254 PROCEEDINGS OF THE AMERICAN ACADEMY. 



larse. 1.00000 crara of salt yielded 1.74155 grams of silver bromide. 

 Of this silver bromide, 0.00252 gram correspond to 0.00138 gram of 

 'sodium bromide contained in the salt. If both these corrections are 

 subtracted, the atomic weight of iron calculated from the quantities 

 remaining is 55.871. 



This value is slightly lower than the result of the analysis of ferric 

 oxide, 55.883. But this was to be expected, for it was pointed out at 

 the time the earlier results were published, that the value obtained was 

 rather too hii2,h than too low, since the errors which would affect deter- 

 minations made by reduction of the oxide, i. e., those due to the presence 

 of non-reducible impurities in the material and to incomplete reduction, 

 would have raised the atomic weight above its true value. As a matter 

 of fact, it was found impossible to prove that either of these errors 

 existed. Occluded gases, which would have produced the opposite effect, 

 were conclusively proved to be absent. It is interesting to note that 

 Analysis 6 of the earlier series, in which most material was used and 

 which consequently should be the most accurate, gave a result, 55,870, 

 practically identical with that of this research. On the other hand, the 

 presence in the ferrous bromide of a trace of ferric salt would have 

 lowered the atomic weight. The average of the two series, 55,877, 

 must represent very closely the true atomic weight in question. 



The value 55.88 for the atomic weight of iron thus receives fresh 

 support. It is hardly conceivable that sufficient undiscovered impurity 

 could have been present in the material used for either series of analyses 

 to have affected the second figure of decimals. Nevertheless, further 

 work on the subject is planned with the hope of increasing the purity 

 of the material used for analysis. 



Attention has been recently called to the possibility that the atomic 

 weights of the magnetic metals might be affected by the attraction of the 

 earth's magnetism. Obviously the error introduced by this weak 

 attraction could not be large, but nevertheless the ascertaining of the 

 order of inaccuracy is a point of some interest to the chemist. 



The intensities of the earth's magnetic field in various places are well 

 known, but the calculation of the attractive action of this field upon a 

 given magnet depends, of course, upon the intensity and length of the 

 magnet as well as upon the field. If the magnet is so short that its two 

 poles are essentially the same distance from the earth's magnetic north 

 pole, it is clear that the repelling effect of the earth upon one pole will 

 practically counterbalance the attracting effect upon the other pole ; 

 hence the effect of the earth's magnetism will be simply directive, and 



