Narrative Identity in the Psychosis Spectrum: A Systematic Review and Developmental Model

Abstract

Individuals with schizophrenia-spectrum disorders face profound challenges as they attempt to maintain identity through the course of illness. Narrative identity—the study of internalized, evolving life stories—provides a rich theoretical and empirical perspective on these challenges. Based on evidence from a systematic review of narrative identity in the psychosis spectrum (30 studies, combined N = 3,859), we argue that the narrative identities of individuals with schizophrenia-spectrum disorders are distinguished by three features: disjointed structure, a focus on suffering, and detached narration. Psychotic disorders typically begin to emerge during adolescence and emerging adulthood, which are formative developmental stages for narrative identity, so it is particularly informative to understand identity disturbances from a developmental perspective. We propose a developmental model in which a focus on suffering emerges in childhood; disjointed structure emerges in middle and late adolescence; and detached narration emerges before or around the time of a first psychotic episode. Further research with imminent risk and early course psychosis populations would be needed to test these predictions. The disrupted life stories of individuals on the psychosis spectrum provide multiple rich avenues for further research to understand narrative self-disturbances.

“What scared me the most was a sense that I had lost myself, a constant feeling that my self no longer belonged to me.…the real ‘me’ is not here anymore.…my thoughts, my emotions, and my actions, none of them belong to me anymore.…in my opinion, schizophrenia is ultimately a disorder of the self” (Kean, 2009, p. 1034).

Since the earliest descriptions by Bleuler and Kraepelin, a disturbed sense of self has been considered a hallmark of the subjective experience of schizophrenia (Bleuler, 1911; Kraepelin, 1919; Lysaker & Lysaker, 2010). This self-disturbance is profound, reaching from basic senses of agency and ownership over experience (de Vries et al., 2013; Henriksen & Parnas, 2012; Hur et al., 2014; B. Nelson, Parnas, et al., 2014; Sass & Parnas, 2003) to lifelong struggles to build personal identity and find meaning amid lost roles, relationships, and internalized stigma (Andresen et al., 2003; Ben-David & Kealy, 2019; Conneely et al., 2020; Roe & Davidson, 2005). “My illness eradicated my sense of self,” one individual with schizophrenia described, “and now I am engaged in the lifelong process of obtaining, maintaining, and slowly modifying my sense of who I am” (Anonymous, 1994). We argue that this lifelong challenge can be captured through narrative identity, a rich, well-established theoretical and empirical approach to studying the narrative self in personality and developmental psychology. In this paper, we systematically review the literature on narrative identity in psychosis, describe the fully formed narrative identity deficits observed in schizophrenia-spectrum disorders, propose a developmental model for how these deficits may emerge in adolescence and early adulthood, and suggest specific hypotheses and future directions for research in this area.

The Narrative Self in Psychotic Disorders

What do we mean by “the self”? Several theorists have distinguished the minimal self, the consciousness of oneself as a subject of experience with agency and ownership in the present moment, from the narrative self, the temporally-extended consciousness of oneself as a individual with a past, present, and future (Damasio, 1999; Gallagher, 2000; Kircher et al., 2003; B. Nelson, Parnas, et al., 2014; Singer et al., 2013; Zahavi, 2008). The minimal self operates in the moment, monitoring unconscious representations of bodily states and the external environment and producing subjective experience with a sense of agency and a sense of ownership (Damasio, 1999; Gallagher, 2000; Northoff & Panksepp, 2008; Sass & Parnas, 2003). Over time, an extended or narrative self emerges as the minimal self triggers encoding and recall of self-relevant memories, producing a continuous sense of self that links present experiences to the remembered past self and the imagined future self (Damasio, 1999; Gallagher, 2000; B. Nelson, Parnas, et al., 2014).

Phenomenologically, individuals with schizophrenia experience dramatic disturbances in the minimal self. The minimal self loses its vitality and stability as a subject of awareness (diminished self-affection) while at the same time aspects of self-experience may seem external or alien (hyper-reflexivity) (Sass & Parnas, 2003). These disturbances in the minimal self are well-established features and perhaps even core symptoms of schizophrenia (de Vries et al., 2013; Henriksen & Parnas, 2012; Hur et al., 2014; B. Nelson, Whitford, et al., 2014; B. Nelson, Parnas, et al., 2014; Parnas & Handest, 2003; Sass & Parnas, 2003). How might these minimal self-disturbances affect the narrative self? If the narrative self emerges over time from coactivation of the minimal self with autobiographical memory, then it should be affected by disturbances in the minimal self. For instance, diminished self-affection might lead to a sense that one’s memories and imagined future are fading or unstable, while hyper-reflexivity might lead to a sense that unfamiliar events have been somehow inserted into one’s life story. Disturbances in the minimal self will ripple outward to disturb the narrative self.

A growing body of evidence supports this claim, showing that schizophrenia interrupts the incremental development of the adult self and leaves behind a fragmented life story dominated by psychiatric illness. The most personally significant memories of individuals with schizophrenia are more likely to include psychotic episodes, hospitalizations, and traumatic events, whereas normative adults’ significant memories are more likely to include achievement and relationships (Berna et al., 2011a, 2011b; Raffard et al., 2009, 2010; Wright et al., 2019). Individuals with pronounced delusions also describe delusion-related memories (e.g., memories of being spied on or conspired against) as being unusually vivid, emotionally intense, and central to the self, highlighting the personal impact of these unusual experiences (Berna et al., 2017). Even more fundamentally, the basic developmental trajectory of the life story is altered by schizophrenia. Significant memories occur earlier in life (age 15-24 and especially age 15-19, compared to age 20-29 and especially age 20-24 in normative adults) (Holm et al., 2017; Raffard et al., 2009, 2010; Ricarte et al., 2017). This shift in the “reminiscence bump” (Koppel & Berntsen, 2015) is likely caused by the onset of psychosis during this crucial developmental stage: the proportion of significant memories increases steadily in the years immediately preceding a psychotic diagnosis, then drops to almost nothing in the years following diagnosis (Holm et al., 2017). Individuals with schizophrenia struggle to maintain coherent selves and identities throughout the life course in the face of altered life experiences and interrupted developmental trajectories.

Narrative Identity

To understand disturbances in the narrative self in the psychosis spectrum, how are we to operationalize the narrative self? In personality and developmental psychology, researchers have recognized that the psychological self operates simultaneously at multiple levels. Not only does the self function as a social actor defined by traits, and a motivated agent defined by goals and values, but also as an autobiographical author mining personally significant experiences to construct a life story connecting the present self to the remembered past and the imagined future (McAdams, 2013a, 2015; McAdams & McLean, 2013). Human beings evolved to make sense of themselves and their worlds through language and storytelling (Donald, 1991; Dor, 2015; Leary & Buttermore, 2003; McAdams, 2019; McAdams & Cowan, 2020). Psychological research has found that life stories not only record the facts of our lives, but also provide our lives with meaning, purpose, and continuity through time (Adler et al., 2016; McAdams, 2013a; McAdams & McLean, 2013). The study of narrative identity, the experience of self as an autobiographical author crafting an “internalized and evolving life story” (McAdams & McLean, 2013, p. 233), has extended personality and developmental research into the arena of life stories, generating a robust empirical and theoretical literature on the psychological processes that make up the narrative self as author of a life story.

Three distinct dimensions of narrative identity have been identified in nonclinical populations (McLean et al., 2020). The first dimension captures basic structural elements that provide the facts, context, and temporal sequencing to make a life narrative comprehensible. The second dimension captures motivational and affective themes in the recalled experiences of the narrated self. Autobiographical storytelling conveys episodic information about the experiences of an autobiographical protagonist, or narrated self (McAdams & McLean, 2013). Many narrative identity variables reflect specific motivational and affective characteristics of this narrated episodic experience. For instance, is a personal story dominated by positive or negative emotion; is the narrated self an active or passive participant in events? The third dimension captures the metacognitive efforts of the narrating self, or the autobiographical narrator, to make sense of experiences and connect them to the present day. In addition to conveying episodic information, autobiographical storytelling extracts semantic details about an episode to situate it within a broader life story and draw conclusions about the self and the world (McAdams & McLean, 2013). Many narrative identity variables reflect specific processes of semantic, autobiographical reasoning about a narrated experience. For instance, does the individual express a lesson learned or insight gained from an experience; do they feel that the experience revealed or changed something about themselves?

These three dimensions—structure, motivational/affective themes, and autobiographical reasoning—were derived empirically in several factor analytic studies (Adler et al., 2018; Cowan, Chen, et al., 2019; McLean et al., 2020), most notably in a large-scale study examining 16 narrative identity variables in 2,565 autobiographical episodes narrated by 855 participants (McLean et al., 2020). The three dimensions differentially relate to important psychological outcomes, with motivational and affective themes linked to well-being (Adler et al., 2016; McLean et al., 2020), and autobiographical reasoning linked to developmental processes such as generativity, wisdom, and ego integrity (Adler et al., 2016; McLean & Pratt, 2006; Pals, 2006). This “Narrative Big Three” provides a taxonomy of psychological processes supporting narrative identity in healthy populations. The three dimensions can also organize findings on dysfunctions of narrative identity in clinical populations (Adler & Clark, 2019), as in a recent review of narrative identity processes in personality disorders (Lind et al., 2020). The current review extends the three-dimensional model into the psychosis spectrum, considering narrative self-disturbances as disruptions of narrative identity along each of the three dimensions.

