STEIN, PINSKER-ASPEN, HILSENROTH

PAI AND BPD

Borderline Pathology and the Personality Assessment

Inventory (PAI): An Evaluation of Criterion

and Concurrent Validity

Michelle B. Stein, Janet H. Pinsker-Aspen, and Mark J. Hilsenroth

Derner Institute of Advanced Psychological Studies

Adelphi University

In this study, we examined how patients diagnosed with borderline pathology (BP) would re-
spond on the Personality Assessment Inventory (PAI; Morey, 1991) Borderline (BOR) scales in
relation to patients without BP pathology. In addition, we examined whether the PAI BOR
scales would be related to variables on the Social Cognition and Object Relations Scale
(SCORS; Hilsenroth, Stein, & Pinsker, 2004; Westen, 1995) derived from early memory narra-
tives. Results indicate that outpatients with a Diagnostic and Statistical Manual of Mental Dis-
orders (4th ed. [DSM–IV]; American Psychiatric Association, 1994) diagnosis of BP scored
significantly higher on the PAI BOR Total (BOR–Total) score, Identity Problems, and Self-
Harm scales in comparison to a Non-BP clinical sample. The overall correct classification rate
for the presence or absence of BP using the BOR Total scale (T

≥

70) was 73%. In addition,

there were several significant relationships between dimensional PAI BOR scales and the pres-
ence versus absence of DSM–IV BP. Moreover, both the BOR-Total and Affect Instability
scales were significantly related to the SCORS variable Complexity of Representations. We
provide clinical examples to illustrate these research findings in an applied manner.

Prior research has shown that the Personality Assessment In-
ventory (PAI; Morey, 1991) is a useful self-report measure in
the assessment of borderline pathology (BP; Bell-Pringle,
Pate, & Brown, 1997; Kurtz & Morey, 2001; Kurtz, Morey, &
Tomarken, 1993; Morey, 1991, 1996; Trull, 1995). The PAI
Borderline (BOR) scales were initially developed to target
specific characteristics that are listed in the Diagnostic and
Statistical Manual of Mental Disorders (3rd ed., rev.
[DSM–III–R]; American Psychiatric Association, 1987) for
borderline personality disorder (BPD; Bell-Pringle et al.,
1997; Morey, 1991, 1996). In one investigation of clinical va-
lidity, Bell-Pringle et al. (1997) were able to differentiate
BPD patients from unscreened controls with 80% accuracy
using the BOR scale. Bell-Pringle et al. also demonstrated
that the BOR scale of the PAI was more accurate in assessing
BPD in a patient population than using profile configurations
of the Minnesota Multiphasic Personality Inventory–2
(MMPI–2; Hathaway & McKinley, 1989). In a second clini-
cal study, Trull (1995) demonstrated that the BOR scale can
be successfully applied to many areas of functioning that in-
dividuals with BPD traditionally struggle with such as de-
pression, coping, and interpersonal problems. In addition,
Kurtz et al. (1993) established both convergent and

discriminant validity between the PAI BOR scales and
MMPI (Hathaway & McKinley, 1967) Personality Disorder
scales (Morey, Waugh, & Blashfield, 1985) in a nonclinical
sample.

The PAI has also been used to discriminate BPD in

comorbid samples. Kurtz and Morey (2001) compared two
groups of outpatients (comorbid major depressive epi-
sode/BPD patients and major depressive episode patients
without BPD) against a control group. The number of Axis I
and Axis II diagnoses were similar for the BPD and non-BPD
groups. Kurtz and Morey found that BPD patients scored sig-
nificantly higher on the PAI BOR raw score (BOR–Total),
Affective Instability (BOR–A), Identity Problems (BOR–I),
and the Negative Impression Management (NIM) scales
when compared to non-BPD patients. In sum, Kurtz and
Morey’s study supported the construct validity of this self-
report measure in assessing BP among comorbid samples.

Previous research has also shown that self-report mea-

sures of personality and psychopathology have demonstrated
important concurrent validity with reliably rated patient nar-
rative data. Acklin, Bibb, Boyer, and Jain (1991) found that
relationship episodes were reliably coded for a clinical and
nonclinical sample. In addition, early memory (EM) based

JOURNAL OF PERSONALITY ASSESSMENT, 88(1), 81–89

Copyright © 2007, Lawrence Erlbaum Associates, Inc.

scores were correlated with self-report measures of mood
(Profile of Mood States; McNair, Lorr, & Droppleman,
1971), attachment style (Separation-Individuation Test of
Adolescence; Levine, Green, & Millon, 1986), and clinical
symptomatology (MMPI; Hathaway & McKinley, 1967;
Symptom Checklist–90–Revised; Derogatis, 1983).

EM-based narrative rating scales have also demonstrated

important relations with BPD pathology. Nigg, Lohr,
Westen, Gold, and Silk (1992) evaluated EMs of borderline
patients (BPD) with and without major depressive disorder
(MDD; BPD/MDD) as well as MDD patients and nonclinical
participants to assess malevolent object relations and affect
tone. Nigg et al. found that pure BPD and comorbid
BPD/MDD participants had significantly higher percentages
of extreme malevolent responses than did MDD participants
and nonclinical controls. BPD patients recalled their rela-
tionships in EMs as more injurious and unhelpful. Last, af-
fect tone scores of BPD patients were more malevolent
compared to all other groups. Moreover, Fonagy et al. (1996)
found that there were significant differences between BPD
and non-BPD outpatients on the Reflective Self-Function
Scale (RF; Fonagy, Moran, Steele, & Higgitt, 1991). Based
on interview narrative data, the RF score was lower for BPD
patients, which indicated greater pathology. In sum, Fonagy
et al.’s (1996) and Nigg et al. (1992) studies found that the
greater the BPD pathology, the more likely such patients
were to perceive others in a malevolent and limited fashion.

