Neurocognitive consequences of adolescent sleep disruptions and their relationship to psychosis vulnerability: a longitudinal cohort study

The study investigates how adolescent sleep disruptions relate to psychosis vulnerability through neurocognitive mechanisms. Prior research shows adolescence is a critical period for maturation of cognitive functions such as inhibition and working memory, with deviations associated with psychopathologies including schizophrenia-spectrum disorders. Cognitive deficits are detectable before onset, peak at first episode, and often deteriorate over the chronic course, with response inhibition and working memory posited as core etiological factors in some models. Sleep disturbances are highly prevalent in psychosis and may causally influence positive and cognitive symptoms; psychotic-like experiences (PLEs) are common in youth and predict later psychosis and other psychiatric disorders. The authors hypothesize that poorer sleep in a given year would impair cognitive functions (especially inhibitory control and spatial working memory), and that worse cognitive performance would be associated with concurrent increases in PLE frequency. They further hypothesize that inhibitory control and spatial working memory mediate the relationship between sleep parameters and PLEs across adolescent development.

The paper reviews evidence linking cognitive deficits to psychosis risk, highlighting response inhibition and working memory as central domains. Patients with schizophrenia show impaired accuracy and reaction time on go/no-go tasks; inhibitory deficits have been proposed to directly contribute to hallucinations (e.g., Waters model). Working memory impairments are present in at-risk individuals and may predict conversion; delayed recall and perceptual reasoning also deteriorate with symptom severity and around first-episode psychosis, though studies disagree on which domains are earliest or most predictive. Sleep disturbances are common in psychosis (e.g., high rates of insomnia; longer sleep latency; reduced sleep efficiency) and, when untreated, exacerbate positive and cognitive symptoms and reduce quality of life. Adolescents at ultra-high risk show increased wake after sleep onset and reduced sleep efficiency; longitudinal work indicates sleep often precedes psychotic symptoms except possibly paranoia (bi-directional). Sleep disruptions are linked to a fourfold increased risk of PLEs, and PLEs themselves are associated with a 3.5-fold increased risk of conversion and broader psychopathology. Sleep is essential for higher-order cognition; sleep loss in childhood predicts poorer adolescent inhibition; adults with insomnia show deficits in working memory, episodic memory, and problem solving. Experimental sleep restriction studies suggest sleep loss causally impairs cognition; evidence indicates sleep disruptions hinder inhibition more than cognitive deficits cause sleep problems. These literatures motivate testing cognition as a mediator linking sleep to PLEs.

Design and cohort: Data are from CoVenture, a 5-year population-based longitudinal cohort (grades 7–11) from 31 Montreal high schools (data collection 2012–2018). Initial recruitment included 3801 students (about 51% female; mean age 12.8, SD 0.45). After data quality control, 3517 participants with complete sleep, cognitive, and PLE data for at least two consecutive time points were included; 262 were excluded for missing baseline or fewer than two consecutive assessments. A subset of 617 participants received brief (3-hour) CBT-based prevention sessions for substance misuse during the trial; the authors note blinding prevents assessing their effects on secondary outcomes (sleep, cognition, PLEs) but expect minimal impact due to small group size and lack of evidence of effects on these outcomes. Ethical approval was obtained; youth assented and parents consented. Measures: Sleep parameters were assessed via a 1-month retrospective questionnaire capturing sleep quality (4 very well to 1 very badly), sleep latency (4 <15 min to 1 >60 min), and wake after sleep onset frequency (4 none to 1 ≥3 times/week). A Composite Sleep Score (CSS) summed these items; lower scores reflect poorer sleep. PLEs over the past 12 months were measured using the PLEQ-C (nine items; 0–2 scale; Cronbach’s α 0.77–0.80 across waves), summed to a dimensional PLE score. Cognition was assessed across four domains: response inhibition via the Passive Avoidance Learning Paradigm (PALP), using number of commission errors as the primary outcome; spatial working memory via the Find the Phone task, using revisit errors as the outcome; perceptual reasoning via a nine-item subset of Cattell’s Culture Fair Intelligence Test, scored as number correct adjusted for time; delayed recall via the Wechsler Child Memory Scales Dot Location test, scored by ability to reproduce the first three trial sequences after a 30-minute delay. Analytical approach: Four separate multilevel models (one per mediator: PALP, SWM, CFT, CMS) tested the role of cognitive performance in the relationship between sleep parameters and PLEs. Models included random intercepts and slopes at the individual level, controlled for sex at the between-person level, and included time coded as wave (with analyses described as controlling for time squared in the abstract). Independent variables were person-mean centered to distinguish within-person from between-person effects. Models estimated (a) between-person associations (mean levels over 5 years), (b) within-person concurrent associations (deviations from individual means in the same year), and (c) within-person lagged associations (predicting next-year outcomes). Significant within-person concurrent/lagged associations informed tests of mediated paths. A Bayesian estimator handled missing data. Analyses used Mplus 8.3.

