You are on page 1of 14

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/232532870

Differences between Taiwanese and U.S.


cultural beliefs about ideal adult attachment

ARTICLE in JOURNAL OF COUNSELING PSYCHOLOGY MARCH 2006


Impact Factor: 3.23 DOI: 10.1037/0022-0167.53.2.192

CITATIONS READS

40 256

2 AUTHORS, INCLUDING:

Brent Mallinckrodt
University of Tennessee
101 PUBLICATIONS 3,776 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Brent Mallinckrodt
letting you access and read them immediately. Retrieved on: 29 February 2016
Journal of Counseling Psychology Copyright 2006 by the American Psychological Association
2006, Vol. 53, No. 2, 192204 0022-0167/06/$12.00 DOI: 10.1037/0022-0167.53.2.192

Differences Between Taiwanese and U.S. Cultural Beliefs About Ideal


Adult Attachment
Chia-Chih DC Wang and Brent S. Mallinckrodt
University of MissouriColumbia

Some researchers believe that important tenets of attachment theory are culturally universal, whereas
others claim that key constructs are rooted in Western values and should not be generalized further. To
explore possible cultural differences in adults, undergraduates from Taiwan (n 280) and the United
States (n 268) were asked in the present study to complete a self-report measure of adult attachment,
the Experiences in Close Relationships Scale (K. A. Brennan, C. L. Clark, & P. R. Shaver, 1998), as they
believed an ideally emotionally and psychologically healthy person of your own gender in your culture
would respond. Findings suggested significant differences by cultural group, gender, and Gender
Culture interactions. Taiwanese women and men endorsed more avoidance in beliefs about ideal adult
attachment than their U.S. counterparts, and Taiwanese men endorsed more anxiety than U.S. men. These
cultural differences were not explained by group differences in independent and interdependent self-
construal.

Keywords: adult attachment, cross-cultural comparison, gender differences, individualism, collectivism

Attachment theory has been described as one of the most influ- infants or children classified in the three attachment categories
ential conceptual frameworks guiding contemporary social devel- Ainsworth and her colleagues identified. Findings of these studies
opment research (Cassidy & Shaver, 1999). However, as attach- generally indicated that the same core aspects of secure attachment
ment theory has received attention from an increasingly broader developed by an infant with primary caregivers can be observed
range of researchers, the universality of key attachment theory across different cultures and that infants and children with secure
concepts and the appropriateness of applying the theory to non- attachment appear to be the numerical majority in widely differing
Western cultures have become the focus of a growing controversy. cultures (van IJzendoorn & Sagi, 1999). Many contemporary at-
Bowlby (1969) maintained that the core hypotheses of attachment tachment researchers believe that the core components of attach-
theory are culturally universal and apply to all members of the ment theory are culturally universal, while acknowledging that
human species. Claims of cultural universality were given initial specific attachment behaviors may be expressed differently across
empirical support when results of observational attachment studies cultural contexts (Grossmann, Grossmann, & Kepler, 2005;
were replicated in samples of motherinfant dyads from Uganda Posada & Jacobs, 2001).
and the United States (Ainsworth, 1967; Ainsworth, Blehar, Wa- However, Rothbaum, Weisz, Pott, Miyake, and Morelli (2000)
ters, & Wall, 1978). In the past three decades, cultural comparison questioned the applicability of attachment theory in Japanese cul-
studies (e.g., Grossmann, Grossmann, Huber, & Wartner, 1981; ture. They argued that because the fundamental assumptions and
Vereijken, Riksen-Walraven, & Kondo-Ikemura, 1997) have ex-
philosophies underlying most attachment constructs are deeply
amined the universal applicability of attachment theory in infancy
rooted in Western culture, and the bulk of empirical studies sup-
and early childhood. These studies used the standard Strange
porting attachment theory have been conducted in the United
Situation protocol and classification system developed by Ains-
States and Europe with White middle-class subjects, attachment
worth et al. (1978) to compare across cultures the percentages of
theory concepts are not as culturally universal as proponents claim.
Critics of Rothbaum et al.s arguments have responded that Japa-
Chia-Chih DC Wang and Brent S. Mallinckrodt, Department of Educa-
nese culture is not homogenous, and the assertion that secure
tional, School, and Counseling Psychology, University of Missouri attachment leads to dependence in Japan rather than exploration
Columbia. has not been empirically tested (Gjerde, 2001); that although
This article was based on the doctoral dissertation completed by Chia- culture-specific socialization may create differences in how attach-
Chih DC Wang, with the supervision of Brent S. Mallinckrodt. We grate- ment outcomes, such as competence or secure base behavior, are
fully acknowledge Chan-Tai Chang, Yuan-Hsien Lin, and Amanda Rose expressed, attachment may still promote these outcomes in every
for their assistance with data collection and Katherine Daly for her feed- culture (Posada & Jacobs, 2001); and that Rothbaum et al. over-
back on an earlier version of this article.
looked a great deal of cross-cultural research evidence supporting
Correspondence concerning this article should be addressed to Chia-
Chih DC Wang, who is now at the Division of Counseling and Educational the universalist position (van IJzendoorn & Sagi, 2001).
Psychology, School of Education, University of MissouriKansas City, Other critics of the universal viewpoint have argued that meth-
Room 220, 615 East 52nd Street, Kansas City, MO 64110. E-mail: ods used to assess attachment in the United States may not be valid
wangdc@umkc.edu when applied to other cultures (Miyake, Chen, & Campos, 1985).

192
CULTURAL BELIEFS ABOUT IDEAL ADULT ATTACHMENT 193

For instance, Trnavsky (1998) found that distinctly different pro- responses from over 1,000 undergraduates to more than 300 items
totypes of attachment emerge in standard laboratory assessment taken from all self-report measures of attachment available at that
protocols for Chinese and American motherinfant dyads. time. Two orthogonal dimensions of attachment avoidance and
Trnavsky questioned the appropriateness of classifying children anxiety were identified. The Avoidance subscale of the ECRS taps
from non-Western cultures on the basis of attachment prototypes fear of intimacy, discomfort with self-disclosure, reluctance to
derived from Western research, while ignoring the influence of seek support or depend on others, and reticence about expressing
culture-specific parenting practices. feelings. The Anxiety subscale assesses worries about being alone,
The controversy about the cultural universality of childhood concern about a partners disapproval, frustration when a partner is
attachment constructs shows little sign of resolution (e.g., Posada not available, and a desire to be closely merged with ones partner.
& Jacobs, 2001; Rothbaum, Weisz, Pott, Miyake, & Morelli, Low levels of both avoidance and anxiety are interpreted as
2001). At the same time, similar questions have been raised with evidence of adult attachment security. In fact, among attachment
respect to the applicability of Western-based adult attachment theorists and researchers, a tacit understanding appears to have
concepts to non-Western cultures. Compared with the small num- developed, often not explicitly acknowledged, that lower levels of
ber of cross-cultural comparisons of attachment security in chil- anxiety and avoidance uniformly indicate higher attachment secu-
dren, there has been even less research examining the universality rity, better social competence, more ideally healthy interpersonal
of attachment constructs for adult close relationships across cul- relationships, and higher life satisfaction and psychological well-
tures. A PsychInfo database search for articles indexed under the being (Bartholomew & Horowitz, 1991; Lopez & Brennan, 2000;
terms attachment behavior and cross-cultural differences identi- Mallinckrodt, 2001; Mikulincer & Shaver, 2003; Simpson &
fied 45 empirical studies of adults, but only 6 of these are directly Rholes, 1998).
relevant to cross-cultural comparisons of adult attachment. In a A close examination of the construct of adult attachment secu-
survey study, You and Malley-Morrison (2000) found that Korean rity, as it is defined and measured by scales like the ECRS,
young adults scored higher on preoccupied attachment and lower suggests that separate components of affect, cognition, and behav-
on intimacy and friendship expectations than their European ior are tapped by the items. We are concerned that items tapping
American counterparts. The authors speculated that strong family each of these three aspects of adult secure attachment may embody
ties in Korean culture may contribute to the observed differences. Western cultural ideals about the socially approved forms of
The finding that Korean adults reported higher scores on preoc- emotional expression, intimacy, and the optimal balance between
cupied attachment than U.S. respondents was confirmed in another dependence and independence in adult attachments. These West-
study examining the association of adult attachment styles with ern cultural norms do not describe the values of all cultures.
perceptions of elder abuse (Malley-Morrison, You, & Mills, 2000). Differences may be especially evident with regard to overt behav-
A third comparative study reported that Chinese students were ior. For example, in the traditional Chinese culture in Taiwan,
significantly lower in attachment security toward romantic part- emotional restraint is valued, and deeply personal, direct self-
ners than their Canadian counterparts (Ditommaso, Brannen, & disclosure is discouraged even in romantic couples who have
Burgess, 2005). A study investigating adult daughters ratings of been together for many years (Moore, 1998). ECRS items such as
their relationship to their mothers found that Asian Indians in the I prefer not to show a partner how I feel deep down describe
United States scored significantly higher than their European strong cultural values for traditional Taiwanese but are scored as
American counterparts on both attachment closeness and depen- behavioral evidence of higher avoidance (i.e., insecurity) by the
dence (Rastogi & Wampler, 1999). A fifth study comparing three ECRS. Conversely, behavior tapped by items such as I tell my
groups of female university students, Hindi-speaking and English- partner just about everything may be seen by traditional Taiwan-
speaking Asian Indians and Dutch-speaking Belgians, found no ese as representative of a quite immature and selfish burdening of
differences across groups in the level of fit for a causal model ones partner with what should remain a private concern (Cho &
predicting life satisfaction from adult daughters bonds with moth- Cross, 1995), whereas it is interpreted as a relative absence of
ers (Prishnee & Poortinga, 2005). Finally, in a large-scale cross- avoidance (i.e., a sign of security) when scoring the ECRS. Sim-
cultural study examining differences in styles, Schmitt et al. (2004) ilarly, with regard to affect and cognition, the value placed on
reported that secure adult attachment was normative in the major- close connection to family, interdependence, and mutual obliga-
ity of the 62 cultures examined, the ratios of preoccupied attach- tion in traditional Chinese culture would be taken as evidence of
ment were higher in East Asian cultures than in other regions, and attachment anxiety if these values lead to endorsing ECRS items
behavior profiles of romantic attachment in different cultures were such as I worry about being alone or I worry a fair amount
correlated with the corresponding sociocultural norms. about losing my partner.
Clearly, more empirical cross-cultural comparison studies are Cultural comparison research depends on the equivalence of the
needed to inform the debate regarding the universality of attach- measures used to assess the constructs of interest. Lonner and
ment theory as applied to adults. However, an important first step Ibrahim (1996) described different types of equivalence. In the
is to understand how secure attachment is defined in the cultures specific case of adult attachment, functional equivalence requires
to be compared. In the United States, conceptions about what that attachment is used by persons in both cultures to achieve the
constitutes secure adult attachment are reflected in the instruments same ends. Metric equivalence (sometimes termed scalar equiva-
that researchers use to assess adult attachment. One of the most lence) requires that the same scale value would indicate the same
widely used self-report adult attachment measurements is the level or intensity of the construct in both groups (Hui & Triandis,
Experiences in Close Relationships Scale (ECRS; Brennan, Clark, 1985). Conceptual equivalence means that adult attachment is
& Shaver, 1998), which was developed from a factor analysis of defined in both cultures similarly. A variant of conceptual equiv-
194 WANG AND MALLINCKRODT

