Spontaneous Self-Descriptions and Ethnic Identities in [304649]

[anonimizat] (TST) was administered in Seoul and New "fork, to 454 students from 2 [anonimizat]. Responses, coded into 33 categories, were classified as either abstract or specific and as either autonomous or social. These 2 dichotomies were more independent in Seoul than in New York. The New York sample included Asian Americans whose spontaneous social identities differed. [anonimizat], or listed it once or twice. Unidentified Asian Americans' self-[anonimizat]-Americans' self-concepts, and twice identified Asian Americans' self-[anonimizat]' self-concepts, [anonimizat]-sociality. Differential acculturation did not account for these results. [anonimizat]-categorization, and acculturation theory are discussed.

[anonimizat]-concept.1 [anonimizat]-independent dimension. [anonimizat]-[anonimizat]-[anonimizat]-lectivistic cultures (Bochner, 1994; Markus & Kitayama, 1991; Triandis, 1989b). [anonimizat] a concrete-abstract dimension. [anonimizat]-istic cultures produce more abstract (not situated) [anonimizat] (Bond & Cheung, 1983; Cousins, 1989; Shweder & Bourne, 1984). However, independent-inter-[anonimizat]-changeably in this literature.

[anonimizat] "[anonimizat], James S. Uleman, and Robert J. Roman, [anonimizat] \brk University; Hoon K. Lee, [anonimizat], [anonimizat], South Korea.

This work was supported in part by a [anonimizat]. [anonimizat], [anonimizat]. We would also like to thank Magda Garcia for her help in developing the coding scheme and Susan LoBosco for coding the data.

[anonimizat], New York University, 6 Washington Place, 7[anonimizat], New York 10003. Electronic mail may be sent via the Internet to [anonimizat].

selves, what the aspects of the self are that concern the person" (McGuire & McGuire, 1981, p. 149). [anonimizat]-[anonimizat], [anonimizat], [anonimizat]. A [anonimizat]-neous ethnic identities.

Individualism, Collectivism, [anonimizat], [anonimizat]-scriptions offered by English speakers from the United States, Australia, or Britain and those offered by Chinese, Japanese, or Malaysian participants. Cousins (1989) found that Americans' self-descriptions contain more traits than do Japanese descrip-tions, whereas Japanese students' self-descriptions contain more concrete attributes and social categories. Bond and Cheung (1983) also found that Japanese students used fewer traits than American students.2 Trafimow, Triandis, and Goto (1991) found that native English speakers from the United States gave more private self-descriptions (e.g., kind and

The issue of whether individualism-collectivism is a single dimen-sion or two dimensions is not addressed here.

However, their Hong Kong students did not differ from American students in trait use. As the authors noted, "Hong Kong is a colonized territory, composed in large part ofa refugee population. Many Chinese see themselves as a mix of Chinese and western cultures" (Bond & Cheung, 1983, p. 155).

142

honest) than did Chinese participants, who gave more collective self-descriptions (e.g., brother and student) than native English speakers. Bochner (1994) found that Malaysians gave more col-lective and fewer idiocentric self-descriptions than did Austra-lians and British.

These differences in descriptions have been attributed to differences in the extent to which the cultures referred to are individualistic or collectivistic in orientation. Individualism de-scribes cultures, such as those of the United States and Western Europe, in which people maintain "loose ties" with each other and a belief in the inherent separateness of people. These cul-tures emphasize autonomy, emotional independence, individ-ual initiative, a right to privacy, primacy of personal goals over in-group goals, behavior regulated by attitudes, and acceptance of confrontation (Hofstede, 1980; Triandis, 1989a; Triandis, McCusker, & Hui, 1990). Furthermore, individualistic cultures emphasize personal goals and preferences, personal autonomy, and distinctive personal characteristics. These cultural impera-tives lead one to view people, including oneself, in terms of both autonomy (independent of specific others) and abstract dispo-sitions (internal attributes that are invariant over time and context; Markus & Kitayama, 1991; Miller, 1988). Newman (1993) found that individualists are more likely to make trait inferences spontaneously.

In contrast, people in collectivistic cultures are integrated into strong, cohesive in-groups. These cultures emphasize col-lective identity, emotional dependence, in-group solidarity and harmony, duties and obligations, behavior regulation by in-group norms, family integrity, and strong in-group-out-group distinctions (Kim, 1994; Triandis, 1989a; Triandis et al., 1990). Thus, people from collectivistic cultures view the "per-son" not as an autonomous being with abstract qualities but in terms of specific relationships to significant others, which Shweder and Bourne (1984) labeled "sociocentrism." The fo-cus is on specific, situation-bound behaviors and social catego-ries rather than on abstract personality traits, because the per-son may change across social contexts (Markus & Kitayama, 1991; Miller, 1984). People from collectivistic cultures do de-scribe personal attributes, such as abilities, opinions, judg-ments, and personality characteristics, but these attributes are understood as situation specific, as sometimes elusive and unre-liable, and as not particularly diagnostic. According to Triandis (1989b), people from non-Western cultures are more likely to possess a "collective self" because their child-rearing practices emphasize the in-group.

A Korean sample offers the opportunity to extend these find-ings in two ways. Thefirstis by a change in sample. Korea has been described as one of the most collectivistic cultures (Bond, 1988; Hofstede, 1980), and a replication of the results obtained from Chinese and Japanese students could strengthen earlier interpretations. The second is by determining whether the two dimensions often referred to as part of an individualistic-col-lectivistic difference (abstract-specific and autonomous-social) are identical.

group because they may participate in both an individualistic and a collectivistic culture. In preliminary work, we found that spontaneous ethnic identities can be assessed with the Twenty Statements Test (TST; Rhee, Uleman, & Roman, 1991), which asks respondents to reply 20 times to the prompt "I am." In a sample of 61 Asian Americans, social identities fell into three groups. Some (32%) did not identify themselves as being Asian, Asian American, or of any particular nationality (although we could so identify them from another questionnaire). Others (50%) identified themselves by either ethnicity or nationality, and the rest identified themselves by both ethnicity (Asian or Asian American) and nationality (Chinese, Indian, or Korean). Our hypothesis in the present study was that the self-descrip-tions of doubly identified Asian Americans would resemble those of Koreans most closely, that the self-descriptions of un-identified Asian Americans would resemble those of Euro-Americans most closely, and that the self-descriptions of singly identified Asian Americans would fall between the preceding two categories. We also expected these differences to be related to acculturation to the United States.

