BEHAVIOUR

RESEARCH AND
THERAPY

PERGAMON

Behaviour Research and Therapy 37 (1999) 545557

Assumptions in borderline personality disorder: specicity,

stability and relationship with etiological factors
Arnoud Arntz *, Roos Dietzel, Laura Dreessen
Department of Medical, Clinical and Experimental Psychology, University of Maastricht, P.O. Box 616, 6200 MD
Maastricht, The Netherlands
Received 21 August 1998

Abstract
The specicity and stability of a set of assumptions hypothesized to be characteristic of Borderline
Personality Disorder (BPD) was investigated. BPD patients (n = 16) were compared to cluster-C
personality disorder patients (n = 12) and to normal controls (n = 15). All subjects were female and
diagnosed with SCID-I and -II. Subjects rated a short version of the Personality Disorder Beliefs
Questionnaire (PDBQ), with six sets of 20 assumptions each, hypothesized to be characteristic of
avoidant, dependent, obsessive-compulsive, paranoid, histrionic and borderline personality disorder. The
BPD assumptions (Cronbach alpha = 0.95) proved to be the most specic to BPD patients. Subjects
rated the shortened PDBQ again after viewing an emotional video fragment one week later. Despite
increased negative emotions, the PDBQ ratings remained relatively stable. Conrming the cognitive
hypothesis, regression analyses indicated that the BPD assumptions mediate the relationship between
self-reported etiological factors from childhood (sexual abuse and emotional/physical abuse) and BPD
pathology assessed with the SCID-II. It is suggested that a set of assumptions is characteristic of BPD,
and is relatively stable despite the instability of the behaviour of people diagnosed as having BPD.
# 1999 Elsevier Science Ltd. All rights reserved.

1. Introduction
In cognitive views of personality disorders a central role is given to schemas, generalized
representations of the self, others, and the world, of the relationships between these elements,
and of the main strategies for survival (Beck, Freeman, & Associates, 1990; Young, 1990).
These schemas are hypothesized to result from the interaction between dispositional and
environmental factors during childhood and, once formed, to strongly inuence informational
* Corresponding author. E-mail: arnoud.arntz@mp.unimaas.nl
0005-7967/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 5 - 7 9 6 7 ( 9 8 ) 0 0 1 5 2 - 1

546

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

input by selective and interpretative processes. Though the content of these schemas are
hypothesized to be mainly constituted by tacit knowledge, therefore unaccessible for direct
introspection, it is believed in cognitive views that important aspects of schemas can be
represented in so-called assumptions, verbal circumscriptions of fundamental beliefs, like ``I'm
inherently bad'' and ``If you do not dominate the other person, he/she will dominate you''. By
letting Ss rate the degree they believe in these propositions it is hypothesized that the content
of their schemas becomes measurable. Various suggestions have been made about the
assumptions that are central in personality disorders: Beck et al. (1990) oer an extensive list
of assumptions for most DSM-III-R personality disorders; Young (1990) has oered a
somewhat dierent view by suggesting that 18 themes like self-sacrice and entitlement are
fundamental dimensions in personality pathology (see Schmidt, Joiner, Young, & Telch, 1995,
for a psychometric evaluation of Young's Schema Questionnaire).
Despite the fact that the Borderline Personality Disorder (BPD) is one of the most severe
and common disorders in psychiatry, almost nothing is known about the assumptions that may
be central in this disorder. Beck et al. (1990), for instance, do not give a list of assumptions
characteristic of BPD, stating that BPD is an exception among the personality disorders by not
being characterized by a specic set of beliefs. Young (see Beck et al. (1990), p. 185) has
suggested that nine of his `early maladaptive schemas' are playing a role in BPD, but the
number is quite large to consider this as the essence of the core beliefs in BPD. Despite the
omission of a list of BPD assumptions in the Beck et al. (1990) book, the chapter on BPD
makes some specic suggestions: BPD individuals are believed to view themselves as inherently
bad, powerless and vulnerable, and to view the world (we would say others) as dangerous and
malignant. Arntz (1994) has elaborated on this suggestion by hypothesizing that these
fundamental beliefs are related to childhood etiological factors common in BPD: sexual abuse
by intimates in particular, but also emotional and physical abuse and neglect (Bryer, Nelson,
Miller, & Krol, 1987; Stone, Unwin, Beacham, & Swenson, 1988; Coons, Bowman, Pellow, &
Schneider, 1989; Herman, Perry, & van der Kolk, 1989; Ogata, Silk, Goodrich, Lohr, Westen,
& Hil, 1990; Weaver & Clum, 1993). In short, it is hypothesized that if members of the nuclear
family (or its substitute) in which the child grows up (sexually/physically/emotionally) abuse
the child, especially in the 612 yr range, the formation of the idea to be inherently bad, that
nobody can be trusted and that others will abuse, punish or leave you when you become
intimate and dependent, are fostered. It has also been hypothesized that BPD is characterized
by a cognitive-emotional developmental stagnation on areas related to these issues, resulting,
for instance, in dichotomous thinking (see Veen & Arntz, 1997, for a test of the schemaspecicity of dichotomous thinking in BPD).
The rst aim of the present study was to test whether a set of assumptions (hereafter to be
referred to as BPD assumptions) derived from the work by Beck et al. (1990) and Arntz (1994)
is characteristic of BPD. To test this issue a 3-group design was chosen: a group of BPD
patients was compared with a psychiatric control group and a healthy control group. The
DSM-III-R (and DSM-IV) distinguishes three clusters of PDs: the `odd' cluster-A (paranoid,
schizoid, and schizotypal PDs); the `dramatic' cluster-B (histrionic, narcissistic, borderline, and
anti-social PDs); and the `anxious' cluster-C (avoidant, dependent, obsessive-compulsive, and
passive-aggressive PDs). Since there is considerable comorbidity between BPD and cluster-A
PDs and other cluster-B PDs, it was decided in this stage of research to compare BPD with

