RICHARDS AND ARCHIBALD. — ATOMIC WEIGHT OF CAESIUM. 467 



IX. The Atomic Weights of Caesium, Potassium, and Nitrogen. 



In the preceding description are recorded analytical results involving 

 the atomic weights of caesium, potassium, nitrogen, chlorine, bromine, 

 silver, and oxygen. Of these, by common consent, the value for oxygen 

 is fixed ; the others may be calculated in many ways from the data in 

 connection with the data obtained by other experimenters. 



No attempt was made to explore any such complete experimental field 

 as that suggested by Clarke * in his recent paper on the calculation of 

 atomic weights. This omission was due not so much to lack of time as 

 to the feeling that some of the results would have been fruitless. Ex- 

 perimental skill in attaining uniformity of conditions can almost always 

 reduce the so called "probable error" of manipulation to a vanishingly 

 small quantity ; but constant errors of a chemical nature are much 

 harder to avoid and much more serious when perpetrated. Hence the 

 " probable error" is little or no clue to the trustworthiness of the results, 

 and can rarely be used as a reliable measure of the relative preponder- 

 ance to be ascribed to the respective members of a series of conflicting 

 figures. f Instead of attempting to observe the weight of every element 

 which combines with a given weight of every other without discrimina- 

 tion, it seems much wiser to select those operations in which the constant 

 errors, by long and thorough study, have been discovered and rendered 

 susceptible of elimination. A single series of experiments, and thor- 

 oughly investigated and properly executed, is of more value than ten 

 series containing unknown and incorrigible errors. 



The preceding figures involve seven different ratios for determining 

 the atomic weight of caesium, assuming in each case one or more of the 

 other atomic weights to be known. Five of these are obvious in the 

 results given ; the other two refer the salts of caesium to those of potas- 

 sium. These latter ratios are of interest because they eliminate, at least 

 in part, any possible constant errors in the processes immediately em- 

 ployed, by virtue of the parallelism between the experiments with the 

 two metals. The atomic weight of caesium should therefore be influenced 

 only by chose constant errors which already affect the assumed atomic 

 weight of potassium. 100.000 parts of silver were found to be equivalent 

 to 155.964 and 69.115 parts of caesic and potassic chlorides respectively, 

 hence these two weights must be equivalent to one another, or 100.000 



* Am. Chem. J., 27, 321 (1902). 



t Richards, A Table of Atomic Weights, Am. Cliem. J., 20, 543 (1898). 



