Personal Relationships, 14 (2007), 129–147. Printed in the United States of America.

Copyright Ó 2007 IARR. 1350-4126=07

Adoption, attachment, and relationship concerns:

A study of adult adoptees

JUDITH A. FEENEY, a NOLA L. PASSMORE, b

a
AND CANDIDA C. PETERSON
a
University of Queensland, Australia and bUniversity of Southern Queensland,
Australia

Abstract
We examined the impact of adoptive status and family experiences on adult attachment security and how attachment
predicts relationship outcomes. Adults adopted as infants (N ¼ 144) and a sample of nonadoptees (N ¼ 131) completed
measures of attachment security at recruitment and 6 months later; other measures assessed parental bonding and
adoptees’ reunion experiences (Time 1), and relationship variables (e.g., loneliness, relationship quality; Time 2).
Insecurity was higher for adoptees and those reporting negative childhood relationships with parents. For adoptees only,
recent relationship difficulties also predicted insecurity. Attachment dimensions were more important than adoptive
status in predicting relationship variables and mediated the effects of adoptive status. The results support the utility of
attachment theory in understanding adoptees’ relationship concerns.

Consistent with the suggestion that adoption is overrepresentation in mental health settings
a risk factor for general adjustment problems and small increases in rates of behavior prob-
(Verrier, 1993), some studies have shown that lems (Juffer & van IJzendoorn, 2005; Nickman
adoptees are overrepresented in clinical groups et al., 2005; van IJzendoorn, Juffer, & Klein
(Wierzbicki, 1993), or fare worse than non- Poelhuis, 2005). Hence, the claim that adop-
adoptees on variables such as self-esteem, de- tion is a risk factor for general adjustment dif-
pression, and anxiety (e.g., Borders, Penny, & ficulties remains controversial. Further, the
Portnoy, 2000; Cubito & Obremski-Brandon, reason for the mixed findings is unclear: Meth-
2000; Tieman, van der Ende, & Verhulst, odological (sampling and measurement) dif-
2005). Other studies comparing adoptees and ferences between studies may be one factor,
nonadoptees, however, have found no differ- but another possibility is that the link between
ences in psychological distress (Collishaw, adoption and adjustment depends on a range
Maughan, & Pickles, 1998) or life satisfaction of factors, biological (e.g., prenatal drug ex-
(Borders et al., 2000). Indeed, recent reviews posure), social (e.g., functioning within the
and meta-analyses suggest that most adoptees adoptive family), and social structural (e.g.,
are well adjusted, although there is evidence of financial well-being of the family).
Despite the mixed findings for general
Judith A. Feeney, School of Psychology, University of adjustment, there are convincing arguments
Queensland, Australia 4072; Nola L. Passmore, Depart- linking adoption to increased risk of interper-
ment of Psychology, University of Southern Queensland, sonal problems. Issues concerning loss are
Toowoomba, Australia 4350; Candida C Peterson, School
of Psychology, University of Queensland, Australia 4072. inherently relational and are central to the
The research was supported by an Australian Research adoption experience: Adoptees have lost their
Council Discovery Grant (DP0344557). birth parents, and more generally, a sense of
Correspondence should be addressed to Judith A.
Feeney, University of Queensland, School of Psychology, being biologically tied to significant others
Australia 4072, e-mail: [Link]@[Link]. (Brodzinsky, 1990; Jones, 1997; Schechter &

129
130 J. A. Feeney, N. L. Passmore, and C. C. Peterson

Bertocci, 1990); further, there is a ‘‘status loss’’ life are likely to influence adults’ relational
associated with being different (Brodzinsky). adjustment.
Moreover, in cases where attempted reunions Recent literature on the lifelong importance
with birth relatives are difficult or unsuccess- of attachment supports this claim. For exam-
ful, adoptees may experience further loss and ple, a wealth of data indicates that needs for
rejection. Finally, in comparison to parental security and a sense of belonging apply across
loss through death or divorce, adoption-related the life span (Baumeister & Leary, 1995;
losses have unique features that may predis- Weiss, 1991). A growing body of research also
pose the individual to relational problems. supports the tenets of adult attachment theory,
Specifically, the losses are often unacknowl- as first expounded by Hazan and Shaver
edged or downplayed (a situation associa- (1987). That is, adults’ close personal relation-
ted with little support provision) and may ships share important emotional and behav-
entail a sense of abandonment and rejection ioral similarities with the bonds that form
(Brodzinsky; Jones; Nickman, 1985). between infants and caregivers. Further, the
concept of attachment style applies to both
types of relationships; that is, early social
Adoption and attachment
experiences (including loss and rejection) cre-
Despite the considerable literature dealing ate individual differences in security, which
with adoption, loss, and abandonment, few shape relational attitudes and behaviors. In
studies have assessed the impact of adoption the context of adult attachment, these individ-
on adults’ peer relationships. Further, research ual differences have been variously concep-
in this area has lacked a strong theoretical tualized in terms of categories (styles) or
framework. Some years ago, a theoretical paper continuous dimensions, such as avoidance
by Edens and Cavell (1999) proposed the util- and relationship anxiety (Shaver & Mikulincer,
ity of attachment principles in the study of 2004). Regardless of their conceptual basis,
adoption. (Attachment refers to the affectional however, studies have shown systematic links
bond that characterizes many close relation- between adult attachment measures and key
ships and that promotes a sense of comfort relational processes and outcomes. For exam-
and security.) The authors noted that infants ple, secure attachment is associated with more
who are adopted may be at greater risk of open expression of thoughts and feelings, and
prenatal or birth complications and parental with higher relationship quality (see Feeney,
psychopathology; in turn, such factors may 1999, for a review).
increase the likelihood of particular attach- To date, however, Borders et al. (2000) are
ment behaviors (e.g., dislike of being held) the only researchers to have systematically
that make the role of adoptive parents more explored the link between adoption and adult
difficult. Adoptive parents may also experi- attachment security. These researchers studied
ence lack of role autonomy and uncertainty a sample of adoptees and their nonadopted
about role obligations, which may interfere with friends and found differences between these
the bonding process. Further, disclosure of two groups in attachment and perceptions of
adoptive status may (depending on its timing social support. Specifically, adoptees were
and manner) lead adoptees to feel that they do underrepresented in the secure attachment
not ‘‘belong’’ in the adoptive family (Edens & group and overrepresented in the preoccupied
Cavell). Attachment theory suggests possible and fearful groups, and reported less support
long-term implications of these difficulties: (from both family and friends) than their non-
Experiences with caregivers are gradually adopted counterparts. This study provided an
internalized in the form of working models, important first step in linking adoption and adult
or generalized expectancies about the self in attachment but was limited by its cross-sectional
relation to close others. As working models are nature, its reliance on a categorical measure of
relatively stable and play an active role in attachment, and its failure to fully consider the
guiding later interactions (Shaver, Collins, & role of early parenting and ongoing relationship
Clark, 1996), negative experiences early in experiences in relation to attachment.
Adoption, attachment, and relationship concerns 131

The present study belonging’’ (Brodzinsky, 1990; Jones, 1997).

For this reason, adoption may be linked with
No study to date has thoroughly explored the
negative working models of attachment; that
impact of adoption on attachment security and
is, with perceptions that the self is unworthy of
relationship outcomes in adulthood or the pos-
love and attention or that other people are un-
sible moderating role of family experiences.
available, uncaring, and rejecting (Bartholomew
The present study addressed these issues by
& Horowitz, 1991).
recruiting a sample of adults who had been
At the same time, it is important to recog-
adopted as infants and a comparison sample
nize that the association between adoptive sta-
of adults from intact biological families, and
tus and insecurity in adulthood may not be
following them over a 6-month period.
strong and that family experiences are also
The relevance of infant adoption stems
likely to have an impact on working models.
from the fact that from the early 20th century
Previous studies have reported greater vari-
through to the 1970s, the dominant type of
ability on measures of psychosocial well-being
adoption in many Western countries (e.g.,
within adopted than nonadopted samples (e.g.,
England, the United States, Australia, New
Borders et al., 2000) and have provided evi-
Zealand) involved infants who were born out
dence that open and affectionate relationships
of wedlock and whose adoption was based on
within the adoptive family may act as a buffer
secrecy and anonymity (Brodzinsky, 2005;
against adjustment difficulties (e.g., Levy-
Hoksbergen & ter Laak, 2005; Ryburn, 1994).
Shiff, 2001; Passmore, Fogarty, Bourke, &
The losses associated with such early adoption
Baker-Evans, 2005). Further, given that search
placement are subtle and covert, emerging
and reunion experiences are often emotionally
gradually over time with the individual’s
intense and are linked to perceptions of accep-
growing awareness of the meaning of having
tance and rejection (Verrier, 1993), these
been adopted (Brodzinsky, 1990). Specifi-
experiences may also shape working models
cally, as children come to realize the impli-
of self and others. Hence, as part of our first
cations of having been adopted, they often
broad aim, we were interested in assessing the
experience a loss of self (or identity) and a loss
importance of both adoptive status and family
of stability in their relationships with adoptive
experiences (with adoptive parents and birth
parents (Brodzinsky, 1990; Nickman et al.,
mothers) as predictors of adult attachment
2005). These various losses (together with
security.
the loss of biological ties) may leave adoptees
Another issue relevant to the effects of
feeling ‘‘incomplete, alienated, disconnected,
adoption on attachment security concerns pat-
abandoned, or unwanted’’ (Brodzinsky, 1990,
terns of stability and change. This issue has
p. 7). These arguments further support the rel-
theoretical and applied significance as it
evance of attachment theory, given its focus on
addresses questions about the malleability of
working models of the self in relation to close
working models. Previous studies point to
others. As detailed next, the present study had
moderate rates of attachment stability in adult
two broad aims.
samples over intervals from 1 week to 25
years, although stability is higher when more
Adoption as a risk factor for insecure attach- reliable measures (e.g., multiple-item scales)
ment. The first aim was to assess the extent are used (Feeney & Noller, 1996). Given that
to which adoption represents a risk factor for attachment measures are not perfectly stable
insecure attachment in adulthood. As noted (even when their limited reliability is taken
earlier, there are sound reasons for suggesting into account), researchers have considered
that the losses associated with the adoption the causes of change over time. For example,
experience may predispose individuals to rela- change in attachment style may be most likely
tional difficulties, including a sense of insecu- in response to major relationship changes,
rity. Specifically, being adopted may be such as the formation and dissolution of couple
associated with a sense of having been rejected bonds (Hammond & Fletcher, 1991; Kirkpatrick
or abandoned by birth parents, and of ‘‘not & Hazan, 1994). Attachment measures may
132 J. A. Feeney, N. L. Passmore, and C. C. Peterson

