Six-year predictors of problems in a national sample of children and youth: II. Signs of disturbance

OBJECTIVE

To identify 1986 and 1989 variables that significantly predicted signs of disturbance assessed in 1992.

METHOD

1986 parent reports and 1989 parent, teacher, and self-reports of syndromes, competencies, family variables, and stressful experiences were tested as predictors of 1992 reports of academic problems, school behavior problems, receipt of mental health services, suicidal behavior, police contacts, substance abuse, and the sum of these six signs.

RESULTS

The predictors accounted for large percentages of variance in most signs and predicted fairly accurately which members of case-control samples would manifest specific signs. Overall predictive accuracy was similar for both sexes, but many predictors differed for boys versus girls. The six signs were weakly associated with each other but were strongly associated with particular syndromes.

CONCLUSIONS

Signs of disturbance were predictable over a 6-year period despite the diversity of a national sample. Previous manifestations of certain signs were modest predictors of the same signs. The Delinquent Behavior and Attention Problems syndromes, plus stressful experiences, predicted the most signs. Sex differences in predictors argue against generalizing findings and inferences from one sex to the other.

Empirically based assessment and taxonomy of psychopathology: cross-cultural applications. A review

This paper provides an overview of empirically based assessment and taxonomy, as illustrated by cross-cultural research on psychopathology. The empirically based approach uses standardized assessment procedures to score behavioral and emotional problems from which syndromes are derived by multivariate analyses. Items and syndromes are scored quantitatively to reflect the degree to which individuals manifest them, as reported by particular informants. Although the approach to assessing problems and to constructing taxonomic groupings differs from the ICD/DSM approach, there are no inherent contradictions between either their models for disorders nor the criterial features used to define disorders. Cross-cultural comparisons have yielded relatively small differences in problem rates and syndrome structure, plus considerable similarity in associations of problems with sex and SES, as well as similar correlations between reports by different types of informants. Research on variations in problems in relation to culture, sex, age, SES, and type of informant can contribute to improving both the ICD/DSM and empirically based approaches and to a more effective synthesis between them.

Six-year predictors of problems in a national sample of children and youth: I. Cross-informant syndromes

OBJECTIVE

To illuminate the development of psychopathology by tracing 6-year predictive paths to outcomes assessed in terms of empirically based syndromes.

METHOD

A national sample assessed at ages 4 through 12 years via parent reports was reassessed 3 and 6 years later via parent, teacher, and self-reports.

RESULTS

For syndromes having the clearest DSM counterparts, cross-informant predictive paths revealed similar traitlike patterns for Aggressive Behavior in both sexes; Delinquent Behavior was less traitlike, with greater sex differences in predictive paths; the Attention Problems syndrome was developmentally stable, but, surprisingly, it was associated with more diverse difficulties among girls than boys; conversely, Anxious/Depressed was associated with more diverse difficulties among boys than girls.

CONCLUSIONS

Quantification of problems via empirically based syndromes can detect important sex, age, and developmental variations that may be masked by uniform diagnostic cutoff points for both sexes and diverse ages. This may be especially true for diagnostic cutoff points derived mainly from clinical cases of one sex, such as depression for girls versus attention and conduct disorders for boys.

Diagnosis, assessment, and comorbidity in psychosocial treatment research

This paper identifies problems in prevailing terminology and conceptual models that may hinder research on treatment. To avoid the multiple meanings of diagnosis, the term assessment is used in reference to identifying the distinguishing features of individual cases, while taxonomy is used to designate the grouping of cases according to their distinguishing features. Treatment research requires clear specification of the behavioral/emotional problems and competencies targeted for intervention. Artifactual comorbidity can be avoided by specifying treatment targets at several levels, including competencies, specific problems, syndromes, profiles of syndrome scores, and global problem scores. To select subjects for treatment research and to evaluate outcomes, multisource data can be coordinated by using a cross-informant computer program, taxonomic decision tree, and averaging of multisource standard scores.

