Introduction

People from East Asian cultures tend to demonstrate self-enhancement less and to report lower self-regard than those from Western, especially North American, cultures (e.g., Heine & Hamamura, 2007). For instance, East Asians have lower scores than Westerners on self-report scales of self-esteem, with Japanese respondents scoring the lowest in a study of 53 nations (Schmitt & Allik, 2005). This tendency is related to social pressure, including cultural norms of modesty (e.g., Cai et al., 2007). This norm influences East Asians’ behavior and feelings; however, the modest behavior enacted by Japanese people has yet to be explored in detail.

Modesty has mainly been studied in personality psychology as a trait (e.g., Costa et al., 2011) and in social psychology as an impression management tactic, for example, as a self-presentation strategy (e.g., Leary, 1995). This study follows the latter approach.

Modesty is “the under-representation of one’s positive traits, contributions, expectations, or accomplishments” (Cialdini & De Nicholas, 1989, p. 626). In Western and Eastern cultures, individuals who emphasize their strengths or attribute their success to their abilities or efforts are seen as arrogant or boastful and thus less likable than those who do not (Bond et al., 1982; Vonk, 1999).

Yoshida, Kojo, and Kaku (1982) found that Japanese children aged 7–8 years old viewed others with a self-effacement presentation as more likable than those with self-enhancement presentation and themselves gave a more modest self-presentation in public than in private. Watling and Banerjee (2007) also found that English children aged eight and older viewed modest self-presentation as leading to better social evaluation than immodest self-presentation.

Modest behavior is observed in both Western and Eastern cultures but is thought to be more highly valued in East Asian collectivist cultures than in Western individualist ones owing to its role in group harmony (Bond et al., 1982). Furthermore, social pressure in favor of modesty might be stronger in Eastern cultures than in Western ones. For instance, Chinese college students reportedly had lower cognitive self-evaluations and global self-esteem than their North American counterparts and stronger modesty underlay these differences (Cai et al., 2007). Thus, modesty is a key concept in understanding East Asians’ behavior and feelings.

Chen et al. (2009) developed the Modest Behavior Scale (MBS) to measure behavioral aspects of everyday modesty, viewing modest behavior as a presentation tactic that reflected a cultural norm rather than a personality trait. Using the MBS, these authors investigated modest behavior in different cultural contexts. First, they collected different behavioral aspects of modesty from the literature and consultations with Chinese cultural informants and performed a scale validation with a sample of 392 Chinese college students. Through an exploratory factor analysis (EFA), a three-factor model was created: Self-Effacement (SE), namely, restraint in pursuing self-interest or denigrating oneself; Other-Enhancement (OE), that is, care for others’ self-esteem; and Avoidance of Attention-Seeking (AA), namely, the avoidance of public self-promotion. In a second study, the researchers added new items to refine the scale and performed an EFA on responses from 405 Chinese college students. A similar, albeit clearer, three-factor structure (SE, OE, and AA) was obtained. The correlations among the three subscale scores and other measures used to test the MBS’s construct validity were generally congruent with the authors’ expectations. These findings indicate that the MBS has a high validity.

In their third study, 204 Canadian undergraduates were surveyed using the MBS items and the same set of predictors as in the second study to examine whether the MBS was applicable in Western cultures. To avoid an imposed etic analysis of behavior (Berry, 1989), the authors used a derived-etic approach by adding emic items to the MBS to represent Canadian modest behavior. Thus, an integrated measure with 39 items was used. Overall, similar results were seen in the Canadian respondents. Therefore, it was confirmed that the MBS and its factor structure were applicable to Western culture, at least in a Canadian context. The Cronbach’s alpha for the three subscales ranged 0.72–0.79 for Chinese participants and 0.73–0.80 for Canadian participants.

However, it remains an open question as to whether modesty has similar behavioral manifestations in other East Asian cultures. Modesty is a traditional virtue in Japanese society, but its factor structure has not been investigated in Japanese populations. Self-effacement has been examined in Japan using scenario-related (Heyman et al., 2010), experimental (e.g., Yamagishi et al., 2012), and questionnaire (e.g., Muramoto & Yamaguchi, 2003) approaches. Muramoto and Yamaguchi (2003) identified social contexts in which people are likely to show self-effacement. However, these authors used a single item to measure self-effacement or other-enhancement as modest behavior. There is still no properly validated Japanese scale for measuring modesty or modest behavior. Validating a Japanese version of the MBS (MBS-J) would encourage empirical investigations of modesty, including its relationship to self-esteem and general psychology.

