I9I0.] OF OXYGEN AND SILVER. 363 



above 40 thousandths (0.04) are mainly due to the fact that we 

 have taken all reactions used without excluding those known to be 

 imperfect and which are generally excluded by others. 



It will also be noticed that the mean departure for the greatest 

 number of cases is only one or a few thousandths ; it is without sig- 

 nificance in the question here considered. 



This question can now be fully answered. Bearing in mind that 

 increased atomic weight necessarily brings a slight increase in the 

 departures for experimental reasons, we must admit that the depar- 

 tures for oxygen and for silver are essentially alike, so that the values 

 0^16 and Ag=io8 stand and fall together. If oxygen is 16, 

 then silver is 108, with nearly the same degree of precision. Again, 

 if the value 108 is denied for silver, then the value 16 for oxygen 

 is equally untenable. 



Finally, it appears to me that it has been fully demonstrated that 

 the assumption of immunity from error for oxygen is as false in 

 fact as it is absurd in philosophy ; possibly that accounts for the 

 tenacity with which the school has clung to the same. 



St. Louis, Mo. 



