Drifting between passive and anticausative. True and alleged accent shifts in the history of Vedic yapresents Текст научной статьи по специальности «Языкознание и литературоведение»

Discussion Articles / Дискуссионные статьи

Leonid Kulikov

Leiden University / Institute of Linguistics (Moscow)

Drifting between passive and anticausative.

True and alleged accent shifts in the history of Vedic -ya-presents

This paper focuses on the system of the Vedic present formations with the suffix -ya- and middle inflexion, paying special attention to the attested accent patterns. On the basis of a study of the paradigmatic and syntactic features of this verbal formation we can conclude that the traditional analysis of some members of this class in terms of the passive/nonpassive (anticausative) opposition is inadequate. I will offer a short overview of the history of this class, concentrating, in particular, on several accent shifts which account for a number of exceptions to the general correlation between the semantics and accent placement (passives: accent on the suffix vs. non-passives: accent on the root). Some of these shifts can be dated to the prehistoric (Common Indo-Aryan?) period (cf. suffix accentuation in such non-passives as mriyate 'dies'), while some others must be features of certain Vedic dialects, dating to the period after the split of Common Indo-Aryan.

Keywords: Sanskrit, Vedic, Rgveda, Atharvaveda, Yajurveda, Indo-European, passive, anticau-sative, middle voice, accentuation, accent shift.

1. Passive, reflexive, anticausative: preliminary remarks and definitions

The distinguishing between closely related intransitive derivations, such as passive, reflexive, anticausative (decausative), is one of the most intricate semantic and syntactic issues in languages with polysemous intransitive markers. Both anticausative and passive derivations entail the promotion of the initial direct object (= Patient) and the demotion of the initial subject (= Agent). This common syntactic feature accounts for their similar morphological marking in many languages (see e.g. Comrie 1985: 328ff.; Haspelmath 1987: 29ff.). In the cases where the markers of the passive and anticausative (at least partly) overlap, passives without an overtly expressed agent can be distinguished from anticausatives only by semantic criteria. This semantic opposition is characterized, for instance, by Comrie (1985: 326) as follows:

Passive and anticausative differ in that, even where the former has no agentive phrase, the existence of some person or thing bringing about the situation is implied, whereas the anticausative is consistent with the situation coming about spontaneously.

This general definition is also relevant for a description of the system of intransitive derivations in a number of Ancient Indo-European languages, such as Ancient Greek or (Vedic) Sanskrit.

Journal of Language Relationship • Вопросы языкового родства • 6 (2011) • Pp. 185-215 • © Kulikov L., 2011

In what follows I will focus on the Vedic verbs with the suffix -ya-, particularly on the genesis of their accentuation. Generally, the -ya-presents with the accent on the suffix are passives (kriyäte 'is made', ucyäte 'is called', stüyäte 'is praised', hanyäte 'is killed'), whereas the -ya-presents with root accentuation behave as non-passive intransitives (cf. pädyate 'falls', bud-hyate 'wakes', nyate 'flows'). However, a few -ya-formations are generally regarded as exceptions to this regularity. The parade examples include mriyäte 'dies' (root mr) and its semantic counterpart jayate 'is born' (root jan).

2. Non-passive -ya-presents with suffix accentuation: the type mriyäte

One of the most debated Vedic verbal formations relevant for a study of the accentual history of the -ya-presents is mriyäte 'dies'. While its semantic opponent, jayate (on which see next Section), is regarded as a passive by meaning, non-passive by form, mriyäte is taken as a passive by form, but non-passive by meaning, being quoted in all Vedic and Indo-European grammars as a handbook example of the non-passive usage of a -ya-present with suffix accentuation.1 A few attempts to analyse this present as a passive proved unsuccessful. For instance, Negelein (1898: 38) treated it as the passive of the transitive mf (<*melH-) 'crush, destroy' ["Der Inder mag sich den Hergang des Todes sehr wohl als ein Zermalmtwerden (mr malmen) vorgestellt haben"], which is etymologically impossible. Hartmann in his book Das Passiv. Eine Studie zur Geistesgeschichte der Kelten, Italiker und Arier (Hartmann 1954: 186ff.) even assumed a particular passive conceptualisation of death in Ancient India. The fact that two verbs which belong to one and the same semantic domain, jayate 'is born' and mriyäte 'dies', show such a striking dissimilarity in accentuation, which generally corresponds to the functional opposition "passive/non-passive", did not escape his attention. But his conclusions from this remarkable fact in the vein of Geistesgeschichte are untenable:

Trotz gewisser Übereinstimmungen im Gefühlswert beider Verba kann jedoch kein Zweifel darüber bestehen, daß das Ausmaß des ,passiven' Einschlages jayate geringer gewesen sein muß als bei mriyäte , da das Gefühl des Ausgeliefertseins an eine außerhalb des Subjektes liegende Macht bei einem Ausdruck für das Zurweltkommen einer Seele nicht so groß gewesen kann wie beim Sterben.

