Early language impairment and young adult delinquent and aggressive behavior

Early language impairment and young adult delinquent and aggressive behavior

E.B. Brownlie

Early childhood speech and/or language (S/L) impairment has been associated with a range of social, emotional, and behavioral problems. S/L impaired children have an increased risk of psychiatric disorders and of internalizing and externalizing behavior problems (Baker & Cantwell, 1987; Beitchman, Wilson, Brownlie, Walters, Inglis et al., 1996). Both epidemiological studies and studies of children referred to speech and language clinics have shown that the most common types of psychiatric symptoms and disorders of S/L impaired children are attention-deficit/hyperactivity disorder and internalizing disorders, rather than conduct disorder, delinquency, or aggression (Beitchman, Brownlie et al., 1996; Silva, Williams, & McGee, 1987).

However, recent studies using clinic and forensic samples have reported language impairment (LI), often not previously diagnosed, in populations with conduct problems or other forms of antisocial behavior. For instance, in a psychiatric clinic sample of 7- to 14-year-olds, two-thirds of children diagnosed with conduct disorder were language impaired (Cohen, Menna, et al., 1998). Further, a high percentage of adolescents in residential treatment for conduct problems or externalizing behavior have been diagnosed with LI (Giddan, Milling, & Campbell, 1996; Warr-Leaper, Wright, & Mack, 1994). In addition, incarcerated adolescent boys have displayed language deficits in comparison with controls (Davis, Sanger, & Morris-Friehe, 1991).

It is possible that language or speech impairment increases the risk of antisocial behavior, but that these effects are not realized until late adolescence or early adulthood, except among individuals with behavior problems as severe as those in psychiatric clinic samples. Few studies have followed children with S/L impairment to adulthood. Rutter and Mawhood (1991) reported on the social outcomes in adulthood of adults diagnosed with S/L impairments in childhood. However, they did not address antisocial behavior, delinquency, arrests, or convictions. Other follow-up studies of S/L impaired children have not yet reported the psychosocial outcomes of their participants beyond early adolescence, or have not distinguished participants according to their early language and speech functioning (Newman et al., 1996). Prospective longitudinal studies following S/L impaired children to adulthood are needed to determine whether children with LI or speech impairment are at increased risk of antisocial outcomes.

The present study is the first to examine adult antisocial outcomes in a community sample of children with early S/L impairment. Briefly, a random community sample of children in the Ottawa-Carlton area of Ontario, Canada, was assessed at age 5 for language and speech functioning. During extensive testing by speech/language pathologists, children were classified according to their speech and language development.

Children were identified as language impaired if they had substantial deficits or delays in expressive or receptive syntax, morphology, or semantics, on the basis of standardized language tests. Language impairment can involve problems expressing or understanding language. Children with LI may have difficulty with language referring to events, articles, and individuals that are not immediately present or salient, such as the time concepts yesterday or tomorrow, or the spatial concepts behind or beside. Children with expressive language limitations may use language that is vague, nonspecific, and lacking in detail, for example, substituting “it,” “stuff,” “things” in place of specific events or objects. Children with receptive language deficits may have problems understanding and following directions at school and at home. Some language impaired children may have difficulty keeping up with verbal exchanges among their peers, and consequently may become frustrated in social situations (Johnson & Beitchman, 1999a).

Children were diagnosed with speech impairment, which is distinct from LI, if they had substantial difficulties physically articulating spoken language, including dysarthria, stuttering, and voice disorders. Children with speech impairment have difficulty producing developmentally expected speech sounds, which may interfere with their being understood. They may be teased by peers, and in severe cases may experience frustration or reluctance to speak (Johnson & Beitchman, 1999b).

Children with S/L impairment and controls were given a battery of psychosocial, cognitive, developmental, and demographic assessments, and were followed and reassessed at ages 12 and 19. In this 14-year, three-wave longitudinal study, we examine whether children with early language or speech impairment had higher levels of delinquency or aggression in young adulthood than children with normal language and speech. We compare participants who were classified at age 5 as (a) language impaired, including both children with pure LI and children with mixed speech and language impairment; (b) pure speech impaired, which included children with speech impairment without LI, and (c) controls, which included children with no evidence of language or speech impairment. We also consider a number of potentially confounding variables that may account for an association between early LI and delinquent or aggressive behavior.

The first variable we examine is verbal IQ. We test for a specific effect of language impairment, by determining whether any associations between language or speech impairment and antisocial outcomes remain, when controlled for verbal IQ. A large literature has demonstrated a link between verbal IQ and antisocial outcomes, including conduct disorder, delinquency, and aggression, particularly in boys and men (e.g., Farrington, 1995; Lynam, Moffitt, & Stouthamer-Loeber, 1993; Schonfeld, Shaffer, O’Connor, & Portnoy, 1988). LI is associated with lower verbal IQ scores (Silva et al., 1987); however, this is not always the case with children with pure speech impairment (Beitchman, Wilson, Brownlie, Walters, & Lancee, 1996). Although verbal IQ and language ability are related, the two constructs are conceptually distinct. The differences between verbal IQ and language ability are reflected in their measurement and scoring. Language ability includes, at minimum, receptive and expressive semantics, morphology, and syntax. Verbal IQ measures do not systematically assess these aspects of language. For instance, Wechsler verbal IQ scales focus on acquired knowledge rather than language ability. Grammatical errors or mispronunciations are generally ignored in scoring, one-word answers are often acceptable, and aspects of expressive language, particularly expressive morphology and syntax, are not assessed (Cohen, 2001).

In addition to verbal IQ, we consider a number of family and demographic variables that may confound the relation between language or speech impairment and antisocial behavior; socioeconomic status, parent distress, parental history of criminal conviction, and family structure (single parent vs. dual parent family). Using hierarchical regression analyses, we assess whether early language or speech impairment conveys additional risk over and above the effects of both verbal IQ and relevant demographic and family variables.

Finally, using structural equation modeling with observed variables, we examine the role of children’s academic and behavioral functioning in relations between language or speech impairment and antisocial behavior. Academic difficulties are associated with antisocial outcomes (Hinshaw, 1992). In addition, LI is associated with academic difficulties, most notably in reading (e.g., Cohen, Barwick, Horodezky, Vallance, & Im, 1998; Schachter, 1996). Thus, we consider two variables related to academics: age 12 standardized reading achievement scores, and age 12 teacher-rated school performance, in accounting for the antisocial outcomes of children with and without LI or speech impairment. We chose to use teacher ratings rather than achievement test scores to measure academic performance. Performance in the classroom may differ from standardized test performance (e.g., Kimball, 1989), and participants’ experiences of relative success at school are better captured with ratings of classroom performance than with standardized achievement test scores. We consider whether poor literacy skills or poor school performance account for relations between early language or speech impairment and adult delinquency or aggression. Finally, early conduct problems predict antisocial outcomes, and may account for the relationship between some domains of cognitive functioning and antisocial outcomes (e.g., Fergusson & Lynskey, 1997; Maughan, Pickles, Hagell, Rutter, & Yule, 1996; Williams & McGee, 1994). Therefore, we consider age 5 and age 12 conduct problems, as well as age 12 reading achievement and age 12 academic performance, in examining indirect and direct relations between early language and speech impairment and antisocial outcomes.

