How individuals monitor their level of comprehen-sion during study is a question of growing interest to researchers of metacognitive processes (see, e.g., Benja-min, Bjork, & Schwartz, 1997; Bjork, 1999; Dunlosky & Nelson, 1994; Koriat, 1997, 1998). Indeed, some argue (e.g., Bjork, 1999; Jacoby, Bjork, & Kelley, 1994) that the readings learners take on their level of comprehen-sion while studying can be as important as their actual comprehension, because, in part, such readings influence their decisions on how to allocate learning resources. On the basis of such readings, for example, students may de-cide to review one chapter rather than another, or to spend more time studying one set of materials versus another in preparation for an examination. Learners, however, can be far from accurate in taking such readings and can be subject to illusions of compre-hension (see, e.g., Bjork, 1999; Jacoby et al., 1994). They can, for example, be led to think that their level of com-prehension or skill is greater than it actually is, owing to conditions of learning (such as massed practice) that en-hance or support performance during study or training but actually impair long-term retention and/or transfer (Bjork, 1999; Simon & Bjork, 2001). Koriat (1997) has stated this slightly differently, arguing that learners can suffer from illusions of competence because they are relatively insensitive to factors in the learning environment (such as repeated presentations) that can enhance performance on later retention tests, while being overly sensitive to fac-tors (such as the perceived association between cues and targets when both are present during study) that do not necessarily enhance performance on later retention tests. In the present article, we describe research exploring the sensitivity, or lack thereof, of learners to the memorial benefits of one such factor: generation. In the sections below, we first define and illustrate the generation effect or advantage, describe two accounts of it that are consistent with a wide body of relevant findings, and then discuss our more recent research addressing the general issue of learners’ sensitivity to the memorial ben-efits of generation and whether—if made sensitive to this benefit—they might then adopt more effective encoding strategies in the processing of new information.Generation As a Condition of LearningWhen learners take an active part in generating the information they are learning, as opposed to having it provided to them, they tend to remember it better. If, for example, learners generate the word banana from a word fragment (e.g., b–n–n–), as opposed to being given the intact word to read, they will recall it better on a later test. Or, if required to generate the exemplar banana to a category-plus-letter-stem cue (e.g., Fruit–ba___) versus 207 Copyright 2007 Psychonomic Society, Inc.Learning how to learn: Can experiencing the outcome of different encoding strategies enhance subsequent encoding?ElizabEth ligon bjorkUniversity of California, Los Angeles, CaliforniaPatricia ann dEWinstanlEyOberlin College, Oberlin, OhioandbEnjamin c. stormUniversity of California, Los Angeles, CaliforniaResearch on how individuals monitor their level of comprehension during study paint a picture of learners as insensitive to many of the factors or conditions of learning that can enhance long-term retention and transfer. In the present article, we discuss research examining the sensitivity, or lack thereof, of learners to one such factor: generation. More specifically, we discuss research addressing the question of learners’ sensitivity to the memorial benefits of generation and whether—if given the opportunity to experience this benefit in their own recall performance—they might then go on to develop enhanced encoding strategies in the processing of new to-be-learned information.Psychonomic Bulletin & Review2007, 14 (2), 207-211E. L. Bjork, [email protected] Bjork, deWinstanley, and stormbeing given the intact pair to study, they will recall banana better in response to the cue Fruit on a later test. This memorial benefit of generation (see, e.g., Jacoby, 1978; Slamecka & Graf, 1978) has proved to be both ro-bust and to extend to a variety of learning materials, in-cluding lists of words, trivia questions (e.g., deWinstanley, 1995), and mathematical problems (e.g., McNamara & Healy, 1995a, 1995b; Pesta, Sanders, & Murphy, 1999). Under certain conditions, however, the generation effect can be diminished or even eliminated. McNamara and Healy (1995a, 1995b), for example, found that genera-tion advantages do not occur for arithmetic problems un-less retrieval strategies that reinstate procedures employed during study are evoked again at the time of test. Similarly, deWinstanley, Bjork, and Bjork (1996) demonstrated that even when participants learn the same materials, generation advantages may or may not occur, depending on the match between the information strength-ened during the generation task and the type of informa-tion required for optimal performance on a later test. More specifically, the conditions of learning were manipulated to force the processing of different types of information in order to generate targets for the same set of cue–target pairs. In one condition, the pairs were blocked into cat-egories, leading participants to focus on target–target rela-tional information (to which free recall tests are assumed to be most sensitive) rather than cue–target relational in-formation (to which cued recall tests are assumed to be most sensitive) in order to perform the generation task. On subsequent tests, participants showed a generation ad-vantage when given a free recall test, but not when given a cued recall test. In the other condition, pairs were not blocked by category, essentially eliminating target–target relational processing as a basis for generating targets and forcing participants to rely on cue–target relational infor-mation instead. On subsequent tests, these participants showed a generation advantage on a cued recall test, but not on a free recall test. In other words, a striking reversal was observed in the relative levels of free and cued recall for targets that had been generated versus read, depending on the type of information that participants were forced to use to generate the targets during learning.Whether generation effects occur can also be influenced by