Event-related potentials of emotional memory: Encoding pleasant, unpleasant, and neutral pictures

Copyright 2002 Psychonomic Society Inc 252

Cognitive Affective amp Behavioral Neuroscience2002 2 (3) 252-263

Emotional events tend to be remembered better thannonemotional events (Bradley Greenwald Petry amp Lang1992 Christianson 1992) The neural correlates of thismemory-enhancing effect of emotion and their temporalaspects in particular are not well understood Taking ad-vantage of the temporal resolution of event-related po-tentials (ERPs) we investigated the time course of theelectrophysiological correlates of encoding pleasant un-pleasant and neutral pictures

During the last decade the field of cognitive neuro-science of emotion has grown dramatically Most studieshave focused on the perception and evaluation of emotionalstimuli (emotional processing) and on the effects of emo-tion on memory formation (emotional memory) A criticaldistinction in this literature is the one between two affectivedimensions emotional arousal and emotional valenceArousal refers to a continuum that varies from calm to ex-citement whereas valence refers to a continuum that variesfrom pleasant to unpleasant with neutral as an interme-

diate value (see Bradley amp Lang 1994 for methods of as-sessing these two dimensions)

Different approaches ranging from behavioral and phar-macological to electrophysiological and functional neuro-imaging have tried to define the anatomical and functionalcorrelates of emotional processing and emotional mem-ory (Cahill 1996 Davidson amp Irwin 1999 Lane amp Nadel2000 LeDoux 1993 Phelps amp Anderson 1997) In thefollowing paragraphs we briefly review studies in thesetwo lines of research by presenting both general back-ground information and specific ERP evidence At the endof each section we highlight the shortcomings of previousERP studies and present the rationale for our approach

Neural Correlates of Emotional ProcessingLesion (Adolphs Tranel Damasio amp Damasio 1994

Bechara Damasio Damasio amp Lee 1999) electrophys-iological (Davidson amp Tomarken 1989 Wheeler David-son amp Tomarken 1993) and functional neuroimaging(Canli Desmond Zhao Glover amp Gabrieli 1998 David-son 1995 Davidson amp Irwin 1999) evidence supportsthe role of the prefrontal cortex (PFC) and the amygdalain the evaluation of the emotional content of stimuli Theroles of other structures such as ventral striatum anteriorcingulate posterior parietal and insula regions have beenalso explored (Davidson amp Irwin 1999 Lane amp Nadel2000) In the case of the PFC different hypotheses havebeen proposed concerning the involvement of differentPFC regions in emotional processes The right-hemisphere

This research was supported by HFMR (Alberta Canada) andNSERC (Canada) FD was supported by Chia PhD Scholarship andDissertation Fellowship from the University of Alberta (Canada) Wethank Annette Colangelo Sanda Dolcos and Eileen Noel for assistancewith data collection and analysis and Kevin LaBar Liz Phelps and twoanonymous reviewers for their comments on a previous version of themanuscript Correspondence concerning this article should be addressedto F Dolcos Center for Cognitive Neuroscience B203 LSRC DukeUniversity Durham NC 27708-0999 (e-mail fdolcosdukeedu)

Event-related potentials ofemotional memory Encoding pleasant

unpleasant and neutral pictures

FLORIN DOLCOSDuke University Durham North Carolina

and University of Alberta Edmonton Alberta Canada

and

ROBERTO CABEZADuke University Durham North Carolina

Emotional events tend to be remembered better than nonemotional events We investigated this phe-nomenon by measuring two event-related potential (ERP) effects the emotion effect (more positiveERPs for pleasant or unpleasant stimuli than for neutral stimuli) and the subsequent memory effect(more positive ERPs for subsequently remembered items than for subsequently forgotten items) ERPswere measured while subjects rated the emotional content of pleasant unpleasant and neutral picturesAs was expected subsequent recall was better for pleasant and unpleasant pictures than for neutral pic-tures The emotion effect was sensitive to arousal in parietal electrodes and to both arousal and valencein frontocentral electrodes The subsequent memory effect at centroparietal electrodes was greater foremotional pictures than for neutral pictures during an early epoch (400ndash600 msec) This result suggeststhat emotional information has privileged access to processing resources possibly leading to bettermemory formation

ERPS OF EMOTIONAL ENCODING 253

hypothesis proposes that the right hemisphere is special-ized for the perception expression and experience of emo-tion (eg Borod Koff amp Caron 1983 see also Borodet al 1998) The valence hypothesis proposes that the lefthemisphere is primarily associated with processing ofpleasant emotions whereas the right hemisphere is pri-marily associated with processing of unpleasant emotionsThis hypothesis is supported by electrophysiological(Davidson amp Tomarken 1989) and functional neuroimag-ing (Canli et al 1998 see also Davidson 1995 Davidsonamp Irwin 1999) evidence The amygdala has been stronglyassociated with emotional processing by both lesion(Adolphs et al 1994) and functional neuroimaging (Irwinet al 1996 Lane et al 1997 Morris et al 1996 Schnei-der et al 1997) studies For instance lesion studies showthat patients with bilateral amygdalar damage are im-paired in detecting emotional facial expression (Adolphset al 1994) and imaging studies on neurologically intactpeople have reported amygdalar activations associatedwith processing of both pleasant and unpleasant stimuli(Irwin et al 1996 Lane et al 1997 Morris et al 1996Schneider et al 1997)

ERP studies of emotion have identified an emotion ef-fect ERPs for emotional stimuli (pleasant or unpleasant)tend to be more positive-going than ERPs for neutral stim-uli (Carretie Iglesias amp Garcia 1997 Cuthbert SchuppBradley Birbaumer amp Lang 2000 Johnston Miller ampBurleson 1986 Naumann Maier Diedrich Becker ampBartussek 1997 Vanderploeg Brown amp Marsh 1987)The emotion effect has been found in different ERP com-ponents including the P300 component (eg Johnstonet al 1986 Naumann Bartussek Diedrich amp Laufer1992 see also Diedrich Naumann Maier amp Becker1997) the N300 component (Carretie Iglesias amp Garcia1997 Carretie Iglesias Garcia amp Ballesteros 1997) andthe slow wave (SW) component (Cuthbert et al 2000Diedrich et al 1997) but most consistently the emotioneffect is expressed in a P300ndashSW complex The P300component may be more sensitive to emotion under in-tentional emotional processing (eg Johnston et al 1986Naumann et al 1992) whereas the N300 component ap-pears to be more sensitive during incidental emotionalprocessing (Carretie Iglesias amp Garcia 1997 CarretieIglesias et al 1997) However the evidence is inconclu-sive (Naumann et al 1992) The neural generators of theemotion effect are not known but it is reasonable to as-sume that it reflects interactions between the amygdalaand cortical regions (Amaral Price Pitkanen amp Car-michael 1992 LaBar et al 1999 Lane et al 1997)

There are two main unsolved issues concerning theemotion ERP effect First it is unclear whether the emo-tion effect is sensitive only to arousal (emotional vs neu-tral) or to both arousal and valence (pleasant vs unpleas-ant) Whereas most studies have only found differencesowing to arousal (see Diedrich et al 1997 for a brief re-view) a few recent studies have also reported differencesthat could be attributed to emotional valence (Cuthbertet al 2000 Diedrich et al 1997) A possible explanation

of why previous studies have not found clear differencesowing to valence could be the use of word stimuli whichtend to generate smaller emotion effect overall (3ndash 4 m Vin amplitude eg Naumann et al 1997) as comparedwith more complex visual stimuli such as pictures (5ndash6 m V or more eg Cuthbert et al 2000) The failure toclearly dissociate the arousal and the valence could alsobe due to the fact that most studies either have comparedonly unpleasant and neutral items or have not equated thepleasant and the unpleasant stimuli for arousal Alsomany studies have used a relatively small number of stim-uli per condition (20 or fewer) which may not give enoughstatistical power for fine dissociations Second if theemotion effect is sensitive to both arousal and valence itis unclear whether these two dimensions can be dissoci-ated in their topographical andor temporal characteristicsThere is some evidence that valence-related differencestend to be larger in anterior electrodes (Cuthbert et al2000 Diedrich et al 1997) but most studies have em-ployed too few electrodes (eg nine) to allow a clear dis-sociation in topography