METHOD

Participants

Wave 1: Age 5

A one-in-three random sample of the 1977 birth cohort of all English speaking children in the Ottawa-Carlton region participated in a speech and language screening. Children who fell below the screening threshold were selected for intensive testing by speech/language pathologists. The final sample consisted of 142 children who failed the intensive testing and 142 controls, matched by sex, age, and classroom or school. For a more detailed description of the procedures used to select the original sample, see Beitchman, Nair, Clegg, and Patel (1986).

The wave 1 sample included 183 (64.4%) boys and 101 (35.6%) girls assessed at a mean age of 5 years, 6 months (hereafter referred to as age 5). Sixteen percent lived with a single parent and 84% had parents who were partnered, either married or cohabiting. Mean nonverbal (performance) IQ was 106.7 (SD = 14.3) for the language or speech impaired group and 119 (SD = 10.8) for the control group. Although participants were selected from a community sample, the sample was not culturally diverse, with more than 90% identified as Caucasian.

Wave 2: Age 12

Of the 284 children who participated in the first wave study, 86% participated at the first follow-up when they were 12 or 13 years of age (hereafter referred to as age 12). Language, cognitive, behavioral, and developmental data were collected from 244 children (113 impaired and 131 control). For a more complete description of the second wave study sample and methods, see Beitchman et al. (1994).

Wave 3: Age 19

At wave three, 258 (90.8%) of the original sample participated. The mean age of participants at the time of testing was 18 years, 10 months (SD = .38), hereafter referred to as age 19. Complete data for the present analyses were available for 168 participants (110 boys and 58 girls). At age 5, 29 of the 110 boys (26.4%) had LI (including mixed language and speech impairment), 20 (18.2%) had speech impairment only, and 61 (55.5%) were controls. Of the 58 girls, 20 (34.5%) had LI, 7 (12.1%) had speech impairment only, and 31 (53.5%) were controls.

Attrition

Although there has been no differential attrition over the years as a function of gender, follow-up participants differed from nonparticipants. Nonparticipants were more likely to be language or speech impaired and to come from low socioeconomic status, single parent families, and they had lower nonverbal IQ scores than follow-up participants. Thus, our sample may have been higher functioning than would have been the case with full, continued participation from wave 1.

Measures of Language and Speech at Age 5

Test of Language Development (TOLD: Newcomer & Hammill, 1977). The TOLD is a battery of seven subtests measuring expressive and receptive dimensions of semanties, syntax, and phonology, designed for children aged 4:0-8:11. In addition to subtest scores, the test provides a Spoken Language Quotient standard score, which is a general index of language development based on the five language subtests: Grammatic Completion, Grammatic Understanding, Oral Vocabulary, Picture Vocabulary, and Sentence Imitation. The test has two additional subtests measuring aspects of speech development: Word Discrimination and Word Articulation. In the normative sample, the test correlated substantially with several other language tests and distinguished language impaired children from children with normally developing language. Five-day test-retest reliability coefficients for all subtests exceeded .80.

Peabody Picture Vocabulary Test–Revised (Dunn & Dunn, 1981). The Peabody Picture Vocabulary Test–Revised is a test of single word receptive vocabulary. Participants indicate which of four pictures correspond to each auditorily presented word. The correlation between alternate forms of the test administered to 5- and 6-year-olds within 14 days was .77. Split-half reliability coefficients for the 5- and 6-year age groups ranged from .73 to .84.

Goldman-Fristoe-Woodcock Auditory Memory Tests (Goldman, Fristoe, & Woodcock, 1974). The Goldman-Fristoe-Woodcock Auditory Memory Tests are among 12 tests in the Goldman-Fristoe-Woodcock Auditory Skills Battery. Two auditory memory tests were used to define language impairment: a test of short-term memory for content, and a test of short-term memory for sequence. Standardization of test administration is maximized by the use of audio-taped auditory stimuli, and training procedures that familiarize children with the test procedures and with the stimuli and vocabulary used in the test. Test performance distinguished children with speech and language problems from children with normally developing language. Internal consistency for the 12 tests ranged from .78 to .97 for the youngest (age 3 to 8) normative sample.

Checklists for Voice Disorders, Stuttering, and Dysarthria. Speech language pathologists assessed children for speech disorders using checklists for voice disorders, stuttering, and dysarthria (articulation disorder). The checklist for voice disorders, based on Wilson (1979), assessed children on ten aspects of voicing using a 7-point severity scale. The assessment was based on both spontaneous speech and number repetition. The stuttering checklist assessed seven levels of stuttering with clear inclusion criteria for each, including criteria to distinguish normal disfluency from incipient stuttering. Stuttering was assessed during spontaneous speech, and during letter, number, and word repetitions. Dysarthria was assessed using a checklist adapted from Darley, Aronson, and Brown (1975).

Definition of Language and Speech Impairment at Age 5

During the initial 1982 study, children were classified as language or speech impaired if they met criteria for LI, speech impairment, or both language and speech impairment. LI, including receptive and/or expressive impairment, was defined as one or more of the following: (a) 1 SD below the mean on the TOLD Spoken Language Quotient; (b) 2 SD below the mean on any TOLD language subtest; (c) 1 SD below the mean on the Peabody Picture Vocabulary Test-Revised; (d) 1 SD below the mean on both content and sequence subtests of the Goldman-Fristoe-Woodcock Auditory Memory Tests. Speech impairment was defined as one or more of the following: (a) 2 SD below the mean on the Test of Language Development Word Articulation subtest; (b) 2 SD below the mean on the TOLD Word Discrimination subtest: (c) evidence of a voice disorder, stuttering, or dysarthria according to the Speech/Language Pathologist conducting the assessment. The 1 SD cutoff defining language and speech impairment is relatively liberal; however, this cutoff has been shown to correspond to the clinical judgments of speech/language pathologists (Aram, Morris, & Hall, 1993; Johnson et al., 1999; Records & Tomblin, 1994).

Measures of Delinquency, Aggression, and Conduct Problems–Ages 5, 12, and 19

Child Behavior Checklist (CBCL; Achenbach, 1991a)–Age 19, Problem scales of the CBCL were used to assess parent-rated delinquency and aggression symptoms. The CBCL is a standardized instrument that has been used to differentiate children referred for psychiatric treatment from nonclinical children. The Delinquency scale consists of 13 items including “lying or cheating” and “steals at home.” The Aggressive Behavior scale consists of 20 items including “physically attacks people” and “gets into many fights.” The 2-week test-retest correlation for behavior problem scores was .93, and the correlation between mothers’ and fathers’ scores was .76. The CBCL has been normed for use with children and adolescents aged 4 through 18 years of age. Raw scores were used in analyses to avoid problems scoring participants older than available norms and to facilitate comparison of girls’ and boys’ scores. The raw scores were square root transformed in order to better approximate normality.

