Remembering and knowing: Two different expressions of declarative memory

Journal of Experimental Psychology:Learning, Memory, and Cognition1995. Vol. 21. No. .VfiW-710

In the public domain

Remembering and Knowing:Two Different Expressions of Declarative Memory

Barbara J. KnowltonUniversity of California, San Diego

Larry R. SquireVeterans Affairs Medical Center, San Diego, and

University of California, San Diego

Amnesic patients and a control group were given a recognition test 10 min after studying words.For each recognized word, participants indicated whether they remembered it (R) or whethersimply they knew that the word was presented but had no recollections about it (K). The patientswere impaired for both R and K responses, performing like a control group tested after 1 week.Another control group was tested both 10 min and 1 week after study. The proportion of wordsinitially eliciting an R response and later eliciting a K response exceeded the proportion of Kresponses that shifted to R responses. These data are accounted for if items initially eliciting Rresponses can also elicit K responses. We conclude that the R-K distinction does not reflect theoperation of explicit and implicit memory but reflects instead a distinction within declarativememory. Thus, K responses depend on brain structures damaged in amnesia; R responses dependon these same structures and also on the frontal lobes for contextual information.

Recall and recognition memory have been proposed todepend on two different kinds of experience. When an old itemevokes a recollection of having specifically encountered theitem, an individual is said to "remember" (R). By contrast,when one has a sense of familiarity about a previouslypresented item, without actually recollecting a specific previ-ous encounter with the item, one is said to experience"knowing" (K; Tulving, 1985). One knows that the item waspresented before but is unable actually to remember the itemas part of the past.

In certain respects, the distinction between rememberingand knowing appears similar to the distinction between explicitand implicit memory (or declarative and nondeclarativememory; for reviews of this distinction, see Schacter, Chiu, &Ochsner, 1993; Squire, Knowlton, & Musen, 1993; Tulving &Schacter, 1990). This distinction contrasts the capacity forconscious recollection about recently occurring facts andevents with the capacity for a variety of nonconscious memoryabilities including priming, skill and habit learning, and simpleforms of classical conditioning. The suggestion that remember-ing and knowing are related to explicit and implicit memory,respectively, is based on the finding that R responses aresensitive to levels-of-processing effects, whereas K responses

Barbara J. Knowlton, Department of Psychiatry, University ofCalifornia, San Diego; Larry R. Squire, Psychiatry Service, VeteransAffairs Medical Center, San Diego, and Departments of Psychiatryand Neurosciences, University of California, San Diego.

This research was supported by the Medical Research Service of theDepartment of Veterans Affairs, National Institute of Mental Health(NIMH) Grant MH24600, the Office of Naval Research, the Mc-Knight Foundation, and an NIMH postdoctoral fellowship.

We thank Nicole Champagne and Joyce Zouzounis for researchassistance.

Correspondence concerning this article should be addressed toLarry R. Squire, Psychiatry Service (116A), Veterans Affairs MedicalCenter, 3350 La Jolla Village Drive, San Diego, California 92161.Electronic mail may be sent via Internet to [email protected].

are not (Gardiner, 1988), as well as the finding that Rresponses are reduced when study items are presented duringdivided attention, but K responses are not (Gardiner & Parkin,1990). It has also been shown that R responses and K.responses do not simply reflect changes in memory strength ordifferences in the confidence that individuals have aboutwhether test items are old or new (Gardiner & Java, 1990;Parkin & Walter, 1992). Finally, R responses show substantialforgetting across a 1-day interval, whereas K responses declinemuch more gradually (Gardiner & Java, 1991). In each of thesecases, similar dissociations have been demonstrated betweenexplicit and implicit memory (for reviews, see Richardson-Klavehn & Bjork, 1988; Schacter et al., 1993). These observa-tions have suggested that R responses reflect a component ofmemory performance dependent on explicit memory, whereasK responses reflect a component of memory performance thatderives from implicit memory (e.g., perceptual representationsystems [PRS]; Tulving & Schacter, 1990; for additionaldiscussion, see Gardiner & Java, 1993).

Yet there is a different way to understand the distinctionbetween remembering and knowing. Both experiences mightdepend on the medial temporal lobe-diencephalic brain struc-tures that support declarative memory, but remembering mayrequire some additional processing beyond what is requiredfor knowing (Tulving, 1989). Specifically, in addition to medialtemporal lobe-diencephalic structures, R responses may re-quire the contribution of the frontal lobes to access specificinformation, including temporal information, that was associ-ated with the presentation of each item during learning.Frontal lobe damage has been shown to impair source memory,that is, memory for when and where information was acquired(Janowsky, Shimamura, & Squire, 1989), and memory forsource should be important for making R responses. Consis-tent with this view, a significant correlation was obtained inolder individuals between the probability of an R response andtheir performance on the Wisconsin Card Sorting Test

699

700 BARBARA J. KNOWI.TON AND LARRY R. SQUIRT.

Table 1Characteristics of Amnesic Patients

Patient

A.B.a

P.H.W.H.L.J.J.L.

N.A.M.G.N.C.b

R.C.b

N.F."V.F.b

P.N.b

J.W.h

M

Age(years)

5469695472

545948755773645561.8

WAIS-RIQ

10411511398

116

109HI90

10694

1039998

104.3

Attention

Hippocampal

8711788

105122

Verbal

formation

6267728373

Diencephalon

10211362

115919381

10498.5

678980766277776573.1

WMS-R

Visual

7283826083

898460977365737076.2

General

5470676974

688669805367675767.8

Delay

< 5057

< 50< 50

58

7163

< 5072

< 5064535757.3

Note. The WAIS-R and the WMS-R indexes yield a mean score of 100 in the normal population with astandard deviation of 15. The WMS-R does not provide scores for individuals who score below 50.Therefore, the four scores below 50 were scored as 50 for calculating a group mean. WAIS-R = WechslerAdult Intelligence Scale—Revised (Wechsler, 1981); WMS-R = Wechsler Memory Scale—Revised(Wechsler, 1987).aThe lesion site for the patient has not been confirmed radiologically but is strongly supported by theetiology of amnesia (see text). These patients had diencephalic amnesia as a result of alcoholicKorsakoff's syndrome.

(WCST; Nelson, 1976), a measure of frontal lobe dysfunction(Parkin & Walter, 1992).

This idea is related to the view that remembering andknowing are operations of episodic and semantic memory,respectively (Tulving, 1989). It has previously been suggestedthat both episodic and semantic memory depend on medialtemporal lobe and diencephalic brain structures and thatepisodic memory depends additionally on the frontal lobes(Shimamura & Squire, 1987; Squire et al., 1993; Tulving,1989).

A neuropsychological analysis based on the organization ofbrain systems provides a favorable way to explore the distinc-tion between R and K responses. Bilateral damage to medialtemporal lobe or diencephalic structures causes an amnesiccondition that selectively impairs declarative memory. Amne-sic patients fail conventional tests of recall or recognition, butthey are fully intact on many other tasks that assess priming,skill learning, and other forms of nondeclarative (implicit)memory (Hintzman, 1990; Squire et al., 1993; Weiskrantz,1988). Accordingly, the performance of amnesic patientsshould reveal the relative contributions of declarative (ex-plicit) memory to R and K responses. If amnesic patientsexhibit a reduction in the accuracy of R responses duringrecognition memory testing, compared with controls, with noreduction in the accuracy of K responses, then K responsesmust depend substantially on a kind of memory (presumably,implicit memory) that is intact in amnesia. However, if amnesicpatients exhibit a similar reduction in both R and K responses,then R and K responses must both depend on the kind ofmemory (i.e., declarative or explicit memory) that is impairedin amnesia.

