COPPER. 93 



other indirect results, similarly corrected, gave 79.900 per cent. Cu in 

 CuO, or Cu = 63.603. If we assign all five experiments equal weight, 

 and judge their value by the two detailed above, the mean percentage 

 becomes 79.900, dt .0038. This figure need not be combined with the 

 data given by previous observers, so far as practical purposes are con- 

 cerned ; but as this work is, in part at least, a study of the compensation 

 of errors, it may not be wasted time to effect the combination, as follows : 



Berzelius 79.823, .0020 



Erdmann and Marchand 79.8645, .0038 



Millon and Commaille 79-7787, .0043 



Hampe 79-8347, db .0013 



Richards 79-9, .0038 



General mean 79-8355, .0010 



This result is practically identical with that of Hampe, whose work 

 receives excessive weight, as does also that of Berzelius. The oxide of 

 copper is evidently of doubtful value in the measurement of this atomic 

 weight. 



The composition of the sulphate has been studied, not only by Hampe, 

 but also by Baubigny* and by Richards.f Baubigny merely ignited 

 the anhydrous salt, weighing both it and the residual oxide, as follows : 



4.022 grm. CuSO 4 gave 2.0035 CuO. 49.813 per cent. 



2.596 1.293 " 49.807 " 



Mean, 49.810, .002 



The same ratio, in reverse that is, the synthesis of the sulphate from 

 the oxide was investigated by Richards (p. 275), who shows that the 

 results obtained are vitiated by the same errors which affect the copper 

 oxide experiments previously cited. The weights given are reduced to 

 vacuum standards. The percentage of oxide in the sulphate is stated in 

 the third column of figures. 



1.0084 grm. CuO gave 2.0235 S rm - CuSO 4 . 49-835 P er cent. 



2.7292 5-4770 " 49.83 " 



1.0144 2.35 49.848 " 



Mean, 49.838, .0036 



The two series combine thus : 



Baubigny 49.810, .0020 



Richards.. , 49.838, .0036 



General mean 49.816, dr .0017 



Here, plainly, the rigorous discussion gives Baubigny 's work weight 

 in excess of its merits. 



* Compt. Rend., 97, 906. 1883. 

 fProc. Amer. Acad., 26, 240. 1891. 



