De Saussure, Fourier, M. Pouillet, and Mr. Hopkins have developed this speculation, and drawn from it consequences of the utmost importance; but it nevertheless rested upon a basis of conjecture. Indeed some of the eminent men above-named deemed its truth beyond the possibility of experimental verification. Melloni showed that for a distance of 18 or 20 feet the absorption of obscure rays by the atmosphere was absolutely inappreciable. Hence, the _total_ absorption being so small as to elude even Melloni's delicate tests, it was reasonable to infer that _differences_ of absorption, if such existed at all, must be far beyond the reach of the finest means which we could apply to detect them. [Sidenote: DIFFERENCES OF ABSORPTION BY GASES.] This exclusion of one of the three states of material aggregation from the region of experiment was, however, by no means satisfactory; for our right to infer, from the deportment of a solid or a liquid towards radiant heat, the deportment of a gas, is by no means evident. In both liquids and solids we have the molecules closely packed, and more or less chained by the force of cohesion; in gases, on the contrary, they are perfectly free, and widely separated. How do we know that the interception of radiant heat by liquids and solids may not be due to an arrangement and comparative rigidity of their parts, which gases do not at all share? The assumption which took no note of such a possibility seemed very insecure, and called for verification. My interest in this question was augmented by the fact, that the assumption referred to lies, as will be seen, at the root of the glacier question. I therefore endeavoured to fill the gap, and to do for gases and vapours what had been already so ably done for liquids and solids by Melloni. I tried the methods heretofore pursued, and found them unavailing; oxygen, hydrogen, nitrogen, and atmospheric air, examined by such methods, showed no action upon radiant heat. Nature was dumb, but the question occurred, "Had she been addressed in the proper language?" If the experimentalist is convinced of this, he will rest content even with a negative; but the absence of this conviction is always a source of discomfort, and a stimulus to try again.