1 On Reactivity to Immediate Judgments of Learning Doctoral dissertation by Franziska Ingendahl Submitted in fulfillment of the requirements for the degree of Doctor rerum naturalium (Dr. rer. nat.) at the Department of Human Sciences of the Technical University of Darmstadt First reviewer: Prof. Dr. Monika Undorf Second reviewer: Prof. Dr. Beatrice G. Kuhlmann Darmstadt 2025 Ingendahl, Franziska: On Reactivity to Immediate Judgments of Learning Darmstadt, Technische Universität Darmstadt, Year thesis published in TUprints: 2025 DOI: https://doi.org/10.26083/tuda-7532 URL: https://tuprints.ulb.tu-darmstadt.de/handle/tuda/14728 URN: https://nbn-resolving.de/urn:nbn:de:tuda-tuda-147288 Date of the viva voce: 10.11.2025 Published under Copyright III Contents Summary V Zusammenfassung VII Manuscripts IX 1 Introduction 1 2 Theoretical Foundations 5 2.1 Reactive Effects of Immediate JOLs 5 2.2 Theoretical Accounts of Immediate JOL Reactivity 7 2.2.1 Cue-Strengthening Account 8 2.2.2 Dual-Task Account 9 2.2.3 Changed-Goal Account 10 2.2.4 Other Accounts 11 2.3 Immediate JOL Reactivity: Where We Stand 12 3 Establishing JOL Reactivity 15 3.1 Establishing Reactivity: Empirical Complexities and Methodological Obstacles 15 3.2 On the Phenomenon of JOL Reactivity 17 4 Explaining JOL Reactivity 25 4.1 Explaining Reactivity: Empirical Inconsistencies and Theoretical Limitations 25 4.2 On Learning Strategies as a Proximal Mechanism of JOL Reactivity 26 4.2.1 Spontaneous Learning Strategy Use 27 4.2.2 Instructed Learning Strategy Use 32 5 Generalizing JOL Reactivity 37 5.1 Generalizing Reactivity: Widespread Discussions and Absence of Empirical Tests 37 5.2 On the Educational Relevance of JOL Reactivity 39 6 General Discussion 43 6.1 Implications 44 6.1.1 Toward an Integrated Framework of JOL Reactivity 44 6.1.2 Toward a Critical Test of Educational Relevance 49 6.1.3 Toward a Deliberate Use of Metacognitive Judgments 50 6.2 Future Research Directions 51 6.3 Conclusion 53 7 Bibliography 55 A Acknowledgments 65 B Statement of Originality 67 C Copies of the Manuscripts 69 V Summary Measurement is an indispensable part of our everyday life. We use thermometers to gauge outside temperatures or clocks to keep on time. While many of those everyday measurements do not alter the observed object, this is not true for all types of measurements – a phenomenon commonly referred to as reactivity. Due to the assumption that learners monitor their learning spontaneously either way, research on people’s knowledge and cognition about their learning and memory (metamemory) has long neglected the possibility that assessing metacognitive monitoring might alter learning and memory itself. However, over the last decade, the examination of reactivity of predictions of one’s future memory performance (immediate judgments of learning, JOLs) has become a hot topic in metamemory research. Extensive research now demonstrates that soliciting JOLs can alter memory performance across diverse study and test conditions. Yet, the field faces several challenges pertaining to establishing JOL reactivity as a phenomenon, explaining its underlying mechanisms, and determining its generalization to educational settings. The aim of this dissertation was to systematically address these challenges and thus to advance the field of JOL reactivity research. I established JOL reactivity as a robust empirical phenomenon and identified the moderating variables that delineate its boundary conditions by conducting a comprehensive meta-analysis on reactivity of immediate JOLs to memory performance. Through theory-driven moderator analyses that test the principal theoretical explanations of JOL reactivity (cue-strengthening account, dual-task account, and changed-goal account), the meta-analysis revealed that none of these accounts provides a comprehensive explanation of the phenomenon. To address the resulting need for new theoretical developments, I uncovered changes in learning strategy use as a proximal mechanism underlying negative effects of JOLs and demonstrated that learning strategy instructions targeted at reducing JOL-induced learning strategy changes eliminate these negative effects. Finally, I tested whether JOL reactivity generalizes to educationally relevant study and test conditions. Using the example of general knowledge facts tested in short-answer and multiple-choice tests, this examination revealed no generalization of reactivity effects, thereby questioning the much-discussed educational relevance of JOL reactivity. The three projects in this dissertation aim to advance the understanding of JOL reactivity by establishing reactivity as a robust yet heterogeneous phenomenon that varies in its size and direction (positive or negative), delineating its boundary conditions, uncovering changes in learning strategy use as a mechanism for negative reactivity, and demonstrating that JOL reactivity does not reliably generalize to educationally relevant study and test conditions. This research emphasizes that future theoretical developments must account for the heterogeneity and context dependency of JOL reactivity and that further tests of the educational relevance of JOL reactivity are important directions for future research. Moreover, it demonstrates that reactivity can pose a serious methodological obstacle for metamemory research, requiring researchers to carefully consider reactivity effects when drawing conclusions based on metacognitive judgments. Summary VI Ultimately, this dissertation illustrates that Heisenberg’s (1958, p. 58) insight – that “what we observe is not nature itself but nature exposed to our method of questioning” – applies as profoundly to the study of metamemory as it does to quantum mechanics. VII Zusammenfassung Messungen sind ein unverzichtbarer Bestandteil unseres Alltags. Wir verwenden Thermometer, um die Außentemperatur zu messen, oder Uhren, um pünktlich zu sein. Während viele dieser alltäglichen Messungen das beobachtete Objekt nicht verändern, gilt dies nicht für alle Arten von Messungen – ein Phänomen, das allgemein als Reaktivität bezeichnet werden. Aufgrund der Annahme, dass Lernende ihr Lernen so oder so spontan überwachen, hat die Forschung zum Wissen und zur Überwachung von Menschen über ihr eigenes Lernen und Gedächtnis (Metagedächtnis) lange Zeit vernachlässigt, dass die Erfassung metakognitiver Überwachung das Lernen und Gedächtnis selbst verändern könnte. Jedoch beschäftigen sich Forschende seit zehn Jahren zunehmend mehr mit der Untersuchung der Reaktivität der Erfassung von Vorhersagen über die zukünftige Gedächtnisleistung der eigenen Person (judgments of learning, JOLs). Zahlreiche Untersuchungen zeigen inzwischen, dass die Erfassung von JOLs die Gedächtnisleistung unter verschiedenen Lern- und Testbedingungen verändern kann (JOL-Reaktivität). Dennoch steht das Forschungsgebiet zu JOL-Reaktivität aktuell vor mehreren Herausforderungen, was die Etablierung von JOL-Reaktivität als Phänomen, die Erklärung ihrer zugrunde liegenden Mechanismen und die Bestimmung ihrer Generalisierbarkeit auf Lernen in Bildungskontexten betrifft. Das Ziel dieser Dissertation ist es, diese Herausforderungen systematisch zu adressieren und damit die Forschung im Bereich der JOL-Reaktivität voranzubringen. Im Rahmen der Dissertation wird JOL-Reaktivität als robustes empirisches Phänomen etabliert und moderierenden Variablen und Randbedingungen identifiziert, indem eine umfassende Metaanalyse zu Reaktivität von JOLs auf die Gedächtnisleistung durchgeführt wurde. Durch theoriegeleitete Moderatoranalysen, die die wichtigsten theoretischen Erklärungen für die JOL-Reaktivität (Cue-Strengthening-Ansatz, Dual-Task-Ansatz und Changed-Goal-Ansatz) testen, zeigt die Metaanalyse dabei, dass keine dieser Erklärungen eine umfassende Erklärung für das Phänomen liefert. Um dem daraus resultierenden Bedarf an neuen theoretischen Entwicklungen gerecht zu werden, identifiziert die Dissertation Veränderungen in der Verwendung von Lernstrategien als proximalen Mechanismus für negative Reaktivität, und zeigt, dass Lernstrategieinstruktionen zur Reduktion JOL-induzierter Veränderungen in Lernstrategien diese negativen Reaktivitätseffekte eliminiert. Abschließend untersucht die Dissertation, ob sich JOL-Reaktivität auf bildungsrelevante Lern- und Testbedingungen verallgemeinern lässt. Anhand einer exemplarischen Betrachtung von Allgemeinwissensfakten und Kurzantwort- und Multiple-Choice-Tests, ergab diese Untersuchung keine Verallgemeinerung der Reaktivitätseffekte, wodurch die viel diskutierte bildungsrelevante Bedeutung der JOL-Reaktivität in Frage gestellt wird. Die drei Projekte dieser Dissertation zielen darauf ab, das Verständnis von JOL- Reaktivität zu verbessern, indem sie Reaktivität als ein robustes, aber heterogenes Phänomen etablieren, das in seiner Größe und Richtung (positiv oder negativ) variiert, ihre Randbedingungen etablieren, Veränderungen in der Verwendung von Lernstrategien als Mechanismus für negative Reaktivität aufdecken und zeigen, dass JOL-Reaktivität nicht VIII zuverlässig auf bildungsrelevante Studien- und Testbedingungen verallgemeinert werden kann. Diese Forschung betont, dass zukünftige theoretische Entwicklungen die Heterogenität und Kontextabhängigkeit der JOL-Reaktivität berücksichtigen müssen und dass weitere Tests der Bildungsrelevanz von JOL-Reaktivität wichtige Richtungen für die zukünftige Forschung sind. Darüber hinaus veranschaulicht sie, dass Reaktivität ein ernsthaftes methodisches Hindernis für Forschung zu Metagedächtnis darstellen kann, weshalb Forscher Reaktivitätseffekte berücksichtigen sollten, wenn sie Schlussfolgerungen auf der Grundlage metakognitiver Urteile ziehen. Letztendlich zeigt diese Dissertation, dass Heisenbergs (1958, S. 58) Erkenntnis „what we observe is not nature itself but nature exposed to our method of questioning“ für die Erforschung von Metagedächtnis ebenso zutrifft wie für Quantenmechanik. IX Manuscripts This cumulative dissertation is based on three manuscripts published or accepted for publication and one manuscript submitted for publication. The four manuscripts aim to extend the understanding of the phenomenon, the underlying mechanisms, and the educational relevance of reactive effects from assessing metacognitive judgments during learning (immediate judgments of learning, JOLs) on memory performance. To this end, this dissertation includes a comprehensive assessment and synthesis of immediate JOL reactivity and its boundary conditions (Manuscript I, Ingendahl et al., in press). It further provides evidence for learning strategies as a proximal mechanism of negative JOL reactivity, allowing for a test of multiple contemporary theoretical accounts of immediate JOL reactivity (Manuscripts II, Ingendahl & Undorf, 2025a, and III, Ingendahl & Undorf, 2025b). Lastly, this dissertation provides an empirical test of the much-discussed educational relevance of immediate JOL reactivity (Manuscript IV, Ingendahl & Undorf, 2024). In the following synopsis, I present an overview and overarching framework for these manuscripts. By summarizing and discussing the results of this research, I refrain from elaborating on detailed statistical and empirical aspects of the research reported in the manuscripts, which are appended to this dissertation in the same order as listed below. Manuscript I Ingendahl, F., Halamish, V., & Undorf, M. (in press). Do immediate judgments of learning alter memory performance? A meta-analytical review. Psychological Bulletin. https://doi.org/10.1037/bul0000487 Manuscript II Ingendahl, F., & Undorf, M. (2025a). Changes in learning strategies contribute to negative reactivity of immediate judgments of learning. Journal of Experimental Psychology: Learning, Memory, and Cognition. Advance online publication. https://doi.org/10.1037/xlm0001475 Manuscript III Ingendahl, F., & Undorf, M. (2025b). Instructed learning strategy use eliminates negative reactivity of immediate judgments of learning. Manuscript invited for revision at Psychonomic Bulletin & Review. Manuscript IV Ingendahl, F., & Undorf, M. (2024). On the educational relevance of immediate judgment of learning reactivity: No effects of predicting