Journal of the American Academy of Child and Adolescent Psychiatry

Jan, 2002

Risk and protective factors as predictors of outcome in adolescents with

psychiatric disorder and aggression.(Statistical Data Included)

Author/s: J. Eric Vance

ABSTRACT

Objective: To identify predictors of behavioral outcomes in high-risk adolescents with

aggression and serious emotional disturbance (SED).

Method: Three hundred thirty-seven adolescents from a statewide North Carolina

treatment program for aggressive youths with SED were followed between July 1995 and
June 1999 from program entry ([T.sub.1]) to approximately 1 year later ([T.sub.2]).
Historical and current psychosocial risk and protective factors as well as psychiatric
symptom severity at [T.sub.1] were tested as predictors of high and low behavioral
functioning at [T.sub.2]. Behavioral functioning was a composite based on the frequency
of risk-taking, self-injurious, threatening, and assaultive behavior.

Results: Eleven risk and protective factors were predictive of [T.sub.2] behavioral
functioning, while none of the measured [T.sub.1] psychiatric symptoms was predictive. A
history of aggression and negative parent--child relationships in childhood was predictive
of worse [T.sub.2] behavior, as was lower IQ. Better [T.sub.2] behavioral outcomes were
predicted by a history of consistent parental employment and positive parent--child
relations, higher levels of current family support, contact with prosocial peers, higher
reading level, good problem-solving abilities, and superior interpersonal skills.

Conclusions: Among high-risk adolescents with aggression and SED, psychiatric symptom
severity may be a less important predictor of behavioral outcomes than certain risk and
protective factors. Several factors predictive of good behavioral functioning represent
feasible intervention targets. J. Am. Acad. Child Adolesc. Psychiatry, 2002, 41(1):36-43.

Keywords: aggression, risk and protective factors.

Children with problems of severe aggression superimposed on serious emotional

disturbance (SED) constitute a familiar challenge to most mental health service systems.
The most severely affected of these children are often excluded from special education
settings, ejected from community-based mental health programs, and sometimes placed
in highly restrictive residential, correctional, or institutional settings. Unfortunately,
longitudinal studies of aggression have indicated that early childhood aggression has a
relatively high likelihood of persistence over time (Loeber, 1982; Olweus, 1979). While a
few interventions have begun to show promise in such youths (Borduin et al., 1995; Tate
et al., 1995), it remains a significant challenge to the field of child and adolescent
psychiatry to identify factors associated with the attenuation of aggression and severe
behavioral disorders.

Although the traditional goal of psychiatric care has been to identify diagnostic
syndromes and attempt symptom reduction, it is not clear whether treating psychiatric
symptoms is necessary or sufficient to improve behavioral outcomes in aggressive
youths. Reviews of treatment effectiveness in programs for violent youths have revealed
that narrowly focused treatment approaches provide less enduring positive changes than
comprehensive interventions that impact on multiple realms of the adolescent's life (Tate
et al., 1995). This is not surprising, given the multifactorial origins of youth aggression
(Lewis et al., 1979), and it explains the success of treatment programs that
simultaneously target multiple risk factors for antisocial and aggressive behaviors

1
(Henggeler et al., 1992). While these comprehensive treatment approaches have begun
to show promise, it remains important to discover what other factors might be
responsible for the persistence or desistence of childhood aggression. Elucidation of such
fac tors would certainly further the evolution of effective intervention strategies for
aggressive youths.

The understanding of developmental psychopathology has been strengthened in recent

years by the emergence of risk and resilience as an explanatory framework to predict
psychosocial outcomes of high-risk youths. In spite of definitional variations of risk and
protective factors in the literature (e.g., Bogenschneider, 1996; Gilgun, 1996; Loeber,
1990; Werner, 1990), a large number of individual and environmental factors that are
predictive of developmental outcomes in children and adolescents have been consistently
documented (Fraser, 1997). Risk and resilience researchers have consistently noted the
deleterious effects of an accumulation of four or more risk factors in increasing the
likelihood of developing childhood psychiatric disorder (Rutter, 1979; Sameroff, 1985).

On the other hand, the ability of certain protective factors to mitigate the negative effects
of risk factors has also been documented. The phenomenon of "invulnerability," or
psychosocial resilience, was first well described in longitudinal studies of children who
were expected to do poorly, having grown up in families with psychiatrically ill parents
(Anthony, 1974). Subsequently, a number of other longitudinal studies of child
development in community-based populations have validated the notion of resilience,
finding that a certain fraction of children exposed to high levels of risk nonetheless have
positive life outcomes (Garmezy et al., 1984; Luthar, 1991; Rutter, 1979; Werner and
Smith, 1982, 1992; Wyman et al., 1992). These studies have consistently pointed to a
variety of specific psychosocial protective factors, which seem to confer resilience in the
face of risk.

Nonetheless, some researchers have pointed out that the concept of resilience may be
less useful in the face of overwhelming risk, which almost invariably leads to poor
outcomes (Tolan, 1996). It has not been shown, however, whether or not the presence of
protective factors might attenuate the severity of problems incurred by adolescents
exposed to overwhelming risk. Studies being conducted in a population of adolescents in
North Carolina are beginning to document that risk and protective processes may operate
differently for children and adolescents who possess overwhelming levels of risk, i.e., well
over four or more risk factors (Bowen, unpublished report to Willie M. Program Evaluation
Branch, 1999; Bowen, unpublished dissertation, 1999; Vance et al., 1998). In this
population, which is the focus of the current study, protective factors appear to take on
an increasingly significant role in predicting whether individuals are able to function and
individual risk factors or cumulative level of risk becomes l ess predictive.

In the current study, an extensive list of well-established risk and protective factors and
scores from the Brief Psychiatric Rating Scale for Children (BPRS-C) (Overall and
Pfefferbaum, 1982) were tested to identify which characteristics or statuses at the time of
adolescents' entry into North Carolina's Willie M. Program were predictive of high or low
behavioral functioning approximately 1 year later. On the basis of previous research on
the Willie M. population, we hypothesized that psychiatric symptoms would be less
predictive of outcomes than risk and protective factors and that risk factors would be less
predictive of outcomes than protective factors. Identifying predictive factors that
represent potential intervention targets in this population of youths with aggression and
SED may contribute to an understanding of the causes of aggression and appropriate
treatment targets for what remains one of the biggest challenges to our mental health
systems.

2
METHOD

Study Population

In 1981, a federally enforced class action lawsuit (the Willie M. lawsuit) compelled the
State of North Carolina to achieve a good-faith effort to treat and rehabilitate youths with
severe aggression and SED. Class membership was defined as any child under age 18
who became state-certified as severely or chronically aggressive, who had a neurological
or psychiatric disorder, and who had been placed in public custody or excluded from
access to needed treatment or educational services. The certification process involved
multiagency nomination of affected children to an independent state review panel,
consisting of licensed mental health clinicians (psychiatrists, psychologists, clinical social
workers), for review of relevant psychiatric history, diagnostic and testing materials, and
required legal criteria, to determine eligibility as a Willie M. class member. Eligible youths
for review and certification must have had a history of repeated or severe aggression and
must have received, in previous evaluations by a qualified mental health practitioner, a
diagnosis of a DSM-IV psychiatric disorder (American Psychiatric Association, 1994) or
medically documented neurological disorder.

Depending on the individualized treatment and habilitation needs of the certified child,
Willie M. Program services always included assertive case management, as well as some
combination of various psychotherapies (cognitive-behavioral, insight-oriented,
supportive, sex offender-specific, group, family), psychopharmacological intervention,
psychiatric or neurological care, paraprofessional services, special education, day
treatment, therapeutic foster home or group home placement, or even secure residential
services. Whenever possible, treatment was delivered in the least restrictive setting, in
the home community, and often consisted of intensive "wrap-around" services. The
quality and availability of services varied significantly between local provider agencies
across the state, but a centralized system of funding, training, quality monitoring, and
referral existed to attempt to achieve the state's goal of appropriate treatment for every
class member.

Data collected for the purpose of monitoring the Willie M. Program indicate the
prevalence and severity of problems in the population served (North Carolina Department
of Human Resources and Department of Public Instruction, 1998). Within the state,
roughly one child in every 1,200 became a certified Willie M. class member during the
tenure of the program. The average age at certification was 13 years, 53% of the class
were white, 42% were African American, and the remainder was composed of other
race/ethnic groups. While structured diagnostic instruments were rarely administered to
class members, the working clinical diagnoses were reviewed annually by qualified
mental health clinicians, revealing one third of the youths had three or more psychiatric
diagnoses, while more than two thirds had at least two diagnoses. Finally, the average
number of psychosocial risk factors for Willie M. certified youths was approximately 13,
far exceeding most definitions of high risk.

Analyzed Sample

Of 1,464 adolescent cases with annual assessment data as of July 15, 1999, 337 (23%)
who met the following criteria were included in the current study: (1) time 1 ([T.sub.1])
and time 2 ([T.sub.2]) assessment data were available, (2) subjects were aged 13 years
or older at the time of their first assessment, (3) cases had complete data on the four
behavior variables used to create the [T.sub.2] behavioral functioning measure, and (4)
the first assessment occurred within 90 days of certification.

No significant differences were found in the gender or race/ethnic composition of the

analyzed sample and the Willie M. adolescent population with assessment data. The
average age of the analyzed sample, however, was about 5 months lower than the

3
average age of the population. Because this difference was statistically significant,
caution should be exercised in generalizing results from the sample to the entire
population.

About 80% of the analyzed sample were males and 21% were females. Whites
constituted just more than half (53%) of the analyzed sample. About 90% of the
remaining analyzed sample were African Americans. The average age of the analyzed
sample was 14.8 years.

Measures

A four-section instrument called the Assessment and Outcomes Instrument (AOI) (Willie
M. Program, 1996) was used by the Program Evaluation Branch of the Willie M. Division
starting in 1995. Program participants were assessed at the time of certification and at
approximately 1-year intervals thereafter. The annual assessment protocol constituted
the first longitudinal evaluation of a large clinical population that attempted to document
the presence or absence of both current and historical risk and protective factors in each
child, while also documenting psychosocial and behavioral functioning, quality of social
environment, and psychiatric symptoms.

The first section of the AOI, the Developmental Risk Assessment, contains 82
dichotomous items assessing risk and protective factors during infancy, the preschool
stage, early school years, and adolescence. The second section of the AOI is a modified
version of the BPRS-C (Overall and Pfefferbaum, 1982), containing 24 items measured
with 7-point scales. The BPRS-C contains subscales related to (1) behavior problems, (2)
depression, (3) thinking disturbance, (4) motor agitation, (5) withdrawal, (6) anxiety, (7)
organicity, and (8) socialization and a total score. The modification consists of the
addition of the Socialization subscale (mistrust of others, problems getting along with
adults, problems getting along with other children). Subscale and total scores were
obtained by summing the scores of the appropriate indicators. The third section of the
AOI, the Child Interview, is completed by the professional (clinician or case manager)
most familiar with the child while interviewing the child. Special instructi ons are provided
for obtaining information from younger children and children with developmental
disabilities. The child's perceptions of his or her behavior, personal strengths, and social
environment are assessed using 35 dichotomous and ordinal-level items. The final section
of the AOI, the Functional Domain Assessment, contains 59 categorical and dichotomous
or ordinal-level questions assessing treatment team, teacher, and parent perceptions of
children's current behavior and functioning, social skills, and social relationships. IQ
testing, most often with the WISC-III (Wechsler, 1991), was routinely performed for
purposes of certification or treatment planning and is recorded on the Functional Domain
Assessment. Teacher report of reading level relative to grade level is also recorded.
Individual and composite items from all four sections of the AOI were examined in the
current study.

The assessment instrument was administered annually by practicing clinicians and case
managers and therefore provides a naturalistic view of assessment and outcome tracking
as it can occur in existing mental health service systems. A 1998 study conducted by
Thompson and Kaval for the Willie M. Program Evaluation Branch (unpublished report)
revealed high interrater reliability between the area program case managers, who
typically complete the instrument, and Regional Service Managers for the program and
State Department of Public Instruction Consultants. Interrater agreement on dichotomous
questions averaged 95%, and Pearson correlations on ordinal questions averaged 0.93.

Data Analysis

The current analysis was designed to determine how well [T.sub.2] behavioral functioning

4
of Willie M. youths could be predicted by [T.sub.1] risk and protective status and
psychiatric symptoms. [T.sub.2] behavioral functioning was determined by generating
factor scores from four [T.sub.2] variables measuring the frequency of risky behaviors
(e.g., unprotected sex, involvement with a gang or "bad crowd," or running away to risky
situations), self-injurious behaviors, threatening behaviors, and assaultive behaviors.
These variables were coded in the positive direction, so higher numbers indicated better
behavior. The distribution of the behavior factor scores was divided into thirds.
Adolescents with scores in the top and bottom thirds of the distribution were assigned to
the "high" behavioral functioning group ("high performers"), and "low" behavioral
functioning group ("low performers"), respectively. To maximize the potential to find
significant group differences on characteristies, sample members in the mid dle third of
the distribution were excluded from the analyses. A total of 221 adolescents were
included in the comparisons; of these 109 were low performers and 112 were high
performers at [T.sub.2]. Analyses were not conducted separately for white and African-
American youths because a previous study (Bowen, unpublished dissertation, 1999)
demonstrated that cross-sectional and longitudinal relationships between risk and
protection status and behavior do not vary by race/ethnicity.

The [T.sub.2] groups with high and low behavioral functioning were compared on 76
[T.sub.1] risk and protective factor variables and composites from the AOI and the
[T.sub.1] BPRS-C subscale scores and total score (Table 1). Cross-tabulations and Pearson
[chi square] tests were used to identify significant group differences for [T.sub.1]
variables with two or three response options. We used t tests to determine whether the
average group scores differed significantly for high and low performers on variables with
four or more response options that had at least an ordinal level of measurement. A two-
railed significance level of .05 was used as the criterion for establishing statistical
significance.

Based on the sample sizes for the two groups examined, the current study had adequate
power (0.80 or greater) to detect effect sizes of approximately 0.40 in the [chi square]
tests described above (Cohen, 1988). The study had adequate power to detect effect
sizes of 0.30 in the t test analyses. This means that the tests of ordinal variables,
including the BPRS-C indicators, were more sensitive to group differences than the
analyses of most risk and protective factor variables. Effect sizes of 0.30 and 0.40 lie
between Cohen's definitions of "small" (0.20) and "medium" (0.50) effect sizes. For the
[chi square] proportions, the effect size of 0.40 meant the study had adequate power to
detect minimum percentage differences of between 9 and 20 points for the high and low
performance groups. These power levels were satisfactory for the exploratory purposes of
the current study.

RESULTS

Adolescents with high and low behavioral functioning at [T.sub.2] differed significantly on
11 (14.5%) of the 76 variables tested. The findings are presented in Table 2 ([chi square]
test results) and Table 3 (t test results). The significant variables and composites included
three historical risk factors, one current risk factor, two historical protective factors, and
five current protective factors. Psychiatric symptoms as assessed by the BPRS-C at
[T.sub.1] were not associated with behavioral functioning at [T.sub.2].

As shown in Table 2, youths in the high behavioral performance group at [T.sub.2] were
less likely to be classified as having a problem with aggression at the preschool and
school-age stages, based on a dichotomous risk variable. (Teenage aggression was not a
significant predictor at the .05 level because of the lack of variance on the measure.
Almost all of the youths in both performance groups had this risk factor.)

The presence of a negative parent--child relationship at the three developmental stages

5
was also assessed with dichotomous variables. Although negative parent--child
relationships were common in both groups, high performers were significantly less likely
to have had a negative relationship with their parents during their school-age years.

Current Risk Factors

Analysis of a three-category IQ measure (55-69,70-99, 100 and over) revealed that high-
performing teenagers at [T.sub.2] were significantly less likely than low performers to
have IQs in the lowest category (Table 2).

Historical Protective Factors

Dichotomous variables were used to indicate the presence of positive parent--child

relationships and consistent parent employment at each developmental stage. As shown
in Table 2, high performers were more likely than low performers to have had a positive
relationship with their parents and to have had a parent who was consistently employed
during their school-age years.

Current Protective Factors

High and low performers differed on three current protective factors related to individual
skills: problem-solving, reading, and interpersonal skills. A dichotomous variable assessed
teacher perceptions that the child had problem-solving and reasoning skills for academic
and nonacademic problems. High performers were more likely to have this protective
factor (Table 2). As shown in Table 3, high performers also had superior scores on an
ordinal measure assessing youths' reading level relative to grade level and on a
continuous factor score composite measuring interpersonal skills (likability, empathy, and
sense of humor) as reported by parents and teachers.

High and low performers also differed significantly on two current protective indicators
related to the social environment: family support and quality of peer affiliation. Both
measures were assessed with 5-point ordinal scales. As shown in Table 3, youths who
were high performers at [T.sub.2], on average, enjoyed higher levels of family support at
[T.sub.1] and had more friends who stayed out of trouble did than low performers.

DISCUSSION

This study found a number of significant [T.sub.1] predictors of favorable behavioral

outcomes after a year of treatment for youths with SED and aggression. The finding that
a history of predominantly negative parent--child interaction in the home was predictive
of poor behavioral outcomes is consistent with prevailing theories implicating family
relationship factors among those contributing to aggressive behaviors (Farrington, 1991).
In contrast, having a parent who was consistently employed during the preteen years
seemed to exert a protective effect on behavioral outcomes. Insofar as employment may
be a marker for parental competence, this finding builds on other research indicating that
parental competency is a protective factor in relation to longitudinal functional outcomes,
at least in children with attention-deficit hyperactivity disorder (Fischer et al., 1993).

Certain cognitive features of the child were also predictive of behavioral outcomes. Lower
IQ predicted poorer outcomes. This finding supports the documentation of frequent
neurocognitive deficits among conduct-disordered and aggressive youths (Lewis et al.,
1979). Conversely, better reading skills, as measured relative to expected grade level,
were positively predictive of good behavioral out-comes in the current study population.
This supports the findings that reading skills serve as a protective factor among high-risk

6
children in community samples (Werner and Smith, 1992).

Better behavioral outcomes were also predicted by several factors that suggest superior
interpersonal skills (being likable, being empathetic, having a sense of humor, and being
a good problem-solver), as perceived by parents and teachers. This finding is consistent
with other research pointing to the protective effects of humor (Masten, 1986), likability
(Garmezy et al., 1984; Wyman et al., 1992), empathy (Luthar, 1991; Werner and Smith,
1992), and interpersonal problem-solving abilities (Shure and Spivack, 1988) among high-
risk youths.

The other protective factors that were shown in this study to be predictive of better
behavioral outcomes were related to social support networks, including the amount of
current emotional support provided by the family for the adolescent and affiliation with
predominantly prosocial peers. The possible beneficial effect of family support for youths
with aggression was also suggested in a follow-up study of previously incarcerated violent
delinquents who returned to their family homes (Lewis et al., 1994). After controlling for
level of psychobiological vulnerabilities and early juvenile violence of the youths, those
who returned home to families had lower rates of recidivism to violence than those who
went to penal or therapeutic placements. The protective effect of prosocial peer affiliation
is consistent with the converse findings that peer affiliation with antisocial youths predicts
persistence of delinquency (Huizinga et al., 1993). Other studies have noted the
protective effect of positive peer support for high-risk youths (Werner and Smith, 1992).

It is notable that in spite of multiple, comorbid psychiatric conditions often diagnosed in

Willie M. youths, none of the clinician-rated BPRS-C psychiatric symptoms identified at
[T.sub.1] were predictive of eventual behavioral outcome. In spite of considerable
variation within and between symptom subscales on the BPRS-C among study subjects,
initial psychiatric symptom severity was not predictive. However, early childhood
aggression during preschool or preteen years, a psychiatric symptom that was measured
elsewhere on the AOI, was a significant predictor of poor behavioral outcomes, supporting
other findings on the persistence over time of behavioral problems, especially aggression,
in the wake of early childhood aggression (Loeber, 1982; Olweus, 1979).

Limitations

The current study had several limitations. First, the population studied was a unique and
extremely impaired group of adolescents, which limits the generalizability of the findings.
Second, the 1-year follow-up was a relatively short period of time in which to draw
conclusions about durable changes in level of functioning. Also, because the study is
neither prospective nor experimental in design, the relationships between [T.sub.1]
predictors and [T.sub.2] behavioral outcomes can only be viewed as associational, not
causal. The study had adequate power to detect small-medium effect sizes, but not small
effect sizes. If larger sample sizes had been available, additional variables may have
been identified as predictors of [T.sub.2] behavior. Although all study subjects received
case management services, we were unable to control our findings in relation to the
varying amounts of treatment and habilitation services received across study subjects.
However, in other analyses, differences in levels of individual s ervices were not highly
predictive of outcomes (Bowen, unpublished report to Willie M. Program Evaluation
Branch, 1999). Finally, the assessment instruments used for this study were administered
by case managers and clinicians in the field, rather than trained research personnel. This
may somewhat limit the reliability of the findings but at the same time may increase the
practical applicability of the findings for clinicians working in community-based treatment
settings.

Conclusions and Clinical Implications

7
This study suggests that important predictive characteristics of the relative risk and
resilience of adolescents with SED can be reliably ascertained by clinicians working in the
field, using a cross-sectional and retrospective assessment tool. The finding that several
well-known psychosocial risk and protective factors significantly predicted behavioral
outcomes in this particular group of adolescents suggests that risk and resilience
mechanisms continue to moderate outcomes, even at extremely high levels of risk and in
the presence of severe disorder.

Most of the predictors of better behavioral outcomes in this population were protective
factors, in contrast to the findings of studies of community-based samples, which have
generally found risk factors to be more predictive of outcomes than protective factors
(Fraser et al., 1999; Jessor et al., 1995; Pollard et al., 1999). As a result of the findings in
community-based samples, some researchers have cautioned practitioners against
relying exclusively on protection-building interventions and encouraged a focus on risk
reduction (Fraser et al., 1999). However, among extremely high-risk youths with
aggression and SED, it may be that building protective factors serves better to attenuate
behavioral problems than attempting to reduce risk. In the current study this observation
may be partly attributable to the fact that most assessed risk factors were historical and
immutable factors, while protective factors represented feasible intervention targets.
Nonetheless, where risk reduction represents a targetable g oal, interventions might be
designed around it.

With regard to intervention strategies for adolescents with aggressive disorders, this
study identifies a number of potential opportunities to moderate risk factors, while
building protective factors in youths with SED and aggression. Secondary prevention
strategies, which intervene to reduce early childhood aggressive behavior, decrease
negative parent--child interactions, and enhance parental competence and employability,
may have positive long-term consequences for the child with aggression. Successful
intervention with adolescents with aggressive disorders may require building social
support bridges to estranged families or prosocial peers. Interpersonal problem-solving
and other social skills might need to be taught in the context of therapeutic or mentor
relationships. School interventions should focus on building reading skills but can also be
effective in building interpersonal skills and "emotional intelligence" (Goleman, 1995).
Children with SED and aggression pose a great challenge to the therapeu tic community
but have been helped in programs with a multidimensional approach (Borduin et al.,
1995). The results of this study suggest a number of risk and protective factors that
might be targeted by treatment providers to build resilience and increase the odds of
positive outcomes for youths with aggression.

Addressing Challenges, Creating Opportunities: Fostering Consumer Participation in

Medicaid and Children's Health Insurance Managed Care Programs. Christine Molnar, MS

With more than half of the Medicaid population enrolled in managed care and with
enrollment in Medicaid managed care and Children's Health Insurance Programs
continuing to accelerate, policy makers and program administrators must quickly come to
consensus on the role of the consumer in the planning, implementation, and oversight of
Medicaid and S-CHIP managed care. This article explores the barriers to consumer
participation and makes concrete policy and programmatic recommendations to increase
consumer participation in public managed care programs. To fully comply with existing
federal and state requirements for consumer participation and for this participation to
have an impact on the quality of managed care programs, states need to provide more
education and support to consumers on: how to choose a health plan, how to access the
health plan services and our-of-network benefits; and how to use grievance procedures.
In addition, more work needs to be done to make the growing availability of health plan
and p rovider performance data meaningful and available to Medicaid and S-CHIP
consumers. Finally, states must address the lack of resources, lack of information, and
lack of training that prevent many consumers from being able to sustain their

8
involvement in system-level advisory and oversight bodies. J Ambulatory Care Manage
2001;24:61-67. Reproduced with permission from Aspen Publishers, Inc.

REFERENCES

American Psychiatric Association (1994), Diagnostic and Statistical Manual of Mental

Disorders, 4th edition (DSM-IV). Washington, DC: American Psychiatric Association
Anthony EJ (1974), The syndrome of the psychologically invulnerable child. In: The Child
in His Family: Children at Psychiatric Risk, Vol 3, Anthony EJ, Koupernik C, eds. New York:
Wiley, pp 529-544

Bogenschneider K (1996), An ecological risk/protective theory for building prevention

programs, policies, and community capacity to support youth. Fam Relat 45:127-138

Borduin CM, Mann BJ, Cone LT et al. (1995), Multisystemic treatment of serious juvenile
offenders: long-term prevention of criminality and violence. J Consult Clin Psychol 63:569-
578

Cohen J (1988), Statistical Power Analysis for the Behavioral Sciences. Hillsdale, NJ:
Erlbaum

Farrington DP (1991), Childhood aggression and adult violence: early precutsora and later
life outcomes. In: The Development and Treatment of Childhood Aggression, Pepler DJ,
Rubin K, eda. Hillsdale, NJ: Erlbaum, pp 5-29

Fischer M, Barkley RA, Fletcher KE, Smallish L (1993), The adolescent outcome of
hyperactive children: predictors of psychiatric, academic, social, and emotional
adjustment. J Am Acad Child Adolese Psychiatry 32:324-332

Fraser MW ed. (1997), Risk and Resilience in Childhood: An Ecological Perspective.

Washington, DC: National Association of Social Workers Press

Fraser MW Richman JM, Galinsky MJ (1999), Risk, protection, and resilience toward a
conceptual framework for social work practice. Soc Work Res 23:131-143

Garmezy N, Masten A, Tellegen A (1984), The study of stress and competence in children.
Child Dev 55:97-111

Gilgun JF (1996), Human development and adversity in ecological perspective, part 1: a

conceptual framework Families in Society September: 395-402

Goleman D (1995), Emotional Intelligence. New York: Bantam Books, pp 261-287

Henggeler SW, Melton, GB, Smith LA (1992), Family preservation using multisystemic
therapy: an effective alternative to incarcerating serious juvenile offenders. J Consult Clin
Psychol 60:953-961

Huizinga D, Loeber R, Thornberry TP (1993), Longitudinal study of delinquency, drug use,

sexual activity, and pregnancy among children and youth in three cities. Public Health
Rep 108:90-96

Jessor R, Van Den Bos J, Vanderryn J, Costa FM, Turbin MS (1995), Protective factors in
adolescent problem behavior: moderator effects and developmental change. Dev Psychol
31:923-933

9
Lewis DO, Shanok S, Pincus JH, Glaser G (1979), Violent juvenile delinquents: psychiatric,
neurological, psychological, and abuse factors. J Am Acad Child Psychiatry 18:307-319

Lewis DO, Yeager CA, Lovely R, Stein A, Cobbam-Portorreal CS (1994), A clinical follow-up
of delinquent males: ignored vulnerabiliries, unmet needs, and the perpetuation of
violence. J Am Acad Child Adolesc Psychiatry 33:518-528

Loeber R (1982), The stability of antisocial and delinquent child behavior: a review. Child
Dev 53:1431-1446

Loeber R (1990), Development and risk factors of juvenile antisocial behavior and
delinquency. Clin Psychol Rev 10:1-41

Luthar SS (1991), Vulnerability and resilience: a study of high-risk adolescents. Child Dev
62:600-616
Masten AS (1986), Humor and competence in school age children. Child Dev 57:461-473

North Carolina Department of Human Resources and Department of Public Instruction

(1998), Report to the Governor and the General Assembly on the Willie M Program 1997-
1998. Raleigh: North Carolina Department of Human Resources and Department of Public
Instruction

Olweus D (1979), Stability of aggressive reaction patterns in males: a review. Psychol Bull
86:852-875

Overall JE, Pfefferbaum B (1982), The Brief Psychiatric Rating Scale for Children.
Psychopharmacol Bull 18:10-16

Pollard JA, Hawkins JD, and Arthur MW (1999), Risk and protection: are both necessary to
understand diverse behavioral outcomes in adolescence? Soc Work Res 23:145-158

Rutter M (1979), Protective factors in children's responses to stress and disadvantage. In:
Primary Prevention of Psychopathology, Vol 3, Kent MW, Rolf JE, eds. Hanover, NH:
University Press of New England, pp 49-74

Sameroff AJ (1985), Environmental factors in the early screening of children at risk In:
Early Identification of Children at Risk: An International Perspective, Frankenburg WK,
Emde RN, Sullivan JW, eds. New York: Plenum, pp 21-44

Shure MB, Spivack G (1988), Interpersonal cognitive problem solving. In: 14 Ounces of
Prevention, Price RH, Cowen EL, Lotion RP, Ramos-McKay J, eds. Washington, DC:
American Psychological Association, pp 69-82

Tate DC, Repucci ND, Mulvey EP (1995), Violent juvenile delinquents: treatment
effectiveness and implications for future action. Am Psychol 50:777-781

Tolan PH (1996), How resilient is the concept of resilience? Community Psychologist

29:12-15

Vance JE, Fernandez G, Biber M (1998), Educational progress in a population of youth with
aggression and emotional disturbance: the role of risk and protective factors. J Emotional
Behav Disord 6:214-221

Wechsler D (1991), Manual for the WISC-III. New York: Psychological Corporation

10
Werner EE (1990), Protective factors and individual resilience. In: Handbook of Early
Intervention, Meisels SJ, Shonkoff JP, eds. Cambridge, England: Cambridge University
Press, pp 97-116

Werner EE, Smith RS (1982), Vulnerable But Invincible: A Longitudinal Study of Resilient
Children and Youth. New York: McGraw-Hill

Werner EE, Smith RS (1992), Overcoming the Odds. High Risk Children From Birth to
Adulthood. Ithaca, NY: Cornell University Press

Willie M. Program (1996), Willie M Assessment and Outcomes Instrument. Raleigh, NC:
Willie M. Program Evaluation Branch

Wyman PA, Cowen EL, Work WC et al. (1992), Interviews with children who experienced
major life stress: family and child attributes that predict resilient outcomes. J Am Acad
Child Adolesc Psychiatry 31:904-910

TABLE 1

Time 1 Variables Tested for Relationship to Time 2 Behavior

Protective Factors Risk Factors

Protective factors from preschool, Risk Factors from preschool,

school-age, and teenage school-age, and teenage
developmental stages developmental sages
Positive discipline (DRA) (a) Aggression (DRA)
Daily structure (DRA) Poverty (DRA)
Positive parent-child Divorce, separation, or single
relationship (DRA) parent (DRA)
Family church involvement (DRA) Negative parent-child relationship
(DRA)
Parent consistent employment (DRA) Witness to family violence (DRA)
Parent high school education Neglect (DRA, assessed only at
or higher (DRA) preschool and school-age stages)
Access to alternative child Physical abuse (DRA)
care (DRA)
Additional current protective Sexual abuse (DRA)
factors
Believe parent cares (CI) (b) Composite of parent mental health,
substance use, criminal behavior
(DRA)
Family support (FDA) (c) Additional current risk factors
Lives at home (FDA) Low IQ (FDA,teacher)
Problem-solving ability (FDA, Medical status (FDA)
teacher)
Reading level (FDA, teacher) Needed medical attention in past 3
months (FDA)
Peer support (CI) Needed psychiatric attention in
past 3 months (FDA)
Peer trouble avoidance (CI) Number of residential moves in past
3 months (FDA)
Positive school attitude (CI) Psychiatric symptoms
Internal locus of control (CI) Eight subscales of BPRS (d) (sum of
3 subscale items for each of 8
disorders)
Use of faith (CI) Total score on BPRS (sum of 24
items)
Believe treatment helps (CI)
Adult support sources composite
(count of adult friend at school,
outside school, and involvement
with mentor or community
activities, CI, FDA)
Number of competencies (CI)
Interpersonal skills (based on
regression factors scores for
parent and teacher reports of
likability, empathetic, sense of
humor; FDA, teacher; FDA,
parent)
Involvement composite (based on
regression factor scores for

11
 child's involvement in
extracurricular activities,
church, community activities,
CI, FDA)

(a)Developmental Risk Assessment; completed by the case manager, who

consults other sources as appropriate.

(b)Child Interview; completed by case manager or clinician during an

interview with the child.

(c)Functional Domain Assessment; completed by the case manager in

consulation with team members and records, and during interviews with
parents or teachers. Parent or teacher sources are indicated where
appropriate in the table.

(d)Brief Psychiatric Rating Scale; completed by clinician who is

familiar with the child.
TABLE 2

Significant Time 1 Differences on Categorical Variables Between High and

Low Performers

(Based on Level of Time 2 Behavioral Functioninig)

Low Performers High

Performers
Risk or Protective Factor No. % No.

Preschool aggression 50 53.8 35

Schoo1-age (a) aggression 98 93.3 88
School-age negative relationship
with parent 74 70.5 54
IQ of 55-69 33 36.7 13
Positive school-age relationship
with parent 29 28.2 47
Parent consistently employed during
child's school-age stage 69 67.0 86
Problem-solving and reasoning skill 27 28.1 48

High
Performers
Risk or Protective Factor % [chi square] p

Preschool aggression 36.5 5.72 .02

Schoo1-age (a) aggression 84.6 4.06 .04
School-age negative relationship
with parent 52.9 6.74 .01
IQ of 55-69 14.0 12.51 .00
Positive school-age relationship
with parent 47.5 8.03 .01
Parent consistently employed during
child's school-age stage 81.9 6.09 .01
Problem-solving and reasoning skill 49.5 9.27 .00

(a)School-age is defined as from the time a child starts school through

age 12.
TABLE 3

Significant Time 1 Differences on Ordinal and Continuous Variables

Between High and Low Performers

(Based on Level of Time 2 Behavioral Functioning)

Low Performers High

Performers
(n = 109 (a)) n = 112
(a))
Risk or Protective Factor Mean (SD) Mean

Reading level (range: 1-5) 2.53 (1.50) 3.29

Interpersonal skills (factor score
composite with mean = 0) -0.13 (0.84) 0.11
Family support (range: 1-5) 4.07 (1.11) 4.40
Proportion of peers who stay out of
trouble (range: 1-5) 2.74 (1.28) 3.15

High
Performers
n = 112 (a))
Risk or Protective Factor (SD) t p

12
Reading level (range: 1-5) (1.57) -3.41 .00
Interpersonal skills (factor score
composite with mean = 0) (0.74) -2.25 .03
Family support (range: 1-5) (0.92) -2.40 .02
Proportion of peers who stay out of
trouble (range: 1-5) (1.19) -2.43 .02

(a)n's for individual analyses varied because of missing values.

COPYRIGHT 2002 Lippincott/Williams & Wilkins
in association with The Gale Group and LookSmart. COPYRIGHT 2002 Gale Group

Journal of the American Academy of Child and Adolescent Psychiatry

May, 2002

Group intervention for children bereaved by the suicide of a relative.(Statistical

Data Included)

Author/s: Cynthia R. Pfeffer

Children suffer bereavement after death of a close relative in approximately one third of
30,000 suicides in the United States annually. While the death of a close relative is one of
the most stressful events for children (Ness and Pfeffer, 1990; Osterweiss et al., 1984),
there is a paucity of research on naturalistic outcomes and interventions of childhood
bereavement after the suicide of a relative.

Controlled studies of children bereaved by parental or sibling deaths from illnesses

suggest significantly more psychosocial problems including anxiety and depression within
the first 2 years after death compared with non-bereaved children (Weller et al., 1991;
Worden and Silverman, 1996). Children bereaved after suicide of relatives also suffer
psychosocial problems. Parental reports of 36 children bereaved by parental suicide and
36 age and gender-matched nonbereaved children indicated that bereaved children had
significantly greater anxiety, aggression, or withdrawal within weeks after the suicide
(Shepherd and Barraclough, 1976). Direct evaluations of 22 children within a year after
parental or sibling suicide indicated they had significantly higher anxiety and depression
compared with scores previously collected from normative child samples (Pfeffer et al.,
1997). During a 3-year follow-up of 25 siblings of 20 adolescent suicide victims and 25
adolescents not exposed to suicide of a peer, bereaved sibli ngs were at 7-fold increased
risk for developing major depressive disorder within 6 months after death but not at
increased risk for subsequent major depressive episodes, posttraumatic stress disorder,
or suicidal behavior (Brent et al., 1996).