Accordingly, the present study aims to verify the validity and reliability of the MBS-J in a sample of Japanese college students. In our validation, we also examined whether a similar factor structure would emerge in a Japanese cultural setting. Because modest behavior is likely influenced by culture, the derived-etic approach was used, entailing the addition to the MBS of emic items generated from a Japanese sample, with the authors’ permission.

Overview of the Current Studies

Two preliminary and three studies were conducted herein. First, a survey collected representative Japanese modest behaviors; a second survey examined the content validity of the original MBS items and the collected new items. Second, we surveyed Japanese undergraduates using the retained items to validate the factor structure of the MBS-J and evaluate construct validity using the correlations with the scores of the theoretically related constructs, namely, trait modesty, independent and interdependent self-construal, self-efficacy, self-esteem, and approval motivation (Study 1). Third, test–retest reliability was evaluated in a two-wave survey (Study 2). Fourth, the construct validity of the MBS-J was examined in terms of self-presentation strategy and dialectical self (Study 3).

Translation

The original MBS items (Chen et al., 2009) were translated, with the authors’ permission, as follows. First, the English version was translated into Japanese by the first author. The Chinese version was also translated into Japanese by a Japanese-Chinese bilingual. The resulting Japanese translations were unified by the first author in consultation with a native English speaker and a Chinese-Japanese bilingual. Second, the Japanese version was back-translated to English by a Japanese-English bilingual and to Chinese by another Chinese-Japanese bilingual. The original and back-translated items were reviewed, and the Japanese version was revised by the first author with the same consultants until there were almost no discrepancies.

Preliminary Surveys

A survey was conducted of 19 Japanese participants to create items to identify Japanese modest behaviors. The participants were asked to freely describe all the modest behaviors that they saw in daily life, producing 99 descriptions, which were sorted by three undergraduates using the KJ method3Footnote 1 (Kawakita, 2017). The resulting items were grouped into the three categories: SE, OE, and AA. From these, 22 items were added to the original 39. Further, a second survey was conducted to test the content validity of the 61 items, and another 15 Japanese participants rated the extent to which each statement described modest behavior on a scale ranging from 1 (not at all) to 5 (very much). Twenty items that nine or more participants (60% +) rated 1–3 (1 = not at all; 2 = not much; 3 = don’t know) were deleted. Of the 20 items deleted, 13 were Chen's original items (1 SE item, 9 OE items, and 3 AAS items). Chen's OE items that were removed during this process included several items that were added for use in Canadian culture and reflected Christian (or Puritan) values or showing respect to others such as equality, hard work, and gratitude (e.g., “Treat everyone equally regardless of status”, “Finish workload on time and in an adequate manner”, “Fulfill duties to friends and family”, and “Thank the person who criticizes me”). The concept of modesty might differ slightly between Canadian and Japanese culture. The remaining 41 items were 11 SE items, 17 OE items, and 13 AAS items (of which Chen's items were 6 items, 9 items, and 11 items, respectively). Because of the large number of OE items, four items similar to other items were deleted to achieve balance (e.g., “Try to find and praise others’ strengths” and “Do things considering the other party's feelings”). This left 37 items (11 SE, 13 OE, and 13 AAS items).

Study 1

Study 1 aimed to validate the factor structure and evaluated the construct validity of the MBS-J with a sample of Japanese undergraduates. The construct validity of the MBS-J subscales was assessed using the correlations with Trait Modesty, Independent and Interdependent Self-Construal, Generalized Self-Efficacy, Self-Esteem, and Approval Motivation.

First, we hypothesized that three-factor structure of the MBS-J would emerge in Japanese undergraduates. Second, following Chen et al. (2009), the subscale scores were hypothesized to be positively correlated with Trait Modesty. Third, like Chen et al. (2009), we predicted that the subscales would be positively correlated with interdependent self-construal, namely, a view of the self as fundamentally connected to others (Markus & Kitayama, 1991), but negatively or not correlated with independent self-construal, that is, a view of the self as separate from others (Markus & Kitayama, 1991). Interdependent self-construal leads people to behave modestly to promote harmonious relationships with others; independent self-construal encourages expression of individual desires, preferences, or abilities (Chen et al., 2009; Markus & Kitayama, 1991). Fourth, SE was hypothesized to be negatively correlated with the self-reported Self-Esteem and Self-Efficacy. Self-esteem is “an individual’s subjective evaluation of his or her worth as a person” (Orth & Robins, 2014, p. 381); self-efficacy is an individual’s beliefs about their ability to achieve a task (Bandura, 1977). Those with greater self-effacement are expected to rate themselves lower on self-report scales because they downplay their abilities and achievements. Finally, the MBS-J was hypothesized to be positively correlated with Approval Motivation because modesty is a strong social norm in East Asian culture, so higher approval motivation would induce more conformity with the norms of modesty. Trait modesty, independent and interdependent self-construal, and self-efficacy were used by Chen et al. (2009); the constructs of self-esteem and approval motivation were introduced in this study.

