When sleep pressure builds but sleep won't come
Somewhere between 40% and 93% of autistic people report sleep problems. The most common complaint isn’t difficulty staying awake. It’s the opposite — being exhausted but unable to fall asleep.
New research from West Virginia University reveals the mechanism behind this paradox. Using mouse models of autism, researchers found that sleepiness builds normally but the ability to actually enter sleep breaks down.
Here’s what happened. BTBR mice — a strain that models idiopathic autism — were kept awake for six hours. During this forced wakefulness, their brains accumulated sleep pressure exactly like typical mice. The biological markers of tiredness increased at the same rate. The drive to sleep built up properly.
But when allowed to sleep again, BTBR mice took significantly longer to fall asleep than control mice. And when they finally did sleep, they spent less time in REM — the sleep stage crucial for processing and consolidation.
The system registered exhaustion. It just couldn’t access rest as easily.
This mirrors clinical insomnia in autistic people. Not the “I’m not tired” kind. The “I’m desperately tired but my body won’t let me sleep” kind. The kind where sleep pressure exists but the gate won’t open.
Researchers measured sleep pressure through brain wave patterns during deep sleep. When you’re tired, these waves become slower and stronger — like your brain is saying “I need proper rest, now.” BTBR mice showed these patterns building normally during wakefulness.
The drive to sleep was there. The ability to initiate and maintain sleep was not.
Interestingly, fmr1-KO mice — which model Fragile X syndrome, the most common genetic cause of autism — showed no sleep problems at all. Normal sleep pressure. Normal ability to fall asleep. Normal recovery.
This heterogeneity matters. Not all autism-related conditions affect sleep the same way. The mechanisms differ even when the diagnostic category overlaps.
The endocannabinoid system — the broken oscillator behind the paradox
Your brain produces its own cannabis-like molecules called endocannabinoids. These aren’t recreational compounds. They’re signalling molecules that regulate everything from appetite to pain to sleep.
In neurotypical brains, endocannabinoid levels oscillate across the 24-hour day. They rise and fall in patterns that align with sleep and wake states. This oscillation acts like a gatekeeper — modulating when your brain is receptive to sleep signals versus when it maintains wakefulness.
Think of it as a biological dimmer switch. During wake periods, the switch stays brighter. During sleep periods, it dims. The transitions happen smoothly because the oscillation is coupled to your actual state.
In autism models, endocannabinoid oscillation uncouples from sleep and wake states.
Previous research from the same lab showed that endocannabinoid patterns differ across the day in both BTBR and fmr1-KO mice compared to typical mice. The molecules are present. They’re just not rising and falling in sync with when the animal is actually asleep or awake.
Endocannabinoids work by dampening inhibitory signals in the brain. Inhibition is essentially your brain’s brake system — it prevents neurons from firing constantly and helps regulate when different brain regions are active versus quiet.
During sleep, inhibition typically increases in sensory processing areas. This helps your brain tune out external stimuli so you can maintain sleep rather than waking at every sound or movement.
When endocannabinoid oscillation breaks, this inhibitory modulation stops tracking properly with sleep and wake. The brake system still exists. It’s just not applying and releasing at the right times.
Both BTBR and fmr1-KO mice showed disrupted endocannabinoid patterns. But only BTBR mice developed the insomnia phenotype.
This suggests endocannabinoid disruption is necessary but not sufficient for autism-related insomnia. Something else — some additional mechanism interacting with the broken oscillator — determines whether sleep access actually fails.
The dissertation attempted to track endocannabinoid levels in real-time during sleep and wake using fluorescent biosensors. Technical issues prevented clear results, but the question remains crucial: if the oscillation is uncoupled, what does the actual pattern look like? Random? Inverted? Stuck?
Why autistic sleep problems appear under load
Under normal conditions, both BTBR and fmr1-KO mice looked fine. They slept the right amount at the right times. Their sleep cycles followed typical patterns. Baseline measurements showed nothing obviously wrong.
The insomnia phenotype only emerged when researchers challenged the system via sleep deprivation.
