AMTPU

Bansb

Mepa

Riopc

Cte

Kc

Edc(ttamhnptcpol

F

AA

R

0d

utomatic Mood-Congruent Amygdala Responses toasked Facial Expressions in Major Depression

homas Suslow, Carsten Konrad, Harald Kugel, Daniel Rumstadt, Pienie Zwitserlood, Sonja Schöning,atricia Ohrmann, Jochen Bauer, Martin Pyka, Anette Kersting, Volker Arolt, Walter Heindel, anddo Dannlowski

ackground: Cognitive theories of depression predict mood-congruent negative biases already at automatic stages of processing,lthough several behavioral studies seem to contradict this notion. That is, depression should potentiate emotional reactivity toegative emotional cues, whereas it should reduce reactivity in response to positive emotional stimuli. Assessing neurobiologicalubstrates of automatic emotion processing might be a more sensitive challenge for automatic negative bias in depression thanehavioral measures.

ethods: In 30 acutely depressed inpatients and 26 healthy control subjects, automatic amygdala responses to happy and sad facialxpressions were assessed by means of functional magnetic resonance imaging (fMRI) at 3 Tesla. To examine automatic responses, aresentation paradigm using subliminal, backward-masked stimuli was employed. A detection task was administered to assess participants’wareness of the masked emotional faces presented in the fMRI experiment.

esults: Detection performance was at chance level for both patients and healthy control subjects, suggesting that the neurobiolog-cal reactions took place in absence of conscious awareness of the emotional stimuli. A robust emotion by group interaction wasbserved in the right amygdala. Whereas healthy control subjects demonstrated stronger responses to happy faces, depressedatients showed the opposite. Furthermore, amygdala responsiveness to happy facial expression was negatively correlated withurrent depression severity.

onclusions: Depressed patients exhibit potentiated amygdala reactivity to masked negative stimuli along with a reduced responsivenesso masked positive stimuli compared with healthy individuals. Thus, depression is characterized by mood-congruent processing ofmotional stimuli in the amygdala already at an automatic level of processing.

ey Words: Amygdala, depression, emotion, faces, fMRI, mood-ongruent bias

arly cognitive theories of depression highlight the impor-tance of negative biases affecting several, if not all, cog-nitive processes. Beck (1), for instance, suggested that

epressogenic schemas operate in all aspects and stages ofognition, favoring or facilitating the processing of negativemood-congruent) stimuli in major depression. Similar predic-ions can be derived from Bower’s network model (2). Accordingo these theories, mood congruency effects should be observablelready at early, automatic processing stages. Pervasive negativeood states in depressed patients might contribute to an en-anced emotional reactivity to negative emotional cues (i.e.,egative potentiation) and reduce reactivity in response toositive emotional stimuli (i.e., positive attenuation) (3–5). Otherheories propose that depression is rather characterized byognitive biases in late or controlled stages of informationrocessing (6,7). Indeed, findings of negative biases in late stagesf attention or explicit memory appear quite consistently in theiterature, whereas findings of automatic emotion processing

rom the Departments of Psychiatry (TS, CK, DR, SS, PO, JB, AK, VA, UD),IZKF-Research Group 4 (CK, SS, MP, UD), Clinical Radiology (HK, WH), andDepartment of Psychology II (PZ, UD), University of Münster, Münster,Germany.

uthors TS and CK contributed equally to this article.ddress correspondence to Thomas Suslow, Ph.D., Department of Psychia-

try, University of Münster, Albert-Schweitzer-Str. 11, 48149 Münster,Germany; E-mail: suslow@uni-muenster.de.

eceived Mar 3, 2009; revised Jun 4, 2009; accepted Jul 10, 2009.

006-3223/10/$36.00oi:10.1016/j.biopsych.2009.07.023

biases in major depression are rare (8). However, most studiesinvestigating cognitive bias in depression made use of reaction-time paradigms with behavioral measures as dependent variable.Neurobiological analyses might provide a more sensitive assess-ment of automatic emotion processing in depression, comparedwith behavioral testing.

