Testing Sleep Hygiene One Change at a Time

What each one is actually for

PDSA Cycle

Plan-Do-Study-Act is a four-beat loop for finding out whether a change actually does what you think it does, on a scale small enough that being wrong costs almost nothing. Plan: state the change and — this is the beat people skip — write down the result you predict, in numbers if you can get numbers. Do: run it, and note anything odd that happens while you run it. Study: put the real result next to the prediction you wrote and look at the gap. Act: adopt it, adapt it, or abandon it, then start the next loop. It came out of W. Edwards Deming's work on industrial quality and is now standard in hospital improvement teams. The engine of it is the written prediction. Without that, 'study' collapses into remembering how you felt about it, and memory is generous to changes you have already invested in.

Sleep Hygiene

Sleep hygiene is a set of behaviours and room conditions meant to make falling asleep and staying asleep easier: get up at the same time every day including weekends, stop screens and bright light an hour before bed, keep the room cool and properly dark, cut caffeine after early afternoon, and use the bed only for sleep and sex so that lying in it stops being associated with being awake and frustrated. Two different things are going on inside that list. Some of it works on your body clock — the internal timer that decides when you get sleepy, which is trained mostly by light and by consistency. Some of it works on arousal — how switched-on your nervous system and your thinking are at the moment your head hits the pillow. And one item, the bed-for-sleep-only rule, works on plain association: the more hours you spend awake and agitated in bed, the more reliably the bed itself becomes a cue for being awake and agitated.

What you get out of pairing them

Here is the thing that makes the pairing worth your evening. Sleep advice is unusually noisy and unusually personal. Two people follow the same ten rules; one gets an extra forty minutes, the other gets nothing but a resentment about their phone. And because sleep varies enormously night to night for reasons that have nothing to do with what you changed — a hard conversation at four in the afternoon, a heavy meal, the neighbour's dog, where you are in your cycle, a mild cold you did not notice you were fighting — a single good night after a single change proves nothing at all, and neither does a single bad one. Most people run into this and draw the wrong conclusion in one direction or the other. They adopt everything, feel better for a week because they are paying attention to sleep at all, and never learn which part they could safely drop. Or they try the whole bundle, get nothing by Thursday, and file sleep hygiene under things that do not work for them. What a prediction-and-comparison loop gets you is a way of learning from noisy evidence without being fooled by it. You are not looking for a good night. You are looking at whether the change moved your average by roughly the amount you thought it would, over enough nights that one bad Tuesday cannot swing the answer.

What PDSA Cycle rests on

• Feedback against prediction corrects belief plausible mechanism
  Comparing the actual result of a change against the predicted result before adopting it corrects mistaken beliefs faster than adopting or rejecting changes based on intuition alone.
• Small scale testing precedes scaling observational studies
  Running a change through a small planned test cycle before wider adoption reveals real-world effects while limiting the risk of being wrong.

What Sleep Hygiene rests on

• Consistent schedule entrains circadian rhythm clinical practice
  Keeping a fixed sleep-wake time trains the body's internal clock to expect sleep at that time.
• Reduced presleep arousal eases sleep onset clinical practice
  Dimming light avoiding screens and cooling the room lowers the physiological and cognitive arousal that would otherwise delay sleep onset.
• Stimulus control strengthens bed sleep association clinical practice
  Using the bed only for sleep strengthens the learned association between bed and sleep rather than wakeful activity.

Where they actually meet

It is worth being straight about the shape of this pairing, because it is not the shape most of these pages have. These two are not two methods doing the same move on different material. One is a container and the other is contents. The plan-do-study-act loop has nothing to say about sleep specifically; it is indifferent to what you put in it. Sleep hygiene has nothing to say about testing; it is a list of recommendations that arrives already decided. Held that way, the honest description is that they do not converge on a shared insight — they slot together, the way a method slots into a subject. But the slot is not arbitrary, and this is where it gets interesting. Sleep is close to the ideal subject for this particular loop. It gives you a fresh data point every single day, the outcome is something you can rate in five seconds on waking, and the changes are individually small and individually reversible. Most domains where people want to run small tests on themselves — a career move, a way of handling conflict — give you a result every few months and no clean measure. Sleep hands you a nightly readout. That is why the fit is real even though the principles underneath the two practices do not overlap.

