Journal of Consulting and Clinical Psychology 1994. Vol. 62, No. 1. 147-156
Copyright 1994 by the American Psychological Association, Inc. 0022-006X/94/S3.00
Enhancing Treatment Gains in a School-Based Intervention for Children of Divorce Through Skill Training, Parental
Involvement, and Transfer Procedures
Arnold L. Stolberg and Jeffrey Mahler
The school-based Children’s Support Group procedure teaches skills to cope with divorce-related events and provides strategies for mastering disrupted developmental tasks. Ss were 103 3rd- through 5th-grade children of separated or divorced parents who were assigned to 1 of 3 treatment groups: support; support and skill building; support, skill building, transfer, and parent training procedures; or no-treatment control. Twenty-six children from intact homes served as nonstressed controls. The two skill-building conditions yielded durable improvements in adjustive behaviors in the home. Transfer components yielded additional improvements in affect, but the absence of substantial in- crements in benefits suggests the need for a closer look at the format and expectations of the transfer vehicle. The benefits of the support-alone condition were experienced most by children who entered the intervention with significant problems, with the greatest reductions in clinical symptomatology at follow-up being found in this group.
The alarming divorce rate in the United States has placed a severe demand on children’s educational and mental health facilities. Among the strategies developed to meet the needs of children of divorce are child-centered groups offered in schools and social service agencies (Grych & Fincham, 1992). The child-centered groups have received substantial attention both because of their practical advantages and because of the increas- ing body of empirical literature supporting their effectiveness (Kalter, Pickar, & Lesowitz, 1984; Pedro-Carroll & Cowen, 1985; Stolberg & Garrison, 1985). Because most children at- tend school daily, school-based groups have the advantage of serving the largest number of children. Peers in the group may serve as active intervention elements through the unique health- promoting impact of their social support and shared perspec- tives and experiences.
A predominant conceptual theme that underlies several school-based prevention strategies (Pedro-Carroll & Cowen, 1985; Stolberg & Garrison, 1985) attempts to mitigate the harmful effects of environmental changes and interparent hos- tility that often accompany divorce. The processes of change and conflict are significant sources of distress and confusion to children (Stolberg & Anker, 1984) and serve to disrupt chil- dren’s mastery of important developmental tasks (Wallerstein, 1983). Reduced parental availability, limited financial re- sources, and altered living circumstances may result in in- creased feelings of anger and frustration. Interparent hostility
Arnold L. Stolberg and Jeffrey Mahler, Department of Psychology, Virginia Commonwealth University.
This research was supported by a grant from the William T. Grant Foundation. We thank the staffof the elementary schools of Chesterfield County, Virginia, for their assistance in this project.
Correspondence concerning this article should be addressed to Ar- nold L. Stolberg, Department of Psychology, Virginia Commonwealth University, 810 West Franklin Street, Richmond, Virginia 23284-2018.
and new life demands facing parents often interfere with par- enting roles and effectiveness. Stolberg and his colleagues Chil- dren’s Support Group and the Children of Divorce Intervention Program (CODIP) by Pedro-Carroll and her colleagues respond to these conceptual underpinnings. Anger control, relaxation, and communication skills are taught to help children cope with the stressors associated with divorce. Alternative systems to as- sist children in the mastery of developmental tasks, such as identity and internal controls are provided.
Program evaluation results support current programs and provide direction for their evolution. The Divorce Adjustment Project (Stolberg & Garrison, 1985) investigated the impact of three interventions on 82 children of divorce whose parents had been separated for 9 to 33 months. Seven- to 13-year-old chil- dren were assigned to a children’s support group condition, a single parents’ support group, to concurrent children’s support and single parents’ support groups, or to a no-treatment control group. The children’s groups provided structured peer support, worked to clarify misunderstandings about the divorce, and taught cognitive-behavioral skills to assist children in coping with divorce-related stressors (e.g., anger control, communica- tion skills) and in mastering developmental tasks that may have been disrupted by the divorce (e.g., impulse control skills, self- definition, social skills). Parents’ groups attempted to foster adults’ divorce adjustment, provided support, and taught skills to facilitate adult development. Secondary efforts provided in- formation and training to enhance single parenting. Evaluation data suggested that participation in the children’s support group alone yielded increased self-esteem immediately after treatment and improved social skills at the 5-month follow-up. Participation in the single parents’ support groups resulted in adult adjustment gains in the absence of parenting skill and child adjustment gains. Parent participation, it was concluded, yields immediate benefits to the children only if the interven- tion is directly aimed at improving parenting competence. En-
148 ARNOLD L. STOLBERG AND JEFFREY MAHLER
hancing adult adjustment did not yield immediate improve- ments in parenting competence.
