MODELS OF PERSONALITY HETEROGENEITY AMONG INDIVIDUALS WITH PTSD

By
Katherine M. Thomas

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF ARTS
Psychology
2011

ABSTRACT
MODELS OF PERSONALITY HETEROGENEITY AMONG INDIVIDUALS WITH PTSD
By
Katherine M. Thomas
Recent typological research assessing relations between personality and PTSD
has consistently replicated three clusters: low pathology, internalizing, and externalizing.
However, the manner in which personality relates to psychopathology generally, and to PTSD
specifically, may be contingent on the model of personality being considered. In particular,
PTSD-personality relations may be direct, pathoplastic, or a hybrid combination of these direct
and pathoplastic relations, but these possibilities have not been examined because of the limited
range of personality models that have been employed to conceptualize personality heterogeneity
in PTSD. The aim of the present study was to examine personality heterogeneity among
individuals diagnosed with PTSD using three personality models hypothesized to show varying
typologies. Participants were recruited using data from the Collaborative Longitudinal
Personality Disorder Study, and were clustered using personality disorder (PD) symptom counts,
Five-Factor Model (FFM) traits, and the Interpersonal Circumplex (IPC) traits. Empirical
validation procedures yielded a three cluster solution for both the PD and FFM models and a four
cluster solution for the IPC. These models all demonstrated unique and incremental relations to
one another, to PTSD, and to other forms of psychopathology. These results indicate that the
nature of the personality-PTSD relation depends on how personality is measured and
conceptualized, and that different models of personality increment one another in describing
individuals with PTSD. The potential of these personality models to help explain etiology,
anticipate co-occurring psychopathology, and tailor treatment planning is also discussed.

ACKNOWLEDGEMENTS
This research was made possible by data collected in conjunction with the Collaborative
Longitudinal Personality Disorders Study (CLPS). Funding for the CLPS is provided by National
Institutes of Health Grants MH050837 to Brown University Department of Psychiatry and
Human Behavior, MH050839 to Columbia University and New York State Psychiatric Institute,
MH050850 to Yale University School of Medicine, MH050840 to Harvard Medical School and
McLean Hospital, and MH050838 to Texas A&M University. I would like to thank the following
CLPS investigators for approving my research proposal and generously sharing data for this
thesis: Emily B. Ansell, Carlos M. Grilo, John C. Markowitz, Thomas H. McGlashan, Leslie C.
Morey, Meghan E. McDevitt-Murphy, Charles A. Sanislow, M. Tracie Shea, Andrew E. Skodol,
and Mary C. Zanarini. I would also to thank my master’s thesis committee for providing
consistent guidance and support throughout my completion of this project: M. Brent Donnellan,
Alytia Levendosky, and particularly my advisor and committee chair, Christopher J. Hopwood.

iii

TABLE OF CONTENTS
LIST OF TABLES ...........................................................................................................v
LIST OF FIGURES .........................................................................................................vi
INTRODUCTION ...........................................................................................................1
Direct Models.......................................................................................................3
Pathoplastic Models .............................................................................................7
Hybrid Models .....................................................................................................10
The Current Study ................................................................................................13
METHOD ........................................................................................................................17
Participants ...........................................................................................................17
Measures ..............................................................................................................18
Statistical Analyses ..............................................................................................19
RESULTS ........................................................................................................................25
Selecting the Optimal Number of Clusters .........................................................25
Final Cluster Solutions........................................................................................28
Testing Study Hypotheses...................................................................................34
DISCUSSION ..................................................................................................................38
Three Distinct Models of Personality and PTSD .................................................39
Relations between Personality Models ................................................................43
Change in Personality Models over Time ............................................................45
Methodological Considerations ...........................................................................47
Limitations, Implications, and Future Directions ................................................50
REFERENCES ................................................................................................................78

iv

LIST OF TABLES
Table 1. Means and Standard Deviations for the number of PD symptoms for individuals with
PTSD .......................................................................................................................55
Table 2. Average Kappa Agreement between sub-samples A and B for Cluster Solutions in Each
Model ......................................................................................................................56
Table 3. Optimal Number of Clusters for Each Personality Model based on Four Empirical
Procedures ...............................................................................................................57
Table 4. PD Clusters’ Mean within Sample T-scores and Standard Deviations for PD, FFM, and
IPC variables ...........................................................................................................58
Table 5. FFM Clusters’ Mean within Sample T-scores and Standard Deviations for PD, FFM,
and IPC variables ....................................................................................................59
Table 6. IPC Clusters’ Mean within Sample T-scores and Standard Deviations for PD, FFM, and
IPC variables ...........................................................................................................60
Table 7. Demographic Characteristics for each cluster ...................................................62
Table 8. Crosstabs between PD and Other Model Clusters .............................................63
Table 9. Crosstabs between FFM and Other Model Clusters…………………………...64
Table 10. Crosstabs between IPC and Other Model Clusters ..........................................65
Table 11. Percent of Individuals in Each Cluster with Axis I Disorder Diagnoses .........66
Table 12. Crosstabs between Baseline and Follow-up PD Clusters ................................67
Table 13. Crosstabs between Baseline and Follow-up FFM Clusters .............................68
Table 14. Crosstabs between Baseline and Follow-up IPC Clusters ...............................69

v

LIST OF FIGURES
Figure 1. The Interpersonal Circumplex (IPC) ................................................................70
Figure 2. Expected Comorbidity Patterns with IPC-based Clusters ................................71
Figure 3. IPC octants........................................................................................................72
Figure 4. Fusion Coefficient (Y-Axis) Plot for the Number of Clusters (X-Axis) in the
PD Model ................................................................................................................73
Figure 5. Fusion Coefficient (Y-Axis) Plot for the Number of Clusters (X-Axis) in the
FFM Model .............................................................................................................74
Figure 6. Fusion Coefficient (Y-Axis) Plot for the Number of Clusters (X-Axis) in the
IPC Model ...............................................................................................................75
Figure 7. Dominance and Warmth Z-score Projections for the Four IPC Clusters .........76
Figure 8. Dominance and Warmth Z-scores for each PD and FFM cluster Projected
on the IPC ...............................................................................................................77

vi

Models of Personality Heterogeneity among Individuals with PTSD
The relationship of personality to psychopathology is complex, and it is
through the dismantling of this complexity that continued progress will be made
[in understanding the role of personality in diagnosis] – Widiger & Smith, 2008,
p. 762
Personality can relate to psychopathology in a variety of ways (Klein, Wonderlich, &
Shea, 1993; Widiger & Smith, 2008). In discussing this relation, Widiger and Smith (2008)
noted: “It would be surprising if the presentation, course, or treatment of an impairment in a
psychologically important component of thinking or feeling (a disorder) were not significantly
affected by a person’s characteristic manner of thinking and feeling (i.e. the individual’s
personality)” (p. 744). However, the relation between personality and psychopathology may
differ depending both on the personality system and the psychopathology being considered.
Models used to describe associations between personality and psychopathology can be
broadly organized as direct, pathoplastic, or hybrid. Direct models posit that personality and
psychopathology are linearly and etiologically related. Alternatively, pathoplastic models assert
that personality influences the expression of psychopathology and that psychopathology
influences personality expression in a bidirectional relationship, but that both domains are
etiologically independent. A third, hybrid, model combines these two frameworks by assuming
that some traits will tend to be directly related to disorders while others will tend to be
pathoplastic, depending on which traits and which disorders are being considered. Hybrid
models can contain personality traits that may relate directly to psychopathology as well as
personality traits that show pathoplastic relations with psychopathology. Therefore distinct
hybrid models of personality can vary greatly in their relation both to psychopathology as well as
in their relation to other hybrid models of personality.

1

These models allow for different kinds of predictions regarding comorbid conditions,
etiological factors, and treatment processes. Direct models can provide useful information for
researchers and clinicians who are interested in etiological personality factors that may render
individuals vulnerable to particular disorders, whereas researchers and clinicians interested in
differential expressions of the same disorders may find more utility in pathoplastic models.
Clinically, direct models are useful for making predictions about how personality might
predispose certain kinds of symptoms, whereas pathoplastic models help explain heterogeneity
among people with the same disorder. Pathoplastic models are also more likely than direct
models to increment diagnoses in clinical predictions because they are generally independent of
disorder severity. Finally, because direct models are linked to disorder severity, they may tend to
be less stable as symptoms increase or remit; conversely the stability of pathoplastic models
should be independent of the degree to which disorders wax and wane. Hybrid models combine
elements of both direct and pathoplastic models in explaining both vulnerability to and
heterogeneity in the expression of disorders and consequently may have utility in many of the
aforementioned domains.
Although these models have been applied to a number of disorders, limited research
exists on how they can be applied to the development and expression of posttraumatic stress
disorder (PTSD). Most previous models of the personality-PTSD relation have used traits that
lead to direct or hybrid effects, and limited attention has been paid to pathoplastic models, which
might help explain heterogeneity among individuals with PTSD that is independent of symptom
severity. As such, our understanding of personality-PTSD relations remains limited. The
purpose of the present study is to test the degree to which three common models of personality –
personality disorders (PDs), five-factor model traits (FFM), and the interpersonal circumplex

2

(IPC) – demonstrate unique direct, pathoplastic, or hybrid relations to PTSD. While the type of
relation (e.g., direct, pathoplastic, or hybrid) that each of these three models of personality shows
with PTSD is of interest, another focus of this study is to determine the degree to which each of
these three models show unique and overlapping relations to one another among individuals
diagnosed with PTSD. Beginning to formulate answers to the question of model overlap may
allow for more thoroughly articulated empirical questions regarding how personality can relate to
psychology broadly and to PTSD specifically. Understanding the unique manners in which
different personality models relate to PTSD could also aid clinicians as different types of
personality models may prove more or less useful for varying aspects of diagnostic formulation
and treatment planning.
Direct Models
Direct models, also referred to as vulnerability/predisposition/scar models (e.g. Klein,
Wonderlich, & Shea, 1993) or causal/etiological models (e.g. Widiger & Smith, 2008), propose a
linear and etiologically linked relation between personality and psychopathology. The
fundamental assumption of direct models is that one condition develops because another
condition promotes risk factors that can either trigger it or reduce the threshold for enduring it
(Klein, Wonderlich, & Shea, 1993). However, while these models presume linearity, they do not
necessarily ascribe directionality. Thus, certain personality traits may serve as a risk factor for
developing pathology or certain pathologies may alter the personality system. Consistent with a
scar model, Hermann (1992; Herman & Van der Kolk, 1987) argues that complex trauma can
lead to personality changes including disturbance of identity formation, difficulty relating to
others, and increased vulnerability to suffer both self-imposed and other-imposed harm in the

3

future. Essentially, she asserts that complex trauma can lead to “characteristic personality
changes” (1992, p. 119).
Other researchers have primarily conceptualized direct relations between these domains
as evidence that personality may influence the development of PTSD. Broadly, there is
empirical support for the concept that personality traits can predispose psychopathology in
general and PTSD in particular. Research suggests that a propensity towards negative emotional
responses to various stimuli may be considered a broad risk factor for the development of what is
commonly termed the “internalizing disorders” (e.g. Krueger, 1999) or the “distress disorders”
(e.g. Clark, Watson, & Mineka, 1994), which generally encompass the anxiety and depressive
disorders (Miller, 2003; Widiger & Smith, 2008). This propensity for negative emotional
experiences has been conceptualized as “neuroticism” (Costa and McCrae, 1992; Eysenk, 1947),
“negative emotionality” (Tellegen, 1982), “negative temperament” (Clark, 1993), and
“emotional instability” (Goldberg, 1993)

1

.

A 2005 meta-analysis by Malouff and colleagues

found that large average effect sizes (i.e., Cohen’s d > 1; Cohen, 1998) across 33 unique studies
differentiate levels of neuroticism in individuals with mood and anxiety disorders from
individuals without these disorders. High levels of neuroticism may also partially account for
the comorbidity commonly observed among the internalizing disorders (Hettema, Neale, Myers,
Prescott, & Kendler, 2006; Krueger, 1999; Lahey, 2009). Further, longitudinal studies have
found that high levels of neuroticism predict future psychopathology, (Fanous, Neale, Aggen, &
Kendler, 2007; Kendler, Neale, Kessler, Heath, & Eaves, 1993; Kendler, Gatz, Gardner, &
Pederse, 2006; van Os & Jones, 2001) perhaps suggesting a causal role of this personality trait on
1

I will adopt the term “neuroticism” throughout this manuscript to retain consistency with the
FFM, which is one of the three models being assessed in the present study. Note, however, that
results from previous research are often based on measures that contain this general propensity
for negative emotional experiences using measures other than those based on the FFM.
4

the development of psychopathology. Neuroticism has also been found to increase vulnerability
to severe consequences of psychopathology, such as suicide attempts, by as much as 225%
(Fergusson, Woodward, & Horwood, 2000). In examining various models of personality and
psychopathology, Widiger and Smith (2008, p. 758) noted:
It is now perhaps well established that the broad domain of neuroticism (or
negative affectivity) provides a personality disposition or vulnerability to a wide
range of psychopathology…within this general context, the particular mental
disorder a person high in neuroticism develops will likely be due in part to other
contributing variables.
PTSD is unlike many psychiatric disorders in that a traumatic event or events represent a
clearly-defined etiological antecedent. However, whereas prevalence estimates suggest that up
to 90% of the general population experiences a trauma during their lifetime, less than 10% of
these individuals go on to develop PTSD (Breslau, Davis, Andreski, & Peterson, 1991; Breslau
et al., 1998; Kessler, Sonnega, Bromet, Hughes, & Nelson, 1995). This implies the potential role
of a neuroticism diathesis in determining who will develop PTSD following exposure to trauma
(Miller, 2003). A growing body of empirical evidence suggests that individuals with higher
levels of neuroticism are more prone to developing PTSD (Hyer et al., 1994; McDevitt-Murphy
et al., in review; Miller, 2003; Miller, Greif, & Smith, 2003; Miller, Kaloupek, Dillon, & Keane,
2004; Miller & Resick, 2007; Sellbom & Bagby, 2009). Further, genetic vulnerability to the
development of PTSD (Afifi, Asmundson, Taylor, & Lang, 2010) may be partially accounted for
by the robust relation between neuroticism and PTSD. This is not entirely surprising given that
in the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV;
American Psychiatric Association, 1994) criterion A2 of PTSD, which is required for the
diagnosis, is met by having a negative emotional response [of fear, helplessness, or horror] to a
traumatic event. Thus, having a negative emotional response to a traumatic event is necessary to

5

the development of PTSD, and is presumably more likely to the extent that neuroticism is high.
In this way, trauma may be alternatively conceptualized as a “contributing variable” in someone
who has a propensity towards negative moods and anxiety. As Miller (2003) noted, if PTSD “is
an extreme manifestation of normal personality processes, then one might hypothesize that
trauma exposure serves to accentuate pathogenic traits present in the pretrauma personality” (p.
384, emphasis in original).
PDs also show a direct relation with PTSD (Ansell et al., in review; Bradley, Heim, &
Westen, 2005; Westen & Heim, 2003; Yen et al., 2002; Zlotnick et al., 2003). Of the ten PDs in
the DSM-IV, Borderline (BPD) tends to have the highest rates of co-occurrence with PTSD
(Westen & Heim, 2003; Yen, et al., 2002), with reported comorbidity rates as high as 68% (Shea,
Zlotnick, & Weisber, 1999). History of childhood trauma and symptom overlap are often noted
as explanations for the similarity and comorbidity between PTSD and BPD, however another
mechanism linking the two disorders may involve high levels of neuroticism. In meta-analyses
examining relations between personality traits and personality disorders, neuroticism has been
found to consistently and positively relate to all ten PDs, with the strongest correlations found
with BPD (Samuel & Widiger, 2008; Saulsman & Page, 2004). BPD is also at least moderately
correlated with all six facets comprising the higher-order trait of neuroticism and showed the
highest weighted mean effect size correlations across studies for two facets that are congruent
with DSM-IV criteria for the disorder: angry hostility (r = .48) and impulsiveness (r = .34;
Samuel & Widiger, 2008). Researchers using data from the Collaborative Longitudinal
Personality Disorders Study (CLPS; Gunderson et al., 2000), which will be utilized in the
present thesis, also found the highest rates of neuroticism in individuals with BPD (Morey et al.,
2002). Thus, the frequent comorbidity between PTSD and BPD may be due in part to a shared