Comparative research with 26 children bereaved by parental suicide and 332 children
bereaved by parental nonsuicidal death suggested that suicide-bereaved children had
greater anxiety, anger, and shame within the postdeath year but similar rates of
posttraumatic stress or suicidal behavior to nonsuicide-bereaved children (Cerel et al.,
1999). Within 18 months of parental death from cancer or suicide, the 16 suicide-
bereaved children reported significantly more severe depressive symptoms than the 64
age-matched cancer-bereaved children (Pfeffer et al., 2000). Parental reports of the
suicide- and cancer-bereaved children's social competence and behavior were similar to a
normative sample.

These empirical results support the need for interventions to promote children's healthy
adjustment after family suicide. This article compares outcomes of children who had or
had not received a manual-based group intervention targeted to assist children with
bereavement resulting from parental or sibling suicide. A literature review revealed no
reports of empirical studies of efficacy of group intervention for children bereaved by the
suicide of a relative. It was hypothesized that reduction in anxiety, depressive, and
posttraumatic stress symptoms and improvement in social adjustment would be
significantly greater among children who had compared with those who had not received
the intervention.

13
METHOD

Sample

One hundred twelve families with children were identified from medical examiners' lists
of consecutive suicide victims from January 1996-November 1999. Initial contacts with
families were by letter sent in collaboration with the medical examiners and telephone
discussions 3 weeks later to determine whether the deceased was a parent or sibling of
children. Ten families refused to participate and 27 could not be located. At the initial
visit, surviving parents and children (75 families, 102 children) signed institutional review
board-approved consent and assent forms, respectively, agreeing to participate. Eligibility
was determined at this visit. Children, aged 6 through 15 years, were eligible except if
they did not speak English, had clinically estimated mental retardation, did not know the
cause of death was suicide, or did not have a participating parent/caretaker. Children
with current psychiatric disorders were excluded because this study evaluated efficacy of
the intervention to decrease bereavement-rel ated symptoms rather than those related
to psychiatric disorders.

After eligibility was established, families were assigned in alternating order to or not to
receive the intervention. If there was more than 1 month's wait to recruit a family, the
next eligible family was assigned to intervention to avoid delay in beginning intervention.
In this case, once at least two families were available for intervention, it began and the
next family was assigned to not receive the intervention. Those who received the
intervention did not receive other interventions. Those who did not receive the
intervention could receive other interventions but participated in the research
assessments. Such families received bimonthly brief telephone calls to maintain contact.

Research Measures

Standard, reliable research instruments were used to measure psychosocial variables at

initial ([T.sub.1]) and outcome ([T.sub.2]) assessments occurring approximately 12 weeks
apart. Semistructured interviews of children and parents using standard research
instruments were conducted by master's-level psychologists trained by the primary
author (C.R.P.) to have interrater reliability (Cohen, 1960) of [kappa] [greater than or
equal to] 0.9. Interviewers were blind to intervention status.

The Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present
State (K-SADS-IVR) (Ambrosini and Dixon, 1996; Kaufman et al., 1997) was administered
at initial visit to assess eligibility by determining whether children had current psychiatric
disorders. This was rated in semistrucrured interviews of children and separately of
parents who reported on their children's psychiatric symptoms. Consensus ratings of
presence of psychiatric disorders were obtained by means of parents' and children's
reports.

The Beck Depression Inventory (BDI) (Becker al., 1961), a questionnaire completed by
parents, assessed severity of parental depressive symptoms. Scores [greater than or
equal to] 20 indicated clinically significant symptoms.

The Childhood Posttraumatic Stress Reaction Index (CPTSRI) (Pynoos et al., 1987),
administered in semistructured interviews to children, identified severity of posttraumatic
stress. Scores [greater than or equal to]25 indicated clinically significant symptoms.

The Children's Depression Inventory (CDI) (Kovacs, 1992), a questionnaire completed by

children, assessed severity of depressive symptoms. T scores [greater than or equal

14
to]60 indicated clinically significant depressive symptoms.

The Revised Children's Manifest Anxiety Scale (RCMAS) (Reynolds and Richmond, 1985),
a questionnaire completed by children, assessed severity of anxiety symptoms. T scores
[greater than or equal to]63 indicated clinically significant anxiety symptoms.

The Social Adjustment Inventory for Children and Adolescents (SAICA) (John et al., 1987)
was administered in semistructured interviews to children and separately to parents, who
reported on children's current social adjustment. Consensus ratings of children's social
adjustment were obtained by using parents' and children's reports. Scores [greater than
or equal to]2 indicated clinically significant maladjustment.

Bereavement Group Intervention

The bereavement group intervention (BGI) was offered in ten 1.5-hour group sessions
weekly to bereaved children from two to five families and separately but simultaneously
to parents. Each group, led by a trained master's-level psychologist, was composed of
two to five children of similar developmental levels and grouped by age, i.e., 6 through 9
years, 10 through 12 years, and 13 through 15 years. Siblings were assigned to the same
group unless problems discussing concerns in the presence of siblings or developmental
differences were present. Such siblings were included in different groups.

An intervention manual comprehensively documented the methods of conducting

developmentally specific sessions with children and parents. Additional information about
the manual may be obtained by writing to the primary author. The intervention aimed to
help children cope during bereavement after parental or sibling suicide and to reduce
morbid outcomes. Theoretical models of attachment (Bowlby, 1980), responses to loss
(Ness and Pfeffer, 1990; Parks, 1996; Stroebe et al., 1993), and cognitive coping (Lazarus
and Folkman, 1984) were used in developing the intervention. The main themes focused
on children's understanding of and responses to the death of a parent or sibling, unique
features of suicide, and loss of personal and environmental resources. Prior to this study,
the manual was reviewed by two child and adolescent psychiatrists and two psychologists
who offered suggestions for revision. Subsequently, the intervention was pilot-tested
twice with three families for each intervention.

Psychoeducational components focused on discussing children's concepts of death and

its permanence, identifying feelings of grief, defining what is suicide, discussing why
people commit suicide, discussing prevention of children's suicidal urges, and enhancing
children's skills in problem-solving. Supportive components facilitated children's
expressions of grief and their identification with positive attributes of the deceased but
avoidance of suicidal urges and hopelessness. Children were empowered to feel more
optimistic, manage traumatic thoughts and stigmatizing concerns about the suicide, and
develop new supportive interpersonal relationships.

Psychoeducational and supportive components for parents helped them to understand

childhood bereavement, foster children's expressions of grief, discuss the suicide, identify
children's morbid reactions, and promote children's emotional and social functioning.
Support was provided for parents to ventilate their grief.

Adherence to Intervention

Training leaders to administer the group intervention consisted of extensive review of the
intervention manual with the primary author (C.R.P.). When leaders administered the
intervention, weekly supervision was provided by the primary author. Intervention
sessions were videotaped and rated by a trained evaluator, who had excellent interrater
reliability (intraclass correlation coefficient = 0.85) with the primary author, for ratings on

15
the Therapist Performance Scale of therapist adherence. This scale, a modification of a
reliable treatment adherence instrument (Elkin et al., 1985), rated the group leader's
ability to provide interventions described in the intervention manual. In this study, the
instrument had internal consistency (coefficient [alpha]) of .7 and interrater reliability
(intraclass correlation coefficient) of 0.9. A score [greater than or equal to]4 indicated
adequate adherence of group leaders. If Therapist Performance Scale ratings were below
adherence criteria on any session, more intensiv e discussions with group leaders
occurred.

Statistical Analysis

Effectiveness of assignment was evaluated for demographic and clinical variable

balances at initial assessment ([T.sub.1]) of children and parents who had or had not
received the intervention. Fisher exact tests for categorical variables and t tests for
continuous variables were used. Outcome variables to evaluate the intervention's efficacy
were children's anxiety (RCMAS), depressive (CDI), and posttraumatic stress (CPTSRI)
symptoms; social adjustment (SAICA); and parents' depressive (BDI) symptoms. Variables
confounding outcomes were identified as a variable significantly associated with initial
([T.sub.1]) scores of an outcome variable and if the variable at initial assessment
([T.sub.1]) was significantly different between children who had or had not received the
intervention. A variable, identified as confounding or unbalanced at [T.sub.1] between
children who had or had not received the intervention, was included in efficacy analyses.

A change measure was operationally defined as the difference between [T.sub.2] and
[T.sub.1] scores of specific clinical variables, The difference in this measure between
children who had or had not received the intervention indicated intervention efficacy.
Analysis of covariance (ANCOVA) was used to model outcome variables at [T.sub.2] as a
function of receiving or not receiving the intervention, outcome variable's initial
([T.sub.1]) scores, children's age, and time between [T.sub.1] and [T.sub.2]. Efficacy was
estimated as adjusted mean change score ([T.sub.2] scores minus [T.sub.1] scores) for
children who had or had not received the intervention. Two other change measures
included difference score ([T.sub.2] minus [T.sub.1] scores) divided either by initial
([T.sub.1] score of the outcome variable or time between [T.sub.1] and [T.sub.2]
assessments. The former represented change relative to initial ([T.sub.1]) status while
the latter adjusted for differential time intervals between [T.sub.1] and [T.sub.2] and
provided an indicator of rate of change per unit time. Since some families had multiple
children, codependence of siblings was controlled using mixed-effects models (Gibbons
cc al., 1993) that allowed evaluating correlated outcomes by using the family as a unit.
Within-group effects among children who did or did not receive the intervention to
evaluate differences between [T.sub.1] and [T.sub.2] scores were conducted with
repeated-measures multiple analysis of variance controlling for variable's initial
([T.sub.1]) scores, children's age, and time between [T.sub.1] and [T.sub.2].

16
RESULTS

Sample

Among 75 families (102 children) screened within a year after the death, 52 (69%)
families with 75 children, aged 6 to 15 years, were eligible (Fig. 1). Twenty-seven
noneligible children had mood and anxiety disorders. Twenty-seven families (39 children),
including 19 (70%) families (28 children) who experienced the death of the children's
parent, were assigned to the intervention (Fig. 1). Twenty-five families (36 children),
including 15 (60%) families (22 children) who experienced the death of the children's
parent, were assigned to not receive the intervention. Table 1 shows comparisons of
demographic variables for 75 eligible children assigned to or not to receive the
intervention.

The 39 children (mean age = 9.6 [+ or -] 2.9 years) assigned to the intervention were
significantly younger at study entry than the 36 nonintervention children (mean age =
11.4 [+ or -] 3.5 years) ([t.sub.73] = 2.4, p [less than or equal to] .02). There were no
significant differences between the 39 or 36 children for other demographic variables or
methods of relatives' suicide, which included gunshot (37%), hanging (27%), overdose
(12%), jumping (10%), and other (14%). No children witnessed the suicide.

Nonintervention children had poorer initial social adjustment (mean [T.sub.1] SAICA = 1.7
[+ of -] 0.3) than children assigned to the intervention (mean [T.sub.1] SAICA = 1.5 [+ or
-] 0.2) ([t.sub.68] = 2.9, p [less than or equal to] .005) (Table 2).

Children older than 13 years had higher initial anxiety ([t.sub.34] = 5.0, p [less than or
equal to] .0001), depression ([t.sub.37] = 3.2, p [less than or equal to] .0003), and
posttraumatic stress than younger children ([t.sub.37] = 2.1, p [less than or equal
to] .04).

Significantly higher dropout rates occurred among nonintervention children (27, 75%)
than for children assigned to the intervention (7, 18%) ([[chi square].sub.1] = 24.6, p
[less than or equal to] .0001) (Fig. 1). Dropout for nonintervention children occurred
because families were too busy to schedule appointments (60%), did not want to talk
about the loss (20%), or sought intervention elsewhere (20%), such as pastoral or school
counseling or individual psychotherapy. These families were not willing to have an
outcome ([T.sub.2]) assessment. Dropouts of families assigned to the intervention
occurred after initial assessment while they were waiting to begin the intervention. These
families did not return for outcome ([T.sub.2]) assessments because they had difficulty
keeping appointments (55%) or did not want to talk about the death (45%). All retained
children assigned to the intervention completed the total number of intervention
sessions. All retained children completed [T.sub.2] assessments.

Retained children were representative of the eligible children assigned to or not to

receive the intervention as suggested by lack of significant differences in demographic
and psychosocial variables for the groups of assigned, retained, and dropout children.

The 32 children retained to receive the intervention (9.8 [+ or -] 3.0 years) were
significantly younger than the 9 retained nonintervention children (12.2 [+ or -] 3.3)
([t.sub.39] = 2.0, p [less than or equal to] .05) (Table 1). There were no significant
differences in other demographic variables between these two groups of children.
Retained children entered the study within the year after death. Eighteen (75%) families
retained to receive the intervention (23 children) and 4 (80%) retained nonintervention
families (8 children) had experienced death of the children's parent.

Nine retained nonintervention children had significantly higher initial depression (mean
CDI scores = 53.7 [+ or -] 11.8) and poorer social adjustment (mean SAICA scores = 1.9

17
[+ or -] 0.4) than the 32 children retained to the intervention (mean CDI scores = 46.5 [+
or -] 8.7, mean SACIA scores = 1.5 [+ or -] 0.3) ([t.sub.39] = 2.0, p [less than or equal
to] .05 for depression, [t.sub.9.6] = 2.3, p [less than or equal to] .05 for social
adjustment) (Table 2). These nine children did not receive any intervention in the
community.

Therapist Adherence

Therapist Performance Scale scores [less than or equal to]4 were identified for each
therapist for each intervention session, suggesting adequate leader adherence in
administering the intervention.

Efficacy of the Bereavement Group Intervention

Efficacy analyses controlled for effects of confounder variables involving children's entry
age and initial psychiatric symptom scores. Development of final ANCOVA efficacy
models were guided by fitting generalized additive models that used nonparametric
functions to estimate relationships between outcome and predictor variables. Results
suggested that age had a curvilinear relation to children's anxiety (RCMAS) and
depression (CDI) scores. Therefore, [age.sup.2] was a variable included to control for this
nonlinear relation. Since time from initial to outcome assessments ranged from 2.5
months to 4.5 months, efficacy analyses accounted for effects of this time interval.

Intervention efficacy was suggested by greater reduction in anxiety symptoms for

children who received the intervention compared with nonintervention children. Rates of
reduction of anxiety (RCMAS-[T.sub.2] minus RCMAS-[T.sub.1] divided by time from
[T.sub.1] to [T.sub.2]) were significantly greater for children who received the
intervention (mean rate RCMAS change = -0.08 [+ or -] 0.11) than for children who did
not (mean rate RCMAS change = 0.02 [+ or -] 0.03) (model [F.sub.4,31] = 3.7, p [less
than or equal to] .01, [R.sup.2] 0.3, BGI versus no BGI [1,31] = 5.2, p [less than or equal
to] .03, effect size = 0.6, RCMAS-[T.sub.1] [F.sub.1,31] = 8.2, p [less than or equal
to] .007, [T.sub.1] age [F.sub.1,31] = 3.0, p [less than or equal to] 0.9, [age.sup.2] [1,31]
= 3.6, p [less than or equal to] .07).

The reduction in children's anxiety (RCMAS-[T.sub.1] scores) was significantly greater for
children who received the intervention (mean RCMAS change = -0.2 [+ or -] 0.2) than for
children who did not (mean RCMAS change = 0.09 [+ or -] 0.1) (model [F.sub.5,30] = 4.3,
p [less than or equal to] .004, [R.sup.2] = 0.4, BGI versus no BGI [1,31] = 7.8, p [less than
or equal to] .009, effect size = 0.8, RCMAS-[T.sub.1] [F.sub.1,31] = 5.5, p [less than or
equal to] .03, [T.sub.1] age [F.sub.1,31] = 3.1, p [less than or equal to] .09, [age.sup.2]
[1,31] = 3.9, p [less than or equal to] .06, time [T.sub.1] to [T.sub.2] [F.sub.1,31] = 0.7, p
[less than or equal to] .4).

Intervention efficacy was indicated by lower outcome anxiety for children who received
the intervention (mean RCMAS-[T.sub.2] score = 39.6 [+ or -] 10.6) than for children who
did not (mean RCMAS-[T.sub.2] score = 56.5 [+ or -] 10.2) (model [F.sub.5,30] = 9.l, p
[less than or equal to] .0001, [R.sup.2] = 0.6, BGI versus no BGI [1,30] = 8.9, p [less than
or equal to] .006, effect size = 0.9, RCMAS-[T.sub.1] [1,30] = 7.9, p [less than or equal to]
.009, [T.sub.1] age [1,30] = 2.6, p [less than or equal to] .1, ag[e.sup.2] [1,30] = 3.4, p
[less than or equal to] .07, time [T.sub.1] to [T.sub.2] [1,30] = 1.0, p [less than or equal
to] .3) (Table 2 and Fig. 1). Figure 2, indicating examples of children's [T.ub.1] ages (i.e.,
6, 11, and 14 years) and RCMAS-[T.sub.1] scores (i.e., 40, 50,60,70), shows that children
who received the intervention had lower outcome anxiety ([T.sub.2]) than children who
did not receive the intervention and that younger children (i.e., 6-year-olds) and young
adolescents (i.e., 14-year-ol ds) had less reduction in anxiety than older school-age
children (i.e., 11-year-olds). These results reflected curvilinear relations between age and

18
children's anxiety.

Intervention efficacy was suggested by greater reduction in depressive symptoms for

children who received compared with children who did not receive the intervention. Rates
of reduction of depression (CDI-[T.sub.2] minus CDI-[T.sub.1] scores divided by time from
[T.sub.1] to [T.sub.2]) were significantly greater for children who received the
intervention (mean rate CDI change = -0.02 [+ or -] 0.06) than for nonintervention
children (mean rate CDI change 0.0009 [+ or -] 0.04) (model [F.sub.4,34] = 6.4, p [less
than or equal to] .0006, [R.sup.2] = 0.4, BGI versus no BGI [1,34] = 7.4, p [less than or
equal to] .01, effect size = 0.7, CDI-[T.sub.1] [F.sub.1,34] = 22.9, p [less than or equal to]
.0001, [T.sub.1] age [F.sub.1,34] = 8.4, p [less than or equal to] .007, ag[e.sup.2] [1,34]
= 9.0, p [less than or equal to] .005).

The reduction in children's depression (CDI-[T.sub.2] minus CDI-[T.sub.1] scores divided

by CDI-[T.sub.1] scores) was significantly greater for children who received the
intervention (mean CDI change = -0.05 [+ or -] 0.2) than for children who did not (mean
CDI change = 0.03 [+ or -] 0.2) (model [F.sub.5,33] = 6.4, p [less than or equal to] .0003,
[R.sup.2] = 0.5, BGI versus no BGI [1,33] = 5.4, p [less than or equal to] .03, effect size =
0.7, CDI-[T.sub.1] [F.sub.1,33] = 23.9, p [less than or equal to] .0001, [T.sub.1] age
[F.sub.1,33] = 11.6, p [less than or equal to] .002, time [T.sub.1] to [T.sub.2] [F.sub.1,33]
= 0.8, p [less than or equal to] .4).

Intervention efficacy was indicated by lower outcome depression for children who
received the intervention (mean CDI-[T.sub.2] score = 44.1 [+ or -] 8.7) than for children
who did not (mean CDI-[T.sub.2] score = 53.9 [+ or -] 7.8) (model [F.sub.5,33] = 9.0, p
[less than or equal to] .0001, [R.sup.2] = 0.6, BGI versus no BGI [1,33] = 5.3, p [less than
or equal to] .03, effect size = 0.7, CDI-[T.sub.1] [1,33] = 3.4, p [less than or equal to] .08,
[T.sub.1] age [1,33] = 9.6, p [less than or equal to] .004, ag[e.sup.2] [1,33] = 10.4, p
[less than or equal to] .003, time [T.sub.1]-[T.sub.2] [1,33] = 0.4, p [less than or equal to]
.6) (Table 2 and Fig. 3). Figure 3, indicating examples of children's [T.sub.1] ages (i.e., 6,
11, and 14 years) and CDI-[T.sub.1] scores (i.e., 40, 50, 60, 70), shows that children who
received the intervention had lower outcome depression ([T.sub.2]) than children who
had nor. Figure 3 illustrates that younger children (i.e., 6-year-olds) and young
adolescents (i.e., 14-year-aids) had less reduction in depression than older school-age
children (i.e., 11-year-olds). These results reflected curvilinear relations between age and
children's depression.

To evaluate the impact of violation of the independence assumption of siblings' scores,

the same models were fitted to include one randomly selected child per family. Results
did not change. Therefore, analyses are reported for all children. The bootstrap method
(Efron and Tibshirani, 1993), a nonparametric ANCOVA, was used to evaluate robustness
of the findings, which were shown to be robust. Finally, there were no significant
differences in outcome scores ([T.sub.2]) or rates of change for children's posttraumatic
stress (CPTSRI), children's social adjustment (SAICA), or parents' depression (BDI) for
those who had or had not received the intervention.

Within-group repeated-measures multiple analysis of variance controlling for outcome

variable's initial ([T.sub.1]) scores, children's age, ag[e.sup.2], and time between
[T.sub.1] and [T.sub.2] suggested that children who received the intervention had a
significant decrease from [T.sub.1] to [T.sub.2] of symptoms of anxiety ([F.sub.1,25] =
4.6, p [less than or equal to] .04, mean RCMAS-[T.sub.1] = 49.1 [+ or -] 10.0, mean
RCMAS-[T.sub.2] = 39.6 [+ or -] 10.6) and depression ([F.sub.1,26] = 10.6, p [less than or
equal to] .003, mean CDI-[T.sub.1] = 46.8 [+ or -] 8.7, mean CDI-[T.sub.2] = 44.1 [+ or -]
8.7) and a trend lowering of postraumatic stress symptoms ([F.sub.1,26] = 3.9, p [less
than or equal to] .06, mean CPTSRI-[T.sub.1] = 25.1 [+ or -] 12.4, mean CPTSRI-[T.sub.2]

19
= 19.6 [+ or -] 11.4). No differences were identified for nonintervention children.

DISCUSSION

This controlled study examined efficacy of a manual-based BGI to enhance psychosocial

adjustment of children after recent parental or sibling suicide. The screened, eligible,
assigned, and retained children were representative of children in the community who
suffered parental or sibling suicide. Similar to U.S. vital statistics, their deceased relatives
consisted of a greater number of white male suicides, predominantly by firearms. Similar
to reports of suicide-bereaved children (Brent et al., 1996; Cerel et al. 1999), the majority
of eligible children had no psychiatric disorders and had low levels of social adjustment
problems.

Intervention efficacy was suggested by the greater reduction in anxiety and depressive
symptoms for children receiving the intervention than for nonintervention children.
Among nonintervention children, outcome anxiety was greater and outcome depression
was similar to that at initial assessment. Children who received intervention and whose
preintervention anxiety and depression symptom scores were rated as clinically
significant (scores [greater than or equal to]63, [greater than or equal to]60,
respectively) had postintervention anxiety and depression scores below clinically
significant 1evels. Children's anxiety and depression were reduced after intervention
despite the persistence of posttraumatic stress symptoms or lack of change in social
maladjustment. Reports of traumatized children suggested that posttraumatic stress
symptoms prolong depressive symptoms (Goenjian et al., 1995). This was not evident
among children who received the intervention. Studies of children with major depressive
disorder sugges ted persistent social adjustment problems after resolution of depression
(PuigAntich et al., 1985). Perhaps social maladjustment of children in this study reflects
features of mourning, and improvement in social adjustment may occur in subsequent
phases of the bereavement process. Longer followup is needed to evaluate this.

Lack of distinctions in changes in depression between parents who did or did not receive
the intervention may be attributed to the main focus of the parental intervention, which
was to educate parents to assist their bereaved children rather than focus on parental
bereavement needs.

Study Limitations

This study highlighted recruitment and retention problems, despite extensive outreach
efforts. Challenges in recruiting bereaved families were reported (Brent et al., 1996; Cerel
et al., 1999). The use of alternating rather than random assignment of participants was
necessary because of the relatively small number of participants eligible at a particular
time and the need to avoid delay in starting the intervention. This method of assignment
may have created some biases, such as differences for age and time from death to study
entry among children assigned to receive or not receive the intervention. There was a
significantly greater dropout among nonintervention families (75%) than those who
received the intervention (18%). Nonintervention families felt too overwhelmed to
participate when not offered intervention. A more active control format may have
facilitated maintaining families. Those who were assigned to intervention and dropped
out, did so before beginning the intervention because they were overwhelm ed by talking
about the deceased. The eligible, assigned, and completer children were representative
of suicide-bereaved children in that their deceased relatives were predominantly white
males with firearm deaths.

Data analyses were compromised by the small retained sample size. Because children
who dropped out did not return for an outcome assessment, only data for children who
completed the study could be analyzed. The impact of violation of the independence

20
assumption of multiple siblings per family on ANCOVA was examined. Although mixed-
effects models (Gibbons et al., 1993) allowed for modeling correlated outcomes by
choosing the family as the unit of analysis, numerical solutions of mixed-effects models
were not attainable because of the limited number of families with multiple children.
Another method to evaluate outcomes with multiple children per family was to average
the outcome within each family. This method introduced biases because ages of children
within families varied and were associated with the outcome variables. Thus sensitivity
analyses that randomly selected one child per family were conducted to evaluate effects
of codependence of outcomes of multiple children in families. Results suggested simil ar
findings to those including all children.

Since children with psychiatric disorders were excluded, research is needed to evaluate
the intervention for children whose co-occurring psychiatric disorders are treated
concurrently with other interventions.

Clinical Implications

A targeted BGI is a strategy for reducing suicide-bereaved children's psychosocial distress

and potentially preventing morbidities. Reported developmental variations in children's
grief (Christ, 2000) were supported by this study's observations of persistence of higher
anxiety and depressive symptoms in younger children and young adolescents.
Developmental understanding of children's grief can guide evaluations, treatment
planning, and research with bereaved children.

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

TABLE 1

Demographic Characteristics of Assigned Children and Those Retained

Assigned

Intervention
Variable (n = 39 children)

Age (yr): mean [+ or -] SD 9.6 [+ or -] 2.9 (a)

Gender: n(%)

Male 16 (41.0)
Female 23 (59.0)

Race/ethnicity: n(%)

White 28 (90.9)
African American 7 (9.1)
Hispanic 4 (0)

Social status (c): n(%)

I 6 (15.4)
II 11 (28.2)
III 15 (38.5)
IV 6 (15.4)
V 1 (2.5)
Time from death to initial assessment
(days): mean [+ or -] SD 306 [+ or -] 376
No. of children/family: mean 1.6 [+ or -] 0.64
[+ or -] SD

Assigned

No Intervention
Variable (n = 36 children)

21
Age (yr): mean [+ or -] SD 11.4 [+ or -] 3.5

Gender: n(%)

Male 12 (33.3)
Female 24 (66.7)

Race/ethnicity: n(%)

White 24 (66.6)
African American 5 (14.0)
Hispanic 7 (19.4)

Social status (c): n(%)

I 5 (13.9)
II 6 (16.7)
III 12 (33.3)
IV 7 (19.4)
V 6 (16.7)
Time from death to initial assessment
(days): mean [+ or -] SD 515 [+ or -]1,008
No. of children/family: mean 1.3 [+ or -] 0.6
[+ or -] SD

Retained

Intervention
Variable (n = 32 children)

Age (yr): mean [+ or -] SD 9.8 [+ or -] 3.0 (b)

Gender: n(%)

Male 13 (40.6)
Female 19 (59.4)

Race/ethnicity: n(%)

White 25 (78.1)
African American 5 (15.6)
Hispanic 2 (6.3)

Social status (c): n(%)

I 6 (18.8)
II 9 (28.1)
III 11 (34.4)
IV 5 (15.6)
V 1 (3.1)
Time from death to initial assessment
(days): mean [+ or -] SD 266.2 [+ or -] 326.1
No. of children/family: mean 1.5 [+ or -] 0.7
[+ or -] SD

Retained

No Intervention
Variable (n = 9 children)

Age (yr): mean [+ or -] SD 12.2 [+ or -] 3.3

Gender: n(%)

Male 5 (55.6)
Female 4 (44.4)

Race/ethnicity: n(%)

White 8 (88.9)
African American 1 (11.1)
Hispanic 0 (0)

Social status (c): n(%)

I 0 (0)
II 2 (22.2)
III 2 (22.2)
IV 3 (33.4)
V 2 (22.2)
Time from death to initial assessment
(days): mean [+ or -] SD 373.1 [+ or -] 485.5

22
No. of children/family: mean 1.6 [+ or -] 0.9
[+ or -] SD

(a)Assigned to intervention versus no intervention: [t.sub.73] = 2.4, p

[less than or equal to] .02.

(b)Received intervention versus no intervention: [t.sub.39] = 2.0, p

[less than or equal to] .05.

(c)Social status reference: Hollingshead and Redlich, 1958.

TABLE 2

Psychosocial Variables of Assigned Children and Parents and Those

Retained

Assigned

Intervention: No Intervention:
Variable [T.sub.1] [T.sub.1]

Anxiety 36, 49.5 [+ or -] 9.6 28, 51 [+ or -] 10.1

Depression 37, 46.8 [+ or -] 8.9 27, 51.7 [+ or -] 13.1
Posttraumatic stress 36, 25.3 [+ or -] 12.2 21, 28.9 [+ or -] 13.6
Social adjustment 39, 1.5 [+ or -] 0.2 (a) 31, 1.7 [+ or -] 0.3
Parent depression 37, 14.7 [+ or -] 8.3 20, 15.4 [+ or -] 12.0

Retained

Intervention
Variable [T.sub.1]

Anxiety 31, 49.3 [+ or -] 9.9

Depression 32, 46.5 [+ or -] 8.7 (b)
Posttraumatic stress 31, 25.1 [+ or -] 12.4
Social adjustment 32, 1.5 [+ or -] 0.3 (c)
Parent depression 32, 14.6 [+ or -] 8.7

Retained

Intervention No Intervention
Variable [T.sub.2] [T.sub.1]

Anxiety 30, 39.6 [+ or -] 10.6 (d) 8, 52.6 [+ or -] 6.5

Depression 31, 44.1 [+ or -] 8.7 (c) 9, 53.7 [+ or -] 11.8
Posttraumatic stress 31, 19.6 [+ or -] 11.4 9, 22.1 [+ or -] 7.0
Social adjustment 32, 1.6 [+ or -] 0.2 9, 1.9 [+ or -] 0.4
Parent depression 32, 11.1 [+ or -] 10.5 8, 14.9 [+ or -] 9.9

Retained

No Intervention
Variable [T.sub.2]

Anxiety 6, 56.5 [+ or -] 10.2

Depression 8, 53.9 [+ or -] 7.8
Posttraumatic stress 8, 17.8 [+ or -] 9.1
Social adjustment 9, 1.8 [+ or -] 0.4
Parent depression 7, 9.7 [+ or -] 4.5

Note: Values represent n, mean [+ or -] SD.

(a)[T.sub.1] assigned to intervention versus no intervention: [t.sub.68]

= 2.9, p [less than or equal to] .005.

(b)[T.sub.1] retained to intervention versus no intervention: [t.sub.39]

= 2.0, p [less than or equal to] .05.

(c)[T.sub.1] retained to intervention versus no intervention:

[t.sub.9.6] = 2.3, p [less than or equal to] .05.

(d)[T.sub.2] retained to intervention versus no intervention: [t.sub.34]

= 3.6, p [less than or equal to] .001.

(e)[T.sub.2] retained to intervention versus no intervention: [t.sub.37]

= 2.9, p [less than or equal to] .006.
Accepted December 18, 2001.

REFERENCES

23
Ambrosini PJ, Dixon JF (1996), Schedule for Affective Disorders and Schizophrenia for
School-Age Children (K-SADS-IVR)-Present State and Epidemiological Version. Medical
College of Pennsylvania and Hahneman University, Allegheny University of Health
Sciences, 3200 Henry Avenue, Philadelphia, PA 19129-1137

Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J (1961), An inventory for measuring
depression. Arch Gen Psychiatry 4:561--571

Bowlby J (1980), Attachment and Loss: Vol 3. Loss: Sadness and Depression. London:
Hogarth

Brent DA, Moritz G, Bridge J, Perper J, Canobbio R (1996), The impact of adolescent
suicide on siblings and parents: a longitudinal follow-up. Suicide Life Threat Behav
26:253--259

Cerel J, Fristad MA, Weller EB, Weller RA (1999), Suicide-bereaved children and
adolescents: a controlled longitudinal examination. J Am Acad Child Adolesc Psychiatry
38:672--679

Christ GH (2000), Healing Children's Grief: Surviving a Parent's Death From Cancer New
York: Oxford University Press

Cohen J (1960), A coefficient of agreement for nominal scales. Educ Psychol Meas 20:37--
46

Efron B, Tibshirani RJ (1993), An Introduction to the Bootstrap. New York: Chapman and
Hall

Elkin IE, Parloff MB, Hadley SW, Autry AH (1985), NIMH Treatment of Depression
Collaborative Research Program: background and treatment plan. Arch Gen Psychiatry
42:305-316

Gibbons RD, Hedeker D, Elkin I et al. (1993), Some conceptual and statistical issues in
analysis of longitudinal psychiatric data. Arch Gen Psychiatry 50:739-750

Goenjian AK, Pynoos RS, Steinberg AM et al. (1995), Psychiatric comorbidity in children
after the 1998 earthquake in Armenia. J Am Acad Child Adolese Psychiatry 34:1174-1184

Hollingshead AB, Redlich P (1958), Social Class and Mental Illness. New York: Wiley

John K, Gammon GD, Prusoff BA, Warner V (1987), The Social Adjustment Inventory for
Children and Adolescents (SAICA): testing of a new semi-structured interview. J Am Acad
Child Adolesc Psychiatry 26:898-911

Kaufman J, Birmaher B, Brent D et al. (1997), Schedule for Affective Disorders and
Schizophrenia for School-Age Children-Present and Lifetime Version (K-SADS-PL): initial
reliability and validity data. J Am Acad Child Adolesc Psychiatry 36:980-988

Kovacs M (1992), Children's Depression Inventory Manual North Tonawanda, NY: Multi-
Health Systems

Lazarus RS, FolkmanS (1984), Stress, Appraisal, and Coping. New York Springer

Ness D, Pfeffer CR (1990), Sequelac of bereavement resulting from suicide. Am J

24
Psychiatry 147:279-285

Osterweiss M, Solomon F, Greene M, eds (1984), Bereavement: Reactions, Consequences.

and Care. Washington, DC: National Academy Press

Parks CM (1996), Bereavement: Studies of Grief in Adult Life. New York: Penguin

Pfeffer CR, Karus D, Siegel K, Jiang H (2000), Child survivors of parental death from
cancer or suicide: depressive and behavioral outcomes. Psychooncoloy 9:1-10

Pfeffer CR, Martins P. Mann J et al. (1997), Child survivors of suicide: psychosocial
characteristies. J Am Acad Child Adolesc Psychiatry 36:65-74
Puig-Antich J, Lukens E, Davies M, Goetz D, Brennan-Quatrrock J, Todak G (1985),
Psychosocial functioning in prepubertal major depressive disorders, II: interpersonal
relationships after sustained recovery from affective episodes. Arch Gen Psychiatry
42:511-517

Pynoos RS, Frederick C, Nader K et al. (1987), Life threat and post-traumatic stress in
school-age children. Arch Gen Psychiatry 44:1057-1063

Reynolds CR, Richmond BO (1985), Revised Children's Manifes Anxiety Scale (RCMAS)
Manual Los Angeles: Western Psychological Services

Shepherd DM, Barraclough BM (1976), The aftermath of parental suicide for children. Br J
Psychiatry 129:267-276

Stroebe MS. Stroebe W, Hansson RO (1993), Handbook of Bereavement: Theory,

Research, and Intervention. New York Cambridge University Press

Weller RA, Weller EB, Fristad MA, Bowes JM (1991), Depression in recently bereaved
prepubertal children. Am J Prychiatry 148:1536-1540

Worden JW, Silverman PR (1996), Parental death and the adjustment of school age
children. Omega J Death Dying 33:91-102

Dr. Pfeffer is Professor of Psychiatry Dr. Kakuma is Associate Research Professor of

Biostatistics in Psychiatry Ms. Jiang is Data Analyst, Ms. Hwang and Ms. Metsch are staff
of else Childhood Bereavement Program at Weill Medical College of Cornell University,
New York.

