Event-related potential evidence for a specific recognition - Brain

Brain (1998),121,1919–1935

Event-related potential evidence for a specificrecognition memory deficit in adult survivors ofcerebral hypoxiaAxel Mecklinger, D. Yves von Cramon and Gabi Matthes-von Cramon

Max Planck Institute of Cognitive Neuroscience, Leipzig, Correspondence to: Axel Mecklinger, Max Planck InstituteGermany of Cognitive Neuroscience, Inselstrasse 22–26, D-04103

Leipzig, Germany. E-mail: [email protected]

SummaryTransient global ischaemia due to cardiac arrest maylead to profound neuropsychological disorders. Recentresearch indicates that memory processes are particularlyimpaired after hypoxic brain injury. Visual recognitionmemory functions were examined in these patients bymeans of event-related potential (ERP) and performancedata. Eight chronic hypoxic patients, matched withcontrols for sex and age, performed a visual recognitionmemory task requiring recognition judgements for eitherobject forms or spatial locations and a visual classification(i.e. oddball) task that imposed negligible memorydemands. Reliable P300 oddball effects were obtainedboth for patients and for controls, whereas the two groupsdiffered in P300 latency and P300 scalp topography.In the memory task, old/new effects (i.e. larger ERP

Keywords: recognition memory; transient global ischaemia; event-related potentials; old/new effects; P300

Abbreviations: EOG 5 electrooculogram; ERP5 event-related potential; WMS-R5 Wechsler Memory Scale—revised

IntroductionTransient global ischaemia due to cardiac arrest may causeprofound neuropsychological disorders. The deficits includeimpairments in memory and executive functions (Volpeet al.,1986), visuospatial deficits (Wilson, 1996) and even severeintellectual or other widespread cognitive impairments(Parkin et al., 1987). Although cognitive abnormalities areoften defined simply as the inability to return to the prelesionoccupation, there is increasing evidence for the view thattransient global ischaemia leads to significant memoryimpairment and a variable degree of impairment of othercognitive abilities (Volpeet al., 1986; Kapur, 1988; Wilson,1996). Wilson (1996) examined a group of 18 patientssuffering from cerebral hypoxia and found a large variationin cognitive functioning, the largest group of patients (n 56) showing deficits in memory and executive functioning.Hopkinset al. (1995) reported severe memory disorders andintact attentional functions in three hypoxic patients aftercoma of 10–14 days.

© Oxford University Press 1998

waveforms for previously studied than for unstudieditems) were found for the controls. In contrast, in patientsthese old/new effects were absent or even inverted inpolarity while recognition performance was well abovechance level, except for one patient. These results suggestthat recognition, based on the retrieval of an item’s studyepisode, is degraded in patients who have suffered aperiod of transient global ischaemia. In the light of thepatients’ above-chance level of recognition performanceand the outcome ofpost hocanalysis of practice-relatedchanges in recognition performance, it is argued that thepatients’ memory disorders are best characterized as adegradation of explicit memory functions such as episodicretrieval of a study episode. Implicit functions such ascognitive skill learning were intact.

The functional characteristics of post-hypoxic amnesiawere examined in more detail by Volpeet al. (1986) (seealso Volpe and Petito, 1985; Hirstet al., 1986). These authorsexamined recognition and recall performance for high-frequency words in six patients with global hypoxic ischaemicinjury and a group of age-matched controls. The patients hadmemory impairments without loss of general intelligenceor sensorimotor functions, as revealed by the WechslerIntelligence Score and the Wechsler Memory Quotient.Contrasting memory performance in recognition and recalltests at variable study–test delays, the authors found similarforgetting rates in patients and controls in the recognitionmemory test, but a more pronounced decay in recallperformance in patients than in controls. A selective deficitin the patients’ recall performance was obtained even whenthey were given extra study time to attempt to equate theirperformance with that of the controls at the shortest study–

1920 A. Mecklingeret al.

test delay. Intact recognition memory after hypoxic braininjury due to cardiac arrest has also been reported byKapur (1988).

Selective decline in recall performance has been taken tosupport the view that recognition is reliant on perceptualprocesses that are preserved in post-hypoxic amnesic patients,whereas recall depends to a larger extent than recognition onother cognitive activities, such as the retrieval of an earlierstudy episode, which is selectively impaired in these patients.This view, however, is challenged by the fact that recognitionalso benefits from the processes assumed to contribute torecall performance. Also, the extent to which perceptualprocesses, such as judgements of an item’s familiarity,contribute to recognition performance is generally unclear.In support of this latter notion, Haistet al. (1992) found thatrecall and recognition were impaired to the same extent in agroup of 12 amnesic patients, some of whom became amnesicafter hypoxic brain injury. These results, in contrast to thoseof Volpe et al. (1996), suggest that the processes underlyingrecall and recognition are functionally related and reliant onthe integrity of the same brain structures (for similararguments see Moscovitch, 1992). Some evidence for thelatter view was also provided by a recent study (Knowltonand Squire, 1995) that examined recognition memory inamnesic patients using the R/K technique (Tulving, 1985).The subjects indicated whether they explicitly remembereda test item (R response) or simply knew that it was presentwithout conscious recollection (K response) of the studyepisode. The patients were similarly impaired in the twotypes of response, suggesting that the memory processescontributing to both response types depend on brain structuresdamaged in amnesia.

Although it is unclear what accounts for the differentialfindings with respect to recognition memory impairments inamnesic patients, one possibility is the differences in thepatients’ neuropathology. Aggleton and Shaw (1996)examined the extent of recognition memory deficits in 112amnesics reported in the literature by grouping the patientsaccording to their neuropathology. It was found that amnesicpatients with multiple sites of brain pathology including thefrontal lobes showed severe impairments on a standardrecognition memory test, whereas patients with focal brainlesions, including those with hippocampal ischaemia, wereonly mildly impaired in a recognition memory test comparedwith controls. This indicates that patient groups withhomogenous neuropathology and aetiology are required inorder to examine the functional characteristics of memorydisorders in amnesic patients.

Though transient global ischaemia is characterized by alarge variety of neuropathological changes and can lead tolocalized and non-localized lesion patterns, certain humanbrain structures have been found to be selectively vulnerableto the lack of blood and oxygen supply. Transient globalischaemia regularly gives rise to brain damage in thehippocampus, cerebellum, striatum and neocortex. In thecortex, damage is accentuated over the boundary zones of

the cerebral hemispheres (Brierleyet al., 1969; Cervo´s-Navarro and Diemer, 1991; Auer and Benveniste, 1997). Ina recent PET study (Kuwertet al., 1993), significantlyreduced glucose consumption was found in the medialtemporal cortex and thalamic projection zones in a group ofseven patients whose hypoxia was due to cardiac arrest. Nosignificant changes in glucose metabolism were obtained infrontal or parietal neocortical regions. Comparable metabolicchanges in the medial temporal cortex following transientglobal ischaemia have been reported by Ruprightet al.(1996).

More precise information on one crucial locus of braindamage was provided by post-mortem neuropathologicalanalyses (Zola-Morganet al., 1986; Rempel-Cloweret al.,1996). Patient R.B., reported by Zola-Morganet al. (1986),developed marked anterograde amnesia with little retrogradeamnesia after ischaemic hypoxia. Neuropathological analysisafter his death revealed selective neuronal loss in the CA1field of the hippocampus with only minor pathologicalchanges in other parts of the brain. A similar relationshipbetween selective damage to the CA1 field within thehippocampus and anterograde amnesia was reported forpatient G.D. after an assumed ischaemic episode due tocardiac arrhythmia (cf. Rempel-Cloweret al., 1996). Theseresults indicate that lesions to the hippocampus proper aresufficient to cause the memory impairments typically foundafter transient global ischaemia. Notably, like other amnesicpatients, R.B. showed impaired memory performance inexplicit memory tests, such as story recall or paired-associatelearning, in combination with intact general intellectualcapacities and implicit memory functions (i.e. word-stemcompletion). The view that ischaemic hypoxia leads tohippocampal damage and associated anterograde amnesia isalso supported by animal data, which show, for instance, thatexperimentally induced ischaemia in rats causes selectivedamage in the CA1 region of the hippocampus and deficits innew learning abilities (Volpeet al., 1984; Daviset al., 1986).

