Many researchers have suggested that rumination, or “dwelling in the past,” is associated with worse emotional and mental health issues (for a review, see
Aldao et al., 2010), even though such thinking is common (
Papageorgiou & Wells, 2001b). Although there have been efforts to understand variations in the associations between rumination and mental health (e.g.,
Ayduk & Kross, 2010;
Ciarocco et al., 2010;
Pennebaker & Graybeal, 2001), such studies have been conducted mostly within Western contexts. However, a growing body of cross-cultural research suggests that rumination may not have the same maladaptive effects across cultures. While people in East Asian cultures engage in rumination more compared with Western cultures, it is associated with less harmful correlates in East Asian cultures (
Chang et al., 2010;
Kwon et al., 2013). Despite such findings, the underlying mechanism is not fully understood yet.
Culture and Perception of Change
Based on cross-cultural comparisons of East Asians and European North Americans,
Peng and Nisbett (1999) suggested that cognitive differences exist between Western and Asian cultures in the degree to which they are oriented toward the interconnections in the universe (i.e., analytical vs. dialectical thinking). One difference is in how people perceive change. In East Asian cultural contexts, people tend to engage in dialectical thinking, which involves a belief that objects, events, and states of being in the world are continuously alternating between two extremes or opposites (
Peng & Nisbett, 1999). Thus, they perceive various states and objects as malleable. On the contrary, in European American cultural contexts, people tend to engage in analytical thinking, and expect the states of the world to be more stable and changes to occur in a linear trend. Such varying assumptions about the world have been found to affect how one predicts whether a current situation will change in the future (
Ji et al., 2001). Researchers found that Chinese participants predicted a greater likelihood of change in a variety of scenarios (e.g., a person who has been winning would lose the next game; adversaries will become lovers) compared with Americans. Furthermore, Americans believed that their happiness across time was more or less linear, whereas Chinese believed that their life happiness was nonlinear (
Ji et al., 2001).
Studies have also shown that dialectical thinking leads East Asians to tend to perceive personal attributes and behavior to have less consistency and stability across contexts, domains, and time, compared with Westerners (
Spencer-Rodgers et al., 2010). For instance, when asked to predict what would happen to someone’s traits, abilities, and behaviors in the future, Chinese participants believed a person’s traits, behaviors, and abilities would change more than did Canadian participants (
Ji & Zhang, 2003). In addition, compared with European Americans, East Asians perceive greater changes in their own dispositions than European Americans (
Spencer-Rodgers et al., 2009). Similarly, building on implicit theories of attributes (
Dweck, 2006;
Dweck et al., 1995), researchers have suggested personal attributes, such as intelligence, are considered to be more malleable in East Asian cultures than in Western cultures (cf.
Heine et al., 2001). Taken together, there is considerable evidence on cultural differences in the extent to which people see change in various aspects of the world, including personal attributes and behavior of others and their own.
The Role of Attribution of Rumination
Cultural differences in perception of change may influence the extent to which people attribute the act of rumination to motivation for change and to improve or self-doubt over one’s ability. If Asians are more likely than Westerners to perceive change and flexibility in personal attributes and behaviors, they may be more likely to see the room for change after a negative experience (e.g., difficult exam, poor impression during an interview). Thus, ruminative thinking may be considered to be a way to think about how one could overcome and avoid past failure the next time one encounters a similar situation, and be perceived as driven by one’s motivation to do better (i.e., self-improvement attribution). On the contrary, if Westerners are less likely than Asians to perceive changes in personal attributes and behaviors, they may see less room for change after a negative experience. Therefore, the act of ruminating may be considered to be less productive and perceived as a manifestation of doubt about one’s ability (i.e., self-doubt attribution). It is important to note that attributions to self-improvement and self-doubt are not mutually exclusive or opposites. Because one can simultaneously be motivated to overcome the current situation while holding doubt over one’s ability to do so, people may infer that both self-improvement and self-doubt can be reasons for rumination. Thus, people may attribute ruminative thinking to both self-improvement and self-doubt, just to different degrees.
