Palmer — Pressure Coefficient of Mercury Resistance. 5 



investigation the resistances in the various arms of the bridge 

 were so proportioned as to give a maximum of sensitiveness, 

 and a movement of the. spring contact on the wire equal to 

 one-tenth of a division was always sufficient to reverse the 

 direction of the galvanometer deflection when balance was 

 obtained. During the measurements at 9° C. no difficulty was 

 experienced from thermoelectro- motive forces, since the water 

 cooler was long enough to keep all the joints at the same tem- 

 perature, but when the steam jacket was employed they caused 

 so much trouble that it became necessary to replace the copper 

 connections inside of the piezometer by iron wires. Disturb- 

 ances of this nature were thus reduced to a minimum and 

 trustworthy resujts could be obtained by closing first the galva- 

 nometer and then the battery circuit. The temperature of the 

 room and of the standard resistances, determined by a small 

 mercurial thermometer placed between the coils, remained 

 nearly constant during the actual time of observation but 

 varied considerably from day to day. 



If K represents the resistance of the mercury thread and W 

 that of the standard of comparison and if x and x' are the 

 readings of the bridge-wire micrometer for the position of bal- 

 ance before and after they are interchanged, we have 



U = W + r(x-x') 

 where r has the meaning and value assigned to it above and 

 all the connection resistances are eliminated except those 

 between R and W and the commutator. To determine these 

 the mercury tube was replaced by a thick copper wire soldered 

 to the same connecting wires and measurements were then 

 made under as nearly as possible the same conditions of pres- 

 sure and temperature that were used with the mercury. The 

 mean of a large number of observations gave '0632 ohms with 

 the copper connections used at low temperature and -5095 ohms 

 with the iron ones used«at the boiling point, and no variation 

 with the pressure could be detected. In reducing the resist- 

 ances to the standard temperature of the Queen box the bridge 

 wire was assumed to have the same temperature coefficient and 

 to be always at the same temperature as the standard coils. 

 This assumption could introduce no appreciable error in the 

 results, since the factor r(x— x') was always less than 0*1 ohm 

 and the temperature of the room never varied much from that 

 of the box, but it greatly simplified the calculation of the cor- 

 rections. It was further found that the slight variations in the 

 temperature of the mercury thread, from 9° C. in one case and 

 from 100° C. in the other, introduced errors that could not be 

 neglected and corrections were introduced using '0009 as the 

 temperature coefficient of mercury. Finally the effect of 

 changes in the volume of the glass tube, due to compression, 