Competencies and problems reported by parents of Greek and American children, ages 6-11

This study compared parent-reported competencies and behavioral/emotional problems in demographically-matched samples of Greek and American children, ages 6-11. Parents of 356 children of each nationality completed the Child Behavior Checklist (CBCL). Competence scores were higher for American children, except on Academic Competence, where scores were higher for Greek children. Greek scores were significantly higher than American scores on the Withdrawn, Anxious/Depressed, Attention Problems, Delinquent Behavior, Aggressive Behavior, Internalizing, Externalizing, and Total Problem scales. On the Anxious/Depressed syndrome, nationality accounted for 14% of the variance. There were few main effects for sex and age and fewer interactions. The higher problem scores in the Greek sample were partly due to the tendency of Greek parents to use extreme item scores. When items were scored present v. absent, Greek scores were higher only on Withdrawn, Anxious/Depressed, Internalizing, and Total Problems, while American scores were higher on Somatic Complaints and Thought Problems. Nationality differences in rates of referral for mental health services and sample differences in exclusion criteria for prior mental health services may have contributed to differences in problem scores. Results are compared to findings from other cross-cultural studies.

Comorbidity of empirically based syndromes in matched general population and clinical samples

Comorbidity of deviance on eight empirically based syndromes was compared in matched general population and clinical samples of 2,705 children aged 4-18, using a bidirectional formula to avoid confounding effects of differential base rates. Syndromes were assessed via parent ratings on the CBCL, teacher ratings on the TRF, and self ratings on the YSR. Significantly higher comorbidity rates were obtained for clinical than general population samples for all 28 pairings of CBCL syndromes, 15 pairings of TRF syndromes, and 22 pairings of YSR syndromes. Bidirectional comorbidity rates for empirically based syndromes were compared to pairings of comparable DSM-III diagnoses.

Epidemiology of behavioral and emotional problems among children of Jamaica and the United States: parent reports for ages 6 to 11

Research on child behavior problems requires standardized methodology in order to identify similarities and differences between societies. The present study compared parent-reported behavior problems of 360 Jamaican and 946 U.S. children aged 6 to 11. It revealed few differences in individual, total, internalizing (e.g., depression), and externalizing (e.g., fighting) problem scores as a function of nationality, gender, or age. Findings from this and other studies indicate the feasibility of a common methodology in cross-national studies of children's problems, but also the need for further refinement.

Three-year course of behavioral/emotional problems in a national sample of 4- to 16-year-olds: 3. Predictors of signs of disturbance

We tested parent-reported family variables, problems, competencies, and stress as predictors of (a) academic problems, (b) school behavior problems, (c) receipt of mental health services, (d) child's need for additional help, (e) suicidal behavior, (f) police contacts, and (g) the sum of these 6 outcomes. Included in the study were 995 cases manifesting at least 1 sign of disturbance (from Outcomes a-f) and 995 matched controls from a national sample of 2,479 children assessed twice over a 3-year interval. Path analyses identified predictors that were significant across age and sex plus those specific to particular groups. The predictive models accounted for medium to large effects in Time 2 signs of disturbance. Time 2 Child Behavior Checklist scores were significantly associated with all Time 2 disturbance scores. The multiple significant risk factors associated with signs of disturbance indicated variations in pathways leading to particular signs of disturbance.

Epidemiological comparisons of American and French children: Parent reports of problems and competencies for ages 6-11

We compared problems and competencies reported by parents on the Child Behavior Checklist for 1842 demographically matched nonreferred French and American children. American children tended to receive higher competence scores, while French children tended to receive higher problem scores. Scores were higher for French than American children on more Internalizing items than Externalizing items. Nearly all nationality differences were small according to Cohen's (1988) criteria. There were few main effects of sex or age group, and fewer significant interactions. Findings are compared with findings on the Child Behavior Checklist in Holland, Australia, Puerto Rico, and Thailand.

Are American children's problems getting worse? A 13-year comparison

OBJECTIVE

To determine whether the prevalence of children's behavioral/emotional problems changed significantly over a 13-year period.

METHOD

Problems and competencies reported by parents and teachers for a random sample of 7 to 16 years old assessed in 1989 were compared with those reported by parents for a 1976 sample and by teachers for a 1981 to 1982 sample. Parent reports were obtained with the Child Behavior Checklist; teacher reports were obtained with the Teacher's Report Form.

RESULTS

Problem scores were higher and competence scores were lower in 1989 than in the earlier assessments. The secular changes were small but included diverse problems, syndromes, and competencies. Changes did not differ significantly by age, gender, socioeconomic status, nor black/white ethnicity. Correlations of 0.97 to 0.99 between rankings of item scores across 7.5- and 13-year intervals support the stability of the assessment procedures. Despite increases in problem scores, the 1989 U.S. scores were not higher than those in several other cultures.