Acculturation is a multifaceted process that refers to individ-ual changes over time in identification, attitudes, values, and behavioral norms through contact with different cultures (Berry, 1980; Berry, Trimble, & Olmedo, 1986). Even though cultural identity is an important aspect of acculturation, there is no research on how it influences people's self-concepts and which self-characteristics they consider most important (Phinney, 1990). What is clear is that the process of accultura-tion is an uneven one. For instance, people may subjectively identify with their ethnic group and have ethnic friends but not speak the ethnic language.

Both acculturation theory and social identity theory predict differences in self-descriptions among the three groups (Berry, 1980; Sue & Wagner, 1973; Tajfel, 1978a, 1978b). Three of Berry's (1976, 1980) four modes of acculturation may be rele-vant to Asian American's spontaneous ethnic identities. Un-identified Asian Americans, who mentioned neither their ethnic group nor their national group on the TST, may be Berry's as-similators. Singly identified Asian Americans, who mentioned either their ethnic or national group, may be integrators. Dou-bly identified Asian Americans, who mentioned both ethnic and national groups, may be Berry's rejectors, strongly asserting their ethnic distinctiveness. If this is correct, and if acculturative influences are reflected in spontaneous self-descriptions, then unidentified Asian Americans should describe themselves sim-ilarly to Euro-Americans and doubly identified Asian Ameri-cans should resemble Korean participants. Singly identified Asian Americans should incorporate aspects of each culture into their self-descriptions, falling between the other groups.

Social identity theory generates similar predictions. Ethnic identity3 is defined as a person's subjective awareness and ac-ceptance of membership in a particular social category (Tajfel, 1978a). People have multiple social identities; however, ethnic identity is particularly relevant in multicultural societies such

Asian Americans, Acculturation, and Social Identity

There is little research on Asian Americans' spontaneous self-descriptions, but Asian Americans are a particularly interesting

We use the term ethnic identity to refer to both ethnic and national identity.

144 RHEE, ULEMAN, LEE, AND ROMAN

as the United States, where many minority groups coexist with a dominant social group (Ethier & Deaux, 1990; Gibbs & Hu-ang, 1989;Phinney, 1990; Sue & Wagner, 1973;Tajfel, 1978b). According to social identity theory, identity is based on two kinds of social comparisons: those between people and those between groups. Ethnic identity seems to be neither salient nor important to unidentified Asian Americans, so they should make primarily interpersonal comparisons, referring more of-ten to abstract, autonomous categories such as traits and less often to intergroup categories such as social roles. In contrast, ethnic identity seems to be very important to doubly identified Asian Americans, so they should make more intergroup com-parisons, referring to more social identities and fewer traits. Sin-gly identified Asian Americans should fall between the preced-ing two groups.

The Present Study

We predicted that the self-descriptions of participants with Western European backgrounds (referred to herein as Euro-Americans) would be more abstract and more autonomous than those of Koreans. In addition, we examined the relation-ship between these two dimensions among Euro-Americans and Koreans. We also expected that other self-description differ-ences found in prior cross-cultural research would be evident, thus extending these findings. Koreans should qualify those traits that they do use more than Euro-Americans because Ko-reans give more weight to contextual factors (Cousins, 1989). Consistent with previous findings, Koreans should use more global self-descriptions (e.g., "human being") than Euro-Americans (Bond & Cheung, 1983; Cousins, 1989).4 Euro-Americans should express more emotions than Koreans be-cause the norm in Asian cultures is to restrain expression of feelings (Markus & Kitayama, 1991).

Our Asian American participants were divided into ethnically unidentified, singly identified, and doubly identified categories on the basis of the TST. We predicted that this three-level ethnic identity variable would predict how much their self-descriptions resembled those of the individualistic Euro-Americans and col-lectivistic Koreans. We also expected these differences to be re-lated to the degree of acculturation to the United States.

Method

Participants

The sample consisted of 105 psychology majors from Vonsei Univer-sity in Seoul, Korea (68 men and 37 women), and 97 Euro-American (37 men and 60 women) and 151 Asian American (54 men and 97 women) college students from introductory psychology courses at New \brk University. The Asian American students included Chinese Americans, Indian Americans, and Korean Americans. Euro-Ameri-cans were so classified if at least three of their grandparents were from Northern or Western Europe or if their grandparents were Americans. (Our U.S. sample also included 38 men and 63 women with Eastern or Southern European backgrounds by the three-grandparent criterion. These Southern-Eastern Euro-Americans were analyzed separately [see Footnote 15].)

The mean age ofthe Korean sample was 20.5 years; Euro-Americans averaged 18.9 years of age, and Asian Americans averaged 18.9 years.

Seventy-one Asian Americans were first-generation immigrants, 51 were second-generation immigrants, and 1 was a third-generation im-migrant. Ninety-eight ofthem considered themselves bilingual, whereas 32 spoke only English. Generational information was unavailable for 28 Asian Americans, and some of the language information was unavail-able for 9.

The educational and socioeconomic levels of the students were roughly equivalent. Yonsei University is one of the most prestigious pri-vate universities in South Korea; it is located in a large urban setting, with students of varying socioeconomic backgrounds. New York Uni-versity is a highly ranked university, and the socioeconomic diversity of its students is similar to that of Yonsei University; however, New York University is less selective in its admission standards.

The U.S. students' ethnicity was established from a personal infor-mation questionnaire given shortly before the TST. This questionnaire also included questions about parents' socioeconomic status, the stu-dents' own generational status, bilingualism,firstlanguage learned, age at which English wasfirstlearned, and the extent of English use at home, at school, in the community, and in general, both while growing up and at present. The Korean students completed only the TST and a questionnaire about their parents' socioeconomic status.

Materials

The TST instructed participants to provide 20 answers to the ques-tion "Who are you?" This was followed by 20 blank lines beginning with the stem "I am." It was given to the U.S. students, in a mass testing session at the beginning ofthe fall term in introductory psychology, after they had completed a personal information questionnaire. The TST was given to the Korean students as thefirstquestionnaire in a small survey during a regular class. All materials were translated into Korean and back-translated into English to confirm the semantic equivalence ofthe questions for the Korean sample (Brislin, 1970, 1980).

