Self-defining Memories Predict Engagement in Structured Activity in First Episode Psychosis, Independent of Neurocognition and Metacognition

Abstract

Background

Self-defining memories (SDMs) are vivid personal memories, related to narrative identity. Individuals with schizophrenia report less specific, more negative, and extract less meaning from these memories compared to control groups. SDMs have been shown to be predicted by neurocognition, associated with metacognition, and linked to goal outcomes in healthy controls. As neurocognition and metacognition are known predictors of poor functioning in psychosis, SDMs may also be a predictor. No study has assessed the relationship to functioning or pattern of SDMs in first episode psychosis (FEP).

Methods

This was a cross-sectional study involving 71 individuals with FEP and 57 healthy controls who completed an SDM questionnaire. FEP participants completed measures of neurocognition, metacognition (Metacognitive Assessment Interview), functional capacity (The University of California, San Diego [UCSD] Performance-Based Skills Assessment), and functional outcome (Time-Use Survey).

Results

SDMs reported by individuals with FEP were less integrated compared to healthy controls. Within the FEP sample, holding less specific memories was associated with engagement in significantly fewer hours of structured activity per week and specificity of SDMs mediated the relationship between neurocognition and functional outcome, independent of metacognition.

Conclusion

This is the first study to assess SDMs in FEP and to explore the important role of SDMs on clinical outcomes, compared to healthy controls. This study suggests that elaborating on specific SDMs is a valid therapeutic target and may be considered a tool to improve daily functioning in FEP.

Introduction

Self-defining memories (SDMs) are vivid, intense, and well-rehearsed personal memories,¹ related to narrative identity and “ingredients” for the life story.² Prior research has considered 4 dimensions of interest for SDM³: (1) specificity—the ability to provide a clear and detailed memory, (2) integration—capacity to learn from, and incorporate, the memory into self-knowledge, (3) type of event—linked to a general theme (eg, achievements) or unresolved conflict (eg, mental health crises),¹ and (4) content valence—the strength of affective response when recalling the SDM.

Self-defining, or autobiographical, memories reported by individuals with schizophrenia have been found to be less specific,^4–6 more negative,⁷ and individuals extract less meaning from these types of memories compared to control groups,⁷ despite cues.⁸

In terms of the separate dimensions of SDMs, research on autobiographical memories (ATMs) highlights that specific ATM can be impacted by neurocognitive deficits in psychosis.^9,10 They may also be linked to negative symptoms in schizophrenia,¹¹ which are related to avoidance of trauma memories, hence, a lower likelihood of reporting specific memories. Specific memories are suggested to be associated with impairment in executive control and functional avoidance,¹² and influence goal outcomes in control participants^1,3 as individuals who report specific memories are better able to use appropriate cognitive-affective information to achieve their goals.¹ Conway and Pleydell-Pearce’s Self-Memory System¹³ suggested that ATM contain knowledge at 3 hierarchical levels of specificity: lifetime period, general events, and event-specific knowledge, which make up the hierarchy of ATM structures. These knowledge structures are joined with the working self, which enables an individual to draw on their memory to achieve goals. Within psychosis, Mehl et al¹⁴ demonstrated that ATM specificity predicted social performance, involving role-play tasks, over neurocognition and symptoms. This current study aimed to understand the role of specific SDMs in predicting functional outcome, associated with mental health recovery.

Second, in terms of integration, individuals with psychosis extract less meaning or learn fewer lessons from the self-defining events they report, compared to healthy controls.^4,8,15,16 Greater integration is associated with greater optimism and attainment of goals within healthy controls,^1,3 and may be associated with neurocognition.¹³

Finally, SDM for individuals with psychosis tend to be more negative in content,¹⁷ and focused on illness.^7,15,18 This may be linked to lower self-esteem and negative outlook.^3,7 However, unlike specificity and integration, content valence has not previously been associated with neurocognition nor outcomes. While Raffard et al⁷ demonstrated that memories reported by individuals with psychosis tend to be focused on hospitalization/illness, this study was conducted in an inpatient unit that may influence the memory recalled, due to contextual cues.^19,20 This hospital-related contextual cue, coupled with known memory difficulties in schizophrenia,^21–23 may have biased the individual to report a memory focused on hospitalization/illness. Another study demonstrated illness-related SDMs in a group of outpatients; however, this group included long-term schizophrenia patients,⁸ who may have integrated their illness within their self, compared to an early psychosis sample. This current study aimed to investigate the pattern of SDMs in first episode psychosis (FEP) and healthy controls.