Cultural differences in attributions of rumination, in turn, may play a role in moderating the relationship between rumination and the subsequent outcomes. While rumination may be related to negative outcomes (e.g., depression) in general, such association may be weaker for those who attribute the act of rumination more to self-improvement or less to self-doubt. By believing that rumination is more for reasons of improvement or less for doubt, the act of ruminating itself may not be as detrimental for the individual. Rather, they may engage in such a thinking process as an opportunity to develop and improve.
In line with our theorization, studies on implicit theories of attributes (
Dweck, 2006;
Dweck et al., 1995) have shown positive effects of having a growth mindset (i.e., perceiving personal attributes as malleable and susceptible to growth) compared with a fixed mindset, such as dampened associations between stressful life events and psychological distress (
Schroder et al., 2015). Furthermore, more relevant to rumination, other researchers have shown that rumination can have weaker associations with negative outcomes when it was action-focused rumination (i.e., focusing on correcting past mistakes and active goal achievement) than state-focused rumination (i.e., focusing on the failure;
Ciarocco et al., 2010). Although action-focused/state-focused rumination and self-improvement/self-doubt attribution differ in whether they target content or attribution of rumination, they share some similarities as well; both focus on possible positive and negative aspects of ruminating about the negative event. Considering such similarities, making more self-improvement and/or less self-doubt attribution about rumination may lead to reduced associations between rumination and psychological correlates, thereby providing a potential mechanism underlying cultural differences in the association.
Study 1
Study 1 was conducted to develop and validate the Attribution of Rumination scale and to examine whether people’s attributions for rumination vary by cultural background. Several studies suggest that how people perceive other’s thinking and beliefs may influence their own psychological processes over and beyond their own thoughts and beliefs (e.g.,
Chiu et al., 2010). In addition, considering cultural differences in self-serving attribution (
Mezulis et al., 2004), asking questions about the self may confound such a tendency (i.e., European Americans may be less likely to endorse negative statements and more likely to endorse positive statements about the self). Therefore, the Attribution of Rumination scale asks respondents to make attribution of another student’s act of rumination rather than that of themselves. We predicted that compared with European Americans, East Asian descents would be more likely to attribute rumination to self-improving motivation (i.e., self-improvement attribution) and less likely to attribute rumination to doubt over one’s ability (i.e., self-doubt attribution).
We report the psychometric properties of the scale as well as results from exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) to validate the hypothesized factor structure of the scale. In addition, we report the results from measurement invariant testing to assess the psychometric equivalence of a construct across our two cultural groups of interest (i.e., European Americans and East Asians). Finally, we report the results that tested the cultural differences in the pattern of attributing rumination.
Method
Item generation and pilot studies
We focused on a negative experience that would be most relevant to students: academic stress (i.e., exams). The scale presents a hypothetical individual ruminating over a stressful academic situation (i.e., doing poorly on an exam) and asks respondents to judge the extent to which self-improvement or self-doubt are potential reasons why the individual is ruminating. Self-improvement items focused on possible changes and improvement in their grades or performance (e.g., “The student wants to improve his or her grades”; “The student wants to do better on the next exam”) through motivation (e.g., “The student is motivated to do better”). These items were relevant to perceived malleability in one’s ability to perform on the exam. On the contrary, self-doubt attribution items involved doubt over their ability (e.g., “The student is doubting if he or she has the capability needed for the class”; “there is nothing he or she can do to do better”) and doubt over better future outcome (e.g., “he or she will not be able to get a better grade”). These items reflect the (low) belief in malleability of one’s ability. Each item was rated on a 7-point Likert-type scale (1 = very unlikely, 7 = very likely).
To test the internal consistency and robustness of the initial items, we administered the scale across three separate pilot samples (Sample 1: 16 University of Wisconsin [UW]–Madison students, Sample 2: 84 Mechanical Turk [MTurk] workers,
1 and Sample 3: 101 MTurk workers
2). With Sample 3, the internal consistency suggested that the scale was appropriate for broader dissemination. The final scale appears in the
appendix.
