22 



Starkweather — RegnauWs Calorie and Our 



— •0051 and the minimum — -0035, the sum of the squares of 

 the differences from the average being almost nil. If, on the 

 other hand, we take the remaining seventeen experiments, 

 lying between 179° and 195°, the average fractional deviation 

 is + -0022, the maximum +'004:1, the minimum —-0007. Here 

 the extreme differences from the average are twice as great as 

 in the preceding, while the sum of the squares of the differences 

 from the average is considerable, thus showing their greater 

 unreliability, apart from Regnault's own statement to that effect. 

 The writer has therefore considered it proper to change 

 these average deviations by making the total heat for 190° a 

 little smaller than that given by Regnault's formula. The 

 equation formed is 



H = 603-2 + 0*356 t — 0*00021 f 



to hold only above 100°. In the fifth column of the table 

 (page 20) there are given the fractional deviations of the cor- 

 rected total heats from this formula. It will be seen that up 

 to 26 the experiments are satisfied as closely as possible. From 

 26 to 57 the deviations gradually increase on the negative side,, 

 while from 57 to 73 they are still more marked on the positive 

 side. But 26 to 57 are greater in number than the latter, are 

 more consistent in themselves, and, as we have seen, are more 

 reliable. This justifies the greater weight given to them. The 

 average deviation in the last seventeen experiments is +'0038. 

 Even if the corrections due to the variation of the specific heat 

 of water are not allowed, the formula still represents the 

 experiments better than Regnault's. 



The formula is correct at 100°, and I think we can say that 

 at 200° it cannot be more than 0*4 per cent out, corresponding 

 to an error in the latent heat of 0*6 per cent. 



Consider now Regnault's experiments below 100°. In the 

 first six of the twenty-three from 88° to 63° the calorimeter 

 temperatures averaged about 7° and 19°, hence the quantities 

 of heat are expressed in c xs as a unit. Experiments 7 to 13 

 have for a calorimeter range 9° to 21°, or are in c 15 as a unit. 

 The remaining experiments have the same calorimeter range 

 as the first six. Taking account of these facts, the fractional 

 deviations of the experiments from Regnault' 

 given below : 



•004 

 •005 

 •004 

 •001 

 •000 

 •005 

 •001 

 •005 



1 



88-11 



o 



87-83 



3 



85-97 



4 



85-24 



5 



85-20 



6 



84-88 



7 



83-08 



8 



82-66 



•001 



•001 



•006 



•005 



•000 



•005 



•005 



•001 



9 



81-03 



10 



80-60 



11 



80-37 



12 



80-17 



13 



79-55 



14 



78-28 



15 



76-50 



16 



71-35 



ilt's 



formula 



are as 



17 



t 

 17-1 



-•008 



18 



70-49 



— •001 



19 



69-70 



— •001 



20 



68-01 



-•007 



21 



66-30 



— •002 



22 



64-33 



— •004 



23 



63-02 



•000- 



