Construct Validation of Narrative Coherence: Exploring Links with Personality Functioning and Psychopathology

Abstract

Narrative coherence serves as an index of the unity in an individual’s sense of self—integrating their past self with their present self and allowing them to pursue meaningful goals for their future. It has been assessed using the Life Story Interview. Personality functioning is used to describe an individual’s ability to develop stable and integrated representations of the self and others, as well as their ability to develop and maintain stable, intimate, and affiliative relationships, and meaningfully empathize with others. We studied the links between narrative coherence (based on Life Story Interviews) and personality functioning (as indexed by clinician ratings using the Levels of Personality Functioning Scale) in a psychiatric sample (N = 134), and more generally studied the nomological net surrounding narrative coherence. Contrary to predictions, results revealed that narrative coherence does not serve as a marker of personality functioning. However, we found evidence of an association between narrative coherence and measures of extraversion and psychosocial functioning. This study represents an important step in integrating narrative identity with empirically-derived structural models of personality pathology and psychopathology. Implications for future research are discussed.

The narrative that we construct about our lives—i.e., the way we see ourselves currently, how we’ve reached this point, and where we see ourselves heading in the future—forms a core aspect of our personality known as narrative identity (McAdams & Pals, 2006). A robust narrative identity hinges on one’s ability to build an autobiographical sense of self from the episodic details of his or her life. This occurs through a process of meaning-making that fosters the development of a sense of unity and purpose – two key components of narrative identity. Coherence is a structural element of life story narratives that indexes whether a narrative flows in a consistent, logical manner, communicates the significance of events adequately, and meaningfully integrates these events into the individual’s sense of self (Adler et al., 2007; Baerger & McAdams, 1999). Specifically, unity in individuals’ identities is represented by the level of coherence in their personal narratives and overall sense of self (Adler, 2012; McAdams & Pals, 2006).

Interestingly, similar themes have been identified in the study of personality functioning, which is defined by one’s ability to develop stable and integrated representations of the self and others and to maintain stable, intimate interpersonal relationships (Livesley & Jang, 2000). Central to these key life tasks is the ability to develop and use integrated, coherent personal narratives that provide one with a cohesive autobiographical sense of self, serving to establish one’s identity and shaping how one interacts with others (Livesley et al., 2016). Deficits in personality functioning and these related abilities can manifest as identity diffusion, i.e., “sudden and dramatic changes in self-concept, goals, personal values, career plans…” (p. 389, Lind et al., 2019) . Recent developments in personality disorder (PD) science have led to a new conceptualization of PD, instantiated by the Alternative Model of Personality Disorders (AMPD) in DSM-5, in which such deficits are understood to be the core of personality psychopathology (APA, 2013).

The links between narrative coherence and personality functioning are theoretically interesting but have yet to be studied thoroughly in the literature. To date, only two studies have explicitly investigated links between narrative coherence and aspects of personality pathology, revealing a mixed pattern of results with respect to links with Borderline Personality Disorder (BPD) pathology (Adler et al., 2012; Lind et al., 2019). Thus, more work is needed to examine the proposed theoretical links between narrative coherence and personality functioning. The primary goal of this study was to explicate the link between these two constructs in a clinical psychiatric sample, in order to develop a novel approach to understanding and predicting deficits in this core feature of personality pathology. Secondarily, the current study aimed to further explore the nomological network of narrative coherence through its relations with maladaptive and normal-range personality variables.

Narrative Coherence and Personality

Narrative coherence is assessed based on responses to the Life Story Interview (LSI; McAdams, 2008), a semi-structured interview designed to inquire about specific, formative events in the lives of individuals. Individuals are asked to think about their life as a story or book comprised of various chapters, and then are asked to discuss the main chapters in their life. Narratives can then be coded for various underlying themes (e.g., coherence, agency, communion). This study focuses on coherence and conceptualizes this construct as an aspect of personal narratives that indexes unity of the self across time and situations. The original coding system developed by Baerger and McAdams (1999) is best suited for this purpose and is used in this study. This system assesses life story coherence across four dimensions: orientation, structure, affect, and integration (see Table 1; Baerger & McAdams, 1999). As noted by Baerger and McAdams (pg. 74; 1999), “the fragments of a life do not spontaneously cohere in an integrated, harmonious manner… it requires a creative and sustained effort on the part of the life story narrator to synthesize the pieces of a life into a story.” Therefore, coherence not only captures structural aspects of a narrative but also whether the individual is making meaning from the story and incorporating it into how they see themselves. This conceptualization of coherence aligns with specific elements of personality functioning as found in the AMPD. Specifically, identity as the “experience of oneself as unique, with clear boundaries…” and “stability of self-esteem and accuracy of self-appraisal” and self-direction as “pursuit of coherent and meaningful” goals, and the “ability to self-reflect productively.” Whereas some authors have focused on the process of building coherence as a developmental task that facilitates psychological well-being and adaptive functioning (Vanden Poel & Hermans, 2019; Waters & Fivush, 2015), we are interested in coherence of the life story as a marker or indicator of personality functioning. In other words, to what extent does level of coherence predict level of personality functioning and the subdomains of personality functioning?

Table 1.

