226 PROCEEDINGS OF THE AMERICAN ACADEMY. 



introduced by the glass envelope is only 2.5 per cent as against 60 per 

 cent in determinations of the compressibility of mercury. Hysteresis 

 and other irregular action will appear, therefore, simply as perturba- 

 tions of this 2.5 per cent correction. There are a number of smaller 

 sources of error, which, even though very obvious, will be mentioned 

 as occasion presents, because in the justification of a new standard it 

 seems well to record all the sources of error that were considered or 

 guarded against. 



The electrical measurements were carried out on a bridge of the 

 Carey Foster type provided with an eight point mercury switch. 

 The variable mercury resistance took the place of one extension coil, 

 and the other was a manganin coil of approximately ten ohms. Meas- 

 urements were made by setting the slider for no deflection, this being 

 preferable to measuring the current by ballistic or steady throw of the 

 galvanometer. A D'Arsonval galvanometer of low resistance was used, 

 of sensitiveness great enough to indicate changes in the position of the 

 slider of less than -fa millimeter. Extension coils and balancing coils 

 were of seasoned manganin, all approximately ten ohms. In com- 

 paring together two mercury resistances the same balancing and ex- 

 tension coils were used, the bridge being provided through leads of | 

 in. copper wire of negligible resistance with two slide wires one meter 

 long. The slide wires w T ere interchanged by mercury switches fre- 

 quently cleaned. The resistance of the extension and balancing coils, 

 as well of the bridge wire, was measured against standard manganin 

 coils known to be correct to 0.01 per cent, which were kindly loaned 

 for the purpose by Professor B. O. Peirce. The bridge wire was cali- 

 brated for uniformity by stepping off on it a resistance equivalent to 

 approximately 10 cm. at 3 cm. intervals. The maximum correction 

 of one wire was 0.4 mm., of the other 0.7 mm. The average arith- 

 metic correction of the first was 0.17 mm., of the latter 0.4 mm. Ap- 

 proximately 33 cm. of either wire has a resistance of one ohm. All the 

 connections in the circuit were either soldered without acid for a flux 

 or were through mercury cups, except two connections at the insulating 

 plug leading to the mercury resistance, which were made with nuts. 

 As it was found that induction effects were unnoticeable, the bridge 

 was operated with the galvanometer circuit permanently closed, thus 

 eliminating the principal sources of thermal currents. Two readings 

 of every resistance with the extension coils interchanged were really 

 unnecessary, therefore; but they were always made so as to secure the 

 increased accuracy of two independent settings. Current was sup- 

 plied by a single Samson cell of about one volt, and was decreased by 

 inserting 100 ohms in the battery circuit. The current through the 



