Proto-Indo-European Syntax

< previous section | Jump to: next section >

4. Verbal Modifiers

4.1. On Elements Introduced through the Q Component.

4.1.1. Introduction.

The grammar proposed in Chapter 1 generates elements, like interrogation and negation, which relate to the entire sentence, by means of the Q constituent of the rule Σ → Q Prop. Although such elements have been discussed in earlier theoretical works, they have not been completely and unambiguously identified. An earlier statement, with some bibliographical review was given by Otto Jespersen (1924: 301-337). Theoretical statements like those of Jespersen, which are based largely on logical analysis with reference to European languages, will have to be augmented by investigations of the elements concerned in a broad span of languages (Hope 1972, Lehmann 1973a) and by closer analysis of well-known languages (Calbert 1971; Pflueger, Saad, and Lehmann, forthcoming). But already some elements have been hierarchically related as well as clearly identified. For the surface expression in many consistent VSO languages and (S)OV languages reflects with remarkable exactness the hierarchical structure of Q elements in surface features associated with the verb.

Before examining the verbal phrase of PIE we will review briefly the central Q features which we expect to find in languages. We will also note syntactic features which are introduced in the P constituent and, like Q features, often attached to verbs. And since PIE was OV in structure, we will examine briefly the verbal system of characteristic OV languages; for we may expect these to contribute to our understanding of the verbal system of PIE.

4.1.2. Some Characteristics of OV Verbal Systems.

As we have noted in Chapter 1, verbal inflection of OV languages is accomplished by means of suffixation. Moreover, OV languages are generally agglutinative (Lehmann 1973a). To understand the PIE verbal system it is also highly important to be aware of the OV characteristic that a verb form may have a minimal suffix, as well as a group of suffixes conveying a complex verbal meaning. We illustrate this characteristic by means of examples from Turkish.

In Turkish a form for stating generally valid facts is expressed with a suffix, (VOWEL)r, followed by deictic suffixes indicating person and number. This form contrasts with one which indicates a momentary action, The forms of these two inflections, which are also called Present I and Present II, made from the root gel- (infinitive gelmek ‘come’) are as follows:

    General Action   Momentary Action
1 sg.   gelirim ‘I come’   geliyorum ‘I am coming’
2 sg.   gelirsin   geliyorsun
3 sg.   gelir   geliyor
1 pl.   geliriz   geliyoruz
2 pl.   gelirsiniz   geliyorsunuz
3 pl.   gelirler   geliyorlar

Besides the agglutinative characteristics illustrated by these forms, Turkish exhibits another frequent characteristic of OV languages: vowel harmony. In its eight-vowel system a suffix like im selects one of the four high vowels in accordance with the preceding vowel; a suffix like ler selects either the front variant e or the back a. Some elements, like the yor of the suffix indicating momentary action, however, are themselves not affected by vowel harmony. The eight-vowel system is as follows:

i   ü   ı   u
e   ö   a   o

In the illustrations of further verb forms only the first singular and the, third singular will be given.

The interrogative (affix mi, mu) of the momentary inflection:
1 sg. geliyormuyum ‘Am I coming?’
3 sg. geliyormu  
The negative (affix mi):
1 sg. gelmiyorum ‘I am not coming.’
3 sg. gelmiyor  
The interrogative of the negative:
1 sg. gelmiyormuyum ‘Am I not coming?’
3 sg. gelmiyormu  
The optative (affix e) of gelmek, for the first person singular:
Simple Optative Interrogative Negative
geleyim ‘I may come.’ geleyimmi gelmeyim
The necessitative (affix meli) of gelmek for the first person singular:
Simple Necessitative Interrogative Negative
gelmeliyim ‘I must come.’ gelmelimiyim gelmemeliyim

These selected forms illustrate the structure of an OV verb system. Suffixes are appended one to another to convey compound meanings. But the compound suffixes are not mandatory. Rather, a simple form may consist only of a stem, such as gelir-, and a person suffix, such as -im. A similar form has survived in the Vedic injunctive.

Injunctive forms consist only of the verbal stem plus a suffic indicating person. Illustrations by means of forms which occur in Vedic texts are given for the root sthā- ‘stand’ in its present and aorist stem:

    Present   Aorist
1 sg.   ——   sthām
2 sg.   tiṣṭhaḥ   sthāḥ (ḥ < s)
3 sg.   tiṣṭhat   sthāt

For further details see Karl Hoffmann, who lists the occurrences of injunctives and discusses their meaning in one of the most impressive publications among recent works on IE (1967).

As Panini recognized, the injunctive is not a distinct inflection comparable with the present indicative, optative, imperative, and so on of Vedic. Rather, injunctive forms are survivals of the PIE verb system at a time when its mode of formation resembled that of systems like that of Turkish.

Recent analysis has provided further insights into the earlier system. The verb endings in the inflections of the dialects and of late PIE are composites, as Indo-Europeanists have held for more than a century. After the authoritative statement by Rudolf Thurneysen (1885), it has been generally agreed that the primary endings mi, si, ti are composed of the personal endings m, s, t found in the Vedic injunctives, plus a suffix i (see Watkins 1969:45-46 for details). The discovery of Hittite, with its further such suffixes, and our greater understanding of PIE phonology, which resulted largely from the laryngeal theory, provided additional credence for the reconstruction of the earlier forms of IE verbal endings. I illustrate these by means of the first-person singular endings of the middle in Indo-Iranian; the data are taken from Warren Cowgill (1968:24-31).

On the basis of the cited forms, the Proto-Indo-Iranian endings for the first singular may be reconstructed as follows:

    Athematic   Thematic   Vedic Athematic   Thematic
Aorist indicative (secondary)   *-i   *-ai   náṃśi   vóce
Present indicative (primary)   *-ai   *-ai   yuñjé   bháre
Subjunctive   *-āi   *-āi   stávai   pṛcchai
Optative   *-īya   *-aiya   bhakṣīyá   voceya

By comparing endings in other dialects and by internal analysis of the Indo-Iranian endings, the PIE form of the optative ending may be reconstructed with the suffixes:

*-yeʔ- + x + o
optative   first-person   middle
suffix   suffix   suffix

The subjunctive ending is also a composite of three suffixes, followed by the primary suffix i/y:

o + x + o + i(y)
subjunctive   first-person   middle   primary
suffix   suffix   suffix   suffix

Such reconstructions, which are based on the work of many Indo-Europeanists who have examined the forms solely on the basis of morphological evidence in the dialects, illustrate that the endings found in late PIE and the dialects are reflexes of sequences of suffixes similar to those in Turkish and in other OV languages.

In the early PIE sequence of verbal suffixes we distinguish suffixes for the deictic category of person and for number. We also distinguish suffixes for the Q categories. These suffixes and the inflections in which they are used have been given traditional labels in IE grammars, such as middle, optative, and subjunctive. It is one of the obligations of a syntactic analysis of PIE to describe the functions of the suffixes so labeled in accordance with a long and idiosyncratic tradition and to determine the categories in accordance with general linguistic theory. Before we do so, however, we will note an important difference between the verb system of OV languages like early PIE and that of languages like Greek, Sanskrit, and the other dialects, for which we posit compound endings.

In the verbal system of OV languages, each affix is to be regarded as an independent segment which may or may not be employed in a given sentence. It would therefore be misleading to equate Vedic -eyam, the equivalent in the active of Vedic -īya, with Turkish -eyim, even though the meaning of a specific form like Turkish geleyim may be equivalent to that of a specific Vedic form like gácheyam. For -īya and -eyam are composite units. Turkish -eyim, on the other hand, is an aggregate of suffixes to which further suffixes may be added, as illustrated above with negative and interrogative forms of the Turkish optative. The Japanese forms given in Chapter 1 may also be compared as illustrations of the agglutinative structure of verbal affixes in an OV language.

Moreover, it would also be misleading to equate the meaning of the Vedic optative suffix with that of the Turkish equivalent. For the Vedic optative endings are composites with compound meanings that contrast as composites with other endings of the Vedic verb, such as the subjunctive, imperative, and indicative. On the other hand, each of the Turkish suffixes adds a distinct meaning to the verb form, as we may illustrate with an example of the Turkish potential -bil- (which is suffixed to the gerund form -e- of a root):

1. Insan bunu bilmiyebilir
  person this know-not-potential-suffix

The sentence means: ‘It is possible that a person doesn't know this’ (Jansky 1954:128). If the suffixes formed a composite unit with a compound meaning like the Vedic endings, we might expect the meaning ‘It is not possible that a person should know this.’ For, in the composite endings of PIE, one element affects other elements as well as the stem. The composite structure and the compound meanings of IE verb endings have had a profound effect on the syntax of late PIE and the dialects. For, in contrast with Turkish and many other languages, one element–person–has come to dominate other elements and also the entire clause, including complements. Before we note the effects of this dominance of person in the late PIE verb, we will review the major Q elements.

4.1.3. On the Identification of Q Elements.

A major problem in the identification of Q elements results from their variety of expression in language. This surface expression may be made in the inflectional system of a language, in the derivational system, in the lexical system, or by means of suprasegmental morphemes. The variety of expression has led to problems in analysis and terminology, as we may note from discussions of aspect.

When we consider the opposition between imperfective and perfective action, we find that Russian is a language in which the contrast is indicated by means of derivational processes in the verbal system, as in:

Imperfective smotret′ ‘see’
Perfective posmotret′ ‘look’

In English the contrast between these two categories is expressed lexically, as the glosses ‘see’ and ‘look’ may indicate. In Homeric Greek the contrast was indicated by means of inflection, with the present indicating imperfective aspect and the aorist indicating perfective aspect.

As another example we may cite the feature Caus., which in Arabic is indicated by inflectional processes in the verbal system. As the following examples may indicate, one of the causative conjugations also carries the feature Vol. Using the symbols F, M, L for the first, middle, and last consonants of the typical trisyllabic root, we may illustrate the inflectional modifications as follows:

FaM a La (simple form), e.g., hasuna ‘be good’

FaMMaLa (volitional, causative), e.g., darrasa ‘teach, instruct’
ʔaFMaLa (unintentional agency), e.g., ʔajlasa ‘cause to sit down’

In English, on the other hand, the causative feature is expressed in a few nonproductive derivational pairs, such as set versus sit, lay versus lie. But generally it is expressed by means of phrases, such as those in He made her sit down; He had the eggs put in the basket.

Discussions of the features concerned and of the means of expressing them abound in earlier work on language as well as in recent studies. Moreover, some of the features concerned have attracted the interest of logicians and have led to intensive attention to modal logic. This is not the place to comment on the current state of modal logic, which by one definition “has to do with the structure of statements which are in grammatical moods other than the indicative” (Snyder 1971:2). This definition is obviously based on the perspective of IE languages. Clearly a great deal of work will be necessary on the “modal” elements of language, by linguists as well as logicians. Instead of commenting directly on such work, at this point I am merely interested in coming to terms with some of the earlier linguistic discussions.

The Q features labeled Perf. and Imperf. are among those most thoroughly studied by linguists. When such features are expressed in the inflectional system, they have been referred to as aspect. When, on the other hand, they are expressed lexically, they have been referred to as mode of action, Aktionsart. The complexities of the features, coupled with the varying possibilities of expression, make their treatment one of the most difficult problems in the study of language, even without such terminological complications, which are based on surface phenomena. For, if we regard only the simplest portion of the problem, that is, the means of expression, we find that even a language like Russian with well-developed means for indicating aspect, also makes the distinction lexically, as in the contrast between:

Imperfective govorit′ ‘say’
Perfective skazat′ ‘tell’

Moreover, any verbal prefix affects the aspectual meaning in Russian and in earlier languages with systematic means of expression for aspect, such as Homeric Greek and Gothic. Discussions of the problems involved have been not only voluminous but also frequently at variance with one another.

For improved understanding of aspect and other Q elements, we must deal with them as underlying elements rather than with their surface means of expression. The work in modal logic is promising, because logic is essentially concerned with underlying elements. As one important insight, modal logicians have recognized that deictic categories like tense differ from other modalities, that, in the words of D. Paul Snyder, “time reference of statements ... seems not to be a modal matter at all” (1971: 7). For Snyder, “modal matters” include “alethic modality” (possibility, necessity, entailment, compatibility), “deontic modality” (obligation, permission), “epistemic modality” (knowledge, belief), and “intentional modality” (wishes, hopes, desires, intentions). These are only some of the Q features, notably those expressed by the “moods” of IE languages.

The recent attention to modalities in linguistic works has suffered from excessive preoccupation with English, as indicated by the relating of modalities to English function words: be, have, cause to, put, and so on. It has also suffered from the awkwardness resulting when these are regarded as individual sentences in a P marker. For besides requiring many intricate reduction processes, such as equivalent-NP deletion, the P markers must then equate the modality elements with the sentence structures of a given language. The treatment of modality features, or Q elements, will be greatly advanced by their study in a wide range of languages. It will profit further from the study of their analysis in earlier linguists' works, such as that of Sibawayyi. While such materials will be referred to here, the chief basis for further illumination of the Q features in PIE will come from an analysis of their roles in successive periods of the various early IE languages.

Before examining the individual Q features we may note briefly the inflectional forms for them in Vedic Sanskrit. In part to illustrate the richness of its inflectional system, the forms are limited to those of one root, yaj- ‘make an offering, sacrifice’, in one work, the Taittirīya Brāhmaṇa; they are taken from Ananthanarayana (1970a).

The modality in which the speaker is concerned with expressing a fact (declarative; epistemic) is marked by the indicative, yajati. The unmarked form is the injunctive, yajanta. For interrogative and negative modality, syntactic devices other than inflection are used.

For reflexive modality the middle is used, yajate. The moods other than the indicative express deontic, alethic, and intentional modalities. The imperative indicates the necessitative, yaja, supplemented in the first and third persons by the subjunctive, yajāti. The optative expresses the volitional, yajet. Further, a secondary conjugation has been developed in Sanskrit to express the intentional modality, the desiderative, iyakṣasi.

Aspects are expressed by the perfect, īje, to indicate perfective modality, as opposed to the imperfective indicated by the present system. The aorist, ayāt, conveys the further meaning of momentary, perfective aspect.

Inflections for the tenses may be mentioned here for the sake of completeness, though as mentioned above these are not included under modalities: imperfect ayajat and future yakṣyasi.

For additional modalities the forms are taken from William D. Whitney (1885): causative yājayati, intensive yāyaj-.

Among the other Indo-European dialects the inflectional forms for indicating modalities are fewer, as is well known. Greek is closest to Sanskrit in this respect. In the other dialects many of the modalities are expressed by means of lexical items, as are interrogation and negation even in Vedic, or by derivation, as is aspect in Slavic. The use of such devices is found also in the Prakrits, that is, that stage of Indic which is chronologically parallel with the forms we have from many of the other IE dialects. For early stages of PIE, however, inflectional elements were the primary elements expressing modality, as we note in the following sections.

4.1.4. The Declarative (Dec.), Interrogative (Int.), and Negative (Neg.) Qualifiers.

When using the Q element Dec., the speaker is primarily concerned with expressing a fact, without wishing “to exert an influence on the will of his hearer directly through his utterance” (Jespersen 1924: 302). Its surface indication is frequently suprasegmental, as in English, with the intonation contour /(2) 3 1 #/ in contrast with the nonfactual /(2) 3 3 ||/.

2. 2He is 3sleeping1 # vs. 2He is 3sleeping3 ||

In some languages, however, such as the VSO Squamish with č and the OV Lisu with , Dec. is indicated by means of a segmental element (Kuipers 1967:157, Hope 1972:158-161):

3. Squamish č-n ʔi'tut vs. ʔi'tut-an
Dec. I-sleep   sleep-I
‘I indeed sleep.’   ‘Shall I sleep?’

4. Lisu ása dye-a̪ vs. ása dye mà-u̪
Asa go-Dec.   Asa go not-Dec.
‘Asa is going.’   ása dye mà ø ása dye
‘Asa didn't go.’

Since PIE was an OV language, we would expect it to have a final suffix expressing Dec. This is the value we posit for the -i of the so-called primary endings.

Ever since Thurneysen's brilliant paper on the imperative (1885:172-180), scholars concerned with the earlier forms of verbal endings in PIE have assumed that -i must be identified as a distinct affix. The analysis is clearest when we compare the so-called primary present indicative endings of Vedic with those of the injunctive.

  Injunctive Indicative
2 sg. bharas bharasi
3 sg. bharat bharati

The injunctive has long been identified as a form unmarked for mood and marked only for stem and person. It may thus be compared with the simplest form of OV languages. By contrast the present indicative indicates “mood.” We associate this additional feature with the suffix -i, and assume for it declarative meaning.

Yet it is also clear that, by the time of Vedic Sanskrit and, we assume, late PIE, the injunctive no longer contrasted directly with the present indicative. We must therefore conclude that the declarative qualifier was expressed by other means in the sentence. We assume that the means of expression was an intonation pattern. For, in normal unmarked simple sentences, finite unaccented verbs stood finally in their clause, as did the predicative elements of nominal sentences; Delbrück's repeatedly used example may be cited once again to illustrate the typical pattern:

5. víśaḥ kṣatríyāya balíṃ haranti
  villagers to-prince tribute pay
  ‘The villagers pay tribute to the prince.’

Since the verb haranti was unaccented, i.e., had no high pitch, we may posit for the normal sentence an intonation pattern in which the final elements in the sentence were accompanied by low pitch.

We support this assumption by noting that a distinctive suprasegmental was used in Vedic to distinguish a contrasting feature, interrogation or request (Wackernagel 1896:296-300). This marker, called pluti by native grammarians, consisted of extra length, as in ágnā3i ‘O fire’ (3 indicates extra length). But a more direct contrast with the intonation of simple sentences may be exemplified by the accentuation of subordinate clauses. These have accented verbs, as in the following line from the Rigveda:

6. RV 8.48.2. antáś ca prā́gā áditir bhavāsi
inside if you-have-entered Aditi you-will-be
‘If you have entered inside, you will be Aditi.’

As the pitch accent on ágā indicates, verbs in subordinate clauses maintained high pitch, in contrast with verbs of independent clauses like bhavāsi. We may conclude that this high pitch was an element in an intonation pattern which indicated incompleteness, somewhat like the 232 | pattern of contemporary English.

Evidence from other dialects supports our conclusion that, in late PIE, Dec. was indicated by means of an intonation pattern with a drop in accentuation at the end of the clause. In Germanic verse, verbs of unmarked declarative sentences tend to occupy unaccented positions in the line, notably the final position (Lehmann 1956:74). Although the surface expression of accentuation patterns in Germanic is stress, rather than the pitch of Vedic and PIE, the coincidence of accentuation pattern supports our conclusions concerning PIE intonation.

The Q feature interrogation (Int.) was apparently also indicated by means of intonation, for some questions in our early texts have no surface segmental indication distinguishing them from statements, for example, Plautus Aulularia 213:

7. aetatem meam scis
  age my you-know
  ‘Do you know my age?’

Only the context indicates to us that this utterance was a question; we may assume that the spoken form included means of expressing Int., and in view of expressions in the later dialects we can only conclude that these means were an intonation pattern.

Questions have generally been classified into two groups: those framed to obtain clarification (Verdeutlichungsfragen), and those framed to obtain confirmation (Bestätigungsfragen — Delbrück 1900:259-271; see Jespersen 1924:302-305 for the labels proposed for the two kinds of questions). For both kinds we posit the feature Int. This feature accompanies statements in which a speaker sets out to elicit information from the hearer. It may be indicated by an intonation pattern, as noted above, or by an affix or a particle, or by characteristic patterns of order, as in German Ist er da? ‘Is he here?’ When Int. is so expressed, the surface marker commonly occupies second position among Q elements, if the entire clause is questioned. The Turkish sentences given above provide examples. If Dec. is not indicated segmentally, the surface expression for Int. stands closest to the end of a clause. In OV languages it would occupy final position, as does ka in Japanese and mi in Turkish. Such means of expression for Int. are found in IE languages, though not in the dialects which are attested earliest.

A notable example is Lat. -ne, which, according to Minton Warren “occurs about 1100 times in Plautus and over 40 times in Terence” (1881: 50). Besides expressions like egone ‘Me?’, sentences like the following occur (Plautus Asinaria 884):

8. Parasite: Aúdin quid ait?   Artemona: Aúdio.
you-heard-Int. what he-says     I-hear
‘Did you hear what he is saying?’     ‘Yes.’

