implicit and explicit evaluation: fmri correlates

Implicit and Explicit Evaluation fMRI Correlatesof Valence Emotional Intensity and Control

in the Processing of Attitudes

William A Cunningham Carol L Raye and Marcia K Johnson

Abstract

amp Previous work suggests that explicit and implicit eval-uations (goodndashbad) involve somewhat different neuralcircuits that process different dimensions such as valenceemotional intensity and complexity To better understandthese differences we used functional magnetic resonanceimaging to identify brain regions that respond differentially tosuch dimensions depending on whether or not an explicitevaluation is required Participants made either goodndashbadjudgments (evaluative) or abstractndashconcrete judgments (notexplicitly evaluative) about socially relevant concepts (eglsquolsquomurderrsquorsquo lsquolsquohappinessrsquorsquo lsquolsquoabortionrsquorsquo lsquolsquowelfarersquorsquo) After scan-ning participants rated the concepts for goodness badness

emotional intensity and how much they tried to control theirevaluation of the concept Amygdala activation correlatedwith emotional intensity and right insula activation correlatedwith valence in both tasks indicating that these aspects ofstimuli were processed by these areas regardless of inten-tion In contrast for the explicitly evaluative goodndashbad taskonly activity in the anterior cingulate frontal pole and later-al areas of the orbital frontal cortex correlated with ratingsof control which in turn were correlated with a measureof ambivalence These results highlight that evaluations arethe consequence of complex circuits that vary depending ontask demands amp

INTRODUCTION

Sometimes evaluation is relatively straightforward Forexample most people do not need to deliberate muchbefore deciding whether concepts such as death hap-piness or freedom are good or bad Other evaluationsrequire more ref lection where multiple factors areconsidered and appropriately weighed Often thesemore complex attitudes about concepts such as guncontrol abortion sex education taxes or preemptivewar can lead to heated debates with others or even withoneself Of course people differ in the amount ofreflection they engage in about particular conceptssome individuals have a rapid and unambiguous re-sponse to abortion and some find disadvantages tofreedom and merits even in death and taxes

Given the strong theoretical links between attitudinaland affective processes the brain region most clearlyassociated with affect the amygdala has been proposedto play a critical role in the evaluation of the environmentand attitudinal processes more generally Across multiplestimulus modalities greater amygdala activation is ob-served to bad than good stimuli (Phelps OrsquoConnor

Gatenby et al 2001 LeDoux 2000 LaBar GatenbyGore LeDoux amp Phelps 1998 Zald Lee Fluegel ampPardo 1998 Zald amp Pardo 1997 Morris Frith et al1996) Moreover amygdala activation has been ob-served when participants are consciously unaware ofthe stimulus being processed (Cunningham JohnsonRaye et al in press Morris Ohman amp Dolan 1998Whalen et al 1998) and when evaluative judgmentsare not explicitly requested (Cunningham JohnsonGatenby Gore amp Banaji 2003) suggesting that theamygdala may be involved in automatic or unconsciousforms of evaluation

Yet the implication that the amygdala is the lsquolsquoattituderegionrsquorsquomdashactivating as a direct function of stimulusnegativitymdashis contradicted by recent research findingsIn addition to activating to negative stimuli the amyg-dala activates to positive stimuli when compared withneutral stimuli (Liberzon Phan Decker amp Taylor 2003Hamann Ely Hoffman amp Kilts 2002 Hamann amp Mao2002 Garavan Pendergrass Ross Stein amp Risinger2001 see Zald 2003 for a review) Such findings suggestthat amygdala activation may be more a function ofemotional intensity than valence This is supported byfurther evidence from a patient who despite bilateralamygdala damage showed negative responses to Blackcompared to White faces on an indirect measure of raceattitudes (Phelps Cannistraci amp Cunningham 2003)

Yale UniversityCurrently at the University of Toronto

D 2004 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 1610 pp 1717ndash1729

This finding suggests that other structures also play arole in processing valence

In the emotion literature emotion has been pro-posed to be a function of two theoretically separableconcepts valence (overall badness relative to goodnessoften defined as bad minus good) and intensity (Rus-sell 2003) At the same time there is evidence thatnegative stimuli are on average experienced as moreintense emotionally than positive stimuli (Ito Caciop-po amp Lang 1998) Thus in neuroimaging studies ithas often not been clear whether amygdala activationreflects the processing of valence or emotional inten-sity Greater activation for negative than positive stimulicould simply reflect the greater average intensity ofnegative stimuli rather than valence per se Support forthis suggestion comes from work looking at neuralprocessing of taste and odor where amygdala activa-tion is associated with intensity and other regions areassociated with valence (Anderson et al 2003 Smallet al 2003) Moreover a patient with bilateral amyg-dala damage rated valence of stimuli similar to controlsbut gave quite different ratings of the emotional inten-sity of these stimuli (Adolphs Russell amp Tranel 1999)

Furthermore in addition to intensity and valenceattitudes differ in complexity Although some attitudesmay be simplemdashmurder is badmdashothers are more com-plex and require reflection for evaluation Such stimulimay be experienced as having both positive and negativeaspects (ie we may feel ambivalent) and reflection maybe necessary to arrive at an evaluative judgment As inmonitoring memories (eg Johnson Hashtroudi ampLindsay 1993) the functional role of more reflectiveor controlled processes in evaluation may be to with-hold responding until more information is availableintegrate multiple sources or dimensions of informationandor retrieve additional information Also whereassimple valence may be processed implicitly explicitevaluation may be necessary for consideration ofconflicting or additional evaluatively relevant informa-tion People are not solely influenced by their initialresponse to stimuli they can control processing to bringevaluations in line with higher-order personal values orsituational constraints

Previous research suggests two neural systems in-volved in evaluation (Cunningham Johnson Gatenbyet al 2003) Greater activity was observed in the amyg-dala and right inferior prefrontal cortex (PFC) (BA 45insula) for stimuli later rated as lsquolsquobadrsquorsquo than those laterrated as lsquolsquogoodrsquorsquo regardless of whether or not the taskrequired an evaluative judgment suggesting evaluativeprocessing that occurs regardless of perceiver intentIn contrast activity in the medial (BA 10) and ventro-lateral (BA 47) PFC was greater when the task requiredan explicit evaluative judgment than when it did notIn addition these areas active during explicit evaluationappeared to be involved in processing evaluative com-plexity They were most active during evaluation of

stimuli with competing positively and negatively va-lenced information These findings suggest two sys-temsmdashone sensitive to stimulus valence and anotherespecially likely to be recruited under conditions ofattitudinal complexity where more deliberation may benecessary Consequently evaluations of the same stim-ulus can result in quite different subjective evaluativeexperiences and perhaps even different evaluative out-comes for example a positive attitude in one circum-stance and a negative one in another Consistent withthis related research shows that brain activity associ-ated with more automatic evaluative processing thatoccurs when stimuli are presented too brief ly forconscious report can be inhibited or modified whenthere is the opportunity for more conscious processing(Cunningham Johnson Raye et al in press)

