Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

The development of research into cross-cultural psychology over the past several decades has involved a search for the most appropriate way in which to describe and analyse cultural differences. Measurement of respondents’ identity has played an increasingly central role in that search. The early stages of cross-cultural research most typically comprised simple comparisons of empirical results between two or more cultures, with cultures being arbitrarily defined as coexistent with national boundaries. It quickly became apparent that nations differ on a multitude of attributes, each of which might account for any differences that had been identified. By the 1970s, calls were already being formulated in favour of ‘unpackaging’ culture, in order to determine which attributes of cultural difference are most directly implicated in those variations in performance that are of interest to psychologists (Rohner, 1984; Whiting, 1976). It was soon proposed that the way in which most individuals within a given culture characterise themselves is a key element of this type. Subsequent attempts to understand cultural variations in identity have involved a continuing interplay between theorising about identity and proposals for how to measure identity in ways that are culturally valid. This chapter follows the historical sequence that has ensued, moving from research that has drawn on open-ended self-descriptions to structured surveys and experimental research.

Beginnings: The Twenty Statements Test

Bond and Cheung (1983) compared the spontaneous self-concept of students in Hong Kong, Japan and the US, using the Twenty Statements Test (TST) pioneered by Kuhn and McPartland (1954). This test asks respondents to complete 20 sentences that begin with the phrase ‘I am… ’. The responses of Japanese students included many fewer direct references to qualities of oneself (e.g., ‘I am friendly’) than the Hong Kong and US respondents. TST responses from a wide variety of nations were subsequently compared (e.g., Bochner, 1994; Triandis, McCusker, & Hui, 1990; Watkins et al., 1998). Studies of this type posed two types of problem that required resolution, if comparative studies of identity were to become fruitful. First, as the TST elicits open-ended responses, some theoretical framework is required in order to interpret the responses obtained. Second, some consideration is necessary of whether the manner in which the TST itself is formatted can be considered as culture-free.

Contexts for Identity: Individualism and Collectivism

The TST researchers were influenced by the emerging conceptual framework adopted by cross-cultural researchers, which seeks to delineate national differences in terms of a series of dimensions. The analysis of individual-level survey data aggregated to the national level by Hofstede (1980) was particularly influential. Among the dimensions first identified by him, the contrast between individualism and collectivism proved particularly attractive, perhaps because it contrasted rich Western nations with less rich nations in other regions of the world. Individualist nations were said to be those in which one’s identity is defined by one’s individual attributes and goals. Collectivist nations were said to be those in which one’s identity is more strongly defined in terms of long-lasting group memberships. Consistent with this distinction, TST researchers devised schemes for content analysis that distinguished self-descriptions in terms of individual traits and abilities from self-descriptions in terms of one’s relationships with others and membership of social entities. Initial results were encouraging. US respondents employed many more trait-like self-descriptions, whereas East Asians referred more frequently to social categories. However, as the range of locations sampled increased, it has become clear that there is no simple correspondence between the predominance of different types of TST content and nations’ positioning on Hofstede’s dimension of individualism–collectivism (Del Prado et al., 2007; Watkins et al., 1998).

Of course, this divergence may be partly due to the unreliability of Hofstede’s measures, but it seems important also to scrutinise the implicit assumptions inherent in the TST procedure. The TST protocol typically asks the respondent repeatedly to complete the stem ‘I am…’. The principal difficulty with this format is that it provides no context within which respondents can locate their response. For members of individualistic cultures, this is not especially problematic. However, collectivistic cultures are conceptualised in terms of individuals’ adherence to the norms and conventions of the groups within which they are located. A more collectivist person would therefore be expected to have difficulty in defining themselves in the absence of a specified context. Cousins (1989) tested this expectation in a study contrasting US and Japanese students. When he used the normal format, he found that US students again used more trait-like self-descriptions. However, when he adapted the TST format to specify context (e.g., ‘When with my friends, I am… . ’, ‘When at home, I am… . ’), the results were quite different. Japanese now used more trait-like descriptions, whereas the US respondents more often qualified their responses in a way that suggested that although they acted in a certain way in this setting, this was not an indication of their overall self.

This finding poses the question of how the decontextualised identities often elicited by the traditional TST procedure relate to the situated identities elicited by Cousins. Tafarodi, Lo, Yamaguchi, Lee, and Katsura (2004) addressed this question to respondents themselves. Respondents were asked whether the beliefs that they held about themselves remained the same in different situations. A total of 65% of Canadians said yes, but only 46% of Japanese and only 28% of Hong Kong Chinese did so. In a similar way, Suh (2002) found that the way that Koreans characterised themselves across five different situations was much less consistent than US responses. Thus, while all persons’ identities will change over long time periods (e.g., Kroger & Marcia, Chapter 2, this volume), persons within collectivist cultures report more short-term variation in their experienced identities as they move between different social contexts. However, variation between contexts is not the same as instability over time. English and Chen (2007) compared self-descriptions of Asian American and European American students. As expected, the Asian Americans showed greater variability in how they described themselves in differing relationship contexts. Crucially, the self-descriptions by each group showed no difference in test–retest consistency 25 weeks later. Thus, Asian Americans show greater situational variability, not greater measurement instability. English and Chen interpreted this result in terms of an ‘if-then’ model for Asian Americans (‘If I am in situation X, then I am like this’).

