A Grammar of Proto-Germanic

Winfred P. Lehmann

Jonathan Slocum, ed.

I. INTRODUCTION

1.1. Definition of Proto-Germanic

Proto-Germanic (PGmc) is the reconstructed language from which the attested Germanic dialects developed; chief among these are Gothic (Go.) representing East Germanic, Old Norse (ON) representing North Germanic, and Old English (OE), Old Saxon (OS), and Old High German (OHG) representing West Germanic. PGmc is distinguished from the other Indo-European languages by phonological innovations such as the change of consonants characterized by Grimm's Law, by morphological innovations such as the introduction of the dental preterite and the n- declension of adjectives, by syntactic innovations such as the large number of modal auxiliaries, and by numerous additions to its lexicon.

As a reconstructed language, Proto-Germanic is not attested in texts; the material on which it is based is found in the attested dialects that developed from it. A yet earlier stage, reconstructed as Proto-Indo-European (PIE), includes means to account for and also to explain the reconstruction. That is to say, the beginnings of PGmc are assumed to overlap with the late stages of PIE, and data from later developments in Germanic dialects compared with evidence from PIE provides the basis for a grammar of PGmc comparable to those for languages spoken today, if not so detailed. PGmc may be dated from approximately 2500 B.C. to the beginning of our era, a period during which it underwent numerous changes.

Our grammar is arranged in three traditional systems:

• the phonological system (chapter 2);
• the syntactic system (chapters 3, 4, 5);
• the semantic system (chapter 6).

The semantic system is presented in relation to the cultural context of the reconstructed language.

Our knowledge of the phonological system and of the morphological component of the syntactic system is relatively good because much of the energy devoted to Germanic linguistics in the nineteenth and twentieth centuries was directed at these areas. Sentence patterns and the semantic system have received far less treatment; as a result, presentation of these requires considerable attention, especially their interpretation in accordance with general linguistic principles that have been developed in recent years. While the resulting grammar of Proto-Germanic may be less assured in some respects than are grammars of attested languages, it is represented here in compact form on the basis of the available data.

1.2. The Available Data

The only textual material contemporary with [late] Proto-Germanic is recorded in classical authors, or maintained in borrowings into other languages as exemplified by Finnish kuningas 'king'. Classical texts chiefly include proper names, such as Khariomēros in Greek and Langobardi in Latin texts. This material has been assembled and interpreted, as by Kluge (1913:5-47). Also important are the earliest Runic inscriptions; while they tend to be longer than the discrete Germanic items recorded in other languages, they are restricted in content and structure. Their language is archaic, though many can be dated only a few centuries before the time of other materials recorded in Germanic dialects such as Old English; few precede the time of our Gothic texts.

These materials provide the earliest data, but the most comprehensive data are provided in texts of Gothic, Old Norse/Old Icelandic, Old English, Old Saxon, and Old High German written in the first millennium A.D. Other dialects, such as Old Frisian, have fewer materials. The modern Germanic languages have generally developed so far from Proto-Germanic that they provide little evidence for its description; for most purposes, only the earliest texts can be used. In the reconstruction of the phonological system, morphologically isolated forms in the "everyday vocabulary" are highly important. An example is Go. faíhu, ON fé, OE feoh, OS fehu, OHG fehu 'cattle', on the basis of which PGmc fehu is reconstructed. When possible, as here, the reconstruction is shown with comparable forms in other languages, such as Lat. pecu 'herd', Skt páśu, Lith. pẽkus.

Inflected forms in the everyday vocabulary are similarly important for reconstructing the morphological system; among these are verbs like ON bīta, OE bītan, OS bītan, OHG bītan 'bite' and Go. (and-)beitan, 3rd sg. pret. and-bait, 3rd pl. pret. and-bitan, past ptc. and-bitans. On the basis of these, PGmc bītan- and comparable forms are reconstructed and supported by cognates such as Gk pheídomai 'I separate'. Another example is Go. bairan, bar, baurun, baurans, and comparable forms for ON bera, OE beran, OS beran, OHG beran, from which PGmc beran and the other forms are reconstructed, supported by cognates such as Lat. ferō 'I bear'. For these and other verbs, as well as nouns, adjectives and pronouns, the entire set of grammatical forms is reconstructed. In this way the morphology of Proto-Germanic is identified, as well as its phonology.

