Psychological Assessment: Copyright 199] by the American Psychological Association, Inc.

A Journal ol Consulting and Clinical Psychology 1040-3590/91/$3.00

1991. Vol. 3, No. 2, 232-238

Discriminant Validity of the MMPI-Borderline Personality Disorder Scale

Timothy J. Trull
University of Missouri-Columbia

The discriminant validity of the Minnesota Multiphasic Personality Inventory (MMPI) Borderline
Personality Disorder scale (MMPI-BPD) was investigated in a sample of psychiatric inpatients by
comparing the MMP1-BPD scores of a criterion group of patients who received a discharge diag-
nosis of borderline personality disorder (BPD) with the scores of several other DSM-III-R diagnos-
tic groups that did not overlap with the BPD group. Results indicated that the MMPI-BPD scale
scores discriminated the BPD group from an "other personality disorders" comparison group and
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

from a schizophrenia-schizoafi'ective group. MMPI-BPD scores of 2 mood-disorder comparison

This document is copyrighted by the American Psychological Association or one of its allied publishers.

groups, however, were not significantly different from those of the BPD criterion group. The
discriminative ability of the MMPI-BPD scale was compared with that of individual MMPI clini-
cal scales as well as several MMPI codetypes. Implications for the development of scales to opti-
mize the differential diagnosis of BPD are discussed.

Borderline personality disorder (BPD) is the most frequent ual of Mental Disorders, third ed ition (DSM-III; American Psy-
personality disorder diagnosis made in both outpatient and in- chiatric Association, 1980) were introduced by Morey, Waugh,
patient settings (Widiger & Trull, in press). As such, there is a and Blashfield (1985). The MMPI-BPD scale consists of 22
great need for a self-report instrument that is a reliable and true-false items that were judged to represent DSM-III criteria
valid indicator of BPD to aid in assessment. Although a num- for BPD and that discriminated between high- and low-scorers
ber of studies have compared Minnesota Multiphasic Personal- on the total scale score. Several studies have demonstrated that
ity Inventory (MMPI; Hathaway & McKinley, 1983) clinical the MMPI-BPD scale discriminates between patients diag-
scale scores of patients diagnosed with BPD with scores of other nosed with BPD and patients receiving other personality dis-
psychiatric patients, there is at present no consensus as to an order diagnoses (e.g., Dubro & Wetzler, 1989; Morey, Blashfield,
MMPI codetype specific to BPD (Gartner, Hurt, & Gartner, Webb, & Jewell, 1988).
1989; Morey & Smith, 1988; Widiger, Sanderson, & Warner, The discrimination of BPD from other personality disorders
1986). For example, Morey and Smith (1988) reviewed 12 stud- is of substantial interest (Morey & Smith, 1988). However, the
ies that reported MMPI scale scores for BPD subjects and con- MMPI-BPD scale's ability to discriminate between BPD and
cluded that the only two consistent findings were (a) an elevated
Axis I diagnostic groups has not yet been reported despite the
F scale, and (b) a high degree of overall profile elevation. Al-
fact that the overlap of BPD and Axis I diagnostic groups, espe-
though elevations on Scales 2, 4,6, 7, and 8 occur frequently in
cially mood disorders, has been the subject of much research
BPD samples, research suggests that these scales are sensitive
and controversy (e.g., Akiskal, Vferevanian, Davis, King, &
to, but not specific to, the BPD diagnosis. Perhaps this is not too
Lemmi, 1985; Gunderson & Elliot, 1985; Widiger, 1989). A
surprising because the original MMPI scales were developed in
number of researchers have posited a close relationship be-
the context of an older psychiatric classification system that
tween BPD and mood disorders. For example, Akiskal et al.
does not correspond well with contemporary psychiatric no-
(1985) reported that comparisons of phenomenology, biological
menclature. In sum, the search for a specific BPD-MMPI code-
markers, family history, and outcome between BPD and mood-
type has not, to date, been fruitful.
disordered groups suggest that BPD may in fact represent a
It is conceivable, however, that responses to MMPI items can
"subaflectivc" disorder, making distinctions between mood
reliably discriminate BPD patients (or patients diagnosed with
disorders and BPD difficult.
any of the other personality disorders) from other diagnostic
In addition to the clinical overlap between mood disorders
groups. With this in mind, MMPI scales to assess the 11 person-
ality disorders identified in the Diagnostic and Statistical Man- and BPD, differentiating these disorders on the basis of self-re-
port inventory scores may be difficult because state mood fac-
tors (especially depression) appear to have a substantial effect
Portions of this article were presented at the Annual Meeting of the on self-reports of personality (Hirschfeld et al, 1983; Reich et
American Psychological Association in Boston, Massachusetts, Au- al, 1986). Several researchers have suggested that acute mood
gust, 1990.
symptoms may cause psychiatric patients to overreport mal-
I thank Thomas A. Widiger, Meg A. Klein, Cynthia Sanderson, Pa-
adaptive personality traits. For example, Hirschfeld et al. (1983)
tricia Frazier, Eric Martin, and Kim Breaux for their help at various
stages of this project. In addition, I acknowledge the thoughtful com- assessed depressed patients before and after treatment and re-
mpnti and supppstinns nrnvirlpH hv tliri1^ annnvTYi/inc r<n/!*a\««»rc ported that self-ratings of personality changed significantly fol-
lowing the alleviation of depressive symptoms. Although de-
pressed, patients tended to score in a more pathological direc-
 MMPI-BPD SCALE 233

