ATOMIC WEIGHTS 139 



This final mean is almost identical with the value derived from ratio 

 29, which gives the composition of silver nitrate. That ratio, moreover, 

 is presumably the best of all, and has the smallest probable error. It 

 dominates the entire combination ; but its rejection would only raise the 

 atomic weight of silver to 107.883. If we should reject all the values 

 for silver dependent upon analyses of chlorates, bromates and iodates, 

 which are generally high, the final mean becomes 107.877. It is clear, 

 therefore, that the true value cannot be very far from the general mean 

 of all, namely, 



Ag = 108.880. ±.00029 



As for the widely aberrant values, especially the first two and the last 

 four, their probable errors are so large that it is a matter of no moment 

 whether they are retained or rejected. Their influence is negligible. 



With the aid of the value thus found for silver, we can now compute 

 twenty values for the atomic weight of chlorine, as follows : 



From ratios 12, 18, and Ag CI = 35.4150, ± .0080 



" 4, 53, and Ag 35.4186, ± .0058 



" 1, 3, and 36 35.4269, ± .0029 



" 3, 31, 38, and Ag 35.4279, ± .0012 



" 3, 38, and 39 35.4401, ± .0012 



" 3, 20, 38, and Ag 35.4483, ± .00096 



" 14 and Ag 35.4502, ± .00035 



" 1, 3, 32, and Ag 35.4556, ±: .0010 



" 3, 38, and 40 35.4569, ± .0022 



" 4, 54 and Ag 35.4575, ± .00093 



" 3, 30, and Ag 35.4610, ± .0043 



" 5 35.4643, ± .00039 



" 1, 3, 33, and Ag 35.4661, ± .00051 



" 3 and 37 35.4717, ± .0030 



" 3, 10, 21, 38, and Ag 35.4745, ± .0032 



" 1, 3, 10, and 34 35.4772, ± .0023 



" 3, 7, and 38 35.4813, ± .0013 



" 10 and Ag 35.5100, ± .0061 



" 3, 7, 21, 38, and Ag 35.5235, ±: .0018 



" 11, 17, and Ag 35.5240, ± .0102 



General mean, CI = 35.4584, ± .0002 



Here, again, the extreme values are evidently of no real significance, 

 and have practically no effect upon the final result. The rounded-off 

 figure, 35.458, is in good agreement with the determinations made by 

 Noyes and Weber, and also with the ratio between silver and chlorine 

 as measured by Eichards and Wells. 



For bromine, using the new value for chlorine in place of that given 

 by ratio 5, eleven values are deducible: 



