Psychology of Metamemory Research Paper

What makes the investigation of metacognition scientific is that the theories of metacognition attempt to account for empirical data about metacognition. Related to that, one of the oldest topics in psychology is the topic of consciousness, and early twentieth-century textbooks about psychology frequently defined psychology as the scientific investigation of consciousness. Theories of an individual’s cognitions about his or her own cognitions may appear to be similar to what some people would regard as consciousness, especially self-consciousness. This is not surprising, because the development of theories of consciousness can be affected by the empirical findings about metacognition, in at least two ways (Nelson 1996): first, the empirical findings pose a challenge to theories of consciousness insofar as such theories should be able to account for the empirical findings about how people monitor their own cognitions, and hence such theories can sometimes be disconfirmed by particular empirical findings; second, the empirical findings may provide clues that can inspire new theories of consciousness (e.g., see Flanagan 1992). Thus the interplay between metacognition and consciousness can be expected to be symbiotic, with each of those affecting the other (e.g., special issue of Consciousness and Cognition, 2000).

Two major subdivisions of metacognition are (a) metacognitive knowledge (i.e., what people know about their own cognitions as based on their life history) and (b) on-line metacognitions comprised of metacognitive monitoring and metacognitive control of one’s own cognitions.

The first of those subdivisions includes autobiographical facts such as ‘I remember things better when I see them than when I hear them’ or ‘I usually remember the gist of the text better than the exact words that were in the text.’ Metacognitive knowledge, especially the development of metacognitive knowledge, has been studied extensively in children (e.g., Kreutzer et al. 1975).

The second of those subdivisions involves questions about the way in which people monitor their on-going cognitions and also the way in which people control their on-going cognitions. The key notion is that the distinction between the meta level versus the object level is relational rather than absolute. Put another way, no particular aspect of cognition is always at the meta-level in any absolute sense. Instead, if one aspect of cognition is monitoring or controlling another aspect of cognition, then we regard the former aspect as metacognitive in relation to the latter aspect. An example may help to clarify this and may also help illustrate the kinds of metacognition that currently are being researched. Imagine that you are asked what the capital of Australia is. You might recall the name of an Australian city, and after you say it, you might be asked how confident you are that the answer you recalled is correct. Thus your confidence judgment (e.g., ‘50 percent’) is at the meta-level, relative to your recall response (e.g., ‘Sydney’). However, if you are then asked to tell how accurate that particular confidence judgment is, you might put an interval around the confidence judgment (e.g., ‘somewhere between 40 and 60 percent’ or ‘somewhere between 30 and 70 percent’); then the original confidence judgment of ‘50 percent’ is at the object level, relative to the confidence interval. Thus a given cognition can be either an object-level cognition (if it is the object of some other cognition that is monitoring or controlling it) or a meta-level cognition (if it is monitoring or controlling some other cognition).

In the heavily researched area of metacognition referred to as ‘metamemory,’ the monitoring and control are of one’s own memory during the acquisition of new information and during the retrieval of previously acquired information. Before about 1950, many researchers conceptualized people as blank slates, and the way that acquisition was believed to occur was that the individual was assumed to be passive, having little or no control over his or her own acquisition. However, since the 1950s, researchers began to conceptualize the individual as having substantial control over acquisition and as being active rather than passive, both during the acquisition of new information and during the retrieval of previously learned information.

Consider this concrete example. Suppose that a student is studying for an examination that will soon occur on French–English vocabulary such as chateau = castle. We suppose that several monitoring and control processes will be activated while the student is learning the new vocabulary and while the student is attempting to retrieve the answers during the subsequent examination. Some of those monitoring and control processes are discussed in the next section (a theoretical framework that integrates these processes into an overall system can be found in an article by Nelson and Narens 1990).

1. Metacogniti e Monitoring

The various metacognitive monitoring processes are differentiated in terms of when they occur during acquisition and retrieval, and also in terms of whether they pertain to the person’s future performance (prospective monitoring) or the person’s past performance (retrospective monitoring). Consider each of those in turn.

