From whichever side I view the matter, I am unable to comprehend how μείζους and μείζονος could have both been formed in the same dialect from usilovoos, and am accordingly of the opinion that only the form usilovs arose by a phonetic process from the primitive form uɛílooos (Sanskrit máhīyasas), while usilovos is a formation by analogy from the nominative μsícov. How the n can be explained in the latter is, indeed, a point where there is room for disagreement. (V. the detailed discussion of BRUGMAN in Kuhn's Zeitschrift, 24, page 1 seq.)

We are now prepared to give a comprehensive answer to the question proposed in the beginning, viz., do phonetic laws as such admit of no exceptions?

We have seen where we may expect to meet with such laws. Certainly not in the collective mass of any existing speech, whether it be a popular dialect or a literary language. For it is not probable that all the individuals within a linguistic community will speak precisely alike. Therefore we can only expect to find these laws in the case of the single individual, or rather, if we wish to be quite exact, only in the average speech of an individual at any one moment. Now from what an individual speaks or would speak at a definite moment of his life, if he allowed the whole mass of his vocabulary to pass through his vocal organs, we must first subtract all that can be regarded as borrowed (in the broadest sense), and then all phonetic formations which depend upon the action of analogy. When this is done, the form which remains is the result of phonetic change alone. Here, and only here - leaving out of account the possible fluctuations of a transitional stage-we may expect complete uniformity in the treatment of all analogous cases, and in this sense we must assert that phonetic laws as such admit of no exceptions.

At the same time, it must be confessed that complete uniformity of phonetic change exists nowhere in the world of actual fact; but there are sufficient grounds for assuming that regularly occurring phonetic change is one of the factors to whose united action the empirical form of language is due. In single instances, it is true, it will only be possible to approximately reproduce this factor in its purity.

DELBRÜCK, Introduction to the Study of Language.

9

We can see at once from the above discussion whether and how far we are able to speak of "laws", or still more, of "natural laws", within the field of phonetics.

It has been shown that the phonetic laws which we postulate are nothing but uniformities which appear in a certain language and period, for which alone they are valid. Whether the expression "law" is really applicable here is doubtful. Yet I avoid entering upon a discussion of the notion "law", as employed in natural science and statistics, because I find that the term "phonetic law" has become so fixed by usage that it cannot be eradicated, and furthermore, because I can propose no better expression in its stead. It is also a harmless term, if we keep in mind that it can have no other sense than that defined above.

I cannot approve of characterizing phonetic laws as "natural laws". These historical uniformities can evidently bear no resemblance to chemical or physical laws. Language is a result of human action, and consequently phonetic laws are not based upon the regularity of natural processes, but upon that of apparently arbitrary human activities.

CHAPTER VII.

THE SEPARATION OF THE RACES.

As we mentioned on page 1, Sir WILLIAM JONES, as early as the year 1786, remarked that every philologist who compares Sanskrit, Greek and Latin necessarily arrives at the conclusion that these three languages must be derived from a common source, which perhaps no longer exists, while there are no such decisive grounds for assuming the same relation for Gothic and Celtic. We found that FRIEDRICH SCHLEGEL took a backward step in comparison with JONES, since he came to the conclusion that the Sanskrit language is the older, the

others younger and derived from it. Even BOPP, in the beginning of his literary career, does not always express himself correctly; thus in his Conjugationssystem, page 9, he speaks of the languages which "spring from the Sanskrit, or with it from a common mother"; but later he rightly characterizes the relation as a sisterly one. He is also on his guard not to overestimate the originality and antiquity of Sanskrit. Thus in the first edition of the Comparative Grammar there is a note (subsequently omitted) to § 605, which runs as follows:

"In my Conjugationssystem, and in the Annals of Oriental Literature (London, 1820), I have called attention to the fact that the Sanskrit second person plural tutupá is a mutilated form, and in the earlier sections of this book allusion has often been made to the fact that in single instances the Sanskrit is at a disadvantage compared with its European sister-idioms. It therefore surprised me that Prof. HöFER in his work Beiträge etc., page 40, made the sweeping assertion that the new investigators have not succeeded in wholly emancipating themselves from the unhappy delusion that the Sanskrit has preserved its original perfection of structure with inviolable fidelity'. I for my part never ascribed to the Sanskrit such fidelity to its original structure, and it has always been a pleasure to me to call attention to the cases in which it must yield the palm to its European sisters" etc.

