Characterising the inconsistency between perceived and actual sleep and its impact upon cognition and mood Authors: Hannah Windmill, Alexander Dyer, Alastair D Smith, Matthew E Roser, Ashwin Dhanda, Stephen D Hall Published: 2024 ABSTRACT AbstractSleep profoundly influences human behaviour across cognition, affect, and daily experience. This study evaluated how subjective reports and objective measures of sleep capture the interaction between sleep quality and quantity on cognition and affect. We collected subjective sleep reports using the Pittsburgh Sleep Quality Index and objective seven-day actigraphy recordings from 83 participants. A test battery, including the Stop Signal, Digit Span, and Emotional Bias Tasks, measured response inhibition, working memory, and affect. Mood was evaluated using the Positive and Negative Affect Schedule, State-Trait Anxiety Inventory, and Beck’s Anxiety Inventory. We reveal that subjective sleep reports do not predict objectively measured sleep duration or quality. We demonstrate that objective measures predict cognitive performance on executive function and memory tasks for the upper and lower sleep quartiles, while subjective reports do not. Moreover, we demonstrate that subjective reports, but not objective measures, are strongly predicted by emotional state. These findings suggest that while subjective measures do not accurately index actual sleep, they are a reliable proxy for emotional well-being. We propose that combined subjective and objective measures are required to characterise the multifaceted interaction between sleep, cognition and emotion. We discuss the implications of these findings for understanding the cause-effect relationship of sleep-disturbance in neurological and psychiatric disorders.Keywords: Sleep, Actigraphy, PSQI, Cognition, Affect, Well-being.Subject terms: Psychology, Human behaviour Introduction Optimal sleep is crucial for maintaining physical, mental and cognitive well-being, exerting profound influence on mortality and morbidity rates1. Sleep duration and quality are often influenced by a variety of factors, including emotional stress, lifestyle choices, and socioeconomic conditions2. Such lifestyle choices include the rapid proliferation and widespread use of blue light-emitting technology, such as smartphones3. Therefore, unravelling the intricate interplay between sleep and well-being is crucial for developing effective guidance and interventions to promote healthier sleep and associated well-being. This bidirectional relationship underscores the importance of understanding how sleep quality acts as a protective factor against the detrimental effects of emotional stress on subjective well-being4. Altered sleep, measured both subjectively and objectively, is a common symptom in a wide array of neurological and psychiatric disorders, complicating the task of monitoring disease progression and treatment efficacy5–7. Various methods are available for measuring sleep, including polysomnography, actigraphy devices, and subjective sleep questionnaires such as the Pittsburgh Sleep Quality Index (PSQI)8. While polysomnography is often considered the gold standard in sleep measurement, studies have reported high accuracy and specificity in actigraphy measurements of sleep duration and wakefulness compared to polysomnography9,10. Subjective methods, such as the PSQI, are commonly used in clinical trials and medical studies due to their ease and cost efficiency, with the PSQI demonstrating high validity for measuring sleep in clinical populations11. There are, however, disparities between objective and subjective sleep measures, particularly in studies that have employed both approaches simultaneously. Specifically, subjective reports tend to underestimate awakenings, in comparison to objective polysomnography data12. This discrepancy may stem from the retrospective nature of subjective methods, such as the PSQI, which rely upon a participant’s memory of their sleep over the past month13. In turn, memory is itself subject to influence from cognitive and affective states. A lack of correlation between objective and subjective sleep measures appears to be further influenced by various sleep and lifestyle characteristics. Studies have found greater correlation between measures on weekdays compared to weekends14, as well as differences in perception between habitual and disrupted sleep cycles15. Importantly, these studies focus on total objective sleep durations and efficiencies14alongside objective measures of latency, sleep staging and wakefulness15 without accounting for influences of sleep variability over time. The relationship between sleep