The 3 A.M. File

1. Turn the clock away. Stop doing the math.

You already did the math once. Four hours and change. Do not do it again in ten minutes, and again in twenty.

Tang, Schmidt and Harvey ran an experiment on precisely this (2007, Journal of Behavior Therapy and Experimental Psychiatry, 38, 40 to 55). Sixty people — thirty good sleepers, thirty bad ones — were told either to watch a clock while trying to sleep or not to. The clock-watchers reported more worry beforehand and took longer to get to sleep, good sleepers included.

Then they ran it again on thirty-eight people with insomnia, against a box built to look identical to the clock that showed random digits instead of the time. This is the half nobody quotes, and it is better than the half everybody quotes. The clock group worried more and said the night had taken longer. The wrist monitors disagreed. Both groups fell asleep in about twelve minutes.

The clock did not cost them sleep. It cost them an honest account of the night — they overshot how long they had been awake by twenty-four minutes, the other group by ten.

So I am not going to tell you the clock keeps you awake, because in that experiment it did not. Two more caveats on top: it measured falling asleep at bedtime, not getting back to sleep at three, and nobody has run this on our exact situation.

What the clock does is make the night feel worse than it is. At 3:14, that is reason enough.

Face down, or in a drawer. Not across the room. Across the room turns into a project.

2. Do not turn on the ceiling light.

If you get up, use the lowest light in the house and keep it below your face. Hallway light, phone torch pointed at the floor. Not the overhead.

Gooley and colleagues compared 116 healthy volunteers living under ordinary room light against dim light in a research unit (2011, Journal of Clinical Endocrinology and Metabolism, 96, E463 to E472). Room light shortened the window where melatonin was elevated by roughly 90 minutes, and during sleep hours knocked melatonin down by more than half in 85% of trials.

Here is the part that gets cut off. What they measured was a hormone, not a night's sleep — and it was a retrospective look at people already living in the unit, not a coin flip. Light really does suppress melatonin. That does not automatically mean the bathroom switch at three costs you sleep, because nobody measured the sleep. Using a real mechanism to promise an unmeasured result is the most common trick in this whole subject. Keeping the light low is free and might help. That is the entire claim.

3. Give it about twenty minutes, then get out of the bed.

Not dramatically. Sit in a chair somewhere else, wrap up in something, be a bored woman in a dim room. Go back when you feel heavy rather than when you feel ready.

The twenty-minute rule comes out of stimulus control, one component of CBT-I, the talking-therapy protocol for insomnia. The honest version, which I have never seen in a post about 3 a.m.: what got tested is the whole package over weeks with a trained person, not this one instruction on one night. Trauer and colleagues pooled 20 randomised trials and 1,162 people (2015, Annals of Internal Medicine, 163, 191 to 204). CBT-I cut time awake in the middle of the night by about 26 minutes, and time to fall asleep by about 19.

Twenty-six minutes. Sit with that, because every promise you read tonight will be bigger than what the real thing delivers.

4. Get the thought out of your head and onto paper.

Not journaling. One line, in the dark, on whatever is nearest. Call the bank. She was right and I hate it. Then put the pen down and stop.

I did this for two years and I still do it. It is the thing that broke the loop for me, and the long version is in I Tried To Think My Way Out For Two Years. It worked for me. I'm one person, and one is not a study.

The research is smaller than you would like, and it is not quite research on this. Frattaroli pooled 146 randomised studies of expressive writing (2006, Psychological Bulletin, 132, 823 to 865) and got an average effect of r = .075 — tiny, near the floor of what a pooled analysis can see. Not zero. Tiny.

And what those studies handed people was three to five sessions of fifteen to twenty minutes, writing all the way down into an upsetting thing. Not six words on the back of a receipt in the dark. I am borrowing a number from a different exercise, and you should hold it that loosely.

Both are true at once: in the numbers it is small, and at three in the morning it was not small for me. I am not dropping the inconvenient half.

5. Tomorrow at nine at night, skip the drink.

Not for tonight. It is a candidate for why tonight is happening, and it has more behind it than anything above.

Ebrahim, Shapiro, Williams and Fenwick reviewed 20 studies of alcohol and sleep in healthy volunteers (2013, Alcoholism: Clinical and Experimental Research, 37, 539 to 549). The pattern held across doses: alcohol shortens the time it takes to fall asleep, deepens and consolidates the first half of the night, pushes the first REM period later at every dose tested, then disrupts the second half.

The second half is where you are sitting right now.

Nobody says this on the way to the bar, so I will. The drink that put you under at eleven is a strong suspect for why you are awake at three. I have four years of not testing that personally, which is a fact about me and not evidence about you.