CODIP (Alpert-Gillis, Pedro-Carroll, & Cowen, 1989; Pe- dro-Carroll & Cowen, 1985) expanded the Children’s Support Group program to include younger children and children from economically disadvantaged homes and improved on the for- mat of the program. The structure of the sessions was given a game-like format, thus increasing children’s interest and active participation in the program. Comparisons of CODIP (Pedro- Carroll & Cowen, 1985) and the Children’s Support Group (Stolberg & Garrison, 1985) procedures and evaluation results indicated a wider range of psychological gains in the former project, perhaps resulting from the more engaging format of CODIP. Teachers rated treatment-group children as displaying fewer shyness, anxiety, and learning problems and greater adap- tive assertiveness, peer sociability, rule compliance, and toler- ance to frustration than control-group children (Pedro-Carroll & Cowen, 1985). Treatment-group children’s self-reported anx- iety was significantly reduced, and their parents reported sig- nificant increases in overall adjustment.
The influence of participants’ age on program content is most clearly seen in the Children of Divorce Developmental Facilita- tion Group (Kalter et al., 1984). The program is based on the conceptualization that older children experience divorce as a predominantly cognitive experience (Kalter, 1989). Thus, pro- gram content was oriented toward dynamic, psychosocial, and sex role issues. Support was provided by peers, reflecting ado- lescents’ normal reliance on peers over parents for emotional assistance. Improvements in boys’ aggression and externalizing problems and improvements in all children’s sad and insecure feelings were found (Plunkett & Kalter, 1984).
Current evaluation models for prevention programs for chil- dren of divorce surfer from four major limitations: (a) Idiosyn- cratic measures lack requisite demonstrations of validity and reliability (Emery, Hetherington, & Dilalla, 1984); (b) Adjust- ment gains are typically defined by unidimensional instruments that hamper the assessment of the program’s overall impact (Warren & Amara, 1984); (c) Assessment strategies frequently use only one rater of children’s adjustive gains, giving limited and perhaps biased views of program benefits; finally, (d) al- though programs may report statistical improvement of sub- jects, the clinical relevance of these findings is unclear. Using standardized, multidimensional instruments, that are com- pleted by raters who view children’s behaviors in different set- tings and that include a linkage to clinical dimensions may re- solve the aforementioned problems.
Five important directions for modification of interventions for children of divorce are drawn from the previous studies. First, adult involvement should be directed toward parenting skill development to yield immediate, adjustive gains for chil- dren. Intentional procedures to transfer therapeutic gains to real-world demands should be integrated into the intervention strategy (Goldstein, 1981). Overlearning and repeated rehearsal (Kanfer, 1979), the use of therapeutic homework (Shelton, 1979), maximizing parents’ use of “real-life reinforcers” (Nay, 1979), and increasing social support and modifying the social environment (Price, 1979) are relevant examples. Third, the role of support as an active intervention ingredient should be explored. The game-like format of program elements to maxi-
mize children’s interest is the fourth revision element. Improv- ing evaluation methodology is the fifth direction for revision.
Our study builds on earlier evaluation studies (Stolberg & Garrison, 1985; Pedro-Carroll & Cowen, 1985). Added to ex- isting intervention strategies are therapeutic home workbooks for children and parents and the increased use of game-like ac- tivities to engage participants. The workbooks have been inte- grated into the program to facilitate the transfer of children’s therapeutic gains, to increase parent support, and to enhance parenting behaviors specifically along dimensions that are per- tinent to their children’s divorce adjustment. The relative con- tributions of peer and parent support on program outcomes is being investigated. The assessment strategy has been expanded to include clinically relevant dimensions of adjustment, rated by individuals who observe the children in different settings.