The amygdala is a central structure in the limbic emotion-processing circuit (9). In addition to a slower, cortical route, theamygdala receives direct projections from the thalamus, allowinga rapid response to emotionally salient stimuli, even beforeconscious cortical representations have been formed (10). Amyg-dala hyperactivity has been implicated in the pathogenesis ofmajor depression, probably by causing negatively biased emo-tion processing (11,12). Therefore, the amygdala was selected asthe region of interest in the present study. Functional magneticresonance imaging (fMRI) was used to examine differencesbetween depressed patients and healthy control subjects inautomatic amygdala reactivity to facial emotions. Emotional faceswere presented briefly and masked backward by neutral faces toprevent conscious emotion processing. Facial expression servesas an important social signal that modulates interpersonal inter-actions (13). The amygdala has been shown to be activatedreliably by facial emotions, even when facial expressions werepresented below the threshold of conscious awareness (14–17).According to a recent meta-analysis of neuroimaging studies onamygdala activation, there is evidence for hemispheric lateraliza-tion during the processing of emotional stimuli. The right amyg-dala may subserve a high-speed detection role for unconsciousstimuli (18).

We hypothesized that relative to healthy control subjects,

acutely depressed patients show stronger automatic (right)

BIOL PSYCHIATRY 2010;67:155–160© 2010 Society of Biological Psychiatry

ae(

M

S

dnDwcttTo4aAimpu((gTd

TSRD

ASHHTANL

DEEEELLLL

tM

f

156 BIOL PSYCHIATRY 2010;67:155–160 T. Suslow et al.

w

mygdala activation in response to mood congruent (sad) facialxpressions and less automatic activation to mood-incongruenthappy) facial expressions.

ethods and Materials

ubjectsDatasets of 30 right-handed inpatients with an acute major

epressive episode according to DSM-IV criteria (19) and diag-osed with the Structured Clinical Interview for DSM-IV Axis Iisorders (SCID-I) interview (20) and 26 healthy control subjectsere analyzed (Table 1). Exclusion criteria were any neurologi-

al abnormalities, substance abuse, former electroconvulsiveherapy, mood stabilizers, and neuroleptic or benzodiazepinereatment. All patients were under antidepressant treatment (seeable S1 in Supplement 1 for details), which was coded in termsf dose and treatment duration into medication levels from 1 to, according to the suggestions of Sackeim (21). The study waspproved by the Ethics Committee of the University of Münster.fter complete description of the study to the subjects, written

nformed consent was obtained. Only patients with primaryajor depression were included (indicated by earlier onset andredominant symptoms). Secondary lifetime diagnoses werendifferentiated somatoform disorder (n � 4), social phobian � 4), dysthymia (n � 3), obsessive-compulsive disorder (OCD)n � 2), panic disorder (n � 1), specific phobia (n � 1),eneralized anxiety disorder (n � 1), and pain disorder (n � 1).wo patients had two and one patient had three comorbid

able 1. Clinical, Sociodemographic, and Affective Characteristics oftudy Participants and Mean Reaction Times (in msec) and Evaluativeesponses to the Neutral Face Mask During the fMRI Experimentependent on Prime Condition (Mean [SD])

Patients(n � 30)

ControlSubjects(n � 26)

p Value,According

to �2 or t Tests(two-tailed)

ge 38.8 (11.4) 36.2 (13.4) .44ex (M/F) 17/13 12/14 .59AMD 24.8 (4.9) .6 (.8) �.001AMA 21.1 (5.8) 1.5 (1.7) �.001otal Education Time 13.6 (1.7) 14.2 (1.8) .24ntidepressant Potencya 2.6 (1.2)umber of Episodes 2.7 (2.0)ifetime Hospitalization

(Weeks) 7.6 (8.7)uration of Illness (Months) 72.2 (75.0)valuation Sad Prime �.13 (.23) �.01 (.33) .12valuation Happy Prime �.08 (.33) �.04 (.27) .60valuation Neutral Prime �.10 (.25) �.03 (.24) .29valuation No-Face Prime �.07 (.29) �.03 (.28) .58atency Sad Prime 1403.0 (537.7) 1319.6 (317.6) .51atency Happy Prime 1398.0 (575.7) 1315.6 (345.2) .54atency Neutral Prime 1400.3 (543.4) 1286.1 (338.2) .38atency No-Face Prime 1395.3 (511.6) 1303.3 (322.1) .45

F, female; fMRI, functional magnetic resonance imaging; HAMA, Hamil-on Rating Scale for Anxiety; HAMD, Hamilton Rating Scale for Depression;

, male.aCoded in terms of dose and treatment duration into medication levels

rom 1 to 4 according to Sackeim (21).

isorders.

