The social side of shame: approach versus withdrawal

ABSTRACT

At present, the consequences and functions of experiences of shame are not yet well understood. Whereas psychology literature typically portrays shame as being bad for social relations, motivating social avoidance and withdrawal, there are recent indications that shame can be reinterpreted as having clear social tendencies in the form of motivating approach and social affiliation. Yet, until now, no research has ever put these alternative interpretations of shame-motivated behaviours directly to the test. The present paper presents such a test by studying the extent to which shame motivates a preference for social withdrawal versus a preference for social approach. Two studies (N = 148 and N = 133) using different shame inductions both showed people experiencing shame to prefer to be together with others (social approach) over being alone (social withdrawal). In addition, the preference for a social situation was found to be unique for shame; it was not found for the closely related emotion of guilt. Taken together, these findings provide direct empirical support for the idea that shame can have positive interpersonal consequences.

KEYWORDS:

• Shame
• interpersonal behaviour
• approach behaviour
• social withdrawal
• guilt

Shame is a social emotion, typically felt after failures, inadequacies, and moral or social transgressions (Izard, 1977); when people fail to live up to social or moral standards and when others are aware – or could be aware – of this failure. When people experience shame, they think of others who disapprove of them or who will evaluate them negatively (Bain, 1875). Shame is also a self-conscious emotion: It is associated with a heightened sense of self-awareness (Izard, 1977; Welten, Zeelenberg, & Breugelmans, 2012) and with feeling small, worthless, incompetent, unworthy, and disgusted with oneself (Lewis, 1971, 1992). The self-conscious nature of shame is closely linked to its social nature. Shame relates to social standing (Gilbert, 1997), appearance in the eyes of others (Lewis, 1971), and feeling inferior to others (Keltner & Harker, 1998).

The interpersonal consequences of experienced shame are considered to be predominantly negative. Shame has been argued to disrupt ongoing activities, to motivate an inability to talk, act, or think clearly (Gilbert, 1997; Lewis, 1992), and to promote a passive self (Barrett, 1995). At an interpersonal level, shame has been associated with the tendency to socially withdraw, hide, disappear, inhibit social interactions, and to isolate oneself from others (e.g. Haidt, 2003; Keltner & Harker, 1998; Lewis, 1971, 1992; Tangney, 1991).

Interestingly, a recent line of research has focussed on the potential positive consequences of shame, suggesting that it would motivate positive interpersonal behaviours to restore one's threatened self (De Hooge, Zeelenberg, & Breugelmans, 2010, 2011). Indeed, empirical research has repeatedly demonstrated shame to be able to induce prosocial behaviours, such as donating, cooperating, or gift giving, notably in situations where people are together with others towards whom they feel ashamed (De Hooge, 2014; De Hooge, Breugelmans, & Zeelenberg, 2008). Similar findings have been reported for experiences of group-based shame (Gausel & Leach, 2011; Gausel, Leach, Vignoles, & Brown, 2012; Welten et al., 2012). Taken together, current research on the behavioural effects of shame suggests that shame leads to both social withdrawal and social approach. The question remains, however, whether shame motivates one over the other.

Surprising as it may sound, to our best knowledge, no research has directly addressed this question. Empirical research supporting a withdrawal perspective has used self-report rating scales, measuring only one behaviour at the time (e.g. “moving away from people or things”, Scherer & Wallbott, 1994), or it has focussed on shame-proneness instead of experienced shame (e.g. Tangney, 1991; see also De Hooge et al., 2008). Empirical research supporting an approach perspective has mostly put shame-experiencing people in social situations, and subsequently studied how social they would behave (e.g. De Hooge, 2014), or has used self-report rating scales that measure tendencies to engage in a single behaviour (e.g. Gausel et al., 2012).

This previous research missed out on a crucial element of social situations that is not only important for understanding shame in everyday life, but also for comparing social withdrawal with social approach tendencies: The element of situation choice. In everyday life, people can often choose whether they want to enter a social situation or prefer being alone. In other words, before deciding how to behave towards others (e.g. selfish or prosocial), people may choose to avoid social situations in the first place. None of the previously mentioned studies has offered respondents such a choice. We therefore argue that previous research has not been able to directly test whether shame motivates social withdrawal or social approach.

