Abstract
Cognitive models are usually conceptualized at the individual level, but are often analyzed at the group level. The level at which analysis can be carried out is dictated by the experimental design, which traditionally has been chosen for pragmatic reasons. Defining the model at the group level allows the incorporation of individual difference variables, which are of interest to many researchers, into the model structure. The nested groups design, with participants nested under the classificatory variables, is appropriate when the model is defined at the group level. That design is illustrated with a study testing a multiplicative model of anticipated compliance, in which medical patients were grouped according to the symptoms and prognoses associated with their diseases.