PERCEPTUALLY BASED FALSE RECOGNITION OF NOVEL … · 2015-09-10 · also prompted investigations of such errors in persons with impaired memory (for review, see ... level of orthographic

PERCEPTUALLY BASED FALSE RECOGNITION OFNOVEL OBJECTS IN AMNESIA EFFECTS OF

CATEGORY SIZE AND SIMILARITY TO CATEGORYPROTOTY PES

Wilma Koutstaal and Daniel L SchacterHarvard University Cambridge USA

Mieke VerfaellieVeterans Affairs Medical Center and Boston University School of Medicine Boston USA

Carolyn Brenner and Eric M JacksonHarvard University Cambridge USA

Previous research suggests that amnesics may show impaired semantically based false recognition underconditions where control participants show high levels of gist-based errors but little or no impairmentwhen controls show less robust false recognition Using abstract novel objects we examined perceptu-ally based false recognition in amnesics under conditions designed to induce differing levels of false rec-ognition in controls Whereas amnesics showed significantly impaired false recognition for categoryprototypes and numerically impaired false recognition when many perceptually similar exemplars werestudiedmdashconditions where controls showed high rates of illusory recognitionmdashamnesics and controlsshowed lower and comparable levels of false recognition when few related exemplars were studied orlures were at a far transformational distance from the prototype Although amnesics may be able to ex-tract some information regarding the perceptual ldquogistrdquo of studied items they appear to do so less effi-ciently than controls

INTRODUCTION

The strategy of using illusory recollection as ameans to probe the nature of memory encoding andretrieval processes has a long and diverse history inpsychology (for general review and discussion seeEstes 1997 Johnson Hashtroudi amp Lindsay1993 McClelland 1995 Roediger 1996Schacter 1995 Schacter Norman amp Koutstaal1998a) Although many studies have focused on

participants with intact memory the uncovering ofrobust forms of false recollection and the effort tounderstand the bases of illusory recollection havealso prompted investigations of such errors inpersons with impaired memory (for review seeSchacter 1997 Schacter et al 1998a)

Several studies have focused on amnesicpatients Individuals who as a consequence of dam-age to medial temporal and diencephalic brainregions arising from a variety of possible causes (eg

COGNITIVE NEUROPSYCHOLOGY 1999 16 (345) 317ndash341

Oacute 1999 Psychology Press Ltd 317

Requests for reprints should be addressed to Wilma Koutstaal Harvard University Department of Psychology 33 KirklandStreet Cambridge MA 02138 USA (Tel (617) 495-3856 Fax (617) 496-3122 E-mail wkwjhharvardedu)

This research was supported by National Institute on Aging grant AG08441 the Human Frontiers Science Program andNational Institutes of Mental Health grant R01 MH57681

anoxia encephalitis Korsakoffrsquos syndrome) showmarked impairments in long-term memory fornewly encountered facts and events yet retain per-ceptual and linguistic functions within the normalrange (Parkin amp Leng 1993Squire 1994) Impor-tantly these studies have revealed that under someconditions amnesics may prove to be less suscepti-ble to illusory recollection than are individuals withintact memory with their impairment in veridicalmemory in some way buffering them against theerrors of illusory memory shown by control partici-pants In other instances however the reverse hasbeen found Under some conditions amnesics haveshown increased rates of false recognition provingto be more prone to such errors than are individualswith unimpaired memory

One significant factor that may mediate whichof these two patterns is observedmdashdepressed versuselevated false recognition among amnesics relativeto controlsmdashis the extent to which the nonstudiedlure items share semantic or conceptual featureswith previously encountered stimuli Several recentstudies have examined false recognition in amnesiaunder conditions where multiple items that aresemantically or conceptually related to a particulartheme word are presented (Schacter Verfaellie ampAnes 1997 Schacter Verfaellie Anes amp Racine1998b Schacter Verfaellie amp Pradere 1996) Akey observation from these studies is that underthese conditionsmdashwhere multiple items mightserve to reinforce the conceptual or semantic gist ofthe listmdashamnesic patients show both impairmentin correct recognition and decreased false recogni-tion For example Schacter et al (1996) comparedthe recognition performance of amnesics and con-trols in the converging verbal associates paradigmdeveloped by Deese (1959) and Roediger andMcDermott (1995) In this paradigm participantsare presented several lists of words where each ofthe words in a given list is associatively related to asingle (non presented) ldquocritical lurerdquo (eg partici-pants might be asked to remember the words bedrest awake tired dream etc without being pre-sented the word sleep the critical lure to which eachof these list words is associated) As expected theaverage rate of veridical recognition of control par-ticipants (83) substantially exceeded that of

amnesics (46) however amnesics also showedimpaired false recognition Whereas controls falselyrecognised 89 of the critical lures amnesicsrsquo falserecognition rate was 57 Largely similar findingswere reported by Schacter et al (1998b) and bySchacter et al (1997) In the latter study using dif-ferent word lists comprised of fewer semantic asso-ciates amnesics showed impaired true recognitionin two experiments and impaired false recognitionin one experiment (Expt 1) but not in another(Expt 2) however in the second experiment con-trol participants also showed particularly low levelsof false recognition

These observations of impaired false recognitionin amnesics under conditions where multiplesemantic associates are presented stand in markedcontrast to an early report indicating that com-pared to non-memory-impaired controls amnesicsmight show elevated rates of false alarms under con-ditions where the lures are related to only one of thestudied items (rather than multiple study items) Inan early study Cermak Butters and Gerrein(1973) found that when only a single related itemwas presented in a continuous recognition para-digm Korsakoff amnesics were more likely thanalcoholic controls to show false recognition of newitems that were similar to previously presenteditems Compared to controls Korsakoff amnesicsshowed significantly greater false recognition ofassociatively related words (eg study table testedwith chair) and also of homonyms (eg bearndashbare)amnesics also showed a nonsignificant tendency tomore often false alarm to synonyms (robberndashthief)and to unrelated items than did controls Morerecently in the study by Schacter et al (1996)amnesic patients demonstrated both a pattern ofldquoimpairedrdquo false recognition of semantically relateditems for which many study items were presentedand increased baseline rates of false alarms for itemsthat were unrelated to study items

What might account for these differing pat-terns Schacter et al (1996) suggested that whenmultiple related associates are presented false rec-ognition among the control participants dependson memory for the associative or semantic infor-mation that also supports veridical memory Theyproposed that because control participants were

KOUTSTAAL ET AL

318 COGNITIVE NEUROPSYCHOLOGY 1999 16 (345)

able to retain and integrate or ldquobind togetherrdquo thestudied items they formed a strong andwell-organised representation of the semantic gistof each set of associatesmdashthereby rendering rejec-tion of the semantically similar lures quite difficultbut also and simultaneously facilitating rejectionof lures that were unrelated to any of the lists Incontrast amnesic patients may have formed only aweak or degraded representation of the semanticgist of the list items Such weak and less well-integrated representations would render amnesicsless likely to show false recollection of the relatedlures but also relatively less likely to reject unre-lated new items successfully (which would not ldquofailto cohererdquo with their memory of the lists in thesame manner that unrelated items for the controlparticipants did) Based on these considerationsSchacter et al predicted that by systematicallyvarying the number of related items that were pre-sented ldquoit should be possible to specify a crossoverpoint where amnesics shift from enhanced toreduced false recognitionsrdquo (Schacter et al 1996 p329 emphasis added)

These outcomes each involved the false recogni-tion of lures that were conceptually or semanticallyrelated to the studied items Importantly howeverin contrast to these converging findings from stud-ies involving similarity of meaning the findingsrelating to the effects of perceptual similarity on falserecognition of amnesics are notably less clearWhereas some studies have suggested thatamnesics may also show impaired false recollection

when items share perceptual similarity (Schacter etal 1997) other studies have not found a deficit(Kolodny 1994) ormdashin instances where perceptualfeatures of studied stimuli are recombined to formnew items (Kroll Knight Metcalfe Wolf ampTulving 1996 Expt 2 also cf Reinitz Verfaellieamp Milberg 1996)mdashhave even reported signifi-cantly higher false recognition among amnesicindividuals1

Evidence that perceptually based false recogni-tion in amnesics also might be depressed comparedto that found for control participants was reportedby Schacter et al (1997) In addition to conditionsmanipulating the conceptual similarity of studieditems these researchers included a condition inwhich they manipulated the perceptual similarity ofwords exposing amnesic and control participantsto some study lists where the words shared a highlevel of orthographic and phonemic similarity (egfor the related lure fate the study list words werefade fame face fake mate hate late date and rateExperiment 1) or were all rhyming words (Experi-ment 2) In both experiments amnesics showedimpaired true recognition for such perceptuallysimilar items and also impaired false recognition(after correcting for baseline levels of false alarms)In one experiment the level of corrected false rec-ognition of amnesics did not exceed chance in thesecond both amnesics and controls showedabove-chance levels of corrected false recognitionbut perceptual false recognition of amnesics fell sig-nificantly behind that of controls

COGNITIVE NEUROPSYCHOLOGY 1999 16 (345) 319

PERCEPTUALLY BASED FALSE RECOGNITION

1The studies of Kroll et al (1996) and Reinitz et al (1996) are of particular interest in that in agreement with many researchers

they underscore the notion that the hippocampal system may play a key role in allowing the ldquocohesionrdquo or ldquobindingrdquo of different ele-ments into a single trace (eg Moscovitch 1994 Rudy amp Sutherland 1994 Schacter Church amp Bolton 1995 Shapiro amp Olton1994 Sutherland amp Rudy 1989) For example Reinitz et al (1996) found that both amnesics and controls often falsely recognisedcompound words that were comprised of a conjunction of words embedded in previously studied items (eg study ldquotoothpasterdquo andldquoheartacherdquo test with the lure ldquotoothacherdquo) however whereas controls showed substantially greater veridical recognition for actuallystudied items (that is items that preserved both the elements and their relation with one another) amnesics showed largely similar lev-els of false recognition of recombined conjunction lures and veridical recognition of the target words These researchers proposed thatamnesics may have especial difficulty retaining information concerning global stimulus structure so that although they can rememberindividual features or parts of items they have marked deficits in tasks that require access to information about how the parts or featuresare interrelated to one another Similar findings of especially heightened susceptibility to conjunction errors among amnesics were re-ported by Kroll et al (1996) both for words comprised of recombined syllables (eg study ldquovalleyrdquo and ldquobarterrdquo test with ldquobarleyrdquo) andfor schematic pictures of faces Nonetheless it is also important to note that unlike the experiments of both Kroll et al (1996) andReinitz et al (1996) in the present study the lures were not created by recombining previously presented features but were comprisedof new elements that were perceptually similarmdashbut not identicalmdashto presented items Because of this difference we focus on studieswhere perceptual similarity was manipulated without the recombination of previously studied elements

