Exaggerated Affect-Modulated Startle During
Unpleasant Stimuli in Borderline Personality Disorder
Erin A. Hazlett, Lisa J. Speiser, Marianne Goodman, Marcela Roy, Michael Carrizal, Jonathan K. Wynn,
William C. Williams, Michelle Romero, Michael J. Minzenberg, Larry J. Siever, and Antonia S. New
Background: Excessive emotional responding is considered to be a hallmark of borderline personality disorder (BPD). The affect-modu-
BPD patients (n ? 27) and healthy control subjects (n ? 21).
instructed to think about the meaning of the word for them personally while eyeblink responses were assessed.
affect. This finding remained significant when we controlled for comorbid diagnoses, including generalized anxiety disorder and post-
traumatic stress disorder. Greater symptom severity was associated with greater affective-startle difference scores (unpleasant-neutral).
emotional stimuli by BPD patients.
tion, startle eyeblink
(frequently called “emotion dysregulation”), which a recent
study indicates is the most prevalent and enduring diagnostic
criteria over time (McGlashan et al. 2005). Emotion dysregulation
is a combination of an inability to regulate emotional responses
(Gunderson and Zanarini 1989; Linehan 1993, 1995) with a high
sensitivity to emotional stimuli and unusually strong and long-
lasting reactions (Donegan et al. 2003; Linehan 1993). To date,
there are only a handful of studies examining emotion dysregu-
lation in BPD, and the majority were exclusively based on the
patients’ self-report of affect experience.
The affective startle modulation paradigm is unique in that it
requires minimum cooperation from the participant and can
yield a nonverbal, objective measure of emotional valence (Lang
et al. 1990). The startle reflex consists of a set of involuntary
responses to a sudden, strong sensory stimulus, for example, a
brief noise burst. The amplitude of the eyeblink component of
the human startle reflex is highly modifiable by environmental
stimuli that precede the reflex-eliciting stimulus. The startle-
evoking stimulus can be presented at any given time during the
presentation of a salient emotional stimulus, functioning as a
probe of changes in emotions and thus providing an index of
core feature of borderline personality disorder (BPD) is
affective dysregulation (Gunderson and Phillips 1991;
Linehan 1993; Siever et al. 2002; Skodol et al. 2002a)
ongoing affective information processing. As might be antici-
pated, phobic subjects show greater than normal startle potenti-
ation when they view pictures of a phobic object (Hamm et al.
1997). In contrast, incarcerated psychopaths—particularly those
characterized by the interpersonal-emotional facet of psychopa-
thy (“fearless dominance”)—show diminished startle probe po-
tentiation when viewing unpleasant picture stimuli (Benning et
al. 2005; Patrick et al. 1993). Affect-modulated startle has been
used to investigate emotional reactivity in patients with anxiety
disorders (Grillon 2002) and depression (Kaviani et al. 2004).
However, only one study has used the affect-modulated startle
paradigm with International Affective Picture System (IAPS)
stimuli and reported no differences between healthy control
subjects and BPD patients in affective startle and self-report
measures of emotion (Herpertz et al. 1999).
The aim of this study was to extend previous work by
examining affect-modulated startle in BPD patients and healthy
control subjects with a set of unpleasant stimuli that involve
themes prevalent in BPD pathology (e.g., abandonment) as well
as neutral stimuli. We used word stimuli previously validated in
a behavioral study of directed forgetting (Korfine and Hooley
2000). Given that affective dysregulation in BPD is thought to be
characterized by enhanced reactions to emotional stimuli
(Donegan et al. 2003; Linehan 1993), we thought that the
BPD-salient unpleasant words used by Korfine and Hooley
(2000) might be particularly well-suited to detect normal-BPD
group differences in affective startle. We hypothesized that,
consistent with this idea, BPD patients, compared with healthy
control subjects, would show exaggerated startle eyeblink (SEB)
magnitude during the unpleasant words compared with neutral
words. For exploratory purposes, we conducted a correlational
analysis to investigate whether greater affective startle is associ-
ated with greater BPD symptom severity.