tion on ratings of personality features. These results were in BPD on Axis II (BC; n - 36); those receiving a primary Axis I diagnosis
contrast to a control group of depressed patients whose depres- of schizophrenia or schizoaffective disorder but not BPD on Axis II
sive symptoms persisted after treatment and whose self-ratings (SZ; n = 96); and those receiving a primary Axis II diagnosis of BPD
of personality remained fairly consistent. Finally, there is also (n = 61). The MMPI scale scores and MMPI-BPD scores of the Axis I
diagnostic groups were compared with those of the borderline group.
some evidence that self-reports of BPD symptoms may be state
In separate analyses, MMPI clinical scale scores and MMPI-BPD
dependent (Hurt et al, 1984; Piersma, 1987), potentially cloud-
scores of the BPD group were compared with those from a comparison
ing any distinction between BPD and mood disorders.
group composed of patients receiving a personality disorder diagnosis
In summary, results from previous studies raise the question other than BPD (OPD; n = 63). These latter analyses were conducted
of whether BPD patients and patients diagnosed with Axis I separately because approximately one half of the OPD patients were
disorders (especially mood disorders) can be discriminated on also included in the Axis I comparison groups. Diagnoses in this OPD
the basis of their self-reports on the MMPI. The present study group included nine non-BPD personality disorder diagnoses (pas-
addressed this question in three ways. First, the mean MMPI sive-aggressive personality disorder was not represented) as well as the
clinical scale scores of BPD inpatients were compared with DSM-III-R diagnosis of personality disorder not otherwise specified.
those scores from patients in one of several Axis I and Axis II
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

comparison groups. Second, the discriminant validity of sev-

This document is copyrighted by the American Psychological Association or one of its allied publishers.

eral MMPI codetypes previously identified as suggestive of a Results

BPD diagnosis was evaluated. Finally, the reliability and dis-
criminant validity of the MMPI-BPD scale were evaluated by Table 1 presents the demographic and diagnostic characteris-
(a) examining the internal consistency of the MMPI-BPD scale tics of the total sample. The mean age of the total sample was
31.37 years; 61% of the sample was female, 82% of the sample
in a large psychiatric sample, and (b) assessing this scale's con-
was White, and the median length of stay in the hospital was
vergent and discriminant validity in a sample of psychiatric
27.00 days. The comparison groups differed on a few of the
inpatients with various Axis I and Axis II diagnoses. In contrast
demographic variables. The BPD group had a significantly
to previous studies, the MMPI-BPD scores of clinically diag-
higher proportion of women than did all other comparison
nosed BPD patients were compared with the scores of several
groups (all p's < .05), and the MD group contained a signifi-
DSM-HI-R (American Psychiatric Association, 1987) diagnos-
cantly higher proportion of women than did the SZ group (p <
tic comparison groups: (a) major depression or dysthymia, (b)
bipolar disorder or cyclothymia, (c) schizophrenia or schizoaf- .05). Only two other significant differences were found: The
BPD group was significantly younger than the MD group (p <
fective disorder, and (d) other (non-BPD) personality disorders.
.05), and the SZ group had a longer length of stay in the hospital
than did the MD group (p < .05).
Method
Table 2 presents comparisons of the mean AT-corrected
The total sample comprised 395 consecutive adult psychiatric inpa- MMPI scale scores of the BPD group and the Axis I diagnostic
tients who were admitted to the acute-care units at a university medical groups. These groups were compared on each MMPI scale
center in a large metropolitan area, who completed the MMPI shortly through an analysis of variance (ANOVA), and significant find-
after admission (typically within 1 week), and who did not meet exclu- ings were followed up with post hoc Scheffe / tests for pairwise
sionary criteria for the study. Patients were excluded from the study if comparisons. Age was significantly related to scale scores on F,
they omitted more than 30 questionson the MMPI or were less than 18
years old. Discharge diagnoses were assigned to all patients according
to the diagnostic criteria of the DSM-III-R. Diagnoses were made by
the treating therapist after extensive consultation with a team of psychi-
Table 1
atrists, residents and interns, and psychologists. All team members
Demographic and Diagnostic Characteristics
had observed and interacted with the patient in question, and clini-
of Total Sample (N = 395)
cians at this institution have been trained extensively in the use of
DSM-fll-R. This method of diagnostic practice approximates the
%of
LEAD (Longitudinal Expert evaluation using All Data) standard of Variable n sample
diagnosis (Skodol, Rosnick, Kellman, Oldham, & Hyler, 1988; Spitzer,
1983). Previous comparisons of discharge diagnoses and diagnoses Age (years) M= 31 .37 (SD= 10.12)
resulting from systematic chart review at this hospital have shown that Sex
these two methods are comparable (Fyer, Frances, Sullivan, Hurt, & Male 156 39
Clarkin, 1988). Each MMPI was scored on the traditional clinical Female 239 61
scales as well as on the Morey et al. (1985) MMPI-BPD scale. Consis- Race
White 324 82
tent with previous studies, the 22-item (overlapping) version of the
Black 33 8
MMPI-BPD was used, and the raw MMPI-BPD scale scores were 7
Hispanic 29
used as the dependent variable in the analyses involving this scale. Other 9 3
Patient scores on the MMPI-BPD items were not known to the team Length of stay Mdn = 27.00 days
members because these scores were calculated by me after the patient Primary Axis 1 diagnosis
was discharged and were not presented in any of the psychological test Bipolar disorder or cyclothymia 39 10
reports. Major depression or dysthymia 87 22
From the total sample, several comparison groups were formed: Schizophrenia or schizoaffective 99 25
those receiving a primary Axis 1 diagnosis of major depression or Primary Axis II diagnosis
Borderline 61 15
dysthymia but not BPD on Axis II (MD; n = 70); those receiving a
Other personality disorder 63 16
primary Axis I diagnosis of bipolar disorder or cyclothymia but not
 234 TIMOTHY J. TRULL

