Childhood Adversity, Attachment

Security, and Adult Relationships:

A Preliminary Study
Elizabeth M. Hill, James P. Young, and Janet L. Nord
The University of Michigan Alcohol Research Center, Department of Psychiatry,
University of Michigan, Ann Arbor
Several evolutionary theorists have linked early rearing context to later reproductive
strategy, hypothesizing that strategies differentiate during development as functional
responses to ecological characteristics, by individuals or through parental manipula-
tion. Attachment security has been proposed as a mediator. In this study, 40 young
adults were given a multidimensional assessment, including the Hazen and Shaver
Adult Attachment Questionnaire. Twenty-four subjects were classified as having se-
cure attachment styles, 16 as nonsecure. The magnitude and predictability of parental
investment during childhood was classified as lower if there was a brief intersibling
interval, parental divorce, fewer economic resources, or less nurturing parents (i.e.,
more childhood adversity). Several such indicators were present for 17 people, 12 of
whom were nonsecure, compared to only 4 of the 23 others. The nonsecurely attached
subjects were less likely to have attained enduring marriages. The 6/16 nonsecure who
had a marriage or cohabitation began them at a younger age and after a shorter
courtship period than did the 1924 secure with such relationships. Separations or
divorces had already occurred in the relationships of 4/6 nonsecure versus S/15 secure.
Attachment security was associated with childhood adversity and adult relationships
for both men and women, when analyzed separately. A retrospective study cannot
address cause and effect, because poor adult relationship outcomes might bias recall of
parental behavior. However, results are consistent with theories that unpredictable
early environments foster short-term rather than long-term mating strategies, possibly
through affecting attachment styles.
KEY WORDS: Attachment; Reproductive strategies; Parental investment; Childhood re-
source adversity; Relationships
INTRODUCTION
Evolutionary theory predicts that life history strategies will be related to
environmental conditions (Emlen and Oring 1977; Krebs and Davies 1984).
Received June 15, 1994; revised September 17, 1994.
Address reprint requests and correspondence to: Elizabeth M. Hill, Ph.D., The University of
Michigan Alcohol Research Center, Department of Psychiatry, 400 E. Eisenhower, Ann Arbor,
MI 48104 U.S.A.
Ethology and Sociobiology 15: 323-338 (1994)
0 Elsevier Science Inc., 1994
655 Avenue of the Americas, New York, NY 10010
0162-3095/94/$7.00
324 E. M. Hill et al.
The components of reproductive life histories that may vary include the
amount of parental effort expended, sex differences in parental effort, and
distribution of investment between own reproduction and helping ones kin.
The optimal timing of mating and parental effort during the life cycle is a key
variable that differentiates reproductive life-history strategies (Hirschfield and
Tinkle 1975; Williams 1966). Variability in behavior, both between and within
populations, may be affected by environmental conditions; the spatial and
temporal distribution of resources affects which reproductive strategies are
most likely to be successful (Emlen and Oring). For example, long-term
cooperation of a mating pair in rearing offspring may be advantageous for both
parents when environmental resources are uniformly distributed and defensi-
ble; however, resources sufficient to support a pair cannot be defended if
distributed too sparsely and widely (Emlen and Oring). Using a life-history
perspective, Borgerhoff Mulder (1992) has recently reviewed factors that affect
costs and benefits of several parental care-allocation schemes for humans.
Several evolutionary theorists have hypothesized that several distinctive
patterns of reproductive investment exist in humans (Dickemann 1986; Belsky
et al. 199 1; Draper and Belsky 1990; Draper and Harpending 198 1; Lancaster
and Lancaster 1987). These authors have proposed that early rearing context
affects later reproductive strategy, such that individuals respond functionally to
ecological characteristics of their developmental environment (recently articu-
lated by Belsky et al.; Dickemann; Draper and Belsky). These authors propose
that stressful, unpredictable environments set people on a course to reproduce
early and to have parental or child-rearing relationships that are only tran-
siently cooperative (Belsky et al.; Draper and Belsky). These arguments are
consistent with the work of Surbey (1990), which documented early menarche
by young women from homes with absent fathers.
Studies in the ethology and behavioral ecology of animal life histories
support the notion that environmental unpredictability affects reproductive
strategies. When the environment is unpredictable, ones future chances for
reproducing are also unpredictable; the best strategy may be to reproduce when
some resources become available (Hill and Low 1992; Rubenstein 1982). In
these situations, it is not advantageous to save resources for future use or invest
them in ones own further growth and development (Rubenstein). When there
are high adult failure rates for various reasons (mortality, competitive failure),
an alternate strategy of opportunistic breeding when young may be best
(Dickemann 1986). Burton (1990) interpreted the pattern seen commonly in
poor neighborhoods-of early childbearing with grandmaternal care of chil-
dren-as a functional response to low survival rates in an environment where
traditional patterns are not feasible. In this alternative life-course strategy, there
are additional elements of separation of marriage from reproduction, and
childbearing by only some siblings in a generation, with others working and
contributing resources (Burton).
Draper, Belsky, and colleagues (Belsky et al. 1991; Draper and Belsky
1990) proposed a mediating role for attachment security in the relationship
Adversity, Attachment, and Relationships 325
between early environment and ones own reproduction, while Dickemann
(1986) described mediation through multiple gender identities. Draper and
Belsky proposed that the internal working model of relationships that
children develop from early interactions with caretakers affects their later
expectations that other relationships will be rewarding and enduring, in partic-
ular, whether a marital partner can be counted on over the long haul (Draper
and Belsky, p. 152). Empirical research has documented some of these associ-
ations. Insecure or mistrustful internal models of relationships are more proba-
ble when behavior of caretakers has been harsh or unpredictable; in turn, such
parental behavior is more likely when conditions are stressful (Ainsworth et al.
