Journal of Personality and Social Psychology Copyright 1997 by the American Psychological Association, Inc.

1997, Vol. 73, No. 5, 1092-1106 0022-3514/97/S3.00

Adult Attachment in a Nationally Representative Sample

Kristin D. Mickelson Ronald C. Kessler
Harvard Medical School and Institute for Social Research, Harvard Medical School
University of Michigan

Phillip R. Shaver
University of California, Davis

The explosion of adult attachment research in the last decade has been limited by its reliance on
college student and distressed samples. Using a large nationally representative sample of American
adults, the authors examined the relation of sociodemographics, childhood adversity, parental repre-
sentations, adult psychopathology, and personality traits to adult attachment in an effort to replicate
previous findings and extend the theory. Distribution of adult attachment styles was similar to that
in prior studies: 59% secure, 25% avoidant, and 11% anxious. Adult attachment was associated with
several sociodemographic variables (e.g., income, age, race) not previously studied. Childhood
adversities of an interpersonal nature were strongly related to insecure adult attachment. Various
types of adult psychopathologies and personality traits were also strongly related to adult attachment.
Implications for adult attachment theory and future research are discussed.

Beginning with Hazan and Shaver (1987; see Shaver & Clark, mew & Horowitz, 1991; Shaver, Collins, & Clark, 1996) and
1994, and Shaver & Hazan, 1993, for reviews), a number of (b) these models both shape an individual's beliefs about
social psychologists during the past decade have extended whether the self is worthy of love and whether others can be
Bowlby and Ainsworth's work (Bowlby, 1982; Ainsworth, trusted to provide love and support and also influence the kinds
Blehar, Waters, & Wall, 1978) on infant-caregiver attachment of interactions individuals have with others and their interpreta-
to the study of adult romantic attachment styles. The major tions of these interactions (Bowlby, 1973).
premises of this work are that (a) the emotions and relational Hazan and Shaver (1987, 1990) based their measure of adult
behavior patterns of adults, like those of infants, are guided romantic attachment on three patterns, or styles, of behavior
by internal working models of self and relationship partners identified in infants' relationships with their primary caregivers
constructed from prior relationship experiences (Bartholo- (usually mothers; see Ainsworth et al., 1978). The three styles
are often termed secure, avoidant, and anxious (or anxious-
ambivalent) . Adults with a secure attachment style are comfort-
Kristin D. Mickelson, Department of Health Care Policy, Harvard able depending on others and find it easy to get close to others.
Medical School, and Institute for Social Research, University of Michi-
Those with an avoidant attachment style are uncomfortable be-
gan; Ronald C. Kessler, Department of Health Care Policy, Harvard
Medical School; Phillip R. Shaver, Department of Psychology, University
ing close to others and find it difficult to trust them. Those with
of California, Davis. an anxious attachment style see others as reluctant to get close,
Preparation of this article was supported by National Institute of worry that others do not really care about them, and are often
Mental Health (NIMH) Training Grant T32-MH16806. The National viewed by others as clingy.
Comorbidity Survey (NCS) is a collaborative epidemiologic investiga- A number of empirical studies, using Hazan and Shaver's
tion of the prevalence, causes, and consequences of psychiatric morbidity (1987) or other measures of adult attachment, have found that
and comorbidity in the United States. The NCS is supported by the
the distribution of adult attachment styles is similar to those
NIMH (Grants MH46376, MH49098, and MH52861) with supplemen-
found for infants, such that approximately 55% of individuals
tal support from the National Institute on Drug Abuse (through a supple-
ment to NIMH Grant MH46376) and the W. T. Grant Foundation (Grant are classified as secure, 25% as avoidant, and 20% as anxious
90135190). Ronald C. Kessler is the principal investigator. A complete (see reviews by Shaver & Clark, 1994, and Shaver & Hazan,
list of NCS publications, study documentation, interview schedules, and 1993). Scores of cross-sectional studies have shown that these
the raw NCS public use data files can be obtained directly from the adult attachment styles are related to a broad array of social
NCS homepage by using the following URL: [Link] psychological variables, including relationship functioning
~ncsum/. We thank Vicki Helgeson and Chris Davis for their comments (Brennan & Shaver, 1995), personality (Collins & Read, 1990;
on drafts of this article and Bill Cunningham and Clark Brayton for Shaver & Brennan, 1992), depression (Carnelley, Pietromo-
preparation of the tables. naco, & Jaffe, 1994), social support (Simpson, 1990; Simpson,
Correspondence concerning this article should be addressed to Kristin Rholes, & Nelligan, 1992), religiosity (Kirkpatrick & Shaver,
D. Mickelson, Department of Health Care Policy, Harvard Medical
1990), substance use (Brennan & Shaver, 1995; Senchak &
School, 180 Longwood Avenue, Boston, Massachusetts 02115. Elec-
tronic mail may be sent via the Internet to mickelson@[Link].
Leonard, 1992), and domestic violence (Dutton, Saunders, Star-
[Link]. zomski, & Bartholomew, 1994). Whether attachment styles are

1092
 ADULT ATTACHMENT IN THE UNITED STATES 1093

causes, consequences, or merely concomitants of these corre- are exploratory in nature, are the first to be reported based on
lated variables is unclear from these largely cross-sectional stud- a broad sample of American adults.
ies. However, evidence from a smaller number of longitudinal
studies shows that attachment styles established early in life can
predict later social outcomes in childhood, adolescence, and Childhood Adversity
adulthood (see Rothbard & Shaver, 1994, for a review; and
Given the premise that adult attachment styles are based on
Waters, Merrick, Albersheim, & Treboux, 1995, for a recent
earlier interpersonal experiences, it is surprising that little prior
longitudinal study spanning 20 years), suggesting that they may
research has comprehensively evaluated the extent to which
play a causal role in relationship processes. Still, whether attach-
childhood traumas such as sexual abuse, physical violence, or
ment styles are causal is a matter of considerable debate and will
death of a parent are related to insecure romantic attachment.
be resolved only with long-term longitudinal or experimental-
However, a small number of studies have examined the relation
intervention studies.
of selective traumas to adult romantic attachment. Shaver and
Prior to conducting these studies, however, a more basic issue
Rubenstein (1980), prior to the operationalization of adult at-
needs to be settled—namely, whether these findings regarding
tachment styles, found that loss in childhood was linked to adult
adult attachment can be replicated and extended in the general
loneliness. Among women, sexual abuse and incest have been
population. A major limitation of extant research on adult ro-
found to be related to anxious attachment in relationships, less
mantic attachment styles is that the vast majority of studies
willingness to depend on others, and more reluctance to get
have been based on samples of college students, with the few
close to others (Alexander, 1993; Mallinckrodt, McCreary, &
remaining studies being focused either on volunteers recruited
Robertson, 1995). Among college students, poor parental mari-
from newspaper ads (Hazan & Shaver, 1987) or distressed
tal quality (but not divorce) and parental drinking problems
adults such as incest survivors (Alexander, 1993) or alcoholics
have been found to be related to insecure romantic attachment,
(Joyce, Sellman, Wells, & Frampton, 1994). It is not clear
especially avoidant attachment (Brennan & Shaver, 1993; Bren-
whether the results based on such samples also apply to the
nan, Shaver, & Tobey, 1991). Finally, it should be noted that
general adult population. The purpose of the present article is
researchers using the Adult Attachment Interview (Main,
to address this problem by examining the distribution and corre-
Kaplan, & Cassidy, 1985) have more fully examined the nega-
lates of adult attachment styles in a large nationally representa-
tive consequences of childhood traumas on adult attachment
tive sample of adults. Five broad classes of theoretically relevant
representations (e.g., Cicchetti, Toth, & Lynch, 1995; Lutz &
correlates were examined: sociodemographic variables, retro-
Hock, 1995). These studies have shown that childhood mal-
spective reports of childhood adversities, retrospective reports
treatment of various kinds has significant consequences for adult
of childhood relationships with parents, adult psychopathology,
attachment organization. But the focus of this research tends to
and adult personality traits.
be more on how the resolution of these adversities affects future
parental caregiving rather than on how the childhood experi-
Sociodemographics ences affect adult romantic attachment. In an effort to extend
prior research, we examined the relation of a detailed set of
Possibly because of sample restriction, previous research has retrospectively reported childhood adversities, including separa-
failed to identify important sociodemographic correlates of tion-loss events, interpersonal traumas, parental psychopatholo-
adult attachment styles. For example, although there is sugges- gies, and other nonsocial traumas to the adult attachment styles
tive evidence from a few previous studies that married respon- of our respondents.
dents are more likely than nonmarrieds to report a secure attach-
ment style (e.g., Hazan & Shaver, 1987), most studies have
focused on college students (the majority of whom are single) Parental Bonding
and consequently have not investigated correlations between
adult attachment and marital status. Although most college stu- Adult romantic attachment styles have been shown to relate
dent studies have failed to find sex differences in adult attach- systematically to mental representations of parents (Bringle &
ment styles (Shaver & Hazan, 1993; Shaver, Papalia, et al., Bagby, 1992; Hazan & Shaver, 1987; Levy, Blatt, & Shaver,
1996), this might reflect the fact that sex differences in many 1996; Rothbard & Shaver, 1991). Whereas secure attachment
domains are smaller in samples of college students than in gen- is related to warm, positive representations of parents, avoidant
eral population samples. Presumably, this finding is due to selec- attachment is related to cool and rejecting representations, and
tion bias coupled with the fact that the roles of men and women anxious attachment is related to a mixture of positive and nega-
are more equal among college students than in the rest of the tive representations. Another common finding is that, among
adult population (Tavris & Wade, 1984). We have been unable college students, the relationship with one's mother is related
to locate any published studies that examined the relation of more strongly to adult attachment styles than the relationship
adult attachment styles to other important sociodemographic with one's father—often termed the maternal strength hypothe-
variables such as race or socioeconomic status (SES). We do sis (Brennan & Shaver, 1993; Levy et al., 1996; Main et al.,
know, however, that children in lower SES homes experience 1985). To determine whether these previous findings would rep-
less continuity in their attachments than children in higher SES licate in a large representative sample, we used a common mea-
homes (see review by Shaver & Hazan, 1993), suggesting that sure of adults' retrospective representations of relationships with
lower SES may be related to insecure attachment. Our analyses their parents, the Parental Bonding Instrument (PBI; Parker,
of these sociodemographic correlates of adult attachment, which Tupling, & Brown, 1979).
1094 MICKELSON, KESSLER, AND SHAVER

