Copyright UMI Company 1999. All Rights Reserved.
Copyright Sage Publications, Inc. Sep 1999

Research shows that feedback concerning a person's prior performance is an important determinant of self-efficacy and subsequent work activity. In addition, several recent cultural models posit that people use different aspects of their environment in assessing their self-concepts. In this article, the authors explore Triandis's sampling-probability hypothesis of cultural influence by examining the relationship of an individual's cultural values and performance feedback referents to an individual's self-efficacy. A laboratory experiment is used to test hypotheses concerning the nature of self-efficacy and feedback referent (self vs. group) in relation to individualism-collectivism. The results show that, depending on cultural values held, participants relied on different combinations of individual- and group-based feedback. The results are discussed with regard to a general model of self-efficacy and culture in an organizational environment.

There is an increasing emphasis on the role of self-efficacy to organizational behaviors such as work performance (Gist, 1987; Gist & Mitchell, 1992; Locke, Frederick, Lee, & Bobko, 1984; Wood & Bandura, 1989). Self-efficacy, or an individual's cognitive estimate concerning his or her capacity to perform a given task, is an effective predictor of an individual's level of effort and task performance across a range of tasks. However, the origins of these efficacy estimates are not altogether clear. For example, what information referents are used by a person in forming his or her self-efficacy beliefs? Although several organization, task, and individual characteristics relevant to self-efficacy have been identified (Gist & Mitchell, 1992; Meyer & Gellatly, 1988), the cultural context in which people work is only beginning to receive research attention. In our study, we describe the relation of one important component of cultural values, individualism-collectivism, to selfefficacy formation as an extension and clarification of existing research (e.g., Earley, 1994).

Individualism-collectivism reflects an individual's values and beliefs toward himself or herself in existing social relationships (Hofstede, 1991; Schwartz, 1993). Individualism-collectivism is of key importance to social relationships in work organizations, and it serves to shape people's selfconcepts and actions.1 The construct of individualism-collectivism has been expanded and developed dramatically during the past decade. In one of the prominent current forms (Earley dc Gibson, 1998; Y. Kashima, personal communication, July 1995; Kim, 1994; Schwartz, 1990; Triandis, 1995), individualism-collectivism is conceptualized as multidimensional and multilevel. In this study, we focus on a particular facet of individualismcollectivism, namely, values concerning group-based work activities. This conceptualization is comparable to what Triandis (1995) described as vertical and horizontal collectivism, or an emphasis on group interdependence with a differentiated view of self from others. That is, people endorse (to varying degrees) interdependent work while seeing themselves as different (to varying degrees) from others in their social environment.

Several researchers argue that collectivists derive, in part, their sense of self based on the actions and reactions of important others, whereas individualists do so based on their self-evaluations of personal achievements (Markus dc Kitayama, 1991; Trafimow, Triandis, & Goto, 1991; Triandis, 1989). This suggests, for example, that a collectivist worker evaluates his or her actions based on the outcomes that the worker's group may receive, whereas an individualist worker does so on the basis of personal work attainments and recognition received (Triandis, 1989; Wagner & Moch, 1986).

This argument is consistent with an information processing view of self. According to this view, people are information seekers in an environment capable of providing a variety of information from various sources and referents (Ashford & Cummings, 1983; Ashford & Tsui, 1991; Greller 8 Herold, 1975). People seek out information about their actions, consciously and unconsciously, to structure and interpret their worlds, and information that helps to maintain one's self-concept receives central attention. Information sought by people is tied to their self-concepts, which in turn are tied to cultural values such as individualism-collectivism (Erez & Earley, 1993). For example, Markus and Kitayama (1991) suggested that a collectivist's selfconcept is tied to group actions and outcomes, whereas an individualist's self-concept is most heavily tied to individual-based evaluations and information. Likewise, Kulik and Ambrose (1992) suggested that demographic characteristics (e.g., gender, race) may influence a person's choice of a "referent other." This referent becomes the key source of information for the person as he or she evaluates himself or herself and others.