Other evidence for a postponed particle for expressing Int. is found in Avestan, in which -na is suffixed to some interrogatives, as in kas-nā ‘who (then)?’; and in Germanic, where na is found finally in some questions in Old High German. Old Church Slavic is more consistent in the use of such a particle than are these dialects, as in chošteši li ‘Do you wish to?’ This particle is also used in contemporary Russian. In view of the lack of such particles in the earliest dialects, we may have to assume that they were introduced after PIE, as a result of influence from OV languages (see Schmitt-Brandt 1971:242).

Yet it is remarkable that the particle used to express Int. in Latin, Avestan, and Germanic is homophonous with the particle for expressing negation. Indo-Europeanists have been greatly concerned with determining the origin of this particle, whether it was related to the negative particle *ne or to the Vedic particle of comparison na (Delbrück 1897: 537–540; 1900:263). While we are only secondarily concerned with the origin of surface markers, it is noteworthy that the OV languages Turkish, with mi, and Quechua, with ču, use homophonous elements for the interrogative and the negative. It is not unlikely therefore that PIE ne of questions is the same particle as that used for the negative. As the interrogative particle, however, it has been lost in most dialects. Its loss is one of the indications that late PIE was not a consistent OV language. Rather than a postposed particle, intonation was used to express the Q component Int., as well as particles that were placed early in clauses, often initially. These will be examined below.

Indications of negation (Neg.) in the early dialects are also like the means of expression in VO languages. Neg., by which the speaker negates the verbal means of expression, commonly occupies third position in the hierarchy of Q elements. In Japanese, as illustrated above, it precedes the element for Int., e.g., yomanai ka ‘Does he not read?’ In Classical Hebrew the expression for Neg. follows that for Int. For PIE, however, we can only posit the particles ne and , neither of which is normally postposed after verbs; see however Example 70 below. Neg., as well as Int., is accordingly expressed by surface elements that indicate VO structure for late PIE. Unlike the means of expression for qualifiers that remain to be discussed, neither Neg. nor Int. was incorporated in the inflectional endings which are found in the dialects.

4.1.5. Reflexivization (Refl.) and Reciprocity (Recip.).

By Refl. the action expressed by a verb is made to refer to the subject of that verb, e.g., He saw himself. A further category, Recip., may be combined with reflexivization when two or more persons are involved; by Recip. the action of one person is made to reflect on another, e.g., They saw each other. In SVO languages both categories are commonly expressed by means of pronouns, as in English. In OV languages, however, and in VSO languages, they are commonly expressed by means of verbal affixes.

Quechua provides an excellent example, as in the following forms (Bills, Vallejo, and Troike 1969:73, 150), in which ku indicates Refl.; na, Recip.:

9. tapukusan
  ask Refl. Continuative (Cont.) 3 sg.
  ‘He is asking himself.’

10. qunakusayku
  give Recip. Refl. Cont. 1 pl. exclusive
  ‘We are giving [it] to each other.’

The order of these affixes is of interest. In Quechua the affix for Refl. occupies Position 9 of the thirteen “modal-suffix” positions identified by Bills, Vallejo, and Troike (1969:333-334). All of these precede Neg. and Int. as well as Dec., which Bills et al. label “the factual suffix.” Similarly, in VSO languages, the affix expressing Refl. follows the means for expressing Int. and Neg., as in the Classical Hebrew sentence:

11. lo: hit̄labbe:š
  Neg. Refl. dress
  ‘He did not dress himself.’

We account for the survival of the affix expressing Refl. and Recip. in PIE by assuming that, as the language became less consistently OV, the final Q expressions on V were placed preverbally or indicated suprasegmentally. The affix expressing Refl. and Recip., the suffix for the so-called middle, was however maintained until Refl. and Recip. came to be expressed in accordance with the patterns of SVO languages (Lehmann 1973b).

Examination of expression for the middle in the early dialects illuminates the method of indicating Q constituents in earlier stages of PIE, as we noted above in § 4.1.2. The suffix for the middle, o, was placed after the person suffix, but before the suffix expressing Dec., i. It is curious that this device was adopted again in Hittite, where -ri and -t(i) are facultative affixes of the mediopassive. Whether we should account for these characteristic Hittite suffixes as a continuation of the OV patterning of PIE or as an independent development resulting from the influence of OV languages in and near Anatolia is a problem for Hittite scholars. Whatever the explanation for them, the Hittite forms have supported the analysis of endings proposed earlier by Thurneysen—an analysis in keeping with the method of formation well known in OV languages.

As we indicated earlier in this chapter, not all elements are required with every finite verbal form. Thus Quechua may have simple verb forms consisting of verbal root and personal ending:

12. hamu - nku
  come they
  ‘They (will) come.’

13. hamu - sa - nku
  come Cont. they
  ‘They are coming.’

Verb forms may also be further extended:

14. hamu - sa - nku - ču
  come Cont. they Neg.
  ‘They are not coming.’

15. yača - či na - ku - λa - sa - nku
  know Caus. Recip. Refl. Delimitative Cont. they
  ‘They are only teaching each other.’

If at some time in the course of development of a language the means of expression for some of the Q categories are placed before the verb, as were those for Int. and Neg. in late PIE, the remaining categories may be expressed with frozen elements. The combination of these with verbal roots may then constitute a fixed paradigm, such as we find in Sanskrit (Whitney 1896:208-212). From the “schemes of normal endings” presented here we may note those for the third-person singular, active and middle:

Primary   Secondary   Perfect   Imperative   Subjunctive   Optative
Act. Mid.   Act. Mid.   Act. Mid.   Act. Mid.   Act. Mid.   Act. Mid.
ti te   t ta   a e   tu tā́m   at(i) ate   yā́t ītá
ātai   et eta

The compound morphemes of Sanskrit and the other dialects, which are referred to in IE studies as endings, incorporated a limited set of markers for features. Besides person and number, these represented voice (active as opposed to middle), mood (imperative, optative, subjunctive), aspect (perfect, present, aorist), and subsequently tense. This development of compound morphemes, rather than successions of relatively independent affixes as in Quechua, led to considerable changes in the IE syntactic system, as we note below, especially in § 4.2. But on the basis of intensive analysis, the earlier affixes can be identified from relatively transparent forms of the early dialects (Brugmann 1913; 1916; Meillet 1937; Watkins 1969; Cowgill 1968). The categories indicated by these elements will be reviewed in the following sections.

4.1.6. The IE Moods: Necessitative (Nec.) and Voluntative (Vol.).

Nec. indicates that the person involved in the action is required to carry it out. Examples may be given with Turkish meli/malı forms:

16. yapmalı
  ‘(he, she, it) must do’

17. oturmalısınız
  ‘you must sit down’

The PIE subjunctive may resemble in meaning an obligative (Obl.). Both of these modalities are included in modal logic under alethic modality. Yet each feature may be expressed separately in a language, as in Quechua, where Obl. is indicated by a periphrastic construction (Bills, Vallejo, and Troike 1969:139) and Nec., or the imperative, by an inflection (ibid.: 13 et passim). In spite of such possible differences, the imperative and subjunctive are here introduced through Nec.

Apart from unmarked forms in the second person, the imperative was expressed by means of a -u affix, as may be illustrated by Skt. bharatu, bharantu. The subjunctive was expressed by means of an affixed e/o, as in the first person *-o-x-o-y as opposed to the (primary) indicative ending *-x-o-y (Cowgill 1968:27; § 4.1.2). The first affix of the subjunctive form indicates the mood; the second, the person; the third, the middle voice; and the last represents Dec. The subjunctive merged with the imperative in Indic; the merger supports the assumption of a similar meaning. In other dialects it underwent other developments, partly as a result of the subsequent structure of the verbal system, partly as a result of phonological developments, such as the loss of laryngeals. These developments also affected the mood which expressed the volitional (Vol.).

By Vol. the subject expresses a wish. The affix in PIE was -yeʔ-, as in the Indic athematic optative first singular ending *-iya, from *-iʔ-x-o < *-yeʔ-x-o and the thematic *-aiya < *-o-yʔ-x-o. The developments by which the optative merged with other verb forms or was lost as a distinct inflection will be indicated below.

As we have noted above, wishes have been included under intentional modality, along with desires, hopes, and intentions. Expressions for desire are found in some languages, as in the Japanese desiderative, e.g., yomitai ‘he wants to read.’ Sanskrit too has an inflection for the desiderative, as in the form iyakṣasi noted above. The marker and the uses of the desiderative are similar to those for the future, as in Skt. yakṣyasi. The evidence is unclear for PIE, in which these affixes may have been sporadic. We may therefore view expressions for desire as a part of the derivational system in PIE and their subsequent developments in the dialects as features of the individual dialects rather than of the parent language.

4.1.7. Aspects.

If we use the term aspect for semantic features, disregarding their manner of expression in a given language, we may posit two basic categories: Imperf. and Perf. (or -Perf. and +Perf.). Use of Perf. indicates that the action is assumed to be completed. As a consequence, Imperf. commonly indicates incomplete action; yet it is more precise to state that, with an Imperf. expression, there is no implication that the action is completed. As Wolfgang Dressler noted in his capable monograph (1968:43), imperfective forms are predominant in negative sentences.

The contrast in aspect is accompanied by other connotations, which have been extensively discussed. Thus Perf. also carries the connotation of punctual action, or even totality (ganzheitlich). And Imperf. carries the connotation of linear, continuative, or durative action and lack of totality (nicht ganzheitlich).

Moreover, Giacomo Devoto has pointed out the quantitative connotation of aspect (1958:396-398). It is this connotation which Dressler has explored in his monograph on plurality. For continuative aspect may be associated with distributive or iterative meaning; it may also be related to intensive aspect, which in turn may imply emphatic aspect.

Support for the assumption of such interrelationships may be provided by noting the IE -ske suffix and its varied uses. In Hittite it is primarily iterative (J. Friedrich 1960:74-75, 140-141); though, as Friedrich illustrates with well-chosen examples, it may also have distributive or repetitive meaning. By Kammenhuber's analysis it is a durative-distributive (1969:220). In Latin it came to have a durative-ingressive meaning, and subsequently indicated inchoative action (Leumann 1965:298). In Greek the suffix did not acquire the prominence it had in Latin; a notable development, however, is its use to indicate iterative preterites in Ionic (Schwyzer 1939:706-712). With this use we may compare its development as an aorist suffix in Armenian (Meillet 1936:115). Aspect accordingly provides difficult problems for IE studies, in part because the inflections of the verb systems in the various dialects came to indicate tense distinctions rather than aspect, but largely because of the broad implications of the categories themselves.

For PIE we posit a separate inflection for perfective aspect—the form which has developed into the perfect in Sanskrit and Greek and into various preterite forms in other dialects. The perfect was characterized in PIE by special endings in the singular (-xe, -the, -e, as opposed to -m, -s, -t) and by o grade of the root. Besides their basic perfective meaning, forms inflected in the perfect indicated a state resulting from previous or completed action, as in the often cited example—Gk. oȋda, Goth. wait, Skt. veda ‘I know’—in contrast with the imperfective meaning of the present inflection, as in:

18. RV 2.35.2. kuvíd asya védat
certainly it he-will-understand
‘Gewiß wird er es verstehen.’ (Geldner 1951-1957:I, 321)

Since the perfect stood in direct contrast with the imperfective present, the basic aspectual contrast in PIE was between perfective and imperfective.

Perfective aspect was in turn distinguished for -momentary versus +momentary aspect. As the cited Vedic example (18) indicates, and as is suggested by the characterization stative for the perfect, its perfective aspect was -momentary. By contrast, momentary aspect was indicated by means of the so-called aorist. This was characterized by zero grade of the root and secondary endings. The contrasts may be illustrated by the three following examples from the Rigveda:

19. RV 6.24.9. sthā́ ū ṣu ūrdhvá (aorist)
stand at-once straight  
‘Stand up straight at once.’  

20. RV 10.27.14. tasthaú mātā́ (perfect)
has-stood-up mother  
‘The mother is standing.’
‘Die Mutter steht still.’ (Geldner (1951-1957:III, 167)

21. RV 8.19.10.
  yásya tvám ūrdhvó adhvarā́ya tíṣṭhasi (present)
  whose you straight for-religious-celebration you-are-standing  
  ‘For whose religious celebration you are standing erect.’
  ‘Zu dessen heiliger Handlung du aufrecht darstehst.’ (Geldner 1951-1957: II, 320)

As the further context indicates, in Example 21 the present form implies continued support for the celebrant; in Example 20, by contrast, the perfect indicates a completed action, without an implication of momentary activity. Example 19 simply implies a momentary, completed action.

We posit this aspectual relationship for PIE, with a hierarchy as follows:

Indications for aspect were made partly by suffixation, as for the perfect with its suffix -e and personal endings, partly by form of the stem, as for the aorist as well as the perfect. Details will be discussed in § 4.4. below for these and other aspects of PIE.

Tense was not indicated in the early IE verb system, but it came to be the prominent characteristic in the early dialects. As the use of the prefix e- may indicate, the prominence of tense must have developed in subgroups of late IE. Prefixation of the augment vowel also indicates that at this time there was a tendency away from an OV structure. The shift from an aspect system to a tense system is accordingly one of the developments that must be examined individually for sets of dialects like Armenian, Greek, and Indo-Iranian, or for Germanic and Italic. Although the changes then are topics for the history of the dialects rather than the parent language, they will be dealt with in outline below.

4.1.8. The Causative (Caus.) and the Passive (Pass.).

Causation was expressed by suffixes in the early dialects, and presumably in late PIE. Yet the predominant suffix, -éyo-, must be late because of its vocalism and because of its form in Baltic and Slavic (Brugmann 1904a:535). Further, this meaning of -éyo- attached to o grade roots may have developed from an iterative-frequentative, as in Gk. phoréō ‘I carry back and forth’ in contrast with phobéō ‘I came to flee’ (Delbrück 1897:118-119). As Brugmann states (1904a:536), the causative meaning is particularly prominent in Indic, Germanic, and Slavic.

Caus. was also indicated by an -n- affix, as is particularly notable in Hittite. Vedic forms also provide evidence for this meaning, such as invati ‘cause to go’ beside eti ‘go’. The exact function of these two causative suffixes remains to be determined, if indeed there is adequate data in our surviving texts. It may be that one was inherently volitional, like the Arabic Form II, e.g., darrasa ‘instruct’, and the other nonvolitional, like Arabic Form IV, e.g., ʔajlasa ‘make sit down’.

Among other problems to investigate is the relationship of each to the so-called Vedic passive, a present inflection with -yá- suffix. This form expresses an intransitive, stative meaning, especially in early stages of Sanskrit. Since “passives” are formed from causatives in -éyo- (Whitney 1885:238), the essential meaning of the -éyo- causatives may be non-voluntative.

Besides the possible further illumination of the IE verb system resulting from additional study of the causative formations, we may note from their restricted forms the similarity of the early IE verb system to that of OV languages. It is well known that causatives in -éyo-, like other derived forms, “belonged only to the present stem” (Brugmann 1904a:535). This restriction indicates to us that such suffixes were added to the basic root of the verb and that their formation was similar to that of the aorist and perfect, much as “modal affixes” in an OV language like Quechua are comparable to one another.

Only when the IE verb system developed a set of standard inflections and fixed endings were derived forms like those in -éyo- treated like simple stems. This is the situation well known from Classical Greek, Latin, Germanic, and the other late dialects, including Classical Sanskrit. By the last half of the first millennium B.C., the verbal system of all the dialects had developed considerably away from that of early PIE (Wackernagel 1926:170). We can understand the earlier system only by reconstruction from Vedic, Hittite, and Homeric Greek, from relic patterns in other dialects, and from comparison with verbal systems in other languages of OV structure.

4.1.9. Congruence and Deixis Categories.

Indications for congruence and for deixis differ considerably from those of Q categories. As one important difference, expressions for such categories are often lacking in verb systems, especially with respect to OV languages; Japanese for example has no inflectional indications of either person or number. Among the important differences when congruence and deixis are expressed is the lack of hierarchy with relation to the expressions of Q categories. Among the Q categories, as we have noted, expressions for Int. and Neg. are placed closer to the sentence boundary than are expressions for moods and aspects. The position of expressions for congruence and deixis is independent of that for any Q categories. This independence has obscured the analysis of some endings, as we will note below.

In PIE the deixis categories referred to as first, second, and third person were expressed in verb forms, as were the congruence categories singular, dual, and plural. Since the basic functions of these are well known and relatively uncomplicated, they will not be discussed further here. The subsequent role of person in the verb came to affect its entire system and to bring about one of the characteristic features of IE languages, as we will note in the following section.

4.2. The Subjective Quality of the Verb in lE.

Among characteristics of IE, in contrast with other languages, is the sharp distinction between the noun and the verb. Meillet states emphatically that the distinction is more precise in IE than in other languages (1937:187). As a further important characteristic, the IE verb is characterized by a feature which Ernst Lewy has labeled subjective (1942:24). This designation refers to the inclusion in the finite verbal forms of an indication for the actor or subject. Thus, though they lack indications of mood and tense, the injunctive forms in Vedic include indications of the person involved in the action.

Many languages, on the other hand, do not include expression of person in finite verb forms. For example, a Japanese verb form such as yomu ‘read’ may be used when the subject is ‘I, you, he, she, it, we, they’ or nouns. The same is true of other finite forms such as yonda ‘read (past)’. The Japanese system accordingly lacks indication of the subjective feature in its simple forms. Another contrast with the IE verb system may be illustrated by Basque, in which the verb is characterized by differing intransitive and transitive forms which are accompanied by different inflected forms of the subject. As the dominant characteristic of the Basque verb, Lewy sees the “passive feature (Fassung)” of the transitive. A further contrasting system is illustrated by Samoyede, in which Lewy sees as an essential characteristic the possessive feature of the transitive verb; for the transitive expression corresponding to ‘I cut’ is like that of the nominal possessive construction ‘my boat’, literally therefore, ‘my cutting’. As these examples may illustrate, the requirement that the acting subject be expressed in the verb form differentiates the IE verb system characteristically from that of many other languages.

The incorporation of the expression for subject has also had an important influence on the IE verbal system and on its categories. For the included subject came to dominate not only the verb and its categories, but also the entire clause; in this way it has also determined the form and meaning of embedded elements. This domination has modified especially the meaning of verbal forms like the subjunctive and optative, as well as nonfinite forms, as we will see below.

Before examining these effects, we may note that verbal systems are capable of undergoing fundamental modifications. Modern Hindi, for example, has a system more like that of Basque and Dravidian than like that of Vedic. And Hungarian has given up some characteristics of the Finno-Ugric system, apparently under influence of the neighboring IE languages, notably German. PIE itself includes indications that the verbal system reconstructed from the dialects was fundamentally modified at one point in its history. For a large number of “impersonal” verbs point to an earlier system in which the actor, or subject, is not expressed; for any given verb form, the subject then may have been as varied as for the Japanese verb. Thus the Latin verb paenitet means ‘there is woe for...’ rather than ‘he, she, it undergoes woe, is sorry’. For more precise specification a pronoun is necessary, for example, paenitet me ‘I am sorry’. Similarly the impersonal passives common in older Latin have no reference to person, as in line 273 from Plautus's Pseudolus:

22. Quid agitur Calidore? Amatur atque egetur acriter;
  What is-carried-out Calidorus is-loved and is-missed vehemently
  ‘How are things, Calidorus?’ ‘I’m in love, and in great agony.’

On the basis of such verbs, which are particularly prominent in reporting natural events, e.g., pluit ‘is raining’, we may conclude that at an earlier time the IE verb was not subjective.

In the course of time the incorporated subject dominated not only the meaning of the root but also that of the Q element, as we may illustrate with virtually any verb form. The Vedic optative first plural vánāmahai accordingly means ‘we want to’ plus ‘(we) win’, that is ‘we want to win’, By contrast the desiderative element in Japanese may merely dominate a further segment of the verb form, as in the form seraretashi of the formal written style. The simple desiderative, as in shitashi, may be translated ‘I want to do’. The desiderative suffix may be added to the passive, of which the simple form serareru may be translated ‘be done’. The passive also may have a potential meaning; because such a meaning suggests “politeness,” the passive may be used to refer to the action of the person addressed. Where the desiderative is made of the passive, as in seraretashi, the form is translated ‘I want you to do’. The desiderative accordingly dominates only the “passive” affix, not the entire verb form. A similar example of restricted dominance was given above for Turkish. In late PIE such a restricted field of dominance was found in only a few forms; in the dialects these were lost, or modified in accordance with the subjective principle.