Experiment 1

We tested patients with bilateral damage to the hippocam-pal formation, patients with circumscribed damage to themidline diencephalon, and patients with alcoholic KorsakofFssyndrome who have diencephalic damage together with somefrontal lobe atrophy. The recognition memory performance ofthese amnesic patients at a 10-min retention interval wascompared with the performance of a matched control grouptested after the same interval. Finally, to determine how R andK responses behave when memory is weakened throughforgetting, we also tested a control group after a 1-weekretention interval.

Method

Amnesic patients. Thirteen amnesic patients (9 men and 4 women)were tested (Tables 1 and 2). Six of the patients had alcoholicKorsakoffs syndrome. They had participated in either a magneticresonance imaging study (Squire, Amaral, & Press, 1990; for N.F.,unpublished observations) or a quantitative computed tomographystudy (Shimamura, Jernigan, & Squire, 1988). These demonstratedmarked reductions in the volume of the mammillary nuclei, reducedthalamic tissue density, and frontal lobe atrophy. Patient M.G.(female) sustained a bilateral thalamic infarction. Patient N.A. sus-tained a penetrating brain injury involving primarily the left dience-phalic region (Squire, Amaral, Zola-Morgan, Kritchevsky, & Press,1989). Of the 5 other patients, 4 (P.H., W.H., L.J., and J.L.) hadbilateral hippocampal damage identified by magnetic resonance imag-ing (for W.H. and J.L., Squire et al., 1990; for P.H., Polich & Squire,1993; for L.J., unpublished observations). Patient A.B., who wasunable to participate in magnetic resonance imaging studies, became

REMEMBERING AND KNOWING 701

Table 2Performance on Standard Memory Tests by Amnesic Patients

Patient

A.B.P H .W.H.L.J.J.L.N.A.M.G.N.C.R.C.N.F.V.F.P.N.J.W.

MControl (n = 8)

M

Diagramrecall

43131

1700348243.8

20.6

10000001000100.2

5.6

Pairedassociates

10000000000100.2

7.6

21000221020121.0

8.9

Wordrecall(%)

3327404040493323193627292932.7

71

Wordrecognition

{%)

8384849393937171857691839084.4

97

50words

3236293331343031372827312931.4

41.1

50faces

3334242920423437302731313431.2

38.1

Note. The diagram recall score was based on delayed (12-min) reproduction of the Rey-OsterriethComplex Figure (Osterrieth, 1944; maximum score = 36). The average score for copying the figure was26.6, a normal score (Kritchevsky, Squire, & Zouzounis. 1988). The paired-associate scores were thenumber of word pairs recalled on three successive trials (maximum score = 10 per trial). The word recallscore was the percentage of words identified correctly on five successive study-test trials (Rey, 1964). Theword recognition score was the percentage of words identified correctly by yes-no recognition across fivesuccessive study-test trials. The scores for words and faces were based on a 24-hr recognition test of 50words or 50 faces (modified from Warrington, 1984; maximum score = 50, chance = 25). The mean scoresfor normal controls shown for these tests are from Squire and Shimamura (1986). Note that patient N.A.was not severely impaired on the two nonverbal memory tests because his brain injury is primarily leftunilateral.

amnesic in 1976 after an anoxic episode and was presumed to havehippocampal damage on the basis of this etiology.

The 13 patients averaged 62.5 years of age and 13.5 years ofeducation. Individual IQ scores on the Wechsler Adult IntelligenceScale—Revised (WAIS-R; Wechsler, 1981) and scores on the Wechs-ler Memory Scale—Revised (WMS-R; Wechsler, 1987) appear inTable 1. Immediate and delayed (12 min) recall of a short prosepassage averaged 4.8 and 0 segments, respectively (21 total segments;Gilbert, Levee, & Catalano, 1968). Scores on other memory testsappear in Table 2. The mean score on the Dementia Rating Scale(Mattis, 1976) was 132.2 (range = 125-143; maximum score = 144).Most of the points were lost on the memory subportion of the test(mean points lost = 7.0). The mean score on the Boston Naming Test(Kaplan, Goodglass, & Weintraub, 1978) was 55.2 (range = 48-59;maximum score = 60). Scores for normal individuals on these tests canbe found elsewhere (Janowsky, Shiniamura, Kritchevsky, & Squire,1989; Squire et al., 1990).

Other participants. Two control groups were tested (Group 1:n = 14, 6 men and 8 women; Group 2: n = 11, 6 men and 5 women).They were either employees or volunteers at the San Diego VeteransAffairs Medical Center or members of the retirement community ofthe University of California, San Diego. The participants in bothgroups were selected to match the amnesic patients with respect to age(Group 1: 62.2 years, range = 54-73; Group 2:62.9 years, range = 51-73), education (Group 1: 15.1 years; Group 2: 13.3 years), andWAIS-R subtest scores for Information (22.2 for Group 1; 21.4 forGroup 2; and 20.6 for the amnesic patients) and Vocabulary (56.5 forGroup 1; 56.7 for Group 2; and 54.7 for the amnesic patients). ForGroup 1, immediate and delayed recall of the short prose passage was7.8 and 6.0 segments, respectively; for Group 2, immediate anddelayed recall of the short prose passage was 6.8 and 5.9 segments,respectively.

Materials. Two sets of 72 six-letter English words were constructedfrom a pool published previously (Gibson & Watkins, 1988). Thewords were printed in capital letters on 3 x 5 in index cards. The firstset of words averaged 24 occurrences per million, and the second set 21per million.

Procedure. Testing was done individually. Words were presentedfor 2 s each with instructions to pay attention to each word becauselater "you will be asked to remember them." Each participant studied36 of the 72 words, and the other 36 were used as distractor items onthe later test. The half of the items that were study items and the halfthat were distractor items were counterbalanced across participants.The amnesic patients and the first control group were tested 10 minafter the study phase. The second control group was tested after 1week.

After the scheduled delay interval (10 min or 1 week), participantswere given a recognition memory test involving the 36 study items andthe 36 distractors. Participants were first instructed that they would bemaking two kinds of recognition responses: R (remember) and K(know). The difference between R and K responses was then ex-plained according to the method of Gardiner and Parkin (1990).' Itwas explained that an R response would be appropriate in a circum-stance such as when one remembers a recent television program and isable to recollect specific details about the experience such as when andwith whom it was viewed. A K response was described as appropriatewhen one had the experience of recognizing a person, being sure thatthe person is known, but being unable to recollect any specific detailsat all about the person, such as the person's name. Participants were

'We thank J. Gardiner for providing the printed instructions that heused to collect R and K responses. We followed these instructions asclosely as possible.

702 BARBARA J. KNOWLTON AND LARRY R. SQUIRE

2.4

2.0

1.6

s8

I 1.2|

| 0.8um5

0.4

CON-10 MIN CON -1 WEEK AMN -10 MIN

R K R K R K

Figure 1. The d' scores for items labeled remember (R; open bars) orknow (K; closed bars) by amnesic patients (AMN) and two controlgroups (CON). The amnesic patients were tested 10 min after thestudy phase, and the control groups were tested 10 min or 1 week afterthe study phase. Brackets show standard errors of the mean.

then asked to give examples of memories that they would associatewith R and K responses.