Participants and procedure
A sample of at least 500 is recommended when conducting CFA using maximum likelihood robust estimation (
Bandalos, 2014). The final sample consisted of 1,468 UW-Madison college students taking an introductory psychology course:
M age = 18.64 (
SD = 1.18), 59.06% female, 73.30% White/Caucasian, 17.92% Asian, 1.29% African American/Black, 2.79% Hispanic/Latinx, 0.68% Arab/Middle Eastern, 0.27% Native American/American Indian/First Nation, 0.07% Pacific Islander/Native Hawaiian, 3.61% Multiple, and 0.07% Other. When testing measurement invariance (i.e., testing whether the scale is compatible across different groups) and our predictions, we focused on European American and East Asian descents (
N = 1,339) because our predictions and theoretical background are based on the comparison between European Americans and East Asians. East Asian participants were defined as those with ancestral backgrounds from China, South Korea, or Japan. European Americans were defined as those who are considered Caucasian, and their native language is English. If the participant identified as being from a background other than East Asian and/or European American background, or both, they were excluded from the final sample. Participants completed the questionnaire as part of a mass survey within the first 2 to 3 weeks of the semester.
Results
Psychometric properties of the Attribution of Rumination: Means, variance, and internal consistency
The mean, variance, and skewness of each attribution facet were examined to see how individuals responded to each facet. The mean score of self-improvement attribution was 5.63 (
SD = 0.96) and that of self-doubt attribution was 4.91 (
SD = 1.16). Both attribution factors have high internal consistency; Cronbach’s alpha was .78 for the self-improvement attribution and .81 for the self-doubt attribution (
Table 1). The two factors were weakly positively correlated,
r = .09. Responses for each attribution facet did not pass the tests of normality as they were negatively skewed. To deal with the non-normality of the data, we used the maximum likelihood robust estimation for CFA.
EFA
We used random sampling to split the sample into Sample 1 (
n = 740) for EFA and Sample 2 (
n = 728) for CFA. There was no significant difference in age and gender between the two randomized samples. We first started with EFA to assess the underlying factor structure of the scale using maximum likelihood estimation and oblique solution. The decision on the number of factors to extract was based on parallel analysis (
Horn, 1965). EFA results using Sample 1 suggested retaining two factors that accounted for a meaningful variance.
CFA
CFA using maximum likelihood robust estimation (
Rosseel, 2012) was conducted to evaluate the EFA-informed a priori theory about the measure’s factor structure and psychometric properties (
Costello & Osborne, 2005;
Henson & Roberts, 2006;
Worthington & Whittaker, 2006). Full information maximum likelihood (FIML) was used to treat missing values (
Brown, 2006). To assess the absolute model fit, we used root mean square error of approximation (RMSEA), standardized root mean square residual (SRMR), and comparative fit index (CFI) as our criterion (
Chen & West, 2008). RMSEA is a measure of discrepancy between the observed and model implied covariance matrices per degree of freedom. Based on
Browne and Cudeck (1993), RMSEA values of 0.05 or less indicate good fit, and values of 0.08 or less indicate adequate fit. SRMR is a measure of the average of the standardized fitted residuals (
Hu & Bentler, 1999). A value of less than 0.08 indicates a good fit (range = 0.00–1.00). CFI is derived from the comparison of a restricted model (one in which a structure is imposed on the data) with a null model (one in which each observed variable represents a factor;
Bentler, 1990). The CFI provides a measure of complete covariation in the data; a value of larger than 0.90 indicates adequate fit to the data.
CFAs confirmed a two-factor structure of the Attribution of Rumination scale. Items loaded on one factor were negative items, involving doubt over their ability (e.g., “The student is doubting if he or she has the capability needed for the class”), helplessness (e.g., “The student feels helpless”), and lack of change in the future (e.g., “The student thinks he or she will not be able to get a better grade”). Items loaded on to the second factor were positive items, focusing on the motivation to improve (e.g., “The student wants to improve his or her grades”) and grow from their failure (e.g., “The student wants to learn from his or her mistakes”). Following such loading pattern, we labeled each factor as self-doubt attribution and self-improvement attribution, respectively (fit statistics:
Table 2; factor loadings and residuals of each item:
Figure 1). We also directly compared our CFA model (i.e., two-factor model) with an alternative model (i.e., single-factor model) to test whether our theorized two-factor model is a better fit for the data compared with a single-factor model. We used the Δχ
2 (chi-square change), which directly compares the fit of the two models after adjusting for differences in the degrees of freedom. Results show that the Δχ
2 was significant (
p < .001), suggesting the superiority of the hypothesized two-factor model over the one-factor model.