alence, linguistic equivalence, applies to the special case in which Thus, both self-construal and adult attachment theory are fun-
instruments are presented in different languages to assess the damentally concerned with predictable patterns in relationships
construct in different groups. Other writers prefer the term seman- between self and others. According to attachment theory, an indi-
tic equivalence when referring to this special type of conceptual viduals childhood interactions with caregivers give rise to work-
equivalence because a word-for-word linguistically accurate trans- ing models consisting of deeply ingrained mental representations
lation may not convey the intended meaning (Flaherty et al., 1988; of self and others (Bowlby, 1969, 1988). Working models of self
Hambleton, 1993; Mallinckrodt & Wang, 2004). This typology is and others persist into adulthood and govern appraisals about the
useful in framing the debate between the universalist and relativist level of comfort and support one can generally expect in relation-
viewpoints. In Lonner and Ibrahims terms, relativists claim that ships and ones own level of worth and esteem in the eyes of
attachment serves different functions in different cultures. Univer- others. A positive working model of others is believed to form the
salists disagree. The design of the present study did not generate basis for relatively low adult attachment avoidance, whereas a
findings relevant to this question because we did not examine positive working model of self is necessary to promote low levels
functional differences in the two cultures we examined. Rather, the of adult attachment anxiety (Mikulincer & Shaver, 2003). Because
focus was on the conceptual dimension of equivalence. We hy- these working models do not develop in a cultural vacuum, this
pothesized that young adults in Taiwan and the United States think study is based on the assumption that expectations about self,
differently about adult attachment and define it differently, espe- others, and interactions in close relationships that are part of
cially with respect to behavioral indicators of secure attachment. individuals working models are constrained by cultural beliefs,
Although our primary focus was on the conceptual dimension, it is for example, about the proper balance between independence and
also likely to have important implications for the metric dimension obligation to others, appropriate expectations for support, or how
as well. Specifically, we expected mean differences across cultures one establishes esteem and self-worth. That is, culturally based
for subsets of the ECRS items that tap cultural differences in the attitudes about norms of proper behaviorsuch as how a romantic
conceptual definition of secure adult attachment. couple should communicate their needs and feelings to one an-
other, how and when to negotiate and resolve conflicts, and the
amount and types of support partners in a romantic relationship are
Self-Construal and Adult Attachment justified in expecting from each otherall probably influence
cultural beliefs about what constitutes secure adult romantic
Self-construal may be one of the factors that influence how attachment. In the domains of attachment affect, cognition, and
young adults in Taiwan and the United States think about adult behavior, these proscriptive cultural norms may exert their stron-
attachment. The concept of self-construal has been widely used in gest influence on behavior because attachment behaviors are the
the literature, describing differences in cultural orientations (Fiske, most observable and amenable to censure or praise. Therefore,
Kitayama, Markus, & Nisbett, 1998). Self-construal is a traitlike beliefs about adult attachment security, especially how it is behav-
disposition involving beliefs, feelings, attitudes, and actions in iorally enacted, may be, to some degree, culture specific. The
social relations, especially the degree of separation or connected- differences between Chinese and Western cultural beliefs about
ness between the self and others (Markus & Kitayama, 1991; ideal adult attachment may stem from underlying cultural differ-
Triandis, 1996). Two distinct types of self-construal have been ences in interdependent versus independent self-construals.
described: independent and interdependent. Western culture gen- The main purpose of this study was to extend inquiry about
erally values individualism and emphasizes the uniqueness, inde- cultural universality of attachment theory concepts to the burgeon-
pendence, and rights of an individual to act free of the constraints ing literature on adult attachment. We studied college undergrad-
of others. Thus, most persons who ascribe to Western cultural uates in Taiwan and the United States because the primary devel-
values develop a relatively strong independent self-construal char- opmental task of these individuals is the struggle to establish
acterized by distinct self other boundaries, celebration of unique intimate attachments (Erikson, 1963). In this study, we asked
personal abilities or characteristics, the pursuit of individual self- participants to respond as an ideally emotionally and psycholog-
interest, and an emphasis on direct self-expression. In relationship ically healthy person of your own gender in your culture would
terms, people with an independent self-construal tend to consider answer. This approach offered two important methodological ad-
others and themselves as separate units (Markus & Kitayama, vantages. First, because we were interested in comparing cultural
1991). In contrast, traditional Chinese culture holds collectivistic values about ideals of attachment security, we reasoned that struc-
values, placing a strong emphasis on interdependence and con- turing the rating task to ask how a hypothetically ideal person
nectedness among individuals. These cultural values favor devel- would respond offered a more direct way to tap the data of interest
opment of a relatively strong interdependent self-construal. The than asking for self-reports of actual attachment. Second, at a
self is defined as inseparable from ones relational context. Har- practical level, because college students in Taiwan tend to have
monious interpersonal relationships and meeting ones social ob- relatively less dating experience than U.S. students at the same
ligation are valued. Ones esteem and status is defined by the age, asking about actual attachment experience would introduce a
views of other members in ones social group. People with an potential confound. We reasoned that students in each of the two
interdependent self-construal are more willing to sacrifice their cultures were likely to have developed well-formed opinions about
personal goals for the needs of their social reference group. Inter- what constitutes ideal attachment security, regardless of possible
dependent cultural values define the self as part of a set of differences in actual romantic experience.
relationships deeply embedded in the surrounding social context On the basis of previous research and our personal knowledge of
(Markus & Kitayama, 1991). each culture, we expected that Taiwanese undergraduates, relative
CULTURAL BELIEFS ABOUT IDEAL ADULT ATTACHMENT 195

