But a classification by language is far more satisfactory than any other sort of classification of nations. For when we think of nations we do not think first of all of their _physique_. The most important thing to know about them is not their hair was dark or red, their eyes brown or blue. What we care most to learn are their national character, their thoughts, their beliefs, their forms of social life. And for the days when we have no national literature, no history, to guide us, almost the only means of gaining reliable information upon these points is by a study of the language of the people in question. Language holds within it far better than do _tumuli_ or weapons, or articles of pottery or woven-stuffs or ornaments, the records of long-past times, records of material civilization and mental culture likewise. It holds these records, as a chemist would say, in solution in it; not visible perhaps to the mere passer-by; but if we know how to precipitate the solution it is wonderful what results we obtain. No sooner has he finished his classification of languages than a mine of almost exhaustless wealth then opens before the philologist--a mine, too, which has at present been only broached. He soon learns the laws governing the changes of sound from one tongue into another. We have noted experimentally some of these laws in the more simple relationships of language, as between English and German, where ‘tag’ becomes ‘day,’ ‘dorf’ ‘thorpe,’ and the like; and all relationships of language are answerable to similar rules. There are laws for the change of sound from Sanskrit into the primitive forms of Greek, Latin, German, English, etc., just as there are laws of change between the first two or the last two.[34] So we soon learn to recognize a word in one language which reappears in altered guise in another. And it may be well imagined how valuable such knowledge can be made. If we find a word common say to Greek and Latin, signifying some simple object, a weapon, a tool, an animal, a house, it is not over-likely that it will have changed from the time when it was first employed: the words of this kind which are now in use have, we know, little tendency to change. So that the time when this word was first used is in all probability the time when the _thing_ was first known to primitive man; and if the word is common to the whole Aryan family, or if it is peculiar to a portion only, then it is argued that the thing was known or unknown before the separation of the Aryan folk. I do not, of course, say that rule is never at fault, only that this is a better criterion than any other sort of research would afford us, and that by this method of word-comparison we get no bad picture of the world of our earliest Aryan ancestors.