Narrative Coherence Coding Guidelines

Index	Definition	Criteria	Rating
Orientation	Narrative has temporal, social, and personal background and gives the listener a frame for understanding the context behind the story being told	no orientation	0
		one orienting detail	1
		more than one orienting detail, but reader feels less than fully psychologically oriented	2
		reader is fully psychologically oriented	3
Structure	Narrative includes key structural components - initiating event, internal response, an attempt, and a consequence – that facilitate the flow of the story	the narrator does not recount something that could be considered a story	0
		there is evidence of a topic, but the flow of the story is impossible to discern	1
		there is evidence of a topic and some clear temporal or causal connections between episodes, but also important gaps in these connections	2
		story has a clearly identifiable beginning, middle, and end, with clearly interpretable temporal and/or causal connections between the episodes	3
Affect	Narrative includes emotions that infuse the story with meaning and significance, emotions typically convey an evaluative stance	there is no affective language used in the narrative	0
		there is affective language present in the interview, but not directly connected to the story being told	1
		there is at least some affective language directly connected to the story being told	2
		the use of affective language clearly underscores the importance of the story	3
Integration	Narrator has incorporated narrative into how they see themselves, i.e., their sense of self, and we can see how it contributes to their identity	the narrative does not include any language connecting the story to the self	0
		the narrative includes any self-event connection only following direct prompt	1
		the narrative includes any self-event connection without direct prompting, without substantial elaboration	2
		the narrative includes elaborated connections between past experiences and the self, evidence of clear autobiographical reasoning	3

Note. Adapted from Adler et al., 2012

Two separate investigations in samples with BPD features have found somewhat conflicting evidence with respect to the purported associations between narrative coherence and BPD pathology. It is hypothesized that the identify problems characteristic of BPD, such as instability in one’s sense of self, chronic feelings of emptiness, and affective dysregulation, may lead to pronounced impairments in coherence. In an examination of the link between aspects of narrative identity and personality pathology, Adler and colleagues (2012) found that individuals with BPD (compared to those with no BPD features) exhibited lower levels agency, narrative coherence, and communion fulfillment (i.e., the extent to which the need for communion/connection is fulfilled in the individual) in their narratives. Indeed, these differences were attributed to the prominent role of identity disturbance in BPD. However, in a sample of adolescents with BPD features, Lind and colleagues (2019) found that increased levels of identity diffusion were not significantly associated with lower levels of narrative coherence. Additionally, multiple regression analysis revealed that identity diffusion and not narrative coherence was significantly associated with BPD features when both were included in the model. This discrepancy coupled with the limited literature in this field signals the need for more empirical investigations to fully explicate the nomological net of narrative coherence.

Apart from the work cited above, there have been no other investigations to date examining the incremental associations between narrative coherence and personality problems, specifically. To this end, the current study aims to address this gap with a focus on the relations between narrative coherence and variables indexing both maladaptive and normal-range personality, as well as measures of psychosocial impairment. To the extent that narrative coherence reflects unity in one’s narrative identity, we would expect higher levels of coherence to be associated with less impairment in one’s identity and by extension, personality functioning.

Personality Functioning

As noted earlier, a separate but related literature on personality functioning defines this construct based on stable and well-integrated representations of the self and others, which are supported by a strong sense of personal chronicity. Personality functioning is captured by Criterion A of the AMPD and features disturbances in self and interpersonal functioning. These are further subdivided into four sub-domains: identity and self-direction, and intimacy and empathy, respectively. Personality functioning parallels components from established theories of PD that emphasize dysfunction in interpersonal behaviors (Benjamin, 2005; Pincus, 2005) and split-off internalized representations of the self and others (Kernberg & Caligor, 2005). Livesley and Jang (2000) refer to personality dysfunction as an adaptive failure in three fundamental life tasks: (1) failure to establish stable and integrated representations of self and others, (2) maladaptive interpersonal functioning, and (3) maladaptive societal functioning marked by deficits in prosocial behavior and lack of cooperative relationships. Mastery of these fundamental life tasks hinges on one’s ability to develop a cohesive autobiographical sense of self through the construction of a coherent, integrated narrative of one’s life (Livesley et al., 2016). Therefore, a lack of a coherent autobiographical narrative may manifest as incoherence in one’s self and interpersonal functioning.

Therefore, coherence – “the fundamental story criterion” – is argued to be necessary for healthy personality functioning (pg. 30, Adler et al., 2018) and to serve as an index of the level of impairment in personality functioning. If such a link between narrative coherence and personality functioning can be definitively shown, it would be beneficial in the clinical context, as coherence can be informally assessed based on the initial unstructured clinical interview completed by most mental health practitioners. By paying attention to the overall coherence of their patient’s story, clinicians may gain a sense of the presence and severity of personality pathology. Emerging work supports the validity of personality functioning as a “severity” indicator of PDs (Morey et al., 2013) and demonstrates that Criterion A ratings are associated with both the presence and number of diagnosed PDs, as well as PD criterion counts (Bastiaansen et al., 2013). That said, Criterion A remains under-investigated relative to Criterion B (maladaptive personality traits) of the AMPD, with 84.8% of all AMPD publications being specifically focused on Criterion B (Zimmermann et al., 2019) with open questions about the nature of personality functioning (Meehan et al., 2019; Morey, 2019; Sleep et al., 2020).

Current Study

The primary aim of the current study is to evaluate narrative coherence as a marker of personality functioning, both at the zero-order level and controlling for important demographic variables that could be associated with coherence, such as sex, education, income, and age (Adler et al., 2007; Baerger & McAdams, 1999). We hypothesize that narrative coherence will predict impairments in personality functioning, above other demographic variables. Given the theory underpinning narrative coherence (Adler et al., 2016), we hypothesize that higher levels of coherence will be associated with less impairment in personality functioning, particularly with respect to ratings of identity with small effect sizes predicted (rs = .10 - .30). We predict that the magnitude of the effect sizes will be reduced due to the divergent methods used to rate narrative coherence (i.e., coding of recorded life story interviews) and level of personality functioning (i.e., clinician ratings based on multiple interviews and self-report questionnaires).

