Download USING DYNAMIC FACTOR ANALYSIS TO MODEL

The Pennsylvania State University
The Graduate School
College of the Liberal Arts
USING DYNAMIC FACTOR ANALYSIS TO MODEL INTRAINDIVIDUAL
VARIATION IN BORDERLINE PERSONALITY DISORDER SYMPTOMS
A Dissertation in
Psychology
by
William DeLoache Ellison
© 2013 William DeLoache Ellison
Submitted in Partial Fulfillment
of the Requirements
for the Degree of
Doctor of Philosophy
December 2013
ii
The dissertation of William D. Ellison was reviewed and approved* by the following:
Kenneth N. Levy
Associate Professor of Psychology
Dissertation Adviser
Chair of Committee
Aaron L. Pincus
Professor of Psychology
Reginald B. Adams, Jr.
Associate Professor of Psychology
Peter Molenaar
Distinguished Professor of Human Development
Melvin M. Mark
Professor of Psychology
Head of the Department of Psychology
*Signatures are on file in the Graduate School.
iii
ABSTRACT
Borderline personality disorder (BPD) is a highly prevalent, debilitating, and costly form
of mental illness that involves instability in self-concept, emotions, and behavior, including
chronic suicidality and self-injury. The essential psychological dynamics underlying the disorder
are poorly understood. In particular, the role of identity disturbance in the disorder is largely
unexplained, although several prominent theories have been advanced. Psychodynamic theories
posit that identity disturbance underlies emotional and behavioral dysregulation, whereas
biosocial theory suggests that identity disturbance is the result and largely not the cause of these
other symptoms. Research to date has been largely cross-sectional and based on retrospective
report, making it difficult to untangle the temporal dynamics of BPD symptoms. In addition,
new research based on ecological momentary assessment (EMA), which could potentially shed
light on this question, has focused on groupwise hypotheses and analyzed interindividual data.
This approach does not account for potential heterogeneity in these processes, whereas personspecific methods based on intraindividual variation can account for this heterogeneity. The
current study uses dynamic factor analysis, a person-specific modeling approach, to investigate
the longitudinal covariation of anger, impulsivity, and identity disturbance. 11 psychiatric
outpatients who were diagnosed either with BPD (n = 4) or with a mood or anxiety disorder, but
not BPD (n = 7) completed a 21-day EMA protocol by rating these symptoms six times per day
at quasi-random times at roughly 2-hour intervals. Cubic spline interpolation was used to
produce time series with equal spacing between measurements, and models were created to
describe the relationship between these symptoms, both in synchronous ratings and at successive
time points. Models were created using examination of modification indices from a baseline
autoregressive model, and multiple fit indices were used to determine good fit. Results revealed
iv
extensive variability between individuals in the dynamics of anger, impulsivity, and identity
disturbance, although a simple autoregressive model fit data well for six participants. The results
support neither psychodynamic nor behavioral theories of BPD symptom dynamics but imply
that each may account for symptom variation in different individuals with the disorder. Results
also show that a person-specific approach to modeling EMA data is feasible and may support the
development of theories of psychological processes in BPD. This class of methods may also be
useful for the study of psychotherapy process and outcome and may aid in treatment planning,
outcome monitoring, and diagnostic assessment in clinical settings.
v
TABLE OF CONTENTS
List of Tables………………………………………………………………………………..........vi
List of Figures………………………………………………………………………………...….vii
Acknowledgements…………………………………………………………………………..…viii
Chapter1. INTRODUCTION……………………………………………....…...…………………1
Borderline Personality Disorder……….………...………………………………………..1
Heterogeneity within the BPD Category………………………………………………….7
Research Based on Retrospective Self-Report....................................................................9
Research Based on Cross-Sectional Data..........................................................................11
Ecological Momentary Assessment in BPD......................................................................12
Person-Specific Analysis of Intraindividual Variation......................................................16
Chapter 2. METHOD.....................................................................................................................19
Participants.........................................................................................................................19
Procedure...........................................................................................................................22
Materials............................................................................................................................23
Data Preparation.................................................................................................................28
Periodic Trends..................................................................................................................30
Dynamic Factor Models.....................................................................................................32
Chapter 3. RESULTS.....................................................................................................................35
Sample Characteristics.......................................................................................................35
Compliance with Prompted Surveys..................................................................................36
Person-Specific Results.....................................................................................................36
General Results..................................................................................................................55
Chapter 4. DISCUSSION..............................................................................................................58
References......................................................................................................................................71
vi
LIST OF TABLES
Table 1. Demographic and Diagnostic Characteristics of the Current Sample (N = 11)........ 20
Table 2. EMA Compliance and Fit of Dynamic Factor Models................................................37
vii
LIST OF FIGURES
Figure 1. Flow chart of participants through the study..........................................................21
Figure 2. Interpolated control/impulsivity data across 19 days of responses for participant 9.....31
Figure 3. 3-variate autoregressive dynamic factor model (AR[1] model) describing the
relationship between anger, impulsivity, and identity disturbance, simultaneously and at
successive time points, as implemented in LISREL (y submodel)......................................33
Figure 4. Scores on baseline self-report questionnaires for study sample (N = 11)......................35
Figure 5. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #1....................39
Figure 6. Dynamic factor model describing the relationship between anger, impulsivity, and the
residual values of identity disturbance (after accounting for the circadian trend in identity
disturbance), at synchronous and successive time points, for participant #2................................41
Figure 7. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #3....................43
Figure 8. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #4....................44
Figure 9. Dynamic factor model describing the relationship between anger, impulsivity, and the
residual values of identity disturbance (after accounting for the circadian trend in identity
disturbance), at synchronous and successive time points, for participant #5................................46
Figure 10. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #6....................48
Figure 11. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #7....................49
Figure 12. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #8....................51
Figure 13. Dynamic factor model describing the relationship between anger and impulsivity, at
synchronous and successive time points, for participant #9..........................................................53
Figure 14. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #10..................54
Figure 15. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #11..................56
viii
ACKNOWLEDGEMENTS
This dissertation would not have been possible without the help, guidance, and support of
many, many people. I am particularly grateful for the undergraduate students of the Laboratory
for Personality, Psychopathology, and Psychotherapy Research for their ongoing work and their
dedication to this research. Many thanks also go to graduate students involved in the project,
especially Wes Scala, Emily Dowgwillo, Mike Roche, Ross MacLean, and Lauren Szkodny. I
would also like to express my gratitude to my fellow graduate students in the lab, Lori Scott,
Rachel Wasserman, Joe Beeney, Christina Temes, Tracy Clouthier, and Wes Scala, whose
collegiality, good spirits, and curiosity made work a pleasure. Thanks and love also go to my
wonderful graduate cohort, especially Amanda Rabinowitz, Aaron Fisher, and Sam Nordberg.
I am also very grateful for institutional supporters and sponsors of this research project,
including the staff and faculty of the Penn State Psychological Clinic and Survey Research
Center. This project also could not have succeeded without funding support from the
International Psychoanalytical Association, American Psychoanalytic Association, Social
Science Research Institute, and the Research and Graduate Studies Office. I am honored to have
had Aaron Pincus, Reg Adams, and Peter Molenaar on my dissertation committee. Their advice
and guidance have greatly improved this project along the way.
I am most grateful to have worked with Ken Levy for the past several years. He is a
dedicated mentor, a first-rate scholar, an expert supervisor, a model of the scientist-practitioner,
and a good friend. I am immensely thankful for his hard work and dedication, and I am both
proud and humbled to anticipate a career that will embody his example.
Finally, I would like to express my immense love and gratitude to my family. I am
forever indebted to my wife, Carol, whose patience, emotional support, and good sense have
made it possible to do something that often made me a little cranky, distracted, and absent.
Special thanks also go to my mother, who was unexpectedly instrumental in my dissertation
defense. To her, to my father, and to Carol: thank you.
1
Chapter 1
Introduction
Borderline Personality Disorder
Borderline personality disorder (BPD) is a costly, debilitating, and common psychiatric
disorder. Studies estimate its prevalence at about 3% of the general population (Coid, Yang,
Tyrer, Roberts, & Ullrich, 2006; Lenzenweger, Lane, Loranger, & Kessler, 2007; Samuels et al.,
2002; Torgersen, Kringlen, & Cramer, 2001; Trull, Jahng, Tomko, Wood, & Sher, 2010).
Individuals with BPD are also highly prevalent in clinical practice, making up about 10-20% of
psychiatric outpatients (Korzekwa, Dell, Links, Thabane, & Webb, 2008; Zimmerman,
Chelminski, & Young, 2008; Zimmerman, Rothschild, & Chelminski, 2005) and 15-40% of
inpatients (Grilo et al., 1998; Widiger & Frances, 1989; Zimmerman, Chelminski, & Young,
2008). In general, those with BPD consume a large amount of costly mental health services in
comparison with individuals with other disorders (Ansell, Sanislow, McGlashan, & Grilo, 2007;
Bender et al., 2001; Hörz, Zanarini, Frankenburg, Reich, & Fitzmaurice, 2010; Sansone,
Zanarini, & Gaither, 2003; Zanarini, Frankenburg, Khera, & Bleichmar, 2001; Zanarini,
Frankenburg, Hennen, & Silk, 2004). A high proportion of individuals with BPD commit
repeated self-injurious acts (Clarkin, Widiger, Frances, Hurt, & Gilmore, 1983; Gunderson et al.,
2011), and up to 10% eventually commit suicide (Stone, 1993). BPD is frequently comorbid
with other mental disorders (Nurnberg et al., 1991; Skodol et al., 2002; Zanarini et al., 1998,
2004), and the presence of BPD negatively affects outcome in the treatment of depression
(Levenson, Wallace, Fournier, Rucci, & Frank, 2012; Shea, Widiger, & Klein, 1992), anxiety
disorders (Chambless et al., 1992; Cloitre & Koenen, 2001; Feeny, Zoellner, & Foa, 2002;
Mennin & Heimberg, 2000), and eating disorders (Cooper, Coker, & Fleming, 1996). Recent
2
developments in the treatment of BPD itself have given clinicians empirically-supported
psychotherapeutic options to use with this demanding population (Bateman & Fonagy, 2008,
2009; Blum et al., 2008; Clarkin, Levy, Lenzenweger, & Kernberg, 2007; Giesen-Bloo et al.,
2006; Gregory et al., 2008; Linehan et al., 2006; McMain et al., 2009; McMain, Guimond,
Streiner, Cardish, & Links, 2012). Nevertheless, treatment remains a significant challenge, and
only about 50% of individuals respond to any given psychotherapy (Levy, 2008). Given these
facts, it is evident that BPD is a serious public health concern and that understanding this severe
form of psychopathology is important.
Current conceptualizations of BPD suggest a multifaceted disorder involving emotion
dysregulation, impulsive and self-harming behaviors, and instability in relationships and selfconcept. The description of BPD in the American Psychiatric Association’s Diagnostic and
Statistical Manual of Mental Disorders (DSM; American Psychiatric Association, 1994, 2013) is
the most influential description of the disorder. It specifies 9 symptoms of BPD: behavioral
efforts to ward off abandonment by others; intense and unstable relationships; unstable selfimage or sense of self; impulsivity, such as reckless driving, substance abuse, impulsive
spending or sex, and binge eating; recurrent suicidal behavior or gestures, or self-mutilating
behavior without suicidal intent; affective instability or excessive mood reactivity; chronic
feelings of emptiness; inappropriate, intense anger; and transient, stress-related paranoia or
dissociation. Five or more of these symptoms, or criteria, are required to be present in order
make a diagnosis of BPD. In addition, the symptomatic pattern should be deviant from cultural
norms and expectations, inflexible and pervasive across situations, distressing or impairing to a
clinically significant degree, temporally stable, and of early developmental origin.
3
Thus, BPD is understood to involve instability in several domains, namely affect,
behavior, and self-concept. However, it is not at all clear how these different aspects of the
disorder interrelate. In particular, the role in the disorder of “identity disturbance,” defined in the
DSM-5 as a “markedly and persistently unstable self-image or sense of self” (American
Psychiatric Association, 2013, p. 663), is not well understood, but a few prominent theories have
been proposed from varying theoretical perspectives. One possibility is that individuals with
BPD have a temperamental vulnerability to emotional extremes, which makes impulsive and
self-harming behaviors more likely. This erratic emotional and behavioral experience, in turn,
leads to relationship instability and also makes it difficult for these individuals to construct a
coherent view of themselves when they reflect on their experiences. This view is most consistent
with behavioral or “biosocial” models of the disorder that hold emotion regulation difficulties to
be the core symptom of the disorder (Crowell, Beauchaine, & Linehan, 2009; Linehan, 1993;
Zanarini & Frankenburg, 2007). From this perspective, the identity disturbance observed in
individuals with BPD is secondary to emotional vulnerability and behavioral dysregulation and
would not serve as a good prospective predictor of impulsivity, emotionality, and other important
features of the disorder when other symptoms are taken into account. Accordingly, treatments
developed with this model of psychopathology intervene to help the individual modulate affect
and regulate behavior, assuming that change in the rest of the individual’s symptoms will follow
(e.g., Linehan, 1993).
In contrast, psychoanalytic and psychodynamic theories of BPD, most notably those from
an object-relations framework, posit that having an incoherent and diffuse sense of self decreases
the ability of those with BPD to consistently use positive self-relevant thoughts to modulate their
emotional reaction to even minor stressors. Thus, the lack of a coherent self-concept is thought
4
to result in extreme affect, erratic behavior, and greater vulnerability to self-harm (Bender &
Skodol, 2007; Clarkin, Yeomans, & Kernberg, 2006; Fonagy & Target, 2006; Levy et al., 2006).
These theories regard the emotional and behavioral symptoms of the disorder as secondary to
identity disturbance in individuals with BPD. In these models, both self-concept instability and
erratic affective experience would be important prospective predictors of impulsivity.
Psychoanalytic treatments for BPD, which are based on this general model of the disorder,
intervene to help the individual develop a more coherent sense of self, which then theoretically
leads to more adaptive emotional and behavioral patterns (e.g., Bateman & Fonagy, 2004;
Clarkin, Yeomans, & Kernberg, 2006). In short, psychodynamic and biosocial theories posit an
opposite role for identity disturbance in the overall mechanism of BPD pathology: one as the
cause, and the other as the result, of the affective and behavioral dysregulation that is typical of
the disorder.
Complicating these models is a possible distinction between identity disturbance as a
stable and lasting individual-differences variable and identity disturbance as a subjective sense of
self-incoherence that can vary over time within an individual. The former variable is described
in the psychodynamic literature as a consequence of temperamental vulnerability to aggression,
the use of immature defensive strategies, and insecure attachment (e.g., Fonagy, Target, Gergely,
Allen, & Bateman, 2003; Kernberg, 1975; Levy, Meehan, Weber, Reynoso, & Clarkin, 2005),
which produce a chaotic and incoherent representation of self. This self-structure is thought to
be relatively permanent and to leave the individual vulnerable to BPD, and its presence is
generally assessed via clinical observation of conflicting self-representations (Kernberg, 1984).
At the same time, many clinical authors and researchers have also suggested that
individuals with BPD can experience a temporary clarity in their self-concept at certain times.
5
For example, Fonagy and colleagues propose that “a stable sense of self is illusorily achieved”
for a short time in those with BPD when their maladaptive defenses allow them to disavow
aspects of their self-concept that are painful (Fonagy et al., 2004, p. 359). The authors also point
out that these temporary states of self-clarity eventually exact a heavy price, as these defenses
may bring about negative interpersonal consequences. Likewise, Kernberg (1984) theorizes that
individuals with BPD tend to keep conflicting representations of self separate, leading to
oscillations between contradictory self-concepts that may temporarily be held with rigid
conviction until outside reality disrupts the active self-representation. In a study of clinical
observations of identity disturbance, Wilkinson-Ryan and Westen (2000) found that a major
component of the construct, as observed by clinicians in everyday practice, was “a tendency to
make temporary hyperinvestments in roles, value systems, world views, and relationships that
ultimately break down” (p. 529). It is reasonable to postulate that these temporary
hyperinvestments in self-defining roles and values might temporarily lend an individual a
subjective sense of clarity in their self-concept, in line with the views advanced by clinical
writers. These accounts converge in the idea that this seemingly coherent sense of self is
maladaptive, as it reflects an unstable and unrealistic self-representation that may lead to poorly
modulated behavior.
In social-personality psychology, a similar plurality of self-structure variables has
developed. The consciously accessible, subjective sense of clarity and stability in the selfconcept is generally referred to as “self-concept clarity.” This variable is usually operationalized
via self-report (Campbell et al., 1996) and is very similar to identity disturbance; in fact,
measures of self-concept clarity show some validity as measures of identity disturbance in BPD
(Pollock et al., 2001; Roepke et al., 2011). Although originally thought of as a stable individual-
6
differences variable, self-concept clarity has been shown to vary from day to day within
individuals and in dynamic relationship with life events and mood (Ayduk, Gyurak, & Luerssen,
2009; Lavallee & Campbell, 1995; Nezlek & Plesko, 2001), suggesting that fluctuations in this
variable may be important. In contrast, structural and more enduring aspects of stability in the
self-concept structure are often assessed through laboratory tasks, such as card-sort procedures
thought to evoke representations of implicit self-structure (e.g., Linville, 1987).
At least one research study suggests that these trait-like and state-like self-concept
variables have a dynamic relationship similar to the psychodynamic models described above.
Boyce (2008) has found that individuals with low trait self-concept clarity (i.e., high trait-level
identity disturbance) could be temporarily induced to experience a state of high self-concept
clarity. When this happened, they showed a compartmentalized structure of self-representations
– the hypothesized organization underlying borderline personality pathology, according to
Kernberg (1984) – as assessed by a laboratory card-sorting measure. In contrast, when
individuals with high trait self-concept clarity were in a state of high self-concept clarity, they
showed an integrated self-structure, which is associated with resilience (Showers & Kling,
1996). Thus, it is possible that momentary states of low subjective identity disturbance, or high
self-concept clarity, may leave individuals with BPD vulnerable to negative outcomes by
distorting their representations of themselves and of others.
Despite the prominent standing of these different models of BPD and the possible
theoretical importance of identity disturbance in BPD, direct empirical tests of the proposed
processes within the disorder are scant, and the essential nature of the disorder is still poorly
understood. There is also no indirect evidence for the suitability of either psychodynamic or
biosocial models of the disorder in terms of differential efficacy or effectiveness of
7
psychotherapies developed in accordance with them (Levy, 2008). Limitations of the DSM
definition of BPD and of the methodology of most of the research to date leave open these
questions and will be described below. These limitations are high levels of heterogeneity within
DSM-IV BPD and a research literature based primarily on retrospective, cross-sectional data.
Heterogeneity within the BPD Category
First, it is unclear whether Borderline Personality Disorder as defined in the DSM has a
single essential nature. Latent class analyses using the BPD criterion set generally show
evidence for a single class of individuals who can be diagnosed with the disorder among large
samples (Clifton & Pilkonis, 2007; Fossati et al., 1999; Shevlin, Dorahy, Adamson, & Murphy,
2007), although the distinction between those who have the disorder and those who do not is
difficult to draw given the frequency of many of the BPD symptoms among those who do not
formally qualify for the diagnosis (Shevlin et al., 2007). Nevertheless, there is abundant
evidence for heterogeneity among individuals with BPD in studies that take variables outside the
DSM system into account. Many variables have been shown to delineate BPD subtypes,
including specific patterns of interpersonal problems (Leihener et al., 2003; Salzer et al., 2013;
Wright et al., 2013), dependency and self-criticism (Kopala-Sibley et al., 2012), psychopathic
traits (Newhill, Vaughn, & DeLisi, 2010), effortful control (Hoermann, Clarkin, Hull, & Levy,
2005), personality variables (Bradley, Conklin, & Westen, 2005; Tramantano, Javier, & Colon,
2003), attachment style (Levy, Meehan, Weber, Reynoso, & Clarkin, 2005), demographic
characteristics (Johnson et al., 2003; Stevenson, Meares, & Comerford, 2003), other cooccurring disorders (Ferrer et al., 2008) and various combinations of these factors (Digre et al.,
2009).