TST Coding System

We developed a coding system for the TST that preserved traditional coding categories and included the two dimensions noted earlier (referred to here as abstract-specific and autonomous-social).5 The present coding system was based on Cousins's (1989) elaboration of work conducted by McPartland, Cumming, and Garretson (1961). Cousins used four basic categories, each representing a different level of abstraction. The present system extended these categories to eight; also,

subcategories were developed to be as inclusive as possible. They are shown in Table 1.

Each subcategory was classified as either abstract or specific (see Ta-ble 1); past practices were followed whenever possible (especially those of Cousins, 1989). For example, traits were divided into abstract (autonomous) and specific (social) states, the social traits being more specific because they referred to specific interpersonal contexts. The au-tonomous-social categories were designated in the same way, relying heavily on Markus and Kitayama's (1991) description of independent and interdependent selves. Self-descriptions that are invariant over time and context and that constitute an internal repertoire of thoughts, feel-ings, actions, desires, preferences, and abilities were classified as auton-omous. Self-descriptions that refer to social context, other people, time,

This category is similar to Bond and Cheung's (1983) "self and so-cial category," although the present coding did not include names.

We refer to the abstract-concrete distinction as abstract-specific to avoid the negative connotation that concrete has among academics. We refer to independent-interdependent as autonomous-social for greater precision.

Table 1

Coding System for the Twenty Statements Test

Category and subcategories

Traits

Pure (kind, friendly)

Qualified

Contextualized (with someone, at home)

Temporal (sometimes, a little)

Social identities

Role-status (student, major)

Family information (sister, son, last child)

Ethnicity/race/nationality

Gender (boy, woman)

Self-ascribed identities (musician, dancer)

Origin (from Hong Kong)

Religion (Christian, child of God)

Occupation (salesperson)

Negation (not a Christian)

Name

Specific attributes

Preferences (interests, likes, dislikes, fond of)

Autonomous (like books)

Social (like children)

Aspirations (wishes, hopes, wants)

Autonomous (be a doctor)

Social (help people)

Activities (activities, habits)

Autonomous (take the bus)

Social (visit friends)

Evaluative descriptions (abilities, evaluations, beliefs)

Autonomous (good in math, may not know myself)

Social (good listener, have many friends)

Physical descriptions

Descriptive (cute, short, sexy)

Age

Factual (height, weight, eye color)

Physical condition (near sighted)

Emotional states

Autonomous (worried, afraid)

Social (in love)

Peripheral information

Immediate situations, states (tired, hungry)

Present residence (live at home)

Other's descriptions

Possessions (clothes, ears)

Global descriptions

Universal-oceanic (human being, earthling)

Existential (me, myself)

Mean

proportion

30.0

24.7

5.3

3.2

2.1

21.6

8.3

3.4

2.5

2.4

1.2

0.9

0.9

0.7

0.2

0.1

13.1

6.0

5.2

0.8

4.9

3.9

1.0

2.2

2.0

0.1

o 7

O. /

6.5

2.3

6.3

3.1

1.6

1.2

0.4

5.0

4.9

0.1

3.6

2.4

0.7

0.3

0.2

2.2

1.4

0.9

Abstract or Autonomous

specific or social

Abstract Autonomous

Specific Social

Specific Social

Specific Social

Specific Social

Specific Social

Specific Social

Specific Social

Specific Social

Specific Social

Specific Social

Specific Social

Specific Autonomous

Specific Autonomous

Specific Social

Specific Autonomous

Specific Social

Specific Autonomous

Specific Social

Specific Autonomous

Specific Social

Specific Autonomous

Specific Autonomous

Specific Autonomous

Specific Autonomous

Abstract Autonomous

Specific Social

Specific Autonomous

Specific Social

Specific Social

Specific Autonomous

Abstract Social

Abstract Autonomous

Note. Proportions are rounded upward. Responses such as "I am a whale" and repeated descriptions were coded as nonsense.

and specific locations and are thus situation bound were classified as social.

Following Bond and Cheung (1983), our unit of analysis was mean-ings rather than responses, because one response may contain several meanings. For instance, "I am a good student" contains "I am a stu-dent" and "I am good." Such responses that contained an evaluation and a role were coded for the more distinctive role rather than the less distinctive evaluation. However, if "I am a student" had been a previous response, then "I am a good student" was coded as an evaluative de-scription. When several closely related descriptions were given in a re-sponse (e.g., "I am friendly and kind" or "I am a daughter and sister"), only thefirstdescription was coded, and the subsequent elaboration was ignored. If a response was repeated, the second one was ignored.

Scoring the TST

Responses from the Korean sample were translated into English and then back-translated into Korean. Translation reliability was .99 on a subset of 20% of the responses (approximately 400 of 2,000). The first author and a native-English-speaking coder unaware of the hypotheses assessed coding reliability on a subset (20%) of all of the responses (approximately 1,800 of 9,000). Coders were unaware of all informa-tion not on the TST. Interrater reliabilities for the two dimensions' cat-egories were .97 for abstract, .96 for specific, 1.00 for autonomous, and

.84 for social. Reliabilities for the specific categories were .98 for pure traits, 1.00 for qualified traits, .95 for social identities, 1.00 for specific attributes, .83 for evaluative descriptions, 1.00 for physical descriptions,

146 RHEE, ULEMAN, LEE, AND ROMAN

.76 for emotional states, 1.00 for peripheral information, and 1.00 for global descriptions. The native-English-speaking coder's scores were used in data analyses.

Fifty Asian American students (33%) did not identify themselves on the TST as Asian, Asian American, or of a particular nationality (Chinese, Indian, or Korean); 82 (54%) identified themselves either by ethnicity (8) or nationality (74); and 19 (13%) identified themselves as Asian, as Asian American, and by nationality.6 This subjective ethnic identification was used to categorize the Asian Americans as unidenti-fied, singly identified, or doubly identified.

To control for variations in the number of TST meaning units, we calculated proportions of responses in each category and subcategory on the basis of the total number of meaning units for each participant, excluding ethnic identifications. Table 1 presents the mean proportions of each category and subcategory for the entire sample.

