Cross-cultural Continuities and Discontinuities in Shame, Guilt, and Pride: A Study of Children Residing in Japan, Korea and the USA

Abstract

In a study of 144 Japanese, 180 Korean, and 688 US children, grades 3–6, differential item functioning analysis supported the cross-cultural equivalence of the TOSCA-C measure of shame, guilt, and pride. Substantial differences were observed in the mean levels of shame, guilt and pride, with Japanese children scoring highest on shame, Korean children scoring highest on guilt, and US children scoring highest on pride. The pattern of correlations, however, was more similar than different across cultures. In all groups, shame-proneness was positively correlated with aggression-relevant constructs, whereas guilt-proneness was associated with a tendency to take responsibility for failures and transgressions.

Keywords:

• Culture
• Guilt
• Pride
• Self-conscious emotions
• Shame

Notes

1. We have chosen to use the term “culture” throughout the manuscript to describe the group-related construct. We do so with caution, recognizing that we have not directly measured “culture,” but rather are using country as a proxy for culture. Although the USA, Japan, and Korea can be differentiated along multiple dimensions (e.g., cultural, political, spatial, economic), our primary interest is in cultural factors that may impact emotion, as suggested by theory. Theory does not predict cross-political, cross-economic or cross-spatial differences in self-conscious emotions. For example, Japan could become part of Australia tomorrow, but children attending fifth grade in Tokyo would still experience shame and guilt the same way. We opted to use the term “culture,” mindful of the caveats, rather than using a term that conveys no information (group) or a term tied to a construct irrelevant to theory (e.g., nation).

2. The response choices represent affective, cognitive, and motivational/behavioral dimensions associated with shame and guilt, respectively, drawn from the clinical, theoretical, and empirical literature. This is not, as some have suggested, merely a measure of behaviors (hiding vs. amending) in lieu of affect or emotion. For example, of the 15 shame items on the TOSCA-C, only one refers to a motivated behavior per se (“I would run upstairs”) and another refers to a behavioral response in conjunction with an affective reaction (“I would slide down in my chair, embarrassed”). Seven shame items refer to cognitions (e.g., “My other friends might think I'm weird … ”), four refer primarily to affect (e.g., “I'd probably feel really lousy about myself”), and two reflect a mix of cognitions and affect. Similarly, of the 15 guilt items on the TOSCA-C, none refer to present or future behaviors, eight refer to cognitions (e.g., “I should have been more careful”), and seven refer primarily to affect (e.g., “I'd feel sorry, very sorry because … ”). Phenomenological research (e.g., Lindsay-Hartz, 1984; Tangney, 1992; Tangney et al., 1996; Wicker et al., 1983) has indicated that feelings of regret are a hallmark of guilt, not shame; such research indicates that worrying about other people's evaluations of the self is a hallmark of shame.

3. In the process of translation, some scenarios were changed when the original situations were not common practice in Japan or Korea. The alternative situations were carefully chosen, keeping the nature of situations as similar as possible. For example, since “patrol duty” is not often practiced in Japan, the original scenario that read: “You are on patrol duty and you turn in three kids.” was changed into “You witnessed three classmates cheating on a quiz and you turn them in”, keeping the theme of “telling on friends”.

4. Age was not available at individual level in Japanese and Korean samples, so age-corrected grade was used as a control variable in these analyses. US children in Grades 4, 5, and 6 were on average 9.7, 10.6, and 11.6, respectively. Similarly, based on information provided by the school the Japanese children from Grades 3 and 4 were on average 8.5 and 9.5 years of age, respectively. Korean children from Grades 4 and 5 were on average 10.5 and 11.5 years of age, respectively. Thus, Korean fourth and fifth graders were recoded as fifth and sixth graders. In no case did age emerge as a statistically significant covariate. To further rule out the possibility of an age confound, secondary analyses were conducted on the subset of participants in nation-grade groups most comparable in age, with results very similar to those observed in the full sample. For example, Japanese and US fourth graders were equivalent in age. Korean fourth and fifth (recoded as fifth and sixth) were comparable to the US fifth and sixth graders. Results were essentially identical to those observed in the full sample. Finally, because school grade is only an approximate index of age, the relation of measured age to shame, guilt, and pride was examined within the US sample. Within the US sample, age was negligibly related to shame, guilt, or pride (rs = −.05, −.02, and .05, respectively, all ps > .05). Thus, although children's age was not precisely equivalent across cultural groups, it unlikely to account for group differences in self-conscious emotions. Three sets of analyses converge: within the age range considered, age does not appear to be relevant to the domain examined here. Individual-level data on socioeconomic status was not available for the Japanese and Korean samples. To assess the likelihood that SES might represent a confound with culture as it relates to self-conscious emotions, we examined the relationship of family income to shame, guilt, and pride within the US sample, where individual-level data were available. Within the US sample, income level is not related to shame, guilt, or pride (rs = −.03, .04, and −.04, respectively, all ps > .05). Thus, although SES was not precisely equivalent across cultural groups, it seems unlikely to account for the group differences in emotion observed here. Most relevant to the current paper, data on race/ethnicity were available at the individual level in the two US samples, and both samples were quite diverse. Secondary analyses (two-way analyses of variance; ANOVAs) assessed the main effect of ethnicity and its interaction with gender. In US Sample 2, there was a significant main effect of ethnicity for shame (see Table 5). African Americans were lower in the propensity to experience shame relative to White and Other participants (ps < .05). But this difference was not observed in Sample 1. (There were too few Asian participants to evaluate as a group.) In both samples, African Americans were higher on Pride than Whites. In Sample 1, Other participants were also higher than Whites. No group differences were observed for guilt, and all interactions with gender were non-significant.

5. Because significant gender differences were observed in mean levels of shame and guilt, we first analyzed the degree to which gender moderated the relation of self-conscious emotions (shame, guilt, and pride) to the four constructs of interest and among themselves, within each cultural group. Observed gender effects were no greater than chance. Thus, for the remainder of the paper, results are presented collapsing across gender.

6. The coefficients presented for shame and guilt are part correlations in which shame was factored out from guilt and vice versa. The covariation between shame and guilt reflects the fact that these emotions share a number of features in common, and that shame and guilt can co-occur with respect to the same situation (Tangney & Dearing, 2002a). This part correlational approach allows a more precise examination of unique relations of shame and guilt to theoretically relevant constructs (Paulhus, Robins, Trzesniewski, & Tracy, 2004). Fisher's r to z transformations were used to assess the differences among Japanese, Korean, and US children in the magnitude of correlates.

7. To account for the fact that different teacher measures of aggression and problem behavior were used with US children versus children in Japan and Korea, the two teacher scales were standardized within country group. That is, teacher ratings were transformed so that within each country group, teacher scores were centered at a mean of zero with a standard deviation of 1. The transformed teacher ratings from the Achenbach and Devereux measures were then merged into a single variable indicating how different each child was from his or her group mean.