Triandis (1989) offered a "sampling-probability" explanation of culture and information scanning. According to his sampling hypothesis, an individual's cultural background guides, in part, the type of information that he or she attends to and the frequency with which it is sampled. Triandis suggests that individualists and collectivists look to, or sample from, multiple types of selves-public, private, collective-with different propensities. The public self refers to a generalized "other" person such as a prototypical stranger, the private self refers to oneself, and the collective self refers to one's in-group. According to Triandis, all three of these selves constitute a person's selfconcept, and a person's interaction with the environment is guided by the relative propensity with which he or she relies on information concerning a particular self. Consistent with Markus and Kitayama's (1991) arguments, Triandis (1989) posited that collectivists sample most frequently and directly from a collective self (group-referenced information), whereas individualists sample most frequently and directly from the private self (personally referenced information). For example, Triandis suggested that in families where children are urged to act independently (as often is the case in an individualist culture), the private self is likely to be accessed when the children face new challenges; in families that stress group harmony, the collective self is likely to be accessed. This is not to imply a deterministic perspective inasmuch as cultural background is only one of several significant influences on a person's information processing. Priming effects and relative salience of the situation will influence sampling as well (Trafimow et al., 1991 ). Thus, a person's self-concept can be derived from different referents of information based on cultural background, personal attributes, and contextual influence.

What role, then, does the sampling described by Triandis (1989) and Markus and Kitayama (1991) play in shaping self-efficacy? Bandura (1982, 1986) and Gist and Mitchell (1992) discussed the role of individual performance cues on subsequent efficacy but did not deal with group performance cues to any significant degree. However, drawing on the work of Triandis (1989) and Markus and Kitayama (1991), we argue that group-referenced performance feedback will be especially salient for collectivists and that individual-referenced performance feedback will be salient for individualists. Some empirical evidence examining this issue is reported by Earley (1994), who examined the effect of job training on efficacy and performance in individualist and collectivist cultures. He found that individualists responded better (higher self-efficacy and performance) to individually focused job training than to group-focused job training. The opposite finding was obtained for collectivists. However, Earley's study was a betweensubjects design (participants received no training, individual-focused training, or group-focused training), and it is unclear whether or not a trade-off among information referents would have occurred had participants received concurrently both individual- and group-focused training. Furthermore, Earley did not address performance feedback directly. Thus, it remains unclear whether people differentially attend to individual- versus group-based performance feedback as a function of their cultural values.

To our knowledge, no direct test of this feedback sampling proposition has been undertaken. However, reviews of the effects of feedback suggest that it may be selectively interpreted by individuals because of the receiver's motivational considerations (Balzer, Doherty, dc O'Conner, 1989; Ilgen, Fisher, & Taylor, 1979). For example, Bond, Wan, Leung, and Giacalone (1985) found that Chinese participants were more willing than American participants to accept critical feedback. In a similar vein, Ashford and Cummings (1983) proposed that feedback is a resource that is not inherently valued; rather, it is valued based on the end products it helps to produce. For individualists, self-referenced end products are more likely to be valued than are groupreferenced end products; thus, the individualist might view individual feedback as more instrumental than group feedback and might restrict information use to feedback that pertains only to the most valued individual outcomes.

Based on this review of the research and theory, we sought to test Triandis's sampling-probability hypothesis derived from the construct of individualism-collectivism. More specifically, the following was hypothesized:

Hypothesis 1: Individualists who were provided with both individual- and groupreferenced feedback on a performance task would base their self-efficacy and self-evaluations of performance on individual-referenced feedback. Collectivists who were provided with both individual and group feedback on a performance task would use feedback concerning their group's performance.

To test this hypothesis, we conducted a study with managers from an individualist culture (the United States) and two collectivist cultures (the Czech Republic and the People's Republic of China [PRC]). We chose these three countries, in part, because they provide us with a means for attaining a wide variation on the cultural construct of individualism-collectivism. That is, by sampling from countries whose modal cultural orientation differs on a given dimension, we increase the likelihood that we will sample individuals who range widely on personal endorsement of these cultural values and beliefs.

The United States is a highly individualist culture in which the focus is on individual accomplishment and self-interest (Hofstede, 1980), whereas the Czech Republic and the PRC emphasize a collective orientation (Earley, 1994; McGregor, 1991). It is acknowledged that American culture consists of many intracultural variations; however, trends distinct from other cultures are readily apparent in the United States. For example, Americans are guided by a strong work ethic emphasizing individual achievement and reward as well as a strong, individual goal orientation (Hofstede, 1980). Taken as a whole, ample evidence exists to suggest that Americans generally are individualist compared to members of most other cultures (Hofstede, 1991).

