norms for experiencing emotions in different … relations and group processes norms for...

INTERPERSONAL RELATIONS AND GROUP PROCESSES

Norms for Experiencing Emotions in Different Cultures:Inter- and Intranational Differences

Michael EidUniversity of Magdeburg

Ed DienerUniversity of Illinois at Urbana-Champaign

Within- and between-nations differences in norms for experiencing emotions were analyzed in across-cultural study with 1,846 respondents from 2 individualistic (United States, Australia) and 2collectivistic (China, Taiwan) countries. A multigroup latent class analysis revealed that there were bothuniversal and culture-specific types of norms for experiencing emotions. Moreover, strong intranationalvariability in norms for affect could be detected, particularly for collectivistic nations. Unexpectedly,individualistic nations were most uniform in norms, particularly with regard to pleasant affect. Individ-ualistic and collectivistic nations differed most strongly in norms for self-reflective emotions (e.g., prideand guilt). Norms for emotions were related to emotional experiences within nations. Furthermore, therewere strong national differences in reported emotional experiences, even when norms were held constant.

The cross-cultural perspective has a long tradition in research onemotions. More than 100 years ago, Charles Darwin (1872/1970)based the theoretical considerations in his book The Expression ofthe Emotions in Man and Animals to a large degree on reports hehad received from people living in different cultures of the world.Since that time, cross-cultural comparisons have become the mostimportant ethological research strategy for proving the assumptionthat there are universal, biologically determined programs, partic-ularly for the expression of emotions. Reviewing the results of thisethological research paradigm, Grammer and Eibl-Eibesfeldt(1993) came to the conclusion that there is a universal humansystem not only for producing emotions but also for understandingthe expression of emotions.

From the perspective of cross-cultural anthropology and psy-chology, however, the biological perspective is not sufficient for afull understanding of the experience and expression of emotionsand must be complemented by consideration of the cultural contextin which emotions are experienced, expressed, and perceived (e.g.,Kitayama & Markus, 1995; Markus & Kitayama, 1991; Matsu-moto, 1996; Mesquita & Ellsworth, 2001; Mesquita & Frijda,1992; Scherer, 1994, 1997; Scherer & Wallbott, 1994; Scherer,Wallbott, & Summerfield, 1986; Shaver, Wu, & Schwartz, 1992;Wallbott & Scherer, 1988). If we follow Triandis's (1997) defini-

Michael Eid, Institute for Psychology, University of Magdeburg, Mag-deburg, Germany; Ed Diener, Department of Psychology, University ofIllinois at Urbana-Champaign.

Correspondence concerning this article should be addressed to MichaelEid, who is now at Faehbereich Psychologie, Universitat Koblenz-Landau,Im Fort 7, 76289 Landau, Germany. Electronic mail may be sent [email protected].

tion of a cultural syndrome as a "shared set of beliefs, attitudes,norms, values, and behavior organized around a central theme andfound among speakers of one language, in one time period, and inone geographic region" (p. 443), it is necessary to identify culturalsyndromes that are related to the experience, expression, regula-tion, and socialization of emotions.

According to Frijda and Mesquita (1995), cultural influences onthe emotion process are mediated to a large extent by the signif-icance an emotion has for an individual. Frijda and Mesquitadistinguished among three aspects of emotion that are culturallyinfluenced. First, they considered social consequences of emotionsthat regulate the expression and suppression of emotions. Second,they stressed the importance of norms for experiencing differentemotions. Third, they discussed social-cohesive functions of emo-tions. Several ethnographic studies have shown that there arestrong cultural differences in the social consequences of emotions,particularly in how the expression of emotions is valued. Theexpression of anger, for example, is strongly disapproved of by theUtku Eskimos (Briggs, 1970), whereas the Kaluli are expected andeven encouraged to show their anger (Schieffelin, 1983; for furtherexamples, see Mesquita & Frijda, 1992).

There are strong cultural differences in display rules (Ekman &Friesen, 1969; Izard, 1980) that are learned during the socializationprocess (Saarni, 1999). There are also cultural differences in thenorms for experiencing different emotions. Hochschild (1983), forexample, discussed the role of feeling rules, social norms thatprescribe how people should feel in specific situations (e.g., on awedding day, at a funeral). In addition to situational feeling rules,norms for the experience of emotions can also be present in asociety in the form of generalized expectations. This means thatemotions can differ in their desirability and perceived appropri-

Journal of Personality and Social Psychology, 2001. Vol. 81, No. 5, 869-885Copyright 2001 by the American Psychological Association, Inc. 0022-3514/01/S5.00 DOI: 10.1037//0022-3514.81.5.869

869

870 EDO AND DIENER

ateness across situations. The emotional climate of nations can becharacterized by generalized norms for experiencing differentemotions and the fact that these norms are subject to historicalchange (Stearns, 1994; Steams & Lewis, 1998).

Most of the research on cultural differences in emotions hasfocused on the first aspect, the social consequences of emotionmanifestations and how they regulate the expression versus sup-pression of emotions. Mesquita and Frijda (1992), however, re-viewed evidence that there are cultural similarities and differencesin all components of the emotion process (i.e., antecedent events,event coding, appraisal, physiological reaction patterns, actionreadiness, emotional behavior, and regulation). In explaining cul-tural differences in these components, not only are the norms forthe expression of emotions relevant but also the norms for theexperience of emotions. For example, people who value positiveemotions might be more alert to positive events, might seeksituations that provoke positive emotions, might appraise positiveevents in a more positive way, might stay in positive situationslonger, and might try to maintain their positive feelings or evenenhance them. However, people who think that positive feelingsare inappropriate are likely to avoid situations that cause positivefeelings, might not be able to appraise positive situations in apositive way, might withdraw from positive situations much ear-lier, and might even try to dampen positive feelings. On the otherhand, people who think that negative emotions (e.g., anger) areappropriate might seek anger-provoking situations when they as-sume that these situations would be helpful (e.g., for clarifyingconflicts, for asserting their goals). These people might be moreprone to appraise situations with respect to the situation's potentialfor hindering their goal, they might not withdraw from angersituations, and they might try to maintain their anger as long asthey think this would be necessary for asserting their goals. How-ever, people who believe that the experience of anger is inappro-priate might avoid anger-provoking situations. They might notfocus on components of a situation that hinder a personal goalwhen appraising a situation, and they might try to withdraw froman anger situation early and try to suppress or dampen their angerreaction. Hence, norms for the experience of emotions might havea strong influence on the regulation of one's emotions.

Moreover, values and norms for experiencing emotions mightbe important not only for an individual emotional episode but alsofor regulating emotions in other people. People who think that theexperience of positive emotions (e.g., pride) is inappropriate arelikely to disapprove of those emotional experiences in other peo-ple. This might be particularly important for the socializationprocess (Saarni, 1999). Parents who think that specific emotionsare inappropriate are likely to raise their children in such a waythat the children learn to avoid specific emotional situations andemotional reactions, to develop a specific appraisal style, and toregulate their emotions in such a way that they maintain a "cor-rect" emotional life. Parents who think that specific emotions arepositive are likely to encourage their children to feel and expressthese emotions and to reward emotional behavior that is in linewith their own emotional norms. Saarni (1999) pointed out thatparents are expected to socialize their children to behave and feelaccording to normative beliefs about desirable and appropriateemotional behavior. There are several ethnographic studies that arein accord with these assumptions (for an overview, see Ulich &Mayring, 1992). Finally, knowledge of norms for experiencing

emotions might be very important for cross-cultural communica-tion and the relationships of people in different cultures.

We owe much of our knowledge of cultural influences onemotions to ethnographic and anthropological studies. These stud-ies have produced interesting insights into cultural specificities,but they are limited in several ways. First, they are single-casestudies that focus on one culture that is described in detail. Thesestudies do not discuss whether there are cross-cultural differencesin emotional norms and how these cross-cultural differences areexplained. Second, these studies often refer to small groups, andthe generalizability of the results might be questionable. Third,they focus on aspects of emotion that are typical for a culture. Howstrong within-culture differences are compared with cross-culturaldifferences is not analyzed.

If we take a cross-cultural perspective, at least two questions areimportant: (a) Which cultures should differ in values and norms forexperiencing emotions (according to theoretical models), andwhy? and (b) Can cross-cultural differences in values and normsbe empirically confirmed (according to empirical studies)?

Values and Norms for Experiencing Emotions

Cultural differences in values and norms for experiencing emo-tions can be predicted from cultural differences in self-construals.In cross-cultural psychology, two prototypical self-construals, theindependent and the interdependent self-construal, have been dis-tinguished by several authors (e.g., Gardner, Gabriel, & Lee, 1999;Markus & Kitayama, 1991; Triandis, 1989). The independent selfconsists of a configuration of inner attributes (e.g., dispositions,motives, and values) that make an individual unique from others.In cultures where an independent self is predominant, people areexpected to become independent from others and to pursue andassert individual goals. People with an individual self focus ontheir own attributes, abilities, and preferences and tend to expressthese attributes in public and in private. The personality patternthat characterizes the independent self is also named idiocentrism(Triandis, 1997). Cultures in which idiocentrism is the predomi-nant personality pattern are called individualistic cultures (Trian-dis, 1997). Hence, people with an independent self are found inindividualistic (e.g., Western) cultures more often than in collec-tivistic ones (e.g., Eastern cultures).

The interdependent (relational) self, on the other hand, is char-acterized by the belief that the self cannot be separated from othersor from the social context. The self is part of all-embracing socialrelationships, and people with an interdependent self focus on andare regulated by the emotions, thoughts, and actions of otherpeople. The personality pattern characterizing the interdependentself is also called allocentrism (Triandis, 1997). In collectivisticcultures, in which this construal of the self is predominant, thesocial norm is to maintain harmony with others, to meet socialobligations, and to support the goals of others who are in a socialrelationship with oneself. Thus, in contrast to individualistic cul-tures, the norm is not to become independent from others but tofulfill one's social duties.

Lee, Aaker, and Gardner (2000) showed that differences inself-construals are accompanied by differences in regulatory focus(approach vs. avoidance). Lee et al. (2000) defined regulatoryfocus as "the extension of the basic hedonic principle of approachand avoidance to allow for distinct self-regulatory strategies and

NORMS FOR EXPERIENCING EMOTIONS 871

needs" (p. 1122). People with independent self-construals arepromotion focused. They focus more strongly on information thatis relevant for approaching their own aspirations and wishes.Moreover, they place more emphasis on positive than on negativeinformation regarding themselves. People with an interdependentself, on the other hand, are more prevention focused. They placemore attention on information that prevents them from violatingsocial norms. Furthermore, they emphasize negative informationregarding themselves.

