The Importance of Construct Validity: Alternative Models for the Assessment of Behavioral Disorders

Abstract

Problems in the definition and conceptualization of behavioral disorders have created problems for both researchers and practitioners alike. The implications of these conceptual difficulties for assessment are considered in a quantitative synthesis of correlations between behavioral ratings and classroom observation. The results, based on 89 correlations drawn from 16 studies, suggest significant problems in the areas of construct and criterion-related validity of widely-used measures of social behavior. The mean correlation between teacher ratings and observations of similar behaviors was, 23. Although correlations were somewhat higher when broken down by behavior, analysis of a multitrait, multimethod matrix suggests that minimum requirements for convergent validity may not be met. These results appear to pose serious methodological difficulties for the models of assessment that stress individual pathology. An alternative “components of variance” model is proposed that appears to better account for social and contextual variables present in educational decision making.