The study investigates whether differences in metalinguistic information contained in written corrective feedback (CF) mediate its effectiveness for second language (L2) development. To address this, metalinguistic CF was distinguished into three types in terms of how specific its metalinguistic information was. They were then compared among themselves and with non-metalinguistic direct CF in their efficacy for short- and long-term development of explicit and implicit L2 knowledge. The target features for written CF were the English articles, and explicit and implicit knowledge were measured by the error correction test and dictogloss writings by 93 EFL learners respectively. The results suggest that, for both explicit and implicit L2 knowledge, metalinguistic CF was beneficial for short-term development only when it had high levels of specific information while it was effective for long-term development regardless of its type. The findings are discussed from the perspective of SLA theory, and their pedagogical implications and suggestions for future research are put forth.