150 ROYAL SOCIETY OF CANADA 
ture the ice density is required. The method is, in fact, doubly 
indirect, as the density of the ice is compared with that of the liquid 
used, the specific gravity of this liquid having to be determined by 
separate experiment, thus there is an extra process introduced, involy- 
ing an additional liability to error. The liquid used was petroleum, 
which does not act on ice at temperatures below —1° C., this being 
the highest temperature at which any of the determinations were 
carried on. 
Professor Nichols’ results are open to precisely the same objec- 
tions as Brunner’s, being carried on at temperatures below the freezing 
point and involving a separate determination of the density of the 
liquid used. 

| 
5 
Fic. 4 NICHOLS  DILATOMETER. 
We are now in a position to summarize the relative advantages 
and disadvantages of each of these three general methods. 
The weak point of the first method, that of adjusting the density 
of the suspending liquid until the ice is in equilibrium, lies in the uncer- 
tainty of each separate experiment — the method is, in fact, not sensitive 
enough for the requirements of the determinations. 
The dilatometrie methods used by Kopp, Plücker and Geissler, 
and by Bunsen, each have some uncertain feature peculiar to the 
particular instrument used. 
The objections to the method of weighing are that the experi- 
ments are not carried on at zero, and the density of the liquid has 
to be determined by a separate experiment. The application of the 
coefficient of expansion in order to reduce the density to zero is not 
a serious objection, though it may be possible that this coefficient, 
which was determined for temperatures ranging from —1° to 
—20° C., may not hold between say —1° and 0°, owing to 0° being 
the point where a change of state takes place; nor is the separate 
experiment for determining the specific gravity of the liquid a serious 
objection, as this is done by the same method as the ice density 
determinations, and is capable of the same degree of accuracy. 
From this it will be seen, that it is the method of weighing from 
which one would expect the most accurate results, and, as a matter 
