no between-condition differences in personal disclosures, R² = .01, F(2, 61) = 0.22, p > .250. In the analysis that followed, it was therefore important to ascertain that any social consequences of exposure to deviance can be attributed to deviance gossip specifically.

The social consequences of exposure to deviance

ICCs point to the importance of accounting for dyad-level variation in ratings of shared reality (ICC = .20), social bonding (ICC = .35), and norm clarification (ICC = .28). Therefore, to assess whether exposure to deviance has social consequences when people gossip about it, we again fitted two-level random-effects maximum-likelihood regression models to participants’ ratings, using two dummy variables to represent the deviance conditions as before. Means and confidence intervals are provided in Table 2.

This analysis revealed that the model that included the condition dummy variables provided a significantly better fit of participants’ ratings that their conversation had clarified their understanding of the prevailing norms, compared with the model including random effects only, LR χ²(2, N = 128) = 11.04, p = .004. As expected, participants in the deviance conditions reported that their conversations led to a significantly greater improvement in their understanding of the prevailing norms than did participants in the control condition—negative-deviance condition: b = 1.02, z = 3.42, p = .001; positive-deviance condition: b = 0.61, z = 2.13, p = .033. Participants’ ratings in the deviance conditions did not differ, χ²(1, N = 128) = 1.84, p = .175.

As in Study 1, repeating this analysis did not provide any evidence that exposure to deviance directly affected participants’ social bonding, LR χ²(2, N = 128) = 0.26, p > .250, or sense of shared reality, LR χ²(2, N = 128) = 2.45, p > .250. It is nonetheless possible that exposure to deviance affected social bonding and shared reality indirectly through norm clarification. We tested this possibility in a mediational analysis.

Deviance gossip and norm clarification mediate the impact of deviance exposure

To see whether the increased tendency to gossip about deviance (rather than some other topic) after exposure to deviance may be key to the effects observed previously, we used two-level random-effects maximum-likelihood regression models to regress participants’ social ratings onto the conversation topic times in turn. The unstandardized regression coefficients for these models are provided in Table 3.

Table 3.

Study 2: Predicting Social-Cohesion Ratings From Time Spent Discussing Topics

Predictor	Norm-clarification rating	Social-bonding rating	Shared-reality rating
Litter-related discussion	0.012** [0.007, 0.017]	0.003 [−0.001, 0.008]	0.002 [−0.002, 0.006]
Sitting-group discussion	−0.001 [−0.007, 0.006]	−0.003 [−0.009, 0.003]	0.000 [−0.006, 0.006]
Chain-hopping discussion	−0.005 [−0.017, 0.009]	−0.001 [−0.013, 0.011]	0.000 [−0.010, 0.011]
Nonsocial discussion	−0.004 [−0.016, 0.007]	0.004 [−0.008, 0.013]	0.008 [−0.001, 0.018]
Personal discussion	0.002 [−0.002, 0.005]	0.004* [0.001, 0.007]	0.001 [−0.002, 0.004]
Constant	3.128 [2.455, 3.801]	4.233 [3.609, 4.857]	4.688 [4.124, 5.252]

Note: The table shows unstandardized regression coefficients, with 95% confidence intervals in brackets.

p < .05. **p < .01.

This analysis showed that the model that included the conversation topic times provided a significantly better fit of ratings of norm clarification than the model that constrained the topic parameters to zero, LR χ²(5, N = 128) = 28.25, p < .001. Importantly, the only significant predictor of norm-clarification ratings was the amount of time participants spent gossiping about deviance, z = 4.86, p < .001. Although the different scale metrics meant that the unstandardized coefficients were small, they pointed to sizeable effects: An additional 83 s of deviance gossip translated into a 1-scale-point increase in norm clarification. Repeating this analysis for social bonding, we found a marginal improvement in model fit, LR χ²(5, N = 128) = 10.38, p = .065, and dyads that spent more time disclosing personal information felt significantly more bonded than dyads that spent less time, z = 2.29, p < .022. Repeating this analysis for shared reality did not improve model fit, LR χ²(5, N = 128) = 3.49, p = .625.

In our final analysis, we tested two mediational expectations. The first was that gossiping about deviance would mediate the impact of exposure to deviance on norm clarification. The second was that gossiping about deviance and norm clarification would serially mediate the impact of exposure to deviance on cohesion. To run these tests analyses, we used generalized multilevel structural equation models to map the direct and indirect effects among the deviance-condition dummies, litter gossip, norm clarification, and the two social-cohesion measures (intercepts at the random dyad level were included for the norm-clarification and cohesion measures; the delta method was used to calculate standard errors for nonlinear transformed parameters; Oehlert, 1992).

