This thesis contributes to the investigation of the facilitative role of written CF for L2 development. Framed within information processing and interactionist perspectives, this study not only investigated the effectiveness of written CF on learners’ improved accuracy, but also explored their responses to written CF with the aim of finding out whether or not, and why, written CF was beneficial or not for L2 development. This research was conducted with 157 first year university students in Northwest China. They were not majoring in English. A quasi-experimental study was conducted first to examine the differential effectiveness of five types of written CF (underlining, error code, metalinguistic explanation, direct correction and direct correction plus metalinguistic explanation). Additionally, two potential impacting factors on the effectiveness of written CF were investigated: linguistic types (regular past tense, irregular past tense and prepositions indicating space) and learners’ proficiency levels (higher and lower). The results suggested that more explicit types of written CF (metalinguistic explanation, direct correction and direct correction plus metalinguistic explanation) facilitated improved accuracy when the three linguistic types were considered as one group; however, the effectiveness was not retained over time. When the three linguistic types were looked at separately, only direct correction was found to facilitate improved accuracy of the irregular past tense immediately after feedback had been given. Although higher proficiency learners produced written texts of a higher accuracy than lower proficiency learners throughout the testing occasions, there was no difference between them with regard to the effectiveness of written CF for treating the targeted linguistic types. In a follow-up case study, the research focused on two individuals whose accuracy did not improve immediately after receiving various types of written CF. The results of the in-depth analyses of their correct uses and incorrect uses of the targeted linguistic types suggested that less explicit types of written CF may not be explicit enough for learners to produce the correct forms. Thus, it may explain why less explicit types of written CF did not result in the improved accuracy. However, there was no evidence to show that any type of written CF was ineffective because the errors which had received written CF did not appear again in the subsequent written texts. In addition, scaffolded written CF was provided on each of these learners’ targeted linguistic errors. The results showed that scaffolded written CF was more effective in facilitating improved accuracy immediately and over time; however, the explicitness of the written CF needed by each learner on different errors varied even when the errors were from the same linguistic category. Although the results of the quantitative study did not show statistical significance regarding the moderating role of learners’ proficiency levels on the effectiveness of written CF, the results of the case studies revealed that learners in the higher proficiency group required less explicit written CF assistance than the learners in the lower proficiency group who required more explicit written CF assistance, especially on less rule-governed, idiosyncratic and complex linguistic types.