A Reader in Nineteenth Century Historical Indo-European Linguistics

CHAPTER SIXTEEN

FERDINAND DE SAUSSURE

MÉMOIRE ON THE PRIMITIVE SYSTEM
OF VOWELS IN THE INDO-EUROPEAN LANGUAGES

Mémoire sur le système primitif des voyelles dans
les langues indo-européennes (Paris: Vieweg, 1887),
authorized reprint of the 1879 edition

Editor's Introduction

Saussure's achievement in his Mémoire is phenomenal. Published during his student days, actually in 1878 rather than the indicated 1879, it was far in advance of his time. Applying the method of internal reconstruction to Proto-Indo-European, he proposed the hypothesis that the long vowels had developed from short vowel plus sonant coefficients. His hypothesis was confirmed after Hittite was discovered. J. Kurylowicz in 1927 pointed out that the Hittite consonants transcribed with ḫ corresponded in some cognates to those which Saussure had suggested purely on the basis of phonological analysis of morphological patterns. The Mémoire is accordingly a fine example of the method of internal reconstruction, possibly the most dramatic application that has been made.

The consonants proposed by Saussure were related to Semitic by Hermann Möller in the following year, and have subsequently been known as laryngeals; their position in the phonological system of Proto-Indo-European and pre-Indo-European has subsequently been one of the intriguing questions of Indo-European linguistics. As is clear from the excerpts presented here, Saussure's chief interest was clarification of the Indo-European ablaut relationships generally, not merely of the roots with long vowels. These excerpts also illustrate the various uncertainties about the phonological system of Indo-European during the seventies. Only at the end of the decade was the "law of palatals" understood and with it the vowel system of late Proto-Indo-European as we know it was proposed.

Although the contributions of Ferdinand de Saussure (1857-1913) to historical linguistics were great, those to descriptive linguistics overshadow them. He is known as the founder of modern linguistics. His influence was largely exerted through a posthumous publication based on lecture notes: the celebrated Cours de linguistique générale, edited by C. Bally and A. Sechehaye. A student at Leipzig, he is largely responsible for establishing the eminent linguistic group in Paris, through his position at the École des Hautes Études, 1881-1891, which after him was led so long by his student, Antoine Meillet (1866-1936). He himself found life more congenial in Geneva, where he gave the lectures on which the Cours was based. His contributions have been capably discussed, as by Rulon Wells, "De Saussure's System of Linguistics," Word 3. 1-31 (1947) and Meillet, Linguistique historique et linguistique générale II. 174-183. We may note that his view of language as a system is apparent in the very title of the Mémoire which he wrote when he was barely twenty-one, and that from this view he made his notable analysis.

(1-6) The immediate object of this small work is to study the various forms under which is manifested what is referred to as IE a; the remaining vowels are not taken into consideration except to the extent that the phenomena related to a require. But if after we have come to the end of such a limited field the table of the Indo-European vocalism is little by little modified under our eyes so that we see it grouped entirely around a, and we take a new view of it, clearly it is the system of vowels in its entirety on which our observations will center and which should be indicated at the start.

Nothing is more disputed: the opinions are almost infinitely divided and various authors have rarely made a completely rigorous application of their ideas. In addition, the problem of the a is related to a series of phonological and morphological difficulties, some of which have yet to be solved, but many have not yet been stated. Thus in the course of our peregrination we will often traverse the most unexplored regions of Indo-European linguistics. If nonetheless we set out, though convinced in advance that our inexperience will often lead us into a maze, it is not recklessness, as is often said, that compels anyone who occupies himself with these studies to attack such questions: rather it is a necessity, it is the first school one must pass. For the question is not one of speculations of a transcendent order but of research into elementary facts without which everything drifts, everything is arbitrary and uncertainty.

I must withdraw some opinions which I have published in an article in the Mémoire de la Société de Linguistique de Paris, entitled: An Essay on a Distinction between different IE a's. Particularly the resemblance of Ar with the phonemes arising from r̥ led me to reject, very reluctantly, the theory of vocalic liquids and nasals, to which I now return after mature reflection.

Bopp and those who immediately followed the illustrious author of the Comparative Grammar limited themselves to stating that in regard to the three vowels a e o of the European languages, the Aryan uniformly showed a. The e and o were then considered weakenings characteristic of the idioms of the West and relatively recent developments from a single IE a.

The work of Curtius in the Sitzungsberichte der Kgl. Sächs. Gesellschaft der Wissenschaften (1864) enriched our understanding greatly: Curtius showed that e appeared in the same place in all the languages of Europe, so that it cannot have developed independently in each of them. And departing from the accepted idea that the mother language only possessed the three vowels a i u, he concluded that all the European peoples must have passed through a common period, during which they still spoke the same language. Also, that during this period a part of the a's were weakened to e, under an unknown influence, while the rest persisted as a. Later the various languages, separately from one another, had carried out a second split of the a, which yielded o. Yet in southern Europe this vowel must have arisen before the end of the Greco-Italic period, in view of the agreement of the o of the two classical languages, notably in the declension of the masculine stems in -a (Gk híppos = equos).

