Neural correlates of attachment trauma in borderline personality
disorder: A functional magnetic resonance imaging study
Anna Buchheima,⁎,1, Susanne Erkb,1, Carol Georgec, Horst Kächelea, Tilo Kircherd,
Philipp Martiuse, Dan Pokornya, Martin Ruchsowf, Manfred Spitzerg, Henrik Walterb
aDepartment of Psychosomatic Medicine and Psychotherapy, University Ulm, Am Hochstraess 8, 89081 Ulm, Germany
bDepartment of Psychiatry, Division of Medical Psychology, University of Bonn, Germany
cDepartment of Psychology, Mills College, Oakland CA, United States
dNeuroimaging-L&F Experimental Psychopathology, Clinic for Psychiatry and Psychotherapy, University, RWTH Aachen, Germany
ePsychosomatic and Psychiatric Hospital, Bad Wiessee, Germany
fHospital for Psychiatry and Psychotherapy Christophsbad, Göppingen, Germany
gDepartment of Psychiatry III, University Ulm, Germany
Received 25 March 2007; received in revised form 4 July 2007; accepted 4 July 2007
Functional imaging studies have shown that individuals with borderline personality disorder (BPD) display prefrontal and
amygdala dysfunction while viewing or listening to emotional or traumatic stimuli. The study examined for the first time the
functional neuroanatomy of attachment trauma in BPD patients using functional magnetic resonance imaging (fMRI) during the
telling of individual stories. A group of 11 female BPD patients and 17 healthy female controls, matched for age and education,
told stories in response to a validated set of seven attachment pictures while being scanned. Group differences in narrative and
neural responses to “monadic” pictures (characters facing attachment threats alone) and “dyadic” pictures (interaction between
characters in an attachment context) were analyzed. Behavioral narrative data showed that monadic pictures were significantly
more traumatic for BPD patients than for controls. As hypothesized BPD patients showed significantly more anterior midcingulate
cortex activation in response to monadic pictures than controls. In response to dyadic pictures patients showed more activation of
the right superior temporal sulcus and less activation of the right parahippocampal gyrus compared to controls. Our results suggest
evidence for potential neural mechanisms of attachment trauma underlying interpersonal symptoms of BPD, i.e. fearful and painful
intolerance of aloneness, hypersensitivity to social environment, and reduced positive memories of dyadic interactions.
© 2007 Elsevier Ireland Ltd. All rights reserved.
Keywords: Borderline Personality Disorder; fMRI; Attachment disorganization; Anterior cingulate cortex; Superior temporal sulcus;
Borderline personality disorder (BPD) is character-
ized by extreme and enduring emotional instability
involving a range of intense affects, including rage,
panic, emptiness, loneliness, and characteristically
Available online at www.sciencedirect.com
Psychiatry Research: Neuroimaging 163 (2008) 223–235
⁎Corresponding author. Tel.: +49731 50061804;fax:+49731 5006
E-mail address: firstname.lastname@example.org (A. Buchheim).
1Both authors contributed equally to this study.
0925-4927/$ - see front matter © 2007 Elsevier Ireland Ltd. All rights reserved.
multifaceted emotional pain and fear of abandonment
(Lieb et al., 2004). Childhood maltreatment by a
caregiver (emotional neglect, physical and sexual
abuse) is one of the most important psychosocial risk
and prognostic factors for BPD pathology (Zanarini,
2000; Zanarini et al., 2006).
Clinically, an essential dimension of BPD patients is
their dysfunction of emotion-regulation systems com-
bined with the inability to adjust emotional responses
(Lieb et al., 2004). Studies using the startle reflex as a
measure for emotional hyper-reactivity reported evidence
that favored (Ebner-Priemer et al., 2005) and failed
(Herpertz et al., 1999) to support this hypothesis. Two
recent fMRI studies reported emotional hyper-reactivity
to emotional pictures (Herpertz et al., 2001) or faces
(Donegan et al., 2003). Two further recent studies
investigated brain activation during processing of auto-
biographical memory. One study found less activation in
other study looking at unresolved life events compared to
resolved life events found, among other regions,
increasing activation of amygdala and anterior cingulated
cortex (Beblo et al., 2006). Finally, reductions in
amygdala (and hippocampal) volume have been reported
for BPD patients (e.g. Driessen et al., 2000; Tebartz van
Elst et al., 2001; Irle et al., 2005). PET studies showed
prefrontal dysfunction in BPD patients in response to
listening to personal scripts of abandonment and abuse
(Schmahl et al., 2003, 2004).
No patient study to date has examined neural patterns
in relation to attachment, a basic behavioral system that
processes relationship-based emotional experience and
Attachment theory provides a powerful framework
for understanding the nature of close relationships, the
links between mental representations in patterns of
emotion regulation and psychopathology (Westen et al.,
2006). Researchers have used two measurement strat-
egies for assessing adult attachment, based on narrative
assessment or self-report. In the present study we refer
on the narrative tradition using interview assessments
(George et al., 1996; George and West, 2003; Main et
al., 1985). This approach classifies attachment through
examination of the person's state of mind with respect to
attachment as expressed in linguistic qualities of the
narratives. Classification falls into two main attachment
groups: organized/resolved and disorganized/unre-
solved. Disorganized/unresolved individuals are
flooded with painful affect, often evidenced through
verbal descriptions of intense fear or linguistic disori-
entation (Main et al., 1985). Studies concur that the
unresolved attachment classification predominates in
BPD patients, related particularly to lack of resolution of
physical and sexual abuse (Fonagy et al., 2000; Agrawal
et al., 2004). Attachment disorganization is considered
to be one core feature in understanding BPD psycho-
pathology in the context of affective and interpersonal
problems (Fonagy et al., 2003; Gabbard, 2005).
