Proto-Indo-European Syntax

7. Syntactic Developments from PIE to the Dialects

7.1. Changes in the Sentence Pattern.

By the time of Classical Greek and Latin, the OV syntactic pattern of PIE had been largely modified to a VO pattern. Yet the VO pattern was still inconsistent in the early classical period of both Greek and Latin, as is illustrated by such relic patterns as OV comparatives, and also in Classical Sanskrit. In spite of such inconsistencies, this stage of development is that “reconstructed” by comparison and represented for the parent language in the standard handbooks, such as Brugmann's. It is virtually an automatic outcome of comparative studies relying on the texts of the classical periods of the three widely studied ancient dialects, Indic, Greek, and Italic. But the Old Hittite texts and the Archaic and Strophic hymns of the Rigveda reflect a language with far more OV characteristics than are proposed for the original IE language in such handbooks. As indicated in the preceding chapters, PIE must be reconstructed as basically OV. In addition to reconstructing PIE in this way, Indo-Europeanists must determine the developments from PIE to the individual dialects. This chapter will sketch some of those developments.

Rather than by reconstructed texts, PIE may be represented by some of the most archaic Hittite materials, such as those found in the Chronicles and Laws—for example, the following sentence from the Chronicles (KBo III 28. 17f.):

In this passage both clauses have final position of the verb. The relationship between the first and second clauses is not specified; yet, as the translation suggests, the first clause may be taken as a relative, indicated as such simply by preposing. Such a preposed relative construction is one of the notable characteristics of OV languages. The verb forms are in accordance with OV structure; huišnušket contains a nu causative affix added to huiš- ‘live’, followed by the -šk- iterative-durative suffix, which is also found in papriškir. Literally, then, it means: ‘he continued causing to live’. As a noteworthy feature of this sentence the negative applies to the root huiš- , not to the entire verb form. In this way the clause represents a construction earlier than the subjective system of PIE, one comparable to the Turkish sentence cited above, Chapter 4, Example 1. If the subjective domination had been in effect, the negative of this form would have meant ‘he was not causing to live’ rather than ‘he was causing not to live’, i.e., ‘he killed’. Morphological constructions like huišnušket must be assumed for PIE, that is, loosely constructed verb forms in which the various affixes maintained their independence.

Reflexes of such verb forms appear in other early dialects, as in the “Ionic iterative preterites” attested chiefly in Homer (Schwyzer 1939: 710-712). Two such -sk- forms are found with other archaisms in Odysseus's account of Tantalus's torment:

Besides maintaining the PIE meaning of the -sk- suffix, the Ionic iterative preterites preserved the archaism of being made from “aorist” as well as “present” stems. Moreover, the lack of an augment in the preterites indicates their archaic construction. It must not be taken as evidence for a compositional derivation, with an otherwise unattested *skon ‘I was’. The explanations proposed for the Ionic iterative preterites before the discovery of Hittite must be abandoned. Rather, the forms must be regarded as reflexes of the PIE verb system, in which -sk- functioned as in Hittite. In addition to these archaic forms, the Homeric passage is also of interest for its preposed complement piéein and for other early constructions, such as the use of the optative aorist kúpsei in the volitional sense.

Before examination of the later constructions, another Hittite sentence may be cited, with preposed relative and complement clauses; the sentence is from the Old Hittite Proclamation of Anittaš (KBo III 22, 49-51; Raman 1973:63):

In this passage the relative clause, unlike that in Example 1, has a relative marker, kuiš. The relationship of the second clause however is unspecified; it may be interpreted as a relative, as it is by Raman: ‘the [one who] resettles Hattusa’. The introductory nu however suggests a more specific interpretation, such as the conditional of the translation. Whatever the interpretation, the modifying clauses are still preposed, and the pattern of arrangement is OV for the clauses as well as for the postposition appan and the preposed genitive nepišaš. Other preposed complements have been illustrated in sentences cited above, such as akuwanna before the verb pianzi in the Hittite example, Chapter 2, Example 17. Later, as early as the Homeric texts, such complements follow.