After the concept of both R and K recognition responses was clear,participants were told that they would be using R and K responses toreport the words that they recognized from the study phase. Partici-pants were then given a list of 72 words arranged in four columns andwere told to circle each word that they confidently recognized from thestudy cards. Beside each circled word they were asked to write either Ror K. They were further told that an R response should signify that therecognized word evoked specific recollections such as what wasthought of when the word appeared, what it looked like on the card, orwhen it occurred in the series. A K response should signify that oneknew that the word had been presented but had no specific recollec-tions about it. A summary of these instructions was printed on a cardthat was in full view during the entire test. Participants were instructedthat, as they worked through the list, they should not return to anyprevious items. Finally, after the first R and the first K responses weremade, participants were asked to explain their choices. Participantswere also asked to explain their first R and K responses from thesecond half of the test. This procedure provided a check that R and Kresponses were being used in accordance with the instructions.

Both the amnesic patients and the 10-min delay control group weretested in two separate sessions using different sets of materials. Thescores obtained for the two sessions were averaged for each partici-pant. For the amnesic patients, the two tests were separated by anaverage interval of 41 days; for those in the control group, the two testswere separated by an average of 40 days. The control group given a1-week delay was tested only once using the second set of materials.

Results

The amnesic patients and members of the control groupswere all able to give valid examples of R and K responsesduring the instruction period before the test, and they wereable to justify their R and K responses appropriately duringthe test. There were no detectable differences between thejustifications given by the control groups and those given by theamnesic patients. Participants usually justified an R responseby describing what they had thought of when they had seen the

word on the card during study. They justified a K response bystating that they were confident that the word was presentedbefore because they "just knew it was there," because they"had a feeling" it was there, or because it looked familiar,although they did not remember actually seeing the word.

Figure 1 shows discriminability (d') scores for items thatwere labeled R and K by each of the three groups. All threegroups successfully discriminated study items from distractoritems for both response types; that is, they gave more R re-sponses and more K responses to study items than to distractoritems (fs > 4.0, ps < .05). Table 3 shows the percentage oftargets and distractors labeled R and K by each of the threegroups. For the groups receiving both sets of materials (thecontrol group tested after 10 min and the amnesic patients),the scores for the first and second test sessions differed by nomore than 5% for either R or K responses and were combined.

Control groups. A 2 x 2 (10-Min Delay vs. 1-Week-Delay x R vs. K Responses) analysis of variance (ANOVA)performed on d' scores for the controls showed a main effect ofgroup, F(l, 23) = 37.94, p < .01, MSE = 0.32, and aninteraction between group and response type, F(l, 23) = 7.20,p < .01, MSE = 0.24. A simple-effects analysis of the two typesof responses was also performed, with the corrected signifi-cance level set at .025. This analysis showed that individuals inthe control group tested after a 1-week delay had significantlyless R response accuracy, F(l, 44) = 41.15,/? < .01, MSE =0.28, and less K response accuracy, F(l, 44) = 8.45, p < .01,MSE = 0.28, than those tested after 10 min. The pattern ofresults was similar when a 2 x 2 ANOVA was performed onthe percentage of hits for each response type corrected bysubtracting the percentage of false alarms. There was a maineffect of group, F(l, 23) = 40.52, p < .01, MSE = 116.48, andan interaction between group and task, F(l, 23) = 7.20, p <.05, MSE = 208.23. A simple-effects analysis showed that therewas a significant difference between the groups in terms of Rresponse accuracy, F(l, 41) = 35.54, p < .01, MSE = 162.36,and no significant difference for K responses, F(l, 41) = 2.77,p = . 10, MSE= 162.36.

Finally, because there were 11 individuals in the 1-weekdelay group and two sets of materials, one set of materials wasused six times, and one set was used five times. To remove anypotential effect of this imbalance, we reanalyzed the data byusing data from only the first 10 individuals tested. The patternof results was the same as in the previous analyses: The 1-weekdelay group continued to show impaired performance for both

Table 3Percentage of Test Items Receiving R and K Responsesin Experiment 1

Response typeR

TargetsDistractors

KTargetsDistractors

Control groups

10-mindelay

45.9 ± 3.51.8 ± 0.6

21.2 ± 3.84.4 ± 1.2

1-weekdelay

24.2 ± 3.610.7 ± 2.7

24.6 ± 4.316.3 ± 2.8

Amnesic patients:10-min delay

17.3 ± 4.88.9 ± 3.0

16.1 ± 4.19.5 ± 2.9

Note. Scores are means ± SEMs. R = remember; K = know.


R and K responses (for d' scores), Fs( 1,42) > 12.4, ps < .01,MSE = 0.224.

Amnesic patients. A 2 x 2 ANOVA (10-Min Controls vs.Amnesic Patients x Response Type) performed on d' scoresshowed a main effect of group, F(l, 25) = 73.53, p < .01,MSE = 0.23, and an interaction with group and response type,F(l, 25) = 14.34, p < .01, MSE = 0.24. A simple-effectsanalysis, again with the corrected significance level set at .025,showed that the amnesic patients were significantly impairedat both R response accuracy, F( 1,49) = 75.84, p < .01, MSE =0.23, and K response accuracy, F(l, 49) = 10.90, p < .01,MSE = 0.23. By contrast, the amnesic patients performedsimilarly to the controls tested after a 1-week delay ( 2 x 2ANOVA comparing d' scores, Fs < 1). An analysis comparingthe 10-min control and amnesic groups based on hit ratescorrected by subtracting false alarms showed a pattern ofresult identical to that obtained for the d' data. There was amain effect of group, F(l, 25) = 95.33,p < .01, MSE = 74.46,and an interaction between group and task, F(l , 25) = 11.87,p < .01, MSE = 184.31. Finally, the simple-effects analysisshowed that the amnesic patients were impaired at R responseaccuracy, F(l, 42) = 66.34, p < .01, MSE = 129.38, and Kresponse accuracy, F(l, 42) = 5.43,p < .025, MSE = 129.38.

Because 13 amnesic patients were tested, one set of materi-als was used six times, and one set was used seven times in eachtest session. To remove any potential effect of materials, wereanalyzed the data by using only the data from the first 12patients tested. The pattern of results remained the same. Theamnesic patients were significantly impaired in terms of both Rand K response accuracy (for d' scores), Fs(l, 47) > 9.3.ps <.01, MSE = 0.239.

The 6 amnesic patients with KorsakofFs syndrome performedsimilarly to the 7 other amnesic patients. The patients withKorsakofFs syndrome gave 23% R responses to study items,12% K responses to study items, 12% R responses to distractoritems, and 7% K responses to distractor items. The correspond-ing values for the other amnesic patients were 13%, 19%, 6%,and 11%. None of these pairwise comparisons approachedsignificance (ts < 1.07,/>s > 0.1). In addition, for the patientswith KorsakofFs syndrome, the d' ± SEM score for R re-sponses was .34 ± .12 and for K responses was .18 ± .32. Forthe other amnesic patients, these values were .40 ± .15 and.43 ± .19, respectively. A 2 x 2 ANOVA indicated no signifi-cant effects of group or interactions with response type (Fs < 1).

Discussion

The amnesic patients exhibited a significant impairment inrecognition memory as measured by both R responses and Kresponses. Their performance matched rather closely theperformance of the control group tested after a delay of 1 weekinstead of 10 min. These findings suggest that both R and Kresponses (i.e., both remembering and knowing) depend onthe medial temporal lobe and diencephalic brain structuresthat are damaged in amnesia and that are essential fordeclarative memory. The results do not support the view thatknowing depends substantially on a perceptual representationsystem (or other form of implicit memory) that is intact inamnesia.