As the following examples indicate, the moods in PIE are controlled by the person indicated in the finite form (Delbrück 1900:425-427):

23. RV 5.54.15.
idáṃ me maruto haryatā váco
this well for-me O-Maruts you-must-accept word
yásya tárema tárasā śatáṃ hímāḥ
whose we-may-pass through-strength 100 winters
  ‘Receive favorably this my word, O Maruts, through whose strength we would like to survive a hundred years.’
24. RV 9.101.9. ā́ bhara... rayíṃ yéna vánāmahai
hither you-must-bear... riches by-which we-want-to-win
‘Bring riches here, through which we wish to conquer.’

Examples of the dominance of the incorporated person marker over participles are also common, such as:

25. RV 4.2.3. antár īyase aruṣā́ yujānó
between thou-goest two-roans yoking
‘You travel among [them] after yoking your two roans.’

Nonfinite verb forms could be used in this way to maintain the subject of the dominant verb.

As a consequence of dominance by the subjective component of the principal verb, the inflected verbs of embedded clauses lost their independent qualifier meanings; the qualifier affixes then indicated subordination. This effect is clearest with regard to the optative and the subjunctive, which in the course of time were used largely in embedded clauses and lost the meaning found for them in independent clauses of the earliest texts (see Macdonell 1916:362-363).

Further effects will be noted in greater detail below, such as the well-known “sequence of tenses” in the classical languages and modifications of person in embedded clauses, e.g.:

26. I told him that he should come.
  From: I told him: (You) come!

At this point I have been primarily concerned with indicating the subject's dominance over individual verb forms and its consequences for the development of the verb systems in late PIE and the dialects. In view of these modifications we must determine the meaning of the PIE inflected forms from their uses in simple sentences, as we will do in the following section.

4.3. Modified Forms of the PIE Simple Sentence.

4.3.1. Declarative Sentences.

As we have seen above, the simplest form of the sentence in late PIE was characterized by normal word order and by an intonation pattern with final drop. Such a sentence could consist of a verb alone, as in the frequently cited expressions for natural phenomena (Brugmann 1904a:624-626):

27. Skt. vāti (cf. RV 4.40.3)
  ‘Blows’ = ‘The wind is blowing.’

Other intransitive verbs could also make up a complete sentence, as in the Hittite Muršilis Sprachlähmung (Götze and Pedersen 1934:4, § 3) with sentence connective nu:

28. nu nahhun
  ‘And I got scared.’

As Meillet and others have indicated, in PIE such sentences were complete and did not require a separate subject. When a “subject” was included, it is to be regarded as appositional to the subjective element of the verb form, as in Meillet's example:

29. mūgit taurus
  he-bellows bull
  ‘The bull is bellowing.’

When transitive verbs were used in sentences, an object was required. Next to sentences consisting of verb alone, object plus verb made up the simplest form of the sentence. The object could stand in any one of the oblique cases, as exemplified in Chapter 2. An example of a sentence with an accusative object is from Muršilis Sprachlähmung (Götze and Pedersen 1934:6, § 15):

30. nu GUD puhugarin únuu̯er
  and bull scape- they-adorned
  ‘And they adorned a scape-bull.’

If in addition a subject was specified, the sentence observed the order SOV, as in the next line (Götze and Pedersen 1934:6, § 16):

31. nu- šan DUTUŠI ŠU-an daiš
  Ptc.-Ptc. (my) sun = I hand put
  ‘And I placed my hand on it.’

Such simple sentences could have many surface varieties, among them coordinate expressions, as in the following sentence from a Strophic poem of the Rigveda. The object consists of two coordinate genitives; the verb kir is inflected in the intensive subjunctive.

32. RV 4.39.1. divás pṛthivyā́ utá carkirāma
sky earth and we-praise
‘We want to praise the sky and the earth.’

Whatever the varieties, the normal unmarked form of the PIE sentence has the order (S)OV.

With marked order the position of these sentence constituents is changed. A fine example may be taken from the Hittite Laws; in a sequence that has SOV order when the S refers to a specific kind of culprit, the order is OSV when S refers to an indefinite actor (J. Friedrich 1959:68, § 25):

33. takku SIG4 kuiški taiezzi
  if bricks someone he-steals
  ‘If someone steals bricks...’

Or the verb may stand in first position, as in:

34. RV 4.48.1. vihí hótrā
you-enjoy poured-offerings
‘Enjoy the libations!’

In such sentences the verb is often an imperative, as in Example 34.

Imperatives and vocatives may also make up a complete sentence or a complete clause, as is indicated by their accentuation in the Rigveda when they are embedded or when each is taken as a separate clause:

35. RV 4.9.1. ágne mṛḷá mahā́ṅ asi
O-Agni you-be-gracious great thou-art
‘Be gracious, Agni. You are great.’

Like other verb forms, imperatives may be accompanied by objects.

Besides the sentences consisting of a verb, with or without an object and an explicit subject, sentences could be expanded by means of optional noun phrases. These noun phrases could stand in any of the oblique cases, according to the sense of the sentence. Their position was determined by stylistic reasons. Because such expansions are similar to one another regardless of the case, only a few examples will be given here (J. Friedrich 1959:68, § 25):

36. takku šamanaz NA4HI.A kuiški taiezzi
  if from-foundation stones someone he-steals
  ‘If someone steals stones from a foundation...’

37. RV 4.16.5. ubhé ā́ paprau ródasī mahitvā́
  both up he-filled worlds with-great-size
  ‘(Indra) filled up both worlds with his great size.’

The Hittite example contains an ablative in addition to the object and subject; the Vedic example contains an instrumental. Examples with datives, locatives, and genitives might also be included here; they would merely provide further evidence of the variety of adverbial elements that could be included in simple sentences by the addition of noun phrases in oblique cases.

4.3.1.a. Nominal Sentences.

Equational sentences, in which a substantive is equated with another substantive, an adjective, or a particle, make up a second type of sentence in PIE. In previous treatments they have been called nominal sentences, and accordingly the term is maintained here.

Examples are as follows:

38. Old Persian (Kent 1953:116). adam Dārayavauš
I Darius
‘I am Darius.’

39. RV 7.12.3. tváṃ váruṇa
you Varuna
‘You are Varuna.’

40. Hittite (J. Friedrich 1960:117). attaš aššuš
father good
‘The father is good.’

41. RV 4.28.5. evā́ satyám maghavānā
thus true benefactors
‘Thus is the truth, O benefactors.’

As these examples illustrate, a verb is not required in such sentences. The sentence consists simply of a topic in first position plus a description or predicate in second.

The basic form of nominal sentences has, however, been a matter of dispute. Some Indo-Europeanists propose that the absence of a verb in nominal sentences is a result of ellipsis and assume an underlying verb ʔes- ‘be’ (Benveniste 1950). They support this assumption by pointing to the requirement of such a verb if the nominal sentence is in the past tense:

42. Hittite (J. Friedrich 1960:118). ABU.I̯A genzuu̯alaš ešta
my-father merciful he-was
‘My father was merciful.’

Yet, as we have noted, tense was not a feature of the verbal system of PIE; sentences like Example 42 are accordingly post-Indo-European. Time of the state or action could have been indicated earlier by particles. Accordingly I follow Meillet (1906-1908) in the view that nominal sentences did not require a verb but that a verb might be included for emphasis.

This conclusion may be supported by noting that the qualifiers which were found in PIE could be used in nominal sentences without a verb. As an example we may cite a Hittite sentence which is negative and imperative (J. Friedrich 1960:117):

43. 1-aš 1-edani menahhanda idāluš
  one to-another towards not evil
  ‘One should not be evil toward another one.’

Yet, if a passage was to be explicit, a form of the root ʔes could be used, as in the following:

44. RV 4.30.1. nákir indra tvád úttaro jyā́yāṅ asti
no-one O-Indra from-you higher not greater he-is
‘No one is higher than you, Indra, nor greater.’

In the course of time a nominal sentence required a verb; this development is in accordance with the subjective characteristic of PIE and the endings which came to replace the individual qualifier markers of early PIE. The various dialects no longer had a distinct equational sentence type. Verbs might of course be omitted by ellipsis. And, remarkably, in Slavic, nominal sentences were reintroduced, as Meillet has demonstrated (1906-1908). The reintroduction is probably a result of influence from OV languages, such as the Finno-Ugric. This phenomenon illustrates that syntactic constructions and syntactic characteristics must be carefully studied before they can be ascribed to inheritance. In North Germanic too an OV characteristic was reintroduced, with the loss of prefixes towards the end of the first millennium A.D. (Lehmann 1970a). Yet in spite of these subsequent OV influences, nominal sentences must be assumed for PIE.

4.3.1.b. Sentences Expanded by Means of Particles.

In addition to expansions by means of additional nouns in nonrequired cases, as illustrated above in Examples 36 and 37, sentences could be expanded by means of particles.

Three subsets of particles came to be particularly important. One of these is the set of preverbs, such as ā in Example 37. Another is the set of sentence connectives, such as Hitt. nu. The third is the set of qualifier expressions, e.g., PIE ‘(must) not’. An additional subset, conjunctions introducing clauses, will be discussed below in the section on compound clauses.

Preverbs are distinctively characterized by being closely associated with verbs and modifying their meaning. In their normal position they stand directly before verbs (Watkins 1964), as in the following line from an Archaic hymn:

45. RV 8.48.8. mā́ no aryó anukāmám párā dāḥ
not us of-foe desire away you-give
‘Do not desert us in accordance with the desire of our enemy.’

As in this example, the preverb is accented in independent clauses (Macdonell 1916:468-469); but in subordinate clauses it is unaccented, as is saṃ below:

46. RV 8.48.1.
víśve yáṃ devā́ utá mártiāso
all which gods and mortals
mádhu bruvánto abhí saṃ cáranti
honey saying about together they-come
  ‘About which all gods and mortals, calling it honey, assemble.’

In marked order, on the other hand, the preverb stands at the beginning of the clause:

47. RV 8.48.2. ánu rāyá ṛdhyāḥ
towards wealth you-thrive
‘Prosper us with wealth.’

Such constructions are also prominent in Homeric Greek and in Hittite, as in the following passage from the Hittite Laws (J. Friedrich 1959:76, § 56):

takku annaš TÚG-ZU IBILA-ši edi nāi nu-za-kan
if mother garment-his son-her away she-directs Ptc.-Refl.-Ptc.
DUMUMEŠ -ŠU parā šūīzzi
son-her forward she-pushes
  ‘If a mother throws out the garment of her son, then she expels her son.’

Preverbs in this way make up a distinctive syntactic combination with their verbal roots. As the translations given above indicate, for example that of Skt. párā dāḥ (Example 45), these combinations had specific meanings differing from those of the root. Moreover, they also developed distinct syntactic properties, as in the transitive combination (Example 47), when anu is used with the root ṛdh-, which normally has an intransitive meaning.

In the course of time these combinations were conflated to single words, as in Classical Greek and Latin and in the other late dialects. This development took place over a long period of time, as is evident from combinations like Gothic:

49. ga-u-laubjats
  ‘Do you two believe?’

in which the preverb is separated from the verb by the interrogative particle. The IE pattern has also left reflexes in Modern German, as in Vórmund ‘guardian’, OHG fóramundo, in which the preverbal element was accented in accordance with compound-noun patterning. In verliéren ‘lose’, on the other hand, the preverb is unaccented. By its vowel it indicates the Germanic and IE situation in which its etymon, maintained also in Hitt. parā (Example 48), Skt. prá, Gk. pró, and so on, was a separate word in Germanic until the time when a stress accent was fixed on the first syllable of words. And the German “separable-prefix” construction, as in:

50. Er liest vor. ‘He is lecturing.’

maintains the pattern in which the preverb had its own accent.

The PIE combination, in which each of the elements of the preverbverb syntagm was independent, though the combination functioned as a unit, required the classification of preverbs as a distinct class of particles and of sentences in which such combinations occurred as a distinct subclass. (See also Delbrück 1888:44-51, 432-471; 1897:103-109.) The subsequent history of these elements, when the preverb came to be attached to verbs in some combinations and developed independently to adverbs and to prepositions, varies from dialect to dialect and must be separately described for each.

While the syntactic patterning of preverbs is well understood, the role of sentence connectives is not. Delbrück had pointed out that particles like Skt. sá, nú, and were used as connectives in clauses (1888:215-216, 514-519). Moreover, a connective frequently introduces sentences. When Hittite was discovered, it was noted that cognates of these particles, nu, ta, šu, were consistently used to introduce sentences (J. Friedrich 1960:155-161). Although E. H. Sturtevant brought this use dramatically to the attention of Indo-Europeanists by proposing that the demonstrative pronoun, as in Skt. sás, sā́, tad, was a reflex of these particles combined with anaphoric pronouns, the syntactic significance for PIE of the sentence connectives was not widely discussed until Albrecht Götze and Myles Dillon linked the Hittite syntactic patterning with the use of sentence-initial particles in Celtic (Sturtevant 1942:26, Dillon 1947:15-24). Yet the syntactic patterning of such elements has not been precisely described, in part because we do not yet know how sentence connectives function in the various language types.

Sentence connectives are prominent in the VSO language Squamish (Kuipers 1967:154-187, 169). If sentence connectives are to be expected in VSO languages, we might look for a similar syntactic pattern in the VSO Insular Celtic languages, where they indeed occur. If typological features are to be expected in a specific language type, the sentence connectives and their patterning in the Anatolian and possibly even in the Indic languages may then have been borrowings. Accordingly an interpretation of the use of sentence connectives faces many problems. For the understanding of PIE syntax, the use of sentence connectives will have to be investigated in each of the dialects. See also the urging of Erich Neu (1970:61) concerning the need of investigations.

Carol Raman has carried out such a study for Hittite. She has concluded that in clauses with nu the same noun is topic as in the main or earlier clause. This function may be illustrated in a passage from Ein althethitisches Gewitterritual (Neu 1970:10, 6). After the introduction of the subject, with a modifying clause, three verbs without a specified subject are introduced by nu (or nu + e ‘they’), as follows: “The UBĀRU people and whichever lord happens to be sitting before the king...”

51. ne šara tienzi nu appa tienzi ne aranda
up they-step   back they-step   they-remain-standing
  “they get up, they step back and remain standing.”

This function of nu may have developed because the topic noun is commonly deleted from the matrix clause and kept in the embedded clause, as here. An English chronicler would phrase the above sentence: “The UBĀRU people and the lord who happens to be sitting before the king get up, step back, and remain standing.” For the understanding of this pattern in Hittite syntax we may note that nu in this use becomes much more frequent in the later texts (J. Friedrich 1960:157).

Johannes Friedrich also points out (1960:157-158) that nu is not used in certain clause and sentence patterns, for example, at the beginning of a new section, in commands and exclamations after negative commands, in emphatic sentences, in parenthetical supplementary clauses, and in the somewhat similar clauses indicating state (Zustandssätze). In all of these patterns there is characteristically a shift in the topic; accordingly the absence of nu would support Raman's conclusion. Furthermore, it is tempting to see in the expanding use of nu in later Hittite an influence from neighboring VSO languages, such as the Semitic.

If we account for the use of nu in this way, we still need to consider the comparable use of ta and, less frequently, šu in the older language (J. Friedrich 1960:161). These are the particles that seem to be cognate with Indic ta and sa and thus led Sturtevant to propose his etymology of the demonstrative *so, sā, tod. There is no doubt about the occurrence of these Hittite sentence connectives; but both are far less frequent than is nu in the later texts. Since however the particles are used similarly in both Hittite and Vedic, we may well assume etyma for them in PIE and the use of sentence connectives in some of its sentences. But such use was probably similar to that noted by Raman for Hittite; sentence connectives accordingly were found in only a small number of sentences. Such sentences are not to be characterized as a specific subclass; the sentence connective in PIE should rather be classed with conjunctions.

We are then left with the problem of accounting for introductory particles like nū́ in Vedic (Delbrück 1888:514-515) and the sentence connectives that Dillon and Götze noted in Old Irish (Dillon 1947). Vedic nū́ simply introduces clauses, with the meaning ‘now’. From such a use the Hittite pattern might well have been expanded, and that of Old Irish no as well (Watkins 1962:113-114;1963:10-17). As noted above, “empty preverbs” to which enclitic pronouns are suffixed may well be characteristic of VSO languages. Such an assumption may be supported by a remarkable parallelism in the use of the Squamish “clitics /namʔ/ ‘go’ and /mʔi/ ‘come’” and forms of Hitt. pāi- ‘go’ and uu̯a- ‘come’ (Kuipers 1967: 207; J. Friedrich 1960:159); in both Squamish and Hittite these verbs stand initially, functioning like particles in that other particles may be affixed to them. The occurrence of sentence connectives is then a problem for general grammar as well as for the grammar of PIE.

The particles which signal Q characteristics, like the interrogative and negative, must however be regarded as indicating specific clause types. These clause types will be discussed individually in the following sections.

4.3.2. Interrogatives.

As noted above, § 4.1.3, interrogative sentences could be distinguished by intonation alone, as in Example 7, or by the use of an interrogative particle or other interrogative words, as in Example 8. The device of order was apparently not used to indicate questions in PIE. But both other devices were used in each of the types, clarification questions and confirmation questions (see Delbrück 1900:259-288).

Clarification questions generally have the interrogative pronoun or particle in initial position, as in the following Hittite example (J. Friedrich 1960:147):

52. kuit apāt
  what that
  ‘What's the purpose of that?’

Initial position is the general pattern in the dialects; but in marked patterns the interrogative word could stand elsewhere, as in the following Strophic passage:

53. RV 4.5.8. pravā́cyaṃ vácasaḥ kím me asyá
to-be-reported of-word what to-me of-it
‘What of this word is to be reported to me?’

This may also be the reason for the position of the interrogative in the following example (J. Friedrich 1960:147):

54. nu namma kuit
  Ptc. still what
  ‘What is there in addition?’

More than one interrogative word could also be used in clarification questions, as in the following question (see also Delbrück 1893:511; 1900:256-260):

55. Odyssey 1.170. tís póthen eis andrôn?
who whence are-you of-men
  ‘Who among men are you, and where are you from?’ (= ‘Who is your father?’)

This pattern, found also in Polish (Wachowicz 1974), is not possible in many IE dialects, including English. In general, clarification questions have one interrogative word and this is initial in its clause, as also in:

56. Odyssey 1.158. kaí moi nemesḗseai
? now with-me you-will-be-angry
‘Will you be angry with me?’

Confirmation questions also have an initial or near-initial interrogative, unless they are without an overt means of expression, as in the following Hittite example (J. Friedrich 1960:146):

57. ŠEŠ-IA-za malāši
  brother-my-Ptc. Refl. you-consent
  ‘Are you agreed, brother?’

As noted above, § 4.1.4, specific particles are found in the various dialects, such as nu in Sanskrit, or in Greek, ne, nonne, and num in Latin, u in Gothic, li in Slavic, and so on.

58. John 9:35. pisteúeis eis tòn huiòn toȗ theoȗ
  Gothic. þu ga-u-laubeis du   sunau gudis?
you be-?-lieve   in-the son of-god
‘Do you believe in the Son of God?’

It has often been pointed out that these particles are located initially or in near-initial position as a result of a transportation rule (Streitberg 1920:161). The rule was proposed by Wackernagel in an influential article (1892:333-434); it will be discussed below in § 4.5.

In the later dialects, interrogative particles are associated with presuppositions. Greek ou presupposes a positive answer, as do nonne in Latin and niu in Gothic. Presumably this is also the force of the initial “negative” particle in Hittite (J. Friedrich 1960:146):

59. UL-u̯ar-an-kan tuētaza memii̯anaz kuennir
  "not" quotes-him-away from-you from-word they-killed
  ‘Didn't they kill him at your word?’

On the other hand, Gk. mḗ plus presupposes a negative answer, as do Lat. num and Goth. ibai.

60. Odyssey 9.405. mḗ tís seu mȇla brotȏn aékontos elaúnei
? no one your sheep of-mortals unwilling he-drives
  ‘It is scarcely true that some one of mortals is driving off your sheep against your will?’

Since the particles differ from dialect to dialect, it is difficult to argue that interrogative particles with specific presuppositions should be ascribed to PIE. The pattern may have been inherited, however, and the surface forms changed; for as Hittite natta (generally written as Akkadian UL) and Greek ou illustrate, such particles were commonly changed or new particles were introduced.