While studies have demonstrated differences in SDMs between individuals with psychosis and healthy controls, and studies in healthy controls show these memories may predict goal outcomes, no study has assessed the impact of SDMs on outcome in psychosis. Functional outcome is a measurable aspect of an individual’s activities of daily living. This has been measured using the Time-Use Survey,^24,25 which captures the number of hours in structured activity per week. Time spent in structured activity is on average 63.5 hours in a healthy population, 25.2 hours in an FEP sample, and 19.7 hours in a delayed recovery group.²⁶ There is clear interest in the identification of those at risk of poor functioning, to target interventions to reduce this disability.

This study will combine 3 theoretical frameworks of (1) cognitive and neurocognitive underpinnings of functional outcome in psychosis, (2) metacognition as a mediator of functional outcome, and (3) sense of self in psychosis. These theories will be explored in turn, to develop the rationale for the hypothesis that SDMs and metacognition may impact on functional outcome.

First, models of functional outcome in psychosis suggest neurocognition, functional capacity, and negative symptoms influence functional outcome.^27–29 However, the picture is complex as cognitions and negative symptoms are shown to have a synergistic interaction that impacts functioning,³⁰ and the path between neurocognition and functioning has been shown to be mediated by functional capacity and cognitive processes.^29–31 Second, these cognitive processes include defeatist performance beliefs and self-stigma,³² and, recently also metacognition,³³ termed “thinking about thinking,”|³⁴ or the way one thinks about one’s experience.³⁵ Metacognition partly mediates the link between neurocognition and functional capacity, and fully mediates between functional capacity and functional outcome³¹ independently from symptoms.^36–38

Finally, SDMs are most relevant to narrative identity² and metacognitive ability has been associated with forming complex ideas of one’s life as a narrative across a lifetime.^10,39,40 It may be that SDMs overlap with metacognitive ability, but involve a distinct reflective process, focusing on one memory, which, when compromised, may impact on functioning. To support this, SDMs, like metacognition, are proposed to use cognitive information to help goal outcomes.¹ Negative content and poor integration might impact on optimism toward reaching a goal, and poor specificity might limit the detail available regarding actions to reach goals. Reflection on the self is shown to impact goal performance,⁴¹ which in turn impact on motivation,⁴² hope,⁴³ and functional outcome. On the basis of the Beck and Rector functional outcome model²⁷ and literature within SDMs, cognitive processes could extend to SDMs and have an independent role on functioning, alongside metacognition.

SDMs may contain different levels of specificity,¹³ which are integrated into the sense of self. These SDMs may be used by the individual, drawing on cognitive and affective information about the self, to engage in functional activities. Following the earlier research , it is hypothesized that SDM will be less specific, less integrated, and more negative in FEP compared to healthy controls. SDM (specificity and integration) may be associated with neurocognition and metacognition. Finally, SDMs might contribute to difficulties in functioning in FEP. This is the first study to assess the role of SDMs in the relationship between neurocognition and functional outcome, independent of metacognition, in FEP.

Methods

Participants

Seventy-one young people with FEP were recruited, via a convenience sample, from outpatient Early Intervention in Psychosis services in United Kingdom. All had been given a formal diagnosis of FEP by a psychiatrist. Participants with a primary diagnosis of substance misuse disorder or organic neurological impairment were excluded.

Fifty-seven healthy control participants were matched on age, gender, and education to the earlier psychosis sample (see table 1 with difference statistics). Participants were recruited through advertisement within the local community. Participants with current mental health problems or history of psychosis were excluded following screening questions.

Table 1.

Sample Characteristics Summary Table

	FEP	Healthy Control	Differences Test
Age, years (SD)	25.93 (5.55)	24.84 (6.34)	F(1, 125) = 36.78, P = .31
Gender (Male/Female)	44/27	41/16	Ӽ2 (1, 128) = 1.41, P = .24
Symptoms (positive)	11.77 (3.46), range 7–19
Symptoms (negative)	13.21 (4.85), range 7–36
Symptoms (general)	27.94 (6.49), range 16–43
Education (years)	12.8 (1.7), range 11–17 years	13.37 (1.58)a, range 11–17 years	F(1, 127) = 2.86, P = .09

Note: FEP, first episode psychosis; SD, standard deviation.