Needless to say that this explanation hardly deserves any serious discussion: mriyäte never functions as a passive (see e.g. Jamison 1983: 150, fn. 92), and, semantically, belongs with the root-accented '-ya-presents of change of state, together with its counterpart jayate 'is born'. An explanation of the abnormal suffix accentuation can be given in phonological terms, and it holds also for a few other 'formal passives' ('pseudo-passives') made from the roots of the type Cr, where the suffix ('passive') accentuation must be of secondary nature. This class also includes dhriyäte 'stays' and a-driyäte 'heeds'. From the semantic and syntactic point of view and in spite of their suffix accentuation, the presents -driyä-te, dhriyä-te, mriyä-te (as well as ghriya-te and -sriya-te, which do not occur unambiguously accented) can be easily grouped with the nonpassive middle -ya-presents of the three main classes, which include:2

1 Cf. Delbrück 1874: 167f.; Whitney 1889: 277, §277; Macdonell 1910: 333, §444a.

2 The non-passive character of these three presents was probably the main reason which has caused the Indian grammarians to group these formations (together with the post-Vedic -priyate [with the preverb a] 'be occupied, employed'; cf. DhP VI 109) with class VI presents; cf. Pan. 1.3.61. In Western scholarship this analysis was

1) verbs denoting changes of state3 (mostly of spontaneous and non-controllable character): jari 'be born' — jaya-te RV+, pya 'fill, swell' — pyaya-te RV+, budh (a)4 '(a)wake' budhya-te RV+, IT (a) 'dissolve' — liya-te RVKh.+;

2) verbs of motion and body posture: pad 'fall, move' — padya-te RV+, 1ya 'drive, speed' — iya-te RV+, rT 'whirl, swirl' — nya-te RV, VS1x, IT (p) 'adhere'--lTya-te Br.+;

3) verbs of mental activities, constructed with the accusative: ka 'long (for), yearn' — kaya-te RV1x, budh (p) 'perceive' — budhya-te AV+, man 'think, respect' — manya-te RV+, mrs 'forget' — mrsya-te RV+.

No doubt, the similar morphological marking of these presents reflects their semantic affinity within the Vedic verbal system. Note that for all these semantic types, middle voice marking is typical in the world languages (see Kemmer 1993; 1994: 182f. et passim). In spite of the small range of classes (1-3), their relevance within the Vedic verbal system is obvious. These types determine which meanings are productive (and, hence, "morphologically influential") in the class of middle -ya-presents, and which are not. For instance, the relevance of type (2) may account for the secondary and more recent usage of budhya-te, which originally (in the RV) could only be used in the intransitive usage (a), meaning '(a)wake'; after the RV, when the class I present bodha-ti 'perceive' disappears, budhya-te takes over its function, thus being grouped together with such present formations as manya-te 'think, respect' or mrsya-te 'forget'.

All verbs of the type mriya-te perfectly fit the three semantic classes listed above. mriya-te denotes a change of state (note, particularly, the parallelism with jaya-te, which will be discussed at length below, in Section 3);5 dhriya-te (together with the hapaxes ghriya-te and -sriya-te) belongs with verbs of motion and body posture; -driya-te refers to a mental activity. Moreover, even the secondary meaning of dhriya-te (p) 'decide, determine', attested from Middle Vedic (Brahmanas) onward, perfectly fits class (3), too. Thus, within the verbal system, all these Criya- presents belong with the middle -ya-presents, and even their later developments are determined by the semantic skeleton (1-3), as shown in Table 1 below:6

Thus, the suffix ("passive") accentuation in the first three presents of the type mriyate must be of secondary origin. All these formations are derived from Cr roots and, together with -ya-passives of the same structure (kriyate 'is made', bhriyate 'is carried' etc.), represent a specific development of r before the present suffix -ya-. Most likely, the regular reflex of *CfiV-was such that it disturbed the morphological transparency of the formation (for instance, **muryate)7 The only way to preserve the transparency of the form was to introduce the accent on the suffix: *Cr-ia—> Criya- (i.e.: *mr-ia—> mriya- etc.). Here the type kriyate (where -ri- goes

usually regarded as mere misunderstanding (see e.g. J. Schmidt 1875: 244ff.; cf. also Benfey 1866a: 198f.). In fact, however, the segmentation mriy-a- is the only possible synchronic solution of the descriptive conflict between the "passive" form and the non-passive meaning of these presents: class VI is the only thematic present with the accent on the thematic vowel (cf. ksi 'dwell' — ksiy-a-ti).

3 See Levin 1993: 240ff., with bibl.

4 Hereafter I use Greek characters (a, |3) to refer to different meanings of polysemous -ya-presents.

5 Cf. M. Leumann 1940: 232 [= Kl.Schr., 323]; Gonda (1951: 92): "the two verbs [= mriyate and jayate. — L. K.] 'formed a pair' and influenced each other".

6 This must also hold true for the presents ghriya-te 'drip' and -sriya-te 'stretch', which do not occur accented, but, by virtue of their phonological structure can only bear accent on the suffix: *ghriya-te, *-sriya-te.

7 Cf. durva- < *dfueH- (Lubotsky 1997: 148, with fn. 29). Note that the -ya-presents (including -ya-passives) derived from roots ending in long sonants (such as purya- 'become full' < *prH-ia-) are not discussed in this paper. On the Avestan reflexes of *Cfi- (Crii-, Cirii-), see Beekes 1999: 64.

Table 1. The main semantic classes of the middle '-ya-presents and the corresponding presents of the type Criyd-te

-ya-presents passives (yujyate, dïyate, hanyate, kriyate, ...)

dhriyate (a), -sriyate, mriyate -driyate, dhriyate (13) ghriyate

middle -ya-presents change of state (jayate, budhyate (a), ...) motion and body posture (padyate, nyate, ...) mental activities (manyate, bûdhyate (|3), ...)

(1) (2) (3)

back to the accentless -r- before -z-) may have served as a model. Due to this accent rule, presents of the type mriyate, which originally belonged with middle '-ya-presents, formally fell together with -ya-passives.8

3. jayate 'is born' — anticausative or former passive?

According to the opinion widely spread in earlier Indo-European and Indo-Iranian studies, jayate (as well as its Old-Iranian cognate, Avestan zaiieiti) is the original passive, with the secondary accent shift in Vedic. Whitney in his seminal Sanskrit grammar (1889: 273, §761b) called it 'altered passive'; likewise, Macdonell in his Vedic grammar (1910: 333, §444a) claims that the original passive has been "transferred to the radically accented ya- class": *jayate—> jayate. Similar statements can also be found in later studies.9 There is no sufficient evidence for such a hypothesis, however. Although a passive interpretation ('is born by smb.') is possible per se, it cannot be supported by the syntactic features of jan. Witness the following examples from the Rgveda and Satapatha-Brahmana:

(1) RV 6.7.3a

tvad vipro ja-ya-te vajy agne

you:ABL poet:NOM.SG bear-YA-3SG.MED prize-winner:NOM.SG fire:VOC.SG 'From you, o fire, is born the poet, the prize-winner.'

