related Aymara, are still the native dialects of a considerable part of South America; with the Tupi-Guarani, already referred to, on the east, in the valleys of the Amazons and its tributaries.

The condition of American languages is thus an epitome of that of the language of the world in general. Great and wide-spread families, limited groups, isolated and perishing dialects, touch and jostle one another. Such, in the vicissitudes of human affairs, must be the history of races and of their dialects. What families, once covering great tracts of the earth's surface, have been wiped out without a trace, what others have been reduced to mere fragments, what have started from a narrow beginning, and, by prosperous growth and by working in parts of other races, have risen to prominence on such points as these we must remain forever only imperfectly informed. We need to guard against supposing that, when we have succeeded in classifying all existing languages and determining their relations, we shall have gained a complete outline of the history of human language: the darkness of the past may hide a great deal of which we do not even catch a glimpse.

Some of the questions bearing on this point will engage our attention in the next chapter.

CHAPTER XIII.

LANGUAGE AND ETHNOLOGY.

Limitations to the scope of linguistic science: materials of speech not analyzable to the end; annihilation, transmutation, new creation, possible in it; cumulative character of evidences of relationship. Impossibility that language can prove either unity or variety of the human race. Relation of language to race, as transmitted institution only; exchange of language accompanying mixture of blood. Insolubility of the ethnological problem. Contributions to it of archæology and linguistics; merits of the latter; importance of the testimony of language to race. Reconciliation of the various lines of ethnological evidence. Inferior value of other classifications of language as compared with the genetic.

THE classification of languages given in the preceding chapter has confessedly represented only the present state of knowledge, and is liable to amendment hereafter, as further investigation shall bring more light. But its main features will probably stand unaltered. The leading independent families will continue separate to the end. One and another of those now recognized, it is true, may hereafter assume a dependent place, as branches of a wider and more comprehensive family, but there is no reasonable ground for anticipating that such will ever be the case with them all. To maintain this is not so much to limit the future of linguistic science, as, rather, to recognize the limits which in the

nature of things are set to its progress; as a brief and simple exposition will show.

We must not fail to appreciate the essential difference between the material of the physical sciences and that of our subject; that we have to deal with the usages of men, in all of which intervenes that indefinite element, the human will as determined by circumstance, by habit, by individual character; and that these do not admit an analysis penetrating to the ultimate elements. There is no natural substance which the chemist may not aspire to analyze; into whatever new forms and combinations an element may enter, he has tests which will detect its presence; neither new creation nor annihilation is possible; all change is but recombination of material always existing; there is no transmutation of one element into another. But it is altogether different with speech. A word, a whole family of words, perishes by simple disuse, and is as if it had never been, unless civilization is there to make a record of its departed worth. A whole language, or family of languages, is annihilated by the destruction of the community that spoke it, or by the adoption of another language by that community. When the Gauls learned Latin, there was nothing saved which, without the aid of external evidences, should show what their primitive speech had been; when the Etruscans were Latinized, but for the scattering words which they had written down, their speech passed out of all reach of knowl edge and many a dialect has doubtless gone out in a like way, leaving no such telltale records. The actual creation of the new in speech is, as we have seen, very rare; yet there is nothing whatever to prevent it save men's preferences. And it amounts, for all purposes analysis, to a new creation, when a derivative word gets

of

so far from its primitive, in form and meaning, that the tie between them is traceable only by external, historical evidence and of such cases all language is full. A formative element is annihilated when it is worn off from every form which it once made; such a one is created when it is fully established in its derived and subordinate use: no process of analysis that we have or of which we can conceive would ever find the lost masi of our first persons plural, or detect the presence of did in loved: there is wanted the historical support, for lack of which a host of other like cases cannot be accounted for.

The changes of linguistic usage are all the time separating in appearance what really belongs together: bishop and évêque are historically one word; so are eye and auge; so are I and je and ik and eyóv and aham; though not one of them has an audible element which is found in any other. And then, the same changes are bringing together what really belongs apart: the Latin locus and the Sanskrit lokas, 'place, room,' have really nothing to do with one another, though so nearly identical and in closely-related languages; likewise Greek onos (holos) and English whole; and so on. We may take the English language (as too many do), and compare it with every unrelated dialect in existence, and find a liberal list of apparent correspondences; which then a little study of the English words will prove unreal and fallacious. This is, above all others, the decisive fact which stands in the way of a comparison that shall penetrate to the bottom of the matter. If there were no resemblances in either the material or the structure of language save such as have a historical basis, we might let them be swept away as much as they would; what was left, if anything were left, would

suffice to prove relationship. As it is, the process of proof is not direct and absolute, but cumulative; the result comes from a sufficient number of particulars of which each, taken by itself, would prove nothing. We have had expressly to allow that two dialects may diverge from a common original so far that all sign of their kinship shall be lost; there may be a plenty of the altered products of common material in them both; but if it have gotten into the condition of bishop and évêque, it is of no use to the linguist. Accidental correspondences are capable of rising to a certain percentage; if all that appear stand at or near this figure, the case is one hopeless of settlement.

This cumulative character of the signs of relationship, the uncertain value of any single item, and the need of historical evidence to support their interpretation, set limits to the reach and competence of linguistic investigation. Thus far, the recognized families are such as have had a common development. There are even some of which the sole uniting tie is a common style of structure. If we cannot prove the American languages related except by the characteristic of polysynthetism, nor the southeastern Asiatic except by that of monosyllabism, it is obviously impossible to prove American and Chinese related by the material correspondences of their roots. In the present stage of linguistic science, root-comparisons are surrounded with too many uncertainties and dangers to have any value. All that have been made thus far are worthless; whether the future will show anything better, we may leave for the future to determine. There is no harm in any one's rating even too highly the possibilities of a progressive science like linguistics, provided he do not let his sanguineness warp his judgment as to what shall