Thirty Taiwanese English-major students, grouped according to their proficiency test (NETPAW) scores and the feedback pattern that they had to use in revising their compositions, were asked to write descriptive essays by comparing any two of their personal belongings. The word count and T-units were listed in each of their four drafts, with each draft rated by three English teachers using a scoring scale. Results showed that except in the first draft, word count and T-units had significant relationships, and except in the final draft, ratings and NETPAW (National English Test in Proficiency for All on the Web) scores had significant relationships. However, their ratings per draft did not exhibit any significant difference. No significant difference existed among the ratings received by the three groups of students who were exposed to the feedback patterns. Moreover, there was no interaction effect between the drafts and the feedback patterns used. The best predictors for the final draft rating were draft 1(D1) word count, D3 ratings, and NETPAW scores.