To explore these issues we investigated the effect ofarousal and valence on ERPs recorded from 32 electrodeswhile subjects rated the emotional content of 60 pleas-ant 60 unpleasant and 60 neutral pictures Pictures wereselected so that both pleasant and unpleasant pictures dif-fered from neutral pictures in terms of both arousal andvalence whereas pleasant and unpleasant pictures dif-fered from each other only in terms of valence

The Enhancing Effect ofEmotion on Memory Formation

Behavioral (Bradley et al 1992) pharmacological(Cahill amp McGaugh 1998 McGaugh Cahill amp Roozen-daal 1996) lesion (Adolphs Tranel amp Denburg 2000Cahill Babinsky Markowitsch amp McGaugh 1995 LaBaramp Phelps 1998) and neuroimaging (Cahill et al 1996Canli Zhao Brewer Gabrieli amp Cahill 2000 CanliZhao Desmond Glover amp Gabrieli 1999 Hamann ElyGrafton amp Kilts 1999) studies have suggested that thebeneficial effect of emotion on memory is related to botharousal and valence with arousal as the most critical fac-tor Pharmacological studies with animals have investi-gated the role of peripheral adrenergic systems and theirinteraction with other neurotransmitter systems (egcholinergic GABA-ergic opiod-peptidergic) in the amyg-dala (Cahill amp McGaugh 1998 McGaugh et al 1996)The results of these studies have suggested that theamygdala exerts its enhancing role on memory for emo-tionally arousing stimuli through the modulation of hip-pocampal activity (Cahill amp McGaugh 1998 McGaughet al 1996)

Lesion (Adolphs et al 2000 Cahill et al 1995 LaBaramp Phelps 1998) and functional neuroimaging (Cahillet al 1996 Canli et al 2000 Canli et al 1999 Hamannet al 1999) studies on humans have supported the roleof the amygdala in the formation of emotional memoryFor instance lesion studies have shown that patients with

254 DOLCOS AND CABEZA

bilateral amygdala damage do not remember emotionalstory segments better than the neutral ones (Cahill et al1995) and temporal-lobectomy patients unlike controlsubjects do not exhibit an increase in memory for arous-ing words over time (LaBar amp Phelps 1998) Also sub-jects with left-amygdalar damage are impaired in theirmemory for emotional stimuli despite entirely normalmemory for neutral stimuli (Adolphs et al 2000) Imag-ing studies using emotional stimuli have shown that thenumber of emotional items remembered was correlatedwith activity in the amygdalar (Cahill et al 1996 Hamannet al 1999) and hippocampal (Hamann et al 1999) re-gions Also a more recent event-related fMRI study (Canliet al 2000) showed that left-amygdalar activity during en-coding predicted memory for unpleasant but not for neu-tral pictures

In the ERP literature there are a considerable numberof studies on emotion (for reviews see Diedrich et al1997 Halgren amp Marinkovic 1995) and on memory (forreviews see Rugg 1995 Wagner Koutstaal amp Schacter1999) but very few of them have investigated the modula-tory effect of emotion on memory Some of the latter haveexamined the effect of emotion on memory retrieval (egMaratos Allan amp Rugg 2000 Maratos amp Rugg 2001Windmann amp Kutas 2001) but only a couple of studiesfocused on emotional encoding One study (Palomba An-grilli amp Mini 1997) measured ERPs during the encod-ing of emotional and neutral pictures and found that theamplitude of ERPs recorded from a parietal electrode (iePZ) was positively correlated with the number of subse-quently remembered slides Another study (LaBar et al1999) combined ERP and blood flow (fMRI) measure-ments during the encoding of unpleasant and neutral pic-tures and identified a neural network (including the pre-frontal parietal and occipitotemporal cortices) potentiallyinvolved in the encoding of unpleasant pictures

A limitation of these studies is that they did not usewhat is probably the most powerful method available toinvestigate the neural correlates of memory encoding thesubsequent memory paradigm In this paradigm encodingtrials are sorted according to whether the item presentedin each trial was remembered or forgotten in a subsequentmemory test (Brewer Zhao Desmond Glover amp Gabrieli1998 Paller Kutas amp Mayes 1987 Wagner et al 1998)ERPs for items that are subsequently remembered tendto be more positive-going than ERPs for items that aresubsequently forgotten (eg Fabiani Karis amp Donchin1986 Paller et al 1987 Sanquist Rohrbaugh Syndulkoamp Lindsley 1980 for reviews see also Rugg 1995 andWagner et al 1999) This difference is known as the sub-sequent memory effect (or Dm effect Paller et al 1987) Itcan occur over frontal midline or parietal locations andit can onset as early as 300ndash400 msec and can persist be-yond 1200 msec (Rugg 1995)

The main goal of the present study was to investigatethe relationship between the emotion effect and the sub-sequent memory effect When the memory advantage foremotional stimuli and the ERP positivity associated with

both the emotion effect and the subsequent memory effectare considered together it is reasonable to predict that thebeneficial effect of emotion on memory performancewill appear as an interaction between the emotion effectand the subsequent memory effect That is if emotionalstimuli are better remembered because they are better en-coded (as has been suggested by the aforementioned le-sion pharmacological and neuroimaging evidence) thesubsequent memory effect should be modulated by theemotional content of the encoded stimuli Yet the only ERPstudy to our knowledge that directly investigated the sub-sequent memory effect for emotional stimuli (LeiphartRosenfeld amp Gabrieli 1993) failed to find a subsequentmemory effect or a modulatory influence of emotion onthe subsequent memory effect This null finding could berelated to two aspects of this study First as was suggestedby the authors it could be related to the use of a shallowencoding task which contrasts with the use of deep en-coding tasks in other subsequent memory studies (egPaller et al 1987) Second the null finding could be re-lated to the use of words as stimuli As was mentionedabove words tend to elicit a weaker emotional effect thanpictorial stimuli do thereby reducing the chances of find-ing a significant emotion-related difference in the sub-sequent memory effect

Thus we investigated the modulatory influence of emo-tion on the subsequent memory effect using a deep en-coding task (pleasantness rating) and pictures as stimuliERPs were recorded while subjects rated random seriesof pleasant unpleasant and neutral pictures on a 5-pointpleasantness scale After the encoding phase the subjectsperformed a free recall task in which they wrote a de-scription for each picture they could remember EncodingERPs were separately averaged for pleasant unpleasantand neutral pictures and within each category for subse-quently remembered and forgotten items

We made two predictions concerning emotional pro-cessing and two predictions concerning emotional mem-ory Concerning emotional processing we expected thatthe emotion effect would be reflected mainly in the mod-ulation of the P300ndashSW complex and would be sensitiveto both arousal and valence and that valence-related dif-ferences would have an anterior distribution possiblyleading to clear topographical dissociation from arousal-related differences (Cuthbert et al 2000 Diedrich et al1997) Concerning emotional memory we predicted thatmemory for emotional pictures would be better than thatfor neutral pictures (Bradley et al 1992) and that con-sistent with this behavioral difference the subsequentmemory effect would be more pronounced for emotionalthan for neutral pictures

METHOD

SubjectsFifteen healthy right-handed female university students partici-

pated in the experiment in exchange for payment or credits Femalesubjects were chosen because previous studies using the same kindof stimuli have found that women show more physiological reac-


tivity with valence judgments than do men (Lang GreenwaldBradley amp Hamm 1993) and because women are more likely to re-port intense emotional experiences (Shields 1991) The experimen-tal data were collected with the understanding and written consentof each subject and the procedures used for data collection adheredto generally accepted practices for experimental research on humanparticipants

MaterialsStimuli consisted of a pool of 180 pictures (60 pleasant 60 un-

pleasant and 60 neutral) selected from the International AffectivePicture System (IAPS Lang Bradley amp Cuthbert 1997) IAPSpictures are rated on a 9-point scale in terms of both arousal (1 =calm and 9 = excited) and valence (1 = unpleasant 5 = neutral 9 =pleasant) using the Self-Assessment Manikin method (Bradley ampLang 1994) The average arousal and valence scores of the selectedpictures1 were respectively 60 and 71 for pleasant pictures 62and 23 for unpleasant pictures and 29 and 50 for neutral picturesPleasant and unpleasant pictures differed from each other in termsof emotional valence but not in terms of emotional arousalwhereas both pleasant and unpleasant pictures differed from neutralpictures in terms of both arousal and valence These two facts wereconfirmed by two analyses of variance (ANOVAs) First an ANOVAon the valence scores of pleasant unpleasant and neutral picturesyielded a significant main effect of valence [F(2118) = 12923p 0001] and post hoc tests showed significant differences be-tween the valence scores of all three categories (p 05) Second anANOVA on the arousal scores of pleasant unpleasant and neutralpictures yielded a signif icant main effect of arousal [F(2118) =12462 p 0001] and post hoc tests showed significant differencesbetween the arousal scores of both the pleasant and the unpleasantpictures and those of the neutral pictures (p 05) but no signifi-cant difference between the pleasant and the unpleasant pictures(p 05) The pictures were formatted so that on the screen theyhad a height of between 10 and 13 cm and a width of between 10 and17 cm The 180 pictures were divided into six sets of 30 pictures (10pleasant 10 unpleasant and 10 neutral) which were assigned to sixstudyndashtest blocks