In addition, a number of studies that have utilized EM nar-

ratives have been particularly important in examining the re-
lated issues of object representations and affect of BPD
patients. One of the first studies in this area was by Frank and
Paris (1981) who interviewed BPD patients regarding their
parents and assessed EMs for parental acceptance of child-
hood dependent and independent behaviors. The three
groups in Frank and Paris’s study were a nonclinical sample,
an outpatient sample consisting of individuals with neuroses
and personality disorders (non-BPD) as well as the third
group of BPD patients. The findings demonstrated that BPD
patients had significantly more negatively toned memories
than did the nonclinical sample. In addition, Frank and Paris
found differences between the nonclinical and borderline
group with respect to the quality of EMs (BPD fathers were
remembered as less approving and more disinterested). In a
second study, Arnow and Harrison (1991) also examined the
relational narratives of BPD individuals through the affective
components of their EMs. Participants included outpatients
that were placed in three diagnostic groups: neurotic pathol-
ogy, BPD, and paranoid schizophrenia. Results revealed sig-
nificant differences in the number of positive memories, with
BPD patients reporting the fewest as compared to both pa-
tients with neurosis and paranoid schizophrenia. Another
significant difference was found in the number of negative
memories, with the BPD patients providing more negative
affect in relation to the other two groups. Furthermore, the
BPD patients revealed more negative affect even when asked

for feelings associated with their happiest EM. Overall,
Frank and Paris’s and Arnow and Harrison’s studies found
that BPD patients had a greater amount of negative affect as-
sociated with EMs when compared to other clinical and
nonclinical groups.

One narrative rating measure that has shown a pattern of

clinical validity is the Social Cognition and Object Relations
Scale (SCORS; Hilsenroth, Stein, & Pinsker, 2004; Westen,
1995). In particular, the SCORS can be used to rate narrative
data and is particularly sensitive to BPD pathology. The afore-
mentioned research has assessed constructs similar to the
SCORS variables Complexity of Representations (COM;
Nigg et al., 1992; Fonagy et al., 1996) and Affective Quality of
Representations (AFF; Frank & Paris, 1981; Arnow & Harri-
son, 1991). For instance, Fonagy et al.’s (1996) RF scale as-
sesses the clarity of individual representation of mental states
of the self as well as the other, much like that of the COM vari-
able on the SCORS. In addition, the SCORS has also been used
successfully to examine EM narratives. Fowler, Hilsenroth,
and Handler (1995) utilized EM-based SCORS ratings in both
categorical and dimensional analyses with a clinical outpa-
tient and nonclinical sample. Results indicated significant dif-
ferences between the clinical and nonclinical groups’ EM
narratives using the COM and AFF scales of the SCORS, with
the clinical sample having lower, more maladaptive, scores.
The clinical outpatient group reported more negative affect
accompanied by victimization at the hands of malevolent
adults (AFF) as well as characterizations of the self and other
that tended to be one-dimensional and simplistic (COM).
Fowler et al.’s (1995) study also assessed the validity of the
SCORS variables COM and AFF with similar, explicit, self-
report constructs measured by the Ego Strength and Anger
scales of the MMPI–2. Results supported the concurrent valid-
ity among EM-based SCORS variables (i.e., COM and AFF)
and similar MMPI–2 scales of psychological functioning (i.e.,
Ego Strength and Anger).

Research has also been conducted utilizing SCORS rat-

ings of Thematic Apperception Test (TAT; Murray, 1943)
narratives in outpatients diagnosed with BPD, antisocial per-
sonality disorder, narcissistic personality disorder (NPD),
and Cluster C personality disorder groups (Ackerman,
Clemence, Weatherhill, & Hilsenroth, 1999). Significant dif-
ferences were found between the four groups across all eight
SCORS variables. The BPD group scored significantly lower
(i.e., more pathological) than the NPD group on all eight
SCORS variables. The BPD patient scores on AFF, Emo-
tional Investment in Moral Standards (EIM), Experience and
Management of Aggressive Impulses (AGG) as well as Iden-
tity and Coherence of Self (ICS) variables were all signifi-
cantly lower than the Cluster C personality disorder group.
Finally, the SCORS variable Self-Esteem (SE) had a signifi-
cant relationship with the MMPI–2 BPD Nonoverlapping
scale (in which items are found only on the BPD scale;
Colligan, Morey, & Offord, 1994; Morey et al., 1985). These
MMPI–2 items were most related to BPD. Therefore, these

82

STEIN, PINSKER-ASPEN, HILSENROTH

results also support the concurrent validity among narrative
based SCORS variables and a self-report scale specifically
designed to assess borderline psychopathology.

Based on prior research, we hypothesized that patients

with DSM (4th ed. [DSM–IV], American Psychiatric Associ-
ation, 1994) BP would score higher on the PAI BOR scales
than individuals without BP. In addition, we anticipated that
borderline psychopathology as measured by the PAI would
have a significant, negative relationship with three SCORS
variables: COM, AFF, and SE. Specifically, we hypothe-
sized that the greater borderline psychopathology on the PAI
would be related to lower ratings on the COM, AFF, and SE
variables of the SCORS. Such negative relationships were
expected, as higher scores on the PAI indicate greater pathol-
ogy, whereas lower scores on the SCORS indicate more
maladaptive responses.

METHOD

Participants

Participants included 58 patients at a university-based outpa-
tient treatment clinic (for additional methodological or proce-
dural information, please see Hilsenroth, 2002). As can be ob-
served in Table 1, patients were predominantly female and
single. The mean age for the sample was 30 years (SD = 10).
This sample consisted of primarily patients with mood disor-
ders with relational problems manifested in either Axis II or
subclinical traits/features of Axis II. Of these patients, 20 were
diagnosed with BP; 9 met full (five or more individual BPD
criteria) criteria for DSM–IV (American Psychiatric Associa-
tion, 1994) BPD; 10 exhibited prominent traits/features (i.e.,
three or four individual criteria) of BPD as part of a personality
disorder not otherwise specified (PD NOS) diagnosis in which
these BPD symptoms were seen as the primary personality pa-
thology (see Grillo et al., 2005; Gunderson et al., 2000;
Zimmerman, Rothschild, & Chelminski, 2005 for similar pro-
cedures); and 1 patient exhibited prominent subclinical
traits/features (i.e., three individual criteria) of BPD. Of the re-
maining 38 patients, 11 met DSM–IV criteria for an individual
or PD NOS Axis II diagnosis, and 9 patients exhibited
subclinical traits/features (i.e., three or four individual crite-
ria) from another, non-BPD, personality disorder. The range of
DSM–IV Axis I and II diagnoses can be seen in Table 1. Finally,
3 patients were not included in the PAI analyses, as they had
completed an EM protocol but did not complete the PAI, bring-
ing our sample used for PAI analyses to 55. One of the patients
who did not complete the PAI met full diagnostic criteria for
BPD and 2 met criteria for PD NOS (primary BPD features).

