The effect of graded aversive stimuli on limbic and visual activation

The e�ect of graded aversive stimuli on limbic and visualactivation

p

Stephan F. Taylora,*, Israel Liberzona, b, Robert A. Koeppec

aDepartment of Psychiatry, University of Michigan Medical Center, UH 9D Box 0118, 1500 E. Medical Center Drive, Ann Arbor, MI 48109-0118,

USAbDepartment of Psychiatry, Veterans Administration Medical Center, Ann Arbor, Michigan, USA

cDepartment of Internal Medicine, Division of Nuclear Medicine, University of Michigan, Michigan, USA

Received 26 July 1999; received in revised form 5 January 2000; accepted 21 January 2000

Abstract

Activation studies have shown that in response to evocative visual stimuli, brain activity increases in the visual cortex and

limbic areas. However, non-a�ective characteristics of these images, such as color composition and visual complexity, confoundthe interpretation of these results. To address this issue, we had subjects rate over 100 images on aversive intensity (facialmutilation, dead bodies) and semantic complexity (number of objects subjects could name). From these images, we assembled

digitized image sets of non-aversive, mild and strong intensity, balanced on semantic complexity and content (human faces and®gures), and adjusted for color composition. A fourth condition consisted of a ®xation cross on a blank screen. Thirteensubjects underwent eight positron emission tomography scans using the [15O] water methodology. Measurement of skin

conductance was recorded simultaneously. All picture conditions, relative to the blank screen, activated the amygdalae andbilateral orbitofrontal cortex, while we found activation trends associated with increasing aversive content in the sub-lenticularregion. Skin conductance increased during all picture conditions. Relative to the non-aversive pictures, aversive image contentcaused modulation of occipital and occipital-temporal cortex. These results demonstrated activation of the amygdala to salient,

arousing stimuli, and not just aversive stimuli. In addition, they suggest that pictorial complexity, as indexed by our semanticmeasure, does not account for the modulation of visual cortex by aversive, emotional stimuli. 7 2000 Elsevier Science Ltd. Allrights reserved.

Keywords: Emotion; Amygdala; Occipital cortex; Positron emission tomography; Limbic system

1. Introduction

The use of emotional probes coupled with func-tional neuroimaging technologies can provide apotent tool to map neural pathways involved inemotional responses. Probes may consist either ofevocative stimuli or human faces. While several

groups have taken advantage of well-characterizedfacial expressions [4,29,33], it remains an experimen-tal question whether or not the perception of anemotional signal recruits the same anatomicalregions as the direct experience of that emotion. Inorder to elicit a direct emotional response, picturesor ®lms designed to evoke speci®c emotions, suchas sadness, happiness, or disgust, have elicited acti-vation of limbic regions, such as orbitofrontal cor-tex, temporal poles, amygdaloid complexes andinsula [18,20,22,23,36]. With regard to the amygdala,a structure associated with emotions in animal stu-dies and human lesion studies [1,8], activation hasbeen reported for aversive images compared to neu-

Neuropsychologia 38 (2000) 1415±1425

0028-3932/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved.

PII: S0028-3932(00 )00032 -4

www.elsevier.com/locate/neuropsychologia

pThis work was previously presented at the 5th International Con-

ference on Functional Mapping of the Human Brain, in Dusseldorf,

Germany in June of 1999.

* Corresponding author. Tel.: +1-734-936-9760; fax: +1-734-936-

4401.

E-mail address: [email protected] (S.F. Taylor).

tral images [18,30,42,43], to happy as well as sad®lms [21], only to aversive images [18], or not atall [20,31]. The question of whether the amygdalaresponds only to negative stimuli, or diverse cat-egories of stimuli, was raised by Breiter et al., whofound that neutral and fearful faces activated theposterior amygdaloid areas [4].

Areas outside traditional limbic regions also playa role in the brain's response to a�ective content.Some groups, including our own, have found thatemotional stimuli activate visual cortex[20,22,23,36,42]. Lane et al. found activation in thevisual cortex in response to unpleasant evocativeimages compared to neutral pictures [22], and theyfound no visual activation for pleasant pictures,relative to neutral ones. On the other hand, Langet al. did ®nd that both pleasant and unpleasantvisual stimuli activated posterior, visual processingregions [23]. One possible interpretation of thesestudies is that a general, arousal-like response toemotional stimuli mediates activation. However, inall of these studies, no speci®c e�orts were made toaddress the confound of visual complexity, e.g.matching the neutral and aversive stimuli on thenon-emotional content of the image. For instance,in debrie®ng interviews, subjects have reportedgreater di�culty identifying objects in the visualscene for aversive pictures, relative to non-aversiveones (Taylor and Liberzon, unpublished obser-vations). While clearly a problem in the interpret-ation of activation in the visual cortex, di�erentialcomplexity also presents a problem in the interpret-ation of limbic activation, since salience, in general,may drive activity in structures such as the amyg-dala [10,19].

In the present study, we have sought to addressthese issues in several ways. We have employed a para-metric design, in which we have systematically variedthe aversive content of evocative stimuli. Finding astep-wise increase in activity across three or more con-ditions of increasing aversiveness would reinforce theconnection between processing of aversive content andspeci®c regional activity, a strategy successfullyemployed in studies of face processing [29,33]. In ad-dition to the parametric design, we also sought to con-trol for e�ects of general stimulus complexity andphysical characteristics of the images by analyzing andmatching the stimulus sets on relevant measures. As a®nal control, we introduced a condition with minimalvisual content to test the hypothesis, suggested by Brei-ter et al. [4], that at least part of the amygdala wouldrespond to all complex visual stimuli. As our resultsshow, we did ®nd evidence for graded activation in thevisual cortex and sublenticular area, and we alsofound activation in the posterior amygdala in responseto complex visual stimuli without aversive content.

