280 PROCEEDINGS OF THE AMERICAN ACADEMY. 



from H to give L, was in each case accomplished independently of, and 

 to a minor extent in disagreement with, the judgment of the other, 

 but in each case the greatest difference between the computed values 

 and those actually observed by Knoblauch, Linde, and Klebe 27 was 

 under two tenths of one per cent, and in each case the average of the 

 deviations was about one tenth of one per cent and they were nearly 

 equally divided between plus and minus. It is probable that some of 

 these deviations may properly be attributed to errors in the observed 

 values. 



The accuracy of the new H formula can now be estimated. It has 

 been pointed out that the self-consistency of the computed points indi- 

 cates a precision of the order of a twentieth of one per cent. The 

 actual error is probably larger than this because of systematic errors 

 in Knoblauch's specific heats. It is possible that these will ultimately 

 be raised enough to make Hi$ a tenth or even a sixth of one per cent 

 larger. Inasmuch, however, as the other two tests which have been 

 applied, based on Henning's direct measurements of H and on Knob- 

 lauch, Linde, and Klebe's volume measurements, have both led to an 

 estimated accuracy of a tenth of one per cent or better, a part of the 

 outstanding disagreement in each case being furthermore reasonably 

 attributable to possible errors on the observed as well as on the com- 

 puted side of the comparison, it would seem that a claim of a tenth of 

 one per cent for the accuracy of the new H formula between 100° and 

 190° is justified. 



B. The value of H XO o- — In what is to follow a suitable value for H 1W 

 will be necessary. Henning's work has already been shown to lead to 

 the value 



H l00 = 639.26 mean calories (Henning). 



Another available value is that of Joly 28 who compared the latent 

 heat of steam at 99.96° with the mean specific heat of water between 

 11.89° and 99.96°. The latter number is 0.99949 according to the 

 curve used in the steam tables already mentioned. The resulting value 

 of i/ioo is 



Hioo = 638.82 mean calories (Joly). 



In this determination of H 1W Regnault's measurements will not be 

 considered at all. They show unmistakable evidence of running 

 lower than they should, probably for the same reason that makes 



•» Forschungsarb., 1905, 21, 33. 

 28 hoc. cit., on page 268. 



