Abstract
Error feedback in writing has long been a divisive topic resulting from different conditions in variables. A more systematic and controlled study with such differences decreased is obviously necessary. This present study is therefore intended to fill this need, employing a text where eight kinds of treatable errors' are implanted beforehand as target forms. Focusing on variables - feedback type, error type, the subject of feedback effect, and learner's metalinguistic knowledge of target form - three hypotheses were tested: (1) that any feedback causes error repair/editing more than no feedback; (2) that feedback on forms with more metalinguistic knowledge causes more error repair/editing; (3) that feedback effect depends on a form itself rather than its level/type. Results suggested that some feedback on some form can be effective on repair/editing to some extent more than no feedback, depending on the conditions of the variables.
