Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
The Pennsylvania State University
The Graduate School
College of the Liberal Arts
COMPARING METHODS TO MODEL STABILITY AND CHANGE IN
PERSONALITY AND ITS PATHOLOGY
A Dissertation in
Psychology
by
Aidan Gregory Craver Wright
© 2012 Aidan Gregory Craver Wright
Submitted in Partial Fulfillment
of the Requirements
for the Degree of
Doctor of Philosophy
August 2012
ii
The dissertation of Aidan Gregory Craver Wright was reviewed and approved* by the following:
Aaron L. Pincus
Professor of Psychology
Dissertation Advisor
Chair of Committee
David E. Conroy
Professor of Kinesiology and Human Development and Family Studies
Kenneth N. Levy
Associate Professor of Psychology
D. Wayne Osgood
Professor of Sociology
Melvin M. Mark
Professor of Psychology
Head of Department of Psychology
*Signatures are on file at the Graduate School.
iii
ABSTRACT
As it stands now, the psychopathology of personality disorders (PD) is at a crossroads,
and there is little agreement on the best way to conceptualize and define PD. This lack of
consensus has led to problems not only in the basic definition of PD, including the accurate
description of the structure, course, and risk/protective factors of the disorders, but this
disagreement also threatens the advance of future science, and imperils attempts to develop
appropriate assessment and effective interventions for this debilitating group of disorders. The
current work builds on past cross-sectional work that has shown that PD and personality traits
are consistently and significantly related, and longitudinal work that has shown that both PD and
personality are plastic and change across time. Three studies using the Longitudinal Study of
Personality Disorders were conducted to address questions about the long-term stability of
interpersonal aspects of personality, the implications of PD symptom distribution on models
relating personality and PD, and the longitudinal relationship between personality and PD. Each
of these questions has important bearing on the manner in which we understand the development
of personality, PD, and the relationship between the two. The results of these studies demonstrate
that 1) interpersonal style is highly stable but mutable, depending in part on how change and
stability are operationalized; 2) non-normal distributional assumptions provide a better fit for
models of the relationship between PD and normative personality structure; and 3) individual
growth in personality is associated with concurrent growth in avoidant PD symptoms. Results of
the proposed study have implications for the ongoing efforts to establish the appropriate
definition, diagnosis, and treatment of PD.
iv
TABLE OF CONTENTS
PAGE
List of Figures…………………………………………………………………………………..vi
List of Tables.….….….….….…………………………………………………………………vii
Acknowledgements ..…………………………………………………………………………viii
Dedication………...……………………………………………………………….……………x
Chapter 1: General Introduction…………………………………………………………..……1
Chapter 2: Interpersonal Development, Stability, and Change in Early Adulthood…………...4
Introduction…………………………………………………………………………….4
Standard Approaches to Measuring Development, Stability, and Change……..7
Circumplex Parameters: Multivariate Tests of Development, Stability, and
Change………………………………………………………………………....11
Method.………………………………………………………………………………..16
Participants.…….……………………………………………………………...16
Procedure.……………………………………………………………………...17
Measures……………………………………………………………………….17
Analysis and Results…………………………………………………………………...18
Standard Analyses of Stability/Change………………………………………...18
Stability and Change of Circumplex Parameters………………………………23
Discussion……………………………………………………………………………...26
Limitations……………………………………………………………………..30
Future Directions…………………………………………………………….…31
Conclusion………………………………………………………………….…..32
Chapter 3: An Empirical Examination of Distributional Assumptions Underlying the
Relationship between Personality Disorder Symptoms and Personality Traits…….…40
Introduction…………………………………………………………………………...40
Continuity and Discontinuity in Personality and its Pathology……………….40
Abnormal Personality, Non-Normal Distributions, and Alternative Models.....43
The Current Study…………………………………………………………..…46
Method……………………………………………………………………….……..…48
Participants………………………………………………………………….....48
Procedure…………………………………………………………………..…..48
Measures…………………………………………………………………….....49
Results………………………………………………………………………………....50
Model Fit……………………………………………………………………....51
PAGE
v
Substantive Comparison of Models…………………………………………...52
Discussion……………………………………………………………………………..55
Implications for Modeling…………………………………………………….55
Implications for the Relationship between Personality and PD………………56
Limitations…………………………………………………………………….60
Conclusion…………………………………………………………………….60
Chapter 4: A Parallel Process Growth Model of Avoidant Personality Disorder Symptoms and
Personality Traits……………………………………………………………………...70
Introduction…………………………………………………………………………...70
The Current Study…………………………………………………………….73
Method………………………………………………………………………………..74
Participants……………………………………………………………………74
Procedure……………………………………………………………………...75
Measures………………………………………………………………………75
Data Analysis………………………………………………………………….76
Results…………………………………………………………………………………78
Discussion……………………………………………………………………………..79
Limitations…………………………………………………………………….82
Chapter 5: General Conclusion………………………………………………………………..87
References……………………………………………………………………………………..92
vi
LIST OF FIGURES
FIGURE PAGE
2.1 The interpersonal circumplex…………………………………………………………38
2.2 Example of structural summary parameters of a cosine curve……………………….39
3.1 Normal distribution fit to observed narcissistic personality disorder features……….66
3.2 Poisson and negative-binomial distributions fit to observed LSPD narcissistic
personality disorder features………………………………………………………….67
3.3 Representation of negative-binomial hurdle model for LSPD narcissistic
personality disorder symptoms……………………………………………………….68
3.4 Scatter plots of personality trait scores and NPD features…………………………...69
4.1 Conceptual diagram of the parallel process growth model…………………………..86
vii
LIST OF TABLES
TABLE PAGE
2.1 Descriptive statistics and rank order stability coefficients for the
interpersonal scales…………………………………………………………………….33
2.2 Growth models for the interpersonal scales……………………………………………34
2.3 Descriptive statistics for ipsative and circular variables……………………………….35
2.4 Correlations of circumplex measures with flux……………………………………….36
2.5 Correlations of structural summary statistics and IPC dimensions with spin,
pulse, D2, and q-correlations…………………………………………………………...37
3.1 Summary of Akaike and Bayesian information criteria for estimated models………..62
3.2 Summary of coefficients from models regressing personality disorder
symptoms on personality traits………………………………………………………...64
4.1 Parameter estimates and indices of fit for the five estimated parallel process
growth models…………………………………………………………………………85
viii
ACKNOWLEDGEMENTS
Agency without communion is no virtue; there are many who share this accomplishment
with me and who deserve my thanks. The data for these papers was graciously provided by
Mark F. Lenzenweger, who is remarkable for his generosity, kindness, and continues to be a
phenomenal collaborator. I am also indebted to the participants of the Longitudinal Study of
Personality Disorder, wherever they may find themselves. More broadly, I am grateful for the
faculty, staff, and graduate students of the Pennsylvania State University Department of
Psychology and Psychological Clinic, it is an idyllic environment in which to study and grow.
The contributions of clients and patients deserve mentioning, their lessons are in some ways the
most valuable and they permeate my thinking. I am honored and thankful to have David E.
Conroy, Kenneth N. Levy, and Wayne Osgood on my dissertation committee, and for their help
and advice along the way. This project would not have been possible without funding provided
by the National Institute of Mental Health (F31MH087053), the help provided by the PSU
Grants Office, and many others who supported that process.
All of my friends have taught me much, and I appreciate all the fun we have had over the
years. I appreciate the camaraderie and friendship offered by the graduate students from the
Personality Laboratory, and the additional mentorship offered by Emily B. Ansell, Nicole M.
Cain, and Mark R. Lukowitsky. Special thanks are due to Christopher J. Hopwood, who, as fate
would have it, was not a labmate, but continues to be a cherished colleague and friend.
I am indebted to my wife, Rachel L. Bachrach, for sharing her kindness, wit, humor, and
sharp mind. She propels me to work hard, and I owe her more than anyone for her love and
support. I look forward to our life together every day. Through some combination of genes and
environment, I have acquired from my parents and grandparents a strong, but late-blooming
ix
work-ethic, a knack for both pragmatism and flights of fancy, an analytical mind, and a keen
interest in the human condition. For these gifts, and their continued love and support, I am
grateful.
Most of all, I thank Aaron L. Pincus. Luckily for me, he can be coaxed in to taking
gambles from time to time, and I think this one has paid off. I cannot express how grateful I am
for his mentoring; he shares his compassion, wisdom, time, and effort generously, and expects
relatively little in return. He has been an unparalleled teacher, collaborator, advisor, and friend.
x
DEDICATION
To Rachel
1
CHAPTER 1
GENERAL INTRODUCTION
At the time of this writing, the psychopathology of personality disorders (PD) is very
much in flux. It has been just over a year since the workgroup charged with revising and
updating the section on personality and PD for the Diagnostic and Statistical Manual of Mental
Disorders, 5th
Edition (DSM-5) released their initial proposed changes (Skodol et al., 2011). The
proposal has been met with a strong reaction from the field. Three prominent journals have
either published (Journal of Personality Disorders; Personality Disorders: Theory, Research,
Treatment) or have special issues in press (Journal of Personality Assessment) devoted to
commentary, critique, and rebuttal addressing the suggested changes. These changes include a
“two-step” process of diagnosis, whereby the presence of a PD is first determined and rated for
severity, followed by a description of stylistic expression of the PD. Additionally, to aid in the
second step, a trait-based dimensional system is proposed to provide the descriptive content for
PD style (Krueger et al., 2011). In what some have construed as a contradictory approach
(Livesley, 2010), a set of PD types are also offered for matching a given patient to a paragraph
description.
Each of these suggested changes are a sharp departure from the existing nosology that
defines PD with 10 discrete categories that have corresponding symptom sets to allow for
polythetic diagnosis of each disorder. Aspects of the current nomenclature’s approach have been
met with sharp criticism over the years, including high rates of co-occurrence among the
categorically defined disorders (Krueger & Tackett, 2003; Widiger & Clark, 2000), boundary
definition issues (Widiger & Clark, 2000), the frequent use of PD not otherwise specified in
practice (Verheul & Widiger, 2004), temporal instability of symptoms (Lenzenweger, Johnson,
& Willett, 2004; Skodol, 2008) and the arbitrary nature of symptom cutoffs (Huprich &
2
Bornstein, 2007). It is clear that the workgroup endeavored to address these issues with the
proposed changes, but it remains unknown how successful the implementation will be. Indeed, it
remains unknown what the final articulation of PDs in DSM-5 will be, and psychopathologists
and practitioners interested in PD are anxiously awaiting the results of field trials and further
discourse from the workgroup.
In part because of the collective concern about both the existing nomenclature and
proposed changes, this is a vibrant time to be conducting PD research. Perhaps now more than
ever are the basic definitional issues of PD being discussed and argued across a number of
scientific venues. Separate from the scientific climate, there have been considerable
methodological advances that can be harnessed to address some of the core questions in the
psychopathology of PD. The past half century has seen advances in structural equation and
multi-level regression modeling that allow researchers to ask complex questions about the
manner in which variables are related to each other. Questions about individual trajectories
across time can now be posed and answered in ways that were previously unavailable or very
difficult. Additionally, powerful statistical software can be employed to model the distribution
of variables allowing for more specificity and sophistication in quantitative investigations.
Taken together, these advances provide the tools to begin addressing scientific questions that
were previously mere hypotheses.
The overarching goal of this dissertation is to address empirically some of the issues
related to the way PD is defined and understood. Broadly, this work seeks to clarify the manner
in which PD is related to personality. The underlying assumption here is that any theory or
understanding of PD should ultimately be unified or rooted in basic conceptions of personality.
To analogize, the cardiologist’s understanding of cardiac pathology is rooted in the same
understanding of the heart muscle that the basic anatomist has. More specifically, the
3
relationship between PD symptoms and personality traits will be examined cross-sectionally and
longitudinally.
The structure of this dissertation is as follows. The middle three chapters are each
distinct papers, written to stand alone and in a style intended for journal publication. As such,
there are certain redundancies the reader should expect (e.g., the description of subject
characteristics; measure description), especially since each study uses the same sample and
measures. Furthermore, these papers were not written in an entirely linear fashion. In particular,
the papers that are encompassed in Chapters 2 and 3 do not have direct bearing on each other. In
contrast, the paper that is the body of Chapter 4 builds upon Chapters 2 and 3, which were the
necessary preliminary steps. Chapter 2 investigates the longitudinal stability and structure of
interpersonal traits. Chapter 3 examines the relationship between PD symptomatology and
personality traits cross-sectionally, but the approach is novel in that the distribution of PD
symptoms and the implications of these distributions are given a central focus. Finally, Chapter
4 serves as an exemplar for longitudinal work examining the relationship between PD and
personality. The focus in this final empirical chapter is on avoidant personality disorder, but the
method is general and can be extended to other disorders and covariates in the future.
4
CHAPTER 2
Interpersonal Development, Stability, and Change in Early Adulthood
No man ever steps in the same river twice, for it is not the same river and he is not the same man.
– Heraclitus (~500, B.C.E.).
In recent decades, there has been much empirical investigation and ensuing debate on
whether, and if so when and how personality changes or develops across the life-span (Costa &
McCrae, 1997; Roberts, Wood, & Caspi, 2008). It now seems to be incontrovertible that
individuals’ personalities are highly stable, while not being entirely so, and the level of stability
depends, in part, on how stability is defined. In adulthood, rates of mean change in broad
personality traits are modest (Roberts, Walton, & Viechtbauer, 2006), individual differences in
these traits are generally maintained (Roberts, Caspi, & Moffit, 2001; Roberts & Delvecchio,
2000), and individuals appear to mostly preserve their intraindividual profile over time (i.e.,
ipsative stability; Donnellan, Conger, & Burzette, 2007; Roberts et al., 2001). However, high
stability is not stasis, and both normative and idiosyncratic development and change occurs. For
example, individuals demonstrate significant interindividual heterogeneity in intraindividual
trajectories around the population’s mean rate of change (Mroczek & Spiro, 2003; Vaidya, Gray,
Haig, Mroczek, & Watson, 2008), rank-order stability coefficients are significantly different
from unity (Srivastava, John, Gosling, & Potter, 2003), and there is a consistent minority of
individuals whose profile changes drastically over time (Donnellan et al., 2007; Roberts et al.,
2001; Robins, Fraley, Roberts, & Trzesniewski, 2001). Thus, the accumulation of findings has
pointed to a nuanced picture of personality development and change. Research that describes the
basic rates and patterns of stability and change in personality is important because it serves as a
5
necessary base from which to launch further excursions that investigate the determinants and
consequences of such stability and change.
Although there is evidence for development throughout the life span, early adulthood has
been identified as a time of marked and vibrant personality development (Roberts et al., 2006).
Unsurprisingly, these years, which bracket the transition from adolescence to adulthood, have
been the focus of much theoretical and empirical interest (Arnett, 2000). It is no doubt that these
years are an important time to investigate change and development, as this time period, which
includes the college years, has been recognized for the unique developmental challenges posed to
individuals (Arnett, 2000; Rindfuss, 1991). Of primary importance for individuals in our society
during this period is learning how to effectively navigate the tasks of getting ahead while getting
along (Hogan, 1983; 1996). During this time many leave home for the first time, take their first
jobs, enroll and attend college, begin their first serious romances, incur debt, and become
ultimately responsible for their behaviors.
The rapid accumulation of studies of personality development in early adulthood over the
last decade reflects the importance of this time period (e.g., Donnellan et al., 2007; Robins et al.,
2001; Vaidya et al., 2008). However, a notable absence from this literature has been the
Interpersonal Circumplex (IPC) model of personality (see Figure 2.1). Indeed, despite its
prominence in the broader personality literature (e.g., Horowitz & Strack, 2010; Pincus &
Ansell, in press), to date no study has used the IPC to investigate the long-term development,
stability, and change of personality in any age group. This absence is not trivial, given that
longitudinal investigations of extraversion and agreeableness (reviewed below) have thus far
demonstrated equivocal mean change findings, especially in early adulthood. The IPC maps
interpersonal functioning (i.e., personality; Leary, 1957; Sullivan, 1953; Wiggins, 1991) using
the primary orthogonal domains of Dominance and Affiliation (see Pincus & Ansell, 2003 for a
6
review). Dominance and Affiliation share close conceptual and empirical relationships to the Big
Five (Goldberg, 1990) traits of Extraversion and Agreeableness (Ansell & Pincus, 2004; McCrae
& Costa, 1989; Pincus, 2002). However, the traits mapped by the IPC are comprised of the
primarily interpersonal aspects of personality and therefore serve as a conceptually related but
somewhat distinct analytic framework.
The majority of the work on personality development has either used, or been
summarized within, the framework of the Big Five traits (e.g., Roberts et al., 2006). Mean
change of extraversion has been associated with inconsistent results, with some investigators
finding increases, and others finding stability in scores over time (e.g., Schuerger, Zarrella, &
Hotz, 1989; Vaidya et al., 2008). These results have been clarified by separating extraversion
into subcomponents of social dominance and vitality, with the former increasing and the latter
not (Roberts et al., 2006). Relatedly, some studies find robust increases in agreeableness over the
college years (Neyer & Lehnart, 2007; Robins et al., 2001; Vaidya et al., 2008). Yet others have
found decreases in agreeableness (Neyer & Asendorpf, 2001) or in similar variables (e.g., Social
Closeness; Donnellan et al., 2007; Roberts et al., 2001). And, as might be expected given the
equivocal nature of individual study results, the meta-analytic result has been one of no change
in this time period (Roberts et al., 2006). These ambiguous results for agreeableness might be
clarified by a more fine-grained analysis of affiliation, as has been the case for extraversion.
Bleidorn, Kandler, Riemann, Angleitner, and Spinath (2009) have taken a facet-level analytic
approach to extraversion and agreeableness in adults primarily in their 30’s, but similar
investigations do not exist in the early adult age-groups.
In fact, the vast majority of the research on personality development that has been
reported has occurred at the broad domain level of analysis, with more nuanced investigations of
the component parts or facets of these domains lagging behind. This more detailed level of
7
analysis is likely to be informative, and has already been demonstrated invaluable as it pertains
to certain traits which contain features that develop at differing rates (e.g., extraversion, Roberts
et al., 2006; conscientiousness, Jackson et al., 2009). Furthermore, different aspects of
personality functioning have been associated with differential patterns of change and stability
(e.g., affective traits; Vaidya et al., 2008). Prior research has used a number of models to
investigate the development and change in personality over time, but none have been as focally
interpersonal as the IPC. Leaving home, moving in with new people, choosing a major, starting
to work, and navigating new bosses, friends, and lovers are all associated with, if not driven by
interpersonal functioning. As such, interpersonal development, stability, and change are
important targets of investigation during this time period. Thus, the research reported here aims
to provide a focused and detailed examination of interpersonal development in early adulthood.
This sample is drawn from the Longitudinal Study of Personality Disorders (LSPD;
Lenzenweger, 2006), a large, prospective multiwave study of personality and its disorder. Three
waves of data have been collected thus far charting the development of basic personality and its
disorder across the college years (i.e., 18-22). This is the only sample I am aware of that has
assessed the participants using a well validated IPC based measure, the Revised Interpersonal
Adjective Scales (IAS-R; Wiggins et al., 1988). The IAS-R provides an assessment of
Dominance and Affiliation at the broad domain level and the more specific component parts of
such domains as assessed by the octants of the IPC. Therefore, the current paper answers the call
for more detailed investigations of personality development that focus on lower-order personality
traits (Roberts et al., 2008; Roberts et al., 2006). Additionally, as will be described below, the
current investigation moves beyond univariate approaches to studying personality development
by expanding into multivariate approaches based on the geometric structure of the IPC.
Standard Approaches to Measuring Development, Stability, and Change
8
Five complementary approaches to measure stability and change in personality
development have been routinely employed (e.g., Donnellan et al., 2007; Roberts et al., 2001;
Robins et al., 2001). Structural stability refers to the stability in the pattern of covariation in
variables across time. In other words, do the variables of interest relate to each other in the same
way at each time point of the study? Commonly this is equated with measurement invariance
across time, and is generally seen as a prerequisite for conducting further analyses of
development and change over time. No prior research has investigated the measurement
invariance of the IAS-R over time periods of multiple years.
Rank-order stability reflects the maintenance of interindividual position, or individual
differences over time. This is assessed using the correlation between scores at two time points,
and prior meta-analytic results have found rank-order stability values of r = .54 for the age-group
investigated here (Roberts & Delvecchio, 2000). Differential stability appears to vary by age-
group investigated (with stability increasing with age; Roberts & Delvecchio, 2000; Vaidya et
al., 2008), by personality trait (Roberts et al., 2001), and by population of interest (e.g.,
borderline personality disorder is associated with less stability; Hopwood et al., 2009).
Absolute or normative stability refers to changes in mean level over time. Changes in the
average level of personality dimensions over time are not necessarily related to changes in
differential stability. Absolute change refers to the group change, irrespective of the individual
shuffling that may occur. Significant observed mean change in personality traits is often thought
to map maturational and basic developmental processes be they biological, socialized, or a
combination of the two. As reviewed above, mean change in personality traits associated with
interpersonal functioning has been somewhat equivocal in the age group charted here. Based on
results reviewed above, it is difficult to predict whether domain level Dominance or Affiliation
will demonstrate mean change, as they are broad variables that blend content such as social
9
dominance, gregariousness, warmth, arrogance and agreeableness. It may be that Affiliation will
increase as higher warmth and communion is associated with increased functional maturity
(Roberts et al., 2001, 2003). However, the more detailed analysis of the IPC octant scales may
shed light on some of the past equivocal results associated with agreeableness reviewed above.
One of the important contributions of this study is the ability to chart change among
lower-order personality constructs, a level of analysis not previously pursued in this age group. I
expect that there will be an increase in octant level Assuredness-Dominance, and an associated
decrease in Unassuredness-Submissiveness, which isolate the poles social dominance. In
contrast, I do not expect that Gregariousness-Extraversion to increase and Aloof-Introversion is
expected to remain stable as well, as these are conceptually akin to the social vitality variables
that have previously demonstrated considerable stability. I anticipate finding that Arrogance-
Calculatingness to decline and Unassumingness-Ingenuousness to increase. This would also be
associated with increased functional maturity and less antagonistic, brash, and self-centered
behavior, which decrease as individuals mature (Hogan & Roberts, 2004). Pure warmth and
affiliation, marked by the octants of Warm-Agreeable and, inversely, Cold-Hearted, remain
somewhat of a question. Warmth is an aspect of social vitality, which is not expected to change;
yet affiliating with others would seem to follow the principle of increased maturity. Readers
likely have noted that the individual octants can be thought of as marking the poles of bipolar
dimensions, allowing for a separate study of each end of these continua. Although I expect that
the poles of each of the four axes to be highly entrained in their trajectories across time, the IPC
scales offer the opportunity to observe that directly, and evaluate whether that is indeed the case.
This is an attractive feature of the IPC structure not generally offered by most existing
personality measures.
Individual stability examines the variation in individual trajectories of change over time.