A secondary aim is to extend the limited extant literature and explore the nomological network of narrative coherence with respect to personality pathology, psychopathology and psychosocial functioning. We will do this by studying the convergent and discriminant relations between coherence and various personality disorder variables, including maladaptive and normal-range personality traits as measured by the PID-5 (Krueger et al., 2012), BFI (John & Srivastava, 1999), PD criterion counts based on the official diagnostic system of the DSM-IV/DSM-5 (SCID-II; First, 1997), and psychopathology as indexed by the IDAS-II (Watson et al., 2012). We hypothesize that coherence will correlate positively with conscientiousness, openness, and extraversion based on the limited past work (Adler et al., 2007; Lodi-Smith et al., 2009). Furthermore, based on the current conceptual understanding of the maladaptive traits and their relation to aspects of personality, we hypothesize that coherence will be negatively associated with levels of negative affectivity which has been shown to overlap with self-pathology (Clark & Ro, 2014). We also predict that aspects of disinhibition and psychoticism will be negatively associated with coherence. Disinhibition is considered to be the pathological counterpart to conscientiousness (Wright & Simms, 2014), whereas psychoticism, which has been shown to be associated with thought disorder, would not facilitate a clear and well-organized story (Kotov et al., 2017). We predict small effect sizes for these associations (rs = .10 - .20). However, given that the associations between coherence and maladaptive personality traits have not yet been explored, these analyses are largely exploratory.

Additional exploratory analyses also will be presented to the examine the convergent and discriminant relations between coherence and interpersonal problems as measured by the Inventory of Interpersonal Problems (IIP; Alden et al., 1990; Horowitz et al., 1988), and psychosocial impairment as indexed through the World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0; Üstün, 2010) and the Range of Impaired Functioning Tool (LIFE-RIFT; Leon et al., 1999). These measures, which tap aspects of physical and psychological health and family relationships, serve to amplify and add breadth to our index of psychological well-being – facilitating the exploration of its links to narrative coherence (Adler et al., 2016; Reese et al., 2011).

Our hypotheses were not pre-registered through any open access forums, and all exploratory work has been identified as such.

Methods

Participants

The current analyses are based on a sub-sample of 134 participants drawn from a validation study of the Comprehensive Assessment of Traits Relevant to Personality Disorder (CAT-PD; Simms et al., 2011) for whom narratives were collected using an abbreviated version of the LSI. The protocol was approved by the University at Buffalo, SUNY, Institutional Review Board (IRB) in accordance with established ethical guidelines. Participants were recruited from mental health clinics in the Western New York area. All participants provided written informed consent and were eligible to participate if they (a) were 18 years of age or older, (b) were in psychiatric treatment currently or within the past two years, and (c) demonstrated sufficient comprehension of English to complete the measures. The mean age of the sample was 44.4 years (SD = 12.2). The sample was primarily female (67.2%; n = 90) and Caucasian (72.4%; n = 97), with smaller proportions of Black/African Americans (23.9%; n = 32) and Asians (1.5%, n = 2). The modal level of education completed was some college (ranging from minimum 9^th grade to graduate or professional degree attained). Based on the Structured Clinical Interview for DSM-IV Axis II PDs, 35% of participants met criteria for Antisocial PD (n = 47), 29% met criteria for Borderline PD (n = 39), 27% met criteria for Paranoid PD (n = 36), 20% met criteria for Avoidant PD (n = 27), 15% met criteria for Obsessive-Compulsive PD (n = 20), 11% met criteria for Schizotypal PD (n = 15), and less than 5% of the sample met criteria for Histrionic, Narcissistic, Dependent, and /or Schizoid PD. Overall, 33% of participants did not meet criteria for any PD (n = 44), 25% met criteria for only one PD (n = 34), and 42% met criteria for two or more comorbid PDs (n = 56).

Procedures

Participants completed all questionnaires in a laboratory on computers with shielding for privacy. Participants first completed several screening measures followed by the full CAT-PD assessment. This was followed by the administration of a structured diagnostic interview and an abbreviated LSI, conducted by trained and supervised clinical psychology doctoral students. (see following section for further details). Each LSI was about 30 minutes in length. Finally, participants were asked to complete additional self-report questionnaires. The measures relevant to the current study are described below.

Measures

Life Story Interview (LSI; McAdams 2008).

The LSI is a semi-structured interview designed to inquire about specific, formative events in the lives of individuals. Participants were administered an abbreviated version of the interview, asking them to think about their life as a story or book comprised of various chapters and to discuss the main chapters in their life. Subsequently, participants were asked to talk about several key scenes in their life, including a high point, low point, turning point, positive childhood memory, negative childhood memory, and “The Next Chapter.” As per guidelines for administration of the LSI, participants were not redirected or asked probing questions that could influence their responses (Adler et al., 2017). Interviews were video-recorded with the consent of the participant.

The videotaped life story interviews and accompanying transcribed narratives were reviewed and rated for narrative coherence by two sets of raters. The coders met with the study lead (J.D.) for four separate training sessions. They were introduced to the construct of narrative coherence and instructed to rate sample narratives from the literature (Adler, 2012; Baerger & McAdams, 1999). Ratings were subsequently reviewed with the study leader and discrepancies were discussed with the group until consensus was reached. The coherence coding system implemented was a modification of the original guidelines proposed by Baerger and McAdams (1999; Table 1) developed by Adler and colleagues (Adler, 2012).