Adult Psychopathology with an external locus of control quit a job or task sooner after
encountering an obstacle than those with an internal locus of
Although there is a substantial comorbidity between mood control (similar to insecurely attached individuals' higher
and substance abuse disorders (Kessler et al., 1997), there are breakup rates). Thus, we included locus of control in our analy-
hints in the attachment literature that mood disorders (with or ses of personality and adult attachment with the expectation
without substance abuse disorders) may be more related to anx- that internal locus of control would be related to secure adult
ious than to avoidant attachment, whereas substance abuse dis- attachment and external locus of control would be related to
orders (with or without mood disorders) may be more related insecure adult attachment.
to avoidant than to anxious attachment. However, as with child- To summarize, unlike prior research, our study examined
hood adversity, most prior studies have neglected to search for adult attachment in a nationally representative sample of adults.
such patterns. Hazan and Shaver (1990) found that adults with Our purpose was twofold: (a) to see if previous findings, based
insecure attachments reported more depression and physical mainly on college students, would replicate in a large representa-
symptoms than those with secure attachments. Carnelley et al. tive sample of adult Americans and (b) to determine whether,
(1994) found that mildly depressed women were insecure in when sufficient variation is present, adult attachment styles are
their romantic relationships. Furthermore, other studies have related to five major categories of variables, namely, sociodemo-
found depression to be linked specifically with anxious attach- graphics, childhood adversity, adult representations of childhood
ment (Pettem, West, Mahoney, & Keller, 1993; Zuroff & Fitzpa- parental bonding, adult psychopathology, and adult personality.
trick, 1995). Patrick, Hobson, Castle, and Howard (1994) found Although our research was primarily exploratory, given the
that other disorders are also linked with specific attachment dearth of prior representative studies, it was guided by attach-
styles: Patients with borderline personality disorder were more ment theory and by previous findings obtained mainly with col-
likely to have an anxious attachment style, whereas those with lege samples.
dysthymia were more likely to have an avoidant attachment
style. With respect to substance abuse disorders, Senchak and Method
Leonard (1992) found that men with an avoidant attachment
Sample
style were more likely to be heavy drinkers than men with a
The data came from the National Comorbidity Survey (NCS; Kessler
secure or anxious attachment style. We examined the relation
et al., 1994). The NCS is a nationwide household survey of the U.S.
of four categories of disorders: mood, anxiety, substance abuse,
population between ages 15 and 54, designed to produce data on the
and other disorders (schizophrenia, conduct, antisocial) to adult prevalence and correlates of psychiatric disorders. The sample was based
attachment. on a stratified, multistage area probability sampling frame of the nonin-
stitutionalized civilian population in the 48 coterminous states, with a
supplemental sample of students living in campus group housing. The
Adult Personality Traits 8,098 respondents who participated in the NCS were selected using
probability methods from 1,205 block-level segments. The segments
Shaver and Brennan (1992), in a study of the Big Five person- were created within a stratified sample of small areas in 172 counties
ality traits, found that, among college students, secure individu- in 34 states throughout the United States. The survey was administered
als were less neurotic and more extroverted than avoidant and face-to-face in the homes of respondents by trained interviewers. The
anxious individuals. Secure individuals were also more agree- response rate was 82.4%.
able than avoidant individuals. Moreover, self-esteem has been Respondents in the 15-24 age range were oversampled compared to
found to be higher in securely attached individuals than in inse- older respondents in order to guarantee that the number of respondents
curely attached individuals (e.g., Collins & Read, 1990; in this age range would be roughly equal to the numbers in each of the
Feeney & Noller, 1990). We assessed neuroticism, extroversion, other age ranges: 25-34, 35-44, and 45-54. These respondents would
openness to experience, and self-esteem to see if we could find otherwise have been underrepresented because so many of them live
with their parents.' The survey also included a supplemental sample of
similar relations to adult attachment style in a representative
students living in campus group housing because there are approximately
sample.
two million such students in the United States who would be unrepre-
Connections between attachment styles and forms of religios- sented in a more conventional household survey that excluded persons
ity have been suggested by Kirkpatrick (e.g., 1994). In prelimi- living in institutional housing (U.S. Bureau of the Census, 1991 ). 2
nary studies, secure adults had the highest probability of a se-
cure sense of God, as if models of parents had been extended 1
The oversampling of 15-24-year-old respondents was carried out in
to include God as a supportive ' 'father,'' whereas avoidant adults a random 25% of households throughout the entire field period. This
were more likely to be atheists, and anxious adults were more fractional oversampling scheme was established on the basis of analysis
likely to engage in emotional expressions of religiosity, such as of household composition data from the U.S. National Health Interview
speaking in tongues (Kirkpatrick & Shaver, 1990, 1992). We Survey (NHIS; U.S. Department of Health and Human Services, 1992)
assessed these traits to see whether previous findings would and was designed to generate a sample with roughly equal numbers of
respondents in the four 10-year age ranges 15-24, 25-34, 35-44, and
replicate in a representative sample.
45-54.
Although prior studies have not directly examined the relation 2
We did not include supplemental samples of other institutional popu-
of locus of control to adult attachment, such a relation has been lations (e.g., prisons, hospitals, nursing homes) or of the homeless popu-
suggested. Miller, Lefcourt, Holmes, Ware, and Saley (1986) lation both because of the expense of doing so and the fact that these
found that adults with an internal locus of control dealt with people make up such a small proportion of the total U.S. population in
marital problems directly (hence, perhaps more securely), the target age range of the NCS that their inclusion would have no
whereas Seligman and Schulman (1986) showed that people meaningful effect on the total population estimates.
 ADULT ATTACHMENT IN THE UNITED STATES 1095

Table 1
NHIS-NCS Part I Demographic Comparisons (N = 8,098}
U.S. population NCS weighted NCS unweighted
Variable (NHIS; %) (%) (%)

Sex
Male 49.1 49.5 47.5
Female 50.9 50.5 52.5
Age (years)
15-24 25.5 24.7 21.8
25-34 30.8 30.1 32.4
35-44 25.9 27.1 27.7
45-54 17.8 18.1 18.1
Marital status
Married-cohabiting 59.8 62.9 54.4
Divorced-separated-widowed 10.1 10.0 15.5
Never married 30.1 27.1 30.1
Race
White 75.0 75.3 75.1
Black 11.9 11.5 12.5
Hispanic 8.6 9.7 9.1
Other 4.5 3.5 3.3
Education (years)
0-11 22.5 22.3 18.2
12 36.8 37.4 33.1
13-15 21.2 21.7 26.3
^16 19.5 18.6 22.4
Region
Midwest 24.6 23.8 25.6
Northeast 20.0 20.2 19.2
West 21.7 19.6 19.6
South 33.7 36.4 35.6
Urbanicity
Major metropolitan 71.2 67.8 68.9
Other urban 8.1 7.5 6.5
Rural 20.7 24.7 24.6
Total N 65,244 8,098 8,098

Note. NHIS = National Health Interview Survey; NCS = National Comorbidity Survey.

The data were weighted to adjust for variation in probabilities of respondents ages 15-24, (b) all older Part 1 respondents who were
selection across households and within households. The data were post- positive on initial questions in one or more diagnostic sections of the
stratified by means of an iterative procedure to approximate the national interview (possible cases), and (c) a one-in-six random subsample of
population distributions of the cross-classification of age, sex, race- all remaining Part 1 respondents. The Part 2 interview took an average
ethnicity, marital status, education, living arrangements, region, and ur- of 68 min to administer and, in most cases, was administered during
banicity as defined by the 1989 U.S. NHIS.3 the same interview session, although a separate visit or telephone inter-
To demonstrate the representativeness of the sample, Table 1 reports view was sometimes required.
the distributions of the NCS along a variety of sociodemographic dimen-
sions and compares the distributions with population distributions ob-
tained from the 1989 U.S. NHIS. The NHIS was used as a population Interviewers and Interviewer Training
comparison rather than the U.S. Bureau of the Census Current Population
Survey because it includes a much wider array of sociodemographic All interviewers went through a rigorous program of training in gen-
variables for purposes of poststratification. The second column in Table eral interviewing technique (Guenzel, Berckmans, & Cannell, 1983), a
1 reports the adjusted distributions after poststratification, whereas the 7-day study-specific training program for the NCS, and periodic re-
third column reports the NCS distributions before poststratification. fresher courses. Supervisors reviewed and edited all interviews, sending
Clearly, our sample is more representative of American adults than any back errors or omissions to interviewers, who recontacted respondents
group studied previously by adult attachment researchers. to obtain the necessary information. Supervisors also recontacted a sam-

Procedure 3
A nonresponse survey was carried out in which selected nonrespon-
A two-phase sample design was used in the NCS. In the first phase, dents were offered a substantial financial incentive to complete a short
the Part 1 diagnostic interview was administered to all 8,098 respon- form of the diagnostic interview. Significantly higher rates of both life-
dents. This interview took an average of 55 min to administer. In the time and current psychiatric disorders were found among these initial
second phase, a Part 2 risk factor interview was administered to a nonrespondents than among respondents in the main survey. The survey
probability subsample of 5,877 respondents consisting of (a) all Part 1 data were weighted to compensate for this nonresponse.
1096 MICKELSON, KESSLER, AND SHAVER