In addition to the simple questions illustrated above, PIE very likely had disjunctive questions. Delbrück gives examples with pluti (1900: 268); in Greek the particle ȇ/ēé was used, either between the two queries or initially in both.

61. Iliad 10.534. pseúsomai étumon eréō?
I-lie or truth I-tell
‘Should I lie or should I tell the truth?’

In a nominal sentence, only the predicate nouns are essential, as in:

62. Odyssey 1.226. eilapínē ēè gámos
revel or wedding-feast
‘Is it a drinking party or a wedding feast?’

In such disjunctive sentences the function of the particle has been argued. It has been interpreted as an interrogative particle. But for Delbrück it was simply a disjunctive conjunction (1900:268-269), and the interrogative itself was signaled by intonation. If so, the Greek examples cited here would be similar to Sanskrit disjunctive questions. Since Hittite disjunctive questions are also indicated by a disjunctive particle rather than by an interrogative particle, Delbrück's conclusion seems the correct one.

Apart from the final interrogative particles attested in Latin, Slavic, and Germanic, the indications for questions are characteristic of those in VO languages. In addition to these particles, only the particle movement rules suggest that a final interrogative particle may have been present in PIE, in keeping with other OV patterns. By the time of the dialects, however, the patterns for interrogation were comparable to those of most current IE dialects and consisted either of a characteristic intonation pattern or of interrogative words.

4.3.3. Sentence Negation.

The archaic IE dialects, such as Vedic with mā́ and na, distinguish between two sentence particles for the negative; on the basis of their forms in other dialects, such as Gk. mḗ and Lat. ne, these are reconstructed *mḗ and *ne. Hittite has a lateral in the first, , in accordance with the sporadic change of nasals to l, as in lāman ‘name’; Greek has ou, ouk of unknown origin rather than *ne. Despite the difference in surface form, in both dialects the functions of the negative pair correspond closely with those in Sanskrit. The change in form provides another illustration that syntactic patterns may be maintained although changes are introduced in surface expressions.

In some dialects, such as Tocharian B, Germanic, and Latin, the distinction between the two particles has been lost. Tocharian B has only , but in Tocharian A mar corresponds to PIE * and to *ne.

Elements may also be combined with the negative particles, for “strengthening,” in the words of Jespersen (1924:335), as in Hitt. natta, Skt. nēd, Lat. nōn, and so on. On the basis of an extended form with -i in several dialects, Lat. nei, nī, Lithuanian (Lith.) neĩ, Slavic (Slav.) ni, Delbrück (1897:524-529) assumes *nei for PIE. PIE would then have had three negative particles: *, *ne, and *nei, an emphatic form of *ne; *nei however must have been only a variant of *ne.

In negative prohibitions * was used, *ne (nei) elsewhere. In the Rigveda, the verb used with mā́ was an injunctive:

63. RV 3.33.8. étad váco jaritar mā́pi mṛṣṭhāḥ
this word O-singer (do)-not-away you-forget
‘Do not forget this word, singer!’

Besides its use with the other moods, could also be used with the injunctive; the injunctive then expressed a general statement (Hoffmann 1967:99-100):

64. RV 3.3.1. átha dhármāṇi sanátā dūduṣat
thus laws from-old not he-forgets
‘Thus he (Agni) never forgets the laws.’

Other dialects support the basis of the assumed distribution of the negative particles in PIE, even though they have verb forms other than the injunctive with reflexes of *.

Although the functions of the negative particles are clear, the position of the particles in the PIE sentence must be determined. In the dialects the negative particles are commonly placed before the verb (Delbrück 1897:521; Jensen 1959:188 for Armenian; J. Friedrich 1960:145), as in the following Tocharian A sentence (Krause 1955:38):

65. ñatsey-ac kumnäṣ
  distress-to not he-comes
  ‘He won't fall into distress.’

Preverbal placement is apparently archaic; it survives in Latin verbs like nōlo < ne vōlo, in Goth. nist, generally in Lithuanian, and in some Slavic forms. But in later texts of most dialects the negative particle stands at or near the beginning of the sentence, a placement pattern also found frequently in the archaic dialects:

66. RV 4.31.9. nahí ṣmā te śatáṃ caná rā́dho váranta ārnúraḥ
not-for at-all your 100 and-not giving they-check enemies
  ‘For they do not check your giving at all, not even a hundred enemies.’

Since the placement pattern varies in this way, we assume that placement in either of the two positions results from a late placement rule and that earlier the negative particle was postverbal, as may be expected in an OV language. K. J. Dover (1960:14) assumes such a placement rule for Greek, noting that it is observed more rigorously in the earlier texts.

Postverbal negatives are attested in Latin, as in the following negative question in Plautus's Curculio 423:

67. nostin?
  ‘Do you not recognize [it]?’

To this question Lyco replies:

68. Quidni noverim?
  ‘Why shouldn't I recognize it?’

Thus the negative particle is postposed after interrogatives as well as verbs, as also in Truculentus 723, where the response is simply:

69. Quidni?
  ‘Why not?’

Since it is precisely the emphatic < *ne + *i which is used in these passages, I would like to propose that the. “strengthening” element is that used to indicate declarative meaning (§ 4.1.4). The two elements would have contracted when *ne was postverbal and was followed in turn by the declarative *-i.

There is additional evidence for postverbal position of the negative. In Hittite, both negative particles can stand finally. Johannes Friedrich interprets such a use as emphatic (1960:146); this interpretation supports the proposed origin for *nei:

namma-ma-kan KUR URUHapalla kuenta-i̯a UL
for-however-down country Hapalla you-smote-and not
epta-i̯a-at UL (= natta)
you-took-and-it not
  ‘For you didn't however conquer the country Hapalla, and you didn't capture it.’

Schwyzer also gives examples of the negative particles in final position in Greek (1950:596-597); again the use is for emphasis.

Normally, however, in Greek and in the other dialects, the negative particles are placed preverbally; if negation were not expressed postverbally in OV languages, the small amount of evidence for it in the IE dialects would scarcely seem adequate to propose such a position for PIE. Because of this small amount of evidence, it may be that negation was indicated by a postverbal marker only in some negative patterns.

The various dialects have introduced further new negative patterns. Correlative patterns came to be expressed with special negatives, such as Gk. oúte and Lat. neque, rather than the repeated simple negative as in the Hittite sentence (Example 70) above. The privative syllable *ne, *n, which was confined to nominal compounds in PIE, came to be used more extensively; it was also used with participles, and eventually it could be prefixed to any word, leading to the situation in the dialects where there is no distinction between word negation and sentence negation.

In the dialects sentence-negative particles were typically placed early in the sentence, though after interrogatives, as the second question from the Odyssey (Example 72) illustrates:

71. Odyssey 3.249. poȗ Menélaos éēn?
where Menelaos he-was
‘Where was Menelaos?’

72. Odyssey 3.251. ouk Árgeos ȇen Akhaiikoȗ
? not Argos he-was Achaean
‘Was he not in Achaean Argos?’

This is the normal position for negative particles in the dialects. When subjects were included in sentences, negative particles were placed after them:

73. Odyssey 3.313. Kài sú, phílos, mḗ dēthà dómōn ápo tȇl’ alálēso,
and you friend not long houses from far you-travel
‘Then don't, O friend, travel long far from your home.’

Negative particles could also be placed before the words to be negated:

74. Odyssey 4.64. epeì ke kakoì toioúsde tékoien
for not indeed lowly such they-might-beget
‘For lowly men surely wouldn't beget such [sons].’

In time placement was regulated by the specific rules in each dialect.

4.3.4. The Reflexive and the Reciprocal.

As indicated above, § 4.1.5, the categories for reflexivity and reciprocity are commonly expressed with verbal affixes in OV and also in VSO languages. Only in SVO languages can we expect to find pronominal forms for use in expressing these categories. The recognition that these are Q categories expressed in this way permits us to account for the IE middle inflection and for the development of pronominal means of expression as dialects came to be SVO in structure.

The functions of the middle were recognized by the Indian grammarians, and are well reviewed by Delbrück (1897:412-432). These functions are also found in Greek and in Hittite (Neu 1968:92-116). As an example of reflexive meaning, Johannes Friedrich cites unuttat ‘she adorned herself’; as an example of reciprocal meaning, šarrandat ‘they parted from one another’ (1960:135). Since Hittite in this way corroborated the uses in the other dialects, there can be no question of the basic uses of the middle in PIE. As noted above, it was expressed by an -o-affix. In Hittite, further middle suffixes were commonly added after the middle endings inherited from PIE: -ri, -ru, -di, -t (Sturtevant 1951:145-148). These are characteristics that we would expect of OV languages.

Basically it is the function of the middle to indicate that the verbal meaning, whether action or state, is to be interpreted with reference to the subject. Delbrück gives excellent illustrations (1897:412-432), as with forms of the Sanskrit root yaj- ‘sacrifice’:

75. RV 4.24.5. ā́d íd dha néma indriyáṃ yajanta
then indeed Ptc. some Indra's-power they-sacrifice
  ‘Then some sacrifice to the power of Indra for their own benefit.’

Active forms indicate that an action, such as a sacrifice, is being carried out with no special reference to the subject; middle forms indicate that the action has some reference to the subject. Depending on the context, this could be reflexive:

76. RV 1.134.3. vāyúr yun̄kte róhitā... ráthe
Vayu he-yoked the-two-tawny-ones... to-chariot
‘Vayu yoked the tawny pair to his chariot.’

The meaning in other contexts is reciprocal:

77. RV 10.88.17. yátrā vádete ávaraḥ páraśca
where they-speak lower upper-and
‘Where the upper and lower speak to one another.’

In such examples, the function of the middle is like that of reflexive and reciprocal forms in a language like Turkish.

Some roots in PIE came to be used generally or exclusively with middle endings, e.g., Skt. ā́ste ‘he sits’, Gk. hḗstai ‘he sits’, in contrast with forms of PIE sed-, which characteristically has active endings, as in Skt. sádati.

78. RV 7.11.1. ny àgne hótā prathamáḥ sadehá
down Agni priest first sit-here
‘Sit down here, Agni, first as priest.’

Agni clearly is not asked to sit for his own benefit. But in another hymn of Book 7, priests who wish to get benefits from Indra are described as sitting at the soma sacrifice like bees or flies on honey:

79. RV 7.32.2. mádhau mákṣa ā́sate
on-honey like flies, bees they-sit
‘They sit [for their own benefit] like bees on honey.’

Roots which came to be inflected consistently in the middle, such as PIE *key- ‘lie’, but with “active” meaning, are referred to as deponents, notably in Latin and Celtic grammar, but also in Hittite (J. Friedrich 1960: 135). As the handbooks inform us, the middle meaning of deponents and of roots inflected only in the middle was eventually lost; Skt. śéte and Gk. keȋtai scarcely mean ‘he is lying down for his own benefit’, but simply ‘he is lying down’. Similarly, Lat. sequor means ‘I follow’. The force of the middle came to be lost as the marker was no longer distinct; this development led to other innovations than that of deponents and of active meanings.

A development in all the dialects was the rise of reflexive and reciprocal pronouns. In PIE, as in many OV languages, there was no reflexive or reciprocal pronoun. In the dialects an adjective *sew(e)-, meaning ‘own’, was taken over as a reflexive adjective; it was also adapted as the basis for the reflexive pronoun. The lack of a reflexive pronoun in Sanskrit and the variation in forms from dialect to dialect leave no doubt about the late development of these forms. Statements in the handbooks ascribing a reflexive pronoun to PIE (Brugmann 1904a:409) must be revised. The late origin of the reflexive pronoun may also be demonstrated by noting its defective set of forms in the dialects, even in a contemporary dialect like German. A full paradigm never was developed. In the third person a single form, sich, is used without distinction in number or gender. In the first and second person, forms of the personal pronoun are used; on the basis of parallelism with these, a genitive seiner was developed for sich. The reflexives in the various dialects could be cited similarly to illustrate that reflexive pronouns and reciprocal pronouns are recent in the IE languages.

The shift from a special verb form, such as we would expect in an OV language, to reflexive pronouns and adjectives is also illustrated by an observation Delbrück made about changes in Sanskrit (1897:413). When sva- ‘own, one's own’ came to be used, the accompanying verb form might simply be in the active:

80. RV 1.46.9. sváṃ vavríṃ kúha dhitsathaḥ
own body where you-desire-to-place
‘Where do you intend to deposit your own bodily form?’

For this line Karl Geldner has an interesting note (1951-1957:I, 57), suggesting that the verb may be active but have a middle sense; the line would then ask the Aśvins: ‘Where do you intend to put on your natural bodily form?’

Another indication of innovations to indicate reflexivity is the use in Homer of the adjective philós, often to be translated ‘dear’, to convey a reflexive meaning (Benveniste 1969:335-353):

81. Odyssey 12.331. phílas ti kheȋras híkoito
own what ever hands it-comes-to
‘whatever might come into their hands’

This use of phílas also reminds us of the shift in syntactic order which necessitated the introduction of new elements. As we have noted above, an innovation was also introduced in the word for ‘not’, ou, ouk; a third notable innovation is found for the conjunction ‘and’, kai. It has not been determined whether these elements are native neologisms or whether they were borrowed. They are powerful indicators however of a shift in syntactic devices.

As a part of the syntactic shift, the passive was developed as a category contrasting with the active. The process has been well described. For some verbal roots the reflexive sense is close to a passive; a root like Skt. muc- ‘release’ would, as Delbrück illustrates (1897:432-435), scarcely be distinguishable as a reflexive, ‘gain a release with reference to oneself’, or as a passive, ‘be released’.

82. RV 1.31.4. śvātréṇa yát pitrór múcyase páry
with-strength when parents you-free-yourself (were-freed) about
‘As soon as you were freed from your parents with force.’

See also Neu 1968:109-116. The history of the development of the passive category must be pursued in each of the dialects. It must be recognized as distinct from the middle when an agent comes to be included regularly in its clause.

The middle itself was gradually replaced by constructions made up of verbal forms combined with reflexives and reciprocals, and by intransitives. Its basic meaning in PIE may have been weakened even in the earliest texts, as Neu has illustrated in his fine study. A Hittite sentence, such as the following with declarative marker -i and a direct object, may have been interpreted with little of the middle force:

83. Neu 1968:55. nan-kan hattešna GAM-anta hattari
Ptc.-it-Ptc. (down) in-pit downward he-stabs
‘And he stabs it downward into the pit.
(And he butchers the piglet in the pit [for his own benefit].)’

Middle forms were also accompanied by interrogative and negative qualifiers, but, since these functioned as they did with active forms, no examples will be given here.

4.3.5. Necessitative and Obligative.

Forms with necessitative modality indicate that the action should or must be carried out. The meaning is straightforward and transparent in positive sentences; in the negative, however, and in questions and dependent clauses, the meaning is less straightforward and accordingly more difficult to interpret. Of forms indicating necessitative modality, those in the positive of the imperative are the simplest to interpret. Imperative forms in the third singular are also most clearly distinguished from forms indicating declarative and reflexive modality, as in Sanskrit:

  Present Indicative Middle Imperative
  (Declarative) (Reflexive-Reciprocal)  
3 sg. bharati bharate bharatu
3 pl. bharanti bharante bharantu

In these endings, accordingly, there was a direct contrast between the -u of the imperative and the -i of the present indicative, as well as the -o(+ i > -e) of the middle.

In the other persons the surface forms are not parallel. The second singular of the imperative was distinguished by the bare root, as in Lat. fer, fac, dic, duc; by the stem, as in Skt. bhara, Gk. phére; or by the suffix -dhi; the second plural was distinguished by the suffix -te. Subsequently the imperative came to be restricted in most dialects to second singular and plural forms. But, in Vedic and in Hittite, forms indicating necessitative meanings are found for all persons and numbers. In Vedic the forms are subjunctive.

Indo-Europeanists have differed in their views on the meaning of the subjunctive in PIE, as on that of the optative. The differences may be due to the development of these moods in the various dialects. Any attempt to equate the uses of the subjunctive and optative in Sanskrit or Greek with the uses of their earlier forms in PIE will lead to serious misinterpretations. For both moods came to be used largely in subordinate clauses. In PIE, however, subordination was indicated by other constructions (§ 4.8). The meanings of the earlier forms of the subjunctive and the optative must therefore be determined from their uses in simple sentences of the early dialects.

Before examining these meanings with the aid of examples from Vedic Sanskrit and Homeric Greek, we may exemplify the effect of the subjective quality of the PIE verb. The dominance of the subject affected the entire sentence, as may be illustrated by means of a Vedic example of the subjunctive in a yad- clause:

84. RV 7.27.1.
índraṃ náro nemádhitā havante
Indra men opposing-one-another they-call-on
yát pā́ryā yunájate dhíyas tā́ḥ
so-that decisive he-accept prayers those
  ‘Men call on Indra in battle so that he will accept their prayers as decisive.’

In this example and in many other passages, the verb form and the pronoun or person of yunájate have been modified in accordance with the well-known rules of complementation in the early dialects. But the subjunctive retains its basic PIE meaning. For if instead of the subordinate clause introduced by yad the poem had continued with a direct quotation, this would have been obligative, comparable to:

‘Men call on Indra in battle, saying:
“You must accept our prayers as decisive.”’

In time the obligative meaning of the subjunctive came to be subsidiary to its function of indicating subordination.

The obligative meaning of the subjunctive has been maintained especially in the first person:

85. Odyssey 8.133. Deȗte, phíloi, tòn kseȋnon erṓmetha
hither friends the stranger we-must-ask
‘But come, friends, let us ask the stranger.’

This use, often called hortatory, is apparent also in the following disjunctive question:

86. TS táṃ vy àcikitsaj juhávānī3 mā́ hauṣā́3m iti
him Ptc. he-pondered I-offer (Subj.) not I-offer Ptc.
  ‘With reference to him he pondered: “Should I perform an offering or should I not?”’

It is significant that here and in similar passages in Greek, the negative * is used, for this is the negative expression found in hortatory statements (§ 4.3.3).

87. Odyssey 3.55.
  Klȗthi, Poseídaon gaiḗokhe, mēdè megḗrēis
  listen (Imper.) Poseidon Earth-holder and-not you-grudge (Subj.)
  ‘Listen, Poseidon Earth-holder, and do not be unwilling.’

The use of the same negative for imperative and subjunctive, and the sequence of the two mood forms, as in this passage, argues strongly in favor of assuming a comparable meaning.

This meaning is also found in questions:

88. Iliad 1.150.
pȏs tís toi próphrōn épesin
why anyone you, or Encl. willing with-words
peíthētai Akhaiȏn
he-should-be-persuaded (trust in) (Subj.) of-Greeks
  ‘Why should anyone of the Greeks willingly believe your words?
(In what way is it necessary that anyone...)’

The hortatory meaning may be clearest in the second person, as in the following passage in which the subjunctive follows an imperative:

89. RV 7.67.4.
ā́ vāṃ vahantu sthávirāso āśvāḥ
hither you-two they-drive powerful horses
píbātho asmé súṣutā mádhūni
you-two-should-drink with-us nicely-produced juices
  ‘Let the powerful horses bring you two here. The two of you must drink with us the nicely extracted soma juices.’

The imperative vahantu indicates a clear request, in keeping with its basic necessitative meaning. But the subjunctive also involves entailment; after the Aśvins have arrived, they are required to join in drinking the soma. There is no doubt that a hortatory use also existed in the third person:

90. RV 8.43.22. imáṃ naḥ śṛṇavad dhávam
this our he-must-hear call
‘He must listen to this our call.’

These examples then indicate a close relationship between the subjunctive and the imperative.

Subjunctive forms have also been compared with injunctives; but, as Hoffmann has demonstrated (1967:249-253), a note of purpose or will accompanies subjunctives, while injunctive forms in the first singular of an aorist stem express an action of the immediate future. The three subjunctives in the following passage illustrate this use:

91. RV 4.33.5.
jyeṣṭhá āha camasā́ duā́ karéti ( karā + iti)
oldest he-said beakers two I-must-make-Ptc.
kánīyān trī́n kṛṇavāméty āha
younger three we-must-make-Ptc. he-said
kaniṣṭhá āha catúras karéti
youngest he-said four I-must-make-Ptc.
‘The oldest said: “I must make two beakers.”
The younger said: “let us make three.”
The youngest said: “I will make four.”’