2. Methods

2.1. Subjects

From community advertisements, we recruited four-teen subjects who were without serious medical/neuro-logical illness, not taking medication and free of allAxis I psychiatric disorders according to a StructuredClinical Interview for Diagnosis [11]. The mean age ofthe subjects was 32.1 years (SD=13.1 years), and thesample included ®ve women. The purpose and risks ofthe study were explained to all subjects, who gave writ-ten informed consent to participate in the protocol, asapproved by the local institutional review board andthe committee on human use of radioisotopes.

2.2. Graded image preparation

Three sets of images with di�ering degrees of aver-sive content, from completely non-aversive to veryaversive, were selected. We matched the image sets oncontent items not related to emotional valence, such asthe presence of faces and human ®gures, and ameasure of overall visual content. We began by select-ing 115 images, including photographic slides from theInternational A�ective Picture System (IAPS; [24]). Toobtain enough images of su�cient aversive content, weselected additional images from pathology and surgicaltextbooks with similar aversive content (e.g. facialmutilation, corpses) to supplement the IAPS (01/3 ofthe 100 images).

The selection of graded images proceeded in twosteps. To determine aversive content and measures ofvisual complexity, we obtained subjective ratings of allthe images from 11 subjects who did not participate inthe PET study. In this pilot study, subjects ®rst viewedeach picture on a back-projecting slide-projector for10 s and named as many objects as they could identifyin the image. Responses were recorded and later codedto give an index of semantic content for each slide(average number of labels per image). Subjects viewedthe slides a second time to rate each slide on aversivecontent on a 5-point scale (1=``not unpleasant'',5=``extremely unpleasant''). From the set of 115slides, we selected three subsets, of 18 images each,strati®ed according to the degree of aversive contentobtained in the pilot phase (mean aversiveness ratings,2SD; non-aversive=1.0120.05; Mildly aversive=2.692 0.59; strongly aversive=4.06 2 0.58). We carefullymatched these subsets with respect to semantic content(mean number of labels 2 SD; non-aversive=4.02 20.93; mildly aversive=3.93 2 0.77; strongly aver-sive=3.9920.37). In addition, we matched each subsetfor human faces and human ®gures. We selected non-aversive images with moderately high arousal levels forthe IAPS slides, and, as expected, signi®cant di�er-

S.F. Taylor et al. / Neuropsychologia 38 (2000) 1415±14251416

ences on arousal did distinguish the image subsets(mean arousal2SD; non-aversive=3.8820.85; mildlyaversive=5.6020.61; strongly aversive=7.1020.21).

Lastly, images were digitized, and color adjustmentswere performed to match the subsets with respect toluminance and color. Using Photoshop 3.0 (AdobeSystems), we calculated median pixel values for overallluminance and three color channels (red, green, blue).We then performed color adjustments on the images inorder to bring the sets into the same ranges for overallluminance and for each color channel (red, green andblue).

2.3. Behavioral task

Subjects viewed pictures from the three sets ofgraded aversive content, plus a ®xation cross on ablank screen, for a total of four conditions in blockedpresentation (blank, non-aversive, mildly aversive andstrongly aversive ). The blank condition provided acomparison against which to evaluate the non-aversivepicture conditions for limbic activation. Subjects wereinstructed to view each picture and make a rating of``unpleasantness'', as in the pilot study. Digitized pic-tures were displayed on a 150 computer monitor, sus-pended approximately 180 from the subject. A stimulusdisplay program (SuperLab, Cedrus, Inc) presentedeach picture for 5 s, and vocal responses were recordedvia a microphone. For the blank condition, subjectswere instructed to attend to the ®xation cross and saythe number `1' every time the ®xation cross changedcolor, which it did every 5 s.

After a brief practice session to ensure understand-ing of the instructions, subjects viewed a set of non-aversive images under conditions identical to an actualscan, i.e. injection of radiotracer and presentation ofstimuli occurred, but image acquisition was not per-formed. Instead, a small dose of radiotracer (8 mCi)was used to time the arrival of the bolus in the brain,and the ``sham'' scanning conditions allowed subjectsto acclimate to the scanner environment. For the eightactual scans, subjects viewed one set of each of thefour conditions, in pseudorandomized order, followedby another set of the four conditions in the identicalorder.

2.4. Psychophysiological recording

Psychophysiological data were recorded using anMP-100 psychophysiological monitoring system (Bio-Pac Systems, Santa Barbara, CA). Skin conductancewas recorded using Ag/AgCl electrodes ®lled with iso-tonic NaCl unibase electrolyte attached to the volarsurface of the second phalanx of the ®rst and third ®n-gers of the left hand. Ampli®cation utilized a constantvoltage technique to measure absolute conductance,

calibrated to 20 mmhos/volt. To permit monitoring ofgaze deviation away from the images, in six subjectswe also recorded horizontal electro-oculograms(EOG), by means of Ag/AgCl electrodes a�xed to thelateral canthus. To calibrate EOG voltage change witheye-position, subjects tracked a target within theboundaries of the monitor. The MP-100 module per-formed analog-to-digital conversion of the ampli®edsignals and passed data to a Macintosh computer run-ning Acknowledge software (BioPac Systems, SantaBarbara, CA) for o�-line analysis of waveforms.

2.5. PET image acquisition

The subjects underwent PET scans in a Siemens CTI931/08-12 scanner (CTI Inc, Knoxville, TN) whichacquires 15 slices over an axial length of 010 cm.Before emission scanning, subjects had a transmissionscan measured by external 68Ge/68Ga ring sources.Lights were dimmed and ambient noise (from coolingfans) was minimal. For each emission scan, subjectswere given an iv bolus injection of 50 mCi of[15O]H2O. Data were collected in a single 60-s framebeginning 5 s after the arrival of radioactivity in thebrain. Eight PET scans separated by 10-min intervalswere acquired for each subject. Stimuli presentationbegan 15 s prior to image acquisition, and continuedfor the ®rst 30 s of each scan (uptake phase of the tra-cer), so as to maximize changes in the PET signals as-sociated with the task of interest [6]. The second 30 sof each scan always contained blank stimuli with the®xation cross.

