25
This article was downloaded by: [University of Auckland Library] On: 16 October 2014, At: 19:04 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK European Journal of Cognitive Psychology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/pecp20 Memory monitoring: How useful is self- knowledge about memory? Ian Maynard Begg a , Lisa A. Martin a & Douglas R. Needham a a Department of Psychology , McMaster University , Hamilton, Ontario, Canada Published online: 08 Nov 2007. To cite this article: Ian Maynard Begg , Lisa A. Martin & Douglas R. Needham (1992) Memory monitoring: How useful is self-knowledge about memory?, European Journal of Cognitive Psychology, 4:3, 195-218, DOI: 10.1080/09541449208406182 To link to this article: http://dx.doi.org/10.1080/09541449208406182 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms- and-conditions

Memory monitoring: How useful is self-knowledge about memory?

Embed Size (px)

Citation preview

This article was downloaded by: [University of Auckland Library]On: 16 October 2014, At: 19:04Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

European Journal of CognitivePsychologyPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/pecp20

Memory monitoring: How useful is self-knowledge about memory?Ian Maynard Begg a , Lisa A. Martin a & Douglas R. Needham aa Department of Psychology , McMaster University , Hamilton,Ontario, CanadaPublished online: 08 Nov 2007.

To cite this article: Ian Maynard Begg , Lisa A. Martin & Douglas R. Needham (1992) Memorymonitoring: How useful is self-knowledge about memory?, European Journal of Cognitive Psychology,4:3, 195-218, DOI: 10.1080/09541449208406182

To link to this article: http://dx.doi.org/10.1080/09541449208406182

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with, in relation to or arisingout of the use of the Content.

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

EUROPEAN JOURNAL OF COGNITIVE PSYCHOLOGY. 1992, I ( 3 ) 195-218

Memory Monitoring: How Useful is Self-knowledge about Memory?

Ian Maynard Begg, Lisa A. Martin and Douglas R. Needham Department of Psychology, McMaster University, Hamilton, Ontario,

Canada

The article reports two experiments in which we investigated the value of memory monitoring. Although the act of monitoring may not in itself be a basis for improved memory, it is reasonable to expea that the act of monitoring will provide useful information to the subject. For example, if monitoring identifies items that are inadequately en&, then the inade- quacy can be redressed if the items are repeated. We expected that repeat- ing an item immediately after the subject had evaluated whether the item would succeed or fail later would be more valuable than merely repeating the item without any interpolated monitoring. But it wasn’t. Nor did items people expected to forget benefit more from repetition than ones they expected to remember. Because the findings were unexpected, we recast the issues, and discovered that we should not have been surprised. On the whole, we would be better advised to teach subjects effective ways to study than to teach them techniques that accurately idenm inadequately encoded items; knowing an item is inadequate is of Little value in the absence of skills that will remediate the inadequacy.

INTRODUCTION

Our interest is with people’s self-monitoring of items that they studied earlier. Does self-monitoring help people to improve their memory for the items? A plausible answer is “Self-monitoring prepares subjects to make

Requests for reprints should be addressed to Ian Be%. Department of Psychology, McMaster University, Hamilton, Ontario L8S 4K1, Canada.

Funding was provided by Grant OGwoo8122 from the Natural Sciences and Engineering Research Council of Canada to I.M. Begg; a certificate of e t h i d approval is on file. Pan of the research was conducted as an honours B.Sc. thesis by L.A. MPrtin, supervised by 1.M. Begg. We are especially grateful for the advice of Ann Anas and James Taylor. Lee Brooks, Rohan Robertson and Jeff Toth also provided helpful discussions. Parts of the data were presented by I.M. Begg in a symposium organised by Gene Winognd at the International Conference on Memory. Lancaster, England, July 1991.

0 1592 Lawrence Erlbaum Associates Limited

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

196 BEGGETAL

effective use of feedback; repetition is especially beneficial right after a monitoring trial, primarily because items that people expect to forget show marked improvements when they are repeated.” However, the answer is wrong. In our experiments, monitoring had no measurable value. Repeti- tion effects were as large for unmonitored items as they were for monitored items, and repetition effects were as large for items that people expected to remember as for items they expected to forget. These results led us to rethink the issues. Why were we surprised? We begin by considering the bases on which we expected to find that monitoring is valuable. Then we report an experiment in which the expected results failed to materialise. Next we will reintroduce the research, recasting the issues in the way we should have foreseen but did not. Finally, we report the results of another experiment, in which yesterday’s surprise becomes tomorrow’s prediction.

WHY METAMEMORY SHOULD BE VALUABLE It is commonly thought that it would be useful to have accurate knowledge about one’s memory. For example, Leonesio and Nelson (1990, p. 464) claimed that “self-monitoring is critical for providing input to self-directed control processes such as the allocation of study time”. Manoni, Comoldi and Marchitelli (1990, p. 1%) noted that the “ability to predict future memory performance is an important aspect of the monitoring and control of cognitive processes”. If subjects can accurately idenhfy items that are inadequately encoded, they can allocate resources so that the items receive the additional processing they need to become memorable. We will discuss the accuracy of memory predictions later. First, we will illustrate our procedure for assessing the value of metamemory, and share the reasons that led us to expect to find some.

The Task

Our procedure is a laboratory analogue of the real world of students who are exposed to some material, then we review this material, monitoring memory to determine which items are adequately learned and which need extra study. Our subjects reviewed previously studied items and predicted whether each one would succeed or fail on a future test of memory. We compared two kinds of review, cue review and pair review. In cue review, a subject who previously studied RAILROAD-MOTHER reviews RAIL- ROAD and predicts whether RAILROAD will later succeed or fail as a cue for recalling the missing partner; pair review asks the same question, but the subjects review RAILROAD-MOTHER rather than just RAIL- ROAD. Begg et al. (1989) found that predictions were more highly correlated with later recall if the reviewed stimuli were cues rather than

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

MEMORY MONITORING 197

pairs (0.84 > 0.63), but that both sorts of predictions were reasonably accurate.

How can we tell if monitoring has value for memory? Monitoring has value if it is associated with good recall of monitored items. In the data reported by Begg et al. (1989), the subjects who reviewed pairs did not recall more pairs than the control subjects who restudied the pairs without assessing memorability (0.39 vs 0.44). The subjects who reviewed cues were no more successful in recall than were the control subjects who restudied the cues without making predictions (0.15 vs 0.16), and recall of pairs with reviewed cues was no better than recall of pairs that were not reviewed at all (0.14). However, a fair test of the value of cue review is not whether cue review is itself a good way to study, but whether a subject who has reviewed a cue is prepared to make effective use of opportunities to devote extra processing to the pairs that need it. What if, right after the cue review trial, the entire pair was presented again? Will these repeated pairs be recalled better than ones that are reviewed as pairs or ones that are merely repeated? Thus the contrast of interest is with the level of recall achieved after

people have studied previously presented pairs like RAILROAD- MOTHER a second time. The interesting case is cue review, in which the second appearance of RAILROAD-MOTHER is on the next trial after people predict whether RAILROAD will be a successful or unsuccessful cue for recall of the missing MOTHER. In pair review, the second appearance of RAILROAD-MOTHER is the review trial on which people predict whether they will recall or fail to recall MOTHER with RAILROAD as the cue. For control pairs, RAILROAD-MOTHER is merely repeated without attention to memorability. Will the effect of repetition be most pronounced after cue review?