• Direct sleep–PLEs associations: Poorer CSS was associated with higher PLE frequency at the within-person concurrent level (B = −0.115, SD = 0.047, p = 0.005); no significant lagged within-person effect (B = −0.052, SD = 0.059, p = 0.215). Between-person association showed poorer sleep linked to higher PLEs (B = −1.063, SD = 0.122, p = 0.001).
• Inhibition (PALP): Poorer sleep (lower CSS) predicted more commission errors within-person, both concurrently (B = −0.011, SD = 0.004, p < 0.001) and with a 1-year lag (B = −0.013, SD = 0.005, p < 0.001). Commission errors were associated with PLEs within-person concurrently (B = −0.411, SD = 0.170, p = 0.005); between-person association was also significant (B = 1.186, SD = 0.469, p = 0.005). A mediated path indicated an indirect effect of sleep on PLEs via commission errors (Table 3: B = −0.005, SD = 0.003, p = 0.005; described as a lagged “L-shaped” path linking prior-year sleep to current inhibition and current PLEs).
• Spatial working memory (SWM): Within-person deficits were associated with more PLEs both concurrently (B = −0.046, SD = 0.018, p = 0.005) and at 1-year lag (B = −0.051, SD = 0.023, p = 0.010). However, SWM showed no significant within-person associations with sleep (concurrent B = −0.019, SD = 0.021, p = 0.177; lagged B = −0.004, SD = 0.030, p = 0.458). Between-person associations showed poorer sleep linked to poorer SWM (B = 0.103, SD = 0.040, p = 0.007) and SWM linked to PLEs (B = −0.084, SD = 0.040, p = 0.012).
• Perceptual reasoning (CFT) and delayed recall (CMS): No significant within-person concurrent or lagged associations with sleep or PLEs. Between-person: sleep associated with perceptual reasoning (B = −1.538, SD = 0.456, p < 0.001) and perceptual reasoning associated with PLEs (B = 0.022, SD = 0.007, p < 0.001). Delayed recall showed no significant associations with PLEs or sleep.
• Overall: Findings support that inhibitory control specifically mediates the sleep–PLE relationship, while working memory relates to PLEs independently of sleep at the within-person level. No within-person evidence for roles of perceptual reasoning or delayed recall.

The results address the hypotheses by demonstrating that sleep disruptions are linked to increased PLEs and that this association is specifically mediated by response inhibition rather than general cognitive impairments. Poor sleep was associated with both concurrent and lagged increases in inhibition errors, consistent with the idea that inadequate sleep hinders the developmental trajectory of inhibitory control during adolescence. Inhibition was concurrently associated with PLEs, aligning with models positing a causal role of inhibitory deficits in hallucinations and broader psychotic phenomena. The mediated, lagged path suggests that sleep problems can have lasting effects on psychosis risk via disruptions to inhibitory control. Working memory deficits were concurrently and prospectively associated with higher PLEs, supporting its role as a potential marker or contributor to psychosis development. However, the lack of within-person association between sleep and working memory implies that SWM–PLE links may reflect shared vulnerabilities (genetic or environmental) or compensatory neural mechanisms maintaining SWM performance despite sleep loss in adolescents. The absence of within-person links involving delayed recall and perceptual reasoning suggests these domains are less directly implicated in the short-term interplay between sleep and PLEs during adolescence, though between-person associations indicate common vulnerabilities across individuals. Collectively, the findings highlight inhibitory control as a plausible mechanistic pathway by which sleep disruptions elevate psychosis risk and underscore potential intervention targets (sleep improvement, inhibitory control training) for prevention.

Inhibition mediates the relationship between sleep disruptions and increases in psychotic-like experiences (PLEs) in adolescents. Working memory deficits show concurrent and lagged associations with increased PLEs but are not linked to sleep at the within-person level. Sleep disruptions exhibit concurrent and lagged associations with PLEs, with the lagged impact operating indirectly via disrupted response inhibition. These results suggest that targeting sleep quality and inhibitory control may enhance psychosis prevention strategies and inform adolescent mental health interventions.

• Sleep assessment relied on a brief, 1-month retrospective composite score without objective measures (e.g., actigraphy) or comprehensive instruments (e.g., PSQI, Consensus Sleep Diary), limiting granularity on specific sleep dimensions and phases.
• The 1-month sleep window per year may not represent annual sleep patterns; longer or chronic sleep disruptions might have broader, more enduring cognitive effects.
• Temporal mismatch between sleep (1-month retrospect) and PLEs (12-month retrospect) complicates precise interpretation of lagged effects, though mediation timing partially mitigates this.
• Potential confounding by general psychopathology, as PLEs and cognitive dysfunctions relate to multiple psychiatric conditions; findings may reflect broader mental health processes.
• A subset of participants received brief CBT-based preventive sessions for substance use; although unlikely to bias outcomes, blinding prevents definitive evaluation of their effects on sleep, cognition, or PLEs.
• Between-person associations may reflect shared unobserved vulnerabilities (e.g., substance use such as cannabis) rather than causal pathways.