Methods and Materials
We studied 27 patients who met DSM-IV criteria (American
Psychiatric Association 1994) for BPD and 21 healthy control
participants. The groups did not differ in age or gender (Table 1).
Participants gave written informed consent for this study, which
From the Department of Psychiatry (EAH, LJS, MR, MC, MJM, LJS, ASN),
Mount Sinai School of Medicine; Bronx Veterans Affairs Medical Center
(MG, MR, JMJ, LJS, ASN), New York, New York; Department of Psychiatry
and Biobehavioral Sciences (JKW), University of California, Los Angeles
and Mental Illness Research, Education and Clinical Center, VA Greater
ment of Psychology (WCW), Eastern Washington University, Cheney,
Address reprint requests to Erin A. Hazlett, Ph.D., Mount Sinai School of
Medicine, Department of Psychiatry, Box 1505, New York, NY 10029;
Received March 2, 2006; revised August 10, 2006; revised October 5, 2006;
accepted October 5, 2006.
BIOL PSYCHIATRY 2007;62:250–255
© 2007 Society of Biological Psychiatry
was approved by the institutional review board, and were paid
for their time.
Participants were rigorously screened and received a struc-
tured interview as described in our other studies (Hazlett et al.
2005; New et al. 2002). All participants received an interview
with a psychologist using the Structured Clinical Interview for
DSM-IV Axis I disorders (First et al. 1996) and the Structured
Interview for DSM-IV Personality Disorders (SIDP-IV) (Pfohl
1996). Healthy control subjects with an Axis I or II diagnosis in
either themselves or a first-degree relative were excluded.
Exclusion criteria for all participants included: severe medical
illness, neurological illness, head injury, history of alcohol/
drug dependence, substance abuse in the past 6 months, and
use of any psychoactive medications at the time of the study.
The patients were free of psychiatric medications for at least 2
weeks before study. Patients with a history of schizophrenia,
psychotic disorder, bipolar type I affective disorder, or current
(in the last 6 months) major depressive disorder (MDD) were
excluded. Thirteen of the patients had a past history of MDD,
5 had PTSD (2 current and 3 past), and 6 had generalized
anxiety disorder. The mean total number of DSM-IV BPD
criteria met was 6.7 ? 1.2 (of the 9 criteria). All of the patients
met DSM-IV criteria of affective instability (item 6), and 22 met
criteria of impulsive behavior (item 4 and/or 5). In a finding
consistent with reports from other groups (Skodol et al.
2002a), patients who met criteria for BPD also met criteria for
other Axis II diagnoses, including schizotypal personality
disorder (n ? 7; 26%) and antisocial personality disorder (n ?
8; 30%). The mean number of non-BPD personality disorder
diagnoses the patient group met was 2.2 ? 1.7. One patient
plete the Barratt Impulsivity Scale, and another did not complete
any of the scales.
The BPD patients were primarily a community sample re-
cruited for the study through advertisement in local newspapers
(90%). A small percentage of the patients were recruited through
referrals from outpatient psychiatric clinics (10%) at the Bronx
Veterans Affairs Medical Center (Bronx, New York) and Mount
Sinai School of Medicine (New York, New York). Although our
BPD patients were recruited from the community, most of them
received some previous type of treatment (1 previously received
antipsychotic drugs, 10 previously received selective serotonin
reuptake inhibitors, 9 previously received counseling/psycho-
therapy, 7 were never previously treated).
Participants completed psychometric self-report measures of
aggression (Buss-Durkee Hostility Inventory [BDHI]; Buss and
Durkee 1957), impulsivity (Barratt Impulsivity Scale [BIS]; Barratt
1965), affective lability (Affective Lability Scale [ALS]; Harvey et
al. 1989), and affective intensity (Affective Intensity Measure
[AIM]; Larsen and Diener 1985).
Three patients and two control subjects were eliminated
before any statistical analysis, because they were considered to
be eyeblink nonresponders. We used a criteria similar to our
other work (Hazlett et al. 2001; 2003), in which nonresponders
failed to exhibit blinks on over 25% of the baseline startle
Participants were seated in a sound-attenuated testing room
ideally suited for psychophysiological recording. The experi-
menter and all necessary psychophysiological equipment were
located just outside the room. A low-light camera was used for
continuous monitoring of the participant during the procedure.