one’s memory for general knowledge facts. Journal of Applied Research in Memory and Cognition, 13(1), 113-123. https://doi.org/10.1037/mac0000113 “We have to remember that what we observe is not nature itself but nature exposed to our method of questioning” Werner Heisenberg (1958) 1 1 Introduction In everyday life, we rely on various tools and methods to observe, measure, and assess the world around us. For example, thermometers help us gauge outdoor temperatures, yardsticks help us measure the height of children, and clocks help us keep track of time. In many of these indispensable everyday measurements and assessments, the mere fact that we observe, measure, or assess does not alter the object of observation: Temperatures do not rise because we use a thermometer, children do not grow higher because they are stood against a yardstick, and time does not pass faster because we glance at a clock. While this is true for many observations, measurements, and assessments, it is certainly not true for all and, specifically, not for scientific measurements and assessments. Various examples show that scientific measurement can indeed alter the observed object: In optics, observing molecules with optical microscopes can change or even eliminate their fluorescent properties (e.g., Song et al., 1995). In quantum mechanics, where Heisenberg’s uncertainty principle (Heisenberg, 1927) serves as a landmark example of measurement- induced change, continuous observation can impede the decay of unstable particles (e.g., Misra & Sudarshan, 1977). In medicine, the measurement of blood pressure by a doctor can induce hypertension in patients with perfectly normal blood pressure (e.g., Franklin et al., 2013). Hence, it is no surprise that measurement-induced changes, so-called reactivity effects, are particularly important in psychology, in which both observers (i.e., researchers) and those being observed (i.e., participants) are humans who actively perceive and engage with their surroundings. In psychology, reactivity refers to individuals modifying their behavior or performance as a consequence of being observed or measured (Double & Birney, 2019). One of the most prominent examples of reactivity in psychology is the Hawthorne effect, which describes that individuals alter their typical behavior in response to being observed (e.g., McCarney et al., 2007). Reactivity has been discussed in many areas of psychology, from fundamental areas of psychology, such as cognitive psychology (e.g., Fox et al., 2011) and developmental psychology (e.g., Reynolds et al., 2016), to applied areas of psychology, such as educational psychology (e.g., Beymer & Robinson, 2022), clinical psychology (e.g., Runhardt, 2021), or health psychology (e.g., French & Sutton, 2010), to psychological methods (e.g., Arslan et al., 2021). The area of psychology that has probably been shaped most by reactivity is the area of cognitive psychology (for a review, see Fox et al., 2011). Already in the early days of scientific psychology of the Würzburg School at the beginning of the 20th century, critics such as Wilhelm Wundt (e.g., Wundt, 1907) voiced their concerns about using introspection for studying cognitive processes (for an overview, see Hackert & Weger, 2018). Also, the reintroduction of self-reports for examining thinking and reasoning during the cognitive revolution in the 1950s and 1960s gave rise to critical discussions on the reactive side effects of self-reports. Critics argued that assessing one’s thoughts by self- reports might alter one’s thinking (e.g., Nisbett & Wilson, 1977), particularly when self- reports prompt people to attend to thoughts they would otherwise not attend to (Ericsson & Simon, 1980). The debate on reactivity of self-reports contributed significantly to a 1 Introduction 2 paradigm change and new methodological developments in the study of thinking and reasoning. Researchers introduced forms of self-reports less prone to reactivity, such as think- alouds (e.g., Ericsson & Simon, 1980, 1993), and developed methods that allow for tracing cognitive processes without requiring verbal reports, such as reaction-time analyses (e.g., Pieters, 1985) or eye tracking (e.g., Ball, 2013). Although reactivity is now part of almost all textbooks on cognitive psychology (Jäkel & Schreiber, 2013), it has not been considered equally in all areas of cognitive psychology. One area in which reactivity has only recently received attention is the area of metacognition (Double & Birney, 2019). Metacognition is people’s “knowledge and cognition about cognitive phenomena” (Flavell, 1979, p. 706). It provides insights into what people know about their cognitions and how they monitor and control their cognitive processes (Nelson & Narens, 1990). Metacognition is highly important in everyday life as it helps people make informed decisions about which information to restudy for an exam or when to give up or seek help. Training metacognition improves academic performance (e.g., Eberhart et al., 2025), and metacognitive abilities predict academic achievement above and beyond intelligence (Vrugt & Oort, 2008). Because metacognition focuses on people’s own beliefs and cognitions, research on metacognition has primarily employed self-reports and metacognitive judgments specifically for studying metacognitive processes (Dunlosky et al., 2016). In the area of metamemory – the cognition about one’s learning and memory (Nelson & Narens, 1990) – the immediate judgment of learning (JOL; Arbuckle & Cuddy, 1969) is one of the most widely used metacognitive judgments. Immediate JOLs are prospective judgments of the likelihood of remembering a recently studied item on a future memory test provided during learning (Rhodes, 2016). When predicting memory performance, people have been shown to base their JOLs on cues (e.g., characteristics of the task or items) they perceive during learning (cue-utilization account; Koriat, 1997). As such, JOLs provide insights into the cues people use for predicting their memory (e.g., Koriat et al., 2004; Undorf et al., 2018, 2022), whether these cues are valid with regard to memory performance (e.g., Besken, 2018; Koriat & Bjork, 2006; Rhodes & Castel, 2008), and whether people are generally accurate in predicting their future memory (e.g., Bröder & Undorf, 2019; Koriat et al., 2002; Rhodes & Tauber, 2011). Over the last decades, research on immediate JOLs has focused on examining JOLs as a dependent variable of metacognitive monitoring (e.g., see Rhodes, 2016). However, studying the reactive effects of providing immediate JOLs on memory performance has become a hot topic in metamemory research (Halamish & Undorf, 2021). Although there were already critical voices (Spellman & Bjork, 1992) and anecdotal evidence (e.g., Zechmeister & Shaughnessy, 1980) about reactive effects of assessing metacognitive judgments more than 30 years ago, research has only recently begun to systematically examine whether assessing immediate JOLs alters memory performance (Mitchum et al., 2016; Soderstrom et al., 2015). About a decade after the initial articles on immediate JOL reactivity by Soderstrom and colleagues (2015) and Mitchum and colleagues (2016), numerous studies show that assessing people’s metacognitive monitoring with immediate JOLs can alter memory performance compared to studying without providing JOLs (e.g., Halamish & Undorf, 2023; Janes et al., 2018; Maxwell & Huff, 2022; Myers et al., 2020; Rivers et al., 2021; Tauber & Witherby, 2019; Undorf et al., 2024; Witherby & Tauber, 1 Introduction 3 2017) – a phenomenon that has been termed JOL reactivity (cf. Double et al., 2018). The rapid development of JOL reactivity research, with nine published articles in 2024 alone, underscores the increasing attention and perceived importance of JOL reactivity within the field. However, this development poses several challenges and open questions to the field concerning establishing JOL reactivity as a phenomenon, its explanation, and its generalizability. This dissertation aims to address these challenges by presenting a comprehensive synthesis of the phenomenon of JOL reactivity, an alternative explanation of JOL reactivity by introducing a proximal mechanism of JOL reactivity, and an empirical test of the generalizability of JOL reactivity to educationally relevant study and test conditions. The dissertation is structured in six chapters, with Chapter 1 serving as an introduction. Chapter 2 presents the theoretical overarching framework for this dissertation by reviewing the current state of research on immediate JOL reactivity and contemporary theoretical accounts of the underlying mechanisms, and introducing the challenges that the field of JOL reactivity research currently faces. Chapters 3 to 5 address the three key challenges in the field of JOL reactivity. In Chapter 3, I address the first challenge of establishing JOL reactivity by presenting a meta- analysis of immediate JOL reactivity (Manuscript I, Ingendahl et al., in press) that allows for a comprehensive and in-depth assessment of JOL reactivity as a phenomenon overall. In Chapter 4, I provide a new perspective on the second challenge of explaining JOL reactivity by showing that changes in learning strategy use serve as a proximal mechanism of JOL reactivity effects. To provide a broad insight into the role of learning strategies in JOL reactivity, I present evidence from spontaneous learning strategy use (Manuscript II, Ingendahl & Undorf, 2025a) and instructed learning strategy use (Manuscript III, Ingendahl & Undorf, 2025b). In Chapter 5, I tackle the third challenge of the untested generalizability of JOL reactivity to educational settings by showing that JOL reactivity does not generalize to educationally relevant materials using the example of general knowledge facts (Manuscript IV, Ingendahl & Undorf, 2024). The final Chapter 6 discusses and integrates the findings from the four manuscripts on the phenomenon, a proximal mechanism, and the educational relevance of JOL reactivity, and presents an integrated framework of processes underlying JOL reactivity. This dissertation concludes by discussing how reactivity should be considered in educational settings and future metamemory research, and by providing directions for future research on JOL reactivity. 5 2 Theoretical Foundations As outlined in Chapter 1, reactivity is a ubiquitous phenomenon in the empirical sciences and of utmost importance in psychological research. Over the last decade, the study of reactivity has also become increasingly important in metacognition and metamemory research (see Double & Birney, 2019). This is particularly true for the widely used immediate JOLs (Halamish & Undorf, 2021). But how can reactivity of immediate JOLs be characterized and what do we know about its underlying mechanisms? Finally, where does research on immediate JOL reactivity stand, and where is it heading? 