Although suggestive these findings are some-what difficult to interpret because the manipulationof perceptual similarity involved familiar wordsand thus may have invoked both semantic and per-ceptual processing This is especially important asthese findings conflict with the outcome of a studyin which perceptual similarity was manipulated forabstract novel materials that did not possess priorsemantic associations Kolodny (1994) examinedrecognition performance of amnesics and controlsfor ldquorandom dot patternsrdquomdashpatterns that were eachcomprised of nine simple black dots arrayed in dif-ferent statistically determined configurations (cfPosner Goldsmith amp Welton 1967 Posner ampKeele 1968) Participants saw multiple exemplarsderived from three different ldquoprototyperdquo patternsand then were given a yesno recognition test foritems that had been presented previously as wellas for new items derived from the prototypes andthe prototypes themselves False recognition ofamnesics numerically exceeded that of controlswith false recognition of amnesics (69) actuallysomewhat greater than true recognition (59) incontrast controls showed the reverse pattern (52false recognition 67 true recognition) A similarpatternmdashwith amnesics showing numericallygreater false recognition (73) than controls(63)mdashwas observed for a subset of the new itemsthat were prototypes or low-level distortions of theprototype that is items for which all else beingequal perceptually based similarity- or gist-basedresponding might be expected to be most pro-nounced Moreover a second experiment reportedby Kolodny (1994) involving pictures with repre-sentational content (eg paintings with religious orbiblical themes) but where lures probably sharedboth conceptual and perceptual similarity with tar-get items again showed elevated false recognitionamong amnesics compared to controls

Does false recognition derived from the percep-tual similarity of lures with previously encounteredtarget items differ in some way from that derivedfrom shared semantic similarity Direct compari-sons of these studies are hampered by the manymethodological differences across the paradigmsAlso because the Kolodny (1994)dot pattern studydid not include an estimation of baseline levels of

false recognition comparisons with the results ofSchacter et al (1997) using perceptually similarwords are difficult Nonetheless the differing pat-terns for these studies clearly suggest the need forfurther evidence regarding the influence of percep-tual similarity on illusory recognition in amnesicsrelative to controls Evidence is particularly neededunder conditions where (1) the stimuli are novelobjects or patterns for which contributions frompre-existing semantic knowledge are minimisedand (2) factors that manipulate the degree to whichgeneral similarity might contribute to performanceare used The possibility that there might be across-over pattern in the level of false recognitionshown by amnesics versus controls as a function ofthe degree of false recognition induced among con-trols is especially of interest because documentingsuch a pattern for novel perceptually related stimuliwould provide evidence for the similarity of gistextraction mechanisms across perceptual and con-ceptual domains

The experiment reported here was designed toaddress these issues We examined false recognitionerrors arising from perceptual similarity in a para-digm where the stimuli were abstract novel objectscomplex multi-featured shapes constructed from aprototype We employed two forms of manipula-tion thatmdashbased on findings from other experi-mentsmdashare known to affect ldquogeneral similarityrdquo(Curran Schacter Norman amp Galluccio 1997) orldquogist basedrdquo (Brainerd Reyna amp Kneer 1995Reyna amp Brainerd 1995) false recognitionresponding in controls One manipulation was cate-gory size Robust effects of category size have beenfound in several previous studies of false recogni-tion in individuals with intact memory functioningwith higher rates of false recognition found follow-ing the presentation of increasing numbers ofsemantically related words (eg Arndt ampHirshman 1998 Robinson amp Roediger 1997Shiffrin Huber amp Marinelli 1995) categorisedpictures of everyday objects (eg Koutstaal ampSchacter 1997 Koutstaal Schacter Galluccio ampStofer in press) and abstract patterns (eg HomaCross Cornell Goldman amp Schwartz 1973Omohundro 1981) The second manipulationinvolved the degree of perceptual similarity of the

KOUTSTAAL ET AL


lures transformational distance from a category pro-totype High rates of false recognition of prototypeitems have been reported among individuals withunimpaired memory (eg Franks amp Bransford1971Solso amp Raynis 1979) with false recognitionrates decreasing with decreasing similarity to theprototype

Based on the findings from Kolodny (1994) onemight expect no impairment among amnesics com-pared to controls in false recognition of perceptuallyrelated items (at least for the uncorrected data) andequivalent or even enhanced false recognition of theprototypes However based on the findings ofSchacter et al (1997) using perceptually similarwords and also on the findings from studies thatused multiple semantically related items (Schacteret al 1996 1997 1998b)mdashall of which pointed toimpaired gist-memory among amnesicsmdashit mightbe expected that amnesics would show a deficit infalse recognition relative to controls particularlyunder conditions where there was an opportunityfor a strong build-up of gist (for example for catego-ries where many related items were presented dur-ing the study phase and for category prototypes) Incontrast under conditions promoting a less strongbuild-up of gist in control participants (for exam-ple where only one or a few related exemplars hadbeen presented or for exemplars at a far distancefrom the category prototypes) one would not nec-essarily expect false recognition of amnesic individ-uals to fall behind that of control participantsWhen only a few categorically related exemplarshad been presented control participants would nothave formed a robust gist-like representation of thecategory and so should not be strongly inclined toshow false recognition of lure items from the cate-gory because of their general similarity to studieditems (cf Homa et al 1973 Koutstaal amp Schacter1997 Koutstaal et al in press Omohundro 1981Robinson amp Roediger 1997) Further when only afew categorically related items had been presentedcontrols may be more able to retain and draw uponmore ldquoitem-specificrdquo representations of the actuallypresented items so as to differentiate studied fromnonstudied items and thus successfully ldquoopposerdquo orldquosuppressrdquo false recognition responses (cf Schacteret al 1998a b) For these items the combination

of a less robust gist representation and possiblygreater retention of item specific informationamong the control participants might result in sim-ilar levels of false recognition for amnesics and con-trols indeed as in the case of Cermak et al (1973)for the smaller category sizes false recognitionamong amnesics might even exceed that ofcontrols

METHOD

Participants

Twelve amnesic patients (9 male 3 female) and 12individuals with intact memory functioning (8male 4 female) took part in the experiment Bothamnesics and their controls were screened at theMemory Disorders Research Center of the BostonVeterans Affairs Medical Center A subgroup of sixamnesic patients were of mixed aetiology (anoxiaencephalitis and thalamic infarct) and six had adiagnosis of alcoholic Korsakoffrsquos syndrome Allpatients were matched to a corresponding controlparticipant on the basis of age education and ver-bal IQ (Wechsler Adult Intelligence Scale-Revised) the Korsakoff patients were matched toindividuals with a history of alcoholism (alcoholiccontrols) whereas mixed aetiology patients werematched to controls with no history of alcoholism(non-alcoholic controls) Table 1 presents the char-acteristics of the individual amnesic patientsincluding their performance on the WechslerMemory Scale-Revised and where available ana-tomical lesion information Individual lesion datafor the Korsakoff patients is not available howeverseveral of these patients took part in a larger mag-netic resonance imaging study (Jernigan SchaferButters amp Cermak 1991) that used quantitativeimage-analytic methods to compare Korsakoffpatients to age-matched alcoholic controls andage-matched normal controls In this study on agroup level and relative to a larger normative sam-ple of 55 normal volunteers both alcoholic controlsand Korsakoff patients showed considerablyreduced diencephalic grey matter volumes how-ever the volume loss of Korsakoff patients was



significantly greater than that of alcoholic controlsin anterior regions of the diencephalon in additionrelative to alcoholic controls Korsakoff patientsshowed significant reductions in grey matter inmesial temporal structures and orbitofrontal cortex

The controls for the mixed amnesic subgrouphad a mean age of 507 years with an average of142 years of education and a verbal IQ of 1042Alcoholic controls had a mean age of 675 an aver-age of 128years of education and an average verbalIQ of 992

Design

The experimental design included a between-subjects factor of group (amnesic vs control) and

two within-subjects factors transformational dis-tance and category size Transformationaldistance had three levels for studied items (nearmiddle far) and four levels for nonstudied items(prototype near middle far) Category size hadfour levels for studied items one three six ornine related items presented at study (termedsingle small medium and large categoriesrespectively) and five levels for nonstudied itemsthe four levels above plus ldquonovel category itemsfor which no related items were presented atstudy these latter items provided an estimate ofbaseline levels of false alarms In additionnon-categorised items (termed ldquounrelatedrdquo seefollowing) were included both as studied itemsand as new items

KOUTSTAAL ET AL


Table 1 Characteristics of Amnesic Patients

WMS-Re

ndashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashPatient Aetiology Age Ed c Verbal IQd GM Delay Atn

CW Thalamic infarct bilateral 56 12 87 79 80 89anterior nucleia

AB Encephalitis mild 58 16 105 76 51 92diffuse cortical atrophyb

PD Anoxia enlarged ventricles 60 20 109 65 61 89diffuse cortical atrophyb

JM Anoxia bilateral 48 12 89 70 52 92hippocampal complexb

PS Anoxia bilateral 39 14 95 90 50 115hippocampal complexb

RL Anoxia 68 18 103 68 66 93Mean 548 153 980 747 600 955

AA Korsakoff 70 9 93 76 62 109LB Korsakoff 63 11 90 99 61 99PB Korsakoff 71 14 87 82 60 93RD Korsakoff 67 12 83 66 50 99WR Korsakoff 69 7 88 76 53 96RM Korsakoff 77 14 112 91 68 95Mean 695 112 922 817 590 985

Anatomical lesion information is provided where availablea Lesions assessed by CTb Lesions assessed by MRIc Ed = years of formal educationd Verbal IQ from the Wechsler Adult Intelligence Scale (Revised)e WMS-R = Wechsler Memory Scale-Revised scores are presented separately for the indices of general

memory (GM) delay and attention (Atn) The WMS-R does not provide scores below 50 and 50was the lowest score used to compute means

Stimuli

The stimuli were colour depictions of complexmulti-featured abstract objects created using acomputer graphics program (Aldus Freehand)Most of the stimuli were ldquocategorisedrdquo items gen-erated by first creating a novel prototype accordingto a specific set of construction rules (describedlater) and then generating additional exemplarsthat belonged to the same category through manip-ulations that distorted the initial prototype ingreater or lesser degree Also non-categorised orldquounrelatedrdquo items were created that did not followthe rules of construction for the categorised items

Prototypes for 18 different categories were firstcreated according to a set of construction rulesEach prototype consisted of a large central form(the ldquomainrdquo component) together with threesmaller features All of the prototypes were two-dimensional and were created so as to form a singleunified object with multiple parts The prototypeswere created so as to be as dissimilar from oneanother as possible Each category had a set of fourunique features associated with it these featureswere assigned in all possible groupings of three tocreate four ldquoprototypesrdquo per category with the onlydifference between the prototypes for a given cate-gory being the features they contained Specificallyusing the letters a b c and d to designate the fourpossible features of a category the prototypes for agiven category consisted of onemdashinvariantmdashmaincomponent with superimposed features of eitherabc abd acd or bcd The features associated with agiven category were selected to possess someldquopairwiserdquo similarity (ie the features within a cate-gory comprised two pairs of two features with thefeatures in each pair being relatively similar to oneanother) The placement of features on the mainunit was constant for a given feature such that if thedifferent prototypes of a category shared a givenfeature it appeared in a similar place

The initial prototypes were then manipulated invarious prescribed ways including alterations ofshape colour outline and size to create additionalexemplars that possessed varying degrees of simi-larity to the prototypes These manipulations speci-fied particular types and ranges of manipulations

and were applied in an algorithmic fashion to itemswithin each category The placement of features onthese exemplars was also altered but within a speci-fied range (eg a given feature would generallyappear more towards the top or side of an objectbut its precise placement varied somewhat)