to their U.S. counterparts, would describe an ideally healthy adult For the Taiwanese comparison sample, participants were solicited from
attachment as having higher levels of avoidance and anxiety. Our five Chinese literature classes on two university campuses in Taiwan, one
second hypothesis held that high levels of independent self- in northern Taiwan and another in southern Taiwan. This course was
construal and lower levels of interdependent self-construal among selected because it is required of every Taiwanese university student. Thus,
a broad selection of majors and geographic areas were represented in this
Taiwanese undergraduates would be associated with beliefs about
sample. No special incentive or extra credit was offered for students
ideal attachment more similar to the Western ideal, that is, featur-
participation, but students were given time to complete the surveys in class.
ing low avoidance and low anxiety. Finally, consistent with pre- All students who attended these classes on the day of data collection
vious research (Markus & Kitayama, 1991; Triandis, 1996), we volunteered to participate in the survey study. A total of 316 Taiwanese
expected that U.S. undergraduates, compared with their Taiwanese students completed surveys, a sample that the instructors estimated was
counterparts, would exhibit significantly higher levels of in- over 95% of the total students who enrolled in the courses. Among all
dependent self-construal and lower levels of interdependent participants, usable data were obtained from 280 (89%) students, which
self-construal. included 127 (45%) women and 153 (55%) men. Their mean age was 19.76
years (SD 1.44, range 1729). In terms of class level, 75% (n 209)
were freshmen, 21% (n 60) were sophomores, 2% (n 5) were juniors,
Method
and 1% (n 2) were seniors. Four students did not specify their class
Participants levels.

Two parallel sets of data, respectively drawn from two distinct cultural
groups of college students, were collected for this study. First, because we Measures
used nonstandard instructions for the ECRS, the reliability of ratings The ECRS (Brennan et al., 1998). The ECRS was used with modified
collected with the modified method was checked with the retest samples of instructions to assess the cultural ideal of healthy adult attachment. The
Taiwanese and U.S. students assessed at two points in time. In addition, ECRS was developed from responses of over 1,000 undergraduate students
two larger samples of Taiwanese and U.S. students were surveyed on one (from a large public university in Texas) to more than 300 items taken from
occasion to provide the data for cross-cultural comparisons. the most frequently used self-report adult attachment instruments. In the
Retest samples. The retest sample of U.S. students (n 26) was original ECRS, respondents are directed to rate how they generally expe-
recruited from a class in interviewing skills for nonmajors and contained 24 rience romantic relationships, not just what may be happening in a current
(92%) women and 2 (8%) men. Their mean age was 20.62 years (SD relationship. We used a 7-point, fully anchored Likert-type scale (1
1.37 years, range 18 24). Regarding ethnic identification, 22 (85%) disagree strongly, 2 disagree somewhat, 3 disagree slightly, 4
indicated White/Caucasian, 2 (8%) indicated Black/African American, neutral/mixed, 5 agree slightly, 6 agree somewhat, 7 agree
and 2 (8%) indicated Asian American/Pacific Islander. Students received strongly). The original ECRS is only partially anchored at response points
extra credit for completing the survey. Packets containing the ECRS with 1, 4, and 7. Factor analyses identified two orthogonal factors of attachment
the modified instructions described below were distributed and collected in Anxiety and Avoidance, with each subscale containing 18 items. Brennan
one class period. A second packet identical to the first was distributed and et al. reported internal reliability (coefficient alpha) of .91 and .94 for the
collected 1 week later. All students in the class volunteered to provide both Anxiety and Avoidance subscales, respectively. Brennan, Shaver, and
test and retest data. Clark (2000) reported that retest reliabilities (3-week interval) for both
The retest sample of Taiwanese students (n 38) was recruited from an subscales were .70. Evidence of validity is provided by significant corre-
undergraduate Chinese history course required for every university student. lations in expected directions with other measures of adult attachment and
The sample consisted of 22 (58%) women and 16 (42%) men. Their mean with measures of preferences for touch and sexual behavior (Brennan et al.,
age was 19.16 years (SD 0.86, range 18 21). Following the same 1998, 2000).
procedure used for the U.S. sample, packets containing the Chinese version In the present study, the ECRS was used with these modified instruc-
of the ECRS with modified instructions were distributed and collected in tions:
class on the first occasion, and again 1 week later. Students were offered
extra credit toward their course grade as an incentive for participation. A The following statements concern how people experience and feel in
total of 42 students completed the first survey, but 4 were absent from class romantic relationships. Please respond to each item according to your
on the second occasion, resulting in a 90% research compliance rate beliefs of how an ideally emotionally and psychologically healthy
(n 38). person of your own gender in your culture would answer. Choose a
Cultural comparison samples. U.S. undergraduate students were so- response from the scale below indicating how much this culturally
licited from a large public university in the Midwest. Survey packets were ideally healthy person would agree or disagree with the statement.
distributed in an undergraduate developmental psychology course. (Note
that data were collected before students in this course had received any In addition to this instruction at the beginning of the instrument, the
instruction about attachment theory.) Students were offered extra credit following reminder appeared at the halfway point (after item #18): Please
toward their course grade as an incentive to participate. A total of 291 U.S. remember to respond to items according to how an ideally emotionally and
undergraduate students completed surveys. A final sample of 268 (92%) psychologically healthy person of your own gender in your culture would
provided usable data. Among them, 204 (76%) were women, 64 (24%) answer. In this study, internal reliability (coefficient alpha) for the ECRS
were men. Their mean age was 20.62 years (SD 1.02, range 19 23). using these instructions in the U.S. sample was .93 and .92 for the
Regarding ethnic identification, 242 (90%) indicated White/Caucasian, 8 Avoidance and Anxiety subscales, respectively. Retest reliability after a
(3%) indicated Black/African American, 5 (2.0%) indicated Asian 1-week interval in the U.S. retest sample (n 26) was .95 and .93 for the
American/Pacific Islander, 5 (2.0%) indicated a mixture of race/ethnici- Avoidance and Anxiety subscales, respectively.
ties, 2 (0.7%) indicated Hispanic, Chicano, Latino or Latina, 1 (0.3%) Experiences in Close Relationships Scale-Chinese Version (ECRS-C).
indicated Native American, and 5 (2.0%) did not indicate an ethnic The ECRS was adapted into Mandarin Chinese by Mallinckrodt and Wang
identification. In terms of class level, 1% (n 3) were freshmen, 48% (n (2004) using back-translation methods, followed by quantitative verifica-
129) were sophomores, 35% (n 95) were juniors, and 15% (n 41) were tion procedures in which retest and internal reliabilities of the original
seniors. ECRS and the ECRS-C were systematically compared in a bilingual
196 WANG AND MALLINCKRODT

