Most people nap wrong. They set a two-hour alarm, sink into deep sleep at the forty-minute mark, get dragged out of it by the alarm at one-twenty, and spend the next half hour feeling worse than before they laid down. Then they conclude that naps do not work for them.
Naps work fine. The timing is broken.
The Cycle
Human sleep runs in roughly ninety-minute cycles. Light sleep, deep sleep, REM, back to light. Rinse, repeat, four to six times per night. The transitions between cycles are where you are closest to waking — your natural exit points.
Waking up at the boundary between cycles feels like surfacing gently. Waking up in the middle of deep sleep feels like being pulled from underwater with a chain. Same total sleep time. Radically different experience.
This is not opinion. It is polysomnography. EEG measurements show that sleep inertia — that groggy, disoriented, cannot-remember-your-own-name feeling — correlates almost perfectly with which sleep stage you were in when the alarm fired. Deep sleep interruption produces the worst inertia. Light sleep or REM boundary interruption produces almost none.
The Two Valid Naps
If you understand the cycle, there are exactly two naps that make sense.
The power nap: twenty to thirty minutes. You stay in light sleep. You never descend into deep sleep at all. You wake up easily because you were never very far under. Marginal recovery, but zero inertia. Good for alertness. Bad for actual physical repair.
The full cycle: ninety minutes. You descend through all the stages and surface naturally at the end of the cycle. Deep sleep happens, REM happens, and you come out the other side at the light sleep boundary. Real recovery. Minimal inertia. Your body actually got to do some work.
Everything between thirty and eighty minutes is the danger zone. Long enough to enter deep sleep, short enough to guarantee you wake up in the middle of it. This is where most people set their alarms. This is why most people think naps make them feel worse.
The Infrastructure Parallel
This is remarkably similar to how database maintenance windows work. You either do a quick cache flush that takes seconds and barely interrupts service, or you do a full vacuum that completes its entire pass and leaves the system cleaner than before. What you do not do is start a vacuum and kill it halfway through. That leaves the system in a worse state than if you had done nothing.
An interrupted maintenance cycle is not partial maintenance. It is corruption with extra steps.
Sleep is the same. A nap interrupted mid-cycle is not partial rest. It is a confused brain with elevated adenosine and no idea what time it is.
The Longer Problem
Napping past ninety minutes introduces a different risk: you start a second cycle. If you sleep for two or three hours, you are rolling the dice on where in the second cycle you wake up. Sometimes you luck into another boundary and feel great. Sometimes you catch deep sleep again and wake up feeling like you lost a fight.
There is also the night problem. A long afternoon nap reduces sleep pressure — the adenosine buildup that makes you feel tired at bedtime. Burn off too much sleep pressure at 3 PM and you are staring at the ceiling at midnight, which creates the exact deficit you were trying to fix.
Cap it at ninety. One cycle. In, through, and out. Then protect the night.
The Uncomfortable Truth
None of this is new information. Sleep researchers have been publishing this data for decades. The ninety-minute cycle was documented in the 1960s. And yet the default advice is still "take a nap" with no guidance on duration, as if sleep were a binary state you toggle on and off rather than a multi-phase process with specific timing requirements.
We build systems with careful attention to state transitions, graceful shutdowns, and clean restarts. Then we treat our own operating system like it has a power button and nothing else.
Respect the cycle. Twenty minutes or ninety. Nothing in between.