2.1 Reactive Effects of Immediate JOLs The idea that assessing immediate JOLs might alter learning and memory dates back to the early work by Arbuckle and Cuddy (1969). When comparing groups of participants providing metacognitive judgments at study and test or only at test, Arbuckle and Cuddy (1969) observed enhanced recall for participants providing metacognitive judgments at study and test. Despite this early anecdotal evidence of reactivity, subsequent research largely overlooked reactivity. This may have been partly due to other earlier studies observing no reactivity (e.g., Kelemen & Weaver, 1997; but see Zechmeister & Shaughnessy, 1980). However, the more important reason was probably the prevailing assumption that learners monitor their learning either way to effectively control their learning and achieve their desired goals (Jordano & Touron, 2018), as implied by cyclical models of metamemory (e.g., Nelson & Narens, 1990). Building on the observation by Ericsson and Simon (1980) that self-reports are only reactive when they require people to attend to cognitive processes they would otherwise not attend to, researchers assumed that assessing immediate JOLs should not be reactive (Halamish & Undorf, 2021). Put differently, if people monitor themselves during learning either way, asking them about their monitoring should not alter learning. However, recent research questions this assumption by emphasizing that metacognitive thoughts may not occur spontaneously as often as assumed (for a review, see Jordano & Touron, 2018) and that the use of metacognitive monitoring needs to be trained (e.g., Dignath et al., 2023; Eberhart et al., 2025). Today’s systematic examination of reactivity effects in metamemory dates back to two fundamental studies by Soderstrom and colleagues (2015) and Mitchum and colleagues (2016). In what is now considered the standard paradigm of reactivity of immediate JOLs, Soderstrom and colleagues (2015) examined whether soliciting immediate JOLs during learning impacts memory performance. Across several experiments, they compared two groups of participants (see Figure 1 for an illustration of the paradigm). Whereas one group, the no-JOL group, studied related (e.g., “eye – color”) and unrelated word pairs (e.g., “penguin – coffee”) for a fixed duration, the other group, the JOL group, studied the word pairs for the same duration but also gave an immediate JOL by predicting their future memory performance halfway during studying. When examining the performance in a subsequent cued-recall test (e.g., “eye – ?”), Soderstrom and colleagues (2015) observed that the size of the effect of word-pair relatedness on recall (i.e., the relatedness effect, Rabinowitz et al., 1982; Mueller et al., 2013) differed between groups. Specifically, they observed that the JOL group consistently showed a larger relatedness effect on cued-recall 2 Theoretical Foundations 6 performance than the no-JOL group. This increased relatedness effect stemmed from the JOL group recalling more related word pairs than the no-JOL group – a finding that Soderstrom and colleagues (2015) termed positive reactivity. In contrast, cued-recall performance for unrelated word pairs did not differ between groups. That is, there was no reactivity for unrelated word pairs. Mitchum and colleagues (2016) employed a similar paradigm in which participants self-paced their study time and provided JOLs immediately after the item presentation (see Figure 1). Like Soderstrom and colleagues (2015), they observed that the JOL group showed an increased relatedness effect, which, however, was driven by reduced cued-recall performance for unrelated word pairs in the JOL group compared to the no-JOL group, so-called negative reactivity. Soderstrom and colleagues (2015) and Mitchum and colleagues (2016) provided important initial evidence that immediate JOLs can alter memory in the form of an increased relatedness effect in cued recall. Further, they showed that the reactive effects of immediate JOLs can be beneficial (positive reactivity) or detrimental (negative reactivity) to memory and that JOL reactivity does not inevitably occur under all conditions (no reactivity). Figure 1 Illustration of the JOL Reactivity Paradigm and the Types of JOL Reactivity Note. The item types, test types, and study times presented here are for illustrative purposes only and do not represent all item types, test types, and test durations studied in JOL reactivity paradigms. (a), (b), and (c) refer to different procedural variations of the onset of the JOL prompt in the JOL group. eye - color 8 s / self-pacedNo -J O L gr ou p JO L gr ou p eye – color eye – color P(recall)? ___ % 4 s 4 s eye – color P(recall)? ___ % self-paced self-paced eye – color P(recall)? ___ % 8 s / self-paced eye – ? _____ (a) JOL after item onset with item present Soderstrom et al. (2015) (b) JOL simultaneously with item onset Maxwell & Huff (2022) (c) JOL after item presentation Mitchum et al. (2016) eye – ? _____ JOL Reactivity Paradigm Study Test % re ca lle d Positive reactivity % re ca lle d Negative reactivity % re ca lle d No reactivity Types of Reactivity 2 Theoretical Foundations 7 Building on these initial studies, researchers have obtained extensive evidence of immediate JOL reactivity (Double & Birney, 2019). Positive reactivity for related word pairs in cued-recall tests is now a well-replicated finding (e.g., Chang & Brainerd, 2022; Double, 2023; Halamish & Undorf, 2023; Janes et al., 2018; Li et al., 2023; Li, Shanks, et al., 2024; Maxwell & Huff, 2022, 2023, 2024; Myers et al., 2020; Tauber & Witherby, 2019; Witherby et al., 2023; Witherby & Tauber, 2017) that has been found under experimenter-paced (e.g., Halamish & Undorf, 2023; Janes et al., 2018; Witherby et al., 2023; Witherby & Tauber, 2017) and self-paced study conditions (Double, 2023; Maxwell & Huff, 2022, 2023, 2024). It has been demonstrated with retention intervals of a few minutes (e.g., 2 or 3 minutes, e.g., Janes et al., 2018; Tauber & Witherby, 2019; Undorf et al., 2024) and of 2 days (Witherby & Tauber, 2017). Positive reactivity for related word pairs has also been observed irrespective of when during studying the JOL prompt appears (see Figure 1 for an overview). This includes immediate JOLs prompted during the item presentation, either simultaneously with the item onset (Maxwell & Huff, 2022, 2023) or after some delay (e.g., Halamish & Undorf, 2023; Janes et al., 2018; Witherby & Tauber, 2017), and immediate JOLs prompted after the item presentation (Li, Shanks, et al., 2024; Murphy et al., 2024; Myers et al., 2024). Importantly, positive reactivity has been found to generalize to other types of word pairs tested in cued-recall tests, including backward-related pairs (Maxwell & Huff, 2022, 2023; but see Mitchum et al., 2016; Undorf et al., 2024), identical pairs (Chang & Brainerd, 2022; Halamish & Undorf, 2023), semantically mediated pairs (Maxwell & Huff, 2024), or category- cued pairs (Rivers et al., 2023). Positive reactivity was also evident in recognition tests of related word pairs (Maxwell & Huff, 2024; Myers et al., 2020, 2024), single words (Li et al., 2022; Li, Pastötter, et al., 2024; Maxwell, 2025; Zhao et al., 2022; Zhao, Li, et al., 2023; Zheng, Li, et al., 2024), and images of objects and scenes (Shi et al., 2022). Furthermore, recent research indicates that positive reactivity is not limited to the predominantly studied age group of young adults but extends to children in grades 1 through 5 (Zhao et al., 2022). However, making JOLs does not always enhance memory. For unrelated word pairs, making JOLs has been shown to either impair (Halamish & Undorf, 2023; Li, Shanks, et al., 2024; Undorf et al., 2024; Witherby et al., 2023) or not impact cued-recall performance (e.g., Janes et al., 2018; Maxwell & Huff, 2023; Myers et al., 2020; Rivers et al., 2021). Similarly, studies examining single words tested in free-recall paradigms report mixed results: Some demonstrate positive reactivity (Chang & Brainerd, 2024; Senkova & Otani, 2021), while others find no reactivity (Kaya & Mulligan, 2024; Maxwell, 2025; Zhao, Yin, et al., 2023). Negative reactivity effects were consistently revealed in research on memory for temporal order (Zhao, Li, et al., 2023; Zhao, Yin, et al., 2023), and no reactivity was recently observed in source-memory tests (Myers et al., 2024). Furthermore, there is inconsistent evidence on whether positive reactivity generalizes to older adults (Murphy et al., 2024; Tauber & Witherby, 2019; Zheng et al., 2025). In summary, recent research shows that providing immediate JOLs during learning can enhance or reduce memory performance under various conditions but does not always do so. 2.2 Theoretical Accounts of Immediate JOL Reactivity Several theoretical accounts have been proposed to explain why providing immediate JOLs alters learning and memory (Janes et al., 2018; Mitchum et al., 2016; Rivers et al., 2021; Senkova & Otani, 2021; Soderstrom et al., 2015; Zhao et al., 2022; Zhao, Li, et al., 2023; Zhao, Yin, et al., 2 Theoretical Foundations 8 2023; also see, Double & Birney, 2019). Three of these accounts – the cue-strengthening account (Soderstrom et al., 2015), the dual-task account (Mitchum et al., 2016), and the changed-goal account (Mitchum et al., 2016) – are most discussed in the scholarly literature on immediate JOL reactivity and are also of particular importance to this dissertation. Therefore, this chapter presents these accounts and their empirical evaluation in detail. Other accounts are only introduced briefly and revisited in the General Discussion in Chapter 6. 2.2.1 Cue-Strengthening Account The cue-strengthening account was proposed by Soderstrom and colleagues (2015) as an explanation for their initial observation of positive reactivity. Based on the idea that people base their JOLs on cues they perceive during learning (Koriat, 1997), the account assumes that making JOLs strengthens the processing of cues (e.g., relatedness) on which JOLs are based. If the later memory test (e.g., a cued-recall test) is sensitive to these strengthened cues, JOLs should enhance memory. For related word pairs, for example, making JOLs is assumed to enhance the processing of the semantic association between cue and target words because people base their JOLs on this semantic association (relatedness effect on JOLs; Mueller et al., 2013). According to the cue- strengthening account, the strengthened semantic association results in positive reactivity for related word pairs in cued-recall tests because cued-recall tests are sensitive to this cue; that is, a strong association is helpful for recalling the target word (e.g., “color”) when presented with the cue word (e.g., “eye”). No reactivity is proposed for unrelated word pairs because they contain no inherent semantic association that JOLs could strengthen. Although initially proposed for explaining reactivity for related word pairs tested in cued-recall tests, the cue-strengthening account has been suggested to explain positive reactivity for other types of materials, such as single words (Maxwell & Huff, 2024) for which JOLs are presumably based on cues other than semantic association. Several studies aimed to test the assumptions and predictions of the cue-strengthening account. Some studies provide support for the cue-strengthening account: In line with the assumption that making JOLs strengthens the cues on which JOLs are based, it has been found that making JOLs enhanced the processing of relatedness of word pairs and elevated cued recall for related and identical word pairs (Halamish & Undorf, 2023; also see Chang & Brainerd, 2022). Findings of positive reactivity for immediate JOLs but no reactivity for JOLs provided before the item presentation (Chang & Brainerd, 2022; Li, Shanks, et al., 2024), so-called prestudy JOLs (Castel, 2008), also support the idea that positive reactivity is due to enhanced processing of cues. Because prestudy JOLs are assessed before studying each item, they can only rely on limited cue information compared to immediate JOLs and, thus, should impact memory to a lesser extent (Chang & Brainerd, 2022). Further, positive reactivity for category-cued word pairs (e.g., “a type of gem – jade”) can be explained by the cue-strengthening account when assuming that making JOLs enhances the processing of the association between categories and exemplars (Rivers, Dunlosky, et al., 2023). Similarly, positive reactivity for mediated word pairs (e.g., “lion – stripes”, Maxwell & Huff, 2024), whose relatedness is established through a shared semantic association (e.g., “tiger”), is consistent with enhanced processing of the relational information connecting cue and target words. Supporting the assumption that immediate JOLs only enhance memory performance in memory tests that are sensitive to the strengthened cues, positive reactivity for 2 Theoretical Foundations 9 related word pairs has been observed in cued-recall tests but not in free-recall tests that are insensitive to relatedness information (Chang & Brainerd, 2022; Myers et al., 2020). However, not all studies support the cue-strengthening account. Positive reactivity is not consistently observed in cued-recall tests when study lists consist solely of related word pairs (Janes et al., 2018; but see, e.g., Maxwell & Huff, 2023; Witherby & Tauber, 2017), and the observed type of reactivity for weakly related word pairs depends on the other item types in the study list (Chang & Brainerd, 2022; Soderstrom et al., 2015). According to the cue-strengthening account, however, making JOLs should strengthen cues regardless of the presence or type of other studied items, as immediate JOLs pertain only to the just-studied item. Also, positive reactivity in cued-recall tests has been found for backward-related word pairs (e.g., “woman – nurse”) that share no strong semantic association from cue word to target word (Maxwell & Huff, 2022, 2023), in contrast to the assumption that JOL reactivity should only appear when the memory test is sensitive to the strengthened cues. Similarly, robust positive reactivity for related word pairs in target-only recognition tests has been demonstrated despite this test being insensitive to the primarily strengthened cue (i.e., the semantic association between cue and target words; Maxwell & Huff, 2024; Myers et al., 2020, 2024). Furthermore, the cue-strengthening account is inconsistent with any observation of negative reactivity (e.g., Li, Shanks, et al., 2024; Mitchum et al., 2016; Witherby et al., 2023), as it is confined to explaining positive effects of immediate JOLs and, therefore, needs to be combined with other accounts to explain both types of reactivity (Janes et al., 2018; Undorf et al., 2024). In summary, the cue-strengthening account has received considerable support, particularly from studies examining related word pairs tested in cued-recall tests. At the same time, in its original form, it fails to explain findings beyond the standard paradigm of relatedness and is generally inconsistent with negative reactivity effects. 