CONCLUSIONS

Viewed categorically in terms of caseness, more untreated children in the 1989 than the 1976 sample would be considered to need help. Multicohort longitudinal studies now in progress will test predictors of within- and between-cohort change.

Epidemiological comparisons of American and Dutch adolescents' self-reports

OBJECTIVE

To determine the cross-cultural applicability of self-reports of adolescents' problems and competencies.

METHOD

Youth Self-Reports (YSRs) were completed by 803 American and 803 Dutch nonreferred adolescents matched by gender, age, and socioeconomic status.

RESULTS

Comparisons revealed that American adolescents reported many more problems than did Dutch adolescents. This finding contrasts sharply with the American/Dutch similarities in parents' and teachers' reports of children's problem behaviors. Competence scores also were much higher for the American than the Dutch sample.

CONCLUSIONS

Cutpoints for discriminating the normal and the clinical range of YSR scores need to be adjusted to local norms for the Dutch sample. Cultural factors may be responsible for the cross-national differences.

Three-year course of behavioral/emotional problems in a national sample of 4- to 16-year-olds: 3. Predictors of signs of disturbance

We tested parent-reported family variables, problems, competencies, and stress as predictors of (a) academic problems, (b) school behavior problems, (c) receipt of mental health services, (d) child's need for additional help, (e) suicidal behavior, (f) police contacts, and (g) the sum of these 6 outcomes. Included in the study were 995 cases manifesting at least 1 sign of disturbance (from Outcomes a-f) and 995 matched controls from a national sample of 2,479 children assessed twice over a 3-year interval. Path analyses identified predictors that were significant across age and sex plus those specific to particular groups. The predictive models accounted for medium to large effects in Time 2 signs of disturbance. Time 2 Child Behavior Checklist scores were significantly associated with all Time 2 disturbance scores. The multiple significant risk factors associated with signs of disturbance indicated variations in pathways leading to particular signs of disturbance.

Behavioral and emotional problems among Thai and American adolescents: parent reports for ages 12-16

We studied adolescents' behavioral and emotional problems in the United States and in Thailand, a Buddhist country in which, reportedly, aggression is discouraged and self-control, emotional restraint, and social inhibition are encouraged. Standardized parent reports on 118 problems revealed 45 Thai-U.S. differences. Thai adolescents were reported to show more overcontrolled problems (e.g., shyness, compulsivity, inhibition of talking, fearfulness, and constipation) than American adolescents (p < .0001). The two groups did not differ reliably in total undercontrolled problems, but Americans showed higher levels of direct, overt, and interpersonally aggressive undercontrol (e.g., fighting and bullying), whereas Thais showed more indirect and subtle undercontrol that was not interpersonally aggressive (e.g., sulking and sullenness). The findings suggest that different cultures may be linked to different styles of adolescent problem behavior.

Nine-year outcome of the Vermont intervention program for low birth weight infants

Twenty-four low birth weight children who had received an experimental intervention (LBWE) during the neonatal period, 31 control children who had received no treatment (LBWC), and 36 normal birth weight children were compared. The intervention involved seven hospital sessions and four home sessions in which a nurse helped mothers adapt to their LBW babies. At age 9, LBWE children scored significantly higher than LBWC children on the Kaufman Mental Processing Composite, Sequential, Simultaneous, Achievement, Arithmetic, and Riddles scales, after statistical adjustments for socioeconomic status. The LBWE children had also advanced more rapidly in school than had LBWC children. Parent (Child Behavior Checklist) and teacher (Teacher's Report Form) ratings of school functioning were more favorable for LBWE than LBWC children, with especially strong effects on Teacher's Report Form scores for academic performance and the attention problems syndrome. At age 9, LBWE children were not significantly inferior to normal birth weight children on any measure. These results bear out a progressive divergence between the LBWE and LBWC children that first became statistically significant in cognitive scores at age 3. The findings suggest that the intervention prevented cognitive lags among LBW children and that this eventually had a favorable effect on academic achievement, behavior, and advancement in school. The progression from no significant differences between LBWE and LBWC children on early cognitive and achievement scores to significant and pervasive differences in later functioning argues for long-term follow-up periods to evaluate properly the power of behavioral interventions to compensate for biological risks.