The Psychosis Spectrum

Substantial research evidence now supports the idea that psychotic experiences form a continuum in the population, ranging from mild and transient psychotic-like experiences to debilitating symptoms of chronic psychotic disorders. The prototypical and most common DSM-5 schizophrenia-spectrum disorder is schizophrenia, which combines positive symptoms such as hallucinations and delusions, negative symptoms such as apathy and diminished emotional expression, and/or disorganized symptoms such as disorganized speech and thought (American Psychiatric Association, 2013; Kotov et al., 2020; Reininghaus et al., 2019). Schizophrenia is one of the leading causes of disability worldwide (Vos et al., 2015) and costs the global economy billions of dollars per year (Chong et al., 2016). Yet schizophrenia is relatively rare. Its lifetime prevalence is around 0.5% (Moreno-Küstner et al., 2018; Saha et al., 2005; Simeone et al., 2015), which rises to roughly 0.8% when including related diagnoses of schizophreniform disorder (symptoms of schizophrenia with a duration less than 6 months) and schizoaffective disorder (symptoms of schizophrenia plus pronounced mood symptoms) (Moreno-Küstner et al., 2018; Simeone et al., 2015).

By contrast, nonclinical psychotic-like experiences are surprisingly common in the general population: 10-25% of adults have experienced hallucinations (Johns & van Os, 2001), and one meta-analysis estimated that 17.5% of adults fall within a broad psychosis phenotype (J. van Os et al., 2000). In fact, 75 to 90% of all reported psychotic-like experiences are transitory and do not require clinical attention (J. van Os et al., 2009). Convergent genetic, neuropsychological, social, and environmental findings support a “fully dimensional” model of psychosis or schizotypy, in which nonclinical psychotic-like experiences and diagnosable psychotic disorders form opposite ends of a continuous distribution (Claridge & Beech, 1995; DeRosse & Karlsgodt, 2015; M. T. Nelson et al., 2013). This distribution is both phenomenological, with individuals at different points along the continuum experiencing different severities of psychotic-like experiences, and temporal, with psychotic-like experiences becoming more severe over time in those who eventually develop formal psychotic disorders (Barrantes-Vidal et al., 2015; Claridge & Beech, 1995; M. T. Nelson et al., 2013; Jim van Os & Reininghaus, 2016). Finally, a range of neuromotor, neuropsychological, and physiological endophenotypes appear in attenuated form in relatives of individuals with psychotic disorders (Allen et al., 2009), suggesting that latent biological mechanisms are similar throughout the psychosis spectrum.

For purposes of this review, the fully dimensional psychosis spectrum can be subdivided into three sections. The mildest or earliest section is characterized by trait-like vulnerability to psychosis. This section would include samples labelled, for example, as nonclinical psychosis, schizotypy, psychometric risk, or unaffected family members. Trait vulnerability to psychosis captures a pre-existing pool of individual differences that confer elevated risk for psychosis, such as genetic factors, cognitive style, and personality traits (M. T. Nelson et al., 2013; Jim van Os & Reininghaus, 2016). These variables are typically observed physiologically or psychometrically rather than clinically, as their associated psychotic-like experiences are not significant enough to warrant clinical attention. The middle section of the psychosis continuum is characterized by imminent risk for psychotic disorders. Terminology is evolving, but typical labels for samples in this section would be clinical high risk for psychosis, ultra high risk for psychosis, at-risk mental state, or DSM-5 Attenuated Psychotic Syndrome. Individuals at imminent risk for psychosis may exhibit brief or attenuated psychotic symptoms and/or functional decline, warranting clinical attention but not meeting criteria for a psychotic disorder. Roughly 10-40% of these individuals will develop a diagnosable psychotic disorder within 24 months of identification (Fusar-Poli et al., 2013, 2016). Finally, the most severe section of the spectrum is characterized by formal schizophrenia-spectrum disorders (SSD; most commonly schizophrenia, schizoaffective, and schizophreniform disorders), as discussed above. Within SSD, a subcategory of first-episode psychosis or early-course psychosis has unique developmental relevance. Diagnosis is somewhat fluid in these samples, with individuals often receiving provisional diagnoses at intake (Addington et al., 2006; Salvatore et al., 2009; Wright et al., 2019); however, these individuals typically receive schizophrenia-spectrum diagnoses within the first year of treatment (Addington et al., 2006; Salvatore et al., 2009; Verma et al., 2012). From a developmental perspective, first-episode psychosis samples is an important subgroup within SSD, as the acute impact of a disorder at the time of diagnosis may be different from its long-standing impact over the following decades.

A fully dimensional model of psychosis has important implications for the study of narrative self-disturbances. It suggests that narrative self-disturbances and issues with lifespan identity development, observed most severely in SSD, probably also occur at varying intensities throughout the psychosis continuum. Moreover, individuals typically pass through intermediate stages of trait vulnerability and imminent risk before developing a full-fledged psychotic disorder (Barrantes-Vidal et al., 2015; M. T. Nelson et al., 2013; Jim van Os & Reininghaus, 2016). Research findings in these populations can therefore shed light on the developmental trajectory of narrative self-disturbances, which likely follow the same pattern as other psychotic symptoms, originating in earlier or milder forms of psychosis before becoming severe enough to warrant clinical attention in SSD (Barrantes-Vidal et al., 2015; Debbané & Barrantes-Vidal, 2015). This review therefore encompasses the entire psychosis spectrum, divided into four categories: trait vulnerability, imminent risk, first-episode psychosis, and SSD.

Narrative Identity in The Psychosis Spectrum

The symptoms of psychotic disorders are known to impact many of the processes that maintain narrative identity. These include deficits in episodic and semantic memory (Berna, Potheegadoo, et al., 2016; Chen et al., 2016; Danion et al., 2005, 2007; Dimaggio et al., 2012; Kircher et al., 2003; Raffard et al., 2016; Ricarte et al., 2014, 2017; Riutort et al., 2003; Zhang et al., 2019), temporal perception and indexing (Allé, Gandolphe, et al., 2016; Ben Malek et al., 2019; Berna, Potheegadoo, et al., 2016; Kircher et al., 2003; Martin et al., 2014; Pauly et al., 2014; Potheegadoo et al., 2012), event segmentation (Zalla et al., 2004), imagining hypothetical futures (Chen et al., 2016; D’Argembeau et al., 2008; Hazan et al., 2019; Raffard et al., 2013, 2016), speech complexity (B. Buck et al., 2015; B. Buck & Penn, 2015; Minor et al., 2015), and metacognition (Lysaker, Molly Erickson, et al., 2010; Lysaker et al., 2013; Morrison et al., 2007). Each of these deficits likely impacts individuals’ ability to construct rich, coherent self-narratives. Moreover, the fully dimensional model of psychosis predicts that milder or earlier forms of psychotic-like experiences would also impact self-narratives in an attenuated fashion. Many authors have therefore begun to examine life stories and narrative identity in the psychosis spectrum.

Many of the narrative identity variables studied in healthy populations have now also been studied in SSD. However, these variables have not yet been systematically organized under the rubric of narrative identity. While several English-language reviews of autobiographical memory in the psychosis spectrum have included life narrative studies (Berna, Potheegadoo, et al., 2016; Kwok et al., 2021; Ricarte et al., 2017; Zhang et al., 2019), none have synthesized these findings within a narrative identity framework. This has left several important questions unresolved. Are all three narrative dimensions disrupted in psychotic disorders? Are some more disrupted than others? What is the precise nature of disruptions on each dimension? Do narrative identity difficulties occur with varying severity throughout the psychosis spectrum? Developmentally, what would this suggest about their potential origins in earlier, milder forms of psychotic-like experiences, prior to the onset of a psychotic disorder? Clinically, are narrative dimensions linked to specific clinical symptoms or functional outcomes? And methodologically, do some methods of assessing narrative identity produce more reliable effects than others?

To address these questions, we extend the three-dimensional taxonomy to systematically review the literature on narrative identity in psychosis. Based on the reviewed literature, we then propose a developmental model of narrative identity in the psychosis spectrum, connecting the three narrative identity dimensions to developmental processes, psychotic-like experiences, and psychotic symptoms.

Scope of the Systematic Review

This review presents a systematic, qualitative synthesis and theoretical integration of research on narrative identity in the psychosis spectrum. Relevant studies were defined as quantitative, empirical studies that included at least one narrative identity variable (for overviews of common narrative identity variables, see Adler et al., 2017; McAdams & McLean, 2013; McLean et al., 2020; for narrative identity variables included in this review, see Appendix Table A.1). A prototypical narrative identity study would follow this sequence: 1., participants are interviewed using questions designed to elicit personally relevant life stories; 2., trained raters score narrative identity variables from the interview transcripts (e.g., a rater might assign a score from 1-5 on the variable of “agency”; sometimes dichotomous ratings, computerized ratings, or self-ratings are also used); 3., narrative identity variables are then analyzed through descriptive and inferential statistics. See Adler (2017) for an accessible guide to narrative identity research methods. Qualitative designs are often highly informative for hypothesis generation and theory building (see, e.g., Josselson, 2009; Toolis & Hammack, 2015); however, the current review focuses on quantitative hypothesis-testing studies.

