EJJSHA MITCHELL SCIENTIFIC SOCIETY. 79 



ment? If oxygon wtre made 100, as in tlio Borzelius tabic, or 

 10 or 1, as have been ])ropose(], then the present nninl)ers, as 

 referred to hydrogen, wouhl be entirely (;hange<l and lost sight 

 of. The plan is to change thetn as little as possible, giving oxy- 

 gen the number sixteen, which was once regarded as the ratio 

 between it and hydrogen, and, ac(;ording to the views of some, 

 may still be it. Then the number for hydrogen will vary very 

 little from unity, and the whole table is nearly based upon it as 

 the unit. This number will change from time to time with new 

 determinations, but such changes will involve no others. Oxy- 

 gen, the standard, will then be fixed, and our task lies in the 

 accurate determination of the others by it. 



Meyer and Seubert cling to the idea that if r)xygen be adopted 

 it must be taken equal to unity, maintaining that = 16 is 

 neither flesh nor fowl, and by no means a unit. It is true that 

 the standard or basis of a series of physical constants has in the 

 past usually been taken equal to unity, but I cannot conceive the 

 power of this habit to be sufficiently strong to force us into 

 inconveniences or inaccuracies. That it is not regarded as a bind- 

 ing rule has been shown by the choice of some recent standards, 

 especially in the new science of electricity. 



The atomic weights are but relative numbers. To be in any 

 respect constants, they must be relative to but one single element. 

 With but few exceptions the ratio to oxygen can be determined. 

 In revision of atomic weights, then, this should receive the 

 chief attention and the oxygen ratio should be most carefully 

 and directly determined. 



Where the intermediation of another element is made use of 

 (even though this be one which "may be counted among those 

 of which the atomic weights are already known with the nearest 

 approach to exactness," as Dr. Mallet recommends) it must be 

 borne in mind that the result is subject to a certain error, which 

 is generally m^dtiplied several times and hence cannot give con- 

 cordant results with the direct oxygen ratio, and less stress must 

 be placed upon it. If the well determined ratio H : O is subject 

 to an error of 0.3 per cent., how much greater is the error in the 

 case of ratios less well known? 