An attitude system should be sensitive to valence andintensity (Eagly amp Chaiken 1998) These have not beenindependently assessed in neuroimaging studies of theevaluation of socially relevant stimuli moreover valencemay have been confounded with emotional intensity inneuroimaging studies of attitudes and evaluation Inaddition an attitude system must be able to deal witha range of complexity Here we investigated whetheremotional intensity and valence are processed by differ-ent neural components and whether the pattern ofneural activity associated with intensity and valencediffers for implicit and explicit evaluation We alsoinvestigated the neural components associated withthe control of attitudes as in complex attitudes thathave both positive and negative aspects

During rapid event-related functional magnetic reso-nance imaging (fMRI) participants made either goodndashbad or abstractndashconcrete judgments about 144 concepts(eg lsquolsquomurderrsquorsquo lsquolsquoloversquorsquo lsquolsquogun controlrsquorsquo lsquolsquoabortionrsquorsquo)Across trials participants made one goodndashbad and oneabstractndashconcrete judgment about each concept Brainregions that subserve lsquolsquoimplicitrsquorsquo evaluation should re-spond to evaluative aspects of the stimuli during bothtasks similarly That is because automatic evaluation bydefinition occurs without intention it should be presentfor both the goodndashbad and abstractndashconcrete tasks Incontrast regions associated with lsquolsquoexplicitrsquorsquo evaluationshould show greater activity in the goodndashbad taskcompared to the abstractndashconcrete task That is theseregions should be more active when participants havethe intention to form an evaluation

The items used in the study were chosen to widelysample concepts that vary on several dimensions Afterscanning participants rated each of the concepts forbadness goodness emotionality of their response andthe degree to which they typically seek to control theirinitial responses to the concept For each participantlsquolsquoattitude valencersquorsquo for each concept was computed asthe difference between the bad rating and good ratingfor that concept (bad good) lsquolsquoemotional intensityrsquorsquowas the rating given on the emotionality scale and

1718 Journal of Cognitive Neuroscience Volume 16 Number 10

lsquolsquocontrolrsquorsquo was the rating given on the control scale Weexpected control ratings to be highest for conceptsabout which people felt most ambivalent (ie bothpositively and negatively)

Using these ratings we looked for brain regionswhose activity varied with each dimension for each taskIf a region is associated with processing valence weshould observe increasing activation as badndashgood in-creases if a region is associated with processing emo-tional intensity we should observe increasing activationfor concepts as they are rated from less to more emo-tionally intense Regions associated with control shouldshow increasing activation as control ratings increaseFurthermore regions that correlated with ratings on adimension similarly for both tasks reflect implicit pro-cessing of that dimension whereas regions that cor-related with a dimension significantly more during thegoodndashbad than the abstractndashconcrete task are by def-inition associated with explicit evaluative processing ofthat dimension

RESULTS

Analysis Overview

In two hierarchical regression analyses we examined therelationship between participantsrsquo ratings on each of ourdimensions of interest (attitude valence emotional in-tensity control) with brain activity for the goodndashbad taskand for the abstractndashconcrete task Compared withbivariate correlational approaches the regression pa-rameters from these analyses reflect the unique rela-tionship between the particular dimension (egvalence) and brain activity while controlling for theother dimensions (eg emotional intensity) Thus wewere able statistically to unconfound the influence ofcorrelated predictors First we examined the relation-ship between brain activity and emotional intensity andvalence (calculated as bad good) ratings for each itemIn a second analysis we added participantsrsquo ratings ofcontrol to the regression analysis to determine addition-al brain regions involved in attitudinal conflict and in-tentional manipulation of attitudinal informationFurthermore for each analysis we classified activationsas being involved in (a) implicit evaluationmdashsignificantand equal correlations in both the goodndashbad and theabstractndashconcretetask(b)explicitevaluationmdashsignificantcorrelation for the goodndashbad task only that was alsogreater than the correlation for the abstractndashconcretetask or significant correlations for both tasks but sig-nificantly greater for the goodndashbad than the abstractndashconcrete task

Task Effects

Activations obtained simply by subtracting one task fromanother are presented in Table 1 For each significant

region we additionally report the significance for theregion comparing activity against a fixation baseline foreach task (positive t values reflect activations and neg-ative t values reflect deactivations relative to fixation)Replicating our previous findings several areas weremore active during the goodndashbad than the abstractndashconcrete task including areas of the medial PFC and theventrolateral PFC In contrast several areas of lateral PFCactivity were greater for the abstractndashconcrete than forthe goodndashbad task

Valence and Emotional IntensityRegression Step 1

For all analyses valence was computed as the differencebetween a participantrsquos ratings of bad and good suchthat larger valence scores reflected a more negativeattitude As expected and depicted in Figure 1 theemotional intensity of concepts rated as bad was higherthan the emotional intensity of concepts rated as goodAs goodness or badness increased emotional intensityincreased for both but more so for the bad conceptsThus the regression model in which both emotionalintensity and valence are entered simultaneously isnecessary to separately assess the brain regions associ-ated with these dimensions

Emotional Intensity

Consistent with the idea that amygdala activation isassociated with emotional intensity irrespective of va-lence we found a region of the left amygdala (Figure 2A)associated with rated emotional intensity [t(19) = 443p lt 001 MNI coordinates 20 4 16] but notvalence [t(19) = 54 p = ns t(19)difference = 363 p lt001] Interestingly this area of the left amygdala is thesame region we previously found associated with lsquolsquobad-nessrsquorsquo relative to lsquolsquogoodnessrsquorsquo when we were not able todifferentiate between emotional intensity and valence(Cunningham Johnson Gatenby et al 2003) A largearea of the orbital frontal cortex also showed more ac-tivation for more emotionally intense stimuli [t(19) =464 p lt 001 MNI 12 32 20] As can be seen inFigure 2A and B the correlation between emotionalintensity and neural activity for both of these regionswas similar for both the goodndashbad and abstractndashconcretetasks suggesting that emotional intensity is processedimplicitly

In addition consistent with the hypothesis that ex-plicit evaluation may involve additional brain regionsseveral regions were correlated with emotional intensityratings only when participants were making explicitgoodndashbad judgments For example additional regionsof the right orbital frontal cortex and the temporal polewere associated with rated emotional intensity onlyduring the explicit evaluative task (Figure 2C and D)

Cunningham Raye and Johnson 1719

The full set of areas correlated with emotional intensityis presented in Table 2

Attitude Valence

Further dissociating the processing of valence fromemotional intensity we found regions associated withvalence and not emotional intensity (see Figure 3)Specifically we found activity in an area of the right

inferior frontalinsular cortex (Figure 3A) that correlatedsignificantly with bad good valence [t(19) = 494 p lt001 MNI 32 20 0] but not with emotional intensityAlso an area of anterior cingulate cortex correlated withvalence [t(19) = 491 p lt 001 MNI 4 20 40] As canbe seen in Figure 3A and B these relationships weresimilar for both the goodndashbad and abstractndashconcretetask suggesting implicit processing of valence Addition-ally in the goodndashbad task (but not the abstractndashconcrete