The aim of the present study was to examine the functionalcharacteristics of memory impairments by means of event-related potential (ERP) and performance measures in patientswho have suffered a period of transient global ischaemia.While previous studies have used different tasks (i.e. recalland recognition) to infer selective post-hypoxic memoryimpairments, our approach was to examine memoryimpairments in these patients within the same task, i.e. avisual recognition memory task. ERPs are small-voltageoscillations measured at the scalp that are time-locked to theprocessing of external events. They involve a sequenceof deflections (i.e. components) that mark the passage ofinformation through the brain. Differences in the timing andscalp topography of ERP components allow inferences to bemade about the temporal and spatial characteristics of brainactivity involved in cognitive processing (Hillyard and Kutas,1983; Mecklinger and Mu¨ller, 1996).

ERPs have been used extensively to examine recognitionmemory functions (for an overview see Rugg, 1995). Aconsistent finding of these studies is that repeated words or

Recognition memory and cerebral hypoxia 1921

figural stimuli evoke more positive ERPs than unrepeated(new) stimuli. These old/new effects start at ~300 ms andhave a broad temporal and topographical extent. Old/neweffects between 300 and 600 ms are also obtained acrossmodalities, i.e. when items are studied in the auditorymodality and tested in the visual modality (Feldsteinet al.,1987). They have also been found to be larger when olditems are correctly assigned to their study context than whenold items are incorrectly assigned (Wildinget al., 1995).Based on these findings, a number of authors have suggestedthat old/new effects are evoked when recognition isaccompanied by the retrieval of information formed at anitem’s initial presentation, i.e. conscious recollection (Smithand Halgren, 1989; Paller and Kutas, 1992; Rugg, 1995; butsee Potteret al., 1992). Since old/new effects in scalp-recorded ERPs are attenuated or diminished in patients withdamage to the mediobasal temporal lobes (Smith and Halgren,1989; Rugget al., 1991; Johnson, 1995), these effects canbe considered as an index of memory processes mediated bythese brain structures (cf. Wilding and Rugg, 1996). Giventhese functional and neuroanatomical characteristics, ERPold/new effects appear to be a valuable method of examiningthe nature of recognition memory impairments caused bytransient global ischaemia.

For the current study we employed a recently developedvisual recognition memory task (cf. Mecklinger, 1998;Mecklinger and Meinshausen, 1998) in which the processesunderlying recognition memory for object forms and objectlocations can be examined using the same study phase andtest stimuli for both types of information. Subjects studyfamiliar objects presented in various positions of a spatialmatrix. Just prior to the test phase, a cue is presentedindicating that recognition judgements for either object forms(object memory condition) or object locations (spatialmemory condition) will be required. Previous experimentsemploying this paradigm have shown that there is acontribution from working memory processes in this task, inthat rehearsal can be used to bridge the study–test intervals.(Mecklinger and Meinshausen, 1998). In the present study,eight patients and age- and sex-matched control subjectsperformed both memory tasks while ERPs were recorded from19 scalp sites. Two recognition conditions were employed inorder to examine whether potential recognition deficits areinformation-specific (i.e. confined to either object forms orspatial locations) or rather reflect a general deficit in visualrecognition memory functions.

Previous experiments employing this task used arandomized order presentation of the object and spatialconditions, such that subjects had to encode and rehearseboth kinds of information until the cue was presented (cf.Mecklinger, 1998). To adapt to the needs of the brain-injuredsubjects, this experiment used a block presentation of thetwo conditions. We assumed that the extent to which thepatients were able to retrieve either object forms or spatiallocations from the study episode would be indexed by reliableERP old/new effects. In contrast, the absence of old/new

effects in either condition would be indicative of a selectivedeficit in retrieving either object forms or spatial locationsfrom memory. Another modification concerned the delaybetween repetitions of items. This delay is usually not ofmajor relevance in ERP studies on recognition memory andcan vary from ~30 s (Rugget al., 1991) to several minutes(Paller and Kutas, 1992). To ensure that the patients wereable to perform the recognition task, we decided to use rathershort intervals (20 s) between repetitions of items. Previousstudies have shown that amnesic patients as well as monkeyswith medial temporal lobe lesions perform well in memorytasks with very short retention intervals (,5 s) but areimpaired at retention intervals ofù15 s (Sidmanet al., 1968;Alvarez-Royoet al., 1992). This had led to the view thatmemory tasks in which information has to be retained forø8 s mainly tap into short-term memory processes whereastasks with retention intervals ofù15 s can be considered aslong-term memory tasks (Alvarez-Royoet al., 1992). Giventhis, we considered a mean retention interval of 20 s as agood compromise that takes into account the patients’memory deficits but still examines the retrieval from episodiclong-term memory.

In order to ascribe possible changes in the ERPs evokedin the recognition task to memory deficits, ERPs were alsorecorded in a visual classification (i.e. oddball) task. In thistask, subjects were presented with a series of standard andrare target events. The subjects’ task was to count the raretargets and to ignore the standards. It has been shownrepeatedly that neurologically unimpaired subjects display alarge, parietally focused P300 component to rare target events(i.e. the P300 oddball effect). Comparing the P300 oddballeffect and the old/new effects in the recognition memorytask in the patients and their age-matched controls wouldenable us to examine the extent to which transient globalischaemia has a general effect on ERP components evokedin cognitive tasks or can be ascribed to a particular memorydeficit in these patients.

MethodSubjectsA group of eight patients, all of whom had suffered a periodof transient global ischaemia due to cardiac arrest, and 24controls participated in the experiment. The main clinicaldata and the medication applied at the time of the study arereported in Table 1A and B. The mean age was 42 years(range 19–60 years). Two female and six male right-handedpatients participated. A selection of relevant neuropsycho-logical data is displayed in Table 1C. Neuropsychologicaltesting of the eight patients revealed a broad variance incognitive performance. The difference between the WIP, areduced version of the Wechsler Intelligence Scale (Dahl,1996), and the Wechsler Memory Scale—revised score(WMS-R) provides an index of the severity of the patients’memory impairment. In normal subjects the two scores are


Table 1 Individual patient information for the eight hypoxic patients under investigation

Patient

22 32 92 51 14 2 67 160

(A) Main clinical dataAge (years) 48 46 46 19 60 43 39 37Sex M M M F M F M NCardiac arrest 1 1 1 1 1 1 1 1Structural/functional causes Toxic DCM CHD/MI EI CHD Toxic CHD/MI CHDImplementation of CPR Delayed Delayed Delayed Delayed Instantly Unknown Instantly DelayedTime since cardiac arrest (months) 9 12 36 5 9 31 6 5

(B) Medication (mg) at time of studyCitalopram ... ... ... 10 ... ... ... ...Paroxetine ... 30 ... ... ... ... ... ...Valproic acid ... ... ... 1200 ... ... ... ...Bromocriptine ... ... 10 ... ... ... ... ...Selectol ... ... ... ... 1 ... ... ...Pergolide 2 ... ... ... ... ... ... ...Sotalol ... ... 80 ... ... ... ... ...Enalapril 10 7.5 ... ... ... ... ... ...Captopril ... ... ... ... ... ... 60 ...Lisinopril ... ... ... ... ... ... ... 5

(C) Main neuropsychological dataWIP 130 82 61 85 125 100 90 95WMS-R 50 50 62 70 50 71 89 83Difference –80 –32 1 –15 –75 –29 –1 –12Digit span 7 10 6 6 6 9 6 6Visual span 12 8 6 6 8 6 10 5

Part A: CHD5 coronary heart disease; CPR5 cardiopulmonary resuscitation; DCM5 dilated cardiomyopathy; EI5 electrical injury;MI 5 myocardial infarction; toxic5 proarrhythmic drugs. Part C: Digit span, visual span5 raw scores of the forward versions from theWMS-R; WIP 5 reduced version of the Wechsler Intelligence Scale (Dahl, 1986); WMS-R5 Wechsler Memory Scale—revised.

equivalent. As is apparent from Table 1C, there was apronounced difference between intelligence and memoryscores in patients 22, 32, 51, 14, 2 and 160. Comparablescores, indicative of no selective memory impairments, wereobtained for patients 92 and 67. Notably, for patient 92 thelatter result arose from a low intelligence score, whereas forpatient 67 memory and intelligence scores were on a similar,high level. The scores for digit and visual span, indicativeof short-term storage capacity (Lezak, 1995), were between12 and 6 and thus within the normal limits for all except onepatient (patient 160). Patients 22, 32 and 92 were diagnosedas demented according to the guidelines of DSM IV (Sasset al., 1996).