With regard to the PRC, ample evidence exists attesting to the strong collectivist orientation of individuals from China (Schwartz, 1993). In China, there exists an emphasis on extended family and friendships as well as an emphasis on work units or danwei (Chen, 1995). Employees come from a society that historically has been focused on collective actions and groupism (Bond,1988; Hsu,1981; Li,1978). For example, Bond, Leung, and Wan (1982) found that Hong Kong Chinese students tended to allocate rewards on the basis of an equality rule rather than an equity rule, particularly if there were ample resources to be distributed. This tendency is consistent with a highly collectivist value orientation (see Chen,1995, for an alternative interpretation).

The Czech Republic is a rapidly emerging capitalist system from a strong tradition of socialism and communism. The Czech Republic, formerly part of Czechoslovakia, was a member of the general alliance of Communist countries having been created by the former Soviet Union. In 1989, Czechoslovakia underwent the "Velvet Revolution," after which the country split itself into two semi-autonomous republics, Czech and Slovakia, and created separate legislatures (Machann, 1991; McGregor, 1991). Although the pace of capitalist reform in the Czech Republic is quite high, management practices, organizational, and societal culture reflect a strong orientation toward group identification and collectivism (McGregor, 1991). Although a number of Czech state-owned companies have been fully or partly privatized, evidence suggests that such reforms still are being met with some degree of skepticism. For example, a strike in the fall of 1994 of more than 8,000 workers from the Skoda automobile plant (one of the single largest state-owned companies in the republic) reflected a growing concern among employees that the capitalist reforms might be taking place at the cost of economic sovereignty. In this instance, Skoda employees were expressing skepticism regarding the proposed privatization of Skoda in an alliance with Volkswagen.

Research conducted by Holda and Cermakova (1980) and described by McGregor ( 1991 ) further demonstrates that the movement toward capitalism has not supplanted the generally collective orientation of employees and managers. What must be stressed, however, is that the collective orientation in the Czech Republic does not refer to an endorsement of communism or socialism; rather, it reflects a team orientation and an emphasis on groupbased welfare (McGregor, 1991). Holda and Cermakova (1980) found, for example, that Czechs endorsed a number of values including living in a happy family, helping those close to oneself, having good friendships, and living and working in peace. These dominant values reflect a general collectivist orientation toward group loyalty and team action but a rejection of a traditional emphasis on political involvement. Indeed, the value of being politically active was ranked lowest.

In summary, this article examines the relationship of individualismcollectivism values to the type of feedback paid attention to by managers. We test our hypothesis using a performance appraisal simulation in which managers read short descriptions of fictitious employees and rated their effectiveness.

METHOD

PARTICIPANTS

A total of 228 managers (92 American, 66 mainland Chinese, and 70 Czech) participated in the study on a voluntary basis. The American participants were recruited from a management training course that they were attending on the Organizations and Human Resources Function. The Chinese participants were recruited from management training programs in the northern and southern regions of the PRC. The Czech participants were recruited from management training courses hosted by a business school located just outside of Prague. All of the managers were natives of the countries in which they were attending their training and were employed in full-time management positions, and most of them were sponsored by their organizations for the course. The country groups were compared on a number of variables including age, education level, gender, job tenure, and company size (see Table 1). An examination of age, education, and company size with a one-way analysis of variance (ANOVA) using country demonstrated no significant differences for age or tenure, but the Chinese sample had a somewhat lower level of education and some individuals in the American sample came from a few very large companies, as reflected by the large variance in Table 1. An examination of gender using a chi-square analysis demonstrated no significant differences among the samples.

DESIGN AND TASK

The study consisted of a 2 (high or low individual feedback) x 2 (high or low group feedback) fully crossed factorial design with individualismcollectivism as a continuous variable. Individual feedback (high vs. low) referred to a person's own performance on the experimental task. Group feedback (high vs. low) referred to the average performance (per person) of a given subject's concocted group. Specifically, group feedback referred to an average score reported for a person's group reported in a metric consistent with the individual feedback manipulation. All of the materials used in the study were translated and back-translated into Chinese (Czech) for the participants in the PRC (Czech Republic) with the aid of two professional translators familiar with business materials and terms.