also be prone to more short-term instability: level of risk (e.g., Rholes, Simpson, & Grich
Individuals may have multiple attachment Stevens, 1998; Shaver & Hazan, 1993; Simpson,
orientations, based on their varied relationship 1990). Further, both experimental and field
experiences, and situational factors operating studies show that working models of attach-
at any given time may affect responses (Baldwin, ment shape relational expectations, behaviors,
Keelan, Fehr, Enns, & Koh-Rangarajoo, 1996). and memories (Pietromonaco, Greenwood, &
Finally, vulnerability factors (e.g., family dis- Feldman Barrett, 2004; Shaver & Mikulincer,
ruptions and psychopathology) may make 2002).
some people more susceptible to change, by Much of the research in this area has been
creating more tentative views of self and cross-sectional, but longitudinal designs are
others (Davila, Burge, & Hammen, 1997). particularly useful when examining relational
What does this mean for the adoption expe- adjustment. For example, Shaver and Brennan
rience? Attachment theorists propose that the (1992) were able to demonstrate that attach-
expectations embodied in working models ment measures showed only modest relations
solidify over time and tend to be self-fulfilling with personality variables and that attachment
(Shaver et al., 1996). They also, however, measures were more predictive of relationship
accept that attachment style may be affected outcomes (8 months later) than were personal-
by powerful relationship experiences. As ity measures. Such findings show that associ-
noted earlier, adoptees face unique relational ations between attachment and relationship
issues pertaining to the loss of biological ties. functioning do not simply reflect common
As they become aware of the implications of method variance or global response sets.
having been adopted, they may also experi- In this study, we tested the ability of attach-
ence a loss of identity, a destabilizing of rela- ment measures to predict relationship varia-
tionships with adoptive parents, emotional bles assessed 6 months later, after controlling
challenges linked to the search process, and for the effect of adoptive status. For complete-
difficulty in juggling relationships with adop- ness, we also compared patterns of prediction
tive and birth families. Hence, adoptees’ work- for adoptees and nonadoptees, although the
ing models may be more malleable than those effects of attachment security are usually
of other adults. In particular, their working mod- robust across diverse variables such as race
els may be reactive to salient events in romantic (Cooper, Shaver, & Collins, 1998) and sexual
relationships, such as relationship deterioration: orientation (Feeney & Noller, 2004). Given
Romantic relationships are the prototypical the expected link between adoptive status
attachment bond for adults (Ainsworth, 1989; and attachment insecurity, together with the
Weiss, 1991) and play a unique role in meet- active role of working models in shaping
ing attachment needs (Doherty & Feeney, relational attitudes and behaviors, we also
2004). tested whether the association between adop-
tive status and relationship functioning was
The role of attachment in predicting relational mediated by attachment. This important ques-
adjustment. Our second broad aim was to tion addresses underlying processes involved
assess the role of adult attachment in predict- in the link between adoption and adjustment;
ing relationship attitudes and behaviors. As that is, attachment may be one mechanism
noted earlier, considerable evidence points to through which adoptive status influences later
the link between security and positive rela- relationship experiences.
tional processes and outcomes. For example,
secure attachment has been linked to high
Hypotheses
levels of interdependence, commitment, and
satisfaction, and to low levels of loneliness; In terms of predicting attachment, we expected
conversely, relationship anxiety has been linked that adults who were adopted as infants would
to conflict and coercion, and avoidance has report more insecurity than those who grew up
been linked to emotional distancing and to the with both biological parents (Hypothesis 1).
perception that intimate bonds involve a high We also, however, expected higher insecurity
Adoption, attachment, and relationship concerns 133

among those perceiving more negative child- groups were almost exclusively Caucasian.
hood relationships with parents (biological or For both groups, the largest recruitment source
adoptive) (Hypothesis 2) and, in the case of (n ¼ 160 of the total 275) involved advertise-
adoptees, less satisfying reunions with birth ments in local media. Smaller numbers were
mothers (Hypothesis 3). (In testing Hypothesis recruited via social networks available to the
2, we also checked for interactive effects of researchers (n ¼ 62), undergraduate psychol-
adoptive status and parenting on attachment ogy pools at the University of Queensland and
security, given previous suggestions that a the University of Southern Queensland (n ¼
well-functioning adoptive family may act as 30), and brochures placed on university cam-
a buffer against adjustment difficulties.) Fur- puses and in community centers (n ¼ 10). In
ther, we expected that in comparison to non- addition, a small number of adoptees heard
adoptees, adoptees’ responses to attachment about the study through adoption-related sup-
measures would be more reactive to recent rela- port groups (n ¼ 6) or through Internet sites
tionship events (Hypothesis 4). dealing with these issues (n ¼ 7). Those inter-
With regard to relationship outcomes, we ested in participating were informed of the
expected that attachment insecurity (assessed purpose and confidential nature of the study,
at Time 1) would predict more negative rela- and were mailed a questionnaire package with
tional attitudes and behaviors at follow-up, after a prepaid envelope for returning the materials.
controlling for adoptive status. That is, we Each questionnaire was assigned a unique
expected insecurity to predict higher scores on identifying code, so that it could be matched
loneliness and perceived risk in intimacy, and to the follow-up questionnaire that was mailed
lower scores on relationship quality (Hypothesis out 6 months later. For both questionnaires, the
5); these effects were expected to apply to both various measures were presented in one of six
adoptees and nonadoptees, although for com- random orders. The comparability of the two
pleteness, we checked for interactive effects of groups and the extent of attrition are consid-
attachment and adoptive status. Finally, we ered in the Procedure section.
expected attachment variables to mediate the
link between adoptive status and relationship
Measures: Time 1
functioning (Hypothesis 6).
The first questionnaire assessed background
variables, attachment security, and parental
Method
bonding (detailed below). In addition, adopt-
ees answered open-ended and structured ques-
Participants
tions about their adoption (e.g., age at adoption,
Two groups of participants were recruited. age at learning adoptive status), and their search
Both groups were restricted to individuals and reunion experiences.
who were at least 18 years of age, had been
born in Australia, and had lived in a two-parent Background variables. Participants were
family (either biological or adopted) for the asked to indicate their age, gender, relation-
first 16 years of life. At the beginning of the ship status (single, cohabiting, married, and
study, the adopted group consisted of adults divorced or separated), parental status (chil-
who had been adopted as infants, by nonkin dren, no children), education level (did not
(N ¼ 144; 111 women and 33 men), and the complete school, completed school only,
comparison group consisted of adults who technical college certificate or diploma, some
grew up with their biological parents (N ¼ university education), and employment status
131; 96 women and 35 men). Age at adoption (full time, part time, not currently employed).
ranged from 1 week to 78 weeks (M ¼ 5.98
weeks, SD ¼ 10.91). Attachment security. To provide a com-
Because there are unique issues in Australia prehensive assessment of current attachment
associated with indigenous adoptions, these security, two measures were used. First, partici-
were not included in the study, and both pants completed the Relationship Questionnaire
134 J. A. Feeney, N. L. Passmore, and C. C. Peterson