Before examples of such patterns are cited, a sentence from Middle Indic will be given, from the Pali report of the Buddha's Fire-Sermon. The only finite verb in this sentence, pakkāmi, follows its object cārikaṁ; but subordinate expressions are placed after as well as before this verb.

This comparatively simple example of Middle Indic prose is considerably more complex syntactically than the Vedic prose passage given above in Chapter 3, Example 64, especially in its modifiers of the principal clause. It is important to account for the basis of the syntactic changes which have taken place in Indic, as well as those in the other dialects.

Explanations for specific syntactic changes have been proposed by suggesting influences from neighboring languages or internal developments. The influence of Dravidian may well have been important in the development of morphological constructions in Middle Indic, such as the compounds in this sentence. But since the Dravidian languages are OV they cannot have brought about changes in order of complements and possibly not the gerundial and participial constructions which characterize Middle Indic prose. Such characteristics are also found in the IE dialects of Europe, as a passage from Saint Augustine's Confessions, cir. 400 A.D., may illustrate. The passage is Augustine's well-known observation of how he learned to speak, Confessions 1.8:

Since passages in other dialects, like this sentence of St. Augustine's, also contain the typical characteristics of VO structure, I suggest that the major cause for the shift to a VO structure, as of complements and subordinate constructions from preverbal to postverbal position, was internal. The shift to VO structure resulted from the development of the subjective quality of the late PIE verb.

7.2. Reasons for the Shift to VO Structure, Illustrated by Complements.

Section 2 of Chapter 4 has illustrated that the PIE verb, and also the entire clause or sentence, came to be dominated by the person marker. The Pali example given above (Example 4) also illustrates such domination. There is no question about the “subject” of the gerund viharitvā, or of the participial abhirantaṁ, or of the postposition saddhiṁ; similarly, the “subject” of praebentēs in Example 5 is clearly hominēs. Other nouns than the subjects, to be sure, had their own attributive and appositional modifiers, indicated by means of nominal inflections, as do bhikkusaṁghena, ordine, and so on. But the entire Pali sentence obviously relates incidents referring to the subject Bhagavā with its third-person singular verb. In addition to Augustine's sentence, Greek texts could be cited for further examples of such “subjective dominance,” as could texts from other dialects.

With such clarity of focus, sentences could have complements follow verbs as well as precede them, as in the examples given in § 4.8. The change in position of complements has been discussed by various Indo-Europeanists, for example Sommer (1959). His compact treatment proposes that infinitives used as complements first had case forms like the accusative and dative, which in the OV period would have preceded the principal verb, e.g., Lat. piscātum īre ‘go with the aim of fishing’ (Sommer 1959:92-93). Thereupon patterns arose, like Gk. bȇ d’ímen ‘he set out to go’, exemplifying “infinitives of purpose” (ibid.:98-99). Participles were also used as complements after verbs, as in Sommer's example (1959:103):

As a further development the complements were introduced by markers for infinitives and then also for finite verbs, as the following examples with hṓste illustrate (Smyth 1956:510-511).

Comparable to Thucydides's use of an infinitive after hṓste to indicate result is Xenophon's use of a finite verb form:

In this last example the result clause is like those in Modern English and other VO languages, as is also the position of the objects with apékrupse, and other syntactic characteristics, such as the preposed conjunctions kaì...kaì.

These examples illustrate how complements postposed after the principal verb came to be expressed by means of complete clauses as well as nominal forms of verbs. Examples of the uses of postposed clauses in other dialects are cited by Delbrück (1900:319-338, 420-447; 1897: 463-475; see also Grace 1971:372-374). As Indo-Europeanists have pointed out (Delbrück 1900:295-406, 423-435), such postposed complementary clauses were also introduced by relative pronouns, or by particles based on relative markers. In this way complements came to use syntactic devices developed for nominal modifiers. Moreover, sequences of modal and tense forms also served to indicate relationships between clauses in complex and compound sentences. These sequences, like the ellipses, assimilations, and mixtures of constructions discussed by Brugmann (1904a:289-705), are characteristic of the dialects after they came to be VO in structure rather than of PIE.