Procedure

The clinicians who conducted the psychological assessment
(i.e., clinical interview, EM protocol, etc.) and psychother-
apy sessions were 18 advanced doctoral students (6 male

and 12 female) enrolled in an American Psychological As-
sociation

approved

Clinical

PhD

program.

The

psychological assessment consisted of four steps including
three meetings between the patient and clinician and one
patient appointment to complete a battery of self-report
measures (including the PAI). The three meetings included
(a) a semistructured diagnostic interview and EM protocol,
(b) interview follow-up, and (c) a collaborative feedback
session (for more details on the assessment process, see
Hilsenroth, Peters, & Ackerman, 2004). Interviewers in-
quired about and assigned DSM–IV ratings for the Axis II
disorder diagnosis under consideration when Axis II disor-
ders or significant subclinical trait/features appeared to be
present (i.e., three or four individual criteria for a given
personality disorder). After the clinical interview, each par-
ticipant received an interpretive/feedback session lasting 1
to 1.5 hr, also videotaped and organized according to a
therapeutic model of assessment (Finn & Tonsager, 1992,
1997; Fischer, 1994).

PAI AND BPD

83

TABLE 1

Demographic Information and Descriptive

Data for Study Variables

Variable

N

%

M

SD

Gender

Male

17

30

Female

41

70

Age (years)

30

10

Marital status

Single

33

58

Married

10

17.5

Divorced

13

23

Widowed

1

1.5

Primary Axis I diagnosis

Adjustment disorder

7

12

Anxiety disorder

6

10

Eating disorder

1

2

Mood disorder

37

65

Substance related disorder

1

2

V code relational problem

4

7

None

1

2

Axis II diagnosis

30

53

Axis II trait/features

10

17

Axis V GAF

60

6

PAI BOR variables (T scores)

BOR–Total

63

11

BOR–A

62

11

BOR–I

65

8

BOR–N

62

12

BOR–S

53

9

SCORS Variables

COM

3.21

.32

AFF

3.88

.44

SE

3.71

.30

Note.

N = 58. GAF = Global Assessment of Functioning scale; PAI =

Personality Assessment Inventory; BOR = Borderline scale; BOR–Total =
BOR Total score; BOR–A = Affective Instability; BOR–I = Identity
Problems; BOR–N = Negative Relationships; BOR–S = Self-Harm; COM =
complexity of representations; AFF = affective quality of representations;
SE = self-esteem.

Ratings of Axis II pathology were based on information

gathered during the semistructured interview and feedback
sessions. Ratings provided by the clinician were based on the
patient’s level of relational functioning at the time of evalua-
tion (i.e., semistructured interview and feedback). External
raters then independently rated the Axis II pathology for each
participant immediately after viewing videotapes of the clini-
cal interview and feedback sessions. For all cases, scoring of
the measures by the external raters was completed independ-
ent of the clinician’s ratings.

Assessment Measures

The PAI.

The PAI is a 344 item self-report measure of

personality in which examinees select the response that best
pertains to them. Responses to be endorsed are either
False/not at all true, Slightly True, Mainly True, or Very
True. We used the BOR raw score (BOR-Total) and subse-
quent BOR subscales in this study (e.g., BOR–A, BOR–I,
BOR–N, and BOR–S). The PAI has been shown to be a reli-
able measure in assessing BP (Morey, 1991, 1996).

Early Memories Protocol.

The EM protocol used in

this study (Early Memories Protocol; Fowler et al., 1995) was
based on Mayman’s (1968) original work and further devel-
oped by Fowler et al. (1995) to include novel queries based on
object relations theory (Winnicott, 1971). The complete set of
eight queries includes earliest memory; second earliest mem-
ory; earliest memory of mother; earliest memory of father; ear-
liest memory of the first day of school; earliest memory of feel-
ing warm and snug; earliest memory of eating, feeding, or
being fed; and earliest memory of a special (transitional) ob-
ject. Obtaining multiple EMs allowed researchers to obtain a
broader and more representative sample of the patients’ rela-
tional experiences and object representations.

SCORS–global ratings.

SCORS–global

ratings

(Hilsenroth, Stein, et al., 2004; Westen, 1995) consists of
eight variables that are scored on a 7-point anchored scale in
which lower scores (e.g., 1, 2, or 3) indicate more pathologi-
cal responses and higher scores (e.g., 5, 6, or 7) indicate
healthy responses. The eight variables are COM, AFF, Emo-
tional Investment in Relationships, EIM, Understanding of
Social Causality, AGG, SE, and ICS. More thorough descrip-
tions of the eight SCORS variables, global rating method,
and various training examples are provided in the manuals

1

developed by Hilsenroth, Stein, et al. (2004) as well as
Westen (1995).

The SCORS was used to rate all EM narratives from each

patient. Raters (authors M. Stein & J. Pinsker-Aspen) were
two advanced graduate students enrolled in an American
Psychological Association approved Clinical Psychology

doctoral program. Prior to rating EM narratives in this study,
the two coders underwent supervised training in the use of
the SCORS manuals (Hilsenroth, Stein, et al., 2004; Westen,
1995) by rating several TAT and EM narratives (not from the
sample used in this study). The two raters were evaluated for
reliability on the SCORS prior to examining the EM narra-
tives used in this study. Each rater achieved a “good” (> .60)
to “excellent” (

≥

.75) level of interrater reliability (one-way

random effect model intraclass correlation coefficients [ICC]
Model 1 [ICC1]; Shrout & Fleiss, 1979) on all SCORS vari-
ables for these practice TAT and EM narratives. Following
the establishment of good interrater reliability on these
scales, judges (authors M. Stein & J. Pinsker-Aspen) began
rating EM narratives of patients in the research study. Both
judges rated all EMs reported by each patient on the SCORS,
and then we used an overall average across all EMs for each
patient in the analyses. We first transcribed each EM narra-
tive, and then we rated them in random order. For all cases,
scoring of EM narratives was completed independently by
the raters who were unaware of patient diagnosis, all other
assessment data (including the PAI), treatment variables, and
other raters’ scoring. We rated EMs reported by each patient
on the SCORS and then we used an overall average score
across all EMs for each patient in the analyses.