4,6,8, and 9. Therefore, analyses for these scales used age as a codetype. Table 4 presents these results using only the patients
covariate. composing the BPD, MD, BC, SZ, and OPD comparison
Consistent with previous findings, mean MMPI scale scores groups (« = 298). The total number of subjects in these analyses
for the BPD group were 2-70T on the F scale as well as 6 of the is less than 328 because some patients in the OPD group also
clinical scales, indicating a high degree of overall profile eleva- received an Axis I diagnosis of MD or BC.
tion. Significant differences between groups were found for 8 of All codetypes had moderate sensitivity (SENS) and specific-
the 13 MMPI scales. With regard to the clinical scales, BPD ity (SPEC), low positive predictive power (PPP), and high nega-
patients scored significantly higher than did both the SZ and tive predictive power (NPP). Of particular interest are the low
BC patients on Scales 2, 3, 4, and higher than the BC patients PPP values, because this index is directly related to the diagnos-
did on Scale 0 (all p's < .05). In addition, patients diagnosed tic decision-making process that confronts the clinician when
with BPD scored significantly lower on Scale 5 than did SZ viewing an MMPI codetype. The last column in Table 4 pres-
patients. Interestingly, the MD and BPD patients did not obtain ents the odds ratio contrasting the prevalence of a BPD diagno-
significantly different scores on any MMPI clinical scale. In sis given the codetype with the prevalence of a BPD diagnosis in
summary, only 5 of the 10 MMPI clinical scales discriminated those not producing the MMPI codetype (Fleiss, 1981). The
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

between BPD patients and SZ or BC patients, and the BPD and odds ratio approximates relative risk and was calculated as a/b
This document is copyrighted by the American Psychological Association or one of its allied publishers.