The Reasons

One reason to expect that cue review will give the best recall of repeated items is by analogy with the generation effect (Begg & Snider, 1987; Slamecka & Graf, 1978). Cue review entails attempted generation of the target, whereas pair review and mere repetition allow reading of the target. When RAILROAD is reviewed, the subject can attempt to retrieve the missing MOTHER, and base the prediction on the success or failure of the attempt. The advantage of generating over reading is of the same magni- tude whether the attempted generation succeeds or fails, so long as there is feedback following the attempt (Kane & Anderson, 1978; Slamecka & Fevreiski, 1983). Hence it is reasonable to expect that the attempt to retrieve a missing target will help memory in the same way that other attempts to generate a complete target from partial information do.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

198 BEGGETAL

Another basis for expecting that cue review will enhance the value of repetition is by analogy with problem solving. Needham and Begg (1991) found that subjects spontaneously used the solution principles from train- ing problems to solve new target problems if they had first tried to solve the training problems; their attempts nearly always failed, but the act of failing prepared them to incorporate the information in the experimenter- provided explanation that followed the attempt. They proposed that the initial attempt is analogous to tilling the soil before planting a seed. Tiffling the soil has potential value that will be in vain without the seed; the seed has potential for growth, but the potential may be unrealised without the advance preparation that tilling provides. Ausubel(1960) has long stressed the importance of advance organisers in learning. Perhaps cue review is an advance organiser; it poses a problem that people try to solve, and the reapptaranct of the pair on the next trial is like the feedback in those experiments.

Whether cue review entails attempted generation followed by feedback or attempted problem solving foflowed by feedback, it is reasonable to expect that cue review will prepare the cognitive system so that a re- presentation of the pair arrives in a fertile cognitive environment. A plausible way for the advance preparation to work is to identify items that are inadequately encoded. Upon monitoring these items, subjects should correctly rate them as least memorable; these items are likely to fail later unless they receive further processing, in which case they should be especially likely to improve.

EXPERIMENT 1 The subjects in Experiment 1 studied pairs like RAILROAD-MOTHER and were tested by cued recall (RAILROATkMOTHER). Between study and test, the subjects judged whether the targets (MOTHER) would be remembered or forgotten in cued recall (RAILROAlh?) . The judge- ment task was either cue review or pair review. Cue review was exactly like the test, in that the cue was presented alone and the subjects predicted whether they would recall or forget the missing target later. Pair review asked the same question, but the entire pair was presented on the review trial.

Our question concerns the mnemonic value of review. If people know that an item will fail but make no use of that knowledge to aid retention, we would be led to question the value gained from possessing accurate information. Previous research has shown that reviewed pairs are not recalled better than unreviewed pairs, but the act of review may have future rather than present value. If so, pairs seen twice with a review trial between the appearances will be recalled better than pairs seen twice

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

MEMORY MONITORING 199

without an interpolated review. If review provides mnemonically useful information, it would make sense for subjects to direct processing to pairs that are inadequately encoded. If so, repetition should aid r e d of pairs that people expect to forget more than pairs they expect to remember.

In summary, we compared the accuracy and value of different ways of reviewing pairs of words. Monitoring is accurate if subjects can predict which pairs will succeed and fail on the later test and valuable if it is associated with high levels of recall. Does monitoring create a cognitive environment in which repetition is especially effective? Does monitoring enable the strategic use of processing to favour needy pairs rather than pairs that are already rich in the right stuff for memory? Subjects studied a running list in which study trials and review trials were interspersed. Each study trial presented a cue-target pair. Review trials presented previously studied pairs (pair review) or just the left-hand members (cue review); the subjects predicted whether they would later recall the target member when given the cue. Half the review trials were followed immediately by a re- presentation of the pair. Each list included unreviewed pairs that were studied either once or twice. Then the subjects were given the left-hand members of each pair as cues for recall of the targets.

Method Subjects. The subject group comprised 43 students of Introductory

Psychology at McMaster University, whose participation met a course requirement. The subjects were tested in four groups, of which two were assigned at random to each experimental condition. There were 22 subjects in the cue review condition and 21 in the pair review condition; an additional 5 subjects were discarded because they failed to recall any words on the test.

Materials. The study list presented % pairs of words for a total of 192 trials; 16 pairs (32 trials) were fillers, with half at each end of the list. The other 80 pairs (160 trials) comprised four conditions, each represented by 20 pairs. There were two sets of control pairs: 20 were presented once and 20 were presented twice, for a total of 60 trials. There were two sets of reviewed pairs: 20 were studied and reviewed (40 trials), and 20 were studied and reviewed but with an extra presentation of the entire pair on the trial immediately following the review trial (60 trials). The body of the list consisted of 11 blocks of 16 trials, 2 of each of 8 types. Three were controls; these included once-studied pairs, the first study of pairs that would be repeated in the next block, and the second study of pairs whose first study was in the preceding block. Two were study trials for pairs that would be reviewed in the next block, with or without repetition after

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

200 BEGGETAL

review. The other three were reviews of items studied in the preceding block; these included the review trial for pairs that would not be repeated, the review trial for pairs that would be repeated, and the repetition of the latter. The trials were arranged in successive blocks so that the lag between the first study of a pair and either its second study or its review ranged from 13 to 20 trials, with a mean of 17. Early and late blocks included fillers to balance the list.

Each trial appeared on a television monitor for 4 sec with a 1-sec blank interval between trials. Review trials included a ?, and were numbered 1 to 48 to correspond to an answer sheet. The subjects circled a numeral from 1 to 5 for each reviewed stimulus: 1 = least likely to remember, 5 = most likely to remember. Two versions of the list were recorded, based on whether review trials included only the cues or entire cue-target pairs.

The final test had 160 words typed in columns. There were 80 new words; these were included to allow conditionalised analyses of recall, but these analyses are not reported because recognition was too good to be interesting. The other 80 words were the left-hand members of all 80 studied pairs. The 160 words were in 10 randomised blacks of 16. To the left of each word was Y(es) and N(o); to its right was a blank space.

Items. The items included 272 nouns with imagery (4 values between 4.00 and 6.50 and Thomdike-Lorge frequency values of 11 or greater, chosen from Paivio, Yuille and Madigan (1968) and from an extended norm list provided by A. Paivio. Thirty-two words made 16 filler pairs, 80 were distractors at test, and 160 made four sets of 20 pairs for the study list. The words were sorted into sets of 20 that were equated for imagery (mean I values were 5.41 or 5.42) and frequency (11 words had F < 50 in each set). These sets were assigned to treatments at random.