The greater responsiveness of identities to context in collectivist cultures is also reflected in language structures and language usage. Some languages (e.g., Arabic) do not employ the personal pronoun ‘I’. Furthermore, many of the languages spoken in collectivist nations permit pronoun drop (for instance, omission of ‘I’) from sentences (Y. Kashima & Kashima, 1998). Consequently, if TST type tests are used to study identity, their format must be modified to accommodate locally prevailing linguistic conventions. For instance, it would be better to ask respondents to list ways of describing themselves that are important to them.

These considerations indicate that if identity is to be studied validly across cultures, it needs to be addressed in ways that take full account of variations in respondents’ context. Researchers have addressed this issue in three different ways, which are considered in turn in the succeeding sections of this chapter. Construal of oneself in terms of concepts explicitly derived from individualism is considered first. Alternative bases of self-construal such as hierarchical position and relatedness are then examined, followed by studies in which self-construal is manipulated experimentally.

Independent Versus Interdependent Self-Construal

The concepts of independence and interdependence were first popularised by Markus and Kitayama (1991). These authors proposed that Americans typically construe themselves as relatively autonomous individuals, while Japanese typically construe themselves as interdependent with the membership groups within which they are embedded. As used by Markus and Kitayama, these terms are conceptually parallel to individualism and collectivism, and some authors have used them interchangeably. However, it is preferable to use individualism and collectivism to describe the culture of large-scale entities such as nations, and independence–interdependence to describe individuals’ self-construals. Cultures are characterised by the interrelatedness of their various components and are consequently more than the simple aggregate of the individuals within them. Measures at the two levels of analysis will therefore sometimes have differing structures (Hofstede, 1980; Schwartz, 1994; Smith, Bond, & Kağıtçıbaşı, 2006).

Markus and Kitayama used their literature review as the basis for a series of propositions as to the consequences of independent versus interdependent self-construal for processes such as cognition, emotion and motivation. Although they did not publish measures of self-construal, their influential formulation has provoked others to develop such measures. The most widespread procedures have entailed the creation of self-report measures using Likert scales. Singelis (1994) created a 30-item survey, comprising separate measures of independent and interdependent self-construal. The Singelis scales provide some advance on measures based on the TST, because the items tapping interdependence refer explicitly to the respondent’s relatedness to others. For instance, one interdependence item reads: ‘It is important to me to maintain harmony within my group’. Another reads: ‘I will stay in a group if they need me, even when I am not happy with the group’. However, it is notable that one of these items describes a value, while the other describes a behaviour. Few of the items refer explicitly to the respondent’s identity, but some do touch closely on issues exposed by the studies discussed above. For instance, one of the independence items reads: ‘I am the same person at home as I am at school’. Another reads: ‘My personal identity independent of others is important to me’. These scales have been used frequently by cross-cultural researchers, and Singelis (1994) has been cited nearly 500 times. However, perhaps because of the heterogeneity of the scale items, they do not always achieve adequate levels of internal consistency. Critics have identified multifactorial solutions (Hardin, Leong, & Bhagwat, 2004; Levine et al., 2003). Studies that have compared TST responses with the Singelis scales have found only weak correlations between them (Bresnahan et al., 2005; Del Prado et al., 2007). Thus, the question of which procedure is preferable must rest on their ability to show meaningful associations with other indices.

Self-Construals as Predictors

Scores on the Singelis self-construal scales have been shown to relate in plausible ways to other measures of how persons think about themselves. At the level most directly relevant to the focus of this chapter, there is evidence that self-construal is linked with identification with one’s nationality. Using a measure similar to the Singelis scales, Jetten, Postmes, and McAuliffe (2002) found that among American students those who identified more with being American scored higher on independence, whereas among Indonesian students those who identified more with being Indonesian scored higher on interdependence. In further studies, these authors showed that when respondents were encouraged to identify with a group that had an individualistic or a collectivistic culture, their self-construals became more independent and interdependent, respectively. Thus, self-construal can be a function of the groups that one associates with, rather than a stable trait-like quality. It is perhaps paradoxical that identifying strongly with an individualistic group leads to construing oneself as more independent.

Self-construals also significantly predict whether respondents believe that personality traits or social context are the best predictors of behaviour. Church et al. (2006) found that American respondents scored significantly higher on belief in traits as causal and significantly lower on contextual beliefs than Malays, Mexicans, Asian Australians, Filipinos and Japanese. Among US respondents, independent self-construal was correlated with trait beliefs whereas interdependent self-construal was correlated with contextual beliefs. However, the results from other nations were confusing. For instance, in Japan, independence predicted contextual beliefs and interdependence predicted trait beliefs. Some of this confusion may be due to the fact that the Singelis scales do not control for cultural differences in acquiescent responding, in other words, the tendency of some respondents to agree with all the items in a survey.