8 For a detailed discussion of this morphological type, see Kulikov 1997. On the secondary accent shift in mriyate, see also Szemerényi 1964: 184, fn. 1. It is worth mentioning that a number of Indo-Europeanists and San-skritists, without explicitly formulating the conditions of this process, have suggested the secondary character of the suffix accentuation in this present type; cf., for instance, the remark by Kellens (1984: 121, note (8)): "Le sens ne permet pas de considérer mriya- comme le passif de mara-: l'accent suffixal paraît donc secondaire". On the partial overlapping of the -ya-stems built on some Cr and CRÏ roots (vacillation CRïya-/CRiya-), see Kulikov 2005.

9 jayate is qualified as an original passive, e.g. in Mayrhofer's grammar (1965: 93, 93), albeit not consistently; see Hauschild 1965: 216; cf. also Hartmann 1954: 186f.; Etter 1985: 215, fn. 290; 245; Kellens 1984: 126ff., note (15); Werba 1997: 288 ("intr. Pr. [=Pass.]").

(2) SB 5.3.5.17

agner vai dhUmo ja-ya-te,

fire:ABL.SG verily smoke:NOM.SG bear-YA-3SG.MED

dhUmad abhram abhrad vrstih

smoke:ABL.SG cloud:NOM.SG cloud:ABL.SG rain:NOM.SG

'Verily, from the fire the smoke arises, from the smoke the cloud, from the cloud the rain.'

The most important piece of evidence for a non-passive analysis of jayate is the lack of constructions with the instrumental of the agent (= the one who begets), which would be typical for a true passive construction (see Hock 1985-86: 90, fn. 5), as in (1a):

(1a) *tvaya vipro ja-ya-te

you:lNS poet:NOM.SG bear-YA-3SG.MED 'The poet is born by you (o fire).'

Besides, there are no good phonological reasons which could explain the supposed accent shift: *jayate—> jayate. Most likely, jayate belonged with anticausatives, not with passives, from the very beginning, meaning 'come into being, arise'. Then, how the widely spread passive analysis of jayate can be explained? I presume it may have emerged under the influence of the passive morphology of its translations in European languages, such as Engl. is born, Germ. ist geboren, Fr. est ne. Note, incidentally, that the Russian translation of this Vedic verb seems to be free of such dangerous side effects: Rus. рождаться 'be born' is a nonpassive intransitive (anticausative), which cannot be employed in passive constructions of the type 'X is born by smb.'

4. -ya-presents with fluctuating accentuation

There are some twenty Vedic -ya-presents attested with boot root and suffix accentuation, cf. mucya-te / mucya-te 'be released, become free', ksiya-te / ksiya-te 'perish, disappear', etc. (hereafter referred to as '"-ya-presents"). According to standard Vedic grammars, this fluctuation is not random only in case of pacyate 'is cooked' vs. pacyate 'ripens' (as in RV 1.135.8 pacyate yavah 'the barley ripens'). In what follows, I will concentrate on synchronic features and diachronic origins of this verbal class.

4.1. Historical distribution of accentuation in Vedic texts

As noticed above, the accent fluctuation of the type mucya-te / mucya-te does not follow any semantic regularity (except for pacya-te). The few attempts at explaining the place of the stress in terms of the passive/non-passive distinction (cf. Gonda 1951: 98f.), parallel to the opposition pacyate 'is cooked' vs. pacyate 'ripens' clearly faltered. We find forms with different accentuation in nearly identical contexts and even parallel passages which differ only in accentuation; cf. RV 10.152.1 jiyate = AV 1.20.4 jiyate. The accentuation of the '-ya-presents is not random, however, as Table 2 below shows (the numbers in superscript indicate the number of accented occurrences):

Table 2. Accentuation of the -ya-presents with fluctuating accentuation in Vedic texts

-ya-presents

-ya-

attestations with root accentuation

-ya-

attestations with suffix accentuation

Verbs of destruction and destructuring ('entropy increase')

fdhyate 'is successful'; + vi 'loses' TS1, SB1 TS1, MS1, SB1, SBK1

ksiyate 'perishes' RV1, TS2, SB2 (BAU), TÀ (act.)1 AV1, SB6, TB1

chidyate 'breaks, is cut off' TS1, ÔBK1 MS2, SB3

jiyate 'suffers loss' RV1, RV-SV1, TS1, SB1 (BAU) AV3, MS2

diryate 'cracks, is split' TS2, +MS2 SB4

puryate 'becomes full' RV1, MS1, TBm2, TÀ (act.)2 MS1, SB12, TB1

bhidyate 'breaks' RVKh.1, TS2 MS2, SB10, SBK1, KathÀ1

miyate 'is damaged, perishes' RV2, TS7, TB3, TÀ2 MS4, SB1

mucyate 'becomes free' RV1, RVKh.1, AV1, TS2 AV3, SB9, SBK3

ricyate 'is emptied'; + ati, pra 'surpasses; is left over' TS7, TB9, (+TÀ1 (?)) MS6, MS-KS1, SB12, SBK5

lupyate 'is damaged, torn' TS1 AV1, TBm1

sisyate 'is left over' +Ш1, TS2, TB3 MS1, SB10, SBK4

siyate 'falls (off)' TS-TB1, MS2, ЙВ1, TB1 SB1, TBm3

siryate 'breaks, collapses' SB1 ^U) MS1, KS1

hiyate 'is left, abandoned' TS4, SB1 MS2, SB2

Verbs of heating

tapyate 'heats; suffers' VS (act.1, med.1), TSm1, MSm1, SB3, TB2 AV6, Tgm1, SB4

dahyate 'burns' RVKh.2 TS1, MS-KS1, SB4

pacyate 'is cooked; ripens, is digested' 'ripen': RV1, +RVKh.1, SB1, TB1 'is cooked': RV1, RV-VS-TS- MS1, RV-AV1, AV2; 'ripens': MS1, SB11, SBK1

