BORON AND FLUORINE. 57 



contained in Ramsay and Aston's material) the atomic weight 

 of boron would have become 11.052. This figure is identical with 

 the value just derived above (Is), upon the same assumption, 

 from Ramsay and Aston's work. It is, furthermore, of the 

 same order of magnitude as that calculated in expression 

 (II), which gave B = 11.061 and, according to argument, also 

 involved borax containing 0.3 per cent, of water and a water- 

 free end-product (in this case silver chloride). True, these 

 values of 11.05, or thereabouts, are incorrect. Nevertheless 

 their concordance with each other appears to furnish additional, 

 though indirect, proof of the essential correctness of our original 

 assumptions. 



This completes the argument concerning Ramsay and Aston's 

 work. The reasoning seems conclusive and completely deprives 

 Ramsay and Aston's results of incongruities which would scarcely 

 yield as plausibly to any other theories or assumptions. True 

 enough, the quantities of moisture called for by our explanation 

 seem unusually large, not to say excessive. These quantities, 

 however, have been shown not to exceed reasonable limits in this 

 particular case. That the worker in atomic weights must ever 

 be on guard against traces of water in the substances which go 

 to make up the ratio sought has been pointed out repeatedly, and 

 with emphasis, by T. W. Richards. The present case seems to 

 be an unusually forceful illustration of this contention. The 

 percentages of moisture involved in this instance are not merely 

 traces, but quantities which should not be neglected in a careful 

 every-day analysis. We should probably search the entire field 

 of atomic weight determinations in vain for another case so 

 obstinate in this respect and so disastrous in its effect as that 

 presented by borax. 



These remarks, of course, are not meant to suggest that the 

 corrections for moisture should be applied to Ramsay and Aston's 

 results for practical ends; they are too hypothetical for a rigid 

 computative revision. However, the corrections are far from 

 arbitrary and, if applied only for the sake of argument, not only 

 dispose of a grave and puzzling inconsistency, but also lead to 

 values practically identical with the one found in the present 

 investigation. Such considerations seem to offer an acceptable 

 excuse for the digression. 



