438 



Mr. A. Tribe. 



[June 16, 



led to their substitution by others of unglazed earthenware. These 

 •consisted of the sides of a large rectangular porous cell ground to as 

 uniform a thickness as possible. With this alteration, but with all 

 the other above-mentioned conditions, the following experiments were 

 made. 



Analysers were successively placed in the positions (all perpen- 

 dicular to the electrodes) shown in the annexed diagram, exhibiting 

 Sb horizontal section drawn to about one-fifth the scale. 



The analysers a, b, c, d, 2 millims. from electrodes, recorded parallel 

 distributions. 



The analysers e, /, g, h, 2 millims. from diaphragm, recorded not 

 absolutely but very nearly parallel distributions. This slight non- 

 parallelism was not noticed in the analogous experiment with the 

 parchment diaphragms, and I am disposed to attribute it to a greater 

 diffusion in this case occasioning a less sharp line of demarcation 

 between the media. 



The analyser i, midway between the diaphragms, recorded a non- 

 parallel distribution of a most pronounced character. The degree of 

 •curvature showed that the energy had been refracted through some 30°, 

 while the position of the ions proved that the bending was towards a 

 perpendicular to the refracting surface. Furthermore, the symmetry 

 of the curve showed that the plane of refraction was the same as that 

 of the incident energy. 



The analysers j, 7c, indicated a result almost identical with i. 



The analysers Z, m, showed a much smaller deviation from the 

 original course of the energy. 



The positions of the relatively good and bad conducting media were 

 now reversed. The 1 per cent, solution was placed in the transmitting 

 cell, and the concentrated in the refracting cell. An analyser at i 

 recorded a non-parallel distribution, and the curvature and position of 

 the ions showed that the course of the energy had been bent about 

 15°. But in this case the refraction was from a perpendicular to the 

 refracting surface. 



It was now natural to anticipate that the refraction, with a given 



