True but not false memories produce a sensory signature in human lateralized brain potentials

True But Not False Memories Produce aSensory Signature in Human Lateralized BrainPotentials

Monica Fabiani Michael A Stadler and Peter M WesselsUniversity of Missouri

Abstract

amp False memories (eg recognition of events that did notoccur) are considered behaviorally and subjectively indistin-guishable from true memories We report that brain activitydiffers when true and false memories are retrieved Stronglyassociated lists of words were presented to one or the othercerebral hemisphere at study This led to lateralized brainactivity for these words during a centrally presented recogni-tion test reflecting their lateralized encoding This activity wasabsent for nonstudied but strongly associated words falsely

recognized as studied items These results indicate that studiedwords leave sensory signatures of study experiences that areabsent for false memories In addition hemifield effectsemerged including a slower reaction time (RT) for falserecognition of nonstudied words whose associated lists werepresented to the left hemifield (ie right hemisphere) Thesefalse recognition responses were accompanied by frontal slowwave activity which may reflect a differential ability of the twohemispheres with respect to semantic processing amp

INTRODUCTION

False memories occur when we believe we rememberevents that did not actually happen (Roediger 1996)They are one of many types of illusion that can give usclues about the nature of cognitive processes (Rama-chandran 1998) Information about the genesis of falsememories may also provide insights into lsquolsquorecovered-memoryrsquorsquo phenomena and therefore contribute to thecurrent debate about the validity of such memories (egShobe amp Kihlstrom 1997) In this article we demon-strate the existence of differential brain activity elicitedby false and true recognition in a laboratory setting inwhich these memories are typically indistinguishable onthe basis of behavioral and subjective indices (StadlerRoediger amp McDermott 1999 Duzel Yonelinas Man-gun Heinze amp Tulving 1997 Gallo Roberts amp Seamon1997 Johnson et al 1997 Payne Elie Blackwell ampNeuschatz 1996 Payne Neuschatz Lampinen amp Lynn1997 Schacter 1996a Roediger amp McDermott 1995)

Recent studies have used a simple laboratory techni-que for producing false memories (Deese Roediger ampMcDermottrsquos paradigm or DRM Duzel et al 1997Johnson et al 1997 Schacter Verfaellie amp Pradere1996b Roediger amp McDermott 1995 see also Deese1959) Subjects are presented with a list of words allassociatively related to a nonpresented critical targetFor example the words DOOR GLASS PANE SHADELEDGE SILL HOUSE OPEN CURTAIN FRAME VIEWBREEZE SASH SCREEN SHUTTER are all related to

WINDOW When subjects are asked to recognize wordsfrom such a list on a later test they erroneously recog-nize the nonpresented false target (WINDOW) as havingoccurred on the study list

From a behavioral standpoint it has proven difficult todiscriminate between veridical and false memories usingthis paradigm In fact the probability of false recognitionof the critical targets is approximately the same as theprobability of correctly recognizing words that wereactually on the list (Stadler et al 1999 Payne et al1996 Roediger amp McDermott 1995) Subjects areequally confident in making true and false recognitionjudgments (see Miller amp Wolford 1999 and Roediger ampMcDermott 1999 for a discussion) and typically reportsimilar conscious experiences in the two cases (Payneet al 1997 Roediger amp McDermott 1995 but seeMather Henkel amp Johnson 1997 and Norman amp Schac-ter 1997 for evidence of differences in specific aspects ofconscious experience for true vs false recognition)Even when instructed about the likelihood of makingfalse recognition errors subjects have difficulty suppres-sing them (McDermott amp Roediger 1998 Gallo et al1997 Roediger amp McDermott 1995 but see SchacterIsrael amp Racine 1999)

Because of the impressive behavioral and subjectivesimilarities between true and false memories research-ers have attempted to find brain indices that woulddiscriminate between them with the assumption thatmemory retrieval involves in part reactivation of sen-

copy 2000 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 126 pp 941ndash949

sory information that was present during processing ofthe original event Since false memories involve eventsthat did not occur there can be no reactivation ofsensory experience hence there should be brain activity(ie sensory signatures) that differentiates true fromfalse memories in regions of the brain associated withsensory experience