The next several paragraphs briefly describe the broad range of cross-cultural studies of social emotions and behaviours in which the Singelis measures have been used to explain national differences. In these studies, differences between samples are first identified in mean self-construal scores and in means on the variable of interest. The effect of self-construal on the variable being studied is then discounted statistically. If the difference between the nations in adjusted means is reduced or entirely eliminated, self-construal is said to partially or wholly account for cross-national differences of interest. For instance, Singelis, Bond, Sharkey, and Lai (1999) compared respondents’ ratings of how embarrassed they would be in each of a set of scenarios with which they were presented. Samples were from Hong Kong, Hawaii and mainland US. Mainland US respondents scored significantly higher on independence and significantly lower on interdependence than members of the other samples. These measures partially accounted for group differences in ratings of one’s embarrassability, both at the level of differences between the different national samples, and also between ethnic groups within each sample separately. In a similar way, Oetzel et al. (2001) compared two types of concerns about loss of face in Germany, China, Japan and the US. Independent self-construal was found to explain differences in the level of concern about one’s own loss of face, while differences in concerns about loss of face by others with whom one is interacting were explained by interdependent self-construal.

Kwan, Bond, and Singelis (1997) reported that the effect of relationship harmony on life satisfaction was explained by scores on interdependent self-construal, both in Hong Kong and in the US. Across Korea, Japan, Hawaii and mainland US, Kim et al. (1996) compared favoured forms of communication style, which they referred to as constraints. Differences in endorsement of a task constraint (in other words, a belief that one should focus on the task) by respondents from Korea, Japan, Hawaii and mainland US were accounted for by independent self-construal, whereas sample differences in endorsement of a relationship harmony constraint (preference for focusing on good relationships) were accounted for by interdependent self-construal.

One of the most striking sets of this type of results has been provided by Earley (1993), who compared social loafing in simulated work groups in China, Israel and the US. The Chinese and Israelis worked harder when they believed that they were working in a team with whom they had affinity. Performance in a team with whom they had no affinity did not differ from performance when working individually. Americans worked harder when they believed that they were working alone. The type of team in which they were working had no differential effect on performance. The differences in work levels between the samples from different nations were fully explained by measures of independent and interdependent self-construal.1

These results indicate that an increasing range of authors have found that self-construal measures can explain cross-national differences of interest. However, not all studies have found such effects and mean national differences in self-construal scores are sometimes absent or in non-predicted directions (Matsumoto, 1999). Consequently, there is continuing confusion as to the circumstances in which self-construal measures may be validly employed. In an influential review, Oyserman, Coon, and Kemmelmeier (2002) compared scores across nations on the Singelis scale, as well as on eight other scales that had been defined by their authors as measuring independence–interdependence or the related concepts of individualism–collectivism, as applied to individuals. Oyserman et al. concluded that the means of the various measures included in their meta-analysis indicated that European Americans are more individualistic than persons from other nations, and less collectivistic than Chinese, but not less than Japanese or Koreans. These conclusions have been challenged on methodological grounds. None of the scales included in this meta-analysis included balanced sets of positively and negatively worded items. Consequently, they are vulnerable to the risk of acquiescent response bias. Acquiescence is known to differ consistently across nations (Smith, 2004). Oyserman et al.’s comparison of means is therefore as likely to have detected the incidence of acquiescence as the incidence of independence or interdependence. Schimmack, Oishi, and Diener (2005) reanalysed a sub-sample of Oyserman et al.’s data, comprising those scores for which it was possible also to include a control for acquiescence. Their analysis now showed that the data confirmed the contrasts between nations that had been first identified by Hofstede (1980). Respondents in individualistic nations do predominantly construe themselves in independent ways, whereas respondents in more collectivistic nations predominantly construe themselves in interdependent ways.

The differential prevalence of these two types of self-construal across cultures has also been addressed by developmental researchers. Parental models of infant care in Germany, Greece, China, Mexico, India, Costa Rica, the US and Cameroon show wide variations (Keller et al., 2006). These authors have shown that a measure of family allocentrism (equivalent to interdependence) accounted for national differences in mothers’ models of parenting. Middle-class mothers in the US, Germany and Greece favour parenting that emphasises the development of independence. Mothers from rural areas in India and Cameroon favour parenting that emphasises the development of relatedness. Urban women in the remaining samples favour an intermediate form of parenting that was described as autonomous–relational. Earlier studies by this group have shown that differing models of parenting are associated with observed differences, such as frequency of body contact, object stimulation, holding and smiling, in parental behaviours towards infants (Keller, Borke, Yovsi, Lohaus, & Jensen, 2007; Keller et al., 2003).

The development of measures of independent and interdependent self-construal has benefited the field, because it has enabled some fruitful attempts at the unpackaging of key elements within very broad and very complex concepts such as that of national culture. No one would propose that independence and interdependence make up the sum total of ways in which individuals define their identity, but this specific contrast has been particularly helpful to cross-cultural psychologists, because it parallels the dimension of cultural variation that has so far been most fully investigated: individualism–collectivism. These concepts are also valuable precisely because they can be measured at the individual level. Nation-level contrasts are likely to prove adequately interpretable only in studies that have sampled 20 or 30 nations. Practical constraints determine that most cross-cultural studies can span no more than a handful of nations. Moreover, the populations of nations are by no means homogeneous. Thus, although it is the case that the majority of persons within a nation such as China will be found to exemplify interdependent self-construal, even within a sample drawn from a more individualistic nation, some persons will be identified who also exemplify interdependent self-construal. This is illustrated, for example, by the way Singelis et al. (1999) found that interdependence could explain differences not only between samples but also between ethnic groups within each sample in their study of embarrassment. Because this level of diversity exists in most nations, studies sampling only a few nations can still contribute to the current progress of cross-cultural investigation. A greater problem at the present time is that it has tended to be the same few nations that have been repeatedly sampled. We need to be sure that the full range of ways in which persons construe themselves is being sampled.

