Evidence of Abnormal Amygdala Functioning inBorderline Personality Disorder: A Functional MRI Study

Sabine C. Herpertz, Thomas M. Dietrich, Britta Wenning, Timo Krings,Stephan G. Erberich, Klaus Willmes, Armin Thron, and Henning Sass

Background: Intense and rapidly changing mood statesare a major feature of borderline personality disorder(BPD); however, there have only been a few studiesinvestigating affective processing in BPD, and in partic-ular no neurofunctional correlates of abnormal emotionalprocessing have been identified so far.

Methods: Six female BPD patients without additionalmajor psychiatric disorder and six age-matched femalecontrol subjects underwent functional magnetic resonanceimaging (fMRI) to measure regional cerebral hemody-namic changes following brain activity when viewing 12standardized emotionally aversive slides compared to 12neutral slides, which were presented in random order.

Results:Our main finding was that BPD subjects but notcontrol subjects were characterized by an elevated bloodoxygenation level dependent fMRI signal in the amygdalaon both sides. In addition, activation of the medial andinferolateral prefrontal cortex was seen in BPD patients.Both groups showed activation in the temporo-occipitalcortex including the fusiform gyrus in BPD subjects butnot in control subjects.

Conclusions: Enhanced amygdala activation in BPD issuggested to reflect the intense and slowly subsidingemotions commonly observed in response to even low-level stressors. Borderline subjects’ perceptual cortex maybe modulated through the amygdala leading to increasedattention to emotionally relevant environmental stimuli.Biol Psychiatry 2001;50:292–298 ©2001 Society of Bio-logical Psychiatry

Key Words: Borderline personality disorder, emotion,amygdala, functional MRI

Introduction

Quality and intensity of affective responses to environ-mental events influence mood and basic features of

personality functioning, such as the organization of socialrelationships and impulse control. Consequently, rigid andpoorly adapted affective responses are seen as a centralfeature of personality disorders. Borderline personalitydisorder (BPD), in particular, is thought to arise fromaffective vulnerability (Linehan 1993). The inability toregulate one’s affective responses leads to marked, rapidlychanging mood states and predisposes patients to variouskinds of self-destructive behavior (Herpertz et al 1997).There have only been a few studies investigating theprocessing of emotional information in BPD. Using anumber of self-report items of affective processing, Levineet al (1997) found significantly lower levels of emotionalawareness and more intense negative responses to stan-dardized everyday life events. In a study using affectivestimuli related to the BPD subjects’characteristic fear ofbeing abandoned, self-ratings indicated more intense emo-tional experiences and an increased sensitivity to evenlow-level emotional stimuli in subjects with impulsiveself-harming behavior (most of whom met the diagnosticcriteria of BPD) compared to other types of personalitydisorders (Herpetz et al 1997). In addition to the cognitiveevaluation of subjective emotional experiences, psycho-physiological responses during experimental emotionswere recently reported in BPD, which, however, gave noevidence for general affective hyperarousal in BPD (Her-pertz et al 1999).

So far, the understanding of BPD is limited to the extentthat no neurofunctional correlates of abnormal emotionalprocessing have yet been identified. Neuroimaging studiesin normal and depressed patients indicate greater subcor-tical than cortical involvement in the processing of emo-tions and a crucial role for the amygdala in processingnegative emotions (Irwin et al 1996; Morris et al 1998;Schneider et al 1995, 1997). A number of positronemission tomography (PET) and functional magnetic res-onance imaging (fMRI) studies based on the induction ofnegative affect in volunteers in response to visual stimuli

From the Department of Psychiatry and Psychotherapy (SCH, BW, HS), Interdis-ciplinary Center for Clinical Research–Central Nervous System (SCH, TMD,TK, SGE, KW, AT), Department of Neuroradiology (TK, SGE, AT), andSection Neuropsychology at the Clinic of Neurology (KW), Medical Faculty ofAachen Technical University – RWTH Aachen, Germany.

Address reprint requests to Sabine C. Herpertz, MD, Medical Faculty of AachenTechnical University, RWTH Aachen, Pauwelsstr. 30, Aachen D-52057,Germany.

Received September 20, 2000; revised December 27, 2000; accepted January 4,2001.