This is the critical methodological insight. Many autism sleep studies measure baseline conditions — how much people sleep on a typical night, whether they maintain normal circadian rhythms, whether sleep architecture looks disrupted, etc.
BTBR mice would pass all those tests. Their baseline sleep appeared completely normal. You’d only discover the problem by keeping them awake then watching what happened during recovery.
Clinical populations show the same pattern. Autistic people often report that sleep problems worsen under stress, during transitions, when demands increase, and when routines change. The system functions adequately under stable conditions but breaks down under load.
Measuring only unstressed baselines misses where capacity actually fails.
The sleep deprivation protocol used in this research — six hours of gentle handling to prevent sleep — mimics the kind of challenge that happens naturally in human life. A long day. Travel. Disrupted routine. Periods where you can’t sleep even though you need to.
For typical individuals, recovery from such challenges is predictable. You fall asleep faster than usual. You sleep longer. Your brain waves show the signature of accumulated sleep pressure being discharged.
BTBR mice showed the pressure accumulating normally. But the discharge failed. Sleep pressure built up with nowhere to go.
This suggests the problem isn’t generating the drive to sleep. It’s converting that drive into actual sleep. The insomnia isn’t about lacking tiredness. It’s about tiredness that can’t translate into rest.
What this means for autistic sleep (and why typical advice fails)
Standard sleep advice assumes the problem is sleep pressure not building adequately. Set an alarm in the morning, and stick to it each morning. Go to bed at the same time each night. Avoid screens for a few hours before bed. Create a dark, cool room. Exercise during the day. Build sleep pressure through routine and deprivation. Et cetera.
But BTBR findings show sleep pressure builds fine. The problem is access.
This explains why autistic people often report that typical sleep hygiene recommendations don’t work. The advice targets pressure accumulation when the actual failure point is somewhere else in the system. It’s not discipline. It’s not routine adherence. It’s not trying hard enough.
The mechanism is neurobiological. Endocannabinoid oscillation uncouples from sleep/wake states. Inhibitory signalling stops modulating properly. The gate that should open when sleep pressure builds stays partially closed.
Cannabis use in autistic populations makes mechanistic sense through this lens. Exogenous cannabinoids can force the gate open that endogenous regulation fails to unlock. This isn’t addressing the root cause — the broken oscillation — but it’s overriding the access failure.
Research confirms exogenous cannabinoids increase sleep time and bout length by activating the same receptors that endocannabinoids normally regulate. When your endogenous system can’t oscillate properly, external supplementation bypasses the regulatory failure.
But it also creates dependency. When you stop, your endogenous system still can’t regulate properly. The access problem remains.
The heterogeneity question remains unresolved. Why does fmr1-KO show endocannabinoid disruption without insomnia while BTBR shows both? What additional mechanism determines whether broken oscillation translates into sleep access failure?
The answer probably involves how different genetic architectures interact with the same disrupted signalling system. Fragile X syndrome involves loss of a specific protein that regulates synaptic plasticity. Idiopathic autism has no single genetic cause. The same endocannabinoid uncoupling might produce different downstream consequences depending on what else is altered in the system.
This matters for intervention development. If the target is just “fix sleep in autism,” you’ll design one-size-fits-all approaches that fail most people. If the target is “identify which subgroups have which specific mechanisms, then match interventions accordingly,” you might actually help people sleep.
The research reveals something else crucial for studying autism more broadly. Baseline measurements under optimal conditions will systematically miss phenotypes that only emerge under challenge. If you want to understand where systems fail, you need to stress-test them.
As long as we only measure unstressed baselines, we’ll continue finding that things look mostly fine when they’re not.
Citations
Rodriguez, B. (2025) — Sleep, Endocannabinoid Signaling, and Cortical Processing in Mouse Models of Autism Spectrum Disorder
Bridi et al. (2020) — Daily Oscillation of the Excitation-Inhibition Balance in Visual Cortical Circuits
Bridi et al. (2025) — Daily oscillation of the excitation/inhibition ratio is disrupted in two mouse models of autism