The psychosis spectrum was defined as a dimension encompassing trait vulnerability, imminent risk, first episode psychosis, and SSD. See “The Psychosis Spectrum” above for definitions and discussion of these terms. To highlight core psychotic experiences, which have been the focus of research on self-disturbances, this review did not include other disorders with psychotic features (e.g., depression or bipolar disorder with psychotic features). Although our search criteria included all DSM-5 schizophrenia-spectrum disorders, all of the SSD studies we located used samples diagnosed with schizophrenia and schizoaffective disorder, the two most common non-substance-related DSM-5 schizophrenia-spectrum disorders (American Psychiatric Association, 2013; Moreno-Küstner et al., 2018; Simeone et al., 2015). For the purposes of this review, therefore, schizophrenia-spectrum disorders (SSD) refers to schizophrenia and schizoaffective disorder.

Methods

The design of the systematic review followed PRISMA guidelines (Moher et al., 2009). We carried out literature searches of the Web of Science, MedLine, and PsycInfo databases up to May 2021 for combinations of the search term root words self-knowledge, narrative identity, self narrative, life narrative, life story, narrative self, or extended self, and psychosis, psychotic, delusional, schizophrenia, schizoaffective, or schizophreniform. We also examined the reference sections and citing literature of relevant articles. Articles were screened by title and abstract, then full-text articles were assessed for eligibility. See Figure 1 for a PRISMA flowchart of records excluded at each stage. Inclusion criteria were: 1., quantitative, empirical articles; 2., written in English; 3., which asked participants to reflect on personally relevant life stories (studies could include hypothetical future narratives, but could not exclusively focus on the future); 4., with a sample fitting the psychosis spectrum as defined above.

Figure 1.

Records excluded at each stage of the systematic review and screening process.

Results and Discussion

Review of the Literature

We identified 245 unique records through literature searches and other methods. The full texts of 66 articles were examined, 36 of which were excluded, with the most common reason being that they focused on a related subject such as memory, metacognition, or linguistic complexity, and did not include measures of narrative identity (see Figure 1). This left 30 quantitative, empirical studies of narrative identity in the psychosis spectrum which were included in the review. Appendix Table A.1 shows the characteristics of included studies. After accounting for instances where multiple studies reported on the same participant sample, the reviewed studies included 978 individuals with SSD and 2,881 individuals without SSD, for a total of 3,859 unique individuals.

The literature review is organized around the three narrative identity dimensions of structure, motivational/affective themes, and autobiographical reasoning. Each dimension contains multiple components (e.g., the dimension of motivational/affective themes contains components of emotional tone and agency). Results are organized by narrative identity components, nested within dimensions. Within each component, results are organized by study population, starting with SSD and working backward through the psychosis spectrum to examine first episode psychosis, imminent risk, and trait vulnerability. The overall pattern of results is summarized in Figure 2 and Table 1.

Figure 2.

Theoretical structure of narrative identity in schizophrenia-spectrum disorders. Latent variable structure based on McLean, K. C. et al. (2020). The Empirical Structure of Narrative Identity: The Initial Big Three. Journal of Personality and Social Psychology, 119(4), 920-944.

Table 1.

Narrative identity across the psychosis spectrum

	Trait Vulnerability	First-Episode Psychosis	Schizophrenia-Spectrum Disorders
Structure	Intact memory specificity (in self-defining memories)	Intact memory specificity (in self-defining memories) Specificity mediates cognition’s effect on functioning	Intact memory specificity (in self-defining memories) Less context, detail, and idea density Less temporal coherence
Motivational/Affective Themes	Possibly more negative emotional tone Less agency Less redemption	Possibly more negative emotional tone	More negative emotional tone Less agency Less communion More themes of unfulfilled agency and communion Redemption more common in illness narratives than non-illness narratives
Autobiographical Reasoning	Intact self-event connections and meaning-making More self-focus and exploratory processing	Less meaning-making	Fewer self-event connections Less meaning-making overall Meaning-making partially preserved in illness narratives Less thematic and causal coherence

Note: Unless otherwise noted, all findings are compared to healthy control groups. See the corresponding sections of the text for details including citations. Thus far, no studies have examined narrative identity in imminent risk samples (e.g., clinical high risk, attenuated psychotic syndrome).

Structural Variables: Disjointed Structure

Structural processes of narrative identity in SSD are disjointed, marked by temporal incoherence and a lack of detail and context, which render the life story incomprehensible to listeners.

Specificity, detail, and context.

A coherently structured narrative requires a certain amount of specific, contextualized detail. Memory specificity is known to be impaired in various recall tasks in SSD. Interestingly, in a recent review of autobiographical memory in SSD, Zhang and colleagues (2019) located only three studies which reported no differences in specificity between SSD and controls, two of which were narrative identity studies of self-defining memories (i.e., vivid, familiar memories which occurred at least one year in the past, have been frequently recalled, are personally significant, and help the individual know who they are as a person) (Raffard et al., 2009, 2010). Since that review was published, a third study has reported no differences in specificity for self-defining memories between first episode psychosis and healthy controls (Wright et al., 2019). Similarly, one study of trait vulnerability found no differences in memory specificity between high and low trait vulnerability groups (Hazan et al., 2019). All four studies dichotomized narratives as “specific” (happening in a particular time and place) vs. “nonspecific”, which has lower statistical power compared to rating the level of specificity on an ordinal scale; however, some autobiographical memory studies (e.g., Neumann et al., 2007) have found significant results in SSD using similar dichotomous methods. The self-defining memory prompt may facilitate specific recall, as it asks participants to identify personally significant memories rather than arbitrary memories such as those associated with a particular emotion (c.f. Neumann et al., 2007).

However, specificity is only one component of a structurally cohesive narrative. Narrative structure also entails a certain amount of detail and context. One of the self-defining memory studies which reported no group differences in specificity (Raffard et al., 2010) also scored contextual coherence (background information to contextualize a story, including information about location and time) using trained raters (Reese et al., 2011), finding impaired context coherence in the SSD group. Similarly, Gruber and Kring (2008) found that narratives of emotional events were less elaborated, with less clear meanings, in individuals with SSD compared to controls. Studies using computerized analysis of life stories in SSD have also found impairments in idea density (the ratio of propositions/assertions divided by the word count) (Moe et al., 2016, 2018) and deep semantic cohesion (an index of details required for story comprehension) (Willits et al., 2018). Even when individuals with SSD recall specific memories in response to narrative identity prompts, they still struggle to elaborate on detail and context.

These deficits have been linked to clinical symptoms and functional outcomes. While one small study (N = 20) reported no correlations between specificity and clinical variables (Raffard et al., 2009), a better powered follow-up study (N = 81) found a correlation between poorer specificity and negative symptoms (Raffard et al., 2010). A third study linked specificity to functional outcomes in first episode psychosis, reporting that specificity mediated the relationship between impaired neurocognition (memory performance) and functional outcomes (Wright et al., 2019). Richness of detail has been examined in a series of studies by Lysaker and colleagues, finding that impairments in detail are associated with poorer insight (Lysaker et al., 2002; Lysaker, France, et al., 2005) and poorer cognitive ability and quality of life (Lysaker et al., 2008) in individuals with SSD. This series of studies did not find significant correlations between detail and negative symptoms; however, their point estimates fell within the 95% confidence interval of the larger study which linked specificity to negative symptoms (Raffard et al., 2010). These data would be consistent with small to medium correlations between negative symptoms and richness of detail. In sum, individuals who experience more severe negative or cognitive symptoms appear more likely to tell incomprehensible life stories missing key facts and contextual details. However, these findings have been somewhat inconsistent and would benefit from further study.

Temporal coherence.

A coherently structured narrative also requires events to be temporally ordered in a logical sequence. Raffard and colleagues (2010) reported lower levels of chronological coherence in the narratives of individuals with SSD compared to controls, assessed by trained raters using the Narrative Coherence Coding scheme (Reese et al., 2011). In two studies, Allé and colleagues (2016; 2015) measured local temporal coherence (how clearly any given episode can be placed in the life course) and global temporal coherence (how clearly various episodes can been sequenced in relation to one another), finding that global temporal coherence, but not local temporal coherence, was impaired in SSD. In other words, personal narratives contained enough facts that raters could roughly place them in the life course of individuals with SSD, but participants did not effectively sequence the narratives in relation to one another. This ineffectiveness may be explained by higher rates of anachronies (deviations from a narrative’s temporal sequence) in individuals with SSD, observed in both studies (Allé, Gandolphe, et al., 2016; Allé et al., 2015). Individuals with SSD seem to have difficulty narrating their life stories with a coherent temporal structure. For instance, an individual might tell a life story that skips mercurially between painful childhood memories, mundane present experiences, and grandiose future plans, leaving listeners adrift and unable to follow the temporal flow.

