Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Apr 1.
Published in final edited form as: J Abnorm Psychol. 2021 Feb 4;130(3):260–272. doi: 10.1037/abn0000656

Symptoms as Rapidly Fluctuating Over Time: Revealing the Close Psychological Interconnections among Borderline Personality Disorder Symptoms via Within-Person Structures

Malek Mneimne 1, Leah Emery 1, R Michael Furr 1, William Fleeson 1
PMCID: PMC8274974  NIHMSID: NIHMS1716284  PMID: 33539116

Abstract

Despite the clinical emphasis on processes happening within individuals, investigations into the psychological, structural connections between mental health symptoms have almost exclusively analyzed differences between people. These investigations have revealed important findings; however, they do not reveal the close connections among symptoms in an individuals’ psychology. This study thus examined the psychological connections between symptoms directly, using borderline personality disorder (BPD) symptoms as an example. Participants (252; 74 with BPD) reported their momentary BPD symptoms five times daily. In support of personalized medicine (Wright & Woods, in press), individuals’ BPD symptom structures differed considerably from each other and from the between-person structure. A novel technique revealed that differences were greater than expected from chance. Within-person structures tended to exhibit more symptom granularity (more factors and lower variance explained) and differing symptom meanings (patterns of loadings). For example, some individuals exhibited close connections between relationship turmoil and identity uncertainty, whereas other individuals exhibited close connections between relationship turmoil and impulsivity. Thus, conceptions of any given person’s psychopathological processes using between-person structural findings will most likely be inaccurate.

General Scientific Summary.

Individuals differed considerably not only in the amount they experienced symptoms, but also in the structure and nature of psychological connections between the symptoms. Thus, conceptions of any given individual based on standard between-person analyses will likely be wrong. Individualized conception of disorders may be required for accurate assessment and treatment.

Keywords: structure, P-technique, within-person, Borderline personality disorder

The purpose of this research is to discover the patterns in which borderline personality disorder (BPD) symptoms repeatedly flare up and subside together within a person over the course of hours, days, or weeks; that is, to discover the within-person structure of BPD symptoms (Brose & Ram, 2012; Fleeson et al., 2019; Nesselroade, 2007; Wright et al., 2015). Secondly, this research evaluates how well the structure of symptoms revealed in traditional between-person factor analyses represents this within-person structure of symptom variation. Thirdly, this research tests whether the psychological connections among symptoms differ significantly across people.

There are at least three reasons this research is important. First, many clinical psychological theories are developed to explain within-person symptom processes in psychopathology. Clinical researchers wonder increasingly about symptom dynamics – how and when symptoms manifest in real time and their contextual triggers e.g., Borsboom et al., 2018; Contreras et al., 2019; Hopwood, 2018; Roche et al., 2016; Russell et al., 2007; Woods et al., in press). If two symptoms covary in a person – continually surging and receding together – then they are part of the same events and psychological experiences. Identifying such close connections between symptoms constitutes a core component of cognitive-behavioral therapies (Haynes et al., 2009) and is essential to advancing definitions of disorders (American Psychiatric Association, 2013) and theories on the nature of psychopathology (Borsboom et al., 2018).

Second, most empirical information about symptom structures (e.g., Contreras et al., 2019; Kotov et al., 2017) is not based on changes over time but on differences between people. Unless the unlikely condition of ergodicity applies, within-person and between-person symptom structures are unlikely to be the same (Fisher et al., 2018; Molenaar, 2004). Symptoms are probably not connected to each other in daily functioning in the way implied by between-person structures.

Third individuals may not even have the same structures as each other, because disorder processes may depend on individual biopsychosocial histories, psychological makeups, and current contexts (e.g., Borsboom et al., 2018; Haynes et al., 2009). The growing trend toward personalized medicine recognizes that different people may have different pathology processes, etiologies, and manifestations of the purportedly same disorder (Fisher, 2015; Wright & Woods, in press). Not only do symptoms differ across people in level or presence, but also in their core meaning and operation. Recently, a few creative and important studies have begun to demonstrate that models applied to individuals’ changes over time reveal heterogeneity in process (Dotterer et al., in press; Ellison et al., in press; Woods et al., in press; Wright et al., in press). In this paper, we contribute to this newly emerging research, and extend it by employing a large sample of individuals, many of whom have BPD, assessing within-person variations in the complete set of BPD symptoms, and using a factor analytic approach. To move beyond description, we present an innovative method for the first test of the statistical significance of these differences – a test needed to demonstrate that heterogeneity is more than random variation but rather represents reliable differences in actual BPD structures.

Close Psychological Connections Between Symptoms

For two variables to be closely psychologically connected within a person, activation of one variable must be accompanied by activation of the other variable at the same times. This requires two conditions. First, the variables must have times they are activated and times they are deactivated within the person (i.e., there must be within-person variation). Second, the variables must become activated and deactivated at the same times (i.e., there must be within-person covariation). For example, if identity uncertainty and relationship turmoil are closely psychologically connected within individuals, then each would have times of more and times of less activation, and the times identity uncertainty was relatively activated would be the times that relationship turmoil was relatively activated.

Such covariation reveals that two symptoms are parts of the same psychological experiences. The covariation might indicate that the symptoms arise from the same contextual triggers. Or it might mean that one symptom causes the other. Or the covariation may arise because the two symptoms have shared meanings for the person. For example, within-person covariation between identity uncertainty and relationship turmoil could arise from identity uncertainty and relationship turmoil reacting to the same events, from identity uncertainty stimulating relationship turmoil or vice versa, or from relationship turmoil and identity uncertainty having shared meanings. In all of these cases, identity uncertainty and relationship turmoil would be closely connected in the individual’s psychology.

Using Between-Person Factor Analysis to Reveal the Structure of Psychopathology

The familiar method for revealing structures of psychological connections between psychopathology variables – between-person factor analysis –does not assess within-person covariation (Fisher, et al., 2018; Fleeson et al., 2019; Molenaar, 2004). Rather, it assesses covariances of differences between individuals, revealing whether individuals who have higher scores than other people on one symptom tend to also have higher scores on other symptoms.

Such analyses of individual differences identify the symptoms that tend to be elevated conjointly in people, and thereby reveal and differentiate disorders (e.g., Kotov et al., 2017). They can determine the coherence of symptoms of a given disorder and the number of dimensions underlying differences between people in the symptoms. For example, many between-person factor analyses of BPD symptoms revealed that individual differences in the nine definitional symptoms of BPD covaried together such that a single dimension explained most of the between-person symptom variance (Hawkins et al., 2014; Clifton & Pilkonis, 2007; Sanislow et al., 2002), although some studies found more than 1 factor (e.g., Becker et al., 2006). A single factor means that people who were higher than other people in any one of the symptoms (efforts to avoid abandonment, relationship turmoil, identity uncertainty, impulsivity, self-harm/suicidality, intense emotions, feelings of emptiness, difficulty controlling anger, and dissociative thoughts) tended to be higher than other people in the other symptoms.