6

neuroticism diathesis. In sum, normative personality trait and PD models both suggest that high
levels of neuroticism directly relate to PTSD.
Pathoplastic Models
Research indicates that some features of personality, like neuroticism, may represent risk
factors for developing PTSD or may increase following a traumatic exposure; however
heterogeneity among individuals with this diagnosis cannot be explained by this direct relation.
Personality traits with minimal zero-order correlations to disorders and limited relations to
functional severity carry the potential to account for heterogeneity in the expression of mental
illness. Such traits can be broadly construed as pathoplastic. Pathoplasticity conceptualizes
personality and psychopathology as etiologically independent, but dynamically related such that
both domains mutually influence one another (Klein, Wonderlich, & Shea, 1993; Pincus,
Lukowitsky, & Wright, 2010; Widiger, Verheul, & van den Brink, 1999). Interpersonal traits,
organized around the meta-concepts of agency and communion, are an example of personality
traits that have demonstrated pathoplastic relations to several forms of psychopathology. These
traits are commonly operationalized using the Interpersonal Circumplex (IPC; Leary, 1957;
Wiggins, 1996), which arranges these dimensions in a circular fashion (see Figure 1). Various
interpersonal theorists have used different terms to represent the more behaviorally anchored
manifestations of agency and communion. One common usage, which will be employed
throughout the present study, is that the meta-concept of agency can be measured along a
continuum of interpersonal dominance to submission and the meta-concept communion can be
measured along a continuum ranging from interpersonal coldness to warmth.
Although the pathoplasticity of interpersonal traits and PTSD has not been examined, this
model has been successfully applied to other anxiety disorders. For example, research using the

7

IPC to reflect variability in interpersonal problems (Alden, Wiggins, & Pincus, 1990) among
individuals with Generalized Anxiety Disorder (GAD) identified four types that did not differ in
age, comorbid diagnoses, or the severity of their GAD symptoms (Salzer et al., 2008). This
finding suggests that these individuals may become anxious as a result of different interpersonal
stressors. For instance, individuals with GAD who are submissive may tend to become more
anxious in situations when they are forced to assert control. Conversely, individuals with GAD
who are dominant might become more anxious in situations in which they are required to
relinquish control. While the severity of anxiety in these two groups of individuals may not
differ, their differing interpersonal dispositions will likely result in variable reactions to similar
situations. In this way, differing interpersonal styles may both moderate the expression of
psychopathology as well as “precipitate maladaptive behavior for different people, even if those
people have the same psychiatric diagnosis” (Horowitz, 2004, p. 106).
Interpersonal characteristics also appear to be pathoplastic to depression, as suggested by
conceptualizations of sociotropic/autonomous (e.g. Beck, 1983) or anaclitic/introjective (e.g.
Blatt, 1982; Blatt, Shahar, & Zuroff, 2001) subtypes. These models assert that individuals may
develop depression as a result of failure in achieving personal goals (i.e., agency) or failure in
achieving social satisfaction (i.e., communion). In fact, recent research suggests that individuals
with depression report more difficultly than individuals without a psychiatric diagnosis as a
result of being too interpersonally submissive and/or cold (Barrett & Barber, 2007). Given that
coldness (e.g. withdrawal) and submissiveness (e.g. disinterest, passivity) are symptomatic of
depression, it seems likely that increased membership in particular IPC quadrants may be related
to characteristic features of the pathology considered. Indeed, disorders that do not have
symptoms that are explicitly interpersonal, such as eating disorders, tend to show even

8

distributions across the four IPC quadrants (Hopwood, Clark, & Perez, 2007; Ambwani &
Hopwood, 2009). Though research (Barrett & Barber, 2007) suggests that individuals with
depression are more likely than their non-depressed counterparts to have specific interpersonal
difficulties (e.g. cold and submissive), pathoplastic research with other disorders generally finds
clusters of individuals with distinct interpersonal styles that do not differ with regard to comorbid
Axis I pathology, including depression (e.g., Hopwood et al., 2007; Kachin et al., 2001; Salzer et
al., 2008).
Conversely, some personality disorders have tended to show direct relations with the IPC
(Pincus, Lukowitsky, & Wright, 2010). Histrionic PD, for example, which involves dramatic,
attention-seeking interpersonal behavior, tends to project empirically into the warm-dominant
quadrant of the IPC (Wiggins & Pincus, 1989). Individuals with Avoidant PD, who tend to
avoid working in situations in which they could be criticized and are generally inhibited in
relationships, tend to cluster in the cold-submissive quadrant of the IPC (Wiggins & Pincus,
1989). Thus, an individual’s interpersonal tendencies are likely to indicate potential personality
pathology, particularly when these tendencies are rigid and pervasive.
Pathoplastic models may be particularly helpful for treatment planning because they
identify individual differences in response to social situations, which includes the therapy
context (Anchin & Pincus, 2010; Benjamin, 2003, 2005). For instance, Kachin, Newman, and
Pincus (2001) found that submissive individuals with GAD fared better after treatment than
dominant individuals and suggested that one mechanism for this relation may be that individuals
who are submissive towards the therapist are more congruent with typical client-therapist roles in
cognitive-behavioral therapy. Overall, interpersonal features have often been found to explain

9

heterogeneity and tailor treatment planning among individuals who have the same diagnosis, but
vary in their interpersonal style. However, this issue has not yet been investigated in PTSD.
Hybrid Models
Many models of personality aim to be comprehensive, and thus measures of personality
often contain some traits that tend to show direct relations and other traits that tend to show
pathoplastic relations with given forms of psychopathology. More comprehensive models can be
described as “hybrid” when they contain both of these kinds of traits. Furthermore, as these
measures and models often differentially weight various aspects of personality that may show
more or less direct or pathoplastic relations to psychopathology, hybrid models can often vary
from one another, such that different types of hybrid models may show unique relations both
with one another and with different forms of psychopathology. Because hybrid models integrate
direct and pathoplastic traits and may lead to varying conceptualizations, they carry the potential
to capitalize on the strengths of both of these frameworks, as well as one another. For example,
one measure of personality may contain more traits that directly relate to psychopathology
whereas another measure may contain more traits that tend to show pathoplastic relations. In
such a case, though both models would be hybrid, they still have the potential to increment one
another given their potentially unique relations with the form of pathology being considered.
Most research on personality models of PTSD has utilized measures that capture both
direct and pathoplastic traits, and accordingly has found hybrid relations. Using various
measures, these studies have replicated three clusters of individuals with PTSD: low pathology,
internalizing, and externalizing (Flood et al., 2010; McDevitt-Murphy et al., in review; Miller,
2003; Miller et al., 2003; Miller et al., 2004; Miller & Resick, 200; Sellbom & Bagby, 2009).
Miller and colleagues followed his (2003) review with three replication and extension studies

10

assessing male combat veterans (Miller et al., 2003, 2004) and female victims of sexual assault
(Miller & Resick, 2007), and employed a different personality measure in each study (the
Multidimensional Personality Questionnaire (MPQ; Tellegen & Waller, 2008), the Minnesota
Multiphasic Personality Inventory (MMPI) Psy-5 Scales (Harkness, McNulty, & Ben-Porath,
1995), and the Schedule for Nonadaptive and Adaptive Personality (SNAP; Clark, 1993),
respectively). The internalizing, externalizing, and low pathology clusters were retrieved in each
sample. In all three studies, neuroticism and rates/severity of PTSD were both higher in the
internalizing and externalizing groups than in the low pathology group, evidencing a direct effect
of that trait on PTSD development. Specifically, in the 2003 MPQ study, only 50% of veterans
in the low pathology cluster developed PTSD as a result of their combat exposure, while 90% of
veterans in the internalizing and externalizing clusters met PTSD criteria. While neuroticism
was directly related to PTSD in these studies, positive emotionality (PEM) and constraint (CON)
demonstrated a pathoplastic relation. Participants in the internalizing cluster had the lowest rates
of PEM whereas those in the externalizing cluster had the lowest rates of CON in all three
reports. Thus, individuals with high neuroticism and low PEM prior to trauma exposure appear
at increased likelihood of developing an internalizing PTSD response, marked by social
avoidance, anxiety, and internalizing depression, whereas those with high neuroticism and low
CON prior to trauma exposure appear to be at risk for developing an externalizing PTSD
response, marked by impulsivity, aggression, antisociality, and substance abuse. Similarly, other
differences were found between the two groups with high rates of PTSD, with individuals in the
internalizing group more likely to have co-occurring depression and panic disorder and 50%
more likely to report childhood sexual abuse and individuals in the externalizing group more
likely to have a history of substance dependence and antisocial behaviors.

11

Follow up investigations of Miller’s three cluster typology of PTSD by other researchers,
including one analysis using SNAP data from the CLPS (McDevitt-Murphy et al., in review), has
replicated low pathology, internalizing, and externalizing clusters defined by the three
dimensions of neuroticism PEM, and CON (see also Flood et al., 2010; Sellbom & Bagby,
2009). Hybrid models illustrate the importance of understanding the personality system being
explored in relation to the pathology being considered. In this vein, more inclusive models of
personality are more likely to find complex relations with psychopathology because they include
traits that tend to be directly related to psychopathology as well as traits that tend to modify the
expression of disorders. Research on PTSD using hybrid models of personality suggests that the
three traits of neuroticism, PEM, and CON account for heterogeneity in both the vulnerability to
and the expression of PTSD. An understanding of these hybrid effects can be read in Miller’s
(2003, p. 384, emphasis in original) summary statement:
One model for the influence of these dimensions on the response to trauma
suggested by this review is that [neuroticism] represents the primary personality
risk factor for PTSD which confers a direct vulnerability to the development of
the disorder following trauma exposure, while low PEM and low CON serve as
moderating factors that influence the form and expression of the posttraumatic
response.
However, while this “big 3” model of personality demonstrates a hybrid relation with
pathology, the pathoplastic traits (PEM and CON) may not fully capture the notable
heterogeneity in individuals with PTSD. For instance, personality models that incorporate the
IPC have generally found four pathoplastic groups, as opposed to two. As such, research using
the IPC could potentially increment models in which assessment of pathoplastic relations was
limited to PEM and CON in terms of capturing heterogeneity among individuals with PTSD that
is unrelated to clinical severity. Although the empirical literature supports a hybrid model
relating personality and PTSD in general, models that are intentionally hybrid using measures

12

which capture more inclusive aspects of interpersonal traits may expand current understandings
of personality and PTSD relations.
The Current Study
The purpose of this study was to compare three models of personality as each relates to
the development and/or expression of PTSD. These models were based on three distinct
personality measures that have not been previously tested in regard to their relation with PTSD.
One model was based on symptoms of DSM-IV PDs, a second model was based on the FFM,
and the final model was based on the IPC. Both the PDs and FFM, directly or indirectly, contain
aspects of neuroticism, while the IPC and FFM both contain variants of agency and communion.
As such, distinctions between these models will permit description of specific personality-PTSD
associations using previously unstudied personality models and provide a test of the general
hypothesis that the nature of the personality-psychopathology relation depends upon the
personality and psychopathology characteristics being considered.
For multiple reasons, relations between personality and PTSD were assessed using cluster
analytic procedures. This approach is consistent with the majority of research that has
empirically examined heterogeneity in individuals exposed to trauma. Additionally, there are
multiple advantages to person-centered approaches to analysis (Donnellan & Robins, in review;
von Eye & Bogat, 2006). Specifically, cluster analysis allows for descriptions of types of
individuals as opposed to descriptions of an individual’s traits. Grouping people according to
types allows for comparisons across individuals that can be based on the dynamic interplay of
common configurations of personality traits within individuals. Additionally, conveying
information to clinicians at the level of types can provide meaningful and ecologically valid
ways of understanding important individual characteristics (Donnellan & Robins, in review;

13

Shedler & Westen, 2007). Of particular importance to understanding the role of personality in
PTSD, person-centered approaches are conceptually useful because “major life events…happen
at the level of the person” (Donnellan & Robins, in review).
Hypotheses
I expected that severity-based clusters would be identified when measures that directly
(i.e., the FFM) or indirectly (i.e., the PDs) contain neuroticism were the basis for clustering, and
that these models would be hybrid in also producing clusters that were high in PTSD severity but
differed in unique ways, such as propensities for co-occurring internalizing or externalizing
pathology. I expected that clustering based upon normative interpersonal traits that have not
shown systematic correlations with psychopathology (i.e., the IPC) would lead to a pathoplastic
model in which clusters did not differ in severity. Specifically, I hypothesized:
1. Clusters based on PD or FFM traits will differ in clinical severity but clusters based
on the IPC will not.
1a. Based on previous research, I anticipated that the PDs would yield three
clusters defined by: (1) low severity, (2) high severity and mostly internalizing pathology,
and (3) high severity and mostly externalizing pathology.
1b. I anticipated that the FFM traits would yield clusters defined by both low vs.
high severity, as well as clusters with similar levels of severity but other, stylistic
differences (e.g., low vs. high agreeableness). Because the FFM contains traits in addition
to PEM and CON, the personality traits which have uniquely defined high severity PTSD
clusters in previous research, I explored the types and number of clusters this model
produced.

14

1c. Despite the absence of previous research assessing interpersonal traits of
individuals with PTSD, I anticipated that the IPC would yield four clusters that did not
differ in severity but did differ in their placement across the four quadrants of the
circumplex. This prediction was based on three factors: 1) the relative lack of
interpersonal symptoms required for a diagnosis of PTSD, 2) the fact that BPD, which
shows a moderate association with PTSD, has not shown systematic relations with a
specific IPC quadrant (Hopwood & Morey, 2007), and 3) research showing pathoplastic
relations between the IPC and other, similar disorders such as GAD.
2. Clusters based on PDs, the FFM, and IPC traits will show differing comorbidity
patterns.
2a. I anticipated that a low pathology cluster in both the PD and FFM models
would have the fewest comorbid disorders; an internalizing cluster would have the
highest rates of depression, GAD, and depressive, avoidant, and obsessive-compulsive
PDs; and an externalizing cluster would have the highest rates of substance use disorders
and antisocial and borderline PDs.
2b. Based on research using the IPC with other disorders, I expected that PTSD
would be present in all four quadrants of the IPC, and that there would be no significant
differences in rates of comorbid Axis I disorders across these groups. Conversely,
because analyses were conducted on a sample of individuals with PDs, I expected the
following comorbidity patterns with the four PDs of interest in the CLPS: rates of
schizotypal pathology would be higher in a cluster of individuals that was cold respective
to the other three clusters; rates of avoidant pathology would be higher in a cluster of
individuals that was submissive respective to the other three clusters; rates of obsessive-

15

compulsive pathology would be higher in a cluster of individuals that was dominant
respective to the other three clusters, and rates of borderline pathology would not
significantly differ across clusters (see Figure 2).
3. Clusters based on IPC traits will be more stable than clusters based on PDs or FFM
traits.
3a. In McDevitt-Murphy et al.’s (in review) study which clustered CLPS
participants with PTSD using the SNAP, 35% of participants changed groups over the
first six months of the study. The low pathology cluster was the most stable, with 74% of
individuals clustering in this group at both baseline and six month follow up. This trend
is concurrent with the decrease in PD and general clinical severity observed over the
course of the CLPS study, and suggests that as PTSD severity or clinical severity more
generally remits, individuals move from more to less severe personality clusters. Thus, I
expected that clusters based on PDs would similarly show relatively unstable cluster
membership, particularly with individuals moving from higher to lower severity clusters.
3b. Conversely, because clusters based on pathoplastic traits were not expected to
relate to symptom severity, I expected that group membership would be relatively stable
from baseline to follow up for clusters based on IPC traits.
3c. Cluster stability with the FFM was exploratory because there are divergent
reasons to expect both more and less stability in this model over time. On one hand, the
FFM contains neuroticism, which was expected to directly relate to PTSD severity.
Therefore, similar to the PD model, it could have been anticipated that as individuals
showed a general improvement in functioning, they would move from more to less severe
clusters. On the other hand, FFM traits have been shown to be relatively stable over time

16

in adult populations (McCrae & Costa, 2003), implying that cluster membership based on
FFM traits might be less likely to change.
Method
Participants
CLPS data was used to test study hypotheses. Because the aim of the CLPS study was to
better understand the course of personality disorders, numerous measures of personality were
collected, which allows for examinations of various models of personality. The sample
represents individuals from a clinical population diagnosed with BPD, avoidant (AVPD),
obsessive-compulsive (OCPD), or schizotypal (STPD) PD, or current major depressive disorder
(MDD) without a PD; participants also showed marked Axis I and II diagnostic co-occurrence
(McGlashan et al., 2000). Participants were recruited from clinical sites in four Northeastern
cities, and excluded on the basis of psychosis, current intoxication or withdrawal, IQ below 85,
age younger than 18 or older than 45, or a confused state due to organic disorders. The baseline
sample consisted of 733 participants, however, for the purposes of this study, I focused
exclusively on individuals who met DSM-IV criteria for PTSD (N = 156) at baseline. Of these
participants, the mean age was 33.31 (SD = 7.64) and the majority were female (75%, N = 117).
The racial composition of the sample was as follows: 72% Caucasian (N = 112), 18% AfricanAmerican (N = 28), 8% Hispanic (N =12), and 2% Asian (N = 3). The average number of
symptoms for the four PDs of primary interest in the CLPS are presented in Table 1. Baseline
data were used to cluster and to examine relations of external variables to cluster membership.
Baseline and 2-year follow-up data were used to assess differential stabilities across cluster
solutions.

