Sleep Problems in Autism? Look at What Happens During the Day
Key Findings
Out of 308 activity, sleep, and circadian measures fed into a machine learning model, 52 were selected as associated with autism. The model could distinguish autistic from non-autistic participants with an AUC of 0.812 (validated at 0.756 in a separate dataset). That's a solid signal.
The surprise: the strongest markers weren't sleep variables. They were physical activity features - time spent sedentary, duration of light activity, time in mixed activity. Autistic adults were more sedentary, for longer stretches, with less light and moderate physical activity throughout the day. Sleep differences were subtler than expected. Total sleep duration wasn't significantly different between groups. The variability of sleep duration was.
Here's the key interaction: for every hour of sedentary time during the day, autistic participants lost about 23 minutes of sleep that night. Non-autistic family members lost about 17 minutes. After controlling for age, sex, and family effects, the difference shrank but remained statistically significant (p = 1.13e-10).
This means the activity-to-sleep pipeline works differently in autistic brains. The same amount of sitting around costs more sleep if you're autistic. The researchers estimated that based on the average sedentary time in their autistic sample (about 11 hours), autistic participants slept roughly 28 to 71 minutes less than their non-autistic relatives.
Of the 52 autism-associated measures, 27 (over half) were measures of day-to-day variability, not averages. It wasn't just that autistic adults were more sedentary on average. Their patterns were also less consistent: more variable sleep durations night-to-night, more erratic activity levels, lower sleep regularity scores.
The Sleep Regularity Index (a measure of how likely you are to be in the same state - asleep or awake - at the same time on consecutive days) was significantly lower in autistic participants across both study cohorts. SPARK participants with reported sleep problems scored even lower. Regularity itself correlated with physical activity: more activity, more regular sleep.
When the researchers compared their activity data to autism trait questionnaires, lower levels of mixed physical activity correlated with more social communication challenges as measured by the Social Responsiveness Scale. More wakefulness after falling asleep (WASO) also correlated with greater social difficulties.
This doesn't prove that inactivity causes social problems. But the association adds another thread to the pattern: physical activity, sleep quality, and core autism traits all move together. Touch one and you may affect the others.
Why It Matters
If you have an autistic family member with sleep problems, you've probably tried the standard advice: consistent bedtime, no screens before bed, melatonin. Some of that helps. But the most common conversation about autism and sleep focuses on what happens at night. This research redirects the lens to what happens during the day.
The data here suggest that autistic people's bodies are more sensitive to daytime activity levels when it comes to sleep. Sitting for long periods doesn't just mean a less active day - it means a worse night, and the penalty is steeper for autistic people than for their non-autistic family members sharing the same household.
This aligns with what autistic adults themselves report: that their bodies and brains have a harder time transitioning between states. A body that's been still all day doesn't flip a switch to "sleep mode" easily. And the day-to-day inconsistency finding matters too. It's not just about one bad day of sitting around. It's about the pattern being unpredictable, which messes with circadian rhythms.
None of this means sleep-focused interventions are wrong. It means they're incomplete. The daytime side of the equation deserves equal attention.
The Fine Print
The study measured activity and sleep at the same time. It cannot tell us whether inactivity causes poor sleep, whether poor sleep causes inactivity, or whether both are downstream of something else entirely (like medication side effects, depression, or sensory overload driving both stillness and insomnia).
The researchers know this. They explicitly call for interventional studies. But the framing of the paper leans toward "activity affects sleep" more than the data alone supports. The day-by-day analysis (today's activity predicting tonight's sleep) is suggestive of direction, but it's not a randomized trial.
The non-autistic comparison group is family members of the autistic participants. They share genetics and often share environments. The researchers acknowledge they need unrelated non-autistic controls to truly validate these findings.
Family controls have a real advantage (shared environment reduces some confounds) but also a real weakness: if autistic traits run on a spectrum in families (which they do - Issue #9 covered this exact dynamic), these "non-autistic" family members may themselves have subclinical activity and sleep differences. That would make the observed differences between groups smaller than the true autistic vs. general population gap. The actual signal might be larger than what this study found.
45% of the autistic participants also had ADHD. When the researchers ran their machine learning model against ADHD status instead of autism, it still performed significantly (AUC 0.704). It didn't work for anxiety (0.564) or depression (0.545), but the ADHD overlap is real.
This doesn't invalidate the findings. The high co-occurrence of ADHD and autism is the reality of this population. But it does mean we can't cleanly separate "this is an autism thing" from "this is an ADHD thing" or "this is a neurodivergent thing." For families, that distinction may not matter. For the science, it matters a lot.