Differences in self-construal and regulatory focus might bestrongly related to norms for experiencing emotions. In particular,it can be expected that there are strong cultural differences inself-conscious or self-reflective emotions—emotions that reflecton the individual's own doing (Tangney & Fischer, 1995) and thatare, therefore, important for self-regulation. Self-conscious emo-tions that indicate that personal goals have been successfullyapproached might be more important and desirable for people withindependent self-construals, who are more promotion focused.Therefore, the self-conscious emotions that arise from succeedingby one's own efforts (e.g., pride) are highly valued in individual-istic societies. However, self-conscious emotions that indicate thatone's controllable actions are wrong or insufficient (negative in-formation regarding oneself) might be more important and desir-able for people with interdependent self-construals, who are moreprevention focused. Consequently, the self-conscious emotionsthat are caused by violating social norms and by failing to fulfillsocial obligations (e.g., guilt) are likely to be more highly valuedin collectivistic societies. Cross-cultural differences in norms foremotions might be less important for other emotions that are notself-conscious in nature and that arise more from external causesthan from internal, controllable sources (e.g., anger, worry, joy).

There have been few empirical cross-cultural studies on normsfor experiencing emotions. Most cross-cultural studies on emo-tions refer to cross-cultural differences in the frequency, intensity,or expression of emotions (see, e.g., Kitayama, Markus, & Matsu-moto, 1995; Stipek, 1998; Stipek, Weiner, & Li, 1989). Thesestudies allow only indirect and, therefore, limited conclusionsabout how emotions are generally valued in these cultures. Forexample, people might think that positive emotions are very de-sirable. If they are not able to seek positive situations or to react topositive events with positive emotions, however, they will not feelpositive emotions frequently or intensely. Thus, norms and feel-ings are distinct components of the emotion process, and, conse-quently, variables characterizing the feeling of emotions are onlyindirect indicators of norms and attitudes. In the following, theresults of some previous studies on norms for affect aresummarized.

Stipek (1998) compared the value of pride between Chinese andAmericans. She found that for the Chinese, pride is more accept-able for achievements that benefit others than for achievementsthat are due to personal accomplishments. Sommers (1984) ex-plored the values of different emotions by asking six questions, butonly with small samples. Sommers found that there were strongcross-cultural similarities with respect to the emotions of love,happiness, and joy, which were considered desirable in all cultures.Also, hate, terror, and rage were considered as dangerous anddestructive in all cultures. Furthermore, guilt, frustration, fear,shame, and embarrassment were consistently considered aversive.Cross-cultural differences were shown with respect to specific

emotions. The Americans valued enthusiasm very strongly, theGreeks highly valued respect, and the West Indians valued pride.The Chinese considered more negative emotions to be useful andconstructive than did the three other nations.

Although they give valuable insights into the norms for emo-tions, previous studies are limited in two ways. First, they haveoften focused either on a limited number of emotions and nationsor on small samples. Second, differences between cultures havetypically been analyzed by comparing mean values (or otherstatistics) between the cultures without testing whether the as-sumption of measurement equivalence across cultures was ful-filled. Comparing means, for example, presumes that individualsuse the scales in the same way (assumption of measurementinvariance). This means that there are not individual responsestyles such as the preference for one response category (e.g., themiddle one) or the avoidance of response categories (e.g., theextreme ones). If individual response styles are present, meanvalues can only be compared if the response styles are equallydistributed in the different cultures. The absence of interindividualdifferences in response styles can be statistically tested in theframework of item response theory by demonstrating that the itemparameters do not differ between individuals or subgroups ofindividuals (e.g., Drasgow & Kanfer, 1985). The assumption thatall individuals use the scales in the same way, however, can bequestioned with regard to two issues: (a) Several analyses thathave been undertaken during the past years have shown that thisassumption is often not fulfilled even within individual cultures(e.g., Eid & Rauber, 2000; Rost, Carstensen, & von Davier, 1997).(b) Moreover, as Leung and Bond (1989) pointed out, it is possiblethat different response styles exist between the varying cultures(see also Bond, 1996). Consequently, the measurement invarianceassumption must be tested within each culture as well as betweendifferent cultures to ensure that these mean differences are not ofan artificial nature. Previous studies on norms for emotions, how-ever, have not tested whether the assumption of measurementinvariance holds. One aim of the present study is to test whetherthe same norm structure can be found in different cultures (theproblem of structural equivalence in cross-cultural psychology;van de Vijver & Leung, 1996, 1997).

In cross-cultural psychology, several methods have been appliedto test the assumptions of measurement and structural equivalence(for an overview, see van de Vijver & Leung, 1996, 1997). Mosttypically, these assumptions have been tested by dimensionalmodels such as confirmatory factor analysis or latent trait modelsof item response theory. Dimensional models, however, are notappropriate for testing the measurement equivalence of norms foremotions in the present research context. Dimensional modelsassume that all individuals and items can be ordered on a contin-uum. This assumption is very strong if we consider norms foremotions. It is very reasonable that, within cultures, individualsdiffer in the types of emotions that they consider desirable andundesirable. For example, there might be people who think thatanger is desirable but guilt is not, and there might be people forwhom the opposite is true. These differences in norm patterns canbe adequately assessed in a typological model. A typologicalmodel is also important from an emotion regulatory standpoint.Following Tangney, Wagner, Fletcher, and Gramzow's (1992)research, for instance, it can be hypothesized that people who rateanger as an unacceptable emotion and guilt as an acceptable one

872 EID AND DIENER

tend to ascribe negative events to themselves, whereas for peoplewith the opposite rating pattern, it is expected that they tend toblame other people for negative events. These individual profiles,however, are not visible if we compare means across cultures. Amean structure that, for example, implies that anger is not accept-able in a country yet guilt is acceptable does not indicate howuniform this difference is within cultures. Cross-cultural researchershave hitherto been frustrated by the fact that they want to examinecultural differences but recognize that very large variations alsoexist within cultures. To answer this challenge and interweavecultural and individual differences, we suggest the use of multi-group latent class analysis (e.g., Lazarsfeld & Henry, 1968; Mc-Cutcheon, 1987). This type of analysis is able to address issues ofscale equivalence as well as within-nation variability.

Latent class analysis is a procedure for categorical responsevariables and is based on four assumptions (e.g., Clogg, 1995; Eid,2001; Langeheine & Rost, 1988). It is first assumed that a popu-lation is not homogeneous; rather, it is composed of subpopula-tions. These subpopulations are called latent classes because theyare not directly observable. Each individual in the population canand must belong to only one subpopulation; that is, the classes aredisjunctive and exhaustive. The membership of a population is notknown a priori, and only the membership probabilities (assignmentprobabilities) and the class sizes can be estimated. Second, all ofthe people belonging to the same class are homogeneous withrespect to the class conditional response probabilities for thevarious categories of items. Third, the class structure holds for allitems analyzed. Fourth, local stochastic independence is assumed.This means that the responses on items are independent, givenlatent class membership (hence, local independence). This as-sumption implies that the class structure explains all associationsbetween the observable responses. In the framework of latent classanalysis, measurement invariance exists when the response prob-abilities of the latent classes do not differ between the variouscultures.

The typological combination of norms for different affects istheoretically more interesting than the simple comparison of meanvalues across cultures. Consequently, a typological approach thatconsiders different rating patterns within a culture and betweencultures is an appropriate starting point for the analysis of inter-and intracultural differences. If we compare typological structuresacross cultures, however, we have to ensure that the typologicalstructures (measurement model) are identical in different cultures(measurement invariance). Thus, we need a typological modelwhose cross-cultural generalizability (measurement invariance)can be statistically tested. Multigroup latent class analysis is anappropriate model for testing typological structures between cul-tures. Compared with other methodological approaches that aretraditionally applied in cross-cultural psychology, latent class anal-ysis has a further strong methodological advantage. In methodstraditionally applied in cross-cultural psychology, it is typicallyassumed that the parameters of a model (e.g., factor loadings) arethe same for all individuals of a culture. This means that a culturehas to be homogeneous and that heterogeneity is only allowed forindividuals belonging to different cultures. If the assumption ofmeasurement equivalence (e.g., assumption of equal factor load-ings) across cultures must be rejected, this means that individualsfrom different cultures cannot be compared, for instance, withrespect to a mean score. One strong advantage of latent class

analysis is that the assumption of homogeneous cultures is notmade. Moreover, there might be subgroups within one culture thatdiffer with respect to the parameters of a model. These subgroupsmight exist in different cultures, but there might be subgroups thatare so culture specific that they do not occur in other cultures.Thus, latent class analysis is able to detect universal types thatexist in all cultures and culture-specific types that exist only inspecific cultures. For example, there might be (universal) patternsof norms for emotions that can be found in all nations but alsopatterns of norms for emotions that are so specific that they can befound only in single nations. The assumption of measurementinvariance would then hold only for the universal types (sub-groups) that can be found in all cultures. Culture-specific classes,however, reveal structural differences between cultures and indi-cate individuals who are most typical for the uniqueness of aculture. For researchers interested in indigenous aspects of a cul-ture, these people might be of major interest. The capability oflatent class analysis to detect universal and culture-specific classesgoes far beyond other statistical models for cross-cultural compar-isons. Although multigroup latent class analysis appears to beperfectly suited for cross-cultural emotion psychology, as far as weknow this type of analysis has not been used in this field until now(for other applications of multigroup latent class analysis in soci-ology and political sciences, see McCutcheon, 1998; McCutcheon& Hagenaars, 1997; McCutcheon & Nawojczyk, 1995). In thefollowing, we show how latent class analysis can be used toanalyze and test cross-cultural differences in emotion norms in amuch stronger way than is possible by comparing mean values.

In addition to this methodological aim, we pursue four moresubstantive aims. The first aim is to analyze whether explicitlymeasured norms differ between individualistic and collectivisticcultures in the way predicted by the cultural differences in self-construals and regulatory focus that we outlined earlier. Differ-ences in the cultures should primarily be found for the self-evaluative emotion pride. Collectivistic countries should differnotably from individualistic countries in such a way that types(latent classes) that are characterized by the undesirability of prideoccur more frequently. Concerning the negative emotions, typesfor which guilt is undesirable should be found more frequently inthe individualistic nations (i.e., the United States and Australia)than in collectivistic ones. On the other hand, types with theopposite tendencies should be found more often in collectivisticcountries.

Our second aim is to analyze whether norms for emotions arerelated to the frequency and intensity of emotions and whetherthese relations can be consistently found in different cultures. Aswe outlined above, there might be only a weak relation betweennorms and the experience of feelings, but this question has notbeen sufficiently explored with empirical data.