This study was supported by Nanette L. Laitman and the William and Mildred Lasdon
Foundation, a fund established in The New York Community Trust by DeWitt-Wallace the
William T. Grant Foundation, the Klingenstein Third Generation Foundation, and the Rodd
D. Brickell Foundation. The authors thank Charles Hirsch, M.D., and Milard Hyland M.D.,
Chief Medical Examiners of New York City and Westchester County, respectively; David
Schomburg Director of Medicolegal Investigations, Chief Medical Examiner Office, New
York City, and Sylvia Martin, Coordinator of Medical Records, Chief Medical Examiner
Office, Westchester County, for their collaboration; and Tamara Jachimowicz, B.A., for
research assistance.

Reprint requests to Dr. Pfeffer, New York Presbyterian Hospital, 21 Bloomingdale Road,
White Plains, NY 10605.
COPYRIGHT 2002 Lippincott/Williams & Wilkins
in association with The Gale Group and LookSmart. COPYRIGHT 2002 Gale Group

Journal of the American Academy of Child and Adolescent Psychiatry

25
March, 2002

Children's responses to low parental mood; II: Associations with family

perceptions of parenting styles and child distress.(Statistical Data Included)

Author/s: Tytti Solantaus-Simula

ABSTRACT

Objective: In an earlier article (part I) the authors identified four patterns of children's
responses to parental low mood: Active Empathy, Emotional Overinvolvement,
Indifference, and Avoidance. They then hypothesized that these response patterns were
related to parenting styles and to discrepancies in family members' perceptions of
parenting and child mental distress.

Method: A normal population sample of 990 twelve-year-old Finnish children and their
mothers (843) and fathers (573) was used. Within-family multivariate analyses conducted
in mother-father-child triads (470) were used to examine whether quality of parenting
varied according to children's responses and whether parents' and children's perceptions
of parenting and child distress were different.

Results: Children in the Active Empathy and Indifference groups experienced more
positive parenting than those in the other two groups. Discrepancies in family members'
perceptions of child distress and mothering and fathering were especially characteris tic
of the Emotional Overinvolvement group. Typical for the Avoidance group was a within-
family agreement on poor parenting and severe child distress.

Conclusions: Children's response patterns as regards parental low mood are related to
family dynamics. The study suggests that discrepancies in parents' and children's
perceptions of parenting and child distress can be meaningful in understanding family
interactions and child development and well-being. J. Am. Acad. Child Adolesc. Psychiatry,
2002, 41(3):287-295.

Key Words: parental low mood, children's responses, parenting, within-family analyses.

In an initial report of children's responses to low parental mood (Solantaus-Simula et al.,

2002), we identified four response patterns in a community ample. The Active Empathy
group made efforts to console and help their parents. The Emotional Overinvolvement
group were also prosocial but were also highly affected by negative emotions. The
Indifference group seemed to have little awareness of parental moods and showed low
empathy and prosocial behavior. The Avoidance group showed active avoidance and little
involvement in parental low moods. The response groups also differed in psychological
adaptation: the Emotional Overinvolvement and Avoidance groups showed higher levels
of emotional and behavioral symptoms than the two other groups. Thus the children's
responses to parental low mood could be distinguished by the levels of prosocial
involvement and symptoms of distress.

We were interested in the dynamics that might characterize families in the response
groups and contribute to the understanding of children's adaptive and nonadaptive
responses. Research shows that parental depression is a risk factor for children's
depression and other psychiatric problems (Beardslee et al., 1998). However, it seems
that it is not the symptoms of depression themselves but their effects on parenting and
family communication that contribute to the transmission of depression (Beardslee and
Podorefsky, 1988; Coyne et al., 1988; Hops et al, 1987). Parenting styles have been
documented to also have an influence on prosocial development (Eisenberg and Mussen,
1997). This prompted us to look first at parenting and then at family communication.

26
There is evidence that parental depression is mediated to children in coercive, impulsive
parenting and reduced negotiations and warmth (Davies and Windle, 1997; Jain et al.,
1996). Thus we hypothesized that children's responses to parental low mood varied with
differences in parenting styles. More punitive and less authoritative parenting is expected
in the symptomatic response groups, i.e., the Emotional Overinvolvement and the
Avoidance groups.

It has also been documented that punitive parenting is related to less empathy and
caring in children (Bryant and Crockenberg, 1980; Hoffman, 1963; Koestner et al., 1990;
Olweus, 1980) and authoritative parenting with prosocial behavior toward others
(McFarlane et al., 1995; Zahn-Waxler et al., 1979). Although these studies have focused
on the impact of parenting on children's prosocial behaviors to others outside the family,
we expected the results also to apply to family relationships. This means that the parents
of the Avoidance and Indifference groups would use more punitive and less authoritative
parenting than those of the more prosocial groups, i.e., the Active Empathy and
Emotional Overinvolvement.

These hypotheses are contradictory concerning the parenting in the Emotional

Overinvolvement group. The parents are expected to be punitive and less authoritative,
because poor parenting is typical among symptomatic children, but they are also
expected to be more authoritative and less punitive, because their children showed high
prosocial behavior. Hence, it may well be that parenting is highly discrepant in these
families; one parent might be very authoritative, and the other more punitive.

We then turned to family communication. Earlier research has pointed out how false
beliefs and miscommunications are characteristic in depressed families (Beardslee and
Podorefsky, 1988; Downey and Coyne, 1990; Keitner and Miller, 1990) and that they may
even contribute to the negative impact parental depression has on child mental health
(Beardslee and Podorefsky, 1988). This means that family members might misinterpret
and, therefore, hold discrepant views of each other's experiences and behaviors. Children
especially have difficulties in understanding their depressive parents' behavior if family
members do not communicate about their problems (Beardslee and Podorefsky, 1988).
This might lead to avoidance of or anxious overinvolvement with the parent on the part of
children. Therefore, we hypothesized that there would be discrepant views between
parents and children concerning parenting behaviors, especially in the Emotional
Overinvolvement and the Avoidance groups.

It has also been documented that children's mental problems go unrecognized in families
with depressed parents (Beardslee et al., 1998; Cummings and Davies, 1993), which
could lead to discrepancies in parental and child reports of child distress. This might also
explain why, in the initial report of this study (Solantaus-Simula et al., 2002), children's
response patterns were associated only with their own reports of emotional and
behavioral problems, but not with their parents'. Therefore, we were interested in
possible discrepancies in children's and their parents' perceptions of children's emotional
and behavioral problems and expected these to occur, especially in the symptomatic
response groups.

On the other hand, earlier research has shown that children's and parents' perceptions
are quite discrepant even in normative samples, including reports on children's
behavioral and emotional distress (Achenbach et al., 1987; Seiffge-Krenke and Kollmar,
1998) and parents' perceptions on mothering and fathering (Deal et al., 1989). There is,
overall, very little systematic research on children's and parents' perceptions of each
other and what the possible discrepancies might mean for children.

Taken together, the following three hypotheses were examined:

27
1. Children's responses will be related to parenting styles.

2. Children's responses will be related to discrepant parenting behaviors between

mothers and fathers in the Emotional Overinvolvement group.

3. Children's responses will be related to discrepancies in family members' perceptions of

parenting and child distress, especially in the nonadaptive groups, i.e., the
Overinvolvement and the Avoidance groups.

METHOD

Subjects

This was a cross-sectional study that included 990 Finnish school children, 502 girls and
488 boys. The description of the total sample and the study procedure is given in part I of
this study (Solantaus-Simula et al., 2002). For the within-family analyses, we used a
subset of the total sample composed of only full triads (the mother, the father and the
index child), which were 470 in number. This means that there are no single-parent
families in these analyses. The children's mean age was 12.6 years (SD = 0.03).
Compared with the full sample, the triads did not differ in child mental health, but they
were better off socioeconomically ([chi square] [30,836] = 6l.29, p = .001) than the rest
of the sample.

Measures

The children's responses to their mothers' and fathers' low moods were elicited in two
questions tapping into emotional and behavioral responses. The cluster analysis revealed
four clusters: the Indifference, Active Empathy, Emotional Overinvolvement and
Avoidance patterns. (For details see Solantaus-Simula et al., 2002.)

Child distress was measured by externalizing and internalizing symptoms reported by

children, fathers, and mothers (the Youth Self-Report and the Child Behavior Checklist)
(Achenbach, 1991) and by depressive symptoms reported by children (Children's
Depression Inventory) (Kovacs, 1992). (For details see Solantaus-Simula et al., 2002.)

The quality of parenting was assessed on the Behavior and Affect Rating Scale: Young
Adult Perception of Parent Hostility and Warmth and the Child Rearing Practices Scale.
These measures were developed for the Iowa Youth and Families Project in Iowa
University. They have been widely used in the Project (e.g., Conger et al., 1994; Simons
et al., 1992), although they have not been validated.

The original scale for parents was considered too long for the multifaceted questionnaire,
and it was shortened from 27 to 20 items. The mothers and the fathers indicated how
often they expressed different feelings to and discussed different things with the target
child and how they disciplined the child. The children indicated on two identical scales
how they perceived their mother's and father's parenting, e.g., listened to their opinion,
acted in loving way, or became angry with them. Both parents and children answered the
questions on a 5-point scale (1 = extremely often, 5 = never). Missing values were
replaced by the variable mean. The responses were then recoded so that the highest
value (5) reflected optimal parenting. Eight items constituted authoritative ([alpha] = .85
for the mothers and the fathers) and punitive ([alpha] = .63 for the mothers and .61 for
the fathers) mothering and fathering. From the child-reported parenting scale, 8 items
constituted authoritative ([alpha] = .88 For mothering and .90 for fathering), and 5 items
constituted punitive ([alpha] = .80 for mothering and .83 for fathering). The psychometric
properties of the Finnish scales are reported elsewhere (Leinonen et al., in press).

28
Statistical Analyses

To test the first hypothesis, between-subject multivariate analyses of variance

(MANOVAs) with univariate analyses were use to study the association between children'
response patterns and parenting quality separately among children (N = 990), mothers
(N = 843) and fathers (N = 573). To test the second and third hypotheses, about the
discrepant perceptions of parenting and child distress, GLM repeated measures
(MANOVAs) (SPSS10 program) were applied in the sample including only the triads (N =
470). The dependent variables were the scores of each family member on authoritative
and punitive mothering and fathering and children's internalizing and externalizing
symptoms. The dependent variables were standardized (mean = 0, SD = 1) in order to
allow comparisons between them. Orthogonal contrast analyses were conducted to
detect specific hypothesized comparisons between the parent- and child-reported
variables within each response pattern (Rosenthal et al., 2000).

RESULTS

Children's Responses and Parenting Styles

To examine whether the quality of parenting varied according to the children's response
patterns, 4 (response patterns) X 2 (child gender) between-subject MANOVAs were
separately applied on the children's perceived parenting (authoritative and punitive
mothering and fathering), and self-reported mothering and fathering (authoritative and
punitive parenting).

The results showed significant associations of response patterns (Wilks [lambda],

[F.sub.12,2569.32] = 12.05, p < .0001) and child gender (Wilks [lambda], [F.sub.6,948] =
3.54, p < .002) with child-perceived parenting. Subsequent ANOVAs were significant only
for response patterns and are shown in Table 1. Tukey-b post hoc analyses showed that
the children in the Emotional Overinvolvement and the Avoidance groups perceived their
mothers as more punitive than the children in the Indifference and Active Empathy
groups. The Avoidance group perceived their mothers also as less authoritative than the
children in the other groups.

The post hoc comparisons showed somewhat different results for child-perceived
fathering. Children in the Avoidance group regarded their fathers as the least
authoritative, those in the Active Empathy group regarded their fathers as the most
authoritative, and those in the Emotional Overinvolvement and Indifference groups fell in
the middle. The Emotional Overinvolvement group reported their fathers to be highly
punitive, whereas the Indifference group reported the lowest level of punitive fathering.
The Active Empathy and Avoidance groups were in the middle.

A MANOVA for the self-reported mothering scores showed a significant interaction effect
between response patterns and the child's gender (Roy's largest root, [F.sub.3,715] =
3.84, p < .01), and an association with response patterns (Wilks [lambda],
[F.sub.9,1735.40] = 2.l2,p p < .02). Post hoc comparisons showed that self-reported
authoritative mothering was high in both the Emotional Overinvolvement and Active
Empathy groups, as compared with the Avoidance group. The interaction effect revealed
that this was especially true concerning girls. As far as self-reported fathering was
concerned, the MANOVA revealed only an association with gender (Wilks [lambda],
[F.sub.3,489.00] = 5.56, p < .001). The fathers reported that they were more punitive
with their daughters than with their sons ([F.sub.1,491] = ll.49, p < .001).
To conclude, the children's response patterns were associated with parenting as
experienced by the child and reported by the mother, but not by the father. Both the
children and the mothers in the Avoidance group reported low authoritative and high
punitive mothering. In the Emotional Overinvolvement group, in turn, children perceived

29
the mothers and fathers as highly punitive, and the mothers reported high authoritative
parenting.

Within-Family Perceptions of Parenting

We hypothesized that within-family discrepancies would occur, especially in the

symptomatic response groups, i.e., in the Emotional Overinvolvement and Avoidance
groups. Within-subject MANOVA models with focused contrast interaction effects were
applied to locate the discrepancies. Separate models were constructed for mothering
(child- and mother-reported authoritative and punitive parenting as dependent variables)
and for fathering (child- and father-reported authoritative and punitive parenting as
dependent variables). The means of parenting and significant focused contrasts between
family members (informants) are illustrated in Figures 1 and 2. (It should be noted that
for punitive parenting, the higher scores indicate less punitive behavior.)

A significant response pattern X family member interaction effect was found for
mothering (Wilks [lambda], [F.sub.9,1728.10] = 4.04, p < .0001). Figure 1 shows
significant contrasts in the symptomatic groups: the children in both the Emotional
Overinvolvement and the Avoidance groups perceived their mothers less authoritative
(contrast values were p = .01 in both groups) and more punitive (contrast values were p
= .001 in both groups) than the mothers themselves. No significant focused contrasts
were found in the Indifference and Active Empathy groups.

A significant response pattern X family member interaction effect was also found for
fathering (Wilks [lambda], [F.sub.9,1728.10] = 4.84, p < .0001). Figure 2 shows
significant contrasts in both the symptomatic and nonsympromatic groups. Children in
the Emotional Overinvolvement group perceived their fathers as more punitive than the
fathers themselves (p = .04). In contrast, the Indifference group perceived their fathers
as less punitive (p .04), and the Active Empathy children more as authoritative, than the
fathers themselves (p = .01) group. No significant contrasts were found in the Avoidance
group.

Finally, a within-subject MANOVA with focused contrasts was conducted to test the
discrepancies between fathering and mothering (parent-reported authoritative and
punitive mothering and fathering as dependent variables). A significant response pattern
X parent interaction effect was found (Roy's largest root, [F.sub.3,472] = 2.80, p <.04).
The mothers of the Emotional Overinvolvement group reported a higher levels of
authoritative (p .04) and nonpunitive (p = .02) parenting than the fathers. No significant
contrasts were found in the other groups.

Within-Family Perception of Child Distress

We hypothesized that discrepancies in reports on child distress would occur, especially in

the symptomatic response patterns. Within-subject MANOVA models with focused
contrast interaction effects were applied to locate the discrepancies. Separate models
were tested for children's internalizing and externalizing symptoms, including both child-
reported and mother- and father-perceived symptoms scores as dependent variables.

Figure 3 presents the results for externalizing symptoms (response patterns X family
member interaction effect was significant: Wilks [lambda], [F.sub.9,1131.84] = 2.28, p
< .02). The children in the Emotional Overinvolvement group reported more symptoms
than both their fathers (p = .003) and mothers (p = .03) perceived. In contrast, children
in the Indifference group reported less externalizing symptoms than their fathers
perceived (p = .02). The children and the parents in the Active Empathy group agreed on
low levels, and those in the Avoidance groups on high levels, of externalizing symptoms
in children.

30
A significant response patterns X family member interaction effect was found also for
internalizing symptoms (Wilks [lambda] [F.sub.9,1131.84] = 2.96, p < .002). Figure 4
shows significant contrasts both in the Emotional Overinvolvement and Indifference
groups. In the first group children reported more internalizing symptoms than their
mothers (p = .00 1) and fathers (p = .00 1) perceived. In the latter group, the children
reported less internalizing symptoms than their mothers (p = .01) and fathers (p = .002)
perceived. The family members in the Active Empathy and Avoidance groups did not
show any contrasts between their distress perceptions.

DISCUSSION

Children's responses to their parents' low moods were related to family interactions as
indicated by the family members' perceptions of parenting and child distress. As an
overall pattern, the children in the Active Empathy and the Indifference (the
nonsymptomatic) groups experienced more positive parenting than those in the
Emotional Overinvolvement and Avoidance (the symptomatic) groups. Our study joins
with others showing an association between parenting and children's behaviors and
attitudes.

We expected to find discrepancies in parents' and children's perceptions, especially in the

families belonging to the symptomatic (Emotional Overinvolvement and Avoidance)
groups. The picture turned out to be more complex. There were experiential and
perceptual discrepancies in both the symptomatic and nonsymptomatic groups of
children. However, these discrepancies were not random, and a closer look at the
symptomatic groups on one hand, and the nonsymptomatic on the other, revealed
interesting dynamics and hypotheses for further research.

Comparison of the nonsymptomatic groups revealed that the children in the Indifference
group, on one hand, systematically deemphasized negative aspects of themselves and of
their patents: they reported less paternal punitiveness and less externalizing and
internalizing symptoms than the parents did. They also reported less parental low mood
than other children. The Active Empathy group, on the other hand, if they disagreed with
their parents, seemed to emphasize positive aspects of their parents: they reported more
supportive fathering that the fathers did. A major difference between the responses of
these two groups was in feelings of empathy and prosocial behavior toward the parents:
the Active Empathy group excelled over the Indifference group. Our measures were
limited regarding intrapsychic processes, but the data suggest that the Indifference group
might use a certain level of denial of hardships in their lives. This might also reflect a
family pattern, because their parents, on their part, might activ ely hide their moods from
the children. This family pattern is not associated with psychological symptoms, but it
might be an obstacle to prosocial development.

Comparison of the two symptomatic groups revealed high familial discrepancies in the
Emotional Overinvolvement group and consistent negative perceptions of fathering and
child well-being in the Avoidance group. The children in the Emotional Overinvolvement
group reported higher levels of both internalizing and externalizing problems than their
parents did. The children either overemphasized their own problems, or the parents failed
to appreciate their children's distress. The children also experienced both mothering and
fathering more negatively than the parents did. As hypothesized, substantial differences
in mothers' and fathers' parenting styles were also characteristic of this response group,
as the mothers reported significantly more authoritative, and fathers more punitive,
parenting than the other parent. Overall, the family members' perceptions of parenting
ranged from very positive to very negative in the Emotional Overinvolvement group.

These comprehensive discrepancies in family members' experiences imply false beliefs

and miscommunications between parents and children and a lack of joint reality in the

31
Emotional Overinvolvement group. This is likely to be both emotionally and cognitively
very demanding for children and could undermine their sense of security. It helps to
explain the diversity and intensity of their responses to parental low mood, as they
ranged from empathetic concern to feeling low, guilty, frightened, and angry.

The situation was different in the Avoidance group. There was little discrepancy in the
perceptions of the children and the parents. According to all three family members,
parenting was dominated by negative emotionality, and the children had high levels of
externalizing and internalizing symptoms. These children responded to their parents'
sadness with avoidance and lack of empathy, which might mirror their own family
experiences. They also had difficulties describing their feelings and behaviors, indicating
problems in metacommunication skills. This could be related to the low levels of mutual
problem solving and supportive discussions in their families.

In this study, family members disagreed on the level of children's mental distress in two
of the four response groups: the Indifference group reported fewer problems, and the
Emotional Overinvolvement group more problems, than their parents did. Earlier research
has also documented discrepant parental and child reports of child distress (Achenbach
et al., 1987). The reasons for and implications of this are still unknown, and it calls for
more research on children's own experiences of family life. It has been generally
accepted that studies assessing children's adjustment should gather information from
different sources. This has been related to efforts to find the more objective truth: the
more the informants agree, the closer to the truth the assessment is supposed to be. Our
findings offer another view: the discrepancies might be illustrative and important as
regards family dynamics and children's well-being. This is an intriguing hypothesis for
future research.

Limitations

We applied here a within-family analysis with focused contrasts for comparing family
members' perceptions and experiences. In earlier research, within-person correlations
(Seiffge-Krenke and Kollmar, 1998) and clustering procedures (O'Connor et al., 1998)
have been applied. Hops and colleagues (1987) applied a method similar to ours and
based their comparisons of family interactions on repeated measures MANOVAs with
planned orthogonal comparisons. Our research setting focused on associations between
categorical (response patterns) and linear (parenting and mental distress) variables, and
ANOVA was a natural choice. Within-family analyses involve a complex methodology, and
future studies should attempt to replicate the findings concerning the dynamics of
discrepant family perceptions and contrasting mothering and fathering profiles.

This study was cross-sectional, and the reported associations are likely to be
bidirectional. Although it is evident that children respond to their parents' moods, the
parents are likely to be influenced by their children's responses as well. For instance,
children's overinvolvement in parental affairs might induce both protective and irritable
responses from the parents, and children's active avoidance might foster withdrawal in
the parents. Parent--child relationships are likely to be transactional; therefore,
longitudinal studies are needed.

The measure for children's responses to parental low mood was developed for the
purpose of the study, and the parenting measures were not validated. These limit the
generalizability of the findings, and the results must be replicated in other samples.

Clinical Implications

Our study joins with Beardslee and Podorefsky (1988) in pointing out how a joint
perception of reality within the family is beneficial for child development. Very discrepant

32
views of parents and children suggest communication problems and might undermine
children's sense of security and compromise their development. However, a joint reality
is, of course, not enough. In the Avoidance group, the families very much agreed on their
perceptions of each other, but the family atmosphere was negative. This was related to
both children's mental distress and low prosocial behavior.

Our results emphasize the importance of family dynamics in children's behavior and
attitudes and support the use of family approaches in clinical interviews, evaluations, and
diagnostic procedures, as well as in research. Understanding parental and child distress
in the transactional context acknowledges children as active and influential members of
the family.

REFERENCES

Achenbach TM (1991), Integrative Guide for the 1991 CBCL/4-18, YSR, and TRF Profiles.
Burlington: University of Vermont Department of Psychiatry

Achenbach TM, McConoughy SH, Howell CT (1987), Child/adolescent behavioral and

emotional problems: implications of cross-informant correlations for situational
specificity. Psychol Bull 101:213-232

Beardslee WR, Podorefsky D (1988), Resilient adolescents whose parents have serious
affective and other psychiatric disorders: the importance of self-understanding and
relationships. Am J Psychiatry 145:63-69
Beardslee WR, Versage EM, Gladstone TR (1998), Children of affectively ill parents: a
review of the past 10 years. J Am Acad Child Adolesc Psychiatry 37:1134-1141

Bryant BK, Crockenberg SB (1980), Correlates and dimensions of prosocial behavior: a

study of female siblings with their mothers. Child Dev 51:529-544

Conger RD, Ge X, Elder GH Jr. Lorenz FO, Simons RL (1994), Economic stress, coercive
family processes, and developmental problems of adolescents. Child Dev 65:541-561

Coyne JC, Kahn J, Gotlib IH (1988), Depression. In: Family Interaction and
psychopathology: Theories, Methods, and Findings. Applied Clinical Psychology, Theodore
J, ed. New York: Plenum, pp 509-533

Cummings ME, Davies PT (1993), Maternal depression and child development. J Child
Psychol Psychiatry 35:73-112

Davies PT, Windle M (1997), Gender-specific pathways between maternal depressive

symptoms, family discord, and adolescent adjustment. Dev Psychol 33:657-668

Deal J, Halverson C, Wampler K (1989), Parental agreement on child-rearing orientations:

relations to parental, marital, family and child characteristics. Child Dev 60:1025-1034

Downey G, Coyne JC (1990), Children of depressed parents: an integrative review.

Psychol Bull 108:50-76

Eisenberg N, Mussen P (1997), The Roots of Prosocial Behavior in Children. Cambridge,

England: Cambridge University Press

Hoffman M (1963), Parent discipline and the child's consideration for others. Child Dev
34:573-588

33
Hops H, Biglan A, Sherman L, Arthud J, Friedman L, Osteen V (1987), Home observations
of family interactions of depressed women.] Consult Clin Psychol 55:341-346

Jain A, Belsky J, Crnic K (1996), Beyond fathering behaviors: types of dads. J Fam Psychol
10:431-442

Keitner GI, Miller 1W (1990), Family functioning and major depression: an overview. Am J
Psychiarty 147:1128-1137

Koestner R, Franz C, Weinberger J (1990), The family origins of empathic concern: a 26-
year longitudinal study. J Pers Soc Psychol 58:709-717

Kovacs M (1992), Children's Depression Inventory. New York: Multi-Health Systems

Leinonen J, Solantaus T, Punamaki, RL (in press), Economic hardship and the quality of
parenting: parents' mental health and marital relationship as mediating and moderating
factors. Int J Dev Behav

McFarlane AH, Bellissimo A, Norman GR (1995), Family structure, family functioning and
adolescent well-being: the transcendent influence of parental style. J Child Psychol
Psychiatry 36:847-864

O'Connor TG, Hetherington EM, Reiss D (1998), Family systems and adolescent
development: shared and nonshared risk and protective factors in nondivorced and
remarried families. Dev Psychopathol 10:353-375

Olweus D (1980), Familial and temperamental determinants of aggressive behavior in

adolescent boys: a causal analysis. Dev Psychol 16:644-660

Rosenthal R, Rosnow RL, Rubin DB (2000), Contrasts and Effect Sizes in Behavioral
Research: A Correlational Approach. Cambridge, England: Cambridge University Press

Seiffge-Krenke, I, Kollmar F (1998), Discrepancies between mothers and fathers'

perceptions of sons' and daughters' problem behaviour: a longitudinal analysis of parent-
adolescent agreement on internalising and externalizing problem behaviour. J Child
Psychol Psychiatry 39:687-697
Simons RL, Lorenz, FO, Conger RD, Wu CI (1992), The support from spouse as mediator
and moderator of the disruptive influence of economic strain on parenting. Child Dev
63:1282-1301

Solantaus-Simula T, Punamaki R-L, Beardslee WR (2002), Children's responses to low

parental mood, I: balancing between active empathy, overinvolvement, indifference, and
avoidance. J Am Acad Child Adolesc Psychiatry 41:278-286

Zahn-Waxler C, Radke-Yarrow M, King RA (1979), Child rearing and child's initiations

towards victims of distress. Child Dev 50:319-330

[Figure 1 omitted]

[Figure 2 omitted]

[Figure 3 omitted]

[Figure 4 omitted]

34
TABLE 1

Means and Standard Deviations of Measures of Parenting Styles as

Perceived by Family Members According to the Response Patterns

Response Patterns

Indifference Active
Empathy
Mean SD Mean

Authoritative parenting
Child-perceived/mother 4.20 (a) 0.05 4.31 (a)
Child-perceived/father 4.00 (ab) 0.04 4.11 (b)
Mother-reported 4.16 (ab) 0.04 4.20 (b)
Father-reported 3.82 0.06 3.79
Punitive parenting
Child-perceived/mother 4.20 (a) 0.05 4.21 (a)
Child-perceived/father 4.39 (a) 0.05 4.19 (b)
Mother-reported 3.88 0.04 3.88
Father-reported 4.00 0.04 3.95

Response Patterns
Emotional
Active Overinvolvement
Empathy
SD Mean SD

Authoritative parenting
Child-perceived/mother 0.03 4.16 (a) 0.07
Child-perceived/father 0.04 3.85 (a) 0.08
Mother-reported 0.03 4.30 (b) 0.07
Father-reported 0.04 3.77 0.09
Punitive parenting
Child-perceived/mother 0.03 3.84 (b) 0.07
Child-perceived/father 0.03 3.91 (c) 0.07
Mother-reported 0.02 3.96 0.05
Father-reported 0.03 3.90 0.06

Response Patterns

Avoidance F Statistics
Mean SD df

Authoritative parenting
Child-perceived/mother 3.90 (b) 0.04 3,953
Child-perceived/father 3.61 (c) 0.05 3,953
Mother-reported 4.04 (a) 0.04 3,715
Father-reported 3.68 0.05 3,491
Punitive parenting
Child-perceived/mother 3.87 (b) 0.04 3,953
Child-perceived/father 4.06 (bc) 0.04 3,953
Mother-reported 3.85 0.03 3,715
Father-reported 3.91 0.03 3,491

Response
Patterns

F Statistics
F Value

Authoritative parenting
Child-perceived/mother 27.08 (***)
Child-perceived/father 25.21 (***)
Mother-reported 5.46 (**)
Father-reported 1.40
Punitive parenting
Child-perceived/mother 15.02 (***)
Child-perceived/father 12.92 (***)
Mother-reported 1.01
Father-reported 1.12

Note: Means within rows not sharing the same superscript were
significantly different at p < .05 in Tukey-b tests.

(**) p < .01.

(***) p < .001.

RELATED ARTICLE: Effectiveness of a Home Intervention for Perceived Child Behavioral
Problems and Parenting Stress in Children With In Utero Drug Exposure. Arlene M. Butz,
RN, ScD, Margaret Pulsifer, PhD, Nicole Marano, BS, Harolyn Belcher, MD, Mary Kathleen

35
Lears, MPH, MSN, Richard Royall, PhD

Objective: To determine if a home-based nurse intervention (INT), focusing on parenting

education/skills and caregiver emotional support, reduces child behavioral problems and
parenting stress in caregivers of in utero drug-exposed children.

Design: Randomized clinical trial of a home-based INT.

Settings: Two urban hospital newborn nurseries; homes of infants (the term infant is used
interchangeably in this study with the term child to denote those from birth to the age of
36 months); and a research clinic in Baltimore, Md.

Participants: In utero drug-exposed children and their caregivers (N= 100) were
examined when the child was between the ages of 2 and 3 years. Two groups were
studied: standard care (SC) (n = 51) and INT (n = 49).

Intervention: A home nurse INT consisting of 16 home visits from birth to the age of 18
months to provide caregivers with emotional support and parenting education and to
provide health monitoring for the infant.

Main Outcome Measures: Scores on the Child Behavior Checklist and the Parenting Stress
Index.

Results: Significantly more drug-exposed children in the SC group earned c scores

indicative of significant emotional or behavioral problems than did children in the INT
group on the Child Behavioral Checklist Total (16 [31%] versus 7 [14%]; P = .04),
Externalizing (19 [37%] versus 8 [16%];P = .02), and Internalizing (14 [27%] versus 6
[12%] P = .05) scales and on the anxiety-depression subscale (16 [31%] versus 5 [10%];
P = .009). There was a trend (P = .06) in more caregivers of children in the SC group
reporting higher parenting distress than caregivers of children in the INT group.

Conclusions: In utero drug-exposed children receiving a home-based nurse INT had

significantly fewer behavioral problems than did in utero drug-exposed children receiving
SC (P = .04). Furthermore, those caregivers receiving the home-based INT reported a
trend toward lower total parenting distress compared with caregivers of children who
received SC with no home visits. Arch Pediatr Adolesc Med 2001;155:1029-1037.
Copyright 2001, American Medical Association.
COPYRIGHT 2002 Lippincott/Williams & Wilkins
in association with The Gale Group and LookSmart. COPYRIGHT 2002 Gale Group

36
Sex Roles: A Journal of Research
May, 1999

Power and Psychological Well-Being in Late Adolescent Romantic Relationships.

Author/s: Renee V Galliher

Using Olsen and Cromwell's (1975) three facets of interpersonal power (power in
resources, process, and outcomes), we examined the balance of power in late adolescent
romantic relationships and the associations between interpersonal power, adolescents'
self-esteem, and depression. Participants were 61 adolescent couples who were primarily
European-American (90% of girls and 93% of boys), with the remainder being comprised
of approximately 2-3% each of Native American, Asian, African American, and Hispanic
individuals. We assessed power in terms of access to emotional resources, control during
video-taped interaction, and control over decision-making. Overall, couples were more
likely to be described by themselves and by independent observers as egalitarian, with
the majority of couples equally contributing emotional resources, sharing power in
interaction, and sharing decision-making responsibility. However, romantic relationships
characterized by inequality in the contribution of emotional resources and in decision-
making were associated with greater psychological symptomatology, especially for
females.

Gender and power, central constructs in feminist discourse, are pervasive aspects of all
social relationships (Unger, 1976; 1978; Yoder & Kahn, 1992). Power is defined as the
means by which a person (or group) gets what is desired, despite opposition. In male-
dominant societies, more men than women possess power in the form of social status,
resources, safety, respect, authority, and positive self-regard (Pratto, 1996). The gender
gap in power filters down to the dyadic relationship between men and women. While a
substantial theoretical and empirical literature base exists exploring the development of
power-related gender differences in early peer interactions and in adult romantic/marital
relationship, there has been relatively little research on power in adolescent romantic
relationships.

Adolescent romantic relationships are viewed as an important context for the

development of one's sense of self or identity (Erikson, 1968) in relationship to others
(Sullivan, 1953). So it is surprising that, despite interest in gender differences in access to
power, little attention has been given to the distribution of power in these early romantic
relationships. In the following, we review the developmental literature on power in male-
female relationships using Olson and Cromwell's (1975) conceptualization of the three
aspects of power: resources, process, and outcome. Following this review, we describe
interpersonal power as manifested in a sample of adolescent romantic relationships and
explore the association between power and adolescent psychological well-being.

The first aspect of power, based on the principles of equity theory, emphasizes the social
exchange of resources. One basic tenet of equity theory states that individuals in
relationships will seek to maximize their own rewards or benefits in their interactions
(Laursen & Jensen-Campbell, in press; Sprecher, 1998) and conceptualizes the most
powerful members in relationships as those with the greatest access to valued resources
(Steil, 1994). Researchers have found females to be less likely to have access to valued
resources across the lifespan. Powlishta and Maccoby (1990) concluded that girls were
less likely than boys to gain access to scarce resources (at least in the absence of an
adult monitor) in a play situation. Further down the developmental continuum, research
on access to resources in marital relationships (Rodman, 1967; Safilios-Rothschild, 1976;
Vanfossen, 1977) has repeatedly found that spouses with greater educational and
economic resources are more likely to dominate and that women with greater access to
these resources are likely to be in more egalitarian relationships. Also, women in abusive
marriages, which are characterized by extreme gender differences in authority, are more
likely to remain in the relationship if they lack access to resources (Gelles, 1976).

37
The allocation of resources has not been empirically assessed in adolescent romantic
relationships. However, Laursen and Jensen-Campell (in press) have presented a
theoretical model that applies a developmental lens to traditional equity theory.
Adolescent romantic relationships differ from adult romantic relationships in that the goal
of these relationships is not necessarily long-term pair bonding. These early romantic
relationships are characterized as less exclusive and more voluntary than adult romantic
relationships, thus they are likely to be more vulnerable to dissolution when one partner
feels under-benefited by the relationship. Relationships that are characterized by inequity
are expected to dissolve quickly. Laursen and Jensen-Campbell predict, therefore, that
adolescents are more likely to avoid potential costs to the relationship, such as conflict
and inequity, since they perceive the relationship as vulnerable. In addition, external
constraints prohibit the development of some of the more traditional gendered power
discrepancies. Since adolescents remain dependent on their parents for material
resources and decision making, independence is limited for both partners. Given that
adolescents are often dependent on parents for resources, we might expect smaller
gender differences in interpersonal power in adolescent couples than in adult couples.