Procedure. The subjects were encouraged to follow instructions and to try hard on tests. They were given the rating sheets for the predictions and instructed on their use. They were told to imagine the two members of each pair interacting and to project their images to an imaginary frame above the television monitor. They were told that there would be review trials intersperxd throughout the list, that they would be predicting recall of targets (right-hand members in pair review, absent items in cue review), and that this test would actually be administered. They were also told that half the review trials would be followed by a re-presentation of the full pair. After the list was finished, the review sheets were collected and replaced with cued-recall test sheets; the test was self-paced, and required less than 15 min to complete.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

MEMORY MONITORING 201

Results and Discussion The results will be presented separately for r e d , predictions and recall conditionalised on the predictions. The alpha level for inferences is 0.05 throughout the article, including computation of least si@cant differ- ences (a) to evaluate simple effects. The means are accompanied by MSe values from analyses of variance.

Cued Recull. The mean proportions of items recalled on the hal test are shown in Table 1. The subjects in the cue review condition studied some pairs once, of which some were reviewed later and some were not; recall of these two sets of items was 0.19 vs 0.15. Other pairs were repeated immediately after being reviewed, or were repeated without an interpolated review; recall of these two sets of pairs was 0.34 vs 0.31. The only reliable effect in analysis of variance was a main effect of repetition [F(1,21) = 50.7, MSe = 0.01OJ. Hence reviewing cues did not reliably enhance recall of targets, even if the pairs were repeated after the review.

The subjects in the pair review condition reviewed whole pairs, of which half received an immediate repetition and half did not; recall of these two sets was 0.32 vs 0.32. Analysis of variance revealed an interaction between repetition and review [F(1,20) = 5.35, MSe = 0.011]; recall was better if targets were seen more than once than if they were seen only once, but reviewing had only modest and unreliable advantages for recall.

Memory Predictions. The accuracy of memory predictions was asses- sed by Goodman-Kmskal gamma coefficients (G is 1 if all recalled items received higher ratings than all unrecalled items and 0 if recall rate is equal for all ratings). G was calculated for each subject (6. Nelson, 1984), then analysed by analyses of variance. Cue review enabled more accurate predictions than pair review (0.82 > 0.40; F(1,41) = 13.4, MSe = 0.1391, even if pairs were repeated immediately after the predictions had been made [0.70 > 0.38; F(1,41) = 6.20, MSe = 0.1791.

TABLE 1 Cued Recall of Reviewed and Unreviewed Pain in Experiment 1

Cued Recall

Review Task No €ma SNdy h t r a Study

No review Cue review Pair review

0.1s 0.31 0.19 0.34 0.32 0.32

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

202 BEGGETAL

TABLE 2 R-11 of Pairs People Expected to Forget YS Remember d e n They Reviewed Cues in

Experiment 1 ~

Conditional Recall

“Forget“ “Rmumbrr”

Review task cue review cue review with feedback

0.03 0.19

0.44 0.63

Vdue of feedbrck 0.16 0.19

Noie: “Forget” pain were rated 1 or 2 and “Remember” pain WCR rated 4 or 5 on the memory monitoring t u k .

Conditional Recall. The next contrast is recall of pairs people expected to forget (rating = l,2) or to remember (rating = 4 3 . As shown in Table 2, the subjects in the cue review condition remembered more pairs they expected to remember than pairs they expected to forget [0.53 > 0.11; F(1,21) = 154, MSe = 0.0251 and they remembered more repeated than unrepeated pairs j0.41 > 0.23; F(1.21) = 20.6, MSe = 0.0321. There was no interaction between the variables; the effect of repetition for pairs people expected to forget was 0.19 - 0.03 = 0.16, compared to 0.63 - 0.44 = 0.19 for pairs they expected to remember. Therefore, people did not devote more useful processing to pairs they expected to forget than to pairs they expected to remember if the pairs were re-presented right after their cues were monitored.

The conditionalised means show why C was so higb for cue review; the subjects rarely (0.03) recalled targets they expected to forget. The accuracy of predictions capitalises on a simple fact; if people can’t recall a target at the time of review, they won’t recall it later either unless some form of extra processing occurs. Conventional wisdom has it that “nothing suc- ceeds like success”. These results suggest, instead, that “nothing fails Like failure”.

We also analysed conditional recall in the pair review condition; the only reliable effect was that people remembered more pairs they expected to remember than ones they expected to forget [0.47 > 0.21; F(1.20) = 37.3, MSe = 0.0371.

Summan/ and Conclusions Experiment 1 compared cue review with pair review. The subjects who reviewed pairs achieved better recall than the subjects who reviewed only the cues, which is not surprising because pair review gives subjects an extra

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

MEMORY MONITORING 203

chance to study the target in relation with the cue. Cue review came up to the level of pair review if the pair was repeated immediately after the cue was monitored, but it did not exceed it, and neither review condition led to recall that was reliably above pairs that were studied twice with no review. The only factor that affected recall was how often the pair was presented in its entirety; there was no reliable effect of having review rather than mere presentation. Memory predictions were also not helpful in directing sub- jects so that they could gain benefits from extra study of pairs they expected to forget rather than remember; both achieved the same benefit from repetition. In summary, memory monitoring detects inadequately encoded pairs, but the detection does not help memory directly, or prepare the subject to gain from repetition, or inform the subject to devote effective processing to pairs that need it. As in previous research, predictions based on cue review were more

accurate than predictions from pair review. The accuracy of the predictions about cues was largely because cue-review subjects almost never recalled targets they expected to forget; we will expand this point later.

REINTRODUCTION

In light of the results of Experiment 1, we now reorient the issues, and introduce the problem in the way we should have introduced it in the first place. Previous investigations of memory monitoring have been concerned with the accuracy of the predictions people make about future tests.

Accuracy

There have been many investigations of the accuracy of various types of memory predictions. Hart (1965; 1%7), for example, found that people have a modest ability to predict which unrecalled answers they will recognise later. Other studies have shown that people can anticipate that some kinds of items will be remembered better than others (e.g. Lovelace, 1984; Vesonder & Voss, 1985). For example, Underwood (1966, pp. 468- 470) found correlations of 0.92 and 0.91 between judged ease of learning and actual learning of items like BUG and VXK; however, learning was also highly correlated with rated meaningfulness and pronounceability (0.88 and 0.90), which makes the accuracy of predictions less impressive.

Although people can discriminate between items that will succeed and those that will fail, their predictions are often less accurate than are other prognosticators. For example, future recognition of unrecalled answen is predicted better on the non-metamemorial, actuarial basis of how many people know the answer than it is by people’s personal “feeling-of- knowing” judgements (Nelson, Leonesio, Landwehr & Narens, 1986).

