964 Proceedings of Royal Society of Edinburgh. [skss. 
If the values just given for C be compared with those obtained 
when elevations above the boiling point of the solvent were used, 
as in the calculations on page 955, it will be obvious they are in 
much greater harmony for the same series. It will also be seen 
that the values of C for one series, when compared wuth those for 
another on the same salt, are in greater harmony ; and the increased 
agreement is maintained when the values for series on different 
salts are considered. The agreement of the values with theory is 
also much greater. 
Most of the values obtained for C indicate an increase of the 
value with increase of concentration ; but it would be premature 
to draw that inference, as in solutions so dilute as some of those 
used, a small ionization error would cause a considerable variation ; 
but in addition it has to be remembered that for some of the 
dilutions used, one-thousandth of a degree variation of boiling 
temperature would have made a difference of 10 in the value of 
C, and for any of the concentrations a few thousandths of a degree 
would have made a considerable change. It has also to be borne 
in mind that the barometer used, although an excellent instru- 
ment, only read to the equivalent of about four thousandths of a 
degree, and was estimated only to two thousandths. The first 
five values obtained for sodium nitrate in the first series are 
very approximately the same : during the time this series of 
observations was taken, no barometric change was detected. For 
potassium bromide the barometer remained steady for the last 
three experiments of the second series ; and here, again, there is 
closer agreement than usual in the values of C. Also for the first 
series on potassium nitrate the barometer was steady, except in 
one observation, and here likewise the harmony of the values is 
increased. It would therefore seem probable that barometric 
changes, even after eliminating their effects by the ordinary 
method, have had considerable influence in producing variations 
in the value of C ; consequently, were these deducted, the changes 
of value would be proportionately diminished. This is a source of 
error which it will be possible more fully to eliminate in later 
experiments. 
The advantages of the method adopted become more apparent 
when it is applied to find molecular weights. 