Between-person structures may provide clues regarding the underlying psychological processes that produce disorders such as BPD. They suggest that symptoms on the same between-person factor may share a causal process and may be psychologically connected to each other. Such causal processes are typically, but not necessarily, suggested to be single causes underlying each factor (Borsboom et al., 2018).

However, because between-person analyses are based on differences between people, they may not actually reflect the psychological connections or processes within people. It is only an inference to assume that within-person variation would be structured similarly to between-person variation, and a tentative one at that (Brose & Ram, 2012; Molenaar, 2004; Nesselroade, 2007; Wright & Woods, in press). Statistically, variations across people in symptom levels are independent of variations across time in symptom occurrences, even for the same symptoms. Unless the relatively rare condition of ergodicity is present, generalizations from group structure to individual structure risk Simpson’s paradox or the ecological fallacy (Fisher et al., 2018).

It is tempting to risk the inference that the between-person structure describes the close psychological connections between symptoms. Postulating one set of causal processes may be simpler than postulating two or more parallel sets of causal processes for the same symptoms, especially if symptom levels are aggregates of momentary symptom experiences (Fleeson et al., 2019). Additionally, connections between symptoms might reflect universal processes, whereas individual differences may occur only in symptom levels, not in definitional structures.

Nonetheless, there are several conceptual reasons to believe that between-person symptom structures may not generalize to within-person symptom structures. First, symptom occurrences might have different causal origins than individual differences in symptom levels. For example, individual differences in symptom levels might arise from differing developmental histories, significant events, environmental stresses, and/or genetic dispositions (Crowell et al., 2009). By contrast, momentary flare-ups in symptoms (e.g., an angry outburst at dinner with one’s partner) are likely caused by short-term contextual triggers, such as an interpersonal conflict (Hepp et al., 2016; Miskewicz et al., 2015). If the causes of individual differences in symptom levels differ from the causes of within-person variability in symptom occurrences, they are likely to lead to different patterns of variation and covariation of individual differences in symptom levels versus within-person variation in symptom occurrences. Because factor structures result from covariations, the structures may also differ.

Second, greater differentiation between symptoms (symptom granularity) is likely in daily life than is suggested by a single between-persons factor. When experiencing a moment of relationship turmoil, it is unlikely that most people would simultaneously experience an avalanche of identity uncertainty, impulsivity, emotional intensity, dissociative thoughts, efforts to avoid abandonment, difficulty controlling anger, and feelings of emptiness. Some people might, some times, but probably not everyone, all the time. Rather, the causes of within-person variance seem differentially specific to symptoms. For example, the causes of flare-ups in relationship turmoil (e.g., an argument) appear to differ from the causes of identity uncertainty (e.g., negative feedback on a cherished ability). Differing causes would lead to symptoms occurring at different times, reducing covariation.

Third, symptom meanings and causes may differ across individuals. For some people, identity may be tied to relationships, such that relationship turmoil triggers identity uncertainty. For others, identity uncertainty may be disorienting, producing dissociative thoughts. Discovering such diversity of symptom and trigger meanings is precisely a focus in cognitive-behavioral therapy (Haynes et al., 2009). This diversity suggests that different individuals will have different within-person structures. If different individuals have different symptom structures, then between-person and within-person symptom structures will not always match.

Only very recently have studies begun to probe within-person processes of clinically relevant phenomena. Studies of anxiety, personality disorders, and BPD have found that networks of connections between clinically relevant experiences showed heterogeneity across subjects (Ellison et al, in press; Fisher et al., 2017; Lane et al., 2019; Woods et al., in press; Wright et al., 2016; Wright et al., 2019). However, no study assessed a full complement of BPD symptoms and none evaluated the degree to which within-person structures of the disorder itself matched between-person structures of the disorder. If they do not match, the definitions of disorders are potentially wrong, the dimensions underlying the disorders are potentially wrong, the structural models of purported connections among a disorder’s symptoms are potentially wrong, and the assumed processes constituting the disorders are potentially wrong.

Identifying Within-Person Structures and Comparing Them

To identify individual structures, most studies have used network analysis, such as GIMME (e.g., Woods et al., in press). Network analysis is excellent for identifying the network aspects of within-person processes, but also differs from the technique typically used to analyze between-person data.

By contrast, in this study we identify within-person structures with the same analytic technique used to identify between-person structures: factor analysis. Similarly to between-person factor analysis, within-person factor analysis analyzes the same variables (the symptoms) with the same goals (revealing patterns of covaration to identify factors and factor loadings). The differences are (i) the sources of variation and covariation – differences between people in symptom levels versus changes across moments in symptom occurrences, and (ii) the sources of the data – many people versus one person at many times. Within-person factor analysis reveals how symptom activations and deactivations covary across time for a single person. With multiple participants, there are separate outputs for each person. Using the same analytic technique allows evaluating whether individuals’ structures differ from the between-person structure while holding the analytic technique constant; it also allows for evaluating heterogeneity on familiar factor analytic features.

We use several familiar descriptive measures for comparing individual factor structures. Because symptoms might not typically all co-occur simultaneously and may have differing causes at the momentary level, individuals may differ in the number of factors in their structures of symptoms. Individuals with a single factor are those who tend to experience all symptoms simultaneously in onslaughts of aversive experiences. Individuals with more factors exhibit symptom granularity, experiencing only subsets of symptoms together.

For similar reasons, individuals may differ in and typically have less variance explained by the first factor in their within-person structures. People with lower explained variance have weaker connections among symptoms, potentially experiencing only one or two symptoms at a time.

Differential causes across symptoms and differential meanings and contexts across individuals would be evident in differing patterns of loadings. Dissimilar patterns across individuals indicate that symptoms connect to each other differently. Similar patterns indicate that symptoms connect to each other in similar ways. For example, an individual for whom identity uncertainty and emptiness were central symptoms would have relatively high loadings for those two symptoms and would not match an individual for whom emotional instability and relationship turmoil were central symptoms.

How to Test for Statistical Significance of Structure Differences between People

Although such comparisons reveal the varied meanings of the symptoms for individuals, they do not afford statistical tests. Unfortunately, as Borkenau and Ostendorf (1998) discovered, the number of reports required for a sufficiently powerful test of differences in factor structures is larger than can be reasonably obtained from people. Relatively small samples of occasions per individual mean that sampling error may produce spurious differences between individual structures. Even one-factor structures comprised of eight variables summarize 28 pairwise associations at once. Under ideal conditions, 100 occasions per person would be needed, usually more (Meade & Bauer, 2007). This impedes multi-group confirmatory factor analyses (in which each person is treated as a group). Thus, despite interest in this question for over two decades, researchers have not yet determined whether differences across within-person factors are significant.