Method

Participants

The participants were 278 Japanese university students (168 male, 108 female, and 2 unspecified), mainly first and second year students taking Introduction to Psychology at Tokyo City University, Yokohama Campus. The mean age was 19.51 years (SD = 2.22).

Measures

MBS-J

The MBS-J comprises 37 items that remained after the preliminary surveys. This scale measured SE, OE, and AA. The respondents were asked to report their daily behaviors. All the responses were given on a five-point Likert scale (1 = does not describe me; 5 = describes me).

Modesty Subscale of the Revised NEO-Personality Inventory (NEO-PI-R)

The modesty facet of the Agreeableness factor was measured with a Japanese version of the NEO-PI-R (Costa et al., 2011), which has eight items. The participants responded to a five-point Likert scale (1 = strongly disagree; 5 = strongly agree).

Independent and Interdependent Self-Construal Scale

Cultural self-construal was measured using the Independent and Interdependent Self-Construal Scale (Takata, 1999)—the most commonly used cultural self-construal scale in Japan, comprising 20 items, 10 measuring the independent self and 10 measuring the interdependent self. The responses for both scales were given on a seven-point Likert scale (1 = strongly disagree; 7 = strongly agree).

Generalized Self-Efficacy Scale

Of the 23 items on the Japanese Generalized Self-Efficacy Scale (α = .88; Narita et al., 1995; Sherer et al., 1982), the 12 that showed higher factor loadings in Narita et al. (1995) were chosen to assess generalized self-efficacy to reduce the burden on participants. The responses were given on a five-point Likert scale (1 = disagree; 5 = agree).

Self-Esteem Scale

Yamamoto, Matsui, and Yamanari’s (1982) Japanese version of the 10-item Self-Esteem Scale (Rosenberg, 1965) was used. The participants responded to a four-point Likert scale (1 = disagree; 4 = agree). This scale is widely used in Japan, but many studies exclude item 8 because its factor loading is reportedly poor; this was also the case in the present research. Therefore we used nine items.

Martina Larsen Approval Motivation Scale

Approval motivation was measured using a 20-item Japanese version of the Martina Larsen Approval Motivation Scale (Larsen et al., 1976; Ueda & Yoshimori, 1991). Each item was measured on a five-point Likert scale (1 = strongly disagree; 5 = strongly agree).

Procedure

The questionnaires were administered in a classroom setting. The participants were informed that the questionnaires were anonymous, that their response would be used for research, and that there was no penalty for not answering. All the participant groups filled out the MBS-J. Five other measures were administered to two sub-samples to test construct validity: the Modesty Subscale of NEO-PI-R and the Martina Larsen Approval Motivation Scale to a sub-sample of 98 undergraduates (66 male, 31 female, and 1 unspecified; mean age = 18.60 years, SD = 1.22) and the Self-Construal Scale, Self-Esteem Scale, and Generalized Self-Efficacy Scale to a sub-sample of 180 undergraduates (102 male, 77 female, and 1 unspecified; mean age = 20.00 years, SD = 2.47). Each survey took approximately 15 min.

Results

Preliminary Analysis

The data were analyzed using SPSS 25.0 and Amos 23.0. Before performing an EFA, descriptive statistics for each item, the Kaiser–Meyer–Oklin (KMO) measure and Bartlett’s test were used to evaluate the adequacy of the data for a factor analysis. Three items were dropped owing to ceiling effects, where more than 30% of the respondents chose the highest score. The skewness and kurtosis values for the retained items were acceptable (< ± 2) for the assumption of univariate normality. The KMO value was 0.82, and Bartlett’s test of sphericity was statistically significant, indicating the suitability of the factor analysis (Tabachnick & Fidell, 2013).