(Bartholomew & Horowitz, 1991). This mea- Reunion experiences. Adoptees answered
sure consists of single-paragraph descriptions a series of questions about their reunion expe-
of each of four attachment styles: fearful, pre- riences. This report considers only three ques-
occupied, secure, and dismissing. Participants tions assessing relationships with birth mothers.
are asked to read the descriptions and choose These questions tapped satisfaction with the
the one that best describes the way they gen- initial reunion and satisfaction with the current
erally are in close relationships. relationship (1 ¼ extremely dissatisfying to 6 ¼
Second, participants completed the At- extremely satisfying) and emotional closeness
tachment Style Questionnaire (ASQ; Feeney, of the current relationship (1 ¼ extremely dis-
Noller, & Hanrahan, 1994). Although the ASQ tant to 6 ¼ extremely close); the three ratings
can yield scores on five attachment scales, the were analyzed separately.
two major factors are avoidance and relation-
ship anxiety (Alexander, Feeney, Hohaus, &
Measures: Time 2
Noller, 2001). Avoidance (16 items) assesses
the tendency to be uncomfortable with inti- At the 6-month follow-up, participants com-
macy and to have difficulty in depending on pleted the same two measures of attachment
relationship partners. A sample item is ‘‘I security included in the first questionnaire. In
find it difficult to depend on others.’’ Rela- addition, they completed the following mea-
tionship anxiety (13 items) assesses fears of sures of relationship attitudes and relationship
being rejected or abandoned, and concerns functioning.
about the extent of partners’ love and commit-
ment (e.g., ‘‘I worry that others won’t care Risk in intimacy. The 10-item Risk in Inti-
about me as much as I care about them.’’). macy scale (Pilkington & Richardson, 1988)
The response format is from 1 (totally dis- assesses perceptions of the risks and dangers
agree) to 6 (totally agree). Alpha coefficients associated with being close to others (e.g.,
for these scales at Times 1 and 2 ranged from ‘‘I’m afraid to get really close to someone
.87 to .90. because I might get hurt.’’) Items use a 6-point
response scale, from 1 (strongly disagree) to 6
Parental bonding. The Parental Bonding (strongly agree), and formed a reliable scale
Instrument (PBI; Parker, Tupling, & Brown, (a ¼ .91).
1979) contains 25 items describing various
parental attitudes and behaviors. This mea- Loneliness. The 37-item Social and Emo-
sure requires participants to think back over tional Loneliness Scale for Adults (DiTommaso
the first 16 years of life and to rate each item & Spinner, 1993) contains three subscales:
(separately for mother and father) from romantic (12 items; e.g., ‘‘I find myself wishing
0 (very unlike this parent) to 3 (very like this for someone with whom to share my life.’’),
parent). Adoptees answered these questions family (11 items; e.g., ‘‘I feel alone when I’m
with respect to their adoptive parents, with my family.’’), and social loneliness (14
whereas those in the comparison group items; e.g., ‘‘What’s important to me doesn’t
answered with respect to their biological seem important to the people I know.’’). The
parents. The PBI yields scores on the dimen- response format ranges from 1 (strongly dis-
sions of care (12 items, e.g., ‘‘spoke to me agree) to 7 (strongly agree). Alpha coefficients
with a warm and friendly voice’’) and over- exceeded .90 for each scale.
protection (13 items, e.g., ‘‘tried to control
everything I did’’). A large body of work by Relationship quality. Participants who were
Parker and his colleagues (e.g., Parker, 1983) currently in a romantic relationship (n ¼ 199)
supports these constructs as the two basic reported the length of that relationship and
dimensions of parenting that impact on the completed three measures assessing relation-
bonding process. Both scales were highly ship quality. Relationship satisfaction and
reliable, with alpha coefficients exceeding commitment were assessed using subscales
.90 for each parent. (five- and seven-item, respectively) of the
Adoption, attachment, and relationship concerns 135

Investment Model Scale (Rusbult, Martz, & whether that relationship existed at the start of
Agnew, 1998). Sample items are ‘‘My rela- the study). Four dichotomous variables were
tionship is close to ideal’’ (satisfaction) and thus created (no ¼ 0, yes ¼ 1), one for each
‘‘I am committed to maintaining my relation- type of relationship change. For each item,
ship with my partner’’ (commitment); response space was also provided for respondents to give
options range from 0 (do not agree at all) to 8 a brief description of the relationship change.
(agree completely). Both subscales were highly
reliable (alpha coefficients exceeded .95).
Procedure: Assessing group comparability
Dyadic trust was assessed with a short (13-item)
and attrition
version of the Trust in Close Relationships Scale
(Boon & Holmes, 1992). A sample item is ‘‘My Comparability of the groups. Before pre-
partner is truly sincere in his/her promises.’’ senting the main data analyses, it is important
Response options range from 1 (strongly dis- to consider the comparability of the two
agree) to 7 (strongly agree), and reliability groups and the extent of attrition. To assess
was very high (a ¼ .96). Preliminary analyses whether the two groups of participants were
showed that the three measures of relationship similar at the beginning of the study, we com-
quality were strongly interrelated (rs ranged pared them on the background variables of
from .70 to .81). Hence, we formed a compo- age, gender, relationship status, parental sta-
site measure of relationship quality by summing tus, education level, and employment status
standardized scores on the three scales (a ¼ (see Table 1). Analysis of variance indicated
.97). that the groups were similar in terms of age,
F(1, 272) , 1. In addition, frequency compar-
Relationship changes. Finally, we asked isons revealed no significant differences on the
respondents about any changes in romantic remaining variables: gender, v2(1) , 1; rela-
relationships over the 6-month period between tionship status, v2(3) ¼ 3.68; parental status,
assessment sessions. That is, participants indi- v2(1) ¼ 1.94; education level, v2(3) ¼ 1.35;
cated whether they had experienced one or and employment status, v2(2) ¼ 3.84 (all ps
more of the following changes: ending of ..10). Overall, the subjects were relatively
a relationship, formation of a new relationship, well educated and predominantly female but
and deterioration or improvement in the qual- represented a wide range in terms of age and
ity of an existing relationship (regardless of occupational status.

Table 1. Participant characteristics

Variable Adopted (n ¼ 144) Comparison (n ¼ 131)

Age (years) M ¼ 37.76 M ¼ 37.67
Gender (%) 77.08 female 73.28 female
Relationship status (%) 25.40 single 31.71 single
12.70 cohabiting 10.57 cohabiting
42.86 married 46.34 married
19.05 divorced 11.38 divorced
Parental status (%) 57.64 children 49.23 children
Education (%) 9.72 , Year 12 9.16 ,Year 12
9.03 Year 12 6.11 Year 12
15.97 technical college 13.74 technical college
65.28 some 70.99 some
university study university study
Employment (%) 50.00 full time 43.08 full time
34.72 part time 32.31 part time
15.28 unemployed 24.62 unemployed
136 J. A. Feeney, N. L. Passmore, and C. C. Peterson

Attrition. To minimize attrition, we Second, a mixed-design multivariate anal-

informed participants of the importance of ysis of variance was conducted on the two
their continued involvement in the study. We scales of the ASQ, with adoptive status varied
also telephoned them prior to the second mail- between subjects, and time (Time 1, Time 2)
out, to check the accuracy of contact details varied within. The multivariate tests revealed
and to encourage them to complete the materi- only a significant main effect of adoptive sta-
als. The rate of attrition was minimal for both tus, multivariate F(2, 262) ¼ 10.23, p , .001,
groups. Specifically, of the original 144 re- g2 ¼ .07. Univariate tests showed that this
spondents in the adopted group, 138 com- significant group difference applied to both
pleted the follow-up questionnaire. Similarly, avoidance, F(1, 273) ¼ 6.96, p , .001, and
of the 131 comparison respondents, 128 com- anxiety, F(1, 273) ¼ 16.82, p , .001. Consis-
pleted the study. Given the very low numbers tent with Hypothesis 1, adoptees obtained
of respondents who did not complete the sec- higher scores than nonadoptees on both attach-
ond questionnaire, it was not feasible to con- ment dimensions (see Table 2).
duct statistical analyses to check that attrition
was random. Nevertheless, all subsequent data Role of parental relationships. To further
analyses were restricted to those who com- explore predictors of the two attachment
pleted both assessments. dimensions (assessed at Time 1), we con-
ducted hierarchical regression analyses. Based
on the natural chronology of events, we
Results
entered adoptive status (adopted ¼ 0, compar-
The results are presented in two sections, cor- ison ¼ 1) at Step 1, and reports of parental
responding to the two broad aims of the study. bonding at Step 2 (see Table 3 for bivariate
First, we investigated the roles of adoptive sta- correlations and regression weights).1 At Step
tus and family variables in predicting attach- 1, adoptive status was inversely related to
ment measures, together with patterns of avoidance (R 2 ¼ .04, p , .001) and to anxiety
stability and change. Second, we examined (R 2 ¼ .07, p , .001), that is, adoptees reported
the role of the attachment scales in predicting more insecurity than nonadoptees. At Step 2,
relationship variables assessed at follow-up, the effect of adoptive status remained signifi-
including the issue of mediation. cant for both dependent variables. Consistent
with Hypothesis 2, however, the parental
bonding scales were also strongly predictive;
Adoption as a risk factor for
R 2 increment ¼ .18 for avoidance and .10 for
insecure attachment
anxiety, p , .001 in each case. Specifically,
Role of adoptive status. The association maternal (b ¼ 2.24) and paternal care (b ¼
between adoptive status and attachment secu- 2.21) were negatively linked to avoidance,
rity was assessed in two ways. First, a fre- and paternal care (b ¼ 2.16) was negatively
quency comparison was conducted, relating linked to anxiety. We also checked for inter-
adoptive status (comparison vs. adopted) to active effects of adoptive status and parenting
the four-group (forced-choice) measure of scales, but no significant effects emerged. For
attachment style completed at Time 1. The completeness, we also repeated the analyses
association was highly significant, v2 (3) ¼
21.31, p , .001. Results showed that the num-
bers of insecure participants were greater in 1. The effects of adoptive status on perceptions of parent-
the adoption group than in the comparison ing were not a focus of this paper. Nonetheless, a mul-
group: preoccupied, 27 versus 16, respec- tivariate analysis of variance revealed an interaction
between adoptive status and gender of parent on PBI
tively; dismissing, 30 versus 21; and fearful, scores, multivariate F(2, 266) ¼ 5.01, p , .01. Adopt-
33 versus 12. Conversely, there were fewer ees perceived their mothers as less caring (M ¼ 2.07,
secure participants in the adoption group than SD ¼ 0.81) than did nonadoptees (M ¼ 2.28, SD ¼
0.73), and as more overprotective (Ms ¼ 1.29 and 1.02,
in the comparison group (49 vs. 80). These SDs ¼ 0.76 and 0.65, respectively); adoptive status was
results support Hypothesis 1. unrelated to perceptions of fathering.
Adoption, attachment, and relationship concerns 137