Pilot research that examined yesno recognitionof the exemplars derived from these algorithmicallyapplied alterations suggested that the items weretoo similar to one another Participants were largelyunable to discriminate studied from nonstudiedexemplars Accordingly the stimuli were furtheraltered to increase the within-category distinctive-ness of the stimuli This involved several furthersteps first each item in a category was furthermanipulated in a specified manner (eg one exem-plarrsquos shape was ldquobentrdquo a certain amount anotherexemplarrsquos outline was increased in thickness etc)then following these manipulations three itemswithin each category that were still relatively similarto one another were selected for additional alter-ations that rendered the three items less similar toone another and to other exemplars within the cate-gory These latter alterations employed overallmanipulations (eg rendering the entire shape intoa somewhat ldquoswirledrdquo or ldquowarpedrdquo shape) and ifnecessary were followed by yet a further manipula-tion in which particular portions of the shape werealtered Finally because these further alterationprocedures resulted in non systematic variations inthe initial ldquodistance metricrdquo for the exemplars thesimilarity of the new items to the prototypes wasassessed by asking six raters to place the exemplarsin each category on a distance metric correspondingto how similar they were to the prototype Thetransformational distance of a given item was thendetermined by the average of the six ratersrsquo judge-ments in conjunction with the further ratings oftwo of the experimenters (EJ and WK) in caseswhere the judgesrsquo ratings did not point to a clearoutcome In addition the two exemplars that com-prised the clearest representatives of the three dis-tances (near middle and far) were chosen tocomprise the critical stimulus items (to be counter-balanced across study and test status see following)

Examples of the categorised items are providedin Fig 1 Shown in the upper two panels are



KOUTSTAAL ET AL


Fig 1 Examples of categorised stimuli (A B) and unrelated items (C) The upper two panels show examples from two different categoriesincluding from left to right a category prototype and items from the near middle and far transformational distances respectively Notethat although for illustrative purposes here the stimuli are shown in black and white the stimuli as shown to participants were presentedin colour with colour comprising an important attribute that was varied both within and across categories

exemplars from two categories including (from leftto right) a prototype and items from the near mid-dle and far distances In addition examples ofunrelated items are shown in the bottom panelunrelated items were never comprised of four com-ponents and could be either two- or three-dimensional figures

For each category the critical exemplars of eachdistance were randomly assigned to one of two sets(A or B) These sets were subsequently used forcounterbalancing the critical exemplars across stud-ied and nonstudied status In order to avoid con-founding the number of related exemplars that werepresented at study with the number of items thatwere tested only a subset of all the items in each cat-egory was tested for the three- six- and nine-itemcategories three old items (one from each distance)and four new items (one from each distance plusthe prototype) were tested for categories whereonly one item was presented only this single targetitem (always a middle distance item) and two newitems (one from the middle distance plus the pro-totype) were tested Category size at study wasmanipulated by systematically excluding some ofthe noncritical items depending on the categorysize to be used no noncritical items were excludedfor nine-item categories but a given set of threenoncritical items (one at each distance) wasexcluded for six-item categories and six noncriticalitems (two at each distance) were excluded forthree-item categories For categories where onlyone item was presented at study only one middledistance item from the critical items was presented(from Set A or Set B depending on the counterbal-ancing condition)

Templates for the study and test lists were cre-ated such that within each third of the list thenumber of items from each category type categorysize (including singles and unrelated items) andtransformational distance (near middle far) werebalanced in addition for the test lists the numberof novel items and prototypes was also balancedacross the test thirds Finally additional templateswere created employing different orderings of theitems within each of the test thirds

For counterbalancing purposes the 18 catego-ries were assigned to 6 stimulus subsets comprised

of 3 categories each these subsets were used todetermine across subjects whether a given cate-gory was shown as studied items or as novel itemswith 1 stimulus subset assigned to each of the largemedium small and novel category-size conditions(each of which were thus represented by 3 catego-ries) and 2 stimulus subsets assigned to the singlecondition (thus represented by 6 categories) Coun-terbalancing across subjects ensured that each cate-gory occurred in each of the 6 conditions once witheach category represented once with the ldquoArdquo andldquoBrdquo exemplars as the studied and nonstudiedexemplars Thus a counterbalancing required 12subjects Amnesics and their matched controls wereassigned the same counterbalancing lists

Procedure

The overall procedure involved three phasesincluding a study phase in which participants wereexposed to the stimuli under an incidental encodingtask a brief retention interval and the test phaseAll participants were tested individually in a singlesession of approximately 40 to 50 minutes Thestimuli at both study and test were presented via apowerbook computer with a colour monitor usingthe PsyScope experimental presentation programThe stimuli appeared in the centre of the screenwith prompts for responding to the encoding taskor the recognition test displayed beneath

In the study phase participants were shown atotal of 78 items (72 critical items preceded andfollowed by 3 buffer items critical items = 27 large18 medium 9 small 6 single and 12 unrelateditems) Each item was presented for 6 seconds andparticipants were asked to rate the ldquooverall com-plexityrdquo of the stimulus on a 9-point scale where ldquo1rdquoindicated the stimulus was not at all complex andldquo9rdquo indicated it was extremely complex In makingtheir complexity ratings participants wereinstructed to consider all aspects of the stimulusincluding both different dimensions of the stimulus(eg shape colour size and outline) and all of thecomponents of the stimulus Participants enteredtheir complexity ratings on the computer keyboardusing the number keys 1ndash9 The study items were



preceded by a brief practice session that includeditems from two further categories not employed inthe experiment and additional unrelated items

The study phase was followed by a 5-minuteretention interval during which participants per-formed an unrelated number search task In the testphase participants were shown a subset of the itemsshown in the study phase together with new itemsand were asked to designate each item as ldquooldrdquo (pre-viously presented) or ldquonewrdquo (never previously pre-sented) The test list included 117 items of which45 were ldquooldrdquo and 72 were ldquonewrdquo The old itemsconsisted of 3 items from each of the studied cate-gories with the exception of the single categorieswhere only the single studied item was presentedplus the 12 unrelated items (ie old items = 3 times 3large category items 3 times 3 medium category items3 times 3 small category items 6 times 1 single categoryitems and 12 unrelated items) The new items werecomprised of 3 related lure items from each of thestudied categories (or for the single categories 1new item) together with one prototype from eachcategory in addition there were 3 items plus theprototype from each of 3 novel categories and 12new unrelated items

RESULTS

All analyses are first presented combining acrossthe two amnesic subgroups (ie combining acrossthe mixed aetiology and Korsakoff subgroups)with results first considered as a function oftransformational distance and then as a function ofcategory size Thereafter the two subgroups arebriefly considered separately The Appendix pres-ents the central true and false recognition outcomesfor individual patients in both the mixed aetiologyand Korsakoff groups The results for the unrelateditems are presented first To increase the normalityof the distributions and to make ldquoequal differencesequally detectablerdquo (Rosenthal amp Rosnow 1991Snedecor amp Cochran 1989) all analyses ofproportions were performed after arcsinetransformation

Combined Groups

Unrelated ItemsWe first considered performance for the unrelateditems that is items that were markedly differentfrom the categorised items and that did not followthe basic rules of construction for category items(see Table 2) For these items amnesics showedsubstantially impaired veridical recognition meansof 65 vs 43 for controls and amnesics respec-tively F(122) = 502 MSe = 009 P = 04 Falsealarms to unrelated items were numerically but notsignificantly more frequent for amnesics (24) thancontrols (18) F lt 1 After correcting for these(nonsignificant) baseline differences in false alarmsto unrelated lures veridical recognition of controlparticipants also significantly exceeded that ofamnesics F(122) = 1052 MSe = 005 P = 004

Analyses by DistanceTable 3 presents the proportion of ldquooldrdquo responsesgiven to studied items (true recognition) andnonstudied items (false recognition) for amnesicsand controls with responses shown separately as afunction of distance from the prototype (near mid-dle far) Also shown are the proportion of ldquooldrdquoresponses to the prototypes (averaged for study cat-egories comprised of three six and nine items) andthe baseline rate of false alarms to novel items(Note that the results for ldquosinglerdquo items for whichonly one categorised item was presented at studyand which was always from the ldquomiddlerdquo distanceare not considered here but are considered in thesubsequent section considering the effects of cate-gory size)

KOUTSTAAL ET AL


Table 2 Recognition Responses for Unrelated (Noncategorised)Items

Hits False Alarms Hits - False Alarms

Amnesic 43 24 20Korsakoff 40 14 26Mixed 46 33 13

Control 65 18 47Alcoholic 56 18 38Non-alcoholic 74 18 56

From Table 3 it can be seen that whereas truerecognition of amnesics was largely unaffected bythe distance of studied exemplars from the proto-type (means of 60 59 and 56 for near middleand far respectively) true recognition of the con-trols was more strongly affected by distance Con-trol participants showed a high level of correctrecognition for near and middle distance items(80) but considerably reduced recognition foritems that were far from the prototype (62) Table3 also shows that false recognition for bothamnesics and controls declines with increasing dis-tance but with controls showing higher false recog-nition for near and middle distance items (meansfor near middle and far respectively for amnesicsof 58 50 and 39 compared to 69 62 and 41for controls) Importantly from Table 3 it can alsobe seen that the group differences in false recogni-tion are most pronounced for the prototypes Rela-tive to the controls amnesics show a falserecognition impairment of 20 for the prototypeitems compared to an impairment of 11 12 and2 for lures from the near middle and far distancesrespectively Amnesics and controls showed mod-erately high and very similar baseline rates of falsealarms ldquooldrdquo responses to novel category items of35 and 38 respectively False alarms to these novelcategory itemsmdashthat is items that followed thesame general rules as all of the categorised items butwere not from the categories that were pre-sentedmdashprovide a measure of participantsrsquo generalwillingness to respond ldquooldrdquo under conditions of

uncertainty but where the stimuli share broad simi-larity with the types of items that were studied

The true and false recognition scores followingcorrection for differences in baseline levels of falsealarms (Hits ndash Novel Category False Alarms andRelated False Alarms ndash Novel Category FalseAlarms or ldquonovel-correctedrdquo scores) are shown inTable 4 Focusing first on the true recognitionresults an initial 2 (Group amnesic or control) times 3(Distance near middle far) analysis of variance(ANOVA) on the novel-corrected true recognitionscores showed a significant effect of DistanceF(244) = 436 MSe = 003 P = 02 but no effectof Group F = 21 and no Group times Distance inter-action F lt 16 Pairwise comparisons showed thataveraging across groups novel-corrected true rec-ognition for near and middle distance itemsexceeded that of far distance items (smallestF = 666) but near and middle distance items didnot differ from one another (F lt 1) Combining thenear and middle distance items and comparingnovel-corrected true recognition for these itemswith that for the far distance items and also includ-ing Group as a factor again showed a main effect ofDistance F(122) = 1049 MSe = 002 P = 004however in this more focused analysis the differ-entially enhanced performance of the controls forthe near and middle distance items was morestrongly apparent F(122) = 408 MSe = 002P = 06 for the Group times Distance interaction

Next considering false recognition a 2(Group) times 4 (Distance prototype near middle



Table 3 True and False Recognition Responses by Transformational Distance

True Recognition False Recognitionmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdash mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashNear Middle Far Mean Near Middle Far Mean Prototype Novel