There were 24 control subjects, all of whom were right-handed, with a mean age of 41 years (range 20–61 years).They were divided into four groups of six according to age,and each patient was then matched to one of these groups.Group A comprised six females with a mean age of 21 years(range 20–24 years) and served as a control group for patient51. Group B included six males (mean age 39 years; range34–43 years) and served as a control group for patients 2,67 and 160. Group C comprised six males with a mean ageof 47 years (range 43–53 years) and served as a controlgroup for patients 22, 32 and 92. Group D included six maleswith a mean age of 59 years (range 55–61 years) and served

as a control for patient 14. Each subject gave informedconsent prior to participation in the study, which was approvedby the ethics committee of the Max Planck Institute ofCognitive Neuroscience, Leipzig, Germany.

StimuliAll stimuli were presented on a 17-inch VGA monitor underthe control of a 486 computer. The stimuli of the visualoddball task were 16 geometrical figures (cross, circle,triangle, ring etc.). The stimuli in the memory task consistedof line drawings of 12 familiar and simple objects drawnfrom a standardized set of 260 line drawings (Snodgrass andVanderwart, 1980). The objects were spectacles, pipe, key,hammer, table, scissors, sock, book, lamp, envelope, cup andhat. All objects were presented in one of 12 equally spacedsquares of a 43 3 grid (dimensions 27.5 cm and 21.5 cm)in blue against a light grey background.

ProcedureEach subject was comfortably seated in an acoustically andelectrically shielded, dimly lit chamber 0.9 m from a computermonitor, and held a small response box on his or her lap.Each subject performed in one session, which included the


oddball task and the object and spatial recognition memorytests. In the oddball task 300 objects were presented in thecentre of the screen, each for 200 ms and with an interstimulusinterval of 1200 ms. Twenty-five per cent of the objects wereeasily discernible through an opening (e.g. a ring) and thesubject’s task was to count these particular objects. Next, therecognition memory tasks were performed, each of whichconsisted of 25 study–test blocks. In the study phases, fourline drawings of familiar objects were presented sequentiallywith a duration of 500 ms and an interstimulus interval of2000 ms at random positions of a 43 3 spatial matrix. Thewords ‘bitte warten’ (please wait) were presented for 3000 msat an interval of 1000 ms after the end of the study phase.Thereafter a cue was presented for 1000 ms, which indicatedwhether in the following test phase, which started 1500 msafter the cue had been removed, object-based or spatial-basedrecognition judgements were required. The cues ‘Objekte’and ‘Positionen’ were used to indicate the object and spatialrecognition memory condition, respectively. The test phasesin both conditions consisted of eight objects presentedsequentially for 1000 ms at random positions of the spatialmatrix. For each trial, in the object condition the subjectswere required to indicate whether or not the object currentlybeing presented was one already seen on the study list(irrespective of its spatial position in the spatial matrix),whereas in the spatial condition the subjects were requiredto indicate for each trial whether or not the object currentlybeing presented was occupying one of the positions in thespatial matrix seen in the study list (irrespective of theobject’s identity). In both conditions the same stimuli werepresented and subjects pressed the left button to respond‘old’ and the right button to respond ‘new’. Old and newresponses were equiprobable in the two conditions andfeedback was provided after each response (correct, false, noresponse). The maximum response time was 2500 ms. Withthis procedure the average delay between study items andtheir repetition in the test phases was 20 s. In the objectcondition, two of the four old/new objects were presented atpositions that also occurred in the study phase; the othertwo old/new objects were presented at unstudied positions.However, old objects were never presented at exactly thesame spatial positions as during the study phase. The sameconstraints were applied to the spatial condition. The fourstudy list items and the eight test items within each blockwere presented with the restriction that spatial positions inthe left and right half of the grid were equiprobable. Theobject condition was always performed first and four practiceblocks were given to the subjects at the beginning of boththe object condition and the spatial condition. Includingelectrode application and removal, each session lasted ~2.5 h.

ERP recordingThe EEG activity was recorded with tin electrodes mountedin an elastic cap (Electrocap International) from the 19electrode sites of the 10–20 system referenced to the left

mastoid. The ground electrode was positioned 10% of thenasion–inion distance anterior to Fz. The vertical electro-oculogram (EOG) was recorded from electrodes locatedabove and below the right eye. The horizontal EOG wasrecorded from electrodes positioned at the outer canthus ofeach eye. Electrode impedance was kept below 5 kΩ. TheEEG and EOG were recorded continuously in both taskswith a bandpass from DC to 70 Hz and were A–D convertedwith 16-bit resolution at a sampling rate of 250 Hz.

Data analysisERPs time-locked to rare and frequent objects (oddball task)and to correctly classified old and new items (memoryconditions) were computed for each subject at all recordingsites, with epochs extending from 200 ms before stimulusonset until 1000 ms thereafter. The average voltages in the200 ms preceding the test items served as a baseline, i.e. itsmean value was subtracted from each data point in thewaveforms. Prior to averaging, each epoch was scanned forEOG and other artefacts. Whenever the SD in a 200 ms timeinterval exceeded 50µV the epoch was rejected. The numberof rejected trials did not differ between patients and controls.The subject average ERPs were digitally low-pass filtered at12 Hz (cut-off frequency).

The P300 components in the oddball task were measuredas mean voltages in the 300–600 ms time intervals. The peaklatency was defined as the time point of the maximal positivedeflection within this time interval. For technical reasons onepatient (patient 14) did not perform the oddball task. Thus,the ERP analysis for this task will be restricted toseven patients and the respective three subgroups of controls.The time intervals for the quantification of the old/new effectswere determined after visual inspection of the ERP waveformsand were based on previous studies (Mecklinger, 1998;Wilding and Rugg, 1996). The time windows were 300–600 ms in the object condition and 300–500 ms in the spatialcondition. For statistical analysis of the ERP data, nineelectrodes, including the three midline sites Fz, Cz and Pzas well as bilateral frontal (F3, F4), central (C3, C4) andparietal (P3, P4) recording sites, were selected. The ERPdata were subjected to ANOVA (analysis of variance) withappropriate Huynh–Feldt corrections (Huynh and Feldt, 1970)for inappropriate degrees of freedom due to violations of thesphericity assumption. If not specified otherwise,post hoccomparisons were performed by means of a modifiedBonferroni procedure (Keppel, 1991)

ResultsOddball taskThe ERP waveforms evoked by target and standard stimulifor the hypoxic patients and all control subjects are displayedin Figs 1 and 2. One patient (patient 32) was unable to countthe target objects but could discriminate between target and


Fig. 1 Controls: oddball. Grand average ERPs for targets and standards in the visual oddball task. Thevertical lines indicate the onset of appearance of the objects, which were presented for 200 ms. Thevertical EOG (EOGV) is plotted in the upper left corner. Solid line5 target stimuli; broken line5standard stimuli.

standard objects. In both groups, the target stimuli evokedlarge P300 components which peaked earlier in the controlgroup than in the patient group. The mean latencies of thetarget P300 at the Pz electrode for the two groups were419 ms (SEM 5.47 ms) and 445 ms (SEM 7.99 ms),respectively. As revealed by at test for independent samples,this difference was highly significant [t(21) 5 –2.56,P ,0.01]. Though more pronounced in the control group, largerP300 amplitudes to targets than to standards (i.e. the P300oddball effect) were obtained in both groups.