© 2001 Society of Biological Psychiatry 0006-3223/01/$20.00PII S0006-3223(01)01075-7

showed leftsided (Morris et al 1998; Schneider et al 1997)or bilateral activation in the amygdala (Breiter et al 1996;Irwin et al 1996). The amygdala is known to receive majorvisual input from sensory areas of the cortex as well asdirectly from the thalamus, the latter providing fast re-sponses to simple perceptual and associative aspects ofexternal stimuli (LeDoux 1996). In addition to subcorticalpathways of emotional processing, which are thought toact automatically even without awareness of the stimuli(Whalen et al 1998), medial prefrontal cortical structuresare involved in assigning meaning to emotional stimuli(Teasdale et al 1999), or, more generally, in consciouslyexperiencing emotion (Lane et al 1997; Reiman et al1997). The ventrolateral and orbital prefrontal cortex hasstrong interconnections with subcortical areas implicatedin emotional behavior and may play a role in correctingemotional responses (Drevets 1998).

This fMRI study, which made use of standardizedaversive emotional stimuli, aimed to clarify brain struc-tures mediating abnormal emotional responses in BPD.We hypothesized that BPD subjects, in contrast to healthyvolunteers, would show a higher degree of activation inlimbic/paralimbic structures, which are known to mediateintense negative emotional responses, the amygdala inparticular. We were further interested in those corticalareas that are directly interconnected with the amygdala.

Methods and Materials

SubjectsSix right-handed female BPD inpatients and six age-matchedright-handed female healthy volunteers participated in the fMRIstudy. BPD subjects were consecutively admitted to an inpatienttreatment program and were free of medication. Assessment ofBPD was according to DSM-IV by two independent raters (B.W.and S.H.) using a structured interview, the International Person-ality Disorder Examination (IPDE; Loranger et al 1996). Bothraters were trained in this interview, and interrater reliability forBPD diagnosis was 1.00. To secure a homogeneous group ofaffectively unstable BPD patients, only those subjects who metthe criteria of affective instability (Item 6) were included. Theinvestigation took place the week before discharge to ensure thatdefinite BPD trait features would be investigated rather than stateconditions of serious affective disturbance. Only female subjectswere chosen, because gender may influence affective responsesto specific stimuli. Using a standardized clinical interview basedon DSM-IV diagnostic criteria (Margraf et al 1994), BPDsubjects were checked to ensure that they were neither sufferingfrom additional Axis 1 disorders (schizophrenia, major depres-sion, anxiety disorders including posttraumatic stress disorder)nor showing any signs of current alcohol or drug abuse. Nosubject scored higher than 5 on the Hamilton Depression RatingScale (Hamilton 1960). Because of these strict requirements, outof nine consecutive subjects screened for the study, only six wereincluded. Control subjects were recruited through bulletin board

announcements and consisted of college students and nonclinicalhospital staff. They were checked to ensure that they had nohistory of psychiatric treatment nor medical problems, by meansof a comprehensive clinical assessment. Both groups were highlycomparable in terms of age (BPD 26.26 8.1 years, control subjects27.2 6 4.5 years) and education (BPD 11.56 1.2 years, controlsubjects 12.06 1.5 years). Subjects with BPD scored significantlyhigher on the trait subscales of the State-Trait Anxiety Scale (STAI;Spielberger et al 1970) and of the State-Trait Anger ExpressionScale (STAXI; Spielberger et al 1985) [trait anxiety:t(10) 5 5.38,p 5 .0003; trait anger:t(10)5 2.27,p 5 .04]. Thus, self-report datareflected a pervasive readiness to react with intense emotions, suchas intense anxiety or anger. All subjects gave written informedconsent to their participation in the study after receiving a fulldescription of the study. The protocol was approved by the localethics commission.

Affect Elicitation and Stimulus PresentationStimulus material was taken from a standardized pool of slides,the International Affective Picture System (Center for the Studyof Emotion and Attention 1998). It included 12 highly arousingunpleasant slides (e.g., mutilated bodies, crying children, scenesof violence and danger1) and 12 neutral slides (householdobjects, plants2). The visual stimuli were projected via laptop andbeamer onto a rear screen that was viewed through a mirrormounted on a standard bird-cage head coil. Subjects viewedneutral slides for 1–2 min before the actual experiment so thatimages could be focused and centered in the visual field of eachsubject. The 24 slides appeared for 6 sec each in random order,followed by a 10-sec rest, during which a blank screen appeared.This inter-stimulus interval was selected according to the resultsof a small pilot study with three subjects that indicated that in theregion of the amygdala 16 sec are needed to ensure that theresponse curve has returned to baseline before the next imageoccurs. Each subject was instructed to attend to the series ofslides that—to a greater or lesser extent—may induce emotionsand to avoid moving as much as possible while in the scanner.One hundred ninety-six brain volumes were acquired in a totalscanning time of 9:48 min. Subjective ratings of valence (plea-sure vs. aversion) and arousal (activation vs. calmness) weretaken immediately after the imaging session, showing each slideagain and using the Self-Assessment Manikin, a nine-point visualanalog scale (Lang 1980). In addition, the emotional state wasassessed before the experiment began using the same visualanalog scale.

