Recasts have continued to be the object of intensive empirical and theoretical inquiry following Nicholas, Lightbown, and Spada's (2001) review. The current article identifies a number of problems with this research and the supporting theory. These problems concern the fact that recasts can take many different forms and perform a variety of functions (not all corrective), which makes definition difficult. Also, recasts, when corrective, can vary in terms of whether they constitute an implicit or explicit corrective strategy and in whether they afford negative or positive evidence. Researchers have almost exclusively examined recasts from a cognitive perspective, ignoring their social and sociocognitive aspects. The significance of learner repair following recasts also remains controversial. Little is currently known about the role that the learner's developmental readiness plays in determining whether recasts work for acquisition. Researchers have not clearly distinguished between intensive and extensive recasts, nor have they considered their differential benefits. The acquisitional value of recasts in comparison to other forms of corrective feedback might have been overestimated. This article emphasizes the need for research that examines the specific properties of recasts and the social and instructional conditions in which they occur.