and Law in Electro-optics. 167 



12. Dividing the square of each of the preceding potentials 

 by the corresponding optical effect, and keeping the quotients 

 true to half a unit by excess or defect, we find the eight num- 

 bers 



72 73 69J 72 73 70J 69J 71. 



There is here a very fair approximation to equality, though 

 it is hardly so close as might have been expected. 



13. Later Measurements. — In the next three sets of mea- 

 surements the work was still of the same empirical and tedious 

 kind as formerly (8, 9, 10). The weather was exceptionally 

 favourable both for electrical and optical work ; and the range 

 of potential in the second set was almost as large as the in- 

 struments could reach in any circumstances. 



In the actual work, I considered that each of the observed 

 values of optical effect was very probably true to less than ±1, 

 and that none of them was possibly in error by so much as ±2. 



There is no equality to be expected between the results of 

 one set of measurements and those of another set ; for the 

 constancy of the electro- optic field was not carefully attended 

 to at present, except in each day's work by itself. 



14. On the day after the decisive experiments of (10) I 

 obtained the following set of results (potentials are given as 

 formerly in the first line, and optical effects in the second) : — ■ 



60 



70 



90 



120 



130 



150 



50 



67 



113 



197 



234 



310 



Dividing the squares of the successive potentials by the 

 corresponding optical effects, and keeping each quotient true 

 to half a unit, we find the six numbers 



72 73 71J 73 72 72J. 



Applying the number 72 \ as divisor to the squares of the 

 successive potentials, we find the following set of calculated 

 values of optical effect, each written under the corresponding 

 potential. For comparison, the observed values of optical 

 effect are repeated in the third line, and the differences be- 

 tween observed and calculated values are written in the fourth 

 line. 



60 



70 90 



120 



130 



150 



49-6 



67-6 111*7 



198-6 



233-1 



310-3 



50 



67 113 



197 



234 



310 



•4 



-•6 1-3 



-1-6 



•9 



-■3 



The differences in the third and fourth columns are rather 

 large (13). 



15. Next day I obtained the following set of results. The 



