ARTICLE
Sleep restriction therapy: a 4-week guide
The most counterintuitive insomnia protocol, the one most likely to work, and the one most people quit before it starts working.
If you're reading this, you've probably already had the thought: you want me to sleep less? That reaction is rational. The protocol is counterintuitive. It is also one of the two evidence-strongest components of CBT-I, paired with stimulus control.
What follows is the actual protocol, with the calculations, not a summary. Four weeks. A wake time that doesn't move. A bedtime that does. A 5-hour floor that is non-negotiable. And a sleep-efficiency target of 85% that decides when the window expands.
What sleep restriction actually is
The premise is that chronic insomnia involves spending too much time in bed trying to compensate for poor sleep. Hours nine through twelve in bed don't become more sleep — they become more lying-awake time, and lying awake in bed is the thing that built the broken bed-sleep association in the first place.
The name is misleading. You aren't restricting sleep. You're restricting the opportunity to sleep — narrowing the window of time you allow yourself in bed, so that whatever sleep does happen has to consolidate inside it. Most people who sleep five hours scattered across nine hours in bed end up sleeping more, or at least more deeply, when forced into a six-hour window.
Mechanistically: it builds homeostatic sleep pressure, narrows the bed-sleep cue, and eliminates the lying-awake reinforcement loop. The protocol is usually run together with stimulus control — they address adjacent mechanisms in the same dysfunction, and the combination outperforms either alone. If you want the broader picture of where this sits among the three insomnia patterns, our pillar guide on insomnia covers them.
The calculation: finding your sleep window
The protocol begins with a measurement, not a target. Spend the first week tracking — a pencil and a notebook are better than an $800 ring for this, and we'll get to why in a minute.
Each morning, write down four numbers: when you went to bed, roughly how long you took to fall asleep, roughly how many times you woke during the night, what time you got up. You don't need precision. The protocol doesn't trust the second decimal place anyway.
After seven nights, calculate Total Sleep Time (TST) — your actual sleep, not your time in bed — then average it across the week. That number, with one important caveat, is your starting sleep window.
The caveat is the floor. Never go below 5 hours, regardless of what the tracking gave you. If your average TST is 4.5h, your starting window is 5h. If it's 3h — which happens — your starting window is still 5h. The floor exists because below 5 hours, cognitive and emotional regulation degrade in ways that make the protocol unsustainable. You won't hold it. The medicine has a minimum dose.
Then pick a fixed wake time you can sustain seven days a week. Anchor everything to wake time, not bedtime. If you want to wake at 6:30am and your starting window is 6h, your bedtime is 12:30am. Yes, that feels late. Yes, you are going to be tired. That is the medicine.
Sleep efficiency: the metric that drives the protocol
The metric that drives this protocol is sleep efficiency — total sleep time divided by total time in bed, as a percentage. Eight hours of sleep in eight hours of bed is 100%. Five hours of sleep in nine hours of bed is 55%. The threshold that matters is 85%.
You don't expand the sleep window until your weekly average efficiency hits 85%. The metric tells you when the body is ready for more. Expand too early and you reintroduce lying-awake time; the curve flattens; you go backward.
How to track without becoming obsessive: rough estimates, not minute-by-minute logs. The same notebook from the measurement week. A sleep tracker is the wrong tool here. Orthosomnia — anxiety about the tracker's numbers — is a documented effect that undoes the protocol, and consumer trackers misclassify sleep stages routinely. The second-decimal-place precision they give you isn't more accurate. It's just more anxiety-producing.
Week 1: the hard week
Week one is the worst week. There is no version of the protocol where it isn't. Daytime sleepiness is significant. Concentration runs slower. Mood often drops. This is sleep pressure building, which feels exactly like sleep deprivation because at this stage it is sleep deprivation — just deployed deliberately, on the right side of a four-week ledger.
Most people who quit this protocol quit during week one. They quit at the exact moment the sleep pressure is being built. Two or three more nights and the curve starts bending. Most never get there.
Practical guardrails. If your job involves driving long distances, operating heavy machinery, or making high-stakes safety decisions, do not run this protocol unsupervised — wait for a stable stretch or do it with a clinician. If your job is normal sedentary office work, you'll function on coffee and adrenaline. Many people have. So have I.
The single rule that holds the whole thing together: no matter how tired you are, no matter how much you accidentally dozed on the couch at 8pm (you will, once or twice), do not go to bed before your scheduled bedtime. The pressure is the medicine. Going to bed early consumes the pressure and resets the curve.
Week 2: titration begins
At the end of week one, calculate the week's average sleep efficiency. The number determines what week two looks like.
If efficiency is 85% or higher, extend the window by 15 minutes — move bedtime 15 minutes earlier, leave wake time fixed. If efficiency is between 80 and 85%, hold the window steady another week. If efficiency is under 80%, contract the window by 15 minutes — move bedtime later. This last case is rare. It still happens.
The 15-minute increment is the boring choice. People who feel better at the end of week one want to jump 30 or 60 minutes. Don't. The slow titration is what locks the new association in. Aggressive expansion regresses the gains within a week, and you've lost the time you put in.
The pattern repeats. Each Sunday: calculate the week's average efficiency, adjust the window by 15 minutes one way or the other, hold for another week. Most people add 15 to 30 minutes per week through weeks two and three. The math feels fussy. It works because the body is learning to trust the new schedule, and trust takes more nights than impatience wants to allow.
Weeks 3-4: consolidation
By week three, most people are sleeping six to seven hours per night with high efficiency inside a 6.5- to 7.5-hour window. The daytime fog of week one has cleared. Cognitive function returns. Mood lifts back toward baseline, often above it.