ProcedureThe experiment was conducted on a PC using the SuperLab soft-

ware (Cedrus Inc) The experimental session consisted of a prac-tice session and six studyndashtest blocks Eight different block orderswere employed and the order of the pictures within each block wasrandomized for each subject Encoding trials consisted of four eventsblank (500 msec) cross fixation (500 msec) picture (2000 msec)and rating screen (until response) The subjects were instructed toexperience the feelings or thoughts elicited by each picture and torate the pictures according to their initial reaction using a 5-pointscale (1 = very unpleasant 2 = slightly unpleasant 3 = neutral 4 =slightly pleasant and 5 = very pleasant) The subjects were also in-structed to remember the pictures for a subsequent memory test (in-tentional learning) After each study phase the subjects were askedto recall as many pictures as possible during 6 min They were pro-vided with response sheets and were instructed to write down a de-scription for each picture using one line per picture They wereasked to provide enough details so that an outsider could identifyeach picture (eg diver vs skier) and could differentiate it fromsimilar pictures (eg platform diver vs springboard diver) Onlypictures whose descriptions were detailed enough to allow bothidentif ication and differentiation were classified as remembered

ERP RecordingEEG was recorded from 32 AgAgCl electrodes embedded in a

cap (Quikcap Neurosoft Inc Sterling VA) with respect to a linkedmastoid reference Recording sites were based on a variation of the1020 international electrode placement system (Jasper 1958) in-

cluding the following electrodes frontopolar (FP1 FP2) frontal(F3 F7 FZ F4 F8) frontocentral (FC3 FCZ FC4) central (C3 CZC4) centroparietal (CP3 CPZ CP4) parietal (P3 PZ P4) fronto-temporal (FT7 FT8) temporal (T7 T8) temporoparietal (TP7TP8) and occipital (O1 OZ O2) Horizontal eye movements weremonitored with bipolar electrodes placed symmetrically on the outercanthus of each eye and vertical movements were monitored withbipolar electrodes situated on the left supraorbital and infraorbitalridges All channels were amplified with a filter bandwith of 003ndash50 Hz and were sampled on line at an AD rate of 500 Hz Therecording epoch was 1500 msec beginning 100 msec prior to pictureonset Trials with values above 100 m V or below ndash100 m V possiblyowing to eye movement artifacts were eliminated from the analy-sis To ensure an adequate signal-to-noise ratio in the ERPs sub-jects who had fewer than 16 artifact-free trials per condition wereexcluded from analysis and were replaced (a total of 4 subjects wereexcluded and replaced)

RESULTS

Behavioral ResultsValence ratings The average valence scores of the pic-

tures as rated by the subjects were 411 for pleasant pic-tures (SD = 037) 147 for unpleasant pictures (SD =024) and 295 for neutral pictures (SD = 024 1 = very un-pleasant 3 = neutral and 5 = very pleasant) An ANOVAon the valence rating scores of pleasant unpleasant andneutral pictures yielded a significant main effect of valence[F(228) = 31248 p 0001] and post hoc tests showedsignificant differences between the valence scores of allthree categories (all ps 01) Thus our subjectsrsquo ratingscores were consistent with the standard valence scoresfrom IAPS (see the Method section)

Memory performance As was expected memoryperformance was affected by the emotional content of thepictures The proportion of pictures recalled was 57 forpleasant pictures (SD = 091) 58 for unpleasant pictures(SD = 095) and 48 for neutral pictures (SD = 078) Aone-way (emotion pleasant unpleasant and neutral)ANOVA yielded a significant main effect of emotion[F(232) =1644 p 0001] and post hoc contrastsyielded significant differences between pleasant andneutral pictures (p 05) and between unpleasant and neu-tral pictures ( p 05) but not between pleasant and un-pleasant pictures ( p 05) Thus memory performancewas higher for emotional than for neutral pictures withno difference between pleasant and unpleasant pictures

ERP ResultsTo investigate the emotion effect ERPs were sepa-

rately averaged for pleasant unpleasant and neutral pic-tures To investigate the modulatory influence of emo-tion on the subsequent memory effect ERPs in eachemotional category were further subdivided into ERPs forremembered and forgotten items Averaged data were an-alyzed using repeated measures ANOVAs and post hocTukeyndashKramer tests

Emotion effect The overall shape of encoding ERPswas similar for pleasant unpleasant and neutral picturesand it was characterized by a N200ndashN300ndashP300bndash


SW complex (Figure 1) As was expected there was anemotion effect ERPs for emotional pictures were morepositive-going than ERPs for neutral pictures Overallthe emotion effect reflected mainly differences in arousalwhich where most prominent in the amplitude of SWsand at parietal electrodes However during an earlierepoch (ie 500ndash800 msec) and at frontocentral electrodes(eg FCZ in Figure 2) the emotion effect was sensitivemainly to valence As Figure 2 suggests the emotion effectwas different at parietal versus frontocentral electrodes Atparietal electrodes the emotion effect appeared at thesame time (500 msec) for both categories of high-arousingpictures (ie pleasant and unpleasant) whereas at fronto-central areas it appeared earlier for pleasant pictures(500 msec) than for unpleasant pictures (after 800 msec)No noticeable valence-related hemispheric asymmetrywas present (Figure 1)

To test these ideas ANOVAs were computed on ERPsfrom PZ and FCZ at two time windows (see Figure 2)These locations and time windows were chosen on thebasis of previous evidence that showed a tendency for aposterioranterior topographical dissociation of thearousalvalence effects at similar locations and ERPcomponents (Cuthbert et al 2000)

To verify whether the emotion effect was different atparietal versus frontocentral electrodes at 500ndash800 mseca 2 (electrode site PZ vs FCZ) 3 3 (emotion pleasantunpleasant neutral) ANOVA was computed It yielded

significant main effects of both electrode site [F(114) =1625 p 0012] and emotion [F(228) = 2784 p 0001] and a significant electrode site 3 emotion inter-action [F(228) = 457 p 019] To elucidate this inter-action separate one-way (pleasant unpleasantneutral)ANOVAs were computed on PZ and FCZ At PZ theANOVA yielded a significant effect of emotion [F(228) =3102 p 0001] and the post hoc tests showed a signif-icant difference between pleasant pictures and neutral pic-tures ( p 05) and between unpleasant and neutral pic-tures ( p 05) but not between pleasant and unpleasantpictures ( p 05) At FCZ in contrast the ANOVAyielded a significant effect of emotion [F(228) = 1829p 0001] and post hoc tests showed a significant dif-ference between pleasant pictures and both unpleasantand neutral pictures ( ps 05) but not between un-pleasant and neutral pictures ( p 05) Thus the twoseparate ANOVAs confirmed a difference between emo-tional (both pleasant and unpleasant) and neutral picturesin parietal regions but a difference between pleasant andboth unpleasant and neutral pictures in frontocentral re-gions (compare FCZ and PZ graphs for the early epoch inFigure 2)

To verify whether the emotion effect at FCZ was differ-ent over time a 2 (epoch 500ndash800 vs 800ndash1200 msec) 33 (emotion pleasant unpleasant and neutral) ANOVAwas performed It yielded significant main effects ofboth epoch [F(114) = 5653 p 0001] and emotion

Figure 1 Emotion effect ERPs for both pleasant and unpleasant pictures were more positive-going than ERPsfor neutral pictures The electrode locations are topographically arranged as viewed from above the subjectsrsquoheads Vertical and horizontal scales represent the voltage amplitude (in microvolts) and the time from stimulusonset (in milliseconds) respectively


