The University of Texas at Austin; College of Liberal Arts
Hans C. Boas, Director :: PCL 5.556, 1 University Station S5490 :: Austin, TX 78712 :: 512-471-4566
LRC Links: Home | About | Books Online | EIEOL | IE Doc. Center | IE Lexicon | IE Maps | IE Texts | Pub. Indices | SiteMap

Proto-Indo-European Syntax

Winfred P. Lehmann

5. Syntactic Categories of PIE

5.1. Means of Expression for Syntactic Categories.

Syntactic categories may be expressed by any of the syntactic devices: arrangement, selection, intonation, or modification. They may also be expressed lexically; in early PIE, for example, the root ʔed- ‘eat’ expresses durative aspect, while the root peγ- > pō- ‘take a drink, swallow’ expresses nondurative aspect. The previous three chapters have dealt with the PIE sentence, its basic elements and their modifiers, noting especially syntactic features and their means of expression through patterns of selection, with some reference to arrangement and intonation. The term feature is used here to refer to universal syntactic phenomena, such as interrogation and negation. Classes representing such features in a given language are referred to as categories. This chapter examines the principal syntactic categories of PIE. It will accordingly deal further with some of the topics discussed in the three preceding chapters. Moreover, since the means of expression for syntactic categories may change in the history of a language, an account of the syntactic categories of PIE involves consideration of such means of expression in various stages of PIE and its dialects. The changes in syntax will however be sketched more explicitly in Chapter 7.

The expression of syntactic categories by means of modificational devices has been thoroughly studied in Indo-European, particularly the ablaut variations and how they came to be utilized to express verbal aspects (Saussure 1879; Hirt 1921-1937:II; Lehmann 1952; Kurylowicz 1968). Among the most distinct uses of ablaut is that of zero grade to express nondurative aspect, as in wid- ‘recognize’ as opposed to weyd- ‘see’. An-other distinct use is that of the o grade to express stative, resultative aspect, as in woid- ‘know < have seen’. Yet not all instances of these modifications are used to express these same syntactic categories, no more than a given ending, such as the verbal ending -te, is always used to express one and the same syntactic category. For example, the o grade in roots of causatives made with -eyo- suffixes does not express resultative aspect. Nor does the zero grade in the plural of perfects express nondurative aspect, or even stative aspect. In much the same way other means of expression for a given syntactic category are not necessarily used solely to indicate that category. Accordingly in the study of syntax one must start from an examination of the categories rather than from the surface expressions.

Yet the examination of syntactic categories cannot be limited to those which are expressed by inflection, however central this device may be. For, as Wackernagel indicated with reference to Brugmann, some categories, such as tense, could be expressed lexically in early Greek and Vedic, and presumably in PIE as well (Wackernagel 1926:47, 158-159; see also Delbrück 1897:265-268). To understand PIE syntax and its development we must be prepared for dealing with the means of expression for syntactic categories which may be vestiges of an earlier system and also with those which, like ablaut, have come to be central.

Comparison with selected OV languages may illustrate further possibilities of expression of syntactic features. In Lisu “the various logical predicates are all expressed by verbs in the surface structure,” as in the following example (Hope 1972:147-148):

1. ...dzwa kə̪le tyə̪ nyi̪ -a happen cause try -Dec.
  ‘try to cause it to happen that x helps’

Negation is also expressed lexically in Lisu; the negative is placed “in front of the leftmost verb” (Hope 1972:155-156), as in:

2. ása tə̌ yæ̪̀ ye tyǎ
  Asa not run descend go Cont.
  ‘Asa is not in the process of running away downhill.’

The lexical expression for ‘not’ is , except in imperative sentences, which have thà. Aspect markers, e.g., tyǎ in Example 2, are also lexical; unlike the negative, they are not moved in front of the verb. The marker for the declarative, -a̪ in Example 1, is however a bound element, as are the markers for interrogative and imperative; Hope labels these markers performatives (1972:158). In contrast with Lisu, the OV languages Japanese and Turkish express the negative by means of an inflectional suffix, a device used also for other verbal categories in these languages (Lehmann 1973a). Accordingly we cannot expect to find the same means of expression for syntactic categories from language to language, nor even in typologically similar languages. PIE resembles Lisu in having lexical elements to express negation and in having these placed before verbs. On the other hand it resembles Turkish in having the feature of reflexivity expressed by means of an inflectional suffix. In order to survey the syntactic categories of PIE systematically, the verbal features will be discussed below in the sequence of primary categories, followed by voice, mood, aspect-tense, and thereupon the deictic categories. Syntactic features associated with substantives will then be treated. When features and their means of expression have been discussed extensively above, corresponding syntactic categories will be treated only briefly here.

5.2. Syntactic Categories Expressed in Conjunction with Verbs.

The primary syntactic features associated with verbs are declarative, interrogative, and negative. Sentences with verbs expressing these features may be statements or questions, both positive and negative; when marked any of these may be exclamations.

Relationships between verbs and nominal elements may also be expressed by means of verbal categories. Characteristic relationships with the subject are expressed through reflexive or reciprocal features. Relationships with objects, such as transitivity, were lexically expressed in PIE, and causative relationships were indicated through various affixes. Categories that express attitudes by means of verbs are referred to as moods. For PIE the characteristic modal categories were the imperative, the subjunctive, and the optative. A prominent derivational suffix also indicated the desiderative. And in late PIE the affix used earlier to mark +Dec. came to mark the indicative as opposed to other moods.

Since verbs are the elements which are generally used to express action (or -action, i.e., state) and process, additional features are commonly associated with them, having to do with degree of completion of the action or state. Categories expressing these features are known as aspects or, when measured by time, as tenses. PIE and the early dialects were marked for aspect.

In the generation of sentences, features of tense are introduced through nominal elements; the categories of inflection for tense were, however, expressed with verbs in PIE. By contrast, other verbal features came to be expressed nominally, notably reflexivity. Some Indo-Europeanists have also assumed the ergative relationship between nouns and verbs in early Indo-European (Vaillant 1936; cf. Lehmann 1958:190), but evidence for that assumption is inadequate. Apparently then the relationship between the kind of action expressed by the verb, or its voice, and the subject was not marked by nominal inflections.

From late PIE the categories of person and number were expressed with verbs, and accordingly Indo-Europeanists often assume these categories to be characteristic or even necessary for verbal forms. Yet in many languages they may not be expressed at all or may be expressed lexically, even with nouns. Thus in Japanese the category of person is indicated by means of honorifics, that is, by lexical devices; lexical devices also indicate number. From internal evidence we assume that in early PIE person and number were also expressed lexically, or not at all.

These categories, as well as the nominal categories of case and gender, will be examined in detail below, as well as their roles in PIE syntax.

Syntacticians have dealt with these categories and their means of expression according to various points of view, some emphasizing to a greater extent surface forms, for example Brugmann (1897-1916, 1925) and Eugene A. Nida (1949); others emphasizing to a greater extent underlying forms, for example Jespersen (1924) and Wackernagel (1926-1928). Whatever their emphasis, they have noted that direct correlations between surface forms and underlying features are only partial, as is strict delimitation of the various categories. The treatment of English passive constructions in transformational grammar provides a recent illustration. The earliest treatments were based largely on surface phenomena (Chomsky 1957). Subsequent treatments have concentrated more heavily on underlying phenomena (Chomsky 1965 and others). This shift in approach of the interpretation of a relatively simple syntactic construction which was made in a short time even under a severely rigorous methodology may serve to illustrate that any systematic presentation is based to some extent on arbitrary decisions, since the relationship between surface expression as represented through syntactic categories and underlying phenomena viewed through features is not direct, and since that relationship varies from language to language as well as from one stage of a language to another.

5.2.1. The Primary Syntactic Categories Associated with Verbs.

In PIE the declarative feature was expressed by the affix -i, in contrast with a zero affix. The contrast is still evident in Vedic texts which have maintained injunctive forms. As Hoffmann has pointed out (1967:278-279, 1970), the injunctive serves to mention an action (Erwähnung), while the indicative, which differed only by the addition of -i, indicates a report (Bericht). At any early stage of PIE, -i was accordingly comparable to declarative affixes in Squamish and in Lisu (§ 4.1.4).

By the time of the early dialects the verb forms marked with -i, that is, the indicative, indicated reality or neutrality with regard to mood (Wackernagel 1926:224). The function of -i had accordingly been modified, as is also evident from the abandonment of the use of -i, or primary endings, in the subjunctive. As subjunctive forms came to be characteristic of subordination, they had secondary rather than primary endings (Brugmann 1904a:552-554). And since the use of the indicative to indicate reality was confined to the present system, even the indicative inflection cannot be defined as representing the declarative feature. Rather, the declarative was expressed by other devices.

In late PIE and the early dialects this device was intonation. For expressing -Dec., particles were used, such as án and ke in Greek (Wackernagel 1926:224, Brugmann 1904a:615, 620; 1925:86).

Marking of -Dec. is found in conjunction with the primary category of interrogation. For example, cognates of an were used to indicate interrogation in Italic and Germanic. Generally, in Gothic, an is used in conjunction with interrogative markers like hwas ‘who’, but, as in the following passage, an may be found without such markers (Streitberg 1920:219).

3. John 18:37. an nuh þiudans is þu
? indeed king are you
‘Are you a king then?’

By the time of the dialects, then, specific lexical devices were used for expressing -Dec., and these were also used to indicate interrogation.

Clarification questions were marked lexically in PIE by forms of *kwo-, kwi-, kwu- (Brugmann 1904a:402; 1925:193, 224) and by other interrogatives. It is unclear whether at an early period these were accompanied by a specific interrogative particle placed at the ends of sentences, such as ka in Japanese, and whether confirmation questions as well were so marked. The use of final interrogative particles is attested in late dialects, especially Latin and Slavic, though it is sporadic. We may most safely assume that confirmation questions were first marked in PIE by means of a special interrogation pattern and only later by presuppositional particles. This assumption is supported by the use of the pluti pattern in some Vedic questions (Brugmann 1925:221) and by various particles in the several dialects. Questions presupposing a “yes” answer generally have a negative particle, na in Indic, ou in Greek, nonne in Latin, ni in Gothic, and so on; questions presupposing a “no” answer have a variety of particles, mḗ in Greek, num in Latin, ibai in Gothic (§ 4.3.2; Brugmann 1925:222-224). Particles, then, as Brugmann pointed out (1925: 224), came to be highly prominent in expressing interrogation.

To express negation, lexical devices were used. Before the development of inflectional forms to indicate necessitative and obligative features, and while injunctive forms were still in use, a distinct negative, PIE , was used to express the negative of these features; *ne was used elsewhere (Brugmann 1904a:582, 612). The PIE contrast between and ne was maintained in the most archaic dialects, in some with changed surface markers (§ 4.3.3). These dialects were Hittite, Indic, Greek, and another which retained the OV pattern, Armenian.

But even in Vedic and Greek the PIE contrast was modified. In Vedic, na was used with the subjunctive even when it had necessitative force, thus assuming one of the uses of PIE . In Greek, on the other hand, was maintained in sentences having subjunctive verb forms with necessitative meaning, though not with subjunctives expressing likelihood (Brugmann 1904a:581-582). The distinction between negative modal statements and simple negative sentences was accordingly no longer expressed as it had been in PIE, even though reflexes of the negative particles were maintained. In some dialects, as in Latin with contrasting with ne, differing negative particles had to do with emphasis.

Emphatic, or marked, variants of declarative, interrogative, and negative features were characterized lexically by means of particles or by marked word order (see Brugmann 1904a:614-622 for a list of such particles, and also Delbrück 1897:497-511). Initial position, especially of verbs, conveyed emphasis for the element in question, as in many examples.

