OBJECTIVES OF THIS RESEARCH WERE TO (1) INVESTIGATE THE EFFECTS OF THE RESPONSE MODE IN PROGRAMED INSTRUCTION, (2) EXTEND THE THEORETICAL PRINCIPLES ON WHICH PROGRAMED INSTRUCTION IS BASED, AND (3) DEVELOP EFFECTIVE METHODS OF TRAINING RECOGNITION SKILL. A SERIES OF FOUR EXPERIMENTS, INVOLVING SIXTH-GRADE PUPILS, WERE REPORTED. IN EXPERIMENT 1, STUDENTS WORKED THROUGH A LINEAR PROGRAM OF 2 VERSIONS, CONSTRUCTED-RESPONSE AND STRAIGHT TEXT. IMMEDIATE AND DELAYED TEST SCORES ON AN OBJECTIVE CRITERION TEST WERE HIGHER FOR THOSE STUDENTS WHO RECEIVED THE CONSTRUCTED-RESPONSE VERSION. EXPERIMENT 2 SHOWED THAT PERFORMANCE ON A CONSTRUCTED-RESPONSE TEST WAS SUPERIOR FOR THE CONSTRUCTED-RESPONSE TRAINING MODE. PREDICTED INTERACTION BETWEEN THE TRAINING MODE AND TYPE OF TEST ITEM WAS SIGNIFICANT. FINDINGS FROM A THIRD EXPERIMENT SUGGESTED THAT USING A COMBINATION OF MODES IN TRAINING MAY LEAD TO CONFUSION AND REDUCE THE EFFECTIVENESS OF A PROGRAM. EXPERIMENT 4 CONCERNED RECOGNITION OF LETTER-LIKE FORMS, ONE OF THE PRIMARY STEPS IN LEARNING TO READ. DATA SUGGESTED THAT A RELATIVELY LARGE AMOUNT OF TIME MIGHT WELL BE DEVOTED TO PRACTICE ON DISCRIMINATING LETTERS FROM THEIR TRANSFORMATIONS. (JC)