8
This heterogeneity is not surprising given the polythetic nature of the DSM diagnostic
algorithm for BPD. No symptom must be present for the diagnosis to be made, and instead the
presence of any five is considered sufficient. This approach to diagnosis was the result of a
decision by authors of the third edition of the manual to group disorders by descriptive features,
rather than by putative etiological factors or underlying mechanisms, in order to achieve
acceptability by researchers and clinicians of differing theoretical backgrounds (Spitzer, 2001).
Thus, for example, neither emotional variability nor identity disturbance must be part of the
overall symptom picture of someone with BPD. The result of the American Psychiatric
Association’s decision is that there are 256 unique ways for an individual to meet criteria for
BPD within the DSM diagnostic system, and in fact, many of these criterion combinations
actually occur in practice. For example, Johansen, Karterud, Pedersen, Gude, and Falkum
(2004) found 136 different criterion patterns among 252 individuals with DSM-IV BPD in
Norwegian day hospitals. This variety within the diagnostic category seems to be important:
individuals with the disorder who meet differing numbers of BPD criteria (Asnaani, Chelminski,
Young, & Zimmerman, 2007) or different criterion patterns among the 256 combinations
(Cooper, Balsis, & Zimmerman, 2010) have been shown to differ in terms of their levels of
pathology and functioning. In all, then, it is evident that a diagnosis of BPD does not hold the
same implications for every individual to which it is applied. This high degree of heterogeneity
raises the possibility that BPD symptoms might interact in different ways in different individuals
and that different theories affording primacy to one factor or another might apply to different
subgroups of individuals with the diagnosis.
9
Research Based on Retrospective Self-Report
Scientific knowledge about the mechanisms of BPD is also constrained by limitations of
research design. One important such limitation of research into the validity of various models of
BPD is that much of it is based on retrospective self-report. In this method of data collection,
individuals are asked to recall and summarize their past experience and to characterize their
general thoughts, feelings, and behaviors, most often in the context of an unstructured or semistructured interview or using a questionnaire measure. This strategy is cost-effective and easy to
implement, and it also mirrors the typical diagnostic encounter in clinical practice. However,
retrospective recall for psychological experiences can be biased and inaccurate (Bolger, Davis, &
Rafaeli, 2003; Ebner-Priemer & Trull, 2009; Fahrenberg, Myrtek, Pawlik, & Perrez, 2007;
Ptacek, Smith, Espe, & Raffety, 1994; Solhan et al., 2009; Shiffman, Stone, & Hufford, 2008;
Smith, Leffingwell, & Ptacek, 1999; Todd et al., 2003). Interviewees and survey respondents are
often unable to recall with accuracy the frequency, sequence, or quality of their experiences.
This is true for the sort of inner emotional experiences that characterize many of the symptoms in
the DSM (Robinson & Clore, 2002) but also holds for relatively objective events, such as the
number of cigarettes smoked within a certain period (Shiffman, 2009). More complex
information, such as the context in which an experience occurs or the individual’s reasons for
carrying out a certain behavior, is also notably difficult to relate to a clinician or researcher with
accuracy (e.g., Todd et al., 2003; Todd et al., 2005). Indeed, linkages between symptoms and
their context and between behaviors and their motivations are often the focus of DSM criteria
(e.g., “frantic efforts to avoid… abandonment” or “stress-related paranoid ideation” in the case
of BPD; American Psychiatric Association, 2013, p. 663) and of interview questions and
questionnaire items based on the DSM criteria. Thus, it is questionable whether retrospective
10
interview-based or questionnaire data result in an accurate picture of an individual’s
psychological problems as they actually occur in the person’s life.
Difficulties in retrospective recall on questionnaires and in interviews may be particularly
problematic for self-report data from individuals with BPD due to their chaotic, rapidly shifting
inner experience. Ebner-Priemer and colleagues (2006), for example, gathered the quality and
intensity of self-reported emotional experience in individuals with BPD and healthy control
participants every 10-20 minutes over a 24-hour period. Afterwards, they asked participants to
rate the intensity of various positive and negative emotions. Individuals with BPD showed a
negative recall bias: they retrospectively recalled their negative emotions as more intense, and
their positive emotions as less intense, than they had indicated at the time. Importantly, this
tendency was not due to comorbid major depressive disorder or post-traumatic stress disorder, as
it was present regardless of these other diagnoses. Other investigations comparing retrospective
recall of affective states and immediate ratings of these states in individuals with BPD have
revealed similarly low correspondence between the two (Links, Heisel, & Garland, 2003; Solhan,
Trull, Jahng, & Wood, 2009).
Other recent research suggests that inconsistency in endorsement of questionnaire items
may also be particularly pronounced in BPD. Hopwood and colleagues (Hopwood & Morey,
2007; Hopwood et al., 2009; Hopwood & Zanarini, 2010) have found that individuals with BPD
or clinically significant BPD features tend to show inconsistent response patterns within
personality questionnaires relative to individuals with other personality disorders, even when
controlling for the presence of a major depressive disorder diagnosis. The authors note that these
findings may reflect a defensive response pattern among individuals with BPD, actual trait
instability, or other factors.
11
Finally, as a recent review paper notes (Santangelo, Bohus, & Ebner-Priemer, in press),
retrospective recall is a particularly problematic data-collection strategy when dealing with
unstable psychiatric symptoms, as concordance between retrospective and immediate ratings for
this type of experience is particularly low. Several symptoms of BPD are characterized by
particular instability over time, especially behavioral manifestations of the disorder such as selfinjury and frantic efforts to avoid abandonment (McGlashan et al., 2005; Shea et al., 2002).
These experiences often have acute environmental determinants (Gunderson et al., 2003),
whereas other symptoms, and indeed functional impairment, are more stable (Skodol et al.,
2005) and may be easier for the individual to summarize with accuracy. In addition, because
many symptoms of BPD involve variability, change, and instability in other variables (e.g., affect
or self-concept), retrospectively reporting on them involves rating changes or differences
between separate experiences rather than rating a single experience (Ebner-Priemer & Trull,
2009). Research has shown that retrospective ratings of change and instability are more prone to
inaccuracy than ratings of mean values or general frequencies (Ebner-Priemer, Bohus, & Kuo,
2007, as cited in Ebner-Priemer & Trull, 2009; Stone, Broderick, Shiffman, & Schwartz, 2004).
For all of these reasons, it is not clear that an empirical literature that relies heavily on
retrospective report can allow researchers to come to an accurate picture of the dynamics and
mechanisms of BPD.
Research Based on Cross-Sectional Data
A final limitation of previous research into the relationships between different BPD
symptoms is that most findings are based on cross-sectional data, reflecting ratings of BPD
symptoms across many individuals at one point in time. For example, numerous factor analyses
and correlational studies of BPD criteria have been conducted, but the results vary with respect
12
to the relationships between affective lability, instability in self-concept, and behavioral
disturbances. Some research suggests that instability of affect and identity show a closer
relationship with each other than with other symptoms, such as impulsivity and suicidality
(Benazzi, 2006), whereas other studies suggest that self-concept instability is related to both
impulsivity and affective instability to a greater degree than the latter two relate to each other
(Becker, McGlashan, & Grilo, 2006). Still other studies are not able to discriminate between the
strength of these relationships (Blais, Hilsenroth, & Castlebury, 1997; Fossati et al., 1999;
Johansen et al., 2004; Koenigsberg et al., 2001; Sanislow et al., 2002). One recent crosssectional study directly tested the unique contributions of different constructs to BPD symptoms.
Cheavens, Strunk, and Chriki (2012) measured self-reported emotion dysregulation,
interpersonal problems, and sense of self among college students and Internet users and found
that only emotion dysregulation uniquely predicted variance in self-reported BPD symptoms.
Because cross-sectional research consists only of the analysis of responses at a single
point in time, it cannot illuminate how various phenomena unfold and interrelate over time
within individuals. The lion’s share of research into the nature of BPD has been cross-sectional,
and accordingly, the temporal interplay between different symptoms of BPD has largely been
neglected. Moment-to-moment variability in emotion, self-concept, and impulsivity is especially
important to understand in BPD, as at least one empirical account suggests that diurnal variance
in these symptoms tends to be more extreme in BPD than either day-to-day variance within
individuals or interindividual variability (Nisenbaum, Links, Eynan, & Heisel, 2010).
Ecological Momentary Assessment in BPD
One method for circumventing both the biases of retrospective recall and the limitations
of cross-sectional data is ecological momentary assessment (EMA). In this type of assessment,
13
individuals complete repeated, unobtrusive records of their experiences, close in time to the
events that are being rated. Thus, EMA both minimizes error due to recall bias and allows for
the longitudinal analysis of multivariate psychological processes (Bolger et al., 2003; Shiffman
et al., 2008; Conner, Tennen, Fleeson, & Barrett, 2009). As such, it is a promising set of
methods for studying psychopathology (Myin-Germeys et al., 2009), and this advantage is
particularly helpful for BPD, which by definition involves variability over time in several
presumably interrelated symptoms.
The study of affective variability in BPD using intensive, repeated paper-and-pencil
measures of mood has a modest history, especially in inpatient settings (e.g., Cowdry, Gardner,
O’Leary, Leibenluft, & Rubinow, 1991; Stein, 1995). In recent years, EMA studies have taken
advantage of the proliferation of handheld digital organizers and mobile “smartphone” devices to
study these processes in outpatients and community samples. These electronic devices have
several advantages: they allow researchers to deliver random prompts to participants to enter
self-report data, reducing practice and habituation effects. They also allow for responses to be
electronically time-stamped, which increases participant adherence to the survey regimen and
helps researchers model responses with greater temporal fidelity. EMA studies of BPD have
grown more common in the past decade (for reviews of previous studies, see Nica & Links, 2009
and Santangelo, Bohus, & Ebner-Priemer, in press), and this method is generally seen as
promising due to improvements in the quality, availability, and price of mobile devices and to
advances in approaches to modeling the resulting data (e.g., Miller, 2012).
Most of the studies using EMA in BPD have focused on the nature of affective
dysregulation in the disorder. For example, Trull and colleagues (2008) used electronic diaries
to prompt outpatients with BPD and outpatients with major depressive disorder (MDD), but
14
without BPD, to complete records of their mood at random intervals six times per day for one
month. They found similar mean levels of positive and negative affect across the two groups.
However, overall variability of positive and negative affect was greater in the BPD group, as was
the instability (differences between successive ratings) of hostility, fear, and sadness. EbnerPriemer and colleagues (2007) found that individuals with BPD reported more complex emotions
(i.e., the presence of more than one emotion at a time), with greater intensity of complex
negative emotions, than healthy control participants. In general, the best-conducted studies
suggest that individuals with BPD show greater instability in affect than healthy controls or
individuals with MDD (Santangelo, Bohus, & Ebner-Priemer, in press).
Other symptoms of BPD are less frequently studied but include instability in
interpersonal behavior (Russell, Moskowitz, Zuroff, Sookman, & Paris, 2007), dissociation and
paranoia (Glaser, van Os, Thewissen, & Myin-Germeys, 2010; Stiglmayr et al., 2008),
suicidality (Links et al., 2007; Links, Eynan, Heisel, & Nisenbaum, 2008) and rejection and rage
(Berenson, Downey, Rafaeli, Coifman, & Paquin, 2011). A few studies have also examined the
contexts in which BPD symptoms arise. For example, Stiglmayr and colleagues (2008) linked
dissociative symptoms to states of stress in BPD patients as well as in clinical and healthy
control participants; the BPD group experienced more stress-linked dissociation than either
control group. Likewise, Glaser et al. (2010) found that stress was related to psychotic
experiences in individuals with BPD to a greater extent than those with active psychosis, cluster
C personality disorders, or healthy controls. These studies have thus provided support for the
validity of the “stress-related paranoid ideation or severe dissociative symptoms” criterion of
DSM-5 BPD (American Psychiatric Association, 2013, p. 663). Berenson and colleagues (2011)
have used EMA methods to relate momentary feelings of rage (which the authors conceptualized
15
as the experience of BPD criterion of intense, uncontrolled anger) and perceptions of rejection in
individuals from the community with BPD and healthy control participants. Results showed that
the experience of intense rage in the context of perceived rejection was much stronger in the
BPD group than in the healthy control group. Although the EMA data did not allow the authors
to determine the directionality of the rejection-rage contingency, a priming task in the laboratory
with the same sample suggested that the mental activation of rejection-related constructs
triggered cognitions associated with rage.
Thus, a number of studies have examined symptoms of BPD using EMA methods, which
have helped to validate clinical observations that the disorder involves affective instability,
extreme anger responses to interpersonal slights, and dissociation and paranoia under distress.
Further studies have elucidated the naturalistic context of the emotional and behavioral
experiences associated with BPD. To date, however, no studies have examined impulsivity or
identity disturbance in BPD using EMA.1 In addition, very few studies have examined the
relationships between different BPD symptoms, although some EMA studies have examined the
social, environmental, and psychological contexts that give rise to a single symptom.
In addition, the extant EMA studies of borderline personality pathology have examined
groupwise hypotheses, comparing a BPD group with a group of healthy control participants or a
clinical group with other diagnoses. These studies generally use multilevel modeling,
generalized linear modeling, or another approach to modeling variation in the symptom or
process of interest. In the most common approach, mean levels of these variables are compared
across groups in order to test hypotheses about the uniqueness of the borderline personality
1
In their review, Santangelo, Bohus, & Ebner-Priemer (in press) cite two EMA studies examining variability in selfesteem and its relationship to BPD features (Tolpin, Gunthert, Cohen, & O’Neill, 2004; Zeigler-Hill & Abraham,
2006). However, variability of self-esteem is not very consistent with the description of identity disturbance in
DSM-IV or DSM-5, where it is described as involving shifts in goals, values, roles, sexual identity, aspirations, and
other non-evaluative aspects of self and identity.
16
syndrome. However, this groupwise approach to the analysis of EMA data does not account
very well for the high levels of potential, and actual, heterogeneity within the class of individuals
with BPD. Multilevel modeling using longitudinal data, for example, may account for individual
variation around certain parameters of interest, but this variation is generally understood as
quantitative variation around a mean value. Thus, this approach does not allow for qualitative
variation in the characteristic processes of BPD (Sterba & Bauer, 2010). Nevertheless, despite
this pervasive focus on groupwise hypotheses and interindividual variation, it is clear that
intensive longitudinal data collected through EMA have the potential to illuminate heterogeneity
within diagnostic categories through the examination of intraindividual variation in BPD
symptoms as they occur over time.
Person-specific analysis of intraindividual variation
In recent years, it has become more apparent that the relationships between variables
within individuals over time are not necessarily the same as the relationships between variables
when these constructs are measured in several persons at one point in time (Kenny, Kashy, &
Bolger, 1998; Tennen & Affleck, 2002). For example, it has been shown that the personality
factors of Neuroticism and Conscientiousness are generally negatively related in factor analyses
of interindividual variation (Mount, Barrick, Scullen, & Rounds, 2005), but they tend to be
positively correlated over time within individuals (Beckmann, Wood, and Minbashian, 2010).
Similar divergence has been shown for the interpersonal dimensions of agency and communion,
which are unrelated in cross-sectional studies but positively associated over time within
individuals (Roche, Pincus, Hyde, Conroy, & Ram, 2013). Thus, the dynamics of interindividual
(or intergroup) variation, where patterns of behaviors, emotions, or other responses are compared
between persons, and the structure of intraindividual variation, where different responses are
17
compared over time within persons, are very often distinct (Borsboom, Mellenbergh, & van
Heerden, 2003; Molenaar, 2004).
Importantly, it is also often the case that substantial heterogeneity exists between
individuals in the intraindividual covariation of variables over time, so that the nomothetic
models that describe individual or group differences do not generally hold implications for the
processes that obtain for a single person, even when the same sample is used for both types of
analysis. It has been shown, for example, that the intraindividual covariance structure of
responses to questionnaires that are designed to assess the Five-Factor Model of personality does
not usually resemble the five factors that are derived from the analysis of their interindividual
variation, and substantial differences between subjects have been found in terms of the number
and composition of the factors required to describe the intraindividual data (Hamaker, Dolan, &
Molenaar, 2005; Molenaar & Campbell, 2009). The same has been shown of items on the
PANAS, a common and robustly replicated measure of positive and negative affect (Rovine,
Molenaar, & Corneal, 1999), and in simulated behavioral genetics data (Molenaar, Huizenga, &
Nesselroade, 2003). Empirical studies (e.g., Hamaker et al., 2005) and a mathematical proof
(Kelderman & Molenaar, 2007) demonstrate that a single, stable, well-fitting model of betweensubjects variation can be derived that does not apply to the intraindividual variation of any of the
individuals in the sample.
Because previous investigations of BPD symptom dynamics using EMA have relied on
between-persons and between-groups approaches to data analysis, they have not revealed the
extent to which these relationships vary across individuals. A person-specific approach to
modeling EMA data would allow for the examination of this heterogeneity and would still
produce meaningful tests of the models put forth in the clinical literature. The current study
18
seeks to apply a person-specific modeling approach (dynamic factor modeling; Molenaar, 1985)
to EMA ratings of disturbances in affect, behavior, and identity among individuals with BPD and
among clinical control participants. Within each individual, competing hypotheses are tested
that derive from clinical theories of BPD from the psychodynamic and behavioral/biosocial
traditions. These hypotheses are: 1) identity disturbance will prospectively predict later selfreported affective and behavioral dysregulation, as predicted by psychodynamic theory and 2)
affective and behavioral dysregulation will predict later identity disturbance, as “biosocial”
theory predicts. In addition, the person-specific approach and diagnostically mixed sample
allows for an exploration of two additional hypotheses: 3) that the relationships among identity
disturbance, anger, and impulsivity will not uniformly resemble hypotheses 1 and 2 but will
instead vary between individuals; and 4) the relationships between these symptoms will be
unique to individuals with DSM-IV borderline personality disorder and not present in individuals
with other DSM-IV disorders.
19
Chapter 2
Method
Participants
Participants were 11 adult individuals participating in outpatient treatment at the Penn
State Psychological Clinic, a community mental health center and the primary training clinic for
the graduate program in clinical psychology at the Pennsylvania State University. In order to be
eligible to participate in the study, participants had to be aged at least 18 years; could not be
diagnosed with schizophrenia, schizoaffective disorder, bipolar I disorder, delusional disorder,
delirium, dementia, amnestic disorder, cognitive disorder NOS, current substance dependence,
mental retardation, or borderline intellectual disability; and had to self-report normal or
corrected-to-normal vision (in order to read questionnaires on the smartphone’s LCD screen).
Characteristics of the participants can be found in Table 1.
Participants were recruited from two separate studies with slightly different diagnostic
inclusion criteria. In the first study, clinicians who were conducting routine diagnostic
interviews with individuals presenting for treatment were asked to mention a study about
“personality in daily life” to clients. If a client indicated interest in the study, he or she could
sign a consent form giving research staff permission to access their diagnostic interview data to
verify their eligibility based on diagnosis and to contact them for additional eligibility screening
over the telephone. The diagnostic inclusion criterion for this study, in addition to the diagnostic
exclusion criteria above, was the presence at intake of either BPD or a DSM-IV anxiety disorder.
