Half a Million Sleep Reports Reveal When Our Moods Rise and Fall

What happens to the human mood between midnight and dawn has long been a matter of clinical conjecture, pieced together from small laboratory studies and the linguistic fingerprints people leave on social media. Now a team of researchers has turned to an unlikely source of insight: a popular online brain-training platform. In one of the largest datasets of directly self-reported behavioral health ever assembled, more than 468,000 Lumosity users aged 20 to 79 recorded how they felt and how long they had slept, producing an unprecedented real-time map of mood and sleep across the hours of the day and the days of the week. The findings, published in PLOS Digital Health, offer both confirmation of long-suspected rhythms and some striking surprises about who sleeps how much, and when.

The scale of the analysis is what sets it apart. Traditional sleep and mood research typically depends on small cohorts recruited into laboratory protocols, where participants sleep under monitoring equipment for a handful of nights. Social media studies offer volume but rely on indirect inference, decoding emotional states from word choice rather than asking people directly. The new study sidesteps both limitations. Users of the cognitive-training platform voluntarily reported their mood on a scale from minus two to plus two and their sleep duration in categories ranging from five hours or less to more than nine hours. Because the reports came from ordinary people going about their daily lives, the data capture something laboratory work cannot: the natural texture of mood and sleep as they unfold across real schedules, real workweeks, and real weekends.

The temporal pattern of mood that emerged is both intuitive and sobering. Mood reports were highest in the afternoon and evening, and lowest from late night through the morning hours. In other words, the darkest emotional hours are literally the dark ones, with negative affect peaking when most of the world is asleep or struggling to be. This circadian trough is well documented in smaller studies, but seeing it reproduce across hundreds of thousands of self-reports gives it a new weight. The authors note that the late-night peak in negative affect aligns with a substantial body of literature on nocturnal suicide risk, a convergence that lends the finding public health significance, even though this nonclinical sample cannot directly address clinical risk or the need for crisis services.

Age told its own story. Mood followed a U-shaped curve across the lifespan, with younger adults and older adults reporting better moods than those in the middle, and this pattern held consistently across every day of the week. The midlife dip in mood is a familiar motif in psychology, sometimes described as a U-shaped curve of well-being, but its appearance in a dataset collected incidentally through a brain-training platform suggests it is robust enough to surface even when nobody is being asked about happiness directly. The researchers highlight midlife as a key intervention period, arguing that the demographic window between young adulthood and older age may be where sleep and mood support could do the most good.

Sleep, unsurprisingly, was tightly bound to mood throughout the data. Longer reported sleep duration was linked to better mood reports at every hour of the day and every day of the week, a relationship so stable that it persisted regardless of when a user happened to log in. Perhaps more intriguing was the interaction between sleep and the weekly cycle: the difference in mood between weekdays and weekends was larger among users who reported longer sleep. This suggests that the psychological payoff of a restful weekend is amplified for those who actually manage to extend their sleep, hinting at what sleep researchers sometimes call social jetlag, the mismatch between biological clocks and socially imposed schedules that many people only partially repay on their days off.

Sleep duration itself was not constant across the clock or the calendar. Users reporting in the afternoon and evening recorded longer sleep than those reporting overnight or in the morning, and weekend reports exceeded weekday ones, a pattern consistent with people catching up on rest when work schedules loosen their grip. The weekly rhythm of sleep is one of the clearest signatures of the tension between biological need and social obligation, and its appearance in self-reported data from a gaming platform underscores how deeply the structure of the workweek is etched into modern sleep behavior.

The demographic disparities in the data are among its most consequential findings. Women reported longer sleep overall than men, a result consistent with prior survey research but now documented at a scale few studies can match. Black users reported shorter sleep than other groups, a disparity that echoes a substantial and growing literature linking racial differences in sleep duration to environmental, occupational, and social stressors. Sleep duration also followed a life-course trajectory, declining from young adulthood through midlife before rising slightly among older users, mirroring the U-shaped mood curve and reinforcing the picture of midlife as the period when both sleep and mood reach their low ebb.

The authors are careful to flag the limitations that come with digital trace data. The sample is likely skewed toward people with higher educational attainment and greater engagement with health-related behaviors, which means the disparity estimates probably understate the true population-level differences rather than exaggerating them. Self-reported sleep duration is also less precise than polysomnography, and the mood scale, while simple, captures only a coarse snapshot of emotional state at the moment of reporting. Yet these caveats cut both ways: if sleep disparities between demographic groups are this visible even within a relatively advantaged, health-engaged sample, the underlying differences in the broader population may be considerably larger.

What the study ultimately demonstrates is the value of large-scale digital platforms as instruments of public health surveillance. A brain-training game, designed for entertainment and cognitive exercise, quietly accumulated a behavioral health dataset of a size and granularity that conventional research infrastructure would struggle to match. The hourly and weekly rhythms it reveals, the demographic gaps it documents, and the midlife vulnerability it highlights all point toward the same conclusion: the patterns of daily life, when measured honestly and at scale, can tell us where to look for problems and when to intervene. As digital platforms continue to weave themselves into daily routines, they may become an essential complement to clinics and laboratories, offering a continuously updated portrait of how a population sleeps, feels, and copes with the demands of the week.

Subject of Research: Large-scale self-reported patterns of mood and sleep duration across time of day, day of week, and demographic groups

Article Title: Large-scale patterns of self-reported mood and sleep across the day and week on an online cognitive-training platform

Article References: Mendoza, R. A., Tubbs, A. S., Perlis, M. L., Fernandez, F.-X., & Grandner, M. A. (2026). Large-scale patterns of self-reported mood and sleep across the day and week on an online cognitive-training platform. PLOS Digital Health, 5(9), e0001754. https://doi.org/10.1371/journal.pdig.0001754

Image Credits: AI Generated

DOI: 10.1371/journal.pdig.0001754

Keywords: mood, sleep duration, circadian rhythm, digital health, cognitive training, demographic disparities, midlife, weekend sleep, self-report, public health, Lumosity, PLOS Digital Health