Inconsistent links have been reported between temporal coherence and clinical symptoms. Allé and colleagues found no such relationships in one of their studies of local and global temporal coherence (2015). Working from the Narrative Coherence Rating Scale, Lysaker and colleagues linked temporal coherence to hopelessness, cognitive symptoms, and impaired insight in one study (2002), but to positive symptoms in another study (2005). Complicating matters further, in a separate pair of studies, Holm and colleagues adopted a temporal macrostructure coding scheme following Habermas (2009) that analyzes the coherence of the life story’s beginning and end as well as the chronology of events within the life story. They found no differences between individuals with SSD and controls on any of these variables (Holm et al., 2016, 2018), but they reported that coherent beginnings of life stories were associated with less negative symptoms in one of their SSD samples (Holm et al., 2016). This mix of findings could reflect variation between patient samples or differences between the various coding systems used. Future research comparing these coding systems to one another in the same sample (see, e.g., Adler et al., 2018) would determine their intercorrelations and their common and unique links to symptoms.

Outside SSD samples, one study has examined temporal coherence using a self-report questionnaire in a trait vulnerability sample, finding no difference between high and low trait vulnerability on temporal coherence (Hallford & Burgat, 2014).

Summary of structural variables.

Individuals with SSD show marked deficits in their ability to construct structurally coherent and comprehensible narratives. Their disjointed life narratives are lacking in detail, context, and temporal coherence. In SSD, these deficits may relate to negative and cognitive symptoms. Structural deficits have not been observed in trait vulnerability samples, and the limited data available in first-episode psychosis suggest that structural deficits may play a role in functional outcomes.

Motivational/Affective Themes: A Focus on Suffering

The motivational and affective themes of narrative identity in SSD are marked by painful emotions, passivity, constraint, and alienation from others. The autobiographical protagonist is suffering in these life stories. However, there may be hope for redemption through narratives of recovery and desires for agency and communion.

Emotional tone.

The emotional tone of a personal narrative is a basic, primary affective theme. Positive emotional tone is robustly related to mental health and well-being in healthy adults (Adler et al., 2016; Cowan, Chen, et al., 2019; McLean et al., 2020) and adults with personality disorders (Lind et al., 2020). Several studies have examined personal narratives’ emotional tone in SSD. Allé and colleagues (2016) found that past memories and imagined future events were self-rated as more negative by individuals with SSD than healthy controls. Holm and colleagues (2016) reported that past life chapters were self-rated as more negative in SSD compared to controls, while self-defining memories were self-rated similarly in the two groups (although point estimates were similar for the two types of memories and statistical power was higher for life chapters). Berna and colleagues (2011a, 2011b) reported no differences in self-rated emotional tone for self-defining memories between SSD and healthy controls. Finally, Bennouna-Greene and colleagues (2012) reported that memories were more negative when assessed by trained raters, but not when self-statements associated with the memories were rated by participants. In sum, personal narratives take on a more negative tone for individuals with SSD, particularly when assessed by trained raters.

Two studies have also compared SSD and clinical comparison samples. Jensen and colleagues (2020) reported that past life chapters were more negative in SSD compared to healthy controls, but not in SSD compared to depression. Moe and Docherty (2014) reported that narrative self-evaluations were even more negative in SSD than in bipolar disorder with psychotic features (both of which were more negative than healthy controls). This second study excluded bipolar individuals in a current depressive episode, so its results would be consistent with the first study, suggesting that the extent of negative emotional tone in life stories may be roughly similar for depression and SSD. In other words, similar to individuals with depression, an individual with SSD might tell a life story about personal failures and losses suffused with sadness and regret. One study has examined the valence of self-defining memories in a first episode psychosis sample, finding no significant difference from controls (Wright et al., 2019). However, this study dichotomized narratives as positive or negative, which limited its statistical power, and reported a small to medium effect size in the direction of more negative tone in the SSD group (Cox logit d = .32).

With respect to earlier or milder forms of psychosis, two studies on self-defining memories in trait vulnerability found conflicting findings, with one reporting more negative valence in the high trait vulnerability group (Berna, Göritz, et al., 2016) and another reporting no group difference between high and low trait vulnerability groups (Hazan et al., 2019). Although evidence is currently much weaker outside SSD, negative emotional tone may be present to some extent in milder or earlier forms of psychosis.

Method variance likely played a role in findings on emotional tone. Several studies dichotomized memories as “positive” or “negative”, whereas a more statistically powerful approach would be to rate the extent of positive and negative emotional tone. Two studies also reported no group differences between SSD and healthy controls in affective experience before and after memory retrieval (pre-post changes in PANAS scores) (Raffard et al., 2009, 2010). However, affective changes during autobiographical narration do not necessarily reflect a story’s valence (Turner et al., in press) and may not be a reliable proxy for narrated emotional tone. Finally, several studies asked participants to rate their own stories’ emotional tone. The use of self-ratings with varying definitions may have increased measurement error and suppressed true effects in these studies. A recent study compared self-ratings and trained coders’ ratings of emotional tone in nonclinical samples, finding medium to large correlations between self-ratings and trained coders’ ratings, and suggesting that variation in the specific instructions given to participants may affect their responses (Dunlop, Harake, et al., 2020). In future research, it would be most reliable for trained raters to assess emotional valence ordinally from a validated coding manual rather than using alternative methods such as positive/negative dichotomization, changes in PANAS scores, or self-ratings.

Agency and communion.

Agency and communion are two of the most fundamental motivational themes expressed in life narratives, capturing basic human drives toward individual action and accomplishment (agency) and relational affiliation and connectedness (communion). Agency and communion have been studied in life narratives in various studies, with agency being particularly closely linked to mental health and well-being (Adler et al., 2016; Cowan, Chen, et al., 2019; McLean et al., 2020). The process of recovery from SSD has also been described as discovering a more agentic self (Davidson & Strauss, 1992), suggesting that agency at the level of the narrative self may be particularly relevant in the psychosis spectrum. Several studies have compared narrated agency between SSD and healthy controls, finding that individuals with SSD narrate their lives with less agency than controls (Bennouna-Greene et al., 2012; Lysaker, Wickett, et al., 2005; Moe & Docherty, 2014). Two of these studies also reported lower levels of communion compared to healthy controls (Lysaker, Wickett, et al., 2005; Moe & Docherty, 2014). Two studies have compared individuals with SSD against clinical control groups, finding that individuals with SSD report lower levels of agency and communion than HIV-positive individuals (Holm et al., 2020), and similar levels of agency and communion to depressed individuals (both groups reported less agency and communion than healthy controls; Jensen et al., 2021). Notably, these studies used various prompts and coding methods to converge on similar findings. Impaired narrative agency has been linked to a host of clinical variables including negative symptoms (Jensen et al., 2021; Lysaker, Wickett, et al., 2005), impaired cognition (Jensen et al., 2021; Lysaker, Wickett, et al., 2005), functional impairment and lower quality of life (Jensen et al., 2021; Lysaker et al., 2006; Lysaker, Ringer, et al., 2010), lower self-esteem (Holm et al., 2020), and greater hopelessness (Holm et al., 2020; Lysaker et al., 2006; Lysaker, Ringer, et al., 2010).

Holm and colleagues (2018) added an important nuance by separating desires for agency and communion from fulfillment of agency and communion. They found no group differences in desires for agency and communion, but they found that agency and communion were less likely to be fulfilled in the life stories of individuals with SSD compared to healthy controls. For instance, one individual with SSD felt education was personally important, but “it has been a while since I have pursued an education because so many things have gone wrong for me. Every time I start something new, then after 6 months or so, things start to go astray” (quoted in Holm et al., 2018). SSD may impact individuals’ ability to construct motivated agentic and communal life stories rather than their desire to do so.

Finally, two studies have examined agency in trait vulnerability samples (Hazan et al., 2019; See et al., 2020). Both studies found lower agency in the high trait vulnerability group compared to the low trait vulnerability group, suggesting that non-agentic or passive life stories may appear throughout the psychosis spectrum.

Redemption.

Finally, redemption is a motivational/affective theme which captures movement in a story from suffering toward freedom, growth, or enhancement. Strongly shaped by culture—as in American redemptive narratives of emancipation or upward social mobility—redemption is distinct from overall emotional tone: “redemptive narratives are not simply happy stories; rather, they are stories of suffering and negativity that turn positive in the end” (McAdams, 2013b, p. 27). Redemptive themes are even better predictors of well-being than emotional tone in nonclinical samples, and they have been strongly linked to generativity, a concern for and commitment to the well-being of future generations (McAdams, 2013b). Only two studies have examined redemption in the psychosis spectrum. In a small study of self-defining memories in outpatients with SSD (n = 24), Berna and colleagues (2011a) reported no group differences from healthy controls on overall redemptive content, but found that redemptive content was more likely to be present in narratives about illness in the SSD sample whereas it was more likely to be present in narratives about other topics in the control sample. At the other end of the psychosis spectrum, See and colleagues (2020) reported that healthy individuals with high trait vulnerability express less redemptive imagery than individuals with low trait vulnerability. There is a substantial conceptual distance between these two study populations. Perhaps redemption in the narrative self is in some way a protective factor against serious mental illness, such that its absence in trait vulnerability indicates risk and its presence in illness narratives in SSD indicates recovery. However, this interpretation is speculative, and further research would be required to define the role of redemption in the narrative self in the psychosis spectrum.

Summary of motivational/affective variables.

Individuals throughout the psychosis spectrum tell life stories in which the autobiographical protagonist suffers through painful emotions, passively accepting his or her inability to change events or meaningfully connect with others. This focus on suffering is seen not only in SSD but also in trait vulnerability and first episode psychosis. Two encouraging signs suggest that all hope is not lost in these life stories: individuals may still desire agency and communion even when these desires are unfulfilled; and may still craft redemptive stories of recovery from illness.