EFA

An EFA using maximum likelihood with promax rotation was conducted for the 34 retained items to examine the factor structure and determine the most useful items for measuring each dimension of modesty. The initial eigenvalues were 6.66, 2.90, 2.43, and 1.52 for the first four factors. The scree plot and interpretability of the factors indicated a three-factor solution comprising AA, SE, and OE, which corresponded to the results of Chen et al. (2009). These accounted for 35.29% of the total variance. The three factors had moderate positive correlations (r = 0.31–0.39). After deleting the items loaded lower than 0.30 on all factors or high on two or more factors one by one, the solution, with 27 retained items, showed that each item had high loading on the assumed factor. From the remaining 27 items (9 for AA, 10 for SE, and 8 for OE), we selected 24 (8 for each factor) for the final scale to improve usability. During this selection, we accounted for the appropriateness of the expression, content, and factor loading of each item as well as whether it was original, preferentially retaining Chen et al.’s (2009) items to enable cross-cultural comparison in the future. Specifically, item 34 ("I don't brag about myself until people ask me about it") was the only new item for AA, so it was deleted, and the original items were retained in priority. In addition, because Items 2 ("Try to make people around me think I am not superior") and 25 ("Try to make myself look inferior to the other party by talking about my failures"), which measure SE, were deleted because they are extreme in content and imply self-deprecation rather than self-modesty. Thus, three items were dropped, leaving 24 for the confirmatory factor analysis (CFA).

CFA

A maximum-likelihood CFA was performed on the 24 items to assess the overall fit of the hypothesized correlated three-factor model. Data from 265 undergraduates (156 male, 107 female, and 2 unspecified) who had no missing values for the items were used. According to Wolf et al. (2013), the minimum sample size for three-factor, six-indicator (for each factor) model with factor loadings of 0.50 is 190. Our sample size satisfied the criterion. The following criteria were used to evaluate the models: goodness of fit index (GFI) ≥ 0.90, root mean square error of approximation (RMSEA) ≤ 0.06, and standardized root mean square residual (SRMR) ≤ 0.08 (Hu & Bentler, 1999), adjusted goodness of fit index (AGFI) ≥ 0.85 (Schermelleh-Engel et al., 2003), Chi-square/degrees of freedom (χ2/df) ≤ 2 (Schreiber et al., 2006). In addition, the Akaike information criterion (AIC; Akaike, 1987) of smaller values was used to assess a better fit.

The fit indices for the hypothesized model were as follows: χ2/df = 2.09, GFI = 0.86, AGFI = 0.83, RMSEA = 0.06, SRMR = 0.08, AIC = 621.77. Some indices showed a good fit to the data; others did not. Therefore, the model was modified to improve fit. In determining the model modifications, we preferred to retain, wherever possible, the original items of Chen et al. (2009). Using the modification indices, we excluded three items and added two correlated errors. Specifically, we deleted item 27 (“I let others go ahead of me”), which was a new item belonging to OE, as an additional path for AA was recommended and its estimated factor loading was significant (0.21, p = 0.001). This results in a simple structure where each item has a high loading for only one factor. Item 22 ("I don't talk about my efforts to succeed in something like an exam", a new item) had the second lowest factor loading for SE and was also deleted because the content of "efforts for exams" was deemed inappropriate for future use with the elderly. Item 12 (“In front of others, I attribute success to luck rather than my own ability”) was also removed following the modification indices. It had the lowest factor loadings (0.32) on SE, which showed a large difference from the factor loadings of other items (0.49–0.61). Finally, within-factor error covariances between items 13 and 18 as well as between items 14 and 15 were included in the model because the modification indices indicated that these error covariances would lead to substantial reductions in χ2, and the pairs had similar content or wording. For e13 and e18, it would be reasonable to set a covariance for their errors in the sense that one of the specific actions of item 13 (“Lead people around me to acknowledge my superiority”) is the content of item 18, which is to post awards where people can see. For e14 and e15, it would make sense to set a covariance for their errors, since they have in common the denial of praise. This modified model was tested; the indices were χ2/df = 1.96, GFI = 0.89, AGFI = 0.86, RMSEA = 0.06, SRMR = 0.07, AIC = 454.23, showing a clear improvement. All indicated good fit except GFI, which suggested further research was needed. The standardized factor loadings ranged from 0.30 to 0.79 (ps < 0.001). The three factors had moderate positive correlations (r = 0.26–0.55). The descriptive statistics and factor loadings of the 21 items in the final model are shown in Table 1.