We believe we are representing exactly the system of Curtius by the following table:

[It is necessary, however, to add the following remark of the Grundzüge, p. 54: "the original dualism (Zweiklang) gan (Skt ǵan-â-mi) and gân (Skt perf. ǵa-ǵân-a), bhar (Skt bhar-â-mi) and bhar (Skt bhâra-s 'bundle') arose by an imperceptible substitution at the start: gen gan, bher bhar, then gon (Gk genésthai, gégona), bher, bhor (Gk phérō, phóros). But nothing can make us believe that there had ever been a time when Gk gen and gon, pher and phor would have been interchanged arbitrarily, of a kind so that one might have said Gk gonésthai, phórō or inversely gégena, phéros." Here accordingly the learned professor admits an original distinction of e and o, and derives the o of Gk ǵegona from IE ā.]

Fick's statement, Spracheinheit der Indogermanen Europas, p. 176ff., reproduces in general the preceding system. The ancient a is divided into a and e in the European period. When a word shows e in all the languages, it is necessary to assume that the change of its a to e goes back to this period. On the contrary it seems for a or o, that although this appears in a single language, it is necessary to admit that a still remained at the time of the community. The ablaut of Gk dérkomai dédorka, but above all of Gmc ita, at, is an admirable utilization of the splitting of a. On this last point see Curtius in the quotation above.

The system of Schleicher is different. Admitting in each vocalic series two degrees of reinforcement produced by the addition of one or two a's, he places for the series of a the three expressions: a aa āa.

He finds these three degrees in Greek: a is represented ordinarily by ε (e.g. Gk édō), but also by o (Gk podós) and by α (Gk ákōn). The first reinforcement, a + a, is represented by o when it is produced from ε, i.e. "Gk gé-gon-a, the first form: ga-gān-a; Skt ǵaǵān-a beside Gk e-gen-ómēn." This same degree is transmitted under the form of ā, ē, when it has a for base: Gk élakon, lélāka. The second reinforcement is o: érrōga. Gothic possesses the three degrees too; the other languages have confused the two reinforcements.

Since the genealogical tree of languages as Schleicher constructed it was not that which most of the other scholars had adopted and did not include a European period, it is clear that the e of the languages of Europe does not go back for him to a common origin. In particular, Goth. i has a different place from that of Gk e in his Compendium; the latter is considered the regular representative of IE a, Goth. i as an abnormal weakening. In formulating the following scheme according to Schleicher's system we therefore avoid the idea of a historically common development of the European vocalism:

It is also necessary to note that Gk a and Lat. a are not mentioned as reinforced degrees.

In a small work entitled: "Die Bildung der Tempusstämme durch Vocalsteigerung," Berlin, 1871, the Germanist Amelung, prematurely lost to our science, has attempted to apply the system of Schleicher in a very consistent manner and to combine it with the fact of common European e. In his eyes, this e is the normal representative of the non-reinforced a. The European a -- with which he also includes o as Curtius had done -- goes back to the first reinforcement which he designates by ā; and the second reinforcement (â) is long ā in the European languages. Presents, such as Goth. fara, Gk ágō, ózō accordingly show a reinforced vowel, and it is necessary to admit that they are denominatives. -- In a word, the dualism: e a is original, and the relation existing between them is that of simple to reinforced vowel. Note the table:

The debate which Amelung had on this question with Leo Meyer in KZ 21 and 22 did not bring any essential modification to this system, which has been presented in detail a second time in the Zeitschrift für deutsches Altertum 16. 161ff.

Brugman, Studien 9. 367ff. KZ 24.2 traces the existence of e, as a distinct vowel from every other, to the Indo-European period, without pretending by this that its pronunciation had been that of an e from its origin; and he designates its proto-type by a₁. Concurrently with this vowel the same scholar finds in Greek, Latin, Slavic o = Lithuanian, Gothic a = Sanskrit ā (at least in open syllables) a stronger phoneme which he calls a₂ and the origin of which was determined by the accent.

In accordance with this theory one generally arranges the following table, which nonetheless is certainly not that approved by Brugman himself, for he alludes (Studien 9.381) to the possibility of a great number of original a's:

In sum, one sees that for the languages of the West, the different authors, whatever their point of view, operate with three entities: the e, a, ā of the European languages. It will be our task to clarify the fact that there are really four different units, not three; that the languages of the North confused two fundamentally distinct phonemes still distinguished in southern Europe: a, a simple vowel, opposed to e; and o, a reinforced vowel, which is merely e in its higher form of expression. The dispute between those who favor the split (original a weakened partially to e) and those who favor a twofold original a (a₁, a₂ becoming e and a) -- this dispute, it is necessary to state, gets us nowhere, because by the a of the languages of Europe is understood an aggregate which has no organic unity.

These four kinds of a which we are going to try to find at the basis of the European vocalism we will pursue further still and arrive at the conclusion that they even belonged to the mother language from which the languages of the East and West arose.

Chapter I. The sonant liquids and nasals.

Before beginning the study of a it is necessary to determine carefully the limits of its domain, and at this point the question of the sonant liquids and nasals is presented. For anyone who admits these phonemes for the mother language will consider a number of vowels of the historical periods of the language recent and distinct from the question of the a.