10
This is conceptually related yet distinct from the differential stability described above. Analyses
of individual stability build upon initial descriptions of differential stability by quantifying the
amount of interindividual variation associated with these intraindividual rates of change. This
provides a quantification and statistical test of the heterogeneity in trajectories, and allows one to
determine if there is significant variability in trajectories. Individual stability has been assessed
in a number of ways. However, individual growth curve (IGC) modeling offers the most
sophisticated approach to charting the variability in individual trajectories, but requires more
than two assessment points (Singer & Willett, 2003). Given that the LSPD has had three
assessment points, I will examine heterogeneity in linear rates of change. Past studies have found
significant interindividual variability in rates of change over time (Mroczek & Spiro, 2003;
Vaidya et al., 2008) and I anticipate finding similar results here.
Ipsative stability assesses the stability of an individual’s personality profile across time.
As such, it is a person-centered approach to change, capturing intraindividual variability or
stability in personality organization. Most commonly, ipsative stability has been measured using
Cronbach & Gleser’s (1953) D2 statistic or the q-correlation (i.e., the product-moment
correlation of individual profiles) across time-points. These approaches provide similar but
slightly different information. The D2 statistic is a direct index of total difference between an
individual’s profiles at two time-points, is unbounded on the upper end, and is calculated as the
sum of the squared differences between individual scales in the profile. Therefore it is a gross
measure of the difference between two profiles, sensitive to changes in elevation, scatter, and
shape. In contrast, the q-correlation controls for mean level and scatter in the profiles, providing
a measure of consistency in the patterning (shape) of two profiles. Regardless of the method, it is
common to find high levels of ipsative stability in personality profiles across time on the average
(Donellan et al., 2007; Robins et al., 2001), and similar results are expected with the IPC scales.
11
Circumplex Parameters: Multivariate Tests of Development, Stability, and Change
IPC based measures are interesting in that they provide a framework for more specific
tests and investigations based on circumplex structure that go beyond the standard measures of
stability and change across time. The studies reviewed above generally have looked at
development in personality traits separately (except for ipsative change that is predicated on a
patterning of scales). This is perhaps not surprising as the oft studied broad traits are putatively
orthogonal in the population, and thus univariate approaches might make the most sense.
However, the scales that comprise the IPC are specified to have a very precise structure that can
be used to extend the study of development, stability, and change in multivariate space. Although
often summarized with the primary dimensions of Dominance and Affiliation, in actuality the
IPC arises from a specific pattern of multivariate relationships among the more fine-grained
octant-level interpersonal variables. This feature can be contrasted with other personality models
and measures which do not define any specific structure among the component scales. Take for
example the NEO Personality Inventory Revised (Costa & McCrae, 1992) facet scales; although
they are expected to be correlated within a factor, there is no more specific structure offered. The
same is true of other models and measures such as the HEXACO (Ashton & Lee, 2004).
A brief review of the structure of IPC measures is warranted. Modern circumplex
measures, of which the IAS-R is one of the most popular, generally divide the full breadth of
interpersonal content in to eighths, assigning a scale to each octant of the circle. This level of
analysis offers a balance of fidelity and reliability, allowing for relatively fine discriminations of
interpersonal content, while also providing sufficiently reliable measurement scales necessary for
the construction of IPC structure. This structure is highly defined, with the conceptual and
empirical relationship between two scales determined by the inverse of their angular distance. In
other words, scales that are closer together in the circle are conceptually more related, and the
12
relationship between any two scales diminishes as the angles between them grows, with the
lowest correlation occurring between scales at 180° (i.e., opposite sides of the circle). This
pattern of associations, when represented in correlations, gives rise to a circulant correlational
pattern first defined by Guttman (1954). The IAS-R (see Figure 2.1) was constructed such that its
eight scales possess a circumplex structure, with the scales at 90° being approximately
orthogonal, and those at 180° being strongly negatively correlated. The two main orthogonal
domains of Dominance and Affiliation are derived from a weighted combination of the octants
based on their theoretical location on the circle (Wiggins et al., 1989). Note that IPC based
measures use a two-letter shorthand to denote angular location of the octant scales (e.g., PA, BC,
etc.). This system is akin to directional coordinates (e.g., NE, SW, etc.) in geographical Cartesian
planes, and was instituted by Leary (1957) to allow for easy communication of content across
measures and levels of functioning (e.g., motivations, behavior, cognitions). The nature of
circumplex scales allows for interesting investigations of unique trigonometric parameters.
Structural Summary /Cosine Curve Modeling is an approach to summarizing circumplex
data that builds on the structure described above (Gurtman & Balakrishnan, 1998; Wright,
Pincus, Conroy, & Hilsenroth, 2009). Just as the pattern of correlations among circumplex scales
is expected to result in a circular array, an individual’s scores are also expected to conform to
this pattern. Taking an individual’s highest scale score, the predicted pattern of scores on the
remaining scales would be slightly lower for scales measuring conceptually related content and
decreasing as the angular distance increases. To illustrate, envision an individual who describes
themselves as highly dominant; they are unlikely to also describe themselves as submissive at the
trait level. However, to the extent that they describe themselves as dominant, they are likely to
describe themselves with similar but slightly lower levels of related features, such as arrogance
or gregariousness (i.e., adjacent octants). If the prototypical predicted pattern of scores were
13
perfectly met, their profile would be precisely sinusoidal in form (see Figure 2.2). This is
because the scales conceptually and semantically constrain most individual’s patterns of
responses.
Figure 2.2 illustrates how such a curve can be comprehensively described by reducing it
to three structural parameters, angular displacement, elevation, and amplitude. The derivation of
these parameters has previously been well summarized and is not repeated here for reasons of
space (see Gurtman & Pincus, 2003 and Wright et al., 2009 for accessible reviews).
Nevertheless, the interpretation of these parameters merits some discussion. A profile’s angular
displacement refers to the location on the IPC associated with an individual’s predominant
interpersonal “theme” or “typology” (Kiesler, 1996; Leary, 1957) and is most commonly
reported in degrees from 0°. Elevation represents the average score across scales, and is
anticipated to be zero in IPC measures without a substantive first factor, like the IAS-R, because
opposing scale scores should cancel each other out. Individual profiles on the IAS-R with an
elevation are most likely produced by specific response styles (e.g., acquiescence). Amplitude
refers to how differentiated the profile is. It captures how much an individual discriminates
between interpersonal content in describing their interpersonal style. Stated otherwise, it is the
degree to which someone endorses that their interpersonal style is one way, and not other ways.
As can be seen in Figure 2.2, amplitude is the distance between the elevation (i.e., mean score),
and the peak of the curve (i.e., the angular displacement, or interpersonal theme of the profile).
Finally, the prototypicality of a profile, or the degree to which it matches a perfect cosine curve,
can be quantified by measuring the goodness-of-fit between an observed profile of scores and
those that would be predicted from a curve created using the structural summary parameters.
This goodness-of-fit between the observed and predicted cosine curve is labeled, R2.
Amplitude is identical, mathematically, to vector length, which was originally used to
14
summarize an interpersonal profile (e.g., Leary, 1957; Wiggins, Phillips, & Tranpnell, 1989).
Vector length, and by extension amplitude, is highly relevant in the current context because it is
associated with predictions about stability/rigidity in interpersonal behavior (Tracey, 2005;
Tracey & Rohlfing, 2010). Specifically, amplitude, which is associated with a more extreme
profile (Wiggins et al., 1989), has been hypothesized to predict rigidity or a narrower sampling
of interpersonal behaviors over time (Tracey, 2005). However, results from studies that have
investigated this have been equivocal (Tracey 2005; Tracey & Rohlfing, 2010; Erickson,
Newman, & Pincus, 2009). Although the current study is not examining stability in specific
interpersonal behaviors across time, related hypotheses might be associated with these structural
variables as they pertain to longer-term interpersonal stability. In particular, a related hypothesis
in this context would be that individuals with more differentiated interpersonal profiles show less
change in interpersonal style over time. Similarly, those individuals who have more prototypical
interpersonal profiles might show less change over time. It is unclear whether either of these will
be the case; however it stands to reason that those individuals who show more characteristic and
well defined profiles are more likely to maintain their interpersonal style over time.
This type of analysis will be the first of its kind. It is somewhat akin to exploring ipsative
stability, but unlike traditional measures of ipsative change (e.g., D2 or q-correlations), there is
an ideal profile pattern based on circumplex structure whose substantive meaning (if any) will be
tested. In addition, differentiation and prototypicality, as variables, can be subjected to some of
the standard change analyses described above—namely, mean, rank-order, and individual
stability. Most meaningful, perhaps, will be the results from the mean change in each of these
parameters. This will test whether if as individuals mature their interpersonal profile becomes
more or less differentiated and prototypical. It is easy to imagine that as someone matures, they
become surer of themselves and who they are, and thus their differentiation and prototypicality
15
increase. On the other hand, it may be that as individuals mature they become more aware of
their multi-faceted nature, they view and describe themselves in less certain terms, and thus
differentiation and prototypicality decrease. These contrasting hypotheses will be tested.
Flux, Pulse, and Spin (Moskowitz & Zuroff, 2004) are three recently developed measures
of variability based on a circumplex data structure. Flux refers to individual variability in
specific interpersonal content (e.g., Assured-Dominance) across time, pulse refers to variability
in amplitude, and spin refers to the variability in angular location around an individual’s mean.
Moskowitz and Zuroff (2004, 2005) pioneered these statistics to capture the variability in
behaviors measured intensively across time, and they have not been previously applied to the
large temporal distances sampled here. Nevertheless, these measures of net interpersonal
variability might be informative, even at this longer multi-year time-scale. This is in part because
individuals do not always chart linear change across time, and may show decreases in a
personality variable between two time points, only to reverse course and change in the opposite
direction by the next sampling session. Even the sophistication of growth curve modeling is not
tuned to capture this type of lack of stability.
Additionally, pulse and spin capture multivariate variability, because an individual’s
amplitude and angular displacement are multivariate measures that are calculated from the
relationship among the octant scales. Spin, summarizes the net variability in content of an
individual’s interpersonal style, without anchoring it in any specific style (e.g., Dominance).
Calculating the descriptive statistics for these variables will be informative in ways that standard
measures of change have not. Recall that interpersonal style, which is described by angular
location on the IPC, is continuous, and is only coarsely summarized by the individual dimensions
of Dominance and Affiliation. Thus, spin between time points will quantify the amount of
change in interpersonal style that can be expected. This takes in to account simultaneous
16
change/stability in both Dominance and Affiliation, and will provide a comprehensive summary
of change in interpersonal type—change in any one dimension is only part of the story. It is
difficult to be confident in any a priori predictions of this type of stability, as it has not been
examined before at this time scale. Also, it is difficult to make predictions because spin is
separate from change in any particular content that might follow a maturational pattern. The
logic is similar for pulse.
However, specific predictions can be made regarding the relationship between the
structural variables described above (e.g., differentiation and prototypicality), and
change/stability in style. Specifically, it is anticipated that amplitude and R2 will be negatively
associated with angular shift, as those individuals who describe themselves in more
differentiated terms and in a more prototypical manner are likely to shift less in the way they
describe themselves. Returning to pulse, or variation in profile differentiation, it is again difficult
to make predictions. In past studies using individual behaviors as the level of analysis, pulse has
failed to show robust associations to other variables (e.g., Moskowitz & Zuroff, 2004, 2005;
Russell et al., 2007). Thus, I offer no specific predictions here, and anticipate that pulse at these
larger intervals is likely to be similarly unrelated to other variables.
In summary, this study will be the first to examine the development, stability, and change
in interpersonal style in early adulthood using the IPC framework. I will explore the standard
approaches to measuring personality development, but will expand beyond these by including
analyses of structure, stability, and change in interpersonal circumplex parameters.
Method
Participants
The 258 participants in the LSPD (Lenzenweger, 1999; Lenzenweger et al., 1997) were
drawn from a population consisting of 2,000 first-year undergraduate students. Extensive detail
17
concerning the initial participant selection procedure and sampling is given elsewhere
(Lenzenweger, 2006; Lenzenweger et al., 1997). The 258 participants consisted of 121 males
(47%) and 137 females (53%). The mean age of the participants at entry into the study was 18.88
years (SD = 0.51). Participants were subsequently assessed at their second and fourth years of
college. Of the initial 258 participants, 250 completed all three assessment waves and are
included in these analyses. Six left the study, and two died in automobile accidents. Of these
individuals 53% were female, 3.6% were African-American, 4.8% Hispanic/Latino, 72%
Caucasian, 17.2% Asian/Pacific-Islander, 0.8% Native-American, and 1.6% Other.
Procedure
Structure of the LSPD Data. As noted above, the LSPD has a prospective multiwave
longitudinal design with participants initially evaluated at three points in time (i.e., first, second,
and fourth years in college). At each time point, participants completed self-report measures of
personality. The average age of study of participants at the assessment waves were 18.88 years
(SD = 0.51) for Wave I (T1), 19.83 years (SD = 0.54) for Wave II (T2), and 21.70 years (SD =
0.56) for Wave III (T3). The mean time between entry into the study (T1) and T2 and T3 was
0.95 years (SD = 0.14) and 2.82 years (SD = 0.23), respectively. The LSPD data are balanced, in
that all participants have three waves of data, and are time structured such that each participant
was assessed repeatedly on the same three wave schedule, although the time between
assessments varies from case to case.
Measures
Revised Interpersonal Adjective Scales (IAS-R; Wiggins et al., 1988). This study uses the
64-item IAS-R which consists of eight scales assessing the eight octants of the IPC, which in
turn can be converted into scores for the two primary dimensions of the IPC, Dominance and
Affiliation, using standard scale weights (see Wiggins et al., 1989). We use the common three
18
letter abbreviations of DOM and LOV respectively for these dimensions in keeping with prior
IAS-R publications. Participants responded to each trait descriptive adjective (e.g., dominant,
coldhearted) on an 8-point scale at each wave of the LSPD. Internal consistency (α) for the
octant scales at each wave of assessment is provided in Table 2.1.
Analysis and Results
Standard Analyses of Stability/Change
Structural Stability
To test for structural stability among the interpersonal scales over time, we used multi-
group structural equation modeling to compare two models. The baseline model was estimated
with individual latent factors for each octant scale that were defined by fixing the loading of the
observed scales to 1.00 and the error variance of the scales at 0.00 and allowing the factor
correlations to be freely estimated within and across each wave of data collection. This creates a
pattern of factor correlations that are equivalent to the manifest matrix within each wave, and a
fully saturated model (i.e., df = 0; Δχ2 = 0.00; p = 1.00). In the second, more constrained model,
factor correlations were fixed to be invariant across time-points. A non-significant chi-square
change (Δχ2) between the baseline and constrained models would be indicative of structural
stability. This represents a stringent test of structural stability as all corresponding correlations
are tested for equivalence across each of the three time points. The resulting change in model fit
indicated that the IAS-R was structurally invariant across all three time-points (df = 72; Δχ2 =
64.82; p = .71).
Rank-Order Stability
Rank-order stability was assessed using the correlations between time-points on the
interpersonal scales. Results are summarized in the three rightmost columns of Table 2.1. In
general, each octant scale showed considerable rank-order stability regardless of the time
19
between assessment points (range of r’s = .68 - .86). Stability decreased as a function of time
between assessment points, with the relationship between T1 and T2 scores, the shortest time
distance, being the highest. The most stable octants in terms of rank-ordering were FG and NO,
the poles of the Introverted-Extraverted axis of the IPC. The least stable were DE and JK, but
this was only relatively so, even these octants demonstrated considerable stability in individual
differences. The dimensions of DOM and LOV exhibited even higher differential stability over
the course of the study. These results point to higher stability than results from previous studies
over this same time period, with the meta-analytic population estimate being only r = .54
(Roberts & Delvecchio, 2000), and others finding stability coefficients for Extraversion and
Agreeableness of .63-.72 and .59-.60 respectively (Robins et al., 2001; Vaidya et al., 2008).
Mean and Individual Level Stability
Mean-level and individual-level stability were studied using an individual growth curve
(IGC) approach within a multilevel modeling framework. ANOVA is an unattractive approach
for investigating mean change in this sample due to the variability in assessment timing for each
individual. Multi-level models are unencumbered by this limitation, and treat time as a
continuous variable. IGC analyses allow for the investigation of within person change over time
in personality traits (Singer & Willet, 2003). In this analytical framework, measurement
occasions (Level 1) are treated as nested within individuals (Level 2). Therefore, each individual
has a trajectory of change over time. The Level 1 model contains two important estimated
growth parameters—the intercept and slope. The individual intercept parameter represents the
mean elevation of the slope at the origin of the time scale. The individual slope parameter
represents the rate of change per unit of time. IGC modeling allows the coefficients for these two
parameters to vary randomly if there is significant interindividual variation in intercept and slope
in the sample. That is to say, each individual is allowed to take on their own values for intercept
20
and slope, which in turn can be explained by introducing between-person predictors at Level 2 in
the model. The general equation in multi-level format for the models estimated is given here:
where, is the outcome score (i.e., personality trait score) for individual i at time t; is the
intercept parameter of the hypothesized growth trajectory for individual i; is the slope
parameter for individual i (that is, the rate of change [yearly] in level of personality trait over
time); is the time at which assessment t of subject i took place, measured in years, and
centered on each individual subject’s age at entry into the study; and is a level 1 residual, or
the unexplained portion of the outcome, across all occasions of measurement, for individual i in
the population. It is assumed to be normally distributed with a mean of zero and a variance
defined by . is the average intercept (i.e., the mean score at the start of the study);
is the average slope ( ; i.e., rate of change); and are the level 2 residuals that
represent the deviation in individual values in intercept and slope. Their variances are
represented by and , and their covariance by .
Of importance for interpreting the results of IGC models are the fixed and random (i.e.,
variance) coefficients. The fixed coefficients ( and ) can be interpreted straightforwardly
in much the same way as basic multiple regression coefficients. These test whether the mean of
the coefficients (i.e., intercept and rate of change) are significantly different from zero. The
random effects ( and test whether significant variability remains unexplained in the
outcome variable (i.e., is there significant interindividual heterogeneity in intercept and slope).
Additionally, the covariance between the intercept and slope is reported, but it is not a
21
focus in this study. In these analyses, the models were fitted employing full maximum likelihood
estimation using HLM-6 (Raudenbush, Bryk, Cheong, Congdon, & Du Toit, 2004).
The results of the growth curve analyses are summarized in Table 2.2. Prior to all
analyses scales were standardized using the original sample, thus all values are in standard units.
The fixed effects for the intercept ( ) serve to compare the sample to the normative sample. On
average, this sample is more Affiliative (LOV = .64, p < .001), but no more Dominant (DOM
= -.05, p = .47) than the original normative group. This is mentioned only briefly here, as
these are provided as descriptive statistics for the interested reader. The main focus is on the
fixed ( ) and random effects ( ) associated with the slope. At the broadest level of analysis,
the sample showed mean increases in LOV ( = .04, p = .02), but no mean change in DOM
( = .00, p = .99). As noted above, these dimensions suffer from the same difficulties as other
scales that have given previous equivocal results, namely they are quite broad. By examining the
results of the octant scales the full and clearer picture emerges.
Four of the octant scales demonstrate mean change over the course of the study. These
can be neatly summarized by noting that PA increases while HI decreases and JK increases while
BC decreases. The results for PA accord well with prior findings of an increase in social
dominance. However, a more nuanced picture emerges when considering that there is no change
in NO, the Gregarious-Extraverted octant, but there are significant declines in BC, the Arrogant-
Calculating octant. Each of these octants is adjacent to PA and contains considerable dominant
content, but this is moderated by the affiliative content of the scale resulting in very different
mean trajectories. The inverse of this process can be found on the other side of the IPC with the
decrease in HI, but increase in JK. Thus, through early adulthood individuals become more
assertive, self-assured, and confident, while also becoming less boastful, cocky, and calculating.
22
Opposite patterns of mean growth in these adjacent octants also highlights that the association
between variables at any given time point (i.e., cross-sectionally) is not necessarily their
relationship across time. Separate processes are captured when variables are investigated
longitudinally which have important implications for individual development and maturation. It
is worth noting that although the mean changes catalogued in Table 2.2 appear modest, these
capture rate of change per year, and thus they do not represent the total change over the study.1
Further, mean change says nothing about the variability in that change, which I turn to next.
The variance components in Table 2.3 represent the variability of residuals around the
mean ( ) and slope ( parameters. For all of the octants and two dimensions of the IPC there
is significant variability in intercept and slope, indicating that there is rich interindividual
heterogeneity in the trajectories of interpersonal development. Thus, the modest (and often non-
existent) mean change exhibited at the group level should be understood in the context of wide
variability at the individual level. There are those for whom the yearly change is starkly
different. For example, the SD associated with a variance of .03 would be .17, which when taken
over 3 years would be over one half of a scale’s SD of change (i.e., .51). It follows that 32% of
the sample is changing over .5 SDs over three years. When examined at the individual level, a
picture emerges that is consistent with a significant degree of instability in each trait over time.
Ipsative Stability/Change
To measure ipsative stability, I employed Cronbach & Gleser’s (1953) D2 and the q-
correlation. Descriptive statistics for each between time point can be found in Table 2.3. Values
of D2 are difficult to directly interpret because they are not standardized or bounded. However,
these will be used below in correlational analyses. In contrast, q-correlations (rq) are readily
1 All IGC analyses were re-run using gender and age of entry to the study as level 2 predictors and neither was found
to predict rate of change in any scale.
23
interpretable in the same way as all product-moment correlations. The values in Table 2.3
demonstrate that on the average there is high stability in individual profile patterns, although
there is considerable range in stability. Only a small minority of these correlations were negative
(rq12 = 2.4%; rq23 = 3.2%; rq13 = 2.0%), and the majority exceeded rq = .80 (rq12 = 72.6%; rq23 =
68.4%; rq13 = 57.2%). Note that there is a gradual trend towards less stability as the distance
between measurement occasions increases. However, on the whole, there is a great deal of
stability in interpersonal profiles.
Stability and Change of Circumplex Parameters
Circumplex Structure
Circumplex structure was evaluated at each wave using RANDALL (Tracey, 1997); a
computer program based on Hubert and Arabie’s (1987) randomized ordering of hypothesized
relationships. In a set of eight scales presumed to conform to a circumplex, 288 predictions are
made about the relationships between scales. RANDALL calculates the number of predictions
satisfied and compares them to a distribution of randomized patterns derived from the observed
matrix to obtain a p value. The results of the RANDALL analyses indicate good conformity to
circumplex structure at each time point. Time 1 (288/288; CI = 1.00; p < .001), Time 2 (288/288;
CI = 1.00; p < .001), and Time 3 (285/288; CI = .99; p < .001).
Structural Summary and Circular Statistics
The structural summary parameters of angular displacement (i.e., interpersonal style),
elevation, amplitude (i.e., differentiation) and goodness-of-fit to a cosine curve (i.e.,
prototypicality) were calculated for each individual (see Gurtman & Balakrishnan, 1998 for a
summary of the methods). Table 2.3 contains the descriptive statistics for the sample at each
time point. As is readily seen in the table, at each time point the range of angular locations spans
the entire circumference of the circle although the mean is in the LM, or Warm-Agreeable
24
octant. It is worth noting that follow-up analyses that applied the structural summary
methodology to the sample as a whole indicated that there was significant variability around this
mean. The average elevation was very close to 0, as expected. Additionally, the average
amplitude suggests that individuals generally have well differentiated profiles (i.e., 1.37 SDs
between the mean and peak scores), but the range indicates that there are those for whom their
profile is flat and undifferentiated, and others who are remarkably differentiated. The full range
of possible goodness-of-fit/prototypicality was observed, with the median ranging between .82
and .76. These results suggest that on the whole individuals have prototypical profiles, but there
is a slight decrease in the prototypicality of interpersonal profiles over time. To test the pattern of
growth in differentiation and prototypicality, individual amplitude and R2 values were subjected
to IGC analyses. The results of these analyses can be found on the bottom of Table 2.2. On the
average there was stability in differentiation, but significant interindividual heterogeneity in
trajectories was found. There was a small average decrease in prototypicality of profiles over the
course of the study, but with significant variability in individual trajectories.
