370 MESSRS. R. T. GLAZEBROOK AND T. C. FITZPATRICK 



Table VIII. gives the results of the observations on the one-half and one-third unit 

 tubes. One other filling of I., which is not given, was taken, but it was clear that 

 some accidental error had been made in the measurement of I or of W, for the result 

 differed from the others by over one in a thousand. This is the only filling in the 

 whole series which has been entirely rejected. The result of V. 1 is clearly too low. 



If we give the result of all the fillings in Table VIII. equal weight, we obtain 

 as the mean value r = '95342 B.A. unit, and the average error is '000029. 



The difference between this and the result derived from the one unit tubes is, at 

 first sight, considerable, but it must be remembered that for the tubes I. and III. an 

 error in the measurement of resistance of '00003 or one thirty-thousandth of a B.A. 

 unit gives rise to an error of one ten-thousandth in the result. This, of course, is an 

 extremely small quantity. Moreover, the small uncertainty which attaches to the 

 corrections SL and 81 would produce a larger effect in these large tubes, and our 

 observations tend to show that the coefficient "82 in the value for SL is possibly 

 rather too great. 



It will be noticed, however, that the values of R, the actual observed resistance of 

 a tube, differ among themselves by extremely small quantities. 



In taking a final mean, it was clearly unfair to weight the observations equally, and 

 we came to the conclusion that the probable accuracy was roughly inversely propor- 

 tional to the area of the cross section. We have, therefore, attached the weights 3, 

 2, and 1 to the results from the one unit, half unit, and third unit tubes respectively, 

 and arrive at the final result that the resistance of a column of mercury 1 metre long, 

 1 square millimetre in section, at is 



95352 B.A. unit. 



If we give equal weights to all the observations, the result will be '95351, so that 

 the effect of the weighting is hardly appreciable. 



It remains to consider the four special fillings of tubes VI. and VIII., which have 

 been omitted. In two of these, VI. 4 and VIII. 4, the attempt was made to fill the 

 tube quite full, when measuring I ; a very small bubble of mercury was left protruding 

 from each end of the tube when it was placed under the reading microscopes, and 

 then flat pieces of brass were brought up simultaneously against the ends, there being 

 a layer of thin paper between the brass and the mercury. It was hoped in this way 

 to squeeze out the superfluous mercury and leave a column with flat ends exactly 

 filling the tube. It seems probable in the case of VI. 4 that this was successfully 

 accomplished, for the value of r found from the experiment is '95354 B.A. unit. 



With VIII. 4, however, it was clear, on looking through the microscopes at the 

 mercury column, that the ends in contact with the paper were slightly curved, and 

 this was still more obvious when the brass and paper were removed. The resulting 

 value of r is accordingly too low, being '95342 B.A. unit. 



In the case of the two fillings, VI. 7 and VIII. 6, mercury was used which had 