7.3. The Postposing of Relative Clauses.

When a relative clause directly precedes its antecedent, the relationship is clear from the arrangement alone. The same is true when it directly follows its antecedent, as in English sentences in which the equivalent noun of the relative clause is the object.

But when relative clauses do not directly follow the antecedent, some other syntactic device must be used to identify it, as in:

By means of the intonation pattern the English relative construction can be specified as referring to color as well as to car, whichever has the primary stress. The syntactic device of selection may be used similarly to specify the antecedent of a relative clause, as in German:

This example illustrates the basis for the course of development of the position of relative clauses in PIE. When a marker was introduced, the relative clause could follow as well as precede the matrix clause.

Raman has thoroughly examined the development of relative clauses in Hittite. In most passages, even in the Old Hittite texts, the relative clause accompanied by a marker already has a referent in the matrix clause, such as -an of Example 3 in this chapter; a referent is found also in Example 1, which has no relative marker. This sentence may be considered an example of the stage directly after that in which preposed relative clauses directly preceded their referents, as illustrated in Example 17 of Chapter 3. Specified by a marker, relative clauses could follow their matrix clauses as well as precede them. The postposed position would have been suggested by the postposed position of complements.

Already in late Hittite, relative clauses have come to be postposed, as they are in the other dialects (Raman 1973:167-201). Raman cites a particularly interesting passage which is preserved in two versions, the older with a direct statement rather than a relative clause, the later cited here with relative marker kuit and postposed arrangement; the passage is from ritual texts, which, according to Professor Kammenhuber, were modernized by continuous use in cult ceremonies (Raman 1973:182):

Raman accounts for the VO characteristics in the later Hittite text through borrowing; this explanation is especially cogent for another Hittite text with VO characteristics, which was written in Egypt, for it is quite plausible that scribes who also knew Egyptian or Akkadian would have taken over their VO patterns (Raman 1973:188). But such an explanation for the shift to VO structure is less weighty for the texts in the other IE dialects. Already in the Rigveda almost half of the relative clauses are postposed, and in later texts, such as those of Homeric Greek, the percentage increases. Concerning these we have no basis for assuming modifications by bilinguals who also knew VO languages. Unless further materials become available for the languages which Greek, Latin, and other dialects displaced, the best explanation for the shift to VO relative constructions is that of internal innovation. Such an explanation is supported by the parallel shift to VO structure of relative clauses, though with differing relative markers in the various dialects. For some of the dialects, however, as for Hittite, the shift may have been advanced by influence from VO languages.

7.4. Change in the Position of Attributive Adjectives and Genitives.

In view of their relationship with modifying relative clauses, change in the position of adjectives and genitives would be expected with that of relative clauses. The change was slow; in the Italic branch, for example, adjectives and genitives are consistently postposed only in the Romance languages, though the main clause pattern of Late Latin was VO already at the time of Saint Augustine. The position of attributive adjectives in the sentence cited from his Confessions (Example 5) is not fixed; variīs is postposed after vocibus but preposed before mōtibus.

During the course of the change a remarkable construction arose in the northern languages, Slavic, Baltic, and Germanic: the so-called weak adjective inflection. In this inflection the relative marker of reduced relative clauses was maintained as a suffix in Slavic and Baltic, e.g., OCS vino novoje ‘wine (which is) new’, in contrast with vino novo ‘new wine’, and Lith. geràsis ‘the good one’ as opposed to gẽras ‘good’. The weak or definite forms of adjectives can be accounted for through their reduction from relative constructions, in contrast with the strong declensions, which continue the regular IE adjective inflection (Lehmann 1970b). Germanic, which lacked reflexes of the yo marker, generalized the n inflection for definite adjectives. Thus the weak inflections provide further insights into changes which took place as IE dialects moved to a VO structure.

