Clark and Watson's tripartite model of emotion (Clark & Watson, 1991) specifies a general factor, negative affect (NA), which represents a shared influence on anxiety and depression, and further specifies two specific factors, physiological hyperarousal (PH), common to anxiety, and (low) positive affect (PA), common to depression (Clark & Watson, 1991). Initially proposed to account for the high comorbidity of depressive and anxiety disorders and symptoms (Clark, 1989), the tripartite model has become increasingly well-specified and has found accumulating empirical support within adult and child samples (e.g., Brown, Chorpita, & Barlow, 1998; Chorpita, Albano, & Barlow, 1998; Joiner, Catanzaro, & Laurent, 1996; Lonigan, Hooe, David, & Kistner, 1999; Watson et al., 1995).
The model has a number of important implications with respect to psychopathology. First, evidence suggests that NA and PA are heritable, temporally stable risk factors (Clark, Watson, & Mineka, 1994; Lonigan & Phillips, 2001), such that these variables may ultimately shed light on the development of anxiety and depressive disorders. For example, NA and PA may be of considerable value in predicting the future emergence of affective pathology, and measuring such variables with precision precludes outlining the mechanisms by which these putative risks lead to dysfunction. Second, the model allows for the improved understanding of the comorbidity among affective disorders, offering a theoretical basis for the high co-occurrence of anxiety and depression in children as well as adults. Third, increased understanding of the core dimensions of affect and arousal related to anxiety and depressive disorders offers the possibility for an improved understanding of the nosology for classifying these syndromes. For example , knowledge about whether generalized anxiety disorder in children is characterized primarily by negative emotionality versus autonomic arousal has important implications for articulating the key features of that disorder within a diagnostic rule system. Finally, recent tests of the model in the context of dimensions of anxiety and depressive disorders have revealed the benefit of moving beyond a single-order assessment strategy to a hierarchical one. That is, it appears important to assess not only the symptoms of particular disorders or syndromes but also the core affective and arousal dimensions related to those syndromes. In this way, the degree to which the symptoms are related to general features of temperament or physiology can be determined, which has implications for course, chronicity, and emerging comorbidity. This is consistent with the notion of heterotypic continuity, whereby the symptomatic expression of an underlying, temporally stable pathology differs over the course of development (e.g., Ca spi, Elder, & Bem, 1988).
To address these important considerations, several recent investigations have begun to evaluate the tripartite model in child inpatient, outpatient, and school samples. For example, Lonigan, Carey, and Finch (1994) found in a sample of 233 clinically referred children that measures related to low PA discriminated children with depressive disorders from those with anxiety disorders. Joiner et al. (1996) conducted an exploratory factor analysis of self-report measures completed by 116 child and adolescent psychiatric inpatients and outlined three factors conceptually similar to those of the tripartite model. Chorpita et al. (1998) conducted a similar investigation with the use of confirmatory factor analysis of child and parent measures in 216 children and adolescents with anxiety and mood disorders and found corroborative support for a three-factor model of negative emotions. Recently, Lonigan et al. (1999) examined the relations of PA and NA measures with anxiety and depression measures in a school sample of 365 children and adolescents. Lonigan et al. (1999) found that NA and PA measures performed in a manner consistent with findings from adult samples, and that such findings were uniform across children and adolescents.
Although the tripartite model originally conceptualized anxiety as a unitary construct, recent investigations have begun to establish that the relation of tripartite dimensions to anxiety may not be uniform across specific anxiety syndromes or disorders (Mineka, Watson, & Clark, 1998). For example, Brown et al. (1998) evaluated the relations of tripartite dimensions to generalized anxiety, depression, social phobia, panic, and obsessions/compulsions in 350 adults with anxiety and mood disorders, and found that PH was significantly positively related to panic only. Further, in that same model, PH was significantly negatively related to measures of generalized anxiety.