4. RV 1.10.1. gā́yanti tvā gāyatríṇó ’rcanty arkám arkíṇaḥ
they-sing you singers they-chant song praisers
‘Celebrating you are the singers; singing a song of praise, the praisers.’

Since interrogative particles and pronouns are often emphatic, they generally stand in initial position in the clause, as in Examples 52 and 55 of Chapter 4; similarly, negation can be emphasized by means of initial negative particles, as in Example 170 of Chapter 4.

Although declaratives, interrogatives, and negatives may in this way be emphatic, emphatic sentences are particularly associated with exclamations and commands, as in Brugmann's example (1904a:680).

5. RV 1.14.12.
yukṣvā́ hy áruṣī ráthe haríto deva rohítaḥ
you-yoke now ruddy to-chariot yellow O-god red-horse
tā́bhir devā́n̐ ihā́ vaha
with-them gods here you-bring
‘Yoke the ruddy, tawny, red horses to the chariot, god!
With them bring hither the gods!’

The frequent use of emphatic patterns for commands led Brugmann to classify sentences into three principal types; exclamatory, declarative, and interrogative (1904a:647-649) ; this classification, according to Brugmann, was based on fundamental psychological functions. The identification of exclamatory sentences as a separate class has the advantage of accounting for abbreviated utterances, as well as vocatives and interjections (1904a:687-697). Yet, since all kinds of sentences, declarative and interrogative as well as imperative, volitional, and so on, can be emphatic, the analysis of sentences into marked and unmarked types provides a more satisfactory classification than does that of Brugmann, removing at the same time the need for the exclamatory type.

Summing up the syntactic devices for expressing the primary syntactic categories in late PIE, we note that these devices were for the most part intonation and lexical. The extensive system of verbal inflection was used to express other verbal categories.

5.2.2. Voice and Its Means of Expression.

Relationships between verbs and nouns were expressed primarily through case forms of nouns. Accordingly early PIE did not have inherently transitive or intransitive verbs. The lack of such a distinction is reflected in dialects as late as Latin, in which the accusative can be used to indicate the goal of the action with “intransitive” verbs (Hale and Buck 1903:203, 206-207):

6. Vergil, Aeneid 1.2. Ītaliam vēnit
‘[to] Italy he-came’
7. Cicero, Catiline 4.1.2. multa tacuī
‘[with regard to] many things I-have-been-silent’

In the early dialects specific case forms came to be associated with specific verbs, such as the accusative for direct objects. From this time verbs are classified as transitive or intransitive; the classification, though in part related to derivational affixes, is for the most part lexical.

The distinction between + and -transitive verbs, however, came to be associated to an increasing extent with inflectional classes. Since the perfect inflection indicated state, accusative nominal forms were less commonly associated with it than with verb forms of the m class. Moreover, accusatives were generally not used with middles. Both of these inflections are comparable in this way to intransitives; this characteristic has led to the assumption of a relationship between the perfect and the middle. As the statements presented above indicate, however, the relationship is only secondary: the middle, with its basic meaning of reflexivization and reciprocality, is comparable with the resultative perfect primarily in lacking an accusative object. Since however transitivity was not a category of the PIE verbal system, the middle and perfect should not be related to one another but treated as categories independent of one another.

Some prominent derivational affixes, notably -yo- and -ē- < -eʔ-, expressed state and accordingly were comparable to intransitives (Brugmann 1904a:492, 527; Delbrück 1897:26-40, esp. 34-35). Yet it is instructive to note that, in keeping with the autonomy of individual words and with the patterning of the case system, verbs with these suffixes may be accompanied by nouns in the accusative. Delbrück gives excellent examples, such as the following for the Sanskrit root as- + -ya- ‘hurl, throw’:

8. RV 2.24.8. tásya sādhvī́r íṣavo yā́bhir ásyati
his accurate arrows with-which he-shoots
‘Accurate are his arrows, with which he shoots.’

9. RV 10.168.1. utó eti pṛthivyā́ reṇúm ásyan
and it-goes on-earth dust throwing
‘And [the wind] goes along the earth, throwing up dust.’

The essential of such -yo- verbs is the expression of a state or process, whether they are intransitive and accompanied by noun in an adverbial sense, as in Example 8, or transitive and accompanied by a noun that can be regarded as the object, as in Example 9. With these possibilities, both affixes were employed in passive formations, after the passive was introduced as a verbal category: the -yo- especially in Indic, the -ē- in Greek. The passive as a separate category developed, however, primarily from the middle.

As we have noted above, § 4.3.4, the middle was a distinct voice category in PIE. It has at least the following meanings. Some roots always had the middle affix, and in these the action or process was viewed as taking place with reference to the subject; see Delbrück (1897:417-425) for examples of such roots. The middle affix was also used in contrast with the active to convey reflexive or reciprocal meaning, as in Examples 76 and 77 of Chapter 4. When late PIE began to assume VO characteristics, pronominal elements were introduced for reflexive and reciprocal meaning. The central meaning of the middle became that of passive voice.

A passive proper requires the presence of an agent, whether animate or inanimate (Gonda 1951a:73–78). In our early texts we find examples of middle verbs used with agents and thus having passive force (Delbrück 1897:432-435), such as the following with the verb epeígomai ‘set oneself in motion, hasten, be beset’:

10. Iliad 6.363. allá g’ órnuthi toȗton, epeigésthō kài autós
but you Ptc. rouse him he-should-hurry Ptc. and self
‘But you stir him up, and he himself should make haste.’

11. Iliad 5.622. epeígeto gàr beléessi
he-hastened, he-was-beset for with-missiles
‘For he was beset with missiles.’

In Example 11 the inclusion of an instrument gives a passive meaning to a verb which when not accompanied by a noun indicating an agent or instrument is simply a middle. From such uses the passive arose as a separate category.

A passive is not likely to arise as a separate category, however, except when a language includes causative constructions; for these require the presence of an agent or instrument as well as a direct object. Parallelism between the passive and increasing frequency of causatives is evident in our Sanskrit texts (Whitney, JAOS 13 : xxxiv). The number of roots found only in the oldest texts with the passive is less than half that of roots found with the passive in both earlier and later texts (37 vs. 105); and this number is lower than that of roots with the passive found only in the later, epic and classical language (117). The proportions between attested causative forms are comparable: 111 vs. 247 vs. 207. The increase in the number of passives made from causatives is even more striking: 9 vs. 28 vs. 110. These figures support the assumption that the passive developed only after PIE, in the various dialects.

This assumption is also supported by the diversity of passive formations, such as: the middle forms in Gothic and Greek, supplemented in Greek with -ē- and -thē- forms; the -yo- forms with middle endings in Sanskrit; the r forms in Italic and Celtic (Leumann 1963:306-308). Such forms are concerns of studies dealing with the individual dialects. Here we have been examining the evidence for voice categories in PIE. On the basis of verbal inflection we may assume only a middle and an active. Derivational affixes could be used to indicate other meanings comparable to voices, such as the causative and the factitive (§ 4.4.3). These, like the middle, underwent various modifications in the individual dialects, leading to different alignments of categories for voice.

5.2.3. Moods and Their Means of Expression.

The inflectional suffixes which were most prominent in PIE indicated moods: -e/o- with necessitative and obligative force, -yeʔ- with voluntative (§ 4.3.5-4.3.6). The dubitative and voluntative meanings of these suffixes became highly frequent, and from these uses the two moods were named subjunctive and optative. The subjunctive and optative were employed particularly in subordinate clauses, in which their meanings were determined in part by the relationship of the subordinate clause to the principal clause, a relationship which was also indicated by introductory particles. In view of the effects of these interrelationships and of particles such as conjunctions, the uses of modal forms in the dialects provide some of the greatest complexities in Indo-European grammar (see such specialized monographs as Goodwin 1893, Gonda 1956, and the bibliographical survey of Calboli 1966-1968). We have assumed however that the subjunctive in PIE had basically an obligative meaning, the optative a volitional, and that these meanings were modified as the individual dialects developed.

The uses of the optative and subjunctive were associated with that of a third mood, the imperative, marked by distinctive endings and used to express commands. The difficulties involved in attempting to determine the uses of these three moods in PIE are compounded by the presence of derivational affixes and by syncretism of the moods among one another and with derivational forms.

The most notable derivational affix with modal meaning was -s-, which expressed desiderative force (Brugmann 1904a:518-519, 529). In Indic, forms with -s- assumed the status of a modal category, although desiderative forms never were very frequent. In other dialects which maintained -s- forms, these fell together with the tense system, expressing the future, or with the subjunctive (Delbrück 1897:242-255, 306-308, 321-330).

In Sanskrit the subjunctive merged with the imperative, so that this mood had first-person forms (see Whitney 1896:215-220 on this merger and the uses of the modes in general). Hittite underwent a similar merger, though here the first singular -l- forms were based on a derivational affix with “imperative” force (Solta 1970:44-84). These mergers extended the meanings of the modal forms which survived in the various dialects, whether the so-called subjunctive based on an IE optative in Germanic or the so-called imperative in Hittite; and because of such broadening of meaning, it is difficult to be certain of the meanings of the moods in PIE.

Besides inflectional and derivational affixes, modal meanings were also expressed lexically. In some dialects such lexical expression serves primarily to support the meaning of a modal form, as does in the following (see Grassmann 1872:1664-1665 and also Denniston 1966b:xxxvii).

12. RV 5.4.6. vadhéna dásyum prá cātáyasva
with-weapon dasyu away indeed you-repel
‘Drive away the Dasyu with your weapon.’

In Hittite such particles have independent meaning, as does man for the expression of potentiality (J. Friedrich 1960:140):

13. man- u̯ar-aš-mu MUTI.I̯A kišari
  ‘may’ quote-he-me spouse he-is
  ‘He could become my spouse.’

From such modal uses of particles in independent sentences we may assume lexical indications of mood in PIE.

Such an assumption is supported by the use of reflexes of *, especially in Vedic. For with injunctive forms is the sole indication of modal meanings, chiefly necessitative but also voluntative (Hoffmann 1967:46-92, 101-106):

14. RV 8.48.14. mā́ no nidrā́ īśata, mótá jálpiḥ
may not us sleep it-overpowers may-not-and chatter
‘May sleep not overpower us, nor chatter.’

The first clause illustrates a necessitative use of mā́ with the modally unmarked injunctive. The second, verbless clause suggests that alone could have negative modal force; here however one may assume conjunction reduction of the second use of īśata, though Hoffmann gives other examples of without a verb (1967:65). Although no modal particle with positive meaning is as prominent in the early texts as are the reflexes of PIE mḗ, such examples as Gk. án and ke support the assumption of lexical as well as inflectional indication of modal meaning in PIE (Wackernagel 1926:224). The primary indicators of mood in PIE however were the inflectional categories known as the subjunctive, optative, and imperative.

5.2.4. Aspects and Their Means of Expression.

As noted in § 4.4, the primary contrast in aspect was one of imperfective versus perfective. In early PIE the distinction was expressed through the contrasting endings; m s t versus x the ø. Another aspectual contrast, that between momentary and durative, was expressed primarily through forms of the root: durative by an accented and accordingly e vowel in the root; momentary by lack of principal accent on the root and accordingly ø vowel, or o vowel when the root received secondary accent on shift of the principal accent. Through subsequent syncretism and shift from aspect to tense, this early aspect was disrupted.