17

Measures
Participant Selection
Structured Clinical Interview for DSM-IV (SCID; First, Spitzer, Gibbon, & Williams,
1996). The SCID assesses DSM-IV Axis I disorders and was used to determine which
individuals in the CLPS sample met criteria for a PTSD diagnosis. The SCID was administered
by experienced clinical interviewers with advanced degrees in a mental health discipline. The
kappa coefficient for inter-rater reliability of PTSD was .88 (Zanarini et al., 2000).
Personality Models
Diagnostic Interview for DSM-IV Personality Disorders (DIPD-IV; Zanarini,
Frankenburg, Sickel, & Yong, 1996). The DIPD-IV is a semi-structured interview that assesses
diagnostic criteria for DSM-IV PDs. The DIPD-IV was used to select patients for the CLPS and
was administered by research assistants trained by the test developer and masked to other study
data. Kappa coefficients for inter-rater reliability of the ten disorders in a baseline sub-sample
(N = 52) ranged from .86 – 1.0, except for paranoid PD (k = .52; Zanarini et al., 2000). I used
DIPD-IV symptom counts to cluster individuals with PTSD for the PD model.
NEO Personality Inventory, Revised (NEO-PI-R; Costa & McCrae, 1992). The NEO-PIR is 240-item self-report measure with a five-point likert scale that assesses the FFM personality
traits: Neuroticism, Conscientiousness, Agreeableness, Extraversion, and Openness to
Experience. Internal consistency reliabilities for the five domains ranged from alphas of .87 –
.92 (Morey et al., 2007). The NEO-PI-R trait scores were used to cluster individuals with PTSD
based on the FFM.
NEO-PI-R, Interpersonal Circumplex Octants. Scores on the eight octants of the IPC
(see Figure 3) can also be derived from the NEO-PI-R (Traupman et al., 2009). Internal

18

consistencies for these eight scales ranged from .58 – .77. Validity of these eight scales as IPC
constructs were also tested with RANDALL (Tracey, 1997), which assesses circumplex structure
based on 288 predictions regarding the intercorrelations of octant scales. Results yielded perfect
RANDALL structure (288/288 predictions met, Correspondence Index = 1.00, p < .001), which
suggests that this measure provides a valid representation of a circumplex model. These raw
octant scales were used to cluster individuals with PTSD based on IPC dimensions.
Validating Measures
DSM-IV Global Assessment of Functioning (GAF). GAF scores, which range from 0–
100 (lowest – highest), index the severity of psychiatric symptoms and functioning. This
measure was used as a global indicator of functional severity for each individual in the sample.
Six-month test-restest reliability of GAF ratings for the full CLPS sample was .79 (Boggs et al.,
2005).
SCID. Four categorical SCID diagnoses were also used to evaluate comorbid Axis I
disorders in the sample. Kappa coefficients for inter-rater reliability of Axis I disorders of
interest in a baseline CLPS subsample (N = 52) were as follows (Zanarini et al., 2000): alcohol
abuse/dependence (k = 1.0), drug abuse/dependence (k = 1.0), GAD (k = .63), and MDD (k =
.80).
DIPD-IV. Dimensional diagnoses from the DIPD-IV were also used to evaluate
comorbid Axis II disorders in the FFM and IPC models.
Statistical Analyses
Participants with PTSD were clustered into personality typologies using raw scores
(Aldenderfer & Blashfield, 1984; Asendorpf et al., 2001) from the PD, FFM, and IPC variable
sets. Cluster analysis is a frequently used statistical method for distinguishing groups of

19

individuals that are relatively homogenous within a heterogeneous sample, and has been the
predominant statistical procedure used in describing models of personality and PTSD (e.g.,
Flood et al., 2010; McDevitt-Murphy et al., in review; Miller et al., 2003; Miller et al., 2004;
Miller & Resick, 2007; Sellbom & Bagby, 2009). The overall goal of cluster analysis is to
minimize variance within clusters while maximizing variance between clusters (Aldenderfer &
Blashfield, 1984; Donnellan & Robins, in review). However, the determination of the optimal
number of clusters is controversial, and strictly empirical methods for making this determination
have been somewhat unreliable in previous research (Aldenderfer & Blashfield, 1984; Morey,
Blashfield, & Skinner, 1983). Accordingly, I employed four techniques to determine the optimal
number of clusters for each of the three models. I evaluated two, three, four, and five cluster
solutions for each personality model. This range was selected to add one cluster on either side of
the expected three (PD and FFM) and four (IPC) cluster solutions.
Selecting the Optimal Number of Clusters
1. First, I examined two, three, four, and five cluster solutions for each model using a
program (Clustan; Wishart, 2006) that completes a hierarchical agglomerative cluster
analysis on 1,000 random permutations of the dataset. This resampling technique
minimizes any problems created by the arbitrary ordering of participants in the data
file or other sample-specific limitations of the data. The cluster solution resulting in
the greatest improvement in the within group sums of squares was considered optimal
for each model (Schmitt et al., 2007).
2. Following methods outlined by Asendorpf et al. (2001), I randomly split the sample
in half to create sub-samples A and B. I then computed two, three, four, and five
cluster solutions using Ward’s (1963) hierarchical method. Ward’s method begins by

20

treating each individual as a cluster and then combines the two most similar clusters,
defined as those that result in the smallest squared Euclidian distance between
clusters, in a stepwise manner until the specified number of clusters has been
obtained. Ward’s method has proven to be an optimal technique for minimizing
variance within clusters (Aldenderfer & Blashfield, 1984; Morey et al., 1983). A
weakness of Ward’s method, however, is that clusters combined in one step remain
together in all subsequent steps, potentially creating non-optimal clusters
(Aldenderfer & Blashfield, 1984; Asendorpf et al., 2001). To compensate for this
weakness, the cluster centers for each solution using Ward’s method were used as
initial cluster centers for a k-means iterative partitioning clustering procedure. Kmeans is a non-hierarchical approach which assigns individuals to the cluster based
on the Euclidian distance between their scores and the nearest cluster mean
(Aldenderfer & Blashfield, 1984; Asendorpf et al., 2001). After each assignment,
cluster means are recomputed and the process is repeated. This continues until the
largest change in any cluster center is less than 2% of the smallest distance between
any two initial cluster centers (the default criterion in SPSS/PASW; Asendorpf et al.,
2001). The resulting clusters were considered the final solution for each sub-sample
in each of the three personality models. Next, individuals in each sub-sample were
assigned to the best fitting cluster in the other sub-sample based on their squared
Euclidian distance from the cluster mean in the other half of the sample (i.e.,
subsample A participants were assigned based on cluster B coefficients and
subsample B participants were assigned based on cluster A coefficients). Cohen’s
kappa (k) was used to compare agreement between cluster assignments (i.e., to

21

compare agreement in subsample A across subsample A and B clustering
assignments) and the resulting k’s were averaged. Asendorpf et al. (2001) have
demonstrated the utility of this procedure in empirically validating cluster solutions,
particularly with larger sample sizes (Asendorpf et al., 2001). In contrast, because I
am comparing the validity of four different cluster solutions for each model, the
solution (either two, three, four, or five) resulting in the highest k was favored for
each respective model.
3. I also plotted the number of possible clusters on the y axis of a graph and the fusion
coefficient values obtained using a Ward’s method clustering procedure on the x axis.
Fusion coefficients represent that value at which each case has merged to form a new
cluster in a hierarchical clustering procedure (Aldenderfer & Blashfield, 1984). Thus,
large fusion coefficients suggest substantial increases in within-cluster heterogeneity,
which is undesirable in cluster analysis. This procedure is analogous to the scree test
(Cattell, 1966) often employed in factor analysis, in which factors that provide limited
covariation among scales are unlikely to be reliable. Using this procedure, the
optimal number of clusters was considered to be the point at which the line created by
plotting the graph breaks from linearity.
4. A final method used for determining the optimal number of clusters was Mojena’s
(1977) procedure, in which the cut-point for determining the number of clusters is the
number of values that are three or more standard deviations above the mean fusion
coefficient value. Thus, if four values are more than three standard deviations above
the mean fusion coefficient value, then the optimal number of clusters in those data
would be four.

22

The methods described above are listed in relative order of their empirical support and
regularity of use in recent literature, thus the number of clusters determined by the initial
methods were given more priority than those determined by the latter if methods disagree. Given
the difficulty of empirically validating the optimal number of clusters for a solution, it was
anticipated that these four methods would not yield perfect agreement. In this case, theory and
prior research also guided my selection. Specifically, I preferred three cluster solutions using the
PDs and NEO as all previous research on personality types within PTSD has either specified or
found a three cluster solution. I preferred a four cluster solution using the IPC given previous
pathoplasticity research with other Axis I disorders, such as GAD and eating disorders. A
strength of this iterative method of determining the correct number of clusters is that cluster
solutions other than those I expected to find were also evaluated and considered.
Extracting Clusters
Once the optimal number of clusters was selected for each model, I employed the same
pattern of Ward’s and K-means clustering procedures described in Asendorpf et al.’s (2001)
procedure above on the full sample (not splitting the data into sub-samples A and B). This twostage analytic approach was chosen because it optimizes strengths of hierarchical and iterative
partitioning methods of analysis initially recognized in a Monte Carlo study to most consistently
produce valid cluster solutions (Milligan, 1981). This approach has more recently been formally
proposed by Asendorpf and colleagues (2001) and frequently employed in subsequent studies
(Asendorpf, 2003; Avdeyeva & Church, 2005; Boehm, Asendorpf, & Avia, 2002; De Fruyt et
al., 2004; Schnabel, Asendorpf, & Ostendorpf, 2002). This procedure has also been successfully
utilized in studies examining personality clusters of individuals diagnosed with PTSD
(McDevitt-Murphy et al., in review; Miller et al., 2003; Miller & Resick, 2007). The final

23

cluster solutions for each model using this two-stage procedure were used for all subsequent
study analyses.
Testing Study Hypotheses
Hypothesis 1. To test the hypothesis that clusters derived from the PDs and FFM would
differ in clinical severity but clusters derived from the IPC would not, I analyzed group
differences in each clustering model in an ANOVA framework using GAF scores as the
dependent variable and cluster membership as the independent variable. If the ANOVA yielded
significant results, post-hoc Duncan analyses were conducted to determine which clusters
significantly differed.
Hypothesis 2. To test the second hypothesis regarding differential comorbidity patterns
across clusters, I conducted an ANOVA, with potentially co-occurring PD dimensions serving as
dependent variables and cluster membership [in each of the three models] as the independent
variable. Post-hoc Duncan analyses were conducted to determine which clusters significantly
differed from one another when the ANOVA yielded significant results. I conducted chi-square
analyses with cluster membership and potentially co-occurring categorical Axis I disorders (i.e.,
substance abuse/dependence, GAD, and MDD).
Hypothesis 3. To test the final hypothesis of stability in cluster membership across time,
cluster weights obtained using baseline data were applied to follow-up data so that clustering
algorithms remained constant over time, and changes in cluster membership were due solely to
changes in personality (and not changes in the clustering algorithms themselves). I used a
binomial test to determine if the percentage of individuals who stayed in the same cluster based
on the initial assessment and a two-year follow-up assessment differed across the three models. I
also tested whether individuals moved from more to less severe clusters, as anticipated by

24

findings from McDevitt-Murphy et al. (in review), by comparing the relative proportions of
individuals moving in each direction with a binomial test.
Results
Prior to computing analyses, item responses were summed to compute overall raw scores
for all of the dimensions in each of the three personality models. One participant was missing
NEO-PI-R data, and was therefore removed from all cluster solutions. This resulted in a total
sample of 155 individuals diagnosed with PTSD.
Selecting the Optimal Number of Clusters for Each Model:
Re-sampling
The Clustan software package (Wishart, 2006) was used to determine the optimal number
of clusters for each model by validating hierarchical cluster solutions against 1,000 random trials
using a re-sampling procedure. One primary advantage of resample procedures is the ability to
correct for small sample sizes, making this procedure particularly appealing for the present
study. Consistent with expectations for each model, this re-sampling procedure yielded a three
cluster solution as the most valid for both the PD and FFM models and suggested a four cluster
solution as the most valid for the IPC.
Two-stage
As detailed in the Methods section, the sample was randomly split in half using SPSS to
create sub-samples “A” and “B.” There were no significant differences between these subsamples on age, gender, or ethnicity. Next, two, three, four, and five cluster solutions were
created by first conducting a Ward’s hierarchical cluster analysis, and then using the cluster
centers obtained from this procedure as the initial cluster centers for a k-means clustering
procedure. Profile scores for each personality model for individuals in sample A were then

25

applied to the clusters weights obtained in sample B, and the level of agreement between
clustering algorithms in each of these sub-samples was assessed with kappa. This process was
then repeated in the other half of the sample by applying individuals profile scores from sample
B to clusters weights obtained in sample A. The two kappa coefficients were then averaged, and
the cluster solution yielding the highest kappa was deemed optimal for that model.
For both the PD and FFM models, the highest agreement was obtained for the two cluster
solution. For the IPC, the four cluster solution produced the highest kappa coefficient (see Table
2). These three clustering solutions all produced a kappa of .60 or greater, the suggested value
for considering the clustering solution valid (Asendorpf et al., 2001).
The sample size in the current study was smaller than typical sample sizes in other
studies that have successfully used Asendorpf et al’s (2001) method. Consequently, after
completing these analyses, it appeared that Asendorpf’s procedure may have been particularly
sensitive to sample size, making high kappa agreements more difficult to obtain when using
solutions with a greater number of clusters. In Asendorpf et al.’s (2001) study, kappa agreement
between sub-samples was greater in clusters with larger sample sizes (N = 730) than smaller
samples (N=155). For example, when clustering sample A for the FFM, the number of
participants in each cluster in the two cluster solution ranged from 31 to 46, whereas the number
of participants in each cluster in the five cluster solution ranged from 9 to 22. Further, having
more clusters in a solution created more opportunity for any given participant to switch clusters.
This may have led to the two cluster solution superiority for both the PD and FFM models. In
light of these considerations, the superiority of the four cluster solution for the IPC model using
this method is notable. Ultimately, this two-step procedure appears useful for cluster validation,

26

but may not have been optimal for determining the best number of clusters given the sample size
in the current study.
Fusion Coefficients
For this procedure, I plotted the last 20 fusion coefficient values obtained using Ward’s
hierarchical method of clustering for each model (see Figures 4-6). This was done for ease of
visual interpretation, given that plots produced by the earlier fusion coefficients were linear for
all models. These plots can then be interpreted in the same way as scree plots are interpreted in
factor analytic procedures (Cattell, 1966), where the break in plot linearity is considered
indicative of the number of clusters (or factors) for that model. Cattell (1966) noted that data can
often produce more than one pile of scree, resulting in two or more breaks in linearity. In such
cases, later breaks in linearity may still be interpreted as representing meaningfully distinct
factors; however, later breaks may be somewhat less reliable due to a greater likelihood of
contamination due to measurement error (Cattell & Vogelmann; 1977). In the current study, the
first break in linearity for all three models was greater than six, suggesting more than six clusters
as the optimal solution. Given that I was evaluating the validity of two-five cluster solutions, the
second break in linearity based on the second slope was considered indicative of the optimal
number of clusters for each model. The second breaks in linearity indicated a five cluster
solution for both the PD and FFM models, and a three cluster solution for the IPC.
Mojena’s Method
Lastly, Mojena’s (1988) method suggests that the optimal number of clusters is the
number of fusion coefficients that are more than three standard deviations above the mean fusion
coefficient value in a hierarchical clustering procedure. The same fusion coefficients that were
obtained using Ward’s procedure and plotted in the method described above were used for this

27

procedure. Results indicated a four cluster solution for the FFM and a three cluster solution for
both the PD and IPC models.
Final Cluster Solutions
Number of Clusters for Each Model
The optimal number of clusters for each model based on the four procedures described
above is presented in Table 3. There were notable discrepancies across all models, particularly
the FFM, in which no two empirical derivation procedures agreed. Accordingly, I performed all
four procedures outlined above with the FFM facets (i.e., the 30 lower order scales of the NEOPI-R) in an attempt to clarify the optimal number of clusters for this model. However, again
none of the four procedures yielded agreement regarding the optimal number of clusters for this
model.
I also explored a second way of scoring IPC data because previous cluster analytic
research using the IPC has often analyzed individuals’ vector (i.e., agency and communion)
scores. I originally chose to analyze at the level of octant scores in order to attain a more
nuanced description of each cluster, however, given previous research using vector scores and to
maintain consistency with analyzing FFM data at two different levels (i.e., trait and facet), all
four empirical derivation procedures were also recomputed using the IPC vector scores.
Solutions from these procedures using the IPC vector scores were generally consistent with
solutions using the IPC octant scores, with the four cluster solution emerging twice as the
optimal solution.
Given discrepancies across these four empirical optimization procedures, theory and
conceptual weighting of the procedures deemed the most appropriate given the sample size of the
study served as the basis for determining the optimal number of clusters for each model. For the