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

204 BEGGETAL

Also, the accuracy of memory predictions is less than the accuracy of predictions based on the number of trials to acquisition, even though that information is potentially available to the subjects (Leonesio & Nelson, 1990). Begg et al. (1989) compared several predictors of future recogni- tion, and found that explicit memory predictions were less accurate predic- tors than were ratings of ease of imagery; memory predictions were more highly correlated with rated familiarity than they were with actual recogni- tion memory.

One factor that increases the accuracy of memory predictions is to base them on memory; predictions are more accurate if the items were pre- viously studied (King, Zechmeister & Shaughnessy, 1980) or tested (Love- lace, 1984). But memory-based predictions are less correlated with final recall than is immediate recall ( B e g et al., 1989), even when subjects have the opportunity to use immediate recallability as the basis for their predic- tions. A limit to accuracy is that predictions tend not to anticipate which ways to study will produce higher levels of retention (e.g. Cutting, 1975; Perlmuner, 1978; Rabinowitz, Ackerman, Craik & Hinchley, 1982; Shaughnessy. 1981; Shaw & Craik, 1989; Zechmeister & Shaughnessy, 1980).

Processing

Begg et al. (1989) proposed that memory predictions are heuristic attnbu- tions based on subjective experiences that occur when items are processed. Subjects expect to remember items that are easiest to process and to forget items that are hardest to process. But ease of processing is a non-analytic basis for judging memorability because not all Causes of easy processing are associated with good retention. H e n u predictions are most accurate when the factors that cause items to succeed and fail in memory also Cause items to be processed differentially easily on the predictive test, and least accurate when the factors that cause differences in memory have different effects on ease of processing. For example, low-frequency items are recognised better than high-frequency items, but high-frequency items, which are easier to process, are predicted to be more memorable than low- frequency items.

If inter-test correlations reflect the similarity of the processes entailed by the tests, then predictions should be most accurate when the predictive tasks resemble the final test. Memory predictions for previously studied items are more accurate than predictions made at study because predic- tions at review entail retrieval, and items that are more and less easily retrieved upon review will also differ in memorial success; items that people can’t retrieve at review will not be remembered later without remediation. Begg et al. (1989) expected that predictive accuracy would

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

MEMORY MONITORING 205

depend on whether the predictive task and the test entail the Same processes, not on the nominal predictive question. Some subjects reviewed RAILROAD-MOTHER and predicted whether they would later r e d MOTHER with RAILROAD as the cue; this is the right question but, by including the entire pair at review, the predictive task entails the wrong process. Other subjects reviewed only RAILROAD and judged whether they would recognise the absent MOTHER later; this is the wrong ques- tion, but subjects need to perform the right process to answer it. Accuracy reflected whether the process was right but not whether the right question was asked.

Presenting the whole pair for review reduces the accuracy of predictions because it spoils the value of the experience of retrieving the target. Jacoby and Kelley (1987) used a similar logic with anagrams. Their subjects rated the difficulty of anagrams that were presented alone or were accompanied by the solution (e.g. SRACF vs SRACF-SCARF). If the solutions were present, the subjects rated anagrams as easier, and the ratings were less highly correlated with solution times and with actual difficulty than if the anagrams were presented alone. Why? The subjective experience of solv- ing an anagram is correlated with difficulty, but the presence of the solution spoils the diagnosticity of subjective experience. In a similar fashion, predicting memory for RAILROAD-? can be based on the ease of retrieving the missing target, but this expcrience is spoiled if the question is asked about RAILROAD-MOTHER. The latter is a post- diction about whether one could have recalled MOTHER had MOTHER been absent; it is well known that postdictions are subject to hindsight bias (Fischhoff, 1975).

Accuracy Revisited

Attributions based on subjective experiences may predict outcome mea- sures, but their accuracy is fragile and dependent on being lucky enough that current experiences are caused by test-appropriate processes. Further- more, we must not confuse a high G with accurate knowledge. In the simple case of a 2 x 2 contingency table comparing predictions (remember vs forget) with outcomes (success vs failure), G = 1 if either of two cells is 0; that is, if subjects never forget items they expect to remember (this would be impressive, but is rarely observed), or if they never remember items they expect to forget (this is less impressive, but is close to the truth). Given this character of the statistic, an optimal strategy for achieving a high G is to predict later failure for all currently unmemorable items, and to adopt a lenient criterion for assessing current memorability. Then all items currently thought to be unmemorable will likely be unmemorable. Such a strategy would merely increase the correlation between predictions

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

206 BEGGETAL

and outcomes; it would not truly indicate increased accuracy of self- knowledge.

But why should later recall be predictable? Accuracy cannot be perfect because future success or failure is not fully determined at the time when predictions are made. Furthermore, subjective experiences cannot give perfectly accurate self-knowledge, because they are effects of processing; reasoning from effect to cause is a logical error, called affirming the consequent. Items may be easy to process for many reasons, some of which are side-effects without functional significance. By analogy, the sound caused by a car has no functional value for locomotion. Are effects that can be witnessed important, or are they side-effects? Does witnessing the effects provide useful information about the processes that caused them?

Labouring in Vain The major reason we should have expected to find that explicit monitoring is of little value is on the basis of what Nelson and Ltonesio (1988) called the “labour-in-vain” effect; when people allot more study time to some items than to others, they usually don’t remember more of the ones they spent more time on. Mazzoni et al. (1990) found that predictions reliably discriminated recalled from unrecalled items and that their subjects allot- ted study time differently to items thought to differ in memorability. However, the items that the subjects studied longer were not better recalled than those studied for less time (see also Mazzoni & Cornoldi, in press). If people give extra study to deficient items, but extra study fails to overcome the deficiency, it would be more useful to teach subjects how to study than to help them improve the accuracy of their monitoring of deficiencies.

With the benefit of hindsight, we now have the foresight to propose that self-monitoring is itself labour in vain. In making a prediction about memory, the cogniser intentionally directs attention to the contents of mind, and evaluates their adequacy to support a future memory test. Self- monitoring will give accurate information about the later task if the processes used when monitoring are the same as the processes needed to succeed on that task. But bringing a memory to mind as the object of attention has no direct value for later memory. If cognitive processes are the drive shaft of the mind, then their conscious effects are the noise of the engine; they indicate that something is happening, but do not participate in that something beyond witnessing it.

On the basis of this account, we predict that even when monitoring accurately discriminates between items that will succeed and fail, monitor- ing will have no mnemonic value; reviewed items will not be remembered better than unreviewed items even if the items are repeated after the

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

MEMORY MONITORING 207

query, and the increased memorability of repeated items will be no greater for items predicted to fail than for items predicted to succeed. Predictions are a form of introspective witness; even when they accurately indicate the state of the system, they have no value for memory.

In Experiment 2, we investigated the mnemonic value of memory monitoring in three ways. First, a monitoring trial is a study occasion; if items are reviewed after study, are they remembered better than items that are merely repeated without attention to memorability? Second, monitor- ing allows advance preparation so that additional study opportunities are especially effective; are repeated items remembered better if there was a predictive review before the repetition than if there was not? Third, monitoring informs the subject which items need remedial processing; do the items people expect to forget gain more from repetition than the items they expect to remember? We expected that each of these questions would have the same answer-no.