In the current research, we induced shame and subsequently gave people the choice between being alone or being with others, revealing a preference for either social withdrawal or social approach. This design allows us to directly compare both perspectives on the social side of shame. Experiment 1 compared the effect of shame to that of guilt (and a control condition), to examine the extent to which effects are shame-specific. Experiment 2 focussed on the effects of shame in relation to the people one could be together with. We report all data exclusions (if any), all manipulations, and all measures in the two experiments.

Experiment 1

Method

Tilburg University students (111 females and 37 males, M_age = 19.91, SD = 3.20) participated in a series of unrelated studies in partial fulfilment of a course requirement. They were randomly assigned to the Shame, Guilt, or Control condition, with withdrawal/approach choice as the dependent variable. Respondents entered the laboratory in groups of eight to twelve, and were seated in separate cubicles. Emotions were induced with the well-validated autobiographical recall procedure followed by a self-report manipulation check (e.g. De Hooge, 2014; De Hooge et al., 2008; De Hooge, Zeelenberg, & Breugelmans, 2007; Ketelaar & Au, 2003; Nelissen, Dijker, & De Vries, 2007). Respondents in the Shame/Guilt condition were asked to thoroughly think about, and briefly write down on a note, a personal experience in which they felt very ashamed (“schaamte” in Dutch) / very guilty (“schuld” in Dutch). In the Control condition, respondents were asked to think about and briefly describe their experiences on a regular weekday. As an emotion manipulation check, respondents subsequently indicated how much shame, guilt, regret, disappointment, fear, anger, and dissatisfaction they felt in the described situation (0 = “not at all”, 10 = “very strongly”).

After this induction, respondents chose their next task: One that they would do alone or one that they would do together with another respondent. The first task was described as

“This task you will do alone. You will see pictures consisting of 180 white and black squares. The number of white and black squares varies across pictures; your task is to every time indicate the number of black squares as soon as possible”.

The second task was described as

“This task you will do together with another person. Both persons receive 10 coins that they can keep or can give to the other person. Your task is to indicate how many coins you will keep and how many coins you will give to the other person”.

Respondents indicated which task they wanted to do (our DV: Withdrawal/Approach choice) and continued with their preferred task. After completion of all tasks, respondents threw away their notes, were thanked, and debriefed.

Results

The emotion manipulation was successful. One-way ANOVAs revealed that both self-reported shame and guilt differed significantly based on condition, F(2, 144) = 191.22, p < .001 and F(2, 144) = 70.58, p < .001, respectively. Follow-up contrasts with a Bonferroni corrected alpha of .025 revealed that respondents in the Shame condition reported more shame (M = 8.88, SD = 1.17) than respondents in the Guilt condition (M = 7.31, SD = 2.38), t(144) = 4.19, p < .001, d = 0.84, and the Control condition (M = 1.60, SD = 1.91), t(144) = 18.70, p < .001, d = 4.60 (see Table 1). Moreover, respondents in the Guilt condition reported more guilt (M = 8.44, SD = 1.87) than respondents in the Shame condition (M = 5.50, SD = 3.56), t(144) = 5.46, p < .001, d = 1.03, and the Control condition (M = 1.87, SD = 2.46), t(144) = 11.88, p < .001, d = 3.01. Respondents in the Shame condition reported more shame than any other emotion, ts(49) > 5.53, ps < .001, and respondents in the Guilt condition reported more guilt than any other emotion, ts(51) > 3.44, ps ≤ .001, with a Bonferroni corrected alpha of .008.

Table 1. Emotion manipulation checks as a function of emotion condition in Experiment 1.

	Emotion condition
Dependent variable	Shame M (SD)		Guilt M (SD)		Control M (SD)
Shame feelings	8.88 (1.17)	>	7.31 (2.38)		1.60 (1.91)
	∨		∧
Guilt feelings	5.50 (3.56)	<	8.44 (1.87)	>	1.87 (2.46)
			∨
Regret feelings	5.40 (3.55)		6.79 (2.70)		2.36 (2.85)
Disappointment feelings	5.50 (3.33)		5.96 (3.01)		3.27 (3.03)
Fear feelings	2.98 (3.11)		5.19 (3.21)		2.00 (2.51)
Anger feelings	3.40 (3.51)		3.65 (3.13)		2.51 (2.70)
Dissatisfaction feelings	6.34 (3.20)		6.17 (3.15)		3.93 (2.98)

Note: The emotion manipulation check could range from “not at all” (0) to “very strongly” (10). Arrows indicate significant differences between the mean of that column and all means of rows or columns following after the arrow at p ≤ .001.