The principle underneath both

If there is a single idea that makes this combination land, it is not shared by both practices — it belongs to the loop, and sleep hygiene simply happens to be a place where it bites unusually hard. The idea is that a belief only gets corrected when you commit to what you expect *before* you find out. Write down 'I think a fixed six-forty wake time will cut how long it takes me to fall asleep from around forty minutes to around twenty-five, within two weeks.' Now you cannot quietly move the goalposts. If it lands at thirty-eight minutes, you noticed something you would otherwise have talked yourself out of noticing. If it lands at twelve, you learned that the thing is a bigger lever for you than you thought, which is worth knowing too. Sleep hygiene needs this more than most subjects, because sleep is the domain where retrospective judgement is least trustworthy. You are assessing, on Wednesday morning, from inside a mildly sleep-deprived brain, how well you slept across a fortnight you barely remember. That is not a measurement. A prediction written down two weeks ago, next to a number you recorded each morning while still in bed, is.

What that makes the real goal

Stated so it does not belong to either tradition: to know which of the things you do around sleep are actually doing something for you, and to know it with enough confidence that you can drop the rest without anxiety. That is the goal. Notice that it is not 'sleep better' — it is a knowledge goal, and it has a different finish line. You can reach it and discover your caffeine cut-off matters enormously and your screen rule does not, which means you get an hour of your evening back. You can reach it and find that none of the ten rules moves your nights, which is genuinely useful information: it points you toward the sleep problems that behaviour does not fix, such as untreated apnoea, a medication effect, or the kind of persistent insomnia that responds to a proper clinical programme rather than a checklist. Any practice that gets you to that knowledge counts as a route to the objective. The loop is one. It is not the only one.

Other practices that serve the same goal

Several other things aim at the same target, and one of them carries far more evidence than either practice here. Cognitive Behavioural Therapy for Insomnia — usually shortened to CBT-I, a structured programme of four to eight sessions — contains a component called sleep restriction that is, functionally, a tightly specified version of the same loop. You keep a daily diary, calculate the fraction of time in bed you actually spent asleep, and then move your allowed time in bed by fifteen minutes a week depending on which side of a threshold that fraction lands. Prediction, measurement, adjustment, repeat — but with the adjustment rule fixed in advance by people who ran the trials, rather than left to your judgement. It has strong randomised evidence for chronic insomnia and outperforms sleep hygiene advice given alone. If your nights are genuinely bad rather than merely improvable, that is the route to take first, and self-experimentation is the thing you do afterwards to fine-tune. Two other routes: a formal single-patient trial, sometimes called an n-of-1 trial, where you alternate a change on and off in blocks so you can see whether the effect follows the change rather than the calendar — useful for anything you can switch off, like a supplement; and simple actigraphy, a wrist tracker, which replaces your morning guess with a number that does not care how you feel about it. The tracker is a mixed blessing, and there is a real failure mode: some people get more anxious about sleep the more precisely they measure it, which is itself a cause of bad sleep.

A tradition that already does this

This has been done, and by the people you would most expect to have done it. Sleep restriction therapy, developed by Arthur Spielman in the 1980s and now a core component of the standard insomnia programme, is a titration loop with the decision rule written down: measure the ratio of time asleep to time in bed each week, and if it is above a set threshold you add fifteen minutes to your window; if it is below, you subtract. That is not a metaphor for the plan-do-study-act cycle. It is the same operation, arrived at independently in a clinical setting, and it is one of the better-evidenced behavioural treatments in medicine. Which tells you two things. First, the move is not novel, so you should not expect it to be magic. Second, and more usefully, it tells you where the sharp edge is: what makes sleep restriction work is that the threshold and the adjustment size were fixed in advance by someone other than the person who wants the answer to come out a particular way. When you run your own loop, you are both the experimenter and the person who badly wants the last two weeks to have been worth it. Writing your prediction down before you start is the cheapest available protection against that.

