Why your mind races at night, and what a sleep lab found about writing

Fifty-seven young adults sat in a controlled sleep laboratory with a sheet of paper and five minutes to write, then went to bed wired to a polysomnograph. Half listed what they still had to do. Half listed what they had already finished, and took about nine minutes longer to fall asleep. If your mind races the moment the light goes off, one culprit the researchers name is the pile of unfinished business. At that hour it has nowhere else to be. The quickest way to set it down, on the night they measured, was not the one most bedtime journaling advice recommends.

Here is the experiment, what it was compared against, and where it stops.

What the experiment actually did

Scullin, Krueger, Ballard, Pruett and Bliwise (2018) recruited 60 healthy adults aged 18 to 30, 57 of whom made the final analysis. Each was assigned at random to one of two five-minute writing tasks at bedtime. One group wrote a to-do list of what they needed to remember over the coming days. The other wrote about activities they had completed that day and over the previous few days. Lights went out at about half past ten, and sleep was recorded objectively, by polysomnography, in the laboratory.

Two design choices carry the whole result. The first is the measure. Sleep onset latency here is not how long the night felt to the person lying there; it is the electrode reading of when sleep actually began. Most of the research on sleep and writing has relied on what people report the next morning, which is what makes this record the harder kind of evidence.

The second choice is the more interesting one: the control. The completed-list group was not a group that did nothing. They also wrote. They also spent five minutes with a pen, at the same hour, in the same room, under instructions to write about their tasks. The only variable left standing is the direction the writing pointed in time.

That matters more than it sounds. Compare a to-do list against no writing at all and the result is ambiguous, because five quiet minutes at bedtime might be doing the work by themselves. Holding the writing constant and flipping only its tense is what makes the nine minutes attributable to something narrow enough to act on.

Nine minutes, and what made the difference

The to-do list group fell asleep in 15.82 minutes on average. The completed-list group took 25.09, a gap that came out at t(55) = 2.32, p = .02 and a medium effect size, d = .63. Detail mattered as well. Within the to-do group, longer and more specific lists went with falling asleep faster, r(26) = −.39, p = .04.

Nine minutes is not a trivial difference for a single change in behavior measured under electrodes. It is also not a night rescued. Both things are worth holding at once.

The specificity result is the part that usually gets dropped in the retelling. Writing “work stuff” is not the same act as writing out the five things work stuff actually consists of. In this sample, the second version was the one that tracked with faster sleep onset.

The completed-list group showed the mirror image. The more they detailed what they had already done, the longer they took to fall asleep, r(27) = .34, p = .07. That one did not reach significance. Read it as a hint, not a finding.

What the researchers think the nine minutes mean

Scullin and colleagues read their result as offloading. Ask why your mind races at night and their answer is custody: an intention with nobody holding it keeps checking in. An unfinished task has no natural stopping point inside your head, so it resurfaces on a loop until something takes it off your hands. Putting the task on paper transfers custody, and the more precisely you transfer it, the less there is left to circle.

This is where the analysis has to slow down. The polysomnograph recorded when people fell asleep. It did not record why. Offloading is the authors’ own reading, argued in their discussion section, and it remains a reading. Any mechanism you attach to those nine minutes, that one included, is inference laid over a stopwatch.

What the data support is a direction. Forward-pointing writing beat backward-pointing writing on this measure, and precision helped. It is a narrow claim, which is exactly what makes it usable.

The bedtime advice that runs the other way

The completed-list condition was not a straw man. Participants wrote about the activities they had finished that day and the days before it — the written recap, in other words, which is close to what most bedtime journaling advice recommends. That was the slower of the two conditions, and the more detailed version of it trended slower still.

That is an uncomfortable result for anyone who writes about journaling for a living, this site included. It is better stated plainly than buried under qualifications.

It also has to be stated accurately, because this one gets stretched in the retelling. The two conditions were tasks ahead and tasks behind. Writing three things you are glad for was never one of them: the instruction was about tasks, not thanks. The experiment says nothing about gratitude journaling, for the simplest reason, which is that it did not measure it. Gratitude has a literature of its own, with effects more modest than the promises around it, and what it does and does not support belongs there rather than here.

The hour is not incidental either. What was measured was five minutes of writing at about half past ten, in a bed with the electrodes already on. The same recap at seven in the evening never entered the comparison. So the practical reading is scheduling, not abolition: move the recap earlier if you want it, and keep the last five minutes pointed forward.

Making that forward list specific enough to be boring is a reasonable bet rather than an established finding; it comes from the thinnest analysis in the paper. A short session with a clear ending is the shape most of the writing evidence rewards anyway, and it is where a journaling habit is easiest to start.

What the study does not show

It does not show that a to-do list will fix your nights. Fifty-seven people, one night each, all healthy, all between 18 and 30, all sleeping in a laboratory for the first time. It was not built to say anything about people whose difficulty falling asleep is long-standing, and the authors name a randomized controlled trial as the next step.

The correlations are the thinnest part of it. Each rests on roughly 28 people, and the authors state outright that the sample was sized for a controlled experiment, not for an individual-differences analysis. Read the specificity findings as a lead worth following rather than a rule.

The single night matters too. Everything above comes from a first night in an unfamiliar room with sensors attached, which is not a neutral setting for anyone. Whether five minutes of writing still buys nine minutes on night forty, in your own bed, was not measured. The authors say so themselves, and name naturalistic settings as the open question.

One clean result, then, obtained once, under conditions that were not yours. On that night, in that room, writing forward beat writing backward by nine minutes, and how specifically people wrote carried part of the difference. That is less than the headlines made of it and more than nothing. Finding out whether it transfers to your own room costs five minutes and a pen. The honest version of that experiment includes the outcome where it does not.