We present a novel solution to this problem (introduced briefly in Mneimne et al., 2017). Its basis is the recognition that factor structures emerge from pairwise associations. If one or more pairwise associations differ significantly, then the structures including those associations, by definition also differ for reasons other than sampling error. This is critical, because sufficient power to test the differences between pairwise associations requires many fewer reports per participant.

The differences between individuals in their within-person associations can be tested with multilevel modeling (MLM). The significance test on the variation of within-person associations across persons reveals whether heterogeneity across people is significant, that is, represents more than sampling error. In MLM, the power of this test is substantial. For example, with only 1000 total occasions (e.g., 20 subjects × 50 occasions each), the power to detect between-person variation in associations exceeds .90 (Martin et al., 2011). This power gives space for comfortable corrections for familywise error rate when testing multiple associations, such as interpreting overall patterns rather than single associations or using stricter alpha levels. Increasing to only 50–100 subjects would easily allow such corrections.

Thus, in this paper we test for individual differences in the many pairwise associations between symptoms. If these differ significantly across individuals, then we can be confident that the revealed heterogeneity in the structures that are made out of these pairwise associations are not due to sampling error.

The Person-General Close Psychological Connections

Although heterogeneity is one primary focus of this study, the other primary focus is revealing the close psychological connections between symptoms evident within-person. By concatenating all within-person data together and removing between-person differences, it is possible to identify the close psychological and within-person connections among symptoms for a kind of abstracted person-in-general (Cattell & Scheier, 1961, p. 169).

The resulting factor analysis improves on traditional (between-person) factor analysis by directly revealing close psychological connections among variables. It reveals the close psychological connections between symptoms that occur during daily functioning, rather than distant connections between differences between people. Thus, we will also report this person-general factor analysis.

Wave 1

Method

Participants.

One subsample was recruited from a local outpatient psychiatry clinic and from fliers recruiting participants who experienced moodiness, distrust, being out of control, and difficult relationships (n = 187; see Hawkins et al., 2014). From this subsample, participants had to endorse seven items on the McLean Screening Instrument for BPD to qualify (Zanarini et al., 2003). In the other subsample, participants were recruited from the local community (n = 95) without regard for symptoms. Inclusion criteria for both subsamples included 1) age 18–65, 2) proficiency with English, and 3) within 50 miles. Exclusion criteria included 1) a court-appointed guardian, 2) current alcohol or substance dependency, 3) current psychotic disorder, 4) urgent suicidal ideation, 5) score below 24 on the Mini-Mental Status Examination, 6) inability to complete experience-sampling reports, and 7) arrest for a violent crime.

Of 282 recruited participants, 252 completed at least 15 valid reports. The sample (67.5% female; Mage = 44 years, SD = 11.1 years) was racially diverse (60.3% White, 33.7% Black, 4.4% Mixed, 1% Asian, 1% American Indian) and economically diverse (median annual income approx. $25,000–$29,000, range = below $0 to $149,000). Seventy-four participants met criteria for BPD according to their scores on the Structured Interview DSM-IV Personality (SIDP-IV; Pfhol et al., 1997).

Procedure.

Participants completed questionnaires used for other purposes and partook in diagnostic interviews using the MINI (Sheehan et al., 1998) and SIDP-IV. Participants carried a Palm-Pilot for 14 days and were instructed to complete experience sampling reports at five specific times per day (i.e., 10 a.m., 1 p.m., 4 p.m., 7 p.m., and 10 p.m.). They could complete reports up to 3 hours after the specified time, but were instructed to report on the 60 minutes before the prompt. Participants received up to $175 in gift cards. Participants also participated in other aspects of the study not reported here.

Experience Sampling Method (ESM).

Participants answered questions about the previous 60 minutes. Momentary BPD symptoms were measured using a 6-point scale from 1 (does not describe me at all) to 6 (describes me very well). Each of the nine symptoms was measured by two items (e.g., “I felt hollow inside in the last 60 minutes”), except self-harm, which was measured by one or three items for another study. Cronbach’s alphas on within-person centered symptoms ranged from .28 (self-harm/suicidality) to .88 (feelings of emptiness) and on mean levels of symptoms they ranged from .69 (dissociative thoughts) to .98 (feelings of emptiness). Because self-harm had both low frequencies and low within-person reliabilities, its covariance was negligible and we did not include it in the analyses. Eight situational triggers were measured by one item each (e.g., “Someone rejected me or left me out,” “I was alone and isolated from others”), and trigger intensity was the mean of these triggers.

Results

Between-person factor structure.

The between-person factor structure was based on the variation between people in average levels of each symptom across all reports. Thus, the factor analysis of these levels was a between-persons analysis built on the same data used in the within-person analyses, but it analyzed the between-person variation in those data (Cattell & Scheier, 1961).

A principal components factor analysis revealed only a single factor with an eigenvalue over 11, confirmed by the scree plot (a similar but slightly different analysis was reported in Hawkins et al., 2014). This factor explained 66% of the variance in the symptoms (additional factors would each explain 1–9%); communalities ranged from .55 (impulsivity) to .75 (identity uncertainty). Thus, differences between individuals in symptom levels were strongly associated with each other, such that people who experienced high levels of one symptom tended to experience high levels of the other symptoms. Figure 1 presents the factor loadings, revealing the highest loadings for identity uncertainty, dissociative thoughts, relationship turmoil, and unstable emotions. These symptoms were the most strongly related to the other symptoms, and reflect a central meaning of the factor.

Figure 1.

Figure 1.

Open in a new tab

Loadings of symptoms from between-person principal components exploratory factor analyses, within-person principal-components exploratory factor analyses, and a multilevel confirmatory factor analysis (MCFA). The within-person structures come from data concatenated across all individuals. Darker lines show between-person structures; lighter lines show within-person structures. The patterns were very similar across all analyses, but the within-person structures had lower magnitude loadings.

Do individuals differ in their symptom structures?

We next conducted complete, distinct factor analyses on each individual. Because there is no between-person variation in one person, the resulting factor structure is within-person and reveals the symptom interconnections for a single individual.

Only participants with at least some variation on all of the symptoms (not including the excluded self-harm variable) were included in this analysis. A finding already strikingly indicative of heterogeneity in symptom structure was that 137 of the 252 participants had no variance at all on at least one symptom and six participants had no variance on at least one symptom across the non-missing reports of another symptom.

