THIS STUDY SOUGHT TO DETERMINE WHETHER SERIAL LEARNING EFFICIENCY DEPENDS ON THE RELATIONSHIP BETWEEN PART SIZE AND THE EMPIRICALLY DETERMINED MEMORY SPAN OF THE LEARNERS. ISSUES EXPLORED WERE THE TRADE-OFF BETWEEN REHEARSABILITY AND EFFORT IN OPTIMIZING THIS RELATIONSHIP, AND FEASIBILITY OF A FLEXIBLE PART SIZE. A MEASURE OF MEMORY SPAN WAS COMPUTED, AND THEN 65 GRADE 6 STUDENTS WERE RANDOMLY ASSIGNED TO 3 TREATMENT GROUPS, 2 VARYING PART SIZE AND 1 A SELF-DIRECTED (COMPLETE FREEDOM IN METHOD OF STUDY) GROUP. A LIST OF LAST NAMES OF THE FIRST 30 U.S. PRESIDENTS WERE PRESENTED IN WHOLE OR IN PART ON 6 TRIALS. SCORES AT 3 MEMORY SPAN LEVELS SHOW A GREATER EFFECTIVENESS OF A PART SIZE SUBSTANTIALLY LARGER THAN MEMORY SPAN, SUGGESTING THAT EFFORT INDUCED BY STRETCHING THE LEARNER'S MEMORY SPAN MAY BE MORE IMPORTANT THAN GUARANTEE OF IMMEDIATE REHEARSAL. BUT OTHER DATA PRESENTED CONTRADICT THIS AND SUGGEST AN ALTERNATE, UNKNOWN EXPLANATION. (LH)