Daily stress did not consistently predict poorer sleep, but a better-than-usual night of subjective sleep was linked to lower perceived stress the following day, highlighting how stress–sleep relationships differ within and between individuals.
The relationship between stress and sleep quality: An intensive longitudinal study. Image Credit: Prostock-studio / Shutterstock
A new Scientific Reports study conducted an intensive longitudinal analysis to capture the bidirectional dynamics between stress and sleep in daily life, using both objective and subjective measures.
The Association Between Stress and Sleep Quality
Stress has been associated with poorer subjective and objective sleep quality. The former reflects an individual’s satisfaction with their sleep, while the latter is based on physiological or behavioral data. They only moderately correlate, despite being relevant to health, which motivates assessing both to holistically understand the sleep experience.
The relationship between stress and sleep is complicated: poor sleep is associated with higher perceived stress, but hormonal balance, impairments in emotion regulation, and cognitive function can also explain this association. Existing research is mainly based on cross-sectional or laboratory studies. While such studies provide valuable insights into causal mechanisms through polysomnography and stress induction, these controlled settings limit ecological validity.
Generalizability can be improved via daily-life studies, but an important limitation is that these studies rely heavily on self-reported stress and sleep. There is limited use of physiological or objective measures. Few studies have examined the bidirectional, day-to-day dynamics of sleep and stress while accounting for temporal fluctuations.
Scientists Examined Stress–Sleep Dynamics via Wearable Sensors and Smartphone Self-Reports
The hypothesis at the within-person level was that higher perceived stress during the day is associated with lower subsequent subjective and objective sleep quality. Associations between sleep quality and subsequent stress were also studied, given the bidirectional relationship.
Over 7 days, electrocardiogram and accelerometry data were measured for 106 university students, as this group is known to frequently experience sleep disturbances and high levels of stress. About 86% were female, and the mean age of the students was 22.65 years.
Self-report data were completed by participants on their smartphones every two hours. Subjective sleep quality was assessed every morning using a question, “How well did you sleep last night?”, which was answered on a 5-point Likert scale.
Objective sleep quality was assessed using wearable-sensor data and the device’s actigraphy-based sleep-staging algorithm. Both momentary and daily measures of perceived stress were included to capture stress across different temporal levels. Additional heart rate variability reductions (AddHRVr), which account for concurrent metabolic demand, were used as an indicator of activity in stress-related physiology. Alcohol, nicotine, and caffeine use near bedtime were included as a combined covariate in sensitivity analyses.
Study Design and Measurement
Weak and Inconsistent Links Between Stress and Sleep
At the between-person level, bivariate Spearman’s correlations were computed between subjective and objective sleep quality indicators, and overall, a small positive correlation was observed. AddHRVr, averaged daily perceived stress, and mean momentary stress were positively correlated with one another. Generally, persons with higher perceived stress showed longer sleep onset latency, poorer subjective sleep quality, longer wake after sleep onset, shorter total sleep time, and lower sleep efficiency, although these associations were inconsistent. These bivariate correlations were descriptive, whereas the primary cross-lagged models provided clearer evidence, primarily of an association between higher perceived stress and poorer subjective sleep quality.
At the within-person level, there was little evidence of stress affecting subsequent sleep quality. Sleep effects were small for both daily stress and mean momentary stress. Similarly, associations between AddHRVr and subsequent sleep quality were almost zero, suggesting no within-person relationship between stress and subsequent sleep quality. In contrast, when subjective sleep quality was better than usual, participants tended to report lower stress the following day, although these effects were small and somewhat uncertain.
Concerning objective sleep parameters, derived from self-reported nocturnal timeframes, model results indicated that associations between objective sleep quality and stress indicators were generally highly uncertain and very small in magnitude. Across multiple indicators, including total sleep time, sleep efficiency, sleep onset latency, and wake after sleep onset, most estimates were close to zero, and their confidence intervals indicated no consistent association.
In sensitivity analyses, models were adjusted using a combined measure of alcohol, nicotine, and caffeine consumption near bedtime, and complete-case data were used. The results were similar to the main analyses. The effect sizes remained very small, and credible intervals largely overlapped zero for a range of stress indicators and sleep quality parameters. The findings were robust to both the inclusion of covariates and the handling of missing data.
Multiverse analyses were conducted to compare different operationalizations of nocturnal timeframes (self-reported, fixed, and accelerometry). Nonparametric Friedman tests were used to examine the variation across the three operationalizations. Differences were noted in wake after sleep onset, sleep onset latency, total sleep time, and sleep efficiency.
Follow-up tests showed that, for the fixed nocturnal timeframe, wake after sleep onset, sleep onset latency, and total sleep time were higher than those from the self-reported or accelerometry-based timeframes. However, sleep efficiency was lower. Additionally, compared to accelerometry, wake after sleep onset was higher in the self-reported nocturnal timeframe.
One exception emerged in the fixed-timeframe models, where people reporting higher daily stress showed longer sleep onset latency at the between-person level. However, this association was not consistently reproduced using the other nocturnal timeframe definitions and should therefore be interpreted cautiously. The authors noted that the fixed 10 p.m. start could partly capture later bedtimes rather than greater difficulty falling asleep once participants intended to sleep.
The overall pattern of weak and uncertain associations between objective sleep parameters and stress remained consistent across different analytic choices.
Conclusions
These findings show that within- and between-person relationships can differ, underscoring the need for intensive longitudinal studies to capture these nuances. The observed variability across different operationalizations of nocturnal timeframes further highlights how methodological choices can shape results in sleep research.
Several limitations should be considered. Stress levels during the monitoring period were generally low to moderate, which may have limited the ability to detect stronger associations, and participants may have been less likely to respond to prompts during periods of high stress. The observational design also precludes causal conclusions, and the predominantly female university-student sample limits generalizability.
Future research should investigate potential non-linearities in the sleep–stress relationship, improve measurement of both stress and sleep quality, and explore additional mechanisms, such as sleep hygiene and perseverative cognition, that may influence these associations.
Journal reference:
- Bamert, M., Schwerdtfeger, A., Rominger, C., & Inauen, J. (2026). The relationship between stress and sleep quality: An intensive longitudinal study. Scientific Reports.DOI:10.1038/s41598-026-63953-7, https://www.nature.com/articles/s41598-026-63953-7