The attachment relationship is an essential biological
system that influences motivational and emotional
processes related to survival (Bowlby, 1969). Animal
studies suggest that limbic structures are involved in
attachment deprivation (Insel, 1997; Bauman et al.,
2004). Structural neuroimaging studies show reduced
hippocampus and amygdala volumes in patients report-
ing traumatic attachment histories (Tebartz van Elst
et al., 2003; Wignall et al., 2004).
Functional imaging studies investigating social
attachments have focused on healthy subjects so far.
Pictures of loved ones (e.g., spouse versus friend or own
versus other baby) (e. g. Bartel and Zeki, 2004;
Leibenluft et al., 2004) evoked cortical and subcortical
responses, including the cingulate cortex, insula, basal
ganglia, and orbitofrontal cortex. No fMRI studies have
examined brain activation while subjects tell stories
when the attachment system is activated.
fMRI data gathered while participants were speaking
continuously demonstrated that this approach can be
reliably applied to healthy controls and schizophrenic
patients with severe formal thought disorder (Kircher
et al., 2001). Recently, we measured attachment
representation in an fMRI environment in which healthy
participants told stories in response to the Adult
Attachment Projective (AAP), a validated attachment
measure described in detail below. We found robust
activation of visual, motor and language related areas
while talking to AAP pictures and activation of the right
amygdala related to attachment status and involvement
in the course of the task (Buchheim et al., 2006).
One key feature of interpersonal problems in BPD
patients is their intolerance of aloneness (Gunderson,
1996).Ina recent BPD study using the AAP measure ina
we examined different narrative responses to “monadic”
attachment pictures (characters facing attachment threats
alone) and “dyadic” attachment pictures (interaction
related traumatic dysregulation was operationally defined
as the frequency of occurrence of “traumatic fear
indicators” in the narratives. The results showed a higher
frequency of these words in unresolved patients than
to dyadic pictures.
224 A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235
In this study, we were interested to further investigate
traumatic dysregulation in borderline patients by analyz-
especially responses to stories associated with loneliness
and abandonment (monadic pictures). According to
Bowlby's (1969) conceptualization being alone is the
single most frightening experience for primates. Thus,
representations of being alone are thought to be the
strongest activators of the attachment system. On the
linguistic level, we predicted to find the same narrative
patterns as in our behavioral study (Buchheim and
George, in press). On the neural level we expected that
patients would show greater activation of brain regions
associated with fear and pain, for example, the amygdala
or the anterior cingulate cortex, in response to monadic
Thirteen female BPD patients were recruited from an
inpatient psychiatric hospital and compared to 21
healthy female volunteers, matched for age and
education. Controls were recruited for the study by an
advertisement in a local newspaper and leaflets
distributed in the Hospital of the University of Ulm.
All control subjects were physically healthy, without a
history of psychiatric disorder and did not use any
medication. Clinical diagnoses of BPD patients were
assessed by a trained psychiatrist (P.M.)2using the
Structured Clinical Interview for DSM-IV (SCID-I and
SCID-II) and the International Personality Disorder
Examination (IPDE). Exclusion criteria of all subjects
were serious medical or neurological illness (including
comorbid psychotic disorders, bipolar disorder, PTSD
and Dissociative Disorder), left handedness, metal in
body, and language problems. We examined the groups
in relation to important variables related to this study:
movement parameters, balance of attachment classifi-
cation groups in each sample, and patient medication.
Six subjects were excluded from our main analysis:
four controls (movementN2mm, see below), and two
patients classified as resolved (not enough to allow any
substantial group inferences). Inclusion of this subgroup
in a control analysis did not alter our results.
Control analyses also demonstrated no influence for
medication, therefore medicated and un-medicated BPD
patients were combined into a single group for the main
analyses. The final sample consisted of 11 BPD patients
did not affect group homogeneity with respect to age
(BPD: 27.8years±6.7, controls: 28.4years±7.5) and
education (BPD: 10.8years±1.4, controls: 10.9years±
1.6). Current depressive episode, current drug and/or
alcohol dependency or abuse were exclusion criteria.
Comorbidity in the final patient group included depres-
sion (n=6), anxiety or panic disorder (n=2), and
somatoform disorder (n=3). Seven patients (53.8%) had
lifetime depressive episode(s), four (30.8%) had lifetime
drug or alcohol abuse, five (38.5%) met current somato-
anxiety disorder, one patient (7.7%) fulfilled current
dissociative disorder (44.5%). With respect to traumatic
experiences in life cycle we documented life events,
which havebeenidentifiedaspotentialriskfactorsfor the
development of BPD (Paris, 1994). All but one patient
reported one or more of these experiences (n=9 sexual
abuse, n=3 violence, n=4 parental neglect, n=5
separation from parents, n=5 psychiatric morbidity of
parents, n=1 single traumatic life event); but none of
them fulfilled PTSD criteria. 45% (5/11) of the patients
were treated with psychotropic medication, including
low doses of neuroleptics (perazin, promethazine and
chlorprothixene, n=3), serotonin-reuptake inhibitors
(n=2) and lithium (n=1). After complete description of
the study to the subjects, written informed consent was
obtained. The protocol was approved by the local
institutional ethics committee.
2.2. Clinical assessment
The Dissociative Experience Scales (DES) (Bern-
stein and Putnam, 1986; Freyberger et al., 1999)
(absorption, dissociative amnesia, depersonlization/de-
realization subscales) was applied as a measure of
severity of dissociative symptoms. Severity of impul-
siveness was assessed by using the Barratt Impulsive-
ness Scale (BIS-10) (Barratt, 1985).
2.3. Attachment measure
The Adult Attachment Projective (AAP) is a vali-
dated measure to assess narrative patterns using a set
of eight pictures, one neutral and seven attachment
scenes. The pictures depict theory-derived attachment
events and are administered as follows: #2 “Child at
Window”; #3 “Departure”; #4 “Bench”; #5 “Bed”; #6
2P.M. has been trained and certified for reliable diagnosis in the
SCID-rating by Prof. Wittchen, Munich, Germany and by A.