Longitudinal Flux, Pulse, and Spin
Flux was calculated as the standard deviation of an individual’s scores in the individual
octants across the three study waves. The flux scores for PA (M = .45; SD = .26), BC (M = .49;
SD = .31), DE (M = .43; SD = .30), FG (M = .40; SD = .28), HI (M = .45; SD = .25), JK (M =
.59; SD = .36), LM (M = .47; SD = .32), and NO (M = .45; SD = .30) were all of a similar
magnitude—close to half of a standard unit—and all were significantly different from zero.
These results suggest that on the average, individuals show significant variability around their
mean score on each octant. Pulse (M = .32; SD = .20), or the standard deviation of amplitude was
slightly more modest. Finally, spin (M = .86; SD = .18) was calculated as the circular standard
deviation following Moskowitz & Zuroff (2004). Given the manner in which circular standard
25
deviation is calculated (see Mardia & Jupp, 1999), the resulting metric of spin makes it difficult
to grasp the exact amount of change in interpersonal style. Therefore I provide the descriptive
statistics of angular change between individual time points in Table 2.3. Although the range
indicates that there are those whose score changes dramatically (i.e., almost 180°), the average
change in interpersonal style is more modest.
Predicting Development, Stability, and Change
In order to test whether the structure of an individual’s interpersonal profile (i.e., profile
differentiation and prototypicality) is related to development and stability I adopted a number of
approaches. First, to determine if either amplitude or R2 was related to linear growth in the
interpersonal scales I estimated a series of conditional growth models with each variable
included as a Level 2 predictor of rate of change in the IAS-R scales. With the exception of the
model for growth in LM, neither amplitude nor R2
were significant predictors of the rate in
structured individual growth. Individuals with more prototypical profiles showed a decreased
rate of growth in LM across time (γ11 = -.18, p = .02). Given that this is the sole significant result
and is a small effect, it is difficult to place much confidence in its meaning. Furthermore, it was
anticipated that these variables would be unrelated to structured change. I next correlated
amplitude and R2 with flux in the four cardinal points of the circumplex (i.e., PA, DE, HI, LM)
and results can be found in Table 2.4. In order to control for the inherent dependency in scores
that would result from using the same time-point’s scores in calculating both amplitude and R2
and flux, these flux variables represent the absolute change in an octant between only two waves
of data. Therefore, the amplitude or R2 parameter is not calculated from the same scores as are
included in the flux parameter. This approach is adopted for all subsequent analyses. The results
presented in Table 2.4, show that neither amplitude nor R2 are correlated with flux scores. Thus,
profile differentiation and prototypicality are unrelated to stability in any specific interpersonal
26
content. Finally, I tested whether profile differentiation and prototypicality are related to the
stability in the overall interpersonal profile. To test this, I correlated amplitude and R2 with spin,
pulse, and the measures of ipsative stability, D2 and q-correlations. Results are summarized in
Table 2.5. Interestingly, both differentiation and prototypicality were consistently correlated with
spin and q-correlations, but not pulse or D2. Individuals with more differentiated and prototypical
profiles had less angular change and higher q-correlations between assessments. What
differentiates spin and q-correlations from pulse and D2 is that the former are pure measures of
the stability in the pattern of a profile, or stated otherwise, the idiographic relationship between
scales. Thus, those individuals with more differentiated and prototypical profiles maintain their
idiographic profile more over time, regardless of changes in level or extremity.
As a final set of analyses, I also explored the relationship between dominance and
affiliation and stability by correlating DOM and LOV scores with flux, pulse, and spin scores.
The results for flux can be found at the bottom of Table 2.4. On the whole, LOV was related to
more stability (i.e., negatively related to flux) in LM and DE. However, LOV was unrelated to
stability in PA or HI. In contrast, DOM was generally unrelated to flux, with the exception of the
flux in PA and HI between times 1 and 3. Therefore, greater affiliativeness is related to less
change in warmth and unrelated to changes in dominance, whereas dominance is generally
unrelated to change in either. The results for pulse and spin can be found on the bottom of Table
2.5. Results suggest that on the whole, specific interpersonal style is unrelated to these types of
change, although some results suggest that higher warmth may be related to higher stability, but
the effects are inconsistent and very modest.
Discussion
This investigation studied the development, stability, and change in interpersonal aspects
of personality across the early adulthood years. Although a number of recent studies have begun
27
to focus on personality development during this time-period, none have used the IPC as the
organizing framework nor have there been any that are so focally interpersonal. Using standard
approaches to studying personality consistency and development, the results reported here are,
on the whole, largely consistent with those found using other models of personality. I found that
the IPC structure is stable across these years, individuals show a high degree of rank-order
stability, and mean changes are gradual and specific to certain aspects of interpersonal
functioning. Tests of ipsative stability suggest that an individual’s idiographic profile is highly
stable. At the same time, individuals do show significant heterogeneity in the rate of change in
interpersonal functioning, and there are those who show very dramatic shifts in their traits and
profile over this time-period. Therefore, the commonly found result of both stability and change
in personality traits appears to hold for the interpersonal traits as well. Among the most novel of
the results found here are those associated with the structural variables derived from the IPC. I
found that individuals who had more differentiated and prototypical profiles showed more
stability in the patterning of that profile over time, but this is unrelated to stability in any specific
interpersonal domain.
On the whole, it appears that rank-order stability in interpersonal functioning is very
high. Past meta-analytic results that have reported findings in the context of the Big Five traits
are consistent with these results, showing that extraversion and agreeableness are the most
differentially stable of the traits (Roberts & DelVecchio, 2000). However, the average 3-year
stability coefficient from the current results (r = .78) exceeds those reported by Roberts and
DelVecchio (r = .54). Although it is unclear why the differential stability found here for
interpersonal traits is higher than that for similar variables (i.e., extraversion and agreeableness)
as measured by similarly reliable instruments, one possibility might be the relative lack of
affective content of the IAS-R. Vaidya and colleagues (2008) demonstrated that affective traits
28
are much less stable over similar time periods than personality traits. It bears mentioning that the
affective traits in that study were measured using self-rated adjectives, just as in this study where
interpersonal functioning was measured using self-rated interpersonal adjectives. Thus, it may be
that individual differences in interpersonal style are maintained, even as other aspects of
personality are associated with more “shuffling of the deck.”
An attractive feature of this investigation was that this study was not limited to the broad
domain level of personality traits, but also examined the component parts of Dominance and
Affiliation. Perhaps the type of development for which this is the most illuminating is in mean
change. As reviewed in the introduction, there have been equivocal results associated with the
more interpersonal traits in the past. I replicated past results that have found mean increases in
social dominance (i.e., Assured-Dominant and Unassured Submissive octants), but stability in
social vitality (i.e., Gregarious-Extraverted and Aloof-Introverted octants). Interesting results
also came out of the lower-order examination of traits associated with agreeableness. Pure
warmth (i.e., Warm-Agreeable and Cold-Hearted octants) was remarkably stable, while there
was an increase in warm-submissive aspects of interpersonal functioning (i.e., Unassuming-
Ingenuousness and Arrogant-Calculating octants). Based on these results it seems that
individuals become less self-serving, argumentative, and disagreeable, but they do not
demonstrate average change in how charitable, kind, and sympathetic they are. This pattern of
development, points to an interesting picture of how these octants relate to each other.
Oftentimes, the octants that are blends of the two primary dimensions (i.e., BC, FG, JK, and NO)
are treated as just that, blends, not being unique constructs in their own right. However, the
results of this study would suggest that these are not merely blends that are reducible to the two
primary dimensions of dominance and affiliation. Instead, the continuous dimension of
interpersonal style that forms the circumference of the IPC is more of a qualitative one, with
29
marked shifts occurring as it is circumnavigated. There are fundamental differences in these
constructs that emerge when measured longitudinally. This is not to say that these variables do
not share close conceptual and empirical relationships, but rather that interesting differences
emerge in the “blends” that create a new recipe, not merely a sum of the interpersonal flavors.
Moving from the longitudinal pattern of adjacent octants to those that oppose each other,
a very consistent pattern emerges. Because the IPC’s structure allows for the separate
measurement of the opposing poles of its component dimensions, I was able to examine whether
stability and change were consistent across these “axes.” In general, the same (but inverse)
pattern replicates across each pole of the dimensions, but the magnitude of change is stronger for
those octants that are associated with less socially desirable behavior (i.e., HI and BC). Each pole
is remarkably well entrained with its partner, and the mean development is highly similar for
each pair. Past work (Donnellen et al., 2007; Roberts et al., 2006) has noted that the general trend
in mean change during this time period is associated with functional maturity. This was found
here as well. For example, individuals become more assured and confident, on the average, but
decrease in boastfulness and cockiness. It is easy to see how this relative reorganization allows
for more effective functioning across important adult situations. It is interesting that no average
increases in how neighborly, friendly, charitable, and sympathetic individuals are. As individuals
mature in young adulthood, they seem to maintain how distant or close they like others to be.
However, regardless of the content of the scales, all octants demonstrated significant
heterogeneity in individual rates of change. This suggests that although there are those who are
developing in a manner that is consistent with personality maturation, others are not, instead
taking trajectories that are perhaps better described as “regression” in the case of those who are
changing in the opposite direction, or “stagnation” for those who do not change at all (see also
Wright, Pincus, & Lenzenweger, 2011b). The determinants of these trajectories are of high
30
interest for future research.
Although IGC models are able to capture the individual across time, they are still limited
in that they focus on one trait at a time. Ipsative stability moves beyond this to capture the
stability in an individual’s profile over time. The q-correlations suggest that individuals are
highly stable at the level of their profile. Joining these traditional approaches to assessment are
the multivariate circumplex parameters that have not previously been applied to the temporal
distances studied here. The median change in interpersonal style is less than one octant’s width,
suggesting that individuals generally maintain the same style across time. Yet, there are those
who literally “do a 180°” in terms of their style over time. The IPC also offers a way in which to
quantify the relative structure of an individual’s interpersonal profile via differentiation and
prototypicality. Differentiation captures the degree to which an individual specifies themselves
as a certain type as opposed to other types. It is as if those with highly differentiated profiles are
saying, “This is what I am like!” In a similar vein, prototypicality captures whether an
individual’s profile follows the conceptual pattern associated with a particular style. Is their
profile patterned in a way that is consistent throughout, or are there idiosyncratic peaks and
valleys that defy the standard patterning. Interestingly, both differentiation and prototypicality
were associated with stability in the structure of one’s profile, but not specific types or absolute
degree of change. These initial results suggest that these structural variables are associated with
change over time, and beckon new investigations in to what other aspects of functioning they
might be related to.
Limitations
As with all studies, a number of limitations remain to be addressed in future
investigations. Notably, these results have very little to say about the mechanisms involved in the
development of interpersonal style over this time-period. Emerging results from other studies
31
have pointed to the influence of both genetics and environment in the development of personality
traits during this same time period (Hopwood et al., 2011). The higher-order traits of
Agency/Dominance and Communion/Affiliation are presumed to be associated with individual
differences in basic neurobiological structure and functioning, linked with incentive reward
systems (i.e., dopaminergic), and affiliative neuroendicrine functioning (e.g., vasopressin and
oxytosin; Depue & Collins, 1999; Depue & Lenzenweger, 2005; Depue & Marrone-Strupinsky,
2005). It would seem to be a safe assumption that the influences are multiple and complex, with
basic socialization and biological maturation each playing a role in orchestrating the harmonics
of development.
Additionally, one must always be mindful about the operationalization of personality.
Here I have adopted a trait approach to personality, with individuals providing self-ratings of
interpersonal style. Although similar pictures emerge when the perceptions of others are included
(see Donnellen et al., 2007 or Jackson et al., 2009), important differences may emerge when
these results are augmented with a second rater’s perception of an individual’s style. Are our
own perceptions more stable than the manner in which others perceive us? Moreover, this has
focused on only one level of functioning. Interpersonal theory explicitly states that functioning
occurs at multiple levels (e.g., biological, motivational, cognitive, behavior; Pincus & Wright,
2010). Here only one of these is captured—self-representation—but future studies would be wise
to capture more of these levels of functioning.
Future Directions
I am unaware of parameters in any other personality model that are conceptually akin to
the IPC’s differentiation and prototypicality. To the extent that these prove interesting, it may be
worth developing similar parameters for other measures and models. One approach might be to
break dimensions in to their polar scales and quantify how much they follow a similar pattern
32
across the poles in much the same way as is done here. Emotional functioning is also often
measured using a circumplex model, might it be that these parameters also serve to offer insight
in to the trajectories an individual charts in their emotional functioning across time? It may be
that these variables are not merely structural parameters, but have substantive interpretability as
well. I would be eager to investigate the substantive interpretation of these variables. It may be
that they have implications for identity, basic self-construal, and social cognition more generally.
The early adult years are interesting because they are a time of high-growth, but it is clear
that individuals continue to develop and change across other eras of the lifespan (Roberts et al.,
2008). A fourth wave of data collection for the LSPD is currently in the planning phases, with
the hopes that these same individuals, who are now in their mid-thirties, will provide us with the
insight into longer term stability of interpersonal functioning, extending beyond early adulthood.
Conclusion
The current study was the first to examine the development, stability, and change of the
interpersonal system as mapped by the IPC in any age group. The results using standard
articulations of stability and change are highly consistent with the results of others studies
following individuals during early adulthood. However, this investigation probed beyond the
broad domain level to study change in the lower-order interpersonal traits, a level of analysis that
is needed to fully understand the highly-nuanced development of personality. Finally,
approaches that capitalize on the circumplex structure of interpersonal variables were brought to
bear on these issues of development and shed new light on stability, change, and the structure of
personality.
33
Table 2.1. Descriptive Statistics and Rank Order Stability Coefficients for the Interpersonal Scales
Time 1 Time 2 Time 3 αT1 αT2 αT3 r12 r23 r13
Assured-Dominant (PA) 0.00 (1.09) 0.01 (1.05) 0.10 (1.00) 0.86 0.85 0.82 0.82 0.75 0.71
Arrogant-Calculating (BC) -0.64 (1.17) -0.89 (1.13) -0.91 (1.16) 0.91 0.91 0.92 0.80 0.77 0.73
Cold-hearted (DE) -0.31 (1.03) -0.37 (1.00) -0.40 (0.96) 0.88 0.88 0.87 0.78 0.73 0.68
Aloof-Introverted (FG) -0.46 (1.32) -0.51 (1.05) -0.45 (1.10) 0.92 0.91 0.91 0.84 0.80 0.75
Unassured- Submissive (HI) -0.29 (1.11) -0.37 (1.09) -0.49 (1.05) 0.89 0.89 0.87 0.83 0.78 0.74
Unassuming-Ingenuous (JK) 0.71 (1.26) 0.91 (1.30) 0.93 (1.28) 0.83 0.85 0.85 0.76 0.70 0.69
Warm-Agreeable (LM) 0.18 (1.12) 0.24 (1.13) 0.23 (1.10) 0.89 0.90 0.89 0.79 0.71 0.72
Gregarious-Extraverted (NO) 0.34 (1.27) 0.35 (1.20) 0.37 (1.21) 0.91 0.91 0.91 0.86 0.80 0.75
Dominance (DOM) -0.03 (1.17) -0.09 (1.13) -0.04 (1.09) -- -- -- 0.88 0.82 0.78
Affiliation (LOV) 0.60 (1.27) 0.75 (1.23) 0.75 (1.25) -- -- -- 0.85 0.81 0.78
Note. N = 250. Standard deviations presented in parentheses. All correlations significant at p < .01.
34
Table 2.2. Growth Models for the Interpersonal Scales
Elevation (Intercept)
of Individual Trajectory
Rate of Change (Slope)
of Individual Trajectory
Variance Components
p ES r
p ES r
p
p
p
p -2LL
IPC Dimensions
DOM -0.05 0.47 0.05 0.00 0.99 0.00 0.15 0.00 1.24 0.00 0.04 0.00 -0.09 0.00 1643
LOV 0.64 0.00 0.46 0.04 0.02 0.15 0.23 0.00 1.37 0.00 0.03 0.00 -0.05 0.04 1859
IPC Octants
PA -0.01 0.89 0.01 0.04 0.05 0.13 0.20 0.00 1.00 0.00 0.03 0.00 -0.08 0.00 1696
BC -0.71 0.00 0.53 -0.08 0.00 0.27 0.26 0.00 1.08 0.00 0.03 0.00 -0.04 0.11 1854
DE -0.32 0.00 0.31 -0.03 0.10 0.10 0.21 0.00 0.86 0.00 0.03 0.00 -0.07 0.00 1685
FG -0.49 0.00 0.41 0.01 0.57 0.04 0.17 0.00 1.07 0.00 0.03 0.00 -0.06 0.01 1677
HI -0.30 0.00 0.26 -0.07 0.00 0.24 0.19 0.00 1.06 0.00 0.03 0.00 -0.07 0.00 1705
JK 0.77 0.00 0.53 0.06 0.01 0.18 0.41 0.00 1.24 0.00 0.03 0.00 -0.04 0.20 2088
LM 0.20 0.01 0.18 0.01 0.55 0.04 0.28 0.00 1.01 0.00 0.02 0.00 -0.04 0.06 1831
NO 0.33 0.00 0.26 0.01 0.49 0.04 0.20 0.00 1.39 0.00 0.04 0.00 -0.09 0.00 1832
Structural Summary
AMP 1.37 0.00 0.90 -0.00 0.94 0.00 0.11 0.00 0.35 0.00 0.01 0.00 -0.03 0.00 1145
R2
0.73 0.00 0.96 -0.01 0.01 0.16 0.02 0.00 0.03 0.00 0.002 0.00 -0.001 0.29 -306
Note. N = 250. DOM = Dominance; LOV = Affiliation; PA = Assured-Dominant; BC = Arrogant-Calculating; DE = Cold-Hearted;
FG = Aloof-Introverted; HI = Unassured-Submissive; JK = Unassuming-Ingenuous; LM = Warm-Agreeable; NO = Gregarious-
Extraverted; AMP = Amplitude or profile differentiation; R2 = Prototypicality or goodness-of-fit to cosine curve; = Fixed effect
coefficient for intercept; = Fixed effect coefficient for slope; = Level 1 residual variance; = Random effect for intercept;
= Random effect for slope; = covariance between intercept and slope; -2LL = -2 log likelihood, also known as the deviance, an
index of fit. Tabled values represent the final estimates of the fixed effects with robust standard errors. The fixed effects and variance
component parameters were tested to determine if they differ from zero. ES r, effect size r, .10 = small effect, .24 = medium effect, .37
= large effect (Rosenthal & Rosnow, 1991, p. 446). For all models -2LL statistics are based on 6 estimated parameters. Model
estimation was done using full maximum likelihood with the HLM-6 program. Significant fixed effects values (p < .05) bolded.
35
Table 2.3. Descriptive Statistics for Ipsative and Circular Variables.
Minimum Maximum Range Mean SD
Ipsative Statistics
0.32 32.25 31.92 4.08 3.80
0.17 71.78 71.61 4.93 6.03
0.53 74.53 73.99 5.67 6.70
rq12
-0.39 1.00 1.38 0.80 0.24
rq23
-0.79 0.99 1.79 0.79 0.25
rq13
-0.94 0.99 1.94 0.74 0.29
Structural Summary Variables
θ Time 1
0° 360° 360° 4° 69°a
θ Time 2
0° 359° 359° 357° 63°a
θ Time 3
1° 360° 359° 4° 65°a
Elevation Time 1
-0.66 0.92 1.59 -0.06 0.23
Elevation Time 2
-1.33 0.53 1.86 -0.08 0.24
Elevation Time 3
-0.78 0.64 1.42 -0.08 0.23
Amplitude Time 1
0.05 4.50 4.46 1.37 0.67
Amplitude Time 2
0.06 4.18 4.11 1.37 0.67
Amplitude Time 3
0.07 3.77 3.70 1.37 0.65
R
2 Time 1
0.00 0.99 0.99 0.82
b 0.21
R2 Time 2
0.01 0.99 0.98 0.78
b 0.24
R2 Time 3
0.01 1.00 0.99 0.76
b 0.24
Angular Change Δθ12
0° 163° 163° 17°b 30°
Δθ23
0° 177° 177° 17°b
33°
Δθ13
0° 177° 177° 20°b
32°
Note. N = 250. D2 = Cronbach’s D
2 statistic; rq = q-correlation; θ = Angle in
Degrees; R2 = Goodness-of-fit/Prototypicality of curve.
Numeral subscripts after D, Q, and Δθ statistics indicate the two related time points.
Amplitude is equivalent to vector length. a Values reported are for angular variance, not standard deviation.
b These scores are skewed, and thus the median is provided.
36
Table 2.4. Correlations of circumplex measures with flux.
PA12
PA13
PA23
HI12 HI13 HI23 DE12 DE13 DE23 LM12 LM13 LM23
Amplitude
AMP Time1 -- -- 0.07 -- -- 0.09 -- -- 0.09 -- -- -0.02
AMP Time2 -- 0.10 -- -- 0.16* -- -- 0.05 -- -- 0.09 --
AMP Time3 0.12 -- -- 0.06 -- -- 0.10 -- -- -0.08 -- --
Prototypicality
R2 Time 1 -- -- 0.09 -- -- 0.08 -- -- 0.15* -- -- 0.04
R2 Time 2 -- 0.07 -- -- 0.12 -- -- -0.06 -- -- -0.00 --
R2 Time 3 0.09 -- -- 0.01 -- -- -0.04 -- -- -0.10 -- --
Axes
DOM Time 1 -- -- -0.07 -- -- -0.09 -- -- 0.06 -- -- 0.05
LOV Time 1 -- -- -0.05 -- -- -0.04 -- -- -0.28* -- -- -0.23*
DOM Time 2 -- -0.20* -- -- -0.19* -- -- 0.01 -- -- 0.06 --
LOV Time 2 -- -0.04 -- -- -0.06 -- -- -0.24* -- -- -0.20* --
DOM Time 3 -0.10 -- -- -0.07 -- -- 0.10 -- -- -0.02 -- --
LOV Time 3 -0.01 -- -- 0.01 -- -- -0.18* -- -- -0.04 -- --
Note. AMP = Amplitude; DOM = Dominance Dimension; LOV = Affiliation Dimension; PA = Assured-Dominant Octant; HI =
Unassured-Submissive Octant; DE = Cold-Hearted Octant; LM = Warm-Agreeable Octant. Numeral subscripts after octant initials
denotes absolute difference between listed time points (e.g., PA12 = absolute difference between PA score at time1 and time 2).
