The psychometric notion of construct validity has evolved greatly sinc
e its introduction by Cronbach and Meehl (1955). Its current status is
to be found in a carefully formulated article by Messick (1989) in wh
ich he embeds construct validity in a realist conception of social sci
ence. In this article the authors examine several facets.of Messick's
approach. They take issue with the emphasis placed on consequential va
lidity as a component of construct validity, the inadequate treatment
of the link between tests and scores, and the focus on construct valid
ity in large-scale standardized test development. The authors show the
profitability Of separating the construct-indicator link from the ind
icator-score link and suggest that although the latter is not always n
ecessary, its treatment in the construct validity literature has gener
ally been inadequate.