Youth Self-Report (YSR; Achenbach, 1991b)–Age 19. Problem scales of the YSR were used to measure age 19 self-reported delinquency and aggression. The YSR is a parallel measure to the CBCL. The YSR Delinquency and Aggressive Behavior scales are almost identical to the corresponding CBCL scales. The 2-week test-retest correlation among 15-18-year-olds is .91 for total problems. Raw scores were used in analyses to facilitate comparison of girls’ and boys’ scores and to avoid problems scoring participants older than available norms.

Conners Teacher Rating Scale (CTRS; Conners, 1969)–Age 5 and Age 12

Participants’ teachers completed this 35-item behavior problems scale to assess conduct problems at waves 1 and 2. Conduct problems were measured using the conduct factor subscale derived by Trites, Blouin, and Laprade (1982), on the basis of a random sample of 9583 school children in the Ottawa-Carlton region of Ontario, Canada. The conduct subscale consists of 7 items, including “steals” and “defiant.” Cronbach’s alpha for the conduct factor was .93. The coefficient of congruence for the conduct factor, based on factors computed on random half samples, was .87. The age 5 and age 12 conduct factor scores were square root transformed in order to better approximate normality.

Participant Interviews–Age 19. Trained interviewers administered a structured participant interview that included questions about participants’ arrests, convictions, and incarcerations. In four cases, parents’ reports were substituted because participants did not complete the interview.

Measures of Academic Performance, Reading Achievement and Verbal IQ–Age 5 and Age 12

Kaufman Test of Educational Achievement (Kaufman & Kaufman, 1985)–Age 12. Reading ability was measured using the reading composite standard scores of the Kaufman Test of Educational Achievement. The correlation between parallel forms of the reading composite for 12- and 13-year-olds in the normative sample was .97, and 1-week test-retest correlation was .94. The reading composite was correlated with other individually administered achievement test reading subscales, with correlations ranging from .75 to .86.

Teacher Report Form (Achenbach & Edelbrock, 1986)–Age 12. We measured academic functioning with the Academic Performance scale of the Teacher Report Form. The scale comprised teachers’ ratings of students’ performance in six academic subjects, relative to grade level. Raw scores rather than standard scores were used to allow direct comparison of girls and boys.

Wechsler Preschool and Primary Scale of Intelligence (Wechsler, 1967)–Age 5. Verbal IQ was measured using the Wechsler Preschool and Primary Scale of Intelligence. This test has excellent reliability and validity.

Demographic and Family Assessments at Age 5

Brief Symptom Inventory (Derogatis & Spencer, 1982). The Brief Symptom Inventory, a shortened version of the SCL-90, is a self-report measure of psychological symptoms and distress. Mothers rated the 53 symptoms on the inventory on a 5-point scale from 0 “not at all,” to 4 “extremely,” according to the degree of distress they experienced in the last week associated with each symptom. Test-retest reliability of the Global Severity Index, a global measure of distress, was .90. Many more mothers than fathers completed the BSI. Thus, mothers’ reports were used in analyses where available. In six cases, only fathers’ reports were available and these were included in the analyses.

Parent Interviews. When the children were 5 years of age, parents were interviewed about the child’s developmental and medical history, as well as demographic and family characteristics. Family structure, family income, parent occupations, and parental history of arrests and convictions were among the information collects. Family income was reported on a 13-point scale.

Data Analysis

Hierarchical Linear Regression

In addition to zero order relation between age 5 language and speech status and age 19 antisocial outcomes, hierarchical linear regression assessed whether any such relations remained when other variables were considered. These variables included age 5 verbal IQ, and age 5 demographic and family variables. Family income was selected to measure socioeconomic status because it was the best predictor of age 19 delinquency and aggression, and therefore provided a more stringent test of the robustness of relations between language and speech impairment and antisocial outcomes. Parents of seven male participants declined to report family income. For those participants, income was estimated based on a regression of income on two alternate socioeconomic indicators: primary parent education and occupational socioeconomic status ratings based on a socioeconomic index for occupations in Canada (Blishen, Carroll, & Moore, 1987).

Structural Equation Modeling With Observed Variables

Structural equation modeling with observed variables was used to examine the interrelationships among key variables at three points in time. The Lisrel program (Joreskog & Sorbom, 2001) was used to conduct these analyses. The antecedent variables were age 5 LI and age 5 teacher-rated conduct problems. Age 5 LI was a dichotomous variable (language impaired versus nonlanguage impaired). The intervening variables were age 12 standardized reading achievement, age 12 teacher-rated academic performance, and age 12 teacher-rated conduct problems. The outcome variables were age 19 parent-rated delinquency symptoms and aggression symptoms. Intervening variables functioned as predictors and as dependent variables within different parts of the model.

We assumed that earlier variables could predict later variables, whereas the reverse was not possible. We assumed that the effects of LI temporally preceded reading achievement and academic performance, since language and speech impairments were identified when the children were in kindergarten, as they entered an academic environment. No direction of effect was specified for the two antecedent variables, LI and age 5 conduct problems, which were allowed to correlate. Similarly, delinquency and aggression were both outcome variables, and no directional pathway between these two variables was specified. Instead, in order to reflect the possibility that a third variable or variables caused both delinquency and aggression, the residuals of the two variables were allowed to correlate, which is statistically equivalent to a directional pathway (Lee & Hershberger, 1990). Finally, the three age 12 intervening variables (reading achievement, academic performance, and conduct problems) were measured at the same time and could mutually influence each other in a number of ways. No causal order was specified among these three variables. Instead, the residuals of the three age 12 variables were allowed to correlate.

For each model that was considered, all possible pathways (i.e., pathways that did not violate the temporal order of the variables) were included in a preliminary, saturated model. The saturated model was then modified by removing selected pathways according the following criteria: (a) if the pathway did not differ significantly from 0 ([alpha] = .05), (b) if the resulting model fit was adequate; (c) if the removal of the pathway did not substantially reduce model fit, and (d) if the exclusion of the pathway was theoretically justifiable (Bollen, 1989). A modified model was selected that showed good fit with the data, had zero and nonzero pathways that were theoretically plausible, and was as parsimonious as possible. In addition, we considered theoretically plausible nonrecursive models, in which bidirectional effects could be included.

Model Fit

Model fit was assessed using root mean square error of approximation (RMSEA). This measure represents the expected discrepancy, per degree of freedom, between the observed covariance matrix and the population matrix, assuming it were available and that parameters were optimally selected. Unlike [chi square] measures, RMSEA does not assume that the model holds exactly in the population. Values less than .05 indicate good fit, and values above .10 represent poor fit (Byrne, 1998). We considered models fitting adequately if the 90% confidence interval for RMSEA did not contain values at or above .10.