These observations were confirmed by statistical analysis.In a first step we examined whether the waveforms evoked bytarget and standard stimuli were different in the three controlsubgroups. An ANOVA with the between-subject factorsubgroup (three levels) and the within-subjects factorselectrode (nine levels) and stimulus type (target versusstandard) revealed a main effect of stimulus type [F(1,15)5

103.50,P , 0.0001], and the interactions stimulus type3subgroup, [F(2,15)5 8.44,P , 0.003] and stimulus type3subgroup3 electrode [F(8,16)5 4.75,P , 0.004]. Based onthese interactions, separate ANOVAs (stimulus type3electrode) were performed for each of the subgroups. Stimulustype3 electrode interaction was found for the young controlsubjects (i.e. subgroup A) (P, 0.004) but not for the two othersubgroups (P . 0.18). This last result indicates that age hada systematic influence on the P300 scalp topography of thecontrols. In order to take these age-related changes in P300scalp topography into account in the comparison of P300 inpatients and controls, an analysis of covariance was performedwith group (patients versus controls), stimulus type andelectrode as factors and age as covariate. This analysis revealeda main effect of stimulus type [F(1,22)5 36.41,P , 0.0001],whereas the group3 stimulus type interaction [F(1,22) 51.70, P , 0.20] and the group3 stimulus type3 electrode


Fig. 2 Patients: oddball. Grand average ERPs for seven patients for targets and standards in the visualoddball task. For details see legend of Fig 1.

interaction [F(8,176) 5 1.30, P , 0.28] did not reach thesignificance level. These results suggest that the P300 oddballeffect was statistically not different for patients and controls.Interestingly, a group3 electrode interaction [F(8,176) 52.98, P , 0.02] was obtained, suggesting that P300 scalptopographies for both targets and standards were different forpatients and controls even when age effects on P300topography were controlled for. The group3 electrodeinteraction was also significant when between-groupdifferences in P300 amplitude were removed [F(8,176)5 3.44,P , 0.02] (cf. McCarthy and Wood, 1985), indicating thatthe P300 components in the patients and controls arise fromdifferent neuronal sources.

The different P300 scalp topography in the control groupsand the patients is further illustrated in Fig. 3, which displaysthe peak amplitude for the target P300 at the three midlineelectrodes for the patients and the subgroups of controls. Toallow a better evaluation of topographical changes, P300

amplitudes in this figure were normalized by converting allP300 amplitudes to the percentage of the P300 at the Pzelectrode (cf. Johnson, 1993). As is apparent from the figure,target P300s increased in amplitude from frontal to parietalrecording sites for the three control groups; this topographiceffect was substantially smaller for the two middle-agedgroups than for the young control group (i.e. subgroup ofpatient 51). Notably, while the largest P300s were obtainedat Pz for the three subgroups of controls, the P300 maximumfor all patients was shifted from the parietal (Pz) to central(Cz) or frontal (Fz) recording sites.

Memory taskPerformance measuresMean reaction times, the proportion of correct responses, hitrates and false alarm rates for the eight patients and the four


Table 2 Performance results (reaction times/proportion of correct responses) in both memory conditions for all eightpatients and the four control subgroups

Patient no. Patients Controls

Proportion correct Hits (%) FA (%) RT (ms) Proportion correct Hits (%) FA (%) RT (ms)

Object memory22 66* 66 35 138932 56 50 38 1044 91 90 7 78692 68* 74 39 133451 67* 61 27 1105 96 93 3 67614 63* 47 21 967 93 89 4 9232 92* 91 8 1269

67 76* 69 18 920 95 94 3 782160 88* 78 2 813

Spatial memory22 54 32 24 128632 47 32 38 1038 94 93 4 74592 64* 59 31 129651 87* 82 8 1063 98 99 2 54614 80* 71 12 870 93 90 4 8772 97* 71 12 1230

67 85* 85 15 693 96 95 2 700160 85* 78 8 905

FA 5 false alarm rate; Hits5 hit rate; RT5 reaction time. *Proportion correct differs significantly from chance performance (0.50).

Fig. 3 Amplitude measures for the P300 evoked by targets in theoddball task. The amplitude measures are plotted for patients 51,2, 67, 160, 22, 32 and 92 and the three age-matched subgroups ofcontrols (SG A, SG B, SG C). Note that P300 amplitude is scaledas percentage of the individual (patients) or group (controls) P300at the Pz electrode.

age-matched control groups are displayed in Table 2. In bothconditions the patients had delayed response times and fewercorrect responses than the control groups. For the eightpatients the mean proportion of correct responses was 72%in the object condition (controls, 94%) and 74% in the spatialcondition (controls, 95%). An ANOVA performed for reactiontimes with group (patients versus controls) as the between-subjects factor, condition (two levels) as the within-subjectsfactor and response type (two levels), revealed longer reactiontimes for the patients than the controls [F(1,30) 5 22.11,P , 0.001]. Responses were also faster in the spatial conditionthan in the object condition [F(1,30) 5 11.30,P , 0.002],and for old responses than for new responses [F(1,30) 530.10, P , 0.001]. The same analyses for performanceaccuracy revealed better performance for controls than for

patients [F(1,30) 5 44.18,P , 0.001]. Accuracy was alsohigher for old responses than for new responses [F(1,30) 536.03, P , 0.001]. Moreover, a group3 response typeinteraction [F(1,30)5 11.02,P , 0.002] was obtained.Posthoccomparisons indicated that for both groups the proportionof correct new responses was lower (and consequently falsealarm rates higher) in the object condition than in the spatialcondition (P , 0.07), whereas there were no differences inthe proportion of correct old responses (i.e. the hit rate)between the two conditions (P , 0.30).

In the light of the low performance of the patient groupwe examined whether the proportion of correct responses foreach patient was significantly different from chance. Valuesof χ2(1) were calculated for each patient and condition andcompared with theχ2(1) values atα 5 0.05. As is apparentfrom Table 2, the proportion of correct responses wassignificantly different from chance in both memory conditionsfor patients 67, 51, 14, 92, 2 and 160 and in the objectmemory task for patient 22, whereas the performance ofpatient 32 was not different from chance level in eithercondition.

ERP measuresThe ERP waveforms of the control subjects evoked by oldand new judgements in both tasks are displayed in Fig. 4(object condition) and Fig. 5 (spatial condition). Thecorresponding ERPs of the patients are shown in Figs 6 and7. Given that recognition memory performance was at chancelevel for patient 32 in both conditions and for patient 22in the spatial recognition condition, the respective ERP


Fig. 4 Controls: object condition. Grand average ERPs evoked by old and new objects at the nineelectrode sites also considered for statistical analysis. The vertical lines indicate the onset ofappearance of the objects which were presented for 200 ms. The vertical EOG (EOGV) is plotted inthe upper left corner. Solid line5 old; broken line5 new.

waveforms of these patients were excluded from Figs 6 and7 and from further analysis. For the controls, the waveformsin both conditions started to get more positive for oldresponses than for new responses at around 300 ms. Theseold/new effects amounted to 2–3µV and had a duration of~300 ms in the object condition and of 200 ms in the spatialcondition. While in the spatial condition the old/new effectswere largest at parietal recording sites, they were moreequally distributed across the scalp in the object condition.This latter result arose from a frontally focused negativityevoked by new objects peaking around 500 ms. In sharpcontrast, these old/new effects were virtually absent in thepatients’ ERPs.

