What is melatonin, and what does it actually do?

Ask most people what melatonin is and you get some version of "the sleep hormone". Which is close enough to be useful and wrong in one specific way, and the way it is wrong happens to matter enormously if you are the person standing in a nursery at three in the morning.

Melatonin does not knock you out. It is not a sedative your pineal gland brews up at bedtime. It is a signal, and the information it carries is not go to sleep now but it is dark, and this is roughly what time it is.

That distinction sounds academic. It is the difference between a fix that works and one that does not.

Where it comes from

Melatonin is made in the pineal gland, a structure about the size of a grain of rice sitting near the middle of your brain. It is built from tryptophan, the amino acid you get from food, which becomes serotonin, which becomes melatonin by way of an enzyme whose activity climbs steeply after dark^[1][2].

What decides when that happens is not the pineal gland itself. It is a cluster of around twenty thousand neurons in the hypothalamus called the suprachiasmatic nucleus, the master clock. The clock keeps its own time, roughly but not exactly twenty four hours, and it is corrected every day by light arriving through your eyes^[1].

So the chain runs: light hits your retina, the clock updates, the clock tells the pineal gland when it is night, the pineal gland releases melatonin, and melatonin tells the rest of your body what the clock thinks. Your liver, your gut and your immune system have no windows. Melatonin is how they find out.

The shape of a normal night

Through the day you have almost none of it. Levels sit close to the floor^[2]. Then, a couple of hours before you would normally fall asleep, production starts climbing.

Sleep researchers call that moment dim light melatonin onset, and it is the most reliable marker we have of where your internal clock actually sits. In one study of 104 healthy adults, assessed under dim light in the hours before bed, onset came at a median of 21:54, roughly two hours before their scheduled bedtime. Melatonin then stayed up for a little over ten hours, clearing around 08:23^[4].

Notice what the curve does and does not do. It does not spike at bedtime and collapse when you wake. It rises slowly, crests in the small hours, and tails off through the morning. The information sits in the timing of the whole wave rather than in the height of any one point on it.

Notice also where the peak sits. Two to four in the morning. Hold that thought.

Why "sleep hormone" is the wrong name

Here is the piece of evidence that settles it. If melatonin were genuinely a sedative, the people best placed to say so would recommend it as one. They do not.

The American Academy of Sleep Medicine, in its clinical guideline on treating chronic insomnia with drugs, suggests that clinicians should not use melatonin for sleep onset or sleep maintenance insomnia, a recommendation it grades as weak^[5]. The 2023 European insomnia guideline reaches the same place and says fast release melatonin is not recommended for insomnia^[6].

And yet the same organisation positively recommends melatonin, carefully timed, for delayed sleep phase disorder and for the non 24 hour rhythm that affects totally blind people^[7]. For jet lag a Cochrane review found nine of ten trials positive^[8].

Read those two positions together and the picture resolves. The same molecule is advised against as a sleeping pill and recommended as a clock adjuster. It is poor at making you sleepy and good at telling your body what time it is. Josephine Arendt, who spent a career on this, puts it more bluntly: in nocturnal animals, which are awake at night, melatonin still rises at night. It is a darkness hormone, not a sleep hormone^[1].

If melatonin is a clock hand rather than a sedative, then the useful question stops being "how do I get more of it" and becomes "what am I doing that moves the hand".

Light is the thing that moves the hand

Your eyes contain a type of cell that has nothing to do with seeing. They are called intrinsically photosensitive retinal ganglion cells, they contain a pigment called melanopsin, and they run a direct line from the retina to the master clock, a pathway first characterised in rats^[9] and since mapped in people^[10][11]. You cannot feel them working. They report light levels whether you are paying attention or not.

Two independent teams mapped how sensitive that system is to each colour of light. Both landed in the blue: Brainard's group put the most potent region between 446 and 477 nanometres, and Thapan's best fit came out at roughly 459^[10][11]. That is some way off where your colour vision is sharpest, which tells you it is a separate system answering a separate question.

The amount of light it takes is the part that catches people out. In one study the average person needed only about 25 lux to halve their melatonin output, which is less than most kitchens. Individuals ranged from 6 lux to 350 lux, a spread of more than fiftyfold, so some people are dramatically more sensitive than others and nobody knows which they are until it matters^[12].

Which brings us back to three in the morning

Most advice about light and sleep concerns the evening. Dim the lamps, put the phone down, that sort of thing. Useful, and not the scenario we are discussing.

The scenario we are discussing is light arriving during the peak. You are woken at three, you switch something on so you can see, and the light lands at the exact point on that curve where melatonin is at its highest.