The task chosen for this study was designed to provide individual-level performance opportunities. We chose a managerial simulation in which the participants read short descriptions of employees and evaluated their performance using an appraisal form consisting of several evaluation items. The "correct" evaluation of an employee was sufficiently ambiguous to make the contrived performance feedback believable. However, this precluded us from using actual performance as a dependent variable in the study.

It is important to note that the task was designed to be performed by a single individual in what can be characterized as a "pooled" interdependence form (Van de Ven & Ferry, 1980). That is, people work on comparable activities independently, but the results of these efforts could be pooled or added together for a group total. We used this type of task so that group performance feedback would be meaningful. However, it is important to note that this was not a group task, and our concocted groups were used for the group feedback manipulation alone. The meaningfulness of the group manipulation was assessed through a manipulation check described in the Results section.

CULTURAL VALUES MEASURE

Individualism-collectivism was assessed using four items on the 5-point modified Likert scale (1 = strongly disagree, 5 = strongly agree) listed in Earley (1994). We chose those items dealing with group interaction and context given that this was the focus of our intervention. These items were as follows:

1. If the group is slowing me down, it is better to leave it and work alone. (reverse scored)

2. A person does better work working alone than in a group. (reverse scored)

3. Problem solving by groups gives better results than does problem solving by individuals.

4. Cooperation among team members usually helps to solve problems.

Responses to the scale were coded so that a high score indicates collectivist values and a low score indicates individualist values. A principal components analysis demonstrated that the items loaded on a single factor representing an endorsement of group-based activities having an eigenvalue of 2.19, accounting for 55% of the total variance (factor loadings ranged from .57 to .81). Subsequently, the items were averaged for a composite score having a reliability (Cronbach's alpha) of .71.

PROCEDURE

All participants followed the same experimental procedure. In the first session of the training course, managers completed a demographics questionnaire and the questionnaire measuring individualism-collectivism. The managers were told that they were assigned to work teams of four to six members based on self-reported demographic similarities including common work experiences, geographic locales, and general interests. The experimenter emphasized to the managers that their groups reflected a number of common interests and backgrounds among group members. In actuality, these groups were based on a random assignment, and group members across the three countries had a similar amount of social interaction prior to the simulation. Groups then were given class time to get to know one another by talking about themselves and their relevant work experiences. Throughout the remainder of the multiweek course, group members participated together as a unit during classroom exercises and met outside the classroom for study sessions, case analyses, and the like. This procedure was designed to contribute to the consistent formation of cohesive teams across the three countries.

During the second session of the course, the experimenter asked participants to write down on a piece of paper what percentages typically represent "high performance," "low performance," and "average performance" for employees in their workplace. This procedure allowed the experimenter to anchor the feedback manipulations in terms of percentage scores that were appropriate in the particular culture being studied. The U.S. participants reported that 92% represented a high score, 75% represented a low score, and 84% represented an average score. On average, the Czech participants reported that 84% represented a high score, 62% represented a low score, and 75% represented an average score. In the PRC, it was felt by the researcher (a native of that country) that adjectives describing performance (e.g., good, average, poor) would be more meaningful than a percentage value because of the participants' relative lack of familiarity with such percentage ratings. We used experts from the PRC who carefully scaled the percentages to adjectives for the purposes of the feedback manipulation. Thus, the manipulations of performance had functional equivalence across the three countries even though they were not empirically identical.2

The experiment was conducted during a normal class session. The exercise was introduced by the experimenter (with the assistance of a colleague in each country fluent in the local language) as a managerial assessment of a person's ability to evaluate others. A packet of materials containing an instruction sheet and performance descriptions for 20 employees was distributed to the participants. The packet contained the following instructions:

In the following exercise, you are going to make judgments concerning the general capabilities and performance of employees. You will be given a short description of an employee , and then you will be asked to evaluate the worker. Although you [might] not have time to finish all of the evaluations, please try to finish as many of the evaluations as you can in the time you are provided. After you are instructed by your professor to stop, he will collect your evaluations and give you feedback concerning how well you rate the employees according to a panel of personnel management experts. In addition, you will find out how well your fellow study group members did as they worked on this task compared [to] how the entire class did on it.