Three-year course of behavioral/emotional problems in a national sample of 4- to 16-year-olds: II. Predictors of syndromes

This study examined relations between parents' ratings of children's behavioral/emotional problems, family variables, and stressful experiences as predictors of 3-year outcomes in a nationally representative sample of American children. Outcomes were measured by time 2 parent, teacher, and self ratings on eight empirically derived cross-informant syndromes. Path analyses indicated that parent ratings of each time 1 syndrome predicted parent ratings of the same time 2 syndrome. Family variables and intervening stressful experiences predicted parent and self ratings, but not teacher ratings of syndromes. The number of family members receiving mental health services was the family variable that predicted the most time 2 syndromes. Parent reports of stress predicted parent ratings of time 2 syndromes, whereas child reports of stress predicted self-ratings of time 2 syndromes.

Three-year course of behavioral/emotional problems in a national sample of 4- to 16-year-olds: I. Agreement among informants

Quantitative and categorical indices of psychopathology are reported for a nationally representative longitudinal sample assessed via eight empirically derived cross-informants syndromes, internalizing, externalizing, and total problems. Results showed medium to large stabilities for parents' ratings during a 3-year interval on all comparable scales. Predictive correlations between time 1 parents' ratings and time 2 teacher and self-ratings were weaker than parent-to-parent correlations. Classification of children as deviant showed weaker predictive relations than did quantitative scores. Odds ratios showed that children classified as deviant by parents' time 1 ratings were much more likely to be deviant at time 2 on corresponding parent, teacher, and self-ratings than were children initially classified as nondeviant.

Seven-year outcome of the Vermont Intervention Program for Low-Birthweight Infants

We compared 24 low-birthweight subjects of an experimental intervention (LBWE), 32 no-treatment controls (LBWC), and 37 normal birthweight (NBW) subjects. The intervention involved 7 hospital sessions and 4 home sessions in which a nurse helped mothers adapt to their LBW babies. At age 7, LBWE scored significantly higher than LBWC on the Kaufman Mental Processing Composite (p less than .001), Sequential (p = .02), and Simultaneous (p = .001) Scales, after statistical adjustments for socioeconomic status. LBWE did not differ from NBW (F less than 1). These results bear out a divergence between the LBWE and LBWC that first became statistically significant at age 3. The findings suggest that the intervention prevented cognitive lags among LBW children, and that long-term follow-ups are needed to evaluate the developmental effects of efforts to overcome major biological and environmental risks.

Problems and competencies reported by parents of Australian and American children

We compared problems and competencies reported for 2600 randomly selected nonreferred children in Sydney and the U.S.A. Sydney children scored significantly higher on 82 problem items, with a mean total problem score of 31.6 versus 20.1 for the U.S.A. Nevertheless, item scores correlated 0.92 between countries, most differences between competence scores were small, and patterns were similar for sex, age, socioeconomic status, and internalizing versus externalizing problems. Although higher clinical cutoffs may be needed in Sydney, the similarity of patterns in scores permits calibration of the Child Behavior Checklist between Sydney and the U.S.A. Sex differences found in seven cultures were summarized.

Epidemiological comparisons of Puerto Rican and U.S. mainland children: parent, teacher, and self-reports

U.S. mainland and Puerto Rican nonreferred samples were compared via the Child Behavior Checklist (ages 4 to 16), Teacher's Report Form (ages 6 to 16), and Youth Self-Report (ages 12 to 16). Problem scores were significantly higher in parent and teacher ratings of Puerto Rican than mainland subjects, but were significantly lower in self-ratings by Puerto Rican adolescents. Adolescents in both cultures reported significantly more problems than their parents or teachers did. Most of the significant cross-cultural differences in parent, teacher, and self-ratings of competencies showed more favorable scores for the mainland subjects. High referral rates, a high prevalence of DSM diagnoses, and low scores on the Children's Global Assessment Scale are consistent with the high problem rates reported by Puerto Rican parents and teachers but not with the lower rates reported by adolescents. Different clinical cutoffs may be needed for all assessments in the mainland versus Puerto Rico.