Our third aim is to scrutinize whether there are cross-culturaldifferences in the intensity and frequency of reported emotionalexperiences if we compare people who have the same norms butbelong to different cultures. This is another as yet unexploredquestion. These analyses go beyond traditional studies, whichcompare the intensity and frequency of emotions between cultureswithout measuring norms. The intensity and frequency of emotionsbetween cultures might be partly or totally due to differences in thenorms for emotions. However, if we correct for differences in thenorms, we can compare individuals who are homogeneous with


respect to what emotions they consider desirable. If there arecross-cultural differences in the intensity and frequency of emo-tions between people who have the same norm, this might be moreindicative of cross-cultural differences in regulation abilities, ge-netic differences, or life circumstances that influence emotionalfeelings.

Method

Participants

We analyzed four subsamples of a large study on emotions in 41countries (Suh, Diener, Oishi, & Triandis, 1998). In the following, we referto the subsamples of college students from the United States (N = 443),Australia (N = 292), Taiwan (N = 553), and the People's Republic ofChina (N = 558). We selected these four countries on the basis of thefollowing criteria: First, the countries investigated should differ in theirindividualistic versus collectivistic orientation. According to Harry Trian-dis's judgments (personal communication, May 1, 1993), the countries canbe classified in the following way on a 10-point individualism scale, where10 = most individualistic and 1 = least individualistic: United States = 10points, Australia = 9 points, Taiwan = 5 points, and China = 2 points.Second, the subsamples must be sufficiently large to perform latent classanalyses. The design of the larger study is described in detail by Suh et al.(1998).

Materials

The participants were required to complete a questionnaire that con-tained items for assessing their life satisfaction, experience of emotions,and personality. Chinese and Taiwanese students completed a MandarinChinese version of the questionnaire. The English questionnaire was trans-lated into Chinese by a native Chinese-speaking bilingual scientist. Aretranslation was performed by a native Chinese-speaking bilingual scien-tist with a Ph.D. in psychology. The retranslation was compared with theoriginal version by three native English-speaking people and was rated ona 7-point scale (1 = totally incorrect, 7 = completely correct). Theretranslation was rated a 6 by two people and a 7 by one person.

One of the questions dealt with the norms for emotions. It read, "In thefollowing question we would like you to indicate how appropriate ordesirable it is to experience certain emotions. Please use the following scaleto give your answer to each emotion."1 The following emotions werepresented: joy, affection, pride, contentment, anger, fear, sadness, andguilt. The theoretical background for selecting these emotions was de-scribed by Diener, Smith, and Fuijta (1995). The statements could beanswered on 7-point scales with the following response categories: ex-tremely desirable and appropriate, desirable and appropriate, slightlydesirable and appropriate, neutral (neither desirable nor undesirable),slightly undesirable and inappropriate, undesirable and inappropriate, andextremely undesirable and inappropriate.

The frequency of emotions was assessed by the following instruction:"Using the scale below, indicate how often you feel each of the emotionslisted below. Put a number from 1 to 7 to accurately reflect how much ofthe time when you are awake you feel that emotion. How much of the timeduring the past month have you felt each emotion?" The same list ofemotions was given. The participants rated each emotion on a seven-category response scale ranging from never (1) to always (7).

The intensity of emotions was assessed by the instruction, "Now use thescale below to indicate the intensity of these emotions WHEN YOU DOFEEL THEM. That is, when you do experience this emotion, no matterhow rarely, typically how INTENSE is your emotional experience?" Theintensity of each emotion from the emotion list described above was ratedon a 7-point scale ranging from none—/ never experienced it (1) toextremely intense (7).

Procedure

In the first step, the items measuring norms for affect were analyzed bylatent class analysis. The multiculture (multigroup) latent class modelswere analyzed with the computer program PANMARK. (van de Pol,Langeheine, & de Jong, 1996). To reduce the cells of the multidimensionalcontingency table, we used two strategies: First, the three desirable and thethree undesirable categories of the norm items were reduced to one cate-gory each. Thus, response variables with three categories underlie thelatent class analyses. Second, the positive emotions of joy, affection, pride,and contentment were analyzed separately from the negative emotions ofanger, fear, sadness, and guilt. The positive and negative emotions wereanalyzed separately because of past evidence that the two are based onsystems that show some degree of independence (e.g., Cacioppo & Gard-ner, 1999).

To analyze the class structure, we first determined, separately for eachnation, the common minimal class number for all the participating nationsthat provided a satisfactory model fit. Then we conducted a multigrouplatent class analysis including all nations. We compared several solutionsusing four different goodness-of-fit criteria: the likelihood ratio test, thePearson chi-square test, the Cressie-Read test, and the Akaike informationcriterion (AIC; for a description of these statistics, see Read & Cressie,1988). Because of sparse tables, we applied the bootstrapping methodologyand used 300 bootstrapping analyses (see Langeheine, Pannekoek, & vande Pol, 1996). In the case of sparse tables, the p values of the test statisticsmight not be valid because the assumption that the test statistics aredistributed according to a chi-square distribution might be violated. Boot-strapping analysis cures this problem because the distributions of the teststatistics can be estimated, and therefore the estimation of the p values ismore valid (Langeheine et al., 1996). We report the original p values thatwere estimated on the basis of the assumption that the test statistics aredistributed according to a chi-square distribution. In addition, we alsoreport the bootstrapping p values. However, we interpret only the boot-strapping p values because we have sparse tables. To test the assumptionof measurement invariance, we restricted all response probabilities to beequal across nations. We evaluated the fit of the restricted models by fourcriteria. First, the models should not have been rejected by any of the threetests (likelihood ratio test, Pearson chi-square test, Cressie-Read test) usinga bootstrapping alpha of .05. Second, we considered the likelihood ratiodifference test. This means that we calculated the differences between thevalues of the likelihood ratio test of the unrestricted models and the valuesof the likelihood ratio test of the models with measurement invariance.Because we have sparse tables we applied Holt and Macready's (1989)recommendation and considered likelihood ratio difference tests as signif-icant if the probability value was smaller than .01. Because PANMARKdoes not provide bootstrapping results for likelihood ratio difference tests,we used traditional p values for these tests. Third, we made the decisioneven more rigorous by comparing the restricted solutions with the unre-stricted ones. If there was any hint that relaxing the strong assumption oftotal measurement invariance would lead to a more appropriate model, weanalyzed a model with relaxed measurement invariance assumptions (par-tial measurement invariance) and tested this model against the unrestrictedone. Fourth, we compared the AIC values of all models considered andchose the model with the lowest AIC value. Hence, we used a veryconservative test strategy to ensure that the models we selected represented

1 One reviewer argued that "appropriate" and "desirable" might not havethe same meaning. In a recently conducted German study (Mohiyeddini &Eid, 2001), we assessed norms for emotions by separate scales, includingan appropriate item and a desirable item. We found that the associationsbetween the appropriateness judgments and the desirability judgmentswere very high for the different emotions (median of the coefficients ofcontingency for the two items = .77), showing that respondents do notdifferentiate much between the concepts.

874 EID AND DEENER

true cross-cultural consistencies and variations and that we did not over-look substantive cultural differences.

To analyze whether and how the latent classes of respondents differ intheir frequency and intensity of experienced emotions, we applied thefollowing strategy. First, all participants were assigned to the latent normclasses for which their assignment probabilities were maximum. BecausePANMARK does not calculate the individual assignment probabilities, weused the computer program LEM (Vermunt, 1993) for these analyses.Then, differences between classes and differences between nations wereanalyzed by analysis of variance and the Kruskal-Wallis test, because theassumption of equal variances between cells was violated in some cases(see below).

Results

Latent Class Analyses of the Norm Items

The goodness-of-fit coefficients for the different latent classmodels are given in Table 1. We first consider the unrestrictedanalyses. The latent class analyses revealed that a five-class struc-ture for positive emotions and a six-class structure for negativeemotions showed a good model fit for all nations. In these analy-ses, no restrictions about the nations were added, which means thatthe response probabilities can differ between nations. For bothpositive and negative emotions, a model with perfect measurementinvariance (equal response probabilities between all nations)shows appropriate fit coefficients for the positive emotions but isinappropriate for the negative emotions. A detailed inspection ofthe results revealed that the perfect measurement invariance as-sumption for all nations might be too strong. Rather, a model inwhich the measurement invariance hypothesis in one class of theChinese population was removed was superior to a model withperfect measurement invariance but not significantly worse thanthe unrestricted model according to the criteria described in theMethod section. Because this was also true for positive emotions,we accepted a solution with partial measurement invariance forboth types of affect. The model with partial measurement invari-ance assumes that there is perfect measurement invariance be-tween Australia, Taiwan, and the United States. Furthermore, allclasses in the Chinese population, with the exception of one class,showed measurement invariance with the three other nations.However, there was one class for both types of affect that wasspecific to China.

Positive emotions. Before the results of the positive emotionsare presented, we would like to reiterate how our theoreticalconsiderations about cultural differences in self-evaluative emo-tions are related to our data set. We expected that differences in thecultures should primarily be found for the self-evaluative emotionpride. The collectivistic country, China, should differ notably fromboth the individualistic countries, the United States and Australia,in such a way that types (latent classes) that are characterized bythe undesirability of pride occur more frequently. The prevalencerate of these types in Taiwan should lie between the rate for Chinaand both the other countries, because with regard to theindividualism-collectivism variable, Taiwan can be classified be-tween China and the other two countries.

The latent class solution for the positive emotions is illustratedin Table 2. The rows refer to the five different classes. The firstfour classes are based on the measurement invariance hypothesisfor all the countries, Class 5a is invariant for the United States,Australia, and Taiwan, and Class 5b is specific to China. In the lastrow, the concentration coefficient is given. We come to thiscoefficient later. The first four columns contain the responseprobabilities of the three categories for the four emotions: desir-able and appropriate, neutral, and undesirable and inappropriate.For example, the response probabilities for joy in Class 1 indicatethat people belonging to this class think (with a probability of 1.00)that joy is desirable and appropriate. For Class 2, the probability ismaximum (.74) for the category neutral and smaller for the cate-gories desirable/appropriate (.19) and undesirable/inappropriate(.09). Hence, people belonging to this class can be characterized aspeople with a neutral attitude to joy. The response probabilities forthe other emotions and classes can be interpreted analogously. Thelast four columns indicate the relative class sizes in the fournations. A class size of .00 means that the class does not exist ina nation as a result of the empirical analysis. The empty cell inRow 5a indicates that the measurement invariance assumption hasbeen relaxed for China and that the probabilities in the Chinesesample were not restricted to be equal to the other nations. Theprobability of Class 5 in China is presented in a separate row tomake sure that Class 5 in China refers to a different responsepattern. Therefore, the empty cell in Row 5a indicates that the sizeof Class 5 in China can be found in another row. Conceptually, a

Table 1Goodness-of-Fit Coefficients for Different Latent Class Models

Model

No measurement invarianceMeasurement invariancePartial measurement invariance

No measurement invarianceMeasurement invariancePartial measurement invariance

df

144264256

108252244

Likelihood ratio

Value

102.75258.40219.65

145.59354.29321.50

P

1.00.59.95

.01< .01< .01

test

p(B)

Pearson chi-square

Value p

Positive emotions

1.00 102.78 1.00.27 314.45 .02.67 316.29 .01

Negative emotions

1.00 128.91 .08.05 331.19 < .01.32 298.94 .01

test

p{B)

1.00.36.60

1.00.10.35

Cressie-Read

Value

96.02271.24250.37

130.04327.73295.83

P

1.00.37.59

.07< .01

.01

test

KB)

1.00.31.61

1.00.07.32

Akaike informationcriterion

12,813.2812,728.9312,706.19

18,042.9217,963.6117,946.83

Note. p(B) represents the probability values calculated on the basis of 300 nonnaive bootstrapping analyses.


