PLEASE SCROLL DOWN FOR ARTICLE

!"#$%&'(#)*+%,&$%-.,/*.&-+-%012%34#**5.'+6%7#**#&8%9:%;:<=/2%>?%@+0'A&'1%>BCBD))+$$%-+(&#*$2%D))+$$%9+(&#*$2%3$A0$)'#E(#./%/A80+'%FC?>G?GG?<HA0*#$"+'%H$1)".*.51%H'+$$I/J.'8&%K(-%L+5#$(+'+-%#/%M/5*&/-%&/-%7&*+$%L+5#$(+'+-%NA80+'2%CBO>F?P%L+5#$(+'+-%.JJ#)+2%Q.'(#8+'%R.A$+6%GOSPC%Q.'(#8+'%;('++(6%K./-./%7C!%GTR6%U4

V.5/#(#W+%N+A'.$)#+/)+HA0*#)&(#./%-+(&#*$6%#/)*A-#/5%#/$('A)(#./$%J.'%&A(".'$%&/-%$A0$)'#E(#./%#/J.'8&(#./2"((E2XX,,,:#/J.'8&,.'*-:).8X$8EEX(#(*+Y)./(+/(Z(FBF??FPC>

V+'+0'&*%).''+*&(+$%.J%&815-&*&%'+$E./$+$%-A'#/5%/./S)./$)#.A$E+')+E(#./%.J%J&)#&*%&JJ+)(%#/%&-.*+$)+/(%&/-%E'+S&-.*+$)+/(%)"#*-'+/7#**#&8%9:%;:%4#**5.'+%&[%9+0.'&"%D:%\A'5+*A/S!.--%0&%R&'W&'-%Q+-#)&*%;)"..*6%]+*8./(6%QD6%U;D%0%U/#W+'$#(1%.J%U(&"%;)"..*%.J%Q+-#)#/+6%;&*(%K&^+%V#(16U!6%U;D

@#'$(%EA0*#$"+-%./2%>G%;+E(+80+'%>BBF

!.%)#(+%("#$%D'(#)*+%4#**5.'+6%7#**#&8%9:%;:%&/-%\A'5+*A/S!.--6%9+0.'&"%D:_>BCB`%aV+'+0'&*%).''+*&(+$%.J%&815-&*&'+$E./$+$%-A'#/5%/./S)./$)#.A$%E+')+E(#./%.J%J&)#&*%&JJ+)(%#/%&-.*+$)+/(%&/-%E'+S&-.*+$)+/(%)"#*-'+/a6%V.5/#(#W+N+A'.$)#+/)+6%C2%C6%GG%b%PG6%@#'$(%EA0*#$"+-%./2%>G%;+E(+80+'%>BBF%_#@#'$(`!.%*#/^%(.%("#$%D'(#)*+2%9=I2%CB:CBcBXCO?ccF>BFBG>PGF?OULK2%"((E2XX-d:-.#:.'5XCB:CBcBXCO?ccF>BFBG>PGF?O

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial orsystematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply ordistribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contentswill be complete or accurate or up to date. The accuracy of any instructions, formulae and drug dosesshould be independently verified with primary sources. The publisher shall not be liable for any loss,actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directlyor indirectly in connection with or arising out of the use of this material.

COGNITIVE NEUROSCIENCE, 2010, 1 (1), 33–43

© 2009 Psychology Press, an imprint of the Taylor & Francis Group, an Informa businesswww.psypress.com/cognitiveneuroscience DOI: 10.1080/17588920903243957

PCNS Cerebral correlates of amygdala responses during non-conscious perception of facial affect in

adolescent and pre-adolescent children

Adolescent Masked Affect William D. S. KillgoreHarvard Medical School, Belmont, MA, USA

Deborah A. Yurgelun-ToddUniversity of Utah School of Medicine, Salt Lake City, UT, USA

During nonconscious perception of facial affect, healthy adults commonly activate a right-lateralized pathwaycomprising the superior colliculus, pulvinar, and amygdala. Whether this system is fully developed prior to adult-hood is unknown. Twenty-three healthy adolescents underwent functional magnetic resonance imaging (fMRI)while viewing fearful, angry, and happy faces, backward masked by neutral faces. Left amygdala activation dif-fered among the three affects, showing reductions to masked anger and increases to masked fear and happy faces.During masked fear, left amygdala activation correlated positively with extrastriate cortex and temporal poles andnegatively with precuneus and middle cingulate gyrus. Responses of the left amygdala to masked anger correlatedpositively with right parahippocampal gyrus and negatively with dorsal anterior cingulate. Amygdala responses tomasked happy faces were uncorrelated with other brain regions. Contrary to the right-lateralized pathway seen inadults, adolescents show evidence of a predominantly left-lateralized extrastriate pathway during masked presen-tations of facial affect.

Keywords: fMRI; Neuroimaging; Masked affect; Nonconscious; Face perception; Limbic system; Adolescence;Development.

INTRODUCTION

Human survival is vitally dependent on the ability todetect and respond rapidly to biologically relevantstimuli in the environment. The visual processing sys-tem comprises two separate neural pathways thatfacilitate flexible appraisal by permitting emotionallyrelevant visual information to be evaluated andresponded to at multiple levels of perceptual aware-ness (LeDoux, 1996). Specifically, evidence suggeststhat conscious visual perception of affective informa-tion relies on the geniculostriate system, a relativelyslow pathway from the retina to the lateral geniculatenucleus to the primary visual cortex and then to higherlevel visual association regions (Kimura, Yoshino,

Takahashi, & Nomura, 2004). Simultaneously, a second,more rapid, system comprising the superior collicu-lus, pulvinar nucleus, and amygdala appears to oper-ate below the level of conscious awareness to permitnear instantaneous appraisal and enhance the prepar-edness of the individual to respond to emotionallysalient stimuli (Kimura et al., 2004; LeDoux, 1996;Ohman, Carlsson, Lundqvist, & Ingvar, 2007).Because this second system bypasses the primary vis-ual cortex and association regions, it is hypothesizedto operate automatically by processing visualinformation at a very crude level of spatial resolutionwithout engaging the direct conscious awareness ofthe individual (Ohman, 2005). This non-consciousvisual perception system has been invoked to explain

Correspondence should be addressed to: William D. “Scott” Killgore, Neuroimaging Center, McLean Hospital, 115 Mill Street, Belmont,MA 02478, USA. E-mail: killgore@mclean.harvard.edu

Downloaded By: [Killgore, William D. S.] At: 19:09 25 February 2010

34 KILLGORE AND YURGELUN-TODD

the phenomenon of blindsight, a condition where cor-tically blind individuals retain the capacity to respondaccurately to some aspects of visual stimuli such asmovement or emotional content in facial expressions,despite lack of conscious perception of the stimuli(Morris, DeGelder, Weiskrantz, & Dolan, 2001).