^aData for healthy controls were captured as categories (eg, General Certificate of Secondary Education [GCSE], A-level, degree, higher degree), which was subsequently converted into years of education to match the FEP group.

Measures

Self-defining Memories.

SDM questionnaire⁴⁴ asked the participant to provide 3 descriptions of SDM. The participant was asked to provide a memory that was at least 1-year-old, remembered very clearly, important to the individual, one that helped the individual to understand themselves as a person, leading to strong feelings, and familiar like a picture or a song.⁴⁵ The participant had to provide a title, age at the time of the event, and a description of the event.

All scripts were coded by the first author through consultation with the classification system and scoring manual of self-defining ATMs.⁴⁵,⁴⁶ This manual was previously shown to have inter-rater reliability (Cohen’s kappa (κ) = 0.54–0.98) for students⁴³ and clinical groups.^7,15

Only the first memory was coded into specificity (nonspecific or specific), integration (integrated or nonintegrated), type of event, and content valence (positive or negative). Details are in supplementary materials. A second independent rater, blind to the scope of the study, coded responses for 12% of the total scripts (15 scripts). Reliability between the 2 raters was good (specificity, integration, and content valence; Cohen’s κ coefficient was .84, P < .001, and for type of event, Cohen’s κ was .83, P < .001).

Neurocognition.

Participants completed a battery of neurocognitive measures, including executive function (verbal fluency⁴⁷ and trail-making task⁴⁸), memory (logical memory and letter-number sequencing subscales [Wechsler Memory Scale-III]⁴⁹), and IQ (vocabulary and matrix reasoning tasks⁵⁰). All raw scores were converted into z scores using age-scaled population means and standard deviations.

Metacognition.

The Metacognitive Assessment Inter view⁵¹ requires the participant to reflect on a recent difficult interpersonal experience and asked a series of questions to assess (1) monitoring, identification of feelings, and thoughts; (2) differentiation, distinguishing between dreams, beliefs, or assumptions; (3) integration, reflection on different mental states, and rules governing them; and (4) decentralization, describing the mental state of the other, which is independent of their own view. These 4 subscales are each scored between 0 and 5, depending on spontaneity, use of aids/prompts, and the sophistication of the answer. The scores are averaged to provide one multidimensional construct. This measure has demonstrated good inter-rater reliability and internal consistency (α = .90 for total metacognition), factorial validity, and reliability (r = .62–.90).⁵¹

Measures of Functioning.

Functional Outcome

Time Use Survey (adapted from Short⁵²) provides an objective measure for hours spent engaging in structured activity per week.²⁴ This measure has been used within schizophrenia⁵³ and FEP sample,²⁶ has good inter-rater reliability⁵⁴ (intraclass correlation coefficient = 0.99),²⁶ and good validity at different stages of psychosis.^26,55

Functional Capacity

The UCSD Performance-Based Skills Assessment⁵⁶ (UPSA) provides a total score for real-life performance skills based on simulated tasks. This measure demonstrates high internal consistency (α = .88), good validity with other scales (direct assessment of functional status [DAFS] r = .86), and good test–retest reliability (r = .91).^57,58

Symptoms.

The Positive and Negative Syndrome Scale⁵⁹ (PANSS; clinical participants only) was included, a standardized instrument for assessing symptom severity in psychosis. This measure has good internal consistency (r = .69–.94), construct validity (r = .77), and inter-rater reliability (0.83–0.87).^59,60

Design

This study involved a cross-sectional design, with measures assessing neurocognition, metacognition, SDM, and functional outcome in FEP. Additional measures can be reviewed in the article by Davies et al.³³ Data from the SDM measure were compared between participants with FEP and healthy control participants.

Procedure

Ethical approval was obtained from London-Camden and Islington NHS Research and Ethics Committee (Ref:11/LO/1877). All participants provided informed consent.

Analysis Plan

Missing data were considered as “Missing at random”. For regression analyses, list-wise deletion was used, as recommended.⁶¹ For mediation analysis, full information maximum likelihood was used, which combines available information to estimate population parameters.^62,63

Hypothesis Testing.

Chi-squared analyses assessed differences in memory reported for specificity, integration, and content valence between individuals with FEP or healthy controls.

