E. W. Morley — Volumetric Composition of Water. 221 



Physik.* Chaptal and Berthollet made a report on the 

 memoir to the Academy, which is contained in the Annales de 

 Chimie et de Physique.f Humboldt and Gay-Lussac made 

 twelve experiments with an excess of hydrogen, and twelve 

 with an excess of oxygen. They determined the amount of 

 nitrogen in their oxygen by absorption with an alkaline sul- 

 phide, and with this oxygen determined the amount of nitro- 

 gen in the hydrogen. The mean error of the measurement of 

 the residue after explosion with an excess of hydrogen was 

 one part in five hundred, and in the experiments with ex- 

 cess of oxygen, one part in two hundred and fifty. From 

 the experiments with excess of hydrogen they deduce 1*9989 

 as the measure of the ratio sought ; from the experiments with 

 excess of oxygen, they infer that both series together jus- 

 tify the conclusion that one hundred volumes of oxygen com- 

 bine with very nearly two hundred volumes of hydrogen. 

 They do not compute the ratio from the experiments with 

 excess of oxygen : it would be 1*982. Since I began experi- 

 ments on the matter, Scott has published several statements 

 of his results. In his first paper,;}; he gives the results of 

 twenty-one experiments. He gives two sets of values of the 

 ratio sought ; one computed on the assumption that the im- 

 purities found in the residue after explosion were originally 

 distributed proportionally between the two gases, and the 

 other on the assumption that all the impurity was contained 

 in the oxygen. From the whole twenty-one experiments, he 

 gets the two values 1*9857 and 1*9941 respectively ; exclud- 

 ing two experiments in which the impurity was very great, 

 he gets 1*9897 and 1*9959; from the best four experiments 

 he gets the values 1*9938 and 1*9964 ; from the six best, 

 he gets the values 1*9938 and 1*9967. The mean error of a 

 determination was one part in two hundred and fifty on the 

 first assumption, and one part in Hve hundred on the second. 

 Rejecting the two worst experiments, the mean errors become 

 one part in five hundred and one part in seven hundred and 

 fifty parts. He gives the value 1994 as the most probable 

 value of the ratio sought ; but from a consideration of the 

 same experiments, Young§ judges that the value of the ratio 

 is between 1*996 and 1*998, and may perhaps be taken as 1*997. 

 In the autumn of 1887, Scott] stated that he had then made 

 over thirty experiments, and gave the most probable value of 

 the ratio as 1*996 to 1*997. In the spring of 1888, Scottf pub- 

 lished four other experiments with a new and larger apparatus ; 

 their mean is 1*997; the gases used sometimes contained as 



* Vol. xx. p. 38, 1805. f Vol. liii, p. 239, 1805. 



X Proceedings, R. S., vol. xlii, p. 398. § Nature, vol. xxxvii, p. 390, 1888. 



|| Br. Assoc. Trans., 1887, p. G68. . *-j[ Nature, vol. xxxvii, p. 439, 1888. 