RESULTS

Interrater Reliability

We calculated ICC 1 and Spearman Brown corrected one-
way random effects model (1, 2). Shrout and Fleiss (1979) re-
ported the magnitude for interpreting ICC values in which
poor is < .40, fair = .40 to .59, good = .60 to .74, and excellent
=

≥

.74. The interrater reliability for the classification of per-

sonality pathology across three dimensions—(a) presence of
a personality disorder, (b) presence of subclinical traits/fea-
tures, and (c) absence of a personality disorder—was excel-
lent (i.e., ICCs > .74; Fleiss, 1981) for this project
(Hilsenroth et al., 2000; Peters, Hilsenroth, Eudell-Simmons,
Blagys, & Handler, 2006). One-way random effects model
ICC ([1,1]; Shrout & Fleiss, 1979) in this sample for Axis II
pathology was .85, and Spearman-Brown correction was .92
(ICC [1,2]; Shrout & Fleiss, 1979). In addition, interrater re-
liability for the presence or absence of BP was also quite high
(

κ

= .74).

In addition, we used ICCs to calculate reliability on EM

narratives for the two raters. ICC 1s fell in the excellent range
for AFF, good range for COM, and fair range for SE with rat-
ings of .83, .63, and .59, respectively. The Spearman Brown
ICC (1,2) corrected values fell in the excellent range for both
AFF and COM and in the good range for SE with scores of
.91, .83, and .74, respectively. In summary, the ratings of the
COM, AFF, and SE variables we used in this study were
highly reliable.

84

STEIN, PINSKER-ASPEN, HILSENROTH

1

For a copy of the manual please email this address hilsenro@

adelphi.edu.

Descriptive Data

The mean T scores for the PAI BOR scale and subscales are
given in Table 1. These scores reflect a sample that might be
“seen as moody, sensitive and having some uncertainty about
life goals” (Morey, 1991, p. 17; see Table 1 for subscale
scores). The means for the SCORS variables in Table 1 re-
flect a mild to moderate range of pathology (e.g. 3 to 4)
within the sample of EM narratives. This range of pathology
may manifest in less mature relationship patterns, more sim-
plistic representations of self and other as well as lower self-
esteem than scores in the 5 to 6 range (for additional informa-
tion regarding descriptive data, see Pinsker-Aspen, Stein, &
Hilsenroth, in press).

Criterion Validity

We used analysis of variance to compare mean T-score dif-
ferences between the BP (DSM–IV diagnoses; n = 17) and
non-BP (n = 38) groups. We calculated Cohen’s d using
pooled standard deviations to determine effect size (d
>.20–.50 = small, .50–.80 = medium, and >.80 = large; Co-
hen, 1988) for these group comparisons. Results revealed
that the PAI BOR scale was significantly higher for patients
diagnosed with BP (M = 67, SD = 10) than without that diag-
nosis (M = 60, SD = 11), F(1,53) = 5.65, p = .02, d = .70.
Please note that the non-BP group included some patients
with Axis II disorders as well as other non-BP Cluster B dis-
orders (i.e., antisocial personality disorder, NPD, histrionic
personality disorder). The BP group means and standard de-
viations for the BOR subscales were as follows: BOR–A M =
66, SD = 11; BOR–I M = 68, SD = 7, BOR–N M = 64, SD = 8;
and BOR–S M = 57, SD = 9. Comparatively, the non-BP
group means were as follows: BOR–A M = 60, SD = 9;
BOR–I M = 62, SD = 8; BOR–N M = 59, SD = 12; and
BOR–S M = 49, SD = 8. We found between-group differ-
ences on the following BOR subscales: BOR–I, F(1, 53) =
4.24, p = .04, d = .61 and BOR–S, F(1, 53) = 5.75, p = .02, d =
70. There was a trend toward significance with BOR–A, F(1,
53) = 3.05, p = .08, d = .51. However, BOR–N was not signif-
icant, F(1, 53) = 1.03, p = .31, d = .30. These significant dif-
ferences between the BP and non-BP groups revealed moder-
ate effect sizes, which demonstrated the validity of the PAI
even in the differentiation of outpatients with other Axis I
and II disorders from those with BP. With regard to overall
functioning, we compared Global Assessment of Func-
tioning (GAF) Scale scores for the BP (M = 58.25, SD = 4.8)
and non-BPD (M = 61.47, SD = 6.4) groups, and we found a
significant difference, F(1, 56) = 3.93, p = .05, d = .55; the
GAF for the BP group was lower. Please note that GAF rat-
ings were available for all 58 patients and not 55 as in other
analyses.

We conducted a diagnostic efficiency analysis to further

examine the clinical use of the PAI BOR scale in accurately
assessing DSM–IV BP (Streiner, 2003). The PAI BOR–Total

scale was dichotomized at

≥

70, and the subsequent overall

correct classification rate for BP patients was 73%. This sup-
ports the PAI clinical utility quite well given the general clin-
ical and substantial interpersonal distress of both these
groups.

We also calculated point biserial correlations to examine

the relationship between the presence (1) or absence (0) of
DSM–IV ratings of BP with the PAI BOR scales. The PAI
BOR scales; BOR–Total, BOR–I, and BOR–S were signifi-
cantly correlated with the DSM–IV diagnosis of BP. There
was a trend toward significance with BOR–A (see Table 2).
These results provide a different expression of the mean
comparisons we reported previously.

