166 SMITHSONIAN MISCELLANEOUS' COLLECTIONS VOL. 54 



If we combine Eichards' three series into a general mean separately,, 

 the value found for the ratio becomes 339.402, ±.0040. The other de- 

 terminations, having high probable errors, affect this mean but slightly, 

 and it makes little difference whether they are retained or rejected. 



We now have the following ratios from which to compute the atomic 

 weight of copper : 



(1). Percentage of Cu in CuO, 79.836, ± .0010 



(2). Percentage of Cu in CuSO^, 39.795, ±: .0036 



(3). Percentage of Cu in CUSO4, 5H„0, 25.451, ± .0011 



(4). Percentage of CuO in CuSO^, 49.816, ± .0017 



(5). Cu:Na,C03: : 100: 166.838, ± .0035 



(6). CuiNa-SO^: : 100: 223.525, ± .0098 



(7). BaS04:Cu::93.289:25.448 



(8). 2AgBr:Cu::100:16.924, ± .0007 



(9). Cu :2Ag:: 100: 339.423, ±: .0038 



Eatio 7 rests upon a single experiment, and must be arbitrarily 

 weighted. For this purpose, the value for copper derived from it may 

 be given double the probable error of the highest among the other de- 

 terminations. To reduce the ratios we have — 



Ag =107.880, ±.00029 C = 12.0038, ± .0002 



CI =: 35.4584, ± .0002 Na = 23.0108, ± .00024 



Br = 79.9197, ± .0003 Ba =137.363, ±.0025 



S = 32.0667, ± .00075 H = 1.0079, ± .00001 



Hence, 



From ratio 1 Cu = 63.3493, ± .0032 



" 4 63.4796, ± .0039 



" 2 63.4993, ± .0069 



" 3 63.5497, ± .0033 



" 5 63.5499, ± .0024 



" 8 63.5664, ± .0027 



" 9 63.5667, ± .00075 



" 6 63.5714, ± .0029 



" 7 63.6765, ± .0138 



General mean, Cu = 63.5550, ± .00063 



This value is possibly, but not certainly, a little too low. The re- 

 jection of the first value, derived from copper oxide, raises the general 

 mean to 63.564, which may be nearer the truth. 



