Previous feedback research for accuracy in second language writing has mainly employed directcorrective feedback. However, this type of feedback tailored to individual learners is notpractical for a large cyber university writing course. Thus, this study which targets Englisharticles explored the use of models as one common feedback method for all students' writings,with textual input enhancement and written metalinguistic explanations on the articles as addedcomponents of the feedback. Three groups were formed: one which received models only asfeedback, another with models where articles were textually enhanced, and the last withtextually enhanced models followed by metalinguistic explanations on the articles. All groupsengaged in narrative writing tasks consisting of individual writing (pre-tests), online relaywriting, individual revisions, and new individual writing (post-tests). The results showed somesignificant within-group differences. That is, all groups significantly improved on their revisionswith both types of articles, but on the post-tests, the models-only group improved significantlywith the definite articles from the pretests, while showing decreased accuracy for the indefinitearticles. The other groups improved with both types of articles on the post-tests, but thedifferences were not significant. Discussions of the findings and future research suggestions areprovided.