2.6. Data analysis

Analysis of the PET data ®rst employed a standard-ization process for each image which enabled aver-aging of image data both within and across subjects.Scans were ®rst reconstructed with a Parzen ®lter witha 0.45 Hz cuto� to yield an e�ective full width at half-maximum (FWHM) of 10±11 mm. Automated rou-tines proportionally normalized images to the mean ofglobal activity (arbitrarily set to 1000), aligned imageswithin subjects (to the 4th scan) and then applied anon-linear transformation to each image to standardizeindividual anatomy according to a reference brain [26±28], which corresponds to the 1988 Talairach andTournoux brain [41]. Standardized images weresmoothed with a 3D Gaussian ®lter, yielding a ®nal,e�ective FWHM of around 12±14 mm. Mean di�er-ences between conditions for each voxel for all subjectswere expressed as Z-scores, using the between-subjectsvariance averaged among all the analyzed voxels as anerror estimate [12,44].

The analysis used a priori regions of interest(ROI), as well as hypothesis-generating searches of

S.F. Taylor et al. / Neuropsychologia 38 (2000) 1415±1425 1417

the entire image set. The regions of interest weredetermined by our own experience with this acti-vation paradigm and published reports with similarparadigms (reviewed above). The ®rst step involvedthe placement of spherical ROIs placed at stereotac-tic coordinates of the anatomically standardizedimages: bilateral amygdaloid complexes de®ned bythe Talairach and Tournoux atlas (15.75 mm diam-eter; x, y, z=224, ÿ4, ÿ16); thalamus & hypothala-mus (15.75 mm; ÿ6, ÿ6, 0; ÿ1, ÿ8, ÿ14) derived fromour prior work (Liberzon et al. [25]). One-way,repeated measures ANOVAs were used to test for sig-ni®cant activation over all four conditions. We alsotested for linear trends (ordering the conditions: blank,non-aversive, mildly aversive and strongly aversive)using single degree-of-freedom polynomial contrasts.This provided the most stringent test to reject that hy-pothesis that no limbic activation occurred in our task.However, this ROI approach has poor sensitivity,since slight sample variation in anatomic location maylead to Type II errors.

In the second step, we performed a more sensitivevoxel-map analysis which would identify limbic fociwhere they occurred in our sample. We obtained Z-score maps of di�erence between the conditions,searching for activation foci within the followingareas of interest: amygdaloid regions, ventral stria-tum, hypothalamus, thalamus, insula, ventromedialfrontal cortex, anterior cingulate cortex and occipitalcortex, plus adjacent temporal cortex known to berecruited by ventral visual processing streams(``visual cortex''). Within these areas, we accepted

any foci with Z > 3.09 (P < 0.001, uncorrected).Within a FWHM of any activation foci in our regionsof interest, we also reported activation trends for fociwith 3.09 r Z > 2.33, in order to minimize Type IIerrors. We also searched for negative Z-scores (Z <ÿ2.33), i.e. where activity was greater in the mild rela-tive to the strong condition, since we would not expectto ®nd negative peaks in our regions of interest forgraded activation.

In addition to foci within these regions, we alsoreported any foci outside the a priori regions whichexceeded the image-wide threshold of signi®cance,generally corresponding to Z > 4.4±4.9. In order todisplay the spatial extent of activation foci, for all®gures, we have used a liberal threshold of Z > 1.65(P < 0.05) which conveys important informationabout foci that just missed chosen cuto� thresholds,permitting comparison with other studies.

Analysis of region of interest data was performedwith Systat (Systat, Inc, Evanston, IL) and StatView(Abacus Concepts, Berkeley, CA) statistical software.

3. Results

3.1. Behavioral results

Subjects tolerated exposure to the images withoutsigni®cant di�culty, with the exception of one subjectwho averted his gaze from the mildly and stronglyaversive images and elected to leave the study aftercompleting six scans. Because he did not maintain hisgaze on the images, we excluded his data from theanalysis. Analysis of the horizontal EOG data from sixsubjects indicated that these subjects maintained theirhorizontal gaze within the bounds of the monitorscreen. One other subject fell asleep during one run ofthe blank condition, and this scan was not included inthe analysis. Online ratings of aversive stimuli veri®edthe intended manipulation of content: 1.13 2 0.14(mean2SD) for the non-aversive pictures; 3.3520.70for the mildly aversive pictures; 4.20 2 0.93 for thestrongly aversive pictures. This was a highly signi®cantmain e�ect of condition (F[2,24]=186, P < 0.0001),and post-hoc testing also revealed a signi®cant di�er-ence between the strong and mild stimuli (t[12]=8.6, P< 0.0001), and between the mild and non-aversivestimuli (t[12]=13.7, P<0.0001).

3.2. Psychophysiological results

The subjects showed similar magnitudes of increasein skin conductance for the three picture conditions, asdepicted in Fig. 1. Non-parametric analysis of variancerevealed a signi®cant e�ect of condition (Kruskal±Wallis H = 8.09, P = 0.02), a ®nding due largely to

Fig. 1. Skin conductance measurement shows the increase in conduc-

tance from baseline, prior to stimuli presentation, during each con-

dition.


the di�erence between the picture conditions and theblank condition.

3.3. PET activation in limbic regions

Using a priori amygdaloid regions, we con®rmed anoverall e�ect of conditions in the left (F[3,36]=4.49, P= 0.003) and right (F[3,36]=4.28, P = 0.01) amygda-loid regions (Fig. 2). Each amygdaloid region alsoexhibited a signi®cant, positive linear trend, fromblank to the non-aversive to the aversive conditions(left: F[1,12]=9.46, P=0.01; right: F[1,12]=4.67, P=0.05). As with the skin conductance data, this e�ectwas driven largely by the di�erence between the blankcondition and the picture conditions.

