Content uploaded by Timothy J Trull
Author content
All content in this area was uploaded by Timothy J Trull
Content may be subject to copyright.
JACKSON AND TRULL
PAI-BOR FACTOR STRUCTURE
THE FACTOR STRUCTURE OF THE
PERSONALITY ASSESSMENT
INVENTORY-BORDERLINE FEATURES (PAI-BOR)
SCALE IN A NONCLINICAL SAMPLE
Kristina M. Jackson, PhD, and Timothy J. Trull, PhD
We evaluated the fit of Morey’s (1991) proposed 4-factor structure on Per-
sonality Assessment Inventory-Borderline Features Scale (PAI-BOR;
Morey, 1991) items in a sample of approximately 5,000 nonclinical par-
ticipants . The proposed model did not fi t the data well. Re sults from a se-
ries of exploratory and confirmatory factor analyses suggested that a
6-factor model provided the best fit to the PAI-BOR item covariances.
A number of self-report and structured interview assessment instruments
target borderline personality disorder (BPD). A relatively recent self-report
measure of BPD features is the Personality Assessment Inventory-Border-
line Features (PAI-BOR) Scale (Morey, 1991). PAI-BOR items tap features of
severe personality pathology that are commonly associated with BPD.
Based on a review of the historical conceptualizations of BPD, as well as on
empirical studies of borderline patients, potential PAI-BOR items were gen-
erated to reflect core “factors” of the construct (Morey, 1991). These factors
are affective instability, identity problems, negative relationships, and
self-harm (Morey, 1991). Final item selection was guided by both the con-
ceptual nature of the items as well as the items’ psychometric properties.
The final version of the PAI-BOR consists of 24 items that are rated on a
4-point scale (0 to 3; false, slightly true, mainly true, very true).
Preliminary studies have supported the reliability and validity of total
PAI-BOR scores in indexing the degree to which borderline personality fea-
tures are present (Morey, 1991; Trull, 1995, 2001). To date, however, only
Morey (1991) has provided data addressing the factor structure of PAI-BOR
items. Morey (1991) tested the proposed 4-factor structure of the PAI-BOR
by performing a confirmatory factor analysis on PAI-BOR item data from a
clinical sample (n= 1,246) that was predominantly male and between the
ages of 30 and 49. Morey (1991) reported that the proposed 4-factor struc-
ture (6 PAI-BOR items per factor) provided an excellent fit to the data
Journal of Personality Disorders, 15(6), 536-545, 2001
© 2001 The Guilford Press
536
From the University of Missouri-Columbia.
This research was supported by National I nstitute of Mental Health Grants R5 5MH52695 and
R01 MH52695.
Address c orrespondence to Timothy J. Trull, PhD, D epartment of Psyc hological Science, 1 06C
McAlester Hall, University of Missouri, Columbia, MO 65211; E-mail: TrullT@missouri.edu.
(Bentler-Bonett Normed Fit Index = .98; Comparative Fit Index = .98). Morey
(1991) also presented some preliminary data supporting the validity of the
four PAI-BOR subscales, and he has advocated using elevations on all four
of these subscales (i.e., ³70T score on each) as a psychometric marker for a
diagnosis of BPD.
The purpose of our study was to attempt to replicate Morey’s (1991) con-
firmatory factor-analysis findings and proposed PAI-BOR factor structure
in a large, nonclinical sample. Such a replication would be important be-
cause factor invariance across groups (e.g., clinical vs. nonclinical) cannot
be assumed, and findings suggesting lack of invariance will have implica-
tions for the use of subscales as measures of the “core” features of border-
line personality.
METHOD
Approximately 5,000 18-year-old freshmen at the University of Missouri
particip ated in a scree ning assessment for a prospective study on the dev el-
opment of borderline personality disorder features in young adults during
the acad emic years 19 97-1998 and 19 98-1999 (Trull, 2001). P articipants in
the screening phase of the study were contacted through mailings, classes,
telephone calls, and electronic messages (e-mails), and were scheduled to
complete the screening battery during supervised sessions held in dormito-
ries, fraternities and sororities, and classrooms. Potential participants were
scheduled up to ten times to complete the assessment or until they indi-
cated their desire to not participate. All those who completed the screening
received five dollars or research credit if enrolled in Introduction to Psychol-
ogy.
