ABSTRACT
In the Industrial/Organizational (I/O) PhD program at Purdue and other universities in the 1960s, the textbooks and most courses taught that selection procedure validities were situationally specific. The dominant theory held that the validity of the same test for what seemed to be the same job varied from employer to employer, region to region, across time periods, and so forth. It was believed and taught that the same test could have high validity in one location or organization and at the same time have zero validity in another for the same or very similar jobs. This belief was supported by the finding that observed validity coefficients for similar tests and jobs varied substantially across different validity studies and the finding that some of these validity coefficients were statistically significant and others were not. The explanation for this puzzling variability was that jobs that appeared to be the same differed in important but subtle ways in what was required to perform them. The conclusion, we were taught, was that the validity of selection procedures had to be estimated anew for each different situation or setting by a validity study conducted in that setting. It was impossible to generalize validity from one setting to others.
