A clock that’s always a little late
Stand near a public clock long enough and you notice something odd: it can be wrong in a steady, repeatable way. Not random minutes here and there, but the same drift year after year. There isn’t one single famous clock tower that all sources agree “loses exactly twelve minutes every year,” but the pattern is real. Big mechanical clocks in towers like Big Ben in London, the Prague Astronomical Clock, and the Zytglogge in Bern are constantly managed to avoid it. When that management slips, the clock’s “personality” shows up as a consistent gain or loss. The core mechanism is plain. A pendulum or balance wheel runs slightly off its intended rate, and the error accumulates.
Why twelve minutes is the kind of number you can get
Twelve minutes per year sounds dramatic until you translate it into the clock’s daily life. A year is 525,600 minutes, so being off by twelve minutes is an error of about 0.0023%. That’s roughly 2 seconds per day. A tower clock can miss that by just having its effective pendulum length change by a tiny amount, or by experiencing slightly more friction than it “should.” The clock doesn’t need to “break” to do this. It just needs to run at a rate that’s a hair slow, day after day, while still sounding confident on the hour.
That steady drift is the giveaway that something in the timekeeping element is consistently different from its ideal condition. Not a one-time shock. Not vandalism. Not a power cut. A slow clock that loses the same amount every year is usually a clock that is stable in the wrong direction.
The slow creep: temperature, air, and friction
Most tower clocks historically used pendulums, and pendulums are sensitive to temperature. Warmer metal expands, making the pendulum effectively longer, which slows the swing. Colder air does the opposite. If a clock is in a drafty, sun-facing tower space that warms in a predictable seasonal cycle, it can spend more of the year running a little slow than a little fast. Air density also changes drag on the pendulum bob, and that drag is small but relentless.
Then there’s friction. Thickening oil in winter, dust in a gear train, a slightly tight bushing, or a pallet face that isn’t quite happy can all nibble away at the rate. People often focus on the big visible hands and bells. The overlooked detail is the escapement surfaces and their lubrication. A tiny change there can alter the “impulse” the pendulum gets each swing, and a tiny change repeated hundreds of thousands of times becomes minutes on the dial.
Hands, dials, and the extra load no one thinks about
A tower clock is not just a timekeeper. It’s also a machine hauling big hands around a big dial, sometimes several dials at once through long shafts and linkages. Wind pushes on large hands. Ice can form. Birds can perch. A hand that is slightly bent or rubbing a dial ring adds load. That load can steal energy from the train and change the behavior of the escapement, which changes the rate. The clock may still run. It may still strike. It just runs a touch slow all year.
This is one reason an indoor regulator clock can be wonderfully accurate while the tower clock built on the same basic principles drifts. The timekeeping element might be decent, but the rest of the system is a moving outdoor sign with a lot of drag. Even the minute hand’s counterweight matters. If it’s off, the torque demand changes across the rotation, and that uneven demand can show up as subtle rate changes.
Why “without fail” usually means “someone keeps letting it happen”
The most interesting part of a precise yearly loss is that it often reflects human routine as much as physics. Many tower clocks are inspected and corrected on schedules: weekly, monthly, or seasonally. Big Ben, for example, has historically been adjusted by changing the effective pendulum rate, famously using coins on the pendulum. That sort of correction culture exists because these clocks do drift, and caretakers learn the drift pattern. If a clock is only reset once a year—say, around an annual festival, a civic ceremony, or a maintenance contract renewal—the “twelve minutes per year” can simply be the amount it drifts between those resets.
It can also be made to look precise by the way people remember it. A caretaker might notice it’s “about ten to fifteen minutes slow by spring” and repeat the same fix each time. Over a few years that becomes “twelve minutes, every year.” The dial doesn’t tell you how the clock behaved day to day. It only shows where it ended up when someone finally compared it to a phone, a radio time signal, or another trusted clock and decided it was time to pull it back.