Contrasts between the aspects in the early system are clearest in Vedic injunctives, as opposed to aorists and perfects of a given root or of an extended root like *wey-d- < *wey- + -d-. The simple injunctive form vet < *weyd-t expresses imperfective, durative aspect; its single attestation, RV 10.53.9, may be translated ‘he understands’. The aorist vidát < *wid-ét expresses perfective, +momentary aspect; as in RV 1.62.3, it may be translated ‘he found’. The perfect véda < *wòid-e expresses perfective, -momentary aspect; as in RV 10.169.2, it may be translated ‘he knows < he has seen and accordingly he understands’. These contrasts may be illustrated more precisely by citing the sentences in which these forms occur:

15. RV 10.53.9. tváṣṭā māyā́ ved
Tvaṣṭr changes understands
‘Tvastr understands the changes.’

16. RV 1.62.3.
índrasyā́n̄girasāṃ ceṣṭhaú
of-Indra-of-Angiras and-at-request
vidát sarámā tánayāya dhāsím
he-found Sarama for-descendants nourishment
‘At the request of Indra and the Angiras
Sarama found nourishment for their descendants.’

The perfective, momentary meaning is equally implicit if one follows the interpretation of Geldner (1951-1957:I, 80): ‘Als Indra und die Angiras' auf der Suche waren, fand...’ The stative, resultative meaning of the perfect form véda is pointed up by the instrumental íṣṭyā in RV 10.169.2, for this indicates that ‘finding’ was made at a definite time in the past, leading to knowledge:

17. RV 10.169.2. yā́sām agnír íṣṭyā nā́māni véda
whose Agni through-sacrifice names he-knows
  ‘Whose names Agni has found out (and thus knows) through the sacrifice.’

By contrast the injunctive form simply indicates a situation; in Hoffmann's term it has a “memorative” meaning, in T. Elizarenkova's a “mentioning,” an interpretation like Hoffmann's Erwähnung (Hoffmann 1967: 279, Elizarenkova 1972:250-253, with reference to her earlier work of 1960). And the aorist form in RV 1.62.3 refers to a single action that has been completed.

While these distinctions were expressed unambiguously in the distinct, contrasting forms of the singular, the dual and plural forms were not unambiguous; they had neither contrasting endings throughout for each of these aspectual categories nor characteristic root forms. The zero grade of the root was found in the dual and plural of the perfect as well as in the aorist, and in athematic presents as well, a vocalism applying also to injunctives. Accordingly the contrasts between ±perfective and ±momentary were not marked throughout a paradigm (present versus aorist/perfect; aorist versus perfect). The lack of contrast in many forms of the dual and plural numbers resulted in syncretism also in the singular, as is evident from the reflexes of these aspectual forms in the dialects. In some dialects the distinction between the +perfective and the -perfective became predominant, as in Slavic, which has a verbal system with principal distinction between these two verbal aspects. In other dialects, as in Sanskrit and Greek, the contrast between -momentary (durative) and +momentary became predominant, leading to a contrast between ongoing action, or present tense, on the one hand, and punctual and stative action, or aorist and perfect tense, on the other hand. Whichever features predominated, all dialects showed syncretism; and in most dialects the aspectual system gave way to a tense system.

Other aspects, such as iterative, intensive, frequentative, and inchoative, never occupied as prominent positions in the PIE verb systems as had ±perfective and ±momentary. These other aspects were expressed by verbal affixes, some of which became productive in specific dialects but are found only sporadically in others. An iterative with -ske- affix is especially prominent in Hittite (J. Friedrich 1960:140-141), where it has durative meaning and indicates action carried out in several steps, as in:

18. anniškimi kuin
  ‘which I carry out (in several ritual acts)’

This example and others given by Friedrich illustrate that even in Hittite the -ske- affix had various functions. We may make the same assumption for PIE, although the iterative meaning may have predominated in sk- forms.

By Dressler's view, the intensive, distributive, and durative developed from the iterative (1968:233). In accordance with this view the various meanings exhibited by sk- formations in the several dialects had their origin in the iterative, such as the inchoatives of Latin, which presumably arose from distributive uses like that in the Hittite example above (18) (see also Brugmann 1913:350-361; Delbrück 1897:59-64). And the causative formation in Tocharian marked by this affix was attributed by Holger Pedersen to an iterative developing through a factitive (1941:168).

Frequentative, iterative, and also intensive aspect was expressed by reduplication, a process which came to be productive especially in Sanskrit (Delbrück 1897:16-26), as in jigā́ti, Gk. bíbāmi ‘walk, stride’, as opposed to the simple root in the aorists ágāt, ébē ‘put one's foot down’. Aspectual categories may then be regarded as compounds of features.

Such a view of aspects as composites rather than single semantic features has been explored by H. J. Verkuyl (1972), as well as the interrelationships of aspects with elements which, by our model of language, are introduced through the nominal constituents of the proposition (see also Anderson 1971). For Verkuyl, durative aspect is composed of “elementary semantic verbal” features, such as “MOVEMENT, AGENTIVE, PERFORM, TAKE, ADD” (1972:96); moreover, it also involves a “nominal node ... containing quantificational information.” Durative aspect then allows the introduction of “durative adverbials.” And a verb with durative aspect is associated with the feature “unspecified quantity,” while a nondurative is associated with the feature “specified quantity.” We may illustrate something of the “compositional nature of the aspects” by comparing examples of the durative verb ‘eat’ and the nondurative ‘bite’.

19. The dogs were happily eating their biscuits.

20. The dogs quickly bit three of the chilren.

As in the numerous examples given by Verkuyl, the perfective verb of Example 20 is associated with a specified quantity, unlike the durative “were eating” of Example 19. Moreover, if perfective meaning is included in a sentence including ‘eat’, as through an adverbial element like ‘quickly’, the simple past is used:

21. The dogs quickly ate their biscuits.

By contrast, the past progressive with ‘quickly’ implies repeated, iterative action, as in:

22. The dogs were quickly eating their biscuits.

These examples illustrate that aspect, though conveyed lexically in English, involves contrasts between manners of action which can be readily introduced; we may make a similar assumption for PIE, though expression of such contrasts in PIE was primarily suffixal. These English examples also illustrate the close relationship between aspect and tense; as Verkuyl cautiously puts it, “the underlying V-node involved in the composition of the Nondurative Aspect has something to do with what I have called ‘the temporalization of abstract entities’” (1972:176). Aspects, then, as Dressler and Verkuyl have demonstrated, cannot be viewed as simple features associated only with the verb; they extend to other elements of the proposition as well and include temporal features as well as features pertaining to manner of action. Because of these relationships to all elements of the proposition, and because of their various means of expression, they are best treated as Q constituents.

In the course of syntactic change a given feature may come to predominate, somewhat as a given phonological feature may change in sound. In late PIE, features of tense became predominant: present tense rather than durative aspect, preterite rather than nondurative. The aspectual meanings thereupon were expressed lexically or by derivational processes. Forms in which the shift from a predominant aspectual to a tense meaning was not carried out provide excellent evidence for the development. Among such forms are the Germanic preterite-presents. In preterite-presents, like Gothic kann ‘I am able to do < I have acquired knowledge’, the shift was carried out lexically. Various reflexes of the perfect of *weyd- provide illustrations in a number of dialects. Vedic véda, Gk. oȋda, Goth. wait ‘I know’ reflect the PIE resultative, -momentary meaning lexically; Latin vīdi ‘I have seen’ on the other hand illustrates the normal shift to a “preterite” tense, a shift also exemplified in Armenian egit, which corresponds to Skt. ávidat ‘he found’. Thus, both of the PIE perfective aspect forms, the aorist and the perfect, were shifted to preterite-tense forms as opposed to present-tense forms, which normally had their origin in imperfectives.

In some dialects, notably Sanskrit, Greek, Italic, and Baltic, the tense system was amplified by a future, which developed from forms indicating doubt, such as s- suffixed forms and subjunctives. PIE itself had no future tense, in contrast with the views of some scholars (see Brugmann's interesting observations [1917] on the lack of a word for ‘tomorrow’ in PIE, though words for ‘today’ and ‘yesterday’ can be reconstructed; see also Gonda 1971:92-93). Reference to time of the action was indicated by means of adverbial or nominal elements; the verbal forms associated with these elements were aspectual. Verbal affixes and endings accordingly indicated aspects in PIE; but the meanings of individual forms were determined in part by lexical features, as we will note in Chapter 6, and also by the case forms of substantives in the proposition.

5.3. Syntactic Categories Expressed in Conjunction with Substantives.

In late PIE, inflection of nouns, pronouns, and adjectives, that is, substantival inflection, was made for case, gender, and number. The primary uses of gender and number were for indication of congruence, gender in nominal constructions, number in nominal and verbal constructions. Number also carried lexical meaning, both with verbs and with substantives. Case was particularly important in specifying the lexical meaning of nominal elements and of verbal elements associated with them, though with some verbs specific case forms were regulated by patterns of government (Meillet 1937:338-341). Yet the central uses of cases marked nominal forms for specific functions in sentences, as the selected examples in the following section illustrate.

5.3.1. Uses of the Case Categories.

With its large number of inflections, the PIE substantive had overt markings for many of the features introduced in the nominal constituent of the proposition.

Of the eight cases, only one, the genitive, served primarily to indicate relationships between substantives. Another, the vocative, marked nouns as being independent of other elements. In this way the vocative preserved longest the earlier situation in which nominal forms with characteristic endings had a self-contained meaning, independent of other elements in the sentence (Meillet 1937:151). It is the case of address, with both independent and included nouns. When independent the vocative was always accented; when included in a clause, it was not. Accentuation however did not affect its meaning.

The six remaining cases expressed meanings largely with regard to the verb. The accusative indicated the target of the verbal action; the dative indicated the receptor. When the agent of the action was animate, it was indicated by the nominative; when inanimate, by the instrumental. As passive forms were introduced, the instrumental was also used for animate agents. Closely related to its use for the instrument of the action is that for accompaniment; accordingly the instrumental is also referred to as the sociative. To indicate a point in time or place, the locative was used; to indicate contrast with an element or departure from a point in time or place, the ablative. These six cases correspond closely with underlying syntactic categories referred to as target, receptor, agent, instrument, place, time, source. Yet they also had additional uses, as described in the handbooks (Delbrück 1893:173-400; Meillet 1937:341-349; and many others, such as Krahe 1972:56-108). Moreover, their uses with specific verbs which draw on their central meanings are illustrated in Chapter 2.

If these additional uses are treated independently of their contexts, or if a case is analyzed from an item approach rather than a process approach, each case will be described as having a multitude of meanings. When, for example, the genitive was analyzed in this way, it was assumed to have more than a score of uses. A study by Benveniste, however, indicates both how a case is to be described in a syntactic discussion and how the genitive can be characterized in a unified way (1966:140-148).

Starting from an article on the Latin genitive by A. W. de Groot, who proposed “eight regular grammatical uses” after a review of the “thirty or so distinct uses” recorded in earlier classifications, Benveniste defines the genitive as follows: “The genitive is the case that in a construction consisting of two nouns takes on the function which either the nominative or the accusative has in a sentence with a finite verb; all other uses... derive from this use” (1966:148; 1971:127). If the verb is intransitive, the noun accompanying it must be in the nominative; and if the verbal construction is changed to a nominal construction, the nominative becomes the so-called subjective genitive, as in patientia animi ‘patience of spirit’ from animus patitur ‘the spirit is patient’. If the finite verb is transitive, the result of a modification of the accusative is the so-called objective genitive, as in patientia doloris ‘endurance of suffering’ from pati dolorem ‘to endure suffering’ (see also Kurylowicz 1964:186-189). In his discussion Benveniste also indicates how the other uses of the genitive in Latin are related to its basic function and how they are to be treated under specific syntactic constructions; thus the “genitive of locality” is to be treated in connection with toponyms, for it occurs only with these. And the “genitive of quality” is a “syntactic derivation...of the normal use of the genitive,” as in aedes regis ‘palace of the king’, with a following infinitive (Benveniste 1971:125):

23. pauperis est numerāre pecus
  of-the-poor-man it-is to-count cattle
  ‘It is a characteristic of the poor man to count cattle.’