The screening battery included items from the PAI-BOR scale (Morey,
1991) as well as validity items from the Personality Diagnostic Question-
naire-Revised (PDQ-R; Hyler & Rieder, 1987) aimed at identifying individu-
als who tended to present themselves in an overly favorable light (Too Good
subscale), to respond in a random or haphazard way (Suspect Question-
naire subscale), or to respond in a way that suggested deceitfulness (Lie
subscale). Finally, items that assessed demographic information were also
included.
RESULTS
The factor analytic results reported here are based on the 4,682 partici-
pants with complete PAI-BOR data who were not excluded based on thei r re-
sponses to the validity items. This sample was predominantly female (63%),
most participants were white (82%), and most were from middle-class fami-
lies (64%; family income between $25,000-$100,000). The average total
PAI-BOR score in this sample was 24.71 (SD = 10.56), and the mean total
scores for men and women, respectively, did not differ significantly. Table 1
shows means and standard deviations for the 24 PAI-BOR scale items for
the full sample (and the calibration and validation sample, respectively; see
below).
PAI-BOR FACTOR STRUCTURE 537
CONFIRMATORY FACTOR ANALYSES
The major goal of this study was to attempt to replicate Morey’s (1991) pro-
posed factor structure of PAI-BOR items. Toward this goal, we conducted a
confirmatory factor analysis (CFA) using data from the full sample (n=
4,682). For this analysis, we used a maximum likelihood estimation proce-
dure included in the statistical package MPlus (Muthen & Muthen, 1999).
The fit of Morey’s proposed 4-factor model (affective instability, identity dis-
turbance, negative relationships, self-harm) was poor, c2(246, N= 4,682) =
8,279.50; RMSEA = .08; Comparative Fit Index (CFI) = 0.74; Tucker-Lewis
Index (TLI) = .70. Figure 1 provides standardized factor loadings for
PAI-BOR items and correlations between factors for this 4-factor model.1
Although the chi-square was approximately twice the degrees of freedom,
the relatively low fit indices, particularly the CFI and the TLI, warranted
538 JACKSON AND TRULL
TABLE 1. Item Content, and Means (and Standard Deviations) for the 24 PAI-BOR
Items, for the Full Sample, for the Calibration Sample, and for the Validation Sample
Full
sample
Calibration
sample
Validation
sample
Scl Item (n= 4682) (n= 2341) (n= 2341)
AI 1. Mood shifts 1.35 (0.87) 1.33 (0.86) 1.37 (0.88)
ID 2. Attitude about self changes 1.07 (0.87) 1.07 (0.86) 1.06 (0.87)
NR 3. Relationships stormy 0.72 (0.84) 0.72 (0.84) 0.72 (0.83)
AI 4. Moods intense 1.19 (0.96) 1.18 (0.95) 1.21 (0.96)
ID 5. Feel empty 1.01 (0.98) 1.00 (0.97) 1.02 (0.99)
NR 6. Let people know they’ve hurt me 1.16 (1.02) 1.17 (1.02) 1.15 (1.02)
AI 7. Mood steady* 1.51 (0.92) 1.50 (0.92) 1.51 (0.92)
ID 8. Worry about people leaving 0.8 6 (0.94) 0.86 (0.94) 0.86 (0.95)
NR 9. People let me down 1.31 (1.05) 1.32 (1.05) 1.29 (1.05)
AI 10. Little control over anger 0.46 (0.71) 0.46 (0.71) 0.46 (0.71)
ID 11. Wonder about life 1.4 9 (1.07) 1.50 (1.07) 1.48 (1.07)
NR 12. Rarely lonely* 1.39 (0.97) 1.40 (0.97) 1.38 (0.97)
SH 13. Do things impulsively 0.78 (0.90) 0.78 (0.90) 0.78 (0.90)
AI 14. Happy person* 0.92 (0.86) 0.92 (0.86) 0.92 (0.86)
ID 15. Can’t handle separation 1.23 (0.95) 1.23 (0.95) 1.23 (0.96)
NR 16. Mistakes in picking friends 0.59 (0.87) 0.61 (0.89) 0.57 (0.85)