The capacity of the brain to process affectiveinformation outside of normal conscious awarenesshas been explored using a technique known as back-ward stimulus masking. This technique presents theindividual with an emotional stimulus below the nor-mal threshold of conscious perception. For example,when an emotional facial expression is presented foran extremely brief duration (e.g., ! 30 ms) and fol-lowed immediately by a “mask” image of a face dis-playing a neutral emotion, the target emotionalexpression is usually not reliably discriminated abovechance levels (Williams et al., 2004). Nonetheless,such masked affective stimuli produce reliable pat-terns of activation within the amygdala and associatedaffect processing regions during functional neuroim-aging (Killgore & Yurgelun-Todd, 2004a; Morris,Ohman, & Dolan, 1998; Whalen et al., 1998). Maskedexpressions of fear have been by far the most exten-sively studied, with most of the evidence suggestingthat nonconscious perception of fearful faces is asso-ciated with greater activity within the amygdala rela-tive to other unseen affective expressions (Suslow et al.,2006; Whalen et al., 1998). Under normal circum-stances, this responsiveness of the amygdala mayenhance survival by increasing vigilance and appro-priate threat-related responses (LeDoux, 1996). Con-versely, extreme levels of amygdala responsivenesshave also been associated with excessive anxiety(Killgore & Yurgelun-Todd, 2005; Rauch et al.,2000). Thus, understanding this circuitry is critical tofurthering our knowledge regarding brain function,social responses, and anxiety.

Findings from previous research suggest severalpredictions regarding the functional neuroanatomythat may be activated by masked presentations offacial affect. Because masked visual stimuli arebelieved to primarily activate the superior colliculus–pulvinar extrastriate pathway, thus bypassing areasinvolved in conscious visual cortical processing, suchstimuli are expected to produce very little activationin primary visual cortex and areas involved in affectregulation such as the prefrontal cortex (Morris,Ohman, & Dolan, 1999; Ohman, 2005). While thereis considerable support for this model in healthyadults (Killgore & Yurgelun-Todd, 2004a; Morris et al.,1998; Whalen et al., 1998), it remains unclear whethernon-conscious activation of the amygdala and deacti-vation of cortical regions by masked affective stimuli

is evident prior to the age of adulthood and, therefore,consistent across development. In the present study,we examined activation of the amygdala and its func-tional correlation with other regions of the brain in asample of adolescent and pre-adolescent children view-ing a series of masked affective faces. The maskedfacial stimuli included expressions from three differ-ent affect categories, including fear, anger, and happi-ness. Based on studies of masked affect presentationsin adults, we hypothesized that the amygdala wouldrespond most strongly to negatively valenced maskedaffective presentations, would be lateralized to theright, and that this activation would be positively cor-related with activity in subcortical limbic structuresand extrastriate visual processing regions but nega-tively correlated with cortical regions involved incognitive control and affect regulation (i.e., prefrontalcortex, anterior cingulate gyrus).

METHODS

Subjects

Twenty-three healthy children (9 male; 14 female)ranging in age from 8 to 18 years (M = 13.0, SD = 3.3)volunteered to undergo functional magnetic reso-nance imaging (fMRI). This age range was selected inorder to encompass the full range of pre- to post-ado-lescent development. The age of the participants didnot differ between male (M = 12.7, SD = 3.1) andfemale (M = 13.3, SD = 3.3) groups, t(21) = 0.44, ns.Participants were predominantly right-handed (fournon-right-handed) by self-report. The children werescreened for psychopathology using a structured clini-cal interview and were included only if they were freefrom any current or past history of psychiatric illness,neurologic problems, or illicit substance or alcoholuse. Children had normal visual acuity (or correctedto normal with contact lenses) and were recruitedfrom the local community of Belmont, MA. All pro-cedures and potential risks were explained and writteninformed consent/assent was obtained from the chil-dren and their parents prior to participation. Familieswere given a small financial payment for their time.

FMRI stimulation paradigm

During fMRI, the children completed three task runsthat involved masked presentations of angry, fearful,and happy affective face in a counterbalanced order.Each run comprised five 30-s blocks that alternatedbetween masked neutral and masked affect conditions.

Downloaded By: [Killgore, William D. S.] At: 19:09 25 February 2010

ADOLESCENT MASKED AFFECT 35

Each block consisted of 10 trials each lasting 3 s. Duringeach trial, a facial expression “target” image was pre-sented for 30 ms followed immediately by a neutralfacial expression “mask” for 170 ms. This durationwas chosen based on extensive data from our ownlaboratory (Killgore & Yurgelun-Todd, 2004a, 2007)as well as detailed psychophysical studies from others(Esteves & Ohman, 1993; Williams et al., 2004). Trialswere separated by a 2800 ms period of blank screen.Both stimuli in each target–mask pair were posed bythe same actor (Ekman & Friesen, 1976). All trialswere identical except for the affect displayed on the“target” expression. Neutral trials presented a neutraltarget followed by a neutral mask whereas affectivetrials presented an affective target followed by a neu-tral mask. Participants were not informed about thebackward masking of the stimuli and were told onlythat they would observe photographs of faces pre-sented briefly and would be required to indicatewhether each face was male or female by pressing abutton with the right hand. The tasks were programmedin Psyscope 1.2.5 software (Macwhinney, Cohen, &Provost, 1997) running on a Power Macintosh G3computer. Stimuli were back-projected onto a screenplaced at the rear of the scanner and viewed via a mirrormounted on the head coil.

