The present study explores the relative efficacy of recasts over metalinguistic feedback on the learning of the English "that"-trace filter and how working memory capacity (WMC) is related to the extent to which learners can benefit from recasts and metalinguistic feedback. Fifty-four Korean English as a foreign language (EFL) learners from six intact classes at a university formed two experimental groups (recasts and metalinguistic feedback) and one control group and carried out two first language (L1) working memory (WM) span tasks (reading span and operation span tasks). The two experimental groups participated in two information gap activities over two treatment sessions, during which they were required to ask questions involving the "that"-trace filter and received corrective feedback (either recasts or metalinguistic feedback) on their erroneous utterances. Two dependent variable measures (a written production test and a grammaticality judgment test) were administered in each test session (pretest and immediate posttest). Results showed that recasts were as effective as metalinguistic feedback in facilitating the acquisition of the target construction. This may, to some extent, be attributable to the blocking of modified output opportunities specifically designed in this study to prevent modified output from playing a potential role as a confound. Also, individual differences in WMC significantly predicted, and thus mediated the effects of, recasts but not metalinguistic feedback, on the acquisition of the "that"-trace filter. This suggests that executive attention or attention control (considered as a critical component of WMC) is involved in the noticing of recasts, but not in the noticing of metalinguistic feedback.