sample. Strong evidence for the semantic equivalence of the two scales was Results
reported in that coefficient alpha was statistically equivalent for same-
language versus mixed-language forms of the instrument. These semantic Preliminary Analyses
equivalence comparisons were based on standard instructions for the scale Data-cleaning procedures were conducted before the prelimi-
adapted into Mandarin Chinese. For the present study, the modified in- nary analyses. For the U.S. sample, 11 surveys were dropped
structions described above were also translated. Retest reliabilities (n 38, because of missing data, and eight were eliminated because of
1-week interval) using the modified instructions were .90 and .92 for the out-of-range responses on one or more of the items used to detect
Avoidance and Anxiety subscales, respectively. Internal reliabilities (co- careless or random responding. Data from 3 international students
efficient alpha) in the Taiwanese cultural comparison sample using the
were also excluded. Finally, using procedures to screen for uni-
modified instructions were .76 and .85 for the Avoidance and Anxiety
variate and multivariate outliers suggested by Tabachnick and
subscales, respectively.
Fidell (2001), data from one U.S. student were excluded from
Self-Construal Scale (SCS; Singelis, 1994). The SCS was administered
further analyses. A similar data-cleaning procedure was conducted
to assess participants individualistic/collectivistic cultural orientations.
The SCS consists of two subscales designed to measure an individuals
with Taiwanese students. Surveys from 36 (11%) of the original
independent/individualistic and interdependent/collectivistic self-construal, students were excluded because of missing data (n 19) or
which have been shown to be orthogonal dimensions. Each subscale out-of-range responses to random-response items (n 17).
contains 12 items, with a 7-point Likert-type scale ranging from 1 (strongly A chi-square test revealed that the proportion of men and
disagree) to 7 (strongly agree). Internal reliability (coefficient alpha) was women in each cultural sample was significantly different, 2(1,
reported to be .70 for the Independent subscale and .74 for the Interdepen- N 548) 52.90, p .001. Consequently, separate analyses
dent subscale. Evidence of construct validity was provided by significant were conducted for men and women to ensure that any cultural
score differences between groups in directions consistent with their cultural differences obtained were not because of potential gender differ-
characteristics and by correlations in expected directions with other vari- ences in the samples. Calculation of retest and internal consistency
ables (Singelis, 1994). The Chinese version of the SCS was developed by reliabilities (reported in the Method section) for the modified
Singelis and his colleagues following a translation and back-translation ECRS/ECRS-C instructions suggested that respondents could
procedure (Singelis, Bond, Sharkey, & Lai, 1999). In the present study, complete the rating task and reliably report their beliefs about the
coefficient alphas for the English SCS were .65 and .63, and for the romantic attachment of an ideally emotionally psychologically
Chinese SCS, .66 and .78, for Independent and Interdependent subscales, healthy person. A further test of the discriminant validity of this
respectively. rating task was conducted by comparing ideal attachment ratings
with actual self-ratings of the Taiwanese students. Significant
differences in repeated measures comparisons would suggest that
Procedure
participants could successfully differentiate between their ideal
Rigorous efforts were made to follow identical data collection proce- beliefs and assessment of their actual attachment style. Ratings of
dures for the U.S. and Taiwanese samples. After obtaining the instructors attachment avoidance and anxiety served as dependent variables in
permission, a researcher came to classes in person to describe the study. a 2 (gender) 2 Type of Rating (ideal vs. self ratings) multivariate
Students in both groups were given survey packets, which included the analysis of variance (MANOVA). Results suggested no significant
consent form, demographic items, the ECRS (or the ECRS-C) with the direct effect for gender, F(2, 266) 1.66, ns, and no significant
modified instructions, described above, and the SCS (or the Chinese- Gender Type of Rating interaction, F(2, 266) 2.63, ns.
language equivalent). Packets given to Taiwanese students also included a However, there was a significant effect for the repeated measures
second version of the ECRS with standard instructions so that we could factor of rating type, F(2, 266) 15.50, p .001, 2 .104.
examine whether the modified instructions resulted in different responses Table 1 shows results of repeated measures t tests comparing ideal
than the standard version. Thus, data on both ideal attachment beliefs and beliefs with actual attachment self-ratings. Differences were sig-
actual self-reported adult attachment were collected from students in Tai- nificant for women with regard to both avoidance and anxiety. The
wan. We intended to follow the same procedure with U.S. students, but the differences for men between ideal beliefs and actual self-ratings
course instructor wanted to keep the interruption of class time to a mini- also approached statistical significance ( ps .062). On the basis
mum. Thus, survey packets for U.S. students contained only the modified of the significant repeated measures main effect, we concluded that
ECRS. Taiwanese students could differentiate assessment of their actual
For Taiwanese students, the ECRS-C with modified instructions was
attachment status from their beliefs about ideal adult attachment
presented first, followed by the SCS. The ECRS-C with standard instruc-
although this inference is stronger for women than for men.
tions was presented last in the packet to assess students ratings of actual
The present study is the first time the ECRS-C has been used
adult attachment. Because U.S. students did not complete the standard
with a relatively large group of persons whose first language is
ECRS, a fixed rather than a counterbalanced order was used for Taiwanese
students so that both groups were presented first with the task of reporting
Chinese. Because we were concerned about potential bias caused
their ideal attachment beliefs. Students who did not wish to participate by the differences between our Taiwanese and U.S. samples in
were instructed to simply return their survey packet at the same time other using the 7-point response scale, we conducted an analysis to
students returned their packets. No personally identifying information was examine differences in response styles between these two samples.
requested on the survey. The entire procedure was completed in class and For each cultural group, seven variables were created, consisting
required about 30 40 min for Taiwanese students and 20 min for U.S. of the number of times a given response scale point was chosen for
students. Both Taiwanese and U.S. student packets contained six items the 36 items. (Items were analyzed without reverse keying.) Thus,
throughout the survey directing students to Please mark 9 for this item to if all seven response scale points were used equally, then a mean
guard against careless or random responding. of 5.14 (36/7) would be expected for each variable. Because scores
CULTURAL BELIEFS ABOUT IDEAL ADULT ATTACHMENT 197

Table 1
Repeated Measures Comparison of Beliefs About Ideal Attachment With Self-Ratings for
Taiwanese Women and Men

Belief about ideal Attachment


attachment self-ratings

Sex/attachment variable M SD M SD df t

Women (n 124)
Avoidance 3.30 0.72 3.54 0.89 123 4.73***
Anxiety 3.97 0.85 4.13 0.91 123 2.26*
Men (n 146)
Avoidance 3.26 0.74 3.35 0.85 145 1.90
Anxiety 4.15 0.99 4.25 1.06 145 1.88

Note. Data from 10 students concerning attachment self-ratings were missing.


p .06. * p .05. *** p .001.

for any one of these seven variables is linearly dependent on the anxiety and more avoidance than U.S. men. Beliefs about ideal
other six, a MANOVA could not be used. A series of seven t tests attachment of Taiwanese women also exhibited significantly more
suggested that Taiwanese respondents, relative to their U.S. coun- avoidance but not more anxiety than U.S. women. In general,
terparts, were significantly less likely to use the extremes of the Table 2 shows a greater number of significant cross-cultural dif-
response scale (i.e., Points 1 and 7). U.S. students chose the ferences (shown in horizontal comparisons of cells) than within-
highest or lowest response option roughly twice as frequently as culture gender differences (shown in vertical comparisons). In fact,
Taiwanese students. For the response point disagree strongly, the only significant within-culture gender difference was that the
Taiwanese M 2.84 (SD 4.15), whereas U.S. M 5.60 (SD ideals of U.S. men exhibited significantly more attachment avoid-
2.84), t(546) 6.82, p .01; for the response point agree ance than the ideal attachment of U.S. women. Because cultural
strongly, Taiwanese M 2.96 (SD 3.73), whereas U.S. M differences were more pervasive than gender differences and be-
4.92 (SD 3.72), t(546) 6.14, p .01. To correct for this cause the former were detected even after controlling for possible
difference in response style, for the remainder of the analyses gender differences, our first hypothesis received strong support.
reported in this article, a modified scoring scheme was used for the The relative positions of the mean differences for these four
ECRS and ECRS-C in which the two most extreme scale points groups are shown in Figure 1. All data for this figure were
were combined (i.e., 1 2 and 7 6) for both cultural groups. converted to z scores on the basis of the grand mean and pooled
standard deviation of the combined sample of both cultures and
Tests of Research Hypotheses sexes. Circles represent the 95% confidence interval around point
estimates. Attachment anxiety is plotted on the y-axis, with lower
Our first hypothesis held that Taiwanese students would exhibit
relatively higher avoidance and higher anxiety in their beliefs scores at the top of the scale so that the figure depicts increasing
about ideal romantic attachment than U.S. students. Although we attachment security (low avoidance, low anxiety) in the upper
did not propose a specific hypothesis about differences between lefta convention with an established tradition in attachment
men and women, because the proportion was significantly differ- research (Bartholomew & Horowitz, 1991; Brennan et al., 1998).
ent in each sample and because previous research (Sagarin & In keeping with previous research (Fraley & Waller, 1998), the
Guadagno, 2004; Searle & Meara, 1999) has found significant figure depicts attachment as two continuous dimensions rather
gender differences in emotional expressiveness and feelings of than as four categories demarked by arbitrary cutting scores.
jealousy in romantic relationships, we decided to examine poten- Figure 1 indicates that beliefs about ideal attachment for both
tial gender differences in conjunction with the test of our first Taiwanese men and women diverge from their U.S. counterparts in
hypothesis to guard against the possibility that gender differences the direction of higher Anxiety and Avoidance subscale scores. In
would be confounded with cultural differences. Thus, a two-way addition, beliefs about ideal attachment of U.S. women, compared
Gender Cultural Group MANOVA was used to test this hypoth- with U.S. men, involve much less avoidance and somewhat higher
esis. Results suggested a significant direct effect for cultural group, levels of anxiety. For the Taiwanese men and women, actual
F(2, 543) 47.12, p .001; a significant direct effect for gender, self-reported attachment is shown in Figure 1 as well as their
F(2, 543) 5.93, p .01; and a significant Gender Group attachment ideals. It can be seen that for both Taiwanese men and
interaction, F(2, 543) 9.23, p .001, which supported our women, their idealized attachment incorporates lower levels of
decision to examine gender differences in conjunction with cul- Anxiety and Avoidance than their actual self-reported adult
tural differences. attachment.
Table 2 reports results of independent sample t tests used as Our second hypothesis asserted that Taiwanese students inde-
follow-up tests to examine differences across each of the four pendent self-construal would be negatively associated with Avoid-
possible combinations of gender and culture. Cultural beliefs about ance and Anxiety in their ideal model of attachment (i.e., more
ideal attachment for Taiwanese men exhibited significantly more similarity to the Western ideal of attachment security), whereas the
198 WANG AND MALLINCKRODT