Standard instructions describing the procedure were read to the
participant, and electrodes were attached for the recording of
SEB (Hazlett et al. 2003). Before beginning the task, two exam-
ples of the startle-eliciting stimulus were presented in order to
familiarize the participant with the stimulus and ensure adequate
elicitation of SEB. Participants were instructed that their primary
task was to read each word silently (Figure 1) when it appeared
on the computer screen and to think about the meaning of the
word for them personally the entire time it appeared. They
were also instructed that sometimes during the viewing of the
word and/or between words they would hear a brief burst of
static noise through the headphones. During the inter-trial
interval, the screen was blank, except that 3 sec before word
onset a “?” sign appeared on the screen to remind partici-
pants to keep their gaze on the screen and prepare for the next
word. Participants were told that during this time they could
stop thinking about the meaning of the previous word. Lastly,
participants were told that at the end of the session they would
be asked to recall as many words as they could remember but
that this was not their primary task. After task completion,
participants recalled aloud as many words as they could.
Participants rated how each of the 16 unpleasant and 16
neutral words made them feel when thinking about the
meaning of the word for them personally with the 9-point
self-assessment manikin scale (SAM) (Lang et al. 1997) for
valence (1 ? very pleasant to 9 ? very unpleasant, 5 ?
neutral) and arousal (1 ? not arousing to 9 ? very arousing,
5 ? neutral; note that the original SAM arousal scale has 9 ?
not arousing, but for the purposes of graphing the two SAM
scales in a consistent manner, we inverted this scale).
Table 1. Demographics, Baseline Startle Response, and Self-Report Ratings in Healthy Control Subjects and BPD Patients
Healthy Control SubjectsBorderline PD Patients
Statistic, p Valuen Mean SDn Mean SD
Baseline Startle (?V)
Affective Intensity Measure
Affective Lability Scale
Barratt Impulsivity Scale
Buss-Durkee Hostility Inventory
?2? 1.01, df ? 1, p ? .32
t ? 1.10, df ? 46, p ? .28
t ? .77, df ? 46, p ? .44
t ? 2.05, df ? 45, p ? .046
t ? 6.87, df ? 45, p ? .0000
t ? 1.89, df ? 44, p ? .07
t ? 8.69, df ? 45, p ? .0000
aSignificant between-group difference.
E.A. Hazlett et al.
BIOL PSYCHIATRY 2007;62:250–255 251
Our word list was taken from a published behavioral study
and consisted of 32 words (16 unpleasant, 16 neutral; see Korfine
and Hooley 2000 and Figure 1 for details). Each word was
presented for 6 sec on a computer screen in a mixed, fixed
pseudorandom order. The inter-trial interval between words was
21–31 sec (mean ? 26 sec). Of the 32 word trials, 24 included a
presentation of the acoustic startle stimulus at one of two
long-lead intervals: 4000 or 5000 msec. There were 12 presenta-
tions (6 during neutral and 6 during unpleasant words) for each
lead interval. The remaining eight words were presented without
a startle stimulus. There were also 12 presentations of the startle
stimulus during the inter-trial interval that serve as our baseline
measure of startle magnitude.
Psychophysiological Recording and Scoring of SEB
The startle-eliciting stimulus was a 104-dB sound pressure
level (SPL) (A) white noise burst of 50 msec in duration with a
near-instantaneous (? 1 msec) rise/fall time. Startle eyeblink was
measured by recording electromyographic (EMG) activity with a
pair of miniature silver-silver chloride electrodes (3 mm in
diameter) filled with standard electrolyte gel and positioned over
the orbicularis oculi muscle just below and slightly lateral to the
participant’s lower right eyelid. Eyeblink responses were re-
corded as EMG activity with a Contact Precision Instruments
system (London, England) that interfaces with a Pentium 4
computer. The EMG activity was filtered with a 30-Hz high-pass
filter and a 500-Hz low-pass filter, along with a 60-Hz notch filter.
