194 



PROCEEDINGS OF THE AMERICAN ACADEMY. 



The average ratio of silver to silver chloride is within less than four 

 thousandths of a percent of that obtained by Richards and Wells, 31 

 0.752634:1.000000. Part of the discrepancy is due to the fact that 

 in a few analyses, notably Nos. 12, 13 and 14, the platinum sponge 

 crucible allowed traces of solid silver chloride to pass through. The 

 difficulty was not detected in time to determine the extent of the error, 

 but in one of these analyses silver chloride was found to have settled 

 out of the mother liquor on long standing. This is undoubtedly the 

 reason why the atomic weight of praseodymium determined in these 

 analyses is slightly higher than in most other experiments. Need- 

 less to say, the crucible was repaired and in Analyses 15 and 16, made 

 subsequently with the same fraction of material as Analyses 12, 13 

 and 14, the results are in accord with the general mean of all the de- 

 terminations. These defective analyses as well as a very few other 

 discrepant determinations, have not been omitted, in computing 

 averages, since, because of the large number of experiments, their 

 influence is not large. 



Results Averaged by Fractions. 



When the results are averaged by fractions, further evidence is 



31 Pub. Car. Inst., No. 28 (1905); Jour. Amer. Chem. Soc, 28, 456. 