Table 2
Comparison of Ideal Attachment Beliefs by Sex and Culture

Taiwanese U.S.
Cells
Attachment variable M SD M SD compared df t

Womena,b
Avoidance (a) 3.37 0.62 (b) 2.56 0.64 ab 329 11.20***
Anxiety (c) 3.98 0.73 (d) 3.82 0.96 cd 329 1.55
Menc,d
Avoidance (e) 3.35 0.62 (f) 3.00 0.88 ef 215 3.36***
Anxiety (g) 4.12 0.84 (h) 3.66 0.84 gh 215 3.66***
ae 278 0.18
cg 278 1.50
bf 266 4.32**
dh 266 1.20

Note. Results of univariate 2 2 analysis of variance (ANOVA) for attachment anxiety means shown above:
main effect for cultural group, F(1, 544) 14.39, p .001, 2 .026; main effect for gender, F(1, 544) 0.01,
ns, 2 .000; Group Gender interaction, F(1, 544) 3.53, ns, 2 .006. Results of univariate 2 2
ANOVA for attachment avoidance: main effect for cultural group, F(1, 544) 86.94, p .001, 2 .138; main
effect for gender, F(1, 544) 11.63, p .001, 2 .021; Group Gender interaction, F(1, 544) 13.13, p
.001, 2 .024.
a
For Taiwanese women, n 127. b For U.S. women, n 204. c For Taiwanese men, n 153. d For U.S.
men, n 64.
** p .01. *** p .001.

level of interdependent self-construal would be positively associ- ther explore these differences, we compared the correlation be-
ated with both Avoidance and Anxiety (i.e., more discrepancy tween the Anxiety and Avoidance ECRS subscales for U.S. stu-
from the Western concept of ideal attachment). Table 3 shows dents (r .21) and for Taiwanese students (r .01), using an
results of bivariate Pearson correlations between all the measured r to z transformation procedure. Results suggested that the ob-
variables in this study, separately by cultural group. As expected, served difference in this correlation between cultures was statisti-
attachment anxiety was associated with interdependent self- cally significant at p .05.
construal for Taiwanese students. However, contrary to expecta- The third hypothesis held that Taiwanese students would have
tions, attachment avoidance was not positively associated with significantly higher interdependent self-construal and lower inde-
independent self-construal. In fact, the association was negative pendent self-construal than U.S. students. A 2 (gender) 2 (cul-
and statistically significant for Taiwanese students. We also found tural group) MANOVA was used to test this hypothesis. Results
an unexpected significant negative correlation between Avoidance suggested a significant main effect for cultural group, F(2, 537)
and interdependent self-construal for Taiwanese students. To fur- 18.51, p .001, but no significant main effect for gender, F(2,

Figure 1. Cultural differences in ideal attachment beliefs.


CULTURAL BELIEFS ABOUT IDEAL ADULT ATTACHMENT 199

Table 3
Correlations by Cultural Group

Variable 1 2 3 4 5 6

Taiwanese undergraduates (n 278)


1. Independent self-construal
2. Interdependent self-construal .52**
3. Ideal attachment avoidance .16** .17**
4. Ideal attachment anxiety .09 .18** .01
5. Actual attachment avoidance .23** .21** .75** .05
6. Actual attachment anxiety .01 .17** .06 .70** .02
U.S. undergraduates (n 265)
1. Independent self-construal
2. Interdependent self-construal .06
3. Ideal attachment avoidance .08 .09
4. Ideal attachment anxiety .07 .21** .17**

Note. All correlations with attachment Avoidance and Anxiety are based on the modified English- and
Chinese-language version of the Experiences in Close Relationships Scale scoring, with recoding of 7 6, 1
2 for the response scale.
** p .01.

537) 1.17, ns, or a Gender Group interaction, F(2, 537) controlling for the effects of self-construal, group and gender
1.70, ns. Table 4 reports univariate effect sizes for these analyses differences in ideal attachment beliefs remained.
and illustrates that our third hypothesis was supported with regard
to both interdependent and independent self-construal. Follow-up Analyses
Given these findings, we wondered whether underlying differ-
ences in self-construal accounted for most of the cultural differ- Given that significant differences in cultural beliefs about ideal
ences in ideal attachment. Accordingly, we repeated the 2 (gen- attachment could not be explained on the basis of differences in
der) 2 (cultural) group analyses shown in Table 2, but this time self-construal, we conducted a more fine-grained exploratory anal-
as a multivariate analysis of covariance (MANCOVA), with inter- ysis. We hoped to identify the particular items on the ECRS/
dependent and independent self-construal entered as covariates. ECRS-C that contributed most to the observed cultural differences
Results were unchanged from the essential findings reported in in subscale scores. We hoped to identify unifying themes that tied
Table 2. In the MANCOVA, there was a significant main effect for these items together conceptually. To accomplish this, direct dis-
group, F(2, 535) 41.81, p .0001; a significant main effect for criminant function analyses were conducted separately for men
gender, F(2, 535) 6.04, p .001; and a significant Gender and women. Again, we corrected for cultural differences in the use
Group interaction, F(2, 535) 9.09, p .001. Thus, even after of the ECRS/ECRS-C response scales by collapsing the extreme
points (7 6), (1 2) for both cultural groups. The 36 items were
entered simultaneously as a block. Wilkss lambda criterion was
Table 4 used to test statistical significance. The canonical discriminant
Comparison of Self-Construal by Sex and Culture function for women was statistically significant, 2(36, N
331) 357.75, p .001, and accounted for 84% of the variance
Taiwanese U.S. in cultural groups. The canonical discriminant function for men
was also statistically significant, 2(36, N 217) 134.87, p
Self-construal M SD M SD df t
.001, and accounted for 73% of the variance in groups. Table 5
Womena,b shows items that met two criteria: (a) a loading of .30 on the
Independent 4.76 0.65 4.93 0.67 324 2.21* canonical discriminant function and (b) a significant t test differ-
Interdependent 5.14 0.66 4.97 0.70 324 2.21* ence at p .05 in the item means across the two groups. In total,
Menc,d
Independent 4.78 0.82 5.10 0.75 214 2.78**
four items for women and five items for men met these criteria.
Interdependent 5.19 0.89 4.89 0.63 214 2.48* In terms of the themes that emerged in these items, results
suggest that compared with U.S. women, Taiwanese women were
Note. Results of a univariate 2 2 analysis of variance (ANOVA) for more likely to endorse as attachment ideals two Anxiety subscale
independent self-construal means shown above: main effect for cultural items involving resenting time ones partner is away and forcing
group, F(1, 538) 13.52, p .001, 2 .021; main effect for sex, F(1,
538) 1.87, ns, 2 .003; Group Sex interaction, F(1, 538) 1.46, partners to show more commitment and feelings. Taiwanese
ns, 2 .003. Univariate 2 2 ANOVA for interdependent self-construal: women were less likely to endorse as attachment ideals two
main effect for cultural group, F(1, 538) 11.69, p .001, 2 .023; reverse-keyed items from the Avoidance subscale involving turn-
main effect for sex, F(1, 538) 0.04, ns, 2 .000; Group Sex ing to ones partner for comfort and reassurance in times of need.
interaction, F(1, 538) 0.88, ns, 2 .002.
a
For Taiwanese women, n 125. b For U.S. women, n 201. c For
Taiwanese men were less likely than their U.S. counterparts to
Taiwanese men, n 152. d For U.S. men, n 64. endorse the same two reverse-keyed Avoidance subscale items
* p .05. ** p .01. chosen by Taiwanese women about relying on ones partner in
200 WANG AND MALLINCKRODT

Table 5
Discriminant Analysis of Individual ECRS Items, Conducted Separately for Women and Men

Taiwanese U.S.
Loading M M Subscale ECRS item and number
a,b
Women

.580 4.21 3.21 Anx. 36. I resent it when my partner spends time away from me.
.442 4.88 5.80 Avd. 35. I turn to my partner for many things, including comfort
and reassurance. (R)
.371 4.52 5.81 Avd. 33. It helps to turn to my romantic partner in times of
need. (R)
.303 4.13 3.39 Anx. 20. Sometimes I feel that I force my partners to show more
feeling, more commitment.