As the next stanza of the hymn indicates, the Ribhus actually made four beakers; accordingly, the subjunctive first singular karā of the third line means something like: ‘I think we must make four.’ The basic obligative meaning is accordingly apparent also in the first-person subjunctives.

From this use a future meaning could readily develop, as the translation of the third line above may indicate. Yet the use of the subjunctive after a past tense indicates that purpose is involved, not merely a future action (Hoffmann 1967:244-245):

92. RV 1.32.5.
áhan vṛtráṃ...
he-slew Vṛtra...
áhiḥ śayata upapṛ́k pṛthivyáḥ
serpent he-must-lie related-to earth
  ‘[Indra] slew Vṛtra; ...the serpent must lie flat on the earth.’

While interpreting upapṛk as closer in meaning to the root pṛc- ‘fill, fructify’, Hoffmann points out the necessitative meaning of the subjunctive, as does Geldner, in his note if not in his translation (1951-1959:I, 37). This passage may also serve to indicate how the subjunctive came to be used in subordinate clauses. In this use, which is prominent in the dialects, it is generally accompanied by a conjunction (Macdonell 1916: 355-360 gives ample examples, as do other grammars).

The obligative meaning that we propose for the subjunctive for PIE is apparent with middle as well as active verb forms. It is also found with interrogatives and with negatives (Macdonell 1916:354-355):

93. RV 2.29.3. kím ū vaḥ kṛṇavāma
what Ptc. Ptc. you we-must-do
‘What should we do for you?’

94. RV 6.28.3. tā́ naśanti
not they they-should-vanish
‘They must not disappear.’

The subjunctive and the imperative must then be reconstructed as forms having distinct meanings, as they are distinguished by means of distinct affixes: e/o for the subjunctive, u and other affixes for the imperative. In accordance with this analysis, the declarative suffix might be expected with some subjunctive forms. And, as is well known, the subjunctive may take primary or secondary endings (see Hoffmann 1967: 268, n. 4; 276; 278, n. 22; Brugmann 1916:524-526). It is noteworthy that the primary endings are particularly prominent in the middle in Indic (Brugmann 1916:533, 643), possibly because the reflexive-reciprocal meaning lends itself to an obligative use:

95. RV 10.108.9. svásāraṃ tvā kṛṇavai
sister you I-will-make
‘I will treat you as my sister [for my own benefit].’

Possibly because the subjunctive endings were so close in form to the indicative, the subjunctive was lost as a separate category in the dialects. Its endings were maintained only where they came to be especially prominent, as in Italic and Celtic (Brugmann 1916:539-542) and possibly in Germanic (Lehmann 1943b). In many dialects the characteristically dependent verb forms which developed were based on PIE optative forms as well as on subjunctives, for the optative forms were more clearly distinguished.

The subjunctive survived with obligative, necessitative meaning notably in the first person of the Sanskrit imperative and also of the Hittite. In Hittite the endings were reshaped, especially by the attachment of -u to the first singular: Act. -(a)llu, Mid. -haharu. In a similar way the declarative -i was attached in Sanskrit to the first singular, -āni, and to the second singular, -dhi, maintained as a simple dental in Hittite (Sturtevant 1951:142), as in it ‘go’. The Hittite first singular, -l(u), may therefore be equated with Skt. -n(i), with l corresponding to n, as in Hitt. lāman, Skt. nāman ‘name’. In Sanskrit and Hittite, then, a merged conjugation developed which corresponded to PIE necessitative forms. In the other dialects the imperative alone was maintained; obligative meaning was expressed largely by means of compound sentences.

4.3.6. Voluntative.

Especially in early Vedic, wishes were expressed with optative forms, particularly in the first person:

96. RV 7.71.1. vāṃ huvema
you-two we-want-to-call
‘We want to call you two [Aśvins].’

97. RV 1.92.8. úṣas tám aśyāṃ yaśásam... rayím
O-dawn that I-want-to-obtain glorious... wealth
‘Dawn, I would like to obtain that glorious wealth.’

98. Odyssey 13.42.   amúmona d’ oíkoi ákoitin
noble Ptc. at-home wife
nostḗsas heúroimi sùn arteméessi phíloisin
having-returned I-would-like-to-find with safe friends
  ‘On my return I hope to find my noble wife and friends safe at home.’

Wishes such as these are very similar to expressions of likelihood. Especially in Greek, the optative generally expressed possibility or likelihood, even in first-person forms:

99. Odyssey 8.216. prȏtós k’ ándra báloimi oisteúsas
first Ptc. man I-would-hit having-shot-arrow
‘I would be first to shoot and hit a man.’

Because these two meanings are found in the earliest texts of those dialects that alone have maintained the optative as a distinct formal and semantic category, the central meaning of the optative has been disputed.

Delbrück has contributed most to the view that the basic meaning of the optative was voluntative (1871:13); he was strongly opposed by William W. Goodwin (1893:373-389), who viewed the optative as “a weak future form” (1893:388). Yet both the potential and the “weak future” uses are found in subordinate clauses, or, as in Example 99, with a particle indicating potentiality: ke. These meanings may then be assumed to be secondary to the voluntative meaning, which is found in independent sentences, as illustrated above.

The development may be illustrated by the following passage:

100. RV 8.44.23.
yád agne syā́m aháṃ tváṃ
if O-Agni I-am (Opt.) I you
tváṃ ghā syā́ ahám
you Ptc. Ptc. you-are (Opt.) I
syúṣ ṭe satyā́ ihā́śíṣaḥ
they-would-be your true now-prayers
  ‘If, Agni, I were you, and you were I, your prayers would now be carried out.’
Without yád the sentence would mean: ‘Agni, I wish I were you...’ With yád it is potential, as for the most part in early Greek. For the potential use, with subordinate force, largely replaced the volitional meaning in Homer. The potential meaning is also evident in passages like the following in Vedic; only the accent on the optative gives formal evidence of subordination:
101. RV 7.71.2.
yuyutám asmád ánirām ámīvāṃ
you-two-ward-off from-us listlessness disease
dívā náktam mādhvī trā́sīthām naḥ
by-day during-night honeyed you-two-protect (Opt.) us
  ‘Ward off from us listlessness and disease, day and night, lovers of honey, you should protect us.’

The potential meaning of the optative is found in the statement of Eurymachus telling Telemachus that the gods will decide who will be king in Ithaca:

102. Odyssey 1.402.
ktḗmata d’ autòs ékhois kaì dṓmasin
possessions Ptc. self you-have (Opt.) and in-house
hoȋsin anássois
in-own you-rule (Opt.)
  ‘But you yourself may keep your possessions and you may rule over your own house.’

But the volitional sense is clear in the first optative of the two in the following examples:

103. Odyssey 1.47.
  hōs apóloito kaì állos, hótis toiaȗtá ge rhézoi
  thus he-perishes (Opt.) also other whoever such Ptc. he-does (Opt.)
  ‘So may anyone else perish who may do such things.’

104. RV 8.23.19. imáṃ ghā vīró amṛ́taṃ dūtáṃ kṛṇvīta mártyaḥ
this Ptc. man immortal messenger he-makes (Opt.) mortal
‘A mortal man should make this immortal his messenger.’

Because of its voluntative use in simple sentences of Vedic, as well as in Greek, I assume that Delbrück's analysis was correct, and I take the earlier form of the optative to be the major means of expression for the voluntative in PIE. This view receives support from the survival of optative forms in Lithuanian with imperative sense (Stang 1966:421-422).

Indo-Europeanists have wondered why the optative has secondary endings, inasmuch as primary endings came to predominate in the subjunctive. I assume that these simply reflect the age of the optative; it must have been a distinct inflection when the secondary endings were still standard. Moreover, since the optative did not express certainty, there was no basis for adding to the optative secondary endings the declarative affix i. This assumption that the optative is an archaic form has semantic support. It is the only inflection which may manifest the situation illustrated above (Example 1): that the affixes may be independent of one another. Thus the optative form of Example 103 may be interpreted: ‘I wish he would perish.’ Delbrück made this point clear in the first of his syntactic investigations (1871:13). It would be intriguing to speculate on the reasons for a first-person subject with the affix -yeʔ-. Here the point will only be made that this characteristic of the optative demonstrates its archaic character, as does the restricted use of secondary endings.

This conclusion leads to the question of why there is no evidence for the optative in Hittite. For an answer we may look to the characteristic suffix of the optative. In its full form this was -yeʔ-. On loss of -ʔ-, with compensatory lengthening, the suffix became -yē- when accented and -ī- when unaccented. Moreover, after the thematic vowel -o-, it became -i-. Reflexes of the three forms of the suffix are apparent in Skt. yā, ī, e, and in Gk. iē, ī, oi, with further changes. In Hittite the -ʔ- would not have contracted with the preceding vowel, and accordingly -ye- and -i- would have been the expected reflexes, possibly with further reduction (J. Friedrich 1960:27). Apparently these reflexes were inadequate to maintain a distinct paradigm. Hittite merged the forms for the necessitative and voluntative, maintaining only one mood other than the indicative, the so-called imperative. This preserved even more strongly the first-person forms than did the Sanskrit imperative, as well as the various meanings of the merging categories. In Hittite the imperative had voluntative and hortatory as well as necessitative force (J. Friedrich 1960: 139). When desired, the potential was indicated by particles, as also in Greek; moreover, particles such as yadi ‘if’, indicating subordination, are more frequent with the optative in the Brahmanas than in the Veda.

In questions the optative indicates potentiality, as in the following three lines, which have a subjunctive, an optative, and an injunctive in succession:

105. RV 7.86.2.
kadā́ antár váruṇe bhuvāni
when Ptc. within Varuna I-am (Subj.)
kím me havyám áhṛṇāno juṣeta
what my libation without-anger he-enjoys (Opt.)
kadā́ mṛḷīkáṃ sumánā abhí khyam
when mercy having-good-spirits about I-see (Inj.)
‘When should I come to be in close touch with Varuna?
What libation of mine would he enjoy free from anger?
When will I in good spirits perceive his mercy?’

It may be hazardous to make much of such contrasts; the subjunctive however indicates a specific point in time; the optative, a potential; and the injunctive, an event to be fulfilled whenever Varuna has mercy on the suppliant.

This passage may indicate the modifications of modal meanings in questions: the subjunctive indicates that the speaker is doubtful about an event; the optative, that the event is potential. These uses compare with those of the optative (and also the subjunctive) in dependent sentences, as in the following negative relative clause, embedded in a clause expressing a wish:

106. RV 8.48.10.
ṛdūdáreṇa sákhiā saceya
wholesome friend I-associate (Opt.)
who me not he-injures (Opt.)
  ‘I wish to associate with a wholesome friend, who would not injure me [when he is drunk].’

Since this became the primary use of the optative, in the dialects the potential meaning is most prominent. Both the meaning and the subordinate use are results of the kind of complementation found in dialects (see § 4.7).

These patterns result from the indication of subject in the late PIE verb and the increasing prominence of a system of tense. The tense system, like the aspect system indicated by the form of roots with attached affixes in PIE, is derived from categories in the proposition. These two categories will be discussed in the following section with other aspectual expressions.

4.4. Tense, Aspect, and Related Categories.

In PIE, tense and the time of the action were not indicated by means of verbal affixes. Indications of the time of the action were given by means of particles or adverbs or were implicit in the aspects of the verb forms. Sanskrit and Greek have preserved patterns in which particles indicate the time of action of the verb, as Brugmann (1904a:571-572) and Wackernagel (1926:47, 156-162) have pointed out. For example, the present middle of sac ‘accompany’ is used with purā́ in the following Strophic passage to indicate past time:

107. RV 7.88.5.
kvà tyā́ni nau sakhyā́ babhūvuḥ
whereto these for-us friendships they-have-become
sácāvahe yád avṛkám purā́ cit
we-accompany-each-other for without-enmity earlier Ptc.
  ‘Where have these our friendships vanished?
  For in the past we have been associating with each other without enmity.’

In these and other examples, such as the frequently cited line from the Iliad, the past time of the action is indicated by a particle, usually a cognate of purā́ and pró:

108. Iliad 1.70. hòs ḗidē t’ eónta t’ essómena pró t’ eónta
who knew those Ptc. being those Ptc. will-be before Ptc. being
  ‘Who knew the things happening now, those that will happen and those that have happened.’

Such evidence and the system of verbal forms indicate that tense was not a grammatical category in PIE.

4.4.1. Aspect: Imperfective versus Perfective, Momentary versus Durative.

Rather than tense, verb forms indicated aspect, that is, state of the action or process expressed by the verb. As pointed out above (§ 4.1.7), this characteristic of the PIE verb system may be determined most clearly in injunctive forms of Vedic; for the difference in verbal stem and endings indicates the difference in state of the action. We assume a basic contrast between imperfective and perfective aspect, and a further contrast between momentary and durative aspect.

The contrast between imperfective and perfective aspect was indicated either by means of affixes or by characteristic forms of the verbal root. The imperfective forms developed into the present forms in the dialects; the perfective developed into the perfect of late PIE, Sanskrit, and Greek. Because of this realignment of the verbal system, the earlier forms and their meanings have been debated at length, especially after the discovery of Hittite. For the Hittite mi- conjugation corresponds to the present of Sanskrit and other dialects; and the hi- conjugation corresponds to the perfect. Yet the correspondences of the Hittite verb system are not exact, either with the forms in the non-Anatolian dialects or with the conjectured aspectual meanings. For Hittite, like the other dialects, failed to maintain completely the system proposed here for PIE.

Evidence that the earlier form of the perfect was a “perfective present” has long been supported by means of forms like Skt. veda, Gk. oȋda, Goth. wait, Middle English (ME) wot ‘I know’; for as the shapes of these forms indicate, and as their cognate, Lat. vīdi ‘I have seen’, confirms, the Sanskrit, Greek, and Germanic forms are perfects of the extended root weyd- ‘see’. When the change from an aspectual to a tense system took place, these perfect forms of weyd- and such other forms as the preterite-presents of Germanic underwent a lexical rather than a grammatical shift (Lehmann 1943a:25). In Hittite, similar shifts took place when the mi- and hi- conjugations came to be parallel in indicating tense rather than aspect.

Yet the earlier distinction of meaning is still apparent in Hittite verbs which are inflected both in the mi- conjugation and the hi- conjugation (J. Friedrich 1960:106-108). One of these is šunna-, from the adjective šu ‘full’ (Kronasser 1966:563-564). Forms are found in the Hittite Laws, a hi- conjugation, third singular present šunnai (J. Friedrich 1959:46, § 96, 97), and a mi- conjugation, third singular present šunizzi (J. Friedrich 1959:74, § 51). The form šunnai refers to a single act of filling a receptacle with grain. The mi- present, by contrast, is translated ‘sows’, presumably from “filling” a field with grain; this meaning suggests that it refers to a continued action and is imperfective. Although the Hittite verb system requires far more study before we can claim to understand it in the various periods of Hittite, such details as agreement of the use of šunna- with perfective meaning and perfective forms provide welcome support for the widely held view that the IE perfect developed from a perfective present (see also Wackernagel 1926:150, based on Appolonius, second century A.D., and 166-168).

According to Louis Renou (1925:7, passim) the PIE perfect indicates a fulfilled state, in contrast with the present, which indicates a process that is under way. Wackernagel (1926:166-168) cites various groups of verbs which have a present meaning in Homer when inflected in the perfect: verbs indicating sounds, such as bébrūkhe ‘he bellows’; action of the sense organs, such as ópōpa ‘I see’; emotions, such as gégēthe ‘he rejoices’; gestures, sésēre ‘he grins’; and others fitting still other meaning classes, such as tethēlṓs ‘blooming’ (see also Munro 1891:31-32). As a common feature of the verbs, Wackernagel sees the intensity or repetition of an action. Yet each of the classes fits the characteristic meaning of the perfect proposed by Renou: ‘bellow, see, rejoice, grin, bloom’ indicate states resulting from previous activity. Since Wackernagel finds similar meanings in other dialects, he considers this the basic meaning of the perfect inherited from PIE. It is noteworthy that many of the Hittite hi- conjugation verbs correspond in meaning: halzāi- ‘shout’; au- ‘see’; uak- ‘bite’. The root ak- ‘die’ is also inflected in the hi- conjugation; compare Gk. téthnēke ‘he is dead’. The Hittite hi- conjugation accordingly provides semantic support for the assumption that it developed from a PIE inflection that indicated a state in the present resulting from previous action.

Hittite also provides formal support for assuming that the PIE h- inflection was a distinct paradigm indicating stative meaning. The characteristic perfect affix was -e. This affix corresponded to other affixes like the -o- of the middle. Moreover, the perfect had characteristic person affixes in contrast with those of the imperative present:

The Two Sets of Person Affixes
    m- Inflection   h- Inflection   h + Perfect
1 sg.   m   x   xa
2 sg.   s   th +e > tha
3 sg.   t   ø   e

To these endings the declarative -i could be affixed, as noted above for m s t. The combined affixes yielded the endings mi, si, ti, found in Hittite as well as Sanskrit and other dialects.

The second set of person affixes plus the perfect affix e, but without a final declarative affix, gave rise to the endings of the perfect forms in Greek and other dialects:

Person Affixes with i, e(a), e(a) + i, o
    Primary   Perfect   Greek   Hittite hi-   Middle (Secondary)
1 sg.   mi   *xa > a   *hai > hi   xo
2 sg.   si   *tha > tha   *thai > ti   so
3 sg.   ti   *e > e   *ai > i   to

Before examining further implications of these forms for PIE, we may observe that the a of the first and second singular xa and tha was modified from e because of contiguity with a colored laryngeals; -a then replaced -e in the third singular in pre-Hittite. The Hittite hi- conjugation endings have the declarative affix i added to these endings. This derivation for Hittite hi, ti, i was proposed by Bernhard Rosencranz (1953), and expanded by Neu (1968:125-131).

Although the m s t person affixes came to be generalized and as a result are viewed as the predominant characteristics for person in the IE verbal systems, it is important to observe that x rather than m was used in the first singular middle. The x th ø affixes must not then be interpreted as characteristic of the perfect or as restricted to the perfect; the middle endings in the Hittite hi- conjugation have the x th ø person affixes, while the middle endings of the other dialects show a combination of these with the m s t affixes. In both Sanskrit and Greek, the second singular secondary endings are based on -th; the Sanskrit first singular endings are also based on the first-person affix x.

Middle Endings
    Hittite hi-   Sanskrit   Greek
    Primary Secondary   Primary Secondary
1 sg.   hari/hahari   e i   mai mēn
2 sg.   ta/tati   se thās   (s)ai (s)o/thēs
3 sg.   a/ari   te ta   tai to

While this chart illustrates the wide use of the x th ø affixes, it also makes clear the modifications which the verbal endings underwent. The spread of the m s t person affixes is illustrated by the Greek primary endings; details for Greek and other dialects are available in the handbooks and are not pertinent to the basic argument presented here. The forms listed here are cited to illustrate the formal relationships between the PIE perfect and the middle, on the one hand, and the uses of the person affixes m s t and x th ø, on the other. The endings in this chart are found after the stage of the language when the middle with -o- and the perfect with -e- were competing forms, each with a characteristic meaning.

This situation of PIE is reflected in the Hittite lack of a distinction in the middle between mi- verbs and hi- verbs. In the Hittite present active there is a contrast in form and meaning as also in PIE; but the distinction is minimal in the preterite active and throughout the middle. Hittite accordingly has a verbal system similar to that proposed here for PIE, in that the middle contrasts on the one hand with the hi- conjugation and on the other hand with the mi- conjugation. In the other Indo-European dialects the systematization has proceeded further; the middle contrasts basically with the active, notably in Greek and Sanskrit. Subsequently other contrasts become more prominent, such as that of the passive versus the active. These subsequent contrasts are well known and well described and are of pertinence here only for the insights they provide on the increasing modifications away from the verb system of PIE.

The greater similarity between the Hittite system and that of PIE does not provide an argument in favor of the Indo-Hittite hypothesis; it merely reflects the greater antiquity of Hittite. Moreover, like the other IE dialects, Hittite underwent a progressive development away from the PIE system, notably in the tendency to merge the mi- and hi- conjugations. The tense distinction then came to predominate in later Hittite, as it also did in later Greek, Latin, Germanic, and other dialects.

The verbal system of early Hittite supports accordingly the conclusions based on analysis of the forms in other early dialects concerning a distinct perfect inflection with a characteristic e- affix in the indicative. In the other moods, the endings are those expected (Whitney 1896:292-295, Brugmann 1913:494).

