A bed, a table, a chair, an exercise bike.
A kitchenette, a bathroom through the far wall.
No window. No radio. No newspaper. No clock, and no hands on the one left on the shelf as furniture.
Two steel doors at the entrance, and only one can be opened at a time.
Everything in the room was chosen so a person inside cannot tell the hour, except one object left in deliberately that wrecked the answer.
The switch on the lamp is a message to the brain
The lamp is the flaw.
Light is not illumination here. It is the input that sets the clock, and the effect depends entirely on when it arrives.
Light in what your brain calls late evening pushes the clock later. Light in early morning pulls it earlier. Between them sits the low point of core temperature, and that is the hinge.
Now give somebody a switch and no way of knowing the hour.
They stay up reading. They turn the lamp on before the temperature minimum and off after, because that is what anyone does at the end of a long evening.
Every cycle delivers a delay and almost never an advance. The subject is not revealing a slow clock. They are slowing it themselves, a little, every day, and the record cannot tell the two apart.
What was actually built under the hillside
It went in beside a monastery in upper Bavaria in the mid sixties, after pilot work in a wartime bunker nearby.
Two identical apartments, and a control room staffed around the clock.
One apartment was shielded against electric and magnetic fields and the other was not, which turns out to matter. Walls were about three feet thick, a figure that comes from a magazine feature rather than a methods section, and no source gives the depth below ground.
Food came through a lock at shuffled hours, so a delivery could never serve as a cue, with samples going back the other way.
Subjects pressed a button on waking and another at bedtime. Sensors in the floor and bed frame logged movement, and a continuous temperature trace ran throughout, which is the real record.
The numbers from both eras
Across 147 subjects the mean free running period was 25.0 hours, give or take half an hour, individuals spread from roughly 24 to 27.
Split by room, it splits too. The unshielded apartment averaged 24.87 hours across 34 people, the shielded one 25.26 across 50.
A gap of 23 minutes between two rooms in one building should warn you off treating 25 as a constant.
The modern figure comes from a different design. Subjects live on an imposed 28 hour schedule, going to bed four hours later each day for a month, in light under two foot candles so no dose can shift anything.
Spread that way, light falls evenly across every phase and the feedback cancels. The period lands at 24 hours 11 minutes, the same in twenty year olds and sixty year olds, with nearly nine tenths of estimates between 24.00 and 24.35.
What the bunker got right, and it was not the period
In about 30 percent of subjects something strange happened.
The sleep and wake cycle came unstuck from the temperature cycle, and the two ran at different lengths in one body.
One subject went to a sleep cycle of 33.4 hours while his temperature carried on near 25. Re analysis of the archive found sleep cycles stretching to 68 hours in some people and shortening to 12 in others, with temperature near 25 throughout.
That finding survived. It is why the human clock is modelled as more than one oscillator, and it came out of the rooms that produced the wrong headline number.
Worth remembering what they worked without. The part of the brain that keeps the body clock was not identified until 1972, so the apparatus was built to measure something nobody could yet locate.
It sits in a line of isolation work that keeps producing different answers, from the group who spent forty days in a French cave to the climber who went 500 days underground.
Why the wrong number outlived the right one
The correction is 27 years old, and it is still losing.
Twenty five hours appears in sleep books, revision cards, wellness copy and plenty of science journalism, because it is the better story. A clock that wants a longer day explains hard mornings.
The real figure explains it too, less dramatically. Eleven minutes a day is enough drift to need correcting every morning, which is what a sunrise is for, as the 1999 paper and the announcement beside it said.
The bunker ran until 1989 and the building is still there, derelict the last time anybody looked.
What it proves is narrower than it looks. Take every cue from a person and they do not reveal their true rhythm. They start writing one, with the only tool you left them, and it takes thirty years and a boring lighting protocol to take it back.
Read the whole thing?
Get the week's signal, not the noise
Our sharpest reporting on energy, climate and nature — free, once a week.