Autobiographical Reasoning: Detached Narration

Autobiographical reasoning in SSD is marked by disconnection between events and the present self; an absence of meaning, lessons, or insights; and thematic and causal incoherence. The autobiographical narrator is largely absent from these life stories, relinquishing any potential benefits from metacognitively making sense of experience.

Self-event connections.

One basic element of autobiographical reasoning is drawing connections between the events in a personal narrative and the present-day self. Did the narrated events illustrate or reveal something important about the self, or did they cause the self to change in some way? Two studies have found lower levels of self-event connections in the self-defining memories of individuals with SSD compared to controls (Allé, d’Argembeau, et al., 2016; Raffard et al., 2010). For instance, an individual might narrate a self-defining memory of a first psychotic episode and hospitalization as if it bears no relation to the present self who lives with a schizophrenia-spectrum diagnosis.

This deficit in self-event connections seems to depend to some degree on clinical insight. Studies in which participants have rated self-event connections, personal impact, or subjective centrality in their own narratives have found no differences between individuals with SSD and controls (Allé. et al., 2015; Berna et al., 2011a, 2011b; Holm et al., 2016). However, one study has compared self-rated and experimenter-rated self-event connections in a SSD sample, finding fewer self-event connections when rated by experimenters but not when self-rated by participants (Bennouna-Greene et al., 2012). This result suggests that individuals with SSD may have incomplete metacognitive insight into thematic links between memories and the present self. In other words, even when they do connect past experiences to the present self, they do not notice that they have done so. Supporting this interpretation, Raffard and colleagues (2010) found that SSD outpatients with poorer insight and more negative symptoms discussed fewer self-event connections in their self-defining memories. Impaired insight may impact individuals’ ability both to construct self-event connections and to metacognitively evaluate those connections.

Interestingly, one study has reported more negative self-event connections (in which the narrative changed the self for the worse) in individuals with SSD compared to healthy controls, and no differences in positive self-event connections (Jensen et al., 2020). This finding is at odds with the overall lower levels of self-event connections observed in SSD. However, the study also included a group of individuals with depression, who reported even more negative self-event connections than the SSD group. This pattern was consistent with other measures of emotional tone in the study, suggesting that the specific question (“has the chapter caused you to change in a negative way?”; Jensen et al., 2020, p. 5) may have prompted emotional tone rather than self-event connections.

In trait vulnerability samples, two studies found no differences between high and low trait vulnerability groups on self-rated event centrality (Berna, Göritz, et al., 2016) or experimenter-rated self-event connections (See et al., 2020). The ability to draw connections between personal narratives and the self does not seem to be affected throughout the psychosis spectrum; rather, it appears to be specifically affected in SSD.

Meaning-making.

Meaning-making, the tendency to derive lessons or insights from past events, also appears to be impaired in SSD. Studies of meaning-making in self-defining memories consistently find lower meaning-making among individuals with SSD compared to healthy controls (Berna et al., 2011a, 2011b; Raffard et al., 2009, 2010). One study has also reported lower levels of meaning-making in a first episode psychosis sample (Wright et al., 2019). In both early and chronic phases of psychotic disorders, individuals are less likely to extract lessons or insights from their personal experiences. For instance, an individual might tell a story of being hospitalized after stopping their medication regimen, without deriving any lessons to guide future behaviour.

Several studies have tested links between meaning-making and clinical variables, finding that impairments in meaning-making are not associated with neurocognition (Wright et al., 2019) but are associated with severity of negative symptoms (Berna et al., 2011b; Raffard et al., 2010). Interestingly, one trait vulnerability study found no differences in meaning-making between high and low trait vulnerability groups, and no relationship between meaning-making and a nonclinical measure of negative symptoms (negative subscale of the Community Assessment of Psychic Experiences) (Berna, Göritz, et al., 2016). This suggests that impairments in meaning-making may not run throughout the psychosis spectrum. Rather, they may be associated specifically with the negative symptoms in SSD.

Thematic and causal-motivational coherence.

In narrative coherence rating scales, thematic and causal-motivational coherence—the extent to which a narrator elaborates on themes and interprets causal links to other autobiographical experiences—are most closely tied to autobiographical reasoning (Adler et al., 2018; Allé et al., 2015). Several studies have found that these constructs are impaired in SSD. One study assessed both thematic and causal-motivational coherence in SSD, finding that both forms of coherence were lower than in healthy controls (Allé et al., 2015). That study’s authors published another study (which is technically beyond the scope of this review as it is not written in English) that links impaired thematic and causal-motivational coherence to executive dysfunction (Allé, Danion, et al., 2016). Raffard and colleagues (2010) also reported lower levels of thematic coherence in individuals with SSD compared to controls. Finally, one study asked participants to self-rate causal links between narratives and later experiences, finding no differences from controls (Holm et al., 2016), although this nonsignificant result may be attributable to the use of self-ratings. Overall, thematic and causal-motivational coherence are likely to be impaired in SSD.

Evidence is considerably weaker in trait vulnerability samples: one study found a trend suggesting possible lower causal coherence in a high trait vulnerability compared to a low trait vulnerability group (Hazan et al., 2019), while another found no group difference between high and low trait vulnerability (Hallford & Burgat, 2014). Further evidence would be required to draw conclusions about thematic and causal-motivational coherence in trait vulnerability.

Self-focus and exploratory processing.

Exploratory processing refers to the tendency to scrutinize, reflect on, or engage with complex aspects of one’s past experiences. Preliminary evidence suggests that individuals with mild psychotic-like experiences may engage in more exploratory processing than controls. One study (Hallford & Burgat, 2014) found that individuals at higher trait vulnerability reported thinking about their lives more frequently and retrieving more autobiographical memories to maintain a sense of self-continuity. Another study replicated this effect, and found that the effect partially mediated the relationship between psychotic-like experiences and lower self-concept clarity (Berna, Göritz, et al., 2016). Although these participants thought more intently about their past experiences, this paradoxically led to greater feelings of uncertainty about themselves. Both of these studies relied on an autobiographical memory questionnaire (Thinking About Life Experiences Scale) and did not ask participants to recall specific personal narratives. Nevertheless, exploratory processing may play a role in the development or maintenance of psychotic symptoms and/or narrative identity disturbances. Alternatively, exploratory processing could accompany mild psychotic-like experiences that are more indicative of general distress than an incipient psychotic disorder (Cowan & Mittal, 2020; Johns & van Os, 2001). Future studies examining exploratory processing in imminent risk and SSD samples could clarify this issue.

Summary of autobiographical reasoning variables.

Individuals with SSD, both in the early and chronic phases, have difficulty interpreting life events, assigning meaning to them, and connecting them to each other and to the present self. Notably, these deficits have not been observed in trait vulnerability samples. Thus, the available evidence suggests that autobiographical reasoning deficits accompany the symptoms of SSD, and do not occur throughout the psychosis spectrum. Intriguingly, meaning-making may be partially preserved in narratives about illness, suggesting a potential way forward for therapeutic interventions aimed at restoring autobiographical reasoning.

Structural Model of Narrative Identity in Schizophrenia-Spectrum Disorders

Key findings from the systematic review are shown in Table 1, grouped by narrative dimension and study population. The overall structure of narrative identity in SSD is shown in Figure 2. In SSD, multiple components of all three narrative identity dimensions are altered. Structural variables are characterized by disjointed structure; motivational/affective themes are characterized by a focus on suffering; and autobiographical reasoning (or lack thereof) is characterized by detached narration. The three dimensions of narrative disturbances may correlate in SSD, forming vicious circles in which each form of disturbance serves to maintain the others. For instance, detached narration of experiences may maintain a disjointed structure as an individual neglects to search for thematic and causal links between autobiographical experiences. Notably, such intercorrelations would confound relationships between symptoms and narrative dimensions when relationships with any one dimension are mediated by the other two dimensions. Links to symptoms may appear different when statistically controlling for the other narrative dimensions, when using comprehensive symptom measures and comprehensive assessments of narrative identity, or when examined in earlier or milder forms of psychosis.

Proposed Developmental Model of Narrative Identity in Schizophrenia-Spectrum Disorders

Only some narrative identity disturbances have been observed in individuals with trait vulnerability for psychosis. This raises an important question: when in the progression of schizophrenia-spectrum disorders might narrative identity disturbances emerge? By examining the available psychosis-spectrum data in relation to the normative development of narrative identity, can we suggest anything about timing or mechanisms for narrative identity disturbances?

Emerging adulthood, a transitional developmental stage between adolescence and adulthood (typically defined as ages 18 to 25) (Arnett, 2007), is a key developmental window for both normative narrative identity and psychotic illness. Normatively, identity development accelerates in adolescence (Steinberg & Morris, 2001) as deeper capacities develop to reason about others’ inner states (Sebastian et al., 2008); self-focus, self-evaluations, and self-conscious emotions become prominent (Sebastian et al., 2008; Steinberg & Morris, 2001); self-beliefs solidify and the self-concept becomes more complex and differentiated (Cole et al., 2001; Steinberg & Morris, 2001); and existential concerns emerge about proceeding into an uncertain future (Adamson et al., 1999). Simple forms of autobiographical consciousness and situated storytelling can be observed in parent-child interactions in childhood (McLean et al., 2007; K. Nelson & Fivush, 2020), and in early adolescence individuals begin to autobiographically reason about single experiences (McAdams, 2015; McAdams & McLean, 2013). As young people move into emerging adulthood in their late teens and early twenties, their use of autobiographical reasoning expands: autobiographical reasoning is now used not only to explain single experiences, but also to tie together a lifetime of experiences into an integrative and cohesive life story (McAdams, 2013a; McAdams & McLean, 2013). At this stage, a young adult can be said to have a fully formed narrative identity, one that explains how they came to be the person they are, and how they are becoming the person they will be in the future, providing life with some degree of unity and purpose (McAdams, 2013a).