Subjects and Conditions
Letters describing the intervention program were sent to parents of all children in third through fifth grades in 11 suburban elementary schools. This procedure yielded all of the 103 subjects of separated or divorced parents. Twenty-six intact-family children were recruited sep- arately from participants in a school-based program to prepare children for entrance into one of the county’s middle schools and served as a nonstressed, intact-control group. Some of the divorce-group children participated in the middle school preparation program. Sample com- position was 55% female (n = 71) and 45% male (n – 58). Subjects ranged in age from 8 to 12 years, with a mean age of 9.8 years. Racial composition of the sample was 86% White and 13% African-American; 1 % were of another race. Thirteen percent of the sample were third graders, 29% were fourth graders, and 57% were fifth graders.
Eighty percent of the subjects (n = 103) consisted of children whose parents had been separated or divorced for less than 48 months. Chil- dren whose parents had been separated or divorced for more than 4 years or had a history of using mental health services for more than 1 year before participation in the study were excluded. The average time between the initial marital separation and beginning the intervention was 37.26 months. Parents had divorced in 47% of these families, were separated in 32% of the families, and had remarried in 21% of the fam- ilies. Mothers were custodians in 74% of the cases, and fathers were custodians for 9% of the children. Joint custody was established in 17% of the families. The nonstressed, intact-control group appeared to be a normally distributed group of children along adjustment dimensions. Preintervention, Child Behavior Checklist (CBCL) means and standard deviations are equivalent to the distributions for the normative samples on the CBCL (M = 50, SD = 10).
Average family income ranged from below $5,000 to over $75,000, with a mean of $26,100. Sixty-two percent of the families earned less than $23,000 annually, and 23% had annual incomes of between $23,000 and $45,000. The remaining 15% of the sample earned annual incomes in excess of $45,000.
Analysis of preintervention, Child Assessment Schedule (CAS) in- terviews indicated that 58% of the subjects (n = 75) did not meet criteria for diagnosis according to the Diagnostic and Statistical Manual of Mental Disorders (3rd ed.; DSM-III; American Psychiatric Associa- tion, 1980), whereas 42% (n = 54) received a clinical diagnosis (see Ta- ble 1). Prevalence of diagnosis differed significantly by marital status, with 48% of the children of divorce (n = 50) and 15% of intact-family children (n = 4) receiving a diagnosis, X
2(l, N = 129) = 9.38, p < .01.
INTERVENTIONS FOR CHILDREN OF DIVORCE 149
Table 1 Frequencies of Diagnoses
Type of diagnosis
Attention deficit disorder with hyperactivity
Attention deficit disorder without hyperactivity
Separation anxiety disorder Overanxious disorder Oppositional disorder Enuresis Major depression Dysthymic disorder Simple phobia Obsessive-compulsive
disorder No diagnosis
1 20 6 6 2 4 5 3
0.8 15.5 4.7 4.7 1.6 3.1 3.9 2.3
Groups were jointly led by doctoral students in clinical psychology and by school guidance counselors and vice principals. Graduate stu- dents and school personnel were trained in the administration of pro- gram procedures and used procedure manuals to ensure consistency of program procedures across groups. Student co-leaders received weekly supervision.
The comprehensive, 14-week intervention consisted of three major components: support or special topics; skill building; and skills transfer. For consistency of intervention procedures, all program procedures were manualized in the LeadersGuide (Stolberg, Zacharias, & Cam- plair, 1991). The support or special topics component of the Children’s Support Group procedure operated as an independent, eight-session in- tervention in the support-alone condition and as half of eight of the sessions in the transfer, skills, and support condition and in the skills and support condition. The discussion of specific themes for each ses- sion was accomplished through the use of cartoon and pictorial stimulus material, writing newspaper articles, and games developed for the inter- vention.
The skill building component consisted of part of five sessions dedi- cated to teaching children to label feelings, to associate them with causal events, and to combine the feelings and events into statements to others. Part of four sessions were devoted to teaching children self-control and problem-solving skills and to helping children apply these skills to di- vorce-related problems. Children were taught to determine whether problems were solvable and whose responsibility it was to solve them. Anger control was the focus of part of the final two treatment sessions. Three sessions were devoted to introductory and termination activities.
In summary, the transfer, skills, and support condition and the skills- and-support condition were composed of two initial sessions that were devoted to introductory activities. Half of the next eight sessions were devoted to support or special topics activities. The other half of these sessions focused on teaching children about feelings and self-control techniques. Self-control and anger control procedures were taught in all of the next three sessions. The final session was devoted to termination activities.