Table 1 Descriptive statistics and factor loadings in a confirmatory factor analysis for the MBS-J (Study 1)
Full size table

Four alternative models were also tested with the same 21 items. Specifically, these were a two-factor model combining AA and SE into one factor, a two-factor model combining SE and OE into one factor, a two-factor model combining AA and OE into one factor, and a one-factor model where all items are loaded onto one factor. The three-factor model showed a much better fit, supporting the three-factor structure of the MBS-J (Table 2). The descriptive statistics, Cronbach’s α, and correlations among MBS-J scores are shown in Table 3.

Table 2 Fit index of each factor model for the MBS-J (Study 1)
Full size table
Table 3 Descriptive statistics, Cronbach’s α, and correlations among MBS-J scores (Study 1)
Full size table

Overall, these results indicate that the 21-item MBS-J has good factorial validity among Japanese undergraduates. Cronbach’s alpha values indicate appropriate internal consistency reliability for each scale.

Construct Validity

Pearson’s correlation coefficients for the 21-item MBS-J with other scales and the Cronbach’s alphas are shown in Table 4. Trait Modesty was positively and highly correlated with AA and SE but not with OE. This supports Chen et al. (2009). For Self-Construal, Interdependent Self was positively correlated with SE and OE. Independent Self was negatively correlated with SE and positively correlated with OE. AA was correlated with neither of these. Furthermore, as expected, SE was negatively correlated with both Self-Efficacy and Self-Esteem. Approval Motivation had positive correlations with SE, AA, and OE, as hypothesized. These results support the construct validity of the MBS-J.

Table 4 Descriptive statistics and correlations between the MBS-J and other scales (Studies 1, 2 and 3)
Full size table

Gender Differences

Though it is desirable to validate the measurement invariance across gender using multi-group CFA, the number of men and women in this study was too small to perform the CFA for the three-factor model with six indicators for each factor adequately. Therefore, an independent-samples t-test was conducted to examine gender differences. Women (M = 3.83, SD = 0.55) had higher scores than men (M = 3.61, SD = 0.58) for OE, t (265) = 3.16, p < 0.01, d = 0.40. There were no significant gender differences for AA (M = 3.41, SD = 0.61 for women, M = 3.35, SD = 0.60 for men, t (269) =  − 0.80, ns) or for SE (M = 3.23, SD = 0.69 for women, M = 3.26, SD = 0.62 for men, t (271) =  − 0.47, ns).

Discussion

Upon investigating the factor structure of the MBS-J in a sample of Japanese undergraduates, in line with Chen et al. (2009), the correlated three-factor scale structure was validated in our sample via an EFA and CFA. Our three-factor model fit the data better than a two-factor or one-factor model. This suggests that modest behavior manifests in the aspects of avoidance of attention-seeking, self-effacement, and other-enhancement in the Japanese cultural setting. To avoid an imposed etic behavior analysis, we added emic items generated from a Japanese sample on the MBS (Berry, 1989). Nevertheless, most of the items retained for AA and OE were drawn from Chen et al. (2009), confirming the similarity of these factors between Chinese and Japanese respondents. However, for SE, four items of six were new. Four original items were deleted because of the ceiling effect, cross-loading, or lower factor loading. However, an overall meaning of the content appeared that nearly matched Chen et al. (2009). Thus, the three components of modest behavior evidenced in the Japanese sample were similar to prior research.

Previous studies have not found consistent gender differences in the MBS: there were no gender effects in a sample of Chinese and Canadian college students in Chen et al., (2009; Study 2). In another investigation, Chinese men were more other-enhancing and avoiding attention-seeking than women (Chen et al., 2017). In our study, Japanese women were more other-enhancing than men. Reportedly, women and girls are expected to have more consideration and warmth for others (e.g., Eguchi & Hamaguchi, 2012). The results herein are congruent with such studies. However, since measurement invariance across gender has not been validated, it needs to be examined in the future.

Our results also indicated that the MBS-J had good internal consistency reliability and construct validity. Construct validity was examined using the correlations with theoretically related measures, such as trait modesty, independent and interdependent self-construal, self-efficacy, self-esteem, and approval motivation. The result that OE is positively correlated with independent self was somewhat unexpected. However, considering the characteristics of independent and interdependent self-construal, i.e., that those with independent selves value to express one’s internal needs and capacities, whereas those with interdependent selves value to be receptive to others, to adjust to their needs, and to restrain one’s own inner desires (Markus & Kitayama, 1991), it makes sense that OE, care for others’ self-esteem such as actively praising others, is positively correlated with independent self-construal.