Traditionally, equity theory has emphasized partners' perceptions of equity in material

and symbolic resources, such as money, goods, services, respect, and status. In romantic
relationships, however, love has been identified as a central resource (Foa, 1971).
Recently, some researchers and theorists (Grauerholz, 1987; Sprecher, 1985) have begun
to argue that assessing interpersonal resources, such as understanding and support,
companionship, love, affection, and commitment, provides an important view of relational
power. This argument is supported by empirical findings that both men and women
describe commitment, attention, and pleasant company among the resources they most
value in their romantic relationships (Van Yperen & Buunk, 1990).

The relative allocation of these emotional resources has been described as a reflection of
the power structure of the relationship. According to the Principle of Least Interest
(Waller, 1937) romantic partners who are more invested, committed, and dependent on
their relationships are less powerful (Sprecher & Felmlee, 1997). One recent investigation
(Felmlee, 1994), examining relative emotional involvement in young adult romantic
relationships, reported that far fewer than half (39%) of dating partners described their
relationships as equal in terms of emotional involvement. Females were twice as likely as
males to be described as the more involved couple member.

In the current study, interpersonal resources were examined in terms of relative levels of
commitment to and dependency on the relationship. Empirical work assessing access to
material resources in children's peer contexts (Powlishta & Maccoby, 1990) and
examining interpersonal power in young adult romantic relationships (Felmlee, 1994)
supports the prediction that couple members would characterize themselves as unequal
in terms of commitment to and dependency on the relationship, and that females will be
more often perceived as the under-benefitted partner. However, theoretical contributions
suggest that developmental factors should temper such an expectation (Laursen &
Jensen-Campbell, in press). Thus, we expected greater egalitarianism in these adolescent
couples in terms of the allocation of emotional resources.

The second aspect, the process of power, is often examined by researchers using
observational methodology. Gender differences in this aspect of power have been
identified across the lifespan and in various types of relationships. Research on early peer
interactions in childhood indicates that at very young ages, girls and boys demonstrate
different behaviors related to exerting control and influence in their interactions. More
specifically, researchers have found that boys use more conversational tactics, such as
initiation and attention-getting devices, which are associated with higher status in our
society, while girls are more likely to use reinforcers and strategies aimed at facilitating
the conversations of boys (Berghout-Austin, Salehi, & Leffler, 1987). In classroom

38
settings, boys have also been observed using more direct attempts to influence their
interaction partners and are more successful than girls in getting their way (Serbin,
Sprafkin, Elman, & Doyle, 1982). Contrastingly, girls show more passive behavior and
often reduce their levels of social behavior when paired with boys in a play situation
(Jacklin & Maccoby, 1978). Comparable findings were reported with adult heterosexual
dyads (not romantic partners). Males have been found to actively control conversations,
both with females they have just met and with females they know, by using a
conversational style which includes interruptions and "cut-offs." Females, on the other
hand, have been found to be more silent in conversations with males than in
conversations with other females (West & Zimmerman, 1983; Zimmerman & West, 1975).

Similar patterns have been found in the romantic and marital relationship of adults.
Howard, Blumstein, and Schwartz (1986) reported that "weaker" tactics for influencing
one's partner, such as manipulation and supplication, were more likely to be used by
women in heterosexual relationships, while men were reported to use the "stronger"
conversational tactics of bullying and autocracy. Observational work with family
interaction has found patterns of dominance in husband-wife conversations to be similar
to those observed in parent-child interactions, with husbands more likely to interrupt
their wives and more likely to control the conversation (West & Zimmerman, 1977).

In this study, the process of power in adolescent romantic relationships was assessed
during videotaped interaction. Although developmental considerations tempered
expectations regarding gender differences in perceived discrepancies in power, previous
research examining mixed-sex interaction in childhood and in various adult relationships
indicates that males are more likely to dominate interaction across developmental levels.
Thus, we hypothesized that adolescent couple members would not participate equally
and that male couple members would be more likely to control the direction of the
discussion.

The third aspect, the outcome of power, is defined as control over decision making.
Again, the research describes a developmental phenomenon in which females
increasingly surrender control in decision making and adopt a passive stance in their
relationships with males. By adolescence, girls in mixed-sex pairs with their classmates
were more likely to relinquish decision making control to their male partners in problem
solving tasks (Lind & Connole, 1985). A few empirical studies have examined the
outcome dimension of power in late adolescent and early adult romantic couples. The
Boston Couples Study, which was conducted in the 1970s, found that males were more
likely to be described as the more powerful couple member and were more likely to
dominate in decision making matters (Peplau, 1979). More recently, Felmlee (1994), in
her large-scale questionnaire study of power in heterosexual relationships, found females
more likely than males to describe their relationships as equal, while males were more
likely to describe themselves as the more powerful couple member. However, fewer than
half of the males and females in this study reported equality, both on a global measure of
power and on a more specific measure of decision making, and males were twice as likely
to be described as the more powerful couple member in non-egalitarian dyads. Thus,
investigations of decision making power in romantic relationships portray a pattern of
male dominance. Developmental considerations (Laursen & Jensen-Campbell, in press),
however, might predict greater decision making equality in early romantic relationships.

Power and Psychological Health

Our second goal was to examine the relationship between power in adolescent romantic
relationships and individual adolescent psychological well-being. We measured
depressive symptomatology and self-esteem as indices of well-being because of their

39
significance during adolescence (Powers, Hauser, & Kilner, 1989). Petersen and
colleagues (1993) reviewed the research on depression and found that in 30 studies of
nonclinical adolescent samples, 20-35 percent of boys and 25-40 percent of girls reported
depressed mood. In longitudinal studies, depressed affect peaks around 17 or 18 years of
age and then declines for boys and levels off for girls. Low self-esteem has been
identified as part of a depressive syndrome and often precedes the full development of
this syndrome. A number of other problem behaviors such as poor peer relationships is
also strongly correlated with low self-esteem (Harter, 1993).

The gender differences in self-esteem and depression that increase in adolescence have
been well documented (Allgood-Merten, Lewinsohn, & Hops, 1990; Brage & Meredith,
1994; Nolen-Hoeksema, 1994; Petersen et al., 1993). Likewise, depression, self-esteem,
and poor interpersonal functioning are associated (Blatt & Zuroff, 1992); however, the
sparseness of research on adolescent romantic relationships leaves unanswered
questions regarding the associations between self-esteem and depression and
interpersonal power dynamics in these dyads.

One central tenet of equity theory states that individuals who find themselves
participating in inequitable relationships will become distressed (Sprecher, 1998).
Furthermore, the greater the inequity in the relationship, the greater the expected
distress. Those who are over-benefiting from the relationship are expected to experience
less overall distress, though they are expected to experience more guilt. Conversely,
those who are under-benefiting are expected to experience anger, depression, and
frustration. In studies examining the relationship between power and psychological health
in marriages, the position of powerlessness in one's intimate relationship has been
related to various negative outcomes (Aida & Falbo, 1991; Felmlee, 1994; Gray-Little &
Burks, 1983; Kollock, Blumstein, & Schwartz, 1985; Maccoby, 1990; Serbin, Sprafkin,
Elman, & Doyle, 1982; Steil & Turetsky, 1987; Vanfossen, 1981). Wives who described
their marriages as "reciprocal" or whose husbands were more likely to help with
household management reported fewer depressive symptoms (Ross, Mirowsky, & Huber,
1983; Steil & Turetsky, 1987; Vanfossen, 1981). Also, women perceived by their friends
as having egalitarian marriages reported higher self-esteem than women described as
traditional wives (Healey, 1980). Clearly, power inequality in adult romantic relationships
has been associated with poorer psychological well-being; however, we do not know the
extent to which this finding applies to adolescents.

Based on the previous research with adult couples, we hypothesized that inequality in
each of the three components of power (resources, process, and outcome) would be
associated with poorer psychological health in these adolescent couples. By comparing
male-dominant, female-dominant, and egalitarian couples, we were able to ask two
related questions: 1) Is the position of powerlessness relative to one's partner related to
psychological well-being?, and 2) Is inequality, in either direction, related to psychological
well-being?

In summary, this study explored three aspects of interpersonal power (resources,

process, outcome) believed to be operative in adolescent romantic relationships and
examined the relationship between the balance of power in the dyad and individual
psychological health. Previous research examining power in young children's peer
relationships and adult romantic relationships supports the prediction that these
adolescent relationships would more likely be characterized as non-egalitarian and that
males would more likely be described as the more powerful partner. However,
developmental considerations (Laursen & Jensen-Campbell, in press) predict greater
equality in these adolescent couples than has been reported in the literature on adult
couples. We also hypothesized that inequality in these adolescent romantic relationships
would be associated with higher depression and lower self-esteem across the three facets

40
of interpersonal power.

METHOD

Participants

Sixty-one target adolescents and their romantic partners participated in our study. Target
adolescents were 18 or 19 years of age and were recruited either through high school
year books and newspaper listings of recent high school graduates (n = 37) or through
freshman-level psychology courses (n = 24). Couples were invited to participate if the
target adolescent's partner was between 16 and 20 years old and the couple had been
dating for a minimum of four weeks. The age range was considered to be broad enough
not to restrict the number of couples that would be eligible and the length-of-
relationships requirement ensured that couples were in a somewhat committed
relationship. The length of dating relationships ranged from four weeks to five years
(median = eight months). Couples contacted through high school yearbooks or
graduation announcements were paid $60. Participants contacted through college
courses received extra credit for their participation.

Most of the participants lived with two parents (72% of girls and 57% of boys). Although
many of the participants held part-time jobs (49% of girls and 28% of boys), almost all
were enrolled in either high school or college (80% of girls and 74% of boys). The couples
in the sample were primarily European-American (90% of girls and 93% of boys), with the
remainder being comprised of approximately 2-3% each of Native American, Asian,
African American, and Hispanic individuals.

Procedure and Measures

Couples came to our laboratory for a total of 4 1/2 hours of data collection. They did this
in either one or two sessions depending upon their schedules. Our lab consists of a suite
of 3 separate rooms so that couples had privacy from our staff during the video-taping
portion and from each other during the interview and questionnaire portions of our study.
Couples were offered juice, soft drinks, and snacks during their sessions to facilitate
attentiveness and cooperation.

Assessing Power

Access to Valued Resources. Two techniques assessed the balance of power in adolescent
relationships in terms of access to emotional resources. Couple members participated in
a semi-structured interview adapted from an interview developed by Orlofsky (1993),
which assesses intimacy in close relationships. The interview was conducted as part of a
larger project and portions of the interview are presented in the current study. Each
participant was asked a series of questions to determine whether he, she, both, or neither
partner was more dependent on the relationship. Four trained coders listened to the
couple members' interviews and categorized the interviewees into one of four groups
based on each participant's perception of his or her relative dependence on the
relationship. Coders determined whether couple members perceived themselves, their
partners, both couple members, or neither couple member as concerned about being too
dependent in the relationship. Fifteen tapes were coded by all four coders; kappas
computed for each possible coder pair ranged from .60 to 1.0 (X = .77), indicating
adequate inter-rater reliability.

The second technique for assessing the balance of emotional resources examined
partners' level of commitment. Couple members completed the Dimensions of
Commitment Inventory (DCI), developed for use with married couples (Adams & Jones,
1997), and adapted for use with young unmarried couples (Rostosky, Welsh, Kawaguchi,

41
& Vickerman, in press). The modified DCI assesses the degree to which an individual
intends to maintain his or her romantic relationship. This intention is manifested in three
ways: (a) as an individual's devotion to and satisfaction with his or her romantic partner
(Commitment to the Partner Scale), (b) an individual's belief in the ideals of the romantic
relationship and a personal sense of obligation to maintain the relationship (Commitment
to the Relationship Scale), and (c) an individual's desire to avoid financial or social
penalties that might result from ending the relationship (Feelings of Entrapment Scale).
Adolescent couple members' responses on the modified DCI have been found to be
correlated with the stability of their relationships (Rostosky et al., in press; Rostosky,
1997). Cronbach's alphas for the current sample were .89 for the Commitment to Partner
scale, .80 for the Commitment to Relationship scale, and .82 for the Feelings of
Entrapment scale.

To assess power imbalances reflected by differing levels of commitment in these

adolescent relationships, difference scores were computed by subtracting one couple
member's commitment scores from the other member's scores. Since being less
committed to the relationship than one's partner is thought to place a couple member in
the more powerful position, difference scores were computed for individuals such that a
positive difference score indicates that the individual is less committed than his or her
partner and, therefore, is more powerful in this domain than the partner. Thus, in
analyses predicting males' psychological functioning, difference scores were computed by
subtracting males' scale scores from females' scale scores. For analyses predicting
females' psychological functioning, difference scores were computed by subtracting
females' scores from males' scores.

Power in Process. The process component of power was assessed using an observational
methodology. Couples were video-taped for twenty-two minutes having two
conversations about issues designed to elicit engaging conversation. In the first
conversation, couples were asked to imagine that it was 20 years in the future and they
were married to each other and had adolescent children of their own. They were
instructed to discuss how they would parent their adolescent children, what they would
like their relationship with each other to be like, and how their imagined family would be
similar to or different from their own families of origin. For the second conversation,
couples were asked to discuss a hypothetical dilemma that has been developed and used
by others (Gilligan, Kohlberg, Lerner, & Belenky, 1971) and was modified only slightly to
fit within contemporary adolescent language norms. The dilemma involved a high school
female whose parents were out of town for the weekend. While she was home alone, her
boyfriend unexpectedly visited. A series of questions asked how she should behave under
a variety of circumstances. For each discussion, couples were given instructions and a
written description of the conversation topic and were left alone to have the
conversation.

Two female graduate student coders (aged 25 and 44) globally rated the conversations,
determining if conversations were dominated by males, dominated by females, or were
participated in equally by both couple members. Conversation dominance was
determined by the relative amount of time each couple member talked and by evaluating
which couple member was most likely to control the direction of the conversation. Couple
members who were more likely to control the direction of conversation through their
greater participation were regarded as the more powerful partner. Eighteen of the tapes
were coded by both coders to assess inter-rarer reliability. Cohen's Kappa was .74,
indicating adequate agreement between the coders.

Outcomes of Power. The outcome component of power was assessed using a decision-
making questionnaire designed for use in a larger study. Three items asked couple
members: 1) When you and your partner disagree on something, who usually wins?, 2)
When you and your partner talk about important things, who usually makes the final
decision?, and 3) When it comes to decisions about sex, who usually has the final say?

42
Participants responded using a seven point Likert-type rating scale in which 1 = Partner
Always Does, 4 = We Both Do Equally, and 7 = I Always Do. Participants ratings were
collapsed and recoded to facilitate analyses. Thus, females' ratings of 1, 2, or 3 and
males' ratings of 5, 6, or 7 were recoded as 1 to indicate that the male was considered to
hold greater decision-making power. Ratings of 4 by either couple member were recoded
as 2 to indicate that the couple was perceived by the participant to share decision-
making power. Females' ratings of 5, 6, or 7 and males' ratings of 1, 2, or 3 were recoded
as 3 to indicate that the female was considered to hold greater decision making power.
Each couple was then categorized as either male dominant, egalitarian, or female
dominant from each partner's perspective for each of the three questions. For example, if
one girlfriend reported that her boyfriend was more likely to make final decisions about
important matters, the couple was categorized as male dominant from her perspective.
Her boyfriend's categorization, however, may or may not agree with hers.

Assessing Adolescent Psychological Functioning

Well-being was assessed using the Center for Epidemiology - Depression Scale (CES-D;
Radloff, 1977) and the Rosenberg Self-esteem Scale (Rosenberg, 1965). These scales
have been widely used and have demonstrated adequate psychometric properties. The
CES-D is a 20-item four-point Likert-type scale that assesses current depressive
symptomatology in non-psychiatric populations. Respondents indicate how well each
statement describes how they have felt during the past week. Examples of items from
the CES-D include: "I felt depressed.", "I had crying spells.", and "I could not get going."
Chronbach's alphas for the CES-D were .88 for female participants and .80 for male
participants in this study.

The Rosenberg Self-esteem Scale is a ten-item four-point Likert-type scale. The

respondent is asked to indicate his or her level of agreement, from strongly agree to
strongly disagree, with each statement describing feelings about the self. Examples of
items from the Rosenberg Self-esteem Scale include: "I feel I do not have much to be
proud of.", "I certainly feel useless at times.", and "I feel that I have a number of good
qualities." Chronbach's alphas for the Rosenberg Self-esteem Scale were .90 for females
and .86 for males in this study.

RESULTS

Preliminary Analyses

Preliminary descriptive analyses were performed to examine the distribution of couples

on several variables that were thought to be relevant to the study of power in couples.
Since the majority of participants were students (80% of girls and 74% of boys), access to
material resources was considered to be primarily under parental control. Preliminary
analyses focused on other traditional markers of power. Thirty-two (52%) of the 61
couples reported that male partners were older than female partners. In the remaining 29
couples, partners were the same age or females were the older couple member.

Couple members were also asked about their dating activities in order to access
adherence to traditional gendered dating roles: 1) Who drives when you go out?, and 2)
Who pays for dating activities? Fifty-nine percent of females and 62% of males reported
that males were more likely to drive on dates. Fifty-four percent of females and 56% of
males reported that males were more likely to pay for dating activities. Thus, the majority
of couples endorsed at least some traditional gender roles in dating behavior.

Three Aspects of Power

Resources/Interpersonal Power. Significant chi-square analyses performed on the ratings

43
of couple members' interview reports of relative dependency on the relationship revealed
a non-chance distribution from both couple members' perspectives (males: [X.sup.2] (3,
N = 56) = 12.14, p [less than] .01; females: [X.sup.2] (3, N = 53) = 16.21, p [less
than] .001). The modal rating of couple members regarding dependency in the
relationship was that neither member was concerned about being too dependent in the
relationship. Additional two-group chi-squares were performed to examine pair-wise
differences in non-egalitarian couples. Of non-egalitarian couples, females were more
likely to describe themselves as the overly dependent partner ([X.sup.2] (1, N = 21) =
10.71, p [less than] .001), while males were equally likely to describe themselves or their
girlfriends as too dependent ([X.sup.2] (1, N = 23) = .043, n.s.). (See Table I for
percentages of couple members coded in each category.)

Our second measure of interpersonal power was the calculated differences between male
and female reported commitment on the three scales measured by the Dimensions of
Commitment Inventory. The mean scores for the three commitment scales did not differ
significantly for males and females (Commitment to Partner Scale: t (59) = 1.39, n.s.;
Commitment to Relationship Scale: t (59) = -.90, n.s.; Feelings of Entrapment Scale:
[TABULAR DATA FOR TABLE I OMITTED] t (59) = -.26, n.s.) and the means of the
calculated difference scores hovered around zero (Commitment to Partner Scale: X = -
1.44, SD = 8.01; Commitment to Relationship Scale: X - 1.13, SD = 9.74; Feelings of
Entrapment Scale: X = .32, SD = 9.51), indicating that overall, males and females did not
report significantly different levels of commitment. However, difference scores examining
within-couple disparity in commitment demonstrated a large range (Commitment to
Partner Scale: range = -22 to 18; Commitment to Relationship Scale: range = -17 to 22;
Feelings of Entrapment Scale: range = -27 to 22), indicating that for some couples, large
discrepancies in the level of commitment existed. In approximately half of the couples,
girlfriends reported more commitment than boyfriends, while the opposite pattern
emerged in approximately half. Thus, group differences were canceled out and no overall
gender differences in commitment were apparent.

Process. Regarding the distribution of conversation in the video-taped interaction, the chi-
square analysis was significant ([X.sup.2] (2, N = 59) = 14.48, p [less than] .001), in that
the majority of couples (55%) were rated as egalitarian by the trained observers. A two-
group chi-square performed on the 26 nonegalitarian couples revealed that males and
females were equally likely to be described as the more dominant couple member
([X.sup.2] (1, N = 26) = 1.39, n.s.). (Refer again to Table I for percentages.)

Outcomes. The outcome component of power was defined as decision making power and
was assessed via three Likert-scale items, which were each analyzed separately for males
and females. The first question asked couple members who is most likely to win the
argument when they disagree on something. A non-significant chi-square indicated that
females were divided on this question, with about one third of girlfriends in each of the
three categories (male dominant, egalitarian, or female dominant; [X.sup.2] (2, N = 59) =
.64, n.s.). The chi-square examining males' responses, on the other hand, was significant
([X.sup.2] (2, N = 60) = 11.20, p [less than] .01). The majority of males reported that
both couple members were equally likely to win the argument. A two group chi-square
examining couples who were nonegalitarian from the males' point of view revealed an
equal distribution of male-dominant couples and female-dominant couples ([X.sup.2] (1,
N = 28) = .571, p [less than] n.s.).

The second question asked couple members who usually makes the final decisions on
important matters. Significant chi-squares indicated that both males ([X.sup.2] (2, N =
60) = 29.10, p [less than] .001) and females ([X.sup.2] (2, N = 61) = 56.50, p [less
than] .001) were more likely to describe themselves as egalitarian in this respect, with
79% of females and 65% of males describing themselves as equally likely to make final
decisions. Of the non-egalitarian couples, males were more likely to describe themselves
as the more powerful couple member ([X.sup.2] (1, N = 21) = 3.86, p [less than] .05),

44
while females in non-egalitarian couples were equally likely to describe themselves or
their partners as the powerful member ([X.sup.2] (1, N - 13) = .08, n.s.). The final
question in regards to decision making in these adolescent couples asked participants
who usually has the final say in decisions about sex. Again, the chi-square was significant
for both males and females (males: [X.sup.2] (2, N = 59) = 14.48, p [less than] .001;
females: [X.sup.2] (2, N = 58) = 33.66, p [less than] .001), with the majority of
participants reporting that they make decisions about sex equally. However, among non-
egalitarian couples, females ([X.sup.2] (1, N = 20) = 12.80, p [less than] .001) were more
likely to report that they had more say in decisions about sex than their boyfriends, while
males who perceived inequality in decisions about sex were more equally divided
([X.sup.2] (1, N = 26) = 1.39, n.s.). Table I contains percentages of couples placed in
each category for each item and from each couple member's perspective.

Power and Adolescent Psychological Functioning

Depression and Self-Esteem. Paired t-tests were performed to examine gender

differences in reported levels of depression and self-esteem. Consistent with previous
research (see review by Petersen et al, 1993), females' scores on the CES-D
([Mathematical Expression Omitted]; SD = 9.38) were significantly greater than males'
CES-D scores ([Mathematical Expression Omitted]; t (59) = 2.70 p [less than] .01), while
males reported higher scores on the Rosenberg Self-esteem Scale [Mathematical
Expression Omitted] than their girlfriends [Mathematical Expression Omitted]; t (59) = -
3.01, p [less than] .01).

Psychological Functioning and Interpersonal Power. Analyses of Variance (ANOVAs) were

performed to determine the relationship between adolescents' perceptions of their
relative dependence on the romantic relationship and self-esteem and depression. For
males, their perceptions of the balance of dependence in the relationship were related to
their reported self-esteem (F(3, 54) = 3.01, p [less than] .05). Tukey's post hoc analyses
indicated that those males who reported that both partners were worried about being too
dependent upon the relationship reported the lowest self-esteem, significantly lower than
those males who reported that neither partner was concerned about being too
dependent. The interview measure of relative dependence did not predict adolescent
girls' self-esteem, nor was it associated with depressive symptoms for males or females.

A series of regression analyses were performed to determine if differences in couple

members' commitment were related to adolescent psychological well-being. Regressions
were computed separately for each of the three scales of the DCI and for males and
females. Two variables were entered into each equation: 1) the linear terms of the
difference scores were entered to test the hypothesis that providing more emotional
commitment to the relationship than one's partner (or being less powerful in this domain)
was associated with poorer psychological functioning, and 2) the quadratic terms of the
difference scores were entered to test the hypothesis that inequality, regardless of the
direction of inequality, was associated with poorer psychological health.

Our efforts to predict Rosenberg Self-Esteem scores from differences in commitment

were successful for females. For both the Commitment to the Relationship scale
([R.sup.2] = .14, F(2,57) = 4.57, p [less than] .05), which encompasses a sense of
responsibility to stay in the relationship, and the Feelings of Entrapment scale ([R.sup.2]
= .17, F (2,57) = 5.87, p [less than] .01), which encompasses a sense of being obligated
to stay in the relationship and fear of social sanctions for leaving the relationship, linear
findings indicated that those females who reported higher levels of commitment on these
two scales than their boyfriends (or were less powerful than their boyfriends in this arena)
reported lower self esteem (Commitment to the Relationship: beta = .39, t = 3.02, p [less
than] .01; Feelings of Entrapment: beta = .38, t = 3.13, p [less than] .01). No relationship
was found between differences in the Commitment to the Partner scale and girls' self-

45
esteem, nor were any of the three difference scores related to boys' self-esteem.

Scores on the CES-D were somewhat related to differences in one of the three
commitment scales for males. A linear trend for the Feelings of Entrapment Scale
([R.sup.2] = .05, F(1,58) = 3.19, p [less than] .10; beta = -.24, t = - 1.84, p [less
than] .07) suggested that males who perceived more external barriers to dissolving the
relationship than their girlfriends may have reported higher depression.

Predicting Psychological Functioning From Power in Process. Neither CES-D scores nor
Rosenberg Self-Esteem scores were related to observers' ratings of dominance in the
videotaped conversations for either males or females.

Predicting Psychological Functioning From Power in Outcomes. A series of ANOVAS were

performed to assess the relationship between decision making power and the
psychological well-being of the adolescents. Two items that assessed decision making
power in the relationship predicted females' self-esteem. The first item asked couple
members who was most likely to win an argument if they disagreed (F(2,58) = 3.37, p
[less than] .05). Tukey's post hoc analyses indicated that those females who felt that they
were more likely to win arguments with their boyfriends reported higher self-esteem than
those who reported that their boyfriends were more likely to win the arguments. Females
who were in egalitarian relationships reported levels of self-esteem between the two
other groups and not significantly different from either. The second item that predicted
females' self-esteem asked couple members who had the most say in decisions about sex
(F(2,57) = 3.44, p [less than] .05). Tukey's post hoc analyses indicated that those females
who felt that they had less say than their boyfriends in decisions about sex reported
lower self-esteem than either those females in egalitarian relationships or those who
reported greater power than their boyfriends in decisions about sex. The level of males'
self-esteem was not related to their perceptions of inequality in decision making power,
nor were any of the decision making items related to depression for either boys or girls.

DISCUSSION

On June 9, 1998, the Southern Baptist Convention decreed that a wife should "submit
herself graciously" to her husband's leadership (Kelly, 1998). Perhaps this action
caricatures a larger and more widespread tension that still exists between conflicting
notions regarding egalitarian versus hierarchical male/female relationships. In this study,
the majority of adolescent romantic relationships were perceived by both observers and
participants as egalitarian. This contrasts with the literature regarding adult married
couples which concludes that, although equal partnership is most often acknowledged as
the ideal, the majority of married couples are not characterized as equal (Steil, 1994).
These findings raise questions regarding the differences between these young, relatively
short-term relationships and the more permanent bonds of marriage. Our results are
consistent with predictions based on the application of developmental considerations to
traditional equity theory (Laursen & Jensen-Campbell, in press). Perhaps adolescent
couples share power more equally than married couples because they have not yet
established the power relations typical of adult relationships (Schwartz, 1994; Steil,
1994). Once married, couples may tend to adopt more traditional gender roles that are
based on economic realities. Even though most marriages have two wage-earners,
husbands still typically have higher incomes than wives and wield more decision making
power as a result. These issues may be less salient in courtship, as implied in findings
from one study of college students indicating that commitment and dependency were
more strongly related to perceived egalitarianism in dating relationships than
comparative resources (Grauerholz, 1987).

Alternatively, perhaps these young couples represent a shift in our society toward more

46
egalitarian romantic relationships that are less gender role stereotyped. That is, perhaps
adolescent couples share decision-making power and contribute emotional resources
equally to a greater extent than past generations. Much of the literature that has
examined gender differences in access to power is somewhat dated (e.g., Jacklin &
Maccoby, 1978; Rodman, 1967; Safilios-Rothschild, 1976; Vanfossen, 1977). Recent
analyses indicate that women in the United States are more likely to endorse egalitarian
gender role attitudes than in the previous two decades (Harris & Firestone, 1998).
However, this finding may be evidence of a shifting norm rather than shifting behavior.
The fact that our observers also rated the majority of the adolescent couples as
egalitarian in the video-taped conversations lends support to the conclusion that these
couples not only endorse egalitarian ideals, but also "practice what they preach."
Longitudinal investigations are needed to clarify whether the egalitarianism of these
adolescent relationships represents a cohort effect or a developmental trajectory.

It is possible that methodological limitations produced a picture of power in these couples

which was biased in favor of egalitarianism. The conversation topics used in the
interaction portion of the study concerned domains which are considered traditionally
feminine: romantic relationships and parenthood. Perhaps the girlfriends in our study
were more practiced and confident in their discussions of these topics. Other studies of
adolescent interaction should incorporate a problem-solving task or activity as a
comparison.

Although the bulk of the adolescent couples in our study were characterized as
egalitarian, both by couple members and by independent observers, interesting patterns
of results emerged among the non-egalitarian couples. Contrary to our predictions, males
were not consistently perceived as the more powerful couple member in non-egalitarian
couples. Males were more likely to describe themselves as dominant in decision-making
about important matters and were less likely to be perceived by girlfriends as too
dependent on the relationship. However, in many areas, such as winning arguments,
imbalances were just as likely to favor the females as males. These findings challenge our
predictions that males would be more likely to be the power-holders in couples with
power imbalances. Contrary to findings in older couples, power does not appear to be as
gender-linked in these adolescent dating couples. These findings indicate that there are
advantages to employing a more comprehensive assessment of power that considers
various domains and perspectives. For these adolescent couples, at least, possessing
interpersonal power in one domain did not necessarily imply power in other domains, and
couple members did not always agree on the distribution of power in their relationship.

Among the non-egalitarian couples, girlfriends in this sample rated themselves more
likely to have the final say in decisions about sex. Thus, despite overall endorsement of
egalitarian ideals, girlfriends continue to perceive a behavioral inequality when it comes
to sexual decisions. The traditional sexual script in which males perform the initiator role
and females perform the refusal role has been found to dominate the interaction patterns
of young couples (Grauerholz & Serpe, 1985; McCormick & Jessor, 1983; Perper & Weis,
1987). These roles require males to push for sex and women to resist their advances,
effectively creating a sexual script based on conflict and power struggle rather than
communication, empathy, and mutuality.

Power and Adolescent Psychological Functioning

Our analyses examining the relationship between power and adolescent psychological
adjustment also yielded interesting findings. We found some support for the position that
interpersonal powerlessness in romantic relationships was associated with psychological
symptomatology, especially for girls. Females who reported that they won more
arguments, had more influence over sexual decisions, and were less committed to their
romantic relationships than their boyfriends had higher self-esteem. With our

47
correlational data, we cannot determine the causal nature of the relationship between
self-esteem and interpersonal power in romantic relationships. However, these findings
do raise speculation that interventions aimed at empowering young women in their
romantic relationships might be beneficial. Perhaps empowering adolescent girls would
help lay the foundation for achieving egalitarian marital relationships. To develop
strategies for empowering adolescent girls, it is crucial to understand how power
operates in their romantic relationships and how they are affected by it (Gage, 1997).
Only then will we be able to help them develop skills and capacity for achieving control
over their own lives and equal heterosexual partnerships. Of course, the romantic
relationships of young women are not self-contained. Rather, they are impacted by
historical, economic, and socio-cultural contexts that serve to maintain power imbalances
in close relationships.

Findings supporting the relationship between powerlessness and psychological

symptomatology in male adolescents were less clear. In contrast to the females, only
concerns about dependency were related to males' self-esteem. Further, males who felt
that both couple members were too dependent demonstrated the lowest self-esteem, so
it was not the position of being powerless relative to their girlfriends that was associated
with the poorer outcome. In addition, one tenuous relationship was established between
males' perceptions of inequality in commitment and their reported depression. Males who
felt more trapped in their relationships than their girlfriends tended to report more
depressive symptoms. These findings are preliminary, but suggest that emotional
equality may have been more salient to boys in our sample, while both concerns over
emotional equality and concerns over the more traditionally assessed aspects of power
(decision-making) were salient to girls. This finding is consistent with our previous work
(Welsh, Galliher, Kawaguchi, & Rostosky, 1997), and that of others (Perry, Schmidtke, &
Kulik, 1998), which suggests that the lens of power is more salient for females than it is
for males. Of course, we recognize that the distribution of power is only one aspect of
romantic relationships and that romantic relationships are only one part of the social
world of adolescents. A more complete model of adolescent psychological well-being
would examine various aspects of adolescents' important relationships.

Several limitations regarding the current study should be noted. Our sample was quite
small and consisted of couples who had been dating for at least one month and were
committed to one another enough to agree to participate in a lengthy and somewhat
arduous investigation of their relationship. The criteria for involvement in our study may
have precluded participation by couples who were particularly non-egalitarian, resulting
in a sample biased toward equal distribution of power. Although the intensive study of
couple members' perceptions of power has advantages, using a less demanding research
design or assessing individuals' perceptions of power in their relationships may provide
additional information that yields a more complete and representative picture of the
distribution of power in adolescent romantic relationships.

Another limitation of our sample was that it was composed entirely of male-female
couples with no gay or lesbian couple participants. One of the few studies of lesbian
relationships concluded that power dynamics can not be reduced to a function of sexual
orientation or ideology; rather, power involves a complex social exchange process
involving both personal and interpersonal resources (Caldwell & Peplau, 1984). Certainly,
more studies examining power in resources, processes, and outcomes in same-sex
romantic relationships are needed.

We consider these findings to be preliminary as we have just begun to consider the

nature of power in the romantic relationships of adolescents. We hope to stimulate
research that will further clarify power relations in adolescence, examine the
developmental trajectory of power balances and imbalances, and explore the ways in

48
which power dynamics impact the psychological health of young women and men.

REFERENCES

Adams, J. M., & Jones, W. H. (1997). The conceptualization of marital commitment: An

integrative analysis. Journal of Personality and Social Psychology, 72, 1177-1196.

Allgood-Merten, B., Lewinsohn, P. M., & Hops, H. (1990). Sex differences and adolescent
depression. Journal of Abnormal Psychology, 59, 55-63.

Berghout-Austin, A. M., Salehi, M., & Leffler, A. (1987). Gender and developmental
differences in children's conversations. Sex Roles, 16(9/10), 497-510.

Blatt, S., & Zuroff, D. (1992). Interpersonal relatedness and self-definition: Two prototypes
for depression. Clinical Psychology Review, 12, 527-562.

Brage, D., & Meredith, W. (1994). A causal model of adolescent depression. Journal of
Psychology, 128, 455-468.

Caldwell, M. A., & Peplau, L. A. (1984). The balance of power in lesbian relationships. Sex
Roles, 10, 587-599.

Erikson, E. H. (1968). Identity: Youth and crisis. New York: W. W. Norton.

Felmlee, D. H. (1994). Who's on top? Power in romantic relationships. Sex Roles, 31(5/6),
275-295.

Foa, U. G. (1971). Interpersonal and economic resources. Science, 171, 345-351.

Gage, A. J. (1997, April). Female empowerment and adolescent demographic behavior.
Paper presented at Female Empowerment and Demographic Processes: Moving Beyond
Cairo. Seminar conducted at the meeting of the IUSSP Committee on Gender and
Population and Programme on Population and Development, Lund, Sweden.

Gelles, R. J. (1976). Abused wives: Why do they stay? Journal of Marriage and the Family,
38(4), 659-668.

Gilligan, C., Kohlberg, L., Lerner, J., & Belenky, M. (1971). Moral reasoning about sexual
dilemmas. Technical Report of the Commission on Obscenity and Pornography (Vol. 1).
Washington, DC: U.S. Government Printing Office.

Grauerholz, E. (1987). Balancing the power in dating relationships. Sex Roles, 17, 563-
571.

Grauerholz, E., & Serpe, R. T. (1985). Initiation and response: The dynamics of sexual
interaction. Sex Roles, 12, 1041-1059.

Gray-Little, B., & Burks, N. (1983). Power and satisfaction in marriage: A review and
critique. Psychological Bulletin, 93, 513-538.