Table 2Norms for Positive Emotions: Conditional Response Probabilities,Class Sizes, and Concentration Coefficients

Class andresponsecategory

1D/ANUD/IA

2D/ANUD/IA

3D/ANUD/IA

4D/ANUD/IA

5aD/ANUD/IA

5bD/ANUD/IA

CON

Joy

1.00.00.00

.19

.74

.08

.81

.11

.09

.79

.10

.11

.96

.04

.01

.13

.24

.64

Conditional

Affection

1.00.00.00

.11

.89

.00

.77

.16

.08

.72

.17

.11

.92

.06

.02

.00

.24

.76

response probabilities

Pride Contentment

.92

.04

.04

.00

.83

.17

.55

.45

.00

.08

.00

.92

.41

.34

.25

.20

.16

.64

.99

.01

.00

.05

.90

.05

.54

.24

.22

.36

.16

.48

.90

.11

.00

.19

.27

.54

AU

.83

.00

.00

.03

.14

.29

Class

US

.83

.01

.14

.03

.00

.29

sizes

CH

.09

.09

.32

.34

.16

.74

TW

.32

.06

.00

.06

.57

.56

Note. Concentration (CON; Wickens, 1989) = 1 - sum of the squared probabilities. The minimum concen-tration in this analysis (five categories of the latent class variable) was zero, and the maximum concentration was.80. AU = Australia; US = the United States; CH = China; TW = Taiwan; D/A = desirable and appropriate;N = neutral; UD/IA = undesirable and inappropriate.

value of .00 and an empty cell mean the same thing, namely thatthe pattern does not exist in a nation.

How can the classes be characterized and how are they distrib-uted across the different nations? Class 1 consists of people whorated all positive emotions as desirable and appropriate because theprobability for the category desirable and appropriate is close to 1.Eighty-three percent of the Australian sample and 83% of theAmerican sample belong to this class, but only 9% of the Chinesesample and 32% of the Taiwanese sample do so. Class 2 consistsof people who rated all the emotions as neutral. This class is smallin all nations, indicating that people have a (positive or negative)attitude toward the pleasant emotions. The sizes of this class aresignificantly different from zero in China and Taiwan but trivial inthe two other nations. Class 3 is characterized by the high desir-ability of joy and affection. Pride and contentment are also to someextent desirable, yet people belonging to this class are ratherindifferent with respect to these emotions. Thirty-two percent ofthe Chinese sample and 14% of the American sample belong tothis class. For the other countries, this class is not of importance.Class 4 differs from Class 3 essentially in that the emotion of prideis explicitly rated as undesirable and inappropriate, whereas con-tentment tends to be rated rather indifferently. This class is mostrepresentative of the Chinese sample, because 34% of that samplecan be classified here. Class 5 differs in its response probabilitiesbetween China and the three other nations. Class 5a, which is not

found in China but is found in all other nations, is characterized bya high desirability of joy, affection, and contentment, whereaspride is rated rather indifferently. This class is prototypical forTaiwan, with more than half of the Taiwanese sample belonging tothis class. Class 5a looks very similar to Class 3 in profile, but thetwo classes differ in the numerical values of the response proba-bilities, in particular for the category desirable and appropriate.The Chinese-specific class, Class 5b, stands out because all of theassessed emotions are rated as undesirable and inappropriate, with16% of the Chinese sample thinking that positive emotions areundesirable.

How can these results be interpreted with respect to our theo-retical considerations? Concerning the positive emotions, pride isthe most relevant. Class 4 can be interpreted as a typical collec-tivistic pattern in which pride is considered undesirable. After all,one third of the Chinese sample belongs to this group, thussupporting the assumption that collectivistic cultures considerpride more undesirable than do individualistic ones. Class 5a alsopoints in this direction with its indifference with respect to pride.This class is typical for Taiwan, which is currently on its way frombeing a collectivistic country to being an individualistic country. Ingeneral, the analyses confirm that the classes are differentiatedprincipally with respect to pride.

In the last row in Table 2, the concentration coefficients aregiven for each nation. The concentration coefficients were calcu-

876 EED AND DEENER

lated on the basis of the class sizes (probabilities of the latentclasses). The concentration coefficient is a measure for the vari-ability of probabilities (see Wickens, 1989). It is zero if one latentclass has a probability of 1 and, consequently, all other classeshave a probability of zero. It is maximum if all latent classes havethe same probabilities. Hence, the larger the concentration is, thelarger is the within-nation heterogeneity. The concentration coef-ficients show that the heterogeneity is very large in China andrelatively small in Australia and the United States, with Taiwan inbetween. This result indicates that the individualistic nations ofAustralia and the United States are very homogeneous nations withrespect to norms for positive emotions, whereas the more collec-tivistic nations of China and Taiwan are rather heterogeneous.

Negative emotions. Concerning the negative emotions, we ex-pected that in the individualistic nations (the United States andAustralia), types of people for whom guilt is undesirable should befound more frequently than in China. On the other hand, types withthe opposite tendencies should be found more often in China. Therelative frequencies in Taiwan should once again lie between thoseof the individualistic countries and that of China. No predictionswere formulated with respect to the other emotions.

The results of the latent class analyses with respect to anger,fear, sadness, and guilt are shown in Table 3. Class 1 identifies a

group that rates all of the negative emotions as desirable andappropriate. This class is approximately the same size (with a classsize of about 22%) for all four nations. Class 2 is characterized bythe fact that all four emotions are considered undesirable andinappropriate. This class is the largest in the United States, Aus-tralia, and Taiwan, with class sizes between 35% and 44%. Thisclass, however, is comparatively small in China (14%). Thus, thereare rather few people in China who think that negative emotionsare generally undesirable and inappropriate. In Class 3, the modalresponse probability for all emotions is the neutral, middle cate-gory. This class is larger in both Asiatic countries than in theUnited States or Australia. In sum, the first three classes representdifferences in the level of desirability (i.e., desirable, neutral,undesirable) that are similar for all negative emotions. The remain-ing classes, however, portray typological differences between spe-cific emotions. In Class 4, guilt is rated as desirable and appropri-ate, whereas the modal value for anger is undesirable andinappropriate. Furthermore, sadness is considered desirable,whereas fear is undesirable. This class is comparably small in allcountries, albeit somewhat larger in China and Taiwan than in theother two countries. Class 5 is characterized by the contrast be-tween fear, which is desirable, and sadness, which is undesirable.It is relatively small in all the countries and is most common in

Table 3Norms for Negative Emotions: Conditional Response Probabilities,Class Sizes, and Concentration Coefficients

Class andresponsecategory

1D/ANUD/IA

2D/ANUD/IA

3D/ANUD/IA

4D/ANUD/IA

5D/ANUD/IA

6aD/ANUD/IA

6bD/ANUD/IA

CON

Anger

.82

.10

.08

.06

.08

.86

.11

.60

.29

.32

.19

.49

.49

.15

.36

.51

.20

.29

.45

.27

.28

Conditional

Fear

.89

.09

.01

.00

.16

.84

.12

.71

.16

.04

.34

.62

.88

.00

.12

.31

.26

.43

.01

.40

.59

response probabilities

Sadness

.94

.06

.00

.00

.11

.89

.00

.90

.10

.64

.35

.01

.05

.21

.74

.56

.12

.32

.31

.00

.69

Guilt

.74

.18

.08

.05

.09

.86

.03

.76

.21

.68

.32

.00

.38

.21

.42

.00

.08

.92

.25

.20

.55

AU

.24

.35

.03

.08

.03

.27

.74

Class

US

.22

.44

.06

.05

.09

.13

.73

sizes

CH

.22

.14

.12

.14

.15

.23

.82

TW

.21

.41

.17

.11

.04

.06

.74

Note. Concentration (CON; Wickens, 1989) = 1 - sura of the squared probabilities. The minimum concen-tration in this analysis (six categories of the latent class variable) was zero, and the maximum was .83. AU =Australia; US = the United States; CH = China; TW = Taiwan; D/A = desirable and appropriate; N = neutral;UD/IA = undesirable and inappropriate.


China. Finally, Classes 6a and 6b point to a typical pattern forindividualistic countries—that is, the undesirability of guilt. Thereare structural differences between China and the other three coun-tries: Mainly, guilt is less undesirable in China than in the othercountries, and sadness is somewhat undesirable in China yet mod-erately desirable in the other three countries. This class is seldomfound in Taiwan, whereas it is relatively common in the othercountries. In sum, there are strong national differences in thedesirability of guilt. In China, the classes in which the probabilityof the undesirable category for guilt is close to 1 (Classes 2 and 6a)are comparatively sparsely occupied (14%) or do not exist,whereas 62% of Australians, 57% of Americans, and 47% ofTaiwanese belong to these classes.

The concentration coefficients indicate that the national differ-ences in heterogeneity are smaller for negative affect than forpositive affect. The concentration coefficients are all rather large,showing that there is strong intranational heterogeneity with re-spect to norms for negative affect. Again, China is the nation withthe largest intranational heterogeneity, with a concentration coef-ficient that is close to the maximum possible value.

In conclusion, regarding the results for positive and negativeaffect norms, it can be said that the self-conscious emotions prideand guilt differentiate between the nations most clearly. We returnto this result in the Discussion.