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

547

cluster-C PDs. In this way clearly distinguishable groups could be compared. As a rst test of
the specicity hypothesis the present study compared BPD patients with cluster-C patients and
normals, with respect to assumptions hypothesized to be specic to BPD as well as with
respect to ve groups of assumptions hypothesized to be specic to other PDs (avoidant,
dependent, obsessive-compulsive, paranoid and histrionic). To avoid possible sex inuences and
because of the high proportion of women in BPD populations in psychiatric settings, only
women were studied. Given the high level of pathology and the high number of comorbidity of
personality disorders in BPD patients it was expected that BPD patients would believe to a
degree comparable to cluster-C patients in cluster-C beliefs, but would surpass these patients as
to cluster-A and -B beliefs. It was specically expected that the group of assumptions
hypothesized to be characteristic of BPD would discriminate BPD patients most clearly from
the other groups.
The second aim was to test the stability of the beliefs. BPD is said to be a disorder
characterized by strong uctuations, and it can therefore be doubted whether BPD
assumptions are stable. Perhaps the belief in various assumptions is as uctuating as the mood
and the behaviour of these patients. Instead of waiting for spontaneous uctuations, an
attempt was made to induce a negative emotional state and to prime BPD characteristic
schemas (Safran, Segal, Hill, & Whien, 1990) by letting the Ss view an emotional video
fragment foregoing the second test. Belief ratings of the second test were compaired with those
of the rst test.
Lastly, it was attempted to test whether the relationship between retrospective reports of
childhood traumas and BPD assumptions support the hypothesis that these assumptions may
originate from childhood traumas. More specically, it was tested whether the relationship
between trauma reports (in particular sexual abuse) and BPD as assessed with a semi
structured interview (SCID-II) was mediated by these assumptions.

2. Method
2.1. Subjects
Patients were recruited from the Community Mental Health Center Maastricht, from the
Psychiatric Hospital `Vijverdal' in Maastricht and from the Psychiatric Hospital `Johan Weijer'
in Amsterdam. Patients were screened with SCID interviews and only patients with either a
Borderline Personality Disorder or a cluster-C PD were invited to participate in the
investigation. There were 16 patients with BPD and 12 patients with a cluster-C PD. In both
cities normal controls (n = 15) were recruited by means of advertisements. Approximately half
of each group came from either city. For all subjects the following inclusion criteria were
employed: (i) age between 21 and 50; (ii) woman. Exclusion criterion was mental retardation.
BPD patients had to meet the DSM-III-R criteria for BPD as assessed by the SCID-II; clusterC patients the criteria for minimal one cluster-C PD as assessed by the SCID-II and R2 BPD
criteria; the normal controls had to report satisfaction with present life conditions, and no
Axis-I disorder, PD or any BPD criterion (assessed with the SCIDs).