The results of this analysis are depicted in Figure 2. As can be seen, after accounting for time spent gossiping about litter, we found that exposure to deviance was no longer significantly associated with differences in perceived norm clarification, relative to the control condition. The associated indirect effects of exposure to deviance on norm clarification through deviance gossip were indeed significant—negative deviance: indirect effect = 0.71, z = 3.67, p < .001; positive deviance: indirect effect = 0.28, z = 2.43, p = .015.

Fig. 2.

Results of the structural equation model used in Study 2 to map the direct and indirect effects among the deviance-condition dummies, time spent gossiping about litter, and ratings of norm clarification, shared reality, and social bonding. Values shown are unstandardized regression coefficients. Solid lines and asterisks indicate significant paths (*p < .05, **p < .01), and dotted lines indicate nonsignificant paths.

Turning to social cohesion, we replicated the earlier findings by showing that exposure to deviance does not boost shared reality or social bonding directly (indeed, positive deviance was a significant negative predictor of shared reality, which points to a possible suppression effect). Notably, though, there was evidence of serial mediation, whereby exposure to negative deviance had significant indirect effects on cohesion through time spent gossiping about deviance and then norm clarification—shared reality: indirect effect = 0.20, z = 2.27, p = .023; social bonding: indirect effect = 0.17, z = 2.11, p = .034. The indirect effects of exposure to positive deviance on cohesion through time spent gossiping about deviance and norm clarification were marginal—shared reality: indirect effect = 0.08, z = 1.86, p = .063; social bonding: indirect effect = 0.07, z = 1.77, p = .077.

General Discussion

We provide evidence that one person’s deviance can shape other people’s social interactions. In particular, participants who were exposed to one of the deviance videos expressed a stronger desire to talk about the video than participants who saw the control video. When given the opportunity, almost all of the participants who saw the deviant act chose to spontaneously gossip about it. We were also able to show that in spurring people to gossip, deviance may have beneficial social consequences. While we found that mere exposure to deviance was sufficient for norm clarification, our findings also suggest that gossiping about deviance may build these perceptions. In particular, the impact of exposure to deviance on a sense that a conversation created a clearer understanding of social norms was fully mediated by the length of time that participants spent gossiping about the deviant act. Further evidence for the importance of gossiping about deviance comes from our finding that exposure to deviance indirectly improved cohesion through norm clarification when (and to the extent that) participants were able to share deviance gossip.

Together, these findings support our claim that gossip may be a mechanism through which deviance can have important downstream social consequences. In particular, while our research replicates the well-established finding that negative deviant behavior is derogated and positive deviant behavior is celebrated (Heckert & Heckert, 2015; Kam & Bond, 2009; Marques, Yzerbyt, & Leyens, 1988), it shows that the consequences of deviance are not limited to the person committing the deviant behavior (Jetten & Hornsey, 2014). In this way, our findings align with Durkheim’s (1895/1964) claim that people who engage in deviant acts make an important contribution to the functioning of societies by drawing people’s attention to, and clarifying their understanding of, the existing social norms. Our findings also align with his suggestion that this greater normative understanding supports better societal unity. Notably, our work builds on these ideas by specifying one mechanism through which deviance may have these effects.

At the same time, our work suggests that gossip may not be necessary for all downstream social consequences. Clearly, just witnessing a deviant act can change a person’s understanding of the behaviors that are typical in a particular social context, which suggests that there is merit in considering the intrapersonal processes that may be sparked by a deviant act and their likely consequences. However, such intraindividual processes have spatial and temporal limits that gossip does not. The desire to gossip about deviance may lead people to indirectly expose others to the deviant event, thus spreading information about the event through a social network. In this way, when gossip about deviance does come into play, it has the potential for widespread social consequences. In future work, it is important to show that gossip about deviance plays a causal role in processes such as these; among other things, such a finding would support claims that the social fitness that accompanies gossip could underpin the evolution of syntactically complex language (e.g., Dunbar, 1996).

A final notable aspect of this study is our focus on positive as well as negative deviance. Although positive deviance is attracting increased research attention, its consequences are poorly understood. In general, our findings support claims that positive and negative deviance may have similar consequences, at least in some domains (Ben-Yehuda, 1990). In particular, it seems that whether a behavior deviates from a social norm positively or negatively, it throws the norm into sharp relief, which confers the attendant social benefits. At the same time, although participants in our studies found the positive and negative deviance equally salient, they spent about twice as long gossiping about the negative deviant act. It is possible, therefore, that the gossip that clarifies social norms typically concerns negative deviance.

In a 1971 interview, musician Frank Zappa said, “I think that progress is not possible without deviation” (Kiers; 40:53). Our results certainly suggest that, without deviance, our conversations would be rather emptier and our social understanding somewhat weaker. They also suggest that to investigate the consequences of deviance for social change, it is important to consider the essential role that our daily gossip may play.

Footnotes

Acknowledgements

We thank Kristen Webb for assistance in collecting Study 1 data and conversation coding.

Action Editor

Jamin Halberstadt served as action editor for this article.

Declaration of Conflicting Interests

The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.