Table 2. Mean scores (and standard devia- Time 2 attachment dimensions; perceptions
tions) on attachment according to group and of a less satisfying reunion were related to later
time avoidance (r ¼ 2.32, p , .01) but not to later
anxiety (r ¼ 2.12, ns).
Attachment factor Adopted Comparison
Avoidance Stability and change in attachment. The
Time 1 3.12 (0.83) 2.83 (0.71) stability of attachment was investigated using
Time 2 3.08 (0.83) 2.84 (0.72) both categorical and continuous measures of
Anxiety attachment security. For the categorical mea-
Time 1 3.29 (1.00) 2.75 (0.88) sure, we calculated the proportion of partici-
Time 2 3.23 (1.03) 2.79 (0.88) pants who endorsed the same attachment style
at both times. The proportion was similar
Note. Scales scores (divided by the number of items) could for adoptees (n ¼ 79, 64.75%) and nonadopt-
range from 1 to 6. ees (n ¼ 79, 65.55%), v2(1) ¼ .02, ns. Further,
test–retest correlations for the scales assessing
with Time 2 attachment dimensions as the avoidance and anxiety were similar for both
dependent variables: Introducing the time lag samples, ranging from .77 to .80.
between the independent and dependent measures Despite the relative stability of the attach-
produced only small decreases in the amount of ment measures, adoptees’ working models
explained variance, and the pattern of prediction may be more sensitive to recent relationship
was the same as that already reported. events. To address this question, we conducted
For those adopted persons who had had regression analyses to predict attachment dimen-
contact with their birth mothers (n ¼ 84), we sions at Time 2, separately for each group and
also correlated the attachment dimensions each type of relationship change. The numbers
(Time 1) with the three ratings assessing per- of participants who had ended or formed a new
ceptions of relationships with birth mothers, relationship were too small to warrant analy-
controlling for time elapsed since the first con- sis; hence, analyses were restricted to the
tact. Perceptions of a less satisfying reunion measures of deterioration and improvement.
were associated with higher levels of avoid- Initial scores on attachment and relationship
ance (r ¼ 2.31, p , .01) and anxiety (r ¼ length were entered as control variables at Step
2.24, p , .05). This finding supports Hypoth- 1, and a dichotomous measure of relationship
esis 3. In contrast, ratings of current satisfac- change (i.e., whether a romantic relationship
tion and emotional closeness were unrelated to had deteriorated, or improved, in the 6-month
attachment (rs ranged from 2.10 to .21). interval) was entered at Step 2. (Participants
Again, these analyses were repeated using who had not had a romantic relationship at any

Table 3. Predicting Time 1 attachment from Parental Bonding Instrument scales

Avoidance Anxiety
Predictor r b r b
Step 1
Adoptive status 2.20** 2.20*** 2.26*** 2.26***
Step 2
Adoptive status 2.20** 2.16** 2.26*** 2.24***
Mother care 2.39*** 2.24** 2.25*** 2.05
Father care 2.30*** 2.21** 2.24*** 2.16*
Mother control .31*** .12 .29*** .12
Father control .20** 2.08 .26*** .07

Note. *p , .05. **p , .01. ***p , .001.

138 J. A. Feeney, N. L. Passmore, and C. C. Peterson

point during the study were omitted from these predict each relationship variable assessed at
analyses.) It is important to note that the num- follow-up (risk in intimacy; romantic, family,
bers of participants reporting recent relation- and social loneliness; relationship quality), from
ship changes did not differ according to group; adoptive status and Time 1 attachment security.
deterioration was reported by 26 adoptees and
19 nonadoptees, and improvement was reported Correlational data. Table 4 shows the cor-
by 37 adoptees and 35 nonadoptees. Further, relations among the variables used in the
the descriptions of relationship changes were regression analyses, separately for adoptees
similar for each group. Specifically, for both and nonadoptees. Associations between the
groups, the common difficulties were with- independent and dependent variables were gen-
drawal (by one or both partners), increased con- erally moderately strong, and there was no evi-
flict, lack of open communication, and outside dence of multicollinearity. As relationship
stressors such as work and study; conversely, quality was relevant only to participants who
the common areas of improvement were in- were currently in a romantic relationship, it is
creased closeness and commitment, and more not included in Table 4. For those participants
open and less conflicted communication. in a romantic relationship, however, correla-
As would be expected from the relative sta- tions between the independent variables and
bility of attachment, strong prediction of later relationship quality were as follows: for avoid-
attachment was provided by Time 1 attach- ance, r ¼ 2.24, p , .02 (adoptees) and 2.15,
ment scores and relationship length (R 2 ranged ns (nonadoptees); for anxiety, r ¼ 2.19, p ,
from .51 to .65). In the comparison group, nei- .06 (adoptees) and 2.06, ns (nonadoptees).
ther relationship deterioration nor improve- Mean scores for relationship quality were
ment added to the prediction of avoidance or 20.08 (SD ¼ 2.95) for adoptees and 0.14 (SD
anxiety at Step 2. For adoptees, however, ¼ 2.64) for nonadoptees.
reports of recent relationship deterioration pre-
dicted higher levels of anxiety (b ¼ .19, p , Regression analyses. In the regression
.005), and avoidance (b ¼ .15, p , .01). Fur- analyses, adoptive status (group) was entered
ther, the regression weight for the effect of at Step 1, and avoidance and anxiety were
relationship deterioration was significantly entered at Step 2. The two-way interaction
larger for adoptees than for nonadoptees in terms (Avoidance
Anxiety, Group
Avoid-
predicting both anxiety, t(196) ¼ 3.77, p , ance, Group
Anxiety) were entered at Step
.001, and avoidance, t(196) ¼ 2.33, p , 3, and the three-way interaction term was
.001. Similar analyses were also conducted, entered at Step 4. Interaction terms were cal-
assessing the effect of a composite measure culated using centered scores, as recommen-
of relationship distress (defined as the experi- ded by Aiken and West (1991). These results
ence of either relationship dissolution or dete- are summarized in Table 5; for ease of presen-
rioration). Again, this variable predicted later tation (and given the scarcity of interaction
avoidance (b ¼ .13, p , .05) and later anxiety effects, as noted later), only the first two steps
(b ¼ .13, p , .05) for adoptees only. These are summarized.2
results support Hypothesis 4.