Within-Nation Differences Between Classes

To analyze differences in the reported intensity and frequency ofexperienced emotions, we assigned all participants to the latentclasses for which their assignment probabilities were maximum.The mean classification (assignment) probabilities for the differentlatent classes are given in Table 4. The mean classification prob-abilities can be interpreted as reliability coefficients (maximumvalue = 1.00). High values indicate high reliability. The meanclassification probabilities in our study are rather large, showingthat a reliable assignment was possible. Next, mean differences inthe frequency and intensity of emotions between these classeswere analyzed by analysis of variance. Because the assumption ofequal variances between cells (tested by the Levene test) wasviolated in several analyses and because a violation of this as-sumption, in combination with unequal cell sizes, questions thevalidity of the F test (Stevens, 1996), we also tested differencesbetween classes with the Kruskal-Wallis test, a nonparametricstatistical test. With this test, differences in the mean ranks be-tween classes were analyzed. We only report the mean valuesbecause most readers might be unfamiliar with mean ranks. Themean values, the correlation between norm classes and the fre-quency and intensity judgments, the F tests, and the Kruskal-Wallis tests for within-nation differences are given in Tables 5, 6,7, and 8. If the Levene test indicates that the equal varianceassumption is violated, the results of the Kruskal-Wallis test

should be interpreted. Moreover, it is important to note that someclasses are very small in some nations (see Tables 2 and 3) andthat, therefore, the F test should generally be interpreted cau-tiously. However, as the results show, the F test and the Kruskal-Wallis test led us to the same conclusion with respect to thesignificance of the results in many cases, showing cross-analysesconsistency of the findings.

Positive emotional experience. Between-classes differences inthe reported frequency and the intensity of emotions are given forall nations in Tables 5 and 6. In general, the correlations betweenthe norm classes and the frequency and intensity judgments arelow to medium sized when they are evaluated with respect toCohen's (1988) classification of effect sizes. They are comparablylarge in Taiwan. The mean differences are significant in mostcases, showing that differences in norms are related to differencesin emotional experiences. Generally, the mean frequency andintensity values are larger in classes in which these emotions areconsidered desirable than in classes in which these emotions areconsidered neutral or undesirable. In particular, in the two collec-tivistic nations, the frequency and intensity values of pride arerather small in Class 4, in which all positive emotions are consid-ered desirable with the exception of pride and contentment.

Negative emotional experiences. The results concerning thefrequency and intensity of negative emotions across latent normclasses are less consistent across nations, and the correlations arealso low to medium sized (see Tables 7 and 8). When we considerthe Kruskal-Wallis test only, 10 out of 32 associations betweennorms and experience are significant. Moreover, with only oneexception, significant associations were only found for the twocollectivistic nations. This result shows that norms for positiveemotions are more closely linked to emotional experiences thanare norms for negative emotions.

Between-Nations Differences Within Classes

In the next step, we analyzed between-nations differences in thereported experience of affect within each of the latent classes. Inthese analyses, differences between nations in the norms for emo-tions are controlled. Thus, mean differences in the frequency andintensity of emotions between nations do not reflect differences inthe desirability of emotions. The most interesting question regard-ing these analyses is whether individuals who do not differ in theirattitudes toward emotions between nations do differ in the inten-sity and frequency of their emotions. The results of these analysesare given in Tables 9, 10, 11, and 12. The mean values on whichthese analyses are based are the mean values reported in Tables5-8. Again, we only focus on the most interesting results.

Positive emotional experiences. Significant international dif-ferences were found for all classes, with the exception of Class 5aand Class 5b. Class 5a exists only in Australia and Taiwan, andwithin this class the two nations are not different. Class 5b could

Table 4Mean Classification Probabilities

Norm items

Positive affectNegative affect

Class 1

.81

.89

Class 2

.86

.88

Class 3

.92

.81

Class 4

.93

.73

Class 5

.85

.79

Class 6

.79

878 EID AND DIENER

Table 5Frequency of Positive Emotions: Within-Nation Mean Differences

Emotion

JoyAffectionPrideContentment




1

3.864.473.484.02

3.032.762.282.24

4.444.393.523.90

4.094.424.044.16

2

4.006.001.003.00

3.102.292.222.18

3.312.912.462.97

2.004.001.003.50

Class

3

3.312.672.492.43

4.003.503.504.00

3.853.703.593.43

4

3.333.503.172.67

3.112.412.062.06

3.873.501.942.63

5

Australia

4.143.432.293.29

China

3.022.662.412.49

Taiwan

4.254.442.613.66

r

.07

.16

.18

.18

.09

.11

.14

.22

.24

.26

.40

.23

United States

3.904.503.203.80

.13

.14

.18

.16

F(djs)

5.222.65*3.09*2.76*

1.091.592.58*2.41*

8.14**9.92**

25.25**7.63**

2.513.07*4.67**3.67*

(3, 284)(3, 283)(3, 283)(3, 284)

(4,531)(4, 533)(4, 533)(4,531)

(4, 526)(4, 528)(4, 528)(4, 528)

(3,430)(3,432)(3,431)(3, 432)

L

******

****

*

2.157.639.86*8.06*

6.557.51

10.26*8.14

28.70**38.40**88.01**30.57**

7.059.61*

11.68**9.86*

(3)(3)(3)(3)

(4)(4)(4)(4)

(4)(4)(4)(4)

(3)(3)(3)(3)

Note. Reported are the mean values (possible range = 1 to 7). Empty columns indicate that the class does notexist in a culture, r represents the correlation between the class variable and the dependent variable (calculatedby the square root of if). The chi-square test we used was the Kruskal-Wallis test. The sample size for thechi-squares can be computed using the degrees of freedom of the F tests: N = df\ + df2 + 1. L = Levene testof equal error variances (asterisks indicate the significance of violations of the homogeneity assumption).*p < .05. ** p < .01.

not be compared between nations because it exists only in China.The first interesting result concerning the frequency and intensityof positive emotions is that China has, in all classes with exceptionof Class 2 and for all positive emotions, the lowest mean value (seeTables 5 and 6). This shows that China is the nation with thelowest frequency and intensity of positive emotions even if wecorrect for national differences in the desirability of emotions. Themean values of the other nations depend on the latent classes. Ingeneral, Australia, Taiwan, and the United States are quite similar,with the exception of the frequency and intensity of pride inClass 4, the class in which pride is undesirable. In this class, themean values of the frequency and intensity of pride are lower inTaiwan than in Australia and the United States. This is in line withthe role pride plays in more collectivistic countries.

Negative emotional experiences. The results for the negativeemotions are given in Tables 11 and 12. For the negative emotions,the national differences are rather strong for anger, and the differ-ences are stronger for the intensity than for the frequency judg-ments. For the intensity judgements, the international differencesare rather strong in Class 2, in which all negative emotions areconsidered undesirable, in Class 4, in which only sadness and guiltare desirable, and in Class 5, in which anger and fear are desirablebut sadness and guilt are not. In general, China has the lowestmean values in the majority of classes. This result implies that,

controlling for norms for negative affect, the Chinese report lessfrequent and less intense unpleasant emotions. Considering thefrequency of the negative emotions, Taiwan is similar to Australiaand the United States, but with respect to the intensity of negativeemotions, there are stronger differences between Taiwan and thetwo individualistic countries. Hence, there is a tendency for peoplein the more individualistic countries of Australia and the UnitedStates to feel negative emotions more intensely than do people inthe more collectivistic countries of China and Taiwan.

Discussion

The analyses reveal a number of interesting insights into thestructure of norms for emotions and the cross-cultural generaliz-ability of this structure. We discuss the results in five areas: (a) thestructure of norms for emotions, (b) intranational variability innorms for affect, (c) norms and emotional experiences, (d)between-nations differences in emotional experience, and (e) theuse of latent class analysis in cross-cultural studies on emotions.

The Structure of Norms for Emotions

Considering the results of the norms for positive and negativeemotions together, the main differences between nations can be


Table 6Intensity of Positive Emotions: Within-Nation Mean Differences

Emotion





1

4.894.994.034.35

3.903.553.093.06

5.124.774.074.57

4.815.064.464.43

2

3.005.002.004.00

3.432.902.752.69

3.773.512.833.43

2.002.501.503.50

Class

3

3.843.543.153.03

5.004.003.003.50

4.434.373.983.35

4

3.834.003.173.17

3.633.232.512.64

3.884.132.382.81

5

Australia

4.294.002.433.71

China

3.413.022.782.69

Taiwan

4.814.593.014.17

r

.18

.17

.22

.16

.14

.14

.18

.14

.32

.23

.41

.32

United States

4.104.903.403.70

.20

.20

.19

.25

F(4fs)

3.16=2.74'4.93'2.49

2.76:

2.914.592.57

14.747.60

* (3,* (3,* (3,

(3,

* (4,* (4,**(4,* (4,

**(4,**(4,

26.89** (4,14.76

6.236.245.429.33

**(4,

**(3,**(3,**(3,**(3,

284)284)284)284)

532)532)532)530)

528)527)528)528)

432)431)431)432)

L

**

*

*

**

7.087.07

13.63**7.08

11.75*11.05*18.86**11.27*

49.44**34.37**90.76**52.82**

11.98**15.57**10.69*21.99**

(3)(3)(3)(3)

(4)(4)(4)(4)

(4)(4)(4)(4)

(3)(3)(3)(3)

Note. Reported are the mean values (possible range = 1 to 7). Empty columns indicate that the class does notexist in a culture, r represents the correlation between the class variable and the dependent variable (calculatedby the square root of if). The chi-square test we used was the Kruskal-Wallis test. The sample size for thechi-squares can be computed using the degrees of freedom of the F tests: N = d^ + df2 + 1. L = Levene testof equal error variances (asterisks indicate the significance of violations of the homogeneity assumption).*p < .05. **p < .01.

found in pride and guilt. In more collectivistic cultures guilt ismore important, whereas in individualistic cultures pride is ofgreater relevance. Hence, the results confirm our theoretical pre-dictions that the group of emotions that are often prone to cross-cultural differences is the self-conscious or self-reflective emo-tions—emotions that reflect on the individual's own actions (seeTangney & Fischer, 1995). The approach-oriented, individualisticcultures think that self-reflective emotions about a person doingwell are good, whereas the collectivistic, Confucian cultures in thePacific Rim believe that self-reflective emotions that indicate thatone's controllable actions are wrong or insufficient are desirable.This is in line with cultural differences in regulatory focus. Forcultures that are promotion focused, pride is more important,whereas for cultures that are prevention focused, guilt is moredesirable.

Generally, the results show that not all "positive" emotions areconsidered positive (i.e., desirable, appropriate) by most individ-uals. Furthermore, not all "negative" emotions are considerednegative (i.e., undesirable, inappropriate) by most individuals. Thisfinding has consequences for the concept of social desirability. Instudies scrutinizing whether affect judgments are biased by asocial desirability response style, it is typically assumed that pos-itive emotions are considered desirable and negative emotions areconsidered undesirable. However, this basic assumption seems to

be wrong. A stronger test for the assumption that affect judgmentsare (not) distorted by social desirability would be to take intoaccount group differences in the desirability of emotions andanalyze the influence of desirability judgments on affect judg-ments in different subgroups.