[F(228) = 226 p 0001] and a significant epoch 3emotion interaction [F(228) = 1028 p 0004] To fur-ther quantify these effects an additional one-way (emo-tion pleasant unpleasant and neutral) ANOVA was per-formed at the late epoch of FCZ It yielded a significantmain effect of emotion [F(228) = 2524 p 0001] andpost hoc tests showed a significant difference for bothpleasant and unpleasant pictures when compared withneutral pictures ( p 05) but no significant differencebetween pleasant and unpleasant pictures ( p 05)

Finally to check for hemispheric differences in theemotion effect 3 (emotion pleasant unpleasant and neu-tral) 3 2 (hemisphere left vs right) ANOVAs were com-puted on ERPs from left (F3 FP1) and right (F4 FP2)frontal locations The ANOVAs performed on ERPsfrom F3F4 did not yield a significant emotion 3 hemi-sphere interaction neither during an early epoch (500ndash800 msec) nor during a late epoch (800ndash1200 msec)suggesting that the emotion effect at these locations wassimilar across hemispheres However the same 3 3 2 de-sign computed on ERPs from FP1FP2 and during theearly epoch (500ndash800 msec) yielded a significant emo-

tion 3 hemisphere interaction [F(228) = 394 p 032]suggesting that the emotion effect was different acrosshemispheres during this early interval Follow-up ANOVAsand post hoc analyses showed that this interaction wasdue to a larger emotion effect for pleasant pictures in theleft hemisphere as compared with the right [F(114) =1123 p 005] Although this f inding is consistentwith the valence hypothesis of hemispheric asymmetry(Davidson 1995) because the analyses did not yieldanalogous results for unpleasant pictures (larger emotioneffect in the right hemisphere as compared with theleft) it does not provide clear ERP evidence for the va-lence hypothesis However the fact that this lateraliza-tion was seen in the frontal electrodes is consistent withthe proposed role of the frontal lobes in the processing ofemotional valence (Davidson 1995 Heller 1993)

Taken together these results show that the emotion ef-fect was different at parietal versus frontocentral locationsand suggest that at parietal locations the emotion effectwas sensitive mainly to arousal (pleasant = unpleasant neutral) whereas at frontocentral locations it was sensi-tive mainly to valence during an earlier epoch (pleasant

A FCZ

B PZ

Pleasant Unpleasant Neutral


20

0

ndash20

20

0

ndash20

0 500 1000 msec

0 500 1000 msec

V

V

Early epoch (500ndash800 msec) Late epoch (800ndash1200 msec)


V V

VV

6

5

4

3

2

1

0

14

12

10

8

6

4

2

0

10

8

6

4

2

0

10

8

6

4

2

0

Figure 2 The emotion effect was different at frontocentral electrodes than at parietal electrodes (A) At frontocentral elec-trodes it appeared earlier for pleasant pictures (500ndash800 msec) than for unpleasant pictures (after 800 msec) (B) At parietalelectrodes it appeared at the same time (500ndash800 msec) for both pleasant and unpleasant pictures Light gray pleasant darkgray unpleasant black neutral


unpleasant = neutral) and then later to overall arousal(pleasant = unpleasant neutral)

Subsequent memory effect As is illustrated by Fig-ure 3 the distribution of the subsequent memory effectwas different during an early epoch than during a lateepoch During an early epoch (400ndash600 msec) the sub-sequent memory effect at midline locations (eg CPZCZ FCZ) was greater for pleasant and unpleasant picturesthan for neutral pictures The difference was maximal atCZ and it can be clearly observed in the subtraction data(remembered 2 forgotten) depicted in Figure 4 Duringa late epoch (600ndash800 msec) however the distributionof the subsequent memory effect was more similar forpleasant unpleasant and neutral pictures It extended overcentroparietal and frontocentral regions and it was max-imal at midline electrodes In addition to the overall sim-ilarity in the general pattern of activation some differ-ences in the subsequent memory effect can also benoticed As Figure 3 suggests the subsequent memoryeffect at frontal electrodes seems to be bilateral for

pleasant pictures right-lateralized for unpleasant picturesand left-lateralized for neutral pictures

To investigate the effect of arousal on the latency of thesubsequent memory effect ANOVAs were conducted onsubtraction data (remembered 2 forgotten) from CZ attwo different epochs 400ndash600 and 600ndash800 msec (seeFigure 4) These locations and time windows were se-lected because as was mentioned above the subsequentmemory effect is likely to occur at midline electrodes andas early as 300ndash 400 msec after stimulus onset (Rugg1995) Since a preliminary ANOVA yielded no differencebetween pleasant and unpleasant pictures either at the early[F(114) = 0002 p 97] or at the late [600ndash800 msecF(114) = 011 p 75] epoch statistical power was in-creased by collapsing pleasant and unpleasant conditionsinto one emotional category The 2 (epoch 400ndash600 vs600ndash800 msec) 3 2 (arousal emotional vs neutral)ANOVA yielded a significant main effect of epoch[F(114) = 2476 p 00025] and a reliable epoch 3arousal interaction [F(114) = 556 p 035] To clarify

Unpleasantpictures

Pleasantpictures

Neutralpictures

Early epoch(400ndash600 msec)

Late epoch(600ndash800 msec)

Figure 3 Subtraction (recalled 2 forgotten) topomaps comparing the subsequent memory effectfor pleasant unpleasant and neutral pictures at different epochs The subsequent memory effectover centroparietal areas occurred earlier for pleasant and unpleasant pictures (400ndash600 msec) thanfor neutral pictures (after 600 msec) The maximal difference occurred at CZ


this interaction separate ANOVAs were computed onearly (400ndash600 msec) and late (600ndash800 msec) epochs Aone-way (arousal emotional vs neutral) ANOVA showedthat the subsequent memory effect was greater for emo-tional than for neutral pictures in the early epoch [F(114) =601 p 028] but not in the late epoch [F(114) = 01p 8] Consistent with these results separate ANOVAsdirectly comparing the waves for remembered and for-gotten pictures for each emotion condition (pleasant un-pleasant and neutral) and epoch (early and late) showedthat the subsequent memory effect (remembered vs for-gotten) for both pleasant and unpleasant pictures was sig-nificant during both early [F(114) = 1029 p 0064for pleasant and F(114) = 956 p 0081 for unpleas-

ant] and late [F(114) = 807 p 0132 for pleasantand F(114) = 1414 p 0022 for unpleasant] epochswhereas for neutral pictures it was significant during thelate epoch [F(114) = 1213 p 0038] but not during theearly epoch [F(114) = 085 p 3732 see also the re-memberedforgotten wave graphs for pleasant unpleas-ant and neutral pictures in Figure 4A] In other wordsthe subsequent memory effect at CZ occurred earlier foremotional than for neutral pictures

To investigate the possible lateralization of the subse-quent memory effect at frontal electrodes ANOVAswere conducted on ERPs for remembered and forgottenpleasant unpleasant and neutral pictures from left (F3) andright (F4) electrodes and for both early (400ndash600 msec) and

Figure 4 Subsequent memory effect at CZ (A) Wave graphs comparing the ERPs for re-membered (R) and forgotten (F) pictures (B) Subtraction (recalled 2 forgotten) wave andbar graphs comparing the subsequent memory effect for pleasant unpleasant and neutralpictures During an early epoch (400ndash600 msec) the subsequent memory effect was larger for emotional than for neutral pictures Light gray pleasant dark gray unpleasant blackneutral


late (600ndash800 msec) epochs A 3 (emotion pleasant un-pleasant and neutral) 3 2 (memory remembered vsforgotten) 3 2 (hemisphere left vs right) repeated mea-sures ANOVA design did not yield any significant hemi-sphere 3 memory emotion 3 memory or hemisphere 3emotion 3 memory interactions during the early epoch(400ndash600 msec) these results suggest that the subsequentmemory effect was not different across emotion conditionsor hemispheres However during the late epoch (600ndash800 msec) similar 3 3 2 3 2 ANOVAs yielded significantmain effects of both memory [F(114) = 2733 p = 0001]and emotion [F(228) = 1101 p = 0003] and a signifi-cant hemisphere 3 emotion 3 memory interaction[F(228) = 568 p 009] Additional ANOVAs meantto elucidate this complex interaction computed on thesubtraction (remembered 2 forgotten) data showed thatthe subsequent memory effect differed across hemispheresonly for neutral pictures [left right F(114) = 512 p 041] but not for pleasant [F(114) = 33 p 575] andunpleasant [F(114) = 155 p 234] pictures Also di-rect comparisons of the subsequent memory effect forpleasant unpleasant and neutral pictures did not yieldany significant differences suggesting that despite thehemispheric differences encountered for neutral pic-tures the subsequent memory effect for emotional andneutral pictures was similar at these frontal locationsThese results suggest that the ERPs from these locationsmay reflect comparable involvement of the same frontalneural generators as those associated with successful en-coding in functional neuroimaging studies (eg Breweret al 1998 Wagner et al 1998)