Loranger MD, New York, regarding IPDE. In seven patients an
experienced Master's level psychologist conducted a second SCID
interview. Agreement between the raters was kappa=1.0 for both
BPD and lifetime depressive episode.
225A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235
“Ambulance”; #7 “Cemetery”; #8 “Corner”. There are
four “monadic” and three “dyadic” scenes (Fig. 1).
Individuals are instructed to tell a story: “Tell me what
led up to that scene, what are the characters thinking or
feeling, and what might happen next?” (George and
West, 2001, 2003). Individuals are classified on the
basis of verbatim narratives into one of two attachment
groups: resolved and unresolved. Unresolved attach-
ment in the coding system is defined as an individual's
failure to contain any frightening or threatening nar-
rative material, including words and phrases such as
death, attack, or devastation. This is termed attachment
dysregulation (George and West, 2003). Stories are
considered resolved when dysregulation is contained,
when characters utilize internal or relationship resources
that provide help or care (Table 2).
A large-scale psychometric investigation of the AAP
with 144 participants (George and West, 2003) showed
excellent inter-judge reliability, test–retest reliability
(retest after three months), discriminant validity and
construct validity using the established Adult Attach-
ment Interview (AAI) (George et al., 1996). The AAI is
a validated semi-structured interview asking individuals
to describe autobiographic childhood experiences with
caregivers (e. g. separations, loss, abuse).
In this study, two blind, reliable AAP judges inde-
Inter-rater agreement was 100%. AAP validity was tested
Group comparison of clinical scales and attachment trauma scales
Variable BPD (n=11) Controls (n=17)ES Exact U-test
M S.D.M S.D.ZP
State anxiety T1 (before fMRI)
State anxiety T2 (after fMRI)
GSI (SCL-90 General Symptom Index)
Barrett Impulsivity Scale Total Score
Dissociative Experience Scale Total Score
Attachment trauma scales in the Adult Attachment Interview
Score for loss experiences (scale 1–9)
Score for abuse experiences (scale 1–9)
P = significance of the two-tailed Exact test,⁎P≤0.05,⁎⁎P≤0.01,⁎⁎⁎P≤0.001.
Fig. 1. Examples of two attachment pictures from the Adult Attachment Projective © George and West (2003): “Bed” (dyadic picture) and
“Cemetery” (monadic picture). The AAP pictures depict events that according to theory and research activate the attachment system, for example,
illness, solitude, separation, loss and abuse. The black and white line drawings contain only sufficient detail to identify an attachment scene. Facial
expressions and other details are omitted or drawn ambiguously. The drawings were developed carefully to avoid gender and racial bias.
226A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235
based on convergent classifications with the AAI,
administered one month after fMRI acquisition and
classified by a blind trained AAI judge. The correspon-
dence between the AAP and AAI resolvedvs. unresolved
categories was highly significant (kappa=0.70).
Beyond overall classification we studied on a more
detailed level, what kind of words with respect to
attachment fear and trauma patients and controls used in
their AAP stories. AAP judges differentiated between so
called “normative” and “traumatic” fear indicators
according a detailed manual (George and West, 2004).
“Normative” fear indicators are defined as those
typically present because of AAP picture “pull”, based
on evaluations of several hundred stories in normative
and clinical samples: statements like “talking to the
deceased” in “Cemetery” or a character frightened by
separation in “Bench” are coded as “normative” fear
indicators (Table 2).
These markers do nothave thesame terrifyingquality
as “traumatic” fear indicators, such as the “deceased
talking back to the living” in “Cemetery” or the girl in
“Bench” described as suicidal and incarcerated. The two
judges agreed 100% on these narrative indicators. The
data we report here focus on the traumatic fear indicators
trauma (Fonagy et al., 2000).
2.4. Attachment task in an fMRI environment
of the AAP. The detailed procedure is described else-
where (Buchheim et al., 2006). Subjects were first
trained in the AAP story telling task prior to entering the
scanner using two non-AAP “neutral” (i.e., not attach-
ment scenes) pictures. The training procedure was
repeated two more times, if necessary. During scanning,
subjects were presented the standard AAP instruction
(“what led up to that scene, what are the characters
thinking or feeling, and what might happen next?”) for
10s and a fixation cross for 10s. This was followed by
one of the seven AAP pictures (120s). Subjects were
instructed to talk about the picture for 2min or as long as
presentation until beginning a new cycle of instruction
and picture presentation. The total procedure included 9
pictures, 2 neutral and 7 standard AAP attachment
stimuli. For detailed trial structure, see Fig. 2.
2.5. Data acquisition
A 1.5-Tesla Siemens Magnetom Symphony scanner
(Siemens, Erlangen, Germany), 64×64 voxels, FoV
192mm, slice thickness 4mm/1mm gap, 25 slices, TE/TR
40ms/2500ms, total acquisition time 25min (= 598
volumes, one session). The paradigm was presented
with fMRI compatible video-goggles (Resonance Tech-
nologies, Northridge, CA).Speech was digitally recorded
using an fMRI compatible microphone and Cool Edit
Pro (Syntrillium Software Cop. Phoenix, Arizona as
software). Head movement was minimized by using
padded earphones fixating the head within the gradient
Transcript example of a “Resolved” and “Unresolved” story to a “monadic” AAP picture “Bench
Resolved AAP storyUnresolved AAP story
(example of a control subject)
“Awomen is afraid, feels bad, had a fight with a friend, sits on a bench to
be alone and by herself. She is sitting and crying. Her friend was very
disappointed that she has not told him the truth for several times, so he
broke up with her. Now she feels abandoned and is afraid of the future.