*p < .05.
37
Table 2.5. Correlations of structural summary statistics and IPC dimensions with spin, pulse, D2, and q-correlations.
Spin Pulse
D2
rq
Δθ12 Δθ13 Δθ23
Pulse12 Pulse13 Pulse23 D2
12 D2
13 D2
23 rq12 rq13 rq23
Amplitude
AMP Time1 -- -- -0.23* -- -- 0.04 -- -- 0.13* -- -- 0.22*
AMP Time2 -- -0.35* -- -- 0.09 -- -- 0.10 -- -- 0.36* --
AMP Time3 -0.25* -- -- -0.08 -- -- -0.04 -- -- 0.30* -- --
Prototypicality
R2 Time 1 -- -- -0.27* -- -- -0.02 -- -- 0.04 -- -- 0.12
R2 Time 2 -- -0.39* -- -- -0.00 -- -- 0.00 -- -- 0.32* --
R2 Time 3 -0.24* -- -- -0.10 -- -- -0.10 -- -- 0.23* -- --
Axes
DOM Time 1 -- -- 0.09 -- -- 0.05 -- -- 0.04 -- -- -0.06
LOV Time 1 -- -- -0.11 -- -- -0.23* -- -- -0.11 -- -- 0.14*
DOM Time 2 -- 0.04 -- -- 0.06 -- -- -0.03 -- -- 0.01 --
LOV Time 2 -- -0.09 -- -- -0.20* -- -- -0.08 -- -- 0.09 --
DOM Time 3 0.07 -- -- -0.02 -- -- 0.04 -- -- -0.05 -- --
LOV Time 3 -0.10 -- -- -0.04 -- -- -0.04 -- -- 0.16* -- --
Note. Δθ = Change in Angular Location; AMP = Amplitude; DOM =
Dominance; LOV = Affiliation.
Numeral subscripts indicate the two time points being compared.
*p < .001. No asterisk denotes p > .05.
38
Figure 2.1. The Interpersonal Circumplex.
39
Figure 2.2. Example of Structural Summary Parameters of a Cosine Curve.
40
CHAPTER 3
An Empirical Examination of Distributional Assumptions Underlying the Relationship between
Personality Disorder Symptoms and Personality Traits
Personality disorder (PD) researchers and theorists have called for an integration of
normal and abnormal personality functioning within comprehensive dimensional models of
personality (Depue & Lenzenweger, 2001; Millon, 2011; Pincus & Hopwood, in press; Widiger
& Simonsen, 2005), and significant empirical attention has been focused on this issue over the
past two decades (e.g., Samuel & Widiger, 2008; Saulsman & Page, 2004; Wiggins & Pincus,
1989). Based on the results of this work, it has been argued that the PDs may best be represented
as specific trait profiles that capture the key features of the disorders (Lynam & Widiger, 2001;
Widiger, Trull, Costa, Sanderson, & Clarkin, 2002), and that normative and abnormal personality
functioning exists on a continuum, with PDs representing “maladaptive expressions” of basic
traits (Widiger & Trull, 2007). The work in this area has relied primarily on correlations and
basic linear regression to model the relationship between personality traits and symptoms of PD.
However, these analytic tools suffer from certain limitations when the underlying distribution of
the variables is severely non-normal, as is the case with PD symptoms in the population.
Alternative approaches that better account for the actual distribution of PD symptoms may
provide better estimates of the relationship between personality and its disorder, and may offer
new insight in to the nature of that relationship.
Continuity and Discontinuity in Personality and its Pathology
The Diagnostic and Statistical Manual of Mental Disorders, 4th
edition, text revision
(DSM-IV-TR, American Psychiatric Association, 2000) offers a categorical model that treats
41
PDs as “qualitatively distinct clinical syndromes” (p. 689) from normative functioning and each
other. Traditionally PDs have been studied as binary outcomes—present or absent (e.g., Shea et
al., 2002); although this distinction has been criticized as arbitrary (Widiger & Clark, 2000).
Alternatively, researchers have modeled the diagnostic criteria as markers for a continuous
dimension of PDs where each criterion counts as an indicator of an incremental increase in the
presence of the disorder (Kass et al., 1985; Lenzenweger & Willet, 2007), and this approach has
proved more reliable and valid (Morey et al., 2007). Therefore, the arbitrary categorical
distinctions made by the official nomenclature lack robust scientific support and dimensionally
defined disorders perform better by empirical standards. However, measuring disorders
dimensionally does not directly speak to their continuity with normal functioning. Dimensional
approaches can make varied distributional assumptions that may have relevance for advancing
understanding of the relationship between normal and abnormal personality.
Meta-analytic studies (see Samuel & Widiger, 2008 and Saulsman & Page, 2004)
summarize results of a large body of research demonstrating that basic personality traits exhibit
significant and replicable relationships to PD. The results of these studies, along with the
evaluation of expert clinicians and researchers, suggest that PD might be well defined using
basic five-factor model trait profiles (e.g., Lynam & Widiger, 2001; Widiger et al., 2002).
Nonetheless, it is not clear that basic traits can fully account for PD and its dysfunction (Morey
et al., 2007). As Samuel and Widiger (2008) point out, the meta-analytic association between
traits and the PDs are generally only modest in size. Furthermore, entering all of the five-factor
traits or even the more fine-grained facets simultaneously as predictors of PD symptoms in
regression models generally only explains a minority of their variance (Bagby, Costa, Widiger,
Ryder, & Marshall, 2005; Reynolds & Clark, 2001). PDs, when treated dimensionally as
42
symptom counts explain more of the variance in important functioning variables (Skodol et al.,
2005). Thus, normal personality traits and PD are not interchangeable representations of
functioning (Krueger et al., 2011), despite what might appear to be easily recognizable shared
content (Widiger & Trull, 2007).
Although personality and PDs are best characterized as dimensional, one possible
explanation for the incongruence is that they may not be entirely continuous. Indeed, DSM-IV-
TR offers a general definition of PD that is intended to help distinguish PDs qualitatively from
normative personality functioning. The DSM-5 workgroup on personality and PD’s proposal
(Skodol et al., 2011) places more emphasis on a general definition, suggesting that the manual
adopt a clear two-step diagnostic process of first determining presence of personality disorder,
followed by clarifying the stylistic manifestation. This fits well with theoretical articulations
(e.g., Kernberg, 1984; Livesley & Jang, 2005; Parker et al., 2004; Pincus, 2005). In this vein, a
number of researchers (e.g., Hopwood et al., in press; Morey et al., 2002; Saulsman & Page,
2004) have found that PD is primarily characterized by higher Neuroticism, lower
Conscientiousness, and lower Agreeableness, with relatively little in the way of distinction
beyond that core profile. Thus, it may be that DSM-IV PDs are not simply the extreme tails of
normative distributions of traits but rather that there is a particular combination of traits that give
rise to or augur for personality pathology more generally. Taken together, these results leave an
unclear picture of how PD and personality traits are related to each other. On the one hand, it
may be that there is a continuous relationship between the two. On the other hand, it may be the
presence of PD pathology relative to none at all is what drives the association, consistent with a
general PD trait profile. Alternatively, that boundary is important, but severity of PD beyond its
presence drives more nuanced relationships. What is clear is that the key theoretical goal of
43
conclusively integrating normative personality traits and PD remains elusive.
Abnormal Personality, Non-Normal Distributions, and Alternative Models
The key theoretical questions of how personality and PD relate to one another are also
inherently questions of methodology. The vast majority of research examining the relationship
between personality traits and PD relies on standard correlation and ordinary least squares (OLS)
regression. These approaches are robust analytic techniques, but nevertheless make a number of
important assumptions (i.e., normality of residuals, homoscedasticity, linearity of relationship,
independence), that, when violated can lead to two forms of bias in estimation (Cohen, Cohen,
West, & Aiken, 2003). In the less serious form, the standard errors, and by extension the
significance test for parameters may be incorrect, although the estimate for the effect is correct.
However, the more serious violation occurs when the actual effect of a relationship is
misestimated. Moving away from reliance on null hypothesis testing might protect somewhat
against the first of these errors, but not the second. A major contributing source to the violation
of these assumptions is the distribution of the variables being modeled.
In the population the actual distribution of psychiatric symptoms is highly positively
skewed with a large number of individuals suffering from no symptoms. PD is no exception.
Figure 3.1 provides an example of such a distribution using the narcissistic personality disorder
(NPD) features in the first wave of the Longitudinal Study of Personality Disorders (LSPD; see
Lenzenweger, 2006), the dataset used for the analyses I report on here. A normal curve has also
been plotted based on the NPD feature distribution’s mean and standard deviation. It is readily
apparent that the observed distribution is non-normal, leptokurtic (i.e., highly kurtotic), with a
strong positive skew. Two additional features of this distribution are worth noting. First, there
are no values below zero, and all are positive integers (i.e., whole numbers), as is the case with
44
all psychiatric and physical symptoms (i.e., one cannot have a negative number of symptoms).
This distribution is characteristic of a count distribution. It is also readily apparent that the
normal curve fit to these data would predict negative values, which is not possible given the
nature of these data. A second feature of note is the large number of zeros in the dataset. This
abundance of zeros has important implications for modeling the relationship between the
symptoms and other variables of interest. Each of these features of the data is discussed in turn.
A number of distributions can describe count data more accurately than the standard
normal distribution presumed in OLS regressions and Pearson correlations. The most basic
approach to modeling counts is with a Poisson distribution, but this also has associated
limitations. Namely, a Poisson presumes that the variance of the distribution is equal to the
mean, a highly constrained assumption in practice (Atkins & Gallop, 2007; Coxe, West, &
Aiken, 2009). Fortunately, other alternatives are better suited to capture counts in real-world
data. One option is the negative-binomial (NB) distribution, which accounts for a larger variance
by estimating an additional parameter for the “over-dispersion” beyond what is anticipated by
the Poisson. Figure 3.2 plots a histogram for the same individual NPD feature counts, but now
both Poisson and NB probability distributions have been fit to the data. Being much better suited
for the symptom counts, neither distribution predicts negative values. Additionally, although the
Poisson fails to account for the large number of zero’s in the data, the NB is better able to
account for the observed values with the flexibility of the second dispersion parameter.
Although the NB is conveniently flexible, a large zero mass may be better addressed using other
approaches.
For such data, as is commonly the case with psychiatric phenomena, models can be
estimated that specifically account for this “inflation” of zeros (see Atkins & Gallop, 2007 for an
45
introduction to these issues). A number of data transformations can attenuate extreme skewness
(e.g., square root, log transformation). However, no transformation will disperse or mitigate a
large number of zero’s in the data (e.g., √0 = 0), and other difficulties associated with common
transformations may arise that often result in serious misspecifications which can lead to
erroneous conclusions (e.g., wrong magnitude or even wrong sign of coefficients; Cohen et al.,
2003; Coxe et al., 2009; King, 1988). Researchers often feel compelled to discard, or leave
unanalyzed those individuals with zero criteria, with the assumption that they offer little
information about the substantive questions of interest. On the contrary, these zeros contain
important information (i.e., which participants are asymptomatic) and this characteristic of
population based data has the potential to mark the “boundary” between normality and
pathology. The jump from zero to one criterion met is potentially a significant threshold and
potentially qualitatively different from each criterion increase thereafter.
Alternative models have been specifically developed to deal with “zero-inflated” data.2
Zero-inflated Poisson (ZIP) and Negative Binomial Hurdle (NBH) models3 are forms of mixture
models that combine two distributions to account for the patterning of the data. Each models the
“zero, not-zero” portion of the distribution (i.e., the difference between having no symptoms vs.
any symptoms) with logistic regression (binary outcome) and the count portion of the
distribution (i.e., degree or severity of PD, given its presence) with Poisson or NB regression.
ZIP and NBH models differ slightly in their treatment of zeros as well. A ZIP assumes some
individuals will have a zero count in the distribution but estimates a class for the excess of zeros,
whereas the NBH treats all zeros as distinct from one-criterion met and beyond. Figure 3.3
2 Terms such as “over-dispersion” and “zero-inflated” imply violations of highly constrained patterns constructed by
statisticians, not violations of the natural state of affairs (P. T. Costa Jr., personal communication, September 22,
2010). 3 Zero-inflated NB and Poisson Hurdle models also exist and are estimable. The focus on the models presented here
is motivated by modeling preferences, pedagogical purposes, and in part to limit the number of models presented.
46
provides an example of a NBH model. The dark column of zeros is differentiated from the
lighter columns of symptoms, which are modeled using a NB distribution. The ZIP/NBH models
can be thought of as two concurrent regressions with a separate set of regression coefficients for
each part. By providing essentially two results, one that models the threshold of presence and
one that models the count portion, there is a direct examination of threshold between pathology
and non-pathology, and the severity of the pathology. Further, it must be emphasized that the
coefficients in each portion of these models are allowed to differ. Thus, a different profile or
pattern may emerge between each portion of the model. In other words, there may be different
processes and variables that distinguish between individuals who meet zero criteria and those
who meet one or more, as opposed to those that predict how many criteria one has once they
meet any at all. For example, there may be normal personality traits that distinguish between
individuals who meet zero criteria and those who meet one or more, while different traits may be
associated with increasing liability for criteria in a disorder.
The Current Study
The goal of the current study is to explore the relationship of personality traits to PD
symptoms using regression models that are capable of more closely approximating the actual
distribution of symptoms in the population. The sample I draw upon is the LSPD, which has a
distribution of PD that closely matches the distribution found in epidemiological samples
(Lenzenweger, 2008; Lenzenweger et al., 1997; Lenzenweger et al., 2007). Unlike samples
selected based on shared diagnostic status or for high levels of pathology, the distributions
approximate those found in the population at large. The LSPD dataset is ideal for the types of
investigations pursued here because it captures the boundary between those individuals whose
personalities function well and those who evidence dysfunction.
47
I have overall two aims. First, I briefly evaluate whether the underlying distribution of
the dependent variable affects the fit of the estimated models to any appreciable degree.
Towards this aim, I run a series of regression models that predict PD symptoms from personality
trait scores, but vary the distribution of the dependent variable, testing the relative fit of normal
continuous, Poisson, ZIP, NB, and NBH based models. The second and more substantive aim
involves comparing the pattern of significant regression coefficients associated with the trait
dimensions to determine the effect of varying distributional assumptions on the relationship
between these variables.
A number of hypotheses follow from the approaches implemented here. First, I expect
the normal distributions to offer the poorest fit to the data. Among the remaining models, I
anticipate that the Poisson distribution will achieve the next worst fit, due to its inability to
adequately account for the large number of zeros and dispersion of the symptom counts. Based
on the plots in Figure 3.2, it is also unlikely that a ZIP model is most appropriate for the types of
distributions observed here. However, it is difficult to predict whether either the NB or NBH
model will emerge as the clearly preferable model based on fit alone. Nevertheless, as noted
above, the NBH models may ultimately be preferred because they offer the ability to test the
difference in the effect of predictors on the “presence” vs. “severity” of PD symptomatology, an
important question in its own right, and one of the novel analytic approaches offered here.
Additional competing hypotheses are offered based on the NBH analyses. First, it is
possible that for each disorder the same predictors that achieve significance for presence (i.e., the
hurdle step) are the same that significantly predict the severity (i.e., the NB step). This is
consistent with a truly continuous dimensional view of traits and PD, with the expectation being
that the same traits that differentiate those with no symptoms from those with symptoms also
48
predictive of how many symptoms one exhibits. An alternative hypothesis would be that a
different pattern of significant predictors are associated with each portion of the model. Thus,
consistent with a qualitative distinction, the variables that distinguish the presence vs. absence of
symptoms differ from the variables that predict how many symptoms are present once any are
present in an individual. A pattern like this could occur in a number of ways, but two appealing
hypotheses are, 1) a consistent pattern of variables is predictive of the logistic (presence vs.
absence) portion of the model regardless of diagnostic category, e.g., (+) Neuroticism, (-)
Conscientiousness, (-) Agreeableness, or 2) a single variable like Neuroticism consistently
predicts the logistic portion, acting like a “gatekeeper.” Finally, it is possible that there is some
hybrid of these hypotheses with some variables emerging as predictors of both parts of the
model, and some as only significant in one part or the other. No prior study has systematically
tested the competing hypotheses, even though they are directly related to the central issue of how
the field understands the relationship between personality and PD.
Method
Participants
Extensive detail concerning the initial participant selection procedure in the LSPD and
sampling is given elsewhere (Lenzenweger, 2006; Lenzenweger et al., 1997). The 250
participants are balanced on gender (53% Females) and the mean age of the participants at entry
into the study was 18.88 years (SD = 0.51). All participants gave voluntary written informed
consent and received an honorarium of $50.00 at each wave. These data were collected and
analyzed with the full approval of the Institutional Review Boards at Cornell University and the
Pennsylvania State University respectively.
Procedure
49
Participants completed self-report measures of personality and clinical assessments were
conducted by experienced Ph.D. or experienced M.S.W. clinicians. Only the data from the initial
assessments are used in the analyses reported here. The LSPD oversampled for PD by selecting
approximately half of the included individuals based on putative positive PD status as assessed
by a self-report measure of PD symptoms (see Lenzenweger et al., 1997). This method was
employed to ensure an adequate sampling of PD pathology in a non-clinical population. Based
on clinical interviews, 11% of the participants qualified for an Axis II diagnosis of some sort.
The raw rates of diagnosed PDs in the LSPD sample were as follows: paranoid = 1.2%, schizoid
= 1.2%, schizotypal = 1.6%, antisocial = 0.8%, borderline = 1.6%, histrionic = 3.5%, narcissistic
= 3.1%, obsessive-compulsive = 1.6%, passive-aggressive = 0.8%, avoidant = 1.2%, dependent =
0.8%, and not otherwise specified = 4.3%. Importantly, these rates closely mirror the rates of PD
found in large epidemiological samples (Lenzenweger, 2008; Lenzenweger et al., 2007).
Measures
International Personality Disorder Examination (IPDE). The IPDE (Loranger, 1988;
1999) was used as the PD measure in this study. The IPDE has excellent psychometric
properties, and it has been shown to be robust as a diagnostic assessment tool even in the face of
mental state (anxiety, depression) changes. The DSM-III-R criteria were assessed in this study
because these were the criteria in effect at the time the LSPD was undertaken. I note the DSM-II-
R and DSM-IV criteria bear considerable resemblance to one another and the fundamental PD
constructs are the same in both nomenclatures. The interrater reliability for IPDE assessments
(based on intraclass correlation coefficients) was excellent, ranging between .84 and .92 for all
PD dimensions. The interviewers (a) were blind to the putative PD group status of the subjects,
(b) were blind to all prior LSPD PD assessment data, and (c) never assessed the same subject
50
more than once. The PD dimensional scores were used for this study. For each symptom, an
individual may receive a score of 0 (Absent or Normal), 1 (Exaggerated or accentuated), 2
(Criterion or Pathological). These values are summed within each disorder to create a “count” of
disorder related features.
Revised Interpersonal Adjective Scales – Big Five (IASR-B5). The IASR-B5 (Trapnell &
Wiggins, 1990) is an extended version of the IAS-R (Wiggins, Trapnell, & Phillips, 1988). The
64-item IAS-R consists of eight scales assessing the eight octants of the IPC, which in turn can
be converted into scores for the two primary dimensions of the IPC: Dominance and Affiliation
using standard scale weights. In addition to the IPC scales and dimensions, the IASR-B5
contains 20-item markers for each of the three dimensions of Conscientiousness, Neuroticism,
and Openness. These three scales of the IASR-B5 correlate highly with the corresponding scales
on the NEO Personality Inventory (r’s = .76, .74, and .67, respectively; Trapnell & Wiggins,
1990) and contain similar levels of affective and behavioral content (Pytlik Zillig, Hemenover,
Diemstbier, 2002). Participants responded to each of 124 adjectives (e.g., dominant, coldhearted,
anxious, organized) on an 8-point scale. Coefficient alphas ranged from .82 to .96. In this study I
use the scores for Dominance, Affiliation, Conscientiousness, and Neuroticism (i.e., the
consensus big four; Widiger & Simonsen, 2005).
Results
A series of regression models were estimated in Mplus 6.1 (Muthén & Muthén, 2010). In
each model, the count of PD features for a given diagnoses was regressed on the four IASR-B5
dimensions in sequence. In other words, each PD’s count and the total PD count of symptoms
were regressed on Dominance, Affiliation, Conscientiousness, and Neuroticism scores
separately. A model was estimated for each personality trait dimension separately in keeping
51
with past literature, and because the dimensions are assumed to be orthogonal in theory, but in
practice often exhibit relationships that attenuate regression coefficients when entered
simultaneously in a model. For each pair of variables, a set of models were estimated with a
different specified distribution for the dependent variable (i.e., PD counts). Normal continuous
(i.e., OLS regression)4, Poisson, ZIP, NB, and NBH models were run in sequence. Relative fit
for each model was assessed using the Akaike Information Criterion (AIC) and the Bayesian
Information Criterion (BIC). The AIC and BIC penalize models for lack of parsimony, and
therefore allow for comparison of model fit across non-nested models.
Table 3.1 reports the AIC and BIC for each model. Table 3.2 reports the regression
coefficients for each model transformed in to effect sizes and their significance. OLS model
coefficients were standardized, coefficients for the count potions of the models were
exponentiated and now represent rate ratios, and the logistic regression coefficients (e.g., binary
“present vs. absent”) were exponentiated and are now represented as odds ratios. It is worth
noting that no formula exists for transforming all of these to the same type of effect size for
direct comparison across models. As a result, what remains most informative is the sign and
significance level of each coefficient.5
Model Fit
For each set of variables, as expected, the OLS models, based on a normal distribution of
PD symptoms, fared the worst in terms of fit. In each case the next worst relative model fits were
4 Although maximum likelihood estimation is used for these models, with a continuous outcome the estimates are
identical to OLS regression. 5 The signs for each of the logistic regression coefficients were reversed prior to exponentiating them because Mplus
6.1 predicts the “zero-class” in these models as opposed to predicting the class with the presence of symptoms. Had
we not changed these, the coefficients of the mixture models would be predicting a) the absence of any symptoms,
and b) the severity, as opposed to a) the presence, and b) the severity. Conceptually that would make the results less
accessible for the purposes here, and the inverse of a logistic coefficient retains the same association, but in the opposite direction, allowing for a direct transformation.
52
associated with the Poisson distributions, followed by the ZIP distributions. This is not
surprising, given the example distributions presented in Figures 3.1-3.3. Recall that the
distribution of NPD symptoms presented in these plots is highly representative of each of the
other PDs in this data set and psychiatric symptoms in general. Although the Poisson based
models fit more poorly than those models based on a NB distribution, they evidenced much
better fit than the OLS models. The NB distribution had equivalent fit to the NBH when
considering the AIC (Mean difference = .03). In terms of the BIC, the two distributions provided
comparable fit, although the NB models performed slightly better (Mean difference = 7.07). This
is to be expected as the BIC assesses a steeper penalty for lack of parsimony, and the NBH
models involve the estimation of 2 more parameters (5 vs. 3). Ultimately, however, relative
model fit is only one of the criteria I considered in the selection of the preferred model, and there
is a conceptual reason to favor the “two-step” NBH model. Indeed, NBH allows for the
exploration of potential qualitative differences in the association of personality with presence of
pathology versus the severity of pathology.