Harris, R. J., & Firestone, J. M. (1998). Changes in predictors of gender role ideologies
among women: A multivariate analysis. Sex Roles, 38, 239-252.

Harter, S. (1993). Cause and consequences of low self-esteem in children and

adolescents. In R. F. Baumeister (Ed.), Self-esteem: The puzzle of low self-regard. New

49
York: Plenum Press.

Healey, D. E. (1980). Attitudes, status, and psychological functioning of women in

egalitarian and traditional marriages. Dissertation Abstracts International, 40, 5812-5813.

Howard, J. A., Blumstein, P., & Schwartz, P. (1986). Sex, power, and influence tactics in
intimate relationships. Journal of Personality and Social Psychology, 51, 102-109.

Jacklin, C. N., & Maccoby, E. E. (1978). Social behavior at thirty-three months in same-sex
and mixed sex dyads. Child Development, 49, 557-569.

Kelly, M. (1998, June 10). Southern Baptists adopt statement on the family. Baptist Press,
p. 1.

Kollock, P., Blumstein, P., & Schwartz, P. (1985). Sex and power in interaction:
Conversational privileges and duties. American Sociological Review, 50, 34-46.

Laursen, B., & Jensen-Campbell, L. A. (in press). The nature and functions of social
exchange in adolescent romantic relationships. In W. Furman, B. B., & Feiring, C. (Eds.),
Contemporary perspectives on adolescent romantic relationships. New York: Cambridge
University Press.

Lind, P., & Connole, H. (1985). Sex differences in behavioral and cognitive aspects of
decision control. Sex Roles, 12(7/8), 813-823.

Maccoby, E. E. (1990). Gender and relationships. American Psychologist, 45, 513-520.

McCormick, N., & Jessor, C. J. (1983). The courtship game: Power in the sexual encounter.
In E. R. Allgeier & N. B. McCormick (Eds.), Changing boundaries: Gender roles and sexual
behavior. Palo Alto, CA: Mayfield.

Nolen-Hoeksema, S. (1994). An interactive model for the emergence of gender

differences in depression in adolescence. Journal of Research on Adolescence, 4, 519-534.
Olsen, D. H., & Cromwell, R. E. (1975). Methodological issues in family power. In R. E.
Cromwell & D. H. Olson (Eds.), Power in families. New York: Wiley.

Orlofsky, J. L. (1993). Intimacy status: Theory and research. In J. E. Marcia, A. S.

Waterman, D. R. Matteson, S. L. Archer, & J. L. Orlofsky, Ego identity: A handbook for
psychosocial research. New York: Springer-Verlag.

Peplau, L. A. (1979). Power in dating relationships. In J. Freeman (Ed.), Women: A feminist

perspective (2nd ed.). Palo Alto, CA: Mayfield.

Perper, T., & Weis, D. L. (1987). Proceptive and rejective strategies of U.S. and Canadian
college women. Journal of Sex Research, 23, 455-480.

Perry, E. L., Schmidtke, J. M., & Kulik, C. T. (1998). Propensity to sexually harass: An
exploration of gender differences. Sex Roles, 38, 443-460.

Petersen, A. C., Compas, B. E., Brooks-Gunn, J., Stemmler, M., Ey, S., & Grant, K. E.
(1993). Depression in adolescence. American Psychologist, 48, 155-168.

Powers, S. I., Hauser, S. T., & Kilner, L. (1989). Adolescent mental health. American
Psychologist, 44, 200-208.

50
Powlishta, K. K., & Maccoby, E. E. (1990). Resource utilization in mixed-sex dyads: The
influence of adult presence and task type. Sex Roles, 23(5/6), 223-240.

Pratto, F. (1996). Sexual politics: The gender gap in the bedroom, the cupboard, and the
cabinet. In D. M. Buss & N. M. Malamuth (Eds.), Sex, power, conflict: Evolutionary and
feminist perspectives. New York: Oxford University Press.

Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the
general population. Applied Psychological Measurement, I, 385-401.

Rodman, H. (1967). Marital power in France, Greece, Yugoslavia, and the United States: A
cross-national discussion. Journal of Marriage and the Family, 29, 320-324.

Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, N J: Princeton

University Press.

Ross, C., Mirowsky, J., & Huber, J. (1983). Dividing work, sharing work, and in-between:
Marriage patterns and depression. American Sociological Review, 48, 809-823.

Rostosky, S.S., Welsh, D. P., Kawaguchi, M. C., & Vickerman, R. C. (in press). Commitment
and sexual behaviors in adolescent dating couples. In W. Jones & J. Adams (Eds.), The
handbook of interpersonal commitment and relationship stability. New York; Plenum.

Rostosky, S. S. (1997). Power, sex, and relationship quality in late adolescent dating
relationships. Unpublished doctoral dissertation, The University of Tennessee, Knoxville.

Safilios-Rothschild, C. (1976). A macro- and micro-examination of power and love: An

exchange model. Journal of Marriage and the Family, 38, 355-362.

Schwartz, P. (1994). Peer marriage: How love between equals really works. New York: The
Free Press.

Serbin, L. A, Sprafkin, C., Elman, M., & Doyle, A. (1982). The early development of sex-
differentiated patterns of social influence. Canadian Journal of Behavioral Science, 14,
350-363.
Sprecher, S. (1985). Sex differences in bases of power in dating relationships. Sex Roles,
12, 449-462.

Sprecher, S. (1998). Social exchange theories and sexuality. The Journal of Sex Research,
35, 32-43.

Sprecher, S., & Felmlee, D. (1997). The balance of power in romantic heterosexual
couples over time from "his" and "her" perspectives. Sex Roles, 37, 361-379.

Steil, J. M. (1994). Equality and entitlement in marriage: Benefits and barriers. In M. J.

Lerner & G. Mikula (Eds.), Entitlement and the affectional bond: Justice in close
relationships. New York: Plenum Press.

Steil, J. M., & Turetsky, B. A. (1987). Is equal better? The relationship between marital
equality and psychological symptomatology. Applied Social Psychology Annual, 7, 73-97.

Sullivan, H. S. (1953). The interpersonal theory of psychiatry. New York: Norton.

Unger, R. K. (1976). Male is greater than female: The socialization of status inequality.

51
The Counseling Psychologist, 6, 2-9.

Unger, R. K. (1978). The politics of gender: A review of relevant literature. In J. Sherman &
F. Denmark (Eds.), The psychology of women: Future directions of research. New York:
Psychological Dimensions.

Van Yperen, N., & Buunk, B. (1990). A longitudinal study of equity and satisfaction in
intimate relationships. European Journal of Social Psychology, 20, 287-309.

Vanfossen, B. E. (1977). Sexual stratification and sex-role socialization. Journal of

Marriage and the Family, 39, 563-574.

Vanfossen, B. E. (1981). Sex differences in the mental health effects of spouse support
and equity. Journal of Health and Social Behavior, 22, 130-143.

Waller, W. (1937). The rating and dating complex. American Sociological Review, 2, 727-
734.

Welsh, D. P., Galliher, R. V., Kawaguchi, M. C., & Rostosky, S. S. (in review). Shared
realities: Adolescent couples' subjective understanding of their interaction and its
relationship to their mental health.

West, C., & Zimmerman, D. H. (1977). Women's place in everyday talk: Reflections on
parent-child interaction. Social Problems, 24, 521-529.

West, C., & Zimmerman, D. H. (1983). Small insults: A study of interruptions in cross-sex
conversations between unacquainted persons. In B. Thorne, C. Kramarae, & N. Henley
(Eds.), Language, gender and society. Rowley, MA: Newbury House.

White, J. W., & Koss, M. P. (1991). Courtship violence: Incidence in a national sample of
higher education students. Violence and Victims, 6, 247-256.

Yoder, J. D., & Kahn, A. S. (1992). Toward a feminist understanding of women and power.
Psychology of Women Quarterly, 16, 381-388.

Zimmerman, D. H., & West, C. (1975). Sex roles, interruptions, and silences in
conversation. In B. Thorne & N. Henley (Eds.), Language and sex: Differences and
dominance. Rowely, MA: Newbury House.
COPYRIGHT 1999 Plenum Publishing Corporation
in association with The Gale Group and LookSmart. COPYRIGHT 2000 Gale Group

52
Journal of Abnormal Child Psychology
August, 2002

Fear of abandonment as a mediator of the relations between divorce stressors

and mother-child relationship quality and children's adjustment problems.

Author/s: Sharlene A. Wolchik

Although the negative effects of parental divorce on adjustment problems have been
extensively documented, the processes through which divorce leads to these outcomes
have not been well articulated. A considerable body of literature has identified both
social--environmental and intrapersonal factors that affect the development of
adjustment problems in children following parental divorce. However, studies have not
examined pathways to the development of these adjustment problems that involve the
joint influence of social--environmental and intrapersonal factors. Identification of such
pathways has clear implications for theories of the etiology of adjustment problems for
children following parental divorce and should provide guidance for the design of
effective prevention and treatment programs. Given that over 1 million children in the
United States experience parental divorce each year (Cherlin, 1992), the public health
implications of such programs are significant.

This study uses a prospective longitudinal design to examine the plausibility of a model in
which children's fear that they will not be cared for (i.e., fear of abandonment) mediates
the relations between two empirically supported correlates of children's postdivorce
adjustment problems: mother-child relationship quality and divorce stressors. First, the
research on children's postdivorce adjustment problems is discussed. Next, the literature
on the relations between divorce stressors, as well as mother-child relationship quality,
and postdivorce adjustment problems is briefly reviewed, and the limited empirical work
on fear of abandonment is discussed. Finally, plausible linkages between divorce
stressors, mother-child relationship quality, fear of abandonment, and children's
postdivorce adjustment problems are articulated and theoretical support for a
mediational model is provided.

Potential adjustment and social adaptation problems of children who have experienced
parental divorce include increased levels of aggression, depression, and anxiety; poor
academic performance; school drop-out; peer relationship problems; drug and alcohol
use; early sexual behavior; and adolescent pregnancy (e.g., Amato & Keith, 1991a;
Hetherington et al., 1992). Although for some children the effects of this transition in
family structure are mild and short lived, for other children, divorce leads to clinically
significant and lasting adjustment problems during childhood and adolescence (see
Amato & Keith, 1991a). Further, several longitudinal studies have shown elevated rates of
mental health problems in adults who experienced parental divorce as children (e.g.,
Chase-Landale, Cherlin, & Kiernan, 1995; Rodgers, Power, & Hope, 1997; Ross &
Mirowsky, 1999). For example, in a prospective study, Rodgers et al. found the odds ratio
of being above the clinical level on mental health problems for parental divorce to be
1.70 at age 23 and 1.85 at age 33.

The research focused on predictors of variation in children's postdivorce adjustment

problems has consistently found that two social--environmental factors, divorce stressors
and custodial parent--child relationship quality, are significantly associated with
postdivorce adjustment problems. It is well documented that divorce often involves a
wide array of disruptions or stressors, including increased fights between parents,
exposure to parental distress, changes in residence and schools, involvement with
parents' new partners, and loss of time with one or both parents, as well as extended
family members (e.g., Sandler, Wolchik, Braver, & Fogas, 1986). There is considerable
evidence indicating a significant relation between divorce stressors and children's
postdivorce adjustment problems (e.g., Sandler, Wolchik, Braver, & Fogas, 1991; Stolberg
& Anker, 1983; Wolchik, Wilcox, Tein, & Sandler, 2000). It also is well documented that

53
changes in parenting, such as decreased warmth and affection, poorer communication,
and erratic discipline, commonly occur after divorce (e.g., Hetherington, Cox, & Cox,
1982; Peterson & Zill, 1986; Simons et al., 1996). Researchers have consistently
documented that high levels of warmth and affection in the custodial mother-child
relationship are negatively related to postdivorce adjustment problems (e.g.,
Hetherington et al., 1992; Simons, Lin, Gordon, Conger, & Lorenz, 1999; Wolchik et al.,
2000). Further, several researchers have shown that divorce stressors and mother--child
relationship quality interact to affect children's postdivorce adjustment problems, such
that the relation between divorce stressors and adjustment problems is mitigated at high
levels of warmth and affection (e.g., Camara & Resnick, 1987; Wolchik et al., 2000).

The current study tests whether the effects of both of these social--environmental factors
can be accounted for through a common mediating pathway, their joint effect on an
intrapersonal factor, children's fear of being abandoned. From a motivational theory of
stress and coping (Skinner & Wellborn, 1994, 1997), stressors affect children's
adjustment problems because they threaten one or more of three basic needs:
relatedness, competence, and autonomy. Sandler (2001) proposed that the effects of
both stressors and protective resources work through their effects on children's
perceptions of satisfaction of these basic needs. He proposed that stressors lead to
higher adjustment problems by threatening basic need satisfaction, whereas protective
resources reduce adjustment problems either by directly promoting need satisfaction or
by decreasing the negative effects of stressors on need satisfaction. This paper proposes
that postdivorce stressors particularly threaten one basic need, children's need to be part
of a caring and stable social group (Baumeister & Leary, 1995) and that the protective
resource of a high quality relationship with the primary residential parent reduces this
threat.

Several researchers have suggested that divorce threatens children's need to be part of a
caring social group. For example, Kurdek and Berg (1987) note that children whose
parents divorce may believe that they will lose contact with their residential, as well as
nonresidential, parent. Similarly, Gardner (1976) observes that children who experience
the departure of one parent from the home wonder what is to prevent the remaining
parent from also leaving. Wallerstein (1985) notes that divorce can cause a pervasive
sense of vulnerability for children as the protective, nurturing aspects of the family
diminish. She also observes that children often experience fears of being lost in the
shuffle and have concerns that their needs will be disregarded because their parents are
so focused on their own needs.

Of the multiple theoretical perspectives that focus on central social relationships, the two
most relevant to the current study are need for relatedness (Baumeister & Leary, 1995)
and attachment (e.g., Bowlby, 1973, 1980). Although these theories differ in many
respects, they converge in predicting that fear of being abandoned by one's primary
caregivers leads to adjustment problems, and thus provide support for the importance of
the model that is being tested. As articulated by Baumeister and Leary (1995), need for
relatedness includes both a need for frequent personal contact that is primarily
affectively positive and free from negative affect and a sense that an interpersonal bond
characterized by affective concern will endure over time. From an attachment
perspective, the hallmark of secure attachment involves open and relaxed
communication between the parent and child and the perceived availability of and
reliance on the attachment figure when distressed (Bowlby, 1969/1982). Increasing
evidence suggests that humans have a need for a sense of felt security in their
relationships with parents, peers, and intimate partners and that these relationships have
significant influence on a variety of developmental and behavioral outcomes (see Bowlby,
1980; Bretherton & Munholland, 1999; Dozier, Stevenson, Lee, & Velligan, 1991). For
example, Baumeister and Leary (1995) review evidence that individuals who lack a sense
of belongingness experience higher levels of mental and physical illnesses, such as
depression, somatic problems, and decreased immunocompetence. Further, researchers

54
have demonstrated that insecure attachment serves as a risk factor for maladjustment in
the context of risk factors from multiple domains, such as family stress and low child IQ
(e.g., Greenberg, 1999).

It is important to note that the current study is not a test of either of these theoretical
perspectives. Neither internal working models of attachment nor need for relatedness is
assessed. Rather, children's fear of being abandoned, which is likely related to both
attachment and need for relatedness, is examined. The limited empirical work on fear of
abandonment has defined this construct as including worries about the stability of
children's relationships with their parents, as well as continuity of living arrangements.
Kurdek and Berg (1987) examined relations between several divorce-related beliefs (peer
ridicule and avoidance, maternal blame, paternal blame, self-blame, hope for
reunification, fear of abandonment) and mother, teacher, and child reports of adjustment
problems. Only fear of abandonment was significantly related to children's reports of
anxiety in their sample of White, middle class children. Using an inner-city, predominantly
ethnic minority sample, Wolchik, Ramirez, Sandler, Fisher, Organ ista, and Brown (1993)
examined the relations between children's postdivorce adjustment problems and fear of
abandonment, paternal blame, maternal blame, and hope for reconciliation. Significant
relations were found only for fear of abandonment, with higher scores being significantly
related to both mother and child reports of children's adjustment problems.

At a theoretical level, it is plausible that the relations between divorce stressors and
adjustment problems, as well as between mother-child relationship quality and
adjustment problems, are mediated by fear of abandonment. As noted earlier, divorce
often sets in motion a multitude of stressors and changes in the mother-child
relationship. The experience of stressors that disrupt children's social connections to their
primary residential or nonresidential parent, involve conflict between their parents, or
indicate vulnerability of their parents is likely to create concerns about the ability or
willingness of their family to continue to care for them. On the other hand, the
interactions that occur in a high quality relationship between the child and the residential
parent provide evidence that the child will be cared for and may either directly reduce
concerns about being abandoned or mitigate the effects of divorce stressors on fear of
abandonment.

The current study tests the plausibility of a model in which the relations between
children's adjustment problems and both divorce stressors and mother-child relationship
quality are accounted for by a common intrapersonal factor, children's fear that they will
be abandoned. In this model, divorce stressors and mother-child relationship quality
relate to fear of abandonment and fear of abandonment relates to adjustment problems.
Further, the relations between divorce stressors and mother-child relationship quality and
adjustment problems are mediated through fear of abandonment. Given empirical and
theoretical work on the stress-mitigating effects of high quality mother-child relationships
in divorced families (e.g., Camara & Resnick, 1987; Sandler, 2001; Wolchik et al., 2000),
the model also tests whether divorce stressors and mother-child relationship quality
interact to predict fear of abandonment. It was predicted that the relation between
divorce stressors and fear of abandonment will be weaker for child ren with high mother-
child relationship quality than that for children with low mother-child relationship quality.

Two methodological aspects of the current study are noteworthy. First, the study utilizes
a prospective longitudinal design in which Time 1 divorce stressors, mother-child
relationship quality, and fear of abandonment predict Time 2 adjustment problems,
controlling for Time 1 adjustment problems. Because prospective longitudinal data satisfy
the condition of time precedence, they are particularly useful in testing the plausibility of
causal directionality between variables. Second, to reduce concerns that observed
relations might be due to shared method variance across the measures or self-report
negativity bias and to allow the examination of the robustness of the findings across

55
models, mother as well as child reports of mother-child relationship quality and children's
adjustment problems were used.

METHOD

Participants

The sample consists of 216 children who experienced parental divorce within the
previous 2 years and their primary residential mothers. These families were participants
in the Divorce Adjustment Project (Sandler, Tein, & West, 1994), a longitudinal study of
children's postdivorce psychological adjustment. The primary goal of this study was to
identify short-term longitudinal correlates of postdivorce adjustment problems that could
be used to guide the development of prevention programs for children who lived
primarily with their mothers, the residential arrangement that characterizes 80% of
divorced families (U.S. Bureau of Census, 1998). Thus, neither primary residential fathers
nor noncustodial fathers were interviewed. The time period of 2 years was used given
that restabilization of the family usually occurs 2-3 years after divorce (Hetherington,
1999). Only families who participated in both Time 1 and Time 2 assessments, which
occurred 5.5 months apart, were included. The 5-month time interval was used b ecause
it was long enough to allow for change in mental health problems and short enough to
detect the prospective effects of stress and adaptation processes that occur at Time 1
(see Sandler et al., 1994; Sheets, Sandler, & West, 1996, for other examples of
prospective longitudinal effects across this time period).

Court records were used to identify potential participants. A random sample of 1,236
families with children was identified from the countywide records of divorces granted in
the last 2 years. Participation in the study was solicited by an initial mailing and a follow-
up phone call. Forty-nine percent of selected families were reached by phone, and of
these, 73% met the following eligibility criteria: the family contained a child between the
ages of 8 and 12; the mother had not remarried and did not have a live-in partner; the
child resided with her/his mother at least half the time; mother and child were fluent in
English; the family lived in and expected to remain in the greater Phoenix metropolitan
area for the study period; and child's residential status (i.e., primary residence with
mother) was expected to remain stable over the study period. The primary reasons for
ineligibility were that the mother had remarried (44%), the family had moved outside of
the greater Phoenix metropolitan area (44%), and the child lived with the mother less
than half of the time (9%). In families where there was more than one child in the
targeted age range, one child was randomly selected to ensure independence of
response.

Fifty-eight percent (n = 256) of the families who were eligible and invited to participate in
the study participated in the Time 1 assessment. Children interviewed at Time 1
averaged 9.59 years of age (SD = 1.19); 44% were female. Eighty-six percent of the
children had at least one sibling living with them. The majority of the mothers were
Caucasian/non-Hispanic (87%); 9% were Hispanic, 2% were Black, and 3% were of
another racial or ethnic background. Mothers averaged 35.3 years of age (SD = 5.5).
Twenty-four percent of the mothers had completed college or attended graduate
programs; 40% had taken some college courses or completed technical school; 28% had
completed high school; and 8% had less than a high school education. Mother's average
yearly income fell in the range of $20,001-$25,000. The average time since physical
separation was 26.6 months (SD = 13.3); the average time since divorce was 13.4
months (SD = 6.5). In 63% of the families, the mothers had sole legal custody; the rest
had joint legal cust ody. Mothers reported that 40% of the children typically had
unrestricted contact with their fathers, 26% saw their fathers on a regular basis, 27% saw
their father only occasionally, and 7% had no contact with their fathers at all.

56
Eighty-four percent (n = 216) of the families who completed the Time 1 assessment also
completed the Time 2 assessment. The reasons that families attritted or were attritted
from Time 2 assessment were (a) referral for treatment by project staff due to children
scoring above the clinical cutoff on the Child Depression Inventory (Kovacs, 1981) or
reporting current suicidal ideation (n = 19), (b) moving out of the Phoenix metropolitan
area (n = 5), (c) becoming ineligible (n = 1), or (d) refusing to participate at Time 2 (n =
15). Demographic data, which were collected at Time 1, for those families who
participated in both assessments are as follows: These children averaged 9.64 years of
age at the first assessment (SD = 1.20); 44% were female. The majority of the mothers
were Caucasian/non-Hispanic (86%); 9% were Hispanic, 2% were Black, and 3% were of
another racial or ethnic background. Mothers averaged 35.5 years of age at the first
assessment (SD = 5.7); 25% of them had completed college or attended gradua te
programs; 39% had taken some college courses or completed technical school; 27% had
completed high school; and 9% had less than a high school education. Mother's average
yearly income fell in the range of $20,001-$25,000. The average time since physical
separation was 26.3 months (SD = 13); the average time since divorce was 13.3 months
(SD = 6.5). In 63% of the families, the mothers had sole legal custody; the rest had joint
legal custody. Mothers reported that 41% of the children typically had unrestricted
contact with their fathers, 26% saw their fathers on a regular basis, 26% saw their father
only occasionally, and 6% had no contact with their fathers at all.

Attrition analyses were conducted on the Time 1 variables to compare families who
completed the Time 2 assessment to those families who did not. Chi-square analyses
were applied to test the categorical variables and t statistics were applied to test the
continuous variables. Mothers from families who completed the Time 1 assessment but
did not complete the Time 2 assessment were older (M = 35.54) than those who
completed the Time 2 assessment (M = 33.80, t = 2.10, p < .05). Also, children from
families who did not complete the Time 2 assessment reported higher fear of
abandonment (M = 0.93), more divorce stressors (M = 4.55), and higher depression
scores (M = 12.37) than children who completed the Time 2 assessment (M = 0.47, t =
2.76, p < .01; M = 3.12, t = 3.00, p < .01; and M = 6.22, t = 3.60, p < .01, respectively).

Procedure

Mothers and children were interviewed separately by trained interviewers. After

confidentiality was explained, mothers signed informed consent forms and children
signed assent forms indicating their willingness to participate. Families received $50
compensation for each assessment.

Predictors

Fear of Abandonment. Children completed the 6-item Fear of Abandonment subscale of

the Children's Beliefs about Parental Divorce Scale (Kurdek & Berg, 1987). This subscale
assesses concerns about the stability of relationships with parents and continuity of living
arrangements. Responses are dichotomous (true; false). Kurdek and Berg obtained a 9-
week stability coefficient of .52 for this subscale. Given the dichotomous response format
and highly skewed responses, confirmatory analysis with MPlus (Muthen & Muthen, 1998)
rather than Cronbach alpha was used to test the factor structure. A key feature of MPlus
is its ability to model factor structure with response variables that are binary,
nonnormally distributed, or both. The analysis showed that a 4-item measure fit the data,
[chi square](df = 2) = 3.41, ns, better than the 6-item measure, [chi square](df = 9) =
28.52, p < .001. These results are consistent with the results of Kurdek and Berg's factor
analysis that indicated that the two items included in the 6-item but not the 4-item scale
had much lower factor loadings than the other items. The following four items were used:
I worry that my parents will want to live without me; It's possible that my parents will
never want to see me again; I worry that I will be left all alone; I think that one day I may

57
have to live with a friend or relative. Reliability was assessed using a confirmatory factor
analytic approach that incorporates both latent theoretical constructs and measured
variables into a single structural equation model (Bollen, 1989; Hayduk, 1987). The
average reliability (squared correlation of the observed variable and its latent variable)
was .53. Thirty-three percent of the children endorsed one or more of the items (22%
endorsed one item, 8% endorsed two items, 2% endorsed three items, and 1% endorsed
four items).

Divorce Stressors. Children reported on the number of negative divorce events that
occurred within the last 3 months on the Divorce Events Schedule for Children (DESC;
Sandler et al., 1986), a "tailor-made" life events scale designed to assess a
representative sample of stressors that children may experience after divorce. Child
report was used because children are the best reporter of their awareness of the
occurrence of negative events, and theoretically, awareness of stressful events is
necessary for primary appraisals of threat, which leads to stress arousal (e.g., Lazarus &
Folkman, 1984). For this scale, stressors were defined as events that typically occur to a
child or in a child's environment following parental divorce and would generally be
perceived as negative by the child. Knowledgeable key informants (i.e., parents and
children who had experienced divorce, mental health professionals, and lawyers who
worked with divorced families) identified over 200 events that they believed had an
important im pact on children after divorce. The research team used these events to
develop nonoverlapping events that did not involve a symptom of a psychological
disorder or physical problem and were primarily beyond the child's control. This process
yielded 62 events (see Sandler et al., 1986, for additional information on scale
development). In a separate sample of children who had experienced parental divorce,
children rated whether each event occurred within the past 3 months and whether the
event was positive, neutral, or negative. To minimize possible contamination of
participants' adjustment and their assessment of the valence of events (e.g., Monroe,
1982), scores were derived using consensually based classification (Sandler et al., 1991;
i.e., events were classified as consensually negative or positive if 80% or more of the
children in the scale development sample who had experienced the event rated it in that
direction). Sixteen of the 62 events were consensually classified as negative; the number
of negative events that occurred is the divorce stressors score. Similar to other life events
scales, the events are heterogenous in content. Examples of negative events are
"Relatives said bad things about mom/dad"; "Dad missed scheduled visits"; "Mom and
dad argued in front of me"; "Parents physically hit/hurt each other"; "I had to give up
pets/toys/things I like." The divorce stressor score correlates with internalizing and
externalizing problems in cross-sectional and short-term longitudinal studies (Sandler et
al., 1986, 1991). Two week test-retest reliability has been shown to be adequate (r = .85;
Sandler, Wolchik, & Braver, 1988).

Mother-Child Relationship Quality. Shortened versions of the acceptance and rejection

subscales of the Children's Report of Parental Behavior Inventory (CRPBI; Schaefer, 1965)
were used to assess child and mother reports of parenting behaviors during the last
month. The CRPBI was originally developed as a child report measure. For the current
project, the items were reworded (i.e., "my mom" replaced with "I") to obtain mother
reports as well. Ten items from the 16-item acceptance subscale were used. Child report
items include "My mom enjoys talking things over with me" and "My mom is able to make
me feel better when I am upset." Ten items from the 16-item rejection subscale were
used as well. Items include "My mom isn't very patient with me" and "My mom forgets to
help me when I need it." These shortened versions contained those items with the
highest item-total correlations and highest factor loadings in a confirmatory factor
analysis conducted with another sample of about 200 divorced families. The CRBPI has
been shown to discriminate normal boys from delinquents and to have adequate internal
consistency (Schaefer, 1965). Fogas, Wolchik, and Braver (1987) report adequate test-
retest reliabilities for mother and child reports of acceptance and rejection. In this
sample, Cronbach's alpha ([alpha]) for mother and child report of acceptance was .81 and

58
.87, respectively; [alpha] for mother and child report of rejection was .79 and .75,
respectively. The acceptance and rejection subscales were combined, with items on the
rejection subscale being reverse coded. Three scores for mother-child relationship quality
were used: child report, mother report, and a composite score, which was formed by
computing the mean of the standardized scores for mothers' and children's reports. The
composite score was used to reduce the problem of rater bias (Schwarz, Barton-Henry, &
Pruzinsky, 1985; Schwarz & Mearns, 1989). In addition, studies have shown that
composite scores are more valid than individual scores as predictors of external criteria
(Eron, 1980; Horowitz, Inouye, & Siegelman, 1979; Moskowitz & Schwarz, 1982).

To examine the possibility that mother-child relationship quality and fear of abandonment
might represent a single latent construct, we tested a model with the fear of
abandonment items, acceptance scale, and rejection scale loading on a single dimension
versus a two-dimensional model in which the fear of abandonment dimension had four
indicators and the mother-child relationship quality had acceptance and rejection
subscales as its indicators. Mplus was applied to test the models because the items for
fear of abandonment were dichotomous. The results indicated that fear of abandonment
is best considered as a separate construct from mother-child relationship quality. For
mother report of the CRPBI, the fit of the two-dimensional model [chi square](df = 4) =
3.33, ns, was much better than the fit of the one-dimensional model, [chi square](df = 5)
= 53.86, p < .001. The difference in fit between the two models was highly significant
([delta][chi square] = 50.53, [delta]df = 1, p < .001). Similarly, for child re port of CRPBI,
the fit of the two-dimensional model [chi square](df = 4) = 3.51, ns, was much better
than the fit of the one-dimensional model, [chi square](df = 5) = 37.966, p <.001. The
difference in fit between the two models was also highly significant ([delta][chi square] =
34.46, [delta]df= 1, p < .001).

Criteria

Externalizing Behavior Problems. Mothers reported on externalizing problems during the

last month on the 33-item Externalizing subscale of the Child Behavior Checklist (CBCL;
Achenbach & Edelbrock, 1983). A large body of research provides support for the
reliability and validity of the CBCL (Achenbach, 1991). In the current sample, [alpha]
was .98. Children reported on externalizing behavior problems that occurred in the last
month using the 28-item Divorce Adjustment Project Externalizing Scale (Arizona State
University Preventive Intervention Research Center, 1985). In this sample, [alpha]
was .96. Composite scores of mothers' and children's reports of externalizing problems
were created. The Time 1 externalizing problems composite score was an average of the
Time 1 standardized scores for mothers' and children's reports. A parallel procedure was
used for the Time 2 externalizing problems composite score. However, to preserve the
metric at Time 1, the Time 2 data were standardized based on Time 1 means and
standard deviations (Pitts, West, & Tein, 1996).

Internalizing Behavior Problems. Mothers reported on internalizing behavior problems

during the last month on the 31-item Internalizing subscale of the CBCL. A large body of
research provides support for the reliability and validity of this subscale (Achenbach,
1991). In the current sample, [alpha] was .98. Children provided reports of depression
and anxiety. Child report of depression was measured using the 27-item Child Depression
Inventory (CDI; Kovacs, 1981), which assesses depressive symptoms during the past 2
weeks. The CDI discriminates clinically depressed and nondepressed psychiatric patients
(e.g., Lobovits & Handel, 1985); it has adequate internal consistency (e.g., Kovacs, 1981)
and test-retest (e.g., Reynolds, 1992) reliability. In this study, [alpha] was .82. Children
completed the 28-item Revised Children's Manifest Anxiety Scale (RCMAS; Reynolds &
Richmond, 1978). The RCMAS has adequate internal consistency (e.g., Reynolds &
Richmond, 1978), test-retest reliability (e.g., Reynolds & Paget, 198 1), and construct
validity (e.g., King, Gullone, Tonge, & Ollendick, 1993). In this study, [alpha] was .84.
Children's reports of internalizing problems were formed by taking the average of the

59
standardized CDI and RCMAS scores to derive a measure that was similar to the measure
of mother report of internalizing problems. Composite internalizing problems scores that
included mother and child report were calculated for Time I and Time 2 using the
procedure described for the externalizing problems composite score.

RESULTS

Means, Standard Deviations, and Descriptive Information

Table I shows the descriptive statistics for all study variables and the correlations
between the study variables. Table II provides the correlations between the study
variables and the correlations of potential covariates (i.e., child age, child gender, family
income, length of separation, and type of access fathers had to their children) with fear of
abandonment and internalizing and externalizing problems. As shown, fear of
abandonment had small-to-moderate significant relations with all measures of
internalizing and externalizing problems at Time 1 and Time 2. Gender was significantly
related to mother and mother/child reports of Time 1 and Time 2 externalizing problems;
boys had higher externalizing problems than did girls. Age was significantly related to
child reports of Time 2 internalizing problems; younger children reported more
internalizing problems. In addition, children who had less access to their fathers had
higher mother/child report of Time 1 internalizing problems.

As shown in Table II, the correlations between Time 1 and Time 2 measures of
internalizing problems and externalizing scores ranged from .52 to .78. Although a
significant proportion of children showed a decrease or no change in problems over time,
there was a tendency for more children to have lower scores at Time 2. On mother report
measures of internalizing and externalizing problems, 59 and 55% of the children had
lower scores at Time 2, respectively. On child report measures of internalizing and
externalizing problems, 48 and 66% had lower scores at Time 2, respectively. There were
no significant differences between Time 1 and Time 2 scores on any measure of
internalizing or externalizing problems.

The mean T-score on mother report of externalizing problems was 56 (SD = 10.0) at Time
1 and 56 (SD = 9.3) at Time 2. (4) The mean T-score on mother report of internalizing
problems was 57 (SD = 10.3) at Time 1 and 56 (SD = 9.3) at Time 2. (4) The proportion of
children who scored in the clinical range (T > 63; Achenbach, 1991) on the CBCL
Externalizing subscale was 26% at Time 1 and 21% at Time 2. (4) The proportion of
children whose scores were in the clinical range (T > 63; Achenbach, 1991) on the CBCL
Internalizing subscale was 31% at Time 1 and 27% at Time 2. (4) The proportion of
children whose scores on the RCMAS exceeded the clinical significance cutoff (T > 60,
Reynolds & Richmond, 1978) was 0.5% at Time 1 and 1% at Time 2. The proportion of
children whose scores exceeded the clinical significance cutoff for the CDI (raw score >
19; Kovacs, 1992) was 0% at Time 1 and 1% at Time 2. Because normative data on the
child report measure of externalizing problems are not available, the percentage of
childre n with clinical significant scores could not be computed. Forty-two percent of the
children scored in the clinical range on one or more measures at Time 1; 36% of the
children scored in the clinical range on one or more measures at Time 2. Time 1 scores
on fear of abandonment were marginally related to having a clinically significant
internalizing or externalizing score at Time 1 (r = .11, p < .10) and significantly related to
having a clinically significant internalizing or externalizing score at Time 2 (r = .14, p
< .05).

Outlier Analyses

A search for multivariate outliers using Cook's distance and Mahalonobis' distance (see
Cook, 1977; Stevens, 1984) was conducted. The results showed that all the data

60
appeared to fit the model and should not bias parameter estimates.

Mediational Model and Analytic Plan

As shown by the thicker lines in Fig. 1, we hypothesized that Time 1 fear of abandonment
mediates the effect of Time 1 divorce stressors and mother--child relationship quality on
Time 2 internalizing and externalizing problems, controlling for Time 1 internalizing or
externalizing problems. Time 1 high divorce stressors and low mother--child relationship
quality were expected to be associated with high levels of Time 1 fear of abandonment,
which, in turn, were expected to lead to increases in internalizing and externalizing
problems at Time 2. Furthermore, we hypothesized that Time 1 mother--child relationship
quality would moderate the relation between Time 1 divorce stressors on Time 1 fear of
abandonment; the relation was expected to be weaker for children with high mother--
child relationship quality than that for children with low mother--child relationship quality.
We tested the mediational and moderational relations of the variables with structural
equation modeling software EQS (Bentler, 1995).

