The Neuropsychology of Memory Illusions: False Recall and Recognition in Amnesic Patients

JOURNAL OF MEMORY AND LANGUAGE 35, 319±334 (1996) ARTICLE NO. 0018

DANIEL L. SCHACTER Department of Psychology, Harvard University, Cambridge, Massachusetts; and Memory Disorders Research Center, Veterans Affairs Medical Center, Boston

MIEKE VERFAELLIE Memory Disorders Research Center, Veterans Affairs Medical Center, Boston; and Boston University School of Medicine, Boston

AND

DAN PRADERE Memory Disorders Research Center, Veterans Affairs Medical Center, Boston

Memory illusions have been explored extensively in cognitive studies of normal memory, but hardly at all in neuropsychological research with amnesic patients. The present experiment examined false recall and recognition of nonstudied words that are preceded by a list of strong associates. We used the Deese (1959) paradigm, recently revived by Roediger and McDermott (1995), in which people frequently claim that nonstudied words appeared on a presented list. Results showed that amnesic patients were less susceptible to false recognition than were matched controls and showed different patterns of false recall. To account for the observed differences between amnesics and controls, we suggest that false recognition of nonstudied words preceded by numerous associates depends on the same kinds of semantic and associative information about study list words that also supports accurate recognition. Amnesic patients do not retain such information, resulting in poor recollection of study list words and decreased susceptibility to false recognition. ᭧ 1996 Academic Press, Inc. Despite a century's worth of psychological amnesic patients have had a profound impact research concerning memory distortions and on theorizing about normal memory. Findings illusions (Ceci & Bruck, 1993; Schacter, of spared implicit memory in patients with 1995a) and scattered observations of confabu- impaired explicit memory have led to numer- lations and related false memories in brain- ous proposals about dissociable memory pro- damaged patients (Johnson, 1991; Mosco- cesses and systems (cf., Bowers & Schacter, vitch, 1995; Schacter & Curran, 1995), theo- 1993; Cermak & Verfaellie, 1992; Cohen & retical understanding of memory illusions has Eichenbaum, 1993; Johnson & Chalfonte, been almost entirely uninformed by neuro- 1994; Moscovitch, 1994; Schacter, Chiu, & psychological observations. This contrasts Ochsner, 1993; Shimamura, 1986; Squire, sharply with other sectors of human memory 1992; Tulving & Schacter, 1990), and demon- research, where neuropsychological studies of strations of temporal gradients in retrograde amnesia have illuminated the nature of consol- Supported by Grants NS26985 and NS27950 from the idation processes (e.g., Cohen & Eichenbaum, National Institute of Neurological Disorders and Stroke. 1993; McClelland, McNaughton, & O'Reilly, We thank Tim Curran and Marcia Johnson for helpful 1995; Squire, 1992). comments and Kimberly Nelson for help with preparation of the manuscript. Address reprint requests to Daniel L. These and other areas in which studies of Schacter, Department of Psychology, Harvard University, amnesia have in¯uenced the analysis of nor- 33 Kirkland Street, Cambridge, MA 02138. mal memory are all characterized by a com-

319 0749-596X/96 $18.00 Copyright ᭧ 1996 by Academic Press, Inc. All rights of reproduction in any form reserved.

AID JML 2457 / a002$$$$21 04-11-96 00:16:57 jmla AP: JML 320 SCHACTER, VERFAELLIE, AND PRADERE mon feature: empirical and theoretical atten- enced by, studies of amnesic patientsÐin tion focuses on the quantity of information sharp contrast to the extensive interactions that amnesic patients and control subjects do that have characterized quantity-oriented re- or do not remember. On explicit tests of mem- search. ory for recent experiences, amnesic patients One of the few attempts at examining mem- remember less than do control subjects, ory distortion in amnesic patients is found in a whereas on implicit tests they produce similar study of false recognition reported by Cermak, numbers of study-list items. In studies of ret- Butters, and Gerrein (1973). Cermak et al. rograde amnesia, inferences are based on the used the false recognition paradigm developed amount of information that patients and con- by Underwood (1965), in which lure items trols recall from different time periods. on a continuous recognition test are preceded Koriat and Goldsmith (1994, in press) have either by unrelated words or by words that recently characterized this emphasis on the bear an associative, semantic, or physical rela- quantity of retained information as an expres- tion to the lure. Underwood reported a modest sion of a ``storehouse'' metaphor of memory. but signi®cant increase in false alarms to re- Koriat and Goldsmith point to the existence of lated lures in comparison to unrelated lures. an alternative correspondence metaphor that In the Cermak et al. study, Korsakoff amnesics focuses less on the amount retained and more and controls encountered four different types on the qualitative characteristics of what peo- of lures: unrelated, associates (e.g., table± ple rememberÐhow well memory corre- chair), synonyms (e.g., robber±thief), and ho- sponds to reality. This metaphor originated in mophones (e.g., bear±bare). Amnesic patients the work of Bartlett (1932), and continued in produced signi®cantly more false alarms to the research of psychologists who focused on associates and homophones than did control such issues as eyewitness testimony, where subjects, and they also showed a slight, non- the qualitative correspondence between what signi®cant trend for more false alarms to syn- a person remembers and what actually hap- onyms and unrelated words. pened is more important than the sheer amount We will say more about these results later of remembered material (e.g., Loftus, 1979; (see General Discussion), but for now we note Wells & Loftus, 1984). that the Cermak et al. data suggest that amne- Studies of memory illusions and distortions sic patients are sometimes more prone to false are closely tied to a correspondence metaphor alarms than are nonamnesic controlsÐthat is, of memory. During the 1970s, many cognitive not only do amnesic patients remember less psychologists focused intensively on investi- than controls do, but what they claim to re- gating various kinds of memory distortions, member may also be less accurate than what including errors attributable to schema-based nonamnesic individuals claim to remember. inferences (e.g., Sulin & Dooling, 1974), ab- Other studies of recognition memory have straction and generalization processes (e.g., shown that amnesics sometimes exhibit a Bransford & Franks, 1971), and misleading higher false alarm rate than control subjects suggestions (e.g., Loftus, Miller, & Burns, even to nonstudied words that have no particu- 1978). More recently, memory illusions and lar relation to studied words (e.g., Knowlton & distortions have become the subject of re- Squire, 1995; Verfaellie & Treadwell, 1993; newed interest, in part because of real-world for discussion, see Roediger & McDermott, controversies about the accuracy of traumatic 1994). Together with their reduced hit rates, memories recovered in psychotherapy (cf., the elevated false alarm rates of amnesic pa- Herman & Harvey, 1993; Lindsay & Read, tients in the latter studies may re¯ect an inabil- 1994; Loftus, 1993; Ofshe & Watters, 1994; ity to discriminate between studied and non- Schacter, 1995b, 1996) However, this corre- studied items, resulting in haphazard guessing. spondence-oriented research has had little ef- This pattern is similar to the mirror effect in fect on, and has been almost entirely unin¯u- recognition exhibited by normal subjects,