Results

The results are organized as follows. Table 2 presents the cor-relations between the two dimensions in each subsample. Table 3 presents the proportion of self-descriptions that fell within each category. Figure 1 displays the proportion of each sub-group's self-descriptions that were abstract or specific, and Fig-ure 2 displays the proportion that were autonomous or social. These results point to (a) the need to distinguish between the two dimensions and (b) the similarity between the Korean and the doubly identified U.S. groups and the difference between these two subgroups and all others.

Correlations Between the Two Dimensions

The structure of the scoring system almost ensured that the proportion of abstract self-descriptions would correlate posi-tively with the proportion of autonomous self-descriptions and that specific self-descriptions would correlate positively with so-cial self-descriptions. Three of the four abstract subcategories (representing about 95% of the abstract responses) were also autonomous, and 18 of the 29 specific subcategories (repre-

Table 2

Correlations Between Poles ofEach

Dimension by Ethnic Group

*/>«.05.

senting about 60% of the specific responses) were also social (see Table 1). However, this did not prevent correlations from differing among subsamples. Accordingly, the proportion of ab-stract self-descriptions was correlated with the proportion of autonomous self-descriptions within each subsample, specific descriptions were correlated with social descriptions, and these correlations were compared between subsamples.7 Cohen's (1988) effect sizes for differences between two correlations were calculated to test the significance of differences between correlations.

The use of abstract and autonomous categories and specific and social categories was highly correlated among Euro-Ameri-cans and unidentified and singly identified Asian Americans but only moderately correlated among doubly identified Asian Americans and Koreans (see Table 2). These correlations were significantly higher among Euro-Americans than Koreans for abstract and autonomous categories, q{ 104) = .67, p < .01, and for specific and social categories, q( 104) = .56, p < .01. They were also higher among unidentified and identified Asian Americans than among Koreans for abstract and autonomous categories, q(67) = .62, p < .01, and 0(92) = .60, p < .01, respectively, and for specific and social categories, q(67) = .56, p < .01, and <?(92) = .54,/? < .01, respectively. None ofthe other comparisons were significant, partly because some samples were quite small. These results indicate that the abstract-spe-cific and the autonomous-social dimensions overlap to different degrees in different cultural subgroups and that they are not synonymous.

Ethnic and Subsample Differences on the Two Dimensions

There were significant differences in proportions of abstract self-descriptions, F(4, 348) = 24.79, p < .001 (see Figure 1 ).8 As predicted, this was partly the result of a significant linear trend among the groups, F(\, 348) = 62.46, p < .001, with Euro-Americans and unidentified Asian Americans at the high end of abstract descriptions and doubly identified Asian Ameri-cans and Koreans at the low end.9 Cohen's (1988) effect sizes for differences between proportions were used to test differences among subsamples. As expected, Euro-Americans used ab-stract descriptions more than Koreans did, h( 101) = .43, p < .01, and unidentified Asian Americans used them more than

There were no major differences among the Chinese American, In-dian American, and Korean American participants. Thus, these groups are referred to by a more general label, Asian Americans.

We focused on these rather than the abstract-social and specific-autonomous correlations because the other pairs are often treated as synonymous in the literature. Clearly, these four correlations are re-lated, but they are not redundant because the coding system does not include all responses.

Analyses of specific categories yielded essentially the mirror image of the results for abstract categories, so they are omitted here. For the same reason, analyses of social categories are omitted.

There was also a significant cubic trend, F( 1, 348) = 15.83, p <

.001, because Euro-Americans used fewer abstract categories than expected.

Note. Decimal points are omitted. Proportions are rounded up, and, as a result of rounding errors, some do not add to 1.00. Proportions with different subscripts in each row are significantly different from each other at the .05 level by protected t tests.

doubly identified Asian Americans, #(28) = .56, p < .05, or Koreans, A (68) = .61, p > .01. Singly identified Asian Ameri-cans differed from Koreans, h(92) = .32, p < .05, whereas dou-bly identified Asian Americans did not, h(32)= .05, ns.

To examine differences in the use ofabstract-specific descrip-tions within groups, we calculated difference scores for each group and tested against the null hypothesis of no difference. As expected, Koreans used specific more often than abstract self-descriptions, h'( 105) = \Al,p< .01. Singly identified and dou-bly identified Asian Americans also used specific more often than abstract self-descriptions, /i'(82) = 0.75, p < .01, and h'(19) = 1.46, p < .01, respectively. Unidentified Asian Ameri-cans showed no difference in the use of specific and abstract self-descriptions, /i'(50) = 0.11. Unexpectedly, Euro-Americans used specific more often than abstract self-descriptions, h'(91) = 0.40,p<.01.

There were also significant differences among the groups in autonomous descriptions, F(4, 348) = 6.84, p < .001 (see Fig-ure 2). This was due in part to the predicted linear trend among groups, F(\, 348) = 36.06, p < .001.10 However, no groups differed significantly from each other in their use of autono-mous self-descriptions (As ranged from .02 to .33). All groups except the doubly identified Asian Americans, /?'(18) = 0.26, used autonomous more often than social self-descriptions (h's ranged from 0.55 to 1.36, ps < .01).

Ethnic and Subsample Differences on the Six Categories

Even though groups generally differed on the two dimensions as predicted, it is informative to examine results for each of six coding categories. Table 3 shows the proportions in each group and between-groups differences on each category." As ex-pected, Koreans used fewer traits than Euro-Americans and un-identified and singly identified Asian Americans; the largest difference was between Koreans and unidentified Asian Ameri-cans, F(4, 348) = 21.09, p < .001. This same pattern was true of pure traits, which are unqualified by context or time. Again,

the largest difference was between Koreans and unidentified Asian Americans, F(4, 348) = 24.36, p < .001.12

The three groups of Asian Americans were arrayed as ex-pected, with unidentified Asian Americans using the most traits and doubly identified Asian Americans using the fewest. This provides clear evidence that subsamples of Asian Americans, defined in terms of their spontaneous ethnic identity, differ pre-dictably in other (logically unrelated) aspects of their spontane-ous self-description.

