the Mercury Unit ivith the British Association Unit. 27 



and the current reversed, in order to eliminate any small 

 thermal current. 



Benoit has determined that the resistance of mercury is 

 decreased by the diffusion of copper amalgam from the ends 

 of copper connecting rods. Similar experiments, performed 

 at this laboratory, have demonstrated the fact that, by leaving 

 the rods dipping in the mercury in the end-pieces for twenty- 

 four hours, the resistance was decreased one part in twenty- 

 four hundred. Therefore, the decrease due to this cause in 

 our investigations must be inappreciable. 



In the cases of nine fillings, a small thermometer was 

 placed in the mercury cups immediately after removing the 

 rods. In no case was the temperature thus obtained greater 

 than 1°C, and the average of the nine sets of observations 

 was o, 5 C. Supposing that 6 centim. of the tube were at 

 this temperature, in the worst case the error would be '00003 

 high ; while, in the other cases, it would be about '00002 high. 



Table III. (p. 26) gives our final results. 



Owing to the tubes being entirely unprotected, great care 

 was necessary in handling them. Tube III. was broken in 

 our preliminary work, in trying to remove the end-pieces, 

 which were then fitted with perforated rubber stoppers. 

 These were after this replaced by common corks. 



Tube I., whose bore was 1'76 millim. in diameter, allowed 

 the soft rubber thimbles to sink into it and thus cause the 

 column of mercury weighed to be smaller than it should have 

 been. A correction should be applied on this account and 

 might have been obtained by jacketing the tube with a water- 

 jacket and then taking a number of alternating observations, 

 using the thimbles and then glass plates. Unfortunately, this 

 tube was broken after all the other observations had been 

 completed. Its average — manifestly low — is not included in 

 the final result. 



Only one observation is neglected in the remaining tubes — 

 II., IV., V., VI. An error of one part in a thousand is 

 apparent in observation No. 1 of tube IV. , which is the case 

 referred to. 



In the final result there are, therefore, on 

 Tube II., 7 observations. 



„ V., 5 

 „ VI., 4 



Total . . 19 



?; 



VI 4 



