The purpose of this study was to investigate the strategy of reforming learners' autogenetic erroneous criteria called ru. The rule on the sum of interior angles in a triangle was adopted. Beforehand, the fifth grade subjects had applied the rule only to a limited extent. Three groups called A, B, C were formed and were given 3 different instructions respectively. Group A subjects were instructed to form examples which they misjudged as exceptions and were to verify them according to the rule. Group B verified the rule with the examples which they had to choose one out of six previously set up under the same instruction as Group A. Group C were to do the same verification task as of Group A, though without being given a specific instruction in the forming task. The results suggested that it was difficult for learners to form the examples they spontaneously misjudged as exceptions. In post-test, Groups A, B, and C made good marks in the same order. Such results showed that the instruction making the learner verify the rule on a self made example had effects on reforming ru.