Neuroimaging methods

Data were collected at 3.0 T on a Siemens Trio MRIscanner and a quadrature RF head coil (TR = 3000ms, TE = 30 ms, flip angle = 90°). At the outset, threedummy images were acquired and discarded to obtaina steady state. For each run, 50 coronal echoplanarimages were acquired over 40 slices (5 mm thick,interleaved), using a 20 cm field of view and a 64 " 64acquisition matrix, providing a resolution of 3.125 " 5" 3.125 mm. Matched T1-weighted high-resolutionimages were acquired at the outset of each scanningsession.

Image processing

Data were preprocessed and analyzed in SPM99(Friston et al., 1995). Functional images were motioncorrected, realigned in three axes using the standardalgorithms in SPM99, and normalized to the three-dimensional space of the Montreal Neurological Insti-tute (MNI). Data were spatially smoothed with an iso-tropic Gaussian kernel (full width half maximum[FWHM] = 10 mm), and resliced to 2 " 2 " 2 mmvoxels using sinc interpolation.

Statistical analysis

The analysis followed a two-step random-effectsapproach in SPM99. First, functional data for eachparticipant were convolved to a boxcar waveformbased on the block design of the task (i.e., maskedaffect vs. masked neutral) and the canonical hemody-namic response function. For each individual, theblood-oxygen-level-dependent (BOLD) signal wasconvolved with the alternating neutral and affectconditions to produce a contrast image evaluatingwithin-subject effects of each masked affect vs.masked neutral baseline. These contrast images werethen analyzed with random effects models in thesecond stage of analyses. To identify amygdalaregions showing variability across the three affectconditions, a priori search territories were placed atthe location of the left and right amygdala. Thesesearch territories were defined according to the speci-fications of the automated anatomical labeling atlas(Tzourio-Mazoyer et al., 2002) as implemented withinthe WFU PickAtlas utility for SPM (Maldjian, Laurienti,Kraft, & Burdette, 2003). Next, constraining the ana-lysis to these amygdala search territories, we con-ducted a voxel-wise omnibus analysis of variance(ANOVA) in SPM99 across the three affect condi-tions (masked anger, fear, happy). Because the pur-pose of this analysis was to identify voxels of interestwithin the amygdala for further analysis, activationwas considered significant if it exceeded an uncor-rected threshold p < .05, and spatial extent of k = 10.This analysis yielded an activation cluster of 37 voxelswithin the left amygdala which was then used as aregion of interest (ROI) in subsequent analyses.Although no activation was found within the rightamygdala for the omnibus F-test, for completeness theright amygdala was also interrogated by placing amirror image ROI at the same MNI coordinates in theright hemisphere (i.e., flipping the x-axis coordi-nates). Data from these functionally defined ROIswithin the amygdala were extracted and entered into a2 (side) " 3 (anger, fear, happy) repeated measuresANOVA in SPSS 12. Furthermore, activation withineach of these ROIs was compared among the threemasked affects using paired t-tests in SPM99 at p <.05, family-wise error (FWE) corrected, k = 10. Thecorrelation between the activation within each ROIand the age of the participants was also calculated.Finally, to examine the functional relationshipbetween the amygdala and whole-brain activation foreach affect condition, functional data within the leftamygdala ROI was extracted and used as a seedregion for subsequent whole brain correlation analysisat p < .001, uncorrected, k = 20. Because of the large

Downloaded By: [Killgore, William D. S.] At: 19:09 25 February 2010

36 KILLGORE AND YURGELUN-TODD

age range in the participants, chronological age wasentered as a nuisance variable using the multipleregression option in SPM99 to control for age-relatedvariance. This was done separately for masked anger,masked fear, and masked happy conditions. Findings ofinterest were overlaid onto the canonical high-resolutionbrain image in SPM for visualization.

RESULTS

Omnibus F-test for affect conditions

Figure 1 shows that within the bilateral amygdalasearch territories, a voxelwise analysis of varianceacross the three masked affect conditions yielded asignificant cluster of activation within the left amy-gdala, 37 voxels; F(2, 65) = 6.76; MNI coordinates:x = –24, y = –6, z = –14. In contrast, no activation dif-ference was observed in the right amygdala acrossthese three masked affect conditions.

Lateralized amygdala activation by affect condition

To further clarify the nature of the group differenceswithin the left amygdala, the activation cluster identi-fied in the previous voxelwise omnibus F-test wasdefined as an ROI for subsequent group comparisons.Although the previous analysis showed no main effectof condition in the right amygdala, for completenessa mirror image ROI for the right amygdala was cre-ated by flipping the x-coordinates of first ROI. Func-tional data were extracted from these bilateral ROIsfor each of the affect conditions and subjected to a 2(hemisphere) " 3 (affect condition) repeated measuresANOVA. There was no significant main effect of

affect or hemisphere, but as evident in Figure 2, therewas a significant hemisphere " affect condition inter-action, F(2, 42) = 6.06, p = .005. Post-hoc compari-sons showed that mean BOLD signal intensity withinthe left amygdala ROI was lower for masked angerthan masked fear (p = .008) or masked happy expres-sions (p = .023). Furthermore, while masked angershowed greater deactivation within the left amygdalathan the right (p = .04), masked happy presentationswere associated with greater deactivation in the rightamygdala relative to the left (p = .004).

Voxelwise paired comparisons between affect conditions

Within the left and right ROIs, voxelwise compari-sons were tested among the three masked affect con-ditions. As is evident in Table 1, both masked fear andmasked happy conditions showed significantlygreater left amygdala activation relative to maskedanger expressions, whereas there was no significantdifference in left amygdala responses to masked fearand happy face presentations. For the right amygdala,none of the pair-wise comparisons among affect con-ditions emerged as significantly different.