MD groups were not distinguishable on any MMPI scale. divided by c/d, in which a = prevalence of BPD diagnosis in
Table 3 presents the comparisons between the BPD and the group with particular codetype, b = 1 — a, c = prevalence of
other personality disorders (OPD) group (i.e., Axis II). Age was BPD diagnosis in group without particular codetype, d= 1 — c
significantly related to scores on Scales 3 and 9, and therefore (Fleiss, 1981). An odds ratio of 1 would indicate no association
was used as a covariate in these respective analyses. Significant between a BPD diagnosis and a codetype, whereas an odds ratio
differences were found only for Scales K and 5 (both p's < .05), of two would indicate that the odds of a person producing a
with the OPD group scoring significantly higher on both of particular codetype and receiving a BPD diagnosis is twice that
these scales. of a person not producing the codetype. The statistical signifi-
In their review of MMPI research on BPD, Morey and Smith cance of each odds ratio was tested by computing a chi-squared
(1988) listed six codetypes that were frequently observed in statistic with one degree of freedom (Fleiss, 1981). As indi-
BPD patients. The sensitivity, specificity, positive predictive cated, no odds ratio was significant, suggesting that these code-
power (PPP), and negative predictive power (NPP) of these co- types were not specific indicators of the BPD diagnosis (i.e.,
detypes for the BPD diagnosis were examined in the present these codetypes were also prevalent among patients in the other
sample. Sensitivity is the proportion of BPD patients producing diagnostic groups).
the respective MMPI codetype; specificity is the proportion of Finally, the MMPI-BPD scores of the sample were consid-
non-BPD patients not producing the codetype; PPP is the con- ered. Using data from the total sample (n = 395), the internal
ditional probability of a BPD diagnosis given the MMPI code- consistency coefficient (KR-20) of the MMPI-BPD scale was
type; and NPP is the conditional probability of not receiving a calculated, and it equaled .69. This value is approximately equal
BPD diagnosis given that one does not produce the MMPI to that obtained by Morey et al. (1985) in the original validation

Table 2
K-correeled Minnesota Multiphasic Personality Inventory T Scores for Borderline Personality
Disorder (BPD), Major Depression or Dysthymia (MD), Bipolar Disorder or Cyclothymia (BC),
and Schizophrenia or Schizoaffective Disorder (SZ) Diagnostic Groups

BPD MD BC SZ
(n = 61) (« = 70) (n = 36) (n = 96)
Signifleant
Scale M SD M SD M SD M SD F contrasts

L 48.30 7.30 50.00 7.71 49.22 8.32 53.04 9.37 4.73** SZ > BPD
F 72.77 16.00 69.09 14.43 67.08 17.31 74.04 18.44 1.70
K 49.16 9.41 51.26 9.33 52.33 10.82 52.19 9.32 1.43
1 64.30 13.72 68.00 14.02 62.03 16.60 62.09 13.90 2.62
2 80.43 14.49 84.90 16,17 67.22 18.98 71.10 16.43 14.75*** MD, BPD > SZ, BC
3 70.05 11.35 72.50 11.95 61.56 13.84 63.88 11.16 11.14*** MD, BPD > SZ, BC
4 79.75 12.94 76.31 13.39 69.81 13.10 73.67 10.84 5.41*** BPD > SZ,BC
5 49.90 15.15 54.73 16.44 56.50 15.16 60.15 13.76 5.96*** SZ > BPD
6 73.15 11.87 72.87 12.38 68.39 13.59 71.70 15.10 1.23
7 72.34 13.03 77.73 14.06 65.78 17.83 70.45 14.87 6.05*** MD > SZ, BC
8 79.20 16.44 79.59 17.43 73.17 19.75 80.12 19.42 1.64
9 63.87 11.97 59.86 11.15 68.31 13.18 65.40 14.08 3.48* BC>MD
0 62.70 12.24 64.94 11.16 53.28 11.71 58.49 9.92 10.73*** MD, BPD >BC
MD>SZ

Note. Analyses of variance for Scales K4,6, 8, and 9 used age as a covariate (i.e., these were ANCOVAs).
*P<.05. **p<.0l. "*p<.0(n.
 MMPI-BPD SCALE 235

Table 3 6.82, p< .001. Post hoc Schefle tests revealed that only the BPD
K-correcled Minnesota Multiphasic Personality Inventory Scores and SZ groups differed significantly on MMPI BPD scores. A
for Borderline Personality Disorder (BPD) and Other (Non-BPD) separate comparison between BPD patients and OPD patients
Personality Disorder (OPD) Diagnostic Groups was made. An ANCOVA (with age as a covariate) was signifi-
cant, F(\, 121) = 6.55, p < .05, indicating that the BPD and
BPD OPD
OPD groups differed significantly on MMPI-BPD scores.
(n = 61) (n = 63)
As Morey and Smith (1988) have noted, one important aspect
Scale M SD M SD F of the MMPI-BPD scale that is in need of further evaluation is
that of appropriate cutoff points. T scores for the MMPI-BPD
L 48.30 7.30 50.49 7.74 2.64 scale have been developed using normative data from the origi-
F 72.77 16.00 70.86 15.19 0.47
K 49.16 9.41 53.29 9.67 5.78'
nal validation study. Previous studies (e.g., Dubro, Wetzler, &
1 64.30 13.72 68.63 16.21 2.58 Kahn, 1988) have used a T-score of 70 or more on the MMPI-
2 80.43 14.49 80.60 19.94 0.00 BPD (raw score > 15)to indicate the presence of BPD. In the
3 70.05 11.35 71.08 13.50 0.01 present study, using a cutoffof 707"resulted in a sensitivity rate
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

4 79.75 12.94 77.79 12.85 0.72

of .25, a specificity rate of .83, a PPP rate of .27, and an NPPrate
This document is copyrighted by the American Psychological Association or one of its allied publishers.