Perhaps the most striking aspect of the self-construal measures thus far discussed is that they treat persons’ identity simply as a stable quality acquired through socialisation, which can subsequently guide our understanding of their emotions and behaviours on particular occasions. Self-construal is taken as consciously accessible and as capable of summary through a single, contrasting pair of concepts. The following sections discuss studies that have sought to broaden the scope of these measures and to allow for their temporal variability.

Additional Dimensions of Self-Construal

The studies outlined in the preceding section focused on the extent to which respondents defined their identity in terms that are associated with Hofstede’s contrast between individualism versus collectivism and Markus and Kitayama’s parallel distinction between independence and interdependence.

Focus on Hierarchy

In addition to his focus on individualism–collectivism, Hofstede (1980) identified three further dimensions of cultural variation. In principle, each of these could also provide a basis for identifying variations in how persons construe themselves. His dimension of power distance concerns the extent to which a culture is organised on the basis of hierarchy. It can be expected to differentiate those who construe themselves as equal to others (low power distance) from those who see themselves in terms of either submission or dominance (high power distance). An individual-level self-construal measure addressing these types of distinctiveness was devised by Singelis, Triandis, Bhawuk, and Gelfand (1995), comprising items describing the extent to which one’s relations with others were based on equality (‘horizontal relationships’) or on hierarchy (‘vertical relationships’). The items in this measure referred also to independence and interdependence (termed individualism and collectivism by Singelis et al.). Thus, the measure has four scales: vertical individualism, horizontal individualism, vertical collectivism and horizontal collectivism. These measures have also been shown to explain cross-national differences found in some studies. For instance, Thomas and Au (2002) found that horizontal individualism explained the stronger effects of job dissatisfaction and the availability of alternatives on intention to leave one’s job that they found in New Zealand compared to Hong Kong. High power distance and collectivism are strongly correlated, at least at the nation level (Hofstede, 2001). In other words, collectivistic nations or cultures are frequently the more hierarchical ones. Consequently, it is possible that the creation of four dimensions of self-construal over Singelis’ (1994) previous two scales has not enhanced predictive validity. No study has yet made a direct comparison.

Focus on Relatedness

Hofstede’s concept of collectivism has been interpreted in a variety of divergent ways, ranging from cultures characterised by life-long identification with a single group to cultures characterised by a generalised affinity for working in groups. Much less attention has been given to the dimension of cultural variation that he named as masculinity versus femininity. Most probably, this is because his labelling of this dimension has been interpreted as sexist. He defined masculine cultures as those in which persons strive for achievement and recognition, and feminine cultures as those in which greater priority is given to enhancing the quality of interpersonal relationships. Thus, Hofstede’s definition of individualism–collectivism rests on variations in attachment to groups, while his definition of masculinity–femininity rests on variations in relatedness to specific others.

These two bases for defining cultural variation are both important in considering the individual-level concept of interdependent self-construal (Smith & Long, 2006). Some of the items in Singelis’ (1994) scales refer to ‘my group’, while others refer to relations with specific persons such as one’s professor and one’s parents, and yet others specify generalised attitudes towards other persons. Responses to these varied items may go well together in some cultural contexts, as they did in Singelis’ original study. In other contexts, it is likely that the priority given to relations with a long-term in-group and relations with persons from other groups will not be closely associated. These variations may contribute to the low reliability that has been found for the Singelis scales in many subsequent studies. Georgas, Berry, van de Vijver, Kağıtçıbaşı, and Poortinga (2005) used the Singelis items in a study spanning 30 nations, and found that the reliabilities varied so greatly that the data from this scale could not be used in their main analyses.

Brewer and Gardner (1996) proposed that the concept of interdependence, as formulated by Markus and Kitayama (1991), is primarily focused on relatedness with other individuals. In their view, it was preferable to distinguish this concept from collectivism, which has more to do with one’s relation to specific groups or other social entities. Personal, relational and collective identities should therefore be distinguished. Following this initiative, Brewer and Chen (2007) made a content analysis of the items comprising all available scales that have been influenced by the concepts of individualism and collectivism. They conclude that it is desirable to make a distinction between items that refer to what they call relational collectivism and those that refer to group collectivism (Sedikides & Brewer, 2001; Chen, Boucher, & Kraus, Chapter 7, this volume; Spears, Chapter 9, this volume). This distinction holds great promise, since it may help to clarify the meaning of some of the more puzzling results in the existing literature. For instance, some researchers have reported the finding that their US respondents endorsed interdependence more strongly than their Japanese respondents (Matsumoto, 1999). This could be because US respondents are more collective and Japanese more relational. In comparing Japan and the US, Yamagishi and Yamagishi (1998) have proposed that Japanese are more often preoccupied with the assurance provided by the state of relationships within their in-group, whereas the more fluid nature of US culture encourages a stronger focus on the status of relations between one’s own group and other groups. Empirical evidence supports this formulation (Takemura, Yuki & Ohtsubo, 2010; Yuki, 2003). Whether this particular reasoning is correct or not, it is important to explore more fully the utility of scales that distinguish relational interdependence from collective interdependence.