Measurement invariance
In addition, to examine whether the developed scale measures the same construct across European Americans and East Asian descents (i.e., measurement invariance), further analyses were needed. Using the sample of European Americans and East Asian descents, multiple-group confirmatory factor analysis (MGCFA) using maximum likelihood robust estimation (
Rosseel, 2012) was conducted to test measurement invariance. Measurement invariance across cultures was tested at different levels (
Putnick & Bornstein, 2016): configural invariance (i.e., each group has the same factor structure, but loadings, intercepts, and residual variance can vary), metric invariance (i.e., loadings are fixed to be equal across groups), and scalar invariance (i.e., loadings and intercepts are fixed to be equal across groups). We first determined whether the model for configural invariance had adequate fit. Once that model was supported, we further tested measurement invariance. Specific standards to determine model fit followed suggestions from
Putnick and Bornstein (2016). We used CFI as the main criterion and supplemented with RMSEA or SRMR. When testing for metric invariance, the cutoff point used for CFI was −0.020, RMSEA was 0.015, and SRMR was 0.030, and when testing for scalar invariance, the cutoff point used for CFI was −0.010, RMSEA was 0.015, and SRMR was 0.010.
The results from MGCFA are presented in
Table 3. First, the model for configural invariance showed adequate fit: RMSEA = 0.091, SRMR = 0.059, and CFI = 0.909. The model for metric invariance showed adequate fit as well (RMSEA = 0.093, SRMR = 0.061, and CFI = 0.908), and based on the measurement invariance criterion, metric invariance was supported, suggesting that the factor loadings were equal across the two groups. We further tested for scalar invariance (i.e., constrained loadings and intercepts to be equal across groups), but the model was not supported. Inspection of the modification indices suggested that freeing the constraints for two items (Items 5 and 6) would improve the fit of the model. After relaxing the equality constraints of these intercepts, the model showed adequate fit, RMSEA = 0.091, SRMR = 0.062, and CFI = 0.905,
3 and passed the invariance cutoff criterion. Overall, these findings provide strong support for the two-factor structure of the scale and also show that the scale is compatible across the two cultural groups of interest.
Cultural differences in attribution of rumination
After the scale was validated and considered compatible across our two groups of interest (i.e., European American and East Asian descent), we further tested our predictions by conducting a linear mixed-effects model with type of attribution (self-improvement vs. self-doubt) as a within-subject variable and culture (European American vs. East Asian descent) as a between-subject variable, while controlling for age and gender. While the culture by attribution type interaction was not significant,
b = 0.18,
SE = 0.10,
t(1324.54) = 1.84,
p = .066, post hoc analyses show that, as predicted, European Americans scored higher on self-doubt attribution (
M = 4.94,
SD = 1.12) compared with East Asian descents (
M = 4.67,
SD = 1.23),
b = −0.23,
SE = 0.05,
t(1330.23) = −2.98,
p = .003, Δ
R2 = .005 (
Figure 2). However, there was no cultural difference in self-improvement attribution (European Americans:
M = 5.64,
SD = 0.94; East Asian descent:
M = 5.59,
SD = 0.99),
b = −0.05,
t(1332.43) = −0.78,
p = .436.
Discussion
Study 1 validated the Attribution of Rumination scale, which measures attributions for rumination within a specific context (i.e., after an exam). Results also provided evidence for measurement equivalence of the scale across European Americans and East Asians. The data supported our hypothesis that European Americans would be more likely to attribute rumination to doubt than East Asian descents. Results did not, however, support the prediction that East Asian descents would be more likely to attribute rumination to self-improvement reasons than European Americans. The lack of difference in self-improvement attribution indicates that, across cultures, people perceive that the motivation to do better contributes to rumination. Rather, cultural difference was confined to attributing rumination to doubt in one’s ability to progress. Although speculative, this may be because a belief that rumination is a coping mechanism (
Papageorgiou & Wells, 2001b) may contribute and override the cultural differences in self-improvement attribution. We will further discuss this in the “General Discussion” section. Based on our findings, we focused on self-doubt attribution as the focal mediator, with additional analyses using self-improvement attribution, in Study 3.