pie of respondents throughout the study to verify responses and guaran- cluded being kidnapped or held captive), rape, sexual molestation, child-
tee high quality interviewer performance. hood pregnancy,5 or the individual witnessed interpersonal trauma be-
tween the parents (poor parental marital quality, parents violent to each
other). Parental psychopathology refers to an individual's father or
Measures
mother being reported by the respondent to have suffered from depres-
Attachment style. Adult attachment style was measured in two re- sion, anxiety, substance abuse, suicidal behavior, and/or antisocial be-
lated ways: qualitative (categorical) and quantitative (scaled), using havior during the individual's childhood. The fourth and final category
Hazan and Shaver's (1987) attachment style measure. Each respondent contained adverse events that could not be classified as separation-loss,
provided a self-rating on three scales: secure, avoidant, and anxious interpersonal, or parental psychopathology (e.g., financial adversity, life-
(see Shaver & Brennan, 1992, for a precedent). The ratings were made threatening accident, natural disaster, witnessing a trauma).
on 4-point scales, with 1 = not at all like me, 2 = a little like me, 3 = The operational definitions of these adversities were based primarily
somewhat like me, and 4 = a lot like me. The three scales were signifi- on previously developed instruments. In all cases our goal was to select
cantly correlated: secure was negatively correlated with the two insecure instruments that were already validated and known to be subject to the
scales (avoidant, r = - . 3 8 ; anxious, r = - . 1 2 ) ; avoidant and anxious least possible recall bias. The majority of the adversities were taken
were positively correlated (r = .27). These rating scales were used in from the Posttraumatic Stress Disorder (PTSD) section of the Diagnostic
correlation and regression analyses. Interview Schedule (DIS; Robins, Helzer, Croughan, & Ratcliff, 1981).
Because a self-categorization of attachment style was not requested Rather than use the administration procedure in the DIS, however, a
in the interview, an inferential qualitative (categorical) style assessment strategy of listing the traumas one by one (used by Resnick, Kil-
was derived to examine the distribution of attachment by sociodemo- patrick, & Lipovsky, 1991) was adopted. The traumas also were referred
graphic characteristics. Our coding of attachment style was derived as to by number rather than name (e.g., "Did event number five ever
follows. If one of the respondent's three ratings was higher than the happen to you?" rather than "Were you ever raped?") because previous
other two (i.e., if one was closer to a lot like me), he or she was assigned research on sensitive items suggests that this questioning method im-
to the corresponding attachment category (secure, avoidant, or anxious, proves accuracy (Bradburn & Sudman, 1979). Parental divorce-separa-
whichever received the uniquely high score). Such an assignment was tion, long absence of a parent, and parents' marital quality were based on
possible for 6,387 respondents. If, instead, a respondent recorded two an expansion of standard Institute for Social Research (ISR) questions
relatively high scores that were tied, and one of the high categories was concerning childhood living arrangements and similar questions used in
secure, the respondent was assigned to the equally high insecure cate- the Home Environment Interview (Robins, 1985). Parental violence
gory, on the assumption that, if anything, the self-ratings might be biased questions were taken from the Conflict Tactics Scale (Straus, 1979),
in the secure direction (Brennan, Clark, & Shaver, in press). This coding the most widely used measure of family violence. Measures of parental
rule was applied to 1,087 respondents, 721 assigned to the avoidant psychopathology were taken from the Family History Research Diagnos-
category and 366 assigned to the anxious category. If a respondent had tic Criteria interview (Endicott, Andreason, & Spitzer, 1978), a widely
two relatively high and tied insecure ratings (i.e., if a person's avoidant used method for obtaining informant reports about the mental health
and anxious ratings were tied and were higher than the secure rating), of relatives. Financial adversity during childhood was assessed with a
he or she was assigned to the less common anxious category, in order subjective measure asking individuals to rate whether their childhood
not to muddy the avoidant category with people who were in fact rela- family's standard of living was better, similar, or worse than those in their
tively eager for closeness and intimacy. This coding rule was applied to community (relative financial adversity). Two objective assessments of
242 respondents.4 Finally, 364 respondents gave the same rating to all a respondent's standard of living during childhood were also included:
three attachment-style descriptions; no attempt was made to assign them education and occupation of the individual's major financial supporter
to one of Hazan and Shaver's (1987) three categories. These individuals during childhood.
were labeled unclassified. Of the 7,716 respondents who could be as- Parental bonding. A subset of questions concerning parent-child
signed to an attachment category using the rules just described, only relationships was taken from the Parental Bonding Instrument (PB1;
approximately 15% were assigned to a category on the basis of tie- Parker et al., 1979), a widely used scale that assesses current representa-
breaking procedures; the other 85% provided one rating that was clearly tions of a respondent's childhood relationship with parents. In the NCS,
higher than the other two. Therefore, our tie-breaking procedure did not three dimensions were assessed separately for mother and father:
affect the results very much. Furthermore, as mentioned above, this warmth, overprotectiveness, and consistency.
categorization of attachment styles was used only to examine the distri- Diagnostic assessment. Diagnostic and Statistical Manual of Men-
bution of attachment by sociodemographic characteristics; all other anal- tal Disorders (DSM-III-R; American Psychiatric Association, 1987)
yses were performed on the continuous rating scales. diagnoses were based on a modified version of the Composite Interna-
Sociodemographics. Eight sociodemographic characteristics were tional Diagnostic Interview (CIDI; World Health Organization [WHO],
assessed: sex, age, marital status, race, income, education, region, and
urbanicity. Region refers to the geographic region in which a person
4
lives (Midwest, Northeast, West, South). Urbanicity refers to the size Analyses using the categorical classifications were also performed
of the population where a person lives (major metropolitan, other urban- without these respondents; the results remained the same, suggesting
suburban, rural). that the findings were not influenced by our coding decision.
5
Childhood adversities. For purposes of the present study, adversities Although we acknowledge that childhood pregnancy is sometimes
were classified as occurring in childhood if the adverse event was experi- the result of a consensual sexual relationship, our survey did not distin-
enced prior to the age of 16 years. We examined four kinds of adversities: guish between desired and undesired childhood pregnancy. However,
separation-loss events, interpersonal traumas, parental psychopatholo- past research has shown that childhood pregnancy is often a negative
gies, and other nonsocial traumas. Separation—loss events were defined experience. For example, Beardslee, Zuckerman, Amaro, and McAllister
as events in which the individual experienced loss of a parent or child (1988) found that adolescent pregnancy was significantly related to
prior to the age of 16 years (death of a parent, parental divorce-separa- depression, with the rate of depression being approximately twice that
tion, long absence (6 months or more) of a parent, childhood miscar- found in prior studies of nonpregnant adolescents. In addition, onset for
riage, childhood abortion). Interpersonal traumas were defined as events depression occurred most often during pregnancy and continued long
in which another person directly harmed the individual: serious assault, after the pregnancy. In general, we believe that childhood pregnancy is
physical abuse, serious neglect, threatened with a weapon (which in- reasonably classified as a childhood adversity.
 ADULT ATTACHMENT IN THE UNITED STATES 1097

1990a). The CIDI is a structured diagnostic interview developed in a Results

collaborative project between WHO and the Alcohol, Drug Abuse, and
Mental Health Administration (Robins, Wing, Wittchen, & Helzer,
Distribution of Adult Attachment Styles
1988). The CIDI was developed in modular form to allow researchers
to use only a subset of diagnostic sections.6 WHO field trials of the
As shown in the first row of Table 2,59.0% of the respondents
CIDI have documented excellent interrater reliability (Wittchen et al.,
1991), good test-retest reliability (Wacker, Battegay, Mullejans, & were classified as secure, 25.2% as avoidant, and 11.3% as
Schlosser, 1990), and good validity relating to concordance with clinical anxious; 4.5% were unclassified because their three attachment
diagnoses (Spengler & Wittchen, 1989). The NCS used the sections style rating scores were equal. This distribution was found to
that yield four groups of diagnoses: mood disorders (major depressive vary significantly as a function of each of the eight sociodemo-
episode, manic episode, dysthymia), anxiety disorders (panic disorder, graphic variables included in the remainder of the table. Respon-
agoraphobia, social phobia, simple phobia, PTSD, generalized anxiety), dents classified as secure were more likely than other respon-
psychoactive substance use disorders (alcohol abuse, alcohol depen- dents to be female, older, married, white, better educated, and
dence, drug abuse, drug dependence), and other disorders (schizophre- better off financially. Respondents classified as avoidant were
nia, conduct disorder, antisocial disorder). For a fuller discussion of the more likely than other respondents to be male, in the age range
diagnostic assessment used in the NCS, see Kessler et al. (1994). 25-44, married or previously married (as opposed to never
Personality measures. Several personality traits were assessed in the married), and to be either Black or "other." Respondents classi-
NCS: self-esteem, locus of control, neuroticism, extroversion, openness
fied as anxious were more likely than other respondents to be
to experience, and religiosity. Given the enormous scope of the NCS,
young, previously married, Black or Hispanic, less well edu-
however, it was not possible to use full measures of these constructs.
cated, and less well off financially. No clear pattern of associa-
Instead, short forms of the scales were selected based on data analyses
from previous studies designed to select the best subset of items to tion between attachment styles and region or urbanicity was
operationalize an entire scale. A short form of the Rosenberg (1965) discernible.
self-esteem scale was used (a = .89), as well as short forms of Leven- To gain an understanding of how these predictors function
son's (1973) locus of control scales (internal, a = .63; luck, a = .69; jointly in a metric sense (rather than an explained variance
powerful others, a = .76) and Goldberg's (1992) neuroticism ( o = sense), we constructed a composite profile of the "security-
.89), extroversion (a = .84), and openness to experience (a = .80) enhancing" characteristics. Using only respondents who could
scales. All scores on these scales were standardized prior to analysis. be classified as secure, avoidant, or anxious (i.e., excluding
The religiosity measures comprised items used for many years in unclassified respondents), the seven characteristics most
surveys at the ISR. Christian fundamentalism was operationalized by strongly associated with the secure attachment category (female,
three dichotomous questions; (a) "Have you been 'born again,' that is,
married-cohabiting, age 45 or older, white, 13+ years of educa-
had a turning point in your life when you committed yourself to Jesus
tion, $20,000+ of income, from the Midwest) were combined
Christ?"; (b) "Do you ever try to encourage people to believe in Jesus
into a profile measure in which individuals were assigned a
and accept Him as their Savior?"; (c) "Please tell me whether you
agree or disagree with the following statement: 'The Bible is the actual number based on how many of these security-enhancing charac-
Word of God and is to be taken literally, word for word.' " Responses teristics they possessed. When the distribution of attachment
were summed and standardized to obtain a fundamentalism score (a = styles was re-examined using this profile measure, three note-
.68). Overall religiosity was assessed with four questions: (a) "In gen- worthy trends were found. First, there was a monotonic increase
eral, how important are religious or spiritual beliefs in your daily life in the proportion of secures as the number of security-enhancing
(very, somewhat, not very, or not at all)?"; (b) "How often do you characteristics increased. Although only half of those with none
usually attend religious services (more than once a week, about once a of the characteristics were classified as secure, close to three-
week, one to three times a month, less than once a month, or never)?"; quarters of individuals with five or more characteristics were
(c) "When you have problems or difficulties in your family, work, or classified as secure. Second, there was little change in the pro-
personal life, how often do you seek spiritual comfort (almost always, portion of avoidants as the number of characteristics increased
often, sometimes, rarely, or never)?"; (d) "When you have decisions (none of the characteristics: 20.3%; five or more characteristics:
to make in your daily life, how often do you ask yourself what God
21.3%). Third, there was a sharp decline in the proportion
would want you to do (almost always, often, sometimes, rarely, or
of anxious respondents as the number of security-enhancing
never)?" Responses were summed and standardized to obtain an overall
characteristics increased. Approximately 30% of people who
religiosity score (a = .85), with higher scores indicating greater
religiosity. did not possess any of the characteristics were classified as
anxious compared with only 6% of those who possessed five or
more of the characteristics.

