did not make any sure induction on the subject. In the present day, the relation of the refractive index to the density and atomic weight of a substance is becoming a matter of theory; yet there remain specific differences of refracting power known only on empirical grounds, and it is curious that in hydrogen an abnormally high refractive power has been found to be joined to inflammability.
The science of chemistry, however much its theory may have progressed, still presents us with a vast body of empirical knowledge. Not only is it as yet hopeless to attempt to account for the particular group of qualities belonging to each element, but there are multitudes of particular facts of which no further account can be given. Why should the sulphides of many metals be intensely black? Why should a slight amount of phosphoric acid have so great a power of interference with the crystallisation of vanadic acid?4 Why should the compound silicates of alkalies and alkaline metals be transparent? Why should gold be so highly ductile, and gold and silver the only two sensibly translucent metals? Why should sulphur be capable of so many peculiar changes into allotropic modifications?
There are whole branches of chemical knowledge which are mere collections of disconnected facts. The properties of alloys are often remarkable; but no laws have yet been detected, and the laws of combining proportions seem to have no clear application.5 Not the slightest explanation can be given of the wonderful variations of the qualities of iron, according as it contains more or less carbon and silicon, nay, even the facts of the case are often involved in uncertainty. Why, again, should the properties of steel be remarkably affected by the presence of a little tungsten or manganese? All that was determined by Matthiessen concerning the conducting powers of copper, was of a purely empirical character.6 Many animal substances cannot be shown to obey the laws of combining proportions. Thus for the most part chemistry is yet an empirical science occupied with the registration of immense numbers of disconnected facts, which may at some future time become the basis of a greatly extended theory.