7.5. Change in Verblike Elements and Verbal Modifiers.

After the arrangement of relative clauses was modified, the change in position of attributive adjectives and genitives is readily understandable, inasmuch as these constructions are reduced forms of relative clauses. But verblike governing elements, such as postpositions, and their patterning may seem not to be closely related to verbs. New High German however provides an example of their relationship in change. After an OV order was adopted in subordinate clauses around 1500 A.D., postpositions were introduced. The entities chosen for postpositions are not of syntactic importance, taken as they are from various sources; wegen ‘because of’, for example, used postpositionally since around 1600, is in origin a dative plural; another, entlanc ‘along’, first attested as postposition in 1751, developed from a phrase. The German phenomena therefore suggest that verblike elements, such as postpositions and prepositions, are placed in accordance with the position of the verb with regard to its object, as stated in the principle of § 1.3. Since comparative constructions are essentially verbal, they too would be remodeled in accordance with the principle.

In this way the verblike constructions of the IE dialects were changed to those of VO pattern, much as the nominal modifying constructions had been. Prepositions came to replace postpositions; comparative constructions came to have the pattern of Latin maior quam tū ‘bigger than you’ rather than the relic pattern tē maior (‘you-from bigger’). Moreover, adverbs typically came to follow verbs, as did the verbal complements. Such changes varied in time of adoption and spread from dialect to dialect; investigations into the reasons for the variation must be pursued by specialists in the several dialects.

Among the verbal innovations of greatest interest for pursuing the changes in the verbal system are the newly introduced reflexive and reciprocal phrases with pronouns. These developed when the suffixed verb system no longer agreed with the changing pattern of arrangement of the clause to VO (Lehmann 1973a, 1973b). The OV middle formation was then lost or modified in meaning. The new verbal phrases are among the earliest analytic verbal constructions. Other analytical constructions were introduced in innovating such paradigms as the perfect passive forms of Latin. Subsequently, as the verbal endings were weakened, analytical devices were also used for tense and mood forms. The system of verb markers thus changed from a system of suffixes, as expected in an OV language, to a system making use of auxiliaries and function words.

7.6. Changes in Selectional Categories.

Among the most significant changes in selectional categories was the thoroughgoing systematization of paradigms. Already in late PIE the three persons had distinctive endings in the plural as well as the singular, as the result of increasing systematization of category markers. But only in the dialects were verbal paradigms systematized to such an extent that grammars can propose principal parts for verbs. The Latin verb system may be the most obvious example; three of the four conjugations have parallel sets of forms that can be constructed on the basis of four principal parts, e.g., laudō, laudāvī, laudātum, laudāre ‘praise’. Even the remaining conjugation shows few traces of the unpredictability of the PIE verb, in which a large number of forms could be made from one root, with no means of predicting the particular forms, even from principal parts, as examination of the forms given by Whitney for Vedic roots (1885) will indicate. The various later dialects systematized in this way not only verbal inflections but also those of nouns, adjectives, and pronouns. In the course of remodeling the IE forms into symmetrical paradigms, new, or weak, inflections were introduced, such as the Latin v perfect exemplified in laudāvī and the k perfect of Greek, the d preterite of Germanic, and so on. The source of most of these formations is still disputed and may never reach general agreement. In proposing theories for their origins a syntactic model is particularly important. For, as is revealed by the investigation of any one form, such as the Germanic dental preterite, scholars have often looked at apparent parallels, such as the composite forms in the Romance languages, without considering whether the syntactic situation was the same in the two differing languages. By the view of syntactic structure held here for PIE and the early dialects, such weak preterites must have arisen by generalization of suffixes rather than from periphrastic formations (Lehmann 1943a). Scholars who hold a different point of view must support it with arguments based on general syntactic theory and a particular framework of IE syntax, not merely with unmotivated parallels in other languages.

