Repeated statements are more frequently judged to be true. One position relates this so-called "truth effect" to meta-cognitive experiences of fluency, suggesting that repeated statements are more frequently judged to be true because they are processed more fluently. While most prior research focused on why repetition influences truth judgments, considerably less is known about when fluency is used as information. The present research addresses this question and investigates whether reliance on fluency is moderated by learning experiences.Specifically, we focus on changes in the reliance on fluency over the course of time. A series of experiments reveals that fluency is more likely to be used in truth judgments when previous reliance on fluency has resulted in valid judgments, compared to when previous reliance on fluency was misleading. These findings suggest that reliance on fluency in judgments is a finely-tuned process that takes prior experiences with fluency-based judgments into account. Prior research has demonstrated that repeated statements are rated more frequently as true than new statements (e.g., Arkes, Hackett, & Boehm, 1989; Hasher, Goldstein, & Toppino, 1977; Hawkins & Hoch, 1992). This effect was denoted truth effect (see Hasher, et al., 1977, for a first description; Schwartz, 1982) and has been reliably replicated across research domains. One prominent explanation holds that repeated statements are rated more frequently as true because, compared to new statements, they can be processed more fluently (Reber & Schwarz, 1999). Specifically, it is assumed that the fluency experienced while encoding a statement is interpreted and used as a metacognitive cue with respect to the statement's truth. Corroborating this fluency-truth hypothesis, it has been demonstrated that increasing processing fluency independent of repetition, also produces truth effects. For instance, it has been shown that statements printed in high as compared to low color contrast are processed more fluently, and that this processing fluency translates into increased ratings of truth (e.g., Reber & Schwarz, 1999;Unkelbach, 2007). Moreover, statements are rated as more probably true when prior semantic activation (priming) facilitates processing (Kelley & Lindsay, 1993). Because fluency is manipulated independently of repetition, both of these findings suggest that repetition is not a necessary condition for truth effects. Rather, it appears that repetition-based truth effects are mediated via processing fluency (cf., Reber & Schwarz, 1999; see Unkelbach, Bayer, Alves, Koch, & Stahl, 2011; for a discussion of the independent effects of fluency and positivity). Subsequently, participants' naïve theories on whether positive or negative events are more difficult to retrieve were manipulated. The ease with which events from childhood could be retrieved influenced participants' evaluations of their childhood-however, this impact depended on whether participants believed that positive or negative childhood events come to min...