28

PD model, the three cluster solution was selected for three reasons: this solution emerged as
optimal in two of the four empirical derivation procedures; based both on previous research (e.g.,
Miller, 2003) as well as the DSM-IV PD clusters, theory suggests a three cluster PD solution;
and Clustan, which generally appeared to be the most accurate empirical derivation procedure
across all models for the current study, suggested three clusters for the PD model.
The optimal number of clusters for the FFM model was less evident based on empirical
procedures, but ultimately a three cluster solution was chosen. The reasons for this decision
paralleled those for the PD model. First, though no two empirical procedures were in accord, the
Clustan re-sampling procedure indicated a three cluster solution. Second, while no cluster
analytic research has been done using FFM scores of participants with PTSD, theory generally
suggests a three cluster solution for this model (e.g., Donellan & Robins, 2009).
The optimal number of clusters for the IPC model was more readily apparent than for the
other two models. The four cluster solution emerged as the most valid for the two empirical
procedures given the most priority based on current empirical support (Clustan re-sampling and
Asendorpf et al.’s (2001) procedure). Further, theory unambiguously supports four cluster IPC
solutions for conditions without clear and specific interpersonal content. The four cluster
solution was also the most frequently valid solution across both IPC octant and IPC vector
scores.
Cluster Characteristics for Each Model
Within sample T-scores computed for all variables used in the clustering procedures are
presented for each personality model in Tables 4-6. There were no significant age, gender, or
racial differences between clusters in any of the three models (see Table 7). The three clusters
recovered for the PD model fall in line with expected patterns of internalizing, externalizing, and

29

low severity pathology groups. The low severity cluster had the lowest score on all ten PD
symptoms counts. Additionally, this cluster had the lowest FFM neuroticism score and the
highest FFM agreeableness and conscientiousness scores; a pattern of FFM traits which has been
linked to low PD severity (Morey et al., 2002). Consistent with internalizing patterns, cluster
two had the highest scores on the avoidant, dependent, and obsessive-compulsive PD scales
(Hopwood, Ansell, Fehon, & Grilo, in press; Paris, 2003). These PDs are similarly grouped
together under the “Cluster C” label in the DSM-IV because they are broadly defined by high
levels interpersonal anxiety. The internalizing cluster was also defined by the highest levels of
neuroticism based on FFM data. Cluster three displayed patterns consistent with externalizing
dimensions, with the highest scores on most of the other PD scales, and notably high rates of
antisocial personality pathology (Krueger, 1999). In accord with expected patterns, these
individuals also had the lowest rates of agreeableness on the FFM and low warmth coupled with
high dominance on the IPC. Thus, the final picture that emerged when clusters were based on
PD scores was that individuals in this sample could be captured according to three broad
typologies: 1) individuals with minimal personality pathology, 2) individuals who tend to have
high levels of distress and are often either overly avoidant of or dependent on others (or who
may be in conflict regarding these competing motives), and 3) individuals who have high rates of
destructive behavior who are generally controlling of and not interested in getting along with
others.
For the FFM, the three clusters appear to be primarily severity based, with one low
severity cluster and two higher severity clusters, both of which had high neuroticism scores but
differed on other FFM traits. The two severe groups appeared to be distinguished by the
presence/absence of protective factors. Given that neuroticism is generally related to levels of

30

distress, and that other traits can serve protective functions (Hopwood, Thomas & Zanarani, in
review) the three FFM clusters were labeled: “low distress,” “resilient distress,” and “vulnerable
distress.” The low distress cluster was defined by the lowest neuroticism scores and the highest
conscientiousness scores. This cluster also displayed the lowest rates of PD pathology, and in
general, had high scores on IPC warmth and dominance. The resilient distress cluster had high
neuroticism scores, but also had relatively high scores on extraversion, openness, and
agreeableness, traits which may serve to buffer somewhat against the generally high levels of
distress individuals in this group often experience. This group of individuals also had somewhat
higher rates of PD symptomatology than individuals in the low distress cluster, but lower rates of
PD pathology than individuals in the vulnerable distress cluster. The third, vulnerable distress,
cluster had similarly high neuroticism scores to the resilient distress cluster, however this group
also had the lowest scores on all of the other four FFM traits. This group of individuals also had
the highest rates of all ten PDs, and were generally interpersonally cold and submissive.
Ultimately, clusters produced using the FFM yielded three types that are non-redundant with the
three PD types. Individuals clustered using the FFM can generally be distinguished based on
their tendencies to be either: 1) relatively free from global distress and more conscientious,
agentic and communal than others within this sample, 2) high in distress, but relatively resilient
due to other protective personality traits such as conscientiousness, extraversion, and
agreeableness, or 3) high in both distress and personality pathology with an absence of other
protective personality traits and with limited resources to either get along (i.e., communion) or to
get ahead (i.e., agency).
The four IPC clusters arrayed across the four circumplex quadrants. Cluster one was
warm-dominant, cluster two was warm-submissive, cluster three was cold-submissive, and

31

cluster four was cold-dominant. All of the octant patterns fit theoretical expectations (e.g., the
cold-dominant and cold-submissive clusters both had relatively high DE scores, only the coldsubmissive cluster had a relatively high FG score, etc.), with the sole exception of the colddominant cluster having a relatively high mean NO score. This indicates that individuals in this
cluster are generally cold and dominant, but that the stronger interpersonal pattern for the
individuals in this cluster is a tendency to be dominant. In addition to their IPC traits,
individuals in the warm-dominant cluster generally had the lowest rates of all DSM-IV
personality pathology. They also showed low levels of neuroticism and high levels of the other
four FFM traits, a pattern which generally corresponds with less risk of maladaptive personality
trait configurations (Samuel & Widiger, 2008; Saulsman & Page, 2004). Individuals in the
warm-submissive cluster also tended to have low rates of PD pathology and, consistent with IPC
expectations, high levels of FFM agreeableness. Individuals who were in either of the “cold”
clusters generally had higher rates of PD symptoms than individuals in either “warm” cluster.
More specifically, individuals who were cold and also submissive had the highest levels of
neuroticism and substantially lower rates of extraversion than any other group of individuals.
Lastly, in addition to high rates of PD symptoms, individuals who were cold and dominant also
displayed the lowest levels of FFM agreeableness. Ultimately, the IPC clusters did not prove to
be purely pathoplastic, with neuroticism showing a direct link with cold-submission and the PDs
demonstrating a direct relation with coldness. Though the four IPC clusters did differentially
align with the IPC quadrants, cluster placements were not wholly symmetrical. Specifically, the
cold-dominant cluster was more defined by a high dominance score whereas the warm-dominant
cluster was more defined by high warmth (see Figure 7).

32

Cluster Membership across Models
Cluster membership for the three personality models was compared in order to determine
trends and consistencies across these models (comparisons of each of the three models are
presented in Tables 8-10). Given the severity effect found in both the PD and FFM models, the
number of individuals who were in both the “low pathology” PD cluster and the “low distress”
FFM cluster was of particular interest. Of the 68 individuals who were in the low severity FFM
cluster, 44 were also in the low pathology PD cluster (64.71%). Individuals in the low pathology
PD cluster were also more likely to be in a warm than in a cold IPC cluster. Individuals in the
internalizing PD cluster were relatively evenly spread across clusters for both the FFM and IPC
models. This suggests that individuals may be vulnerable to internalizing symptoms for a variety
of reasons, and that the typical profile of an individual with internalizing personality pathology
and PTSD may be less specific than that of individuals with low or externalizing personality
pathology and PTSD. Individuals in the externalizing PD cluster were less likely to be in the
resilient distress FFM cluster, suggesting that these individuals are more likely to have low
distress, or, if they have higher distress, to have fewer protective resources. Further, half of the
individuals in the externalizing PD cluster were in the cold-dominant IPC cluster, suggesting
general interpersonal problems related to being controlling and unfriendly, with problems most
specifically related to a lack of warmth. This trend was expected, as previous research has linked
antisocial personality pathology with cold-dominance using the IPC (Hopwood & Morey, 2007;
Wiggins & Pincus, 1989).
Comparing individuals in the FFM model with the PD and IPC models, those who were
originally in the low distress FFM model were most likely to be warm-dominant and least likely
to be cold-submissive in the IPC. Individuals in the resilient distress FFM cluster were most

33

likely to be internalizers and very unlikely to be externalizers in the PD model. Individuals in
the resilient distress cluster were also more likely to be in a warm IPC cluster than in a cold IPC
cluster. Individuals in the vulnerable distress FFM cluster were most likely to be internalizers in
the PD model. Notably, no one in this group was in the warm-dominant IPC cluster; the
majority were cold-submissive.
Although I anticipated pathoplasticity when cluster solutions were based on the IPC,
there was some evidence of direct relations when comparing individuals’ IPC cluster placement
with cluster placement in the other models. Specifically, the majority of individuals in the
warm-dominant cluster were in the low pathology and low distress PD and FFM clusters.
Individuals in the warm-submissive cluster were also somewhat more likely to be in the low
distress than other FFM clusters, and tended not to be in the externalizing PD cluster. Half of the
individuals in the cold-submissive cluster were in the internalizing PD cluster and more than half
were in the vulnerable distress FFM cluster. Finally, cold-dominant individuals were least likely
to be in the low pathology PD cluster but most likely to be in low distress FFM cluster. This
may suggest that this group of individuals have particularly high rates of PD symptoms, but are
generally less distressed about their pathology, as evidenced by relatively low levels of
neuroticism. In sum, comparing individuals’ cluster placement across the three models
highlights that knowing a person’s pathological symptoms (e.g., PDs), personality traits (e.g.,
FFM), and interpersonal style (e.g., IPC) generally provides non-redundant information.
Testing Study Hypotheses
Hypothesis 1
My first hypothesis was that both the PD and FFM models would yield severity-based
clusters in which a lower severity cluster would have higher GAF scores than more severe

34

clusters, whereas the IPC would yield clusters that were not defined by severity, and thus would
not have significantly differing GAF scores. For the PD model, a one-way ANOVA with cluster
membership as the independent variable and baseline GAF score as the dependent variable
yielded significant differences across the three clusters, with the low pathology cluster having
higher GAF scores than either the internalizing or externalizing clusters (see Table 4). It was
similarly predicted for the FFM that a low severity cluster would have a higher GAF score than
other clusters produced in this model. Results from a one-way ANOVA failed to support this
hypothesis, with none of the clusters significantly differing in GAF score (see Table 5).
However, based on other descriptive patterns, individuals in the low distress cluster generally
seemed healthiest while individuals in the vulnerable distress cluster generally seemed the most
at risk. Consequently, GAF scores may not have been an optimal measure for detecting finergrain differences in functioning between these groups of individuals. For the IPC model, no
cluster was anticipated to have a higher mean GAF score than any other cluster. As expected, a
non-significant one-way ANOVA failed to provide support for GAF differences across IPC
clusters (see Table 6).
Hypothesis 2
The second study hypothesis was that there would be differing patterns of comorbidity
across the three personality models. Patterns of comorbidity with Axis I disorders were assessed
with chi-square tests (see Table 11). There were no significant differences in rates of MDD
across clusters for any model. However, this likely related to the high base rates of MDD in the
sample and their relatively healthy status, both of which are a function of sample selection. As
such, this null result would be unlikely to generalize to other samples. In accord with study
hypotheses, the externalizing PD cluster had the highest rates of alcohol and substance use

35

disorders (AUD and SUD, respectively). Contrary to predictions, the internalizing cluster did
not have higher rates of generalized anxiety disorder (GAD). For the FFM, the vulnerable
distress cluster had the highest rates of GAD, AUD, and SUD. No differences in Axis I
disorders were anticipated for the IPC clusters, however varying rates of GAD and AUD were
found across clusters, with individuals in the warm-dominant cluster having lower rates of both
of these disorders than individuals in any of the other three clusters.
Personality pathology based on PD symptom counts was discussed when describing
cluster characteristics; however it will be detailed further in light of specific study hypotheses.
Predictions regarding comorbid personality pathology for each of the three models were tested
using one-way ANOVAs with cluster membership (for each model) as the independent variable
and dimensional DIPD-IV scores as the dependent variables (results are presented in Tables 4-6).
For the PD model, it was anticipated that the internalizing cluster would have the highest rates of
AVPD and OCPD and an externalizing cluster would have the highest rates of ASPD and BPD.
Indeed, the internalizing cluster had significantly higher rates of AVPD than either of the other
two clusters. Rates of OCPD did not significantly differ between the internalizing and
externalizing cluster, however OCPD was higher in both of these groups than in the low
pathology cluster. This same pattern was found for BPD, with rates not differing between the
internalizing and externalizing clusters, but being significantly higher in both of these clusters
than in the low pathology cluster. Rates of ASPD were significantly higher in the externalizing
cluster than in either of the other two clusters. Notably, there were significant differences across
clusters for all ten PD scales.
The same personality pathology patterns were expected for the FFM model as for the PD
model; however the clusters derived based on FFM traits did not correspond in the same ways to

36

internalizing and externalizing patterns as they did for the PD model. Instead, clusters derived
using the FFM traits corresponded more closely to a spectrum model ranging from low distress
to high distress with and without the presence of other protective personality traits. In general
alignment with study hypotheses, rates of AVPD were significantly higher in both high distress
clusters; whereas rates of ASPD were significantly higher in the vulnerable distress cluster than
in either of the other two clusters. Unexpectedly, there were no significant differences across
clusters in rates of BPD or STPD.
For the IPC, it was anticipated that colder clusters would have higher rates of STPD.
Consistent with this prediction, both the cold-dominant and cold-submissive clusters had higher
rates of STPD than either of the warm clusters. Rates of AVPD were expected to be higher in
submissive clusters. This hypothesis was supported, however rates of AVPD were unexpectedly
also high for the cold-dominant cluster. The hypothesis that OCPD would be highest in
dominant clusters was supported, with the highest rates present in both the cold-dominant and
warm-dominant clusters. Lastly, rates of BPD were not expected to differ across clusters.
Consistent with expectations, non-significant results from a one-way ANOVA failed to support
different rates of BPD across clusters.
Hypothesis 3
The final study hypothesis was that cluster membership based on IPC traits would be
more stable than clusters based on PD or FFM traits. Specifically, given the overall
improvement observed during the course of the CLPS, it was expected that individuals would
move from more to less severe PD clusters, and that individuals who originally placed in the low
pathology PD clusters would not generally move to clusters defined by more severe pathology
(i.e. internalizing or externalizing). Stability using the FFM model was exploratory given that,

37

on one hand, it was expected that, similar to the PD model, individuals would move from more
to less severe clusters at follow-up; however on the other hand, FFM traits have generally shown
high stability in adult populations (McCrea & Costa, 1990). Follow-up data were available for
132 participants from the original sample for the PD model and for 122 participants for the FFM
and IPC models. The specific aim of this hypothesis was to understand the direction of change
in cluster membership across these models. Therefore, I examined whether the percentage of
people who moved to a healthier cluster was significantly greater than the percentage that
remained in or moved to less healthy clusters. Based on results from previous analyses, the
“healthy” clusters for each model were considered the low pathology PD cluster, the low distress
FFM cluster, and the warm-dominant IPC cluster. Results from a binomial test comparing the
number of individuals who moved to a healthier cluster to the number of individuals who
remained in or moved to less healthy clusters were significant for all three models (p < .05).
This finding was consistent with predictions for the PD model, but not for the IPC model. Rates
of agreement between baseline and follow-up clusters as well as the percent of individuals in
each cluster at baseline compared to follow-up are presented for each model in Tables 12-14.
The percentage of individuals who placed in the healthy cluster at follow-up rose by at least 25%
for all models. Conversely, membership in all of the less healthy clusters decreased from
baseline to follow up across all three models.
Discussion
The purpose of this study was to better understand variability among individuals with
PTSD in the context of three models of personality. Each model was hypothesized to uniquely
relate to functional severity, personality pathology and psychopathology, and stability in a
sample of individuals diagnosed with PTSD. Specifically, personality disorder (PD) and five-

38

factor (FFM) models were expected to show hybrid relations with PTSD, whereas an
interpersonal circumplex (IPC) model was expected to demonstrate a pathoplastic relation to
PTSD. Each of these personality models was tested among individuals in the Collaborative
Longitudinal Personality disorder Study (CLPS) sample who were all diagnosed with PTSD in
addition to either a PD or MDD. Scales from each of the three personality models were cluster
analyzed in order to derive groups of individuals with PTSD with distinguishing personality
characteristics. Overall, and consistent with the study’s broad hypothesis, the resulting
typologies varied meaningfully, suggesting that personality variability in PTSD depends on how
personality is conceived. More specifically, results from the present study can be organized into
four broad categories, which are each discussed in turn: (1) characteristics of these three distinct
models of personality heterogeneity in PTSD, (2) relations between personality models, (3)
change in personality models over time, and (4) methodological considerations. Limitations,
implications, and future directions are also addressed.
Three Distinct Models of Personality and PTSD
Results from the present study indicate that the PD, FFM, and IPC models of personality
each demonstrate a unique relation to severity and psychopathology in a sample of individuals
diagnosed with PTSD. Consequently, this study provides important implications for building on
the current understanding of relations between personality and PTSD. The PD model yielded a
three cluster solution whose clusters, as expected, were characterized by low pathology,
internalizing, and externalizing characteristics and comorbidity patterns. These clusters are
consistent with Miller’s (2003) typology of personality clusters of individuals with PTSD, as
well as with subsequent research that has examined personality-based clusters of PTSD using
personality models that mix normal and abnormal traits. In this model, individuals in the low