The mean old/new effects for the patients and conditionsin which recognition was different from chance level and theircontrol groups in both recognition conditions are displayed in

Fig. 8. The old/new effects in the object condition wereeither absent (patients 14, 22, 67, 92 and 160) or of muchsmaller magnitude (patients 51 and 2) than those of thecontrols. A similar pattern of results was obtained for thespatial condition: here the old/new effects were virtuallyabsent (patients 67, 92 and 160) or substantially reduced inmagnitude (patients 14, 51 and 2) compared with those ofthe controls.

In parallel to the ERP analysis in the oddball task, we firstexamined whether the ERPs in the memory task were differentfor the four control subgroups. The mean amplitude measuresbetween 300 and 600 ms (object condition) and between 300and 500 ms (spatial condition) were used for quantificationof the old/new effects. The ANOVAs with the factorssubgroup (four levels), response type (two levels) andelectrode (nine levels) revealed main effects of response type


Fig. 5 Controls: spatial condition. Grand average ERPs evoked by old and new spatial locations at thenine electrode sites also considered for statistical analysis. For details see legend of Fig 4.

in the object condition [F(1,20) 5 43.51,P , 0.0001] andthe spatial condition [F(1,20)5 45.75,P , 0.0001] as wellas interactions among response type and electrode [objectcondition: F(8,24) 5 4.81, P , 0.02; spatial condition:F(8,24) 5 7.33, P , 0.001]. However, no interactionsinvolving the subgroup factor were obtained (P . 0.20). Inthe light of this result, the waveforms were collapsed acrossthe four control subgroups.

To examine whether object-based and spatial-based old/new effects differed in scalp topography, an ANOVA withthe factors condition (two levels) and electrode (nine levels)was performed on the differences between old and newresponses in the 300–600 ms (object condition) and the300–500 ms (spatial condition) time interval. This analysisrevealed a significant condition3 electrode interaction[F(8,184) 5 3.45, P , 0.03]. This interaction was alsosignificant [F(8,184) 5 2.89, P , 0.04] when the meanamplitude measures were normalized such that amplitude

differences between the two conditions were removed (cf.McCarthy and Wood, 1985) indicating that different neuronalstructures contribute to the old/new effects in both recognitionconditions (Johnson, 1993).

Contrasting the old/new effects of the controls (n 5 24)and the patients with above-chance recognition performancerevealed highly significant interactions between response typeand group for the object condition [F(1,29) 5 6.36, P ,0.02] and the spatial condition [F(1,28)5 13.53,P , 0.001],indicating that old/new effects were present for the controlsbut not for the patients.

DiscussionThis study examined visual recognition memory in a groupof eight hypoxic brain-injured patients and 24 age-matchedcontrols using ERPs and performance measures. All subjectsperformed a visual oddball task and two versions of a memory


Fig. 6 Patients: object condition. Grand average ERPs for old and new objects for the seven patientswho showed above-chance recognition performance in the object condition. For details see legend ofFig. 4.

task, in which recognition judgements were required eitherfor line drawings of familiar objects or their respective spatiallocations within a two-dimensional spatial matrix. The mainresults can be summarized as follows. First, for all patientsunder investigation reliable P300 oddball effects wereobserved. However, the P300 was prolonged and displayeda different scalp topography in the patients compared withthe controls. Secondly, in both recognition conditions thereaction times and accuracy were substantially degraded in thepatients compared with the controls, but, with the exception oftwo patients (both tasks for one patient and the spatialcondition for the other), recognition performance wassignificantly above chance level. Thirdly, the controlsdisplayed reliable ERP old/new effects for both types ofrecognition judgement, these effects being parietally focusedfor spatial-based judgements and more broadly distributedfor object-based judgements (see also Mecklinger and

Meinshausen, 1998). This difference in scalp topography canbe taken as evidence that different brain regions mediate theold/new effects in the two conditions. Interestingly, the old/new effects were virtually absent in both conditions for allpatients. These results will now be discussed with respect tothe functional characteristics of memory impairments afterischaemic hypoxic encephalopathy.

For the P300 component in the oddball task, differencesand similarities were found for patients and controls. Bothgroups showed significant oddball effects, i.e. larger P300sfor targets than for standards, these effects being statisticallyindistinguishable between the two groups. It has been arguedthat P300 amplitude is related to stimulus categorizationprocesses and reflects the degree to which a current modelof the environment is modified or updated once sensoryinformation has been analysed (Donchin and Coles, 1988;Mecklinger and Ullsperger, 1993). In oddball tasks, P300


Fig. 7 Patients: spatial condition. Grand average ERPs for old and new spatial locations for the sixpatients who showed above-chance recognition performance in the spatial condition. For details seelegend of Fig. 4.

latency is assumed to reflect the speed of these stimulus-related processes (Polich, 1991). In the light of these modelsthe present results suggest that the cognitive processesassociated with memory updating during visual classificationare not affected by transient global ischaemia. However,despite these similar oddball effects, P300 components weredelayed for ~50 ms in the patient group. Between-groupdifferences in P300 latency of similar magnitude have beenreported in studies examining P300 across the lifespan (Pictonet al., 1984; Pfefferbaumet al., 1984). For example, Pictonet al. (1984) found a delay of 70 ms in the auditory P300 ofsubjects aged 70 years compared those aged 20 years. DelayedP300 components in oddball tasks have also been found ina variety of ERP studies of demented patients (Goodinet al.,1978; Johnsonet al., 1991; Polich, 1991; Johnson, 1992).On the basis of these studies, it can be assumed that stimuluscategorization processes, though functionally comparable

with normal controls, are significantly delayed by ischaemic–hypoxic encephalopathy.

In addition to differing in latency, the P300 also differedin scalp topography between patients and controls. Consistentwith other studies, the target P300 in the control group waslargest at the parietal recording sites but acquired a morefrontal scalp distribution in subjects of increasing age (Pictonet al., 1984; Friedman and Simpson, 1994; Friedmanet al.,1997). In showing a central or frontocentral maximum, thepatients’ P300 was topographically clearly dissociable fromthe P300 of the control group. This topographical dissociationwas also obtained when the effects of age were taken intoaccount. Recent evidence from intracranial ERP recordings(Halgren et al., 1995a, b), from combined functional MRIand ERP recordings (Menonet al., 1997) and from ERPsrecorded from patients with circumscribed brain lesions(Knight et al., 1989; Verlegeret al., 1994; Knight, 1997)


Fig. 8 Mean amplitude measures of the old/new effects at the Czelectrode in the object and spatial conditions for patients showingabove-chance recognition performance and the correspondingmean values (11 SE) for the controls. The size of the old/neweffects of patients 22 and 32, which were excluded fromstatistical analysis, were 0.15µV (patient 22) and 2.0µV (patient32) (spatial condition) and 1.08µV (Patient 32) (objectcondition).

suggests that multiple brain regions contribute to thegeneration of the parietal maximal P300, some of theseregions being in the primary and secondary cortices. Thus,it is conceivable that the parietal attenuation of the patients’visual P300 arose from neuronal cortical alterations alongthe cortical arterial boundary zones in the posterior cortex,typical of those found in cerebral hypoxia (Cervo´s-Navarroand Diemer, 1991; Auer and Benveniste, 1997).