The second group of participants was recruited from among outpatient clients from the same
clinic who had participated in previous studies and had asked to be contacted about further
20
Table 1
Demographic and Diagnostic Characteristics of the Current Sample (N = 11)
M
SD
37.0
11.48
Global Assessment of Functioning, baseline
56.82
13.47
IPDE BPD dimensional score, baseline
8.45
6.06
GAPD score, baseline
2.62
0.48
SCCS score, baseline
2.98
0.82
MSI-BPD score, baseline
5.27
2.10
BSCS score, baseline
2.50
0.50
ALS total score, baseline
2.47
0.81
Characteristic
N
%
Age
Female
9
81.8
African-American
2
18.2
Asian
0
0
American Indian/ Alaska Native
1
9.1
Caucasian
9
81.8
Hispanic or Latino
1
9.1
Native Hawaiian/ Pacific Islander
0
0
Single
6
54.5
Dating
2
18.2
Married
2
18.2
Divorced
2
18.2
Separated
0
0
Widowed
0
0
Bisexual
2
18.2
Heterosexual
9
81.8
Homosexual
0
0
1
Ethnicity
Marital Status
2
Sexual Orientation
1
Participants could list more than one ethnicity.
Participants could list more than one marital status (e.g., divorced and dating).
Note. IPDE = International Personality Disorder Examination; BPD = Borderline Personality
Disorder; GAPD = General Assessment of Personality Disorder; SCCS = Self-Concept Clarity
Scale; MSI-BPD = McLean Screening Inventory for BPD; BSCS = Brief Self-Control Scale;
ALS = Affective Lability Scales.
2
21
research. These individuals were called and, if they expressed interest, were screened for the
study via telephone in the same way as the first group. Diagnostic eligibility for these previous
studies was based on the presence of either BPD or a DSM-IV MDD or dysthymia diagnosis, in
addition to the same exclusion criteria as above. Thus, participants were eligible for the current
study if they met criteria for either BPD or a unipolar mood or anxiety disorder (but not BPD).
Figure 1 shows the flow of the 25 recruited individuals through the current study. In all, 11
eligible individuals agreed to participate and submitted enough EMA data to be analyzed.
6 contacted from previous
research studies
19 referred from clinic intakes
4 declined participation
15 screened
6 ineligible
- Schizophrenia (1)
- active Alcohol Dependence (1)
- Mental Retardation (1)
- No BPD or anxiety disorder (3)
9 eligible
6 screened
6 eligible
2 could not be reached
7 entered study
6 entered study
1
Dropped out
6 completed study
Figure 1. Flow chart of participants through the study.
1
5 completed study
22
Procedure
After being recruited as above, participants completed baseline questionnaire measures
and received extensive training in the use of the smartphone device. For 21 days following this
laboratory session, participants completed three kinds of surveys on their smartphones: surveys
in response to block-randomized auditory prompts from the device (“prompted surveys”),
surveys after every interpersonal interaction lasting at least three minutes, and surveys within an
hour of going to sleep for the night. The three types of surveys varied in their content, and the
current report only concerns the prompted surveys.
Before the 21-day EMA sampling period, participants chose between three 12-hour
intervals for daily prompts: 6:00am to 6:00pm, 8:00am to 8:00pm, or 10:00am to 10:00pm.
Participants received 6 prompts per day to complete surveys, each of which occurred at a
pseudo-random time within a 2-hour block during the overall 12-hour interval. Participants were
instructed to complete a prompted survey immediately upon hearing the prompt, although they
could decline to do so if they were engaged in an activity that would make it dangerous to attend
to the survey (for example, driving a car). In this case, they had the option of completing a
survey at a later time. Each survey was time-stamped by the date and time it was begun. Once a
survey was begun, the participant had 10 minutes to complete it before the phone “timed out.”
Surveys were transmitted to a central, encrypted server as soon as they were completed, as long
as the participant had access to a wireless data network. Compliance with all surveys was
monitored by the research team, and phone calls were placed to participants who had completed
a low number of surveys to resolve any technological problems that might have interfered with
survey completion.
23
At the end of the 21-day EMA sampling period, participants returned to the laboratory to
return the phone and receive payment. They could receive up to $150 in compensation for this
portion of the study: $50 for the baseline questionnaires and training session and $2.35 for each
day of participation, which increased to $100 if they met a threshold of compliance with the
survey protocol (participation for 18 of 21 days, with 85% completion of prompted surveys and
an average of 6 interaction-contingent surveys per day). The study was approved by the
Institutional Review Board of the Pennsylvania State University.
Materials
Diagnostic Interviews. Participants completed semi-structured diagnostic interviews as
part of their clinic intake or as part of previous research studies.
Anxiety Disorders Interview Schedule (ADIS; Brown, DiNardo, & Barlow, 1994).
Those recruited for the study from clinic intakes were diagnosed with Axis I disorders using an
augmented version of the ADIS, a semi-structured interview for the diagnosis of substance use
disorders, anxiety disorders, mood disorders, eating disorders, somatoform disorders, and
psychotic disorders. Within the clinic, interrater reliability of diagnoses based on the ADIS is
good (Cohen’s κ = .673 for mood disorders and .655 for anxiety disorders; Nordberg,
McAleavey, Castonguay, & Levy, 2009).
Structured Clinical Interview for DSM-IVAxis I Disorders, Clinician Version (SCID-I;
First, Spitzer, Gibbon, & Williams, 1997). Individuals recruited from previous research
studies were diagnosed with Axis I disorders using the SCID-I, a semi-structured interview
covering mood disorders, psychotic disorders, substance use disorders, anxiety disorders,
somatoform disorders, and eating disorders. Previous analyses of interrater reliability within this
24
research sample have shown high concordance of SCID-I diagnoses between raters (κ ranges
from .64 to 1.0; Scott, Levy, & Granger, 2013).
International Personality Disorders Examination (IPDE; Loranger, 1999). All
individuals in the sample, whether recruited from clinic intakes or previous research studies,
were diagnosed with personality disorders using the IPDE. The IPDE provides both a criterion
count and a dimensional score for each personality disorder. The dimensional score for BPD
includes both symptoms above criterion level (which receive a score of 2) and symptoms that are
elevated but below the criterion threshold (which receive a score of 1). The interrater reliability
of these diagnoses within the clinic is good (Cohen’s κ = .677 for all Axis II disorders and .679
for BPD; Nordberg et al., 2009). Interrater reliability of these diagnoses within the research
sample (Scott et al., 2013) was excellent (Cohen’s κ values ranged from .71 to 1.0 for all Axis II
disorders; for BPD, κ = .88), as was the reliability of the number of BPD criteria met (ICC = .94)
and of the BPD dimensional scores (ICC = .98).
Baseline Questionnaires. Participants completed baseline questionnaires during the
laboratory session at which they received training in use of the smartphone. They completed
these questionnaires via an online survey-hosting website (surveymonkey.com). The following
questionnaire measures are considered in the current study:
General Assessment of Personality Disorder (GAPD; Morey et al., 2011). The GAPD is
a 65-item self-report questionnaire rated on a 4-point Likert-type scale ranging from 1 (“Strongly
Disagree”) to 4 (“Strongly Agree”). The measure is designed to assess a general level of
personality functioning, as articulated by the DSM-5 Personality Disorders Work Group (e.g.,
Skodol et al., 2011). This construct is not part of the diagnosis of personality disorders in DSM-5
but is contained within the manual’s alternative model for personality disorders (American
25
Psychiatric Association, 2013, Section III). An initial study of the GAPD (Morey et al., 2011)
showed that the 65-item scale could be described with a unidimensional structure and
discriminated between individuals with differing numbers of DSM-IV personality disorders. Its
latent trait dimension also showed strong relationships with the identity disturbance and
impulsivity criteria of BPD, even taking other DSM-IV PD criteria into account. In the current
study, the unweighted mean of responses was used to characterize individuals’ level of
personality functioning. Responses were keyed so that higher numbers indicated a higher level
of personality functioning (i.e., less personality pathology). The internal consistency of the
GAPD in the current sample was excellent (α = 0.97).
Self-Concept Clarity Scale (SCCS; Campbell et al., 1996). The SCCS is a 12-item selfreport measure designed to assess the extent to which a person's self-beliefs are clearly and
confidently defined, internally consistent, and stable. Items are rated on a 5-point Likert-type
scale ranging from “Strongly disagree” to “Strongly agree.” The SCCS has been shown to relate
to the actual consistency of individuals’ self-attribute ratings (Campbell et al., 1996). Although
the measure was developed for use with normal samples (Campbell et al., 1996; Campbell,
Assanand, & Di Paula, 2003), it has also been used as a self-report measure of self-concept
stability among individuals with BPD (Pollock et al., 2001; Roepke et al., 2011). The internal
consistency of the SCCS is good, both in the initial validation study (α = 0.86; Campbell et al.)
and in the current study (α = 0.88).
Affective Lability Scales (ALS; Harvey, Greenberg, & Serper, 1989). The ALS is a 54item self-report instrument designed to measure lability in anxiety, depression, anger, and
hypomania, and labile shifts between anxiety and depression and hypomania and depression.
Items are rated on a 4-point Likert-type scale ranging from “Very characteristic of me, extremely
26
descriptive” to “Very uncharacteristic of me, extremely undescriptive.” Scores on the ALS have
been found to differentiate individuals with BPD from those with bipolar II disorder (Henry et
al., 2001) and depression (Solhan, Trull, Jahng, & Wood, 2009). In addition, large-scale
research shows strong correlations between the ALS and other markers of borderline personality
traits (Ellison & Levy, 2012; Trull, 2001). The individual scales of the ALS are highly
intercorrelated, suggesting that they measure a general tendency toward emotional lability
(Harvey et al., 1989), and the overall scale had high internal consistency in the current sample (α
= 0.99). Scores were coded so that higher values on the 1-4 scale would indicate more affective
lability.
Brief Self-Control Scale (BSCS; Tangney, Baumeister, & Boone, 2004). The BSCS is a
13-item self-report measure of effortful self-control. It has most often been used in non-clinical
samples, where it shows a robust relationship with measures of adjustment, self-esteem,
substance abuse, and interpersonal skills (Tangney, Baumeister, & Boone, 2004) and relates to
real-world outcomes such as academic performance (Duckworth, Tsukayama, & May, 2010) and
criminal and deviant behavior (Holtfreter, Reisig, Piquero, & Piquero, 2010; Reising & Pratt,
2011). Importantly, self-control scores on the BSCS are negatively related to self-reported
symptoms of BPD (Tangney et al., 2004). The internal consistency of the BSCS was good in
Tangney et al. (α = 0.84) and adequate in the current sample (α = 0.68).
McLean Screening Inventory for Borderline Personality Disorder (MSI-BPD;
Zanarini et al., 2003). The MSI-BPD is a 10-item screening questionnaire for BPD that can be
used in interview or self-report contexts. Its items are face-valid inquiries about DSM-IV BPD
criteria (with one additional item relating to interpersonal mistrust), scored on a yes/no basis.
Research generally supports its validity and utility as a screener for BPD (Noblin, Venta, &
27
Sharp, in press; Patel, Sharp, & Fonagy, 2011; Zanarini et al., 2003), although its performance as
a screener in a recent study was mixed (Chanen et al., 2008). The MSI-BPD has also been used
as an outcome measure in a study of psychotherapy for BPD (Williams, Hartstone, & Denson,
2010). Zanarini et al. (2003) found the MSI-BPD had an optimal combination of sensitivity and
specificity with a “caseness” cutoff of 7 endorsed items for adults. The internal consistency of
the MSI-BPD was good in validation studies (α = 0.74 in Zanarini et al., 0.73 in Noblin et al.,
and 0.94 in Patel et al.) and adequate in the current sample (α = 0.66).
Prompted smartphone surveys. Participants completed 6 randomly-prompted
smartphone surveys per day, which contained 46 questions and were designed to take about 5
minutes each. Participants completed smartphone surveys on Motorola Droid Razr devices.
Questions asked about participants’ current affect, symptoms and functioning, repetitive
thoughts, cravings to use substances, self-control capacity, values, thoughts of suicidality and
self-harm, and self-concept and self-concept clarity. The current study concerns questions
relating to 3 symptoms of BPD: anger, impulsivity, and self-concept disturbance.
Anger data was collected using the prompt, “How angry do you feel right now?”, which
was rated on a visual analog scale (VAS) on the touch-sensitive screen of the smartphone.
Participants chose a point along a continuum ranging from “Not at all” to “Extremely,” which
was encoded as an integer value from 0 to 100. Impulsivity was operationalized as participants’
answers to the prompt, “Please rate how you see yourself RIGHT NOW using the following
scales” on a VAS ranging from “Impulsive” to “In control.” These responses were encoded on a
0-100 scale, with 100 representing “in control.” Responses to this question were then reversescored so that higher scores would represent higher levels of impulsivity. Identity disturbance
was measured using the prompt, “RIGHT NOW I have a clear sense of who I am and what I
28
am.” Respondents used a VAS with the anchor points “Strongly disagree” and “Strongly agree,”
and a 0-100 scale was used to encode responses. This item was also reverse scored, so that
higher scores represented higher levels of identity disturbance (lower levels of self-concept
clarity). This item was adapted from the 12-item SCCS (Campbell et al., 1996) to refer to state
(not trait) identity disturbance. A similar modification has been used in a daily diary study to
study fluctuations in self-concept clarity (Ayduk, Gyurak, & Luerssen, 2009).
Single-item measures were used to measure the constructs of interest in the current study
because the primary advantage of EMA studies is their ecological validity relative to laboratorybased retrospective report. For this reason, questionnaire batteries in EMA studies tend to be
brief to allow unobtrusive measurement of phenomena of interest, under the presumption that
larger batteries of questions would disrupt the participant’s natural, lived experience. However,
brief instruments run the twin risks of unreliability and inadequate coverage of important
constructs. As methodologies for ecologically valid assessment have been developed,
researchers have increasingly commented on the tension between maximizing reliability and
content validity of assessments by using psychometrically sound measures and maximizing
ecological validity by minimizing burden to participants (Bolger et al., 2003; Csikszentmihalyi &
Larson, 1987; Shiffman et al., 2008). Relative priority in the current study was given to
maximizing ecological validity.
Data preparation
The dynamic factor analyses used in the current project require data with equal time
intervals between measurement occasions (Molenaar & Rovine, 2011). Because the
smartphones used in the current project prompted participants to complete surveys at random
times within 2-hour blocks, and because participants varied in how quickly they responded to
29
prompts, this requirement was violated. Thus, cubic spline interpolation (Forsythe, Malcolm, &
Moler, 1977) was used to generate evenly-spaced data points to be used in the current analyses.
To do this, the data were divided by participant and then by day. An initial zero point was
chosen for each day’s data for each participant at the quarter hour time (e.g., 8:45am, 9:00am,
9:15am) before that day’s initial survey start timestamp. An endpoint for the interpolation was
then chosen for each participant in order to best cover the group of observation intervals across
days and produce even time intervals for all data points. Separate interpolations were then
performed for each participant’s data on each day, yielding 6 evenly-spaced data points per day
per participant per variable. Days with at least 5 completed surveys were submitted to
interpolation, and a visual comparison of the raw and interpolated data was used to ensure that
the interpolation produced an accurate representation of the original time series. The interval
between interpolated observations varied slightly between participants depending on
participants’ actual timing of survey completion. Cubic spline interpolation was carried out
within the R program, version 2.11.1, using the splines package (Bates & Venables, 2010).
Each participant’s interpolated data across the 3-week diary period was used to derive
estimates of the sequential covariance of identity disturbance, impulsivity, and anger. To do this,
each participant’s data were treated as a 3-variate time series of length 6 (the number of
observations per day) with a number of replications K equal to the number of days. BlockToeplitz covariance matrices were created describing the intraindividual covariation of anger,
impulsivity, and identity disturbance across successive time points (t and t + 1). The blockToeplitz matrices were created under the assumption of “weak stationarity,” meaning that the
means and variances of the data were assumed not to vary depending on location in the time
series, and covariances of variables with equal relative spacing were constant over time. Thus,
30
the covariance matrices contained equal variance and covariance values for the three variables of
interest regardless of their position at time t or at t + 1. This stationarity assumption is necessary
for dynamic factor modeling (Molenaar & Rovine, 2011). The total number of observations
within each factor was calculated as
in order to account for the mean levels estimated when creating the block-Toeplitz covariance
matrix (that is, a loss of K – 1 degrees of freedom). These covariance matrices were then used to
evaluate the fit of the models to each participant’s data.
Periodic Trends
The presence of periodic trends in the time series data is a potential concern, because if
the variables are influenced by such a trend within the time period sampled, estimates of
relationships between variables across time points could be affected. Research has also
demonstrated trends in affective variables within days for individuals with BPD (Nisenbaum et
al., 2010). Periodic trends were checked by a visual inspection of the time series data (see Figure
2) and by a trend analysis. The trend analysis was conducted by regressing the interpolated time
series data for each participant simultaneously onto sine and cosine functions. These functions
were given a period equal to that participant’s window for prompted surveys, or roughly 12
hours (Refinetti, Lissen, & Halberg, 2007). Statistically significant coefficients relating these
sine and/or cosine functions to anger, impulsivity, or self-concept clarity point to the possible
presence of periodic trends in these variables. This method also allows for the derivation of
descriptive statistics regarding each participant’s periodic oscillation in these symptoms. For
example, the amplitude A of the periodic oscillation is given by the formula
31
where b1 and b2 are the unstandardized regression coefficients for the sine and cosine functions,
respectively. The time of day t0 at which the estimated minimum or maximum of the function
occurs is
where t0 is in fractions of a day. Given the frequency of 1 day for these functions, the opposite
extreme is then given by t0 + ½ (Stolwijk, Straatman, & Zielhuis, 1999). If the regression
analysis suggested the presence of circadian trends in the data, a sensitivity analysis was
performed in the model-fitting step by substituting the regression residuals from the trend
analysis for the original time series data and comparing the model fit results with the original
model’s results.
Figure 2. Interpolated control/impulsivity data across 19 days of responses for participant 9. The
trend analysis showed a circadian trend in these data with an amplitude of 7.98 and a maximum
impulsivity about 2/3 of the way between the first and second rating of the day.
32
Dynamic Factor Models
Model fitting was conducted using LISREL, version 8.12 (Jöreskog & Sörbom, 1993).
An autoregressive model with a lag of 1 occasion, or an AR(1) model, was used as the baseline
model for each participant. In the AR(1) models, levels of identity disturbance, impulsivity, and
anger at each time point were used to predict levels of these variables at the next survey.
Correlations between simultaneously rated variables were allowed. If this baseline model did not
show good fit to the data, modification indices were used to guide the addition of additional
parameters (i.e., cross-lagged regression weights between one variable at time t and another
variable at time t + 1). These less-restricted models are termed vector autoregressive models, or
VAR(1) models when a lag of 1 occasion is used. Each model was evaluated via a combination
of fit statistics, namely the Standardized Root Mean Square Residual (SRMR), Root Mean
Square Error of Approximation (RMSEA), Non-Normed Fit Index (NNFI), and Comparative Fit
Index (CFI). These fit statistics were used in place of the chi-square statistic in dynamic factor
modeling because the sequential dependence in the ratings violates the assumption of
independence underlying the distribution of chi-square (Molenaar, 1985). In contrast, these
alternative indices do not have a theoretical sampling distribution and performed well in a small
simulation study (Molenaar & Rovine, 2011). Models were considered adequate fits to the data
when the following conditions were met: RMSEA ≤ 0.08, SRMR ≤ 0.1, NNFI ≥ 0.95, and CFI ≥
0.95. Chi-square values are reported as supplementary information. Figure 3 represents the
baseline AR(1) model as implemented in LISREL.