Substantive Comparison of Models
Table 3.2 catalogues the coefficients for each model. Odds and rate ratios of 1.0 indicate
no effect, and those that are below 1.0 are indicative of a negative effect between the predictor
and the outcome. Recall that an odds ratio in logistic regression with a continuous predictor is
the change in the odds of some outcome occurring relative to the other (in this case having at
least one PD symptom) per unit increase in the predictor. The rate ratio is the factor by which the
predicted counts increase per one unit rise in the predictor. Here the trait predictors are
standardized on the original IASR-B5 sample. Trait Dominance and Affiliation can be
conceptualized as rotational variants of Extraversion and Agreeableness (McCrae & Costa, 1989;
53
Pincus, 2002), although the difference in rotation leads to a somewhat different pattern of
correlations.
Given the considerable number of results with each PD modeled five ways, I highlight
notable results here in the text, and refer readers to Table 3.2 for a more detailed account. First,
taking a broad view, similar patterns of coefficients emerge across all models. For example,
radical differences, such as a change in sign, do not occur. However, upon close inspection the
models differ in interesting ways that offer insight in to how personality and different types of
PD may function.
Paranoid: Across all models, paranoid symptoms were associated with lower Affiliation
and higher Neuroticism. The effect for Conscientiousness also differed. In the OLS model the
effect is the same as the meta-analytic result: small and non-significant. Yet in the NBH model
low Conscientiousness emerges as a strong significant predictor of symptom presence, but is not
predictive of severity.
Schizoid: Across models, there was a consistent significant association with lower
Dominance and Affiliation. The NBH model would suggest that low Dominance primarily
differentiates those who have symptoms from those who do not, whereas low Affiliation is
associated with the presence and severity of symptoms.
Schizotypal: Low Dominance, low Affiliation, and high Neuroticism were associated
with schizotypal symptom severity across models, and with both presence and severity in the
NBH model.
Antisocial: Across models, low Affiliation and higher Neuroticism were associated with
antisocial features. However, the NBH model suggests that low Affiliation is predictive of
antisocial feature presence, but Neuroticism predicts the severity of the dysfunction beyond that
54
initial symptom
Borderline: A pattern of low Affiliation, low Conscientiousness, and high Neuroticism
was characteristic of all models. This pattern also predicted the presence of any symptoms in the
NBH model, whereas only low Affiliation and Neuroticism predict severity beyond this.
Histrionic: Higher Dominance and lower Neuroticism were associated with histrionic
features across models, but the modest effect observed for lower Conscientiousness in the OLS
model disappeared in the NBH model. Additionally, Dominance is only predictive of the first
symptom, but Neuroticism is predictive of severity along the continuum of features.
Narcissistic: Across models, the results change substantially. Low Conscientiousness,
which was non-significant in the OLS model, emerges as a significant predictor of presence
along with low Affiliation and higher Neuroticism. But, severity is predicted entirely by the
interpersonal traits.
Avoidant: Significant effects for lower Dominance and Affiliation and higher
Neuroticism were consistent across models. The modest effect for low Conscientiousness in the
OLS models did not emerge in the NBH model. Additionally, although low Affiliation and high
Neuroticism are associated with the presence of any symptom, low Dominance was predictive of
both presence and severity.
Dependent: Lower Dominance and Conscientiousness, along with higher Neuroticism
were consistent predictors across models, but only predictive of presence in the NBH models.
Higher Neuroticism was the sole predictor of severity beyond the first symptom.
Obsessive-Compulsive: In the OLS model, lower Affiliation and higher Neuroticism were
predictive of obsessive-compulsive features. However, in the NBH model although each
predicted both presence and severity, lower Dominance was also a significant predictor of
55
symptoms presence.
Total PD: Pooling all symptoms together in to one count was associated with lower
Affiliation, lower Conscientiousness, and higher Neuroticism. And although each predicted
severity, the sole significant trait that predicted presence of Total PD was Neuroticism.
Discussion
The current study was designed to address a limitation in much of the prior work that has
linked personality traits and PD—specifically, although PD is not a normally distributed
phenomenon in the population, it has consistently been modeled as so. First, goodness-of-fit was
evaluated for traditional models that treated PD symptoms as normally distributed as compared
to models that employed a variety of count distributions. Second, the coefficients relating
personality traits to PD were compared across models. Overall, the study results demonstrate that
treating PD symptomatology as normally distributed results in the poorest estimation as
evidenced by the consistently worst model fit, that count based distributions that can
accommodate “overdispersion” are the most appropriate, and the traits that distinguish those who
have symptomatology from those who do not are not always identical to the traits that are
associated with symptom severity. Taken together, these results suggest that the relationship
between personality and PD symptoms may not be a simple linear dimension (i.e., solely
extreme levels of traits), and the processes that are associated with the presence of PD are not
necessarily those associated with its severity.
Implications for Modeling
Perhaps the most convincing evidence in favor of the modeling approaches adopted here
are the plots of NPD features in Figures 3.1-3.3, which are representative of all the PDs. As can
be seen, the normal curve fit to these data would predict negative values, which is impossible.
56
Post-hoc evaluations of the predicted values resulting from the actual estimated models in Tables
3.1-3.2 confirm that negative values are predicted. This bodes poorly for any type of actuarial
model that might be based on those analyses. Ultimately these violations are reflected in the
AICs and BICs. The fitted Poisson, NB, and NBH distributions in these plots also presaged the
model fit results, with the NB and NBH best able to capture the actual distributions. In the past,
distributions of the types implemented here might have been more challenging, but a number of
readily available statistical packages now include these as standard features. I used Mplus, but
Atkins and Gallop (2007) report that R, SAS, and SPSS are also able to handle some or all of
these approaches respectively.
The symptoms of PD are not normally distributed in the population and most individuals
do not have clinically significant personality impairment. These characteristics have not been
given adequate attention in research linking personality and PD, and depending on the actual
distribution of the sample, it can lead to bias and erroneous conclusions (Atkins & Gallop, 2007;
Cohen et al., 2003; Coxe et al., 2009). As King (1988) notes, these modeling issues are not
esoteric minutiae, but potentially have important substantive implications in applied research.
Among the more serious violations is the prediction of negative values when none exist in the
sample (as is the case here), a misestimation of the magnitude of the effect, and possibly a
misestimation of the direction of the effect. These types of difficulties have the potential to
impede continued advances towards an integrated psychological science that encompasses
normative and abnormal functioning within one framework. As the field attempts to bridge the
gap, it is important that it is done in quantitatively defensible ways.
Implications for the Relationship between Personality and PD
Moving beyond the specifics of model estimation and fit to more substantive issues, this
57
study also investigated whether the traits that differentiate those with pathology from those
without are the same that predict the severity of pathology once present. In keeping with prior
research, I focused on the four dimensions of the Five-Factor traits that are commonly linked
with personality pathology. Other trait approaches are undoubtedly viable and informative (e.g.,
Depue & Lenzenweger, 2001; Tellegen & Waller, 2008), however the traits here have the
greatest similarity to those used in much of the published work in this area (e.g., Samuel &
Widiger, 2008). The two-step mixture models (e.g., ZIP, NBH) offer exciting new possibilities to
evaluate longstanding questions concerning continuity vs. discontinuity in the personality/PD
domain. As such, I tested a series of hypotheses about the pattern of associations across the two
parts of these models. If I had identified the same pattern of traits predicting both presence and
severity, this would be consistent with a continuous dimensional model of normal and abnormal
personality. Different patterns of traits predicting presence versus severity of PD would be
inconsistent with a continuous view, and more suggestive of a qualitative difference. A final
possibility was that a combination more in line with a hybrid model could emerge. As is often
the case, the more nuanced and complex result of a combination of both seems to be most
representative of the pattern across disorders and the Total PD count.
I note that the patterns of results associated with the normally distributed models are
highly consistent with prior work (see e.g., Samuel & Widiger, 2008), suggesting that these
results are representative of the effects estimated in other samples. In terms of the NBH models,
expectedly, Neuroticism was associated with almost every disorder in some way, either the
presence or severity. And in most cases it was associated with both. Across disorders, low
Affiliation was generally associated with both presence and severity, whereas low
Conscientiousness was generally predictive of the presence of specific disorders but offered little
58
prediction of severity beyond that. The failure for Obsessive-Compulsive PD to correlate with
Conscientiousness here is notable, although not surprising, as Samuel and Widiger (2008)
showed this is an inconsistent effect depending on the nature of the measures. Interview
assessments show no association between the two (Interestingly, this is true whether the
interview is of PD or traits). The Total PD count suggests that Neuroticism is the trait that best
differentiates those with any pathology from those without although it also predicts severity
along with low Affiliation and low Conscientiousness. This pattern of associations with severity
was the same as that found by Hopwood and colleagues (in press) in a clinical sample. However,
they did not investigate which variables served to distinguish those who had no-pathology from
any pathology at all. The results for Neuroticism are notable because it has predominantly been
associated with borderline PD, and this is view is furthered in the DSM-5 proposal. These results
suggest that it is the key trait associated with the presence and severity of any PD, not merely
borderline, with other traits characterizing the nuanced variability in phenotypic expression.
Differences in the level of traits across the spectrum of PD has previously been shown to
be in some respects non-linear (O’Connor, 2005), indicating that PD is not merely a sum of trait
features, but may be better thought of as an emergent property at certain levels of traits, or vice
versa. To help clarify this issue, Figure 3.4 provides the scatter plots of NPD features with each
trait. Because the NPD features are counts, the traits values occur in “stripes.” NPD was selected
here in keeping with previous examples, but also because certain traits are predictive of both
presence and severity (Affiliation), presence alone (Conscientiousness and Neuroticism), and
severity alone (Dominance). Notice the far left column, in each case the trait values of those
without any symptom occur across the spectrum, indicating that knowing someone’s trait level
without knowing their pathology is often not diagnostically informative. For example, there are
59
individuals at all levels of Dominance, including high levels, which do not have any narcissistic
pathology. Yet, once there is any narcissistic pathology, a rising trend is observable associated
with increases in Dominance. The opposite is true with Neuroticism, those without narcissistic
pathology are lower on average, but once pathology is present, Neuroticism is not predictive of
severity. These types of more nuanced relationships suggest that these disorders are not reducible
to sums of basic traits (cf. Miller et al. 2005), but are more complex in their structure.
A number of notable areas of convergence emerged out of the NBH analyses that were
not evident when the outcomes were treated as normally distributed— avoidant, obsessive-
compulsive, and schizotypal PDs, and paranoid, borderline, and narcissistic respectively each
were groups having the same pattern of significant traits predicting the presence of any
symptom, but each could be differentiated by the traits associated with severity beyond that
initial symptom. This offers a distinct view on issues of co-occurrence and shared features across
the DSM defined disorders. If traits are taken as “underlying” or to have primacy in these
relationships, one could hypothesize that these groups share similar etiological risk diatheses, but
their phenotypic manifestation then is determined or varies as a function of other traits.
Nevertheless, I caution readers from drawing any strong conclusions in this regard, as regression
based analyses like these cannot establish primacy of this type.
These results highlight that those processes that confer risk for pathology are not always
the same as those that are associated with severity and maintenance of pathology. Additionally,
these results serve to generate hypotheses about shared temperamental etiologies, along with
targets for intervention. For example, it may be that what is most advisable clinically given a
severe PD is to address aspects of their personality that are associated with the severity first, with
the hopes that some of the most severe and damaging behaviors can be avoided in the short term,
60
even though lasting and transformative change may be more gradual.
Limitations
Several caveats must also be considered with these data. First, the present sample was
more homogenous in age, educational achievement, and social class than the U.S. population at
large. Thus, it may be that the LSPD is representative of the population in terms of PD
distributions, but is less representative in other respects. Second, given that the LSPD subjects
were selected from a population of first-year university students, the sample may have been
somewhat censored for individuals affected by some of the most severe PDs. However, one must
be cautious in ascribing undue levels of mental health to subjects who happen to be selected for
academic achievement, as such selection does not confer immunity to psychopathology. To this
end, I note that 16% (or 1 in every 6) of the LSPD sample subjects was diagnosed with a formal
Axis II disorder by the end of the study period using the highly conservative IPDE. Many other
subjects met intermediate levels of PD criteria (e.g., 2 or 3 criteria) that fell short of DSM
diagnostic threshold but indicated some degree PD disturbance of clinical intensity nonetheless
according to the IPDE. I also note that 45.2% of the LSPD subjects had a lifetime (or current)
Axis I disorder by the end of college, and these data are broadly consistent with the distribution
of Axis I disorders in the U.S. population (see Kessler, Chiu, Demler, & Walters, 2005).
Conclusion
The findings here offer a new perspective on the relationship between personality and its
pathology. First, the distributions of symptoms are non-normal, and to most appropriately model
their effect requires the use of more sophisticated approaches than standard Person correlations
or OLS regression. When these approaches are adopted, a more nuanced and complex picture
emerges suggesting that PD is not merely the tail end of a distribution of normal traits, and the
61
processes that are associated with the presence of pathology are not always those that are
associated with increasing severity. Although I do not argue that these results are definitive, I
suggest that these analytic approaches are more appropriate, will lead to more trustworthy
results, and have the potential to elucidate some of the issues associated with continuity and
discontinuity in personality and its pathology, and inform quantitative and qualitative distinctions
in this area (Wright, in press).
62
Table 3.1. Summary of Akaike and Bayesian Information Criteria for Estimated Models
Normal
Poisson
Zero-Inflated
Poisson
Negative-
Binomial
NB Hurdle
Disorder AIC BIC AIC BIC AIC BIC AIC BIC AIC BIC
Paranoid
DOM 901.66 912.23
677.15 684.19
576.51 590.60
533.16 543.72
534.42 552.03
LOV 886.44 897.00
635.51 642.55
555.62 569.70
515.00 525.56
516.04 533.65
CONS 898.08 908.65
666.14 673.19
569.11 583.19
529.50 540.06
526.61 544.21
NEUR 867.71 878.27
583.41 590.46
518.18 532.27
493.98 504.55
498.52 516.13
Schizoid
DOM 832.50 843.07
550.42 557.47
465.34 479.43
456.27 466.83
444.73 462.34
LOV 809.79 820.35
506.55 513.59
445.69 459.77
440.57 451.14
440.99 458.60
CONS 846.85 857.41
592.38 599.42
487.88 501.97
466.64 477.21
470.38 487.98
NEUR 848.86 859.42
598.58 605.62
490.21 504.30
468.97 479.54
470.87 488.48
Schizotypal
DOM 1035.10 1045.66
867.97 875.01
771.53 785.62
720.45 731.01
722.86 740.46
LOV 1035.23 1045.79
874.57 881.61
771.96 786.05
715.07 725.64
715.27 732.88
CONS 1058.94 1069.50
940.92 947.96
804.64 818.73
737.05 747.61
739.91 757.51
NEUR 1045.70 1056.26
899.85 906.89
785.02 799.10
725.39 735.95
728.47 746.08
Antisocial
DOM 1155.26 1165.82
984.54 991.58
742.84 756.93
650.66 661.23
654.42 672.03
LOV 1143.01 1153.58
930.70 937.74
720.59 734.68
636.55 647.11
638.17 655.77
CONS 1156.78 1167.35
993.84 1000.88
746.68 760.77
651.44 662.00
655.42 673.03
NEUR 1154.67 1165.24
982.71 989.76
734.38 748.46
647.93 658.50
651.60 669.20
Borderline
DOM 1067.89 1078.46
948.12 955.16
738.18 752.27
680.66 691.23
684.57 702.18
LOV 1049.73 1060.30
889.87 896.91
709.93 724.02
667.19 677.76
669.62 687.23
CONS 1062.01 1072.57
927.19 934.24
727.55 741.64
675.72 686.29
673.75 691.36
NEUR 1032.29 1042.86
832.21 839.25
677.68 691.77
646.31 656.88
648.64 666.25
Histrionic
DOM 1087.97 1098.54
989.40 996.45
775.38 789.46
743.87 754.44
741.06 758.66
LOV 1096.89 1107.45
1021.76 1028.81
787.01 801.09
750.61 761.17
750.61 768.22
CONS 1090.91 1101.48
1001.43 1008.47
779.25 793.33
746.96 757.53
745.72 763.32
NEUR 1072.13 1082.70
942.31 949.35
756.46 770.55
731.71 742.27
725.81 743.42
Narcissistic
63
DOM 1163.73 1174.29
1081.02 1088.06
820.98 835.07
758.72 769.29
759.20 776.81
LOV 1152.28 1162.84
1045.52 1052.56
807.88 821.97
749.24 759.80
748.62 766.22
CONS 1172.11 1182.68
1119.64 1126.68
836.87 850.95
764.37 774.93
764.05 781.65
NEUR 1148.25 1158.82
1024.83 1031.87
801.12 815.21
745.28 755.85
731.75 749.36
Avoidant
DOM 898.18 908.75
659.29 666.33
597.91 611.99
590.53 601.10
588.31 605.91
LOV 923.68 934.25
724.22 731.26
625.01 639.10
604.02 614.58
604.10 621.71
CONS 936.87 947.44
754.84 761.89
639.27 653.36
617.42 627.99
620.45 638.06
NEUR 905.70 916.26
678.56 685.60
599.29 613.38
591.52 602.09
580.82 598.43
Dependant
DOM 934.02 944.58
740.09 747.13
654.04 668.13
611.63 622.19
612.16 629.76
LOV 938.50 949.06
752.95 759.99
660.45 674.53
615.68 626.24
617.42 635.03
CONS 932.47 943.03
736.09 743.13
651.86 665.94
609.87 620.43
609.67 627.28
NEUR 889.08 899.65
631.22 638.26
591.70 605.78
563.91 574.48
560.53 578.14
Obs.-Comp
DOM 1035.10 1045.67
807.31 817.88
804.63 818.72
772.45 783.01
770.64 788.24
LOV 1004.85 1015.42
775.64 786.20
771.80 785.88
747.57 758.14
748.85 766.45
CONS 1034.53 1045.09
806.68 817.25
808.40 822.49
771.89 782.46
775.01 792.62
NEUR 1014.50 1025.06
792.78 803.34
782.52 796.61
755.04 765.60
753.09 770.70
Total PD
DOM 1962.46 1973.02
3849.63 3856.68
3240.01 3254.10
1669.64 1680.20
1673.13 1690.73
LOV 1929.41 1939.98
3445.55 3452.59
2906.81 2920.90
1643.63 1654.20
1646.34 1663.94
CONS 1954.96 1965.53
3746.20 3753.24
3130.53 3144.61
1663.19 1673.76
1665.44 1683.05
NEUR 1910.19 1920.75
3207.19 3214.23
2797.59 2811.67
1624.55 1635.11
1625.19 1642.80
Note. N = 250. AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; NB = Negative-Binomial;
NB = Negative-Binomial. All models estimated in Mplus 6.1 (Muthén & Muthén, 2010). Normal Continuous, Poisson,
and Negative-Binomial models each had 3 free parameters, whereas Zero-Inflated Poisson and Negative-Binomial
Hurdle Models each had 5 free parameters.
64
Table 3.2. Summary of coefficients from models regressing personality disorder symptoms on personality traits.
Normal
Poisson
Zero-Inflated Poisson
Negative-
Binomial
Negative-Binomial Hurdle
Disorder β p
RR p
OR(i) p(i) RR(P) p(P)
RR p
OR(h) p(h) RR(NB) p(NB)
Paranoid
DOM -0.021 0.741
0.954 0.799
0.847 0.327 1.054 0.770
0.965 0.801
0.882 0.389 1.058 0.758
LOV -0.244 0.000
0.642 0.000
0.631 0.037 0.782 0.018
0.544 0.000
0.572 0.001 0.591 0.021
CONS -0.121 0.054
0.776 0.034
0.633 0.013 1.014 0.910
0.773 0.051
0.670 0.004 1.029 0.895
NEUR 0.357 0.000
2.020 0.000
1.324 0.353 1.984 0.001
2.296 0.000
2.044 0.000 2.275 0.002
Schizoid
DOM -0.252 0.000
0.563 0.002
0.428 0.000 0.933 0.726
0.628 0.024
0.437 0.000 0.949 0.807
LOV -0.380 0.000
0.516 0.000
0.696 0.051 0.641 0.000
0.502 0.000
0.560 0.000 0.579 0.001
CONS -0.090 0.155
0.811 0.117
1.015 0.933 0.764 0.129
0.780 0.110
0.885 0.358 0.731 0.161
NEUR 0.003 0.967
1.006 0.969
1.260 0.197 0.875 0.325
1.006 0.969
1.165 0.318 0.836 0.276
Schizotypal
DOM -0.308 0.000
0.613 0.000
0.708 0.015 0.760 0.001
0.680 0.000
0.650 0.002 0.714 0.004
LOV -0.308 0.000
0.651 0.000
0.572 0.001 0.803 0.008
0.600 0.000
0.542 0.000 0.683 0.015
CONS -0.068 0.280
0.894 0.260
0.891 0.447 0.946 0.578
0.890 0.262
0.878 0.314 0.919 0.576
NEUR 0.237 0.000
1.452 0.000
1.318 0.067 1.247 0.006
1.448 0.000
1.418 0.009 1.429 0.003
Antisocial
DOM 0.120 0.055
1.334 0.134
1.168 0.287 1.188 0.308
1.259 0.128
1.215 0.159 1.213 0.308
LOV -0.248 0.000
0.631 0.000
0.614 0.004 0.770 0.000
0.562 0.000
0.579 0.001 0.696 0.004
CONS -0.092 0.144
0.817 0.072
0.890 0.413 0.882 0.227
0.791 0.091
0.865 0.276 0.796 0.250
NEUR 0.130 0.039
1.326 0.006
1.069 0.648 1.359 0.008
1.480 0.007
1.168 0.222 1.600 0.016
Borderline
DOM 0.052 0.415
1.103 0.410
1.080 0.593 1.046 0.621
1.095 0.410
1.091 0.521 1.070 0.625
LOV -0.269 0.000
0.658 0.000
0.702 0.023 0.754 0.008
0.638 0.001
0.648 0.003 0.727 0.019
CONS -0.161 0.010
0.754 0.006
0.629 0.002 0.971 0.762
0.741 0.011
0.638 0.001 0.961 0.765
NEUR 0.367 0.000
1.853 0.000
1.435 0.037 1.590 0.000
2.143 0.000
1.692 0.000 2.034 0.000
Histrionic
65
DOM 0.189 0.002
1.392 0.001
1.425 0.014 1.131 0.098
1.350 0.005
1.456 0.007 1.157 0.114
LOV 0.027 0.668
1.047 0.741
1.006 0.962 1.041 0.729
1.038 0.744
1.017 0.894 1.041 0.734
CONS -0.156 0.013
0.781 0.027
0.815 0.130 0.883 0.126
0.815 0.031
0.795 0.083 0.875 0.130
NEUR 0.308 0.000
1.605 0.000
1.744 0.000 1.214 0.008
1.644 0.000
1.808 0.000 1.292 0.016
Narcissistic
DOM 0.201 0.001
1.464 0.004
1.214 0.174 1.237 0.024
1.343 0.012
1.266 0.088 1.294 0.042
LOV -0.289 0.000
0.651 0.000
0.692 0.007 0.799 0.002
0.616 0.000
0.666 0.002 0.699 0.000
CONS -0.088 0.161
0.857 0.132
0.772 0.053 0.990 0.897
0.866 0.152
0.775 0.049 0.988 0.911
NEUR 0.313 0.000
1.672 0.000
2.250 0.000 1.135 0.133
1.706 0.000
2.273 0.000 1.194 0.138
Avoidant
DOM -0.409 0.000
0.499 0.000
0.572 0.001 0.694 0.000
0.552 0.000
0.499 0.000 0.659 0.000
LOV -0.279 0.000
0.653 0.000
0.559 0.000 0.857 0.010
0.583 0.000
0.543 0.000 0.759 0.056
CONS -0.167 0.007
0.748 0.020
0.859 0.319 0.822 0.102
0.763 0.023
0.795 0.084 0.793 0.102
NEUR 0.377 0.000
1.866 0.000
2.524 0.000 1.250 0.077
1.870 0.000
2.596 0.000 1.276 0.058
Dependent
DOM -0.138 0.027
0.775 0.035
0.758 0.102 0.898 0.402
0.791 0.044
0.736 0.026 0.881 0.416
LOV 0.037 0.558
1.076 0.556
1.111 0.584 1.015 0.932
1.081 0.540
1.105 0.431 1.023 0.916
CONS -0.159 0.011
0.754 0.007
0.705 0.035 0.907 0.361
0.752 0.013
0.693 0.007 0.857 0.415
NEUR 0.425 0.000
2.044 0.000
2.425 0.000 1.498 0.004
2.221 0.000
2.818 0.000 1.802 0.001
Obsessive-Compulsive
DOM -0.044 0.483
0.936 0.562
0.742 0.022 1.051 0.494
0.953 0.570
0.757 0.032 1.078 0.402
LOV -0.340 0.000
0.654 0.000
0.690 0.016 0.763 0.000
0.632 0.000
0.629 0.001 0.686 0.000
CONS -0.065 0.300
0.908 0.331
0.929 0.598 0.945 0.498
0.916 0.331
0.911 0.467 0.932 0.511
NEUR 0.284 0.000
1.489 0.000
1.721 0.001 1.196 0.041
1.476 0.000
1.784 0.000 1.239 0.045
Total PD
DOM -0.051 0.423
0.942 0.521
0.857 0.239 0.967 0.676
0.958 0.524
0.857 0.239 0.971 0.676
LOV -0.355 0.000
0.694 0.000
0.798 0.159 0.724 0.000
0.687 0.000
0.797 0.157 0.684 0.000
CONS -0.179 0.004
0.815 0.002
1.055 0.704 0.820 0.001
0.820 0.003
1.055 0.705 0.795 0.001
NEUR 0.437 0.000
1.616 0.000
1.998 0.000 1.478 0.000
1.647 0.000
2.002 0.000 1.571 0.000
Note. Bold = p < .05. DOM = Dominance; LOV = Affiliation; CONS = Conscientiousness; NEUR = Neuroticism; OR = Odds Ratio;
RR = Rate Ratio; P = Poisson; NB = Negative-Binomial; i = inflation class; h = hurdle class; PD = Personality Disorder.