Concurrent Validity

We initially hypothesized that the COM, AFF, and SE vari-
ables of the SCORS would be significantly correlated with
the PAI BOR scale. However, we found that only the SCORS
variable COM demonstrated a significant relationship with
the BOR scale (see Table 2). We then examined whether
COM was related to any of the four PAI BOR subscales. We
found that there was a significant correlation with the
BOR–A. Trends toward significance were also noted for the
BOR–N and BOR–I. The negative correlations observed
were expected because higher scores on the PAI indicate
greater psychopathology, whereas lower ratings on the
SCORS indicate more maladaptive responses.

We also examined the relationship between word count

(i.e., response productivity) of the EM narratives and COM,
which was significant (r = .55, p < .0001). However, as can
be seen in Table 2, all the analyses that demonstrated a sig-
nificant relationship remained significant after the effects for
response productivity were controlled for.

PAI AND BPD

85

TABLE 2

Relationship of PAI BOR Scales With

the Presence of

DSM–IV BP and COM

BP

COM

PAI BOR Scale

r

pb

p

r

pr

p

BOR–Total

.31

.02

–.30

–.33

.02/.01

BOR–A

.23

.09

–.33

–.30

.01/.02

BOR–I

.27

.04

–.24

–.23

.07/.08

BOR–N

.14

.32

–.23

–.25

.08/.06

BOR–S

.31

.02

–.05

–.18

.71/.19

Note.

N = 55. PAI = Persoanlity Assessment Inventory; BOR = PAI

Borderline scale; DSM–IV = Diagnostic and Statistical Manual of Mental
Disorders (4th ed.); BP = borderline pathology; COM = complexity of
representations; BOR–Total = BOR Total score, BOR–A = Affective
Instability; BOR–I = Identity Problems; BOR–N = Negative Relationships;
BOR–S = Self-Harm. r

pb

= point biserial correlation for which the presence

of BP is given a 1 and absence of BP is given a 0; pr = partial correlations
examine the relationship between COM and the PAI BOR scales (e.g.,
BOR–Total, BOR–A, BOR–I, BOR–N, BOR–S; raw scores), controlling for
the effects of word count.

DISCUSSION

The findings from this study indicate that the PAI BOR scale
is significantly associated with DSM–IV BPD pathology
such that patients with BP scored significantly higher on this
scale than did our clinical comparison sample. The presence
of patients with both Axis I and II disorders represents an ex-
tremely stringent comparison sample. Therefore, the PAI
BOR scale appears to be targeting BP specifically rather than
assessing global distress or general Axis II pathology. We
also found between-group differences on two PAI subscales,
BOR–I and BOR–S, and BOR–A trended toward signifi-
cance. These findings are also noteworthy, as even with a
comparison sample of Axis II patients (including other Clus-
ter B disorders), individuals in this study with BP experi-
enced greater difficulty with identity and self-harm than did
the participants in the comparison sample and a trend for
greater affective instability. BOR–N was not significant, per-
haps because the non-BP group had a substantial number of
Axis II disorders included, and individuals with Axis II dis-
orders by definition have difficulties in interpersonal relat-
ing. In addition, the overall correct classification for the pres-
ence or absence of BP was 73%, which is similar to Bell-
Pringle et al.’s (1997) findings (80%). In sum, participants
with a DSM–IV diagnosis of BP scored higher on several PAI
BOR scales than did non-BP participants, which further
supports the usefulness of the PAI in assessing borderline
specific pathology.

Even in comparison with other clinical outpatients, our

findings indicate that BP patients have more difficulty in the
areas of affective regulation (BOR–A), identity development
(BOR–I), and impulsivity (BOR–S), which thus provided
additional support for the prominent role of emotional
dysregulation, identity disturbance, and impulsivity in BP.
The ability to differentiate BP from non-BP patients, includ-
ing patients with other Axis II disorders other than BP, is ex-
tremely important for clinical assessment, and these results
demonstrate the applied clinical utility of the PAI in outpa-
tient settings.

Our study supports past research in that EM narratives can

be reliably rated using the SCORS. The SCORS variable
COM demonstrated a significant, negative relationship with
the PAI BOR scale. One way to interpret this finding is that
the individuals who scored higher on BOR often split repre-
sentations, which makes it difficult to see both a positive and
negative aspect of a person or situation. These types of narra-
tives would be rated more maladaptively on the SCORS, as
this splitting of good and bad objects severely impacts the
perception of interpersonal environments. COM also demon-
strated a significant, negative relationship with the BOR–A
subscale. This finding suggests that patients with higher
scores on this scale experienced less integrated affect in rela-
tion to themselves and others in addition to having a greater
tendency toward experiencing emotions as sudden, intense,
and short lived. These findings regarding COM support past

research (Ackerman et al., 1999; Fonagy et al., 1996; Fowler
et al., 1995; Nigg et al., 1992) that has demonstrated a rela-
tionship between borderline psychopathology and the inabil-
ity to elaborate on the self and other from patient narrative
data.

We also found some limited relationships between the

SCORS variable COM and the PAI BOR–N and BOR–I
subscales. This trend toward significance with BOR–N and
BOR–I suggests individuals who scored lower on COM
might have experienced themselves and others negatively or
even damaged in some way. Other research on EM narratives
has supported BPD patients’ tendency to view significant
others in their world as less approving, unhelpful, disinter-
ested, and abusive as well as having an increased negative
self-appraisal (Frank & Paris, 1981; Nigg et al., 1992; Arnow
& Harrison, 1991).

A limitation of the study is that some meaningful relation-

ships might have been missed due to limited statistical
power. It would be useful to replicate findings with a larger
sample to see if additional relationships exist. However, even
with the study’s limited power, this research contributes to
the BP literature in multiple ways. Most notable is the use of
a treatment-seeking naturalistic comparison group with Axis
I and II psychopathology. This non-BP clinical comparison
group allows for greater generalizability of the findings to
varied outpatient settings. This is important, as contempo-
rary research has indicated that Axis I problems are often sig-
nificantly intertwined with Axis II personality processes
(Westen & Arkowitz-Westen, 1998; Westen, Novotny, &
Thompson-Brenner, 2004). As such, our naturalistic sample
lends greater credence to findings that the PAI BOR scales
can effectively differentiate BP individuals from other treat-
ment-seeking individuals.