For each of the remaining 109 participants, we conducted a principal components factor analysis (with oblimin rotation when required). Descriptively, the factor structures differed from each other and from the between-person structure. Figure 2 shows results for two participants. These participants differed from each other and from the between-person structure in the number of factors, pattern of loadings, and highest loading variables. Even though the first participant’s structure consisted of only one factor with all symptoms loading on that factor, the pattern of loadings was unique, e.g., avoiding abandonment was nearly unrelated to this participant’s other symptoms. The second participant’s symptom dynamics were best captured by three factors. The first factor was similar to the between-person factor, except that impulsivity, difficulty with anger, and dissociative thoughts were not connected to relationship turmoil, identity uncertainty, and intense emotions. The second factor revealed that efforts to avoid abandonment was a relatively isolated symptom, and the third factor revealed a distinctly negative connection between impulsivity and difficulty with anger.

Figure 2.

Figure 2.

Open in a new tab

Example individual within-person factor structures. Subject 1 had a single factor, but with a different configuration than the between-persons structure. Subject 2 had three factors, with the first one configuraly similar to the between-persons structure but the second factor connecting abandonment avoidance and relationship turmoil to each other, and the third factor connecting difficulty with anger and impulsivity together negatively.

Individual differences in symptom granularity.

The number of factors indicates the degree to which the symptoms formed distinct clusters of interconnected symptoms. Figure 3 shows the number of participants whose structures consisted of each number of factors. Only nine participants needed only one factor. Instead, nearly all participants’ variance was best explained by more factors than in the between-person analysis. Thus, individuals exhibited more symptom granularity in their on-going psychological functioning than was implied by the between-person structure.

Figure 3.

Figure 3.

Open in a new tab

The psychological functioning of individuals consisted of more distinctions among symptoms than implied by between-person factor structures.

Individual differences in the psychological closeness of symptoms.

The amount of variance explained by the first factor provides a more continuous indicator of symptom closeness. This indicator also varied considerably across participants (Figure 4). For a small number of participants (14), the first factor explained > 60% of the variance, as implied by the between-person structure. For most participants, < 45% of the variance was explained by the first factor, such that the symptoms were connected to each other less closely than suggested by the between-person structure.

Figure 4.

Figure 4.

Open in a new tab

The overall amount of psychological connection among symptoms varied considerably across participants.

Individual differences in the meanings of factors and symptoms.

To assess the degree to which meanings of factors differed across individuals, we correlated each individuals’ patterns of loading with the between-person pattern. Each correlation summarizes the degree to which the within-person factor had a similar meaning to the between-person factor. To get the best chance of finding a positive match, the closest match was selected for each participant. Closest matches correlated between −.27 and .84, M = .37. Thus, some individuals had at least one factor with a similar meaning to the between-person factor, but many had completely unrelated meanings. The average closest match in meaning to the between-person factor was only modest.

Are differences between individuals reliable?

To test whether these individual differences in structure were greater than would be expected from chance, we conducted an MLM for each pair of symptoms, with one symptom predicting another, and symptom reports nested within participants. Symptom relationships were allowed to vary across individuals, because the amount and significance of variation in these relationships indicates whether individuals differed from each other.

For every symptom pair, variation across subjects was significant (Table 1, first column).2 Thus, the psychological connections between symptom pairs differed across participants. Therefore, the structures consisting of these pairs also differed significantly across participants.

Table 1.

Individual differences in slopes, by symptom pair and required minimum amount of variance.

Minimum Variance Required to be Included in the Analysis
Symptom Pair >0 .1 .5 1.0
1 → 2 .23*** .16** .15* .15†
1 → 3 .22*** .19*** .19** .21*
1 → 4 .24*** .19*** .19** .15
1 → 6 .21*** .17** .16* .15†
1 → 7 .27*** .22** .21* .18†
1 → 8 .22*** .20** .20* .14
1 → 9 .21*** .19*** .19* .18
2 → 3 .26*** .24*** .24*** .27**
2 → 4 .27*** .26*** .28*** .28**
2 → 6 .30*** .26*** .26*** .27***
2 → 7 .30*** .28*** .28*** .30**
2 → 8 .34*** .28*** .24*** .22**
2 → 9 .21*** .19*** .20** .23*
3 → 4 .31*** .23*** .18** .20*
3 → 6 .39*** .34*** .30*** .24**
3 → 7 .37*** .32*** .26*** .26**
3 → 8 .42*** .35*** .27** .26*
3 → 9 .29*** .23*** .20** .22*
4 → 6 .35*** .32*** .30*** .28**
4 → 7 .27*** .24*** .23** .26**
4 → 8 .41*** .39*** .34*** .28**
4 → 9 .21*** .20*** .22** .25*
6 → 7 .31*** .30*** .29*** .28***
6 → 8 .31*** .28*** .24*** .21***
6 → 9 .19*** .18*** .16*** .16*
7 → 8 .32*** .31*** .30*** .28**
7 → 9 .24*** .21*** .20*** .24**
8 → 9 .18*** .16*** .18** .20*
Open in a new tab

Note. Table entries are standard deviations of slopes. They reveal the amount individuals differ from each other in the association between the two symptoms indicated on the left. N’s were successively smaller in the columns to the right and varied across symptom pairs (see Table 2).

***

p < .001

**

p < .01

*

p < .05

†

p < .10

This analysis included all participants, even those with little to no variance in the two symptoms. Covariances might have differed across participants largely because variances differed, however. To evaluate this possibility, we repeated the analyses on participants with more variance in the two variables. This also concentrated the sample on those who had the highest levels of symptoms. However, it also shrank the sample, lowering power to detect differences. Table 2 shows the number of participants available for each symptom at different minimum variances. As can be seen in Table 1, most differences between people remained high and significant across increasingly stringent inclusion limits. Most non-significant variances involved the symptom of avoiding abandonment. Thus, even limited to participants with considerable symptom variation, within-person structures differed significantly across people.

Table 2.

Number of participants having at least a minimum variance in the symptom, for different minimum amounts.

Minimum Variance
Symptom 0 .1 .5 1.0
1. Avoiding Abandonment 252 112 68 41
2. Relationship Turmoil 252 150 95 62
3. Identity Uncertainty 252 124 75 42
4. Impulsivity 252 146 83 53
6. Intense Emotions 252 189 139 105
7. Feelings of Emptiness 252 152 106 73
8. Difficulty with Anger 252 154 99 67
9. Dissociative Thoughts 252 121 62 29
Open in a new tab

Note. Table entries are numbers of participants with at least the minimum variance on the indicated variable. Analyses will have fewer participants because participants must have sufficient variance on both of the involved variables in order to be included in the analysis.

The person-general structure of the within-person covariances.

All 11,146 within-person reports were listed consecutively, as though they were from the same individual. Between-person variation was removed by centering each symptom within person – subtracting each person’s mean for a symptom from each of that person’s instances of the symptom. This allowed identifying the person-general but still within-person connections between daily variations in symptoms.

