24 



Starkweather— -RegnauW s Calorie and Our 



satisfy Regnault's determination at 100°, this being by far the 

 most certain of any, the possibility of a linear formula is thus 

 precluded, although it is evidently nearly such. 



If we attempt to form an expression for H to satisfy 

 Dieterici's value at 0°, Griffiths' values from 25° to 50°, and 



Regnault's value at 100°, it will make —77- positive instead of 



at 



negative, will not agree with Regnault's determinations from 



60° to 90°, and will make an abrupt break at 100° where it 



joins the formula for the region above 100°. If one neglects 



Dieterici's determination and forms a formula to satisfy Reg- 



nault's value at 100° and Griffiths' determinations, and to join 



smoothly with the formula above 100°, it will give a value at 



0° far below Dieterici's. The writer has therefore formed the 



following equation, using Dieterici's value at 0°, Regnault's H 



jit 



at 100°, and making — at 100° equal to that by the formula 

 at 



used above 100° : 



II = 598-9 + 0-442 t — 0-00064 f. 



A comparison of Regnault's experiments with this formula 

 follows, the fractional deviations being given : 



1 



+ •0018 



6 



-•0040 



11 



-•0029 



16 — -0031 



21 +-0002 



2 



+ •0018 



7 



— •0040 



12 



+ •0001 



17 — '0061 



22 — -0016 



3 



— •0051 



8 



•0000 



13 



+ •0011 



18 +-0009 



23 +-0025 



4 



— •0041 



9 



— •0029 



14 



— •0037 



19 +-0009 





-5 



+ •0009 



10 



— •0039 



15 



+ •0004 



20 — '0050 





The agreement is far better than by Regnault's formula. 

 Indeed a much better agreement could not be obtained by any 

 formula which joins smoothly with one which fits the experi- 

 ments above 100°. If there is a question as to which set of 

 experiments to fit best, those above 100° or those below, the 

 answer is immediately those above, for not only are the 

 experiments below 100° in less concordance with one another, 

 as shown by the deviations just tabulated, but Regnault him- 

 self says that they were more difficult to conduct, the boiling 

 not taking place steadily, but in great puffs at intervals. The 

 expression for H will be considered only as provisional, how- 

 ever; the subject of total and latent heats below 100° will be 

 discussed again in another paper. The formula does not agree 

 any too well with Griffiths' latent heats (adding to them A), the 

 fractional deviations of the latter from the former being about 

 -•005. 



It should be observed that this formula, as well as the 

 formula for total heats above 100°, holds independently of the 

 judgment made that Regnault's calorie is equal to the specific 