AID JML 2457 / a002$$$$22 04-11-96 00:16:57 jmla AP: JML AMNESIA AND MEMORY ILLUSIONS 321 where manipulations that lower hit rates tend hits to studied words. In addition, subjects also to increase false alarm rates (e.g., were asked to make remember/know judg- Glanzer & Adams, 1990). ments about studied and nonstudied words, Evidence that amnesic patients sometimes where a ``remember'' response indicates that make more false positive responses than con- subjects possess a speci®c, vivid recollection trol subjects raises the possibility that they of having encountered a word during the study might be unusually susceptible to memory il- list, and a ``know'' response indicates that lusions that are expressed by false alarms to word just seems familiar (cf., Gardiner & nonstudied items. Consistent with this idea, Java, 1993; Tulving, 1985). Roediger and Mc- Reinitz, Verfaellie, and Milberg (1996) report Dermott found that subjects often claimed to that amnesic patients are more prone than con- remember having encountered the critical trols to false alarms based on illusory memory lures on the study list. In fact, subjects pro- conjunctions, where subjects claim to have vided remember responses to critical lure seen a new stimulus when in fact they have words just as often as they provided remember seen only its component features (Reinitz, responses to studied words. Lammers, & Cochran, 1992), and Kroll, In the present experiment, we examined Knight, Metcalfe, Wolf, and Tulving (1996) whether amnesic patients are also subject to report similar ®ndings in patients with left, the memory illusions embodied in false recall right, or bilateral hippocampal damage. and recognition of nonpresented associates Roediger and McDermott (1995; see also such as sleep. On the one hand, the observed Read, 1996) have recently found another strik- tendency of amnesic patients to false alarm ing memory illusion. Using a paradigm devel- more often than control subjects in some con- oped initially by Deese (1959), Roediger and ditions leads to the prediction that they would McDermott presented subjects with lists of be especially susceptible to this memory illu- words that are all strong associates of a criti- sion. On the other hand, however, false recall cal, nonpresented word. For example, when and recognition of critical lures in the Roe- the nonpresented word was sleep, the pre- diger and McDermott paradigm might depend sented list of associates consisted of bed, rest, on very different processes than those that un- awake, tired, dream, wake, night, blanket, derlie false alarms in other situations. Spe- doze, slumber, snore, pillow, peace, yawn, and ci®cally, false recall and recognition of words drowsy. Following presentation of a series of such as sleep may depend on retaining and such lists, subjects were given a free recall remembering associative or semantic informa- test and a yes/no recognition test. Roediger tion about the list of presented words. If it and McDermott, like Deese, found that sub- does, then amnesic patients may be less prone jects often intruded the nonpresented word on to false memories of sleep because they may a free recall test. In addition, they found that fail to remember the semantic or associative subjects made an extraordinarily high number information that ordinarily misleads non- of false alarms to critical lures such as sleep memory-impaired subjects into claiming that on the recognition test. In one experiment, for they remember something that never hap- example, the hit rate for studied items was .79 pened. when the recognition test was preceded by a To examine these possibilities, we exposed recall test and .65 when the recognition test amnesic patients and control subjects to a se- was preceded by unrelated distractor activity. ries of lists containing strong associates of a The corresponding false alarm rates to critical nonpresented critical word. Immediately after lures were actually higher than the hit ratesÐ presentation of half the lists, patients and con- .81 and .72, respectively. Subjects expressed trols attempted free recall; after the other half high con®dence in their false alarms; indeed, of the lists, they carried out unrelated arithme- they were just as con®dent about their false tic problems. Finally, all subjects were given alarms to critical lures as they were about their yes/no recognition tests for studied words,

AID JML 2457 / a002$$$$22 04-11-96 00:16:57 jmla AP: JML 322 SCHACTER, VERFAELLIE, AND PRADERE critical lures, and unrelated lures. In addition and a target word (critical lure) that was not to indicating whether a word had appeared presented for study. The study words were all previously on a study list, subjects also indi- highly associated to the critical lure and were cated whether they actually remembered the ordered such that the strongest associates oc- prior presentation of the words or whether curred ®rst in the list. The 24 lists were subdi- they just knew that the word had been pre- vided into three sets for counterbalancing pur- sented previously. poses. Design and procedure. All participants METHOD were tested in two conditions, a Study / Re- Participants. Twelve amnesic patients (10 call condition and a Study / Arithmetic con- male, 2 female) and 12 individuals with intact dition, which were administered in two ses- memory functioning (10 male, 2 female) par- sions separated by at least 1 week. Half of the ticipated in the experiment. The amnesic pa- subjects received the Study / Recall condi- tients had all been screened at the Memory tion during the ®rst session and the Study / Disorders Research Center of the Boston Arithmetic condition during the second ses- VAMC. Six patients had a diagnosis of alco- sion. For the other half of the participants, this holic Korsakoff syndrome and 6 patients had order was reversed. a variety of nonalcoholic etiologies (anoxia, A set of eight lists was used in each condi- encephalitis, thalamic infarct). Because the al- tion. The remaining eight lists were not stud- coholic and nonalcoholic amnesics performed ied. Four of these appeared on the recognition similarly on the experimental task, they are test that accompanied the Study / Recall con- further treated as a single amnesic group. They dition, whereas the other four appeared on the had a mean age of 57.2 years and an average recognition test that accompanied the Study of 13.5 years of education. Their overall level / Arithmetic condition. Lists were counter- of intellectual functioning was in the average balanced so that they were used equally often range, as indicated by a mean Verbal IQ of in the Study / Recall condition, in the Study 98.9 on the Wechsler Adult Intelligence / Arithmetic condition, and as nonstudied ScaleÐRevised. Likewise, their attentional lists. Nonstudied lists were counterbalanced capabilities were intact, as indicated by a across the two recognition tests. All partici- mean score of 103.8 on the Attentional Index pants were tested individually. Before presen- of the Wechsler Memory ScaleÐRevised tation of each study list, they were told that a (WMS-R). In contrast, they consistently ex- series of words would be presented via a tape hibited severe de®cits on a variety of explicit player and that they should try and remember memory tasks. On the WMS-R, they obtained the words. The words were recorded in a fe- a mean General Memory Index of 82.1 and a male voice and were presented at a rate of mean Delayed Memory Index of 58.5. Details approximately 1.5 s per word. Immediately on individual patients are presented in Table following presentation of a study list, partici- 1. The control group consisted of 12 individu- pants were asked to say out loud as many of als who were matched to the amnesics in terms the words as they could remember (Study / of age, education and overall level of intelli- Recall condition) or to perform simple addi- gence. Six of these individuals had a history tion and multiplication problems (Study / of alcoholism and the other 6 had no history Arithmetic condition). Approximately 1 min of alcoholism. Their mean age was 52 years; was given for either of these tasks, after which they had an average of 13.4 years of education the next study list was presented. Eight study and a Verbal IQ of 107.7. lists in the same condition (Study / Recall or Materials. The target materials consisted of Study / Arithmetic) were presented during a 24 lists of 16 words, identical to those used single session. Presentation of all eight study by Roediger and McDermott (1995). Each list lists took approximately 20 min. contained 15 words to be presented for study The recognition test was administered ap-

AID JML 2457 / a002$$$$22 04-11-96 00:16:57 jmla AP: JML AMNESIA AND MEMORY ILLUSIONS 323

TABLE 1

CHARACTERISTICS OF AMNESIC PATIENTS

WMS-R

Etiology Age VIQ GM ATT DLY

Korsakoff 67 93 76 109 62 Korsakoff 60 90 99 99 61 Korsakoff 73 123 104 116 56 Korsakoff 66 88 76 96 53 Korsakoff 68 87 82 93 60 Korsakoff 64 83 66 99 õ50 Anoxia 57 109 65 89 61 Encephalitis 44 111 81 107 69 Anoxia 31 95 65 120 õ50 Anoxia 36 95 90 115 õ50 Encephalitis 67 126 102 114 õ50 Thalamic 53 87 70 89 80 Mean 57.2 98.9 82.1 103.8 58.5