The simple regularity, which immediately follows from the above table can be formulated as follows:

• in the Rg-Veda (together with the RV-Khilani) and in the texts of the Taittiriya school (Taittiriya-Samhita, Taittiriya-Brahmana and, probably, Taittiriya-Aranyaka), '-ya-pres-ents show root accentuation;

• in the Atharva-Veda, Maitrayani Samhita, Satapatha-Brahmana and, most likely, in the texts of the Kathaka school, '-ya-presents show suffix accentuation.

While evidence provided by the RV, AV, MS, SB and the accentuated texts of the Taittiriya school is quite sufficient to make decisive conclusions on the accentual patterning of the '-ya-presents in these texts, the case of the Kathaka (KS) is less clear. The overwhelming majority of the -ya-presents occur in the unaccentuated parts of the text in ed. Schroeder; evidence consists of only three attestations (dahyamana KS 10.5:130.6-7, ati-ricyate KS 14.10:209.6, api-sTryate KS 35.16:62.2). Yet, in spite of the scarcity of attestations, the three accented occurrences (to which one form in the Katha-Aranyaka may be added) as well as the close affinity of the language of the Kathaka and Maitrayani schools lead to the assumption that the corresponding dialects belong together as far as the accentual patterning of the '-ya-presents is concerned.

Still more problematic is the position of the dialect of the Vajasaneyins. The only '-ya-pres-ent which occurs accented in the VS is tapya-te (tapyamanaya VS 39.12).10

The aforementioned distinction holds foremost for the larger semantic class of '-ya-pres-ents, which includes verbs referring to (spontaneous) destruction and some related processes (for a detailed semantic analysis of the '-ya-presents see Section 4.4 below). More intricate is the case of the second, smaller, semantic class, which includes verbs of heating. In the RV, the place of the accent in pacya-te depends on its meaning ('be cooked/ripen'). The SB and MS have generalized the suffix accentuation (except for one root-accented occurrence in the SB), as in the case of the verbs of destruction; the root accentuation of the only accented occurrence in the TB matches both its semantics ('ripen') and the rule of accent placement in the Taittiriya and therefore does not prove anything. Likewise, dahya-te (unattested in the RV) essentially follows the model of the verbs of destruction, except for one occurrence in the TS. Most complicated is the situation with tapya-te. In the AV, all the accented forms bear accent on the suffix, whilst in the TB the accent is on the root, which meets our rule. The accent placement in the Samhitas of the Yajurveda (TS, MS) seems to be random; note, however, that both occurrences which do not meet the rule (TSm 3.2.8.2 anu-tapyamana- « anu-tapyamana-MSm 2.3.8:37.1) appear in a mantra — that is, in the older language (which may represent an earlier situation as compared to what we observe at the later stage of the development of the same Vedic dialects, in Vedic prose). All the three root-accented occurrences attested in the SB are imperatives (met with in one passage), while the suffix accentuation is attested in indicative forms.

From the rule formulated above it immediately follows that (i) the suffix accentuation of the '-ya-presents in the AV, MS and SB does not suggest their passive value or any particular semantic difference from the corresponding forms with the root accentuation attested in the RV(Kh.) and Taittiriya — contra Gonda 1951; (ii) there are no good reasons to emend the suffix accentuation in these texts on the basis of non-passive semantics (cf. Insler 1987: 62f. on AV ksTyate). The accent fluctuation does not depend on the semantics of the -ya-presents in question, but represents a difference between Vedic schools/dialects.

4.2. Exceptions to the general rule

Exceptions to our rule are relatively few among the 'entropy increase' verbs (shown with the outline letters in Table 2 above); for convenience, they are summarized in Table 3 below:

10 The nonce formation tápyate (dat.sg. act.prt.), attested in the same passage, is a secondary replacement of the class I active participle (tápate) and thus cannot serve as evidence for the accentual patterning of tápyá-te in this text; moreover, it may even have triggered the root accentuation of the adjacent tápyamanaya.

Table 3. -ya-presents with fluctuating accentuation: exceptions to the general accentual pattern

-ya-presents -ya- root accentuation instead of the expected suffix accentuation -ya- suffix accentuation instead of the expected root accentuation

Verbs of destruction and destructuring ('entropy increase')

fdhya-te 'is successful'; + vi 'loses' vy-fdhyai SB 2.1.2.4 rdhyate TS 1.5.2.2

ksiya-te 'perishes' ksiyate SB 14.4.2.28, ksiyeta SB 14.4.3.7 apa-ksïyate TB 1.5.10.5

chidya-te 'breaks, is cut off' vy-ava-chidyai SBK 2.8.3.18

jiya-te 'suffers loss' jiyate SB 14.4.3.23

dirya-te 'cracks, is split' +diryeta MS 2.1.82* : 9.14, 15

purya-te 'becomes full' prati-puryeta MS 3.2.2:17.11 a-puryate TB 1.5.10.5

mucya-te 'breaks' mucyatai AV 8.8.6

lupya-te 'is damaged, torn' lupyate TBm 2.8.8.2

sisya-te 'is left over' +uc-chisyatai AV 2.31.3

siya-te 'falls (off)' ati-siyante MS 2.6.12" : 64.1, 6, ava-siyante SB 3.2.6.8 ava-sïyante TBm 3.12.7.2-33*

sirya-te 'breaks, collapses' siryate SB 14.6.9.28 etc.