Overall, the MBS-J subscales showed adequate correlations with those measures, supporting the hypotheses. In addition, despite a moderate correlation among the subscales, the three subscale scores could be distinguished by their relationship with other measures. These variations in the subscales ensured MBS-J’s effectiveness and validity. In summary, the MBS-J is a valid and reliable measure for modest behavior.

Study 2

Study 2 examined the test–retest reliability of the MBS-J.

Method

Participants and Measure

Some undergraduates who participated in Study 1 completed the MBS-J twice, three weeks apart (N = 45). The second sample comprised 23 male and 22 female participants. The mean age was 20.71 years (SD = 4.45). In order to detect a Pearson’s correlation coefficient of r = 0.60 with 80% power (alpha = 0.001, two-tailed), G*Power (Faul et al., 2009) suggests we would need 39 participants. Since we had 45 participants, the statistical power was calculated to be 0.76 (α = 0.001, r = 0.55, two-tailed) and 0.89 (α = 0.001, r = 0.60, two-tailed), respectively.

Procedure

The survey was conducted as in Study 1. Data for those who completed the survey at both points were analyzed. There were no missing values.

Results and Discussion

The Cronbach’s alpha coefficients for AA, SE, and OE were, respectively, 0.76, 0.77, 0.71 for Time 1, and 0.72, 0.73, 0.72 for Time 2. Pearson’s correlations were used as indicators of test–retest reliability: these were 0.60, 0.55, and 0.59 for AA, SE, and OE, respectively. All test–retest coefficients were significant at p < 0.001. It means that the scale will yield the same result in the same situation at a different point in time, i.e., it is stable over time. Thus, the MBS-J has good test–retest reliability.

Study 3

The MBS-J’s construct validity was examined in terms of self-presentation and the dialectical self (Peng & Nisbett, 1999). Modest behavior is considered a self-presentational strategy for gaining acceptance from others (Leary, 1995). Although SE was significantly negatively correlated with Rosenberg’s self-esteem in Study 1, it is unclear as to whether those who scored higher on SE rated their self-esteem lower because of a self-presentational strategy or because of actual lower self-esteem. The present study used both explicit and implicit measures of self-esteem to determine whether the significant negative correlation between SE and Rosenberg’s self-esteem was due to self-presentational motivation. The Implicit Association Test (IAT; Greenwald & Farnham, 2000) minimizes the influence of self-presentational effects (Banaji & Greenwald, 2013; Kobayashi & Greenwald, 2003). If people who score higher on SE have lower self-esteem in actuality, they would be more likely to have lower scores for implicit self-esteem. Specifically, SE would be negatively correlated with implicit self-esteem. If those with higher self-effacement tendencies evaluate themselves lower on a self-esteem scale because of their self-presentation strategy, SE should not be correlated with implicit self-esteem or should be less correlated with implicit self-esteem than with explicit self-esteem. This study posited the latter hypothesis.

Second, studies on naïve dialecticism (Peng & Nisbett, 1999), a concept that explains cultural differences, have accumulated over the last two decades. Naïve dialecticism is East Asians’ lay belief that reality is changeable and contradictory and that parts cannot be understood separately from their relationship to the whole because all phenomena are complexly intertwined (Peng & Nisbett, 1999). Individuals with dialectical thinking believe that opposite or contradictory concepts such as “good” and “bad” coexist in everything, including the self, and tend to show more tolerance of contradiction (Spencer-Rodgers et al., 2010). The Dialectical Self Scale was developed to measure the dialectical thinking in self-perception (Spencer-Rodgers et al., 2015).

In hierarchical Japanese society, individuals with greater cultural modesty norms tend not to object to others’ opinions, particularly their superiors, for fear of violating the code of conduct. Thus, they tend to experience a large discrepancy between their private feelings and public behavior. They must accept this discrepancy to successfully adjust to social life. Accordingly, modest behavior and the dialectical self are thought to be closely connected in that modest people strengthen their dialectical thinking in the self or that dialectical thinking encourages modest behavior. Therefore, we hypothesized that the dialectical self is positively correlated with modest behavior, particularly with SE and OE.