Amygdala correlations with age

To examine the relationship between age and amy-gdala responses, the mean extracted data from eachamygdala ROI were entered into a Pearson correlationanalysis for each affect condition separately. For themasked anger vs. neutral condition, age was notsignificantly correlated with fMRI signal in the left(r = .14, ns) or right (r = .11, ns) amygdala ROIs.However, for the masked fear vs. neutral condition,

Figure 1. Within the bilateral amygdala search territories, a voxelwise ANOVA yielded a cluster of 37 voxels in the left amygdala that differedsignificantly across the three masked affect conditions. This cluster can be seen in axial (left panel) and coronal slices (middle panel). Parameter esti-mates extracted from the peak voxel (MNI coordinates: x = –24, y = –6, z = –14) are plotted for visualization and show a decline in left amygdala acti-vation for the masked anger vs. neutral contrast relative to the increases shown for the fear vs. neutral and happy vs. neutral contrasts (right panel).

0

0.05

0.10

0.15

–0.15

–0.10

–0.05

–0.20Anger Fear Happy

Am

ygda

la P

aram

eter

Est

imat

eL RL

R

Downloaded By: [Killgore, William D. S.] At: 19:09 25 February 2010

ADOLESCENT MASKED AFFECT 37

there was a significant negative correlation betweenparticipant age and fMRI signal in the ROI for theleft (r = –.48, p = .02) but not the right amygdala (r =–.17, ns). Finally, there was no significant correlationbetween age and amygdala ROI responses for eitherthe left (r = .10, ns) or right (r = .32, ns) sides duringthe masked happy vs. neutral condition.

Left amygdala correlations across whole brain

Mean activation from the left amygdala cluster ROIidentified by the omnibus F-test was extracted for

each subject and used as a seed region for correlationanalysis with other regions across the entire brain.After controlling for age effects, partial correlationswith whole brain were evaluated for each maskedaffect condition separately. Figure 3 shows the voxelclusters that correlated with left amygdala responsesto masked affect; the coordinates of cluster maximaare listed in Table 2.

Masked anger

During masked anger trials, activation within the leftamygdala was positively correlated with activationwithin secondary visual processing regions, includingbilateral inferior occipital gyri, as well as bilateral cer-ebellum and temporal pole regions (see Figure 3). Incontrast, left amygdala responses during maskedanger were negatively correlated with activation inregions of the supplementary motor cortex/precentralgyrus, predominantly on the right, and a large clusterof activation encompassing the middle/posterior cin-gulate gyrus and precuneus (see Figure 3).

Masked fear

As is evident in Figure 3, greater responses within theleft amygdala during the masked fear presentationswere positively correlated with a single cluster of acti-vation in the contralateral right anterior hippocampusextending inferiorly into the parahippocampal gyrus.

Figure 2. Based on the left amygdala cluster of activation produced by the omnibus ANOVA, identical mirror-image ROIs were placedwithin each amygdala. The left panel shows an axial view of the location of the ROIs (left = red; right = blue). The right panel shows that arepeated measures ANOVA on the extracted data from each ROI for each masked affect condition yielded a significant hemisphere by condi-tion interaction (p = .005).

p = 0.008

p = 0.04

p = 0.004

p = 0.02

TABLE 1 Regions showing differences in left amygdala ROI activation

among the masked anger, fear, and happy conditions via paired t-tests

ContrastActive voxels

MNI coordinates

FWE corrected px y z t-score

Anger > Fear # # # # # #Fear > Anger 23 #28

##24

#6 #12 3.13 .014Anger > Happy # # # # #Happy > Anger 11 #4 #14 2.80 .025Fear > Happy # # # # # #Happy > Fear # # # # # #

Notes: None of the comparisons listed above were significantfor the right amygdala. Significance was set at p < .05, family-wiseerror (FWE) corrected within the amygdala search territory, k = 10.

Downloaded By: [Killgore, William D. S.] At: 19:09 25 February 2010

38 KILLGORE AND YURGELUN-TODD

Activation in the left amygdala was negatively corre-lated with responses in a large cluster within the rightanterior cingulate gyrus, and a number of smaller andwidely distributed regions including the left middlecingulate gyrus, supramarginal gyrus, and middletemporal pole. Interestingly, left amygdala responsesto masked fear were also negatively correlated withactivation in the right amygdala and cerebellum.

Masked happy

No regions of the brain showed any significant posit-ive or negative correlations with left amygdalaresponses to masked happy faces.

DISCUSSION

In the present study of pre-adolescent and adolescentchildren we found that: (1) although amygdalaresponses were observed, contrary to the predominantfindings in adults, activation to masked affective faceswas more responsive in the left than in the right amy-gdala; (2) the left amygdala activation during maskedaffect presentations showed a differential pattern oflateralized activation depending on the specific affectcondition, but this pattern was not systematicallyrelated to valence; (3) the activation of the left amy-gdala correlated differentially with several brain regionsdepending on the specific masked affect condition,suggesting that the amygdala may be associated with

Figure 3. The left amygdala was used as a seed region to determine whole-brain correlations with amygdala activation during each affectcondition. The three leftmost panels in each row show maximum intensity projections or “glass brain” views, while the rightmost image showsa representative overlay of activation on the canonical brain template. After controlling for age, activation of the left amygdala during maskedanger presentations was associated with (a) increased activation within extrastriate visual cortex, cerebellum, and temporal pole regions, and(b) reduced activation within posterior medial cortex including the precuneus and middle cingulate gyrus. During masked fear presentations,left amygdala activation (controlling for age) was associated with (c) increased activation within the right parahippocampal gyrus and (d)reduced activation within the dorsal anterior cingulate gyrus, left middle cingulate gyrus, supramarginal gyrus, and middle temporal pole.

b) Anger: Amygdala Negative Correlation

a) Anger: Amygdala Positive Correlation

d) Fear: Amygdala Negative Correlation

c) Fear: Amygdala Positive Correlation

–82

–8

0

12

Downloaded By: [Killgore, William D. S.] At: 19:09 25 February 2010

ADOLESCENT MASKED AFFECT 39

modulating different aspects of affective processingdepending on the specific emotion perceived. Anumber of implications regarding affective processingduring development can be derived from these findings.