ww.sobp.org/journal

fMRI MethodsFacial stimuli in the fMRI experiment consisted of grayscale

normalized sad, happy, and neutral expressions of 10 individuals(22). Emotional and neutral faces were presented for 33 msecand masked by neutral faces of the same individuals. To avoididentity of prime and mask in the neutral face condition,vertically mirrored faces were used as neutral primes. That is,neutral prime faces were produced by mirror-inversion (left toright) of neutral mask faces. Eighty trials were shown: 20 withsad, 20 with happy, and 20 with neutral prime faces; in 20 trials,no prime faces were presented. Faces were shown in two fixedpseudorandom sequences with the restriction of no repetition ofan individual and no more than one repetition of a primecondition on consecutive trials. Each trial lasted 9 sec. A fixationcross presented for 800 msec preceded a prime face shown for 33msec, which was followed by the corresponding neutral facemask presented for 467 msec. A blank screen followed for 7.700msec. During this time period, subjects had to evaluate whetherthe neutral (mask) face expressed rather negative or positivefeelings, by pressing one of four buttons (�1.5, �.5, �.5, and�1.5). In each hand, participants held a fiber optic response padwith two buttons (the positive or the negative response keys).One half of the sample gave positive responses with the lefthand, the other half with the right hand. Judgments and reactiontimes were registered. Images were projected to the rear end ofthe scanner (Sharp [Osaka, Japan] XG-PC10XE with additionalhigh frequency [HF] shielding). T2* functional data were ac-quired at a 3 tesla (3T) scanner (Gyroscan Intera 3T, PhilipsMedical Systems, Best, The Netherlands), using a single-shotechoplanar sequence with parameters selected to minimizedistortion in the region of central interest, while retaining ade-quate signal-to-noise ratio (S/N) and T2* sensitivity (23). Vol-umes consisting of 40 axial slices were acquired (matrix 642;resolution 3.5 � 3.5 � 3.5 mm; repetition time [TR] � 3 sec; echotime [TE] � 30 msec; flip angle [FA] � 90°). Functional imagingdata were motion-corrected, spatially normalized to standardMontreal Neurological Institute (MNI) space and smoothed(Gaussian kernel, 8 mm full-width at half maximum [FWHM])using Statistical Parametric Mapping (SPM5; Wellcome Depart-ment of Cognitive Neurology, London, United Kingdom, http://www.fil.ion.ucl.ac.uk/spm). An event-related analysis designwas used. For each subject, trials were averaged separately foreach prime condition, reducing the data to four average trials persubject. Brain responses to the prime stimulus categories wereisolated by convolving a vector of onset times of the emotionaland neutral primes and the no-face condition with a canonicalhemodynamic response function. Since the two baseline condi-tions (neutral face and no face) did not differ with respect toamygdala responses elicited across both groups (p � .05, uncor-rected), only the crucial contrasts of happy versus neutral andsad versus neutral expressions are reported. These two individ-ual first-level contrasts were entered into an analysis of variance(ANOVA) using the flexible factorial model implemented inSPM5, with emotion as within-subjects factor and group asbetween-subjects factor. A third factor, subjects, was also in-cluded in the model to account for the individual constants. Totest whether the masked presentation of sad and happy expres-sions resulted in activation of the amygdala, the main effect ofeach emotion condition (versus the neutral prime baseline) wasassessed separately within the whole sample. The model wasused to calculate the main effects of group (patients vs. controlsubjects), emotion (happy vs. sad), and the crucial group �

emotion interaction at a threshold of p � .05, familywise error

(awMewsmuivatbipp

ytrwmts

D

tetosiappte

Ffd

T. Suslow et al. BIOL PSYCHIATRY 2010;67:155–160 157

FWE) corrected for the amygdalae. The amygdala was definedccording to Tzourio-Mazoyer et al. (24) and the amygdala maskas created by means of the Wake Forest University (School ofedicine, Winston-Salem, North Carolina) Pickatlas (25). Toxplore the nature of the group � emotion interaction, t testsere used to investigate the effect of emotion within each group

eparately (paired t tests) and to compare the activation due toasked sad and happy faces between groups at p � .05,ncorrected. At the location of maximal group � emotionnteraction, the contrast values of happy versus neutral and sadersus neutral faces were extracted and used for exploratorynalyses, controlling for potential effects of gender, detectionask performance, medication level, number of episodes, comor-idity status, and duration of illness. These variables werencluded either as factor (gender) or covariate (detection taskerformance) in the group � emotion ANOVA or (within theatient group) correlated with the extracted contrast values.

