A SERIES OF EXPERIMENTS WAS DESIGNED TO STUDY THE VALUE OF TEACHING CHILDREN SUCH A COMPLEX PROBLEM-SOLVING STRATEGY AS TO USE KNOWLEDGE OF RESULTS AFTER A PERFORMANCE TRIAL TO REJECT MORE THAN ONE HYPOTHESIS AT A TIME (MULTIPLE HYPOTHESIS STRATEGY). APPROXIMATELY 150 THIRD-GRADE CHILDREN WERE DIVIDED INTO 3 GROUPS--THOSE TAUGHT THE MULTIPLE HYPOTHESIS STRATEGY, THOSE TAUGHT TO SELECT A HYPOTHESIS AND STICK WITH IT UNTIL PROVEN WRONG (SINGLE HYPOTHESIS STRATEGY), AND THOSE GIVEN NO STRATEGY INSTRUCTION WITH AND WITHOUT OPPORTUNITY FOR PRACTICE. INSTRUCTION AND TESTING WERE CARRIED OUT BY MEANS OF AUTOINSTRUCTIONAL PROGRAMS WITH COMPUTER-GENERATED SEQUENCE. THE EVIDENCE INDICATED THAT INSTRUCTION IN PROBLEM-SOLVING STRATEGIES WAS SUPERIOR TO UNGUIDED DISCOVERY. RELIABLE DIFFERENCES BETWEEN MULTIPLE AND SINGLE HYPOTHESIS STRATEGIES WERE NOT OBTAINED. MULTIPLE HYPOTHESIS STRATEGY INSTRUCTION WAS CONCLUDED TO BE NOT DEPENDABLY INFERIOR TO THE SINGLE HYPOTHESIS STRATEGY. INDIVIDUAL DIFFERENCES IN ABILITY AND AGE, HOWERVER, CORRELATED WELL WITH ALL THE TESTED STRATEGIES. (WN)