Compared with the 10-min delay control group, the amnesicpatients and the 1-week delay control group exhibited greaterimpairment for R responses than for K responses. The amnesicpatients performed similarly to the 1-week delay controlgroup. However, because the K response d' scores for amnesicpatients and 1-week delay controls were close to zero, it isdifficult to compare them. This could be important because ifamnesic patients could exhibit better K response accuracy thanthe 1-week delay control group, such a finding would beconsistent with an implicit-memory contribution to K re-sponses. However, such a contribution of implicit memory toK responses could not have been substantial because theamnesic patients were clearly impaired at making K responsesin comparison to the controls tested after a 10-min delay. Itseems more likely that the R and K responses elicited byamnesic patients depend on their residual declarative memory.

Because the frontal lobes appear to be important for makingR responses (Parkin & Walter, 1992), one might expect thatpatients with KorsakofFs disease, which is often accompaniedby some degree of frontal lobe damage, would exhibit impairedR responses relative to the performance of non-Korsakoffamnesic patients. In our study, these two subgroups of patientsperformed similarly. It is possible that no difference emergedbecause both subgroups were already performing so poorlybecause of their declarative memory impairment. In addition,the patients with KorsakofFs syndrome exhibited a consider-able range of frontal lobe dysfunction such that, as a group, theextent of frontal lobe dysfunction might not have been severeenough to differentiate them from the other patients. Neverthe-less, there was a significant correlation for Korsakoff patientsbetween the R response d' scores and scores on the WCST(number of categories achieved; r = .85, p < .05), consistentwith the idea that the frontal lobes are involved in making Rresponses. For the non-Korsakoff patients, this correlation wasnot significant {r = .32, p > .20). Finally, there was not asignificant correlation between d' scores for K responses andperformance on the WCST for either group of amnesicpatients (rs < .45,ps > .20).

Both the amnesic patients and the control group that wastested after 1 week exhibited elevated false-alarm rates for Rand K responses. This finding suggests that the elevatedfalse-alarm rates reflect weakened memory and are not due tothe amnesic syndrome per se. At the same time, it is peculiarthat the false-alarm rate that we observed in controls after 1week is higher than what was observed in an earlier, similarstudy of normal individuals that examined R and K responsesas a function of retention interval (Gardiner & Java, 1991). Inthat study, false-alarm rates at the 1-week retention intervalwere only 5% for R responses and 8% for K responses (ourcontrols at one week: 11% for R responses and 17% for Kresponses). Is it possible that the finding of reduced recogni-tion memory performance in amnesic patients, as assessed byboth R and K responses, depends on finding anomalously highfalse-alarm rates?

One difference between this study and the earlier one(Gardiner & Java, 1991) is that our study group was mucholder (mean age = 62.9 years for the control group tested after1 week) than the undergraduate students tested in the earlierstudy. Thus one possibility is simply that older individuals

704 BARBARA J. KNOWI.TON AND LARRY R. SQUIRE

2.4i')

Sco

re

b>

b

llty

((

a

| 0.8(O

O0.4

n

10 MIN

I1 WEEK

IT

K

Figure 2. The d' scores for items labeled remember (R: open bars) orknow (K; closed bars) by two groups of young individuals tested at twodifferent intervals after the study phase. Brackets show standard errorsof the mean.

often have high false-alarm rates compared with young people.A second possibility is that individuals in our study useddifferent criteria for making their R and K responses than theindividuals tested by Gardiner and Java (1991). This possibilitydeserves consideration because our early experience with thistest procedure indicated that R and K responses, as well asfalse-alarm rates, are quite sensitive to the particular instruc-tions that are delivered. To address this issue, we nextrepeated the test procedure from Experiment 1 but withyounger individuals.

Experiment 2

The question of interest was what pattern of R and Kresponses, and what false-alarm rate, would be exhibited byyounger people. A finding similar to what was reported earlierfor young people (Gardiner & Java, 1991) would validate ourtesting procedure and indicate that the high false-alarm rateobserved in Experiment 1, for the controls tested after 1 week,was probably due to the age of the group.

Method

Participants. Participants were recruited from the San DiegoVeterans Affairs Medical Center and from the local community andwere assigned to one of two groups. For Group \{n = 13,3 men and 10women), the participants averaged 24.5 years of age (range = 22-27years) and 16.3 years of education. Group 1 was tested 10 min after thestudy phase. For Group 2 (n = 12, 3 men and 9 women), theparticipants averaged 25.2 years of age (range = 22-33 years) and 15.6years of education. Group 2 was tested 1 week after the study phase.

Materials and procedure. The second set of words from Experiment1 served as test materials for Experiment 2. The study and testprocedures were the same as in Experiment 1.

Results and Discussion

Figure 2 shows the d' scores for items that were labeled Rand K, and Table 4 lists the proportion of target items and

distractor items that were given each of the two types ofresponses. Both groups were able to discriminate study itemsfrom distractor items for both response types (fs > 3.5.ps < .01). With respect to their R and K responses to studyitems, the young groups resembled the older control group inExperiment 1. Thus, the young group tested after 10 min gavemore R responses than K responses to study items, l(\2) = 4.6,p < .01, and the young group tested after 1 week gave aboutthe same number of R and K responses to study items, i(l 1) =1.1, p > 0.1.

The finding of interest was that the false-alarm rate obtainedby the young group in Experiment 2 was quite similar to thefalse-alarm rate reported for young individuals by Gardinerand Java (1991) and was lower than the false-alarm rateobtained by the older group in Experiment 1 who were testedafter 1 week (for R responses, 4% vs. 11%), 1(21) = 2.15,p <.05; for K responses (7% vs. 17%), t(2\) = 2.90, p < .01. Inother respects, the young group tested after 1 week alsoperformed similarly to the young group tested by Gardiner andJava (1991). Our young group gave R and K responses to 23%and 18% of the study items and 4% and 7% of the distractoritems, respectively. For their young group, the correspondingfour scores were 25%, 23%, 5%, and 8%. The scores for thegroup tested after 10 min were also similar in the two studies(our scores: 47%, 14%, 1%, and 2%; their scores: 49%, 26%,0%, and 5%). These findings provide direct evidence that thehigh false-alarm rates in Experiment 1 of our study resultedfrom the fact that we tested older individuals. Moreover, ourtest protocol yielded results in normal individuals similar towhat has been observed in other laboratories. This findingprovides additional support for the conclusion of Experiment 1that R and K responses both depend on the kind of (declara-tive) memory that is impaired in amnesia.

Experiment 3

If both R and K recognition responses depend on declara-tive memory, the question remains about how the two kinds ofrecognition are related. One possibility, which is based on theframework discussed by Jones (1987), is that the memorycomponents giving rise to R and K responses are redundant(Figure 3). Redundancy effectively means that all items elicit-ing R responses have the potential to elicit K responses. Thisview of R and K responses is consistent with the proposal thatR and K responses reflect a distinction within declarativememory between episodic and semantic memory, respectively


Response type

RTargetsDistractors

KTargetsDistractors

Young group

10-min delay

47.0 ± 5.50.7 ± 0.5

14.2 ± 3.22.4 ± 0.9

1-week delay

22.9 ± 2.84.2 ± 1.5

18.1 ± 3.07.0 ± 1.7

Note. Scores are means ± SEMs. R = remember; K = know.


CE>REDUNDANCY INDEPENDENCE EXCLUSIVITY

Figure 3. Venn diagrams describing three possible relationshipsbetween the processes that lead to R (remember) and K (know)responses. The shaded regions indicate the conditions required for anR response, and the open regions indicate the conditions required fora K response. In the case of independence, an R response occurs onlywhen two independent processes are active.

