CHAPTER 11 Metacognition and

Asher Koriat

Abstract reviews research addressing these questions, emphasizing its implications for issues con- The study of metacognition can shedlight on cerning consciousness; in particular, the gen- some fundamental issues about conscious- esis of subjective experience, the function of ness and its role in behavior. Metacognition self-reflective consciousness, and the cause- research concerns the processes by which and-effect relation between subjective expe- people self-reflect on their own cognitive rience and behavior. and processes (monitoring) and how they put their rnetaknowledge to use in regulating their and Introduction behavior (control). Experimental research on metacognition has addressed the follow- There has been a surge of interest in ing questions. First, what are the bases of metacognitive processes in recent years, metacognitive judgments that people make with the topic of metacognition pulling in monitoring their , remembering, under one roof researchers from tradition- and performance? Second, how valid are ally disparate areas of investigation. These such judgments and what are the factors areas include memory research (Kelley that affect the correspondence between sub- & Jacoby, 1998; Metcalfe & Shimamura, jective and objective indexes of knowing? 1994; Nelson & Narens, 1990; Reder, 1996), Third, what are the processes that underlie developmental psychology (Schneider & the accuracy and inaccuracy of metacogni- Pressley, igq7), social psychology (Bless tive judgments? Fourth, how does the out- & Forgas, 2000; Jost, Kruglanski, & Nel- put of metacognitive monitoring contribute son, 1998; Schwarz, 2004), judgment and to the strategic regulation of learning and decision making (Gilovich, Griffin, & remembering? Finally, how do the metacog- Kahneman, 2002; Winman & Juslin, 2005), nitive processes of monitoring and con- neuropsychology (Shimamura, zooo), trol affect actual performance? This chapter forensic psychology (e.g., Pansky, Koriat, ZYO THO C,\MRRIDGb HANDBOOK OF CONSCIOUCNFCS

& Goldsmith, 2005; Perfect, zooz), edu- trol). The object level, in contrast, has no cational psychology (Hacker, Dunlosky, & control over the level and no access Graesser, 1998), and and to it. For example, the study of new mate- creativity (Davidson & Sternberg, 1998; rial involves a variety of basic, object-level Metcalfe, 1998a). The establishment of operations, such as text processing, compre- metacognition as a topic of interest in its hending, rehearsing, and so on. At the same own right is already producing synergy time, metacognitive processes are engaged among different areas of investigation con- in planning how to study, in devising and cerned with monitoring and self-regulation implementing learning strategies, in moni- (e.g. Fernandez-Duque, Baird, & Posner, toring the course and success of object-level 2000). Furthermore, because some of the processes, in modifying them when neces- questions discussed touch upon tradition- sary, and in orchestrating their operation. In ally ostracized issues in psychology, such the course of studying new material, learners as the issues of consciousness and free are assumed to monitor their degree of com- will (see Nelson, 1996)~a lively debate prehension online and then decide whether has been going on between metacognitive to go over the studied material once again, researchers and philosophers (see Nelson how to allocate time and effort to different & Rey, 2000). In fact, it appears that the segments, and when to end studying. increased interest in metacognition research We should note, however, that the derives in part from the feeling that perhaps distinction between cognitive and metacog- this research can bring us closer to dealing nitive processes is not sharp because the with (certainly not resolving) some of the same type of cognitive operation may occur metatheoretical issues that have been the at the object level or at the meta level, and province of philosophers of thc mind. in some cases it is unclear to which level a particular operation belongs (Brown, 1987). Definition Research Traditions Metacognition concerns the study of what people know about in general, Historically, there have been two main and about their own cognitive and mem- lines of research on metacognition that ory processes, in particular, and how they proceeded almost independently of each put that knowledge to use in regulating their other, one within developmental psychology information processing and behavior. Flavell and the other within experimental memory (1~71)introduced the term "," research. The work within developmen- which concerns specifically the monitoring tal psychology was spurred by Flavell (see and control of one's learning and remem- Flavell, 1979; Flavell& Wellman, 1977), who bering. Metamemory is the most researched argued for the critical role that metacog- area in metacognition and is the focus of this nitive processes play in the development chapter. of memory functioning (see Flavell, 1999). Nelson and Narens (1990) proposed Within memory research, the study of a conceptual framework that has been metacognition was pioneered by Hart's adopted by most researchers. According to (1965) studies on the feeling-of-knowing them, cognitive processes may be divided (FOK), and Brown and McNeill's (1966) into those that occur at the object level work on the tip-of-the-tongue (TOT). and those that occur at the meta level: The There is a difference in goals and object level includes the basic operations methodological styles between these two traditionally subsumed undcr the rubric of research trahtions. The basic assumption information processing - encoding, rehears- among developmental students of mctacog- ing, retrieving, and so on. The meta level is nition is that learning and memory per- assumcd to oversee object-level operations formance depend heavily on monitoring (monitoring) and return signals to regulate and regulatory proficiency. This assump- them actively in a top-down fashion (con- tion has resulted in attempts to specify the METACOGNI'rION AND CONSCIOUSNESS 9l

components of metacognitive abilities, to In addition to the developmental and trace their development with age, and to the experimental-memory lines of research, examine their contribution to memory func- there has been considerable work on tioning. Hence a great deal of the work metacognition in the areas of social psy- is descriptive and correlational (Schneider, chology and judgment and decision mak- 1985). The focus on age differences and indi- ing. Social psychologists have long been vidual differences in metacognitive skills has concerned with questions about metacog- also engendered interest in specifying "defi- nition, although their work has not been ciencies" that are characteristic of children at explicitly defined as metacognitive (see Jost different ages and in devising ways to rem- et al., 1998). In particular, social psychol- edy them. This work has expanded into the ogists share the basic tenets of metacog- educational domain: Because of the increas- nitive research (see below) regarding the ing of the critical contribution of importance of subjective feelings and beliefs, metacognition to successful learning (Paris & as well as the role of top-down regula- Winograd, ~gqo),educational programs have tion of behavior. In recent years social psy- been developed (see Scheid, 1993) designed chologists have been addressing questions to make the learning process more "metacog- that are at the heart of current research in nitive." Several authors have stressed specif- metacognition (e.g., Winkielman, Schwarz, ically the importance of metacognition to Fazendeiro, & Reber, 2003; Yzerbyt, I,ories, transfer of learning (see De Corte, 2003). & Dardenne, 1998; see Metcalfe, 1998b). The conception of metacognition by Within the area of judgment and decision developmental psychologists is more com- malung, a great deal of the work concern- prehensive than that underlying much of ing the calibration of probability judgments the experimental work on metacognition. (Fischhoff; 1975; Lichtenstein, Fischhoff, & It includes a focus on what children know Phillips, 1982; Winman & Juslin, 2005) is about the functioning of memory and par- directly relevant to the issues raised in ticularly about one's own memory capac- metacognition. ities and limitations. Developmental work has also placed heavy emphasis on strate- Research Questions gies of learning and remembering (Bjorklund & Douglas, 1997; Brown, 1987; Pressley, This chapter emphasizes the work on Borkowslu, & Schneider, 1987). In addition, metacognition within the area of adult many of the issues addressed in the area of memory research. It is organized primarily theory of mind (Perner & Lang, 1999) con- around the five main questions that have cern metacognitive processes. These issues been addressed in experimental research on are, perhaps, particularly important for the metamemory. First, what are the bases of understanding of children's cognition. metacognitive judgments; that is, how do In contrast, the experimental-cognitive we know that we know (e.g., Koriat & Levy- study of metacognition has been driven Sadot, lqqq)? Second, how valid are subjec- more by an attempt to clarify basic questions tive intuitions about one's own knowledge; about the mechanisms underlying monitor- that is, how accurate are metacognitive ing and control processes in adult mem- judgments, and what are the factors that ory (for reviews, see Koriat & Levy-Sadot, affect their accuracy (e.g., Schwartz & 1999; Nelson & Narens, 1990; Schwartz, Metcalfe, lyyq)? Third, what are the pro- iqyq). This attempt has led to the emergence cesses underlying the accuracy and inaccu- of several theoretical ideas as well as spe- racy of metacognitive judgments? In par- cific experimental paradigms for examining ticular, what are the processes that lead the monitoring and control processes that to illusions of knowing and to dissociations occur during learning, during the attempt between knowing and the feeling of know- to retrieve information from memory, and ing (e.g., Benjamin & Bjork, 19qG; Koriat, following the retrieval of candidate answers i995)? Fourth, what are the processes under- (e.g., Metcalfe, 2000; Schwartz, zooz). lying the strategic regulation of learning 2 QZ THE CAMBIIIDGE IIANDBOOK OF CONSCIOUSNESS

and remembering? In particular, how does the importance of subjective processes and the output of monitoring affect control top-down (see Koriat, processes (e.g., Barnes, Nelson, Dunlosky, 2ooob). The study of metacognition is gen- Mazzoni, & Narens, 1999; Son & Metcalfe, erally predicated on a view of the person as ~OOO)?Finally, how do the metacognitive an active organism that has at its disposal an processes of monitoring and control affect arsenal of cognitive operations that can be actual memory performance (e.g., Koriat applied at will toward the achievement of & Goldsmith, 1996a; Metcalfe & Kornell, various goals. The strategic choice and reg- 2003)? ulation of these operations are assumed to Although these questions focus on be guided in part by the person's subjective relatively circumscribed processes of mem- beliefs and subjective feelings. ory and metamemory, they touch upon Embodied in this view are two metathe- some of the issues that are at the heart oretical assumptions (see Koriat, zoo2 j. of the notions of consciousness and self- The first concerns agency - the assumption consciousness. Thus, the study of the sub- that self-controlled processes have measur- jective monitoring of knowledge addresses a able effects on behavior. Although most defining property of consciousness, because researchers would acknowledge that many consciousness implies not only that we cognitive processes, including some that are know something but also that we know subsumed under the rubric of executive that wc know it. Thus, consciousness binds function, occur outside of cpnsciousness, together knowledge and metaknowledge there is also a recognition that the per- (Koriat, 2ooob). This idea is implied, for son is not a mere medium thi.ough which example, in Rosenthal's (2000) "higher- information flows. Rather, peo order " (HOT) philosophical theory freedom and flexibility in of consciousness: A "lower-order" men- their cognitive processes durindlearning and tal state is conscious by virtue of there remembering. Furthermore, id is assumed being another, higher-order mental state that that such self-regulation proc4sses deserve makes one conscious that one is in the lower- to be studied not only because they can have order state (see Chapter 3). Clearly, the considerable effects on performance but also subjective feelings that accompany cognitive because they are of interest in their own processes constitute an essential ingredient right. of conscious awareness. Rather than taking This assumption presents a dilemma these feelings (and their validity) at their face for experimental researchers because self- value, the study of metacognition attempts controlled processes have been tradition- to uncover the processes that shape subjec- ally assumed to conflict with the desire tive feelings and contribute to their validity of experimenters to exercise strict experi- or to their illusory character. Furthermore, mental control. Of course, there are many the study of monitoring-based control has studies in which learning and remembering implications for the question of the function strategies have been manipulated (through of conscious awareness, and for the benefits instructions) and their effects investigated and perils in using one's own intuitive feel- (e.g., Craik & Lockhart, 1972). Unlike such ings and subjective experience as a guide to experimenter-induced strategies, however, judgments and behavior. self-initiated strategies generally have been seen as a nuisance factor that should be avoided or neutralized. For example, labo- Basic Assumptions about Agency ratory studies typically use a fixed-rate pre- and Consciousness sentation of items rather than a self-paced The increased interest in metacognition presentation (see Nelson & Leonesio, 1988). seems to reflect a general shift from Also, in measuring memory performance, the stimulus-driven, behavioristic view of sometimes forced-choice tests are peferred the person to a view that acknowledges over free-report tests to avoid having to AND COUSCIOTTSNtSS "93

deal with differences in "guessing," or else Winhelman & Berridge, 2004). However, some correction for guessing procedure is by and large, much of the experimental used to achieve a pure measure of "true" research in metacognition is predicated on memory (see Koriat & Goldsmith, 1qq6a; the tacit assumption that the metacognitive els son & Narens, 1994). Needless to say, processes studied entail conscious control. people in everyday life have great freedom in Nonetheless, although the term "metacog- their memory processes, and the nition" is generally understood as involving challenge is to find ways to bring these self- conscious awareness, it should be acknowl- controlled metacognitive processes into the edged that monitoring and control pro- laboratory (Koriat, 2oooa; Koriat & Gold- cesses can also occur unconsciously (Spehn smith, iqqha). & Reder, 2000). The second assumption concerns the role I now review some of the experimental of self-reflective, subjective experience in work on metamemory, focusing on research guiding controlled processes. This is, of that may have some bearing on general course, a debatable issue. It is one thing to questions about phenomenal experience and equate col~trolledprocesses ~vithconscious conscious control. processes (e.g., Posner & Snyder, 1975); it is another to assume that subjective expe- Experimental Pavadigms in the Study rience plays a causal role in behavior. Stu- of Online Metamemory dents of metacognition not only place a heavy emphasis on subjective experience A variety of metacognitive judgments have but also assume that subjective feelings, such been studied in recent years that ought to be as the feeling of knowing, are not mere included under the umbrella of metacogni- epiphenomena, hut actually exert a causal tion (Metcalfe, 2000). Among these are ease- role on information processing and behavior of-learning judgnlents (Leonesio & Nelson, (Koriat, 2ooob; Nelson, 1996). 1990)~judgments of comprehension (Mak A similar growing emphasis on the role of & McCuire, 2002)~remember/know judg- subjective feelings in guiding judgments and ments (Gardiner, & Richardson-Klavehn, behavior can be seen in social-psychological ~OOO),output monitoring (Koriat, Ben-Zur, research (Schwarz Pr Clore, 2003) and in & Sheffer, 1y88), olfactory metacognition decision malung (Slovic, Finucane, Peters, & (J6nsson & Olsson, loo?), and source moni- MacGregor, 2002). Also, the work on mem- toring (Johnson, 19~7).However, the bulk of ory distortions and false brings to the experimental work has concerned three the fore the contribution of phenomenolog- types of judgments. ical aspects of remembering to source mon- First are judgments of learning (JOLs) itoring and reality monitoring (see Kelley & elicited following the study of each item. Jacohy, 1998; Koriat, Goldsmith, & Pansky, For example, after studying each paired- 2000; Mitchell &Johnson, 2000). associate in a list, participants are asked to It should be stressed, however, that not all assess the likelihood that they will be able to students of metacognition subscribe to the the target word in response to the cue assumptions discussed above. In particular, in a future test. These item-by-item judg- Reder (1987) has argued that a great deal of ments are then compared to the actual recall strategy selection occurs without conscious performance. deliberation or awareness of the factors that Second are FOK judgments elicited influence one's choice. Of course, there is following blocked recall. In the Recall- little doubt that many monitoring and con- Judgment-Recognition (WR) paradigm trol processes occur without consciousness introduced by Hart (19651, participants (Kentridge & Heywood, ~OOO),SO the ques- are required to recall items from memory tion becomes one of terminology, like the [typically, the answers to general knowledge question whether feelings must be conscious questions). When they fail to retrieve or can also be unconscious (Clore, 1994; the answer, they are asked to make FOK 94 THE CAMBRIDGE HANDBOOK OF CONSCIOUSNESS judgments regarding the likelihood that and can even sense its imminent emergence they would be able to select the correct into consciousness. What is peculiar about answer from among several distractors in a this experience is the discrepancy between forced-choice test to be administered later. subjective and objective knowing. So how The validity of FOK judgments is then can people monitor the presence of infor- evaluated by the correspondence between mation in memory despite their failure to these judgments and performance on the retrieve it? In reviewing the verbal learning recognition test. Finally, after retrieving literature more than 30 years ago, Tulving an answer from memory or after selecting and Madigan (1970)~in fact, argued that one an answer, the subjective confidence in of the truly unique characteristics of the correctness of that answer is elicited, memory is its knowledge of its own knowl- typically in the form of a probability judg- edge. They proposed that genuine progress ment reflecting the assessed likelihood that in memory research depends on understand- the answer is correct. Whereas JOLs and ing how the memory system not only can FOK judgments are prospective, involving poduce a learned response or retrieve an predictions of future memory prformance, image but also can estimate rather accurately confidence judgments are retrospective, the likelihood of its success in doing it. A involving assessments about a memory that great deal of research conducted since 1970 has been produced. has addressed this question. Many different variations of these general paradigms have been explored, including View variations in the type of memory stud- The Direct-Access ied (semantic, episodic, autobiographical, A simple answer to the question about eyewitness-type events, etc.), the format of the basis of feelings of knowing is pro- the memory test (free recall, cued recall, vided by the direct-access view according to forced-choice recognition, etc.), and the which people have direct access to mem- particular judgments elicited (item-by-item ory traces both during learning and during judgments or global judgments, using a prob- remembering and can base their metacog- ability or a rating scale, etc.). nitive judgments on detecting the presence and/or the strength of these traces. For example, in the case of JOLs elicited dur- ing study, it may be proposed that learners How Do We Know That We Know? can detect directly the memory trace that is The Bases of Metacognitive Judgments formed following learning and can also mon- itor online the increase in trace strength that As we see later, metacognitive judgments occurs in the course of study as more time is are accurate by and large. JOLs made for spent studying an item (e.g., Cohen, Sandler, different items during study are generally & Keglevich, 1~~1).Of course, to the extent predictive of the accuracy of recalling these that learners can do so, they can also decide items at test. FOK judgments elicited fol- to stop studying (under self-peed condi- lowing blocked recall predict the likelihood tions) when trace strength has reached a of recalling or recognizing the elusive target desirable value (Dunlosky & Hertzog, 1998). at some later time, and subjective confidence A direct-access account has also been in the correctness of an answer is typically advanced by Hart (1965) with regard to diagnostic of the accuracy of that answer. FOK. Hart proposed that FOK judgments Thus, the first question that emerges is, How represent the output of an internal moni- do we know that we know? tor that can survey the contents of This question emerges most sharply with ory and can determine whether the regard to the tip-of-the-tongue [TOT) state, of a solicited memory target exists in stor in which we fail to recall a word or a name, Thus, the feeling associated with the TO and yet we are convinced that we know it state may be assumed to stem from direc METACOGNITION AiUD CONSCIOUSNESS 295 privileged access to the memory trace of degree of validity in predicting objective the elusive target (see also Burke, MacKay, memory performance (Benjamin & Bjork, Worthley, & Wade, 1991; Yaniv & Meyer, 1996). To the extent that such indeed is 1~87).Hart stressed the functional value of the case, then the accuracy of metacogni- having such a monltor, given the general fal- tive judgments is not guaranteed, but should liblllty of the memory system. If the mon- depend on the validity of the cues on which itor "signals that an item is not in storage, it rests. then the system will not continue to expend Inferential, cue-utilization accounts gen- useless effort and time at retrieval; instead, erally distinguish between information- input can he sought that will put the item based (or theory-based) and experience- into storage" (Hart, 1965; p. 214). based metacognitive judgments (see Kelley Direct-access (or trace-access) accounts, & Jacoby, 1996a; Koriat & Levy-Sadot, which assume that monitoring involves a 1999; Matvey, Dunlosky, & Guttentag, 2001; direct readout of information that appears Strack, 1992). This distinction parallels a dis- in a ready-made format, have two merits. tinction between two modes of thought that The first is that they can explain not only has been proposed in other domains (see the basis of JOLs and FOK judgments but Kahneman, 2003, and see further below). also their accuracy. Clearly, ifJOLs are based Thus, it is assumed that metacognitive judg- on accessing the strength of the memory ments may be based either on a deliber- trace that is formed following learning, then ate use of beliefs and memories to reach an they ought to be predictive of future recall, educated guess about one's competence and which is also assumed to depend on memory , or on the application of heuristics strength. Similarly, if FOK judgments moni- that result in a sheer subjective feeling. tor the presence of the memory trace of the unrecalled item, they should be expected to Theory-Based Monitoring predict the future recognition or recall of that item. Consider first theory-based metacognitive The second merit is that they would judgments. Developmental students of cog- seem to capture the phenomenal quality of nition placed a great deal of empha- metacognitive feelings: the subjective feel- sis on what Flavell called "metacognitive ing, such as that which accompanies the knowledge;" that is, on children's beliefs tip-of-the-tongue state, that one monitors and intuitions about their own memory directly the presence of the elusive target in capacities and limitations and about the memory and its emergence into conscious- factors that contribute to memory perfor- ness (James, 1890). In fact, metacognitive mance (Brown, 1987) Such beliefs have feelings are associated with a sense of self- been found to affect the choice of learning evidence, which gives the impression that strategies, as well as people's pedictions of people are in direct contact with the con- their own memory performance (see Flavell, tents of their memories and that their intro- 1999; Schneider & Pressley, 1997). spections are inherently accurate. In contrast, the experimental research on adult metacognition contains only scattered references to the possible contribution of The Cue-Utilization View of theories and beliefs to metacognitive judg- Metacognitive Judgments ments. For example, in discussing the bases Although the direct-access view has not of JOLs, Koriat (1997) proposed to distin- been entirely abandoned (see Burke et al., guish between two classes of cues for theory- 1991; Metcalfe, zooo), an alternative view based online JOLs, intrinsic and extrinsic. has been gaining impetus in recent years. The former includes cues pertaining to the According to this view, metacognitive judg- perceived a priori difficulty of the stucLed ments are inferential in origin, based on a items (e.g., Rabinowitz, Ackerman, Craik, variety of cues and heuristics that have some & Hinchley, 1982). Such cues seem to affect 2 96 THI: CAMBRIDGE HANI)BOOK OF CONSCIOUSNESS