For exploratory reasons, a supplementary whole-brain anal-sis was conducted at p � .001, uncorrected, with a clusterhreshold of k � 25 voxels (see Table S2 in Supplement 1 foresults). Finally, amygdala responsiveness to happy and sad facesas correlated with depression severity and anxiety level aseasured with the Hamilton Depression Scale (HAMD) (26) and

he Hamilton Anxiety Scale (HAMA) (27) within the patientample.

etection TaskAfter the fMRI experiment, all subjects were asked whether

hey had noticed anything extraordinary during the scannerxperiment. Then, subjects took part in a forced-choice detectionask outside the scanner. The detection task consisted of 40 trialsf the same stimulus presentation conditions and the sametimuli as in the fMRI experiment. Subjects were told thatmmediately before the face with the neutral expressionnother face was shown so briefly that it was very difficult toerceive. They should recognize the expression of the faceresented before the neutral face. Subjects were informed thathe covert face could have a happy, a sad, or a neutralxpression and in some cases no covert face was shown.

igure 1. Left: coronal view (y � 4), depicting significant group � emotion inor the amygdala volume. Right: bar graphs depicting the mean contrast v

ependent on emotion and study group.

Subjects should indicate via button press which prime condi-tion (happy, sad, neutral, no face) was presented before theneutral mask. A’ was calculated as nonparametric measure ofsensitivity, taking into account hit rate and false alarm rate,with A’ � .5 indicating chance level (28).

Results

Detection TaskAfter the fMRI experiment, all subjects reported that they had

not recognized any briefly presented emotional faces, even afterbeing informed about their presence. In the detection task,average sensitivity of healthy control subjects and patients didnot differ significantly from chance level, either for happy(control subjects: A’ � .49; patients: A’ � .51), sad (controlsubjects: A’ � .54; patients: A’ � .51), or neutral prime faces(control subjects: A’ � .48; patients: A’ � .43), according to ttests (all ps � .2). Importantly, both groups did not differconcerning their sensitivity indexes for emotional or neutralfaces (all ps � .55).

Behavioral ResultsPatients and control subjects did not differ in their evaluative

ratings of the neutral mask faces or their reaction times, irrespec-tive of prime condition (Table 1).

fMRI ResultsAcross both groups, the paradigm successfully elicited auto-

matic amygdala responses to both masked happy and maskedsad faces, compared with masked neutral faces, despite subjec-tive unawareness of the emotional primes (see Figure S1 inSupplement 1). The ANOVA based on the contrasts of happyversus neutral and sad versus neutral expressions yielded nosignificant main effect of emotion or group within the amygdala.However, the hypothesized emotion � group interaction wasobserved in a cluster within the right amygdala, x � 30, y � 4,z � �20; t(54) � 3.56, d � .97; puncorrected � .00039; pFWE-corrected �.016; cluster size k � 7 (Figure 1). To determine the amygdalasubregion where differential processing of emotion faces oc-

tion in the right amygdala, thresholded at p � .05, familywise error correctedfor happy-neutral and sad-neutral extracted from x � 30, y � 4, z � �20,

teracalue

www.sobp.org/journal

cNGtDfpr�Cts1s�d�mdf

ssHwycpasofs

D

esOcmitidnaafaawu

rrtgebop

158 BIOL PSYCHIATRY 2010;67:155–160 T. Suslow et al.

w

urred, the SPM Anatomy Toolbox Version 1.5 (Institut füreurowissenschaften und Medizen, Forschungszentrum Jülich,ermany) (29) was administered. The emotion � group interac-

ion was located in the lateral and basal nuclei of the amygdala.epressed patients showed higher amygdala responses to sad

aces than to happy faces, x � 28, y � 4, z � �20, t (29) � 2.91,� .003, d � 1.08, k � 20, and compared with amygdala

esponses to sad facial expressions in healthy control subjects, x30, y � 2, z � �22, t (54) � 2.34, p � .011, d � .64, k � 29.

onversely, healthy control subjects showed the opposite pat-ern, with stronger amygdala responses to happy compared withad faces, x � 32, y � 0, z � �26, t (25) � 2.74, p � .006, d �.10, k � 94, and with the depressed patients’ amygdala re-ponses to happy faces, x � 34, y � 4, z � �20, t (54) � 2.26, p

.016, d � .62, k � 5. Introducing gender as additional factor oretection performance (A’) as covariate did not alter the emotiongroup interaction. Furthermore, there were no associations ofedication level, number of episodes, comorbidity status, oruration of illness on amygdala responsiveness to happy or sadaces in the patient group (all ps � .17).