(Tulving, 1989). An R response occurs when an item evokes anepisodic memory involving the item. (Episodic memory cannotbe retrieved without also retrieving semantic memory.) A Kresponse occurs when one knows that the item had occurred(semantic memory) but can retrieve no episodic memoriesinvolving the item. By this scenario, when an R response to anitem is lost, the item becomes a K. response on the route tobecoming a miss.

A second possibility is that R and K responses arise fromindependent memory components (Figure 3). For example,consider two independent processes supporting context memoryand semantic memory, respectively. Under one view of twosuch processes, contextual memory about an item and seman-tic memory for the item can be acquired independently. Onemight remember that one of the items was printed on a creasedindex card (context memory) but not recognize the item itself(semantic memory). Conversely, one might know that an itemwas on the list but forget that it was printed on a creased card.An R response would be given whenever one had bothsemantic memory for an item and some memory for thecontext in which the item occurred. A K response would begiven when one had semantic memory for an item but did notremember any contextual information. The independenceview resembles the redundancy view in that all items receivingR responses also have the potential to elicit K responses (i.e.,when context memory is lost). The independence view differsfrom the redundancy view in that those items that initiallyelicited R responses can also become "context-only" responsesand thereby become misses. For example, if an item elicitingan R response (because it had both semantic and contextualcomponents) were to lose its semantic component, the contex-tual information that remained (e.g., memory that one of theitems had been presented on a creased card) would not besufficient for recognition of the item. As a result, the itemwould be missed.

A third possibility (Gardiner & Java, 1990,1993; Gardiner &Parkin, 1990) is that the relationship between the memoryprocesses leading to R and K responses is exclusive (Figure 3).By this view, items give rise either to conscious recollections(and R responses) or to a perceptually driven sense offamiliarity (and K responses). By the exclusivity view, thefrequency of K responses need not change when R responsesare reduced. This view appears to gain support from the find-ing that K response levels can remain quite stable under manycircumstances in which R responses are dramatically reduced.

For example, when attention was divided at study. R responsesdeclined markedly, but K responses remained ;it about thesame level (Gardiner & Parkin, 1990). This finding is strikingbecause one might suppose a priori that many items elicitingan R response (i.e., when individuals remember somethingabout the specific item) also have the potential to elicit a Kresponse after the more specific information needed for an Rresponse is lost (i.e., because even if individuals cannotremember anything about the item they might at least knowthat the item was on the list). Yet, no increase in K responseswas observed with the divided-attention manipulation. Itwas concluded that the process giving rise to R re-sponses is exclusive of the process that gives rise to K re-sponses. However, another possibility is that an increase in Kresponses did occur when R responses were lost under dividedattention. This increase in K responses might not have beenobserved because it was matched by a loss of K responses, thatis, a reduction in the number of items supported by the Kprocess alone. In other words, the number of items that wereconverted by divided attention from an R response to a Kresponse may have been compensated for by a similar numberof items that were converted from a K response to a miss.

The question of what happens when R responses are lost isimportant because the different views about how R and Kresponses are related make different predictions. The exclusiv-ity view predicts that, as R responses decline, no appreciableconversion of R responses to K responses should occurbecause R and K responses are processed separately. Theredundancy and independence views both predict that if Rresponses are decreased, some of the items that had originallyelicited R responses should now elicit K responses. Finally, asR responses decline, the redundancy view predicts moreconversion of R to K responses than does the independenceview. To test these possibilities, a method is needed to trackthe fate of individual items in the same individual, so that it canbe determined directly what happens when the frequency of Rresponses is reduced. Do items that elicit an R response in onecondition subsequently elicit a K response or does this essen-tially never happen? In Experiment 3, we tested the samegroup of individuals 10 min after study and again 1 week afterstudy. The fate of the items given R and K responses duringthe first recognition test was examined at the 1-week test, at atime when R responses were expected to decline significantly.A finding that the rate of conversion of R responses to Kresponses is insignificant, or no larger than the rate ofconversion of K responses to R responses, would suggest thatthe processes contributing to R responses and to K responsesare exclusive. However, a finding that a substantial number ofitems receiving R responses at the first recognition testsubsequently receive K responses after a 1-week delay wouldsuggest that an item receiving an R response also has thepotential to elicit a K response. If so, the processes underlyingR and K responses cannot be exclusive. They must beredundant or independent.

Method

Participants. The participants (n =24, 13 men and 11 women)were either volunteers at the San Diego Veterans Affairs Medical


2.0£8« 1.6

j>«= 1.2jO(0c| 0.8&<n5

0.4

n

101WIN 1 WEEK

T

TT T


R K R K

Figure 4. The d' scores for items labeled remember (R; open bars) orknow (K; closed bars) by a control group tested at 10 min and again at 1week after the study phase.

Center or members of the retirement community of the University ofCalifornia, San Diego. None had participated in Experiments 1 or 2.The participants averaged 64.8 years of age (range = 47-80) and 14.1years of education. Scores on the Vocabulary and Informationsubscales of the WAIS-R were 54.1 and 21.9, respectively. Immediateand delayed recall of the short prose passage was 7.1 and 5.5 segments,respectively.

Materials. The stimuli consisted of a subset of the words used inExperiment 1 (three sets of 36 words each). For each participant, oneset served as target items and were presented to the participants forstudy exactly as in Experiment 1, a second set served as distractor itemsfor the recognition test that was given after a 10-min delay, and a thirdset served as distractor items for the recognition test that was givenafter a 1-week delay. Which set of words served as targets ordistractors was counterbalanced across participants. The constructionof the study cards and the test lists was the same as in Experiment 1.

Procedure. Using the same procedure as in Experiment 1, wepresented 36 target words and 10 min later gave a 72-word recognitiontest. All participants were also given a second recognition test after 1week. At that time, the instructions from the earlier recognition testwere repeated, emphasizing that judgments should be made on thebasis of "the words you saw on the cards, not the words you saw on thelist last week." Again, just as in Experiment 1, participants were askedto make R or K responses for the words that they recognized.

Results

Figure 4 shows d' scores for R and K responses after the10-min delay and after the 1-week delay, and Table 5 lists theproportion of study items and distractors that were labeled asR and K responses at each delay. Participants successfullydiscriminated study items from distractor items for bothresponse types; that is, they gave more R and K responses tostudy items than to distractor items at both study-test delays(fs > 5.9, ps < .01). These results, obtained by a single grouptested at two different delays, are similar to the resultsobtained when different groups were tested at the same twodelays (Experiment 1, Figure 1). Specifically, like the controlgroup in Experiment 1, the group in Experiment 3 produced

Response type 10-min delay 1-week delay

RTargetsDistractors

TargetsDistractors

33.9 ± 3.62.9 ± 1.0

18.6 ± 2.83.8 ± 0.8

17.5 ± 2.81.7 ±0.6

19.9 ± 3.26.7 ± 1.7

Note. Scores are means ± 5/TAfs. R = remember; K = know.

on average significantly more R responses than K responses tostudy items at the 10 min study-test delay (34% vs. 19%),f(23) = 3.0, p < .01, and produced a similar number of R andK responses to study items at the 1-week study-test delay (18%vs. 19%), ((23) = 0.45,/? > 0.1. The effect of the delay was toreduce the frequency of R responses, t(23) = 6.62, p < .01,without affecting the frequency of K responses. Discriminabil-ity (d' ) scores (Figure 4) showed this same pattern—R vs. Kresponses at the 10-min study-test delay: 1.52 vs. 0.83, ((23) =4.66, p < .01; R vs. K responses at the 1-week study-test delay: 0.99 vs. 0.75, r(23) = 1.39,p > 0.1.