SH 17. When upset, hurt self 0.19 (0.52) 0.18 (0.51) 0.20 (0.54)
AI 18. Can’t express all of anger 0.95 (1.04) 0.94 (1.03) 0.95 (1.05)
ID 19. Don’t get bored* 1.68 (0.95) 1.70 (0.95) 1.66 (0.94)
NR 20. Stay friends with people* 1.06 (0.74) 1.05 (0.73) 1.07 (0.76)
SH 21. Too impulsive 0.54 (0.76) 0.54 (0.76) 0.54 (0.76)
SH 22. Spend money easily 1.41 (1.06) 1.41 (1.06) 1.42 (1.05)
SH 23. Reckless person 0.38 (0.65) 0.38 (0.65) 0.38 (0.65)
SH 24. Careful about money* 1.47 (0.97) 1.45 (0.97) 1.48 (0.97)
Note. Scl = Morey (1991) subscale to which the item was originally assigned. AI = Affective Instability; ID =
Identity Problems; NR = Negative Relationships; SH = Self-Harm. *Indicates this item is reverse-scored.
These are brief descriptions of items, not actual PAI-BOR items.
1. Following Morey (1991), the model allowed estimates of covariances between the four latent
factors but did not estimate any item error covariances.
concern. In addition, the correlation between identity disturbance and neg-
ative relationships factors was r= .91, which was much higher than desired,
and a number of factor loadings were low (i.e., £.40).
Given these problems with model fit, we examined modification indices to
determine whether the addition or removal of parameters would result in an
improved model fit. Modification indices suggested a significant improve-
ment i n model fit with the addition of parameters (i.e., paths) from item 17 to
each of the other factors (i.e., negative affectivity, identity disturbance, and
negative relationships). The wording of item 17 suggests that it may tap
both negative affect and self-harm. As such, we concluded that item 17 was
factorially complex, and we dropped it from the model. In addition, we noted
that allowing the error terms for items 22 and 24 to covary would increase
model fit dramatically. However, given the content overlap between items 22
and 24 (both refer to spending money recklessly), and their high
intercorrelation (r= .74), we decided to drop item 24 in the revised model. A
CFA on this revised model (i.e., with items 17 and 24 removed) was con-
ducted. Although model fit improved, c2(203, N= 4,682)= 4454.49; RMSEA
= .07; CFI = 0.83; TLI = .81, fit indices remained below the desired value (at
or below .06 for RMSEA, and at or above .95 for CFI and TLI; Hu & Bentler,
1999). Further, the factors of identity disturbance and negative relation-
ships remained highly correlated (r= .91).
Next, we hypothesized that, given their high intercorrelation (r= .91),
identity disturbance and negative relationships might be better modeled as
a single factor. We combined these two subfactors into a single factor, and
reestimated the model using all 24 items. Model fit was not improved,
PAI-BOR FACTOR STRUCTURE 539
FIGURE 1. Standardized factor loadings for a confirmatory factor analysis of Morey’s 4-factor
PAI-BOR model.
c2(249, N= 4,682) = 8,350.18; RMSEA = .08; CFI = 0.73; TLI = .71. Finally,
we noted that some of the individual items were skewed (e.g., for item 17,
skew = 3.19; kurtosis = 10.92). As such, we computed polychoric correla-
tions between items and reestimated the initial model proposed by Morey
(1991) using the Calis procedure in SAS (SAS Institute, 1999). Model pa-
rameters were quite similar to those of the maximum likelihood model
shown in Figure 1, and model fit was not improved.