Amnesic 60 59 56 58 58 50 39 49 52 35Korsakoff 65 69 56 63 56 54 50 53 61 35Mixed 56 50 56 54 61 46 28 45 43 35

Control 80 80 62 74 69 62 41 57 72 38Alcoholic 85 82 69 78 72 65 46 61 72 47Non-alcoholic 74 78 56 69 65 59 35 53 72 28

Items for categories where only one item was shown at study are omitted because for these categories all exemplars were from themiddle distance The prototype means shown are also based only on categories where three six or nine exemplars were shownNovel indicates the baseline level of false alarms to ldquonovelrdquo category exemplars (0 exemplars shown at study)

far) ANOVA performed on the novel-correctedfalse recognition scores revealed a main effect ofDistance F(366) = 1534 MSe = 002 P lt 0001no effect of group F lt 15 and a slight trendtoward a Group times Distance interactionF(366) = 204 MSe = 002 P = 12 Pairwisecomparisons showed that except for the prototypeand near items novel-corrected false recognitiondecreased regularly with increasing distance fromthe prototype (prototype = near gt middle gt far forthe near middle and far comparisons smallestF = 577) Critically however although there wasno overall effect of group an analysis restricted tothe novel-corrected false recognition scores for theprototypes alone revealed significantly impaired falserecognition among amnesics relative to controlsF(122) = 436 MSe = 005 P = 05

Analyses by Category SizeTable 5 presents the proportion of ldquooldrdquo responsesfor the amnesic and control groups separately as afunction of category size (one three six or ninerelated exemplars shown at study) for both studieditems (ldquotrue recognitionrdquo) and nonstudied items(ldquofalse recognitionrdquo) Also shown are the propor-tion of old responses to novel items (ie the baselinemeasure of false alarms) In order to reduce thenumber of comparisons and to increase the powerof individual comparisons we combined the tworelatively smaller category sizes (one and three

related items presented) and the two relativelylarger categories (six and nine related items pre-sented) contrasting the two larger and two smallercategories

We first considered veridical recognition as afunction of category size after correcting for novelfalse alarms (see Table 6) For novel-corrected truerecognition a 2 (Group) times 2 (Category Size largervs smaller) ANOVA revealed no overall effect ofGroup F = 19 an effect of Category SizeF(122) = 1252 MSe = 001 P = 002 and aGroup times Category Size interaction F(122) = 471MSe = 001 P = 04 Overall novel-correctedveridical recognition was greater for larger (32)than smaller (23) categories but this difference waslargely carried by control participants (24 vs 20 foramnesics compared to 41 vs 25 for controls)

We next conducted a similar 2 (Group) times 2(Category Size larger vs smaller) analysis on thenovel-corrected false recognition scores This anal-ysis showed a main effect of Category SizeF(122) = 1451 MSe = 002 P = 001 with noeffect of Group F lt 1 and no CategorySize times Group interaction F = 20 False recogni-tion was greater for lures from many-exemplar cat-egories (21) than for lures from few-exemplarcategories (07) with both amnesics and controlsshowing this pattern (for amnesics 17 vs 08 dif-ference of 9 for controls 26 vs 06 difference of20)

KOUTSTAAL ET AL


Table 4 Novel-corrected True and False Recognition Responses by Transformational Distance

True Recognition False Recognitionmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdash mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashNear Middle Far Mean Near Middle Far Mean Prototype

Amnesic 26 25 21 24 24 15 04 14 17Korsakoff 30 34 21 28 21 19 15 18 26Mixed 21 15 21a 19a 26 12 - 07 10 08

Control 42 42 25 36 31 25 03 20 35Alcoholic 38 34 21a 31 25 18a - 01 14a 25Non-alcoholic 46 50 28 41 37 32 07 25 45

Items for categories where only one item was shown at study are omitted because for these categories all exemplars were fromthe middle distance The prototype means shown are also based only on categories where three six or nine exemplarswere shown

P lt 05 for comparisons against zeroa Marginal effects 05 lt P lt 10 for comparisons against zero

We also considered novel-corrected false rec-ognition of the prototypes separately as a functionof the number of exemplars that were studied A 2(Group) times 2 (Category Size larger vs smaller)ANOVA revealed a marginal effect of Groupwith controls tending to show greater overall falserecognition of the prototypes on this measure thanamnesics F(122) = 334 MSe = 008 P = 08(Note that because this analysis includes singlecategories it differs somewhat from the analysesreported earlier for transformational distancewhich excluded the single item categories andwhere amnesics were significantly impaired onthis measure) There was also a main effect of

Category Size F(122) = 1051 MSe = 005P = 004 and a trend toward a Group times CategorySize interaction F(122) = 410 MSe = 005P = 06 Comparing the level of novel-correctedfalse recognition of the prototypes for theamnesics versus controls within each category sizeseparately showed that amnesics were significantlyimpaired in false recognition of prototypes fromlarge categories F(122) = 492 MSe = 012P = 04 marginally impaired in false recognitionof prototypes from medium categoriesF(122) = 275 MSe = 015 P = 11 and notimpaired for prototypes from small or single-itemcategories Fs lt 1



Table 5 True and False Recognition Responses by Category Size

True Recognition False Recognitionmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdash mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashndashmdashmdashmdashmdashmdashmdashndashndashmdashmdashmdashndash

Category Items Category Items Prototypemdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdash mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdash mdashmdashmdashmdashmdashmdashndashndashmdashmdashmdashndash9 6 3 1 Mean 9 6 3 1 Mean 9 6 3 1 Novel

Amnesic 62 56 57 53 57 55 48 44 40 47 56 44 56 29 35Korsakoff 76 57 56 56 61 67 50 43 42 50 72 61 50 25 35Mixed 48 54 59 50 53 43 46 46 39 44 39 28 61 33 35

Control 81 76 65 60 70 68 58 45 42 53 86 70 61 32 38Alcoholic 83 82 70 64 75 65 63 56 50 58 89 61 67 36 47Non-alcoholic 78 70 59 56 66 70 54 35 33 48 83 78 56 28 28

False recognition responses to the prototype are also shown separately as a function of category size (nine six three or one itemshown at study) Also shown are false alarms to ldquonovelrdquo items

Table 6 Novel-corrected True and False Recognition Responses by Category Size

True Recognition False Recognitionmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdash mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashndashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashndash

Category Items Category Items Prototypemdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdash mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashndash mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashndash9 6 3 1 Mean 9 6 3 1 Mean 9 6 3 1 Mean

Amnesic 27 21 23 18 22 20 13 10 06 12 21a 10 21 - 06 12Korsakoff 41 23a 21a 21a 26 32 15 08 07 16 38 26a 15 - 10 17Mixed 13 19 25 15 18a 08 12 12 04 09 04 - 07 26 - 01 06

Control 43 38 27 22 33 30 21 08 04 16 49 32 24 - 06 25Alcoholic 36 34 23 17 28 18 16 08 03 11 42 14 20 - 11 16a

Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33

Results are shown for true and false recognition following correction for baseline levels of false alarms to novel category items P lt 05 for comparisons against zeroa Marginal effects 05 lt P lt 10 for comparisons against zero

Subgroup Analyses

From the means shown in Table 3 and Table 5 itcan be seen that in absolute terms the Korsakoffpatients generally showed higher levels of both trueand false recognition than did mixed amnesics(average true recognition for all four category sizesfor Korsakoff vs mixed amnesics respectively of 61vs 53 average false recognition of 50 vs 44)Korsakoff patients also showed greater false recog-nition of the prototypes than did mixed amnesicswith this effect especially pronounced for items forwhich nine or six related exemplars had been stud-ied (Korsakoff patients means of 72 and 61mixed amnesics means of 39 and 28 respectively)However a largely similar pattern was found for thetwo subgroupsrsquo respective control groups with thealcoholic controls on average showing numericallyhigher true recognition than did the mixed aetiol-ogy controls (means for all four category sizes of 75vs 66 respectively) and also higher false recogni-tion (means of 58 vs 48) An exception howeverconcerned unrelated items (see Table 2) wherealcoholic controls showed relatively less true recog-nition than did mixed controls (56 and74respectively)

Because the alcoholic controls also showed anespecially high rate of false alarms to novel items(47 compared to 35 for the Korsakoff amnesics)after correction for baseline levels of false alarmsthere were relatively small and unsystematic differ-ences in corrected true and false recognition for theKorsakoff group compared to their controls (eg forall four category sizes overall novel-corrected truerecognition of 26 vs 28 for Korsakoff patients andalcoholic controls respectively overall novel-corrected false recognition of 16 vs 11 likewisefor false recognition of the prototypes novel-corrected false recognition averages of 26 vs 25respectively) In contrast mixed amnesics and theircontrols showed a less pronounced difference inbaseline rates of false alarms to novel categoriseditems (35 and 28 respectively)

Focusing on the performance of the mixedamnesia subgroup Table 3 and Table 5 show thatthe recognition performance of the mixed amnesicpatients was quite consistent with the expectation

that under conditions of relatively little or weakgist amnesics might show levels of false recogni-tion similar to or exceeding that shown by con-trols False recognition of the mixed amnesicsnumerically exceeded that shown by controls fornovel items (35 vs 28) for single-item category lures(39 vs 33) and for three-item category lures (46 vs35) However also as expected under conditionsof relatively greater build-up of gist this patternreversed with the control participants showinghigher levels of false recognition than the mixedamnesics for larger category sizes where either sixor nine items had been presented (for six-item cate-gories false recognition of 46 vs 54 for mixedamnesics versus their controls respectively fornine-item categories 43 vs 70) This differencewas also observed for the prototypeswhere the falserecognition rate of mixed amnesics fell nearly 30behind their controls (combining across categoriesof three six and nine items means of 43 vs 72)Indeed within the prototypes themselves a similarldquocrossoverrdquo in performance as a function of categorysize (and presumably ldquogistrdquo) was observed withmixed amnesics falling well behind their controls infalse recognition of the prototypes for the six- andnine-item categories (39 vs 83 and 28 vs 78respectively) but with the pattern reversed for thethree-item and single-item categories (61 vs 56and 33 vs 28 respectively)

For the novel-corrected data subgroup analysesrevealed no significant differences for the Korsakoffamnesics compared to their controls For the mixedamnesics significant or nearly significant differ-ences in novel-corrected true recognition werefound for near-distance items [means of 21 vs 46F(110) = 498 MSe = 005 P = 05] for middle-distance items [means of 15 vs 50 F(110) = 464MSe = 009 P = 06] items from large categories[means of 13 vs 50 F(110) = 964 MSe = 005P = 01] and for unrelated items [means of 13 vs56 F(110) = 1301 MSe = 005 P = 005] Fornovel-corrected false recognition significant ornearly significant differences were found for mid-dle-distance items [means of 12 vs 32 F(110) = 366 MSe = 004 P = 08] for items fromlarge categories [means of 08 vs 43F(110) = 527 MSe = 008 P = 04] prototypes

KOUTSTAAL ET AL


for large categories [means of 04 vs 56F(110) = 655 MSe = 016 P = 03] prototypesfor medium categories [means of ndash 07 vs 50F(110) = 1009 MSe = 012 P = 01] and for theprototypes from large medium and small catego-ries combined [means of 08 vs 45 F(110) = 731MSe = 006 P = 02] The pattern of depressedfalse recognition for the larger categories and forprototypes was quite consistent across patients (seeindividual patient data in the Appendix) for thelarger category sizes and for the nearmiddletransformational distances five of the six mixedaetiology patients showed false recognition scoresthat fell below the mean of the controls for the pro-totypes all of the mixed aetiology patients showedlower false recognition rates than the mean for theircontrols

