Actually, No: Counting Sheep

Counting sheep is the only sleep technique your grandmother, your GP and a cartoon lamb could all have agreed on. It is also, if you have read a newspaper at any point since 2002, the one that science has formally thrown out. You will have seen the line: researchers put counting sheep up against imagining a waterfall, and the waterfall won by twenty minutes.

That study exists. We went and read it. There are no sheep in it.

The belief

Count sheep and you will drop off. One after another over a gate, in order, until the counting takes the wheel and you are gone before you notice.

It deserves more respect than it gets, because it encodes something the research has since confirmed. Lying awake is very often a problem of the mind still being at work, and the folk remedy’s instinct is to give that mind a small, dull, harmless job so it stops doing the large one. Whoever first said it had noticed the right thing. The argument is only about whether counting is any good at it.

The verdict

Hyped, and unusually, hyped from both ends at once. The folk version oversells counting, which has never been shown in a controlled trial to help anyone fall asleep. The debunking oversells the debunking, because the single study cited as proof of failure did not test counting at all.

What the famous study actually did

Harvey and Payne, 2002, in Behaviour Research and Therapy. Forty-one people with insomnia, one experimental night each, randomly assigned to one of three instructional sets before bed: distract yourself using imagery, distract yourself by whatever means you like, or no instruction at all.¹

The imagery group was asked to picture a scene that was interesting and pleasant and absorbing enough to hold them. That group reported falling asleep faster than the group given nothing. The general-distraction group, who chose their own methods, did not, and the researchers’ prediction that vague distraction would actively make matters worse was not supported either.

Nowhere in the design is there an arm called counting sheep. Some people in the free-choice group may well have counted something. The paper does not say, because it was never the question. The twenty-minute figure that has followed the study around for a quarter of a century is attached to a comparison nobody ran. The sheep-versus-waterfall framing was added somewhere on the way to the newspapers in January 2002.

The study itself is also thinner than its reputation suggests. Forty-one participants, a single night, and a sleep onset the participants reported themselves rather than one anybody measured in a lab. What it supports is modest and worth having: a specific, engaging imagery task shortened self-reported sleep latency compared with no instruction. What it does not support is the sentence everybody quotes.

The kernel is real

Underneath the folklore there is a genuine, measurable finding. What is running through your head at bedtime does predict how long you take to fall asleep, and you can show it without asking anyone’s opinion of their own night.

Wicklow and Espie put a voice-activated tape recorder beside twenty-one people for three consecutive nights, sixty-three nights of data in total, and had them speak their spontaneous pre-sleep thoughts aloud as they occurred. Sleep latency was measured by actigraphy rather than by asking. Two categories of thought stood out as among the strongest predictors of how long sleep took to arrive: thinking about sleep itself, and thinking about the consequences of sleeping badly.²

That is a small sample, and it is correlational. It shows that the target the folk remedy is aiming at exists. It does not show that any particular technique hits it.

Counting has its own quiet history

Something very close to counting was not laughed out of the clinic. It walked in through the front door. In 1991 a group at Cambridge proposed articulatory suppression — silently repeating a simple word over and over, to a rhythm — as an insomnia technique, on a sensible theoretical footing borrowed from working-memory research. If intrusive thought is largely verbal, occupying the verbal channel should crowd it out.³

The evidence behind it never grew up. It amounts to an uncontrolled preliminary case series and a single-case experiment set against other techniques, and the authors were careful to describe their own findings as preliminary. Their indications were mainly about staying asleep rather than getting to sleep in the first place. No randomised trial ever followed. Thirty-five years on, that is still where the counting evidence sits.

What does have a trial behind it

The closest thing to a proper test of the deliberate-refocusing idea is Gellis and colleagues in 2013. Fifty-one university students with poor sleep were randomised either to a single session of cognitive refocusing treatment plus sleep hygiene, or to sleep hygiene on its own. The refocusing instruction is essentially the grown-up version of the sheep: choose material that is interesting enough to hold your attention and boring enough not to wake you up, and return to it when your mind wanders off.⁴

At one-month follow-up the refocusing group had improved significantly more on insomnia severity. The measure that was supposed to explain why — a drop in cognitive arousal before bed — moved in the right direction without reaching significance. The authors called their result preliminary support and they were right to. One session, self-report outcomes, students rather than a clinical population.

And then the result that spoils the tidy story

If vivid imagery were the active ingredient, it should beat the alternatives. Digdon and Koble tested that, more or less by accident. Forty-one students were randomised by email to one of three week-long bedtime exercises: constructive worry, in which you write down your worries and a next step for each earlier in the evening; imagery distraction; or a gratitude exercise.⁵

All three reduced pre-sleep worry and arousal and improved sleep against the participants’ own baseline week. None of them beat the others. It is a pilot study with no untreated control and online diaries for outcomes, so it settles nothing on its own. It does suggest that the mechanism people reach for is the wrong one. Richer imagery may not be what does the work. Giving the mind a defined, non-threatening job for the last twenty minutes probably is, and there appear to be several ways to do that.

The ceiling on all of this

A 2020 systematic review in Sleep Medicine Reviews went through the whole literature on deliberately managing what you think about at bedtime: mental imagery, hypnosis, paradoxical intention, articulatory suppression, ordinary thought suppression, distraction. Its conclusion about their effects was that they are “either beneficial, negligible, or detrimental”.⁶

Effects running in all three directions, after decades of work, is the honest state of play. One note of disclosure, since we are leaning on this paper: one of its authors develops and sells an app built on this family of techniques. It does not change what the review found, and it is the sort of thing you should be told.

The part that is quietly remarkable

The same review sets out what an ordinary, uneventful sleep onset looks like from the inside. As you go under, higher-order thinking switches off, and sensory imagery starts arriving on its own: fragments of scene and sound and movement that you did not compose and cannot steer. Hypnagogic imagery, in the literature. The pictures your brain makes when nobody is driving.⁶

Set against that, the pre-sleep thinking reported by people with insomnia is planning, problem-solving and rehearsing how the day might have gone differently. Daytime work, in a body that had begun to shut down for the night.

Which puts the whole business in a slightly different light. The techniques with any evidence behind them are the ones that imitate what your mind was about to start doing anyway. You cannot order sleep to arrive, though you can rehearse its opening move.

The study that supposedly debunked counting sheep never tested counting sheep. What it tested was one specific, absorbing scene, set against a vague instruction to distract yourself and against no instruction at all.

What this might mean for your evening

If you count sheep and you like it, nothing here is a reason to stop. Nothing in this literature suggests it does harm, and a technique you will actually use at half past eleven beats a better one you will not.

What is worth borrowing is the shape rather than the sheep. The instruction that has done best under test is to give your attention something specific, absorbing and mild. Specific, because a vague intention to think about nothing leaves the floor open. Absorbing, because attention that is only half spent will wander back to tomorrow. Mild, because anything with stakes in it will wake you up. Recalling a walk you know well in real detail qualifies. So does replanting a garden in your head. So, on the evidence, does writing your worries down at eight o’clock with a next step beside each, which does the same job earlier in the evening by other means.

And it is worth holding all of this loosely. The strongest study here is one session with fifty-one students and self-reported outcomes. This is a small, honest corner of sleep science, not a solved problem, and the review that summarises it says the effects go in every direction. If you would like something to keep an evening company while you find out what suits you, our interactive soundscape is free to sit with, and the rest of the Lab is where we carry on working through claims like this one.

← Back to The Lab