hiya-te 'is left, abandoned' hiyate SB 3.6.2.14 ~ 3.6.2.15

Verbs of heating

tapya-te 'heats; suffers' anu-tapyamanah MSm 2.3.8:37.1, tapyadhvam, tapyasva2x SB 6.1.3.2-4 anu-tapyamana TSm 3.2.8.2

Most exceptions fall into one of the following types:

(1) eleven non-indicative forms with suffix accentuation instead of root accentuation:

• subjunctives: vy-rdhyai SB 2.1.2.4, +vy-ava-chidyai SBK 2.8.3.18, mucyatai AV 8.8.6, +uc-chisyatai AV 2.31.3;

• optatives: ksiyeta SB 14.4.3.7, +diryeta MS 2.1.82x : 9.14, 15, prati-puryeta MS 3.2.2:17.11;

• imperatives: tâpyadhvam, tâpyasva2x SB 6.1.3.2-4.

The tendency to bear the accent on the root in the non-indicative forms of -yâ-presents in the AV, MS and SB was by no means a strict rule, however: we find subjunctives and optatives with the accent on the suffix as well, cf. jïyéta MS 1.6.10:103:2, dïryéta SB 4.5.10.72x, bhidyéyûtûm SBK 4.9.4.15, etc. On the assumption that in the corresponding Vedic dialects the accent was retracted from the suffix to the root, the root-accented forms listed above can be regarded as preserving the original accentuation (see Section 5.2 below).

(2) six occurrences attested in the YVic mantras: anu-tápyamanah MSm 2.3.8:37.1 « anu-tapyá-mänä TSm 3.2.8.2, lupyáte TBm 2.8.8.2,11 ava-siyánte TBm 3.12.7.2-33x. One may assume that the accent shift in some YVic dialects was completed by the beginning of the Brahma-na period, while the mantras attest the transitional period and vacillation in accent placement.

(3) four exceptions in the last chapters of book 14 of the SB [= BÄU] (ksiyate SB 14.4.2.28 [= BÄUM 1.4.28], ksiyeta SB 14.4.3.7 [= BÄUM 1.5.7],12 jiyate SB 14.4.3.23 [= BÄUM 1.5.23 = BÄUK 1.5.15], siryate SB 14.6.9.28 [= BÄUM 3.9.28] = SB 14.6.11.16 [= BÄUM 4.2.6] = SB

14.7.2.27 [= BÄUM 4.4.27]) must be due to the late character of the text, which not infrequently gives erroneous accents; cf. íyámana- SBv 14.7.1.14 [= BÄUMv 4.3.14], manyasai SB 14.6.9.26 [= BÄUM 3.9.26], manyáte SB 14.9.2.7 [thus mss.; ed. Weber mányate], sájyate SB

14.6.9.28 [v.l. apud ed. Weber].

(4) for siya-te, there may have existed additional semantic rules which determined accent placement in some usages, see Kulikov 2011, s.v. for details.

(5) only four exceptions seem unmotivated: rdhyáte TS 1.5.2.2, hiyate SB 3.6.2.14 « 3.6.2.15, apa-ksiyáte and a-püryáte in TB 1.5.10.5.

4.3. Instances of semantically motivated accent shift

To sum up, for the majority of '-yá-presents the accent fluctuation does not involve any semantic or syntactic features. Thus, the standard explanation of the accent shift in -yá-pas-sives as motivated by the non-passive (reflexive or anticausative) syntax13 finds no or little support in the linguistic facts. The only clear instance of an opposition correlated with the place of accent is pácyá-te, employed in the sense 'be cooked' or 'ripen', depending on its accentuation (on the suffix vs. on the root; for references, see Kulikov 2001, 2011, s.v.). This correlation seems to hold true only for the language of the RV, however (where, incidentally, the root accentuation is attested only once, at RV 1.135.8, against three instances of suffix accentuation). Note, furthermore, that the semantic opposition 'be cooked' ~ 'ripen' does not amount to the passive/non-passive distinction, but suggests an idiomatic change (lexi-calization).

Another instance of semantic motivation may be dáhyá-te, which occurs with the root accentuation in the RVKh. ('burn [by itself]') and with the suffix accentuation in the Vedic prose (TS, MS-KS, SBK) ('be burned [by fire]'), but this semantic distinction is too subtle and evidence rather scant. Besides, four of the five occurrences follow the accentual patterning attested for verbs of destruction, the only exception being TS 5.5.2.3 dahyámana.

4.4. Semantics of the '-ya-presents

The middle -ya-presents with fluctuating accentuation represent, in a sense, a 'bridge' between -yá-passives and non-passive middle '-ya-presents. This small class reveals a remarkable semantic and structural similarity.14

11 The suffix accentuation of lupyate could also be triggered by the adjacent -ya-passives appearing in the same passage.

12 This occurrence belongs to the first group of exceptions as well.

13 Cf. Panini's sutra 6.1.195 acah kartr yaki 'before [the passive suffix] -ya- [in verbs with the roots ending] in a vowel (aC-) [the root optionally bears the accent if the verb is employed] in the reflexive [usage]'.