1978; Bowlby 1969; Sroufe 1979).
Chisholm (1993) has presented a similar conceptual model of the relation-
ship between local mortality rate, attachment style, and short- versus long-term
mating patterns. He elaborated on the role of mortality rate, using proposals of
Promislow and Harvey (1990, 1991) that life-history traits are affected by the
intensity and age-specificity of mortality rates. In this model, high mortality
produces proximate stress for parents, which results in emotional stress for
children. In turn, early psychological stress affects attachment patterns, predis-
posing one toward a reproductive strategy of lower parental investment (psy-
chologically, a less involved style of parenting; Chisholm 1993, p. 10). The
concept of early stress allows incorporation of existing psychological litera-
ture, but it is vague. The interconnections in this model would be clearer if the
amount and predictability of parental investment and other aspects of the
resource environment are used instead of early stress. Chisholm also as-
sumes that children do not directly sense mortality rates, while they certainly
do witness violence and accidents, and are aware of the deaths of significant
individuals or neighbors (Be11 and Jenkins 1993).
Our model of the interrelationships among early environment, attachment
security, and adult relationships is shown in Figure 1. The salient dimensions
of early environment are the predictability and amount of parental investment.
Adult relationships are expected to vary by timing and duration. Measures are
discussed in detail below. Early childhood experiences are hypothesized to
correlate with attachment security, which influences adult relationships. This
model has similarities to the one published by Draper and Belsky (1990),
which was further refined by Belsky et al. (1991). Parent-child relationships
figured in their model, which emphasized father absence. Their reproductive
measures included timing of puberty, stability of adult-pair bonds, degree of
parental investment in children, and r-selected versus k-selected life-history
strategies. In refinements, discussion of r- and k-strategies have been elimi-
nated, and Belsky et al. focused instead on variation along a continuum of
parental investment of quality versus quantity. Our model, rather, focuses on
the effect of early environmental unpredictability, whatever the source, on
timing of first attempt to form marital reproductive relationships.
During a study comparing 20 young adults who each had an alcoholic
parent to 20 young adults with social-drinking parents, subjects were given
326 E. M. Hill et al.
EARLY ENVIRONMENT 0 ATTACHMENT 0 ADULT
) STYLE
RELATIONSHIPS ~
(Predictability and amount (Security) (Short vs long-term)
of parental investment)
- SE.5 of family
. occupation of mother and father
* perceived poverty
_ Resarce of caretakers
* divorce I separation
- Nurturing behavior of pa.~.nts
-positive relationship
-abusive punishment
- Within family variation
- interbirth interval with siblings
. favoritism by parents
* secwe / itLsccluc
-discomfort with closmess
- discamfott with insufftcient
closeness
- comfort with closeness
. attainment of marriage I
cohabitating relationship
* age at first relationship
* duration of courtship
- intemtpted relationships
*marital social adjustment
- friendship social adjusmunl
J
FIGURE 1. Conceptual model of the relationships among attachment security, early envi-
ronment, and adult relationships.
additional assessments to address some aspects of our model of the interrela-
tionships of early rearing environment, attachment security, and reproductive
behavior as a young adult. Originally, we expected that children of an alcohol-
dependent parent would have less secure attachment, and that attachment
security would correlate with the participants drinking problems. Neither
hypothesis was supported; so groups were collapsed for the analysis presented
here. This paper contrasts those subjects with secure attachment (n = 24) to
those nonsecurely attached (n = 16), showing the relationship of adult attach-
ment security to early childhood adversity and to some adult reproductive
measures.
METHODS
Participants
Young-adult subjects were 20 men and 20 women, ranging in age from 18-30
(X = 24.8 + 3.8). Most completed high school and were currently employed
with incomes in the $lO,OOO-14,999 or $15,000-19,999 range. All families
were Caucasian, except one. The original groups of young adults and their
parents have been described in detail elsewhere (Hill et al. 1992). Parents were
in their mid-fifties, and most had more than one marriage. Demographic
differences between alcohol-dependent and social-drinker parents were negli-
gible, except that alcohol-dependent parents tended to be less educated. These
Adversity, Attachment, and Relationships 327
two groups were pooled for the analyses reported here, since alcohol-depen-
dence in a parent did not affect the subjects attachment security, as noted
above.
Procedure
Participants were recruited by contacting eligible parents who were already
participating in research studies at the University of Michigan Alcohol Re-
search Center (UMARC). UMARC participants are recruited by a variety of
means, including placing advertisements in local newspapers, interviewing
patients in treatment at several local alcoholism treatment programs, and direct
mailings to various groups, such as university alumni or service organizations.
Eligible subjects from this pool were asked to consent to contact children and
to provide telephone numbers. The consent rate of the 45 parents contacted
was 89% (n = 40). We recruited 20 parents who were alcohol dependent and
20 who were social drinkers, as determined by structured diagnostic interview.
When a family had more than one child as a potential participant, the one
closest to age 25 was recruited; if they were equidistant to age 25, one was
chosen randomly. The participants were informed that the purpose of the study
was to examine alcohol use and childhood experiences in a variety of families.