Analysis Procedures

The data were analyzed using linear and logistic regression analysis Childhood Adversity
with the attachment scales (in the linear models) or classifications (in
the logistic models) as the outcome variables. As a result of the complex The remainder of the analyses were performed on the three
sample design and weighting, estimates of standard errors were obtained attachment rating scales, with the scores standardized prior to
using the method of Jackknife Repeated Replication (Rust, 1985). An
SAS macro was used to implement this procedure by computing esti-
mates in each of 42 subsample pseudoreplicates and manipulating these
estimates to arrive at design-based standard errors. These estimates take 6
Diagnoses were generated with the CIDI diagnostic program (World
into account both the clustering and weighting in the study's design. Health Organization, 1990b).
1098 MICKELSON, KESSLER, AND SHAVER

Table 2
Distribution of Adult Attachment Types by Sociodemographic Characteristics
Secure Avoidant Anxious Unclassified

Variable N SE SE SE SE X2 df p

Total sample 8,080 59.0 1.0 25.2 0.7 11.3 0.6 4.5 0.3
Sex
Male 3,997 56.8b 1.1 27.6. 1.0 10.8, 0.7 4.8, 0.5
Female 4,083 61.2. 1.2 22.8b 0.8 11.7, 0.7 4.2, 0.4 28.0 3 .001
Age (years)
15-24 2,000 58.7b 1.3 19.8C 1.3 17.4. 1.2 4.2, 0.6
25-34 2,435 56.9b 1.5 27.9, 1.2 10.7b 0.9 4.6, 0.6
35-44 2,189 58.7b 1.6 28.4, 1.2 8.6b.c 0.8 4.3, 0.7
45-54 1,456 63.6, 1.7 23.3 b 1.5 8.0c 1.0 5.1. 0.9 145.5 9 .001
Marital status
Married—cohabiting 5,077 61.4, 1.1 26.5. 0.9 8.1 b 0.6 4.1 b 0.4
Divorced - separated - widowed 811 48.2C 2.5 25.7, 1.7 18.2. 1.6 7.9. 1.3
Never married 2,192 57.6b 1.2 22. l b 1.0 16.2, 0.9 4.2 b 0.6 184.9 6 .001
Race
White 6,087 60.8a,b 1.0 25.0,,b 0.8 10.2b 0.6 4.1 b , c 0.3
Black 926 5O.8c 2.9 28.1, 2.2 16.0, 1.5 5.2C 1.1
Hispanic 783 57.8b 3.0 21.0,, 2.3 15.3, 2.1 5.9C 1.2
Other 284 51.9b,c 3.4 31.5, 3.8 9.2b 2.1 7.4,,c 2.2 79.2 9 .001
Income ($)
0-19,999 2,053 49.2b 1.8 26.6, 1.2 18.3, 1.4 6.0, 0.6
20,000-34,999 1,985 61.1, 1.4 24.0, 1.2 1.0 3.9b 0.5
35,000-69,999 2,851 63.0, 1.2 24.1, 1.0
n.o
8.8bb 0.8 4.1 b 0.5
2=70,000 1,190 63.0, 2.3 27.4, 2.2 5.7C 0.8 3.9b 1.1 203.7 9 .001
Education (years)
0-11 1,802 55.4C 1.7 21.7C 1.3 16.9, 1.4 6.0, 0.8
12 3,020 55.0c 1.5 27.8, 1.0 11.9b 1.0 5.4. 0.6
13-15 1,756 61.4b 1.8 25.7,,b 1.5 9.9 b 1.1 3.1 b 0.6
^16 1,502 68.8, 1.6 23.8b,c 1.5 5.0c 0.5 2.5 b 0.5 199.1 9 .001
Region
Midwest 1,632 62.0, 2.3 23.1, 1.7 9.7, 1.5 5.2, 0.8
Northeast 1,918 59.4.,b 1.2 25.4, 0.7 12.1, 1.0 3.1 b 0.7
West 2,947 57.2b 1.7 25.6, 1.2 12.1, 0.9 5.1, 0.6
South 1,582 58.9,,b 2.4 26.3, 1.7 10.3, 1.1 4.5,, b 0.6 28.2 9 .001
Urbanicity
Major metropolitan 3,647 59.0, 1.6 25.2, 1.1 11.9, 1.0 3.9b 0.5
Other urban 2,688 59.8, 1.7 24.7, 1.3 10.9, 1.2 4.5,, b 0.5
Rural 1,744 57.9, 1.1 25.7, 1.0 10.6, 0.7 5.8, 0.9 12.9 6 .044

Note. Common subscripts in each column block indicate that means do not significantly differ a t p < .05.

analysis.7 Linear regressions were performed on each childhood parents' marital quality as poor, and witnessing violence be-
adversity to predict the three attachment ratings. Table 3 reports tween parents were all related negatively to the secure rating.
the metric regression coefficients and corrected standard errors All nine interpersonal traumas were related positively to the
for each adversity and the three attachment scales. The first avoidant rating, and all but serious assault were related posi-
column reports the proportion of the population who experi- tively to the anxious rating. None of the analyses involving the
enced each adversity. For separation-loss events, only one ad- avoidant-anxious difference score were significant. In general,
versity was found to relate significantly to any of the attachment then, the effects in the interpersonal traumas section of Table 3
scales: Parental divorce-separation was related negatively to were mostly due to differences between securely and insecurely
the secure attachment rating and positively to the anxious attach-
ment rating. To test whether parental divorce-separation related
7
differently to the two forms of insecure attachment, the differ- Parallel analyses (i.e., proportions, logistic regressions) were per-
ence between the avoidant and anxious ratings was used as a formed on the three attachment categories for each predictor variable.
These results are not discussed here; they were similar to those found
dependent variable in a regression analysis. The difference be-
with the continuous scales, but the linear regression results were stronger
tween the avoidant and anxious slopes was marginally signifi-
and more consistent. Moreover, the continuous ratings more accurately
cant (p < .10), such that parental divorce-separation tended reflect the individuals' self-perceptions as opposed to the constructed
to be related more to anxious than to avoidant attachment. categories, which forced individuals into one category if two attachment
As shown in Table 3, most of the interpersonal traumas were scores were tied. If interested, readers may obtain tables of the categori-
significantly related to the attachment ratings. Physical abuse, cal analyses from Kristin Mickelson or from the NCS homepage (see
serious neglect, being threatened with a weapon, perceiving author note).
 ADULT ATTACHMENT IN THE UNITED STATES 1099

Table 3
Association Between Adult Attachment Ratings and Childhood Adversities
Avoidant-
Secure Avoidant Anxious anxious"
Adversity % SE b SE b SE b SE b SE

Separation-loss
Childhood miscarriage 0.36 0.1 -0.01 0.20 0.18 0.18 0.12 0.29 0.05 0.23
Childhood abortion 0.44 0.1 0.25 0.17 0.31 0.31 -0.01 0.26 0.32 0.49
Parental divorce-separation 18.46 0.7 -0.15** 0.05 0.07 0.06 0.18*** 0.04 -0.11 0.06
Death of a parent
Mother 3.35 0.4 -0.01 0.10 0.13 0.11 0.18 0.14 -0.05 0.16
Father 5.86 0.5 0.11 0.10 -0.10 0.08 0.02 0.08 -0.11 0.12
Mother-father11 -0.12 0.14 0.22 0.13 0.16 0.15
Long absence (>6 months)
Mother 0.50 0.1 -0.24 0.17 -0.05 0.31 0.19 0.31 -0.25 0.20
Father 6.57 0.4 0.10 0.08 -0.08 0.07 0.08 0.07 -0.16 0.07
Mother-father -0.34* 0.17 0.03 0.32 0.11 0.31
Interpersonal traumas
Serious assault 2.27 0.2 -0.18 0.11 0.21* 0.10 0.16 0.09 0.05 0.11
Physical abuse 6.05 0.4 -0.19* 0.08 0.32*** 0.07 0.31*** 0.06 0.00 0.08
Serious neglect 2.82 0.3 -0.32*** 0.08 0.61*** 0.11 0.75*** 0.09 -0.14 0.12
Threatened with a weapon 3.13 0.3 -0.34*** 0.08 0.40*** 0.08 0.24** 0.09 0.16 0.09
Rape 1.88 0.2 -0.19 0.12 0.39*** 0.11 0.52*** 0.13 -0.13 0.14
Sexual molestation 6.24 0.4 -0.03 0.06 0.18** 0.06 0.25** 0.08 -0.08 0.10
Childhood pregnancy 0.66 0.1 -0.19 0.22 0.54** 0.18 0.58** 0.22 -0.04 0.20
Parents' relationship
Poor marital quality 10.97 0.6 -0.21** 0.07 0.20*** 0.06 0.16** 0.06 0.04 0.05
Parents violent to each other 18.76 0.7 -0.19*** 0.05 0.29*** 0.05 0.33*** 0.05 -0.03 0.06
Parental psychopathology
Depression
Mother 22.68 1.0 -0.02 0.04 0.20*** 0.03 0.10* 0.04 0.10* 0.04
Father 14.44 0.6 -0.03 0.05 0.14** 0.05 0.10* 0.04 0.04 0.07
Mother-father6 0.01 0.06 0.06 0.07 -0.00 0.05
Anxiety
Mother 17.31 0.9 -0.07 0.04 0.25*** 0.04 0.18*** 0.04 0.07 0.05
Father 12.26 0.7 0.01 0.06 0.15** 0.05 0.10* 0.05 0.05 0.06
Mother-father -0.08 0.07 0.11 0.06 0.08 0.05
Substance abuse
Mother 6.28 0.5 -0.07 0.07 0.17* 0.07 0.07 0.07 0.09 0.07
Father 18.70 0.9 -0.05 0.04 0.17*** 0.05 0.08 0.05 0.09 0.05
Mother-father -0.02 0.07 -0.00 0.07 -0.00 0.07
Suicidal behavior
Mother 3.05 0.3 -0.03 0.09 0.12 0.09 0.26* 0.13 -0.14 0.11
Father 1.57 0.3 0.06 0.14 0.36* 0.17 0.39** 0.14 -0.03 0.27
Mother-father -0.09 0.16 -0.24 0.20 -0.13 0.18
Antisocial behavior
Mother 1.20 0.1 -0.26** 0.11 0.57*** 0.15 0.41* 0.17 0.16 0.18
Father 4.84 0.4 -0.24* 0.11 0.14 0.08 0.10 0.09 0.04 0.11
Mother-father -0.03 0.16 0.43** 0.16 0.31 0.17
Other traumas
Life-threatening accident 4.85 0.3 -0.17 0.09 0.12 0.08 0.19* 0.08 -0.07 0.09
Natural disaster 7.35 0.4 -0.04 0.08 0.14* 0.07 0.05 0.04 -0.08 0.10
Witnessing a trauma 7.23 0.5 -0.06 0.07 0.10 0.08 0.07 0.07 0.03 0.08
Relativefinancialadversity 10.82 0.6 -0.01 0.08 0.19** 0.06 0.15* 0.06 0.04 0.07
Parent's occupation" 0.00 0.00 -0.00 0.00 -0.00 0.00 0.00 0.00
Parent's education 0.01 0.00 -0.02 0.00 -0.02* 0.01 -0.00 0.00
* Childhood adversity was used to predict the difference between avoidant and anxious ratings. b Regressions were performed on the difference
between attachment ratings for maternal events and attachment ratings for paternal events. c Duncan socioeconomic index.
* p < . 0 5 . **p < .01. ***p < .001.