33
ζ1 (t)
ζ1 (t + 1)
η1 (t)
= η1
ψ21
β41
1
η1 (t + 1)
= η4
1
y1(t)
ψ54
y1(t + 1)
ψ31
ζ2 (t)
ψ64
ζ2 (t + 1)
η2 (t)
= η2
β52
η2 (t + 1)
= η5
ψ32
1
1
y2(t)
ψ65
y2(t + 1)
ζ3 (t)
ζ3 (t + 1)
η3 (t)
= η3
1
y3(t)
β63
η3 (t + 1)
= η6
1
y3(t + 1)
Figure 3. 3-variate autoregressive dynamic factor model (AR[1] model) describing the
relationship between anger, impulsivity, and identity disturbance, simultaneously and at
successive time points, as implemented in LISREL (y submodel).
34
In the dynamic factor modeling approach, the residual variance of variables at time t + 1,
ζt + 1, represents the degree to which an individual’s anger, impulsivity, and self-concept clarity
are predicted by earlier levels of these variables. These “innovation” parameters thus quantify
the level of sequential structure in these three BPD symptoms and can, in extensions of the
current methodology, be used to quantify the effects of interventions upon the idiographic
processes (e.g., Fisher, Newman, & Molenaar, 2011) and can be treated as individual difference
variables in their own right (Ram, Brose, & Molenaar, 2013). In the current study, the
innovation parameters are used as an index of the stability of the processes modeled.
35
Chapter 3
Results
Sample Characteristics
Thirteen participants volunteered for the study, completed baseline measures, and began
the smartphone portion of the protocol. Of these, two participants stopped providing prompted
survey responses shortly into the 3-week smartphone data collection period (after two and six
days of participation, respectively). The remaining 11 participants constitute the sample for the
current study. Their demographic, diagnostic, and baseline psychological characteristics can be
found in Table 1. Figure 4 shows participants’ scores on baseline surveys.
5
Participant
4.5
1
4
2
Score
3.5
3
3
4
2.5
5
2
6
7
1.5
8
1
9
0.5
10
11
0
GAPD
SCCS
BSCS
ALS
Measure
Figure 4. Scores on baseline self-report questionnaires for study sample (N = 11). GAPD =
General Assessment of Personality Disorder; SCCS = Self-Concept Clarity Scale; BSCS = Brief
Self-Control Scale; ALS = Affect Lability Scales. Higher scores on the GAPD, SCCS, and BSCS
and lower scores on the ALS indicate better functioning in these domains.
36
Compliance with Prompted Surveys
Because participants were explicitly told that they could forego responding to a
smartphone prompt if they could not do so safely and could initiate a survey at a later time to
compensate for the omission, compliance was measured as the percentage of the expected 126
prompted surveys (21 days with 6 surveys per day) completed. Surveys completed outside of the
21-day sampling window (for example, before the laboratory visit on the day when the phone
was to be returned) were not counted in this total, because they were not used in data analysis.
Compliance ranged from 61.9% to 100%, with a mean of 89.0% and a median of 90.4%. Table
2 shows compliance for each participant, as well as the fit statistics for each participant’s final
dynamic factor model.
Person-Specific Results
Participant 1.
Diagnosis and baseline characteristics. Participant 1 was a 32-year-old married
Caucasian woman who identified as heterosexual. She was recruited through previous
participation in research and diagnosed via the SCID with major depressive disorder and past
PTSD. She met two BPD criteria (unstable and intense interpersonal relationships and affective
instability) and had an IPDE dimensional score of 5 (with one subthreshold score in addition to
the two BPD criteria) and a Global Assessment of Functioning (GAF) score of 65. Her scores on
baseline self-report measures were as follows: GAPD = 2.83, SCCS = 2.5, BSCS = 2.69, ALS =
2.94. She endorsed 4 items on the MSI-BPD.
Dynamic factor model. Participant 1’s data were described after interpolation by data
points with lags of 6000s. According to the fit indices chosen a priori for the evaluation of
dynamic factor models in LISREL, the autoregressive model (AR[1]) fit the data poorly, χ2(6) =
37
Table 2
EMA Compliance and Fit of Dynamic Factor Models
Participant
Days
χ2
df
p
CFI
NNFI
SRMR
RMSEA
2.36
4
0.67
1.00
1.00
0.043
0.0
1
14
Compliance
(%)
76.2
2
20
97.6
5.75
6
0.45
1.00
1.00
0.055
0.0
3
20
96.8
4.72
4
0.32
1.00
0.99
0.021
0.042
4
19
88.9
6.47
5
0.26
0.99
0.96
0.050
0.056
5
18
100
3.23
5
0.66
1.00
1.00
0.042
0.0
6
10
61.9
2.95
6
0.82
1.00
1.00
0.046
0.0
7
17
88.1
3.75
6
0.71
1.00
1.00
0.038
0.0
8
20
90.4
6.27
6
0.39
1.00
1.00
0.051
0.021
9
19
94.4
0.15
2
0.93
1.00
1.00
0.013
0.0
10
19
84.1
2.33
6
0.89
1.00
1.00
0.036
0.0
11
17
96.8
1.75
6
0.94
1.00
1.00
0.034
0.0
Note. Days = number of days with EMA data usable for dynamic factor analyses; Compliance =
percentage of 126 expected prompted surveys completed; CFI = Comparative Fit Index; NNFI =
Non-Normed Fit Index; SRMR = Standardized Root Mean Residual; RMSEA = Root Mean
Square Error of Approximation.
38
21.98, CFI = 0.80, NNFI = 0.50, SRMR = 0.14, RMSEA = 0.20. Two additional cross-lagged
parameters were required to achieve good fit on all fit indices. In the final model (Figure 5), a
vector autoregressive model (VAR[1] model), anger at time t positively predicted levels of
impulsivity at time t + 1, unstandardized β = 0.17, t = 2.21, p = .03. In addition, identity
disturbance at time t negatively predicted levels of impulsivity at time t + 1, unstandardized β = 0.25, t = 3.16, p < .001. All autoregressive parameters were significant with the exception of
anger, which did not show a relationship between values at successive time points,
unstandardized β = -0.04, t = 0.31, p = .76. In addition, all synchronic correlations between
variables (that is, correlations between variables measured at the same occasion) were significant
with the exception of the relationship between identity disturbance and impulsivity (at time t,
unstandardized ψ = 12.30, t = 1.67, p = .10. The residual variance parameters suggested that
time t ratings accounted for a negligible portion of the variance in anger at time t + 1, 25% of the
variance in impulsivity at time t + 1, and 20% of the variance in identity disturbance at time t +
1.
Periodic trends. No significant periodic trends were observed in this participant’s anger,
impulsivity, or identity disturbance data.
Participant 2.
Diagnosis and baseline characteristics. Participant 2 was a 60-year-old single Caucasian man
who identified as heterosexual. He was recruited through clinic intake and diagnosed via the
ADIS with mood disorder NOS and alcohol abuse on Axis I. On Axis II, he was diagnosed with
narcissistic personality disorder and BPD. He met five criteria for BPD: identity disturbance,
chronic feelings of emptiness, inappropriate anger, unstable interpersonal relationships, and
39
1.00
ANG
t
- 0.04
ANG
t+1
0.40***
- 0.27*
0.43**
0.31**
0.75
0.27*
IMP
t
0.23*
IMP
t+1
0.02
- 0.32**
- 0.20
0.80
ID DIST
t
0.45***
ID DIST
t+1
Figure 5. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #1. Numbers
represent completely standardized maximum likelihood parameter estimates. Dashed lines
indicate nonsignificant parameters retained in the final model (p > .05). ANG = anger; IMP =
impulsivity; ID DIST = identity disturbance. Interval between t and t + 1 is 6000s.
* p < .05. ** p < .01. *** p < .001.
affective instability. He had no other subthreshold scores on the IPDE, giving him a BPD
dimensional score of 10. He endorsed 9 items on the MSI-BPD and had a GAF score of 45. His
scores on baseline self-report measures were as follows: GAPD = 2.29, SCCS = 2.42, BSCS =
2.92, ALS = 2.91.
Dynamic factor model. The autoregressive model fit participant 2’s data well (see Figure
5), so no further modifications were made to the AR(1) model. However, most individual
parameter estimates in the model were not significantly different from zero, including the
autoregressive relationships between impulsivity and identity disturbance and these variables at
40
the next measurement occasion (an interval of 6840s separated time points for this participant).
Anger at time t did predict levels of anger at the next time point, unstandardized β = 0.25, t =
2.68, p = .009. Anger was also positively related to impulsivity in synchronic ratings, ψ = 72.55,
t = 3.02, p = .003.
Periodic trends. A significant periodic trend was found for participant 2’s identity
disturbance data over the 20 days of the diary period, bcos = 2.79, t = 2.30, p = 0.02. The trend
analysis suggested that this participant’s identity disturbance peaked roughly halfway through
the day and that the oscillation had a fairly small amplitude (2.95 on a 0-100 scale). A sensitivity
analysis of this participant’s data using the residuals from the trend analysis suggested that the
above dynamic factor model did not fit as well after taking this trend into account, χ2(6) = 7.81, p
= 0.25, CFI = 0.94, NNFI = 0.85, SRMR = 0.060, RMSEA = 0.055. The greatest modification
index from this analysis was the lagged relationship between identity disturbance (at time t) and
anger at time t + 1. Adding this parameter to the model improved fit to acceptable levels, χ2(6) =
7.81, p = 0.25, CFI = 0.94, NNFI = 0.85, SRMR = 0.060, RMSEA = 0.055, although the added
parameter was not significantly different from zero, β = 0.08, t = 1.65, p = .10. Time t variance
accounted for a relatively small portion of t + 1 variance in anger, impulsivity, and residual
identity disturbance ratings (6%, 2%, and 1%, respectively). Figure 6 shows participant 2’s final
model.
Participant 3.
Diagnosis and baseline characteristics. Participant 3 was a 46-year-old married
Caucasian woman who identified as heterosexual. She was recruited through clinic intake and
diagnosed with BPD without a diagnosis on Axis I. She met six BPD criteria: unstable and
intense interpersonal relationships, identity disturbance, chronic suicidality, affective instability,
41
0.94
ANG
t
ANG
t+1
0.26**
0.26**
0.32**
0.16
0.11
0.98
IMP
t
0.14
IMP
t+1
0.15
0.13
0.16
0.99
ID DIST
resid.
t
0.04
ID DIST
resid.
t+1
Figure 6. Dynamic factor model describing the relationship between anger, impulsivity, and the
residual values of identity disturbance (after accounting for the circadian trend in identity
disturbance), at synchronous and successive time points, for participant #2. Numbers represent
completely standardized maximum likelihood parameter estimates. Dashed lines indicate
nonsignificant parameters retained in the final model (p > .05). ANG = anger; IMP =
impulsivity; ID DIST resid. = residual values of identity disturbance from trend analysis.
Interval between t and t + 1 is 6840s.
* p < .05. ** p < .01. *** p < .001.
feelings of emptiness, and intense anger. She also received subthreshold scores on two other
criteria, giving her a dimensional score of 14 on the IPDE, and she endorsed 9 items on the MSIBPD. Her GAF score was 55. Her scores on baseline self-report measures were as follows:
GAPD = 1.82, SCCS = 1.92, BSCS = 2.08, ALS = 3.19.
Dynamic factor model. The AR(1) model for participant 3’s data showed poor fit on
most fit indices, χ2(6) = 22.32, CFI = 0.95, NNFI = 0.88, SRMR = 0.15, RMSEA = 0.16. A stepby-step modification of the model based on modification indices resulted in the addition of two
42
cross-lagged parameters. This partial VAR(1) model suggested that, for this participant, higher
levels of anger predicted higher levels of impulsivity and higher levels of identity disturbance
6480 seconds later. All variables were positively related in synchronic ratings. In addition, this
model suggested that participant 3 was not consistent in levels of anger, impulsivity, and identity
disturbance from one occasion to the next, as autoregressive relationships were not significant.
Periodic Trends. The trend analysis showed a significant periodic trend in participant 3’s
identity disturbance scores, bsin = - 4.38, t = 3.01, p = 0.003. The trend predicted a peak in
identity disturbance in mid-morning with an amplitude of 4.5 additional units of identity
disturbance on a 0-100 scale. In addition, a significant periodic trend was observed in this
participant’s impulsivity scores, bsin = - 2.76, t = 2.16, p = 0.03, with an amplitude of 3.02 units
and a maximum in impulsivity about one-third of the way through the day. The dynamic factor
model fit to participant 3’s raw data also showed good fit when the residual identity disturbance
and impulsivity values were used, χ2(4) = 2.80, CFI = 1.00, NNFI = 1.00, SRMR = 0.024,
RMSEA = 0.0. Thus, the model was retained (Figure 7). Time t variance accounted for a
negligible portion of variance in participant 3’s t + 1 anger ratings but did account for substantial
portions of t + 1 impulsivity (15% of the variance) and identity disturbance (23% of the
variance).
Participant 4.
Diagnosis and baseline characteristics. Participant 4 was a 34-year-old single Caucasian
woman who identified as bisexual. She was recruited through clinic intake and diagnosed with
recurrent MDD of mild severity, chronic PTSD, and generalized anxiety disorder on Axis I. She
also met three criteria for BPD on Axis II (impulsivity, suicidality or parasuicidality, and chronic
43
1.00
ANG
t
ANG
t+1
0.03
0.70***
0.43***
0.71***
0.59***
0.85
IMP
t
0.60***
IMP
t+1
- 0.05
0.53***
0.37***
0.71***
0.77
ID DIST
t
0.15
ID DIST
t+1
Figure 7. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #3. Numbers
represent completely standardized maximum likelihood parameter estimates. Dashed lines
indicate nonsignificant parameters retained in the final model (p > .05). ANG = anger; IMP =
impulsivity; ID DIST = identity disturbance. Interval between t and t + 1 is 6480s.
* p < .05. ** p < .01. *** p < .001.
feelings of emptiness) and received subthreshold scores on five other criteria, giving her an IPDE
dimensional score of 11. She endorsed 4 items on the MSI-BPD and had a GAF score of 65.
Her scores on baseline self-report measures were: GAPD = 3.22, SCCS = 3.67, BSCS = 2.92,
ALS = 2.06.
Dynamic factor model. The baseline AR(1) model showed inadequate fit to participant
4’s data, χ2(6) = 16.10, CFI = 0.92, NNFI = 0.79, SRMR = 0.070, RMSEA = 0.13. Modification
led the addition of three cross-lagged parameters. In the final model (Figure 8), participant 4’s
initial levels of impulsivity predicted impulsivity 5520 seconds later, and impulsivity also
44
predicted anger at the later time point. Although identity disturbance at time t did not
significantly predict identity disturbance at time t + 1, it did negatively predict later anger. There
was a significant autocorrelative relationship between anger at successive time points. The
model also required a cross-lagged regression of identity disturbance on earlier anger scores,
although this parameter value itself was not significantly different than 0, unstandardized β =
0.13, t = 1.75, p = .08. Time t values accounted for 20%, 4%, and 4% of time t + 1 variance in
anger, impulsivity, and identity disturbance, respectively.
0.80
ANG
t
ANG
t+1
0.23*
0.33***
0.28*
0.37***
0.96
IMP
t
- 0.04
0.51***
IMP
t+1
0.19*
- 0.33**
- 0.05
0.49***
0.15
0.96
ID DIST
t
0.12
ID DIST
t+1
Figure 8. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #4. Numbers
represent completely standardized maximum likelihood parameter estimates. Dashed lines
indicate nonsignificant parameters retained in the final model (p > .05). ANG = anger; IMP =
impulsivity; ID DIST = identity disturbance. Interval between t and t + 1 is 5520s.
* p < .05. ** p < .01. *** p < .001.
45
Periodic trends. The trend analysis did not suggest the presence of a periodic trend in
anger, impulsivity, or identity disturbance for participant 4.
Participant 5.
Diagnosis and baseline characteristics. Participant 5 was a 27-year-old Caucasian
woman who identified as bisexual and gave her relationship status as “dating.” She was
recruited through clinic intake and diagnosed with MDD and social phobia on Axis I and
avoidant personality disorder on Axis II. She met two BPD criteria: chronic feelings of
emptiness and recurrent suicidal behavior. Her BPD dimensional score on the IPDE was 7, and
her GAF score was 45. Her scores on baseline self-report measures were as follows: GAPD =
2.37, SCCS = 2.75, BSCS = 1.85, ALS = 2.59. She endorsed 5 items on the MSI-BPD.
Dynamic factor model. The baseline AR(1) model showed good fit for participant 5’s
data. However, the autoregressive parameter values were not significant, suggesting that
participant 5’s levels of anger, impulsivity, and identity disturbance did not relate to levels of
these variables 7000 seconds later. Anger was not significantly related to either synchronic
impulsivity or synchronic identity disturbance in the final model, but impulsivity and identity
disturbance were modestly and positively related.
Periodic trends. The trend analysis suggested a periodic trend in participant 5’s identity
disturbance levels, bsin = - 4.45, t = 2.52, p = 0.01. The periodic function had an estimated
amplitude of 4.81 and a peak identity disturbance in late morning. Applying the residual values
from this function to the dynamic factor modeling procedure, the AR(1) model no longer showed
adequate fit on all indices, χ2(6) = 6.97, CFI = 0.92, NNFI = 0.81, SRMR = 0.060, RMSEA =
0.042. Modification indices suggested an additional cross-lagged relationship between identity
disturbance at time t and impulsivity at time t + 1, after which the model fit well on all indices,
46
χ2(5) = 3.23, p = 0.66, CFI = 1.00, NNFI = 1.00, SRMR = 0.042, RMSEA = 0.0. However, this
additional parameter itself was not significant, unstandardized β = - 0.27, t = 1.97, p = .052.
Overall, variance in time t values accounted for 2%, 5%, and 0% of the t + 1 variance in anger,
impulsivity, and identity disturbance, respectively. Figure 9 shows participant 5’s final dynamic
factor model.
0.98
ANG
t
- 0.14
ANG
t+1
- 0.14
- 0.10
0.14
0.95
IMP
t
0.14
0.24*
0.07
IMP
t+1
0.22*
- 0.21
1.00
ID DIST
resid.
t
0.15
ID DIST
resid.
t+1
Figure 9. Dynamic factor model describing the relationship between anger, impulsivity, and the
residual values of identity disturbance (after accounting for the circadian trend in identity
disturbance), at synchronous and successive time points, for participant #5. Numbers represent
completely standardized maximum likelihood parameter estimates. Dashed lines indicate
nonsignificant parameters retained in the final model (p > .05). ANG = anger; IMP =
impulsivity; ID DIST resid. = residual values of identity disturbance from trend analysis. Interval
between t and t + 1 is 7000s.
* p < .05. ** p < .01. *** p < .001.
47
Participant 6.
Diagnosis and baseline characteristics. Participant 6 was a 20-year-old single Caucasian
woman who identified as heterosexual. She was recruited through clinic intake and was
diagnosed with social phobia; additionally, bulimia nervosa and binge eating disorder were
included as diagnoses to be ruled out. On Axis II, participant 6 qualified for obsessivecompulsive personality disorder and met one BPD criterion (impulsivity). She received a score
of 2 on the IPDE BPD dimension. Her scores on baseline self-report measures were: GAPD =
2.75, SCCS = 3.25, BSCS = 2.38, ALS = 2.56. She endorsed 4 items on the MSI-BPD and had a
GAF score of 80.
Dynamic factor model. The baseline AR(1) model showed good fit for participant 6’s
data. However, impulsivity was the only autoregressive parameter to reach significance,
unstandardized β = 0.28, t = 2.29, p = .03. In addition, anger was related to impulsivity in
synchronic ratings, unstandardized ψ = 245.08, t = 2.84, p = .006. In all, time t variance
accounted for 3%, 8%, and 4% of variance in t + 1 anger, impulsivity, and identity disturbance,
respectively. Figure 10 shows the final dynamic factor model for this participant.