Participants also were given basic information concerning factors commonly used (for all three countries) for making performance evaluations (e.g., the worker should show that he or she is hard-working and dedicated to the job). They then were allowed to work on the 20 performance descriptions for a period of 15 minutes. The following was a typical performance description:

Alice Adams didn't feel very good about herself compared to others because she never did anything special and never got any recognition from the organization. Also, she didn't feel vary good about herself because she never seemed willing to make the effort to learn something new. She just didn't have the confidence to start a new hobby from scratch, even in the privacy of her own home. Alice liked balance. She personally liked it when things changed a little and also stayed somewhat the same.

(Note that the names were translated into local equivalents during the translation part of the study.) Based on such a description, participants were asked to evaluate the employee on a 5-point Likert scale with regard to three factors: (a) the performance level of the employee (1 = not at all satisfactory, 5 = extremely satisfactory), (b) how the employee compares to other employees (1 = much worse, 5 = much better), and (c) the general level of motivation and drive of the employee (1 = low, 5 = high).

At the end of the first 15-minute trial, the packets were collected and sorted according to study group. Participants were told that they could take a short break while the experimenter "scored" the performance evaluations based on a comparison of the evaluations to the evaluations given by the panel of personnel management experts. At this point, the experimental manipulation was enacted. Feedback forms reporting the percentage of evaluations that the participants evaluated correctly, the percentage that their study group evaluated correctly on average, and the average for the entire class were prepared for each participant; however, this feedback was contrived such that participants randomly received either (a) high individual feedback and high group feedback, (b) high individual feedback and low group feedback, (c) low individual feedback and low group feedback, or (d) low individual feedback and high group feedback. All participants within a given country received the same "class average" based on the percentage reported as representing average performance (84% for the United States, 75% for the Czech Republic, and a corresponding adjective in Chinese for the PRC). The percentages used to represent high versus low performance feedback matched the percentages reported as representative of typical high and low performance within each culture (e.g., high performance for Czech participants was reported as 84%).

When participants returned from a 10-minute break, they were handed their feedback forms and asked not to discuss the results with others. They were told that they would be given another opportunity to complete a second packet of performance evaluations and were asked to think about how they felt they would do on the next round. Participants then were asked to complete the questionnaire containing the dependent variables measures (selfefficacy, individual-based performance beliefs, and satisfaction with performance). After completing the questionnaire, the participants were given 15 minutes to complete a second packet of performance descriptions for 20 employees. At the end of this second performance trial, participants turned in their packets, were given a questionnaire containing manipulation checks, and were debriefed as to the objectives of the experiment. It was stressed to the participants that the feedback they had received had been concocted to test which feedback referent (individual or group) they paid the most attention to during the performance trials.

DEPENDENT MEASURES

Three dependent variables were measured: self-efficacy, individual-based performance beliefs, and satisfaction with performance. To measure selfefficacy, participants were asked to rate their self-efficacy for five levels of performance-correctly completing 4, 8, 12, 16, and 20 evaluations in a 15minute performance period-using a 100-point certainty scale (0 = certain the performance level cannot be achieved, 100 = certain the performance level can be achieved). For subsequent analyses, the responses to the scale were averaged for a composite self-efficacy score having a reliability (Cronbach's alpha) of .71.

In an attempt to assess efficacy beliefs more generally, an additional type of performance expectation was measured. To avoid confusion with the selfefficacy variable, we call this variable an individual-based performance belief. Individual-based performance beliefs were measured using four items rated on a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree). These items assess performance beliefs derived from self-referenced feedback:

1. I feel capable to do evaluations based on the feedback I received.

2. I feel competent to perform this evaluation task.

3. I feel a sense of mastery over the material that I worked on.

4. Based on my personal feedback, I feel confident that I can perform well.

For subsequent analyses, the responses to the scale were averaged for a composite score having a reliability (Cronbach's alpha) of .80.

Satisfaction with performance was assessed using two items, each rated on a 5-point scale (1 = not at all satisfied/happy, 5 = completely satisfied/ happy):

1. How satisfied are you with your performance from the first trial?

2. How happy do you feel about yourself based on your performance on the first trial?

For subsequent analyses, the responses to the scale were averaged for a composite score having a reliability (Cronbach's alpha) of .79.