The meaning of the perfect in PIE must be proposed on the basis of the thoroughly explored Vedic and Homeric texts, though the other dialects confirm the stative, resultative meaning, as in the Germanic preterite-presents. Support is also provided by the long-observed alignment between middle and perfect forms, such as the Greek perfect dédorka ‘I see’ and the middle dérkomai ‘I see’ (Wackernagel 1926:168). This alignment must be interpreted on the basis of the resultative implications of the two forms. The middle, as noted above, indicates that the result of action expressed by the verb has an impact for the subject (‘I see with some impact on my further action’; also ‘I see myself’).

109. Iliad 16.10.
  dakruóessa min poti dérketai óphr’ anélētai
  weeping but her towards she-looks (Mid.) so-that she-will-pick-up (Mid.)
  ‘Weeping [the girl] looks at her [mother] so that she will pick her up.’

Many examples of the perfect in Homer and the Vedic could be cited to indicate that an action has taken place with a specific result:

110. Iliad 22.94-95.
édu min khólos ainós
it-entered Ptc. Ptc. him anger terrible
smerdaléon dédorken
fiercely Ptc. it-gazes (Perf.)
  ‘And terrible anger entered it [the snake], and it gazes fiercely.’

Since both the perfect and the middle in this way have implications based on the result of an action, their forms show a natural relationship. But, apart from their relationship in sharing resultative meaning, they should not be more closely aligned, as if the perfect were a preterite to a middle present. On the contrary, the two inflections are parallel derivations from verbal roots.

The subsequent alignment of the aorist and the perfect also developed from similarities in meaning. As is commonly assumed in IE studies, an aorist should not be posited for PIE as a separate verbal category comparable to the aorists of Sanskrit and Greek. The aorist forms in these dialects have simply developed from PIE roots with punctual meaning. A characteristic shape of the root is in zero grade, as of wid- for the extended root weyd-. Suffixed only with secondary endings, the verb forms built on such roots had punctual, perfective meaning. In this way they contrasted with the forms of the perfect, which were -punctual or -momentary. The contrast may readily be noted in many occurrences in Homer, such as the perfect in Iliad 1.85 and the aorist in 1.148.

111. Iliad 1.85. eipè theoprópion, hóti oȋstha
you-say revelation which you-have-seen (Perf.)
‘Tell us the revelation which you know.’

112. Iliad 1.148.
tòn d’ ár hupódra idṑn
him Ptc. Ptc. from-beneath-the-eyebrows glancing (Aor.)
proséphē pódas ōkùs Akhilleús
he-addressed feet swift Achilles
‘With a scowl swift-footed Achilles addressed him.’

Since the distinction between the perfect and the aorist is so thoroughly established, further examples will not be given here. Moreover, examples are unnecessary to demonstrate that, like the perfect, the aorist came to be used to indicate past time when the shift from an aspectually based to a tense-based verbal system was carried out in the various dialects.

The independent status of the aorist was most clearly developed in those dialects in which the augment was used: Indo-Iranian, Armenian, and Greek. Slavic to be sure has an aorist; but it contains no perfect. The aorist in Slavic refers to events in the past; more than “40% of the attested OCS [Old Church Slavic] aorists are imperfective” (Lunt:1955: 136, based on Dostál). The aorist accordingly was maintained as a category comparable with that in PIE in only a small number of dialects.

Since in these dialects the augment must have had an effect on the meaning of the aorist, our most secure evidence for determining the meaning of the aorist in PIE will be taken from augmentless forms, like that in Example 112, or injunctives in the Vedas.

Aorist injunctives have as one of their prominent meanings a “resultative statement of confirmation” (Hoffmann 1967:214-235). A passage like the following confirms a situation which has resulted from previous action:

113. RV 6.19.2. índram evá dhiṣáṇā sātáye dhāt
Indra indeed Dhiṣaṇa for-gain he-placed
‘Dhiṣaṇa (has) designated Indra for gain.’

The resultative use of the aorist is found also in Greek (Wackernagel 1926:182-184). It illustrates how the aorist could have been amalgamated with the perfect, especially when used with the augment. For the “actual past” the augmented aorist is used in the Rigveda (Hoffmann 1967:219); between this use, that of the perfect in Latin, and that of the preterite in Germanic there was little difference. In each of these dialects the inherited aspectually distinguished forms were adapted for a tense system.

Presumably the resultative meaning led to the coincidence of s- formations with the aorist characterized by zero grade of the root. For the suffixed verbs with s that have survived outside the aorist have a resultative sense, such as Skt. dvéṣṭi ‘he hates’ (the nonsuffixed root is found in Gk. dei-, as in deídō ‘I fear’); Avestan yāstē, Lith. juóste ‘he girds’ (versus Skt. yuváti ‘he ties on’); Skt. úkṣati ‘he grows up’; Goth. wahsjan ‘grow’ (versus Lat. augēre ‘increase’). The s- suffix apparently came to be used with some roots as a marker for the aorist (Brugmann 1913: 390-392). As Delbrück concluded at the end of his examination of the various aorist forms in Sanskrit (Delbrück 1876:88), all the aorist forms were used similarly in Vedic. The surface affix s and its variants simply came to indicate the aorist category, in addition to the zero-grade form of roots. While the aspect system remained in force the aorist contrasted with the perfect primarily through its momentary meaning; and both the aorist and the perfect contrasted with the present through their perfective meaning.

The close association between the aorist and the perfect, as well as the structure of the PIE verb system, may still be demonstrated by means of the aorists in the Rigveda, which are based on perfect stems, as Hoffmann puts it (1967:64-65). Examples are paptas and vocas, as in the following passages:

114. RV 10.95.15. púrūravo mā́ mṛthā mā́ prá papto
Pururavas not you-die not Ptc. you-run-away
‘Pururavas, you shall not die; you must not run away.’

115. RV 6.52.14. mā́ vo vácāṅsi paricákṣyāṇi vocaṃ
not to-you words disregarded I-speak (Aor.)
‘May I not speak to you words that are disregarded.’

These forms demonstrate that aorists could have been based on presents in x th ø as well as on the imperfective presents in m s t. They also permit us to understand why the aorist did not develop as a separate category in Hittite and why it maintained itself in Indo-Iranian and Greek. In Hittite the perfective developed a separate conjugation parallel with the imperfective mi- conjugation. Each had a preterite. Each also had an imperative, which was used in preventive sentences; there was accordingly no need for a separate aorist, especially in preventive sentences—one of its important uses in Vedic. Like Indo-Iranian, Greek maintained an aspectual basis for its verb system. Tense was indicated largely by particles, notably the augment. The various aspectual forms were accordingly maintained in both of these dialects. Eventually however the category of tense—which in Sanskrit and Greek was secondary to that of aspect, as indicated by the separate imperfect in the present system as well as the augmented forms in the aorist and Greek perfect systems—came to predominate over the aspect system. The interplay between aspect and tense, which was manifested differently in each of the dialects, is one of the intriguing topics to pursue in their individual histories.

4.4.2. Secondary Aspects.

Similar to the imperfective aspects are indications of continuity. Expressions for continuative, iterative, distributive, and intensive aspect can be proposed for PIE. These expressions however are not as prominent as those for perfective and imperfective aspect, and accordingly they are here labeled secondary.

Dressler, noting Devoto's relating these aspects to quantitative expressions examined them under the topic plurality (1968:43). Viewing these aspects in this way does much to clarify their position in language and their treatment in grammar. Plurality of verbal action, as in Hitt. akkušk- ‘drink repeatedly’ versus eku- ‘drink’, must be viewed in the same way as plurality expressed by nouns, as in Hitt. watar ‘water’ (sg.), widār (pl.). Plurality may well be interpreted as a semantic feature. This feature can be applied to verbs as well as nouns.

Combined with imperfective aspect, plurality yields continuative aspect. With perfective aspect, it yields iterative or distributive aspect. When used with emphasis it indicates an intensive. Such an analysis provides an understanding of affixes which seem to have a diversity of meanings and even widely different developments in the Indo-European dialects. The most notable of these is the -sk- affix.

The -sk- affix (-ch- in Sanskrit) was identified by Delbrück as the marker of one of four characteristic present classes (1897:59-64). His characterization of its meaning as “terminative,” however, was not accurate and led him even in his brief discussion to difficulties of interpretation. One of the -sk- formations that must be posited for PIE is that of the root gwem- ‘come, go—move’; forms are found in Greek as well as Sanskrit. Among passages cited by Delbrück (1897:60-61) is the following:

116. RV 10.155.3. téna gacha parastarám
with-that go farther-off
‘With that go away in the distance.’

This imperative, used to a witch, can scarcely be interpreted as terminative. The same is true of the use of the equivalent imperative in Greek, even though Zeus is commanding the dream to go to the ships of the Achaeans:

117. Iliad 2.8. básk’ íthi
go there
‘Go away.’

Rather than terminative aspect, the Sanskrit and Greek presents in -sk- indicate continuative aspect, a meaning which developed from the earlier meaning still apparent in the Greek iteratives in -skon.

Delbrück discusses these iteratives at greater length than the presents (1897:62-64). His analysis of their meaning is remarkably accurate and is corroborated by the subsequently discovered Hittite iterative-durative in -šk- (J. Friedrich 1960:74-75). The iteratives in -šk- are of further interest because of the clarification they provide for the reconstruction of the PIE verb system. It is remarkable that they take secondary endings. Delbrück considers it a matter of chance that no present form pháskō was attested beside éphaskon/phásken ‘said’. Furthermore, he concludes from the Latin -sk- form of es ‘be’, escit, that Gk. éskon must once have had a present beside it. It is important however to account for the forms that are attested. The restriction of secondary endings cannot be accidental, whether the forms are accompanied by an augment, as in éphaskon, or not, as in phásken. These iteratives are simply reflexes of PIE roots affixed with -sk- which maintained the earlier system when the secondary endings were the basic person markers. In the course of time the -sk- forms were incorporated into the verb systems of the various dialects and came to express meanings which differed from the original iterative. Analysis of these is a matter of concern for the individual dialects and has been thoroughly discussed (Schwyzer 1939:706-712; Leumann 1963:341, 1965:298).

To understand the situation in PIE we must make use of the earliest evidence, such as that in Greek, notably Ionic Greek, and in Hittite. Used with an imperfective root, such as PIE gwem- ‘come, go, proceed’, -sk- indicated continuative action; used with a perfective root, such as PIE ṛ-, it indicated iterative, distributive action, as in Hittite ar- ‘reach’, arnu- ‘move to’ and the following Strophic passage:

118. RV 4.16.3. vṛ́ṣā yát sékaṃ vivipānó árcāt
bull when drink swallowing he-may-sing
  ‘When Indra may join in the song [from time to time], gulping in the soma.’

A further example of an -sk- form from this intriguing root is found in the Odyssey:

119. 14.23. autòs d’ amphì pódessin heoȋs aráriske pédila
he Ptc. about feet his he-was-fitting sandals
‘But he was fitting sandals on his feet.’

The PIE root (< γer-) meaning ‘reach, move’ came to have wide application in its suffixed form r-sk-, as in Vedic to ‘rays of light, peaks of thunder, bursts of song, grinding of the soma’; but basically it indicated a plurality of individual actions of the meaning of the simplex.

Such a meaning also could be interpreted as an intensive (Brugmann 1904a:493-494, passim), as in Hittite azzikk- ‘feast’ beside ed- ‘eat’. Intensive meaning was also expressed by means of reduplication, as in the Sanskrit secondary conjugation (Whitney 1896:362-372). In other dialects there was no distinct inflection for intensive action. We may assume that in PIE it was expressed by means of a plurality affix, whether reduplication or the -sk- suffix.

4.4.3. The Causative.

In a causative construction the subject of the verb brings it about that someone else carries out the action of the verb. Causatives may thus be accompanied by more than one noun, or by none if the aim of the action is understood; causatives must be analyzed in accordance with these patterns. In Japanese, for example, causatives may have no accompanying noun, or a noun indicating the aim of the action, or a noun indicating the subject of the action and another noun to indicate the aim (see Yamagiwa 1942:121).

120. Kodomo ni wa maiban ha o migakasenasai
  children by Ptc. nightly teeth Obj. Ptc. cause-to-brush
  ‘Make children brush their teeth every night.’

Such causatives have given rise to an analysis by which the causative is accounted for through raising to a sentence with a pro-verb, in a derivation which may be expressed as follows:

When the causative has a surface marker, this is affixed to the verb of the lower sentence, as is Japanese -se- to migaka-. The agent of the lower sentence is often expressed by means of an instrumental or an instrumental phrase, like kodomo ni. As in the English sentence below, however, the causative force may be expressed lexically, and the agent may have no special indication other than order.

121. Teach the children their lessons.

In such sentences the verb of the lower sentence is transitive (The children learn their lessons) in contrast with sentences having intransitive verbs (The children sleep).

In PIE, causatives were largely derived from intransitive verbs. The causative affix was -n-, which was infixed between a root and its affix, as in PIE γer-, γr-ew-, γr-n-ew-. An example of this root with causative affix is found in the following sentence:

122. RV 1.174.2. ṛṇór apó
stir-up waters
‘Stir up the waters.’

The causative force of the affix is especially prominent in Hittite, as in the form arnu- ‘move to’, from ar- ‘reach’:

123. J. Friedrich 1959:20, 19a. Éir-šet-pat arnuzi
house-his-Ptc. he-moves-to
‘He brings [him] to his own house.’

The previous example illustrates the dominant pattern for causatives in PIE, whether with the -n- affix or with the subsequently more widespread affix -eyo- attached to a root in o grade.

124. RV 5.5.10. tátra havyā́ni gāmaya
thither offerings cause-to-go
‘Make the offerings go there.’

When the verb of the lower sentence is transitive, the agent in the surface sentence is expressed in the accusative.

125. RV 2.37.6. uśán devā́n̐ uśatáḥ pāyaya havíḥ
desirous gods desirous cause-to-drink offering
‘Desirous one, make the desirous gods drink the offering.’

Subsequently, in the Brahmanas, the agent of the causative is expressed in the instrumental. It is this construction that is often considered the causative proper, as in the Japanese sentence given above (Example 120). If an expressed agent is required in causative constructions, early PIE cannot be said to have included a causative in its verb system. The -n- infix forms and the -eyo- forms may rather be labeled factitive. In Hittite this factitive use of the -n- infix is particularly prominent; for the suffix -nu- is used to make factitives from adjectives and nouns, Lg., tepnu- ‘diminish’ from tepu- ‘little’ (J. Friedrich 1960:74). Such verbs are similar to transitives, in Hittite as well as the other dialects. See Delbrück's fine comparison of Skt. r̥ṇóti ‘set in motion’ as opposed to íyarti ‘move’ (1897:21-22, 41) as well as his discussion of the -n- formations in general (1897:40-59). Saussure's view of the various -n- formations as resulting from one PIE class is now generally recognized (Schwyzer 1939:690-701), though with a slightly different morphological analysis (Lehmann 1952:24, Strunck 1967:25). We assume that the basic meaning of the -n- forms was factitive and that the causative became prominent only in late PIE and the early dialects.

It is to the absence of a causative proper that we ascribe the lack of a passive in PIE. A passive is a construction in which the agent of a causative is treated as subject. Inasmuch as such agents were not found in PIE, there was no basis for a passive. The forms labeled “passive” (Whitney 1896:275-277) are without agents in the early language and accordingly are basically intransitives, even in the late portions of the Rigveda, such as the wedding hymn:

126. RV 10.85.13.
aghā́su hanyante gā́vó
among-Aghas are-killed cows
’rjunyoḥ páry uhyate
among-the-two-Arjunas to she-is-led-home.
‘Among the Aghas cows are killed;
Among the two Arjunas she is brought home.’

The so-called aorist passive of the Vedas is also without an agent and must be viewed as a resultative even when used with an instrumental, as in the following line:

127. RV 5.1.1. ábodhy agníḥ samídhā jánānām
is-awake Agni by-kindling-wood of-men
‘Agni is awake because of the kindling wood of men.’

This statement is not a passive in the strict sense, as may be noted from Geldner's translation (1951-1957:II, 1): “Agni ist durch das Brennholz der Menschen wach geworden in Erwartung der Uṣas.” But through the use of instrumentals in such constructions, passives could develop, as may be even clearer in the following passage:

128. RV 3.61.6. ṛtā́varī divó arkaír abodhi
order-loving of-sky through-songs is-awake
  ‘The order-loving [goddess] of heaven is awake because of songs.’
  ‘The order-loving [goddess] of heaven was awakened by the songs.’

When agents as well as instruments come to be used in such constructions and in causative constructions, the passive must be included among the categories of the verbal systems. Its development has been discussed by various scholars and from various points of view (Schwyzer 1942, Gonda 1951a, Hartmann 1954).

Since we do not assume either a passive or a causative verbal category for early PIE, the place of the forms in -n- and of the “passives” discussed above must be noted. I assume that these forms are lexically derived. PIE was rich in affixes to carry out derivation. The affixes could be adopted as means of expression in the inflectional system, as was the s in the aorist and other affixes in individual dialects (k in Greek, v in Latin, dh in Germanic, and so on; Lehmann 1942:125-132; 1943a:19-26; see also Brugmann 1913:178-392). But, until they were incorporated in the inflectional system, they must be described as a part of the lexicon, much as are the nominal affixes and indicators of deictic relationships.

Affixes to indicate number and person are accounted for under lexical items, as are gender in participles and other nominal elements. These congruence and deictic categories will accordingly be discussed in Chapters 5 and 6.

4.5. Transportation Processes.

These processes, in the early dialects and in PIE, are notable especially for particles, pronouns, and verbs. Wackernagel has written the classical essay on the topic (1892); see also Dover (1960) for the situation in Greek. According to Wackernagel's fundamental conclusion, generally known as Wackernagel's law, the elements in question were unaccented and were typically placed in second position in the clause. Dover, labeling these elements q and elements like nouns M (1960:14), assumed for Greek the following arrangements: MqM, MMq; and, with more than one q, a preferred arrangement Mq(q...) M(M...). Since verbs characteristically stood in final position in the clause, as we have noted above, Wackernagel's law is the result of a transportation process, by which unaccented verbs were frequently placed in second position in the clause; other unaccented elements too were generally placed in this position. Wackernagel's hesitant suggestion (1892:427) that PIE had the same verb positions as does modern German must be rejected. Rather, it is to be assumed that verbs and other elements labeled q by Dover were sentence-final and subsequently shifted to an earlier position in the sentence.

In dealing with PIE it is one of the major problems to determine the elements that are transported to second position. Dover has proposed for his class q numerous subclasses of particles and pronouns, including also the verb eȋnai ‘be’ in its copulative sense (1960:12-14). He contrasts these elements with those of a group labeled p, which “never, or only in certain specifiable circumstances, end a clause.” Among the p elements are conjunctions like allá ‘but’, mḗ ‘lest’, relatives, articles, and prepositions. Prepositions, as Delbrück pointed out, were postpositives in the early dialects; and accordingly these p elements would have been q in PIE. The articles and relatives were not found in PIE; neither were many of the conjunctions. And as Watkins indicated (1962), the conjunction-like elements that must be assumed for PIE are to be viewed as sentence connectives, not conjunctions like those in Classical Greek and Latin. Among Dover's other elements we are left with “the simple negative,” which according to him “is not easy to classify,” but which he did “not treat ... as p” (1960:14). This examination of Dover's p class and the studies of Wackernagel lead us to conclude that at an earlier stage of the language the class of q elements was more comprehensive than it was in Greek, including also elements which corresponded to some of those labeled p for Greek by Dover.

Hittite provides evidence for including the negative particle among the q elements, as do other dialects; examples have been given above, § 4.3.3. As in the examples cited there, sentence negation is typically indicated postverbally in OV languages, as are indicators for the interrogative. Another remarkable agreement of PIE with OV languages is the use of the same particle for indicating interrogation and negation. For just as -ne is used in Latin as a final interrogative indicator (and the transported Gk. ou, Hitt. U-UL, etc.), so are Turkish -mi and Quechua ču, which is labeled a nonfactual suffix by Bills, Vallejo, and Troike (1969: 333). We may conclude therefore that, when PIE was strongly OV, interrogative and negative affixes were placed postverbally in the sentence. Further, when the order was being changed to VO, particles indicating these modalities were placed earlier in the sentence, often preverbally.