Eight narratives (5%) from the study sample were coded during the training process. Inter-rater reliability (IRR) was assessed using intra-class correlation coefficients (ICCs) through a one-way random effects model (Case 1; Shrout & Fleiss, 1979). The coders achieved excellent inter-rater reliability on average (ICC = .90) before proceeding to code the full dataset. Coding of the full dataset was completed in two stages across two semesters. During the first stage of coding, five coders each were assigned a separate portion of the dataset to code until each narrative in the dataset was coded once (i.e., the data was single-coded). The second set of nine coders were not identical to the first set, as there was turnover of lab personnel from the first to the second semester. These coders each were assigned to code a separate portion of the already-coded dataset until each narrative had been coded twice. Thus, a total of fourteen coders, blind to the study hypotheses, coded the dataset. This means that each narrative was coded by a different subset of two coders. For this reason, we used Case 1 and conducted a one-way random effect model to determine IRR. Good IRR (Cicchetti, 1994) was achieved for the entire dataset (ICC = 0.66). Given that each of the four dimensions were intercorrelated (rs = .22 - .81), an overall score of coherence was computed as the average of the four dimensions.

Personality Inventory for DSM-5 (PID-5; Krueger et al., 2012).

The PID-5 is a 220 item self-report questionnaire that consists of 25 personality trait scales. All items are presented with response options that range from “very false or often false” to “very true or often true” (Krueger et al., 2012). Based on exploratory factor analyses conducted, five factors have been identified that underlie responses and have been labelled as follows: negative affect, detachment, antagonism, disinhibition, and psychoticism (Krueger et al., 2012). The median alpha coefficient was .87 (range = .73 - .96) for each of the facets in the current study. The construct validity of the PID-5 is well-documented in past work (Anderson et al., 2013; Hopwood et al., 2013; Wright & Simms, 2014).

Structured Clinical Interview for DSM-IV Axis II PDs (SCID-II; First, 1997).

The SCID-II is a semi-structured interview that assesses the 10 DSM-IV PDs. Each PD criterion is scored using a scale with the following options: 0 (absent), 1 (present but subthreshold), or 2 (present) rating. Clinicians rate the presence of each symptom and determine whether sufficient criteria are met for a specific PD diagnosis. For the current study, PD criterion counts for each diagnosis were used as opposed to categorical determinations indicating the presence or absence of a disorder. The median single-measures ICC for PD criterion ratings was .62 for avoidant PD, .62 for depressive PD, .47 for obsessive-compulsive PD, .62 for paranoid PD, .51 for schizotypal PD, .44 for schizoid PD, .58 for histrionic PD, .49 for narcissistic PD, .54 for borderline PD, and .84 for antisocial PD.

Levels of Personality Functioning Scale (LPFS; Morey et al., 2011).

The LPFS is used to characterize the severity of personality impairment across four dimensions based on clinician ratings: identity (M = 1.99, SD = 1.09), self-direction (M = 1.75, SD = 1.13), empathy (M = 1.50, SD = 1.13), and intimacy (M = 2.16, SD = 1.22).The LPFS criteria used were excerpted directly from the AMPD in Section III of the DSM-5 (American Psychiatric Association, 2013). Each dimension is rated on a scale of 0 (healthy functioning) to 4 (extreme impairment), with a rating of 2 or higher (moderate impairment) indicating clinically significant impairment in personality functioning. LPFS ratings were conducted by clinical psychology doctoral students and a project coordinator with a Master’s degree, under the supervision of a licensed clinical psychologist with considerable experience in the characterization and diagnosis of personality disorder (LJS). Participants were assessed with the Life Story Interview, a trait interview based on the self-report CAT-PD, the SCID-II, and an interview designed to follow-up on participants responses to the Range of Impaired Functioning Tool (LIFE-RIFT) – a measure of psychosocial impairment. The LPFS was then rated by the same clinical rater who conducted the interview and represented a holistic judgment based on all of the knowledge collected. Clinical raters were trained to rate levels of personality functioning using the AMPD Criterion A anchors/rating scheme found in the DSM-5. Training took place over multiple sessions and was conducted by the second author of the paper (LJS). A 25% subset of the data was selected for double-coding to facilitate calculations of inter-rater reliability using the intra-class correlation coefficient (ICC). The single-measures ICC for the overall rating of personality dysfunction was 0.53.

Big Five Inventory (BFI; John & Srivastava, 1999).

The BFI is designed to measure normal-range personality traits as per the five-factor model (i.e., extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience; John & Srivastava, 1999). The measure consists of 44 items where participants rate the extent to which they believe the item describes them on a 5-point scale ranging from “disagree strongly” to “agree strongly” (John et al., 1991). The median alpha coefficient was .80 (range = .75 - .82) for each of the domains in the current study. Past work has also documented the construct validity of this measure (Rammstedt & John, 2007).

Inventory of Depression and Anxiety Symptoms-II (IDAS-II; Watson et al., 2012).

The IDAS-II is an expanded revision of the Inventory of Depression and Anxiety Symptoms (IDAS; Watson et al., 2007). It is self-report measure that consists of 99 items grouped into nonoverlapping scales of depression and anxiety symptoms. Participants are asked indicate the extent to which they have experienced each item in the past two weeks on a 5-point scale ranging from “not at all” to “extremely.” The median alpha coefficient was .84 (range = .74 - .89) for each of the scales in the current study. This measure also shows strong convergent and discriminant validity when compared to other measures of anxiety, depression, and mania (Watson et al., 2007).

Inventory of Interpersonal Problems – Short Circumplex (IIP-SC; Alden et al., 1990; Horowitz et al., 1988).

The IIP-SC is a 32-item self-report inventory of interpersonal style and behaviors that either are “hard for you to do” or behaviors that “you do too much” using a 5-point Likert scale ranging from “not at all” (0) to “extremely” (4). The median alpha coefficient was .81 (range = .67 - .89) for the scales in the current study. The internal structure of this measure aligns with the eight domains of the interpersonal circumplex, and it demonstrates adequate convergent validity with other measures of interpersonal style and attachment, although evidence of discriminant validity is lacking (Bailey et al., 2018; Hopwood et al., 2008).

World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0; Üstün, 2010).

