B. A. Unit of Resistance. 233 



larger concentric cylinder. The terminals are soldered to the 

 copper bottoms of mercury cups, arranged in two circles around 

 the hard rubber ring which closes the annular space between 

 the inner and outer copper cylinders. The inner cylinder is 

 filled with water. The connections of the ten coils can be 

 varied at pleasure : they can all be thrown in series, in parallel, 

 or in any intermediate arrangement. There were two compara- 

 tors used, with the coils 100 and 10 ohms respectively. 



To keep the temperature constant, spirals of lead pipp were 

 placed around the Elliott coil and in the inner cylinders of the 

 comparators, through which there was a constant flow of water 

 from the city supply. This answered its purpose admirably ; 

 the temperature varied only a degree or so from day to day 

 even. The water, was of course, allowed to flow some hours 

 before beginning observations. 



The terminals of all resistances were brought to large mer- 

 cury cups (m, m) each having an amalgamated copper disc lying 

 on the bottom. The main current did not flow through any 

 part of the circuit of the induced current ; a short bridging 

 piece (d) is used, as Lord Rayleigh found necessary. 



The resistances used were all compared several times by 

 different observers with the standard : this was a Warden 

 Muirhead 10 ohm coil, whose value was determined at the 

 Cavendish Laboratory in 1887 ; it was 9*99416 B.A. units at 

 16*5°, with temperature coefficient of '000292 per ohm per 

 degree. 



In taking the observations, the aim was to adjust the resist- 

 ances first so that there should be only a small deflection ; 

 after a number of galvanometer readings for this "balanced " 

 arrangement had been taken, the resistance " c " was changed 

 so as to give a deflection of 10 divisions (say) ; readings were 

 taken for this " unbalanced " arrangement ; the original " bal- 

 anced" was then restored and readings taken. If nothing had 

 changed sensibly since the beginning of the experiment, the 

 average deflections for the two " balanced " would agree ; of 

 course, this condition was only approximated to. The " un- 

 balanced" set gives the data for correcting for the small deflec- 

 tion of the " balanced." 



Each experiment then consists of the galvanometer, speed, 

 and temperature readings pertaining to the three arrangements 

 of resistances : these three arrangements are called " A," " B," 

 and "A 7 " in the order taken, irrespective of the magnitude of 

 the deflections. In general, R x and R 3 (subscripts 1, 2, 3, refer 

 to A, B, A.', respectively : R is the " effective " resistance) are 

 the same, and the corresponding deflections are small ; B is in 

 this case used to correct both Rj and R 3 , and the mean of the 

 corrected values is used. When, however, the deflection for 



