1091 
worse as more air was condensed and therefore had passed through 
the capillary. Usually the disturbance could be reducetl or obviated 
entirely by operating the stirrer while the air was being compressed 
into the observation-tube. 
The uncertainty in the computation of the density arises in the 
following manner: beside the observation-tube a long piece of the 
glass ¢apillary is inside the cryostat, in our experiments about 35 cms 
with a volume of a quarter of the tube. Only a part of this length 
is immersed in the cold liquid; as previous experiments have shown, 
the temperature of the tube above the liquid increases pretty rapidly 
and near the top approaches the normal. A considerable correction 
has to be applied to the measurements for the gas inside this capil- 
lary. Assuming that the experiments are exclusively conducted with 
compression, the air which enters the capillary will retain its com- 
position in the incompletely cooled part and will have to be taken 
into account as air. In the lower part on the other hand the air 
will separate into liquid and vapour and, if the liquid flows down 
properly, this part will finally contain saturated vapour. Still, owing to the 
circumstance, that thorough stirring can only take place in the obser- 
vation-tube at the bottom, there is no guarantee, that the vapour in 
the capillary has the correct composition, while on the other hand 
the stirring in the capillary has the disadvantage of a partial mixing 
of the gases in the cold and the warmer parts of the capillary. As 
the extent to which these factors come into play is unknown, it is 
impossible to take them into account and the upper portion has to 
be taken as air, the lower portion as saturated vapour. In our case 
there was the additional difficulty, that the density of the saturated 
vapour was not yet accurately known, as it can only be found by 
interpolation from measurements of the vapour-density of a number 
of mixtures. In the mean time we had to be satisfied with an 
estimate. It may be added, that in a determination of a vapour- 
density, ie. with only a trace of. liquid in the tube, the entire 
cooled portion (observation-tube + part of capillary) has to be con- 
sidered as containing saturated vapour; similarly in a measurement 
near the critical point, when on stirring the liquid surface flattens 
out and disappears, the same volume has to be assumed in the cal- 
culation as being filled homogeneously. The various uncertainties 
arising -from the sources mentioned show themselves in small ir- 
regularities in the results which were obtained. 
On the above grounds it is our intention in future experiments 
to return to the ordinary method of stirring notwithstanding the 
clumsy dimensions of the electro-magnet which it involves: the glass 