The WHODAS 2.0 is a 36-item self-report assessment of health and disability the covers six domains of functioning, including: cognition, mobility, self-care, getting along, life activities, and participation. Scores were computed for each of the six domains, as well as, a single global score, with higher scores indicating greater disability. The median alpha coefficient was .90 (range = .80 - .96) for the domains in the current study.

Range of Impaired Functioning Tool (LIFE-RIFT; Leon et al., 1999).

The LIFE-RIFT is another self-report measure that assesses functioning across four domains: work, interpersonal relationships, satisfaction and recreation. Participants are asked to rate their level of impairment on a scale of 0 to 6 (for work, satisfaction, and recreation) or 0 to 8 (for interpersonal relationships) – with 0 representing “not applicable” and the highest score 6/8 representing “no information”. Otherwise, lower scores (1- 5/7) indicate none to low impairment and higher scores indicate greater severity of impairment.

Power Analysis

Power analyses were conducted to identify the sample size needed to detect an increment in variance of 5% at an α level of .05. Assuming that 5% of the variance would be accounted for by the demographic variables entered at Step 1 of the regression, if narrative coherence accounts for an additional 5% of the variance in personality functioning, then a sample of 144 participants will provide 80% of the power needed to detect this effect. The sample size of the present study closely approximates these values.

Results

All analyses were conducted in R (R Core Team, 2017) using the following packages: irr (v0.84.1; Gamer et al., 2019), psych (v1.9.12.31; Revelle, 2019), car (v3.0-2; Fox & Weisberg, 2019), lme4 (v1.1-20; Bates et al., 2015), ez (v4.4.-0; Lawrence, 2011), stats (v3.4.3; R Core Team, 2017), and Hmisc (v4.2-0; Harrell Jr, 2020). All data and the accompanying R scripts are available through Open Science Framework.

Descriptive statistics

Table 2 presents the descriptive statistics and intercorrelations among the narrative coherence variables and correlations between narrative coherence and demographic variables that may be associated with our outcome of interest (e.g., age, education, gender, and income; Adler et al., 2007). All intercorrelations were significant and ranged from r = .22 to .81. Although not presented here, intercorrelations among the criterion personality and psychopathology measures were generally significant and of sufficient magnitude. With respect to the demographic variables, narrative coherence, orientation, and structure correlated weakly to moderately with income (rs = .19 - .25). A one-way analysis of variance revealed a significant difference across race subgroups on narrative affect (F(2, 128) = 3.00, p = .05). Tukey post-hoc t-tests indicated that African Americans (M = 1.18, SD = 0.05) used less affect in their narratives than Caucasians (M = 1.40, SD = 0.46; d = .49).

Table 2.

Narrative Coherence descriptive statistics, intercorrelations and Pearson correlation coefficients between Narrative Coherence and demographic variables

Variable	Narrative Coherence
	Narrative Coherencea	Orientation	Structure	Affect	Integration
Mean (SD)	1.44 (0.32)	1.76 (0.36)	1.53 (0.39)	1.34 (0.47)	1.11 (0.51)
Minimum	0.7	0.5	0.6	0.2	0
Maximum	2.15	2.5	2.4	2.4	2.5
Skewness	−0.11	−0.32	0.08	−0.16	0.27
Narrative Coherence Coding
Narrative Coherence
Orientation	.76**
Structure	.76**	.64**
Affect	.67**	.31**	.22**
Integration	.81**	.44**	.52**	.38**
Demographic Variablesb
Gender	.04	−.03	.07	.07	0
Age	.01	.02	−.02	.09	−.06
Education	.15	.16	.16	.14	.02
Income	.23**	.19*	.25**	.14	.13
LPFS
Overall	−.08	−.04	−.07	−.06	−.08
Identity	.09	.11	.09	.08	.01
Self-direction	−.16	−.11	−.18*	−.08	−.12
Empathy	−.20*	−.10	−.15	−.17	−.16
Intimacy	−.02	.01	−.02	−.001	−.05

Note. LPFS = Level of Personality Functioning Scale.

^aNarrative coherence was computed as the average of the four components: Orientation, Structure, Affect and Integration.

^bDue to missing values in the demographic variables and LPFS, the number of participants is smaller than the total sample size and varies from 129 - 134.

^*p < .05,

^**p < .01

Predicting personality functioning

A series of hierarchical linear regressions were conducted to test our hypothesis that narrative coherence incrementally predicts personality functioning over age, gender, education, and income (Adler et al., 2007; Baerger & McAdams, 1999). The models were constructed by first entering the demographic variables in the first step predicting each of the personality disorder variables of interest: personality functioning (LPFS) overall score and scores on the specific elements of personality functioning (identity, self-direction, empathy and intimacy). Significant variables from this step were retained, and then narrative coherence was added in the next step. Additional exploratory regressions were conducted with the specific dimensions of narrative coherence (orientation, structure, affect, integration) entered individually in the final step. Regression results (see Table 3) indicated that income significantly predicted impairment in the overall personality functioning score, self-direction, empathy, and intimacy. The stepwise addition of narrative coherence did not incrementally predict personality functioning or its sub-domains as hypothesized. Similar null results were found across additional models conducted in which dimensions of narrative coherence (i.e., orientation, structure, affect and integration) were added in a stepwise fashion.

Table 3.

Hierarchical Regression Analyses of Narrative Coherence and its subcomponents relative to demographic variables in the prediction of Personality Functioning and subdomains^a

	Dependent Variables
Step 1: Demographic Variables	Personality Functioning	Identity	Self-Direction	Empathy	Intimacy
Age	−.05	−.20*	−.12	.04	.02
Gender	.03	.09	.06	−.02	.03
Education	−.08	.05	.05	−.16	−.18
Income	−.28**	−.16	−.29**	−.19*	−.20*
R2	.07	.02	.07	.07	.06
F2	3.46**	1.79	3.40**	3.31**	2.95*
Step 2: Narrative Coherence	−.03	.10	−.08	−.11	.03
R2	.07	.03	.07	.07	.05
ΔR2	−.01	.01	.00	.00	−.01
ΔF	−0.69	−0.11	−1.1	−0.35	−0.58

Note.