While Study 1 supported our prediction that attributions for rumination differ by culture, it is unclear whether it is related to other constructs relevant to perception of change (i.e., dialecticism, growth mindset). Thus, an online survey was conducted to test the association of these constructs with the Attribution of Rumination scale.
General Discussion
Through these studies, we present a novel approach to understanding the factors underlying cultural differences in the association between rumination and outcomes. We developed a scale to capture our proposed construct and found cultural differences in attributing rumination to self-doubt. Furthermore, the developed scale was related, but not redundant, with associated constructs (i.e., growth mindset, dialecticism; Study 2). In Study 3, we not only provided evidence supporting previous cross-cultural findings in the link between rumination and depressive symptoms (
Chang et al., 2010;
Kwon et al., 2013) but also further identified a mechanism that partly underlies such cultural differences. Finally, Study 4 provided supporting evidence for acculturation underlying the findings of Study 3.
While the primary aim of developing the scale was to provide a mechanism to explain cultural differences, we also believe that this scale can be used to explain within-culture variance in the link between rumination and outcomes. In fact, there have been contradicting findings in the effects of rumination. Several researchers have demonstrated that ruminating about and making meaning out of negative experiences has been generally considered helpful (e.g.,
Greenberg, 2005;
Rachman, 1980), while others have suggested that such “dwelling in the past” is associated with a range of negative outcomes, such as depression (e.g.,
Nolen-Hoeksema et al., 2008). The effort to understand such variance has generally been focused on how people engage in rumination (e.g.,
Kross & Ayduk, 2011) or what type of rumination people engage in (e.g.,
Ciarocco et al., 2010). Our approach takes a different perspective, focusing on how people perceive and attribute the act of rumination, therefore contributing to the existing literature on rumination.
We predicted cultural differences in people’s attribution of rumination to both self-improvement and self-doubt, but only found cultural differences in the latter. We speculate that other beliefs or mechanisms may contribute and override cultural differences in self-improvement attribution. For example, studies done in Western cultures found that patients with recurrent major depression believe that rumination is a helpful coping mechanism for them to solve problems, gain insight, and prevent future mistakes and failures (
Papageorgiou & Wells, 2001a,
2001b). It is possible that even non-depressed individuals within Western cultures recognize such reasons for people to ruminate, which could have led to a relatively high attribution of rumination to self-improvement even in Western cultures. In addition, the findings of Study 4 point to the possibility that both American and Asian cultural values may contribute to self-improvement attribution potentially via different routes. Future research needs to entangle such potential factors underlying self-improvement attribution.
We would like to clarify that we do not claim that one style of thought is better or more adaptive than the other. In fact, it is possible that there are contexts where East Asians’ way of cognitive processes could be associated with worse outcomes compared with that of Westerners, such as social sharing of negative experiences (
Kim et al., 2008). In fact, in Study 3, East Asian descents showed higher depressive and anxious symptoms compared with European Americans. Such findings suggest that while rumination may not be as disadvantageous for Asians, there must be other maladaptive factors among East Asian descents that are contributing to their higher depressive and anxious symptoms. Future research, therefore, is needed to examine other potential factors that may account for such negative outcomes.
Despite our theoretical assumption that dialectical thinking contributes to perception of positive changes following a negative experience (e.g., potential future improvement after failure in exam), it is yet unclear how optimism may also play a role in the cultural variation in attribution to rumination. Previous studies on cultural differences in the level of optimism have provided some mixed findings. Whereas several studies found higher optimism among European Americans than among Asian Americans (e.g.,
Chang, 1996), there is also evidence showing higher optimism in response to severe acute respiratory syndrome (SARS) outbreaks among Chinese than among Canadians (
Ji et al., 2004; for similar results in the context of the recent COVID-19 outbreak, see
Ji et al., 2021). Thus, it is possible that Asians may show optimism in the context of a specific negative event. Furthermore, the role of optimism in East Asian cultures relative to American culture has been found to be relatively complex (
Chang, 1996;
Ji et al., 2004;
Peng & Nisbett, 1999). It would be fruitful for future research to examine the role of optimism in attribution to rumination and in the links to psychological adjustments across cultures.