MRI AcquisitionFunctional magnetic resonance imaging was performed on a 1.5Tesla Gyroscan ACS NT Powertrak 6000 (Philips MedicalSystems, Best, Netherlands) equipped with echo-planar imagingcapabilities. The subjects were scanned using the standard headcoil while the head was immobilized using Velcro straps and

1 Negative stimuli: No 1300, 2900, 3000, 3030, 3060, 3102, 3110, 3170, 3301,6230, 6313, 9410.

2 Neutral stimuli: No 5500, 7002, 7004, 7006, 7009, 7010, 7025, 7100, 7140, 7217,7224, 7235.

Abnormal Amygdala Functioning in Borderline PD 293BIOL PSYCHIATRY2001;50:292–298

foam rubber pads. Field homogeneity was optimized for eachsubject before each scan using automatic shimming sequences.After localizing images, a high-resolution, T1-weighted volumet-ric anatomic scan was acquired for three-dimensional (3-D)reconstruction of the brain and for Statistical Parametric Map-ping (SPM99) input (3-D fast field echo sequence [FFE] withrepetition time [TR]: 30 m/sec, echo time [TE]: 4.5 m/sec, flipangle: 30°, matrix: 2563 256, field of view (FOV): 2203 176mm, 90 contiguous 2-mm slices). This sequence was obtainedbefore the functional scans to rule out gross pathology and tofamiliarize subjects with the scanner. For functional imaging, ablood oxygenation level dependent (BOLD) contrast multislicesingle-shot echo planar T2*-weighted gradient-echo sequencewas used (imaging parameters: TR: 3000 msec, TE: 40 m/sec,flip angle: 90°, matrix: 643 64, FOV: 2303 230). Twentycontiguous, 5-mm thick axial slices were acquired during eachTR. Slice orientation was slightly angled to the anterior commis-sure-posterior commissure (AC-PC) line to cover the wholehippocampal region and to reduce susceptibility-induced artifactsof the skull base (Krings et al 1999). Scans of the dorsalprefrontal areas could not be included because of technicallimitations of the scanner at the time the investigations tookplace.

Image Analysis and Statistical AnalysesImage analysis was done by assessing condition-specific effects,thereby subtracting neutral trials from negative trials. Imageanalysis was done on a Sun Microsystems (Palo Alto, CA)workstation using SPM99 (Wellcome Department of CognitiveNeurology, London, UK) to identify significant areas of activa-tion. The spatial movement (x, y, z-translation and x, y, z-rotation) of the subject during the experiment was estimated byleast-square minimization between the first volume as referenceand all volumes of each scan using SPM99 realignment algo-rithm. The six estimated movement parameters were used toreslice each volume with the rigid body transformation ofSPM99. All participants’ MRI volumes were coregistered withthe first image. For spatial normalization of data, 12 parameters(six linear and six nonlinear) were used as provided by theSPM99 echo-planar imaging (EPI)-template. Then, functionalimages were smoothed with a 103 103 10-mm Gaussian filter.Finally, MNI (Montreal Neurologic Institute) coordinates weretransformed to the Talairach space (Talairach and Tournoux1988) by using a correction procedure (Medical Research Coun-cil 1999). The SPM single_subj_T1.img template, a T1-weightedMRI of a representative individual, supplied with SPM99, wasused for the anatomical projection of the data.

Statistical parametric maps of condition-specific group effectsin the BPD and control group were done for a voxelwisesignificance level ofp 5 .0001, that is the corrected type-I errorlevel for multiple testing for the entire volume analyzed, usingfixed-effects statistical analyses. In addition to assessing con-trasts by subtracting neutral from negative trials in each group,we also looked for an interaction effect between the negative–neutral conditions and the BPD–control subects groups (groupby condition interaction) by subtracting the negative–neutralcontrast of the BPD group from the negative–neutral contrast of

the control group. Owing to the small sample size, this interac-tion effect was tested on a corrected voxelwise significance levelof p 5 .05. Fixed-effects analyses assume that each subjectmakes the same, fixed contribution to the observed activation andtherefore discount random variations from subject to subject(Friston et al 1999). Therefore, in a second step a random-effectsanalysis was used in the clinical group where “random” meansthat the study allowed the expression of each subject’s activationto be modeled as a random variable, comparing the averageactivation to the variability of that activation over subjects(Friston et al 1999). With regard to the small sample size,however, the analyses were done on a voxelwise significancelevel of an uncorrectedp 5 .01, so that inference about thepopulation from which the subjects came can only be made withcaution. Finally, single-case analyses of the BPD subjects weredone to check how group effects that were found in fixed-effectsanalyses but could not be replicated in random-effects analyseswere present in each single subject.