^aAll standardized regression coefficients are from the final step in the analyses. Due to missing values in the demographic and personality functioning variables, the number of participants in this analysis is smaller than the total sample size and varies from 126 – 132.

^*p < .05,

^**p < .01

Exploring the nomological net of narrative coherence

Zero-order correlations were computed to explore the nomological network of narrative coherence and its dimensions through its relations with personality, personality pathology, and psychopathology variables (see Table 4). Correlations were considered significant at α = .01 to account for Type I error. Most of these correlations were relatively small and non-significant. However, several significant and trend-level patterns emerged that deserve attention. First, narrative coherence as a composite yielded a small positive association with BFI extraversion and small negative association with unusual beliefs and experiences from the PID-5. Second, beyond the composite score, the dimensions of coherence showed a more nuanced pattern of relations with personality and psychopathology. Narrative orientation showed a small negative association with grandiosity significant at α = .01. Narrative structure yielded a small positive association with the Avoidant PD criterion count, and small negative associations with the Schizotypal PD and Antisocial PD criterion counts. With respect to aspects of psychopathology, narrative structure showed a small negative association with IDAS euphoria, as well as a small negative association with IDAS claustrophobia. Narrative affect also showed a small positive association with the Avoidant PD criterion count and a small negative association with PID-5 unusual beliefs and experiences. Narrative integration showed a small significant positive association with extraversion from BFI, a small negative association with the Schizoid PD criterion count, and a small negative association with IDAS well-being significant at α = .01. Overall, there was a pattern of significant positive associations between integration with extraversion and well-being, punctuated by trend-level negative associations with personality pathology characterized by social withdrawal, thought disorder, and unusual beliefs.

Table 4.

Pearson correlation coefficients between Narrative Coherence and Measures of Personality and Personality Pathology

Variable	Narrative Coherence
	Narrative Coherence	Orientation	Structure	Affect	Integration
SCID-II Criterion Countsa
Avoidant PD	.17	.07	.21*	.20*	.03
Dependent PD	.06	−.08	−.03	.16	.10
Obsessive Compulsive PD	.03	−.03	−.05	.16	−.02
Paranoid PD	−.10	−.11	−.14	−.02	−.05
Schizotypal PD	−.14	−.11	−.19*	−.06	−.08
Schizoid PD	−.15	−.04	−.12	−.01	−.26**
Histrionic PD	−.05	−.02	−.07	−.11	.04
Narcissistic PD	.03	.07	−.02	.08	−.04
Borderline PD	.04	.03	.01	.04	.03
Antisocial PD	−.09	−.04	−.18*	−.05	−.03
Big Five Inventoryb
Neuroticism	.07	−.02	.11	.06	.06
Extraversion	.20*	.13	−.05	.15	.30**
Conscientiousness	−.03	−.06	−.12	.08	−.03
Agreeableness	−.10	−.08	−.05	−.12	−.06
Openness to Experience	.03	.07	.00	−.07	.11
PID-5 Domainsc
Negative Affectivity	.02	−.08	−.02	.04	.08
Detachment	−.09	−.13	0	.00	−.13
Antagonism	−.01	−.13	−.02	−.01	.09
Disinhibition	.01	.03	.04	−.05	.02
Psychoticism	−.11	−.05	−.06	−.16	−.05
IDASd
General Depression	.05	.03	.14	.03	−.03
Dysphoria	.07	.04	.10	.09	−.02
Lassitude	.06	.03	.11	.08	−.01
Insomnia	.04	.04	.07	.03	−.01
Suicidality	−.08	−.01	−.02	−.11	−.08
Well-Being	.12	.06	−.10	.07	.27**
Ill-Temper	.06	.01	.07	−.02	.10
Mania	−.04	−.09	−.13	.02	.03
Euphoria	−.09	−.11	−.23*	−.03	.06
Panic	−.06	−.02	−.04	−.07	−.04
Social Anxiety	−.02	−.07	−.01	.08	−.05
Claustrophobia	−.15	−.12	−.09	−.20*	−.06
Trauma Intrusions	.00	.01	−.02	−.02	.04
Trauma Avoidance	−.16	−.13	−.09	−.12	−.14

Note. SCID-II = Structured Clinical Interview for DSM-IV Axis II PDs; PID = Personality Inventory for the DSM-5; IDAS = Inventory of Depression and Anxiety Symptoms.

Due to missing values, the sample size is lower than the total sample size for the measures above and varies from:

^aN = 128-133,

^bN = 106,

^cN = 119 – 120,

^dN = 113 – 116.

^*p < .05,

^**p < .01

Predicting psychosocial impairment/functioning

Zero-order correlations also were conducted between narrative coherence, its dimensions, and variables indexing psychosocial impairment and functioning (i.e., LIFE-RIFT and WHODAS), as well as a measure of interpersonal behaviors and functioning (i.e., IIP). See Table 5 for a complete summary of these correlations. Results revealed significant moderate associations between overall narrative coherence and the employment scale of the LIFE-RIFT, indicating that greater impairment in the employment domain was associated with decreases in narrative coherence (Table 5). For participants with children, decreases in narrative coherence were associated with greater impairment in their relationships with their children. Results also revealed small negative associations between the self-care scale of the WHODAS and narrative coherence, such that greater impairments in self-care were associated with decreases in overall narrative coherence (Table 5). Furthermore, there was a moderate negative association between orientation and the LIFE-RIFT employment scale. Narrative structure showed negative associations with the LIFE-RIFT employment scale, impairments in relationships with one’s children, but small positive associations with several domains from the IIP, including: Nonassertive, Exploitable, and Overly Nurturant, as well as indices of Angular Displacement and Elevation, and a small negative association with the R-Squared, Fit Index. Lastly, integration showed a small to moderate negative associations with the LIFE-RIFT employment scale and the WHODAS self-care scale. Overall, there was a pattern of negative associations between narrative coherence and its dimensions with impairments in employment, interpersonal functioning and the ability to care for oneself.