Note. VIQ, Verbal IQ from the Wechsler Adult Intelligence Scale (Revised); WMS-R, Wechsler Memory Scale- Revised; scores are presented separately for the indices of General Memory (GM), Attention (ATT), and Delay (DLY). proximately 2 min after completion of the re- positions 1, 8, and 10. The nonstudied words call or arithmetic task that followed the ®nal consisted of the critical lures corresponding study list. Participants were presented with a to each of the eight studied lists, the critical list of words and were asked to indicate for lures corresponding to four nonstudied lists, each word whether or not it had been pre- and the items in serial positions 1, 8, and 10 sented on one of the audiotapes. In case they of these nonstudied lists. thought a word had previously been presented, they were then asked to indicate whether they RESULTS remembered or knew the word. Instructions Free recall. We ®rst analyzed the mean pro- for the remember/know judgment were similar portion of study list words and critical lures to those used by Roediger and McDermott produced by amnesic patients and control sub- (1995). Participants were told that a remember jects on the free recall test, averaged across judgement should be made if they could spe- the eight target lists. As expected, amnesic ci®cally recollect hearing a word on the tape patients recalled on average a much smaller recorder. It was explained that such recollec- proportion of study list words (.27) than did tion might include remembering something controls (.52), t(22) Å 5.10, p õ .0001. In about the speaker's voice or about the addition, amnesics intruded a nonsigni®cantly thoughts they had when they heard the word. smaller proportion of critical lures (.29) than They were told that a know judgement should did controls (.33), t(22) õ 1. Thus, amnesics be made if they felt or knew that a word was produced about the same proportion of study presented earlier on the tape recorder, but if list targets and critical lures (.27 vs .29), they could not recollect anything speci®c whereas control subjects recalled signi®cantly about the word or its occurrence. more study list targets than critical lures (.52 The recognition test contained 48 words, 24 vs .33), t(1) Å 3.26, p õ .01). A combined studied words and 24 nonstudied words. The ANOVA that included Type of Item (studied studied words were obtained by selecting for vs critical lure) as an independent variable re- each of the eight study lists the items in serial vealed a signi®cant Subject Group 1 Type of

AID JML 2457 / a002$$$$22 04-11-96 00:16:57 jmla AP: JML 324 SCHACTER, VERFAELLIE, AND PRADERE

One issue that complicates the interpretation of the above analyses is the fact that amnesics made on the average many more noncritical lure intrusions (1.32/list) than did control subjects (.55/list), t(22) Å 2.40, p õ .05. Furthermore, the nature of these intrusions was somewhat different for the two groups. For amnesics, 22% of their intrusions were unrelated to the studied lists, 48% were related to the just studied list and 30% were related to previously studied lists. For the controls, the corresponding percentages were 19, 79, and 2%. Thus, while the two groups were equally likely to produce unrelated in- FIG. 1. Probability of correct recall in amnesics and trusions, t(22) Åõ1, amnesic patients were control subjects as a function of serial position. less likely than controls to produce intrusions related to the just studied list, t(22) Å 2.72, Item interaction, F(1,22) Å 7.39, MSE Å .017, p õ .05, and, correspondingly, more likely to p õ .05, con®rming the above ®ndings. They produce perseverations to previously studied provide evidence for a difference in the rela- lists, t(22) Å 2.65, p õ .05. Because of differ- tion between critical lure intrusions and recall ences between the two groups in the number of target items in amnesic patients and con- as well as the pattern of intrusions, we com- trols, respectively. puted for each subject the proportion of criti- Figure 1 presents the proportion of recalled cal lures intruded as a function of the total target items as a function of serial position in number of intrusions (critical lures / unre- the study list. Following Roediger and McDer- lated intrusions). According to this adjusted mott (1995), we compared intrusions of criti- measure, amnesic patients intruded a signi®- cal lures to the mean proportion of words re- cantly smaller proportion of critical lures called from the nonrecency and nonprimary (.14) than did control subjects (.27), t(22) Å portions of the serial position curve (positions 2.15, p õ .05. 4±11). For control subjects, the proportion of One question addressed by Roediger and words recalled from these positions (.44), like McDermott concerns whether critical lure in- the overall recall rate, was higher than the trusions arise from associative processes op- proportion of lures intruded (.33), although erating during the recall test itself. If so, then this difference did not reach signi®cance, t(11) one might expect more critical lure intrusions Å 1.64, p Å .13. By contrast, for amnesic when subjects recall many targets from a list patients the proportion of words recalled from than when they recall few targets. To address the middle positions of the serial position the issue, we examined recall of target items curve (.17) was smaller than the proportion of as a function of whether or not subjects pro- critical lures intruded (.29); t(11) Å 2.27, p õ duced the critical lure (only 10 amnesic pa- .05. Consistent with these ®ndings, an AN- tients were included in this analysis because OVA including Type of Item (Studied vs Crit- 2 never produced a critical lure). Amnesic pa- ical Lure) as a factor revealed a signi®cant tients recalled more target items from lists for Group 1 Type of Item interaction, (F(1,22) which they produced the critical lure (mean Å 7.40, MSE Å .021, p õ .05). Once again, Å 4.8) than from lists for which they did not it looks as though the relation between recall produce the critical lure (mean Å 3.9; t(9) Å of studied words and intrusion of critical lures 2.44, p õ .01). Control subjects, in contrast, differs in amnesic patients and control sub- showed comparable levels of target recall jects. whether they produced the critical lure (mean

AID JML 2457 / a002$$$$22 04-11-96 00:16:57 jmla AP: JML AMNESIA AND MEMORY ILLUSIONS 325

TABLE 2

RECOGNITION DATA FOR STUDIED AND NONSTUDIED TARGET WORDS AND STUDIED AND NONSTUDIED CRITICAL LURES IN AMNESIC PATIENTS AND CONTROL SUBJECTS

Proportion of old responses

Overall R K

Item type/Condition A C A C A C

Studied Studied / recall .54 .85 .30 .71 .24 .14 Studied / arithmetic .46 .83 .25 .71 .21 .12 Nonstudied .34 .18 .14 .06 .20 .12 Critical lure Studied / recall .60 .83 .38 .70 .22 .13 Studied / arithmetic .57 .89 .30 .83 .27 .06 Nonstudied .43 .29 .17 .11 .26 .18

Note. R and K refer to remember and know responses, respectively. A and C refer to amnesic patients and control subjects, respectively. In the text, values reported are collapsed across the recall and arithmetic conditions.