Schacter et al (1996a) used positron emission to-mography (PET) with a blocked design (ie withblocks of test trials consisting of words all in the sameexperimental condition) and reported increased neuralactivity in the left temporo-parietal region for truememories but not for false memories However thesefindings were not replicated in subsequent studieswith randomized word presentation in which brainactivity was recorded by means of functional magneticresonance imaging (fMRI Schacter Buckner Kout-staal Dale amp Rosen 1997) and event-related brainpotentials (ERPs Duzel et al 1997 Johnson et al1997) A problem with these studies is that sensorysignatures like the original sensation may only beseen in transient brain activity and may therefore notbe readily visible in the slower hemodynamic responsemeasures obtained with fMRI and PET These signa-tures may also correspond to a relatively small portionof the neural processing of verbal material Thereforespecial controls may be required to make them moreapparent even when measures with higher temporalresolution such as ERPs are used

Using nonverbalizable stimuli Gratton Corballis andJain (1997) showed that the ERPs elicited by correctlyrecognized line patterns during test were systematicallylateralized according to their side of presentation duringstudy Although recognition performance was largelydependent on whether the study and test hemispherematched subjects did not have a conscious recollectionof the side at which the item had been studied Thissuggests that the lateralized brain activity was related tothe operation of a hemispherically organized sensorymemory whose functioning may be independent of thesubjectrsquos conscious control (see also Fabiani Gratton ampHo submitted Gratton Fabiani Goodman-Wood ampDeSoto 1998)

This contralateral-control procedure may be particu-larly useful for demonstrating the presence of sensorysignatures in the false memory paradigm In fact thelateralization effect visible at retrieval is related to aphysical property of the stimulus at encoding (ie thehemifield of presentation) which can only exist if thestimulus was actually presented However in theGratton et al (1997) study the lateralization effectwas elicited using nonverbal stimuli It was an openquestion whether a similar phenomenon would alsooccur in a paradigm in which highly associated wordsare used

In this experiment subjects studied words from asso-ciated lists (Stadler et al 1999) presented randomly to

the left or right of fixation (with the constraint thatwords from the same associative list were all displayed inthe same hemifield see Figure 1A) The study phase wasfollowed by a test phase in which words were presentedat the center of the screen and ERPs elicited by eachword were recorded (Figure 1B) During test subjectswere asked to indicate via a choice button-presswhether each word was lsquolsquooldrsquorsquo (ie studied) or lsquolsquonewrsquorsquo(recognition task) Studied words (ie true targets)were randomly intermixed with nonpresented buthighly associated words (ie false targets) and othernonpresented words from similar but nonstudied lists(control words) We predicted that true target (studied)words would leave sensory signatures on the ERPselicited at test which would be lateralized consistentlywith the side of presentation at study (Gratton et al1997) Brain responses to related but nonpresented

+

+

+

+

+

table

sit

nap

desk

snore

C ha ir L is t S lee p L is t

S tu dy P h as eA)

B)

+

+

+

+

door

nap

window

chair

T RU E TA R G E T C O N T R O LL is t N o t S tu d ied

T RU E TA R G E TL is t S tu d ied R ig h t

FA L S E TA R G E T C O N T R O LL is t N o t S tu d ied

FA L S E TA R G E TL is t S tu d ied Le ft

ldquo N E W rdquo

ldquoO L D rdquo

ldquoN EW rdquo

ldquoN EW rdquo

Condition

CorrectResponse

Tes t P h ase

Figure 1 (A) Schematic representation of the study phase (B)Schematic representation of the test phase

942 Journal of Cognitive Neuroscience Volume 12 Number 6

words (false targets) that are falsely recognized shouldnot be lateralized thus distinguishing true from falserecognition

RESULTS

Reaction Time and Accuracy

The behavioral results (Figure 2A and B) showed arobust false memory effect The percentage of lsquolsquooldrsquorsquoresponses correctly given to true targets did not differfrom that erroneously given to false targets (t(13) lt 1)whereas the rate of old responses erroneously given tocontrol words was significantly lower than those given tofalse targets (t(13) = 967 p lt 001) Reaction times(RTs) were longer for control words that were erro-neously classified as old (896 msec SD = 137) than forthe other two groups given lsquolsquooldrsquorsquo responses (752 msecSD = 111 for true targets t(13) = 545 p lt 001 and782 msec SD = 115 for false targets t(13) = 431 p lt001) The prolonged RTs to false targets with respect totrue targets (t(13) = 231 p lt 05) were due to the RTsto false targets whose associated lists were presented tothe left hemifield (and were therefore first processed bythe right hemisphere) This finding suggests that theremay be a difference in the processing of the false targetsdepending on the hemisphere to which associatedwords were first presented

