The fact that there is a very definite connexion between the values of the atomic weights, and the properties, of the elements, lends some support to the hypothesis that the substances we call, and are obliged at present to call, elements, may have been formed from one, or a few, distinct substances, by some process of progressive change. If the elements are considered in the order of increasing atomic weights, from hydrogen, whose atomic weight is taken as unity because it is the lightest substance known, to uranium, an atom of which is 240 times heavier than an atom of hydrogen, it is found that the elements fall into periods, and the properties of those in one period vary from element to element, in a way which is, broadly and on the whole, like the variation of the properties of those in other periods. This fact suggests the supposition--it might be more accurate to say the speculation--that the elements mark the stable points in a process of change, which has not proceeded continuously from a very simple substance to a very complex one, but has repeated itself, with certain variations, again and again. If such a process has occurred, we might reasonably expect to find substances exhibiting only minute differences in their properties, differences so slight as to make it impossible to assign the substances, definitely and certainly, either to the class of elements or to that of compounds. We find exactly such substances among what are called the _rare earths_. There are earth-like substances which exhibit no differences of chemical properties, and yet show minute differences in the characters of the light which they emit when they are raised to a very high temperature. The results of analysis by the spectroscope of the light emitted by certain elements at different temperatures may be reasonably interpreted by supposing that these elements are separated into simpler substances by the action on them of very large quantities of thermal energy. The spectrum of the light emitted by glowing iron heated by a Bunsen flame (say, at 1200° C. = about 2200° F.) shows a few lines and flutings; when iron is heated in an electric arc (say, to 3500° C. = about 6300° F.) the spectrum shows some two thousand lines; at the higher temperature produced by the electric spark-discharge, the spectrum shows only a few lines. As a guide to further investigation, we may provisionally infer from these facts that iron is changed at very high temperatures into substances simpler than itself.