2. Possible effects of gender were examined in prelimi-

The role of attachment in predicting nary regression analyses, in which gender and adoptive
relationship outcomes status were entered at Step 1, and their product term
was entered at Step 2; no main or interactive effects of
Our second aim was to assess the role of attach- gender were obtained. Further, although relatively few
ment dimensions in predicting relational adjust- participants were recruited through adoption-related
support groups or Internet sites, these participants
ment; within this broad question, we were may experience particular difficulty in dealing with
interested in possible interactive effects of their adoptive status, and hence may unduly influence
group and attachment (i.e., in the relative pre- the results. For this reason, the regression analyses were
repeated, deleting these participants. The results of
dictive power of attachment for each group). these analyses were almost identical to those reported
Hence, regression analyses were conducted to in Table 5.
 Adoption, attachment, and relationship concerns 139

At Step 1, adoptive status afforded signifi-

Note. In each cell, the first entry is for adoptees (N ¼ 144); the second entry is for nonadoptees (N ¼ 131). Scales scores (divided by the number of items) could range from 1 to 6 (avoidance,
2.63 (1.14),
2.26 (1.04)
loneliness
6. Social
cant prediction of three of the five dependent

—
measures, although the amount of explained
variance was small. Specifically, adoptive status
predicted risk in intimacy (R 2 ¼ .02, p , .02),
and both family (R 2 ¼ .02, p , .02) and social
loneliness (R 2 ¼ .03, p , .01). In each instance,

.46***, .43***
loneliness
5. Family

being adopted was associated with higher scores

2.17 (1.33),
1.81 (1.08)
(more negative outcomes) on these variables.
—
At Step 2, the attachment dimensions pro-
vided a significant increase in explained vari-
ance in all dependent measures, supporting
Hypothesis 5. The increment in explained var-
iance was substantial for measures of more
.54***, .39***
.34***, .36***
4. Romantic

general relationship experiences (e.g., .44 for

loneliness

2.67 (1.76),

risk in intimacy, .27 for social loneliness) but

2.74 (1.64)
—

was smaller for the measure of the quality of

a specific romantic relationship (.04). The
regression weights indicated that avoidance
was associated with perceptions of more risk
in intimacy, more loneliness (all three scales),
.38***, .50***
.61***, .53***
Table 4. Descriptive statistics and correlations among the focal variables (full sample)

and less relationship quality; anxiety predicted

.35***, .22*
3. Risk in
intimacy

greater risk in intimacy and more loneliness

2.61 (1.13),
2.30 (0.96)
—

(again, all three scales). Interestingly, adoptive

status no longer predicted any of the dependent
variables, once the attachment dimensions
were included (this point is revisited shortly,
in the section on mediation).
Interactive effects were very scattered.
.45***, .38***

.39***, .37***
.27**, .38***
.29***, .22*

Avoidance and anxiety interacted to predict

2. Anxiety
(Time 1)

3.29 (1.00),
2.75 (0.88)

perceived risk in intimacy (b ¼ .12, p ,

—

.05). Simple slopes analysis indicated that

the effect of avoidance was more marked at
high levels of anxiety (regardless of group);
those high in both avoidance and anxiety per-
anxiety, risk in intimacy), or from 1 to 7 (loneliness scales).

ceived the most risk in intimacy. Only one

.51***, .40***
.65***, .63***

.42***, .44***
.51***, .50***
.33***, .23**
1. Avoidance

product term involving group (of a possible

(Time 1)

3.12 (0.83),

15) was significant. Specifically, the three-

2.83 (0.71)
—

way interaction predicted family loneliness

(b ¼ .18, p , .02), with follow-up analyses
revealing a two-way (Avoidance
Anxiety)
*p , .05. **p , .01. ***p , .001.

interaction for nonadoptees only. High levels

of family loneliness were reported by non-
1. Avoidance (Time 1)

4. Romantic loneliness

adoptees who were high in both avoidance

5. Family loneliness
2. Anxiety (Time 1)

and anxiety (cf. fearful); in contrast, for adopt-

6. Social loneliness
3. Risk in intimacy

ees, high levels of family loneliness were

reported by those high in avoidance, regardless
of level of anxiety (cf. fearful or dismissing).
Variable

M (SD)

Attachment dimensions as mediators of the

effects of adoptive status. Results reported
140 J. A. Feeney, N. L. Passmore, and C. C. Peterson

Table 5. Summary of explained variance and significant predictors of relationship variables

Dependent variable Step 1 Step 2

Risk in intimacy R 2 ¼ .02*, adoptive R 2 increment ¼ .44***,
status b ¼ 2.15* avoidance b ¼ .59***,
anxiety b ¼ .16**
Romantic loneliness R 2 ¼ .01, ns R 2 increment ¼ .10***,
avoidance b ¼ .22**,
anxiety b ¼ .17*
Family loneliness R 2 ¼ .02*, adoptive R 2 increment ¼ .20***,
status b ¼ 2.15* avoidance b ¼ .37***,
anxiety b ¼ .15*
Social loneliness R 2 ¼ .03**, adoptive R 2 increment ¼ .27***,
status b ¼ 2.15* avoidance b ¼ .42***,
anxiety b ¼ .20***
Relationship quality R 2 ¼ .01, ns R 2 increment ¼ .04*,
avoidance b ¼ 2.18*

Note. Adoptive status was entered at Step 1; avoidance and anxiety were entered at Step 2. The analysis predicting
relationship quality was restricted to those who were currently in a romantic relationship.
*p , .05. **p , .01. ***p , .001.

to this point indicate that (a) adoptive status comes, supplementary analyses were con-
was related to attachment dimensions, (b) attach- ducted to explore the effects of age-related
ment dimensions were related to the dependent variables. First (and in light of the wide age
variables, and (c) all significant associations range of the sample), three age groups were
between adoptive status and the dependent formed: young adults (aged 18–30 years, n ¼
variables (risk in intimacy, family and social 83), middle adults (aged 31–40 years, n ¼ 96),
loneliness) became nonsignificant when the and older adults (aged 41 and older, n ¼ 95).
attachment dimensions were entered. These re- Supplementary regressions showed that the
sults are suggestive of mediated effects (Baron effects of adoptive status on the relationship
& Kenny, 1986). variables were generally similar across age
To provide a further check on possible groups, but adoptive status explained more
mediated relationships, the Sobel test was variance in social loneliness for young adults
used—this test evaluates whether the indirect than for all others (R 2 ¼ .09 and .01, respec-
effect of an independent variable on a depen- tively; z ¼ 2.30, p , .025). For adoptees, we
dent variable via a proposed mediator is sig- also examined the effects of age at adoption
nificantly different from zero (Howell, 2002). and age at learning adoptive status. (These
The Sobel tests provided strong and consistent variables showed substantial positive skew;
support for attachment as a mediator of the hence, a log transformation was applied.)
effects of adoptive status (Hypothesis 6). Spe- The only significant links with relationship
cifically, both avoidance and anxiety mediated variables were that later adoption placement
the effects of adoptive status on risk in inti- was associated positively with perceived risk
macy, and on social and family loneliness. in intimacy (r ¼ .18, p , .05) and family
The z values for the six tests ranged from loneliness (r ¼ .21, p , .02).
–3.11 to –3.79, and the associated p values
ranged from .002 to .0002.
Discussion
Supplementary analyses: Age-related effects. This study used attachment theory as a frame-
As adoptive status explained only small work to examine the relationship experiences
amounts of variance in the relationship out- of a sample of adults who were adopted as
Adoption, attachment, and relationship concerns 141

infants and a comparison sample of adults who values of kinship and maternal instinct cause
grew up with their biological parents. The us to view child relinquishment in terms of
results suggest that insecure attachment may rejection and abandonment. It is also possible
be more widespread among adoptees than that adoptees for whom the adoption experi-
among nonadoptees. Adoptees scored higher ence is a salient concern tend to attribute rela-
than comparison participants on avoidance tionship difficulties (arising for whatever
and anxiety, which have consistently emerged reason) to that experience. These arguments
as the two key dimensions underlying mea- do not deny the reality of adoption-related los-
sures of adult attachment (Brennan, Clark, & ses, however, or the possibility of some degree
Shaver, 1998). Adoptees were also overrepre- of loss linked to the experience of being par-
sented in the insecure attachment categories, ented by nonkin (Leon). In short, there is clear
particularly the fearful style. According to evidence that many adoptees feel rejected or
attachment theory, fearful attachment repre- abandoned by birth parents (e.g., Brodzinsky,
sents the most problematic pattern of working 1990; Jones, 1997); hence, the attachment per-
models, involving negative perceptions both spective is relevant, given its focus on working
of self-worth and of the availability and models of the self and others.
responsiveness of others (Bartholomew & Consistent with Hypothesis 2, however,
Horowitz, 1991). Further, empirical research results clearly indicated that attachment secu-
on the characteristics of the four styles sup- rity was not a function of adoptive status alone.
ports this assertion (e.g., Feeney et al., 1994). In fact, self-reports of parental bonding were
These findings on the attachment characteris- more powerful predictors of the attachment
tics of the samples support Hypothesis 1 and dimensions than was adoptive status, and the
corroborate the suggestion that attachment patterns of prediction did not differ according
theory provides a useful perspective on rela- to group. (Similarly, research by Passmore
tionship issues that arise for adoptees, includ- et al., 2005, has shown that parental variables
ing loss, search, and reunion (Edens & Cavell, are more important than adoptive status in pre-
1999). dicting self-esteem.) Perceptions of care and
At the same time, it is important to affection from mothers and fathers during
acknowledge that adoptive status explained childhood were particularly relevant in pre-
only 7% of the variance in attachment, despite dicting adult attachment security. This finding
the statistical significance of the associations. fits with a large body of literature linking
It is also important to note that adoptees in this attachment security to experiences of sensitive
sample were adopted as young infants. For this and responsive caregiving (e.g., Rothbard &
reason, a specific attachment between infant Shaver, 1994) and highlights the influential
and birthmother was unlikely to have formed role of both parents in the bonding process.
before the child was relinquished (and hence, The fact that patterns of prediction were sim-
unlikely to have been lost); theory and re- ilar for both groups highlights the importance
search suggest that attachment bonds develop of adoptees’ experiences with their adoptive
through a history of repeated, mutual interac- parents and is consistent with research show-
tions and involve the infant’s conscious aware- ing that open and caring relationships in the
ness of the attachment figure as a distinct adoptive family facilitate adjustment (Kelly,
person (Noller, Feeney, & Peterson, 2001). Towner-Thyrum, Rigby, & Martin, 1998).
Further, although Verrier (1993) argued (from Despite the undisputed importance of rela-
a psychoanalytic perspective) that separating tionships with adoptive parents, the role of
an infant from the mother inflicts lasting dam- interactions with birth relatives should not be
age by disrupting prenatal attachment and the ignored. With recent changes in legislation in
infant’s emerging sense of self; this claim is many countries (regarding disclosure of iden-
highly contentious, and some suggest that tifying information), increasing numbers of
adoption-related losses may be largely socially adult adoptees are searching for birth relatives
constructed. Leon (2002), for example, pro- and subsequently having reunions; moreover,
posed that deep-seated cultural beliefs in the although some search and reunion experiences
142 J. A. Feeney, N. L. Passmore, and C. C. Peterson