Benveniste's conclusions are accordingly based on a generative syntactic approach, as are also those of Jerzy Kurylowicz. These conclusions have important theoretical implications for the view of grammar maintained here.

The characterization of the genitive as derived from either the nominative or the accusative indicates that the genitive occupies a particular hierarchy in the system of cases and in this way demonstrates that there is indeed such a hierarchy. In this hierarchy the underlying case categories differ from verbal categories and especially from those introduced under Q. For the genitive case is subordinate to other cases, and these in turn are subordinate to the verbal element with its categories. Only the vocative, which is independent, is an exception. The position of the genitive, then, supports the analysis of the proposition into a verb and abstract case categories, as well as the analysis of the sentence into a Q and a Prop. constituent. Moreover, this approach clarifies the positing of a central meaning for a case.

Benveniste's conclusions on the “meaning” of the genitive are much like those of earlier Indo-Europeanists who have proposed “basic meanings” (Grundbegriffe) for the cases (Delbrück 1893:185-187, 333-335, with references to Grimm and Gaedicke; Sommer 1959:10, cited by Krahe 1972:56-57). Thus, for Ferdinand Sommer, “the genitive denotes the sphere of a verbal process or that of a further noun.” With greater use of a generative approach, Benveniste's definition is more precise, though still comparable with that of Sommer. Moreover, in contrast with some earlier Indo-Europeanists, Benveniste is not giving the “basic meaning” of the genitive in a supposedly primeval stage of Indo-European; rather, the meaning proposed by him is that added by the selection of a genitive case form in a particular construction. More elaborate statements of the choice of a particular case in specific constructions, and in connection with specific lexical elements, such as those given at length by Delbrück (1893:308-333), belong in the lexicon rather than the syntactic component of an account of PIE or of any language. The same procedure must be used in the study of other case forms.

Care must also be taken to distinguish possible differences between the basic use of the genitive in PIE and that in a dialect like Latin. In his study of the genitive as examined by other scholars, including Wackernagel, Kurylowicz, and Benveniste, and as occurring in other languages, Calvert Watkins concluded that Roman Jakobson's analysis of the meaning of the case is to be assumed for PIE (Watkins 1967:2191-2198). The primary function of the genitive in PIE then relates to “the extent of participation of the entity in the message”; the transformationally derived subjective and objective genitives are secondary to a “genitive of belonging.” The question is partly one of chronology, partly one of focus. The meaning “extent of participation” is a semantic characterization, in contrast with Benveniste's syntactic definition. We may relate the definitions of Watkins and Benveniste by noting that the genitive, as a nominal form derived from nominatives or accusatives with given verbs, indeed indicates the extent of participation of a given substantive in a verbal action. If one seeks a still earlier use of the nominal forms which are the etyma of the genitive in PIE, one may find it in connection with a definition of the nominal suffix -s in pre-IE; -s was used to specify an individual (Lehmann 1958:190). In late PIE the genitive had the syntactic function indicated by Benveniste. This then was its “basic meaning” for PIE, comparable to the basic meanings given for the vocative and the six other cases at the beginning of this section.

From these basic meanings secondary uses arose in specific contexts, and other means of expression than case suffixes, as we may note briefly. We have proposed in Chapter 2 that a verb is characteristically accompanied by one case form in PIE, in addition to the person expressed in the verbal ending. Meillet characterized the accusative as “determining the meaning of the verb” (1937:343). But as the following examples with Skt. vidh- demonstrate, other cases also function in this way. Accompanied by the accusative, vidh- may be translated ‘honor’

24. RV 8.96.8. śúṣmaṃ ta enā́ havíṣā vidhema
power of-you with-this with-offering we-would-honor
‘We would honor your power with this offering.’

Various other examples could be cited with a similar use of the accusative, including the examples given by Meillet, such as those with the Greek verb bállō ‘throw’:

25. Odyssey 2.80. potì skȇptron bále gaíēi
toward Ptc. staff he-threw to-ground
‘And he threw the staff down to the ground.’

Complementing or determining the meaning of the verbs bállō and vidh-, the accusatives indeed fulfill the function stated by Meillet. But they also specify the element which is the target of the verbal action. This use as target is then the basic meaning of the accusative. It applies also in connection with other verbs, such as verbs of motion like Lat. īre in Rōmam īre ‘go to Rome’. Especially when discussed independently of verbs, such uses are taken as secondary and treated independently, often with separate labels, but, as Meillet indicates, these uses derive from the basic meanings, in association with certain verbs. We may illustrate such developments with other cases as well, as in the dative accompanying vidh-.

Like Example 24, RV 8.48.13 would be complete with the verb form and the dative (te > ta); tásmai is appositional to ta. The vocative Indo is independent.

26. RV 8.48.13. tásmai ta indo havíṣā vidhema
that, such-a you Indu with-oblation we-would-worship
‘We would worship you, such a one, O Indu, with oblation.’

As Meillet has stated (1937:347), the dative does not indicate the target of the action, as would the accusative, but rather the person (or object) in view of which the action was carried out. In Example 26 then vidh- is specified for the receptor of the act of worship. The dative may thus determine the meaning of the verb as does the accusative. Moreover, the dative is characteristically found with certain verbs, like those of ‘hearing’. This use simply supports the proposal that the basic meaning of the dative is as receptor.

Both Example 24 and Example 26 have vidh- with its primary noun accompanied by an instrumental. In these sentences the instrumental has its primary meaning, indicating the instrument of the action (Schwyzer 1943:16). As another example in this poem indicates, this meaning could readily be used in the sense of accompanied by. The minimal sentence of this example states that the speaker would partake of the pressed-out soma, bhakṣ- governing the genitive te, which is accompanied by the appositional sutásya:

27. RV 8.48.7. iṣiréṇa te mánasā sutásya bhakṣīmáhi
devoted you with-mind pressed we-would-partake-of
  ‘With devoted mind we would partake of you pressed-out.’

In this passage the instrumental indicates “accompanying circumstances” (Delbrück 1893:238-240), though its relationship to the basic meaning is transparent. In the following stanza the instrumental suasti, though indicating the instrument of the action, is virtually equivalent to a dative, indicating purpose:

28. RV 8.48.8. sóma rājan mṛḷáyā naḥ suastí
Soma king you-be-gracious to-us through-means-of-well-being
‘King Soma, be gracious to us for our welfare.’

These examples taken from an Archaic hymn may illustrate how the instrumental came to be used to express meanings other than its central meaning.

The ablative likewise is used in accordance with its basic meaning of the starting point of the action. As Gonda points out (1971:119), it is excellently characterized by Panini (1.4.24-31) as the “case of the item which remains in its own place when another item is removed from it.” This meaning is figurative as well as literal; the figurative use is found widely in comparison, with or without a comparative suffix on the adjective:

29. RV 1.57.4. nahí tvád anyó girvaṇo gíraḥ sághat
not-indeed from-you other desiring-praise songs-of-praise can-endure
‘For no one but you can endure the songs of praise, you who desire praise.’

In another figurative use it also expresses the reason for an action (Macdonell 1916:318):

30. RV 7.89.5. mā́ nas tásmād énaso deva rīriśaḥ
not us from-that from-sin O-god you-punish
‘Do not punish us because of that sin, O god.’

These secondary uses of the ablative conform with its limitations of a distinct form only in the singular, notably the comparative, which represents a contrast between two items. Its limitation to the singular may have been largely responsible for its loss, in both its primary and its secondary meanings.

For the other cases, any of these secondary meanings could become more prominent, as study of the dialects indicates. Thus in Lithuanian and the Slavic languages the instrumental is used predicatively as a resultative, as in the example below (see Delbrück 1893:262-268 and Senn 1966:429, who gives examples such as the following):

31. sniẽgas pavir̃sta vándeniu
  snow it-becomes water
  ‘The snow turns to water.’

Such developments of individual case forms would require extensive discussion if the various details were to be presented, as they have been by Delbrück (1893:200-400). Here only the principle behind the developments is presented: when a second or third case form was used in sentences, its meaning was extended beyond the basic meaning, and eventually the secondary meanings might become prominent.

The meaning involved could also be specified by particles, which accompanied case forms in particular passages. An example from the cited Archaic hymn may illustrate such uses of postpositions; in the following example, the locative máda < máde would adequately express the meaning ‘in...’, but the postposition ā́ ‘in’ is included, apparently to specify this meaning more clearly:

32. RV 8.48.6.
áthā te máda ā́ soma mánye
then for your intoxication in Soma I-think-of-myself
revā́m̐ iva
being-rich as
  ‘For then in intoxication with you, Soma, I consider myself rich.’

The development of particles like ā́ as postpositions, subsequently prepositions, has been described as thoroughly as have the uses of the case forms (see for example Delbrück 1893:643-774; Meillet 1937:345-349).

In his extensive treatment Delbrück has pointed out how particles were combined with verbs as well as nouns. Thus ánu in the following line is interpreted as a verbal particle accompanying the root tan- ‘extend’ and the particle ā, with the accusative maintaining its independent meaning:

33. RV 8.48.13. ánu dyā́vāpṛthivī́ ā́ tatantha
over heaven-earth to you-have-extended
‘You have extended yourself over heaven and earth.’

In the following example ánu may be interpreted as a postposition as readily as a verbal particle:

34. RV 10.97.19. víṣṭhitāḥ pṛthivī́m ánu
spread earth over
‘spread throughout the earth’

When such particles came to be placed before nouns, they developed into prepositions, such as those in the later dialects; the cognates of ánu are attested in numerous dialects: Avestan ana, Gk. aná, Lat. an, Goth. ana, Slav. na, Lith. .

35. Odyssey 9.209. hén dépas emplḗsas húdatos anà eíkosi métra kheȗ
one cup having-filled water over twenty measures he-poured
  ‘after filling one cup [with wine] he poured it on twenty measures of water’

In this way particles first supplemented case forms as postpositions, then as prepositions, and eventually in many IE dialects as the sole markers of relationships between verbs and nouns. Such postpositional and prepositional phrases have been equated with inflectional forms representing underlying case categories. The shift from inflected nouns to noun phrases including postpositions or prepositions is thus a matter of surface syntax.

There are various reasons for the shift from an inflectional system marking cases to the use of prepositions; these reasons were phonological and morphological and, not least, syntactic. The phonological reasons are primarily a result of a shift to stress accent, because of which characteristic final endings were weakened and merged or lost. The phonological syncretism was intensified by systemic morphological realignment of the eight-case singular system with the smaller system of the plural and dual. Most influential in the placement of particles in preposed position and accordingly prepositional constructions was the shift to a VO structure; for prepositional phrases are the characteristic nominal constructions to indicate case relationships in VO languages.

Although these changes led to the indication of substantival relationships to verbs largely through prepositional phrases, the indication was marked in part by arrangement, notably of the subject and object. Already in the early dialects the subject or object relationship of duals and of neuter nouns to verbs in the third person was determinable chiefly from their position and clarified by the presence of other elements in the clause which were unambiguous. When first- and second-person forms of the verb merged with third-person forms, the position of a substantive often was the only means of distinguishing its role. Word order then became fixed. Thus in spite of the unambiguous syntactic use of he as opposed to him, one must put the English subject initially in translating the following German sentence:

36. Die Tóchter hat er gesehen.
  He saw the dáughter. NOT The dáughter he saw.

The changing role of case categories thus is related to general syntactic changes, and these also involve the other nominal categories.