Conclusions may be supported by examination of borrowed forms, such as OE scrīfan, OS scrīban, OHG scrīban 'write'. Since these are inflected like inherited forms such as PGmc bītan-, they may seem to be native and not borrowed forms. But the Latin cognate scrībere has the same medial consonant as OHG scrīban; moreover the initial consonantal cluster scr is rare in Germanic, so that the assumption of a borrowing is supported. Such borrowings also support identification of the phonological elements in the items involved, like the -ī- in the Germanic forms, since the phonological structure of Latin is well known.

Other borrowings are often difficult to interpret, such as those taken from Germanic into Finnish. It has been assumed that some of these were adopted in Finnish before the Germanic consonant shift. By this assumption, Finnish kana versus the Proto-Germanic form of ON hane 'rooster' would have maintained the voiceless velar stop of Proto-Indo-European before the Germanic consonant shift. But the k- of kana can also be interpreted as a substitute for PGmc x- < h-, a phoneme not found in Finnish. By the interpretation of Finnish k- as a substitute for the shifted phoneme, the borrowing may have been relatively late. Secure conclusions can therefore be based on borrowings only when there is corrobative information of a relationship between the groups of speakers concerned.

The reconstructed syntax is based on sentences in the earliest texts, especially Runic inscriptions such as the Gallehus inscription of the early 5th century:

ek hlewagastiz holtijaz horna tawido
I Hlewagastir (the) Holtijan (the) horn I-made
'I Hliugast of Holt made the horn'.

The line is representative of Germanic word order, even though it is constructed in accordance with poetic principles as may be noted by comparing lines in the Old Icelandic Song of Weland such as line 4:

drósir suðrœnar, dýrt lín spunno,
women southern dear linen they-were-spinning
'southern women, they were spinning expensive linen'.

In both examples, the verbs occupy final position in accordance with the arrangements of SOV (Subject Object Verb, often referred to as OV) languages. Metrical requirements may have led to other patterns, such as the placement of the adjective suðrœnar after its noun, in contrast with the typical pre-positioning of adjectives in OV languages as exemplified by dýrt. As in these examples, conclusions regarding syntactic patterns are examined in accordance with typological principles; these assist in identifying patterns modified for stylistic or metrical reasons. Typological principles have been applied in the reconstruction of the phonological and morphological as well as syntactic components.

The semantic system is similarly reconstructed. The presence of words for cattle, sheep, goats, horses, etc., indicates use of domestic animals, which in turn provides evidence for the social context in which Proto-Germanic was spoken. Moreover, the word for linen is found in all the Germanic languages, lein in Gothic and līn in the others; it can therefore be reconstructed for Proto-Germanic, where its presence evidences the cultivation of flax for use in the production of clothing.

The early texts, then, provide ample data for reconstructing the phonological, grammatical and semantic systems of Proto-Germanic. These are proposed on the basis of well-established methods, as stated in the next section. The data for the social context are supplemented by descriptions in classical texts, chiefly Caesar's Gallic War and Tacitus' Germania, and by information from archeological discoveries.

1.3. The Linguistic Methods

Three methods are used to identify earlier elements: the comparative method, the method of internal reconstruction, and the examination of residues; the results of these are then considered in accordance with typological principles determined in the general study of languages. These principles are especially important in the reconstruction of syntax, but they apply also for the other components: expletives like "hmpf", for example, would not be taken into consideration when a phonological system is reconstructed. Morphological features and paradigms are reconstructed with reference to patterns that are attested in the declensional and conjugational systems of many languages.