ERPs

The main hypothesis of this study concerned the occur-rence of encoding-related lateralizations in response totrue but not false targets which was tested using adouble-subtraction method (Gratton 1998 see alsoMethods)

As predicted the ERPs recorded at test showedlateralized brain activity for the true targets (t(13) =243 p lt 05 two-tailed central and posterior electrodesites 210ndash700 msec time window) but not for the falsetargets (t(13) = ndash 021 ns) This activity can be visua-lized in both the original data (Figure 3A) and in the dataderived by means of contralateral-control subtractions(Figures 3B and 4)

The interaction between these two effects (ie thedifference between the lateralizations for true andfalse targets) was tested with a nonparametric signtest because of the difference in variability for thetwo groups of trials due to the difference in thenumber of trials included in the averages This inter-action was significant for the early part of the wave-form (210ndash400 msec p lt 05) and only marginallysignificant when the entire interval (210ndash700 msec)was considered (p lt 10) Note that differentiallateralized activity cannot be due to stimulus orresponse requirements at test because all stimuliwere words presented at fixation and all were judgedto be lsquolsquooldrsquorsquo by the subjects Therefore this lateralizedactivity can be interpreted as an expression of thememory trace formed at study The absence of thisactivity for the false targets correctly indicates thattheir memory traces did not include a sensory signa-ture

Because the paradigm involved repeated studyndashtestblocks it could be hypothesized that exposure to re-cognition test trials could alter subsequent encodingand help the subjects develop strategies that may lead tothe critical results If this were the case one wouldexpect to see larger lateralization effects for the studiedwords in the second-half of the experiment than in thefirst In fact there was no significant difference betweenthe first- and second-half of the study (t lt 12 for bothlatency ranges)

Figure 3A shows that in addition to a difference inlateralization there were also differences in the ampli-tude of the ERP response to true and false targets Thesedifferences can be seen more clearly in Figure 5 whichshows true targets false targets and new trials at all therecording sites From this figure it is evident that thepositivity elicited by false targets was reduced in ampli-tude with respect to the activity elicited by true targets(midline electrode sites F(113) = 486 p lt 05)However when this effect was examined more closelyit was apparent that the reduced positivity was due tothose false targets whose associated lists were firstprocessed by the right hemisphere (F(226) = 523p lt 05) as can be seen more clearly in Figure 6 Thisfinding possibly suggests that the semantic processingunderlying the DRM effect may be carried out predomi-nantly by the left hemisphere (Metcalfe Funnell ampGazzaniga 1995 Phelps amp Gazzaniga 1992) and there-fore may be more readily elicited by words presented inthe right visual field

Figure 2 (A) A false memory effect is evident in the percentage oflsquolsquooldrsquorsquo responses given to true targets (white bar 7260 SD = 793)and false targets (hatched bar 7119 SD = 1405) and is coupledwith a low overall false alarm rate to other nonpresented control words(black bar 2007 SD = 1296) (B) Mean RT for each response type

Fabiani Stadler and Wessels 943

Results obtained with PET (Schacter et al 1996a) andfMRI (Schacter et al 1997) have pointed to the anteriorprefrontal cortex as an area involved in true and false

recognition Data obtained in the present study suggestthat there may be a difference in both RT and ERPactivity for false targets whose associated lists were first

Figure 4 Grand average fieldmap of the memory-relatedlateralizations for true targets(left) and false targets (right)obtained with the contralateral-control procedure at a latencyof 500ndash600 msec after stimulusOnly one hemisphere is shownbecause given the analyticprocedure the two hemi-spheres will display effects thatare identical in magnitude butof opposite sign