Å 7.6) or not (mean Å 8.0; t(11) Å 1.04; ns). arithmetic. However, ANOVAs revealed that We also examined the output position of the type of task (free recall vs arithmetic) did not critical lure, reasoning that a relatively late yield any signi®cant main effects or interac- output position for the critical lure would tend tions, so all subsequent analyses are collapsed to indicate a role for associatively related across the free recall and arithmetic condi- items produced previously during the recall tions. Although the table entries provide sepa- test (cf., Roediger & McDermott, 1995). For rate proportions for the recall and arithmetic amnesic patients, the average output position conditions, in the text we refer to mean pro- for critical lures was 4.8 (of 5.8 words proportions averaged across these two conditions. duced for lists in which there was a critical Analyses of studied words and correspond- lure intrusion). For control subjects, the aver- ing distractors revealed a signi®cantly higher age output position for critical lures was also hit rate in control subjects (.84) than in amne- 4.8, even though the total number of items sic patients (.50), t(22) Å 5.94, p õ .0001, produced for these lists was considerably together with a signi®cantly higher false alarm higher (8.6). Thus, in relative terms the amne- rate in the amnesics (.34) than in the controls sic patients produced the critical lure later than (.18), t(22) Å 2.26, p õ .05. Adopting the did the control subjects. Although not conclu- standard high-threshold correction procedure, sive, these last two analyses are consistent we subtracted the false alarm rate from the hit with the possibility that associative processes rate. This analysis revealed, as expected, that during the recall test play a more prominent recognition accuracy in amnesic patients (.16) role in the critical lure intrusions of amnesic was signi®cantly lower than in control sub- patients than control subjects. jects (.66), t(22) Å 9.60, p õ .0001. We ob- Recognition. The ®rst column of Table 2 tained a nearly identical pattern of results presents the proportion of old responses to when we analyzed critical lures (nonpresented studied words, critical lures, and their corre- words preceded by a list of high associates) sponding distractors for amnesic patients and and their corresponding distractors (words control subjects. The results are presented sep- drawn from the same pool of critical lures that arately as a function of subjects' activities were not preceded by a list of high associates). prior to the recognition test: free recall or Overall, amnesic patients made many fewer

AID JML 2457 / a002$$$$23 04-11-96 00:16:57 jmla AP: JML 326 SCHACTER, VERFAELLIE, AND PRADERE false alarms to critical lures (.59) than did nonsigni®cant F(1,22) Å 1.43, MSE Å .021, control subjects (.86), t(22) Å 3.85, p õ .001. p Å .25. By contrast, the amnesic patients made more However, this analysis is complicated by false alarms to the distractors for critical lures the fact that false alarm rates of both amnesics (.43) than did the control subjects (.29), just and controls were higher to distractors that as they did in the preceding analysis of dis- were drawn from the pool of critical lures than tractors for studied words. However, this dif- to ``ordinary'' distractors for studied items. ference was not statistically signi®cant, t(22) An ANOVA that included Type of Distractor Å 1.35, p Å .19, largely because the present as a factor revealed a signi®cant main effect analysis was based on fewer items than the of this variable, F(1,22) Å 8.91, MSE Å .027, preceding analysis, which led to increases in p õ .01, along with a nonsigni®cant interac- between-group variance. Most importantly, tion with Subject Group (F õ 1). Roediger however, a combined ANOVA that included and McDermott (1995) reported a similar ef- Type of Item (Critical Lure vs Distractor) as fect. To take into account these differing false an independent variable revealed a highly sig- alarm rates, corrected recognition scores for ni®cant Subject Group 1 Type of Item inter- studied items and critical lures were com- action, F(1,22) Å 28.79, MSE Å .017, p õ pared. Whereas amnesic patients' corrected .0001. This crossover interaction con®rms that recognition scores were the same for studied amnesic patients and control subjects re- items and critical lures (.16), controls' cor- sponded in a qualitatively different manner rected recognition scores were higher for stud- to the critical lures and their corresponding ied items (.66) than for critical lures (.57), distractors. although the effect fell just short of the con- Despite their tendencies to make more false ventional signi®cance level t(11) Å 2.17, p Å alarms to distractors, amnesic patients made .052. These results imply that control subjects fewer false alarms to critical lures than did remembered speci®c information about stud- control subjects. Indeed, when we subtracted ied items above and beyond the information the proportion of old responses to distractors that drove their critical lure responses, from the proportion of old responses to critical whereas amnesic patients did not. However, lures, the corrected proportion was much when an ANOVA was performed on these smaller in amnesic patients (.16) than in con- corrected scores, it failed to reveal a signi®- trol subjects (.57), t(22) Å 5.36, p õ .0001. cant Subject Group 1 Item Type interaction, Amnesic patients did, however, make signi®- F(1,22) Å 1.99, MSE Å .013, p Å .17. cantly more old responses to critical lures than For the recognition tests that were preceded to distractors, t(11) Å 3.05, p õ .05, indicating by free recall, it is possible to examine re- that the presentation of associatively related sponses to studied words and critical lures as items reliably in¯uenced their recognition per- a function of whether or not they were pro- formance. duced on the recall test. Table 3 presents the A comparison of studied items and critical relevant data. There was a main effect of Re- lures revealed that both amnesic patients and call F(1,22) Å 16.66, MSE Å .047, p õ .001, control subjects provided more old responses indicating that subjects were more likely to to critical lures than to studied items (.59 vs call old those studied words and critical lures .50 for amnesics and .86 vs .84 for controls), that were produced in the recall test than those replicating a ®nding by Roediger and McDer- studied words and critical lures that were not mott (1995). An ANOVA that included Item produced. Consistent with previous analyses, Type (Studied vs Critical Lure) as a factor there was also a main effect of Subject Group, revealed that the main effect of this variable F(1,22), 18.21, MSE Å .097, p õ .001. In approached but did not attain signi®cance, addition, however, there was a signi®cant Re- F(1,22) Å 3.59, MSE Å .021, P Å .07. The call 1 Type of Item interaction, F(1,22) Å Subject Group 1 Item Type interaction was 4.50, MSE Å .041, p õ .05, indicating that

AID JML 2457 / a002$$$$23 04-11-96 00:16:57 jmla AP: JML AMNESIA AND MEMORY ILLUSIONS 327

TABLE 3 Remember vs know responses. We subdi-

PROPORTION OF ITEMS CALLED OLD ON THE RECOGNI- vided the proportion of old responses made TION TEST BY AMNESIC PATIENTS AND CONTROL SUBJECTS by amnesic patients and controls into remem- IN THE STUDY / RECALL CONDITIONS AS A FUNCTION OF ber and know responses, respectively (Table WHETHER THE WORDS WERE PRODUCED ON THE FREE 2). Roediger and McDermott found that sub- RECALL TEST jects claimed to remember critical lures as of- Production ten as they claimed to remember studied rate in free words, and our data show a similar pattern recall Recognition for both control subjects and amnesic patients. Overall, control subjects provided remember Condition A C A C responses to .71 of the studied words and .77 Studied of the critical lures; the corresponding propor- Produced .27 .52 .74 .98 tions for amnesic patients are .28 and .34. Our Not produced .73 .48 .44 .74 data also reveal that differences in recognition Critical lure performance between amnesics and controls Produced .29 .34 .58 .96 Not produced .71 .66 .59 .76 are attributable to these marked between group differences, which were highly signi®- Note. A refers to amnesic patients and C refers to con- cant for both studied words,t(22) Å 7.76, p õ trol subjects. .0001, and critical lures, t(22) Å 5.71, p õ .0001. However, these effects are mitigated by the fact that neither the control subjects the tendency to call critical lures old was in- nor the amnesic patients used know responses ¯uenced less by recall/nonrecall than was the more often for studied words than for their tendency to call studied items old. corresponding distractors (t õ 1 for both am- Although the three-way interaction of Re- nesic patients and controls). Similarly, neither call 1 Type of Item 1 Subject Group did group used know responses more often for not achieve statistical signi®cance, F(1,22) Å critical lures than for their corresponding dis- 2.34, MSE Å .041, p Å .14, inspection of Ta- tractors (t õ 1 for both amnesics and controls). ble 2 reveals that the Recall 1 Type of Item These analyses imply that both subjects interaction is entirely attributable to the fact groups tended to use the know response when that for amnesic patients, critical lures that they were just guessing that a word had ap- were produced on the free recall test and those peared on the list, and used the remember re- that were not produced received a nearly iden- sponse whenever they felt a degree of cer- tical proportion of old responses on the recog- tainty that a word had been studied previously. nition test (.58 vs 59, respectively). In con- In the Roediger and McDermott (1995) exper- trast, critical lures that had been produced on iment, subjects did provide more know re- the recall test by control subjects received sponses to studied than nonstudied words. more old recognition responses than those that Since our instructions were nearly identical to had not been produced (.96 vs .76), t(11) Å theirs, it is unclear why our subjects showed 2.87, p õ .05. Likewise, studied words that a general bias to use the remember response had been recalled by amnesic patients re- for previously studied items. Whatever the ceived more old recognition responses than reasons, the analysis of remember responses did studied words that had not been recalled tells us little more than the analysis of old (.74 vs .44), t(11) Å 6.70, p õ .0001, and the responses, so we cannot make much of our same was true for studied words in control remember/know data. subjects (.98 vs .74), t(11) Å 6.68, p õ .0001. These analyses imply that production of criti- GENERAL DISCUSSION cal lures did not in¯uence recognition in the We found that non-memory-impaired sub- same way for amnesics and control subjects. jects, like college students in previous studies,

