10 Mr. E. H. Griffiths and Miss Marshall on the 



rate of revolution of the stirrer), and that this relation was 

 sufficiently close for the reduction of experiments in which the 

 value of r was somewhere between 5 and 6 per second. 



The value of Q s was found to increase rapidly as 6^ (the 

 temperature of the calorimeter) diminished. Our recent re- 

 determinations of Q s have explained a discrepancy which caused 

 an uncertainty in the former determination of Q s at 20° and 

 50°, and we now find that it was due to an arithmetical error 

 in the reduction of the observations. Fortunately, the resulting 

 correction in no way affects the conclusions arrived at in 

 Paper W ; and our present investigation confirms the accuracy 

 of the experimental determinations of Q s obtained in 1894. 

 As the extreme differences between any of our determinations 

 of Q s at temperatures above 20° (after the arithmetical correc- 

 tions at the 20° and 50° points) do not exceed 1 part in 70, 

 even if w r e include the values found in 1^94, and as the mean 

 probable error at any temperature is much below 1 in 100, it 

 is obvious that the values of Q s have been ascertained with 

 more than sufficient precision. 



8. tq = GoAW -V) + («* '-*)>. 



We have previously indicated (sect. 4, p. 6) the manner 

 in which 6 J and 0J' (the initial and final temperature of the 

 walls of the surrounding chamber) were determined by the 

 direct observation of mercury-thermometers. The value of d 

 (the difference between 6 X and 6 , i. e. the calorimeter-tempe- 

 rature and that of the surrounding walls) was ascertained by 

 the differential platinum-thermometers previously referred to, 

 and was always small — rarely greater than *01° C. It was 

 usually determined by the reading of the galvanometer-swings 

 without altering the position of the contact-maker on the 

 bridge- wire. 



When evaporating water, the values of d were so small that 

 an approximate value of the swing (in terms of a length of the 

 bridge-wire) was sufficient. During these benzene experi- 

 ments, however, we found it impossible to maintain the 

 thermal balance with such perfection. The only values of d 

 which were of consequence were d' and d' 1 (the initial and 

 final values), and, especially in the earlier experiments, the 

 internal temperature (^) rose so rapidly as soon as the last 

 drop of benzene had evaporated, that it was found impossible 

 to switch, off the electric current from the calorimeter- coil at 

 the precise moment necessary to reduce d lf to negligible dimen- 

 sions. As a consequence, the final swings (which were read 

 by means of a micrometer-eyepiece) were often considerable, 

 and therefore it was necessary to determine their value on each 

 occasion; for, although throughout a series of experiments at 