39

2

pathology cluster tended to have the lowest rates of co-occurring psychopathology, relatively
better global functioning, and greater variability in interpersonal personality styles. The
internalizing PD cluster had the highest rates of DSM-IV “cluster C” personality pathology as
well as the highest levels of neuroticism, suggesting that a typical individual in the internalizing
cluster is likely to be significantly distressed and will tend to have particular problems related to
maladaptively avoiding and/or depending on others. The externalizing PD cluster could
primarily be described by notably high rates of antisocial personality pathology, with other
results indicating that these individuals are characteristically antagonistic, unfriendly, and
dominant, and tend to have higher rates of alcohol and substance use disorders than individuals
in other clusters.
Results using the FFM also yielded a three cluster solution; however groups produced
using FFM traits showed only modest overlap with those produced using PD symptoms. Similar
to the low pathology PD group, the FFM yielded a low distress group characterized by low levels
of neuroticism and high levels of conscientiousness. Roughly half of the sample was classified as
low distress using the FFM, while the remaining half was split nearly evenly between two
distinct, high distress clusters. Both high distress clusters had similarly high levels of
neuroticism, but were otherwise distinguished by the relative presence or absence of other
generally protective personality traits, namely: extraversion, agreeableness, and
conscientiousness. Individuals in the “resilient distress” cluster had relatively high levels of the
other, generally adaptive, FFM traits. In contrast, individuals in the “vulnerable distress” cluster

2

It is important to note that, as this study used a sample of individuals diagnosed with at least
two DSM-IV disorders, the low pathology group is not healthy in a general sense, but rather
these individuals display greater mental well-being and lower rates of additional
psychopathology relative to other individuals within the sample. This is true for the “healthy”
cluster in all three personality models.
40

had low levels of all of the other traits. These vulnerable distress individuals had the highest rates
of personality pathology and the greatest level of difficulties with diagnostic comorbidity related
to anxiety, alcohol, and substance use.
Although the PD and FFM models both demonstrated hybrid relations with PTSD, in
many ways the FFM more closely resembled a direct model. The internalizing and externalizing
clusters in the PD model each displayed worse outcomes than the low pathology cluster, but did
not generally differ in functional severity or rates of comorbid psychopathology. In contrast, a
severity effect was evident across the three clusters of the FFM. Low distress individuals were
generally the healthiest and vulnerable distress individuals generally had the highest rates of
comorbid psychopathology, with individuals in the resilient distress group tending to fall
between these two clusters on most measures of functional severity and psychopathology. This
severity effect, however, cannot be solely attributed to high neuroticism, as the levels of
neuroticism did not differ between the resilient distress and vulnerable distress clusters.
Corresponding with expectations from interpersonal theory, empirically derived clusters
using the IPC arrayed across the quadrants of the circumplex, with the number of individuals in
each cluster fairly balanced across groups. These clusters were accordingly labeled with respect
to the presence/absence of interpersonal warmth and dominance. Individuals in the warmdominant cluster generally are disposed to get along well with others (i.e., communion) and to
achieve personal goals (i.e., agency). Additionally, these individuals tend to have the lowest
rates of pathology and global distress, coupled with high levels of extraversion, agreeableness,
and conscientiousness. Individuals in the warm-submissive cluster are generally friendly,
agreeable, and passive in their interactions with others, and display relatively low rates of
personality pathology. Individuals in the cold-submissive cluster tended to have relatively high

41

levels of general distress (i.e., neuroticism) and limited interest in engaging with others,
evidenced by low levels of both IPC warmth and FFM extraversion. Individuals in the colddominant cluster tended to have the highest rates of personality pathology and a general tendency
to be antagonistic when dealing with others.
As demonstrated by the nature of the cluster solutions in each model, all three models
capture unique correlates and stylistic elements that may be useful in understanding personality
heterogeneity among individuals with PTSD. This indicates that clinical diagnoses can be
informatively supplemented when considered in parallel with models of personality, and that it
may be useful to consult multiple domains of personality for the most thorough representation of
relevant personality attributes. For example, knowing that an individual with PTSD and MDD
also has high levels of internalizing personality pathology provides incrementally useful
information; namely, that these two diagnoses may largely be due to a shared relation to a
general internalizing disposition. Likewise, knowing that an individual with PTSD and MDD
has a trait profile connoting “resilient distress” based on the five high-order personality traits
provides a more complete understanding than knowing her DSM diagnoses alone. With this
additional information, a clinician might infer that, despite her high levels of general distress
(i.e., neuroticism), she is comfortable around and agreeable with others and may be effectively
able to work towards goal directed activities (i.e., extraverted, agreeable, and conscientious). In
the same vein, clinicians could better treat and understand the stressors of an individual with
comorbid PTSD and MDD diagnoses if they also know that she has a warm and submissive
interpersonal style. While each model of personality can usefully increment diagnosis in the
aforementioned ways, knowing an individual’s personality type based on all three of these
models provides the most complete pictures of who she is and what her most significant

42

problems in living are likely to be. This information can assist clinicians in developing
hypotheses about the most beneficial ways to address these problems therapeutically.
Specifically, an individual with PTSD and MDD that can be broadly classified according to the
personality types of “internalizing,” “resilient distress,” and “warm-submissive” is likely to have
general problems related to high levels of distress and more specific problems related to
avoidance and dependence. She is also likely to be motivated to be close to others, but to
display minimal interpersonal control in her interactions. An effective treatment approach for
this individual might be utilizing techniques aimed at reducing her mood and anxiety symptoms,
reinforcing her own resilient personality traits, particularly as they relate to her interest in being
with others, and increasing interpersonal assertiveness.
Relations between Personality Models
An interesting aspect of this study was the ability to compare how the same individuals
could be categorized according to different models of personality. In this study, each of the three
models tested demonstrated a hybrid relation with PTSD. However, despite their common
hybrid categorization, these models all related to one another differently, suggesting that each
model assessed unique aspects of personality-psychopathology relations. For instance, relations
between the PD and IPC models were largely influenced by interpersonal warmth, such that
individuals who were in either warm IPC cluster tended to have lower levels of personality
pathology than individuals in either cold cluster. A different direct pattern emerged between the
IPC and FFM models in which relations between the IPC and FFM were more influenced by
dominance. Specifically, on all FFM traits aside from agreeableness, the warm-dominant and
cold-dominant clusters did not significantly differ. This suggests that relations between the FFM
and IPC may relate more systematically to dominance than to warmth. Though further research

43

is needed to better understand relations across these models, patterns of covariation in this study
between the PD, FFM, and IPC models may suggest that high levels of global distress are
generally more associated with interpersonal submission whereas high levels of personality
disorder pathology relate most strongly with interpersonal coldness.
Further analyses of the relations between these models indicated that the externalizing
cluster was primarily driving interpersonal differences in the PD model and the vulnerable
distress cluster was primarily driving differences in the FFM (see Figure 8, which locates each
PD and FFM cluster within the IPC). In the PD model, the externalizing cluster was both colder
and more dominant than either the internalizing or low pathology cluster, which did not
significantly differ from one another (see Table 4). In the FFM, the vulnerable distress cluster
was interpersonally colder and more submissive than either the low pathology or the protected
distress cluster, which did not significantly differ from one another (see Table 5). This pattern of
findings suggests that individuals who have high rates of externalizing personality pathology are
most likely to have difficulties or limited interest in getting along with others coupled with a
tendency to be interpersonally controlling, whereas individuals who have high levels of general
distress (i.e., high neuroticism) are most likely to be protected from the adverse influences of
negative emotionality as a consequence of their interpersonal warmth and dominance as
compared to their peers who also have high levels of general distress but lack these protective
interpersonal traits. Overall, these patterns suggest that broad distress, in the form of
neuroticism, is contextualized by the patterning of other normal and abnormal characteristics. In
this way, the direct relation between neuroticism and PTSD was not as simple, or linear, as
expected. This suggests that personality traits, such as neuroticism, that directly relate to

44

psychopathology when studied in isolation, may show hybrid relations with psychopathology
when considered in conjunction with other direct or pathoplastic personality traits.
Change in Personality Models over Time
Previous CLPS studies have shown that individuals in this sample improved in terms of
diagnostic status and functioning over time. For instance, previous examinations of PD stability
using the CLPS sample found that over half of the participants initially recruited on the basis of a
PD diagnosis showed full remission two years later (Grilo et al., 2004; Shea et al., 2002; Skodol
et al., 2005). Likewise, in the present study participants tended to move from more to less severe
clusters in all three personality models from baseline to follow-up (see Tables 12-14). This
general improvement in personality functioning subsequent to a PTSD diagnosis may suggest,
with regard to the scar model (Herman, 1992), that as PTSD symptoms remit, individuals’
personality “scars” begin to heal.
Overall, the movement towards healthier clusters was anticipated for the PD model.
However, descriptively interesting distinctions between the internalizing and externalizing
clusters also emerged here. The majority of individuals who were classified as internalizers at
baseline were in the low pathology group at follow-up. Conversely, half of the individuals who
were in the externalizing cluster at baseline remained in this cluster at follow-up, with the other
half moving to both the low pathology and internalizing clusters. This may suggest that
individuals in the externalizing cluster generally had lower rates of improvement relative to low
pathology and internalizing individuals. It is also possible that this trend was due to the
relatively smaller sample size within this cluster. In evaluating the FFM, it was initially striking
that no one could be classified as having resilient distress at follow-up; however this is quite
consistent with the good prognosis conferred by having these protective traits. Indeed, the

45

overwhelming trend of individuals who were initially classified as having resilient distress was a
movement to the low distress group (24 of 29 participants; 83%). Many individuals in the
vulnerable distress group at baseline also moved to the low distress cluster at follow-up, though,
as expected based on having a less adaptive configuration of FFM traits at baseline, this
percentage was smaller than that of the resilient distress group. This seems to indicate that the
resilient distress cluster within the FFM model fairly accurately captures individuals who might
be most amenable to improvement, despite their high distress, and partly because of other
protective traits.
In examining cluster stability over a two-year period, it is important to highlight
discrepancies between the current study and McDevitt-Murphy et al.’s (in review) study that also
examined clusters of individuals in the CLPS sample diagnosed with PTSD. They found cluster
membership to be stable from baseline to follow-up for 65% of individuals, a much higher
percentage than was found in the present study. These discrepant results could be in part due to
different measures utilized (the SNAP versus the DIPD-IV and NEO-PI-R); however they are
most likely due to different approaches to computing follow-up clusters. McDevitt-Murphy and
colleagues re-computed clusters using follow-up data, and found three clusters that roughly
corresponded with low pathology, internalizing, and externalizing characteristics. In contrast, I
chose to examine cluster stability by applying follow-up data to clusters computed using baseline
data. In this way, stability was based on whether individuals changed over a two-year period,
and not on whether clusters changed over a two-year period. As expected, less stability was
observed when the question was framed as one of individual change. Despite these differences,
both studies demonstrated greater movement towards generally healthier and lower pathology
clusters among those whose cluster membership changed.

46

Methodological Considerations
As anticipated, the PD and FFM models both displayed a hybrid relation with PTSD. For
each of these models a quantitative severity effect separated the lower severity groups from two
unique, higher severity groups. However, within this level of quantitatively high severity there
was a qualitative distinction between the two types of groups in each model. It seems that the
two high distress clusters in the PD model convey whether a person is more likely to experience
internalizing or externalizing distress, whereas the two high distress clusters in the FFM indicate
how capable a person is likely to be in tolerating her distress. For example, knowing whether an
individual with PTSD can be classified as internalizing or externalizing indicates the type of
comorbid psychopathology he is most likely to have. The high distress FFM clusters provide
unique information in that, regardless of whether an individual can be categorized as having an
internalizing or an externalizing disposition, he is more apt to be able to cope with his potentially
pathological characteristics if he is also conscientious, extraverted, and agreeable.
An unexpected finding was that the IPC was not a purely pathoplastic model in this
study. The relatively even distribution of individuals across the four clusters initially indicated
that this model showed more pathoplasticity than either the PD or FFM model; however, similar
to both of these models, IPC clusters showed direct relations to other forms of pathology as well
as to functioning outcomes. This is particularly surprising in light of previous research using the
IPC with other populations, which has generally not found systematic relations between IPC
clusters and disorders such as GAD and bulimia. Further, the IPC model, similar to the other
two models, demonstrated a general pattern of individuals moving to healthier clusters at follow-

47

up. The general pattern that emerged across a variety of tests was that the warm-dominant
cluster was healthier than the other three IPC clusters. The hybrid nature of the IPC in this study
is most likely due to a combination of two factors: (1) measurement and (2) systematic relations
between the IPC and personality pathology.
Despite the NEO-PI-R derived IPC items proving adequately reliable and conforming
well to circumplex structure, it nonetheless was not an ideal measure of the IPC. As with several
other IPC-based measures, items used to measure each of the octants using the NEO-PI-R varied
in terms of adaptive valence, with more desirable items on the warm side and less desirable items
on the cold side of the IPC. For example, items corresponding to the NO (warm-dominant)
octant were “sometimes I bubble with happiness” and “I laugh easily.” Examples of items
loading on the FG (cold-submissive) octant were “I am not a cheerful optimist” and “I am not as
lively as other people.” These cold-submissive items clearly contain elements of
psychopathology not present in the warm-dominant items, particularly depressive
symptomatology. Indeed, there has long been a tension in interpersonal theory between the
theoretical postulate that interpersonal warmth can be problematic and interpersonal coldness
adaptive and the measurement difficulties associated with developing good items to assess
pathological warmth and adaptive coldness (Ethier, Sadler, & Woody, 2009; Hatcher & Rogers,
2009). Many studies that have demonstrated purely pathoplastic relations between the IPC and
various forms of psychopathology have used a version of the Inventory or Interpersonal
Problems (Horowitz, Rosenberg, Baer, Ureno, & Villasenor, 1988), a measure that deals with
this problem by removing elevation, which is essentially an indicator of global levels of
interpersonal distress, and ipsatizing directional vector scores around individual’s personal
means. This limits the relation between distress and the nature of interpersonal problems,

48

perhaps also minimizing hybrid aspects of the IPC model. As such, pathoplasticity is more
likely to be observed using IPC models when IPC measures do not conflate psychiatric severity
or distress with interpersonal style.
A second possibility is that difficulty measuring maladaptive warmth reflects the
overarching truth that warmth is more adaptive than coldness. Theoretical conceptualizations
have considered high levels of and strivings towards both agency and communion as a general
recipe for optimal living (Bakan, 1966; Freud, 1930). In this way, human satisfaction can be
viewed as effectively loving (i.e. communion) and working (i.e. agency). In this study,
personality pathology was related to the tendency to be interpersonally cold. Previous research
has similarly linked personality pathology primarily to differentiation with regard to agency, and
only minimally to communion (Morey, 1985). This research was, however, also conducted with
a measure (the Interpersonal Checklist; Laforge & Suczek, 1955) in which traits related to
warmth were generally more desirable than traits related to coldness. Ultimately, answers to the
important question of whether warm-dominance connotes a relatively healthy interpersonal style
will require better articulation and measurement of the primary associated benefits and problems
of all four quadrants of the IPC (Hopwood, Koonce, & Morey, in press).
A final methodological consideration that emerged in the present study was the utility of
examining personality types as opposed to personality traits. The empirical effectiveness of
types remains ambiguous as some research suggests types do not increment traits in predicting
functioning or psychopathology (Asendorpf, 2003; Costa et al., 2002) whereas other research
indicates that types do indeed provide incremental validity over and above traits, particularly
with regard to long-term prediction (Asendorpf & Denissen, 2006; Hart, Atkins, & Fegley,
2003). However, aside from predictive power, describing personality at the level of types may

49

be a more efficient and effective means of conveying meaningful aspects of personality,
particularly for psychologists who are less familiar with nuances in the personality literature
(Donnellan & Robins, in review). It is more efficient to tell a clinician that her patient,
diagnosed with PTSD, can be classified as having vulnerable distress than to tell the same
clinician that her patient has high levels of FFM neuroticism and low levels of extraversion,
openness, agreeableness, and conscientiousness. Perhaps more important than efficiency,
categorization at the level of types permits description of the important ways that personality
traits can relate to one other; a possibility that can be lost in simple trait approaches to
description. For instance, even though having high levels of neuroticism and low levels of the
other four traits connotes a vulnerable distress profile, the dynamic interplay between these traits
is more readily apparent using a typological description. Ultimately, the debate as to whether
personality is best conceptualized at the level of types or at the level of traits in some ways
parallels the debate regarding whether personality relates to psychopathology in a direct or a
pathoplastic manner. Just as hybrid models of personality-psychopathology relations carry the
potential to capitalize on the strengths of both direct and pathoplastic models by combining these
frameworks, it can also be useful to discuss personality with regard to both types and traits. For
example, two people could both be typologically categorized as having vulnerable distress when
one individual has a neuroticism trait score of 60T and the other has a meaningfully higher
neuroticism trait score of 110T. In this way, types and traits can be usefully considered in
parallel instead of in competition.
Limitations, Implications, and Future Directions
Though this study augmented the current empirical understanding of relations between
personality and PTSD in several important ways, some limitations are noteworthy. With regard