To examine changes of emotional self-ratings as a function ofaffective stimulus valence in addition to group effects, repeated-measures anayses of variance (ANOVAs) were done on theexperimental data (i.e., subjective ratings of valence and arous-al), with diagnostic group (BPD, normal control subjects) as thebetween-subjects factor and slide valence category (unpleasant,neutral) as the within-subject factor.T tests were used to test forgroup differences in psychometric data.

Results

Self-Report

According to self-ratings, the emotional state did not differbetween the two groups before the beginning of theexperiment. Post-scan self-report ratings indicated thatslide stimuli evoked the intended feelings, because repeat-ed-measures ANOVA showed a strong overall slide va-lence effect [valenceF(1,10)5 47.69,p 5 .0001; arousalF(1,10)5 52.87,p 5 .0001], unpleasant slides elicitingfeelings that were significantly more negative and morearousing than those reported in response to neutral slidesin the total sample. Self-report data revealed no overallgroup effect in the subjective slide evaluation and nogroup3 slide valence interaction effect.

Comparison of Brain Activation in Response toNegative and to Reference Neutral Pictures.

Number of voxels in the cluster (k) and voxel (T) levels aswell as Talairach coordinates of the significant activationareas are illustrated in Table 1 for the BPD group as wellas for the control group. As shown in Figure 1, fixed-effects analysis on a significance level of correctedp 5.0001 revealed strong bilateral BOLD fMRI signal inten-sity changes in the amygdala in BPD subjects (rightamygdala: k5 31, T 5 8.40; left amygdala: k5 22, T 57.57). The amygdala was not activated in the control group

294 S.C. Herpetz et alBIOL PSYCHIATRY2001;50:292–298

when subjects were viewing negative rather than neutralpictures (neither on a significance level of correctedp 5.0001 nor of correctedp 5 .05). Both groups, BPD andcontrol subjects, showed activation in the temporo-occip-ital cortex on both sides; however, in BPD subjectsactivation was related to a larger ventral temporal region,including, particularly, the fusiform gyrus on both sides.

By subtracting the negative–neutral contrast of the BPDgroup from the negative–neutral contrast of the controlgroup, we found a significant group by condition interac-tion (Figure 2), with only the BPD group showing asignificant activation of the amygdala that met the strictcorrected significance criterion (right amygdala: k5 23,T 5 6.21, correctedp 5 .05; left amygdala: k5 9, T 55.58; correctedp 5 .05). Results of the random-effectsanalysis in BPD, which was performed despite the small

sample size, also indicated a significant bilateral activationof the amygdala (right amygdala: k5 123, T5 7.29; leftamygdala: k5 107, T 5 6.33; uncorrectedp 5 .01) inaddition to responses in the temporo-occipital cortex, thefusiform gyrus in particular, on the left (k5 290, T 512.28; uncorrectedp 5 .01) and on the right side (k5 411,T 5 11.83, uncorrectedp 5 .01).

According to a fixed-effects analysis in the BPD group,a further area of activation was located in the right frontalinferior gyrus (Brodmann area [BA] 47) and in the leftmedial frontal gyrus (BA 10) of the BPD subjects.Looking at single cases, area 47 was activated in four, area10 in three out of the six BPD subjects.

Discussion

To our knowledge, this is the first functional neuroimagingstudy in BPD patients. While processing standardized

Figure 1. Response to the presentation of emotionally aversiveslides in female patients with borderline personality disorder(left) and in normal volunteers (right). Activation in the bilateralamygdala and in the fusiform gyrus is shown on the axial slice (zaxis5 216) in borderline personality disorder, but not in healthyvolunteers: Data result from fixed-effects analyses with activa-tion set at a threshold of correctedp 5 0.0001.

Table 1. Brain Regions Showing Significant Activation in Borderline Personality Disorder (BPD)and in Control Subjects (Data from Fixed-Effects Analyses)

Region

Talairach co-ordinates (mm)

k Tx y x

Patients with BPD (n 5 6)Fusiform gyrus, BA 37, Ra 40 251 217 571 11.62Fusiform gyrus, BA 37, La 240 251 217 418 10.90Amygdala R 16 25 217 31 8.40Amygdala L 220 28 213 22 7.57Frontal inferior gyrus, BA47, R 48 27 2 11 6.74Frontal medial gyrus, BA 10, L 28 55 12 26 6.72

Control subjects (n 5 6)Medial temporal gyrus, BA 39, R 44 258 10 137 9.24Medial occipital gyrus, BA 39, L 252 273 7 167 8.09

Voxelwise significance level of correctedp 5 .0001.aClusters also include more dorsally located regions of the medial occipital gyrus (right side: x5 52, y 5 270, z5 7, T 5

11.06, BA 39; left side: x5 252, y 5 273, z 5 7, T 5 9.33, BA 39). k, number of voxels in cluster; T, voxel levels; BA,Brodmann area; R, right; L, left.