RESULTS

Antisocial Outcomes of Speech, Language, and Control Cohorts

Table 1 shows age 19 parent-reported and self-reported delinquency and aggression raw scores, separately by gender and age 5 speech and language functioning. We compared participants with age 5 LI, including mixed language and speech impairment, to participants with speech impairments but normal language (pure speech impairment) and controls. A two-way ANOVA of (square root transformed) parent-rated delinquency scores showed a significant interaction between gender and group (language impaired, pure speech, and controls) [F(2, 162) = 5.101, p = .004]. Whereas mean parent-rated delinquency scores were higher among language impaired boys than boys with normal language (d = .86), delinquency scores of language impaired and nonlanguage impaired girls were similar (d = -.08). Because the pattern of relationships between language and speech impairment and antisocial outcomes looked quite different for the boys as compared to the girls, we could not justify combining them into a single sample; therefore we consider the boys and the girls separately.

Boys with age 5 LI scored higher on age 19 parent-rated delinquency than boys with pure speech impairment and controls. Language impaired boys also scored higher on parent-rated aggression than the boys with pure speech impairment. Since the pure speech impaired boys were not different from the controls in terms of their age 19 outcomes, speech impairment was omitted from further analyses for the boys. Subsequent analyses of LI compared language impaired boys to boys without LI, including both controls and boys with pure speech impairment.

In contrast to parent ratings, boys’ YSR self-reported delinquency and aggression symptoms were not related to language or speech impairment. However, self-reported arrests were more prevalent among language impaired boys (41.5%) than controls (20.2%), [chi square](1, N = 142) = 6.88, p = .009. Similarly, language impaired boys were more likely to report having been convicted for a crime (28.9%) than controls (14.5%), [chi square](1, N = 142) = 4.60, p = .032.

Among the girls, the speech/language groups did not differ in parent-reported or self-reported delinquent or aggressive behavior ratings. In the remaining analyses, we examine the relation between age 5 LI and age 19 parent-reported delinquency and aggression for the boys only.

Hierarchical Regression

Hierarchical regression analyses predicting boys’ age 19 parent-rated delinquency and aggression were conducted in order to consider the effects of potentially confounding variables. Results of these analyses are summarized in Table II. In the first step, LI (language impaired vs. nonlanguage impaired) was entered. In the second step, verbal IQ was added. In the final step, four age 5 demographic and family variables were added: family income, parent ever convicted of a crime, parent distress, and single parent family structure. Although the regression coefficients for LI appeared attenuated with the addition of verbal IQ, verbal IQ did not account for the relation between LI and delinquency, F(1, 143) = 1.404, p = .238, or aggression, F(1, 143) = .703, p = .403. Further, the relation was not accounted for by age 5 demographic or family variables. However, age 5 parent distress did contribute unique variance to age 19 delinquency and aggression ratings.

Structural Equation Modeling With Observed Variables Predicting Boys’ Parent-Rated Delinquency and Aggression Symptoms

We conducted structural equation modeling with observed variables in order to examine the relation between early LI and parent-rated delinquency and aggression symptoms. The correlation matrix for variables included in the path models for the boys who had complete data on all path variables is shown in Table III. For purposes of comparison, correlations for the girls who had complete data are also shown, with each correlation coefficient shown immediately below the corresponding correlation coefficient for the boys. The saturated structural equation model, which includes all possible pathways among the variables (i.e., pathways not violating the temporal order of the variables), is shown in the appendix.

Figure 1 shows the modified model, in which nonsignificant pathways have been removed. In this model, age 5 LI directly predicted age 19 delinquency. Age 5 LI indirectly predicted aggression, through age 12 academic performance. The removal of the direct pathway between LI and delinquency resulted in substantially poorer fit (RMSEA = .08, 90% confidence interval = 0.15; [chi square] (1) for change in model fit = 8.86, p < .01.) Improvement indices (changes to the model [X.sup.2] that would result if omitted pathways were included in the model) are shown in Table IV. None approached the value of [chi square](1) = 3.84, p = .05. Thus, in no case did the exclusion of a pathway from the modified model result in significantly poorer fit than if the pathway were included.

Alternate Models

We also considered nonrecursive models with bidirectional causal pathways among variables. However, there was no nonrecursive model that had both adequate fit and significant pathways among all variables.

Model Fit

Model fit indices were consistent with an adequate to good fit between the model and the data (RMSEA = 0.00; 90% confidence interval = 0, .095).

DISCUSSION

This study demonstrated a direct effect of childhood LI on late adolescent delinquency symptoms in a longitudinal community sample. Boys diagnosed with LI at age 5 scored higher than controls on age 19 parent-rated delinquent behavior. This effect was distinct from verbal IQ; the association between LI and delinquency symptoms remained even when controlled for verbal IQ, and for demographic and family variables. An association between verbal skills and antisocial behavior is not a new finding. However, although verbal IQ and language ability are closely related concepts, this study suggests a specific effect of LI.

In addition, the use of structural equation modeling allowed us to investigate models of the interrelationships of behavioral, language, and academic performance at ages 5, 12, and 19. The data are correlational, showing predictive rather than causal relationships. Nevertheless, the combinations of direct and indirect pathways allowed us to address multiple explanations for this relationship. First, the apparent relationships between delinquency and LI could be spurious, appearing only because of a correlation between LI and early conduct problems. This hypothesis was not supported for delinquency, as LI predicted delinquency symptoms over and above early conduct problems.

Second, the association may be due to the negative impact of LI on school functioning. The timing of the initial study allowed us to assess the impact of LI on children’s schooling, as children participated in the study at age 5 before they had academic experience beyond kindergarten. Consistent with the findings of Fergusson and Lynskey (1997) that poor reading is not associated with delinquency once confounding variables are controlled, reading performance did not account for the relationship between boys’ LI and delinquency symptoms. Further, the effects of reading on later antisocial outcomes were mediated by school performance. The importance of school performance, rather than measured literacy attainment, suggests that negative school experiences may play an important role in the development of boys’ antisocial behavior. Boys with LI are at risk for lack of success at school. However, the direction of effect between age 12 conduct and age 12 academic performance cannot be determined with correlational data.

It is possible that another antecedent variable not addressed in the current study accounted for both LI and delinquency. One candidate antecedent variable is neurodevelopmental abnormality or immaturity (Beitchman, 1985). Research by Pine and colleagues (1997) suggested that left brain abnormalities may underlie both LI and behavior problems in boys. However, in a prospective study with multiple assessments of neuropsychological indicators from birth, environmental factors rather than neurodevelopmental factors predicted adolescent antisocial behavior (Aguilar, Sroufe, Egeland, & Carlson, 2000). Another possible link between LI and delinquency is through executive function. Language and executive functioning are difficult to untangle because language is a key component of self-regulation (Berk, 1992). In particular, the development of internal, self-directed speech facilitates the rehearsal of rules, the ability to consider, and to modify ongoing behavior with respect to its consequences, and the ability to form appropriate plans for future action.

Finally, the cumulative effects of compromised communication skills may account for the increased levels of delinquency in language impaired boys. Language is central to navigating social life. Thus, language skills may affect children’s experiences in multiple social contexts. At home, the comprehension and expression difficulties experienced by language impaired children may be frustrating for parents, creating an ongoing challenge to parent/child relationships, and interfering with positive socialization (Patterson, 1982). At school, language impaired children are often rejected by their peers (Gertner, Rice, & Hadley, 1994) and viewed negatively by their teachers (Redmond & Rice, 1998).