50

to measurement, IPC octants derived from NEO-PI-R scales, though validated in previous
research (Traupman et al., 2009) and corresponding perfectly to circumplex structure according
to the RANDALL procedure (Tracey, 1997), may not have provided the best representation of
interpersonal theory, as these items were not constructed with the IPC in mind. Additionally, the
FFM and IPC models shared items and this may have differentially impacted how these models
related to one another relative to each of their relations with the PD model. Further, data for the
FFM and IPC models were self-reported, whereas data for the PD model was obtained via a
semi-structured interview. This may have added to distinctions between these models that were
due to measurement as opposed to differential aspects of personality.
Another category of limitations can be organized around the study sample. The primary
aim of this study was to better understand personality-PTSD relations; however individuals in
this sample all had co-morbid diagnoses. Consequently, results from the present study may not
generalize to individuals who only have more isolated difficulties related to PTSD. Despites this
loss in generalizability, however, these results may translate adequately to clinical practice given
the high rates of co-morbidity consistently observed in clinical settings (Shea, Widiger, & Klein,
1992; Westen, Novotny, & Thompson-Brenner, 2004). Secondly, study results, particularly with
regard to empirical validation procedures for clustering methods, may have benefited from a
larger sample size. Particularly, results from Asendorpf et al.’s (2001) procedure may have been
more consistent with Clustan results with a larger sample size.
Finally, and perhaps most importantly, it was beyond the scope of this thesis to evaluate
trauma with the same level of scrutiny as I evaluated personality. For example, some
individuals in the sample may have received a PTSD diagnosis based on an isolated trauma,
while others may have developed PTSD as the result of more chronic, severe or complex

51

traumas. Given that direct models evidence both that personality influences the development of
PTSD (e.g., Miller, 2003) and that complex trauma alters the personality system (e.g., Herman,
1992), critical examinations of personality and PTSD relations could benefit from a more
nuanced assessment of the severity and types of trauma(s) an individual has experienced.
Despite these limitations, results from the present study are important and novel for
several reasons. Of primary importance, this study demonstrated that distinct personality models
based on personality disorders, five-factor traits, and interpersonal traits show varying relations
to PTSD and other forms of psychopathology. This general finding illustrates that clinicians
would benefit from considering multiple levels of personality when assessing individuals with
PTSD. Miller’s (2003) typology of three personality types of individuals with PTSD (low
severity, high-severity internalizing, and high-severity externalizing) was recovered when
clusters were based on the PD model; however the FFM and IPC models each provided unique
representations of how personality can relate to PTSD.
Knowing how a person with PTSD fits into these models of personality can be beneficial
in a variety ways. For instance, military organizations could group soldiers according to FFM
personality traits to better prepare those with vulnerable personality profiles for potentially
traumatic service assignments. Conversely, a clinicians’ understanding of how to best treat an
individual with PTSD may be improved by knowing his interpersonal personality style. Indeed,
psychotherapy research has demonstrated that individuals with different interpersonal styles
respond to treatment in different ways (Anchin & Pincus, 2010; Blatt et al., 2001; Kiesler, 1996;
Pincus, Lukowitsky, & Wright, 2010), indicating the potential utility of the IPC model, when
used in conjunction with a patient’s presenting problems, to guide therapy. For example,
treatment approaches for an individual with PTSD who otherwise has low rates of personality

52

pathology, low global distress, and a warm-dominant interpersonal style may be different from
approaches for treating an individual with PTSD, low rates of personality pathology and global
distress, but a cold-submissive interpersonal style, despite shared overlap in three of four
important domains of personality and psychopathology. Further, the development of PTSD may
differ depending on whether an individual has a general personality disposition towards
internalizing pathology and vulnerable distress or whether she has minimal personality pathology
and low global distress. One possibility that merits future empirical attention is that individuals
with more resilient personality profiles (e.g., low pathology, low distress, and/or warmdominant) require more severe or repeated traumatic exposure in order to develop PTSD.
Though there were more systematic relations between membership in some clusters than
others across models (e.g., individuals who were warm-dominant were more likely to also have
low personality disorder pathology than individuals who were cold-dominant) the fact that
individuals differed in a variety of ways regarding their cluster placement across these models
further indicates that these distinct conceptualizations of personality are indeed capturing unique
aspects of individual motivation and behavior. Future research should focus on the aspects of
each of these models that are overlapping and the aspects of these models that are separable. For
example, recent research indicates that personality traits and personality disorders generally
provide non-overlapping and incrementally useful information (Morey et al., 2007; Morey et al.,
in review). However, the majority of this research has focused on differential information
provided by PD symptoms and FFM traits, with less work comparing how each of these models
relates to the IPC. Accordingly, more research aimed at understanding the relations between
various direct, pathoplastic and hybrid models of personality could meaningfully improve our
understanding of the ways that personality can relate to psychopathology.

53

In sum, findings from this study indicate the importance of separating and considering the
various ways that personality can be conceptualized and measured and, as a result, the various
ways that personality relates to psychopathology. This research shows that it is important for
researchers and clinicians to know what type of personality model is being utilized when
conducting research on personality-psychopathology relations and when using personality as a
framework to guide treatment approaches. Though multiple models of personality need not
always be considered, it is important for psychologists to understand that the construct
personality does not systematically mean the same thing across models merely because each
model includes the same descriptive label. Rather, personality is inherently intertwined with the
measures being utilized and evaluated, and these measures are influenced heavily by the
theoretical models that underlie them. The present study demonstrates that personality disorders,
five-factor personality traits, and interpersonal personality traits each provides non-redundant
and clinically important information regarding heterogeneity among individuals with PTSD.

54

Table 1. Means and Standard Deviations for the number of PD symptoms for individuals with
PTSD
AVPD

BPD

OCPD

STPD

M

3.87

5.10

2.73

2.39

SD

2.23

2.66

1.97

2.01

Note: AVPD = Avoidant Personality Disorder; BPD = Borderline Personality Disorder; OCPD
= Obsessive Compulsive Personality Disorder; STPD = Schizotypal Personality Disorder. The
maximum number of symptoms differed for each disorder: AVPD (7), BPD (9), OCPD (8),
STPD (8).

55

Table 2. Average Kappa Agreement between sub-samples A and B for Cluster Solutions in Each
Model

2 cluster solution
3 cluster solution
4 cluster solution
5 cluster solution

PD
0.862
0.562
0.759
0.469

FFM
0.736
0.228
0.311
0.472

IPC
0.095
0.465
0.597
0.482

Note: PD = Personality Disorder Model; FFM = Five-Factor Model; IPC = Interpersonal
Circumplex Model.

56

Table 3. Optimal Number of Clusters for Each Personality Model based on Four Empirical
Procedures
Clustan
PD
FFM
IPC

Asendorpf

Fusion Plot

Mojena

3
3
4

2
2
4

5
5
3

3
4
3

Number
Selected
3
3
4

Note: PD = Personality Disorder Model; FFM = Five-Factor Model; IPC = Interpersonal
Circumplex Model.

57

Table 4. PD Clusters’ Mean within Sample T-scores and Standard Deviations for PD, FFM, and
IPC variables.

DSM-IV PDs
(clustered variables)
Paranoid
Schizotypal
Schizoid
Histrionic
Borderline
Narcissistic
Antisocial
Avoidant
Dependent
ObsessiveCompulsive
Five Factors
Neuroticism
Extraversion
Openness
Agreeableness
Conscientiousness
IPC octants
PA
BC
DE
FG
HI
JK
LM
NO
IPC vectors
Dominance
Warmth
Functioning variable
GAF
* p < .05

Cluster 1
Low
Pathology
N = 68

Cluster 2
Internalizing

Cluster 3
Externalizing

F

Post-hoc
Duncan

N = 65

N = 22

45.57 (8.70)
46.99 (10.37)
47.78 (10.40)
45.67 (5.77)
45.05 (8.77)
46.26 (4.98)
44.87 (3.94)
43.42 (8.19)
44.28 (5.21)
46.86 (10.19)

53.34 (9.20)
51.65 (8.49)
51.65 (8.49)
52.15 (9.47)
54.85 (7.72)
52.00 (11.01)
48.40 (5.54)
56.79 (6.93)
55.79 (9.84)
52.73 (10.00)

53.82 (10.99)
54.43 (10.58)
52.53 (9.90)
57.02 (15.07)
54.04 (10.99)
55.65 (13.92)
70.57 (7.78)
50.26 (9.47)
50.57 (11.90)
51.64 (7.08)

13.89*
6.57*
3.15*
15.87*
29.42*
10.79*
207.10*
47.85*
30.53*
6.51*

1<2&3
1<2&3
1<3
1<2<3
1<2&3
1<2&3
1<2<3
1<3<2
1<3<2
1<2&3

45.61 (8.42)
50.46 (9.14)
49.78 (9.44)
52.49 (9.80)
53.42 (9.98)

54.74 (8.24)
48.59 (10.49)
50.48 (10.79)
50.33 (9.22)
47.18 (9.81)

49.54 (13.00)
52.75 (10.79)
49.24 (8.83)
41.34 (8.30)
47.77 (7.40)

16.70*
1.56
0.16
11.83*
7.71*

1&3<2

50.48 (8.81)
48.89 (8.16)
47.86 (10.71)
49.15 (10.20)
48.66 (9.32)
51.00 (9.58)
50.86 (9.43)
50.56 (9.82)

48.28 (10.45)
49.13 (10.48)
50.82 (9.25)
50.98 (9.03)
52.97 (9.44)
50.50 (9.81)
50.79 (10.02)
48.84 (9.26)

53.61 (11.36)
55.99 (11.93)
54.20 (8.42)
49.75 (12.14)
45.36 (11.36)
45.43 (11.03)
45.05 (10.65)
51.71 (12.50)

2.52
4.83*
3.85*
0.56
6.25*
2.78
3.24*
0.87

2<3
1&2<3
1<3

50.44 (8.06)
51.87 (10.53)

47.75 (9.81)
49.72 (9.05)

55.29 (11.24)
45.04 (9.63)

5.03*
4.08*

1&2<3
3<1&2

56.10 (8.63)

50.66 (10.36)

50.36 (10.46)

6.24*

1<2&3

58

3<1&2
2&3<1

1&3<2
3<1&2
3<1&2

Table 5. FFM Clusters’ Mean within Sample T-scores and Standard Deviations PD, FFM, and
IPC variables.

Five Factors
(clustered variables)
Neuroticism
Extraversion
Openness
Agreeableness
Conscientiousness
DSM-IV PDs
Paranoid
Schizotypal
Schizoid
Histrionic
Borderline
Narcissistic
Antisocial
Avoidant
Dependent
ObsessiveCompulsive
IPC octants
PA
BC
DE
FG
HI
JK
LM
NO
IPC vectors
Dominance
Warmth
Functioning variable
GAF
* p < .05

Cluster 1
Low
Distress
N = 75

Cluster 2
Resilient
Distress
N = 41

Cluster 3
Vulnerable
Distress
N = 39

F

Post-hoc
Duncan

43.74 (7.75)
53.60 (9.40)
49.65 (9.15)
50.38 (10.15)
57.36 (7.09)

54.99 (7.44)
52.01 (7.99)
57.83 (7.10)
53.20 (7.90)
44.37 (5.56)

56.79 (8.89)
40.98 (7.25)
42.45 (7.47)
45.90 (10.50)
41.25 (6.86)

45.26*
29.55*
34.92*
5.78*
97.55*

1<2&3
3<1&2
3<1<2
3<1&2
3<2<1

47.88 (8.48)
48.99 (9.83)
49.08 (9.37)
49.07 (9.18)
48.80 (10.21)
50.05 (11.52)
48.01 (8.59)
46.54 (9.19)
47.19 (8.63)
50.66 (9.92)

49.50 (10.27)
49.62 (10.55)
48.62 (10.76)
50.17 (8.83)
50.58 (10.26)
48.23 (7.00)
49.08 (8.97)
52.63 (9.29)
52.34 (9.58)
47.82 (10.09)

54.61 (11.10)
52.36 (9.58)
53.23 (9.89)
51.69 (9.22)
51.69 (9.22)
51.75 (9.41)
54.80 (12.01)
53.88 (10.17)
52.95 (11.54)
51.01 (9.97)

6.28*
1.51
2.81
0.82
1.17
1.24
6.62*
9.85*
6.18*
1.35

1&2<3

53.21 (9.73)
51.10 (10.17)
49.04 (10.50)
47.05 (10.24)
46.32 (9.80)
50.30 (9.47)
50.83 (9.74)
52.43 (9.57)

50.86 (8.79)
49.03 (9.45)
47.23 (8.67)
49.20 (8.09)
51.53 (8.98)
50.24 (9.85)
53.08 (8.06)
52.77 (8.23)

42.91 (8.17)
48.90 (10.25)
54.77 (8.84)
56.52 (8.39)
55.46 (8.56)
49.18 (11.30)
45.16 (10.80)
42.42 (8.76)

16.61*
0.88
6.83*
13.59*
13.17*
0.17
7.32*
18.36*

3<1&2

53.57 (10.22)
51.67 (9.28)

50.36 (7.59)
53.14 (9.12)

42.75 (7.86)
43.48 (9.50)

18.50*
13.17*

3<1&2
3<1&2

54.37 (8.49)

51.05 (9.49)

52.44 (12.65)

1.57

59

1&2<3

1&2<3
1<2&3
1<2&3

1&2<3
1&2<3
1<2<3
3<1&2
3<1&2

Table 6. IPC Clusters’ Mean within Sample T-scores and Standard Deviations PD, FFM, and IPC variables.
Cluster 1
WarmDominant
N = 38

Cluster 2
WarmSubmissive
N = 44

Cluster 3
ColdSubmissive
N = 40

IPC Octants
(clustered variables)
PA 54.86 (7.94) 44.42 (5.06) 43.41 (7.84)
BC 46.37 (7.29) 45.79 (6.72) 47.62 (8.00)
DE 40.48 (5.93) 44.60 (6.24) 57.61 (6.20)
FG 41.27(7.54)
49.89 (6.27) 60.10 (7.15)
HI 43.76 (8.27) 55.06 (5.91) 56.39 (7.80)
JK 50.60 (9.32) 56.54 (7.51) 48.58 (8.79)
LM 54.90 (9.92) 52.85 (5.56) 44.36 (11.61)
NO 56.85 (8.47) 48.68 (5.95) 40.26 (6.84)
IPC vectors
Dominance 55.97 (4.93) 44.18 (5.10) 41.19 (6.06)
Warmth 59.81 (6.80) 54.91 (4.65) 41.02 (6.57)
DSM-IV PDs
Paranoid 44.96 (6.87) 48.47 (9.19) 55.67 (10.75)
Schizotypal 46.77 (10.15) 47.15 (8.50) 54.34 (10.66)
Schizoid 44.80 (6.86) 47.40 (9.31) 55.60 (9.52)
Histrionic 48.28 (7.61) 49.96 (8.47) 45.46 (5.49)
Borderline 49.95 (9.22) 50.01 (10.88) 47.31 (9.13)
Narcissistic 48.77 (8.96) 48.05 (6.58) 47.29 (5.28)
Antisocial 46.46 (6.48) 48.08 (7.90) 50.39 (9.82)
Avoidant 43.24 (8.79) 51.68 (9.41) 54.36 (9.67)
Dependent 48.61 (9.14) 52.71 (12.04) 48.46 (8.61)
Obsessive- 49.00 (11.35) 46.91 (9.92) 52.42 (9.96)
Compulsive

60

Cluster 4
ColdDominant
N = 33

F

Post-hoc
Duncan

59.84 (8.60)
62.68 (8.26)
58.93 (6.25)
47.99 (8.84)
42.68 (9.66)
42.30 (9.39)
47.39 (8.68)
55.69 (9.15)

44.16*
39.92*
85.06*
43.21*
32.17*
17.25*
11.01*
38.98*

3&2<1<4
1, 2 & 3 < 4
1<2<3&4
1<2&4<3
1&4<2&3
4<1&3<2
3&4<1&2
3<2<1&4

61.58 (7.01)
43.05 (6.95)

104.03*
81.77*

3<2<1<4
3&4<2&1

50.97 (9.89)
52.27 (8.60)
52.66 (10.58)
57.52 (13.90)
53.29 (10.10)
57.31 (15.00)
56.16 (13.12)
50.26 (8.77)
49.86 (9.19)
52.32 (7.55)

9.22*
6.16*
11.24*
11.23*
2.21
8.76*
7.08*
10.34*
1.66
3.02*

1&2<3&4
1&2<3&4
1&2<3&4
3 < 2 & 4; 1 < 4
3<4
1, 2 & 3 < 4
1, 2 & 3 < 4
1 < 2, 3 & 4
2<3&4

Table 6 (cont’d)
Five Factors
Neuroticism
Extraversion
Openness
Agreeableness
Conscientiousness
Functioning
variable
GAF
* p < .05

46.75 (7.42)
58.19 (6.70)
52.36 (9.65)
55.73 (7.52)
53.62 (9.22)

49.81 (7.42)
47.51 (5.62)
49.61 (9.73)
56.66 (6.77)
48.34 (8.61)

54.17 (9.89) 48.92 (10.67)
39.08 (5.27) 57.13 (7.96)
45.71 (9.45) 52.99 (9.41)
45.99 (9.45) 32.39 (6.50)
46.61 (11.88) 52.15 (8.54)

54.11 (9.59)

51.07 (10.77) 53.45 (10.53)

61

53.79 (8.62)

4.03*
76.07*
4.54*
48.49*
4.37*

0.80

1&4<3
3<2<1&4
3<1&4
4<3<1&2
3 < 1 & 4; 2 < 1

Table 7. Demographic Characteristics for each cluster
Mean age

% Female

% Caucasian

PD clusters
Low pathology
Internalizing
Externalizing
Test statistics

34.66 (7.52)
32.0 (7.76)
32.77 (7.39)
F(2) = 2.09

77.94%
76.92%
59.09%
(2)
η = .15

66.18%
76.92%
77.27%
(2)
η = .12

FFM clusters
Low distress
Resilient distress
Vulnerable distress
Test statistics

33.96 (7.74)
31.83 (7.93)
33.49 (7.19)
F(2) = 1.05

78.67%
70.73%
71.79%
(2)
η = .09

64.0%
80.49%
79.49%
(2)
η = .18

IPC clusters
Warm-dominant
Warm-submissive
Cold-dominant
Cold-submissive
Test statistics

32.05 (7.85)
34.41 (8.11)
34.50 (6.71)
31.70 (7.74)
F(3) = 1.47

78.95%
81.81%
65.0%
72.73%
(2)
η = .15

63.16%
86.36%
67.50%
69.70%
(2)
η = .21

Note: There were no significant demographic differences (p < .05) between clusters in any of
the three models. Age was assessed using a one-way ANOVA; gender and race were assessed
using a chi-square test.