Collectively the subsequent memory data show that ascompared with forgotten pictures remembered pictureselicited larger potentials over centroparietal and fronto-central regions for both emotional and neutral picturesHowever at central locations (ie CZ) and during anearly epoch (400ndash600 msec) the subsequent memory ef-fect was larger for emotional than for neutral picturessuggesting faster encoding for emotional pictures

DISCUSSION

The study yielded two main results both of which wereconsistent with our predictions First there was a cleartopographical difference between arousal-related and valence-related components of the emotion effect Theemotion effect was sensitive to arousal at parietal siteswhereas it was sensitive to both arousal and valence atfrontocentral sites Second there was a difference be-tween the subsequent memory effects for emotional andneutral pictures The subsequent memory effect was greaterin an early epoch for emotional than for neutral stimulisuggesting better encoding of high-arousing than of low-arousing stimuli

Emotion EffectAt parietal sites the emotion effect reflected differ-

ences in arousal whereas at frontocentral sites it reflected

differences in both arousal and valence These findingsare consistent with lesion (see Heilman 2000) and neuro-imaging (LaBar et al 1999) evidence about the role ofparietal areas in the processing of emotional arousal andwith ERP evidence that the valence-related differencestend to be larger at anterior electrodes than at posteriorelectrodes (Cuthbert et al 2000 Diedrich et al 1997)The findings are also consistent with a neuropsychologi-cal model developed by Heller and collaborators (Heller1993 Heller amp Nitschke 1998) which proposes the ex-istence of two distinct neural systems involved in the pro-cessing of emotional information one involved in the pro-cessing of emotional arousal located in the parietal lobesand one involved in the modulation of emotional valencelocated in the frontal lobes

It is possible that the ERP differences between emo-tional and neutral pictures reflected attentional differencesthat follow from viewing emotional content rather thandirect effects of emotion The fact that ERPs for emotionaland neutral pictures differed mainly in amplitude consti-tutes evidence for different levels of engagement of thesame neuralfunctional processes (Allan Wilding ampRugg 1998) Consistent with this idea there is evidenceshowing that because of their motivational significanceemotionally salient stimuli are selected by the brain forsustained attentional processing (see Lang Simons ampBalaban cited in Cuthbert et al 2000)

The neural generators of ERP differences reported areunclear but one may speculate that arousal-related differ-ences in the emotion effect primarily reflected the contri-bution of amygdalandashcortical interactions (Amaral et al1992) whereas valence-related differences primarily re-flected the contribution of the prefrontal cortex (David-son 1995 Heller 1993) In addition because the emotioneffect could reflect attentional modulation these differ-ences may also be related to activity in ldquoattentionrdquo areasfrom parietal and prefrontal regions (eg HopfingerBuonocore amp Mangun 2000)

The suggestion of a possible involvement of the amyg-dala as the neural generator of the emotion effect at pari-etal electrodes seems to be in disagreement with evidencesuggesting that the amygdala is important in arousal-mediated memory effects rather than in arousal per se(Cahill et al 1995 LaBar amp Phelps 1998) That is it hasbeen shown that amygdala patients did not show enhancedmemory to emotional items despite normal skin con-ductance responses to emotional words (LaBar amp Phelps1998) or normal arousal ratings of emotional scenes of anarrated slide show (Cahill et al 1995) One possibleexplanation is that the amygdala is involved in emotionalarousal but is not necessary for it Thus damage to theamygdala may not impair emotional arousal even if thisregion participates in emotional arousal in the normalbrain (Dolcos Graham LaBar amp Cabeza in press) An-other possible explanation is that the ERP emotion effectfrom parietal electrodes reflects activity in other brainareas involved in the processing of emotional arousalsuch as the parietal lobes This interpretation is consistent


with the proposed role of the parietal areas in the pro-cessing of emotional arousal (Heller 1993) and can rec-oncile the above-mentioned lines of evidence

The valence-related difference in the emotion effect atfrontocentral electrodes could reflect a difference inarousal between the pleasant and the unpleasant picturesbut the fact that pleasant and unpleasant pictures wereequated for arousal suggests that this difference could re-flect a real valence difference at the cortical level of af-fective processing (Cuthbert et al 2000) This differenceoccurred because the emotion effect started earlier forpleasant than for unpleasant pictures (Figure 2) and couldreflect a preference toward pleasant pictures This inter-pretation has ecological validity in the sense that nor-mally people are more likely to show preference forpleasant than for unpleasant stimuli (the Pollyanna effectMatlin amp Stang 1978) It is also consistent with the no-tion proposed by Diedrich et al (1997) that the emotioneffect in the 400- to 600-msec epoch reflects the process-ing of the emotional qualities of the stimuli depending onthe amount of attention paid to the emotional content

This earlier emotion effect for pleasant pictures mayseem surprising if one considers the evidence coming fromwork on the potentiated startle reflex that suggests en-hanced response to unpleasant stimuli relative to pleasantstimuli (eg Lang Bradley amp Cuthbert 1990 SchuppCuthbert Bradley Birbaumer amp Lang 1997) A simpleexplanation of these differences could be that they may re-flect different processes Whereas the affective modula-tion of the startle reflex is thought of as reflecting moti-vational priming in which a defensive reflex is augmentedwhen the ongoing motivational state is aversive in nature(Lang 1995 cited by Schupp et al 1997) the modula-tion of the ERP may be interpreted as reflecting earlierprocessing owing to preference Given these differencesit would be interesting to design studies intended to dis-sociate these phenomena

In sum the emotion effect results show a clear ERPtopographical dissociation of arousal and valence Theseresults are consistent with previous ERP evidence (Cuth-bert et al 2000 Diedrich et al 1997) and with evidencesupporting the general notion that parietal areas are in-volved in the processing of emotional arousal (Heilman2000 Heller 1993 LaBar et al 1999) whereas prefrontalareas are involved in the evaluation of the emotional va-lence of stimuli (Davidson amp Irwin 1999 Davidson ampTomarken 1989 Heller 1993)

Subsequent Memory EffectThe main result of the present study is the finding that

the subsequent memory effect was larger during an earlyepoch for emotional than for neutral pictures There is aconsiderable amount of evidence that emotional stimuliare processed faster than neutral stimuli (Ohman Flykt ampLudqvist 2000) If we assume that shorter ERP latenciesreflect privileged access to processing resources (GanisKutas amp Sereno 1996) and that this advantage in pro-cessing leads to better encoding the present finding of anearlier subsequent memory effect for emotional pictures

could be one of the mechanisms underlying better memoryfor emotional than for neutral events

However a question still remains Why would an ear-lier engagement of processing resources confer an encod-ing advantage to emotional stimuli According to Salt-house (1996) two distinct mechanisms are responsible forthe relationship between speed and cognition the limitedtime mechanism and the simultaneity mechanism The lim-ited time mechanism is based on the idea that more pro-cessing results in higher performance and that the amountof processing is greater when the speed of processing isfaster The simultaneity mechanism is based on the ideathat higher level processing depends on the amount ofproducts of early processing simultaneously available forfurther processing Consequently the faster the speed ofprocessing the more information that is simultaneouslyavailable for high-level processing and thus the higherthe performance Adapted to our experiment these prin-ciples could explain our subsequent memory results forhigh-arousing pictures The emotional content granted aprivileged access to processing resources that resulted ina better encoding of high-arousing pictures This in turnresulted in better retrieval of pleasant and unpleasant pic-tures as compared with neutral pictures If true this in-terpretation could explain the beneficial effect of emotionon memory performance