Menc,d

.434 4.12 3.32 Anx. 36. I resent it when my partner spends time away from me.
.373 4.16 5.40 Avd. 35. I turn to my partner for many things, including comfort
and reassurance. (R)
.344 3.91 3.07 Avd. 11. I want to get close to my partner, but I keep pulling
back.
.325 4.82 3.85 Anx. 8. I worry a fair amount about losing my partner.
.325 4.77 5.57 Avd. 33. It helps to turn to my romantic partner in times of
need. (R)

Note. ECRS Experiences in Close Relationships Scale; Avd. Avoidance subscale; Anx. Anxiety
subscale; R indicates that an item is ordinarily reversed keyed when computing ECRS Anxiety or Avoidance
subscales. However, for clarity of presentation in this table, means are presented without reverse keying. Higher
mean scores in all cases indicate stronger agreement with the item on a 5-point scale that ranged from 2 to 6.
Lower means for Taiwanese students on reverse-keyed items indicate more Anxiety or Avoidance. From
Attachment Theory and Close Relationship, by J. A. Simpson and W. S. Rholes, 1998. Copyright 1998 by
Guilford Press. Reprinted with permission.
a
For Taiwanese women, n 117. b For U.S. women, n 191. c For Taiwanese men, n 136. d For U.S.
men, n 60.

times of need. Taiwanese men were more likely than their U.S. the Taiwanese perspective would involve behavioral norms sug-
counterparts to endorse an additional Avoidance item about want- gesting more Anxiety and Avoidance than norms concerning ide-
ing to get close but pulling back. Finally, they were more likely ally secure attachment from the U.S. perspective. This hypothesis
than U.S. men to endorse as ideals two Anxiety items involving received clear support. In addition to the cultural differences, we
resentment of partners for spending time away and worries about also found statistically significant gender differences in beliefs
losing ones partner. about ideal attachment and significant Gender Culture interac-
A final follow-up analysis was based on the observation that tions. Because of the different cultural norms between Chinese and
estimates of internal reliability (coefficient alpha) for the ECRS-C American societies, we also hypothesized that Taiwanese young
in our Taiwanese sample were lower than for the ECRS in our U.S. adults would report higher interdependent and lower independent
sample. For the Avoidance and Anxiety subscales, respectively, self-construal than their U.S. counterparts; Table 4 shows general
these values were .74 and .83 for the Taiwanese students and .91 support for this hypothesis. These results are congruent with other
and .91, respectively, for U.S. students. Using Feldts (1969) cultural comparison research (e.g., William-Doherty, Hatfield,
formula for statistically comparing coefficient alpha, as described Thompson, & Choo, 1994).
by Mallinckrodt and Wang (2004), the difference in alpha for However, although we found mean differences across cultures
Avoidance in the present study was significant, F(267, 279)
in support of our third hypothesis about interdependence and
2.66, p .01, and the difference in alpha for Anxiety was also
independence, our second hypothesis was only partly supported.
significant, F(267, 279) 1.66, p .01.
Attachment Anxiety was positively associated with interdependent
self-construal as expected, but the relationship between Avoidance
Discussion and independent self-construal was significantly negative for the
The chief purpose of this study was to compare Taiwanese and Taiwanese group, instead of the positive association we hypothe-
U.S. young adults with regard to their beliefs about the attachment sized. These correlations were small in magnitude (rs .18; see
of an ideally emotionally and psychologically healthy person of Table 3) and may have been attenuated because of the marginal
your own gender in your culture. On the basis of our understand- reliability of both SCS subscales for Taiwanese respondents and
ing of Chinese culture in Taiwan and from the results of limited for the Independent subscale for U.S. students. It is noteworthy
available research, we expected that ideal adult attachment from that only the Avoidance subscale behaved counter to expectations
CULTURAL BELIEFS ABOUT IDEAL ADULT ATTACHMENT 201

in these analyses. An important component of independent self- difference described above. For example, I turn to my partner for
construal is a willingness to express oneself without self-censoring many things, including comfort and reassurance and It helps to
for the sake of maintaining social harmony. Attachment avoidance turn to my romantic partner in times of need were endorsed much
involves a reluctance to self-disclose because of fear of emotional less by Taiwanese than by U.S. students, perhaps because these
intimacy. Perhaps the observed negative correlation was the result items imply a direct expression of needs that is not culturally
of a tendency toward restrained emotional expression that is char- acceptable in Taiwan. Furthermore, the finding that Taiwanese
acteristic of both high attachment avoidance and low independent students were more likely to endorse as attachment ideals items
self-construal. such as I resent it when my partner spends time away from me
In the follow-up analyses, we believe that the comparison of and I worry a fair amount about losing my partner may reflect a
differences in internal reliability provide additional support for our relation-based self-identity.
first hypothesis. Although these ECRS-C internal reliability esti- With regard to the universality debate, the strongest test of the
mates for Taiwanese students are acceptable in absolute terms, relativist position would be provided by a study designed to
they were significantly lower than the corresponding estimates for examine Lonner and Ibrahims (1996) functional dimension; that
the U.S. sample in this study. The equivalence of the semantic is, do Taiwanese and U.S. students use their attachment systems to
content of the ECRS and ECRS-C items was quantitatively veri- achieve different goals? The present study does not offer evidence
fied by Mallinckrodt and Wang (2004), who found that internal on this point; therefore, we believe the cultural differences ob-
reliability was equivalent in comparisons of same-language versus served in this study should not be interpreted as supporting the
split-language versions administered to a bilingual sample of Tai- relativist position. Rather, the evidence, at Lonner and Ibrahims
wanese international students in the United States. However, conceptual level of equivalence, suggests that young adults in
Mallinckrodt and Wang did not conduct a cross-cultural compar- these two cultures tend to conceptualize ideal attachment differ-
ison. The finding that coefficient alpha was significantly different ently. The magnitude of the mean differences we did find should
for the two cultural groups examined in the present study suggests be kept in perspective. On the basis of the rough benchmarks
that for Taiwanese students, the ECRS-C items do not measure described by Cohen (1969), an effect size of 2 .04 for cultural
cohesive, unitary constructs to the same extent that attachment differences in mean Anxiety and 2 .16 for Avoidance are best
anxiety and avoidance are measured by the ECRS items for U.S. interpreted as small and moderate effects sizes, respectively
students. (Vacha-Hasse & Thompson, 2004).
People in Taiwan tend to derive their identity from their group Because these cultural differences remained significant even
affiliations and network of relationships. It is a common belief that after controlling for the effects of self-construal in the
the self is inseparable from the relational context and that true MANCOVA analysis, underlying differences in self-construal do
satisfaction and lasting happiness are attained mainly through not appear to be the basis for cultural differences in the way young
achieving a harmonious balance in ones social context. Further- adults in our samples think about ideal attachment. Perhaps other
more, in contrast to the dominant culture in the United States, culture-related variables (e.g., emotional expressivity, filial piety,
which values explicit expression of ones inner feelings and direct or norms of proscribed behavior in close romantic relationships)
communication of ones emotional needs, in traditional Taiwan- account for the observed differences in ideal attachment beliefs.
ese/Chinese culture, such direct expression is generally considered More research certainly is needed to identify possible underlying
immature, unacceptable, and threatening to the primary cultural factors contributing to these cultural differences.
value of interpersonal harmony (Bagozzi, Wong, & Yi, 1999). In terms of practical significance, the findings of this study
Research suggests that Taiwanese partners in close relationships suggest that some behaviors and beliefs valued by Taiwanese
attempt to anticipate each others needs without relying on direct regarding healthy ideal attachment could be misperceived from the
requests of one another (Gao, Ting-Toomey, & Gudykunst, 1996). U.S. standpoint as reflecting enmeshment, blurred interpersonal
In addition, the widespread emphasis in the United States on boundaries, and a preoccupation with abandonment. Similarly,
emotional separation from parents to give priority to ones spouse Taiwanese values of restraint in directly disclosing ones emo-
is antithetical to Taiwanese cultural values, which emphasize a tional needs or asking for help are likely to be misperceived as
lifelong close emotional connection between parents and children, attachment avoidance from the Western perspective. It is very
especially a mother and her children (Ho, 1989; Moore, 1998). It important for counselors and psychologists not to pathologize but
is important to note that the enhanced avoidance in Taiwanese rather to consider these evident differences as diverse ways of
ideal attachment found in the present study was specific to expressing ideal attachment across different cultures (Fiske et al.,
romantic relationships and may not be found in parent child 1998; S. Sue, 1999). Research suggests that the Western ideal of
relationships. secure attachment is related to positive psychotherapy processes
Given this cultural context, the higher avoidance and higher for U.S. clients (Mallinckrodt, Porter, & Kivlighan, 2005; Meyer
anxiety endorsed in Taiwanese ideal adult attachment may reflect & Pilkonis, 2001), but it is important to avoid the harm that can be
Chinese interpersonal cultural norms especially with regard to done to clients of other cultures from inappropriate generalization
overt behavior concerning social inhibition, emotional restraint, (D. W. Sue, Arredondo, & McDavis, 1992; D. W. Sue & Sue,
and discouragement of directly disclosing ones emotional needs 1990).
as well as relation-referent self-identity. Results of the discrimi- Findings of this study also suggest significant gender differ-
nant function analyses appear to support these inferences. The ences in ideal attachment, and significant Gender Culture inter-
specific ECRS/ECRS-C items that most strongly differentiated the actions. U.S. men endorsed significantly more avoidance as part of
two groups in this study seem to involve the themes of cultural their attachment ideal than did U.S. women, whereas Taiwanese
202 WANG AND MALLINCKRODT