5.3.2. Gender.

Although gender distinctions are expressed in the substantives of all Indo-European dialects, it is clear that the gender category was developed late in PIE. The decisive evidence for this conclusion was given by Johannes Schmidt in his study of the plural formations of IE neuters (1889; see also Lehmann 1958; Kurylowicz 1964:207-226). According to Schmidt's study, neuter plurals originated from singular collectives. Since feminine singular collectives had the same forms and the same meanings as did neuter plurals, we must assume that an earlier system of endings did not indicate gender categories but rather specific semantic meanings. These meanings I have characterized as follows: -s = individual, -m = resultative (or result), -h = collective (Lehmann 1958: 192). As the nominal system of late PIE developed, these endings were aligned in the gender, number, and case paradigms found in the various IE dialects.

In examining gender, we must note that in PIE and the dialects its function was the indication of congruence. Gender may also have different uses, as T. F. Mitchell has recently demonstrated in an important paper on concord, especially in Sindhi and Cairene Arabic (1973). Concord in Arabic is determined by number as well as gender; and gender conveys the additional force of “marking the role and status relations of interlocutors,” as Mitchell illustrates. In IE, however, gender is a means to express congruence, primarily a syntactic category to indicate the bounds of the nominal phrase when nouns, adjectives, and adjectival pronouns are involved; gender also permits precision of pronominal reference (see also Brugmann 1891).

Many studies have been carried out in attempts to account for the origin of gender as a category in PIE, generally relating it to sex distinctions. These were based on the earlier active-stative distinction in Proto-Indo-European. The situation is still maintained in Hittite, with its distinction between common and neuter nominal classes. But in the other dialects o-stems came to indicate masculine gender, and ā-stems feminines. The distinction extended to nouns with other stems as well as to adjectives and pronouns, establishing threefold congruence in the nominal system. The gender distinction of the dialects other than Hittite was accordingly based on syntactic purposes.

It is readily understandable that expression for congruence is less important in OV languages than in VO languages. For, in OV languages, nominal modification, whether of demonstratives, adjectives, or relative constructions, is indicated simply by preposing. Thus neither Turkish nor Japanese has markers to indicate congruence, nor did Classical Armenian (Jensen 1959:47); I assume that early PIE also lacked expression for congruence. The gender distinctions as we know them, especially from the early dialects (Delbrück 1893:89-133), were developed when nominal modifiers and anaphoric elements came to stand in other positions than the prenominal. For when adjectives are placed after nouns, or when anaphoric pronouns stand at some distance from them, expressions to mark the relationship are highly useful. The syntactically simple first line of the Odyssey may serve as illustration:

37. Odyssey 1.1. ándra moi énnepe, moȗsa, polútropon, hòs mála pollà
man me you-tell O-Muse much-turned who very many
‘Tell me of the man, O Muse, of many devices, who [traveled] through very many places.’

The masculine endings of polútropon and hós relate these modifiers to ándra, excluding possible relationship with another substantive, such as moȗsa. Although we cannot account for innovations through poetic use, inasmuch as poets normally utilize the devices their languages provide rather than introduce new ones, nonetheless, it is difficult to conceive of Vedic poetry as we find it without the device of congruence. For virtually every stanza would be unintelligible if there were no congruence markers to relate nominal expressions. We account for the development of the IE gender system, then, as one process in the shift from an OV structure toward a VO structure.

The development of the surface markers of the gender system has been well described (Brugmann 1911:82-109, Kurylowicz 1964:207-226). It is also well known how the expression for gender meshes with that for case and number. Such morphological details are not of concern in determining the reasons for the introduction of gender distinctions in PIE. The situation in Hittite may however still be noted.

Hittite has no expression for the feminine gender. Following established terminology, some Hittite scholars labeled the two congruence classes of Hittite substantives masculine and neuter (Sturtevant 1951: 82-83). Today, however, the terms common and neuter are generally used. As Sturtevant pointed out, however, a neuter ending is found only for a stems, which have -n in the nominative. And nouns may be used adjectivally without indication of congruence; kurur means ‘inimical’ as well as ‘enemy’ and ‘enmity’. Hittite apparently, then, maintained an older stage in which the feminine category had not yet been introduced, though Hittite did introduce innovations by marking a neuter category in one nominal inflection and in pronouns (Sturtevant 1951:111). Hittite was then even more archaic than Vedic, in which neuter nouns could “function syntactically as singulars as well as plurals” (Gonda 1971:62; see also Delbrück 1888:80).

38. RV 1.64.5. ū́dhar divyā́ni
udder heavenly (pl.)
‘celestial udders’

Moreover, Hittite has a neuter plural -a, which corresponds to the ending identified by Schmidt as a collective; yet neither this ending nor the neuter category, is as well developed as in the other early dialects (J. Friedrich 1960:115-118). The “nominative-accusative neuter (of pronouns) both singular and plural, often refers to plural antecedents of either gender” (Sturtevant 1951:111). Hittite therefore has a nominal system in which the so-called neuter plural resembles a collective rather than the gender marker known from the other early dialects. Accordingly we may conclude that “the lack of a special feminine in Hittite is to be considered as an archaism” rather than “as an innovation” (Kurylowicz 1964:217). Hittite as a relatively consistent OV language had not yet developed fully the congruence markers which are found in the nominal phrase of other IE dialects. But in pronouns, where anaphoric referents were often widely separated from antecedents, as in relative constructions, an incipient congruence system was introduced.

In other dialects this system came to be extended. Whether it was ever extended in Ancient Armenian is unclear. But by the time of the texts which have survived in the other early dialects, a threefold congruence system had been developed for the nominal phrase, though it was not marked in all nominal forms. Through archaisms in the various dialects we can determine that this gender system arose in late PIE. As a syntactic congruence category, it is marked most clearly in those case forms which have essentially grammatical functions: the nominative and the accusative. In the late declensions, notably the o stems and the ā stems, however, the gender markings have been extended also to the oblique cases, notably the genitive.

5.3.3. Number.

Like gender, number is a congruence category; in PIE and the dialects it differs from gender in applying to verbs as well as substantives. Number, then, is a device for relating subjects with their verbs as well as the constituents of nominal expressions and anaphoric markers. Late PIE had three numbers: singular, dual, and plural. For the most part the uses of each of these categories corresponded with their natural meanings. Dual forms were used in connection with two items, especially when these formed a pair, such as the eyes or a divine couple (Meillet 1937:188, 338-339). Plural forms were used for more than two items; singular forms for one. As Meillet points out, however, a “singular” item composed of two or more elements could be in the plural, such as Homeric Gk. zeiaí ‘spelt(s)’; Vedic dúraḥ, Lat. forēs, etc. ‘door(s)’. Variations from natural uses also reflect older patterns of the language or foreshadow new ones (Delbrück 1893:133-172).

The system of verb endings clearly points to an earlier period in which there was no verbal inflection for number (see Kurylowicz 1964:57, 150). For the dual and plural endings are obviously defective. We cannot reconstruct endings in these two numbers which are as well supported as are those of the singular, except for the third plural.

39.   m inflection h inflection
  1 sg. m x
  2 sg. s th
  3 sg. t ø
  3 pl. nt r

Only the third plural, as indicated, can be posited for an early period of PIE. The development of precisely third-person forms to express number supports the assumption that the number category was used for congruence.

The number system is defective in substantival as well as in verbal inflection. The personal pronouns never did introduce expressions for plurality, as suppletive paradigms indicate, e.g., Hitt. uk ‘I’, uēš ‘we’, etc., in contrast with demonstratives, e.g., kāš, kē ‘this, these’, and nouns, e.g., antuhšaš, antuhšeš ‘man, men’. In the system of nominal inflection the most notable reflex of an older prenumber system is the ending -a, used in the nominative singular feminine and the nominative-accusative plural neuter. This use, discussed at length by J. Schmidt (1889), reflects a collective ending -h, which was lost in late PIE with the change of the preceding -e- to -a- (Lehmann 1958:191). The earlier use has left reflexes in various dialects, such as: the Hittite and Greek neuter plural subjects accompanied by verbs in the singular; Latin neuter plurals like loca accompanying masculines, i.e., locus ‘place’; the use of neuter plurals in Germanic for collectives consisting of masculine and feminine nouns (Lehmann 1957). Number accordingly was not consistently applied in late PIE and the early dialects in accordance with natural reference. Subsequently application became more regular, and number congruence was carried out for both substantives and verbs.

When the category of number came to be more fully established in the inflectional system, a threefold set of forms was introduced for comparison of adjectives: the comparative when two persons or items were involved, as in the dual; the superlative for more than two, as in the plural; and the positive for one. The absence of markers for comparison in Hittite and the variation in markers from dialect to dialect and within dialects like Greek indicate that comparison was late (Brugmann 1904a: 320-324); other evidence is found in suppletive forms throughout the dialects, some surviving as late as NE good, better, best.

The late development of the number system in the noun is also clear from the lack of parallelism between the forms of the dual, for which only three forms developed, and the forms of the singular and plural. Moreover, the oblique endings of the plural and the dual instrumental, dative, and ablative were added in accordance with the rules for external rather than internal sandhi (Macdonell 1916:39). Accordingly the dual was never developed as fully as were the other two numbers. The formal aberrancy and its restriction in use to pairs apparently led to its early loss. The number system, both in the noun and in the verb, came to be reduced to two categories, with eventual loss of uses and inflections corresponding to dual number, such as a distinct adjectival comparative as opposed to a positive and a superlative.

5.4. Surface Classes in Which These Categories Are Marked.

In late PIE the two major classes of words, verbs and nouns, were sharply distinguished, as Meillet emphasized (1937:187-192). Person was the dominant category in verbs, case in nouns. Accordingly, even though the two classes shared the category of number, and though they could be built on the same roots, verbs were characterized by their categories and the means of expression for these. Nouns and the other two substantival classes, pronouns and adjectives, were marked for the nominal categories. The numerals were in part nominal, in part adjectival.

5.4.1. The PIE Root.

Indo-Europeanists have been greatly concerned with the description and role of the PIE root. Two brilliant studies, both published in 1935 (Kurylowicz 1935; Benveniste 1935), dealt with the description of the root, taking into consideration the huge amount of earlier work as represented in the standard handbooks (Hirt 1921-1937; Meillet 1937; and many others). The monographs of Kurylowicz and Benveniste may be viewed as definitive, because they were able to solve perplexing problems by the assumption of laryngeals in the PIE phonological system. Accordingly, roots which had seemed aberrant, such as dhē- ‘place’ and es- ‘be’, were shown to conform to the pattern found in bher- ‘bear’ and were reconstructed dheʔ- and ʔes-. The canonical root form for PIE was accordingly CeC-.

Roots could be extended by means of suffixes, either suffixes accompanied by a vowel or consonantal suffixes. Such suffixation resulted in a restricted number of extended roots for verbs. The possibilities were summarized in Proto-Indo-European Phonology (Lehmann 1952:17-18); only a chart of the possible canonical root forms of primary verbs will be presented here:

40. Unextended root: CeC-
  Extended root: I   II
CeC-C-   CC-eC
IIa. CC-eC-C
IIb. CC-n-eC-

If further suffixes were added, the form was nominal rather than verbal.

In opposition to these conclusions it has been proposed that this scheme is far too mathematical, that languages do not conform so strictly to canonical shapes. Any Indo-Europeanist would admit exceptions, some probably results of borrowing. Yet the preponderance of the IE vocabulary can be analyzed in accordance with the assumptions on which the forms in Example 40 are based. We may therefore conclude that this description is correct and that the IE lexicon should be analyzed as based on roots which were affixed as indicated in Example 40.

The skepticism concerning this conclusion has resulted partly from statements on the role of the proposed root. The statements which seem dubious were based on the assumption that the reconstructed parent language was primitive, that if taken back far enough it could enlighten us on an early period in which man spoke in monosyllables. Archeological work in the past decades, or even the past century, has totally destroyed any assumption that language was introduced only a brief time before the period of PIE. Accordingly, PIE must not be viewed as a form of language close to the primeval period of man, nor even as a primitive language; actually we have no evidence of any primitive language, even though this term is still encountered in publications. We are then apparently left with the alternative of either considering the PIE root as a convenient fiction for IE grammatical study or trying to account for it in relation to similar elements in other languages.

