Child Development, May/June 2000, Volume 71, Number 3, Pages 690–694
Continuity and Discontinuity of Attachment from Infancy
through Adolescence
Claire E. Hamilton
This study reports relations between infant Ainsworth Strange Situation classifications, negative life events,
and Adolescent Attachment Interview classifications. Overall, the stability of secure versus insecure classifications
was 77%, and infant attachment classification was a significant predictor of adolescent attachment classification.
Chi-square analyses indicate that negative life events are significantly related to change in attachment
classification. The sample (n ϭ 30) is drawn from the Family Lifestyles Project (FLS), an ongoing longitudinal
study of children’s development within the context of nonconventional family lifestyles. The distribution of
family lifestyles within this study, unlike those in the full FLS sample, represent a higher proportion of conventional
two-parent families (40%). There were no differences between adolescents reared in conventional or
nonconventional families in the distribution of adolescent attachment security, the experience of negative life
events, or the continuity of attachment from infancy through adolescence.
INTRODUCTION
This study examines the stability and continuity of attachment
security from infancy through adolescence.
It extends our understanding of attachment by exploring
continuity beyond childhood in adolescents
who have been reared in both conventional and nonconventional
family contexts. By studying attachment
in this group we can examine how attachment
functions within a broader definition of family and
family life. The adolescents in this study are drawn
from the Family Lifestyle Project (FLS),1 ongoing longitudinal
study of the influence of nonconventional
family lifestyles on children’s development. The FLS
families at the beginning of the project were evenly
distributed across five major lifestyles: single
mothers, domestic living groups (loosely affiliated
groups of nonrelated adults and children), creedal
communal groups, unmarried cohabitating couples
(“social contract” couples), and a comparison group
of conventional two-parent families. The subsample
of families and adolescents included in this study represent
a higher proportion of conventional families
(40%) than that found in the full FLS Project. Nevertheless,
the sample does differ from the more traditional
family contexts usually included in developmental
research.
Generally it is assumed that attachment should
remain stable over time. Various models may account
for this stability. Bowlby (1982) proposed that infants
in their development of attachment relationships also
form internal working models of themselves and the
social world. Although change in this internal working
model is possible, over the course of early childhood
the internal working model becomes less flexible
and consciously accessible and so may be less
susceptible to change. An alternative model focuses
on the environmental supports for attachment; stability
in this framework rests on the notion that environments
tend to be relatively stable and so continuity of
attachment is maintained by the environment rather
than a within-person characteristic (Lamb, Thompson,
Gardner, & Charnov, 1985). These two models are difficult
to isolate in the real world and are not directly
tested in the current study. However, the study does
examine environmental circumstances that might
support or disrupt attachment continuity. In general,
we would expect that factors influencing change in
attachment would include disruption in the caregiver–
child relationship through separation or loss and general
family stressors, which may reduce the caregiver’s
ability to provide. In this study, these factors
are more likely to occur in the families with nonconventional
lifestyles.
© 2000 by the Society for Research in Child Development, Inc.
All rights reserved. 0009-3920/2000/7103-0014
1 FLS began in 1973 under the leadership of Bernice T. Eiduson
and colleagues (Alexander, 1978; Eiduson, Cohen, & Alexander,
1973; Eiduson & Weisner, 1978). The project is currently
directed by Thomas S. Weisner, Department of Psychiatry, University
of California, Los Angeles.
This is one of three long-term longitudinal studies assessing
infant attachment. See Waters, Hamilton, & Weinfield, “The Stability
of Attachment Security from Infancy to Adolescence and
Early Adulthood: General Introduction,” for an overall view of
study design, measures, and supporting references.
Hamilton 691
There were no initial differences in the distribution
of infant attachment classifications across the various
family lifestyles. Over the sample as a whole the proportions
of secure and insecure infant attachment
classifications were similar to those found generally
in the United States (Bernstein, Zimmerman, & Eiduson,
1981). Although potential risk factors associated
with nonconventional lifestyles did not seem to influence
initial infant attachment security, these potential
risks may have a greater impact on the continuity of
attachment over time. Nonetheless, changes in family
circumstances occurred at a much higher rate in nonconventional
families than would be expected in more
traditional families with young children. Throughout
their early childhoods, children in the nonconventional
families experienced more frequent changes in
family composition (e.g., by the time children were 4
years old, 84% of the conventional families remained
intact versus only 52% of the social-contract families;
Eiduson, Kornfein, Zimmerman, & Weisner, 1982).
