SOMK MKIMVKY STAN'DAKDS OF IlKXlSTANCi:. II. 



7 



majority of cases \\ere designated by the numbers 0, 5, 10, 15, ... and multiples 

 of 5. The variations in the mean cross-sections of these 5-centim. lengths was thus 

 obtained and enabled p.' (1st series) to be evaluated. (See Table XVIII.) 



Next, the thread lengths for the positions 1, 6, 11, 16, Ac., centims. were measured, 

 thus dividing the tul>e into a new series, the p.' of which should le approximately 

 equal to that of the first. Similarly, the third, fourth, and fifth series were 

 determined, the whole of the measurements occupying alniut nix hours. 



No correction for varying temperature was necessary, for the room in which the 

 calibration was conducted was maintained at a constant temperature by a silver 

 ammonium-chloride thermostat. The small variations due to the proximity of the 

 olwerver were found to be too small to affect the final results. 



With respect to the uniformity of the tul)es, the following table, giving the values 

 of the conical corrections for mercury standards constructed by different observers is 

 of interest. The same method of calculation has not, however, l)eeu generally 

 employed. 



TABLE II. 



1>< termination of L cind W. 



The best approximate positions for the extremities of the mercury standard were 

 decided from a preliminary evaluation of L' : /W in conjunction with the formula given 

 on p. f)3. \Vhene\er possible. ./ and // were chosen so as to lie equal and of opposite 



K -2 



