When the Lights Go Out

Somewhere near you there is a building with a mental health service inside it. Tonight, at about five o'clock, the last person in it will put on a coat, set an answerphone, and go home to their own family. This is not a scandal. It is a working day ending, the way working days do.

The difficulty is that nothing inside the people on that service's list ends at the same time.

This issue is about the hours after that door closes. Not the drama of them. The arithmetic of them: what is actually open, what is actually known, and what happens in the specific, unglamorous stretch between the last bus and the first light, when a person is awake and the institutions that exist for them are not.

Our last issue argued that a mood disorder is a swing nobody bought a ticket for, and that what decides its damage is whether anyone notices. This one narrows to the hardest case of that argument. At two in the afternoon, noticing is possible. There are colleagues, receptionists, a number that gets answered. At four in the morning there is a ceiling, a phone, and whatever a person has already been told to do.

I want to be careful here, because this subject attracts a particular kind of writing. There is a familiar paragraph about the terror of three in the morning, and it is usually built out of atmosphere rather than evidence. We went looking for the evidence instead, and the first thing we found was that the most repeated claim in the entire field is wrong.

So this issue does what the last one did. It argues, and it shows its working. Every figure carries the body that published it, the period it covers and the population it describes. Where a number everybody repeats turned out to be untraceable, or true of a different country, or true only after an adjustment nobody mentions, we have said so and left it out. There is a list of those at the back. It is longer than the list of things we could prove.

That is not a weakness in the reporting. It is the finding.

You will have read that suicides peak at three in the morning. It is in the features, the podcasts and the awareness copy. It is not true, and the way it became untrue is worth more than the claim ever was.

Start with the harder problem underneath it. In this country we do not know what time people die by suicide, because nobody writes it down.

Hold that for a moment, because it has a consequence. Any British statistic you have ever seen about suicide and the small hours was not derived from British national data, for the simple reason that the data does not exist. It came from somewhere else, and usually from the United States.

What the American data shows is genuinely striking, and it is not what the headline says. Researchers took nearly eighty thousand suicides and asked a better question than the usual one. Not how many happened at each hour, but how much risk each waking hour carried, once you account for the fact that almost everyone is asleep at three in the morning and almost nobody is asleep at noon.

And now the sentence that almost never travels with it.

This matters beyond pedantry. A service planner reading the popular version moves resources to the hours with the highest counts and concludes the night is quiet. A service planner reading the actual finding understands that the night is thin: fewer people awake, each of them at considerably higher risk, and almost nothing open.

Where Britain does keep records, they point the same way. The largest study of self harm presentations in English hospitals looked at more than ten thousand episodes across Oxford, Manchester and Leeds and found the commonest time of arrival was between ten at night and two in the morning. The data is from 2000 and 2001, which we state plainly because it is old, and because nobody has published a better British series since.

So the demand is nocturnal, the risk per waking hour is nocturnal, and the provision, as the next pages show, is not.

The folk explanation for why four in the morning feels worse is that you are tired. The evidence says something stranger and more precise: the hour itself is doing some of the work, independently of how long you have been awake.

This was established by an experiment with an unpleasant premise. Volunteers were put on schedules of twenty eight and thirty hours, long enough that their body clocks came loose from the clock on the wall, so that the researchers could separate two things that ordinarily move together: how long a person has been awake, and where they are in their own circadian cycle.

That is the honest core of the thing, and it is enough. It says the feeling arriving at four in the morning is not a failure of character or a lapse in perspective. It is a physiological state with a timetable, and it will pass on that timetable whatever you conclude about your life while it is happening.

Around that core sits a much larger claim, and we are going to mark exactly where the evidence stops. In 2022 a group of sleep researchers proposed what they called the Mind After Midnight hypothesis: that being awake during the biological night shifts attention toward the negative, flattens reward, and loosens the brain's ordinary restraint, and that this combination is what makes nocturnal wakefulness dangerous.

What is far better established is the direction of travel between sleep and illness, and it runs both ways.

Read that as a relative risk, not an absolute one. These are multiples of a small starting number, in populations, over years. It does not mean one insomniac in two becomes depressed, and a publication that lets you believe it does has misled you while quoting a real study.

The useful half of the finding is that it runs in the other direction too, which is where the treatment evidence lives.

And the limit of it, stated as plainly as we can.

In 2019 the health service in England made a promise about exactly these hours, and the promise was unambiguous.

The route itself now exists. In August 2024 the health service announced that the mental health option on 111 was available everywhere in England, which is a real achievement and worth saying so. The question this issue asks is what happens when the number is dialled.