CLINICAL EXAMPLES

To provide a more clinically applied example of the findings
from this study, we now examine PAI data in relation to EMs
of the first day of school from two patients: one with and one
without BP. Fowler et al. (1995) reported that the memory of
the first day of school “assesses the degree to which separa-
tion-individuation

from

mother

and

family

remains

conflictual” (p. 81). This is particularly salient for individu-
als with BPD, as they often struggle between separation indi-
viduation and merger of their interpersonal conflicts that re-
volve around these two extremes. BPD patients also typically
exhibit impairment in their ability to modulate their level of
connectedness with others, and interpersonal events are often
experienced as being “too distant” or “too engulfing.” Mean
PAI BOR T scores as well as SCORS ratings (mean of the
two raters) are presented in Table 3 for each patient.

Patient B was a 31-year-old single woman diagnosed with

substance abuse disorder and PD NOS with four DSM–IV
BPD symptoms (2, 4, 6, and 8; American Psychiatric Associ-

86

STEIN, PINSKER-ASPEN, HILSENROTH

ation, 1994, p. 654). Her EM of the first day of school was the
following:

I went to the Montessori school and remember being dropped
in front of the school in front of these big, intimidating doors.
I remember saying, “Hey” in order to get one of the Nun’s at-
tention because I could not remember their names. The Nun
humiliated me and replied, “Hay is for horses” and did not
address me after that. I remember feeling humiliated after the
incident.

In this narrative, there is limited complexity, as internal

states are minimally elaborated and simplistic. The basic
theme of the narrative can be shortened to “I attempted to
get someone to help me and was humiliated.” The average
(of the two judges) SCORS rating for COM was 3. In fact,
six out of the eight SCORS variables had a mean of 3,
which falls in the low-to-moderate functioning range. This
woman’s PAI scores on the BOR–Total and BOR–A
subscale were elevated, with T scores of 74 and 85, respec-

tively, which is 2 to 3½ SDs above the normative mean.
The other BOR subscales, BOR–I, BOR–N, and BOR–S,
were 62, 66, and 64, respectively, which were all over 1 SD
above the normative mean (refer to Morey, 1991, p. 17, for
interpretation of T scores). In summary, the preceding
SCORS variables and PAI scales suggest an individual who
is sensitive to issues of separation, distance, and closeness.
She may perceive such relational issues as bad, negative,
harmful, and rejecting and engage in highly conflictual
struggles to modulate the amount of contact she desires in
relationships (e.g., distancing vs. merger). This may leave
Patient B with feelings of inadequacy, helplessness, uncer-
tainty, low self-worth, sudden shifts in mood, and high re-
activity to others.

Patient D was a 48-year-old divorced woman with MDD.

She did not have any Axis II diagnosis, traits, or features. Her
earliest memory of the first day of school was the following:

I was in kindergarten and I wet my pants. We were learning
colors and I must have been having trouble with the colors.
My teacher, Miss Anderson, had paper with balloons and she
wanted to test me and asked me to point to the red balloon. I
got it right. Then she lined us up to take us to the bathroom.
We all lined up, went to the bathroom and then went back to
class. I went into the bathroom and giggled, never went to the
bathroom. Then back in the classroom I had to go, the teacher
said, “No.” I wet my pants, but no one knew it. I sat with my
skirt spread out around me so no one could tell. I was embar-
rassed. Everyone got up to leave and I just sat there. Miss An-
derson asked why I was sitting there. I told her I wet my
pants. She told me to stay there. She went to get me clothes
and helped me out. She was real sensitive to my embarrass-
ment and protected me from the other kids.

One can see in Patient D’s EM that there are varying internal

states of the self and the other. There is a nuanced representa-
tion of the other, and there is a range of negative and positive
representations in this narrative. The COM SCORS variable is
also higher (4.5), which again indicates a more adaptive re-
sponse. In this example, the eight SCORS variables ranged
from 3.5 to 4.5. The means of five of the eight variables were
either 4 or 4.5, which is again reflective of a healthier narrative.
All of the PAI BOR scales were also well in the normative
range, which is indicative of less BPD pathology. The
BOR–Total score and subscales BOR–A, BOR–I, BOR–N,
and BOR–S were all in the average range, with T scores of 51,
57, 50, 43, and 48, respectively (see Table 3).

It is evident in these two examples how both the PAI and

EM narratives are both quantitatively and qualitatively dif-
ferent from individuals who are diagnosed with and without
borderline psychopathology. This was consistent across both
the PAI BOR scales and SCORS variables. These findings
provide ideographic examples that support this and prior re-
search that has demonstrated individuals with higher BP
have more difficulty integrating positive and negative as-
pects of the self and other in their environment.

PAI AND BPD

87

TABLE 3

PAI T Scores and SCORS Variables

of Patients With and Without Borderline

Pathology

PAI Scale

T Score

SCORS

Variables

M

a

Patient B

b

with BP

BOR-Total

74

Complexity

3.0

Affect

3.0

BOR-A

85

Relationships

3.0

BOR-I

62

Morals

3.5

BOR-N

66

Causality

3.0

BOR-S

64

Aggression

3.0

Self-Esteem

3.0

Identity

4.0

Patient D

c

with no BP

BOR-Total

51

Complexity

4.5

Affect

4.5

BOR-A

57

Relationships

4.5

BOR-I

50

Morals

3.5

BOR-N

43

Causality

4.0

BOR-S

48

Aggression

3.5

Self-Esteem

3.5

Identity

4.5

Note.

PAI = Personality Assesment Inventory; SCORS = Social Cognition

and Object Relations Scale; BOR = PAI Borderline scale; BOR-Total = BOT
Total score; BOR-A = Affective Instability; BOR-I = Identity Problems;
BOR-N = Negative Relationships; BOR-S = Self-Harm; Complexity =
Complexity

of

Representation;

Affect

=

Affective

Quality

of

Representations; Relationships = Emotional Investment in Relationships;
Morals = Emotional Investment in Values and Moral Standards; Causality =
Understanding of Social Causality; Aggression = Experience and
Management of Aggressive Impulses; Identity = Identity and Coherence of
Self.

a

Mean SCORS ratings for earliest memory of the first day of school across

two raters.

b

Patient was a 31-year-old single woman with a diagnosis of

substance abuse disorder and personality disorder not otherwise specified
with four borderline personality disorder symptoms (2, 4, 6, and 8).

c

Patient

was a 48-year-old divorced womean with a diagnosis of major depressive
disorder.