She thinks about the fight and realizes that she has to say sorry. But she
is afraid that her friend would not talk to her, like her mother often did
when she was young. She is afraid. She is sitting there for a long time,
thinking about the problem. After a while she gets up and is trying to
get in contact with the friend to talk about everything.”
(example of a control subject)
“She is very sad, wants to hide herself under the bench, she is very
frightened, feels abandoned by everybody. Life can be so cruel. Her
friend does not love her anymore, because she has overweight. Her
mother broke up contact with her because she is not interested in her life
anymore. She is frightened about the future and she doubts that she ever
will meet someone who finds her attractive. I have no idea how this could
end. I think she sits there for ever, I really don't know.”
(example of a borderline patient)
“She feels homeless, it seems that she is incarcerated in jail, wants to
escape from this isolation, she thinks about suicide. It is also possible
that she is in a mental institution, because she has already tried to
commit suicide and now she has to be alone in an empty room. Nobody
helps her, and she has no relatives or friends. I have no idea. (long pause)
I think she only dreams of running away.”
“Normative fear indicators” are underlined italics, “traumatic fear indicators” are bold.
227A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235
Fig. 2. Paradigm, modeling and main effects: right side: at the upper right the trial structure is depicted, demonstrating one of the monadic pictures. In the middle, the digital speech recording is shown.
In the right lower part, the main effects of “picture” (see text) in the control andpatient group are shown(one sample t-test, Pb0.001 uncorrected, extent threshold 5 voxels, see also Table 4). On the left
side the model in SPM is depicted. Pre-narrative, narrative and post-narrative were modeled as separate regressors. The fourth regressor modeled onset of every single word, the following two
regressors modeled instruction and baseline. Moreover there are six regressors for movement parameters.
A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235
2.6. Data analysis
2.6.1. Behavioral scales
Group differences were analyzed using the Kruskal–
Wallis H-test and the exact Mann–Whitney U-test
(SPSS version 14). Non-parametric test procedures were
used because of the non-normal distribution of the
dependent variables. The magnitude of the group differ-
ences was expressed by the effect size (ES, Cohen's d).
2.6.2. Neuroimaging data
Analyses were carried out with SPM2 (www.fil.ion.
ucl.ac.uk) and MATLAB 6.1 (MathWorks, Natick,
Massachusetts). Preprocessing steps: 1) Motion correc-
tion by realigning them to the first volume of each
session, 2) spatial normalization (3×3×3mm), and 3)
smoothing (FWHM 8mm). Subjects with head move-
ment of N2mm within a trial cycle were excluded from
further analysis (four controls, no patients).
The regression model is depicted and explained in
detail in the legend of Fig. 2. All regressors except those
for motion were convolved with a function that modeled
a prototypical hemodynamic response. The variance of
each voxel was estimated for each trial according to the
General Linear Model. Individual regionally specific
effects of interest were calculated for each participant
using linear contrasts, resulting in a t-statistic for every
The effects of interest in this study were those for
monadic and dyadic pictures. The contrast for monadic
pictures included #3, #5 and #6. Picture 8 was excluded
line, for each subject, thereby including any mental
processes that occurred before the speaking phase.
Group differences were assessed at a second level
using random effects analysis. Three analyses were
performed. Analysis 1: main effects of “picture” were
calculated using one sample t-tests. Analysis 2 (main
analysis): one-way ANOVAs with three groups was
calculated for monadic and dyadic “pictures” contrasts-
resolved controls (n=10), unresolved controls (n=7)
and unresolved patients (n=11, medicated and un-
medicated). Within the ANOVAs, patients were con-
trasted against both control groups combined as well
as with each control group separately. A three-group
analysis with two conditions was calculated to test for an
interaction effect of group by picture category. Analysis
was calculated in order to control for effects of
attachment status and medication ((resolved controls
(n=10), unresolved controls (n=7), resolved patients
(n=2, medicated), unresolved medicated patients (n=4)
and unresolved un-medicated patients (n=7)).
T-statistics for each voxel were set at a threshold
of Pb0.001, uncorrected for multiple comparisons.
Results were corrected for extent threshold, resulting in
Pb0.05 at the cluster level. Brain areas were identified
using atlases (Talairach and Tournoux, 1988; Duvernoy,
3.1. Behavioral data
BPD patients differed significantly from controls in
all clinical scales (Table 1). The AAP classification
Three-group-comparison: occurrence frequency of traumatic fear indicators in the AAP
M S.D.M S.D.M S.D.P
Monadic (alone) pictures
Mann–Whitney U-test: P = significance of the two-tailed Exact test; Kruskal–Wallis H-test: P = significance estimated in 100000 Monte-Carlo
229A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235
distribution was 10 resolved and 7 unresolved controls,
and 2 resolved (included only in the fMRI control
analysis) and 11 unresolved patients. The difference
between the groups was significant. AAI analyses (Main
and Goldwyn, 1985) showed that patients were sig-
nificantly more unresolved with respect to sexual abuse
and loss through death of a significant person compared
to controls (Table 1).
Main effect of picture (pre-speech plus picture) from Analysis 1 for all, monadic and dyadic pictures in both groups, controls and patients
Region BA ControlsPatients
All MonadicDyadic AllMonadicDyadic
48,12,42, 3.43 54,3,42, 3.51
54,27,−6, 3.7954,27,−6, 3.77
Lateral prefrontal cortex
Superior frontal gyrus60,6,6,63, 3.970,3,63, 3.76 0,18,63, 3.99
Medial prefrontal cortex3,30,42, 3.68
8 0,30,36, 3.75
R 7 24,−72,48, 4.82 24,−72,48, 4.89
Precentral gyrusL 6
−42,−9,33, 4.39 −45,−6,33,
−42,−9,33 4.00 −51,−9,39, 3.96 −51,−9,36, 4.00
54,−3,27, 3.93 51,−3,24, 3.60
−63,−9,18, 3.63Postcentral gyrus
Superior parietal lobe
Superior temporal gyrus
Medial temporal gyrus
Parahippo campal/ lingual
−15,−102,3, 3.67 −21,−102,9, 4.55
−30,−81,12, 3.76 −33,−84,3, 4.41
Occipital cortex 36,−90,9, 6.40
−30,−90,6, 6.27 −30,−90,6,
36,−90,9, 6.54 36,−90,6, 5.93 30,−84,9, 4.90
For significant activated regions Talairach coordinates (x, y, z) as well as Z-value are given (one sample t-tests, Pb0.001 uncorrected).