AID JML 2457 / a002$$$$23 04-11-96 00:16:57 jmla AP: JML 328 SCHACTER, VERFAELLIE, AND PRADERE often falsely claimed to remember encounter- control subjects may generate associates to ing critical lures such as sleep on a study list, each target item and link or bind target items making many more false alarms to critical to each other (Johnson & Chalfonte, 1994), lures than to associatively unrelated lures. thereby generating a well-organized or fo- Amnesic patients also made more false alarms cused representation of the theme of the list to critical lures than to unrelated lures. How- (Norman & Schacter, 1996). Studied words ever, they were far less susceptible to false that ®t this thematic representation are likely recognition of critical lures than were controls, to be judged as old, but so are critical lures. whereas they made more false alarms to unre- Just as old words, critical lures may enable lated lures than control subjects did. On the control subjects to recollect other studied free recall test, controls and amnesics intruded words that ®t the general theme of the list, similar numbers of critical lures, but when we thus enhancing their experience that they are took into account amnesic patients' tendencies actually remembering a previously studied to intrude more noncritical lures than control item. subjects, we found that the amnesics intruded By contrast, amnesic patients retain only a a smaller proportion of critical lures than did degraded and poorly organized semantic rep- controls. resentation of the theme of the study list and Despite the fact that amnesic patients made the individual words in it. Their impaired rec- fewer false alarms to critical lures and a ollection of such information produces both smaller proportion of critical lure intrusions fewer hits to old items and fewer false alarms than did controls, our data also suggest that to critical lures than is observed with control amnesics' recognition and recall were both in- subjects. Because our amnesic patients all had ¯uenced less by memory for particular study damage to medial temporal and/or diencepha- list items than was controls' performance. lic brain structures, and because there are both Corrected recognition scores were higher for theoretical and empirical reasons to believe studied words than for critical lures in control that the hippocampus and related structures subjects, but were identical in amnesic pa- contribute to remembering a recently studied tients. Similarly, control subjects produced word (cf., Johnson & Chalfonte, 1994; Mos- more studied words than critical lures on the covitch, 1994; Schacter, Alpert, Savage, free recall test, whereas amnesics produced Rauch, & Albert, 1996; Schacter, Reiman, identical numbers of studied words and criti- Uecker, Polster, Yun, & Cooper, 1995; Squire, cal lures; when we restricted analysis to the 1992), it seems clear that memory processes middle serial positions, amnesics actually pro- that are impaired in amnesia play an important duced more critical lures than target words. role in false alarms to critical lures. We ®rst discuss the implications of our recog- On the other hand, amnesics are more likely nition data and then turn to the free recall than control subjects to make false alarms to results. unrelated lures. This may occur because con- Although amnesic patients made fewer trols can use their well-organized thematic false alarms to critical lures than did control representation to reject unrelated distractors. subjects, we observed the opposite pattern for Amnesic patients are less able to do so, and false alarms to unrelated lures. False alarms hence make false alarms to unrelated dis- to critical lures appear to be based on memory tractors based on haphazard guessing and per- processes that also support accurate recollec- haps some misattributions based on ¯uent pro- tion of words that were actually presented. cessing (Jacoby, Kelley, & Dywan, 1989) of Just as amnesic patients retain less informa- nonstudied words that overlap in some way tion than controls about words that appeared with studied words (although our unrelated on the list, they also fail to retain the semantic distractors bear no systematic relation to stud- or associative information that supports false ied items, some orthographic or phonological, alarms to critical lures. At the time of study, and semantic features of these words occa-

AID JML 2457 / a002$$$$23 04-11-96 00:16:57 jmla AP: JML AMNESIA AND MEMORY ILLUSIONS 329 sionally overlap with studied words). As Ja- should be possible to produce within a single coby and colleagues have pointed out, if a experimental situation either reduced or en- word is processed ¯uently or easily, then a hanced false recognition in amnesic patients, person may mistakenly attribute this ¯uent depending on the number of prior associates processing to a prior encounter with that word to a lure. When large numbers of associates during the study episode, even though the mislead control subjects into producing illu- word was never presented (Jacoby et al., 1989; sory recollections, amnesic patients should Jacoby, 1991; Kelley & Jacoby, 1990). show reduced false recognition; when a single This latter point can help to illuminate the or small number of related items yield ¯uent pattern of results observed in the earlier study processing that can be opposed by conscious of false recognition in amnesic patients by recollection in control subjects, amnesics Cermak et al. (1973). Using a continuous rec- should show enhanced false recognition. Ex- ognition procedure, they found that amnesics tending this logic a step further, it is known made more false recognition responses to as- that increasing the numbers of study list words sociates and homophones of studied words related to a lure word results in a systematic than did controls. The rates of false recogni- increase in false alarms (Hintzman, 1988; tion responses to these items were not nearly Shiffrin, Huber, & Marinelli, 1995). By sys- as high as those observed in the present study, tematically varying the number of associates probably because critical lures were preceded to critical lure items, it should be possible to by only a single related word. Thus, in contrast specify a crossover point where amnesics shift to the Roediger and McDermott paradigm, from enhanced to reduced false recognitions. subjects did not develop the kind of well-orga- From our perspective, this crossover point nized semantic representations of the list that would indicate when false recognitions in con- drive false alarms to critical lures. We suggest trol subjects are enhanced rather than inhibited that in the Cermak et al. experiment, false by memory for the general theme and individ- recognition to associates and homophones was ual items in a list. instead driven by ¯uency-based processes. Having suggested that false recognition of Normal subjects, we suggest, were able to critical lures depends on memory for the gen- counteract these effects because they could eral theme of the study list and the individual recollect the identity of the presented associ- words in it, we can ask at a more ®ne-grained ates and distinguish them from the lure level of analysis how such illusory recollec- itemsÐthat is, they could use recollection to tions come about. Roediger and McDermott oppose ¯uency (Jacoby, 1991). Amnesic pa- (1995) consider several possible sources of tients, however, could not remember the iden- false alarms to critical lures that are preceded tity of the studied items and thus could not by numerous associates. One draws from Un- oppose the effects of ¯uency (Cermak, Ver- derwood's (1965) classic ``implicit associa- faellie, Sweeney, & Jacoby, 1992). Hence, tive response'' account, which holds that at amnesic patients made more false recognition the time of study, target words activate associ- responses than did controls. Interestingly, am- ates that are later confused with the target it- nesic patients did not make excessive numbers self. One version of this idea holds that associ- of false recognition responses to synonyms in ates are truly activated implicitly, in the sense the Cermak et al. (1973) experiment. Cermak that the subject does not become consciously et al. explained this ®nding with reference to aware of the associate during study; it is gen- impaired encoding processes in amnesia, but erated by spreading activation through an as- since neither amnesics nor controls made sociative network. One problem with this idea, more false recognitions to synonyms than to as Roediger and McDermott (1995) note, is unrelated lures it is unclear whether de®cient that people claim to remember the critical encoding is relevant. lures, whereas responses based on implicit In view of the foregoing considerations, it spreading activation would not be expected