The consent rate was 87% for the 46 young-adult children (20/26 children of
alcohol-dependent parents, 20/20 of social-drinking parents). The young adult
subjects received $25 for completion of an interview and self-report assess-
ments. Interviewers were blind to the parental alcohol-dependence status.
Measures
Participants were given a multidimensional assessment that included an inter-
view for retrospective family, economic, and reproductive life-history mea-
sures, and standardized self-report scales for attachment and social adjustment.
Constructs measured were listed in Figure 1.
Early environment. Nurturing behavior of parents during early life (age
6-12) was assessed using the Home Environment Interview (HEI, Robins et al.
1985). The respondent portions were used for the present study, with the
questions referring to siblings deleted. Another study of children of alcoholic
parents employed a similar instrument (Reich et al. 1988). The present study
used the original adult retrospective version. To create indices, sums were
calculated of positive responses to yes/no questions for both mother and father
separately, then summed for a parental total.
Positive Relationships between parent and child were measured by 13
items, 6 asking whether parents often spent time with the subject working
around the house, playing, in outings, asking about activities, and celebrating
holidays or going visiting, and other items on attending school functions,
showing affection, and being easy to talk to. The subjects were also asked
328 E. M. Hill et al.
about whether parents frequently criticized the child or ever said the child was
not loved, or that he or she should not have been born; scores for these items
were reversed for the positive relationship total. Abusive Punishment by
parents included being (1) hit with a belt or stick, (2) locked in a room for
more than an hour, or (3) thrown out of the house for over an hour, or if the
respondent (4) feared serious harm during parental punishment, or (5) was
punished in front of nonfamily members. Higher scores reflect more abusive
punishment.
Within-family variation in parental investment was represented by (1) one
item from the HEI, parental discrimination (whether a parent was particularly
hard on you, compared to other siblings), and (2) by birth interval between
oneself and other siblings. Two years or less was considered a short interbirth
interval. Other parental investment indicators related to economic resources.
Parental occupations were given socioeconomic status ratings according to
Featherman and Stevens index (1982). Using this scale, an electrician is coded
30, a lawyer 87. In addition, perceived poverty during youth was assessed by
asking about family resources (potential responses ranged from, We were
poor and had a hard time paying for food and housing to We were well off
or wealthy; adapted from Binion, 1982). Participants also noted whether their
parents had separated or divorced during childhood (before age 12).
Attachment security. Participants completed the Hazen and Shaver Adult
Attachment Questionnaire (Hazen and Shaver 1987, 1990). This 13-item
instrument gives scores, based on factor analysis, on three subscales that reflect
these types: comfort with closeness, discomfort with closeness, and discomfort
with insuflcient closeness. In addition, respondents classified themselves on a
three-response, forced-choice item that asks which of three statements (corre-
sponding to avoidant, ambivalent, and secure-attachment types) best describes
them. This item has a test-retest concordance of 89% (personal communica-
tion). Of the 40 subjects in this study, all but four were classified the same type
by using the highest subscale or with the forced-choice item; these four were
classified according to the highest subscale. There were 16 participants with
nonsecure-attachment types (8 avoidant [4 men, 4 women], 5 ambivalent [2
men, 4 women], and 2 who had equal scores on avoidant and ambivalent
subscales). The internal consistency of the scale items with the secure/nonse-
cure classification is shown in Table 1. Most items differentiated between
nonsecure and secure groups. The lack of significance of a few items was due
to obscured differences when avoidant and ambivalent types were collapsed
(e.g.,
merge completely with another person was answered positively by
most of those with ambivalent but only one of those with avoidant attachment
types).
Adult relationships. The third set of measures were of adult relationships
that reflect an enduring, long-term, high-investment reproductive strategy:
attainment and maintenance of a marriage or cohabitation relationship (of at
Adversity, Attachment, and Relationships 329
Table 1. Internal Consistency of Adult Attachment Scale: Percentage Agreeing or Strongly
Agreeing with Statements
Statement Insecure Secure p-value
I am somewhat uncomfortable
being close to others.
I find it to trust others. easy
I find it difficult to depend on others.
I am nervous when anyone gets
too close.
Love partners often want me to be more
intimate than I feel comfortable being.
1 find that others are reluctant to get as
close as I would like.
1 often worry that my panner doesnt
really love me.
I want to merge completely with
another person.
I lind it relatively easy to get close
to others.
I feel comfortable depending on
other people.
I feel comfortable having other people
depend on me.
1 dont often about being abandoned. worry
I dont often worry about someone getting
too close to me.
58.25 16.67 ,035
43.75 83.33 ,062
68.75 17.39 ,004
56.25 08.33 ,006
31.25 16.67 .215
43.75 16.67 ,215
50.00 04.17 ,038
37.50 45.83 ,413
50.00 83.33 ,038
31.25 75.00 ,032
43.75 95.83 .002
56.25 91.67 ,012
56.25 95.83 .OlO
least a years duration), older age at marriage, and longer courtship length.
Interrupted relationships were defined as those with divorces or separations. Only
eight participants had children as yet, precluding analysis of that measure.
The Social Adjustment Scale (Weissman and Bothwell, 1975) was also
used, which is a self-report measure of functioning in various role areas: work,
marital, social and leisure (friends), parental, family, and economic. Scale
scores are averages of frequency-type items on a scale of l-5. A general
social-adjustment score is an average of all items. The scales that relate to
functioning in relationships are the marital and friendship (social/leisure)
relationship scales. For example, marital-relationship questions ask about the
frequency of arguments, talking about feelings or problems, being demanding,
being bossed around, feeling dependent, feeling affection, and having and
enjoying intercourse. Friendship questions ask about the frequency in the past
two weeks of talking with friends on the telephone, talking about problems or
feelings with friends, going out socially, having arguments with friends, having
ones feelings hurt, feeling shy or uncomfortable with people, feeling lonely,
and so forth. Higher scores reflect poorer social adjustment.