Table 5.

Pearson correlation coefficients between Narrative Coherence and the scales of the Range of Impaired Functioning Tool (LIFE-RIFT) and the World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0)

Variable	Narrative Coherence
	Narrative Coherence	Orientation	Structure	Affect	Integration
LIFE-RIFT Employment	−.38**	−.31**	−.33**	−.14	−.36**
LIFE-RIFT Household	−.04	−.11	.04	−.04	.00
LIFE-RIFT Student	.34	.29	.47*	.11	.22
LIFE-RIFT – Spouse	.05	.05	.01	.17	−.08
LIFE-RIFT – Children	−.26*	−.12	−.28**	−.17	−.20
LIFE-RIFT – Other Relatives	−.09	−.10	−.10	−.08	.00
LIFE-RIFT – Friends	−.03	−.02	−.01	−.06	.01
LIFE-RIFT Satisfaction	−.08	−.07	.07	−.10	−.12
LIFE-RIFT Recreation	.02	−.02	.04	.02	.02
WHODAS Cognition	−.03	−.02	.01	−.01	−.06
WHODAS Mobility	−.06	−.01	.01	−.07	−.09
WHODAS Self-Care	−.20*	−.10	−.10	−.18	−.21*
WHODAS Getting Along	−.01	.01	.03	−.01	−.05
WHODAS Housework	.08	.06	.07	.09	.01
WHODAS Work	.10	.07	.17	.06	.03
WHODAS Life Activities	.10	.07	.13	.08	.02
WHODAS Participation in Society	.09	.04	.06	.07	.09
WHODAS Full Scale	.02	.01	.05	.03	−.04

Note. Due to missing values in the LIFE-RIFT and WHODAS variables, the number of participants is smaller than the total sample size and varies from 111 - 132.

^*p < .05,

^**p < .01

Discussion

The aims of the current study were twofold: (1) to evaluate narrative coherence as a marker of personality functioning, both at the zero-order level and controlling for important demographic variables, and (2) to extend our understanding of the nomological network of narrative coherence. Overall, we found very few significant associations that conformed with our a priori hypotheses, and many of the remaining significant associations did not appear to be particularly meaningful. In light of these findings, we discuss implications for our understanding of narrative coherence and its purported ties to personality pathology and psychopathology, as well as methodological considerations that may have affected our outcomes.

Narrative Coherence and Personality Functioning

We found that overall narrative coherence was associated with empathy from the LPFS and narrative structure was associated with self-direction from the LPFS. However, these associations were no longer significant when accounting for key demographic factors, in particular income. These results did not match our expectations. Lower income was found to be associated with greater impairments in personality functioning, and this appears to be a more robust association than the associations with narrative coherence and structure. Indeed, there is a well-documented association between personality problems and deficits in psychosocial functioning, of which income may be a crude index (Skodol, 2018). It may also be the case that, despite the apparent conceptual links between these two constructs, they represent two differing aspects of personality and as such may be used in complement with one another to provide a fuller clinical picture. Whereas clinician-rated personality functioning is a global index of one’s current severity of impairment, narrative coherence involves a skilled re-telling of significant past events in one’s life. Indeed, Lind and colleagues (2019) found that identity diffusion, measured using the Assessment of Identity Development in Adolescence, was incrementally associated with BPD features over and above narrative coherence. Another interesting finding revealed that low levels of identity diffusion were not significantly associated with increased narrative coherence. Therefore, it appears that impairments in self functioning may take multiple forms, and that in addition to using narratives to gather information about personality and psychosocial functioning, other forms of information are still needed to draw more accurate clinical impressions and treatment plans.

As such, it is possible that narrative coherence simply does not map onto personality functioning as we hypothesized. However, it is also possible that there is something about the way narrative coherence is coded (e.g., idiosyncrasies in our dataset or with the coherence coding scheme selected) that obscures any true relationship that might be there. However, we thoroughly reviewed our data and did not find any coding errors. Therefore, we would argue that the former is more likely. Our null results also could be explained by the different methods used in assessing narrative coherence and personality functioning – meaning that we do not see the typical inflation of correlations due to shared method variance¹. Our narrative coherence coding was based only on the transcripts of life story interviews, whereas the LPFS ratings were made based on a host of information about the participants’ clinical characteristics. Given that our findings do not comport with past studies on narrative coherence and personality and the decreases in inter-rater reliability, replication studies are needed in order to gain a greater understanding of the relation between narrative coherence and LPFS. Moreover, it would be important to examine the links between narrative coherence and LPFS rating using alternative methods to those used here. For example, several self-report LPFS measures have been developed recently (e.g., Gamache et al., 2019; Morey, 2017; Weekers et al., 2019) that have the potential to yield different results than those reported in this study. Similarly, different methods can be used to assess and code narrative coherence (Lysaker et al., 2012; Reese et al., 2011). That said, if results are wildly inconsistent across studies and different measurements of these constructs, then additional work will be needed to bring clarity to the measurement of both personality functioning and narrative coherence.