attached individuals, not to differences between kinds of inse- atively to the secure rating. However, only maternal antisocial
cure attachment. behavior was also related positively to the avoidant and anxious
The results in Table 3 also show that only one of the parental ratings. Two disorders yielded significant associations with the
psychopathologies was related to an individual's secure rating: insecure attachment ratings for both parents: depression and
Both maternal and paternal antisocial behavior were related neg- anxiety. Individuals whose mother or father suffered from de-
1100 MICKELSON, KESSLER, AND SHAVER

pression or anxiety were more likely to endorse both avoidant sion coefficients and corrected standard errors for each of the
and anxious attachment styles than those who did not have a three attachment scales. The parent-child relationship scores
parent with these disorders. Only maternal depression distin- were standardized prior to analysis. Three dimensions of the
guished between avoidant and anxious ratings: Maternal depres- parent-child relationship were assessed: warmth, overprotec-
sion was related more strongly to avoidant than to anxious tiveness, and consistency. Maternal and paternal warmth were
attachment. In addition, having a mother or father with a sub- significant predictors of all three attachment ratings, with
stance abuse disorder was related positively to being avoidant, warmth being related positively to the secure rating and nega-
but not to being anxious. Finally, paternal suicidal behavior was tively to the two insecure ratings. There were no significant
related positively to both insecure ratings, whereas maternal differences between avoidant and anxious attachment for paren-
suicidal behavior was related only to the anxious rating. tal warmth. In addition, there were no significant differences
Four of the other nonsocial traumas were related to the adult between maternal and paternal warmth for any of the three
attachment ratings (see Table 3). Having a life-threatening acci- attachment ratings.
dent was related positively to being anxious. Being in a natural Paternal (but not maternal) overprotectiveness was related
disaster was slightly related positively to being avoidant. Rela- positively to the secure rating. Maternal and paternal overprotec-
tive financial adversity was related positively to both the avoid- tiveness were differentially related to being avoidant. Maternal
ant and anxious ratings. Finally, education of the primary bread- overprotectiveness was related positively to being avoidant,
winner was related negatively to being anxious. In other words, whereas paternal overprotectiveness was related negatively to
as the parent's education increased, endorsement of the anxious being avoidant. Only maternal overprotectiveness was related
attachment style decreased. None of the nonsocial traumas dis- to the anxious rating, such that the more overprotective the
tinguished between avoidant and anxious attachment. individual felt his or her mother to have been, the more strongly
A differential analysis was performed to determine whether he or she endorsed the anxious attachment style. However, only
the "maternal strength" pattern found in prior research was paternal overprotectiveness distinguished between avoidant and
replicated in this study. To test whether maternal events and anxious attachment: Paternal overprotectiveness was related
disorders were more strongly related to attachment than paternal negatively to avoidant but positively to anxious. Maternal-pater-
events and disorders, the difference between the attachment rat- nal differences emerged for overprotectiveness on all three at-
ing scores for mother and father was used as a dependent vari- tachment ratings. For secure attachment, paternal overprotec-
able in regression analyses. These analyses, labeled Mother- tiveness was related more strongly than maternal. For avoidant
father in Table 3, revealed two significant differences: (a) mater- attachment, as stated above, maternal and paternal overprotec-
tiveness differed in the valence of their association, not in the
nal antisocial behavior was related more strongly (and posi-
strength. Finally, for anxious attachment, maternal overprotec-
tively ) to avoidant attachment than paternal antisocial behavior
tiveness was related more strongly than paternal.
and (b) long absence of mother was related more strongly (and
negatively) to secure attachment than long absence of father. Parental consistency was not significantly associated with any
of the adult attachment ratings. In addition, neither maternal
nor paternal consistency distinguished between avoidant and
Parental Bonding
anxious attachment. Only one significant maternal-paternal dif-
As with childhood adversity, linear regressions were per- ference was found for consistency. Although maternal consis-
formed on each parent-child relationship scale to predict the tency was related positively to anxious attachment, paternal con-
three attachment ratings. Table 4 reports the standardized regres- sistency was related negatively.

Table 4
Associations Between Adult Attachment Ratings and Parental Representations
Secure Avoidant Anxious Avoidant—anxious"
Parent-child
representation b SE b SE b SE b SE

Parental warmth
Mother 0.11*** 0.02 -0.15*** 0.02 -0.13*** 0.02 -0.02 0.02
Father 0.12*** 0.02 -0.15*** 0.02 -0.10*** 0.03 -0.05 0.03
Mother-father11 -0.01 0.02 0.00 0.02 -0.03 0.02
Parental overprotectiveness
Mother -0.01 0.02 0.04* 0.02 0.05* 0.02 -0.01 0.02
Father 0.06** 0.02 -0.04* 0.02 0.02 0.02 -0.05* 0.02
Mother-father -0.07*** 0.02 0.08*** 0.02 0.04* 0.02
Parental consistency
Mother 0.03 0.02 -0.00 0.02 0.02 0.02 -0.02 0.02
Father 0.03 0.02 -0.02 0.02 -0.02 0.02 0.00 0.02
Mother-father -0.00 0.02 0.02 0.02 0.04* 0.02
a
Parental representation was used to predict the difference between avoidant and anxious ratings. b Regressions were performed on the difference
between attachment ratings for maternal representations and attachment ratings for paternal representations.
* p < .05. **p < .01. ***p < .001.
 ADULT ATTACHMENT IN THE UNITED STATES 1101

Adult Psychopathology osity was related only to being more secure. Self-esteem, inter-
nal locus of control, luck, extroversion, and Christian fundamen-
In Table 5, the lifetime prevalences of the DSM-II1-R psychi- talism discriminated between the two insecure ratings. Except
atric disorders assessed in the NCS are reported. As with the for extroversion, which was related more strongly to the avoid-
previous tables, the metric regression coefficients and corrected ant attachment style, the other three were related more strongly
standard errors relating these disorders to the adult attachment to the anxious attachment style.
styles are presented for each disorder and each of the three
attachment ratings. Adult psychopathology was used to predict
each of the three attachment ratings. Except for alcohol abuse, Discussion
drug abuse, and schizophrenia, all of the psychiatric disorders
were related negatively to being secure and positively to being Distribution
avoidant and anxious. Schizophrenia was not a significant pre-
dictor of the secure rating, and alcohol and drug abuse were not The distribution of adult attachment styles—59% secure,
significant predictors of the anxious rating. Alcohol abuse and 25% avoidant, and 11% anxious—in our nationally representa-
drug dependence were the only two disorders that discriminated tive sample is similar to the distributions found previously in
between avoidant and anxious attachment. Both were related less representative but large-sample studies of college students
more strongly to being avoidant than to being anxious. and newspaper readers (e.g., Brennan et al., 1991; Hazan &
Shaver, 1987; see Shaver & Hazan, 1993, for a review) as well
as in studies of American infants (Campos, Barrett, Lamb,
Adult Personality Traits
Goldsmith, & Stenberg, 1983; van Uzendoorn, 1994). The only
As shown in Table 6, except for overall religiosity, the person- major discrepancy between our distribution and previously re-
ality characteristics were significant predictors of all three at- ported distributions is for anxious attachment; most prior studies
tachment ratings. Self-esteem, internal locus of control, extro- of college students have found approximately twice the number
version, and openness to experience showed the same pattern of anxious individuals as found in our study (see reviews by
of relations. All four traits were related positively to the secure Shaver & Clark, 1994, and Shaver & Hazan, 1993).
rating and negatively to the two insecure ratings. External locus Furthermore, in our study, anxious attachment was related
of control and neuroticism yielded the opposite pattern; they negatively to age (from 17.4% in the 15-24 age range to 8.0%
were related negatively to the secure rating and positively to in the 45-54 age range), a result that is consistent with the
the two insecure ratings. Christian fundamentalism was related higher proportion of anxious individuals found in studies using
positively to all three attachment ratings. However, overall religi- college students. This result could suggest that anxious adults

Table 5
Associations Between Adult Attachment Ratings and Life Prevalences of DSM-III-R Psychiatric Disorders
Prevalence of Avoidant-
disorder Secure Avoidant Anxious anxious*

Type of disorder % SE b SE b SE b SE b SE

Mood
Major depressive episodes 17.08 0.7 -0.23*** 0.04 0.42*** 0.05 0.34*** 0.05 0.08 0.06
Manic episode 1.67 0.3 -0.42* 0.19 0.97*** 0.14 0.87*** 0.16 0.10 0.20
Dysthymia 6.36 0.4 -0.30*** 0.06 0.48*** 0.08 0.35*** 0.07 0.13 0.09
Anxiety
Panic disorder 3.51 0.3 -0.36* k * 0.10 0.51*** 0.13 0.41*** 0.06 0.10 0.13
Agoraphobia 6.68 0.4 - 0 . 3 1 * 1:* 0.07 0.65*** 0.07 0.61*** 0.08 0.04 0.09
Social phobia 13.36 0.7 -0.27*' K* 0.05 0.52*** 0.04 0.52*** 0.05 -0.00 0.05
Simple phobia 11.29 0.6 -0.16*' 0.05 0.37*** 0.05 0.46*** 0.06 -0.09 0.08
PTSD 7.82 0.5 -0.16*" 0.06 0.42*** 0.08 0.47*** 0.07 -0.05 0.09
Generalized anxiety 5.16 0.3 -0.28* 0.08 0.60*** 0.09 0.44*** 0.07 0.16 0.13
Psychoactive substance use
Alcohol abuse 9.46 0.5 -0.12* 0.05 0.17** 0.06 -0.03 0.05 0.20* 0.08
Alcohol dependence 14.10 0.7 -0.13** 0.05 0.28*** 0.05 0.25*** 0.06 0.03 0.07
Drug abuse 4.45 0.3 -0.16* 0.08 0.14* 0.06 0.03 0.05 0.11 0.08
Drug dependence 7.50 0.4 -0.22** 0.08 0.35*** 0.07 0.20*** 0.06 0.15* 0.07
Other
Schizophrenia 1.26 0.2 -0.26 0.17 0.78*** 0.17 0.48** 0.18 0.30 0.18
Conduct disorder 12.96 0.6 -0.21*** 0.04 0.33*** 0.05 0.32*** 0.05 0.01 0.07
Antisocial disorder 3.15 0.2 -0.35*** 0.08 0.57*** 0.08 0.46*** 0.09 0.11 0.11