We expected social identities (e.g., student and sister, exclud-ing ethnicity and nationality) to be most common among Ko-reans and least common among Euro-Americans, but the re-verse occurred, F(4, 348) = 9.76, p < .001 (see Table 3).13 We also expected doubly identified Asian Americans to use more social identities than unidentified Asian Americans, with singly identified Asian Americans falling in between the other two groups. This expectation was confirmed. However, unlike the

Again, there was a significant cubic trend, F( 1, 348) = 20.17, p <

.001, because Euro-Americans used fewer autonomous descriptions than expected.

' Even though there were eight categories in the coding scheme, the results of only six categories are discussed. We had no predictions for the evaluative descriptions or peripheral information categories. How-ever, the F tests for these categories were significant at p < .01 (see Table 3 for proportions).

There were no significant differences among the groups in the use ofthe less frequent qualified traits, F(4, 348) = 0.31.

To examine differences in the use of social identities more closely, we conducted analyses on the most frequent subcategories: role-status and family information. There were significant differences for both the role-status and family subcategories, F(4, 348) = 6.24, p < .001, and F(4, 348) = 3.52, p < .01, respectively. As expected, doubly identified Asian Americans used more than singly identified Asian Americans, who used more than Euro-Americans and unidentified Asian Ameri-cans. However, Koreans used fewer than expected. When self-ascribed identities were excluded from the social identities, Euro-Americans and Koreans no longer differed.

148 RHEE, ULEMAN, LEE, AND ROMAN

EURO UNID SINGLY-ID DOUBLY-ID KOREAN

ABSTRACT SPECIFIC

Figure 1. Proportion of abstract and specific categories (see Table 1) by subgroup. EURO = European Americans; UNID = unidentified Asian Americans; SINGLY-ID = singly identified Asian Americans; DOUBLY-ID = doubly identified Asian Americans.

earlier result for traits, this result is open to the simple and un-interesting alternative interpretation that the use of some social identities (e.g., ethnicity) predicts the use of others such as membership in a family or institutions.

Because specific attributes (preferences, aspirations, and activities) and physical descriptions are specific rather than ab-stract categories, we expected them to be most frequent among Koreans' and doubly identified Asian Americans' self-descrip-tions. As predicted, Koreans used specific attributes (pref-erences, aspirations, and activities) most frequently (26%), whereas the other groups did not differ (less than 13%), F(4,

= 35.97, p < .001. The particular subcategories that ac-counted for this effect were preferences, F( 4, 348) = 12.80,p<

.001, and aspirations, F(4, 348) = 35.76, p < .001. However, there was a significant difference among groups on the physical descriptions category, F(4, 348) = 18.29, p < .001. Mean pro-portions among Asian American subsamples were ordered as predicted, but Euro-Americans and Asian Americans were higher than predicted and higher than Koreans.

We predicted that emotional states would be most frequent among Euro-Americans and least frequent among Koreans and doubly identified Asian Americans. This was confirmed, F(4,

= 4.77, p = .001. On the basis of prior results, we also predicted that Koreans would use the most abstract category, global descriptions, more often than the other groups. This too was confirmed, F(4, 332) = 2.88, p < .05. In particular, Ko-reans used more universal descriptions than the other groups, F(4, 332) = 5.06, p = .001, consistent with Cousins's (1989) findings.

In sum, Euro-Americans used more traits, pure traits, and emotional states than Koreans, as predicted. They also used fewer specific attributes and global descriptions than Koreans, as predicted. The Asian American subsamples were also or-

dered as predicted for four of these categories, with traits, pure traits, and emotional states decreasing and specific attributes increasing from unidentified to doubly identified Asian Ameri-cans. Asian American subsamples were also ordered as ex-pected for three other categories, with social identities and phys-ical descriptions increasing from unidentified to doubly identi-fied Asian Americans.

There were only two significant unexpected findings. Euro-Americans used more social identities and physical descriptions than Koreans. These unexpected differences prompted us to compare our samples' self-descriptions with those from the only other published study of which we are aware that has used some of these same categories to compare individualistic and collec-tivistic cultures (Cousins, 1989).l4 Our Korean sample ap-peared to be comparable to Cousins's Japanese sample in the use of traits (13% and 19%, respectively), social identities (35% and 27%, respectively), and physical descriptions (4% and 5%, respectively). However, our Euro-Americans appeared to use fewer traits (27% vs. 58%), more social identities (38% vs. 9%), and more physical descriptions (12% vs. 2%) than Cousins's American sample. Possible explanations are offered in the sec-tion to follow.15

Direct comparisons on the full measure were not possible because Cousins (1989) reported proportions for thefivemost important attri-butes checked by each participant. To obtain a comparable measure, we recalculated our proportions on the basis ofonly thefirstfiveresponses; we did so on the arguable assumption that these most accessible catego-ries were also the most important.

Because the literature suggests that individuals from Southern and Eastern Europe differ from Euro-Americans (Markus & Kitayama, 1991; Triandis, 1989b), the former participants were compared with Euro-Americans on the self-description dimensions. Participants were classified as Southern-Eastern Euro-Americans if at least three of their

0.8

EURO UNID SINGLY-ID DOUBLY-ID KOREAN

AUTONOMOUS SOCIAL

Figure 2. Proportion of autonomous and social categories (see Table 1) by subgroup. EURO = European Americans; UNID = unidentified Asian Americans; SINGLY-ID = singly identified Asian Americans; DOUBLY-ID = doubly identified Asian Americans.

Acculturation Predictions ofEthnic Identity and Self-Descriptions

Acculturation variables were examined to determine whether they could predict Asian Americans' spontaneous ethnic identities or self-descriptions. The Asian American groups did not differ in generation status, F{2, 116) < 2.0; socioeconomic status, F(2,

< 1.0; or bilingualism, F(2, 139) < 1.0. The three groups also did not differ in English use at home while growing up, at school, in the community, and in general, Fs(2, 139) < 2.0, ps >

. 10. They did differ marginally on present English use at home, F(2, 139) = 2.52, p = .084. The only clear differences were in the age at which they learned English, F( 2, 139) = 4.80, p = .01, and whether English was their first language, F(2, 139) = 6.75, p <

.01. Unidentified Asian Americans learned English at a mean of 3 years of age, whereas singly identified Asian Americans learned it at a mean of 5 years and doubly identified Asian Americans learned it at a mean of 6 years. Unidentified Asian Americans learned English as a first language 83% ofthe time; the correspond-

grandparents were from Southern or Eastern Europe or if their parents were from these countries. Most of the Southern-Eastern Euro-Ameri-cans were Jewish, Greek, Italian, and Russian. Their mean age was 19.4 years. They were divided into unidentified (n = 42) and identified (n =

subgroups. Overall, identified Southern-Eastern Euro-Americans used more specific and more social dimensions than unidentified South-ern-Eastern Euro-Americans and Euro-Americans. Thus, as with the Asian American subsamples, spontaneously self-identified Southern-Eastern Euro-Americans were more collectivistic in their self-descrip-tions than unidentified Southern-Eastern Euro-Americans and Euro-Americans.

ing figures were 45% and 26% for singly and doubly identified Asian Americans.