14 For a discussion of the semantic features of the '-ya -presents, see Kulikov 1998a, 1998b.

The main subclass of the '-ya-presents can be defined in semantic terms as follows. A good deal of these verbs denote processes of spontaneous destruction: breaking, bursting (in the Brahmanas often said of sacrificial vessels), splitting, as well as destruction in general ('disappear', 'perish', etc.), cf. ksiya-te 'perish, disappear'; chidya-te 'break, cut off'; dirya-te 'crack, split, burst'; bhidya-te 'break, split'; miya-te 'damage, perish'; lupya-te 'damage, tear'; sirya-te 'break, collapse'. The definition of another subclass (mucya-te, etc.) poses some problems. Intuitively, the meanings of mucyate 'becomes free', siyate 'falls', sisyate 'is left over', etc. are rather close to the semantic domain of destruction, but their common denominator is difficult to capture. Yet, one may argue that they all denote a process when an element ceases to be incorporated into a system or structure — for instance, some part(s) of an object break off and fall down — which, ultimately, results in the destruction of a system. Specifically, mucyate 'becomes free, is released' can be determined as 'ceases to be bound, included into a bound system'; sisyate 'is left over' and hiyate 'is abandoned, is left over' « 'remains outside a structure'; siyate 'falls (out)' « 'ceases to be included into a structure through falling out of it'. At first glance, rdhyate 'is successful, fulfilled, goes well' does not belong to this semantic type; but its meaning changes to the opposite in compounds with the preverb vi: 'is deprived of [a property], loses', i.e. 'ceases to be connected with some (structural) elements'; cf. jiyate 'suffers loss', which is very close. ricyate belongs here both when employed as a simplex ('is emptied' ~ 'is deprived of its content') and with the preverbs ati, pra ('surpasses, is redundant' « 'goes beyond the scope of a structure'; 'is left over'). For this subgroup I propose the tentative label 'verbs of destructuring'.

In my view, we are able to determine an even more general semantic feature which encompasses the meanings of both 'destruction' and 'destructuring'. All these verbs denote spontaneous "fatal" processes which result in destroying some natural or artificial system or organism, and, to put it in general terms, in the entropy increase.15

In this semantic perspective, instructive is the present siya-te, whose semantics does not amount to falling down. In one of its usages, siya-te refers to a particular kind of falling, which accompanies natural decay, growing old: falling out of hairs, teeth, etc., i.e. typical instances of entropy increase. Cf. also JB 1.1 bhasmavasiyate 'some borings fall down [from the piece of kindling wood being churned]', on which Bodewitz (1973: 21f., note 4) comments that "bhasma refers to wooden dust falling of the wood sticks during the churning, a product of erosion [emphasis is mine. — L. K.]" — again, a typical instance of entropy increase.

The verb pr 'fill' cannot be included into the class of 'entropy increase' verbs in any of its usages. Yet, in the compound with the preverb a it functions as the counterpart of an 'entropy increase' verb, ksi (with the preverb apa): a-pr 'wax' and apa-ksi 'wane' denote opposite changes of the half-moon.

The semantic affinity of the 'entropy increase' verbs is also supported by the fact that they often co-occur in texts. To mention a few passages: TB 1.5.10.5, SB 1.7.2.22, 2.1.3.1, 2.4.4.18, 19, 8.4.1.10, 10.4.2.17 (co-occurrence of apa-ksi and a-pr); PB 6.7.15 (ava-chid, vy-rdh, ji); TS 7.2.1.4, MS 1.6.10, 1.8.7, AV 10.1.32 (muc, ha); SB 3.1.1.3 ((abhy)ati-ric, sis), TS 3.2.9.5, AB 6.2.6 (lup, ha).16

The second, smaller, semantic class of '-ya-presents includes three verbs of heating: tapya-te 'heat, suffer', dahya-te 'burn', pacya-te 'cook; ripen'.

15 For a detailed discussion of this semantic feature, see Kulikov 1998a.

16 Cf. Gonda 1959: 204.

4.5. Paradigmatic features

The most remarkable paradigmatic feature shared by the -ya-presents of entropy increase is their opposition to transitive-causative presents with nasal affixes (cf. ksiyate — ksinati, ksinoti, chidyate — chinatti, lupyate — lumpati, etc.). By contrast, the three '-ya-presents of heating are opposed to class I presents (tapya-te — tapa-ti, dahya-te — daha-ti, pacya-te — paca-ti); for details, see Kulikov 2011, chapter C.III.2.

It seems that the paradigmatic similarity of the -ya-presents could be an important feature of this verbal class, which supported their semantic affinity and, in some cases, could even trigger the rise of secondary transitive presents with nasal suffixes for some -ya-presents of this class; cf. such formations as sinasti, simsati.

4.6. Phonological similarity

Some types of phonological structures are particularly common among the -ya-presents of entropy increase, while some others are unattested. Specifically, five stems (one third) show the structure CTya-, four stems belong to the type CiCya-, three stems show the structure CT/urya-. By contrast, all the three verbs of heating are derived from CaC roots, uncommon among verbs of 'entropy increase'. It is of course impossible to posit a strict correlation between phonological structures and semantic classes; however, the phonological similarity could additionally support the structural affinity of the verbs in question17 and cause accent shift in some -ya-presents of similar structures, even in spite of different semantics, in particular, in -viyante 'are impregnated' TS 6.1.7.1 (Vvya/vT), Tyamana- 'speeding, driving' (Vya) MSm 2.6.11:70.12, v.l. [three mss.], SBv 14.7.1.14 = BAUMv 4.3.14); cf. esp. the parallelism vya/vT, ya/T ~ jya/ji.

4.7. Accent fluctuation of the type mucya-te / mucya-te in a diachronic perspective

The features shared by the -ya-presents (semantics, non-passive syntax, opposition to transitive nasal presents, partial phonological similarity) belong to different layers of the language structure and are essentially independent of each other. This implies that the similarity of -ya-presents cannot be mere coincidence, and they form a morphologically relevant verbal class, rather than a random group. Their semantics (entropy increase, heating) seems to be the main parameter organizing these verbs to a structural class and, eventually, determining their properties. For instance, this feature could trigger the emergence of the nasal presents sinasti, simsati (Br. +), built as transitive-causative counterparts of sisya-te.

