346 PROCEEDINGS OP THE AMERICAN ACADEMY. 



complete the quota. This is quite a different story. It is evident that 

 nearly all the errors mentioned tend to make Winkler's result too high. 



Some singular oversights appear also in the calculation of the results. 

 For example, in one place Winkler compares 0.1 G62 gram of nickel with 

 0.6079206 gram of silver. The small amount of nickel was deposited 

 in a large platinum dish, and its weight could certainly not have been 

 determined more accurately than within 0.1 milligram, hence at least 

 three decimal places of the recorded weight of silver were superfluous, 

 even if the volumetric solution could have been prepared with an error 

 of only one part in six millions. It is perhaps well to mention also that 

 his final results, varying in the case of cobalt from 59.5996 to 59.7480 

 (if O = 15.96) are given as far as four decimal places. 



While a review of this work is necessary in order to explain why the 

 results should be too high, perhaps one should not be severe in one's 

 criticism of it, for Professor Winkler himself rejects it, as well as some 

 later work on cobalt,* in his most recent contribution upon the subject. f 

 In this new paper he pins his faith to another series of determinations 

 made in 1894, with a very ingenious method adopted after sundry fruit- 

 less attempts in other directions. It behooves us then to consider this 

 later work with great care. J 



Evidently many of the errors which render the older investigation 

 untrustworthy were eliminated from that of the subsequent year. The 

 nickel was separated from the platinum dish and afterwards ignited in an 

 atmosphere of hydrogen, and the solubility of argentic chloride does not 

 enter into the question. On the other hand, the unfortunate use of vol- 

 umetric operations and the misuse of figures remained, while to these were 

 added other dangers not present in the older work. The ingenious pro- 

 cedure was as follows: pure nickel was acted upon by pure iodine, and 

 the excess of iodine was determined by sodic tliiosulphate. Many text- 

 books upon volumetric analysis name the process of iodometry as one 

 of the most accurate of titrimetric methods simply because the end 

 point is an extremely sharp one. In reality, the lack of permanence 

 of the necessary solutions render it distinctly unsuitable for very ac- 

 curate work even under the best conditions. When the iodine must 



* Zeitschr. Anorg. Chem., IV. 462. 



t Zeitschr. Anorg. Chem., XVII. 236. 



i 1894 (1895), Winkler, Zeitschr. Anorg. Chem., VIII. 1,291; Ni = 58.85. It 

 must be borne in mind in referring to Winliler's papers that he uses the old stan- 

 dard = 15.96. His values have all been translated into the more convenient nota- 

 tion (0 = 16.000) in this paper. 



