THE VARIOUS WEAKNESSES, DIFFICULTIES, AND "COMMON CONFUSIONS" THAT HAVE CHARACTERIZED THE CRITERIA USED IN TEACHER SELECTION PROCEDURES ARE IDENTIFIED, AND SUGGESTIONS FOR IMPROVEMENT ARE MADE. TEN STEPS IN THE DEVELOPMENT AND EVALUATION OF PROCEDURES ARE OUTLINED--FROM IDENTIFICATION OF LOCALLY VARYING EXPECTATIONS ARISING FROM VARYING VALUES (E.G., IS THE TEACHER TO BE PERMISSIVE OR A DISCIPLINARIAN) THROUGH DRAWING INFERENCES ABOUT THE VALIDITY OF THE PROCEDURES FOR PREDICTING THE (OPERATIONALLY DEFINED) CRITERION BEHAVIORS. CRITERIA MUST BE CLEARLY DISTINGUISHED FROM CRITERION MEASURES. PROBLEMS OF THE VALIDITY AND GENERALIZABILITY OF CRITERION DESCRIPTIONS AND OF THE VALIDITY AND REALIABILITY OF PROCEDURES FOR ESTIMATING CRITERION BEHAVIORS ARE DISCUSSED (E.G., IS THE BEHAVIOR UNI- OR MULTIDIMENSIONAL. ARE THE DIMENSIONS DISCRETE, REPRESENTATIVE, GENERALIZABLE, REPLICABLE). A VARIETY OF APPROACHES TO JUDGING THE VALIDITY OF CRITERION ESTIMATES AND TO DESCRIBING CRITERIA ARE GIVEN, AND SIX YARDSTICKS FOR JUDGING THE USEFULNESS OF A CRITERION DESCRIPTION ARE OUTLINED. THE MOST VALID OF SEVERAL METHODS OF OBTAINING CRITERION ESTIMATES USES SAMPLES OF THE ON-GOING CRITERION BEHAVIORS AND DIRECT ESTIMATION BASED ON OBSERVATION OF THEM IN PROCESS. THIS DOCUMENT APPEARED IN GILBERT, H.B., AND LANG, G., "TEACHER SELECTION METHODS," NEW YORK, 1967. (RP)