A correlation analysis in the patient sample yielded noignificant correlation of HAMD scores and amygdala respon-iveness to sad faces. However, a negative correlation ofAMD scores and amygdala responsiveness to happy facesas observed, again confined to the right amygdala, x � 26,� �8, z � �12 (coordinates of the voxel with the highest

orrelation [r � �.565]); t (28) � 3.62; puncorrected � .00058;

FWE-corrected � .025; k � 98. Thus, patients with weakerutomatic amygdala responsiveness to happy facial expres-ions suffered from higher depression levels. The correlationf HAMA scores and amygdala responsiveness to happy or sadaces yielded no significant results but showed a trend in theame direction.

iscussion

The present neuroimaging data provide evidence for a differ-ntial response pattern of the amygdala to subliminal emotiontimuli in depressed patients compared with healthy individuals.ur findings are consistent with the idea of automatic mood-

ongruent cognitive biases in major depression, as well as inentally healthy subjects. As hypothesized, amygdala reactivity

n depressed patients was increased to masked negative emo-ional stimuli and decreased to masked positive emotional stimulin comparison with healthy control subjects. Thus, it appears thatepression is characterized by a dysregulation of automaticeurobiological responsivity, showing negative potentiationnd positive attenuation. In line with previous research onmygdala responsiveness, the backward-masked emotionalaces successfully elicited robust amygdala responses. Inccordance with previous research, differences in amygdalactivation between groups were found in the right amygdala,hich seems to be specifically important for the processing ofnconscious stimuli (18).

The results of the detection task and the subjective experienceeported by our participants indicate that the neurobiologicaleactions took place in the absence of conscious awareness ofhe emotional stimuli. Healthy control subjects showed a stron-er amygdala response to happy facial expression than to sadxpression. This asymmetry is consistent with fMRI data reportedy Killgore and Yurgelun-Todd (14) and previous research fromur laboratory (30). In a broader methodological context, the

resent data parallel several psychophysiological studies that

ww.sobp.org/journal

demonstrated a positive or protective processing bias in healthysubjects (31,32). Also, in line with previous neuroimaging re-search, compared with healthy control subjects, depressed pa-tients manifested stronger amygdala responses to negative stim-uli (33–36).

The present data suggesting a dysfunctional reactivity of theamygdala to emotion stimuli at an automatic processing level inmajor depression complement behavioral findings that depres-sion is characterized by processing biases in late or controlledstages of information processing (6,7). It appears that neurobio-logical methods such as fMRI may represent a more sensitive wayof assessing automatic emotion processing in depression thanbehavioral testing. Future studies on emotion processing indepression should combine different methods of response as-sessment and evaluate directly the relationship between re-sponse levels for different types of processing (automatic vs.controlled). In contrast with results from previous psychologicalstudies based on the affective priming paradigm (37,38), noevaluative shifts due to masked facial emotions (compared withthe neutral or the no prime condition) were observed in ourexperiment. The absence of priming effects could be due, at leastin part, to the fact that we applied sad instead of angry facialexpression in the negative prime condition. However, the patternof our findings can also be interpreted in the sense that func-tional neuroimaging could represent a more sensitive tool todetect and measure subtle emotion processing compared withbehavioral tests.

Numerous researchers have argued that the amygdala plays acentral role in generating negative emotional experience(16,39,40). From this perspective, greater amygdala responsivityto negative faces in depressed patients could directly affect theirmood state in a negative way. However, in the present study, nocorrelation was observed between severity of current depressionand amygdala response to sad faces. Instead, we found aninverse relationship between intensity of depressive symptomsand amygdala response to happy faces. It is known that theamygdala modulates vigilance to exponentiate subsequent infor-mation processing throughout the brain (9). A low automaticreactivity of the amygdala to positive stimuli could implicate lessengagement in the encoding of positively valenced stimuli orreduced recruitment of attentional resources that can bring(peripheral) emotional stimuli to conscious awareness (41,42).Conversely, high amygdala responsivity to negative stimuli indepression was shown to be associated with automatic evalua-tive biases (43) and could bias negatively attention and highercognitive processes. According to our results, the between-groupdifference in amygdala activation in response to emotion faceswas located in the basolateral nuclei. Findings from previousresearch (15) indicate that basolateral amygdala activation inresponse to masked fearful faces is positively associated with traitanxiety. The basolateral complex could represent an integral partof an amygdala-cortical network for unconscious emotionalvigilance. Recruitment of visual and prefrontal areas by thebasolateral amygdala could enhance the allocation of attentionalresources for processing of sad stimuli and reduce the processingof positive or happy stimuli in depression. The basolateralamygdala is also known to regulate the consolidation of memorythrough its projections to many other brain regions involved instoring new information (44,45). Depressed patients’ differentialresponsivity of the basolateral amygdala to emotion stimuli maycontribute to a preferential encoding of mood-congruent stimuli.

