232 SMITHSONIAX MISCELLANEOUS COLLECTIONS VOL. 54 



The antecedent values for reducing these ratios are : 



Ag = 107.880, ± .00029 S = 32.0G67, ± .00075 



CI = 35.4584, ± .0002 C = 12.0038, ± .0002 



Ba = 137.363, ± .0025 H = 1.00779, ± .00001 



Hence, for magnesium, we have 



From ratio 5 Mg = 24.0162, ± .0020 



" 3 24.0803, -± .0680 



" 1 24.2561, ± .0333 



" 6 24.2865, ± .0020 



" 7 24.3154, ± .0007 



•' 8 24.3344, ± .0033 



"2 24.394, -.0021 



" 4 24.7063, ± .0037 



General mean, Mg = 24.3039, ± .0006 



Tliis final value is possibly a little too low, as compared with the 

 individual values which are presumably the best. The figures are, how- 

 ever, iJeculiarly instructive. Eatios 2, 7 and 8, representing essentially 

 the work of Marignac and Eichards and Parker, were originally reduced 

 with the Stas values for sulphur, silver and chlorine. These values are 

 Ag= 107.93, Cl = 35.457 and S = 32.074: With these figures, and using 

 only jMarignac's data for ratio 2, the following values for magnesium are 

 obtained : 



From MgS04:MsO, Mg = 24.383 



2Ag:MgCL, Mg = 24.382 



2AgCl:MgCl2, Mg = 24.371 



From the general mean represented by ratio 2, Mg = 24.398, a slightly 

 higher value. 



The concordance here is much greater than in the reduction with 

 modern values, and may be interpreted in either of two ways. Either 

 the Stas values are more exact than tlie new values for Ag, CI and S, or 

 the earlier concordance is deceptive. In short, an agreement between 

 determinations of atomic weight made by diverse methods, is dependent 

 in great part upon the antecedent values used in tlie computations. Con- 

 cordance and discordance may be equally deceptive. Illustrations of this 

 statement are not uncommon. 



