68 MR F. K. SMITH ON THK CONSTi;i T (TK)X <>F 



sign. This was conveniently accomplished by integrating the area contained by the 

 whole curve, the extreme onlinates and an abscissa. A mean line representing a mean 

 thread length was thus obtained, and the positions of x and y chosen relatively to 

 this. A second integration of the chosen length followed, necessitating in a few cases 

 a slight adjustment of the favoured positions. 



The tube having been cleaned and dried by the method already descril>ed, 

 a mercury column was introduced. Its length and position were such that the 

 extremities of the column coincided very nearly with the x and y limits. The ends of 

 the tube were next capped with rubber tubing and glass plugs, wire being employed to 

 hind the tubing and so prevent leakage. This latter was an important consideration, 

 as the tubes were often immersed in water for several hours. The measurements of 

 length were conducted by Mr. B. F. E. KEELING, who kindly submits the following 

 report on the method employed and the degree of accuracy attained. 



Determination of Length (Mr. B. F. E. KEELING). 



The measurements of the mercury columns were made in a water-bath comparator 

 furnished with two micrometer microscopes reading directly to O'OOl millim. The 

 tube, with its mercury column, was closed at the ends with short lengths of india- 

 rubber tubing closed with pieces of glass rod, and was then strapped into a groove in 

 a stout brass bar. The clamps were sufficiently slack to allow of free contraction 

 and expansion. This bar and a standard metre of known value and coefficient of 

 expansion were supported side by side in the water bath, and were successively 

 brought under the microscopes. The order of the readings was as follows : Two 

 readings were taken on the metre, then three on the lines of contact of the mercury 

 and glass. The tube was then raised till the tops of the menisci were in focus, and 

 three readings were taken. Next, the tube and metre were put out of adjustment 

 and re-adjusted, and the readings repeated in the reverse order. The temperature of 

 the water was observed four times, two thermometers being used. 



Two metres were employed, one of nickel and one of brass, and both of these were 

 compared with the standard metre of the National Physical Laboratory. 



An example of the otaervations and necessary reductions follows : 