Hierarchical regression analyses of the TST dimensions and cat-egories were performed with the major acculturation variables (socioeconomic status, generational status, bilingualism, English as a first language, and English usage) and selected interactions (Generation X Socioeconomic Status, Generation X Bilingualism, and Generation X English usage) entered first and ethnic identity entered last. The acculturation variables did not significantly pre-dict self-description features, with a few exceptions. More use of spoken English, overall, predicted more use of abstract categories (fi = .94, p < .001). Learning English at a later age predicted more use of physical descriptions (/? = .24, p < .05). The interaction of generation and socioeconomic status predicted the use of social descriptions (0 = -.77, p < .05), with later generations and those of higher socioeconomic status using social descriptions less. No other results were significant.

Spontaneous ethnic identity significantly predicted features of self-descriptions after the effects of acculturation variables had been controlled. It predicted the use of specific descriptions (/? =

.40, p < .001), autonomous descriptions (0 = -.24, p < .05), and social descriptions (jS = .48,p < .001). It also predicted the use of four categories: traits (0 = -.32, p < .01), pure traits (/3 = -.34, p

.001), social identities(j3 = 35, p < .001), and physical descrip-tions (/3 = .29, p < .01). Thus, most acculturation variables used in the present study do not account for the effects of ethnic identity on Asian Americans' spontaneous self-descriptions.16

In an attempt to increase the predictive power of the acculturation variables, we factor analyzed them. All variables except socioeconomic status loaded on Factor 1. However, the means of these variables had fewer significant relations with the self-description category use than their components did.

150 RHEE, ULEMAN, LEE, AND ROMAN

Discussion

The purpose of this study was to examine differences in self-description by members of individualistic (Euro-American) and collectivistic (Korean) cultures, as well as by Asian Ameri-cans who differ in their level of ethnic identity. Past research on individualism-collectivism has emphasized two dimensions of the self-concept: abstract-specific and autonomous-social.

dimensions also show differences among Asian Americans who differ in their spontaneous ethnic identities. Those who identi-fied least with their collectivistic origins were most similar to the Euro-Americans, and those who identified most were most similar to the Koreans. This provides evidence that the strength of spontaneous ethnic identities is systematically related to other aspects ofthe self-description that are characteristic ofthe culture of origin and supports the validity of this measure of ethnic identity.

Results showed that the abstract-specific dimension is em-pirically distinct from the autonomous-social dimension, even though the two are often discussed in the literature as though they were interchangeable. We found the expected positive cor-relations between the use of abstract and autonomous catego-ries and between the use of specific and social categories; how-ever, there was also significant variation among groups. The cor-relations were highest among most U.S. subsamples and relatively low among Koreans. Thus, abstract-specific and au-tonomous-social categories have more similar referents in Euro-American culture than in Korean culture. Descriptively, this was because some categories that are largely specific and autonomous (specific attributes such as preferences, aspira-tions, and activities, as well as evaluative descriptions) or ab-stract and social (global descriptions) were more frequent in the Korean sample. As clearly shown in Figure 1, Koreans' sponta-neous self-descriptions were predominantly specific. Con-versely, pure traits and emotional states, which are abstract and autonomous categories, were more frequent in the Euro-Amer-ican sample. Figure 2 shows that Euro-Americans' spontaneous self-descriptions were predominantly autonomous.

The predominance of autonomous self-descriptions among Euro-Americans is very consistent with, and perhaps explained by, their culture's individualistic emphasis. The predominance of specific self-descriptions among Koreans may result from the in-group focus of collectivistic cultures and the value placed on not standing out from other in-group members. Highly specific self-descriptions distinguish one from other members of the in-group (which a self-description must do) while providing only narrow bases for that distinction. Highly specific self-descrip-tions thereby allow a person to be distinctive but not generally different from others. Conflating the abstract-specific and au-tonomous-social dimensions obscures the distinction between them and, particularly, oversimplifies the characterization of self-descriptions from collectivistic cultures.

Note that specific and autonomous descriptions were most frequent overall (see Figures 1 and 2) and were never less fre-quent than abstract or social descriptions in any group. This simple fact is also obscured ifthe two dimensions are conflated.

As predicted, we also found a significant linear decrease in both abstract and autonomous self-descriptions when moving from Euro-Americans through unidentified Asian Americans, singly identified Asian Americans, and doubly identified Asian Americans to Koreans. This suggests that both dimensions are useful and accurate in characterizing differences between indi-vidualistic and collectivistic cultures. More interestingly, both

Euro-Americans and Koreans differed significantly on each of the eight coding categories, although not always in the pre-dicted direction. As expected, Euro-Americans used traits (especially pure traits) and emotional states more than Ko-reans; Koreans used specific attributes and global descriptions more than Euro-Americans. However, Euro-Americans unex-pectedly used social identity and physical description categories more than Koreans. What might account for this? One possi-bility is that ethnic identities were particular salient to our U.S. sample. New York City and New York University are more eth-nically diverse than the University of Michigan, where Cousins (1989) gathered his data. New York University arts and science undergraduates are about 33% Asian American, 10% African American, and 10% Latino. According to the distinctiveness postulate (McGuire, McGuire, Child, & Fujioka, 1978), situa-tional distinctiveness (on visible features such as gender, ethnic-ity, and hair color) influences self-description. In the more eth-nically diverse New York group, ethnicity and associated char-acteristics such as physical attributes are more salient. Such factors may account for their frequent use by our Euro-Americans.