are very positive, others involve perceptions of insecurity are likely to interfere with the reso-
further loss and rejection. In the present study, lution of relationship problems. In short, the
for adoptees who had met their birth mothers, present data highlight the fact that working
dissatisfaction with the initial reunion was models are somewhat malleable, even in adult-
associated with higher levels of avoidance hood, and provide indirect support for Davila
and anxiety. This finding is consistent with et al.’s (1997) claim that factors such as paren-
Hypothesis 3, but only the link with avoidance tal adjustment difficulties and family disrup-
was robust over time. Hence, the nature of the tion may affect the coherence of working
association between attachment anxiety and models.
dissatisfaction with the reunion is unclear—it The fifth hypothesis proposed that insecurity
is possible that anxiety colors reports of the would predict relationship variables assessed 6
reunion experience, rather than (or in addition months later. This hypothesis was supported.
to) the reverse. After controlling for adoptive status, the attach-
Interestingly, only reports of the initial ment dimensions provided a significant increase
reunion with birth mothers (as opposed to cur- in explained variance in all relationship mea-
rent closeness and satisfaction) were linked to sures. The increase in explained variance was
adult attachment security. This finding may quite large for measures of general relationship
reflect the highly emotional nature of the attitudes and experiences (risk in intimacy, and
search and reunion experience (Verrier, 1993); the three forms of loneliness) but was smaller
further, adoptees bring particular hopes and for the measure tapping perceptions of a specific
expectations to the reunion experience, and couple bond. One explanation for these findings
failure to fulfill these may negatively impact is that ratings of a specific relationship are
working models. Perceptions of current rela- influenced more strongly by actual partner
tionships with birth mothers may have less characteristics and by the course of couple inter-
impact on attachment models for at least two action, than by the individual’s working models
reasons. First, some of these relationships do of attachment. In addition, excluding partici-
not extend beyond the initial reunion or its pants who were not currently in a romantic rela-
immediate aftermath. Second, even in cases tionship resulted in some loss of statistical
where there is some ongoing contact, these power.
relationships may be less formative than those There were few interactions with adoptive
that are established early in life with primary status in predicting relationship variables, sug-
caregivers and involve very regular and sus- gesting that the implications of attachment for
tained interaction. relationship functioning are similar for adopt-
The attachment measures were relatively ees and nonadoptees. An exception was the
stable for both samples over the course of the finding that family loneliness was related to
study. These results are consistent with the high avoidance in adoptees (a main effect)
claim that in adulthood, mental models of but to the combination of high avoidance and
attachment ‘‘tend to persist relatively un- high anxiety in nonadoptees. Perhaps, adop-
changed’’ (Bowlby, 1973, p. 235), although tion-related losses tend to sensitize avoidant
Bowlby recognized that these models could adoptees to the possibility of rejection by fam-
be accommodated to fit a changing social real- ily members, resulting in perceptions of
ity. Indeed, within the adopted sample only, ‘‘being alone’’ within the family (regardless
reports of recent relationship deterioration of level of anxiety).
predicted later avoidance and anxiety, after In contrast to the very scattered interactive
controlling for initial attachment scores. This effects, adoptive status itself predicted per-
finding supports Hypothesis 4 and suggests ceived risk in intimacy and reports of family
that adoptees’ working models are more sen- and social loneliness. It is interesting to spec-
sitive to recent relationship events that involve ulate on why adoptees and nonadoptees dif-
challenge or perceived threat, such as relation- fered on these relational measures but not the
ship conflict and emotional distancing. This remaining ones. The remaining measures
result is noteworthy, given that increases in focused on romantic relationships, either in
Adoption, attachment, and relationship concerns 143

terms of relating to a specific partner (relation- Hence, it has been suggested that researchers
ship quality) or in terms of perceiving a should measure preadoptive risk history, con-
bond with ‘‘one special person’’ who provides sider the interdependent nature of biological
love and encouragement (romantic loneliness). and environmental risk factors, and examine
Adoptees may perceive fewer challenges with the cumulative nature of risk factors (Brooks,
these issues than with developing closeness Simmel, Wind, & Barth, 2005).
and trust in the family context, feeling ‘‘in
tune’’ with others generally, and being confi-
Limitations and strengths of the study
dent that others will not be hurtful or rejecting.
Given that health and well-being are predicted The present study had several limitations.
by perceived family support and by a sense of First, it involved convenience sampling as
belonging to a strong social network (e.g., there was no register of adoptions occurring
Cunningham & Barbee, 2000), these latter in this era. Further, the samples were predom-
issues are clearly important. inantly female. Although this gender imbal-
A key finding of this study was the consis- ance is typical in studies of adult adoptees, it
tent support for the mediating role of attach- may have influenced some of the results. For
ment in the association between adoptive example, the four-group measure of attach-
status and relationship outcomes (Hypothesis ment often shows gender differences (Feeney
6). That is, the negative associations of adop- & Noller, 1996); hence, the distribution of
tive status with risk in intimacy and family and attachment styles may differ somewhat in a
social loneliness were fully mediated by more balanced sample. Gender was not, how-
avoidance and anxiety. This finding points to ever, related to adoptive status or relationship
the pivotal role of insecurities regarding the outcomes and did not moderate the effect of
worthiness of the self and the dependability adoptive status on these outcomes; hence, the
of others and lends further support to the rele- gender imbalance is unlikely to affect interpre-
vance of attachment principles for the study of tation of the focal results. The fact that the
adoptees’ relationships in adulthood. study was conducted in a single country raises
In short, our findings support the claim that concerns over the generality of the findings as
adoption may represent a risk factor for rela- the acceptability of adoption varies in different
tional difficulties later in life, at least in terms countries and cultures; further, more recent
of family relationships, general perceptions of adoptions in Australia are less likely to involve
intimacy and belonging, and reactivity to rela- secrecy and anonymity, and more likely to
tionship stressors. The amount of variance involve children from other countries and eth-
explained by adoptive status was small, how- nic backgrounds. As noted earlier, however,
ever, and many other factors are likely to similar adoption practices to those discussed
impact on adoptees’ relational adjustment. In in this paper occurred in many western coun-
the present study, later adoption placement tries for much of the 20th century.
was linked to family loneliness and perceived In terms of adoptees’ relationship history,
risk in intimacy. Moreover, adoptive status the focus was on relationships with adoptive
explained more variance in social loneliness parents and birth mothers, rather than on cat-
for young adults than for older adults, suggest- egorical variables such as search status (hav-
ing that some of the effects of adoption may ing searched or not searched for birth relatives)
decrease with the passage of time or with or reunion status (having been reunited with
increased relationship experience. Although one or more birth relatives). Other studies sug-
beyond the scope of this study, the extent of gest that these variables also predict adjust-
secrecy surrounding the adoption and the man- ment outcomes (e.g., Borders et al., 2000;
ner of learning one’s adoptive status are also Cubito & Obremski-Brandon, 2000). In addi-
likely to be relevant (Brodzinsky, 2005). In tion, to maintain adequate sample sizes, we
addition, biological factors (both prenatal and did not investigate relationships with other
genetic) may impact on adjustment (Brodzinsky, birth relatives with whom reunions were less
1990) and also affect attachment security. common, such as fathers and siblings—these
144 J. A. Feeney, N. L. Passmore, and C. C. Peterson