Problems reported for clinically referred American and Dutch children

Child Behavior Checklists completed by parents of 50 children of each sex at each age from 4 to 16 seen in 28 American and 21 Dutch mental health services (N = 2,600) were examined. Analyses controlling for sex, age, and socioeconomic status showed somewhat higher total problem scores for American than Dutch children, with a mean of 57.27 versus 53.18 on a scale ranging from 0 to 240. This nationality difference accounted for less than 1% of the variance in total scores. American children obtained higher scores on more externalizing items than Dutch children. Small but significant differences in total, externalizing, and internalizing problems found between the two samples may reflect nationality differences in referral patterns. Competence scores were significantly higher for American than Dutch referred children, but did not differ as much as in comparable normative samples. Competence scores may reflect cultural differences more than problem scores do.

Replication of empirically derived syndromes as a basis for taxonomy of child/adolescent psychopathology

To advance the empirical identification of child/adolescent syndromes, principal components analyses were performed on four sets of parents' ratings of 8,194 6- to 16-year-olds referred to American and Dutch mental health services. The following syndromes replicated well for both sexes at ages 6-16: Aggressive, Anxious/Depressed, Attention Problems, Delinquent, Somatic Complaints, and Withdrawn. For both age ranges, a syndrome designated as Socially Inept replicated well among boys, and one designated as Mean replicated well among girls. Evidence was also found for a Schizoid syndrome in all sex/age groups and a Sex Problems syndrome among girls at ages 6-11. Syndrome scores discriminated well between nationwide normative and clinical samples. The replicated syndromes contribute to the empirical basis for a taxonomy of the kinds of disorders commonly seen between the ages of 6 and 16.

Epidemiology of behavioral and emotional problems among Thai and American children: teacher reports for ages 6-11

As a sequel to comparison between reports by parents, we compared behavioral/emotional problems of 6-11-yr-old Thai and American children reported by teachers. These revealed higher ratings for Thai than for American children on nearly all problems showing significant cross-national differences. Thai children were rated higher on both overcontrolled and undercontrolled behavior and had more overcontrolled than undercontrolled problems (p less than 0.0001). Boys were higher than girls on all 48 problems than showed significant sex differences. The findings underscore (1) the impact of culture on children's problems in the school setting and (2) the importance of surveying teacher as well as parent perspectives.

A comparison of syndromes derived from the child behavior checklist for American and Dutch girls aged 6-11 and 12-16

Child Behavior Checklists completed by parents of 1848 clinically referred American and Dutch girls aged 6-11 and 12-16 were subjected to principal components analyses with varimax rotations. For the 6-11 yr age group, seven of the nine empirically derived syndromes showed cross-national correlations ranging from 0.80 to 0.98. For 12-16-yr-old girls, all eight of the American syndromes were firmly replicated for Dutch girls with correlations ranging from 0.84 to 0.97. Furthermore, cross-national similarities in the distribution of scores for normative samples supported the use of the same syndrome scales by clinicians and researchers in the two countries.

Multiaxial empirically based assessment: parent, teacher, observational, cognitive, and personality correlates of child behavior profile types for 6- to 11-year-old boys

The purpose of this study was to determine whether profile patterns derived from parents' ratings of clinically referred boys were significantly associated with differences in functioning identified via teachers' ratings, direct observations, cognitive tests, and personality measures. Subjects were 185 6- to 11-year-old boys who were referred to an outpatient psychiatric service or school psychologist and whose Child Behavior Checklist (CBCL) profiles correlated significantly with types previously identified through cluster analysis. Comparisons of boys classified by six CBCL profile types and Internalizing versus Externalizing groupings of these types showed many differences that are not apt to be evident in linear correlations between the different sources of data. The findings provide a basis for research on differences between profile patterns with respect to etiology, course, prognosis, and responsiveness to interventions. They also indicate the potential importance of higher-order patterns of functioning that extend across multiple sources of data.

Minimizing adverse effects of low birthweight: four-year results of an early intervention program

The outcome of an early intervention program for low-birthweight (LBW) infants was examined in this study. The intervention consisted of 11 sessions, beginning during the final week of hospitalization and extending into the home over a 3-month period. The program aimed to facilitate maternal adjustment to the care of a LBW infant, and, indirectly, to enhance the child's development. Neonates weighing less than 2,200 grams and under 37 weeks gestational age were randomly assigned to experimental or control conditions. A full-term, normal birthweight (NBW) group served as a second control. 6-month analyses of dyads who completed all assessments over a 4-year period (N's = 25 LBW experimental, 29 LBW control, and 28 NBW infant-mother dyads) showed that the experimental group mothers reported significantly greater self-confidence and satisfaction with mothering, as well as more favorable perception of infant temperament than LBW control group mothers. A progressive divergence between the LBW experimental and LBW control children on cognitive scores culminated in significant group differences on the McCarthy GCI at ages 36 and 48 months, when the LBW experimental group caught up to the NBW group. Possible explanations for the observed delay in the emergence of intervention effects on cognitive development and the mediating role of favorable mother-infant transactional patterns are discussed in light of recent evidence from the literature.