Method

Participants

The participants were 62 Japanese college students at the Tokyo City University, Yokohama Campus. Participants were recruited in psychology classes and computer rooms. The participants were given course credit or volunteered. Following Greenwald and Farnham (2000), eight participants were excluded for having error rates on the IAT higher than 20%. The data for the remaining 54 undergraduates (38 male and 16 female) were used in the subsequent analysis. The mean age was 20.93 years (SD = 1.16). There were no missing values. In order to detect a Pearson’s correlation coefficient of r = -0.45 with 80% power (α = 0.01, two-tailed), G*Power suggests that 53 participants are needed. In Study 1, the correlation between SE and Rosenberg Self-Esteem was -0.45, so this value was used to calculate the sample size. There were 54 participants, so the statistical power was calculated to be 0.67 (α = 0.05, r = -0.32, two-tailed).

Measures

MBS-J

Modest behavior was measured using the 21-item MBS-J developed in Studies 1 and 2. In this study, the Cronbach’s alpha coefficients for AA, SE, and OE were .77, .68, and .68, respectively.

Explicit Self-Esteem

A validated Japanese version of the 10-item Self-Esteem Scale (Rosenberg, 1965) by Mimura and Griffiths (2007) was used. The participants responded to a four-point scale (1 = strongly disagree; 4 = strongly agree). We used all the items to produce a total score because all factor loaded properly (.20–.71).

Implicit Self-Esteem

The self-esteem IAT (Greenwald & Farnham, 2000) was administered via personal computer. This categorization task assesses automatic associations of the self with pleasant or unpleasant words. The participants were asked to press either the “E” key with their left forefinger or the “I” key with their right forefinger to identify the stimulus on the monitor as either self or other or as either pleasant or unpleasant as quickly as possible. Five self-descriptive stimuli (watashi, watashi-ni, watashi-no, jibun-no, and jibun-ni (I, me, my, mine, and self)), other-descriptive stimuli (karera-ni, kanojo-ra, tanin, tanin-no, and hokano-hito (them, they, others, their, and other)), pleasant stimuli (yoi, akarui, kirei-na, kasikoi, and uresii (good, cheerful, beautiful, smart, and happy)), and unpleasant stimuli (warui, tsurai, kitanai, oroka-na, and kanasii (bad, painful, dirty, stupid, and sad)) were used. The descriptive stimuli were taken from Kobayashi and Greenwald (2003) with minor modifications.

The program randomly selected items from the appropriate stimuli lists. When an incorrect response was made, an “X” was presented at the center of the bottom of the monitor, and the participant was asked to press another key. Each stimulus item was displayed until a correct response was made; the response time was recorded. The interval of time between trials was 250 ms. The IAT comprised 7 blocks, with 20 trials for practice blocks and 40 for data-collection blocks. Blocks 4 and 7 were the critical blocks used to calculate the IAT effect. The data-collection block order (i.e., whether the Self + Positive block was encountered first or second) was counterbalanced. The program was improved and revised using open source web-based IAT program code (Mason et al., 2016).

Dialectical Self

A 32-item Japanese version of the Dialectical Self Scale (Suzuki et al., 2012; Spencer-Rodgers et al., 2015) was used to assess dialectical thinking in the domain of self-perception (e.g., “When I hear two sides of an argument, I often agree with both”). The participants responded to a seven-point Likert scale (1 = strongly disagree; 5 = strongly agree).

Procedure

In each session, 2–11 participants participated. After sitting at a personal computer separate from the other participants and signing an informed consent form, each participant completed the self-esteem IAT and then completed a self-report questionnaire that assessed self-esteem, modest behavior, and the dialectical self. Finally, the participants were debriefed. The process took approximately 20 min in total.

Results and Discussion

Following Greenwald, McGhee, and Schwartz (1998), the first two trials in each block were discarded; individual response times faster than 300 ms or slower than 3000 ms were recoded to 300 ms and 3,000 ms, respectively; and individual response time was log-transformed to normalize the distribution. Each participant’s implicit self-esteem was computed by subtracting the mean response latency for the Me + Positive block from that for the Me + Negative block. Higher scores indicated greater positive implicit self-esteem.

One sample t-test for explicit and implicit self-esteem against the scales’ theoretical midpoints (2.5 and 0, respectively) showed that positivity bias was significant for implicit self-esteem (t(53) = 15.32, p < 0.001, d = 2.08) but nonsignificant for explicit self-esteem (t(53) = 0.27, ns), supporting the findings of previous studies that employed Japanese samples (Yamaguchi et al., 2007). This indicated that Japanese people implicitly regard themselves in a positive manner, although they do not explicitly grant themselves high evaluations.