men and women did not differ in ideals for either attachment nonequivalence of the translation rather than the result of true
subscale, and U.S. men and women did not differ in their ideal differences in the samples. All data were vulnerable to the biases
beliefs about attachment anxiety. Apparently, U.S. women expect inherent in self-report measurement. Although the data were kept
an ideally attached woman to be more emotionally expressive, anonymous from the beginning of collection, ratings of ideal
engage in more self-disclosure, and to desire more closeness in attachment may have been susceptible to social-desirability bias.
attachments than men expect of an ideally healthy attachment for Conversely, it could be argued that we were actually attempting to
a man. This result is consistent with research suggesting that assess the most socially desirable form of attachment within a
women in the United States are more willing to express their culture. The analysis of retest data suggested that students ratings
feelings and affective experiences and are more expressive about of ideal attachment were stable across time, but more research is
emotion in close romantic attachments than men (Brody & Hall, clearly needed to substantiate the validity of procedures that ask
1993; Searle & Meara, 1999; Sprecher & Sedikides, 1993). Of participants to provide ratings of hypothetical others.
most interest, these gender differences are not evident in Taiwan. The primary findings of the present study suggest the presence
Perhaps American culture may foster a greater gender difference in of culture-specific beliefs about adult secure attachment between
attitudes about ideal attachment than Taiwanese culture. In Amer- the two cultural groups. Further research is needed to replicate and
ican society, boys at an early age are often encouraged to be generalize the findings in order to identify the underlying mech-
independent, strong, responsible, and exhibit control over expres- anisms accounting for the observed differences. Qualitative re-
sions of emotional vulnerability. Boys can be easily shamed by search methods (e.g., directly asking participants to describe their
being labeled like a girl. In contrast, in Taiwanese culture, both ideal attachment relationship) could be helpful to further clarify
boys and girls are encouraged to remain psychologically and the basis for the cross-cultural conceptual differences in ideal
physically close to their mothers until adolescence, and they are attachment observed in this study. The technique of asking infor-
expected to be obedient and compliant with the advice given by mants about their view of an ideal state may offer a promising
parents and other family elders (Ho, 1989). Thus, it is possible that approach for cultural comparison of values in other domains. A
fewer differences between the socialization of boys and girls in particularly fruitful area may be further investigation of how
Taiwan result in relatively more similarity in the attachment ideals gender, culture, and developmental experience influence how as-
of young women and men in Taiwan than in the United States. pirations for ideal attachment differ from actual self-perceptions.
However, much more research is needed in this area before a solid Another direction would be to examine the relations of ideal and
conclusion could be made. actual adult attachment to psychological symptoms, well-being, or
Several important methodological limitations of this study must successful adaptation. If this beginning effort at comparing how
be noted. All U.S. participants were undergraduates drawn from a ideal adult attachment is conceptualized across cultures can be
developmental psychology course in the Midwest. Their beliefs extended to examine functional aspects of adult attachment in each
about ideal attachment may have differed from other undergradu- culture, then the community of attachment researchers could draw
ates in other parts of the United States. Similarly, although the closer to forming evidence-based conclusions in the debate about
recruitment of Taiwanese participants had a good geographical universality of attachment as applied to adults.
balance across the country, their beliefs may not have been similar
to other young adults in Taiwan. In addition, perhaps the slight
References
difference in age between the two cultural groups may have had an
impact on ideal attachment. Furthermore, although information Ainsworth, M. D. S. (1967). Infancy in Uganda: Infant care and the growth
regarding the representations of different majors between U.S. and of love. Baltimore: Johns Hopkins University Press.
Taiwanese samples is not available, the Taiwanese sample is likely Ainsworth, M. D. S., Blehar, M., Waters, E., & Wall, S. (1978). Patterns
to include students with various majors (because the literature of attachment. Hillsdale, NJ: Erlbaum.
course in Taiwan is required of all college freshman and sopho- Bagozzi, R. P., Wong, N., & Yi, Y. (1999). The role of culture and gender
more students), and the U.S. sample is most likely to overrepresent in the relationship between positive and negative affect. Cognition &
Emotion, 13, 641 672.
students majoring in psychology or education. Because it is a
Bartholomew, K., & Horowitz, L. M. (1991). Attachment styles among
practical impossibility to collect data from members of two dif- young adults: A test of a four-category model. Journal of Personality
ferent cultures under precisely the same conditions, it is important and Social Psychology, 61, 226 244.
to be aware that any of the differences between the collected Bowlby, J. (1969). Attachment and loss: Vol. I. Attachment. New York:
samples may have contributed to the observed differences in this Basic Books.
study. Bowlby, J. (1988). A secure base: Clinical applications of attachment
It is also possible that the ideal attachment beliefs reported by theory (pp. 137157). New York: Routledge.
the university students in either culture, most of whom were not Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measure-
married, are not representative of adults in later stages of devel- ment of adult attachment: An integrative overview. In J. A. Simpson &
opment. In addition, one should not assume that all individuals W. S. Rholes (Eds.), Attachment theory and close relationships (pp.
46 76). New York: Guilford Press.
from a particular culture are homogeneous with regard to their
Brennan, K. A., Shaver, P. R., & Clark, C. A. (2000). Specifying some
values and beliefs. Therefore, generalization to all young adults in mediators of attachment-related anxiety and avoidance. Unpublished
either culture is uncertain at best. Despite our efforts to document manuscript, State University of New York College at Brockport.
equivalence of the ECRS-C to the English-language ECRS Brody, L. R., & Hall, J. A. (1993). Gender and emotion. In M. Lewis & J.
(Mallinckrodt & Wang, 2004), it is also possible that cultural Haviland (Eds.), Handbook of emotions (pp. 447 460). New York:
group differences found in this study were the result of remaining Guilford Press.
CULTURAL BELIEFS ABOUT IDEAL ADULT ATTACHMENT 203

