48 Reynolds and Moorby, Equivalent of Heat 



(Pi) used in the discharge pipe the corresponding rise was 

 o*oo66*^ per inch of mercury. 



The pressures under which the initial temperature 

 thermometer (Q2) worked were different, according to the 

 load carried by the brake. Taking these varying pressures 

 into account, a correction factor to be applied to the final 

 result was obtained. Jts value was (i— 0'00037). After 

 applying this correction it is extremely improbable that 

 any error exceeding yip^'^F., or one part in 18,000, can 

 have existed in the calculated mean rise of temperature of 

 the water in any trial. 



7. The main experiments. 



The trials were begun on Feb. 5th, 1896. The accepted 

 trials were made in 1 1 different series, each series con- 

 taining generally 3 pairs of heavy and light trials. Between 

 each series, some slight alteration was made in the 

 apparatus or method of working, all the alterations leading 

 up to the methods finally adopted, and which have been 

 described. 



Series L 



Contained 4 pairs of trials, giving 4 separate deter- 

 minations of K. This series differed froni all that followed 

 it in that the outer brass skin of the brake was exposed 

 directly to the atmosphere, and consequently the radiation 

 constant had the large value of 36'86. 



The mean value of K obtained from the series was 

 777'8i. The maximum value being 778"56 and the mini- 

 mum 777*02. This series of trials has not been allowed 

 any weight in the final result, because in some of the 

 trials the necessary observations of the temperature 

 gradient in the shaft were not taken, and, consequently, 

 the conduction correction could only be approximated to 

 by assuming this gradient the same in all the trials, since 

 the supply of cold water to the stuffing-box on the shaft 