relationships may also have an important many studies in this area have relied on cross-
impact on some adoptees. sectional data, which are prone to the problem
All measures in the study involved self- of common method variance. In addition, the
reports from the same informant, leading to broad method of recruitment is an important
the possibility of informant bias. There is, feature; much research on adoption has relied
however, growing evidence of concordance on samples recruited largely through clinical
between self-reports and observers’ ratings of settings or support groups, which are unlikely
relational behavior (e.g., Hahlweg, Kaiser, to be representative of the wider population.
Christensen, Fehm-Wolfsdorf, & Groth, 2000). Further, the attachment perspective provides
Similarly, many studies indicate that self- a strong theoretical framework that recognizes
reports of adult attachment do not simply tap both the formative nature of early social expe-
a generalized tendency to perceive events riences and the possibility of later revision of
more or less favorably. For example, these working models.
reports have been linked to independent rat-
ings of behavior (e.g., Rholes et al., 1998),
Implications of the findings
corroborative reports from friends and partners
(e.g., Kobak & Hazan, 1991), and indexes of A key finding of this study was that adoptive
unconscious processes, including physiologi- status no longer predicted relational adjust-
cal measures of arousal and affect regulation ment when attachment dimensions were
(e.g., Diamond & Hicks, 2005). included in the analyses. In other words, it
As already noted, adoptive status explained seems that the greater levels of loneliness
small amounts of variance in the relational and risk in intimacy reported by adoptees
measures, although the attachment dimensions reflect their negative working models of self
provided relatively strong prediction. As Whitley or others. This finding has important implica-
(1996) noted, however, even a small effect tions for those working in counseling and clin-
can have practical significance if it (a) applies ical settings. In recent years, attention has been
to a large population, (b) is cumulative, or (c) directed to developing interventions based
informs theory. All these factors are relevant directly on attachment principles (e.g., Davila,
to studies of adoption and attachment. First, 2003; Johnson, 2003). This work recognizes
although it is impossible to establish exact the traumatizing nature of loss and rejection,
numbers of adults who were adopted in the the self-fulfilling nature of working models,
past under systems involving secrecy and ano- and the therapeutic effect of interventions that
nymity, the numbers are likely to be large; in focus on the need for secure emotional con-
terms of current statistics, between 2% and 4% nections. Such interventions may offer a useful
of American families include an adopted child approach for those struggling with adoption-
(Brooks et al., 2005), and 6 in 10 Americans related issues (including reunions) and those
have had some personal experience of adop- dealing with other complex family experiences
tion (including that of friends or family mem- such as foster care (Levy & Orlans, 2003).
bers; Evan B. Donaldson Institute, 1997). Other aspects of our findings are also rele-
Further, variables that increase insecurity tend vant to practitioners. First, the small effect
to have cumulative effects, insofar as working sizes for adoptive status fit with previous
models are self-fulfilling; for instance, some- research, which suggests that most adoptees
one who fears rejection is likely to approach are relatively well adjusted (e.g., Nickman
others defensively, eliciting further rejection et al., 2005). Hence, practitioners should not
and exacerbating insecurity (Feeney, 1999). assume that adoptees invariably face major
Finally, studies of adoption and attachment adjustment problems. Similarly, our findings
add to the growing body of literature support- point to the importance of various family rela-
ing the relevance of attachment principles to tionships as influences on adjustment. Percep-
adults’ close relationships. tions of relationships with adoptive parents
A notable strength of the present study is are clearly important. In addition, reunions with
its short-term longitudinal design. In contrast, birth relatives are salient experiences, and
Adoption, attachment, and relationship concerns 145

adoptees’ expectations of this process are Boon, S. D., & Holmes, J. G. (1992, July). Walking the
tightrope: Risk in dating relationships. Paper pre-
likely to impact on their satisfaction with the sented at the Sixth International Conference on Per-
outcome. Practitioners should also be sensitive sonal Relationships, Orono, Maine.
to the important features of different types of Borders, L. D., Penny, J. M., & Portnoy, F. (2000). Adult
adoptees and their friends: Current functioning and
adoption, including ‘‘traditional’’ (as studied psychosocial well-being. Family Relations, 49, 407–
in the present research), international, and spe- 418.
cial-needs adoptions; although the theme of Bowlby, J. (1973). Attachment and loss: Vol. 2. Separa-
tion: Anxiety and anger. New York: Basic Books.
loss is central to adoption, many other issues Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-
are specific to the context of the adoption. report measurement of adult attachment: An integra-
tive overview. In J. A. Simpson & W. S. Rholes (Eds.),
Attachment theory and close relationships (pp. 46–76).
Conclusions New York: Guilford.
Brodzinsky, D. M. (1990). A stress and coping model of
This study employed a short-term longitudinal adoption adjustment. In D. M. Brodzinsky & M. D.
design to investigate the relationship experi- Schechter (Eds.), The psychology of adoption (pp. 3–
ences of adopted and nonadopted adults. The 24). Oxford, UK: Oxford University Press.
Brodzinsky, D. M. (2005). Reconceptualizing openness in
results support the view that adoption may be adoption: Implications for theory, research, and prac-
a risk factor for negative relational attitudes tice stress. In D. M. Brodzinsky & J. Palacios (Eds.),
and relationship difficulties in adult life, par- Psychological issues in adoption: Research and prac-
tice (pp. 145–166). Westport, CT: Praeger.
ticularly in terms of attachment security. The Brooks, D., Simmel, C., Wind, L., & Barth, R. P. (2005).
effects were generally small, however, and Contemporary adoption in the United States: Implica-
relationships with adoptive parents and reun- tions for the next wave of adoption theory, research
and practice. In D. M. Brodzinsky & J. Palacios (Eds.),
ions with birth mothers also emerged as factors Psychological issues in adoption: Research and prac-
that impact on adoptees’ attachment security. tice (pp. 1–25). Westport, CT: Praeger.
Overall, the findings support the call to view Collishaw, S., Maughan, B., & Pickles, A. (1998). Infant
adoption: Psychosocial outcomes in adulthood. Social
the adoption experience as a ‘‘lifelong pro- Psychiatry and Psychiatric Epidemiology, 33, 57–65.
cess’’ (Borders et al., 2000) and suggest that Cooper, M. L., Shaver, P. R., & Collins, N. L. (1998).
attachment principles can be usefully applied Attachment styles, emotion regulation, and adjustment
in adolescence. Journal of Personality and Social Psy-
in examining these issues. chology, 74, 1380–1397.
Cubito, D. S., & Obremski-Brandon, K. (2000). Psycho-
References logical adjustment in adult adoptees: Assessment of
distress, depression, and anger. American Journal of
Aiken, L. S., & West, S. G. (1991). Multiple regression: Orthopsychiatry, 70, 408–413.
Testing and interpreting interactions. Newbury Park, Cunningham, M. R., & Barbee, A. P. (2000). Social sup-
CA: Sage. port. In C. Hendrick & S. S. Hendrick (Eds.), Close
Ainsworth, M. D. S. (1989). Attachments beyond infancy. relationships: A sourcebook (pp. 273–285). Thousand
American Psychologist, 44, 709–716. Oaks, CA: Sage.
Alexander, R. P., Feeney, J. A., Hohaus, L., & Noller, P. Davila, J. (2003). Attachment processes in couple therapy:
(2001). Attachment style and coping resources as pre- Informing behavioral models. In S. M. Johnson & V.
dictors of coping strategies in the transition to parent- E. Whiffen (Eds.), Attachment processes in couple and
hood. Personal Relationships, 8, 137–152. family therapy (pp. 124–143). New York: Guilford.
Baldwin, M. W., Keelan, J. P. R., Fehr, B., Enns, V., & Davila, J., Burge, D., & Hammen, C. (1997). Why does
Koh-Rangarajoo, E. (1996). Social-cognitive concep- attachment style change? Journal of Personality and
tualization of attachment working models: Availability Social Psychology, 73, 826–838.
and accessibility effects. Journal of Personality and Diamond, L. M., & Hicks, A. M. (2005). Attachment style,
Social Psychology, 71, 94–109. current relationship security, and negative emotions:
Baron, R. M., & Kenny, D. A. (1986). The moderator- The mediating role of physiological regulation. Jour-
mediator variable distinction in social psychological nal of Social and Personal Relationships, 22, 499–
research: Conceptual, strategic, and statistical consid- 518.
erations. Journal of Personality and Social Psychol- DiTommaso, E., & Spinner, B. (1993). The development
ogy, 51, 1173–1182. and initial validation of the Social and Emotional
Bartholomew, K., & Horowitz, L. M. (1991). Attachment Loneliness Scale for Adults (SELSA). Personality
styles among young adults: A test of a four-category and Individual Differences, 14, 127–134.
model. Journal of Personality and Social Psychology, Doherty, N., & Feeney, J. A. (2004). The composition of
61, 226–244. attachment networks throughout the adult years. Per-
Baumeister, R. F., & Leary, M. R. (1995). The need to sonal Relationships, 11, 469–488.
belong: Desire for interpersonal attachments as a fun- Edens, J. F., & Cavell, T. A. (1999). A review and refor-
damental human motivation. Psychological Bulletin, mulation of adoptive relationships from an attachment
117, 497–529. perspective. Adoption Quarterly, 3, 43–70.
146 J. A. Feeney, N. L. Passmore, and C. C. Peterson