JOLs, particularly during the first study trial, the same test; those who believed that the as suggested by the observation that norma- test measured abstract reasoning ability (on tive ratings of ease of learning are predic- which they had rated themselves highly) tive both of JOLs and of recall of different estimated that they had achieved higher items (e.g., Koriat, 1997; Leonesio &Nelson, scores than did those who thought that 1990; Underwood, 1966). The second class they had taken a computer programming includes extrinsic factors that pertain either test. This was so despite the fact that the to the conditions of learning (e.g., number two groups did not differ in their actual of times an item has been presented, pre- performance. sentation time, etc., Mazzoni, Cornoldi, & Another finding that points to the effects Marchitelli, 1990; Zechmeister & Shaugh- of one's a priori beliefs comes from stud- nessy, 1980) or to the encoding operations ies of the relationship between confidence applied by the learner (e.g., level of pro- and accuracy. People's confidence in their cessing, interactive imagery, etc.; Regg, Vin- responses is generally predictive of the accu- ski, Frankovich, & Holgate, 1991; Matvey racy of these responses in the case of general et al., 2001; Rabinowitz et al., 1982; Shaw knowledge questions but not in the case of & Craik, 1989). For example, participants' eyewitness memory (Perfect, 2002). Perfect JOLs seem to draw on the belief that gen- (2004) provided evidence that this occurs erating a word is better for memory than because people's confidencr is based in part it (Begg et al., 1991; Matvey et al., on their preconceptions about their abilities. 2001). Koriat (1997) proposed that JOLs are Such preconceptions are generally valid in comparative in nature. Hence, they should the case of knowledge questions, for be more sensitive to intrinsic cues pertaining which people have had considerable feed- to the relative recallability of different items back and hence know their relative standing. within a list than to factors that affect over- Such is not the case with eyewitness mem- all performance (see Begg, Duft, I,alonde, ory, for which they lack knowledge about Melnick, & Sanvito, 1989; Carroll, Nelson, & how good they are and, by implication, how Kirwan, 1997; Shaw & Craik, 1989). Indeed, confident they ought to be. Thus, people's he obtained evidence indicating that, in confidence in their performance seems to be making JOLs, the effects of extrinsic factors based in part on their preconceived beliefs are discounted relative to those of intrinsic about their own competence in the domain factors that differentiate between different of knowledge tested. items within a list. Evidence for the effects of beliefs and Another major determinant of people's theories also comes from studies of correc- metacognitive judgments is their perceived tion processes in judgment. People often self-efficacy (Bandura, 1977). In fact, peo- base their judgments directly on their sub- ple's preconceived notions about their slulls jective feelings (see Schwarz & Clore, 1996; in specific domains pedict their assessment Slovic et al., 2002). However, when they of how well they did on a particular task. realize that their subjective experience has For example, when students are asked to been contaminated by irrelevant factors, tell how well they have done on an exam, they may try to correct their judgments they tcnd to overestimate greatly their per- according to their beliefs about how these formance on the test, and this bias derives judgments had been affected by the irrel- in part from the tendency of people to evant factors (Strack, 1992). For example, base their retrospective assessments on their in the study of Schwarz, Bless, Strack, preconceived, inflated beliefs about their Klumpp, Rittenauer-Schatka, and Simons sk~llsin the domain tested, rather than on (lq91), participants who were asked to their specific experience with taking the recall many past episodes demonstrating test (Dunning, Johnson, Ehrlinger, & Kruger, self-assertiveness reported lower self-ratings 2003). In a study by Ehrlinger and Dun- of assertiveness than those who were asked ning (zooj), two groups of participants took to recall a few such episodes, MhTALOLNIIION AND CONSCIOUSNk55 2Y7

because of the greater difficulty experienced tors. Brainrrd, Reyna, Wright, and Mojardin in recalling many episodes. However, when (2003) also discussed a process termed "rec- led to believe that the experienced difficulty ollection rejection" in which a distractor that had been caused by background music, par- is consistent with the gist of a presented item ticipants relied more heavily on the retrieved may be rejected when the verbatim trace of content, reporting higher ratings under the that item is accessed. However, they argued many-episodes condition than under the that this process can occur automatically, few-episodes condition. These and other outside conscious awareness. findings suggest that the correction pro- The evidence reviewed thus far supports cess is guided by the person's beliefs about the idea that metacognitive judgments may the factors that make subjective experi- be based on one's beliefs and theories. For ence an unrepresentative basis for judgment. example, the subjective confidence in the Although most researchers assume that the correctness of one's memory product (e.g., a correction process requires some degree of selected answer in a quiz) can be based on a awareness (see Gilbert, zooz), others sug- logical, analytic process in which one evalu- gest that it may also occur unconsciously ates and weighs the pros and cons (Cigeren- (Oppenheimer, 2004). zer, Hoffrage & Kleinbolting, 1991; Koriat, More recent work in (see Lichtenstein, & Fischhoff, 1980). FOK judg- Schwarz, 2004) suggests that the conclu- ments, too, may draw on theories or beliefs sions that people draw from their metacog- resulting in an educated guess about the like- nitive experience, such as the experience lihood of retrieving or recognizing an elusive of fluent processing, depend on the naive word in the future (Costermans, Lories, & theory that they bring to bear. Further- Ansay, 1~92).Such judgments may not be more, people can be induced to adopt qualitatively lfferent from many predic- opposite theories about the implications of tions that people make in everyday life processing fluency, and these theories modu- late experience-based judgments. These sug- Experience-Based Monitoring gestions deserve exploration with regard to judgments of one's own knowledge. Expcrience-based metacognitive judgments, Another line of evidence comes from in contrast, are assumed to entail a quali- studies that examined how people deter- tatively different process from that under- mine that a certain cvent did not happen. lying theory-based judgments. Consider, for Strack and Bless (199~)proposed that deci- example, the TO1 experience. The strong sions of nonoccurrence may be based on conviction that one knows the elusive target a metacognitive strategy that is used when is based on a sheer subjective feeling. That rememberers fail to retrieve any feature of feeling, however, appears to be the prod- a target evcnt that they have judged to uct of an inferential process that involves be highly memorable. In contrast, in the the application of nonanalytic heuristics (see absence of a clear recollection of a non- Jacoby & Brooks, 1984; Kelley & Jacoby, memorable event, people may infer that 1qy6a; Koriat & Levy-Sadot, 199y) that oper- the cvent had actually occurred (but had ate below full consciousness and give rise to a been forgotten). Indeed, non-occurrence sheer subjective experience. Indeed, the idea decisions are made with strong confidence that subjective experience can be influenced for events that would be expected to be and shaped by unconscious inferential pro- remembered (e.g., one's name, a salient cesses has received support in the work of item, etc.; Brown, Lewis, & Monk, 1977; Jacoby, Kelley, Whittlesea, and their asso- Ghetti, 2003). On the other hand, study- ciates (see Kelley & lacoby, 1998; Whittle- ing material under conditions unfavorable sea, 2004). Koriat (1993) argued that the for learning (or expecting fast forgetting, nonanalytic, unconscious basis of metacog- Ghetti, 2003) results in a relatively high rate nitive judgments is responsible for the phe- of false alarms for non-memorable distrac- nomenal quality of the feeling of knowing as representing an immediate, unexplained learning or on the ease with which they intuition, similar to that which is associ- are retrieved. Both of these types of cues ated with the experience of perceiving (see become available in the course of learning Kahneman, 2003). According to this view, and hsclose the memorability of the stud- sheer subjective experience, which lies at ied material. Such cues have been assumed the core of conscious awareness, is in fact to give rise to a sheer feeling of knowing. the end product of processes that lie below Indeed, there is evidence suggesting that awareness. JOLs monitor the ease with which stud- Several cues have been proposed as deter- ied items are processed during encoding minants of JOL, FOK, and subjective confi- (Begg, et al., 1989; Koriat, 1997; Matvey dence. These cues have been referred to col- et al., 2001). For example, Begg et al. (1989) lectively as "" cues (Koriat, 1997). reported results suggesting that JOLs are With regard to JOLs and FOK, these cues sensitive to sevcral attributes of words (e.g., include the ease or fluency of processing of concreteness-abstractness) that affect ease a presented item (Begg et al., 1989), the of processing. Other findings suggest that familiarity of the cue that serves to probe JOLs are affected by the ease and probabil- memory (Metcalfe, Schwartz, & Joaquim, ity with which the to-be-remembered items 1993;Reder & Ritter, 1992; Reder & Schunn, are retrieved during learning (Benjamin & 1996), the accessibility of pertinent partial Bjork, 1996; Benjamin, Bjork, & Schwartz, information about a solicited memory tar- 1998; Koriat &Malayan, 2005). For example, get (Dunlosky & Nelson, 1992; Koriat, 1993; Hertzog, Dunlosky, Robinson, and Kidder Morris, lqqo), and the ease with which (2003) reported that JOLs increased with information comes to mind (Kelley & Lind- the speed with which an interactive image say, 1993; Koriat, 1993; Mazzoni & Nelson, was formed between the cue and the target 1995). Subjective confidence in the correct- in a paired-associates task. Similarly, Matvey ness of retrieved information has also been et al. (2001) found that JOLs increased with claimed to rest on the ease with which infor- increasing speed of generating the targets mation is accessed and on the effort experi- to the cues at study. These results are con- enced in reaching a decision (Kelley & Lind- sistent with the view that JOLs are based say, 1993; Nelson & Narens, 1990; Robinson on mnemonic cues pertaining to the fluency &Johnson, 1998; Zakay & Tuvia, 1998). of encoding or retrieving to-be-remembered These cues differ in quality from items during study. those underlying theory-based judgments. With regard to FOK judgments, sev- Whereas the latter judgments draw upon eral heuristic-based accounts have been pro- the content of domain-specific beliefs and posed. According to the cue familiarity knowledge that are retrieved from memory, account, first advanced by Reder (1987; see the former rely on contentless mnemonic also Metcalfe et al., iyqj), FOK is based on cues that pertain to the quality of process- the familiarity of the pointer (e.g., the ques- ing, in particular, the fluency with which tion, the cue term in a paired-associate, etc., information is encoded and retrieved. As see Koriat & Lieblich, 1977) that serves to Koriat and Levy-Sadot (1999) argued, "The probe memory (Reder, 1987). Reder argued cues for feelings of knowing, judgments of that a fast, preretrieval FOK is routinely and learning or subjective confidence lie in struc- automatically made in response to the famil- tural aspects of the information processing iarity of the terms of a memory question system. This system, so to speak, engages in to determine whether the solicited answer a self-reflective inspection of its own oper- exists in memory. This preliminary FOK ation and uses the ensuing information as a can guide the question answering strategy. basis for metacognitive judgments" (p. 496). Indeed, the latency of speeded FOK judg- Consider experience-based JOLs. These ments was found to be shorter than that of have been claimed to rely on the ease providing an answer. Furthermore, in sev- with which the items are encoded during eral studies, the advance priming of the AND CONSCIOUSNESS "99