Table 1 Significant Areas of Activation Associated with Task

Size Area BA t LR x y z AC GB

GoodndashBad gt AbstractndashConcrete

691 Orbital frontal gyrus 11 770 mid 0 28 8 745 ns

Medial frontal gyrus 1032 660 L 8 52 20 471 442

Medial frontal gyrus 10 654 R 12 56 8 ns 504

32 Amygdalahippocampus na 490 R 16 8 20 491 504

86 Precuneus 23 595 L 4 56 28 ns 429

297 Inferior temporal gyrus 2021 775 R 64 12 32 ns 608

Middle temporal gyrus 2021 771 R 64 16 24 ns 589

Temporal pole 38 623 R 32 20 24 ns 644

109 Middle temporal gyrus 2021 522 L 64 8 28 ns 389

Temporal pole 20 491 L 48 16 44 ns 434

Inferior temporal gyrus 20 425 L 52 24 20 406 457

27 Superior frontal gyrus 9 428 R 20 36 44 448 529

90 Angular gyrus 39 637 L 56 60 24 510 ns

Middle temporal gyrus 37 413 L 64 60 4 ns 404

AbstractndashConcrete gt GoodndashBad

193 Inferior frontal gyrus 45 1068 R 48 40 16 762 491

Middle frontal gyrus 46 544 R 44 56 12 558 ns

Inferior frontal gyrus 44 501 R 52 16 32 875 693

85 Middle frontal gyrus 46 549 L 48 52 8 520 ns

Middle frontal gyrus 46 537 L 44 52 20 518 446


45 Inferior temporal gyrus 37 462 R 56 56 24 649 620

Inferior temporal gyrus 37 459 R 60 60 12 658 628

1860 Lingual gyrus 18 808 L 12 92 8 724 648

Calcarine gyrus 18 769 L 20 72 8 804 660

Inferior parietal lobe 740 692 R 32 56 40 798 678

Note Table shows local maxima p lt 001 with an extent threshold of 15 voxels BA = Brodmannrsquos area RL = right or left hemisphere t = maximalt-statistic for the statistical difference x y and z the 3-D coordinates of the activation within normalized MNI space AC = t value for regioncompared to fixation for abstractndashconcrete trials GB = t value for region compared to fixation for goodndashbad trials For each region the signifi-cance for the region comparing activity against fixation baseline for each task (positive t values ref lect activations and negative t values ref lectdeactivations relative to fixation) AC = abstractndashconcrete task GB = goodndashbad task ns = nonsignificant


task) we found that an area of the left lateral orbitalfrontal cortex (BA 47) correlated with valence [t(19) =543 p lt 001 MNI 44 28 12] Thus whereas leftamygdala activity appears to reflect emotional intensityof the stimulus activity in areas of the right inferiorfrontalinsula and anterior cingulate appears to reflectan automatic or implicit response to the valence ofthe concept (ie the degree of the difference in badvs good ratings)1 The full set of areas is presented inTable 3

Control Regression Step 2

In many dual-process models of social cognition thefunction of explicit evaluative processes is to control ormanipulate attitudinal information (see Chaiken ampTrope 1999) Such reflective processing is recruitedfor explicit evaluation (ie induced by an evaluativeagenda) and should be especially likely given attitudinalcomplexity Examining the postscan ratings provided byparticipants provides information about attitudinallycomplex stimuli As can seen in Figure 4A stimuli ratedas most likely to elicit control are ones for whichparticipants indicated a roughly equal amount of posi-tivity and negativity (resulting in a lsquolsquoneutralrsquorsquo attitudescore when subtracting good from bad) but indicateda high emotional intensity Further breaking downthis relationship (see Figure 4B) these stimuli are theones for which participants as expected indicatedboth strong positivity and strong negativity that isambivalence (eg Preister amp Petty 1996 Cacioppo ampBernston 1994)

Interestingly all regions positively associated withcontrol ratings ( p lt 05) for abstractndashconcrete trialswere also qualified by an interaction with task in thatpartial correlations with rated control were significantlylarger for the goodndashbad trials than for the abstractndashconcrete trials This suggests that control was greaterunder a reflective agenda of explicit evaluation than inimplicit evaluation Specifically we found that areas ofthe anterior cingulate (BA 32) [t(19) = 542 p lt 001MNI 4 24 32 see Figure 5A] and the right anteriorPFC (BA 10) [t(19) = 663 p lt 001 MNI 40 56 4 seeFigure 5B]mdashareas frequently associated with tasks re-quiring cognitive controlmdashwere correlated with controlratings Additional areas correlated with control werethe bilateral lateral orbital PFC (BA 47) [left t(19) =663 p lt 001 MNI 52 28 8 right t(19) = 502 p lt001 MNI 48 32 4] and an area of the medial PFC[t(19) = 554 p lt 001 MNI 8 52 16] Overall thepattern of results suggests that control is especiallyengaged under conditions of an agenda to evaluateand when ambivalence is detected The full set ofcorrelations is presented in Table 4

We also computed an index of ambivalence using theequations specified by the Gradual Threshold Model(Preister amp Petty 1996) As might be expected theseratings were highly correlated with the ratings of controland were not found to predict any brain regions aboveand beyond the other ratings It is important to notehowever that the bivariate correlation between BOLDsignal and computed ambivalence identified an area ofthe right inferior frontal cortex consistent with ourprevious findings (Cunningham Johnson Gatenbyet al 2003)

DISCUSSION

Evaluation is not a single process Rather evaluationvaries in the degree to which it is implicit (relativelyautomatic and perhaps unconscious) or more explicit(deliberate controlled and conscious) (Greenwald ampBanaji 1995) Moreover we would expect implicit andexplicit processes to differentially subserve different as-pects of evaluation such as emotional intensity valenceand the desire to modify or control an initial responseConsistent with this idea we identified brain areas thatwere associated with evaluative processing of conceptsregardless of the participantrsquos task and other areas thatwere active only (or more so) when participants had thereflective agenda to evaluate concepts Furthermore weidentified regions that were correlated with participantsrsquoratings of specific dimensions of the stimuli pointing todistributed systems for evaluative processing

Dissociating Valence and Emotional Intensity

Social psychologists have for some time emphasizedtwo aspects of attitudes emotional arousalintensity

Figure 1 Relationship between valence and emotional intensityratings of valence (calculated as bad good such that larger scores

represent a more negative attitude) are plotted against ratings of

emotional intensity in black collapsed across items and participants

Plotted in gray is the fitted relationship between the variables


and valence (Eagly amp Chaiken 1998 Osgood Suci ampTannenbaum 1954) In the present study these twoaspects were dissociated during implicit evaluationActivation in the left amygdala was correlated with the

rated lsquolsquoemotional intensityrsquorsquo of concepts and an area ofthe right inferior frontalinsular cortex was associatedwith lsquolsquoattitude valencersquorsquo scores (bad good) That bothof these relationships were observed with and without

Figure 2 Correlation with emotional intensity random effects contrasts showing partial correlations ( p lt 001) between brain activityand rated emotional intensity controlling for rated valence for (A) the left amygdala (B) the medial orbital frontal cortex [A and B show

similar correlations for the two tasks] (C) the right lateral orbital frontal cortex and (D) the right temporal pole [C and D show

significant correlations only for the goodndashbad task] Parametric timelines were generated for significant voxels significant at p lt 001 andaveraged across participants separately for the goodndashbad and abstractndashconcrete trials


intention to generate an evaluation provides supportfor the suggestion that arousalintensity and valenceare basic aspects of evaluative processing that likely oc-cur automatically