66
Figure 3.1. Normal Distribution Fit to Observed Narcissistic Personality Disorder Features.
67
Figure 3.2. Poisson and Negative-Binomial Distributions Fit to Observed LSPD Narcissistic Personality Disorder Features.
68
Figure 3.3. Representation of Negative-Binomial Hurdle Model for LSPD Narcissistic Personality Disorder Symptoms.
69
Figure 3.4. Scatter plots of personality trait scores and NPD features.
70
CHAPTER 4
A Parallel Process Growth Model of Avoidant Personality Disorder Symptoms
and Personality Traits
Historically, the personality disorders (PD) have been construed as highly stable forms of
psychopathology, and this is emphasized in the definition of this class of disorders (American
Psychiatric Association, 2000). However, empirical results from a number of prospective,
multiwave, longitudinal studies suggest that in actuality, symptoms of PD demonstrate instability
and plasticity over time (Johnson et al., 2000; Lenzenweger et al., 2004; Skodol et al., 2005;
Zanarini et al., 2003). These findings have ushered in a shift in understanding and
conceptualization of PD symptomatology, and beckon for new theoretical models of personality
pathology that can account for this observed change over time. Avoidant personality disorder
(AVPD) is among those disorders that evidence considerable longitudinal change (Grilo et al.,
2004; Lenzenweger, 1999) with significant heterogeneity in intraindividual (i.e., within-person)
symptom trajectories over time (Lenzenweger et al., 2004). Yet, it remains unknown which other
aspects of psychological functioning are associated with this interindividual (i.e., between-
person) variability in AVPD trajectories. Indentifying other psychological systems that change in
tandem with AVPD would provide strong evidence that those systems play an important and
potentially causal role in AVPD and personality pathology more generally. I propose that normal
personality traits may demonstrate just this pattern, and test whether individual trajectories in
AVPD and personality traits are linked dynamically over time.
Prior work investigating the cross-sectional relationship between AVPD and basic
personality traits finds a consistent pattern of results (see Alden et al., 2002 for a review), and
71
two recent meta-analyses have summarized the relationship between AVPD, along with other
PDs, and the five-factor model of personality traits (Samuel & Widiger, 2008; Saulsman & Page,
2004). In short, AVPD shows a strong positive relationship with neuroticism, or the tendency to
experience negative emotions (e.g., anxiety, anger, depression, guilt), and a strong negative
relationship with extraversion, or the tendency to be outgoing, gregarious, and experience
positive emotions. These results converge with earlier work that employed interpersonal
circumplex based trait models, and found that AVPD was related to low trait dominance and
affiliation (Wiggins & Pincus, 1989) and socially-avoidant interpersonal problems (Pincus &
Wiggins, 1990; Soldz et al., 1993). These relationships are consistent with the diagnostic features
of the disorder which center on a keen sensitivity to interpersonal rejection, exquisite fears of
humiliation and judgment, and accordingly, avoidance of social and interpersonal situations,
especially when it involves new people or new situations. However, the wealth of cross-
sectional associations cannot speak to the longitudinal relationship between basic personality
traits and AVPD, about which very little is known.
As noted above, a number of large-scale prospective studies have assessed the
longitudinal stability of PD diagnoses and symptom criteria. In both clinical (Shea et al., 2002;
Skodol et al., 2005; Zanarini et al., 2003) and community (non-clinical) samples (Cohen et al.,
2005; Johnson et al., 2000; Lenzenweger, 1999) results reveal that there are significant declines
in the number of individuals meeting diagnostic threshold, and significant declines in symptoms
when treated as dimensions, across the PDs. Using a patient sample, the remission rate of
categorically diagnosed AVPD was 50% after two years (Grilo et al., 2004). In a non-clinical
sample, the Longitudinal Study of Personality Disorders (LSPD; Lenzenweger, 2006)
demonstrated that on the average, AVPD symptom counts decline modestly but significantly
72
(Lenzenweger et al., 2004). But, more importantly, further analyses that employed an individual
growth curve analytic framework, which estimates a separate trajectory per case in the study,
indicate that the mean decline in the whole sample masks significant interindividual
heterogeneity in symptom change over time (Lenzenweger et al., 2004). In other words, although
on the average individuals decline in the number of criteria they meet, some show stability over
time, and others increase in criteria over time. Therefore, it is important to elucidate the
influential aspects of individuals and their environments that act as risk and protective factors in
augmenting or mitigating an individual’s trajectory of change. This raises the general question,
what other aspects of psychological functioning are related to individual trajectories in AVPD
symptom change? More specifically, are the same personality traits that are associated with
AVPD at a given time point (i.e., cross-sectional links) also associated with the rates of change
over time?
The empirical literature on stability and change in normal range personality traits and the
processes they represent has developed separately from that of PD, but striking similarities have
emerged. Once thought to be entirely stable (James, 1890), it is now understood that individual’s
personality traits are indeed highly stable, but not fixed (Wright et al., 2011c). Rates of mean
change in broad personality traits are modest but significant, and on the average neuroticism
declines, while agreeableness, conscientiousness, and social dominance increase over adulthood
(Roberts et al., 2006). However, as is the case for PD, normative change masks significant
interindividual heterogeneity in intraindividual trajectories around the population’s mean rate of
change (Mroczek & Spiro, 2003; Vaidya et al., 2008; Wright et al., 2011c).
Thus, basic personality science and the psychopathology of PD converge on the finding
that each construct is not as stable as once thought. Importantly, when examined at the level of
73
the individual, rich heterogeneity emerges in the direction and rates of change observed. Given
that this is the case, these fundamental domains of individual functioning have the potential to be
related longitudinally. To date, this question remains mostly unexplored and therefore the
longitudinal relationship between these systems remains unknown. The lone study to examine
AVPD and personality traits over time did find that change in personality traits predicted
subsequent (i.e., at a later time) AVPD change (Warner et al., 2004). Although informative, this
work did not test whether the rates of change in each system corresponded. If the rate of change
in PD and personality could be shown to be significantly related, it would represent an important
advance in the science of personality and its pathology, suggesting that these constructs are
developmentally linked. This would provide much needed evidence in favor of a unified science
of personality and psychopathology, allowing for more confident assertions that normal
personality traits and PD comprise manifestations of the same psychological system (e.g., Depue
& Lenzenweger, 2005; Clark, 2007; Pincus & Hopwood, in press; Widiger & Trull, 2007).
The Current Study
The current investigation is structured to evaluate whether the trajectories of change in
PD and personality are linked longitudinally. I base these analyses on the LSPD sample. The
LSPD is ideally suited to study this question, as it is a naturalistic study, drawing form a non-
clinical sample designed in such a way as to include individuals who were at a putative risk for
PD pathology at the outset of the study, but also individuals who exhibited no significant
pathology, but might develop symptoms over the course of the study (see Lenzenweger, 2006).
These analyses focus on AVPD symptomatology and five commonly assessed personality traits,
dominance, affiliation, conscientiousness, neuroticism, and openness. Additionally, as is detailed
below, I adopt an analytic approach that is designed to measure multivariate change (i.e.,
74
simultaneous change in multiple dimensions or systems)—parallel process growth curve
modeling (see e.g., Bollen & Curran, 2006). The parallel process growth curve framework allows
us to test whether the change that is occurring in each set of variables, PD symptoms and
personality traits, is related to each other. To my knowledge, this is the first study to examine the
conjoint change in personality and AVPD symptoms using parallel process growth curve models.
In addition to answering the basic question of whether PD and personality are related in their
rates of change over time, I will answer whether the same traits that demonstrate significant
cross-sectional relationships are those that are dynamically related to change in AVPD
symptoms.
Method
Participants
The 258 participants in the LSPD were drawn from a population consisting of 2,000 first-
year undergraduate students. Extensive detail concerning the initial participant selection
procedure and sampling is given elsewhere (Lenzenweger, 2006; Lenzenweger et al., 1997). The
258 participants consisted of 121 males (47%) and 137 females (53%). The mean age of the
participants at entry into the study was 18.88 years (SD = 0.51). Participants were assessed at
their first, second, and fourth years of college. All participants gave voluntary written informed
consent and received an honorarium of $50.00 at each wave. These data were collected and
analyzed with the full approval of the Institutional Review Boards at Cornell University and the
Pennsylvania State University respectively. Of the initial 258 participants, 250 completed all
three assessment waves and are included in these analyses. Six left the study, and two died in
automobile accidents.
Procedure
75
Structure of the LSPD Dataset. As noted above, the LSPD has a prospective, multiwave,
longitudinal design with participants evaluated at three points in time (i.e., first, second, and
fourth years in college). At each time point, participants completed self-report measures of
personality and clinical assessments were conducted by experienced Ph.D. or experienced
M.S.W. clinicians. The LSPD oversampled for PD by selecting approximately half of the
included individuals based on putative positive PD status as assessed by a self-report measure of
PD symptoms (see Lenzenweger et al., 1997). This method was employed to ensure an adequate
sampling of PD pathology in a non-clinical population. At Wave 1, 11% of the participants
qualified for an Axis II diagnosis of some sort based on clinical interviews. The raw rates of
diagnosed PDs in the LSPD sample at Wave 1 were as follows: paranoid = 1.2%, schizoid =
1.2%, schizotypal = 1.6%, antisocial = 0.8%, borderline = 1.6%, histrionic = 3.5%, narcissistic =
3.1%, obsessive-compulsive = 1.6%, passive-aggressive = 0.8%, avoidant = 1.2%, dependent =
0.8%, and not otherwise specified = 4.3%.
Measures
International Personality Disorder Examination (IPDE). The IPDE (Loranger, 1988, 1999) was
used as the PD measure in this study. The IPDE has excellent psychometric properties, and it has
been shown to be robust as a diagnostic assessment tool even in the face of mental state (anxiety,
depression) changes. The DSM-III-R criteria were assessed in this study because these were the
criteria in effect at the time the LSPD was undertaken. I note the DSM-III-R and DSM-IV
criteria bear considerable resemblance to one another and the fundamental PD constructs are the
same in both nomenclatures. Clinically experienced interviewers received training in IPDE
administration and scoring by Dr. A. W. Loranger and were supervised throughout the project by
the third author (M.F.L.), who was blind to the participants’ identity, putative PD status, and all
76
prior assessment information. The interrater reliability for IPDE assessments (based on intraclass
correlation coefficients) was excellent, ranging between .84 and .92 for all PD dimensions. The
interviewers (a) were blind to the putative PD group status of the subjects, (b) were blind to all
prior LSPD PD assessment data, and (c) never assessed the same subject more than once. The
IPDE PD dimensional scores were used for this study.
Revised Interpersonal Adjective Scales – Big Five (IASR-B5). The IASR-B5 (Trapnell &
Wiggins, 1990) is an adjective-based, 124-item measure that provides scores for the personality
trait dimensions of dominance, affiliation, conscientiousness, neuroticism, and openness.
Participants responded to each of 124 adjectives (e.g., dominant, coldhearted, anxious,
organized) at each wave of the LSPD. Coefficient alphas for all scales at all waves ranged from
.82 to .96.
Data Analysis
Latent growth curve models (LGM) in a structural equation modeling (SEM) framework
offer a flexible approach to study average rates of growth (i.e., change) and individual
differences in trajectories over time (Bollen & Curran, 2006). In LGM, an individual’s scores at
each time point are modeled as a function of latent growth factors that are estimated from the
observed scores. With three time points, linear rates of change can be estimated with one latent
factor for the intercept, or level of the curve at the measurement time of the intercept, and
another latent growth factor is estimated for the slope, or rate of change. These factors can be
regressed on covariates to control for the effect of other variables (e.g., demographics). The
flexibility of LGMs allows them to be extended to multivariate models, often referred to as
parallel process LGMs. Multivariate LGMs are so called because they chart trajectories of
change or growth processes in two or more variables in parallel. By allowing the growth factors
77
of each parallel set of variables to correlate, it is possible to examine whether the intercept and
growth in one is related to the intercept and growth in the others. This offers a very powerful
analytic approach for the study of stability, change, and development across time.
I estimated parallel process LGMs in Mplus 6.1 (Muthén & Muthén, 2010). Figure 4.1
provides a conceptual diagram of the estimated models with the paths labelled for ease of
communication. The loadings for the waves of measurement on the slope factor were fixed to 0.0
for Wave 1, thereby setting the start of the study as the intercept, followed by the mean
assessment time between waves for the remaining two loadings (i.e., .95 for Wave 2, 2.82 for
Wave 3). In this study the effect of sex and age of entry to the study are included as covariates of
the trajectories of change by regressing the latent growth factors on each. In addition, the study
group is included to account for the study’s sampling strategy. The residual variances of the
growth factors (i.e., remaining variance after controlling for sex and age) were allowed to freely
intercorrelate. Of most interest here is the covariance between the slope factors (i.e., growth
rates), or to what degree the change in personality dimensions and AVPD symptoms are
associated with each other (Path B in Figure 4.1). The additional parameters capture the
relationship between intercepts (Path A), the covariance between initial levels and rates of
change within each system (Paths C and D), and the effect of initial status on personality and
AVPD on the rate of change in the other (Paths E and F). The AVPD symptom counts were
modeled using a Poisson distribution. Symptom counts of psychiatric disorders rarely follow a
normal distribution (which is assumed in standard SEM), and significant violations of this
assumption can lead to problems in estimation and severe misspecification of model parameters
if a more appropriate distribution is not employed (e.g., the prediction of negative values;
Atkinson & Gallop, 2007; King, 1988; Wright et al., 2011a).
78
Results
Five parallel process LGMs were estimated, one for each personality dimension of the
IASR-B5. Individual models were required because attempts to estimate a model with growth in
all five personality dimensions and AVPD symptoms modeled simultaneously led to problems
with specification (i.e., negative residual variances), which is not uncommon in very large
structural models. When growth is modeled using a Poisson distribution, traditional SEM fit
indices (e.g., chi-square, root mean square error of approximation [RMSEA], comparative fit
index [CFI], etc.) are not available. Therefore the log-likelihood and indices of relative fit, the
Akaike (AIC) and Bayesian information criteria (BIC) are provided. There were no apparent
sources of model strain based on a careful evaluation of the resulting parameter estimates.
Finally, Table 4.1 provides the raw parameter coefficients and standard errors. The parameters
associated with the AVPD symptoms are on the logit scale, making direct interpretation difficult.
When exponentiated these values provide the estimated counts, but leave the study parameters on
vastly different scales. Therefore, to aid with interpretability, the estimated effect size on the r
metric is also provided (where r = √((t2)/(t
2 + df)) see Rosenthal & Rosnow, 1991).
Past work with the LSPD sample has shown that on the average, AVPD symptoms
decrease over the course of the study, but there is significant heterogeneity in individual
trajectories (Lenzenweger et al., 2004). Also in this sample Wright and colleagues (2011b) found
that on the average, affiliation, conscientiousness, and openness each increase over the course of
the study waves, whereas dominance shows no mean change, and neuroticism decreases
significantly. However, each personality dimension evidenced significant heterogeneity in
individual trajectories. I have briefly summarized these findings here to provide a context for this
study’s results.
79
The resulting parameter values and fit statistics are presented in Table 4.1. Age, as a
covariate, was never a significant predictor of the intercepts and slopes in either personality or
AVPD. Sex (females = 0, males = 1) was only predictive of lower affiliation and neuroticism
intercepts and conscientiousness slope, but were otherwise not predictive of intercepts or slopes.
Those individuals selected for the study based on a positive putative pd status demonstrated
higher AVPD and neuroticism intercepts, lower affiliation and conscientiousness intercepts, and
more steep declines in neuroticism, but otherwise study group was unassociated with study
parameters. AVPD intercept was significantly associated with lower dominance, affiliation, and
openness, and higher neuroticism intercepts (A Paths). The rate of change in AVPD was
significantly negatively related to the rates of change in dominance and affiliation, and positively
to neuroticism (B Paths). At a less stringent level of significance, AVPD slope was also
negatively related to slope in openness (p < .10). The rate of change in AVPD was never related
to initial status (D Paths), whereas initial personality status was significantly negatively related to
slope for all traits with the exception of Affiliation (C Paths). AVPD intercept was unrelated to
personality trait trajectories (E Paths). Finally, initial status in openness was positively associated
with rate of change in AVPD symptoms (F Paths).
Discussion
In this study, I tested whether the intraindividual rates of change in AVPD symptoms and
five broad personality trait dimensions were related over the course of approximately 4 years.
The findings revealed that individual trajectories in AVPD symptoms were indeed associated
with rates of change in personality symptoms. This is the first study to demonstrate such a
relationship for any PD, and provides an important contribution to the science of personality and
the psychopathology of PD, suggesting personality traits and PD symptoms are developmentally
80
linked.
Specifically, I found that at the outset of the study, an individual’s level of AVPD
symptoms was related to higher neuroticism, but lower dominance, affiliation, and openness,
replicating the results of many prior studies (see Alden et al., 2002; Saulsman & Page, 2004;
Samuel & Widiger, 2008 for reviews). These well known associations, observed again in this
sample, provide confidence in the novel results associated with the growth factor relationships,
the primary target of this study. In terms of the relationship between the growth factors, I found
that the rate of change in AVPD symptoms is associated with the rate of change in dominance,
affiliation, neuroticism, and to a lesser extent openness. Notably, these relationships are in the
same directions as was found in the intercepts. One’s AVPD symptoms decrease as they become
more dominant (i.e., more self-assured, assertive, and less submissive), more affiliative (i.e.,
more cooperative, engaging, less aloof), less neurotic (i.e., less inclined to experience negative
emotions and anxiety), and open (i.e., amenable to new ideas and experiences). The reverse is
also true, such that as these traits decline, AVPD symptoms increase.
An individual’s initial rate of AVPD was not predictive of the rates of change in
personality traits, nor was an individual’s initial personality trait level predictive of AVPD
change, with the exception of openness. Individuals lower in openness at the outset of the study
showed more rapid declines in AVPD symptoms over the course of the study. This relationship is
best understood in the context of the mean trend of decline in AVPD symptoms that has been
reported earlier (Lenzenweger, 1999), and is likely reflective of the law of initial values. This
was the only trait for which this occurred.
By looking longitudinally within a LGM framework, this study adopts a more person-
centered approach in the study of the relationship between personality and its pathology. These
81
longitudinal data capture the very important dynamic interplay between an individual’s
personality and AVPD symptoms. Change in each occurs within an individual across time,
whereas between individuals these rates and patterns of change vary. Past work has examined the
heterogeneity in AVPD and personality growth separately, but this is the first evidence that they
are dynamically linked in the paths they take. Furthermore, the longitudinal relationships
observed here are consistent with hypotheses based on the diagnostic features of AVPD, the
aspects of personality measured by these traits, and prior cross-sectional work.
From a psychopathology perspective, these results offer compelling evidence that normal
personality and personality pathology should be understood as expressions of the same system.
This implication is not minor, as the modern empirical literatures in each are often blind to the
other’s advances. A number of theorists have now called for more unification in the
understanding of basic personality and personality disorder. Theoretical proposals calling for
more unity range in perspective from the neurobiological (Depue & Lenzenweger, 2005),
temperament (Clark, 2007), psychometric (Widiger & Trull, 2007), and interpersonal (Pincus &
Hopwood, in press; Wright, in press). Each of these perspectives, although distinct in some
respects, is also quite similar in others (Widiger & Simonsen, 2005). The findings reported here
bolster this argument, and provide a previously unavailable piece of evidence for this notion that
personality and its disorder belong within a coherent and comprehensive model of normative and
non-normative functioning.
Clinically, cognitive, behavioral (e.g., Alden & Capreol, 1993), and psychodynamic (e.g.,
Barber et al., 1997) psychotherapies have all been shown to effectively reduce AVPD symptoms.
Interestingly, cognitive and behavioral approaches tend to target the symptoms of AVPD,
whereas psychodynamic approaches have targeted interpersonal functioning in the context of an
82
individual’s personality. The current results do not elucidate the direction of causation (i.e., is
personality change driving symptom change or vice versa), but taken together with the results of
existing treatment studies support the notion that the relationship is bidirectional between
personality and PD, and both are valid treatment targets. Regardless, one potential implication of
these specific results is that generally increasing dominance (e.g., through assertiveness training)
and affiliation in the context of emotional regulation may catalyze symptomatic change in AVPD
while also generally increasing the level of an individual’s functioning.