terms of a question was found to enhance attributes, and so on (see Koriat, Levy-Sadot, speeded, preliminary FOK judgments with- Edry, & de Marcas, 200;; Miozzo & Cara- out correspondingly increasing the proba- mazza, 1997). These partial clues may give bility of recall or recognition of the answer rise to a sheer feeling that one knows the (Reder, 1987, 1988). Schwartz and Metcalfe answer. An important assumption of the acc- (1~92)extended Reder's paradigm to show essibility account is that participants have no that cue priming also enhances (unspeeded) direct access to the accuracv of the oartial FOK judgments elicited following recall clues that come to mind, and therefore both failure. Additional evidence for the cue- correct and wrong partial clues contribute to familiarity account comes from studies using the FOK. a proactive-interference paradigm (Metcalfe Support for the accessibility account et al., 1993). Remarkable support was also comes from a study on the TOT state (Koriat obtained using arithmetic problems: When & Lieblich, 1977). An analysis of the ques- participants made fast judgments whether tions that tend to induce an overly high they knew the answer to an arithmetic prob- FOK suggested that the critical factor is lem and could retrieve it, or whether they the amount of information they tend to had to compute it, Know judgments were elicit. For example, questions that contain found to increase with increasing frequency redundancies and repetitions tend to pro- of previous exposures to the same parts of duce inflated feelings of knowing, and so are the problem, not with the availability of the questions that activate many "neighboring" answer in memory (Reder & Ritter, 1992). answers. Thus, accessibility would seem to This was true even when participants did be a global, unrefined heuristic that responds not have enough time to retrieve an answer to the mere amount of information irrespec- (Schunn, Reder, Nhouyvanisvong, Richards, tive of its correctness. Because people can & Stroffolino, 1997; see Nhouyvanisvong & rarely specify the source of partial informa- Reder, 1998, for a review). tion, they can hardly escape the contami- Consistent with the cue-familiarity nating effects of irrelevant clues by attribut- account are also the results of studies of ing them to their source. Such irrelevant the feeling-of-not-knowing. Glucksberg and clues sometimes precipitate a strong illusion McCloskey (1981) and Klin, Guzman, and of knowing (Koriat, 1995, 1998a) or even Levine (1997) reported results suggesting an illusory TOT state - reporting a TOT that lack of familiarity can serve as a basis for state even in response to questions that have determining that something is not known. no real answers (Schwartz, 1998), possibly Increasing the familiarity of questions because of the activations that they evoke. for which participants did not know the Indeed, Schwartz and Smith (1997) answer increased the latency of Don't Know observed that the probability of reporting responses as well as the tendency to make a a TOT state about the name of a ficti- Know response erroneously. tious animal increased with the amount According to the accessibility account of of information provided about that animal, FOK, in contrast, FOKis based on the overall even when the amount of information did accessibility of pertinent information regard- not contribute to the of recall- ing the solicited target (Koriat, 1993). This ing the name of the animal. In addition, account assumes that monitoring does not FOK judgments following a commission precede retrieval but follows it: It is by try- error (producing a wrong answer) are higher ing to retrieve a target from memory that than following an omission error (Koriat, a person can appreciate whether the target 1995; Krinsky & Nelson, 1985; Nelson & is "there" and worth continuing to search Narens, igqo), suggesting that FOK judg- for. This occurs because, even when retrieval ments are sensitive to the mere accessibilitv fails, people may still access a variety of of information. clues and activations, such as frag- In Koriat's (1993) study, after participants ments of the target, semantic and episodic studied a nonsense string, they attempted to 30° TIIE Ct\MBRII?GE HANDB OOK OF CONSCIOUSNESS

recall as many of the letters as they could cate that FOK decreases with the number and then provided FOK judgments rcgard- of pre-experimental, neighboring ing the probability of recognizing the cor- that are linked to a cue, suggesting- - - that these rect string among lures. The more letters judgments are sensitive to the competition that participants could access, the stronger between the activated elements. was their FOK regardless of the accuracy Subjective confidence in the correctness of their recall. When the number of letters of one's answers has also been assumed to accessed was held constant, FOK judgments rest sometimes on mnemonic cues deriv- also increased with the ease with which ing from the process of recalling or select- information came to mind, as indexed by ing an answer. Thus, people express stronger recall latency. confidence in the answers that they retrieve If both correct and incorrect partial infor- more quickly, whether those answers are cor- mation contribute equally to the feeling rect or incorrect (Nelson & Narens, 1990). that one knows the elusive memory target, Similarly, in a study by Kelley and Lindsay how is it that people can nevertheless mon- (~yqj),retrieval fluency was manipulated itor their knowledge accurately? Accord- through priming. Participants were asked to ing to Koriat (19~;)this happens because answer general information questions and to much of the information that comes sponta- indicate their confidence in the correctness neously to mind (around y 0%; see Koriat & of their answers. Prior to this task, partici- Goldsmith, ~yqha)is correct. Therefore, the pants were asked to read a series of words, total amount of partial information accessi- some of which were correct answers and ble is a good cue for recalling or recogniz- some were plausible but incorrect answers ing the correct target. Thus, the accuracy of to the questions. This prior exposure was metamemory is a byproduct of thc accuracy found to increase the speed and probabil- of memory: Memory is by and large accu- ity with which those answers were pro- rate in the sense that what comes to mind is vided in the recall test and, in parallel, to much more likely to be correct than wrong. enhance the confidence in the correctness A third account still assumes a combined of those answers. Importantly, these effects operation of the familiarity and accessibility were observed for both correct and incorrect heuristics. According to this account both answers. These results support the view that heuristics contribute to FOK, but whereas retrospective confidence is based in part on a the effects of familiarity occur early in simple heuristic: Answers that come to mind the microgenesis of FOK judgments, those easily are more likely to be correct than those of accessibility occur later, and only when that take longer to retrieve. cue familiarity is sufficiently high to drive The imagination inflation effect also illus- the interrogation of memory for potential trates the heuristic basis of confidence judg- answers (Koriat, & Levy-Sadot, 2001; Ver- ments. Asking participants to imagine some non & Usher, 2003). This account assumes childhood events incrcased confidence that that familiarity, in addition to affecting FOK these events did indeed happen in the judgments directly, also serves as a gating past (Garry, Manning, Loftus, & Sherman, mechanism: When familiarity is high, par- 19~6).Merely aslung about the cvent twice ticipants probe their memory for the answer, also increased subjective confidence. Possi- and then the amount of information acces- bly imagination of an event and attempt- sible affects memory performance. Mihen ing to recall it increase its retrieval fluency, familiarity is low, the effects of potential which in turn contributes to the confidence accessibjlity on FOK are more limited. that the event has occurred (see also Hastie, It should be noted, however, that results Landsman & Loftus, 1y78). obtained by Schreiber and Nelson (1998) In sum, although metacognitive judg- question the idea that FOK judgments are ments may be based on explicit inferences sensitive to the mere accessibility of partial that draw upon a priori beliefs and knowl- clues about the target. These results indi- edge, much of the recent evidence points to AIYD CONSCIOUSNESS 3 01

the heuristic basis of such judgments, sug- access to memory traces comes from obser- gesting that feelings of knowing are based vations documenting a dissociation between on the application of nonanalytic heuris- subjective and objective indexes of know- tics that operate below conscious awareness. ing. Several such dissociations have been These heuristics rely on mnemonic cues reported. These dissociations also bring to pertaining to the quality of processing and the fore the effects of specific mnemonic result in a sheer noetic experience. Thus, it cues on metacognitive judgments. would seem that sheer subjective feelings, With regard to JOLs, Begg et al. (1989) such as the feeling of knowing, which are found that high-frequency words, presum- at the core of subjective awareness, are the ably fluently processed, yielded higher JOLs product of unconscious processes (Koriat, but poorer recognition memory than low- 2ooob). frequency words (see also Benjamin, 2003). The distinction between information- Narens, Jameson and Lcc (iqyq) reported based and experience-based processes has that subthreshold target priming enhanced important implications that extend beyond JOLs, because it facilitated the pro- metacognition. It shares some features with cessing of the target, although it did not the old distinction between and emo- affect eventual recall. tion (see Damasio, 19943, but differs from Bjork (1999) described several conditions it. It implies a separation between two com- of learning that enhance performance dur- ponents or states of consciousness - on the ing learning but impair long-term reten- one hand, sheer subjective feelings and intu- tion and/or transfer. According to Bjork itions that have a perceptual-like quality and Rjork (lqqz), these manipulatiolls facil- and, on the other hand, reasoned cognitions itate "retrieval strength but not "storage that are grounded in a network of beliefs and strength." As a result, the learners, fooled explicit memories. It is a distinction between by their own performance during learning, what one "feels" and "senses" and what one may experience an illusion of competence, "knows" or "thinks." The extensive research resulting in inflated predictions about their in both and social psy- future performance. For example, massed chology (e.g., Jacoby & Whitehouse, 1989; practice typically yields better performance Strack, iqyz) indicates that these two com- than spaced practice in the short term, ponents of conscious awareness are not only whereas spaced practice yields considerably dissociahle, but may actually conflict with better performance than massed practice in each other, piling judgments and hehavior the long term. Massed practice, then, has in opposite directions (Denes-Raj & Epstein, the potential of leading learners to over- lgyq). The conflict between these compo- estimate their future performance. Indeed, nents is best illustrated in correction phe- Zechmeister and Shaughnessy (1980) found nomena (e.g., Jacoby & Whitehouse, 1989; that words presented twice produced higher Strack, ~yyz),which suggest that when JOLs when their presentation was massed people realize that their subjective experi- than when it was distributed, although the ence has been contaminated, they tend to reverse pattern was observed for recall. A change their judgments so as to correct for similar pattern was reported by Simon and the assumed effects of that contamination Bjork (2001) using a motor-learning task: (Strack, 199~). Participants asked to learn each of sev- eral movement patterns under blocked con- ditions predicted better performance than Dissociations between Knowing when those patterns were learned under and the Feeling of Knowing random (interleaved) conditions, whereas actual performance exhibited the opposite The clearest evidence in support of the idea pattern. that metacognitive judgments are based on Benjamin et al. (1998) reported sev- inference from cues rather than on direct eral experiments documenting a negative 3 O2 TIIE CAMBRIDGE HANDBOOK OF CONSCIOUSNESS

relation between recall predictions and reviewed above. These findings indicate that actual recall performance, presunlably deriv- manipulations that enhance the familiarity ing from reliance on retrieval fluency when of the terms of a question enhance FOK retrieval fluency was a misleading cue for judgments associated with that question future recall. For example, they had partici- without correspondingly affecting actual pants answer general information questions recall performance. A similar dissociation, and assess the likelihood that they would be inspired by the accessibility account, has able to free recall each answer in a later test. been demonstrated by Koriat (1995): The The more rapidly participants retrieved an results of that study suggest that FOK judg- answer to a question, the higher was their ments for general information questions estimate that they would be able to free tend to be accurate as long as these ques- recall that answer at a later time. In reality, tions bring to mind more correct than incor- however, the opposite was the case. rect partial information. However, deceptive Another type of dissociation was reported questions (Fischhoff, Slovic, & Lichtenstein, by Koriat, Bjork, Sheffer, and Bar (2004). 1977)~which bring to mind more incorrect They speculated that, to the extent that than correct information, produce unduly JOLs are based on processing fluency at the high FOK judgments following recall failure time of study, they should be insensitive to and, in fact, yield a dissociation to the extent the expected time of testing. This should that FOK judgments are negatively corre- be the case because the processing fluency lated with subsequent recognition memory of an item at the time of encoding should performance. not be affected by when testing is expected. With regard to confidence judgments, Indeed, when participants made JOLs for Chandler (1994) presented participants with tests that were expected either immediately a series of target and non-target stimuli, each after study, a day after study, or a week after consisting of a scenic nature picture. In a study, JOLs were entirely indifferent to the subsequent recognition memory test, a dis- expected retention interval, although actual sociation was observed such that targets for recall exhibited a typical forgetting func- which there existed a similar stimulus in the tion. This pattern resulted in a dissociation non-target series were recognized less often, such that predicted recall matched actual but were endorsed with stronger confidence recall very closely for immediate testing. For than targets for which no similar non-target a week's delay, however, participants pre- counterpart was included. Thus, seeing a dicted over 5 0% recall, whereas actual recall related target seems to impair memory while was less than 20%. enhancing confidence. That study also demonstrated the impor- Busey, Tunnicliff; Loftus, and Loftus tance of distinguishing between experience- (2000) had participants study a series of based and theory-based JOLs: When a new faces appearing at different luminance con- group of participants were presented with ditions. For faces that had been studied in a all three retention intervals and asked to dim condition, testing in a bright condition estimate how many words they would recall reduced recognition accuracy, but increased at each interval, their estimates closely confidence, possibly because it enhanced mimicked the forgetting function exhib- their fluent processing during testing. ited by the first group's actual recall. Thus, In sum, several researchers, motivated the effects of forgetting on recall per- by the cue-utilization view of metacogni- formance seem to emerge under condi- tive judgments, have deliberately searched tions that activate participants' beliefs about for conditions that produce a dissociation memory. between memory and metamemory. Inter- Dissociations have also been reported estingly, all of the manipulations explored between FOK judgments and actual mem- act in one direction: inflating metacogni- ory performance. First are the findings tive judgments relative to actual memory in support of the cue-familiarity account performance. Some of the experimental METACOGNITION 4ND CONSCIOUSNESS 303

conditions found to engender illusions of tion or bias can also be assessed by elicit- knowing are ecologically unrepresentative, ing global or aggregate predictions (Hertzog, even contrived. However, the demonstrated Kidder, Powell-Moman, & Dunlosky 2002; dissociations clearly speak against the notion Koriat, Sheffer, & Ma'ayan, 2002; Liberman, that metacognitive judgments rest on privi- 2004), for example, by asking participants to leged access to the contents of one's own estimate how many answers they got right memory and comparing that estimate to the actual number of correct answers. It should be stressed that calibration can The Validity of Metacognitive be evaluated only when judgments and per- Judgments formance are measured on equivalent scales. Thus, for example, if confidence judgments How valid are subjective feelings of know- are made on a rating scale, calibration can- ing in monitoring actual knowledge? How not be evaluated unless some assumptions accurate are people's about are made (e.g., Mazzoni &Nelson, 1995). their memory? Earlier research has sought to Such is not the case for the second establish a correspondellce between know- aspect of metacognitive accuracy, resolution ing and the feeling of knowing as an attempt (or relative accuracy). Resolution refers to to support the trace-access view of metacog- the extent to which metacognitive judg- nitive judgments. Later studies, in contrast, ments are correlated with memory perfor- inspired by the inferential view, have con- mance across items. This aspect is com- centrated on producing evidence for mis- monly indexed by a within-subject gamma correspondence and dissociation, as just correlation between judgments and perfor- reviewed. Although the conditions used in mance (Nelson, 1984). For example, in the these studies may not be ecologically rep- case of JOLs and FOK judgments, resolu- resentative, the results nevertheless suggest tion reflects the extent to which a partic- that the accuracy of metacognitive judg- ipant can discriminate between items that ments is limited. Furthermore, these results she will recall and those that she will not. In point to the need to clarify the for the case of confidence, it reflects the ability accuracy and inaccuracy and to specify the to discriminate between correct and incor- conditions that affect the degree of corre- rect answers. spondence between subjective and objective The distinction between calibration and measures of knowing. resolution is important. For example, in Two aspects of metacognitive accuracy monitoring one's own competence dur- must be distinguished. The first is calibra- ing the preparation for an exam, calibra- tion (Lichtenstein et al., 1982) or "bias" or tion is pertinent to the decision when to "absolute accuracy" (see Nelson & Dunlosky, stop studying: Overconfidence may lead to 19qi), which refers to the correspondence spending less time and effort than are actu- between mean metacognitive judgments ally needed. Resolution, in turn, is rele- and mean actual memory performance and vant to the decision how to allocate the reflects the extent to which metacogi?i- time between different parts of the mate- tive judgments are realistic. For example, if rial. Importantly, resolution can bc high, confidence judgments are elicited in terms even perfect, when calibration is very poor. of pobabilities, then the mean pobability Also, calibration and resolution may be assigned to all the answers in a list is com- affected differentially. For example, Koriat pared to the proportion of correct answers. ct al. (2002) observed that practice study- This comparison can indicate whether ing the same list of items improves res- probability judgments are well calibrated olution but impairs calibration, instilling or whether they disclose an overconfidence underconfidence. bias (inflated confidence relative to perfor- We should note that much of the exper- mance) or an underconfidence bias. Calibra- imental work on the accuracy of JOLs and 3 "4 THE CAMBRIDGE HANDB()OK OF CONSCIOUSNESS

FOK judgments has focused on resolution. the JOL-recall gamma correlation averaged In contrast, research on confidence judg- ,~4across several studies that used lists of ments, primarily the work carried out within paired-associates that included related and the judgment and decision tradition, has unrelated pairs (Koriat et al., 2002). In con- concentrated on calibration. trast, in Dunlosky and Nelson's (1~94)study, With regard to JOLs elicited during study, in which all pairs were unrelated, the gamma the results of several investigations indi- correlation averaged .20. cate that by and large item-by-item JOLs Monitoring seems to be particularly poor are well calibrated on the first study-test when it concerns one's own actions. When trial (e.g., Dunlosky & Nelson, 1994; Maz- participants are asked to perform a scries zoni & Nelson, 1995). Judgments of com- of minitasks (so called self-performed task) prehension, in contrast, tend to be very and to judge the likelihood of recalling these inflated. One reason for this is that in mon- tasks in the future, the accuracy of their pre- itoring comprehension people assess famil- dictions is poor, and much lower than that iarity with the general domain of the text for the study of a list of words (Cohen et al., instead of assessing knowledge gained from 1991). It has been argued that people some- that text (Glenberg, Sanoclu, Epstein, & times have special difficulties in monitoring Morris, 1987). their own actions (e.g., Koriat, Ben-Zur, & Two interesting trends have been Druch, 1991). reported with regard to the calibration of However, two types of procedures have JOLs. First is the aggregate effect. When been found to improve JOL resolution. The learners are asked to provide an aggregate first procedure is repeated practice study- judgment (i.e., predict how many items ing the same list of items. As noted ear- they will recall), their estimates, when lier, although repeated practice impairs cal- transformed into percentages, are substan- ibration, it does improve resolution (King, tially lower than item-by-item judgments. Zechmeister, & Shaughnessy, 1980; Koriat, Whereas the latter judgments tend to be 2002; Mazzoni et al., 1990). Thus, in Koriat relatively well calibrated or even slightly et al.'s (2002) analysis, in which the JO1.- inflated, aggregate judgments tend to yield recall gamma correlation averaged .54 for underconfidence (Koriat et al., 2002, 2004; the first study-test cycle, that correlation Mazzoni & Nelson, 1995). A similar effect reached .82 on the third study-test cycle. has been observed for confidence judgments Koriat (1997) produced evidence suggesting (Griffin & Tversky, 1992). that the improved resolution with practice Second is the underconfidence-with- occurs because (a) with increased practice practice (UWP) effect (Koriat et al., 2002): studying a list of items, the basis of JOLs When learners are presented with the same changes from reliance on pre-experimental list of items for several study-test cycles, intrinsic attributes of the items (e.g., per- their JOLs exhibit relatively good calibration ceived difficulty) toward a greater reliance on the first cycle, with a tendency toward on mnemonic cues (e.g., processing fluency) overconfidence. However, a shift toward associated with the study of these items, marked underconfidence occurs from the and (b) mnemonic cues tend to have greater second cycle on. The UWP effect was found validity than intrinsic cues, being sensitive to to be very robust across several experimen- the immediate processing of the items dur- tal manipulations and was obtained even for ing study. Rawson, Dunlosky, and Thiede a task involving the monitoring of memory (2000) also observed an improvement in for self-performed tasks. judgments of comprehension with repeated Turning next to resolution, the within- reading trials. person correlation between JOLs and sub- A second procedure that proved effec- sequent memory performance tends to be tive in improving JOL accuracy is that relatively low, particularly when the stud- of soliciting JOLs not immediately after ied material is homogeneous. For example, studying each item, but a few trials later. METACOCNITION AND CUNSCIOUSNRSS 305