The PIE root agrees nicely with comparable forms of OV languages. If, for example, we compare Japanese, we may note that two types of roots can be derived from verbs: roots ending in consonants, like yom- from yomu ‘read’, and roots ending in vowels, like ki- from kiru ‘put on, clothe’. These roots may take various inflectional and derivational affixes, which are added in accordance with the phonological principles of Japanese; every syllable must end in a vowel, or in vocalic . Thus the past tense, suffixed with -ta, is yonda for yomu and kita for kiru. Other suffixed forms may be examined in any description of Japanese. Another OV language, Turkish, also has simple roots to which affixes are added, for example, ver- ‘give’, gör- ‘see’, and so on. Like Japanese roots, the Turkish roots are found in their simplest forms in verbs and in nominal compounds. Thus, in Japanese the compounds with mono ‘thing’ of the two verbs cited are kimono ‘wear-thing > clothing’ and yomimono ‘reading matter’. These examples may clarify the role of the PIE root. It was not first and foremost a free form, though free forms consisting only of the root were possible. In Japanese ki ‘wearing’ can be a free form. The PIE root was used without suffixes in some imperative forms, as Latin fer ‘carry’ indicates. Although such free forms occur, the root is primarily the basic component of lexical elements which carries the fundamental lexical meaning. The PIE root then is as realistic in PIE grammar as are roots in Japanese, Turkish, and other OV languages.

From any root, extended forms of various kinds could be constructed. The analysis of these is one of the achievements of nineteenth-century linguistics and is amply treated in the handbooks (Brugmann 1906; 1913; Meillet 1937:146-173, 197-223, 252-325; Watkins 1969). When using handbooks and other treatments of PIE, great care must be applied in sorting out forms for various stages of the language. Recent scholarship has been concerned with the identification of such forms for various stages of the language. In general we cannot discuss these problems here, nor even deal with the various derived forms of nouns and verbs. Since any such discussions would greatly extend this presentation, we limit ourselves to brief definitions of verbs and nouns.

5.4.2. The PIE Verb.

The verb in PIE was characterized by inflection for person, number, aspect, mood, and voice. The roots and stems on which verbal forms were constructed were strictly limited, as noted in Example 40; these bases were in general simpler than those of nouns. Yet denominative forms of verbs were also possible, particularly in later stages of the language. These were common in the so-called weak inflections. In time they led to a different type of verbal inflection from that of PIE, an inflection characterized by secondary suffixes and composite forms. The early PIE verb is illustrated most clearly in injunctive forms, of which examples have been cited earlier. Forms with the four endings of the m inflection given in Example 39 are attested in the Veda, as for dhā- (for dhām see Hoffmann 1967:150).

41. 1 sg. dhām véda
  2 sg. dhāḥ < dhās véttha
  3 sg. dhāt véda
  3 pl. dhuḥ < dhur vidús

Because of subsequent changes, the endings of the h inflection have become modified in all the dialects; but they may be reconstructed from Vedic forms like those cited in Example 41 for the extended root vid-, as well as from forms in other dialects. More complex finite verb forms have been treated above. Like the forms cited in Example 41, each finite form was inflected for first, second, or third person; for singular, dual, or plural number; for ±momentary aspect; and for one of the moods or voices.

5.4.3. The PIE Noun.

The noun in PIE was characterized in inflection for case, gender, and number, Endings indicating these were generally added to extended forms of roots, but some root nouns are attested in the early dialects, such as forms in the Rigveda of the masculine noun pád- ‘foot’ (Example 42). (The forms in parentheses may be posited because of attestations of other nouns; root forms with vocatives have the same forms for these as for nominatives.) The three possible dual forms of pád- are attested; in the singular and the plural the nominative and dative forms are not attested, as might be expected for a nonanimate noun.


(See Meillet 1937:254-255 for other root nouns.) These inflected forms express the nominal categories which have been discussed above.

5.4.4. Pronouns.

Personal pronouns were inflected for case and number. Demonstrative and anaphoric pronouns, which supplied the forms for the third personal pronoun, were also inflected for gender (Delbrück 1893: 460-521; Meillet 1937:332-338). We have noted above that the gender category was late. Pronouns provide evidence for the conclusion that the category of number was also introduced late, as differing roots of personal pronouns in the singular, dual, and plural indicate: *eg- ‘I’, *ne- ‘we two’, *we- ‘we’; *tew- ‘thou’, *we- ‘you two’, *yew- ‘you’. Moreover, a different root is used in the cases of the first-person pronoun other than the nominative: *me- ‘me’, *nes- ‘us’. This shift in root may be compared with stem changes between the nominative and the oblique cases in heteroclitic nouns, such as the neuter Hitt. watar, genitive wetenaš ‘water’, or nouns with -en- suffix in the oblique cases, such as Lat. homō, accusative hominem ‘man’ (Meillet 1937:262-263; see also Sturtevant 1951:102-108). The shifts are archaisms maintained from the period before development of the case system, and they provide evidence that inflection for case categories was also relatively late.

The reflexive pronoun was late. It developed after the loss of the middle from a root *sew- ‘own’. Meillet has assembled illustrations of the use of this root in compounds and as an adjective (1937:336-338). Patterned after forms of *tew- ‘thou’, *sew- developed a set of singular case forms other than the nominative; these were used to refer to plural as well as singular subjects.

The pronouns also differ from nouns in having enclitic forms. The enclitic pronouns were placed in second position in the sentence and are especially prominent in the Anatolian languages, where sentence-connecting particles are followed by these and other enclitics (Carruba 1969).

Rather than a third-person pronoun parallel with the first and second, PIE had a number of demonstrative stems and, in addition, interrogatives, which were used as the basis of indefinites (Brugmann 1904b; Meillet 1937:325-329; Brugmann 1904a:399-406). These developed a full set of case forms in the singular, dual, and plural, with the obvious exception of the vocative. The relative lateness of their inflection is clear from the parallelism with nominal o- and ā- stems, the most recent PIE inflection. Demonstratives and interrogative indefinites differ from the personal pronouns in having adjectival as well as nominal use. Moreover, they developed characteristic endings in part; see Brugmann's chart for details (1904a:406); also see Sturtevant (1951:108-115).

The simple demonstrative, or anaphoric pronoun, was *so-, sā-, to-. Besides this a variety of roots was the basis for a demonstrative indicating objects close to the speaker, notably roots with k-, and a demonstrative indicating more remote objects, notably roots with n. PIE then had threefold deixis in demonstratives. Brugmann discusses this threefold deixis with reference to the personal pronouns, noting an “I-deixis”, a “you-deixis” and the more remote reference deixis (1904b). By his approach the personal pronouns are closely connected with demonstratives. The interrogatives and indefinites were based on roots with kw-.

The use of relative pronouns, like that of reflexives, depends on the structure of the language. Since early PIE was OV in structure, relative constructions were marked simply by preposing. When a marker was necessary to indicate relative clauses (Raman 1973), the kw- forms were adapted in some dialects: Anatolian, Italic, Slavic, and Baltic; other dialects used demonstrative forms: y- forms in Indic and Greek, s- forms in Germanic and Celtic (Gonda 1971:75, 138). Since our Celtic materials are from the Insular VSO languages, relative constructions are generally marked by position, or by special verb forms, rather than by pronouns, in keeping with the patterning of other VSO languages.

5.4.5. Adjectives.

In late PIE and the early dialects, adjectives were distinguished from nouns and pronouns by inflection for each of the three genders. Many adjective classes, however, did not have distinct inflections for the three genders, as is evident even from Latin grammar; some adjectives of the third declension had distinct endings only for neuter nominative, accusative, and vocative: m.f. gravis, n. grave ‘heavy’; others, including present participles, had the same ending for the three genders in the nominative and vocative singular, such as memor ‘mindful’, amāns ‘loving’. Adjective inflection accordingly developed only late in IE (Delbrück 1893:400-460).

For an early period of PIE we may assume that adjectives were uninflected, as relic constructions in Hittite indicate; kurur, as we have noted above, § 5.3.2, could be used adjectivally by simply preposing it to nouns.

Besides their inflection for the three genders, adjectives also were distinguished by their inflection for comparison. This too was late. In early PIE, adjectives were not a distinct inflectional class by reason of either gender inflection or comparison.

5.4.6. Numerals.

The role of adjectives in early PIE may be inferred from that of the cardinal numerals. These were used adjectivally as well as nominally (Brugmann 1904a:362-372; Delbrück 1893:522-536). As inflection became more prominent, the numerals most clearly associated with the number category came to be inflected. The numeral for ‘one’ based on *oy-, was inflected like an adjective in the singular; so was the root *sem- ‘one’, which was used in Greek. The numeral for ‘two’, based on *dwō-, was inflected as a dual, subsequently as a plural. The numerals for ‘three’ and ‘four’ were also inflected as plurals. The remaining cardinal numerals were uninflected, though in some dialects they came to be inflected (Brugmann 1904a:362-370). Ordinals were always treated like adjectives (Brugmann 1904a:370-372; Szemerényi 1960: esp. 92); as Oswald Szemerényi has demonstrated definitively, they are late adjectival forms based on the cardinals. The numerals then provide us with important insights into the development of inflected adjectival and pronominal forms from noninflected, preposed modifiers.

5.4.7. Participles and Other Nominal Forms Derived from Verbs.

Nominal modifiers were also derived from verbal forms. Based on stems with various meanings, they incorporated the category of voice and the aspectual distinctions (Delbrück 1897:218-220, 250-251, 476-499). We may illustrate the inclusion of such categories in participles with the often-cited line from the Iliad, in which participles of *es- incorporate active and middle voice, as well as continuous present and future meanings.

43. Iliad 1.70.
hòs ḗidē t’ eónta t’ essómena
who he-knew those and being those and will-be-happening
pró t’ eónta
before and being
  ‘who knew what is, what will be, and what was in the past’

Such nominal forms seem to correspond in use to adjectives and nouns. Yet they retain the functions of finite verbal forms in many of their early uses, as we may indicate by comparing them with participles in a VSO language like Ancient Egyptian.

A well-known use of participles, in which they correspond to clauses of circumstance or time, may be illustrated with a frequent formula in Homer:

44. Odyssey 1.63. tḗn d’ apameibómenos proséphē
her Ptc. answering he-said
‘Then answering her he said...’

Such uses are impossible in Ancient Egyptian, according to Alan Gardiner (1950:270). We may view the IE constructions as archaisms from the earlier OV syntactic patterns. As we have noted in § 4.8, Example 145, Japanese has participal forms in a wide array of meanings; like apameibómenos in Example 44, these are placed before the clause with the finite verb, though, like other nominal forms in Japanese, they are not inflected. We may assume that such forms were found in early PIE and that they subsequently came to be inflected, as congruence markers were extended to nominal modifiers. This assumption is supported not only by the large number of uses of participles in IE but also by the archaic inflectional patterns of nt- participles, as we have noted for the Latin present participle with its single nominative ending and no distinction for gender. When participles came to be viewed as nominal forms, nominal inflections may also have been extended to them because of the subjective dominance of verb forms, as noted in § 4.2.

The recent addition of inflection to nominal forms of verbs may also be observed in infinitives and supines (Delbrück 1897:440-475; Sturtevant 1951:74-75 148-149; J. Friedrich 1960:142-144). Infinitival forms could simply be combined with nouns, as complements; see Delbrück (1897:465-467), who gives among others the following example:

45. RV 1.113.16. ā́raik pánthāṃ yā́tave sū́ryāya
she-cleared way for-going for-sun
‘She cleared the way for the sun to go.’