An exploratory factor analysis was conducted on the concatenated data first in order to discover whether a structure emerged that was similar to the between-person structure. A single factor emerged. As can be seen in Figure 1, the pattern of loadings was also very similar to the pattern in the between-person structure (configural similarity = .69, Hartley & Furr, 2017). However, the overall magnitude of the loadings was markedly lower (general saturation similarity = .19). This factor similarly explained only 40% of the variance in symptoms.

Because such an analysis is relatively untested, we verified this finding with another, confirmatory technique. A multilevel confirmatory factor analysis (MCFA) takes advantage of the multilevel, nested nature of the data to evaluate the fit of a model simultaneously at both the between-person and the within-person level. All symptoms were set to load on a single factor at both levels, with no residual correlations at either level, using the lavaan package version 0.6–4 in R (Rosseel, 2012). Fit indices were marginal (χ2 = 1,796.017, p < .001, CFI = .884, TLI = .860, RMSEA = .064), with the largest modification indices at the within-person level.

With two freely estimated residual correlations at the within-person level, the modified unidimensional model fit well (χ2 = 966.756, p < .001, CFI = .941, TLI = .927, RMSEA = .047). Thus, the unidimensional model fit very well at the between-person level but less well at the within-person level. Additionally, the loadings were again much lower at the within-person level than at the between-person level (see Figure 1). Both techniques converged on the finding that the average within-person covariation produces a similar pattern of – but substantially suppressed – loadings to the between-person structure.

Trigger-factor contingencies.

In our model, within-person factors reflect clusters of symptoms that occur jointly as short-term events. The proximal mechanisms of BPD are the flare-ups of these factors in response to triggers such as being rejected or offended (Fleeson et al., 2019; Miskewicz et al., 2015). For example, a person might become intensely emotional, angry, and have dissociative thoughts when another threatens to leave a relationship. Although the purpose of this paper was to identify within-person structures, we conducted additional analyses to explore the contingencies of these factors on triggers as an illustration. This analysis overlaps with analyses presented in Miskewicz et al., (2015), but here we test whether the newly-identified and individually-unique factors are responsive to triggers.

Factor scores were saved for each moment for each individual based on the individual’s factor solution. MLM predicted variation in each person’s first factor scores from the trigger intensity the person experienced at each moment. The fixed slope of this analysis was significant, b = .68, p < .001, supporting the suggestion that these factors were sensitive to triggers. Subjects also differed significantly in the slope, such that some people responded minimally with their first factor to triggers and others responded strongly, presumably reflecting differing meanings of the factors across individuals, SD = .26, p < .001 (b’s thus ranged from about .42 to about .94 for about 2/3 of the sample).

Wave 2

Method

Wave 2’s method was very similar. Approximately 18 months after Wave 1, 165 of the original 282 plus 17 replacement participants completed at least 15 valid reports. Participants also completed other measures not reported here.

The ESM items were the same, except one item about relationship turmoil was reworded and participants described the previous 3 hours rather than 60 minutes. The ESM lasted only one week (some participants continued past one week). Cronbach’s alphas (centered within-person): .43 (dissociative thoughts) to .85 (feelings of emptiness); on mean levels, .71 (dissociative thoughts) to .99 (feelings of emptiness).

Results

Between-person structure.

Principal components factor analysis of symptom mean levels revealed a single factor with an eigenvalue over 1. This factor explained 68% of the variance (additional factors each explaining 1–10% of the variance); communalities ranged from .48 (impulsivity) to .81 (identity uncertainty). Wave 2 thus replicated the strong associations between individual differences in BPD symptoms (Figure 1).

Individual differences in structures.

Of the 165 participants, 104 had no variance on at least one variable and two had no variance on at least one symptom across the non-missing reports of another symptom. A principal components factor analysis with oblimin rotation analyzed each of the remaining 59 individuals’ symptoms. Only 5 participants had only one factor with an eigenvalue > 1, 20 participants had two, 25 participants had three, 8 participants had four, and 1 subject needed five factors to explain his or her symptom variance. Thus, participants exhibited more symptom granularity than the between-person factor structure suggested.

Similarly, the amount of variance explained by the first factor varied considerably across participants (M = 45.6%, SD = 14.6%, Range = 24.2% – 82.8%). Participants differed quite a bit in how closely the symptoms were linked together.

Each participant’s loadings on each factor were correlated with the between-person loadings and the highest match was selected for each person. Best matches ranged from −.89 to .90 (M = .05; SD = .44).3 Thus, the average person’s best matches to the between-person factor were only moderate.

These individual differences in structures were significant in the novel pairwise comparison technique. Table 3 shows the results of MLMs with one of the symptoms predicting another symptom. The first two columns show that the variation across subjects in covariances was significant when including all participants. In Wave 2, the very small number of participants with high variance on both of the two associated variables (Table 4) reduced power. Consequently, fewer of these variances were significant. In sum, the close psychological connections between pairs of symptoms differed across participants. Thus, heterogeneity in the structures made out of these pairs was not produced by sampling error.

Table 3.

Wave 2 individual differences in slopes, by symptom pair and required minimum amount of variance.

Minimum Variance Required to be Included in the Analysis
Symptom Pair >0 .1 .5 1.0
1 → 2 .30*** .23** .21† a
1 → 3 .27*** .27** .31† .30
1 → 4 .27*** .22** .23 a
1 → 6 a .19† .18 a
1 → 7 .26*** .23** .24 .20
1 → 8 .36*** .30* .24 .22
1 → 9 .26*** .21* .26 .23
2 → 3 .28*** .27*** .21† .20
2 → 4 .28*** .20* .06 a
2 → 6 .36*** .29*** .28** .22†
2 → 7 .33*** .28** .27† .18
2 → 8 .41*** .34*** .28** .17
2 → 9 .26*** .22** .20† .17
3 → 4 .38*** .31** .26† .14
3 → 6 .64*** .34** .24* a
3 → 7 a .30** .25* .23
3 → 8 .71*** .39** .32* .30
3 → 9 .44*** .27** .17 .10
4 → 6 .40*** .36*** .30** .24†
4 → 7 .27*** .24** .23† .a
4 → 8 .46*** .45*** .40** .38*
4 → 9 .21*** .21** .23* .10
6 → 7 .30*** .28*** .25** .23*
6 → 8 a a a a
6 → 9 .27*** .21*** .21* .15
7 → 8 .44*** .41*** a a
7 → 9 .30*** .28*** .29** a
8 → 9 .27*** .21** .21* .17
Open in a new tab

Note. Table entries are standard deviations of slopes. They reveal the amount individuals differ from each other in the association between the two symptoms indicated on the left. N’s were successively smaller in the columns to the right and varied across symptom pairs (see Table 4).

***

p < .001

**

p < .01

*

p < .05

†

p < .10

a = failed to converge after 1000 iterations.

Table 4.