E. S. Kashima and Hardie (2000) developed three 10-item scales in Australia tapping personal, relational and collective orientations. All items were positively worded, and the three scales were found to correlate positively with one another. When each of the other scale scores was partialled out, the personal and collective scales were shown to link in predictable ways with the self-construal measures discussed in the preceding sections. After partialling out the other two scales, the relational scale correlated only with a measure of attachment closeness.

Del Prado et al. (2007) included in their six-nation survey an Aspects of Identity questionnaire, which was devised in the US by Cheek and Tropp (1997; see also Cheek, Smith, & Tropp, 2002). This survey distinguishes between personal (‘my personal values and moral standards’), relational (‘My relations with people I feel close to’), social (‘My reputation’) and collective (‘My race or ethnic background’) forms of identity. Respondents were asked to rate items concerning the importance to them of each of these identities. Del Prado et al. tested the ability of the Singelis measures of independence and interdependence to predict each type of identity in each of four nations. In the two individualistic nations, the US and Australia, independence predicted personal and relational identities and interdependence predicted social and collective identities. In Mexico, independence predicted personal and relational identities and interdependence predicted collective and relational identities. However, in the Philippines, independence predicted the importance of all four forms of identity.

The results obtained by E. S. Kashima and Hardie and by Del Prado et al. both underline the need to control for acquiescent responding, particularly in collectivist cultures such as the Philippines, where it is more prevalent. Until this has been achieved, it is difficult to interpret the variations in the results that have been obtained. A clear separation between measures of personal identity and relational identity in differing cultural contexts has not yet been achieved.

Focus on Category Inclusiveness

An alternative approach to this problem was attempted by Harb and Smith (2008). An instrument was constructed that asks about the respondent’s degree of involvement at four different levels of social inclusiveness, labelled as personal, relational, collective and humanity-as-a-whole. Relational categories are defined as those that involve dyadic relations, or an interconnected set of dyadic relationships (e.g., ‘friends’). Collective categories are defined as those in which the individual is an interchangeable exemplar of a larger scale social category (e.g., ‘students at my university’). Following the inclusion by Singelis et al. (1995) of items referring to hierarchical relations, a distinction is also made between vertical and horizontal relationships within the relational and collective levels, making six dimensions in total. Harb and Smith selected entities which best represent each of these categories and asked students from four nations to complete five Likert scales describing their involvement in each of the six social categories. Confirmatory factor analyses supported the retention of the six separate self-construal indices. Each type of self-construal was found to be significantly related to measures of identification with the equivalent social category, and inclusion of others in oneself, and with endorsement of a distinctive profile derived from Schwartz’s (1992) values survey. For instance, horizontal relatedness predicted identification with friends, inclusion of friends in one’s self and endorsement of benevolence, stimulation and hedonism values. The sample was drawn from the UK and three Arab nations. UK students scored significantly higher on personal and on relational-horizontal self-construal. Syrian students scored significantly higher on collective-horizontal, collective-vertical and relational-vertical self-construal. Students in Jordan and Lebanon had intermediate scores. This procedure does succeed in providing a more clearly differentiated set of self-construals, but its validity rests on selection of adequately distinctive exemplars for each of the relational and collective categories. An alternative possibility is that persons might be able to construe the same exemplar in different ways, in which case a measure would be required that differentiates styles of construal rather than targets of construal.

Focus on Agency

Each of the projects discussed in the preceding section has sought to achieve separation between relatedness and other dimensions of self-construal. Kağıtçıbaşı (2005) has argued that the reason why this has proved difficult is that the concept of independence–interdependence, as usually defined, includes two quite separate dimensions of self-construal. She identifies these as a dimension of interpersonal distance, named as separation versus relatedness, and a dimension of agency, named as autonomy versus heteronomy. Drawing on her earlier studies of parenting (Kağıtçıbaşı, 1996), she defines autonomy as ‘a state of being a self-governing agent’ (Kağıtçıbaşı, 2005, p. 404), which places her view close to that of self-determination theorists (see Soenens & Vansteenkiste, Chapter 17, this volume). Heteronomy is defined as reliance on others as a source of guidance. In terms of these dimensions, independent self-construal is characteristic of persons who construe himself or herself as high on separation from others and high on agency. Interdependent self-construal would be characteristic of a person who construes themselves as high on relatedness and high on heteronomy. The distinction between these two dimensions permits the formulation of two further types of self-construal. Autonomous-relational self-construal would characterise persons high on both autonomy and relatedness. Such a person would be an active initiator of actions, while retaining membership with a cohesive network of relatedness. Kağıtçıbaşı (2005) does not discuss the fourth possible type, which would entail heteronomy and separateness. However, she cites evidence supporting her contention that autonomous-relational self-construal is characteristic of urban populations within collectivist cultures, in contrast to rural populations in collectivist cultures, who would be more likely to show heteronomous-relational self-construal. Kağıtçıbaşı has yet to publish results of measures employing her concepts, but it could be predicted that measures of autonomous-relational self-construal would be correlated with Harb and Smith’s measure of horizontal collectivism, while her measure of heteronomous-relational self-construal would link with their measure of vertical collectivism.