EXPERIMENT 2 Experiment 2 included study, review and test, with several changes from the procedure used in Experiment 1. Study and review were separate tests rather than interspersed. The subjects first studied a list of pairs, and then they received a review list of cues or of pairs and predicted whether they would remember or forget each one. In Experiment 1, feedback was unpredictable, because only half the review trials were followed immedi- ately by a repetition of the pair. In Experiment 2, feedback was predict- able; for some subjects, every reviewed item was repeated immediately after the review trial. Finally, after the review list, there was a second study list that repeated some reviewed and unreviewed pairs.

We compared recall of pairs reviewed in different ways with recall of pairs seen the same number of times without any review. Once-studied pairs included control pairs that were studied once without review and pairs whose cues were reviewed. Twice-studied pairs included control pairs studied twice without review, cue review with feedback, cue review with delayed repetition, and pair review. There were also pairs that were seen three times; these included pair review with immediate repetition, pair review with delayed repetition, and cue review with both immediate and delayed repetition.

We expected that memory predictions would be most accurate in cue review conditions because the experience of retrieving the missing target is unspoiled by having the target physically present. Final recall should track the number of times the entire pair was studied, with bigger benefits from delayed than from immediate repetition. But will the different types of review influence recall beyond the level achieved by mere repetition? Will

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

208 BEGGETAL

it matter if the second appearance of a target is after a cue review, during a pair review, or without review at all? If predictions are accurate, recall will be higher for pain people expect to remember than ones they expect to forget. But will repetition be especially beneficial for pairs people expect to forget?

Method Subjecfs. Altogether, 86 students of Introductory Psychology at

McMaster University served as subjects to meet a requirement of the course. They were tested in eight groups, of which two were assigned at random to each of four experimental conditions, with from 17 to 25 subjects in each one. Six other subjects were discarded because of failure to recall any words on the test.

Materials and Procedure, The item pool was 100 pairs of nouns, with five subsets of 20 pairs each. The 200 words were selected from Paivio et al. (lW), with I values between 4.00 and 6.50, and frequency values of 11 or greater. The nouns were distributed into 10 sets of 20 that were matched for frequency and imagery; each set had 9 words whose frequency was 50 or greater and 11 whose frequency was between 11 and 49 inclusive, and the mean I values ranged from 5.41 to 5.52. The sets were selected and paired at random, yielding five sets of 20 pairs.

The five sets of pairs were sorted into three lists-the initial study list, the review list and the final study list. Two sets were in the initial study list and in the review list; one set was reviewed but was not repeated in the final study list, and the other was reviewed and then was repeated in the final study list. Control pairs appeared in the two study lists but not in the review list; these included once-studied pain that were only in the initial study list, once-studied pairs that were only in the final study list, and twice-studied pairs that were in both study lists. Hence the initial study list included 80 pairs, the review list included 40 of them, and the final study list included 60 pairs; 20 pairs had been initially studied then reviewed, 20 pairs had been studied but not reviewed, and 20 pairs were appearing for the first time. Each list was block randomised so that all sets were equally represented in each block; the size of the blocks was eight pairs for the initial study list and the review list, and six pairs for the final study list.

The lists were recorded on video-cassette from a computer-driven dis- play; each stimulus was on the screen for 4 sec, with a 1-sec blank interval between stimuli. The study lists displayed pairs with the members side by side. Four versions of the review list were recorded. In the cue review conditions, the left-hand member appeared, followed by dashes and a question mark. The items were numbered from 1 to 40 to correspond to a

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

MEMORY MONITORING 209

rating sheet, on which the subjects were to circle a numeral from 1 (least likely to remember) to 5 (most likely to remember). The pair review task was the same except that the right-hand member replaced the dashes. In both cases, the subjects judged whether they would be able to recall the right-hand member with the left-hand member as the cue. The other variable was feedback. One group receiving each type of review saw the complete pair in the next stimulus interval; hence the lists with feedback required twice the presentation time as the lists without feedback.

After all three lists had been presented, the subjects received a test. The 100 left-hand members of the pairs were cues for written r e d of the right- hand members. The test was self-paced, and took about 10 min.

The subjects were fully instructed about each requirement of the task, and they were encouraged to do their best to follow all instructions, and to try hard on all tests. They were told that they would see three lists, that the second list would include questions about some items, that some items would be repeated over lists, and that there would be a test at the end. All of the subjects were instructed to imagine the two members of each pair interacting with each other, and they were told that the imagery procedure would help them remember many of the pairs. Groups receiving the four Merent review lists were told what form the query would take. Rating sheets were handed out before the review list, and taken back in before the final study List, after which the recall test sheets were handed out; the imagery instructions were repeated before the final list.

In summary, review type (cue vs pair) and feedback (present vs absent) were between-subjects factors and delayed repetition (present vs absent) was a within-subjects factor. Also within subjects, there were unreviewed control pairs that appeared once or twice.

Results and Discussion In all, four groups of subjects studied 100 pairs that comprised five sets of 20 pairs that were treated differently within subjects. The following sec- tions present final recall, the accuracy of predictions and recall conditional- ised on predictions.

Recall. Mean final cued recall is shown in Table 3; comparisons be- tween control YS reviewed pairs are reliably different if the mean difference is 0.064 or greater (MSe = 0.0109), and comparisons within the reviewed pairs are reliably different if the two means differ by 0.056 or more (MSe = 0.0084). The control pairs have been averged over the four review condi- tions because there were no reliable differences among them. The recall of the three sets of control items differed from each other [F(2,164) = 161, MSe = 0.01141. Repeated pairs were recalled best (0.45). Of once-studied

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

210 BEGGETAL

TABLE 3 Cued Recall of R e v i s w e d and Unreviewed Pairs in Experiment 2

Cued R e d

R w i m Tptk No FdShrdy With F d S h c d y

No mieW cue review Cue rrviear with feedback P8ir revim Pair review with feedback

0.15 0.15 0.37 0.36 0.44

0.45 0.48 0.61 0.64 0.65

pairs, recall was better if that one study was in the final list rather than the initial list (0.27 < 0.15); only the latter is included in Table 1.

Explicit monitoring had no mnemonic value. Cue review had no direct value for recall; pairs whose cues were reviewed were not recalled better than unreviewed pairs (0.15 ws 0.15). Cue review did not prepare the system to take advantage of delayed repetition; pairs studied in both lists were not recalled better if their cues were reviewed between the repetitions than if they were merely repeated (0.48 ws 0.45). Cue review did not prepare the system to take advantage of feedback; recall was no better if cue review was followed immediately by a re-presentation of the pair than if the pair was reviewed (0.37 ws 0.36), and these two conditions remained nearly equal if the pairs received an extra presentation in the final list (0.61 ws 0.64).