548

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

Mean age of the BPD group was 29.6 yr, of the cluster-C group 32.6 yr, and of the normal
controls 35.0 yr. (NS, KruskallWallis). BPD patients had on the average 2.81 Axis-I
disorders, cluster-C patients 2.33 (NS, KruskallWallis). BPD patients were found to have
more PDs (m = 2.94) than cluster-C patients (m = 1.83; KruskallWallis w 2=3.42, p = 0.03,
d.f. = 1). The three groups did not dier signicantly with respect to location, educational
level (three levels were distinguished), and proportion living with a partner.
2.2. Materials
A Dutch version of the SCID-II was used to diagnose PDs (Psychiatrisch Centrum
Bloemendaal, 1991). This version has proved to have good inter-rater reliability (average
Kappa = 0.80; Kappa for BPD = 0.79; ICC for BPD trait score = 0.91; Arntz et al., 1992;
testretest ICC for traitscore = 0.72; Dreessen & Arntz, 1998). The Dutch version of the
SCID-I was used to assess Axis-I disorders (Koster van Groos, 1985; Arntz, Bogels, &
Hoekstra, 1992). Interviewers either participated in the original reliability study of the SCID-II
or had been trained by these interviewers.
The Personality Disorder Beliefs Questionnaire (PDBQ, Dreessen & Arntz, 1995) was used
to assess strength of belief in a series of assumptions, hypothesized to be specic to various
PDs. Each PD is represented by 20 assumptions, partly formulated on the basis of Beck et al.
(1990; with permission), partly hypothesized by the constructors. All assumptions hypothesized
to be specic to BPD were constructed by the authors, since the appendix of Beck et al.'s
(1990) book does not give a list of BPD assumptions (see Arntz, 1994, for hypothesized themes
of BPD assumptions). Assumptions of six common PDs were selected and randomized,
yielding a 120-item questionnaire. Subjects were instructed to rate strength of belief in each
assumption by placing a vertical mark on 100 mm Visual Analogue Scales (VASs), with
anchors ``I don't believe this at all'' and ``I believe this completely''. Ratings were expressed in
mm, a higher score meaning stronger belief in the assumption. For each of the six subscales an
averaged belief score was derived (range 0100). Internal consistencies proved to be excellent in
the present sample: Avoidant (Cronbach alpha = 0.96); Dependent (0.92); ObsessiveCompulsive (0.92); Paranoid (0.97); Histrionic (0.88); and BPD assumptions (0.95).
A Dutch version of the Childhood Trauma Questionnaire (CTQ; Bernstein & Fink, 1993;
Bernstein, Fink, & Handelsman, 1994; Arntz & Wessel, 1996) was used to assess traumatic
experiences during childhood. Internal consistencies of the subscales assessed with Cronbach
alpha proved to be excellent in the present sample: Physical and Emotional Abuse (0.96),
Emotional Neglect (0.96), Physical Neglect (0.87), Sexual Abuse (0.95).
To get more factual information about sexual abuse, a one-page self-report checklist asked
subjects to indicate whether they had ever been sexually approached against their wish, by
whom, at which age(s), the duration of the abuse, and what had happened. A sexual abuse
sumscore was constructed out of the number of perpretators, the age-level at time of the abuse
(the younger the higher the score), the duration (the longer the duration the higher the score),
and the severity of what had happened. In case of several abuse reports, each contributed to
the sumscore separately. The internal consistency was good (Cronbach alpha = 0.80).
On a 5-item questionnaire subjects rated present emotional state on 100 mm Visual
Analogue Scales. Degree of happiness, sadness, anxiousness, anger and restlessness were scored

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

549

with a high score meaning high negative emotion, and averaged. Reliabilities of the three
resulting averaged scores were good (week 1 pre-test Cronbach alpha = 0.82; week 2 pre-test
Cronbach alpha = 0.79; post-test Cronbach alpha = 0.86).
A 10 min fragment of the movie ``Sleeping with the enemy'' on video tape was used to
inuence subjects' state (schema activation and induction of negative emotions). This fragment
shows the disbalance in power between a suspicious man and his wife. The husband tries to
control his wife in a violent way. When he suspects her without any ground of adultery he
beats her. After this he behaves as if nothing has happened, and she doesn't dare to raise the
issue.
2.3. Procedure
Before subjects participated they signed an informed consent form. Next, the SCIDs, if not
already administered, were administered by one of the authors. Finally, subjects lled out the
PDBQ and the emotion questionnaire. One week later subjects participated in a second session,
starting with the CTQ, the one-page self-report form on sexual abuse and the emotion
questionnaire. After a 10 min break the video fragment was displayed on a television set.
Subjects were instructed to watch the video for approximately 10 min. Nothing about the
content was mentioned, nor about the title. Subjects were told that they would not be
interviewed about the lm fragment, so that they did not had to remember specic issues.
Following this, subjects lled out the emotion list and the PDBQ. Lastly, subjects were
debriefed and thanked for participation.

3. Results
3.1. Specicity of BPD assumptions
Mean scores on the six PDBQ subscales and test statistics are summarized in Table 1. BPD
patients made on the average the highest belief ratings on all six PDBQ subscales, with the
cluster-C patients somewhere in between the normal controls and the BPD patients. However,
BPD patients did not dier signicantly from cluster-C patients as to their mean belief in
avoidant, dependent and obsessive-compulsive beliefs (Table 1). BPD patients scored
signicantly higher ( p < 0.05) than cluster-C patients on paranoid, histrionic and BPD
assumptions. After covarying out the contribution of paranoid PD pathology by ANCOVA
with number of paranoid traits assessed with the SCID-II as covariate, the dierence between
BPD and cluster-C patients on the paranoid subscale of the PDBQ was no longer signicant
(Table 1). The same held for the histrionic subscale of the PDBQ: after covarying out the
contribution of SCID-II histrionic traits, the dierence between the patients groups was no
longer signicant (Table 1). A stepwise discriminant analyses on the three groups was
subsequently executed with weighted prior probabilities (based on group sizes), minimizing
Wilk's Lambda as selection rule, and signicance level to enter of 0.05. The PDBQ BPD
assumptions scale was the only subscale that was entered (F(2, 40) = 23.96, p < 0.001). Of the

550

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

Table 1
Mean scores of the three groups on the 6 assumption scales and test statistics of two contrasts
Normals

C-cluster

BPD

BPD vs. normals

BPD vs.
C-cluster

PBDQ scale

mean

S.D.