Cassidy, J., & Shaver, P. R. (Eds.). (1999). Handbook of attachment: for clinical and community psychology (pp. 89 117). Chichester, En-
Theory, research, and application. New York: Guilford Press. gland: Wiley.
Cho, W., & Cross, S. E. (1995). Taiwanese love styles and their association Mallinckrodt, B., Porter, M. J., & Kivlighan, D. M., Jr. (2005). Client
with self-esteem and relationship quality. Genetic, Social, and General attachment to therapist, depth of in-session exploration, and object
Psychology Monographs, 121, 283309. relations in brief psychotherapy. Psychotherapy: Theory, Research,
Cohen, J. (1969). Statistical power analysis for the behavioral sciences. Practice, and Training, 42, 85100.
New York: Academic Press. Mallinckrodt, B., & Wang, C. (2004). Quantitative methods for verifying
Ditommaso, E., Brannen, C., & Burgess, M. (2005). The universality of semantic equivalence of translated research instruments: A Chinese
relationship characteristics: A cross-cultural comparison of different version of the Experiences in Close Relationships Scale. Journal of
types of attachment and loneliness in Canadian and visiting Chinese Counseling Psychology, 51, 368 379.
students. Social Behavior & Personality, 33, 57 68. Markus, H. R., & Kitayama, S. (1991). Culture and the self: Implications
Erikson, E. H. (1963). Childhood and society (2nd ed.). New York: Norton. for cognition, emotion, and motivation. Psychological Review, 98, 224
Feldt, L. S. (1969). A test of the hypothesis that Cronbachs alpha or 253.
Kuder-Richardson coefficient twenty is the same for two tests. Psy- Meyer, B., & Pilkonis, P. A. (2001). Attachment style. Psychotherapy:
chometrika, 45, 99 105. Theory, Research, Practice, and Training, 38, 466 472.
Fiske, A. P., Kitayama, S., Markus, H. R., & Nisbett, R. E. (1998). The Mikulincer, M., & Shaver, P. R. (2003). The attachment behavioral system
cultural matrix of social psychology. In D. T. Gilbert, S. T. Fiske, & G. in adulthood: Activation, psychodynamics, and interpersonal processes.
Lindzey (Eds.), The handbook of social psychology: Vol. 2 (4th ed., pp. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol.
915981). New York: McGraw-Hill. 35, pp. 53152). New York: Academic Press.
Flaherty, J. A., Gaviria, F. M., Pathak, D., Mitchell, T., Wintrob, R., & Miyake, K., Chen, S., & Campos, J. J. (1985). Infant temperament, moth-
Richman, J. A. (1988). Developing instruments for cross-cultural psy- ers mode of interaction, and attachment in Japan: An interim report.
chiatric research. Journal of Nervous and Mental Disorders, 176, 257 Monographs of the Society for Research in Child Development, 50(12),
263. 276 297.
Fraley, R. C., & Waller, N. G. (1998). Adult attachment patterns: A test of Moore, R. L. (1998). Love and limerence with Chinese characteristics:
the typological model. In J. A. Simpson & W. S. Rholes (Eds.), Attach- Student romance in the PRC. In V. C. de Munck (Ed.), Romantic love
ment theory and close relationships (pp. 77114). New York: Guilford
and sexual behavior: Perspectives from the social sciences (pp. 251
Press.
283). Westport, CT: Praeger Publishers.
Gao, G., Ting-Toomey, S., & Gudykunst, W. (1996). Chinese communi-
Posada, G., & Jacobs, A. (2001). Childmother attachment relationships
cation processes. In M. H. Bond (Ed.), The handbook of Chinese
and culture. American Psychologist, 56, 821 822.
psychology (pp. 280 293). New York: Oxford University Press.
Prishnee, M. A., & Poortinga, Y. H. (2005). Recalled early maternal
Gjerde, P. F. (2001). Attachment culture and amae. American Psychologist,
bonding and mother and self-related attitudes in young adult daughters:
56, 826 827.
A cross-cultural study in India and Belgium. International Journal of
Grossmann, K. E., Grossmann, K., Huber, F., & Wartner, U. (1981).
Psychology, 40, 324 338.
German childrens behavior toward their mothers at 12 months and their
Rastogi, M., & Wampler, K. S. (1999). Adult daughters perceptions of the
fathers at 18 months in Ainsworths Strange Situation. International
mother-daughter relationship: A cross-cultural comparison. Family Re-
Journal of Behavioral Development, 4, 157181.
lations, 48, 327336.
Grossmann, K. E., Grossman, K., & Kepler, A. (2005). Universal and
Rothbaum, F., Weisz, J., Pott, M., Miyake, K., & Morelli, G. (2000).
culture-specific aspects of human behavior: The case of attachment. In
Attachment and culture: Security in the United States and Japan. Amer-
W. Friedlmeier, P. Chakkarath, & B. Schwarz (Eds.), Culture and
human development: The importance of cross-cultural research for the ican Psychologist, 55, 10931104.
social sciences (pp. 7597). Hove, England: Taylor & Francis. Rothbaum, F., Weisz, J., Pott, M., Miyake, K., & Morelli, G. (2001).
Hambleton, R. K. (1993). Translating achievement tests for use in cross- Deeper into attachment and culture. American Psychologist, 56, 827
national studies. European Journal of Psychological Assessment, 9, 829.
57 68. Sagarin, B. J., & Guadagno, R. E. (2004). Sex differences in the contexts
Ho, D. Y. F. (1989). Continuity and variation in Chinese patterns of of extreme jealousy. Personal Relationships, 11, 319 328.
socialization. Journal of Marriage and the Family, 51, 149 163. Schmitt, D. P., Alcalay, L., Allensworth, M., Allik, J., Ault, L., Austers, I.,
Hui, C. H., & Triandis, H. C. (1985). Measurement in cross-cultural et al. (2004). Patterns and universals of adult romantic attachment across
psychology: A review and comparison of strategies. Journal of Cross- 62 cultural regions: Are models of self and of other pancultural con-
Cultural Psychology, 16, 131152. structs? Journal of Cross-Cultural Psychology, 35, 367 402.
Lonner, W. J., & Ibrahim, F. A. (1996). Appraisal and assessment in Searle, B., & Meara, N. M. (1999). Affective dimensions of attachment
cross-cultural counseling. In P. B. Pedersen, J. G. Draguns, W. J. styles: Exploring self-reported attachment style gender and emotional
Lonner, & J. E. Tremble (Eds.), Counseling across cultures (4th ed., pp. experience among college students. Journal of Counseling Psychology,
293322). Thousand Oaks, CA: Sage. 46, 147158.
Lopez, F. G., & Brennan, K. A. (2000). Dynamic processes underlying Simpson, J. A., & Rholes, W. S. (Eds.). (1998). Attachment theory and
adult attachment organization: Toward an attachment theoretical per- close relationship. New York: Guilford Press.
spective on the healthy and effective self. Journal of Counseling Psy- Singelis, T. M. (1994). The measurement of independent and interdepen-
chology, 47, 283300. dent self-construals. Personality and Social Psychology Bulletin, 20,
Malley-Morrison, K., You, H. S., & Mills, R. B. (2000). Young adult 580 591.
attachment styles and perceptions of elder abuse: A cross-cultural study. Singelis, T. M., Bond, M. H., Sharkey, W. F., & Lai, C. S. Y. (1999).
Journal of Cross-Cultural Gerontology, 15, 163185. Unpackaging cultures influence on self-esteem and embarrassability:
Mallinckrodt, B. (2001). Interpersonal processes, attachment, and devel- The role of self-construals. Journal of Cross-Cultural Psychology, 30,
opment of social competencies in individual and group psychotherapy. 315341.
In B. R. Sarason & S. Duck (Eds.), Personal relationships: Implications Sprecher, S., & Sedikides, C. (1993). Gender differences in perceptions of
204 WANG AND MALLINCKRODT

emotionality: The case of close heterosexual relationships. Sex Role, 28, Shaver (Eds.), Handbook of attachment: Theory, research, and clinical
511530. applications (pp. 713734). New York: Guilford Press.
Sue, D. W., Arredondo, P., & McDavis, R. (1992). Multicultural counsel- van IJzendoorn, M. H., & Sagi, A. (2001). Cultural blindness or selective
ing competencies and standards: A call to the profession. Journal of inattention. American Psychologist, 56, 824 825.
Counseling & Development, 70, 477 486. Vereijken, C. M. J. L., Riksen-Walraven, J. M., & Kondo-Ikemura, K.
Sue, D. W., & Sue, D. (1990). Counseling the culturally different: Theory (1997). Maternal sensitivity and infant attachment security in Japan: A
and practice (2nd ed.). New York: Wiley. longitudinal study. International Journal of Behavioral Development,
Sue, S. (1999). Science, ethnicity, and bias: Where have we gone wrong? 21, 35 49.
American Psychologist, 54, 1070 1077. William-Doherty, R., Hatfield, E., Thompson, K., & Choo, P. (1994).
Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th Cultural and ethnic influences on love and attachment. Personal Rela-
ed.). Needham Heights, MA: Allyn & Bacon. tionships, 1, 391398.
Triandis, H. C. (1996). The psychological measurement of cultural syn- You, H. S., & Malley-Morrison, K. (2000). Young adult attachment styles
dromes. American Psychologist, 51, 407 415. and intimate relationships with close friends: A cross-cultural study of
Trnavsky, P. (1998). Strange situation behaviors in Chinese infants. Child Koreans and Caucasian Americans. Journal of Cross-Cultural Psychol-
Study Journal, 28, 69 88. ogy, 31, 528 534.
Vacha-Hasse, T., & Thompson, B. (2004). How to estimate and inter-
pret various effect sizes. Journal of Counseling Psychology, 51,
475 481. Received March 7, 2005
van IJzendoorn, M. H., & Sagi, A. (1999). Cross-cultural patterns of Revision received January 5, 2006
attachment: Universal and contextual dimensions. In J. Cassidy & P. R. Accepted January 18, 2006

You might also like