or less in common. Certainly, the trait is barely suggested in Eastern Europe, that is, in the Slavic and Finnic languages. Perhaps this geographical fact was the only reason that precluded its use as an argument for the pre-Germanic substratum theory. Since this article is to be restricted to a scrutiny of the arguments for that theory, I might omit the analytic trend from the discussion altogether. But this at least I wish to suggest, that the old controversy whether the compound tenses were transferred from Germanic to Vulgar Latin or vice versa is hardly to the point. They developed independently in the two groups at a time when the phonetic weakening of the endings had paved the way to such an innovation. Nothing in IndoEuropean led up to it, but the system as such is such a frequent device in languages the world over that we do not have to seek for any explanation other than the natural logic of these formations. Or should we suppose that the trait was particularly strong in Western Europe on account of the analytic tendency of a language, or perhaps a group of languages of pre-IndoEuropean aborigines, of which Basque is the last descendant? Certainly, this language has carried the system of compound verb forms to a much higher degree of complication than any Indo-European language. We do not know the age of those numerous Basque compound verb forms; but they are hardly of later origin than those of the Germanic and Romance languages. Also, they are so different from the latter that they could not have served as direct models. But it is remotely possible, I admit, that the habits of thought in Basque and kindred languages were carried over into the Indo-European languages of Europe to a considerable extent, perhaps through the medium of Celtic. Unfortunately, we cannot venture the remotest guess at the grammar of continental Celtic. The compound tenses of Irish seem to stand midway between those of Basque and the Romance languages.

On the whole, then, the facts are distinctly against the substratum theory. There is nothing, I am convinced, that would justify it in the form in which it has been advanced heretofore, namely, that Germanic is the result of a modification

of a common Indo-European language through the influence of an earlier population of Northern Europe. The most important Germanic traits are not innovations, but systematic developments of Indo-European linguistic tendencies. It is only in historical times that a few real innovations appear. But they do not belong to the Germanic languages alone, but appear in neighboring languages as well. They may, but need not, be due to a substratum. If they are, it is an ethnic element that spread, more or less, over all of Western Europe.

But while I cannot accept the substratum theory in its present form, I am not altogether skeptical about it in principle. Germanic, it is true, I consider a straight descendant from Indo-European without any pre-historic layer. But what is Indo-European after all? The Indo-Europeans themselves were doubtless a highly mixed group; perhaps the Germanic group was, for geographical reasons, more homogeneous than the other groups, but even for them the Roman reports about their racial purity are certainly greatly exaggerated. It would be preposterous to assume that the Indo-European language was created ex ovo, after the racial elements had been blended into a sort of national unit. Was, perhaps, Indo-European the original possession of one of the racial components and, during the process of amalgamation, constantly transferred to the others? The racial mixture was such a complicated one, and its developments spread over so many millenniums that we cannot assume that linguistically one of the components was always the giver and never, or rarely, the taker. Let us consider this: Our reconstruction of Indo-European would date the common language back some five or six thousand years. Surely, European languages as such are many times older than that. Would anyone suppose for a moment that, for instance, the Cro-Magnon man, some thirty or fifty thousand years ago, had no language, in the usual sense of the word? Unfortunately, very little is known of the racial genealogy of the population of Europe. The terms 'Nordic, Alpine, Mediterranean,' that have come into such general acceptance of late, are hardly more than terms of convenience for the present results of the blending. Dixon's 25 painstaking analysis of cranial types does give us a 25 Roland B. Dixon, Racial History of Mankind. New York, 1923.

glimpse into the complicated character of the process itself, but it is a highly problematic one, on which no linguistic theory can be built up. Still, in all of this complication, the obvious fact remains that that there exists some common factor of racial affinity over Western Europe, another over Central, and a third, over Eastern Europe. This is doubtless in part a result of national groupings and geographical and cultural conditions, but at least to some extent it appears to be the consequence of differences in racial mixture. Every part of Europe has racial stocks from many sources; but a different strain predominates in each of its main divisions.