Figure 2. Response to the presentation of emotionally aversiveslides in borderline personality disorder on a coronar slice (left;y axis5 24) and on an axial slice (right; z axis5 216). Resultsfrom the analysis of the group by condition interaction, subtract-ing the negative–neutral contrast of the borderline group from thenegative–neutral contrast of the control group, show activation inthe right amygdala (x5 16, y 5 21, z 5 217) and in the leftamygdala (x5 220, y 5 25, z 5 213) (correctedp 5 .05).

Abnormal Amygdala Functioning in Borderline PD 295BIOL PSYCHIATRY2001;50:292–298

negative emotional stimuli, female BPD patients showedan intense activation pattern in the amygdala on both sides.In the control group, activation of the amygdala, which isthought to mediate intense emotions, was not found usingthe same set of emotional slides. Although only six BPDsubjects were scanned, the prominent role of the amygdalain the processing of negative emotions was revealed in allstatistical analyses, including the analysis of the group bycondition interaction (i.e., reducing the responses to thenegative minus neutral slides of the control group from thecorresponding contrast of the BPD group). This analysisstill indicated bilateral amygdala activation. A furtherinteresting finding was the strong bilateral activation ofthe fusiform gyrus in BPD subjects compared to controlsubjects, which was supported by random-effects analysis.Besides, the BPD group as opposed to the control groupshowed an activation of the left frontal medial gyrus (BA10) and of the right frontal inferior gyrus (BA 47). Theseareas of activation found in the prefrontal cortex were lessprominent and cannot be generalized, because they werenot found in all BPD subjects and could not be identifiedin random-effects analyses.

In comparing our finding of enhanced amygdala acti-vation in BPD to those reported for other psychiatricdisorders, a similar amygdala activation was noted inpatients with posttraumatic stress disorder (Rauch et al1996), and in obsessive-compulsive disorder patients(Breiter et al 1996) during fMRI scanning of their pro-voked symptoms. Amygdala activation in both anxietydisorder studies was interpreted to be owing to assessmentof aversive or threat-related events and recall of emotionalmemories (Breiter et al 1996). Several studies have sug-gested that the amygdala is not only activated directly bysensory information originating in thalamic and corticalareas but also by thoughts and memories organized in thehippocampus (Aggleton 1992).

Activation of the amygdala may be regarded as amanifestation of a neurobiological fear reaction, the moreas the BPD subjects of our study were shown to have ahigh trait anxiety; however, amygdala involvement hasalso been reported in depressive subjects, in which in-creased activity has been found in the limbic–thalamo–cortical circuit, comprising the amygdala, the thalamus,and the orbital prefrontal cortex (Drevets 1998, 1999).Therefore, amygdala activation may be a biological indi-cator of intense aversive emotions in the sense of a highintensity of affective associations to given stimuli.

Similar to other neuroimaging studies of experimentallyinduced aversive emotions (Lane et al 1997; Reiman et al1998; Teasdale et al 1999), medial prefrontal activationwas also seen in three of the six BPD subjects. This areahas been suggested to be associated with processingaffect-related meanings, either those accessed by external

stimuli or those related to memory. Therefore, the medialprefrontal cortex appears to mediate subjective emotionalresponses.

An abnormal elevation of cerebral hemodynamics in theventrolateral prefrontal cortex, which was found in fourBPD subjects of our group, has also been reported duringinduced aversive emotional states in patients sufferingfrom anxiety disorders or depression (Drevets 1999).These parts of the prefrontal cortex are directly connectedwith the basal nucleus of the amygdala. They are regardedas a gateway for distinctive sensorial information and maymodulate or inhibit amygdala-driven emotional responsesand thus provide top-down control of the amygdala (Dre-vets 1999; Morgan et al 1993; Rauch et al 1998).