The first major finding was that the left amygdalawas the most differentially responsive to the threemasked affect conditions, whereas there was no evid-ence of such responsiveness within the right amy-gdala, regardless of whether voxelwise or ROIanalyses were performed. Several studies focusing onadult samples have reported either bilateral (Williams

et al., 2006a) or greater right amygdala activation inresponse to masked affect presentations (Kugel et al.,2008; Morris et al., 1998, 1999; Williams et al.,2006b), a finding supported by a recent meta-analysis(Costafreda, Brammer, David, & Fu, 2008). Interest-ingly, not all studies show right lateralized activationof the amygdala during masked affect presentations inadults. Notably, Whalen and colleagues presentedadult subjects with images of “eye-whites” rather thanfull facial stimuli and found significant responses inthe left amygdala for fearful relative to happy eye-whites (Whalen et al., 2004). The similarity betweenthe responses of children viewing masked full facesand adults viewing incomplete (i.e., crude) facialstimuli raises the possibility that the left amygdalaresponse pattern may reflect a more primitive ordevelopmentally immature aspect of nonconsciousfacial feature processing.

The reason for the reversed lateralization observedin the present child through adolescent sample isunclear but the outcome may reflect a developmentalphenomenon. For instance, we have previously foundthat adolescents show a pattern of reversed lateral-ized amygdala and prefrontal activation relative toboth younger children and adults viewing theovertly presented affective face stimuli (Killgore &Yurgelun-Todd, 2004b). It is well known that theperiod of adolescence is characterized by dramaticchanges in cerebral organization (Hasan et al., 2007;Shaw et al., 2008), neuronal structure (Rabinowicz,Petetot, Khoury, & de Courten-Myers, 2009), andvolume changes in the amygdala (Yurgelun-Todd,Killgore, & Cintron, 2003). Our previous work raisedthe possibility that affective processing during adoles-cence may be associated with shifts in the lateraliza-tion of cortical and limbic processing (Killgore &Yurgelun-Todd, 2004b), similar to that observed forother cognitive and affective capacities (Everts et al.,2009; Killgore, Gruber, & Yurgelun-Todd, 2007).Consistent with that possibility, we found a negativecorrelation between the age of the participants andresponsiveness of the left amygdala to masked fear-ful faces. The negative correlation of left amygdalaactivation with age is similar to our previous find-ings for overtly presented faces (Killgore, Oki, &Yurgelun-Todd, 2001), and may suggest that the amy-gdala responses may become more right-lateralizedwith maturation through adolescence. While it ispossible that the present finding may reflect a similarpattern of reversed lateralization that extends to non-conscious affect processing in adolescents, furtherresearch will be needed to establish the reliability ofthis preliminary finding. Furthermore, there is someevidence that functional activity within the amygdala

TABLE 2 Regions showing activity correlations with left amygdala

responses to masked affect

Active voxels

MNI coordinates

Regions of activation x y z t-score

Anger: Positive correlation with left amygdala search territoryL. inf occipital gyrus 34 #38 #82 #10 5.62L. amygdala/sup temporal

pole99 #28 #2 #20 4.71

R. mid/sup temporal pole 61 46 14 #30 4.62R. cerebellum (area 6) 195 12 #76 #14 4.56L. cerebellum (area 7b) 85 #16 #74 #44 4.42R. inf occipital gyrus 38 42 #78 #6 4.39R. cerebellar crus 47 36 #66 #30 4.25

Anger: Negative correlation with left amygdala search territoryR. suppl motor/sup frontal

gyrus184 18 #16 58 5.45

L. mid cingulate gyrus 241 #10 #44 40 4.89R. precuneus 186 16 #40 42 4.55R. postcentral gyrus 51 56 #14 40 4.46R. precentral gyrus 20 40 #18 54 4.25

Fear: Positive correlation with left amygdala search territoryR. parahippocampal gyrus 44 14 #2 #14 4.28

Fear: Negative correlation with left amygdala search territoryR. ant cingulate gyrus 218 12 22 28 7.11L. mid cingulate gyrus 22 #10 #18 40 5.48L. supramarginal gyrus 24 #58 #38 24 5.12L. mid temporal pole 24 #42 10 #34 4.17R. amygdala 25 32 #4 #28 4.04Cerebellar vermis (area 8) 24 #2 #68 #44 3.95R. cerebellum (area 6) 26 36 #56 #30 3.94

Happy: Positive correlation with left amygdala search territoryNone # # # # #

Happy: Negative correlation with left amygdala search territoryNone # # # # #

Notes: L = left hemisphere, R = right hemisphere, sup = superior,mid = middle, inf = inferior, ant = anterior, suppl = supplementary.Atlas coordinates are from the MNI standard atlas, such that xreflects the distance (mm) to the right or left of midline, y reflectsthe distance anterior or posterior to the anterior commissure, and zreflects the distance superior or inferior to the horizontal planethrough the AC–PC line. All values exceed a threshold of p < .001,k = 20.

Downloaded By: [Killgore, William D. S.] At: 19:09 25 February 2010

40 KILLGORE AND YURGELUN-TODD

may be inversely related to prefrontal activation(Whalen et al., 2004), and such prefrontal activationduring affective processing may increase with matu-ration throughout adolescence (Yurgelun-Todd &Killgore, 2006).

The second major finding was that the lateraliza-tion of amygdala responses was modulated by thespecific affect category presented. For masked angryfaces, both the left and right amygdala showed evid-ence of deactivation relative to masked neutral faces,but the magnitude of deactivation was significantlygreater for the left vs. the right amygdala. Further-more, the response observed within the left amygdalawas significantly lower for masked anger relative tothe masked fear and masked happy conditions. Arecent study also reported decreased activation withinthe right amygdala during presentations of maskedangry vs. masked neutral faces in healthy children andadolescents but showed that youths with pediatricgeneralized anxiety disorder showed elevated rightamygdala responsiveness during this task (Monk et al.,2008). Together, these studies suggest that reducedamygdala activation to masked angry faces may be anormal response in older children and adolescents,while exaggerated responses to these stimuli may beassociated with heightened anxiety in this age group.Interestingly, masked fear, while showing a trendtoward greater right amygdala responses relative tothe other affect conditions, failed to show significantlaterality effects, a pattern that differs from the com-monly observed right lateralized effect in adults(Costafreda et al., 2008; Morris et al., 1999). Our previ-ous study with a group of adolescents viewing maskedsad faces found significant activation within the rightbut not left amygdala, a finding that was stronger inadolescents than in adults (Killgore & Yurgelun-Todd, 2007). On the other hand, in the present study,masked happy expressions yielded greater activationin the left amygdala and reduced activation in theright, a pattern opposite from that of masked anger. Ina previous study of similarly aged adolescents, wefailed to observe significant responses in either amy-gdala during masked happy presentations using a dif-ferent set of stimuli and faster presentation parameters(Killgore & Yurgelun-Todd, 2007), suggesting con-siderable variability in how this age group may per-ceive masked images of happy expressions. Overall,these findings argue against a simple valence effectwhereby faces displaying negative affect would beexpected to elicit greater amygdala responses thanthose expressing positive affect, suggesting instead thatthe amygdala responses may have been affected bysome other feature that distinguished among the emo-tional expression categories (e.g., arousal, dominance,