Data Analysis
An index of multiple risks was created by summing the number of relatively
independent early environmental risks experienced. Parental occupation and
330 E. M. Hill et al.
perceived poverty were significantly related, as were positive relationship with
mother and abusive punishment (see below). Therefore, five risks were cumu-
lated into an index of multiple risks: parental occupation below the group
median (58.5) positive relationship with mother below the median (1 1.5),
intersibling birth interval of two years or less, separation or divorce of parents,
and discrimination by a parent. A comparable cumulative measure of adult-
relationship pattern was not created; several measures apply only to persons
with marital or cohabiting relationships. However, one measure was considered
representative of the construct and applicable to all subjects: friendship adjust-
ment (SASR social/leisure subscale).
Data were analyzed by calculating associations among the three sets of
variables. A variety of tests were used, based on the distribution of the
measures, including Spearmans correlations, Students I tests, Wilcoxons
rank sum tests and chi-square tests (likelihood ratio chi-square). Because the
Social Adjustment Scale gives lower scores for better adjustment, the signs
of correlations with this measure have been reversed to facilitate interpretation.
Since specific associations were predicted between attachment security and
early environment and adult relationships, no corrections for multiple tests
were made. However, when associations within constructs were analyzed, such
as interrelationships of early environment measures, the Bonferroni correction
was used. Tables of results show either the mean and standard deviation or the
median and interquartile range, depending on the statistical test employed.
Since the sample size was restricted, only one multivariate analysis was
conducted (linear regression). The continuous measure of friendship adjust-
ment was tested for an independent association with attachment security
beyond its association with the multiple risk measure.
RESULTS
Associations of Early Environment with Attachment Security
Attachment security was affected by the early environmental indicators of
predictability and amount of parental investment during childhood (Table 2).
Individually, these indicators had small effects that were in the predicted
Table 2. Associations Between Attachment Securitv and Earlv Parental Investment
Measure
Occupation of mother and father
Perceived poverty
Caretakers divorced or separated
Positive relationship with mother
Abusive punishment
Short interbirth interval
Parents were especially hard
Insecure Secure
(n = 16) (n = 24)
55.20 2 23.15 56.13 2 18.56
56.25% (906) 33.33% (8.24)
31.25% (506) 12.50% (3/24)
9.12 + 2.76 11.50 -t 1.84
1.31 2 2.12 1.13 ? 1.78
56.25% (9/16) 37.50% (9/24)
31.25% (5/16) 08.33% (2/24)
p-value (stat)
.S. (t)
,151 (x2)
ns. (x2)
,003 (t)
ns. (t)
n.s. (x2)
.063 (~2)
Two risks
Three risks
75.00% (12/16) 45.83% (I l/24) ,068 (x2)
50.00% (806) 8.33% (2/24) ,003 (x2)
Adversity, Attachment, and Relationships 331
Table 3. Associations Between Attachment Security and Adult Relationships
Measure
Insecure Secure
(n = 16) (n = 24) p-value (stat)
Attained married/cohabiting relationship
Age at first relationship
Duration of courtship
Interrupted relationships
Social adjustment:
Marital
Friendship
Overall
37.5% (6/16) 62.5% (15/24) n.s. (x2)
21 (19-22) 23 (21-25) ,050 (Wz)
6.5 (5-12) 24 (12-36) .021 (Wz)
66.7% (4/6) 33.3% (5/15) n.s. (x2)
2.55 k 1.10 1.59 2 0.33 ,012 (t)
2.20 2 0.66 1.72 + 0.30 ,014 (t)
1.96 2 0.43 1.65 k 0.20 ,003 (t)
direction. Trends were present for nonsecurely attached subjects to have a
perception of a poorer early lifestyle, a worse relationship with mother during
years 6-12, and to feel a parent was particularly hard on me, of all the
children. Of the seven who reported that a parent was especially hard on
them, only two showed a secure attachment style.
Environmental risk measures were relatively independent, except that
parents occupations and perceived poverty were significantly related; the
average prestige score was 47.5< 2 18.7 for those who perceived themselves to
be poor, compared to 61.9 ? 19.5 for those who were comfortable (t(38) =
-2.35, p = .024). However, the two economic measures were not significantly
related to other investment measures. Perceived parental discrimination ap-
peared to be associated with parental occupational prestige; participants who
felt that parents were especially hard on them reported lower prestige scores
(38.7 t 19.4) than participants who did not so feel (61.8 +- -13.1; t(17) =
3.11, p = .031), but the small sample of only seven who felt unfavored (among
the 18 who felt parents showed discrimination) limits the reliability of this
finding. Participants with a short intersibling interval tended to experience
mote abusive punishment by parents (1.89 + -2.42) than participants with
longer intersibling intervals (0.64 + 1.09; t(38) = -2.03, p = .054). Having
experienced parental separation or divorce was not significantly related to
other measures.
The multiple risk index also differed for the secure and nonsecure groups.
When several risks were present, the effect on attachment security was strong
(Table 2). Two risks were present for 23 people, 12 of whom were nonsecure,
compared to only 4 nonsecure of the 17 others. Three or more risks were
present for 10 people, 8 of whom were nonsecure, compared to 8 nonsecure of
the 30 without three or more risks.
