In 1831 Graham began the investigation of the peculiar phenomenon observed by Doebereiner. Repeating Doebereiner's experiments, Graham found that a portion of the hydrogen in the cracked vessels passed outwards through the small fissures, and a little air passed inwards: the water therefore rose in the jar, tube or flask, because there was a greater pressure on the surface of the water outside than upon that inside the vessel. Any gas lighter than air behaved like hydrogen; when gases heavier than air were employed the level of the water inside the vessel was slightly lowered after some hours. Graham found that the passage of gases through minute openings could be much more accurately studied by placing the gas to be examined in a glass tube one end of which was closed by a plug of dry plaster of Paris, than by using vessels with small fissures in the glass. The _diffusion-tube_ used by Graham generally consisted of a piece of glass tubing, graduated in fractions of a cubic inch and having a bulb blown near one end; the short end was closed by a thin plug of dry plaster of Paris (gypsum), the tube was filled with the gas to be examined, and the open end was immediately immersed in water. The water was allowed to rise until it had attained a constant level, when it was found that the whole of the gas originally in the tube had passed outwards through the porous plug, and air had passed inwards. The volume of gas originally in the tube being known, and the volume of air in the tube at the close of the experiment being measured, it was only necessary to divide the former by the latter number in order to obtain the number of volumes of gas which had passed outwards for each one volume of air which had passed inwards; in other words to obtain the _rate of diffusion_ compared with air of the gas under examination. Graham's results were gathered together in the statement, "The diffusion-rates of any two gases are inversely as the square roots of their densities." Thus, take oxygen and hydrogen: oxygen is sixteen times heavier than hydrogen, therefore hydrogen diffuses four times more rapidly than oxygen. Take hydrogen and air: the specific gravity of hydrogen is 0.0694, air being 1; the square root of 0.0694 is 0.2635, therefore hydrogen will diffuse more rapidly than air in the ratio of 0.2635:1.