The shift in placement is particularly clear with relation to the negative. For indicators of it were placed before enclitic pronouns in the early dialects, in a pattern which has become more and more prominent, as is evident from languages like English, with contemporary forms none, nothing, nowhere, etc. The negated pronouns are more transparent in earlier forms of the language, as in Vedic nákis, Gk. oú tis, mḗ tis, Lat. nēmo, OHG nioman ‘no one’, and so on. The negative element ne was not used in compounding in PIE (Brugmann 1904a:310); n had this function. The presence of negated “pronouns” in the early dialects, with the pronominal element differing from dialect to dialect, e.g., kis in Vedic, *hemō, homo in Latin, man in Germanic, indicates that the negative movement was effected dialectally and not in the protolanguage. This conclusion is highly important for our understanding of the transportation processes in the late period of PIE and the early dialects.

The subsequent placement rules for the interrogative and the negative are a matter of concern for each of the dialects and will not be treated here. As Delbrück has pointed out (1897:521-524), the negative was placed particularly before verbs and before indefinite pronouns; some conflations, like Goth. nist < ni ist ‘is not’ and Lat. nōlo < ne volo ‘I don't wish’, indicate that these verbs are characteristically those to which Wackernagel's law applied. Like the negative indefinite pronouns, the interrogative indicators in sentences vary from dialect to dialect and must be described for each dialect (Delbrück 1900:260-267). And from this variation in form, as well as the postposed position of interrogative particles like Vedic , Goth. u, and Russian li, we may assume that the dialect patterns for the interrogative came to differ considerably from those of PIE.

Moreover, there is evidence for proposing that other particles were placed postverbally in PIE (Delbrück 1897:497-519). Delbrück has classified these in a special group, which he labels particles. They have been maintained postpositively primarily in frozen expressions: ē in Gk. egṓnē, ge in égōge ‘I’ (Schwyzer 1939:606). But they are also frequent in Vedic and early Greek; Delbrück (1897:498-511) discusses at length the use of Skt. gha, Gk. ge, and Skt. sma, Gk. mén, after pronouns, nouns, particles, and verbs, as in the following sentence:

129. RV 6.44.18. índra sūrī́n kṛṇuhí smā no ardhám
O-Indra lords-of-offering make Ptc. our half, part
‘Indra, take up the cause of our sacrificial lords.’

As here, these particles stand after the element they emphasize.

The particle which has maintained postposed position most widely in the dialects is the conjunctive, Skt. ca, Gk. te, Lat. que, Goth. -u(h) (Delbrück 1897:512-516). The disjunctive is also maintained in this use in Vedic , Gk. ēwé < Homeric ḗ and Lat. ve (Leumann 1965:521):

130. Ennius Annals 250. prudenter qui dicta loquive tacereve posset
prudently who sayings speak-or be-silent-or he-could
  ‘Whoever could prudently report hearsay or be silent about it.’

In discussing these particles it is again important to note the syntactic use and to consider secondarily the surface forms. For Hittite with its postposed particles may illustrate a more archaic pattern than do the other dialects (J. Friedrich 1960:147-155), as in the following sentence:

131. Madduwattaš 2.64. nu-u̯ar-an šannatti-i̯a munnāši-i̯a-u̯ar-an
Ptc.-Ptc.-him conceal-and not hide-and-Ptc.-him not
‘Neither conceal him nor hide him.’

The postposed negatives of this line, lē ... lē, and the postverbal conjunctives, i̯a ... i̯a, may then be taken to exemplify the postverbal position of qualifier expressions which we assume for PIE.

If these particles followed the typical order for OV languages, we would assume verbal sequences in PIE consisting of the root followed by affixes for qualifier categories like reflexivity, by aspect affixes, and by expressions for negation, interrogation, and the declarative, as in the examples cited in § 1.4.4. When VO order was introduced, expressions for interrogation and negation were shifted to a preverbal position. Interrogative particles and pronouns came to stand initially for the most part; negative particles were preposed to indefinites and to verbs and were often placed initially (Delbrück 1897:521-524). The postverbal expressions that were not transported, that is, the indicators of aspect, subsequently tense, reflexivity, mood, and the congruence categories, became fixed in the concretions known as verb endings. After these endings were fixed, the transportation processes applied largely to unaccented elements, except for stylistic purposes, which do not concern us here. The transportation processes in the dialects accordingly differed considerably from those in PIE. For example, Skt. ha, which commonly stood at the end of a clause after a finite verb, is later used in the unaccented second position (Speyer 1886:312-313; see also Macdonell 1916:340-341). We may assume then that particles which in the dialects are placed preverbally may earlier have been postverbal.

By this assumption we may account for the augment e in Indo-Iranian, Greek, and Armenian. It may be the affix for the perfect, shifted to preverbal position and subsequently fixed to verbal forms. In support of this origin for the augment we may note that it is maintained in those dialects which also maintained , also in compounded words like Skt. mā́kis, Gk. mḗtis ‘no one’. Since Hittite has preserved evidence that , the equivalent of , could stand finally (J. Friedrich 1960:146), its subsequent position before verbs as well as other elements provides credence for assuming that the augment e was a transported form of the perfect affix.

In other dialects, notably Slavic but also Baltic, Germanic, and Italic, transposed elements were combined with verbs to mark aspectual rather than temporal relationships (Delbrück 1897:146-170). The process of combination then varied from dialect to dialect (Brugmann 1904a:288, 484-486), as did the transportation rules. The unaccented pronouns, which in early PIE may have been the chief syntactic elements to stand in second position, continued to be placed in this position. But interrogatives, negatives, adverbs, and verbs came to follow placement rules of their own, obscuring the transportation rules of PIE.

4.6. Topicalization.

In topicalization an element in a sentence is marked in such a way that it assumes the most prominent role in the sentence. Differing syntactic devices may be employed to achieve such marking. In Japanese and in Lisu distinctive particles are used. In English distinctive intonation may be used. In PIE the syntactic device used was arrangement. Delbrück has identified the patterns of arrangement which are employed to achieve topicalization (1878:13); these consist of moving a syntactic element towards the front of the sentence, often placing it in initial position (see also Schwyzer 1950:690-691; Watkins 1962).

The elements most prominently affected are nouns or pronouns in the role of subject. As we have noted in Chapter 1, subjects are not essential elements in the PIE sentence. When an element is to be marked, however, a pronoun or noun can be introduced as subject and placed initially, as in the following sentence from a Strophic hymn:

132. RV 7.71.5. yuváṃ cyávānaṃ jaráso ’mumuktaṃ
you-two Cyavāna from-old-age you-two-released
‘You two released Cyavāna from old age.’

This stanza of the hymn, highlighting the two Aśvins, follows one in which the chariot of the Aśvins is emphasized, by a more elaborate pattern discussed below.

The hymn begins with a stanza depicting the time of day, the dawn. Stanza 2, which asks the Aśvins to come to the suppliant, has the verbal action topicalized:

133. RV 7.71.2. upā́yātaṃ dāśúṣe mártiāya
you-two-come-here pious mortal
‘Come here you two to the pious mortal.’

Stanza 3 emphasizes the direction of the action, with the particle ā́ in initial position:

134. RV 7.71.3. ā́ vāṃ rátham... vṛ́ṣaṇo vartayantu
hither your chariot... stallions they-must-turn
‘Hither the stallions should turn your chariot.’

As these sentences illustrate, any syntactic element—noun, verb, or particle—may be topicalized by initial placement.

While any syntactic element can be topicalized in this way, the major focus of topicalization is on an explicit subject of the sentence. An example may be taken from the Old Hittite purification ritual edited by Otten and Souček (1969:24, § 13-14).

adueni akueni nu URUHattuša ii̯annahhé LUGAL-ša
we-eat we-drink Ptc. Hattuša I-go king-however
URUArinna paizzi
Arinna goes
  ‘We eat; we drink; and I go off to Hattuša. The king however goes to Arinna.’

The priest who narrates and carries out the ritual does not need a pronoun to refer to his own action. By contrast, the king is highlighted in the following sentence; since the emphasis is on him rather than on the city Arinna, the word for “king” is put in initial position. Similar constructions may be found elsewhere in the ritual, for example p. 18, § 15. Examples from Vedic prose texts are provided by Delbrück (1878:26-32), also for predicate nouns and for nouns in all cases.

A further device to indicate topicalization may be illustrated by the fourth stanza, in which a relative construction highlights the chariot:

136. RV 7.71.4.
vāṃ rátho nṛpatī ásti voḷhā́
which you chariot lords-of-men it-is conveyor
trivandhuró vásumān̐ usráyāmā
three-seated full-of-riches daybreak-driving
ā́ na enā́ nāsatyā úpa yātam
in us with-that Nāsatyas up you-two-come
abhí yid vāṃ viśvápsnio jígāti
about when you all-nourishing it-may-come
  ‘The chariot which is yours, your vehicle,
  three-seated, rich, leaving at daybreak,
  with that, Nāsatyas, come to us,
  when the all-nourishing [chariot] approaches.’

The topicalization by which subjects came to be included and placed in initial position may be related to the development of the individualizing -s added to nouns indicating individuals (Lehmann 1958). This development led to the nominal declension of late PIE and the dialects. As I indicated in my 1958 article, traces of the earlier system are evident in Hittite and in Vedic. With the development of the nominative case to indicate subjects, the topics of sentences were generally indicated by this syntactic category. The further development of a passive inflection permitted the use of the nominative as the case for the topic, whether it referred to the agent or instrument or to the target.

In Example 136 the element to be topicalized, ráthas, is introduced in a preposed relative construction. This device is also used in Hittite, as in the following example taken from J. Friedrich (1960:168–see others there):

137. nu kuiš tān pēdaš DUMURU nu LUGAL-uš apāš kišaru
  Ptc. who second rank son Ptc. king that-one should-become
  ‘Who is a son of second rank, that one shall become king.’

While such relative clauses are more elaborate devices to indicate topicalization than are preposed elements, they are similar in effect to the patterns discussed above in initial placement. Topicalization in PIE is, then, characterized by placement of the elements topicalized at or near the beginning of the sentence.

4.7. Coordination.

While coordination is prominent in the earliest texts, it is generally implicit. The oldest surviving texts consist largely of paratactic sentences, often with no connecting particles. The Brahmanic passages cited at the end of Chapter 3, notably the passage numbered 56, illustrate such a structure. New sentences may be introduced with particles, or relationships may be indicated with pronominal elements; but these are fewer than in subsequent texts, as examples taken from an Archaic hymn and from Hittite may indicate. Stanzas 3-7 of the Archaic hymn RV 1. 167 are cited and analyzed by Hoffmann (1967:194-197); accordingly only the following stanzas will be included in the last sections of this chapter.

Stanza 8 consists of four sentences, with no connective other than the introductory particle utá in the third line:

138. RV 1.167.8.
pā́nti mitrā́váruṇāv avadyā́c
they-protect Mitra-Varuna from-shame
cáyata īm aryamó ápraṣastān
he-punishes them Aryaman-Ptc. unpraiseworthy
utá cyavante ácyutā dhruvā́ṇi
and they-waver nonwavering firm
vāvṛdhá īm maruto dā́tivāraḥ
he-has-prospered him O-Maruts giver-of-wishes
  ‘Mitra and Varuna protect from shame;
  Aryaman punishes the unpraiseworthy.
  [Though] the nonwavering and the firm waver,
  The worshiper has prospered, O Maruts.’

Neither coordinate subjects, like ácyutā and dhruvā́ṇi, nor sentences are explicitly connected with particles. In his comment on the stanza, Geldner notes that a hypotactic relationship exists here and occasionally elsewhere between sentences connected by utá (1951-1957:I, 244, note to stanza 8, lines b-c). The type of relationship must however be supplied by the reader; utá itself is a coordinating particle.

Similar patterns of paratactic sentences are found in Hittite, with no overt marker of coordination or of subordination. J. Friedrich states that “purpose and result” clauses are not found in Hittite (1960:163), but that coordinate sentences are simply arranged side by side with the particle nu, as in the Hittite Laws (1959:16, § 4):

takku ÌR-an našma GEME-an kuiški walhzi na-aš aki QAZZU
if slave or female-slave anyone he-strikes Ptc.-he he-dies his-hand
waštai apun arnuzi Ù I SAG.DU pai
it-sins that-one he-gives-recompense and one person he-gives
  ‘If anyone strikes a male or a female slave, [so that] the slave dies, his hand is guilty,
and he pays recompense for that one and gives one person.’

Here the conditional relationship of the first clause is indicated by the particle takku. But the result clause na-aš aki does not differ in form from a coordinate clause. Conditional relationships too are found in Hittite with no indication of subordination (J. Friedrich 1960:165).

The subordinate relationships that are indicated, however, have elements that are related to relative particles. Accordingly the subordination found in the early dialects is a type of relative construction, as we will note more fully in § 4.9.

As these examples and these references indicate, no characteristic patterns of order, or of verb forms, distinguish subordinate from coordinate clauses in PIE and the early dialects. Hermann therefore concluded in his celebrated article that there were no subordinate clauses in PIE (1895). The paratactic arrangement which he assumed for PIE, however, is characteristic of OV languages. Hypotaxis in OV languages is often expressed by nonfinite verb forms and by postposed particles, as we will note in the following two sections, where evidence for such patterns in PIE will be cited.

The arrangement of sentences in sequence, as in RV 1.167.8, is a typical pattern of PIE syntax, whether for hypotactic or for paratactic relationships. Expressions for coordination were used largely for elements within clauses and sentences. When used to link sentences, conjunctions were often accompanied by initial particles indicating the beginning of a new clause and also indicating a variety of possible relationships with neighboring clauses. Sentence-connecting particles are, however, infrequent in Vedic and relatively infrequent in the earliest Hittite texts; we may conclude therefore that formal markers of sentence coordination were not mandatory in PIE.

The normal coordinating particle in most of the dialects is a reflex of PIE kwe. This is postposed to the second of two conjoined elements, or to both, in accordance with conjoined constructions in OV languages ( § 4.5; Delbrück 1897:512-519). Hittite -a, -i̯a is used similarly, as in attaš annaš a ‘father and mother’ (J. Friedrich 1960:154-155). The disjunctive particle PIE wē̆ is also postposed like kwe, as in Delbrück's example (1897:516):

140. TS náktaṃ dívā varśati
by-night or by-day or it-rains
‘It rains either by night or by day.’

In Hittite, however, besides the postposed disjunctive particles -ku ... -ku ‘or’, there was the disjunctive particle našma, which stood between nouns rather than after the last, as in Example 139. This pattern of conjunction placement came to be increasingly frequent in the dialects; it indicates that the conjunction patterns of VO structure have come to be typical.

With the change in coordinating constructions, new particles were introduced; some of these, for example, Lat. et, Goth. jah, OE and, have a generally accepted etymology; others, like Gk. kaí, are obscure in etymology. Syntactically the shift in the construction rather than the source of the particles is of primary interest, though, as noted above, the introduction of new markers for the new VO patterns provides welcome lexical evidence of a shift. The syntactic shift also brought with it patterns of coordination reduction (Ersparung) which have been well described for some dialects (Behaghel 1923-1932:III, 497-528). Such constructions are notable especially in SVO languages, in which sequences with equivalent verbs (S, V, O, Conj., S2, V1, O2) delete the second occurrence of the verb (ibid.:525, a MHG excerpt from Tauler):

141. daz einer einez will und ein ander ein anderz
  ‘that one one-thing wants and another an other’

Reduction of equivalent nouns in either S or O position is also standard, as in Beowulf 120-123 (Behaghel 1923-1932:III, 512).

...Wiht unhǣlo,
...creature of-evil
...gearo sōna wæs,
...ready at once he-was
...ond on ræste genam
...and in sleep took
þrītig þegna
thirty thanes
  ‘The creature of evil was ready at once and seized in their sleep thirty warriors.’

But in the paratactic structures characteristic of Hittite, such reduction is often avoided, as in the early lines of Muršilis Sprachlähmung (Götze and Pedersen 1934:4, § 3):

nu-mu-kán memii̯aš KAxU-i anda tepau̯ešta
now-to-me-within word mouth in it-became-little
nu-mu-kán memii̯aš tepu kuitki šarā ii̯attat
now-to-me-within word little somewhat up it-went
  ‘Now my speech in my mouth became infrequent;
  now my speech came upward somewhat slowly.’

In an SVO language the second memii̯as would probably not have been explicitly stated, as in: ‘now my speech came to be halting and was uttered slowly.’ The lack of such reduction, often a characteristic of OV languages, gives an impression of paratactic syntax.

Another pattern seeming to be paratactic is the preposing of “subordinate clauses,” either with no mark of subordination or with a kind of relative particle, as in the concluding passage of Muršilis Sprachlähmung (Götze and Pedersen 1934:10, § 16). The second from last clause has no mark to indicate subordination; the earlier clauses contain a form of relative particle.

IŠTU GIŠBANŠUR-ma-za-kán kuizza azikinun
from table-but-Refl.-Ptc. from-which I-was-accustomed-to-eat
IŠTU GAL-i̯a-kán kuizza akkuškinun
from beaker-and-Ptc. from-which I-was-accustomed-to-drink
šašti-i̯a-za-kán ku̯edani šeškeškinun IŠTU
in-bed-and-Refl.-Ptc. in-which I-was-accustomed-to-sit from
URUDDU10xA-ia-za-kán kuizza arreškinun
basin-and-Refl.-Ptc. from-which I-was-accustomed-to-wash
kuit-i̯a imma ÚNUTU anda u̯erii̯an ešta nu UL
what-and else utensil Adv.-Ptc. mentioned it-was now not
any it-was-taken from god likewise it-was-determined
  ‘The god also determined that nothing more should be used of the table from which I was accustomed to eat, of the beaker from which I was accustomed to drink, of the bed in which I was accustomed to sleep, of the basin in which I was accustomed to wash, and of whatever other article was mentioned.’

In an SVO language like English, the principal clause, which stands last in Hittite, would be placed first. The interpretation of the preceding clause as a result clause is taken from Götze and Pedersen. The initial clauses contain relative particles which indicate the relationship to kuitki of the second-from-last clause; they also contain coordinating particles: a, i̯a. In this passage the clauses, whether coordinate or subordinate from our point of view, are simply arrayed in sequence. Each concludes with a finite verb which provides no evidence of hypotaxis. The sentence connectives which occur—repeated instances of a/ia—heighten the impression of coordination.

Such examples may also provide insights concerning the absence in Hittite of verb forms which are cognates of the Vedic and Greek optative and subjunctive. These forms came to be used largely to indicate subordination. Since Hittite was highly consistent in its OV patterning, such verb forms were not required. Hittite however did not forego another device, which is used to indicate subordinate relationship in OV as well as VO languages, the so-called nonfinite verb forms. These are used for less explicit kinds of complementation, much the way relative constructions are used for more explicit kinds, as we note in the following sections.

4.8. Complementation.

The understanding of compound sentences and sentences with complements has been a problem to Indo-Europeanists since the beginnings of their concern with syntax. A citation from Windisch's investigation of the origin of the relative pronoun in the Indo-European languages may assist in understanding their point of view: “For syntax there follows from what has been said the important result, that the simple sentence was indeed developed before the separation of languages, but compound sentences [Satzgefüge] were not” (1869:205). This point of view may be characterized as reflecting the belief that language has gradually become more complex; man first talked in simple sentences, but in time he developed compound and complex sentences. The process of this development was still transparent in the history of PIE and its early dialects. Indo-Europeanists have completely rejected this view; but they have not departed from the attempted explanations of syntactic constructions, such as relative clauses, which were made in line with it. Hittite has now provided the data for understanding the relative constructions which we posit for late PIE (Raman 1973). This understanding, both of complementary constructions in PIE and of their development in the early dialects, is built on observations of typological patterning in language.

Study of typological characteristics has clarified not only the syntactic characteristics of simple sentences but also those of compound sentences in languages of differing types. Compound sentences may result from the embedding of nominal modifiers; in VO languages embedded nominal modifiers follow nouns, whereas in OV languages they precede nouns. This observation has led to an understanding of the Hittite and the reconstructed PIE relative constructions. If we follow the standard assumption that in relative constructions a second sentence containing an NP equivalent to an NP in the matrix sentence is embedded in that matrix sentence, we may expect that either sentence may be modified, as illustrated in § 3.6. A sentence may also be embedded with a dummy noun; the verb forms of such embedded sentences are commonly expressed with nominal forms of the verb, variously called infinitives, supines, or participles. In OV languages these, as well as relative constructions, precede the verb of the matrix sentence.

