MEASUREMENT OF TEMPERATURE. 
183 
furnace ; this renders the wire very soft. (If the wire is not annealed, its resistance 
will be slightly diminished on the first heating, owing to a change of structure.) It 
is then wound on a glass tube and annealed again, to make it preserve a perfect spiral 
form. 
Measurement of the Resistance. 
If the resistance be too small, chance errors, such as change in the contacts, are 
unduly magnified. To make the resistance large involves either inconveniently fine 
wire or an awkwardly long spiral. In the actual experiments, therefore, spirals were 
used with resistances varying from 5 to 20 ohms. The most rapid and accurate way 
of measuring such a resistance is by the Wheatstone bridge method with an 
ordinary post-office box, and a sensitive Thomson mirror galvanometer with lamp 
and scale. Resistances of 10 and 1000 are taken out in the arms of the balance, so 
that the adjustable resistance is 100 times that of the spiral when the balance is 
obtained. This gives the value direct to four figures, and a fifth may be obtained by 
accurately observing the galvanometer throws and interpolating. 
The box used was a very accurate box of B.A. units by Messrs. Elliott. On the 
extraordinary accuracy of the boxes the reader may be referred to the paper by 
Glazebrook, “ On the Determination of the Ohm in Absolute Measure ” (see ‘ Phil. 
Trans./ 1883, p. 2G2), whence it appears that an uncertainty of 1° C. in the tem¬ 
perature of the box will give an error of about '0004 in the absolute value of a 
resistance, hut that ratios of resistances, if the temperature of the box be uniform 
may be determined with an error of only about *00002. 
The sensitive galvanometer employed had a resistance of about 14^ ohms, being 
approximately that most suited to the given arrangement of resistances. It was 
arranged so as to be extremely sensitive. A deflection of one scale division corre¬ 
sponded to a difference of *0001 ohm when the resistance to be measured was 
10 ohms. It was also nearly dead-beat to facilitate the reading of deflections for 
interpolation. 
The battery consisted of from four to six LeclanchA elements in series; as the 
circuit had a resistance of about 500 ohms, they worked steadily. 
Sources of Error. 
Thermo-electric effects were made as small as possible by careful arrangement of the 
junctions; they were then completely eliminated by using a rocking mercury cup 
commutator in the battery circuit, for rapidly reversing the current and observing 
when the reversal had no effect on the galvanometer ; it is important, however, to 
make the thermo-electric effect small, because, if it is large and variable, it greatly 
interferes with accurate observation of the resistance by making the galvanometer 
unsteady. 
There were only two screw contacts; these were between the thick copper leads 
