90 PROCEEDINGS OF THE AMERICAN ACADEMY. 



This average is after all but little different from the total mean of 

 all the determinations, 137.434; and either average may be taken 

 without serious error for the atomic weight of barium. Indeed, in the 

 present state of chemistry, the third decimal place is of little or no 

 importance in a large atomic weight, and the round number 137.44 

 is sufficient for the most exacting work. 



An excellent proof of the purity of the materials and the accuracy 

 of the work is to be found in the comparison of the weights of silver 

 and argentic chloride. Omitting in each case the less accurate first 

 series, we may repeat the following figures : 



[II.] 2 AgCl : BaCl 2 = 100 : x = 72.6563 

 [III.] 2 AgCl : BaCl 2 = 100 : x = 72.6555 



Average 96.5247 



Hence Ag : AgCl = 72.6559 : 96.5247 



= 107.930 : 143.387 



Therefore the atomic weight of chlorine is found to be 35.457, a 

 quantity exactly identical with that found by Stas in a different way. 

 A similar comparison of the weights of the silver and argentic bromide 

 has been given already in the fifth column in the table on page 29 

 of Volume XXVIII. of these Proceedings. 



It remains only to compare these figures with the older ones 

 already described. Evidently all results for barium based upon the 

 conversion of baric chloride into baric sulphate, as well as all results 

 including water of crystallization, are worthy of no confidence. The 

 incomplete knowledge regarding the method of Gay Lussac mani- 

 fested by Marignac and Dumas in 1858 is quite enough to explain 

 their unsatisfactory and discrepant results. We should expect to find 