Figure 3 (A) Grand averageERPs at test are displayed atelectrodes T5 (left hemisphere)and T6 (right hemisphere) fortrue targets (top row) and falsetargets (bottom row) for listsstudied on the right hemifield(thin line LH = left hemi-sphere) and left hemifield(thick line RH = right hemi-sphere) Amplitude in micro-volts is on the ordinate andtime in milliseconds is on theabscissa Time lsquolsquo0rsquorsquo indicatesstimulus onset The lateraliza-tion for true targets is indicatedby the fact that the ERP ampli-tude in response to wordsstudied on the right (left hemi-sphere) is larger at electrodeT5 whereas the opposite is truefor words studied on the left(right hemisphere) (B) Grandaverage of the lateralizationwaveforms across middle andposterior electrodes Theshaded area indicates the peri-od in which there is a significantlateralization for the true targets(thick line) No lateralization isevident for false targets (thinline)


presented to the left hemifield (ie right hemisphere)Therefore we tested whether true and false targets fromlists studied in the left or right visual fields differed withrespect to the activity they elicited at prefrontal elec-

trode sites (Fp1 and Fp2) at test These results areshown in Figure 7 In a latency window between 700and 1500 msec the false targets associated with listspresented to the left hemifield (right hemisphere) at

Figure 6 Grand average at thePz electrode for test ERPs eli-cited by true targets studied tothe left and right of fixation andby false targets whose asso-ciated lists were studied to theleft or right of fixation Ampli-tude in microvolts is on theordinate and time in millise-conds is on the abscissa Timelsquolsquo0rsquorsquo indicates stimulus onset

Figure 5 Grand average ofERPs at test for new truetargets and false targets at allelectrode sites used in thestudy Amplitude in microvoltsis on the ordinate and time inmilliseconds is on the abscissaTime lsquolsquo0rsquorsquo indicates stimulusonset


study differed significantly from the other true and falsetargets (F(113) = 654 p lt 05)

DISCUSSION

This study demonstrates that with contralateral-controlprocedures it is possible to isolate brain activity repre-senting sensory signatures associated with memory forverbal material We cannot yet localize the area withinthe brain that generates these signatures but the in-volvement of the ventral stream of visual processing issuggested by the position of the electrodes showing thelateralization effects by source localization algorithmsapplied to these data and by converging evidence fromoptical imaging recordings from the visual cortex usingthe same contralateral-control procedure in a differentparadigm (Fabiani et al submitted Gratton et al 1998)At this time we also cannot identify the exact nature ofthe processing that generates these signatures Perhapsfurther research employing different orienting tasks thatmay encourage or discourage the use of semantic ororthographic information could distinguish these possi-bilities

Recently research has shown that making studieditems more distinctive helps subjects avoid false memoryerrors (Schacter et al 1999) Memory traces are as-sumed to be composed of a number of features includ-ing information about the memoryrsquos source that is thecontext in which it was formed (Johnson Hashtroudi ampLindsay 1993) These features if consciously accessiblemay lead to the use of strategies or heuristics that allow

subjects to better distinguish false from true memories(Schacter et al 1999 Johnson et al 1997 Mather et al1997) The effects of the lateralization manipulationused here may be interpreted in terms of distinctivenessTrue memories may leave a sensory signature thatmakes each memory trace distinctive false memoriespresumably lack such a distinctive feature Of course inthis study the distinctive sensory feature did not influ-ence subjectsrsquo recognition judgments probably becausethis feature was not consciously accessible (Grattonet al 1997 Mather et al 1997)

Additional effects were observed in this study sug-gesting that false targets whose associates were firstpresented to the right hemisphere differed from truetargets and from false targets associated with wordsstudied on the right (left hemisphere) Namely falserecognition of these words was accompanied by longerRT smaller P300 amplitude at posterior sites (latency400ndash700 msec) and larger positive slow wave activity atfrontal sites later in the waveform (700ndash1500 msec)This latter finding is consistent with results obtained byWilding and Rugg (1996 see also Trott FriedmanRitter Fabiani amp Snodgrass 1999) showing sustainedprefrontal activity in source-monitoring retrieval tasksIt also converges with ERP results suggesting thatfrontal positive slow waves may be associated withelaboration (Fabiani Karis amp Donchin 1990 KarisFabiani amp Donchin 1984) and with the idea that theleft hemisphere may act as an lsquolsquointerpreterrsquorsquo (Phelps ampGazzaniga 1992) Finally although it is not possible todraw direct inferences between the location of scalp

