The question addressed in this article is this: How do people make judgments of learning? Judgments of learning (JOLs) are assessments that people make about how well they have learned particular information-that is, predictions about how likely they will be to remember a target item when later given a cue. These assessments are then, presumably, used to control further study (Benjamin, Bjork, & Schwartz, 1998;Dunlosky & Hertzog, 1998;Mazzoni, Cornoldi, & Marchitelli, 1990;Metcalfe, 2002;Metcalfe & Kornell, 2003;Nelson & Dunlosky, 1991;Nelson, Dunlosky, Graf, & Narens, 1994;Nelson & Narens, 1990Son, 2004;Son & Metcalfe, 2000). For example, Nelson and Dunlosky argued that "the accuracy of JOLs is critical, because if the JOLs are inaccurate, the allocation of subsequent study time will correspondingly be less than optimal" (p. 267). Because their role in learning is crucial, understanding the mechanisms underlying these judgments is central to understanding people's control of their own mental processes and for finding ways to ameliorate those processes. We will argue here that JOLs made at a delay are based on two distinct processes rather than on only the single retrieval process that previous researchers have emphasized. If so, then understanding both of those processes, and the pitfalls therein, may be pivotal to improving people's performance. In this article, we provide data that challenge the assumption that delayed JOLs are made only by an assessment of retrieval goodness or fluency. We argue and provide evidence that JOLs, like other metacognitive judgments, instead may involve two distinct stages: (1) a quick preretrieval stage based, perhaps, on cue familiarity, which determines whether or not the second stage-retrieval-will occur, and (2) a later stage in which the judgments are based on an assessment of the goodness and/or fluency of retrieval or on other aspects of the target that can become apparent once retrieval has been attempted.In the typical delayed-JOL paradigm that we investigated, participants studied cue-target pairs. Following study, they were presented with only the cue and were asked to judge how confident they were that they would be able to remember the target when they were given the cue on a future memory test. Then, at some later time, they were given a test. The prevailing view of how such JOLs are made is that when given the cue, people try to retrieve the target and base their judgments on their perception of some of the qualities of that attempted retrieval of the target, such as how quickly, easily, or fluently it comes to mind, or on how replete the retrieved information is (Begg, Duft, Lalonde, Melnick, & Sanvito, 1989;Benjamin et al., 1998;Dunlosky, Rawson, & McDonald, 2002;Kelley & Lindsay, 1993;Maki, 1998;Nelson & Dunlosky, 1991;Nelson & Narens, 1990;Nelson, Narens, & Dunlosky, 2004;Spellman & Bjork, 1992;. Three experiments tested the hypothesis that people make judgments of learning (JOLs) by attempting to retrieve the target first. If this were the whole story, then ...