Limitations
Several caveats must also be considered with these data and analyses. First, despite the
impressive ability of these models to capture association between changes over time in both
systems, they do not determine causality. It remains to be determined whether personality
changes drive symptom changes, or vice versa. Or potentially variables external to the
personality system as a whole are driving this change. Undoubtedly, contained within the days,
weeks, and months that make up the years are innumerable interactions and life experiences that
serve to cumulatively push and pull an individual’s trajectory one way or another. Additionally,
the results of this study are at too coarse of a level of analysis to speak directly to person-
environment transaction theories (Caspi & Roberts, 2001).
In terms of the data, the present sample was clearly more homogenous in age, educational
achievement, and social class than the U.S. population at large. Perhaps the most effective way to
assess the generalizability of findings from the LSPD is to evaluate whether prior core LSPD
findings have been replicated, and they have. For example, the LSPD-based estimate for PD
prevalence in the community (11%; Lenzenweger et al., 1997) has now been broadly replicated
several times in U.S. nonclinical community samples (Crawford et al., 2005; Samuels et al.,
83
2002) and the U.S. general population (Lenzenweger et al., 2007). Furthermore, the patterns of
change in mean levels of PD features over time initially reported for the LSPD sample
(Lenzenweger, 1999) were subsequently replicated in both clinical (Shea et al., 2002; Zanarini et
al., 2003) and nonclinical community (Johnson et al., 2000) samples. Thus, although the present
sample is somewhat more compressed in terms of demographic background characteristics, this
has not led to findings at odds with those obtained in other epidemiological or longitudinal PD
research. Second, given that the LSPD subjects were selected from a population of first-year
university students, the sample may have been somewhat censored for individuals affected by
some of the most severe PDs. However, one must be cautious in ascribing undue levels of mental
health to subjects who happen to be selected for academic achievement, as such selection does
not confer immunity to psychopathology. To this end, I note that 16% (or 1 in every 6) of the
LSPD sample subjects was diagnosed with a formal Axis II disorder by the end of the study
period (i.e., by Wave 3) using the highly conservative IPDE. Many other subjects met
intermediate levels of PD criteria (e.g., 2 or 3 criteria) that fell short of DSM diagnostic threshold
counts but indicated the presence of some degree PD disturbance of clinical intensity
nonetheless. An additional strength of these results is that they are based on clinical interviews,
and self-reported personality traits, and therefore the method of assessment is not a confound.
Third, I am mindful that there are undoubtedly many predictors of rates of development
in personality and change in PD symptoms. Some change may be driven by important time-
varying processes of a broad (e.g., other temperament factors) or more specific nature (e.g.,
romantic relationships, developing friendships). I hope to establish such hypotheses and probe
the LSPD database more deeply in the coming years to explore such possibilities as well as
collect additional data relevant to this in upcoming Wave 4 and Wave 5 assessments for which
84
planning is currently underway. These will capture this sample in their 30’s and beyond,
allowing for tests to be extended to more advance ages. Finally, I emphasize that these analyses
measure personality traits broadly as opposed to more specifically. Future analyses might go
beyond the personality trait domains to look at the more specific facet/octant level of analysis.
Past research would suggest (Samuel & Widiger, 2008) that this might be a fruitful avenue for
future investigation.
85
Table 4.1. Parameter estimates and indices of fit for the five estimated parallel process growth models.
Dominance
Affiliation
Conscientiousness
Neuroticism
Openness
Coef. S.E. ES r Coef. S.E. ES r Coef. S.E. ES r Coef. S.E. ES r Coef. S.E. ES r
Model Intercepts
AVPD Intercept -3.514 4.137 0.05 -3.947 4.252 0.06 -3.851 4.364 0.06 -5.33 4.33 0.08 -3.378 4.245 0.05
AVPD Slope -0.024 2.777 0.00 0.141 2.606 0.00 0.116 2.743 0.00 0.861 2.581 0.02 -0.303 2.662 0.01
Personality Intercept -0.617 2.686 0.01 3.635 2.157 0.11 2.076 2.985 0.04 2.828 2.549 0.07 2.246 2.348 0.06
Personality Slope -0.398 0.635 0.04 -0.683 0.578 0.07 -0.329 0.602 0.03 1.104 0.894 0.08 -0.678 0.652 0.07
Growth Factor Covariances Path A -0.831*** 0.153 0.32 -0.344** 0.110 0.19 -0.174 0.137 0.08 0.648*** 0.129 0.30 -0.366*** 0.139 0.16
Path B -0.035* 0.018 0.12 -0.041* 0.020 0.13 -0.019 0.016 0.08 0.068** 0.021 0.20 -0.029† 0.017 0.11
Path C -0.095*** 0.027 0.22 -0.050 0.042 0.08 -0.054* 0.023 0.14 -0.069† 0.040 0.11 -0.086*** 0.024 0.22
Path D 0.056 0.129 0.03 0.052 0.143 0.02 0.067 0.141 0.03 0.035 0.148 0.02 0.063 0.143 0.03
Path E 0.046 0.030 0.09 0.051 0.037 0.09 -0.040 0.028 0.09 -0.001 0.044 0.00 0.023 0.032 0.05
Path F -0.007 0.059 0.01 0.011 0.074 0.01 -0.004 0.060 0.00 -0.055 0.056 0.06 0.169* 0.072 0.15
Regression on Covariates AVPD Intercept Sex 0.136 0.250 0.03 0.149 0.257 0.04 0.191 0.260 0.05 0.202 0.257 0.05 0.169 0.263 0.04
Age 0.077 0.220 0.02 0.096 0.225 0.03 0.090 0.232 0.02 0.169 0.229 0.05 0.064 0.225 0.02
Group 1.330*** 0.263 0.31 1.388*** 0.274 0.31 1.370*** 0.272 0.30 1.323*** 0.271 0.30 1.384*** 0.272 0.31
AVPD Slope Sex -0.201 0.131 0.10 -0.171 0.131 0.08 -0.184 0.132 0.09 -0.174 0.132 0.08 -0.181 0.130 0.09
Age -0.010 0.148 0.00 -0.019 0.139 0.01 -0.018 0.146 0.01 -0.060 0.137 0.03 0.004 0.142 0.00
Group -0.210 0.143 0.09 -0.210 0.142 0.09 -0.192 0.144 0.08 -0.160 0.145 0.07 -0.187 0.141 0.08
Personality Intercept Sex 0.133 0.151 0.06 -0.883*** 0.142 0.37 0.143 0.141 0.06 -0.316* 0.135 0.15 -0.234 0.140 0.10
Age 0.029 0.143 0.01 -0.115 0.115 0.06 -0.095 0.159 0.04 -0.222 0.136 0.10 -0.102 0.125 0.05
Group -0.080 0.143 0.04 -0.803*** 0.134 0.36 -0.464** 0.137 0.21 1.149*** 0.129 0.49 0.152 0.135 0.07
Personality Slope SEX -0.005 0.034 0.01 -0.021 0.038 0.04 -0.071* 0.032 0.14 0.032 0.047 0.04 0.036 0.036 0.06
AGE 0.020 0.034 0.04 -0.039 0.031 0.08 0.022 0.032 0.04 -0.062 0.047 0.08 0.038 0.034 0.07
Group 0.023 0.034 0.04 0.003 0.036 0.01 -0.024 0.032 0.05 -0.096* 0.046 0.13 0.014 0.036 0.02
Model Fit LL -1452.42
-1540.79
-1468.67
-1613.92
-1517.40
AIC 2962.85
3139.58
2995.33
3285.83
3090.80
BIC 3064.97
3241.71
3097.46
3296.02
3189.41
Note. N = 250. AVPD = Avoidant Personality Disorder; LL = Log Likelihood; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion. Coef.
= Raw Estimated Coefficient; S.E. = Standard Error; ES r = Effect Size r. † p < .10; * p < .05; ** p < .01; *** p < .001.
86
Figure 4.1. Conceptual Diagram of Parallel Process Growth Model.
Note. AVPD = Avoidant Personality Disorder Symptoms; P = Personality Trait Score; T1-T3 =
Study Wave 1-3; Single headed arrows denote regression paths, double headed arrows denote
covariances. Path A = Covariance between personality and AVPD intercepts; Path B =
Covariance between personality and AVPD slopes; Path C = Covariance between personality
intercept and growth factors; Path D = Covariance between AVPD intercept and growth factors;
Path E = Covariance between AVPD intercept and personality growth; Path F = Covariance
between personality intercept and AVPD growth.
87
CHAPTER 5
GENERAL CONCLUSION
The three preceding chapters were each written to stand alone as self-contained
manuscripts. Yet each is complementary to the others and speaks to important issues in the
science of personality, the psychopathology of PD, or both. Furthermore, each chapter included
detailed results and discussion sections that reported on the specific findings. Therefore, these
final words will attempt to integrate the findings, apply them to the questions that face the
science of PD, and outline some potential next steps for future investigation.
As stated in Chapter 1, the psychopathology of PD is currently in the middle of an ardent
debate on how best to define the core constructs of the discipline. The DSM-5 workgroup on
personality and PD recently published a set of proposed revisions to the constructs that has
served to catalyze this debate. Although the proposal has been met with hand-wringing by some,
the discourse it has generated is also setting the stage for what I anticipate will be a vibrant time
for the scientific study of PD. The underlying assumption of this dissertation is that the
psychopathology of PD would benefit from a scientific model that encompasses both normative
and non-normative personality functioning. The overarching goal of this dissertation was to
clarify the relationship between personality and PD by applying advanced quantitative analytical
techniques both cross-sectionally and longitudinally to variables of each. The data that were
chosen for these analyses come from the LSPD sample, which has a number of attractive features
for studying personality and PD. First, the longitudinal nature of the data was ideal for the study
of change in interpersonal aspects of personality (Chapter 2) and the conjoint change in
personality and PD (Chapter 4). Second, because of the distributions of PD in this sample, the
boundary between healthy and pathological functioning can be focally targeted (Chapter 3).
88
Chapter 2 stands apart from the subsequent chapters because the research contained
therein does not directly address the relationship between personality traits and PD. In part,
Chapter 2 serves as an important prerequisite for the models eventually tested in chapter 4.
Nevertheless, the results offer insight in to the development of interpersonal functioning in an
important developmental period. In conjunction with prior work with this sample (Wright et al.,
2011b) and a large body of existing research (see Roberts et al., 2006), the average
developmental trend in personality is towards increased functional maturity, especially during
the late adolescent and early adulthood years. However, individuals vary in the rate and direction
of change (see also Mroczek & Spiro, 2003). These results are now well known in basic
personality science, but are also highly informative to the scientific study and understanding of
PD by providing a context for understanding the stability of PD.
Much like early conceptions of personality (James, 1890), PD has been described as
“enduring” or “chronic” in its course. However, given that basic personality is not fixed, this
places the onus on psychopathologists to further specify the degree of stability of PD. Is it
expected to be more stable and enduring than basic personality variables? This might suggest that
there are self-sustaining maintenance factors. Or perhaps PD is expected to be equally stable,
suggesting that PD symptoms are of equivalent structural permanence as personality traits? In
contrast, perhaps PD is expected to be relatively stable compared to other disorders (e.g., panic
disorder), but not as stable as general personality functioning? These questions may seem like
little more than an intellectual exercise given that the LSPD (Lenzenweger, 1999; Lenzenweger
et al., 2004), the Collaborative Longitudinal Personality Disorders Study (CLPS; Skodol et al.,
2005), and others (Johnson et al., 2000; Zanarini et al., 2003) have shown that PD is not very
stable at all. On the contrary, however, I believe these questions are now more germane than
89
ever—especially if they are asked in the context of revising the field’s conceptions of PD. The
clinical observation that PD is a highly stable form of pathology has not held up empirically, but
these studies have been contingent on the DSM’s current articulation of PD, which has been
highly criticized on a number of grounds. Some existing evidence from the CLPS sample
demonstrates that PD symptoms are differentially stable, with some showing trait like stability
and others ebbing much more rapidly (Skodol et al., 2005). Other research with the same clinical
sample has shown that as the defined symptoms wane functional impairment (e.g., social and
occupational) remains.
One implication of this would be to move away from descriptions of PD that focus on
symptomatic flare ups (e.g., suicidal gestures), and instead define it in terms of more central and
enduring pathological tendencies (e.g., mood lability). By considering the results of basic
personality science, much more precise research questions can be formulated. For example,
based on clinical observation, what would we expect the stability coefficient for a given PD to
be? Alternatively, in what respects is PD stable: rank order, structural, individual, etc.? This is
just one manner in which basic psychological science can inform psychopathology. Yet, it also
requires a theoretical assumption; namely, that PD and personality are “made of the same
substance” and operate under similar principles (Wright, in press).
Chapters 3 and 4 address the question of how PD and personality are related more
directly. To be most accurate, the work presented in the prior two chapters serves to clarify the
relationship between personality traits and the established DSM PDs. Alternative conceptions of
personality and PD exist, and have associated theoretical and empirical literatures, although no
other personality or PD frameworks are as actively researched as these. Further, among the
theoretical proposals linking personality and PD, the one that has perhaps received the most
90
empirical attention is the suggestion that PD can be adequately represented using normal trait
profiles (cf. Widiger & Trull, 2007). Thus, this dissertation joins the discourse in each of these
lively fields in the hopes of contributing valuable new information.
The results from the parallel process LGM offer some of the most compelling evidence to
demonstrate the personality and PD are meaningfully related. In fact, the two seem to be
inextricably intertwined and the well known cross-sectional relationships appear to hold up
longitudinally in a very direct way—an individual’s rate of change in PD is related to an
individual’s rate of change in personality traits. Although the focus on this chapter was on
AVPD, I view this as an exemplar, serving to demonstrate a useful methodological approach.
But, I also fully expect that similar results would emerge with other PDs included in the models.
It is worth noting that a cross-sectional relationship does not imply a similar longitudinal
relationship, and therefore these results are all the more striking because they do emerge in such
a consistent pattern. These findings can be added to the accumulating trove of evidence that PD
and personality are best understood within the same scientific framework.
However, despite their importance, the results of Chapter 4 must be understood in the
context of Chapter 3’s results. A strong association between PD and personality traits does not
require that each is continuous with the other, even if both is best understood as dimensional
(Morey et al., 2007). One of the least robust aspects of the trait theories of personality disorder is
the link between normality and pathology, which remains poorly articulated and empirically
undemonstrated (Wright, in press). Some recent work that has employed item response theory to
compare the characteristics of scales that measure normal range personality constructs with
scales that measure maladaptive construct have met with what I interpret as equivocal results (see
Samuel et al., 2010; Walton et al., 2006). The approach adopted here was distinct, using a variety
91
distributions to show first that PD symptoms are not normally distributed, and this has important
modeling implications. And, second, using mixture distributions to demonstrate that what
distinguishes those individuals with pathology from those who have none is not always the same
as what predicts the severity of pathology in those who have it.
Thus, although there are strong benefits from adopting an approach towards personality
pathology that is rooted in basic personality, it should not be simplistic or reductionist. For trait
conceptions of personality in particular, there is a clear necessity to have well defined
articulations of what constitutes a “maladaptive variant” (cf. Widiger & Trull, 2007). Personally,
I believe that maladaptive aspects of personality traits exist in the process of how they are
expressed (Wright, in press). The most parsimonious description of this viewpoint was offered
first by Leary (1957), although it recently has been reviewed elsewhere (Pincus & Hopwood, in
press; Pincus & Wright, 2010; Wright, in press). Trait related behaviors that are expressed rigidly
(i.e., to the exclusion of other more appropriate behaviors), extremely (i.e., with more intensity
than necessary), or that do not match the environment (i.e., choosing a behavior that does not suit
the situational demands) are likely to result in maladaptive functioning. These hypotheses were
not tested here. To empirically test these assumptions data of a different sort must be collected.
Individuals must be sampled repeatedly for behaviors, emotions, and cognitions that span normal
and extreme expressions. This would likely provide further needed clarity on the boundary
between normative and non-normative functioning, and further elucidate the temporal stability of
pathology on a scale (i.e., within day, daily) that can better inform treatment and interventions.
92
References
Alden, L.E., Laposa, J.M., Taylor, C.T., & Ryder, A.G. (2002). Avoidant personality disorder:
Current status and future directions. Journal of Personality Disorders, 16, 1-29.
Alden, L.E. & Capreol, M.J. (1993). Avoidant personality disorder: Interpersonal problems as
predictors of treatment response. Behavior Therapy, 24, 357-376.
American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders
(4th ed), Text Revision. Washington, DC: Author.
Ansell, E.B., & Pincus, A.L. (2004). Interpersonal perceptions of the five-factor model of
personality: An examination using the structural summary method for circumplex data.
Multivariate Behavioral Research, 39, 167-201.
Arnett, J. J. (2000). Emerging adulthood: A theory of development from the late teens through
the twenties. American Psychologist, 55(5), 469-480.
Atkins, D.C. & Gallo, R.J. (2007). Rethinking how family researchers model infrequent
outcomes: A tutorial on count regression and zero-inflated models. Journal of Family
Psychology, 21(4), 726-735.
Bagby, R. M., Costa, P. T., Jr., Widiger, T. A., Ryder, A. G., & Marshall, M. (2005). DSM-IV
personality disorders and the five-factor model of personality: A multi-method examination
of domain--and facet-level predictions. European Journal of Personality. Special Issue:
Personality and Personality Disorders, 19, 307-324.
Barber, J.P., Morse, J.Q., Krakauer, I.D., Chittams, J., & Crtis-Christoph, K. (1997). Change in
obsessive-compulsive and avoidant personality disorders following time-limited supportive-
expressive therapy. Psychotherapy, 34, 133-143.
Bleidorn, W., Kandler, C., Riemann, R., Angleitner, A., & Spinath, F.M. (2009). Patterns and
93
sources of adult personality development: Growth curve analyses of the NEO PI-R scales
in a longitudinal twin study. Journal of Personality and Social Psychology, 97, 142-155.
Bollen, K.A. & Curran, P.J. (2006). Latent curve models: A structural equation perspective.
Hoboken, NJ: John Wiley & Sons.
Caspi, A., & Roberts, B. W. (2001). Target article: Personality development across the life
course: The argument for change and continuity. Psychological Inquiry, 12(2), 49-66.
Clark, L. A. (2007). Assessment and diagnosis of personality disorder: Perennial issues and an
emerging reconceptualization. Annual Review of Psychology, 58, 227-257.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation
analysis for the behavioral sciences. Mahwah, NJ: Erlbaum.
Cohen, P., Crawford, T. N., Johnson, J. G., & Kasen, S. (2005). The children in the community
study of developmental course of personality disorder. Journal of Personality Disorders.
Special Issue: Longitudinal Studies, 19(5), 466-486.
Costa, P. T., & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO–PI–R) and
NEO Five-Factor Inventory (NEO–FFI) professional manual. Odessa, FL: Psychological
Assessment Resources.
Costa, P. T., & McCrae, R. R. (1997). Longitudinal stability of adult personality. In R. Hogan, J.
Johnson, & S. Briggs (Eds.), Handbook of personality psychology (pp. 269–292). San
Diego: Academic Press.
Coxe, S., West, S.G., & Aiken, L.S. (2009). The analysis of count data: A gentle introduction to
Poisson regression and its alternatives. Journal of Personality Assessment, 91(2), 121-
136.
Crawford, T. N., Cohen, P., Johnson, J. G., Kasen, S., First, M. B., Gordon, K., & Brook, J. S.
94
(2005). Self-reported personality disorder in the Children in the Community Sample:
Convergent and prospective validity in late adolescence and adulthood. Journal of
Personality Disorders, 19, 30–52. doi: 10.1521/pedi.19.1.30.62179
Cronbach, L.J. & Gleser, G.C (1953). Assessing similarity between profiles. The Psychological
Bulletin, 50(6), 456-483.
Depue, R. A., & Collins, P. F. (1999). Neurobiology of the structure of personality: Dopamine,
facilitation of incentive motivation, and extraversion. Behavioral and Brain Sciences, 22,
491-569.
Depue, R.A. & Lenzenweger, M.F. (2001). A neurobehavioral dimensional model. In W.J.
Livesley (Ed.), Handbook of personality disorders: Theory, research, and treatment (pp.
136-176). New York, NY: Guilford.
Depue, R. A., & Lenzenweger, M. F. (2005). A neurobehavioral dimensional model of
personality disturbance. In M. F. Lenzenweger, & J. F. Clarkin (Eds.), Major theories of
personality disorder (2nd ed). (pp. 391-453). New York, NY, US: Guilford Press.
Depue, R. A., & Morrone-Strupinsky, J. V. (2005). A neurobehavioral model of affiliative
bonding: Implications for conceptualizing a human trait of affiliation. Behavioral and
Brain Sciences, 28, 313-395.
Donnellan, M. B., Conger, R. D., & Burzette, R. G. (2007). Personality development from late
adolescence to young adulthood: Differential stability, normative maturity, and evidence
for the maturity-stability hypothesis. Journal of Personality, 75(2), 237-263.
Erickson, T. M., Newman, M. G., & Pincus, A. L. (2009). Predicting unpredictability: Do
measures of interpersonal rigidity/flexibility and distress predict intraindividual
variability in social perceptions and behavior? Journal of Personality and Social
95
Psychology 97, 893-912.
Goldberg, L.R. (1990). An alternative “Description of personality”: The Big Five factor
structure. Journal of Personality and Social Psychology, 59, 1216–1229.
Grilo, C.M., Shea, T.M., Sanislow, C.A., Skodol, A.E., Gunderson, J.G.,…McGlashan, T.H.
(2004). Two-year stability and change of schizotypal, borderline, avoidant, and
obsessive-compulsive personality disorders. Journal of Consulting and Clinical
Psychology, 72, 767-775.
Gurtman, M.B., & Balakrishnan, J.D. (1998). Circular measurement redux: The analysis and
interpretation of interpersonal circle profiles. Clinical Psychology: Science and Practice,
5, 344-360.
Gurtman, M. B., & Pincus, A. L. (2003). The circumplex model: Methods and research
applications. In J. Schinka & W. Velicer (Eds.), Handbook of psychology: Research
methods in psychology, Vol. 2 (pp. 407-428). Hoboken, NJ: Wiley
Guttman, L. (1954). A new approach to factor analysis: The radex. In P. F. Lazarsfeld (Ed.),
Mathematical thinking in the social sciences (pp. 258-348). Glencoe, IL: Free Press.
Hogan, R.T. (1983). A socioanalytic theory of personality. In M. Page (Ed.), 1982 Nebraska
symposium on motivation (pp. 55-89). Lincoln, NE: University of Nebraska Press.
Hogan, R.T. (1996). A socioanalytic perspective on the five-factor model. In J.S. Wiggins (Ed.),
The five-factor model of personality: Theoretical perspectives (pp. 163-179). New York,
NY: Guilford Press.
Hogan, R., & Roberts, B. W. (2004). A socioanalytic model of maturity. Journal of Career
Assessment, 12, 204–217.