5 49.90 15.15 60.00 14.86 14.04"*

6 73.15 11.87 71.27 13.30 0.69 of .91. The odds of producing an MMPI-BPD score at or above
7 72.34 13.03 75.48 18.19 1.21 70r and receiving a BPD diagnosis was 1.61. Comparing these
8 79.20 16.44 80.16 19.21 0.09 diagnostic efficiency rates to those presented in Table 4 reveals
9 63.87 11.97 64.92 10.66 0.98 that the MMPI-BPD scale appears to be a more conservative
0 62.70 12.24 59.84 13.06 1.59
indicator of a borderline diagnosis than the MMPI codetypes.
Note. Analyses of variance for Scales 3 and 9 used age as a covariate The MMPI-BPD is a much less sensitive but a more specific
(i.e, these were ANCOVA's). measure of a borderline personality disorder diagnosis. False
*p<.05. "/[Link]. positives were minimized at the expense of false negatives.
A final analysis involved examining cutoff scores on the
MMPI-BPD scale with regard to sensitivity, specificity, and
study (KR-20 =.71). The mean score on the MMPI-BPD scale agreement with clinical diagnosis. Table 6 presents these data.
for the entire sample (n = 395) was 10.67 (SD = 3.83). As for As can be seen, a cutoff score of 13 on the MMPI-BPD resulted
correlates of the MMPI-BPD scores, only age covaried signifi- in the highest diagnostic agreement (i.e_ kappa) between the
cantly (/•=-. 18, p < .01); in general, MMPI-BPD scores were MMPI-BPD and clinical diagnosis. A cutoffof 13 versus 15 (i.e.,
higher for younger psychiatric inpatients. Because of this signifi- >70T) resulted in higher sensitivity, lower specificity, and higher
cant association, age was used as a covariate in analyses involv- diagnostic agreement. However, the overall rate of diagnostic
ing MMPI-BPD scores. The mean MMPI-BPD score for each agreement (kappa = .19) for a cutoffof 13 was quite low.
of the comparison groups is presented in Table 5. Also pre-
sented are results of the respective analyses of covariance (AN-
Discussion
COVAs) and the results of post hoc Scheffe tests. An ANCOVA
(with age as a covariate) comparing the MMPI-BPD scores of In this study, BPD patients obtained significantly higher
the MD, BC, SZ, and BPD groups was significant, F(3, 258) = MMPI scores than did the schizophrenia/schizoaffective (SZ)

Table 4
Sensitivity, Specificity, Positive Predictive Power, Negative Predictive Power, and Odds Ratios of
MMPI Codetypes Hypothesized la Indicate the Presence of Borderline Personality Disorder
and of a Cutoff Score oj'lOT on the MMPI-BPD Scale

Positive Negative
predictive predictive
Sensitivity Specificity power power Odds ratio

MMPI codetype
2, 4, and 8 2= 70r .52 .60 .25 .83 1.65
2 and 7 > 70r .51 .51 .21 .80 1.06
2 and 8 > 70r .57 .52 .24 .83 1.48
4 and 8 > 70T .59 .53 .24 .83 1.61
6 and 8 ^ 70r .59 .54 .25 .84 1.69
7 and 8 > 70r .56 .51 .22 .82 1.29
MMPI-BPD > 70 r
(raw score S; 15) .25 .83 .27 .91 1.61

Note. MMPI-BPD = Minnesota Multiphasic Personality Inventory-Borderline Personality Disorder

scale. No odds ratio significant at p < .05. N = 298, includes all borderline personality disorder, major
depression or dysthymia, bipolar disorder or cyclothymia, schizophrenia or schizoaffective disorder, and
other personality disorder patients.
 236 TIMOTHY J. TRULL

Table 5
Comparison of Diagnostic Groups on the MMP1-BPD Scale

Raw MMPI-
BPD score
Significant
Diagnostic group n M SD ANCOVA F contrast

Comparison with Axis I groups

Borderline (BPD) 61 12.56 3.49 6.82*** BPD > SZ
Bipolar/cyclothymia 36 11.11 3.47
Major depression/dysthymia 70 10.97 3.68
Schizophrenia/schizoaffective (SZ) 96 9.74 3.80
Comparison with Axis II group
Borderline 61 12.56 3.49 6.55* BPD > OPD
Other personality disorder (OPD) 63 10.83 3.74
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Note. MMPI-BPD = Minnesota Multiphasic Personality Inventory-Borderline Personality Disorder

This document is copyrighted by the American Psychological Association or one of its allied publishers.

scale; ANCOVA = analysis of covariance.