The temptation now is to expand faster. Resist. The protocol ends in one of two ways. Either you've titrated up to a window that gives you 7 to 8 hours of high-efficiency sleep and the protocol's job is done. Or you've stabilized at a window shorter than 7 hours — 6.5, say — with high efficiency and good daytime function, and that turns out to be your actual sleep need.
The cultural assumption that everyone needs 8 hours is an average over a population. Individuals vary widely. Some people are biologically 6.5-hour sleepers. They've spent years forcing themselves into 8-hour bed-windows and calling the resulting lying-awake 'insomnia.' The protocol reveals this. It's one of the quieter outcomes.
If you reach week four and your efficiency is still under 80%, the diagnosis might not be straightforward sleep-onset or sleep-maintenance insomnia. Time to see a sleep specialist. Common alternatives: untreated sleep apnea, circadian rhythm misalignment, primary anxiety driving the sleep problem. See our medical disclaimer for guidance on when professional evaluation is the right next step.
Common failure modes
Most articles stop at the protocol. The failures happen after the protocol starts. The four ways week one breaks:
Napping
One 20-minute nap on the couch eats two hours of sleep pressure. Two short naps cost a night. Set alarms on your phone for the hours you know you'll be vulnerable — usually 3-5pm and 8-9pm. Plan a coffee around 2pm. Bright light helps. Cold water on the face helps. Couch is the enemy.
Weekend drift
The fixed wake time is not a Monday-through-Friday thing. Sleeping in to 9am on Saturday after a 6:30am weekday wake is the same as crossing a time zone every weekend — it erases days of titration. The rule is seven days. There is no kind way to put this; weekend lie-ins are why the protocol stalls.
Ignoring the 5-hour floor
People with 3h or 4h baseline TST sometimes try to start at 3h or 4h windows. Don't. Below five hours, your judgment about whether the protocol is working degrades, which leads to bailing on it. The floor exists to keep you on the protocol long enough for it to work.
Doing this during a high-stress life period
Sleep restriction during a divorce, a startup launch, grief, or a major work crunch is setting yourself up to quit by night four. The protocol needs four weeks of relative life stability to land. If life is currently a mess, wait six weeks. The protocol will still be there.
What this isn't going to fix
Sleep restriction is the homeostatic-pressure tool. It rebuilds the bed-sleep association by forcing the sleep into a smaller, denser window. For sleep-onset and sleep-maintenance insomnia, it's one of the strongest behavioral interventions available, and it pairs naturally with stimulus control therapy — most CBT-I programs run them together. If you're choosing between this protocol and prescription medication, our CBT-I vs sleeping pills comparison covers the head-to-head.
It will not fix circadian rhythm misalignment, where the issue is timing rather than duration. The most commonly misdiagnosed version of this is delayed sleep phase disorder — if sleep restriction made you feel worse and your real issue is a late-running clock, that article is the right next read. It will not fix sleep apnea, which is a mechanical problem of the airway. It will not fix severe anxiety disorders that are primarily driving the insomnia rather than being driven by it. If you've run the protocol carefully for four weeks and efficiency has not moved, that's a strong signal the diagnosis isn't straightforward insomnia — see our medical disclaimer.
Sleep restriction is usually paired with stimulus control therapy, the other behavioral CBT-I component. They address adjacent mechanisms; most clinical CBT-I programs run them in parallel.
If your specific issue is waking at 3am rather than failing to fall asleep, our piece on what to do when you can't sleep at 3am covers the middle-of-night version of the same protocols.
Seven worked weekly schedules — sleep-onset vs sleep-maintenance subtypes, shift work, older adults — are at sleep restriction therapy schedule examples.
If DIY runs out of steam, the structured digital programs are compared in best CBT-I online programs.
When digital programs aren't enough, online sleep doctor covers the telehealth options.
If 3am cortisol is part of the picture, 3am cortisol awakening covers the biology.
Tracking adherence during SRT is its own decision — sleep tracker comparison covers which devices help and which feed the loop.
Frequently asked questions
Is sleep restriction safe?
For most adults, yes — when followed correctly, with the 5-hour floor held strictly, and away from high-stakes work during the build phase. It is not safe for: pregnancy, untreated bipolar disorder (sleep deprivation can trigger episodes), untreated sleep apnea (the protocol can worsen apnea before improving the insomnia), or occupations requiring acute safety vigilance. If any of these apply, do the protocol with clinician supervision, or wait.
What if I can't stay awake until my new bedtime?
This is what success looks like, not what failure looks like. If you genuinely can't keep your eyes open at 11:30pm and your scheduled bedtime is 12:30am, the sleep pressure is building exactly the way the protocol wants. Stand up. Walk around the block. Do something boring that requires being upright. Hold the bedtime. Going to bed early consumes the pressure that's about to make tomorrow night work.
Can I do this alongside sleeping pills?
The protocol still works on benzodiazepines and Z-drugs, but the data is noisier and the medication can mask the efficiency signal you're trying to titrate against. The cleaner version is to taper off the medication with your prescriber first, then run the protocol. The taper itself is not something to do alone — discuss it with whoever prescribed.
Should I use a sleep tracker?
No. Consumer trackers misclassify sleep stages routinely, so the precision is largely false, and the anxiety the numbers produce in people already anxious about their sleep is well-documented as orthosomnia. A rough log on paper gives you the signal the protocol needs — sleep latency, awakenings, total — and stops where false precision begins.
What if it works and then I drift back to old patterns?
Relapse is common. Life happens — a stressful stretch, a vacation, a few weekends of drift, and the bed-sleep association loosens again. The protocol can be run a second time, and usually the second run is faster than the first, because the body remembers what it learned and most of the time-in-bed habits that built the original problem have been replaced.