To illustrate these processes, a relatively simple sentence may be cited from an ancient Japanese story, “Taketori no okina no monogatori” [Story of the old man of (= who was) a bamboo-hewer] (Dickens 1888: 50);

Mote itarite kano ura ni oru Wokei ni
having, holding going-to that coast on being Wokei to
kogane wo torasu
money, gold Ptc. cause-to-take

As commentary to the sentence and its structure, F. Victor Dickens's translation is quoted: “So Fusamori took the letter and went down to the coast, and delivered it to Wokei, to whom he likewise gave gold” (1888: 16). Dickens's interpretation results from an analysis by which the first reduced coordinate sentence is mote ‘having’; the second is itarite ‘going to’; these are “coordinate” with torasu, a causative accompanied by an agent and a target. With less interpretation than given by Dickens the sentence may be translated: “Having [the letter] and going to [the coast] he causes Wokei to take the money.” Within this sentence a further sentence is embedded, kano ura ni oru ‘he is on that coast’. A more literal translation than Dickens's might then read: ‘So Fusamori takes the letter and goes and gives [it and] money to Wokei, who is on that coast.’ The nonfinite verb forms mote and torite are loosely connected with the principal verb torasu. The attributive or relative-marked verb form oru is here closely connected with the following noun Wokei; in another wide-spread use, forms corresponding to oru are connected with a following particle.

If we regard the Japanese sentence as an example of a compound sentence in OV languages, we may expect to find in PIE as well nonfinite verb forms like mote and itarite preceding the main clause and also relative constructions preceding their antecedents. The Japanese nonfinite forms correspond to participles in the IE languages, like vásānaḥ in the last lines of the following Strophic hymn:

146. RV 4.5.15.
rúśad vásānaḥ sudṛśīkarūpaḥ
brightly dressing-himself beautifully-hued
kṣitír rāyā́ puruvā́ro adyaut
dwelling like of-riches much-desired gleamed
‘Dressing himself brightly, beautifully hued,
the much-desired [Agni] gleamed like a dwelling of riches.’

The Japanese nonfinite forms also correspond to infinitives in the IE languages, such as srávitavā́i which is used for expressing an embedded complement of result or, according to Macdonell (1916:333), “a final or consequential sense” in the following Strophic hymn:

147. RV 7.21.3. tvám indra srávitavā́ apás kaḥ
you O-Indra for-flowing waters you-make
‘You, O Indra, make the waters to flow.’

Also in the poetic texts such infinitives may follow the main verb, as in the following passages:

148. RV 5.1.2. ábodhi hótā yajáthāya devā́n
he-woke-up priest for-sacrificing gods
‘The priest has awakened to sacrifice to the gods.’

149. RV 5.39.2.
vidyā́ma tásya te vayám
we-wish-to-know your you we
ákūpārasya dāváne
infinite for-giving
  ‘We wish to know you as you [are], infinite in gifts.’

The postposed order may result from stylistic or poetic rearrangement; yet it is also a reflection of the shift to VO order, a shift which is reflected in the normal position for infinitives in the other IE dialects. In the Brahmanas still, infinitives normally stand directly before the verb, except in interrogative and negative sentences (Delbrück 1878:33-35):

150. ŚB tád áśvam ā́netavaí brūyāt
then horse for-bringing he-should-speak
‘He should order [them] to bring a horse then.’

On the basis of the Brahmanic order we may assume that in PIE nonfinite verbs used as complements to principal verbs preceded them in the sentence.

Hittite provides examples of preposed complementary participles and infinitives to support this assumption (J. Friedrich 1960:111-112, 142-155, 164). Participles were used particularly with har(k)- ‘have’ and eš- ‘be’, as in the example given above (144): uerii̯an ešta ‘was mentioned’; the pattern is used to indicate state. Among their various uses, infinitives could indicate result, with or without an object (J. Friedrich 1960:142):

151. 1-aš 1-an kunanna šanhanzi
  one one to-kill not he-tries
  ‘One should not try to kill another.’

Moreover, infinitives could be used to express purpose, as in the following example, which pairs an infinitive with a noun (J. Friedrich 1960: 143):

152. tuk-ma uttar ŠÀ-ta šii̯anna išhiull-a ešdu
  to-you-however this word in-heart for-laying instruction-and it-should-be
  ‘But for you this word should be for taking to heart and for instruction.’

Moreover, the infinitive could be loosely related to its object, as in examples cited by Friedrich (1960:143-144), such as the following:

153. apāš-ma-mu harkanna šan(a)hta
  he-however-me for-deteriorating he-sought
  ‘But he sought to destroy me.’

The complementary infinitive indicates the purpose of the action; as Friedrich points out, it is attached to the verb šanhta plus its object mu in a construction quite different from that in subsequent dialects.

These uses are paralleled by uses in Vedic, as may be noted in the work of Macdonell (1916:333-338), from which the following examples are taken (p. 336):

154. agníṃ páristarītavā āha
  fire for-enclosing he-says
  ‘He says that the fire is to be enclosed.’

155. RV 5.79.10.
etā́vad véd uṣas tvám
so-much or-Ptc. dawn you
bhū́yo dā́tum arhasi
more or give you-can
  ‘You can give either so much or more, O Dawn.’

On the basis of such examples in Vedic and in Hittite, we may assume that infinitive constructions were used to indicate a variety of complements in PIE.

Hittite and Sanskrit also provide examples of participles functioning appositionally or as adjectives indicating state (J. Friedrich 1960:164):

156. ammuk-u̯ar-an akkantan IQ.BI
  to-me-Pte.-indicating-quotation-him dying he-described
  ‘He told me that one had died.’

This pattern had been noted by Delbrück for the Rigveda, with various examples (1900:327):

157. RV 10.42.3. śiśīhí śiśayáṃ tvā śṛṇomi
you-strengthen me being strong, strengthening you I-hear
‘Strengthen me; I hear that you are strong.’

The adjective śiśayá ‘strengthening’ is an adjective derived from the same root as śiśīhí.

Delbrück also noted that such “appositives” are indicated in Greek by means of clauses. Greek accordingly represents a further stage in the development of the IE languages to a VO order. Yet Greek still maintained preposed participles having the same subject as does the principal verb, as in:

158. Odyssey 5.486. tḕn mèn idṑn gḗthēse
it Ptc. seeing he-rejoiced
‘Seeing it, he rejoiced.’

This pattern, preserved in some IE languages to this day, permits the use of two verbs with only one indicating mood and person; the nonfinite verb takes these categories from the finite. As we have noted in examples given above, participles were used in the older period for a great variety of relationships. though also without indicating some of the verbal categories. Dependent clauses are more flexible in indicating such relationships, and more precise, especially when complementary participles and infinitives follow the principal verb. Possibly this was largely the reason for their great expansion in the dialects. The description of these and of their development in the early dialects will be discussed in the following section.

4.9. Subordinate Clauses.

In an OV language a subordinate clause proper consists of a finite verb form preposed to another word, often a particle. For example, the verb oru ‘be’ is preposed to ni ‘in, on, when’ in the following sentence from the story of the old bamboo-hewer (Dickens 1888:49):

kore wo mo awareto mite oru ni Taketori
this Obj.-Ptc. Ptc. sadly seeing be on, when bamboo-hewer
no Okina hashiri irite iwaku
Ptc. old-man running coming-in say
  ‘When she looked with sadness at this, the old bamboo-hewer comes in swiftly and says...’

The form oru is the same as that in the sentence cited above (Example 145). In that sentence the relationship corresponds to that between a noun and a relative clause. By contrast, in the sentence cited here the relationship corresponds to that between a temporal subordinate clause and a principal clause. These Japanese examples may illustrate the relationship between relative clauses and subordinate clauses in an OV language.

Indo-Europeanists have long recognized the relationship between the subordinating particles and the stem from which relative pronouns were derived in Indo-Iranian and Greek. Thus Delbrück has pointed out in detail how the neuter accusative form of PIE yo- was the basis of the conjunction *yod in its various meanings: (1) temporal, (2) temporal-causal, (3) temporal-conditional, (4) purpose (1900:319-333). He also recognized the source of conjunctional use in sentences like:

160. RV 3.48.2.
yáj jā́yathās tád áhar asya kā́me
which you-were-born that day of-it in-desire
’ṅśóḥ pīyū́ṣam apibo giriṣṭhā́m
of-plant milk you-drank be-on-mountain
  ‘On the day you were born you drank the mountain milk out of desire for the plant.’
161. RV 7.86.4.
kím ā́ga āsa varuṇa jyéṣṭhaṃ
what sin it-is O-Varuna terrible
yát stotā́raṃ jíghāṅsasi sákhāyam
which singer you-want-to-kill friend
  ‘What is the terrible sin, O Varuna, because of which you reject the singer, your friend?’

Although Delbrück recognized the source of the conjunction *yod and other conjunctional forms based on the relative-pronoun stem yo- he did not provide an explanation for this development from a relative clause to a subordinate clause.

Comparison with the OV structures illustrated in the Japanese sentences illuminates the relationship. In an OV language the relative-clause relationship is parallel with that in subordinate clauses: Modern Japanese for example simply preposes a finite verb before either a noun or an adverbial particle to indicate some kind of modifying relationship of the clause concluded by the verb. Ancient Japanese provides an illustration of a variant device, for it contained a special, attributive verb ending to mark the relationship, exemplified in the -u ending of the verb ori. If such a situation applied in PIE when it was OV in structure, upon the shift to a VO structure PIE might have been expected to use one and the same device to indicate linking with a following clause. This device was the so-called relative pronoun.

Delbrück's preliminary explanation of relative and subordinate clauses in PIE has been supported by closer analysis of the use of ya in the old sections of the Rigveda (Porzig 1932:210-303) and by parallel developments in Hittite. Walter Porzig's interpretation of the relative clauses in the Rigveda must be viewed in accordance with the clarification that has been provided by the more archaic Hittite texts. But even without drawing on this clarification, Porzig's conclusions are valuable. Arguing in part on the basis of the verbal accent in subordinate clauses, Porzig concludes that relative clauses must have stood before the main clause originally and, further, that the earliest type of subordinate ya- clauses must have been the preposed relative constructions (1932:292). On the basis of his own investigation, then, Porzig supports Delbrück's derivation of a variety of ya- clauses from a relative nexus.

This conclusion receives striking support from Hittite, for in it we find the same syntactic relationship between relative clauses and other subordinate clauses as is found in Vedic, Greek, and other early dialects. But the marker for both types of clauses differs. In Hittite it is based on *kwid rather than *yod; thus, Hittite too uses the relative particle for indicating subordination. The remarkable parallelism between the syntactic constructions, though they have different surface markers, must be ascribed to typological reasons; we assume that Hittite as well as Indo-Aryan and Greek was developing a lexical marker to indicate subordination, in contrast with simple preposing of the subordinate construction like that in Japanese. Since Delbrück has provided many example of ya- constructions (1900:319-333), the citations here are largely Hittite kuit clauses, together with excerpts from the Archaic hymn 1.167.

As does yad in Vedic, Hitt. kuit signals a “loose” relationship between clauses which must be appropriately interpreted; an example from the Old Hittite Šaušgamuwa Treaty may illustrate the Hittite nexus (Kühne and Otten 1971:8, line 8):

tuk-ma-za [m]DIŠTAR-A.A.-an HA-DA-A-NU
you-however-Refl. Šaušgamuwa brother-in-law
kuit DÙ-nun nu DUTUŠI AŠ-ŠUM EN-UT-TI pahši
“that” I-made Ptc. my-sun, me concerning rule you-protect
  ‘Protect me in my rule, [regarding the fact that = because] I made you,
Šaušgamuwa, my brother-in-law.’

This is part of a statement made by the king of the Hittites; the first line recapitulates a statement he has made earlier in the text. In contrast with the two earlier occurrences, the recapitulation is related to the nu clause by means of kuit, which the editors translate ‘because’. As J. Friedrich has stated (1960:163), kuit never stands initially in its clause. Sentences in which it is used are then scarcely more specifically interconnected than are conjoined sentences with no specific relating word, as in examples cited by Friedrich (ibid.):

163. nu taškupāi nu URU-aš dapii̯anzi išdammašzi
  Ptc. you-shout Ptc. city whole it-hears
  ‘Now cry out [so that] the whole city hears.’

Like this example, both clauses in a kuit construction generally are introduced with nu (J. Friedrich 1960:159). We may assume that kuit became a subordinating particle when such connections were omitted, as in Friedrich's example (ibid.):

ABU.KA-mu kuit tuēl ŠUM-an memiškit nu-tta
father-me because your name he-said-repeatedly now-you
appaddan EGIR-an šan(a)hhun
for-that-reason therefore I-cared-for
  ‘Because your father mentioned your name to me again and again,
for that reason I concerned myself about you.’

The relationship expressed by yád in the following Strophic hymn is similar, though yád can also be analyzed as a relative pronoun still:

165. RV 1.131.4.
vidúṣ ṭe asyá vīryàsya pūrávaḥ
they-know you of-this manliness Purus
púro yád indra śā́radīr avā́tiraḥ
cities because, when O-Indra autumnal you-tore-down
  ‘The Purus know this heroic deed of yours, Indra, since you tore down the autumnal cities.’

These examples illustrate that both yád and kuit introduce causal clauses, though they do not contain indications of the origin of this use.

We assume that the use of yád, kuit, and other relative particles to express a causal relationship arose from subordination of clauses introduced by them to an ablative, as in the following passage:

166. RV 7.89.5.
ácittī yát táva dhármā yuyopimá
unknowing that, because your law, order we-have-disturbed
mā́ nas tásmād énaso deva rīriṣaḥ
not us because-of-that because-of-sin O-god you-harm
  ‘Do not harm us, god, because of that sin [that] because unknowingly we have disturbed your law.’

As such relationships with ablatives expressing cause were not specific, more precise particles or conjunctions came to be used. In Sanskrit the ablatival yasmāt specifies the meaning ‘because’.

Further, yadā́ and yátra specify the meaning ‘when’. In Hittite, mān came to be used for temporal relationships, possibly after combined use with kuit; kuitman expressed a temporal relationship even in Late Hittite, corresponding to ‘while, until’, though mahhan has replaced mān (J. Friedrich 1960:164-165 gives further details). The conjunction mān itself specifies the meanings ‘if’ and ‘although’ in standard Hittite. In both Hittite and Vedic then, the “loose” relative-construction relationship between subordinate clauses and principal clauses is gradually replaced by special conjunctions for the various types of hypotactic relationship: causal, temporal, conditional, concessive. Just as the causal relationship developed from an ablative modified by a relative construction, so the temporal and conditional relationship developed from a clause modifying an underlying time node.

The less differentiated and less precisely related subordinate clauses are often still evident, however, as in yád clauses of the Archaic hymn, Rigveda 1.167. For conciseness, only yád clauses will be cited here, with Hoffmann's interpretation of each; the entire stanzas and their translations are given by Hoffmann (1967:194-197).

RV 1.167.5. jóṣad yád īm asuryā̀ sacádhyai
  she-desires when them Asuryan to-follow
‘when the Asuryan will desire to follow them’
RV 1.167.6. arkó yád vo maruto havíṣmān
  song-of-praise whenever, if for-you Maruts accompanied-by-libations
‘if the song of praise accompanied by libations is designed for you, Maruts’
RV 1.167.7.
sácā yád īṃ vṛ́ṣamaṇā ahaṁyú
together because them manly-minded proud
sthirā́ cij jánīr váhate subhāgā́ḥ
rigid though women she-drives well-favored
‘because the manly minded, proud, yet stubborn [Rodasi] brings along other favored women’

In these three stanzas yad introduces subordinate clauses with three different relationships: temporal, conditional, causal. Such multiple uses of yad belong particularly to the archaic style; subsequently they are less frequent, being replaced by more specific conjunctions.

In addition to the greater specificity of subordinate relationship indicated by particles, the early, relatively free hypotactic constructions come to be modified by the dominant subjective quality of the principal verb. The effect may be illustrated by passages like the following from a Strophic hymn, in which the verb of the principal clause is an optative:

168. RV 1.38.4.
yád yūyám pṛṣnimātaro
if, when you having-Prsni-as-mother [Maruts]
mártāsaḥ syā́tana
mortals you-would-be
stotā́ vo amṛ́taḥ syāt
singer your immortal he-would-be
  ‘Your singer would be immortal if [= in a situation when] you Maruts were mortals.’ (That is, if our roles were reversed, and you were mortals, then you would wish me to be immortal.)

This passage illustrates how the use of the optative in the principal clause brings about a conditional relationship in the subordinate clause (see also Delbrück 1900:329-330). Through its expression of uncertainty the optative conveys a conditional rather than a temporal meaning in the yad clause.

Lacking verb forms expressing uncertainty, Hittite indicates conditional relationships simply by means of particles (J. Friedrich 1960: 165-167). Although several particles are used in Hittite to indicate various types of conditional clauses—man ... mān for contrary-to-fact, takku and man for simple conditionals—Hittite did not develop the variety of patterns found in other dialects. These patterns, as well described in the handbooks, are brought about not only by differing particles but also by the uses of the various tense and mood forms. Constructions in the dialects which have developed farthest from those of PIE are those in which the tense, mood, or person is modified in accordance with rules based on the verb form of the principal clause. Such shifts are among the most far-reaching results of the subjective quality of the Indo-European verb (Delbrück 1900:437-445).

Differences between the constructions in the various dialects reflect the changes as well as the earlier situation. In Homer, statements may be reported with a shift of mood and person, as in the following line:

169. Odyssey 3.19. líssesthai min autós, hópōs nēmertéa eípēi
request Ptc. him self that true-things he-may-say
‘You yourself ask him so that he tells the truth.’

The form eípēi is a third-person aorist subjunctive. If the statement were in direct discourse, the verb would be eȋpe, second-person imperative, and the clause would read: eȋpe nēmertéa ‘tell the truth’. Such shifts in person and mood would not be expected in an OV language; in Japanese, with to following quotations, the statement is repeated without change. This is also the situation in Vedic, in which statements are repeated and indicated with a postposed iti, as in the passage (Example 56) of Chapter 3. The shifts in the other dialects, as they changed more and more to VO structure, led to intricate expression of subordinate relationships, through shifts in person, in mood, and in tense, as well as through specific particles indicating the kind of subordination. The syntactic constructions of these dialects then came to differ considerably from that even in Vedic.

The earliest poems of the Vedas are transparent in syntax, as may be illustrated by Stanzas 9 and 10 of Hymn 1.167:

RV 1.167.9.
nahī́ vo maruto ánty asmé
never Ptc. your Maruts near from-us
ārā́ttāc cic chávaso ántam āpúḥ
from-far or of-strength end they-reached
dhṛṣṇúnā śávasā śuśuvā́ṅsó
they bold power strengthened
’rṇo dvéṣo dhṛṣatá pári ṣṭhuḥ
flood like enmity bold against they-stand
‘Never have they reached the limit of your strength, Maruts, whether near or far from us. Strengthened by bold power they boldly oppose enmity like a flood.’
RV 1.167.10.
vayám adyéndrasya préṣṭhā vayám
we today-Indra's most-favored we
śvó vocemahi samaryé
tomorrow we-wish-to-be-called in-battle
vayám purā́ máhi ca no ánu dyū́n
we formerly great and us through days
tán na ṛbhukṣā́ narā́m ánu ṣyāt
that us chief of-men to may-he-be
‘We today, we tomorrow, want to be called Indra's favorites in battle. We were formerly. And great things will be for us through the days; may the chief of men give that to us.’

Although the hymn offers problems of interpretation because of religious and poetic difficulties, the syntax of these two stanzas is straightforward; the verbs in general are independent of one another, in this way indicating a succession of individual sentences. Such syntactic patterns, though more complicated than those of prose passages like Examples 56 and 64 of Chapter 3, lack the complexity of Classical Greek and Latin, or even Homeric Greek. These early Vedic texts, like those of Old Hittite, include many of the syntactic categories found in the dialects, but the patterns of order and relationship between clauses had already changed considerably from the OV patterns of PIE. After we review the categories and their methods of expression (Chapter 5), and after we then examine characteristic lexical patterns of PIE (Chapter 6), we will summarize the principal syntactic developments between PIE and the dialects (Chapter 7).

< previous section | Jump to: next section >

  • Linguistics Research Center

    University of Texas at Austin
    PCL 5.556
    Mailcode S5490
    Austin, Texas 78712

  • For comments and inquiries, or to report issues, please contact the Web Master at