Table 1 shows the comparisons of the mild andstrong conditions with the non-aversive conditions andthe analysis for graded limbic activation. Comparingthe non-aversive to the mild conditions, activation

occurred in the left anterior insula and left sublenticu-lar region. Comparing the strongly aversive to thenon-aversive conditions, a focus appeared in the rightanterior insula/sublenticular region, and activation inthe left insula fell just below our signi®cance threshold.The activity in the right insula was contiguous with asub-lenticular focus, near the substantia innominata(extended amygdala). We noted a homologous left-sided focus in the sub-lenticular area (coordinates:ÿ21, 5, ÿ18), evident in Fig. 3A, but below our signi®-cance threshold (Z = 2.91). In the comparison of thestrong and mild conditions, activation trends appearedin the left sublenticular region, 11±14 mm more me-dial, but with overlap evident on Fig. 3A. We alsonoted an activation trend in the right insular region.There where no negative trends in this comparison, i.e.where the mild condition had more activity than thestrong condition, consistent with the interpretation ofgraded activation.

The a priori VOI analysis suggested that limbic acti-vation occurred in the comparison of the picture con-ditions with the blank conditions; thus we alsoanalyzed the strong and non-aversive conditions rela-tive to the blank condition (see Fig. 3B and Table 2).The voxel-map analysis demonstrated activation in bi-lateral amygdaloid regions and bilateral orbitofrontalcortex. In the comparison of the strong and blank con-ditions, we also noted hypothalamic/basal forebrainactivation. Fig. 3B also shows activation in the sublen-ticular regions, but the foci tended to blend with theamygdaloid peaks and covered the region known asthe extended amygdala. However, we did not ®nd insu-lar activation in this comparison.

Since we expected to ®nd insular activation in thestrong minus blank comparison, we performed an ROIanalysis on both insular peaks, which revealed a pat-tern of activation distinctly di�erent from the predictedmonotonic trend. Normalized activity for 13.5 mmspherical ROIs placed on each peaks is depicted inFig. 4, which shows that insular activity during the

Fig. 2. Activation is demonstrated in a priori regions of interest in

the left (L) and right (R) amygdalae. Normalized activity is shown

for spherical regions of interest placed on individual images, after

anatomic standardization. P-values indicate signi®cant F-tests for the

region of interest.

Table 1

Activation in limbic regions Ð graded aversive intensity

Mildly aversive minus non-

aversive

Strongly aversive minus non-aversive Strongly aversive minus mildly aversive

Region (x, y, z )a Z-scoreb (x, y, z )a Z-scoreb (x, y, z )a Z-scoreb

L sublenticular region ÿ24, 8, ÿ18 3.20 [ÿ21, 5, ÿ18 2.91] ÿ10, 10, ÿ18 3.44

R insula/sublenticular

region

[26, 3, 4 2.42] 26, 1, ÿ4 3.20 [26, ÿ4, ÿ7 2.84]

28, 3, 2 3.18

L insula ÿ28, 19, 2 3.43 [ÿ28, 21, 0 2.90]

a Stereotactic coordinates from Talairach and Tournoux atlas [41], right/left, anterior/posterior and superior/inferior, respectively.b Foci are listed with Z > 3.09 (P < 0.001, uncorrected). For purposes of comparison, we have also listed any focus as a trend, in brackets,

with Z>2.33 (P<0.01, uncorrected) and located within a FWHM of the peak of a focus above threshold in another contrast.


blank condition was nearly equal to or greater thanthe aversive conditions. In the comparison of theblank and non-aversive condition, we identi®ed twofoci, indicating relative decrease, in the right and leftinsula (x, y, z coordinates: 33, ÿ10, 2 and ÿ21, 3, 7,respectively; Z=ÿ6.21, Z=ÿ4.71, respectively).

3.4. PET activation outside limbic regions

Outside of limbic regions, we found signi®cant acti-vation in the visual cortex, along with signi®cantincreases from the mildly aversive pictures to thestrongly aversive pictures. Table 2 and Fig. 5 show thepattern of bilateral activity increase with the presen-

tation of mildly aversive images and strongly aversiveimages, including lingual gyrus, mid-occipital gyrus,fusiform gyrus and inferior/middle temporal gyrus.Relative to the non-aversive pictures, the stronglyaversive pictures demonstrated activation in the samepattern as the mildly aversive pictures, only withhigher magnitude at most peaks. We noted six peakswhere overlap occurred for the mildly and stronglyaversive conditions. In the direct comparison of thesetwo conditions, we found a trend level increases forstrong images in the right and left mid-occipital gyrus,consistent with graded activation. We also noted peaksin the right and left lingual gyrus, although they didnot clearly show a progression of activation from non-

Fig. 3. The ®gure shows activity changes, in the selected coronal planes (A) for the non-aversive and aversive conditions and (B) relative to the

blank condition. Numbers in the bottom row correspond to mm anterior or posterior of the anterior commissure. Activation is depicted as Z-

scores on the color scale, displayed for Z > 1.65 (uncorrected probability of P < 0.05), mapped onto a reference MRI atlas. OF: orbitofrontal

cortex; A: amygdala; I: insula; SL: sublenticular region; H: hypothalamus.


aversive to mild to strong. We did not ®nd any peaksin the visual cortex where activity was greater in themild compared to the strong conditions (Table 3).

4. Discussion

Using graded stimuli, we assessed brain activationto salient, aversive content. Exposure to aversive stim-uli modulated basal forebrain and amygdaloid regions,as predicted, and we found a step-wise increase in acti-vation associated with increasing aversive content inthe visual cortex. We found an activation focus in theleft sublenticular limbic area suggestive of graded acti-vation. The parametric approach, designed to betterisolate the psychological component associated withspeci®c regional activity, also revealed activation in theamygdalae in response to non-aversive stimuli, relative

to blank stimuli. In the discussion below, we will focuson how these ®ndings, interpreted in the context ofwork emerging from several laboratories, delineate thefunctional anatomy of emotional processing in intacthuman subjects.