Basic Problems: Theory and Method

The measures proposed as alternatives to a simple contrast between independence and interdependence enrich our understanding of self-construal, by identifying a fuller range of ways in which persons in differing cultural contexts can choose to identify themselves. Some of this diversity was already apparent from factor analyses of data using the original Singelis (1994) items (Hardin et al., 2004; Levine et al., 2003). However, the more recently devised measures entail more explicit theorising about the range of ways in which self-construals may vary.

Despite this, the recent approaches share with the Singelis scales the weaknesses that are present in any measure that asks respondents to characterise themselves on a series of Likert scales. The principal weakness is that the way that they position themselves on these scales is implicitly comparative. However, we do not know what comparators the respondent will have employed. Judgement could be made relative to one’s own internal aspirations, or relative to those within one’s immediate context, or relative to some salient reference group. It is unlikely that the judgements would be made relative to one’s image of persons from other nations. Consequently, identity measures collected from different nations or different cultural groups may be biased. A member of a collectivist culture may rate himself or herself as highly independent relative to those around him or her, but still be much more interdependent than most members of an individualistic culture. Effects of this kind could explain failures to find predicted differences in mean self-construals between people from different cultures.

This line of reasoning was investigated by Heine, Lehman, Peng, and Greenholz (2002). Using the Singelis scales, they showed that direct comparisons of mean responses from Canada and Japan did not differ. However, when respondents familiar with both cultures were each asked to complete two modified versions of the Singelis items reading ‘Compared to most North Americans, I am… ’ and ‘Compared to most Japanese, I am… ’, the predicted effects were found. This procedure brings into play the stereotypes that members hold about their fellow-nationals and about the other cultural group, but it does not ensure that the resulting data are necessarily more valid, because a frame of reference has been imposed which may not be the respondent’s preferred frame of reference.

An implication of Heine et al.’s critique is that measures of self-construal must either contain explicit scale anchors, or else that they should be used in ways that involve intra-cultural, or better still intrapersonal, data analyses rather than comparing mean levels across cultures. For instance, studies cited earlier such as Kwan et al.’s (1997) study of life satisfaction and Earley’s (1993) analysis of social loafing utilised a series of parallel within-subject hypothesis tests for each cultural group that was sampled. Another instance of this type is provided by the work of Vignoles (Chapter 18, this volume), whose cross-cultural analyses of identity motives focus on within-participant variance across multiple elements of identity.

A much more radical way of addressing the problem stems from the original position adopted by Markus and Kitayama (1991). They did not seek to measure self-construals directly at all, choosing instead to test hypotheses predicting how participants would respond to a variety of tasks, based upon the premise that independence and interdependence pervade particular cultural groups. This position has been explored fruitfully in recent years (Kitayama, Ishii, Imada, Takemura, & Ramaswamy, 2006; Kitayama, Park, Sevincer, Karasawa, & Uskul, 2009), but does not advance our understanding of the nature of self-construal itself. A final way to address the need for within-subject analyses is provided by studies that employ experimental priming. These are considered in the next section.

Experimental Approaches

The approaches to the cross-cultural study of identity that have been discussed in preceding sections treat identity as a relatively stable attribute. Persons are seen as having been socialised to think of themselves in ways that are to a substantial degree compatible with the cultural milieu in which they are located. However, we have abundant evidence from research in social psychology that persons are typically aware of a range of identities, any of which may be elicited by momentary events (e.g., Tajfel & Turner, 1979; Turner, Hogg, Oakes, Reicher, & Wetherell, 1987; for a review, see Spears, Chapter 9, this volume). From this perspective, cultural differences in identity must be thought of as the predominance of that particular set of identities that are frequently elicited by life within the settings that make up a given culture. For instance, Kitayama, Markus, Matsumoto, and Norasakkunkit (1997) show that US cultural settings include many that encourage individuals to feel proud and self-enhancing, whereas Japanese cultural settings include many that encourage individuals to feel modest and self-critical. Several types of priming may be involved.

Language as a Prime

Spoken language is an instance of a constantly recurring cultural prime. As noted earlier, languages that can drop the first person pronoun are more prevalent in collectivist nations (Y. Kashima & Kashima, 1998). Repeatedly speaking in a way that does not require the personal pronoun can be expected at the least to predispose against thinking of oneself as agentic. In bilingual contexts, the choice between spoken languages is frequently an important marker of cultural identity in a given setting (Noels, Clément, & Gaudet, 2004) and of consequent actions that accord with that setting. Comparative studies have shown that among bilinguals, the language in which a survey is completed affects responses. Across 24 nations, respondents who completed a survey in English answered in ways that were closer to the answers by Caucasian respondents than were those of respondents answering in their first language (Harzing, 2005). Sanchez Burks, Lee, Choi, Nisbett, Zhao, and Koo (2003) explored the cross-cultural implications of the finding that North Americans feel that work and personal relations should be kept separate, whereas those from other parts of the world see work and non-work as more closely interwoven. They showed that when Thai–English bilinguals responded to scenarios concerning work difficulties in English, they took no account of personal aspects of the situation, whereas when they responded to the same scenarios in Thai they did take account of personal relationship issues. Thus, the language that was used primed a particular cultural orientation.

