Brugmann's article is included here for two reasons: it illustrates the growing control over articulatory phonetics; it reflects an awareness that the phonological and morphological levels of language are distinct, and that the one can be examined for insights into the other. Only a small portion of the article is given. The remainder is important for comparative Indo-European grammar -- the excerpt presented here, for general theory, especially for the assertion that there were vocalic nasals in Proto-Indo-European.
The recognition that PIE m and n were also vocalic led to considerable clarification of the ablaut in the Indo-European languages. Eventually the six resonants -- y w r l m n -- were classed together, for the clarification of many interrelationships in Indo-European grammar, such as an understanding of the Germanic strong verb bases. Brugmann's formulations are awkward in part -- another reason for merely providing excerpts. But publication of the article eventually led to the general assumption of vocalic resonants.
Verner's explanation of the phonological variation in sets like
OHG ziohan zōh zugum zogenillustrated that phonological change did not occur by morphological sets but rather in similar phonological environments. Accordingly, aberrancies in morphological sets might point to earlier phonological change. Brugmann led off the investigation of vocalic nasals by scrutinizing patterns in a morphological set, the n-stems. His procedure leads to that now known as internal reconstruction; in using it Brugmann is not as precise as is Saussure, but through its use he added conviction to conclusions which were supported by reference to general phonetic observations.
The article illustrates a tremendous number published in the last quarter of the nineteenth century which gradually clarified the important problems in the Indo-European family. Most dealt with minor problems and received little lasting acclaim. But their results led to the great compilations, such as Brugmann's Grundriss and to the grammars of the individual languages, such as Streitberg's, Pedersen's, Meillet's, Hirt's which are still widely used.
Karl Brugmann (1849-1919) is one of the great Indo-Europeanists. His capacity for work was enormous. He produced the Grundriss der vergleichenden Grammatik der indogermanischen Sprachen, second edition 1897-1916, which will never be superseded, for it is reliable, thorough and representative of Indo-European studies when many of the important problems had been clarified. Even though his writing was prolific, Brugmann was, however, not as clear and compact as Verner. His article on vocalic nasals lacks the immediate impact of Verner's, though it was as important for clarification of the Indo-European vowel system as was Verner's for the Indo-European accent.
Osthoff's essay "On the Question of the Origin of the Germanic n-Declension," just published in Paul and Braune's Beiträgen III 1 95., which I became acquainted with several months ago through a special printing, I will not hesitate to characterize as a work that will exert for a long time a most profound influence on research in the field of nominal stem formation and inflection. Its principles, to be sure, are not new, but its application to the given facts is new in many instances, and opens a great many quite new and farreaching perspectives. I will treat the most essential results of the essay at another opportunity, in order to use them as a basis to clear up various phenomena of stem gradation which remain obscure. Here I will deal with a matter which Osthoff handled only in passing, and which led him to a result whose validity I must question.
As is well known, the accusative plural in Old Indic is for the most part a weak case in those consonantal declensions, in which stem gradation occurs. For example, the accusative forms ap-ás 'waters', path-ás 'paths', ukshṇ-ás 'oxen', tudat-ás 'pushing' stand in contrast to the nominative plurals ā́p-as, pánthān-as, uksháṇ-as and tudánt-as, and to the accusative singulars ā́ap-am, pánthān-am, uksháṇ-am and tudánt-am. A different situation exists in the European languages, which along with the Aryan took part in stem gradation, and have this appear often, though never as clearly and openly as does the Old Indic. In the European languages the accusative plural is throughout a strong case, so that Osthoff sets up as basic form for his model the Indo-European stem uks-án-, uksan-as in the Aryan languages and uksán-as in the European languages. Naturally now, only one of these two forms can be considered the original Indo-European form. Either the Aryan or the European has altered the original relationship.
Osthoff decides on pages 35ff. in favor of the Aryan languages, and consequently asserts that the accusative plural in the original language was a weak case with a weak stem form and a stressed case ending. If I understand correctly, three considerations lead him to this assumption:
1. If uksán-as is taken in the original form, then a shift of the accent from the stem suffix to the ending has to be assumed for Old Indic. Shifts of the accent did indeed often occur in Old Indic toward the beginning of the word, but never in the other direction.