How might the progression of an incipient schizophrenia-spectrum disorder interact with these developmental processes? The motivational and affective themes comprising a focus on suffering have been observed throughout the psychosis spectrum, and also in samples with mood disorders (Adler, 2012; Singer et al., 2013) and personality disorders (Lind et al., 2020). In fact, a focus on suffering has even been linked to well-being and subclinical depressive symptoms in healthy adult samples (Adler et al., 2016; Cowan, Chen, et al., 2019). In a sense, motivational/affective themes set a lower bar for narrative self-disturbance. A narrative focus on suffering seems to accompany the general distress inherent in psychopathology, unhappiness, or poor well-being (Adler et al., 2016; Caspi et al., 2014). Distress is also known to impact self-knowledge and cognitive biases at the trait level, for instance in the development of negative beliefs about the self in childhood such as “I am helpless” or “I am unloved” (Dozois et al., 2012; Noone et al., 2015)—beliefs which have also been linked to psychotic-like experiences (Cowan, McAdams, et al., 2019; Noone et al., 2015). Thus, it seems most likely that a focus on suffering would result from some of the same vulnerability factors that confer risk for psychosis, and therefore would predate the emergence of both psychotic disorders and narrative identity (McAdams, 2013a). One key vulnerability factor is general distress experienced early in life, for instance through childhood trauma, bullying victimization, or unsupportive family environments, all of which have been linked to future psychotic experiences (Carol & Mittal, 2015; Catone et al., 2015; Mayo et al., 2019; Wigman et al., 2012). For young people vulnerable to psychosis, the first steps into autobiographical reasoning in adolescence may already rest on painful, passive, or alienated memories and self-beliefs.

However, individuals with trait vulnerability to psychosis do not appear to show deficits in autobiographical reasoning. If anything, they exhibit more self-focus and exploratory processing than their peers (Berna, Göritz, et al., 2016; Hallford & Burgat, 2014). They may be unusually interested in mining their personal experiences for meaning and self-definition. This suggests that autobiographical reasoning deficits are not associated with vulnerability factors and general distress. Developmentally, autobiographical reasoning may emerge at a normative or even excessive level during adolescence for those at risk for psychosis. Normative autobiographical reasoning in this stage would fit with the adolescent “reminiscence bump” (Holm et al., 2017; Raffard et al., 2009, 2010; Ricarte et al., 2017) and the large proportion of self-defining memories that occur in adolescence for individuals who later develop SSD (Holm et al., 2017) The fact that these individuals encoded highly personally significant memories in middle and late adolescence suggests that they were probably engaged in significant autobiographical reasoning during this time.

Yet, paradoxically, narrative self-exploration can leave vulnerable individuals feeling less sure of who they are. The tendency to scrutinize and make meaning of autobiographical experience is associated with poorer self-concept clarity in individuals with trait vulnerability to psychosis (Berna, Göritz, et al., 2016; Cicero, 2017), suggesting that attempts to make meaning of experience actually lead to more confusion about the self. Confusion about the self, indicated by low self-concept clarity, has also been linked to aberrant salience and basic self-disturbances in nonclinical samples (Cicero et al., 2013, 2015, 2016, 2020, 2021; Gaweda et al., 2019). A heightened focus on integrating one’s own experiences may lead to confusion and fragmentation when those experiences fail to cohere with one another due to the perplexing effects of aberrant salience and basic self-disturbances. For young people at imminent risk for psychosis—experiencing attenuated psychotic symptoms and becoming progressively more perplexed by their own experiences—normative attempts at making sense of their lives through autobiographical reasoning may simply make them more aware of all the ways in which their experiences do not fit together. This would seriously interfere with structural integration of life experiences. Autobiographical reasoning may be confusing and largely ineffective at this stage, leading to disjointed structure in the emerging life stories of individuals at imminent risk for psychosis.

Imagine the subjective experience of trying to make sense of painful, disjointed events. The experience would likely be aversive, and in fact autobiographical reflection seems to be experienced as unpleasant for individuals with trait vulnerability for psychosis (Cicero & Kerns, 2011). How long would an individual continue to engage in autobiographical reasoning under these circumstances? Particularly if incipient negative and cognitive symptoms progressively impaired their metacognitive abilities, as is the case around the time of a psychotic diagnosis (Barbato et al., 2013; McLeod et al., 2014; Trauelsen et al., 2016)? Behavioral principles suggest that an individual will stop engaging in any aversive behavior that is not being rewarded. In other words, at some point the individual would likely give up on autobiographical reasoning. By the time they experience a first psychotic episode, individuals are engaged in less meaning-making compared to healthy controls (Wright et al., 2019), a trend which only becomes more pronounced in the chronic phase of SSD. These individuals seem to have “given up” on attempts to autobiographically reason about their lives, which have ultimately proven unsuccessful in establishing a coherent narrative identity. Cognitive feedback loops may then become self-reinforcing (Rector et al., 2005). Individuals may expect that autobiographical reasoning will be painful and that experiences will ultimately not make sense, and conclude that there is no point trying to autobiographically reason about experience.

Thus, by the time an individual develops a diagnosable schizophrenia-spectrum disorder, the three dimensions of narrative identity will have become disjointed structure, a focus on suffering, and detached narration. At this point. the mechanisms which produce and maintain dysfunctions in narrative identity for individuals with SSD may interact and reinforce one another in positive feedback loops. The full proposed developmental model of narrative identity in SSD is shown in Figure 3. This model generates several specific, testable hypotheses about the development of narrative identity deficits in SSD.

Figure 3.

Proposed developmental model of narrative identity in schizophrenia-spectrum disorders. Note that stages of illness and developmental stages (shown at the bottom of the figure) are approximate, e.g., imminent risk samples often include both adolescents and emerging adults.

Specific Hypotheses

Hypothesis 1.

For individuals who go on to develop a psychotic disorder, a focus on suffering will appear first, disjointed structure will appear second, and detached narration will appear last. This hypothesis could be tested by cross-sectional comparisons of narrative identity among different age groups or risk categories of psychosis risk, or by repeated measurements of narrative identity during the imminent risk period and the early course of psychosis.

Hypothesis 2.

Individuals who exhibit a greater focus on suffering early in life (e.g., through negative self-beliefs) will be predisposed to develop disjointed structure and detached narration later in life. This hypothesis could be tested longitudinally, by collecting data on negative self-beliefs from children or adolescents who exhibit trait vulnerability to psychosis and then following up to examine narrative identity later in life. Alternatively, this hypothesis could be tested through retrospective parent report of childhood self-beliefs, or through follow-back studies (see, e.g., Cupo et al., 2021) of related clinical constructs such as the negative self-evaluations captured in commonly used measures of depression.

Hypothesis 3.

Other self-concept variables, notably self-concept clarity, may help to explain narrative identity disturbances, e.g., disjointed structure will likely correlate with low self-concept clarity. This hypothesis could be tested by correlational analyses of narrative identity and other self-concept variables, especially self-concept clarity. Mediation models may show that self-concept variables such as self-concept clarity mediate associations between attenuated psychotic symptoms and disjointed structure in imminent risk groups or in the early phases of SSD.

Hypothesis 4.

If detached narration represents “giving up” trying to make sense of the world, then longitudinal trends in autobiographical reasoning should be curvilinear: for individuals who later develop SSD, autobiographical reasoning should emerge at a normative level in adolescence, then decrease before or around the time of a first psychotic episode. This hypothesis could be tested cross-sectionally by comparing different ages or illness stages, or longitudinally be collecting repeated narratives from the same individuals. Detached narration may also be associated with hopelessness and negative expectancies, which may be tested by correlations in adult schizophrenia samples.

Treatment Implications

A multidimensional, developmental understanding of narrative identity disturbances in the psychosis spectrum suggests several avenues for intervention. The mechanisms and targets identified in this review are well-suited to existing metacognitive and mentalization-based treatments such as Metacognitive Reflection and Insight Therapy (MERIT; Lysaker & Klion, 2017). Metacognitive treatment addresses fragmentation of the self (a state in which thoughts, emotions, and perceptions are experienced as scattered fragments rather than an integrated whole) by building more complex and integrated understandings of the self and others (Hamm et al., 2017; Lysaker et al., 2020). Metacognitive interventions could likely improve narrative structure by promoting narrative cohesion, elaboration of memories, and temporal sequencing; mitigate a focus on suffering by promoting agency and redemption; and improve autobiographical reasoning by promoting self-understanding and meaning-making. Thus, metacognitive interventions seem well-suited to addressing narrative identity deficits. A related treatment, narrative enhancement and cognitive therapy (NECT; Yanos et al., 2011), has also been shown to improve self-esteem and reduce self-stigma, potential indicators of a reduced focus on suffering.