Periodic trends. The trend analysis did not suggest a periodic trend in participant 6’s
data.
Participant 7.
Diagnosis and baseline characteristics. Participant 7 was a 38-year-old single Caucasian
woman who identified as heterosexual. She was recruited through previous participation in
research and diagnosed with mood disorder NOS on Axis I, with past diagnoses of PTSD and
eating disorder NOS. On Axis II, she received diagnoses of BPD and paranoid personality
disorder. She met all nine BPD criteria, giving her an IPDE dimensional score of 18, and she
48
had a GAF score of 35. She endorsed 7 items on the MSI-BPD. Her scores on baseline selfreport measures were: GAPD = 2.49, SCCS = 2.33, BSCS = 1.85, ALS = 1.91.
0.97
ANG
t
0.18
ANG
t+1
0.41*
0.44*
0.06
0.92
0.06
IMP
t
0.28*
IMP
t+1
0.21
0.21
0.96
ID DIST
t
0.21
ID DIST
t+1
Figure 10. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #6. Numbers
represent completely standardized maximum likelihood parameter estimates. Dashed lines
indicate nonsignificant parameters retained in the final model (p > .05). ANG = anger; IMP =
impulsivity; ID DIST = identity disturbance. Interval between t and t + 1 is 6000s.
* p < .05. ** p < .01. *** p < .001.
Dynamic factor model. The baseline AR(1) model fit participant 7’s data well, although
only impulsivity showed a significant autoregressive character, unstandardized β = 0.29, t = 3.17,
p = .002. Impulsivity was reliably (and positively) related to synchronic identity disturbance,
unstandardized ψ = 234.61, t = 3.99, p < .001. Time t values explained 18, 6, and 10 percent of
49
variance in t + 1 ratings of anger, impulsivity, and identity disturbance, respectively. Figure 11
shows the final dynamic factor model for participant 7.
Periodic trends. The trend analysis did not reveal any evidence of periodic trends in
participant 7’s data.
0.99
ANG
t
- 0.09
ANG
t+1
0.22*
0.21
0.16
0.92
IMP
t
0.17
0.29**
IMP
t+1
0.48***
0.48***
1.00
ID DIST
t
- 0.01
ID DIST
t+1
Figure 11. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #7. Numbers
represent completely standardized maximum likelihood parameter estimates. Dashed lines
indicate nonsignificant parameters retained in the final model (p > .05). ANG = anger; IMP =
impulsivity; ID DIST = identity disturbance. Interval between t and t + 1 is 7000s.
* p < .05. ** p < .01. *** p < .001.
50
Participant 8.
Diagnosis and baseline characteristics. Participant 8 was a 50-year-old divorced
Caucasian woman who identified as heterosexual and also noted that she was in a romantic
relationship. She was recruited through previous participation in research. She was diagnosed
with recurrent MDD of moderate severity, dysthymic disorder, social phobia, and eating disorder
NOS. Although she qualified for an alcohol dependence diagnosis based on her recent history of
alcohol use, she was not currently drinking. She met eight BPD criteria (the unmet criterion was
inappropriate, intense anger) and received a dimensional score of 17 on the IPDE. However, she
only endorsed 4 items on the MSI-BPD. She had a GAF score of 50. Her scores on baseline
self-report measures were: GAPD = 2.65, SCCS = 3.33, BSCS = 2.23, ALS = 2.74.
Dynamic factor model. The baseline AR(1) model fit well for participant 8’s data. Each
parameter was significant in the model, with reliable autoregressive relationships for anger,
impulsivity, and identity disturbance across a 6400-second interval, as well as a positive and
moderate correlation between synchronic ratings of anger and impulsivity and positive
correlations between identity disturbance and anger and identity disturbance and impulsivity.
Variance in time t values explained 18, 6, and 10 percent of variance in t + 1 ratings of anger,
impulsivity, and identity disturbance, respectively.
Periodic trends. The trend analysis suggested a significant periodic trend in participant
8’s anger ratings, bsin = -6.13, t = 2.27, p = 0.02. The regression model showed a peak in
participant 8’s anger ratings about 5 hours 20 minutes after the first rating of the day, with an
amplitude of 7.02 points on a 0-100 scale. In addition, this participant’s identity disturbance
values showed a periodic trend, bcos = 5.54, t = 2.84, p = 0.0053, with the wave function showing
a peak identity disturbance value at about 9 hours after the first rating of the day with an
51
amplitude of 5.55 points. Accounting for these trends did not alter the dynamic factor model for
this participant (Figure 12).
0.82
ANG
t
0.42***
ANG
t+1
0.33***
0.38***
0.34***
0.94
IMP
t
0.33**
0.25**
IMP
t+1
0.53***
0.56***
0.90
ID DIST
t
0.31***
ID DIST
t+1
Figure 12. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #8. Numbers
represent completely standardized maximum likelihood parameter estimates. ANG = anger; IMP
= impulsivity; ID DIST = identity disturbance. Interval between t and t + 1 is 6400s.
* p < .05. ** p < .01. *** p < .001.
Participant 9.
Diagnosis and baseline characteristics. Participant 9 was a 32-year-old single Caucasian
man who identified as heterosexual. He was recruited through clinic intake and diagnosed with
social phobia. In addition, a single major depressive episode in partial remission was noted,
along with a rule-out diagnosis of dysthymic disorder. He received no diagnosis on Axis II and
52
received a dimensional BPD score of 3 on the IPDE without meeting any BPD criteria. He
endorsed 3 items on the MSI-BPD and had a GAF score of 55. His scores on baseline self-report
measures were: GAPD = 3.06, SCCS = 4.33, BSCS = 2.23, ALS = 1.20.
Dynamic factor model. Participant 9 rated his identity disturbance at 0 on a 0-100 scale
at every occasion during the three-week diary period, which was consistent with his score on the
12-item SCCS at baseline (the highest in the sample) and the fact that he did not endorse the
identity disturbance item on the MSI-BPD nor meet this criterion based on the diagnostic
interview. Because these ratings did not vary over the course of the three weeks, only anger and
impulsivity were used in the dynamic factor model. An AR(1) model containing only these two
variables at t and t + 1 fit well. Anger and impulsivity both showed significant stability from
one occasion to the next (with a 6000-second interval), unstandardized βanger = 0.42, t = 4.95, p <
.001 and unstandardized βimpulsivity = 0.31, t = 3.40, p < .001. Anger and impulsivity were
moderately and positively correlated, unstandardized ψ = 129.59, t = 3.62, p < .001.
Approximately 18% and 9% of anger and impulsivity variance at t + 1, respectively, was
accounted for by ratings at the previous occasion.
Periodic trends. Participant 9’s anger showed a significant periodic trend, bsin = 9.69, t =
3.23, p = 0.0016. According to the regression model, the amplitude of the wave was 9.86 on a 0100 scale and peaked about 6 hours after the first rating of the day (i.e., mid-afternoon).
Impulsivity also displayed a periodic trend, bsin = - 7.85, t = 2.75, p = 0.007. According to the
model, participant 8’s impulsivity peaked about 75 minutes after the first rating of the day with
an amplitude above the mean of 7.98 points on a 0-100 scale. Adjusting for these trends by
using residual values in dynamic factor models resulted in no changes to model parameters
relative to the model using raw data, so this model was retained (Figure 13).
53
0.82
ANG
t
0.43***
ANG
t+1
0.29**
0.39***
0.91
IMP
t
0.31***
IMP
t+1
Figure 13. Dynamic factor model describing the relationship between anger and impulsivity, at
synchronous and successive time points, for participant #9. Numbers represent completely
standardized maximum likelihood parameter estimates. ANG = anger; IMP = impulsivity.
Interval between t and t + 1 is 6000s.
* p < .05. ** p < .01. *** p < .001.
Participant 10.
Diagnosis and baseline characteristics. Participant 10 was a 28-year-old single woman.
She identified as heterosexual and selected three ethnic categories to describe herself: AfricanAmerican, American Indian/Alaska Native, and Hispanic or Latina. She was recruited through
previous research participation. She was diagnosed with dysthymic disorder and received no
diagnosis on Axis II. She did not meet any BPD criteria nor receive any subthreshold ratings on
the IPDE. She had a GAF score of 75 and endorsed 4 items on the MSI-BPD. Her scores on
baseline self-report measures were: GAPD = 3.32, SCCS = 4.17, BSCS = 3.00, ALS = 1.20.
Dynamic factor model. The baseline AR(1) model fit the data well for participant 10.
However, only anger showed a significant autoregressive relationship across 7000-second
intervals, unstandardized β = 0.47, t = 5.24, p < .001. In addition, impulsivity and identity
54
disturbance were positively correlated in synchronic ratings, unstandardized ψ = 12.79, t = 2.45,
p = .02. None of the other model parameters were significantly different from 0. Anger at time t
accounted for 22% of the variance in anger at t + 1, whereas impulsivity and identity disturbance
showed lower stability.
Trend analysis. The trend analysis did not reveal any evidence of periodic trends in
participant 10’s data, and so the AR(1) model was retained (Figure 14).
0.78
ANG
t
0.47***
ANG
t+1
0.05
0.10
0.99
IMP
t
0.06
0.12
IMP
t+1
0.05
0.25*
0.26*
1.00
ID DIST
t
-0.06
ID DIST
t+1
Figure 14. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #10. Numbers
represent completely standardized maximum likelihood parameter estimates. Dashed lines
indicate nonsignificant parameters retained in the final model (p > .05). ANG = anger; IMP =
impulsivity; ID DIST = identity disturbance. Interval between t and t + 1 is 7000s.
* p < .05. ** p < .01. *** p < .001.
55
Participant 11.
Diagnosis and baseline characteristics. Participant 4 was a 40-year-old divorced
African-American woman who identified as heterosexual. She was recruited through previous
research participation and diagnosed with recurrent major depressive disorder of moderate
severity. She met two BPD criteria, identity disturbance and stress-related paranoid ideation, and
she received an IPDE dimensional score of 6. Her GAF score was 55, and her scores on baseline
self-report measures were: GAPD = 2.05, SCCS = 2.17, BSCS = 3.31, ALS = 3.83. She
endorsed 5 items on the MSI-BPD.
Dynamic factor model. The AR(1) model fit participant 11’s data well, although none of
the autoregressive parameters were significantly different from 0. Only 2% of anger and
impulsivity ratings at t + 1 were accounted for by earlier ratings, and a negligible portion of the
variance in identity disturbance was accounted for by earlier ratings. Only impulsivity and
identity disturbance showed significant relationships in synchronic ratings, unstandardized ψ =
104.10, t = 3.20, p = .002. Figure 15 shows the final dynamic factor model for Participant 11.
Trend analysis. The trend analysis did not reveal any evidence of periodic trends in
participant 11’s data.
General Results
Dynamic factor modeling revealed a wide variety of relationships between anger,
impulsivity, and identity disturbance, both in synchronous ratings and over time intervals of
approximately 1.5-2 hours. For six individuals, an autoregressive model was sufficient to
describe the variation in these symptoms from one measurement occasion to the next, although
the parameter values of the autoregressive relationships themselves varied considerably. For the
other five individuals, cross-lagged relationships between one symptom at one occasion and a
56
0.98
ANG
t
- 0.15
ANG
t+1
0.07
0.09
0.98
IMP
t
0.12
0.14
IMP
t+1
0.11
0.36**
0.37**
1.00
ID DIST
t
0.07
ID DIST
t+1
Figure 15. Dynamic factor model describing the relationship between anger, impulsivity, and
identity disturbance, at synchronous and successive time points, for participant #11. Numbers
represent completely standardized maximum likelihood parameter estimates. Dashed lines
indicate nonsignificant parameters retained in the final model (p > .05). ANG = anger; IMP =
impulsivity; ID DIST = identity disturbance. Interval between t and t + 1 is 7000s.
* p < .05. ** p < .01. *** p < .001.
second symptom at the next occasion were required to achieve good model fit. Within these
VAR(1) models, there was not a consistent relationship between identity disturbance and the
other symptoms. In two individuals, identity disturbance negatively predicted later anger or
impulsivity; in one, identity disturbance followed (but did not precede) feelings of anger; and in
the remaining two participants, identity disturbance did not show a lagged relationship with the
other BPD symptoms. In general, identity problems tended to be positively correlated with
synchronous anger and impulsivity, although the magnitude of these relationships also varied.
Overall, the results suggested that there was not a consistent relationship between identity
57
disturbance and the BPD symptoms of anger and impulsivity, although isolated cases conformed
to the theoretical models posited by psychodynamic (e.g., Kernberg, 1984) and behavioral (e.g.,
Linehan, 1993) theorists. The presence or absence of a DSM-IV BPD diagnosis did not seem to
make a difference in the dynamic factor model adopted.
58
Chapter 4
Discussion
The present study represents a preliminary effort to apply person-specific modeling
approaches to ecological momentary assessment data in individuals with BPD. The aims of the
current study were to investigate the dynamic relationship of subjective identity disturbance to
anger and impulsivity and to explore whether these relationships were similar across individuals
with, and without, a diagnosis of BPD. The dynamic factor models varied in their resemblance
to the models postulated on the basis of prior theories. Hypothesis 1, based on the
psychodynamic notion that identity disturbance is a “core” symptom of BPD that accounts for
the appearance of the other symptoms, was supported in two of 11 participants. For these
individuals, the experience of identity disturbance negatively predicted the later appearance of
anger or impulsivity. Hypothesis 2, that identity disturbance is primarily the result of behavioral
and affective instability rather than their cause, was supported in one individual, whose sense of
self-concept clarity decreased after an earlier increase in anger. Otherwise, there were no
individuals for whom identity disturbance varied with anger and impulsivity in line with these
prominent theories. Thus, the results failed to support either hypothesis in the majority of the
sample. However, results suggested that there is a good deal of heterogeneity between
individuals in the nature and strength of the relationships between these three BPD symptoms
over time, in line with hypothesis 3. Finally, contrary to hypothesis 4, there was little indication
in the present data that the type of model or the extent of the relationships between the three
symptoms depended on the presence or absence of a BPD diagnosis.
The results showed that identity disturbance has a complex and varied temporal
relationship with other BPD symptoms. Interestingly, identity disturbance was sometimes
59
negatively related to later expressions of anger and impulsivity in the dynamic factor models.
The direction of this relationship is consistent with the notion that a temporary sense of selfconcept clarity can precede affective and behavioral dysregulation (Fonagy, Gergely, Jurist, &
Target, 2004). It may be for example, that a temporary, and illusory, subjective sense of selfconcept clarity arises in some individuals with BPD because they engage in the defensive
strategy of splitting (Kernberg, 1984): they keep positive and negative aspects of their selfconcept separate and thus do not have access to self-information that might help them selfregulate (Lumsden, 1993). Thus, they may be more vulnerable to extreme behaviors when their
sense of self is more secure.
The results also provide support for the notion that the subjective experience of selfconcept clarity can be dissociated from the more trait-like and structural construct of “identity
diffusion” or “identity disturbance” proposed by clinical authors and captured in the DSM. This
distinction is in line with findings in social-psychological research (Ayduk, Gyurak, & Luerssen,
2009; Boyce, 2008; Nezlek & Plesko, 2001). The construct assessed in the current study is
likely to conform to the former in its direct assessment of momentary and subjective sense of
self, which did vary substantially in 10 of 11 participants. In contrast, the latter may be best
assessed by an observer (such as a clinician), who can witness seemingly contradictory
statements about goals, values, and other self-defining attributes over time, even as the individual
may not be aware of this variation. A clinician may also infer the presence of structural identity
disturbance from an individual’s history of erratic investment in different roles, vocational and
social pursuits, and relationships (Kernberg, 1984, 2006; Stern et al., 2010), even if the
individual draws a strong sense of identity from these activities at the time.
60
In addition, in one individual, identity disturbance was the sequel of earlier anger (and in
another participant this relationship was present in the final model but did not reach
significance). This relationship provides limited support for the idea that identity disturbance
can be a temporary consequence of emotion dysregulation (Linehan, 1993). Longitudinal tests
of this proposition would be helpful. Also useful would be direct experimental tests of the
relationship between emotional and behavioral dysregulation and later self-concept. For
example, recent studies have shown that impulsivity can be induced in laboratory tasks,
producing a tendency to indulge in maladaptive eating and substance use (Guerrieri, Nederkoorn,
Schrooten, Martijn, & Jansen, 2009; Jones et al., 2011). Anger inductions through exposure to
provocative stimuli, harassment by a confederate, punishment by another participant, stress
interview, and environmental manipulation also have a rich history in laboratory settings
(Lobbestael, Arntz, & Wiers, 2008). It would be instructive to see how momentary self-concept
clarity is affected by these experimental manipulations.
Despite the presence of a few individuals in the sample who showed these hypothesized
relationships, most participants did not show a lagged relationship between identity disturbance
and the other variables, although synchronous relationships were common. For them, symptoms
did not show a dynamic relationship across the time intervals in the current study. This seems
consistent with the model of BPD presented in DSM, in which the disorder is described by a
temporally stable cluster of symptoms that do not necessarily influence one another. However,
the stability of these models, as indicated by the residual variance of anger, impulsivity, and
identity disturbance at t + 1, was generally fairly low, suggesting that the autoregressive models
did not account very well for these variables’ values over time. In addition, the autoregressive
parameters themselves were often not significantly different from zero. It could be that the
61
experience of anger, impulsivity, and identity disturbance is not stable for these individuals or
that it depends on contextual factors or other psychological experiences not measured in the
current study.
It did not seem that the models describing the dynamic relationships between anger,
impulsivity, and identity disturbance depended on the presence or absence of a DSM diagnosis of
BPD, as some participants without BPD demonstrated dynamic relationships between anger,
impulsivity, and identity disturbance, and some participants with the diagnosis did not show
evidence of these relationships. The significance of this finding is unclear, however, given the
small sample size and the absence of explicit statistical comparisons between individuals or
groups. It is also true that many participants without BPD in the current study met some number
of criteria for the disorder. In fact, the median number of BPD criteria met by these seven
individuals was 2. Research indicates that the presence of even one BPD criterion is associated
with greater levels of functional impairment and suicidality and a greater likelihood of
psychiatric hospitalization (Zimmerman, Chelminski, Young, Dalrymple, & Martinez, 2012),
suggesting that even at subthreshold levels, BPD symptoms may have important consequences.
It should also be noted that the DSM differentiates individuals with BPD on the basis of
symptomatic level, not on the covariation of symptoms with one another (as in the current
study). This leaves open the possibility that diagnostic groups could be differentiated in the
current data by the severity of these symptoms or in their functional consequences for the
individual, both of which are clearly important considerations.
Nevertheless, given the presence of subthreshold BPD traits in much the sample and the
fact that dynamic relationships between these symptoms could sometimes be found in the
absence of the disorder itself, the current results raise the possibility that the pathological
62
processes involved in BPD may cross diagnostic lines. This finding has implications for
evolving nosological systems. Disorders in the DSM are often presumed to operate as latent
entities, based in underlying physiological, neuroanatomical, or psychological mechanisms, that
cause the appearance of observable “symptoms.” This assumption is made explicit, for example,
in factor analyses of the symptom clusters that purport to test the validity of these diagnoses
(e.g., Blais, Hilsenroth, & Castlebury, 1997). However, influential critics of the DSM diagnostic
system have recently argued that a scientific nosology should reflect fundamental
neurobiological and psychological mechanisms of dysfunction (e.g., Cuthbert & Insel, 2012;
Kendler, Zachar, & Craver, 2011). The present findings suggest that relationships between
anger, impulsivity, and identity pathology may cut across DSM-IV categories.