On the basis of significant correlations between internalizing or externalizing problems

with age, gender, and the type of access fathers had to their children, these three
variables were included as covariates in prediction of Time 1 fear of abandonment and
Time 2 internalizing and externalizing problems. Income and length of separation were
not significantly related to the study variables and thus neither was included in the
model. As convention for model fit, Time 1 divorce stressors, mother--child relationship
quality, internalizing problems, and externalizing problems were treated as being
correlated. However, for the simplicity of presentation, these correlations are omitted
from the figures. The magnitude of the correlations of these variables was similar to
those of the zero-order correlations depicted in Table II.

To test the moderation effect, an interaction term of divorce stressors by mother--child

relationship quality was specified. To reduce potential multicollinearity, measures of
divorce stressors and mother--child relationship quality were centered (Aiken & West,
1991). A significant interaction would indicate a potential moderation effect of mother--
child relationship quality on the relation between divorce stressors and fear of
abandonment. The direction of the moderation effect would depend on the sign of this
interaction, with a negative sign indicating that the positive relation between stressors
and fear of abandonment is attenuated as mother--child relationship quality increases.

According to recent guidelines for establishing potential mediational pathways (see Judd
& Kenny, 1981b; MacKinnon, Krull, & Lockwood, 2000), two sets of hypotheses should be
supported to test for mediation: (a) Time 1 divorce stressors should significantly relate to
Time 1 fear of abandonment, and (b) Time 1 fear of abandonment should predict Time 2
internalizing/externalizing problems after controlling for the temporal stability of the
problems (i.e., Time 1 score; see MacCallum & Austin, 2000). The same hypotheses apply
to the link between Time 1 mother--child relationship quality, Time 1 fear of
abandonment, and Time 2 internalizing/externalizing problems.

Although the earlier research on mediation (e.g., Baron & Kenny, 1986; Judd & Kenny,
1981b) specified that the independent variable (i.e., divorce stressors, mother--child
relationship quality) by itself needs to be significantly related to the outcome variable,
MacKinnon et al. (2000) argued that under the circumstance where an inconsistent
mediation effect exists, the zero-order correlation between the independent and
dependent variables may not be significant (also see Rosenberg, 1968). An inconsistent
model occurs when the direct and mediated effects of an independent variable on a
dependent variable have opposite signs.

Given support for these hypotheses, the effects of fear of abandonment in mediating the

61
relations between divorce stressors/mother--child relationship quality and
internalizing/externalizing problems are tested for significance using the multivariate-
delta method (MacKinnon, 1994; Sobel, 1982, 1986). This method calculates the standard
error of a product of two random variables (e.g., the product of the path coefficient from
the independent variable to the mediator variable and the path coefficient from the
mediator variable to the outcome variable), which is then used to test the significance of
the mediation effects (i.e., z = ab/[SE.sub.ab]; z [greater than or equal to] 1.645, for p
[less than or equal to] .05, one-tailed test). If the interaction between the stressors and
mother--child relationship quality were significant, the magnitude of the mediational link
between divorce stressors, fear of abandonment, and internalizing/externalizing problems
would depend on mother--child relationship quality. The m oderated mediation effect
would then be tested following the procedures outlined by Tein, Sandler, MacKinnon, and
Wolchik (2000).

To allow an examination of the robustness of the findings, three tests of this model were
conducted. One model used composite scores for mother--child relationship quality,
externalizing problems, and internalizing problems, and children's reports of divorce
stressors and fear of abandonment. The use of composite measures has the advantage of
helping to ensure that the full breadth of the construct is represented and that errors of
measurement and rater bias are minimized for constructs with multiple measures (e.g.,
Epstein, 1983; Schwarz et al., 1985). The second model, the child model, is a within-
reporter model that used child report of all variables. Although all variables in this model
share the same reporter, the method variance due to reporter in the Time 2 adjustment
problems measures is accounted for by controlling for the Time 1 adjustment problems
score. The third model, the cross-reporter model, used mother report of mother--child
relationship quality, internalizing problems, and externalizing pro blems, and child report
of divorce stressors and fear of abandonment. The use of mother report of some variables
and child report of others reduces concerns that the significant effects are due to
common reporter method variance.

Data for participants who had missing scale scores on study variables were imputed with
NORM (Schafer, 1997; Schafer & Olsen, 1998), a multiple imputation (MI) (5) procedure
for multivariate normal data that takes uncertainty of missing data into account. The
imputation process replaces each missing value by a plausible value and then transforms
the data set with missing values into a complete data set. The choice to use imputation
was based on concerns that analyses that include only participants with complete data
have limitations in terms of generalizability and statistical power (e.g., Little & Rubin,
1987; Schafer & Olsen, 1998).

Model Tests

The results showed that all three models were plausible (i.e., Bentler Comparative Fit
Index [CFI] > .90; Bentler, 1990): composite model-[chi square](df = 19) = 40.97, p
< .05, CFI = .96; childmodel-[chi square](df = 19) = 46.93, p < .01, CFI = .94; and cross-
reporter model-[chi square](df = 19) = 31.32, ns, CFI = .98. Figure 2(a) shows the results
of the model using composite variables for mother-child relationship quality, internalizing
problems, and externalizing problems. The paths from the covariates to Time 1 fear of
abandonment, Time 2 internalizing problems, and Time 2 externalizing problems were
omitted for a clearer presentation of the hypothesized mediation and moderation effects.
As shown, Time 1 divorce stressors were significantly, positively related to Time 1 fear of
abandonment (unstandardized path coefficient, b = .17, SE = .05, p < .01) and Time 1
mother-child relationship quality was significantly, negatively related to Time 1 fear of
abandonment (b = -.19, SE = .08, p < .05) after controlling for the covariate variables
(i.e., age, gender, and father access). There were also significant effects of Time 1 fear of
abandonment on both Time 2 externalizing problems(b = .13, SE = .05, p < .01) and
internalizing problems (b = .12, SE = .05, p < .05), after controlling for the covariates and
Time 1 externalizing or internalizing problems. According to the criteria mentioned

62
earlier, these significance tests implied that fear of abandonment mediated the relations
between both divorce stressors and mother-child relationship quality and internalizing
externalizing problems. Results of the multivariate delta method showed that Time 1 fear
of abandonment mediated the effects of Time 1 divorce stressors on both Time 2
externalizing and internalizing problems (z = 2.10, p < .01, for externalizing problems; z
= 1.89, p < .05, for internalizing problems). Similarly, Time 1 fear of abandonment
mediated the effects of Time 1 mother-child relationship quality on internalizing
externalizing problems (z = -1.82, p < .05, f or externalizing problems; z = -1.67, p < .05,
for internalizing problems). The model also indicates that Time 1 fear of abandonment
had a complete mediation effect for Time 1 divorce stressors on Time 2 internalizing and
for Time 1 mother-child relationship quality on Time 2 internalizing and externalizing
problems. As shown in Table II, the zero-order correlations between Time 1 divorce
stressors and Time 2 internalizing problems as well as between Time 1 mother-child
relationship quality and Time 2 internalizing and externalizing problems were significant.
However, after taking the indirect effects through Time 1 fear of abandonment into
account, the direct effects of Time 1 divorce stressors on Time 2 internalizing problems
and Time 1 mother-child relationship quality on Time 2 internalizing and externalizing
problems were no longer significant. The interaction between Time 1 mother-child
relationship quality and Time 1 divorce stressors was marginal (b = -.14, SE = .08, p
< .06); the relation between div orce stressors and fear of abandonment was attenuated
as mother-child relationship quality increased. Age was negatively related to Time 2
internalizing problems (b = -.09, SE = .04, p <.01) after controlling for Time 1
internalizing problems and the other covariates; younger children were likely to have
higher internalizing problems than older children.

The results for the child report model and the cross-reporter model were generally
consistent with those of the model that used composite measures. As in the composite
model, the relation between Time 1 divorce stressors and Time 1 fear of abandonment
was significant and positive in both models. In the child report model, the effects of Time
1 fear of abandonment on Time 2 externalizing and internalizing problems were both
significant after controlling for their respective Time 1 measure and the covariates, b
= .22, SE = .08, p < .01; b = .18, SE = .08, p < .05, respectively; see Fig. 2(b). Results of
the multivariate-delta method indicated that Time 1 fear of abandonment mediated the
effects of Time 1 divorce stressors on both Time 2 internalizing problems (z = 1.89, p
< .05) and externalizing problems (z = 2.12, p < .01). Time 1 fear of abandonment had a
complete mediation effect for Time 1 divorce stressors on Time 2 internalizing problems.
Time 1 mother--child relationship quality was significantly, negatively related to fear of
abandonment (b = -.14, SE = .06, p < .05). The interaction between Time 1 divorce
stressors and Time 1 mother--child relationship quality was not significant (b = -.10, SE
= .07, ns). Time 1 fear of abandonment also mediated the relations between Time 1
mother--child relationship quality and Time 2 externalizing problems (z = -1.73, p < .05)
and Time 2 internalizing problems (z = -1.64, p < .05). Time 1 fear of abandonment
completely mediated the relations between Time 1 mother--child relationship quality on
Time 2 externalizing problems.

For the cross-reporter model, which used mother report of mother--child relationship
quality and children's adjustment problems, there was a significant effect of Time 1 fear
of abandonment on Time 2 internalizing problems after controlling for Time 1
internalizing problems and the covariates, b = .80, SE = .41, p < .05; see Fig. 2(c).
Although it was in the expected direction, the effect of Time 1 fear of abandonment on
Time 2 externalizing problems was not significant (b = .65, SE = .47, ns). Results of the
multivariate-delta method indicated that Time 1 fear of abandonment mediated the
relations between Time 1 divorce stressors and Time 2 internalizing problems (z = 1.71, p
< .05). After controlling for Time 1 internalizing problems, the covariate variables, and
the mediation effect, there was a direct, negative effect of Time 1 divorce stressors on
Time 2 internalizing problems (b = - .74, SE = .31, p < .05). This is referred to as
inconsistent mediation (Mackinnon et al., 2000; Rosenberg, 1968). Time 1 m other--child

63
relationship quality was not related to Time 1 fear of abandonment (b = -.08, SE = .06,
ns). The interaction between Time 1 divorce stressors and Time 1 mother--child
relationship quality was marginal in the negative direction (b = -.10, SE = .06, p < .10).

DISCUSSION

The findings of this prospective study extend our understanding of how the stressors
associated with divorce lead to children's postdivorce adjustment problems. The most
robust findings occurred for internalizing problems. In all three models, fear of
abandonment consistently predicted internalizing problems as reported by both mothers
and children and tests of mediation indicated that the effects of divorce stressors on
internalizing problems were mediated through fear of abandonment. In two models, fear
of abandonment mediated the effects of divorce stressors on externalizing problems as
well. Further, two of the three models provided support for fear of abandonment as a
mediator of the relation between mother--child relationship quality and both internalizing
and externalizing problems.

An important methodological strength of this study is the use of a prospective,

longitudinal design. The consistency of the findings across prospective tests of models
indicates that the findings cannot be accounted for by the effects of internalizing or
externalizing problems on fear of abandonment, divorce stressors, or mother--child
relationship quality. Another strength involves the use of multiple reporters. The child and
composite models are likely inflated because the same informant provides information
about the independent, mediating, and dependent variables. However, concern that the
effect of Time 1 fear of abandonment on Time 2 adjustment problems is accounted for by
shared reporter method variance is mitigated because the reporter method variance is
accounted for by controlling for Time 1 adjustment problems scores. Further, the
consistency of effects across the models and the pattern of findings in the cross-reporter
model indicate that the mediational relations for fear of abandonment cannot be
accounted for by shared reporter method variance.

The current findings provide linkages between several areas of prior work. Specifically,
they tie together clinical observations that divorce-related disruptions in social
relationships and the physical environment evoke threats to children's sense of
belongingness and sense of security (e.g., Wallerstein & Kelly, 1980) and research
documenting cross-sectional and longitudinal relations between divorce stressors and
adjustment problems (e.g., Sandler et al., 1986, 1991; Stolberg & Anker, 1983). Also, the
findings on the mental health implications of fear of not being taken care of by one's
family are consistent with theoretical and empirical work on the relations between need
for relatedness and attachment organization (see Baumeister & Leary, 1995, and
Greenberg, 1999, for reviews) and mental and physical health outcomes. The findings
also provide support for the theoretical proposition that the effects of stressors and the
protective function of mother-child relationships on children's adjustment problems a re
mediated through the pathway of threat to a basic motivational need to be part of a
caring social group (Sandler, 2001).

All three models indicated that fear of abandonment mediated the relation between Time
1 divorce stressors and Time 2 internalizing problems. For the composite and child
models, complete mediation occurred; for the cross-reporter model, partial mediation
occurred. It is interesting to speculate about the processes that explain the mediating
role of fear of abandonment. It is well documented that divorce leads to disruptions in
several salient social relationships (e.g., Sandler et al., 1986, 1991). For many children,
time with one or both parents is reduced markedly because of custody arrangements and
increased work hours. Also, time with relatives and friends is often decreased due to the
restructuring of the residential parent's social network, as well as change in
neighborhoods or schools. The occurrence of many such disruptions is likely to cause

64
children to reevaluate the security or stability of their central social ties. Such
reevaluation may lead to concerns about the continuity of caregiver relations hips,
operationalized in the current study as fear of abandonment. The link between fear of
abandonment and internalizing problems may be explained by ruminations about
whether one's basic needs for care will be met. These ruminations may interfere with
problem-solving efforts, increase self-focus, or lead to feelings of helplessness (Morrow &
Nolen-Hocksema, 1990; Vasey, 1993; Wolchik et al., 1993). It is important to note that
inconsistent mediation occurred (i.e., the direct and mediated effects of divorce stressors
on change in internalizing problems had opposite signs) in one of these three models.
Although it is possible that this finding is artifactual, it is also plausible that high levels of
divorce stressors at Time 1 may mobilize protective factors such as active coping and
social support that decrease internalizing problems over time. It will be important for
future research to test this theoretical proposition.

Fear of abandonment was also a significant mediator of the relations between divorce
stressors and externalizing problems in two of the models tested, the child report model
and the composite model. High fear of abandonment may generate anger toward
mothers, which reduces their ability to implement behavior controls (Allen, Moore,
Kuperminc, & Bell, 1998). It is also plausible that high fear of abandonment may lead
children to seek support from disaffected, antisocial peers, leading to increased behaviors
consistent with the values of these peers (Kaplan, 1984). Alternatively, externalizing
problems may reflect an extreme method of obtaining a response from a caregiver and
thus could be viewed as a way to restore a sense of connection with mothers (e.g., Allen
et al., 1998; Allen & Land, 1999). Further, it is possible that the mediational relation is
due to low levels of monitoring and supervision that may co-occur with high levels of fear
of abandonment and divorce stressors. Future research should examine m odels that
include monitoring and supervision.

Fear of abandonment completely mediated the relation between mother-child

relationship quality and externalizing problems in both the composite and child models
and internalizing problems in the composite model. These findings extend the literature
on mother--child relationship quality and postdivorce adjustment (e.g. Hetherington et al.,
1992; Wolchik et al., 2000) by articulating a plausible mechanism through which this
relationship affects adjustment problems. Warm, affectionate, and affirming mother-child
relationships may provide concrete evidence that this relationship will endure overtime,
which reduces both children's fears of not being cared for and their impact on
internalizing and externalizing problems.

The mediational relation between relationship quality, fear of abandonment, and

externalizing and internalizing problems may also be viewed from the perspective of
attachment theory. It might be argued that children's reports of fear of abandonment are
tapping an aspect of children's internal working models of attachment. Indeed, concepts
of loss and insecurity of attachment relationships are key components of many
conceptualizations of attachment (Bowlby, 1969/1982; Bretherton, 1995; Greenberg,
1999). On the other hand, it seems overly simplistic to believe that a 4-item self-report
scale is an adequate measure of attachment, and it may be that fear of abandonment is a
response to proximal situational stressors and changes in parent--child relationships
following divorce, rather than reflecting a more enduring internal working model of
attachment. Nevertheless, given the empirical support for the importance of fear of
abandonment as a predictor of adjustment problems in children of divorce, further
research should assess the relations between attachment organization, fear of
abandonment, and mother-child relationship quality.

In addition to testing the mediational role of fear of abandonment, the current study
provided some descriptive information on this construct. Fear of abandonment did not
differ across age or gender. Although few children reported high levels of fear of
abandonment, high scores were positively associated with all continuous measures of

65
adjustment problems both cross-sectionally and longitudinally. Of particular importance
from a clinical perspective, high fear of abandonment significantly predicted clinically
meaningful levels of adjustment problems over time. Although the models that were
tested pertain to the full range of changes in adjustment problems over time (both
improvements and decrements) further research might focus on understanding the role
of fear of abandonment in explaining why some children of divorce are on a trajectory
toward serious behavior problems over time and why others adapt well to divorce after
the initial period of disruption and distress.

The results of this study have both applied and theoretical implications. As increasingly
argued by prevention researchers, work that identifies potentially modifiable mediators
and moderators of empirically supported correlates of children's adjustment to stressors
should inform the development of effective programs (Lipsey, 1990; Lorion, Price, &
Eaton, 1989; Sandler, Wolchik, MacKinnon, Ayers, & Roosa, 1997). The current findings
suggest that reducing children's fear of abandonment should be a central focus of such
interventions. One strategy that could be included in mother-focused interventions
involves improving mother-child relationship quality. Illustratively, the New Beginnings
Program, a preventive intervention for divorced mothers, includes components on one-
on-one time with children, enjoyable family activities, and effective listening and
communication skills (Wolchik, West, Westover, Sandler, Martin, Lustig, et al., 1993;
Wolchik, West, Sandler, Tein, Coatsworth, Lengua, et al., 2000). The progr am produced
significant effects on children's adjustment problems that were partially mediated by
improvement in mother--child relationship quality (Tein, Sandler, MacKinnon, & Wolchik,
2000; Wolchik et. al., 1993). It may be that the effect of improved mother--child
relationship quality on adjustment problems is in part mediated by decreases in children's
fears that they will not be cared for following the divorce. Also, children may benefit from
explicit reassurance that the mother--child relationship will endure over time. The results
have implications for identifying children of divorce who are at risk for developing
adjustment problems. Children who experience high levels of divorce stressors, high fear
of abandonment, or low mother-child relationship quality are a particularly high-risk
group that could benefit from interventions.

The limitations of this study suggest several areas for future research. First, the sample
was largely Caucasian/non-Hispanic. Future research that examines ethnic similarities
and differences in mediating mechanisms of the effects of divorce would provide a
foundation for developing culturally competent interventions. Given the increasingly
multicultural nature of our society (America's Children, 1998; U.S. Bureau of Census,
1996), the public health significance of such research is clear. Second, the average length
of time since parental separation was about 2 years. Fear of abandonment is likely to be
higher immediately after parental separation. Future research should examine the
relations of fear of abandonment, divorce stressors, and internalizing and externalizing
problems earlier in the process of divorce. Third, although the current study recruited
participants using a random sample of divorce records, the inability to reach potential
participants, use of numerous eligibility criteria, referral for tr eatment of families with
children with suicidal ideation, refusal, and attrition that occurred between the two
assessments resulted in significant loss of potential participants. The attrition analysis
suggests that the findings should not be generalized to children who experience high
levels of divorce stressors, fear of abandonment, or depression. Tests of the current
model with a more representative sample is an important direction for future work.
Fourth, this study examined only one plausible mediator of the relation between divorce
stressors and adjustment problems. The small-to-moderate magnitude of the relations
between fear of abandonment and internalizing and externalizing problems suggests that
research should examine more complex models that include additional processes that
may contribute to children's postdivorce adjustment. For example, future studies could
include measures of attachment organization, active coping, self-esteem as well as fear
of abandonment. Also, future studies could examine the contribution of father--child
relationship quality to fear of abandonment and assess whether a warm, affectionate

66
father--child relationship mitigates the effect of a rejecting mother--child relationship on
fear of abandonment.
Table I

Descriptive Data for Predictor and Criterion Variables

Measure Reporter Mean Range SD

Divorce stressors 1 Child 3.12 0-12 2.37

Relationship quality 1 Mother 53.60 21-66 4.59
Relationship quality 1 Child 51.37 20-60 7.41
Relationship quality 1 Mother/child 0.00 -2.60-1.31 0.70
Fear of abandonment 1 Child 0.47 0-4 0.78
Internalizing problems 1 Mother 9.87 0-33 6.45
Internalizing problems 1 Child -0.08 -1.46-2.03 0.74
Depression 6.22 0-18 4.40
Anxiety 9.18 0-24 5.57
Internalizing problems 1 Mother/child -0.06 -1.58-1.91 0.69
Externalizing problems 1 Mother 13.77 0-39 8.21
Externalizing problems 1 Child 9.76 0-38 6.76
Externalizing problems 1 Mother/child -0.02 -1.44-2.31 0.80
Internalizing problems 2 Mother 10.05 0-39 6.51
Internalizing problems 2 Child 0.00 -1.18-3.50 0.90
Depression 5.12 0-24 4.66
Anxiety 8.04 0-27 6.34
Internalizing problems 2 Mother/child 0.00 -1.35-2.48 0.75
Externalizing problems 2 Mother 13.73 0-43 7.88
Externalizing problems 2 Child 8.35 0-42 7.48
Externalizing problems 2 Mother/child 0.00 -1.43-2.82 0.79

Measure Kurtosis Skewness Valid N

Divorce stressors 1 0.90 0.90 216

Relationship quality 1 1.74 -1.07 216
Relationship quality 1 3.16 -1.55 199
Relationship quality 1 0.67 -0.84 216
Fear of abandonment 1 3.07 1.79 216
Internalizing problems 1 0.17 0.64 213
Internalizing problems 1 -0.59 0.36 216
Depression -0.372 0.66 216
Anxiety -0.742 0.41 199
Internalizing problems 1 -0.09 0.42 213
Externalizing problems 1 0.29 0.67 214
Externalizing problems 1 0.92 0.91 216
Externalizing problems 1 -0.09 0.59 214
Internalizing problems 2 1.83 1.07 201
Internalizing problems 2 0.98 1.07 216
Depression 2.45 1.41 213
Anxiety 0.08 0.92 212
Internalizing problems 2 0.42 0.81 201
Externalizing problems 2 1.34 0.94 209
Externalizing problems 2 3.15 1.56 213
Externalizing problems 2 0.94 0.86 206

Note. Mother/child reports of relationship quality, internalizing

problems, and externalizing problems were composite scores of the
standardized mothers' and children's reports of the corresponding
measures. Children's reports of internalizing problems were composite
scores of standardized CDI and RCMAS scores.
Table II

Zero-Order Correlations of the Study Variables

1 2 3 4 5 6 7

1. Divorce stressors 1.00

1 - Child report
2. Relationship quality .02 1.00
1 - Mother report
3. Relationship quality -.08 .21 1.00
1 - Child report
4. Relationship quality -.08 .75 .79 1.00
1 - Mother/child report
5. Fear of abandonment .25 -.06 -.15 -.15 1.00
1 - Child report
6. Internalizing problems .04 -.41 -.09 -.32 .14 1.00
1 - Mother report
7. Internalizing problems .24 -.19 -.20 -.28 .38 .15 1.00
1 - Child report

67
 8. Internalizing problems .17 -.41 -.19 -.40 .33 .81 .70
1 - Mother/child report
9. Externalizing problems .13 -.45 -.17 -.42 .14 .49 .25
1 - Mother report
10. Externalizing problems .15 -.11 -.16 -.20 .22 .15 .46
1 - Child report
11. Externalizing problems .17 -.35 -.20 -.39 .23 .40 .44
1 - Mother/child report
12. Internalizing problems -.01 -.34 -.03 -.26 .20 .73 .15
2 - Mother report
13. Internalizing problems .24 -.11 -.08 -.15 .32 .12 .52
2 - Child report
14. Internalizing problems .15 -.31 -.08 -.28 .32 .57 .44
2 - Mother/child report
15. Externalizing problems .07 -.37 -.20 -.40 .16 .42 .23
2 - Mother report
16. Externalizing problems .11 -.06 -.17 -.15 .27 .11 .39
2 - Child report
17. Externalizing problems .11 -.27 -.25 -.36 .27 .33 .40
2 - Mother/child report
18. Child gender -.14 -.11 -.26 -.21 -.11 .01 -.04
19. Child age -.13 .01 .04 .02 -.08 .10 -.04
20. Type of father access .13 -.03 -.11 -.07 .06 .12 .09
21. Length of separation -.05 .07 .04 .07 .06 -.01 -.02
22. Income .01 .15 .06 .16 -.07 -.12 -.04

8 9 10 11 12 13 14

1. Divorce stressors
1 - Child report
2. Relationship quality
1 - Mother report
3. Relationship quality
1 - Child report
4. Relationship quality
1 - Mother/child report
5. Fear of abandonment
1 - Child report
6. Internalizing problems
1 - Mother report
7. Internalizing problems
1 - Child report
8. Internalizing problems 1.00
1 - Mother/child report
9. Externalizing problems .49 1.00
1 - Mother report
10. Externalizing problems .39 .29 1.00
1 - Child report
11. Externalizing problems .55 .82 .79 1.00
1 - Mother/child report
12. Internalizing problems .62 .33 .08 .26 1.00
2 - Mother report
13. Internalizing problems .40 .29 .44 .46 .11 1.00
2 - Child report
14. Internalizing problems .67 .42 .35 .48 .74 .75 1.00
2 - Mother/child report
15. Externalizing problems .44 .77 .25 .64 .47 .21 .46
2 - Mother report
16. Externalizing problems .32 .31 .65 .59 .02 .66 .46
2 - Child report
17. Externalizing problems .48 .68 .57 .78 .32 .56 .58
2 - Mother/child report
18. Child gender -.02 .30 .12 .26 -.00 -.07 -.04
19. Child age .04 .05 .09 .08 -.04 -.16 -.12
20. Type of father access .13 .09 -.02 .05 .07 -.02 .04
21. Length of separation -.02 .01 -.01 -.00 .04 .05 .06
22. Income -.11 -.09 .08 .00 -.12 -.02 -.09

15 16 17 18 19 20 21

1. Divorce stressors
1 - Child report
2. Relationship quality
1 - Mother report
3. Relationship quality
1 - Child report
4. Relationship quality
1 - Mother/child report
5. Fear of abandonment
1 - Child report
6. Internalizing problems
1 - Mother report

68
 7. Internalizing problems
1 - Child report
8. Internalizing problems
1 - Mother/child report
9. Externalizing problems
1 - Mother report
10. Externalizing problems
1 - Child report
11. Externalizing problems
1 - Mother/child report
12. Internalizing problems
2 - Mother report
13. Internalizing problems
2 - Child report
14. Internalizing problems
2 - Mother/child report
15. Externalizing problems 1.00
2 - Mother report
16. Externalizing problems .24 1.00
2 - Child report
17. Externalizing problems .79 .79 1.00
2 - Mother/child report
18. Child gender .30 .05 .22 1.00
19. Child age .03 .05 .04 .05 1.00
20. Type of father access .07 .01 .05 .00 .12 1.00
21. Length of separation .08 .05 .09 -.06 -.02 .16 1.00
22. Income -.09 .06 -.03 -.08 -.06 -.12 -.04

22

1. Divorce stressors
1 - Child report
2. Relationship quality
1 - Mother report
3. Relationship quality
1 - Child report
4. Relationship quality
1 - Mother/child report
5. Fear of abandonment
1 - Child report
6. Internalizing problems
1 - Mother report
7. Internalizing problems
1 - Child report
8. Internalizing problems
1 - Mother/child report
9. Externalizing problems
1 - Mother report
10. Externalizing problems
1 - Child report
11. Externalizing problems
1 - Mother/child report
12. Internalizing problems
2 - Mother report
13. Internalizing problems
2 - Child report
14. Internalizing problems
2 - Mother/child report
15. Externalizing problems
2 - Mother report
16. Externalizing problems
2 - Child report
17. Externalizing problems
2 - Mother/child report
18. Child gender
19. Child age
20. Type of father access
21. Length of separation
22. Income 1.00

Note. p [less than or equal to] .05 when r [greater than or equal to]
.13;

p [less than or equal to] .01 when r [greater than or equal to] .17.
Type of father access was coded as 1 (unrestricted contact), 2 (regular
access), 3 (occasional access), and 4 (no contact at all).

ACKNOWLEDGMENTS

69
This research was supported by National Institute of Mental Health Grant #P30MH39246.
The authors thank Philip Poirier, Linda Ruehlman, Spring Dawson-McClure, Kathleen
Hipke, and Ernest Fairchild for their assistance with this paper. We also thank the
mothers and children for their willingness to share their experiences with our research
team.

Received July 3, 2002; revision received January 10, 2002; accepted January 14, 2002

(4.) The CBCL T-scores and clinical cutpoints are based on norms that use a 6-month
rather than 1-month reporting time frame. Thus, these figures are likely to underestimate
the prevalence of clinically significant behavior problems, given that 1-month scores are
expected to be lower than 6-month scores.

(5.) MI assumes that data that are missing are "missing at random." As a Bayesian
procedure, MI solves a missing data problem by replacing each missing observation in a
stochastic or random fashion. With NORM, the imputation process replaces each missing
value by a plausible value drawn from the predictive distribution of the missing data and
transforms the data set with missing values to a complete data set. Repeating the
imputation process, M complete data sets are generated. Each M data set is then
analyzed independently and the results of the M analyses are subsequently combined,
following the procedure described by Schafer and Olsen (1998), to produce one overall
estimate for each parameter of interest. The resulting estimates incorporate the within-
imputation variation, a measure of ordinary sampling variation, and the between-
imputation variation, a measure of the uncertainty due to the missing data.

REFERENCES

Achenbach, T. M. (1991). Manual for the Child Behavior Checklist/4-18 and 1991 Profile.
Burlington, VT: University of Vermont, Department of Psychiatry.

Achenbach, T. M., & Edelbrock, C. (1983). Manual for the Child Behavior Checklist and
Revised Child Behavior Profile. Burlington, VT: University of Vermont, Department of
Psychiatry.

Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting
interactions. Newbury Park, CA: Sage.

Allen, J. P., & Land, D. (1999). Attachment in adolescence. In J. Cassidy & P. R. Shaver
(Eds.), Handbook of attachment: Theory, research and clinical applications. New York:
Guilford Press.

Allen, J. P., Moore, C. M., Kuperminc, G. P., & Bell, K. L. (1998). Attachment and
adolescent psychosocial functioning. Child Development, 69, 1406-1419.

Amato, P. R., & Keith, B. (1991a). Parental divorce and the well-being of children: A meta-
analysis. Psychological Bulletin, 110, 26-46.

America's Children. (1998). Key national indicators of well-being. Washington, DC: Federal
Interagency Forum on Child and Family Statistics.

Arizona State University Preventive Intervention Research Center. (1985). The divorce
adjustment project externalizing scale. Tempe, AZ. Unpublished document, Arizona State
University, Program for Prevention Research.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator distinction in social

psychological research: Conceptual, strategic, and statistical considerations. Journal of

70
Personality and Social Psychology, 51, 1173-1182.
Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal
attachments as a fundamental human motivation. Psychological Bulletin, 117, 497-529.

Bentler, P. M. (1990). Fit indexes comparative in structural models. Psychological Bulletin,

107, 238-246.

Bentler, P. M. (1995). EQS structural equations program manual. Encino, CA: Multivariate
Software, Inc.

Bollen, K. (1989). Structural equations with latent variables. New York: Wiley.

Bowlby, J. (1973). Attachment and loss: Vol. 3. Loss. New York: Basic Books.

Bowlby, J. (1980). Attachment and loss: Vol. 3. Loss. New York: Basic Books.

Bowlby, J. (1969/1982). Attachment and loss: Vol. 1. Attachment. New York: Basic Books.

Bretherton, I. (1995). Attachment theory and developmental psychopathology. In D.

Cicchetti & S. L. Toth (Eds.), Rochester Symposium on Developmental Psychopathology:
Vol. 6. Emotion, cognition, and representation (pp. 231-260). Rochester, NY: University of
Rochester Press.

Bretherton, I., & Munholland, K. A. (1999). Internal working models in attachment

relationships: A construct revisited. In J. Cassidy & P. R. Shaver (Eds.), Handbook of
attachment: Theory, research and clinical applications. New York: Guilford Press.

Camara, K. A., & Resnick, G. (1987). Marital and parental subsystems in mother-custody,
father-custody, and two-parent households: Effects on children's social development.
Advances in Family Intervention, Assessment and Theory, 4, 165-196.

Cassidy, J., & Shaver, P. R. (1999). Handbook of attachment: Theory, research, and
clinical applications. New York: Guilford Press.

Chase-Lansdale, P. L., Cherlin, A. K., & Kiernan, K. E. (1995). The long-term effects of
parental divorce on the mental health of young adults: A developmental perspective.
Child Development, 66, 1614-1634.

Cherlin, A. J. (1992). Marriage, divorce, remarriage. Cambridge, MA: Harvard University

Press.

Cook, R. D. (1977). Detection of influential observation in linear regression.

Technometrics, 19, 15-18.

Dozier, M., Stevenson, A. L., Lee, S. W., & Velligan, D. L. (1991). Attachment organization
and familial overinvolvement for adults with serious psychopathological disorders.
Development and Psychopathology, 3, 475-489.

Epstein, S. (1983). Aggregation and beyond: Some basic issues on the prediction of
behavior. Journal of Personality, 51, 360-392.

Eron, L. D. (1980). Prescription for reduction of aggression. American Psychologist, 35,

244-252.

Fogas, B. S., Wolchik, S. A., & Braver, S. L. (1987, August). Parenting behavior and

71
psychopathology in children of divorce: Buffering effects. Paper presented at the
American Psychological Association Convention, New York.

Gardner, R. A. (1976). Psychotherapy with children of divorce. New York: Jason Aronson.

Greenberg, M. T. (1999). Attachment and psychopathology in childhood. In J. Cassidy & P.

R. Shaver (Eds.), Handbook of attachment: Theory, research, and clinical applications (pp.
469-496). New York: Guilford Press.
Hayduk, L. (1987). Structural equation modeling with lisrel, essentials and advances.
Baltimore: The John Hopkins University Press.

Hetherington, E. M. (1999). Should we stay together for the sake of the children? In E. M.
Hetherington (Ed.), Coping with divorce, single parenting, and remarriage: A risk and
resiliency perspective (pp. 93-116). Mahwah, NJ: Erlbaum.

Hetherington, E. M., Clingempeel, W. G., Anderson, E. R., Deal, J. E., Stanley Hagan, M.,
Hollier, E. A., et al. (1992). Coping with marital transitions: A family systems perspective.
Monographs of the Society for Research in Child Development, 57(2/3, Serial No. 227).

Hetherington, E. M., Cox, M., & Cox, R. (1982). Effects of divorce on parents and children.
In M. E. Lamb (Ed.), Nontraditional families: Parenting and child development (pp. 233-
285). Hillsdale, NJ: Erlbaum.

Horowitz, L. M., Inouye, D., & Siegelman, E. Y. (1979). On averaging judges' ratings to
increase their correlation with an external criterion. Journal of Consulting and Clinical
Psychology, 47, 453-458.

Judd, C. M., & Kenny, D. A. (1981a). Process analysis: Estimating mediation in treatment
evaluation. Evaluation Review, 5, 602-619.

Judd, C. M., & Kenny, D. A. (1981b). Estimating the effects of social interventions. New
York: Cambridge University Press.

Kaplan, H. B. (1984). Self-attitudes and deviant responses: Toward the validation of a

general theory of deviant behavior. In S. Mednick, M. Halway, & K. Finello (Eds.),
Handbook of longitudinal research: Teenage and adult cohorts (Vol. 2, pp. 141-150). New
York: Praeger.

King, N., Gullone, E., Tonge, B. J., & Ollendick, T. H. (1993). Self-reports of panic attacks
and manifest anxiety in adolescents. Behavior Research and Therapy, 31(1), 111-116.

Kovacs, M. (1981). Rating scales to assess depression in school-aged children. Acta

Pacedopsychiatrica, 46, 305-315.

