This study examines whether revision mediates the effects of its preceding written corrective feedback (CF) in developing second language (L2) knowledge. For this purpose, effects of direct CF and three sub-types of metalinguistic CF with and without revision were compared in the quasi-experiment with 197 EFL university students. They formed 9 groups (i.e., 4 revision groups, 4 non-revision groups, and 1 control group) and completed 4 dictogloss writings. Written CF and revision were implemented after the first two writings. English articles were the target structure for the written CF. The first, third and fourth writings were measured for the accuracy of article use as the pre-test, the immediate post-test and the delayed post-test, respectively. The results reveal that revision did not affect the efficacy of direct CF while it enhanced that of metalinguistic CF to varying degrees with its subtypes. These findings indicate that, if revision is conducted following written CF with particular metalinguistic information, it may facilitate learning new L2 knowledge beyond manifesting existing L2 knowledge.