The 'entropy increase' semantics could also influence the accentual behaviour of a -ya-present even in the cases where it was registered only with some preverbs; cf. rdhya-te 'be successful', which changes its meaning to the opposite ('be deprived, lose') in compounds with vi, and therefore follows the accentual pattern of the verbs of entropy increase. The same explanation probably holds true for the suffix accentuation of the non-passive vi-lipyate (MS) 'comes unstuck [and falls off]' (~ simplex lipya-te 'stick, smear').

17 Recall the old theory of 'rime-words' and 'rime-ideas' (Reimwortbildungen) (Bloomfield 1895; Wood 1907/ 1908; Güntert 1914: 30ff. [on chid/bhid]; 65f. [on ksi/mi]). Note, incidentally, that the first of Wood's lists of rime-words (op.cit., 142f.), labelled 'dwindle' group, includes all the five aforementioned Ciya- presents of the 'entropy increase': ksiyate, [jiyate] (Wood gives only the nasal present jinati), siyate, miyate, hiyate (though Wood does not mention accent fluctuation).

Furthermore, the parallelism between a-pr 'wax' and apa-ksi 'wane' (the latter of which belongs to the 'entropy increase' type) has probably triggered changes in accent patterning of -purya-te, in spite of the fact that this present does not show any meaning related to entropy increase.

On the other hand, some middle -ya-presents with the root accentuation were not grouped with the -ya-presents (and hence did not change their accentuation) if their semantic affinity with the entropy increase class was not supported by other features. Thus, for instance, padya-te 'fall', albeit similar to siya-te in meaning, does not show other features of the -ya-presents referring to 'entropy increase' (note, in particular, the root structure CaC and the lack of a transitive-causative counterpart with the nasal affix) and does not change to padya-te in the dialects of the AV, MS and SB.18

The "intermediate" position of -ya-presents between -ya-passives and non-passive middle -ya-presents probably results from their peculiar semantics. Judging from their non-passive meanings and syntax (see above) as well as from their root accentuation in the Rgveda, originally these formations probably belonged with the class IV presents. Later on, in some contexts they could be re-interpreted as passives (in accordance with the scenario: 'breaks' ^ 'is broken [by smb.]'; 'becomes free' ^ 'is released', or the like) and, due to the increasing productivity of the -ya-passives, undergo accent shift in several Vedic dialects — in particular, in the dialects of the AV, Maitrayani-Kathaka and SB.19

The great number of exceptions in the Yajurvedic mantras and the root accentuation of tapyamana- in the Vajasaneyi-Samhita as against the prevalent suffix accentuation in the Sata-patha-Brahmana, which belongs to the same Vedic school (White Yajurveda), may point to the fact that the accent shift from the root to the suffix in the corresponding Yajurvedic dialects was only completed by the period of Vedic prose (Brahmanas properly speaking).

5. Concluding diachronic remarks on the accentual history of Vedic -ya-presents

5.1. The original accentuation of (middle) -ya-presents

It is commonplace in Vedic studies to assume that all -ya-presents, irrespectively of accentuation and diathesis (i.e. both -ya-passives and class IV presents), go back to one source,20 and the correlation between accentuation and the passive/non-passive distinction is

18 Not counting three occurrences with suffix accentuation in the late SB (books 11 and 14).

19 Our knowledge of the Vedic dialectology is still rather poor (for a systematic treatment of the issue, see Witzel 1989), and we are thus far unable to draw any decisive conclusions on the localisation of the dialects in which this accent shift was operative. Nevertheless, relying upon Witzel's preliminary outline, one might tentatively assume that one of the centres of this process was the Kuru region (KS, early SB), whereas the dialect(s) of the Pancala (where the TS can be tentatively located) were more conservative and have preserved root accentuation.

20 I will not discuss here at length the highly controversial issue of the origin of the present suffix -ya-. Suffice it to mention that the morpheme -ya- must be historically identical in the middle class IV presents (with root accentuation) and -ya-passives, while some of the active -ya-presents may go back to a different (denominative?) source. For a survey of possible sources of the present suffix -ya- (resp. PIE *-ie/o-), see, in particular, Lubotsky 1985 (who demonstrates that the PIE source of Ved. sus, i.e. *H2sus, must be an adjective, not a verbal root); Barton 1986: 143, fn. 27 ("many of the *f/o-intransitives are doubtless denominal in origin"); Rasmussen 1993: 480ff.; and Kort-landt's (1981: 127f.), who advocates the genetic relationship between Vedic -i-aorists, -aya-causatives and -ya-passives: the former may go back to "a deverbative noun of the type *kwori [> Ved. (a)kari. — L. K. ], which could

an Indo-Aryan innovation. See, for instance, F. M. Müller 1864: 582; Delbrück 1874: 168; J. Schmidt 1875: 256f.; Hillebrandt 1880: 342f. [= Kl.Schr., 606f.]; Speijer 1896: 49, §168; Brug-mann 1902: 527f.; 1916 [Grundr.2]: 185; Reichelt 1902: 80; M. Leumann 1940: 231ff. [= Kl.Schr., 321ff.]; Gonda 1951: 721 et passim; 1971: 90f.; Thumb/Hauschild 1959: 333ff.; Strunk 1967: 78. This assumption, based, above all, on the intransitivity of the majority of -ya-presents, immediately raises the question on the original place of accent: did they bear accent on the suffix or on the root? Evidence is controversial: the zero grade of the root may betray the original suffix accentuation, while the increasing productivity of -yá-passives and the archaic non-productive character of many class IV presents rather point to the root accentuation. Most scholars considered the root accentuation in the class IV presents secondary as against the suffix accentuation in -yá-passives; see already Benfey 1865: 1783 [= Kl.Schr. II, 141]; 1866: 196; Saussure 1877; Froehde 1881: 172; Diels 1913: 4.22 Very plausible is Kurylowicz's (1952: 114f.) assumption that accent retraction to the root in class IV presents was due to the influence of class I presents with phonologically regular full grade root, thus: *asy-á--> ásya-,

*pasy-á—> pásya-, etc. on the model of *gacháti ^ gáchati, where a < *n has been reanalysed as full grade (Saussure 1879: 174 [= Rec., 163]);23 cf. also Gonda 1951: 92 ("the accentual differentiation of the -ya-verbs was attended by a partial leaning towards other thematic root-accented presents"); 1971: 91.