Using a plan-do-study-act loop to find out which sleep rules are yours

Do this rather than reading it. Pick one rule — one, not three. Say it is the fixed wake time, because it has the best evidence behind it of anything on the standard list. Write two lines in a note on your phone: 'From Monday, alarm at 06:40 every day including Saturday. I predict my average time to fall asleep drops from about forty minutes to about twenty-five, and I predict I feel worse for the first four days.' The second prediction matters as much as the first, because if you have not predicted the bad early stretch you will read it as failure and quit on Thursday, which is when almost everyone quits. Then, every morning, before you check anything, write two numbers: roughly how long you think it took you to fall asleep, and how rested you feel out of ten. Five seconds. Do not upgrade to a spreadsheet; the upgrade is where people stop. Run it fourteen days. Now sit down with the fourteen pairs of numbers and your written prediction, and look at the second week only — the first week is mostly the adjustment, not the effect. If the sleep-onset average landed within shouting distance of twenty-five minutes, you have found a lever, keep it. If it sat at thirty-eight, you have learned that this rule is not where your problem lives, and you can stop spending willpower defending it. Then start the next loop with the next rule. Two important refinements. Change one thing at a time even though it is slower, because if you change three you will never know which one did it, and you will end up carrying all three forever out of caution. And do not run a loop across a week that is not a normal week — travel, illness, a deadline, a visiting relative. A loop run through chaos returns noise that looks exactly like a result.

What sleep hygiene teaches you about running the loop itself

The traffic runs the other way too, and it corrects the most common way this loop gets misused by people testing things on themselves. Sleep hygiene's own central claim is that consistency *is* the intervention — that a fixed wake time works by training your body clock, the internal timer that decides when you get sleepy, and that training takes days to weeks rather than hours. Which means the effect you are trying to measure does not exist yet during the period most people would call their test. If you run three days and study the result, you have measured the disruption, not the change. Sitting with that fact will fix how you run every other loop you ever run on yourself, in any domain. The question 'how long does this thing need before there is anything to see?' has to be answered in the planning beat, not discovered in the study beat. Sleep hygiene also hands you a second correction, this one about what you measure. The bed-for-sleep-only rule works by weakening a learned association, and associations weaken unevenly — you get a stretch of nights where nothing has changed, then it goes. If you are averaging across a fortnight you will miss that shape entirely. So plot the fourteen points rather than averaging them, and look at where the line bends. A third thing, less obvious and worth more: the standard sleep hygiene list already tells you which of its items are supposed to work fast and which are supposed to work slowly. Room temperature and light are same-night effects. Schedule and stimulus control are slow ones. That distinction gives you your run length for free, before you have collected a single night's data — which is precisely what the planning beat is supposed to produce and what people writing their own plans almost never manage.

The Sunday night you realise you have been obeying eight rules for a year

It is Sunday, quarter past eleven. You have done everything right for about a year — phone out of the bedroom, blackout blind, thermostat down to eighteen, nothing caffeinated after two, the whole list — and you are lying there entirely awake, doing the arithmetic on how many hours are left before the alarm. And what actually surfaces is not despair about sleep. It is a small, irritated question: which of these am I even doing for a reason? You cannot answer it. You adopted the list all at once, in a bad fortnight, from an article. So on Monday you write two lines in a note. 'Keeping everything the same except caffeine — moving my cut-off from two in the afternoon back to eleven in the morning, for two weeks. I predict it takes fifteen minutes off how long I lie there. I predict I am foggy and slightly annoyed until about Thursday.' Every morning after that you add two numbers to the note before your feet touch the floor: minutes-to-sleep, best guess, and rested-out-of-ten. Twelve seconds a day. Days one through four are exactly as predicted and mildly unpleasant, and because you wrote that down in advance it reads as the plan working rather than as evidence you were wrong. Day nine is terrible — a work thing at nine at night — and you note it and move on, because one night was never going to be the answer. Then on the Sunday two weeks later you look at the second seven days next to your prediction. Average around twenty-two minutes, down from thirty-eight. Larger than you guessed. That is your lever, and now you know it is yours rather than a thing you read. And here is the part you did not expect: the following month you run the same loop on the screen rule, and it moves nothing at all. Nothing. So you give yourself your evenings back and read on your phone in bed again, without the low background sense that you are cheating. Two loops, six weeks, one rule kept with confidence and one dropped with confidence.