These patterns of instability in family configuration
persisted throughout childhood. Financial stresses
were also more prevalent in nonconventional families;
overall household income was lower and less
predictable than that found in conventional families.
Nonconventional parents were also more likely to experiment
with and regularly use drugs themselves
and to expose or involve their children in drug use
(Garnier & Weisner, 1992).
Elevated life stresses and changes in family circumstances
may pose challenges for children as they
develop. Although no differences due to family lifestyle
were found in early childhood across a variety
of cognitive, physical, or socioemotional outcomes in
the FLS Project (Weisner, 1986), some differences did
begin to emerge as children entered school. Children
in more unstable families with lower commitment to
their alternative lifestyles had lower academic achievement
than those from conventional families or from
nonconventional families with more commitment to
their alternative values (Weisner & Garnier, 1992).
Children in nonconventional familes were also exposed
to or involved in more parental drug use than
those in conventional families and this exposure was
associated with less competent behavior in elementary
school (Weisner & Garnier, 1992) and in later increased
rates of adolescent drug use (Garnier & Weisner,
1995).
This study examines the continuity of attachment
from infancy through adolescence in a sample of families
many of whom were actively experimenting
with family lifestyles and social roles. This study also
examines the influence of negative life events by
using the common events identified by Waters, Hamilton,
and Weinfield (2000) as well as extended maternal–
child separations and parental drug use. These two
events have particular relevance to this sample. Some
of the communal families featured periods of maternal–
child separation as part of their structure. Parental
drug use has been more frequent among the nonconventional
parents and does seem to have some negative
effect on child outcomes.
METHOD
Sample. The original FLS sample consisted of 205
American families representing conventional twoparent
families (n ϭ 51) and nonconventional families
(53 social contract couples, 26 families in domestic living
groups, 45 single mothers, and 30 families in creedal
communes). Families were recruited during the
mothers’ third trimester of pregnancy (see Weisner &
Garnier, 1992, and Weisner & Wilson-Mitchell, 1990,
for a full description of the sample and sampling
methodology). Through middle childhood there was
very little sample attrition, and 95% of the data on
school achievement are complete through age 12. The
adolescent follow-up of these families was primarily
based on mailed questionnaires and surveys. Limited
resources were available to conduct the in-depth interviews
required for the administration of the Adult
Attachment Interview (AAI), therefore, a subsample
of the FLS adolescents was selected. Three strategies
for subsampling were considered: randomly sampling
from the full FLS sample, sampling based on
representing the overall distribution of infant attachment,
and sampling to overrepresent adolescents
with insecure infant attachment classifications. The
third strategy was used to ensure that the less frequently
occurring insecure classifications were represented
adequately in the analyses.
Method. Families were contacted by mail as part of
the larger follow up of FLS families that was in
progress. Forty-nine adolescents within the full FLS
sample had insecure infant attachment classifications.
Twenty-nine of those adolescents were inaccessible
because of family relocation outside of the project
location. Twenty adolescents with insecure infant
attachment classifications (6 with resistant classifications
and 14 with avoidant classifications) were targeted
for inclusion in the current study. Two families
were not located. Twelve adolescents with avoidant infant
attachment classifications and 6 adolescents with
resistant infant attachment classifications were included
in the final subsample. Twelve adolescents
with secure infant attachment classifications were also
included. In contacting the adolescents with secure attachment
histories, four adolescents initially identi-
692 Child Development
fied for inclusion in the subsample were subsequently
excluded and replaced. In two of these families the
adolescent was attending college outside of the geographical
area, one family declined to participate, and
the fourth family and adolescent, although agreeable,
proved difficult to schedule. One set of fraternal
twins (both boys) was included in the subsample;
both had been classified as avoidant in infancy.2
The distribution of family lifestyles in the subsample
is disproportionate to the full sample in which
only 25% of the families were categorized as conventional.
The distribution of family lifestyles across the
subsample at birth was 17% (n ϭ 5) social contract
couples, 7% (n ϭ 2) domestic living groups, 27% (n ϭ
8) single mothers, 10% (n ϭ 2) creedal communes,
and 40% conventional. Although in the full sample
there was no association between family lifestyle
group and infant attachment classification, in the subsample
infant insecurity in the subsample was significantly
associated with conventional lifestyles.