and cognitive performance has garnered significant attention, particularly using actigraphy in healthy individuals. Studies have highlighted interactions between sleep latency, memory, and verbal task performance16. However, these investigations often focus on either objective (actigraphy) or subjective (self-report) measures of sleep, creating a gap in understanding the complementary roles these measures play in cognitive outcomes. Integrating both approaches is crucial for obtaining a comprehensive view of how sleep impacts cognitive functions. Similarly, systematic reviews of sleep and mood, using either subjective report or objective measurement, suggest a complex reciprocal interaction between sleep and emotional state, with sleep-affect associations appearing to be stronger between subjective than objective measures17,18. Moreover, these reviews highlight the need for further research to determine the potential influence of sleep variability on mood. A more complete understanding of the relationships between, mood, cognition, and sleep, measured objectively and subjectively, is essential for improving the diagnosis and monitoring of conditions in which sleep is a consideration. Therefore, in this study, we aimed to contribute to the understanding of the complex interactions between sleep, mood, and cognitive performance by examining both objective and subjective sleep measures and their interactions with mood and cognitive outcomes in a healthy population. Critically, we included measures of variance in objective sleep, alongside previously investigated average durations and efficiency, derived as median scores, to address a knowledge gap of the role of variable objective sleep on retrospective reporting of subjective sleep quality. We selected the Positive and Negative Affect Schedule (PANAS), the State-Trait Anxiety Inventory (STAI), and Beck’s Anxiety Inventory (BAI) to capture a comprehensive range of self-reported mood states and anxiety levels. PANAS measures the extent of positive and negative affect, providing a broad assessment of mood. STAI and BAI assess both state and trait anxiety, offering insights into temporary anxiety states and more stable anxiety traits. These tools collectively allow for a nuanced understanding of mood and anxiety, which are critical for examining their interaction with sleep and cognitive performance. To complement these subjective measures, we included the Cambridge Neuropsychological Test Automated Battery (CANTAB) Emotional Bias Task, which assesses unconscious mood biases. This task provides an additional layer to recorded affect scores by revealing implicit emotional processing, thus capturing mood dimensions that self-report questionnaires may not capture. For cognitive performance, we utilized the CANTAB Stop Signal Task and Digit Span Task. The Stop Signal Task measures response inhibition, a core executive function involving the ability to suppress a prepotent response. In contrast, the Digit Span Task assesses working memory capacity, which involves holding and manipulating information over short periods of time. These tasks were chosen to capture different aspects of executive functioning, providing a measure of both simple and complex processes. In this study, simple executive functioning refers to cognitive processes requiring more basic control mechanisms, such as response inhibition, which is measured in the Stop Signal Task. Additionally, complex executive functioning involves more higher-order control processes, such as maintaining and manipulating multiple items of information in working memory, which is assessed by the Digit Span Task. Tasks, therefore, differ not only in complexity but also in the specific cognitive domains that we aim to target. This approach aligns with previous studies that report a differential impact of poor sleep on cognitive functioning depending on task complexity of vigilance tasks19. Here we define complexity in relation to performance in response to increased cognitive demand, focusing on the impact of sleep quality on executive function components including response inhibition and working memory. The inclusion of these diverse measures ensures a broader assessment of the interplay between sleep, mood, and cognitive performance, thus offering a representative cross-section for the study. Specifically, we sought to elucidate the extent to which subjective and objective sleep measures are consistent, complementary, or redundant with mood outcomes, shedding light on possible putative mechanisms whereby subjective experiences modulate affective states. By investigating these relationships, we hope to provide insight