26* Method of Measuring Electrolytic Conductivity. 



that taking only one determination, B, is probably correct to 

 1 in 2000 or -fa per cent. The probable error of the mean is 

 about 1 in 7000. 



Another source of error would arise if the heat caused by 

 the passage of the current through the electrolyte had not 

 sufficient time to escape into the liquid in the bath. The 

 current in our experiments would produce a rise of tempe- 

 rature of -J° C. per minute, but the surface exposed is so 

 enormous compared with the mass of electrolyte in the cross- 

 tube that calculation shows that the difference in temperature 

 between bath and electrolyte after vigorous stirring for two 

 minutes is infinitesimal. 



The temperature of the bath was read by means of a ther- 

 mometer graduated in tenths of a degree, and standardized 

 by comparison with a thermometer certified by the Reichs- 

 anstalt. It was, however, difficult to read the thermometer 

 with certainty to less than ^ ° C. The error arising from 

 this cause might amount to fa per cent. 



We estimate the possible error in the calibration of the 

 tubes to amount to not more than ^o P er ^ent., but this 

 source of error arises largely from the fact that we had not 

 in our possession apparatus for measuring the lengths of the 

 tubes to less than 0*01 centim. With suitable apparatus 

 there would be no difficulty in determining the constant for the 

 tubes with much greater accuracy. 



The greatest source of error appears to be a chemical one, 

 viz. : — the preparation of the solutions, for while the one 

 solution prepared in one way by Dr. Ewan gave 1008*3 for 

 the conductivity, the second prepared by a different method 

 gave 1012'2. 



We have also made experiments to see whether the alter- 

 nating current method of measuring electrolytic resistance 

 would be improved by adopting the balancing-cell which we 

 have described. To express the results very briefly, we find 

 that for resistances not greater than 1000 ohms or thereabouts 

 it is distinctly an improvement, enabling dead silence to be 

 obtained in the telephone, while without the balancing-cell 

 there was always a feeble buzz. For resistances much greater 

 than 1000 ohms the introduction of the balancing- cell seemed 

 of no avail whatever. We did not pursue these experiments 

 with alternating currents very far, because it seemed clear 

 that there was very little chance of getting as great accuracy 

 as with the continuous current method described. 



