BETWEEN THE FREEZING AND BOILING-POINTS. 
241 
absolutely air-free water, as described in Section 5, and from this time on the work at 
the higher points progressed more favourably. In this series observations were 
obtained at 30° and 86°, but the latter unfortunately with only one flow. The second 
flow could not be taken on account of the rapid evaporation of water from the tank, 
the constant-level device not being used at that time. Here again, extrapolation for 
the heat-loss from the value at 30° can be made, but the procedure can hardly be 
justified with a greater accuracy than 1 in 1000. The agreement with the later 
results is, however, extraordinarily good. 
Series 8. Nos. LI. to LV. 
This series was made with box 2 and Thermometer E, involving the F.I. correction 
in Group IV. It extends from 32° C. to 92° C., and in many respects is the most 
important series of the whole. The Calorimeter C was refitted throughout so as to 
give an entirely new set of observations. The complete agreement of the measure¬ 
ment of the specific heat at 32° C. with the other measurements with box 1, eliminates 
any possible error due to the box and its connections. The four sets of observations 
at the higher points are exceedingly consistent, and distinctly show that the previous 
trouble to obtain the measurements in this region was due to the effect of the air in 
the water. Not the slightest trouble was experienced with air making its appearance 
in the calorimeter in these experiments. The order in which these observations were 
taken was as follows : 32°, 74°, 92°, 80° and 68°, between which the calorimeter 
cooled down to the temperature of the laboratory, and had to be heated up to the 
desired point each time. The measurement at 86° in Series 7 is in very good agree¬ 
ment with these. The beautiful consistency of this last series of measurements might 
make it desirable to repeat the observations between 50° and 60° C. with air-free 
water. I did not consider this was necessary, however, as the continuity of the 
observations at the two ends of the range is so good, and the divergence in the results 
obtained between 50° and 60° C. is so clearly explained by unforeseen and extraneous 
causes. 
The calculation of the results in the tables just given for the determination of the 
value of the electrical and heat energy has been very much facilitated by the use of 
the Brunsviga calculating machine, which is very much to be recommended for this 
class of work. For the application of the small correction factors, and for the final 
estimation of the values of d and h, the Fuller cylinderical slide-rule has been 
constantly used. The values of the electrical and heat watts given in the summary 
at the foot of each table may therefore be in error by 2 or perhaps 3 in the fourth 
decimal place, but no more, but I feel confident that in the estimation of the mean 
value, upon which the value of d depends, this error tends to disappear, and that the 
value of d given by the measurements in any of the tables, represents the observa¬ 
tions to an accuracy of 1 part in 100,000. 
VOL. CXCIX.—A. 2 I 
