as a Source of small Constant Currents. 357 



It is, however, also true that polarization is never sufficient 

 to cause an uncertainty of 0*5 per cent, in the value of a 

 current of nominally *001 ampere yielded by the kind of cell 

 about to be described. 



The greatest observed uncertainty during three months of 

 continued testing and varying of conditions was 0*6 per cent.; 

 of this probably only half at most was to be traced to 

 polarization. 



On an average the uncertainty did not exceed 0*1 per 

 cent. ; so that we consider that in stating that the Clark cell 

 may be used with security to give currents of a value not 

 differing by more than 1 per cent, from their nominal value we 

 shall leave a really unnecessarily large margin for uncertainty 

 in the actual value of a properly constructed external 

 resistance. 



In May 1888 we began to investigate the behaviour of the 

 Clark cell when giving small currents, with a view to apply- 

 ing it to the calibration at any time of a certain galvanometer 

 which we had designed for one of the hospitals of Sydney. 

 We were aware that the Clark cell had the reputation of being 

 easily polarized, but this seemed to us probably to arise from 

 the fact that the cells as usually made expose very small 

 surfaces of zinc and mercury. We therefore constructed 

 some cells of considerable size in order to find out whether 

 our idea was or was not correct. The cells constructed and 

 tested were the following : — 



(1) A set of 10 test-tube cells of the ordinary pattern. 

 The tube of each cell was three inches long and half an inch 

 in internal diameter. The area of surface of mercury exposed 

 was about 0*2 square inch or 1*3 square centimetre. The 

 cells differed considerably from one another as regards re- 

 sistance, the average resistance being about 40 1. ohms. 

 The positive electrodes were bits of " pure " zinc rod as 

 supplied by the dealers, J inch in diameter, but of rather 

 variable lengths. The areas of the zinc surfaces immersed 

 differed considerably in the different cells. These cells were 

 first set up in June 1886, and have been tested at intervals 

 ever since. Their behaviour is quite satisfactory with the 

 exception of that of No. 18, which was out by about *001 

 volt from the beginning. 



(2) Thirteen cells, similar to the above, made up in 1888 

 for the purpose of this investigation, and charged with a 

 sample of the same mercury as used by Lord Rayleigh, i. e. 

 taken originally from the same bottle. Our object in pre- 

 paring these cells was to connect our work directly with Lord 

 Rayleigh's measurements. 



