330 Dr. J. Kerr's Experiments on the 



readings were taken as the measure of H', the strain- 

 generated retardation of the pencil H. 



(3) The winged plate was then transferred to another 

 table, and put in line with the quartz compensator at zero, 

 between two crossed Nicols each at 45° to the vertical : the 

 given strain was again applied, and its effect compensated ; 

 and the corresponding reading of the quartz compensator was 

 taken as the measure of the difference Y / — B7. 



In the course of the experiments, there were precautions 

 taken and checks applied, which it would be tedious to 

 describe ; and in this way a good many of the measurements 

 were thrown out of account ; but as the result of a day's 

 work there were found to be, among the later measurements, 

 about half a dozen identical sets of readings, 



V' = 55, H'=27i, V'-H' = 96, 



for which all the tests had been fairly satisfied. Reduced to 

 wave-lengths (or divided by 84, 84, 299 respectively), the 

 values are 



•654, -327, -321 ; 



and these measures of V, H', V — H', are sensibly as the 

 numbers 2, 1, 1. A second day's experiments, with 

 smaller strain, gave similarly 



•439, -219, -220; 



and a third day's work, with strain again increased, gave 



•635. -317, -314. 



It appears therefore, by this method, that the two absolute 

 retardations in question are sensibly as 2 to 1. 



10. In a third set of measurements, the winged plate and 

 refractor were used alone, and the range of measured effect 

 extended to three wave-lengths. V passing through the 

 middle of the central pillar, and H through a wing, the plate 

 was strained, and the effect was compensated and measured. 

 The strained plate was immediately moved across the pencils 

 till H took the place of Y ; and in this position the plate was 

 unstrained, and the effect was compensated and measured. 

 The next pair of measurements were generally taken in the 

 contrary order (H' to V'). 



The ratio of V to H' was measured many times by this 

 method, and with great care. The successive determinations 

 were not very concordant : with several exceptions they lay 

 between 2*05 and 1*95 ; the majority lay between the closer 

 limits 2*03 and 1*97 ; and there was no clear preponderance 

 of values above 2, or of values below 2. A number of the 



