THE PAPER REPORTS ON AN ATTEMPT TO DETERMINE EMPIRICALLY, FOR SEVERAL TEST-CONSTRUCTION PROBLEMS, THE AMOUNT OF IMPROVEMENT RESULTING WHEN TESTS ARE TAILORMADE TO FIT ONE PARTICULAR CHARACTERISTIC OF A LOCAL POPULATION--THE BASE RATES OF THE TWO CRITERION GROUPS WHICH THE TEST IS DESIGNED TO SEPARATE. THE BASIC PROCEDURE USED WAS TO CONSTRUCT A SERIES OF TESTS WHICH WERE ALIKE IN THE ITEM POOL AND ITEM-SELECTION TECHNIQUE USED, AND IN THE TWO CRITERION GROUPS WHICH THE TESTS WERE DESIGNED TO SEPARATE, BUT WHICH DIFFERED IN THE RELATIVE BASE RATES OF THE TWO CRITERION GROUPS ASSUMED IN THE CONSTRUCTION OF THE TESTS. CROSS-VALIDATION SAMPLE DATA WERE THEN USED TO ESTIMATE THE VALUE OF EACH OF THE TESTS IN POPULATIONS WITH EACH OF THE ASSUMED BASE RATES. THE PURPOSE WAS TO ESTIMATE, FOR EACH OF THESE POPULATIONS, THE EXTENT TO WHICH THE TEST TAILORMADE FOR THAT POPULATION EXCEEDED IN VALUE TESTS TAILORMADE FOR POPULATIONS WITH DIFFERENT BASE RATES. THE RESULTS SHOWED NO NOTICEABLE DIFFERENCE IN THE VALUES OF THE VARIOUS TESTS. THESE RESULTS WERE CONSISTENT ACROSS FOUR DIFFERENT TEST-CONSTRUCTION METHODS STUDIED, AND ACROSS THREE DIFFERENT SETS OF DATA WHICH DIFFERED IN THE ITEM POOL AND THE CRITERION GROUPS USED. (AUTHOR)