38 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 54 



The density ratios, H : 0, now combine as follows : 



Dumas and Boussingault, corrected 15.9015, ± .031 



Regnault, corrected 15.9105, ± .0044 



Rayleigh, 1888 15.884, ± .0048 



Rayleigh, 1892 15.882, ± .0023 



Cooke 15.890, ± .0067 



Leduc 15.906, ± .0154 



Morley, including all the data 15.8955, ± .0005 



Thomsen 15.8878, ± .0022 



General mean 15.8948, ± .00048 



If we reject all of Morley's data for the density of hydrogen except 

 his third, fourth and fifth series, the mean becomes 



15.8991, ± .00048 



In either case Morley's data vastly outweigh all others. 



If oxygen and hydrogen were perfect gases, uniting by volume exactly 

 in the ratio of one to two, then their relative densities would also indi- 

 cate their relative molecular weights. But, in fact, the two gases vary 

 from Boyle's law in opposite directions, and the true composition of 

 water by volume diverges from the theoretical ratio to a measurable 

 extent. Hence, in order to deduce the atomic weight of hydrogen from 

 its density, or that of oxygen, if the hydrogen scale is preferred, a 

 small correction must be applied which depends upon the amount of 

 the divergence. Until modern times our knowledge of the volumetric 

 composition of water rested entirely upon the determinations made by 

 Humboldt and Gay Lussac ^ early in the last century, which gave a ratio 

 between H and of a little less than 2 : 1, but their data need no farther 

 consideration here. 



In 1887 Scott '' published his first series of experiments, 21 in number, 

 finding as the most probable result a value for the ratio of 1.994 : 1. In 

 March, 1888,' he gave four more determinations, ranging from 1.9962 to 

 1.998 : 1 ; and later in the same year * another four, with values from 

 1.995 to 2.001. In 1893,'* however, by the use of improved apparatus, 

 he was able to show that his previous work was vitiated by errors, and to 

 give a series of measurements of far greater value. Of these, twelve were 

 especially good, being made with hydrogen from palladium hydride, 

 and with oxygen from silver oxide. In mean the value found is 2.00245, 

 ±.00007, with a range from 2.0017 to 2.0030. 



^ Journ. de Phys., 60, 129. 



*Proc. Roy. Soc, 42, 396. 



« Nature, 37, 439. 



* British Assoc. Report, 1888, 631. 



»Proc. Roy. Soc, 53, 130. In full in Philosophical Transactions, 184, 543. 1893. 