Figure 7 Grand average atfrontal sites (Fp1 and Fp2) fortest ERPs elicited by true targetsstudied to the left and right offixation and by false targetswhose associated lists werestudied to the left or right offixation Amplitude in micro-volts is on the ordinate andtime in milliseconds is on theabscissa Time lsquolsquo0rsquorsquo indicatesstimulus onset


electrodes and the underlying brain structures taken atface value these results are also consistent with theinvolvement of the prefrontal cortex in the DRM task asreported by previous studies (Schacter 1996a Schacteret al 1997)

These differential effects between true and false tar-gets were not observed in the Johnson et al (1997) ERPstudy of false memory However in that study wordswere encoded and tested at fixation therefore it ispossible that the results reflected predominantly theinfluence of left-hemisphere processing In fact in thepresent experiment false targets whose associated listswere presented to the right hemifield (left hemisphere)did not differ from true targets either behaviorally orwith respect to the ERP activity they elicited Presum-ably it was our addition of the contralateral-controlprocedure that accounts for the differences in findingsbetween the two studies

Note that the lateralization effects are visible for truerecognition of words encoded at both the left and rightof fixation However the differences in P300 and slowwave activity mostly distinguish the false targets asso-ciated with words processed by the right hemispherefrom the other words Thus these may represent func-tionally distinct phenomena the first of which (laterali-zation) may represent an actual sensory signaturewhereas the second (P300 and slow wave effects) mayreflect the interaction between the encoding manipula-tion and the hemispheric specialization for semanticprocessing

In conclusion we have demonstrated differences inbrain activity at the time of retrieval for true and falsememories Studied words elicited a sensory signaturewhereas words erroneously identified as part of thestudied set did not Thus these results suggest thateven when subjects cannot consciously discriminatebetween true and false memories their brain activitymay reveal information relevant to this discrimination

These results were obtained using a contralateral-control procedure (Gratton 1998) which may be idealfor revealing transient differences related to a smallportion of the processing of verbal material They alsoconfirmed previous evidence that visual memorytraces are hemispherically organized even in the caseof verbal stimuli (Gratton et al 1997 1998) and thatthe left hemisphere may have a prominent role in thesemantic processes that accompany false memoryeffects

METHODS

Subjects

Fourteen right-handed adults (10 females age range 20ndash28) with normal or corrected-to-normal vision signedinformed consent and were paid for their participationin this study Some additional subjects whose perfor-

mance in the recognition task was at chance were notincluded in further analyses

Stimuli and Procedures

The stimuli were 36 15-word associative lists (Stadleret al 1999) Subjects were presented with six studyndashtestblocks During each study phase words from four listswere displayed in random order for 200 msec each 158to either the left or the right of a fixation cross with a1500-msec interstimulus interval (ISI) and with theconstraint that words from the same associative list wereall presented in the same hemifield

The 200-msec stimulus duration used during thestudy was chosen to prevent subjects from performingsaccadic eye movements toward the laterally presentedstimuli Even if subjects were to move their eyes towarda stimulus this should only add to the experimentalnoise and decrease the probability of observing memory-related lateralizations rather than bias the results

Each study list was repeated twice to improve wordreading with the same words presented in the samehemifield on both occasions Subjects were instructed tofixate on the central cross During every test phase 24words were displayed in random order for 200 mseceach with a 2000-msec ISI between words Twelve ofthe test words were true targets (ie words that wereactually studied from positions 1 8 and 10 of theassociative lists) and were randomly intermixed withfour false targets (ie the nonpresented lures associatedto the studied words) six true target controls (wordsfrom positions 1 8 and 10 of the 12 lists that were notstudied) and two false target controls (ie the criticallures associated with the nonstudied lists) Subjectswere asked to indicate via button presses whether ornot each word was part of the studied set The responseassignments were counterbalanced across subjects

ERP Recording and Analysis

ERPs elicited by test words were recorded from the full10ndash20 system montage (19 scalp electrodes) by means ofan electrode cap (ElectroCap International) The leftmastoid was used as reference and an average mastoidreference was computed off-line The data were digitizedat 100 Hz and were filtered on-line using a 001ndash30-Hzband-pass Vertical and horizontal eye movements wererecorded and corrected off-line (Gratton Coles ampDonchin 1983) Each recording epoch started 100 msecbefore stimulus presentation and lasted 1600 msec

ERP data recorded at test were averaged separately foreach subject electrode and experimental conditionAverage data files were digitally filtered with a band-passof 0ndash15 Hz before measurement Mean amplitude mea-sures were derived for each subject condition andelectrode for two time windows (210ndash700 msec and710ndash1500 poststimulus) These extended measurement


time windows encompassing several hundred millise-conds of recording were chosen to avoid use of multi-ple comparisons