The data were collected as raw EMG at 1000 Hz, beginning 500
msec before startle stimulus onset to 300 msec after startle
stimulus onset. Integrated EMG was used to score eyeblink
amplitude with a time constant of 20 msec.
Startle eyeblink amplitude was scored off line with our
standard methods described elsewhere (Hazlett et al. 2003) with
the raw, rectified EMG signal. Pre-startle EMG activity was
defined as occurring 100 msec before startle stimulus onset. Peak
startle activity was defined as a 20-msec average of the peak
activity (10 msec before peak, 10 msec after peak) occurring
within a window of 20–120 msec after startle stimulus onset.
Startle eyeblink amplitude was the difference between the peak
startle activity and the pre-startle EMG activity in microvolts.
Startle eyeblink modification scores were then calculated as
percent change scores: [(startle during word stimulus ? startle
alone)/(startle alone)] ? 100. Thus, a positive change score
indicates startle potentiation relative to baseline eyeblink magni-
tude. Percent change scores are preferred, because microvolt
units are correlated with baseline eyeblink magnitude whereas
percent change scores are not, removing any dependence on
baseline eyeblink amplitude (Blumenthal et al. 2004, 2005).
Statistical analyses were performed with Statistica (StatSoft
2003). We used a mixed model, repeated-measures analysis of
variance (ANOVA) to examine the SEB data with a 2 ? 2 ? 2
design. Diagnostic group (healthy control subjects vs. BPD
patients) was the between-subjects variable. Word type (neutral,
unpleasant) and startle probe position (4000?msec, 5000?msec)
were within-subject variables.
Similar ANOVAs (Group ? Word type) were done to examine
self-report ratings for the words and word recall. The dependent
variables were self-assessment manikin (SAM) ratings for valence
and arousal and percent score for unpleasant and neutral words
Significant interactions involving the group factor were fol-
lowed by planned comparisons to test our a priori hypotheses.
Rom’s procedure (Rom 1990) was used for all groups of fol-
low-up t tests to control for experiment-wise Type I error.
Affective Modulation of Startle
Compared with the healthy control subjects, the BPD patients
showed greater startle potentiation during the unpleasant words
but not during the neutral words [Group ? Word type interac-
tion, F (1,46) ? 6.91, p ? .01, Figure 2A]. Follow-up t tests
confirmed that the patients showed significantly greater mean
startle potentiation during unpleasant words (averaged across
the two probe positions) than the control subjects [t(46) ? 2.57,
p ? .01] but did not differ from normal during neutral words
(p ? .47).
The Group ? Word ? Startle probe position interaction was
not significant (p ? .52). There was a nonsignificant trend for a
group main effect, which indicated that the BPD patients showed
greater overall startle potentiation (averaged across word type
and probe position) than the healthy control subjects [F(1,46) ?
3.94, p ? .053]. However, as can be seen in Figure 2A, this
difference was due to a group difference in startle potentiation
during unpleasant words. The groups did not differ in baseline
startle amplitude during the inter-trial interval (Table 1).
It should be noted that a Kolmogorov-Smirnov test for
normality indicated that the percent change SEB modification
scores showed a normal distribution (all p ? .20) In contrast, the
absolute microvolt SEB values were significantly skewed (all p ?
.01). The ANOVAs on the absolute value data (square root
transformed microvolt values) produced results that were nearly
identical to the percent change data.
Diagnostic Comorbidity in BPD
We conducted parallel analyses in order to assess whether our
main finding of exaggerated affective startle during unpleasant
words could be explained by a comorbid diagnosis of antisocial
Figure 1. Schematic of a portion of the experimental session. A startle
eyeblink (SEB) eliciting probe (104 dB, 50-msec white noise burst) was
presented either 4 or 5 sec after the onset of pre-selected unpleasant and
neutral words. Occasionally the startle probe was presented alone during
the inter-trial interval to measure baseline startle magnitude. The unpleas-
of abandonment, rejection, anger and rage, self-harm, and others being
ant and neutral words were matched on frequency of usage in the English
language and on string length (word list taken from Korfine and Hooley:
angry, useless, despise, misunderstood, enraged, reject, cruel, uncaring;
Neutral: collect, regular, actually, improvise, skeptical, daring, persuade,
moderate, predict, consider, coin, deliberate, talkative, listen, satisfy, view).