Roughly one call in four to the crisis line ends before anyone speaks. We do not know what hour those calls were made, because the published collection contains no time of day field. That absence is now a pattern: it is the third time in this issue that the obvious question cannot be answered because nobody records the hour.

Behind the phone line sits the team, and this is the distinction the section turns on. Researchers reached 233 crisis teams, about ninety five per cent of those mapped in England, and asked what they could actually do in the middle of the night.

We want to be fair about that gap. The direction of travel is upward and the people staffing these teams are not the problem. But the difference matters enormously at the other end of the line. A voice on a telephone can assess, advise, de-escalate and stay with you. It cannot open your door, sit in your kitchen, or take you anywhere. At two in the afternoon that difference is academic. At four in the morning it is the entire question.

For those who do get through to a service, the standard is four hours for a very urgent referral. That standard is moving in the wrong direction.

Which leaves the charities, who never closed, and who are the reason this issue is not bleaker than it is. Samaritans has been answering a telephone through the night since 1953, with around twenty three thousand volunteers across more than two hundred branches. Their own annual report is candid about what that now costs them.

Put it together and the shape is plain. Risk per waking hour is highest when fewest services run. Presentations cluster either side of midnight. One crisis team in three cannot come to you, one call in four to the crisis line ends before it is answered, and three in ten calls to the country's oldest night line go unanswered too. None of this is villainy. It is what a system built around office hours does when the thing it treats does not keep them.

This is the part of the issue where a publication owned by a company writes about that company's product, so the reader is entitled to know exactly how much weight to put on it. The answer is: less than the preceding pages, and we will be specific about why.

Elysium is a conversational mental health companion built by The Threshold Collective. We built the model and we own it. It talks, it listens, and a safety layer sits in front of every exchange, watching for the kinds of things a person says when they are in danger and responding to those first, before anything else in the conversation is attended to.

Here is the honest description of what it is for. It is not therapy. It is not a diagnosis. It is not a clinician, not a crisis team, and not an emergency service, and it is not a replacement for any of them. It cannot come to your door either. What it can do is the one thing the rest of this issue has been describing the absence of: it is awake at four in the morning, and it does not have a coat to put on.

It is also closed to the public at present while we move it onto new infrastructure, and we would rather tell you that than write this page as though it were open. What we claim for it is narrow. The gap in these pages is structural: human services are staffed by humans, who must sleep, and the demand does not respect that. Something awake in those hours does not solve it. It can sit in the gap, reduce the hours a person spends entirely alone in it, and point them firmly at the people who can act. That is the whole claim, it is smaller than this industry usually makes, and it is the only one the evidence supports.

The most useful sentence in this entire issue is not a statistic. It is the observation, from the one experiment that separated the clock on the wall from the clock inside a person, that mood tracks circadian phase.

Translated out of the journal, it means this. What you are feeling at four in the morning is being produced, in part, by the hour. The hour is going to end. It ends every single time, and it has never once failed to end, and the person you will be at nine o'clock will read the conclusions you reached at four and not recognise them.

That is not encouragement. It is a finding, and it is the reason the oldest piece of advice about this is also the best: do not decide anything now. Not because your reasoning is weak, but because the instrument you are reasoning with has a known nightly fault, and it self corrects on a schedule you can rely on.

Dostoevsky, who spent four years in a Siberian prison and knew rather more about endurance than most people who write about it, put a thought in the head of a character waiting to die. He imagined a man who is told he must stand on a square yard of space for the rest of his life, for a thousand years, for eternity, and concludes that he would rather live so than die at once. Only to live, to live and live. Life, whatever it may be.

It is not a comforting passage in its context and we are not going to pretend it is. But it names the only argument that matters at four in the morning, which is not that things will improve. It is that the hour is not permanent and you are not required to settle anything inside it.

The practical version, since this publication prefers those. Ring one of the numbers on the next page, which are staffed right now, tonight, by people who are awake on purpose. Tell someone in the morning what the night was like, because the pattern is the thing a clinician needs and nobody remembers it accurately a week later. Write it down while it is happening if you can. And if you are reading this at a reasonable hour, in good health, on behalf of somebody else: ask them what their nights are like. Almost nobody asks. It is the single question in this issue that costs nothing.

The lights go out every evening in every mental health building in this country, and they will go out again tonight. That is not going to change. What can change is how many people are alone in the hours afterwards, and how quickly somebody notices in the morning.