A systematic review estimates that about 61% of autistic adults use psychotropic medications - drugs that independently affect both sleep and physical activity. Antipsychotics cause sedation and weight gain. SSRIs suppress REM sleep. Stimulants for co-occurring ADHD increase daytime activity while disrupting nighttime sleep. Any of these could produce the exact correlation this study found - not because inactivity causes poor sleep, but because the same pill is changing both.
The paper does not control for medication use. With 45% of their autistic sample also having ADHD (and likely medicated), and 47% reporting depression (and likely on SSRIs), this is not a small oversight. It's the single biggest unaddressed confound.
The researchers found that self-reported sleep efficiency correlated poorly with accelerometer-derived sleep efficiency (r = 0.038, which is basically zero). Total sleep time and sleep onset matched moderately (r = 0.44 to 0.73), but individual-level disagreement was wide.
This is actually an important finding in itself. If your autistic family member says they're sleeping terribly but their doctor says "the sleep study looks fine," both might be right in different ways. Subjective experience and objective measurement capture different things. This study's reliance on wearable data is a strength (objective, long-term, home-based), but it misses the experience of sleep. Feeling rested and being physiologically asleep are not the same.
What to Do With This
Break up long sedentary stretches. The study found that extended periods of inactivity (not just total sedentary time, but long uninterrupted blocks) were among the strongest autism-associated markers. If your kid's default after school is 4 hours of screen time without moving, even brief movement breaks could help. A walk around the block, a trampoline session, a dance break - it doesn't need to be structured exercise.
Track the pattern, not just the night. If you're logging sleep issues, try logging daytime activity too. You might notice that bad sleep nights follow low-activity days. That connection might not be obvious without looking at both sides of the clock.
Aim for consistency. The day-to-day variability finding means that regularity matters. A consistent moderate-activity routine might do more for sleep than occasional bursts of intense exercise followed by nothing. The Sleep Regularity Index data backs this up - more regular daily patterns correlated with more regular sleep.
Meet your kid where they are. "Move more" is easy advice to give and hard advice to follow when your child has sensory sensitivities, motor coordination challenges, or social anxiety about group activities. This isn't about gym class or team sports. It's about finding the kind of movement that works for your specific kid - swimming, bouncing, walking, dancing alone in their room. The research says the type of activity matters less than the consistency. If you want help spotting the connection between daytime activity and nighttime sleep, tools that track both in one place can make the pattern visible - that's part of what we built Brainloot to do.
Your body keeps a running tab on movement. If you've noticed that your sleep gets worse during stretches of low activity, this study gives that observation some objective backing. The autistic participants in this study averaged about 11 hours of sedentary time per day. Even small reductions in that number were associated with better sleep.
Consider a wearable if you don't have one. The disconnect between how autistic adults reported their sleep and what the accelerometers measured was striking. A fitness tracker or smartwatch that logs activity and sleep patterns might reveal connections you can't feel in the moment. Not as a judgment tool - as an information tool.
This is not "just exercise more." We know. Sensory overload at gyms, motor coordination challenges, autistic burnout depleting your energy budget - these are real barriers, not excuses. If you're in a burnout period and barely functioning, adding exercise pressure is the wrong move. But if you're in a stable stretch and noticing your sleep is off, small increases in daytime movement (on your terms, in your way) might be worth trying.
Ask about daytime activity when patients present with sleep complaints. The standard sleep hygiene conversation often misses the daytime half of the equation. For autistic patients, this data suggests the activity-sleep relationship may be stronger than in non-autistic patients, making it more clinically relevant, not less.
Be skeptical of self-reported sleep efficiency. The near-zero correlation between self-reported and accelerometer-measured sleep efficiency means clinical conversations about "how well are you sleeping?" may be capturing experience but not physiology. Objective monitoring (even consumer-grade wearables) could add useful data for treatment planning.
This study used three separate algorithms (GGIR, ChronoSapiens, and Accelerometer) to process the same raw accelerometer data, and the combined model outperformed any single algorithm. The methodological details on elastic net variable selection and the even/odd day validation split are worth studying if you work with wearable data.
The preliminary genetic analysis in the paper (rare variants in 133 clock/sleep genes from the SPARK cohort) is small but points toward an interesting direction: molecular subtyping of sleep disturbances in autism. The NOS1 gene finding, while far from definitive, connects accelerometer phenotypes to known sleep biology.
This study tracked sleep and activity on wearables for 3 weeks. Brainloot does that every day - syncing sleep, activity, and biomarker data from Oura and Fitbit into one dashboard built for neurodivergent families.
Address the root, unlock your loot.