approach/withdrawal). Previous work suggests thatthe amygdala may be responsive to a broad range ofpositive and negative facial affect stimuli (Costafredaet al., 2008; Fitzgerald, Angstadt, Jelsone, Nathan, &Phan, 2006), suggesting that amygdala responses maysimply reflect visual vigilance for biologically salientinformation (Suslow et al., 2006). Clearly, furtherwork will be necessary to identify the specific factorsthat lead to modulation of these responses and whetherthese are similar for adults and children.

Another primary goal of the present study was toexamine the relationship between amygdala activa-tion and responses within other brain regions duringperception of masked affective faces. Of note, wefound that the correlation between amygdala activa-tion and cerebral responses was modulated by theaffective category of the masked expressions. Duringpresentations of masked angry vs. neutral faces, leftamygdala activation was associated with correspond-ing increases in bilateral regions of extrastriate cortex(i.e., inferior occipital gyrus; BA 19), bilateral tempo-ral pole and peri-amygdalar cortex, and cerebellum,and decreases in superior frontal regions, middle cin-gulate gyrus, and precuneus, even after statisticallycontrolling for age. These correlations are generallyconsistent with the notion that masked affective pres-entations involve an extrastriate visual pathway (Morriset al., 1999; Ohman, 2005), and further support theoperation of such a pathway during the period of ado-lescence. In particular, with greater amygdala activa-tion in response to masked angry faces, there was alsogreater activation of areas involved in secondary vis-ual processing, without corresponding increases inprimary visual cortex. This suggests that the amygdalaresponses may be closely associated with a crude vis-ual analysis that bypasses primary visual processing,possibly via the superior colliculus–pulvinar pathwayinvolved in the phenomenon of blindsight (Bittar,Ptito, Faubert, Dumoulin, & Ptito, 1999; Morris et al.,2001; Schoenfeld, Heinze, & Woldorff, 2002). Theenhancement of extrastriate visual processing withgreater amygdala activation is consistent with a reportby Suslow and colleagues, who found that amygdalaresponses were positively correlated with the abilityto subsequently identify previously presented maskedaffect stimuli (Suslow et al., 2006). Interestingly, wealso found that amygdala responses to masked angerwere negatively correlated with activation within pri-marily posterior midline regions such as the middlecingulate gyrus and precuneus. These regions are keystructures in the “default mode” network, which isactivated across a variety of resting states or when anindividual is engaged in self-referential introspectivecognition or visual imagery, and which is attenuated

Downloaded By: [Killgore, William D. S.] At: 19:09 25 February 2010

ADOLESCENT MASKED AFFECT 41

during the engagement of externally focused cogni-tive tasks (Cavanna & Trimble, 2006; Raichle et al.,2001). In other words, those children and adolescentsshowing the greatest amygdala responses to themasked angry faces also tended to have the greatestdecline in these posterior midline regions, perhapsreflecting a shift away from self-focused cognitionand toward greater awareness of the external environ-ment—a form of enhanced readiness in the face ofpotential threat.

A different pattern of amygdala correlationsemerged for masked fear vs. neutral presentations.Signal in the left amygdala ROI was positively corre-lated with a single cluster of activation localized tothe right parahippocampal gyrus, a region of the brainknown to be activated by emotional facial expressions(Fusar-Poli et al., 2009), memory consolidation (Janzen,Jansen, & van Turennout, 2008), and successful mem-ory retrieval (Hayes, Nadel, & Ryan, 2007). In con-trast, left amygdala activation during the masked fearcondition was negatively correlated with a largeregion of the right dorsal anterior cingulate gyrus, aregion involved in error detection (Modirrousta &Fellows, 2008), self-evaluation of decision-makingquality (Hewig et al., 2009), and behavioral inhibition(Beaver, Lawrence, Passamonti, & Calder, 2008).Fear-related left amygdala activation was also nega-tively correlated with a number of smaller regionsincluding the right amygdala, left middle cingulategyrus, left middle temporal pole, left supramarginalgyrus, and cerebellum. Together, these findings sug-gest that amygdala responses during masked fearpresentations were associated with increased activa-tion of systems involved in episodic memory encod-ing, particularly for emotional information andreduced activation in cognitive monitoring and errordetection systems.

Finally, activation within the left amygdala duringmasked happy vs. neutral presentations was not sig-nificantly correlated with any other brain regions,consistent with the general notion that the amygdala ismost involved in activating a neurobehavioralresponse to potentially threatening stimuli (Ohrmannet al., 2007). Our previous work using masked affec-tive stimuli with adolescents failed to find significantamygdala responses to masked happy faces (Killgore& Yurgelun-Todd, 2007). Thus, in adolescents, amy-gdala responses to happy faces seem to be inconsist-ently associated with a particular pattern of changes incerebral activation relative to other masked affects.