In addition, self-categorization theory (Turner, 1987) distin-guishes between personal identity and social identity. Because the self is dynamic (Markus & Wurf, 1987), different aspects of the self are made salient, or sampled, in different situations and times. There are at least three levels of abstraction of self-cate-gorization: (a) one's human identity, (b) one's social identity, and (c) one's personal identity. These identities are based, re-spectively, on interspecies, intergroup, and interpersonal com-parisons between oneself and others. Personal identity is more salient in intragroup contexts, whereas social identity is more salient in intergroup contexts. Thus, even though our Euro-American participants may often think of themselves in terms ofthe abstract traits associated with interpersonal comparisons, the ethnically diverse student body and testing situation may have prompted intergroup comparisons. This may explain their high use of the social identities and physical descriptions rela-tive to Cousins's (1989) U.S. sample from Michigan.

Self-Descriptions Among Asian Americans

The self-descriptions ofthe three subsamples ofAsian Ameri-cans (unidentified, singly identified, and doubly identified) var-ied as expected on the two dimensions and on most categories. Unidentified Asian Americans, like Euro-Americans, used ab-stract descriptions and autonomous descriptions the most,

whereas doubly identified Asian Americans, like Koreans, used these descriptions the least. Singly identified Asian Americans fell between these other Asian American subsamples. Singly and doubly identified Asian Americans also used more specific than abstract self-descriptions. On six of the eight coding categories, especially the most frequent ones, the three Asian American subsamples also responded largely as predicted. Unidentified Asian Americans used more trait and emotional state categories and used fewer social identity, specific attribute, and physical description categories. These findings are consistent with our predictions from both acculturation and social identity theories.17

Interestingly, unidentified Asian Americans were more ex-treme than the Euro-Americans on several categories. They used significantly more traits and fewer social identity and phys-ical description categories than the Euro-Americans. It is as though they "bent over backwards" to adopt a Euro-American perspective in their self-descriptions and, in the process, "over-shot" the norm of the dominant culture (Triandis, Kashima, Shimada, & Villareal, 1986). According to social identity the-ory (Tajfel, 1978a), people are motivated to maintain a positive self-image. If the group they belong to precludes this, they may leave it and try to "psychologically pass" as members of the dom-inant group, identifying with the mainstream culture.

Unexpectedly, the acculturation variables used in this study did not account for differences in the self-descriptions of Asian American subsamples. Perhaps these variables failed because the range of acculturation was attenuated, given that this entire subsample had been admitted to New York University. Al-ternately, perhaps the acculturation concept and variables were simply too remote from the processes that determine self-de-scriptions. By contrast, social identity and self-categorization theories proved to be quite useful, perhaps because they address the determinants of spontaneous self-descriptions more directly.

Future Research

Hofstede's (1980) observation that cultures vary on individ-ualism-collectivism has begun to generate a great deal of re-search, as well as measures at the individual level. The TST is one such measure and is particularly important because of hy-pothesized differences in spontaneous self-descriptions. Al-though prior work (Cousins, 1989;Markus&Kitayama, 1991) has discussed abstract-concrete and independent-interdepen-dent dimensions as unidimensional, the present results suggest that they are two dimensional: abstract-specific and autono-mous-social. Furthermore, these two dimensions are most dis-tinct in collectivistic cultures. Future research should examine the generality of this result.

Our speculations about the effects of a multicultural context on the TST suggest future research on its stability across a vari-ety of testing conditions. Alternatively, the TST's sensitivity to manipulations of individualism-collectivism should be exam-ined. We are currently looking for effects of the language used to administer the TST to bilingual speakers, predicting that lan-guage will invoke other features of the culture and affect the resulting self-descriptions. One might also vary the salience of

in-groups and out-groups or of communal and exchange rela-tionships (Batson, 1993; Clark & Mills, 1979,1993).

The psychometric properties of the TST should also be ex-plored (Wylie, 1974). The test-retest reliability and internal consistency of scores derived from the present coding system and from Cousins's (1989) coding system are unknown. Spon-taneous ethnic identity, assessed by the TST, also appears to be a promising measure. It is systematically related to many other categories of self-description in ways that are theoretically meaningful. However, the same reliability questions can be raised about it. Can the number of "items" be increased by ex-panding the range of responses that indicate the strength of eth-nic identity? How much is it subject to situational influences, particularly multicultural contexts?

Finally, future research should address the relation of this TST measure of ethnic identity to other, theoretically related measures. One would expect relations with acculturation mea-sures, particularly multi-item measures with a larger range than that found in the current sample. Aspects ofboth acculturation and social identity theory also suggest that unidentified Asian Americans, if they have renounced their cultures of origin, may differ on collective self-esteem (Luhtanen & Crocker, 1990).

Conclusion

Individualism-collectivism, as a cultural variable, seems to be related to at least two distinct dimensions of the spontaneous self-description: the abstract-specific dimension and the auton-omous-social dimension. These dimensions were only moder-ately correlated in our Korean and doubly identified Asian American U.S. sample, and they were both significantly related to differences between Euro-Americans' and Koreans' sponta-neous self-descriptions. In addition, our results show that the TST can be used to assess the strength of spontaneous social identities among minority groups in a majority culture. Among our U.S. Asian American subgroups, this measure was reliably related to other features ofthe spontaneous self-description that one would predict from differences between individualistic and collectivistic cultures.

Note that whatever the effect of the assessment context for the U.S. sample, it was the same for all ofthese subgroups. That is, all ofthe U.S. samples, regardless of ethnicity, completed the TST in an ethnically heterogeneous class setting.

References

Batson, C. D. (1993). Communal and exchange relationships: What is the difference? Personality and Social Psychology Bulletin, 19, 677-683.

Berry, J. W. (1976). Human ecology and cognitive style: Comparative

studies in cultural andpsychological adaptation. New \brk: Wiley.

Berry, J. W. (1980). Acculturation as varieties of adaptation. In A. Pad-ilia (Ed.), Acculturation: Theory, models and some newfindings(pp. 9-25). Boulder, CO: Westview Press.

Berry, J. W., Trimble, J., & Olmedo, E. (1986). Assessment ofaccultur-ation. In W. Lonner & J. Berry (Eds.), Field methods in cross-cultural research (pp. 291-349). Beverly Hills, CA: Sage.