Given the relatively poor fit of Morey’s model using our nonclinical sample
data, and the failure to improve upon the model with removal of items, the
combination of factors, or estimation using polychoric correlations, we de-
cided to empirically test the nature of the factor structure using exploratory
factor analysis.
EXPLORATORY FACTOR ANALYSES
We divided our sample into halves, a calibration subsample (N= 2,341) and
a validation subsample (N= 2,341), according to observation (i.e., partici-
pant) number (odd observation numbers were assigned to the calibration
sample, even numbers were assigned to the validation sample). The number
of males and females was roughly equal in each subsample (36.9% male in
the calibration sample, 37.2% male in the validation sample). Table 1 shows
means and standard deviations for the 24 PAI-BOR scale items for the cali-
bration sample (middle column) and validation sample (right column).
Using the calibration sample, we performed an exploratory factor analysis;
with the validation sample, we replicated the derived empirical model using
confirmatory factor analysis.
Calibration Sample. We conducted a principal components analysis
(communality estimates set equal to one) with promax rotation. To deter-
mine the best model (i.e., the number of factors to extract), we used two cri-
teria: (1) the number of factors with an eigenvalue above 1.0, and (2) the
scree plot, which is a visual representation of each factor solution. Six fac-
tors had eigenvalues greater than one.2Although there were two “elbows” in
the scree plot, one suggesting a 4-factor solution and one suggesting 6 fac-
tors, our criteria suggested that a 6-factor solution might be most appropri-
ate.3The 6-factor solution included the following factors: lack of
control/impulsive behavior (items 13, 17, 18, 21, and 23), mood instability
(items 1, 4, 7, and 10), chronic emptiness/loneliness/boredom (items 2, 5,
11, 12, 14, and 19), separation and abandonment concerns (items 6, 8, and
15), negative relationships (items 3, 9, 16, and 20), and reckless spending
(items 22 and 24). Standardized regression coefficients from the rotated
540 JACKSON AND TRULL
2. Eigenvalue s for the first eight facto rs were 5.95, 2.14, 1 .48, 1.32, 1 .22, 1.07, 0.93, and 0.91,
respectively.
3. Examination of the item factor loadings generated by the 4-factor solution revealed many
item cross-loadings, making interpretation of the factors difficult. Therefore, our choice of a
6-factor solution also resulted in a much more interpretable factor solution.
pattern matrix are presented in Table 2. As can be seen, only one item (item
9) loaded greater than .40 on more than one factor.
Validation Sample. Using confirmatory factor analysis, we tested the fit of
this 6-factor model in the other half of the sample. Model fit was fair [c2(237,
N= 2,341) = 2,432.67; RMSEA = .06; CFI = 0.86; TLI = .83; see Figure 2 for
standardized parameter estimates].
In ord er to deter mine whether our fair (but not exceptional) mod el fit in the
validation sample was due primarily to model fit overall or to only a fair rep-
lication of the model, we also conducted a 6-factor CFA on the calibration
sample. Model fit for the calibration sample was nearly identical to that of
the validation sample, c2(237, N= 2,341) = 2,429.12; RMSEA = .06; CFI =
0.86; TLI = .83, suggesting excellent replication of a fair (but not exception-
ally well fitting) model.