Evan B. Donaldson Institute (1997, November). Bench- ment processes in couple and family therapy (pp. 165–
mark adoption survey: Report on the findings. Prince- 190). New York: Guilford.
ton, NJ: Princeton Survey Associates. Levy-Shiff, R. (2001). Psychological adjustment of adopt-
Feeney, J. A. (1999). Adult romantic attachment and ees in adulthood: Family environment and adoption-
couple relationships. In J. Cassidy & P. R. Shaver related correlates. International Journal of Behavioral
(Eds.), The handbook of attachment: Theory, research, Development, 25, 97–104.
and clinical applications (pp. 355–377). New York: Nickman, S. L. (1985). Losses in adoption: The need for
Guilford. dialogue. Psychoanalytic Study of the Child, 40, 365–398.
Feeney, J. A., & Noller, P. (1996). Adult attachment. Nickman, S. L., Rosenfeld, A. A., Fine, P., MacIntyre, J.
Thousand Oaks, CA: Sage. C., Pilowsky, D. J., Howe, R., et al. (2005). Children in
Feeney, J. A., & Noller, P. (2004). Attachment and sexu- adoptive families: Overview and update. Journal of
ality in close relationships. In J. H. Harvey, A. Wenzel, the American Academy of Child and Adolescent Psy-
& S. Sprecher (Eds.), Handbook of sexuality in close chiatry, 44, 987–995.
relationships (pp. 183–201). Mahwah, NJ: Erlbaum. Noller, P., Feeney, J. A., & Peterson, C. (2001). Personal
Feeney, J. A., Noller, P., & Hanrahan, M. (1994). Assessing relationships across the lifespan. Philadelphia: Psy-
adult attachment: Developments in the conceptualization chology Press.
of security and insecurity. In M. B. Sperling & W. H. Parker, G. (1983). Parental overprotection: A risk factor in
Berman (Eds.), Attachment in adults: Theory, assess- psychosocial development. New York: Grune & Stratton.
ment, and treatment (pp. 128–152). New York: Guilford. Parker, G., Tupling, H., & Brown, L. B. (1979). A Parental
Hahlweg, K., Kaiser, A., Christensen, A., Fehm- Bonding Instrument. British Journal of Medical Psy-
Wolfsdorf, G., & Groth, T. (2000). Self-report and chology, 52, 1–10.
observational assessment of couples’ conflict: The Passmore, N. L., Fogarty, G. J., Bourke, C. J., & Baker-
concordance between the Communication Patterns Evans, S. F. (2005). Parental bonding and identity style
Questionnaire and the KPI Observation System. as correlates of self-esteem among adult adoptees and
Journal of Marriage and the Family, 62, 61–67. non-adoptees. Family Relations, 54, 523–534.
Hammond J. R., & Fletcher, G. J. O. (1991). Attachment Pietromonaco, P. R., Greenwood, D., & Feldman Barrett,
styles and relationship satisfaction in the development L. (2004). Conflict in adult close relationships: An
of close relationships. New Zealand Journal of Psy- attachment perspective. In W. S. Rholes & J. A. Simp-
chology, 20, 56–62. son (Eds.), Adult attachment: Theory, research, and
Hazan, C., & Shaver, P. R. (1987). Romantic love concep- clinical implications (pp. 267–299). New York: Guil-
tualized as an attachment process. Journal of Person- ford.
ality and Social Psychology, 52, 511–524. Pilkington, C. J., & Richardson, D. R. (1988). Perceptions
Hoksbergen, R., & ter Laak, J. (2005). Changing attitudes of risk in intimacy. Journal of Social and Personal
of adoptive parents in northern European countries. In Relationships, 5, 503–508.
D. M. Brodzinsky & J. Palacios (Eds.), Psychological Rholes, W. S., Simpson, J. A., & Grich Stevens, J. (1998).
issues in adoption: Research and practice (pp. 27–46). Attachment orientations, social support, and conflict
Westport, CT: Praeger. resolution in close relationships. In J. A. Simpson &
Howell, D. C. (2002). Statistical methods for psychology W. S. Rholes (Eds.), Attachment theory and close rela-
(5th ed.). Pacific Grove, CA: Duxbury. tionships (pp. 166–188). New York: Guilford.
Johnson, S. M. (2003). Attachment theory: A guide for Rothbard, J. C., & Shaver, P. R. (1994). Continuity of
couple therapy. In S. M. Johnson & V. E. Whiffen attachment across the life span. In M. B. Sperling &
(Eds.), Attachment processes in couple and family W. H. Berman (Eds.), Attachment in adults: Theory,
therapy (pp. 103–123). New York: Guilford. assessment, and treatment (pp. 31–71). New York:
Jones, A. (1997). Issues relevant to therapy with adoptees. Guilford.
Psychotherapy, 34, 64–68. Rusbult, C. E., Martz, J. M., & Agnew, C. R. (1998). The
Juffer, F., & van IJzendoorn, M. H. (2005). Behavior prob- Investment Model Scale: Measuring commitment
lems and mental health referrals of international adopt- level, satisfaction level, quality of alternatives, and
ees: A meta-analysis. Journal of the American Medical investment size. Personal Relationships, 5, 357–391.
Association, 293, 2501–2515. Ryburn, M. (1994). Open adoption: Research, theory, and
Kelly, M. M., Towner-Thyrum, E., Rigby, A., & Martin, practice. Aldershot, UK: Avebury.
B. (1998). Adjustment and identity formation in adop- Schechter, M. D., & Bertocci, D. (1990). The meaning of
ted and nonadopted young adults: Contributions of the search. In D. M. Brodzinsky & M. D. Schechter
family environment. American Journal of Orthopsy- (Eds.), The psychology of adoption (pp. 62–90).
chiatry, 68, 497–500. Oxford, UK: Oxford University Press.
Kirkpatrick, L. E., & Hazan, C. (1994). Attachment styles Shaver, P. R., & Brennan, K. A. (1992). Attachment styles
and close relationships: A four-year prospective study. and the ‘‘big five’’ personality traits: Their connections
Personal Relationships, 1, 123–142. with each other and with romantic relationship out-
Kobak, R. R., & Hazan, C. (1991). Attachment in mar- comes. Personality and Social Psychology Bulletin,
riage: Effects of security and accuracy of working 18, 536–545.
models. Journal of Personality and Social Psychol- Shaver, P. R., Collins, N., & Clark, C. L. (1996). Attach-
ogy, 60, 861–869. ment styles and internal working models of self and
Leon, I. G. (2002). Adoption losses: Naturally occurring relationship partners. In G. J. O. Fletcher & J. Fitness
or socially constructed? Child Development, 73, 652– (Eds.), Knowledge structures in close relationships: A
663. social psychological approach (pp. 25–61). Mahwah,
Levy, T. M., & Orlans, M. (2003). Creating and repairing NJ: Erlbaum.
attachments in biological, foster, and adoptive fami- Shaver, P. R., & Hazan, C. (1993). Adult romantic
lies. In S. M. Johnson & V. E. Whiffen (Eds.), Attach- attachment: Theory and evidence. In D. Perlman &
Adoption, attachment, and relationship concerns 147

W. Jones (Eds.), Advances in personal relationships van IJzendoorn, M. H., Juffer, F., & Klein Poelhuis, C. W.
Vol. 4 (pp. 29–70). London: Jessica Kingsley. (2005). Adoption and cognitive development: A meta-
Shaver, P. R., & Mikulincer, M. (2002). Attachment- analytic comparison of adopted and nonadopted child-
related psychodynamics. Attachment & Human Devel- ren’s IQ and school performance. Psychological Bul-
opment, 4, 133–161. letin, 131, 301–316.
Shaver, P. R., & Mikulincer, M. (2004). What do self-report Verrier, N. N. (1993). The primal wound: Understanding
attachment measures assess? In W. S. Rholes & J. A. the adopted child. Baltimore, MD: Gateway Press.
Simpson (Eds.), Adult attachment: Theory, research, and Weiss, R. S. (1991). The attachment bond in childhood
clinical implications (pp. 17–54). New York: Guilford. and adulthood. In C. M. Parkes, J. Stevenson-Hinde,
Simpson, J. A. (1990). Influence of attachment styles on & P. Marris (Eds.), Attachment across the life cycle
romantic relationships. Journal of Personality and (pp. 66–76). London: Tavistock/Routledge.
Social Psychology, 59, 971–980. Whitley, B. E., Jr. (1996). Principles of research in
Tieman, W., van der Ende, J., & Verhulst, F. C. (2005). behavioral science. Mountain View, CA: Mayfield.
Psychiatric disorders in young adult intercountry Wierzbicki, M. (1993). Psychological adjustment of
adoptees: An epidemiological study. American Jour- adoptees: A meta-analysis. Journal of Clinical Child
nal of Psychiatry, 162, 592–598. Psychology, 22, 447–454.