Logistic regression analyses assessed whether neurocognitive and metacognitive ability were associated with likelihood of reporting a specific or integrated SDM. Linear regression analyses assessed whether specificity and integration of SDM predicted functional outcome, controlling for neurocognition and metacognition. Finally, a mediation model was developed to assess whether specificity and integration of SDM mediate the relationship between neurocognition and functional outcome, independent of metacognition. Due to sample size, the model was built through sequential steps: (1) neurocognition to functional outcome with metacognition as a single mediator, (2) with SDM as single mediator, and (3) with all significant mediators.

Results

Sample Characteristics

A total of 71 participants with FEP completed the assessments (mean age = 25.93, SD = 5.55, range 18–39). Fifty-seven healthy control participants completed the SDM measure (mean age = 24.84, SD = 6.34, range 18–39).

Data Checking

All variables were checked for skewness, kurtosis, and outliers. UPSA total was positively skewed and, therefore, transformed using square root transformation.

Frequency and Descriptive Statistics

Sixty percent of FEP and 89% of healthy controls provided 3 SDMs. Due to the limited number who provided all 3 memories, only the first SDM was coded.

Frequency statistics for SDMs are presented in table 2 and supplementary material 2.

Table 2.

The Frequency Statistics for Self-defining Memories

	FEP sample		Healthy Control Sample		Difference Tests
Specific vs nonspecifica	66%	34%	79%	21%	Ӽ2 (1, N = 128) = 2.54, P =.11
Integrated vs nonintegratedb	18%	82%	58%	42%	Ӽ2 (1, N = 128) = 21.52, P <.001
Positive vs negative content	52%	48%	65%	35%	Ӽ2 (1, N = 128) = 2.12, P =.15
Words per first memory, Mean	58.97 (SD = 47.5, median = 39, range 4–202 words)		119.04 (SD = 92.5, median = 98, range 21–491 words)		t(79.38)= −4.45, P < .001
Type of event
Recreation/exploration	22% (87% positive)		33% (46% positive)
Relationship	22% (6% positive)		28% (24% positive)
Achievement/mastery	24% (100% positive)		16% (24% positive)
Guilt/shame	0%		5% (100% negative)
Drug, alcohol, or tobacco use	0%		2% (100% positive)
Hospitalization/stigmatization of illness	6% (100% negative)		0%
Failure	6% (100% negative)		2% (100% negative)
Life threatening event	20% (7% positive)		12% (100% negative)
Event not classifiable	0%		2% (100% positive)

Note: FEP, first episode psychosis; SD, standard deviation.

^a33% of self-defining memories (SDMs) were coded as specific positive in FEP group, compared to 47% in control group.

^b9% of SDMs were coded as integrated positive in FEP group, compared to 40% in control group.

Descriptive statistics for neurocognitive, metacognitive, and outcome variables are presented in table 3.

Table 3.

Descriptive Statistics for Neurocognition, Metacognition, and Functioning

Cognitive/Functioning Measure	Raw Scores		z Scores (Created from Age-Scaled Scores)
	Mean (SD)	Range	Mean (SD)	Range
Immediate verbal memory (0–75)	28.76 (10.93)	10 to 55	−1.25 (1.13)	−3.16 to 1.3
Delayed verbal memory (0–50)	16.76 (8.34)	0 to 35	−1.21 (1.13)	−3.33 to 1.11
Letter-number sequence (0–21)	8.83 (2.47)	4 to 15	−.99 (.96)	−2.78 to 1.18
Verbal fluency (semantic)	18.85 (4.59)	9 to 29	−.34 (.87)	−2.18 to 1.42
Verbal fluency (phonemic)	33.09 (9.4)	15 to 55	−.87 (.97)	−2.85 to 1.5
Vocabulary (0–80)	53.63 (10.82)	31 to 73	−.28 (1.16)	−2.6 to 1.8
Matrix reasoning (0–35)	26.03 (4.16)	13 to 34	.2 (8.5)	−2.5 to 2
Trail-making test (B-A)	46.93 (31.24)	6.64 to 135.60	.97 (2.45)	−3.88 to 8.17
Metacognitive Assessment Interview (MAI) total (0–5)	2.85 (1.2)	.44 to 4.88
UPSA total (0–100)	72.98 (14.5)	36.62 to 95.24
Time-Use Survey (hours in activity per week)	33.97 (26.57)	2.30 to 96.74

Note: FEP, first episode psychosis; SD, standard deviation; UPSA = UCSD performance–based skills assessment.