The skills transfer component comprised a combined KidsBook and ParentsBook (Stolberg et al., 1991) and a series of four parent work- shops. The KidsBook is an activity and home workbook that helps chil- dren prepare for upcoming sessions, provided practice opportunities for
skills learned in the group, and encouraged parent-child communica- tion. Through drawing and writing exercises, the KidsBook encouraged children to examine their thoughts, feelings, and behaviors related to divorce and to evaluate their progress. Material in the KidsBook corre- sponded to the sequencing and content of the in-school sessions.
The ParentsBook encouraged parents to be involved in their child’s experience with the transfer, skills, and support condition. It directed parents to encourage the use of skills learned and to initiate the discus- sion of important divorce-related topics. It recommended procedures to address issues that may have affected the parent-child relationship and suggested ways in which parents could give support and understand- ing to their child. Material in the ParentsBook corresponds to the se- quencing and content of the in-school sessions and KidsBook activities.
Treatment groups were configured in three ways; transfer, skills, and support; skills and support; and support alone. The session content of the two skills and support groups included three meetings entirely de- voted to introductory or termination activities. Eight sessions were equally divided into skill-building and support activities. The remaining three sessions focused on skill-building exercises alone. The skills transfer components supplemented the skills-and-support activities in transfer, skills, and support. The support-alone groups devoted three sessions to combined introductory or termination activities and support activities and all of the remaining five sessions to support activities.
Random assignment of schools to conditions yielded a transfer, skills, and support group (n – 29); a skills and support group (« = 28); a sup- port-alone group (n = 23); and a divorce, no-treatment control group (n = 23). The intact, no-treatment control group comprised 26 children. Because the project spanned 3 years, many schools participated in sev- eral intervention phases. Participating schools were randomly assigned to treatment condition each semester. Thus, one school may have offered different intervention configurations at different times. Once schools had agreed to participate during a semester, children were re- cruited. Children’s acceptance into their school’s assigned intervention was based only on their family’s willingness to participate in all aspects of the intervention and evaluation.
Within each of the divorce intervention cells, five intervention groups of 5 to 8 children were held. Because of parents’ wishes not to partici- pate in the research component of the project, a few groups included children who did not complete evaluation instruments. The total num- ber of nonresearch subjects was less than 5. The additive effect of step- wise increases in intervention elements on adjustment was assessed by adding the skills activities to the support program, yielding the skills and support condition, and by adding the transfer procedures to the later intervention to produce the transfer, skills, and support interven- tion. Inclusion of the divorce, no-treatment control group facilitated determining the effects of time on the adjustment of children of divorce. Comparisons with the intact, no-treatment control group enabled us to determine how close to nonstressed peers the children of divorce were after treatment.
Measures of child adjustment were divided into four clusters: affect, cognition, behavior in the home, and behavior in the school. The affect cluster comprised the State-Trait Anxiety Inventory for Children (STAI-C; Spielberger, 1973) and the Children’s Depression Inventory. The Self-Perception Profiles (SPP) scales made up the cognition cluster. The cluster for behavior in the home included the CBCL scales (Activi- ties, Social, School, Internalize, Externalize, and the standardized, total pathology score Sum t). Six of the Teacher’s Rating Form scales (Aca- demics, Adaptive Functioning, Happy, Learning, Internalize, External- ize, Sum t) constituted the cluster for behavior in the school.
Children’s Depression Inventory (CDI). The GDI (Kovacs, 1981) is a 27-item self-report rating scale that assesses affective, behavioral, so- cial, attitudinal, and vegetative symptoms of depression in children.
150 ARNOLD L. STOLBERG AND JEFFREY MAHLER
Children select one self-rating from a group of three, ranging from nor- mal to clinically significant symptoms that best describes his or her ex- perience over the past 2 weeks. The measure has established norms for both clinical and normal populations (Kovacs, 1981: Smucker, Craig- head, Craighead, & Green, 1986). Internal consistency estimated by coefficient alpha was .84 and 3-week test-retest reliability with fifth graders ranged between .74 and .77 (Smucker etal., 1986). Validity stud- ies support its usefulness in discriminating between emotionally dis- tressed and normal children (Saylor, Finch, Spirito, & Bennett, 1984).