These results also highlight that although the amyg-dala may be critical in automatically processing theemotional intensity of a stimulus it is not the onlystructure involved in implicit evaluation For examplealthough greater amygdala activation to Black than toWhite faces has been associated with greater race biason an implicit measure of race attitudes (CunninghamJohnson Raye et al in press Phelps OrsquoConnor Cun-ningham et al 2000) a patient with bilateral amygdaladamage also showed race bias on this same implicitmeasure (Phelps Cannistraci et al 2003) That a patientwith bilateral amygdala damage showed an automaticrace bias has been suggested to be problematic forinterpreting the neuroimaging results (see Cacioppoet al 2003) Yet if the amygdala is only one componentof an implicit evaluation system then this may not be apuzzle Patients with bilateral amygdala damage mayhave other intact areas that are involved in automaticevaluation such as the area of the right inferior frontalinsular cortex or orbital frontal cortex found in thisstudy Thus a potentially interesting direction for futureresearch would be to examine the qualitative differencesin implicit attitudes that patients with various lesionsmay show

Explicit Processing and Control

In forming evaluations of the world we are not limitedto those processes that unfold automatically and with-out intention When people have a specific goal toevaluate more controlled reflective processes are re-cruited An important function of ref lection is theselectioninhibition and manipulation of activated infor-mation to achieve a desired evaluative outcome or onethat is more congruent with situational constraints Wefound areas that correlated with control ratings (and notother ratings) that have typically been associated withcognitive control such as the anterior cingulate cortexand the right PFC (Mitchell Heatherton Kelley Wylandamp Macrae 2003 Carter et al 1998) Consistent with theidea that control is a function of more explicit processingwe found that these correlations interacted with tasksuch that the correlations were stronger in the goodndashbad task than in the abstractndashconcrete task Thus ex-plicit processing may permit the experience of morecomplex attitudes than simple valence and emotionalintensity

Control processes are especially critical for evaluationwhen information becomes more complex and lessstraightforward Examples of such complexity are in-stances in which both positive and negative informationabout an attitude object are simultaneously activemdashinstances that are understudied in the neuroscientific

Table 2 Areas Significantly Correlated with Rated EmotionalIntensity

Size Area BA t LR x y z

GoodndashBad = AbstractndashConcrete

50 Amygdala na 443 L 20 4 16

Caudate na 431 L 4 4 4

143 Orbitalfrontal gyrus

11 464 R 12 32 20

29 Medialfrontal gyrus

9 444 R 8 48 28

56 Brainstem na 496 mid 0 24 8

18 Middletemporal gyrus

37 479 R 56 72 8

45 Middleoccipital gyrus

39 512 L 48 64 28


19 454 L 40 88 32

35 Vermis na 538 mid 0 68 40

40 Angular gyrus 3940 480 R 40 52 32

98 Precentral gyrus 6 522 L 32 8 60

19 Precentral gyrus 6 439 R 36 8 52

100 Postcentral gyrus 2 551 L 24 44 56


27 Temporal pole 38 518 R 36 24 36

135 Precuneus 23 495 R 12 52 28

49 Cerebulum 18 586 L 12 56 12

GoodndashBad Only

23 Posterior orbitalfrontal gyrus

11 668 R 24 24 16

28 Operculum 38 407 L 60 8 0

16 Inferiortemporal gyrus

20 626 R 52 32 16

49 Inferiorfrontal gyrus

48 478 R 60 4 8


45 453 L 44 36 8

16 Precuneus 23 478 L 16 60 36

65 Supramarginalgyrus

40 555 L 64 48 36

Note Table shows local maxima p lt 001 with an extent threshold of15 voxels BA = Brodmannrsquos area RL = right or left hemispheret = maximal t-statistic for the statistical difference x y and z the3-D coordinates of the activation within normalized MNI space


study of affect and attitude As depicted in Figure 4Bthe concepts that participants reported trying to con-trol the most were the ones for which they reportedhaving both positive and negative attitudes Furtherthese concepts were ones for which a calculation ofvalence would result in a lsquolsquoneutralrsquorsquo attitude and a highdegree of emotional intensity (likely because of thecoactivation of strong positive and negative responses)To deal with this complexity different features mayneed to be noted weighted integrated or resolvedto make binary goodndashbad judgments required duringscanning

In our previous work we found that regions of theventrolateral PFC (BA 47) were active when explicitly

processing information that has both positive and neg-ative characteristics (Cunningham Johnson Raye et alin press Cunningham Johnson Gatenby et al 2003)Similar areas have been found to be associated withsemantic memory selection and perhaps inhibition ofirrelevant information (eg Thompson-Shill DrsquoEspositoAguirre amp Farah 1997) Activity in this region was asso-ciated with ratings of control in the present study sug-gesting that attitude conflict and ambivalence may havebeen involved when participants sought to control theirevaluative responses This should not be surprising At atheoretical level control is necessary when conflicting in-formation or desires are present Consistent with modelsof affect and attitude that propose separable processing

Figure 3 Correlation with valence random effects contrasts showing partial correlations ( p lt 001) between brain activity and rated valence

(bad good) controlling for rated emotional intensity for (A) the right insularinferior frontal cortex (B) the anterior cingulate cortex [A and B

show similar correlations for the two tasks] and (C) the left lateral orbital frontal cortex [C shows a significant correlation only for the goodndashbad

task] Parametric timelines were generated for significant voxels significant at p lt 001 and averaged across participants separately for the goodndashbadand abstractndashconcrete trials


of positivity and negativity (ie Preister amp Petty 1996Cacioppo amp Bernston 1994) our current and previousdata suggest that sometimes positive and negative infor-mation about a stimulus can be active simultaneously

Explicit Processing and Emotion

Explicit evaluation may allow for both more cognitivelycomplex attitudes and more affectively complex atti-tudes Simple emotions (eg fear joy) can arise rela-tively automatically from evaluation accompanyingperceptual processing of a stimulus whereas morecomplex emotions (eg remorse jealously empathy)likely involve more reflective processing (eg Johnsonamp Multhaup 1992) Similarly attitudes may vary from

relatively automatic assessments of valence and intensi-ty to more complex consciously constructed judgmentsIn this study we found some regions that were associ-ated with emotional intensity only when participantswere making a reflective evaluation (an area of theorbital frontal cortex temporal pole) The fact that theseregions are sensitive to emotional intensity only when anagenda to evaluate is active suggests that the emotionalexperience resulting from explicit processing may differfrom the emotional experience resulting from implicitprocessing alone For example for some people implicitprocessing of the concept lsquolsquoaffirmative actionrsquorsquo may elicitrelatively simple feelings of fear or anger whereasemotions experienced during explicit evaluation of lsquolsquoaf-firmative actionrsquorsquo may include guilt jealousy hope andother complex emotions in different combinationsacross individuals

Conclusions

In summary many important attitude objects are notsimply wholly good or wholly bad and the potentialrichness of our evaluativeemotional experience can-not be realized by automatic processes alone Peopleactively manipulate evaluative information in light ofsituational constraints and circumstances givingweight to information depending on its relevance fora current judgment Moreover people can engage inmore controlled processes to override information andeven come to final evaluations that may differ in va-lence from automatic evaluations (eg Fazio 1990)Additionally as more reflective component processesare engaged the emotional experience associatedwith the evaluation can become qualitatively richerand this emotional information itself can influence