Note. DSM-1II-R - Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric Association, 1987); PTSD
posttraumatic stress disorder.
* Adult psychopathology was used to predict the difference between avoidant and anxious ratings.
*p < .05. * * p < . 0 1 . ***/>< .001.
1102 MICKELSON, KESSLER, AND SHAVER

Table 6
Associations Between Adult Attachment Ratings and Adult Personality Characteristics
Secure Avoidant Anxious Avoidant- anxious*

Characteristic b SE b SE b SE b SE

Self-esteem 0.17*** 0.02 -0.32*** 0.02 -0.37*** 0.02 0.05* 0.02

Internal locus of control 0.11*** 0.02 -0.10*** 0.02 -0.18*** 0.02 0.08*** 0.02
External locus of control
Luck -0.04* 0.02 0.13*** 0.02 0.21*** 0.02 -0.08** 0.03
Powerful others -0.07*** 0.02 0.15*** 0.02 0.14*** 0.02 0.02 0.03
Neuroticism -0.16*** 0.02 0.29*** 0.02 0.29*** 0.02 -0.00 0.02
Extroversion 0.31*** 0.02 -0.35*** 0.02 -0.14*** 0.02 -0.21*** 0.03
Openness to experience 0.15*** 0.02 -0.12*** 0.02 -0.10*** 0.02 -0.02 0.02
Christian fundamentalism 0.04* 0.02 0.04* 0.02 0.09*** 0.02 -0.04* 0.02
Overall religiosity 0.08*** 0.02 -0.01 0.02 -0.01 0.01 -0.00 0.02

" Personality characteristics were used to predict the difference between avoidant and anxious ratings.
p<m. "p< .001.

either gradually become secure (by finding or creating a trusting, graphic variables or their correlates on adult attachment (e.g.,
positive marital-romantic relationship) or more self-protec- sex and race, which clearly exist prior to one's attachment
tively avoidant as they age. However, it is also plausible that style). Yet, the direction of effect remains unclear because of
the decline is due to cohort effects. Based on the dramatic the cross-sectional nature of the study. Does attachment style
societal changes occurring since the end of World War II (e.g., affect one's SES, education, and relationship status, or vice
increases in divorce rates, geographic mobility, participation of versa? Or are there reciprocal effects between adult attachment
mothers in the labor force, and rates of psychiatric disorders in and sociodemography?
the general population), attachment styles of people born in Interestingly, although a combination of these security-en-
successive NCS cohorts may be different as a result of their hancing sociodemographic variables was not related to de-
different life experiences. This possibility is consistent with creased avoidant attachment, it was related to decreased anxious
recent findings of an increasing number of anxiety disorders in attachment. Given that anxious attachment is thought to be re-
recent cohorts (Wittchen, Zhao, Kessler, & Eaton, 1994). It is lated most strongly to inconsistent, unpredictable treatment by
not clear, however, why cohort differences would occur mainly attachment figures, and avoidant attachment is thought to be
in the prevalence of anxious attachment and not in the prevalence related most strongly to cool, consistently rejecting, and distant
of avoidant attachment. treatment by attachment figures (Ainsworth et al., 1978; Roth-
Adjudication between these contending possibilities is im- bard & Shaver, 1994), we can speculate that the positive effects
portant, both because it would speak to whether attachment of the security-enhancing sociodemographic variables on attach-
styles are generally stable throughout the adult life course or ment style are due to greater consistency in treatment by others
are subject to change, and because it would shed light on the and not necessarily to lack of rejection. The relative stability of
possibility that secular changes in society have led to an aggre- rates of avoidant attachment across a wide variety of sociodemo-
gate change in the distribution of attachment styles across suc- graphic contexts suggests that avoidance may be rooted more in
cessive cohorts. Given that only the relative prevalence of anx- biological temperament than are secure and anxious attachment
ious and secure styles is affected by age, and given longitudinal (Goldsmith & Alansky, 1987; Seifer, Schiller, Sameroff, Res-
findings reported recently by Klohnen and Bera (1996), we nick, & Riordan, 1996).
tentatively favor the interpretation that anxious attachment be-
comes less common with age. Childhood Adversity
What patterns, if any, can be discerned from this study about
Sociodemographics
the relation of childhood adversities to adult attachment? First,
As mentioned earlier, most previous attachment studies have because not all of the childhood adversities were significantly
not identified significant sociodemographic correlates of adult related to adult attachment, our results suggest that simply expe-
attachment styles. In the present study, which included the full riencing a trauma in childhood does not automatically predestine
range of sociodemographic variability, being white, female, an individual to problems in future relationships. Rather, the
well-educated, middle-class, married, middle-aged; and from data indicate that it is the type of childhood adversity that matters
the Midwest were all associated significantly with an increased most for adult attachment. Second, childhood interpersonal trau-
likelihood of attachment security in adulthood. A few of these mas (e.g., physical abuse, serious neglect) had the most consis-
characteristics, such as being married and being successful at tent associations with insecure adult attachment styles, being
work, may both affect and be affected by attachment security strongly related to increased anxious and avoidant attachment
(Hazan & Shaver, 1990). The most plausible interpretation of as well as to decreased secure attachment. Third, parental psy-
the others is that they reflect an influence of the sociodemo- chopathologies had strong associations with insecure attach-
 ADULT ATTACHMENT IN THE UNITED STATES 1103

ment, but only one significant association with secure attach- ing given the assumption by many researchers that traumatic
ment. One particularly interesting finding is that parental sub- childhood events, especially interpersonal ones, threaten the
stance abuse (in line with findings reported by Brennan et al., bonding process with one's parents, which ultimately leads to
1991) was related to offsprings' avoidant attachment. As men- insecure attachments in adult relationships. Although significant
tioned below, substance abuse was also related to avoidant at- and theoretically consistent patterns were found between adult
tachment for respondents themselves. Although it is easy to attachment and parental warmth, overprotectiveness, and consis-
draw attachment-theoretical connections between certain kinds tency, they were not as strong as those found for actual adverse
of parenting and grown-up children's attachment patterns, it is events. This suggests that childhood adversity may either directly
also worthwhile to consider the possibility that the apparent affect adult attachment or affect it through a process other than
similarities between parental and adult-child pathologies are damaged parental bonding. However, another possibility is that
due in part to inherited temperament (Goldsmith & Alansky, because our measure of parental bonding is a shortened version
1987; Seiferetal., 1996). of the PBI (Parker et al., 1979), it attenuates the true associa-
Only one separation-loss event had a significant association tions found with more complete measures; or, alternatively, ab-
with adult attachment. Unlike previous studies of college stu- stract characterizations may be less valid and more easily biased
dents (e.g., Brennan & Shaver, 1993), which failed to find than recall of specific events.
significant effects of parental divorce on grown-up children's In general, we did not find support for the maternal strength
romantic attachment styles, our study (which included a wider hypothesis (Botens, Shaver, & Levy, 1991; Brennan & Shaver,
range of socioeconomic niches and ages) found that parental 1993). Whereas prior studies of college students have found
divorce was negatively related to attachment security and posi- that maternal characteristics are stronger predictors of attach-
tively related to anxious (but not avoidant) attachment. The ment style than paternal characteristics, in our study maternal
marginal differential association with anxious but not avoidant and paternal events showed similar relations to adult attachment,
attachment suggests that the experience of parental divorce in- suggesting that mothers and fathers have comparable influences
cludes the perception of attachment figures as inconsistently on offsprings' adult attachment styles. In only two instances was
available rather than as cold or rejecting. there evidence for the maternal strength hypothesis: Maternal
Finally, we found that financial adversity during childhood antisocial behavior was related more strongly to avoidant attach-
was related positively to insecure adult attachment. This result ment than paternal antisocial behavior, and long absence of the
is consistent with prior literature that has found lower SES mother was related more strongly to secure attachment than
children experience less continuity in their attachments than long absence of the father. This latter finding is consistent with
children in higher SES homes (see review by Shaver & Hazan, Bowlby's (1973) theory that maternal separation has important
1993). consequences for later social development.
To summarize, these patterns suggest that it is adversities in The instances in which maternal and paternal representations
which an adult directly betrays the trust of a child that are most significantly differed (parental overprotectiveness, parental con-
likely to affect adult attachment. Moreover, personal behavior sistency) were mainly due to a difference in the valence of the
that parents cannot easily control (e.g., depression, substance relation not in the strength of the relation. For example, paternal
abuse, antisocial behavior) tend to be associated with avoidant overprotectiveness was related negatively to avoidant attach-
rather than anxious attachment in adulthood, which suggests ment, whereas maternal overprotectiveness was related posi-
that individuals with troubled parents may learn that it is not tively to avoidant attachment. These findings suggest that when
safe to depend on others. More important, these patterns may differences between mothers and fathers are found for adult
offer a clue about possible mechanisms for the relation between attachment it may be due more to the way in which maternal
childhood adversity and adult psychopathology. For example, versus paternal events or representations affect an adult child's
Kessler and Magee (1994) found that the relation between child- attachment style rather than to the strength of those effects.
hood violence and depression in adulthood was mediated by Alternatively, the lack of support for the maternal strength hy-
adult interpersonal conflicts. In other words, if adult interper- pothesis in our study, and evidence for it in college student
sonal conflicts were absent, there was no association between samples, may suggest that maternal strength is limited to adoles-
childhood violence and adult depression. This finding, along cents. To test this possibility, we conducted post hoc analyses
with those of the present study, suggests that adult attachment examining the youngest age group (15-24), but these analyses
plays a mediating or moderating role in the relation between also failed to support the maternal strength hypothesis. Thus,
childhood adversities and adult psychopathology. If future stud- we can rule out this alternative explanation.
ies find such a relationship to exist between childhood adversi-
ties, adult attachment, and adult psychopathology, interventions
Adult Psychopathology and Personality
could be designed to help improve the relationship skills of
adults traumatized during childhood, which could in turn de- All forms of assessed psychopathology in the respondents
crease the occurrence and severity of adult mental illness. were related significantly to one or more of the adult attachment
ratings, and usually to all three, suggesting that attachment style
Parental Bonding has a pervasive predisposing effect on the development of more
specific psychological disorders, or that having a psychological
We found that abstract characterizations of parents were re- disorder distorts one's internal working models of self and oth-
lated less strongly to respondents' adult attachment ratings than ers. In general, there were few differences between correlates
were specific indications of parental abuse. This result is surpris- of avoidant and anxious attachment. Two important exceptions
1104 MICKELSON, KESSLER, AND SHAVER

