peans on the one hand and Asiatics on the other, and that after the separation certain peculiarities were developed in both groups, as, for example, the European e, which subsequently clung to all the subdivisions of the main group. For the European branch it seemed necessary to make two such subdivisions, the northern and southern, of which the former was again divided into Slavonic and Germanic, the latter into Greek, Italic and Celtic.

The Greek was here the hardest to dispose of. Some scholars assumed that the Celtic first freed itself from the South-European mass, after which the Greek and Italic remained together for a while; others (like SCHLEICHER) advocated the closer community of the Italic and Celtic; others, finally, divided the Greek wholly from Europe, and transferred it to Asia. This is the decision of GRASSMANN (Kuhn's Zeitschrift, 12, page 119), who speaks with great certainty of the many phenomena "in which the far-reaching harmony between a Greek and Aryan (pre-Brahmanic) nature becomes evident to us in language, poesy, mythology and life, and bears witness to the powerful intellectual development which the ancestral Greco-Aryan race passed through after its separation from the other branches." SONNE expresses the same opinion in his apparently forgotten article 1): Zur ethnographischen Stellung der Griechen, Wismar, 1869.

All these hypotheses, so far as they involve the idea of the separation of races or languages, were opposed by JOHANNES SCHMIDT, in an essay on the relationship of the Indo-European languages (Die Verwandtschaftsverhältnisse der indogermanischen Sprachen, Weimar, 1872). JOHANNES SCHMIDT starts from the same point where SCHLEICHER'S opposition to BOPP began, namely, the relation of the Slavo-Lithuanian to the Asiatic, but considers BOPP essentially in the right. It is, indeed, very remarkable that in both groups the k of kantam becomes a sibilant (or something similar), while the k of ka

1) I take the occasion to quote a sentence from this article: "But if in Sanskrit the verb of the main sentence assumes an unaccented form relatively to every preceding objective determination, I think we must recognize in this phenomenon, which is so thoroughly opposed to our European ideas, a remnant of pro-ethnical accentuation." (Page 3.)

"who" remains. Should not this remarkable agreement be explained as a result of common development, and is not SCHLEICHER'S assumption of historical accident inadmissible? If, however, Bopp's view is correct, there is no break between Asia and Europe, but only a "continuous transition" [kontinuierliche Vermittelung]. And SCHMIDT finds the same state of affairs in Europe. He recognizes that Greek, Italic and Celtic are intimately connected; but they do not form a historically distinct group, for as the Italic occupies an intermediate position between Greek and Celtic, the Celtic, on the other hand, is intermediate between Italic and Germanic, and further, the Germanic between Celtic and Slavonic, etc. Thus we can compare the Indo-European languages to a great chain of different rings, so linked together that it has neither beginning nor end. If we begin arbitrarily with the Indo-Iranian, the next ring will be the Slavo-Lithuanian, then the Germanic, Celtic, Italic, until the Greek is finally interlinked with the Indo-Iranian. The Armenian, which has been more accurately investigated only within the last few years, would take its place between the Indo-Iranian and Greek.

It will readily be seen that this transition or "wave-theory" (as its originator christens it, since the progressive movement within the bounds of language can be compared with the motion of the waves) agrees with the ramification theory in giving weight to the points of agreement (some of which have been mentioned) between the separate Indo-European languages, but differs from it in assuming a continous transition in place of ramification. We accordingly must first examine this assumption. I am of opinion that the transition theory is untenable, if it is understood in the sense that a continuous transition takes place between all Indo-European languages, as they are historically transmitted to us. Against it we have the fact that the separate languages form independent unities, each shut off from the others. It is true that we may be in doubt under which group single dialects (e. g. within the Germanic family) are to be ranged; but with the chief languages, as for instance the Germanic in its relation to the Slavonic, the case is different. There could never be a doubt whether a certain linguistic mass were Slavonic or Germanic;

fixed boundaries exist between Germanic and Slavonic, as well as between the other chief languages. We are accordingly led to suppose that formerly, when the Germanic was spoken by fewer people, it constituted an uninterrupted field of intercourse, within which the separate Germanic dialects were developed in the course of time. The same is true of the other languages. And even if we were willing to make the assumption (which it seems to me cannot be proved, in spite of the ingenuity expended upon it) that the neighboring domains of two adjacent languages, like the Slavonic and Germanic, stand in closer relation than those more remote from each other, this would only prove that single peculiarities of the former boundary-region had passed over into the two divided territories, and that the position of the parts of each domain had suffered no great displacement; the assumption would still remain possible that the separate Indo-European languages have been divided from each other for a long period by boundaries preventing intercourse. The transition hypothesis must therefore be understood in the sense that in primitive times the languages did indeed form one connected whole, in the manner described by SCHMIDT, but that then boundaries preventing intercourse were formed, and thus a separate life began, which subsequently gained a rich historical development. This modification of SCHMIDT's hypothesis, which evidently recommends itself by its universal historical probability, is due to LESKIEN (Die Declination im SlawischLitauischen und Germanischen, Leipzig, 1876). It would accordingly seem that the transition and ramification hypotheses do not unconditionally exclude each other, but are to a certain extent compatible.