On the other hand, Delbrück (1897: 435f.) argued for the opposite development ('-ya—> -yá-); cf. also Kümmel in LIV 637, note 2 [ad got. paursjan* 'dürsten'] s.v. *ters-.24 No doubt, the system of -ya-presents was subject to a number of analogical accent shifts of both kinds ('-ya-^ -yá-), even within the historical period; see Section 4 on -ya-presents with fluctuating accentuation and Kulikov 2011 on the supposedly passive origin of -ingya-te (-ángya-te) 'move, stir' and irya-te 'move'.

The difference in accentuation between (middle) class IV presents and -yá-passives is clearly secondary. We can only speculate why the passive subclass has generalized the suffix accentuation (which probably was original), while non-passives have retracted accent to the root. This accent shift may have started in a few old non-passive -ya-presents, in which zero and full grade could not be distinguished (cf. mányate), or where the full grade was introduced instead of the phonetically impossible zero grade (as in padyate ^ **pdyáte) or in order to avoid morphological opacity (nahyati/-te ^ **ahyáti/-te < *nhiáti/-tai (?) or asyati/te ^ *siáti/-tai).25 The root accentuation could be introduced for such presents in accordance with Saussure's rule (reformulated by Kurylowicz for -ya-presents, see above) and subsequently generalized for all non-passives.

5.2. The genesis of accentuation of -ya-presents: a possible scenario

A possible relative chronology of accent shifts in the Vedic middle -ya-presents can be summarized as follows:

itself be used predicatively", whereas causatives and -ya-passives are supposed to be derivatives from this noun. For a comprehensive survey of existing hypotheses, see also Kulikov 2011.

21 "... everybody knows the intimate connection between the -ya- class and the 4th present-class".

22 Saussure (1879: 234 [= Rec., 219], with fn. 1) even assumed that the active -ya-present ranyati (with the irregular suffix accentuation) might be a trace of the original accent placement in this present type.

23 Essentially the same hypothesis was already proposed by Froehde (1881: 172), albeit in a less explicit form.

24 "Offenbar *tfs-ie- mit verschobenem Akzent wie im Ved. (oder deutet dieser Fall auf grundsprachliches Alter des Wurzelakzents bei primären ¿e-Präsentien?)".

25 Cf. LIV 227, note 2 s.v. *Hnedh- and LIV 242f., note 4 s.v. *hies- .

Proto-Indo-Aryan

I. Accent retraction in non-passive -ya-presents. Accentual differentiation of -ya-passives and (middle) non-passive -ya-presents

II. Accent shift in the type mriyate. *Cf-ia—> Criya-

Old Indo-Aryan (Vedic) dialects

III. Accent shift in some (middle) non-passive -ya-presents (in the dialects of the AV, MS-KS and SB)

V0-iá-

V0-iá-te

(e.g. hanyáte 'is killed')

Vé-ia-te (e.g. mányate 'thinks')

V0-iá-te

(hanyáte)

Criyá-te (mriyáte)

Vé-ia-te (mányate)

V0-iá-te

(hanyáte)

Criyá-te (mriyáte)

V0-iá-te

(mucyáte AV, MS)

Vé-i

ia-

(mányate)

te

te

Alexei Kassian

Institute of Linguistics of the Russian Academy of Sciences (Moscow)

Some considerations on Vedic -ya-presents

The paper by L. Kulikov consists of two parts: 1) detailed discussion upon the peculiarities of some specific Vedic verbal stems, and 2) reconstruction of the -ya-present pattern for Proto-Indo-Aryan. Below I will touch upon the second — comparative — portion of the paper (§5).

Kulikov discusses two functions of the Vedic -ya-suffix: medial present tense of the 4th class with nonpassive intransitive function, i.e. the V'-ya-te pattern (however, the frequent active V'-ya-'1 pattern is excluded from the analysis) and the regular passive forms of the V-ya-te pattern. The author concludes that the original Proto-Indo-Aryan pattern of the middle voice was *V-ya-te, which later split into two accentual and semantic types.1

1 Kulikov labels this proto-level as "Proto-Indo-Aryan", but, in fact, some OPers. and Avest. evidence may prove that the

The only explication of such a split proposed by Kulikov is J. Kurytowicz's idea that non-passive mid-

grammaticalization of the ya-suffix as an exponent of the passive voice goes back to the Proto-Indo-Iranian level. On the contrary, if we reject OPers. and Avest. data, an accurate term should be "Proto-Vedic", not Proto-Indo-Aryan in general. On the other hand, the Dardic language Shina shows the same grammatical-ized passive voice in -izh- (= -ij-), Bailey 1924: 29, Schmidt & Ko-histani 2008: 145 ff., 194 f. This fact should prove the Proto-Indo-Aryan antiquity of such a grammaticalization, if Shina -izh- does indeed contain *-ya- (as is suspected by V. A. Dybo, see his reply below, although I would rather suppose that Shina -izh- reflects an innovative formation in Dardic). Below, for the sake of convenience, I will use Kulikov's term "Proto-Indo-Aryan" in regard to the grammaticalized passive voice in -ya-.

In any case, it is important that such a grammaticalization is an inner Indo-Aryan (or Indo-Iranian) innovation. E.g., in the Balto-Slavic group (the closest linguistic relative of Indo-Iranian) j-praesentia normally seem to be associated with transitive or