1.3.1. The Comparative Method

Using the comparative method, comparable forms in related languages are examined and earlier items are reconstructed on the basis of similarity in form, in distribution, and in relation to other elements. Examples were provided in the previous section. The reconstruction of PGmc t on the basis of Gothic, Old English, and Old High German forms provides a more complex example. Five positions are illustrated:

The comparable forms of t in Gothic and Old English, supported by the same forms in the last two Old High German words, provide evidence for reconstructing Proto-Germanic t. Further examination of particular developments in Old High German leads to explanation of the z, zz in the first three words, and in this way supports the assumption of Proto-Germanic t.

1.3.2. The Method of Internal Reconstruction

Through use of internal reconstruction, earlier elements and patterns are identified using paradigmatic variations in a language. The procedures are based on the observation that sound change takes place in specific phonological environments, regardless of morphological classes or paradigms. If phonological alternations are found within morphological paradigms, it may be possible to reconstruct the earlier situation. Examples may be taken from Old English verbs:

The variation in the second consonant of these verbs is not found in the majority of Old English verbs, e.g.

On the basis of lack of variation in the majority of verbs, it may be proposed without comparison of material in other Indo-European languages that the s : r and ð : d developed from earlier single sources. Efforts to identify those sources may be guided by related forms like OE cost 'object of choice' or by examination of subsequent forms like NE seethe. The earlier consonants are then posited as s and ð.

The importance of the method of internal reconstruction lies in its applicability to data in one language alone. Its use permits reconstruction of earlier forms from forms that themselves are reconstructed, as for example in a language like Proto-Indo-European.

1.3.3. The Use of Residues

Residues are elements that are found among the common items of a language. Differing forms of these may be learned by children before they master the systems by which regular elements are constructed, such as the forms of man : men, woman : women in contrast with less common words like span : spans, woolen : woolens. Explanation of the plurals of man and woman and other such forms may then be proposed, such as that they were at one time formed by modification of the stem vowel through the process known as umlaut. Conclusions based on residues may be more problematic than conclusions based on use of the comparative method or the method of internal reconstruction, but they can also be supported by earlier forms of such words if cognates are attested.

Residues may also appear in morphological items. For example, Go. wait 'I know' and witum 'we know' have the forms of the preterite, although the glosses obviously indicate the present tense. An explanation may be found by adducing the weak verb Go. witan 'keep watch over'; comparison with it suggests that the root meaning is 'see'. The forms wait : witum then are accounted for through a shift in meaning from 'I/we have seen' to 'I/we know'. The further assumption then may be drawn that, at an earlier stage, the form providing the preterite in Germanic indicated a state resulting from completed action. In this way the preterite forms with present meaning are accounted for, as well as their basis.

1.4. Examination of the Results of such Applications
by General or Typological Principles

As suggested above in reference to syntactic elements, application of these methods is carried out with constant attention to general principles that are based on analysis and description of all known languages. Before a brief account of these is given, it should be noted that the historical grammars of Proto-Indo-European and the Indo-European languages produced in the past two centuries are based primarily on the grammars of Sanskrit, Greek, and to a lesser extent Classical Latin. In accordance with this basis, Proto-Indo-European is assumed to have had eight cases in its nominal system and an extensive verbal system. As a result of this assumption, explanations were required for the smaller number of cases in the Germanic languages and for its system of (only) two tenses, among other features. This basis can no longer be maintained, in part due to the reason proposed for its assumption.

The preeminent Indo-Europeanists Brugmann and Meillet stated that their central works, Brugmann's Grundriss (1897-1916) and Meillet's Introduction (1937), are not grammars that represent an earlier language but rather are summaries of the data found in the Indo-European dialects. In making that statement, Brugmann added that a historical approach was preferable, but that the time for it had not yet come. More than a century has passed since then, and the time has indeed come. As stated above, historical grammars must now be produced on the same basis as grammars of contemporary languages. Accordingly, a grammar of Proto-Germanic must be a description of the language from approximately 2500 B.C. to the beginning of the common era, as noted above. A grammar of Proto-Indo-European must be a description of the language from approximately 5000 B.C. to 2500 B.C. A grammar of the still earlier stage, Pre-Indo-European, must be a description of the language spoken from approximately 8000 B.C. to 5000 B.C. In reviewing below the current methods for providing and supporting descriptions of all languages, those methods of central importance for reconstructing Pre-Indo-European are given first. They continue procedures that were inaugurated in the 19th century and are generally referred to as "typological."