mean

S.D.

mean

S.D.

t(29)

t(26)

Avoidant
Dependent
Obs-comp.
Paranoid

12.5
16.1
24.7
12.4

9.5
11.4
13.5
7.7

23.7
18.4
18.7
23.5

8.13
4.85
4.10
5.94

<0.001
<0.001
<0.001
<0.001

17.1

7.7

10.9

6.60

<0.001

BPD

11.5

7.1

51.7
40.2
47.1
47.9
40.3a
39.6
37.2b
48.1

16.3
15.8
16.7
21.9

Histrionic

40.3
36.4
42.3
26.8
34.4a
26.5
28.9b
32.7

15.4

8.36

<0.001

1.52
0.58
0.70
2.45
0.68a
2.90
1.64b
2.30

0.14
0.57
0.49
0.022
0.53a
0.007
0.11b
0.029

12.9
19.9

Note. T-tests of planned comparisons.aAdjusted means, t-statistic and p-level after covarying out the number of
paranoid traits (SCID-II).bAdjusted means, t-statistic and p-level after covarying out the number of histrionic traits
(SCID-II).

43 cases 63% was correctly classied; 75% of the BPD patients, 87% of the normal controls,
but only 17% of the cluster-C patients was correctly classied1.
3.2. Stability of assumptions
Table 2 presents mean negative emotion scores on the three measurements. In the whole
sample there was a slight reduction in negative emotions from week 1 to week 2, which failed
to reach signicance (t(42) = 0.98; paired t-test). As intended, negative emotions raised after
the video fragment (compared to week 1: t(42) = 2.00, p = 0.05; compared to immediately
before the video: t(42) = 3.36, p = 0.002; paired t-tests). In general, both patient groups scored
higher on negative emotions than the normal controls (Table 2). BPD patients did not show
stronger emotional response to the video fragment than normal controls (NS), and cluster-C
patients (NS). Table 3 presents changes in PDBQ assumption scores from pretest to posttest;
these were generally small. In both clinical groups they were all negative, implying that the
1
Because of the relatively small sample size and the relatively large number of predictors in the discriminant analysis, a replication was done in a second sample with data from a study by van den Hoorn (1998). As part of a study
on psychometric properties of the Borderline Personality Disorder Severity Index (Arntz, Cornelis, & de Bie, 1997)
van den Hoorn let BPD and cluster-C PD patients ll out the short version of the PDBQ. The sample consisted of
47 patients, 23 with cluster-C PDs, and 24 with BPD, as diagnosed with the SCID-II. BPD patients scored higher
than cluster-C patients on all 6 subscales of the PDBQ (t(45)>2.90; p < 0.01), but the dierence was the most pronounced on the BPD assumption subscale (t(45) = 4.62, p < 0.0001). A stepwise discriminant analyses was subsequently executed with weighted prior probabilities (based on group sizes), minimizing Wilk's Lambda as selection
rule, and signicance level to enter of 0.05. The PDBQ BPD assumptions scale was the only subscale that was
entered (F(1, 45) = 21.32, p < 0.0001). Of the 47 cases 76.6% was correctly classied; 79.2% of the BPD patients
and 73.9% of the cluster-C patients. These results were comparable to those obtained in the rst sample and suggest
that the observed specicity of assumptions hypothesized to be specically related to BPS was not based on chance.

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

551

Table 2
Mean scores on the averaged negative emotions scale on the three measurements
Normals

C-cluster

BPD

Total sample

Measurement

mean

S.D.

mean

S.D.

mean

S.D.

mean

S.D.

week 1 pretest
week 2 pretest
week 2 posttest

17.5
20.1
24.6

15.8
15.8
13.3

42.2
33.8
49.0

20.6
17.6
23.4

45.2
41.0
49.8

20.8
19.1
23.5

34.7
31.7
40.8

22.7
19.4
23.3

video, though increasing negative emotional state, did not lead to increased belief in various
basic assumptions. More specically, BPD patients did not show an increased score in any of
the six groups of assumptions of the PDBQ, including the BPD assumptions (Table 3; since all
assumption subscale scores decreased in the BPD sample statistical testing is redundant). Test
retest correlations (Table 4) also indicate that interindividual dierences on assumption scores
were remarkably stable. Only the cluster-C assumptions scores (avoidant, dependent, obsessivecompulsive) in the BPD patients were somewhat less stable, but the BPD and paranoid
assumptions were very stable in the BPD patient sample (Table 4).
Exploring the changes in emotions after the video, it appeared that of the ve emotions
anger increased most in both clinical groups. Given this relatively specic eect and the
content of the video it was explored whether especially malevolence (of others) assumptions
were increased in BPD patients after the video. A mean score of malevolence assumptions of
the BPD assumptions was constructed, but this score failed to increase from pre-test to posttest in the BPD group (again, a small decrease was observed).

Table 3
Mean assumption scores of the three groups on pretest and posttest
Normals

C-cluster

BPD

PBDQ scale

mean

S.D.

mean

S.D.

mean

S.D.