In paired-ass~ciatelearning, delaying JOLs crepancy between the standard conditions has been found to enhance JOL accuracy of learning and the standard condtions of markedly (~unlosky& Nelson, 1994; Nelson testing. On a typical memory test, people g, Dunlosky, 1991). However, the delayed- are presented with a question and are asked JOL effect occurs only when JOLs are cued to produce the answer. In contrast, in the by the stimulus term of a paired-associate, corresponding learning condition, both the not when cued by an intact stimulus- question and the answer generally appear response pair (Dunlosky & Nelson, 1992). in conjunction, meaning that the assessment It would seem that the condition in which of one's future memory performance occurs JOL~are delayed and cued by the stimu- in the presence of the answer. This differ- luS alolle approximates the eventual crite- ence has the potential of creating unduly rion test, which requires access to informa- high feelings of competence that derive from tion in long-term memory in response to the failure to discount what one now knows. a cue. Indeed, Nelson, Narens, and Dun- This situation is similar to what has been losky reported evidence suggesting referred to as the "curse of knowledgeN- that, in making delayed JOLs, learners rely the ctfficulty in discounting one's privileged heavily on the accessibility of the target, knowledge in judging what a more igno- which is an effective predictor of subsequent rant other knows (Birch & Bloom, 2003). recall. When JOLs are solicited immediately Koriat and Bjork produced evidence suggest- after study, the target is practically always ing that learners are particularly prone to a and hence its accessibility has foresight bias in paired-associate cue-target little diagnostic value. There is still contro- learning when the target (present during versy, however, whether the delayed-JOL study) brings to the fore aspects of the cue effect indeed reflects improved metamem- that are less apparent when the cue is later ory (Dunlosky & Nelson, 1992) or improved presented alone (at test). Subsequent exper- memory (Kimball & Metcalfe, 2003; Spell- iments (Koriat & Bjork, 2006) indicated man & Bjork, 1992). that foresight bias, and associated overconfi- Koriat and Ma'ayan (2005) reported evi- dence, can be alleviated by conditions that dence suggesting that the basis of JOLs enhance learners' sensitivity to mnemonic changes with delay: As the solicitation of cues that pertain to the testing situation, JOLs is increasingly delayed, a shift occurs in including study-test experience, particularly the basis of JOLs from reliance on encocLng test experience, and delaying JOLs. fluency (the ease with which an item is com- Another way in which JOLs can be made mitted to memory) toward greater reliance more sensitive to the processes that affect on retrieval fluency (the ease with which the performance during testing was explored by target comes to mind in response to the cue). Guttentag and Carroll (1998) and Benjamin In parallel, the validity of retrieval fluency (2003). They obtained the typical result in in predicting recall increases with delay and which learners predict superior recognition becomes much better than that of encoding memory performance for common than for fluency. These results suggest that metacog- uncommon words (although in reality the nitive judgments may be based on the flex- opposite is the case). However, when dur- ible and adaptive utilization of ctfferent ing the recognition test learners made post- mnemonic cues according to their relative dictions about the words that they could validity in predicting memory performance. not remember (i.e., judged the likelihood The results of Koriat and Ma'ayan suggest that they would have recognized the word 'hat repeated practice and delay may con- if they had studied it), they actually post- tribute to JOL accuracy by helping learn- dicted superior recognition of the uncom- ers Overcome biases that are inherent in mon words. Furthermore, the act of malung fluency. Koriat and Bjork (2005) postdictions for one list of items was found described an illusion of competence -fore- to rectify predictions made for a second list sightbias - that arises from an inherent dis- of items studied later. As far as the accuracy of FOK judgments probability judgments markedly exceeding is concerned, these judgments are relatively the proportion of correct answers (Lichten- well calibrated (Koriat, 1993) and are mod- stein et al., 1982). This overconfidence has erately predictive of future recall and recog- been claimed to derive from a confirmation nition. Thus, participants unable to retrieve bias (see Koriat et al., 1980; Nickerson, 1998; a solicited item from memory can estimate Trope & Liberman, 1996) -the tendency to with above-chance success whether they build toward a conclusion that has already will be able to recall it in the future, produce been rcached by selectivcly gathering or uti- it in response to clues, or identify it among lizing evldence that supports that conclu- distractors (e.g., Gruneberg R- Monks, 1974; sion. However, it has also been argued that Hart, 1967). In a meta-analysis, Schwartz part of the observed overconfidence may and Metcalfe (1994) found that the accuracy be due to the biased sampling of items by of FOK judgments in predicting subsequent researchers - the tendency to include too recognition performance increases with the many deceptive items. Indeed, when items number of test alternatives. The highest cor- are drawn randomly, the overconfidence bias relations were found when the criterion test decreases or disappears (Cigerenzer et al., was recall. '991). Assuming that metacognitive judgments More recently, attempts have been made are based on internal. mnemonic cues. then to show that confideilce in a decision is their accuracy shoulddepend on the validity based on the sampling of events from mem- of the cues on which they rest. Howevcr, ory, with overconfidence resulting from a only a few studies examined the validity biased sampling (Winman & Juslin, 2005). of the mnemonic cues that are assumed Indeed, Fiedler and his associates (Fiedler, to underlie FOK judgments. Koriat (1993) Brinkmann, Retsch, & Wild, 2000; Freytag showed that the correlation between the & Fiedler, 2006) used a sampling approach amount of partial information retrieved to explain several biases in judgment and about a memory target (regardless of its decision malung in tcrms of the iiotion accuracy) is a good predictor of eventual of metacognitive myopia. According to this memory performance, and its validity is approach, many environmental entities have equal to that of FOK judgments. Whereas to be inferred from the information given in the ovcrall accessibility of information about a sample of stimulus input. Because sam- a target (inferred from the responses of one ples are rarely representative, at1 impor- group of participants) pedicted the mag- tant metacognitive requirement would be to nitude of FOK judgmcnts following recall monitor sampling biases and control for failure, the output-bound accuracy of that them. People's responses, however, arc finely information was predictive of the accuracy tuned to the information given in the (resolution) of these FOK judgments sample, and biased judgments, including (Koriat, 1995). In a similar manner, cue overconfidence, derive from the failure to familiarity may contribute to the accuracy consider the constraints imposed on the gen- of FOK judgments because in the real world eration of the information sample. cues and targets (or questions and answers) It is important to note that over- typically occur in tight conjunction; there- confidence is not ubiquitous: When it fore familiarity with the clue should predict comes to sensory discriminations, partici- familiarity with the target (Metcalfe, 2000). pants exhibit underconfidence, thinlung that Turning finally to retrospective confi- they did worse than they actually did dence judgments, these have received a great (Bjorkrnan, Juslin, & Winman, iqyj). Also, deal of research in the area of judgment whereas item-by-item confidence judg- and decision maluilg. When participants are ments yield overconfidence, aggregate (or presented with general knowledge questioils global) judgments (estimating the num- and are asked to assess the probability that ber of correct answers), as noted earlier, the chosen answer is correct, an overconfi- typically yield underconfidence (Cigerenzer dence bias is typically observed, with mean et al., iyyi; Griffin & Tversky, 1992). The underconfidence for aggregate judgments level of variability among witnesses in exper- may derive in part from a failure to make an imental laboratory studies. Such studies typ- allowance for correct answers likely to result ically maintain the same conditions across from mere guessing (Liberman, 2004). participants. In contrast, under naturalistic A great deal of research has been car- conditions the correlation is ge~lerallymuch ried out also on the confidence-accuracy higher, and it is that type of correlation that (c-A) relation, with variable results. The would seem to be of relevance in a forcn- general pattern that emerges from this sic context (Lindsay, Read & Sharma, 1998). research is that the C-A relation is quite A second reason, mentioned earlier, is that strong when calculated within each partic- retrospective confidence judgments tend to ipant (which is what was referred to as res- be based in part on participants' precon- olution), but very weak when calculated ceptions about their ability in the domain between participants (see Perfect, 2004). tested, and these preconceptions tend to be Consider the latter situation first. Research of low validity when they concern eyewit- conducted in the domain of eyewitness tes- ness memory (e.g., lineup identification). timony, focusing on the ability of partic- Several studies explored the subjective ipants to recall a particular detail from a mnemonic cues that may mediate the crime or to identify the pcrpetrator in a within-person C-A correlation. These cues lineup, has yielded low C-A correlations include retrieval latency and the (Wells & Murray, 1984). That research has of effortless retrieval. The correlation was typically focused on a between-individual higher for recall than for recognition pre- analysis, which is, perhaps, particularly rele- sumably because recall provides more cucs vant in a forensic context: It is important to pertaining to ease of retrieval than recogni- know whether eyewitnesses can be trusted tion (Koriat & Goldsmith, 1996a; Robinson, better when they are confident in the tes- Johnson, & Herndon, 1997). Robinson, timony than when they express low con- Johnson, & Robertson (2000) found that rat- fidence. Similarly, if there are several wit- ings of vividness and detail for a videotaped nesses, it is important to know whether event contributed more strongly to confi- the more confident among them is likely to dence judgments than processing fluency be the more accurate. Thus, in this context and were also more diagnostic of memory the general finding is that a person's confi- accuracy. Attempts to enhance the C-A rela- dence in his or her memory is a poor predic- tion in eyewitness identification by inducing tor of the accuracy of that memory. greater awareness of the and rea- On the other hand, research focusing on soning process involved in the decision pro- within-person variation has typically yielded cess have been largely ineffective or even moderate-to-high C-A correlations. Thus, counterproductive (Robinson & Johnson, when participants answer a number of ques- 1998). tions and for each question report their con- In sum, the accuracy of metacognitive fidence in the correctness of the answer, judgments has attracted a great deal of inter- the cross-item correlation between confi- est because of its theoretical and practical dence and accuracy tends to be relatively implications. The results are quite variable, high (e.g. Koriat & Goldsmith, 1996a). The although by and large JOI,s, FOK judg- same is true when the questions concern ments, and confidence ratings are moder- the episodic memory for a previously expe- ately predictive of item differences in actual rienced event (Koriat, Goldsmith, Schnei- memory performance. der, & Nakash-Dura, 2001). Thus, peoplecan generally discriminate between answers (or memory reports) that are likely to he correct The Control Function and those that are likely to be false. of Metacognition Why are the between-participant correla- tions very low? Several studies suggest that As noted earlier, much of the work in meta- these low correlations stem from the low cognition is pedicatcd on the assumption 30~ THE CAMBRIDGE HANDBOOK OF CONSCIOUSNESS that consciousness is not a mere epiphe- First, Thiede and Dunlosky (1999) showed nomenon. Rather, subjective feelings and that when learners are presented with an subjective judgments exert a causal role on easy goal (e.g., to learn a list of 30 items behavior. In metacognition research this idea with the aim of recalling at least lo of them), has been expressed in terms of the hypoth- they tended to choose the easier rather than esis that monitoring affects control (Nelson, the more difficult items for restudy Thiede 1996). Indeed, several observations suggest a and Dunlosky took these results to imply a causal link between monitoring and control hierarchy of control levels: At a superordi- so that the output of monitoring serves to nate level, learners may plan to invest more guide the regulation of control processes. effort studying either the easier or the more With regard to the online replation of difficult items. This strategy is then imple- learning, it has been proposed that JOLs mented at the subordinate level to control affect the choice of which items to relearn the amount of bme allocated to each item and how much time to allocate to each item. and to select items for restudy. Indeed, it has been observed that under self- Second, Son and Metcalfe (2000) hadpar- paced concltions, when learners are given ticipants learn relatively difficult materials the freedom to regulate the amount of time with the option to go back to materials that spent on each item, they tend to allocate had previously been studied. Under high more time to items that are judged to be &f- time pressure, participants allocated more ficult to learn than to those that are judged study time to materials that were judged as to be easier (for a review see Son & Met- easy and interesting. When the time pres- calfe, 2000). It was proposed that the effects sure was not so great, however, they tended of item difficulty on study time allocation to focus on the more difficult items. are mediated by a monitoring process in These results indicate that study time which learners judge the difficulty of each allocation is also affected by factors other item and then invest more effort in studying than the output of online monitoring. the judged-difficult items to compensate for Indeed, other studies indicated, for exam- their difficulty (Nelson & Leonesio, 1988). ple, that learners invest more study time Dunlosky and Hertzog (1998; see also when they expect a recall test than when Thiede & Dunlosky, 1999) proposed a they expect a recognition test (Mazzoni & discrepancy-reduction model to describe the Cornoldi, 1993) and when the instructions relation between JOLs and study time allo- stress memory accuracy than when they cation. Learners are assumed to monitor stress speed of learning (Nelson & Leonesio, online the increase in encoding strength that 1988). Also, the allocation of study time to occurs as more time is spent studying an item a given item varies according to the incen- and to cease study when a desired level of tive for subsequently recalling that item and strength has been reached. This level, which according to the expected likelihood that is referred to as "norm of studyn (Le Ny, the item would be later tested (Dunlosky & Denhiere, & Le Taillanter, 1972), is preset Thiede, 1998). on the basis of various motivational factors, Altogether, these results suggest that such as the stress on accurate learning ver- study time allocation is guided by an adap- sus fast learning (Nelson & Leonesio, iy88). tive strategy designed to minimize effort and Thus, in self-paced learning, study continues improve learning. until the perceived degree of learning meets With regard to FOK judgments, several or exceeds the norm of study. studies indicated that they predict how long In their review of the literature, Son and people continue searching for a memory tar- Metcalfe (2000) found that indeed, in 35 get before giving up: When people feel that of 46 published experimental conditions, they know the answer or that the answer is learners exhibited a clear preference for on the tip-of-the-tongue, they search longer studying the more difficult materials. There than when they feel that they do not know are two exceptions to this rule, however. the answer (Barnes et al., 1999; Costermans AND CONSCIOUSNI~SC 309