Table 3 Areas Significantly Correlated with Rated Valence



61 Insula (claustrum) 48 494 R 36 20 0

Inferior frontal gyrus 45 473 R 52 20 4

33 Anterior cingulate 32 491 R 4 20 40

GoodndashBad Only

16 Lateral orbitalfrontal gyrus

47 543 L 44 28 12


Figure 4 Relationship between valence emotional intensity and control (A) Behavioral ratings of emotional intensity and computed valenceare plotted against control (B) Behavioral ratings of bad and good are plotted against control For each plot behavioral ratings are collapsed

across items and participants Dark blue represents the lowest control ratings and red represents the highest control ratings


judgments in ways that more primitive emotional re-sponses may not

METHODS

Participants

Twenty participants were paid for their participationParticipants reported no abnormal neurological historyand had normal or corrected-to-normal vision All par-ticipants provided informed consent

Materials

One hundred forty-four concepts were selected for thestudy These concepts varied on multiple dimensions

such as goodndashbad abstractconcrete and complexityExamples of concepts used in the study are lsquolsquomurderrsquorsquolsquolsquoloversquorsquo lsquolsquofreedomrsquorsquo lsquolsquomulticulturalismrsquorsquo lsquolsquotechnologyrsquorsquolsquolsquorecyclingrsquorsquo lsquolsquoimmigrationrsquorsquo lsquolsquoterrorismrsquorsquo and lsquolsquopoetryrsquorsquo

Procedure

During fMRI on each trial participants categorizedconcepts along one of two dimensions (goodndashbad orabstractndashconcrete) and indicated their categorization bymaking one of two button presses with their righthand Using E-Prime for PC stimuli were forwardprojected with an LCD onto a screen at the base ofthe MRI bore A prism mirror positioned over theparticipantsrsquo eyes allowed them to view stimuli All

Figure 5 Correlation with control random effects contrasts showing partial correlations ( p lt 001) between brain activity and rated controlcontrolling for rated emotional intensity and valence for (A) the anterior cingulate (B) the anterior prefrontal cortex and (C) the right lateral

orbital frontal cortex Parametric timelines were generated for significant voxels significant at p lt 001 and averaged across participants separately

for the goodndashbad and abstractndashconcrete trials


stimuli were presented in black letters against a whitebackground

For each trial a 500-msec cue indicated whether thetrial required a goodndashbad or an abstractndashconcrete re-sponse immediately followed by a concept presentedfor 2 sec Additional null fixation trials were used tospace out trials The ratio of critical trials to null fixationtrials was 41 Between all trials a fixation cross re-mained on the screen for 2 4 or 6 sec To synchronizestimulus presentations with functional scanning allfunctional runs were initiated by a trigger sent by theMRI scanner at the beginning of each scan Each of sixruns contained 48 of each trial type (goodndashbad task

abstractndashconcrete task randomly intermixed) All 144concepts were presented before any were repeatedTwo lists were counterbalanced so that each conceptwas first rated as goodndashbad for half the participants andabstractndashconcrete for the other half

After scanning participants completed a question-naire in which all concepts were rated along fourdimensions Participants indicated on a 0 to 9 scale theextent to which the target concept was good the extentto which it was bad the extent to which their responseto the concept was emotional and the extent to whichthey typically try to reflectively control or suppress initial(presumably automatic) responses that they have towardthe concept

fMRI Parameters

All imaging was conducted with a Seimens 3-T scanner atthe Yale Magnetic Resonance Research Center To getwhole-brain functional coverage 32 axial slices (slicethickness 38 mm no skip) were prescribed parallel tothe ACndashPC line with the 11th slice centered on the ACndashPC line Nearly isotropic functional images were ac-quired from inferior to superior using a single-shotgradient echo-planar pulse sequence (TE = 25 msecTR = 2 sec in-plane resolution = 375 375 mmmatrix size = 64 64 and FOV = 24 24 cm)

Data Analysis

Data were analyzed using the general linear model asimplemented by SPM2 (Friston et al 1995) Prior toanalysis data were corrected for slice acquisition timeMotion correction was performed using the INRIAligntoolbox for SPM (Freire amp Mangin 2001) Data werethen transformed to conform to the default EPI MNIbrain interpolated to 4 4 4 mm Functional datawere smoothed using a 12-mm FWHM (full width halfmaximum) kernel Finally a low pass filter removedfrequencies greater than 018 Hz a cutoff that representsthe frequency after which signals as a function ofexperimental effects are no longer expected

Two hierarchical regression analyses were constructedto examine the relationship between participantsrsquo ratingson each of our dimensions of interest (attitude valenceemotional intensity and control) separately for thegoodndashbad task and for the abstractndashconcrete taskIn each analysis a series of regressors were constructedto examine BOLD brain activity to each of the trial typesTwo regressors were used for each trial type theexpected BOLD signal following neural activity and aderivative to model the onset of the neural responseFor the first regression analysis we computed the rela-tionship between emotional intensity and valence foreach trial for both the goodndashbad and abstractndashconcreteconditions Covariation with these parametric regressorsindicate brain regions that are significantly related to

Table 4 Areas Significantly Correlated with Rated Control



36 Anterior frontal pole 10 467 R 32 60 8

32 Middle frontal gyrus 9 474 L 32 28 52

54 Lateral orbital frontalgyrus

47 502 R 48 32 4

16 Medial frontal gyrus 8 456 R 16 32 60

26 Middle frontal gyrus 9 440 R 48 16 48

21 Inferior temporalgyrus

20 466 R 48 4 44

22 Angular gyrus 39 474 R 48 60 28

GoodndashBad Only


Anterior cingulate 32 542 L 4 24 32

Medial frontal gyrus 1032 554 R 8 52 16

Middle frontal gyrus 9 505 R 40 16 52

42 Lateral orbital gyrus 47 680 L 52 28 8

21 Precuneus 31 344 L 12 60 28


20 636 R 60 24 16

30 Middle temporalgyrus

21 594 R 60 44 4


21 527 L 52 40 4

162 Angular gyrus 39 763 R 36 60 28

23 Parahippicampalgyrus

35 553 L 28 24 24


Note Table shows local maxima p lt 001 with an extent threshold of15 voxels BA = Brodmannrsquos area RL = right or left hemisphere t =maximal t-statistic for the statistical difference x y and z the 3-Dcoordinates of the activation within normalized MNI space


increases (or decreases) with the reported ratings foreach participant for each of the judgment types (goodndashbad or abstractndashconcrete) In the second regressionanalysis we added the third parametric regressor ofcontrol Contrast maps were generated for each partici-pant for each analysis of interest

Random effects composite t-maps were generatedusing the individual participant contrast maps as inputRegions of activation were defined as those areas inwhich 15 contiguous voxels were significant at p lt 001This was done for all regions except for the amygdala ana priori region of interest based on previous publishedwork and data from our own laboratory The criterionfor amygdala activation was 5 or more voxels significantat p lt 01 For areas to be described as involved pri-marily in reflective evaluation we additionally requiredthe correlation for the goodndashbad task to be significantlygreater than the correlation for the abstractndashconcretetask p lt 01