GENERAL DISCUSSION

A central finding of this experiment is thatamnesics may show either substantial impairmentof perceptually based false recognition or virtuallyno impairment depending on the nature of theexperimental conditions and the extent to whichthey allow the extraction of gist information amongcontrol participants Under some conditionsmdashpar-ticularly those that allow maximal strengthening ofgist-based representations in controlsmdashamnesicsshowed a substantial impairment of perceptuallybased false recognition for pictures of novel abstractobjects for which they had no pre-existing semanticor conceptual knowledge However under condi-tions where control participants showed relativelyless robust false recognition false recognition ofamnesics was also less impaired The deficit inamnesicsrsquo false recognition was especially apparentfor the prototypes For the control participants theprototypes elicited the highest levels of false recog-nition that were observed for any of the experimen-tal conditions with rates of false recognition thatmatched or for some conditions exceeded thoseseen for true recognition Yet for these itemsamnesic patients showed a substantial and reliabledeficit in false recognition

Although this pattern was observed both in theoverall group analysis (associated with a false recog-nition deficit of 18 after taking baseline falsealarms into account) and in the analysis for themixed aetiology subgroup alone (associated with anovel-corrected false recognition deficit of 37)the unusually high baseline level of false alarmsamong the alcoholic control group largely elimi-nated the pattern observed for the uncorrected orabsolute scores in the subgroup of Korsakoffpatients Thus the conclusions for this subgroup areless clear Ignoring the baseline differencesKorsakoff and mixed aetiology patients generallyshowed not dissimilar patterns for transformationaldistance Korsakoff patients showed comparativelygreater false recognition ldquodeficitsrdquo for the prototypeand near and middle distance items (deficits of 1116 and 11) than for far distance items (deficit ofndash 04) mixed aetiology patients did not show a verymarked deficit for near items (04) but showedstrong deficits for middle distance items (13) andan extremely marked deficit for the prototypes (29)with little deficit for far distance items (07) For thefactor of category size however mixed aetiologyamnesics showed more systematic ldquogist-relatedrdquodeficits than did Korsakoff patients

The nature of this ldquogist-relatedrdquo pattern of defi-cits in mixed amnesics is perhaps most clearly seenby comparing the differences in false recognition ofmixed amnesics versus their controls for the variouscategories These differences were positive show-ing numerically higher false recognition foramnesics than for controls for each of five catego-ries involving few items novel items single itemssingle item prototypes small category items andsmall category prototypes (differences of + 7+ 6 + 5 + 11 and + 5 respectively) inmarked contrast for medium and large categoryitems and medium and large category prototypesthe differences were all negative and in severalinstances of substantial magnitude (ndash 8 ndash 27ndash 50 and ndash 44) now in each case showinghigher false recognition for controls than foramnesics or ldquoimpaired false recognitionrdquo foramnesics A focused comparison contrasting theperformance of mixed amnesics versus their con-trols for the former set of items drawn from



ldquolow-gistrdquo conditions against those for the latter setof items drawn from ldquohigh-gistrdquo conditions yieldedno effect of Group F lt 1 a main effect of Gist(high gist = 55 low gist = 39) F(110) = 915MSe = 002 P = 01 and a highly significantGroup times Gist interaction F(110) = 1440 MSe =002 P = 004

This interaction provides clear evidence for aldquocrossoverrdquo in the false recognition performance ofamnesics versus control participants as a function ofthe number of categorically related items that hadbeen encountered previously Moreover consistentwith the proposal made earlier by Schacter et al(1997) the across-group difference tracks with themagnitude of the false recognition effect in the controlgroup with mixed aetiology amnesics showing thegreatest deficits under conditions where gist-basedresponding among their normal controls is mostrobust namely for the large categories (70) andmedium categories (54) and especially for the pro-totypes from these categories (83 and 78 respec-tively) Mixed amnesics showed no such deficithowever for the small or single item categorieswhere gist-based responding among controls wasalso much lower (35 and 33 respectively)

What might account for these differential pat-terns with amnesics showing little or no impair-ment under some conditions (few or fartransformational distance exemplars) and substan-tial impairment under others (many or neartransformational distance exemplars) A pattern ofthis form might emerge if amnesics showed uni-formly little effect of the manipulations of categorysize or transformational distance but controls weresensitive to these manipulationsmdashacquiring anincreasingly strong sense of familiarity with the cat-egories as more exemplars were encountered andmore often incorrectly succumbing to this sense offamiliarity if a lure item was similar to many previ-ously encountered items or if it was similar to (oractually was) the most representative or ldquoproto-typicalrdquo instance of the category On this account adifferential pattern of impairment across the differ-ent conditions would emerge because in the con-trol participants general familiarity with thecategories was relatively weak in some conditions(and thus matched that of amnesics) but continued

to accumulate in other conditions (and thusexceeded that of amnesics)

This simple account however does not appearto provide a sufficiently modulated portrayal of thedata Although overall the performance of amne-sic patients was less affected by the factors oftransformational distance and category size thanwas found for control participants amnesics werenot entirely immune to the influence of perceptualgist or general similarity information Combiningacross the two amnesic subgroups amnesicsrsquo overalllevel of false recognition for the category lures (47)significantly exceeded their baseline level of falsealarms to novel category items (35)F(111) = 737 MSe = 001 P = 02 Further atleast in numerical terms amnesics showed someldquogist-like systematicityrdquo in their false recognitionresponses to categorised lures showing numericallygreater false recognition of categorised lure itemsthat were closer in similarity to the prototypes(means of 58 50 and 39 for near middle and fardistance items respectively) and also for lures forwhich a larger number of related items had beenpresented at study (55 48 44 and 40 for studycategories comprised of nine six three and oneitems respectively) This result suggests thatamnesics were able to retain at least some ldquocate-gory-specificrdquo information regarding the stimulusitems and that this informationmdashlike that for con-trolsmdashwas influenced by the factors oftransformational distance and category size albeitto a lesser degree

The obtained pattern of results therefore can bebetter understood in terms of a differential build-upof gist information in the amnesics compared to thecontrols If category specific information wasextracted andor retained less efficiently byamnesics than by control participants this differ-ence might be least evident in cases where controlsalso were able to construct only relatively weak gistrepresentations (as when few category exemplarswere presented) but might emerge more strongly asthe gist representations of controls became increas-ingly robust Given the relatively low levels ofwithin-category discriminability shown by bothamnesics (on average approximately 10) andcontrols (less than 20 for the categorised items

KOUTSTAAL ET AL


excluding the prototypes) it is likely that this cate-gory-specific information was of a relatively globaland ldquocoarserdquo form most often involving generalfamiliarity with the nature of ldquocategory-likerdquo itemsrather than specific or richly detailed recollection

This account of the differences in false recogni-tion between amnesics and controls is based on thenotion that an impairment in gist extraction mightselectively affect performance for large categorysizes and near transformational distance exemplarsThe same pattern of results however mightemerge if control participants had stronger gist rep-resentations for the smaller category sizes or fartransformational distance exemplars but if con-trolsmdashunlike amnesicsmdashwere able to counteractfalse recognition responses successfully for theseitems (cf Schacter et al 1998b) That is it is possi-ble that individuals in the control group actuallyhad a stronger gist-like representation than didamnesics for the smaller categories (as they appar-ently did for the larger categories) but they wereable to oppose or suppress gist-based respondingfor these items by relying on more detaileditem-specific memory (which might for examplesuffer from less interference when few rather thanmany overlapping exemplars had been studied cfSchacter et al 1998a)

The overall pattern of performance by the con-trol participants argues against a strong contribu-tion from this factor For a ldquoselective suppressionrdquoaccount of this form to be viable it would beexpected that control participants would demon-strate a high level of item-specific memory in thoseconditions in which suppression was occurringSuch ldquoitem-specificrdquo memory might be shown invarious ways but one measuremdashreadily obtainedfrom the present datamdashis participantsrsquo ability todiscriminate studied items from similar but notstudied lures (ie hits minus related false alarms)Overall controls showed somewhat greater abilitythan did amnesics to discriminate studied targetsfrom related category lures Yet contrary to whatwould be expected on the basis of a selective sup-pression account amnesics and controls showednumerically less pronounced differences in suchwithin-category discriminability for the far dis-tance items (True Target ndash Related False Target

means of 17 for amnesics compared to 21 for con-trols) than for middle distance items (means of 09vs 18)or near distance items (means of 02 vs 11)Similarly for the factor of category size the differ-ences in within category discriminability foramnesics versus controls were not noticeably morepronounced for the smaller categories (means of 13and 19 for amnesics and controls) than for thelarger categories (means of 07 vs 15 for amnesicsand controls)

One qualification concerning the differentialpatterns of false recognition for amnesics versuscontrols that we observed concerns the levels offalse recognition that emerged for lures at a fartransformational distance from the prototype andfor lures from smaller categories where participantswere exposed to only one or three target items percategory For these conditions neither amnesicsnor controls showed rates of false recognition sig-nificantly above those found for novel categorylures (on average for both combined groups falserecognition of far distance items was approximately4 greater than for novel category lures andbetween 6 and 8 for the smaller categories alsosee Table 4 and Table 6 where asterisks indicatethe conditions where novel-corrected true recogni-tion and novel-corrected false recognition exceededchance levels) On the one hand this pattern is ofconcern because where false recognition rates are atfloor or at near-floor levels it would be difficult todetect a difference between the amnesic and controlgroups On the other hand and while acknowledg-ing this as a legitimate concern it might also benoted that the absolute levels of false recognitionfor the far distance lures and smaller category lureswere not inconsiderable ranging between 28 and50 for the four subgroups for far distance itemsand between 33 and 56 for the smaller categorysizes Further the near-equivalence of false recog-nition rates in these cases to the false alarm rateobserved for lures from novel categoriesmdashthat isitems that followed the same general rules of con-struction as the studied categorised items false rec-ognition rates of 28 to 47mdashmay reflect a realfunctional equivalence The overall levels of generalfamiliarity of novel category lures would not beexpected to differ radically from that for lures from



categories where only one or three related exem-plars had been presented Still it is possible thatespecially under conditions involving lower base-line levels of false recognition some differences inthe false recognition rates of amnesics versus con-trols might be observed for these conditions