Our finding of a strong activation of the temporo-occipital cortex during visually induced emotions in bothgroups is consistent with the data of Frederikson et al(1997), who performed a positron emission tomographystudy during anxiety provocation with visual phobogenicstimuli. As in our study, these authors described increasedneural activity to fear-related stimuli compared to neutralstimuli in the secondary visual cortex, which—accordingto their interpretation—suggests increased vigilance pro-voked by the stimuli with higher perceptual and behavioralsignificance; however, the BPD and control group alsoshowed differences in the activation pattern of brain areasprocessing visual information. BPD subjects showed thestrongest temporal activation in BA 37 in the fusiformgyrus, a region implicated in complex visual featuredetection and recognition of facial expression (George etal 1993). The activation of the fusiform gyrus in theaversive versus neutral stimuli comparison found in BPDbut not in control subjects could be due to modulation viaback-projections from the amygdala, which has an ana-tomic connection to the infero-temporal lobe visual re-gions. Modulation of perceptual cortex through the amyg-dala could have the function of strengthening attention toemotionally relevant environmental stimuli (Breiter et al1996). This finding provides further evidence for in-creased sensitivity to emotional aspects of the environ-ment typical of BPD individuals (Herpetz et al 1997).

In conclusion, our data represent preliminary findingsdemonstrating neurofunctional correlates of abnormallyenhanced limbic processing of emotional stimuli in BPD.Taking together neuroimaging findings from differentnosological entities, hyperresponsivity of the amygdalamight represent an oversensitization to aversive emotionalstimuli; however, excessive signal intensity could alsoreflect attenuated habituation of response rather than a trueexaggeration of response within the amygdala of BPDsubjects. A strong habituation of the BOLD fMRI signalintensity in response to emotional stimuli has been re-ported for normal volunteers (Breiter et al 1996; Buechel

296 S.C. Herpetz et alBIOL PSYCHIATRY2001;50:292–298

et al 1996; LaBar et al 1998) and, because regionalhemodynamic changes parallel neuronal activity (Kringset al 1999), this finding is consistent with the suggestionthat the amygdala serves as a rapid, transient informationprocessing pathway for behaviorally relevant stimuli (Le-Doux 1996). Such an adaptive habituation of amygdalaresponses may be disturbed in subjects showing long-lasting, slowly subsiding emotional reactions, thus “a slowreturn to emotional baseline” (Linehan 1993). A temporalanalysis of the hemodynamic amygdalar response couldgive further information. The activation pattern found inthe prefrontal cortex suggests enhanced activation of thoseventrolateral areas that are thought to exhibit top-downcontrol over limbic pathways. Because this result isexclusively based on activation found in a fixed-effectsanalysis and was only obtained in four of the six BPDsubjects, it needs to be replicated in further functionalneuroimaging studies including a higher number of sub-jects. Medial prefrontal activation, seen in half of the BPDsubjects, suggests that enhanced processing of emotionsmay not only include subcortical but also cortical areasmediating conscious emotions. In our study, subjectiveratings of emotional experiences, which were performedimmediately after the scanning session, did not reflectstronger conscious emotional responses in BPD subjectscompared to control subjects.

Because of certain methodical limitations, one shouldgeneralize from these results with caution. The mainlimitation of the present study is the small cohort size;however, even when random-effects analyses were applieddespite the small sample size, the main findings fromfixed-effects statistical analyses could be confirmed. Fur-thermore, the careful recruitment of clinical subjects withsevere BPD but without a major psychiatric co-morbiditymay strengthen our findings. The exclusion of subjectswith an additional major psychiatric disorder, depressionin particular, suggests that the changes observed in thecerebral activation pattern may be a trait feature. Never-theless, the trait character of the findings should be furtherconfirmed in a test-retest study. A number of neuroimag-ing studies showed activation in the amygdala, whengenerating emotions in healthy volunteers (George et al1995; Irwin et al 1996; Schneider et al 1997). Differencesin findings might have methodological reasons: In thesestudies, functional data acquisition was mostly restrictedto slices covering the amygdala, thus yielding a highernumber of functional images in this exclusive region ofinterest, whereas we chose a quasi whole-brain fMRIapproach. Differences might also be consistent with find-ings of other research groups demonstrating rapid amyg-dala habituation to repetitive presentations of emotionalstimuli in healthy subjects (Breiter et al 1996; Buechel etal 1998; LaBar et al 1998). Another limitation that needs

to be considered is susceptibility artifacts that can occur intemporal regions in particular, and which increase withslice thickness. We used the rather large slice thickness toobtain a better signal-to-noise ratio for the activatedsubcortical and cortical areas and to provide a rather fullbrain coverage. Finally, in future studies on female BPDsubjects, the phase of the menstrual cycle should berecorded. Recent data suggest that under estrogen thefemale brain shows a marked increase in perfusion at leastin cortical areas that are thought to result from changes inthe cerebral vascular anatomy (Dietrich et al 2001);however, this estrogen-related change in perfusion appearsto depend on the size of the cerebral area concerned andthus is expected to be minor in small structures as theamygdala.

