Sometime in the 1920s the psychologist Kurt Lewin was sitting in a Berlin café with his students and noticed that the waiter could hold an enormous, complicated order in his head without writing any of it down — and then, once the bill was settled, couldn't recall a thing about it.
One of those students was Bluma Zeigarnik, and she went and tested it properly.
Her 1927 study gave people a series of small tasks — puzzles, assembling things — and interrupted some of them partway through. Afterwards she asked what they remembered. The interrupted tasks came back dramatically better than the finished ones, on the order of ninety percent better in her original results.
Unfinished things stay loaded. Finished things get released.
It's a feature, not a flaw
The instinct is to read this as a bug in human attention, which it isn't. A brain that keeps incomplete work in the foreground is a brain that finishes things. Whatever fires when you're halfway through is the reason you come back and close it out.
Worth saying that the effect has had a mixed replication history, like a lot of early twentieth-century psychology, and the effect size moves around depending on how you set the experiment up. But the core observation has held up well enough to outlive its critics, and more to the point, everyone recognizes it immediately, because everyone has lain in bed at 11:40pm being reminded about an email.
What it costs you now
Zeigarnik's waiter was holding maybe six orders and he got to clear them within the hour.
You're holding somewhere north of forty open loops and most of them won't resolve this week. The renewal. The reply. The thing your kid mentioned. The form. That's forty separate small processes, all running, all with permission to interrupt you, none of them ever getting the completion signal that shuts them off.
Which is why you can be genuinely relaxed on a Saturday and still have a low hum of something-undone going, and why the hum doesn't correlate at all with how much you actually got done. It correlates with how much is open.
Writing it down is most of the fix
The reason externalizing tasks works — and it does work, this is the least controversial idea in the entire field — is that it satisfies enough of what the loop is asking for. The loop wants assurance the thing won't be lost. Put it somewhere reliable and it mostly stops asking.
Mostly. It'll keep asking if it doesn't trust the place you put it, and a list of eighty items you never reach the bottom of is not a trustworthy place. You'll keep the loop open as backup, because some part of you correctly suspects the list isn't going to get there.
A list that measures your day, plans what fits, and gets to the end of itself is a list your brain will actually hand things over to. The loops close because the tasks close.
The waiter forgot the order the second the bill was paid, and that wasn't a failure of memory. It was the system working. Finishing is what makes the noise stop.
Sources: Bluma Zeigarnik, "Das Behalten erledigter und unerledigter Handlungen," Psychologische Forschung, 1927 (English translation); Zeigarnik effect overview.