232 PROCEEDINGS OF THE AMERICAN ACADEMY 



It is not to be expected that our results, or those of our contempo- 

 raries, are final. The most that has beeu accomplished by recent in- 

 vestigation is to show that the ratio of the atomic weight of oxygen to 

 that of hydrogen, deduced from the elementary substances, as we know 

 them in their purest condition, is decidedly less than that of 10 to 1. 

 The evidence as to the exact value of this ratio is still conflicting, and 

 although after our experience we cannot see how greater accuracy 

 could be gained by any variation of our process, we are far from claim- 

 ing that our results have not been vitiated by unknown constant 

 errors. Fortunately, an exact knowledge of the ratio is at present of 

 no practical importance in chemical analysis. The only question on 

 which the actual small indefiniteness has a bearing is the unit of atomic 

 weights, a question that has been much discussed of late ; but here, as 

 it seems to us. one consideration should be conclusive. 



Every one who has worked on the oxygen and hydrogen ratio 

 knows that the resources of experimental science have been taxed to 

 the utmost in these investigations. Greater accuracy is not to be i x- 

 pected with our present appliances, and yet there is an outstanding 

 uncertainty amounting to more than one half of one per cent of the 

 value. This corresponds to a variation of more than one unit in many 

 of the higher atomic weights ; and unless we are willing that these 

 chemical constants should fluctuate to this extent with every new de- 

 termination of the fundamental ratio, we must seek a more invariable 

 basis. The most natural and stable unit is the atomic weio-ht of 

 oxygen, not only on account of the wonderful combining power of this 

 element, but also because the ratio of the combining weights of most 

 of the elements to that of oxygen are known with great precision ; 

 and many cogent reasons could be urged for returning to the system 

 of Berzelius, which referred all the other weights to that of oxvsren, 

 assumed to be 100. But this system would not exhibit to advantage 

 the numerical ratios ou which modern chemical classification is based. 

 Hence for a provisional system we most warmly approve of that which 

 assumes = 16 as its basis, and of the best known atomic weights 

 leaves only the value of H to vary with our changing knowledge. 

 This system has all the stability it is possible at present to secure, and 

 exhibits to advantage the relations which are important in classifica- 

 tion. Moreover, it is no small recommendation to this system that a 

 large number of the weights are whole numbers, within the limits of 

 error of ordinary analytical work, and for this reason can be easily 

 remembered. 



In conclusion, I would express my obligations to my nephew, Dr. 



