This study investigated the effects of negotiation in the process of providing written corrective feedback (WCF) by comparing three different patterns: non-negotiation, minimal negotiation, and extended negotiation. The data were collected from three intermediate academic English classes at a university in Korea. Each class was exposed to different degrees of negotiation while they received WCF for two target forms, the past verb forms and the English articles, in their written production (text re-construction tasks). Their accuracy improvement was measured by text-reconstruction writing tasks and error correction tests. The study found that the use of negotiation affected accuracy improvement of the target forms; however, its effectiveness varied depending on the target forms and measurements. For the past tense, negotiation, regardless of its amount, resulted in more accuracy improvement than non-negotiation in both text re-construction tasks and error correction tests. However, for the English articles, negotiation operated differently: extended negotiation significantly brought about more accuracy development in both measurements, but the difference between minimal negotiation and non-negotiation was only observed in the error correction test, not in the text re-construction task.