252 BIOL PSYCHIATRY 2007;62:250–255
E.A. Hazlett et al.
personality disorder, generalized anxiety disorder, and/or post-
traumatic stress disorder (PTSD). For each of these comorbid
diagnoses, we removed the subgroup of comorbid patients and
repeated the ANOVA described previously. In each case, the
pattern of startle results was nearly identical to those observed
with the full sample of patients and the Group ? Word type
interaction remained significant (all p ? .03).
Self-Report Measures for Unpleasant and Neutral Words
Valence. Both groups rated the unpleasant words as more
unpleasant than neutral words. However, compared with the
healthy control subjects, the BPD patients rated the unpleasant
words as significantly less unpleasant, but group ratings for the
neutral words were similar [significant Group ? Word type
interaction, F (1,46) ? 5.49, p ? .02, Figure 2B]. Follow-up t tests
confirmed that the patients rated the unpleasant words as less
unpleasant [t(46) ? 2.85, p ? .01] but did not differ from normal
on neutral word ratings (p ? .81).
There was a significant group main effect, which indicated
that the BPD patients had a lower overall self-report rating
(averaged across neutral and unpleasant words) compared with
the healthy control subjects [F(1,46) ? 5.20, p ? .03]. This
difference was primarily due to a group difference in the valence
rating of the unpleasant words (Figure 2B).
Arousal. Both groups rated the unpleasant words as more
arousing than the neutral words [main effect of word type,
F(1,46) ? 22.70, p ? .00002], and there were no group differ-
ences (Group ? Word type interaction, p ? .68, Control subjects:
unpleasant mean ? 6.55 ? 1.32, neutral mean ? 5.10 ? 0.93;
Patients: unpleasant mean ? 6.16 ? 1.39, neutral mean ? 4.94 ?
Self-Report Measures of Affective Lability, Impulsivity,
The BPD group had higher self-report scores compared with
the healthy control subjects for affective intensity, affective
lability, and aggression (Table 1). The two groups did not differ
on the self-report rating of impulsivity.
Recall of Words
The two groups did not differ on word recall for unpleasant or
neutral words (Group ? Word type interaction, p ? .43). Both
groups recalled more unpleasant than neutral words (Control
subjects: unpleasant mean ? 65.9 ? 13.2%, neutral mean ?
34.1 ? 13.1%; Patients: unpleasant mean ? 67.4 ? 19.3%, neutral
mean ? 28.8 ? 16.9%), and the main effect of word type was
significant [F(1,46) ? 66.77, p ? .0000; overall mean unpleas-
ant ? 66.6 ? 17.0%; neutral words ? 31.5 ? 15.2%].
Symptomatic Correlates of Affective Startle
We conducted Pearson correlations for the patient group
between three affective startle variables (SEB magnitude during
unpleasant and neutral words, SEB difference score [unpleasant-
neutral word condition] averaged across the two probe posi-
tions), and total number of DSM-IV BPD symptoms, ALS, and
AIM scores. Among the BPD patients, those with higher total
number of DSM-IV diagnostic criteria (each criterion was rated
either 0 ? absent or 1 ? present) showed larger SEB modifica-
tion difference scores (unpleasant-neutral condition), r ? .38,
p ? .049. This pattern indicates that greater symptom severity is
associated with larger SEB magnitude during the unpleasant
word condition relative to the neutral condition. None of the
other correlations reached significance.
To our knowledge, this is the first study to show affective
startle abnormalities in BPD patients. Our main findings are:
1) compared with healthy control subjects, the BPD patients
showed enhanced startle potentiation during the unpleasant
word condition but not during the neutral word condition or
baseline startle amplitude; this pattern of results remained statis-
tically significant when we controlled for comorbid diagnoses of
PTSD, generalized anxiety disorder, and antisocial personality
disorder; and 2) greater symptom severity, defined by total number
of DSM-IV BPD symptoms, was associated with greater affective-
startle difference (unpleasant-neutral) scores in patients.