In addition, the full range of findings from this study sup-

ports the conclusions of past research that have demonstrated
BPD patients tend to provide narratives that are seen as more
malevolent as well as having more difficulties identifying
and elaborating on the self or other in an integrated manner
(Ackerman et al., 1999; & Fonagy et al., 1996; Nigg et al.,
1992). Blais, Hilsenroth, Fowler, and Conboy (1999) identi-
fied that BPD pathology is “related to greater disruptions in
object relations, more primitive defensive functioning (i.e.,
splitting), and the presence of raw primitive aggressive im-
pulses (p. 568)” with regard to free response and narrative as-
sessment data. Furthermore, the results of this study
demonstrate that explicit measures (i.e., PAI) can be used in
conjunction with implicit measures (i.e., EMs) to investigate
BP (e.g., Blais, Hilsenroth, Castlebury, Fowler, & Baity,
2001). This can be useful in treatment with respect to under-
standing how the patients come to perceive their inner world
as well as those around them (e.g., therapist). Recently, Blais
and Bistis (2004) summarized that implicit measures of per-
sonality used in research are “surprisingly consistent with re-
gard to the object relations of borderline patients” (p. 492). In
addition, Blais and Bistis noted that systematically reviewing
implicit and explicit data across multiple measures will help
the clinician identify BP and better describe behaviors that
are severely impacting their performance outside of the test-
ing room. Last, examining individual patient data may aid
clinicians in better identifying the severity of the patient’s
condition and in setting appropriate and realistic goals re-
garding treatment. McClelland, Koestner, and Weinberger
(1989) further posited that explicit measures (self-attributed
measures)

Involve analytic thought in the sense of people making com-
plex judgments as to the degree to which certain statements
apply to them. … [I]mplicit measures are coded to reflect
emotional and motivational themes in the person’s life
unevaluated as to their appropriateness in terms of concepts
of the self, others and what is important. (p. 698)

Therefore, the findings of this study support the use of both
implicit and explicit measures of personality to provide a
comprehensive view of the patient’s level of object represen-
tations. Additional research should continue to examine the
relationship between the PAI scales and implicit measures of
personality. Such research would serve to increase the clini-
cal utility of this measure and inform psychotherapy.

REFERENCES

Ackerman, S., Clemence, J., Weatherhill, R., & Hilsenroth, M. (1999). Use

of the TAT in the assessment of DSM–IV Cluster B personality disorders.
Journal of Personality Assessment, 73, 422–448.

Acklin, M., Bibb, J., Boyer, P., & Jain, V. (1991). Early memories as expres-

sions of relationship paradigms: A preliminary investigation. Journal of
Personality Assessment, 57, 177–192.

American Psychiatric Association. (1987). Diagnostic and statistical man-

ual of mental disorders (3rd ed., rev.). Washington, DC: Author.

American Psychiatric Association. (1994). Diagnostic and statistical man-

ual of mental disorders (4th ed.). Washington, DC: Author.

Arnow, D., & Harrison, R. (1991). Affect in early memories of borderline

patients. Journal of Personality Assessment, 56, 75–83.

Bell-Pringle, V. J., Pate, J. L., & Brown, R. C. (1997). Assessment of border-

line personality disorder using the MMPI–2 and the Personality Assess-
ment Inventory. Assessment 4, 131–139.

Blais, M., & Bistis, K. (2004). Projective assessment of borderline

psychopathology. In M. J. Hilsenroth & D. L. Segal (Eds.), Comprehen-
sive handbook of psychological assessment (Vol. 2, pp. 485–499).
Hoboken, NJ: Wiley.

Blais, M. A., Hilsenroth, M. J., Castlebury, F., Fowler, J. C., & Baity, M. R.

(2001). Predicting DSM–IV Cluster B personality disorder criteria from
MMPI–2 and Rorschach data: A test of incremental validity. Journal of
Personality Assessment, 76, 150–168.

Blais, M. A., Hilsenroth, M. J., Fowler, J. C., & Conboy, C. A. (1999). A

Rorschach exploration of the DSM–IV borderline personality disorder.
Journal of Clinical Psychology, 55, 563–572.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd

ed.). New York: Academic.

Colligan, R. C., Morey, L. C., & Offord, K. P. (1994). The MMPI/MMPI–2

personality disorder scales: Contemporary norms for adults and adoles-
cents. Journal of Clinical Psychology, 50, 168–200.

Derogatis, L. (1983). SCL–90: Administration, scoring & procedures man-

ual for the revised version. Baltimore: Johns Hopkins University Press.

Finn, S. E., & Tonsager, M. E. (1992). Therapeutic effects of providing

MMPI–2 test feedback to college students awaiting therapy. Psychologi-
cal Assessment, 4, 278–287.

Finn, S. E., & Tonsager, M. E. (1997). Information-gathering and therapeu-

tic models of assessment: Complementary paradigms. Psychological As-
sessment, 9, 374–385.

Fischer, C. T. (1994). Individualized psychological assessment. Hillsdale,

NJ: Lawrence Erlbaum Associates, Inc.

Fleiss, J. L. (1981). Statistical methods for rates and proportions (2nd ed.).

New York: Wiley.

Fonagy, P., Leigh, T., Steele, M., Steele, H., Kennedy, R., & Mattoon, G.

(1996). The relations of attachment status, psychiatric classification, and
response to psychotherapy. Journal of Consulting and Clinical Psychol-
ogy, 64, 22–31.

Fonagy, P., Moran, G. S., Steele, M., Steele, H., & Higgitt, A. C. (1991). The

capacity for understanding mental states: The reflective self in parent and
child and its significance for security of attachment. Infant Mental Health
Journal, 13, 200–216.