At the adolescent follow-up, family organization
included 14 two-parent married couples (including
one lesbian couple and two stepparent families), 15
single parents (including two fathers) and one communal
group (both biological parents, now divorced,
and with new partners were present in the household).
The adolescents’ average age was 17½ years
(range ϭ 17 to 19 years).
PROCEDURES AND MEASURES
Infant attachment. Infant attachment was assessed
at 12 months by using the Strange Situation
(Ainsworth, Blehar, Waters, & Wall, 1978). In the full
FLS sample, 76% (n ϭ 153) of the infants were classified
as secure, 15% (n ϭ 30) as avoidant, and 10% (n ϭ
19) as resistant. Reliability was initially established
through pilot samples and careful and systematic
training of raters. Raters obtained 90% agreement before
data collection. Reliability was assessed periodically
throughout the data collection and differences
were resolved by consensus (see Bernstein et al., 1981,
for a full description of the 12-month procedures).
Negative life events. The negative life events were
those identified in Waters et al. (2000). Two additional
negative life events were included because of the particular
nature of this sample: prolonged physical separation
of mother and child in early childhood (greater
than 6 months), which was a feature associated particularly
with some communal groups and parental drug
2 All analyses were run separately including only one of the
twins. Results did not differ from those conducted with the full
sample.
use. Scoring of these events was based on a full review
of all case materials, which included contemporaneous
interviews with parents and adolescents about changes
in their lives, as well as summaries of the interviews
conducted with the parents(s) at child ages 12 months,
3 years, and 6 years; descriptive notes of home observations
and testing at child age 6 years; summaries of
school observations and interviews with teachers at
grades 1/2, 3/4, and 5/6; school records, and logs of
phone contacts made by project staff from birth
through adolescence. The full case history of each adolescent
was reviewed and the presence of each discrete
negative life event was noted. The presence or
absence of negative life events was unambiguous and
all coding was done by the author.
Adolescent Attachment Interview (AAI). The author
interviewed adolescents by using the adolescent version
of the AAI (George, Kaplan, & Main, 1984). All
interviews were audiotaped and later transcribed for
analysis. The author, trained by Mary Main and Erik
Hesse, scored the transcripts in accordance with the
procedures developed by Main and Goldwyn (1991).
Interrater agreement, based on a subset of the cases
(n ϭ 15) was 80% (␬ ϭ .67) across the three scoring categories,
Secure, Dismissing, and Preoccupied. Percent
agreement across the classifications ranged from
75% (Dismissing) to 100% (Preoccupied). The second
rater was trained by the author; in all cases the scores
of the more experienced rater were used.
RESULTS
Adolescent attachment. Nine adolescents were classified
as secure and 21 as insecure (13 Dismissing and
8 Preoccupied). No adolescent received a primary
Unresolved classification.
Attachment stability. Table 1 contains the crosstabulation
of infant and adolescent attachment classification.
Sixteen of the adolescents classified as insecure
in infancy were also classified as insecure in adolescence.
Seven adolescents remained secure. Overall,
Table 1 Infant and Adolescent Attachment Classifications
Adolescent
Attachment
Infant Attachment
Avoidant Secure Resistant Totals
Dismissing 8 4 1 13
Secure 1 7 1 9
Preoccupied 3 1 4 8
Totals 12 12 6 30
Note: t(4) ϭ .23, p Ͻ .01.
Hamilton 693
77% of the adolescents retained the same classification
over time (54% expected by chance, ␬ ϭ .49, p Ͻ
.01). Analysis based on three-way classifications was
similar. Sixty-three percent of the adolescents retained
the same classification. Continuity of attachment
from infancy through adolescence was assessed
by using Goodman and Kruskals ␶. ␶ is a measure of
predictive association and is robust to uneven distributions
in the dependent variable (Reynolds, 1984).
Infant attachment classification predicted adolescent
attachment classification, t(4, n ϭ 30) ϭ .23, p Ͻ .01.
Negative life events. The total number of negative
life events ranged from 0 to 4 (M ϭ 1.27). The incidence
of individual events varied considerably. Divorce
was the most frequently occurring event; fully
half of the families experienced a divorce or mate separation
sometime before adolescent follow-up. Parental
substance abuse also had a high frequency and
was present in nine families. The remaining events
were infrequent (one or two occurrences across the
sample as a whole) and no adolescent experienced
the death of a parent.