BOPP has no fixed name for the one ancestral speech Stammsprache] from which the individual languages were derived. He speaks of the one ancestral speech, of the period of linguistic unity, of the primitive period of language, of the primeval formative period, etc. This one ancestral speech, which no longer exists, was in BOPP's opinion essentially similar to its sister languages. It is especially worthy of mention that he did not claim that it was incapable of change. Instead he assumes "that at the time of the identity of those languages which are now separated, many disturbances had already taken place in the organism of that one ancestral speech". (§ 673.) Thus he assumes that in oldest times the feminine in ā had an -s in the nominative, but had already lost it in the period of linguistic unity. I cannot find that Bopp expressed any conjecture regarding the home of the race which spoke

this primitive language, and he has in general no inclination to view things from an ethnological stand-point. The ethnological point of view was first emphasized by KUHN in the Osterprogramm des Berliner Realgymnasiums for 1845. (Cf. WEBER, Indische Studien, 1, page 323.)

According to BOPP, the individual languages have freed themselves from the "primitive home” by an "individualizing” process. The expression "separation of languages" [Sprachtrennung] also occurs (§ 493). Of the nearer or more remote relationship, i. e. of the order followed in the separation of the languages, BOPP's opinion was as follows: in Asia the Sanskrit and Medo-Persian are intimately connected; in Europe the Greek and Latin. In regard to the position of the Slavonic Bopp's opinion changed in the course of time. First (Vergl. Gram., 1st edition, page 760) he considered the Lithuanian, Slavonic and German as "triplets"; later (Ueber die Sprache der alten Preussen, Abh. der Berl. Akad., 1853, page 80) he defined his view thus: "The separation of the SlavoLithuanian idioms from the Asiatic sister-language, whether we call this Sanskrit or leave it without a name, is of later date than that of the classic, Germanic and Celtic languages, yet prior to the bifurcation of the Asiatic portion of our linguistic domain into the Medo-Persian and Indian branches." He did not assume a special relationship between the languages of the Celts and Romans.

SCHLEICHER was the first to establish a formal system of ramification for the Indo-European languages (under the figure of a genealogical tree). He agreed with BOPP in his assumption of a closer relationship between the Indian and Iranian branches (which is, indeed, irrefutable), and between the Italic and Greek languages, but differed from him in regard to the position of the Slavo-Lithuanian. He attempted to prove that the similarity of phonetic structure, which indubitably exists between the Asiatic languages and the Slavo-Lithuanian, does not date from primitive times, but originated in each group individually. Thus he assumes that the word for "hundred" in the parent speech was kantam, and that from this, after the separation of the primitive race into two, çatam

was developed in the Asiatic division, and suto in the Slavonic, quite independently of each other; so that the similarity between ç and s in this word, in which the Greek and Latin have preserved the old k, could not furnish any basis for genealogical conclusions. (Cf. Beiträge, 1, page 107.) Accordingly, he wholly separates the Slavo-Lithuanian from the Asiatic division, and with JACOB GRIMM places it with the Germanic group. The chief proof of the close relationship of these languages consists in their agreement in the dative plural, where they exhibit an m, while the other languages have bh (e. g. Slavonic vlěkomu and Gothic vulfam, but Sanskrit vṛkebhyas). Further, since SCHLEICHER places the Celtic with the Italic (Beiträge, 1, 437), he obtains the following three groups: 1) Asiatic; 2) Slavo-Germanic; 3) Greco-Italo-Celtic. He defined the historical relation between these groups according to the fidelity with which each (in his opinion) has retained the primitive type. This fidelity seemed to him least in the Slavo-Germanic branch; he therefore assumed that this division was first separated from the primitive race, and then the Greco-Italo-Celtic, so that the Asiatic group alone remained.

It is plain, however, that this chronological classification depends upon a very questionable line of argument. The more advanced phonetic decay of the Slavo-Germanic (if, indeed, it can be regarded as proved) may be simply owing to the fact that the Slavo-Germanic has developed more quickly than its sister-tongues. SCHLEICHER does not, therefore, adduce sufficient grounds for dividing the Slavo-Germanic from the great European mass to which it geographically belongs. That it also belongs there from linguistic considerations was shown by LOTTNER in Kuhn's Zeitschrift, 7, page 18 seq. He establishes two great groups, the Asiatic and the European, the latter being especially characterized by a common 7 in opposition to the Asiatic r (e. g. ñoló, Gothic filu, as opposed to Sanskrit purú). A further characteristic was added by G. CURTIUS, in the e which appears uniformly in many positions, in opposition to the Asiatic a (e. g. pépw, fero, Gothic baira, i. e. běra, as opposed to bhárāmi). Thus the supposition seemed very probable that the Indo-Europeans, who spoke a uniform language while they were together, first split apart into Euro