It also suggests statistical methods for considering issues of validity and reliability. The proposed methods relay on recent developments in two types of models: (1) the ANCOVA model with contextual effects separating total aggregated variable relationships into components at the individual and organizational levels, and (2) structural equation models for the simultaneous analysis of longitudinal data from several cohorts. The methods differ from previous approaches, which generally rely on either subjective factors or objective factors but not both. Copies of the survey instruments to be used for inmates and for staff are presented. Figures, footnotes, lists of measures, copies of a study of staff expectations and perceptions at 6 institutions, and 57 references. (Author summary modified)