The activation of the amygdala to non-aversive pic-tures extends previous functional imaging work inhumans and points to the importance of multiple func-tional roles for the amygdala. While most studies tendto focus on the amygdaloid nuclei and aversive re-sponses, such as fear, work with animal models alsoshows that the amygdalae mediate conditioned re-sponses which enhance information processing to non-aversive stimuli [10,19]. Functional neuroimaging stu-dies in humans have shown a positive correlation ofblood ¯ow in the amygdala with subsequent recall ofpleasant images [16], and in response to cues forcocaine craving [14]. Breiter et al. noted that fearfulfaces, relative to neutral faces, elicited activation in theanterior amygdala, whereas neutral faces, relative to a®xation control, elicited activity in the left posterioramygdala [4]. They speculated that the posterior regionmay respond to complex stimuli (faces in their study),whereas anterior portions reacted to speci®c categoriesof expression, e.g. fear.

This anterior±posterior distinction nicely matchesour data, in so far as activation in the anterior, sub-lenticular region occurred speci®cally to aversive con-tent, whereas more posterior activation occurred forall complex pictures. The graded activation we foundin the sublenticular areas lies in the region of theextended amygdala [2]. Also using aversive picturesfrom the IAPS, Paradiso et al. reported activation ofthis region, which they labeled as nucleus accumbens[30], one of the components of the extended amygdala.However, the limited anatomical resolution of ourtechnique, common to any imaging modality usinginter-subject averaging, should temper any speculationabout the relative roles of anterior and posterior

Table 2

Activation in limbic regions Ð comparison to blanks

Non-aversive minus blanks Strongly aversive minus blanks

Region (x, y, z )a Z-scoreb (x, y, z )a Z-scoreb

R amygdala 17, ÿ10, ÿ16 3.91 15, ÿ8, ÿ14 3.44

L amygdala ÿ28, ÿ10, ÿ16 3.26 ÿ24, ÿ4, ÿ16 3.47

Basal forebrain/hypothalamus ÿ6, ÿ4, ÿ9 3.31

R lateral orbitofrontal cortex 30, 32, ÿ16 3.65 24, 28, ÿ18 3.34

21, 32, ÿ16 3.34

L lateral orbitofrontal cortex ÿ33, 28, ÿ18 5.25 ÿ26, 28, ÿ18 4.68

ÿ24, 23, ÿ20 4.44

ÿ21, 17, ÿ22 4.13

a Stereotactic coordinates from Talairach and Tournoux atlas [41], right/left, anterior/posterior and superior/inferior, respectively.b Foci are listed with Z > 3.09 (P < 0.001, uncorrected). Foci exceeding the image-wide threshold of P < 0.05 (Z > 4.4±4.9, depending on

image) are underlined.

Fig. 4. Activity in the insula during the non-aversive condition

shows deactivation relative to the blank condition. Spherical regions

of interest were de®ned by activation in the aversive conditions rela-

tive to the non-aversive conditions.


regions. In addition, we cannot de®nitively rule out thepossibility that adjacent hippocampal or parahippo-campal regions might contribute to our observed sig-nal. Nevertheless, multiple laboratories have nowdemonstrated amygdaloid activation to non-aversivestimuli, and additional re®nements in technique will benecessary to de®nitively resolve the relevant functionalanatomy in greater detail.

One alternative explanation of our ®ndings is thatour subjects exhibited a conditioned response to thenon-aversive pictures, i.e. they anticipated or associ-ated an aversive picture with the non-aversive images.The skin conductance measurements showed equal re-sponses to all pictures, which may have re¯ected thisassociation. However, we do not believe that antici-pation or a conditioned response to the non-aversive

Fig. 5. Occipital activation for aversive stimuli increases with aversive intensity. Activations are rendered on the cortical surface of a reference

MRI, otherwise identical to Fig. 3.

Table 3

Activation peaks outside limbic regions

Mildly aversive minus non-aversive Strongly aversive minus non-

aversive

Strongly aversive minus mildly

aversive

Region (x, y, z )a Z-scoreb (x, y, z )a Z-scoreb (x, y, z )a Z-scoreb

Visual cortex

R lingual gyrus 15, ÿ89, 0 4.26 15, ÿ87, ÿ2 3.09

L lingual gyrus ÿ8, ÿ85, ÿ9 4.63 ÿ12, ÿ82, ÿ2 3.92

ÿ10, ÿ76, 7 3.11

ÿ26, ÿ73, ÿ7 3.96 ÿ24, ÿ80, ÿ7 3.10

R mid-occipital gyrus [33, ÿ80, 22 2.87] 33, ÿ78, 18 4.35 [28, ÿ76, 16 2.43]

[39, ÿ78, 16 2.86] 42, ÿ80, 16 4.11

L mid-occipital gyrus ÿ33, ÿ85, 9 4.46 ÿ35, ÿ82, 11 5.04 [ÿ28, ÿ80, 20 2.70]

R fusiform gyrus 33, ÿ60, ÿ14 3.28

L fusiform gyrus [28, ÿ42, ÿ16 2.59] ÿ28, ÿ55, ÿ14 4.02

ÿ39, ÿ67, ÿ14 3.76

ÿ46, ÿ69, ÿ11 3.81 ÿ44, ÿ69, ÿ11 3.98

L inf/middle temporal

gyrus

ÿ51, ÿ64, 2 3.49 ÿ57, ÿ64, ÿ2 3.51

R inf/middle temporal

gyrus

[60, ÿ58, ÿ4 2.73] 62, ÿ53, ÿ4 3.50

Other

Left cerebellum ÿ30, ÿ46, ÿ20 4.70

a Stereotactic conditions from Talairach and Tournoux atlas [41], right/left, anterior/posterior and superior/inferior, respectively.b Foci are listed as in Table 1. Foci exceeding the image-wide threshold of P<0.05 (Z>4.4±4.9, depending on image) are underlined.


stimuli can fully account for our ®ndings. We did com-pare subjects who experienced the non-aversive pic-tures before the aversive pictures, and they did notdi�er from those who saw the aversive pictures ®rst.Our subjects did rate the non-aversive pictures as notunpleasant, a fact arguing against a conditioned re-sponse to aversive stimuli, at least at the consciouslevel. Psychophysiological studies of the SCR to IAPSpictures demonstrate activity to salient, arousing stim-uli, regardless of emotional valence [15]. Since wematched our pictures on complexity, in the form ofsemantic information, the pictures in the non-aversiveset had moderate scores on the arousal dimension ofthe IAPS. The salience of image content in the non-aversive pictures may have accounted for the acti-vation of the amygdalae and the observed SCR.

