Reducing the Results of Experiments. 93 



differences of like signs into separate groups would show quite 

 clearly to the most superficial observer that all agreement 

 between the calculated and observed values is absent. 



45-71655 

 48*02648 



50-06998 

 52-02681 

 54-00142 

 56-00914 

 57-85589 

 (59-97052 

 (61-91205 

 64-04535 

 66-02517 

 68-00851 



69-93139 

 72-02232 

 73-98643 



u calc. — u obs. 



-•000717* 



—000404* 



-•000139 

 + -000038 

 + -000218 

 + -000274 

 + -000299 

 + •000037) | 

 + -000159) J 

 - -000158 H 

 - -000257 | 

 -•000278 ' 



> 



I7*J 



- -00018 

 + •000008*1 

 + -000597 ) 



In the above table Mr. Lupton has, somewhat unfairly, 

 included two results (those in brackets), which, as I pointed 

 out (Chem. Trnas.) p. 70, table ii. p. 43, and plate 2 f), were 

 evidently erroneous ; it is immaterial, however, whether they 

 are inserted in the above table or not ; but I have omitted 

 them in the following sketch (p. 94), which gives a far more 

 striking illustration of the untenable nature of Mr. Lupton's 

 contention that " the equation expresses the results with far 

 greater accuracy than could be attained by even the most 

 careful drawing." As the densities themselves are unmanage- 

 able for diagrammatic purposes, I have had to take the 

 differentials deduced directly from them. 



It is certainly extraordinary that Mr. Lupton should have 

 imagined that he could disprove by such means not only the 

 one particular break in question, but also the 101 others 

 described in my paper, and most of which were far better 

 marked than this particular one. Indeed, it is by some strange 

 accident that Mr. Lupton was led to attack the only one 



* These values are not given in Mr. Lupton's table. 



t The tables (loc. cit.), not the plates, must be taken as giving the 

 correct values. The cross in plate 2 at 62 per cent, is misplaced, and 

 should stand for the three diiferential points in its neighbourhood show- 

 ing the accidentally large errors. In the last column of the table on 

 p. 142 for 52-71 read 53-01, and for 46-65 read 46'87. 