Furthermore, our main analyses did not differentiate between the two subscales of rumination, namely, Brooding and Reflective Pondering, because our aim was to focus on ruminative thinking as a whole as our first step in examining the role of attribution of rumination. At the same time, we conducted follow-up exploratory analyses in Study 3 (see the
Supplemental Materials) and found similar patterns across two subscales (i.e., both Culture and Self-Doubt Attribution moderated the association between each subscale of rumination and depressive symptoms in separate analyses) though the mediation effect for reflective pondering was weak. Such patterns are in line with previous findings that found cultural differences in the association between subscales of rumination, that is, Brooding (
Grossmann & Kross, 2010) and Reflective Pondering (
Kwon et al., 2013), and depressive symptoms. Further examination of different types of rumination will benefit better understanding of the role of culture and attribution of rumination in the associations between rumination and depressive symptoms.
A potential limitation of the developed scale is that the scenario is specific and most relevant to students in the academic context (i.e., “After the exam, . . .”). Thus, whether this scale will show similar patterns across populations other than students and different types of negative experiences (e.g., breaking up with a partner, losing a job) is unclear. It is possible that cultural differences in perception of change can be similarly applied to varying experiences; ruminating about the loss of a loved one, for instance, could still be attributed to doubting one’s ability to move beyond the past negative experience. In addition, because the current scale of attribution of rumination used a third-person perspective, one may wonder to what extent it corresponds to self-attribution. Of note is that when placed under a situation presented in the given scenario (i.e., failure after difficult exam), not all individuals may ruminate to begin with, which makes it hard to assess how individuals attribute their own act of rumination without confounding it with the frequency of rumination. It would be fruitful for future research to develop ways to assess self-attribution of rumination and examine its association with the current scale of attribution of rumination.
Furthermore, our current study focused on attribution and did not examine how the construct is related to other constructs that focus on the content of negative thoughts (e.g.,
Ciarocco et al., 2010;
Kross & Ayduk, 2011). It would be fruitful for future research to explore how attribution of rumination, particularly self-doubt attribution, could be related to the content and the type of rumination people engage in. For example, people who attribute rumination to self-doubt may tend to recount the concrete details of the experience (i.e., self-immersion;
Kross & Ayduk, 2011) and to focus on future impacts of their failure (i.e., state-focused rumination;
Ciarocco et al., 2010) when they ruminate about a past negative event.
Another limitation is in the participant sample. Our recruitment of East Asian participants is limited to individuals currently living within the United States as international students or U.S. citizens. By recruiting and comparing East Asians currently living in their own countries may provide a clearer picture of the cultural difference. Another limitation is that the present research is a cross-sectional correlational design. Although we ruled out theory of mind in Study 4, it is hard to rule out all other potential factors playing a role in the association between rumination and negative outcomes. One possible factor is uncontrollability of negative thought, a characteristic of rumination that has been linked to depressive symptoms (
Raes & Williams, 2010). Thus, further research is necessary to examine how the scale is related to such uncontrollability, particularly in relation to depression. In addition, our theoretical and experimental approach is based on the general comparison of two commonly compared groups in literature: European Americans and East Asians. While we are proposing attribution of rumination as a mechanism to explain cultural variation in the association between rumination and outcomes, it is unclear whether there are other cultural factors behind such observation and how well the current findings generalize to people from other cultures. For example, self-distancing has been proposed to underlie cultural differences in the association between rumination and outcomes between Russians and Americans (
Grossmann & Kross, 2010). Therefore, further investigation is needed to examine whether other cultural factors, such as self-distancing, might also underlie cultural differences between East Asians and European Americans and also to explore whether attribution of rumination plays a role among different cultural groups, such as Russians.
Notwithstanding these limitations, the present work provides initial evidence in cultural differences in attributing rumination to self-doubt, and that such cultural differences partially explain cultural variation in the association between rumination and depressive symptoms. As such, our findings work as a starting point to help understand cultural differences as well as individual differences in the magnitude of the association between rumination and depressive symptoms.