Testing Construct Theories

This paper presents and illustrates a novel methodology, construct-specification equations, for examining the construct validity of a psychological instrument. Whereas traditional approaches have focused on the study of between-person variation on the construct, the suggested methodology emphasizes study of the relationships between item characteristics and item scores. The major thesis of the construct-specification-equation approach is that until developers of a psychological instrument understand what item characteristics are determining the item difficulties, the understanding of what is being measured is unsatisfyingly primitive. This method is illustrated with data from the Knox Cube Test which purports to be a measure of visual attention and short-term memory.