As for the neural generators of this ERP effect giventhe aforementioned pharmacological (Cahill amp McGaugh1998 McGaugh et al 1996) lesion (Adolphs et al 2000Cahill et al 1995 LaBar amp Phelps 1998) and neuro-imaging (Cahill et al 1996 Canli et al 2000 Canli et al1999 Hamann et al 1999) evidence linking the amyg-dala to emotional memory formation it is reasonable to as-sume that the earlier subsequent memory effect for emo-tional pictures indirectly reflects activity in this structureAlthough activity in the amygdala is unlikely to volume-conduct to the scalp this region has intimate connectionswith a number of neocortical regions (Amaral et al 1992)and hence it is likely to influence scalp ERP recordingsThe neocortical generators of the emotional modulationof the subsequent memory effect are unclear but they mayinvolve prefrontal and parietal regions associated withemotional processing (Barbas 2000 Davidson amp Irwin1999 LaBar et al 1999 Lane et al 1997 Lang et al1998) Finally the similarity of the subsequent memoryeffect for emotional and neutral pictures at frontal elec-trodes may suggest that the ERPs from these locationsreflect comparable involvement of the same neural gener-ators as those associated with successful encoding in func-tional neuroimaging studies (Brewer et al 1998 Wagneret al 1998)

ConclusionsThe present study yielded two novel results First it

yielded a dissociation between arousal and valence Theemotion effect results clearly showed that the processingof the emotional content of pictures is sensitive to botharousal and valence and that their effects can be dissoci-ated over the scalp ERPs from parietal areas were sen-


sitive to arousal whereas ERPs from frontocentral areaswere sensitive to both arousal and valence Second this isthe first study in which the subsequent memory effect foremotional pictorial stimuli has been investigated The sub-sequent memory effect results show that this effect occurredearlier for pleasant and unpleasant pictures than for neu-tral pictures and suggest that emotional stimuli have priv-ileged access to processing resources possibly resultingin better encoding These results are compatible with pre-vious evidence about the role of different brain areas in theprocessing of emotional information and with evidenceabout the beneficial effect of emotion on memory forma-tion The present study provides the first available evi-dence of a link between two seemingly unrelated ERP phe-nomena the emotion effect and the subsequent memoryeffect

REFERENCES

Adolphs R Tranel D Damasio H amp Damasio A (1994) Im-paired recognition of emotions in facial expressions following bilat-eral damage to the human amygdala Nature 372 699-672

Adolphs R Tranel D amp Denburg N (2000) Impaired emotionaldeclarative memory following unilateral amygdala damage Learningamp Memory 7 180-186

Allan K Wilding E L amp Rugg M D (1998) Electrophysiolog-ical evidence for dissociable processes contributing to recollectionActa Psychologica 98 231-252

Amaral D G Price J L Pitkanen A amp Carmichael S T(1992) Anatomical organization of the primate amygdaloid complexIn J P Aggleton (Ed) The amygdala Neurobiological aspects of emo-tion memory and mental dysfunction (pp 1-66) New York Wiley-Liss

Barbas H (2000) Connections underlying the synthesis of cognitionmemory and emotion in primate prefrontal cortices Brain ResearchBulletin 52 319-330

Bechara A Damasio H Damasio A R amp Lee P L (1999) Dif-ferent contributions of the human amygdala and ventromedial pre-frontal cortex to decision-making Journal of Neuroscience 19 5473-5481

Borod J C Koff E amp Caron H (1983) Right hemisphere spe-cialization for the expression and appreciation of emotion A focus onthe face In E Perecman (Ed) Cognitive processing in the right hemi-sphere (pp 83-110) New York Academic Press

Borod J C Obler L K Erhan H M Grunwald I S CiceroB A Welkowitz J Santschi C Agosti R M amp Whalen J R(1998) Right hemisphere emotional perception Evidence acrossmultiple channels Neuropsychology 12 446-458

Bradley M M Greenwald M K Petry M C amp Lang P J(1992) Remembering pictures Pleasure and arousal in memory Jour-nal of Experimental Psychology Learning Memory amp Cognition 18379-390

Bradley M M amp Lang P J (1994) Measuring emotion The self-assessment manikin and the semantic differential Journal of Behav-ior Therapy amp Experimental Psychiatry 25 49-59

Brewer J B Zhao Z Desmond J E Glover G H amp GabrieliJ D (1998) Making memories Brain activity that predicts how wellvisual experience will be remembered Science 281 1185-1187

Cahill L (1996) Neurobiology of memory for emotional eventsConverging evidence from infra-human and human studies In ColdSpring symposia on quantitative biology (Vol 61 pp 259-264) ColdSpring Harbor NY Cold Spring Harbor Laboratory Press

Cahill L Babinsky R Markowitsch H amp McGaugh J L(1995) The amygdala and emotional memory Nature 377 295-296

Cahill L Haier R J Fallon J Alkire M T Tang CKeator D Wu J amp McGaugh J L (1996) Amygdala activity atencoding correlated with long-term free recall of emotional informa-tion Proceedings of the National Academy of Sciences 93 8016-8021

Cahill L amp McGaugh J L (1998) Mechanisms of emotionalarousal and lasting declarative memory Trends in Neurosciences 21294-299

Canli T Desmond J E Zhao Z Glover G amp Gabrieli J D E(1998) Hemispheric asymmetry for emotional stimuli detected withfMRI NeuroReport 9 3233-3239

Canli T Zhao Z Brewer J Gabrieli J D E amp Cahill L(2000) Event-related activation in the human amygdala associatedwith later memory for individual emotional experience Journal ofNeuroscience 20 1-5

Canli T Zhao Z Desmond J E Glover G amp Gabrieli J D E(1999) fMRI identifies a network of structures correlated with re-tention of positive and negative emotional memory Psychobiology27 441-452

Carretie L Iglesias J amp Garcia T (1997) A study on the emo-tional processing of visual stimuli through event-related potentialsBrain amp Cognition 34 207-217

Carretie L Iglesias J Garcia T amp Ballesteros M (1997)N300 P300 and the emotional processing of visual stimuli Elec-troencephalography amp Clinical Neurophysiology 103 298-303

Christianson S-A (1992) The handbook of emotion and memoryResearch and theory Hillsdale NJ Erlbaum

Cuthbert B N Schupp H T Bradley M M Birbaumer N ampLang P J (2000) Brain potentials in affective picture processingCovariation with autonomic arousal and affective report BiologicalPsychology 52 95-111

Davidson R J (1995) Cerebral asymmetry emotion and affective styleIn R J H Davidson amp K Hugdahl (Eds) Brain asymmetry (pp 361-387) Cambridge MA MIT Press

Davidson R J amp Irwin W (1999) The functional neuroanatomy ofemotion and affective style Trends in Cognitive Sciences 3 11-20

Davidson R J amp Tomarken A J (1989) Laterality and emotionAn electrophysiological approach In F Boller amp J Grafman (Eds)Handbook of neuropsychology (Vol 3 pp 419-441) New York Else-vier

Diedrich O Naumann E Maier S amp Becker G (1997) A frontalpositive slow wave in the ERP associated with emotional slides Jour-nal of Psychophysiology 11 71-84

Dolcos F Graham R LaBar K S amp Cabeza R (in press) Co-activation of the amygdala and hippocampus predicts better recall foremotional than for neutral pictures Brain amp Cognition

Fabiani M Karis D amp Donchin E (1986) P300 and recall in anincidental memory paradigm Psychophysiology 23 298-308

Ganis G Kutas M amp Sereno M I (1996) The search for ldquocommonsenserdquo An electrophysiological study of the comprehension of wordsand pictures in reading Journal of Cognitive Neuroscience 8 89-106

Halgren E amp Marinkovic K (1995) Neurophysiological networksintegrating human emotions In M S Gazzaniga (Ed) The cognitiveneurosciences (pp 1137-1151) Cambridge MA MIT Press

Hamann S B Ely T D Grafton S T amp Kilts C D (1999) Amyg-dala activity related to enhanced memory for pleasant and aversivestimuli Nature Neuroscience 2 289-293

Heilman K M (2000) Emotional experience A neurological modelIn D R Lane amp L Nadel (Eds) Cognitive neuroscience of emotion(pp 328-344) New York Oxford University Press

Heller W (1993) Neuropsychological mechanisms of individual dif-ferences in emotion personality and arousal Neuropsychology 7476-489

Heller W amp Nitschke J B (1998) The puzzle of regional brain ac-tivity in depression and anxiety The importance of subtypes and co-morbidity Cognition amp Emotion 12 421-447