2. One encounters attempts in many places during the historical period of the Indo-European languages to assimilate the nominative and accusative towards each other in form. Now if uksán-as is assumed to be the original form, so that in the original language the nominative and accusative plural were formed alike on the one hand, and the accusative singular and the accusative plural were on the other hand accented alike, then Aryan would have taken exactly the opposite course and would have disturbed the original agreement of cases.
3. It is a quite unprovable hypothesis that the original form of the suffix of our case was -ams or -ans: the whole group of languages points only to -as. This proposition plays a role to the extent that it implies that the accusative plural was not formed by simply attaching the plural -s to the singular form in -am. Accordingly, there would not necessarily need to be agreement between the singular and plural as far as the stem gradation is concerned.
In contrast to these statements let us weigh the following:
1. Among themes which undergo stem gradation the accusative plural often appears in Vedic as a strong case with regard to the form as well as to the accent; for example, ā́p-as beside ap-ás, uksháṇ-as beside ukshṇ-ás, vṛshaṇ- as beside vṛshṇ- as. Among monosyllables without stem gradation, the accusative plural in Vedic is at times accented as a strong case, at other times as a weak case, thus rā́j-ás and rāj-ás, vā́ḱ-as and vāḱ-ás.1 In themes of this sort, in later Sanskrit some words appear with the stress on the stem syllable, such as nā́v-as and vāḱ-as, others with the stress on the case ending, such as mās-ás (Benfey, Vollständige Grammatik p. 318, IV). In Old Bactrian furthermore the accusative plural is probably about as prevalent in the strong form, and thus like the nominative plural in sound as in the weak form; in the strong, for example, in çpānō from çpā 'dog', dātārō from dāter 'giver'. urvānō or urvānō from urvan - 'soul' (See Spiegel Grammar p. 119).2 The Vedic uksháṇs corresponds very closely to Goth. auhsans; similarly, vāḱas and nā́vas to the Gk ópas and nēas. If therefore all the Indo-European languages are familiar with the accusative plural as a strong case and only the Aryan, beside the general Indo-European relationship, exhibits a different one, characteristically peculiar only to itself, it follows as a matter of course that this exclusively Aryan form, which even in the Aryan languages is not regular, is not the original form.
2. The fact that Greek from earliest times on does not use the same form for the accusative and nominative plural, but shows the ending -as (ópas) for the former and -es (ópes) for the latter, remains quite enigmatical in Osthoff's conception. For the view that in the common Indo-European language state, the ending of the nominative plural -as had already undergone weakening to -es, while at the same time the original form of the accusative plural uksan-ás persisted, and that precisely the old stress of the case ending caused Greek to preserve the pure -a-sound is to my mind highly artificial; and one must object to it above all, that the assumption that the high pitch on the end syllable -as prevented any departure from a pure a-sound is absolutely without basis. For where else in Greek is such an influence of the accent to be found? I look in vain for analogies and believe that instances like the genitive op-ós = 0ld Indic vāḱ-ás simply demolish Osthoff's hypothesis.
3. Everything indicates that our Indo-European ending -as actually originated from -ams. The m of pad-am (pedem) is without doubt essentially the same element as the m of akva-m (equum). If now, as no one doubts, the plural of akva-m was originally akva-ms, and from that form akva-ns, perhaps already in the original language; this form, however, differs from the singular only by the addition of the plural characteristic s,3 it is extremely probable that the plural of pad-am was pad-ams. None of the various languages prevents our establishing this as the original Indo-European form; Greek as a matter of fact, points to it most decisively. I will prove the correctness of this assertion below at relatively great extent.
If we are to consider -ams accordingly as the original form of the case suffix, then it necessarily follows that the accusative plural in the original Indo-European period belonged to the strong cases....
(293) We now arrive at the central point of our argument, at the demonstration that no phonetic obstacle exists to setting up -ams as the original form, and that Greek -as must necessarily be derived from -ans.