Emotion conditions differed in observed choices for Withdrawal or Approach, χ² (2, N = 147) = 14.26, p = .001, V = .31 (see Table 2). The majority of respondents in the Shame condition selected working together (i.e. Approach: 62%), more than respondents in the Guilt condition (25%), χ² (1, N = 102) = 14.23, p < .001, V = .37, or the Control condition (42.2%), χ² (1, N = 95) = 3.242, p = .054, V = .20. Marginally more respondents in the Guilt condition selected working alone (Withdrawal: 75%) compared to respondents in the Control condition (57.8%), χ² (1, N = 97) = 3.24, p = .07, V = .18.

Table 2. Withdrawal/approach choice as a function of emotion condition in Experiment 1.

	Emotion condition
Withdrawal/Approach choice	Shame % (n)		Guilt % (n)		Control % (n)
Alone task	38% (19)	<	75% (39)	≥	57.8% (26)
Together task	62% (31)	>	25% (13)	≤	42.2% (19)

Note: Withdrawal/approach choice reflects the percentage (and number) of respondents who chose the together task, χ² (2, N = 147) = 14.26, p < .01, V = .31. A “>” mark indicates a significant difference, while a “≥” mark indicates a marginally significant difference (p = .07).

Discussion

Experiment 1 suggests that shame motivates a preference for social approach over social withdrawal: Being together with others instead of being alone. Interestingly, respondents experiencing guilt preferred not to interact with others.

Although these results provide preliminary evidence for the idea that shame leads to social approach, there are also some limitations. First, the usage of autobiographical recall as emotion induction may have primed shame and guilt associations instead of shame and guilt experiences. Thus, instead of emotion effects we may have found effects of emotion knowledge. Second, the situation choice was confounded. The options not only differed in being individual or social, but also in content (counting dots vs. sharing money). Perhaps the shame induction elicited a preference for money instead of a preference for social approach? Finally, the social task was performed with people unaware of the emotion-evoking event. Previous research has shown that the consequences of shame may differ depending upon whether others know about the shameful event or not (De Hooge et al., 2008; see also Zeelenberg & Pieters, 2006). It is thus possible that the findings would be different when the approach option concerns others who know about the shameful event.

Experiment 2 was designed to overcome these issues. In Experiment 2, shame was induced with a task in the lab, and respondents received a choice between an individual and a social task that did not differ in content. Furthermore, we examined whether the preference for withdrawal or approach depended on whether or not others were aware of the shameful experience.

Experiment 2

Method

Erasmus University students (62 females and 71 males, M_age = 19.95, SD = 2.40) participated in a series of unrelated studies in partial fulfilment of a course requirement. They were randomly assigned to one of the conditions of a 2 (Emotion condition: Shame vs. Control) × 2 (Knowledge about Shameful Event: Yes vs. No) between-subjects design with a continuous task preference measure and a withdrawal/approach choice as dependent variables.

We used the shame-induction of De Hooge et al. (2008). Respondents entered the laboratory in groups of eight to twelve, were seated in separate cubicles, and were informed that they would interact via computers in groups of four. They then completed two intelligence tests, which were ostensibly meant to see whether group members were comparable in knowledge and academic abilities. Next, respondents publicly received bogus feedback about their performance. In the Shame condition, the respondent learned that all group members saw on their computer screen that the respondent earned 9 (labelled “insufficient” in terms of academic abilities) out of a maximum of 20 points, while other group members earned 16, 17 (both labelled “sufficient”), and 19 (labelled “highly sufficient”) points. In the Control condition, the respondent earned 16 (labelled “sufficient”) points, while other group members earned 16, 17, and 19 points. Note that in both conditions respondents received the lowest score in the group. The difference resided in the absolute score level, being “insufficient” in the Shame condition and “sufficient” in the Control condition. As such, this design provides a conservative test of the effects of shame.

Respondents chose which task to do next: An “alone task” (explained to be a task that respondents had to do alone. In this task, they would make monetary decisions) or a “together task” (explained to be a task that respondents would do together with another respondent. In this task, both respondents would make monetary decisions). In the No-knowledge conditions, the together task would be done with another respondent from another group who knew nothing about the intelligence scores. In the Knowledge conditions, the together task would be done together with a group member who knew the intelligence score of the respondent. Respondents indicated with a slider bar to what extent they preferred doing one task over the other (the dependent measure Withdrawal/approach preference, ranging from “0” = alone task to “100” = together task). They also chose which task they wanted to do (the dichotomous dependent measure Withdrawal/approach choice).