Self-esteem or schema-focused treatments may also be helpful, particularly in the imminent risk period (C. D. J. Taylor et al., 2017; H. E. Taylor et al., 2014). These treatments could mitigate a focus on suffering in the life stories of young people, with the potential knock-on effect of making autobiographical reasoning less aversive. Mindfulness-based and third-wave cognitive therapies such as dialectical behavior therapy may also be helpful in improving self-concept clarity (Hanley & Garland, 2017; Roepke et al., 2011), helping individuals to develop normative narrative structure. Finally, trauma is a prominent factor linking self-concept clarity, core beliefs, and metacognition (Appiah-Kusi et al., 2017; Evans et al., 2015; Gaweda et al., 2019; Hamm et al., 2021; Leonhardt et al., 2015; Wigman et al., 2012). Furthermore, the process of recovering coherence is often painful as it may entail growing awareness of past trauma (e.g., Leonhardt et al., 2015), losses (K. D. Buck et al., 2013), and limitations (Conneely et al., 2020). Trauma-focused therapies, including trauma-focused metacognitive therapies, may be beneficial in addressing narrative identity disturbances. Future studies employing narrative identity variables as outcomes for metacognitive, schema-focused, mindfulness-based, and trauma-informed treatments would be highly informative.

Other Current Questions and Future Directions

Research in Imminent Risk and First Episode Psychosis Samples

Most of the reviewed studies focused on individuals in the chronic his phase of SSD. Findings in these populations have been relatively consistent across samples. Significant and nonsignificant effects have generally been in the same direction, and differences in statistical significance may reflect differences in statistical power rather than effect sizes. By contrast, evidence in milder or earlier forms of psychosis is weaker and less consistent. Of the 30 studies in this review, less than a quarter investigated non-chronic SSD samples, with one (3%) study on first episode psychosis and five (17%) studies on trait vulnerability. None of the reviewed studies included imminent risk samples (e.g., clinical high risk, at-risk mental state, attenuated psychosis syndrome). Research with imminent risk and first-episode psychosis samples can illuminate the developmental trajectories and pathogenetic mechanisms that may lead to broad narrative disruptions in SSD. The developmental model proposed above should be re-assessed in light of future cross-sectional or longitudinal research in samples spanning late adolescence and emerging adulthood.

Relationships with Symptoms, Medication, Neural Function, and the Minimal Self

Narrative identity processes are complex, likely recruiting many cognitive, affective, and neural mechanisms. Their relationships with symptoms have been tested in a somewhat ad hoc manner thus far. Some studies have tested relationships to broad symptom measures (e.g,. PANSS), some have tested relationships to specific symptoms, and some have not examined any relationships to symptoms. Future studies which include comprehensive symptom measures (e.g., PANSS symptom dimensions) in larger sample sizes would give more definitive evidence about links between symptoms and narrative processes.

Psychosis spectrum research has also uncovered many disturbances associated with the minimal self, including in self-referential processes, semantic self-concept, and associated neural mechanisms. However, it is currently unclear how these processes interact with narrative identity. Are some elements of narrative identity more closely linked to basic self-disturbances? Are specific elements of narrative identity associated with dysfunction in specific neural circuits? Do intermediate self-concept processes such as self-esteem or self-concept clarity mediate these relationships? Future research could address these questions by defining links between various levels of analysis of self-referential processing.

Antipsychotic medication may also affect narrative identity processes. Antipsychotic medication has been shown to affect function in the medial prefrontal cortex and posterior cingulate cortex, two key brain regions associated with self-referential processing (Wang et al., 2017). However, the effects of these changes are not well understood. Some evidence suggests that antipsychotic medication may facilitate narrative processes, while other evidence suggests it may impair narrative processes. Antipsychotic medication is associated with impairments in working memory (Eitan et al., 1992; Hutcheson et al., 2015; Reilly et al., 2007), which could make it more difficult to hold autobiographical details in mind during autobiographical reasoning. Furthermore, antipsychotic medication may damage the self-concept, with self-doubt, low self-esteem, and self-stigmatization contributing to poor subjective quality of life in individuals prescribed antipsychotic medication (Chou et al., 2014; Moritz et al., 2010; Rodak et al., 2018; Sajatovic & Jenkins, 2007). This process could sharpen a focus on suffering. By contrast, antipsychotic medication is associated with potential improvements in episodic memory (Eitan et al., 1992; Hutcheson et al., 2015; Reilly et al., 2007) and improvements in insight, which could facilitate autobiographical reasoning and narrative structure (Bianchini et al., 2014). Further research would be valuable to examine the potential effects of antipsychotic medication on narrative identity processes.

Comprehensive Assessment of Narrative Identity and Psychopathology

Very few studies so far have assessed all three narrative dimensions on ordinal scales. Future studies including comprehensive assessment of narrative identity on all three dimensions would be helpful to clarify the relative severity and clinical relevance of various narrative identity deficits. Ideally, studies which pair comprehensive assessments of narrative identity with comprehensive symptom assessments (e.g., PANSS symptom dimensions) would be well placed to define the relative importance of various narrative processes and relationships between symptoms and narrative identity. Similarly, another limitation of the reviewed literature is that studies tended to compare individuals with SSD to healthy controls. With a few exceptions (three studies with mood disorder control groups, Holm et al., 2020; Jensen et al., 2020, 2021, and two studies with HIV+ control groups, Moe & Docherty, 2014; Willits et al., 2018), most studies in this area have not been designed to distinguish between general psychopathological deficits vs. specific schizophrenia-spectrum deficits. We have interpreted findings in the psychosis spectrum in relation to findings from other populations, but this comes with the important caveat that these samples have generally not been directly compared to one another using the same methods. Studies comparing SSD to clinical comparison groups are important to test the specificity and relative magnitude of narrative disruptions in various forms of psychopathology.

Trained Raters vs. Other Approaches

Methodologically, the “gold standard” in narrative identity research is for trained raters to score interview transcripts from a common coding manual, assigning numerical ratings on an ordinal scale (Adler et al., 2017). A team of multiple raters (typically 2 or 3 raters) train to a high standard of interrater reliability (e.g., intraclass correlation > .75) by applying a coding manual to transcribed interviews. Coding manuals may have been validated in previous studies, such as the Adler coding manual for agency (Adler, 2012; Adler et al., 2008), in which raters assign scores from 0 to 4 for the extent of agentic language in a narrative. New coding manuals may also be developed as needed. Once raters achieve good interrater reliability, they may then divide up the remaining interview transcripts to code independently, or they may each code all the transcripts, with disagreements resolved by consensus or by taking their mean rating. For complete descriptions of this method, see tutorials by Adler et al. (2017) or Syed and Nelson (2015).

Obviously, this method requires a substantial investment of researchers’ time and effort. It has several notable advantages that often justify the time and effort. Chiefly, it allows multiple coders to demonstrate inter-rater reliability, which enhances results’ internal and external validity (Adler et al., 2017; Syed & Nelson, 2015). Relatedly, it is more statistically powerful than other approaches (Adler et al., 2017; McAdams & McLean, 2013). The reviewed literature supports this second point: significant results were most common when trained raters coded variables on ordinal scales, compared to other methods such as dichotomizing narratives (e.g., as positive/negative or specific/nonspecific) or comparing participants’ questionnaire results before and after an interview. While trained raters will not be appropriate in every study, this method maximizes validity and statistical power, which will often justify the additional research burden.

Participants’ self-ratings are an interesting case. It seems logical that a participant should be able to reflect on and rate his or her own narratives. However, this approach introduces confounding rater effects such as comprehension, motivation, and demand characteristics (Adler et al., 2017; Dunlop, Harake, et al., 2020). Moreover, psychosis-spectrum samples are heterogeneous, with multiple symptom and trait profiles typically present in any given sample (e.g., Cowan et al., 2020; Dickinson et al., 2018; Horan et al., 2008). Even more than in healthy populations, there is likely to be substantial variability in how individuals on the psychosis spectrum might understand and respond to questions about their own narratives. The current review found few statistically significant results from participants’ self-ratings, suggesting that self-ratings were a less statistically powerful approach than ratings scored by trained judges.

Emerging Narrative Identity Concepts

Several schizophrenia researchers have emphasized interpersonal processes in self-disturbances (e.g., Lysaker & Lysaker, 2010; B. Nelson, Parnas, et al., 2014). Likewise, narrative identity researchers have begun to attend to interpersonal factors in narrative identity. Researchers now examine interpersonal narratives, including vicarious narratives (stories we tell about other people) and relationship narratives (stories we tell about our relationships with other people) (Dunlop, Bühler, et al., 2020; Kirkegaard Thomsen et al., 2020). Another important emerging concept is the study of master narratives, cultural scripts for life stories which individuals fill in with details of their own lives (or resist by adopting alternative narratives) (McLean & Syed, 2015). No studies have yet examined cultural or interpersonal aspects of narrative identity in psychosis-spectrum samples, and concepts such as vicarious narratives, relationship narratives, and master narratives could potentially enrich this literature.