Further, LI tends to persist, as do its social consequences. In one longitudinal study, language impaired boys continued to experience social problems in various domains at age 32 (Rutter & Mawhood, 1991). In the present study, over 70% of children with age 5 LI continued to have LI at age 19 (Johnson et al., 1999). Redmond and Rice (1998) argued that the behavior problems of language impaired children reflect their adaptations to situations in which the communication demands exceed their linguistic resources. Within their peer groups, delinquent behavior may represent a specific adaptation for language impaired boys. Delinquent behavior may provide an alternate means of obtaining social status among peers for boys who have difficulty using social speech to further their social goals, including peer hierarchy negotiation (Maccoby, 1986; Zoccolillo, 1993).

[FIGURE 1 OMITTED]

LI may also increase risk for delinquency by interfering with the ability to understand others’ perspectives, affecting both social competence and moral development. Language ability is associated with the development of theory of mind understanding, which requires the ability to take the perspective of others (Astington & Jenkins, 1999). Cohen, Menna, and colleagues (1998) demonstrated that children with LI had deficits in social cognitive skills, including interpreting the feelings of others and social problem solving. In addition to the social consequences of poor social cognition, compromised verbal reasoning skills, and limitations in perspective taking may also interfere with moral development. Young adults with underdeveloped morality may be particularly likely to engage in the prevarication, concealment, and dishonesty that characterize nonviolent delinquent behavior.

In contrast to delinquency, LI did not directly predict aggression. The relationship between LI and aggression was indirect, mediated by reading and academic performance. This is consistent with the early onset and stability of aggressive behavior (Olweus, 1979), since LI is not particularly associated with high levels of aggression in early childhood.

The dependent variables in these analyses, CBCL parent ratings of delinquency and aggression symptoms, are dimensional measures. Further, the study participants were part of a community sample rather than a referred or adjudicated sample. The combination of a community sample and a dimensional approach resulted in a dependent variable that represented less severe antisocial behavior than in many other studies of antisocial behavior. Although some participants met criteria for antisocial personality disorder or were convicted of crimes, including violent crimes, most demonstrated relatively low levels of antisocial behavior. The CBCL Delinquency scale items did not necessarily consist of illegal acts, but included other forms of antisocial behavior, just as the Aggressive Behavior Scale did not necessarily constitute physical aggression, but also included verbal aggression and destruction of property. The scales were empirically derived syndromes, comprised clusters of symptoms that tended to co-occur among subgroups of clinically referred children (Achenbach, 1991a). Thus, the scales were labeled post hoc to describe their contents: they were not designed to measure the predefined constructs of delinquency and aggression.

Nevertheless, the dimensional dependent variables reflected a range of antisocial behavior from mild to severe that might be found in the community, and evidence supports their construct validity. First, the Delinquency and Aggressive Behavior scales differentiate clinically referred children from controls (Achenbach, 1991a). Second, both scales have been shown to be associated with external validators, such as use of mental health services, problems at school, and self-reported symptoms (Gould, Bird, & Staghezza Jaramillo, 1993; Jensen et al., 1996). Third, both scales are associated with the conceptually linked psychiatric diagnosis of conduct disorder, (Biederman, Mick, Faraone, & Burback, 2001). In addition, scores on the CBCL Delinquency and Aggressive Behavior scales have distinguished persistent from short-term conduct disorder (Biederman et al., 2001).

The use of parent ratings in the dependent variable avoided shared method variance with childhood conduct problems, which were rated by teachers. Some evidence suggests that parent ratings are particularly appropriate for measuring externalizing behavior rather than internalizing behavior (Jensen, Salzberg, Richters, & Watanabe, 1993). However, self-report is often preferred to index young adults’ antisocial behavior (Verhulst & Van der Ende, 1992). In the present study, response patterns on the Achenbach Youth Self Report for the language impaired, speech impaired, and control groups failed to correspond to parent reports. Results of the present study were partially validated using participants’ self-reports of their delinquent behavior. The higher proportion of self-reported arrests and convictions among the language impaired boys than the control boys supports the validity of CBCL parent reports. It is possible that the arrests and convictions reported may reflect a greater likelihood of language impaired youth being apprehended or convicted if they do commit delinquent acts. However, the higher prevalence of antisocial personality disorder in the language impaired cohort (Beitchman et al., 2001) lends further support to the validity of parent reports, which endorsed more antisocial behavior among the language impaired boys than controls. Further analysis of interactions among raters would be of interest.

[ILLUSTRATION OMITTED]

Conclusions

This is the first follow-up of a representative community sample of S/L impaired children that has reported on adult antisocial behavior. The link between boys’ LI and delinquency identified in this study extends this finding from clinic-referred and incarcerated samples to a community sample, albeit with milder forms of antisocial behavior. In contrast to LI, speech impairment was not associated with antisocial outcomes.

The structural equation model allowed us to evaluate models of the relation among variables at ages 5, 12, and 19. The sample size of the present study limited statistical power and precluded structural equation modeling for the girls. For the boys, confidence intervals for the degree of model fit indices showed the degree of accuracy associated with parameter estimates for a sample of this size. Using the limited available data on the girls, it appears that the correlation between LI and delinquency was specific to boys. However, further analyses with larger samples are needed. In addition, selective attrition resulted in a healthier sample than would have been the case if all the children in the wave 1 study had participated fully in the wave 2 and wave 3 studies. Given the pattern of differences between continuing participants and participants with incomplete data, it is likely that the effects of attrition underestimated the impact of LI. However, counter evidence was not available. Third, generalizability is limited by the fact that the sample was comprised of predominantly White participants, because of the demographics of the region sampled at the first phase of the study. Fourth, although the participants were drawn from a one-in-three random sample, all available participants with language and/or speech impairments were included in the follow-up study, whereas children with normal speech and language skills were less likely to be included. Thus, the sample is not random, but overrepresents speech and/or language impaired individuals. Therefore, estimates of population parameters must be interpreted with caution. Finally, although the study provided detailed assessments at three points in time, the interval between assessments was wide (6-7 years). The longitudinal design over a long time-period allowed predictions from childhood to late adolescence to be tested. However, we were limited by the particular ages at which data were collected.

Despite the much replicated finding that compromised verbal skills are related to antisocial outcomes, high proportions of youth with LI have been located in treatment facilities for youth with conduct problems (e.g., Giddan et al., 1996; Warr-Leaper et al., 1994). Although some authors (e.g., Moffitt, 1993) have speculated on the specific contribution that verbal difficulties may make in the development of delinquency, assessment for LI is not often incorporated in clinical assessments (Cohen, Menna et al., 1998). LI is particularly likely to be missed when the referral is for highly salient externalizing problems (Cohen, Davine, Horodezky, Lipsett, & Isaacson, 1993). Perhaps one reason that this association has not translated into automatic language assessment and treatment is that the construct of verbal IQ implies a stable level of ability. In contrast, a diagnosis of LI suggests specific linguistic processing problems that may be amenable to treatment. The association of LI in boys with antisocial outcomes both within referred/incarcerated samples and within a community sample underscores the need for language assessment and intervention with emerging antisocial behavior, as well as with attention-deficit/hyperactivity disorder and internalizing problems.

