BETWEEN THE FREEZING AND BOILING-POINTS. 
255 
For Calorimeter E we have the two values at 29°’92 C. and 20°'92, which are 
‘05994 and ’05414. 
These give for the coefficient of t in the radiation expression, the value '000645 ; or, 
applying the first formula, the value of the radiation loss at 29°’92 from the value at 
20 o, 92 = ’06051. This is within 6 parts in 10,000, and is comparable in size with 
the variations from the calculated values for calorimeter C. Doubtless there would 
be slight differences in the temperature coefficient of the radiation loss for different 
calorimeters with different degrees of vacuum. 
In Series VI., for Calorimeter C, the decrease in the radiation loss takes place with 
decrease in temperature well in agreement with the other series until the experiments 
at 0°, when the value of the heat-loss is increased by nearly 3 parts in 1000. The 
two experiments at 1°’35 and 2°’68, both with the inflowing water at 0°T5 C., agree 
however very closely with the formula as regards the temperature change in h. The 
explanation of the apparent increase at these points is not altogether clear, hut may 
be looked for in the very high value of the specific heat of water in the neighbourhood 
of 0°, which would influence the validity of the method adopted of eliminating the 
heat-loss from the large and small flows. A similar increase, although much smaller, 
was noticed in the heat-loss for the same calorimeter at 4°, in Series I. Owing to the 
small conduction effect at the inflow end of the calorimeter, the water in the large and 
small flows enters the flow-tube, where it is heated by the electric current, necessarily 
at a slightly different temperature, as was pointed out before. 
Whereas this would produce no error at a part of the range where the value of d 
t 
was not changing rapidly with the temperature, at the freezing-point, where a very 
small difference in temperature produces a large change in the value of d, it cannot be 
regarded as equal in the difference equations for the two flows for the same value 
of dd. Taking this into consideration, I have calculated the value of d, for the two 
experiments under consideration, by extrapolating for the value of the heat-loss from 
the curve for the other observations in the same series between 20° and 8°. By this 
means, the value of d for each flow in the same experiment differs nearly 1 part in 
1000 in the extreme case. The following are the values so obtained :— 
Date. 
Mean temperature. 
cl large flow. 
cl small flow. 
November 18 ... 
o 
1-35 
+ -0066 
+ -0073 
„ 22 . . . 
2-68 
+ -0051 
+ -0060 
5) ... 
0-67 
+ ’0072 
The mean value for each experiment is larger than the value calculated in the usual 
way, but for the same value of the flow the values of d are very consistent for the 