1.1 Prospecti e Monitoring

1.1.1 Ease-of-learning judgments.

Even prior to initiating the intentional acquisition of to-be-learned items, some metacognitive monitoring occurs. An ease-of-learning judgment is the person’s judgment of how easy or difficult the items will be to acquire. For instance, the person might believe that cheval = horse will be more difficult to learn than chateau = castle. Underwood (1966) showed that learners are somewhat accurate—not perfectly, but well above chance—at predicting which items will be easiest to learn. The learners’ predictions of how easy it would be to learn each item, made in advance of the presentation of those items for study, were positively correlated with subsequent recall after a constant amount of study time on every item. That is, the items people predicted would be easiest to learn had a greater subsequent likelihood of being recalled than items predicted to be hardest to learn.

1.1.2 Judgments of learning.

The next kind of monitoring occurs during or soon after acquisition. The learner’s judgment of learning is his or her prediction of the likelihood that a given item will be remembered correctly on a future test. Arbuckle and Cuddy (1969) showed that the predictive accuracy of people’s judgments of learning is above chance but far from perfect, similar to the situation for ease-of-learning judgments. Research by Leonesio and Nelson (1990) showed that judgments of learning are more accurate than ease-of-learning judgments for predicting eventual recall, perhaps because people’s judgments of learning can be based on what learners notice about how well they are mastering the items during acquisition. Mazzoni and Nelson (1995) showed that judgments of learning are more accurate when the learning is intentional rather than incidental, even when the amount recalled is the same for intentional versus incidental learning.

In regard to intentional leaning, two examples illustrate how widely the accuracy of judgments of learning can vary. First, in situations such as the acquisition of foreign-language vocabulary as discussed above, Nelson and Dunlosky (1991) found that people’s judgments of learning can be almost perfectly accurate if the judgment of learning is made not immediately after studying a given item but rather after a short delay; this finding has been replicated in many laboratories and is called the ‘delayed-JOL effect’ (where ‘JOL’ stands for judgment of learning). The delayed-JOL effect is exciting because it shows that under the proper conditions, people can monitor their learning extremely accurately. However, there currently is controversy over the theoretical mechanisms that give rise to the high accuracy of delayed JOLs (e.g., Dunlosky and Nelson 1992, Kelemen and Weaver 1997, Spellman and Bjork 1992), and this seems a fruitful topic for future research.

Second, however, in situations such as the acquisition of text, JOLs can be extremely inaccurate. Glenberg and his colleagues had people read passages of text and make a JOL after each passage, followed by a test consisting of one true–false inference derived from each passage. The recurring finding was that the JOLs were not above chance accuracy for predicting that test performance; Glenberg et al. (1982) referred to this as an ‘illusion of knowing.’ However, subsequent researchers discovered that those findings had a highly limited domain and that people’s JOL accuracy for assessing their comprehension of text could be well above chance if small changes were made in the tests that assessed the accuracy of the JOLs. In particular, Weaver (1990) found that the accuracy of JOLs increased as the number of true–false inference questions increased, and Maki et al. (1990) found that the accuracy of JOLs increased when the test question was a multiple-choice item rather than a true–false item. Thus the way in which the accuracy of JOLs is assessed can affect conclusions about the degree of metacognitive accuracy.

1.1.3 Feeling-of-knowing judgments.

Another kind of metacognitive monitoring judgment is people’s prediction of whether they will eventually remember an answer that they currently do not recall. This was the first metamemory judgment examined in the laboratory to assess people’s accuracy at predicting their subsequent memory performance. Hart (1965) found that feeling-of-knowing judgments were somewhat accurate at predicting subsequent memory performance. The likelihood of correctly recognizing a nonrecalled answer was higher for nonrecalled items that people said they knew than for nonrecalled items people said they didn’t know. However, people frequently did not recognize answers that they had claimed that they would recognize, and people sometimes did recognize answers that they had claimed they wouldn’t recognize (although in part, these correct recognitions could sometimes be due to guessing factors in the multiple-choice recognition test). Subsequently, the accuracy of predicting other kinds of memory performance such as relearning was investigated by Nelson et al. (1984), who also offered several theoretical explanations for how people might make their feeling-of-knowing judgments. Currently, the most widely accepted explanation (e.g., Koriat 1997, Metcalfe et al. 1993, Reder and Ritter 1992) is that rather than monitoring directly the nonrecalled information in memory (almost as if by magic—see Nelson and Narens 1990), what people do when they make feeling-of-knowing judgments is to assess both their familiarity with the stimulus cue (aka ‘stimulus recognition’) and the partial components that they can recall from the requested response (e.g., tip-ofthe-tongue components such as the first letter or the number of syllables), and then draw an inference based on that assessment. For instance, when people recognize the stimulus che al as having been studied previously and or recall that ‘h’ is the first letter of the requested response, then they might infer that they will recognize the requested response if they saw it in a multiple-choice test item.