Wave 2 number of participants having at least a minimum variance in the symptom, for different minimum amounts.

Minimum Variance
Symptom 0 .1 .5 1.0
1. Avoiding Abandonment 165 64 29 17
2. Relationship Turmoil 165 85 46 24
3. Identity Uncertainty 165 61 25 12
4. Impulsivity 165 89 45 22
6. Intense Emotions 165 114 74 44
7. Feelings of Emptiness 165 81 39 28
8. Difficulty with Anger 165 90 58 40
9. Dissociative Thoughts 165 60 32 14
Open in a new tab

Note. Table entries are numbers of participants with at least the minimum variance on the indicated variable. Analyses will have fewer participants because participants must have sufficient variance on both of the involved variables in order to be included in the analysis.

The person-general, within-person structure.

An exploratory principal components factor analysis on the 5013 person-centered reports revealed clear evidence for a single factor. This factor explained 40.1% of the variance. A MCFA set all symptoms to load on a single factor at both levels. Fit indices were marginal (χ2 = 998.902, p < .001, CFI = .888, TLI = .866, RMSEA = .064. Freeing three within-person residuals again improved fit (χ2 = 966.756, p < .001, CFI = .941, TLI = .927, RMSEA = .047); loadings were again much lower at the within-person level than at the between-person level (see Figure 1). Figure 1 shows the close correspondence to Wave 1 in loading pattern, in two analytic techniques, despite waves occurring about 18 months apart.

Discussion

Despite an emphasis on the individual in mental illness nosologies (e.g., APA, 2013) and in clinical contexts (Haynes et al., 2009; Russell et al., 2007), research investigating the structure of psychopathology has almost exclusively assessed between-person variability in general levels of symptoms. Findings yielded by such studies may not validly reveal how symptoms are connected to each other in individuals’ on-going psychological functioning and thus may not be highly useful to psychological care (Fisher et al., 2018; Molenaar, 2004; Nesselroade, 2007; Wright & Woods, in press).

This study found that, indeed, between-person structures of BPD symptoms appear not to validly reveal the close psychological connections between these symptoms. The way symptoms connected to each other in on-going psychopathological processes was not the way suggested by between-person factors. Conclusions about the nature or processes of BPD on the basis of between-person factor structures are questionable. The definitions of disorders are potentially wrong, the dimensions underlying the disorders are potentially wrong, the structural models of purported connections among a disorder’s symptoms are potentially wrong, and the assumed processes constituting the disorders are potentially wrong.

Although the between-person factor structure replicated the often-found single-factor solution (e.g., Clifton & Pilkonis, 2007; Hawkins et al., 2014), suggesting that the symptoms have close psychological connections to each other, most individuals required more than one factor to account for the variance in symptoms. Thus, most individuals exhibited more and stronger distinctions between symptoms than was suggested by the between-person structure. These findings confirm expectations that BPD symptoms would not all flare up and diminish at the same times for all people, and that the causes of symptom flare-ups are likely to be symptom- and person-specific (Wright & Woods, in press). For example, some individuals’ identities may be interwoven with connections to others, such that identity uncertainty means threats to those interconnections and thus covaries with relationship turmoil. Conversely, for others, focusing on identity may be a calming oasis, while emotional intensity represents a destabilizing threat to identity, resulting in identity uncertainty covarying with emotional intensity. Additionally, individual differences in symptom structures may be caused by individual differences in contextual variables.

The burgeoning personalized medicine approach to understanding psychopathology has begun to reveal that within-person processes in clinical phenomena are heterogeneous across individuals (Dotterer et al., in press; Ellison et al., in press; Woods et al., in press; Wright et al., 2015). This study complements this growing literature by testing heterogeneity in the constitution of BPD. It uses familiar factor analytic techniques rather than network approaches to allow comparability to the typical between-person techniques and to extend across approaches. In addition, it was the first study to employ a large sample of both those with and without BPD, collect two waves of ESM data, and to assess the full set of symptoms of BPD. Even with the same factor analytic techniques used in between-person analyses, and even with assessment of the defining symptoms of BPD, within-person structures differed remarkably across individuals. Thus, heterogeneity cannot be dismissed as tied to a specific analytic method.

Using a new analytic approach, we tested the significance of these distinctions across individuals. Almost every pairwise association between symptoms differed significantly across individuals. Because the factorial structures emerge from pairwise associations, and because the pairwise associations differed significantly across participants, we can conclude that the differences across participants in structures were not due to chance. This was also true when limited to the participants with substantial variance on the symptoms, who would almost entirely be those with high levels of BPD symptoms.

Interestingly, when combining all the within-person data together to create a person-general but still within-person structure of close psychological connections, the pattern of correlations was very similar to the pattern of between-person structure. The main prominent difference was a depressed magnitude of loadings. This raises the intriguing puzzle as to why the person-general pattern was similar to the between-person pattern, but very few individuals’ patterns were similar to the between-person pattern.

The simplest and perhaps most compelling explanation is that the person-general within-person structure is the correct one, and the individual variations from it were normal random variations due to small samples (as in any situation with mean and error variance).4 This would mean that heterogeneity is no more than error variance around the common structure. It would not represent actual differences between individuals in processes. This is one reason the significance tests on the individual differences in pairwise associations were so important. Because most of them were significant, we can finally conclude that heterogeneity was not due to sampling error.

We are currently working on the mathematics to another possible solution, which could ironically be described as the ecological fallacy in reverse. In this solution, the person-general within-person structure is the average of the various within-person covariations. The between-person solution then results from the average of the within-person covariations. There is certainly no mathematical necessity that this would happen, but we are trying to mathematically determine the conditions under which it could be true, and to collect empirical data relevant to those conditions.

Implications for the Nosology of BPD

The findings suggest it is simply inaccurate to conclude that the between-person structure describes the structure of symptoms within individuals. Rather, the constitution of BPD varied across people. The heterogeneity between participants’ factor structures can be understood in terms of individual differences in a variety of distal and proximal contextual variables, including culture, biology, developmental and learning histories, internal and external triggers, and (dys)regulatory processes. For example, one participant’s structure may reflect a learned pattern, where in some situations, he or she feels empty and paranoid and engages in efforts to avoid abandonment (factor 1), but when faced with relationship problems, he or she experiences intense emotions and becomes angry (factor 2).

This does not imply a free-for-all, wherein BPD can mean anything and there is no similarity among BPD sufferers. Although within-person covariances differed considerably across individuals, extremely divergent covariances were rare. As shown in Tables 1 and 3, the standard deviations on the connections were often around .30, meaning that most covariances were in the same direction but differed in magnitude. For example, most people showed positive connections between experiencing intense emotions and relationship turmoil (approximately 2/3 of the participants had a positive beta between .33 and .83). This suggests limits to the forms of BPD that commonly exist.