Intranational Variability in Norms for Affect

The nations in this study do not have one set of norms to whicheveryone adheres. Instead, the countries include people with avariety of views about the normativeness of emotions. The nationson average differ, but this disguises the fact that there is substantialvariability in norms within the societies. Culture seems to influ-ence the number of people adhering to a particular viewpoint, butthe cultures are nevertheless heterogeneous enough that there aredifferences within them. Thus, cross-cultural researchers need toconsider not only average cultural differences but also the vari-ability within cultures.

These within-nation differences can be explained by two con-cepts of cross-cultural psychology: first, the distinction betweenidiocentrism and allocentrism, and, second, the distinction betweentight versus loose cultures (e.g., Triandis, 1989). The concepts ofidiocentrism and allocentrism reflect the distinction between indi-vidualism and collectivism and are concepts we can use to char-

880 EID AND DIENER

Table 7Frequency of Negative Emotions: Within-Nation Mean Differences

Emotion

Class

F(d/s)

Australia

AngerFearSadnessGuilt




2.772.442.882.36

2.502.632.652.27

2.752.772.592.31

3.242.793.062.62

2.672.402.872.11

2.262.312.161.89

2.412.542.622.11

2.932.513.012.39

2.502.402.902.20

2.382.572.382.00

2.522.772.652.21

3.092.482.962.13

1112.962.772.85

2.372.472.622.26

2.362.512.302.34

3.292.672.952.76

3.202.503.402.40

3.132.492.932.36

China

2.542.912.342.04

2.502.222.582.22

Taiwan

2.783.002.612.57

2.872.803.401.87

United States

3.072.672.862.40

3.492.702.972.65

.20

.16

.07

.19

.08

.16

.14

.10

.14

.13

.16

.13

.16

.12

.04

.13

2.34*1.510.302.16

0.642.71*2.151.19

3.32**1.862.87*1.84

2.31*1.300.191.51

(5,284)(5, 283)(5, 284)(5, 284)

(5, 534)(5, 532)(5, 534)(5, 535)

(5, 523)(5, 525)(5, 525)(5, 525)

(5, 420)(5, 422)(5, 422)(5,422)

*

*

***

*

9.155.020.539.69

7.1914.70*11.70*2.78

15.30**9.64

11.85*11.32*

12.48*6.061.498.60

(5)(5)(5)(5)

(5)(5)(5)(5)

(5)(5)(5)(5)

(5)(5)(5)(5)

Note. Reported are the mean values (possible range = 1 to 7). Empty columns indicate that the class does notexist in a culture, r represents the correlation between the class variable and the dependent variable (calculatedby the square root of TJ2). The chi-square test we used was the Kruskal-Wallis test. The sample size for thechi-squares can be computed using the degrees of freedom of the F tests: N = dfi + df2 + 1. L = Levene testof equal error variances (asterisks indicate the significance of violations of the homogeneity assumption).*p < .05. **p < .01.

acterize nations on the individual level. Thus, within individualis-tic and collectivistic nations there can be idiocentric andallocentric individuals. However, in individualistic nations thereshould be more idiocentric individuals, and in collectivistic coun-tries there should be more allocentric individuals. The latent classanalyses revealed that most people of the two individualisticcountries show an idiocentric norm pattern: In the two morecollectivistic nations, both idiocentric and allocentric patterns canbe found, although most of the people belong to the more allocen-tric patterns.

Nations differ not only in their collectivism and individualismbut also in how loose or tight they are. Tight cultures are veryhomogeneous with respect to norms. In tight cultures, there is ahigh pressure on all individuals of a society to follow these norms.Loose cultures do not have such strong norms and tolerate moredeviations. The concentration coefficient might be regarded as ameasure of the tightness versus looseness of a nation regarding thevariables under study. According to this coefficient, Australia andthe United States are relatively tight nations with respect to normsfor positive affect. In these nations, there might be pressure onindividuals to be joyful, happy, and full of love and pride and tomake use of their constitutional right to the pursuit of happiness.Deviations from this norm of happiness might have a strongimpact, and being unhappy might be regarded as failing. People

who are less happy are expected to correct their unhappiness byusing, for instance, psychotherapy. It was unexpected that theindividualistic nations are the most uniform with regard to pleasantaffect norms. Although these nations are "loose" in terms of thenorms for behavior, at the level of emotions they appear not to beloose. Rather, the desirability of happiness seems to be prescribedby the culture. Indeed, one justification for people "doing then-own thing" is that everyone ought to be happy and, therefore,follow their own desires. Thus, allowing for more variability inindividual behavior seems, ironically, to require a strong normabout the desirability of positive experiences.

The two more collectivistic nations seem to be looser countrieswith respect to norms for positive emotions. In particular, China,which was characterized as "a collectivistic, but 'relatively' loosecountry" by Triandis (1989, p. 511), shows a large variety in normtypes. In China, people who are unhappy might also be accepted,because in China there are also individuals who think that positiveemotions are undesirable. Pride is considered undesirable by alarge number of people in the two more collectivistic countries, butthere are still many people who think that feeling pride is accept-able. Thus, the norms for positive emotions (e.g., not to feel pride)seem not to be so strong that all individuals must follow them. Thismight be a sign that there is a movement to more diversity in thesecountries. For norms for negative emotions, the intranational va-


Table 8Intensity of Negative Emotions: Within-Nation Mean Differences

Emotion





1

4.533.684.503.59

3.442.943.272.91

3.933.513.613.03

4.623.714.293.70

2

4.313.764.293.43

2.512.152.812.41

3.443.393.693.04

4.383.554.163.62

Class

3

3.803.804.603.20

2.982.903.032.71

3.213.503.312.90

4.003.523.613.22

4

4.193.964.524.04

2.932.803.112.91

3.082.912.962.75

4.763.814.004.10

5 6

Australia

5.004.404.403.20

4.713.754.513.54

China

3.042.863.042.55

3.102.492.992.74

Taiwan

3.653.703.652.95

4.073.534.533.27

United States

4.793.744.003.48

4.033.574.143.32

r

.17

.08

.08

.12

.20

.20

.09

.11

.21

.14

.20

.08

.16

.07

.10

.12

1.710.410.340.81

4.30**4.30**0.991.45

4.99**2.23

(5,(5,(5,(5,

c (5,• (5 ,

(5,(5,

< (5,(5,

4.38** (5,0.62

2.180.400.971.16

(5,

(5,(5,(5,(5,

284)284)283)283)

534)535)533)533)

525)525)524)524)

421)422)422)420)

L

**

*

9.032.042.023.14

20.41**23.55**

5.808.34

20.51**10.0021.39**

2.63

10.961.625.634.87

(5)(5)(5)(5)

(5)(5)(5)(5)

(5)(5)(5)(5)

(5)(5)(5)(5)

Note. Reported are the mean values (possible range = 1 to 7). Empty columns indicate that the class does notexist in a culture, r represents the correlation between the class variable and the dependent variable (calculatedby the square root of TJ2). The chi-square test we used was the Kruskal-Wallis test. The sample size for thechi-squares can be computed using the degrees of freedom of the F tests: N = rff, + df2 + 1. L = Levene testof equal error variances (asterisks indicate the significance of violations of the homogeneity assumption).*p < .05. **p < .01.

riety differs less strongly between nations. All nations are rela-tively heterogeneous in their norms for experiencing unpleasantemotions, but China is nevertheless the most heterogeneous. Froma methodological point of view, the results show that latent classanalysis is an appropriate methodology to discover internationaldifferences in the tightness versus looseness of nations.

Norms and Emotional Experiences

In terms of emotional experiences, we found that there are low-to medium-sized correlations between norm classes and emotionalexperiences. For positive emotions, we found a consistent patternsuch that the frequency and intensity of positive emotions arerelated to the norm classes in such a way that higher desirabilitygoes along with respondents reporting more experience of thisemotion. For negative affect, however, the results were much moreinconsistent. We discuss the results for both types of affect sepa-rately. However, it should be noted that the variegated pattern offindings suggests that people do not automatically infer normsfrom their own emotional experiences.

For positive emotions, significant associations with respect tothe frequency of emotions were predominantly found for pride andcontentment. This is in line with the latent structure of the normdata because the classes differ mainly between the desirability ofpride and contentment. The association between the desirability

and frequency of positive emotions might point to the influence ofemotion regulation. People who think that a positive emotion isdesirable might seek that emotion more strongly than other people.On the other hand, if a person thinks, for example, that pride isundesirable, that person will avoid situations in which this emotionoccurs. If those situations cannot be avoided, the person willappraise the situation in such a way that pride will not arise; if thisis not possible, it is likely that the person will regulate this feelingdownward. Hence, the association between the frequency of pos-itive affect and norms for these emotions suggests that regulativebehavior is effective. However, although individuals can and doseek such situations, no situation is perfectly under the control ofindividuals, and therefore the associations might not be verystrong.

The associations between norm classes and emotional experi-ences, in particular emotion intensity, might also suggest theimportance of socialization influences. If individuals have learnedthat specific emotions are desirable, they are free to feel theseemotions intensely. However, if they have learned that specificemotions are undesirable or inappropriate, they might have learnedto regulate this feeling downward.

Why are the results so different for negative emotions? Oneinteresting result is that not only people who think that negativeemotions are desirable but also people who think that negative

882 EID AND DIENER

Table 9Frequency of Positive Emotions: Between-Nations MeanDifferences Within Different Norm Types (Classes)

Emotion r


.30

.29

.32

.35

Class 1

31.08** (3, 977)30.95** (3, 979)38.28** (3, 978)44.92** (3, 980)

81.25** (3)75.05** (3)

103.34** (3)109.26** (3)


.19

.45

.22

.33

Class 2

1.08 (3,85)7.19** (3, 85)1.45 (3,84)3.59* (3,85)

5.14 (3)16.78** (3)5.93 (3)

12.37** (3)


.20

.30

.32

.31

Class 3

4.16* (2, 195)9.62** (2, 196)

11.02** (2, 196)10.25** (2, 195)

9.08* (2)24.96** (2)20.32** (2)19.41** (2)


.23

.36

.25

.32

Class 4

2.60 (3, 199)10.20** (3, 201)4.57** (3, 201)7.40** (3, 199)

7.88* (3)22.04** (3)10.18* (3)15.23** (3)

Class 5a


.00

.11

.05

.05

0.072.830.740.62

(1,230)(1,231)(1,231)(1,231)

0.09 (1)2.74 (1)0.56 (1)1.12 (1)

Note. Reported are the mean values (possible range = 1 to 7). Emptycolumns indicate that the class does not exist in a culture, r represents thecorrelation between the class variable and the dependent variable (calcu-lated by the square root of TJ2). The chi-square test we used was theKruskal-Wallis test. The sample size for the chi-squares can be computedusing the degrees of freedom of the F tests: N = dfl + df2 + 1. L = Levenetest of equal error variances (asterisks indicate the significance of viola-tions of the homogeneity assumption).* p < .05. ** p < .01.

emotions are undesirable feel them (moderately) frequently andintensely. One possible explanation for this phenomenon is thatthere might be two groups of people. One group of people mayoften experience negative emotions because, for them, negativeemotions have a positive function, indicating that something isgoing wrong. These people might be able to use this indicator ofnegative emotions to cope with the situations that elicit theseemotions. Additionally, there might be another group of peoplewho frequently experience these emotions but are not able toregulate them—these people might be stuck with the emotionswithout making use of their function. These people might begenetically predisposed to negative affect and therefore experienceit more often and believe that it is not appropriate because theyfind it to be aversive. These assumptions would predict that inter-individual differences in norms might be only weakly related tointerindividual differences in emotional experiences. These as-sumptions, however, are speculative, and they indicate that the

linkage between emotional norms and emotional experiencesmight be rather complex. However, this subject seems to be a veryfascinating topic of future cross-cultural studies on emotions.