In the systematization of paradigms, thematic vowels were highly important. The regular verb conjugations generally involve reflexes of thematic vowels, as in laudāre, or parallels, as in habēre and audīre. Moreover, the most systematic and complete inflections are thematic, such as the adjective inflection illustrated in Latin by bonus, -a, -um ‘good’ in contrast with the inflection of the third declension, as in gravis, -e ‘heavy’. Examination of the grammars of the individual dialects illustrates the pervasiveness of thematic inflections.

Phonological changes, especially the loss of the laryngeals and the realignment of the resonants, however, came to obscure the earlier parallelism between forms even in late PIE. Thus the nominative feminine ending of bona, < -ah < -eh, was earlier parallel with those of the masculine and neuter (-os, -om); but upon loss of the h the parallelism was lost.

Such losses of parallelism contributed to the gradually increasing analytic character of the dialects, especially those dialects in which a stress accent led to weakening of inflectional syllables. Even in the highly inflected period of Classical Latin, forms like rosae (feminine genitive/dative singular, nominative plural) and rosīs (dative/ablative plural) had to be disambiguated by function words or by arrangement; prepositions like sub would distinguish the ablative from among homonymous forms; its role as subject or predicate nominative would distinguish the nominative plural. In the course of time the inflectional syllables were further reduced or lost, partly through lack of distinctiveness, partly through replacement of their function by prepositions and other syntactic markers. Moreover, the development towards analytical structure correlates with departure from an OV structure with its rich possibilities of suffixation. Investigations of the extent to which these various forces were of effect in the selectional changes belong among syntactic problems of general concern. For the understanding of such problems the individual changes in each of the dialects need detailed study.

7.7. Changes in Modification.

Modification, or sandhi, played an important role in the delimitation of PIE words and of phrases consisting of an accented word followed by an enclitic. Meillet concluded that words were syntactic units in PIE, using metrical practices in Indo-Iranian, Greek, and Latin to support his conclusions (1937:136-140). Words were demarcated by the possibility of only one accent and also by features of modification of their final elements.

The final elements were modified in accordance with phonological classes. Presumably even in late PIE the modification of final stops corresponded to that of early Indic, where they consisted only of the implosive segment; without discussing other changes, it may be recalled that in Armenian, Greek, Slavic, Baltic, Germanic, and Celtic such final stops were lost. The final fricative also had a weakened form, as may be determined from its reflexes in Sanskrit, such as ḥ and its merger with vowels, e.g., o < -as.

These and other modification phenomena are of interest here because of the evidence they supply for assuming that the word was a syntactic and morphological unit in PIE and because of the effects of such modifications on syntactic patterning. Since the final segments of words, those subjected to weakening, consisted in large part of the markers for syntactic categories, further weakening or loss led to the greater use of analytic devices for the expression of syntactic relationships. The effects are clearest for dialects which introduced a stress accent, such as Celtic, Germanic, Italic, and Middle Indic. In contrast with the well-preserved inflectional endings of Baltic and Slavic, the dialects with stress accent underwent major losses of inflectional markers. The details of the effects of such losses and their relationships to other changes in the dialects are intricate, and by no means ascertained in spite of admirable studies of “laws of finals.” For syntactic study the effect in reducing the selectional devices rather than precise details on the phonological processes concerned is of primary importance, for it led to greater reliance on function words and other analytic devices in syntax.

7.8. Changes in Intonation.

The information on intonation patterns and their uses in PIE is slight, as noted in § 2.7. Yet, at least in early Vedic and presumably in PIE as well, pitch accent on the verbs of subordinate clauses is one indicator of subordination. When a stress accent was introduced, such a use of intonation must have been disrupted. Jules Bloch states that this device for indicating subordination was lost by the time of Classical Sanskrit (1965:311). A great deal of further study is required, however, to permit more than general statements on the changes in intonation patterns in Indic and other dialects. Such study must also involve the changes in use of enclitics, that is, the loss of application of Wackernagel's law.