Acknowledgments

We thank Steven Most Joseph Bates and Norman Farb forhelpful comments on earlier versions of this manuscript Thisresearch was supported by a grant from the National Instituteof Health (MH 62196)

Reprint requests should be sent to William A CunninghamDepartment of Psychology University of Toronto 100 StGeorge Street Toronto ON M5S 3G3 Canada or via e-mailcunninghampsychutorontoca

The data reported in this experiment have been deposited inthe fMRI Data Center (httpwwwfmridcorg) The accessionnumber is 2-2004-1171A

Note

1 It is possible that the relationship between amygdalaactivation and emotional intensity but not valence could be afunction of power That is a correlation between valence andamygdala activation might exist but we did not have enoughpower to detect the relationship Although possible this seemsunlikely given our data Plotting the relationship betweenamygdala activation and valence shows a U-shaped functionwith the most positive and most negative stimuli (ie the oneswith the highest emotional intensity) showing the most ac-tivation It should be noted however that the activation to themost negative stimuli was greater than to the most positivestimuli showing evidence of the negativity bias (Ito Cacioppoamp Lang 1998)

REFERENCES

Adolphs R Russell J A amp Tranel D (1999) A role forthe human amygdala in recognizing emotional arousalfrom unpleasant stimuli Psychological Science 10167ndash171

Anderson A K Christoff K Stappen I Panitz DGhahremani D G Glover G Gabrieli J D amp Sobel N(2003) Dissociated neural representations of intensity

and valence in human olfaction Nature Neuroscience6 196ndash202

Cacioppo J T amp Bernston G (1994) Relationshipbetween attitudes and evaluative space A criticalreview with emphasis on the separability of positiveand negative substrates Psychological Bulletin 115401ndash423

Cacioppo J T Berntson G G Lorig T S Norris C JRickett E amp Nusbaum H (2003) Just because yoursquoreimaging the brain doesnrsquot mean you can stop using yourhead A primer and set of first principles Journal ofPersonality and Social Psychology 85 650ndash681

Carter C S Braver T S Barch D M Botvinick M MNoll D amp Cohen J D (1998) Anterior cingulate cortexerror detection and the online monitoring of performanceScience 280 747ndash749

Chaiken S amp Trope Y (1999) Dual-process theories insocial psychology New York Guilford Press

Cunningham W A Johnson M K Gatenby J C Gore J Camp Banaji M R (2003) Component processes of socialevaluation Journal of Personality and Social Psychology85 639ndash649

Cunningham W A Johnson M K Raye C L Gatenby J CGore J C amp Banaji M R (in press) Separable neuralcomponents in the processing of black and white facesPsychological Science

Eagly A H amp Chaiken S (1998) Attitude structure andfunction In D T Gilbert S T Fiske amp L Gardner(Eds) The handbook of social psychology (Vol 1 4th edpp 269ndash232) New York McGraw-Hill

Fazio R H (1990) Multiple processes by which attitudesguide behavior The MODE model as an integrativeframework In M P Zanna (Ed) Advances in experimentalsocial psychology (Vol 23 pp 75ndash109) New YorkAcademic Press

Freire L amp Mangin J F (2001) Motion correctionalgorithms may create spurious brain activations in theabsence of subject motion Neuroimage 14 709ndash722

Friston K J Holmes A P Worsley K J Poline J PFrith C D Frackowiak R S J (1995) Statistic parametricmaps in functional imaging A general linear approachHuman Brain Mapping 2 189ndash210

Garavan H Pendergrass J C Ross T J Stein E A ampRisinger R C (2001) Amygdala response to bothpositively and negatively valenced stimuli NeuroReport12 2779ndash2783

Greenwald A G amp Banaji M R (1995) Implicit socialcognition Attitudes self-esteem and stereotypesPsychological Review 102 4ndash27

Hamann S B Ely T D Hoffman J M amp Kilts C D (2002)Ecstasy and agony Activation of the human amygdala inpositive and negative emotion Psychological Science 13135ndash141

Hamann S amp Mao H (2002) Positive and negativeemotional verbal stimuli elicit activity in the left amygdalaNeuroReport 13 15ndash19

Ito T A Cacioppo J T amp Lang P J (1998) Elicitingaffect using the International Affective Picture SystemTrajectories through evaluative space Personality andSocial Psychology Bulletin 24 855ndash879

Johnson M K Hashtroudi S amp Lindsay D S (1993)Source monitoring Psychological Bulletin 114 3ndash28

Johnson M K amp Multhaup K S (1992) Emotion and MEMIn S Christianson (Ed) The handbook of emotion andmemory Research and theory (pp 33ndash66) Hillsdale NJErlbaum

LaBar K S Gatenby J C Gore J C LeDoux J E ampPhelps E A (1998) Human amygdala activation during


conditioned fear acquisition and extinction A mixed-trialfMRI study Neuron 20 937ndash945

LeDoux J E (2000) Emotion circuits in the brain AnnualReview of Neuroscience 23 155ndash184

Liberzon I Phan K L Decker L R amp Taylor S F(2003) Extended amygdala and emotional salienceA PET activation study of positive and negative affectNeuropsychopharmacology 28 726ndash733

Mitchell J P Heatherton T F Kelley W K Wyland C Lamp Macrae C N (2003) The contents of consciousnessThe neural substrates of thought suppression Manuscriptunder review

Morris J S Frith C D Perrett D I Rowland D Young A WCalder A J amp Dolan R J (1996) A differential neuralresponse in the human amygdala to fearful and happy facialexpressions Nature 383 812ndash815

Morris J S Ohman A amp Dolan R J (1998) Consciousand Unconscious emotional learning in the humanamygdala Nature 393 467ndash470

Osgood C E Suci G J amp Tannenbaum P H (1954)The measurement of meaning Urbana IL University ofIllinois Press

Phelps E A Cannistraci C J amp Cunningham W A (2003)Intact performance on an indirect measure of face biasfollowing amygdala damage Neuropsychologia 41203ndash208

Phelps E A OrsquoConnor K J Cunningham W AFunayama E S Gatenby J C Gore J C amp Banaji M R(2000) Performance on indirect measures of race evaluationpredicts amygdala activation Journal of CognitiveNeuroscience 12 729ndash738

Phelps E A OrsquoConnor K J Gatenby J C Gore J CGrillon C amp Davis M (2001) Activation of the left

amygdala to a cognitive representation of fear NatureNeuroscience 4 437ndash441

Priester J R amp Petty R E (1996) The gradual thresholdmodel of ambivalence Relating the positive and negativebases of attitudes to subjective ambivalence Journal ofPersonality and Social Psychology 71 431ndash449

Russell J A (2003) Core affect and the psychologicalconstruction of emotion Psychological Review 110145ndash172

Small D M Gregory M D Mak Y E Gitelman DMesulam M M amp Parrish T (2003) Dissociation ofneural representation of intensity and affective valuationin human gustation Neuron 39 701ndash711

Thompson-Schill S L DrsquoEsposito M Aguirre G K ampFarah M J (1997) Role of left inferior prefrontal cortexin retrieval of semantic knowledge A reevaluationProceedings of the National Academy of Sciences USA94 14792ndash14797

Whalen P J Rauch S L Etcoff N L McInerney S CLee M B amp Jenike M A (1998) Masked presentationsof emotional facial expressions modulate amygdala activitywithout explicit knowledge Journal of Neuroscience 18411ndash418