The emotion manipulation check was identical to Experiment 1. In addition, respondents indicated the extent to which they felt alone, small, the centre of all attention, how much they wanted others not to know about the intelligence score, and how much they wanted to repair the outcomes (all items rated on scales ranging from “0” = not at all to “10” = very strongly; De Hooge et al., 2010). These were averaged to form a shame index (α = .89). After completion of all tasks respondents were thanked and thoroughly debriefed.

Results

The shame manipulation was successful: Respondents in the Shame condition reported more shame (M = 6.46, SD = 2.81) than respondents in the Control condition (M = 3.02, SD = 2.19), t(129) = 7.81, p < .001, d = 1.37. Also for the shame index, respondents in the Shame condition scored higher (M = 6.15, SD = 2.53) than respondents in the Control condition (M = 3.39, SD = 2.04), t(129) = 6.85, p < .001, d = 1.20. Furthermore, respondents in the Shame condition reported more shame than any other emotion, all ts(66) > 5.97, ps < .01.

Logistic regression analysis showed only significant effects of Emotion condition on Withdrawal/approach choice, Wald (1) = 4.01, p = .04 (Knowledge awareness condition and interaction, Walds < 1) (see Table 3). In both the No-knowledge and Knowledge condition, an overwhelming majority of respondents in the Shame condition chose the together task whereas respondents in the Control condition did not clearly choose one task above the other, χ² (1, N = 64) = 4.16, p = .04, ϕ = .26, and χ² (1, N = 67) = 3.71, p = .05, ϕ = .24, respectively. In both the No-knowledge and Knowledge condition, Shame respondents’ selection of the together task also significantly differed from 50%, χ² (1, N = 34) = 11.77, p < .01, and χ² (1, N = 33) = 10.94, p < .01, respectively.

Table 3. Withdrawal/approach choice and withdrawal/approach preference as a function of emotion condition and knowledge condition in Experiment 2.

	Emotion condition
Dependent variable knowledge condition	Shame % (n)		Control % (n)
Withdrawal/approach choice
No-knowledge	78.8% (26)	>	54.8% (17)
Knowledgeable	79.4% (27)	>	57.6% (19)
Withdrawal/approach preference	M (SD)		M (SD)
No-knowledge	71.48 (25.69)	≥	58.52 (32.89)
Knowledgeable	71.47 (29.75)	>	52.45 (31.67)

Note: Withdrawal/approach choice reflects the percentage of respondents who chose the together task. Withdrawal/approach preference reflects the degree to which respondents preferred doing the together task, ranging from 0 (prefer alone task) to 100 (prefer together task). There are significant differences between means separated by an “>” mark with all χ² > 3.71, all ps < .05, and t(127) = 2.59, p = .01. There is a marginal difference between means separated by an “≥” mark, with t(127) = 1.72, p = .09.

A 2 (Emotion condition) × 2 (Knowledge about Shameful Event) between-subjects ANOVA with Withdrawal/approach preference as the dependent variable replicated these findings. A main effect of Emotion condition showed that respondents in the Shame condition preferred the together task more (M = 71.48, SD = 27.61) than respondents in the Control condition (M = 55.39, SD = 32.15), F(1, 127) = 9.25, p < .01,  = .07. There was no significant effect of Knowledge, F(1, 127) = 0.33, p = .56, nor a significant interaction F(1, 127) = 0.33, p = .57.

Discussion

In line with Experiment 1, the findings of Experiment 2 show that shame experiences are associated with a preference for social approach, more than with a preference for social withdrawal. The data also reveal that it did not matter for the social approach tendencies in shame whether other people were aware of the shame-inducing event or not.

General discussion

In spite of a large body of research on the antecedents and consequences of experiences of shame, little was known about the direct, social consequences of this emotion. Based on the literature, we derived two contradictory predictions regarding the social consequences of shame. While some researchers propose that shame motivates social withdrawal (e.g. Haidt, 2003; Keltner & Harker, 1998; Lewis, 1971, 1992; Tangney, 1991), others argue that shame motivates social approach (e.g. De Hooge et al., 2008; Gausel & Leach, 2011; Gausel et al., 2012; Welten et al., 2012). Surprisingly, no empirical research thus far directly compared these two opposing expectations, because existing studies did not include the possibility for respondents to choose between entering social and non-social situations. We examined this issue in two experiments. In both experiments, we found that the majority of respondents experiencing shame preferred to interact with others over being alone. This suggests that shame predominantly motivates social approach.