Attachment Security and Adult Relationships
Attachment security was clearly related to current relationships (Table 3). The
nonsecurely attached subjects were somewhat less likely to have attained
enduring marriages. Only 6 of the 16 nonsecure had married or cohabited 3 of
these were marriages); 15 of the 24 securely attached participants had cohabi-
tation or marital relationships (11 were married). The 6116 nonsecure who had
332 E. M. Hill et al.
a marriage or cohabitation began them at a younger age and after a shorter
dating period than did the 15/24 secure with such relationships. Separations or
divorces had already occurred in the relationships of 4/6 nonsecure versus 5/15
secure. Only 8 of the young adults had children, 6/24 of the secure group (4/13
men, 2/l 1 women) and 2/16 nonsecure (O/7 men, 2/9 women). The nonsecurely
attached subjects reported worse social adjustment both with friends (social
and leisure activities) and with marital partners (based on the 6 nonsecurely
attached who were married).
Measures of adult relationships were not independent. Marital and social
adjustment had a Spearman correlation of .75 (p = .0005). The older the age at
first marriage or cohabitation, the longer was courtship duration (rs = .39, p =
.OS). Those whose relationships were already terminated had begun them at
about age 21, two years younger than those whose relationships were still
intact (20.89 ? 1.96 vs 23.08 2 2.39; t(19) = 2.306, p = .037). General social
adjustment scores were better for those with an older age at first relationship
(rs = .63, p = .002) and a longer dating period (r, = .48, p = .027); it was not
associated with having interrupted marriages, however.
Sex Differences
Since men and women might differ on average in attachment security, relation-
ship patterns, and environmental predictors, associations of interest were
examined within sexes. First, men and women were compared on demographic
measures and attachment security. Male subjects did have about one more year
of education on average (2 = 15.00 -+ 1.85 vs 13.70 2 1.90, t(38) = 2.19, p =
.035), and they were generally in a higher income category (S;- = 8.40 -+ 2.70 vs
6.30 2 2.55, r(38) = 2.52, p = .016; $15,000-19,00O/year vs $7,000-9,999).
Male and female subjects did not differ on number of marriages. The three
subscales of the Hazen and Shaver Adult Attachment Scale did not differ by
sex (all ps > .40; Discomfort with closeness: X = 10.25 2 3.13 Men vs 9.40 2
4.07 Women; Comfort with closeness X = 10.95 & 1.82 Men vs 10.70 % 2.87
Women; and Discomfort with insufficient closeness X = 8.20 t 1.47 Men vs
8.60 2 2.26 Women).
The nonsecure subgroup (n = 16) contained 7 men and 9 women, and the
secure subgroup (n = 24) contained 13 men and I1 women. Early parental
investment measures were lower for nonsecure subjects within sexes. The
effect of early risks on womens attachment was apparent for those with three
or more risks (nonsecure: 5/9; secure: l/l 1, x2 = 5.09, p = .024), but was not
statistically significant with two or more risks (nonsecure: 7/9; secure: 5/l 1, x2
= 2.16, p = 0.142). Most of nonsecure men had experienced two or more risks
(5/7), as did half of the secure group (6/13; x2 = 1.17, p = 0.279). One of the
secure men had three or more risks (l/13), compared to 3/7 of the nonsecure
group (x2 = 3.52, p = .061).
The association of adult relationships with attachment security existed for
both men and women, but varied by relationship measure used. Only 1 of the 7
Adversity, Attachment, and Relationships 333
nonsecure men, and 5 of the 9 nonsecure women, had a marriage or cohabita-
tion relationship (compared to 8/13 secure men, x*(l) = 4.11, p = .043; and
7/l 1 secure women, x2( 1) = 1.35, p = ,714). The 5 nonsecure women had
married at a younger age (median = 21 vs 24, z = -1.717, p = .090), after a
shorter courtship (median months = 6 vs 30, z = -1.959, p = .050). The
differences in social adjustment measures by attachment security within each
sex mirrored the differences between the overall secure and nonsecure groups.
With the small sample size, however, attachment-security group differences in
friendship social adjustment and overall adjustment were significantly only for
women (Friendship X = 2.35 -+ 0.70 Nonsecure vs 1.80 t 0.38 Secure, t(l8) =
2.25, p = .037; Overall X = 2.05 ? 0.35 Nonsecure vs 1.68 ? 0.24 Secure,
r(18) = 2.78, p = .012), not for men (Friendship X = 2.00 + 0.59 Nonsecure vs
1.66 t 0.22 Secure, unequal variance t(6.9) = 1.49, p = .180; Overall Sz = 1.86
? 0.53 Nonsecure vs 1.62 t 0.15 Secure, unequal variance t(6.5) = 1.18, p =
,281).
In summary, the effect of attachment security on adult relationships was
shown for men in the probability of having a relationship of at least one year.
For women, the association occurred for age at the first relationship, duration
of courtship, and social adjustment.
Early Environment and Adult Relationships
From these analyses, attachment security appeared to be an important mediator
of the effects of early environment on adult relationships. However, early
environment might have direct independent associations with adult relation-
ships, regardless of attachment. Therefore, we examined the univariate rela-
tionships of the early environment measures to adult relationships and also
conducted several regression analyses involving both attachment security and
early environmental risk. Most of the univariate relationships were not as
strong as the associations between the measures of adult relationships and the
measure of attachment security; attachment security seems to summarize well
the relevant other variables. None of the interrelationships examined (of 42
associations or correlations) reached a nominal significance level of p < .05.