These considerationsmdashwhich have thus far beenconfined to categorised itemsmdashhave emphasisedthe parallels in the memory processes shown byamnesics and controls with amnesics showing sim-ilar but less efficient gist extraction than that shownby individuals with intact memory However theemergence of clear group differences for thenon-categorised or unrelated items where the mem-ory performance of controls far exceeded that ofamnesics suggests that in addition to building upcoarse or global forms of category informationcontrol participants were also able to encode andretain more detailed item-specific informationEvidence for this latter form of more detailed orspecific recollective informationmdashand the need todifferentiate it from other forms of informa-tionmdashespecially derives from a comparison of par-ticipantsrsquo corrected true recognition for theunrelated items versus their recognition of items forwhich a single categorised item was presentedThese two conditions were similar in that for bothconditions only one item of a given sort was pre-sented at study however whereas the single itemsfollowed the rules of construction for the categor-ised items (and were themselves category items)the unrelated items did not follow those rules andwere noticeably different from the categoriseditems Whereas amnesic and control participantsshowed essentially identical novel-correctedveridical recognition for the single categoriseditems ( 18 and 22 respectively) corrected veridicalrecognition of the two groups diverged markedlyfor the unrelated items For the unrelated itemsamnesics achieved a very similar score to that seenfor single items (ldquounrelated-correctedrdquo mean of20) but control participantsmdashnow presumablyresponding on the basis of detailed item-specificinformationmdashshowed significantly higher levels ofrecognition (47) An analysis contrasting correctedveridical recognition of the single versus unrelateditems for the two groups yielded a significant inter-

action of group with item type F(122) = 562MSe = 003 P = 03 indicating reliably differentialpatterns of responding for these item types foramnesics versus controls

The above pattern of findings with controlsshowing greater item-specific memory for noncategorised items and alsomdashin the case of the pro-totypesmdashgreater gist-based false recognition raisesa question regarding the extent to whichitem-specific memory and gist memory mightcovary Is decreased gist-based false recognition aninevitable consequence of decreased item-specificmemory Although generalisations across differentparadigms and different populations need to bemade very cautiously it might be noted that con-trary to this suggestion across a number of studieswe have observed several different relationsbetween indices that probably primarily reflectitem-specific memory (veridical recognition ofsingle or unrelated ldquoone-of-a-kindrdquo items) andgist-based false memory (false recognition of luresthat are perceptually andor conceptually similar totarget items) For example in other experimentsusing the same abstract novel objects as in the pres-ent experiment but comparing older and youngeradults we observed largely equivalent levels ofgist-based false recognition in older and youngeradults even though like the control participants inthe present study younger adults demonstratedgreater item-specific memory (as indexed by per-formance for unrelated items) than did older adultsIn a paradigm involving categorised pictures ofeveryday objects we found that older adults showreliably greater gist-based false recognition thanyounger adults together with decreased item-specificmemory as measured by their recognition of unre-lated items or items from categories where only asingle related item was presented (Koutstaal ampSchacter 1997 Koutstaal et al in press) Furthermatching older and younger adults on their level ofitem-specific memory by requiring younger partic-ipants to perform a secondary task during the initialencoding of the pictures (Koutstaal Schacter ampBrenner 1998) did not eliminate the age-difference in gist-based responding Under theseconditions the two age groups showed similar lev-els of item-specific memory but older adults con-

KOUTSTAAL ET AL


tinued to show greater gist-based false recognitionThese varied patterns of gist-based false recogni-tion in relation to item-specific memory for olderversus younger adults suggest that the two forms ofresponding may be determined by multiple factorsand though they may sometimes covary (as theydid here) at other times they may not (For moregeneral discussion and findings concerning thebases underlying true and false recognition and theconditions under which they may or may notcovary see Arndt amp Hirshman 1998) These mul-tiple patterns also make it hazardous to predictwhether if the amnesic patients and control partic-ipants were led to have similar levels ofitem-specific memory (by for example testingcontrol participants after a longer study-to-testinterval) they would show false recognition similarto that of the amnesics Although an increasedstudy-to-test interval might decrease item-specificmemory in the control participants it is possiblethat false recognition of categorised items particu-larly category prototypes would remain relativelypreserved (cf McDermott 1996 Payne ElieBlackwell amp Neuschatz 1996 Reyna amp Brainerd1995) with controls still more often falsely recog-nising these items than amnesics

Theoretical Implications

Overall the outcomes of the present experiment areconsistent with the predictions based on prior stud-ies using multiple conceptually or perceptuallyrelated words where under conditions where con-trols showed high levels of illusory recognitionamnesics also showed substantial deficits in falserecognition The present demonstration of a signif-icant impairment in perceptually based false recog-nition of prototypes among amnesics using novelabstract visual stimuli extends these earlier find-ings showing that impairments in amnesic patientsmay be observed in cases where the stimuli have nopre-existing semantic associations and where theextraction of similarities depends on representa-tional and associational processes that occurentirely within the experimental session itself

The parallel patterns found for both perceptu-ally similar items (this study Schacter et al 1997)

and for semantically or conceptually similar items(Schacter et al 1996 1997 1998b) with amnesicsshowing ldquoimpairedrdquo false recognition under condi-tions where individuals with intact memory showhigh levels of false recognition but not under condi-tions where cognitively intact individuals show lesspronounced levels of false recognition suggest thatsome of the processes underlying false recognitionmay be similar across these broad types of similar-ity In particular both forms of false recognitionappear at least in part to derive frommdashand also toreflectmdashlimits on the ldquoexactrdquo or ldquoverbatimrdquo encod-ing retention andor retrieval of information Inboth domains increasing convergence or overlap ofthe stimulus items whether in terms of semantic orassociative attributes (eg Robinson amp Roediger1997 Vogt amp Kimble 1973) or perceptual features(eg Franks amp Bransford 1971 Solso amp Raynis1979) may act to increase the likelihood of falserecognition of nonstudied items that possess simi-lar characteristics

Nonetheless these findings do not imply thatthe processes are entirely parallel across the percep-tual and conceptual domains Although we haveunderscored the congruencies in the processesrelating to these broad types of similarity in regardto certain aspects particularly in relation to thenumber of related items that are presented andtheir ldquodistancerdquo from a central prototype otheraspects of the processing of perceptually and con-ceptually similar materials are quite likely to differThus depending on the particular paradigm usedspecific factors may contribute differently to falserecognition (or other forms of memory errors suchas false recall) in the conceptual versus perceptualdomains For example with semantically or asso-ciatively related items such as the converging asso-ciates used in the DeeseRoediger-McDermottparadigm where all of the items converge upon asingle critical lure errors may arise from a form ofinferential and elaborative processing at the time ofencoding andor retrieval in a way that differs fromthat for novel perceptually similar items wherepre-existing associative and thematic informationis less likely to be strongly evoked Further whereasthe convergence of items in semantic paradigms(during either study or test) may lead to explicit



phenomenological awareness of the nonstudiedlure because this item is generated or automaticallyactivated in response to the multiple semanticallyrelated items (cf Gallo Roberts amp Seamon 1997McDermott amp Roediger 1998 Read 1996Roediger amp McDermott 1995) the convergenceof visually similar items particularly novel patternsor shapes is unlikely to result in the ldquogenerationrdquo ofa ldquolure in quite the same way Although a givenexemplar might remind an individual of earlier pre-sented similar target items participants wouldappear to be unlikely to generate ldquonovelrdquo variants orexemplars of the targets spontaneously (in partbecause pre-existing associational support for suchgeneration is lacking) This suggests that directldquosource confusionsrdquo between items that are pre-sented and items that are implicitly or consciouslygenerated during study (did I hear or read this wordversus only think it) may be more likely to arise forsemantically or associatively related words than forperceptually similar items

Although a number of studies have focused onthe effects of either perceptual similarity or concep-tual similarity on false recognition the relationsbetween perceptual and conceptual similarityincluding the possibility that they may have addi-tive effects have been much less explored In theparadigm using categorised pictures of everydayobjects that we have used to explore false recogni-tion in older and younger adults (Koutstaal ampSchacter 1997 Koutstaal et al in press) theobjects within a particular category (eg catschairs or teddy bears) may share both considerableperceptual similarity (general shape colour etc)and also conceptual (semantic and lexical) informa-tion however in this paradigm perceptual andconceptual similarity are confounded and the rela-tive roles of each in false recognition can be assessedonly indirectly by considering the consequences ofshifting attention to the different components orsimilar manipulations More direct evidence for thepossible additive effects of perceptual and concep-tual similarity has been provided by Henkel andFranklin (1998) These researchers found higherrates of source monitoring errors in a judgementtask requiring participants to differentiate betweenvisually presented versus imagined items if the

items both physically resembled one another andbelonged to the same functional category (egbagelndashdoughnut) than if items shared primarily onlyconceptual similarity (eg pantsndashshirt) or primarilyonly physical similarity (eg magnifying glassndashlolli-pop) The possible effects of combining both per-ceptual and conceptual similarity on falserecognition performance including false recogni-tion of amnesic patients versus controls have notbeen explored

Although our finding that amnesics showed sig-nificantly depressed false recognition for proto-types based on perceptual similarity is consistentwith the results from previous studies of false recog-nition in amnesia using conceptually related words(Schacter et al 199619971998b) or orthograph-ically and phonologically similar words (Schacter etal 1997) this outcome clearly appears to be incon-sistent with the findings reported by Kolodny(1994) particularly those of the random-dot pat-tern study where amnesics showed false recogni-tion rates that exceeded those of controls even foritems that were prototypes or low distortion exem-plars of the categories As noted previously theabsence of baseline rates in that study makes com-parisons difficult and it is possible thatmdashafter cor-rectionmdashthe results from the two studies would beless discrepant In addition the studies differed inmany respects One factor in particular that recentresearch has suggested may have been important isthe number of times that the target items were pre-sented Whereas target items were presented onlyonce in the present study targets were presentedrepeatedly in the Kolodny study This latter factor isimportant because repeated presentation of the tar-gets may have allowed control participantsmdashatleast to some extentmdashto suppress false recognitionof the lures Although within the current experi-ment we found little evidence of ldquosuppression-likerdquoprocesses other experiments using the semanticconverging associates paradigm have demonstratedsuppression among younger persons with intactmemory when the target items have been presentedrepeatedly (Kensinger amp Schacter this issueMcDermott 1996)mdashpossibly because such repeti-tion facilitates the encoding and retention ofgreater amounts of item-specific information or

KOUTSTAAL ET AL


otherwise encourages more careful monitoring ofrecognition responding Importantly in a studycomparing the effects of repeated study presenta-tion on the false recognition of amnesics versuscontrols Schacter et al (1998b) found that whereasrepeated presentations of semantic associates leadto decreasing false recognition in controls amnesicsshowed either increasing rates of false recognitionacross trials (Korsakoff patients) or fluctuating lev-els of false recognition (amnesic patients of mixedaetiology) These findings raise the possibility thatthe repeated target presentations in the study byKolodny may have acted both to decrease false rec-ognition among controls (albeit not too successfullygiven that in absolute terms the rates of false rec-ognition were not inconsiderable) and either toexacerbate or at least to maintain false recognitionin amnesic individuals