Empirically based assessment of the behavioral/emotional problems of 2- and 3- year-old children

The aim was to determine whether ratings of 2- and 3-year-olds could yield more differentiation among their behavioral/emotional problems than the internalizing-externalizing dichotomy found in previous studies. The 99-item Child Behavior Checklist for Ages 2-3 (CBCL/2-3) was designed to extend previously developed empirically based assessment procedures to 2-and 3-year-olds. Factor analyses of the CBCL/2-3 completed by parents of 398 2- and 3-year-olds yielded six syndromes having at least eight items loading greater than or equal to .30 and designated as Social Withdrawal, Depressed, Sleep Problems, Somatic Problems, Aggressive, and Destructive. Second-order analyses showed that the first two were related to a broad-band internalizing grouping, whereas the last two were related to a broad-band externalizing grouping. Scales for the six syndromes, two broad-band groupings, and total problem score were constructed from scores obtained by 273 children in a general population sample. Mean test-retest reliability r was .87, 1-year stability r was .69, 1-year predictive r with CBCL/4-16 scales at age 4 was .63, 2-year predictive r was .55, and 3-year predictive r was .49. Children referred for mental health services scored significantly higher than nonreferred children on all scales. A lack of significant r's with the Minnesota Child Development Inventory, Bayley, and McCarthy indicate that the CBCL/2-3 taps behavioral/emotional problems independently of the developmental variance tapped by these measures.

A comparison of syndromes derived from the Child Behavior Checklist for American and Dutch boys aged 6-11 and 12-16

Child Behavior Checklists completed by parents of 1863 clinically referred American and Dutch boys aged 6-11 and 12-16 were subjected to principal components analyses with varimax rotations. For each age group, the construct validity of seven empirically derived syndromes was supported by cross-national correlations ranging from 0.80 to 0.98. These syndromes were designated as Aggressive, Delinquent, Depressed (ages 6-11 only), Hyperactive, Schizoid (ages 12-16 only), Somatic Complaints, Uncommunicative, and Withdrawal (Social Withdrawal for ages 6-11; Hostile Withdrawal for ages 12-16). Cross-national similarities in the distribution of scores for normative samples supported the use of the same syndrome scales by clinicians and researchers in the two countries. The cross-national construct validity of Obsessive-Compulsive and Immature syndromes did not receive adequate support.

Psychopathology of childhood

We considered problems of assessment and taxonomy that have hindered the study of childhood disorders. Many instruments have been developed for assessing children's behavior as seen by parents, teachers, clinicians, trained observers, peers, and children themselves. Multivariate analyses of behavior problems show consistencies in the identification of a few broad-band patterns and more numerous narrow-band syndromes, despite differences in specific items comprising different instruments. Satisfactory reliability has been achieved in many instances, although the variability of children's problem behavior limits agreement among informants who see children under different conditions. Despite their disagreements, however, each informant's perspective may validly contribute to the comprehensive assessment of a child's needs. The quantitative, relativistic, and developmental aspects of children's behavior disorders argue for replacing categorical approaches to assessment and taxonomy with taxometric approaches that capitalize on quantitative methods for identifying and assessing children's behavioral problems and competencies.

A typology of child behavior profile patterns: distribution and correlates for disturbed children aged 6--16

This article reports the construction of a typology of behavior problem patterns, as scored on the Child Behavior Profile. Hierarchical cluster analysis was used to identify reliable profile patterns that characterize clinically referred boys and girls aged 6-11 and 12-16. Procedures for classifying the profiles of individual children were developed and good agreement was obtained between classifications based on rating by mothers and a clinician. The distribution of profile patterns was determined for each age and sex group. Children classified by different patterns were found to differ significantly in age, race, school performance, and amount and quality of participation in activities and social relationships. Issues in the application of cluster analysis to behavioral data were analyzed and various options for the classification of individual cases were compared.