1.5. Specific Language Types

Specific types have been determined for all the components of languages, but those of special importance concern the semantic system and the syntactic system. Each has an effect on other systems of the language in question, as will be noted in the following sections.

1.5.1. Government and Agreement Languages

Until recently it was taken for granted that all languages were like English, in which government is central. For example, verbs govern nouns and pronouns, as in See her. So do adpositions, as in with her. No distinction is made in English grammar between nouns and verbs with animate and inanimate reference, so that I see the dog and I see the fire have the same pattern. All the major languages spoken today are structured in this way, from Chinese and Japanese through Hindi and Arabic to the languages of Europe. But in turning attention to many of the Amerindian and African languages, and to languages spoken by small groups elsewhere, as in the Caucasus, a different basic structure has been identified. In such languages, agreement is central rather than government. Two fundamentally different types of language are then recognized: Government Languages and Agreement Languages. Much of the investigation leading to this understanding was carried out by Soviet linguists (cf. Klimov 1983); they refer to the approach as "contentive," that is, based on content rather than form. In this approach, the semantic system is central.

Each type has two sub-types. For Government Languages, these are Nominative/Accusative (often referred to as Nominative or as Accusative) and Ergative; for Agreement Languages, these are Active and Class. In Class Languages, nouns and verbs are marked with affixes that represent semantic classes like humans, trees, and tools, such as Sesotho mo- for person in mo-tho, se- for tree in se-fate, n- for dog in n-tjá (Demuth 2000:273). In Active Languages, nouns and verbs are classified either as representing animate/active or as representing inanimate/stative items or processes, such as the prefix o- with active verbs like 'run' o-jan 'is running', and the prefix i- with words like 'good' i-katu 'is good' (Seki 1990:369). Sentences are constructed by pairing animate nouns and verbs that agree in classification rather than by government, as of verbs accompanied by objects. The discovery and description of Agreement Languages is highly important for Indo-European studies because Pre-Indo-European can now be identified as an Agreement Language of the Active type.

In Pre-Indo-European as an Active Language there were three parts of speech: nouns, verbs, and particles. There was no inflection. In the typical sentence pattern, verbs were final: items corresponding to objects preceded them. If there were overt subjects, they preceded objects and any adverbial elements. Relationships were indicated by particles. In the course of time, some particles came to be attached to nouns and others to verbs; the combinations resulted in the inflections of Proto-Indo-European. While Proto-Indo-European was a Government Language, and Proto-Germanic as well, residues of the earlier Active stage may be expected in accordance with section 1.3.3 above. Some of these have been maintained to the time of Proto-Germanic, of which two are noted here.

Active languages distinguish between inclusive and exclusive pronouns, so that a form corresponding to we could either include or exclude reference to the hearer as well as the speaker, that is, 'I and you' or 'I and others, but not you'. Prokosch (1939:282) pointed out that many languages make this distinction, such as "most Australian languages, nearly all of the Austronesian and most of the Dravidian group...", and he proposed that Proto-Indo-European had, as well; in this way he accounted for the use of PIE *we- for the first person plural in Germanic, as in Go. weis 'we' and as second plural pronoun in Latin vōs 'you', and also in the Gothic second person plural dative and accusative form izwis '(to) you' — citing, as well, forms from other Indo-European languages. He provided a description, but not an explanation. An explanation is now provided by the assumption that the twofold use is a residue from the Active stage of Pre-Indo-European, where forms of *we- indicated the inclusive meaning 'I and you'. As another "residue," Proto-Germanic has few adjectives that are inherited from Proto-Indo-European. In Active languages, stative verbs take the place of adjectives. Contentive typology in this way provides explanations for features of Proto-Indo-European and Proto-Germanic that had been noted, as by Prokosch, but not explained. More such features will be observed in the treatment of the syntactic and semantic systems in chapters 5 and 6.