Such a use of the infinitive is found also in Greek, Germanic, Baltic, and Slavic. In Greek the article came to be used with the infinitive (Schwyzer 1950:368-372).

By a further development the infinitive could be used alone in complementation, as noted in § 4.8; it then had the same subject as the principal verb, as in Hittite (J. Friedrich 1960:143):

46. SANGA akuu̯anna u̯ekzi
  priest drink he-requests
  ‘The priest requests to drink.’

This use led to periphrastic constructions, both with participial or adjectival forms of the verb and with infinitival or nominal forms, as already in Hittite (J. Friedrich 1960:136-137):

47. hurtanteš ešir
  cursed they-were
  ‘They were cursed.’

A periphrastic construction with nouns was accompanied by dāi- ‘place’ or tii̯a- ‘step’ (see Götze and Pedersen 1934:21-22):

48. Muršilis Sprachlähmung I.7. nu-mu ... aši memii̯aš tešhaniškiu̯an tii̯at
Ptc.-to-me that situation appear-in-dream began
‘And that situation began to appear in my dreams.’

The nominal elements of the verb thus were employed in various constructions in the dialects (Brugmann 1904a:603-609). These developed from uses in PIE which were like those of adjectives and nouns.

In accordance with these uses, nominal forms of verbs were normally related closely to other elements of the sentence. The various dialects however also contain absolute constructions, that is, case forms of nouns accompanied by participles in constructions which resemble independent clauses. The cases differ from dialect to dialect and also within dialects. Moreover, the uses vary considerably, from manner to time to means, and so on. From this variation we may conclude that the absolute constructions developed independently in each of the dialects. Such independent origins of parallel constructions are plausible only if we can ascribe them to typological reasons. I have proposed that absolute constructions arose as the various dialects were developing from OV to VO languages. During the transition period a participle might be related either to a preceding or to a following finite verb. Accordingly the participle and its noun were used “absolutely,” without reference to a finite verb, in the Sanskrit locative absolute, the Greek genitive absolute, the Latin ablative absolute, and the Gothic, Slavic, and Baltic dative absolute. This explanation for the origin of absolute constructions receives support from the possible location of absolutes before or after the finite verb in the various dialects (Lehmann 1972b:987-989). The following example of a Vedic locative absolute provides a further illustration of its independence, though the relationship to other constituents of the sentence is clear:

49a. RV 1.16.3.
índram prātár havāmaha
Indra early we-call-on
índram prayaty àdhvaré
Indra going-along sacrifice
  ‘We call on Indra early, on Indra as the sacrifice proceeds.’

The locative adhvaré, accompanied by the present participle prayatí, is parallel on the one hand to the adverb prātár, on the other hand to an independent clause. And when one compares the third line of this stanza:

49b. índraṃ sómasya pītáye
  Indra of-soma at-drinking
  ‘on Indra at the drinking of soma’

the participial absolute seems parallel to a verbal noun in the dative. In this way verbal derivatives of various kinds could be used to express categories generally associated with nouns, such as the case category of time in the three lines of Example 49. Other mechanisms for expressing syntactic categories in the sentence were provided by particles and frozen forms of inflected words, generally referred to as adverbs.

5.5. The Use of Particles and Other Uninflected Words as Syntactic Means of Expression

Noninflected words of various functions were used in indicating relationships between other words in the sentence or between sentences. Some were used for modifying nouns, often indicating the relationships of nouns to verbs. Although these were generally placed after nouns and accordingly were postpositions, they have often been called prepositions by reason of their function rather than their position with regard to nouns (Delbrück 1888:433-471). Others were used for modifying verbs, often specifying more precisely the meanings of verbs; these then may be called preverbs. Others, commonly referred to as sentence connectives, were used primarily to indicate the relationships between clauses or sentences (Watkins 1964; Lehmann 1969).

5.5.1. Postpositions.

Surveys of the postpositions in the various dialects are given in the handbooks, as for Hittite (J. Friedrich 1960:129-130) and Vedic (Delbrück 1888:440-470). They are found with specific cases, in accordance with their meanings. Yet in the Old Hittite texts, the genitive rather than such a specific case is prominent with postpositions derived from nouns, such as piran ‘(in) front’ (Neu 1970:10, 59):

50. kuiš LUGAL-ua-aš piran ēšzi
  who king's front he-sits
  ‘whoever sits before the king’

Such postpositions came to be frozen in form, whether unidentifiable as to etymology; derived from nouns, like piran; or derived from verbs, like tirás in Example 51. Further, as the language came to be VO, they were placed before nouns. As case forms were less clearly marked, they not only “governed” cases but also took over the meanings of case categories. The preposition tirás (tiró), derived from the root *tṛ- ‘cross’, illustrates both the etymological meaning of the form and its eventual development as preposition:

51. RV 8.82.9.
yáṃ te śyenáḥ padā́bharat
what for-you eagle with-foot-he-bore
tiró rájāṅsy áspṛtam
crossing, through skies not-relinquishing
píbéd [<píba íd] asya tvám īśiṣe
you-drink-indeed of-it you you-are-master (for-your-benefit)
  ‘What the eagle brought for you in his claws, not dropping it [as he flew] through the skies, of that drink. You control [it for your own benefit].’

The syntactic use of such particles with nouns is accordingly clear.

5.5.2. Preverbs.

Rather than having the close relationships to nouns illustrated above, particles could instead be associated primarily with verbs, often the same particles which were used as postpositions. Thus in the following Hittite passage piran is used as preverb (Szabó 1971:24):

Ptc. to weather-god and his-male-gods to sun-god
Ù DINGIR.LÚMEŠ-ŠU píran zikizzi
and his-male-gods front he-lays-in-turn
  ‘He lays before in turn [the slices of bread] for the weather god and his male gods and the sun god and his male gods.’

See also Example 15 in Chapter 2. Such combinations of particles and verbs came to be treated as units and are found repeatedly in specific uses (Delbrück 1888:433-437).

Preverbs may occupy various positions: if unmarked, they are placed before the verb; if marked, they are placed initially in clauses (Watkins 1964), as in the following examples of úpa and ápa from an Archaic hymn:

53. RV 7.71.4. ā́ na enā́ nāsatyā úpa yātam
to us with-that Nasatyas hither you-two-come
‘Come hither to us with that, Nasatyas.’

And the hymn begins with the following line, in which the departure of Night is emphasized by initial placement of ápa ‘off, away’:

54. RV 7.71.1. ápa svásur uṣáso nág jihīte
away from-sister from-Dawn Night she-departs
‘Night goes away from her sister Dawn.’

In the course of time the preverbs in unmarked position came to be combined with their verbs, though the identity of each element is long apparent in many of the dialects. Thus, in Modern German the primary accent is still maintained on some verbal roots, and in contrast with cognate nouns the prefix carries weak stress: erteílen ‘distribute’, Úrteil ‘judgment’. The steps toward the combination of preverb and verbal root have been described for the dialects, for example, Greek, in which uncombined forms as well as combined forms are attested during the period of our texts.

Preverbs which remained uncombined came to be treated as adverbs. Thus adverbs, as in the sequence:

55. RV 1.147.5. átaḥ pāhi
thence you-protect
‘Protect us from that.’

are similar in use to preverbs like ápa in Example 54. Combinations of preverbs plus verbs, on the other hand, eventually came to function like unitary elements. Already in the Rigveda, compounds of frequent verbs like i ‘go’ may be regarded as independent verbs. Combinations of i plus sam ‘come together, join’, for example, are used with the dative or locative, and they may also be combined with other preverbs:

56. RV 6.19.9. ā́ viśváto abhí sám etu
toward from-all-sides hither together it-should-come
‘It should join [us] here from all sides.’

The two different positions of preverbs in early texts led eventually to different word classes.

5.5.3. Sentence Particles.

Particles were also used to relate sentences and clauses (J. Friedrich 1959:18, § 11):

takku LÚ.ULÙLU-an EL.LUM QA.AZ.ZU našma GÌR-ŠU kuiški
if man free his-hand or his-foot someone
tuu̯arnizzi nušše 20 GÍN KUBABBAR paai
he-breaks Ptc.-to-him 20 shekels silver he-gives
  ‘If anyone breaks the hand or foot of a freeman, then he must give him twenty shekels of silver.’

Particles like the initial word in this example indicate the kind of clause that will follow and have long been well described. The function of particles like nu is not, however, equally clear.

A generation ago Dillon and Götze related nu and the use of sentence connectives to similar particles in Old Irish (Dillon 1947). Such particles introduce many sentences in Old Irish and have led to compound verb forms in this VSO language. Delbrück had also noted their presence in Vedic (1888:23-24), as in the following examples:

58. hovāca gā́rgyaḥ
  Ptc. Ptc.-he-said Gārgya
  ‘Then Gārgya said...’

59. tád u hovācā́suriḥ
  Ptc. Ptc. Ptc.-he-said-Asuri
  ‘Then Asuri said...’

Since introductory šu and ta were more frequent than was nu in the older Hittite texts, scholars assumed that sentences in IE were regularly introduced by these sentence connectives. And Sturtevant proposed, as etymology for the anaphoric pronoun, combinations of *so- and *to- with enclitic pronouns, as in the well-known Hittite sequence ta-at, cf. IE *tod, and so on (see Otten and Souček 1969:89-90 for the use of such particles in one text). Subsequent study has not corroborated assumptions on the consistent use of sentence connectives, even in narrative patterns. Otten and Souček found, for example, that their Old Hittite text contained far more uses of nu than were expected (twenty-two) and only one šu in contrast with thirty-two attested uses of ta. Moreover, sentence connectives seem to be characteristic of VSO languages. Squamish, for example, resembles Hittite in beginning its sentences with “sentence connectives” followed by other enclitics (Kuipers 1967:206-211). Since Hittite was in close contact with the VSO language Akkadian, the widespread use of nu and the other sentence connectives may be the result of influence from other languages. And if, as especially Szemerényi has maintained, Indo-Iranian was influenced by Semitic languages before it was carried further east, the occasional uses of nū, sa, and ta in Vedic as sentence connectives may also be the result of outside influence rather than an inheritance from PIE. The apparent similarity between Old Irish and Hittite may therefore be explained on grounds other than common origin. Further study of the role of sentence connectives in language is therefore essential before we can propose firm conclusions about their possible role in PIE.

It is clear however that sentence connectives were used in Hittite to indicate continued treatment of a given topic (Raman 1973:165). This use was illustrated in Example 59. It is also found with Hittite relative constructions. Such a function may also be ascribed to Vedic and tád, as in Examples 58 and 59. Yet, since this use may be accounted for through post-PIE influences, sentence connectives may have had a minor role in PIE.

Other particles, like Hitt. takku ‘if’, probably had their counterparts in PIE, even if the surface forms were completely unrelated. This is also true for emphatic particles like íd in Example 51; they were used after nouns as well as imperatives like píba (see Speyer 1896:72). Such emphatic particles combined with imperatives suggest the presence of interjections, which cannot be directly reconstructed for PIE but are well attested in the several dialects.

Although sentence-introductory particles like Hitt. takku can be reliably assumed, there is less evidence for coordinate sentence connectives. A coordinating particle can clearly be reconstructed on the basis of Skt. ca, Gk. te, Lat. que, and so on. But its primary function is the coordination of elements in the sentence rather than clauses or sentences. Moreover, when ca is used to connect verbs in the Vedic materials, they are parallel (Delbrück 1888:472-474); Delbrück finds only one possible exception. This situation reminds us of the OV language Japanese; for in an OV language the relating of successive verbs is carried out by means of nonfinite verbs placed before finite. We may then expect that coordinating particles had their primary use in PIE as connectors for sentence elements rather than for sentences.