Conclusion

The life stories of individuals with SSD appear as disjointed collections of events, linked by a focus on suffering, but unintegrated due to detached narration. These narrative deficits are key aspects of a lifelong struggle to maintain a coherent sense of self and identity. Moreover, they may originate at different points along the psychosis spectrum, and we have proposed a developmental model of narrative identity to account for the emergence of disjointed structure, a focus on suffering, and detached narration in schizophrenia-spectrum disorders. Narrative identity provides many promising conceptual and empirical research avenues to understanding and ameliorate disturbances in the narrative self in the psychosis spectrum.

Highlights.

• Individuals in the psychosis spectrum experience lifelong challenges with identity

• These challenges manifest, in part, as disjointed, painful, detached life stories

• Difficult life stories can be understood as disturbances in narrative identity

• Narrative identity disturbances likely appear in adolescence and emerging adulthood

Appendix

Table A.1.

Reviewed studies

Reference	Sample	Narrative prompt/ data collection	Narrative analysis methods	Narrative variables
Schizophrenia-Spectrum Disorders
(Melissa C. Alle et al., 2015)	27 SCZ 26 healthy controls	Life story, plus 7 most important events	Trained raters: Habermas coherence coding system Self-ratings: event centrality	Structure: temporal coherence Autobiographical reasoning: thematic and causal coherence, event centrality
(M. C. Alle, Gandolphe, et al., 2016)	Study 1: 20 SCZ, 21 healthy controls Study 2: 30 SCZ, 28 healthy controls	Life story, plus 7 most important events	Trained raters: Habermas coherence coding system	Structure: temporal coherence Autobiographical reasoning: thematic and causal coherence
(M. C. Alle, d’Argembeau, et al., 2016)	27 SCZ 27 healthy controls	Life story, plus 7 most important events and 3 future events	Trained raters: self-event connections Self-ratings: vividness, valence, centrality	Motivational/affective: emotional tone Autobiographical reasoning: self-event connections
(Bennouna-Greene et al., 2012)	25 SCZ 25 healthy controls	24 events relating to 4 identity statements (i.e., “I am…”)	Trained raters: valence, specificity, active/passive memories, thematic link to identity Self-ratings: thematic link to identity	Structure: specificity Motivational/affective: emotional tone, agency Autobiographical reasoning: self-event connections
(Berna et al., 2011a)	24 SCZ 24 healthy controls	Self-defining memories	Trained raters: meaning-making, redemption, valence Self-ratings: subjective impact, personal relevance, recalled and current emotions (PANAS)	Motivational/affective: emotional tone, redemption Autobiographical reasoning: event centrality, meaning-making
(Berna et al., 2011b)	24 SCZ 24 healthy controls (same sample as Berna et al., 2011a)	Self-defining memories	Trained raters: meaning-making, redemption, emotional tone Self-ratings: subjective impact, personal relevance	Motivational/affective: emotional tone, redemption Autobiographical reasoning: event centrality, meaning-making
(Berna et al., 2017)	Study 1: 17 SSD (SCZ and SZA) Study 2: 83 undergraduates	Memories associated with PDI delusion items; positive and negative memories	Self-ratings: vividness, emotional valence, emotional intensity, event centrality	Motivational/affective: emotional tone Autobiographical reasoning: event centrality
(Gruber & Kring, 2008)	Study 1: 42 SCZ Study 2: 24 SCZ, 19 healthy controls	Memories associated with various emotions	Trained raters: tellability (grammar, elaboration), clarity, temporal sequencing	Structure: Detail, temporal coherence
(Holm et al., 2016)	25 SCZ 25 healthy controls	Life story chapters and self-defining memories	Self-ratings: valence, self-event connections, causal connections to other events, self-centrality, chronology	Structure: temporal coherence Motivational/affective: emotional tone Autobiographical reasoning: self-event connections, causal coherence, self-centrality
(Holm et al., 2017)	25 SCZ 25 healthy controls (same sample as Holm et al., 2016)	Self-defining memories	Trained raters: memories related to illness Self-ratings: age at event	Content of self-defining memories
(Holm et al., 2018)	24 SCZ 24 healthy controls (same sample as Holm et al., 2016)	Self-defining memories	Trained raters: temporal macrostructure, agency, communion, agency fulfilment, communion fulfilment	Structure: temporal coherence Motivational/affective: agency, communion
(Holm et al., 2020)	29 SSD (SCZ or SZA) 29 HIV+ adults	Indiana Psychiatric Illness Interview	Trained raters: agency, communion	Motivational/affective: agency, communion
(Jensen et al., 2020)	20 SCZ 20 depression 20 healthy controls	Past and future life chapters	Self-ratings: positive tone, negative tone, positive change, negative change	Motivational/affective: emotional tone Autobiographical reasoning: self-event connections
(Jensen et al., 2021)	20 SCZ 20 depression 20 healthy controls (same sample as Jensen et al., 2020)	Past and future life chapters	Trained raters: agency, communion	Motivational/affective: agency, communion
(Lysaker, Wickett, et al., 2005)	24 SSD (16 SCZ, 9 SZA) 12 mixed controls (4 depression, 8 legally blind)	Life story, divided into chapters	Trained raters: Scale to Assess Narrative Development	Motivational/affective: Agency, alienation (i.e., low communion)
(Lysaker, France, et al., 2005)	51 SSD (38 SCZ, 14 SZA)	Indiana Psychiatric Illness Interview	Trained raters: Narrative Coherence Rating Scale	Structure: richness of detail, temporal coherence
(Lysaker et al., 2006)	34 SSD (20 SCZ, 14 SZA)	Indiana Psychiatric Illness Interview	Trained raters: Scale to Assess Narrative Development	Motivational/affective: Agency, alienation (i.e., low communion)
(Lysaker et al., 2008)	68 SSD (41 SCZ, 29 SZA)	Indiana Psychiatric Illness Interview	Trained raters: Narrative Coherence Rating Scale	Structure: richness of detail, temporal coherence
(Lysaker, Ringer, et al., 2010)	88 SSD (69 SCZ, 34 SZA)	Indiana Psychiatric Illness Interview	Trained raters: Scale to Assess Narrative Development	Motivational/affective: agency, alienation (i.e., low communion)
(Moe & Docherty, 2014)	50 SCZ 17 Bipolar w psychotic features 24 healthy controls	Self-descriptions	Trained raters: negative-positive self-regard, striving/ambitious, level of self-definitions	Motivational/affective: emotional tone, agency
(Moe et al., 2016)	32 SCZ 15 healthy controls	Indiana Psychiatric Illness Interview	Trained raters: Scale to Assess Narrative Development Computerized: idea density	Structure: detail (idea density) Motivational/affective: Agency, alienation (i.e., low communion)
(Moe et al., 2018)	32 SCZ (same sample as Moe et al., 2016)	Indiana Psychiatric Illness Interview	Computerized: idea density	Structure: detail (idea density)
(Raffard et al., 2009)	20 SCZ 20 healthy controls	Self-defining memories	Trained raters: specificity, meaning-making Self-ratings: PANAS pre- and postinterview	Structure: specificity Motivational/affective: emotional tone Autobiographical reasoning: meaning-making
(Raffard et al., 2010)	81 SCZ 50 healthy controls	Self-defining memories	Trained raters: specificity, meaning-making, self-event connections, Narrative Coherence Coding scheme Self-ratings: PANAS pre- and postinterview	Structure: specificity, context coherence, temporal coherence Motivational/affective: emotional tone Autobiographical reasoning: self-event connections, meaning-making, thematic coherence
(Willits et al., 2018)	200 SSD (SCZ and SZA) 55 adults with HIV	Indiana Psychiatric Illness Interview	Computerized: deep semantic cohesion	Structure: detail/context coherence
First-Episode Psychosis
(Wright et al., 2019)	71 first-episode psychosis (provisional diagnosis, typically unspecified psychotic disorder) 57 healthy controls	Self-defining memories	Trained raters: specificity, integration (meaning-making), emotional valence	Structure: specificity Motivational/affective: emotional tone Autobiographical reasoning: meaning-making
Trait Vulnerability
(Berna, Göritz, et al., 2016)	49 high schizotypy 147 low schizotypy (scores on Community Assessment of Psychic Experiences)	Self-defining memories	Self-ratings: meaning-making, event centrality, emotional valence, emotional intensity, Thinking About Life Experiences scale	Motivational/affective: emotional tone Autobiographical reasoning: event centrality, meaning-making, exploratory processing
(Hallford & Burgat, 2014)	417 general population	Awareness of Narrative Identity Questionnaire	Self-ratings: temporal coherence, causal coherence, thematic coherence, Thinking About Life Experiences scale, Prodromal Questionnaire	Structure: context coherence Autobiographical reasoning: thematic and causal coherence, exploratory processing
(Hazan et al., 2019)	54 high schizotypy 41 low schizotypy (scores on Schizotypal Personality Questionnaire)	Turning point memories	Trained raters: specificity, valence, causal coherence, agency	Structure: specificity Motivational/affective: emotional tone, agency Autobiographical reasoning: causal coherence
(See et al., 2020)	1,542 nonclinical adolescents	Turning point memories	Trained raters: agency, redemption, self-event connections Self-ratings: schizotypal traits (Personality Inventory for DSM-5)	Motivational/affective: agency, redemption Autobiographical reasoning: self-event connections

Note: SSD = schizophrenia-spectrum disorders; SCZ = schizophrenia; SZA = schizoaffective disorder.