230A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235
As predicted, BPD patients showed significantly
more traumatic fear indicators in the monadic sto-
ries, and not in the dyadic ones, as compared to both
control groups (Table 3). These results passed the
Bonferroni criterion of simultaneous inference (Pb
0.0167). The strongest difference was found between
unresolved patients and resolved controls. The Krus-
kal–Wallis test showed significant differences in
monadic pictures Window, Bench, and Cemetery. The
difference for Corner did not reach statistical signifi-
cance. Window, Bench, and Cemetery were selected for
Fig. 3. a, b, c. Group differences for monadic and dyadic pictures. The results are from the second level analysis for monadic pictures (n=3) or dyadic
pictures (n=3), respectively, thresholded at Pb0.001 at the voxel level and Pb0.05 at the cluster level (for exact location and z-values see text). The
figure shows effect sizes, bars indicate 90% confidence interval (and resolved controls n=10, unresolved controls, n=7, unresolved patients, n=11).
Note, that the groups are grey scale coded only for the analysis in which the effect is significant at the chosen level of significance.
231 A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235
3.2. fMRI analysis
Analysis 1: both control and patient groups showed
activations in visual (occipital), motor (precentral
cortex, basal ganglia and cerebellum) and language
related areas (temporal cortex), as well as in the anterior
cingulate, superior and middle frontal gyrus (Fig. 2 and
Table 4). This analysis was calculated to test replicabil-
ity of our pilot study (Buchheim et al., 2006).
Analysis 2 (main analysis, Fig. 3): as hypothesized
BPD patients showed significantly stronger activation of
the anterior midcingulate cortex (aMCC, x=3, y=18,
z=24, Z=4.43) than controls in response to monadic
pictures3. A similar activation trend for dyadic pictures
was not significant.
In response to dyadic pictures, BPD patients showed
less activation of the right parahippocampal gyrus (GH,
x=33, y=−39, z=−15; Z=4.31), and stronger activation
of the rightsuperior temporal sulcus (STS,x=60,y=−45,
z=24; Z=4.52) than controls. Again, similar activation
trends for monadic pictures were not significant. This
two subgroups of controls (resolved, unresolved) in
order to test whether this effect was due to diagnosis
or attachment classification. The results were the same,
indicating a diagnosis effect. No other significant acti-
vations were found.
Analysis 3 (control analysis): the results of analysis 2
remained unchanged when including the two resolved
patients and splitting the patient group by medication.
The effects sizes of medicated and un-medicated
patients did not differ significantly in all three regions
of group differences. The effect sizes of the resolved
patients were in between those of the unresolved
patients and controls. Resolved patients may be more
similar to healthy controls; however, there are too few
patients (n=2) to interpret this finding.
This study investigated the neural correlates of
attachment trauma in BPD patients versus controls
while telling stories in response to attachment-activating
scenes. As expected, BPD patients showed a higher
proportion of unresolved attachment classifications and
more traumatic fear indicators in monadic pictures than
controls. As hypothesized BPD patients showed signif-
icantly more dorsal anterior cingulate cortex activation
than controls in response to “monadic” pictures. In
response to dyadic pictures patients showed significant-
ly more activation of the right superior temporal sulcus
and less activation of the right parahippocampal gyrus.
4.1. Attachment trauma on a narrative level
In accordance with previous research (Fonagy et al.,
2000; Agrawal et al., 2004), the majority of BPD
patients were classified as unresolved. Convergent
classifications between the scanner-administered AAP
and the AAI administered outside the scanner confirm
that the fMRI-AAP procedure was feasible also for BPD
patients. The number of unresolved controls here,
mostly due to loss experiences, is greater than the
average percentage reported in healthy populations
(George and West, 2003). Unresolved patients had
significantly higher ratings for loss and abuse on the
AAI scales compared to unresolved controls, indicating
again that the combination of unresolved loss and abuse
is more likely to contribute to pathological distress than
experiences of loss alone (Lyons-Ruth et al., 2003).
The linguistic analysis of traumatic fear indicators
provides a more specific understanding of attachment
trauma in BPD patients. Unresolved BPD patients
manifested significantly greater traumatic dysregulation
in response to monadic pictures (Window, Bench,
Cemetery), whereas normative dysregulation predomi-
nated in unresolved controls. For example, in Cemetery,
patient stories described isolation, abandonment, mur-
der, suicide, and dissociated imagery (e.g., figures
floating above the ground). Controls predominantly
described typical graveyard contact with the deceased
(visit) or grief talk.
4.2. Neural correlates of attachment trauma
4.2.1. Monadic pictures: anterior cingulate cortex
As expected BPD patients (unresolved) showed
increased ACC activation in monadic pictures where
traumatic dysregulation indicators were present. ACC
activation is observed in response to pain and unpleas-
antness (Schnitzler and Ploner,2000).ACC activation in
healthy subjects is associated with social relationship
stimuli, including intimate relationships (Bartel and
Zeki, 2004), social exclusion (Eisenberger et al., 2003),
and pictures evoking grief (Gündel et al., 2003).
However, the ACC is not homogeneous (Vogt, 2005).
3Results for monadic pictures remain unchanged when including
picture 8 into the analysis, i.e. when including all monadic pictures.