Hypothesis Testing

Hypothesis 1.

A significant difference was found between groups on frequency of integrated memory reported, Χ² (1, N = 128) = 21.52, P < .001. Table 2 highlighted 13 out of 71 individuals with FEP reported an integrated memory, compared to 33 out of 57 healthy control participants (See figure 1).

Fig. 1.

Bar graph for percentage of reported integrated self-defining memories in first episode psychosis (FEP) and healthy control sample. For color, please see the figure online.

No difference was found for frequency of specificity (P = .11) and content valence (P = .15) between the groups.

Hypothesis 2.

To understand the relationship between SDMs and neurocognition, a single neurocognitive factor was created using the z-scores of all neurocognitive variables, following research that assumes a single neurocognitive factor.^33,64–67

Logistic regression analysis assessed whether neurocognition could determine the likelihood of SDM to be specific or nonspecific. This model was significant (Ӽ² = 8.0, df = 1, P = .005). Neurocognition explained 14.8% (Nagelkerke R²) of the variance in specificity and correctly classified 69% of the cases. Neurocognition did not predict integration (P = .28).

Hypothesis 3.

Logistic regression analysis assessed whe ther metacognitive ability could determine the likelihood of SDM to be specific or nonspecific. The model was significant (Ӽ² = 16.16, df = 1, P < .001). Metacognition explained 28.7% (Nagelkerke R²) of the variance in specificity and correctly classified 75.7% of the cases. Metacognition did not predict integration (P > .2).

Hypothesis 4.

Specificity was a significant predictor of functional outcome, while controlling for metacognition (neurocognition was not significant after including metacognition). This model predicted 70.4% (adjusted r² = .70) of the variance in functional outcome score (R²=.70, F(2, 68) = 78.67, P < .001); specificity predicted 1.8% of this variance and improved the baseline model (ΔR²= .02, F(1, 66) = 4.08, P = .047). Individuals who reported a specific SDM had a mean time-use score of 43.3 (SD 3.92) hours within structured activity per week, compared to those with nonspecific SDM, mean of 14.92 (SD 2.44) hours.

When including depression as a covariate, for 21 participants with individualized PANSS scores, depression did not predict functioning (P > .05) and specificity was still a significant predictor of functioning (ΔR² = .38, P = .003).

Integration did not predict functional outcome. Functional capacity was not predicted by any SDM variable.

Mediation Model

The mediation was conducted using Mplus with Multiple Mediation Model (structural equation modeling) using maximum likelihood estimation, bootstrapping, and corrected confidence intervals, following Preacher and Hayes⁶⁸ causal steps of mediation.

A series of mediation models were conducted to identify the indirect mediating effect of specificity of SDM between neurocognition and functional outcome, independent of metacognition. A full multiple mediation model is presented.

We aimed to confirm a single neurocognitive factor solution using a confirmatory factor analysis of neurocognition z scores. However, a confirmatory factor analysis (CFA) demonstrated that a 1-factor solution for neurocognition was not a good fit to the data [χ²(20) = 79.5, P = .00, comparative fit index (CFI) = .75, Tucker Lewis index (TLI) = .65, root mean square error of approximation (RMSEA) = 0.21]. Instead, neurocognition was a 2-factor solution containing factor 1 representing memory: immediate and delayed logical memory and factor 2 representing “other” neurocognition: letter-number sequence, executive functioning, and verbal and performance IQ. The model demonstrated an excellent fit [χ²(19) = 18.92, P = .46, CFI = 1.0, TLI = 1.0, RMSEA = 0.00]. From this point forward, all analyses are conducted first with the memory neurocognitive factor then the “other” neurocognitive factor.

First, the mediating effect of metacognition on the relationship between memory and functional outcome was tested. Significant direct pathways were found between memory and metacognition (β = .62, P < .001) and metacognition and functional outcome (β = .78, P < .001). Metacognition significantly and fully mediated the relationship between memory and functional outcome (β = .48, P < .001, ±95% CI [0.36, 0.6]).