are alcohol abuse and drug dependence, both of which were vice versa, as well as to distinguish between age and cohort
much more characteristic of people with avoidant than with effects.
anxious attachments. According to previous studies that have
found similar relations (Brennan & Shaver, 1995; Dozier &
Conclusion
Kobak, 1992; Simpson, Rholes, & Nelligan, 1992), this result
could be due to adults who cannot express their feelings of In summary, the three adult attachment patterns delineated by
distress to others (including attachment figures) being more Hazan and Shaver (1987; Shaver, Hazan, & Bradshaw, 1988),
likely to attempt to alter their moods with the help of drugs of following distinctions made about infants by Ainsworth and her
various kinds. However, as with the sociodemographic variables, colleagues (1978), were assessed, for the first time, in a large
longitudinal studies are needed to determine whether these rela- representative sample of American adults. The distribution pat-
tions between adult psychopathology and adult attachment are tern of the three attachment styles was similar to what has
bidirectional or unidirectional. been found in previous studies. Unlike previous studies, ours
The findings for personality traits show that all of the person- encompassed wide sociodemographic heterogeneity and allowed
ality measures examined here were related to secure and inse- us to detect associations between adult attachment styles and
cure attachment ratings. Secure attachment was related to higher several sociodemographic categories. Many interesting connec-
self-esteem, internal locus of control, extroversion, and open- tions between adult attachment, on the one hand, and childhood
ness to experience—all of which fit well with the designation adversity and parental characteristics, on the other, were ob-
of security. Both avoidant and anxious attachment were associ- tained, but the strongest connections involved interpersonal trau-
ated with external locus of control, neuroticism, introversion, mas such as physical abuse and serious neglect rather than more
and lack of openness to experience. The only significant differ- abstract characteristics of parents, such as judged warmth and
ences between avoidant and anxious attachment are that avoid- consistency. Attachment styles, especially the primary distinc-
ance was differentially associated with extroversion, and tion between insecure and secure attachment, were related to
anxious attachment was differentially associated with low respondents' personality traits and indicators of psychopathol-
self-esteem, internal locus of control, luck, and Christian funda- ogy. On the few occasions in which avoidant and anxious attach-
mentalism—the latter perhaps involving the kind of religious ment had significantly different correlates, the results fit well
emotionality discussed by Kirkpatrick and Shaver (1990) in with previous findings and theory. Taken together, our results
connection with anxious attachment. Interestingly, overall religi- are compatible with the application of attachment theory to adult
osity was associated positively with attachment security but was relationships (Shaver & Clark, 1994) and suggest new lines of
unrelated to the two forms of insecure attachment. inquiry—for example, concerning the cohort effect or apparent
decline in anxious attachment across the adult years. Finally,
future longitudinal studies of both long- and short-term design
Limitations will be worthwhile because attachment patterns appear to be
Two limitations of this study are important to highlight. First, central organizing factors in personality and social development,
we used a brief measure of adult attachment. Longer and more and attachment theory appears to provide valuable leads about
reliable measures of adult attachment have been proposed re- the underlying reasons for associations among a variety of social
cently by, among others, Brennan et al. (in press), Brennan and and personality variables.
Shaver (1995), Collins and Read (1990), Feeney, Noller, and
Callan (1994), and Griffin and Bartholomew (1994). Hazan References
and Shaver's (1987) three-category conception of attachment
styles has been elaborated by Bartholomew (e.g., Bartholo- Ainsworth, M. D. S., Blehar, M., Waters, E., & Wall, S. (1978). Patterns
mew & Horowitz, 1991) into a two-dimensional, four-category of attachment. Hillsdale, NJ: Erlbaum.
conception that is being used by numerous investigators. Bartho- Alexander, P. C. (1993). The differential effects of abuse characteristics
lomew (e.g., Bartholomew & Horowitz, 1991; Scharfe & Bar- and attachment in the prediction of long-term effects of sexual abuse.
tholomew, 1994) has also created an extensive interview proce- Journal of Interpersonal Violence, 8, 346-362.
American Psychiatric Association. (1987). Diagnostic and statistical
dure that appears to produce more reliable assessments of at-
manual of mental disorders (3rd ed., rev.). Washington, DC: Author.
tachment style. Future large-scale studies will benefit from Bartholomew, K., & Horowitz, L. M. (1991). Attachment styles among
improvements in measuring adult attachment style. These im- young adults: A test of a four-category model. Journal of Personality
provements are important because some of the lack of differenti- and Social Psychology, 61, 226-244.
ation between anxious and avoidant attachment ratings in the Beardslee, W. R., Zuckerman, B. S., Amaro, H., & McAllister, M.
present study may be due to the brevity of the rating scales. (1988). Depression among adolescent mothers: A pilot study. Journal
Perhaps especially in noncollege samples or subsamples, multi- of Developmental and Behavioral Pediatrics, 9, 62-65.
item scales may be necessary to spell out the differences be- Botens, D. D., Shaver, P. R., & Levy, K. N. (1991). Styles of romantic
tween the two (or in Bartholomew's theory, three) kinds of attachment and parental representations. Unpublished manuscript,
insecure attachment. State University of New York at Buffalo.
Bowlby, J. (1982). Attachment and loss: Vol. I. Attachment (2nd ed.).
A second limitation is the cross-sectional nature of the sam- New \brk: Basic Books.
pling procedure. Because of this, it is impossible to make cause- Bowlby, J. (1973). Attachment and loss: Vol. II. Separation: Anxiety
effect statements about adult attachment style and its correlates. and anger. New York: Basic Books.
Longitudinal studies would help us to determine whether adult Bradbum, N. M., & Sudman, S. (1979). Improving interview methods
attachment style predicts the onset of psychiatric disorders or and questionnaire design. San Francisco: Jossey-Bass.
 ADULT ATTACHMENT IN THE UNITED STATES 1105

Brennan, K. A., Clark, C. L., & Shaver, P. R. (in press). Self-report theoretical perspective. Journal of Personality and Social Psychology,
measurement of adult attachment: An integrative overview. In J. A. 59, 270-280.
Simpson & W. S. Rholes (Eds.), Attachment theory and close rela- Joyce, P. R-, Sellman, D., Wells, E., & Frampton, C. M. (1994). Parental
tionships. New 'Vbrk: Guilford Press. bonding in men with alcohol disorders: A relationship with conduct
Brennan, K. A., & Shaver, P. R. (1993). Attachment styles and parental disorder. Australian and New Zealand Journal of Psychiatry, 28, 4 0 5 -
divorce. Journal of Divorce and Remarriage, 21, 161 — 175. 411.
Brennan, K. A., & Shaver, P. R. (1995). Dimensions of adult attachment, •Kessler, R. C , Crum, R. M., Warner, L. A., Nelson, C. B., Schulenberg,
affect regulation, and romantic relationship functioning. Personality J., & Anthony, J. C. (1997). The life-time co-occurrence of DSM-
and Social Psychology Bulletin, 21, 267-283. I1I-R alcohol abuse and dependence with other psychiatric disorders
Brennan, K. A., Shaver, P. R., & Tobey, A. E. (1991). Attachment styles, in the National Comorbidity Survey. Archives of General Psychiatry,
gender, and parental problem drinking. Journal of Social and Personal 54, 313-321.
Relationships, 8, 451-466. Kessler, R. C , & Magee, W. J. (1994). Childhood family violence and
Bringle, R. G., & Bagby, G. J. (1992). Self-esteem and perceived quality adult recurrent depression. Journal of Health and Social Behavior,
of romantic and family relationships in young adults. Journal of Re- 35, 13-27.
search in Personality, 26, 340-356. Kessler, R. C , McGonagle, K. A., Zhao, S., Nelson, C. B., Hughes, M.,
Campos, J. J., Barrett, K., Lamb, M. E., Goldsmith, H. H., & Stenberg, Eshleman, S., Wittchen, H.-U., & Kendler, K. S. (1994). Lifetime
C. (1983). Socioemotional development. In M. M. Haith & J. J. and 12-month prevalence of DSM-IH-R psychiatric disorders in the
Campos (Eds.), Handbook of child psychology, Vol. 2. Infancy and United States: Results from the National Comorbidity Survey. Ar-
psychobiology (pp. 783-915). New York: Wiley. chives of General Psychiatry, 51, 8-19.
Carnelley, K. B., Pietromonaco, P. R., & Jaffe, K. (1994). Depression, Kirkpatrick, L. A. (1994). The role of attachment in religious belief
working models of others, and relationship functioning. Journal of and behavior. In K. Bartholomew & D. Perlman (Eds.), Advances in
Personality and Social Psychology, 66, 127-140. personal relationships: Attachment processes in adulthood (Vol. 5,
Cicchetti, D., Toth, S. L., & Lynch, M. (1995). Bowlby's dream comes pp. 239-265). London: Jessica Kingsley.
full circle: The application of attachment theory to risk and psychopa- Kirkpatrick, L. A., & Shaver, P. R. (1990). Attachment theory and reli-
thology. Advances in Clinical Child Psychology, 17, 1-75. gion: Childhood attachments, religious beliefs, and conversion. Jour-
Collins, N. L., & Read, S. J. (1990). Adult attachment, working models, nal for the Scientific Study of Religion, 29, 315-334.
and relationship quality in dating couples. Journal of Personality and Kirkpatrick, L. A., & Shaver, P. R. (1992). An attachment-theoretical
Social Psychology, 58, 644-663. approach to romantic love and religious belief. Personality and Social
Psychology Bulletin, 18, 266-275.
Dozier, M., & Kobak, R. R. (1992). Psychophysiology and adolescent
attachment interviews: Converging evidence for repressing strategies. Klohnen, E. C , & Bera, S. (1996). Behavioral and experiential patterns
Child Development, 63, 1473-1480. of avoidantly and securely attached women across adulthood: A 30-
year longitudinal perspective. Manuscript submitted for publication.
Dutton, D. G., Saunders, K., Starzomski, A., & Bartholomew, K. (1994).
Intimacy-anger and insecure attachment as precursors of abuse in Levenson, H. (1973). Multidimensional locus of control in psychiatric
patients. Journal of Consulting and Clinical Psychology, 41, 397-
intimate relationships. Journal of Applied Social Psychology, 24,
404.
1367-1386.
Levy, K. N., Blatt, S. J., & Shaver, P R . (1996). Styles of romantic
Endicott, J., Andreason, N., & Spitzer, R. L. (1978). Family history
attachment and parental representations. Manuscript submitted for
research diagnostic criteria. New \brk: Biometrics Research, New
publication.
York State Psychiatric Institute.
Lutz, W. J., & Hock, E. (1995). Maternal separation anxiety: Relations
Feeney, J. A., & Noller, P. (1990). Attachment style as a predictor of
to adult attachment representations in mothers of infants. Journal of
adult romantic relationships. Journal of Personality and Social Psy-
Genetic Psychology, 156, 57-72.
chology, 58, 281-291.
Main, M., Kaplan, N., & Cassidy, J. (1985). Security in infancy, child-
Feeney, J. A., Noller, P., & Callan, V. J. (1994). Attachment style, com-
hood, and adulthood: A move to the level of representation. In I.
munication, and satisfaction in the early years of marriage. In K.
Bretherton & E. Waters (Eds.), Growing points of attachment theory
Bartholomew & D. Perlman (Eds.), Advances in personal relation-
and research. Monographs of the Society for Research in Child Devel-
ships: Attachment processes in adulthood (Vol. 5, pp. 269-308).
opment, 5 0 ( 1 - 2 , Serial No. 209), 66-104.
London: Jessica Kingsley.
Mallinckrodt, B., McCreary, B. A., & Robertson, A. K. (1995). Co-
Goldberg, L. R. (1992). The development of markers for the Big-Five occurrence of eating disorders and incest: The role of attachment,
factor structure. Psychological Assessment, 4, 26—42. family environment, and social competencies. Journal of Counseling
Goldsmith, H. H., & Alansky, J. A. (1987). Maternal and infant temper- Psychology, 42, 178-186.
amental predictors of attachment: A meta-analytic review. Journal of Miller, P. C , Lefcourt, H. M., Holmes, J. G., Ware, E. E., & Saley, W. E.
Consulting and Clinical Psychology, 55, 805-816. (1986). Marital locus of control and marital problem solving. Journal
Griffin, D. W., & Bartholomew, K. (1994). The metaphysics of measure- of Personality and Social Psychology, 51, 161 — 169.
ment: The case of adult attachment. In K. Bartholomew & D. Perlman Parker, G., Tupling, H., & Brown, L. B. (1979). A parental bonding
(Eds.), Advances in personal relationships: Attachment processes in instrument. British Journal of Medical Psychology, 52, 1-10.
adulthood (Vol. 5, pp. 17-52). London: Jessica Kingsley. Patrick, M., Hobson, R. P., Castle, D., & Howard, R. (1994). Personality
Guenzel, P. J., Berckmans, T. R., & Cannell, C. F. (Eds.). (1983). Gen- disorder and the mental representation of early social experience.
eral interviewing techniques: A self-instructional workbook for tele- Development and Psychopathology, 6, 375—388.
phone and personal interviewer training. Ann Arbor: Institute for Pettem, Q, West, M., Mahoney, A., & Keller, A. (1993). Depression
Social Research, Survey Research Center, University of Michigan. and attachment problems. Journal of Psychiatry and Neuroscience,
Hazan, C , & Shaver, P. R. (1987). Romantic love conceptualized as an 18, 78-81.
attachment process. Journal of Personality and Social Psychology, Resnick, H. S., Kilpatrick, D. G., & Lipovsky, J. A. (1991). Assessment
52, 511-524. of rape-related posttraumatic stress disorder: Stressor and symptom
Hazan, C , & Shaver, P. R. (1990). Love and work: An attachment- dimensions. Psychological Assessment, 3, 561-572.
1106 MICKELSON, KESSLER, AND SHAVER