2.2.2 Dual-Task Account The dual-task account (Mitchum et al., 2016) aims to explain why making immediate JOLs impairs memory performance. It assumes that conscious monitoring of one’s learning and memory is a resource-intensive process. The dual-task account proposes that asking participants to solicit immediate JOLs interferes with the primary task of studying and thus impairs memory performance through dual-task costs. Dual-task costs should mainly manifest when the primary task of studying is difficult (Mitchum et al., 2016). Thus, the dual-task account predicts negative reactivity when the primary task demands of studying are high (e.g., when studying unrelated word pairs) but no reactivity when the primary task demands are low (e.g., for related word pairs). In line with the dual-task account, negative reactivity has been observed when the primary task is relatively difficult. This includes studying difficult materials, such as unrelated word pairs (e.g., Halamish & Undorf, 2023; Li, Shanks, et al., 2024; Undorf et al., 2024; Witherby et al., 2023), and performing difficult memory tests, such as temporal order-construction tests (Zhao, Li, et al., 2023). Further, making JOLs has been found to reduce study time efficiency for unrelated word pairs (Mitchum et al., 2016). That is, participants who made JOLs had a lower memory performance for unrelated word pairs than participants who did not make JOLs, despite having studied the word pairs for a similar self-paced study time. However, the dual-task account faces two major obstacles: First, it cannot explain why making JOLs does not always impair memory performance for unrelated pairs (e.g., Janes et al., 2 Theoretical Foundations 10 2018; Maxwell & Huff, 2024; Rivers et al., 2021; Soderstrom et al., 2015). Second, it is inconsistent with the well-replicated finding of positive reactivity. Because the dual-task account focuses on the negative effects of providing JOLs, it must be combined with other theoretical accounts to explain both positive and negative reactivity (see Janes et al., 2018; Undorf et al., 2024). To sum up, whereas recent research supports the assumption of the dual-task account that making JOLs can interfere with learning, the dual-task account fails to provide a comprehensive explanation of positive and negative JOL reactivity. However, it is worth noting that the limited extent of evidence supporting the dual-task account is related to the fact that only few studies (e.g., Mitchum et al., 2016; Zhao et al., 2025) test the dual-task account specifically. This might be partially due to negative reactivity being less frequently reported and not widely considered a valid finding in research on immediate JOL reactivity (Undorf et al., 2024). 2.2.3 Changed-Goal Account The changed-goal account (Mitchum et al., 2016) takes a motivational perspective to explain why making JOLs enhances memory for some materials and/or reduces memory for other materials. It was proposed to explain negative reactivity for unrelated word pairs and positive reactivity for related word pairs based on a single underlying mechanism. The changed-goal account proposes that positive and/or negative reactivity can be explained by a change in the goal learners pursue during learning: Whereas learners initially and under conditions of studying without JOLs tend to adopt the goal of mastering all studied items for the later memory test, providing immediate JOLs during learning (i.e., repeatedly predicting the likelihood of remembering the studied items) makes learners realize that they probably cannot remember all items at the test. Consequently, learners shift their goal from mastering all items to maximizing their performance on the test by focusing on those items they perceive to be easy or moderately difficult (e.g., related word pairs) and neglecting those they perceive to be difficult (e.g., unrelated word pairs). This goal shift results in positive reactivity for items perceived as easier and/or negative reactivity for items perceived as more difficult. Multiple studies align with the changed-goal account. Research with cued-recall tests has shown positive reactivity for materials typically perceived as easy, such as related word pairs (e.g., Halamish & Undorf, 2023; Janes et al., 2018; Rivers et al., 2021), backward-related word pairs (Maxwell & Huff, 2022, 2023), identical word pairs (Chang & Brainerd, 2022; Halamish & Undorf, 2023), as well as negative reactivity for materials typically perceived as difficult, such as unrelated word pairs (e.g., Li, Shanks, et al., 2024; Undorf et al., 2024; Witherby et al., 2023), consistent with the assumption that people focus on items perceived as relatively easier when making JOLs. A shift of focus on items perceived as relatively easier when making JOLs also explains mixed findings of reactivity for weakly related word pairs: When weakly related word pairs are studied alongside more difficult materials, such as unrelated word pairs, making JOLs leads to a focus on weakly related pairs because they appear easier in comparison, resulting in positive reactivity for weakly related word pairs (Chang & Brainerd, 2022). Conversely, when weakly related word pairs are presented alongside easier materials, such as strongly related word pairs, making JOLs causes learners to neglect the weakly related pairs because they appear more difficult in comparison, leading to no positive reactivity for weakly related word pairs (Soderstrom et al., 2015). Highlighting the role of differences in perceived item difficulty as a boundary condition of JOL reactivity, Janes and colleagues (2018) found no reactivity for homogeneous lists of related or 2 Theoretical Foundations 11 unrelated word pairs but observed reactivity when the same items were included in a mixed list (but see Maxwell & Huff, 2023). Further, Myers and colleagues (2020) showed that JOLs enhanced performance for related word pairs in recognition tests (also, see Maxwell & Huff, 2024; Myers et al., 2024), supporting that making JOLs changes learning goals rather than test-specific processes. Other findings are inconsistent with the changed-goal account. Contrary to the assumption that JOL reactivity is due to a changed learning goal because of repeated predictions of one’s own memory, no reactivity has been observed for prestudy JOLs (Chang & Brainerd, 2022; Li, Shanks, et al., 2024) despite also involving repeated predictions of one’s future memory performance. Also, there is mixed evidence whether perceived differences in item difficulty are a central boundary condition of reactivity, with a recent study showing positive reactivity for related word pairs irrespective of whether other item types were included in the study list (Maxwell & Huff, 2023; but see Janes et al., 2018). The changed-goal account can also not explain why reactivity differs across memory tests, such as why making JOLs enhances memory in cued-recall tests (e.g., Janes et al., 2018; Rivers, Dunlosky, et al., 2023; Soderstrom et al., 2015) and recognition tests (e.g., Myers et al., 2020; Shi et al., 2022; Zheng, Li, et al., 2024) but not consistently in free-recall tests (Myers et al., 2020; Senkova & Otani, 2021; Zhao, Li, et al., 2023). Overall, the changed-goal hypothesis provides an explanation for empirical findings of both positive and negative reactivity. Despite having received considerable support, it is contested by mixed evidence for its main assumptions that repeated exposure to JOLs leads to changes in learning goals and that differences in perceived difficulty are a central boundary condition of JOL reactivity. 2.2.4 Other Accounts In addition to the three accounts presented above, other theoretical explanations of the mechanisms underlying immediate JOL reactivity have been proposed. The item-specific processing account (Senkova & Otani, 2021) states that making immediate JOLs enhances memory by inducing deep, item-specific processing, i.e., the processing of the characteristics of an individual item (e.g., whether an item is pleasant or unpleasant or whether it is concrete or abstract). In line with the item-specific/relational framework of memory (e.g., Einstein & Hunt, 1980), the item-specific processing account has been extended with the idea that by inducing item- specific processing, making JOLs disrupts inter-item relational processing and thus impairs memory for the order in which items are studied (Zhao, Li, et al., 2023; Zhao, Yin, et al., 2023). The attentional-reorientation account (Rivers et al., 2021; Zhao et al., 2022), sometimes also referred to as the enhanced learning-engagement account (e.g., Li et al., 2022; Tauber & Witherby, 2019), proposes that making JOLs improves memory because it reorients learners’ attention on the task and, therefore, enhances their learning engagement. It states that learners’ attention fluctuates during learning and that they start mind wandering during a learning task (Rivers et al., 2021; Rivers, Janes, et al., 2023; Shi et al., 2022). Because making JOLs requires learners to refocus their attention on the learning task, it increases actual study time which in turn enhances memory performance. 