These preliminary data encourage the use of functionalneuroimaging techniques in future research on BPD tofurther analyze and understand cerebral processing ofemotional stimuli in abnormal personality.

This research was supported with a grant from the InterdisciplinaryCenter for Clinical Research of the Medical Faculty, RWTH Aachen. Theauthors thank S. Pollrich, W. Reith, and S. Keme´ny for guidance onimage acquisition and analysis.

ReferencesAggleton J (1992):The Amygdala: Neurobiological Aspects of

Emotion, Memory, and Mental Dysfunction. Wiley-Liss: NewYork.

Breiter HC, Etcoff NL, Whalen PJ, Kennedy WA, Rauch SL,Buckner RL, et al (1996): Response and habituation of thehuman amygdala during visual processing of facial expres-sion.Neuron17:875–887.

Breiter HC, Rauch SL, Kwong KK, Kennedy DN, Savage CR,Olivares MJ, et al (1996): Functional magnetic resonanceimaging of symptom provocation in obsessive-compulsivedisorder.Arch Gen Psychiatry53:595–606.

Buechel C, Morris J, Dolan RJ, Friston KJ (1998): Brain systemsmediating aversive conditioning: An event-related fMRIstudy.Neuron20:947–957.

Center for the Study of Emotion and Attention (1999):TheInternational Affective Picture System: Digitized photo-graphs.Gainesville, FL: The Center for Research in Psycho-physiology, University of Florida.

Dietrich T, Krings T, Neulen J, Willmes K, Erberich S, Thron A,Sturm W (2001): Effect of blood estrogen level on corticalactivation patterns during cognitive activation as measured byfunction MRI. Neuroimage13:425–432.

Drevets WC (1998): Functional neuroimaging studies of depres-sion: The anatomy of melancholia.Annu Rev Med49:341–361.

Drevets WC (1999): Prefrontal cortical-amygdalar metabolism inmajor depression.Ann N Y Acad Sci877:614–637.

Frackowiak RSJ, Friston KJ, Frith CD, Dolan RJ, Mazziotta JC

Abnormal Amygdala Functioning in Borderline PD 297BIOL PSYCHIATRY2001;50:292–298

(1997):Human Brain Function. Academic Press: San Diego,43–48.

Frederikson M, Fischer H, Wik G (1997): Cerebral blood flow.J Clin Psychiatry58(suppl 16):16–21.

Friston KJ, Holmes AP, Worsley RJ (1999): How many subjectsconstitute a study?Neuroimage10:1–5.

George MS, Ketter TA, Gill DS, Haxby JV, Ungerfeider LG,Herscovitch P, et al (1993): Brain regions involved inrecognizing facial emotion or identity: An oxygen-15 PETstudy.J Neuropsychol5:384–394.

George MS, Ketter TA, Parekh PI, Horwitz B, Herscovitch P,Post RM (1995): Brain activity during transient sadness andhappiness in healthy women.Am J Psychiatry152:341–351.

Hamilton M (1960): A rating scale for depression.G NeurolNeurosurg Psychiatry23:56–62

Herpertz S, Gretzer A, Steinmeyer EM, Mu¨hlbauer V, Schu¨rkensA, Sab H (1997): Affective instability and impulsivity inpersonality disorder: Results of an experimental study.JAffect Disord44:31–37.

Herpertz S, Kunert HJ, Schwenger UB, Sass H (1999): Affectiveresponse in borderline personality disorder—a psychophysi-ological approach.Am J Psychiatry156:1550–1556.

Irwin W, Davidson RJ, Lowe MJ, Mock BJ, Sorenson JA, TurskiPA (1996): Human amygdala activation detected with echo-planar functional magnetic resonance imaging.Neuroreport7:1765–1769.

Krings T, Erberich SG, Ro¨ssler F, Reul J, Thron A (1999): MRblood-oxygenation-level-dependent signal differences in pa-renchymal and large draining vessels. Implications for func-tional magnetic resonance imaging.Am J Neuroradiol20:1907–1914.

LaBar KS, Gatenby JC, Gore JC, LeDoux JE, Phelps EA (1998):Human amygdala activation during conditioned fear acquisi-tion and extinction: A mixed-trial fMRI study.Neuron20:937–945.

Lane RD, Fink GR, Chau PM, Dolan RJ (1997): Neuralactivation during selective attention to subjective emotionalresponses.Neuroreport8:3969–3972.

