1. Why Did Scientists Stop Using Earth's Spin as the Clock?
For a long time a second was a fraction of a day: one 86,400th of an average day, based on how long Earth takes to turn. That is a natural clock, and it worked well for ships and calendars. But by the middle of the 20th century, astronomers knew that Earth's spin is not perfectly steady, and in 1952 the International Astronomical Union adopted a more even time scale, called ephemeris time, built on the motion of the Earth around the Sun. [3]
Cesium clocks, which started running in 1955, soon confirmed that the length of a day wobbles from one year to the next. [3] A ruler that stretches and shrinks makes a poor ruler.
2. How Did "A Fraction of a Year" Become the Official Second?
Scientists in the 1950s chose a year as the new reference, not a day. A recommendation of the International Committee for Weights and Measures in 1956, confirmed by the General Conference on Weights and Measures in 1960, defined the second as a fraction of the length of the year 1900. [1] Wikipedia's summary gives that fraction as 1/31,556,925.9747. [3] That odd-looking number comes from a formula for the Sun's apparent motion written by the astronomer Simon Newcomb. [3]
A year takes far too long to measure directly every time you need a second. So in practice, observers watched the Moon. According to Wikipedia's summary, against the background stars the Moon's position changes about 13 times faster than the Sun's. [3] That makes the Moon a faster "hand" to read the time from, much as the faster hand of a clock is easier to read precisely. Matching the Moon's position with a printed table of predicted positions gave the time in "ephemeris seconds."
3. What Did the Three-Year Comparison With a Cesium Clock Find?
In 1955 Louis Essen and Jack Parry at the National Physical Laboratory in the United Kingdom got a cesium clock running. For about three years, from mid-1955 to early 1958, it was compared with Moon-based ephemeris time. Wikipedia describes the result, reported by William Markowitz, R. G. Hall, Essen and Parry in 1958, like this: one ephemeris second equals 9,192,631,770 cycles of the cesium radiation, give or take 20 cycles. [3] (We could not check the 1958 paper itself.)
Taking that figure, twenty cycles out of more than nine billion is about 2.2 parts in a billion. A 1965 article in Physics Today gives the uncertainty as "2 or 3 parts in 109." [2] By our own arithmetic, that is a difference of roughly 0.06 to 0.09 second over a whole year. It is tiny, but it is not zero.
4. Was the Doubt in the Atomic Clock or in the Sky?
The doubt came from the astronomical observations. The same Physics Today article says that the limitation "lay in the astronomical observations, not in the atomic device." [2] To get ephemeris time from the Moon you have to line up the observed position against a table of predicted positions. Any wobble in Earth's spin, or any small error in the table, goes straight into the answer. [3]
So the cesium clock was the steadier of the two instruments. It had been used to measure something that was rougher than itself.
5. Why Is 9,192,631,770 a Chosen Number and Not a Measurement Anymore?
In 1967 the General Conference on Weights and Measures replaced the old year-based definition. It decided: "The second is the duration of 9 192 631 770 periods of the radiation corresponding to the transition between the two hyperfine levels of the ground state of the caesium 133 atom." [1] The number was chosen to match the ephemeris second of the 1950s comparison. [3]
From then on the number is not something to be measured; it is the definition. A 1965 writer in Physics Today had already said it should be treated "not as a measured value, but rather as an exactly defined one." [2] Once the second is defined that way, there is nothing left to compare it with. Re-running the Moon test would only produce a fuzzier version of a number that is now exact by agreement.
This is why repeating the 1958 test would not improve the definition. The old method cannot beat the precision of today's clocks, and the thing it measured has been replaced by a decision. (A simple way to picture it: the Moon was a bridge used once to cross from the old second to the new one, and the bridge was not needed after that.) Please note that this page explains the limits that sources describe for the 1950s method; it does not claim that a modern, better astronomical test is impossible.
6. How Steady Is Your Own Second?
Look at a clock with a second hand, or use a stopwatch. Count "one, two, three..." in your head, one number per second, up to 60, and see how much real time has passed. Repeat three times. Your count will probably drift a little, and not always in the same direction.
To dig deeper, read the BIPM resolution of 1967, which is short, and then look up "ephemeris time" and the 1958 paper by Markowitz and colleagues in Physical Review Letters, volume 1, page 105. A good question to take along: how large was the table's error compared with the Moon's, and how would anyone know?