2 Theoretical Foundations 12 2.3 Immediate JOL Reactivity: Where We Stand Over the last ten years, the examination of reactivity of immediate JOLs has emerged as a hot topic in research on metacognition and metamemory. Whereas the initial studies (Mitchum et al., 2016; Soderstrom et al., 2015) provided foundational evidence that soliciting immediate JOLs can alter memory performance, subsequent research shows a complex and multifaceted pattern of findings. Immediate JOL reactivity seems to be diverse in its occurrence and direction. Assessing JOLs does not always alter memory (e.g., Zhao, Li, et al., 2023), and if so, JOLs can enhance or impair later memory performance (e.g., Li, Shanks, et al., 2024; Witherby et al., 2023; also, see Undorf et al., 2024). In addition, multiple theoretical accounts (Mitchum et al., 2016; Rivers et al., 2021; Senkova & Otani, 2021; Soderstrom et al., 2015) approach the question of why immediate JOLs alter memory from different perspectives and have received considerable attention in current research (e.g., Janes et al., 2018; Kaya & Mulligan, 2024; Li, Shanks, et al., 2024; Rivers, Janes, et al., 2023). Despite significant progress, the field faces several key challenges limiting our understanding of immediate JOL reactivity. The rapid increase in research on immediate JOL reactivity, with more than 25 papers published over the last five years, has resulted in greater knowledge about the conditions under which JOL reactivity can occur. However, this rapid expansion has also made establishing a coherent perspective on JOL reactivity increasingly difficult, as it reveals a complex and sometimes contradictory pattern of findings. The multitude of findings and their diversity make it challenging to determine the occurrence, size, and direction of JOL reactivity overall. In addition, the heterogeneity in methodological approaches mask the boundary conditions under which JOL reactivity occurs. Establishing the occurrence, size, and boundary conditions of immediate JOL reactivity by synthesizing the existing evidence remains a key challenge. Another key challenge in research on immediate JOL reactivity is explaining JOL reactivity. Whereas multiple theoretical accounts have been proposed, our understanding of the underlying mechanisms of JOL reactivity remains limited. The similarity and lack of specificity of the predictions of the proposed accounts, as well as the limited overlap of the conditions for which predictions can be derived make it difficult to test the predictions of different accounts against each other (Myers et al., 2024; Undorf et al., 2024). This is further aggravated by the fact that the most widely recognized theoretical accounts (i.e., the cue-strengthening account, Soderstrom et al., 2015, the dual-task account, Mitchum et al., 2016, and the changed-goal account, Mitchum et al., 2016) were already developed in the initial studies with word pairs, questioning their adequacy for explaining reactivity more broadly. On top of this, none of the theoretical accounts can comprehensively explain findings on JOL reactivity. Explaining the underlying mechanisms of immediate JOL reactivity, therefore, presents an essential task for advancing research in this field. The implications of JOL reactivity for educational settings have received much attention (e.g., Chang & Brainerd, 2022; Janes et al., 2018; Maxwell, 2025; Soderstrom et al., 2015; Tekin & Roediger, 2020; Witherby & Tauber, 2017). In particular, recent research discusses whether the positive effects of immediate JOLs on memory performance can be utilized for learning in educational settings to improve academic achievement (e.g., Soderstrom et al., 2015; Tekin & Roediger, 2020; Witherby & Tauber, 2017). Although training the use of metacognitive judgments during learning, such as JOLs, can generally improve learning under educationally relevant 2 Theoretical Foundations 13 conditions (e.g., Händel et al., 2020; for a meta-analysis, see Dignath et al., 2023), it would be premature to draw conclusions about the educational implications of JOL reactivity. All current findings on immediate JOL reactivity stem exclusively from controlled, often lab-based, paradigms with simple materials, such as single words or word pairs. As the educational relevance is increasingly discussed but remains untested, the generalization of immediate JOL reactivity to educationally relevant study and test conditions is a key task for JOL reactivity research. In this dissertation, I address the key challenges and tasks in the field of JOL reactivity. In the following three chapters, I introduce these challenges in detail and address them in three empirical projects. Figure 2 presents an overview of the key challenges and related empirical projects. In the first project on the phenomenon of JOL reactivity, I address the challenge of establishing JOL reactivity as a phenomenon by presenting a comprehensive meta-analysis of immediate JOL reactivity that synthesizes the existing findings and identifies important boundary conditions. In the second project, I approach the challenge of explaining JOL reactivity by introducing a novel perspective on the mechanisms underlying JOL reactivity by identifying learning strategies as a proximal mechanism of JOL reactivity. In the third and final project on the educational relevance of JOL reactivity, I attend to the question of generalizing JOL reactivity by providing an empirical test of the generalizability of JOL reactivity to educationally relevant materials, using the example of general knowledge facts. Together, the projects presented in this dissertation aim to improve our understanding of immediate JOL reactivity. Figure 2 Overview of Challenges in JOL Reactivity Research with Corresponding Projects and Manuscripts Presented in This Dissertation Note. Project refers to three subprojects presented in this dissertation. Challenges in Immediate JOL Reactivity Research Establishing JOL reactivity Explaining JOL reactivity Generalizing JOL reactivity Manuscript I Project 1 On the phenomenon of JOL reactivity Manuscripts II and III Project 2 On a proximal mechanism of JOL reactivity Manuscript IV Project 3 On the educational relevance of JOL reactivity Projects in This Dissertation 15 3 Establishing JOL Reactivity As outlined in Chapter 2, research on immediate JOL reactivity currently faces key challenges that limit the understanding of reactivity. One of these is establishing JOL reactivity – that is, systematically synthesizing existing evidence to determine the occurrence, size, and boundary conditions of JOL reactivity. 3.1 Establishing Reactivity: Empirical Complexities and Methodological Obstacles Research on JOL reactivity has developed rapidly in recent years. Whereas JOL reactivity initially received only gradual attention in metamemory research after the first articles (Mitchum et al., 2016; Soderstrom et al., 2015) were published, interest in reactivity effects has grown strongly over the last years. Only five studies (Janes et al., 2018; Mitchum et al., 2016; Soderstrom et al., 2015; Tauber & Witherby, 2019; Witherby & Tauber, 2017; also, see Halamish, 2018) examined JOL reactivity within the first five years of research. In the following five years, between 2020 and 2024, the number of newly published studies on the topic increased sixfold. This surge in research has contributed greatly to our knowledge of the conditions under which JOL reactivity occurs. Before 2020, research on JOL reactivity mainly focused on examining related and unrelated word pairs tested in cued-recall tests. Only in more recent years studies have started examining JOL reactivity for other materials, such as single words (see Senkova & Otani, 2021) or images (see Shi et al., 2022), and other memory tests, such as recognition tests (see Myers et al., 2020), free-recall tests (see Myers et al., 2020), order-construction tests (see Zhao, Li, et al., 2023), and source-memory tests (see Myers et al., 2024). The extensive research in recent years has advanced the understanding that JOL reactivity extends far beyond studying related and unrelated word pairs for cued-recall tests. At the same time, the rapid accumulation of research poses a problem for a comprehensive description of JOL reactivity as a phenomenon at large. The findings reveal a complex and sometimes contradictory pattern of JOL reactivity: On the one hand, many studies show that JOL reactivity indeed occurs under various conditions. This is particularly true for positive reactivity observed for related word pairs in cued-recall tests (e.g., Janes et al., 2018; Li, Shanks, et al., 2024; Maxwell & Huff, 2022; Myers et al., 2020; Soderstrom et al., 2015; Zhao et al., 2025) and single words in recognition tests (e.g., Li et al., 2022; Li, Pastötter, et al., 2024; Maxwell, 2025; Zheng, Li, et al., 2024). On the other hand, the findings also indicate great heterogeneity: The occurrence and direction of JOL reactivity vary not only between study and test conditions (see Chapter 2) but also within similar conditions. For example, studies examining JOL reactivity for unrelated word pairs in cued- recall tests with experimenter-paced study time sometimes show no (e.g., Janes et al., 2018; Maxwell & Huff, 2023; Soderstrom et al., 2015) and sometimes negative reactivity (e.g., Halamish & Undorf, 2023; Li, Shanks, et al., 2024; Mitchum et al., 2016). The results also differ tremendously for single words tested in free-recall tests; Some studies report positive 3 Establishing JOL Reactivity 16 reactivity (Guo et al., 2024; Senkova & Otani, 2021), others show no reactivity (Maxwell, 2025; Zhao, Li, et al., 2023), and still others reveal trends toward negative reactivity (Kaya & Mulligan, 2024). Even positive reactivity for related word pairs in cued-recall tests is less reliable than assumed: A recent analysis of 17 experiments with related and unrelated word pairs tested in cued-recall tests shows that positive reactivity for related word pairs and negative reactivity for unrelated word pairs was significant only in half of the analyzed experiments (Undorf et al., 2024). This complex pattern of consistent and inconsistent findings makes it difficult to establish the occurrence, size, and direction of immediate JOL reactivity. But what determines whether JOL reactivity is positive, negative, or does not occur at all? What are the boundary conditions of JOL reactivity? Despite ample research on JOL reactivity, the question of the boundary conditions of JOL reactivity remains unanswered. One reason for this is that multiple teams of researchers investigate similar questions independently, sometimes yielding conflicting results. For example, studies examining perceived differences in difficulty as a boundary condition for JOL reactivity have reached opposing conclusions. Janes and colleagues (2018) found that JOLs were only reactive when participants studied heterogeneous lists with related and unrelated word pairs that differed in difficulty, but not when the study list contained only related or unrelated word pairs. In contrast, Maxwell and Huff (2023) observed JOL reactivity for different types of word pairs irrespective of perceived differences in the difficulty of study lists. Because of these conflicting results, the role of perceived difficulty differences as a boundary condition of JOL reactivity remains unclear. Another reason that complicates the identification of boundary conditions of JOL reactivity is the methodological diversity represented in current research. For example, studies differ in whether they manipulate making JOLs within (e.g., Rivers et al., 2021) or between participants (e.g., Rivers, Dunlosky, et al., 2023), whether studying is experimenter- paced (e.g., Soderstrom et al., 2015) or self-paced (e.g., Mitchum et al., 2016), and when during learning JOLs are prompted (see Figure 1). Although the methodological decisions are generally well motivated within the context of the individual studies, the diversity of approaches impedes a clear identification of the boundary conditions of JOL reactivity. For example, it remains unclear whether the above-mentioned contradictory findings on perceived differences in difficulty (Janes et al., 2018; Maxwell & Huff, 2023) are due to differences in study time allocation (experimenter-paced study in Janes et al., 2018 versus self-paced study in Maxwell & Huff, 2023) or variations in the timing of the JOL prompt (JOLs prompted during item presentation after 4 s delay in Janes et al., 2018 versus JOLs prompted during item presentation simultaneously with item onset in Maxwell & Huff, 2023). The field's rapid development has provided critical insights into JOL reactivity. Still, it has also revealed a complex pattern of when and in what form JOL reactivity occurs. In particular, the tension between the homogeneity and the heterogeneity of the findings and the diverse methodological variations of the JOL reactivity paradigm create the need for a comprehensive synthesis of the phenomenon. This need is amplified by the fact that no up- to-date meta-analysis on immediate JOL reactivity is available. An early meta-analysis by Double and colleagues (2018) provided valuable guidance for the emerging research field. 