Zald D H (2003) The human amygdala and the emotionalevaluation of sensory stimuli Brain ResearchndashBrainResearch Review 41 88ndash123

Zald D H Lee J T Fluegel K W amp Pardo J V (1998)Aversive gustatory stimulation activates limbic circuits inhumans Brain 121 1143ndash1154

Zald D H amp Pardo J V (1997) Emotion olfaction andthe human amygdala Amygdala activation during aversiveolfactory stimulation Proceedings of the National Academyof Sciences USA 94 4119ndash4124


This finding suggests that other structures also play arole in processing valence

In the emotion literature emotion has been pro-posed to be a function of two theoretically separableconcepts valence (overall badness relative to goodnessoften defined as bad minus good) and intensity (Rus-sell 2003) At the same time there is evidence thatnegative stimuli are on average experienced as moreintense emotionally than positive stimuli (Ito Caciop-po amp Lang 1998) Thus in neuroimaging studies ithas often not been clear whether amygdala activationreflects the processing of valence or emotional inten-sity Greater activation for negative than positive stimulicould simply reflect the greater average intensity ofnegative stimuli rather than valence per se Support forthis suggestion comes from work looking at neuralprocessing of taste and odor where amygdala activa-tion is associated with intensity and other regions areassociated with valence (Anderson et al 2003 Smallet al 2003) Moreover a patient with bilateral amyg-dala damage rated valence of stimuli similar to controlsbut gave quite different ratings of the emotional inten-sity of these stimuli (Adolphs Russell amp Tranel 1999)

Furthermore in addition to intensity and valenceattitudes differ in complexity Although some attitudesmay be simplemdashmurder is badmdashothers are more com-plex and require reflection for evaluation Such stimulimay be experienced as having both positive and negativeaspects (ie we may feel ambivalent) and reflection maybe necessary to arrive at an evaluative judgment As inmonitoring memories (eg Johnson Hashtroudi ampLindsay 1993) the functional role of more reflectiveor controlled processes in evaluation may be to with-hold responding until more information is availableintegrate multiple sources or dimensions of informationandor retrieve additional information Also whereassimple valence may be processed implicitly explicitevaluation may be necessary for consideration ofconflicting or additional evaluatively relevant informa-tion People are not solely influenced by their initialresponse to stimuli they can control processing to bringevaluations in line with higher-order personal values orsituational constraints

Previous research suggests two neural systems in-volved in evaluation (Cunningham Johnson Gatenbyet al 2003) Greater activity was observed in the amyg-dala and right inferior prefrontal cortex (PFC) (BA 45insula) for stimuli later rated as lsquolsquobadrsquorsquo than those laterrated as lsquolsquogoodrsquorsquo regardless of whether or not the taskrequired an evaluative judgment suggesting evaluativeprocessing that occurs regardless of perceiver intentIn contrast activity in the medial (BA 10) and ventro-lateral (BA 47) PFC was greater when the task requiredan explicit evaluative judgment than when it did notIn addition these areas active during explicit evaluationappeared to be involved in processing evaluative com-plexity They were most active during evaluation of

stimuli with competing positively and negatively va-lenced information These findings suggest two sys-temsmdashone sensitive to stimulus valence and anotherespecially likely to be recruited under conditions ofattitudinal complexity where more deliberation may benecessary Consequently evaluations of the same stim-ulus can result in quite different subjective evaluativeexperiences and perhaps even different evaluative out-comes for example a positive attitude in one circum-stance and a negative one in another Consistent withthis related research shows that brain activity associ-ated with more automatic evaluative processing thatoccurs when stimuli are presented too brief ly forconscious report can be inhibited or modified whenthere is the opportunity for more conscious processing(Cunningham Johnson Raye et al in press)

An attitude system should be sensitive to valence andintensity (Eagly amp Chaiken 1998) These have not beenindependently assessed in neuroimaging studies of theevaluation of socially relevant stimuli moreover valencemay have been confounded with emotional intensity inneuroimaging studies of attitudes and evaluation Inaddition an attitude system must be able to deal witha range of complexity Here we investigated whetheremotional intensity and valence are processed by differ-ent neural components and whether the pattern ofneural activity associated with intensity and valencediffers for implicit and explicit evaluation We alsoinvestigated the neural components associated withthe control of attitudes as in complex attitudes thathave both positive and negative aspects

During rapid event-related functional magnetic reso-nance imaging (fMRI) participants made either goodndashbad or abstractndashconcrete judgments about 144 concepts(eg lsquolsquomurderrsquorsquo lsquolsquoloversquorsquo lsquolsquogun controlrsquorsquo lsquolsquoabortionrsquorsquo)Across trials participants made one goodndashbad and oneabstractndashconcrete judgment about each concept Brainregions that subserve lsquolsquoimplicitrsquorsquo evaluation should re-spond to evaluative aspects of the stimuli during bothtasks similarly That is because automatic evaluation bydefinition occurs without intention it should be presentfor both the goodndashbad and abstractndashconcrete tasks Incontrast regions associated with lsquolsquoexplicitrsquorsquo evaluationshould show greater activity in the goodndashbad taskcompared to the abstractndashconcrete task That is theseregions should be more active when participants havethe intention to form an evaluation

The items used in the study were chosen to widelysample concepts that vary on several dimensions Afterscanning participants rated each of the concepts forbadness goodness emotionality of their response andthe degree to which they typically seek to control theirinitial responses to the concept For each participantlsquolsquoattitude valencersquorsquo for each concept was computed asthe difference between the bad rating and good ratingfor that concept (bad good) lsquolsquoemotional intensityrsquorsquowas the rating given on the emotionality scale and