The present findings may also shed light on the problematic

interpersonal functioning of depressed patients (46). Our results

suttefprm

meab(s(ttdrbomtth

tpFmmf

M

t

o

11

T. Suslow et al. BIOL PSYCHIATRY 2010;67:155–160 159

how that automatic, subcortical reactions of depressed individ-als are strong to negative but only weak to positive socioemo-ional signals. Individuals manifesting a selective bias favoringhe processing of negative facial responses may tend to experi-nce interpersonal failures. Reduced responsiveness to positiveacial expressions, which has already been shown in depressedatients on a behavioral level (47), could lead to disturbedelationships in the sense of less attunement and mutual involve-ent (48,49).Some limitations must be acknowledged. All patients were

edicated, which might represent a confounding factor. How-ver, previous research has repeatedly shown that antidepressantgents reduce amygdala responses to negative stimuli (34,50–52)ut enhance amygdala or subcortical responses to positive faces53,54). Given this consistent pattern, the medication in ourample would rather counteract the effects observed in our studywhich might therefore be even stronger in unmedicated pa-ients). Furthermore, our subjects were severely depressed inpa-ients. Thus, our results might not generalize to more moderatelyepressed nonhospitalized patients who constitute the usuallyecruited study population for behavioral studies of cognitiveias in depression. A larger sample size, with a broader spectrumf illness severity, would be necessary to investigate whether theood congruency effects observed in our sample are restricted

o severely affected patients. To control the effect of hospitaliza-ion on amygdala reactivity, it would be necessary to includeospitalized, nondepressed subjects as a second control group.

In sum, our study provides compelling support for the notionhat depression is characterized by mood-congruent emotionrocessing already at early and automatic processing stages.uture studies should investigate potential effects of psychophar-acological and psychotherapeutic treatment on automaticood-congruent amygdala activation and its relation to clinical

eatures or treatment response.

This research was supported by the Rolf-Dierichs-Stiftung,ünster, Germany (BD193436 to UD, TS, and CK).The authors report no biomedical financial interests or po-

ential conflicts of interest.

Supplementary material cited in this article is availablenline.

1. Beck AT (1967): Depression: Clinical, Experimental, and Theoretical As-pects. New York: Harper and Row.

2. Bower GH (1981): Mood and memory. Am Psychol 36:129 –148.3. Beck AT (2008): The evolution of the cognitive model of depression and

its neurobiological correlates. Am J Psychiatry 165:969 –977.4. Depue RA, Iacono WG (1989): Neurobehavioral aspects of affective dis-

orders. Annu Rev Psychol 40:457– 492.5. Scher CD, Ingram RE, Segal ZV (2005): Cognitive reactivity and vulnera-

bility: Empirical evaluation of construct activation and cognitive diathe-ses in unipolar depression. Clin Psychol Rev 25:487–510.

6. Mathews A, MacLeod C (2005): Cognitive vulnerability to emotionaldisorders. Annu Rev Clin Psychol 1:167–195.

7. Williams JM, Watts FN, MacLeod C, Mathews A (1997): Cognitive Psychol-ogy and Emotional Disorder. Chichester, England: Wiley.

8. Dannlowski U, Kersting A, Donges US, Lalee-Mentzel J, Arolt V, Suslow T(2006): Masked facial affect priming is associated with therapy responsein clinical depression. Eur Arch Psychiatry Clin Neurosci 256:215–221.

9. Davis M, Whalen PJ (2001): The amygdala: Vigilance and emotion. MolPsychiatry 6:13–34.

0. LeDoux JE (1996): The Emotional Brain. New York: Simon and Schuster.1. Phillips ML, Drevets WC, Rauch SL, Lane R (2003): Neurobiology of emo-

tion perception II: Implications for major psychiatric disorders. Biol Psy-

chiatry 54:515–528.

12. Whalen PJ, Shin LM, Somerville LH, McLean AA, Kim H (2002): Functionalneuroimaging studies of the amygdala in depression. Semin Clin Neuro-psychiatry 7:234 –242.

13. Ekman P (1984): Expression and the nature of emotion. In: Scherer KR,Ekman P, editors. Approaches to Emotion. Hillsdale, NJ: Lawrence Erl-baum, 319 –343.