3 Establishing JOL Reactivity 17 However, this meta-analysis only included a fraction of the studies examining JOL reactivity today and considered only a few moderating factors as boundary conditions. Therefore, it cannot meet the need for a comprehensive synthesis of the phenomenon, which is reflected in several attempts to synthesize JOL reactivity across smaller sets of studies. For example, Halamish and Undorf (2023), Kaya and Mulligan (2024), Myers and colleagues (2020), Undorf and colleagues (2024), Witherby and Tauber (2017), and Zheng and colleagues (2024) have each used small-scale meta-analyses to summarize the findings of their experiments. Recently, Zhao, Li, and colleagues (2023) reported a meta-analysis on immediate JOL reactivity for single words across their experiments and eight published experiments by others. To address this need for establishing JOL reactivity as a phenomenon, I present a comprehensive and up-to-date meta-analysis of reactivity of immediate JOLs. 3.2 On the Phenomenon of JOL Reactivity Ingendahl, F., Halamish, V., & Undorf, M. (in press). Do immediate judgments of learning alter memory performance? A meta-analytical review. Psychological Bulletin. https://doi.org/10.1037/bul0000487 The meta-analysis presented in Manuscript I (Ingendahl et al., in press) provides a comprehensive synthesis of immediate JOL reactivity. By synthesizing 344 effect sizes from 175 experiments, this work establishes the occurrence, size, direction, and moderating factors of JOL reactivity that delineate its boundary conditions. Further, the presented meta- analysis evaluates the principal theoretical accounts proposed to explain the mechanisms underlying JOL reactivity. Using a multilevel meta-analytical approach, we examined whether making JOLs alters memory performance in a positive or negative direction, synthesizing findings of relevant studies in the field. Furthermore, we tested whether JOL reactivity differs depending on experimental characteristics, thereby identifying boundary conditions of JOL reactivity. To address these research questions, we conducted a comprehensive meta-analysis on JOL reactivity based on published and unpublished experiments with young and middle- aged adults that include an intentional learning paradigm and a random manipulation of providing immediate JOLs during learning. In addition to the comprehensive meta-analysis, we performed a series of analyses on subsets of experiments with related and/or unrelated word pairs to provide a rigorous test of the principal theoretical accounts of JOL reactivity (i.e., the cue-strengthening account, the dual-task account, and the changed-goal account). The evaluation of the theoretical accounts was centered on the moderating factors of JOL reactivity for related and unrelated word pairs. Related and unrelated word pairs are particularly well-suited for testing the examined accounts, all of which were initially proposed to explain JOL reactivity within the paradigm of the relatedness effect. 3 Establishing JOL Reactivity 18 Table 1 Predictions of the Principal Theoretical Accounts for Moderators of JOL Reactivity for Related and Unrelated Word Pairs Theoretical Account Moderator Cue-strengthening account Dual-task account Changed-goal account Word-pair relatedness Positive reactivity for related word pairs & no reactivity for unrelated word pairs No reactivity for related word pairs & negative reactivity for unrelated word pairs Positive reactivity for related word pairs & negative reactivity for unrelated word pairs Memory test For related word pairs: Positive reactivity in cued-recall tests & no reactivity in other tests Study time allocation For unrelated word pairs: Negative reactivity is stronger for experimenter-paced study than for self- paced study Study time For unrelated word pairs: Negative reactivity is stronger for shorter study times JOL prompt onset For unrelated word pairs: Negative reactivity is stronger for JOLs prompted during study than for JOLs prompted immediately after study Data collection environment For unrelated word pairs: Negative reactivity is stronger in supervised environments than in unsupervised environments Number of item types Reactivitya for study lists with multiple item types & no reactivity for study lists with a single item type Type of JOL manipulation Reactivitya is stronger for between- participants manipulations than for within- participants manipulations Note. The font color indicates the type of predicted reactivity (red: positive reactivity, blue: negative reactivity, gray: no reactivity). a Prediction applies to positive and negative reactivity 19 We adopted a theory-driven approach to investigate the moderating factors that delineate the boundary conditions of JOL reactivity. We focused on experimental characteristics informing at least one of the principal theoretical accounts. Additionally, other moderators, such as the type of publication and language, were tested to explore additional potential sources of heterogeneity of JOL reactivity. Table 1 presents an overview of the tested theory-informing moderators and the corresponding predictions of the cue- strengthening account, the dual-task account, and the changed-goal account. All three theories suggest that JOL reactivity varies with word-pair relatedness. The cue- strengthening account predicts positive reactivity for related word pairs due to enhanced processing of their semantic association but no reactivity for unrelated pairs which lack such associations. In contrast, the dual-task account predicts no reactivity for related word pairs but negative reactivity for unrelated word pairs for which the primary demands of studying are high. The changed-goal account proposes positive reactivity for related word pairs and/or negative reactivity for unrelated word pairs due to a changed focus on studying items perceived as easier. Moreover, each theoretical account specifies distinct moderators of JOL reactivity. The cue-strengthening account predicts that positive reactivity for related word pairs should only occur in memory tests that are sensitive to the semantic association between cue and target words, such as cued-recall tests, but not in other memory tests, including free-recall and recognition tests. The dual-task account proposes that negative reactivity for unrelated word pairs becomes more pronounced as the primary study task becomes more demanding. Specifically, negative reactivity for unrelated word pairs should be more pronounced when the study time allocation is experimenter-paced instead of self-paced, when study time decreases, and when the onset of the JOL prompt is during studying (simultaneously with the item onset or after some delay during the item presentation) than when it is immediately after studying. The dual-task account also predicts that negative reactivity is stronger in supervised data collection environments (e.g., a laboratory) than in unsupervised data collection environments (e.g., online without supervision), because participants are usually more involved in experimental tasks in supervised data collection environments than in unsupervised environments (e.g., Brigden, 2024), and high involvement should intensify interference. The changed-goal account predicts that reactivity should only occur when the study list includes multiple item types, emphasizing the role of perceived differences in difficulty as a boundary condition of reactivity. Finally, it suggests that reactivity is stronger with between-participants than within-participants JOL manipulations, as a general shift in learning goals may carry over to items without JOLs. Our search strategy for identifying eligible experiments included a comprehensive literature search in databases (e.g., PsycInfo or ProQuest) and related reviews (Double et al., 2018; Double & Birney, 2019), as well as a call for unpublished and published data. Based on this initial search, we identified 802 records meeting the eligibility criteria on a surface level. After screening the titles and abstracts and an in-depth coding of the records, we included 175 experiments from 31 published and 18 unpublished records in the meta- analysis. Overall, the included experiments reported 344 effect sizes of JOL reactivity and were based on a total sample size of 15,079 participants. 3 Establishing JOL Reactivity 20 Figure 3 Forest Plot of Effect Sizes Included in the Meta-Analysis on Immediate JOL Reactivity Note. Error bars of the individual effect size estimates display 95% confidence intervals based on the variance of the effect size estimate. The color indicates the direction and the difference from 0, i.e., the point indicating no reactivity (red color: positive reactivity; blue color: negative reactivity; gray color: no reactivity/ confidence interval overlaps with 0). Synthesizing evidence from these 344 effect sizes, multilevel meta-analytical models revealed that making JOLs during learning led to a small, significant improvement of memory performance compared to studying without JOLs, g = 0.22, p < .001, 95% CI [0.17, 0.27]. Our meta-analysis thereby established JOL reactivity as a robust phenomenon. However, substantial heterogeneity emerged across effects, with large variability in the direction (positive or negative), size (ranging from |g| = 0.00 to |g| = 1.81), and significance of individual reactivity effects, as illustrated in Figure 3. Additionally, we identified evidence of publication bias, with unpublished experiments yielding systematically smaller reactivity ⌧2 ⌧1.5 ⌧1 ⌧0.5 0 0.5 1 1.5 2 2.5 Hedges' g (JOL ⌧ No JOL) ⌧2 ⌧1.5 ⌧1 ⌧0.5 0 0.5 1 1.5 2 2.5 Hedges' g (JOL ⌧ No JOL) Study 0.22 [0.17, 0.27]Robust 4⌧Level Model ( k = 344) 3 Establishing JOL Reactivity 21 effects than published experiments. Our meta-analysis thus indicated that while JOL reactivity is a robust cognitive phenomenon, its individual effects tend to be small and highly heterogeneous. This high heterogeneity implies that to fully establish JOL reactivity as a phenomenon, it is necessary to identify the specific conditions under which it occurs. We conducted a series of moderator analyses to examine the sources of this heterogeneity and to identify the experimental characteristics that delineate the boundary conditions of JOL reactivity. We identified four experimental characteristics moderating the occurrence and size of JOL reactivity overall. First, the type of study materials emerged as a key factor for the occurrence of reactivity. Positive reactivity was consistently observed for word pairs and single words but was absent for other materials such as pictures. Second, the type of memory test proved to be an important moderator. Positive reactivity reliably occurred in cued-recall tests and recognition tests. In contrast, no reactivity was observed on average for free-recall tests and other memory tests such as source-memory tests. Third, the onset time of the JOL prompt moderated the size of reactivity. Positive reactivity was stronger when the JOL prompt was presented simultaneously with the item onset than when the prompt was delayed by a few seconds or presented immediately after studying the item. Fourth, the language in which an experiment was conducted was associated with the occurrence of reactivity. On average, JOLs enhanced memory performance in experiments conducted in English or Chinese but not in those conducted in German or Hebrew. Other experimental characteristics such as the type of study time allocation or study time per item were not related to JOL reactivity overall. These findings delineate important experimental conditions under which JOL reactivity is most likely to occur and highlight the boundary conditions that shape its occurrence and size. Whereas these findings were grounded in a comprehensive analysis of all 344 reactivity effects, I now turn to a more focused examination of JOL reactivity for related and unrelated word pairs. This focused analysis provided a direct test of the theoretical accounts of JOL reactivity introduced earlier (see Table 1). Figure 4 provides an overview of the results from the theory-driven moderator analyses conducted to evaluate the cue- strengthening account, the dual-task account, and the changed-goal account. For the jointly proposed moderator of word-pair relatedness, we observed that the type of word pairs was a key moderator of the type of JOL reactivity (see Figure 4, Panel A). Making JOLs enhanced memory performance