Kovacs, M. (1992). Child Depression Inventory: CDI Manual. North Tonawanda, NY: Multi-
Health Systems.

Kurdek, L. A., & Berg, B. (1987). Children's beliefs about parental divorce scale:
Psychometric characteristics and concurrent validity. Journal of Consulting and Clinical
Psychology, 55, 712-718.

Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal and coping. New York: Springer.

Lipsey, M. (1990). Theory as method: Small theories of treatments. In L. Sechrest, E.

Perrin, & J. Bunker (Eds.), Research methodology: Strengthening causal interpretations of

72
non-experimental data (pp. 44-53). Washington, DC: U.S. Department of Health and
Human Services.

Little, R. J. A., & Rubin, D. B. (1987). Statistical analysis with missing data. New York:
Wiley.

Lobovits, D. A., & Hendal, P. J. (1985). Childhood depression: Prevalence using DSM-III
criteria and validity of parent and child depression scales. Journal of Pediatric Psychology,
10, 45-54.

Lorion, R., Price, R., & Eaton, W. (1989). The prevention of child and adolescent disorders:
From theory to research. In D. Schaffer, I. Philips, & N. Enzer (Eds.), Prevention of mental
disorders, alcohol and other drug use in children and adolescents (pp. 55-97).
Washington, DC: Office of Substance Abuse Prevention Monograph-2 (DHHS Publication
No. ADM 90-1646).
MacCallum, R. C., & Austin, J. T. (2000). Application of structural equation modeling in
psychological research. Annual Review of Psychology, 51, 201-226.

MacKinnon, D. P. (1994). Analysis of mediating variables in prevention and intervention

research. National Institute of Drug Abuse Research Monograph Series, 139, 127-153.

MacKinnon, D. P., Krull, J. L., & Lockwood, C. M. (2000). Equivalence of the mediation,
suppression, and confounding effect. Prevention Science, 1, 173-181.

Monroe, S. J. (1982). Life events assessment: Current practices, emerging trends. Clinical
Psychology Review, 2, 435-453.

Morrow, J., & Nolen-Hoeksema, S. (1990). Effects of responses to depression on the

remediation of depressive affect. Journal of Personality and Social Psychology, 58, 519-
527.

Moskowitz, D. S., & Schwarz, J. C. (1982). A validity comparison of behavior counts and
ratings by knowledgeable informants. Journal of Personality and Social Psychology, 42,
518-526.

Muthen, L. K., & Muthen, B. O. (1998). Mplus user's guide: The comprehensive modeling
program for applied researchers. Los Angles: Muthen & Muthen.

Peterson, J. L., & Zill, N. (1986). Marital disruption, parent-child relationships, and
behavior problems in children. Journal of Marriage and the Family, 48, 295-308.

Pitts, S. C., West, S. G., & Tein, J.-Y. (1996). Longitudinal measurement models in
evaluation research: Examining stability and change. Evaluation and Program Planning,
10, 383-350.

Reynolds, W. M. (1992). Depression in children and adolescents. In W. M. Reynolds (Ed.),

Internalizing disorders in children and adolescents. New York: Wiley.

Reynolds, C. R., & Paget, K. D. (1981). Factor analysis of the revised Children's Manifest
Anxiety Scale for blacks, whites, males, and females with a national normative sample.
Journal of Consulting and Clinical Psychology, 49, 352-359.

Reynolds, C. R., & Richmond, B. (1978). What I think and feel: A revised measure of
children's manifest anxiety. Journal of Consulting and Clinical Psychology, 6, 271-280.

73
Rodgers, B., Power, C., & Hope, S. (1997). Parental divorce and adult psychological
distress: Evidence from a national birth cohort: A research note. Journal of Child
Psychology and Psychiatry, 38, 867-872.

Rosenberg, M. (1968). The logic of survey analysis. New York: Basic Books.

Ross, C. E., & Mirowsky, J. (1999). Parental divorce, life-course disruption, and adult
depression. Journal of Marriage and the Family, 61, 1034-1045.

Sandler, I. N. (2001). Quality and ecology of adversity as common mechanisms of risk and
resilience. American Journal of Community Psychology, 29, 19-63.

Sandler, I. N., Tein, J.-Y., & West, S. G. (1994). Coping, stress, and the mental health of
children of divorce: A longitudinal study. Child Development, 65, 1744-1763.

Sandler, I. N., Wolchik, S. A., & Braver, S. L. (1988). The stressors of children's post-
divorce environments. In S. Wolchik & P. Karoly (Eds.), Children of divorce: Empirical
perspectives on adjustment (pp. 111-144). New York: Gardner.
Sandler, I. N., Wolchik, S. A., Braver, S. L., & Fogas, B. (1986). Significant events of
children of divorce: Toward the assessment of a risky situation. In S. M. Auerbach & A.
Stolberg (Eds.), Crisis intervention with children and families (pp. 65-87). New York:
Hemisphere.

Sandler, I. N., Wolchik, S. A., Braver, S. L., & Fogas, B. (1991). Stability and quality of life
events and psychological symptomatology in children of divorce. American Journal of
Community Psychology, 19, 501-520.

Sandler, I. N., Wolchik, S. A., MacKinnon, D., Ayers, T. S., & Roosa, M. W. (1997).
Developing linkages between theory and intervention in stress and coping processes. In
S. A. Wolchik & I. N. Sandier (Eds.), Handbook of children's coping: Linking theory and
intervention. (pp. 3-40). New York: Plenum.

Schaefer, E. S. (1965). Children's report of parental behavior: An inventory. Child

Development, 36, 413-424.

Schafer, J. L. (1997). Analysis of incomplete multivariate data. London: Chapman & Hall.

Schafer, J. L., & Olsen, M. K. (1998). Multiple imputation of multivariate missing-data

problems: A data analyst's perspective. Multivariate Behavioral Research, 33, 545-571.

Schwarz, J. C., Barton-Henry, M. L., & Pruzinsky, T. (1985). Assessing child-rearing

behaviors: A comparison of ratings made by mother, father, child and sibling on the
CRPBI. Child Development, 56, 462-479.

Schwarz, J. C., & Mearns, J. (1989). Assessing parental childrearing behaviors: A

comparison of parent, child and aggregate ratings from two instruments. Journal of
Research in Personality, 23, 450-468.

Sheets, V., Sandler, I. N., & West, S. G. (1996). Assessment of threat appraisals in
children of divorce. Child Development, 67, 2166-2182.

Simons, R. L., Lin, K.-H., Gordon, L. C., Conger, R. D., & Lorenz, F. O. (1999). Explaining
the higher incidence of adjustment problems among children of divorce compared with
those in two-parent families. Journal of Marriage and the Family, 61, 1020-1033.

74
Simons, R. L., and Associates. (1996). Understanding differences between divorced and
intact families: Stress, interaction, and child outcome. Thousand Oaks, CA: Sage.

Skinner, E. A., & Wellborn, J. G. (1994). Coping during childhood and adolescence: A
motivational perspective. In R. Lerner, D. Featherman, & M. Perlmutter (Eds.), Lifespan
development and behavior (pp. 91-133). Hillsdale, NJ: Erlbaum.

Skinner, E. A., & Wellborn, J. G. (1997). Children's coping in the academic domain. In S. A.
Wolchik & I. N. Sandler (Eds.), Handbook of children's coping with common life stressors:
Linking theory, research and interventions (pp. 387-422). New York: Plenum.

Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structure

equation models. In S. Leinhardt (Ed.), Social methodology (pp. 290-312). San Francisco:
Jossey Bass.

Sobel, M. E. (1986). Some new results on indirect effects and their standard errors in
structure equation models. In N. Tuma (Ed.), Social methodology (pp. 159-186). San
Francisco: Jossey Bass.
Stevens, J. P. (1984). Outliers and influential data points in regression analysis.
Psychological Bulletin, 95, 334-344.

Stolberg, A. L., & Anker, J. M. (1983). Cognitive and behavioral changes resulting from
parental divorce and consequent environment changes. Journal of Divorce, 7, 23-41.

Tein, J.-Y., Sandler, I. N., MacKinnon, D. P., & Wotchik, S. A. (2000). Testing of mediation
when program effects interact with pretest mediator or outcome variables. Paper
presented at the 8th annual SPR conference in Montreal.

U.S. Bureau of Census. (1996, May). Resident population of the United States: Middle
series projections, 1996-2000, by Sex, Race, and Hispanic Origin, with Median Age. In
Current Population Reports (Series P25-1130). Washington, DC: Government Printing
Office.

Vasey, M. W. (1993). Development and cognition in childhood anxiety: The example or

worry. In T. H. Ollendick & R. J. Priz (Eds.), Advances in clinical child psychology (pp. 1-
39). New York: Plenum.

Wallerstein, J. S. (1985). Children of divorce: Emerging trends. Psychiatric Clinics of North

America, 8, 837-855.

Wallerstein, J. S., & Kelly, J. B. (1980). Surviving the breakup: How children and parents
cope with divorce. New York: Basic Books.

Wolchik, S. A., Ramirez, R., Sandler, I. N., Fisher, J. L., Organista, P. B., & Brown, C. (1993).
Inner-city, poor children of divorce: Negative divorce-related events, problematic beliefs
and adjustment problems. Journal of Divorce and Remarriage, 19, 1-20.

Wolchik, S. A., West, S., Sandler, I. N., Tein, J.-Y., Coatsworth, D., Lengua, L. J., et al.
(2000). The New Beginnings Program for divorced families: An experimental evaluation of
theory-based single-component and dual-component programs. Journal of Consulting and
Clinical Psychology

Wolchik, S. A., West, S., Westover, S., Sandler, I. N., Martin, A., Lustig, J., et al. (1993). The
children of divorce parenting project: Outcome evaluation of an empirically-based
parenting program. American Journal of Community Psychology, 21, 293-331.

75
Wolchik, S. A., Wilcox, K. L., Tein, J.-Y., & Sandler, I. N. (2000). Maternal acceptance and
consistency of discipline as buffers of divorce stressors on children's adjustment
problems. Journal of Abnormal Child Psychology, 28, 87-102.

Sharlene A. Wolchik (1, 3)

Jenn-Yun Tein (1)

Irwin N. Sandler (1)

Kathryn W. Doyle (2)

(1.) Program for Prevention Research, Arizona State University. Tempe, Arizona.

(2.) Johns Hopkins University, Baltimore, Maryland.

(3.) Address all correspondence to Sharlene A. Wolchik, Department of Psychology,

Arizona State University, Tempe, Arizona 85287-1104; e-mail: sharlene.wolchik@asu.edu.
COPYRIGHT 2002 Plenum Publishing Corporation
in association with The Gale Group and LookSmart. COPYRIGHT 2002 Gale Group

Journal of Abnormal Child Psychology

April, 2002

Sibling collusion and problem behavior in early adolescence: Toward a process

model for family mutuality.(Abstract)

Author/s: Bernadette Marie Bullock

INTRODUCTION

Most social development research focuses on parent--child relationship processes and

neglects sibling relationship dynamics. Systemic views of parenting, however,
acknowledge that parenting does not occur in a vacuum. In fact, the sibling subsystem
can form a unique context within families that potentially promotes or detracts from the
socialization goals of caregivers. Of interest in this study are the sibling interactions that
promote problem behavior, otherwise referred to as "sibling collusion."

Sibling collusion is a process by which siblings in a family form coalitions of deviance that
potentially undermine parents' socialization efforts. Collusion among siblings is
predicated on a pattern of mutual positive reinforcement of deviant talk. Families with a
target child identified as either high risk or normative are compared using direct
observations of sibling collusion to determine the extent to which sibling coalitions
account for variance in deviant peer associations and problem behavior in early
adolescence.

Two sibling processes related to the development of conduct problems have been
extensively investigated: coercive sibling interactions and mutuality of sibling problem
behavior. The purpose of this study is to extend this research by focusing on collusive
interactions in which siblings mutually reinforce each other's deviant behavior through
positive affect.

Coercive Process

The coercion model emphasizes the role of escape conditioning during family interactions
in the development of aggressive and antisocial behavior (Patterson, 1982; Patterson,

76
Reid, & Dishion, 1992). Escape conditioning occurs in the context of family conflict, in
particular, a recurring pattern of using aversive behavior to terminate conflict.

Within families engaged in high rates of conflict and coercion, sibling interaction may
provide another setting in which children learn to use aggression and other forms of
aversion. Parents who do not attend to and manage sibling play may inadvertently allow
conflicts to be solved by means of coercion. Patterson (1984) aptly refers to siblings in
antisocial families as "fellow travelers within a coercive system." Indeed, families seeking
treatment for childhood behavior problems show higher rates of coercive interchanges
among siblings than do nonreferred families (Arnold, Levine, & Patterson, 1975).
Moreover, siblings in families with aggressive children participate in protracted coercive
chains, with problem children exhibiting a greater likelihood of engaging in extended
coercive exchanges (Patterson, 1984, 1986). Coercion, then, is a process that is
embedded within family conflict and negative affect.

Deviant Mutuality

Both environmental and genetic models predict high levels of similarity among siblings in
behavior adjustment (Reiss, Neiderhiser, Hetherington, & Plomin, 2000; Rowe, 1981;
Scarr & Grajek, 1982). The level of association among siblings appears to be moderate-to-
high across studies using somewhat different methods and foci. For example, Lewin,
Hops, Davis, and Dishion (1993) examined sibling similarity using teacher and peer
reports. They found a moderately high level of covariation among 45 pairs of elementary
school-aged siblings for peer ratings of popularisty and teacher ratings of social behavior
in the playground, school adjustment, and academic competence.
Several studies also suggest similarity among siblings in levels of delinquency and
substance use. Sibling drug use was found to have a direct association with adolescent
substance use in several studies (Brook, Whiteman, Gordon, & Brenden, 1983; McKillip,
Johnson, & Petzel, 1973). Siblings are also known to be similar in their levels of delinquent
behavior (Brownfield & Sorenson, 1994; Rowe, Rodgers, & Meseck-Bushey, 1992). In
studies of the emergence of antisocial conduct, it has long been documented that 50% of
the crimes are committed by 5% of families (e.g., Loeber & Dishion, 1983; West &
Farrington, 1977).

There is very little social interactional research that might account for the mutuality of
siblings' problem behavior and the clustering of deviance in families. Siblings may share a
coercive history that would formulate a trajectory toward poor schoolwork and deviant
peer relations. It may also be that, as children in the same family mutually develop
pattems of problem behavior, they actively form coalitions of deviance within the family,
and perhaps, outside the home in shared peer networks. In previous research,
adolescents were found to engage in a process with their peers labeled "deviancy
training," which predicted escalations in substance use, delinquency, and violence
(Dishion, Capaldi, Spracklen, & Li, 1995; Dishion, Eddy, Haas, Li, & Spracklen, 1997;
Dishion, Spracklen, Andrews, & Patterson, 1996). These studies identify a pattern of
mutual reinforcement for deviant talk in friendship relationships, where deviant talk is
entrenched in positive affect.

It is hypothesized that sibling relationships, on occasion, may also be organized around

deviant talk. In the context of families, siblings establish mutuality with respect to
undermining parent influence, breaking family rules, and other forms of problem
behavior. In the absence of caregivers, siblings may support stories and acts of deviance
through laughter, interest, and reciprocity. In the presence of caregivers, siblings may
collude to avoid adult demands, undercut caregiver leadership, and support a coalition of
deviance. This process may evolve in the presence of a positive or a negative sibling
relationship, as the immediate goal is to undermine parental authority and perpetuate
problem behavior. Sibling collusion, presumably, would be highest among children who
are close in age and close to adolescence, where covert forms of problem behavior

77
increase, partly as a function of deviant peer involvement (Dishion, Capaldi, et al., 1995;
Patterson, 1993).

Peer Linkages

Efforts to link the world of peers and siblings focus primarily on the generalization of
interpersonal style (Abramovitch, Carter, Pepler, & Stanhope, 1986; Stormshak, Bellanti,
& Bierman, 1996). One neglected aspect of sibling influence is the extent to which
siblings influence a child's social network. By far, the most prognostic indicator of
problem behavior is an association with deviant peers (Dishion, Capaldi, et al., 1995;
Dishion, Andrews, et al., 1996; Dishion, Eddy, et al., 1997; Elliott, Huizinga, & Ageton,
1985; Patterson & Dishion, 1985). Like all social processes, the hypothesis is that the
peer- sibling connection is bidirectional. Collusion among siblings may be a product of
shared deviant peer influences and may lead to increased exposure and linkages with
antisocial friends.
These findings indicate a direct relationship between the peer group and adolescent
problem behavior. Because we conceptualize the peer group as being a highly salient and
proximal influence for young adolescents, deviant peer association is anticipated to
mediate the relationship between sibling collusion and concurrent problem behavior.

A pattern of conspiring to defy adult influence would also be expected to weaken

communication between the caregiver and adolescents to the extent that they coexist in
two different worlds. Parent monitoring is predicated on accurate information and good
communication within a family, in general, and between a parent and child, in particular
(Dishion & McMahon, 1998). From an ecological perspective, a family system in which
youngsters collude toward deviance would account for a synergistic effect on adolescent
problem behavior, vis-a-vis directly encouraging such behavior and indirectly, through
undermining parent monitoring and increasing access to deviant peer activities.

The Current Study

This study is seen as a first step toward identifying sibling relationship processes that
might account for unique variance in adolescent problem behavior in some families. The
goals of this study were (a) to determine whether sibling collusion processes could be
identified and measured during a videotaped family interaction; (b) to examine whether
siblings are more likely to engage each other in collusive interactions in families with a
child identified as at high risk for delinquency, versus a group of normative comparisons;
and (c) to assess the extent to which collusive interactions in the sibling subsystem are
directly related to problem behavior or mediated by involvement with deviant peers.

METHOD

Participants

This study was part of the larger Project Alliance study (Dishion & Kavanagh, 2000),
designed to prevent the early onset of adolescent problem behaviors (alcohol use,
tobacco use, and delinquency) by supporting middle school families living in high-risk
neighborhoods.

The families used in the present analysis are a sub-sample of a group of 257 families with
a sixth-grade target child (mean age 12.65 years) identified as normative and 152
families with a child (mean age 12.40 years) distinguished as being at high risk for
problem behavior. The term "normative" refers to youth who were well adapted within
the school context.

The sample was selected from three middle schools after reviewing academic records

78
and discipline contacts. Families with a middle school child were invited to participate in
the normative group if their middle school student earned above a C grade point average
and had no attendance problems or discipline contacts for problem behavior at school. Of
those approached, 48% of families with a normative target child agreed to participate, for
a total of 123 families. The sample was recruited to represent gender and ethnicity
equally (European American and African American).

The high-risk sample, which represented a subsample of sixth-grade students involved in

the first cohort of the Project Alliance prevention trial, was selected the following year
from the same three middle schools. Several studies confirm that teacher ratings of child
conduct are among the most sensitive predictors of behavior problems and delinquency
(see Loeber & Dishion, 1983), and are often used as an initial screening device (Dishion &
Patterson, 1993; Loeber & Dishion, 1987).
A teacher rating measure (Teacher Risk Perception), which is based on an instrument
developed in previous research to screen for risk in middle schools (Soberman, 1995),
was used for risk identification. The measure includes 16 items, rated from 1 (low risk) to
5 (high risk). Risk assessment criteria included classroom behavior, attitude regarding
school, negative mood, tobacco use, involvement with troublesome or substance-using
peers, and acceptance by peers. The mean of the 16 items was used to predict risk
status; the standardized item alpha was .91. Children with the top 50 risk scores from
each school, as well as any student who scored 3.0 or greater on the smoking questions,
were assigned to the high-risk group.

Of the students, 152 were identified as high risk and 55% of the families with a high-risk
target child agreed to participate in the family assessment, resulting in a total sample of
83. The mean age of target children in this group was 12.40 years (SD = 0.50). Only
families with siblings aged 10 or older living in the home at the time of assessment were
included in this study.

Each family with a study child identified as high risk, and a sibling aged 10 or older who
participated in the family assessment (n = 26), was matched on child age, gender,
ethnicity, and parental marital status to a family with a normative target child (n = 26),
yielding a final sample of 52 young adolescents (27 girls, 25 boys; see Table I). Target
children in the normative group had a mean age of 12.65 years (SD = 0.75). Mean level
analyses indicated that the age difference between the high-risk and normative groups
was nonsignificant, t (49) = 1.42, p > .15.

Procedures

Questionnaires were administered to the parent(s), teacher, and child of the participating
families. In addition, home visits involved a brief interview with the parent(s) and youth,
as well as a videotaped observation of family interaction, referred to as the FAST. The
family received a $75 reimbursement at the conclusion of this appointment.

Measures

Family Assessment Task

Each family participated in a series of seven video-taped conversations designed to

directly assess family management practices. Topics discussed during the parent-target-
child conversation included (a) planning a fun family activity; (b) discussion of an area in
which parents wanted to see the target child advance; (c) discussion with the target child
about a period during the past month when the child had spent at least 1 hr with peers,
without adult supervision; and (d) discussion of a circumstance during the previous
month in which the parents felt the need to impose limitations on the target child's
behavior.

79
Each of these four conversations lasted 5 min and included the parents and the target
child. Following a 5-min break, siblings over the age of 10 participated in three additional
discussions in which they were asked to (a) work together to solve a family problem that
was identified by the parents during the previous visit (5 min); (b) discuss parent, target
child, and sibling beliefs regarding the use of alcohol, tobacco, and other drugs (8 min);
and (c) plan a family celebration that could occur during the next 2-3 months (7 min).

Sibling Collusion Macroratings

The Sibling Interaction Ratings (SIB-R) instrument was designed to measure collusion in
the sibling subsystem (Bullock & Dishion, 1999). Coders who were blind to family risk
status were trained to determine whether or not the siblings "conspired to engage in
negative behavior," "conspired to avoid parental direction, monitoring or limit-setting,"
"shared stories or mutual experiences involving deviant or rule-breaking behavior,"
"discussed mutual friends or acquaintances that are or have been involved in deviant or
rule-breaking behavior," and "exchanged knowing glances, shared laughter, winks, or
other gestures that appeared to indicate implicit understanding of shared deviant or rule-
breaking behavior."

For each collusion item, each of the sibling's behaviors were independently rated on a 4-
point scale: 0 (clearly not present), 1 (some), 2 (clear example), 3 (multiple examples).
Global ratings were based on the entire FAST. A total sibling collusion score was
computed from the mean of the five items (standardized item [alpha] = .93). Coder
reliability was calculated on 25% of cases, with a percent agreement of 75% (exact) and
91% ([+ or -] 1).

Relationship Process Macroratings

Coders also completed a 79-item Coder Impressions Questionnaire (COIMP; Dishion &
Kavanagh, in press; Forgatch, Fetrow, & Lathrop, 1984) to assess various aspects of
family management and problem solving. Coder reliability was calculated for 20% of
cases, with agreement ranging from 82 to 85%. To investigate sibling relationship
process, ratings of both positive (sibling mutuality) and negative (e.g., coercive)
dynamics were included.

A positive relationship score was computed using the mean of four items (for older male
sibling, older female sibling, younger male sibling, and younger female sibling). Positive
relationship items consisted of behaviors such as supportive talk, empathy, and
understanding of others' feelings (standardized item [alpha] = .65). A negative sibling
relationship score was computed using the mean of four items that correspond to those
above. Negative relationship items include behaviors such as use of guilt, mind reading,
and controlling of others' feelings (standardized item [alpha] = .87).

Child Interview

The Child Interview (Oregon Social Learning Center, 1984) consists of questions regarding
family environment and structure, peer relationships, traumatic events, and antisocial
behavior (including substance abuse). For purposes of this study, an antisocial peer score
was computed from the mean standardized score of 11 items, which were rated on a
scale from 1 (none of them) to 5 (all of them). This score addressed to what extent during
the past year the target child's friends cheated on a test, damaged something that was
not theirs, stole something, threatened to hit someone for no reason, broke into property
with the intent to steal, sold hard drugs, suggested an illegal activity, sold or gave alcohol
to peers, were in trouble with the police, or were involved in gang fighting. The
standardized item alpha for the antisocial peer score was .84.
A peer drug use score was computed using the mean of five items, which were rated on a

80
scale from 1 (very few, 15% or less) to 5 (almost all, 85% or more). Questions included
the percentage of kids at school who smoked or chewed tobacco; drank beer, wine, wine
coolers, or hard liquor; and used marijuana or other drugs. The standardized item alpha
for the peer drug-use scale was .85. Finally, a deviant peers' score was derived using the
mean of peer drug use and antisocial peers' scores (standardized item [alpha] = .76).

Peer Relations and Social Skills

The target child, parents, and teachers completed questionnaires assessing the target
child's peer relations and social skills: Child's Peer Ratings and Social Skills, Parent Beliefs
and Peers, and Teacher Peers/Social Skills (Dishion & Kavanagh, in press). The measures
included questions involving the type of children with whom the child associated, the
degree to which the child was accepted or rejected by peers, and how easily the child
was influenced by peers and adults. Four items were used for this analysis: "percent of
friends who behaved well in school," "percent of friends who misbehaved or broke the
rules," "percent of friends who experimented with smoking and drugs," and "percent of
friends who dressed or acted like a gang member." These items were rated from 1 (very
few, less than 25%) to 5 (almost all, more than 75%). Deviant peer scores from the child,
parent, and teacher reports were computed from means of the four items, with child,
parent, and teacher standardized item alphas of .73, .61, and .80, respectively.

Youth Self-Report

The Youth Self-Report is a standardized, validated, and reliable system for assessing
children's social competence and the presence of self-reported problem behavior. The
measure includes a 112-item checklist, which is based on the Child Behavior Checklist
(Achenbach, 1991). For this analysis, the externalizing summary score was derived using
the mean of delinquent (standardized item [alpha] = .82) and aggressive behavior
(standardized item [alpha] = .80) scales. The standardized item alpha for externalizing
was .85.

Child Behavior Checklist, Parent Form

The Parent Form of the Child Behavior Checklist (Achenbach, 1991) is also a
standardized, validated, and reliable system for assessing children's social competence
and the presence of problem behavior. The measure contains 113 items, which assess
children's social competence and problem behaviors via parent report. The externalizing
summary score was derived using the mean of delinquent (standardized item [alpha]
= .79) and aggressive behavior (standardized item [alpha] = .84) scales, with an overall
standardized item alpha of .84 for externalizing.

Construct Building Strategy

Three constructs (antisocial problem behavior, sibling collusion, and deviant peer group)
were selected to examine whether sibling collusion was directly related to problem
behavior or mediated by involvement with deviant peers. The psychometric properties of
the scales from which these constructs were derived are provided in Tables II and III.
Constructs were developed based on a strategy described by Patterson et al. (1992). A
description of the measures that constitute antisocial problem behavior, sibling collusion,
and deviant peer group constructs follows.
Antisocial Problem Behavior

This construct was developed by computing the mean of parent and child standardized
externalizing scales from the Child Behavior Checklist Parent Report and Youth Self-
Report forms. These measures included items that examined the degree to which the
child gets into trouble, lies or cheats, uses drugs or alcohol, and so forth, on a scale from
0 (not true) to 2 (very true or often true). This construct had significant convergent

81
validity for parent and youth reports (r = .40, p < .005).

Sibling Collusion

This construct was created by computing the mean of the standardized sibling collusion
scale from the SIB-R and the standardized positive and negative relationship scales from
the COIMPs. These scales reflected the degree to which siblings engaged in positive and
negative relationship dynamics and the extent to which overall sibling communication
reflected an implicit understanding of shared antisocial or rule-breaking behavior.
Previous research suggests that both mutuality (positive) and coercion (negative) are
characteristic of the relationships of siblings who engage in deviant acts (Arnold et al.,
1975; Brownfield & Sorenson, 1994; Patterson, 1984, 1986; Rowe et al., 1992). This
construct was designed to capture these relationship processes as they are embedded in
a culture of sibling deviance. The construct was internally consistent and the three scales
used in construct development showed significant convergent validity, p < .01 (see
Tables II and III).

Deviant Peer Group

This construct was created by computing the mean of standardized deviant peer group
scales from the Child Interview and child, parent, and teacher peer relations
questionnaires. Collectively, these measures assessed the extent to which the target
child associates with friends who engage in antisocial dress and conduct, including
substance use. As indicated in Tables II and Ill, the deviant peer group construct was
internally consistent, with scale convergent validity ranging from r = .02 (p > .05) to .57
(p < .01).

Data Analysis Strategy

The data analysis strategy addressed the three questions posed in this study. First, the
viability of the sibling collusion construct, measured using the SIB-R instrument, was
examined. This was accomplished by assessing bivariate correlations among collusion
items and the standardized item alpha for the collusion scale in order to establish the
internal consistency and convergence of rating items. Second, a series of mean level
comparisons were conducted to explore whether or not sibling communication variables
and scales differentiated families with a high-risk versus a normative target child. Finally,
multiple regression procedures were used to evaluate whether sibling collusion was
directly related to problem behavior or mediated by involvement with deviant peers. For
these analyses, predictor variables included the sibling collusion and deviant peer group
constructs; the criterion variable was the antisocial problem behavior construct.
RESULTS

Measurement of Sibling Collusion

As mentioned previously, one goal of this study was to investigate if sibling collusion
could be identified and measured during a videotaped family interaction task. An
assessment of individual items on the SIB-R indicated that the frequency with which
siblings engaged in collusive behavior ranged from 9.5 to 22.0% per item (see Table IV).
Sharing deviant stories, discussing mutual deviant friends, and an implicit understanding
of shared deviant behavior were most commonly rated.

A review of bivariate Pearson correlations for the SIB-R (see Table V) also revealed a high
degree of item convergence. Collusion items (negative behavior, avoids parental
direction, shares deviant stories, discusses mutual deviant friends, and understanding of
shared deviant behavior) were significantly correlated (p < .05) on the whole. In addition,
COIMPs of positive and negative sibling relationship dynamics, and the aggregate of SIB-R
collusion items, were significantly correlated (r = .36 and r = .55, p < .01). Lastly, SIB-R

82
collusion items were found to be internally consistent, with a standardized item alpha
of .93.

Initial analyses were conducted to investigate mean level differences in collusion by each
demographic variable (age, gender, ethnicity, family structure, gross income, and
number of siblings). Univariate analyses of variance (ANOVAs) indicated no significant
differences in collusion for each of these variables, with the exception of family structure
in which marginally significant collusion by structure differences were detected, F(2, 48)
= 2.43, p <.10. Planned orthogonal contrasts revealed mean differences between single-
and dual-parent families t(48) = 2.20, p < .05, with siblings in single-parent families
engaging in significantly more collusion than siblings in dual-parent families (M = 0.22,
SD 1.01 and M = -0.27, SD = 0.32, respectively).

Collusion: High-Risk Versus Normative

Frequency data were examined to assess whether siblings in families with a high-risk
versus normative target child differed in their use of collusion. Only 5 of 25 families (20%)
in the no-risk target child group had siblings who engaged in collusion. Families with a
high-risk target child exhibited nearly three times the rate of collusive behavior, with 15
of 26 families (58%) containing siblings who colluded with each other during 20 mm of
videotaped interaction.

A composite sibling collusion scale was created by computing the mean of all sibling
collusion items for each family from the SIB-R (standardized item [alpha] = .85). The
sibling collusion scale was analyzed first using a one-way ANOVA to determine whether
the aggregate of sibling collusion differed in families with a high-risk versus a normative
target child (as identified by the teacher rating questionnaire).

As anticipated, a main effect for sibling collusion by risk was found. Siblings in families
with a high-risk target child displayed more collusion than siblings in families in which the
target child was not identified as being at risk for delinquency (see Table VI).
To further explore whether sibling processes differed in families with a high-risk versus
normative target child, a comparison of COIMPs of positive and negative relationship
scores by risk was made using a one-way ANOVA. COIMP sibling relationship scores
combine siblings in their ratings. These analyses revealed a main effect for both positive
and negative sibling relationship by risk. Siblings in families with a high-risk target child
displayed significantly more positive and negative relationship dynamics than siblings in
comparison families.

Predicting Problem Behavior

The second aim of this study was to assess whether sibling collusion was directly related
to problem behavior or mediated by involvement with deviant peers. Mean level analyses
suggested that high-risk target children were more likely to be rated by self- and parent
report as engaging in problem behavior than their normative counterparts (see Table VI).
Only marginally significant differences were found between high-risk and normative
children for the association with deviant peers construct.

Multiple Regression Analysis

The hypothesis that deviant peer association would mediate the covariation between
sibling collusion and problem behavior was tested using a multiple regression analysis. As
illustrated in Table VII, the mediation hypothesis was not supported. Although the
variation between sibling collusion and problem behavior was reduced after entering
deviant peers into the model ([beta] = .39-.25), the link between sibling collusion and
problem behavior remained statistically reliable (p < .05). Furthermore, deviant peer
association accounted for unique variance in problem behavior after controlling for sibling

83
collusion ([beta] = .45, p < .001). The total additive model accounted for 33.8% of the
variance in concurrent problem behavior, F(2, 48) = 10.60, p < .001.

DISCUSSION

This study represents the initial phase of an examination of sibling collusion processes
and their relationship to growth in problem behavior. These data are consistent with
investigations of sibling mutuality (Arnold et al., 1975; Brook et al., 1983; Dadds,
Sanders, Morrison, & Rebgetz, 1992; Lewin et al., 1993; McKillip et al., 1973; Patterson,
1984) and further extant research by revealing that siblings in families with a high-risk
target child promote their own deviance by colluding in the presence of adult caregivers.
Moreover, this process is readily observed while videotaping family discussions. Effect
sizes indicate that mean differences in collusion between high-risk versus normative
groups are moderately large (.56). More importantly, these analyses demonstrate that
sibling collusion uniquely predicts adolescent problem behavior above and beyond
association with a deviant peer group.

These results also suggest that sibling interactions in families with a target child at high
risk for delinquency are characterized by both positive and negative relationship
dynamics, compared to sibling exchanges in families with a normative target child.
Although this finding may seem paradoxical, it fits with the literature on both coercive
family process and deviancy training. Coercive interactive styles are functional in family
relationships with high rates of negative affect and conflict, whereas deviancy training is
predicated on positive affect, shared deviance, and humor. The coexistence of high rates
of positive and negative exchanges in families with children at risk for problem behavior
speaks to the chaos and disorganization that is prognostic of long-term behavioral
maladjustment (Loeber & Dishion, 1983).

The growing body of behavioral genetic research suggests that behavioral concordance
among siblings can be accounted for by shared environmental factors, as well as genetic
similarity (Reiss et al., 2000). The data did not permit examination of the contribution of
heritable characteristics of siblings and the siblings' shared environment to the regulation
of collusive processes in the sibling subsystem. However, these findings do suggest that
siblings, and potentially parents, are participants in a family culture that supports the
discussion and enactment of antisocial behavior. It is plausible that the sibling collusion
process, genetic factors, and the adjustment of individual family members interact to
create a family context that fosters sibling maladjustment.

Contrary to our expectations, deviant peer involvement did not mediate the relationship
between sibling collusion and problem behavior. Instead, both deviant peer association
and collusion among siblings contributed additively to the prediction of behavior
problems. The directionality of this relationship between sibling collusion and deviant
peers is difficult to disentangle. Extant research indicates that once children develop an
antisocial repertoire through interactions with siblings and other family members, they
are more likely to establish social networks in which their antisocial conduct is reinforced
(Dishion, Patterson, Stoolmiller, & Skinner, 1991). Consequently, we envision the sibling
context to be a training ground in which young adolescents encourage the antisocial
behavior of their fellow siblings and subsequently seek out similarly reinforcing peer
relationships. It may be that, early in development, children with the strongest sibling
coalitions and a shared deviant peer network are the most difficult to supervise and guide
into adolescence.

These findings have several implications. First, the high degree of similarity of sibling
interpersonal processes related to deviance suggests that problems within the family are
systemic, with all family members contributing to the nature and quality of each family
member's behavior (Minuchin, 1985). Second, in light of the fact that parents were

84
present for the family observation task (during which sibling collusion was measured). It
is possible that those parents may implicitly endorse such behavior through an absence
of effective family management. Sibling collusion may be a byproduct of a distressed
family system and may also serve as an indicator of disrupted or ineffectual family
management practices.