This result was irrespective of the behavioural results, showing that
pictures 2, 4 and 7 differentiate clearly between groups with respect to
traumatic fear indicators.
232A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235
The subgenual ACC is mainly concerned with emotions,
in particular, the representation of autonomic afferences.
The dorsal region posterior to the genu of the corpus
posterior midcingulate cortex (aMCC, pMCC). These
are overlapping pain and fear sites. The aMCC is
innervated by the midline and intralaminar thalamic
nuclei belonging to the medial pain system, and also
receives direct input from the amygdala. Thus, involve-
observed ACC activation in our study was located in the
aMCC. In the context of our study, we interpret this
finding as a neural signature of pain and fear associated
with attachment trauma. This pattern is consistent with
our hypothesis and reports that abandonment fears
are the most persistent long-term symptoms in BPD
(Zanarini et al., 2003).
Our results are consistent with a FDG-PET study
demonstrating increased baseline ACC metabolism in
BPD patients (extending from aMCC into the medial
prefrontal cortex) as compared to healthy controls
(Juengling et al., 2003). However, they are not in
accordance with findings from two recent functional PET
studies (Schmahl et al., 2003, 2004). Women with BPD
and a history of sexual abuse showed significantly less
aMCC activation compared to women with sexual abuse
without BPD. Nevertheless, there is a crucial difference to
our study: subjects in the PET studies listened to pre-
processed, scripted memories reintroduced during their
neuroimaging experiment. The subjects in our study were
instructed to respond spontaneously to attachment-activat-
ing pictures, which prevented anticipatory self-regulation.
found an interaction of increased pain-induced response
in DLPFC and deactivation in the periguneal, ventral part
of the ACC and the amygdala(Schmahletal., 2006). The
authors interpret this pattern as an indicator of successful
antinociception that patients have acquired by their
experience of repetitive self-mutilation. We interpret our
finding of clearly more dorsal aMCC activation as an
indicator of unsuccessful coping with emotional pain.
experiences of sexual abuse compared to non-sexually
abused controls. It may be that severe trauma (though not
fulfilling PTSD criteria), more than a diagnosis of BPD
per se, is associated with increased aMCC activation
during emotional processing.
Furthermore, our specific stimuli indicating alone-
ness did not activate the amygdala compared to studies
using more general emotional or psychophysical stimuli
(Herpertz et al., 2001; Donegan et al., 2003; Schmahl
et al., 2006).
4.2.2. Dyadic pictures: superior temporal sulcus and
There were no specific hypotheses with respect to the
group differences that need to be explained. The STS is
Frith, 2003). It is a crucial part of a network involved in
“thinking about others” (Saxe and Kanwisher, 2003).
Attachment researchers suggest that abusive childhood
experiences of BPD patients lead to the inhibition of
constructive“mentalizing” capacities used to reflect upon
self and others. BPD patients show distorted, blocked or
“hyper-analytical” thinking processes when asked to
describe attachment experiences (Fonagy et al., 2003).
They often demonstrate a misleading hypersensitivity to
others' mental states that facilitates manipulating and
controlling perceived threatening relationships. Based on
this model, we interpret the increased STS activation in
BPD patients as a neural indicator of fear-based
hypervigilance in attachment relationships.
A second finding was the decreased activation of the
parahippocampal gyrus in BPD patients compared to
controls. Along with the hippocampus, this region is in-
volved in memory processes (Eichenbaum, 2000).
Recently, we have shown that this region is associated
with a “subsequent memory effect” for neutral items that
are encoded in a positive emotional context in healthy
subjects (Erk et al., 2003). Reduced parahippocampal
activation in BPD patients, thus, may be explained by
reduced positive valence of memories of dyadic interac-
tions. This interpretation is consistent with the finding that
both resolved and unresolved controls in our study re-
ported greater positive interactions in the dyadic narratives
(i.e., warmth and mutuality) than the BPD patients.
4.3. Limitations and conclusions
Several limitations of our study must be considered.
First, speaking within the scanner is associated with
movement; however, the amount of movement was
comparable to other studies without speaking (Kircher
et al., 2001). We used several measures to account for
residual movement in our model, such as, including
movement parameters as a covariate of no interest as
well as modeling the onset of every spoken word.
Furthermore, the activated regions were not those
typically found for movement artifacts. Second, our
sample size was not large enough to fill all four cells of
the design. The control analyses showed that medication
status and including the two resolved patients did not
change our main results. Nonetheless, our findings must
be examined using larger samples. Third, the influence
233 A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235
of lifetime psychiatric conditions in the patient group
cannot be ruled out, although patients with current
psychosis and substance abuse were excluded.
In conclusion, our behavioral results confirm that
BPD pathology is associated with traumatic attachment
fear related to autobiographic abuse and loss experi-
ences. The monadic attachment pictures representing
aloneness showed narrative and neural patterns of dif-
ferentiation between patients and controls. The fact that
such a differentiation was found only between (unre-
solved) patients and (resolved and unresolved) controls
can be interpreted such as that attachment disorganiza-
tion and disease specific factors have additive effects.
Our findings may provide evidence for possible mech-
anisms related to the fearful intolerance of aloneness in
BPD patients (Gunderson, 1996). The dyadic pictures,
representing the quality of potential attachment interac-
This finding highlights borderline patients' hypersensi-
tive attention to the social environment (Fonagy et al.,
2000) and addresses their poor contextualization of
positive relationship memories (Levy et al., 2006).
Our results suggest evidence for potential neural
mechanisms of attachment trauma underlying interper-
sonal symptoms of BPD. Moreover the findings indicate
that we have developed a sensitive procedure capable to
eliciting differences in patterns of brain activity between
controls and individuals with BPD.