The BPD-related affective startle abnormality suggests that the
patients showed excessive emotional responsivity specifically
Figure 2. (A) Means for SEB modification percent change scores (from
baseline) for the neutral and unpleasant word condition are shown for
healthy control subjects and BPD patients [significant Group ? Word type
ant words, t(46) ? 2.57, p ? .01]. The groups did not differ during neutral
and BPD patients. 1 ? very pleasant, 5 ? neutral, 9 ? very unpleasant. The
the healthy control subjects, but the two groups did not differ on their
p ? .02; *significant group difference for unpleasant words, p ? .01]. The
SEMs. Abbreviations as in Figure 1.
E.A. Hazlett et al.
BIOL PSYCHIATRY 2007;62:250–255 253
during the unpleasant words. This provides objectively measured
physiological evidence of the emotion dysregulation that is
observed clinically to be a core feature characterizing BPD
(Skodol et al. 2002a, 2002b). In further support of emotion
dysregulation and consistent with exaggerated startle during an
unpleasant stimulus: the patients in our study as well as other
studies (Levine et al. 1997) had higher-than-normal AIM scores,
which is a self-report measure thought to assess the intensity with
which feelings are experienced. Intensity on this scale is defined
as the strength of response, and items include manifestations of
reaction such as bodily responses and subjective experiences.
Our affective-startle findings are inconsistent with the only
other BPD study published to date, which showed a normal
pattern of startle modulation in response to unpleasant (IAPS)
pictures (Herpertz et al. 1999). Possible reasons for this discrep-
ancy include our use of different unpleasant stimuli. We chose
words shown to be particularly salient to BPD patients, whereas
Herpertz et al. (1999) employed more generally unpleasant
pictures. Perhaps during passive viewing of generally unpleasant
stimuli (e.g., picture of disfigured baby), BPD patients exhibit
normal startle potentiation. In contrast, stimuli related to BPD-
salient themes combined with the instruction to actively “think
about the meaning of the word for them personally” might be
more powerful triggers of affective dysfunction in BPD. In a
finding consistent with this concept, a study that included
BPD-specific words (negative self-descriptors) in the context of
an emotional Stroop task showed that BPD patients exhibited
interference caused by supraliminally (consciously processed)
presented emotional words (Arntz et al. 2000). Our sample
characteristics also differed from Herpertz et al. (e.g., we had a
community sample that included both genders, whereas Her-
pertz et al. examined female inpatients).
In contrast to the nonverbal, objective affective startle mea-
sure of emotional response, the SAM self-report scores indicated
that the BPD patients rated the unpleasant words as less emo-
tional or unpleasant than healthy control subjects. Thus, the
psychophysiological measure of valence indicated that the BPD
patients exhibited emotional hyper-responsivity, whereas the
self-report indicated emotional hypo-responsivity. Herpertz et al.
(1999) reported a dissociation between psychophysiological and
self-report measures of arousal in BPD patients—lower-than-
normal physiological arousal (skin conductance) during neutral
pictures yet higher-than-normal self-report of their arousal expe-
riences. Some investigators have suggested that psychopaths
learn over time to edit or tailor their verbal responses to conform
to social expectations (Herpertz et al. 2001; Patrick et al. 1993).
Borderline personality disorder patients might also learn over
time to tone down their verbal expressions of emotion in order
to “fit in better.” This idea is consistent with clinical research
(Rosenthal et al. 2005) that suggests that increased psychopa-
thology in BPD is associated with the tendency to chronically
and deliberately attempt to reduce the frequency or intensity of
One of the important issues our study and previous work
(Herpertz et al. 1999) reflects is the inadequacy of self-report
measures of emotional responsiveness. Our findings suggest
startle might be an important nonverbal adjunct to self-reports for
assessment of emotion regulation and treatment outcome in
BPD. As discussed by others (Grillon and Baas 2003), verbal
reports are frequently inaccurate and influenced by demand
characteristics, erroneous interpretations, and misperceptions. In
contrast, startle is a physiological response impervious to volun-
tary control yet sensitive to emotional processing.