As a structure which exchanges projections with het-eromodal and unimodal cortex, the complex of nucleiin the amygdaloid region are positioned to evaluatediverse types of stimuli [3,9,13]. Our complex, visualscenes would entail processing along the ventral aspectof the occipital cortex and temporal cortex, with eachprocessing area sending projections to the amygdaloidnuclei. By comparison, the blank condition, with onlya ®xation cross, recruited far less visual cortex thanthe picture conditions. Activation in the amygdaloidnuclei, probably lateral and baso-lateral [9], mighthave re¯ected information in¯ow from a�erent projec-tions of cortical neurons. Furthermore, the amygdaloidnuclei could process this information in the compu-tation of stimulus salience, as we have implied above,beyond the activity level elicited by a�erent projec-tions. Subjects may have found the non-aversive stim-uli more salient, ambiguous, or interesting than theblank condition, leading to amygdaloid activation.After all, a generalized response to salient stimulishould not be unexpected in a neural system involvedin the identi®cation of salient stimuli and preparationof appropriate responses.

The strongest modulation of cerebral activity byaversive content appeared in the visual cortex. Becauseseveral groups have previously reported modulation ofvisual cortex by emotional content [20,22,23,36,42], itwas important to rule out di�erences in the visualcharacteristics of the stimuli as causes of di�erentialactivation. We balanced our image sets on color spec-trum, luminance and semantic content, although itremains an open question as to how one should matchvisual characteristics and isolate a�ective componentsof an image. We did not attempt to match the imagesets on arousal, which does remain confounded withthe dimension of aversiveness. However, the pictureswith the strongest negative valences are also rated ashigh arousal in the IAPS. This suggests that arousaland negative valence may be inextricable in normalpsychology. Nevertheless, the demonstration of a

graded response from non-aversive to mildly tostrongly aversive images, for at least part of the visualcortex, suggests that physical di�erences between thepictures could not account for the activation pattern.While we cannot de®nitively rule out the possibilitythat di�erential eye-movements might have in¯uencedneuronal activity in the visual cortex, a study by Langet al. with the IAPS pictures demonstrated no di�er-ence in the duration or magnitude of scanning eyemovements when viewing these picture [23].

If activation in visual processing areas does not rep-resent physical characteristics of the stimuli, then iteither represents content-speci®c, bottom-up processingof emotion, or top-down modulatory e�ects on thevisual processing stream. If this is a bottom-up pro-cess, neurons in the visual cortex would responddirectly to emotional content. While an intriguingpossibility, we are not aware of any data to supportthis interpretation. On the other hand, several types oftop-down processing are known to occur. Anatomi-cally, projections from the amygdala extend to all theprocessing steps in the ventral visual stream, includingprimary visual cortex [3]. Arousal to the aversiveimages could have produced the step-wise e�ect wefound for the graded stimuli, which did di�er alongthe arousal dimension, taken from the IAPS. In ourown work, we have found that task instructions toperform a recognition memory task on aversive visualstimuli modulated activity in lingual gyrus [42],suggesting that top-down e�ects do in¯uence the signalin sensory cortex. Furthermore, because subjectsalways rated the pictures, which entails a cognitiveappraisal of an emotional response, we cannot saywhat e�ect this task instruction had on the activationpattern. Modulation of visual processing by selectiveattention is a well-described phenomenon [7,17], andanother possible mechanism. For example, subjectsmay have actively attended more to the emotionalimages. Seven of our 13 analyzable subjects found theaversive slides more interesting. Clearly, additional ex-perimental manipulations will be required to separatetop-down from bottom-up processing, as well as gen-eral arousal from selective attention.

Other anatomic components implicated in emotionalresponses were also activated in our study, and somere¯ected the same pattern as the amygdalae, whileothers showed a more speci®c pattern for emotionalvalence. Orbitofrontal cortex exhibited activation inthe comparison of each picture condition with theblank condition. This region of limbic cortex has beenimplicated in the maintenance of reward contingenciesin operant behavior [38,45] and has been activated inself-induced dysphoria [32], simple phobias [35] andobsessive compulsive disorder [34]. Unlike the amyg-dala, it is not essential to acquisition of conditionedfear responses, although it does exchange extensive


projections with the lateral and basolateral nuclei ofthe amygdala [3]. In humans, lesions of the orbitofron-tal cortex cause a syndrome characterized, in part, bya�ect dysregulation and poor impulse control [37,40].We also found activation of the cerebellum in theblank-to-non-aversive comparison, which ®ts withemerging data suggesting a cognitive/emotional rolefor the cerebellum [39]. Conclusions here are limited,however, since part of the cerebellum lies outside the®eld of view of our PET camera.

The activation foci in bilateral anterior insulademonstrated an intriguing pattern in association withstimulus content. The aversive pictures evoked subjec-tive feelings of disgust in our subjects, and the ®ndingsagree with prior work showing activations for subjectsviewing faces with expressions of disgust, relative tofaces with neutral expressions [33] and during aversiveconditioning [5]. With projections to and from theamygdala [3], the insula is positioned to mediateemotional reactions. Interestingly, both insulae exhib-ited an apparent decrease in activity, from the blankto the non-aversive condition. This pattern of activitymight re¯ect a decrease to the non-aversive images,rather than an increase to the aversive images. What-ever the explanation for these counter-intuitive results,they do illustrate that apparently straightforwarddi�erence measurements may conceal a more complexrelationship only evident when more than one refer-ence condition is employed.