Supplemental Material

Additional supporting information can be found at

Open Practices

All data and materials have been made publicly available via the Open Science Framework and can be accessed at https://osf.io/jqkg7/#. The complete Open Practices Disclosure for this article can be found at https://journals-sagepub-com.web.bisu.edu.cn/doi/suppl/10.1177/0956797617716918. This article has received the badges for Open Data and Open Materials. More information about the Open Practices badges can be found at .

Notes

References

Baumeister

R. F.

Zhang

Vohs

K. D.

(2004). Gossip as cultural learning. Review of General Psychology, 8, 111–121.

Ben-Yehuda

(1990). Positive and negative deviance: More fuel for a controversy. Deviant Behavior, 11, 221–243.

Ben-Ze’ev

(1994). The vindication of gossip. In Goodman

R. F.

Ben-Ze’ev

(Eds.), Good gossip (pp. 11–24). Lawrence: University Press of Kansas.

Dunbar

(1996). Grooming, gossip, and the evolution of language. Cambridge, MA: Harvard University Press.

Durkheim

(1964). The rules of sociological method ( Solovay

S. A.

Mueller

J. H.

, Trans.). New York, NY: Free Press. (Original work published 1895)

Feinberg

Willer

Stellar

Keltner

(2012). The virtues of gossip: Reputational information sharing as prosocial behavior. Journal of Personality and Social Psychology, 102, 1015–1030.

Foster

E. K.

(2004). Research on gossip: Taxonomy, methods, and future directions. Review of General Psychology, 8, 78–99.

Heath

Bell

Sternberg

(2001). Emotional selection in memes: The case of urban legends. Journal of Personality and Social Psychology, 81, 1028–1041.

Heckert

D. M.

Heckert

D. A.

(2015). Positive deviance. In Goode

(Ed.), The handbook of deviance (pp. 80–100). Chichester, England: John Wiley & Sons.

10.

Jetten

Hornsey

M. J.

(2014). Deviance and dissent in groups. Annual Review of Psychology, 65, 461–485.

11.

Kam

C. C. S.

Bond

M. H.

(2009). Emotional reactions of anger and shame to the norm violation characterizing episodes of interpersonal harm. British Journal of Social Psychology, 48, 203–219.

12.

Kiers

. (Director). (1971). A day with Frank Zappa [Motion picture]. The Netherlands: Vrijzinnig Protestantse Radio Omroep.

13.

Maas

C. M.

Hox

J. J.

(2005). Sufficient sample sizes for multilevel modeling. Methodology, 1(3), 86–92.

14.

Marques

J. M.

Yzerbyt

V. Y.

Leyens

J.-P.

(1988). The “black sheep effect”: Extremity of judgments towards ingroup members as a function of group identification. European Journal of Social Psychology, 18, 1–16.

15.

Oehlert

G. W.

(1992). A note on the delta method. American Statistician, 46, 27–29.

16.

Peters

Kashima

(2007). From social talk to social action: Shaping the social triad with emotion sharing. Journal of Personality and Social Psychology, 93, 780–797.

17.

Peters

Kashima

(2015). Bad habit or social good? How perceptions of gossiper morality are related to gossip content. European Journal of Social Psychology, 45, 784–798. doi:10.1002/ejsp.2123

18.

Peters

Kashima

Clark

(2009). Talking about others: Emotionality and the dissemination of social information. European Journal of Social Psychology, 39, 207–222.

19.

Rabe-Hesketh

Skrondal

(2012). Multilevel and longitudinal modeling using Stata (3rd ed.). College Station, TX: Stata Press.

20.

Rabe-Hesketh

Skrondal

Pickles

(2004). Generalized multilevel structural equation modeling. Psychometrika, 69, 167–190.

21.

Rosnow

R. L.

(2001). Rumor and gossip in interpersonal interaction and beyond: A social exchange perspective. In Kowalski

R. M.

(Ed.), Behaving badly: Aversive behaviors in interpersonal relationships (pp. 203–232). Washington, DC: American Psychological Association.

22.

Smith

E. R.

(2014). Evil acts and malicious gossip: A multiagent model of the effects of gossip in socially distributed person perception. Personality and Social Psychology Review, 18, 311–325.

23.

Smith

J. R.

Louis

W. R.

Terry

D. J.

Greenaway

K. H.

Clarke

M. R.

Cheng

(2012). Congruent or conflicted? The impact of injunctive and descriptive norms on environmental intentions. Journal of Environmental Psychology, 32, 353–361.

24.

Stukas

A. A.

Bratanova

Peters

Kashima

Beatson

R. M.

(2010). Confirmatory processes in attitude transmission: The role of shared reality. Social Influence, 5, 101–117.

25.

Wert

S. R.

Salovey

(2004). A social comparison account of gossip. Review of General Psychology, 8, 122–137.

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.

334.54 MB

0.00 MB