To investigate what aspects of practice increase the relative accuracy of judgments of learning (JOLs), this study manipulated both JOL timing and type of practice. Three experiments examined the hypotheses that practice with test but not without test improves the relative accuracy of JOLs, and that a similar process mediates the effects of both delay and practice. The results of the experiments revealed that practice without test does not increase the relative accuracy of JOLs, but practice with test does, and that this advantage is different from the advantage caused by delay. These results are discussed in the context of the retrieval hypothesis of memory as well as theories of JOLs.