The choice of the specific interval encompassed bythe window used for the lateralization analysis wasbased on two criteria The first criterion was the apriori knowledge derived from previous studies Grat-ton et al (1997) found sustained lateralization effectswith a peak latency of approximately 450 msecwhereas optical imaging studies (Fabiani et al sub-mitted Gratton et al 1998) showed lateralizationeffects at earlier latencies in the visual cortex Thesecond criterion was the component structure of theaverage waveform in this study which is typical ofrecognition paradigms and includes a P200 with apeak latency of approximately 250 msec and a P300with a peak latency exceeding 500 msec In studiesexamining priming effects versus aspects of the recol-lective experience (eg Paller amp Kutas 1992) primingeffects are evident at short latencies whereas recollec-tive processes are often riding on the P300 or otherlate components (eg Fabiani et al 1986 1990) Thislatter distinction was the basis for examining the inter-action of the lateralization effects for two separateintervals (200ndash400 and later)

The Contralateral-Control Method

The contralateral-control method makes it possible toisolate brain activity that systematically occurs in one orthe other hemisphere depending on the experimentalcondition (in this case presentation of the word to theleft or right hemifield during study Gratton 1998) Thecomputation of lateralized waveforms is achieved bymeans of a two-step procedure First the ERPs recordedat homologous electrode locations over the left and righthemispheres are subtracted from each other with theactivity from the electrode ipsilateral to the manipulation(ie ipsilateral to the hemifield of word presentationduring study) subtracted from the activity recorded atthe contralateral electrode This step eliminates activitythat occurs symmetrically in both hemispheres Thesecond step is to average the lateralization waveformsobtained for the left- and right-hemisphere conditionsThis step eliminates the influence of electrical activitythat is not lateralized or whose lateralization is indepen-dent of the experimental manipulation

Acknowledgments

This work was supported by a McDonnell-Pew grant to MonicaFabiani and Michael A Stadler We thank Nelson CowanGabriele Gratton Steve Hackley and Jonathan King forcomments on an earlier version of this manuscript

Reprint requests should be sent to Monica Fabiani Universityof Missouri Department of Psychology 210 McAlester HallColumbia MO 65211 USA e-mail fabianimmissouriedu

REFERENCES

Deese J (1959) On the prediction of occurrence of particularverbal intrusions in immediate recall Journal of Experi-mental Psychology 58 17ndash22

Duzel E Yonelinas A P Mangun G R Heinze H J ampTulving E (1997) Event-related brain potential correlates oftwo states of conscious awareness in memory Proceedingsof the National Academy of Sciences USA 94 5973ndash5978

Fabiani M Gratton G amp Ho J (submitted) Multiple visualmemory phenomena in a memory search task

Fabiani M Karis D amp Donchin E (1986) P300 and recall inan incidental memory paradigm Psychophysiology 23 298ndash308

Fabiani M Karis D amp Donchin E (1990) Effects of mnemonicstrategy manipulation in a Von Restorff paradigm Electroen-cephalography and Clinical Neurophysiology 75 22ndash35

Gallo D A Roberts M J amp Seamon J G (1997) Remember-ing words not presented in lists Can we avoid creating falsememories Psychonomic Bulletin and Review 4 271ndash276

Gratton G (1998) The contralateral organization of visualmemory A theoretical concept and a research tool Psy-chophysiology 35 638ndash647

Gratton G Coles M G amp Donchin E (1983) A new methodfor off-line removal of ocular artifact Electroencephalogra-phy and Clinical Neurophysiology 55 468ndash484

Gratton G Corballis P M amp Jain S (1997) Hemisphericorganization of visual memories Journal of Cognitive Neu-roscience 9 92ndash104

Gratton G Fabiani M Goodman-Wood M R amp DeSoto MC (1998) Memory-driven processing in human medial oc-cipital cortex An event-related optical signal (EROS) studyPsychophysiology 35 348ndash351

Johnson M K Hashtroudi S amp Lindsay D S (1993) Sourcemonitoring Psycholological Bulletin 114 3ndash28

