Composition of Dilute Sulphuric Acid. 313 



vation and calculation are really due to errors. They occur 

 in two instances, at all events, in pairs having opposite signs. 

 One of these is near 79 per cent., the other near 65 per cent. 

 They are precisely what might be expected if from any cause 

 one density was more than usually erroneous. Is it not likely 

 that other minor errors are to be explained in a similar way? 



Mr. Pickering appears to me to be on the horns of a dilemma. 

 If we assume the larger values of the maximum error to be 

 correct, then the values of $ — « are for the most part (except 

 at the abnormal points) so well within the limits of the error 

 of experiment that the whole discussion resolves itself (even 

 from his point of view) into a controversy about matters less 

 than the error of experiment. If the lower limit is adopted, 

 then he must explain why, when at 64*50 per cent, the value 

 of 7 — a is some three and a half times greater than the esti- 

 mated maximum .error, the theory of discontinuity is required 

 to account for the much smaller discrepancies found elsewhere. 



Horizontal lines are drawn in Table IV. at the points where 

 Mr. Pickering believes that there are breaks in the continuity 

 of the curve. The differences ft— a show no sign of regularity 

 In these intervals, they are for the most part of the same order 

 of magnitude as the estimated error of experiment ; and any 

 argument which could be drawn from the fact that they 

 exceed the lower values which may be assigned to it is neu- 

 tralized by the fact that in some cases they unquestionably 

 exceed the largest limits Mr. Pickering has specifically men- 

 tioned. Finally, to obtain the last place it has been necessary 

 to strain the calculations, and to carry them in some instances 

 to a place further than Mr. Pickering himself has done. If 

 therefore we adopt the safer course of neglecting figures 

 <10~ 5 , the continuous equation expresses the facte at least 

 as well as the five discontinuous curves. 



It is of course logically open to Mr. Pickering to claim that 

 his case is not disproved until a process similar to that which 

 I have adopted has been successfully applied in a much larger 

 number of instances. So far as I am concerned I can only 

 reply that I think my result places him in the position of 

 having to prove a negative, and that I certainly cannot under- 

 take to devote myself to the multiplication of such evidence. 



I have before stated that I think that the curves in some 

 parts — if not discontinuous — have peculiar features which 

 suggest special physical causes. Thus between 84 and 

 94 per cent, they are obviously nearly straight lines. My 

 doubts have always had reference to the minor changes of 

 curvature and to the use of the ruler in detecting them. They 

 are confirmed by the above investigation. 



Phil. Mag. S. 5. Vol. 32. No. 196. Sept. 1891. Y 