into instances of sleep characteristics driving affective outcomes in healthy populations. Following prior research outcomes, our hypotheses were structured to address these key points: (i) subjective ratings were expected to provide limited predictive consistency for objective measures of sleep, (ii) participants with poorer objective sleep scores were expected to exhibit lower performance on cognitive tests, and (iii) subjective measures of sleep would interact with mood test scores. Method Participants: Participants (n = 83) were recruited to the study, which was conducted remotely using an actigraphy device over seven days. An in-person assessment was completed on day seven at the University of Plymouth (UoP) Brain Research & Imaging Centre (BRIC). Participants were recruited with the aim of capturing a variety of sleep qualities; opportunity sampling techniques were used to recruit those with stereotypically poor sleep including new parents and undergraduate students. Ethical approval was obtained for the study from the Faculty of Health Research Ethics and Integrity Committee (FREIC:2841). All methods were performed in accordance with the relevant guidelines and regulations, in accordance with the Declaration of Helsinki. Informed consent was obtained from all participants following briefing of the experimental procedures, prior to their inclusion in the study, in accordance with relevant guidelines and regulations. All participants satisfied eligibility criteria that excluded a diagnosis of neuropsychiatric disorder, neurological conditions, and sleep disorder. Seven participants were excluded from the final analysis, as their seven-day actigraphy device data was incomplete. We therefore obtained a final dataset of 76 participants (50 female, 24 male, 2 other), with a mean age of twenty-four years (range 20–44, SD 5.1). Procedure Sleep measures Participants were fitted with a wrist-worn actigraphy device (GENEActiv, ActivInsights), which gathered data through triaxial accelerometers and light sensors (50 Hz sampling rate) to monitor sleep-wake and activity cycles. The devices were worn for a period of seven days, acquiring continuous measurements to determine habitual sleep onset time, rise time, time in bed, time awake, number of active periods, sleep duration and sleep efficiency. As sleep efficiency and duration scores are calculated using a subset of the other measurements acquired by the device, we decided to include these measures alone as the addition of contributing measures in isolation held the potential to introduce multicollinearity amongst predictor variables. These data were logged and computed for each day20, and we used median values and ranges of sleep duration and sleep quality as objective measures. Range, rather than variance, was chosen as the measure of sleep variability to capture the impact of extreme change in sleep pattern. On day seven, participants attended the BRIC research facility to complete the in-person portion of the study. The Pittsburgh Sleep Quality Index (PSQI), a self-report questionnaire widely used in clinical settings to capture sleep quality over the previous month, was used to obtain the subjective measures of sleep based upon a global PSQI sleep-quality score8. The singular global PSQI score is summed from 19 self-rated items forming seven ‘component’ scores. A global score below five indicates good sleep quality whereas higher global scores indicate poorer sleep. Cognition and affect measures To capture a range of cognitive measures, participants completed a battery of automated psychological tests displayed on a mobile tablet device (iPad, Apple, USA). A battery of tasks from the Cambridge Neuropsychological Test Automated Battery21 was used to measure response inhibition, working memory and emotional affect using the Stop Signal Task (SST), Digit Span Task (DST) and Emotional Bias Task (EBT), respectively (Table 1). An additional battery of standardised questionnaire assessments was compiled using Qualtrics and completed to obtain self-reported mood measurements. This comprised the Positive and Negative Affect Schedule (PANAS)22, State-Trait Anxiety Inventory (STAI)23and Beck’s Anxiety Inventory (BAI)24. The PANAS and BAI measures require reflection on differing feelings, emotions and anxious symptoms over the past month, including the day of testing. Whereas the STAI-trait measures retrospectively reflect on how an individual ‘generally’ feels, and the STAI-state measures as individuals how they feel in the present moment. Table 1Elaboration of CANTAB tests, as stated by Cambridge Cognition (2019).Test nameTask descriptionOutcome measuresStop Signal