"It doesn't affect my sleep, I drop off fine"

This is the belief that lets the afternoon and evening cups happen, and it's held with real confidence. You put your head down at eleven and you're asleep within twenty minutes, most nights. Coffee clearly isn't keeping you up.

Two things are wrong with the test you're using. The first is that falling asleep isn't the only measure of a night. The second is that you're not in a position to judge the rest of it.

Ian Clark and Hans-Peter Landolt reviewed the trials and population studies on caffeine and sleep in 2017. Caffeine typically made people take longer to fall asleep, but it also reduced total sleep time, reduced sleep efficiency, which is the share of time in bed spent asleep, and worsened how people rated their sleep. Underneath, in the recordings of brain activity, the deep slow-wave sleep was reduced and the light stage-one sleep, wakefulness and arousals were increased. The dose mattered and the timing mattered. Older adults may be more sensitive than younger ones. And there were large differences between people, some of them genetic.

So the picture is of a night that's shallower and more broken and a little shorter, with less of the deep sleep that leaves you restored. You may drop off fine. What follows is a lighter night than you'd have had, and you can't feel the difference from the inside, because you were asleep. All you have is the morning, and in the morning you feel tired, and you've got an explanation for that already, which is that you're a tired person.

The timing is the part people underestimate most. Christopher Drake and colleagues gave people four hundred milligrams of caffeine, about four mugs of instant, either at bedtime, three hours before it, or six hours before it, and measured their sleep at home. All three disturbed sleep compared with placebo, including the dose six hours out. The authors concluded that caffeine six hours before bed has important effects and supports the advice to avoid it for at least that long. The NHS insomnia page says the same, not to drink tea or coffee for at least six hours before bed. Your half past five cup, for an eleven o'clock bedtime, is inside that window. Your three o'clock cup is only just outside it. And Landolt's morning dose, from chapter 4, was sixteen hours out and still changed the night.

There's a plain reason for all this. Caffeine leaves the body slowly. Several hours after a cup, a good fraction of it is still there. A cup at half past five is still partly present at eleven, doing at night exactly what it did in the afternoon, which is muffling the signal that says sleep.

I'd put it this way. You drop off fine because you're tired enough to drop off through the caffeine. That isn't evidence that the caffeine's doing nothing. It's evidence of how tired you are.