6 PROCEEDINGS OF THE AMERICAN ACADEMY 



which has come to our knowledge was made incidentally by B. linger* 

 while studying the chemical relations of Schlippe's salt, NajS^Sb- 

 9 H.,0. An analysis of this salt gave Sb= 1 19.7 fi, when S = 32 and 

 Na = "23 ; hut this value rests on a single experimental determination, 

 by a method from which great accuracy could not be expected. 



As is well known, the value 122 obtained by Dumas is the one 

 which is almost universall}^ accepted as the atomic weight of antimony, 

 and this not only in consequence of the deservedly great authority of 

 this distinguished experimenter, but also because his result so nearly 

 agrees with the previous determination of Dexter. Nevertheless, a 

 careful examination of the work of Schneider, referred to above, will 

 convince every chemist that it is impossible to refer the difference 

 between 120.3, the number he obtained, and 122, to any experimental 

 errors. This difference is nearly one and a half per cent of the whole 

 value ; although it is evident that the probable experimental error 

 does not, in either case, amount to one tenth of one per cent. More- 

 over, it will be found that this investigation of Schneider is a model of 

 its kind. All the details of the experimental work are given, and it 

 is evident that every precaution was taken which the circumstances re- 

 quired. P^urthermore, the method of Schneider has a very great advan- 

 tage over all the processes by which the atomic weight of antimony 

 has been determined, in that he was able to obtain satisfactory evi- 

 dence, not only of the purity, but also of the definite atomic composi- 

 tion of the material he analyzed. In the determination of an atomic 

 weight, it is not only essential that the substance selected for analysis 

 should be pure, in the sense of containing no adventitious elements : it 

 is even more important that neither of the constituents whose atomic 

 ratio is to be determined should be present in excess of the normal pro- 

 portion. But, while experimenters have been most careful to establish 

 the purity of their material in the first sense, they have seldom given a 

 thought to the possibility of such an impurity as the preponderance of 

 one or the other of the proper constituents of the compound practically 

 constitutes. Theoretical considerations, based on the law of definite 

 proportions, might lead a cliemist to believe that impurity in the last 

 sense was impossible in a true chemical compound. But how are we 

 to know that a given material is of the true type? It is certain that 

 the ordinary physical tests of purity, such as the melting point and the 

 boiling point, are insulficient; and the same is equally true of the per- 

 fection of crystalline form, that ciiaracter on which of all others the 



* Arcliiv. der Pliarmacie, 1871, Band 137, S. 191. 



