86 Mr. J. 0. McConnel on the 



itself limited the aperture considerably, and, to stop light 

 reflected from the sides of the Nicol, he put a diaphragm 

 between it and the quartz. The diameter of the aperture, 

 according to the engraving which he gives, was less than 

 half the thickness of the pile of plates. I have tried to 

 approximate to the amount of error involved, and have con- 

 vinced myself that it must have been considerable. He him- 

 self was led to suspect some fault in his standardizer for 

 the following reason : — Let us call the reading of the plates 

 when they neutralized the polarization of the sky on a certain 

 occasion Gj. Then, when the plates were tilted in the 

 opposite direction, they again neutralized the polarization at 

 a reading Gg. The absolute values of the polarization corre- 

 sponding to the readings Gi and G2 should of course have 

 been identical, but according to the standardizer they were 

 always markedly different [e.g. such as to give r="250 and 

 r = '225)*. He was unable to find any explanation of these 

 discrepancies, but they led him to regard his observations 

 as only comparable with each other when the pile had not 

 been dismounted in the interval. Further he specially says 

 (p. 56) that his measures of the polarization are not to 

 be compared with measures obtained with other instruments, 

 or, in other words, that they are not absolute values at all. 

 We can see now that he was right, and that these discre- 

 pancies were only a sign of a much larger error. Owing to 

 some slight want of symmetry, more of the reflected light 

 must have been thrown out of the field of view on one side 

 than on the other. 



When I firLt appreciated the meaning of this oversight in 

 Eubenson's work, I thought it would render his elaborate 

 series of measures quite useless except for intercomparison. 

 But I have since discovered that the polai'izing power of a 

 pile of glass plates varies very slowly with the index of refrac- 

 tion. I do not know hov/ this result will strike others, but 

 to me it was most unexpected. However, it enables us to 

 obtain a fair idea of the real values of the polarization that 

 Rubenson observed. There is a very simple formula^ due to 

 Prof. Stokes t, for the polarizing power of a pile, taking into 

 account all the reflexions. Once found it may be easily 

 verified. Let a be the proportion of light, polarized in the 

 plane of incidence, reflected from a single surface ; then the 



* He aimed at a probable error less than -002, and would no doubt 

 have succeeded if the error had not been systematic. 



t Proc. Ptoy. Soc. 1862. It was rediscovered by Prof. W. G. Adams, 

 Phil. Mao-. March 1871. 