Hopfinger J B Buonocore M H amp Mangun G R (2000) Theneural mechanisms of top-down attentional control Nature Neuro-science 3 284-291

Irwin W Davidson R J Lowe M J Mock B J Sorenson J Aamp Turski P A (1996) Human amygdala activation detected withecho-planar functional magnetic resonance imaging NeuroReport7 1765-1769

Jasper H H (1958) The tenndashtwenty electrode system of the Interna-tional Federation Electroencephalography amp Clinical Neurophysiol-ogy 10 371-375


Johnston V S Miller D R amp Burleson M H (1986) MultipleP3s to emotional stimuli and their theoretical significance Psy-chophysiology 23 684-694

LaBar K S amp Phelps E A (1998) Arousal-mediated memory con-solidation Role of the medial temporal lobe in humans Psychologi-cal Science 9 490-493

LaBar K S Rabinovici G D Ranganath C Gitelman D RParish T B Paller K A amp Mesulam M-M (1999) Spa-tiotemporal dynamics of a neural network for emotional picture en-coding revealed by parallel evoked potential and f MRI measure-ments Society for Neuroscience Abstracts 25 2146

Lane R D amp Nadel L (2000) Cognitive neuroscience of emotionNew York Oxford University Press

Lane R D Reiman E M Bradley M M Lang P J Ahern G LDavidson R J amp Schwartz G E (1997) Neuroanatomical corre-lates of pleasant and unpleasant emotion Neuropsychologia 351437-1444

Lang P J (1995) The emotion probe Studies of motivation and at-tention American Psychologist 50 372-385

Lang P J Bradley M M amp Cuthbert B N (1990) Emotion at-tention and the startle reflex Psychological Review 97 377-395

Lang P J Bradley M M amp Cuthbert B N (1997) Internationalaffective picture system [Pictures] Gainesville NIMH Center for theStudy of Emotion and Attention

Lang P J Bradley M M Fitzsimmons J R Cuthbert B NScott J D Moulder B amp Nangia V (1998) Emotional arousaland activation of the visual cortex An fMRI analysis Psychophysi-ology 35 199-210

Lang P J Greenwald M K Bradley M M amp Hamm A O(1993) Looking at pictures Affective facial visceral and behavioralreactions Psychophysiology 30 261-273

Lang P J Simons R F amp Balaban M T (1997) Attention and ori-enting Sensory and motivational processes Mahwah NJ Erlbaum

LeDoux J E (1993) Emotional memory systems in the brain Behav-ioural Brain Research 58 69-79

Leiphart J Rosenfeld J P amp Gabrieli J D (1993) Event-relatedpotential correlates of implicit priming and explicit memory tasks In-ternational Journal of Psychophysiology 15 197-206

Maratos E J Allan K amp Rugg M D (2000) Recognition mem-ory for emotionally negative and neutral words An ERP study Neu-ropsychologia 38 1452-1465

Maratos E J amp Rugg M D (2001) Electrophysiological correlatesof the retrieval of emotional and non-emotional context Journal ofCognitive Neuroscience 13 877-891

Matlin M W amp Stang D J (1978) The Pollyanna principle Selec-tivity in language memory and thought Cambridge MA Shenkman

McGaugh J L Cahill L amp Roozendaal B (1996) Involvementof the amygdala in memory storage Interaction with other brain sys-tems PNAS Proceedings of the National Academy of Sciences 9313508-13514

Morris J S Frith C D Perrett D I Rowland D YoungA W Calder A J amp Dolan R J (1996) A differential neural re-sponse in the human amygdala to fearful and happy facial expressionsNature 383 812-815

Naumann E Bartussek D Diedrich O amp Laufer M E (1992)Assessing cognitive and affective information processing functionsof the brain by means of the late positive complex of the event-relatedpotential Journal of Psychophysiology 6 285-298

Naumann E Maier S Diedrich O Becker G amp Bartussek D(1997) Structural semantic and emotion-focused processing of neu-tral and negative nouns Event-related potential correlates Journal ofPsychophysiology 11 158-172

Ohman A Flykt A amp Ludqvist D (2000) Unconscious emotionEvolutionary perspectives psychophysiological data and neuropsy-chological mechanisms In R D Lane amp L Nadel (Eds) Cognitiveneuroscience of emotion (pp 296-327) New York Oxford UniversityPress

Paller K A Kutas M amp Mayes A R (1987) Neural correlates

of encoding in an incidental learning paradigm Electroencephalog-raphy amp Clinical Neuropsychology 67 360-371

Palomba D Angrilli A amp Mini A (1997) Visual evoked poten-tials heart rate responses and memory to emotional pictorial stimuliInternational Journal of Psychophysiology 27 55-67

Phelps E A amp Anderson A K (1997) Emotional memory Whatdoes the amygdala do Current Biology 7 R311-R314

Rugg M D (1995) ERP studies of memory In M G H Coles ampM D Rugg (Eds) Electrophysiology of mind Event-related brainpotentials and cognition (pp 132-170) New York Oxford UniversityPress

Salthouse T A (1996) The processing-speed theory of adult age dif-ferences in cognition Psychological Review 103 403-428

Sanquist T F Rohrbaugh J W Syndulko K amp Lindsley D B(1980) Electrophysiological signs of level of processing Perceptualanalysis and recognition memory Psychophysiology 17 568-576

Schneider F Grodd W Weiss U Klose U Mayer K RNagele T amp Gur R C (1997) Functional MRI reveals left amyg-dala activation during emotion Psychiatry Research NeuroimagingSection 76 75-82

Schupp H T Cuthbert B N Bradley M M Birbaumer N ampLang P J (1997) Probe P3 and blinks Two measures of affectivestartle modulation Psychophysiology 34 1-6

Shields S (1991) Gender in the psychology of emotion A selectiveresearch review In K Strongman (Ed) International review of stud-ies on emotion (pp 227-245) New York Wiley

Vanderploeg R D Brown W S amp Marsh J T (1987) Judgmentsof emotion in words and faces ERP correlates International Journalof Psychophysiology 5 193-205

Wagner A D Koutstaal W amp Schacter D L (1999) When en-coding yields remembering Insights from event-related neuroimag-ing Philosophical Transactions of The Royal Society of London Se-ries B 354 1307-1324

Wagner A D Schacter D L Rotte M Koutstaal W Maril ADale A M Rosen B R amp Buckner R L (1998) Buildingmemories Remembering and forgetting of verbal experiences as pre-dicted by brain activity Science 281 1188-1191

Wheeler R E Davidson R J amp Tomarken A J (1993) Frontalbrain asymmetry and emotional reactivity A biological substrate ofaffective style Psychophysiology 30 82-89

Windmann S amp Kutas M (2001) Electrophysiological correlates ofemotion-induced recognition bias Journal of Cognitive Neuro-science 13 577-592

NOTE

1 The IAPS picture numbers were as follows pleasant (2050 44604470 4490 4510 4520 4531 4532 4533 4535 4550 4561 45724598 4599 4607 4608 4609 4611 4640 4653 4658 4659 46604670 4680 4690 5450 5460 5470 5621 5623 5626 5629 59107270 7400 7502 8021 8030 80318034 8041 8080 8090 81308161 8170 8180 8190 8200 8210 8300 8370 8400 8470 84908496 8501 8502) unpleasant (1050 1111 1113 1120 1274 13011310 1930 2053 2730 2800 3051 3062 3061 3150 3180 32203230 3261 3350 3550 6020 6190 6211 6212 6242 6243 62446370 6530 6570 6571 6821 6831 7361 7380 8230 8480 90069040 9042 9050 9160 9181 9250 9300 9420 9490 9561 95709620 9621 9622 9630 9800 9810 9910 9911 9920 9921) and neu-tral (2190 2210 2381 2440 2480 2570 2840 2870 2880 28904150 4250 4605 5130 5500 5510 5520 5530 5533 5534 57406150 7000 7002 7004 7006 7009 7010 7020 7025 7030 70317034 7035 7040 7050 7080 7090 7100 7110 7130 7140 71507160 7170 7175 7180 7185 7187 7217 7224 7233 7235 74917500 7550 7710 7950 9210 9700)