The influence of negative life events on the continuity
of attachment was assessed by using a 2 (no
event versus one or more events) ϫ 4 (stable secure,
stable insecure, secure in infancy-insecure in adolescence,
insecure in infancy-secure in adolescence) ␹2
analysis. Negative life events were associated with
the continuity of attachment, ␹2(3, n ϭ 30) ϭ 10.70,
p Ͻ .05. Negative life events seemed to operate primarily
in terms of maintaining insecure attachments.
Adolescents with stable insecure attachments were
most likely to have experienced one or more negative
life events (Table 2). In contrast, negative life events
were equally likely to have occurred in adolescents
with stable secure classifications.
Family lifestyles. There were no differences in the
distribution of adolescent security in nonconventional
and conventional families, ␹2(1, n ϭ 30) ϭ .52,
ns) or the experience of negative life events, ␹2(1, n ϭ
30) ϭ .81, ns or in the likelihood of retaining the same
attachment classification, ␹2(1, n ϭ 30) ϭ 3.14, ns.
DISCUSSION
The relevance of early experience to later development
is one of the basic questions raised in developmental
research. What, if any, continuity exists across the
lifespan and what processes influence change and
continuity? In this study the quality of the relationship
formed by infants with their mothers was meaningfully
associated with their attachment representations
in adolescence. Change in attachment classification to
some extent was related to the presence or absence of
negative life events. However, negative life events
seemed primarily to support an early trajectory identified
by insecure infant attachment.
Bowlby (1982) has suggested that the continuity of
attachment is maintained primarily through the infant’s
formation of an internal working model of self
and other that is then carried forward by the individual.
The results from the present study do not rule out
the internal working model as the process underlying
continuity but they do suggest that, at least for insecure
infants, the environment may also exert an influence.
Adolescents who retained an insecure attachment
classification were also more likely to have
experienced negative life events. Whether internal or
external forces supported this continuity is unclear.
For infants with secure attachment relationships the
picture is less clear. Half of these children did experience
negative life events, which suggests that secure
attachment may serve as a protective factor and that
these children were possibly more resilient to life
stresses. A more in-depth examination of the stresses
experienced by these adolescents suggests that the
negative life events experienced by secure adolescents
were qualitatively different and potentially less
stressful than those experienced by those who became
insecure.
For example, divorce rates were quite high in this
sample and occurred in both adolescents with insecure
and secure attachment representations. Three
families of adolescents with secure attachment representations
experienced divorce; these divorces happened
late in childhood or adolescence and were reported
by parents to be relatively free of acrimony. In
contrast, divorce in the families of adolescents who
remained or became insecure was more likely to
occur in early childhood (10 out of 13 divorces were in
early childhood) and was often accompanied by contemporaneous
reports by parents and observers of
high degrees of marital conflict (eight families reported
such conflict). Five of the mothers in these
Table 2 Association between Attachment Continuity and Negative
Life Events
Negative Life Events
Infant-adolescent
Attachment None
One or
More Total
Secure-secure 3 4 7
Secure-insecure 1 4 5
Insecure-insecure 1 15 16
Insecure-secure 2 0 2
Note: ␹2(3, n ϭ 30) ϭ 10.70, p Ͻ .05.
694 Child Development
families, and none of the mothers of secure adolescents,
reported incidents of spousal abuse. Divorce, as
well as other negative life events in the families of adolescents
with insecure representations, was likely
tied to a constellation of negative life stressors, which
in turn supported and maintained insecure parent–
child relationships. Negative life events in the families
of adolescents with secure representations were
more likely to be isolated events.
Identifying connectedness between early experience
and later development points to the important
role of attachment in development. This continuity is
not a phenomenon limited to more traditional middleclass
patterns of childrearing. Continuity of attachment
does not seem to be a factor unique to conventional
family structures. What is not clear from this
study and what should be pursued in future are the
processes that maintain continuity and stability.
ACKNOWLEDGMENTS
This study is based on a doctoral dissertation submitted
to the University of California, Los Angeles, in
partial fulfillment of the requirements for the Ph.D.
degree. The Family Lifestyles Project has been supported
by Carnegie Corporation grant B3870 and
USPHS grant MH24947 to the late Bernice T. Eiduson,
principal investigator, and by Carnegie Corporation
grant B4189 and W.T. Grant Foundation grant #92-
1488 to Thomas S. Weisner, principal investigator.