STA1-C. The STAI-C (Spielberger, 1973) is composed of two 20- item scales measuring the child’s level of state and trait anxiety. Chil- dren select the statement that best describes how they usually feel. The measure has norms for elementary school children. Six-week test-retest reliability coefficients for A-Trait scale were .65 for boys and .71 for girls. Evidence for concurrent validity of A-Trait scale was based on its .75 correlation with the Children’s Manifest Anxiety Scale (Spielberger, 1973).
SPP. The SPP (Harter, 1983) is a 36-item scale designed to assess perceived competence in six areas of children’s self-concept: cognitive competence, athletic competence, social acceptance, physical appear- ance, conduct and behavior, and general self-worth. Children rate each self-description on a 4-point scale. Psychometric data suggest that the measure is appropriate for boys and girls in Grades 3 through 9 and for group administration. Factor analysis supports the validity of the subscales across ages. Test-retest reliabilities ranged from .69 to .87 (Harter, 1983).
Child Assessment Schedule (CAS). The CAS (Hodges, Kline, Stern, McKnew, & Cytryn, 1982) is a standardized diagnostic interview for children designed to be administered by interviewers with experience in childhood psychopathology. It is designed for use with children aged 7 to 12 years old and yields a DSM-III diagnosis. The interview is di- vided into three sections: content area questions, onset and duration questions, and interviewer observational judgments. The completed in- terview yields both diagnostic information and a total psychopatholog- ical index.
Studies have demonstrated interrater reliability of the CAS with kap- pas ranging from .47 to .71, indicating moderate-to-good rater agreement. Evidence for the validity of the CAS includes concordance with the Schedule of Affective Disorders and Schizophrenia for School- Age Children (K.-SADS) interview (Hodges, McKnew, Burbach, & Roe- buck, 1987) and correlations with overall psychiatric ratings ranging between .82 to .87 (Verhulst, Berden, & Sanders-Woudstra, 1985). Inter- rater reliability estimates calculated for the current study yielded Co- hen’s kappa values ranging from .51 to .91, with an overall mean of .78.
Child Behavior Checklist-Parent Form (CBCL). The CBCL (Achenbach & Edelbrock, 1983) contains 118 items, for which parents rate frequency of occurrence on a 3-point scale. The measure yields normed scores on nine problem behavior scales for each sex, a normed total pathology scale, two broad-band second-order factors (Externaliz- ing and Internalizing) and three social competence scales. Higher scores are keyed to frequency of the behavior and reflect either greater proso- cial skills or maladaptive behavior. The t scores on internalizing, exter- nalizing, and total pathology (Sum t) and the three prosocial scales were used in this study. Studies have demonstrated the measure’s ability to discriminant between normal and disturbed children (see Achenbach & Edelbrock, 1983, for a review). Test-retest and interparent correlations of .89 and .74 respectively, have been reported (Achenbach & Edel- brock, 1983).
CBCL-Teacher’s Rating Form (TRF). The TRF developed by Achenbach and Edelbrock (1986) was designed to obtain teachers’ re- ports of children’s problem behavior, school performance, and adaptive functioning. Teachers rate the frequency of occurrence of each of the 133 items. Higher scores indicate greater maladaptive behavior, aca- demic achievement, and adaptive functioning. The t scores on the in-
ternalizing, externalizing, and total pathology (Sum t) scales and the four prosocial scales were used in this study. Two-month test-retest co- efficients for boys aged 6-11 averaged .74 for behavior scales (Achen- bach & Edelbrock, 1986). Teacher-teacher aid agreement was .57.
Teacher, parent, and child pretreatment ratings of all children were collected 1 to 2 weeks before the beginning of the programs. Testing of the children and their parents was done in groups of about 20. Teachers completed their ratings during the school day. Project staff who were unaware of the subjects’ group assignment administered the CASs at the school during the school day. Posttesting of all children was per- formed at the conclusion of the intervention. Follow-up data were col- lected 1 year after completion of the intervention. Teachers and parents were not informed of the multiple formats of the intervention and were, thus, not apprised of the unique aspects of the group to which their children had been assigned. Teachers may have been aware of the larger evaluation project because of their school’s involvement with the pro- gram.