Lang PJ (1980): Behavioral treatment and bio-behavioral assess-ment: Computer applications. In: Sidowski JB, Johnson JH,Williams TA, editors. Technology in Mental Health CareDelivery Systems.Norwood, NJ: Ablex, 119–137.

LeDoux JE (1996):The Emotional Brain. New York: Simon andSchuster.

Levine D, Marziali E, Hood J (1997): Emotion processing inborderline personalty disorders.J Nerv Ment Dis185:240–246.

Linehan MM (1993):Cognitive-Behavioural Treatment of Bor-derline Personality Disorder. New York: Guilford Press.

Loranger AW, Susman VL, Oldham HM, Russakoff LM (1996):International Personality Disorder Examination (IPDE): AStructural Interview for DSM-IV and ICD-10 PersonalityDisorders [Mombour W et al, German translation]. WhitePlains, NY: New York Hospital Cornell Medical Center,Westchester Division.

Margraf J, Schneider S, Ehlers A (1994):Diagnostisches Inven-tar fur psychische Sto¨rungen[Diagnostic Inventory for Men-tal Disorders]. Berlin: Springer.

Medical Research Council, Cognition and Brain Sciences Unit(1999):The MNI Brain and the Talairach Atlas. CambridgeImagers Home Page, available at matthew.brett@mrc-cbu.cam.ac.uk.

Morgan MA, Romanski LM, LeDoux JE (1993): Extinction ofemotional learning: Contribution of medial prefrontal cortex[letter]. Neuroscience163:109–113.

Morris JS, Frith CD, Perrett DL, Rowland D, Young AW, CalderAJ, et al (1996): A differential neural response in the humanamygdala to fearful and happy facial expressions.Nature383:812–815.

Morris JS, Ohman A, Dolan RJ (1998): Conscious and uncon-scious emotional learning in the human amygdala.Nature393:467–470.

Rauch SL, Jenike MA, Alpert NM (1994): Regional cerebralblood flow measured during symptom provocation in obses-sive-compulsive disorder using oxygen 15-labeled carbondioxide and positron emission tomography.Arch Gen Psy-chiatry 51:62–70.

Rauch SL, Shin LM, Whalen PJ, Pitman RK (1998): Neuroim-aging and the neuroanatomy of posttraumatic stress disorder.CNS Spectrums3(Suppl 2):30–41.

Rauch SL, van der Kolk BA, Fisler RE, Alpert NM, Orr SP,Savage CR, et al (1996): A symptom provocation study ofposttraumatic stress disorder using positron emission tomog-raphy and script-driven imagery.Arch Gen Psychiatry53:380–387.

Reiman EM, Lane RD, Ahern GL, Schwartz GE, Davidson RJ,Friston KJ, et al (1997): Neuroanatomical correlates ofexternally and internally generated human emotion.Am JPsychiatry154:918–925.

Schneider F, Grodd W, Weiss U, Klose U, Mayer KR, Nagele T,et al (1997): Functional MRI reveals left amygdala activationduring emotion.Psychiatry Res76:75–82.

Schneider F, Gur RE, Alavi A, Harper Mozley L, Smith RJ,Mozley PD, et al (1995): Mood effects on limbic blood flowcorrelate with emotional self-rating. A PET study withoxygen-15 labeled water.Psychiatry Res Neuroimaging61:265–283.

Spielberger CD, Gorsuch RL, Lushene RE (1970):State-Trait-Anxiety Inventory, Manual. Palo Alto, CA: Consulting Psy-chologist Press.

Spielberger CD, Johnson EH, Russell SF, Crane RJ, Jacobs GA,Worden TJ (1985): The experience and expression of anger;construction and validation of an Anger Expression Scale. In:Chesney MA, Rosenman RH, editors.Anger and Hostility inCardiovascular and Behavioral Disorders. Washington, DC:Hemisphere, 5–30.

Talairach J, Tournoux P (1988):Co-Planar Stereotaxic Atlas ofthe Human Brain. 3-Dimensional Proportional System: AnApproach to Cerebral Imaging.Stuttgart: Thieme.

Teasdale JD, Howard RJ, Cox SG, Ha Y, Brammer MJ, WilliamsSCR, et al (1999): Functional MRI study of the cognitivegeneration of affect.Am J Psychiatry156:209–215.

Whalen PJ, Rauch SL, Etcoff NL, McInerney SC, Lee MB,Jenike MA (1998): Masked presentations of emotional facialexpressions modulate amygdala activity without explicitknowledge.J Neurosci18:411–418.

298 S.C. Herpetz et alBIOL PSYCHIATRY2001;50:292–298