Between-Nations Differences in Emotional Experience

There appear to be differences in the experience of emotionsacross nations even when norms are held constant. Thus, it appearsthat factors such as genetics or life circumstances also influenceemotional experiences beyond the influence of norms.

The most important result here is the finding that China consis-tently shows the lowest experiences in almost all norm classes.People in China also have the lowest frequency and intensityscores of both positive and negative affects. This result can beexplained by the general value emotions have in China (for anoverview, see Russell & Yik, 1996). Several authors, such asKlineberg (1938), Potter (1988), and Wu (1982), have pointed out

Table 10Intensity of Positive Emotions: Between-Nations MeanDifferences Within Different Norm Types (Classes)

Emotion F(djs)


.26

.29

.26

.29

Class 1

23.71** (3,980)30.61** (3, 978)23.13** (3, 979)30.03** (3, 980)

68.89** (3)58.48** (3)63.75** (3)77.25** (3)

Class 2


.25

.30

.16

.31

1.962.76*0.793.12*

(3, 85)(3, 85)(3, 85)(3, 85)

7.13 (3)8.03* (3)2.87 (3)

10.57* (3)


.22

.23

.21

.09

Class 3

4.97** (2, 196)5.28** (2, 195)4.67** (2, 196)0.84 (2, 195)

9.49** (2)9.36** (2)9.54** (2)1.27 (2)


.09

.26

.16

.18

Class 4

0.51 (3, 199)4.80** (3,201)1.83 (3,200)2.27 (3,201)

1.28 (3)12.71** (3)2.72 (3)5.28 (3)

Class 5a


.08

.07

.08

.06

1.671.201.490.98

(1,231)(1,231)(1,231)(1,231)

2.09 (1)1.94 (1)1.47 (1)1.78 (1)

Note. Reported are the mean values (possible range = 1 to 7). Emptycolumns indicate that the class does not exist in a culture, r represents thecorrelation between the class variable and the dependent variable (calcu-lated by the square root of if). The chi-square test we used was theKruskal-Wallis test. The sample size for the chi-squares can be computedusing the degrees of freedom of the F tests :N = dfl + df2+ 1. L = Levenetest of equal error variances (asterisks indicate the significance of viola-tions of the homogeneity assumption).* p < . 0 5 . * * p < . 0 1 .


Table 11Frequency of Negative Emotions: Between-Nations MeanDifferences Within Different Norm Types (Classes)

Emotion F(dfi)

Class 1

more similar to China. Considering the frequency of negativeemotions, Taiwan is similar to Australia and the United States, butwith respect to the intensity of negative emotions there are strongerdifferences between Taiwan and the two individualistic countries.These results confirm the role of Taiwan as a nation that ishistorically and philosophically strongly linked to China but nowis more oriented to individualistic values.







.26

.11

.16

.12

.24

.08

.25

.17

.21

.11

.18

.09

.27

.13

.19

.16

.20

.11

.21

.17

.16

.10

.13

.18

Note. Reported are thecolumns indicate that thepniTP 1 a t\ on h<

lated by the

9.78** (3,408) **1.59 (3,408) **3.77 (3,409)1.88 (3,409) **

Class 2

12.35** (3, 591)1.18 (3,593) **

13.20 (3,594)5.83** (3, 595)

Class 3

2.93* (3, 191)0.74 (3, 191)2.12 (3, 191)0.57 (3, 191)

Class 4

5.07** (3, 187)1.12 (3, 188)2.38 (3, 188)1.77 (3, 188)

Class 5

2.00 (3, 136)0.55 (3, 135) **2.10 (3, 136)1.40 (3, 135)

Class 6a

1.80 (2, 134)0.77 (2, 134)1.08 (2, 134)2.32 (2, 134)

mean values (possible range =class does not exist in a culture.

3tween the c ' n 5 5 vnrinhlp and thp Hpnpnfipntsquare root

Kruskal-Wallis test. The

28.45**8.47*

13.95**14.79**

49.63**12.19**41.18**16.69**

9.08*5.157.74**9.26*

16.03**6.217.408.98*

9.36*1.435.917.37

3.971.441.609.26*

(3)(3)(3)(3)

(3)(3)(3)(3)

(3)(3)(3)(3)

(3)(3)(3)(3)

(3)(3)(3)(3)

(2)(2)(2)(2)

1 to 7). Emptyr represents thevariahlp. (c.Pi\n\-

of T}2). The chi-square test we used was thesample size for the chi-squares can be computed

using the degrees of freedom of the F tests: N = df + df2 -test of equaltions of the 1* p < .05.

h l .L = Leveneerror variances (asterisks indicate the significance of viola-lomogeneity** p < .01.

that in China there isdangerous,:irrelevant, (

assumption).

a general attitude to consider emotions as)r illness causing. Moreover, the moderation

or suppression of emotions is generally highly valued in Chin;i. In

Latent iStudies

Class Analysuon Emotions

; in Cross-Cultural

The analyses show that latent class analysis is a helpful meth-odology to explore and test typological structures of emotion. With

Table 12Intensity' of Negative Emotions: Between-Nations MeanDifferences Within Different Norm Types (Classes)

Emotion







r

.31

.24

.32

.23

.43

.35

.31

.26

.24

.23

.25

.12

.46

.34

.37

.34

.49

.31

.28

.27

.23

.07

.12

.07

F(djs) L

Class 1

14.59** (3, 409)8.65** (3,409)

15.11** (3,409)7.50** (3, 407)

Class 2

44.29** (3, 593)28.24** (3, 594) *20.73** (3, 593)14.87** (3, 594)

Class 3

4.07** (3, 190)3.43* (3, 191)4.38** (3, 191)0.92 (3, 189) *

Class 4

16.57** (3, 188) *8.36** (3,188)9.90** (3, 187)7.99 (3,188)

Class 5

14.39** (3, 137)4.80** (3, 136)3.83* (3, 135)3.48* (3, 136)

Class 6a

3.99* (2, 134)0.36 (2, 134)0.98 (2, 134)0.33 (2, 134)

f(df>

41.55**29.06**42.02**21.46**

111.76**72.48**53.08**41.07**

11.50**10.40*10.54*4.84

36.25**18.64**24.74*18.23**

32.71**14.89**11.78**13.75**

9.30*0.912.200.71

(3)(3)(3)(3)

(3)(3)(3)(3)

(3)(3)(3)(3)

(3)(3)(3)(3)

(3)(3)(3)(3)

(2)(2)(2)(2)

particular, the value of a moderated emotional life might explainthe relatively low scores of the Chinese even in the classes inwhich the emotions are considered desirable. An alternative ex-planation is that there are genetical differences in the physiology ofemotional responses between different countries.

Another interesting result is that Taiwan is very similar toAustralia and the United States with respect to almost all positiveemotions, with the exception of pride. Regarding pride, Taiwan is

Note. Reported are the mean values (possible range = 1 to 7). Emptycolumns indicate that the class does not exist in a culture, r represents thecorrelation between the class variable and the dependent variable (calcu-lated by the square root of if). The chi-square test we used was theKruskal-Wallis test. The sample size for the chi-squares can be computedusing the degrees of freedom of the F tests: N = df + df2 + l.L = Levenetest of equal error variances (asterisks indicate the significance of viola-tions of the homogeneity assumption).* p < .05. **p < .01.

884 EED AND DIENER

latent class analysis, it is possible to consider (a) categoricalresponse variables, (b) intra- and international differences in indi-vidual profiles, and (c) universal and culture-specific norm types.Furthermore, it is possible to test the equivalence of structuresacross cultures in a strong way. The results reveal that the strongassumption of measurement invariance had to be rejected for thetotal sample but that there are subgroups of individuals that areequivalent across nations and subgroups that are culture specific.Comparing the nations by one statistic (e.g., the mean value), as isoften done, obscures this cultural variety and complexity.

A model with universal and culture-specific norm classes fitsthe data very well. This clearly shows that there is structuralequivalence and structural diversity between nations. It also dem-onstrates the advantage of latent class analysis in separating uni-versal from culture-specific emotion patterns that go beyond "tra-ditional" methods of cross-cultural psychology. Furthermore, thelatent classes revealed that there are different typologies of norms.In particular, for negative emotions we found quite different normpatterns. One interesting result, for example, is that the desirabilityof anger contrasts sharply with the desirability of guilt. In classesin which anger is desirable, guilt is always undesirable, and inclasses in which anger is undesirable, guilt is always desirable. Afurther interesting result is that there are classes in which allpositive emotions are considered undesirable and all negativeemotions are considered desirable. These different norm patternsdemonstrate that a typological approach is superior to and moreinformative than a dimensional approach.

In the study of norms for emotions presented in this article,generalized norms for affect were analyzed. In future studies, itmight be interesting to use latent class analysis for exploring thestructure of situation-specific feeling rules and to scrutinize howstrongly norms of emotions generalize across situations. More-over, it might be useful to include more nations differing inindividualism and collectivism.

References

Bond, M. H. (1996). Chinese values. In M. H. Bond (Ed.), The handbookof Chinese psychology. Hong Kong: Oxford University Press.

Briggs, J. L. (1970). Never in anger: Portrait of an Eskimo family.Cambridge, MA: Harvard University Press.

Cacioppo, J. T., & Gardner, W. L. (1999). Emotion. Annual Review ofPsychology, 50, 191-214.

Clogg, C. (1995). Latent class models: Recent developments and prospectsfor the future. In G. Arminger, C. C. Clogg, & M. E. Sobel (Eds.),Handbook of statistical modeling in the social sciences (pp. 311-359).New York: Plenum.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences(2nd ed.). Hillsdale, NJ: Erlbaum.

Darwin, C. (1970). The expression of the emotions in man and animals.Chicago: University of Chicago Press. (Original work published 1872)

Diener, E., Smith, H., & Fuijta, F. (1995). The personality structure ofaffect. Journal of Personality and Social Psychology, 69, 130-141.