62

Table 8. Crosstabs between PD and Other Model Clusters
PD cluster membership
Low pathology
Internalizing
Externalizing
FFM cluster membership:
Low distress
Resilient distress
Vulnerable distress
Total
IPC cluster membership:
Warm-dominant
Warm-submissive
Cold-submissive
Cold-dominant
Total

44 (64.7%)
14 (20.6%)
10 (14.7%)
68

23 (35.4%)
23 (35.4%)
19 (29.2%)
65

8 (36.4%)
4 (18.2%)
10 (45.5%)
22

24 (35.3%)
22 (32.4%)
15 (22.1%)
7 (10.3%)
68

12 (18.5%)
18 (27.7%)
20 (30.8%)
15 (23.1%)
65

2 (9.1%)
4 (18.2%)
5 (22.7%)
11 (50.0%)
22

Note: percentages were computed vertically (e.g., 64.7% of individuals in the low pathology PD
cluster were in the low distress FFM cluster).

63

Table 9. Crosstabs between FFM and Other Model Clusters

Low distress
PD cluster membership:
Low pathology
Internalizing
Externalizing
Total
IPC cluster membership:
Warm-dominant
Warm-submissive
Cold-submissive
Cold-dominant
Total

FFM cluster membership
Resilient distress Vulnerable distress

44 (58.7%)
23 (30.7%)
8 (10.7%)
75

14 (34.1%)
23 (56.1%)
4 (9.8%)
41

10 (25.6%)
19 (48.7%)
10 (25.6%)
39

26 (34.7%)
18 (24.0%)
10 (13.3%)
21 (28.0%)
75

12 (29.3%)
16 (39.0%)
6 (14.6%)
7 (17.1%)
41

0 (0%)
10 (25.6%)
24 (61.5%)
5 (12.8%)
39

Note: percentages were computed vertically (e.g., 58.7% of individuals in the low distress FFM
cluster were in the low pathology PD cluster).

64

Table 10. Crosstabs between IPC and Other Model Clusters

Warmdominant

IPC cluster membership:
WarmColdsubmissive
submissive

Colddominant

PD cluster membership:
Low pathology
Internalizing
Externalizing
Total

24 (63.2%)
12 (31.6%)
2 (5.3%)
38

22 (50.0%)
18 (40.9%)
4 (9.1%)
44

15 (37.5%)
20 (50.0%)
5 (12.5%)
40

7 (21.2%)
15 (45.5%)
11 (33.3%)
33

FFM cluster membership:
Low distress
Resilient distress
Vulnerable distress
Total

26 (68.4%)
12 (31.6%)
0 (0%)
38

18 (40.9%)
16 (36.4%)
10 (22.7%)
44

10 (25.0%)
6 (15.0%)
24 (60.0%)
40

21 (63.6%)
7 (21.2%)
5 (15.2%)
33

Note: percentages were computed vertically (e.g., 63.2% of individuals in the warm-dominant
IPC cluster were in the low pathology PD cluster).

65

Table 11. Percent of Individuals in Each Cluster with Axis I Disorder Diagnoses
Generalized
Anxiety Disorder

Alcohol Use
Disorder

Substance Use
Disorder

Low pathology
Internalizing
Externalizing
Eta

19.1%
26.2%
22.7%
.078

36.8%
56.9%
81.8%
.309*

51.5%
55.4%
90.9%
.269*

Low distress
Resilient distress
Vulnerable distress
Eta

14.7%
17.1%
46.3%
.292*

38.7%
58.5%
69.2%
.262*

45.3%
61.0%
82.1%
.305*

Warm-dominant
Warm-submissive
Cold-submissive
Cold-dominant
Eta

7.9%
25.0%
35.0%
21.2%
.233*

31.6%
54.5%
60.0%
60.6%
.234*

50.0%
54.5%
67.5%
63.6%
.142

PD:

FFM:

IPC:

* χ² < .05

66

Table 12. Crosstabs between Baseline and Follow-up PD Clusters
Follow-up cluster membership
Baseline cluster
membership:
Low pathology
Internalizing
Externalizing
Total

Percent of individuals in
each cluster:

Low pathology

Internalizing

Externalizing

Baseline

Follow-up

50 (86.2%)
37
6
93

7
14 (25.9%)
4
25

1
3
10 (50.0%)
14

43.9%
41.9%
14.2%

70.5%
8.9%
3.6%

*percentages in diagonals are the percent of individuals who placed in the same cluster at
baseline and follow-up

67

Table 13. Crosstabs between Baseline and Follow-up FFM Clusters
Follow-up cluster membership
Baseline cluster
membership
Low distress
Resilient distress
Vulnerable distress
Total

Low
distress
49 (81.7%)
24
25
98

Resilient
distress
0
0 (0%)
0
0

Vulnerable
distress
11
5
8 (24.2%)
24

Percent of individuals
in each cluster:
Baseline Follow-up
48.4%
26.5%
25.2%

80.3%
0%
19.7%

*percentages in diagonals are the percent of individuals who placed in the same cluster at
baseline and follow-up

68

Table 14. Crosstabs between Baseline and Follow-up IPC Clusters
Follow-up cluster membership
Baseline cluster
Warmmembership:
dominant
Warm-dominant 26 (100%)
Warm-submissive
24
Cold-submissive
21
Cold-dominant
26
Total
97

Warmsubmissive
0
10 (27.8%)
11
0
21

Coldsubmissive
0
1
1 (3.0%)
1
3

Colddominant
0
1
0
0 (0%)
1

Percent of individuals
in each cluster:
Baseline Follow-up
24.5%
28.4%
25.8%
21.3%

79.5%
17.2%
2.5%
0.8%

*percentages in diagonals are the percent of individuals who placed in the same cluster at
baseline and follow-up

69

Figure 1. The Interpersonal Circumplex (IPC)
Dominant

Warm-Dominant
Agency

Cold-Dominant

Communion

Cold

Cold-Submissive

Warm

Warm-Submissive

Submissive

70

Figure 2. Expected Comorbidity Patterns with IPC-based Clusters
Dominant
OC PD
Cold-Dominant
Warm-Dominant

PTSD & BPD

Cold

Warm

SPD

Cold-Submissive

Warm-Submissive
AV PD
Submissive

*Note: OCPD = Obsessive-Compulsive Personality Disorder; PTSD = Posttraumatic Stress
Disorder; BPD = Borderline Personality Disorder; SPD = Schizoid Personality Disorder;
AVPD= Avoidant Personality Disorder.

71

Figure 3. IPC octants

(PA)
Assertive
(BC)
Manipulative

(NO)
Extraverted

(LM)
Altruistic

(DE)
Selfish

(FG)
Introverted

(JK)
Straightforward
(HI)
Unassertive

72

Figure 4. Fusion Coefficient (Y-Axis) Plot for the Number of Clusters (X-Axis) in the PD Model
8000
6000
4000
2000
0
1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20

73

Figure 5. Fusion Coefficient (Y-Axis) Plot for the Number of Clusters (X-Axis) in the FFM Model
400000
300000
200000
100000
0
1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20

74

Figure 6. Fusion Coefficient (Y-Axis) Plot for the Number of Clusters (X-Axis) in the IPC Model
25000
20000
15000
10000
5000
0
1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20

75

Figure 7. Dominance and Warmth Z-score Projections for the Four IPC Clusters
Dominant
1.5

1

0.5

0

Cold
-1.5

-1

-0.5

Warm
0

-0.5

-1

-1.5
Submissive

76

0.5

1

1.5

Figure 8. Dominance and Warmth Z-scores for each PD and FFM cluster Projected on the IPC
Dominant
1.5

1

PD:
Externalizin
g

0.5

FFM: Low
PD: Low

FFM:
Resilient

0

Cold
-1.5

-1

-0.5

0

PD:
Internalizing
-0.5

FFM:
Vulnerable
-1

-1.5

Submissive

77

0.5

1

1.5

Warm

REFERENCES

78

Afifi, T.O., Asmundson, G.J.G., Taylor, S., & Lang, K.L. (2010). The role of genes and
environment on trauma exposure and posttraumatic stress disorder symptoms: A review
of twin studies. Clinical Psychology Review, 30, 101-112.
Alden, L.E., Wiggins, J.S., & Pincus, A.L. (1990). Construction of circumplex scales for the
Inventory of Interpersonal Problems. Journal of Personality Assessment, 55, 521-536.
Aldenderfer, M.S., & Blashfield, R.K. (1984). Cluster Analysis. Quantitative Applications in the
Social Sciences, No. 44. Beverly Hills, CA: Sage.
Ambwani, S., & Hopwood, C.J. (2009). The utility of considering interpersonal problems in the
assessment of bulimic features. Eating Behaviors, 10, 247-253.
American Psychiatric Association (1994). Diagnostic and statistical manual of mental
disorders—IV. Washington, DC: Author.
Anchin, J.C., & Pincus, A.L. (2010). Evidence-based interpersonal psychotherapy with
personality disorders: theory, components, and strategies. In, Magnavita, J.J. (Ed.),
Evidence-based treatment of personality dysfunction: Principles, methods, and processes.
Washington, DC: American Psychological Association.
Ansell, E.B., et al. (in review). The association of personality disorders with the prospective
seven year course of anxiety disorders.
Asendorpf, J.B. (2003). Head-to-head comparison of the predictive validity of personality types
and dimensions. European Journal of Personality, 17, 327-346.
Asendorpf, J.B., Borkenau, P., Ostendorf, F., & Van Aken, M.A.G. (2001). Carving personality
description at its joints: Confirmation of three replicable personality prototypes for both
children and adults. European Journal of Personality, 15, 169-198.
Asendorpf, J.B., & Denissen, J.J.A. (2006). Predictive validity of personality types versus
personality dimensions from early childhood to adulthood: Implications for the
distinction between core and surface traits. Merrill-Palmer Quarterly, 52, 486-513.
Avdeyeva, T.V., & Church, A.T. (2005). The cross-cultural generalizability of personality types:
A Philippine Study. European Journal of Personality, 19, 475-499.
Bakan, D. (1966). The duality of human existence: Isolation and Communion in Western Man.
Boston: Beacon Press.
Barrett, M.S., & Barber, J.P. (2007). Interpersonal profiles in major depressive disorder. Journal
of Clinical Psychology, 63 (3), 247-266.

79

Beck, A. T. (1983). Cognitive therapy of depression: New perspectives. In P. J. Clayton & J. E.
Barrett (Eds.), Treatment of depression: Old controversies and new approaches (pp. 265290). New York, NY: Raven.
Benjamin, L. S. (2003). Interpersonal reconstructive therapy: Promoting change in
nonresponders. New York, NY: Guilford Press.
Benjamin, L. S. (2005). Addressing interpersonal and intrapsychic components of personality
during psychotherapy. In S. Strack (Ed.), Handbook of personology and psychopathology
(pp. 417-441). Hoboken: John Wiley & Sons Inc.
Blatt, S.J., Quinlan, D.M., Chevron, E.S., McDonald, C., & Zuroff, D. (1982). Dependency and
self-criticism: Psychological dimensions of depression. Journal of Counseling and
Clinical Psychology, 50 (1), 113-124.
Blatt, S.J., Shahar, G., & Zuroff, D.C. (2001). Anaclitic (sociotropic) and introjective
(autonomous) dimensions. Psychotherapy, 38 (4), 449-454.
Boehm, B., Asendorpf, J.B., & Avia, M.D. (2002). Replicable types of personality: Spanish
NEO-PI samples. European Journal of Personality, 16, 25-41.
Boggs, C.D., Morey, L.C., Skodol, A.E., Shea, M.T., Sanislow, C.A., Grilo, C.M., McGlashan,
T.H., Zanarini, M.C., & Gunderson, J.G. (2005). Differential Impairment as an Indicator
of Sex Bias in DSM–IV Criteria for Four Personality Disorders. Psychological
Assessment, 17(4), 492-496.
Bradley, R., Heim, A., & Westen, D. (2005). Personality constellations in patients with a history
of childhood sexual abuse. Journal of Traumatic Stress, 18 (6), 769-780.
Breslau, N., Davis, G.C., Andreski, P., & Peterson, E. (1991). Traumatic events and
posttraumatic stress disorder in an urban population of young adults. Archives of General
Psychiatry, 48, 216-222.
Breslau, N., Kessler, R.C., Chilcoat, H.D., Schultz, L.R., David, G.C., & Andreski, P. (1998).
Trauma and posttraumatic stress disorder in the community: The 1996 Detroit area
survey of trauma. Archives of General Psychiatry, 55, 626-632.
Cattell, R.B. (1966). Handbook of multivariate experimental psychology. Chicago: Rand
McNally.
Cattell, R. B., & Vogelmann, S. (1977). A comprehensive trial of the scree and KG criteria for
determining the number of factors. Journal of Multivariate Behavioral Research, 12,
289-325.
Clark, L.A. (1993). Schedule for Nonadaptive and Adaptive Personality: Manuel for
Administration, Scoring and Interpretation. Minneapolis: University of Minnesota Press.

80

Clark, L.A., Watson, D., & Mineka, S. (1994). Temperament, personality, and the mood and
anxiety disorders. Journal of Abnormal Psychology, 103(1), 103-116.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ:
Erlbaum.
Costa, P.T., Herbst, J.H., McCrae, R.R., Samuels, J., & Ozer, D.J. (2002). The replicability and
utility of three personality types. European Journal of Personality, 16, 73-87.
Costa, P.T., & McCrea, R.R. (1992). NEO Five-Factor Inventory. Lutz, FL: Psychological
Assessment Resources.
De Fruyt, F., Mervielde, I., & van Leeuwen, K. (2002). The consistency of personality type
classification across samples and five-factor measures. European Journal of Personality,
16, 57-72.
Donnellan, M.B., & Robins, R.W. (in review). Personality types: Issues and controversies.
Ethier, N., Sadler, P., & Woody, E. (2009). The good, the bad, and the unwily: constructing a
balanced measure of interpersonal adjectives. Poster presented at the annual meeting of
the Society for Interpersonal Theory and Research, Toronto, Canada.
Eysenck, H.J. (1947). Dimensions of Personality. New York: Praeger.
Fanous, A.H., Gardner, C.O., Prescott, C.A., Cancro, R., & Kendler, K.S. (2002). Neuroticism,
major depression and gender: A population based twin study. Psychological Medicine,
32, 719-728.
Fergusson, D.M., Woodward, L.J., & Horwood, L.J. (2000). Risk factors and life processes
associated with the onset of suicidal behavior during adolescence and early adulthood.
Psychology Medicine, 30, 23-39.
First, M.B., Spitzer, R.L., Gibbon, M., Williams, J.B.W. (1996). Structured clinical interview for
DSM-IV axis I disorders, patient editions (SCID-P). New York State Psychiatric
Institute, Biometrics Research: New York.
Flood, A.M., Boyle, S.H., Calhoun, P.S., Dennis, M.F., Barefoot, J.C., Moore, S.D., & Beckham,
J.C. (2010). Prospective study of externalizing and internalizing subtypes of
posttraumatic stress disorder and their relationship to mortality among Vietnam veterans.
Comprehensive Psychiatry, 51, 236-242.
Freud, S. (1930). Civilization and its Discontents. New York: Norton & Company:
Goldberg, L.R. (1993). The structure of phenotypic personality traits. American Psychologist, 48,
26-34.

