6 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 54 



tioned. Its fairness depends in part upon the fairness of the experi- 

 menter. One chemist, making a series of measurements, gives all of his 

 determinations. Another chemist selects those which are most con- 

 cordant, and suppresses others which seem to him less trustworthy. The 

 latter series, therefore, is likely to receive higher weight than belongs to 

 it; while the former series will be underweighted. The rejection of 

 data, even by the man who is most familiar with them, is always a 

 dangerous proceeding, and one which should be discouraged. 



The other and more usual method of adjusting the atomic weights, 

 that of selecting determinations in accordance with their apparent chem- 

 ical merit, has recently been followed by Brauner. In his excellent and 

 critical discussion of the subject, now appearing in Abegg's Handbuch 

 der anorganischen Chemie, he gives all the determinations for each ele- 

 ment, and then assigns preference to those which most appeal to his 

 judgment. In most instances his findings agree with mine, and there- 

 fore our conclusions reinforce each other. Sometimes we differ, and in 

 such cases it would seem that new determinations are desirable. When 

 values derived from different sources, and computed by different methods 

 are concordant, they may be regarded as probably well established; bui 

 even then certainty is not attained. The history of atomic weight de- 

 terminations bears abundant witness to this assertion. 



For example : Until within very recent years the work of Stas, em- 

 phasized by that of Marignac, was regarded as almost final. Now, how- 

 ever, some of the ratios measured by these chemists are found to be out 

 of harmony with the best modern investigations, and there is a tendency 

 towards rejecting the older work altogether. But the researches of Stas 

 give a homogeneous and concordant group of atomic weights, which 

 cannot be entirely thrown aside without much more evidence against them 

 than as yet exists. It is probable that the silver used by Stas contained 

 occluded oxygen, as was pointed out by Dumas ^ ; and this would account 

 for some, but not all of the variations from recent revisions of the ratios. 

 It is also probable, as Richards has shown, that Stas underrated the 

 solubility of silver chloride. How large these errors may be in Stas' 

 work, assuming them to exist, is uncertain; and to assign zero weight to 

 his determinations would be too extreme a procedure. His data and 

 Marignac's are therefore retained in the present recalculation, with the 

 proper mathematical weight; and the final results seem to be satisfactory. 

 Indeed, the Stas values for silver, chlorine and bromine, applied to the 

 determinations of other atomic weights, sometimes give more concordant 

 results than the modern figures. This is especially true in the cases of 

 caesium, barium and magnesium, although the discrepancies are not large. 



^ Ann. Chim. Phys. (5), 14, 289. See also the Appendix to the first edition of this Recalculation, 

 in which the influence of a correction for occluded oxygen is considered. 