et al., 1992; Gruneberg, Monks, & Sykes, be expected to rate their performance on a 1977; Schwartz, 2001). FOK judgments are quiz of scientific reasoning lower than men also predictive of the speed of retrieving rate themselves. Such was indeed the case, an answer, so that in the case of commis- although in reality there was no gender dif- sion responses the correlation between FOK ference in actual performance. When asked judgments and retrieval latency is positive, later if they would like to participate in a sci- whereas for omission responses the corre- ence competition, women were more likely lation between FOK and the latency of to decline, and their reluctance correlated the decision to end search is negative (see significantly with their self-rated perfor- Nelson & Narens, 1990). mance on the quiz. Thus, their choices were Search time is also affected by other fac- affected by their confidence even when con- tors in addition to FOK judgments: When fidence was unrelated to actual performance. participants are penalized for slow respond- A systematic examination of the con- ing, they tend to retrieve answers faster trol function of confidence judgments was but produce more incorrect answers (Barnes conducted by Koriat and Goldsmith (1994, et al., 1999). 1996a,b) in their investigation of the strate- As noted earlier, Reder (1987 ) proposed gic regulation of memory accuracy. Con- that preliminary FOK judgments also guide sider the situation of a person on the witness the selection of strategies for solving prob- stand who is sworn to "tell the whole truth lems and answering questions. In her stud- and nothing but the truth." To meet this ies, the decision whether to retrieve a solu- requirement, that person should monitor tion to an arithmetic problem (Know) or to the accuracy of every piece of information compute it was affected by manipulations that comes to mind before deciding whether assumed to influence cue familiarity. These to report it or not. Koriat and Goldsmith studies suggest that FOK judgments that are proposed a model that describes the mon- misled by cue familiarity can misguide the itoring and control processes involved. The decision to retrieve or compute the answer. rememberer is assumed to monitor the sub- Retrospective monitoring can also affect jective likelihood that each candidate mem- behavior. When people make an error in per- ory response is correct and then compare forming a task they can detect that without that likelihood to a preset threshold on the an external feedback and can often imme- monitoring output to determine whether diately correct their response. Following the to volunteer that response or not. The set- detection of an error, people tend to adjust ting of the control threshold depends on the their speed of responding to achieve a desir- relative utility of providing as complete a able level of accuracy (Rabbit, 1966). report as possible versus as accurate a report Confidence judgments have also been as possible. Several results provided consis- shown to affect choice and behavior and tent support for this model. First, the ten- do so irrespective of their accuracy. As dency to report an answer was very strongly noted earlier, people are often overconfident correlated with subjective confidence in the in their knowledge. Fischhoff et al. (1977) correctness of the answer (the intra-subject showed that had sufficient faith in gamma correlations averaged more than .95; their confidence judgments that they were Koriat & Goldsmith, 1996b, Experiment 1; willing to stake money on their valilty. see also Kelley & Sahakyan, 2003). This Consider the finding, mentioned earlier, result suggests that people rely completely that when judging how well they have done on their subjective confidence in deciding on a test, people tend to base their judg- whether to volunteer an answer or with- ments on their preconceptions about their hold it. In fact, participants were found to abilities in the domain tested. Ehrlinger rely heavily on their subjective confidence and Dunning (2003) reasoned that because even when answering a set of "deceptive" women tend to perceive themselves as less general knowledge questions, for which sub- scientifically talented than men, they should jective confidence was quite undiagnostic 3 lo THE CAMBRIDGE HANDHOOK OF CONSCIOUSNESS of accuracy (Koriat & Goldsmith, i99hb, Sahakyan, 2003; Pansky, Koriat, Goldsmith, Experiment 2). Second, participants given a & Pearlman-Avnion, 2002) were found to be high accuracy incentive (e.g., "you win one less effective than young adults (college stu- point for each correct answer but lose all dents) in utilizing the optio~lto withhold of your winnings if even a single answer is answers to enhance their accuracy. These incorrect") adopted a stricter criterion than results have implications for the dependabil- participants given a more moderate incen- ity of children's testimony in legal settings. tive (a 1:1 penalty-to-bonus ratio), suggest- Interestingly, results suggest that the rela- ing that the strategic regulation of memory tionship between monitoring and control, reporting is flexibly adapted to the empha- what Koriat and Goldsmith (1q96b) termed sis on memory accuracy. Third, the option "control sensitivity," lnay be impaired to to volunteer or withhold responses (which some extent in aging (Pansky et al., 2002) is often denied in traditional memory exper- and in certain psychotic disorders, such as iments) allowed participants to boost the schizophrenia (Danion, Gokalsing, Robert, accuracy of what they reported, in compar- Massin-Krauss, & Bacon, 2001; Koren et al., ison with a forced-report test. This increase 2004). In the Koren et al. study, for instance, occurred by sacrificing some of the cor- the correlation between confidence judg- rect answers; that is, at the expense of ments in the correctness of a response and memory quantity performance, This implies the decision to volunteer or withhold that that eyewitnesses generally cannot "tell the response was highly diagnostic of the degree whole truth" and also "tell nothing but the of and awareness that schizophrenic truth," but must find a compromise between patients showed concerning their mental the two requirements. Importantly, how- condition - more so than traditional mea- ever, the extent of the quantity-accuracy sures of executive control, such as the Wis- tradeoff was shown to depend critically on consin Card Sorting Task. Patients exhibit- monitoring effectiveness: In fact, when mon- ing low control sensitivity were also less able itoring resolution is very high (that is, when a to improve the accuracy of their responses person can accurately discriminate between when given the option to choose which correct and wrong answers), the accuracy of answers to volunteer and which to withhold. what is reported may be improved signifi- The research reviewed above has direct cantly under free report conditions at little bearing on the question of how people can or no cost in quantity performance. Thus, avoid false memories and overcome the con- in the extreme case when monitoring is per- taminating effects of undesirable influences. fect, a person should be able to exercise a Using fuzzy-trace theory as a framework, perfect screening process, volunteering all Brainerd et al. (2003) proposed amechanism correct items of information that come to for false-memory editing that allows chil- mind and withholding all incorrect items. dren and adults to reject false but gist- Koriat and Goldsmith's model was consistent events. The model also predicts applied to study the strategic regulation of the occurrence of erroneous recollection memory accuracy by school-aged children rejection, in which true events are inappro- (Koriat et al., 2001). Even second-to-third- priately edited out of memory reports. grade children were effective in enhancing Payne, Jacoby, and Lambert (2004) inves- the accuracy of their testimony when given tigated the ability of participants to over- the freedom to volunteer or withhold an come stereotype-based memory distortions answer under a 1:i penalty-to-bonus ratio, when allowed the option of free report. and they were able to enhance the accu- Reliance on subjective confidence allowed racy of their reports even further when given participants to enhance their overall mem- stronger incentives for accurate reporting. ory accuracy, but not to reduce stereotype However, both the children in this study bias. The results suggested that whereas sub- (see also Roebers, Moga, & Schneider, 2001) jective confidence monitors the accuracy and elderly adults in other studies (Kelley & of one's report, stereotypes distort memory METACOGhII ION AND CONSCIOUSNESS 311

through an unconscious-accessibility bias to cision or coarseness) of the information which subjective confidence is insensitive. that is reported (Goldsmith & Koriat, 1999; Hence the effects of stereotypes are difficult Goldsmith, Koriat, & Pansky, 2005; Gold- to control. smith, Koriat, & Weinberg-Eliezer, 2002). of Johnson and her associates For example, when not completely cer- ~h~A A-- on monitoring (see Johnson, 1997; tain about the time of an event, a person Johnson, Hashtroudi, & Lindsay, 19~3)also may simply report that it occurred "late in has important implications for the edit- the afternoon" rather than "at four-thirty." ing of memory reports. According to the Neisser (1988) observed that when answer- source-monitoring framework, there are sev- ing open-ended questions, participants tend eral phenomenal cues that can be used by a to provide answers at a level of general- Tememberer to specify the source of a men- ity at which they are not likely to be mis- tal record, including such mnemonic cues as taken. Of course, more coarsely grained vividness, perceptual detail, and spatial and answers, although more likely to be correct, temporal information. Because mental expe- are also less informative. Thus, Goldsmith riences from different sources (e.g., percep- et al. (2002) found that when participants tion versus imagination) differ on average are allowed to control the grain size of their in their phenomenal qualities (e.g., visual report, they do so in a strategic manner, clarity), these diagnostic qualities can sup- sacrificing informativeness (degree of preci- port source monitoring by using either a sion) for the sake of accuracy when their heuristically based process or a more strate- subjective confidence in the more precise gic, systematic process. Both types of pro- informative answer is low, and talung into cesses require setting criteria for making a account the relative payoffs for accuracy and judgment, as well as procedures for compar- informativeness in choosing the grain size

ing- activated phenomenal information to the of their answers. Moreover, the monitoring criteria. and control processes involved in the reg- The broader implication of the work on ulation of memory grain size appear to be the strategic regulation of memory accu- similar to those underlying the decision to racy (Koriat & Goldsmith, 1996b) is that, to volunteer or withhold specific items of infor- investigate the complex dynamics between mation, implying perhaps the use of com- (a) memory (the quality of the information mon metacognitive mechanisms. A more that is available to the rememberer), (b) recent study by Goldsmith et al. (2005), monitoring, (c) control, and (d) overt accu- which examined changes in the regulation racy and quantity performance, one must of grain size over different retention inter- include a situation in which participants are vals, also yielded results consistent with this free to decide what to report and what not to model: Starting with the well-known finding report. In fact, in everyday life people have that people often remember the gist of an great freedom in reporting an event from event though they have forgotten its details, memory: They can choose what perspective Goldsmith et al. (2005) asked whether to adopt, what to emphasize and what to rememberers might exploit the differential shp, how much detail to provide, and so forgetting rates of coarse and precise infor- forth. Such strategic regulation entails com- mation to strategically regulate the accu- plex monitoring and control processes that racy of the information that they report over go beyond the decision to volunteer or with- time. The results suggested that when given hold specific items of information, and these, control over the grain size of their answers, too, deserve systematic investigation. people tend to proa '1'd e coarser ansaerers at In fact, the conceptual framework of longer retention intervals, in the attempt to Koriat and Goldsmith was extended to incor- maintain a stable level of report accuracy. porate another means by which people nor- In sum, the few studies concerning the mally regulate the accuracy of what they control function of metacognition suggest report: control over the grain size (pre- that people rely heavily on their subjective, 1'2 TNR CAMBRIDGE HANDBOOK OF CONSCIOUSNESS metacognitive feelings and judgments in ple use under self-paced conditions is largely choosing their course of action. In addition appropriate. These and other results were to the monitoring output, however, they also seen to accord with the region of proxi- take into account a variety of other consid- mal learning framework according to which erations, such as the goals of learning and learning poceeds best by attending to con- remembering, time pressure, emphasis on cepts and events that are nearest to one's accuracy versus quantity, and the emphasis current understanding and only later going on accuracy versus informativeness. on to integrate items that are more difficult. Thiede, Anderson, and Therriault (2003) used a manipulation that affected the The Effects of Metacognitive learner's monitoring accuracy in studying Regulation on Memory Performance text. They found that improved accuracy resulted in a more effective regulation of Given the dynamics of monitoring and con- study and, in turn, in overall better test trol processes discussed so far, it is of performance. Thus, learners seem to rely interest to ask, To what extent does the self- on their metacognitive feelings in regulating regulation of one's processing affect actual their behavior, and to the extent that these memory performance? There are only a few feelings are accurate, such self-regulation studies that have examined this issue sys- helps improve memory performance. tematically. As noted earlier, under self- With regard to confidence judgments, as paced learning conditions, when partici- noted earlier, the work of Koriat and Gold- pants are free to allocate study time to smith (1994, 19~36b)indicates that when different items, they tend to divide their given the option of free report, people time unevenly among the items. Does the enhance their memory accuracy consider- self-allocation of study time affect actual ably in comparison to forced-report test- memory performance? Nelson and Leone- ing and do so by relying on the subjective sio (1988) coined the phrase "labor-in-vain confidence associated with each item that effect" to describe the phenomenon that comes to mind. Because confidence is gener- large increases in self-paced study time ally predictive of accuracy, reliance on con- yielded little or no gain in recall. Specif- fidence judgments is effective in enhancing ically, they observed that the amount of accuracy when accuracy is at stake. How- self-paced study time increased substan- ever, the effective regulation of memory tially under conditions that emphasized accuracy comes at the cost of reduced accuracy in comparison with a condition that memory quantity, and both the increase in emphasized speed. However, the increase in memory accuracy achieved under the free- study time resulted in little or no gain in report option and the reduction in mem- recall. ory quantity depend heavily on monitoring Metcalfe and her associates (Metcalfe, effectiveness. 2002; Metcalfe & Kornell, 2003) examined Koriat and Goldsmith (1996b) evaluated systematically the effectiveness of the pol- the effectiveness of the participants' control icy of study time allocation for enhanc- policies given their actual levels of moni- ing memory performance. They found, for toring effectiveness. The participants were example, that learners allocated most time found to be quite effective in choosing a to medium-difficulty items and studied the control policy that would maximize their easiest items first (in contrast to what would joint levels of free-report accuracy and quan- be expected from the discrepancy-reduction tity performance, compared to an "opti- model, Dunlosky & Hertzog, 1998). When mal" control policy that could be applied study time was experimentally manipulated, directly, based on the confidence judg- the best performance resulted when most ments assigned to the individual answers time was given to the medium-difficulty under forced report. The effectiveness of items, suggesting that the strategy that peo- the participants' control of grain size in the AND CONSCIOUSNESS 3'3

Goldsmith et al. (2002) study was much less best highlighted by contrasting experience- impressive, however, perhaps because of the based judgment and theory-based judg- greater complexity of the incentive struc- ments. Similar contrasts have been proposed ture (differential payoffs for correct answers by researchers in both cognitive psychology at different grain sizes, a fixed penalty for and social psychology who drew a distinction incorrect answers, regardless of grain size). between two general modes of cognition (see In fact, one of the interesting findings of that Chaiken & Trope, lgqq), and each of these study was that participants seemed to adopt contrasts highlights a particular dimension. a simple "satisficing" heuristic based on the Thus, different researchers have conceptual- payoff (whether explicit or implicit) and ized the distinction in terms of such polari- confidence for the more precise-informative ties as Nonanalytic versus Analytic cognition answer alone. rather than to comDare the (Jacoby & Brooks, 1984), Associative versus expected subjective utility (confidence mul- Rule-Based Systems (Sloman, 1996)~Expe- tiplied by subjective of potential riential versus Rational Systems (Epstein answers at different grain sizes. Monitor- & Pacini, lqqg), lmpulsive versus Reflec- ing effectiveness for the correctness of the tive processes (Strack & Deutsch, 20041, answers at different grain sizes was, how- Experience-Based versus Information-Based ever, also relatively poor (see also Yaniv & processes (Kelley & Jacoby, 1qq6a; Koriat & Foster, 1997). Thuq it may be that there Levy-Sadot, ~qyq),Heuristic versus Delib- are limits on the complexity and efficiency erate modes of thought (Kahneman, zooj), of both monitoring and control processes and Experiential versus Declarative informa- that in turn place limits on the performance tion (Schwarz, 2004). Stanovich and West benefits that can be achieved through such (2000) used the somewhat more neutral control. terms System 1 versus System 2, which In sum, only a few studies explored the have been adopted by Kahneman (2003) in effectiveness of metacognitive monitoring describing his work on judgmental biases. and control processes in enhancing actual In this cha~terI focused on the contrast memory performance. More work in this between theory-based and experience-based vein is needed. judgments, which seems to capture best the findings in metacognition. As far as metacog- nitive judgments are concerned, the impor- Metacognition and Consciousness: tant assumption is that both experience- Some General Issues based and theory-based judgments are infer- ential in nature. They differ, however, in two respects. First, theory-based judgments draw In concluding this chapter I would like to upon the content of declarative (semantic comment on how the research on metacog- and/or episodic) information that is typically nition relates to some of the fundamental stored in long-term mcrnory. Experience- issues regarding consciousness and its role based judgments, in contrast, are assumed to in behavior. I discuss three issues: the deter- rely on mnemonic cues stemming from the minants of subjective experience, the con- current processing of the task at hand. Such trol function of subjective experience, and cues as fluency of processing or ease of access the cause-and-effect relation between con- pertain to the quality and efficacy of object- sciousness and behavior. level processes as revealed online. Hence, as Koriat (1993) argued, experience-based The Genesis of Subjective Experience FOK judgments, for example, monitor the The study of the bases of metacognitive information accessible in short-term mem- judgments and their accuracy brings to the ory rather than the information available in fore an important process that seems to long-term memory. It follows that the accu- underlie the shaping of subjective experi- racy of theory-based judgments depends on ence. The unique qualities ofthat process are the validity of the theories and knowledge 3l4 THE CAMBRIDGE HANDBOOK OF CONSClOUSNFSS