Some trends occurred, however, that may warrant future investigation. Age at
first relationship tended to be older when parental occupational prestige was
higher (rs = .396, p = .075); friendship social adjustment tended to be better
when ones relationship with mother had been positive (rs = .295, p = .065).
For those who had married or cohabitated, the frequency of relationship
termination tended to be higher when intersibling interval was shorter (short
IBI, 5/7 terminated vs longer IBI, 4/10 terminated, x*(l) = 3.56, p = ,059) and
when respondents felt parents were particularly hard on them (parents hard, 4/4
terminated vs parents not hard, 2/7 terminated, F.E.T. p = .061).
The multiple risk summary measure did show univariate associations with
adult relationship measures. For example, only 1 of 10 participants with three
or more risks had an enduring first marriage or cohabitation (no divorce or
334 E. M. Hill et al.
separation), compared to 1 l/30 with fewer risks (x2(1) = 2.938, p = .087).
Friendship adjustment had a correlation of .412 with multiple risks 0) = .008).
However, when the effect of attachment security was modeled simultaneously,
the risk correlation became nonsignificant. Attachment security had a signifi-
cant independent association with friendship adjustment, after accounting for
the effects of multiple risks (F(1,37) = 5.64, p = ,023; with both in the
equation: F(2,37) = 5.89, p = .006; R* = .24). Overall, attachment security was
a useful construct for linking rearing environment with subsequent relationship
functioning.
DISCUSSION
Results of this study support an association among early environment, adult
attachment security, and adult relationships. Adults with nonsecure attachment
models were much more likely to have experienced several indicators of
unpredictable or scarce resources, compared to secure adults. This study
assessed childhood resource adversity primarily in terms of parental invest-
ment. The magnitude and predictability of parental investment during child-
hood was classified as lower if there was a brief intersibling interval, parental
divorce, fewer economic resources, or less nurturing parents. Several such
indicators were present for 17 people, 12 of whom were nonsecure, compared
to only 4 of the 23 in the lower risk group. Attachment security was associated
with the likelihood of attaining a marital relationship after a lengthy courtship.
The nonsecurely attached participants were less likely to have attained endur-
ing marriages. The 6/16 nonsecure who had a marriage or cohabitation began
them at a younger age and after a shorter courtship period than did the 15/24
secure with such relationships. Separations or divorces had already occurred in
the relationships of 4/6 nonsecure versus 5/15 secure. Results are consistent
with theories that unpredictable early environments foster younger, shorter-
term (rather than later, long-term) mating strategies (or in refraining from
forming a marital relationship), possibly through affecting attachment styles
(Dickemann 1986; Draper and Belsky 1990).
Attachment security was associated with parental investment measures
and adult relationships for men and women, when analyzed separately. Men
and women did, however, show somewhat different responses to early envi-
ronmental factors. For men, the effect of a nonsecure attachment model on
adult relationships was manifested in a lower probability of attaining a lasting
cohabitation or marital relationship. Women with insecure attachment often
had relationships, but the age at first relationship was younger and the duration
of courtship was shorter for them than for women with secure attachment
models, This sex difference may reflect a propensity of men to have relation-
ships that do not involve cohabitation. The instruments used in this study
recorded relationships that involved cohabitation or mutual children, not all
relationships. More probably this difference reflects the greater difficulty for
men than women to obtain a mate, in general. The evolutionary basis of such a
Adversity, Attachment, and Relationships 335
sex difference has been proposed as the sex difference in requisite parental
investment (Trivers 1972). Studies of nonindustrial societies do show that
mens reproductive lives are more variable than womens, in terms of reproduc-
tive success (reviewed in Daly and Wilson 1983).
In our model, the salient dimensions of early environment are the predic-
tability and amount of parental investment. These factors are expected to affect
the timing of first reproduction attempts, a focus which is somewhat different
from that of Draper, Belsky and colleagues (Belsky et al. 199 1; Draper and
Belsky 1990). Instead they focused on variation along a parental investment
continuum of quality versus quantity, which is a model used in many
comparative studies in animal behavior (reviewed in Borgerhoff Mulder,
1992). This focus was criticized by Maccoby (1991). There is little evidence
for a positive correlation of an early onset of reproduction with final family
size, or on the eventual number of offspring of those who begin life with
various levels of environmental unpredictability. The correlation between age
at onset of reproduction and final family size is unclear, particularly because it
may be changing over time (reviewed in Furstenberg et al. 1987, pp. 34-40).
When measured in their early thirties, early childbearers had about S-1.0 more
children than later childbearers, according to several national surveys (re-
viewed in Furstenberg et al.). However, later childbearers may continue longer
and catch up. Early childbearers show a rapid tempo of family building, but
appear to terminate their reproductive careers earlier, particularly in geographi-
cal areas where abortion and sterilization are available (Furstenberg et al.). If
these patterns are common, early reproduction may not lead to more reproduc-
tion, but primarily may represent a shift in timing.