One might expect that because the BPD patients showed
exaggerated startle during the unpleasant words, they might also
recall more unpleasant words than normal, but this was not the
case. Both the healthy control subjects and BPD patients recalled
unpleasant words significantly more than they recalled neutral
words. The lack of a Group ? Word type interaction for recall is
consistent with the findings of Korfine and Hooley (2000) for the
“remember” condition and suggests that despite showing greater-
than-normal affective processing, the BPD patients did not
display a memory bias for the unpleasant words.
Our results suggest that greater symptom severity in the
patient group, defined here by greater total number of DSM-IV
BPD criteria, is associated with greater affective startle (unpleas-
ant-neutral condition). Thus, affective instability and impulsivity
traits as defined by DSM-IV criteria seem to be related to the BPD
patients’ unusual pattern of exaggerated affective startle. This
coupling of biological measures of emotional responsiveness
and personality disorder traits is consistent with other work by
our group showing greater BPD traits related to affective insta-
bility are associated with greater mood responsiveness as in-
dexed by physostigmine response (Steinberg et al. 1997).
An important caveat is that our sample of BPD patients
included more men than women and therefore might be consid-
ered unrepresentative. Although DSM-IV reports that BPD is
diagnosed predominately in women (3:1 ratio of women to
men), recent studies suggest that the differential gender preva-
lence of BPD in clinical settings might largely be a function of
sampling and diagnostic biases of various types (Skodol and
Bender 2003). Studies employing semi-structured interviews
have yielded a more balanced gender distribution, and data from
the Collaborative Longitudinal Personality Disorders Study
(CLPS) suggests that symptoms of BPD are largely similar in men
and women (Johnson et al. 2003).
In conclusion, these findings are consistent with the concept
that BPD patients exhibit emotion dysregulation, which includes
an inability to regulate emotional responses (Gunderson and
Zanarini 1989; Linehan 1993, 1995) with a high sensitivity to
emotional stimuli and unusually strong reactions (Donegan et al.
2003; Linehan 1993). However, replication efforts will need to
address important issues like employing positively valenced
words matched on arousal level to the negatively valenced words
to determine whether the effects found here were due to the
negative valence of the stimuli or to a more general arousal
during any emotional stimuli. Yet, there were no group
differences on reported arousal in the current study. Examin-
ing whether BPD patients exhibit exaggerated affective startle
during generally unpleasant words versus BPD-salient words
would also be useful. A recent behavioral study found that
BPD patients showed enhanced recall of both generally
unpleasant as well as borderline-specific stimuli, despite the
instruction to forget it, suggesting enhanced processing of
negative information (Domes et al. 2006). Future studies
should examine gender differences and include non-BPD
psychiatric control subjects (e.g., schizotypal personality dis-
order) to determine specificity of these findings. Lastly, re-
search integrating psychophysiology and functional neuroim-
aging is needed to examine brain circuitry underlying these
BPD-related affective startle abnormalities.
This research was supported by a Grant from the Borderline
Personality Disorder Research Foundation to LJS, an Indepen-
dent Investigator Award from the National Alliance for Research
254 BIOL PSYCHIATRY 2007;62:250–255
E.A. Hazlett et al.
on Schizophrenia and Depression and MH073911 to EAH, and Download full-text
a VA MERIT Grant (9001-03-0051) to ASN.
American Psychiatric Association (1994): DSM-IV: Diagnostic and Statistical
A test with the emotional Stroop paradigm. J Personal Disord 14:366–
Barratt ES (1965): Factor analysis of some psychometric measures of impul-
siveness and anxiety. Psychol Rep 16:547–554.
lation, and electrodermal reactivity in twin men. Psychophysiology 42:
Blumenthal TD, Cuthbert BN, Filion DL, Hackley S, Lipp OV, van Boxtel A
myographic studies. Psychophysiology 42:1–15.
used to quantify prepulse inhibition of eyeblink responding. Psycho-
Buss AH, Durkee A (1957): An inventory for assessing different kinds of
Domes G, Winter B, Schnell K, Vohs K, Fast K, Herpertz SC (2006): The influ-
ence of emotions on inhibitory functioning in borderline personality
disorder. Psychol Med 36:1163–1172.
et al. (2003): Amygdala hyperreactivity in borderline personality disorder:
ing, context, and neurobiology. Biol Psychiatry 52:958–975.