Following recent trends in outcome research methodology, analyses were divided into two major groupings to provide out- come and normative evaluations of program impact (Kendall & Grove, 1988). The outcome analyses questioned whether treatment had produced any adjustive changes through com- parisons of groups of children of divorce in treatment and no- treatment control conditions. The normative analyses com- pared the status of the treatment groups at postassessment and follow-up assessment periods with that of a normative control group of nonstressed, intact-family peers.
Outcome data were clustered into four dimensions: affect, cognition, behavior in the home, and behavior in the school to provide information on the specific levels of program impact (Kendall & Norton-Ford, 1982; Strupp & Hadley, 1977). In the first step of the outcome analyses, we calculated eight sets of repeated measures, multivariate analyses of variance (MANOVAs) using either measures of affect (GDI, STAI-C: Trait, STAI-C:State), cognition (the six SPP scales, CAS-Total Pathology), behavior in the home (CBCL), or behavior in the school (TRF) as dependent measures and time (either pre- to postassessment [pre-post] or pre- to postassessment and postas- sessment to follow-up [pre-post-follow-up]) and divorce group membership as the independent variables. Analyses were con- ducted separately for pre-post and pre-post-follow-up data be- cause of the reductions in power resulting from subject attrition at follow-up. When the MANOVAs were significant, repeated measures (ANOVAs) and Duncan’s multiple range tests were used as post hoc tests. Comparisons reported are significant at p < .05.
In the second step of the outcome analyses, divorce-group children were rated either as having moderate-to-severe adjust- ment problems or as relatively symptomfree, using pre-inter- vention TRF and CBCL Sum t scores. We then calculated two sets of two chi-square coefficients, first using the sample of chil- dren with adjustment problems and then using the sample that was symptomfree. The four levels of divorce groups and the presence or absence of adjustment problems at posttesting con- stituted the categorization variables. The analyses were re-
INTERVENTIONS FOR CHILDREN OF DIVORCE 151
peated with follow-up adjustment ratings. The calculations us- ing the sample of children with adjustment problems at pretest- ing yielded information on which interventions most helped in reducing adjustment problems. Analyses of well-adjusted chil- dren provided some understanding of the deleterious effects of an intervention.
For the normative comparisons, a series of four M ANOVAs were conducted. Affect, cognition, behavior in the home, and behavior in the school clusters at postassessment and then fol- low-up were used as dependent measures, and four of the child groups, excluding the divorce controls, served as the indepen- dent variable. This analysis provided social validation of pro- gram impact and helped to determine similarities between chil- dren of divorce and their nonstressed peers after treatment (Kazdin, 1977).
Table 2 presents preassessment, postassessment, and follow- up group means and standard deviations for each of the study’s dependent variables, MANOVA Fs for the Group X Time in- teractions for pre-post and pre-post-follow-up repeated mea- sures analyses, and ANOVA Fs for the Group X Time interac- tions for the pre-post comparisons. Repeated measures MANOVAs calculated on pre-post data yielded significant Group X Time interactions for the affect cluster, F(3, 95) = 1.76, p < .05, and the behavior in the home cluster, F(3, 90) = 1.75, p < .05. Repeated measures ANOVAs calculated on the individual affect scales identified a significant Group X Time interaction for the STAI-C: Trait measure, F(3, 95) = 3.68, p < .01. Significant univariate Fs for the Group X Time interactions were found for the scales of behavior in the home: Internalizing, F(3, 90) = 4.50, p < .005; Externalizing, F(3, 90) = 6.20, p < .001; and Sum /, F(3, 90) = 6.12, p < .001. Duncan’s multiple range tests calculated on pre-post change scores for the previous four variables indicated that the greatest decreases in Interna- lizing scores (M = 8.23), Externalizing scores (M – 8.54), and Sum / scores (M = 9.42) were found in the skills and support condition. These decreases were significantly greater than those found in the transfer, skills, and support condition (Externaliz- ing [M = 2.63] and Sum t scores [M = 3.42]), in the support- alone condition (Internalizing [M = .09], Externalizing [M = .22], and Sum t scores [M = .23]), and in the divorce controls (Internalizing [M = 2.05], Externalizing [M = .82], and Sum t scores [M = 1.59]). Decreases in trait anxiety were greatest in the transfer, skills, and support condition (M = 6.50), compared with STAI-C: Trait scores in the skills and support condition (M = 2.48) and the support-alone condition (M = 1.87).