1.5.2. VO and OV Languages

Quite independently of these features in the semantic system of Government Languages, characteristic features have been identified that distinguish the syntactic systems of languages. These are based on the position of the verb with regard to the object. As VO languages, English and the major European languages have the verb placed before the object. Many other languages, such as Japanese and Turkish, place the verb finally: they are OV languages. Subsets of these types are based on the placement of subjects, yielding six syntactic types. Biblical Hebrew, like the Semitic languages in general, places verbs initially, so that it is a VSO language, in contrast with the European SVO languages of today. Similarly there are two types of OV languages, SOV and OSV. Final position of subjects, yielding VOS and OVS languages, is rare, as are languages of the OSV type.

Syntactic and morphological elements are arranged in accordance with the two basic types. As a general principle, modifiers are placed outside the central structure, whether OV or VO, so that adjectives, genitive modifiers and relative clauses typically stand before nouns in OV languages, but after them in VO languages. Similarly, verbal modifiers, like markers of tense, mood and person, stand before verbs in VO languages, vs. after them in OV languages. Further, adpositions, like verbs, are placed before nouns as prepositions in VO languages but after them as postpositions in OV languages. And comparative constructions place the item compared, referred to as the standard, in accordance with the position of the object, often including a particle as in Japanese are yori kare ga takai 'that from this is expensive' = 'this is more expensive than that'.

Few languages are consistent. For example, adjectives stand before nouns in modern English, and older forms of English included more such inconsistencies. They reflect the still earlier situation of Proto-Indo-European, which had OV structure. Observation of the various patterns provides clues on the history of a language. French, for example, places adjectives after nouns, and is accordingly more consistently VO than English. It has developed farther from the OV structure of Proto-Indo-European than has English.

In examining the syntactic development of language, I some time ago proposed that when a language is adopted by many non-native speakers, it tends to become SVO. Modern Spoken Arabic for example, in contrast with VSO Classical Arabic, has become an SVO language. The shift of the Germanic languages from the OV structure of Proto-Indo-European to the VO structure of the modern dialects provides us with data on their social context through the past five millennia or so, as does that of Greek and Latin. We know from historical sources, such as Livy's History of Rome, that Latin was adopted by many speakers of other languages, such as Etruscan; its shift to VO structure may be credited to such adoptions. The shift in the Germanic languages is similarly explained.

1.6. The Phonological Structure of Languages

The principles of phonological structure have long been determined. Elements are grouped by phonetic value and distribution. In this way the [t] of stand is grouped with the [t] of tan, even though the latter is aspirated while the former is not; the (shared) functional sound, or "phoneme," is labelled /t/. Similarly, the [ts] of the Japanese word for one, hitotsu, is grouped with the [t] before [o] because [ts] stands only before [u] while [t] stands before [e a o]; both are variants ("allophones") of the phoneme /t/, in contrast with the different phonetic elements indicated between brackets such as Japanese [ts].

In current historical grammars, the phonological elements proposed are usually phonemes. Brugmann on the other hand proposed some phonetic elements for Proto-Indo-European; for example, on the assumption that some occurrences of Proto-Indo-European s were voiced, he included z as well as s in his system (1897:72). These and his other phonetic elements, like þ and ð, are no longer accepted as distinct phonemes in Proto-Indo-European.

Two subsets of phonemes are proposed for a language: the consonants and vowels are referred to as segmental; pitch and stress are referred to as supra-segmental. While modern English, German and so on have a stress system, the accentual system of early Proto-Germanic was based on pitch, as is that of the Chinese languages today.

1.7. The Syntactic Structure of Languages

Two systems are subsumed by syntactic structure: the morphological system and the system of sentences. The morphological system deals with the inflection and derivation of words. In traditional grammars the inflectional sub-system enjoys by far the most extensive presentation, and so-called exceptions are treated in detail. In this grammar, only the major inflections are presented (chapter 3); occasional forms that may be variations of individual items are assumed to be treated in dictionaries, including etymological dictionaries. The derivational elements are presented similarly (chapter 4).