As noted above, this does not in itself argue against the validity of these diagnoses, which
generally refer more to the mean levels of these symptoms within an individual rather than an
essential pattern of their covariation over time. However, a number of researchers have recently
begun to conceptualize psychopathology from the bottom up, viewing symptoms themselves as
important entities that interrelate over time, may be reciprocally causal, and cross standard
diagnostic boundaries (Borsboom, Cramer, Schmittmann, Epskamp, & Waldorp, 2011;
Bringmann et al., 2013; Kendler, Zachar, & Craver, 2011; van Os, Delespaul, Wigman, MyinGermeys, & Wichers, 2013; Wichers, in press; Wigman et al., 2013). Some of these researchers
have proposed that these symptom networks might eventually supplant diagnostic systems such
as the DSM (e.g., Wichers, 2013). EMA methods may aid in the construction of these symptom
networks given that individual symptoms can be assessed more unobtrusively than disorders (the
diagnosis of which requires inquiry about several symptoms at once), making them easy to
capture in real-life contexts, and because causal modeling generally requires longitudinal
63
assessments of relatively high frequency. The severity of various symptoms could also be
captured in these models (e.g., Bringmann et al., 2013), which would provide valuable
information about which symptoms are most extreme. An approach taking both symptom
severity and covariation into account would likely be of the greatest theoretical and clinical
utility.
The results also point to the heterogeneity of longitudinal variation in BPD symptoms
within individuals. Thus, the current results support the notion that models of intraindividual
variation reveal heterogeneity among individuals that may not be apparent in nomothetic models
of psychological processes. Identity disturbance may positively predict emotional and
behavioral disturbance in some individuals, whereas it may relate negatively to these symptoms
of BPD in others. It may also be a cause of these phenomena in some but an effect of them in
others. In short, it is possible that multiple theoretically derived models of the mechanisms of
BPD hold true simultaneously within different subgroups of BPD patients. In principle, it could
be possible for some of the data derived from previous EMA studies of BPD (e.g., Berenson,
Downey, Rafaeli, Coifman, & Paquin, 2011) to be re-analyzed using person-specific techniques.
This could reveal whether the models derived from the interindividual and groupwise analyses
also describe intraindividual variation with uniformity. The possibility of such a secondary
analysis in any given study depends on the suitability of the data structure, adequate power, and
longitudinal coverage of the variables of interest, but the heterogeneity revealed in the present
data suggests that it could be scientifically important.
Thanks to recent developments in person-specific methods, it may now be possible to
conduct analyses of intraindividual variation that both respect heterogeneity between individuals
and lead to nomothetic models that can be applied to broader populations or subgroups. For
64
example, Gates and Molenaar (2012) describe an iterative method whereby overall group
structures are first derived that work well for the majority of individuals, and then individual
paths are allowed. This approach was shown to work for fMRI data from homogeneous and
heterogeneous groups. In a similar vein, Bringmann et al. (2013) combined an intraindividual
and interindividual approach by creating multilevel VAR models of emotion data, collected
using EMA, among 129 participants with residual depression. The authors then submitted these
models to a network analysis to estimate the average connection strength of the emotion
variables, leading to an overall model of pathology. An approach similar to these, suitably
applied to EMA data, might allow future researchers to account for the person-specificity of
BPD dynamics while also creating models that apply generally to everyone with BPD, or to
subgroups of individuals with the disorder.
The current results also have several clinical implications. First, they highlight that not
every individual with BPD is the same and suggest that clinicians should pay attention to the
idiographic, dynamic relationships between symptoms of the disorder. This would be fairly easy
to integrate into several empirically-supported treatments for BPD, many of which already have
therapists and clients monitor different phenomena as they unfold over time in order to plan
interventions and foster insight. For example, clients participating in Dialectical Behavior
Therapy (DBT; Linehan, 1993) are asked to complete “diary cards” by rating different symptoms
of BPD, affective states, behaviors, impulsive urges, and the like on a daily basis. These records
are then used to analyze chains of behavior in order to identify problematic patterns and to
suggest ways of disrupting them. In this way, DBT interventions are tailored to each individual,
but the process of gathering these diary data is lengthy, taking on the order of months. Using
EMA data to monitor these patterns could make this system more time-effective, and dynamic
65
factor analysis could lend statistical rigor to the normally impressionistic system of behavioral
chain analysis. Interventions could also be introduced into the statistical model to examine the
effect of particular session-to-session therapist actions on the individual’s pathology (see, e.g.,
Gates, Molenaar, Hillary, & Slobounov, 2011).
The advent of increasingly affordable and powerful mobile technologies and the
development of data-analytic techniques to model the resultant data have also led some
researchers to advocate for an entirely person-specific approach to diagnosis, based on reciprocal
interactions between affective states and behaviors instead of on symptom checklists (van Os et
al., 2013). This “precision diagnosis” approach could potentially empower clients as they
participate in collecting and interpreting experience-near diagnostic data and would also lead
naturally into interventions to counter pathological relationships between symptoms. Similar
appeals have been made for the development of person-specific methods of personality
assessment (Caldwell, Cervone, & Rubin, 2008; Haynes, Mumma, & Pinson, 2009). The current
results show that basic person-specific approaches to understanding the dynamics of
psychopathology can be implemented fairly easily using readily available software. The utility
of these methods for diagnosis and assessment in clinical settings may depend on clinicians’
ability to select a limited number of relevant symptoms for a given client, so that ecological
validity can be maximized and intervals between assessments can be minimized.
Individual differences derived from these assessments might be very useful for the
practice of psychotherapy. For example, variability in extraversion over time may reflect an
individual’s reactivity to situational cues calling for extraverted behavior (Fleeson, 2001).
Similarly, the stability of dynamic processes between BPD symptoms may indicate how
amenable a client might be to disruptions in these processes by a skilled therapist (Fisher et al.,
66
2011). If multiple processes are evident, a therapist might select a therapy target according to the
process most likely to be affected by deliberate interventions. Similarly, EMA data might be
used to derive prescriptive recommendations for one therapy over another, depending on the
processes that might be targeted by different therapies. For example, if an individual’s selfconcept incoherence appeared to be driving later impulsive or self-injurious behaviors or
emotion dysregulation, a therapy targeting the client’s sense of self (such as TransferenceFocused Psychotherapy; Clarkin, Yeomans, & Kernberg, 2006) might be most appropriate. On
the other hand, a therapy targeting behavioral or emotional dysregulation directly (such as DBT)
might be the best choice for a client whose identity disturbance appeared to be the product of
these other symptoms. EMA data might also be profitably used to monitor the outcome of
psychotherapy by examining the degree to which therapy clients’ pathological processes
(measured before therapy) change as a result of therapeutic intervention (Fisher et al.; Piasecki,
Hufford, Solhan, & Trull, 2007). Such data might also be used to understand psychotherapy
process by helping to quantify dynamic processes that are hypothesized to mediate between
therapy techniques and outcomes (e.g., Gunthert, Cohen, Butler, & Beck, 2005).
Several limitations of the current study deserve mention. One potential limitation is in
the frequency of the observations taken during the EMA sampling period. In order to accurately
model the processes of interest in BPD, data must be collected at an adequate frequency (Collins,
2006). More precisely, modeling a process requires sampling at twice the frequency of the
process itself (e.g., Shannon, 1949/1998). The 1-2 hour interval between observations in the
current study raises the possibility that important processes that occurred at a greater frequency
were not captured. For example, Ebner-Priemer and Sawitski (2007) have demonstrated, using
an EMA paradigm, that a specific affective process among individuals with BPD is only
67
captured when 15-30 minute sampling intervals are used. Although the current study did not
seek to capture this particular process, future research with smaller sampling intervals, as well as
techniques for analyzing process speed in EMA data (Shiyko & Ram, 2011), would be helpful in
discerning whether smaller sampling intervals are necessary to tap important dynamics among
the BPD symptoms involved in this study. Previous research has also uncovered the existence of
several processes of affective instability at varying frequencies within BPD samples (Nisenbaum
et al., 2010), and it is possible that the frequency of these processes varies from individual to
individual. If the present data are robust, they point to the presence of potentially important
psychological processes occurring over a span of 90-120 minutes involving anger, impulsivity,
and identity disturbance in some individuals. However, it is also possible that at least some of
the observed correlations between variables that were measured synchronously in the current
study actually represent non-contemporaneous relationships between variables that occur at a
smaller timescale.
A second limitation of the current study concerns measurement within the EMA sampling
protocol. The psychometric properties (that is, the reliability and validity) of the self-report
questions given during the EMA portion of the study are difficult to determine. In order to
maximize ecological validity by minimizing the time burden of completing surveys, the
constructs in the EMA surveys were operationalized with only one self-report question each.
Single questions preclude the possibility of bolstering construct validity via aggregation and are
thus not generally optimally reliable or valid measures of constructs. The current study adopted
the single-item strategy in order to maximize the ecological validity of the EMA protocol, and
the tradeoff between ecological validity and construct validity in EMA research is a widely
recognized one (Bolger et al., 2003; Csikszentmihalyi & Larson, 1987; Shiffman et al., 2008)
68
without a consensus solution. However, multi-item measures of anger, impulsivity, and identity
disturbance would be preferable as measures of these phenomena, at least in terms of reliability
of measurement. In addition, the measurement of impulsivity by self-report is particularly
challenging because these measures show an uncertain relationship with behavioral measures of
impulsivity (Cyders & Coskunpinar, 2011; Meda et al., 2009; Reynolds, Ortengren, Richards, &
deWit, 2006). Thus, the correspondence between the self-report ratings of impulsivity used in
the current study and actual impulsive behaviors is not clear.
A third limitation of the current study is its power. The 11 participants in the sample
each provided enough EMA data to estimate the fit of dynamic factor models. However, the
power needed for examination of model fit is not necessarily the same as power needed to avoid
type II errors in the detection of significant parameter values within the model. This might be a
reason why several models required parameter estimates for good fit that were not themselves
significantly different from zero. It is also possible that additional power would afford the
opportunity to detect other synchronous and lagged relationships in these time series that were
not apparent using the current data. In addition, the relatively limited sample of individuals in
the current study precludes any direct statistical estimation of commonalities between these
individuals in their model parameters. Techniques are being developed to allow for the
estimation of parameters that apply across groups, even as person-specific parameters are also
allowed (e.g., Gates & Molenaar, 2012), but the small sample size precludes their use in the
current study.
Finally, the person-specific approach used in the current study is not without its
limitations. For example, as stated above, the dynamic factor modeling approach required the
assumption of weak stationarity in the data, which may not be strictly tenable given that several
69
individuals in the sample had just received preliminary attention in a psychiatric setting, which
can be a time of rapid symptom change (Hansen, Lambert, & Forman, 2002; Howard, Kopta,
Krause, & Orlinsky, 1986; Kopta, Howard, Lowry, & Beutler, 1994). Psychotherapy may have
also had effects on the covariance of these symptoms over the three-week smartphone period. In
addition, the relatively small number of participants in the current sample made comparison
between models and generalization to the broader population difficult. Nevertheless, methods
have been developed to retain a person-specific approach while dealing with several of these
limitations. For example, Molenaar et al. (2009) describe an approach to time series modeling
that does not require weakly stationary data. Another extension of the basic dynamic factor
modeling approach used in the current paper allows for large numbers of variables (Zuur, Fryer,
Jolliffe, Dekker, & Beukema, 2003), which might be helpful for examining transdiagnostic sets
of symptoms instead of beginning from within the BPD criterion set, as in the current paper.
Given the prevalence of BPD as well as its debilitating nature and its direct and indirect
costs, it is vital to understand properly and to treat effectively. The current study is, to my
knowledge, the first to apply dynamic factor modeling to EMA data with individuals with BPD.
As such, it presents a statistically rigorous method for modeling the dynamic processes of the
disorder, one person at a time, using a promising data-collection technique. These models can
then be aggregated to delineate subgroups or to create models that work optimally well for most
individuals (e.g., Gates & Molenaar, 2012), a criterion that is scientifically important due to the
high level of heterogeneity within BPD. This modeling approach also has the potential to fulfill
recent appeals for a psychiatric nosology based on neurobiological and behavioral systems (e.g.,
Cuthbert & Insel, 2012) because of its data-driven, bottom-up construction, and it may be
feasible as a diagnostic or outcome-monitoring system in clinical practice as well. Finally, this
70
approach also makes it possible to test hypotheses derived from clinical theory by modeling
predicted patterns of symptom dynamics over time. The current results do not fully support the
dynamic models derived from either prominent psychodynamic or behavioral conceptualizations
of BPD, although they indicate that certain individuals may obey these patterns. Further research
will be needed to ascertain how often theorized symptom dynamics can be recovered under
different sampling conditions.
71
References
American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders
(4th edition, text revision). Arlington, VA: American Psychiatric Association.
American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders
(5th ed.). Arlington, VA: American Psychiatric Association.
Ansell, E. B., Sanislow, C. A., McGlashan, T. H., & Grilo, C. M. (2007). Psychosocial
impairment and treatment utilization by patients with borderline personality disorder,
other personality disorders, mood and anxiety disorders, and a healthy comparison group.
Comprehensive Psychiatry, 48, 329-336.
Asnaani, A., Chelminski, I., Young, D., & Zimmerman, M. (2007). Heterogeneity of borderline
personality disorder: Do the number of criteria make a difference? Journal of Personality
Disorders, 21, 615-625.
Ayduk, Ö., Gyurak, A., & Luerssen, A. (2009). Rejection sensitivity moderates the impact of
rejection on self-concept clarity. Personality and Social Psychology Bulletin, 35, 14671478.
Bateman, A., & Fonagy, P. (2004). Psychotherapy for borderline personality disorder:
Mentalization-based treatment. Oxford: Oxford University Press.
Bateman, A., & Fonagy, P. (2008). 8-year follow-up of patients treated for borderline personality
disorder: Mentalization-based treatment versus treatment as usual. American Journal of
Psychiatry, 165, 631-638.
Bateman, A., & Fonagy, P. (2009). Randomized controlled trial of outpatient mentalizationbased treatment versus structured clinical management for borderline personality
disorder. American Journal of Psychiatry, 166, 1355-1364.
72
Bates, D. M., & Venables, W. N. (2010). Regression spline functions and classes (Version
2.11.1).
Becker, D. F., McGlashan, T. H., & Grilo, C. M. (2006). Exploratory factor analysis of
borderline personality disorder criteria in hospitalized adolescents. Comprehensive
Psychiatry, 47, 99-105.
Beckmann, N., Wood, R. E., & Minbashian, A. (2010). It depends how you look at it: On the
relationship between neuroticism and conscientiousness at the within- and the betweenperson levels of analysis. Journal of Research in Personality, 44, 593-601.
Benazzi, F. (2006). Borderline personality-bipolar spectrum relationship. Progress in NeuroPsychopharmacology and Biological Psychiatry, 30, 68-74.
Bender, D. S., Dolan, R. T., Skodol, A. E., Sanislow, C. A., Dyck, I. R., McGlashan, T. H., …
Gunderson, J. G. (2001). Treatment utilization by patients with personality disorders.
American Journal of Psychiatry, 158, 295-302.
Bender, D. S., & Skodol, A. E. (2007). Borderline personality disorder as a self-other
representational disturbance. Journal of Personality Disorders, 21, 500-517.
Berenson, K. R., Downey, G., Rafaeli, E., Coifman, K. G., & Paquin, N. L. (2011). The
rejection-rage contingency in borderline personality disorder. Journal of Abnormal
Psychology, 120, 681-690.
Blais, M. A., Hilsenroth, M. J., & Castlebury, F. D. (1997). Content validity of the DSM-IV
borderline and narcissistic personality disorder criteria sets. Comprehensive Psychiatry,
38, 31-37.
Blum, N., St. John, D., Pfohl, B., Stuart, S., McCormick, B., Allen, J., … Black, D. W. (2008).
Systems training for emotional predictability and problem solving (STEPPS) for
73
outpatients with borderline personality disorder: A randomized controlled trial and 1-year
follow-up. American Journal of Psychiatry, 165, 468-478.
Bolger, N., Davis, A., & Rafaeli, E. (2003). Diary methods: Capturing life as it is lived. Annual
Review of Psychology, 54, 579-616.
Borsboom, D., Cramer, A. O. J., Schmittmann, V. D., Epskamp, S., & Waldorp, L. J. (2011).
The small world of psychopathology. PLoS One, 6(11).
doi:10.1371/journal.pone.0027407
Borsboom, D., Mellenbergh, G. J., & van Heerden, J. (2003). The theoretical status of latent
variables. Psychological Review, 110, 203-219.
Boyce, K. (2008). Self-structure and self-concept clarity: Evaluative integration as a means to
increased self-clarity. Doctoral dissertation, University of Oklahoma, Norman, OK.
Bradley, R., Conklin, C. Z., & Westen, D. (2005). The borderline personality diagnosis in
adolescents: Gender differences and subtypes. Journal of Child Psychology and
Psychiatry, 46, 1006-1019.
Bringmann, L. F., Vissers, N., Wichers, M., Geschwind, N., Kuppens, P., Peeters, F., …
Tuerlinckx, F. (2013). A network approach to psychopathology: New insights into
clinical longitudinal data. PLoS One, 8(4). doi:10.1371/journal.pone.0060188
Brown, T., DiNardo, P., & Barlow, D. (1994). Anxiety Disorders Interview Schedule, Adult
Version. Oxford: Oxford University Press.
Caldwell, T. L., Cervone, D., & Rubin, L. H. (2008). Explaining intra-individual variability in
social behavior through idiographic assessment: The case of humor. Journal of Research
in Personality, 42, 1229-1242.
74
Campbell, J. D., Assanand, S., & Di Paula, A. (2003). The structure of the self-concept and its
relation to psychological adjustment. Journal of Personality, 71, 115-140.
Campbell, J. D., Trapnell, P. D., Heine, S. J., Katz, I. M., Lavallee, L. F., & Lehman, D. R.
(1996). Self-concept clarity: Measurement, personality correlates, and cultural
boundaries. Journal of Personality and Social Psychology, 70, 141-156.
Chambless, D. L., Renneberg, B., Goldstein, A., & Gracely, E. J. (1992). MCMI-diagnosed
personality disorders among agoraphobic outpatients: Prevalence and relationship to
severity and treatment outcome. Journal of Anxiety Disorders, 6, 193-211.
Chanen, A. M., Jovev, M., Djaja, D., McDougall, E., Yuen, H. P., Rawlings, D., & Jackson, H. J.
(2008). Screening for borderline personality disorder in outpatient youth. Journal of
Personality Disorders, 22, 353-364.
Cheavens, J. S., Strunk, D. R., & Chriki, L. (2012). A comparison of three theoretically
important constructs: What accounts for symptoms of borderline personality disorder?
Journal of Clinical Psychology, 68, 477-486.
Clarkin, J. F., Levy, K. N., Lenzenweger, M. F., & Kernberg, O. F. (2007). Evaluating three
treatments for borderline personality disorder: A multiwave study. American Journal of
Psychiatry, 164, 922-928.
Clarkin, J. F., Widiger, T. A., Frances, A., Hurt, S. W., & Gilmore, M. (1983). Prototypic
typology and the borderline personality disorder. Journal of Abnormal Psychology, 92,
263-275.
Clarkin, J. F., Yeomans, F. E., & Kernberg, O. F. (2006). Psychotherapy for borderline
personality focusing on object relations. Washington, DC: American Psychiatric
Publishing, Inc.
75
Clifton, A., & Pilkonis, P. A. (2007). Evidence for a single latent class of Diagnostic and
Statistical Manual of Mental Disorders borderline personality pathology. Comprehensive
Psychiatry, 48, 70-78.