Even more pronounced differences between patients andcontrols were found in the memory task. The patients showeddegraded recognition performance that was significantlydifferent from chance in all but one patient. Their proportionof correct responses was 23% lower compared with thecontrols in both conditions, and the patients’ decrease in thenumber of hits was equivalent to their increase in falsealarms. Moreover, the patients, especially those who werediagnosed as demented according to DSM IV (i.e. patients22, 32 and 92) responded substantially more slowly and withlower accuracy than the age-matched controls (Table 2). Inshowing substantially degraded recognition performance inall patients the results are contrary to those reporting intactrecognition memory in amnesic patients. For instance, Volpeet al. (1986) found intact recognition performance for wordsin hypoxic brain-injured patients after a comparable shortdelay of 3 s. Similarly, Hirstet al. (1986) report intactrecognition performance in amnesic patients, even withretention intervals of 5 min. An explanation for thisdiscrepancy could be derived from procedural differences ofthe recognition tasks. Whereas in the studies of Volpeet al. (1986) and Hirstet al. (1986) trial-unique stimuli wereemployed, in the present experiment we used recurrentstimuli, i.e. the study stimuli were drawn from a fixed set.Consequently the latter procedure imposed more demandson working memory because previously studied stimuli aremore likely to generate interference on stimuli that are task-relevant in an actual trial. Given these procedural differences,i.e. the higher working memory demands imposed by the

necessity to inhibit task-irrelevant stimuli, it is conceivablethat the patients’ low recognition performance in the presentstudy resulted from a higher degree of susceptibility toproactive interference compared with the controls. (It couldalso be argued that elevated false alarm rates could beevidence for increased susceptibility to interference in thepatients. However, given that the subjects were providedwith feedback after each recognition judgement, it isconceivable that, besides the target status, the non-targetstatus of the items was also memorized. For this reason it ismore likely that susceptibility to interference resulted fromincreases in the numbers of both misses and false alarms.)We tentatively suggest that susceptibility to interferenceaccounts for the discrepancy of the patients’ recognitionmemory performance in the present experiment and the twostudies mentioned above.

The patients’ low recognition memory performance wasparalleled by reduced or virtually absent ERP old/new effectsfor both kinds of recognition judgement. Prior to discussingthese effects in the light of memory disorders in ischaemic–hypoxic patients, alternative interpretations will beconsidered. First, it could be argued that the patients show ageneral deficit in cognitive ERP components. Given theobservation that all patients under investigation showedreliable P300 oddball effects, the interpretation of a generaldeficit in cognitive ERP components appears to be unlikely.Secondly, it is conceivable that the deficit in recognitionaccuracy and the reduced ERP old/new effect reflect increasedrehearsal demands. It has been found as that old/new effectsget smaller, the more information has to be retained inmemory or with decreases in recognition performance(Mecklinger et al., 1992). However, digit spans and visualspans were within the normal limits except for one patient(patient 160) (Table 1C). Thus, short-term memory capacitycan be assumed to be sufficient to perform the recognitiontask. Moreover, there was no relationship between thepatients’ old/new effects and memory performance. Forexample, in patients 2 and 160, who showed the lowestimpairment in spatial memory performance, thecorresponding old/new effects were negligible (patient 2) oreven inverted in polarity (patient 160). Thus, neither a generaldeficit in cognitive ERP components nor limitations inworking (short-term) memory appear to be adequate toaccount for the absence of old/new effects in the patients’ERPs.

A memory-based interpretation of the degraded recognitionperformance and the absence of ERP old/new effects can bederived from so-called dual process models of recognitionmemory. According to these models an item can be recognizedas ‘old’ following the retrieval of an earlier study episode,i.e. an explicit memory phenomenon, or on the basis of itsfamiliarity or increased perceptual fluency, an implicitmemory phenomenon (Mandler, 1980; Jacobi and Dallas,1981). Based on the view that ERP old/new effects reflectbrain activity mediating the retrieval of an earlier studyepisode, i.e. one of the two modes by which recognition


judgements can be made (cf. Paller and Kutas, 1992; Rugg,1995), it can be assumed that the functional locus ofrecognition memory impairments in ischaemic–hypoxicpatients is in the retrieval of information from earlier studyepisodes. An objection to this interpretation might be thatthe present task contains a contribution from working (short-term) memory processes, because the task permits the use ofrehearsal between the study and the test phase such that theERP results cannot unequivocally be related to long-termepisodic memory retrieval. Although at present there is noobjective time standard that separates short-term memoryfrom long-term memory, two aspects provide argumentsagainst this objection. First, the mean delay between tworepetitions was 20 s. Previous studies have shown thatamnesic patients with lesions of the medial temporal lobeswho perform well in a memory task with very short retentionintervals, i.e. when information can be held in a short-termstore, show degraded memory performance when retentionintervals exceed 15 s (Sidmanet al., 1968; Cave and Squire,1992; see also Ringo, 1993). Given this, it was proposed thatin memory tasks with retention intervals.15 s mainly long-term memory processes mediated by medial temporal lobestructures have to be assumed. Thus, it is reasonable toassume that, in the present task, working memory processes(i.e. rehearsal) were probably restricted to the interval betweenthe study and test phases, and because of these rehearsalprocesses information entered episodic long-term memoryand was retrieved from there in the test phase. Secondly, theold/new effects in the control group as well as in previousstudies employing this paradigm (cf. Mecklinger, 1998;Mecklinger and Meinshausen, 1998) were closely similar intheir temporal and topographical characteristics to thosereported in previous ERP recognition memory experiments.This suggests that the present old/new effects, as in previousstudies, can be considered as an electrophysiological correlateof episodic memory retrieval.

The observation that old/new effects were equally degradedin both recognition conditions provides evidence against aselective recognition deficit for either object forms or objectlocations, and rather indicates that the brain regions mediatingthe retrieval of object forms and spatial locations are equallydamaged by an ischaemic–hypoxic encephalopathy. Theassumption that the conscious recollection of study episodesis impaired in ischaemic–hypoxic patients together with thepatients’ above-chance recognition performance also suggeststhat relying on the products of the retrieval processes is notalways necessary for correct recognition performance. Ratherthan solely supporting recognition judgements, consciouslyrecollected information of an earlier study episode presumablyserves as input for other evaluation or strategic controlsystems that operate on the products of the retrieval process(cf. Moscovitch, 1992).

Conversely, the patients’ above-chance recognitionperformance also implies that some memory processes notrequired for conscious remembering are intact after transientglobal ischaemia. The memory processes usually spared in

amnesic patients are considered implicit (non-declarative)memories. ‘Implicit memory’ refers to a heterogeneouscollection of abilities (cf. Squire, 1994). They enable skillacquisition and priming and are expressed implicitly byfacilitation in performance without the conscious recollectionof previous events (cf. Musen and Squire, 1991; Squire, 1994).To examine the extent to which facilitation in performance,indicative of spared implicit memories actually occurred, wecontrasted memory performance for the first and last five testblocks for patients and controls. If some form of implicitmemory remains intact after transient global ischaemia, wewould expect both patients and controls to show betterperformance in the last than in the first test blocks. In thispost hoc analysis only the performance data for thosepatients and conditions in which recognition performancewas significantly above chance were considered. For thepatients, recognition performance increased from 72 to 78%in the object condition and from 82 to 85% in the spatialcondition. As revealed byt tests for dependent samples, thisimprovement in performance was significant in the objectcondition [t(6) 5 2.79,P , 0.03] and marginally significantin the spatial condition [t(5) 5 1.91, P , 0.10]. Thecorresponding values for the control group were 91 and 95%(object condition) and 93 and 95% (spatial condition), thedifferences being significant in both the object task [t(23) 54.5, P , 0.001] and the spatial task [t(23) 5 2.15, P ,0.001]. This pattern of results suggests that repeated taskperformance led to an improvement in object recognitionand possibly in spatial recognition in the patients, andthat these facilitatory effects were very similar for patientsand controls.

To obtain an estimate of the potential contribution of short-term memory capacity to these practice-related changesin recognition performance, we examined the correlationsbetween the visual span scores from the WMS-R (Table 1C)and the performance differences between the first and lasttest blocks. Only those patients and those conditions in whichrecognition performance was significantly above chance wereconsidered in this analysis. No reliable correlations wereobtained (object condition,r 5 0.14; spatial condition,r 5–0.16,P . 0.47). Nor were there any reliable correlationsbetween the proportion of correct responses in both conditionsand the visual span scores (object condition,r 5 –0.18;spatial condition,r 5 0.29,P . 0.52). These results suggestthat the patients’ differential short-term memory capacities,as revealed by visual span scores, can account neither forpractice-related improvements in recognition performancenor for recognition memory performance in general.

