TABLE XXXVI. 



TABLE XXXVII. 



From Tables XXXVI and XXXVII, \ve see that the probable error of a single 

 difference 



M p 5 = 3-9 X10~ 7 sec. 



M, 7 =4-5 

 21 M =2-5 



Mean value 3-6xlO~ 7 sec. 



A comparison of these figures with the corresponding one of series B 3 shows only 

 a very slight gain in the accuracy of observation. At the same time, a comparison 

 with the observations of Potsdam I is not at all unfavourable, and such difference 

 as is indicated, may readily be referred to the still unsatisfactory action of the spring 

 in the coincidence apparatus or the contact in the clock. In spite of this approximate 

 agreement, the value of e m is 5 3 x 10"' sec. compared with 1 9 X 10~ 7 sec. for Potsdam I, 

 and this is clear proof that the errors here are instrumental rather than observational. 



52 