Implications for Clinical Assessment and Treatment

This research brings structural research closer to clinical approaches (Fleeson et al., 2019; Hopwood, 2018; Roche et al., 2016; Wright et al., 2015). Clinicians often utilize clinical interviews and observation to arrive at an understanding of an individual’s problematic associations. They often do not assume that a simple, single underlying cause is responsible for their patients’ symptoms, but rather that symptoms result from a variety of distal and proximal contextual variables. They often also personalize their approach to the particular individual in front of them (Wright & Woods, in press).

The approach to structure in this paper aligns more closely with and supports such clinical approaches. It assumed and revealed that different individuals had different structures, that these structures arose from unique, person-specific variables, and that understanding the individual requires understanding these structures. When clinicians judge that a particular client is not well-described by the between-person structure, and want to use a personalized approach, their judgments are supported by these findings.

This study also demonstrates the utility of ESMs, coupled with within-person analyses, for providing a rich and direct empirical window into symptom dynamics (Trull, 2018). Integration of this methodology into clinical practice may be challenging, but ultimately rewarding, as it may make psychotherapy more efficient and empirically-based. In this study, symptom factors varied with momentary triggers, and individuals differed in their symptom factor responses to triggers. Following an initial screening, an individualized ESM could be developed and administered in an effort to better understand someone’s symptom dynamics (Haynes, et al., 2009). The results could suggest ideas about proximal treatment targets, such as a person’s most variable symptoms or their highest symptom correlations.

Limitations and Directions for Future Research

Although this study identified psychological connections between symptoms, future research is needed to address the causes of those connections. Symptoms might be connected because the occurrence of one causes another, because they have overlapping meanings, or because they arise from the same contexts. For example, emotional intensity and identity uncertainty might be connected because they both mean being out of control for that person or because the individual typically reacts to rejection with both intense emotions and doubts about identity.

Another limitation is that although we ruled out sampling error as a potential explanation for individual differences in structures, we did not rule out individual differences in measurement error as an explanation. More differentiated factor structures might have been a result of greater measurement error rather than psychological differences among the symptoms. A third limitation was that small case-to-variable ratios utilized for the within-person factor analyses added some unreliability to those analyses, and differed across participants. There was enough reliability to demonstrate that individual differences were significant, but not enough to produce highly stable estimates of individuals’ factor structures. Finally, the nature of BPD as a between-persons concept is also debated, especially its existence as a categorical concept and the possibility that BPD may be a stand-in for more general psychopathology (Hopwood & Krueger, 2018). Additional studies to assess dimensional concepts of BPD and other disorders are needed.

Conclusions

Identifying how symptoms connect in daily life is a core component of cognitive-behavioral therapies and is essential to advancing the definitions and theories of disorders. Most researchers use between-person findings to understand symptom connections; however, doing so requires an inference to be made. This study revealed that inferences about BPD symptom connections that are based upon the between-person structure will usually be inaccurate. Although there was an average within-person structure, most people deviated from it. Furthermore, it is important to stress that these deviations were not in the levels of BPD symptoms, but rather in the constitutions of BPD and in the natures of its symptoms. The novel statistical technique demonstrated that these differences were significant. Similar research on other symptoms and disorders would be beneficial.

Acknowledgments

Research presented in this manuscript was supported by the National Institute of Mental Health of the National Institutes of Health (R01MH70571). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Research was approved by the Wake Forest School of Medicine, IRB00012110. We thank Jennifer L. Wages and Michelle Anderson for their contributions.

Footnotes

1

Using the eigenvalue > 1 criterion is arbitrary, but what matters is that the same criterion was used for the within-person factor structures, allowing fair comparability of solutions. We also consider variance explained by the first factor in order to provide a less arbitrary metric, and present a multilevel confirmatory factor analysis later.

2

A conservative Bonferroni correction for conducting 28 tests would yield an alpha level of .0017, still revealing most of these results to be significant.

3

Without the two negative outliers, there was still substantial variability, M = .45; SD = .23.

4

Cattell, for example, believed this to be the correct explanation (Cattell & Scheier, 1961, p. 153).