Table I. Age 19 Delinquency and Aggression by Gender and Speech/Language

Status

Delinquency Aggression

Rater and Group N M (SD) M (SD)

Boys

Parent (a)

Control 81 1.85 (2.59) 4.44 (5.61)

Speech only impaired 27 2.04 (3.63) 3.63 (6.86)

Language impaired 44 4.07 (4.08) 8.20 (7.97)

Self

Control 85 4.45 (2.91) 8.93 (5.29)

Speech only impaired 28 4.57 (2.91) 8.18 (5.13)

Language impaired 48 5.00 (3.36) 8.77 (5.71)

Girls

Parent

Control 42 1.14 (1.32) 3.69 (4.06)

Speech only impaired 8 3.75 (3.81) 8.88 (7.47)

Language impaired 27 1.96 (2.71) 5.63 (6.55)

Self

Control 49 3.78 (2.37) 8.10 (4.45)

Speech only impaired 10 3.50 (3.63) 7.90 (4.55)

Language impaired 34 2.76 (1.60) 8.12 (3.26)

Note. Multiple comparison statistics shown refer to square root

transformed scores. Non-transformed raw scores are shown in this table

to increase interpretability.

(a) Tukey’s HSD, p SO, C; Aggression: LI > SO.

Table II. Summary of Hierarchical Regression Analysis Predicting Boys’

Delinquency and Aggression (N = 146)

Variable B SE B [beta] p

Dependent Variable:

Parent-Rated

Delinquency

Step 1

Language

impairment .729 .190 .305 .000

Step 2

Language

impairment .540 .247 .226 .020

Verbal IQ -.009 .007 -.122 .237

Step 3

Language

impairment .573 .244 .240 .020

Verbal IQ -.005 .008 -.066 .537

Income -.044 .030 -.131 .151

Parent conviction .278 .256 .091 .274

Parent distress

(BSI) .010 .009 .166 .042

Single parent

Household -.162 .277 -.050 .580

Dependent Variable:

Parent-Rated

Aggression

Step 1

Language

impairment .851 .261 .262 .001

Step 2

Language

impairment .667 .341 .205 .053

Verbal IQ -.004 .010 -.039 .721

Step 3

Language

impairment .667 3.41 .221 .034

Verbal IQ -.008 .010 -.039 .721

Income -.057 .042 -.125 .177

Parent

conviction .152 .354 .037 .668

Parent distress

(BSI) .031 .013 .203 .015

Single parent -.284 .382 -.067 .458

Variable Adjusted [DELTA][R.sup.2] p([DELTA][R.sup.2])

Dependent Variable:

Parent-Rated

Delinquency

Step 1

Language

impairment

.087 .093 .001

Step 2

Language

impairment

Verbal IQ

.089 .009 .238

Step 3

Language

impairment

Verbal IQ

Income

Parent conviction

Parent distress

(BSI)

Single parent

Household

.127 .061 .042

Dependent Variable:

Parent-Rated

Aggression

Step 1

Language

impairment

.062 .069 .001

Step 2

Language

impairment

Verbal IQ

.060 .005 .403

Step 3

Language

impairment

Verbal IQ

Income

Parent

conviction

Parent distress

(BSI)

Single parent

.101 .065 .038

Note. Unadjusted [R.sup.2] for Delinquency: Step 1: [R.sup.2] = .093,

Step 2: [R.sup.2] = .102, Step 3: [R.sup.2] = 163. Unadjusted [R.sup.2]

for Aggression: Step 1: [R.sup.2] = .069, Step 2: [R.sup.2] = .073, Step

3: [R.sup.2] = .138.

Table III. Correlations Between Model Variables for Boys and Girls

Reading Academic Conduct Conduct

Variable (age) LI (age 5) (age 12) (age 12) (age 5) (age 12)

Reading (age 12) -.63***

-.53***

Academic (age 12) -.42*** .60***

-.42*** .48***

Conduct (age 5) .27** -.25** -.34***

-.11 .15 -.02

Conduct (age 12) .17 -.19* -.47*** .30**

-.06 -.13 -.13 .21

Delinquency (age 19) .36*** -.29** -.34*** .33*** .39***

-.04 .15 -.24 .16 .17

Aggression (age 19) .26** -.25** -.36** .35*** .34***

-.08 .14 -.19 .28* .03

Delinquency

Variable (age) (age 19)

Reading (age 12)

Academic (age 12)

Conduct (age 5)

Conduct (age 12)

Delinquency (age 19)

Aggression (age 19) .77***

.61***

Note. Within each cell, correlations for boys (N = 110) are shown

directly above correlations for girls (N = 58). LI = language

impairment; Academic = academic performance; Conduct = teacher-rated

conduct problems; Delinquency = parent-rated delinquency; Aggression =

parent-rated aggression.

*p < .05, **p < .01, ***p < .001.

Table IV. Improvement Indices for Zero paths

Pathway Improvement

Index

Language (age 5) [right arrow] Conduct (age 12) 0.901

Language (age 5) [right arrow] Aggression (age 19) 1.576

Conduct (age 5) [right arrow] Reading (age 12) 1.391

Reading (age 12) [right arrow] Conduct (age 12) 1.491

Reading (age 12) [right arrow] Delinquency (age 19) 0.419

Reading (age 12) [right arrow] Aggression (age 19) 0.047

Academic (age 12) [right arrow] Reading (age 12) 1.697

Academic (age 12) [right arrow] Conduct (age 12) 1.508

Academic (age 12) [right arrow] Delinquency (age 19) 0.554

Conduct (age 12) [right arrow] Reading (age 12) 1.252

Aggression (age 19) [right arrow] Delinquency (age 19) 0.554

Delinquency (age 19) [right arrow] Aggression (age 19) 1.576

ACKNOWLEDGMENTS

This research was supported by Grant #6606-5639-102, awarded to Joseph H. Beitchman by the National Health Research and Development Program. E. B. Brownlie was supported by the Social Sciences and Humanities Research Council. We thank our participants and their parents for their willing participation. We also extend our thanks to the Ottawa research team: Brenda Cavanaugh, Hope Harris, Karen Ivings, and Lisa McAvoy.

Received March 7, 2003; revision received February 20, 2004; accepted February 21, 2004

REFERENCES

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

Achenbach, T. M. (1991b). Manual for the Youth Self-Report and 1991 Profile. Burlington, VT: University of Vermont Department of Psychiatry.

Achenbach, T. M., & Edelbrock, C. S. (1986). Manual for the Teacher’s Report Form and Teacher Version of the Child Behavior Profile, Burlington, VT: University of Vermont Department of Psychiatry.

Aguilar, B., Sroufe, L. A., Egeland, B., & Carlson, E. (2000). Distinguishing the early-onset/persistent and adolescence-onset antisocial behavior types: From birth to 16 years. Development and Psychopathology, 12, 109-132.