For humans, the adaptive problem might thus be whether the present is a
propitous time for parental investment in a child versus other alternatives for
inclusive fitness (Hill and Low, 1992). Those alternatives are investing somatic
effort in oneself now to prepare for better reproductive opportunities later, or
investing in reproduction of kin. When resources are scarce but predictable, the
best strategy would be to save resources for a better time, and to reproduce
when a threshold level is reached. Scarcer resources might result in reproduc-
tive shutdown, as Maccoby (1991) argued, when the best strategy is to divert
resources to other kin and have no children of ones own. In the present study,
many of those with nonsecure attachment styles did not form relationships or
have children. Only 6/16 nonsecure yet had a cohabitation or marital relation-
ship lasting at least a year, compared to 15/24 securely attached persons. When
resources are unpredictable, however, one cannot expect that the future will be
better. If a threshold level of resources happened to become available, ones
best bet would be to reproduce (Dickemann 1986; Rubenstein 1982). These
predictions follow the logic of bet-hedging theory, which tracks reproductive
success by the number of generations that a lineage survives; having fewer
offspring may enhance long-term reproductive success when their probability
of successfully reproducing is increased by investment (Hirschfield and Tinkle
1975; Williams 1966; Wittenberger 1981).
336 E. M. Hill et al.
The interrelationships of resource distribution, parental effort, and invest-
ment in specific offspring are complex. Comparative studies consistently show
that parental investment decreases as resource availability decreases (reviewed
in Clutton-Brock, 1991). However, age at which offspring become independent
and investment is terminated may not be the best measure of parental invest-
ment; duration of investment may be brief in environments with abundant
resources while magnitude of investment remains high. Lee et al. (1991)
proposed a U-shaped relationship between resource availability (maternal
condition) and termination of investment (weaning age), based on analyses of
intraspecific variation among large-bodied mammals. Longer duration of in-
vestment would be associated with mild food limitation, whereas scarce or
abundant resources could be associated with briefer investment periods and
earlier weaning. Clearly, human life history theory must be further elaborated
to develop more specific predictions about the joint effects of resource distri-
bution and the age-specificity of mortality.
A retrospective study cannot address cause and effect, since poor adult
relationship outcomes might bias recall of parental behavior. The role of
attachment security as a psychological mechanism mediating behavior is
unclear from a correlational retrospective study such as this. In the conceptual
model, early environment affects attachment security, which is a stable charac-
teristic that is resistant to later change. However, the environment may change;
an ideal organism might show total flexibility in relationship patterns in order
to respond to current conditions. However, learning during sensitive periods,
rather than completely flexible learning, may be optimal under some condi-
tions, depending on the pattern of environmental change over time (Chisholm
1993). In the present study, attachment security seemed to be a good proxy for
measuring success in relationships. Few of the nonsecurely attached people
had serious relationships of over a years duration, and those who did had
begun them at a younger age after a shorter dating period. The security
measure was more associated with the relationship variables than the early
environment measures alone.
Future research could beneficially address the roles of specific aspects of
childhood adversity and environmental risks in relationship formation.
Chisholm (1993) proposed that high mortality is translated into proximate
stress and less-involved parenting. Attachment theory could predict relation-
ship dysfunction from less-involved parenting and other childhood adversity
without the use of evolutionary theory (Belsky, personal communication).
However, we used life-history theory to create a novel measure of early
environmental resources in terms of parental investment, including the inter-
birth interval, number of sibs, and perceived parental discrimination. Concep-
tualizing resources and parental nurturance as parental investment allows
application of life-history theory to understanding the sequelae of childhood
adversity. In addition, it is important to note that the accumulation of risks is
critical. Thus, losses or inadequacies in one area can be ameliorated by
strengths in another area. Specific risks may not be particularly salient alone
Adversity, Attachment, and Relationships 337
(even father absence), but they usually signal deficiencies in several other areas
(such as low resources and less nurturance).
In addition, differences in adversity for children in the same families have
not been sufficiently explored. Increasingly, research is examining within-
family variation and effects of nonshared environment. In some cases, behavior
patterns may appear to be a product of family-level selection, when one
individuals behavior is seen as the manipulated result of a parents reproduc-
tive strategy. These family-level concepts have been applied to human repro-
ductive strategies by Dickemann (1986). She proposed that in unpredictable or
diverse environments, families as units may differentiate children into highly
divergent strategies to hedge their reproductive bets. Future research could test
expectations derived from this perspective, particularly concerning unpredict-
ability and sibling differences; for example, more parental discrimination and
more variation among children may occur in environments that are highly
unpredictable.
This study was supported in part by Center Grant No. P50 AA07378 from the National Institute of
Alcoholism and Alcohol Abuse. The authors thank Kathleen Singer, Virginia Newton, and Ana Iacob
for assistance in data collection and management, and Patricia Draper, Jay Belsky, James Chisholm,
and Lisa Thomson Ross for their critical comments. Findings were presented at the Human Behavior
and Evolution Society Meeting, August, 1991.
REFERENCES
Ainsworth, M., Blehar, M., Waters, E., and Wall, S. Patterns of Attachment, Hillsdale, NJ:
Erlbaum, 1978.
Bell, C.C., and Jenkins, E.J. Community violence and children on Chicagos Southside. Psychia-
try 56: 46-54, 1993.
Belsky, J., Steinberg, L., and Draper, P. Childhood experience, interpersonal development, and
reproductive strategy: an evolutionary theory of socialization. Child Development 62:
647-670, 1991.
Binion, V.J. Sex differences in socialization and family dynamics of female and male heroin
users. Journal of Social Issues 38: 43-57, 1982.
Bowlby, J. Attachment and Loss: Vol. I.Attachment, New York: Basic Books, 1969.