Avoidant pre
Avoidant post
Dependent pre
Dependent post
Obs-comp. pre
Obs-comp. post
Paranoid pre
Paranoid post
Histrionic pre
Histrionic post
BPD pre
BPD post

12.5
13.0
16.1
14.8
24.7
22.1
12.4
13.1
17.1
19.5
11.5
11.4

9.5
10.1
11.4
11.7
13.5
12.9
7.7
9.3
7.7
8.5
7.1
8.9

40.3
33.8
36.4
33.3
42.3
38.5
26.8
25.2
26.5
26.4
32.7
30.7

23.7
23.7
18.4
16.2
18.7
18.7
23.5
23.6
12.9
14.0
19.9
18.2

51.7
48.7
40.2
34.9
47.1
43.8
47.9
41.2
39.6
34.6
48.1
44.5

16.3
20.1
15.8
13.8
16.7
14.3
21.9
20.1
10.9
9.2
15.4
15.4

552

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

Table 4
Testretest correlations of PDBQ assumption subscales
Subscale

Total sample

Normals

Cluster-C

BPD

Avoidant
Dependent
Obs-comp.
Paranoid
Histrionic
BPD

0.92
0.89
0.88
0.95
0.90
0.95

0.86
0.93
0.91
0.91
0.86
0.91

0.94
0.87
0.91
0.98
0.93
0.93

0.78
0.75
0.69
0.88
0.81
0.87

Note. All correlations p < 0.001, except for R = 0.69 ( p = 0.003).

3.3. Borderline personality disorder, childhood trauma, and cognitive assumptions

In accordance with previous ndings, more BPD patients reported childhood sexual abuse
experiences than Ss of the other groups on our self-constructed sexual abuse checklist: 87.5%
of the BPD patients reported any form of sexual abuse, compared to 41.7% of cluster-C and
33.3% of normal controls (w 2(1) = 6.60, p < 0.01; w 2(1) = 9.57, p < 0.01 respectively; note
that the checklist also comprises milder forms of sexual abuse). The sexual abuse sumscore was
signicantly higher in the BPD group than in the other groups: F(2, 40) = 8.59, p < 0.001;
BPD vs. normals t(29) = 3.95, p < 0.001; BPD vs. cluster-C t(26) = 2.92, p < 0.01. In all
aspects (number of perpretators, age, duration and severity of what had happened) the BPD
group scored higher than the other groups. The BPD group had also signicantly higher scores
on two subscales of the CTQ, physical/emotional abuse ( p < 0.01) and physical neglect
( p < 0.05), compared to the control groups. On the CTQ emotional neglect and sexual abuse
subscales the BPD group did not dier signicantly from the control groups. To test whether
the relationship between BPD and childhood trauma was specically related to sexual abuse,
and whether this was mediated by BPD assumptions a series of regression analyses was
executed.
In the rst series the number of BPD-traits as assessed with the SCID-II was regressed on
the four CTQ subscales and on the sexual abuse sumscore. In a backward analyses two
predictors remained: the sexual abuse sumscore (Beta = 0.37, p = 0.015) and the CTQ
Physical/Emotional Abuse factor (Beta = 0.37, p = 0.016) (42.5% variance explained). When
the BPD assumption score was added as a predictor, a backward procedure showed that sexual
abuse and physical/emotional neglect did no longer contribute to BPD pathology: BPD
assumptions (Beta = 0.65, p < 0.001) and CTQ Physical Neglect (Beta = 0.29, p < 0.01)
accounted for 57% of the variance. A more stringent test of the role of BPD assumptions is to
control for the other ve PDBQ assumption scales. Regression analyses showed however that
only BPD assumptions and CTQ physical neglect related signicantly to BPD pathology, the
other assumption scales did not contribute signicantly to BPD pathology.
The relationship between BPD assumptions and childhood trauma factors was assessed by a
regression analysis with all CTQ scales and the sexual abuse sumscore as predictors and BPD
assumptions score as dependent variable. A backward procedure demonstrated that the sexual

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

553

abuse sumscore (Beta = 0.35, p = 0.033) and CTQ Physical/Emotional Abuse factor
(Beta = 0.27, p = 0.098) were signicantly related to the BPD assumptions, explaining 31% of
the variance.
Lastly, a series of logistic regression analyses were executed to check whether the ndings
could be replicated using a categorical denition of BPD. The sexual abuse sumscore and CTQ
Physical/Emotional Abuse scale were the only predictors of BPD diagnosis (74% correctly
classied). After entering the PDBQ BPD assumptions as predictor, sexual abuse and physical/
emotional abuse were no longer signicant predictors of BPD status. BPD assumptions
( p = 0.001) and, to a lesser degree, CTQ Physical Neglect ( p = 0.03) were related to BPD
diagnosis (88% correctly classied, w 2(2) = 24.43, p < 0.0001).
To summarize, the results support the hypothesis that BPD assumptions mediate the
relationship between childhood traumatic experiences and BPD pathology. BPD assumptions
can be viewed as a more proximate `cause' of BPD pathology, compared to sexual abuse and
other childhood factors. The BPD assumptions as assessed with the PDBQ were clearly related
to the sexual abuse sumscore, and, less strongly, to Physical/Emotional Abuse. The
relationship of one childhood factor with BPD pathology, Physical Neglect, does not appear to
be mediated by the PDBQ BPD assumptions. The resulting model with BPD pathology
expressed as number of traits as dependent variable was tested with structural equation
modeling (LISREL). Path coecients are shown in Fig. 1. Tests indicated good goodness of t
(normed t index = 0.94; non-normed t index = 0.91) and a nonsignicant dierence from
the saturated model (w 2=5.21, p = 0.16, d.f. = 3).