We thank Edgar Schilly and Marco Jahn, Department
of Psychiatry, University of Ulm, for technical assis-
tance in fMRI data analysis, Kathrin Brändle, Depart-
ment of Psychiatry and Department of Diagnostic
Radiology, University of Ulm for assistance in fMRI
acquisition, Dipl.-Psych. Claudia Simons and Justice
Krampen, Department of Psychosomatic Medicine and
Psychotherapy, University of Ulm for translating the
German transcripts into English. We thank Dipl.-Psych
Dagmar Pape, Munich, a certified Adult Attachment
Interviewer in the hospital and PD Dr. Fabienne Becker-
Stoll, Munich, a certified judge, for classifying the Adult
Attachment Interviews. We thank David George and Dr.
Malcolm West for their contributions.
Agrawal, H.R., Gunderson, J., Holmes, B.M., Lyons-Ruth, K., 2004.
Attachment studies with Borderline Patients. A review. Harvard
Review of Psychiatry 12, 94–104.
Barratt, E.S., 1985. Impulsiveness subtraits: arousal and information
processing. In: Spence, J.T., Izard, C.E. (Eds.), Motivation,
Emotion and Personality. Elsevier, North-Holland, pp. 137–146.
Bartel, A., Zeki, S., 2004. The neural correlates of maternal and
romantic love. NeuroImage 21, 1155–1166.
2004. The development of mother–infant interactions after neonatal
amygdala lesions in rhesus monkeys. The Journal of Neuroscience
Beblo, T., Driessen, M., Mertens, M., Wingenfeld, K., Piefke, M.,
Woermann, F.G., 2006. Functional MRI correlates of the recall of
Medicine 36, 845–856.
Bernstein, E.M., Putnam, F.W., 1986. Development, reliability, and
validity of a dissociation scale. Journal of Nervous and Mental
Disease 174, 727–735.
Bowlby, J., 1969. Attachment and Loss.Vol.1: Attachment. Basic
Books, New York.
Buchheim, A., George, C., in press. The representational function of
attachment disorganization in borderline personality disorder and
anxiety disorder. In: Solomon, J., George, C. (Eds.), Disorganiza-
tion of Attachment and Caregiving. The Guilford Press, New York.
Buchheim, A., Erk, S., George, C., Kächele, H., Ruchsow, M., Spitzer,
M., Kircher, T., Walter, H., 2006. Measuring attachment represen-
tation in an fMRI environment: a pilot study. Psychopathology 39,
Donegan, N.H., Sanislow, C.A., Blumberg, H.P., Fulbright, R.K.,
Wexler, B.E., 2003. Amygdala hyperreactivity in borderline
personality disorder: implications for emotional dysregulation.
Biological Psychiatry 54, 1284–1293.
Driessen, M., Herrmann, J., Stahl, K., Zwaan, M., Meier, S., Hill, A.,
Osterheider, M., Petersen, D., 2000. Magnetic resonance imaging
volumes of the hippocampus and the amygdala in women with
borderline personality disorder and early traumatization. Archives
of General Psychiatry 57, 1115–1122.
Duvernoy, H.M., 1999. The human brain. Springer, New York.
Ebner-Priemer, U.W., Badeck, S., Beckmann, C., Wagner, A., Feige, B.,
Weiss, I., Lieb, K., Bohus, M., 2005. Affective dysregulation and
dissociative experience in female patients with borderline person-
ality disorder: a startle response study. Journal of Psychiatric
Research 39, 85–92.
Eichenbaum, H., 2000. A cortical-hippocampal system for declarative
memory. Nature Neuroscience 1, 41–50.
Eisenberger, N., Lieberman, M.D., Williams, K.D., 2003. Does
rejection hurt? An fMRI study of social exclusion. Science 302,
Erk, S., Kiefer, M., Grothe, J., Arthur, P., Wunderlich, A., Spitzer, M.,
Walter, H., 2003. Emotional context modulates subsequent
memory effect. NeuroImage 18, 430–447.
Fonagy, P., Target, M., Gergely, G., 2000. Attachment and borderline
personality disorder: a theory and some evidence. The Psychiatric
Clinics of North America 23, 103–122.
Fonagy, P., Gergely, G., Jurist, E.L., Target, M., 2003. Affect
regulation, mentalization, and the development of the self. Other
Press, New York.
Freyberger, H.J., Spitzer, C., Stieglitz, R.D., 1999. Fragebogen zu
Dissoziativen Symptomen. Huber, Bern, Switzerland.
Gabbard, G., 2005. Mind, brain, and personality disorders. The
American Journal of Psychiatry 162, 648–655.
234 A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235
Gallagher, H.L., Frith, C.D., 2003. Functional imaging of “theory of
mind”. Trends in Cognitive Sciences 7, 77–83.
of a new measure of adult attachment: the Adult Attachment
Projective. Attachment and Human Development 3, 30–61.
George, C., West, M., 2003. The Adult Attachment Projective:
measuring individual differences in attachment security using
projective methodology. In: Hilsenroth, M. (Ed.), Comprehensive
Handbook of Psychological Assessment. Personality Assessment,
vol. 2. Wiley and Sons, New Jersey, pp. 431–448.
George, C., West, M., 2004. Traumatic dysregulation coding.
Unpublished coding instructions. Mills College, Oakland, CA.
George, C., Kaplan, N., Main, M., 1985/1996. The Adult Attachment
of Psychiatry 153, 752–758.
Gündel, H., O'Connor, M.F., Littrell, L., Fort, C., Lane, R., 2003.
Functional neuroanatomy of grief: an fMRI study. The American
Journal of Psychiatry 160, 1946–1953.
Herpertz, S.C., Kunert, H.J., Schwenger, U.B., Sass, H., 1999.
Affective responsiveness in borderline personality disorder: a
psychophysiological approach. The American Journal of Psychiatry
Herpertz, S., Dietrich, T.M., Wenning, B., Krings, T., Erbereich, S.G.,
Willmes, K., Thron, A., Sass, H., 2001. Evidence of abnormal
amygdala functioning in borderline personality disorder: a
functional MRI study. Biological Psychiatry 50, 292–298.