RESULTS

Analysis Overview


Task Effects





Emotional Intensity





Attitude Valence










86 Precuneus 23 595 L 4 56 28 ns 429



















1860 Lingual gyrus 18 808 L 12 92 8 724 648










DISCUSSION























50 Amygdala na 443 L 20 4 16



11 464 R 12 32 20


9 444 R 8 48 28



37 479 R 56 72 8


39 512 L 48 64 28


19 454 L 40 88 32

35 Vermis na 538 mid 0 68 40

40 Angular gyrus 3940 480 R 40 52 32





27 Temporal pole 38 518 R 36 24 36

135 Precuneus 23 495 R 12 52 28

49 Cerebulum 18 586 L 12 56 12

GoodndashBad Only


11 668 R 24 24 16

28 Operculum 38 407 L 60 8 0


20 626 R 52 32 16


48 478 R 60 4 8


45 453 L 44 36 8

16 Precuneus 23 478 L 16 60 36


40 555 L 64 48 36















Conclusions








GoodndashBad Only


47 543 L 44 28 12






METHODS

Participants


Materials



Procedure










fMRI Parameters


Data Analysis









47 502 R 48 32 4




20 466 R 48 4 44

22 Angular gyrus 39 474 R 48 60 28

GoodndashBad Only






21 Precuneus 31 344 L 12 60 28


20 636 R 60 24 16


21 594 R 60 44 4


21 527 L 52 40 4

162 Angular gyrus 39 763 R 36 60 28


35 553 L 28 24 24






Acknowledgments




Note


REFERENCES













































RESULTS

Analysis Overview


Task Effects





Emotional Intensity





Attitude Valence










86 Precuneus 23 595 L 4 56 28 ns 429



















1860 Lingual gyrus 18 808 L 12 92 8 724 648










DISCUSSION























50 Amygdala na 443 L 20 4 16



11 464 R 12 32 20


9 444 R 8 48 28



37 479 R 56 72 8


39 512 L 48 64 28


19 454 L 40 88 32

35 Vermis na 538 mid 0 68 40

40 Angular gyrus 3940 480 R 40 52 32





27 Temporal pole 38 518 R 36 24 36

135 Precuneus 23 495 R 12 52 28

49 Cerebulum 18 586 L 12 56 12

GoodndashBad Only


11 668 R 24 24 16

28 Operculum 38 407 L 60 8 0


20 626 R 52 32 16


48 478 R 60 4 8


45 453 L 44 36 8

16 Precuneus 23 478 L 16 60 36


40 555 L 64 48 36















Conclusions








GoodndashBad Only


47 543 L 44 28 12






METHODS

Participants


Materials



Procedure










fMRI Parameters


Data Analysis









47 502 R 48 32 4




20 466 R 48 4 44

22 Angular gyrus 39 474 R 48 60 28

GoodndashBad Only






21 Precuneus 31 344 L 12 60 28


20 636 R 60 24 16


21 594 R 60 44 4


21 527 L 52 40 4

162 Angular gyrus 39 763 R 36 60 28


35 553 L 28 24 24






Acknowledgments




Note


REFERENCES












































Attitude Valence










86 Precuneus 23 595 L 4 56 28 ns 429



















1860 Lingual gyrus 18 808 L 12 92 8 724 648










DISCUSSION























50 Amygdala na 443 L 20 4 16



11 464 R 12 32 20


9 444 R 8 48 28



37 479 R 56 72 8


39 512 L 48 64 28


19 454 L 40 88 32

35 Vermis na 538 mid 0 68 40

40 Angular gyrus 3940 480 R 40 52 32





27 Temporal pole 38 518 R 36 24 36

135 Precuneus 23 495 R 12 52 28

49 Cerebulum 18 586 L 12 56 12

GoodndashBad Only


11 668 R 24 24 16

28 Operculum 38 407 L 60 8 0


20 626 R 52 32 16


48 478 R 60 4 8


45 453 L 44 36 8

16 Precuneus 23 478 L 16 60 36


40 555 L 64 48 36















Conclusions








GoodndashBad Only


47 543 L 44 28 12






METHODS

Participants


Materials



Procedure










fMRI Parameters


Data Analysis









47 502 R 48 32 4




20 466 R 48 4 44

22 Angular gyrus 39 474 R 48 60 28

GoodndashBad Only






21 Precuneus 31 344 L 12 60 28


20 636 R 60 24 16


21 594 R 60 44 4


21 527 L 52 40 4

162 Angular gyrus 39 763 R 36 60 28


35 553 L 28 24 24






Acknowledgments




Note


REFERENCES
















































DISCUSSION























50 Amygdala na 443 L 20 4 16



11 464 R 12 32 20


9 444 R 8 48 28



37 479 R 56 72 8


39 512 L 48 64 28


19 454 L 40 88 32

35 Vermis na 538 mid 0 68 40

40 Angular gyrus 3940 480 R 40 52 32





27 Temporal pole 38 518 R 36 24 36

135 Precuneus 23 495 R 12 52 28

49 Cerebulum 18 586 L 12 56 12

GoodndashBad Only


11 668 R 24 24 16

28 Operculum 38 407 L 60 8 0


20 626 R 52 32 16


48 478 R 60 4 8


45 453 L 44 36 8

16 Precuneus 23 478 L 16 60 36


40 555 L 64 48 36















Conclusions








GoodndashBad Only


47 543 L 44 28 12






METHODS

Participants


Materials



Procedure










fMRI Parameters


Data Analysis









47 502 R 48 32 4




20 466 R 48 4 44

22 Angular gyrus 39 474 R 48 60 28

GoodndashBad Only






21 Precuneus 31 344 L 12 60 28


20 636 R 60 24 16


21 594 R 60 44 4


21 527 L 52 40 4

162 Angular gyrus 39 763 R 36 60 28


35 553 L 28 24 24






Acknowledgments




Note


REFERENCES

























































50 Amygdala na 443 L 20 4 16



11 464 R 12 32 20


9 444 R 8 48 28



37 479 R 56 72 8


39 512 L 48 64 28


19 454 L 40 88 32

35 Vermis na 538 mid 0 68 40

40 Angular gyrus 3940 480 R 40 52 32





27 Temporal pole 38 518 R 36 24 36

135 Precuneus 23 495 R 12 52 28

49 Cerebulum 18 586 L 12 56 12

GoodndashBad Only


11 668 R 24 24 16

28 Operculum 38 407 L 60 8 0


20 626 R 52 32 16


48 478 R 60 4 8


45 453 L 44 36 8

16 Precuneus 23 478 L 16 60 36


40 555 L 64 48 36















Conclusions








GoodndashBad Only


47 543 L 44 28 12






METHODS

Participants


Materials



Procedure










fMRI Parameters


Data Analysis









47 502 R 48 32 4




20 466 R 48 4 44

22 Angular gyrus 39 474 R 48 60 28

GoodndashBad Only






21 Precuneus 31 344 L 12 60 28


20 636 R 60 24 16


21 594 R 60 44 4


21 527 L 52 40 4

162 Angular gyrus 39 763 R 36 60 28


35 553 L 28 24 24






Acknowledgments




Note


REFERENCES























































Conclusions








GoodndashBad Only


47 543 L 44 28 12






METHODS

Participants


Materials



Procedure










fMRI Parameters


Data Analysis









47 502 R 48 32 4




20 466 R 48 4 44

22 Angular gyrus 39 474 R 48 60 28

GoodndashBad Only






21 Precuneus 31 344 L 12 60 28


20 636 R 60 24 16


21 594 R 60 44 4


21 527 L 52 40 4

162 Angular gyrus 39 763 R 36 60 28


35 553 L 28 24 24






Acknowledgments




Note


REFERENCES












































METHODS

Participants


Materials



Procedure










fMRI Parameters


Data Analysis









47 502 R 48 32 4




20 466 R 48 4 44

22 Angular gyrus 39 474 R 48 60 28

GoodndashBad Only






21 Precuneus 31 344 L 12 60 28


20 636 R 60 24 16


21 594 R 60 44 4


21 527 L 52 40 4

162 Angular gyrus 39 763 R 36 60 28


35 553 L 28 24 24






Acknowledgments




Note


REFERENCES















































fMRI Parameters


Data Analysis









47 502 R 48 32 4




20 466 R 48 4 44

22 Angular gyrus 39 474 R 48 60 28

GoodndashBad Only






21 Precuneus 31 344 L 12 60 28


20 636 R 60 24 16


21 594 R 60 44 4


21 527 L 52 40 4

162 Angular gyrus 39 763 R 36 60 28


35 553 L 28 24 24






Acknowledgments




Note


REFERENCES













































Acknowledgments




Note


REFERENCES











































implicit and explicit evaluation: fmri correlates

Documents