The system of sentences deals with the structure of the sentence and its elements. These are presented (chapter 5) in accordance with the principles discussed in section 1.4 above. The earliest data in the dialects indicate that Proto-Germanic was an OV language; sentences as well as the morphological systems include residues of Indo-European Active structure. The treatment of such elements is critical for understanding the relation of Germanic to the other Indo-European languages, and for understanding its relation to Proto-Indo-European.

1.8. The Semantic Structure of Languages

The semantic structure of a language is least often viewed as a system. Groups of elements are recognized, such as kinship terms, but for the most part the vocabulary is divided into general groups such as terms for nature, for foods, for the household, and so on; in this way vocabulary reflects the culture and the social structure of the speakers, supplementing the information obtained from texts, from archeological findings, and to some extent from genealogical findings.

1.9. The Relationship of Germanic to the Other Indo-European Languages

The relationship of the Germanic language group to other language groups can only be determined by evidence in the languages. The closest language groups to Germanic are the Balto-Slavic, the Italic, and the Celtic. Yet, unlike Indo-Iranian, Greek, and Armenian, which have the augment as a common innovation as well as extensive verbal inflection, these four western groups lack any common phonological or morphological innovations. They share common vocabulary items, more for instance between Germanic and Italic than between Germanic and Celtic. Some of these may be attributed to a relatively late date, such as the name of a grain, either wheat or spelt, Lat. far, ON bǫrr, and the name of the goat, Lat. haedus, Go. gaits, as well as that of the male goat, Lat. caper, ON hafr.

Similarly, the Germanic words in common with Celtic indicate contacts between the two groups, but not major innovations; among them is a word for wagon, OIr. fēn, ON vagn, and a word for traveling, Irish rīadaim, OE rīdan. Among vocabulary items common to Germanic and Baltic are the words for eleven and twelve, which are innovations of the pattern "one/two left over" — Go. ainlif, Lith. vienúolika 'eleven' & Go. twalif, Lith. dvýlika 'twelve' — and words for movement, OE gengan, Lith. źengiù 'go, stride'. Other examples are given by Porzig (1954:106-147), some of which will be examined in the last section of this grammar.

In view of the absence of common innovations shared among other dialects, such as the augment, I assume that Germanic broke off independently — early — from Proto-Indo-European. Its archaic structure has been pointed out variously, as for instance in my article on the conservatism of Germanic phonology (Lehmann, 1953) and in subsequent publications.

1.10. The Development of Germanic

The development of Latin provides the model for understanding the expansion of the other Indo-European dialects. In its early form, it was the language of a small group of speakers in northern Italy in the eighth century B.C. Among other language groups at the time, that of the speakers of Etruscan was probably the largest. In the course of the following centuries, Latin was adopted by those groups, including also speakers of Celtic languages in the north, of Venetic, of Oscan and Umbrian, and even of Greek in the south, so that at the beginning of our era Latin was the most prominent language in the Italian peninsula. The bases for its expansion can only be imagined, but among them was military competence, as may be assumed from the account of the historian Livy. Other Classical historians, among them Herodotus, have provided material on various groups of speakers elsewhere, such as those north of the Black Sea; but for none of their languages do we have information comparable to Livy's for Latin.

The earliest description of the Germanic group of speakers was provided by Julius Caesar for the middle of the last century B.C., in the sixth section of his work on the Gallic wars, which with Tacitus' Germania of 98 A.D. remains central for any description of Germanic culture.

It may be concluded, then, that the Germanic group of speakers developed somewhat independently of the other Indo-European dialect groups. For a long time, the group may have been relatively small; but whatever the size, it was coherent at the time of the Germanic consonant shift for, unlike the later High German consonant shift, the earlier shift was carried through consistently among all speakers of Proto-Germanic, as was also the adoption of the dental preterite for weak verbs. Such consistently adopted changes can only have been introduced and generally carried out in a group that was in close intercommunication. Only after the Germanic shift did sub-groups develop: the speakers of Gothic, Old Norse, Old English, Old High German, and so on. As separate groups, they introduced innovations leading to the dialects that later became independent languages.