When this is worth reaching for

You'll know this is the right audit to run when you're already sleeping tolerably and trying to work out which of your habits are earning their keep, rather than dealing with sleep that's genuinely broken. A few specific situations make it worth doing. You adopted the whole list at once during a bad patch and have never separated the parts since — that's the classic case, and the audit usually returns two or three items doing real work and a tail of items that aren't. Something in your life is about to change the constraints, so the old bundle won't survive anyway: a new baby, shift work, a partner with a different schedule, moving somewhere with a hotter bedroom — knowing which two rules are load-bearing tells you what to protect and what to let go. The routine has started to feel controlling rather than helpful, and the anxiety about keeping it is plausibly costing more than the routine returns. Or someone has told you to do something and you can't tell whether it works — a supplement, a specific mattress, a particular wind-down sequence — because a two-week loop with a written prediction is a cheap way to settle it, and the alternative is carrying the thing forever on the strength of the first good night you happened to have after starting it.

Where this does not apply

Do not run self-experiments on your sleep if your sleep is bad enough to be doing damage. If you are falling asleep during the day when you do not intend to, if you have been told you stop breathing or gasp at night, if you have had months of the kind of insomnia that has started to reorganise your life around it, or if you are on medication that plausibly affects sleep — the loop is the wrong tool at the wrong moment, because it is slow, it is low-powered, and it will keep you occupied for months on a problem that needs assessment. Go and get assessed, then run loops afterwards on what is left. A second caution, less obvious. For some people, measuring sleep makes sleeping worse. There is a recognised pattern where close attention to sleep quality feeds the anxiety that is causing the poor sleep in the first place — clinicians sometimes call the extreme version orthosomnia, meaning the sleep problem is now being driven by the pursuit of perfect sleep. If you notice you are lying awake thinking about what number you are going to write in the morning, stop measuring. The loop is not worth the cost, and the honest move is to abandon it. Finally, some things simply cannot be run through this. If you cannot control your wake time because of shift work, or you cannot control the light or noise because of where you live, testing the rule tells you nothing except that you could not run the test.

What the evidence does and does not support

The two sides here carry very different kinds of backing, and it is worth being exact about it. The plan-do-study-act cycle has almost no randomised evidence as a method — its standing comes from decades of use in industrial and healthcare improvement, and a well-known 2014 review in BMJ Quality and Safety found that most published applications of it did not actually follow the cycle they claimed to, which tells you the idea is easier to endorse than to run. Sleep hygiene is in a stranger position than most people realise: as a standalone package it is the thing that regularly serves as the comparison condition in insomnia trials, and it regularly loses to the fuller cognitive-behavioural programme. Individual components hold up much better than the bundle — fixed wake time and stimulus control have good clinical support, the screen rule has weaker and more contested support than its cultural prominence suggests. So the evidence pattern is: a method with thin formal backing but sound logic, applied to a bundle whose parts are unequally supported, which is precisely the situation where testing them one at a time is worth doing. As for the pairing itself: nobody has run it. To test it you would need to randomise people to receive sleep hygiene as a bundle versus one component at a time with a written prediction and a fourteen-day window, and measure both sleep outcomes and, more interestingly, whether the one-at-a-time group was still doing the useful components six months later. That study does not exist, so what is on this page is a way of organising your own evidence, not a result.