From a broader theoretical perspective ourresults showing impaired gist-based false recogni-tion of prototypes in amnesics bear upon two moregeneral questions about the nature of preserved ver-sus impaired memory functions in amnesics Oneissue concerns the relation between categorisation orclassification tasksmdashthat is tasks involving theacquisition and transfer of knowledge concerningthe classification of novel stimuli including theability to correctly assign novel items to classesbased on experience with other category mem-bersmdashand recognition memory involving theexplicit recollection of stimuli or events Previouslyreported evidence showing a dissociation in theperformance of amnesics relative to control partici-pants on these two types of task with amnesicsshowing impaired recognition memory but intactor near-normal levels of categorisation learning(Knowlton amp Squire 1993 Kolodny 1994 Squireamp Knowlton 1995 also cf Knowlton Mangels ampSquire 1996 Knowlton Ramus amp Squire 1992Knowlton Squire amp Gluck 1994) has provokedconsiderable discussion and research concerningthe factors underlying this dissociation especiallywhether it is necessary to invoke separate memorysystems to account for the difference or if otheraccounts (eg Nosofsky 1988 Nosofsky amp Zaki1998) are possible Our finding of impairedgist-based false recognition of prototypes in

amnesics appears to provide additional evidence forthe distinction between the processes or structuresthat support categorisation performance and thosethat underlie recognition (in this case illusoryrather than veridical recognition) Indeed a grow-ing set of evidence suggests that in contrast tointact classification performance including intactclassification of perceptually based prototypes(Knowlton amp Squire 1993 Kolodny 1994 Squireamp Knowlton 1995) amnesic patients showimpaired false recognition of ldquoprototypicalrdquo butnot-presented target items with such deficitsfound for semantically similar words (Schacter etal 1996 1997 1998b) perceptually similar words(Schacter et al 1997) andmdashin the presentstudymdashperceptually similar novel abstract patternsOn the one hand it is not known if amnesics wouldshow normal or near-normal classification perfor-mance for complex abstract stimulus patterns of theform used here as has been found for the randomdot patterns (Knowlton amp Squire 1993 Kolodny1994 Squire amp Knowlton 1995) and artificialgrammar strings (Knowlton et al 1992 Knowltonamp Squire 1994) used by previous investigators Onthe other hand the findings of the current studytaken together with those of the several earlier stud-ies employing semantically and perceptually similarwords suggest that different processes and possi-bly different representations may supportldquogist-based false recognitionrdquomdashmanifested duringa task requiring the making of explicit recognitiondecisions upon the basis of an episodic query of thepastmdashand categorisation performance requiringthe designation of items as belonging or notbelonging to a specific category without any director necessary involvement of recollection Althoughthe factors underlying the dissociation betweenamnesicsrsquo recognition versus classification perfor-mance are likely to be a subject of continuing inves-tigation a clear implication of the present findingsis that differences in the performance of amnesicsand control participants as manifested in illusory orfalse recognition also comprise one part of the over-all data set for which a satisfactory account must beoffered

Finally and also in a broader perspective itmight be noted that our results provide evidence



against early proposals suggesting that whereasconceptual or semantic processing is impaired inamnesia perceptual processingmdashnot only forimplicit tests but also for explicit testsmdashmight beunimpaired in amnesia (Blaxton 1989 1995) Wefound that under certain conditions perceptuallybased false recognition was clearly and significantlyimpaired in amnesics thereby demonstrating thatthe processes involved in the explicit false recollec-tion of abstract perceptual patternsmdashlike that ofconceptual informationmdashmay also be impaired inamnesics These findings are thus consistent withour earlier findings using perceptually relatedwords (Schacter et al 1997) and with other recentreports indicating that explicit memory of amnesicsis impaired not only for tests that emphasise con-ceptual processing such as free or cued recall butalso on tests that draw more extensively upon per-ceptual processes such as graphemic cued recall(Cermak Verfaellie amp Chase 1995) and wordfragment cued recall (Vaidya Gabrieli Keane ampMonti 1995) That in the present study this per-ceptual memory impairment should assume theapparently paradoxical guise of fewer false recogni-tion errors in amnesic individuals than in controlparticipants shows yet again the versatility andgenerality of the strategy of using false recollectionto uncover the nature of the retrieval and encodingprocesses operating in veridical memory

REFERENCES

Arndt J amp Hirshman E (1998) True and false rec-ognition in MINERVA2 Explanations from a globalmatching perspective Journal of Memory and Lan-guage 39 371ndash391

Blaxton TA (1989) Investigating dissociations amongmemory measures Support for a transfer-appropriateprocessing framework Journal of ExperimentalPsychology Learning Memory and Cognition 15657ndash668

Blaxton TA (1995) A process-based view of mem-ory Journal of the International Neuropsychological So-ciety 1 112ndash114

Brainerd CJ Reyna VF amp Kneer R (1995) Falserecognition reversal When similarity is distinctiveJournal of Memory and Language 34 157ndash185

Cermak LS Butters N amp Gerrein J (1973) Theextent of the verbal encoding ability of Korsakoff pa-tients Neuropsychologia 11 85ndash94

Cermak LS Verfaellie M amp Chase KA(1995) Implicit and explicit memory in amnesia Ananalysis of data-driven and conceptually driven pro-cesses Neuropsychology 9 281ndash290

Curran T Schacter DL Norman KA amp GalluccioL (1997) False recognition after a right frontal lobeinfarction Memory for general and specific informa-tion Neuropsychologia 35 1035ndash1049

Deese J (1959) On the prediction of occurrence ofparticular verbal intrusions in immediate recall Jour-nal of Experimental Psychology 58 17ndash22

Estes WK (1997) Processes of memory loss recov-ery and distortion Psychological Review 104148ndash169

Franks JJ amp Bransford JD (1971) Abstraction ofvisual patterns Journal of Experimental Psychology 9065ndash74

GalloDARoberts MJ amp Seamon JG (1997) Re-membering words not presented in lists Can we avoidcreating false memories Psychonomic Bulletin and Re-view 4 271ndash276

Henkel LA amp Franklin N (1998) Reality monitor-ing of physically similar and conceptually related ob-jects Memory and Cognition 26 659ndash673

Homa D Cross J Cornell D Goldman D ampSchwartz S (1973) Prototype abstraction and clas-sification of new instances as a function of number ofinstances defining the prototype Journal of Experi-mental Psychology 101 116ndash122

Jernigan TL Schafer K Butters N amp CermakLS (1991) Magnetic resonance imaging of alco-holic Korsakoff patients Neuropsychopharmacology 4175ndash186

Johnson MK Hashtroudi S amp Lindsay DS(1993) Source monitoring Psychological Bulletin114 3ndash28

Kensinger EA amp Schacter DL (this issue) Whentrue memories suppress false memories Effects ofageing Cognitive Neuropsychology 16 399ndash415

KnowltonBJMangels JA amp Squire LR (1996) Aneostriatal habit learning system in humans Science273 1399ndash1402

Knowlton BJ Ramus SJ amp Squire LR (1992) In-tact artificial grammar learning in amnesia Dissocia-

KOUTSTAAL ET AL


tion of classification learning and explicit memory forspecific instances Psychological Science 3 172ndash179

Knowlton BJ amp Squire LR (1993) The learning ofcategories Parallel brain systems for item memoryand category knowledge Science 262 1747ndash1749

Knowlton BJ amp Squire LR (1994) The informa-tion acquired during artificial grammar learningJournal of Experimental Psychology Learning Memoryand Cognition 20 79ndash91

Knowlton BJ Squire LR amp Gluck MA(1994) Probabilistic classification learning in amne-sia Learning and Memory 1 106ndash120

Kolodny JA (1994) Memory processes in classifica-tion learning An investigation of amnesic perfor-mance in categorization of dot patterns and artisticstyles Psychological Science 5 164ndash169

Koutstaal W amp Schacter DL (1997) Gist-basedfalse recognition of pictures in older and youngeradults Journal of Memory and Language 37 555ndash583

Koutstaal W Schacter DL Galluccio L amp StoferKA (in press) Reducing gist-based false recogni-tion in older adults Encoding and retrieval manipula-tions Psychology and Aging

Koutstaal W Schacter DL amp Brenner C(1998) Dual task demands and gist based false recogni-tion of pictures in younger and older adults Manuscriptsubmitted for publication

Kroll NEA Knight RT Metcalfe J Wolf ES ampTulving E (1996) Cohesion failure as a source ofmemory illusions Journal of Memory and Language35 176ndash196

McClelland JL (1995) Constructive memory andmemory distortions A parallel distributed processingapproach In DL Schacter JT Coyle GDFischbach MM Mesulam amp LE Sullivan (Eds)Memory distortion How minds brains and societies re-construct the past (pp 66ndash90) Cambridge MA Har-vard University Press

McDermott KB (1996) The persistence of falsememories in list recall Journal of Memory and Lan-guage 35 212ndash230

McDermott KB amp Roediger HL III (1998) At-tempting to avoid illusory memories Robust falserecognition of associates persists under conditions ofexplicit warnings and immediate testing Journal ofMemory and Language 39 508ndash520

Moscovitch M (1994) Memory and working withmemory Evaluation of a component process modeland comparisons with other models In DL Schacteramp E Tulving (Eds) Memory systems 1994 (pp269ndash310) Cambridge MA MIT Press

Nosofsky RM (1988) Exemplar-based accounts ofrelations between classification recognition and typ-icality Journal of Experimental Psychology LearningMemory and Cognition 14 700ndash708

Nosofsky RM amp Zaki SR (1998) Dissociationsbetween categorisation and recognition in amnesicand normal individuals An exemplar-based interpre-tation Psychological Science 9 247ndash255

Omohundro J (1981) Recognition vs classification ofill-defined category exemplars Memory and Cogni-tion 9 324ndash331

Parkin AJ amp Leng NRC (1993) Neuropsychologyof the amnesic syndrome Hillsdale NJ LawrenceErlbaum Associates Inc

Payne DG Elie CJ Blackwell JM amp NeuschatzJS (1996) Memory illusions Recalling recognis-ing and recollecting events that never occurred Jour-nal of Memory and Language 35 261ndash285

Posner MI Goldsmith R amp Welton KE(1967) Perceived distance and the classification ofdistorted patterns Journal of Experimental Psychology73 28ndash38

Posner MI amp Keele SW (1968) On the genesis ofabstract ideas Journal of Experimental Psychology 77353ndash363

Read JD (1996) From a passing thought to a falsememory in 2 minutes Confusing real and illusoryevents Psychonomic Bulletin and Review 3 105ndash111

Reinitz MT Verfaellie M amp Milberg W(1996) Memory conjunction errors in normal andamnesic subjects Journal of Memory and Language 35286ndash299

Reyna VF amp Brainerd CJ (1995) Fuzzy-trace the-ory An interim synthesis Learning and IndividualDifferences 71 1ndash75

Robinson KJ amp Roediger HL III (1997) Associa-tive processes in false recall and recognition Psycho-logical Science 8 231ndash237

Roediger HL III (1996) Memory illusions Journalof Memory and Language 35 76ndash100

Roediger HL III amp McDermott KB (1995) Cre-ating false memories Remembering words not pre-sented in lists Journal of Experimental PsychologyLearning Memory and Cognition 21 803ndash814

Rosenthal R amp Rosnow RL (199 l) Essentials of be-havioral research Methods and data analysis NewYork McGraw-Hill

Rudy JW amp Sutherland RJ (1994) The mem-ory-coherence problem configural associations andthe hippocampal system In DL Schacter amp E



Tulving (Eds) Memory systems 1994 (pp 119ndash146)Cambridge MA MIT Press

Schacter DL (1995) Memory distortion Historyand current status In DLSchacter JT Coyle GDFischbach MM Mesulam amp LE Sullivan (Eds)Memory distortion How minds brains and societies re-construct the past (pp 1ndash43) Cambridge MA Har-vard University Press

Schacter DL (1997) False recognition and the brainCurrent Directions in Psychological Science 6 65ndash70

Schacter DL Church B amp Bolton E (1995) Im-plicit memory in amnesic patients Impairment ofvoice-specific priming Psychological Science 6 20ndash25

