APPENDIX. 267 



teen. Here we have forty elements falling within the limit 

 of variation assigned by Mallet, and twenty-six falling 

 without. The variations we may properly study in some 

 detail. 



Taking first the elements whose atomic weights vary from 

 even multiples of unity by less than a tenth of a unit, we 

 have to consider the following : aluminum, arsenic, barium, 

 bismuth, boron, bromine, cadmium, caesium, calcium, carbon, 

 cobalt, columbium, didymium, fluorine, gallium, hydrogen, 

 iridium, iron, lead, lithium, magnesium, manganese, nickel 

 nitrogen, osmium, oxygen, palladium, phosphorus, scan- 

 dium, selenium, silver, sodium, sulphur, thorium, tin, tita- 

 nium, tungsten, uranium, yttrium, and zinc. Of these, 

 aluminum, arsenic, barium, bismuth, cadmium, calcium, 

 carbon, cobalt, columbium, fluorine, hydrogen, iridium, iron, 

 lithium, magnesium, manganese, nickel, nitrogen, phos- 

 phorus, scandium, sodium, sulphur, tungsten, uranium, 

 yttrium, and zinc have plus variations, while boron, bro- 

 mine, ceesium, didymium, gallium, lead, osmium, palla- 

 dium, selenium, silver, thorium, tin, and titanium fall 

 slightly under the units to which they approximate. Oxy- 

 gen, as the standard of comparison, of course shows here 

 no variation, its possible error having been transferred to 

 hydrogen. 



Of the foregoing elements it will be seen that twenty-six 

 have plus variations from whole numbers, while thirteen 

 are minus. Among the latter, boron, gallium, osmium, pal- 

 ladium, thorium, and titanium have been but roughly deter- 

 mined. Bromine, by Dumas' correction, has its variation 

 diminished. In the cases of lead, caesium, selenium, and 

 tin, the cause of variation, supposing one to exist, remain's 

 to be determined. The value for osmium is undoubtedly 

 several units too high, so that its agreement with Front's 

 law may be considered purely accidental. As for didym- 

 ium, the figure assigned is the mean of all determinations ; 

 whereas Clove's data, calculated with SO3 = 80, make Di = 

 147.021, a variation which, like most of the others, is far 

 within the limits of ordinary experimental error. In the 