Second, the mediating effect of SDMs was tested. A significant direct pathways were found between memory and functional outcome (β = .31, P = .01), memory and specificity of SDM (β = .41, P = .013), and specificity of SDM and functional outcome (β = .61, P < .001). Specificity significantly and partially mediated the relationship between memory and functional capacity (β = .25, P = .021, ±95% CI [0.04, 0.46]).

Finally, a full multiple mediation model was conducted with mediating effect of metacognition and SDMs on the relationship between memory and functional outcome. A significant direct pathway was found between memory and metacognition (β = .62, P < .001) and specificity of SDM (β = .41, P = .013). A significant direct pathway was found between metacognition and functional outcome (β = .58, P < .01), and specificity of SDM and functional outcome (β =.4, P < .001). Metacognition significantly mediated the relationship between memory and functional outcome (β = .36, P < .001, ±95% CI [0.22, 0.5]) and specificity of SDM also significantly mediated the relationship between memory and functional outcome (β = .16, P = .047, ±95% CI [0.02, 0.32]). The direct pathway was nonsignificant, suggesting a full mediation model (See figure 2).

Fig. 2.

Mediation of the effect of neurocognition to functional outcome through two covarying mediators: specificity of self-defining memories and metacognition. ***P < .001, **P < .01, *P < .05. For color, please see the figure online.

Factor 2: “other” neurocognitive factor significantly predicted functional outcome, β = .47, P < .001. However, this factor did not predict specificity and, therefore, not included in the model.

Discussion

This was the first study to demonstrate that individuals with FEP displayed different patterns of SDMs compared to healthy control participants. Those with FEP were less likely to report integrated SDMs, compared to controls. This supports research in chronic schizophrenia cohorts,^8,15 but demonstrates that deficits exist at first episode rather than as a result of chronic illness. Integration may enable the individual to interpret events as meaningful to themselves and define who they are as a person. This may be disrupted in psychosis, as outlined in the “disrupted self” framework.^69,70. Berna et al⁸ demonstrated that individuals with schizophrenia report fewer integrated memories and more trauma-related memories. It may be suggested that trauma memories are not integrated into the self, to avoid continued distress, but consequentially leave a fractured sense of self.

Although nonsignificant, individuals with psychosis reported less specific memories and more negative memories, which focused on (1) negative relationships, (2) trauma, (3) failure, and (4) illness. This is aligned with research that suggests those with psychosis have poorer social relationships^71,72 and more interpersonal, trauma memories.^73–76 The lack of significant difference between the groups may be because these memories may be less prominent in the early stages of psychosis.

A small proportion of participants reported SDMs related to hospitalization/illness, in contrast to Raffard et al.⁷ This may have been triggered by the hospital contextual cues in the study by Raffard et al, while the present study was conducted in a community setting. Alternatively, this FEP group may not have integrated the illness into their identities, compared to a chronic schizophrenia group.

Memory specificity was significantly associated with functional outcome in FEP, independent of neurocognition and metacognition. This supports functional outcome models, which suggest neurocognition^27,28,31,77 and metacognition³³ play an important role in functioning in psychosis, but demonstrates a role of a distinct, reflective process of reporting SDMs on functional outcome. Individuals with FEP who report a specific SDM spent 43.3 hours within structured activity per week, compared to those who reported a nonspecific SDM who were engaged in 14.9 hours. In comparison, Hodgekins et al²⁶ demonstrated that individuals with FEP spent 25.2 hours in structured activity compared to 19.7 hours for a delayed recovery group. The differences reflect important differences in recovery trajectories.

Integration and valence are important aspects of SDMs, and the fewer integrated memories in FEP are an important finding, but these aspects of SDMs did not predict functional outcome in FEP. This may be a power issue, due to the reduced number of integrated memories in FEP, or integration may be more related to trauma, and therefore symptoms.⁷⁸ Blagov and Singer³ and Singer et al⁷⁹ demonstrated a negative correlation between specificity and integration. However, Blagov and Singer³ explicitly requested important memories to one’s life, which may have encouraged a focus on integration, at the expense of specificity, and Singer et al⁷⁹ demonstrated no such relationship in older participants, due to the greater ability of older adults to provide both integrated and specific SDMs. Following Conway and Pleydell-Pearce,¹³ specificity needs to be present within event, general, and lifetime memories in order to describe how the memory was integrated. An individual may need to have a certain level of specificity in SDM, in order to integrate this memory to influence functioning.

