EXPERIMENTS ON KAOLINIZATION. 297 



server. The cups corresponding to m, and conveying the current due to the 

 Daniell E, are, however, filled with mercury to a level a little higher than 

 the rest. Hence, under all circumstances the circuit containing E, and not 

 passing through G^, is closed first. A moment after, however, that contain- 

 ing e and G^ is also completed, but it will be obvious that the effect from 

 e and E, if the directions of these electromotive forces are properly chosen, 

 will act differentially on G^, as was desired. 



The electromotive force e, obtained as above, is never wholly due to 

 the thermo-element e alone, but contains also a disturbing electromotive 

 force, £, resulting from the accidental distribution of temperature in the 

 connections. For a short period of time (that of an observation) e may be 

 considered as nearly constant, or at least varying linearly. In order to elim- 

 inate the latter, very largely at least, Dr. Strouhal and myself, in analogous 

 experiments, inserted the two commutators A and B. In a series of corre- 

 sponding positions of the commutators, alternately opposite, direct meas- 

 urement would give 



+e+£ 

 +E 



_e+£(l+a) 

 -E ~ ' 



+e+e{l+2a)_^ 



+E 



-e+Kl+3^) _ 

 -E ~~ ' 



+e+£(l+4a). 

 ~\-E 



where a is a constant. If now an odd number of observations be made, 

 and if Jf, be the mean of the odd right-hand members, M. the mean of the 

 even right-hand members, 



In the present investigation, the electromotive forces measured being 

 exceedingly small, at least five commutations of both A and B were made 

 for each value of -^t cited. 