1.2 Retrospective Confidence Judgments

In contrast to the aforementioned monitoring judgments in which people attempt to predict their future memory performance, retrospective confidence judgments occur after someone recalls or recognizes an answer. They are judgments of how confident the person is that his or her answer was correct. For instance, if someone were asked for the English translation equivalent of chateau, the person might recall ‘castle’ (the correct answer) or might recall ‘hat’ (the incorrect answer, probably occurring because the person confused chateau with chapeau) and then, without feedback from the experimenter, would make a confidence judgment about the likelihood that the recalled answer was correct. Fischhoff et al. (1977) demonstrated that these retrospective confidence judgments have substantial accuracy, but there is a strong tendency for people to be overconfident, especially when the test is one of recognition. For instance, for the items that people had given a confidence judgment of ‘90 percent likely to be correct,’ the actual percentage of correct recognition was substantially below that. Subsequent research by Koriat et al. (1980) found that people’s accuracy could be increased—and people’s tendency to be overconfident decreased—if at the time of making each retrospective confidence judgment, the people were asked to give a reason that their response (in either recall or recognition) might have been wrong. However, no change in accuracy occurred when people were asked to give a reason that the response might have been correct, so the conclusion of the researchers was that in the usual situation of making retrospective confidence judgments, people have a ‘confirmation bias’ to think of reasons why they were correct and fail to think of reasons why they might have been wrong.

2. Metacognitive Control

Although it is interesting that people can monitor their progress during acquisition and retrieval, this would be little more than a curiosity if it had no other role in learning and memory. However, people can control aspects of their acquisition and retrieval. First, consider what people can control during self-paced acquisition; second, consider what they can control during retrieval.

2.1 Control During Self-paced Acquisition

2.1.1 Allocation of self-paced study time during acquisition.

A student who is learning foreign-language vocabulary can allocate various amounts of study time to the various to-be-learned items, for example, allocating extra study time to the most difficult items. Bisanz et al. (1978) found that the allocation of study time may be related to people’s JOLs. Learners in the early years of primary school make accurate JOLs but do not utilize the JOLs when allocating study time across the items, whereas slightly older children do utilize their JOLs when allocating study time. The older children allocated extra study time to items that they judged not yet to have been learned and did not allocate extra study time to items that they judged to have been learned.

2.1.2 Strategies during self-paced study.

People can control not only how much study time they allocate to various items, but also which strategy they use during that study time. Often there are strategies that are more effective than rote repetition, but do people know about them? People’s utilization of a mnemonic strategy for the acquisition of foreign-language vocabulary was investigated by Pressley et al. (1984). After people learned some foreign-language vocabulary by rote and learned other foreign-language vocabulary by the mnemonic strategy, they chose whichever strategy they preferred for a final trial of learning new foreign-language vocabulary. Only 12 percent of the adults chose the mnemonic strategy if they had not received any test trials during the earlier phase. However, 87 percent chose the mnemonic strategy if they had received test trials during the earlier acquisition phase. Thus, test trials help people to see the effectiveness of different strategies. When the subjects were children instead of adults, they not only needed test trials but also needed experimenter-provided feedback after those test trials so as to know how well they had performed on the rote-learned items versus the mnemonic-learned items. Without both the test trials and the feedback, the children were unlikely to adopt the advantageous mnemonic strategy.

2.2 Control During Retrieval

2.2.1 Control of initiating attempts at retrieval.

Immediately after someone is asked a question, and before attempting to search memory for the answer, a metacognitive decision occurs about whether the answer is likely to be found in memory. If you are asked what the telephone number is for the President of the United States, you probably would decide immediately that the answer is not in your memory. Notice that you do not need to search through all the telephone numbers that you know, nor do you need to search through all the information you have stored in your memory about the President. Consider how different that situation is from one in which you are asked the telephone number of one of your friends.