The vowel of the case suffix -am in Old Indic pā́d-am, Gk pód-a Lat. ped-em, etc. has been called a connecting vowel. For the sake of brevity let us maintain this name provisionally, without wishing to make any statement about the origin of the vowel. It is surely the same vowel which we encounter in the inflection of the verb before endings beginning with -nt, as in the third person plural before -nti, -nt, and -ntai, -nta, when these endings appear on themes which end with consonants, such as the Old Indic third person plural dvish-ánti (cf. first person plural dvish-más). We then call this too a connecting vowel. But now in both Aryan and European a significant difference is shown in the treatment of a when it is a connecting vowel and when it is thematic (part of the stem suffix); this holds true in nouns as well as in verbs, so that we are forced to conclude that the thematic a, for example from ákva-m Gk híppo-e and bhára-nti Gk phéro-nti, was already pronounced differently from the connecting vowel in the original Indo-European, for example pád-am Gk pód-a and as-ánti Gk é-anti. As these Greek forms set beside the original forms show, the difference in this language is still clearly distinct.
In Old Indic, the difference between the two a-sounds can be arrived at from a hard and fast rule, whose operation we will now exarnine more closely.
It is a constant rule that after a thematic a which is followed by a consonant,4 a nasal never disappears without a trace, and conversely that a nasal after a connecting vowel a disappears completely, if its syllable has low tone.
Let us begin with the verb. First, compare the indicative bhára-nti (Class I) and bíbhr-ati (Class III); the imperative bhára-ntu and bíbhr-atu; the participial accusative singular bhára-ntam and bíbhr-atam. Contrast further bíbhr-ati (Class III) and dvish-ánti (Class II), bíbhr-atu and dvish-ántu, bíbhr-atam and dvish-ántam. The law to be noted here is not invalidated by the fact that the third person plural middle of Classes II, V, VII, VIII and IX lacks the nasal, in spite of the accent on the connecting vowel, as in dvish-átē, ḱinv-átē, juñǵ-átē, tanv-átē and jun-átē. The fact that this stress is more recent and that the accent originally stood on the end syllable is proved by such Vedic forms as indh-atē, tanv-atḗ, etc. (Delbrück, Das altindische Verbum p. 74). There is the same type of relationship between the later tanvátē and the Vedic tanvatḗ as there is between the later máti- fem. (mens) and the Vedic form mattí- which also has lost the nasal because of influence of the conditions of the accent; more on this below.
If we now compare the formation of the accusative plural of the a-stems and the consonant stems, we find that áçvān i.e. (*áçva-ns) is like bhára-nti; conversely vā́ḱ-as, i.e. (*váḱ-ans) like bíbhr-ati. We therefore find that ending -as, with which this investigation began, has appeared in complete accordance with the sound laws for Indo-European -ans....
(303) This is the place to go into the articulatory phonetics of our question.... E. Sievers, in his splendid Grundzügen der Lautphysiologie, sets forth the principle, p. 24ff., that the liquids r and l and the nasals ṅ, n, and m can be vowels just as well as consonants. He teaches that, for example, in the usual pronunciation of ritten and handel, rittn and handl, n and l form the whole second syllable, and actually made up a syllable, and are to be designated as actual vowels. Accordingly, a strong distinction should be made between the nasalis sonans as in rittn, ātm and the nasalis consonans as in berittne, ātme; in the first words the nasal carries the accent of the final syllable, while in the second the accent is placed on the e. The sonore nasalis can carry the main stress of the word, as for example in the bisyllabic ńi-nein and ńi-ja as I know them for the expression of unwilling negation and heated asseveration in the Wiesbaden dialect. Now if we designate vocalic nasal m and n in contrast with the consonantal m and n, I am convinced that we have to establish for the original language beside ákva-m, ákva-ms the forms pád-m, pád-ms, and beside bhára-nti, bhára-ntai the forms as-ńti, ā̀s-ntai. By means of svarabhakti, i.e. the appearance of a short vowel from resonant consonants before m and n (cf. Sievers, Lautphysiologie p. 142) all the above-cited forms -m, -ms, -nti, etc., developed. First of all, therefore, svarabhakti developed in such forms as patár-m, and then spread also to those instances where a voiceless sound preceded the nasal.
My friend Osthoff urged me on to this conception. During a conversation in which I told him the main results of my study, he said: "One will probably have to posit the nasal in the original language precisely as a vowel" (in the sense of Sievers).5
With this assumption, we gain a double advantage. First, we can unite as one the double suffix forms -am, -anti, etc., (in consonant stems) with -m, -nti, etc., (in vowel stems). Second, all the qualitative vowel differences in the various languages which were cited above are simply solved, and I hope, some other difficulties too....