The descriptive statistics and correlations between the MBS-J, explicit and implicit self-esteem, and the dialectical self are shown in Table 4. It was expected that SE would not correlate with implicit self-esteem or would not do so as highly as with explicit self-esteem. Table 4 shows that SE had a strong negative correlation with explicit self-esteem, and the correlation between SE and implicit self-esteem was almost zero. These results support the hypothesis, suggesting that those who efface themselves tend to show lower explicit self-esteem scores for a self-presentational motivation rather than because of actual low self-regard. Moreover, it was predicted that modest behavior would be positively correlated with dialectical self, particularly with SE and OE. Table 4 shows that SE and OE had positive correlations with dialectical self, supporting the hypothesis. Thus, those with greater dialectical self tend to deprecate themselves and enhance others.

Taken together, although the statistical power is not enough, this suggests that those with a higher self-effacement tend to score lower on self-report self-esteem scale because of a self-presentational strategy instead of lower self-esteem. Thus, overall, these results demonstrate that the 21-item MBS-J has good construct validity.

General Discussion

The present study examined the factor structure, construct validity, and reliability of the MBS-J in a sample of Japanese college students, indicating that the scale has good factorial validity, construct validity, and internal consistency reliability, and test–retest reliability. Although we added emic items generated from a Japanese sample to the MBS items, the MBS-J showed a three-factor structure consisting of AA, SE, and OE, which was consistent with Chen et al. (2009). Furthermore, construct validity was examined using correlations with theoretically related measures, such as trait modesty, independent and interdependent self-construal, self-esteem, self-efficacy, approval motivation, and dialectical self. The MBS-J subscales had adequate correlations with those measures. For self-esteem, it was hypothesized that those with higher self-effacement tend to report lower explicit self-esteem owing to a self-presentation strategy. To test this hypothesis, explicit and implicit self-esteem, as measured by an IAT cognitive task, was considered (Study 3). The results showed that SE was strongly negatively correlated with explicit self-esteem and not correlated with implicit self-esteem, supporting the hypothesis. These findings confirm the construct validity of this scale for modest behavior performed as self-presentation. This study produced evidence for construct validity using implicit and self-reported measures.

The content of the AAS and SE was the same as Chen’s original scale, although some items were removed and new items were adopted. The content of the OE was the same as that of Chen’s scale, except those items reflecting Christian values and respect for others, which were added for use in the Canadian culture, were excluded in the preliminary survey. Therefore, with regard to OE, it is possible that the concept of modesty slightly differs between Canadian and Japanese culture, but this point was left as a future issue.

Modesty is a key concept in understanding East Asians’ behavior and feelings, but until now there is no properly validated Japanese scale for measuring modesty. This study facilitates future research on modesty within Japanese culture and between Japanese and other East Asian or Western cultures by providing a reliable and valid tool for assessing modesty in Japanese late adolescents.

Limitations and Future Directions

First, the results herein were drawn from university student samples; consequently, it is unclear whether the scale is valid or reliable for other groups. Future studies should investigate the psychometric properties of the MBS-J for various age groups; this would allow researchers to examine how the three aspects of modesty develop or change with age.

For example, regarding the cultural self-construal in Japanese, Takata (1999) found that independence increases from adolescence to late adulthood (ages 60 +), while interdependence declines from adolescence to middle adulthood (ages 40–59). Considering that interdependence leads people to behave modestly to promote harmonious relationships with others, and that SE is positively correlated with interdependence and negatively correlated with independence in this study, self-effacement is expected to decline from adolescence to middle adulthood. This prediction assumes that modesty is well developed by adolescence. Future research is needed to validate it, including whether or not the assumption is correct. Such studies could help indicate how social norms of modesty are internalized and modest behavior manifest.

Second, since measurement invariance across gender was not validated, future research is needed to confirm gender difference. Third, although the MBS-J was shown to have factorial validity and construct validity, the predictive validity of the scale should be examined in terms of how well it predicts actual modest behavior. Lastly, this study suggests that higher self-effacement leads some to rate themselves lower on self-report scales for self-esteem. Therefore, self-effacement should be controlled when comparing self-report self-esteem scores across cultures. The MBS could be used as a control in such investigations. Future studies are required to validate the scale across cultures.