As suggested by the term "sibling collusion," these interactions may dramatically

undermine caregivers' attempts to socialize. Inspection of the videotaped interactions,
however, suggests that for some families, caregivers were also positively involved in
deviant discussions. Indeed, sibling collusion may be indicative of a process in which
multiple family members interact to promote a culture of deviance. As such, the
relationship between sibling collusion, parenting practices, and outcome are likely to be
synergistic, with each factor influencing the other. The process of assessing the potential
multidirectionality of these effects is now underway.
Note that higher levels of sibling collusion were observed in single-parent families. A body
of literature suggests that parenting alone is more difficult and that a modest level of
child problem behavior is observed in single-parent families (Home, 1981). Clearly,
managing multiple children as a single parent presents formidable challenges (Grouter,
McHale, & Perry-Jenkins, 1990; Pfiffner, Jouriles, Brown, Etscheidt, & Kelly, 1990). This
may also be true for families undergoing stepparenting transitions (Capaldi & Patterson,
1991).

Often, there is a pull in one-parent families to connect with adolescents in friendship.

Qualitative accounts of clinical families have described the "sibling parent," in which the
parenting relationship is approached as an extended friendship (Patterson, 1982). In such
families, children, caregivers, and extended family may engage in deviant discussions,
which may set the stage for early-onset and persistent antisocial and criminal behavior.
Future research is needed to identify the sibling and parental relationship processes that
promote deviant peer association and contribute to the escalation of antisocial problem
behavior.

The identification of this collusive process indicates that effective family therapies are
needed that specifically target the sibling context, particularly the mechanism by which
collusive sibling behavior undermines parent attempts to socialize. Previous intervention
research reveals that a focus on family management leads to a reduction in antisocial
behavior for the child and siblings (Arnold et al., 1975). Indeed, ongoing analyses confirm
that sibling collusion is significantly and negatively associated with parental monitoring, r
= -.48, p < .001 (Bullock & Dishion, 2001).

If sibling collusion is thought of as one form of problem behavior, interventions that

support parenting should incorporate information regarding the dynamics of appropriate
sibling interaction. In addition, therapeutic approaches must underscore effective family
management, parental supervision, and the fostering of positive, constructive family
interaction as a means of preventing problem behavior and mitigating involvement with a
deviant peer group. Adolescence, in particular, may present unique challenges to
caregivers, especially when there is more than one child in the family.

Family problem-solving techniques may be particularly useful, as they are intended to

promote prosocial cooperation, positive communication, and constructive relationships
within the family (Forgatch, 1984; Forgatch & Patterson, 1989; Pfiffner et al., 1990).
Given these data, it is not surprising that successful adolescent interventions accentuate
positive family relationships, supporting caregivers in a leadership role, constructive
communication, monitoring, and effective behavior management strategies
(Chamberlain, Patterson, Reid, Kavanagh, & Forgatch, 1984; Dishion, Andrews, Kavanagh,
& Soberman, 1996; Henggeler, Schoenwald, Borduin, Rowland, & Cunningham, 1998).

85
There are several limitations to this initial analysis of sibling collusion and problem
behavior. This was a relatively small sample, which reduced statistical power and limited
the potential of this study to address sibling age, gender, and ethnicity effects. The cross-
sectional nature of these data also precludes the analysis of time-ordered effects and
inferences regarding causality. Longitudinal data collection is continuing in order to
address the long-term, unique impact of sibling interaction on adolescent social
development and the persistence of problem behavior into young adulthood.

To better understand the causal nature of collusive sibling dynamics and the variability of
these processes across family and cultural situations, the next generation of family-based
intervention research will test these ideas experimentally through intervention trials. It
will be especially important to link specific intervention strategies with corresponding
family processes (i.e., collusion among siblings). Such specificity will provide an insight
into potential causal mechanisms and expand the prevention and treatment
armamentarium needed to support family change.

Table I

Participant Demographic Information

Families (N = 52)
High risk Normative
(n = 26) (n = 26)

Target child age (in years)

M (SD) 12.40(0.50) (a) 12.65 (0.75)
Range 12.00-13.00 12.00-14.00
Ethnicity
African American 12 13
European American 10 8
African/European 2 2
American
Other 2 3
Gender of target child
Male 11 14
Female 15 12
Gender of siblings
Older male 8 10
Older female 11 9
Younger male 7 10
Younger female 5 15
Twin 2 1
Number of siblings
participating
M (SD) 1.44 (0.77) 1.81 (0.94)
Family structure
Single parent 10 13
Dualparent 16 13
Gross income
Median gross income $15,000-29,999 (b) $30,000-59,999
<$14,999 9 2
$15,000-29,999 8 5
$30,000-59,999 5 9
$60,000-89,999 0 7
>$90,000 0 3

(a)n = 25.

(b)n = 22.
Table II.

Desription of Constructs and Scales

Construct Measure

Problem behavior
Parent report Child Behavior Checklist (b)
Child report Youth Self-Report (b)
Sibling collusion
Macrosocial coding Sibling Interaction Rating (c)
Macrosocial coding Coder Impressions (d)
Macrosocial coding Coder Impressions (d)
Deviant peer group
Child report Child Interview (e)

86
 Child: Peers/Social Skills (d)
Parent report Parent's Beliefs and Peers (d)
Teacher report Teacher Peers/Social Skills (d)

Construct Scale Cronbach's [alpha]

Problem behavior .32 (a)

Parent report Externalizing .84
Child report Externalizing .85
Sibling collusion .76
Macrosocial coding Sibling collusion .93
Macrosocial coding Positive relationship .65
Macrosocial coding Negative relationship .87
Deviant peer group .62
Child report Peer drug use .85
Antisocial peers .84
Deviant peer group .76
Deviant peer group .73
Parent report Deviant peer group .80
Teacher report Deviant peer group .80

(a)Bivariate correlation.

(b)Achenbach (1991).

(c)Bullock & Dishion (1999).

(d)Dishion & Kavanagh (in press).

(e)Oregon Social Learning Center (1984).

Table III.

Bivariate Correlations for Scales Used to Create Sibling Collusion,

Deviant Peer Group, and Problem Behavior Constructs

Problem behavior PCBC externalizing mom YSR externalizing child

PCBC 1.00
YSR .40 (***) 1.00
COIMP positive COIMP negative
Sibling collusion communication communication

COIMP positive relationship 1.00

COIMP negative relationship .61 (*****) 1.00
SIB-R sibling collusion .36 (***) .55 (*****)

SIB-R
Sibling collusion collusion

COIMP positive relationship

COIMP negative relationship
SIB-R sibling collusion 1.00
Deviant peer group CINT CPRSK PPRSK TRPSK

CINT 1.00
CPRSK .39 (***) 1.00
PPRSK .25 (*) .57 (***) 1.00
TRPSK .41 (***) .12 .02 1.00

Note. PCBC = Child Behavior Checklist, Parent Form;

YSR = Youth Self-Report;

COIMP = Coder Impressions;

SIB-R = Sibling Interaction Ratings;

TC = Target Child;

CINT = Child Interview;

CPRSK = Child: Peers/Social Skills;

PPRSK = Parent Beliefs and Peers;

TRPSK = Teacher Peers/Social Skills.

(*)p < .10.

(***)p < .01.

87
(****)p < .005.

(*****)p < .001.

Table IV.

Percentage of Sibling Collusion by Item

Collusion item Siblings %

Negative Behavior 9.5

Avoid Parental Direction 13.4
Share Deviant Stories 21.1
Discuss Deviant Friends 22.0
Mutual Deviant Behavior 17.3
Table V.

Interitem Pearson Correlations for Sibling Interaction Ratings

Sibling
NB APD SDS DDF

Sibling
Negative 1.00
Behavior (NB)
Avoid Parental .49 (**) 1.00
Direction (APD)
Share Deviant .57 (*****) .71 (*****) 1.00
Stories (SDS)
Discuss Deviant .58 (*****) .53 (**) .68 (*****) 1.00
Friends (DDF)
Mutual Deviant .71 (*****) .22 .51 (**) .18
Behavior (MDB)

Sibling
MDB

Sibling
Negative
Behavior (NB)
Avoid Parental
Direction (APD)
Share Deviant
Stories (SDS)
Discuss Deviant
Friends (DDF)
Mutual Deviant 1.00
Behavior (MDB)

(**)P <.05.

(*****)P <.001.
Table VI

Target Child Externalizing, Sibling Collusion, and Relationship Scores

by Risk

Risk (SD) Norm (SD)

Sibling Collusion (SIB-R) 0.29 (0.45) 0.09 (0.22)

Sibling Positive 2.33 (1.17) 0.77 (0.52)
Relationship (COIMP)
Sibling Negative 2.32 (1.82) 0.38 (0.18)
Relationship (COIMP)
Youth Self-Report (YSR (a)) 16.84 (10.12) 9.56 (7.40)
Parent Report (CBCL (b)) 13.00 (7.93) 5.38 (6.72)
Deviant Peers 0.15 (0.67) -0.21 (0.69)

F ratio

Sibling Collusion (SIB-R) 4.06 (**)

Sibling Positive 38.76 (*****)
Relationship (COIMP)
Sibling Negative 29.37 (*****)
Relationship (COIMP)
Youth Self-Report (YSR (a)) 8.43 (****)
Parent Report (CBCL (b)) 12.98 (*****)
Deviant Peers 3.54 (*)

Note. SD = standard deviations; SIB-R = Sibling Interaction Ratings;

COIMP = Coder Impressions.

88
(a)YSR, Achenbach (1991).

(b)CBCL, Achenbach (1991).

(*)P <.10

(**)P <.05.

(****)P <.005.

(*****)P <.001.
Table VII

Summary of Model Predicting Problem Behavior Using Sibling Collusion

and Deviant Peer Group

Model R [R.sup.2] SE [R.sup.2] change

1. (Constant)
Collusion construct .39 .15 .81 .15
2. (Constant)
Collusion construct .58 .34 .72 .19
Deviant Peer construct

Model F change Unstandardized [beta]

1. (Constant) -.35
Collusion construct 8.78 (****) -.34
2. (Constant) -.22
Collusion construct 13.51 (*****) .22
Deviant Peer construct .56

Model Standardized [beta] t Tol. (a)

1. (Constant) -2.20
Collusion construct .39 2.96 (****) 1.00
2. (Constant) 1.53
Collusion construct .26 2.13 (**) .92
Deviant Peer construct .45 3.675 (*****) .92

(a)Tolerance

(**)p <.05.

(****)p <.005.

(*****)p <.001.

ACKNOWLEDGEMENTS

This research was supported by training Grand No. 5-T32-MH20012 from the National
Institute of Mental Health and Grant No. DA 07031 from the National Institute on Drug
Abuse, to Thomas J. Dishion, PhD. We are grateful to the Project Alliance staff, Alison
Schneiger, Kate Kavanagh, and Patty Harrington for their assistance on this project, to
Kirby Deater-Deckard for his helpful comments, and to Ann Simas for he editorial support.

Received January 15, 2000; revision received December 4, 2000; accepted August 17,
2001

REFERENCES

Abramovitch, P., Carter, C., Pepler, D. J., & Stanhope, L. (1986). Sibling and peer
interaction: A final follow-up and a comparison. Child Development, S7, 217-229.

Achenbach, T. M. (1991). Manual for the Child Behavior Checklist/4-18 and 1991 profile.
Burlington: University of Vermont, Department of Psychology.

Arnold, J. E., Levine, A. G., & Patterson, G. R. (1975). Changes in sibling behavior following
family intervention. Journal of Consulting and Clinical Psychology, 43, 683-688.

89
Brook, J. S., Whiteman, M., Gordon, A. S., & Brenden, C. (1983). Older brother's influence
on young sibling's drug use. Journal of Psychology, 114, 83-90.

Brownfield, D., & Sorenson, A. M. (1994). Sibship size and sibling delinquency. Deviant
Behavior, 15, 45-61.

Bullock, B. M., & Dishion, T. J. (1999). Sibling Interaction Ratings. Unpublished coding
instrument. (Available from Child and Family Center, University of Oregon, 195 West 12th
Avenue, Eugene, OR 97401-3408)

Bullock, B. M., & Dishion, T. J. (2001, April). Sibling collusion, parenting, and peers:
Toward unraveling a multidirectional process. In E. A. Stormshak & B. M. Bullock (Chairs),
Sibling relationships as a context for the development of risk. Paper presented at the
biennial meeting of the Society for Research in Child Development, Minneapolis, MN.

Capaldi, D. M., & Patterson, G. R. (1991). Relation of parental transitions to boys'

adjustment problems: I. A linear hypothesis. II. Mothers at risk for transitions and
unskilled parenting. Developmental Psychology, 27, 489-504.

Chamberlain, P., Patterson, G. R., Reid, J. B., Kavanagh, K., & Forgatch, M. S. (1984),
Observation of client resistance. Behavior Therapy, IS, 144-155.

Crouter, A. C., McHale, S. M., & Perry-Jenkins, M. (1990). Parental monitoring and
perceptions of children's school performance and conduct in dual- and single-earner
families. Developmental Psychology, 26, 649-657.

Dadds, M. R., Sanders, M. R., Morrison, M., & Rebgetz, M. (1992). Childhood depression
and conduct disorder: II. An analysis of family interaction patterns in the home. Journal of
Abnormative Psychology, 101, 505-513.

Dishion, T. J., Andrews, D. W., Kavanagh, K., & Soberman, L. H. (1996). Preventive
interventions for high-risk youth: The Adolescent Transitions Program. In B. McMahon & R.
D. Peters (Eds.), Childhood disorders, substance abuse and delinquency: Prevention and
early intervention approaches (pp. 184-214). Newbury Park, CA: Sage.
Dishion, T. J., Capaldi, D., Spracklen, K. M., & Li, F. (1995). Peer ecology of male
adolescent drug use, Development and Psychopathology, 7, 803-824.

Dishion, T. J., Eddy, M., Haas, E., Li, F., & Sparcklen, K. (1997). Friendships and violent
behavior during adolescence. Social Development, 6, 207-223.

Dishion, T. J., & Kavanagh, K. (2000). An ecological approach to family intervention for
adolescent substance use. In W. De Mey, E. Moens, K. Van Leeuwen, & L. Verhofstadt-
Deneve (Eds,), Prevention of antisocial behavior: Starring at (pre-) conception?
Tendencies in care and family support (pp. 137-154). Brussels, Belgium: University of
Ghent.

Dishion, T. J., & Kavanagh, K. (in press). Adolescent problem behavior: An intervention
and assessment sourcebook for working with families in schools. New York: Guilford.

Dishion, T. J., & McMahon, R. J. (1998). Parental monitoring and the prevention of child
and adolescent problem behavior: A conceptual and empirical formulation. Clinical Child
and Family Psychology Review, 1, 6 1-75.

Dishion, T. J., & Patterson, G. R. (1993). Antisocial behavior: Using a multiple gating
strategy. In M. I. Singer, L. T. Singer, & T. M. Anglin (Eds.), Handbook for screening

90
adolescents at psychosocial risk (pp. 375-399). New York: Lexington Books.

Dishion, T. J., Patterson, G. R., Stoolmiller, M., & Skinner, M. L. (1991). Family, school, and
behavioral antecedents to early adolescent involvement with antisocial peers.
Developmental Psychology, 27, 172-180.

Dishion, T. J., Spracklen, K. M., Andrews, D. W., & Patterson, G. R. (1996). Deviancy
training in male adolescent friendships. Behavior Therapy, 27, 373-390.

Elliott, D., Huizinga, D., & Ageton, S. (1985). Explaining delinquency and drug use.
Beverly Hills, CA: Sage.

Forgatch, M. S. (1984). A two-stage analysis of family problem solving: Global and

microsocial. Unpublished doctoral dissertation, University of Oregon, Eugene.

Forgatch, M., Fetrow, B., & Lathrop, M. (1984). Coder Impressions. Unpublished
instrument. (Available from Oregon Social Learning Center, 160 East 4th Avenue, Eugene,
OR 97401-2426)

Forgatch, M. S., & Patterson, G. R. (1989). Parents and adolescents living together: Vol. II.
Family problem-solving. Eugene, OR: Castalia.

Henggeler, S. W., Schoenwald, S. K., Borduin, C. M., Rowland, M. D., & Cunningham, P. B.
(1998). Multisystemic treatment of antisocial behavior in children and adolescents. New
York: Guilford.

Horne, A. M. (1981). Aggressive behavior in normative and deviant members of intact

versus mother-only families. Journal of Abnornative Child Psychology, 9(2), 283-290.

Lewin, L. M., Hops, H., Davis, B., & Dishion, T. J. (1993). Multimethod comparison of
similarity in school adjustment of siblings and unrelated children. Developmental
Psychology, 29, 963-969.

Loeber, R., & Dishion, T. J. (1983). Early predictors of male delinquency: A review.
Psychological Bulletin, 94, 68-99.

Loeber, R., & Dishion, T. J. (1987). Antisocial and delinquent youth: Methods for their early
identification. In J. D. Burchard & S. N. Burchard (Eds.), Prevention of delinquent
behavior(Vol. 10, pp.75- 89). Newbury Park, CA: Sage.
McKillip, J., Johnson, J.E., & Petzel, T. P. (1973). Patterns and correlates of drug use among
urban high school students. Journal of Drug Education, 3. 1-12.

Minuchin, P. (1985). Families and individual development: Provocations from the field of
family therapy. Child Development, 52, 289-302.

Oregon Social Learning Center. (1984). Child Interview. Unpublished instrument.

(Available from Oregon Social Learning Center, 160 East 4th Avenue, Eugene, OR 97401-
2426)

Patterson, G. R. (1982). A social learning approach: III. coercive family process. Eugene,
OR: Castalia.

Patterson, G. R. (1984). Siblings: Fellow travelers in a coercive system. In R. J. Blanchard

& D. C. Blanchard (Eds.), Advances in the study of aggression (Vol. 1, pp. 173-215). New

91
York: Academic Press.

Patterson, G. R. (1986). The contribution of siblings to training for fighting: A microsocial

analysis. In D. Olweus, J. Block, & M. RadkeYarrow (Eds.), Development of antisocial and
prosocial behaviors (pp. 235-260). Orlando, FL: Academic Press.

Patterson, G. R. (1993). Orderly change in a stable world: The antisocial trait as a

chimera. Journal of Consulting and Clinical Psychology, 61, 911-919.

Patterson, G. R., & Dishion, T. J. (1985). Contributions of families and peers to

delinquency. Criminology, 23, 63-79.

Patterson, G. R., Reid, J. B., & Dishion, T. J. (1992). A social learning approach: IV.
Antisocial boys. Eugene, OR: Castalia.

Pfiffner, L. J., Jouriles, E. N., Brown, M. M., Etscheidt, M. A., & Kelly, J. A. (1990). Effects of
problem solving therapy on outcomes of parent training for single parent families. Child
and Family Behavior Therapy, 12, 1-11.

Reiss, D., Neiderhiser, J. M., Hetherington, E. M., & Plomin. R. (2000). The relationship
code: Deciphering genetic and social influences on adolescent development. Cambridge,
MA: Harvard University Press.

Rowe, D. C. (1981). Environmental and genetic influences on dimensions of perceived

parenting: A twin study. Developmental Psychology, 17, 203-208.

Rowe, D. C., Rodgers, J. L., & Meseck-Bushey, S. (1992). Sibling delinquency and the
family environment: Shared and unshared influences. Child Development, 63, 59-67.

Scarr, S., & Grajek, S. (1982). Similarities and differences among siblings. In M. E. Lamb &
B. Sutton-Smith (Eds.), Sibling relationships: Their nature and significance across the
lifespan (pp.357-381). Hillsdale, NJ: Erlbaum.

Soberman, L. H. (1995). Psychometric validation of a brief teacher screening instrument.

Dissertation Abstracts International Section A: The Humanities and Social Sciences, 56(2-
A), 0514. (University Microfilms International)

Stormshak, E. A., Bellanti, C. J., & Bierman, K. L. (1996). The quality of sibling
relationships and the development of social competence and behavioral control in
aggressive children. Developmental Psychology, 32, 79-89.

West, D. J., & Farrington, D. P. (1977). The delinquent way of life. London: Heinemann
Educational Books.

Bernadette Marie Bullock (1,2) and Thomas J. Dishion (1)

(1.) Department of Psychology, University of Oregon Child and Family Center, Eugene,
Oregon.

(2.) Address all correspondence to Bernadette Marie Bullock, University of Oregon Child
and Family Center, 195 West 12th Avenue, Eugene, Oregon 97401-3408; e-mail:
bullock@darkwing.uoregon.edu.

0091-0627/02/0400-0143/0 [c] 2002 Plenum Publishing Corporation

COPYRIGHT 2002 Plenum Publishing Corporation
in association with The Gale Group and LookSmart. COPYRIGHT 2002 Gale Group

92
Adolescence
Winter, 2000

UNDERSTANDING DYSFUNCTIONAL AND FUNCTIONAL FAMILY BEHAVIORS FOR

THE AT-RISK ADOLESCENT.

Author/s: Don Martin

ABSTRACT

At-risk adolescents and their impact on families and society, as well as characteristics of
both healthy and maladaptive families, are discussed. Cognitive distortions of
dysfunctional adolescents and their effect on family members, along with methods for
intervention and creating more healthy environments, are delineated from a systemic
viewpoint.

In the authors' view, the family is the major social unit for emotional development in
adolescents. Thus, understanding families helps therapists conceptualize how
adolescents develop affectively, behaviorally, cognitively, and psychologically (Vernon,
1998).

The family is an integral social system (Becvar & Becvar, 1988), held together by strong
bonds of affection and caring; at the same time, family members exercise control,
approval, and dissent for each other's actions. As part of this interaction, every family has
a structure, whether dysfunctional or functional, chaotic or rigid. This family organization
helps it to achieve goals within a developmental time frame and to survive as a unit
(Kessler, 1988; Thompson & Rudolph, 1998).

At-Risk Adolescents

It is estimated that over seven million American adolescents--one in four--are extremely

vulnerable to multiple high-risk behaviors and school failure, while another seven million
are at moderate risk (Carnegie Council, 1989; Husain & Cantwell, 1992). In today's
society, adolescents are apt to become involved with damaging behaviors, particularly
those associated with alcohol, drugs, sexual activity, sexually transmitted diseases, and
pregnancy. Whether this is due to cultural conditions or erosion of the family unit is
debatable (Wicks-Nelson & Israel, 1991).

Unfortunately, along with these pressures, many young people lack guidance and
support. The path to adulthood has been described as one of isolation. During
adolescence, exploratory behavior patterns emerge. Many of these behaviors carry high
risks and have resulted, for example, in an unprecedented number of alcohol-related
accidents and school dropouts. The need to develop self-esteem and inquiring minds
among our youth has never been more necessary. The Carnegie Council (1989) and other
researchers (Thompson & Rudolph, 1998), in formulating goals for educating adolescents,
note five characteristics of an effective adolescent:

1. Effective adolescents are intellectually reflective persons who have developing thinking
skills. They are able to express themselves in persuasive, coherent writing as well as
verbally; they know the basic vocabulary of the arts, math, and sciences, and have
learned to appreciate a variety of cultures and languages.

2. They are en route to a lifetime of meaningful work. Work is essential to survival, as well
as an integral part of one's identity. Our youth must be knowledgeable about a variety of
career options and not be restricted by race or gender. Certainly high school graduation
will be a prerequisite for entering the work force and it is hoped that they will understand
the advantages of post-secondary education.

93
3. Adolescents will be good citizens, thus taking responsibility for shaping our world. We
need to develop children who are doers, not just subservers--those who can demonstrate
on a daily basis their commitment to their own character, their community, and their
schools. Also it is hoped that they will understand the basic values of our nation and have
an appreciation for both the western and non-western worlds.

4. Adolescents will be caring individuals who are able to think clearly and critically, and
act ethically. Our youth must recognize that there is a difference between right and
wrong, and must have the courage to act on their convictions. They will model values
that have been associated with good family development--including integrity, tolerance,
and appreciation of others. They will understand the importance of close relationships
with family and friends, recognizing that relationships require effort and sacrifice, and
that without them, life has relatively little meaning and can be filled with insecurity and
loneliness.

5. Our youth will understand the correlation between exercise, diet, and health. These
provide a sense of competence and strength. We must help our youth become proficient
because success is directly related to self-image. The effective adolescent will appreciate
personal strengths and work to overcome weaknesses.

It is our belief that every youth in our nation, poor or rich, advantaged or disadvantaged,
should have the opportunity to achieve success, not just minimum competence, in all of
these areas. This is the challenge to our society as a whole--our educational, community
and social-support systems. However, it also is a direct challenge to individuals to help
families maximize their potential.

Unfortunately, many families are unable to cope with the problems faced by adolescents
(Robin & Foster, 1989; Vernon, 1998). Many adolescents are growing into adulthood
alienated from others, and with low expectations of themselves. There is greater
likelihood that they will become unhealthy, addicted, violent, and chronically poor.
Equally disturbing is that adolescents from the more affluent communities are displaying
similar problems. Too many students are dropping out of school or participating at a
minimal level. Even if they graduate, they have few marketable skills and their parents
are not demanding that they acquire these skills.

Affluent parents seem to send mixed messages--that their lives are too demanding, and
at the same time, because of their affluence, they do not see the needs of their troubled
teenagers. When these problems do hit home, parents' reaction is often shock or dismay.

On the other hand, less advantaged families, in struggling to make a living, do not have
the time to build family relationships. Further, greater mobility in quest of economic
opportunities makes family cohesiveness less attainable. In a time of great change, many
parents are confused about their roles and relationships and are less aware of the new
temptations faced by their adolescents (Wicks-Nelson & Israel 1991).

The Carnegie Council (1989, pp. 22-25) reported that in a recent graduating high school
class, 92% had consumed alcohol, and of those, 56% had begun in the sixth through
ninth grades, while 36% had begun in the tenth through twelfth grades. These numbers
do not include those who had dropped out of school, and who were even more likely to
use alcohol. Problem behaviors are also interrelated. For example, young people who
drink often experiment with illegal drugs. They may smoke and engage in unprotected
sex. These same adolescents are more prone to school failure.

More teenagers are becoming sexually active before the age of 16 (Berns, 1993), and
girls are becoming pregnant at a greater rate and dropping out of school early. Young

94
mothers are usually economically disadvantaged, have limited opportunities, and their
pregnancies lead disproportionately to the birth of low-weight babies who are vulnerable
to many poor outcomes. It has been estimated that one-fourth of all sexually active
adolescents will become infected with a sexually transmitted disease before graduating
from high school, AIDS being the greatest concern (Vernon, 1998).

With the increase in risk-taking behaviors and substance abuse, motor vehicle deaths are
also increasing. This is true particularly among those aged 10 to 14 years. This results
from association with older adolescents who have been drinking. For this same age
group, between 1980 and 1985 the suicide rate doubled. Seriously delinquent activities
are peaking now at the age of 15, and of the 28 million boys and girls aged 10 to 17 in
the U.S., 14 million are at moderate or high risk due to substance use and other
deleterious behaviors (Vernon, 1998). The cost of these behaviors to society is several
billion dollars.

Creating Healthy Families

Problem-solving and communication skills are of particular importance, especially when

one considers that the relationship between adolescents and parents may be conflictual.
If these conflicts are not resolved, it is difficult to restore an equitable pattern of family
functioning. The more conflictual the dispute among family members, the greater the
need for resolution skills. Robin and Foster (1989) indicate that in solution-focused
families, members are able to share their feelings without offending others. They are able
to decipher "hidden" messages. Conversely, verbal attacks, shouting, and other power-
oriented techniques usually provoke anger in the recipients.

Reiss (1991) has been studying families that do not exhibit pathology in an effort to
understand how they communicate, coming up with several hypotheses:

1. These healthy families speak clearly. They are not rigid in their discussions, nor are
they confused and chaotic.

2. They tend to agree more often than disagree, and are able to assert themselves
without offending others.

3. They have a friendly environment and are able to disagree without upsetting other
members.
4. They show variation in affect; they can express happiness or sadness to each other.

5. They have a good sense of humor and have the ability to laugh at themselves.

6. They respect each other's need for privacy, and do not engage in mind reading.

Family systems need versatility, the ability to overcome conflict, and the capability to
develop alternative solutions. Healthy families do not accept just any idea. They are not
impulsive; they negotiate and compromise. In families that function effectively, grudges
are not held very long. Arguments are short and followed by more friendly interactions.

In contrast, families that are unhealthy may find a weaker member to "scapegoat." This
helps other family members to feel important. Scapegoating often occurs in families that
are too rigid (authoritarian) or disorganized (laissez-faire). Both of these family structures
contribute to dysfunctional behavior.

An authoritarian power structure is one in which parents impose their values upon their
adolescent children. These children see the adults in the family as demanding and
restrictive. Adolescents frequently have no alternative but to break the rules. Even as the

95
adolescent grows older, authoritarian parents have difficulty renegotiating outdated
rules. Further, they do not receive much input from the adolescent.

At the other end of the spectrum are the permissive or laissez-faire families in which
parents either are too busy or abdicate their parenting responsibilities to social service
agencies or to the adolescent. This can create enormous difficulties for adolescents who
may be conflictual with their parents as they seek independence, but also need a place
where they feel secure and supported and can receive guidance. The permissive family
does not provide this. Adolescents in these families view their parents as disinterested,
and have to make their own decisions in a very complex world. They may seek love in
maladaptive ways, such as by becoming pregnant or through drug-using peer groups.
Permissive parents often see themselves as close to, and understanding of, their children.
Some even are able to communicate on an informal basis; however, most children in
permissive families have a poor self-image and do not develop the skills required in order
to compete in today's society. Permissive family structures are often confused with more
democratic styles, but they are not the same (Becvar & Becvar, 1988; Robin & Foster,
1989).

The democratic style offers a decision-making method in which the parent is responsible
for final decisions, but utilizes problem-solving skills that produce less conflict and greater
adolescent developmental achievement. These parents encourage adolescents to
participate in matters that are of importance to them. Democratic parents recognize that
adaptation, particularly in a society that is rapidly changing, is important; they see their
families as flexible rather than rigid. Democratic families understand that family
members differ and these differences are respected and encouraged. Children do not
have to exhibit maladaptive behavior in order to gain independence. Each member has a
chance to contribute in family discussions. In family projects, everyone gets involved
whenever possible. It is interesting that these families tend to put a positive light on
negative behaviors. For example, if a child is demanding, they see it as assertive (Reiss,
1971).

Democratic families understand that labels placed on youngsters often stay with them for
a long time, often into adulthood. Thus, when they disagree, they do not resort to
accusations or recriminations, but tend to accentuate constructive exchanges. In
contrast, an unhealthy family will accentuate the negative, rather than applying effective
problem-solving techniques.

Cognitive Distortions and the Maladaptive Family

Counselors need to understand that not all families face issues involving large conflicts.
Reiss (1991) noted in his classic study of families that some do not experience conflict
overtly. Children may accept their parents' values, and, yet, remain independent and
assertive. Thus, it is important to avoid labeling all adolescents as confrontational in their
quest for independence.

Robin and Foster (1989) note that cognitive distortions have a great impact on
parent/adolescent relationships in many ways. These distortions may help establish rigid
positions that increase maladaptive behavior, anger generated by negative attributions,
or illogical thinking, which generally escalates hostility among family members. In their
examination of cognitive distortions among parents and children, Robin and Foster have
developed eight themes that describe this phenomenon:

1. Perfectionism -- when parents expect their children to behave flawlessly. At the same
time, adolescents see their parents as always having the correct answer.

2. Ruination -- the belief that if the adolescent engages in maladaptive behavior, there

96
will always be catastrophic consequences; not only the adolescent's life will be ruined,
but also the lives of the other family members. From the adolescent's perspective,
restrictions placed by parents will ruin his or her life.

3. Fairness -- the belief by the adolescent that parents should always treat him or her
fairly and that life should be fair for everyone.

4. Love and Approval -- based on the concept that no one should have secrets and that
everyone should always approve of others' behavior. If you fail to confide, you are lacking
in love for another human being.

5. Obedience--the parents' belief that no matter what they say or do, the adolescent
should agree without question.

6. Self-blame--the adolescent or parent refuses to accept blame for his or her own
mistakes, instead believing that if the other had provided better information or had acted
differently, the mistake would not have been made.

7. malicious Intent--the view that if a person misbehaves, it is done deliberately to hurt

other family members. Criticism and constructive feedback are seen as hurtful.

8. Autonomy--adolescents' belief that they should be able to do whatever they wish

without any restrictions.

Clinicians must recognize that these cognitions serve a purpose for some unhealthy
families. They may provide a sense of balance or help the family avoid intimacy. These
distortions may even be seen as helpful in improving a particular negative quality.

CONCLUSIONS

During their children's adolescence, parents' decision-making becomes even more

difficult due to the complexity of such issues as discipline, schooling, and intimacy. The
adolescent is constantly requesting changes in the rules, and in the process parents may
disagree with each other. Adolescents are adept at recognizing this ambivalence and may
play one parent against the other. Disagreement among parents is not unusual, but when
they are unable to resolve a conflict, the disagreement may lead to maladaptive behavior
by the adolescent (Haley, 1980; Kessler, 1988).

Most of the problems seen in dysfunctional families with adolescents also occur in normal
families; however, the rate of dysfunction is much higher in families that have
maladaptive methods of solving problems. In fact, in many dysfunctional families
adolescence-related issues are a continuation of prior parental difficulties. However, an
implicit goal for every family, even a very unhealthy one, is the growth and preservation
of its members.

Typically, the emergence of an adolescent in the family's life cycle results in a period of
upheaval (Walsh, 1982). The family's parameters undergo continuous evaluation as the
adolescent goes though a period of change--physiological, cognitive, emotional or
behavioral. This period of change necessitates a series of psychosocial adjustments
within the family, the major one focusing on the adolescent's primary developmental task
of becoming independent from parents (Levant, 1984). How the family reacts to conflict
during this period of adjustment determines whether the normal processes of
adolescence will be resolved or whether they will result in pathology and an at-risk
adolescent (Goldenberg & Goldenberg, 2000).

97
Maggie Martin, Ed.D., Principal, McKinley Elementary School, Lisbon, Ohio.

REFERENCES

Becvar, D., & Becvar, R. (1988). Family therapy: A systemic integration. Needham
Heights, MA: Allyn & Bacon.

Berns, R. (1993). Child, family, and community. New York: Harcourt, Brace, and
Jovanovich.

Carnegie Council on Adolescent Development. (1989). Turning points: Preparing

American youth for the 21st century. Washington, DC: Carnegie Corporation.

Goldenberg, I., & Goldenberg, H. (2000). Family therapy: An overview (5th ed.). Monterey,
CA: Brooks/Cole.

Haley, J. (1980). Leaving home: The therapy of disturbed young people. New York:
McGraw-Hill.

Husain, S., & Cantwell, D. (1992). Fundamentals of child and adolescent psychopathology.
Washington, DC: American Psychiatric Press.

Kessler, J. (1988). Psychotherapy of childhood. Englewood Cliffs, NJ: Prentice Hall.

Levant, R. (1984). Family therapy: A comprehensive overview. Englewood Cliffs, NJ:

Prentice Hall, Inc.

Minuchin, S., & Fishman, H. (1981). Family therapy techniques. Cambridge, MA: Harvard
University Press.

Reiss, D. (1971). Varieties of consensual experience. Family Process, 10, 1-35.

Robin, A., & Foster, S. (1989). Negotiating parent-adolescent conflict: A behavioral family
systems approach. New York: Guilford.

Reiss, D. (1991). The family's construction of reality. Cambridge, MA: Harvard University
Press.

Terkelsen, K. G. (1980). Toward a theory of the family life cycle. In E. Carter, M.

Thompson, & L. Rudolph (1998). Counseling children. Monterey, CA: Brooks/Cole.

Vernon, A. (1998). Counseling children and adolescents. Denver, CO: Love Publishing Co.

Walsh, F. (Ed.). (1982). Normal family processes. New York: Guilford.

Wicks-Nelson, R., & Israel, A. (1991). Behavior disorders of children. Englewood Cliffs, NJ:
Prentice-Hall.
COPYRIGHT 2000 Libra Publishers, Inc.
in association with The Gale Group and LookSmart. COPYRIGHT 2001 Gale Group

98