One difference between the findings from Experiments 1and 3 was that in Experiment 3 the false-alarm rate at the1-week study-test delay was not significantly higher than thefalse-alarm rate at the 10-min study-test delay, as it was inExperiment 1—Experiment 1: for R responses, 2% to 11%; forK responses, 5% to 16%; Experiment 3: for R responses, 3% to2%, f(23) = 0.97,p > .10; for K responses, 4% to 7%, r(23) =1.90, p = .07. Nevertheless, by testing the same group twice attwo different retention intervals we obtained overall rathersimilar results to what was observed when two different groupswere tested at the same two retention intervals. Accordingly,we next conducted an item analysis to determine how indi-vidual study items were reported at the 10-min test (Rresponse, K response, or miss) and what the fate of each targetitem was at the 1-week test.

Table 6 shows the fate of individual study items across thetwo study-test delays. The scores are the percentage of itemsthat began as an R response, a K response, or a miss and laterbecame R responses, K responses, or misses, respectively. Themain question of interest was whether there was a significantpercentage of items that converted from R responses to Kresponses as R response accuracy declined. As Table 6 shows,fully 29% of study items that elicited an R response at the

Table 6/tew Analysis for Experiment 3

Response typefor study items

at 10 min

RKM

Note. Scores aremiss.

R

35.910.14.7

means ±

Fate of items at 1 week

(%)

±4.7±3.3± 1.3

SEMs. P

K(%)

28.5 ± 5.128.1 ± 4.69.8 ± 2.9

'. - remember; K =

35.8 ±61.6 ±85.5 ±

•• k n o w ;

•)

4.55.03.2

M =

REMEMBERING AND KNOWING 7(17

10-min test later elicited a K response at the 1-week test. Thisvalue was significantly above zero, r(23) = 5.59, p < .01, andsignificantly higher than the percentage of study items initiallygiven a K response that later converted to an R response(10%),f(23) = 2.75,p < .05.

We considered the possibility that more items changed fromR to K than from K to R because a bias was present for itemsrecognized from the first test (and not the study phase) to elicitK responses rather than R responses. As a measure of thebaseline tendency to produce R or K responses, we took thefrequency with which items that were missed (M) at the 10-mintest later elicited R or K responses at the 1-week test. Table 6shows that there was a small nonsignificant bias in favor of Kresponses (M to R = 5%, M to K = 10%), r(23) = 1.59,/? >.10. Accordingly, to obtain a corrected score that was free ofthis bias, we subtracted the percentage of items that convertedfrom M to K from the percentage of items that converted fromR to K (29% - 10% = 19%). Similarly, we subtracted thepercentage of items that converted from M to R from thepercentage of items that converted from K to R(10% - 5% = 5%). After this correction, the number of itemsconverting from R to K responses was still greater than thenumber of items that converted from K to R responses, /(23) =2.36, p < .05. Moreover, the corrected number of itemsconverting from K to R responses was not significantly greaterthan zero, t(23) = 1.69, p > .1. Thus, with the passage of timemany items that initially elicited an R response later elicited aK response. Furthermore, there was not a symmetrical ten-dency for items to move from a K response to an R response.Presumably some items were initially recollected clearly andwith sufficient contextual detail to elicit an R response, butlater some of these items lost supporting detail, became simplyfamiliar, and elicited a K response.

Discussion

In Experiment 3, the same individuals were tested after a10-min delay and again after a 1-week delay. Their perfor-mance was similar to that obtained in Experiments 1 and 2 inwhich different groups were tested at the two delays. Althoughthe individuals in Experiments 1 and 3 were matched in termsof age, their performance was closer to that of the youngerindividuals tested in Experiment 2. Namely, the elevatedfalse-alarm rates at the 1-week delay seen in Experiment 1were not seen in Experiment 3. It is probably the case thatexposure to target items on the first test in Experiment 3enhanced memory for the items when they were seen again onthe second test. Thus, the elevated false-alarm rates in olderindividuals seem to occur only under conditions when memoryis weak.

A substantial proportion of items that initially elicited an Rresponse subsequently elicited a K response during the secondrecognition test. In contrast, a much lower proportion of itemsmoved in the opposite direction: from a K response to an Rresponse. The low rate of conversion from K to R responseswas expected because it is unlikely that an item would elicitsome specific, conscious recollection 1 week after study thatwas not also accessible after 10 min. The frequent conversionof R responses to K responses shows directly that, when R re-

sponses are lost, the information that remains can support Kresponses. These results do not support the view thai the-relationship between R and K responses is exclusive becauseexclusivity predicts little or no conversion from R to Kresponses.

Accordingly, R and K responses must be either redundantor independent. Both redundancy and independence predictthat the conversion of R responses to K responses should begreater than zero and greater than the conversion of Kresponses to R responses. Redundancy predicts the higherrate of R to K conversion. Under redundancy, the loss of Rresponses is caused by the loss of specific episodic memory foritems. The semantic information that remains will elicit a Kresponse, unless semantic information is also lost to such anextent that the item will be missed. Independence predicts alower rate of R to K conversions. Under independence, the Rresponses that are lost can become K responses, they canbecome misses because both context memory and semanticmemory are lost, or they can become misses because semanticmemory has been lost.

In summary, the results of Experiment 3 are inconsistentwith the view that R and K responses are exclusive. Instead,the results are best explained by the view that all R responseshave the potential to become K responses. One possibility isthat R responses are redundant with K responses. A secondpossibility is that R responses occur when two independentprocesses are active. One of these processes supports contex-tual memory, and the second process supports K responses.

General Discussion

The present study shows that both remembering (R) andknowing (K) are impaired in amnesic patients and dependenton the medial temporal lobe-diencephalic structures damagedin amnesia. Accordingly, it would appear that both R and Kresponses depend on declarative (explicit) memory, that is, thekind of memory that is impaired in amnesia. If K responsesreflected the operation of nondeclarative (implicit) memory,amnesic patients should be as accurate as controls at making Kresponses. Yet, in Experiment 1 amnesic patients were im-paired in K response accuracy, just as they were impaired in Rresponse accuracy. Moreover, when declarative memory wasweakened in controls by increasing the delay between studyand test from 10 min to 1 week, the pattern of R and Kresponses exhibited by controls was similar to the patternexhibited by amnesic patients.

One difference between the findings of Experiment 1 andresults obtained previously (Gardiner & Java, 1991) was thatour controls exhibited a higher false-alarm rate at the 1-weekdelay. Experiment 2 showed that this difference probably arosebecause our controls were much older than the undergradu-ates tested in the earlier study. When we tested young adults,the false-alarm rates (and the remainder of the data set) weresimilar to what was previously reported. Accordingly, themethods and the materials that we used to assess R and Kresponses appear to be comparable to the methods andmaterials used by others.

Finally, in Experiment 3 we found that a significant propor-tion of study items that elicited an R response at the 10-min


study-test delay subsequently elicited a K response at the1-week study-test delay. These results show directly that Rresponses have the potential to become K responses, and theysuggest an explanation for the observation that certain experi-mental manipulations in normal individuals can reduce Rresponses without appreciably reducing K responses (Gardiner,1988; Gardiner & Java, 1991; Gardiner & Parkin, 1990).Specifically, these manipulations (i.e., levels of processing,divided attention, and retention interval) do reduce K re-sponses, but the reduction in K responses is largely compen-sated for by the conversion of R responses to K responses. Inthis way, one can obtain the result of a sharp decrease in Rresponses accompanied by no apparent decrease in K re-sponses. The opposite dissociation, a change in K responseswithout a change in R responses, would be difficult to explain ifR responses indeed depend on the same kind of memory as Kresponses. In a recent study, maintenance rehearsal of itemsincreased K responses without significantly affecting R re-sponses (Gardiner, Gawlik, & Richardson-Klavehn, 1994).However, R response probability did decrease a little (4%) asK response probability increased (9%). A decrease in Rresponses in parallel with an increase in K responses can beexplained by supposing that some items that would otherwisehave elicited R responses were only able to support Kresponses.