for related word pairs, consistent with the overall JOL reactivity effect identified in our meta-analysis. In contrast, making JOLs impaired memory performance for unrelated word pairs. This dichotomy of positive reactivity for related word pairs and negative reactivity for unrelated word pairs suggests that the nature of the study material relates not only to the size but also to the direction of JOL reactivity. Further, it shows that both positive and negative reactivity are robust findings of JOL reactivity – even though negative reactivity (g = –0.09) appeared to be smaller than positive reactivity (g = 0.44). Although each of the three theoretical accounts predicted that word-pair relatedness moderates JOL reactivity, not all are fully consistent with the observed reactivity pattern. The cue-strengthening account and the dual-task account are each only partially consistent with the observed results. Specifically, the cue- strengthening account aligns with the positive reactivity for related pairs but not with the negative reactivity for unrelated pairs, as it predicts no reactivity for these pairs. Conversely, 3 Establishing JOL Reactivity 22 the dual-task account is in line with the negative reactivity for unrelated word pairs, but it does not fit the positive reactivity for related pairs, where it predicts no reactivity. Only the changed-goal account fully aligns with the overall pattern of results, as it predicts that making JOLs should enhance memory for related word pairs and/or reduce memory for unrelated word pairs. Beyond word-pair relatedness, the results from the remaining theory-specific moderator analyses further tested the individual theoretical accounts. The cue-strengthening account predicted that the type of memory test moderates reactivity for related word pairs. Consistent with this prediction, we observed that reactivity for related word pairs varied by the type of memory test (see Figure 4, Panel B): Positive reactivity was present in cued- recall and recognition tests, but not in free-recall or other memory tests. This finding is partially consistent with the cue-strengthening account, as making JOLs enhanced memory performance in cued-recall tests, which are sensitive to semantic associations between cue and target words, but not in free-recall tests, which lack this sensitivity. However, the observation of positive reactivity in recognition tests is inconsistent with the cue- strengthening account, as recognition tests that focus on recognizing the target word only are generally considered insensitive to semantic associations between cue and target words. The dual-task account proposed that negative reactivity for unrelated word pairs should be stronger when the primary task of studying becomes more demanding. Contrary to this prediction, negative reactivity for JOL reactivity was not moderated by the type of study time allocation (see Figure 4, Panel C), the study time per item (see Figure 4, Panel D), or the onset of the JOL prompt (see Figure 4, Panel E). However, reactivity for unrelated word pairs varied depending on the environment of data collection (see Figure 4, Panel F), as predicted by the dual-task account: Negative reactivity for unrelated word pairs was stronger in supervised environments than in unsupervised environments, where no reactivity was observed. The changed-goal account predicted that JOLs only alter memory performance for related and unrelated word pairs when multiple item types are included in the study list. Contrary to this prediction, reactivity for neither related nor unrelated word pairs depended on the number of item types in the study list (see Figure 4, Panel G). However, the finding that JOL reactivity for related and unrelated word pairs was more pronounced in between- subjects designs than within-subjects designs (see Figure 4, Panel H) is fully consistent with the changed-goal account. Overall, the theory-driven moderator analyses reveal that none of the examined accounts fully explains the observed pattern of JOL reactivity for related and unrelated word pairs. 3 Establishing JOL Reactivity 23 Figure 4 Overview of Results from Theory-Focused Moderator Analyses Note. All panels except panel D present effect size estimates based on categorical moderator analyses. Panel D presents the results of a continuous moderator analysis. Error bars and confidence regions represent a 95% confidence interval. See Table 1 for more details on the moderator variables depicted in panels A to H and predictions of the theoretical accounts of JOL reactivity. Asterisks indicate the significance of the omnibus test of moderation: *** p < .001, ** p < .01, * p <. 05 A Word-Pair Relatedness *** B Memory Test *** E Onset of JOL Prompt F Data Collection Environment * C Study Time Allocation D Study Time G Number of Item Types H Type of JOL Manipulation **/* 3 Establishing JOL Reactivity 24 In summary, the presented meta-analysis addresses the key challenge of establishing JOL reactivity. By systematically synthesizing evidence from 344 effect sizes across 175 experiments, this work showed that soliciting immediate JOLs enhances future memory performance overall. However, the size of positive JOL reactivity appears to be smaller than the published literature suggests, as analyses reveal publication bias favoring larger reactivity effects. Nevertheless, the presented meta-analysis establishes JOL reactivity as a robust psychological phenomenon. At the same time, it indicated that JOL reactivity is not a uniform effect but a heterogeneous phenomenon with respect to its occurrence, size, and direction. Notably, the synthesis of reactivity for related and unrelated word pairs uncovered the dual nature of JOL reactivity. Whereas making JOLs robustly enhanced memory under some conditions, it impaired memory under others. Thus, our meta-analysis highlights that both positive and negative reactivity are valid findings of JOL reactivity, which should be acknowledged and considered in theoretical models and future research alike. Employing a series of moderator analyses, our meta-analysis further identified characteristics of the experimental design that shape JOL reactivity, thereby establishing evidence for the boundary conditions of the phenomenon overall. We tested JOL reactivity across various methodological variants of the reactivity paradigm by synthesizing a broad range of reactivity effects. In doing so, our meta-analysis helps to resolve inconsistencies in previous research, such as the contradictory evidence of whether including multiple item types within the study list is a requirement of JOL reactivity (see Janes et al., 2018; Maxwell & Huff, 2023). Our meta-analytical results revealed that JOL reactivity for related or unrelated word pairs does not depend on including multiple item types within the list, thus resolving the contradiction in previous research. Additionally, the moderator analyses highlight novel factors that have not received much attention, such as the language in which an experiment is conducted (see Undorf et al., 2024). Whereas multiple factors, including culture, sample characteristics, or linguistic aspects, may contribute to the moderation by language, this finding underscores the importance of considering learner characteristics in JOL reactivity. Moreover, our meta-analysis touches on another key challenge in the field – explaining JOL reactivity. Theory-driven moderator analyses testing the cue-strengthening account, the dual-task account, and the changed-goal account reveal a mixed picture. At least one moderator analysis showed the predicted findings and thus provided support for each of the accounts. However, none of the accounts was fully supported by the moderator analyses. For example, the predictions of the cue-strengthening account were only partially consistent with the results; one theory-consistent result of the dual-task account stood in contrast to one partially consistent and three inconsistent results. Similarly, only two out of three predictions of the changed-goal account were met by the results. Thus, none of the examined theoretical accounts explain the empirical findings of JOL reactivity for word pairs in their entirety, questioning the validity of these accounts for explaining JOL reactivity. Overall, JOL reactivity appears to be a multifaceted phenomenon that cannot be explained by a single theoretical account or mechanism. Our meta-analysis thus highlights that explaining the underlying mechanism(s) of JOL reactivity remains a key challenge. 25 4 Explaining JOL Reactivity As shown in Chapter 3, JOL reactivity is a robust phenomenon that occurs under diverse study and test conditions. The question, therefore, arises: Why are JOLs reactive, and what are the underlying mechanisms of this reactivity? Several theoretical accounts have been proposed to answer this question, yet explaining JOL reactivity remains an important task. 4.1 Explaining Reactivity: Empirical Inconsistencies and Theoretical Limitations Several factors contribute to the fact that our knowledge about the underlying mechanisms is still limited, despite repeated attempts to clarify them (e.g., Janes et al., 2018; Li, Shanks, et al., 2024; Myers et al., 2024). A major obstacle is the inconsistent empirical support for the proposed accounts. As introduced in Chapter 2 (see Section 2.2), both supporting and contradicting findings have been reported for the cue-strengthening account (e.g., Halamish & Undorf, 2023; but see, e.g., Myers et al., 2020), the dual-task account (e.g., Mitchum et al., 2016; but see, e.g., Maxwell & Huff, 2022), and the changed- goal account (e.g., Janes et al., 2018; but see, e.g., Li, Shanks, et al., 2024). This inconsistency is compounded by conflicting findings in studies examining the same or similar predictions (e.g., Chang & Brainerd, 2022; Janes et al., 2018; Maxwell & Huff, 2023; Soderstrom et al., 2015). Our meta-analysis in Chapter 3 corroborates this pattern: Focusing on related and unrelated word pairs – the materials for which these theories were initially developed – we found only mixed evidence for each account. Given their moderate explanatory power under presumably optimal conditions, it is questionable whether any of these accounts, in isolation, can fully explain JOL reactivity. The identification of underlying mechanisms is further hindered because existing theories rarely allow for a direct comparison. Only in rare cases, such as for backward-related word pairs or prestudy JOLs, the theoretical accounts can be tested against each other (e.g., Li, Shanks, et al., 2024; Maxwell & Huff, 2022). Our meta-analysis illustrates this strikingly: Among eight theory-driven moderators, the accounts could be directly tested against each other for only one moderator. Differences in the specificity of the accounts, restricting the number of testable predictions and their falsifiability, further complicate the evaluation and ranking of the accounts. This issue is aggravated by the fact that not all theoretical accounts consider positive and negative JOL reactivity. The cue-strengthening account and dual-task account explain only one reactivity type, rendering them a priori incompatible with the other type. Recent research (e.g., Halamish & Undorf, 2023; Li, Shanks, et al., 2024; Undorf et al., 2024; Witherby et al., 2023) and our meta-analysis alike emphasize that positive reactivity and negative reactivity are legitimate findings of JOL reactivity. Theoretical accounts should therefore incorporate mechanisms underlying positive and negative reactivity to explain JOL reactivity comprehensively. 4 Explaining JOL Reactivity 26 As illustrated above, the inconsistency of empirical support, the limited opportunity to test the accounts against each other, and the focus on only one type of reactivity highlight the need for new theoretical developments. A promising direction in this regard is the investigation of so-called proximal mechanisms (e.g., Quintana, 2023; also, see Rivers et al., 2021) that act as direct causes of JOL reactivity, rather than more distal, higher-level causes, such as enhanced processing of cues, interference, or goal changes. Examining proximal mechanisms enables taking a perspective on the phenomenon unrestricted by previous accounts but, at the same time, allows for a test of higher-level explanations. To address the task of explaining JOL reactivity, I present evidence for changes in learning strategy use as a proximal mechanism of JOL reactivity. 