References

  1. American Psychiatric Association (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Arlington, VA: American Psychiatric Publishing. [Google Scholar]
  2. Becker DF, McGlashan TH, & Grilo CM (2006). Exploratory factor analysis of borderline personality disorder criteria in hospitalized adolescents. Comprehensive Psychiatry, 47, 99–105. doi: 10.1016/j.comppsych.2005.07.003. [DOI] [PubMed] [Google Scholar]
  3. Borkenau P, & Ostendorf F (1998). The Big Five as states: How useful is the five-factor model to describe intra-individual variations over time? Journal of Personality Research, 32, 202–221. [Google Scholar]
  4. Borsboom D, Cramer A, & Kalis A (2018). Brain disorders? Not really…why network structures block reductionism in psychopathology research. Behavioral and Brain Sciences, 1–54. [DOI] [PubMed] [Google Scholar]
  5. Brose A, & Ram N (2012). Within-person factor analysis: Modeling how the individual fluctuates and changes across time. In Mehl MR & Connor TS (Eds.), Handbook of research methods for studying daily life (pp. 459–478). New York, N.Y.: Guilford Publications. [Google Scholar]
  6. Cattell RB, & Scheier IH (1961). The meaning and measurement of neuroticism and anxiety. Ronald. [Google Scholar]
  7. Clifton A, & Pilkonis PA (2007). Evidence for a single latent class of Diagnostic and Statistical Manual of Mental Disorders borderline personality pathology. Comprehensive Psychiatry, 48, 70–78. [DOI] [PubMed] [Google Scholar]
  8. Contreras A, Nieto I, Valiente C, Espinosa R, & Vasquez C (2019). The study of psychopathology from the network analysis perspective: A systematic review. Psychotherapy and Psychosomatics, 88, 71–83. [DOI] [PubMed] [Google Scholar]
  9. Crowell SE, Beauchaine TP, & Linehan MM (2009). A biosocial developmental model of borderline personality: Elaborating and extending Linehan’s theory. Psychological Bulletin, 135, 495–510. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Dotterer HL, Beltz AM, Foster KT, Simms LJ, & Wright AGC (in press). Personalized models of personality disorders: Using a temporal network method to understand symptomatology and daily functioning in a clinical sample. Psychological Medicine. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Ellison WD, Levy KN, Newman MG, Pincus AL, Wilson SJ, & Molenaar PCM (in press). Dynamics among borderline personality and anxiety features in psychotherapy outpatients: An exploration of nomothetic and idiographic patterns. Personality disorders. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Fisher AJ (2015). Toward a dynamic model of psychological assessment: Implications for personalized care. Journal of consulting and clinical psychology, 83(4), 825. [DOI] [PubMed] [Google Scholar]
  13. Fisher AJ, Medaglia JD, & Jeronimus BF (2018). Lack of group-to-individual generalizability is a threat to human subjects research. Proceedings of the National Academy of Sciences of the United States of America, 115, E6106–E6115. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Fisher AJ, Reeves JW, Lawyer G, Medaglia JD, & Rubel JA (2017). Exploring the idiographic dynamics of mood and anxiety via network analysis. Journal of Abnormal Psychology, 126(8), 1044–1056. [DOI] [PubMed] [Google Scholar]
  15. Fleeson W, Furr RM, Mneimne M, & Arnold EM (2019). Using basic personality process models to inform the personality disorders: Core momentary stressor-symptom contingencies as basic etiology. In Lynam DR & Samuel DB (Eds.). Using basic personality research to inform personality pathology. Oxford University Press. [Google Scholar]
  16. Hartley AG, & Furr RM (2017). A profile-based framework for factorial similarity and the congruence coefficient. Journal of Personality Assessment, 99, 653–663. [DOI] [PubMed] [Google Scholar]
  17. Hawkins AA, Furr RM, Arnold EM, Mneimne M, & Fleeson W (2014). The structure of borderline personality disorder symptoms: A multi-method, multi-sample examination. Personality Disorders: Theory, Research, and Treatment, 5, 380–389. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Haynes SN, Mumma GH, & Pinson C (2009). Idiographic assessment: Conceptual and psychometric foundations of individualized behavioral assessment. Clinical Psychology Review, 29, 179–191. [DOI] [PubMed] [Google Scholar]
  19. Hepp J, Carpenter RW, Lane SP, & Trull TJ (2016). Momentary symptoms of borderline personality disorder as a product of trait personality and social context. Personality Disorders: Theory, Research, and Treatment, 7(4), 384–393. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Hopwood CJ (2018). Interpersonal dynamics in personality and personality disorders. European Journal of Personality, 32, 499–524. [Google Scholar]
  21. Hopwood CJ, & Krueger RF (2018). Introduction to the Special Section on Theories of Borderline Personality Disorder. Journal of Personality Disorders, 32(2), 145–147. a [DOI] [PubMed] [Google Scholar]
  22. Kotov R, Krueger RF, Watson D, Achenbach TM, Althoff RR, Bagby RM,…Zimmerman M (2017). The Hierarchical Taxonomy of Psychopathology (HiTOP): A dimensional alternative to traditional nosologies. Journal of Abnormal Psychology, 126, 454–477. [DOI] [PubMed] [Google Scholar]
  23. Martin JGA, Nussey DH, Wilson AJ, & Réale D (2011). Measuring individual differences in reaction norms in field and experimental studies: A power analysis of random regression models. Methods in Ecology and Evolution, 2(4), 362–374. [Google Scholar]
  24. Meade AW, & Bauer DJ (2007). Power and precision in confirmatory factor analytic tests of measurement invariance. Structural Equation Modeling, 14(4), 611–635. [Google Scholar]
  25. Miskewicz K, Fleeson W, Arnold EM, Law MK, Mneimne M, & Furr RM (2015). A contingency-oriented approach to understand borderline personality disorder: Situational triggers and symptoms. Journal of Personality Disorders, 29, 486–502. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Mneimne M, Kutz A, & Yoon KL (2017). Individual differences in the motivational direction of anger. Personality and Individual Differences, 119, 56–59. [Google Scholar]
  27. Molenaar PCM (2004). A manifesto on psychology as idiographic science: Bringing the person back into scientific psychology, this time forever. Measurement, 2, 201–218. [Google Scholar]
  28. Nesselroade JR (2007). Factoring at the individual level: Some matters for the second century of factor analysis. In Cudeck R & MacCallum R (Eds.), 100 years of factor analysis (pp. 249–264). Mahwah, NJ: Erlbaum. [Google Scholar]
  29. Pfhol B, Blum N, & Zimmerman M (1997). Structured interview for DSM-IV personality (SIDP). Washington, D.C.: American Psychiatric Association. [Google Scholar]
  30. Roche MJ, Jacobson NC, & Pincus AL (2016). Using repeated daily assessments to uncover oscillating patterns and temporally-dynamic triggers in structures of psychopathology: Applications to the DSM–5 alternative model of personality disorders. Journal of Abnormal Psychology, 125(8), 1090–1102. [DOI] [PubMed] [Google Scholar]
  31. Rosseel Y (2012). lavaan: An R Package for Structural Equation Modeling. Journal of Statistical Software, 48(2), 1–36. [Google Scholar]
  32. Russell RL, Jones ME, & Miller SA (2007). Core process components in psychotherapy: A synthetic review of P-technique studies. Psychotherapy Research, 17, 271–288. [Google Scholar]
  33. Sanislow CA, Grilo CM, Morey LC, Bender DS, Skodol AE, Gunderson JG, … McGlashan TH (2002). Confirmatory factor analysis of DSM-IV criteria for borderline personality disorder: Findings from the Collaborative Longitudinal Personality Disorders Study. The American Journal of Psychiatry, 159(2), 284–290. [DOI] [PubMed] [Google Scholar]
  34. Trull TJ (2018). Ambulatory assessment of borderline personality disorder. Psychopathology, 51(2), 137–140. [DOI] [PubMed] [Google Scholar]
  35. Woods WC, Arizmendi C, Gates KM, Stepp SD, Pilkonis PA, & Wright AGC (in press). Personalized models of psychopathology as contextualized dynamic processes: An example from individuals with borderline personality disorder. Journal of Consulting and Clinical Psychology. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Wright AGC, Beltz AM, Gates KM, Molenaar PCM, & Simms LJ (2015). Examining the dynamic structure of daily internalizing and externalizing behavior at multiple levels of analysis. Frontiers in Psychology, 6. (2016-22150-001). [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Wright AGC, Gates KM, Arizmendi C, Lane ST, Woods WC, & Edershile EA (2019). Focusing personality assessment on the person: Modeling general, shared, and person specific processes in personality and psychopathology. Psychological Assessment, 4, 502–515. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Wright AGC, Hallquist MN, Stepp SD, Scott LN, Beeney JE, Lazarus SA, & Pilkonis PA (2016). Modeling heterogeneity in momentary interpersonal and affective dynamic processes in borderline personality disorder. Assessment, 4, 484–495. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Wright AGC & Woods WC (in press). Personalized models of psychopathology. Annual Review of Clinical Psychology. [DOI] [PubMed] [Google Scholar]
  40. Zanarini MC, Vujanovic AA, Parachini EA, Boulanger JL, Frankenburg FR, & Hennen J (2003). A screening measure for BPD: The McLean Screening Instrument for borderline personality disorder (MSIBPD). Journal of Personality Disorders, 17, 568–573. [DOI] [PubMed] [Google Scholar]

RESOURCES