Unfortunately an objection must be noticed, which proceeds from the stand-point of more recent investigations, and is opposed to both the ramification and the transition hypothesis. That is, it has been discovered, by the investigations of the last few years, that the data from which it was customary to draw conclusions in regard to the closer relationship of individual languages are not so decisive as was hitherto assumed.

In general, it is clear that not every point of identity between two languages can be regarded as an argument for an

original community of life. If, for example, some languages have lost the augment, which is still possessed by others, of course it does not follow that this loss necessarily took place during the common life of these languages. It must also be admitted that identity of vocabulary (unless this appears to an overwhelming extent) cannot be used to prove an original community of life, because the possibility always remains that a word which we only find in certain languages existed also in the others, although it has been effaced by the ravages of time. Our material is sensibly diminished by these considerations, so that, strictly speaking, we have as conclusive evidence only those new formations which are developed in common. Under this head were ranked until recently the division of the unitary Indo-European k into k and s (sz) in both the Asiatic and Slavo-Lithuanian families; the e of the European languages; the r in the middle and passive of Italic and Celtic; and the m in the Slavo-Lithuanian and Germanic dative plural. But another explanation for these facts has very recently presented itself. It is often assumed (as remarked above) that these cases are not examples of new formations in the individual languages, but that the manifoldness must be traced back to the primitive spcech. FICK took the lead with his paper on the linguistic unity of the Indo-Europeans of Europe (Die Spracheinheit der Indogermanen Europa's, Göttingen, 1873), in which, following ASCOLI, he showed that the two sounds of the Asiatic and Slavo-Lithuanian which were previously supposed to have originated from k were really the regular representatives of two different Indo-European k's (v. above, page 52). Then followed the very probable theory (also referred to above) that e belonged to the primitive speech; further, that the r of the middle and passive in Italic and Celtic may possibly stand in connection with the r of the Indian-re, -rate etc. (cf. WINDISCH, Beiträge von Kuhn und Schleicher, 8, page 465, note); and that the m of Slavonic and Germanic perhaps belonged originally not to the bh-suffix, but to another.

If, now, this whole mode of reasoning is justified (as I assume), then from such differences as these, which reach back into the primitive speech, no conclusions can be drawn

respecting the successive ramifications of the Indo-European languages, and it is necessary to adopt a skeptical position with regard to all the groupings hitherto attempted, with the single exception of the Asiatic group, which is held together by the common change of the old e into a.

In fact, I consider this stand-point the correct one at the present stage of investigation, and accordingly I think that our assertions in regard to the whole question of the mutual relation of the separate Indo-European languages must be reduced to the following. It is very probable that the primitive speech was not entirely homogeneous, as there was formerly an inclination to suppose. For if we are right in assuming that this speech passed through a development of centuries, the primitive race must have been very numerous at the time the inflection was fully perfected, and therefore differences in speaking must have already begun to manifest themselves within its limits, as described in general terms above (pages 52 and 59). These differences are the germs of some of the differences which we observe in the Indo-European languages. Others were added to these, after the primitive speech had divided into various individual languages. It is possible that the forefathers of the later Greek, Italic and Celtic nations were formerly settled beside each other in the way we are led to suppose from their present geographical position; but it is also possible that great displacements of the races have occurred, which render their former situation obscure. We will therefore content ourselves for the moment with acknowledging an original community of the Indo-European languages, but must abstain from classifying them into groups, with the exception of the Indo-Iranian.

This is true with regard to the Greco-Italic unity so often assumed. It is impossible to affirm with certainty that this unity did not exist, but it is equally impossible to assert that it can be demonstrated. Of the reasons adduced in its favor1)

1) SCHMIDT has very properly not introduced the word-comparisons of MOMMSEN, as they prove nothing. For a part of the words in question can also be found in other languages (as MOMMSEN himself acknowledges in the later editions of his Roman History), and the others (like milium, rapa, vinum) are possibly or probably borrowed words.