In summary, we have demonstrated multiple com-ponents to the activation pattern observed with aver-sive visual stimuli. Graded activation linked tostimulus valence occurs in parts of the visual cortex,and sublenticular regions. However, the demonstrationof activation in limbic regions for non-aversive stimulisuggests that general stimulus salience also drives ac-tivity patterns, possibly in di�erent limbic regions. Asthe response to salient stimuli is one of the hallmarksof an emotional system, this component deserves ad-ditional study. As the response to salience confoundsthe attempt to study speci®c emotional responses, thiscomponent needs to be carefully accounted for or iso-lated in the design of activation studies.

Acknowledgements

The authors gratefully acknowledge assistance inconstruction of the stimulus sets and data collectionfrom Laura Decker and David Badre. This work wassupported by grants from the National Alliance forResearch in Schizophrenia and Depression and theNational Institute of Mental Health (K08 MH01258)to SFT. Institutional support was also provided by theAnn Arbor Veterans Administration Medical Center.

References

[1] Adolphs R, Tranel D, Damasio H, Damasio AR. Fear and the

human amygdala. Journal of Neuroscience 1995;15:5879±91.

[2] Alheid GF, Heimer L. New perspectives in basal forebrain or-

ganization of special relevance for neuropsychiatric disorders:

The striatopallidal, amygdaloid, and corticopetal components of

substantia innominata. Neuroscience 1988;27:1±39.

[3] Amaral DG, Price JL. Amygdalo-cortical projections in the

monkey (Macaca fascicularis ). Journal of Comparative

Neurology 1984;230:465±96.

[4] Breiter HC, Etco� NL, Whalen PJ, Kennedy WA, Rauch SL,

Buckner RL, Strauss MM, Hyman SE, Rosen BR. Response

and habituation of the human amygdala during visual proces-

sing of facial expression. Neuron 1996;17:875±87.

[5] Buchel C, Morris J, Dolan RJ, Friston KJ. Brain systems med-

iating aversive conditioning: an event-related fMRI study.

Neuron 1998;20:947±57.

[6] Cherry SR, Woods RP, Mazziotta JC. Improved signal-to-noise

in activation studies by exploiting the kinetics of oxygen-15-

labeled water. Journal of Cerebral Blood Flow and Metabolism

1993;13:S714.

[7] Corbetta M, Miezin FM, Shulman GL, Petersen SE. A PET

study of visuospatial attention. The Journal of Neuroscience

1993;13:1202±26.

[8] Davis M. The role of the amygdala in conditioned fear. In:

Aggleton JP, editor. The amygdala: neurobiological aspects of

emotion, memory and mental dysfunction. New York: Wiley-

Liss, 1992. p. 255±395.

[9] de Olmos JS. Amygdala. In: Paxinos G, editor. The human cen-

tral nervous system. San Diego, CA: Academic Press, 1990. p.

583±710.

[10] Everitt BJ, Morris KA, O'Brien A, Robbins TW. The basolat-

eral amygdala-ventral striatal system and conditioned place pre-

ference: further evidence of limbic-striatal interactions

underlying reward-related processes. Neuroscience 1991;42:1±18.

[11] First MB, Spitzer RL, Gibbon M, Williams J. Structured clini-

cal interview for DSM-IV axis I disorders (SCID), non-patient

version: User's guide. Washington DC: American Psychiatric

Press, 1996.

[12] Friston KJ, Frith CD, Liddle PF, Frackowiak RS. Comparing

functional (PET) images: the assessment of signi®cant change.

Journal of Cerebral Blood Flow and Metabolism 1991;11:690±9.

[13] Gallagher M, Holland PC. The amygdala complex: multiple

roles in associative learning and attention. Proceedings of the

National Academy of Sciences, USA 1994;91:11,771±6.

[14] Grant S, London ED, Newlin DB, Villemagne VL, Liu X,

Contoreggi C, Phillips RL, Kimes AS, Margolin A. Activation

of memory circuits during cue-elicited cocaine craving.

Proceedings of the National Academy of Sciences, USA

1996;93:12,040±5.

[15] Greenwald MK, Cook EW, Lang PJ. A�ective judgment and

psychophysiological response: dimensional covariation in the

evaluation of pictorial stimuli. Journal of Psychophysiology

1989;3:51±64.

[16] Hamann SB, Ely TD, Grafton ST, Kilts CD. Amygdala activity

related to enhanced memory for pleasant and aversive stimuli.

Nature Neuroscience 1999;2:289±93.

[17] Heinze HJ, Mangun GR, Burchert W, Hinrichs H, Scholz M,

Munte TF, Gos A, Scherg M, Johannes S, Hundeshagen H,

Gazzaniga MS, Hillyard SA. Combined spatial and temporal

imaging of brain activity during visual selective attention in

humans. Nature 1994;372:543±6.

[18] Irwin W, Davidson RJ, Lowe MJ, Mock BJ, Sorenson JA,

Turski PA. Human amygdala activation detected with echo-pla-


nar functional magnetic resonance imaging. NeuroReport

1996;7:1765±9.

[19] Kapp BS, Whalen PJ, Supple WF, Pascoe JP. Amygdaloid con-

tributions to conditioned arousal and sensory information pro-

cessing. In: Aggleton JP, editor. The amygdala: neurobiological

aspects of emotion, memory and mental dysfunction. New

York: Wiley-Liss, 1992. p. 229±54.

[20] Kosslyn SM, Shin LM, Thompson WL, McNally RJ, Rauch

SL, Pitman RK, Alpert NM. Neural e�ects of visualizing and

perceiving aversive stimuli: a PET investigation. NeuroReport

1996;7:1569±76.

[21] Lane RD, Reiman EM, Ahern GL, Schwartz GE, Davidson

RJ. Neuroanatomical correlates of happiness, sadness, and dis-

gust. American Journal of Psychiatry 1997;154:926±33.