on which they are based, whereas the accu- by attempting to place the metacognitive racy of experience-based judgments should distinction within a broader framework that depend on the diagnosticity of the effective encompasses other similar distinctions. For mnemonic cues. example, research in social psychology sug- Second, they differ in the nature of gests that the interplay between declara- the underlying process. Theory-based judg- tive and experiential information is greater ments are assumed to rely on an explicitly than has been realized so far (see Schwarz, inferential process: The process is assumed 2004). However, little is known about the to be deliberate, analytic, slow, effortful, and possibility that a similar interplay between largely conscious. In contrast, experience- the effects of theories and knowledge and based judgments involve a two-step process: those of mnemonic cues occurs also with A fast, unconscious, automatic inference regard to metacognitive judgments. Also, lit- results in a sheer subjective experience, and tle research has been carried out that exam- that subjective experience can then serve as ines the possible effects of attribution and the basis for noetic judgments. Therefore, misattribution on metacognitive judgments. as Koriat and Levy-Sadot argued (1999), the Furthermore, processing fluency has been processes that take off from subjective expe- shown to affect a variety of phenomenal rience generally have no access to the pro- experiences, such as liking, truth judgments, cesses that have poduced that experience recognition decisions, and so on. Again, in the first place. it is important to examine noetic feelings It is experience-based metacognitive in the context of these other phenomenal judgments that have attracted the atten- experiences. tion of memory researchers who asked the question, How do we know that we know? The Control Function of Subjective (e.g., Hart, 1965;Tulving & Madigan, 1970). Experience Experience-based judgments have the qual- ity of immediate, direct impressions, similar The issue of metacognitive control emerges to what would follow from the trace-access most sharply when we ask, What is the sta- view of metacognitive judgments. However, tus of metacognitive monitoring and con- as argued earlier, this phenomenal quality trol processes within the current distinction could be explained in terms of the idea that between implicit and explicit cognition? In experience-based judgments are based on an light of the extensive research on both of inferential process that is not available to these areas of research, one would expect the consciousness, and hence the outcome of answer to be quite straightforward. How- that process has the phenomenal quality of a ever, such is not the case. In an edited vol- direct, self-evident intuition (see Epstein & ume on Implicit Memory and Metacognition Pacini, 1999). (Reder, 1996), the discussions of the partic- Thus, the work on metacognition nicely ipants revealed a basic : Kelley converges on the proposals advanced by and Jacoby (1996b) claimed that "metacog- Jacoby and Kelley (see Kelley & Jacoby, nitinn and implicit memory are so similar as 1993) and by Whittlesea (2002, 2004) on to not be separate topics" (p. 287). Funnell, the shaping of subjective experience. These Metcalfe, and Tsapluni (1996), on the other proposals also parallel ideas in the area of hand, concluded that "the judgment of what social psychology on the genesis of vari- and how much you know about what you ous subjective feelings (see Bless & Forgas, know or will know is a classic, almost defini- 2000; Strack, 1992). However, although it tional, explicit task" (p. 172). Finally, Reder is theoretically comforting that the distinc- and Schunn (1996) stated, "Given that feel- tion between experience-based and theory- ing of knowing, like strategy selection, tends based metacognitive processes converges on to be thought of as the essence of a metacog- similar distinctions that have emerged in nitive strategy, it is important to defend our other domains, a great deal can be gained claim that this rapid feeling of knowing is actually an implicit process rather than an of immediate feelings, such as experience- explicit process" (p. 5 o) based metacognitive feelings, is to augment Koriat (1998b, 2000b) argued that this self-control; that is, to allow some degree of ambivalence actually discloses the two faces personal control over processes that would of metacognition. He proposed a crossover otherwise influence behavior directly and model that assigns metacognition a piv- automatically, outside thc person's con- otal role in mediating between unconscious sciousness and control. and cui~sciousdeterminants of information processing. Thus, metacognitive judgments The Cause-and-Effect-Relation between were assumed to lie at the interface between Monitoring and Control implicit and explicit processes. Generally spealung, a rough distinctioil can be drawn A final metatheoretical issue concerns the between two modes of operation: In the assumptioil underlying much of the work explicit-controlled mode, which underlies in metacognition (and adopted in the fore- much of our daily activities, behavior is going discussion) - that metacognitive feel- based on a deliberate and conscious eval- ings play a causal role in affecting judgments uation of the available options and on a and behavior. However, the work of Jacoby deliberate and controlled choice of the most and his associates (see Kelley & Jacoby, appropriate course of action. In the implicit- 1998) and of Whittlesea (2004) suggests a automatic mode, in contrast, various fac- process that is more consistent with thc tors registered below full consciousness may spirit of the James-Lange view of emotion influence behavior directly and automati- (see James, 1890): Subjective experience is cally, without the mediation of conscious based on an interpretation and attribution control (see Bargh, 1997; Wegner, 2002). of one's own behavior, so that it follows Metacognitive experiences are assumed rather than precedes controlled processes. to occupy a unique position in this scheme: In fact, the assumption that metacognitive They are implicit as far as their antecedents feelings monitor the dynamics of informa- are concerned, but explicit as far as their tion processing implies that such feelings are consequences are concerned. Although a sometimes based on the feedback from self- strong feeling of knowing or an unmedi- initiated object-level processes. For exam- ated subjective conviction is certainly part ple, the accessibility model of FOK (Koriat, and parcel of conscious awareness, they may 1993) assumes that FOK judgments are themselves be the product of an unconscious based on the feedback from one's attempt to inference, as reviewed earlier. Once formed, retrieve a target from memory. Hence they however, such subjective experiences can follow, rather than precede, controlled pro- serve 3s the basis for the conscious control cesses. Thus, whereas discussions of thefunc- of information processing and action. tion of metacognitive feelings assume that The crossover model may apply to other the subjective experience of knowing drives types of unmediated feelings (Koriat & Levy- controlled action, discussions of the bases Sadot, 1999). Thus, according to this view, of metacognitive feelings imply that such sheer subjective feelings, which lie at the feelings are themselves based on the feed- heart of consciousness, may themselves be back from controlled action, and thus follow the product of unconscious processes. Such rather than precede behavior. feelings represent an encapsulated summary Recent work that addressed the cause- of a varietv of unconscious influences. and it and-effect relation between metacogni- is in this sense that they are informative (see tive monitoring and metacognitive con- Schwarz & Clore, 1996): They contain infor- trol (Koriat, in press; Koriat, Ma'ayan, & mation that is relevant to conscious con- Nussinson, 2006; see Koriat, 20oob) sug- trol, unlike the implicit, unconscious pro- gests that the interplay between them is cesses that have givcn rise to these feelings. bidirectional: Although metacognitive mon- Koriat (2ooob) speculated that the function itoring can drive and guide metacognitive 116 'I'HE CAMBRIDGE HANDBO

control, it may itself be based on the feed- References back from controlled operations. Thus, when control effort is goal driven, greater effort Bandura, A. (1977). Self-efficacy: Toward a uni- enhances metacognitive feelings, consistent fying theory of behavioral change. Psychological with the "feelings-affect-behavior" hypothe- Review, 84, 191-zl j . sis. For example, when different incentives Bargh, J. A. (1997). Thc automaticity of every- are assigned to different items in a study list, day life. In R. S. Wyer, Jr. (Ed.), Advances i+z learners invest more study time on the high- socialcognition (Vol. lo, pp. 1-61). Mahwah, NJ: incentive items and, in parallel, make higher Erlbaum. JOLs for these items than for the low- Barnes, A. E., Nelson, T. O., Dunlosky, J., Maz- incentive items. This is similar to the idea zoni, G., & Narens, L. (1999). An integrative that we run away because we are fright- system of metamenlory components involved ened, and therefore the faster we run away in retrieval. In D. Gopher &A. Koriat (Eds.), andpeiformance XVII: Cognitive re@- the safer we feel. In contrast, when control lation of performance: Interaction of theory and effort is increased effort is corre- data driven, application (pp. 287-313). Cambridge, MA: lated with lower metacognitive feelings, con- MIT Press. sistent with the hypothesis that such feelings Begg, I., Duft, S., Lalonde, P., Melnick, R., & are based on the feedback from behavior. For Sanvito, J. (1989). Memory predictions are example, under self-paced learning the more based on ease of processing. Jounzal of Memory effort learners spend studying an item the and , 2 8, 610-632. lower is their JOL, and also the lower is their Begg, I., Vinski, E., Frankovich, L., & Holgate, subsequent recall of that item. This is simi- B. (1991). Generating makes words memorable, lar to the idea that we are frightened because but so does effective reading. Memory and Coe we are running away, and therefore the faster nition, 19,487-497. we run the more fear we should experi- Benjamin, A. S. (zoo?). Predicting and postdict- ence. Thus, the study of metacognition can ing the effects of word frequency on memory. also shed light on the long-standing issue of Memory and Cognition, jI, 297-3 05 the cause-and-effect relation between con- Benjamin, A. S., & Bjork, R. A. (1996). Retrieval sciousness and behavior. fluency as a metacognitive index. In L. In sum, some of the current research Reder (Ed.), Implicit memory and metacognition in metacognition scratches the surface (pp. 309-3 38). Hillsdale, NJ: Erlbaum. of metatheoretical issues concerning con- Benjamin, A. S., Bjork, R. A,, & Schwartz, sciousness and its role in behavior and is B. L. (1998). The mismeasure of memory: beginning to attract the attention of philoso- When retrieval fluency is misleading as a metamnemonic index. Journal of Experimental phers of mind (see Nelson & Rey, 2000). Psychology: General, 12 7,5 5--68. Birch, S. A. J., & Bloom, P (2003). Children Acknowledgments are cursed: An asymmetric bias in mental- state attribution. Psychological Science, 74,283- 286. The preparation of this chapter was sup- ported by a grant from the German Fed- Bjork, R. A. (1999). Assessing our own competence: Heuristics and illusions. In D. eral Ministry of Education and Research Gopher & A. Koriat (Eds.), Attention andpe?for- (BMBF) within the framework of German- mance XW:Cognitive regulation ofperformance: Israeli Project Cooperation (DIP). The chap- Interaction of theory and application (pp. 435- ter was prepared when the author was a 4 5 9). Cambridge, MA: MIT Press. fellow at the Centre for Advanced Study, Bjork, R. A,, & Bjork, E. I>. (199~).A new theory Norwegian Academy of Science, Oslo. I of disuse and an old theory of stimulus fluctu- am grateful to Morris Goldsmith, Sarah ation. In A. F. Healy, S. M. Kosslyn, & R. M. Bar, and Rinat Gil for their help on this Shiffrin (Eds.), Essays in honor of William K. chapter. Estes, Vol. 1: From learningtheory to connectionist 4ND CONSCIOUSNESS 317

theoy, Vol. 2 : From learningprocessesto cognitive in a modified recognition test. Memoy gl Cog- processes (pp. 3 5-67). Hillsdale, NJ: Erlbaum. nition, 22, 273-280. Bjorklund, D. F., & Douglas, R. N. (1997). Clore, G. L. (1994). Can emotions be noncon- The development of memory strategies. In N. scious. In P. Ehan& R. J. Davidson (Eds.), The Cowan (Ed.), The development of memory in nature of emotion: Fundamental questions. Series childhood (pp. 201-246). Hove, England: Taylor in affective science (pp. 283-2yql. London: & Francis. Oxford University Press. Bjorhan, M., Juslin, P, & Winman, A. Cohen, R. L., Sandler, S. P, & Keglevich, L. (~qyj).Realism of confidence in sensory (199 1). The failure of memory monitoring in discrimination: The underconfidence phe- a free recall task. Canadian Journal of Psychol- nomenon. Perception and Psychophysics, 54, ogy, 45,523-538, 75-81. Costermans, J., Lories, G., &. Ansay, C. (1992). Bless, H., & Forgas, J. P (Eds.). (2000). The mes- Confidence level and feeling of knowing in sage within: The role of subjective experience in question answering: The weight of inferential social cognition and behavior. Philadelphia: Psy- processes. Journal of Experimental Psychology: chology Press. Learning, Memoy, and Cognition, 18, 142-15 0. Brainerd, C. J., & Reyna, V F., Wright, R., & Craik, F. I. M., & Lockhart, R. S. (1972). Levels of Mojardin, A. H. (2003). Recollection rejection: processing: A framework for memory research. False-memory editing in children and adults. Journal of Verbal Learning and Verbal Behavior, Psychological Review, 110, 762-784. 11, 671-684. Brown, A. I>. (1987). Metacognition, executive Damasio, A. R. (1994). Descartes' error: Emo- control, self-regulation, and other more mys- tion, reason, and the human brain. New York: terious mechanisms. In F. E. Weinert & R. Putnam. H. Kluwe (Eds.), Metacognition, motivation, Danion, J. M., Gokalsing, E., Robert, P., Massin- and understanding (pp. 95-11 6). Hillsdale, NJ: Krauss, M., & Bacon, E. (2001). Defective rela- Erlbaum. tionship between subjective experience and Brown, J., Lewis, V J., & Monk, A. F. (1977). behavior in schizophrenia. American Journal of Memorability, word frequency and negative Psychiaty, 158, 2064-2066. recognition. Quarterly Journal of Experimental Davidson, J. E., & Sternberg, R. J. (1998) Smart Psychology, 2 9, 461-47 j problem solving: How metacognition helps. In Brown, R., & McNeill, D. (1966). The "tip of the D. J. Hacker, J. Dunlosky, & A. C. Graesser tongue" phenomenon. Journal of Verbal Learn- (Eds.), Metacognition in educational theory and ingand Verbal Behavior, 5, 325-3 j7. practice (pp. 47-68). Mahwah, NJ: Erlbaum. Burke, D. M., MacKay, D. G., Worthley, J. S., De Corte, E. (2003). Transfcr as the productive & Wade, E. (1991). On the : use of acquired knowledge, skills, and moti- What causes word finding failures in young and vations. Current Directions in Psychological Sci- older adults? Journal of Memoy and Language, ence, 12, 142-146. 308 542-579. Denes-Raj, V., & Epstein, S. (1994). Conflict Busey, T A,, Tunnicliff, J., Loftus, G. R., & Loftus, between intuitive and rational processing: E. (2000). Accounts ofthe confidence accuracy When people behave against their better judg- relation in recognition memory. Psychonomic ment. Journal of Personality and Social Psychol- Bulletin gZ Review, 7,26-48. ogy, 66, 819-829. Carroll, M., Nelson, T. O., & Kirwan, A. (1997). Dunlosky, J., & Hertzog, C. (1998). Training pro- Tradeoff of semantic relatedness and degree of grams to improve learning in later adulthood: overlearning: Differential effects on metamem- Hclping older adults educate themselves. In D. ory and on long-term retention. Acta Psycholog- J. Hacker (Ed.), Metacognition in educational ica, gj, 239-253. theoy and practice (pp. 249-275). Mahwah, Chaiken, S., & Trope, Y. (Eds.). (iyqy). Dual NJ: Erlhaum. process theories in social psychology. New York: Dunlosky, J., &Nelson, 0. (1992). Importance Guilford Prcss. of the kind of cue for judgments of learning Chandler, C. C. (1994). Studying related pictures (JOL) and the delayed-JOL effect. Memoy 81 can reduce accuracy, but increase confidence, Cognition, 20, 374-jXo. 318 TtIF CAMBRIDGE WAN1>BOOK OF CONSCIOUSNESS

Dunlosky, J., & Nelson, T. 0. (1994). Does the Flavell, J. H., X Wellman, H. M. (1977). sensitivity of judgments of learning (JOLs) to Metsmemoqr In R. V. Kail & J. W. Hagen the effects of various study activities depend on (Eds.), Perspectives on the development of inem- when the JOLs occur? Journal of Memory and oy and cognition (pp. 3-33). Hillsdale, NJ: Language, 33, 545-56'5. Erlbaum. Dunlosky, J., & Thiede, K. W (1998). What Freytag, P. & Fiedler, K. (2006). Subjective valid- makes people study more? An evaluation of ity judgements as an index of sensitivity to factors that affect self-paced study. Acta Psy- sampling bias. In K. Fiedler & P. Juslin (Eds.), chologica, 98, 37-5 6. Infomation sampling and adaptive cognition Dunning, D.,Johnson, K., Ehrlinger, J., & Kruger, (pp. 127-146). New York: Cambridge Univer- J. (2003). Why people fail to recognize their sity Press. own incompetence. Current Directions in Psy- Funnell, M., Metcalfe, J., & Tsapkini, K. (1996). chological Science, 12, 8 3 -8 7 In the mind but not on the tongue: Feeling Ehrlinger, J., &Dunning, D. (2303). How chronic of knowing in an anomic patient. In L. M. self-views influence (and potentially mislead) Reder, (Ed.) Implicit memory and metacognition estimates of performance. Journal of Personality (pp. 171-194). Hillsdale, NJ: Erlbaum. and Social Psychology, 84, 5-17 Gardiner, J. M., & Richardson-Klavehn, A. (2000). Remembering and knowing. In Epstein, S., & Pacini, R. (1999). Some basic issues Tulving & F. M. Craik (Eds.), Yhe Oxford regarding dual-process theories from the per- E. 1. handbook of memoy (pp. 229-244). London: spective of cognitive-experiential self-theory. In S. Chaiken X Y Trope (Eds.), Dual process Oxford University Press. theories insocialpsycilology (pp. 462-482). New Carry, M., Manning, C. G., Loftus, E. F., & Sher- York: Guilford Press. man, S. J. (1996). Imagination inflation: Imag- ining a childhood event inflates confidence that Fernandez-Duque, D., Baird, J., & Posner, M. it occurred. Psychonomic Rulletin and Review, 3, (2000). Awareness and metacognition. Con- 208-214. sciousness and Cognition, y, 3 24-3 26 Chetti, S. (2003). Memory for nonoccurrences: Fiedler, K., Brinkmann, Betsch, Wild, B. B., T., Xr The role of metacognition. Journal of Memory (2000). A sampling approach to biases in con- and I,anguage, 48, 722-7 3 9 ditional probability judgments: Beyond base Gigerenzer, G., Hoffrage, & Kleinbolting, rate neglect and statistical format. Journal of U., Experimental Psychology: General, 129, 399- H. (~qql).Probabilistic mcntal models: A 418. Brunswikian theory of confidence. Psycholo@- cal Review, 98, 506-528. Fischhoff, B. (1975). Hindsight is not equal to Gilbert, D. T. (2002). Inferential correction. In T. foresight: The effect of outcome knowledge on Gilovich, D. Griffin, & I).Kahneman (Eds.), judgment under uncertainty. Journal of Exper Heuristics and biases: The psychology of intuitive imental Psychology Human Perception and Per- judgment (pp. 167-164) New York: Cambridge formance, I, 288-299 University Press. Fischhoff, B., Slovic, P., & Lichtenstein, S. (1~~7).Gilovich, T, Griffin, D. & Kahneman, D. (Eds.). Knowing with certainty: The appropriateness (2002). Heuristics and biases: The psychology of of extreme confidence. Journal of Experimental intuitive judgment. New York: Cambridge Uni- Psychology: Htrman Perception and Performance, versity Press. 3, 552-564, Glenberg, A. M., Sanoclu, T, Epstein, W, & Mor- Flavell, J. H. (1971). First discussant's comments: ris, C. (1987). Enhancing calibration of com- What is memory development the devel- prehension. Journal of Experimental Psychology: opment of? Human Development, 14, 272- General, 116, 119-1j6. 278. Glucksherg, S., & McCloskcy, M. (1981). Deci- Flavell, J. H. (1979). Metacognition and cog- sions about ignorance: Knowing that you nitive monitoring: A new area of cognitive- don't know. Journal of Experimental Psychology: developmental inquiry. American Psychologist, Human Learning and Memo ry, 7, j 11-3 2 5 34, 906-911. Goldsmith, M., & Koriat, A. (199~)).The strategic Flavell, J. H. (1999). Cognitive development: regulation of memory reporting: Mechanisms Children's knowledge about the mind. Annual and performance consequences. In D. Gopher Review of Psychology, 50, 21-45. & A. Koriat (Eds.), Axtention and peltomance AND CONSCIOUSNLSS 319