14. Killgore WD, Yurgelun-Todd DA (2004): Activation of the amygdala andanterior cingulate during nonconscious processing of sad versus happyfaces. Neuroimage 21:1215–1223.

15. Etkin A, Klemenhagen KC, Dudman JT, Rogan MT, Hen R, Kandel ER, et al.(2004): Individual differences in trait anxiety predict the response of thebasolateral amygdala to unconsciously processed fearful faces. Neuron44:1043–1055.

16. Nomura M, Ohira H, Haneda K, Iidaka T, Sadato N, Okada T, et al. (2004):Functional association of the amygdala and ventral prefrontal cortexduring cognitive evaluation of facial expressions primed by maskedangry faces: An event-related fMRI study. Neuroimage 21:352–363.

17. Liddell BJ, Brown KJ, Kemp AH, Barton MJ, Das P, Peduto A, et al. (2005):A direct brainstem-amygdala-cortical “alarm” system for subliminal sig-nals of fear. Neuroimage 24:235–243.

18. Costafreda SG, Brammer MJ, David AS, Fu CH (2008): Predictors of amyg-dala activation during the processing of emotional stimuli: A meta-analysis of 385 PET and fMRI studies. Brain Res Rev 58:57–70.

19. American Psychiatric Association (1994): Diagnostic and Statistical Man-ual of Mental Disorders, 4th ed. Washinghton, DC: American PsychiatricAssociation.

20. Wittchen HU, Wunderlich U, Gruschwitz S, Zaudig M (1997): SKID-I.Strukturiertes Klinisches Interview für DSM-IV. Göttingen, Germany:Hogrefe.

21. Sackeim HA (2001): The definition and meaning of treatment-resistantdepression. J Clin Psychiatry 62(suppl 16):10 –17.

22. Ekman P, Friesen WV (1976): Pictures of Facial Affect. Palo Alto, CA: Con-sulting Psychologists Press.

23. Robinson S, Windischberger C, Rauscher A, Moser E (2004): Optimized 3T EPI of the amygdalae. Neuroimage 22:203–210.

24. Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F, Etard O,Delcroix N, et al. (2002): Automated anatomical labeling of activations inSPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 15:273–289.

25. Maldjian JA, Laurienti PJ, Kraft RA, Burdette JH (2003): An automatedmethod for neuroanatomic and cytoarchitectonic atlas-based interro-gation of fMRI datasets. Neuroimage 19:1233–1239.

26. Hamilton M (1960): A rating scale for depression. J Neurol NeurosurgPsychiatry 23:56 – 62.

27. Hamilton M (1959): The assessment of anxiety states by rating. Br J MedPsychol 32:50 –55.

28. Grier JB (1971): Nonparametric indexes for sensitivity and bias: Comput-ing formulas. Psychol Bull 75:424 – 429.

29. Eickhoff SB, Stephan KE, Mohlberg H, Grefkes C, Fink GR, Amunts K, et al.(2005): A new SPM toolbox for combining probabilistic cytoarchitec-tonic maps and functional imaging data. Neuroimage 25:1325–1335.

30. Dannlowski U, Ohrmann P, Bauer J, Kugel H, Arolt V, Heindel W, et al.(2007): Amygdala reactivity predicts automatic negative evaluations forfacial emotions. Psychiatry Res 154:13–20.

31. Rottenberg J, Gross JJ, Gotlib IH (2005): Emotion context insensitivity inmajor depressive disorder. J Abnorm Psychol 114:627– 639.

32. Deldin PJ, Keller J, Gergen JA, Miller GA (2001): Cognitive bias andemotion in neuropsychological models of depression. Cogn Emot 15:787– 802.

33. Siegle GJ, Steinhauer SR, Thase ME, Stenger VA, Carter CS (2002): Can’tshake that feeling: Event-related fMRI assessment of sustained amyg-dala activity in response to emotional information in depressed individ-uals. Biol Psychiatry 51:693–707.

34. Sheline YI, Barch DM, Donnelly JM, Ollinger JM, Snyder AZ, Mintun MA(2001): Increased amygdala response to masked emotional faces indepressed subjects resolves with antidepressant treatment: An fMRIstudy. Biol Psychiatry 50:651– 658.

35. Abler B, Erk S, Herwig U, Walter H (2007): Anticipation of aversive stimuliactivates extended amygdala in unipolar depression. J Psychiatr Res41:511–522.