Overall, the present findings are generally consist-ent with the hypothesis that masked affect presenta-tions bypass the geniculostriate visual processingsystem and directly activate limbic and extrastrate

regions (LeDoux, 1996), presumably via the superiorcolliculus–pulvinar–amygdala pathway (Kimura et al.,2004; Morris et al., 1999), and that this systemappears to be functional by late childhood or earlyadolescence, if not earlier. Additionally, the presentresults further suggest that each affective category isassociated with distinct patterns of activation withinthe amygdala and that the pattern of amygdala laterali-zation during masked affect processing may be differ-ent in adolescent children relative to the eventual rightlateralized pattern that is commonly observed in adults.These findings are consistent with a small number ofstudies suggesting that the adolescent developmentalperiod may be characterized by a shift in lateralizedactivation within the amygdala. Some evidence sug-gests that these patterns may differ between males andfemales. Unfortunately, the sample sizes in the presentstudy precluded adequate power to resolve potentialsex differences. Future studies may clarify the extent towhich these findings reflect stable response patterns ofaffective responses to masked stimuli and the extent towhich they differ as a function of sex.

Manuscript received 1 May 2009Manuscript accepted 4 August 2009

First published online 23 September 2009

REFERENCES

Beaver, J. D., Lawrence, A. D., Passamonti, L., & Calder,A. J. (2008). Appetitive motivation predicts the neuralresponse to facial signals of aggression. Journal of Neu-roscience, 28(11), 2719–2725.

Bittar, R. G., Ptito, M., Faubert, J., Dumoulin, S. O., &Ptito, A. (1999). Activation of the remaining hemispherefollowing stimulation of the blind hemifield in hemi-spherectomized subjects. NeuroImage, 10(3), 339–346.

Cavanna, A. E., & Trimble, M. R. (2006). The precuneus: Areview of its functional anatomy and behavioural corre-lates. Brain, 129(3), 564–583.

Costafreda, S. G., Brammer, M. J., David, A. S., & Fu, C. H.(2008). Predictors of amygdala activation during the pro-cessing of emotional stimuli: A meta-analysis of 385 PETand fMRI studies. Brain Research Review, 58(1), 57–70.

Ekman, P., & Friesen, W. V. (1976). Pictures of facialaffect. Palo Alto, CA: Consulting Psychologists Press.

Esteves, F., & Ohman, A. (1993). Masking the face: Recog-nition of emotional facial expressions as a function ofthe parameters of backward masking. ScandinavianJournal of Psychology, 34(1), 1–18.

Everts, R., Lidzba, K., Wilke, M., Kiefer, C., Mordasini,M., Schroth, G., et al. (2009). Strengthening of lateralityof verbal and visuospatial functions during childhood andadolescence. Human Brain Mapping, 30(2), 473–483.

Fitzgerald, D. A., Angstadt, M., Jelsone, L. M., Nathan, P. J., &Phan, K. L. (2006). Beyond threat: Amygdala reactivityacross multiple expressions of facial affect. NeuroImage,30(4), 1441–1448.

Downloaded By: [Killgore, William D. S.] At: 19:09 25 February 2010

42 KILLGORE AND YURGELUN-TODD

Friston, K., Holmes, A., Worsley, K., Poline, J., Frith, C., &Frackowiak, R. (1995). Statistical parametric maps infunctional imaging: A general approach. Human BrainMapping, 5, 189–201.

Fusar-Poli, P., Placentino, A., Carletti, F., Allen, P., Landi,P., Abbamonte, M., et al. (2009). Laterality effect onemotional faces processing: ALE meta-analysis of evid-ence. Neuroscience Letters, 452(3), 262–267.

Hasan, K. M., Sankar, A., Halphen, C., Kramer, L. A.,Brandt, M. E., Juranek, J., et al. (2007). Developmentand organization of the human brain tissue compart-ments across the lifespan using diffusion tensor imaging.NeuroReport, 18(16), 1735–1739.

Hayes, S. M., Nadel, L., & Ryan, L. (2007). The effect ofscene context on episodic object recognition: Parahip-pocampal cortex mediates memory encoding andretrieval success. Hippocampus, 17(9), 873–889.

Hewig, J., Straube, T., Trippe, R. H., Kretschmer, N.,Hecht, H., Coles, M. G., et al. (2009). Decision-makingunder risk: An fMRI study. Journal of Cognitive Neuro-science, 21(8), 1642–1652.

Janzen, G., Jansen, C., & van Turennout, M. (2008). Memoryconsolidation of landmarks in good navigators. Hippoc-ampus, 18(1), 40–47.

Killgore, W. D. S., Gruber, S. A., & Yurgelun-Todd, D. A.(2007). Depressed mood and lateralized prefrontal activityduring a Stroop task in adolescent children. NeuroscienceLetters, 416(1), 43–48.

Killgore, W. D. S., Oki, M., & Yurgelun-Todd, D. A. (2001).Sex-specific developmental changes in amygdala responsesto affective faces. NeuroReport, 12(2), 427–433.

Killgore, W. D. S., & Yurgelun-Todd, D. A. (2004a). Acti-vation of the amygdala and anterior cingulate duringnonconscious processing of sad versus happy faces.NeuroImage, 21(4), 1215–1223.

Killgore, W. D. S., & Yurgelun-Todd, D. A. (2004b). Sex-related developmental differences in the lateralizedactivation of the prefrontal cortex and amygdala duringperception of facial affect. Perceptual and Motor Skills,99(2), 371–391.

Killgore, W. D. S., & Yurgelun-Todd, D. A. (2005). Socialanxiety predicts amygdala activation in adolescentsviewing fearful faces. NeuroReport, 16(15), 1671–1675.

Killgore, W. D. S., & Yurgelun-Todd, D. A. (2007). Uncon-scious processing of facial affect in children and adoles-cents. Social Neuroscience, 2(1), 28–47.

Kimura, Y., Yoshino, A., Takahashi, Y., & Nomura, S.(2004). Interhemispheric difference in emotionalresponse without awareness. Physiology and Behavior,82(4), 727–731.

Kugel, H., Eichmann, M., Dannlowski, U., Ohrmann, P.,Bauer, J., Arolt, V., et al. (2008). Alexithymic featuresand automatic amygdala reactivity to facial emotion.Neuroscience Letters, 435(1), 40–44.

LeDoux, J. E. (1996). The emotional brain: The mysteriousunderpinnings of emotional life. New York: Simon &Schuster.