Drasgow, F., & Kanfer, R. (1985). Equivalence of psychological measure-ment in heterogeneous populations. Journal of Applied Psychology, 70,662-680.

Eid, M. (2001). Advanced statistical models for the study of appraisal andemotional reaction. In K. R. Scherer, A. Schorr, & T. Johnstone (Eds.),Appraisal processes in emotion: Theory, methods, research (pp. 319-330). New York: Oxford University Press.

Eid, M., & Rauber, M. (2000). Detecting measurement invariance in

organizational surveys. European Journal of Psychological Assessment,16, 20-30.

Ekman, P., & Friesen, W. V. (1969). The repertoire of nonverbal behavior:Categories, origins, usage, and coding. Semiotica, 1, 49-98.

Frijda, N. H., & Mesquita, B. (1995). The social roles and functions ofemotions. In S. Kitayama & H. R. Markus (Eds.), Emotion and culture:Empirical studies of mutual influence (pp. 51-87). Washington, DC:American Psychological Association.

Gardner, W. L., Gabriel, S., & Lee, A. (1999). "I" value freedom, but "we"value relationships: Self-construal priming mirrors cultural differencesin judgment. Psychological Science, 10, 321-326.

Grammer, K., & Eibl-Eibesfeldt, I. (1993). Emotionspsychologie im Kul-turenvergleich [Emotion psychology in cultural comparisons]. In A.Thomas (Ed.), Kulturvergleichende Psychologie. Eine Einfiihrung (pp.289-322). Gottingen, Germany: Hogrefe.

Hochschild, R. (1983). The managed heart. Berkeley, CA: University ofCalifornia Press.

Holt, J. A., & Macready, G. B. (1989). A simulation study of the differencechi-square statistic for comparing latent class models under violation ofregularity conditions. Applied Psychological Measurement, 13, 221-231.

Izard, C. E. (1980). Cross-cultural perspectives on emotion and emotioncommunication. In H. C. Triandis & W. Lonner (Eds.), Handbook ofcross-cultural psychology (Vol. 3, pp. 185-221). Boston: Allyn &Bacon.

Kitayama, S., & Markus, H. R. (Eds.) (1995). Emotion and culture:Empirical studies of mutual influence (pp. 23-50). Washington, DC:American Psychological Association.

Kitayama, S., Markus, H. R., & Matsumoto, D. (1995). Culture, self, andemotion: A cultural perspective on "self-conscious" emotions. In J. P.Tangney & K. Fisher (Eds.), Self-conscious emotions: The psychology ofshame, guilt, embarrassment, and pride (pp. 439-464). New York:Guilford Press.

Klineberg, O. (1938). Emotional expression in Chinese literature. Journalof Abnormal and Social Psychology, 33, 517-520.

Langeheine, R., Pannekoek, J., & van de Pol, F. (1996). Bootstrappinggoodness-of-fit measures in categorical data analysis. SociologicalMethods and Research, 24, 492-516

Langeheine, R., & Rost, J. (1988). Latent trait and latent class models.New York: Plenum.

Lazarsfeld, P. F., & Henry, N. W. (1968). Latent structure analysis.Boston: Houghton Mifflin.

Lee, A. Y., Aaker, J. L., & Gardner, W. L. (2000). The pleasures and painsof distinct self-construals: The role of interdependence in regulatoryfocus. Journal of Personality and Social Psychology, 78, 1122-1134.

Leung, K., & Bond, M. H. (1989). On the empirical identification ofdimensions for cross-cultural comparisons. Journal of Cross-CulturalPsychology, 20, 133-151.

Markus, H. R., & Kitayama, S. (1991). Culture and the self: Implicationsfor cognition, emotion, and motivation. Psychological Review, 98, 224-253.

Matsumoto, D. (1996). Culture and psychology. Pacific Grove, CA:Brooks/Cole.

McCutcheon A. L. (1987). Latent class analysis. Newbury Park, CA: Sage.McCutcheon, A. L. (1998). Correspondence analysis used complementary

to latent class analysis in comparative social research. In J. Blasius & M.Greenacre (Eds.), Visualization of categorical data (pp. 477-488). SanDiego, CA: Academic Press.

McCutcheon, A. L., & Hagenaars, J. A. (1997). Comparative social re-search with multi-sample latent class models. In J. Rost & R. Langeheine(Eds.), Applications of latent trait and latent class models in the socialsciences (pp. 266-277). Munster, Germany: Waxmann.

McCutcheon, A. L., & Nawojczyk, M. (1995). Making the break: Popularsentiment toward legalized abortion among American and Polish Cath-


olic laities. International Journal of Public Opinion Research, 7, 232-252.

Mesquita, B., & Ellsworth, P. (2001). The role of culture in appraisal. InK. R. Scherer, A. Schorr, & T. Johnstone (Eds.), Appraisal processes inemotion: Theory, methods, research (pp. 233-248). New York: OxfordUniversity Press.

Mesquita, B., & Frijda, N. H. (1992). Cultural variations in emotions: Areview. Psychological Bulletin, 112, 179-204.

Mohiyeddini, C , & Eid, M. (2001). [Norms for the expression and theexperience of emotions.] Unpublished raw data.

Potter, S. H. (1988). The cultural construction of emotion in rural Chinesesocial life. Ethos, 16, 181-208.

Read, T., & Cressie, N. (1988). Goodness-of-fit statistics for discretemultivariate data. New York: Springer.

Rost, J., Carstensen, C , & von Davier, M. (1997). Applying the mixedRasch model to personality questionnaires. In J. Rost & R. Langeheine(Eds.), Applications of latent trait and latent class models in the socialsciences (pp. 324-332). Munster, Germany: Waxmann.

Russell, J. A., & Yik, M. S. M. (1996). Emotion among the Chinese. InM. H. Bond (Ed.), The handbook of Chinese psychology (pp. 166-188).Hong Kong: Oxford University Press.

Saarni, C. (1999). The development of emotional competence. New York:Guilford.

Scherer, K. (1994). Evidence for both universality and cultural specificityof emotion elicitation. In P. Ekman & R. J. Davidson (Eds.), The natureof emotion. Fundamental questions (pp. 172-175). New York: OxfordUniversity Press.

Scherer, K. (1997). The role of culture in emotion-antecedent appraisal.Journal of Personality and Social Psychology, 73, 902-922.

Scherer, K. R., & Wallbott, H. G. (1994). Evidence for universality andcultural variation of differential emotion response patterning. Journal ofPersonality and Social Psychology, 66, 310-328.

Scherer, K. R., Wallbott, H. G., & Summerfield, A. B. (Eds.). (1986).Experiencing emotion: A cross-cultural study. Cambridge, England:Cambridge University Press.

Schieffelin, E. D. (1983). Anger and shame in the tropical forest: An affectas a cultural system in Paua New Guinea. Ethos, II, 181-191.

Shaver, P. R., Wu, S., & Schwartz, J. C. (1992). Cross-cultural similaritiesand differences in emotion and its representation. In M. S. Clark (Ed,),Emotion: Review of personality and social psychology (pp. 175-213).Newbury Park, CA: Sage.

Sommers, S. (1984). Adults evaluating their emotions: A cross-culturalperspective. In C. Z. Malatesta & C. Izard (Eds.), Emotions in adultdevelopment (pp. 319-338). Beverly Hills, CA: Sage.

Stearns, P. N. (1994). American cool: Constructing a twentieth centuryemotional style. New York: New York University Press.

Stearns, P. N., & Lewis, J. (1998). An emotional history of the UnitedStates. New York: New York University Press.

Stevens, J. (1996). Applied multivariate statistics for the social sciences(3rd ed.). Mahwah, NJ: Erlbaum.

Stipek, D. (1998). Differences between Americans and Chinese in thecircumstances evoking pride, shame, and guilt. Journal of Cross-Cultural Psychology, 29, 616-629.

Stipek, D., Weiner, B., & Li, K. (1989). Testing some attribution-emotionrelations in the People's Republic of China. Journal of Personality andSocial Psychology, 56, 109-116.

Suh, E., Diener, E., Oishi, S., & Triandis, H. C. (1998). The shifting basisof life satisfaction judgments across cultures: Emotions versus norms.Journal of Personality and Social Psychology, 74, 482-493.

Tangney, J. P., & Fischer, K. (Eds.). (1995). Self-conscious emotions: Thepsychology of shame, guilt, embarrassment, and pride. New York:Guilford Press.

Tangney, J. P., Wagner, P., Fletcher, C , & Gramzow, R. (1992). Shamedinto anger? The relation of shame and guilt to anger and self-reportedaggression. Journal of Personality and Social Psychology, 62, 669-675.

Triandis, H. (1989). The self and social behavior in differing culturalcontexts. Psychological Review, 96, 506-520.

Triandis, H. (1997). Cross-cultural perspectives on personality. In R.Hogan, J. Johnson, & S. R. Briggs (Eds.), Handbook of personalitypsychology (pp. 439-464). San Diego, CA: Academic Press.

Ulich, D., & Mayring, P. (1992). Psychologie der Emotionen [Psychologyof emotions]. Stuttgart, Germany: Kohlhammer.

van de Pol, F., Langeheine, R., & de Jong, W. (1996). PANMARK 3. Panelanalysis using Markov chains—A latent class analysis program. Voor-burg, the Netherlands: Netherlands Central Bureau of Statistics.

van de Vijver, F., & Leung, K. (1996). Methods and data analysis ofcomparative research. In J. W. Berry, Y. H. Poortinga, & J. Pandey(Eds.), Handbook of cross-cultural psychology (2nd ed., Vol. 1, pp.257-300). Chicago: Allyn & Bacon.

van de Vijver, F., & Leung, K. (1997). Methods and data analysis forcross-cultural research. Thousand Oaks, CA: Sage.

Vermunt, J. (1993). LEM: Log-linear and event history analysis withmissing data using the EM algorithm (WORC Paper 93.09.015/7).Tilburg, the Netherlands: Tilburg University.

Wallbott, H. G., & Scherer, K. R. (1988). How universal and specific isemotional experience? Evidence from 27 countries. In K. R. Scherer(Ed.), Facets of emotions (pp. 31-56). Hillsdale, NJ: Erlbaum.

Wickens, T. D. (1989). Multiway contingency table analysis for the socialsciences. Hillsdale, NJ: Erlbaum.

Wu, D. (1982). Psychotherapy and emotion in traditional Chinese medi-cine. In A. J. Marsella & G. M. White (Eds.), Cultural conceptions ofmental health and therapy (pp. 285-301). Dordrecht, the Netherlands:Reidel.

Received November 22, 2000Revision received April 13, 2001

Accepted April 13, 2001 •

norms for experiencing emotions in different … relations and group processes norms for...

Documents