RESULTS

DESCRIPTIVE STATISTICS

The means, standard deviations, and Pearson correlations for efficacy, individual-based performance beliefs, satisfaction with performance, and individualism-collectivism are presented in Table 2.

Two manipulation checks were conducted for the individual and group feedback manipulations using country of origin. Two items were used to assess individual and group feedback received (1 = very poor, 5 = very good):

1. Based on the personal feedback that I received, my performance was ....

2. Based on the group feedback that I received, my group's performance was ....

The results of two ANOVAs for individual feedback by country and group feedback by country demonstrated that the manipulations were effective. We obtained a significant main effect for individual feedback with no additional main or interaction effects for the first analysis as well as a significant main effect for group feedback with no additional main or interaction effects for the second analysis. Finally, a one-way ANOVA was conducted on individualism-collectivism using country of origin, and the results demonstrated that the U.S. sample was significantly more individualist than both the Czech and Chinese samples and that the Czech and Chinese samples did not differ from one another, F(2, 227) = 28.57, p < .01, Ms = 3.36, 4.23, and 3.97 for the United States, Czech, and Chinese samples, respectively).

TEST OF HYPOTHESIS

To test the hypothesis that individualism-collectivism would interact with feedback referent, we conducted a series of moderated regression analyses. For each individual, a mean individualism-collectivism composite score was used, and a country dummy variable was formed coding country as -2, 1, and 1 for the United States, Czech, and Chinese samples, respectively.3 In this analysis, self-efficacy, individual-based performance beliefs, and satisfaction with performance were regressed on demographic variables (Step 1), cultural variable and feedback main effects (Step 2), two-way culture by feedback interactions (Step 3), three-way culture by individual feedback by group feedback interaction (Step 4), and country dummy variable along with its two- and three-way interactions with individual and group feedback. The results reported in Table 3 demonstrated significant main and interaction effects for the dependent variables.

The analyses demonstrated main and interaction effects. The demographic variables accounted for relatively little variance in the dependent variables. The main effect for individual feedback consistently was related to all three dependent variables, with group feedback related to efficacy and with individualism-collectivism related to individual-based performance beliefs and satisfaction. In addition, several interaction effects were significant. A three-way interaction (individualism-collectivism by individual feedback by group feedback) was significant for both types of efficacy estimates, and the individualism-collectivism by individual feedback interaction was significant for satisfaction.

These interaction effects were analyzed further using two approaches. First, a test of simple main effects was conducted within high and low levels of individualism-collectivism (associating high levels with collectivism and low levels with individualism) using a median split and the two types of feedback with a two-way ANOVA. Second, a comparison of means using a least significant difference (LSD) test was conducted, and the results are presented in Table 4. The post hoc LSD test was conducted as a complement to the ANOVA because it is a more refined test than the omnibus F test of simple main effects. The means corresponding to this analysis are presented in Table 4, and the ANOVA results are presented in Table 5.

According to a strict sampling hypothesis, in Table 4, we would expect efficacy and related constructs to be higher in Cells 3 and 4 than in Cells 1 and 2 for individualists (high individual feedback vs. low individual feedback) and higher in Cells 6 and 8 than in Cells 5 and 7 for collectivists (high group feedback vs. low group feedback). In terms of the ANOVAs, this pattern would be reflected in significant effects for individual feedback for individualists and significant effects for group feedback for collectivists. The results of the analyses demonstrated significant main effects for individual feedback for all dependent variables across low and high levels of individualism-collectivism, demonstrating that self-efficacy, individualbased performance beliefs, and satisfaction were higher for participants receiving high individual performance feedback. In addition, group feedback was significant for self-efficacy, but only for the high collectivism subgroup. Only one of the interaction effects within the two subgroups was significant, namely, individual-based performance beliefs for the high collectivism subgroup.