Priming Independence/Interdependence

Language of response provides an implicit cultural prime, but researchers have increasingly also employed a range of other primes related to the concepts of individualism and collectivism, some explicit and some implicit (Oyserman & Lee, 2007, 2008). Of the 67 cultural priming studies identified by Oyserman and Lee, only eight were conducted in more than one nation. Priming has been found to show modest effects on various measures of values and self-concept and larger effects on measures of cognition. An early instance related to priming of self-concept is provided by the work of Trafimow, Triandis and Goto (1991). These authors asked students to spend 2 min thinking either about all the things that made them different from others, or about all the things that they had in common with close others. They were then asked to complete the TST. Those who had spent time thinking about differences scored more highly on statements about their personal self, while those who had thought about similarities scored higher on collective self-representations.

Priming of Biculturals

Priming studies have frequently involved samples of bicultural respondents (see Huynh, Nguyen, & Benet-Martinez, Chapter 35, this volume). For instance, respondents may be asked to evaluate cultural icons such as the Statue of Liberty, the Eiffel Tower or a Chinese dragon, or to rate culturally distinctive advertisements, prior to completing an experimental task (Hong, Morris, Chiu, & Benet-Martinez, 2000). In studies of this type, no direct measures of cultural identity are typically collected. Identity is treated as the hypothesised causal variable accounting for the culturally distinctive effects that are obtained. For instance, Verkuyten and Pouliasi (2002) primed Greek children living in the Netherlands to respond to a survey either in Greek or in Dutch. They were also shown icons such as the national flag and asked for their reactions. Those responding to the Greek primes reported significantly stronger identification with their friends, a more positive social identity, a less positive personal identity and more external attributions for events. These differences are all strongly in the direction of differences that were found between separate control groups of monocultural Greek and monocultural Dutch children. Thus, the primes elicited separate sets of schemata relating to both self-description and description of events whose difference in magnitude was almost as great as the differences found between the two separate monocultural groups. This suggests that cultural priming studies have considerable potential for explaining cultural differences, at least among biculturals. However, if we are to be clear that the effects obtained through priming are attributable to elicitation of specific cultural identities, then direct measurement of identities is preferable. The capacity of identity measures to mediate the effects of cultural priming manipulations could then be tested, just as has been the case in studies using self-construal measures.

National identity is but one of many identities available to an individual, and is one that is much less likely to be elicited in everyday interactions than are those identities that are more proximate. Thus, if priming of identities is to assist our understanding of cultural differences, it is necessary to establish the degree of association between the cultural icons used in priming studies and the cognitive structures thought to be characteristic of members of a given culture. Wan et al. (2007) have provided some initial indications of links between cultural identification and preferred ways of characterising cultures. In three studies, they showed that identification with one’s nation was significantly associated with personal endorsement of values that were perceived by the sample as a whole to be more characteristic of one’s nation (see also Jetten et al., 2002; Schildkraut, Chapter 36, this volume). These effects were replicated when assessing also identification with subsidiary levels of grouping such as identification with one’s university (Wan, Chiu, Peng, & Tam, 2007). Thus, as social identity theorists would predict, cultural identification involves internalisation of a prototype that characterises one’s culture collectively, rather than a simple matching of one’s personal values with those around oneself.

How Do Priming Studies Advance the Field?

It is likely that the cultural differences that have mostly been categorised in terms of simple contrasts between individualism and collectivism entail a whole range of ways of thinking about oneself and others that are differentially accessible to members of differing cultural groups. Some of these may be associated with ways of processing information relevant to one’s cultural identity, whereas others are likely to be associated with ways of processing information relevant to one’s participation in the varying groups and activities that make up one’s day-to-day life. It remains to be determined whether we gain greater benefit from retaining more global concepts such as individualism–collectivism or from differentiating a more semiotic perspective that addresses the whole range of identities espoused by members of a given culture (Y. Kashima, 2009; Vignoles, Chapter 18, this volume).

To choose between these alternatives, we need more evidence of the relationships between different types of priming effects, self-construal and identity. Sui, Zhu and Chiu (2007) made content analyses of self-descriptions by mainland Chinese students in Beijing who had been exposed to a prime comprising either Chinese cultural icons or US cultural icons. The Chinese prime elicited more interdependent self-descriptions, whereas the US prime elicited more independent self-descriptions. This effect could equally be due to the elicitation of self-schemata or the elicitation of knowledge about China and the US. In a subsequent study using the same primes, students undertook a memory test. Those receiving the Chinese prime (and those receiving no prime) were better at remembering words related to their mother than were those receiving the US prime. Thus, the US prime hindered Chinese students’ memory performance in a way that is consistent with thinking of oneself as independent. This result clearly implies that priming achieves its effect through the elicitation of self-schemata, rather than the elicitation of cultural knowledge. Motherhood is much more relevant to self-schemata than it is to knowledge of different cultures.