In a similar way, the Indo-European languages have probably resulted from a variegated compromise between several, perhaps numerous, languages from the very beginning. In what was later the Slavic branch, there prevailed a different linguistic blend from the (later) Celtic branch. Perhaps there never was a time when the ancestors of the later Slavs and Celts could have understood one another. All of them, to be sure, have certain elements of structure and a considerable stock of vocabulary in common. But other structural features are geographically distributed, more or less similar to the racial divisions. For instance, the West is highly analytic, the East (Slavic and Ugro-Finnic) only slightly so. The Western languages (except Basque, where conditions are altogether different in this respect) have a pre-positive article, the Eastern, a post-positive article or some similar pronominal suffix. The North-East (Norse and Balto-Slavic) forms reflexive verbs by suffixing the reflexive pronoun of the third singular to all persons, the West is without any such device. The West abounds in tenses and has only traces of aspects-in the East the tenses are rudimentary and the aspects prevail. In the Eastern languages, the adjective invariably precedes the noun, in the West it follows. Also the generally recognized centum-satem division may be mentioned here, although I attach to it much less importance than to the group of structural features just quoted.

This geographical distribution of linguistic features will become clearer from the following tabulation:

It is important to observe that the geographical distribution of these contrasts is not always the same. The Germanic group shares some of the Western and some of the Eastern features. On the other hand, the Eastern and the Western areas have several features in common, particularly the marked palatalizing and in general assimilating tendencies, while in the Germanic group assimilations are comparatively very rare. Such overlapping is natural. Each of the three main divisions probably had considerable elements of every racial component of the other two, but their strength and importance varied, and with it their linguistic influence. For the sake of convenience, we may tentatively accept the terms 'Nordic, Alpine, Mediterranean,' however unfounded these generalizations may be. It is, of course, altogether arbitrary and doubtless incorrect to ascribe any sort of linguistic unity to any or each of these groups; but I shall do so hypothetically, merely in order to illustrate the point more clearly: The West, let us say, is predominantly Mediterranean, with considerable Nordic and Alpine admixture; in the Center, the Nordic type predominates over the other two, and in the East, in pre-historic times, the Alpine. None of these types spoke 'Indo-European' in the sense of our reconstruction. But one of them (it does not matter which—let us say, the Nordic, merely because we have accepted that as the Central type) was perhaps more tenacious linguistically than the other two. Most of the general elements

of the structure of its language, together with the most fundamental parts of the vocabulary (pronouns, numerals, words of relationship, etc.), may have prevailed everywhere-more or less. But considerable influence was exerted by the other two components; each of the three (or five, or ten, or more) resulting language types can with equal right be called 'Indo-European as far as its relationship to the others is concerned, for each of them had a considerable share in the common structural basis. But they were never alike. In this sense, and only in this, I believe in a substratum.

We cannot hope ever to arrive at a complete understanding of these involved processes; even the assumption of only three component elements is an utterly unwarranted simplification. But the reconstruction of a uniform Indo-European language is merely scientific fiction, albeit an extremely fruitful one. It need not and should not be given up, but it should be conservatively interpreted. The assumption of an Indo-European 'nation' (Meillet, Introduction, p. 375, 398, etc.) is wholly unfounded. Each IE. language possesses a great many elements that are lacking elsewhere. They reflect primary components that were less important in the other groups. But that is not a substratum in the generally accepted sense of the term.

But even though we shall never be able to assign any given linguistic characteristics to a definite racial type, at least one task in this direction is waiting for linguistic science: The existing material must be arranged and compared with the meager anthropological data in existence. What the IndoEuropean languages have in common has been analyzed and assembled with sufficient clearness and completeness. It will now be necessary to scrutinize with equal accuracy their distinguishing features. For instance, we must bring together those traits and elements that distinguish Slavic from all, or most, other IE. languages. This material must then be compared with the corresponding facts in the neighboring languages, whether Indo-European or not (in this instance, with Germanic, Finnic and Caucasian). I am quite certain that in this way we shall find many linguistic features that cover certain terri