Memory Prcdxtions. Predictive accuracy was assessed by gamma coef- ficients (G). Cue review allowed more accurate predictions than pair review, both without feedback (0.86 > 0.51) and with it (0.59 > 0.40); analysis revealed main effects of type of review and of feedback [F( 1.82) = 14.6, 6.88; MSe = 0.1071. Accuracy in the pairs that received an extra repetition revealed no reliable effects (mean G = 0.45, MSe = 0.149).

Condifionul Recall. Mean recall of pairs that people expected to re- member (rating = 4.5) or forget (rating = 1.2) when they reviewed the cues is shown in Table 4. Analysis of variance revealed main effects of all three variables. Recall benefited from feedback [0.53 > 0.37; F(1,36) = 7.74, MSe = 0.1351 and delayed repetition [0.58 > 0.34; F(1,36) = 81.4, MSe = 0.028], and recall was higher for pairs people expected to remem- ber than for pairs they expected to forget (0.62 > 0.29; F(1,36) = 102, MSe = 0.0401. There were no reliable interactions. For pairs that were not

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

MEMORY MONITORING 211

TABLE 4 R e c a l l of Pain People Expeaed to Forget vs Remember when They Reviewed Cue0 in

Experiment 2

~

No F d S d y With FdSNdy

“Forger” “Runnnbc)’ “Forget“ “Remember”

Review task cue review 0.03 0.43 0.38 0.64 Cue review with feedback 0.23 0.62 0.50 0.78

Value of feedback 0.20 0.19 0.12 0.14 ~ ~ ~ ~~~ ~~ -

Norc “Forget” pain were rated 1 or 2 and “Remember” pain were rated 4 or 5 on the memory monitoring task.

repeated on the hal list, feedback had about the same value for pairs people expected to forget (0.23 - 0.03 = 0.20) as for pairs they expected to remember (0.62 - 0.43 = 0.19); for pairs that were repeated on the final list, the value of feedback for “forget” vs “remember” pairs was also about the same (0.50 - 0.38 = 0.12; 0.78 - 0.64 = 0.14). Therefore, feedback was no more valuable for pairs peopie expected to forget than for pairs they expected to remember.

The interaction between “forget” vs “remember” pairs and delayed repetition approached reliability [F(1,36) = 3.66, P = 0.06, MSe = 0.0371. Without feedback, the effect of delayed repetition was greater for “forget” pairs than for “remember” pairs (0.38 - 0.03 = 0.35; 0.64 - 0.43 = 0.21); however, the “forget” pairs had more room to improve, and the relative change was about the same for the two (0.36 vs 0.38). Similarly, if the pairs were repeated right after cue review, the effect of delayed repetition was greater for “forget” than for “remember” pairs in absolute terms (0.50 - 0.23 = 0.27; 0.78 - 0.62 = 0.16) but not in relative terms (0.34 vs 0.42). Thus the effect of delayed repetition is greater for “forget” than “remem- ber” pairs, but the size of the improvement may be an artifact of the available range for improvement; relative to the range, the advantage disappears.

The pair review conditions provided no new results. Feedback (massed repetition) did not reliably help recall, but delayed repetition did (0.65 > 0.45; F(1,46) = 66.5, MSe = 0.0281, and recall favoured “remember” pairs over “forget” pairs t0.70 > 0.40; F(1,46) = 78.0, MSe = 0.0541. The value of delayed repetition was 0.20 for “forget” pairs and 0.19 for “remember” pairs.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

Summary and Conclusions Final recall was a function of the number of times the target and cue had appeared together. Recall of once-studied pairs did not benefit from review of the cue alone. Recall of twice-studied pairs did not benefit from having a cue review trial interpolated between initial study and delayed repetition. Recall of twice-studied pairs was also about the same if the second study was a pair review trial or if it was feedback after a cue review trial. R d of thrice-studied pairs was best if the third presentation was in the final list, but recall was equally good if the second presentation had been a pair review trial or had been feedback after a cue review trial. There is no evidence that reviewing helps memory more than would be expected from the number of times the pair is seen.

Conditional analysts of recall examined the future success or failure of pairs people expected to remember or forget. These data revealed that repetition was equally valuable for pairs people expected to forget as for pairs they expected to remember. In cue review, using memory as an object indicates which pairs will fail unless steps are taken to remediate them when they are repeated. However, subjects either fail to avail themselves of the opportunity any more than they do with pairs they expect to remember, or they invest the labour in these pairs but do so in vain.

The most accurate memory predictions were in the cue review condition, but the higb values of G do not mean that subjects have accurate self- knowledge. They mean only that if a target is irretrievable at review, it will still be irretrievable at test unless something happens. Although G has good psychometric properties for these types of data, interpretation re- quires looking at the values in the cells of the matrix; G is high because subjects rarely recall items they expect to forget. If subjects monitor themselves leniently, they will expect to remember more items than they actually do, with very small values in the “recall items expected to be forgotten” cell. A smart look at the matrices is sufficient to prevent one from mistaking a high G for accurate self-knowledge.

GENERAL DISCUSSION The experiments in this article provide clear support for our thesis; memory monitoring does not make a valuable contribution to memory. Even when people can accurately monitor memory to determine which items are most likely to fail on a final test of recall, those items benefit no more from repetition than do the items people expect to remember. Monitoring is no more valuable than merely re-studying without explicit attention to memorability, either directly or in preparing the system to take advantage of repeated opportunities to study items.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

MEMORY MONITORING 213

Why was monitoring valueless? The simplest explanation is that subjects do not know how to study effectively. They may devote more time and resources to items they think are in need, but the time is used ineffectively and the additional processing they invest reaps no rewards. Why were we surprised by the results? Our surprise likely reflects our mistaken belief that subjects can use their metacognitions to help them tailor and control encoding processes to meet perceived needs.

What Next? A fruitful area for research would be to find ways to eliminate the labour- in-vain effect. The original demonstration by Nelson and Leonesio (1988) was based on instructions for subjects to spend as much time with each item as they needed. Perhaps this instruction tacitly biases subjects to use massed practice for each item. Would labour have been in vain if the subjects had been told to go through the list as open as they needed? This would bias them to distribute their practice. Similarly, the procedures of Mazzoni et al. (1990; Mazzoni & Cornoldi, in press) define labour as massed time spent studying items. Again, if the freedom to allot differen- tial study based on perceived need was freed from the procedural demand to do so all at one fell swoop, perhaps additional study would be of more

We suspect that monitoring would have value if the experiment induced people to use an ineffective process when studying. Begg. Vinski, Frank- ovich and Holgate (1991) found that memory for generated items exceeds memory for read items if the read items are processed by a superficial procedure, such as pronouncing the items, but not if the read items are processed by a meaningful procedure, such as imagining them. The present experiments included imagery as the basic study procedure for all condi- tions. If the basic study procedure was pronunciation, perhaps people would re-pronounce the items they expect to remember but use some other procedure to re-study the items they expect to forget. It might be worth- while to instruct subjects that if an item seems unmemorable, they should study it diffeerenzly upon repetition, but that if it seems memorable, they should stick with the original study procedure; that is, “don’t fix it if it ain’t broke”.