Schacter DL Norman KA amp Koutstaal W(1998a) The cognitive neuroscience of constructivememory Annual Review of Psychology 49 289ndash318

Schacter DL Verfaellie M amp Anes MD(1997) Illusory memories in amnesic patients Con-ceptual and perceptual false recognition Neuro-psychology 11 331ndash342

Schacter DL Verfaellie M Anes MD amp RacineC (1998b) When true recognition suppresses falserecognition Evidence from amnesic patients Journalof Cognitive Neuroscience 10 668ndash679

Schacter DL Verfaellie M amp Pradere D(1996) The neuropsychology of memory illusionsFalse recall and recognition in amnesic patients Jour-nal of Memory and Language 35 319ndash334

Shapiro ML amp Olton DS (1994) Hippocampalfunction and interference In DL Schacter amp ETulving (Eds) Memory systems 1994 (pp 87ndash117)Cambridge MA MIT Press

Shiffrin RM Huber DE amp Marinelli K(1995) Effects of category length and strength onfamiliarity in recognition Journal of ExperimentalPsychology Learning Memory and Cognition 21261ndash287

Snedecor GW amp Cochran WG (1989) Statisticalmethods (8th ed) Ames IA Iowa State UniversityPress

Solso RL amp Raynis SA (1979) Prototype forma-tion from imaged kinesthetically and visually pre-sented geometric figures Journal of ExperimentalPsychology Human Perception and Performance 5701ndash712

Squire LR (1994) Declarative and nondeclarativememory Multiple brain systems supporting learningand memory In DL Schacter amp E Tulving (Eds)Memory systems 1994 (pp 203ndash231) CambridgeMA MIT Press

Squire LR amp Knowlton BJ (1995) Learning aboutcategories in the absence of memory Proceedings of theNational Academy of Sciences USA 92 12470ndash12474

Sutherland RJ amp Rudy JW (1989) Configural as-sociation theory The role of the hippocampal forma-tion in learning memory and amnesia Psychobiology17 129ndash144

Vaidya CJ Gabrieli JDE Keane MM amp MontiLA (1995) Perceptual and conceptual memoryprocesses in global amnesia Neuropsychology 9580ndash591

Vogt J amp Kimble GA (1973) False recognition as afunction of associative proximity Journal of Experi-mental Psychology 99 143ndash145

KOUTSTAAL ET AL




APPENDIX

Novel-Corrected True and False Recognition Data for Individual PatientsTrue Recognition False Recognition

mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashndash mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashndashCategory Size Distance Category Size Distance

mdashmdashmdashmdashmdashmdashmdash mdashmdashmdashmdashmdashmdashndash mdashmdashmdashmdashmdashmdashmdash mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashNear Proto- Near

Patient Aetiology Larger Smaller Middle Far Larger Smaller type Middle Far Novel

CW Thalamic infarct 42 39 47 25 14 11 25 25 - 08 42bilateral anterior nucleia

AB Encephalitis mild 00 14 06 17 - 11 - 08 - 17 - 06 - 28 50diffuse cortical atrophyb

PD Anoxia enlarged 50 53 45 67 39 47 33 56 44 00ventricles cortical atrophyb

JM Anoxia bilateral 03 08 08 14 25 00 14 25 - 08 42hippocampal complexb

PS Anoxia bilateral 03 00 - 03 03 - 14 - 03 - 08 08 - 42 75hippocampal complexb

RL Anoxia 00 06 06 00 06 00 00 06 00 00

Patient Mean 16 20 18 21 10 08 08 19 - 07 35Control Mean 46 30 48 28 34 07 45 34 07 28

AA Korsakoff 33 19 28 00 28 20 22 22 33 33LB Korsakoff 39 45 39 39 22 33 39 17 28 50PB Korsakoff 11 14 11 28 06 - 11 28 11 - 17 17RD Korsakoff 03 03 03 - 03 25 00 20 14 31 58WR Korsakoff 61 28 72 33 33 11 22 39 11 00RM Korsakoff 45 17 39 28 28 - 08 28 17 06 50

Patient Mean 32 21 32 21 24 07 26 20 15 35Control Mean 35 20 36 21 17 06 25 21 - 01 47

Means for prototypes are based on items from category sizes of three six and nine items larger = average of six- and nine-itemcategories smaller = average of one- and three-item categories False alarms for novel category items are shown at the far rightall scores shown here have been corrected for these baseline differences in false alarms Anatomical lesion information isprovided where available

a Lesions assessed by CTb Lesions assessed by MRI

anoxia encephalitis Korsakoffrsquos syndrome) showmarked impairments in long-term memory fornewly encountered facts and events yet retain per-ceptual and linguistic functions within the normalrange (Parkin amp Leng 1993Squire 1994) Impor-tantly these studies have revealed that under someconditions amnesics may prove to be less suscepti-ble to illusory recollection than are individuals withintact memory with their impairment in veridicalmemory in some way buffering them against theerrors of illusory memory shown by control partici-pants In other instances however the reverse hasbeen found Under some conditions amnesics haveshown increased rates of false recognition provingto be more prone to such errors than are individualswith unimpaired memory

One significant factor that may mediate whichof these two patterns is observedmdashdepressed versuselevated false recognition among amnesics relativeto controlsmdashis the extent to which the nonstudiedlure items share semantic or conceptual featureswith previously encountered stimuli Several recentstudies have examined false recognition in amnesiaunder conditions where multiple items that aresemantically or conceptually related to a particulartheme word are presented (Schacter Verfaellie ampAnes 1997 Schacter Verfaellie Anes amp Racine1998b Schacter Verfaellie amp Pradere 1996) Akey observation from these studies is that underthese conditionsmdashwhere multiple items mightserve to reinforce the conceptual or semantic gist ofthe listmdashamnesic patients show both impairmentin correct recognition and decreased false recogni-tion For example Schacter et al (1996) comparedthe recognition performance of amnesics and con-trols in the converging verbal associates paradigmdeveloped by Deese (1959) and Roediger andMcDermott (1995) In this paradigm participantsare presented several lists of words where each ofthe words in a given list is associatively related to asingle (non presented) ldquocritical lurerdquo (eg partici-pants might be asked to remember the words bedrest awake tired dream etc without being pre-sented the word sleep the critical lure to which eachof these list words is associated) As expected theaverage rate of veridical recognition of control par-ticipants (83) substantially exceeded that of

amnesics (46) however amnesics also showedimpaired false recognition Whereas controls falselyrecognised 89 of the critical lures amnesicsrsquo falserecognition rate was 57 Largely similar findingswere reported by Schacter et al (1998b) and bySchacter et al (1997) In the latter study using dif-ferent word lists comprised of fewer semantic asso-ciates amnesics showed impaired true recognitionin two experiments and impaired false recognitionin one experiment (Expt 1) but not in another(Expt 2) however in the second experiment con-trol participants also showed particularly low levelsof false recognition

These observations of impaired false recognitionin amnesics under conditions where multiplesemantic associates are presented stand in markedcontrast to an early report indicating that com-pared to non-memory-impaired controls amnesicsmight show elevated rates of false alarms under con-ditions where the lures are related to only one of thestudied items (rather than multiple study items) Inan early study Cermak Butters and Gerrein(1973) found that when only a single related itemwas presented in a continuous recognition para-digm Korsakoff amnesics were more likely thanalcoholic controls to show false recognition of newitems that were similar to previously presenteditems Compared to controls Korsakoff amnesicsshowed significantly greater false recognition ofassociatively related words (eg study table testedwith chair) and also of homonyms (eg bearndashbare)amnesics also showed a nonsignificant tendency tomore often false alarm to synonyms (robberndashthief)and to unrelated items than did controls Morerecently in the study by Schacter et al (1996)amnesic patients demonstrated both a pattern ofldquoimpairedrdquo false recognition of semantically relateditems for which many study items were presentedand increased baseline rates of false alarms for itemsthat were unrelated to study items

What might account for these differing pat-terns Schacter et al (1996) suggested that whenmultiple related associates are presented false rec-ognition among the control participants dependson memory for the associative or semantic infor-mation that also supports veridical memory Theyproposed that because control participants were

KOUTSTAAL ET AL














KOUTSTAAL ET AL





METHOD

Participants






Design



KOUTSTAAL ET AL



WMS-Re







RL Anoxia 68 18 103 68 66 93Mean 548 153 980 747 600 955




Stimuli









KOUTSTAAL ET AL








Procedure







RESULTS


Combined Groups



KOUTSTAAL ET AL






















KOUTSTAAL ET AL























Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00


















KOUTSTAAL ET AL





METHOD

Participants






Design



KOUTSTAAL ET AL



WMS-Re







RL Anoxia 68 18 103 68 66 93Mean 548 153 980 747 600 955




Stimuli









KOUTSTAAL ET AL








Procedure







RESULTS


Combined Groups



KOUTSTAAL ET AL






















KOUTSTAAL ET AL























Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00










KOUTSTAAL ET AL





METHOD

Participants






Design



KOUTSTAAL ET AL



WMS-Re







RL Anoxia 68 18 103 68 66 93Mean 548 153 980 747 600 955




Stimuli









KOUTSTAAL ET AL








Procedure







RESULTS


Combined Groups



KOUTSTAAL ET AL






















KOUTSTAAL ET AL























Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00









METHOD

Participants






Design



KOUTSTAAL ET AL



WMS-Re







RL Anoxia 68 18 103 68 66 93Mean 548 153 980 747 600 955




Stimuli









KOUTSTAAL ET AL








Procedure







RESULTS


Combined Groups



KOUTSTAAL ET AL






















KOUTSTAAL ET AL























Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00








Design



KOUTSTAAL ET AL



WMS-Re







RL Anoxia 68 18 103 68 66 93Mean 548 153 980 747 600 955




Stimuli









KOUTSTAAL ET AL








Procedure







RESULTS


Combined Groups



KOUTSTAAL ET AL






















KOUTSTAAL ET AL























Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00






Stimuli









KOUTSTAAL ET AL








Procedure







RESULTS


Combined Groups



KOUTSTAAL ET AL






















KOUTSTAAL ET AL























Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00






KOUTSTAAL ET AL








Procedure







RESULTS


Combined Groups



KOUTSTAAL ET AL






















KOUTSTAAL ET AL























Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00











Procedure







RESULTS


Combined Groups



KOUTSTAAL ET AL






















KOUTSTAAL ET AL























Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00








RESULTS


Combined Groups



KOUTSTAAL ET AL






















KOUTSTAAL ET AL























Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00






















KOUTSTAAL ET AL























Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00











KOUTSTAAL ET AL























Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00





















Non-alcoholic 50 43 32 28 38 43 26a 07 06 20 56 50 28 13E-4 33


Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00






Subgroup Analyses






KOUTSTAAL ET AL



GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00







GENERAL DISCUSSION












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00












KOUTSTAAL ET AL













KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00

















KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00










KOUTSTAAL ET AL














KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00


















KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00










KOUTSTAAL ET AL









REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00













REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00







REFERENCES



















KOUTSTAAL ET AL

















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00





















































KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00























KOUTSTAAL ET AL




APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00








APPENDIX










RL Anoxia 00 06 06 00 06 00 00 06 00 00






Documents

PERCEPTUALLY BASED FALSE RECOGNITION OF NOVEL … · 2015-09-10 · also prompted investigations of such errors in persons with impaired memory (for review, see ... level of orthographic