The distinction between R and K responses was originallyproposed to reflect the operation of episodic and semanticmemory, respectively (Tulving, 1985; see also Tulving, 1989,1993). The finding that both R and K response accuracies wereimpaired in amnesia supports this idea. Episodic and semanticmemory are different forms of declarative memory (Squire,1987; Tulving, 1993). Remember responses measure the recol-lection of item information that is embedded within orassociated with the learning episode, whereas know responsesmeasure context-free item memory.

The finding that items initially eliciting R responses can laterelicit K responses (Experiment 3) also supports the idea that Rand K responses reflect the operation of episodic and semanticmemory. Initially, some items are not only recognized ashaving been presented, which is sufficient for a K response, butthey also evoke aspects of the learning episode, which issufficient for an R response. Later, with the passage of time,memory for specific episodic information can be lost whilememory for the simple fact that an item was presented can beretained. Accordingly, some items that initially elicited an Rresponse later elicited a K response. By this account, theprocess underlying R responses is redundant with the processunderlying K responses. That is, recollection of an item as partof an episode necessarily implies the availability of semanticknowledge that the item itself was previously presented.

An alternative view, that R and K responses arise frommemory components that are independent cannot be ruled outby the data. Accordingly, it is important to note that theindependence view is also compatible with the view developedin this article that R and K responses are two differentexpressions of declarative memory. One must only supposethat the memory for contextual information necessary to elicitan R response can sometimes disconnect from item memory(e.g., one might initially recognize an item, as well as the fact

that it was printed on a creased card, give an R response, andthen later remember that one of the cards was creased but notrecognize the item itself as a study item). It should be notedthat, if this independence view of the memory componentsleading to R and K responses is correct, then R responses donot always reflect episodic memory. Instead, R responses canalso result when two kinds of semantic memory are availableconcurrently (memory for an item and memory for some factabout the item).

We conclude that the memory components underlying Rand K responses, whether redundant or independent, bothdepend on the medial temporal lobe and midline diencephalicstructures damaged in amnesia that are important for declara-tive memory (Squire & Zola-Morgan, 1991; Zola-Morgan &Squire, 1993). In addition, we propose that R responses (andepisodic memory) depend on the integrity of the frontal lobes(also see Tulving, 1989). The latter idea gains support from thefinding that frontal lobe damage can produce source amnesia(Schacter, Harbluk, & McLachlan, 1984; Janowsky, Shi-mamura, & Squire, 1989). Source amnesia refers to loss ofmemory for when and where information was acquired, that is,the learning context. Source memory is essential for making Rresponses (and for episodic memory).

Recent electrophysiological findings also support the ideathat R and K responses depend on declarative memory andfurther that the distinction between R and K responses resultsfrom a post-recollective process (Smith, 1993). In a recognitionmemory test, event-related potentials from study items thatelicited R responses and event-related potentials to studyitems that elicited K responses were similar until about 550 msafter item presentation. Yet, items that were endorsed as studyitems (i.e., all of the items that received either R or Kresponses) could be distinguished from items endorsed as newitems beginning about 400 ms after item presentation. Smith(1993) suggested that both R and K responses result from acommon process of recollection, dependent on the hippocam-pus and related structures, which operates from 400 to 550 msafter item presentation. The distinction between R and Kresponses arises only after 550 ms, when individuals attend tothe products of their retrieval efforts. This second stage ofprocessing determines whether any specific recollections areavailable about the learning episode that would merit attach-ing an R response to the recognized item.

Declarative memory stores semantic information aboutitems (thereby yielding K responses) and also episodic (orcontextual) recollections about specific items (thereby yieldingR responses). Dissociations that are observed between differ-ent measures of memory sometimes reflect the distinctionbetween declarative and nondeclarative memory (e.g., thedifferential effects on recognition memory and repetitionpriming of deep vs. superficial encoding or normal vs. dividedattention). In other cases, these same manipulations candissociate within declarative memory the kind of memory thatdepends on frontal lobe function and the kind that does not.The dissociation between remembering and knowing appearsto be an example of this second kind of distinction.

Performance can be influenced in a number of ways by pastexperience (processing can be speeded, preferences can bealtered, and choices can be biased)—all independently of


declarative memory (Schacter et al., J993; Squire et al., 1993).It has been proposed that, in these circumstances, individualsgain perceptual fluency (i.e., processing efficiency) with thestudy items, which results in an experience of "familiarity"(Jacoby, Kelley, & Dywan, 1989; Mandler, 1980; Whittlesea,1993). If one accepts that K responses are an appropriatemeasure of the experience of familiarity, then the finding thatK response accuracy is impaired in amnesia raises a questionas to whether perceptual fluency ordinarily produces a phenom-enological (conscious) experience of familiarity.

At the same time, these considerations in no way contradictthe general idea that perceptual fluency can sometimes beused as a heuristic such that individuals will judge an item asold, even when that item has not previously been presented.For example, when each test item in a recognition memory testwas preceded by the masked, subthreshold presentation of thetest item, the probability was increased of calling the item"old" whether or not it had previously been presented (Forster,1985; Jacoby & Whitehouse, 1989). This effect is thought tooccur because the test items are perceived more fluently,especially when individuals are asked to make speeded recog-nition judgments. Recently, masked prepresentation of testitems was found to increase the probability of K responses butnot R responses (Rajaram, 1993). It is important to note,however, that the accuracy of K responses was not increased.Individuals simply made more K responses to both target itemsand distractor items. Thus, improved perceptual fluency canapparently make old or new items seem familiar under somecircumstances, and this familiarity can give rise to K responses.Under ordinary recognition memory test conditions, however,perceptual fluency does not appear to provide a useful cue forfamiliarity. Specifically, in our study amnesic patients did notmake K responses as accurately as controls, even though theold items presumably had the benefit of perceptual fluency.Indeed, if it had been tested, the amnesic patients couldpresumably have exhibited intact priming for the old items.

It has previously been noted that influences of perceptualfluency on the feeling of familiarity may be rather uncommon(Whittlesea, 1993, p. 1235). It is also worth mentioning thatamnesic patients give no indication of experiencing feelings offamiliarity out of proportion to the extent of their impaired(declarative) memory. In formal tests, confiderice ratings forrecognition judgments and the recognition judgments them-selves are strictly proportional to the severity of the memoryimpairment (Haist, Shimamura, & Squire, 1992; Shimamura &Squire, 1988).

References

Forster, K. I. (1985). Lexical acquisition and the modular lexicon.Language & Cognitive Processes, 2, 87-108.

Gardiner, J. M. (1988). Recognition failures and free-recall failures:Implications for the relation between recall and recognition. Memory& Cognition, 16, 446-451.

Gardiner, J. M., Gawlik, B., & Richardson-Klavehn, A. (1994).Maintenance rehearsal affects knowing not remembering; elabora-tive rehearsal affects remembering not knowing. Psychonomic Bulle-tin & Review, 1, 107-110.

Gardiner, J. M., & Java, R. I. (1990). Recollective experience in wordand nonword recognition. Memory & Cognition, 18, 23-30.

Gardiner. J. M.. & Java. R. 1. (1491). Forgetting in recognitionmemory with and without recollective experience. Memory A Cogni-tion, 19. (117-623.

