The 90-minute rule is borrowed from sleep research

Good advice, shaky citation. Worth separating the two.

Advice to work in ninety-minute blocks almost always arrives with a scientific-sounding justification attached: your body runs on ninety-minute ultradian cycles, so working with them rather than against them is biologically optimal. The block length is genuinely useful. The justification is considerably weaker than the confidence with which it is delivered.

What the research says

The idea traces to Nathaniel Kleitman, the sleep researcher, who proposed a basic rest-activity cycle and hypothesised that it ran during wakefulness as well as sleep. That hypothesis is the whole basis of the productivity version.

It was contested early. By 1977 other investigators were arguing there was no evidence for a basic rest-activity cycle outside sleep, and a 1976 Nature study found the REM cycle to be sleep-dependent rather than an expression of some rhythm running all day. A US defence technical report looking for a ninety-minute rhythm in gross activity did not find one, and concluded that if ninety-minute cycles exist in some behaviours, gross activity is a poor way to see them.

The picture is not entirely negative. More recent work suggests ultradian fluctuations during wakefulness are real, affecting things like focus and reaction time, but operating through different mechanisms than the sleep cycle. The popular 90/20 schedule itself traces to Ernest Rossi, who extended Kleitman’s work well past where Kleitman left it.

What is missing from all of it is the step everyone assumes: a controlled trial showing that ninety-minute work blocks with breaks produce more output than some other arrangement. I could not find one. The chain runs from sleep research, through a contested hypothesis about waking, to a specific schedule, and the last link is an extrapolation.

Why ninety minutes is still a good number

None of this means you should not use it. It means the reason is different, and the real reason is more useful because it tells you when not to.

Ninety minutes is good for work with an expensive start. If it takes you fifteen or twenty minutes to rebuild context, a short block spends most of itself warming up, and ninety is roughly where the ratio stops being wasteful without pushing so far that the end of the block is worthless. That is an argument about your work, not your biology, and it correctly predicts that the right number is different for a code review and for a hard debugging session.

It also means that if you are drifting at sixty minutes, the block is wrong and no amount of respecting your ultradian rhythm will fix it. Shorten it. The schedule is a tool, not a biological fact you are obliged to honour.

More