D O C . 1 4 5 O C T O B E R 1 9 1 9 1 2 3 take into consideration a few not so unimportant things for the determination. Whether his measurements per se were ideal does not seem certain to me either, so that despite the splendid support for the research by the Zeiss Works, I am of the view that the experiment has to be repeated by an experienced researcher. (Harress was a doctoral student at the time!) Once I am back in Berlin, I am going to seek to win over Warburg, for inst. for it seems to me that the Bureau of Standards [Reichsanstalt] is best suited for this purpose, which obviously also has in Gehrcke a very seasoned optics specialist with a genuine interest in moving bodies.[3] This in introduction. Now comes the question: When I said that I put the theory in order, I naturally calculated as if the accel- eration did not noticeably affect the rate of the speed of light. I did not always feel that this was justified. On the contrary, for some time, while I was in some desper- ation about the Harress puzzle, I believed the effect of the acceleration was detect- able. And when I spoke with W. Wien about it at the time, he told me he thought this was quite an unlikely conjecture. The arising centrifugal forces are so negligi- ble that, at least according to the general theory of relativity, with which one may equate them with attractive gravitational forces, no influence would be expected. And that is very clear to me, because in the experiment the centrifugal forces were very unlikely ever to have been 100 times as large as the Earth’s attraction on that [glass] object. I would like to know if you agree with this conclusion. Nothing can be learned from the present analysis it is not accurate enough for that. Since I myself intend to publish a couple of comments on Harress’s experiment, in order to provide whoever repeats the experiment with everything he could need for the theory,[4] I would be obliged to you for a prompt reply. I am sending you simultaneously a copy of Harress’s dissertation. I would also like to add: Whether the ray of light bends in a rotating medium is a side issue. I can show that it cannot have any influence on the observations because of Fermat’s principle of least time in geometrical optics. It could only make a difference if a velocity-independent acceleration term is added to the veloc- ity term in the formula for the speed of light. While writing this I also realize that such an auxiliary term, in order to fall under consideration here, would have to be proportional to the acceleration while it obvi- ously has the dimension of a velocity. Otherwise, it could only depend on the object’s refractive index, which is a pure number, the velocity of light c, and the gravitational constant, whose dimension is . It is, however, mathemati- cally impossible to construct such an expression. On the assumption that the acceleration has no effect, the theory can, moreover, also be presented in such a way that the experiment simply confirms the summand vx/ in the Lorentz transformation m 1– l3t 3– , c2