16 Mr. C. Chree on the 



remarkably small and fully accounted for by errors of 

 reading. 



It is quite true that the greatest deviation of g n from its 

 mean value is only '6 ; but as this mean value is only 

 2 '7, a greater accuracy would seem desirable. Also, in 



log — ^t*L one deviation is as much as '015, which 



exceeds Lorberg's limit. This quantity is further so much 

 of the nature of an average, that large deviations could 

 hardly be expected. I have constructed a table of values of 



log — v ^~ ; and though the mean value deduced for M//, 2 



is nearly the same as Lorberg's, the numbers fluctuate from 

 •1875 to *1522, and so represent a deviation much above what 

 Lorberg allows to be possible. I have also calculated s after 

 Lorberg's method from the formula 



s(l — e~~^) = x n+m — x n e e" i 



putting m=2, 3, 5, 8, and 9, and obtained values of 5 steadily 

 increasing from 7*82 to 8*36. A study of Weber's table 

 also shows that in several cases, allowing Lorberg's full 

 error, we should get 



X n «#.ft-fi < &n+l $71+2) 



which would absolutely make /£ 2 a negative quantity. Weber 

 himself says he had observed the galvanometer-zero to alter 

 as much as 1*2 scale-divisions per minute, and complains of 

 fluctuations occurring, especially in his experiments on water. 

 There is finally somewhat of a dilemma. Lorberg's method 

 would indicate that /x 2 remained constant during the whole 

 range of each experiment, and so was independent of the 

 temperature ; but the values he obtains from the two experi- 

 ments at different temperatures are different, and so would 

 require /uu 2 to increase with the temperature. 



The value Lorberg finds for s would give r about J°, and 

 this, he thinks, might escape detection. I agree with him in 

 thinking it unlikely that the cover and its enclosed atmo- 

 sphere remain at zero temperature throughout, but think it 

 equally unlikely that r should be a constant. Weber gives 

 only one specimen table at each temperature for his experi- 

 ments on water ; thus Lorberg availed himself of all the data 

 at his disposal, as he limited his criticism to water. This 

 seems, however, rather a slender foundation to build an 

 elaborate theory on, especially as the value for k deduced 

 after Weber's method from the typical experiment at the 