Bochner, S. (1994). Cross-cultural differences in the self-concept: A test

152 RHEE, ULEMAN, LEE, AND ROMAN

of Hofstede's individualism/collectivism distinction. Journal of Cross-Cultural Psychology, 25, 273-283.

Bond, M. H. (1988). Finding universal dimensions of individual varia-tion in multicultural studies ofvalue: The Rokeach and Chinese value surveys. Journal of Personality and Social Psychology, 55, 1009-1015.

Bond, M. H., & Cheung, T. (1983). College students' spontaneous self-concept: The effect of culture among respondents in Hong Kong, Ja-pan, and the United States. Journal ofCross-Cultural Psychology, 14, 153-171.

Brislin, R. W. (1970). Back-translation for cross-cultural research.

Journal ofCross-Cultural Psychology, 1, 185-216.

Brislin, R. W. (1980). Translation and content analysis oforal and writ-ten materials. In H. Triandis & J. W. Berry (Eds.), Handbook of cross-cultural psychology (Vol. 2, pp. 389-444). Boston: Allyn & Bacon.

Clark, M. S., & Mills, J. (1979). Interpersonal attraction in exchange and communal relationships. Journal ofPersonality and Social Psy-chology, 37, 12-24.

Clark, M. S., & Mills, J. (1993). The difference between communal and exchange relationships: What is and is not. Personality and Social Psychology Bulletin, 19, 684-691.

Cohen, J. (1988). Statistical power analysisfor the behavioral sciences.

Hillsdale, NJ: Erlbaum.

Cousins, S. D. (1989). Culture and selfhood in Japan and the U.S.

Journal ofPersonality and Social Psychology, 56, 124-131.

Ethier, C, & Deaux, K. (1990). Hispanics in Ivy: Assessing identity and perceived threat. Sex Roles, 22, 427-440.

Gibbs, J. T, & Huang, L. N. (1989). Children of color: Psychological

interventions with minority youth. San Francisco: Jossey-Bass.

Hofstede, S. (1980). Culture's consequences: International differences

in work-related values. Beverly Hills, CA: Sage.

Kim, U. (1994). Introduction to individualism and collectivism: Con-ceptual clarification and elaboration. In U. Kim, H. C. Triandis, C. Kagiticibasi, S. C. Choi, &G. Yoon (Eds.), Individualism and collec-tivism: Theory, method, and application (pp. 19-40). Beverly Hills, CA: Sage.

Luhtanen, R., & Crocker, J. (1990). Collective self-esteem and ingroup bias. Journal ofPersonality and Social Psychology, 58, 60-67.

Markus, H. R., & Kitayama, S. (1991). Culture and the self: Implica-tions for cognition, emotion, and motivation. Psychological Review,

224-253.

Markus, H. R., & Wurf, E. (1987). The dynamic self-concept: A social psychological perspective. Annual Review ofPsychology, 38,299-337.

McGuire, W. J., & McGuire, C. V. (1981). The spontaneous self-con-cept as affected by personal distinctiveness. In M. D. Lynch, A. A. Norem-Hebeisen, & K. J. Gergen (Eds.), Self-concept: Advances in theory and research (pp. 147-172). Cambridge, MA: Ballinger.

McGuire, W. J., McGuire, C. V., Child, P., & Fujioka, T. (1978). Sa-lience of ethnicity in the spontaneous self-concept as a function of one's ethnic distinctiveness in the social environment. Journal ofPer-sonality and Social Psychology, 36, 511-520.

McPartland, T. S., Cumming, J. H., & Garretson, W. (1961). Self-con-ception and ward behavior in two psychiatric hospitals. Sociometry, 21, 111-124.

Miller, J. G. (1984). Culture and the development of everyday social explanation. Journal ofPersonality and Social Psychology, 46, 961 — 978.

Miller, J. G. (1988). Bridging the content-structure dichotomy: Cul-ture and the self. In M. Bond (Ed.), The cross-cultural challenge to social psychology (pp. 245-265). Beverly Hills, CA: Sage.

Newman, L. S. (1993). How individualists interpret behavior: Idiocen-trism and spontaneous trait inference. Social Cognition, 11, 243-269.

Phinney, J. S. (1990). Ethnic identity in adolescents and adults: Review of research. Psychological Bulletin, 108, 499-514.

Rhee, E., Uleman, J. S., & Roman, R. J. (1991). [Spontaneous ethnic identities among Asian American college students]. Unpublished raw data.

Shweder, R. A., & Bourne, E. J. (1984). Does the concept of the person vary cross-culturally? In A. Marsella & G. White (Eds.), Cultural conceptions of mental health and therapy (pp. 97-137). New York: Reidel.

Sue, S., & Wagner, N. N. (1973). Asian-Americans: Psychology perspec-

tives. Ben Lomund, CA: Science and Behavior Books.

Tajfel, H. (1978a). Social categorization, social identity and social com-parison. In H. Tajfel (Ed.), Differentiation between social groups (pp.

61-76). London: Academic Press.

Tajfel, H. (1978b). The social psychology of minorities. London: Mi-nority Rights Group.

Trafimow, D., Triandis, H. C, & Goto, S. G. (1991). Some tests of the distinction between the private self and the collective self. Journal of Personality and Social Psychology, 60, 649-655.

Triandis, H. C. (1989a). Cross-cultural studies of individualism and collectivism. In J. Berman (Ed.), Nebraska Symposium on Motiva-tion: Cross-cultural perspectives (pp. 41-133). Lincoln: University of Nebraska Press.

Triandis, H. C. (1989b). The self and social behavior in differing cul-tural contexts. Psychological Review, 96, 506-520.

Triandis, H. C, Kashima, Y, Shimada, E., & Villareal, M. (1986). Ac-culturation indices as a means of confirming cultural differences. In-ternational Journal ofPsychology, 21, 43-70.

Triandis, H. C, McCusker, C, & Hui, C. H. (1990). Multimethod probes of individualism and collectivism. Journal ofPersonality and Social Psychology, 59, 1006-1020.

Turner, J. C. (1987). Rediscovering the social group: A self-categoriza-

tion theory. Oxford, England: Basil Blackwell.

Wylie, R. C. (1974). The self-concept: A review of methodological con-siderations and measuring instruments (Vol. 1). Lincoln: University of Nebraska Press.

Received February 16, 1994

Revision received October 28, 1994

Accepted November 23, 1994 •