But what is curious is that this threefold classification, and the consequences to which it leads, should not at once have been fully reasoned out, nay, that a system most palpably erroneous should have been founded upon it. We find it repeated again and again in most works on Comparative Philology, that Chinese belongs to the _isolating_ class, the Turanian languages to the _combinatory_, the Aryan and Semitic to the _inflectional_; nay, Professor Pott[8] and his school seem convinced that no evolution can ever take place from _isolating_ to _combinatory_ and from _combinatory_ to _inflectional_ speech. We should thus be forced to believe that by some inexplicable grammatical instinct, or by some kind of inherent necessity, languages were from the beginning created as _isolating_ or _combinatory_, or _inflectional_, and must remain so to the end. It is strange that those scholars who hold that no transition is possible from one form of language to another, should not have seen that there is really no language that can be strictly called either isolating, or combinatory, or inflectional, and that the transition from one stage to another is in fact constantly taking place under our very noses. Even Chinese is not free from combinatory forms, and the more highly developed among the combinatory languages show the clearest traces of incipient inflection. The difficulty is not to show the transition of one stratum of speech into another, but rather to draw a sharp line between the different strata. The same difficulty was felt in Geology, and led Sir Charles Lyell to invent such pliant names as _Eocene_, _Meiocene_, and _Pleiocene_, names which indicate a mere dawn, a minority, or a majority of new formations, but do not draw a fast and hard line, cutting off one stratum from the other. Natural growth, and even merely mechanical accumulation and accretion, here as elsewhere, are so minute and almost imperceptible that they defy all strict scientific terminology, and force upon us the lesson that we must be satisfied with an approximate accuracy. For practical purposes Humboldt’s classification of languages may be quite sufficient, and we have no difficulty in classing any given language, according to the prevailing character of its formation, as either isolating, or combinatory, or inflectional. But when we analyze each language more carefully we find there is not one exclusively isolating, or exclusively combinatory, or exclusively inflectional. The power of composition, which is retained unimpaired through every stratum, can at any moment place an inflectional on a level with an isolating and a combinatory language. A compound such as the Sanskrit +go-duh+, cow-milking, differs little, if at all, from the Chinese _nieou-jou_, _vaccæ lac_, or in the patois of Canton, _ngau ü_, cow-milk, before it takes the terminations of the nominative, which is, of course, impossible in Chinese.