In his presidential address to Section A of the British Association at Southampton in 1882, Lord Rayleigh alluded to an investigation which he had begun on the densities of hydrogen and oxygen, relatively to each other. The object of the research was to discover whether the atomic weights of these gases, determinable from their densities and from the proportions by volume in which they combine, was actually as 1 to 16, or whether some fractional number was necessary to express the weight of an atom of oxygen relatively to that of hydrogen. In 1888 his first account of the determination was published in the _Proceedings_ of the Royal Society. In 1889 he published a continuation of his first paper, and in 1892 he gave his final results; the number obtained was 15·882 for the atomic weight of oxygen, calculated from its density, hydrogen being taken as 1. In 1893 further experiments on densities were published,[25] those of oxygen and nitrogen being specially considered with reference to the density of air. He found the weights of one litre of oxygen, nitrogen, and air to be Oxygen 1·42952 grams “Nitrogen” 1·25718 " Air 1·29327 " A simple calculation leads to the composition of purified air. The percentage of oxygen must be 20·941, and that of “nitrogen” 79·059, in order to give a mixture of which the weight of a litre is 1·29327. Now, this corresponds with the results of the best analyses, quoted above. And the accuracy of these determinations of density is confirmed by this means, as well as by results of other experiments made by Leduc, von Jolly, and Morley. But Lord Rayleigh was not content to prepare his gases by one process only. The oxygen, of which the mean value of the weight of a litre is given above, was prepared in three different ways: by the electrolysis of water, by heating chlorates, and by heating potassium permanganate. The results showed that the only difference which could be detected, and that an extremely minute one, must be attributed to experimental error. The actual weights of the contents of his globe were-- Electrolysis, May 1892 2·6272 grams " " " 2·6271 " Heating chlorates, May 1892 2·6269 " " " June " 2·6269 " Heating permanganate, January 1893 2·6271 " These numbers are subject to a deduction of 0·00056, due to the fact that when the globe was empty of air, its capacity was somewhat reduced, owing to the external pressure of the atmosphere.