*p<.05. **V<.001.

or bipolar/cyclothymia (BC) patients on four clinical scales. No cantly higher on this MMPI scale than did patients in the SZ
significant differences, however, were found between the BPD and OPD groups. This latter result is consistent with previous
and major depression/dysthymia (MD) groups on any of the studies that have compared the MMPI-BPD scores of BPD and
MMPI clinical scales. As for the OPD group, significant differ- OPD groups (Dubro & Wetzler, 1989; Morey et al., 1988). It
ences were found for Scales K and 5, with the OPD group scor- should also be noted that the MMPI-BPD scale does not ap-
ing higher than the BPD group in both cases. The diagnostic pear to be measuring only severity of psychiatric symptoms
efficiency rates (SENS, SPEC, PPP, and NPP) for each MMPI because scores did discriminate between the BPD and SZ
codetype thought to be indicative of a BPD diagnosis were ap- groups, with the SZ group scoring significantly lower on this
proximately equal. PPP values for these codetypes were consis- scale. Finally, the NPP of the MMPI-BPD cutoff score of 70T
tently low; there was little improvement over the base rate of the was higher than that of the MMPI codetypes, indicating that
BPD diagnosis (prevalence = .20). Therefore, an examination the MMPI-BPD may be more useful in ruling out a BPD diag-
of the traditional MMPI scale scores and several codetypes did nosis.
not reveal any pattern of scores distinctive to the BPD diagno- In the present study, however, MMPI-BPD scores failed to
sis, and scores on the traditional MMPI scales did not differen- discriminate BPD patients from those diagnosed with a mood
tiate BPD from depressive disorders or from other personality disorder on Axis I and no BPD diagnosis on Axis II. High scores
disorders. on the MMPI-BPD scale were obtained by those suffering
Results from the present study suggest that Morey et al.'s from mood disorders as well as BPD; the average scores of these
(1985) MMPI-BPD scale discriminates those clinically diag- patients were approximately one or more standard deviations
nosed as BPD from those diagnosed as SZ, as well as from those above the mean score of the normative sample group used in
diagnosed as OPD. In both cases, BPD patients scored signifi- the development of the MMPI personality disorder scales
(Morey & Smith, 1988). These results support the contentions
of Akiskal et al. (1985) and Widiger (1989) that the BPD and
Table 6
mood disorder constructs overlap such that differentiation be-
Cutoff Scores on the Minnesota Multiphasic Personality
tween the two proves difficult. Results suggest that the MMPI-
Inventory-Borderline Personality Disorder Scale (MMPI-BPD)
BPD scale may measure a general construct of affectivity, char-
Raw MMPI-BPD score Sensitivity Specificity Kappa acterized by symptoms of both depressive disorders and bipo-
lar/cyclothymic disorders. Supporting this hypothesis, an
>9 .88 .33 .11 examination of the MMPI-BPD scale reveals that many items
>10 .79 .44 .13
>ll .72 .52 .15
assess affective features of BPD (e.g., I cry easily [True], / very
£12 .64 .61 .17 seldom have spells of the blues [False], I brood a great deal
>13 .56 .69 .19 [True], and / am not easily angered [False]). Therefore, it ap-
>14 .39 .76 .14 pears that many of the current MMPI-BPD items are as char-
>15 .25 .83 .08
acteristic of a mood disorder as they are of borderline pathol-
>I6 .20 .90 .11
&I7 .13 .93 .08 ogy. It is not surprising then that the MMPI-BPD scale fails to
>I8 .08 .97 .07 differentiate borderline and mood pathology.
£=19 .07 .99 .09 Assuming that the distinction between BPD and mood dis-
>20 .00 1.0 .03 order constructs is indeed a valid one, a scale to differentiate
Note. A* = 298, includes all borderline personality disorder, major between BPD and mood disorders could be constructed by
depression ordysthymia, bipolar disorder or cyclothymia, schizophre- emphasizing MMPI items that are shown to empirically dis-
nia or schizoafFective disorder, and other personality disorder patients. criminate between these disorders. Viewed from an empirical
 MMPF-BPD SCALE 237