This study was funded by dissertation grants from
the Psychocultural Study Group, Department of Psychiatry
and Biobehavioral Sciences, University of California,
Los Angeles. Portions of this paper were presented
at the biennial meeting of the Society for
Research in Child Development, Indianapolis, IN,
1995. Appreciation is extended to my dissertation
committee for their support throughout this process
and their thoughtful comments: Carollee Howes
(Chair), Leila Beckwith, Nicholas Blurton-Jones, Aimee
Dorr, Philip B. Ender, and Thomas S. Weisner. I would
also like to thank FLS Project Staff, Helen Garnier, and
Maurine Bernstein for their help in contacting families
and their assistance in negotiating the complex
FLS database. Finally, I am grateful for the continued
involvement of the FLS families.
ADDRESS AND AFFILIATION
Corresponding author: Claire E. Hamilton, Elementary
Education, 427Aderhold Hall, University of Georgia,
Athens, GA 30602; e-mail: chamilt@coe.uga.edu.
REFERENCES
Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S.
(1978). Patterns of attachment: A psychological study of the
strange situation. Hillsdale, NJ: Erlbaum.
Alexander, J. (1978). Marriages without weddings: Changes
in the American family. Unpublished manuscript, Institute
for Clinical Social Work, Los Angeles.
Bernstein, M., Zimmerman, I. L., & Eiduson, B. T. (1981, August).
Attachment behaviors of infants reared in alternative
life styles. Paper presented at the annual meeting of
the American Psychological Association, Los Angeles.
Bowlby, J. (1982). Attachment and loss. New York: Basic Books.
Eiduson, B. T., Cohen, J., & Alexander, J. (1973). Alternatives
in child rearing in the 1970s. American Journal of Orthopsychiatry,
43, 720–731.
Eiduson, B. T., Kornfein, M., Zimmerman, I. L., & Weisner,
T. S. (1982). Comparative socialization practices in traditional
and alternative families. In M. E. Lamb (Ed.), Nontraditional
families: Parenting and child development (pp.
315–346). Hillsdale, NJ: Erlbaum.
Eiduson, B. T., & Weisner, T. S. (1978). Alternative family
styles: Effects on young children. In J. H. Stevens, Jr. &
M. Mathews (Eds.), Mother/Child father/child relationships
(pp. 197–221). Washington, DC: National Association
for the Education of Young Children.
Garnier, H. E., & Weisner, T. S. (1995, April). Countercultural
and nonconventional family lifestyles, family
values, and adolescent substance use. Paper presented at
the biennial meeting of the Society for Research in Child
Development, Indianapolis.
Garnier, H. E., & Weisner, T. S. (1992). Countercultural drug
use then and now: A 16-year longitudinal study of children
growing up in drug-using families. Paper presented
at the annual meeting of the American Educational
Research Association, San Francisco, CA.
George, C., Kaplan, N., & Main, M. (1984). Attachment interview
for adults. Unpublished manuscript, University of
California at Berkeley.
Lamb, M. E., Thompson, R. A., Gardner, W., & Charnov, E. L.
(1985). Infant–mother attachment. Hillsdale, NJ: Erlbaum.
Main, M., & Goldwyn, R. (1991). Adult attachment rating and
classification. Unpublished manuscript.
Reynolds, H. T. (1984). Analysis of nominal data. Newbury
Park, CA: Sage Publications.
Waters, E., Hamilton, C. E., & Weinfeld, N. S. (2000). The
stability of attachment security from infancy to adolescence
and early adulthood: General introduction. Child
Development, 71, 678–683.
Weisner, T. S. (1986). Implementing new relationship styles
in conventional and nonconventional American families.
In W. Hartup & Z. Rubin (Eds.), Relationships and development
(pp. 185-206). Hillsdale, NJ: LEA Press.
Weisner, T. S., & Garnier, H. (1992). Nonconventional family
lifestyles and school achievement. American Educational
Research Journal, 29(3), 605–632.
Weisner, T. S., & Wilson-Mitchell, J. E. (1990). Nonconventional
family life-styles and sex typing in six-year-olds.
Child Development, 61(6), 1915–1933.