In a previous study (Watson, 1960), college students were told that a three-number string ("2, 4, 6") conformed to a rule and that they had to guess the rule by proposing other three-number strings; they were then told whether each guess was right or wrong. However, it was suspected that, since the subject's hypothesis was falsified by an experimenter, the true value of instructing students to falsify would only become apparent if they had to decide whether their hypotheses were right or wrong without being able to ask the experimenter. Psychology students (N=120) were assigned to one of three strategies: (1) confirmatory--proposing strings subjects thought would be correct if their hypotheses were correct; (2) disconfirmatory--proposing strings subjects thought would be incorrect if their hypotheses were correct; and (3) control--testing hypotheses without any suggestions as to how to go about testing. Results indicate that once experimenter feedback concerning the correctness of hypotheses was eliminated, disconfirmatory instructions greatly improved subjects' performance on the string task. Subjects instructed to disconfirm obtained a higher proportion of incorrect strings and were more likely to test a hypothesis with an incorrect string, suggesting that their instructions helped them avoid the "confirmation bias" noted in earlier research. (JN)