The fall in the speed of light from 1928 to 1945
According to Einstein's theory of relativity, the speed of light in a vacuum is invariant: it is an absolute constant. Much of modern physics is based on that assumption. There is therefore a strong theoretical prejudice against raising the question of possible changes in the velocity of light. In any case, the question is now officially closed. Since 1972 the speed of light has been fixed by definition. The value is defined as 299,792.458 ± 0.001 # 2 kilometers per second.
As in the case of the universal gravitational constant, early measurements of c differed considerably from the present official value. For example, the determination by Römer in 1676 was about 30 percent lower, and that by Fizeau in 1849 about 5 percent higher.
In 1929, Birge published his review of all the evidence available up to 1927 and came to the conclusion that the best value for velocity of light was 299,796 ± 4 km/s. He pointed out that the probable error was far less than in any of the other constants, and concluded that 'the present value of c is entirely satisfactory, and can be considered as more or less permanently established.' However, even as he was writing, considerably lower values of c were being found, and by 1934 it was suggested by Gheury de Bray that the data pointed to a cyclic variation in the velocity of light.
From around 1928 to 1945, the velocity of light appeared to be about 20 km/s lower than before and after this period. The 'best' values, found by the leading investigators using a variety of techniques, were in impressively close agreement with each other, and the available data were combined and adjusted by Birge in 1941 and Dorsey in 1945.
In the late 1940s the speed of light went up again. Not surprisingly, there was some turbulence at first as the old value was overthrown. The new value was about 20 km/s higher, close to that prevailing in 1927. A new consensus developed. How long this consensus would have lasted if based on continuing measurements is a matter for speculation. In practice, further disagreement was prevented by fixing the speed of light in 1972 by definition.
How can the lower velocity from 1928 to 1945 be explained? If it was simply a matter of experimental error, why did the results of different investigators and different methods agree so well? And why were the estimated errors so low?
One possibility is that the velocity of light really does fluctuate from time to time. Perhaps it really did drop for nearly twenty years. But this is not a possibility that has been seriously considered by researchers in the field, except for de Bray. So strong is the assumption that it must be fixed that the empirical data have to be explained away. This remarkable episode in the history of the speed of light is now generally attributed to the psychology of metrologists:
The tendency for experiments in a given epoch to agree with one another has been described by the delicate phrase 'intellectual phase locking.' Most metrologists are very conscious of the possible existence of such effects; indeed ever-helpful colleagues delight in pointing them out! . . . .Aside from the discovery of mistakes, the near completion of the experiment brings more frequent and stimulating discussion with interested colleagues and the preliminaries to writing up the work add fresh perspective. All of these circumstances combine to prevent what was intended to be 'the final result' from being so in practice, and consequently the accusation that one is most likely to stop worrying about corrections when the value is closest to other results is easy to make and difficult to refute.
But if changes in the values of constants in the past are attributed to the experimenters' psychology, then, as other eminent metrologists have observed, 'this raises a disconcerting question: How do we know that this psychological factor is not equally important today?' In the case of the velocity of light, however, this question is now academic. Not only is the velocity fixed by definition, but the very units in which velocity is measured, distance and time, are defined in terms of light itself.
The second used to be defined as 1/86,400 of a mean solar day, but it is now defined in terms of the frequency of light emitted by a particular kind of excitation of caesium-133 atoms. A second is 9,192,631,770 times the period of vibration of the light. Meanwhile, since 1983 the meter has been defined in terms of the velocity of light, itself fixed by definition.
As Brian Petley has pointed out, it is conceivable that:
(i) the velocity of light might change with time, or (ii) have a directional dependence in space, or (iii) be affected by the motion of the Earth about the Sun, or motion within our galaxy or some other reference frame.
Nevertheless, if such changes really happened, we would be blind to them. We are now shut up within an artificial system where such changes are not only impossible by definition, but would be undetectable in practice because of the way the units are defined. Any change in the speed of light would change the units themselves in such a way that the velocity in kilometers per second remained exactly the same.