Aram, D. M., Morris, R., & Hall, N. E. (1993). Clinical and research congruence in identifying children with specific language impairment. Journal of Speech and Hearing Research, 36, 580-591.

Astington, J. W., & Jenkins, J. M. (1999). A longitudinal study of the relation between language and theory-of-mind development. Developmental Psychology, 35, 1311-1320.

Baker, L., & Cantwell, D. P. (1987). A prospective psychiatric follow-up of children with speech and language disorders. Journal of the American Academy of Child and Adolescent Psychiatry, 26, 546-553.

Beitchman, J. H. (1985). Speech and language impairment and psychiatric risk: Toward a model of neurodevelopmental immaturity. Psychiatric Clinics of North America, 8, 721-735.

Beitchman, J. H., Brownlie, E. B., Inglis, A., Wild, J., Ferguson, B., & Schachter, D. (1996). Seven-year follow-up of speech/language impaired and control children: Psychiatric outcome. Journal of Child Psychology and Psychiatry, 37, 961-970.

Beitchman, J. H., Brownlie, E. B., Inglis, A., Wild, J., Mathews, R., & Schachter, D., et al. (1994). Seven year follow-up of speech/language impaired and control children: Speech/language stability and outcome. Journal of the American Academy of Child and Adolescent Psychiatry, 33, 1322-1330.

Beitchman, J. H., Nair, R., Clegg, M., & Patel, P. G. (1986). Prevalence of speech and language disorders in 5-year-old kindergarten children in the Ottawa Carlton region. Journal of Speech and Hearing Disorders, 51, 98-110.

Beitchman, J. H., Wilson, B., Brownlie, E. B., Walters, H., Inglis, A., & Lancee, W. (1996). Long-term consistency in speech/language profiles: II. Behavioral, emotional, and social outcomes. Journal of the American Academy of Child and Adolescent Psychiatry, 35, 815-825.

Beitchman, J. H., Wilson, B., Brownlie, E. B., Walters, H., & Lancee, W. (1996). Long-term consistency in speech/language profiles: I. Developmental and academic outcomes. Journal of the American Academy of Child and Adolescent Psychiatry, 35, 804-814.

Beitchman, J. H., Wilson, B., Johnson, C., Atkinson, L., Young, A., & Adlaf, E., et al. (2001). Fourteen-year follow-up study of speech/language impaired and control children: Psychiatric outcome. Journal of the American Academy of Child and Adolescent Psychiatry, 40, 75-82.

Berk, L. E. (1992). Children’s private speech. In L. E. Berk & R. M. Diaz (Eds.), Private speech: From social interaction to self-regulation (pp. 17-53). Hillsdale, NJ: Erlbaum.

Biederman, J., Mick, E., Faraone, S. V., & Burback, M. (2001). Patterns of remission and symptom decline in conduct disorder: A four-year prospective study of an ADHD sample. Journal of the American Academy of Child and Adolescent Psychiatry, 40, 290-298.

Blishen, B. R., Carroll, W. K., & Moore, C. (1987). The 1981 socioeconomic index for occupations in Canada. Canadian Review of Sociology and Anthropology, 24, 465-488.

Bollen, K. A. (1989). Structural equation modelling with latent variables. New York: Wiley.

Byrne, B. M. (1998). Structural equation modeling with LISREL. PRELIS, and SIMPLIS: Basic concepts, applications, and programming. Mahwah, NJ: Erlbaum.

Cohen, N. J. (2001). Language and psychopathology in infants, children, and adolescents. Thousand Oaks, CA: Sage.

Cohen, N. J., Barwick, M. A., Horodezky, N. B., Vallance, D. D., & Im, N. (1998). Language, achievement, and cognitive processing in psychiatrically disturbed children with previously unidentified and unsuspected language impairments. Journal of Child Psychology and Psychiatry, 39, 865-877.

Cohen, N. J., Davine, M., Horodezky, N. B., Lipsett, L., & Isaacson, L. (1993). Unsuspected language impairment in psychiatrically disturbed children: Prevalence and language and behavioral characteristics. Journal of the American Academy of Child and Adolescent Psychiatry, 32, 595-603.

Cohen, N. J., Menna, R., Vallance, D. D., Barwick, M., Im, N., & Horodezky, N. B. (1998). Language, social cognitive processing, and behavioral characteristics of psychiatrically disturbed children with previously identified and unsuspected language impairments. Journal of Child Psychology and Psychiatry, 39, 853-864.

Conners, C. K. (1969). A teacher rating scale for use in drug studies with children. American Journal of Psychiatry, 126, 884-938.

Darley, F. L., Aronson, A. E., & Brown, J. R. (1975). Motor speech disorders. Philadelphia: W.B. Saunders.

Davis, A. D., Sanger, D. D., & Morris-Friehe, M. (1991). Language skills of delinquent and non-delinquent males. Journal of Communication Disorders, 24(4), 251-266.

Derogatis, L., & Spencer, P. (1982). The Brief Symptom Inventory: Administration, scoring and procedure manual–I. Riderwood, MD: Clinical Psychometric Research.

Dunn, L. M., & Dunn, L. M. (1981). Peabody Picture Vocabulary Test–Revised. Nashville, TN: American Guidance Centre.

Farrington, D. P. (1995). The twelfth Jack Tizard memorial lecture: The development of offending and antisocial behavior from childhood: Key findings from the Cambridge Study in Delinquent Development. Journal of Child Psychology and Psychiatry, 36, 929-964.

Fergusson, D. M., & Lynskey, M. T. (1997). Early reading difficulties and later conduct problems. Journal of Child Psychology and Psychiatry, 38, 899-907.

Gertner, B. L., Rice, M. L., & Hadley, P. A. (1994). Influences of communication competence on peer preferences in a preschool classroom. Journal of Speech, Language, and Hearing Research, 37, 913-923.

Giddan, J. J., Milling, L., & Campbell, N. B. (1996). Unrecognized language and speech deficits in preadolescent psychiatric inpatients. American Journal of Orthopsychiatry, 66. 85-92.

Goldman, R., Fristoe, M., & Woodcock. R. W. (1974). Goldman-Fristoe-Woodcock Auditory Memory Tests. Circle Pines, MN: American Guidance Services Inc.

Gould, M. S., Bird, H., & Staghezza Jaramillo, B. (1993). Correspondence between statistically derived behavior problem syndromes and child psychiatric diagnoses in a community sample. Journal of Abnormal Child Psychology, 21, 287-313.

Hinshaw, S. P. (1992). Externalizing behavior problems and academic underachievement in childhood and adolescence: Causal relationships and underlying mechanisms. Psychological Bulletin, 111, 127-155.

Jensen, P. S., Salzberg, A. D., Richters, J. E., & Watanabe, H. K. (1993). Scales, diagnoses, and child psychopathology; I. CBCL and DISC relationships. Journal of the American Academy of Child and Adolescent Psychiatry, 32, 397-406.