Task (SST)The participant must respond to an arrow stimulus, by selecting one of two options, depending on the direction in which the arrow points. If an audio tone is present, the participant must withhold making that response (inhibition).Stop signal reaction time (SSRT).Digit Span Task (DST)The participant will hear a sequence of digits and then immediately be prompted to verbally repeat the sequence. If the participant responds correctly, the next trial presents a longer sequence. The task terminates when participants responded incorrectly on three occasions at a span length. The participant’s span is the longest number of sequential digits that can accurately be remembered.The longest sequence successfully reached and passed.Emotional Bias Task (EBT)Participants view images of faces that are morphed between two emotions of varied intensities. The variants cover continuums from happy to sad, happy to angry or happy to disgust. Each face is displayed for 150ms, followed by a two-alternative forced choice where they must select one of the two emotions.Bias point: proportion of trials selected as happy compared to alternative emotion (0 to 15 scale). Elaboration of CANTAB tests, as stated by Cambridge Cognition (2019). Data analysis Raw bin file actigraphy data was extracted from the device via USB after 7 days of wear. The raw data recordings were processed using a custom GENEActiv (ActivInsights) script utilising the GENEAread and GENEAclassifyPackages in R. This script analysed the data in 24-hour periods, segmenting into predefined variables covering temperature, light, magnitude, and degrees of motion. Default thresholds were used to identify sleep onset and rise times where the default times for sleep onset were set to post-3pm and rise time as pre-3pm to ensure capturing of data specific to each individual night’s sleep. A sleep threshold of 2.5 h was set to identify large blocks of sleep in the data which may represent the nighttime sleep period and was used to filter out shorter naps or sustained inactivity. Median sleep duration and efficiency scores were derived from measures of total time in bed, total sleep time, time awake and the number of active periods. In addition, to capture variability in sleep habits, particularly the influence of extreme changes, ranges were calculated by subtracting maxima and minima across the 7-day recording period. Statistical analysis The predictor variables in this study were objective sleep quality and quantity (medians and ranges) taken from actigraphy devices and subjective sleep quality reflected in global PSQI scores. The response variables were SST reaction times, EBT bias point, DST maximum length, PANAS positive and negative scores, STAI-t, STAI-s and BAI scores. All statistical analyses were completed using R-Studio25. Data met the assumptions (linearity, normality, homoscedasticity, and independence) required to conduct linear regression as determined using Q-Q, scale-location and residuals vs. fitted plots. Multicollinearity was assessed by calculating variance inflation factors of the predictor variables using the ‘car’ package in R – VIF scores ranged between 1.24 and 2.11, adequately passing the assumption that there is no multicollinearity present among these variables26. A multiple linear regression model of main effects used a formula of ‘scale(global PSQI scores) ~ scale(median sleep duration) + scale(range of sleep duration) + scale(median sleep efficiency) + scale(range of sleep efficiencies)’ to investigate relationships between subjective sleep as a response variable and objective sleep variables as predictors. An additional regression analysis investigated the relationship between objective sleep duration and PSQI raw sleep duration scores from the PSQI question ‘During the past month, how many hours of actual sleep did you get at night?’ using the formula ‘lm(scale(PSQI raw duration) ~ scale(median sleep duration)’. Raw self-reported duration is not a validated output of the PSQI, yet this comparison allowed for a more like-for-like comparison on subjective and objective sleep duration scores. Additionally, the relationship between all sleep characteristics as predictors, and cognitive and mood as response variables, was investigated using the following model: ‘scale(DV) ~ scale(median sleep duration) + scale(range of sleep duration) + scale(median sleep efficiency) + scale(range of sleep efficiencies) + scale(global PSQI scores).’ FDR p-value correction was carried out using the Benjamini Hochberg method to address the issue of multiple comparisons. Bayes Factor analyses were conducted to investigate any non-significant findings. This analysis was performed using lmBFin the BayesFactor package in R27. A Bayes value nearing 1 indicated that the data was neither in favour of the null or alternative hypothesis, suggesting a lack of sensitivity to effects. Whereas a value of less than 1 presented support in favour of the null, and a Bayes Factor value of greater than 3 shows substantial evidence for the alternative hypothesis28. Results Relationship between objective and subjective sleep measures Linear regression models were used to investigate: (1) the potential for a relationship between subjective sleep quality and objective median sleep duration and efficiency, (2) ranges of objective sleep duration and efficiency across the week of actigraphy recording as predictors, and (3) raw PSQI sleep duration self-reported scores and objective median sleep duration scores. This analysis identified no significant relationships between these measures (Table 2; Fig. 1). Table 2Investigating relationships between objective sleep measures (actigraphy) and subjective sleep quality (global PSQI scores).Global PSQIAdj R2β [95% CI]pFDRBFSleep DurationMedian− 0.01− 0.01[−0.31, 0.29]0.950.25Range0.12 [−0.16, 0.40]0.820.64Sleep EfficiencyMedian0.04 [−0.26, 0.33]0.950.24Range0.12 [−0.16, 0.41]0.820.67PSQI Reported Sleep DurationSleep DurationMedian0.020.17 [−0.06, 0.40]0.150.60 Investigating relationships between objective sleep measures (actigraphy) and subjective sleep quality (global PSQI scores). Non-significant relationships between subjective sleep quality (global PSQI scores and raw PSQI self-reported sleep duration) and objective sleep measures. Relationship between objective and subjective sleep measures and affect scores Analyses revealed specific sleep-related influences on mood scores. Specifically, positive relationships were observed between global PSQI scores of subjective sleep quality and STAI-t reflecting trait anxiety (β [95% CI] = 0.34 [0.11, 0.57], pFDR = 0.02), prolonged state anxiety as reflected by BAI scores (β [95% CI] = 0.41 [0.19, 0.63], pFDR = 0.002), and negative affect as measured by the PANAS (β [95% CI] = 0.35 [0.12, 0.57], pFDR = 0.01). This demonstrates that poorer sleep quality (higher PSQI score) was accompanied by higher anxiety levels, and more negative and intense mood. Accordingly, we observed a negative relationship between PSQI scores and positive affect measured by the PANAS (β [95% CI] = − 0.30 [−0.52, − 0.07], pFDR = 0.05) (Fig. 2). Bayes Factor values for these significant findings provide additional support highlighting robust evidence of an effect (Table 3). In contrast, frequentist regression analysis identified no significant relationships between objective sleep (actigraphy) measures and affect scores (Table 3). Bayesian evidence suggests an insensitivity to detect effects between ranges of sleep efficiency and emotional bias point measures (1.25). See Table 3 for full detailed results. Significant relationships between subjective sleep quality and mood measures. STAI-t and PSQI: Adj R2 = 0.07, β = 0.34, pFDR = 0.02. PANAS positive and PSQI: Adj R2 = 0.10, β = − 0.30, pFDR = 0.05. BAI and PSQI: Adj R2 = 0.14, β = 0.41, pFDR = 0.002. PANAS negative and PSQI: Adj R2 = 0.09, β = 0.35, pFDR = 0.01. The solid lines represent the best-fit regression lines for eachrelationship. Table 3Main effects of objective and subjective sleep quality on measures of cognition, anxiety and affect. All values are scaled and obtained from linear multiple regression with all models including objective and subjective sleep quality and quantity as predictors. BF denotes Bayes factor values. Objective sleep scores measured using actigraphy devices. Subjective scores measured using the PSQI assessment. FDR p-values presented, corrected using the Benjamini Hochberg method, with significant findings presented in bold.Sleep DurationSleep EfficiencySubjective Sleep QualityMedianRangeMedianRangePSQIAdj R2β [95% CI]pFDRBFβ [95% CI]pFDRBFβ [95% CI]pFDRBFβ [95% CI]pFDRBFβ [95% CI]p FDRBFEmotion Bias Point0.02− 0.14 [−0.39, 0.11]0.440.350.05 [−0.10, 0.19]0.660.480.03 [−0.01, 0.07]0.440.250.20 [−0.10, 0.47]0.441.25− 0.05 [−0.29, 0.18]0.660.23Stop Signal RT0.02− 0.11 [−0.36, 0.14]0.490.44− 0.12 [−0.26, 0.03]0.340.440.004 [−0.04, 0.04]0.830.250.20 [−0.08, 0.48]0.340.30− 0.15 [−0.38, 0.08]0.340.49Digit Span Max Pass− 0.0070.20 [−0.06, 0.45]0.621.11− 0.04 [−0.18, 0.11]0.760.250.002 [−0.04, 0.43]0.920.470.11 [−0.18, 0.39]0.760.240.07 [−0.17, 0.30]0.760.27STAII State− 0.030.01 [−0.25, 26]0.960.25− 0.01 [−0.16, 0.13]0.960.240.01 [−0.03, 0.05]0.960.290.01 [−0.28, 0.30]0.960.240.18 [−0.06, 0.42]0.670.65STAII Trait0.07− 0.01 [−0.26, 0.23]0.930.260.01 [−0.13, 0.15]0.930.250.02 [−0.02, 0.06]0.920.36− 0.03 [−0.31, 0.24]0.930.240.34 [0.11, 0.57] 0.02 12.23Beck’s Anxiety Inventory0.14− 0.13 [−0.36, 0.11]0.430.25− 0.06 [−0.20, 0.07]0.430.370.02 [−0.02, 0.06]0.430.34− 0.09 [−0.35, 0.18]0.520.280.41 [0.19, 0.63] 0.002 36.47PANAS Positive0.10− 0.18 [−0.42, 0.06]0.270.77− 0.09 [−0.23, 0.05]0.270.380.01 [−0.03, 0.05]0.630.260.17 [−0.10, 0.44]0.270.28− 0.30 [−0.52, − 0.07] 0.05 4.23PANAS Negative0.09− 0.12 [−0.36, 0.12]0.560.26− 0.04 [−0.18, 0.10]0.670.240.02 [−0.01, 0.06]0.520.310.06 [−0.22, 0.33]0.670.290.35 [0.12, 0.57] 0.01 15.67 Main effects of objective and subjective sleep quality on measures of cognition, anxiety and affect. All values are scaled and obtained from linear multiple regression with all models including objective and subjective sleep quality and quantity as predictors. BF denotes Bayes factor values. Objective sleep scores measured using actigraphy devices. Subjective scores measured using the PSQI assessment. FDR p-values presented, corrected using the Benjamini Hochberg method, with significant findings presented in bold. Relationship between objective and subjective sleep measures and cognition Analyses identified no significant relationships between subjective sleep quality scores, reflected by each participant’s global PSQI score, and cognitive measures. Furthermore, frequentist regression analyses identified no significant relationships between the objective actigraphy measures and cognition (Table 3). However, Bayesian outcomes suggested an insensitivity to detect effects between the following measures: average sleep duration and digit span scores of working memory (1.11). As the Bayes Factor value nears a score of 1, this outcome deems the findings inconclusive, due to a lack of power to detect - what may be - a smaller effect size.We saw an opportunity to further investigate this outcome during post-hoc analysis after visual inspection of the data revealed a large proportion of data within the inner quartiles, likely overshadowing effects of both higher and lower quantities of sleep. Post-hoc investigations Bayes Factor values indicated inconclusive findings (BF = 1.11) for the relationship between average sleep duration and working memory capacity (i.e., digit span). Given this inconclusive Bayesian evidence, and our research-based hypothesis predicting objective sleep and cognitive outcomes, we explored this relationship further through visually inspecting median objective sleep duration data using a density plot (Fig. 3). This exposed a high frequency of data within the inner quartiles, which led us to carry out post-hoc exploratory analysis, isolating the lower and upper quartiles for further investigation. The resulting lower quartile had a group mean sleep duration of 4.72 h (SD = 0.84) and the upper quartile had a mean duration of 7.53 (SD = 0.36). A significant difference is observed in the maximum digit span sequence between the lower and upper median sleep duration groups (Fig. 4). Individuals with higher sleep duration exhibited superior working memory performance (M = 6.55, SD = 1.19) compared to lower sleep duration (M = 5.45, SD = 1.73, (t (38) = −2.34, p = .02). Density plots of sleep data distributions across the study sample. Maximum Digit Span scores for participants in upper and lower sleep duration quartiles. Participants with higher sleep duration had a mean score of 6.55, while those with lower sleep duration had a mean score of 5.45, (t(38) = −2.34, p = .02). Discussion This study explored the relationship between subjective and objective measures of sleep, with a particular focus on the cross-predictive reliability of PSQI and actigraphy sleep measures in a young, healthy sample population. We aimed to understand the extent to which subjective and objective sleep measures are consistent, complementary, or redundant by exploring their association with cognition and mood measures. Specifically, we analysed response inhibition, working memory, emotional bias, anxiety, and affect, to determine the extent to which they interact with subjective reports and objective measures of sleep. Consistent with prior research12,13, our findings reveal a significant disconnect between subjective reports of sleep and corresponding objective sleep measurements, with no notable relationship between global PSQI scores and actigraphy data. This fascinating observation suggests that (i) people poorly encode or inaccurately retrieve quantitative sleep information, and/or (ii) the datatypes we are measuring are insensitive to, and inaccurately reflect, actual sleep processes, and/or (iii) the subjective interpretation of sleep quality