260 PROCEEDINGS OF THE AMERICAN ACADEMY. 



inclusion) of gases by the oxide is a less serious cause of error in the case 

 of iron than in most other cases, if the temperature of ignition is reason- 

 ably high. This is the chief probable error which would tend to make 

 the results too low. The three most probable errors having the opposite 

 effect are : — first, the possible presence of magnetic oxide in the ferric 

 oxide ; second, the possibility of incomplete reduction during the analysis 

 of this substance; and third, the possible presence of alkaline, siliceous, 

 or other non-reducible material. Wackenroder seems to have been the 

 only experimenter who thought seriously of the first two of these errors; 

 and none of the early experimenters paid any attention to the last, for 

 until recently glass was supposed by most chemists to be wholly insoluble. 

 Wackenroder's work was perhaps the most intelligently carried out of all 

 the older determinations, although his individual analyses did not agree 

 among themselves quite so well as some of the others. His greatest 

 omission was the recognized lack of purity in his hydrogen, although he 

 could not observe an error due to this cause. It is interesting to note 

 that the average of his six results gives the value 55.82 for the atomic 

 weight in question, while a corrected average obtained by rejecting the 

 two most discordant values gives the value 55.92. The result of the 

 present work indicates a value midway between the two averages. 

 Wackenroder's valuable work does not enter into Clarke's average for the 

 atomic weight of iron, for Clarke was unable to find the original paper. 



In order to prepare pure iron for his later work, Berzelius fused his 

 metal with ferrous oxide, but gave no proof of the effectiveness of his 

 treatment. It is not impossible that traces of the oxide may have been 

 held by the iron, as copper dissolves cuprous oxide. Erdmann and Mar- 

 chand made the ferric oxide, which they reduced quantitatively, by the 

 ignition of ferrous oxalate , — a method which invites the presence of 

 magnetic oxide. Maumene's method was the process of filtration used 

 in common analysis, in which the possible errors in both directions are 

 plentiful. 



These detailed criticisms, taken in connection with the general lack of 

 accuracy which is to be observed in the quantitative work of half a cen- 

 tury ago, seem to show that there is nothing improbable in the present 

 result, 55.88. Of course the analysis of a single compound is never con- 

 clusive, so that this result is announced only as a preliminary one, which 

 we shall hope to support or disprove in the near future. 



Harvard University, Cambridge, Mass., 

 October, 1899. 