In the same paper, a second experiment exposed people to ordinary room light, at or below about 200 lux, during the hours they would normally have been asleep. Melatonin fell by more than half in eleven of the thirteen trials. The spread was wide, from almost no effect at one end to 92% of the night’s melatonin gone at the other^[4].

For some people that is most of the night's melatonin, gone, because they needed to see a nappy.

And this is where the usual explanation for lying awake afterwards falls apart. You did not fail to get back to sleep because you are bad at sleeping, or stressed, or too old. You told your clock it was daytime. Your clock believed you.

One honest caveat, because the internet will tell you otherwise: how quickly melatonin recovers once the light goes off is not well established in humans. You will see confident numbers quoted for it. We went looking for the evidence behind them and could not find it, so we are not going to repeat them. What is well established is the suppression itself.

So should you take a tablet?

In the UK the question is partly answered for you. Melatonin is a prescription only medicine here. You cannot buy it in Boots, you cannot legally buy it as a supplement, and the gummies you have seen in American content do not exist as a legal product in Britain^[13].

The licensed adult product, Circadin, is prolonged release 2mg, indicated as monotherapy for the short term treatment of primary insomnia characterised by poor quality sleep, in patients aged 55 or over, for up to thirteen weeks^[14]. That is a narrower door than most people imagine.

Even setting the law aside, the numbers are sobering.

Pooling nineteen trials and 1,683 people, melatonin shortened time to fall asleep by about seven minutes and added about eight minutes of total sleep across a range of primary sleep disorders. The authors describe the benefit as modest, and smaller than other pharmacological treatments for insomnia^[15].

What that number is not is a reason to go looking for it. It is a prescription medicine here, the evidence base is narrower than the marketing abroad suggests, and the dose and timing are the sort of thing that needs a clinician rather than a blog. If you think melatonin might help you, that conversation belongs with your GP.

Does it really fall off with age?

You will have read that melatonin collapses as you get older, usually just before someone offers to sell you some. The evidence is shakier than the confidence of the claim.

A careful study comparing 34 healthy older adults with 98 healthy young men under controlled conditions found no significant difference in 24 hour melatonin, nocturnal peak or total output. The authors concluded the results do not support reduced melatonin as a general feature of healthy ageing^[16].

A later systematic review landed in the middle: total 24 hour production seems broadly stable in healthy ageing, though the nocturnal peak may fall somewhat in the oldest groups^[17]. Much of what looks like an age effect in the wider literature is probably illness, medication and the fact that older people often get less daylight and more light at night.

The clearest decline is actually the one nobody markets at: melatonin drops substantially across childhood and into puberty^[18].

What all of this adds up to

If melatonin were a sedative, the sensible response to a bad night would be to take more of it. Since it is a timing signal, the sensible response is to stop sending the wrong time.

You cannot top melatonin back up after a 3am hallway light has knocked it down. There is no tablet that reverses suppression that has already happened, and in Britain there is no tablet you can buy anyway. Which leaves the light itself as the part of the picture you actually control.

Which, practically, comes down to a handful of things:

• Get daylight in the morning. The clock needs a strong signal at the right time, and the strongest correction available is going outside.
• Treat the three hours before bed as a dimmer switch, not an on and off. Expert consensus suggests keeping evening light low and the bedroom close to dark^[19].
• At night, use the least light that lets you do the job. Brightness matters as much as colour. A dim light pointed at the floor beats a bright one pointed at your face.
• Prefer long wavelength light after dark. The system is most sensitive in the blue and far less so at the red end^[10][11], which is why hospital night shifts and sleep laboratories have used red lighting for decades.
• Assume you might be a sensitive one. With a fiftyfold spread between individuals^[12], your own experience beats any general rule.

That fourth point is the one we built a company around, so treat what follows as interested rather than neutral. Driftback makes two lights for exactly this moment: a motion sensor plug in light for the landing, and a bedside lamp running a true 670nm red with a warm 1700K amber setting. Genuinely long wavelength emitters rather than a red tint over a white LED, and both dim enough to be nearly nothing. Red light is not magic and it is not zero. It is just a great deal further from the peak of that melanopsin curve than the bulb in your hallway.

One last thing

There is something quietly reassuring in all this. A hormone you cannot feel, on a schedule you did not choose, getting overridden by a light switch you pressed without thinking, is not a personal failing. It is a solvable problem with a boring solution.

Nobody is going to fix 3am. The baby will still wake. But the hour of lying there afterwards, staring at the ceiling and doing sums about how long until the alarm, is not a compulsory part of the deal. That bit is often just the light.