Hopwood, C.J., Donnellan, M.B., Blonigen, D.M., Krueger, R.F., McGue, M., Iacono, W.G., &
96
Burt, A.S. (2011). Genetic and environmental influences on personality trait stability and
growth during the transition to adulthood: A three-wave longitudinal study. Journal of
Personality and Social Psychology, 100(3), 545-556.
Hopwood, C.J., Malone, J.C., Ansell, E.B., Sanislow, C.A., Grilo, C.M., McGlashan, T.H.,
…Morey, L.C. (in press). Personality Assessment in DSM-V: Empirical support for
rating severity, style, and traits. Journal of Personality Disorders.
Hopwood, C.J., Newman, D.A., Donnellan, M. B., Markowitz, J. C., Grilo, C. M., Sanislow, C.
A. … & Morey, L. C. (2009). The stability of personality traits in individuals with
borderline personality disorder. Journal of Abnormal Psychology, 118(4), 806-815.
Horowitz, L.M., & Strack, S. (2010). Handbook of interpersonal psychology. Hoboken, NJ: John
Wiley & Sons.
Hubert, L.J., & Arabie, P. (1987). Evaluating order hypotheses within proximity matrices.
Psychological Bulletin, 102, 172-178.
Jackson, J.J., Bogg, T., Walton, K.E., Wood, D., Harms, P.D., Lodi-Smith, J. … Roberts, B.W.
(2009). Not all conscientiousness scales change alike: A multimethod, multisample study
of age differences in the facets of conscientiousness. Journal of Personality and Social
Psychology, 96(2), 446-459.
James, W. (1950). The principles of psychology. New York: Dover. (Original work published in
1890).
Johnson, J. G., Cohen, P., Kasen, S., Skodol, A. E., Hamagami, F., & Brook, J. S. (2000). Age-
related change in personality disorder trait levels between early adolescence and
adulthood: A community-based longitudinal investigation. Acta Psychiatrica
Scandinavica, 102(4), 265-275.
Kass, F., Skodol A.E., Charles, E., Spitzer, R.L., Williams, J.B. (1985). Scaled ratings of DSM-III
97
personality disorders. American Journal of Psychiatry, 142, 627–630.
Kessler, R. C., Chiu, W. T., Demler, O., & Walters, E. E. (2005). Prevalence, severity, and
comorbidity of the 12-month DSM-IV disorders in the National Comorbidity Survey
Replication. Archives of General Psychiatry, 62, 617–627.
Kernberg, O.F. (1984). Severe Personality Disorders: Psychotherapeutic Strategies. New Haven,
CT: Yale University Press.
Kiesler, D. J. (1996). Contemporary interpersonal theory and research: Personality,
psychopathology, and psychotherapy. New York: John Wiley & Sons.
King, G. (1988). Statistical models for political science event counts: Bias in conventional
procedures and evidence for the exponential Poisson regression model. American Journal of
Political Science, 32, 838–863.
Krueger, R.F., Eaton, N.R., Clark, L.A., Watson, D., Markon, K.E., … Livesley, W.J. (2011).
Deriving an empirical structure of personality pathology for DSM-5. Journal of Personality
Disorders. 170-191.
Leary, T. (1957). Interpersonal diagnosis of personality. New York: Ronald Press.
Lee, K., & Ashton, M. C. (2004). The HEXACO Personality Inventory: A new measure of the
major dimensions of personality. Multivariate Behavioral Research, 39, 329-358.
Lenzenweger, M. F. (1999). Stability and change in personality disorder features: The
longitudinal study of personality disorders. Archives of General Psychiatry, 56(11), 1009-
1015.
Lenzenweger, M.F. (2006). The longitudinal study of personality disorders: History, design
considerations, and initial findings. Journal of Personality Disorders, 20(6), 645-670.
Lenzenweger, M. F. (2008). Epidemiology of personality disorders. Psychiatric Clinics of North
America, 31(3), 395-403.
98
Lenzenweger, M. F., Johnson, M. D., & Willett, J. B. (2004). Individual growth curve analysis
illuminates stability and change in personality disorder features: The longitudinal study of
personality disorders. Archives of General Psychiatry, 61(10), 1015-1024.
Lenzenweger, M. F., Lane, M. C., Loranger, A. W., & Kessler, R. C. (2007). DSM-IV
personality disorders in the national comorbidity survey replication. Biological
Psychiatry, 62(6), 553-564.
Lenzenweger, M.F., Loranger, A.W., Korfine, L., & Neff, C. (1997). Detecting personality
disorders in a nonclinical population: Application of a 2-stage for case identification.
Archives of General Psychiatry, 54(4), 345-351.
Lenzenweger, M. F., & Willett, J. B. (2007). Predicting individual change in personality disorder
features by simultaneous individual change in personality dimensions linked to
neurobehavioral systems: The longitudinal study of personality disorders. Journal of
Abnormal Psychology, 116(4), 684-700.
Livesley, W.J. (2010). Confusion and incoherence in the classification of personality disorder:
Commentary on the preliminary proposals for DSM-5. Psychological Injury and Law, 3, 304-
313.
Livesley, W.J. & Jang, K.L. (2005). Differentiating normal, abnormal, and disordered personality.
European Journal of Personality, 19, 257-268.
Loranger, A. (1988). The Personality Disorder Examination (PDE) manual. Yonkers, NY: DV
Communications.
Loranger, A.W. (1999). International Personality Disorder Examination: DSM-IV and ICD-10
Interviews. Odessa, Fla: Psychological Assessment Resources Inc.
Lynam, D.R., & Widiger, T.A. (2001). Using the five factor model to represent the DSM-IV
personality disorders: An expert consensus approach. Journal of Abnormal Psychology,
99
110, 401-412.
McCrae, R.R., & Costa, P.T. (1989). The structure of interpersonal traits: Wiggins's circumplex
and the five-factor model. Journal of Personality and Social Psychology, 56(4), 586-595.
Millon, T. (2011). Disorders of personality: Introducing a DSM/ICD spectrum from normal to
abnormal, 3rd
Ed. Hoboken: Wiley.
Morey, L. C., Gunderson, J. G., Quigley, B. D., Shea, M. T., Skodol, A. E., … Zanarini, M. C.
(2002). The representation of borderline, avoidant, obsessive-compulsive, 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, M. T. … McGlashan, T. H.
(2007). Comparison of alternative models for personality disorders. Psychological
Medicine, 37(7), 983-994.
Moskowitz, D. S., & Zuroff, D. C. (2004). Flux, pulse, and spin: Dynamic additions to the
personality lexicon. Journal of Personality and Social Psychology, 86, 880-893.
Moskowitz, D. S., & Zuroff, D. C. (2005). Robust predictors of flux, pulse, and spin. Journal of
Research in Personality, 39, 130-147.
Mroczek, D.K., & Spiro, A., III. (2003). Modeling intraindividual change in personality traits:
Findings from the normative aging study. The Journals of Gerontology: Series B:
Psychological Sciences and Social Sciences, 58B(3), P153-P165.
Muthén, L.K. and Muthén, B.O. (1998-2010). Mplus User’s Guide. Sixth Edition. Los Angeles,
CA: Muthén & Muthén.
Neyer, F.J., & Asendorpf, J.B. (2001). Personality–relationship transaction in young adulthood.
Journal of Personality and Social Psychology, 81(6), 1190-1204.
100
Neyer, F.J. & Lehnart, J. (2007). Relationships matter in personality development: Evidence
from an 8-year longitudinal study across young adulthood. Journal of Personality, 75(3),
535-568.
O’Connor, B.P. (2005). Graphical analysis of personality disorders in five-factor model space.
European Journal of Personality, 19, 287-305.
Parker, G., Hadzi‐Pavlovic, D., Both, L., Kumar, S., Wilhelm, L., & Olley, A. (2004). Measuring
disordered personality functioning: To love and work reprised. Acta Psychiatrica
Scandinavica, 110, 230‐239.
Pincus, A.L. (2002). Constellations of dependency within the five-factor model of personality. In
P. T. Costa Jr., & T. A. Widiger (Eds.), Personality disorders and the five-factor model of
personality (2nd ed.). (pp. 203-214). Washington, DC, US: American Psychological
Association.
Pincus, A. L. (2005). A contemporary integrative interpersonal theory of personality disorders. In
J. Clarkin & M. Lenzenweger (Eds.), Major theories of personality disorder (2nd Ed.)
(pp. 282-331). New York: Guilford.
Pincus, A.L., & Ansell, E.B. (2003). Interpersonal theory of personality. In T. Millon & M.
Lerner (Eds.), Handbook of psychology: Personality and social psychology, Vol. 5 (pp.
209-229). Hoboken, NJ: John Wiley & Sons Inc.
Pincus, A.L., & Ansell, E.B. (in press). Interpersonal theory of personality. In H. Tennen & J.
Suls (Eds.), Handbook of psychology: Personality and social psychology, Vol. 5 (2nd
ed.).
Hoboken, NJ: John Wiley & Sons.
Pincus, A.L., & Hopwood, C.J. (in press). A contemporary interpersonal model of personality
pathology and personality disorder. In T.A. Widiger (Eds). The Oxford handbook of
personality disorders. New York: Oxford University Press.
101
Pincus, A. L., & Wiggins, J. S. (1990). Interpersonal problems and conceptions of personality
disorders. Journal of Personality Disorders, 4(4), 342-352.
Pincus, A.L., & Wright, A.G.C. (2010). Interpersonal diagnosis of psychopathology. In L.M.
Horowitz & S. Strack (Eds.). Handbook of interpersonal psychology: Theory, research,
and therapeutic interventions (pp. 359-381). Hoboken: John Wiley & Sons Inc.
Pytlik Zillig, L. M., Hemenover, S. H., & Dienstbier, R. A. (2002). What do we assess when we
assess a big 5 trait? A content analysis of the affective, behavioral and cognitive
processes represented in the big 5 personality inventories. Personality and Social
Psychology Bulletin, 28(6), 847-858.
Ram, N. & Gerstorf, D. (2009). Time-structured and net intraindividual variability: Tools for
examining the development of dynamic characteristics and processes. Psychology and
Aging, 24, 778-791.
Raudenbush, S.W., Bryk, A.S., Cheong, Y.F., Congdon, R. (2004). HLM-6: Hierarchical Linear
and Nonlinear Modeling. Lincolnwood, IL: Scientific Software International.
Reynolds, S. K., & Clark, L. A. (2001). Predicting dimensions of personality disorder from
domains and facets of the five-factor model. Journal of Personality, 69, 199-222.
Rindfuss, R.R. (1991). The young adult years: Diversity, structural change, and fertility.
Demography, 28, 493–512.
Roberts, B.W., Caspi, A., & Moffitt, T.E. (2001). The kids are alright: Growth and stability in
personality development from adolescence to adulthood. Journal of Personality and
Social Psychology, 81(4), 670-683.
Roberts, B.W., & DelVecchio, W.F. (2000). The rank-order consistency of personality traits
from childhood to old age: A quantitative review of longitudinal studies. Psychological
102
Bulletin, 126(1), 3-25.
Roberts, B. W., & Jackson, J. J. (2008). Sociogenomic personality psychology. Journal of
Personality, 76, 1523-1544.
Roberts, B.W., & Mroczek, D. (2008). Personality trait change in adulthood. Current Directions
in Psychological Science, 17(1), 31-35.
Roberts, B.W., Walton, K.E., & Viechtbauer, W. (2006). Patterns of mean-level change in
personality traits across the life course: A meta-analysis of longitudinal studies.
Psychological Bulletin, 132(1), 1-25.
Roberts, B.W., Wood, D., & Caspi, A. (2008). The development of personality traits in
adulthood. In O.P. John, R.W. Robins & L.A. Pervin (Eds.), Handbook of personality
psychology: Theory and research (3rd ed.). (pp. 375-398). New York, NY, US: Guilford.
Robins, R.W., Fraley, R.C., Roberts, B.W., & Trzesniewski, K.H. (2001). A longitudinal study
of personality change in young adulthood. Journal of Personality, 69(4), 617-640.
Rosenthal, R., & Rosnow, R.L. (1991). Essentials of behavioral research: Methods and data
analysis (2nd
ed.). New York, NY: McGraw-Hill.
Russell, J. J., Moskowitz, D. S., Zuroff, D. C., Sookman, D., & Paris, J. (2007).Stability and
variability of affective experience and interpersonal behavior in borderline personality
disorder. Journal of Abnormal Psychology, 116, 578-588.
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(8), 1326-1342.
103
Samuels, J. E., Eaton, W. W., Bienvenu, O. J., Brown, C., Costa, P. T., & Nestadt, G. (2002).
Prevalence and correlates of personality disorders in a community sample. British
Journal of Psychiatry, 180, 536–542.
Saulsman, L. M., & Page, A. C. (2004). The five-factor model and personality disorder empirical
literature: A meta-analytic review. Clinical Psychology Review, 23(8), 1055-1085.
Schuerger, J.M., Zarrella, K.L., & Hotz, A.S. (1989). Factors that influence the temporal stability
of personality by questionnaire. Journal of Personality and Social Psychology, 56, 777-
783.
Shea, M. T., Stout, R., Gunderson, J., Morey, L. C., Grilo, C. M., … M. C., & Keller, M. B.
(2002). Short-term diagnostic stability of schizotypal, borderline, avoidant, and obsessive-
compulsive personality disorders. The American Journal of Psychiatry, 159(12), 2036-2041.
Singer, J.D., & Willett, J.B. (2003). Applied longitudinal data analysis: Modeling change and
event occurrence. New York, NY, US: Oxford University Press.
Skodol, A.E., Clark, L.A., Bender, D.S., Krueger, R.F., Morey, L.C., … Oldham, J.M. (2011).
Proposed changes in personality and personality disorder assessment and diagnosis for
DSM-5 Part I: Description and rationale. Personality Disorders: Theory, Research, and
Treatment, 2(1), 4-22.
Skodol, A. E., Pagano, M. E., Bender, D. S., Shea, M. T., Gunderson, J. G., … McGlashan, T.
H. (2005). Stability of functional impairment in patients with schizotypal, borderline,
avoidant, or obsessive-compulsive personality disorder over two years. Psychological
Medicine: A Journal of Research in Psychiatry and the Allied Sciences, 35(3), 443-451.
Soldz, S., Budman, S., Demby, A., & Merry, J. (1993). Representation of personality disorders in
circumplex and five factor space: Explorations with a clinical sample. Psychological
104
Assessment, 5, 41–52.
Srivastava, S., John, O.P., Gosling, S.D., & Potter, J. (2003). Development of personality in early
and middle adulthood: Set like plaster or persistent change? Journal of Personality and
Social Psychology, 84(5), 1041-1053.
Sullivan, H.S. (1953). The interpersonal theory of psychiatry. New York, NY: Norton.
Tracey, T.J.G. (1997). RANDALL: A Microsoft FORTRAN program for a randomization test of
hypothesized order relations. Educational and Psychological Measurement, 57, 164-168.
Tracey, T.J.G. (2005). Interpersonal rigidity and complementarity. Journal of Research in
Personality, 39, 592-614.
Tracey, T. J. G., & Rohlfing, J. E. (2010). Variations in the understanding of interpersonal
behavior: Adherence to the interpersonal circle as a moderator of the rigidity
psychological well-being relation. Journal of Personality, 78, 711-746.
Trapnell, P. D., & Wiggins, J. S. (1990). Extension of the interpersonal adjective scales to
include the big five dimensions of personality. Journal of Personality and Social
Psychology, 59(4), 781-790.
Vaidya, J.G., Gray, E.K., Haig, J.R., Mroczek, D.K., & Watson, D. (2008). Differential stability
and individual growth trajectories of big five and affective traits during young adulthood.
Journal of Personality, 76(2), 267-304.
Warner, M.B., Morey, L.C., Finch, J.F., Gunderson, J.G., Skodol, A.E., Sanislow, C.A., Shea,
M.T., McGlashan, T.H., & Grilo, C.M. (2004). The longitudinal relationship of personality
traits and disorders. Journal of Abnormal Psychology, 113(2), 217-227.
Widiger, T. A., & Clark, L. A. (2000).Toward DSM–V and the classification of psychopathology.
Psychological Bulletin, 126, 946–963.
105
Widiger, T. A., & Simonsen, E. (2005). Introduction to part two of the special section on the
research agenda for the development of a dimensional classification of personality
disorder. Journal of Personality Disorders, 19(3), 211-211.
Widiger, T.A. & Trull, T.J. (2007). Plate tectonics in the classification of personality disorder:
Shifting to a dimensional model. American Psychologist, 62, 71-83.
Widiger, T. A., Trull, T. J., Clarkin, J. F., Sanderson, C., & Costa, P. T., Jr. (2002). A description
of the DSM-IV personality disorders with the five-factor model of personality. In P. T.
Costa, Jr., & T. A.Widiger (Eds.), Personality disorders and the five-factor model of
personality (2nd ed., pp. 9– 99). Washington, DC: American Psychological Association.
Wiggins, J.S. (1991). Agency and communion as conceptual coordinates for the understanding
and measurement of interpersonal behavior. In D. Cicchetti, & W. M. Grove (Eds.),
Thinking clearly about psychology: Essays in honor of paul E. meehl, vol. 1: Matters of
public interest; vol. 2: Personality and psychopathology. (pp. 89-113). Minneapolis, MN,
US: University of Minnesota Press.
Wiggins, J. S., Phillips, N., & Trapnell, P. (1989). Circular reasoning about interpersonal
behavior: Evidence concerning some untested assumptions underlying diagnostic
classification. Journal of Personality and Social Psychology, 56, 296-305.
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.
Wiggins, J.S., Trapnell, P., & Phillips, N. (1988). Psychometric and geometric characteristics of
the revised interpersonal adjective scales (IAS-R). Multivariate Behavioral Research,
23(4), 517-530.
106
Wright, A.G.C. (in press). Qualitative and quantitative distinctions in personality disorder.
Journal of Personality Assessment.
Wright, A.G.C., Pincus, A.L., Conroy, D.E., & Hilsenroth, M.J. (2009). Integrating methods to
optimize circumplex description and comparison of groups. Journal of Personality
Assessment, 91, 311-322.
Wright, A.G.C., Pincus, A.L., & Lenzenweger, M.F. (2011a). An empirical examination of
distributional assumptions underlying the relationship between personality disorder
symptoms and personality traits. Manuscript in preparation.
Wright, A.G.C., Pincus, A.L., & Lenzenweger, M.F. (2011b). Departures and arrivals:
Development of personality and its pathology. Manuscript submitted for publication.
Wright, A.G.C., Pincus, A.L., & Lenzenweger, M.F. (2011c). Interpersonal development,
stability, and change in young adulthood. Manuscript submitted for publication.
Wright, A.G.C., Pincus, A.L., & Lenzenweger, M.F. (2010). Modeling Stability and Change in
Borderline Personality Disorder Symptoms using the Revised Interpersonal Adjective
Scales - Big Five (IASR-B5). Journal of Personality Assessment, 92, 501-513.
Zanarini, M. C., Frankenburg, F. R., Hennen, J., & Silk, K. R. (2003). The longitudinal course of
borderline psychopathology: 6-year prospective follow-up of the phenomenology of
borderline personality disorder. The American Journal of Psychiatry, 160(2), 274-283.
Curriculum Vitae
Aidan G. C. Wright
Education
2011 – 2012 Clinical Internship, Western Psychiatric Institute and Clinic
2006 – 2012 Doctor of Philosophy, Clinical Psychology, The Pennsylvania State University
2004 – 2006 Master of Science, Psychology, Villanova University
1999 – 2003 Bachelor of Arts, Psychology, The Pennsylvania State University
Honors and Awards
2011 Rising Star Award, Association for Research in Personality
2011 Jerry S. Wiggins Student Award for Outstanding Interpersonal Research, Society for Interpersonal
Theory and Research
2010-2012 NIMH Ruth H. Kirschstein National Research Service Award Predoctoral Fellowship
2010 Mary S. Cerney Award for Outstanding Student Research Paper, Society for Personality Assessment
2010 Raymond Lombra Graduate Student Award for Excellence in Research in the Social Sciences, College
of the Liberal Arts, The Pennsylvania State University
2009 Martin T. Murphy Award for Excellence in Clinical Psychology, The Pennsylvania State University
2009 Outstanding Publication by a Graduate Student in Psychology, The Pennsylvania State University
2007-2008 Penn State Quantitative Social Science Initiative Predoctoral Fellowship
Selected Publications
Wright, A.G.C., Pincus, A.L., Conroy, D.E., & Hilsenroth, M.J. (2009). Integrating methods to optimize circumplex
description and comparison of groups. Journal of Personality Assessment, 91(4), 311-322
Wright, A.G.C., Pincus, A.L., Conroy, D.E., & Elliot, A. (2009). The pathoplastic relationship between interpersonal
problems and fear of failure. Journal of Personality, 77(4), 997-1024.
Pincus, A.L., Ansell, E.B., Pimentel, C.A., Cain, N.M., Wright, A.G.C., & Levy, K.N. (2009). The initial
development and derivation of the Pathological Narcissism Inventory. Psychological Assessment, 21(3),
365-379.
Wright, A.G.C., Pincus, A.L., & Lenzenweger, M.F. (2010). Modeling stability and change in borderline personality
disorder symptoms using the Revised Interpersonal Adjective Scales - Big Five (IASR-B5). Journal of
Personality Assessment, 92(6), 501-513.
Wright, A.G.C., Lukowitsky, M.R., Pincus, A.L., & Conroy, D.E. (2010). The higher-order factor structure and
gender invariance of the Pathological Narcissism Inventory. Assessment, 17(4), 467-483.
Wright, A.G.C., Pincus, A.L., & Lenzenweger, M.F. (2011). Development of personality and the remission and
onset of personality pathology. Journal of Personality and Social Psychology, 101(6), 1351-1358.
Wright, A.G.C. (2011). Quantitative and qualitative distinctions in personality disorder. Journal of Personality
Assessment, 93(4), 370-379.
Pincus, A.L. & Wright, A.G.C. (2011). Interpersonal diagnosis of psychopathology. In L.M. Horowitz and S. Strack
(Eds.) Handbook of Interpersonal Psychology: Theory, Research, Assessment, and Therapeutic
Interventions (pp. 359-381). New York: Wiley.
Wright, A.G.C., Pincus, A.L., & Lenzenweger, M.F. (in press). An empirical examination of distributional
assumptions underlying the relationship between personality disorder symptoms and personality traits.
Journal of Abnormal Psychology.
Wright, A.G.C., Pincus, A.L., Hopwood, C.J., Thomas, K.M., Markon, K.E., & Krueger, R.F. (in press). An
interpersonal analysis of pathological personality traits in DSM-5. Assessment. Wright, A.G.C., Pincus, A.L., & Lenzenweger, M.F. (in press). A parallel process growth model of avoidant
personality disorder symptoms and personality traits. Personality Disorders: Theory, Research, and
Treatment.
Wright, A.G.C., Thomas, K.M., Hopwood, C.J., Markon, K.E., Pincus, A.L. & Krueger, R.F. (in press). The
hierarchy of DSM-5 pathological personality traits. Journal of Abnormal Psychology.
Wright, A.G.C., Pincus, A.L., & Lenzenweger, M.F. (in press). Interpersonal development, stability, and change in
young adulthood. Journal of Personality.