486 



Royal Society. 



Appendix B. 

 The following Table shows the degree of concordance obtained 

 in the separate experiments used to determine the unit. The deter- 

 minations were made by observing the deflections of a certain magnet 

 when a coil revolved at a given speed, first in one direction, and then 

 in the opposite direction. The first column shows the speed in each 

 experiment ; the second shows the value of the B. A. unit in terms 



of 10 7 -, as calculated from the single experiments. A differ- 

 second & r 



ence constantly in one direction may be observed in the values ob- 

 tained when the coil revolved different ways. This difference de- 

 pended on a slight bias of the suspending thread in one direction. 

 The third column shows the value of the B. A. unit calculated 

 from the pair of experiments. The fourth shows the error of the 

 pair from the mean value finally adopted. In the final mean 

 adopted, the 1864 determination was allowed five times the weight 

 allowed to that of 1863. 



1. 



Time of 100 revo- 

 lutions of coil, in 

 seconds. 



Value of B. A. unit in 



n , A7 metres 



terms of 10 7 -> 



second 

 as calculated from 

 each experiment. 



3. 



Value from mean 



of each pair of 



experiments. 



Percentage error of 

 pair of observations 

 from mean value. 



17-54 



17-58 



77-62 



76-17 



53-97 



54-53 



41-76 



41-79 



54-07 



53-78 



17-697 



17-783 



17*81 



17'78 



17-01 



16-89 



21-35 



21-38 



21-362 



21-643 



11-247 



16-737 



1-0121 



0-9836 



1-0468 



0-9613 



0-9985 



0-9998 



0-9915 



0-9936 



0-9961 



0-9886 



0-9878 



1-0136 



0-9952] 



1-0174 



1-0191 



0-9895 



1-0034 



1-0011 



0-9968 



1-0096 



1-0424 



0-9707 



0-9978 



0040 

 9992 

 9925 

 9924 

 0007 

 0063 

 0043 

 0022 

 0040 

 9981 



-0-22 

 + 0-40 

 -0-08 

 -0-75 

 -0-76 

 + 0-07 

 + 0-63 

 + 0*43 

 + 0-22 

 + 0-40 

 -0-19 



Probable error of R (1864) =0*1 per cent. 



Probable error of R (1863) =0*24 



Difference in two values 1864 and 1863=0*16 „ 

 Probable error of two experiments = 0*08 „ 