Johnson M K Nolde S F Mather M Kounios J SchacterD L amp Curran T (1997) The similarity of brain activityassociated with true and false recognition memory dependson test format Psychological Science 8 250ndash257

Karis D Fabiani M amp Donchin E (1984) lsquolsquoP300rsquorsquo andmemory Individual differences in the von Restorff effectCognitive Psychology 16 177ndash216

Mather M Henkel L A amp Johnson M K (1997) Evaluatingcharacteristics of false memories Rememberknow judg-ments and memory characteristics questionnaire comparedMemory and Cognition 25 826ndash837

McDermott K B amp Roediger H L III (1998) Attempting toavoid illusory memories Robust false recognition of associ-ates persists under conditions of explicit warnings and im-mediate testing Journal of Memory and Language 39 508ndash520

Metcalfe J Funnell M amp Gazzaniga M S (1995) Right-hemisphere memory superiority Studies of a split-brain pa-tient Psychological Science 6 157ndash164

Miller M B amp Wolford G L (1999) Theoretical commentaryThe role of criterion shift in false memory PsychologicalReview 106 398ndash405

Norman K A amp Schacter D L (1997) False recognition inyounger and older adults Exploring the characteristics ofillusory memories Memory and Cognition 25 838ndash848

Paller K A amp Kutas M (1992) Brain potentials duringmemory retrieval provide neurophysiological support forthe distinction between conscious recollection and primingJournal of Cognitive Neuroscience 4 375ndash391

Payne D G Elie C J Blackwell J M amp Neuschatz J S(1996) Memory illusions Recalling recognizing and recol-lecting events that never occurred Journal of Memory andLanguage 35 261ndash285


Payne D G Neuschatz J S Lampinen J M amp Lynn S J(1997) Compelling memory illusions The qualitative char-acteristics of false memories Current Directions in Psy-chological Science 6 56ndash60

Phelps E A amp Gazzaniga M S (1992) Hemispheric differ-ences in mnemonic processing The effects of left hemi-sphere interpretation Neuropsychologia 30 293ndash297

Ramachandran V S (1998) Consciousness and body imageLessons from phantom limbs Capgras syndrome and painasymbolia Philosophical Transactions of the Royal Societyof London Series B Biological Sciences 353 1851ndash1859

Roediger H L III (1996) Memory illusions Journal of Mem-ory and Language 35 76ndash100

Roediger H L III amp McDermott K B (1995) Creating falsememories Remembering words not presented in listsJournal of Experimental Psychology Learning Memoryand Cognition 21 803ndash814

Roediger H L III amp McDermott K B (1999) False alarms andfalse memories Psychological Review 106 406ndash410

Schacter D L (1996) Illusory memories A cognitive neu-roscience analysis Proceedings of the National Academy ofSciences USA 93 13527ndash13533

Schacter D L Buckner R L Koutstaal W Dale A M ampRosen B R (1997) Late onset of anterior prefrontal activityduring true and false recognition an event-related fMRIstudy Neuroimage 6 259ndash269

Schacter D L Israel L amp Racine C (1999) Suppressing falserecognition in younger and older adults The distinctivenessheuristic Journal of Memory and Language 40 1ndash24

Schacter D L Reiman E Curran T Yun L S Bandy DMcDermott K B amp Roediger H L III (1996a) Neuroana-tomical correlates of veridical and illusory recognitionmemory Evidence from positron emission tomographyNeuron 17 267ndash274

Schacter D L Verfaellie M amp Pradere D (1996b) Theneuropsychology of memory illusions False recall and re-cognition in amnesic patients Journal of Memory andLanguage 35 319ndash334

Shobe K K amp Kihlstrom J F (1997) Is traumatic memoryspecial Current Directions in Psychological Science 6 70ndash74

Stadler M A Roediger H L III amp McDermott K B (1999)Norms for word lists that create false memories Memoryand Cognition 27 494ndash500

Trott C T Friedman D Ritter W Fabiani M amp SnodgrassJ G (1999) Episodic priming and memory for temporalsource Event-related potentials reveal age-related differ-ences in prefrontal functioning Psychology and Aging 14390ndash413

Wilding E L amp Rugg M D (1996) An event-related potentialstudy of recognition memory with and without retrieval ofsource Brain 119 889ndash905


sory information that was present during processing ofthe original event Since false memories involve eventsthat did not occur there can be no reactivation ofsensory experience hence there should be brain activity(ie sensory signatures) that differentiates true fromfalse memories in regions of the brain associated withsensory experience

