ABSTRACT
Lee Cronbach has pointed out a number of critical problems in psychological methodology, especially in measurement and in the conceptualization of research investigation. He showed us that subjects taking tests often choose their response to an item on bases unrelated to the concept we are trying to measure; response sets (Cronbach, 1946, 1950) are systematic components in the measuring process producing a datum. They are intrusive components that vary over subjects, even among several giving the same overt response. So the actual measuring process is often not what the investigator wants it to be, or interprets it as being. A few years later, Cronbach and Meehl (1955) explicated the basic scientific process by which both measuring instruments and constructs are validated in work on rather global constructs such as those used in personality. Then Cronbach (1957) differentiated two contrasting disciplines in scientific psychology, the study of persons and the study of treatments. More recently (Cronbach, 1982a), he developed an analytic framework for program evaluation, specifying the inferences from units (such as persons), from treatments, and from particular measuring operations—a framework that can, I have found, be adapted fruitfully to inferential decisions in other areas of research. He has also cautioned us against being naively optimistic about the future of the course we have been following in psychology and other social sciences (Cronbach, 1975, 1982b, 1986).