Macwhinney, B., Cohen, J., & Provost, J. (1997). The Psy-Scope experiment-building system. Spatial Vision,11(1), 99–101.

Maldjian, J. A., Laurienti, P. J., Kraft, R. A., & Burdette, J. H.(2003). An automated method for neuroanatomic andcytoarchitectonic atlas-based interrogation of fMRI datasets. NeuroImage, 19(3), 1233–1239.

Modirrousta, M., & Fellows, L. K. (2008). Dorsal medialprefrontal cortex plays a necessary role in rapid errorprediction in humans. Journal of Neuroscience, 28(51),14000–14005.

Monk, C. S., Telzer, E. H., Mogg, K., Bradley, B. P., Mai,X., Louro, H. M., et al. (2008). Amygdala and ventro-lateral prefrontal cortex activation to masked angryfaces in children and adolescents with generalizedanxiety disorder. Archives of General Psychiatry,65(5), 568–576.

Morris, J. S., DeGelder, B., Weiskrantz, L., & Dolan, R. J.(2001). Differential extrageniculostriate and amygdalaresponses to presentation of emotional faces in a corti-cally blind field. Brain, 124(6), 1241–1252.

Morris, J. S., Ohman, A., & Dolan, R. J. (1998). Consciousand unconscious emotional learning in the human amy-gdala. Nature, 393(6684), 467–470.

Morris, J. S., Ohman, A., & Dolan, R. J. (1999). A subcorti-cal pathway to the right amygdala mediating “unseen”fear. Proceedings of the National Academy of Sciencesof the United States of America, 96(4), 1680–1685.

Ohman, A. (2005). The role of the amygdala in human fear:automatic detection of threat. Psychoneuroendocrinol-ogy, 30(10), 953–958.

Ohman, A., Carlsson, K., Lundqvist, D., & Ingvar, M.(2007). On the unconscious subcortical origin of humanfear. Physiology and Behavior, 92(1–2), 180–185.

Ohrmann, P., Rauch, A. V., Bauer, J., Kugel, H., Arolt, V.,Heindel, W., et al. (2007). Threat sensitivity as assessedby automatic amygdala response to fearful faces predictsspeed of visual search for facial expression. Experimen-tal Brain Research, 183(1), 51–59.

Rabinowicz, T., Petetot, J. M., Khoury, J. C., & de Courten-Myers, G. M. (2009). Neocortical maturation duringadolescence: Change in neuronal soma dimension. Brainand Cognition, 69(2), 328–336.

Raichle, M. E., MacLeod, A. M., Snyder, A. Z., Powers, W.J., Gusnard, D. A., & Shulman, G. L. (2001). A defaultmode of brain function. Proceedings of the NationalAcademy of Sciences of the United States of America,98(2), 676–682.

Rauch, S. L., Whalen, P. J., Shin, L. M., McInerney, S. C.,Macklin, M. L., Lasko, N. B., et al. (2000). Exaggeratedamygdala response to masked facial stimuli in posttrau-matic stress disorder: a functional MRI study. BiologicalPsychiatry, 47(9), 769–776.

Schoenfeld, M. A., Heinze, H. J., & Woldorff, M. G.(2002). Unmasking motion-processing activity in humanbrain area V5/MT+ mediated by pathways that bypassprimary visual cortex. NeuroImage, 17(2), 769–779.

Shaw, P., Kabani, N. J., Lerch, J. P., Eckstrand, K., Lenroot,R., Gogtay, N., et al. (2008). Neurodevelopmental tra-jectories of the human cerebral cortex. Journal of Neuro-science, 28(14), 3586–3594.

Suslow, T., Ohrmann, P., Bauer, J., Rauch, A. V., Schwindt,W., Arolt, V., et al. (2006). Amygdala activation duringmasked presentation of emotional faces predicts con-scious detection of threat-related faces. Brain and Cog-nition, 61(3), 243–248.

Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D.,Crivello, F., Etard, O., Delcroix, N., et al. (2002). Auto-mated anatomical labeling of activations in SPM using amacroscopic anatomical parcellation of the MNI MRIsingle-subject brain. NeuroImage, 15(1), 273–289.

Downloaded By: [Killgore, William D. S.] At: 19:09 25 February 2010

ADOLESCENT MASKED AFFECT 43

Whalen, P. J., Kagan, J., Cook, R. G., Davis, F. C., Kim, H.,Polis, S., et al. (2004). Human amygdala responsivity tomasked fearful eye whites. Science, 306(5704), 2061.

Whalen, P. J., Rauch, S. L., Etcoff, N. L., McInerney, S. C.,Lee, M. B., & Jenike, M. A. (1998). Masked presenta-tions of emotional facial expressions modulate amygdalaactivity without explicit knowledge. Journal of Neuro-science, 18(1), 411–418.

Williams, L. M., Das, P., Liddell, B. J., Kemp, A. H.,Rennie, C. J., & Gordon, E. (2006a). Mode of functionalconnectivity in amygdala pathways dissociates level ofawareness for signals of fear. Journal of Neuroscience,26(36), 9264–9271.

Williams, L. M., Liddell, B. J., Kemp, A. H., Bryant, R. A.,Meares, R. A., Peduto, A. S., et al. (2006b). Amygdala–

prefrontal dissociation of subliminal and supraliminalfear. Human Brain Mapping, 27(8), 652–661.

Williams, L. M., Liddell, B. J., Rathjen, J., Brown, K. J.,Gray, J., Phillips, M., et al. (2004). Mapping the timecourse of nonconscious and conscious perception offear: An integration of central and peripheral measures.Human Brain Mapping, 21(2), 64–74.

Yurgelun-Todd, D. A., & Killgore, W. D. (2006). Fear-related activity in the prefrontal cortex increases withage during adolescence: A preliminary fMRI study. Neu-roscience Letters, 406(3), 194–199.

Yurgelun-Todd, D. A., Killgore, W. D. S., & Cintron, C. B.(2003). Cognitive correlates of medial temporal lobedevelopment across adolescence: A magnetic resonanceimaging study. Perceptual and Motor Skills, 96(1), 3–17.

Downloaded By: [Killgore, William D. S.] At: 19:09 25 February 2010