(Manuscript received December 21 2001revision accepted for publication July 19 2002)


hypothesis proposes that the right hemisphere is special-ized for the perception expression and experience of emo-tion (eg Borod Koff amp Caron 1983 see also Borodet al 1998) The valence hypothesis proposes that the lefthemisphere is primarily associated with processing ofpleasant emotions whereas the right hemisphere is pri-marily associated with processing of unpleasant emotionsThis hypothesis is supported by electrophysiological(Davidson amp Tomarken 1989) and functional neuroimag-ing (Canli et al 1998 see also Davidson 1995 Davidsonamp Irwin 1999) evidence The amygdala has been stronglyassociated with emotional processing by both lesion(Adolphs et al 1994) and functional neuroimaging (Irwinet al 1996 Lane et al 1997 Morris et al 1996 Schnei-der et al 1997) studies For instance lesion studies showthat patients with bilateral amygdalar damage are im-paired in detecting emotional facial expression (Adolphset al 1994) and imaging studies on neurologically intactpeople have reported amygdalar activations associatedwith processing of both pleasant and unpleasant stimuli(Irwin et al 1996 Lane et al 1997 Morris et al 1996Schneider et al 1997)

ERP studies of emotion have identified an emotion ef-fect ERPs for emotional stimuli (pleasant or unpleasant)tend to be more positive-going than ERPs for neutral stim-uli (Carretie Iglesias amp Garcia 1997 Cuthbert SchuppBradley Birbaumer amp Lang 2000 Johnston Miller ampBurleson 1986 Naumann Maier Diedrich Becker ampBartussek 1997 Vanderploeg Brown amp Marsh 1987)The emotion effect has been found in different ERP com-ponents including the P300 component (eg Johnstonet al 1986 Naumann Bartussek Diedrich amp Laufer1992 see also Diedrich Naumann Maier amp Becker1997) the N300 component (Carretie Iglesias amp Garcia1997 Carretie Iglesias Garcia amp Ballesteros 1997) andthe slow wave (SW) component (Cuthbert et al 2000Diedrich et al 1997) but most consistently the emotioneffect is expressed in a P300ndashSW complex The P300component may be more sensitive to emotion under in-tentional emotional processing (eg Johnston et al 1986Naumann et al 1992) whereas the N300 component ap-pears to be more sensitive during incidental emotionalprocessing (Carretie Iglesias amp Garcia 1997 CarretieIglesias et al 1997) However the evidence is inconclu-sive (Naumann et al 1992) The neural generators of theemotion effect are not known but it is reasonable to as-sume that it reflects interactions between the amygdalaand cortical regions (Amaral Price Pitkanen amp Car-michael 1992 LaBar et al 1999 Lane et al 1997)

There are two main unsolved issues concerning theemotion ERP effect First it is unclear whether the emo-tion effect is sensitive only to arousal (emotional vs neu-tral) or to both arousal and valence (pleasant vs unpleas-ant) Whereas most studies have only found differencesowing to arousal (see Diedrich et al 1997 for a brief re-view) a few recent studies have also reported differencesthat could be attributed to emotional valence (Cuthbertet al 2000 Diedrich et al 1997) A possible explanation

of why previous studies have not found clear differencesowing to valence could be the use of word stimuli whichtend to generate smaller emotion effect overall (3ndash 4 m Vin amplitude eg Naumann et al 1997) as comparedwith more complex visual stimuli such as pictures (5ndash6 m V or more eg Cuthbert et al 2000) The failure toclearly dissociate the arousal and the valence could alsobe due to the fact that most studies either have comparedonly unpleasant and neutral items or have not equated thepleasant and the unpleasant stimuli for arousal Alsomany studies have used a relatively small number of stim-uli per condition (20 or fewer) which may not give enoughstatistical power for fine dissociations Second if theemotion effect is sensitive to both arousal and valence itis unclear whether these two dimensions can be dissoci-ated in their topographical andor temporal characteristicsThere is some evidence that valence-related differencestend to be larger in anterior electrodes (Cuthbert et al2000 Diedrich et al 1997) but most studies have em-ployed too few electrodes (eg nine) to allow a clear dis-sociation in topography

To explore these issues we investigated the effect ofarousal and valence on ERPs recorded from 32 electrodeswhile subjects rated the emotional content of 60 pleas-ant 60 unpleasant and 60 neutral pictures Pictures wereselected so that both pleasant and unpleasant pictures dif-fered from neutral pictures in terms of both arousal andvalence whereas pleasant and unpleasant pictures dif-fered from each other only in terms of valence

The Enhancing Effect ofEmotion on Memory Formation

Behavioral (Bradley et al 1992) pharmacological(Cahill amp McGaugh 1998 McGaugh Cahill amp Roozen-daal 1996) lesion (Adolphs Tranel amp Denburg 2000Cahill Babinsky Markowitsch amp McGaugh 1995 LaBaramp Phelps 1998) and neuroimaging (Cahill et al 1996Canli Zhao Brewer Gabrieli amp Cahill 2000 CanliZhao Desmond Glover amp Gabrieli 1999 Hamann ElyGrafton amp Kilts 1999) studies have suggested that thebeneficial effect of emotion on memory is related to botharousal and valence with arousal as the most critical fac-tor Pharmacological studies with animals have investi-gated the role of peripheral adrenergic systems and theirinteraction with other neurotransmitter systems (egcholinergic GABA-ergic opiod-peptidergic) in the amyg-dala (Cahill amp McGaugh 1998 McGaugh et al 1996)The results of these studies have suggested that theamygdala exerts its enhancing role on memory for emo-tionally arousing stimuli through the modulation of hip-pocampal activity (Cahill amp McGaugh 1998 McGaughet al 1996)

Lesion (Adolphs et al 2000 Cahill et al 1995 LaBaramp Phelps 1998) and functional neuroimaging (Cahillet al 1996 Canli et al 2000 Canli et al 1999 Hamannet al 1999) studies on humans have supported the roleof the amygdala in the formation of emotional memoryFor instance lesion studies have shown that patients with









METHOD












RESULTS



















A FCZ

B PZ



20

0

ndash20

20

0

ndash20

0 500 1000 msec

0 500 1000 msec

V

V



V V

VV

6

5

4

3

2

1

0

14

12

10

8

6

4

2

0

10

8

6

4

2

0

10

8

6

4

2

0







Unpleasantpictures

Pleasantpictures

Neutralpictures












DISCUSSION






















REFERENCES











































































NOTE











METHOD












RESULTS



















A FCZ

B PZ



20

0

ndash20

20

0

ndash20

0 500 1000 msec

0 500 1000 msec

V

V



V V

VV

6

5

4

3

2

1

0

14

12

10

8

6

4

2

0

10

8

6

4

2

0

10

8

6

4

2

0







Unpleasantpictures

Pleasantpictures

Neutralpictures












DISCUSSION






















REFERENCES











































































NOTE












RESULTS



















A FCZ

B PZ



20

0

ndash20

20

0

ndash20

0 500 1000 msec

0 500 1000 msec

V

V



V V

VV

6

5

4

3

2

1

0

14

12

10

8

6

4

2

0

10

8

6

4

2

0

10

8

6

4

2

0







Unpleasantpictures

Pleasantpictures

Neutralpictures












DISCUSSION






















REFERENCES











































































NOTE















A FCZ

B PZ



20

0

ndash20

20

0

ndash20

0 500 1000 msec

0 500 1000 msec

V

V



V V

VV

6

5

4

3

2

1

0

14

12

10

8

6

4

2

0

10

8

6

4

2

0

10

8

6

4

2

0







Unpleasantpictures

Pleasantpictures

Neutralpictures












DISCUSSION






















REFERENCES











































































NOTE








A FCZ

B PZ



20

0

ndash20

20

0

ndash20

0 500 1000 msec

0 500 1000 msec

V

V



V V

VV

6

5

4

3

2

1

0

14

12

10

8

6

4

2

0

10

8

6

4

2

0

10

8

6

4

2

0







Unpleasantpictures

Pleasantpictures

Neutralpictures












DISCUSSION






















REFERENCES











































































NOTE








Unpleasantpictures

Pleasantpictures

Neutralpictures












DISCUSSION






















REFERENCES











































































NOTE











DISCUSSION






















REFERENCES











































































NOTE






DISCUSSION






















REFERENCES











































































NOTE

















REFERENCES











































































NOTE





REFERENCES











































































NOTE








































NOTE



Documents

Event-related potentials of emotional memory: Encoding pleasant, unpleasant, and neutral pictures