Grillon C, Baas J (2003): A review of the modulation of the startle reflex by
affective states and its application in psychiatry. Clin Neurophysiol 114:
Gunderson JG, Phillips KA (1991): A current view of the interface between
borderline personality disorder and depression. Am J Psychiatry 148:
Gunderson JG, Zanarini MC (1989): Pathogenesis in borderline personality.
In: Tasman A, Hales RE, Frances AJ, editors. Review of Psychiatry, vol 8.
Washington, DC: American Psychiatric Press, 25–48.
lation fearful subjects. Psychophysiology 34:97–107.
Harvey PD, Greenberg BR, Serper MR (1989): The affective lability scales:
Development, reliability, and validity. J Clin Psychol 45:786–793.
Hazlett EA, Dawson ME, Schell AM, Nuechterlein KH (2001): Attentional
A startle modification analysis. Psychophysiology 38:669–677.
Hazlett EA, Levine J, Buchsbaum MS, Silverman JM, New A, Sevin EM, et al.
(2003): Deficient attentional modulation of the startle response in pa-
tients with schizotypal personality disorder. Am J Psychiatry 160:1621–
sonality disorder. Biol Psychiatry 58:614–623.
Herpertz SC, Kunert HJ, Schwenger UB, Sass H (1999): Affective responsive-
ness in borderline personality disorder: A psychophysiological ap-
Herpertz SC, Werth U, Lukas G, Qunaibi M, Schuerkens A, Kunert HJ, et al.
(2001): Emotion in criminal offenders with psychopathy and borderline
collaborative longitudinal personality disorders study. ComprPsychiatry
tive modulation of the startle response in depression: Influence of the se-
Korfine L, Hooley JM (2000): Directed forgetting of emotional stimuli in
borderline personality disorder. J Abnorm Psychol 109:214–221.
reflex. Psychol Rev 97:377–395.
Lang PJ, Bradley MM, Cuthbert BN (1997): International Affective Picture
System (IAPS): Technical manual and affective ratings.: NIMH Center for
the Study of Emotion and Attention.
Larsen RJ, Diener E (1985): A multitrait-multimethod examination of affect
Levine D, Marziali E, Hood J (1997): Emotion processing in borderline per-
ity Disorder. New York: The Guilford Press.
The Guilford Press.
McGlashanTH, Grilo CM, Sanislow CA,RalevskiE,MoreyLC,GundersonJG,
et al. (2005): Two-year prevalence and stability of individual DSM-IV
criteria for schizotypal, borderline, avoidant, and obsessive-compulsive
personality disorders: Toward a hybrid model of axis II disorders.
emission tomography response to meta-chlorophenylpiperazine in im-
Startle reflex modulation. J Abnorm Psychol 102:82–92.
Pfohl B (1996): Disease, disorder, syndrome, or symptom: Which shall we
Bonferroni inequality. Biometrika 77:663–665.
Rosenthal MZ, Cheavens JS, Lejuez CW, Lynch TR (2005): Thought suppres-
sion mediates the relationship between negative affect and borderline
personality disorder symptoms. Behav Res Ther 43:1173–1185.
Siever LJ, Torgersen S, Gunderson JG, Livesley WJ, Kendler KS (2002): The
Skodol AE, Bender DS (2003): Why are women diagnosed borderline more
than men? Psychiatr Q 74:349–360.
The borderline diagnosis I: Psychopathology, comorbidity, and person-
ality structure. Biol Psychiatry 51:936–950.
The borderline diagnosis II: Biology, genetics, and clinical course. Biol
StatSoft I (2003): Statistica, 6.0 ed. Available at: http://www.statsoft.com.
Steinberg BJ, Trestman R, Mitropoulou V, Serby M, Silverman J, Coccaro E, et
al. (1997): Depressive response to physostigmine challenge in border-
line personality disorder patients. Neuropsychopharmacology 17:264–
E.A. Hazlett et al.
BIOL PSYCHIATRY 2007;62:250–255 255