Differential prediction of internalizing and externalizing symptomatology

Using data from an eight-year longitudinal study of 214 children's social and emotional development, I conducted three studies to (1) examine patterns of agreement for internalizing (INT) and externalizing (EXT) symptomatology among different informants (mothers, fathers, teachers, and adolescents) using a recently developed structural equation modeling approach for multi-trait, multi-method data; (2) examine the developmental trajectories for INT and EXT and predict individual differences in symptom development using temperament and parenting variables; and (3) describe patterns of INT and EXT co-occurrence and predict these patterns from temperament and parenting. In Study 1, longitudinal invariance was established for mothers', fathers' and teachers' reports over a six-year period. Sex, age, and SES did not substantially moderate agreement among informants, although both sex and age were differentially related to symptomatology depending on the informant. Agreement among teachers and mothers, but not among mothers and fathers, differed by domain of symptomatology, and was greater for EXT than for INT. In Study 2, latent profile analysis, a person-centered analytic approach, did not provide easily interpretable patterns of symptom development, a failure that is likely the result of the relatively modest sample size. Latent growth curve models, an alternative analytic approach, did provide good fit to the data. Temperament and parenting variables were examined as predictors of the latent growth parameters in these models. Although there was little prediction of the slope, effortful control was negatively related to overall levels of EXT, whereas impulsivity and anger were positively related. Mutually responsive orientation, a measure of the parent-child relationship, was a more consistent predictor of EXT than was parental warmth. Furthermore, the relation between mutually responsive orientation and EXT was partially mediated by inhibitory control. Across informants, there were few consistent predictors of INT. In Study 3, latent profile analysis was used to classify individuals into different patterns of INT and EXT co-occurrence. In these models, a similar class structure was identified for mothers and for teachers. When temperament and parenting were examined as predictors of co-occurring symptomatology, few significant interactions were found and results largely replicated prior findings from this data set using arbitrary symptom groups.