Gardiner, J. M., & Java. R. I. (1993). Recognising and rememberine.In A. Collins, S. Gathercole, M. Conway, & P. Morris (Eds).Theories of memory (pp. 163-188). Hillsdale, NJ: Erlbaum.

Gardiner, J. M., & Parkin, A. J. (1990). Attention and recollectiveexperience in recognition memory. Memory & Cognition, IS. 579-583.

Gibson, J. M., & Watkins, M. J. (1988). A pool of 1203 words withunique 2-Ietter fragments. Behavioral Research Method. Instructions.and Computers, 20, 390-397.

Gilbert, J., Levee, R., & Catalano, K. (1968). A preliminary report on anew memory scale. Perceptual and Motor Skills. 27. 277-278.

Haist, F., Shimamura, A. P., & Squire. L. R. (1992). On therelationship between recall and recognition memory. Journal ofExperimental Psychology: Learning, Memoir, and Cognition. IS, 691-702.

Hintzman, D. (1990). Human learning and memory: Connections anddissociations. Annual Review of Psychology, 41, 109-139.

Jacoby, L. L., Kelley, C. M., & Dywan, J. (1989). Memory attributions.In H. L. Roediger & F. I. M. Craik (Eds), Varieties of memory andconsciousness: Essays in honour of Endel Tithing (pp. 391^22).Hillsdale, NJ: Erlbaum.

Jacoby, L. L., & Whitehouse, K. (1989). An illusion of memory: Falserecognition influenced by unconscious perception. Journal of Experi-mental Psychology: General, 118, 126-135.

Janowsky, J. S., Shimamura, A. P., Kritchevsky, M.. & Squire, L. R.(1989). Cognitive impairment following frontal lobe damage and itsrelevance to human amnesia. Behavioral Neuroscience, 103, 548-560.

Janowsky, J. S., Shimamura, A. P., & Squire, L. R. (1989). Sourcememory impairment in patients with frontal lobe lesions. Neuropsy-chologia, 27, 1043-1056.

Jones, G. V. (1987). Independence and exclusivity among psychologi-cal processes: Implications for the structure of recall. PsychologicalReview, 94, 229-235.

Kaplan, E. F., Goodglass, R , & Weintraub. S. (1978). The BostonNaming Test. Boston: E. Kaplan & H. Goodglass.

Kritchevsky, M., Squire, L. R., & Zouzounis, J. A. (1988). Transientglobal amnesia: Characterization of anterograde and retrogradeamnesia. Neurology, 38, 213-219.

Mandler, G. (1980). Recognizing: The judgment of previous occur-rence. Psychological Review, 87, 252-271.

Mattis, S. (1976). Dementia Rating Scale. In R. Bellack & B. Keraso(Eds.), Geriatric psychiatry (pp. 77-121). New York: Grune &Stratton.

Nelson, H. E. (1976). A modified card sorting test sensitive to frontallobe deficits. Cortex, 12, 313-324.

Osterrieth, P. A. (1944). Le test de copie d'une figure complexe [Thetest of copying a complex figure]. Archives de Psychologie, 30,206-356.

Parkin, A. J., & Walter, B. M. (1992). Recollective experience, normalaging, and frontal dysfunction. Psychology and Aging, 7, 290-298.

Polich, J., & Squire, L. R. (1993). P300 from amnesic patients withbilateral hippocampal lesions. EEG Clinical Neuropsychology, 86,408-417.

Rajaram, S. (1993). Remembering and knowing: Two means of accessto the personal past. Memory & Cognition, 21, 89-102.

Rey, A. (1964). L'examen clinique psychologie [The clinical exam inpsychology]. Paris: Presses Universitaires de France.

Richardson-Klavehn, A., & Bjork, R. A. (1988). Measures of memory.Annual Review of Psychology, 39, 475-542.

Schacter, D. L., Chiu, C. Y., & Ochsner, K. N. (1993). Implicitmemory: A selective review. Annual Review of Neuroscience, 16,159-182.


Schacter, D. L., Harbluk, J. L., & McLachlan, D. R. (1984). Retrievalwithout recollection: An experimental analysis of source amnesia.Journal of Verbal Learning and Verbal Behavior, 23, 593-611.

Shimamura, A. P., Jernigan, T. L., & Squire, L. R. (1988). Korsakoff'ssyndrome: Radiological (CT) findings and neuropsychological corre-lates. Journal of Neuroscience, 8, 4400-4410.

Shimamura, A. P., & Squire, L. R. (1987). A neuropsychological studyof fact recall, recognition memory, and confidence ratings. Journal ofExperimental Psychology: Learning, Memory, and Cognition, 14, 763-770.

Shimamura, A. P., & Squire, L. R. (1988). Long-term memory inamnesia: Cued recall, recognition memory, and confidence ratings.Journal of Experimental Psychology: Learning, Memory, and Cognition,14. 763-770.

Smith, M. E. (1993). Neurophysiological manifestations of recollectiveexperience during recognition memory judgments. Journal of Cogni-tive Neuroscience, 5, 1-13.

Squire, L. R. (1987). Memory and brain. New York: Oxford UniversityPress.

Squire, L. R., Amaral, D. G., & Press, G. A. (1990). Magneticresonance measurements of hippocampal formation and mammil-lary nuclei distinguish medial temporal lobe and diencephalicamnesia. Journal of Neuroscience, 10, 3106-3117.

Squire, L. R., Amaral, D. G., Zola-Morgan, S., Kritchevsky, M., &Press, G. A. (1989). Description of brain injury in the amnesicpatient N. A. based on magnetic resonance imaging. ExperimentalNeurology, 105, 23-25.

Squire, L. R., Knowlton, B., & Musen, G. (1993). The structure andorganization of memory. Annual Review of Psychology, 44, 453-495.

Squire, L. R., & Shimamura, A. P. (1986). Characterizing amnesicpatients for neurobehavioral study. Behavioral Neuroscience, 100,866-877.

Squire, L. R., & Zola-Morgan, M. (1991). The medial temporal lobememory system. Science, 253. 1380-1386.

Tulving, E. (1985). Memory and consciousness. Canadian Psychologist,26, 1-12.

Tulving, E. (1989). Remembering and knowing the past. AmericanScientist, 77, 361-367.

Tulving, E. (1993). What is episodic memory? Current Directions inPsychological Science, 2, 67-70.

Tulving, E., & Schacter, D. L. (1990). Priming and human memorysystems. Science, 247, 301-306.

Warrington, E. K. (1984). Recognition memory test. Windsor, England:NFER-Nelson.

Weiskrantz, L. (1988). Some contributions of neuropsychology ofvision and memory to the problem of consciousness. In A. Marcel &E. Bisiach (Eds.), Consciousness and contemporary science (pp.408-417). New York: Oxford University Press.

Wechsler, D. (1981). Wechsler Adult Intelligence Scale—Revised[manual]. New York: Psychological Corporation.

Wechsler, D. (1987). Wechsler Memory Scale—Rexised [manual]. NewYork: Psychological Corporation.

Whittlesea, B. W. A. (1993). Illusions of familiarity. Journal ofExperimental Psychology: Learning, Memory, and Cognition, 19, 1235—1253.

Zola-Morgan, S., & Squire, L. R. (1993). Neuroanatomy of memory.Annual Review of Neuroscience, 16, 547-563.

Received December 6, 1993Revision received July 8, 1994

Accepted July 18, 1994

Remembering and knowing: Two different expressions of declarative memory

Documents

Transcript of Remembering and knowing: Two different expressions of declarative memory