This rapid feeling-of-knowing judgment that precedes an attempt to retrieve an answer was investigated by Reder (1987). She found that people are faster at making a feeling-of-knowing decision about whether or not they know the answer to a general information question (e.g., ‘What is the capital of Australia?’) than they are at answering that question (e.g., saying ‘Canberra’). Thus a metacognitive decision can be made prior to (as well as after) retrieving the answer. Only if people feel that they know the answer will they continue their attempts to retrieve the answer. When they feel they do not know the answer, they don’t even attempt to search memory (as in the aforementioned example of your response to a query for the President’s telephone number).

2.2.2 Control of the termination of retrieval.

People may initially believe that they know an answer, but after extended attempts at retrieval without producing the answer, they eventually terminate searching for the answer. The metacognitive decision to terminate such an extended search of memory was investigated by Nelson et al. (1984). They found that the amount of time elapsing before someone gives up searching memory for a nonretrieved answer is greater when the person’s on-going feeling of knowing for the answer is high rather than low. As an example, someone might spend a long time during an examination attempting to retrieve the English equivalent of chateau (which the person studied the night before) but little or no time attempting to retrieve the English equivalent of boıte (which the person did not study previously). The metacognitive decision to continue versus terminate attempts at retrieving an answer from memory can also be affected by other factors, such as the total amount of time available during the examination. A recent theory of the metacognitive components involved in retrieval has been offered by Barnes et al. (1999).

3. Neuropsychological Aspects of Metacognition

Neuropsychological patients have been investigated to determine if any of them have particular deficits of metacognition. For instance, Korsakoff patients, who have frontal-lobe damage as well as other brain damage (Shimamura et al. 1988), have deficits in the accuracy of their JOLs (Bauer et al. 1984). Also, Korsakoff patients have extremely low feeling-ofknowing accuracy (Shimamura and Squire 1986) but normal retrospective-confidence-judgment accuracy (Shimamura and Squire 1988). Patients with primarily frontal-lobe deficits sometimes show normal recall but reduced feeling-of-knowing accuracy (Janowsky et al. 1989).

Many of the experiments cited in this research paper have been reprinted in a book edited by Nelson (1992). Other books containing more recent findings about metacognition have been edited by Metcalfe and Shimamura (1994), Reder (1996), and Mazzoni and Nelson (1998).

Bibliography:

1. Arbuckle T Y, Cuddy L L 1969 Discrimination of item strength at time of presentation. Journal of Experimental Psychology 81: 126–31
2. Barnes A E, Nelson T O, Dunlosky J, Mazzoni G, Narens L 1999 An integrative system of metamemory components involved in retrieval. Attention and Performance 17: 287–313
3. Bauer R H, Kyaw D, Kilbey M M 1984 Metamemory of alcoholic Korsakoff patients. Society for Neurosciences Abstracts 10: 318
4. Bisanz G L, Vesonder G L, Voss J F 1978 Knowledge of one’s own responding and the relation of such knowledge to learning: A developmental study. Journal of Experimental Child Psychology 25: 116–28
5. Dunlosky J, Nelson T O 1992 Importance of the kind of cue for judgments of learning (JOL) and the delayed-JOL effect. Memory & Cognition 20: 374–80
6. Fischhoff B, Slovic P, Lichtenstein S 1977 Knowing with certainty: the appropriateness of extreme confidence. Journal of Experimental Psychology: Human Perception and Performance 3: 552–64
7. Flanagan O 1992 Consciousness Reconsidered. MIT Press, Cambridge, MA
8. Glenberg A, Wilkinson A C, Epstein W 1982 The illusion of knowing: Failure in the self-assessment of comprehension. Memory & Cognition 10: 597–602
9. Hart J T 1965 Memory and the feeling-of-knowing experience. Journal of Educational Psychology 56: 208–16
10. Janowsky J S, Shimamura A P, Squire L R 1989 Memory and metamemory: Comparisons between patients with frontal lobe lesions and amnesic patients. Psychobiology 17: 3–11
11. Johnson M K, Raye C L 1981 Reality monitoring. Psychological Re iew 88: 67–85
12. Kelemen W L, Weaver C A 1997 Enhanced metamemory at delays: Why do judgments of learning improve over time? Journal of Experimental Psychology: Learning, Memory, & Cognition 23: 1394–409
13. Koriat A 1997 Monitoring one’s own knowledge during study: A cue-utilization approach of judgments of learning. Journal of Experimental Psychology: General 126: 349–70
14. Koriat A, Lichtenstein S, Fischhoff B 1980 Reasons for confidence. Journal of Experimental Psychology: Human Learning and Memory 6: 107–18
15. Kreutzer M A, Leonard C, Flavell J H 1975 An interview study of children’s knowledge about memory. Monographs of the Society for Research in Child De elopment 40: 1–60
16. Leonesio R J, Nelson T O 1990 Do different metamemory judgments tap the same underlying aspects of memory? Journal of Experimental Psychology: Learning, Memory, & Cognition 16: 464–70
17. Maki R H, Foley J M, Kajer W K, Thompson R C, Willert M G 1990 Journal of Experimental Psychology: Learning, Memory, & Cognition 16: 609–16
18. Mazzoni G, Nelson T O 1995 Judgments of learning are affected by the kind of encoding in ways that cannot be attributed to the level of recall. Journal of Experimental Psychology: Learning, Memory, and Cognition 21: 1263–74
19. Mazzoni G, Nelson T O 1998 Metacognition and Cognitive Neuropsychology: Monitoring and Control Processes. Erlbaum, Mahwah, NJ
20. Metcalfe J, Schwartz B L, Joaquim S G 1993 The cue familiarity heuristic in metacognition. Journal of Experimental Psychology: Learning, Memory, and Cognition 19: 851–61
21. Metcalfe J, Shimamura A P 1994 Metacognition: Knowing about Knowing. MIT Press, Cambridge, MA
22. Nelson T O 1992 Metacognition: Core Readings. Allyn and Bacon, Boston
23. Nelson T O 1996 Gamma is a measure of the accuracy of predicting performance on one item relative to another item, not of the absolute performance on an individual item comment. Applied Cogniti e Psychology 10: 257–60
24. Nelson T O, Dunlosky J 1991 When people’s judgments of learning (JOLs) are extremely accurate at predicting subsequent recall: The ‘delayed-JOL effect.’ Psychological Science 2: 267–70
25. Nelson T O, Gerler D, Narens L 1984 Accuracy of feeling-ofknowing judgments for predicting perceptual identification and relearning. Journal of Experimental Psychology: General 113: 282–300
26. Nelson T O, Narens L 1990 Metamemory: A theoretical framework and new findings. In: Bower G H (ed.) The Psychology of Learning and Moti ation: Ad ances in Research and Theory. Academic Press, San Diego, CA, Vol. 26, pp. 125–73
27. Pressley M, Levin J R, Ghatala E 1984 Memory strategy monitoring in adults and children. Journal of Verbal Learning and Verbal Beha ior 23: 270–88
28. Reder L M 1987 Strategy selection in question answering. Cogniti e Psychology 19: 90–138
29. Reder L M 1996 Implicit Memory and Metacognition. Erlbaum, Mahwah, NJ
30. Reder L M, Ritter F E 1992 What determines initial feeling of knowing? Familiarity with question terms, not with the answer. Journal of Experimental Psychology: Learning, Memory, and Cognition 18: 435–51
31. Shimamura A P, Jernigan T L, Squire L R 1988 Korsakoff’s syndrome: Radiological (CT) findings and neuropsychological correlates. Journal of Neuroscience 8: 4400–10
32. Shimamura A P, Squire L R 1986 Memory and metamemory: A study of the feeling-of-knowing phenomenon in amnesic patients. Journal of Experimental Psychology: Learning, Memory, and Cognition 12: 452–60
33. Shimamura A P, Squire L R 1988 Long-term memory in amnesia: Cued recall, recognition memory, and confidence ratings. Journal of Experimental Psychology: Learning, Memory, and Cognition 14: 763–70
34. Spellman B A, Bjork R A 1992 When predictions create realityjudgments of learning may alter what they are intended to assess. Psychological Science 3: 315–16
35. Underwood B J 1966 Individual and group predictions of item difficulty for free learning. Journal of Experimental Psychology 71: 673–79
36. Weaver C A 1990 Constraining factors in calibration of comprehension. Journal of Experimental Psychology: Learning, Memory, and Cognition 16: 214–22