AID JML 2457 / a002$$$$24 04-11-96 00:16:57 jmla AP: JML 330 SCHACTER, VERFAELLIE, AND PRADERE to generate such recollective experiences. Our source monitoring processes that are most rel- data lead to a similar concern about this view, evant to critical lure false alarms. because if responses to critical lures were The ideas that we considered earlier con- based solely on nonconscious activation pro- cerning memory for themes and individual cesses, we might expect amnesic patients to items are closely related to Brainerd and be in¯uenced to the same degree as control Reyna's ``fuzzy trace'' account of false rec- subjects. ognition (e.g., Brainerd, Reyna, & Kneer, By another version of Underwood's ac- 1995; Brainerd, Reyna, & Brandes, 1995; count, subjects consciously think of the word Reyna & Brainerd, 1995). By their view, recog- at the time of study and are later subject to nition memory can be based either on a ``gist source memory confusion (Johnson, Hash- trace'' that preserves the general meanings troudi, & Lindsay, 1993), such that they can and interpretations engendered by studied no longer remember whether they actually items or on a ``verbatim trace'' that preserves heard it or only thought of it themselves. In- speci®c information about the exact identity deed, this sort of source misattribution is cen- of each item (we use the term ``speci®c tral to understanding false recognition of criti- trace'' rather than ``verbatim trace'' to avoid cal lures, because when they commit such the implication of a literal recording of an false recognitions subjects are mistaking prior item or event). In general, hits on a recogni- thoughts for prior perceptions. Because amne- tion test tend to be based on recollection sup- sic patients tend to have more dif®culty reported by the speci®c trace, whereas false membering source information than do con- alarms are based on ``feelings of similarity'' trols (Schacter, Harbluk, & McLachlan, 1984; supported by the gist trace. By this view, Shimamura & Squire, 1987), problems in false alarms to critical lures would be attrib- source monitoring alone cannot be responsible utable to strong feelings of similarity that are for false recognitions of critical lures in our supported by a gist representation. If the gist experiment. That is, if amnesic patients were representation is degraded in amnesic pa- as likely as control subjects to generate the tients, they would be less likely to false alarm critical lure at the time of study, we would to critical lures, as we observed. In addition, expect them to make moreÐnot fewerÐfalse a de®cient speci®c representation would pro- alarms to critical lures. From the perspective duce impaired hit rates. However, fuzzy-trace of the source monitoring framework (Johnson et al., 1993), amnesic patients would have to theory tends to characterize false recognition be either (a) less likely than controls to think responses based on the gist representation as of the critical lure at all during the study phase involving ``vague'' feelings of familiarity or of the experiment, (b) less likely to bind to- similarity (e.g., Brainerd et al., 1995). Fuzzy gether the generated critical lure with other list trace theory may not be entirely applicable items, perhaps because they do not reactivate to the kinds of false alarms observed in the previously studied words and generated criti- Roediger and McDermott paradigm, where cal lures in an attempt to actively rehearse the people claim to vividly remember the critical list, or (c) more likely to forget a critical lure lures, perhaps because this view says little word that they, like controls, generated in re- about the kinds of source misattributions that sponse to associates at the time of study. We likely play a role in critical lure false alarms. know of no empirical basis for invoking the If the gist representation is the basis of false ®rst possibility, whereas there are many theo- alarms to critical lures, then either it is capa- retical and empirical reasons for invoking the ble of giving rise to strong feelings of recol- latter two ideas (cf., Johnson & Chalfonte, lection, or it supports a type of ¯uent pro- 1994; Kroll et al., 1996; Reinitz et al., 1996; cessing that is enhanced in control subjects Schacter, 1994; Squire, 1992). Further re- by recollection of speci®c information about search will be needed to delineate the exact other study list items. Fuzzy trace theory as it

AID JML 2457 / a002$$$$24 04-11-96 00:16:57 jmla AP: JML AMNESIA AND MEMORY ILLUSIONS 331 currently stands does not distinguish between in some way from the critical lures they pro- these possibilities. duced. This could come about if amnesics' The fuzzy trace account does lead to an production of critical lures depends in part on interesting perspective on another feature of associative processes operating during recall. our data: for control subjects, corrected recog- A further question that arises from this anal- nition scores were higher to old items than to ysis concerns the relation between the gist rep- critical lures, whereas for amnesic patients, resentation that Brainerd and Reyna invoke, corrected recognition scores were identical to and which we suggest is degraded in amnesic old words and critical lures. This implies that patients, and the processes that drive false rec- amnesic patients retained no speci®c represen- ognition when only a single associate precedes tation of the studied items and based their rec- a lure item. Does the same gist representation ognition responses entirely on degraded gist drive false recognition in both cases? If so, it representations. This account can simultane- might be surprising that amnesic patients ously accommodate two key features of our show greater false recognition in the experi- recognition data: (1) in absolute terms, amne- mental situation that Cermak et al. investi- sics were less likely to false alarm to critical gated, and impaired false recognition in the lures (because of a degraded gist representa- present paradigm. Such an outcome could tion), and (2) in relative terms, amnesics come about, however, if control subjects in showed greater reliance on a general semantic the Cermak-type paradigm can oppose the ef- match between study and test items than did fects of the gist trace by calling on a speci®c controls (because of an absent speci®c repre- trace, whereas amnesics have no speci®c trace sentation). While we offer this hypothesis ten- available to oppose the effects of the gist trace. tatively, it offers a promising avenue for future Thus, even a degraded gist representation investigation. could result in enhanced false recognition on Some of the same considerations apply to the part of amnesic patients. our free recall data. When their overall intru- Finally, we wish to highlight that our study sion rate was taken into account, amnesic pa- points toward the potential usefulness of neu- tients showed a lower proportion of critical ropsychological data in analyzing and decom- lure intrusions than did control subjects. How- posing memory illusions. Just as studies of ever, at the same time, control subjects re- priming and procedural learning in amnesia called more targets than critical lures, whereas have helped to dissociate underlying systems amnesic patients exhibited the opposite pat- and processes, our results, considered together tern. These qualitative differences are explica- with previous data, support the conclusion that ble if amnesic patients' production of studied distinct underlying processes subserve differ- words on the free recall test depends entirely ent kinds of false recognition effects. Studies on a degraded gist representation, because of other neuropsychological populations im- they have no speci®c representation of partic- plicate additional sources of memory illu- ular items. While an appealing idea, it does sions. For example, Schacter, Curran, Galluc- not readily accommodate one curious feature cio, Milberg, and Bates (in press) have re- of our results: studied words that were pro- cently documented excessive false recognition duced on the free recall test were more likely in a patient with a right frontal lobe lesion, to be recognized by both controls and amne- which they characterize in terms of de®cient sics than were nonproduced words, whereas retrieval processes involved in generating a critical lures that were produced on the free focused description of a study episode (see recall test were more likely to be recognized Norman & Schacter, in press; for a similar than nonproduced critical lures by control sub- false recognition phenomenon, see Parkin, jects but not by amnesic patients. This ®nding Bindschaelder, Harsent, & Metzler, in press). implies that studied words that amnesic pa- In a related vein, previous research and theo- tients produced on the free recall test differed rizing with confabulating patients has high-