Xlfll: Cognitive reg~.llationof performance: Inter- Jacoby, L. L., & Whitehouse, K. (1989). An action of theory and application (pp. 373-400) illusion of memory: False recognition influ- Cambridge, MA: MIT Press. enced by unconscious perception. Journal of Goldsmith, M., Koriat, A., & Pansky, A. (2005) Experimental Psychology: General, I 126- Strategic regulation of grain size in memory 135. reporting over time. Journal of Memory and James, W. (1890). The principles of psychologv. Language, 52, 505-525. New York: Holt. Goldsmith, M., Koriat, A., & Weinberg-Eliezer, Johnson, M. K. (199~).Identifying the origin A. (2002) Strategic regulation of grain size of mental experience. In M. S. Myslohodsky memory reporting. Journal of Experimental (Ed.), The mythomanias: 7he nature ofdeception Psychology: General, I 3 1, 7 3-9 5 and self-deception (pp I 3 3-1 80). Hillsdale, NJ: Griffin, D., &Tversky, A. (1992).The weighing of Erlbaum. evidence arid the determinants of confidence. Johnson, M. K., Nashtroudi, S., & Lindsay, D. S. Cognitive Psychology, 24,411-43 5. (1~~3).Source monitoring. Prychological Hul- Gruneberg, M. M., & Monks, J. (1974). "Feeling letin, "4,3-28, of knowing" and cued recall. Acta Psychologica, Jonsson, F. U., & Olsson, M. J. (2003). Olfac- 38, 257-26. tory metacognition. Chemical Senses, 28, 651- Gruneberg, M. M., Monks, .I., & Sykes, R. N. 658. (19~7).Some metllodological problems with Jost, .J. T., Kruglanski, A. W, & Nelson, T 0. feelings of knowing studies. Acta Psychologica, (1998). Social metacognition: An expansion- 41,365-371. ist review. Personality and Social Psychology Guttentag, R., & Carroll, D. (1998). Memora- Review, 2, 137-154. bility judgments for high- and low-frequency Kahneman, D. (2003). A perspective on judg- words. Memory and Cognition, 2 6, 95 1-95 8 ment and choice: Mapping bounded rational- Hacker, U. J., Dunlosky, J., & Graesser, A. C. ity. American Psychologist, 58, 697-720. (Eds.). (1yq8). Metacognition in educational the- Kelley, C. M., X Jacoby, L. L, (1993). The coil- oly and practice. Mahwah, NJ: Erlbaum. struction of subjective experience: Memory Hart, J. T. (196;). Memory and the feeling-of- attributions. In M. Davies & C;. W Humphreys knowing experience. Journal ofEducationa1Psy- (Eds.), Consciousness: Psychological and philo- chology, 56, 208-216. sophical essays. in mimi and language Hart, J. T. (1967). Second-try recall, recognition, vol. 2, pp. 74-89). Malden, MA: Blackwell. and the memory-monitoring process. Journal of Kelley, C. M., & Jacoby, I.,. L. (1996a). Adult Educational Psycizology, 5 8, 193 -197. egocentrism: Subjective experience versus ana- Hastie, R., Landsman, R., & Loftus, E.F. (19~8). lytic bases for judgment. Journal ofMemory and The effects of initial questioning on subsequent Language, 351 '5 7-75 eyewitness testimony. Jurimetrics Jounzul, ly, Kelley, C. M., & Jacoby, I.. L. (1yy6b). Mem- 1-8 ory attributions: Remembering, knowing, and Hertzog, C., Dunlosky, J., Robinson, A. E., & feeling of knowing. In L. M. Reder (Ed.), Kidder, D. P. (2003). Encoding fluency is a Implicit memory and metacognition (pp. 287- cue used for judgments about learning. .Journal 308). Hillsdale, NJ: Erlbaum. of Experimental Psychology: Learning, Memory, Kelley, C. M., Ek Jacoby, L. 1.. (1998). Subjec- and Cognition, 2 9, 22-34 tive reports and process dissociation: Fluency, Hertzog, C., Kidder, D. P., Powell-Moman, A, knowing, and feeling. Acta Psychologicu, y8, & Dunlosky, J. (2002). Aging and monitor- 127-140, ing associative learning: Is monitoring accuracy Kelley, C. M., & Lindsay, D. S. (1993). Remem- spared or impaired? Psychology and Aging, 17, bering mistaken for knowing: Ease of retrieval 209-22 j. as a basis for confidence in answers to general Jacoby, L. L., & Brooks, L. R. (1984). Nonana- knowledge questions. Journal of Memory and lytic cognition: Memory, perception, and con- Language, 3 2, 1-24 cept learning. 111 G. H. Bower (Ed.), The psy- Kelley, C. M., & Sahakyan, L. (2003). Memory, chologv of learning and motivation: Advances monitoring, and control in the attainment of in research and theory (pp. 1-47], New York: memory accuracy. Journal of Memory and Lan- Academic Press. guage, 48, 704-721. 320 THE CAMBRIDGE HANDROOK OF CONSCIOUSNESS

Kentridge, R. W., & Heywood, C. A. (2000). (Eds.), Applied metacognition (pp. 261-286). Metacognition and awareness. Consciousness Cambridge: Cambridge University Press. and Cognition, y, 3 08-3 12. Koriat, A. (in press). Are we frightened because Kimball, D. R., & Metcalfe, J. (2003). Delay- we run away? Some evidence from metacogni- ing judgments of learning affects memory, not tive feelings. In H.Uttl, N. Ohta, &A. L. Siegen- metamemory. Memoly and Cognition, 31, 918- thaler (Eds.), Memory and emotions: Interdisci- g29. plinary perspectives. Malden, MA: Blackwell. King, J. F., Zechmeister, E. B., & Shaughnessy, J. J. Koriat, A,, Ben-Zur, H., & Druch, A. (1991). (1980). Judgments of knowing: The influence The contextualization of memory for input and of retrieval practice. American Journal of Psy- output events. PsycJ~ologicalResearch, 53, 260- chologv, 93, 329-343. 270. Klin, C. M., Guzman, A. E., & Levine, Koriat, A,, Ben-Zur, H., & Sheffer, D. (1988). W. H. (1997). Knowing that you don't know: Telling the same story twice: Output monitor- Metamemory and discourse processing. Journal ing and age, Journal of Memoly and Language, of Experimental Psychology: Learning, Memoly, 271 23-39. and Cognition, 2 3,13 78-13 93. Koriat, A,, &Bjork, R. A. (2005). Illusions ofcorn- Koren, D., Seidman, L. J., Poyurovsky, M., Gold- petence in monitoring one's knowledge dur- smith, M., Viksman, P, Zichel, S., & Klein, E. ing study. Journal of Experimental Psychology: (2004). The neuropsychological basis of insight Learning, Memoly and Cogmition, 31, 187- in first-episode schizophrenia: A pilot metacog- '94. nitive study. Schizophrenia Research, 70, 195- Koriat, A., & Bjork, R. A. (2006). Illusions of 202. competence during study can be remedied by Koriat, A. (iyyj). How do we know that we manipulations that enhance learners' sensitiv- know? The accessibility model of the feeling of ity to retrieval conditions at test. Memory 82 knowing. Psychological Review, loo, 609-639. Cognition, 34, 959-972 Koriat, A. (19~5).Dissociating knowing and the Koriat, A,, & Bjork, R. A,, Sheffer, L., & Bar, feeling of knowing: Further evidence for the S. K. (2004). Predicting one's own forgetting: accessibility model. Journal of Experimental The role of experience-based and theory-based Psychologv: General, 12 4, 3 11-3 j 3 processes. Journal of Experimental Psychology: Koriat, A. (1997). Monitoring one's own knowl- General, 133, 643-656. edge during study: A cue-utilization approach Koriat, A,, & Goldsmith, M. (1994). Memory to judgmcnts of learning. Journal of Expen'men- in naturalistic and laboratory contexts: Distin- tal Psychology: General, 12 6, 349-370. guishing the accuracy-oriented and quantity- Koriat, A. (1998a). Illusions of knowing: The link oriented approaches to memory assessment. between knowledge and metaknowledge. In Journal of Experimental Psychology: General, Y Yzerbyt, G. Lories, & B. Dardenne (Eds.), 123, 297-315- Metacognition: Cognitive and social dimensions Koriat, A,, & Goldsmith, M. (lqyha). Mem- (pp. 16-34). London: Sage. ory metaphors and the real-life/laboratory Koriat, A. (iqq8b). Metamemory: The feeling of controversy: Correspondence versus store- knowing and its vagaries. In M. Sabourin, F. I. house conceptions of memory. Behavioral and M. Craik, & M. Robert (Eds.), Advances inpsy- Brain Sciences, 19, 167-228. chological science (Vol. 2, pp. 461-469). Hove, Koriat, A,, &Goldsmith, M. (iq96b). Monitoring UK: Psychology Press. and control processes in the strategic regulation Koriat, A. (20008). Control processes in remem- of memory accuracy. Psychological Review, 103, bering. In E. Tulving & F. I. M. Craik (Eds.), 490-517, The Oxford handbook of memoly (pp. j 3 3-346) Koriat, A,, Goldsmith, M., & Pansky, A. (2000). London: Oxford University Press. Toward a psychology of memory accuracy. Koriat, A. (2ooob). The feeling ofknowing: Some Annual Review of Psychology, 51, 48 1-5 3 7 metatheoretical implications for consciousness Koriat, A,, Goldsmith, M., Schneider, W., & and control. Consciousness and Cognition, 9, Nakash-Dura, M. (2001). The credibility of 149-17'. children's testimony: Can children control the Koriat, A. (2002). Metacognition research: An accuracy of their memory reports? Journal of interim report. In T. J. Perfect & B. L. Schwartz Experimental Child Psychology, 79, 405-43 7. AND CONSCIOUSNESS 321

Koriat, A,, & Levy-Sadot, R. (1999). Pro- Liberman, V. (2004). Local and global judg- cesses underlying metacognitive judgments: ments of confidence. Journal of Experimental Information-based and experience-based mon- Psychology: Learning, Memory, and Cognition, itoring of one's own knowledge. In S. Chaiken 30, 729-732- & Y Trope (Eds.), Dualprocess theories in social Lichtenstein, S., Fischhoff B., & Phillips, L. D. psychology (pp. 483-5021. New York: Guilford (1~82).Calibration of probabilities: The state Press. of the art to 1980. In D. Kahneman, P Slovic, & Koriat, A., & Lev-Sadot, R. (2001). The com- A. Tversky (Eds.), Judgment under uncertainty: bined contributions of the cue-familiarity and Heuristics and biases (pp, 306-3 34). New York: accessibility heuristics to feelings of knowing. Cambridge University Press. Journal of Experimental Psychology: Learning, Lindsay, D. S., Read, D. J., & Sharma, K. Memory, and Cognition, 2 7, 34-5 3 (1998). Accuracy and confidence in person Koriat, A,, Levy-Sadot, R., Edry, E., & de Mar- identification: The relationship is strong when cas, S. (2003). What do we know about what witnessing conditions vary widely. psycho lo@^ we cannot remember? Accessing the seman- cal Science, y, 215-218. tic attributes of words that cannot be recalled. Maki, R. H., & McGuire, M. .J. (2002). Mrtacog- Journal of Experimental Psychology: Learning, nition for text: Findings and implications for Memory, and Cognition, 2 9, 109 5-1105. education. In T. Perfect & B. Schwartz (Eds.), Koriat, A,Lichtenstein, S., &Fischhoff, B. (1980). Applied metacognition (pp. 3 9-67). Cambridge: Reasons for confidence. Journal of Experimental Cambridge University Press. Psychology: Human Learning and Memory, 6, Matvey, G., Dunlosky, J., & Guttentag, R. (2001). 107-118. Fluency of retrieval at study affects judgments Koriat, A,, & Lieblich, 1. (1977). A study of mem- of learning (JOLs): An analytic or nonanalytical 15 ory pointers. Acta Psychologica, 41, 1-164 basisforJOLs? Memoyand Cognition, 2 9,222- Koriat, A,, & Ma'ayan, H. (2005). The effects of 233. encoding fluency and retrieval fluency on judg- Mazzoni, G., & Cornoldi, C. (1993). Strategies ments of learning. Journal of Memory and Lan- in study time allocation: Why is study time guage, 52,478-492. sometimes not effective? lournal of Experimen- Koriat, A,, Ma'ayan, H., & Nussinson, R. (2006). tal Psychology: General, 122,47-60. The intricate relationships between monitor- Mazzoni, G., Cornoldi, C., & Marchitelli, G. ing and control in metacognition: Lessons (1990). Do memorability ratings affect study- for the cause -and-effect relation between time allocation? Memory and Cognition, 18, subjective experience and behavior Jour- 196-204. nal of Experimental Psychology: General, 135, 36-69, Mazzoni, G., &Nelson, T. 0. (1995). Judgments of learning are affected by the kind of encoding Koriat, A,, Sheffer, L., & Ma'ayan, H. (2002). in ways that cannot be attributed to the level Comparing objective and subjective learn- of recall. .lournal of Experimental Prychology: ing curves: Judgments of learning exhibit Learning, Memory, and Cognition, 2 1, 12 6 3- increased underconfidence with practice. lour- nu1 of Experimental Psychology: General, 13 I, 1274. 147-162. Metcalfe, J. (1yy8a). Cognitive optimism: Krinsky, R., &Nelson, T 0. (1985). The feeling of Self-deception or memory-based processing knowing for different types of retrieval failure. heuristics? Personality and Social Psychology Acta Psychologca, 58,141-158. Review, 2, 100-110. Le Ny, J. F., Denhiere, G., & Le Taillanter, D. Metcalfe, J. (Ed.). (1yy8h). Metacognition. (1972). Regulation of study-time and interstim- [Special issue]. Persoizality and Social Psycho- ulus similarity in self-paced learning condi- logical Review, 2 tions. Acta Psychologica, 3 6, 280-289. Metcalfe, J. (2000). Metamemory: Theory and Leonesio, R. J., & Nelson, T 0. (~yqo).Do dif- data. In E. Tulving & F, I. M. Craik (Eds.), ferent metamemory judgments tap the same The Oxford handbook of memory (pp. 197-211). underlying aspects of memory? Journal of London: Oxford University Press. Experimental Psychology: Learning, Memory, Metcalfe, J. (2002). IS study time allocated selec- and Cognition, 16,464-470. tively to a region of proximal learning? Journal 322 Till' CAhlBRIDLE HANDB OOK OF CONSCIOUSNEbS