U S .

The Monitoring Process Although memory monitoring did not contribute to improved memory for items, we should not lose sight of the fact that monitoring did provide potentially valuable information to the cogniser. In particular, cue review subjects accurately identified items that would later fail. The processing

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

214 BEGGETAL

account (Beg et al., 1989) proposes that memory predictions are based on differential ease of processing of reviewed stimuli. Subjects identiQ inade- quately encoded items because the monitoring process is similar to the actual recall process. The factors that cause failure in hal cued recall will also, on an earlier d o n , cause retrieval of the target to fail or to be difficult. Predictions were less accurate in pair review conditions because the presence of the target at review makes the determinants of ease of processing different from what they will be at test, when the target is not physically present.

Although differential ease of processing is the heuristic basis for memory predictions, it is not the only basis for predictions. For example, those subjects who were told of the negative relation between recognition and frequency were able to ovemde their bcuristic biases and predict better memory for low- than high-frequency words (Begg et al., 1989). Similarly, predictions at study are usually insensitive to different ways to study, but subjects predict an advantage for interactive imagery over separate imag- ery if both processes are used in the same list (Begg t t al., 1989). Also, generating is more effortful than reading, but subjects predict better memory for generated than for read items (Begg et al., 1991). In summary, memory predictions are based on how easily items are processed, but they can be modified to include additional information.

Additional Concerns

One concern is that memory monitoring may be harmful if the predictions have an inappropriate basis. Begg and Green (1988) speculated that there could be a metamemorial basis for the small effects of massed repetition; perhaps items are so easily processed on immediate repetition that they give the illusion of being highly memorable, and people do not process them beyond noting their reappearance. Identifymg inadequate items will be valuable if people know how to fix the problem, but misidentifying items as being adequate may have a negative effect.

Misidentification may be the rule rather than the exception when judge- ments are based on heuristics like ease of processing. Because items may be easily processed for many reasons, judgements are subject to illusions when the experimenter chooses tasks appropriately. Items that are easily processed at test give the illusion that they have occurred frequently (Tversky & Kahneman, 1973), the illusion that they are presented in the absence of background noise (Jacoby, Allan, Collins & Larwill, 1988), the illusion that they are presented for long exposure durations (Witherspoon & Allan, 1985) and even the illusion that they are old (Jacoby & White- house, 1989; Whittlesea, Jacoby & Girard, 1990). Easily processed items seem familiar, and this familiarity gives repeated statements the illusion of

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

MEMORY MONITORING 215

being true (Bacon, 1979; Begg, Armour & Kerr, 1985; Begg & h o u r , 1991; Begg, Anas & Farina&, in press) and repeated names the illusion of being famous (Jacoby, Kelley, Brown & Jasechko, 1989a; Jacoby, Woloshyn & Kelley, 1989b). Memory monitoring is another case in which the subjective experience of processing an item influences judgements about the item, and it too can lead people to take inappropriate action based on the heuristic, but wrong, effects of ease of proctssing.

Another concern is that people’s ideas about memory contain miscon- ceptions that, if unopposed, wil l inappropriately influence perception, classification, judgement and action. For example, people’s confidence in their memory need not mean it is accurate. John Dean’s confident testi- mony during the Watergate hearings led many to believe that his memory was accurate, but subsequent comparisons between his testimony and the tape-recorded events showed many discrepancies (Neisser, 1981). Simi- larly, the widespread misconception that memory is improved by hypnosis (Loftus & Loftus, 1980) is likely because hypnosis influences confidence (Wells & Murray, 1984). In a review of the relation between confidence and accuracy in hypnotic memory, Krass, Kinoshita and McConkey (1989) concluded that hypnotised subjects are as confident as controls when hypnosis reduces accuracy, and more confident than controls when hypno- sis has no effect on accuracy. Misconceptions about memory may be very difficult to override; for example, Kroll, Schepeler and Angin (1986) found that subjects think they have remembered more bizarre than common images even when they have not. Such misconceptions may even be why we so often forget the special place in which we stored an important object; Winograd and Solway (1986) proposed that people think they will be able to relocate objects if the location makes them think of the object, although it is more often the case that we h o w the object and need to retrieve the location.

Final Words Although we were unable to find any value for memory monitoring, we believe that memory monitoring and other metacognitions are important to investigate. Metacognitions may be wrong, and monitoring may lead to useless or even harmful activities, but people’s ideas are powerful determi- nants of their behaviour. A more complete understanding of the processes of metamemory will help us in our goal of arming our students with useful skills. We hope our results fail to generalise to the real world, so that we can soundly reject our proposal. But we cannot let our hopes blind us to whatever reality emerges in such investigations.

Manuscript received September 1991 Revired rnurusuipt received Febnury 1992

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

216 BEGGETAL

REFERENCES

Ausubel. D.P. (1960). l%e use of advance oq&n in the learning m d retention of

Blcon, F.T. (1979). Credibility of r epu ted statemeots: Memory for trivia. J o d of

Beg. I.M., Ana. A. & Fuinrcd, S. (iu pras). Dissociation of prootsscs in belief: Source moktioo. statement funiliuty. and the illusioo of truth. J o d of Erpc- Psychology: General.

Beg, I. & Anwur, V. (1991). Repetition pnd tbe riog of truth: Biariag commeots. CMOdw J o d of &Iwwrd Scioct. 23,195-213.

Beg. I. & Green, C. (1988). Repetition and trwr interaction: SuperpddiUVity. Memory Md C o g d o n , 16.232-242.

Begg. 1. & Snider, A. (1987). Tbc geocratioo cffcct: Evidcncc for poenlkd inhibition. J o d of Experimcnd Psychology: Lcorning. Memory Md CkgnUion, 13.55>563.

Begg, I . , h o u r . V. & Kerr. T. (1985). 00 belicviog whnt m remember. CpMdw J o d of Behavioral Scimcc, 17, 199-214.

Beg, I., Duft, S.. LAO&, P., Melnidt, R. & Wvito, J. (1989). Memory predictions are baaed on use of procasing. Jownol of M e m o r y and b p o g e , 28,610-632.

Bcgg. I., V i . E.. Fraokovich, L. & Holgate, B. (in press). Generating m a k a words wmomble. but so dots effective mading. Memory and Cognirjon.

Cutting, J.E. (1975). Orienthg tasks affect 4 pcrformaoce more tbao subjective im- pressions of ability to r d . Psycholopad Rrpom, 36,155-158.

F-, B. (1975). Hiadsight & foresight: The effect of outcome knowledge in judgment under uacertainty. J o d of Experinmud Psychology: Human Perceplion and Perform- Mu. I , 288-299.

Hart, J.T. (1965). Memory pod the feeling-of-knowing experience. J o d of E M o n a l Psychology, 56,208-216.

