Guthrie's findings (1967) were not duplicated in Matsuura (1970) and Morimasa et al (1982) where Japanese subjects learned rules in anagram tasks similar to those used in the Guthrie study. The present author assumed that the anagrams used in the latter studies were inappropriate for Japanese subjects and that is what produced the contrary results. In the present study, forty Japanese undergraduates learned revised anagrams under four different conditions: discovery-, discovery-and-expositoryinstruction-, expository-instruction-and-examples-, and control-conditions. The results were basically the same as in the Guthrie study. As predicted, discovery learning produced better transfer than expository instruction, while it provided little benefit for tasks that could be completed only by applying learned rules.