Ng and Han (2009) used a similar procedure to that of Sui et al., but also made fMRI brain scans of 15 mainland Chinese and Hong Kong students. The scans of those receiving the Western prime were subtracted electronically from the scans of those receiving the Chinese prime, in order to reveal the areas of brain activity that differentiated the two experimental conditions. The results were interpreted as showing that, for those with a Chinese prime, the memory tasks involving self and mother activated the same area in the ventral medial prefrontal frontal cortex. However, for those receiving the Western prime, each of the two tasks elicited activation in a separate area of the brain. Thus, there is preliminary evidence that priming effects are interpretable in terms of differing patterns of brain activation that are consistent with the contrast between independence and interdependence.

These studies leave open whether priming effects of these types would always elicit identification with the cultural group that is primed. Using cultural icons as primes could elicit a wide range of reactions, not just independence versus interdependence. Not all Chinese are pro-American, nor are all Americans pro-Chinese. Even among biculturals, Zou, Morris, and Benet-Martinez (2008) question whether culture primes necessarily elicit identification. Biculturals may be ambivalent or indifferent towards one or the other of their available cultural identities, leading to varying levels of identity integration (Huynh, et al., Chapter 35, this volume). Consequently, in some circumstances, priming could lead to effects associated with disidentification. Zou et al. showed that, among Chinese and US students, measures of identification and disidentification with one’s nation were distinct from one another. Among US respondents, a US prime interacted with high US identification leading to enhancement of the typically American tendency to attribute causes to individuals. However, among Chinese respondents, a Chinese prime interacted with high Chinese disidentification, leading to a reduction in the typically Chinese tendency to attribute causes to groups. Thus, in this case, priming and identification achieved their impact interactively, not as main effects.

Lechuga and Wiebe (2009) found that a Spanish language prime increased the reported interdependence of bicultural Hispanics, but also increased their identification with US culture. Thus, in this case, priming affected self-construal and identification in apparently opposing directions. This could be because the complexity of elements comprised within a language prime elicits multiple effects. If primes are to illuminate the causal processes relating to self-construals, they may need to be structured in more precisely theory-driven ways than are provided by language or cultural icons.

The results of priming studies raise a further issue that has not yet been addressed. Most such studies have employed bicultural respondents. Biculturals are by definition likely to have accessible a range of cultural identities, whether these be integrated with one another or not. These are readily available to experimenters, but what implications do the results of such studies have for the broader field of cross-cultural comparisons? One could argue that we are all biculturals, indeed multiculturals, in consideration of the range of multiple identities that social identity theorists have identified. Studies employing monoculturals have certainly yielded significant effects on measures of self-construal that relate to cultural difference (e.g., Ng & Han, 2009; Sui et al., 2007; Trafimow et al., 1991), but we do not know whether these significant effects are of similar magnitude to that which is found between the self-construals of equivalent monocultural populations. Neither do we know whether the ingenuity of experimenters can devise primes that will elicit the more fine-tuned variations in self-construal that survey researchers have begun to identify.

Conclusion

There is a paradox that is central to the study of culture and identity. In cultural groups that are relatively homogeneous, members may only rarely think of themselves in terms of their national or cultural identity. However, cross-cultural psychologists have mostly continued to treat nations as distinguishable cultures. In analysing cultures conceptualised at this macroscopic level, there is a compelling need for explanatory organising concepts that can identify key elements in such overcomplex entities. Survey measures of self-construal and experimental cultural priming are two of the stronger current candidates for this task. Each has its strengths and weaknesses. There have been continuing problems in creating valid ways of measuring self-construal, partly on account of cultural differences in response style, and partly because of difficulty in defining the comparison group to which a respondent’s ratings might relate. The range of respondents’ available identities also means that rating scales may themselves prime respondents in unpredictable ways. Nonetheless, the studies showing mediation of cultural differences by self-construal measures have successfully narrowed the range of available explanations for a wide variety of identified cultural differences.

The experimental basis of priming studies offers the prospect of more firmly established causal explanations. However, global ‘Western’ and ‘Eastern’ primes are not likely to capture the finer detail of existing cultural differences, and are most readily applicable among bicultural populations that have well-developed alternative systems of construing the world. In order to understand better the impact of priming, it will still be necessary to measure more fully the impact of potential intervening variables such as identification, which raises again the difficulties of measurement associated with the assessment of self-construal.

Cultural identity becomes most salient among biculturals, and among the increasing number of persons who are tourists, sojourners, expatriates and immigrants within the contemporary world (see also Jensen, Arnett, & Mackenzie, Chapter 13, this volume). Acculturation psychology has become a major field of investigation (Berry, Phinney, Sam, & Vedder, 2006; Sam & Berry, 2006). These areas of study provide valuable information concerning the impact of numerous variables related to successful acculturation. They also underline the value of treating identification with one’s new culture and identification with one’s culture of origin as separate from one another. However, with the exception of some approaches discussed within the present volume (Huynh, et al., Chapter 35, this volume), this literature does not suggest ways of analysing relations between self and cultural identity additional to those that have been explored within the present chapter. With the continuing global intermingling of cultural groups, we can anticipate a steady increase in salience of both bicultural identities and multiple identities.

Note

  1. 1.

    Earley used the terms individualism and collectivism, but when used to characterise individuals rather than cultures, measures of individualism and collectivism are similar to measures of independence and interdependence.