Hart, J.T. (1967). Memory and the memory-monitoriog process. J o d o f V e r M Learning and V e h d Behavior, 6.685-691.

Jacoby, L.L. k KeUcy. C.M. (1987). Unconscious inllucocts of memory for a prior event. Personality and social Psychology BuuCrin, 13.314-336.

Jacoby, L.L. & Whitehow, K. (1989). An illusioo of memory: False recopition influenced by unco&ous pcraption. J o d of Experimrnral Psychology: General, 118. 126-135.

Jacoby, L.L., Allan, L.G.. Collins, J.C. k LuwiU, L.K. (1968). Memory ioflucocessubjcc- tivc expericocc: Noise judgments. Journal of Experinvntal Psychology: Learning, Mem- ory and Cognition, 14.240-247.

Jacoby, L.L., KeUey, C.M., Brown. J. & Jasechko, J. (1989a). Becomingfamousoverni%t: Limits on the ability to avoid unconscious in8uences of the past. Journal of Persodiry and Social Psychology, S6, 326-338.

Jacoby, L.L., Woloshp. V. k Kclley, C.M. (1989b). Becoming famous without being recognized: Unconscious influences of memory provided by dividing attcntioo. l o u d of Experimental Psychology: General, 118. 115-125.

Kane, J.H. & Andemn, R.C. (198) . Depth of processing and interference effects in the learning and remembering of sentences. J o d of Educational Psychology, 70. 626-635.

King. J .F. . Zechmeister, E.B. & Shaughnessy. J.J. (1980). Judgments of knowing: The influence of retrieval practice. American J o d of Psychology, 93, 329-343.

Krass. J . , Kinoshita. S. & McConkey, K.M. (1989). Hypnotic memory and confident reporting. Applied Cognitive Psychology. 3 , 35-52.

Kroll, N.. Schepelcr. E. & Angin. K. (1986). Bizarre imagery: The misrcmembered mnemo- nic. Jouml of Eiperimmtal Psychology: Lcarning, Memory and Cognition. 12, 42-53.

Leonaio. R.J. & Nelson. T.O. (1990). Do different metamemory judgments tap the same

h g f u l verbal k.ming. J o d of E i k a t w d ' P ~ y c h o i ~ m , 51,267-272.

P$ychology: H w ~ n Lr&g Md M r m ~ r y . 5,241-252.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

MEMORY MONITORING 217

&dying rrpecb of w r y ? J o ~ of Ex- Psycho&gy: w g , Memory and Co@ion. 16.464-470.

k n . ~ m n i c ~ PWW. 35, m u m . Loftus, E.F. & hftus, G.R. (19ao). On the permmenre of storcd information in the human

Lovelaa, E.A. (1984). Metlwmory: Monitoring future redlability during study. J o d of

M u l o n i , G . & Cornddi. C. (in press). Strategies in study time allocation. J o d of Expminund Psychology: General.

Mazmni, G.. Comoldi. C. & Muchitelli, G. (1990). Do memorability ratings affect study- time docation? Memory and Cbgnition, 18,196204.

Needh.m. D.R. & B e g , I.M. (1991). Roblemaientcd processing promotes spontaneous .nJogical transfer: Munory-orientcd procesSiag promotes memory for training. Memory and Cognition, 19,543-557.

Psychology: Learning, Memory and Cognition, 10,756-766.

Neisser. U. (1981). John Dean’s memory: A case study. Cognition, 9.1-22. Nelson, T.O. (1984). A comparison of aurent measures of the accuracy of fteling-of-

knowing prcdictiom. Psychologid BuUmn. 95,109-133. Nebon. T.O. & Leonesio, R.J. (1988). Allocation of self-paced study time and the “labor-in

vain” effect. Jownol of ExperimclJp1 Psychology: L a m i n g , Mmry d Cognition. 14. 676-684.

Nelson, T.O., Leo&. R.J., Landwehr, R.S. & Nuens, L. (1986). A comparison of three predictors of cue individual‘s memory perfomma: The individual’s feeling of knowing versus the normative feeling of knowing vemu base-rate difficuity. J o d of Experimen- rpI Psychology: Lcarning, Memory and Cognition, 12, 279-287.

Paivio. A.. Yuille. J.C. & Madigan, S.A. (1968). Concreteness. imagery. and meaningful- ness values for 925 nouns. J o d of Exprimend Psychology Monograph, 76, No. 1. Part 2.

Perlmutter. M. (1978). What is memory aging the aging of? Developmrmol Psychology, 14. 330-345.

Rabinowitz, J.C.. Ackerman. B.P.. Craik, F.I.M. & Hinchley, J.L. (1982). Aging and mctamemory: The roles of relatedness and imagery. J o d of Gerontology, 37,688-695.

Shaughnessy, J.J. (1981). Memory monitoring accurrcy and modification of rehearsal strate- gies. J o d of Verbal burning and Verbal Behavwr. 20,216-230.

Shaw, R.J. & Craik, F.I.M. (1989). Age differences in predictions and perfomme on a cued recall task. Psychobgy and Apng , 4 , 131-135.

Slamecka, N.J. & Fevreiski, I . (1983). The generation effect when generation fails. Jouml of Verbal Lcarning and Verbal Behavior, 22, 153-163.

Slamecka, N.J. & Graf, P. (1978). The gencntion effect: Delineation of a phenomenon. l o u d of Experitneruol Psychology: Human Lcaming and Memory, 4,592604.

Tvenky, A. & Kahneman. D. (1973). Availability: A heuristic for judging frequency and probability. Cognitive Psychology, 5 , 207-32.

Underwood. B.J. (1966). Exprrimmtd psychology, 2nd edn. New York: Appleton- Century-Crofts.

Vcsonder, G.T. & Voss. J.F. (1985). On the ability to predict one’s own responses while learning. Journal of Memory and h g u a g e , 24. 363-376.

Wells, G.L. & Murray, D.M. (1984). Eyewitness confidence. In G.L. Wells & E.F. Loftus (Eds), Eyewimss testimony: Psychological perspectives. New York: Cambridge Univer- sity Press.

Whittlesea, B.W.A.. Jacoby, L.L. & Girard, K. (1990). Illusions of immediate memory: Evidence of an attributioaal basis for feelings of familiarity and perceptual quality. l o u d of Memory and Lunguage. 29, 716-732.

Winograd. E. & Solway, R. (1986). On forgetting the locations of things stored in special plam. J o u d of Experinuntol Psychology: General, 115.366-372.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014

218 BEGGETAL

Withcnpoon, D. & Allm. L.G. (1985). The effects of a prior presentation on temporal judgments in a perceptual identification task. Memory and Cognition, 13,101-111.

Zedmeher, E.B. & Sbughacssy, JJ. (1980). Wben you bow that you know and when you think YOU knoo, but don’t. Bullriin of the Psychonomic socirry, IS, 41-44.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

uckl

and

Lib

rary

] at

19:

05 1

6 O

ctob

er 2

014