Longitudinal comorbidity models posit some association between constructs over time. Although this association is often operationalized as cross-lagged autoregressive processes, associations between constructs are often expressed as stable interindividual associations between the constructs across measurement occasions. Although such general associations are often modeled as parallel state-trait models or as parallel (often polynomial) growth curves, such an approach risks overlooking the possibility of identification of developmentally limited traits and the implicit measurement model associated with the construct. Such models often present difficulties in terms of poor fit or improper solutions which are remedied ad hoc. In order to identify better fitting alternative, a “right-sizing” approach to development of comorbidity models is proposed wherein the dimensionality, patterning, and mean level of an observed series is first considered in order to identify the model that best represents prospective change over time. Cormorbidity of problem behaviors or psychopathology can then be expressed via covariation between the latent variables and growth models identified. The approach is illustrated with a prospective study of the cormorbidity of psychological distress and alcohol use in college students. Assumption checking procedures for the resulting model are also illustrated.