Robins, L. N. (1985). Early home environment and retrospective recall: Shaver, P. R., & Rubenstein, C. (1980). Childhood attachment experi-
A test for concordance between siblings with and without psychiatric ence and adult loneliness. In L. Wheeler (Ed.), Review of personality
disorders. American Journal of Orthopsychiatry, 55, 27-41. and social psychology (Vol. 1, pp. 42-73). Beverly Hills, CA: Sage.
Robins, L. N., Helzer, J. E., Croughan, J. L., & Ratcliff, K. S. (1981). Simpson, J. A. (1990). The influence of attachment styles on romantic
National Institute of Mental Health diagnostic interview schedule: Its relationships. Journal of Personality and Social Psychology, 59, 971 —
history, characteristics and validity. Archives of General Psychiatry, 980.
38, 381-389. Simpson, J. A., Rholes, W. S., & Nelligan, J. S. (1992). Support-seeking
Robins, L. N., Wing, J., Wittchen, H.-U., & Helzer, J. E. (1988). The and support-giving within couple members in an anxiety-provoking
Composite International Diagnostic Interview: An epidemiologic in- situation: The role of attachment styles. Journal of Personality and
strument suitable for use in conjunction with different diagnostic sys- Social Psychology, 62, 434-446.
tems and in different cultures. Archives of General Psychiatry, 45, Spengler, P., & Wittchen, H.-U. (1989). Procedural validity of standard-
1069-1077. ized symptom questions for the assessment of psychotic symptoms:
Rosenberg, M. (Ed.). (1965). Society and the adolescent self-image. A comparison of the CIDI with two clinical methods. Comprehensive
Princeton, NJ: Princeton University Press. Psychiatry, 29, 309-322.
Rothbard, J. C , & Shaver, P. R. (1991). Attachment styles and the qual- Straus, M. A. (1979). Measuring intrafamily conflict and violence: The
ity and importance of attachment to parents. Unpublished manuscript. Conflict Tactics (CT) Scales. Journal of Marriage and the Family,
Psychology Department, State University of New "fork at Buffalo. 41, 75-88.
Rothbard, J. C , & Shaver, P. R. (1994). Continuity of attachment across Tavris, C , & Wade, C. (1984). The longest war: Sex differences in
the life span. In M. B. Sperling & W. H. Berman (Eds.), Attachment perspective. New Ifork: Harcourt Brace Jovanovich.
in adults: Clinical and developmental perspectives (pp. 31-71). New U.S. Bureau of the Census. (1991). Statistical abstract of the United
York: Guilford Press. States 1991: The national data book. Washington, DC: U.S. Govern-
Rust, K. (1985). Variance estimation for complex estimators in sample ment Printing Office.
surveys. Journal of Official Statistics, 1, 381-397. U.S. Department of Health and Human Services. (1992). National
Scharfe, E., & Bartholomew, K. (1994). Reliability and stability of adult Health Interview Survey, 1989 [Computer file]. Hyattsville, MD: Na-
attachment patterns. Personal Relationships, 1, 23-44. tional Center for Health Statistics.
Seifer, R., Schiller, M , Sameroff, A. J., Resnick, S., & Riordan, K. van IJzendoorn, M. H. (1994). Adult attachment representations, paren-
(1996). Attachment, maternal sensitivity, and infant temperament dur- tal responsiveness, and infant attachment: A meta-analysis on the pre-
ing the first year of life. Developmental Psychology, 32, 12-25. dictive validity of the Adult Attachment Interview. Psychological Bul-
Seligman, M. E. P., & Schulman, P. (1986). Explanatory style as a letin, 117, 387-403.
predictor of productivity and quitting among life insurance sales Wacker, H. R., Battegay, R., Mullejans, R., & Schlosser, C. (1990).
agents. Journal of Personality and Social Psychology, 50, 832-838. Using the CIDI in the general population. In C. N. Stefanis, A. D.
Senchak, M , & Leonard, K. E. (1992). Attachment styles and marital Rabavilas, & C. R. Soldatos (Eds.), Psychiatry: A world perspective
adjustment among newlywed couples. Journal of Social and Personal (pp. 138-143). New York: Elsevier.
Relationships, 9, 51-64. Waters, E., Merrick, S. K., Albersheim, L. J., & Treboux, D. (1995,
Shaver, P. R., & Brennan, K. A. (1992). Attachment styles and the "Big April). Attachment security from infancy to early adulthood: A
Five" personality traits: Their connections with each other and with twenty-year longitudinal study. Poster presented at the biennial meet-
romantic relationship outcomes. Personality and Social Psychology ing of the Society for Research in Child Development, Indianapolis,
Bulletin, 18, 536-545. IN.
Shaver, P R . , & Clark, C. L. (1994). The psychodynamics of adult Wittchen, H.-U., Robins, L. N., Cottier, L. B., Sartorious, N., Burke,
romantic attachment. In J. M. Masling & R. F. Bornstein (Eds.), Em- J. D., Regier, D. A., & Participants in the multicentre WHO/AD-
pirical perspectives on object relations theories (pp. 105-156). AMHA field trials. (1991). Cross-cultural feasibility, reliability and
Washington, DC: American Psychological Association. sources of variance in the Composite International Diagnostic Inter-
Shaver, P. R., Collins, N. L., & Clark, C. L. (1996). Attachment styles view (CIDI). British Journal of Psychiatry, 159, 645-653.
and internal working models of self and relationship partners. In Wittchen, H.-U., Zhao, S., Kessler, R. C , & Eaton, W. W. (1994).
G. J. O. Fletcher & J. Fitness (Eds.), Knowledge structures in close DSM-III-R generalized anxiety disorder in the National Comorbidity
relationships: A social psychological approach (pp. 25—61). Hills- Survey. Archives of General Psychiatry, 51, 355-364.
dale, NJ: Erlbaum. World Health Organization. (1990a). Composite International Diagnos-
Shaver, P. R., & Hazan, C. (1993). Adult romantic attachment: Theory tic Interview (CIDI, Version 1.0.). Geneva, Switzerland: Author.
and evidence. In D. Perlman & W. Jones (Eds.), Advances in personal World Health Organization. (1990b). Composite International Diagnos-
relationships (Vol. 4, pp. 29-70). London: Jessica Kingsley. tic Interview Computer Programs (Version 1.1.). Geneva, Switzerland:
Shaver, P. R., Hazan, C , & Bradshaw, D. (1988). Love as attachment: Author.
The integration of three behavioral systems. In R. J. Sternberg & M. L. Zuroff, D. C , & Fitzpatrick, D. K. (1995). Depressive personality
Barnes (Eds.), The psychology of love (pp. 68-99). New Haven, CT styles: Implications for adult attachment. Personality and Individual
Yale University Press. Differences, 18, 253-365.
Shaver, P. R., Papalia, D., Clark, C. L., Koski, L. R., Tidwell, M. C , &
Nalbone, D. (1996). Androgyny and attachment security: Two related Received March 13, 1996
models of optimal personality. Personality and Social Psychology Revision received November 22, 1996
Bulletin, 22, 582-597. Accepted November 23, 1996 •