This study investigated whether the effect of two forms of computer-automated metalinguistic corrective feedback in drills transferred to subsequent writing tasks. The English simple past tense, a learned structure, was selected as the target structure. Participants included 117 intermediate learners of English as a foreign language assigned to two feedback groups, one no-feedback group, and one control group. These groups completed writing tasks before the drills, immediately after the drills, and two weeks after the drills. In the drills, the feedback groups completed an untimed error correction test (ECT 1) in which they received either metalinguistic feedback or metalinguistic feedback with corrections. After that, the feedback groups completed another untimed ECT (ECT 2). The no-feedback group completed the two ECTs without receiving any feedback. The results showed that the feedback groups performed better than the no-feedback group on ECT 2. However, no effect for group was found on the learners' improvement from the first to the second writing task and from the first to the third writing task.