Cloitre, M., & Koenen, K. C. (2001). The impact of borderline personality disorder on process
group outcome among women with posttraumatic stress disorder related to childhood
abuse. International Journal of Group Psychotherapy, 51, 379-398.
Coid, J., Yang, M., Tyrer, P., Roberts, A., & Ullrich, S. (2006). Prevalence and correlates of
personality disorder in Great Britain. British Journal of Psychiatry, 188, 423-431.
Collins, L. M. (2006). Analysis of longitudinal data: The integration of theoretical model,
temporal design, and statistical model. Annual Review of Psychology, 57, 505-528.
Conner, T. S., Tennen, H., Fleeson, W., & Barrett, L. F. (2009). Experience sampling methods:
A modern idiographic approach to personality research. Social and Personality
Psychology Compass, 3, 292-313.
Cooper, L. D., Balsis, S., & Zimmerman, M. (2010). Challenges associated with a polythetic
diagnostic system: Criteria combinations in the personality disorders. Journal of
Abnormal Psychology, 119, 886-895.
Cooper, P. J., Coker, S., & Fleming, C. (1996). An evaluation of the efficacy of supervised
cognitive behavioral self-help for bulimia nervosa. Journal of Psychosomatic Research,
40, 281-287.
Cowdry, R. W., Gardner, D. L., O'Leary, K. M., Leibenluft, E., & Rubinow, D. R. (1991). Mood
variability: A study of four groups. American Journal of Psychiatry, 148, 1505-1511.
76
Crowell, S. E., Beauchaine, T. P., & Linehan, M. M. (2009). A biosocial developmental model
of borderline personality: Elaborating and extending Linehan's theory. Psychological
Bulletin, 135, 495-510.
Csikszentmihalyi, M., & Larson, R. (1987). Validity and reliability of the experience-sampling
method. Journal of Nervous and Mental Disease, 175, 526-536.
Cuthbert, B. N., & Insel, T. R. (2012). Toward the future of psychiatric diagnosis: The seven
pillars of RDoC. BMC Medicine, 11, 126.
Cyders, M. A., & Coskunpinar, A. (2011). Measurement of constructs using self-report and
behavioral lab tasks: Is there overlap in nomothetic span and construct representation for
impulsivity? Clinical Psychology Review, 31, 965-982.
Digre, E. I., Reece, J., Johnson, A. L., & Thomas, R. A. (2009). Treatment response in subtypes
of borderline personality disorder. Personality and Mental Health, 3, 56-67.
Duckworth, A. L., Tsukayama, E., & May, H. (2010). Establishing causality using longitudinal
hierarchical linear modeling: An illustration predicting achievement from self-control.
Social Psychological and Personality Science, 1, 311-317.
Ebner-Priemer, U. W., Kuo, J., Welch, S. S., Thielgen, T., Witte, S., Bohus, M., & Linehan, M.
M. (2006). A valence-dependent group-specific recall bias of retrospective self-reports: A
study of borderline personality disorder in everyday life. Journal of Nervous and Mental
Disease, 194, 774-779.
Ebner-Priemer, U. W., & Sawitzki, G. (2007). Ambulatory assessment of affective instability in
borderline personality disorder: The effect of the sampling frequency. European Journal
of Psychological Assessment, 23, 238-247.
77
Ebner-Priemer, U. W., & Trull, T. J. (2009). Ecological momentary assessment of mood
disorders and mood dysregulation. Psychological Assessment, 21, 463-475.
Ebner-Priemer, U. W., Welch, S. S., Grossman, P., Reisch, T., Linehan, M. M., & Bohus, M.
(2007). Psychophysiological ambulatory assessment of affective dysregulation in
borderline personality disorder. Psychiatry Research, 150, 265-275.
Ellison, W. D., & Levy, K. N. (2012). Factor structure of the primary scales of the Inventory of
Personality Organization in a nonclinical sample using exploratory structural equation
modeling. Psychological Assessment, 24, 503-517.
Fahrenberg, J., Myrtek, M., Pawlik, K., & Perrez, M. (2007). Ambulatory assessment Monitoring behavior in daily life settings: A behavioral-scientific challenge for
psychology. European Journal of Psychological Assessment, 23, 206-213.
Feeny, N. C., Zoellner, L. A., & Foa, E. B. (2002). Treatment outcome for chronic PTSD among
female assault victims with borderline personality characteristics: A preliminary
investigation. Journal of Personality Disorders, 16, 30-40.
Ferrer, M., Andión, Ó., Matalí, J., Valero, S., Navarro, J. A., Ramos-Quiroga, J. A., … Casas, M.
(2010). Comorbid attention-deficit/hyperactivity disorder in borderline patients defines
an impulsive subtype of borderline personality disorder. Journal of Personality
Disorders, 24, 812-822.
First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. W. (1997). Structured clinical
interview for DSM-IV Axis I disorders: Clinician version. Arlington, VA: American
Psychiatric Publishing.
Fisher, A. J., Newman, M. G., & Molenaar, P. C. M. (2011). A quantitative method for the
analysis of nomothetic relationships between idiographic structures: Dynamic patterns
78
create attractor states for sustained posttreatment change. Journal of Consulting and
Clinical Psychology, 79, 552-563.
Fleeson, W. (2001). Towards a structure- and process-integrated view of personality: Traits as
density distributions of states. Journal of Personality and Social Psychology, 80, 10111027.
Fonagy, P., Gergely, G., Jurist, E., & Target, M. (2004). Affect regulation, mentalization, and the
development of the self. New York: Other Press.
Fonagy, P., & Target, M. (2006). The mentalization-focused approach to self pathology. Journal
of Personality Disorders, 20, 544-576.
Fonagy, P., Target, M., Gergely, G., Allen, J. G., & Bateman, A. W. (2003). The developmental
roots of borderline personality disorder in early attachment relationships: A theory and
some evidence. Psychoanalytic Inquiry, 23, 412-459.
Forsythe, G. E., Malcolm, M. A., & Moler, C. B. (1977). Computer methods for mathematical
computations. Englewood Cliffs, NJ: Prentice Hall.
Fossati, A., Maffei, C., Bagnato, M., Donati, D., Namia, C., & Novella, L. (1999). Latent
structure analysis of DSM-IV borderline personality disorder criteria. Comprehensive
Psychiatry, 40, 72-79.
Gates, K. M., & Molenaar, P. C. M. (2012). Group search algorithm recovers effective
connectivity maps for individuals in homogeneous and heterogeneous samples.
Neuroimage, 63, 310-319.
Gates, K. M., Molenaar, P. C. M., Hillary, F. G., & Slobounov, S. (2011). Extended unified SEM
approach for modeling event-related fMRI data. Neuroimage, 65, 1151-1158.
79
Giesen-Bloo, J., van Dyck, R., Spinhoven, P., van Tilburg, W., Dirksen, C., van Asselt, T., …
Arntz, A. (2006). Outpatient psychotherapy for borderline personality disorder:
Randomized trial of schema-focused therapy vs transference-focused psychotherapy.
Archives of General Psychiatry, 63, 649-658.
Glaser, J.-P., van Os, J., Thewissen, V., & Myin-Germeys, I. (2010). Psychotic reactivity in
borderline personality disorder. Acta Psychiatrica Scandinavica, 121, 125-134.
Grant, B. F., Chou, S. P., Goldstein, R., Huang, B., Stinson, F. S., Saha, T. D., Ruan, W. J.
(2008). Prevalence, correlates, disability, and comorbidity of DSM-IV borderline
personality disorder: Results from the Wave 2 National Epidemiologic Survey on
Alcohol and Related Conditions. Journal of Clinical Psychiatry, 69, 533-545.
Gregory, R. J., Chlebowski, S., Kang, D., Remen, A. L., Soderberg, M. G., Stepkovitch, J., &
Virk, S. (2008). A controlled trial of psychodynamic psychotherapy for co-occurring
borderline personality disorder and alcohol use disorder. Psychotherapy: Theory,
Research, Practice, Training, 45, 28-41.
Grilo, C. M., Levy, K. N., Becker, D. F., Edell, W. S., & McGlashan, T. H. (1996). Comorbidity
of DSM-III-R axis I and II disorders among female inpatients with eating disorders.
Psychiatric Services, 47, 426-429.
Guerrieri, R., Nederkoorn, C., Schrooten, M., Martijn, C., & Jansen, A. (2009). Inducing
impulsivity leads high and low restrained eaters into overeating, whereas current dieters
stick to their diet. Appetite, 53, 93-100.
Gunderson, J. G., Bender, D., Sanislow, C., Yen, S., Rettew, J. B., Dolan-Sewell, R., … Skodol,
A. E. (2003). Plausibility and possible determinants of sudden "remissions" in borderline
patients. Psychiatry, 66, 111-119.
80
Gunderson, J. G., Stout, R. L., McGlashan, T. H., Shea, M. T., Morey, L. C., Grilo, C. M., …
Skodol, A. E. (2011). Ten-year course of borderline personality disorder:
Psychopathology and function from the Collaborative Longitudinal Personality Disorders
study. Archives of General Psychiatry, 68, 827-837.
Gunthert, K. C., Cohen, L. H., Butler, A. C., & Beck, J. S. (2005). Predictive role of daily coping
and affective reactivity in cognitive therapy outcome: Application of a daily process
design to psychotherapy research. Behavior Therapy, 36, 77-88.
Hamaker, E. L., Dolan, C. V., & Molenaar, P. C. M. (2005). Statistical modeling of the
individual: Rationale and application of multivariate stationary time series analysis.
Multivariate Behavioral Research, 40, 207-233.
Hansen, N. B., Lambert, M. J., & Forman, E. M. (2002). The psychotherapy dose-response effect
and its implications for treatment delivery services. Clinical Psychology: Science and
Practice, 9, 329-343.
Harvey, P. D., Greenberg, B. R., & Serper, M. R. (1989). The Affective Lability Scales:
Development, reliability, and validity. Journal of Clinical Psychology, 45, 786-793.
Haynes, S. N., Mumma, G. H., & Pinson, C. (2009). Idiographic assessment: Conceptual and
psychometric foundations of individualized behavioral assessment. Clinical Psychology
Review, 29, 179-191.
Henry, C., Mitropoulou, V., New, A. S., Koenigsberg, H. W., Silverman, J., & Siever, L. J.
(2001). Affective instability and impulsivity in borderline personality and bipolar II
disorders: Similarities and differences. Journal of Psychiatric Research, 35, 307-312.
81
Hoermann, S., Clarkin, J. F., Hull, J. W., & Levy, K. N. (2005). The construct of effortful
control: An approach to borderline personality disorder heterogeneity. Psychopathology,
38, 82-86.
Holtfreter, K., Reisig, M. D., Leeper Piquero, N., & Piquero, A. R. (2010). Low self-control and
fraud: Offending, victimization, and their overlap. Criminal Justice and Behavior, 37,
188-203.
Hopwood, C. J., & Morey, L. C. (2007). Psychological conflict in borderline personality as
represented by inconsistent self-report item responding. Journal of Social and Clinical
Psychology, 26, 1065-1075.
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, 806-815.
Hopwood, C. J., & Zanarini, M. C. (2010). Five-factor trait instability in borderline relative to
other personality disorders. Personality Disorders: Theory, Research, and Treatment, 1,
58-66.
Hörz, S., Zanarini, M. C., Frankenburg, F. R., Reich, D. B., & Fitzmaurice, G. (2010). Ten-year
use of mental health services by patients with borderline personality disorder and with
other Axis II disorders. Psychiatric Services, 61, 612-616.
Howard, K. I., Kopta, S. M., Krause, M. S., & Orlinsky, D. E. (1986). The dose-effect
relationship in psychotherapy. American Psychologist, 41, 159-164.
Johansen, M., Karterud, S., Pedersen, G., Gude, T., & Falkum, E. (2004). An investigation of the
prototype validity of the borderline DSM-IV construct. Acta Psychiatrica Scandinavica,
109, 289-298.
82
Johnson, D. M., Shea, M. T., Yen, S., Battle, C. L., Zlotnick, C., Sanislow, C. A., … Zanarini,
M. C. (2003). Gender differences in borderline personality disorder: Findings from the
Collaborative Longitudinal Personality Disorders study. Comprehensive Psychiatry, 44,
284-292.
Jones, A., Guerrieri, R., Fernie, G., Cole, J., Goudie, A., & Field, M. (2011). The effects of
priming restrained versus disinhibited behaviour on alcohol-seeking in social drinkers.
Drug and Alcohol Dependence, 113, 55-61.
Jöreskog, K. G., & Sörbom, D. (1993). LISREL (Version 8.12). Chicago, IL: Scientific Software
International.
Kelderman, H., & Molenaar, P. C. M. (2007). The effect of individual differences in factor
loadings on the standard factor model. Multivariate Behavioral Research, 42, 435-456.
Kendler, K. S., Zachar, P., & Craver, C. (2011). What kinds of things are psychiatric disorders?
Psychological Medicine, 41, 1143-1150.
Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data analysis in social psychology. In D.
Gilbert, S. Fiske & G. Lindzey (Eds.), Handbook of social psychology (4th ed., pp. 233265). New York: McGraw-Hill.
Kernberg, O. F. (1975). Borderline conditions and pathological narcissism. Northvale, NJ: Jason
Aronson.
Kernberg, O. F. (1984). Severe personality disorders: Psychotherapeutic strategies. New Haven:
Yale University Press.
Kernberg, O. F. (2006). Identity: Recent findings and clinical implications. Psychoanalytic
Quarterly, 75, 969-1004.
83
Koenigsberg, H. W., Harvey, P. D., Mitropoulou, V., New, A. S., Goodman, M., Silverman, J.,
… Siever, L. J. (2001). Are the interpersonal and identity disturbances in the borderline
personality disorder criteria linked to the traits of affective instability and impulsivity?
Journal of Personality Disorders, 15, 358-370.
Kopala-Sibley, D. C., Zuroff, D. C., Russell, J. J., Moskowitz, D. S., & Paris, J. (2012).
Understanding heterogeneity in borderline personality disorder: Differences in affective
reactivity explained by the traits of dependency and self-criticism. Journal of Abnormal
Psychology, 121, 680-691.
Kopta, S. M., Howard, K. I., Lowry, J. L., & Beutler, L. E. (1994). Patterns of symptomatic
recovery in psychotherapy. Journal of Consulting and Clinical Psychology, 62, 10091016.
Korzekwa, M. I., Dell, P. F., Links, P. S., Thabane, L., & Webb, S. P. (2008). Estimating the
prevalence of borderline personality disorder in psychiatric outpatients using a two-phase
procedure. Comprehensive Psychiatry, 49(4), 380-386.
Lavallee, L. F., & Campbell, J. D. (1995). Impact of personal goals on self-regulation processes
elicited by daily negative events. Journal of Personality and Social Psychology, 69, 341352.
Leihener, F., Wagner, A., Haaf, B., Schmidt, C., Lieb, K., Stieglitz, R., & Bohus, M. (2003).
Subtype differentiation of patients with borderline personality disorder using a
circumplex model of interpersonal behavior. Journal of Nervous and Mental Disease,
191, 248-254.
84
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, 553-564.
Levenson, J. C., Wallace, M. L., Fournier, J. C., Rucci, P., & Frank, E. (2012). The role of
personality pathology in depression treatment outcome with psychotherapy and
pharmacotherapy. Journal of Consulting and Clinical Psychology, 80(5), 719-729.
Levy, K. N. (2008). Psychotherapies and lasting change. American Journal of Psychiatry, 165,
556-559.
Levy, K. N., Meehan, K. B., Kelly, K. M., Reynoso, J. S., Weber, M., Clarkin, J. F., & Kernberg,
O. F. (2006). Change in attachment patterns and reflective function in a randomized
control trial of transference-focused psychotherapy for borderline personality disorder.
Journal of Consulting and Clinical Psychology, 74, 1027-1040.
Levy, K. N., Meehan, K. B., Weber, M., Reynoso, J., & Clarkin, J. F. (2005). Attachment and
borderline personality disorder: Implications for psychotherapy. Psychopathology, 38,
64-74.
Linehan, M. M. (1993). Cognitive-behavioral treatment of borderline personality disorder. New
York: The Guilford Press.
Linehan, M. M., Comtois, K. A., Murray, A. M., Brown, M. Z., Gallop, R. J., Heard, H. L., …
Lindenboim, N. (2006). Two-year randomized controlled trial and follow-up of
dialectical behavior therapy vs therapy by experts for suicidal behaviors and borderline
personality disorder. Archives of General Psychiatry, 63, 757-766.
85
Links, P. S., Eynan, R., Heisel, M. J., Barr, A., Korzekwa, M., McMain, S., & Ball, J. S. (2007).
Affective instability and suicidal ideation and behavior in patients with borderline
personality disorder. Journal of Personality Disorders, 21, 72-86.
Links, P. S., Eynan, R., Heisel, M. J., & Nisenbaum, R. (2008). Elements of affective instability
associated with suicidal behaviour in patients with borderline personality disorder.
Canadian Journal of Psychiatry, 53, 112-116.
Links, P. S., Heisel, M. J., & Garland, A. (2003). Affective instability in personality disorders.
American Journal of Psychiatry, 160, 394-395.
Linville, P. W. (1987). Self-complexity as a cognitive buffer against stress-related illness and
depression. Journal of Personality and Social Psychology, 52, 663-676.
Lobbestael, J., Arntz, A., & Wiers, R. W. (2008). How to push someone’s buttons: A comparison
of four anger-induction methods. Cognition and Emotion, 22, 353-373.
Loranger, A. W. (1999). International Personality Disorder Examination. Lutz, FL:
Psychological Assessment Resources.
Lumsden, E. A. (1993). Borderline personality disorder: A consequence of experiencing affect
within a truncated time frame? Journal of Personality Disorders, 7, 265-274.
McGlashan, T. H., Grilo, C. M., Sanislow, C. A., Ralevski, E., Morey, L. C., Gunderson, J. G.,
… Pagano, M. (2005). Two-year prevalence and stability of individual DSM-IV criteria
for schizotypal, borderline, avoidant, and obsessive-compulsive personality disorders:
Toward a hybrid model of axis II disorders. American Journal of Psychiatry, 162, 883889.
McMain, S. F., Guimond, T., Streiner, D. L., Cardish, R. J., & Links, P. S. (2012). Dialectical
behavior therapy compared with general psychiatric management for borderline
86
personality disorder: Clinical outcomes and functioning over a 2-year follow-up.
American Journal of Psychiatry, 169, 650-661.
McMain, S. F., Links, P. S., Gnam, W. H., Guimond, T., Cardish, R. J., Korman, L., & Streiner,
D. L. (2009). A randomized trial of dialectical behavior therapy versus general
psychiatric management for borderline personality disorder. American Journal of
Psychiatry, 166, 1365-1374.
Meda, S. A., Stevens, M. C., Potenza, M. N., Pittman, B., Gueorguieva, R., Andrews, M. M., …
Pearlson, G. D. (2009). Investigating the behavioral and self-report constructs of
impulsivity domains using principal component analysis. Behavioural Pharmacology, 20,
390-399.
Mennin, D. S., & Heimberg, R. G. (2000). The impact of comorbid mood and personality
disorders in the cognitive-behavioral treatment of panic disorder. Clinical Psychology
Review, 20, 339-357.
Miller, G. (2012). The smartphone psychology manifesto. Perspectives on Psychological
Science, 7, 221-237.
Molenaar, P. C. M. (1985). A dynamic factor model for the analysis of multivariate time series.
Psychometrika, 50, 181-202.
Molenaar, P. C. M. (2004). A manifesto on psychology as an idiographic science: Bringing the
person back into scientific psychology, this time forever. Measurement, 2, 201-218.
Molenaar, P. C. M., & Campbell, C. G. (2009). The new person-specific paradigm in
psychology. Current Directions in Psychological Science, 18, 112-117.