Chorpita, Plummer, and Moffitt (2000) conducted an investigation of these issues in a sample of 100 children with anxiety and mood disorders. Children were between the ages of 7 and 17, and were diagnosed with structured clinical interviews. On the basis of the findings of Brown et al. (1998), two main structural models were tested. The first posited NA and PA as higher order factors explaining variance in the anxiety and depression dimensions, some of which in turn influenced PH. The second posited all three tripartite factors as higher order factors, explaining variance in the lower order anxiety and depression dimensions. In the first model, Chorpita, Plummer, et al. (2000) found that NA and PA were related to anxiety and depression in the expected manner, but that such relations were not uniform across specific anxiety syndromes. This was similar to the findings of Brown et al. (1998), in that NA was positively related to all anxiety and depression dimensions, and PH was not significantly related to depre ssion, social anxiety, generalized anxiety, or obsessions/compulsions. Thus, the prediction that tripartite dimensions were related to several of the mood and anxiety disorders in a child population was confirmed and extended an examination of related constructs in adults (Brown et al., 1998). Nevertheless, in terms of model evaluations, both the initial model and second model fit the data well. Because differences in fit of the models were relatively minimal, Chorpita, Plummer, et al. (2000) felt it was premature to conclude that one structure was preferable to another. Thus, some clarification is needed with respect to what is the most interpretable and best fitting model to explain the relation of affective dimensions to symptoms of anxiety and depression.
To examine these issues further, this study involved a larger investigation of the structural relations among tripartite factors and dimensions representing symptoms of selected anxiety disorders and depression. Also, because relatively little is known about these dimensions in the population in general, a large school sample of children and adolescents seemed like the most appropriate starting point for such an investigation. It was hypothesized that general aspects of the tripartite model would be supported (i.e., NA positively related with anxiety and depression, PA negatively related with depression). Further, it was hypothesized that the relations of tripartite factors would not be uniform across all anxiety dimensions. Specifically, PH would show a positive relation to panic and perhaps to separation anxiety (Last, 1991), but not to other anxiety dimensions. Finally, the robustness of such models was evaluated by examining the equivalence of structure and parameters across age and gender, and the intera ction effects of structural parameters with gender and grade level were examined, to determine whether certain parameters differed between boys and girls or across grade level. For example, given suggestions in the diagnostic nosology that worry in young children is characterized by autonomic arousal (e.g., "overanxious disorder," American Psychiatric Association [APA], 1987) more than worry in adults, which is characterized by a relative absence of autonomic arousal (cf. Brown et al., 1998), it was expected that such syndromes as generalized anxiety might show a higher relation to PH at earlier grade levels than that at later grade levels.
Participants were 1,578 children and adolescents recruited from 13 public and private schools in Grades 3 through 12 on O'ahu, HI, (median grade = 7) for a series of studies involving assessment of anxiety and emotional factors. The mean age of the sample was 12.87 years (SD = 2.82; range = 6.17-18.92)(3) and the group consisted of 893 girls 54.4%) and 748 boys (45.6%). Over 20 different ethnicities were identified, including Japanese American (n = 463; 28.2%), Filipino (n = 217; 13.2%), Hawaiian(n = 204; 12.4%), Chinese American (n = 138; 8.4%), Caucasian (n = 133; 8.1%), or multiethnic (n = 276; 16.8%). The remaining participants (n = 210; 12.8%) identified principally with one of the following: Korean, Okinawan, Portuguese, African American, Hispanic American, Samoan, Southeast Asian, Puerto Rican, Native American, Tongan, Fijian, or Guamanian. The group was broadly representative of the economic and geographic diversity of the local population. With the exception of 63 participants excluded because of a l istwise strategy for handling missing data, this sample was the same as the Study 1 sample reported on in Chorpita, Yim, Moffitt, Umemoto, and Francis (2000), described below.
Anxiety and Depression Factors
Children in this investigation were a subset of those participating in a larger series of studies designed to establish and evaluate new measures of tripartite factors and of DSM anxiety and depression (Chorpita, Daleiden, Moffitt, Yim, & Umemoto, 2000; Chorpita, Yim, et al., 2000). Briefly, 1,641 children were administered …