Fig. 1. The relationships between childhood traumas, PDBQ BPD assumptions, and BPD pathology. Standardized
path coecients are shown. PDBQ BPD assumptions mediate the relationship between physical/emotional and
sexual abuse and BPD pathology. Apparently, the contribution of physical neglect is not represented in the PDBQ
BPD assumptions. +p < 0.10; *p < 0.05; **p < 0.01 two-tailed. (1) Sexual Abuse (sumscore) without correction for
Physical/Emotional Abuse: Beta = 0.51 ( p = 0.0005). (2) Physical/Emotional Abuse (CTQ) without correction for
sexual abuse: Beta = 0.47 ( p = 0.0014). (3) Expressed as number of BPD traits assessed with the SCID-II. When
expressed as diagnosis (BPD yes/no) results are comparable and 88% of the Ss are correctly classied
(w 2(2) = 24.43, p < 0.0001).

554

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

4. Discussion
BPD patients were, as hypothesized, characterized by high belief scores on each of the six
investigated PDBQ subscales. The BPD assumption subscale discriminated the three groups the
most strongly, and was the sole discriminating scale in the discriminant analysis. Though at
rst sight paranoid and histrionic assumptions also discriminated BPD patients from cluster-C
patients, these dierences disappeared when the inuence of corresponding PD traits as
assessed with the SCID-II were covaried out. Thus, from the six PDBQ subscales the BPD
assumptions seem to constitute the essential characteristic of BPD. This nding was replicated
in a new sample of cluster-C and BPD patients.
The BPD assumptions appeared to be remarkably stable for such an unstable disorder.
Despite induction of negative emotions by a lm fragment hypothesized to be relevant to BPD
pathology in particular (indeed several BPD patients were very upset as reported by their
therapists), the belief scores did not increase and testretest correlations were high. The lack of
increase in belief ratings after viewing the lm fragment cannot be attributed to a ceiling eect:
the mean scores (Table 3) show that there was enough room for higher ratings. Though the
negative emotions score was generally higher after viewing the lm fragment than at session 1,
the relatively high levels of negative emotions during the rst session may have obscured the
eects of negative emotional state and schema priming on belief scores. Further studies should
clarify this issue, but the present ndings suggest high stability of belief scores in BPD patients.
This is in agreement with our observations in treatment: idiosyncratic beliefs scored at start
and end of each treatment session remain for a long time remarkably stable. Nevertheless,
belief ratings may show more variability in BPD under dierent conditions, e.g.: (i) when
subjects are instructed to rate their present belief strength instead of their general belief (state
vs. trait); (ii) with dierent forms of emotional change (e.g., naturally occurring instead of
experimentally induced); (iii) with anxiety, depression, or happiness instead of anger that was
most profound with the present induction; (iv) with more intense emotions; (v) or with
dierent emotional stimuli. Despite the need for more research, it remains of interest that it is
possible to measure core assumptions specic to BPD on a stable basis.
An explanation for any strong relationship between a measure X and a diagnosis Y is that
measure X consists of the same variables that dene the diagnosis. This would, of course, lead
to purely tautological ndings. Table 5 lists the 20 BPD assumptions of the PDBQ. It is clear
that the majority does not just reect a DSM BPD criterion. However, it could be argued that
items 6, 7 and 8 are closely related to DSM-IV criteria 3 (item 8) and 4 (6 and 7). On the other
hand, an analysis with the BPD assumption score with these items left out showed that
specicity did not change. In a discriminant analysis this corrected score remained capable of
classifying 75% of the BPD patients, 25% of the cluster-C patients, and 87% of the normals
correctly. Thus, it seems unlikely that the present ndings were tautological.
In general, the results supported the idea that BPD pathology is more related to childhood
sexual and emotional abuse, than other personality disorders. The nding that BPD
assumptions mediate (in a statistical sense) the relationship between childhood sexual, physical
and emotional abuse and BPD pathology support the hypothesis that childhood trauma form
the basis upon which dysfunctional schemas can develop which, in turn, may underlie BPD
pathology (see Arntz, 1994). The evidence is, of course, purely correlational and therefore any

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

555

Table 5
BPD assumptions of the PDBQ
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(17)
(18)
(19)
(20)

I will always be alone.