The LSD analysis demonstrated that individualists' efficacy judgments were influenced by individual feedback regardless of group feedback provided (Cells 3 and 4 vs. Cells 1 and 2). Collectivists had their highest efficacy judgments given a combination of high group and high individual performance information (Cell 8 vs. Cells 5 to 7), although this difference merely approached significance (p = .07). Satisfaction with performance was influenced most heavily by individual feedback information in assessing satisfaction with performance regardless of the group feedback received. However, this contrast appears to be stronger for individualists (Cells 3 and 4 vs. Cells 1 and 2) than for collectivists (Cells 7 and 8 vs. Cells 5 and 6). Combining these results with the ANOVA results demonstrated that individual feedback plays a consistent role for both individualists and collectivists, whereas group feedback appeared to be critical only to the judgments formed by collectivists. Thus, these results provide modest support for Triandis's sampling hypothesis.

COUNTRY EFFECTS

Finally, we conducted a set of post hoc tests on the dependent variables using country of origin rather than individualism-collectivism to assess country-level differences in effects. The data were analyzed with ANOVAs conducted separately for each country using individual and group feedback as the independent variables. The means and standard deviations for these analyses are presented in Table 6, and the corresponding ANOVAs are presented in Table 7. The results of the analyses demonstrated significant main effects for individual feedback across all dependent variables for the U.S. sample and no other main or interaction effects. For the Czech sample, group feedback was significant for efficacy and individual-based performance beliefs. Group and individual feedback were significant for efficacy, and individual feedback was significant for satisfaction. No main or interaction effects were obtained for the Chinese sample for individual-based performance beliefs.

DISCUSSION

An important finding from our study is that the most direct prediction derived from current discussions on individualism-collectivism did not receive consistent empirical support. Although Bandura (1986), Markus and Kitayama (1991), and Triandis (1989), among others, posited that selfevaluations such as efficacy are shaped through maturation and socialization experiences based on cultural context, we found mixed support for a sampling hypothesis of feedback. Our findings suggest that self-efficacy beliefs are influenced most strongly by personal, and not group-based, referents for individualists. However, the expected pattern for collectivists was not obtained consistently. Collectivists' judgments of efficacy were influenced by individual and group feedback. More specifically, collectivists' judgments of efficacy were highest if they were provided either high individual or high group performance feedback. This provides mixed support for Triandis's (1989) probability-sampling hypothesis that a collectivist samples most heavily, and directly, from group-referenced information.

Simplified conceptual presentations of individualism-collectivism posit that individualists focus more on individual-based feedback and collectivists focus more on group-based feedback. It is nearly axiomatic that such feedback patterns occur. However, our results fail to support this simplified supposition. Although the anticipated main effect of feedback was observed for individualists, the picture is more complex for collectivists. Individualists appear to sample predominantly from individual-based referents, but collectivists do not sample solely from group-based referents. Therefore, a direct sampling hypothesis does not adequately capture the self-evaluations that occurred in our study.

Perhaps the most important finding from this study is that a collectivist's sense of efficacy is not derived simply from his or her group's success; rather, it is based on a unique combination of individual- and group-referenced feedback. That is, collectivists appear to develop a strong sense of efficacy if either personal or group feedback signals successful performance. This suggests that a collectivist's sense of self is based on both personal and group-based information. Although socially based information is central to a collectivist's efficacy, this information is interpreted in light of personal accomplishment. Not only does a collectivist benefit from knowing that his or her work group has been successful, but he or she needs to know about personal success as well. If we consider the welfare of a social group in a collective society, then this finding makes perfect sense because it is through personal attainment that a group will prosper. Consistent with an "invisible hand" metaphor, a collectivist's sense of self is derived through individual contribution toward collective success.

Another interesting aspect of this study concerns the satisfaction variable. Somewhat inconsistent with the performance expectations, individual feedback was the only significant effect on satisfaction with performance for collectivists. That is, regardless of group feedback, both individualists and collectivists were more satisfied after having received positive individual-level feedback than were those having received negative feedback. One might speculate that this divergence from the performance expectancies might be tied to a difference in cognition versus affect. Perhaps cognitions are more contextually sensitive to group-based cues than are affective reactions that would be associated with satisfaction.