Analyses of pre-post-follow-up data yielded significant Group X Time interaction effects in the repeated measures MANOVAs on the behavior in the home cluster, F(3,65) = 2.07, p < .001 (see Table 2). We did not find significant interaction effects for the affect cluster, F(3, 80) = 1.45, ns; cognition clus- ter, F(3, 40) = 1.14, ns; and behavior in the school cluster, F(3, 57) = 0.74, ns. Significant univariate Fs for the Group X Time interactions were found for individual behavior in the home scales: Internalizing, F(3, 65) = 8.46, p < .001; Externalizing, F(3, 65) = 4.12,p < .01; and Sum /, F(3, 65) = 8.24, p < .001. Duncan’s multiple-range tests calculated on the four preassess- ment-to-follow-up change scores indicated that the greatest im- provements were in the transfer, skills, and support condition Internalizing scores (M = 7.63), Externalizing scores (M =
8.31), and Sum t scores (M = 8.88) and in the skills and support condition Externalizing scores (M = 7.77) and Sum t scores (M = 8.54), compared with the scales for divorce controls: Interna- lizing (M = .32), Externalizing (M = 1.64), and sum t (M = 1.86).
In the second stage of outcome evaluations, divorce-group children were rated either as having adjustment problems at program entry or as well-adjusted. Moderate-to-severe adjust- ment problems at program entry were operationalized by either TRF Sum t scores or CBCL Sum t scores greater than or equal to 60 (n = 66). Children with pre-intervention TRF Sum t and CBCL Sum t scores less than 60 (n = 37) were considered to be relatively symptomfree. Divorce group and adjustment ratings at posttesting were used as categorization variables in chi- square analyses on the sample with adjustment problems, yield- ing significant differences in adjustment ratings by group post- treatment, x2(3, N = 60) = 8.70, p < .05. Eight of 21 children in the skills and support condition displayed clinically significant signs of improvement, whereas only 2 of 20 children in transfer, skills, and support, 2 of 11 children in support alone, and none of the children in divorce controls showed similar clinical im- provement (see Table 3). Chi-squares on the same subgroups and categorizations found significant differences in adjustment ratings by group at follow-up, x2(3, N = 41) = 11.46, p < .01. Four of 11 skills and support children and 7 of 10 support-alone children displayed improvements in adjustment, whereas none of the divorce controls and 3 of 14 transfer, skills, and support children showed such gains.
Investigations of children rated relatively well-adjusted be- fore treatment yielded significant differences in posttreatment adjustment patterns, x2(3, N = 32) = 7.61, p < .05. Four of 9 children in the support-alone condition displayed clinically significant adjustment problems, whereas only 2 of 7 children in the transfer, skills, and support condition and 2 of 6 children in the skills and support condition showed deteriorated func- tioning. Again, no changes in adjustment were noted in chil- dren in the divorce controls condition. No significant differ- ences in adjustment ratings by group preintervention to follow- up intervention were found for this initially well-adjusted group, x2(3, N=2\)= 3.86, ns, perhaps because only 1 child in each of the skill-building conditions met the initial adjustment criteria. Interestingly, 5 of 7 support-alone children displayed significant deteriorations in adjustment from pre-intervention to follow-up.
MANOVAs calculated in the normative evaluation indicated no significant posttesting differences in affect, F(9, 339) = 1.05, ns; cognition, F(21, 156) = 1.40, ns; or behavior in the home, F(21, 258) = 1.51, ns, of children of divorce in the treatment groups and intact-family controls. A significant model effect was found for the behavior in the school cluster at posttesting, F(21, 216) = 1.96, p < .01. See Table 2 for group means and standard deviations. After treatment, teachers rated children of divorce as being more poorly adjusted than their intact-family peers on the three TRF pathology subscales: Sum t, F(3, 81) = 4.45, p < .01; Internalizing, F(3, 81) = 4.81, p < .01; and Externalizing, F(3, 81) = 3.89, p < .01. MANOVAs were also calculated with follow-up scores as the dependent measures. No significant differences were found between treatment-group children and their intact-family peers in affect, F(9,282) = 0.97,
152 ARNOLD L. STOLBERG AND JEFFREY MAHLER