528 Proceedings of the Royal Society of Edinburgh. [Sess. 
are undoubtedly the most accurate data on this subject. The algebraic 
sum of all the temperature differences, divided by the whole number of our 
observations (27), gives the mean divergence from H. and H.’s curve of a 
curve drawn through our observations. This mean divergence is — 0'0063°, 
which is well within the limit of error assigned to our measurement of 
absolute temperature (± 001°). This corresponds to a mean pressure 
deviation of + 008 mm. 
Comparison with other Methods . — We may now institute a comparison 
between the results obtained by our method and those obtained by other 
methods, using Holborn and Henning’s values as the standard of comparison. 
There are only two sets of static determinations in the region 50-90°. 
The mean divergence of Magnus’ results in this region is + 0‘81 mm. (or 
— 0'07°), and that of Batelli’s, +2*3 mm. These are respectively ten and 
thirty times as great as our divergence. Ramsay and Young’s static 
observations began at 120°, and the mean deviation of their values between 
120° and 150° from those of H. and H. is — 0T°, or of the same order as 
that of Magnus. 
There are three sets of observations by dynamic methods in this region, 
one of which is that of H. and H. Regnault’s mean divergence is +0*4 mm. 
(or — 004°) and Wiebe’s — 0 25 mm., respectively five and three times as 
great as ours. 
The deviations of the recalculated values are as follows: Brock +027 
mm., Thieson — O il mm., Eckholm — 0 33 mm. 
Assuming the correctness of Holborn and Henning’s values, this compari- 
son shows that the present method, in addition to the advantages already 
enumerated, possesses also that of giving accurate results. The degree of 
self-consistency of the observations made by this method may be seen by 
inspection of the table, and a more detailed study of this point will be given 
in the following paper, in connection with another example. 
( Issued separately September 7 , 1910 .) 