For Greek, Allen has made preliminary analyses in admirable investigations of the bases of Greek meter (1968:106-124). A further imaginative investigation of changes in the accent patterns of the dialects has been carried out by Paul Kiparsky (1973). These investigations promise greater understanding of PIE intonation patterns and their changes, with resultant changes in syntax. For the time being these investigations are chiefly concerned with individual accentual phenomena; as Allen states, “we know virtually nothing about ‘tonal syntax,’ i.e. the way in which such patterns interacted with one another and with clause- and sentence- intonations in continuous speech” (1968:118). The patient research of Indo-Europeanists like Allen and Kiparsky, based on that of earlier scholars like Kurylowicz and Hirt, may in time amplify our knowledge and provide some improved observations on the use of intonation in PIE syntax and its subsequent changes.

7.9. The Syntactic Changes, with Reference to the Community of PIE Speakers

The overall pattern of changes from PIE to the dialects is from an OV to a VO structure. Individual dialects differ in the time and extent of the changes.

The southern European dialects have become consistently VO: Albanian, Greek, the Romance languages. The group farthest to the west, Insular Celtic, with its VSO structure, has developed farthest from the OV structure of PIE.

Persian and the northern European dialects—Baltic, Slavic, Germanic—are inconsistently VO. These dialects illustrate some of the problems that complicate the understanding of the syntactic changes; for especially Slavic and North Germanic were influenced by neighboring OV languages. As a result they reintroduced OV characteristics. The dialect which, apart from Tocharian, was most heavily influenced by neighboring OV languages was Indic, which was contiguous to Dravidian. Like Slavic and North Germanic, Indic illustrates that an unbroken trend of change, or drift, cannot be assumed for any of the dialects, or even for PIE. In view of our lack of information on the early location of Armenian, we cannot account for its OV structure, whether it was maintained from that of the parent language or reintroduced. The Anatolian languages must have continued the OV structure of the parent language.

PIE itself must have been subjected to various influences. As a result its stages cannot be sketched with certainty. The common trend towards VO patterning, apparent even in Anatolian, indicates however that late PIE must have been changing from an OV structure. To what extent its earlier structure was consistently OV cannot be ascertained, at least for the present. Future possible contributions to the understanding of PIE syntactic developments will depend largely on insights into the processes of syntactic change. Improved understanding may also result from the study of its neighboring languages. Archeological investigations have provided a good measure of certainty on the location of the parent language around 3000 B.C. through identification of the community of PIE speakers with the Kurgan culture of the area north of the Black Sea. Further investigations are beginning to provide information on contacts between this community and other communities. Metallurgical data are giving information both on the spread of the Indo-Europeans and on sources of their innovations.

Gimbutas concludes that the new technique of alloying copper with arsenic or tin was devised in “Transcaucasia during the 5th millennium B.C.” and “transmitted to Europe by the ‘Kurgan’ (Indo-European) people at the end of the 4th millennium B.C.” Presumably the Indo-Europeans were in contact with the “proto-Caucasian (Kartvelian) family” at this time (Gimbutas 1973:174, 207). The contacts may permit inferences on early syntactic structure, as they have on phonological structure (Gamkrelidze 1966).

Whether or not such inferences are possible, a treatment of PIE syntax must be based on the assumption that the language reconstructed was in use at least as early as 3000 B.C. This assumption alone requires revisions of earlier statements of PIE syntax. It also readily permits an understanding of the divergent systems of the dialects, from the Anatolian to those of Europe. Since many of the conclusions concerning PIE are based on recently provided information and recent developments in syntactic and typological theory, a treatment of PIE syntax can only be preliminary at this time. It should however indicate some of the problems and opportunities for further research, while providing a general statement on the syntactic structure of the parent language of a language family for which we have one of the longest and most comprehensive amounts of data.