Two-Factor Model. Finally, given the fit of the 6-factor model, we esti-
mated a 2-factor model on the calibration sample because the eigenvalues
associated with the first two factors were much larger than those of the re-
PAI-BOR FACTOR STRUCTURE 541
TABLE 2. Standardized Regression Coefficients from the Six-Factor Rotated Pattern
Matrix
Factor
1 2 3 4 5 6
Item
Lack of
control/
impulsive
behavior
Mood
instability
Chronic
emptiness/
loneliness/
boredom
Separation and
abandonment
concerns
Negative
relationships
Reckless
spending
1 0.03 0.85* 0.02 0.02 -0.10 0.01
2-0.003 0.33 0.38* 0.27 -0.17 0.04
30.18 0.16 -0.01 0.12 0.41* 0.01
40.17 0.70* -0.03 0.06 -0.001 -0.10
5 0.07 0.10 0.54* 0.32 0.03 -0.08
6 -0.05 0.17 -0.09 0.50* 0.32 0.00
7 -0.14 0.74* 0.21 -0.05 0.04 0. 09
8-0.05 -0.11 0.18 0.70* 0.16 0.05
9-0.02 -0.05 -0.006 0.42 0.58* -0.01
10 0.37 0.42* -0.03 -0.04 0.07 -0.06
11 0.38 -0.19 0.50* 0.14 -0.24 -0.01
12 -0.10 0.05 0.70* 0.13 0.03 0.03
13 0.75* 0.04 -0.13 -0.03 0.04 0. 09
14 0.03 0.16 0.59* -0.20 0.26 -0.06
15 -0.07 0.06 -0.03 0.73* -0.12 0.10
16 0.08 -0.08 -0.07 0.16 0.71* 0.03
17 0.49* -0.12 0.28 0.02 0.11 -0.22
18 0.43* 0.21 0.04 0.10 0.14 -0.15
19 -0.11 0.04 0.55* -0.09 0.06 0.24
20 -0.02 -0.04 0.20 -0.31 0.67* 0.07
21 0.67* 0.10 -0.14 -0.02 0. 04 0.19
22 0.12 -0.003 -0.00 0.15 0. 01 0.85*
23 0.76* 0.00 0.08 -0.18 -0.04 0.17
24 0.06 -0.03 0.11 0.02 0.05 0.88*
Note. Calibration Sample, N= 2341. Coefficients m arked with an asterisk (*) repres ent the factor to which
an item was assigned in the 6-factor solution.
maining factors. According to a 2-factor EFA, the first factor, which repre-
sented a smaller impulsivity/disinhibition factor, contained items 13, 21,
22, 23, and 24. The second factor represented a large negative emotionality
factor and contained all remaining items. We attempted to replicate this
2-factor solution on the validation sample using CFA; however, model fit
was poor, c2(251, N= 2,341) = 4,525.42; RMSEA = .08; CFI = 0.72; TLI = .69.
DISCUSSION
This study attempted to replicate the factor structure of PAI-BOR items pro-
posed by Morey (1991) in a nonclinical sample. Results suggested that
Morey’s (1991) 4 -factor model did not provide an adequate fit to the data. In-
stead, of the models we tested, a 6-factor model appeared to account best for
the covariances between PAI-BOR items. Several of these six factors resem-
bled those proposed by Morey (1991).
According to Morey (1991), PAI-BOR items tap four major features or
traits of borderline personality disorder that are embedded in the DSM-IV
[American Psychiatric Association (APA), 1994] criteria set for BPD and that
are included in historical accounts of this personality disorder. Items tap-
ping affective instability seem well represented in the PAI-BOR. In both clin-
ical (Morey, 1991) and nonclinical samples (present study), factor analyses
of PAI-BOR items revealed a factor reflecting mood shifts, intense mood
states, and anger dyscontrol. Our mood instability factor included four of
Morey’s six affective instability items. However, Morey’s (1991) two remain-
ing affective instability items seem to reflect a lack of happiness and anger
542 JACKSON AND TRULL
.62
Impulsiv ity/
Dyscontrol
Reckless
spendin g
Negative
relations
Mood
instability
Separation
concerns
Chronic
emptiness
13 17 18 21 23 1 4 7 10 2 5 11 12 14 6 8 15 3 9 16 20 22 2419
.45 .29 .93 .79.58.44.67.59.30.57.64.33.76.59.51.70.71.72.66.70.46.37.72
.24.76.68
.57 .71
.41 .56 .72 .24
.40
.66
.17
.23.54
.42
FIGURE 2. Standardized factor loadings for a confirmatory factor analysis on the validation
sample to replicate the 6-factor PAI-BOR model.
expression, and these items loaded on our chronic emptiness and our
dyscontrol factors, respectively.