Jensen, P. S., Watanabe, H. K., Richters, J. E., Roper, M., Hibbs, E. D., & Salzberg, A. D., et al. (1996). Scales, diagnoses, and child psychopathology: II. Comparing the CBCL and the DISC against external validators. Journal of Abnormal Child Psychology, 24, 151-168.

Johnson, C., & Beitchman, J. H. (1999a). Mixed receptive-expressive language disorders. In B. J. Sadock & V. A. Sadock (Eds.), Kaplan and Sadock’s comprehensive textbook of Psychiatry (7th ed., pp. 2639-2644). Baltimore, MD: Lippincott Williams & Wilkins.

Johnson, C., & Beitchman, J. H. (1999b). Phonological disorders. In B. J. Sadock & V. A. Sadock (Eds.), Kaplan and Sadock’s Comprehensive Textbook of psychiatry (7th ed., pp. 2945-2650). Baltimore, MD: Lippincott Williams & Wilkins.

Johnson, C., Beitchman, J. H., Young, A., Escobar, M., Atkinson, M., & Wilson, B., et al. (1999). Fourteen year follow-up of children with and without speech/language impairments: Speech/language stability and outcomes. Journal of Speech, Language, and Hearing Research, 42, 744-760.

Joreskog, K., & Sorbom, D. (2001). Lisrel 8.51. [Computer software]. Scientific Software International Inc.

Kaufman, A. S., & Kaufman, N. L. (1985). Kaufman Test of Educational Achievement. Circle Pines, MN: American Guidance Services.

Kimball, M. (1989). A new perspective on women’s math achievement. Psychological Bulletin, 105, 198-214.

Lee, S., & Hershberger, S. (1990). A simple rule for generating equivalent models in covariance structure modeling. Multivariate Behavioral Research, 25, 313-334.

Lynam, D., Moffitt, T., & Stouthamer-Loeber, M. (1993). Explaining the relation between IQ and delinquency: Class, race, test motivation, school failure, or self-control? Journal of Abnormal Psychology, 102, 187-196.

Maccoby, E. E. (1986). Social groupings in childhood: Their relationship to prosocial and antisocial behavior in boys and girls. In D. Olweus, J. Block, & M. Radke-Yarrow (Eds.), Development of antisocial and prosocial behavior (pp. 343-357). Orlando, FL: Academic Press.

Maughan, B., Pickles, A., Hagell, A., Rutter, M., & Yule, W. (1996). Reading problems and antisocial behavior: Developmental trends in comorbidity. Journal of Child Psychology and Psychiatry, 37, 405-418.

Moffitt, T. E. (1993). The neuropsychology of conduct disorder. Development and Psychopathology, 5, 135-151.

Newcomer, P. L., & Hammill, D. D. (1977). The Test of Language Development. Austin, TX: Empire Press.

Newman, D. L., Moffitt, T. E., Caspi, A., Magdol, L., Silva, P. A., & Stanton, W. R. (1996). Psychiatric disorder in a birth cohort of young adults: Prevalence, comorbidity, clinical significance, and new case incidence from ages 11-21. Journal of Consulting & Clinical Psychology, 64, 552-562.

Olweus, D. (1979). Stability of aggressive reaction patterns in males: A review. Psychological Bulletin, 86, 852-875.

Patterson, G. R. (1982). Coercive family interactions. Eugene, OR: Castalia Press.

Pine, D. S., Bruder, G. E., Wasserman, G. A., Miller, L. S., Musabegovic, A. & Watson, J. B. (1997). Verbal dichotic listening in boys at risk for behavior disorders. Journal of the American Academy of Child and Adolescent Psychiatry, 36, 1465-1473.

Records, N., & Tomblin, J. B. (1994). Clinical decision making: Describing the decision rules of practicing speech-language pathologists. Journal of Speech and Hearing Research, 37, 144-156.

Redmond, S. M., & Rice, M. L. (1998). The socioemotional bahaviors of children with SLI: Social adaptation or social deviance? Journal of Speech, Language, and Hearing Research, 41, 688-700.

Rutter, M., & Mawhood, L. (1991). The long term sequelae of specific developmental disorders of speech and language. In M. Rutter & P. Casaer (Eds.), Biological risk factors in childhood and psychopathology (pp. 233-259). Cambridge: Cambridge University Press.

Schachter, D. C., (1996). Academic performance in children with speech and language impairment: A review of followup research. In J. H. Beitchman, N. J. Cohen, M. M. Konstantareas, & R. Tannock (Eds.), Language, learning, and behavior disorders (pp. 515-529). Cambridge: Cambridge University Press.

Schonfeld, I. S., Shaffer, D., O’Connor, P., & Portnoy, S. (1988). Conduct disorder and cognitive functioning: Testing three causal hypotheses. Child Development, 59, 993-1007.

Silva, P. A., Williams, S., & McGee, R. (1987). A longitudinal study of children with developmental language delay at age three: Later intelligence, reading and behavior problems. Developmental Medicine and Child Neurology, 29, 630-640.

Trites, R. L., Blouin, A. G. A., & Laprade, K. (1982). Factor analysis of the Conners Teacher Rating Scale based on a large normative sample. Journal of Consulting and Clinical Psychology, 50, 615-623.

Verhulst, F. C., & Van der Ende, J. (1992). Agreement between parents’ reports and adolescents’ self-reports of problem behavior. Journal of Child Psychology and Psychiatry, 33, 1011-1023.

Warr-Leaper, G., Wright, N. A., & Mack, A. (1994). Language disabilities of antisocial boys in residential treatment. Behavioral Disorders, 19, 159-169.

Wechsler, D. (1967). Manual for the Wechsler Preschool and Primary Scale of Intelligence. New York: The Psychological Corp.

Williams, S., & McGee, R. (1994). Reading attainment and juvenile delinquency. Journal of Child Psychology and Psychiatry, 35, 441-459.

Wilson, D. K. (1979). Voice problems of children (2nd ed.). Baltimore: William’s and Wilkins.

Zoccolillo, M., (1993). Gender and the development of conduct disorder. Development and Psychopathology, 5, 65-78.

E. B. Brownlie, (1,5) Joseph H. Beitchman, (2) Michael Escobar, (3) Arlene Young, (1,2) Leslie Atkinson, (2) Carla Johnson, (4) Beth Wilson, (2) and Lori Douglas (2)

(1) Psychology Department, Simon Fraser University, Burnaby, British Columbia, Canada.

(2) Child, Youth, and Family Program, Centre for Addiction and Mental Health, Toronto, Ontario, Canada.

(3) Department of Statistics, University of Toronto, Toronto, Ontario, Canada.

(4) Department of Speech-Language-Pathology, University of Toronto, Toronto, Ontario, Canada.

(5) Address all correspondence to E. B. Brownlie, Child, Youth and Family Program, Centre for Addiction and Mental Health: Clarke Site. 250 College Street, Toronto, Ontorio, Canada M5T 1R8.

COPYRIGHT 2004 Plenum Publishing Corporation

COPYRIGHT 2004 Gale Group