Molenaar, P. C. M., Huizenga, H. M., & Nesselroade, J. R. (2003). The relationship between the
structure of interindividual and intraindividual variability: A theoretical and empirical
87
vindication of developmental systems theory. In U. M. Staudinger & U. Lindenberger
(Eds.), Understanding human development (pp. 339-360). New York: Springer.
Molenaar, P. C. M., & Rovine, M. (2011). Introduction to multivariate time series analysis.
Pennsylvania State University. University Park, PA.
Molenaar, P. C. M., Rovine, M. J., & Corneal, S. E. (1999). Dynamic factor analysis of
emotional dispositions of adolescent stepsons towards their stepfathers. In R. K.
Silbereisen & A. Von Eye (Eds.), Growing up in times of social change (pp. 287-318).
Berlin: Walter de Gruyter.
Molenaar, P. C. M., Sinclair, K. O., Rovine, M. J., Ram, N., & Corneal, S. E. (2009). Analyzing
developmental processes on an individual level using nonstationary time series modeling.
Developmental Psychology, 45, 260-271.
Morey, L. C., Berghuis, H., Bender, D. S., Verheul, R., Krueger, R. F., & Skodol, A. E. (2011).
Toward a model for assessing level of personality functioning in DSM-5, part II:
Empirical articulation of a core dimension of personality pathology. Journal of
Personality Assessment, 93, 347-353.
Mount, M. K., Barrick, M. R., Scullen, S. M., & Rounds, J. (2005). Higher-order dimensions of
the big five personality traits and the big six vocational interest types. Personnel
Psychology, 58, 447-478.
Myin-Germeys, I., Oorschot, M., Collip, D., Lataster, J., Delespaul, P., & van Os, J. (2009).
Experience sampling research in psychopathology: Opening the black box of daily life.
Psychological Medicine, 39, 1533-1547.
88
Newhill, C. E., Vaughn, M. G., & DeLisi, M. (2010). Psychopathy scores reveal heterogeneity
among patients with borderline personality disorder. Journal of Forensic Psychiatry &
Psychology, 21, 202-220.
Nezlek, J. B., & Plesko, R. M. (2001). Day-to-day relationships among self-concept clarity, selfesteem, daily events, and mood. Personality and Social Psychology Bulletin, 27, 201211.
Nica, E. I., & Links, P. S. (2009). Affective instability in borderline personality disorder:
Experience sampling findings. Current Psychiatry Reports, 11, 74-81.
Nisenbaum, R., Links, P. S., Eynan, R., & Heisel, M. J. (2010). Variability and predictors of
negative mood intensity in patients with borderline personality disorder and recurrent
suicidal behavior: Multilevel analyses applied to experience sampling methodology.
Journal of Abnormal Psychology, 119, 433-439.
Noblin, J. L., Venta, A., & Sharp, C. (in press). The validity of the MSI-BPD among inpatient
adolescents. Assessment.
Nordberg, S. S., McAleavey, A. A., Castonguay, L. G., & Levy, K. N. (2009, November).
Analysis of the reliability of intake diagnosis at the Penn State Psychological Clinic.
Grand Rounds, Penn State Psychological Clinic.
Patel, A. B., Sharp, C., & Fonagy, P. (2011). Criterion validation of the MSI-BPD in a
community sample of women. Journal of Psychopathology and Behavioral Assessment,
33, 403-408.
Piasecki, T. M., Hufford, M. R., Solhan, M., & Trull, T. J. (2007). Assessing clients in their
natural environments with electronic diaries: Rationale, benefits, limitations, and barriers.
Psychological Assessment, 19, 25-43.
89
Pollock, P. H., Broadbent, M., Clarke, S., Dorrian, A., & Ryle, A. (2001). The Personality
Structure Questionnaire (PSQ): A measure of the multiple self states model of identity
disturbance in cognitive analytic therapy. Clinical Psychology and Psychotherapy, 8, 5972.
Ptacek, J. T., Smith, R. E., Espe, K., & Raffety, B. (1994). Limited correspondence between
daily coping reports and retrospective coping recall. Psychological Assessment, 6, 41-49.
Ram, N., Brose, A., & Molenaar, P. C. M. (2013). Dynamic factor analysis: Modeling personspecific processes. In T. D. Little (Ed.), The Oxford handbook of quantitative methods in
psychology (Vol. 2: Statistical Analysis, pp. 441-457). Oxford: Oxford University Press.
Refinetti, R., Lissen, G. C., & Halberg, F. (2007). Procedures for numerical analysis of circadian
rhythms. Biological Rhythm Research, 38, 275-325.
Reisig, M. D., & Pratt, T. C. (2011). Low self-control and imprudent behavior revisited. Deviant
Behavior, 32, 589-625.
Reynolds, B., Ortengren, A., Richards, J. B., & de Wit, H. (2006). Dimensions of impulsive
behavior: Personality and behavioral measures. Personality and Individual Differences,
40, 305-315.
Robinson, M. D., & Clore, G. L. (2002). Belief and feeling: Evidence for an accessibility model
of emotional self-report. Psychological Bulletin, 128, 934-960.
Roche, M. J., Pincus, A. L., Hyde, A. L., Conroy, D. E., & Ram, N. (2013). Within-person
covariation of agentic and communal perceptions: Implications for interpersonal theory
and assessment. Journal of Research in Personality, 47, 445-452.
Roepke, S., Schröder-Abé, M., Schütz, A., Jacob, G., Dams, A., Vater, A., … Lammers, C.-H.
(2011). Dialectic behavioural therapy has an impact on self-concept clarity and facets of
90
self-esteem in women with borderline personality disorder. Clinical Psychology and
Psychotherapy, 18, 148-158.
Rovine, M. J., Molenaar, P. C. M., & Corneal, S. E. (1999). Analysis of emotional response
patterns for adolescent stepsons using p-technique factor analysis. In R. K. Silbereisen &
A. von Eye (Eds.), Growing up in times of social change (pp. 261-286). Berlin: Walter de
Gruyter.
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.
Salzer, S., Streeck, U., Jaeger, U., Masuhr, O., Warwas, J., Leichsenring, F., & Leibing, E.
(2013). Patterns of interpersonal problems in borderline personality disorder. Journal of
Nervous and Mental Disease, 201, 94-98.
Samuels, J., Eaton, W. W., Bienvenu, O. J., III, Brown, C. H., Costa, P. T., Jr., & Nestadt, G.
(2002). Prevalence and correlates of personality disorders in a community sample. British
Journal of Psychiatry, 180, 536-542.
Sanislow, C. A., Grilo, C. M., Morey, L. C., Bender, D. S., Skodol, A. E., Gunderson, J. G., …
McGlashan, T. H. (2002). Confirmatory factor analysis of DSM-IV criteria for borderline
personality disorder: Findings from the collaborative longitudinal personality disorders
study. American Journal of Psychiatry, 159, 284-290.
Sansone, R. A., Rytwinski, D., & Gaither, G. A. (2003). Borderline personality and psychotropic
medication prescription in an outpatient psychiatry clinic. Comprehensive Psychiatry, 44,
454-458.
91
Santangelo, P., Bohus, M., & Ebner-Priemer, U. W. (In press). Ecological momentary
assessment in borderline personality disorder: A review of recent findings and
methodological challenges. Journal of Personality Disorders.
Scott, L. N., Levy, K. N., & Granger, D. A. (2013). Biobehavioral reactivity to social evaluative
stress in women with borderline personality disorder. Personality Disorders: Theory,
Research, and Treatment, 4, 91-100.
Shannon, C. E. (1998). Communication in the presence of noise. Proceedings of the IEEE, 86,
447-457.
Shea, M. T., Stout, R., Gunderson, J., Morey, L. C., Grilo, C. M., McGlashan, T., … Keller, M.
B. (2002). Short-term diagnostic stability of schizotypal, borderline, avoidant, and
obsessive-compulsive personality disorders. American Journal of Psychiatry, 159, 20362041.
Shea, M. T., Widiger, T. A., & Klein, M. H. (1992). Comorbidity of personality disorders and
depression: Implications for treatment. Journal of Consulting and Clinical Psychology,
60, 857-868.
Shevlin, M., Dorahy, M., Adamson, G., & Murphy, J. (2007). Subtypes of borderline personality
disorder, associated clinical disorders and stressful life-events: A latent class analysis
based on the British Psychiatric Morbidity Survey. British Journal of Clinical
Psychology, 46, 273-281.
Shiffman, S. (2009). How many cigarettes did you smoke? Assessing cigarette consumption by
global report, time-line follow-back, and ecological momentary assessment. Health
Psychology, 28, 519-526.
92
Shiffman, S., Stone, A. A., & Hufford, M. R. (2008). Ecological momentary assessment. Annual
Review of Clinical Psychology, 4, 1-32.
Shiyko, M. P., & Ram, N. (2011). Conceptualizing and estimating process speed in studies
employing ecological momentary assessment designs: A multilevel variance
decomposition approach. Multivariate Behavioral Research, 46, 875-899.
Showers, C. J., & Kling, K. C. (1996). Organization of self-knowledge: Implications for
recovery from sad mood. Journal of Personality and Social Psychology, 70, 578-590.
Skodol, A. E., Clark, L. A., Bender, D. S., Krueger, R. F., Morey, L. C., Verheul, R., …
McGlashan, T. H. (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, 4-22.
Skodol, A. E., Pagano, M. E., Bender, D. S., Shea, M. T., Gunderson, J. G., Yen, S., …
McGlashan, T. H. (2005). Stability of functional impairment in patients with schizotypal,
borderline, avoidant, or obsessive-compulsive personality disorder over two years.
Psychological Medicine, 35, 443-451.
Smith, R. E., Leffingwell, T. R., & Ptacek, J. T. (1999). Can people remember how they coped?
Factors associated with discordance between same-day and retrospective reports. Journal
of Personality and Social Psychology, 76, 1050-1061.
Solhan, M. B., Trull, T. J., Jahng, S., & Wood, P. K. (2009). Clinical assessment of affective
instability: Comparing EMA indices, questionnaire reports, and retrospective recall.
Psychological Assessment, 21, 425-436.
Stein, K. F. (1995). The organizational properties of the self-concept and instability of affect.
Research in Nursing & Health, 18, 405-415.
93
Sterba, S. K., & Bauer, D. J. (2010). Matching method with theory in person-oriented
developmental psychopathology research. Development and Psychopathology, 22, 239254.
Stern, B. L., Clarkin, J. F., Critchfield, K. L., Horz, S., MacCornack, V., Lenzenweger, M. F., &
Kernberg, O. F. (2010). Structured Interview of Personality Organization (STIPO):
Preliminary psychometrics in a clinical sample. Journal of Personality Assessment, 92,
35-44.
Stevenson, J., Meares, R., & Comerford, A. (2003). Diminished impulsivity in older patients
with borderline personality disorder. American Journal of Psychiatry, 160, 165-166.
Stiglmayr, C. E., Ebner-Priemer, U. W., Bretz, J., Behm, R., Mohse, M., Lammers, C.-H., …
Bohus, M. (2008). Dissociative symptoms are positively related to stress in borderline
personality disorder. Acta Psychiatrica Scandinavica, 117, 139-147.
Stolwijk, A. M., Straatman, H., & Zielhuis, G. A. (1999). Studying seasonality by using sine and
cosine functions in regression analysis. Journal of Epidemiology and Community Health,
53, 235-238.
Stone, A. A., Broderick, J. E., Shiffman, S. S., & Schwartz, J. E. (2004). Understanding recall of
weekly pain from a momentary assessment perspective: Absolute agreement, betweenand within-person consistency, and judged change in weekly pain. Pain, 107, 61-69.
Stone, M. H. (1993). Long-term outcome in personality disorders. British Journal of Psychiatry,
162, 299-313.
Tangney, J. P., Baumeister, R. F., & Boone, A. L. (2004). High self-control predicts good
adjustment, less pathology, better grades, and interpersonal success. Journal of
Personality, 72, 271-324.
94
Tennen, H., & Affleck, G. (2002). The challenge of capturing daily processes at the interface of
social and clinical psychology. Journal of Social and Clinical Psychology, 21, 610-627.
Todd, M., Armeli, S., Tennen, H., Carney, M. A., & Affleck, G. (2003). A daily diary validity
test of drinking to cope measures. Psychology of Addictive Behaviors, 17, 303-311.
Todd, M., Armeli, S., Tennen, H., Carney, M. A., Ball, S. A., Kranzler, H. R., & Affleck, G.
(2005). Drinking to cope: A comparison of questionnaire and electronic diary reports.
Journal of Studies on Alcohol, 66, 121-129.
Torgersen, S., Kringlen, E., & Cramer, V. (2001). The prevalence of personality disorders in a
community sample. Archives of General Psychiatry, 58, 590-596.
Tramantano, G., Javier, R. A., & Colon, M. (2003). Discriminating among subgroups of
borderline personality disorder: An assessment of object representations. American
Journal of Psychoanalysis, 63, 149-175.
Trull, T. J. (2001). Structural relations between borderline personality disorder features and
putative etiological correlates. Journal of Abnormal Psychology, 110, 471-481.
Trull, T. J., Jahng, S., Tomko, R. L., Wood, P. K., & Sher, K. J. (2010). Revised NESARC
personality disorder diagnoses: Gender, prevalence, and comorbidity with substance
dependence disorders. Journal of Personality Disorders, 24(4), 412-426.
Trull, T. J., Solhan, M. B., Tragesser, S. L., Jahng, S., Wood, P. K., Piasecki, T. M., & Watson,
D. (2008). Affective instability: Measuring a core feature of borderline personality
disorder with ecological momentary assessment. Journal of Abnormal Psychology, 117,
647-661.
Tyrer, P. (1999). Borderline personality disorder: A motley diagnosis in need of reform. Lancet,
354, 2095-2096.
95
van Os, J., Delespaul, P., Wigman, J., Myin-Germeys, I., & Wichers, M. (2013). Beyond DSM
and ICD: Introducing "precision diagnosis" for psychiatry using momentary assessment
technology. World Psychiatry, 12, 113-117.
Wichers, M. (In press). The dynamic nature of depression: A new micro-level perspective of
mental disorder that meets current challenges. Psychological Medicine.
Widiger, T. A., & Frances, A. J. (1989). Epidemiology, diagnosis, and comorbidity of borderline
personality disorder. In A. Tasman, R. E. Hales & A. J. Frances (Eds.), American
Psychiatric Press Review of Psychiatry (Vol. 8, pp. 8-24). Washington, D.C.: American
Psychiatric Press.
Wigman, J. T. W., van Os, J., Thiery, E., Derom, C., Collip, D., Jacobs, N., & Wichers, M.
(2013). Psychiatric diagnosis revisited: Towards a system of staging and profiling
combining nomothetic and idiographic parameters of momentary mental states. PLoS
One, 8(3). doi:10.1371/journal.pone.0059559
Wilkinson-Ryan, T., & Westen, D. (2000). Identity disturbance in borderline personality
disorder: An empirical investigation. American Journal of Psychiatry, 157, 528-541.
Williams, S. E., Hartstone, M. D., & Denson, L. A. (2010). Dialectical Behavioural Therapy and
borderline personality disorder: Effects on service utilisation and self-reported symptoms.
Behaviour Change, 27, 251-264.
Wright, A. G. C., Hallquist, M. N., Morse, J. Q., Scott, L. N., Stepp, S. D., Nolf, K. A., &
Pilkonis, P. A. (2013). Clarifying interpersonal heterogeneity in borderline personality
disorder using latent mixture modeling. Journal of Personality Disorders, 27, 125-143.
Zanarini, M. C., & Frankenburg, F. R. (2007). The essential nature of borderline
psychopathology. Journal of Personality Disorders, 21, 518-535.
96
Zanarini, M. C., Frankenburg, F. R., Dubo, E. D., Sickel, A. E., Trikha, A., Levin, A., &
Reynolds, V. (1998). Axis I comorbidity of borderline personality disorder. American
Journal of Psychiatry, 155, 1733-1739.
Zanarini, M. C., Frankenburg, F. R., Hennen, J., Reich, D. B., & Silk, K. R. (2004). Axis I
comorbidity in patients with borderline personality disorder: 6-year follow-up and
prediction of time to remission. American Journal of Psychiatry, 161, 2108-2114.
Zanarini, M. C., Frankenburg, F. R., Hennen, J., & Silk, K. R. (2004). Mental health service
utilization by borderline personality disorder patients and Axis II comparison subjects
followed prospectively for 6 years. Journal of Clinical Psychiatry, 65, 28-36.
Zanarini, M. C., Frankenburg, F. R., Khera, G. S., & Bleichmar, J. (2001). Treatment histories of
borderline inpatients. Comprehensive Psychiatry, 42, 144-150.
Zanarini, M. C., Vujanovic, A., Parachini, E. A., Boulanger, J. L., Frankenburg, F. R., &
Hennen, J. (2003). A screening measure for BPD: The McLean Screening Instrument for
Borderline Personality Disorder (MSI-BPD). Journal of Personality Disorders, 17, 568573.
Zimmerman, M., Chelminski, I., Young, D., Dalrymple, K., & Martinez, J. (2012). Does the
presence of one feature of borderline personality disorder have clinical significance?
Implications for dimensional ratings of personality disorders. Journal of Clinical
Psychiatry, 73, 8-12.
Zimmerman, M., Rothschild, L., & Chelminski, I. (2005). The prevalence of DSM-IV personality
disorders in psychiatric outpatients. American Journal of Psychiatry, 162, 1911-1918.
97
Zuur, A. F., Fryer, R. J., Jolliffe, I. T., Dekker, R., & Beukema, J. J. (2003). Estimating common
trends in multivariate time series using dynamic factor analysis. Environmetrics, 14, 665685.
Curriculum Vitae
William D. Ellison
Education
Pennsylvania State University
Harvard Medical School/Massachusetts Mental Health Center
Princeton University
Ph.D., Psychology, December 2013
Predoctoral Internship, 2012-2013
A.B. (History), cum laude, 2003
Selected Honors and Awards
University Graduate Fellowship, Pennsylvania State University
Kenneth I. Howard Memorial Award, North American Society for Psychotherapy Research
Grant Funding
Title: “Modeling Dynamic Processes in Borderline Personality Disorder”
Funding Agency: American Psychoanalytic Association
Dates: 10/2011 – 9/2014
Funding amount: $10,000
Role: Principal Investigator
Title: "Modeling Dynamic Processes in Borderline Personality Disorder"
Funding Agency: International Psychoanalytic Association
Dates: 11/2011 – 10/2014
Funding amount: $4,012
Role: Principal Investigator
Selected Publications
Bernecker, S. N., Levy, K. N., & Ellison, W. D. (In press). A meta-analysis of the relation between patient
adult attachment style and the working alliance. Psychotherapy Research.
Ellison, W. D., Levy, K. N., Cain, N. M., Ansell, E. B., & Pincus, A. L. (2013). The impact of pathological
narcissism on psychotherapy utilization, initial symptom severity, and early-treatment symptom
change: A naturalistic investigation. Journal of Personality Assessment, 95, 291-300.
Ellison, W. D., & Levy, K. N. (2012). Factor structure of the primary scales of the Inventory of Personality
Organization in a nonclinical sample using exploratory structural equation modeling. Psychological
Assessment, 24, 503-517.
Levy, K. N., Ellison, W. D., Scott, L. N., & Bernecker, S. L. (2011). Attachment style. Journal of Clinical
Psychology, 67, 193-203.
Clarkin, J. F., Levy, K. N., & Ellison, W. D. (2010). Personality disorders. In L. M. Horowitz & S. Strack
(Eds.), Handbook of interpersonal psychology: Theory, research, assessment, and therapeutic
interventions (pp. 383-403). New York: Wiley.