is more heavily influenced by other factors. Accordingly, we demonstrate here an indication of the way in which objective and subjective sleep are associated with cognitive and mood measures. Specifically, while objective sleep measures were predictive of our working memory cognitive test outcome in our subset analysis, the subjective measures were not predictive of any cognitive measures. Conversely, and perhaps more important from a clinical perspective, the subjective reports of sleep quality were strongly predicted by mood. Importantly, participant mood was not predicted by the objective measures of sleep, which we propose reflects a causal relationship. Specifically, we propose that poor sleep, defined objectively, does not result in a decline in mood scores, but that low mood scores result in the perception of poorer sleep quality. Similarly, it is plausible to suggest that a perception of poorer sleep may lead to lower mood, fuelling a subjective sleep-mood feedback loop. Yet, it is possible that an unknown factor, not measured in the present study, impacts upon both mood and perception of sleep quality. As previously noted, it is important to consider that the 30-day retrospection required to complete the PSQI may contribute to the disparity between this subjective report and the objective measures from actigraphy, which recorded only the previous 7-days13. It is possible that collection of the actigraphy data over a 30-day period would alter this association, however it is reasonable to assume that the PSQI sleep perceptions will be more heavily weighted to recency. Alternatively, it is possible that objective and subjective measures of sleep may capture differing facets of sleep as previous studies collecting data across the same periods as our own have also failed to identify coherence between subjective and objective measures of sleep quality29. This finding raises concerns regarding the accuracy of sleep assessment tools commonly used in clinical settings, potentially leading to an inadequate understanding of a patient’s actual sleep patterns and health and the associations with symptoms of treatment effects. The present study sought to characterise sleep, cognition, and mood in a cohort of healthy younger adults, representative of the general population. Consequently, the range of sleep measures acquired were over a narrow range without a normal distribution due to the habitual nature of sleep as opposed to manipulating or depriving sleep in this sample. Specifically, sampling in this population is heavily weighted toward ‘good’ sleep, which reduces the sensitivity of analysis of interactions between sleep characteristics and other measures. To address this, we conducted post-hoc analyses to explore the near significant observations, by isolating the lower and upper quartiles to investigate the extremes of habitual sleep. Post-hoc t-test uncovered an effect of objective average sleep duration on working memory capacity, where those in the upper quartile (ranging from 7.17 to 8.38 h of sleep on average, with a mean of 7.53 h) had a greater working memory capacity than those in the study’s lower quartile (ranging from 3.25 to 6.07 h of average sleep duration, with a mean of 4.72 h). This finding from our subset analysis, along with non-significant relationships between objective sleep measures and stop-signal outcomes - reflective of response inhibition - supports known research findings on sleep restriction and deprivation whereby reduced sleep is associated with a decline in complex functioning yet simpler cognitive functions remain intact19. However, when assessing working memory in particular, a decline is commonly seen in reaction times and speed of processing contributing to working memory capacity as opposed to accuracy alone30,31. These observations are of particular significance when considering neurological and other disorders that impact upon quantified sleep quality and which have potential cognitive implications. For example, Parkinson’s disease (PD), while dominated by motor symptoms, is known to impact sleep quality32. While non-motor symptoms such as cognitive decline are a factor of increasing importance in PD, it is unclear to what extent these aspects can be accounted for by sleep quality. We would acknowledge that, given the existing complexity in understanding the relationship between sleep and cognition, our aim of capturing participants’ sleep and behaviour as naturally as possible may pose a slight limitation to the overall conclusions we draw regarding this relationship. As we did not ask participants to modify their lifestyles during the study, unmonitored factors, such