468 PROCEEDINGS OP THE AMERICAN ACADEMY. 



of potassic to 225.659 of caesic chloride. Again, 100.000 parts of 

 argentic chloride were found to result from the precipitation of 117.398 

 and 52.022 parts of the same chlorides, a result which gives the almost 

 equal ratio 100.000: 225.670. The average of these two, 225.665, is 

 given below. 



In the same way, 100.000 parts of nitric anhydride were found to be 

 equivalent to 87.232 and 260.693 parts of potassic and caesic oxides 

 respectively. 



A similar calculation might be performed with the bromide, using 

 Stas's data obtained from potassic bromide ; but this result would not 

 have the same significance because the experimental work would uot 

 then have been done all by the same hands under the same conditions. 



Below is given a table containing the results for the constant in 

 question computed from the several ratios. After each standard sub- 

 stance is given its assumed molecular weight, and after each value of the 

 atomic weight of caesium is given the " mean error " or average deviation 

 from the mean. This is chosen rather than the " probable error" because 

 the latter is in most cases too small to possess important significance. 

 In the cases of the last two ratios the average of the mean errors of the 

 potassium and caesium series is recorded. 



From AgCl (143.385) : CsCl, Cs = 132.876 ± 0.003 



" Ag (107.93) : CsCl, Cs = 132.877 ± 0.003 



« AgBr (187.885) : CsBr, Cs = 132.880 ± 0.002 



" Ag (107.93) : CsBr, Cs = 132.881 ± 0.004 



" N.Os (108.08) : CsA Cs= 132.879 ± 0.005 



" KCl (74.595)* : CsCl, Cs = 132.879 ± 0.002 



" K2O (94.280) : Cs^O Cs = 1 32.879 ± 0.003 



Average 132.879 f 



It is of interest to note that not one of these seven independent 

 figures deviates from the average of all by an amount greater than its 

 mean experimental error. The inference to be drawn from this is that 

 constant errors have been essentially eliminated from the various chemical 

 processes involved. 



For the atomic weight of potassium we have three direct ratios, which 

 give the following results : — 



* This value of potassic chloride is known directly with reference to oxygen 

 from the analysis of potassic chlorate, which (according to Clarke) contains 39.154 

 per cent of oxygen. 



t If O = 15.879, Cs = 131.874. 