of Experimental Psychology: General, 13 I, 349- Nelson, T. O., &Narens, I-. (1990). Metamemory: 363. A theoretical framework and new findings. In Metcalfe, J., & Kornell, N. (2003). The dynam- G. Bower (Ed.), The psychology of learning and ics of learning and allocation of study time to motivatiolz:Advances in research and theory (pp. a region of proximal learning. Journal of 1;xper- 12 5-17 3). New York: Academic Press. imental Psychology: General, 112, 5 3 0-542. Nelson, T. O., & Narens, L. (109~).Why investi- Metcalfe, J., Schwartz, B. L., & Joaquim, gate metacognition. In J. Metcalfe &A. P. Shi- S. G. (1993). The cue-familiarity heuris- mamura (Eds.), Metacognition: Knowing about tic in metacognition. Journal of Experimental knowing (pp. 1-25). Cambridge, MA: MIT Psychology: Learning, Memoly, and Cognition, Press. 19, 851-864. Nelson, ' O., Narens, L., & Dunlosky, J. Metcalfe, J., & Shimamura, A. P. (Eds.). (1994). (2004). A revised methodology for research on Metacognition: Knou~ingabout knowing. Cam- metamemory: Pre-judgment Recall and Moni- bridge, MA: MIT Press. toring (PRAM). Psychological Methods, 9, 5 3- Miozzo, M., & Caramazza, A. (1997). Retrieval of 69. lexical-syntactic features in tip-of-the tongue Nelson, T O., & Rey, G. (Eds.). (2000). Metacog- states. Journal of Experimental Psychology: nition and consciousness: A convergence of Learning, Memo% and Cognition, 23, 1410- psychology and philosophy [Special issue] '423. Conscious~zessand Cognition, 9 (2). Mitchell, K. J., &Johnson, M. K.(2000). Source Nickerson, R. S. (1098). Confirmation bias: monitoring: Attributing mental experiences. In A ubiquitous phenomenon in inany guises. E. Tulving & F. I. M. Craik (Eds.), The Oxford Xevieui of General Psychology, 2, 175-220. handbook of memory (pp. 179-195). London: Nhouyvanisvong, A,, & Reder, L. M. (1998). Oxford University Press. Rapid feeling-of-luiowing: A strategy selection Morris, C. C. (1990). Retrieval processes under- mechanism. In V Y. Yzerhyt, G. Lories, & B. lying confidence in comprehcnsion judgments. Dardenne (Eds.), Metacognition: Cognitive and Journal of Experimental Psychology: Learning, social dimensions (pp. 3 5-52). L.ondon: Sage. Memory, and Cognition, 16, 22 3-2 3 2 Oppenheimer, O. M. (2004). Spontaneous dis- Narens, L., Jameson, K. A,, & Lee, V A, (1994). counting of availability in frequency judgment Subthreshold priming and memory monitor- tasks. Psychological Science, 15, 100-105 ing. In J. Metcalfe & A. P. Shimamura (Eds.), Pansky, A., Koriat, A., & Goldsmith, M. (2005). Metacoflition: Knowing about knowing (pp. 71- Eyewitness recall and testimony In N. Brewer 92). Cambridge, MA: MIT Press. & K. D. Williams (Eds.), Psychologyand law:An Neisser, U. (1985). Time present and time past. In empirical perspective (pp, 93-150). New York: M. M. Gruneberg, P. Morris, & R. Sykes (Eds.), Cuilford. Practical aspects of memory: Current research Pansky, A,, Koriat, A,, Goldsmith, M., & and issues (Vol. 2, pp. 545-560). Chichester, Pearlman-Avnion, S. (March, 2002). Memory England: Wiley. accuracy and distortion in old age: Cognitive, Nelson, 0. (198~).A comparison of current metacogtzitizie, and neurocognitive determinants. measurcs of the accuracy of feeling-of-knowing Poster presented at the 30th Anniversary Con- predictions. Psychological Bulletin, yg, 109-133 ference of the National Institute for Psychobi- Nelson, 'T 0. (I 996). Consciousness and ology, Jerusalem, Israel. metacognition. American Psychologist, 5, lor- Paris, S. G., & Winograd, P. (I~IIO).How metacog- 116. nition can promote academic learning and Nelson, T. O., & Dunlosky, J. (1991). When instruction. In R. F Jones & I., Idol (Eds.), people's judgments of learning (SOLS) are Dimensions of thinking and cognitive instruction extremely accurate at predicting subsequent (pp. 15-51). Hillsdale, NJ: Erlhaum. recall: The "delayed-JOL effect." Psychological Payne, R. K., Jacoby, I.. I,., & Lambcrt, A. J., Science, 2, 267-270. (2004). Memory monitoring and the control Nelson, O., & Leonesio, R. J. (1988). Alloca- of stereotype distortion. Journal of Experimen- tion of self-paced study time and the "labor-in- tal Social Psychology, 40,52-64. vain effect." Journal of Experimental Psychology: Perfect, T. J. (2002). When does eyewitness con- Learning, Memory, and Cognition, 14, 676-686. fidence predict performance? In T J. Perfect AND CONSCIOIISNESS 3=3

& B. I,. Schwartz (Eds.), Applied metacognition The detrimental effects of attention to the (pp. 95-120). Cambridge: Cambridge Univer- identification process. Law and Human Behav- sity Press. ior, 22, 400-428. Perfect, T J. (2004). The role of self-rated abil- Robinson, M. D., Johnson, J. T., & Herndon, F. ity in the accuracy of confidence judgments (1997). Reaction time and assessments of cog- in eyewitness memory and general knowledge. nitive effort as predictors of eyewitness mem- Applied Cognitive Psychology, 18, 157-168 ory accuracy and confidence. Journal of Applied Perner, .I., &Lang, B. (1999). Development ofthe- Psychology, 82, 416-42 5 ory of mind and executive control. Trends in Robinson, M. D., Johnson, J. T., & Robertson, Cognitive Sciences, 3, 3 3 7-344 D. A. (2000). Process versus content in eyewit- Posner, M. J., & Snyder, C. R. R. (1i)jj). Attcn- ness metamemory monitoring. Journal of Exper- tion and cognitive control. In R. L. Solso imental Psychology Applied, 6,207-221. (Ed.), Information processing and cognition: The Roebers, C. M., Moga, N., & Schneider, W, Loyola Symposium (pp. 5 5-85). Hillsdale, NJ: (2001). The role of accuracy motivation on chil- Erlhaum. dren's and adults' event recall. Journal ofExper- Pressley, M., Borkowski, J. G., & Schneide~; imental Child Psychology, 78, 3 13-3 29. W. (1987). Cognitive strategies: Good strat- Rosenthal, D. M. (2000). Consciousness, con- egy users coordinate metacognition and knowl- tent, and metacognitive judgments. Conscious- edge. Annah of Child Development, 4, 89-129. ness and Cognition, y, 203 -2 14. Rabinowitz, J. C., Ackerman, B. P, Craik, Scheid, K. (1993). Helping students become st rate^ E I. M., & Hinchley, J. L. (1982). Aging and gic learners: Guidelinesfor teaching. Cambridge, metamemory: The roles of relatedness and MA: Brookline Rooks. imamnary Journal of Gerontology, 3 7,688-695 Schneider, W. (1985). Developmental trends Rahbitt, P M. A. (1966). Errors and error correc- in the metamemory-memory behavior tion in choice reaction tasks. Journal of Experi- relationship: An integrative review. In D. mental Psychology, 71, 264-272. L. Forest-Pressley, G. E. MacKinnon, & T. Rawson, K. A., Dunlosky, J., & Theide, K. W. G. Waller (Eds.), Metacoflzition, cognition, (2000). The rereading effect: Metacomprehen- and human p&ormance (Vol. 1, pp. 57-109). sion accuracy improves across reading trials. Orlando, FL: Academic Press. Memoy and Cognition, 28, 1004-1010. Schneider, W., & Pressley, M. (1997). Memoy Reder, L. M. (1987). Strategy selection in ques- dewelopment between two and twenty (zd ed.). tion answering. Cognitive Psychology, 19, 90- Mahwah, NJ: Erlbaum. 138. Schreiber, T. A,, & Nelson, D. L. (1998). The Reder, L. M. (1988). Strategic control of retrieval relation between feelings of knowing and the strategies. In G. H. Bower (Ed.), Thepsychology number of neighboring concepts linked to the of learning and motivation: Advances in research test cue. Memory and Cognition, 26, 869- and theoy (Vol. 22, pp. 227-259). San Diego: 883. Academic Press. Schunn, C. D., Reder, 1.. M., Nhouyvanisvong, Reder, L. M. (Ed.). (1996). Implicit nzemoy and A., Richards, D. K., & Stroffolino, P. J. (1997). metacognition. Mahwah, NJ: Erlbaum. To calculate or not to calculate: A source activa- Reder, L. M., & Ritter, F. E. (1992). What deter- tion conhsion model of problem familiarity's mines initial feeling of knowing? Familiarity role in strategy selection. Journal of Experimen- with question terms, not with the answer. Jour- tal Psychology: Learning, ~hfemoy,and Cogni- nal of Experimental Psychology: Learning, Mem- tion, 2 3, 3-29. oy, and Cognition, 18, 43 5-45 1. Schwartz, B. L. (1994). Sources of information in Reder, L. M.& Schunn, C. D. (1996). Metacogni- metamemory: Judgments of learning and feel- tion does not imply awareness: Stratem choice ing of knowing. Psychonomic Bulletin 82 Review, is governed by implicit learning and mem- 357-375. ory In L. M. Reder (Ed.), Implicit memoy Schwartz, B. L. (1998). Illusory tip-of-the-tongue and metacognition (pp. 4 5-77). Mahwah, NJ: states. Memoy, 6,623-642 Erlbaum. Schwartz, B. L. (2001). The relati011 of tip-of-the- Robinson, M. D., &Johnson, J. T. (1998). HOWnot tongue states and retrieval time. Memoy and to enhance the confidence-accuracy relation: Cognition, 2 9, 117-126. 3=4 TIll? CAMBRIDGE HANDBO

Schwartz, B. L. (2002). Tip-of-the-tonguestates: itive judgment (pp. j 97-420). New York: Cam- Phenomenology, mechanism, and lexical bridge University Press. retrieval. Mahwah, NJ: Erlbaum. Son, L. K., & Metcalfe, J. (2000). Metacognitive Schwartz, B. L., & Metcalfe, J. (iyyz). Cue famil- and control strategies in study-time allocation. iarity but not target retrievability enhances Journal of Experimental Psychology: Learning, feeling-of-knowing judgments. Journal of Memory, and Cognition, 26, 204-221. Experimental Psychology: Learning, Memory, Spehn, M. K., & Reder, L. M. (2000). The uncon- and Cognition, 18, 1074-1083 scious feeling of knowing: A commentary on Schwartz, B. L., & Metcalfe, J. (iyyq). Method- Koriat's paper. Consciousness and Cognition, y, ological problems and pitfalls in the study of 187-192. human metacognition. In J. Metcalfe & A. Spellman, B. A,, & Bjork, R. A. (1992). When P. Shimamura (Eds.), Metacognition: Knowing pedictions create reality: Judgments of learn- about knowing (pp. yj-113). Cambridge, MA: ing may alter what they are intended to assess. MIT Press. Psychological Science, 3, 3 15-3 16 Schwartz, R. I,., & Smith, S. M. (1997). The Stanovich, K. E., & West, R. F. (2000). Individ- retrieval of related information influences tip- ual differences in reasoning: Implications for of-the-tongue states. Journal of Memory and the rationality debate. Behavioral and Brain Sci- Language, 36, 68-86. ences, 23,645-665. Schwarz, N. (2004). Meta-cognitive experiences Strack, F, (iyyz). The different routes to social in consumer judgment and decision making. judgments: Experiential versus informational Journal of Consumer Psychology, 14, 3 32-348 strategies. In L. L. Martin & A. Tesser (Eds.), Schwarz, N., Bless, H., Strack, F., Klumpp, G., The construction of social judgments (pp. 249- Rittenauer-Schatka, H., & Simons, A. (1991). 275). Hillsdale, NJ: Erlbaum. Ease of retrieval as information: Another Strack, F., & Bless, H. (lyyq). Memory for look at the availability heuristic. Journal of nonoccurrences: Metacognitive and presuppo- Personality and Social Psychology, 61, 195- sitional strategies. Journal of Memory and Lan- 202. guage, 33,203-217. Schwarz, N., & Clore, G. L. (1996). Feelings and Strack, F., & Deutsch R. (2004). Reflective and phenomenal experiences. In E. T. Higgins &A. impulsive determinants of social behavior Per- W. Kruglanski (Eds.), Social psychology: Hand- sonali~and Social Psychology Review, 8, 220- book of basic principles (pp. 43 3-465). New 247. York: Guilford Press. Thiede, K. W, Anderson, M. C. M., &Therriault, Schwarz, N., & Clore, G. L. (2003). Mood D. (2003). Accuracy of metacognitive monitor- as information: 20 years later. Psychological ing affects learning of texts. Journal of Educa- Inquiry, 14, 296-303 tional Psychology, 95, 66-73, Shaw, R. J., & Craik, F I. M. (1989). Age differ- Thiede, K. W., & Dunlosky, J. (~yyy).Toward ences in predictions and performance on a cued a general model of self-regulated study: An recall task. Psychology and Aging, 4, 13 1-13 5 analysis of selection of items for study and Shimamura, A. P. (2000). Toward a cognitive self-paced study time. Journal of Experimental neuroscience of metacognition. Consciousness Psychology: Learning, Memory, and Cognition, and Cognition, y, 3 13-32 3. 25, 1024-1037. Simon, D. A,, & Bjork, R. A. (2001). Metacogni- Trope, Y., & Liberman, A. (1996). Social hypoth- tion in motor learning. Journal of Experimental esis testing: Cognitive and motivational mech- Psychology: Learning, Memory, and Cognition, anisms. In E. T Higgins & A. W Kruglan- 27, 907-912. ski (Eds.), Social psychology: Handbook of basic Sloman, S. A. (1996). The empirical case for two principles (pp. 239-270). New York: Guilford systems of reasoning. Psychological Bulletin, ~iy, Press. 3-22 Tulving, E., & Madigan, S. A. (1970).Memory and Slovic, P., Finucane, M., Peters, E., & MacGre- verbal learning. Annual Review of Psychology, gor, D. G. (2002). The affect heuristic. In T. 21,437-484, Gilovich, D. Griffin, & Kahneman, D. (Eds.), Underwood, B. J. (1966). Individual and group Heuristics and biases: The psychology of intu- predictions of item difficulty for free learning. METACOGNITION AND CONSCIOIJSNESS 3=5

Journal of Experimental Psychology, 71, 673- judgment. In J. Musch & K. C. Klauer (Eds.), The psychology of evaluation: Affective processes 679, Vernon, D., & Usher, M. (2003). Dynamics in cognition and emotion (pp. 189-217). Mah- metacognitive judgments: Pre- and postre- wah, NJ: Erlbaum. trieval mechanisms. Journal of Experimental Winman, A., & Juslin, P. (200j). "I'm m/n confi- psychology: Learning, Memoy, and Cognition, dent that I'm correct": Confidence in foresight 29, 339-346- and hindsight as a sampling probability. In K. Wegner, D. M. (2002). The illusion of conscious Fiedler & P. Juslin (Eds.), Information sampling ,ill. cambridge, MA: MIT Press. and adaptive cognition. Cambridge, UK: Cam- bridge University Press. wells, G. L., & Murray, D. M. (1984). Eyewit- ness confidence. In G. L. Wells & E. F. Loftus, Yaniv, I., & Foster, D. P. (199;). Precision and (Eds.), Eyewitness testimony: Psychological per accuracy of judgmental estimation. Journal of spectives (pp. 155-170). New York: Cambridge Behavioral Decision Making, lo, 21-j~ University Press. Yaniv, I., & Meyer, D. E. (1987). Activa- whittlesea, B. A. (2002). TWO routes to tion and metacognition of inaccessible stored remembering (and another to remembering information: Potential bases for incubation not). Journal of Experimental Psychology: Gen- effects in problem solving. Journal of Experi- eral, 131, 325-348. mental Psychology: Learning, Memoly, and Cog- Whittlesea, B. W A. (2004). The perception of nition, 13, 187-205 integrality: Remembering through the valida- Yzerbyt, Y, Lories, G., & Dardenne. B. tion of expectation. Journal of Experimental (Eds.). (1998). Metacognition: Cognitive and Psychology: Learning, Memoy, and Cognition, social dimensions. Thousand Oaks, CA: Sage. 30, 891-908. Zakay, D., & Tuvia, R. (1998). Choice latency Winkielman, P, &Beridge, K. C. (2004). Uncon- times as determinants of post-decisional con- scious emotion. Current Directions in Psycholog- fidence. Acta Psychologica, 98, 103-115 ical Science, 13, 120-123. Zechmeister, E. B., & Shaughnessy, J. J. (lyRo). Winkielman, P., Schwarz, N., Fazendeiro, T. A,, When you know that you know and when you & Reber, R. (2003). The hedonic marking of think that you know but you don't. Bulletin of processing fluency: Implications for evaluative the Psychonomic Society, 15,41-44. The Cambridge Handbook of Consciousness

Edited by Philip David Zelazo, Morris Moscovitch and Evan Thompson University of Toronto

CAMBRIDGE UNWERSITY PRESS CAMBRIDGE UNIVERSI IY PRESS Cambridge, New York, Melbourne, Madr~d,Cape Town, Slngaporr, Sio Paulo

Cambridge University Press 32 Avenue of the Americas, New York, NY 10013-2473,USA wwwcambridge.org Information on this title: wwwcambridge.orgIy7K05z~K5743 7

O Cambridge University Press 2007

This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Camhridge University Press.

First published 2007

Printed in the United States of America

A catalog record for this publication is available from the British Libray.

Libray of Congress Catalogmg in Publication Data The Cambridge handbook of consciousness / edited by Philip David Zelazo, Morris Moscovitch, Evan Thompson. p, cm. Includes bihliographical references and index. ISIIN-13:978-0-521-85743-7 (hardhack) ISHN-10:0-521-85743-0 (hardback) issx-ij: 978-0-521-67412-6 (pbk.) IS~IN-IO:0-521-67412-3 (pbk.) I. Consciousness. I. Zelazo, Philip David. 11. Moscovitch, Morris, 1945- 111. Thompson, Evan. IV. Title.

ISBN 978-0-521-85743-7 hardback Ism q78-0-521-67412-6 paperback

Cambridge University Press has no responsibility for the persistence or accuracy of uni.s for external or third-party Internet Web sites referred to in this publication and does not guarantee that any content on such Web sites is, or will remain, accurate or appropriate.