81

Grilo, C.M., Shea, M.T., Sanislow, C.A., Skodal, A.E., Gunderson, J.G., Stout, R.L., Pagano,
M.E., Yen, S, Morey, L.C., Zanarini, M.C., & McGlashan, T.H., (2004). Two-year
stability and change in schizotypcal, borderline, avoidant, and obsessive-compulsive
personality disorders. Journal of Consulting and Clinical Psychology, 72, 767-775.
Gunderson, J.G., Shea, M.T., Skodol, A.E., McGlashan, T.H., Morey, L.C., Stout, R.L.,
Zanarini, M.C., Grilo, C.M., Oldham, J.M., & Keller, M.B. (2000). The collaborative
longitudinal personality disorders study: Development, aims, design and sample
characteristics. Journal of Personality Disorders, 14, 300-315.
Harkness, A.R., McNulty, J.L., & Ben-Porath, Y.S. (1995). The personality psychopathology
five (PSY-5): Constructs and MMPI-2 scales. Psychological Assessment, 7, 104-114.
Hart, D., Atkins, R., & Fegley, S. (2003). Personality and development in childhood: A personcentered approach. Monographs of the Society of Research in Child Development, 68, vii109.
Hatcher, R.L., & Rogers, D.T. (2009). Development and validation of a measure of interpersonal
strengths: The inventory of interpersonal strengths. Psychological Assessment, 21(4),
554-569.
Herman, J.L. (1992). Trauma and Recovery. New York, NY: Basic Books.
Herman, J.L., & van der Kolk, B.A. (1987). Traumatic antecedents of borderline personality
disorder. In B.A. van der Kolk (Ed.), Psychological Trauma (pp. 111-126). Washington
DC: American Psychiatric Press.
Hettema, J.M., Neale, M.C., Myers, J.M., Prescott, C.A., & Kendler, K.S. (2006). A populationbased twin study of the relationship between neuroticism and internalizing disorders.
American Journal of Psychiatry, 163, 857-864.
Hopwood, C.J., Ansell, E.B., Fehon, D.C., & Grilo, C.M. (in press). Personality heterogeneity in
female adolescent inpatients with features of eating disorders. Comprehensive Psychiatry.
Hopwood, C.J., Clarke, A.N., & Perez, M. (2007). Pathoplasticity of bulimic features and
interpersonal problems. International Journal of Eating Disorders, 40 (7), 652-658.
Hopwood, C.J., Koonce, E.A., & Morey, L.C. (in press). An exploratory study of integrative
personality pathology systems and the interpersonal circumplex. Journal of
Psychopathology and Behavioral Assessment.
Hopwood, C.J. & Morey, L.C. (2007). Psychological conflict in borderline personality as
represented by inconsistent self-report item responding. Journal of Social and Clinical
Psychology, 26(9), 1097-1107.

82

Hopwood, C.J., Thomas, K.M., & Zanarani, M.C. (in review). Clinical implications of an
integrative model of borderline personality disorder
Horowitz, L.M. (2004). Interpersonal foundations of psychopathology. Washington, D.C.:
American Psychological Association.
Horowitz, L.M., Rosenberg, S.E., Baer, B.A., Ureno, G., & Villasenor, V.S. (1988). Inventory of
interpersonal problems: Psychometric properties and clinical applications. Journal of
Consulting and Clinical Psychology, 56(6), 885-892.
Hyer, L., Braswell, L., Albrecht, B., Boyd, S., Boudewyns, P., & Talbert, S. (1994). Relationship
of NEO-PI to personality styles and severity of trauma in chronic PTSD victims. Journal
of Clinical Psychology, 50, 699-707.
Kachin, K.E., Newman, M.G., & Pincus, A.L. (2001). An interpersonal problem approach to the
division of social phobia subtypes. Behavior Therapy, 32, 479-501.
Kendler, K.S., Gatz, M., Gardner, C.O., & Pederse, N.L. (2006). Personality and major
depression. Archives of General Psychiatry, 63, 1113-1120.
Kendler, K.S., Neale, M.C., Kessler, R.C., Heath, A.C., & Eaves, L.J. (1993). A longitudinal
twin study of personality and major depression in women. Archives of General
Psychiatry, 50, 853-862.
Kessler, R.C., Sonnega, A., Bromet, E., Hughes, M., & Nelson, C.B. (1995). Posttraumatic stress
disorder in the national comorbidity survey. Archives of General Psychiatry, 52(12),
1048-1060.
Kiesler, D. J. (1996). Contemporary interpersonal theory and research: Personality,
psychopathology, and psychotherapy. New York: John Wiley & Sons.
Klein, M.H., Wonderlich, S., & Shea, M.T. (1993). Models of relationships between personality
and depression: Toward a framework for theory and research. In Klein, Kupfer, & Shea
(Eds.). Personality and Depression (pp. 1-54). New York: The Guilford Press.
Krueger, R.F. (1999). The structure of common mental disorders. Archives of General
Psychiatry, 56, 921-926.
Laforge, R. & Suczek, R.F. (1955). The interpersonal dimension of personality: III. An
interpersonal checklist. Journal of Personality, 24, 94-112.
Lahey, B.B. (2009). The public health significance of neuroticism. American Psychologist,
64(4), 241-256.
Leary, T. (1957). Interpersonal diagnosis of personality. New York: Ronald Press.

83

Malouff, J.M., Thorsteinson, E.B., Rooke, S.E., & Schutte, N.S. (2005). The relationship
between the five-factor model of personality and symptoms of clinical disorders: A metaanalysis. Journal of Psychopathology and Behavioral Assessment, 27, 101-114.
McCrea, R.R., & Costa, P.T. (1990). Personality in Adulthood: A Five-Factor Theory
Perspective. New York, NY: Guilford Press.
McDevitt-Murphy, M.E., Shea, T.M., Yen, S., Grilo, C.M., Sanislow, C.A., &
Markowitz, J.C. (in review). Prospective investigation of a PTSD personality
typology.
McGlashan, T.H., Grilo, C.M., Skodal, A.E., Gunderson, J.G., Shea, M.T., Morey, L.C.,
Zanarini, M.C., & Stout, R.L. (2000). The collaborative longitudinal personality disorders
study: Baseline Axis I/II and II/II diagnostic co-occurrence. Acta Psychiatrica
Scandinavica, 102(4), 256-264.
Miller, M.W. (2003). Personality and the etiology and expression of PTSD: a three-factor
model perspective. Clinical Psychology: Science and Practice, 10, 373-393.
Miller, M.W., Greif, J.L., & Smith, A.A. (2003). Multidimensional personality
questionnaire profiles of veterans with traumatic combat exposure: externalizing
and internalizing subtypes. Psychological Assessment, 15, 205-215.
Miller, M.W., Kaloupek, D.G., Dillon, A.L., & Keane, T.M. (2004). Externalizing and
internalizing subtypes of combat-related PTSD: A replication and extension using
the PSY-5 scales. Journal of Abnormal Psychology, 113, 363-345.
Miller, M.W., & Resick, P.A. (2007). Internalizing and externalizing subtypes in female
sexual assault survivors: implications for the understanding of complex PTSD.
Behavior Therapy, 38, 58-71.
Milligan, G.W. (1981). A monte carlo study of thirty internal criterion measures for cluster
analysis. Psychometrika, 46(2), 187-199.
Mojena, R. (1977). Hierarchical grouping methods and stopping rules – an evaluation. Computer
Journal, 20, 359-363.
Morey, L.C. (1985). An empirical comparison of interpersonal and DSM-II approaches to
classification on personality disorders. Psychiatry, 48, 358-364.
Morey, L.C., Blashfield, R.K., & Skinner, H.A. (1983). A comparison of cluster analysis
techniques within a sequential validation framework. Multivariate Behavioral Research,
18, 309-329.

84

Morey, L.C., Gunderson, J.G., Quigley, B.D., Shea, T.M., Skodol, A.E., McGlashan, T.H. Stout,
R.L., & Zanarini, M.C. (2002). The representation of Borderline, Avoidant, ObsessiveCompulsive, and Schizotypal Personality Disorders by the Five-Factor Model. Journal of
Personality Disorders, 16 (3), 215-234.
Morey, L.C., Hopwood, C.J., Gunderson, J.G., Skodol, A.E., Shea, T.M., Yen, S., Stout, R.L.,
Zanarini, M.C., Grilo, C.M., Sanislow, C.A., & McGlashan, T.H. (2002). Comparison of
alternative models for personality disorders. Psychological Medicine, 37, 983-994.
Morey, L.C., Hopwood, C.J., Markowitz, J.C., Gunderson, J.G., Grilo, C.M., McGlashan, T.H.,
Shea, M.T., Yen, S., Sanislow, C.A., & Skodol, A.E. (in review). Long term predictive
validity of diagnostic models for personality disorder: Integrating trait and disorder
concepts.
Paris, J. (2003). Personality disorders over time: precursors, course and outcome. Journal of
Personality Disorders, 17(6), 479-488.
Pincus, A.L., Lukowitsky, M.R., & Wright, A.G.C. (2010). The interpersonal nexus of
personality and psychopathology. In T. Millon, R. Kreuger, & E. Simonsen
(Eds.), Contemporary directions in psychopathology: Toward DSM-V and ICD11. New York: Guilford.
Salzer, S., Pincus, A.L., Hoyer, J., Kreische, Leichsenring, F., & Leibing, E. (2008).
Interpersonal subtypes within Generalized Anxiety Disorder. Journal of Personality
Assessment, 90 (3), 292-299.
Samuel, D.B. & Widiger, T.A. (2008). A meta-analytic review of the relationships between the
five-factor model and DSM-IV-TR personality disorders: A facet level analysis. Clinical
Psychology Review, 28, 1326-1342.
Saulsman, L.M., & Page, A.C. (2004). The five-factor model and personality disorder empirical
literature: A meta-analytic review. Clinical Psychology Review, 23, 1055-1085.
Schnabel, K., Asendorpf, J.B., & Ostendorf, F. (2002). Replicable types and subtypes of
personality: German NEO-PI versus NEO-FFI. European Journal of Personality, 16, 724.
Schmitt, N., Oswald, F.L., Kim, B.H., Imus, A., Merritt, S., Friede, A., Shivpuri, S. (2007). The
use of background and ability profiles to predict college student outcomes. Journal of
Applied Psychology, 92 (1), 165-179.
Sellbom, M., & Bagby, R.M. (2009). Identifying PTSD personality subtypes in a workplace
trauma sample. Journal of Traumatic Stress, 22(5), 471-475.

85

Shea, M.T., Stout, R.L., Gunderson, J.G., Morey, L.C., Grilo, C.M., McGlashan, T.H., Skodol,
A.E., Dolan-Sewell, R.T., Dyck, I.R., Zanarini, M.C., & Keller, M.B. (2002). Short-term
diagnostic stability of schizotypcal, borderline, avoidant, and obsessive-compulsive
personality disorders. American Journal of Psychiatry, 159, 2036-2041.
Shea, M.T., Widiger, T., & Klein, M. (1992). Comorbidity of personality disorders and
depression: implications for treatment. Journal of Clinical and Consulting Psychology,
60, 857-868.
Shea, M.T., Zlotnick, C., & Weisber, R.B. (1999). Commonality and specificity of personality
disorder profiles in subjects with trauma histories. Journal of Personality Disorders,
13(3), 199-210.
Shedler, J., & Westen, D. (2007). The Shedler-Westen Assessment Procedure (SWAP): Making
personality diagnosis clinically meaningful. Journal of Personality Assessment, 89(1),
41-55.
Skodol, A.E., Gunderson, J.G., Shea, M.T., McGlashan, T.H., Morey, L.C., Sanislow, C.A.,
Bender, D.S., Grilo, C.M., Zanarini, M.C., Yen, S., Pagano, M.E., & Stout, R.L. (2005).
The collaborative longitudinal personality disorders study (CLPS): Overview and
implications. Journal of Personality Disorders, 19(5), 487-504.
Tellegen, A. (1982). Brief Manual for the Mulitdimensional Personality Questionnaire.
Minneapolis: University of Minnesota.
Tellegen, A., & Waller, N.G. (1982). Exploring personality through test construction:
Development of the Multidimensional Personality Questionnaire. In Boyle, Matthews, &
Saklofske (Eds.), The SAGE Handbook of Personality Theory and Assessment, Volume
Two: Personality Measurement and Testing (pp. 261-292). Thousand Oaks California:
Sage Publications.
Tracey, T. J. (1997). RANDALL: A Microsoft FORTRAN program for a randomization test of
hypothesized order relations. Educational and Psychological Measurement, 57, 164-168.
Traupman, E.K., Smith, T.W., Uchino, B.N., Berg, C.A., Trobst, K.K., & Costa, P.T.
(2009). Interpersonal circumplex octant, control, and affiliation scales for the
NEO-PI-R. Personality and Individual Differences, 47, 457–463.
van Os, J., & Jones, P.B. (2001). Neuroticism as a risk factor for schizophrenia. Psychological
Medicine, 31, 1129-1134.
von Eye, A. & Bogat, A.G. (2006). Person-oriented and variable-oriented research: Concepts,
results, and development. Journal of Development Psychology, 52(3), S31-S40.
Ward, J.H. (1963). Hierarchical grouping to optimize an objective function. Journal of the
American Statistical Association, 58, 236-244.

86

Westen, D., & Heim, A.K. (2003). Disturbances of self and identity in personality disorders. In
M.R. Leary & J.P. Tangney (Eds.), Handbook of Self and Identity (pp. 643-664). New
York: Guilford Press.
Westen, D., Novotny, C.N., & Thompson-Brenner, H. (2004). The empirical status of
empirically supported psychotherapies: Assumptions, findings, and reporting in
controlled clinical trials. Psychological Bulletin, 130(4), 631-663.
Widiger, T.A., & Smith, G.T. (2008). Personality and psychopathology. In John, Robins, &
Pervin (Eds.), Handbook of Personality Theory and Research, Third Edition (pp. 743769). New York: The Guilford Press.
Widiger, T. A., Verheul, R., & van den Brink, W. (1999). Personality and psychopathology. In
L. A. Pervin& O. P. John (Eds). Handbook of personality: Theory and research (2nd ed.,
pp. 347-366). New York: Guilford.
Wiggins, J. S. (1996). An informal history of the interpersonal circumplex tradition. Journal of
Personality Assessment, 66, 217-233.
Wiggins, J.S., & Pincus, A.L. (1989). Conceptions of personality disorders and dimensions of
personality. Psychological Assessment: A Journal of Consulting and Clinical Psychology,
1(4), 305-316.
Wishart, D. (2006) Clustan Graphics Primer: A Guide to Cluster Analysis, 4th ed. Edinburgh,
Scotland: Clustan Limited.
Yen, S., Shea, T.M., Battle, C.L., Johnson, D.M., Zlotnick, C., Dolan-Sewell, R., Skodol, A.E.,
Grilo, C.M., Gunderson, J.G., Sanislow, C.A., Zanarini, M.C., Bender, D.S., Rettew,
J.B., & McGlashan, T.H. (2002). Traumatic exposure and posttraumatic stress disorder in
borderline, schizotypal, avoidant, and obsessive-compulsive personality disorders:
Findings from the Collaborative Longitudinal Personality Disorders Study. The Journal
of Nervous and Mental Disease, 190 (8), 510-518.
Zanarini, M.C., Frankenburg, R.F., Sickel, A.E., & Yong, L. (1996). The diagnostic interview for
DSM-IV personality disorders (DIPD-IV). Mclean Hospital: Belmont, MA.
Zanarini, M.C., Skodol, A.E., Bender, D., Dolan, R., Sanislow, C., Schaefer, E., Morey, L.C.,
Grilo, C.M., Shea, T.M., McGlashon, T.H., & Gunderson, J.G. (2000). The Collaborative
Longitudinal Personality Disorders Study: Reliability of Axis I and Axis II diagnoses.
Journal of Personality Disorders, 14 (4), 291-299.
Zlotnick, C., Johnson, D.M., Yen, S., Battle, C.L., Sanislow, C.A., Skodol, A.E., Grilo, C.M.,
McGlashon, T.H., Gunderson, J.G., Bender, D.S., Zanarini, M.C., & Shea, T. (2003).
Clinical features and impairment in women with Borderline Personality Disorder (BPD),
with Posttraumatic Stress Disorder (PTSD), BPD without PTSD, and other Personality
Disorders with PTSD. Journal of Nervous and Mental Disease, 191 (11), 706-714).

87