The present study found evidence for an impulsivity/dyscontrol factor, a
major feature of BPD (APA, 1994). More y’s (19 91) self-h arm factor reflects “a
tendency to engage in impulsive self-destructive behaviors” (p. 72), and four
of his six self-harm factor items loaded on our impulsivity factor. Interest-
ingly, the PAI-BOR does not include any items that directly assess suicidal
behavior, often considered a hallmark feature of BPD. The endorsement of
one PAI-BOR item (item 17) may reflect suicidal behavior or it may reflect
self-harm/self-mutilation without the intent to kill oneself. Therefore, one
potential limitation of the PAI-BOR is that this feature of BPD (i.e., suicidal
behavior) is underrepresented. Morey’s (1991) PAI does include a Suicidal
Ideation scale, but these items do not directly assess suicidal behavior (i.e.,
gestures, attempts).
Morey’s (1991) identity problems and negative relationships factors also
have historical precedents and appear as features listed in the DSM-IV BPD
criteria set. In general, these factors emerged from our analyses (although
we gave the factors different labels). Our chronic emptiness/loneli-
ness/boredom factor included four of the six identity problem items
whereas the remaining two loaded on our separation and abandonment
concerns factor. Although one could certainly make a case for separation
and abandonment concerns reflecting identity problems, our analyses sup-
ported a separate factor. It is also noteworthy that Morey’s identity prob-
lems factor seemed the least coherent in our sample (see Figure 1). Three of
the six items were only modestly related to this latent variable (standardized
factor loadings £.40). Conce rning Morey’s (1991) negative re lationships fac-
tor, four of the six items loaded on a negative relationships factor in our
sample as well. The remaining two items (along with item 15) loaded on a
separation and abandonment concerns factor in our sample.
Along with our separation and abandonment concerns factor, another
somewhat unique factor emerged from our data. Specifically, a sixth factor,
reckle ss spending, was comp rised of tw o items. Although reckles s spending
may be seen as an indicator of impulsivity, these items did not load highly
on this factor in our sample. Although these two items did seem to define an
interpretable factor in our nonclinical sample, the similarities in the word-
ing of these two items may call into question the inclusion of both in the
PAI-BOR. Future research might focus on the necessity of two similarly
worded items.
It is also instructive to compare our results to those from several recent
studies that examined the latent structure of DSM BPD symptoms (e.g.,
Clarkin, Hull, & Hurt, 1993; Fossati, Maffei, Bagnato, Donati, Caterina, &
Novella, 1999; Sanislow, Grilo, & McGlashan, 2000). Although these stud-
ies found evidence for fewer latent factors or latent classes underlying BPD
criteria (because of the smaller number of variables considered), several of
their findings wer e consisten t with the results f rom our stud y. Similar to our
study, previous investigators found evidence for an impulsivity/behavioral
dysregulation factor. Our factor is most similar to that extracted by
Sanislow et al . (2000) in that it includes items related to self -harm. Previous
studies, and our own study, also found evidence for a mood instability fac-
PAI-BOR FACTOR STRUCTURE 543
tor. However, in some studies, this factor was defined more by anger
dyscontrol than other types of affective instability (e.g., Sanislow et al.,
2000). Finally, several studies (Clarkin et al., 1993; Sanislow et al., 2000)
reported a latent factor involving interpersonal/relationship disturbance.
However, our negative relations factor was narrower and did not include
features related to identity disturbance, feelings of emptiness, and aban-
donment concerns.