36. Surguladze S, Brammer MJ, Keedwell P, Giampietro V, Young AW, Travis

MJ, et al. (2005): A differential pattern of neural response toward sad

www.sobp.org/journal

3

3

3

4

4

4

4

4

4

160 BIOL PSYCHIATRY 2010;67:155–160 T. Suslow et al.

w

versus happy facial expressions in major depressive disorder. BiolPsychiatry 57:201–209.

7. Murphy ST, Monahan JL, Zajonc RB (1995): Additivity of nonconsciousaffect: Combined effects of priming and exposure. J Pers Soc Psychol69:589 – 602.

8. Murphy ST, Zajonc RB (1993): Affect, cognition, and awareness: Affec-tive priming with optimal and suboptimal stimulus exposures. J Pers SocPsychol 64:723–739.

9. Abercrombie HC, Schaefer SM, Larson CL, Oakes TR, Lindgren KA,Holden JE, et al. (1998): Metabolic rate in the right amygdala predictsnegative affect in depressed patients. Neuroreport 9:3301–3307.

0. Schaefer SM, Jackson DC, Davidson RJ, Aguirre GK, Kimberg DY, Thomp-son-Schill SL (2002): Modulation of amygdalar activity by the consciousregulation of negative emotion. J Cogn Neurosci 14:913–921.

1. Adolphs R, Tranel D, Hamann S, Young AW, Calder AJ, Phelps EA, et al.(1999): Recognition of facial emotion in nine individuals with bilateralamygdala damage. Neuropsychologia 37:1111–1117.

2. Amaral DG (2002): The primate amygdala and the neurobiology ofsocial behavior: Implications for understanding social anxiety. Biol Psy-chiatry 51:11–17.

3. Dannlowski U, Ohrmann P, Bauer J, Kugel H, Arolt V, Heindel W, et al.(2007): Amygdala reactivity to masked negative faces is associated withautomatic judgmental bias in major depression: A 3 T fMRI study. J Psy-chiatry Neurosci 32:423– 429.

4. McGaugh JL (2004): The amygdala modulates the consolidation ofmemories of emotionally arousing experiences. Annu Rev Neurosci 27:1–28.

5. Chavez CM, McGaugh JL, Weinberger NM (2009): The basolateral amyg-dala modulates specific sensory memory representations in the cere-

bral cortex. Neurobiol Learn Mem 91:382–392.

ww.sobp.org/journal

46. Gotlib IH, Hammen CL (2002): Handbook of Depression. New York: Guil-ford Publications.

47. Suslow T, Dannlowski U, Lalee-Mentzel J, Donges US, Arolt V, Kersting A(2004): Spatial processing of facial emotion in patients with unipolardepression: A longitudinal study. J Affect Disord 83:59 – 63.

48. Bouhuys AL, Geerts E, Mersch PP (1997): Relationship between percep-tion of facial emotions and anxiety in clinical depression: Does anxiety-related perception predict persistence of depression? J Affect Disord43:213–223.

49. Surguladze SA, Young AW, Senior C, Brebion G, Travis MJ, Phillips ML(2004): Recognition accuracy and response bias to happy and sad facialexpressions in patients with major depression. Neuropsychology 18:212–218.

50. Harmer CJ, Mackay CE, Reid CB, Cowen PJ, Goodwin GM (2006): Antide-pressant drug treatment modifies the neural processing of noncon-scious threat cues. Biol Psychiatry 59:816 – 820.

51. Norbury R, Mackay CE, Cowen PJ, Goodwin GM, Harmer CJ (2007): Short-term antidepressant treatment and facial processing. Functional mag-netic resonance imaging study. Br J Psychiatry 190:531–532.

52. Fu CH, Williams SC, Cleare AJ, Brammer MJ, Walsh ND, Kim J, et al.(2004): Attenuation of the neural response to sad faces in majordepression by antidepressant treatment: A prospective, event-re-lated functional magnetic resonance imaging study. Arch Gen Psychi-atry 61:877– 889.

53. Fu CH, Williams SC, Brammer MJ, Suckling J, Kim J, Cleare AJ, et al. (2007):Neural responses to happy facial expressions in major depression fol-lowing antidepressant treatment. Am J Psychiatry 164:599 – 607.

54. Schaefer HS, Putnam KM, Benca RM, Davidson RJ (2006): Event-relatedfunctional magnetic resonance imaging measures of neural activity to

positive social stimuli in pre- and post-treatment depression. Biol Psy-chiatry 60:974 –986.