152 REPORT— 1898. 



shutting a stop-cock, it expanded into one arm of a glass U tube, 

 thereby forcing a mercury index at the bottom up the other arm ; this 

 index cut off the gas supply by closing the aperture of the inlet tube. 

 On cooling the index sank ; the gas was turned on and relighted by a 

 bypass. Regulation of the temperature accordingly did not take place 

 until the path leading to the reservoir was closed, so that regulation at 

 any desired temperature could be obtained by leaving the stop-cock open 

 until that temperature had been reached. In this case, as also in the 

 other thermostat, the bath was not heated directly by the gas jet, but a 

 baffle plate was interposed. The daily variation of temperature with this 

 apparatus was about 0-2° C, but the changes were so slight and so slow 

 that the probable error introduced would be less than that caused by 

 error in reading the thermometer. The Ijath was always kept at a constant 

 temperature for some hours before readings were taken. With these 

 arrangements it was safe to assume that the temperature of the coil was 

 the same as that read off from the thermometer. The terminals of the coil 

 dipped into mercury cups in one end of a pair of stout copper rods, half 

 an inch in diameter, the other ends of which rested in mercury cups on a 

 Carey Foster bridge. The leads from each of the coils were of very 

 small and approximately equal resistance, so that no appreciable error 

 could be introduced by alteration in their resistance with change of 

 atmospheric temperature. Also, as a part of each lead was inside the 

 thermostat, heat lost by conduction along tlie leads would be withdrawn 

 from this part and not from the coil itself. 



The measurements were made with a Carey Foster bridge, the platinum 

 silver slide wire of which had been previously calibrated. This wire was 

 50 centimetres long, and had a resistance of 0-001859 ohms per half 

 centimetre at 13-5° C, and was graduated in lialf millimetres. Correction 

 was made for alteration in the resistance of the wire due to change in its 

 temperature, an increase of 1° C. producing au increase of 0-000011 ohms 

 in the resistance of half a centimetre. Determinations of the difference 

 of resistance between the two coils were made at intervals of about an 

 hour, and if two or three quite consistent readings could be taken these 

 were considered as correct, but where discrepancies occurred the mean of 

 .several results was taken. The slight changes in the temperature of the 

 standard were easily allowed for, since it could be assumed that for such 

 small changes the two coils had the same temperature coefficients. So if 

 the standard, instead of being at the tempei'ature t, were at the tempera- 

 ture t + Z, and if the coil under test were at the temperature t', it was 

 assumed that the standard was at temperature t, and the coil under test 

 at the temperature t' — 8. 



There are four principal sources by which error can be introduced — 

 viz., error in obtaining the correct position of balance, error in the value 

 of the temperature coefficient of the slide- wire, error in reading the 

 temperature of the standard coil, and error in reading the temperature of 

 the coil under test. As regards the first of these, the sensibility of the 

 arrangement was such that a change of half a millimetre in the position 

 of the slider produced a deflection of about a centimetre on the galvano- 

 meter scale, so that balance could easily be obtained correct to 0-05mm. 

 The error due to not knowing the temperature coefficient of the slide- wire 

 with certainty will not be great, as all the measurements were made at 

 temperatures near to 13-5° C, at which temperature its resistance was 

 known. The greatest error is introduced in reading the thei'mometers 