Within the dual-process framework of recognition memory,facilitatory effects of similar kinds in amnesic patients havebeen interpreted as evidence for a spared familiarity orperceptual fluency component of recognition memory.However, preserved familiarity or perceptual fluency isunlikely to be the source of the present behavioural facilitationeffects, mainly because items from a fixed set wereinterchangeably used as targets and non-targets in different


blocks, such that familiarity with target stimuli could notbuild up across the experiment. The latter view is supportedby the observation that some forms of implicit memory inamnesic patients, such as the acquisition of reading skills orpriming of novel verbal stimuli, are item-specific with littleor no transfer to other items or task situations (Musenet al.,1990; Musen and Squire, 1991).

Given this, the behavioural facilitation effects seen in thepatients in the present study are more likely to reflect intactgeneral learning skills. Consistent with this view, a varietyof learning skills are known to be preserved in amnesicpatients. For example, amnesics show learning curves whenreading mirror-reflections of words that are closely similarto those of controls even for unique, non-repeated words(Cohen and Squire, 1980). Thus, it is conceivable that in thepresent experiment patients acquired skills for accessing andretrieving memory information that can be applied to multipleobject forms or object locations and that led to behaviouralfacilitation during recognition memory judgements. Thislatter process is apparently intact in ischaemic hypoxicpatients, whereas the retrieval of previous study episodes,which requires conscious access to a specific memory traceand which is indexed by ERP old/new effects, is degradedin these patients.

The notion that transient global ischaemia causes a selectivedeficit in explicit memory functions like the retrieval of anitem’s study context is also consistent with the cases of R.B.(Zola-Morganet al., 1986) and G.D. (Rempel-Cloweret al.,1996), who showed impairments in explicit memory functionsbut intact short-term memory and implicit memory functionssuch as word priming and cognitive skill learning following‘hypoxia’, with neuronal loss largely restricted to the CA1region of the hippocampus. Notably, in the temporal lobepatients examined by Smith and Halgren (1989) and by Rugget al.(1991) most of the hippocampus, including the head andthe anterior portion of the body, were removed, confirming theview that the integrity of the hippocampus proper seems tobe crucial for intact explicit memory functions (see alsoZola-Morgan and Squire, 1986).

In conclusion, the absence of ERP old/new effects duringvisual recognition memory in ischaemic–hypoxic patientssuggests that one functional locus of memory impairmentsin these patients is on the level of consciously accessingmemory-stored information from earlier study episodes.Practice-related changes in recognition memory performance,indicative of implicit memory processes that occur withoutawareness, are similar to those observed in controls and thusappear to be spared after ischaemic–hypoxic encephalopathy.In the light of their similarity to selective memory disordersafter focal damage to the mediobasal temporal lobes (Squireand Cohen, 1984) or the hippocampus proper (Zola-Morganet al., 1986), the present results support the view that ERPold/new effects can be considered as an index of the explicitmemory processes mediated by mediobasal temporalstructures.

AcknowledgementsWe wish to thank Heike Bo¨thel for his valuable help in dataacquisition and analysis, Erdmut Pfeifer for his technicalsupport in all phases of this project, and Peter Bublak forhelpful comments on earlier versions of the manuscript.

ReferencesAggleton JP, Shaw C. Amnesia and recognition memory: a re-analysis of psychometric data. Neuropsychologia, 1996; 34: 51–62.

Alvarez-Royo P, Zola-Morgan S, Squire LR. Impairment of long-term memory and sparing of short-term memory in monkeys withmedial temporal lobe lesions: a response to Ringo [comment].Behav Brain Res, 1992, 52: 1–5. Comment on: Behav Brain Res1991; 42: 123–34.

Auer RN, Benveniste H. Hypoxia and related conditions. In: GrahamDI, Lantos PL, editors. Greenfield’s neuropathology. 6th ed. London:Arnold; 1997. p. 263–314.

Brierley JB, Brown AW, Excell BJ, Meldrum BS. Brain damage inthe rhesus monkey resulting from profound arterial hypotension. 1.Its nature, distribution and general physiological correlates. BrainRes 1969; 13: 68–100.

Cave CB, Squire LR. Intact verbal and non-verbal short-termmemory following damage to the human hippocampus.Hippocampus 1992; 2: 151–63.

Cervos-Navarro J, Diemer NH. Selective vulnerability in brainhypoxia. [Review]. Crit Rev Neurobiol 1991; 6: 149–82.

Cohen NJ, Squire LR. Preserved learning and retention of pattern-analyzing skill in amnesia: dissociation of knowing how andknowing that. Science 1980; 210: 207–10.

Dahl G. WIP: Handbuch zum reduzierten Wechsler-Intelligenztest.Athenaum: Hain; 1986.

Davis HP, Tribuna J, Pulsinelli WA, Volpe BT. Reference andworking memory of rats following hippocampal damage inducedby transient forebrain ischemia. Physiol Behav 1986; 37: 387–92.

Donchin E, Coles MGH. Is the P300 component a manifestation ofcontext updating? Behav Brain Sci 1988; 11: 357–74.

Feldstein P, Smith ME, Halgren E. Cross modal repetition effects onthe N4. Fourth International Workshop on Cognitive Neuroscience.Paris, 1987.

Friedman D, Simpson GV. ERP amplitude and scalp distribution totarget and novel events: effects of temporal order in young, middle-aged and older adults. Brain Res Cogn Brain Res 1994; 2: 49–63.

Friedman D, Kazmerski V, Fabiani M. An overview of age-relatedchanges in the scalp distribution of P3b. [Review]. Electro-encephalogr Clin Neurophysiol 1997; 104: 498–513.

Goodin DS, Squires KC, Starr A. Long latency event-relatedcomponents of the auditory evoked potential in dementia. Brain1978; 101: 635–48.

Haist F, Shimamura AP, Squire LR. On the relationship betweenrecall and recognition memory. J Exp Psychol Learn Mem Cogn1992; 18: 691–702.


Halgren E, Baudena P, Clarke JM, Heit G, Lie´geois C, Chauvel P,et al. Intracerebral potentials to rare target and distractor auditoryand visual stimuli. I. Superior temporal plane and parietal lobe.Electroencephalogr Clin Neurophysiol 1995a; 94: 191–220.

Halgren E, Baudena P, Clarke JM, Heit G, Marinkovic K, DevauxB, et al. Intracerebral potentials to rare target and distractor auditoryand visual stimuli. II. Medial, lateral and posterior temporal lobe.Electroencephalogr Clin Neurophysiol 1995b; 94: 229–50.

Hillyard SA, Kutas M. Electrophysiology of cognitive processing.[Review]. Annu Rev Psychol 1983; 34: 33–61.

Hirst W, Johnson MK, Kim JK, Phelps EA, Risse G, Volpe BT.Recognition and recall in amnesics. J Exp Psychol Learn MemCogn 1986; 12: 445–51.

Hopkins RO, Gale SD, Johnson SC, Anderson CV, Bigler ED,Blatter DD, et al. Severe anoxia with and without concomitant brainatrophy and neuropsychological impairments. J Int NeuropsycholSoc 1995; 1: 501–9.

Huynh H, Feldt LS. Conditions under which mean square ratiosrepeated measurements designs have exact F distributions. J AmStat Assoc 1970; 65: 1582–9.

Jacoby LL, Dallas M. On the relationship between autobiographicalmemory and perceptual learning. J Exp Psychol Gen 1981; 110:306–40.

Johnson R Jr. Event-related brain potentials. In: Litvan I, AgidY, editors. Progressive supranuclear palsy: clinical and researchapproaches. New York: Oxford University Press, 1992. p. 122–54.

Johnson R Jr. On the neural generators of the P300 component ofthe event-related potential. Psychophysiology 1993; 30: 90–7.

Johnson R. Event-related potential insights into the neurobiologyof memory systems. In: Boller F, Grafman J, editors. Handbook ofneuropsychology. Vol. 10. Amsterdam: Elsevier; 1995. p. 135–63.

Johnson R Jr, Litvan I, Grafman J. Progressive supranuclear palsy:altered sensory processing leads to degraded cognition. Neurology1991; 41: 1257–62.

Kapur N. Memory disorders in clinical practice. London:Butterworths; 1988.

