This study investigated the effects of extensive versus intensive recasts. The focus was on the effect of feedback on learning English articles, which, as nonsalient target structures, have been shown to be difficult for many second language learners. Intensive recasts were operationalized as recasts provided on article errors only, while extensive recasts were provided on any errors including article errors. Forty-eight adult intermediate learners of English as a second language (ESL) were divided into 3 groups: an intensive recast group, an extensive recast group, and a control group. Learners conducted 2 communicative tasks with a native-speaker instructor and received feedback on their errors. They were pretested and posttested (immediately and after 2 weeks) using 3 different outcome measures: an oral picture description task, a written grammaticality judgment task, and a written storytelling task. The results revealed that the extensive recast group significantly outperformed the control group on the oral picture description and the grammaticality judgment tasks, whereas the intensive recast group did not. On the written storytelling task, both recast groups outperformed the control group, but the difference was not statistically significant. These findings point to the advantage of extensive recasts and challenge the assumption that recasts on single errors are necessarily more effective than recasts on a wide range of errors.