This study replicating two earlier studies investigated the relative effectiveness of two techniques for teaching English grammar. In the "garden path" method, learners are induced to produce errors, which are then corrected. The error avoidance method prevents error production by means of grammatical explanations. Subjects were 56 Japanese junior college students of English as a Foreign Language. Both the experimental group (garden path treatment) and control group (error avoidance treatment) were administered two kinds of tests, one in which they were to make grammaticality judgments of the if-clause, and one in which they rewrote sentences using the passive voice correctly. The two tests gave mixed results: in the first, no statistical difference emerged between the two groups; in the second, the experimental group performed better. It also appeared that both treatments had a temporary effect in the first test, while both had long-term effects in the other test. It is concluded that both techniques may work well, at least in the short term, for this population. (MSE)