AID JML 2457 / a002$$$$24 04-11-96 00:16:57 jmla AP: JML 332 SCHACTER, VERFAELLIE, AND PRADERE lighted how disturbances of strategic retrieval of linguistic ideas. Cognitive Psychology, 2, 331± and monitoring processes subserved by the 350. CECI,S.J.,&BRUCK, M. (1993). Suggestibility of the frontal lobes, basal forebrain, and related child witness: A historical review and synthesis. Psy- structures can lead to memory distortions (cf., chological Bulletin, 113(3), 403±439. Johnson, 1991; Moscovitch, 1995; Norman & CERMAK, L. S., BUTTERS,N.,&GERREIN, J. (1973). The Schacter, 1996). Phelps and Gazzaniga (1992) extent of the verbal encoding ability of Korsakoff and Metcalfe, Funnell, and Gazzaniga (1995) patients. Neuropsychologia, 11, 85±94. have shown that memory illusions can arise CERMAK,L.S.,&VERFAELLIE, M. (1992). The role of ¯uency in the implicit and explicit task performance in split-brain patients when recognition re- of amnesic patients. In L. R. Squire & N. Butters sponses are based exclusively on the left (Eds.), Neuropsychology of memory (pp. 36±45). hemisphere, which generates a categorical or New York: Guilford Press. gist-like representation, as opposed to a more CERMAK, L. S., VERFAELLIE, M., SWEENEY, M., & JA- speci®c representation that is associated with COBY, L. L. (1992). Fluency versus conscious recollection in the word completion performance of amne- the right hemisphere. And as noted earlier, sic patients. Brain and Cognition, 20, 367±377. Kroll et al. (1996) and Reinitz et al. (1996) COHEN,N.J.,&EICHENBAUM, H. (1993). Memory, amne- report that amnesics and patients with damage sia, and the hippocampus. Cambridge, MA: MIT to the left or to the right hippocampal system Press. are more prone to illusory memory conjunc- DEESE, J. (1959). On the prediction of occurrence of par- tions than are normal controls. This may re- ticular verbal intrusions in immediate recall. Journal of Experimental Psychology, 58, 17±22. ¯ect impaired consolidation processes that or- GABRIELI, J., FLEISCHMAN, D., KEANE, M., REMINGER, dinarily serve to bind together distinct attri- S., & MORRELL, F. (1995). Double dissociation be- butes of an event into a uni®ed engram. tween memory systems underlying explicit and im- Considered together with our results, all of plicit memory in the human brain. Psychological Sci- these studies point toward the conclusion that ence, 6, 76±82. GARDINER,J.M.,&JAVA, R. I. (1993). Recognizing and a variety of distinct neural and psychological remembering. In A. F. Collins, S. E. Gathercole, processes underlie different kinds of memory M. A. Conway, & P. E. Morris (Eds.), Theories of illusions. We believe that future brain-ori- memory (pp. 163±188). Hove, United Kingdom: Erl- ented studies can help to reveal and disentan- baum. gle them, and thereby establish a cognitive GLANZER, M., & ADAMS, J. K. (1990). The mirror effect in recognition memory: Data and theory. Journal of neuroscience of memory distortion that can Experimental Psychology: Learning, Memory, and illuminate these important but perplexing vul- Cognition, 16, 5±16. nerabilities of human memory. HERMAN, J., & HARVEY, M. (1993). The false memory debate: Social science or social backlash. Harvard Mental Health Letter, 9, 4±6. REFERENCES HINTZMAN, D. L. (1988). Judgements of frequency and BARTLETT, F. C. (1932). Remembering. Cambridge: Cam- recognition memory in a multiple-trace memory bridge Univ. Press. model. Psychological Review, 95, 528±551. BOWERS,J.S.,&SCHACTER, D. L. (1993). Priming of JACOBY, L. L. (1991). A process dissociation framework: novel information in amnesic patients: Issues and Separating automatic from intentional uses of mem- data. In P. Graf & M. E. J. Masson (Eds.), Implicit ory. Journal of Memory and Language, 30, 513± memory: New directions in cognition, development, 541. and neuropsychology (pp. 303±326). New York: Ac- JACOBY, L. L., KELLEY, C. M., & DYWAN, J. (1989). ademic Press. Memory attributions. In H. L. Roediger & F. I. M. BRAINERD, C. J., REYNA,V.F.,&BRANDES, E. (1995). Craik (Eds.), Varieties of memory and conscious- Are children's false memories more persistent than ness: Essays in honour of Endel Tulving (pp. 391± their true memories? Psychological Science, 6, 359± 422). Hillsdale, NJ: Erlbaum. 364. JOHNSON, M. K. (1991). Reality monitoring: Evidence BRAINERD, C. J., REYNA,V.F.,&KNEER, R. (1995). from confabulation in organic brain disease patients. False-recognition reversal: When similarity is dis- In G. P. Prigatano & D. L. Schacter (Eds.), Aware- tinctive. Journal of Memory and Language, 34, 157± ness of de®cit after brain injury: Clinical and theo- 185. retical issues (pp. 176±197). New York: Oxford BRANSFORD, J. D., & FRANKS, J. J. (1971). The abstraction Univ. Press.