Fowler, C., Hilsenroth, M., & Handler, L. (1995). Early memories: An ex-

ploration of theoretically derived queries and their clinical utility. Bulletin
of the Menninger Clinic, 59, 79–98.

Frank, H., & Paris, J. (1981). Recollections of family experience in border-

line patients. Archives of General Psychiatry, 38, 1031–1034.

Grilo, C., Sanislow, C. A., Shea, M. T., Skodol, A. E., Stout, R. L.,

Gunderson, J. G., et al. (2005). Two-year prospective naturalistic study of
remission from major depressive disorder as a function of personality dis-
order co-morbidity. Journal of Consulting and Clinical Psychology, 73,
78–85.

Gunderson, J., Shea, T., Skodol, A., McGlashan, T., Morey, L., Stout, R., et

al. (2000). The collaborative longitudinal personality disorders study: De-
velopment, aims, design, and sample characteristics. Journal of Personal-
ity Disorders, 14, 300–315.

Hathaway, S. R., & McKinley, J. C. (1967). Manual for the Minnesota

Multiphasic Personality Inventory. New York: Psychological Corpora-
tion.

Hathaway, S. R., & McKinley, J. C. (1989). Minnesota Multiphasic Person-

ality Inventory–2. Minneapolis: University of Minnesota Press.

Hilsenroth, M. J. (2002). Adelphi University: Psychotherapy process and

outcome research team. In P. Fonagy, J. Clarkin, A. Gerber, H. Kachele,

88

STEIN, PINSKER-ASPEN, HILSENROTH

R. Krause, E. Jones, R. et al. (Eds.), An open door review of outcome stud-
ies in psychoanalysis (2nd ed., pp. 241–247). London: International Psy-
choanalytic Association.

Hilsenroth, M. J., Peters, E. J., & Ackerman, S. J. (2004). The development

of therapeutic alliance during psychological assessment: Patient and ther-
apist perspectives across treatment. Journal of Personality Assessment,
83, 332–344.

Hilsenroth, M., Stein, M., & Pinsker, J. (2004). Social Cognition and Object

Relations Scale: Global rating method (SCORS–G). Unpublished manu-
script, The Derner Institute of Advanced Psychological Studies, Adelphi
University, Garden City, NY.

Kurtz, J., & Morey, L. (2001). Use of structured self-report assessment to di-

agnose borderline personality disorder during major depressive episodes.
Assessment, 8, 291–300.

Kurtz, J., Morey, L., & Tomarken, A. (1993). The concurrent validity of

three self-report measures of borderline personality. Journal of
Psychopathology and Behavioral Assessment, 15, 255–266.

Levine, J., Green, C., & Millon, T. (1986). The separation-individuation test

of adolescence. Journal of Personality Assessment, 50, 123–137.

Mayman, M. (1968). Early memories and character structure. Journal of

Projective Techniques and Personality Assessment, 32, 303–316.

McClelland, D. C., Koestner, R., & Weinberger, J. (1989). How do self-

attributed and implicit motives differ? Psychological Review, 96,
690–702.

McNair, D., Lorr, M., & Droppleman, L. (1971). Manual for the Profile of

Mood States. San Diego, CA: Educational and Industrial Testing Service.

Morey, L. (1991). Personality Assessment Inventory professional manual.

Odessa, FL: Psychological Assessment Resources.

Morey, L. (1996). An interpretative guide to the Personality Assessment In-

ventory (PAI). Odessa, FL: Psychological Assessment Resources.

Morey, L., Waugh, M., & Blashfield, R. (1985). MMPI scales for the

DSM–III personality disorders: Their derivation and correlates. Journal of
Personality Assessment, 49, 245–251.

Murray, H. A. (1943). Thematic Apperception Test: Manual. Cambridge,

MA: Harvard University Press.

Nigg, J., Lohr, N., Westen, D., Gold, L., & Silk, K. (1992). Malevolent ob-

ject representations in borderline personality disorder and major depres-
sion. Journal of Abnormal Psychology, 101, 61–67.

Peters, E., Hilsenroth, M., Eudell-Simmons, E., Blagys, M., & Handler, L.

(2006). Reliability and validity of the Social Cognition and Object Rela-

tions Scale (SCORS) in clinical use. Psychotherapy Research, 16,
617–626.

Pinsker-Aspen, J. H., Stein, M. B., & Hilsenroth, M. J. (in press). The clini-

cal utility of early memories as a predictor of therapeutic alliance. Psycho-
therapy.

Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations: Uses in assessing

rater reliability. Psychological Bulletin, 86, 420–428.

Streiner, D. (2003). Diagnosing tests: Using and misusing diagnostic and

screening tests. Journal of Personality Assessment, 81, 209–219.

Trull, T. (1995). Borderline personality disorder features in nonclinical

young adults: 1. Identification and validation. Psychological Assessment,
7, 33–41.

Westen, D. (1995). Social Cognition and Object Relations Scale: Q-sort for

projective stories (SCORS–Q). Unpublished manuscript, Department of
Psychiatry, Cambridge Hospital and Harvard Medical School, Cam-
bridge, MA.

Westen, D., & Arkowitz-Westen, L. (1998). Limitations of Axis II in diag-

nosing personality pathology in clinical practice. American Journal of
Psychiatry, 155, 1767–1771.

Westen, D., Novotny, C., & Thompson-Brenner, H. (2004). The empirical

status of empirically supported psychotherapies: Assumptions, findings,
and reporting in controlled clinical trials. Psychological Bulletin, 130,
631–663.

Winnicott, D. W. (1971). Playing and reality. New York: Routledge.
Zimmerman, M., Rothschild, L., & Chelminski, I. (2005). The prevalence of

DSM–IV personality disorders in psychiatric outpatients. American Jour-
nal of Psychiatry, 162, 1911–1918.

Michelle Stein
Derner Institute of Advanced Psychological Studies
Adelphi University
158 Cambridge Avenue
Garden City, New York 11530
Email:[e-mail]

Received June 24, 2005
Revised May 12, 2006

PAI AND BPD

89