perspective on scale construction, the purpose of a self-report The MMPI-BPD items were retained in the recent revision
scale (such as the MMPI-BPD) is not to describe or to character- of the MMPI, the MMPI-2 (Butcher, Dahlstrom, Graham, Tel-
ize the disorder's symptomatology, but rather to optimize dif- legen, & Kaemmer, 1989). Therefore, researchers using the
ferential diagnosis (Wiggins, 1973). Consistent with this ap- MMPI-2 can calculate scores on the MMPI-BPD scale and
proach, MMPI items assessing nonaffective features of BPD determine which differential diagnoses are best addressed by
would be emphasized in a scale designed to differentiate BPD this scale. Future research might also examine individual
from mood disorders. For example, BPD features such as "iden- MMPI-2 items to determine if a subset of these would aid in
tity disturbance," "impulsivity," and perhaps "pattern of unsta- the discrimination between BPD and mood disorders. As
ble/intense interpersonal relationships" might be represented previously noted, this has been a difficult differential diagnosis.
by more items than the affective features of BPD in such a scale. It may in fact be the case that the BPD and mood disorder
Morey et al.'s (1985) MMPI-BPD items do not appear to sample constructs are hopelessly intertwined because of an overlap in
these three features of the borderline construct adequately, diagnostic features (Akiskal et al., 1985; Widiger, 1989). It is
whereas affective features of BPD appear to be overrepresented. conceivable, however, that some assessment instruments may
It is, of course, conceivable that a scale constructed to differ- be able to differentiate BPD and mood disorders. Future stud-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

entiate BPD from mood disorders may not differentiate BPD ies should also examine the discriminant validity of other self-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

from other disorders. Several scales or subscales may be neces- report measures of BPD (e.g, the Millon Clinical Multiaxial
sary because it is unreasonable to expect one set of items to be Inventory-II Millon, 1987; the PDQ-R; Hyler & Rieder, 1987)
optimal for ruling out all possible diagnostic alternatives. For to determine whether these instruments can differentiate BPD
example, the items that optimally discriminate BPD from and mood disorders.
mood disorders will not likely be the same as the items that best In summary, the present study suggests that borderline pa-
differentiate BPD from schizotypal personality disorder. In gen- tients can be distinguished from schizophrenia-schizoaffective
eral, it appears that alternative sets of self-report items are patients on the basis of MMPI clinical scale scores as well as
needed to address questions regarding various differential diag- scores on the MMPI-BPD scale. MMPI clinical scale scores
noses (Widiger & Trull, 1991). were helpful in distinguishing borderlines from bipolar-cy-
There are a few potential limitations to the current study that
clothymia patients; however, the MMPI-BPD scale scores for
should be discussed. The results from the present study would
these two groups were not significantly different. The OPD
have been expected if there was a high comorbidity rate for
group was best differentiated from the borderline group by
mood disorder in the BPD group. An examination of the Axis I
their respective MMPI-BPD scores. Finally, the MMPI scales
comorbid diagnoses, however, revealed that only three BPD's
and the MMPI-BPD scale failed to discriminate between the
(5%) received a diagnosis of bipolar disorder or cyclothymia on
major depression-dysthymia group and the borderline group.
Axis I, and 17 (28%) received an Axis I diagnosis of major de-
Whether these latter two groups of patients can be differen-
pression or dysthymia. Therefore, only about one third of the
tiated by means of self-report psychological test scores remains
BPD patients received comorbid mood disorder diagnoses. To
to be demonstrated.
explore this issue further, an additional analysis was performed
in which scores from a "pure" BPD group (i.e., those patients References
who did not meet criteria for any other Axis I or II disorder, n =
19) were compared with the MD, BC, and SZ groups. The same Akiskal, H. S., Yerevanian, B. I., Davis, G. C, King, D., & Lemmi, H.
pattern of results emerged; only the BPD and SZ groups were (1985). The nosologic status of borderline personality: Clinical and
polysomnographic study. American Journal of Psychiatry, 142.192-
discriminable on the basis of the MMPI-BPD scores. There-
198.
fore, it appears that comorbid mood disorder diagnoses in some
American Psychiatric Association. (1980). Diagnostic and statistical
BPD patients do not completely account for the pattern of re-
manual of menial disorders (3rd ed.). Washington, DC: Author.
sults obtained in the present study. American Psychiatric Association. (1987). Diagnostic and statistical
The present study is limited by its reliance on hospital dis- manual of mental disorders (3rd ed, rev). Washington, DC: Author.
charge diagnoses, which could be criticized as potentially unre- Butcher, J. N., Dahlstrom, W G., Graham, J. R., Tellegen, A., & Kaem-
liable and fallible. The diagnoses in this study were made only mer, B. (1989). Manual for the rstandardized Minnesota Multiphasic
after extensive consultation with all team members and repre- Personality Inventory: MMPI-2. Minneapolis: University of Minne-
sent the consensus of a number of professionals from a variety sota Press.
of disciplines. Thus, it is likely that these diagnoses are more Dubro, A., & Wetzler, S. (1989). An external validity study of the
reliable than a diagnosis assigned by only one professional. In MMPI personality disorder scales. Journal of Clinical Psychology,
addition, Fyer et al. (1988) found a high level of convergence 45, 570-575.
between discharge diagnoses at this hospital and those diag- Dubro, A., Wetzler, S., & Kahn, M. (1988). A comparison of three
self-report questionnaires for the diagnosis of DSM-HI personality
noses assigned following a systematic chart review. Neverthe-
disorders. Journal of Personality Disorders, 2, 256-266.
less, the results should be replicated in groups of psychiatric
Fleiss, J. L. (1981). Statistical methods for rates and proportions (2nd
patients diagnosed by structured interview. Although the pres-
ed.). New York: Wiley.
ent study raises the issue of the overlap between the BPD and Fyer, M., Frances, A., Sullivan, T, Hurt, S., & Clarkin, J. (1988). Comor-
mood disorder constructs, additional studies using different bidity of borderline personality disorder. Archives of General Psychia-
research designs (e.g., concurrent assessments of mood, assess- try, 45, 348-352.
ments when patients are not in an acute state) will be necessary Gartner,)., Hurt, S. W, & Gartner, A. (1989). Psychological test signs of
to ultimately address whether these constructs can be disentan- borderline personality disorder: A review of the empirical literature.
gled. Journal of Personality Assessment, 53, 423-441.
 238 TIMOTHY I. TRULL