On the Structure of Narrative Coherence

In mapping the nomological network of narrative coherence, we found evidence indicating that the components of narrative coherence (i.e., orientation, structure, affect and integration) do not neatly cohere together given their differing patterns of association with the various psychopathology and personality pathology variables. With respect to normal range personality traits, narrative coherence correlated with extraversion as we hypothesized, although only at the level of a trend. That being said, narrative integration significantly correlated with extraversion, such that higher levels of integration in a narrative were associated with greater extraversion. Integration requires the individual to take a step back from the story being told in order to derive specific meaning from the story, and then also to draw a connection between this theme and how they see themselves. It has been hypothesized that due to their proclivity for social interaction, those higher in extraversion may be well-practiced in recounting their stories and as a result, develop more coherence in their narratives over time (Lodi-Smith et al., 2009).

Further, narrative integration also was negatively associated with Schizoid PD (which is itself negatively associated with extraversion) and positively associated with the well-being scale of the IDAS – a scale largely defined by items reflecting high energy and positive affect (Watson et al., 2007). This finding is consistent with the significant correlation observed between integration and extraversion. Given that extraversion also was significantly correlated with well-being (r = .46), we conducted post-hoc linear regressions but found no evidence of an incremental association between narrative integration and well-being. Our findings support the notion that integration is associated with extraversion and that the relationship between integration and well-being can be explained by extraversion. This helps to address gaps in the literature regarding the incremental validity, or lack thereof, of aspects of narrative coherence (Adler et al., 2016; Adler & Clark, 2019).

Overall, we believe that some of our findings lend support to prior work showing that the four dimensions of narrative coherence did not neatly represent one component (Adler et al., 2018). That said, there was no clearly discernible patterns of association between the most of the dimensions of narrative coherence and other variables, apart from the associations between integration with variables indexing aspects of positive affectivity (e.g., extraversion, well-being) seems to set this dimension apart. Whereas we do not view narrative coherence and its dimensions as simply a “derivative of traits” (Adler & Clark, 2019, p. 2), these findings suggest that narrative coherence does not represent one component, and individual differences in personality may account for some of the variation across the four dimensions of this construct.

Narrative Coherence and Psychosocial Functioning

Lastly, we examined its associations with measures indexing physical health, and social and employment functioning. The zero-order correlations revealed that the employment scale of the LIFE-RIFT measure correlated significantly with overall narrative coherence, orientation, structure, and integration, indicating that greater impairment in the employment domain was associated with reduced narrative coherence and its components. Given that narrative coherence is said to represent unity in one’s identity (McAdams & Pals, 2006), and the associations between integration and positive affect, the association with employment functioning appears reasonable. Indeed, there is a large body of evidence attesting to the robust association between personality problems such as dysregulated affect, instability of self-esteem, and compromised reality testing and psychosocial functioning (see Skodol, 2018 for a review). Lastly, we observed a negative association between narrative structure with impairment in participants’ relationships with their children. Reese and colleagues (2011) have cited evidence supporting similarities between mothers’ narratives and that of their young children. It is perhaps the case that greater narrative structure may facilitate better parent-child interactions through providing greater direction or clarity in the relationship. That said, further exploration and replication of this association is needed.

Limitations and Conclusion

This work represents the first direct attempt in the literature to explore the links among narrative coherence, personality functioning, and others aspects of psychopathology and personality pathology in a well-characterized clinical sample. However, it is not without its limitations. Coherence is fundamental to the intelligibility of a narrative, without it a narrative would simply cease to make sense – becoming a jumble of words. That said, coherence has shown to be one of the most difficult narrative identity constructs to operationalize. Unlike constructs such as agency, communion, and redemption, where the field has converged on the underlying theory and coding system, there are multiple coherence coding systems and research showing that the construct itself manifests differently depending on when and how it is measured (Adler, 2012; Adler et al., 2018; McLean et al., 2020). We also used a very abbreviated version of the LSI in this study, which may have had implications for our ability to fully capture narrative coherence, but was done to minimize participant burden. Prior studies also have found no links between narrative word count (a proxy of general expressiveness) and narrative coherence abilities (Adler, 2012; Adler et al., 2007, 2008). And although we achieved good inter-rater reliability, it does reflect difficulties with capturing this construct (Cicchetti, 1994). Second, this study relied on cross-sectional data and as a result, we can make no inferences about the directionality of the effects observed. Third, our sample is primarily Caucasian and African American, and recruited only from the Western New York region. Taken together, these features may affect the generalizability of our findings.

In summary, our aim was to bridge gaps in the literature between a fundamental component of narrative identity, a construct based in models of basic personality science, and personality functioning, which is understood to represent the core of personality pathology in contemporary, empirical models of PDs, such as the AMPD. Indeed, recent work by Adler and Clark (2019) issues a strong call for more integrative research initiatives that serve to incorporate narrative identity into the developing and increasingly prominent structural models of psychopathology and personality pathology. Specifically, they suggest incorporating interviews designed to elicit life story information into broader studies of psychopathology and personality pathology (which we have done here), examining the associations between variation in narrative identity and levels of psychological well-being at the more general domain-level versus specific disorder-level of psychopathology, and refining aspects narrative identity for use as clinical tools that can shape treatment of personality and psychopathology. As such, we feel that this work represents an important first step towards fulfilling these initiatives. However, given that our correlational analyses did not reveal a rich network of associations between narrative coherence and the psychopathology and personality pathology variables included, we were unable to conduct further analyses of the sort advocated by Adler and Clark (2019). Nevertheless, we maintain that this is not a fruitless endeavor given the potential for enriching our current models of psychopathology and personality pathology. However, we would argue that before additional analyses of this sort are undertaken that more empirical work on the nature of narrative coherence is needed. The current study is in line with past work which shows that, despite the high inter-correlations, the dimensions of this construct do not all bear the same associations with other measures. As such, further exploration and replication of these findings is still needed before clear conclusions can be drawn.