Keppel G. Design and analysis. 3rd ed. Englewoods Cliffs (NJ):Prentice-Hall; 1991.

Knight RT. Distributed cortical network for visual attention. J CognNeurosci 1997; 9: 75–91.

Knight RT, Scabini D, Woods DL, Clayworth CC. Contributions oftemporal–parietal junction to the human auditory P3. Brain Res1989; 502: 109–16.

Knowlton BJ, Squire LR. Remembering and knowing: two differentexpressions of declarative memory. J Exp Psychol Learn Mem Cogn1995; 21: 699–710.

Kuwert T, Homberg V, Steinmetz H, Unverhau S, Langen KJ,Herzog H, et al. Posthypoxic amnesia: regional cerebral glucoseconsumption measured by positron emission tomography. J NeurolSci 1993; 118: 10–6.

Lezak MD. Neuropsychological assessment. 3rd ed. New York:Oxford University Press; 1995.

Mandler G. Recognizing: the judgment of previous occurrence.Psychol Rev 1980; 87: 252–71.

McCarthy G, Wood CC. Scalp distribution of event-relatedpotentials: an ambiguity associated with analysis of variance models.Electroencephalogr Clin Neurophysiol 1985; 62: 203–8.

Mecklinger A. Remembering ‘what’ and ‘where’: an ERP study ofthe neurocognitive systems mediating spatial and object recognitionmemory. Hum Brain Mapp 1995; Suppl 1: 336.

Mecklinger A. On the modularity of recognition memory forobject form and spatial location: a topographic ERP analysis.Neuropsychologia 1998; 5: 441–60.

Mecklinger A, Meinshausen R. Recognition memory for objectform and object location: an event-related potential study. MemCogn. In press 1998.

Mecklinger A, Muller N. Dissociations in the processing of ‘what’and ‘where’ information in working memory: an event-relatedpotential analysis. J Cogn Neurosci 1996; 8: 453–73.

Mecklinger A, Ullsperger P. P3 varies with stimulus categorizationrather than probability. Electroencephalogr Clin Neurophysiol 1993;86: 395–407.

Mecklinger A, Kramer AF, Strayer DL. Event related potentials andEEG components in a semantic memory search task.Psychophysiology 1992; 29: 104–19.

Menon V, Ford JM, Lim KO, Glover GH, Pfefferbaum A. Combinedevent-related fMRI and EEG evidence for temporal-parietal cortexactivation during target detection. Neuroreport 1997; 8: 3029–37.

Moscovitch M. Memory and working-with-memory: a componentprocess model based on modules and central systems. J CognNeurosci 1992; 4: 257–67.

Musen G, Squire LR. Normal acquisition of novel verbal informationin amnesia. J Exp Psychol Learn Mem Cogn 1991; 17: 1095–104.

Musen G, Shimamura AP, Squire LR. Intact text-specific readingskill in amnesia. J Exp Psychol Learn Mem Cogn 1990; 16: 1068–76.

Paller KA, Kutas M. Brain potentials during memory retrievalprovide neurophysiological support for the distinction betweenconscious recollection and priming. J Cogn Neurosci 1992; 4:375–91.

Parkin AJ, Miller J, Vincent R. Multiple neuropsychological deficitsdue to anoxic encephalopathy: a case study. Cortex 1987; 23: 655–65.

Pfefferbaum A, Ford JM, Wenegrat BG, Roth WT, Kopell BS.Clinical application of the P3 component of event-related potentials.I Normal aging. Electroenceph Clin Neurophysiol 1984; 59: 85–103.

Picton TW, Stuss DT, Champagne SC, Nelson RF. The effects ofage on human event-related potentials. Psychophysiology 1984; 21:312–25.

Polich J. P300 in the evaluation of aging and dementia. [Review].Electroencephalogr Clin Neurophysiol Suppl 1991; 42: 304–23.

Potter DD, Pickles CD, Roberts RC, Rugg MD. The effects ofscopolamine on event-related potentials in a continuous recognitionmemory task. Psychophysiology 1992; 29: 29–37.


Rempel-Clower NL, Zola-Morgan SM, Squire LR, Amaral DG.Three cases of enduring memory impairment after bilateral damagelimited to the hippocampal formation. J Neurosci 1996; 16: 5233–55.

Ringo JL. Spared short-term memory in monkeys following medialtemporal lobe lesions is not yet established: a reply to Alvarez-Royo, Zola-Morgan and Squire. [Review]. Behav Brain Res 1993;59: 65–72.

Rugg MD. ERP studies of memory. In: Rugg MD, Coles MGH,editors. Electrophysiology of mind. Oxford: Oxford UniversityPress; 1995. p. 132–70.

Rugg MD, Roberts RC, Potter DD, Pickles CD, Nagy ME. Event-related potentials related to recognition memory. Brain 1991; 114:2313–32.

Rupright J, Woods EA, Singh A. Hypoxic brain injury: evaluationby single photon emission computed tomography. Arch Phys MedRehabil 1996; 77: 1205–8.

Sass H, Wittchen H-U, Zaudig M. Diagnostisches und StatistischesManual Psychischer Sto¨rungen. Go¨ttingen: Hogrefe; 1996.

Sidman M, Stoddard LT, Mohr JP. Some additional quantitativeobservations of immediate memory in a patient with bilateralhippocampus lesions. Neuropsychologia 1968; 6: 245–54.

Smith ME, Halgren E. Dissociation of recognition memorycomponents following temporal lobe lesions. J Exp Psychol LearnMem Cogn 1989; 15: 50–60.

Snodgrass JG, Vanderwart M. A standardized set of 260 pictures:norms for name agreement, image agreement, familiarity, and visualcomplexity. J Exp Psychol Learn Mem Cogn [Hum Learn] 1980;6: 174–215.

Squire LR. Declarative and nondeclarative memory: multiple brainsystems supporting learning and memory. In: Schacter DL, TulvingE, editors. Memory systems 1994. Cambridge (MA): MIT Press;1994. p. 203–31.

Squire LR, Cohen NJ. Human memory and amnesia. In: Lynch G,

McGaugh JL, Weinberger NM, editors. Neurobiology of learningand memory. New York: Guilford Press; 1984. p. 3–64.

Tulving E. Memory and consciousness. Can Psychol 1985; 26: 1–12.

Verleger R, Heide W, Butt C, Kompf D. Reduction of P3b potentialsin patients with temporo-parietal lesions. Brain Res Cogn BrainRes 1994; 2: 103–16.

Volpe BT, Petito CK. Dementia with bilateral medial temporal lobeischemia. Neurology 1985; 35: 1793–7.

Volpe BT, Pulsinelli WA, Tribuna J, Davis HP. Behavioralperformance of rats following transient forebrain ischemia. Stroke1984; 15: 558–62.

Volpe BT, Holtzman JD, Hirst W. Further characterization of patientswith amnesia after cardiac arrest: preserved recognition memory.Neurology 1986; 36: 408–11.

Wilding EL, Rugg MD. An event-related potential study ofrecognition memory with and without retrieval of source [publishederratum appears in Brain 1996; 119: 1416]. Brain 1996; 119:889–905.

Wilding EL, Doyle MC, Rugg MD. Recognition memory withand without retrieval of context: an event-related potential study.Neuropsychologia 1995; 33: 743–67.

Wilson BA. Cognitive functioning of adult survivors of cerebralhypoxia. Brain Inj 1996; 10: 863–74.

Zola-Morgan S, Squire LR. Memory impairment in monkeysfollowing lesions limited to the hippocampus. Behav Neurosci 1986;100: 155–60.

Zola-Morgan S, Squire LR, Amaral DG. Human amnesia and themedial temporal region: enduring memory impairment following abilateral lesion limited to field CA1 of the hippocampus. J Neurosci1986; 6: 2950–67.

Received February 5, 1998. Revised May 11, 1998.Accepted June 4, 1998

Documents

Event-related potential evidence for a specific recognition - Brain