146 ROYAL SOCIETY OF CANADA 
the general principle being identical with that of the weight thermo- 
meter, without, of course, involving the coefficient of expansion for 
the glass dilatometer, as the observations before and after were both 
taken at zero Centigrade. Three determinations were made, the 
results showing a remarkable agreement, the mean value being 
0-91685 + -00008. 
This value, determined by three observations alone, has been the 
accepted value of the density of ice since 1870. This quantity, how- 
ever, is of great importance, especially in connection with the use 
of the Bunsen ice calorimeter, and this fact led to a redetermination 
of this value in 1899 by Nichols. 
Professor Nichols first determinations were made by the 
dilatometer method, using a specially prepared instrument, resem- 
bling a combination of both a specific gravity bottle and a Bunsen 
ice calorimeter (see fig. 4). The principle employed in freezing the 
water was identical with that of the Bunsen calorimeter, and the 
method of observing the increase in volume of the water was the 
same as that used by Bunsen in his density of ice determinations, 
the density being arrived at by two different methods of calculation 
from the different weighings, both results, however, being obtained 
from the identical experiments. These two values, 0:92154 and 
0:91631, are far from showing good agreement, this fact being espe- 
cially pointed out by Nichols and used as an argument against the 
dilatometer method, when mercury is used to fill the instrument. 
The great divergence shown by these two results is probably due in 
a large measure to the deformation of the walls of the instru- 
ment, caused by filling the dilatometer completely and afterwards 
partially with mercury. 
These considerations led Professor Nichols to abandon this 
method and to devise another, in which the results would be more 
certain and consistent. The next method tried was that of weighing, 
first in air, and then in refined petroleum. Great precautions were 
taken in the details of the experiments, the density of the petro- 
leum being found by weighing a piece of glass in distilled water, and 
then in the petroleum, the coefficient of expansion of the materials 
being allowed for. All the weighings were likewise reduced to 
weights “in vacuo.” 
The main object of this series of experiments was to determine 
the variation in density, if such existed, of specimens of ice formed 
under different conditions and obtained from different sources. Four 
varieties of ice were thus experimented on, and results obtained 
