SIXTY-FIVE STUDENTS IN TWO CLASSES IN HIGH SCHOOL GEOMETRY WERE ASSIGNED BY STRATIFIED RANDOM PROCEDURE ON THE BASIS OF THE HENNON-NELSON TEST OF MENTAL ABILITY TO FOUR EXPERIMENTAL GROUPS--TWO USING A LINEAR OR A BRANCHING TYPE PROGRAM EXCLUSIVELY, AND TWO SWITCHING PROGRAM TYPE MIDWAY THROUGH THE EXPERIMENT. A THIRD CLASS, TAUGHT BY THE SAME TEACHER, WAS GIVEN CONVENTIONAL INSTRUCTION. BOTH PROGRAMS WERE CONSTRUCTED FROM THE TEXT NORMALLY USED BY THE CLASS, THE LINEAR WITH VERY SMALL STEPS, AND THE AN ACHIEVEMENT TEST WAS ADMINISTERED AS PRETEST, POST-TEST, AND AN AHIEVEMENT TEST WAS ADMINISTERED AS PRETEST, POST-TEST, AND SEVEN WEEK DELAYED RETENTION TEST. AN ATTITUDE QUESTIONNAIRE WAS GIVEN HALFWAY THROUGH THE EXPERIMENT, WITH THE POST-TEST, AND WITH THE RETENTION TEST. ANALYSES OF VARIANCE SHOWED NO SIGNIFICANT DIFFERENCES AMONG EXPERIMENTAL GROUPS IN ACHIEVEMENT OR RETENTION, ALTHOUGH A SUB-GROUP OF HIGH ABILITY PERFORMED SIGNIFICANTLY BETTER THAN ONE OF LOW ABILITY STUDENTS IN ALL CONDITIONS. ALTHOUGH ALL EXPERIMENTAL GROUPS PREFERRED PROGRAMED TO CONVENTIONAL INSTRUCTION DURING AND IMMEDIATELY AFTER THE EXPERIMENT, NO PREFERENCE WAS SHOWN SEVEN WEEKS LATER. THE LINEAR PROGRAM WAS PREFERRED TO THE BRANCHING PROGRAM BY THE ENTIRE GROUP, ALTHOUGH IT TOOK SIGNIFICANTLY LONGER TO COMPLETE THAN THE BRANCHING PROGRAM. (BB)