Specificity, or the ability to report a detailed SDM, may have enabled the individual to reflect on their previous experiences in a coherent manner, to identify important memories to the self and identity. This identity may allow the individual to view themselves as a person with skills and draw specific detail into their SDMs to guide function. This ability may allow them to accurately reflect on their ability and monitor errors, which facilitates engagement in activities. This furthers research within a healthy sample.^1,3

The mean word count for the FEP group was lower than for the healthy control group and lower than that reported by Raffard et al.⁷ The word count was greater than that reported by Jobson and O’Kearney⁸⁰ and similar to Singer and Moffitt.⁴⁴ This FEP sample may have written fewer words due to the lack of specific memories, supported by a previous positive association between word count and specificity,⁷ although it was possible to have memories that were brief (15 words) and specific. Alternatively, the lower word count may have been due to data collection (asking participants to write the memory down) or lack of motivation, previously been linked to functioning.^81,82 Future studies should explore these hypotheses.

The impact of specificity on functioning was independent of metacognition, thus highlighting a novel contribution of SDMs. Metacognition significantly predicted the likelihood of reporting a specific memory, which suggests metacognition is associated with difficulties in recalling ATMs and organizing one’s experience into a coherent narrative.³⁹

Neurocognition also significantly predicted the likelihood of reporting a specific SDM, supporting previous research^9,10; however, both metacognition and SDMs have an independent role in predicting functioning. Theoretically, it is expected that SDMs are particularly pertinent to functioning in psychosis. They are associated with goal outcomes,¹ and the typical age of reported SDM is 20–24 years in controls, compared to 15–19 years in a schizophrenia sample,⁷ which is before the onset of psychosis,⁸³ highlighting the interest in understanding the connection between psychosis, SDMs, and functioning. However, this might apply to ATMs more broadly, not just SDMs. Future studies could include a control condition that asks participants to provide an ATM which is not self-defining.

In terms of clinical implications, Lysaker and Klion⁸⁴ recently outlined the Metacognition Reflection and Insight Therapy, which is specifically aimed at improving metacognition.^85,86 In addition, narrative enhancement and cognitive therapy aim to construct positive narratives about the self.⁸⁷ Given the partial mediation effect of SDMs between neurocognition and functioning, therapies should focus on improving both metacognitive and neurocognition, eg, cognitive remediation, shown to improve both neurocognition and real-life skills.⁸⁸

Limitations

First, the sample was small, particularly for analysis of binary variables, as larger samples are needed for complex mediation models.⁸⁹ Due to the small sample size, we used a series of single mediation models to build into the larger multiple mediation model. Future studies should aim to replicate this finding in a sufficiently powered multiple mediation model. Second, in a subsample, the results remained after controlling for depression. However, as depression is shown to influence functional outcome in schizophrenia^90,91 and specificity of ATMs,⁹² a follow-up study should consider this further in order to replicate this finding.

Third, the SDM questionnaire does not explicitly state that the memory description should explain why this memory is meaningful, characteristic of an integrated memory. This lack of instruction may have influenced integration in their reporting. Future studies should include a spontaneous and cued integration response, akin to Berna et al.⁸ Using the Metacognitive Assessment Scale-Abbreviated (MAS-A),⁹³ assessing one’s acknowledgement of distress and management of difficulties, correlated with social functioning,⁹⁴ may provide different outcomes to Metacognitive Assessment Interview (MAI). Future studies could conduct a sensitivity analysis to replicate and build on these findings using the MAS-A. Due to power, this study was unable to fully separate the variance explained by neurocognition, metacognition, and SDMs. Future studies should explore whether SDM is a distinct factor, or a proxy for neurocognition.

Conclusion

This study was the first study to describe SDMs in FEP and assess the impact on functional outcome. Specificity of SDMs predicted functioning in FEP, independent of metacognition. Individuals who reported a specific SDM were more likely to utilize their real-life functional skills to partake in structured activities. In terms of clinical importance, elaborating on specific SDMs within therapeutic contexts may be useful, and future intervention strategies should explore SDMs as a tool to improve functioning.

Supplementary Material

Supplementary data are available at Schizophrenia Bulletin online.

Funding

Sussex Partnership NHS Foundation Trust and Economic Social Research Council through a PhD studentship (ES/J500173/1 to Ms. Wright).