Moving the caffeine cut-off, and what the sheet did and did not show

Six weeks with nothing after two in the afternoon, then six weeks with no rule at all, in that order, with everything else held as steady as a life allows. I expected a large effect on how long it took to fall asleep. I got a small one, and a larger effect somewhere I was not looking.

Measure	With the rule	Without
Self-reported time to fall asleep, median	18 min	24 min
Nights I recorded as “woke and stayed awake”	4 of 42	11 of 42
Median lights-out	22:54	23:19
Cups per day, mean	2.4	3.6

The bit I did not expect

Falling asleep barely moved. Staying asleep moved a lot — nearly three times as many nights where I woke in the small hours and did not get back off. Whether that is caffeine or something correlated with the kind of day on which I want a fourth cup, I genuinely cannot separate.

Two obvious problems with this comparison. The order was not randomised, and the second block ran through December, which is not a normal six weeks for anyone. I would not have published this as anything more than a note about myself.

Where the extra cups came from

All of the difference is in a single slot: between three and five in the afternoon. Morning intake was identical to within a rounding error across both blocks. So the rule was not really a rule about caffeine; it was a rule about what I do at four o’clock, and what I did instead was make tea and walk to the window.

What I kept

Not the two o’clock rule, which I broke eleven times in six weeks and resented on most of them. I kept the four o’clock walk to the window, which happens with or without a cup in hand, and my afternoon intake has settled around 2.9 cups without me doing anything else about it.

Not advice

An uncontrolled comparison of one person against himself, with self-reported numbers, over twelve weeks that included a holiday. Read it as a diary entry.

Previous: twice a week · next: rebuilding an evening routine