Our findings have a number of potential implications in a managerial context. The most obvious application of these findings is that performance feedback needs to be tailored according to a person's cultural background. If we extrapolate these findings, then we would suggest that certain managerial interventions, such as performance appraisals or 360 deg feedback, have an impact and importance depending on the cultural frame an employee brings to a work setting. Of course, this is a single study using a constrained work context in which to demonstrate possible feedback effects. However, our reliance on a laboratory context has the advantage of providing a way in which to control for a number of factors that might confound or suppress the sometimes subtle effect of cultural values on organizational processes. There are other application issues that can be derived from our study as well. For example, in designing teams in various cultural contexts, a number of cautions should be exercised. Our study suggests that team-based performance appraisal and feedback is not the only way in which to motivate collectivist employees. The Western folk wisdom (i.e., stereotype) that collectivists are selfless and dominated by their in-group is not supported by our findings. Indeed, managers need to use individual-based feedback and a sense of uniqueness and recognition in this cultural context.

The present study is not without inherent limitations as a function of the methods employed. The single most serious limitation is the nature of the task used. The work performance by each participant was based on a pooled form of task interdependence (Van de Ven & Ferry, 1980). This type of task minimizes interpersonal interaction and mutual dependence. Likewise, the measures assessed in the study focus on individual-level perceptions and judgments rather than on group-level ones. Thus, the test provided for in our study was biased toward individually focused information, so it might underrepresent the significance of group-based feedback. In addition, we sampled individuals from three nations differing on at least a single cultural dimension of interest in this study. Given the obvious complexity of culture, such an approach does not fully capture its richness. This concern was addressed, to some extent, with the secondary analyses conducted using country of origin as a grouping variable. Here, the results appeared to mirror the results obtained using the cultural grouping variable.

Another limitation to this study was that the work groups used were not necessarily as psychologically salient to the participants as are other potential groups (e.g., family groupings, long-term work groups), so a stronger, group-referenced sampling effect for collectivists might have occurred in more mature groups (McGrath, 1984). Finally, our task and measures did not include a performance measure as often is the case for research on selfefficacy. We chose a task for which no demonstrably correct solutions were available because we wanted to make certain that performance referent information would not be met with suspicion due to a participant's own sense of how he or she had performed the task. The lack of a "hard" performance measure in this study is not critical, however, given the consistent and robust nature of the relationship between self-efficacy and performance that is reported in the efficacy literature (Bandura, 1986; Wood & Bandura, 1989).

More generally, what this study and others of its type suggest is that it is possible to specifically test cultural values in relation to organizational behavior. Many researchers argue that an individual's workplace identity is shaped by information based on cultural values held by employees. Our findings suggest that caution should be exercised in assuming an overly simple, cultural moderating effect for various organizational practices. A continuing view that individualists are dominated by individually referenced cues whereas collectivists are dominated by group-referenced cues appears unwarranted. That is, collectivist employees partly derive their sense of self from individually referenced cues, just as do individualist employees. As a number of researchers have argued (e.g., Sampson, 1989), a fundamental unit of awareness in a social environment is that of the self.

AUTHORS' NOTE: Requests for reprints and other correspondence should be addressed to P. Christopher Earley, Kelley School of Business, Indiana University, Bloomington, IN 47405. The authors thank John McIntyre and Amy Randel for their invaluable assistance in collecting the data as well as Dean William Pendergast and Joraj Korean of the Czechoslovak Management Center and Dou Sheng-gong at Northeastern University of the People's Republic of China for providing access to the research sites and the translations of experimental materials.

ABOUT THE AUTHORS

P Christopher Earley is the Randall L Tobias Distinguished Professor of Global Leadership at the Kelley School of Business, Indiana University. He received his PhD. in industrial and organizational psychology in 1984 from the University of Illinois-Urbana-Champaign. His research interests include intercultural and international aspects of organizational behavior. Recent books include Face, Harmony, and Social Structure: The Transplanted Executive (with Miriam Erez) and Culture, Self-Identity, and Work (with Miriam Erez).

Cristina B. Gibson, Ph.D., is an assistant professor at the Center for Effective Organizations, University of Southern California. Her research interests include communication, interaction and effectiveness in teams, the impact of culture and gender on work behavior, social cognition, and international management. She has contributed to numerous books on human resource theory and practice, and her research has appeared in various journals. In 1996, she received an extensive 4-year research grant from the National Science Foundation to investigate the implementation of teams in multinational corporations.

Chao C. Chen is an associate professor of organization management at Rutgers University. Hc received his PhD. in organization behavior and human resources from the State University of New York at Buffalo. His current research interests include cooperation, intercultural dynamics, and relationships between cultural individualism and collectivism.