Sleep Schedule vs. Circadian Rhythm: Three Anchors, One Doing Most of the Work

In 2019, Christopher Depner and colleagues published a study in Current Biology that gets cited far more often than it gets read. Thirty-six healthy adults were assigned to nine hours in bed nightly, five hours nightly, or five hours nightly with an unrestricted recovery weekend. The recovery group slept longer and felt better over those two days. Then they went back to restriction, and by the end of the protocol their muscle-specific insulin sensitivity had dropped further than the group that never got a weekend at all.

The study is usually deployed as proof that irregularity is expensive. Fair enough. But it says nothing about which part of the week to hold still, and nearly all the advice built on top of it collapses sleep schedule and circadian rhythm into one dial — the dial you supposedly turn by going to bed at the same time every night.

We wanted to know whether that dial exists, and where it actually is. So we compared the three things people are told to keep constant, on three criteria we'll state up front: direct phase authority (can it move the pacemaker itself, or only the behavior around it), survivability (does it hold up across a week that includes a late dinner and a Saturday), and failure mode (what it costs on the nights you miss).

Anchor one: the fixed bedtime

Bedtime is the anchor people are told to defend, and the one they feel worst about losing.

Mechanically, it is the weakest of the three, because bedtime is an output. Sleep onset happens when homeostatic sleep pressure — Borbély's Process S, accumulating since you woke — crosses a circadian alerting signal that is not under your control. That alerting signal does something counterintuitive in the evening: it peaks. Peretz Lavie's work in the 1980s described a window of roughly two to three hours before habitual sleep onset in which sleep is unusually hard to initiate. He called it a forbidden zone; it's now more often called the wake maintenance zone.

So a person whose pacemaker peaks its alerting drive at 10:30 p.m., told to be in bed with lights off at 10:00, has been instructed to lie down inside the least sleepable window of their day.

What happens next is well characterized. Richard Bootzin's stimulus control work, going back to 1972, treats the bed as a learned cue. Pair it repeatedly with lying awake, checking the clock, and running the arithmetic on how many hours are left, and the bed stops predicting sleep and starts predicting arousal. Perlis and colleagues formalized this in 1997 as a neurocognitive model of insomnia, with conditioned cortical arousal at the center.

There's also the effort problem. The common citation here is Ansfield, Wegner and Bowser (1996), who instructed participants to fall asleep as quickly as possible and found that the instruction lengthened sleep onset, most sharply under high cognitive load. It's a small laboratory study and we'd hold it loosely, but it points at something clinicians see constantly: trying to sleep is an activity, and activities are incompatible with sleep.

Anchor two: the fixed wake time

Wake time has no direct line to the pacemaker either. What it has is control over the input that does.

When you get up, you decide when your retina first sees daylight. That single fact makes wake time the scheduling variable with the most downstream leverage, and it is why sleep restriction therapy and CBT-I protocols fix rise time and let bedtime float rather than the reverse.

The observational evidence is more suggestive than the mechanism deserves credit for. Phillips and colleagues (2017, Scientific Reports) tracked 61 undergraduates and found that irregular sleepers — measured with a sleep regularity index — had dim-light melatonin onset roughly 2.6 hours later than regular sleepers, despite similar total sleep duration. Duration was matched; timing was not. Windred and colleagues (2024, Sleep) took the same regularity measure to roughly 60,000 UK Biobank participants wearing accelerometers and reported that regularity predicted all-cause mortality more strongly than sleep duration did.

That is a large, well-instrumented association. It is still an association.

Anchor three: timed morning light

This is the only one of the three with direct phase authority, and the evidence base is not close.

Light reaches intrinsically photosensitive retinal ganglion cells containing melanopsin, most responsive around 480 nm — blue-toward-cyan. Those cells project through the retinohypothalamic tract to the suprachiasmatic nucleus. Khalsa and colleagues (2003, Journal of Physiology, 23 participants) mapped a human phase response curve to a single bright light pulse, showing that the same light produces a phase advance or a phase delay depending entirely on when it lands.

Intensity matters less than most people assume. Zeitzer and colleagues (2000, Journal of Physiology) found roughly half-maximal melatonin suppression near 100 lux — around the level of ordinary indoor lighting. Timing matters more. Wright and colleagues (2013, Current Biology) sent eight people camping for a week with no electric light and found melatonin onset advanced by about two hours.

How the shift actually happens, in order

1. Light hits the retina. The ipRGCs integrate it over minutes, not milliseconds — brief glances don't count for much.
2. The signal travels the retinohypothalamic tract to the SCN.
3. The SCN reads the timing against your core body temperature minimum, which sits roughly two hours before habitual wake in most adults. Light after that minimum advances the clock. Light before it delays.
4. Over subsequent days, melatonin onset moves earlier. Advances are slower and smaller than delays — the human clock resists being pulled forward.
5. The window of high sleep propensity moves earlier with it.
6. Bedtime follows.

Bedtime is step six. It is a readout of the first five.

Is a consistent wake time better than a consistent bedtime?

For adults with variable weekday and weekend schedules, yes — wake time is the more useful thing to hold constant, because it fixes the timing of your first light exposure, and light is the only variable in this comparison with direct access to the circadian pacemaker. Bedtime cannot shift the clock; it can only place you in bed before or after the clock is ready.

The caveat is real. We could not find a trial that randomized adults to fixed-wake-only versus fixed-bedtime-only and compared outcomes head to head. The wake-time recommendation is an inference from a solid mechanism and consistent observational data, not a finding from a study designed to test it.

An honest rule of thumb

• Hold wake time inside a 60-minute band across all seven days. A one-hour Saturday sleep-in is not a lapse.
• Get outdoors within an hour of waking, for 15 to 30 minutes. Overcast daylight still runs 10 to 100 times brighter than a lit room.
• Go to bed when you're sleepy, not when the schedule says. If you've been awake and wired for about 20 minutes, get up.
• If you do sleep in, get the light anyway — then go back to bed if you want to.

Where each claim stands

Light as the principal zeitgeber for the human circadian pacemaker: well-established, across decades of laboratory phase-shifting work. Sleep regularity predicting health outcomes: a strong observational signal that has not been shown to be causal. Fixed wake time outperforming fixed bedtime: plausible and mechanistically well-grounded, but not directly tested. Fixed bedtime as the main lever on your body clock: folk wisdom. Weekend recovery sleep undoing the metabolic cost of weekday restriction: contradicted by the best experiment we have, with the usual caveat that it was 36 people.¹

Reviewer's note. I've watched the bedtime rule turn otherwise unbothered sleepers into clock-watchers. The mechanism it's supposed to work through isn't the one that does the work, and the anxiety it generates is not a side effect — it's a conditioned response with its own literature.

What we couldn't settle

The tolerable band of wake-time variability is unknown. We wrote 60 minutes because it is the number clinicians use and because it is achievable, not because a study identified it as the threshold; Bei and colleagues' 2016 review in Sleep Medicine Reviews found the field measuring day-to-day variability in inconsistent ways, which makes a clean threshold hard to extract.

Reverse causation also remains open in the regularity-and-mortality data. Illness disrupts sleep timing long before it kills anyone, and accelerometer studies cannot easily separate the two.

The place to look next is dose-response work on light at ecological intensities rather than laboratory brightness, and any trial that treats regularity as a randomized intervention rather than a measured exposure. Until then, hold the morning still and let the night arrive on its own.