THE STUDY EXAMPLE USED IN THE ARTICLE IS A LONGITUDINAL EVALUATION OF A FIELD-BASED COMPENSATROY EDUCATIONAL PROGRAM FOR YOUNG CHILDREN. THE FACTORS TO BE CONSIDERED IN DETERMINING THE NUMBER OF CHILDREN NEEDED TO PARTICIPATE IN A LONGITUDINAL STUDY ARE DISCUSSED. THE LEVEL AT WHICH QUESTIONS ARE TO BE ADDRESSED (SUCH AS WHETHER THE PROGRAM EFFECTIVENESS IS BEING EVALUATED ACROSS SEVERAL SITES OR WITHIN ONE SITE) AND THE NEED TO TEST FOR INTERACTIONS (SUCH TESTING HAS A MULTIPLICATIVE EFFECT ON SAMPLE SIZE) CAN AFFECT SAMPLE SIZE. APPROPRIATE COMPARISON GROUPS MUST BE IDENTIFIED SO THAT EVALUATION QUESTIONS CAN BE ANSWERED WITH A MINIMUM OF BIAS. THE NATURE OF THE TREATMENT AND ITS EXPECTED CONSISTENCY MUST ALSO BE CONSIDERED. ESTIMATES OF EXPECTED EFFECT SIZE SHOULD TAKE INTO ACCOUNT COMPETING TREATMENTS RECEIVED BY THE COMPARISON GROUP, THE EXTENT TO WHICH THE TREATMENT IS INCONSISTENTLY IMPLEMENTED, AND THE EVALUATIONS OF SIMILAR PROGRAMS THE NEXT FACTOR TO BE CONSIDERED IS ATTRITION RATE: AN ATTRITION RATE OF 50 PERCENT CAN BE EXPECTED IN MANY MULTIYUEAR EVALUATIONS. IN THE HYPOTHETICAL FIELD-BASED STUDY, A SAMPLE SIZE OF 60 IN EACH GROUP IN EACH SITE AT POSTTEST WAS DETERMINED; TAKING ATTRITION INTO ACCOUNT RESEARCHERS DOUBLED THE SAMPLE NUMBER TO 120. THE POINT IS MADE THAT NO PANACEAS ARE AVAILABLE FOR THE PROBLEMS RAISED, BUT THEY MUST BE ADDRESSED ON A CASE-BY-CASE BASIS. A DIAGRAM AND EIGHT REFERENCES ARE PROVIDED.