﻿474 Dr. Norman Campbell on 



n x as an unknown, to calculate its values, and subsequently 

 to compare the calculated and the observed values. Ac- 

 cordingly, the following method of successive approximations 

 was adopted, the first of the two theories of the previous 

 paper as to the nature of the negative ion being chosen 

 provisionally. 



It will be seen that, according to (1), the relation between 

 i and p given in any one of the tables should be of the form 



It was assumed that the relation was of this form, and the 

 values of P, R, and /jl calculated from the measurements in 

 each of the tables. The presence of three unknowns made 

 the probable errors of calculation large, and attention was 

 only paid to the values of. jjl obtained. These values, calcu- 

 lated by this method for the various tables, are given in 

 Table XX. It will be observed thai, within the rather large 

 limits of error, there is no systematic difference between the 

 values of //, for gold and for aluminium. Accordingly, all 

 the values of //,, both for gold and aluminium, were taken as 

 giving a single set of equations, similar to (4), whence the 

 values of v and V might be calculated. The figures so 

 obtained are given at the foot of the table. 



13. An approximate value of R can now be calculated for 

 each of the tables from the values of v and V together with 

 the determined value of n . Taking this value as correct, 

 each of the tables now gives a set of equations of the form 



from which P and jjl can be determined. The necessary 

 calculations w^ere carried out, and the results are given in 

 Table XXI. The new values obtained for //, are satisfactorily 

 in accord with those determined before, and it appeared that 

 nothing would be gained by repeating the same process of 

 approximation. The values of i calculated from these values 

 of P and /x, together with the approximate value of R, agreed 

 thoroughly with those determined experimentally. In a few 

 cases the difference between the calculated and observed 

 values was greater than the uncertainty in a single measure- 

 ment of the current, but no systematic divergence of the 

 values for any given table could be detected. It may be 

 confidently concluded that the curve is very nearly of the 

 form assumed, and that the value of R adopted is very nearly 

 correct. 



