14 REPORT 1876. 



Preliminary experiments. — The preliminary experiments gave the differ- 

 ences between the coils and the variation-coefficients approximately. 



The results appeated in some cases different from former measurements, 

 so that it was thought better not to rely on these, but to make a more careful 

 set of experiments on which to found the final comparison. 



Ai^proximate coefficient of " Flat Coil " and Middle Coils. — The variation- 

 coefficient of the " Flat Coil " was taken from the preliminary experiments. 

 This was given by a fairly good series of ex]5eriment8 ; and a first approxi- 

 mation was considered sufficient, since the coil during the final experiments 

 never varied in temperature more tban two degrees, being always bathed in 

 the tap-water. A similar remark applies to the middle coils. The coils used 

 for middle coils were 29 and 43 (F and G) when neither of these was being 

 measured, in which case 2 and 3 (A and B) were used. The coefficients of 

 these coils, so far as required for small temperature-corrections, were taken 

 from the preliminary experiments. 



Method of experimenting. — The method used in the final experiments was 

 as follows : — 



First. AH the coils (the flat coil, the two middle coils, and the coil to bo 

 compared with the flat coil) were bathed in a stream of tap-water, the tem- 

 perature of which was carefully taken by means of a Casella's thermometer 

 (lent us by Mr. Gordon), reading to tenths of a degree Centigrade and easily 

 estimable to hundredths. After the temperature of the stream had been con- 

 stant for twenty minutes or so, the difference between the coil to be compared 

 and the flat coil was found. 



Secondly. Another series of experiments was made in which the flat coil 

 and the middle coils were kept at the temperature of the tap as before ; but 

 the remaining coil was raised by careful nursing, which lasted two hours or 

 more, to the temperature (or to one of the temperatures) at which, ac- 

 cording to the B.A. Report, it is correct. 



Lastly. The coils Avere compared with each other at the standard tempera- 

 tures, the middle coils being kept at the temperature of the tap-water. 



Variation-coefficients, how found. — The first two sets were used to give the 

 variation-coefficients, being peculiarly fitted to do so, because in them the 

 temperature of the flat coil did not alter much in comparison with the altera- 

 tion in the coil compared with it. 



Differences between the coils, how found. — Then using the low-temperature 

 experiments the differences of resistance between the respective coils and the 

 flat coil (all at lO"^ C.) were found. 



Control experinwnts, how used. — From this, of course, the difference 

 between any two coils at any temperatures could be calculated. This was 

 done for the old standard temperatures, and the results compared with the 

 results of direct experiment obtained from our third set of experiments. 

 This gave a test of the accuracy of our work ; and it is on this mainly that 

 we rely in claiming to have stated the temperatures at which the coils are . 

 equal within 0°'l C. in all cases. 



Degree of accuracy. — The degree of accuracy of resistance varies, of course, 

 for the different coils. For the platinum units 0"-l C. corresponds to a 

 variation of "03 per cent, resistance, for the platinum-silver to about -002 

 per cent. 



In the B.A. Eeport, 1865 (p. 303)*, aUthe coils are stated to be accurate at 

 the temperatures indicated within -01 per cent. This corresponds to about 

 one thirtieth of a degree Centigrade for the platinum units. It is not stated 



* Eeprint, p. 137 