Burgerhoff Mulder, M. Reproductive decisions. In Evolutionary Ecology and Human Behavior,
E.A. Smith and B. Winterhalder (Eds.). New York: Aldine, 1992, pp. 339-374.
Burton, L. Teenage childbearing as an alternative life-course strategy in multigeneration black
families. Human Nature 1: 123-143, 1990.
Chisholm, J. Death, hope, and sex: life history theory and the development of reproductive
strategies. Current Anthropology 34: l-12, 18-24, 1993.
Clutton-Brock, T.H. The Evolution of Parental Care, Princeton, NJ: Princeton University Press,
1991.
Daly, M., and Wilson, M. Sex, Evolution, and Behavior, (2nd. Ed.) Belmont, CA: Wadsworth,
1983.
Dickemann, M. Multiple genders and reproductive strategies in humans: steps toward a radical
sociobiology. Paper presented at the Anthropology Colloquim series, University of
Michigan, Ann Arbor, October 7, 1986.
Draper, P., and Belsky, J. Personality development in evolutionary perspective. Journal of
Personality 58: 141-162, 1990.
Draper, P., and Harpending, H. Father absence and reproductive strategy: an evolutionary
perspective. Journal of Anthropological Research 38: 255-273, 1982.
338 E. M. Hill et al.
Emlen, S.T., and Oring, L.W. Ecology, sexual selection, and evolution of mating systems.
Science 197: 215-223, 1977.
Featherman, D.L., and Stevens, G. A revised socioeconomic index of occupational status:
application in analysis of sex differences in attainment. In Social Structure and Behav-
ior: Essays in honor of William Hamilton Sewell, New York: Academic Press, 1982.
Furstenberg, F.F. Jr., Brooks-Gunn, J., and Morgan, S.P. Adolescent Mothers in Later Ltje,
Cambridge, England: Cambridge University Press, 1987.
Hazen, C., and Shaver, P.R. Romantic love conceptualized as an attachment process. J ournul of
Personality and Social Psychology 52: 5 1 l-524, 1987.
Hazen, C., and Shaver, P.R. Love and work: an attachment-theoretical perspective. J ournal of
Personality and Social Psychology 59: 270-280, 1990.
Hill, E.M., and Low, B.S. Contemporary abortion patterns: a life history approach. Ethology and
Sociobiology 13: 35-48, 1992.
Hill, E.M., Nord, _I., and Blow, F.C. Young-adult children of alcoholic parents: protective effects
of positive family functioning. British J ournal of Addiction 87: 1167-1690, 1992.
Hirschfield, M.F., and Tinkle, D.W. Natural selection and the evolution of reproductive effort,
Proceedings of the National Academy of Sciences USA 72: 2227-223 1, 1975.
Krebs, J.R., and Davies, N.B. Behavioural Ecology: An Evolutionurv Approach, Sunderland,
MA: Sinauer Associates, 1984.
Lancaster, J., and Lancaster, C. The watershed: change in parental-investment and family-
formation strategies in the course of human evolution. In Parenting across the Lifespan:
Biosocial Dimensions, J. Lancaster, A. Rossi, .I. Altmann, and L. Sherrod (Eds.). New
York: Aldine, 1987, pp. 187-205.
Lee, P.C., Majluf, P., and Gordon, I.J. Growth, weaning and maternal investment from a
comparative perspective. J ournal of Zoology 225: 99-l 14, 1991.
Maccoby, E.E. Different reproductive strategies in males and females. Child Developmenr 62:
676-681, 1991.
Promislow, D., and Harvey, P. Living fast and dying young: a comparative analysis of life-history
variation among mammals. Journal of rhe Zoological Society of London 220: 4 17-437,
1990.
Promislow, D., and Harvey, P. Mortality rates and the evolution of mammal life histories. Acta
Oecologicu 12: 119-137, 1991.
Reich, W., Earls, F., and Powell, J. A comparison of the home and social environments of
children of alcoholic and non-alcoholic parents. British J ournal of Addiction 83: 831-
839, 1988.
Robins, L., Schoenberg, S.P., Holmes, S.J., Ratcliff, K.S., Benham, A., and Works, J. Early home
environment and retrospective recall: a test for concordance between siblings with and
without psychiatric disorders. American J ournal of Orthopsychiutry 55: 27-41, 1985.
Rubenstein, D.I. Risk, uncertainty and evolutionary strategies. In Current Problems in Sociobio-
logy, Kings College Sociobiology Group (Eds.). Cambridge, England: Cambridge
University Press, 1982, pp. 91-111.
Sroufe, L.A. The coherence of individual development, American Psychologist 34: 834-841,
1979.
Surbey, M.K. Family composition, stress, and the timing of human menarche. In The Socioendo-
crinology of Primate Reproduction. F.B. Bercovitch and T.E. Zeigler (Eds.). New York:
Alan R. Liss, 1990, pp. 71-97.
Trivers, R. Parental investment and sexual selection. In Sexual Selection and the Descent of Man:
1871-1971, B. Campbell (Ed.). Chicago: Aldine, 1972, pp. 136179.
Turke, P. Evolution and the demand for children. Population and Development Review 15:
61-90, 1989.
Weissman, M.M., and Bothwell, S. Assessment of social adjustment by patient self-report.
Archives of General Psychiatry 33: 111 l-l 115, 1975.
Williams G.C. Adaptation and Natural Selection: A Critique of Some Current Evolutionary
Thought, Princeton, NJ: Princeton University Press, 1966.
Wittenberger, J.F. Animal Social Behavior, Boston: Duxbury, 198 1.