as alcohol consumption, could have independently influenced actigraphy measurements by affecting sleep stages not detected by these devices. Moreover, unlike measures such as polysomnography, the actigraphy data does not provide an accurate record of sleep onset latency, which may be an important inclusion for further studies aiming to comprehensively address the role of this feature in cognitive and mood influence. We, nonetheless, propose that the PSQI is an inappropriate clinical measure to capture sleep-related information accurately, given its lack of agreement with objective data and the probable confounding influence of mood on retrospective reports, whereas an objective measure such as actigraphy has the potential to inform on possible underlying mechanisms of cognitive impairment. This study highlights the significant role of affect and anxiety levels in shaping subjective perceptions of sleep quality. We observed significant positive relationships between global PSQI scores and STAI trait anxiety, Beck’s anxiety inventory and negative affect and a negative relationship between global PSQI scores and positive affect. Overall, these findings underscore the impact of emotional well-being on subjective sleep perception, consistent with prior literature17,18. Critically, we believe this to be an influential relationship since there was no relationship between objective sleep measures and mood outcomes. This association suggests that how individuals perceive their own sleep quality may influence their mood, or similarly, if bi-directional, suggests that improving mood states may influence perceptions of sleep regardless of whether their sleep did notably improve. Previous research supports this idea, finding that lower subjective sleep quality, reporting using global PSQI scores, were related to greater emotional awareness - an aspect of interoception involving recognising internal emotional states33. Importantly, we highlight several points of potential value in the consideration of clinical populations. First, while subjective sleep measures may be of limited value in capturing the physiological aspects of sleep quality, they nonetheless provide unique insights into an individual’s perception of their sleep which should be treated as an independent outcome to physiological sleep. This subjective perception is valuable not only as a simple proxy for mood and affective state but also to provide an independent dimension of sleep quality in instances where divergence between perceived and actual sleep is anticipated, for example, in conditions like insomnia where these discrepancies are common. Second, as affective state is often driven by the primary clinical condition, the PSQI may serve as a potentially valuable tool for ongoing monitoring that avoids the stigma associated with direct mental health tests. Finally, if this association is shown to be robust, the PSQI provides a simple metric to evaluate the effectiveness of interventions aimed at improving patient wellbeing. Conclusion In conclusion, this study contributes to the growing body of literature examining the intricate relationship between subjective and objective sleep measures, cognitive functioning, and mood outcomes. This study provides greater evidence for the disparity between objective and subjective sleep scores, following investigation of additional variance-based measures alongside commonly used objective averages. Furthermore, the findings extend our understanding of the association between objective measures and cognitive function, in particular the impact upon complex executive functioning and memory processes and the importance of applying an appropriate objective measure to capture the potential basis for cognitive decline in disorders where sleep quality is impaired. We have also demonstrated that while subjective sleep measures may not directly and accurately reflect true sleep quality, these ratings are directly driven by mood and affective state. This provides a potentially valuable proxy for monitoring patient wellbeing where it may be impacted secondary to another clinical condition. Finally, we underscore the potential value of acquiring both objective and subjective measures to obtain a comprehensive understanding of the participant, albeit from a perspective that differs from that previously proposed. Author contributions H.W. and S.D.H conceptualised the study, H.W and A.D conducted data collection, H.W. analysed the results, H.W wrote the original draft preparation. All authors were involved in reviewing and editing the manuscript. Data availability The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request.