Schacter et al (1996a) used positron emission to-mography (PET) with a blocked design (ie withblocks of test trials consisting of words all in the sameexperimental condition) and reported increased neuralactivity in the left temporo-parietal region for truememories but not for false memories However thesefindings were not replicated in subsequent studieswith randomized word presentation in which brainactivity was recorded by means of functional magneticresonance imaging (fMRI Schacter Buckner Kout-staal Dale amp Rosen 1997) and event-related brainpotentials (ERPs Duzel et al 1997 Johnson et al1997) A problem with these studies is that sensorysignatures like the original sensation may only beseen in transient brain activity and may therefore notbe readily visible in the slower hemodynamic responsemeasures obtained with fMRI and PET These signa-tures may also correspond to a relatively small portionof the neural processing of verbal material Thereforespecial controls may be required to make them moreapparent even when measures with higher temporalresolution such as ERPs are used

Using nonverbalizable stimuli Gratton Corballis andJain (1997) showed that the ERPs elicited by correctlyrecognized line patterns during test were systematicallylateralized according to their side of presentation duringstudy Although recognition performance was largelydependent on whether the study and test hemispherematched subjects did not have a conscious recollectionof the side at which the item had been studied Thissuggests that the lateralized brain activity was related tothe operation of a hemispherically organized sensorymemory whose functioning may be independent of thesubjectrsquos conscious control (see also Fabiani Gratton ampHo submitted Gratton Fabiani Goodman-Wood ampDeSoto 1998)

This contralateral-control procedure may be particu-larly useful for demonstrating the presence of sensorysignatures in the false memory paradigm In fact thelateralization effect visible at retrieval is related to aphysical property of the stimulus at encoding (ie thehemifield of presentation) which can only exist if thestimulus was actually presented However in theGratton et al (1997) study the lateralization effectwas elicited using nonverbal stimuli It was an openquestion whether a similar phenomenon would alsooccur in a paradigm in which highly associated wordsare used

In this experiment subjects studied words from asso-ciated lists (Stadler et al 1999) presented randomly to

the left or right of fixation (with the constraint thatwords from the same associative list were all displayed inthe same hemifield see Figure 1A) The study phase wasfollowed by a test phase in which words were presentedat the center of the screen and ERPs elicited by eachword were recorded (Figure 1B) During test subjectswere asked to indicate via a choice button-presswhether each word was lsquolsquooldrsquorsquo (ie studied) or lsquolsquonewrsquorsquo(recognition task) Studied words (ie true targets)were randomly intermixed with nonpresented buthighly associated words (ie false targets) and othernonpresented words from similar but nonstudied lists(control words) We predicted that true target (studied)words would leave sensory signatures on the ERPselicited at test which would be lateralized consistentlywith the side of presentation at study (Gratton et al1997) Brain responses to related but nonpresented

+

+

+

+

+

table

sit

nap

desk

snore

C ha ir L is t S lee p L is t

S tu dy P h as eA)

B)

+

+

+

+

door

nap

window

chair

T RU E TA R G E T C O N T R O LL is t N o t S tu d ied

T RU E TA R G E TL is t S tu d ied R ig h t

FA L S E TA R G E T C O N T R O LL is t N o t S tu d ied

FA L S E TA R G E TL is t S tu d ied Le ft

ldquo N E W rdquo

ldquoO L D rdquo

ldquoN EW rdquo

ldquoN EW rdquo

Condition

CorrectResponse

Tes t P h ase

Figure 1 (A) Schematic representation of the study phase (B)Schematic representation of the test phase



RESULTS



ERPs



















DISCUSSION












METHODS

Subjects















Acknowledgments



REFERENCES







































RESULTS



ERPs



















DISCUSSION












METHODS

Subjects















Acknowledgments



REFERENCES

















































DISCUSSION












METHODS

Subjects















Acknowledgments



REFERENCES












































DISCUSSION












METHODS

Subjects















Acknowledgments



REFERENCES







































DISCUSSION












METHODS

Subjects















Acknowledgments



REFERENCES











































METHODS

Subjects















Acknowledgments



REFERENCES










































Acknowledgments



REFERENCES






















































True but not false memories produce a sensory signature in human lateralized brain potentials

Documents