Implications

The current findings provide interesting avenues for future research. One potential line of future research concerns the motivations underlying people's choices to select social approach situations when experiencing shame. We believe that the most parsimonious explanation of social approach tendencies in shame is that such behaviours help to cope with the threatened or damaged self that is the central concern when feeling ashamed (Lewis, 1971). The desire to have a positive self-view is one of the most fundamental human motives (e.g. Alexander & Knight, 1971), and a large portion of our self-view comes from our belonging to others and to social groups (Leary & Baumeister, 2000). The threatened self in shame may thus be interpreted as a signal that there is a need to approach others to protect and promote one's belonging to a social group. In line with this idea, some emotion researchers have suggested that shame signals to group members one's awareness of norm-violating behaviour and one's willingness to conform to group standards (e.g. Barrett, 1995; Fessler, 1999; Gausel & Leach, 2011; Gilbert, 1997; Izard, 1977).

Another potential motivation underlying people's choices for social approach tendencies following shame may be submissive appeasement. Various researchers have suggested that appeasement may be a central function of shame (e.g. Fessler, 1999; Gausel & Leach, 2011; Gilbert, 1997), motivating social approach, gaining social approval, or de-escalating social conflict (Gilbert, 1997). Future research is therefore poised to examine the motivations underlying our findings of shame leading to social approach behaviours.

One may wonder whether similar findings would hold for dispositional shame. Both in our theorising and in our experiments, we have focussed upon experiences of shame. Although people's general tendency to experience shame, also called shame-proneness (Tangney, 1991), is related to one's actual shame experience (state shame), it does not always have similar consequences (see De Hooge et al., 2008; Nelissen, Breugelmans, & Zeelenberg, 2013). Shame-proneness has been related to negative aspects such as depression, shyness, anxiety, and anger, and it therefore remains to be seen whether shame-prone people would also opt for social approach when provided with the opportunity of situation choice.

Considering the existence of gender differences in the experiences of self-conscious emotions, it seems relevant to question whether shame's tendency to motivate social approach would depend on gender aspects. Although the current experiments include too few respondents to make any reliable claims on this topic, preliminary analyses did not reveal any consistent pattern of gender differences across the two experiments. Future research might further delve into potential gender differences in social approach consequences of shame.

Limitations

There are at least three remarks that could be made with respect to our findings. Although we think that none of these is critical to our findings, we think it is valuable to shortly discuss each. First, Experiment 2 used a performance failure to induce shame, and did not include other shame experiences such as those that stem from moral transgressions. It is possible that findings for performance failures do not translate to other shame experiences. We currently have no reason to assume that we would find different results for other shame experiences, because the autobiographical recall procedure of Experiment 1 included a range of shame experiences.

Second, one may wonder whether the current findings are specific for shame, or whether the findings would hold for negative feelings in general. Maybe negative emotions elicit tendencies to seek out others as a coping mechanism for negative feelings. Again, we have no compelling reason to believe this is the case because we found no preference for social approach over social withdrawal after inductions of guilt.

A third issue could be raised about the fact that the approach tasks in our studies were relatively neutral tasks. It is possible that shame motivates social approach when such interactions are neutral to or positive for the damaged self, but that shame motivates social withdrawal when interactions would further harm the damaged self. This converges with recent findings that shame motivates restoration of the damaged self, but that this restore motive decreases when situational factors make it too risky or difficult to restore (De Hooge et al., 2010, 2011). Neither of our studies were specifically designed to address these issues, so it might be interesting to address these in subsequent, more targeted research.

Conclusion

Two experiments demonstrated that experiences of shame motivated social approach over social withdrawal. These findings may lead to a re-appreciation of the social function of shame, to be seen no longer as the ugly sibling of guilt, but rather as a self-conscious emotion that regulates how we relate to others in our surroundings. Shame may nurture group relationships and motivate behaviours aimed at preventing exclusion from a group or by others. This is the social side of shame.

Acknowledgements

We thank Maartje Schoolderman for help with data collection of Experiment 2.

Disclosure statement

No potential conflict of interest was reported by the authors.