COUNSELORS OFTEN ADMINISTER TESTS OF QUESTIONABLE VALIDITY. IN RELIABILITY STUDIES, EVERY PRECAUTION IS TAKEN TO STABILIZE THE STIMULUS SITUATION. IN ASSESSING VALIDITY, CONCERN CENTERS ON BEHAVIOR UNDER DIFFERENT STIMULUS CONDITIONS. CRONBACH'S THEORETICAL LIMIT FOR A VALIDITY COEFFICIENT OF A TEST IS THE SQUARE ROOT OF THE RELIABILITY COEFFICIENT. GHISELLI, IN REVIEWING HUNDREDS OF VALIDITY STUDIES COMPLETED BETWEEN 1919 AND 1964, FOUND THAT NONE OF HIS FOUR MAJOR CLASSES OF APTITUDE TESTS FORECAST PROFICIENCY ON ANY JOB WITH A HIGH DEGREE OF ACCURACY. ALTHOUGH TESTS CAN HAVE A SUFFICIENTLY HIGH DEGREE OF PREDICTIVE POWER TO BE OF PRACTICAL VALUE IN PERSONNEL SELECTION, THEY MAY ALSO FRUSTRATE THE COUNSELOR'S CHANCES OF HELPING THE CLIENT SOLVE IMMEDIATE PROBLEMS. IN ACTUARIAL INTERPRETATION AND BEHAVIOR PREDICTION BASED ON TEST DATA, GREATER RELIANCE SHOULD BE PLACED ON MULTIVARIATE ANALYSIS. RIGOROUS COLLECTION, ANALYSIS, AND REPORTING OF PREDICTION AND CRITERION DATA ARE NECESSARY. USE OF GOLDMAN'S MULTIDIMENSIONAL APPROACH CAN HELP THE CLIENT BY DEFINING THE ESSENTIAL ELEMENTS OF THE PROBLEM. UNTIL MORE VALID TESTS ARE DEVELOPED, WE MUST CHOOSE BETWEEN REFUSAL TO USE TESTS, AND USING THEM AS PART OF AN EXTENSIVE DESCRIPTION. THIS SPEECH WAS PRESENTED AT THE AMERICAN PSYCHOLOGICAL ASSOCIATION CONVENTION, WASHINGTON, D.C., SEPTEMBER 2, 1967. (PR)