It is important to highlight several points and acknowledge the limitations
of the present study. First, these results do not challenge the reliability, va-
lidity, or use of the total PAI-BOR score as an indicator of degree of border-
line pathology or features. Our results only suggest that researchers or
clinicians should be cautious in their interpretation of scores on Morey’s
(1991) PAI-BOR subscales as markers for the presence of the “core” features
of BPD-affective instability, identity problems, negative relations, and
self-harm-in nonclinical samples. Our results suggest a different factor
structure; our six factors are similar to but not identical to several of
Morey’s (1991) four factors. Second, in addition to its nonclinical nature,
the present sample differed from Morey’s in several important ways. For ex-
ample, all participants in the present study were 18 years of age and most
were female, whereas, most clinical participants in Morey’s (1991) valida-
tion sample were between 30 and 49 years of age, most were male, and most
received affective disorder or substance use disorder diagnoses. Further, as
would be expected, Morey’s (1991) clinical sample produced higher mean
scores of the PAI-BOR scale and its four subscales than we observed in our
nonclinical sample. Differences in the endorsement or “difficulty levels” of
items between samples can also influence the degree to which their item
covarianc e structures are similar. Finally, two of the six factors identified b y
the present study included a relatively small number of items (i.e., separa-
tion/abandonment concerns and reckless spending). These factors
emerged because their respective items did not covary significantly and in a
systematic fashion with items defining the other four factors. However, the
utility of subscales with so few items is limited. Therefore, the present re-
sults are best seen as a more detailed examination of the structure of the
PAI-BOR items in a nonclinical sample rather than as evidence advocating
the establishment of new PAI-BOR subscales. It will be important for future
research to attempt to test Morey’s (1991) 4-factor model of PAI-BOR items
in additional nonclinical and clinical samples.
In summary, the major implication of this research is that the factor
structure of PAI-BOR reported for clinical samples (Morey, 1991) does not
appear to account adequately for the item covariance structure in a large
nonclinical sample of young adults. Thus, investigators should be cautious
in calculating and interpre ting PAI-BOR su bscale scores in similar samples.
The present results suggest that at least some of Morey’s four subscales
may be multidimensional, and alternative interpretations of scores on these
subscales may be warranted.
544 JACKSON AND TRULL
PAI-BOR FACTOR STRUCTURE 545
REFERENCES
American Psychiatric Asso ciation. (1994). Di-
agnostic and sta tistical ma nual for men-
t a l d is o r d e rs (4t h E d ition ) .
Washington, D C: American Psychiatric
Press.
Clarkin, J. F., Hull, J. W., & Hurt, S. W.
(1993). Factor structure of borderline
personality disorder criteria. J ournal of
Personality D isorders,7, 137-143.
Fossati, A., Maffei, C., Bagnato, M., Donati,
D., Caterina, N., & Novella, L. (1999).
Latent structure analysis of DSM-IV
borderline personality disorder crite-
ria.Comprehensive Psychiatry,40,
72-79.
Hu, L., & Be ntler, P. M. (1999). Cutoff criteria
for fit indexes in covariance structure
analysis: Conventional criteria versus
new alternatives. Struct ural Equation
Modeling, 6, 1-55.
Hyler, S. E., & Rieder, R. O. (1987). PDQ-R
Personality Quest ionnaire. New York:
New York State Psychiatric Institute.
Morey, L. C. (1991). Personali ty assessment
i nv e n t o ry : P r o fe s s i o na l m a n u a l.
Odessa, FL: Psychological Assessment
Resources.
Muthen, L. K., and Muthen, B. (1999). Mplus
user’s guide. Los Angeles: Muthen &
Muthen.
Sanislow, C. A., Grilo, C. M., McGlashan, T.
H. (2000). Factor analysis of the
DSM-III-R bo rderline personality diso r-
der criteria in psychiatric in patients.
American J ournal of Psychiatry,157,
1629-1633.
SAS Institute, Inc. (1999). Cary, NC, USA.
Trull, T. J. (1995). Borderline personality dis-
order features in nonclinical young
adults I: Identification and validation.
Psychological A ssessment,7, 33-41.
Trull, T. J. (2001). Structural relations be-
tween borderline personality disorder
features and putative etiological corre-
lates. Journ al of Abnormal Psychology,
110, 471-481.