There is no one who really cares about me, who will be available to help me, and whom I can fall back on.
If others really get to know me, they will nd me rejectable and will not be able to love me; and they will
leave me.
I can't manage it by myself, I need someone I can fall back on.
I have to adapt my needs to other people's wishes, otherwise they will leave me or attack me.
I have no control of myself.
I can't discipline myself.
I don't really know what I want.
I need to have complete control of my feelings otherwise things go completely wrong.
I am an evil person and I need to be punished for it.
If someone fails to keep a promise, that person can no longer be trusted.
I will never get what I want.
If I trust someone, I run a great risk of getting hurt or disappointed.
My feelings and opinions are unfounded.
If you comply with someone's request, you run the risk of loosing yourself.
If you refuse someone's request, you run the risk of loosing that person.
Other people are evil and abuse you.
I'm powerless and vulnerable and I can't protect myself.
If other people really get to know me they will nd me rejectable.
Other people are not willing or helpful.

conclusion about causality cannot be drawn. It is, for instance, unclear whether childhood
abuse causes the formation of BPD assumptions as measured by the PDBQ, or that these
assumptions were already present in the child and increased the risk of abuse, or that a third
factor causing both was present. General family factors as a third variable have been
suggested, though the results reported by Weaver & Clum (1993) suggest that sexual abuse is
the essentially characteristic childhood factor in BPD. In contrast to Weaver and Clum's
ndings, the present study suggests that physical and emotional abuse have an additional and
equally important relationship with BPD as sexual abuse. This is in accord with the fact that
though very high proportions of BPD patients report sexual abuse, not all of them seem to
have experienced sexual traumatization during childhood. This nding is also in accordance
with Linehan's (Linehan, 1993) theory, stressing the role of lack of emotional validation of the
child by central rearing gures in the etiology of BPD.
Another problem is the retrospective nature of the report of childhood trauma. However, a
recent study by Bernstein, Ahluvalia, Pogge, and Handelsman (1997) suggests that the CTQ is
a valid instrument: CTQ reports by psychiatric adolescents appeared to be strongly related to
ratings of various forms of abuse by therapists who used more objective indicators of abuse (as
Childhood Protective Service reports, etc.). As to the validity of the sexual abuse checklist, this
self-report form asked for very specic sexual acts, and not for subjective experiences. The
more a questionnaire asks for subjective experiences, the more room there is for biased reports,
for instance caused by high negative aectivity and neuroticism in severe patients (van den
Hout, Nelissen, Jansen, Merckelbach, & Schouten, 1997). The self-report checklist seems a
relatively valid method. This is corroborated by the fact that results with this self-report form

556

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

were essentially the same as others have found with lengthy interviews. A puzzling nding was
that the sexual abuse sumscore performed better than the CTQ sexual abuse score in the
analyses. Perhaps our checklist, asking for (serious) facts, measures the more severe forms of
sexual abuse, which may be more characteristic of BPD than the CTQ questions (Bernstein,
personal communication). A (unexpected) third childhood trauma factor, physical neglect,
appeared in the analysis after entering the BPD assumptions. It is dicult to interpret this
nding, since this CTQ factor did not play any role in the regression analyses before entering
the BPD assumptions. Future research is necessary to nd out whether or not childhood
physical neglect is characteristic of BPD. It seems clear, however, that the cognitive schemas
related to this factor are not represented in the BPD assumption scale of the PDBQ. Thus, if
physical neglect during childhood turns out to be related to BPD pathology it is indicated to
adapt the BPD assumptions scale.
Future studies are also needed to address remaining questions. For instance, the present
study investigated only women. It is possible that other assumptions and other childhood
factors are characteristic of male BPD patients. Furthermore, the present study is only a rst
step towards the identication of the core beliefs of BPD patients. Because of the small sample
size, more detailed analyses on item level were for instance not useful. The present ndings are
only a rough indication that it is possible to identify characteristic BPD assumptions. This
study was also not designed to test the specicity of the assumptions of the other PDs. Given
the mixed group of cluster-C PDs in the control group it is not amazing that the discriminant
analysis did predict cluster-C patients so poorly. A much larger scale study is indicated to test
the specicity of all PDBQ subscales. It would be interesting to investigate the relationships
between various hypothesized childhood factors, like parental rearing styles, attachment,
childhood traumas, etc., and fundamental assumptions, in addition to testing the associations
between assumptions and personality disorders.

Acknowledgements
Arthur de Bie (Psychiatric Hospital Maastricht) and Henriette Kuipers (Psychiatric Hospital
Johan Weijer, Amsterdam) are acknowledged for their help in patient recruitment. David
Bernstein (Bronx VA Medical Center, New York) gave Arnoud Arntz and Ineke Wessel
(University of Maastricht) permission to translate the CTQ into Dutch. Erik Schouten
(University of Maastricht) helped with executing the LISREL computations. Aaron T. Beck
gave us permission to use assumptions listed in Beck et al. (1990) for developing the PDBQ.
Marije van den Hoorn (University of Maastricht) gave permission to use her data set for
replication of the discriminant analysis.

References
Arntz, A. (1994). Treatment of borderline personality disorder: a challenge for cognitive-behavioural therapy. Behaviour Research and
Therapy, 32, 419430.
Arntz, A., & Wessel, I. (1996). Jeugd trauma vragenlijst (Dutch version of the childhood trauma questionnaire). Maastricht: Author.