AID JML 2457 / a002$$$$25 04-11-96 00:16:57 jmla AP: JML AMNESIA AND MEMORY ILLUSIONS 333

JOHNSON,M.K.,&CHALFONTE, B. (1994). Binding com- lam, & L. E. Sullivan (Eds.), Memory distortion (pp. plex memories: The role of reactivation and the hip- 226±251). Cambridge, MA: Harvard Univ. Press. pocampus. In D. L. Schacter & E. Tulving (Eds.), NORMAN,K.A.,&SCHACTER, D. L. (in press). Implicit Memory systems 1994 (pp. 311±350). Cambridge, memory, explicit memory, and false recollection: A MA: MIT Press. cognitive neuroscience perspective. In L. M. Reder JOHNSON, M. K., HASHTROUDI, S., & LINDSAY,D.S. (Ed.), Implicit memory and metacognition. Hillsdale, (1993). Source monitoring. Psychological Bulletin, NJ: Erlbaum. 114, 3±28. OFSHE,R.,&WATTERS, E. (1994). Making monsters: KELLEY,C.M.,&JACOBY, L. L. (1990). The construction False memories, psychotherapy, and sexual hysteria. of subjective experience: Memory attributions. New York: Scribner. Mind & Language, 5, 49±68. PARKIN, A. J., BINDSCHAEDLER, C., HARSENT, L., & KNOWLTON,B.J.,&SQUIRE, L. R. (1995). Remembering METZLER, C. (in press). Veri®cation impairment in and knowing: Two different expressions of declara- the generation of memory de®cit following ruptured tive memory. Journal of Experimental Psychology: aneurysm of the anterior communicating artery. Learning, Memory, and Cognition, 21, 699±710. Brain and Cognition. KORIAT,A.,&GOLDSMITH, M. (1994). Memory in natu- PHELPS, E., & GAZZANIGA, M. S. (1992). Hemispheric ralistic and laboratory contexts: Distinguishing the differences in mnemonic processing: The effects of accuracy-oriented and quantity-oriented approaches left hemisphere interpretation. Neuropsychologia, to memory assessment. Journal of Experimental Psy- 30, 293±297. chology: General, 123, 297±315. READ, J. D. (1996). From a passing thought to false memory in 2 minutes: Confusing real and illusory events. KORIAT,A.,&GOLDSMITH, M. (in press). Memory meta- phors and the everyday-laboratory controversy: The Psychonomic Bulletin and Review 3, 105±111. correspondence versus the storehouse conceptions of REINITZ, M. T., LAMMERS,W.J.,&COCHRAN,B.P. memory. Behavioral and Brain Sciences. (1992). Memory conjunction errors: Miscombination of stored stimulus features can produce illusions of KROLL, N. E. A., KNIGHT, R. T., METCALFE, J., WOLF, memory. Memory and Cognition, 20, 1±11. E. S., & TULVING, E. (1996). Cohesion failure as a REINITZ, M. T., VERFAELLIE,M.,&MILBERG, W. (1996). source of memory illusions. Journal of Memory and Memory conjunction errors in normal and amnesic Language, 35, 176±196. subjects. Journal of Memory and Language, 35, LINDSAY,D.S.,&READ, J. D. (1994). Psychotherapy and 286±299. memories of childhood sexual abuse: A cognitive REYNA,V.F.,&BRAINERD, C. J. (1995). Fuzzy-trace perspective. Applied Cognitive Psychology, 8, 281± theory: An interim synthesis. Learning and Individ- 338. ual Differences, 7, 1±75. LOFTUS, E. F. (1979). Eyewitness testimony. Cambridge, ROEDIGER,H.L.,&MCDERMOTT, K. B. (1994). The prob- MA: Harvard Univ. Press. lem of differing false-alarm rates for the process dis- LOFTUS, E. F. (1993). The reality of repressed memories. sociation procedure: Comment on Verfaellie and American Psychologists, 48, 518±537. Treadwell (1993). Neuropsychology, 8, 284±288. LOFTUS, E. F., MILLER,D.G.,&BURNS, H. J. (1978). ROEDIGER, H. L., III, & MCDERMOTT, K. B. (1995). Creat- Semantic integration of verbal information into a vi- ing false memories: Remembering words not pre- sual memory. Journal of Experimental Psychology: sented in lists. Journal of Experimental Psychology: Human Learning and Memory, 4, 19±31. Learning, Memory, and Cognition, 21, 803±814. MCCLELLAND, J. L., MCNAUGHTON, B. L., & O'REILLY, SCHACTER, D. L. (1994). Priming and multiple memory R. C. (1995). Why there are complementary learning systems: Perceptual mechanisms of implicit memory. systems in the hippocampus and neocortex: Insights In D. L. Schacter & E. Tulving (Eds.), Memory sys- from the successes and failures of connectionist mod- tems 1994. Cambridge, MA: MIT Press. els of learning and memory. Psychological Review, SCHACTER, D. L. (1995a). Memory distortion: History and 102, 419±457. current status. In D. L. Schacter, J. T. Coyle, G. D. METCALFE, J., FUNNELL, M., & GAZZANIGA, M. S. (1995). Fischbach, M. M. Mesulam, & L. E. Sullivan (Eds.), Right-hemisphere memory superiority: Studies of a Memory distortion: How minds, brains and societies split-brain patient. Psychological Science, 6, 157± reconstruct the past (pp. 1±43). Cambridge, MA: 164. Harvard Univ. Press. MOSCOVITCH, M. (1994). Memory and working-with- SCHACTER, D. L. (1995b). Memory wars. Scienti®c Ameri- memory: Evaluation of a component process model can, 272, 135±139. and comparisons with other models. In D. L. SCHACTER, D. L. (1996). Searching for memory: The Schacter & E. Tulving (Eds.), Memory systems 1994 brain, the mind, and the past. New York: Basic (pp. 269±310). Cambridge, MA: MIT Press. Books. MOSCOVITCH, M. (1995). Confabulation. In D. L. SCHACTER, D. L., ALPERT, N. M., SAVAGE, C. R., RAUCH, Schacter, J. T. Coyle, G. D. Fischbach, M. M. Mesu- S. L., & ALBERT, M. S. (1996). Conscious recollec-

AID JML 2457 / a002$$$$25 04-11-96 00:16:57 jmla AP: JML 334 SCHACTER, VERFAELLIE, AND PRADERE

tion and the human hippocampal formation: Evi- Evidence for a dissociable memory function. Quar- dence from positron emission tomography. Proceed- terly Journal of Experimental Psychology, 38A, ings of the National Academy of Sciences, 93, 321± 619±644. 325. SHIMAMURA, A. P., & SQUIRE, L. R. (1987). A neuropsy- SCHACTER, D. L., CHIU,C.Y.P.,&OCHSNER,K.N. chological study of fact memory and source amnesia. (1993). Implicit memory: A selective review. Annual Journal of Experimental Psychology: Learning, Review of Neuroscience, 16, 159±182. Memory, and Cognition, 13, 464±473. SCHACTER,D.L.,&CURRAN, T. (1995). The cognitive SQUIRE, L. R. (1992). Memory and the hippocampus: A neuroscience of false memories. Psychiatric Annals, synthesis from ®ndings with rats, monkeys, and hu- 25, 726±730. mans. Psychological Review, 99, 195±231. ULIN OOLING SCHACTER, D. L., CURRAN, T., GALLUCCIO, L. D., MIL- S ,R.A.,&D , D. J. (1974). Intrusion of a thematic idea in retention of prose. Journal of Exper- BERG,W.,&BATES, J. (in press). False recognition and the right frontal lobe: A case study. Neuropsy- imental Psychology, 103, 255±262. chologia. TULVING, E. (1985). How many memory systems are there? American Psychologist, 40, 385±398. SCHACTER, D. L., HARBLUK,J.L.,&MCLACHLAN,D.R. TULVING, E., & SCHACTER, D. L. (1990). Priming and (1984). Retrieval without recollection: An experi- human memory systems. Science, 247, 301±306. mental analysis of source amnesia. Journal of Verbal UNDERWOOD, B. J. (1965). False recognition produced by Learning and Verbal Behavior, 23, 593±611. implicit verbal responses. Journal of Experimental SCHACTER, D. L., REIMAN, E., UECKER, A., POLSTER, Psychology, 70, 122±129. UN OOPER M. R., Y ,L.S.,&C , L. A. (1995). Brain VERFAELLIE, M., & TREADWELL, J. (1993). Status of rec- regions associated with retrieval of structurally co- ognition memory in amnesia. Neuropsychology, 7, herent visual information. Nature, 376, 587±590. 5±13. SHIFFRIN, R. M., HUBER, D. E., & MARINELLI, K. (1995). WELLS,G.L.,&LOFTUS, E. F. (1984). Eyewitness testi- Effects of category length and strength on familiarity mony: Psychological perspectives. New York: Cam- in recognition. Journal of Experimental Psychology: bridge Univ. Press. Learning, Memory and Cognition, 21, 267±287. SHIMAMURA, A. P. (1986). Priming effects in amnesia: (Received July 4, 1995)

AID JML 2457 / a002$$$$25 04-11-96 00:16:57 jmla AP: JML