Insel, T.R., 1997. A neurobiological basis of social attachment. The
American Journal of Psychiatry 154, 726–735.
Irle, E., Lange, C., Sachsse, U., 2005. Reduced size and abnormal
asymmetry of parietal cortex in women with borderline personality
disorder. Biological Psychiatry 57, 173–182.
Juengling, F., Schmahl, C., Heβlinger, B., Ebert, D., Bremner, J.D.,
Gostomzyk, J., Bohus, M., Lieb, K., 2003. Positron emission
tomography in female patients with borderline personality
disorder. Journal of Psychiatric Research 37, 109–115.
McGuire, P.K., 2001.Neural correlates offormal thoughtdisorder in
schizophrenia: preliminary findings from a functional magnetic
resonance imaging study. Archives of General Psychiatry 58,
Leibenluft, E., Gobbini, M.I., Harrison, T., Haxby, J.V., 2004.
Mothers' neural activation in response to pictures of their children
and other children. Biological Psychiatry 15, 225–232.
in the TFP treatmentof borderlinepatientswith transferencefocused
psychotherapy. Journal of Clinical Psychology 62, 481–501.
Lieb, K., Zanarini, M.C., Schmahl, C., Linehan, M.M., Bohus, M.,
2004. Borderline personality disorder. The Lancet 364, 453–461.
Lyons-Ruth, K., Yellin, C., Melnick, S., Atwood, G., 2003. Childhood
experiences of trauma and loss have different relations to maternal
unresolved and Hostile–Helpless states of mind on the Adult
Attachment Interview. Attachment and Human Development 5,
Main, M., Goldwyn, R., 1985/2005. Adult Attachment Classification
System. Unpublished Manuscript. University of California,
Department of Psychology, Berkeley.
Main, M., Kaplan, N., Cassidy, J., 1985. Security in infancy,
childhood, and adulthood: A move to the level of representation.
In: Bretherton, I., Waters, E. (Eds.), Growing points of attachment
theory and research. Monographs of the Society for Research in
Child Development, 50, pp. 66–104 (1–2, Serial Number 209).
Paris, J., 1994. Psychological risk factors for BPD in female patients.
Comprehensive Psychiatry 35, 301–305.
Saxe, R., Kanwisher, N., 2003. People thinking about thinking people.
The role of the temporo-parietal junction in “theory of mind”.
NeuroImage 19, 1835–1842.
Schmahl, C.G., Elzinga, B.M., Vermetten, E., Sanislow, C., McGlashan,
T.H., Bremner, D.J., 2003. Neural correlates of memories of
abandonment in woman with and without borderline personality
disorder. Biological Psychiatry 54, 142–151.
Schmahl, C.G., Vermetten, E., Elzinga, B.M., Bremner, D.J., 2004. A
positron emission tomography study of memories of childhood
abuse in borderline personality disorder. Biological Psychiatry 55,
Schmahl, C.G., Bohus, M., Esposito, F., Treede, R.D., Salle, F.,
Greffrath, W., Ludaecher, P., Jochims, A., Lieb, K., Scheffler, K.,
Henning, J., Seifritz, E., 2006. Neural correlates of antinociception
in Borderline Personality Disorder. Archives of General Psychiatry
Schnell, K., Dietrich, T., Schnitker, R., Daumann, J., Herpertz, S.C.,
2007. Processing of autobiographical memory retrieval cues in
borderline personality disorder. Journal of Affective Disorders 97,
Schnitzler, A., Ploner, M., 2000. Neurophysiology and functional
neuroanatomy of pain perception. Journal of Clinical Neurophys-
iology 17, 592–603.
Talairach, J., Tournoux, P., 1988. Co-Planar Stereotactic Atlas of the
human brain. Thieme, Stuttgart.
Tebartz van Elst, L., Thiel, T., Hesslinger, B., Lieb, K., Bohus, M.,
Henning, J., 2001. Subtle prefrontal neuropathology in a pilot
magnetic resonance spectroscopy study in patients with borderline
personality disorder. The Journal of Neuropsychiatry and Clinical
Neurosciences 13, 511–514.
Tebartz van Elst, L., Hesslinger, B., Thiel, T., Geiger, M., Haegele, K.,
Lemieux, L., Lieb, K., Bohus, M., Henning, J., Ebert, D., 2003.
Frontolimbic brain abnormalities in patients with borderline
personality disorder: a volumetric magnetic resonance imaging
study. Biological Psychiatry 54, 163–171.
Westen, D., Nakash, O., Thomas, C., Bradley, R., 2006. Clinical
assessment of attachment patterns and personality disorder in
adolescents and adults. Journal of Consulting and Clinical
Psychology 74, 1065–1085.
Wignall, E.L., Dickson, J.M., Vaugh, P., Farrow, T., Wilkinson, I.D.,
Hunter, M.D., Woodruff, P., 2004. Smaller hippocampal volume in
patients with recent-onset posttraumatic stress disorder. Biological
Psychiatry 56, 832–836.
Vogt, B.A., 2005. Pain and emotion interactions in subregions of the
cingulate gyrus. Nature 6, 533–544.
Zanarini, M.C., 2000. Childhood experiences associated with the
development of borderline personality disorder. The Psychiatric
Clinics of North America 23, 89–101.
Zanarini, M.C., Frankenburg, F.R., Hennen, J., Silk, K.R., 2003. The
longitudinal course of borderline psychopathology: 6 year pro-
spective follow up of the phenomenology of Borderline Person-
ality Disorder. The American Journal of Psychiatry 160, 274–283.
2006. Prediction of the 10-year course of borderline personality
disorder. The American Journal of Psychiatry 163, 827–832.
235 A. Buchheim et al. / Psychiatry Research: Neuroimaging 163 (2008) 223–235