4.2 On Learning Strategies as a Proximal Mechanism of JOL Reactivity Learning strategies enable learners to control their learning behavior and thus impact their learning success. However, not all learning strategies are equally effective (see Dunlosky et al., 2013). Among the cognitive learning strategies, those that establish connections between the elements of the learning materials, such as imagery, are particularly effective. In contrast, the mere repetition of learning materials is a comparatively less effective cognitive strategy. Metacognition plays an important role in the use of learning strategies (e.g., Dent & Koenka, 2016) because it helps learners select appropriate strategies or flexibly adapt the use of strategies (for a review, see Metcalfe, 2009). Therefore, it is plausible that assessing metacognitive judgments, such as JOLs, may alter the use of learning strategies and subsequent memory performance. Whether changes in the use of learning strategies may underlie JOL reactivity has received little attention so far (for initial research, see Mitchum et al., 2016; Rivers et al., 2021). Therefore, it is unknown whether assessing JOLs during learning alters the use of learning strategies and whether changes in learning strategy use can explain JOL reactivity. The research presented in this section examines the role of learning strategies in JOL reactivity and thus contributes to resolving the task of explaining JOL reactivity. In doing so, we conceptualize learning strategies as a proximal mechanism of reactivity. That is, we examine whether changes in learning strategy use contribute directly and immediately to reactivity effects, without restricting our examination to any overarching mechanism that triggers these changes. This approach allows for a broader and more open analysis of learning strategies. At the same time, it permits testing existing theoretical accounts without focusing on one particular account by selecting the examined contributing factor. To comprehensively examine changes in the use of learning strategies as a proximal mechanism of reactivity, I present evidence from spontaneous (Manuscript II, Ingendahl & Undorf, 2025a) and instructed use (Manuscript III, Ingendahl & Undorf, 2025b) of learning strategies. 4 Explaining JOL Reactivity 27 4.2.1 Spontaneous Learning Strategy Use Ingendahl, F., & Undorf, M. (2025a). Changes in learning strategies contribute to negative reactivity of immediate judgments of learning. Journal of Experimental Psychology: Learning, Memory, and Cognition. Advance online publication. https://doi.org/10.1037/xlm0001475 The six experiments reported in Manuscript II (Ingendahl & Undorf, 2025a) examined whether making JOLs changes the spontaneous use of learning strategies for word pairs and whether these changes can explain positive reactivity for related word pairs and negative reactivity for unrelated word pairs. If changes in learning strategies explain JOL reactivity, spontaneous learning strategy use should differ by the requirement of making JOLs and mediate reactivity. Although no principal theoretical account explicitly identifies learning strategies as a mechanism for reactivity, by conceptualizing learning strategies as a proximal mechanism within the theoretical accounts, we derived predictions about differences in learning strategy use from the cue-strengthening account, the dual-task account, and the changed-goal account. The enhanced processing of relatedness when making JOLs, predicted by the cue- strengthening account, might be due to an enhanced spontaneous use of relational encoding strategies (i.e., mental imagery, sentence generation, and making connections) for related word pairs. The interference of making JOLs when studying unrelated word pairs, suggested by the dual-task account, could manifest in a reduced use of learning strategies for unrelated word pairs, for which interference should be particularly high. Finally, the prioritization of related word pairs over unrelated word pairs when making JOLs predicted by the changed- goal account could appear as an enhanced use of learning strategies for related word pairs and/or reduced use of learning strategies for unrelated word pairs. In all experiments, participants studied related and unrelated word pairs for 6 seconds and provided a JOL (“Likelihood of coming up with the second word when presented with the first word at the later test?”) on a scale from 0% to 100% after 2 s study duration (JOL condition) or did not provide JOLs during studying (no-JOL condition). Afterward, following a short filler task, participants completed a cued-recall test and an item-based assessment of their spontaneously used learning strategies (mental imagery, rote repetition, sentence generation, making connections without forming an image or a sentence, retrieval practice, other strategy, no strategy) for each studied word pair. We varied across experiments whether making JOLs was manipulated between (Experiments 1a, 1b, 2a, and 2b) or within participants (Experiments 3a and 3b), whether the semantic association of the related word pairs was weak (Experiments 1a and 1b) or strong (Experiments 2a, 2b, 3a, and 3b), and whether learning strategies were assessed in a separate block after the cued- recall test (Experiments 1a and 1b) or integrated within the cued-recall test (Experiments 2a, 2b, 3a, and 3b). Further, Experiments 1b, 2b, and 3b tested the robustness of the findings of Experiments 1a, 2a, and 3a under conditions of testing participants’ understanding of the assessed learning strategies. 4 Explaining JOL Reactivity 28 Figure 5 Overview of Results for (A) Cued-Recall Performance, (B) Use of Mental Imagery, and (C) Use of No Strategy in Experiments 1a to 2b in Manuscript II Note. Error bars present one standard error of the mean. Panels B and C present only experiments for which a difference in reported use in the respective learning strategy was observed. The four experiments (Experiment 1a: N = 130; Experiment 1b: N = 128; Experiment 2a: N = 130; Experiment 2b: N = 128) that manipulated the requirement to provide JOLs between participants revealed a clear and consistent pattern of results. As illustrated in Figure 5 (Panel A), the JOL group showed enhanced memory performance for related word pairs but impaired memory performance for unrelated word pairs compared to the no-JOL group. The observed reactivity effects provided ideal conditions for examining the contribution of spontaneous learning strategies to positive and negative reactivity. A B C 4 Explaining JOL Reactivity 29 Critically, the reported use of learning strategies showed clear differences between groups overall, regardless of the strength of the semantic association of the related word pairs and the time at which learning strategies were assessed. Although the differences in learning strategies varied slightly between experiments, the results revealed a clear pattern for two learning strategies across several experiments: In three out of four experiments each, making JOLs was associated with a less frequent use of mental imagery (see Figure 5, Panel B) and a more frequent use of no learning strategy for studying unrelated word pairs (see Figure 5, Panel C). The observed differences in learning strategies for unrelated word pairs suggest that learning strategies may contribute to negative reactivity for unrelated word pairs, as both less frequent use of mental imagery and more frequent use of no strategy are linked to worse cued-recall performance. Multilevel moderated mediation analyses tested the proposed relation between making JOLs, spontaneous learning strategy use, and cued-recall performance for related and unrelated word pairs (see Figure 6). Specifically, we tested whether the direct effect of making JOLs on cued-recall performance is mediated by the indirect effect of making JOLs by spontaneous learning strategy use on cued-recall performance. The analyses confirmed the suggested role of learning strategy use in negative reactivity for unrelated word pairs. Using mental imagery and/or no learning strategy consistently fully mediated negative reactivity for unrelated word pairs but did not mediate positive reactivity for related word pairs. These findings showed that negative reactivity for unrelated word pairs is due to changes in the spontaneous use of learning strategies. In contrast, learning strategies appeared to play no or, at best, a minor role in positive reactivity for related word pairs. Only one experiment, Experiment 1b, found evidence for a partial mediation of positive reactivity by learning strategies, in the form of making connections. 4 Explaining JOL Reactivity 30 Figure 6 Estimates of (A) Direct Effects and (B) Indirect Effects of the Multilevel Moderated Mediation Models in Experiments 1a to 2b in Manuscript II Note. Estimates represent standardized regression weights. Error bars represent 95% confidence intervals. *** p < .001, ** p < .01, * p < .05 Experiments 1a to 2b provided consistent evidence for a contribution of learning strategies to (negative) reactivity. These findings contrast with previous research, finding no differences in learning strategy use by making JOLs (Mitchum et al., 2016; Rivers et al., 2021). This discrepancy can be partially explained by methodological differences in the assessment of learning strategies, such as using post-experimental questionnaires (Mitchum et al., 2016, Experiment 1) instead of an item-based assessment. However, our findings also differ from a recent experiment (Rivers et al., 2021, Experiment 3) using a similar methodological approach. One explanation for this might be that previous work clustered learning strategies as effective and ineffective. Our results, however, showed that making JOLs was not associated with differences in all effective or ineffective learning strategies, but instead with differences in specific learning strategies. Thus, clustering learning strategies as (in)effective may have masked learning strategy differences. Another explanation could lie in the type of JOL manipulation (between-subjects in our Experiments 1a to 2b versus within-subjects in Experiment 3 by Rivers et al., 2021). Changes in the use of learning strategies due to JOLs might generalize to studying without JOLs when manipulating JOLs within participants and thus could obscure differences in learning Group Cued Recall Learning Strategy Word-Pair Relatedness Word-Pair Relatedness A B *** * *** ** ** * *** ** * * Group Cued Recall Learning Strategy Word-Pair Relatedness Word-Pair Relatedness 4 Explaining JOL Reactivity 31 strategies. Moreover, negative reactivity has not yet been observed in within-subjects paradigms of JOL reactivity (e.g., Rivers et al., 2021) – a finding consistent with our meta- analysis (see Chapter 3). At the same time, our findings in Experiments 1a to 2b showed that learning strategies contributed specifically to negative reactivity. This suggests that the absence of negative reactivity in within-subjects paradigms may contribute to the absence of learning strategy differences. We tested these explanations in Experiments 3a and 3b. Experiments 3a (N = 68) and 3b (N = 64) examined whether making JOLs changes spontaneously used learning strategies when participants make JOLs for half of the studied word pairs and study the remaining half without JOLs. The results fully replicated the previous findings with within-subjects paradigms (Rivers et al., 2021): Making JOLs enhanced memory performance for related word pairs, but did not affect cued-recall performance for unrelated word pairs. Crucially, learning strategies did not differ between JOL and no-JOL trials in the absence of negative reactivity, thus supporting the conclusion that changes in the spontaneous use of learning strategies contribute to negative reactivity for unrelated word pairs.