A. Arntz et al. / Behaviour Research and Therapy 37 (1999) 545557

557

Arntz, A., Bogels, S., & Hoekstra, R. (1992). Revisie van de Nederlandse vertaling van het gestructureerd klinisch interview voor
DSM-III-R (SCID-I) (Revision of the Dutch translation of the structured clinical interview for DSM-III-R (SCID)). Maastricht:
Author.
Arntz, A., van Beijsterveld, B., Hoekstra, R., Hofman, A., Eussen, M., & Sallaerts, S. (1992). The interrater reliability of the structured
clinical interview for DSM-III-R personality disorders. Acta Psychiatrica Scandinavica, 85, 394400.
Arntz, A., Cornelis, J., & de Bie, A.J.H.T. (1997). Reliability and validity of the borderline personality disorder severity index. Journal
of Personality Disorders, accepted pending revision.
Beck, A. T., Freeman, A., & Associates (Eds.). (1990). Cognitive therapy of personality disorders. New York: The Guilford Press.
Bernstein, D.P., & Fink, L. (1993). Childhood Trauma questionnaire. Boston: Author.
Bernstein, D. P., Fink, L., & Handelsman, L. (1994). Initial reliability and validity of a new retrospective measure of child abuse and
neglect. American Journal of Psychiatry, 151, 11321136.
Bernstein, D. P., Ahluvalia, T., Pogge, D., & Handelsman, L. (1997). Validity of the Childhood Trauma Questionnaire in an adolescent
psychiatric population. Journal of the American Academy of Child and Adolescent Psychiatry, 36, 340348.
Bryer, J. B., Nelson, B. A., Miller, J. B., & Krol, P. A. (1987). Childhood sexual and physical abuse as factors in adult psychiatric illness. American Journal of Psychiatry, 144, 14261430.
Coons, P. M., Bowman, E. S., Pellow, T. A., & Schneider, P. (1989). Post-traumatic aspects of the treatment of victims of sexual abuse
and incest. Psychiatric Clinics of North America, 12, 325335.
Dreessen, L., & Arntz, A. (1995). The personality disorder beliefs questionnaire (short version). Maastricht: Author.
Dreessen, L, & Arntz, A (1998). Short-term testretest interrater reliability of the structured clinical interview for DSM-III-R personality disorders (SCID-II) in outpatients. Journal of Personality Disorders, 12, 138148.
Herman, J. L., Perry, J. C., & van der Kolk, B. A. (1989). Childhood trauma in borderline personality disorder. American Journal of
Psychiatry, 146, 490495.
van den Hoorn, M. (1998). Betrouwbaarheid en validiteit van de borderline personality disorder severity index (Reliability and validity
of the borderline personality disorder severity index). MA thesis, Maastricht University.
van den Hout, M.A., Nelissen, I., Jansen, A.T., Merckelbach, H., & Schouten, E.G. (1997). Neuroticism, `negative disambiguation',
and self-reported sexual abuse. Submitted.
Koster van Groos, G.A.S. (1985). Nederlandse vertaling van het gestructureerd klinisch interview voor DSM-III-R (SCID-I) (Dutch
translation of the structured clinical interview for DSM-III-R (SCID-I)). Unpublished manuscript.
Linehan, M.M. (1993). Cognitive-behavioral treatment of borderline personality disorder. New York: Guilford.
Ogata, S. N., Silk, K. R., Goodrich, S., Lohr, N. E., Westen, D., & Hil, E. M. (1990). Childhood sexual and physical abuse in adult
patients with borderline personality disorder. American Journal of Psychiatry, 147, 10081013.
Psychiatrisch Centrum Bloemendaal (1991). Nederlandse vertaling van het gestructureerd klinisch interview voor DSM-III-R
Persoonlijkheidsstoornissen (SCID-II) (Dutch translation of the structured clinical interview for DSM-III-R personality disorders).
Bloemendaal: Author.
Safran, J. D., Segal, Z. V., Hill, C., & Whien, V. (1990). Rening strategies for research on self-representations in emotional disorders. Cognitive Therapy and Research, 14, 143160.
Schmidt, N. B., Joiner, Th. E., Young, J. E., & Telch, M. J. (1995). The schema questionnaire: investigation of psychometric properties
and the hierarchical structure of a measure of maladaptive schemas. Cognitive Therapy and Research, 19, 295321.
Stone, M. H., Unwin, A., Beacham, B., & Swenson, C. (1988). Incest in female borderlines: its frequency and impact. International
Journal of Family Psychiatry, 9, 277293.
Veen, G., & Arntz, A. (1997). Multidimensional dichotomous thinking characterizes borderline personality disorder. Submitted for
publication.
Weaver, T. L., & Clum, G. A. (1993). Early family environments and traumatic experiences associated with borderline personality disorder. Journal of Consulting and Clinical Psychology, 61, 10681075.
Young, J.E. (1990). Cognitive therapy for personality disorders: a schema-focused approach. Sarasota: Professional Resource Press.