The probable errors of the mean result are : 



p & = 21-5X10" 7 sec. 

 Pi -- 16-2 



Pa. -~-- 16>4 



p m = 16-6X1Q-' sec. 



From simple inspection of Table XVIII above, it is clear that the clock rate was 

 very uneven, though the mean of the first pair is not very greatly different from the 

 mean of the second pair of observations. 



Apart from this systematic error, it is clear from analysis of Table XIX, which 

 gives the differences of the mean from the individual pendulums, the residual 

 differences and their squares, that other important errors are present. 



TABLE XIX. 



The probable errors of a single difference are therefore, for : 



M p -5 

 M.-7 

 21 - M,, 



the mean of which is 



8 -6 XlO" 7 sec. 

 6-0 

 7-3 

 7-3X10" 7 sec. 



The high value may be due to unevenness of clock rate during the short period 

 occupied in swinging all three pendulums, but the unsatisfactory conditions in which 

 the observations were made are clearly responsible in some measure. 



As will be seen later, it is necessary to neglect the observations made with pendulum 

 No. 5 in arriving at the final result. 



26 