Gunderson, J, & Elliot, O. (1985). The interface between borderline Piersma, H. (1987). The MCMI asa measure of AW-///Axis II diag-
personality disorder and affective disorder. American Journal of Psy- noses: An empirical comparison. Journal of Clinical Psychology, 43,
chiatry. 142, 277-288. 478-483.
Hathaway, S. R, & McKinley, J. C. (1983). The Minnesota Multiphasic Reich, J., Noycs, R., Corycll, W, & O'Gorman, T. (1986). The effect of
Personality Inventory Manual. New York: Psychological Corpora- state anxiety on personality measurement. American Journal of Psy-
tion. chiatry, 143, 760-763.
Hirschfeld, R., (German, G., Clayton, P, Keller, M., McDonald-Scott, Skodol, A., Rosnick, L., Kellman, Oldham, J., & Hyler, S. (1988). Vali-
P., & Larkin, B. (1983). Assessing personality: Effects of the depres- dating DSM-III-R personality disorder assessments with longitu-
sive state on trait measurement. American Journal of Psychiatry, 140, dinal data. American Journal of Psychiatry, 145,1297-1299.
695-699. Spitzer, R. (1983). Psychiatric diagnosis: Are clinicians still necessary?
Hurt, S., Hyler, S., Frances, A., Clarkin, J., & Brent, R. (1984). Assessing Comprehensive Psychiatry, 24, 399-411.
borderline personality disorder with self-report, clinical interview, Widiger, T. (1989). The categorical distinction between personality
or semistructured interview. American Journal of Psychiatry, 141, and affective disorders. Journal of Personality Disorders, 3, 77-91.
1228-1231. Widiger, T. A., Sanderson, C, & Warner, L. (1986). The MMPI, proto-
Hyler, S. E, & Rieder, R. O. (1987). PDQ-R personality questionnaire. typal typology, and borderline personality disorder. Journal of Per-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

New York: New York State Psychiatric Institute. sonality Assessment, 50, 540-553.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Millon, T. (1987). Millon Clinical Mulliaxial Inventory-II: Manual for Widiger, T, & Trull, T. (in press). Borderline and narcissistic personal-
the MCMI-I1. Minneapolis, MN: National Computer Systems. ity disorders. In P. Sutker, & H. Adams (Eds.), Comprehensive hand-
Morey, L, Blashfield, R., Webb, W, & Jewell, J. (1988). MMPI scales for book of psychopathology (2nd ed.). New "York: Plenum.
DSM-II1 personality disorders: A preliminary validation study. Widiger, T. A., & Trull, T. J. (1991). Diagnosis and clinical assessment.
Journal of Clinical Psychology, 44, 47-50. Annual Review of Psychology, 42,109-133.
Morey, L., & Smith, M. R. (1988). Personality disorders. In R. Greene Wiggins, J. (1973). Personality and prediction: Principles of personality
(Ed.), The MMPI: Use with specific populations (pp. 110-158). Phila- assessment. Reading, MA: Addison-Wesley.
delphia: Grune & Stratton.
Morey, L., Waugh, M, & Blashfield, R. (1985). MMPI scales for DSM- Received May 14,1990
III personality disorders: Their derivation and correlates. Journal of Revision received September 24,1990
Personality Assessment. 49, 245-251. Accepted October 11,1990 •