46 PROCEEDINGS OP THE AMERICAN ACADEMY. 



The heavier of the two suspension systems was used, and no attempt 

 was made to attain the highest sensitiveness. With a scale distance not 

 far from 2.5 meters this galvanometer usually gave about 1 cm. deflection 

 for a current of 10 -8 amperes, without reversal. 



The current, about 25 amperes, which was sent through the iron bars, 

 was measured by means of a shunted ammeter, which, tested * in January, 

 1905, appeared to have, in the part of the scale used in the experiments of 

 this paper, an error less than one part in one thousand. Tested two or 

 three months later at the Jefferson Laboratory, on the basis of a Cromp- 

 ton standard resistance and a cadmium cell, it read about one half of one 

 per cent low in the same part of the scale. It is assumed in this paper 

 that the instrument gave correct readings as it was used in 1904. 



Disturbing Currents. 



Much care was taken to avoid accidental thermo-electric currents. 

 For example, unnecessarily thick layers of solder at joints between 

 pieces of copper were studiously avoided. It was not practicable, how- 

 ever, to get rid of mercury contacts, and slight differences of temjjerature 

 near such contacts sometimes gave currents sufficiently great to require 

 especial attention. The mere operation of a commutator having pieces 

 of copper dipping into little mercury wells, the reversals occurring rather 

 oftener than once a minute, appeared to produce a cumulative disturbing 

 effect, the slow increase going on for hours. The various reversals of 

 other connections which were regularly practised had a tendency to elim- 

 inate the effect of such disturbances from the final result ; but in some 

 cases it was necessary to measure the illegitimate current regularly and 

 make allowance for it. 



Deflections of the galvanometer were sometimes noticed which 

 appeared to be due to slight currents entering the galvanometer circuit 

 from some external non-metallic path. For example, when connections 

 were made for getting the reduction factor of the galvanometer, if the 

 circuit of this instrument remained connected at one point with the 

 thermo-electric couple wound on the cold end of either iron bar, a or ft, and 

 if at the same time the commutator leading the main current into the 

 iron bars was closed, the galvanometer showed a considerable illegitimate 

 current, the direction and magnitude of which were not changed by 

 reversing the commutator of the twenty-five ampere current, though the 



* By the kindness of Professor Laws at the Massachusetts Institute of Tech- 

 nology. 