The hypothesis of sonant nasals was first proposed and developed by Brugman, Studien 9. 287ff. In the same work (325) the author also touched on the subject of the sonant liquids, of which apparently the first notion is due to Osthoff.

1. Sonant liquids.

In the Indo-European mother language, the liquid, or liquids, if one accepts two of them, existed not only in the state of consonants but also in the state of sonants, that is to say, that they were able to carry a syllabic accent, capable of forming a syllable. This took place, as is known, in historical times, in Sanskrit. Everything leads one to believe that the sonant liquids never arose except through weakening, because of which the a which preceded the liquid was expelled; but this does not hinder our placing them, as we shall see, on the very same plane with i and u....

(8) 1. Root syllable.

The order adopted here to distinguish the different instances in which r{syllabic} appears is based on a new classification of roots, which can only be justified later but should not confuse the reader in the meantime.

We will deal only with the roots containing e. -- Every root which contains e in the languages of Europe has the ability of expelling this e and in this way taking on a weaker form, on the sole condition that the phonetic combinations so produced can be readily pronounced.

To be arranged under the roots containing e are those in which are found the diphthongs ei and eu and which one is accustomed to cite under their weakened form, deprived of e: thus, kei, sreu, deik, bheugh (ki, sru, dik, bhugh).

The i and u of these roots, as well as the liquid and nasal of roots such as derk bhendh can be called sonant coefficients (coefficient sonantique).

They are parallel in vocalism of the root. Depending on whether the e remains or disappears, their function varies: r l m n develop from consonants to sonants; i and u pass from a symphthongic state to an autophthongic.

A. Roots ending with a sonant coefficient.

Examples kei (weak form ki) sreu (w.f. sru) bher (w.f. bhr) men (w.f. mn).

B. Roots including a sonant coefficient followed by a consonant.

Examples deik (w.f. dik) bheugh (w.f. bhugh) derk (w.f. dr̥k) bhendh (w.f. bhn̥dh).

C. Roots without a sonant coefficient, ended by a consonant.

Examples pet (w.f. pt) sek (w.f. sk) sed (w.f. zd)....

(51) How then could the a and o of the languages of the South have arisen from one and the same original a? By what miracle could this old a be colored to o, and never to a, in all the times that it is found to vary with e? -- Conclusion: the twofold a and o of the classical languages is original, and it must be that in the single a of the North two phonemes were confused.

Confirmation: when a root contains a in Greek or in Latin and this root is found in the languages of the North, one observes in the first place that it there still shows the vowel a, but what is the important fact, that this a never alternates with e, as is the case when Greek corresponds with an o. Thus Goth. vagja = Gk okhéō, hlaf = Gk (ké)klopha are accompanied by viga and hlifa. But agis(a-) = Gk ákhos, or ala = Lat. alo do not have a parent form with e. On the other hand, the roots of the latter type have a characteristic, unknown among the first type: the ability to lengthen their a (agis: ōg, ala: ōl) of which we will have to take account below.

Brugman has designated with a₁ the prototype of European e; his a₂ is the phoneme which we have called o up to now. As to this third phoneme which is Greco-Italic a and which constitutes a portion of the a's of the languages of the North, we will designate it by the letter A, after noting well that it is not the parent of e(a₁) nor of o (a₂). -- Excluding for the time being the other possible kinds of one obtains the following table:

(134-135) § 11. Grammatical role of the phonemes A and O̬. Complete system of the primordial vowels.

When one considers the following cases of the permutation a₁,a₂: Goth. hilfa hlaf, Gk kléptō kéklopha, Gk. híppos híppe, and when one compares with them the following cases of the permutation A Ā: Goth. sake sōk, Gk láskō lélāka, Gk númphā númphă, the temptation is strong, assuredly, to set up the proportion Ā : A = a₂ : a₁. But this would be to get involved in a course without result and to misunderstand the true character of the phenomena. For greater clarity we are going to construct at once the system of vowels such as we understand it. For the time being we are concerned only with root syllables.

The phoneme a₁ is the root vowel of all roots. It can be alone in forming the vocalism of the root or it can be followed by a second sonant which we have called sonant coefficient (p. 8).

Under certain conditions which are not known, a₁ is replaced by a₂; under others, better known, it is expelled.

When a₁ is expelled, the root remains without vowel when it does not contain a sonant coefficient. When it does, the sonant coefficient comes to be alone, or in an autophthongic state (p. 8), and provides a vowel to the root.

The phonemes A and O̬ are sonant coefficients. They cannot appear alone except in the reduced state of the root. In the normal state of the root, it is necessary that they be preceded by a₁, and the combinations a₁ + A, a₁ + O̬ give rise to the longs Ā, Ō̬. The permutation a₁: a₂ takes place before A and O̬ as elsewhere.

Useful designations

For a₁A and a₁O̬ after contraction: Ā₁ and Ō̬₁.

For a₂A and a₂O̬ after contraction: Ā₂ and Ō̬₂.

The theory summed up in this table has been applied to all the types of roots above except those which contain A and O̬....