Another such particle is *wē̆ ‘or’. It is notable that, like *kwe, the particle indicating disjunctive ‘or’ was postposed, in retention of the OV pattern as late as Classical Latin.

We have noted above that particles in PIE may also have corresponded to verbal qualifiers. The most notable of these is *, which carried a negative modal meaning. There is indication of such uses of particles in other patterns, for example, of Vedic purā́ ‘earlier’ to indicate the past, as apparently Brugmann was the first to point out (Delbrück 1888:502), and also Vedic sma, to indicate repeated action in the past (Hoffmann 1967:201, 203). It is curious that sma is also found after mā́ in Vedic (Hoffmann 1967:91, 98). As we have suggested above, such mood- and tense-carrying particles may have been transported from a postverbal to a preverbal position. Some particles may accordingly have been equivalent in an earlier stage of PIE to elements used after verbs to indicate verbal categories.

5.6. On the Generation of Sentences.

Having examined the syntactic categories of PIE and its syntactic rules, we may observe how they account for sentences.

The P rules presented in § 1.3 and subsequently amplified would produce the following underlying structures:

  1. Σ → Conj. Σn (optional)
  2. Σ → Q Prop.
  3. Q [±Dec.] [±Int.] [±Neg.] [±Mid.] [±Nec.] [±Vol.] [±Perf.] [±Mom.] [±Iter.] [±Caus.]
  4. P V (Target) (Receptor) (Agent) (Means) (Source) (Time) (Place) (Manner)

One of the possibilities generated by these rules would be as follows:

(Σ →) +Dec. -Int. -Neg. +Mid. -Nec. -Vol. +Perf. +Mom. -Iter. -Caus. V Agent Means

This string of categorial elements would generate a sentence with a middle verb in the aorist. Since the verb is unspecified for person and number, the verb form would be third singular. The +Dec. would require that it be indicative; and the following minus categories would require it to be a positive statement. The verb would be accompanied by a noun in agential relationship with the verb; such a noun might be expected to be animate, though it could also be an “active” natural force. This nuclear sentence might be accompanied by other nominal forms, possibly in adverbial relationship.

As an example of such a sentence we may cite from an Archaic hymn:

60a. RV 5.52.6. bhānúr arta tmánā
light has-appeared by-itself

The position of the adverbial instrumental indicates its facultative inclusion in the sentence; as discussed in Chapter 2, one or more such elements could be introduced in sentences. The actual line in the hymn includes a further element following tmánā: disáḥ ‘of the sky’, a genitive modifying bhānúr. This nominal modifier would have been introduced by the optional P rule (P Rule 6):

NP → (Det.) N (Σ)

The modifying Σ would be reduced in accordance with processes that are well known and not of interest here.

The basic sentence in Stanza 6 of this Archaic hymn then reports a sunrise, in Geldner's translation:

60b. von selbst bricht der Glanz des Himmels an
  ‘the light of the heavens dawns spontaneously’

Many other such sentences that are attested in our early texts, in the Anatolian, Greek, and other dialects as well as Indo-Iranian, could be generated in much the same way as this sentence.

A further, somewhat more complex example from a Strophic hymn may provide another illustration; again the verb is based on the root ṛ- ‘go’, which will be discussed in Chapter 6.

61. RV 7.56.21. mā́ vo dātrā́n maruto nír arāma
may-not your from-gift O-Maruts out, except we-will-go
[interpretation of verbal phrase with ablative = ‘we will be deprived of’]
‘May we not be deprived of your gift, O Maruts.’

This sentence would be generated from the following categorial set: +Dec. -Int. +Neg. -Mid. -Nec. +Vol. +Perf. +Mom. -Iter. -Caus. V Agent 1 pl. Det. Source, coordinated with a nominal sentence generating the vocative maruto. The +Vol. category is attached to the negative particle as well as to the V, and accordingly determines the selection of mā́. The agent node is attached to the V, generating the first-person plural form. The pronominal vo is generated from the node Det. of the optional P rule generating nominal modifiers.

Further such examples could be given, for complex and compound sentences as well as simple sentences. Rather than an account of the generation of further sentences, some examples will be cited to illustrate the analysis of complicated sentences in accordance with the syntactic rules given above.

5.7. On the Analysis of Sentences.

As an example of a complex sentence we may take the opening of the Hittite Laws (J. Friedrich 1959:16):

takku LÚ-an našma SAL-an šullannaz kuiški kuenzi
if, when man or woman from-quarrel anyone he-kills
apūn arnuzi...
that-one he-brings (recompense)...
  ‘If anyone kills a man or a woman because of a quarrel, he gives recompense for that one...’

In the principal clause the verb is a cognate of Skt. ṛ- ‘go’, with causative meaning indicated by the nu- form. P Rules 3 and 4 would accordingly generate the categories expressed by the principal clause, which may roughly be paraphrased: ‘he causes that recompense goes with regard to him on a given occasion’. The verb is accompanied by categories for the agent, the target, and time. The agent category is expressed in the verb form. The target category is expressed by the demonstrative pronoun. The time category is expressed by the clause introduced and made more specific by the particle takku ‘if’.

The temporal-conditional clause introduced through P Rule 6 is a sentence consisting of a verb accompanied by an agent expressed by the indefinite pronoun and the verb form, as well as by a compound target and a source category. The arrangement of the two clauses with regard to each other is that expected in an OV language. If we view the first clause as subordinate, this sequence of clauses is an example of embedding resulting in a complex sentence.

If we parse the following clause as coordinate with the principal clause, assuming that the obscure Ù means ‘and indeed’, the sentence consists of two coordinate clauses plus the subordinate clause:

62b. ...Ú 4 SAG.DU pāi LÚna-ku SALna-ku
  ...and-indeed 4 heads he-gives man-or woman-or
  ‘...and indeed he pays the equivalent of four persons, either men or women.’

This sentence, introduced through P Rule 1, consists of an agent incorporated in the verb and a target preceded by the Det. expressed by the numeral ‘four’. The noun introduced through the target category is itself modified by the disjunctive apposition. This examination may illustrate that sentences of moderate complexity yield to analysis as readily as simple sentences.

Additional complexities are simply results of further coordination or embedding. For example, relative constructions, attributive adjectives, or genitives are modified forms of embedded sentences, as may be illustrated with citations from an Archaic hymn:

63. RV 2.33.5.
hávīmabhir hávate havírbhir
with-invocations is-supplicated who with-oblations
áva stómebhī rudráṃ diṣīya
down with-songs-of-praise Rudra I-would-give (Aor. Opt.)
  ‘With songs of praise I would appease Rudra, who is supplicated with invocations and oblations.’

The first of these two lines has a middle form in passive sense, with Rudra as target of the verb hū- ‘call’; embedded in the main clause, this undergoes deletion of the target, with substitution of the relative marker yo < yas. Even further reduction is carried out in descriptive genitives and adjectives, as in:

64. RV 2.33.8. gṛṇīmási tveṣáṃ rudrásya nā́ma
we-sing terrible Rudra's name
‘We invoke the terrible name of Rudra.’

The basic sentence consists of nā́ma and gṛṇīmási; in it are embedded constructions corresponding to ‘the name is terrible’ and ‘the name is Rudra's’, each with reduction of the equivalent NP corresponding to nā́ma of the principal clause. Further analyses of complex sentences would simply repeat and amplify the processes noted here.

5.8. Marked Order in Sentences.

Elements in sentences can be emphasized, by marking; the chief device for such emphasis is initial position. Arrangement in this position is brought about by stylistic rules applied after structures have been generated by P rules and transformational rules. The sentence quoted above is a ready example. In Stanza 8 of this hymn the last two lines have initial position of the verb. The stylistic basis for this position is clear; these lines simply rephrase the content of the initial statement, emphasizing the poet's praise of Rudra:

65. RV 2.33.8.
prá babhráve vṛṣabhā́ya śvitīcé
forth for-brown for-bull for-whitish
mahó mahī́ṃ suṣṭutím īrayāmi
of-great great praise I-cause-to-move
namasyā́ kalmalīkínaṃ námobhir
I-will-do-homage radiant with-obeisances
  ‘For the brown and whitish bull I send out a great hymn of praise of the great one. I will do homage to the radiant one with obeisances.’

The last line, given as Example 64, continues the marked pattern both with initial position of the verb and selection of the present tense.

Other sentence elements may also be placed in initial position for marking. An example is the preverb prá in the first line of this stanza (Watkins 1964:1035-1042, Lehmann 1969:9-10). In unmarked position the preverb directly precedes the verb. Changes in normal order thus provide one of the devices for conveying emphasis.

Other devices have to do with selection, notably particles which are postposed after a marked element. These have been identified in the handbooks (e.g., Delbrück 1888:471-540; see esp. 497-501). Emphasis can also be indicated by lexical selection, as in the repetition of mah- in the second line of Example 65. Presumably other modifications might also be made, as in intonation, as we have noted above in the discussion of pluti. The various syntactic devices accordingly provided means to introduce marking in sentences.

5.9. Topicalization with Reference to Emphasis.

If we analyze sentences as consisting of a known and an unknown or new component, that is, of a theme and a rheme, segments of the theme may be sorted out as forming the more important elements. These may be said to form the topic of the discourse. They are distinguished by arrangement. For example, in the sentence from the Hittite Laws (Example 62), the topic is indicated as ‘man or woman’ by initial placement of these elements, rather than ‘as a result of a quarrel’. Similarly in Stanza 8 of the Archaic hymn to Rudra (Example 65), the topic is indicated as ‘the bull (Rudra)’ rather than ‘praise’ by the arrangement of the two phrases with regard to each other. Like emphasis, topicalization is carried out by patterns of arrangement, but the arrangement is applied to coequal elements rather than elements which are moved from their normal order.

Topicalization by arrangement is well known in the study of the early languages, as in the initial lines of the Homeric poems. The Iliad begins with the noun mȇnin ‘wrath’, the Odyssey with the noun ándra ‘man’. These, to be sure, are the only possible nouns in the syntactically simple sentences opening both poems: mȇnin áeide ‘Sing of the wrath’ and ándra moi énnepe ‘Tell me of the man’. Yet the very arrangement of moi and other enclitics occupying second position in the sentence, in accordance with Wackernagel's law, indicates the use of initial placement among nominal elements for topicalization.

The use of topicalization may be illustrated by a more complex set of sentences, such as the first address of Zeus in the Odyssey. Only the first lines of this will be quoted; but these indicate a shift in topic from the ‘gods’ to ‘men’, then to a particular man, Aegisthus, then to Agamemnon, and subsequently (in line 40, not included here) to Orestes.

66. Odyssey 1.32.
Ȏ pópoi, hoȋon dḗ nu theoùs brotoì aitióōntai;
alas how indeed now gods mortals they-blame
eks hēméōn gár phasi kák’ émmenai, hoi kaì autoì
from us for they-say evils are they but and themselves
sphȇisin atasthalíēisin hupèr móron álge’ ékhousin,
own stupidities beyond fate woes they-have
hōs kaì nȗn Aígisthos hupèr móron Atreídao
thus and now Aegisthus beyond fate of-son-of Atreus
gȇm’ álokhon mnēstḗn, tòn d’ éktane nostḗsanta,
he-married spouse wedded him but he-killed having-returned
  ‘Alas, how the mortals are now blaming the gods. For they say evils come from us, but they themselves have woes beyond what's fated by their own stupidities. Thus Aegisthus beyond what was fated has now married the wedded wife of Agamemnon, and killed him on his return.’

As this passage and many others that might be cited illustrate, the basic sentence patterns could be rearranged by stylistic rules, both for emphasis and for topicalization. In this way the relatively strict arrangement of simple sentences could be modified to bring about variety and flexibility.