Research Project
2199520
The social and behavioral sciences are plagued by the existence of literally thousands of measurement instruments. For many of the more developed constructs (reading comprehension, spatial reasoning, extraversion) there are fifty or more published instruments generally varying in construct label, scaling methodology, and item format. Messick (1980) and others have commented on the jingle, jangle and jungle reflected in this diversity and it has been argued that the sterility of much behavioral science theorizing is due, in part, to the chaotic state of our measurement armament. A meta-analytic procedure called "construct generalization" has been proposed to bring order to this chaos. In short, the procedure involves: (1) stating a construct theory (say for extraversion), (2) operationalizing the theory in the form of a specification equation, (3) generating theoretical difficulties or endorsability values for each item in each test purporting to be a measure of the construct in question, (4) aggregating and rescaling the p values (observed difficulties) associated with each item, (5) correlating the observed difficulties with the theoretical difficulties, and (6) synthesizing a general statement of the fit of the construct theory to the observed item values. The proposed project will set forth a complete statement of the method, resolve several theoretical and statistical issues and develop a user friendly computer program for conducting construct generalization studies.