[22] Lane RD, Reiman EM, Bradley MM, Lang PJ, Ahern GL,

Davidson RJ, Schwartz GE. Neuroanatomical correlates of

pleasant and unpleasant emotion. Neuropsychologia

1997;35:1437±44.

[23] Lang PJ, Bradley MM, Fitzsimmons JR, Cuthbert BN, Scott

JD, Moulder B, Nangia V. Emotional arousal and activation of

the visual cortex: an fMRI analysis. Psychophysiology

1998;35:199±210.

[24] Lang PJ, Greenwald MK. The international a�ective picture

system standardization procedure and initial group results for

a�ective judgments: Technical Report IA. Gainesville, FL:

Center for Research in Psychophysiology, University of Florida,

1988.

[25] Liberzon I, Taylor SF, Fig LM, Decker LR, Koeppe RA,

Minoshima S. Limbil activation and psychophysiologic re-

sponses to aversive visual stimuli: interaction with cognitive

task. Neuropsychopharmacology, in press.

[26] Minoshima S, Berger KL, Lee KS, A MM. An automated

method for rotational correction and centering of three-dimen-

sional functional brain images. Journal of Nuclear Medicine

1992;33:1579±85.

[27] Minoshima S, Koeppe RA, Frey KA, Kuhl DE. Anatomic stan-

dardization: Linear scaling and nonlinear warping of functional

brain images. Journal of Nuclear Medicine 1994;35:1528±37.

[28] Minoshima S, Koeppe RA, Mintun MA, Berger K, Taylor SF,

Frey KA, Kuhl DE. Automated detection of the intercommis-

sural (AC-PC) line for stereotactic localization of functional

brain images. Journal of Nuclear Medicine 1993;34:322±9.

[29] Morris J, Frith CD, Perrett DI, Rowland D, Young AW,

Calder AJ, Dolan RJ. A di�erential neural response in the

human amygdala to fearful and happy facial expressions.

Nature 1996;383:812±5.

[30] Paradiso S, Johnson DL, Andreasen NC, O'Leary DS, Watkins

GL, Boles Ponto LL, Hichwa RD. Cerebral Blood Flow

Changes Associated With Attribution of Emotional Valence to

Pleasant, Unpleasant, and Neutral Visual Stimuli in a PET

Study of Normal Subjects. American Journal of Psychiatry

1999;156:1618±29.

[31] Paradiso S, Robinson RG, Andreasen NC, Downhill JE,

Davidson RJ, Kirchner PT, Watkins GL, Ponto LLB, Hichwa

RD. Emotional activation of limbic circuitry in elderly normal

subjects in a PET study. American Journal of Psychiatry

1997;154:384±9.

[32] Pardo JV, Pardo PJ, Raichle ME. Neural correlates of self-

induced dysphoria. American Journal of Psychiatry

1993;150:713±9.

[33] Phillips ML, Young AW, Senior C, Brammer M, Andrew C,

Calder AJ, Bullmore ET, Perrett DI, Rowland D, Williams

SCR, Gray JA, David AS. A speci®c neural substrate for per-

ceiving facial expressions of disgust. Nature 1997;389:495±8.

[34] Rauch SL, Jenike MA, Alpert NM, Baer L, Breiter HC, Savage

CR, Fischman AJ. Regional cerebral blood ¯ow measured

during symptom provocation in obsessive-compulsive disorder

using oxygen 15-labeled carbon dioxide and positron emission

tomography. Archives of General Psychiatry 1994;51:62±70.

[35] Rauch SL, Savage CR, Alpert NM, Miguel EC, Baer L, Breiter

HC, Fischman AJ, Manzo PA, Moretti C, Jenike MA. A posi-

tron emission tomographic study of simple phobic symptom

provocation. Archives of General Psychiatry 1995;52:20±8.

[36] Reiman EM, Lane RD, Ahern GL, Schwartz GE, Davidson

RJ, Friston KJ, Yun L-S, Chen K. Neuroanatomical correlates

of externally and internally generated human emotion.

American Journal of Psychiatry 1997;154:918±25.

[37] Reitman F. Orbital cortex syndrome following leucotomy.

American Journal of Psychiatry 1946;103:238±41.

[38] Rolls ET, Yaxley S, Sienkiewicx ZJ. Gustatory responses of

single neurons in the orbitofrontal cortex of the macaque

monkey. Journal of Neurophysiology 1990;64:1055±66.

[39] Schmahmann JD, Sherman JC. The cerebellar cognitive a�ec-

tive syndrome. Brain 1998;121:561±79.

[40] Starkstein SE, Robinson RG. The role of the frontal lobes in

a�ective disorder following stroke. In: Levin HS, Eisenberg

HM, Benton AL, editors. Frontal lobe function and dysfunc-

tion. New York: Oxford University Press, 1991. p. 288±303.

[41] Talairach J, Tournoux P. A co-planar stereotaxic atlas of a

human brain. Stuttgart: Thieme-Verlag, 1988.

[42] Taylor SF, Liberzon I, Fig LM, Decker LR, Minoshima S,

Koeppe RA. The e�ect of emotional content on visual recog-

nition memory: A PET activation study. NeuroImage

1998;8:188±97.

[43] Whalen PJ. Fear, vigilance, and ambiguity: Initial neuroimaging

studies of the human amygdala. Current Directions in

Psychological Science 1998;7:177±88.

[44] Worsley KJ, Evans AC, Marrett S, Neelin P. A three-dimen-

sional statistical analysis for CBF activation studies in human

brain. Journal of Cerebral Blood Flow and Metabolism

1992;12:900±18.

[45] Zald DH, Kim SW. Anatomy and function of the orbital fron-

tal cortex, I: Anatomy, neurocircuitry, and obsessive compulsive

disorder. Journal of Neuropsychiatry 1996;8:125±38.


The effect of graded aversive stimuli on limbic and visual activation

Documents

Transcript of The effect of graded aversive stimuli on limbic and visual activation