The influence of Dravidian on Indo-Aryan phonetics

Paper read at the XIVth World Sanskrit Conference, Kyōtō 2009 The influence of Dravidian on Indo-Aryan phonetics Ferenc Ruzsa Eötvös Loránd University, Budapest, Hungary 1 It is now a widely accepted opinion that Dravidian languages had an important effect on the development of Sanskrit. Most frequently mentioned aspects of this influence are loan-words, the appearance of retroflexion, the extensive use of gerunds and the iti construction.2 Perhaps the almost complete loss of old syntax (notably of subordinative sentences) and the appearance of a completely new syntactical structure, generally but misleadingly called ‘compounds’3 might be considered even more important. Later in the Prakrits the loss of the ātmanepada conjugation and the dual, the disappearance of past finite verbal forms and the reduction of the modes to optative and imperative only4 can all be explained in this way. In this paper only one aspect, phonetics will be investigated, and in this wider context: from the earliest Vedic up to late Middle Indic. It will appear that all the important developments in Indo-Aryan phonetics during these some twenty centuries could be interpreted as due to a single constant and strong influence – that of a language with a phonetic structure similar to Tamil. 1. Retroflexion The retroflex–dental opposition is very strong in Dravidian languages; and in IndoIranian it was absent, as in all Indo-European languages.5 Retroflex pronunciation of some dentals (puṣṭi) and also retroflex phonemes (gaṇa) appeared already in the Ṛgveda6; this became more and more widespread later, the number of retroflex phonemes increasing at least to the end of the Prakrit age. The most important source of this is the loss of the r or ṣ starting the retroflexion (duṣṭa → duṭṭha, varṇa → vaṇṇa). Also r frequently causes retroflexion in plosives (prati → *praṭi → paṭi) and in later prakrits we find many unexplained retroflexes (e.g. nūṇaṁ in Jaina Mahārāṣṭrī, Jacobi 1886 [1967]; and even ṇūṇaṁ in Mahārāṣṭrī, Bubenik 1996: 60). 1 My researches were supported by the Hungarian National Research Fund, OTKA, project no. K 75550. Burrow 1955: 373–88., Kuiper 1974: 146, Hock 1996: 18; 24–7, Bryant 1999: 61–5, Kobayashi 2004: 17. 3 Killingley–Killingley 1995: 42–7. The Dravidian origin of the long compounds in Indo-Aryan languages has been first seriously suggested only quite recently in an excellent paper by Scharfe (2006). 4 Of course in Pali we still have a past tense, and also the conditional and the injunctive (in prohibitive use only). The loss of the injunctive in later Prakrits seems to be unmentioned in the grammars available to me (Pischel 1981, Woolner 1928 [1996], Bubenik 1996, Hinüber 2001), so I checked it in the Prakrit texts of the akuntalā (Williams 1876 [1961]) and in the Satta-saī (Basak 1971) – it does not occur. 5 At least it is not mentioned for any other Indo-European language in Ramat–Ramat 1998. 6 From now on abbreviated as RV, always referring to the received text; quoted from Aufrecht 1861–63. 2 1 Although in the RV most retroflexes are phonetically determined (pūrṇa), i.e. they could be but allophones of dentals, still some are clearly not (jaṭhara). Especially important is the fact that the loss of the Proto-Aryan voiced sibilants (and they are absent already from our very earliest texts) occurred only after the phonemicization of retroflexion – otherwise we would have nīda and īde instead of nīḍa and īḍe in Sanskrit (nīḷa and īḷe in the RV), from Proto-Vedic *niẓḍa and *iẓḍăi (from *nizda and *izdăi by the RUKI-rule). Similarly such forms as ānaṭ (← *ānaṣṭ ← āna +t), very frequent in the RV, prove that retroflex phonemes were already part of the language when final consonant clusters got reduced to a single consonant. Although Deshpande (1993) showed convincingly that the RV-text we have significantly differs from the original phonetically and notably in the retroflexes, still we can say on the basis of the last two phenomena (nīḍa, ānaṭ) that the earliest form of the language documented in any way already had retroflex phonemes. An important and in the RV frequently mentioned family of ṛṣis, the Kaṇvas had a phonemic ṇ in their name. It may be Pakritism for *Kṛṇva (Mayrhofer 1992–96); but that would only show that their everyday language has suffered even more Dravidian influence (ṛ→a) than what can be shown in Vedic. Of course it is far from improbable that some of the older hymns of the RV were composed in a language earlier than that (in Proto-Vedic or even in Proto-Aryan), but this fact and this phase of the language must remain unknown to us. 2. Vowels Proto-Vedic had the vowels a, ā, i, ī, u, ū, ṛ, ṝ, ḷ and the diphthongs *ăi, *āi, *ău and *āu. Proto-Dravidian had no vocalic ṛ and ḷ nor diphthongs, but had e, ē, o and ō. Already standard Sanskrit replaced *ăi and *ău (and also in sandhi *ăi-a and *ău-a) with ē and ō. The long diphthongs were shortened into ai and au; although not ProtoDravidian phonemes, these do exist in Old Tamil. Then in Middle Indic the remaining diphthongs, ai and au (and also aya and ava in any position) change to ē and ō. In certain environments these are shortened, so we have short e and o; and with the disappearance of vocalic ṛ and ḷ the end result is exactly the original Dravidian set of vowels: a, ā, i, ī, u, ū, e, ē, o, ō. All these changes were complete already by the age of Pali. The reason for the appearance of short e and o in all Middle Indic dialects is the law of mora, according to which long vowels get shortened before consonant clusters and geminates (e.g. Pali sĕṭṭhi and ŏkkamati for Sanskrit rēṣṭhin and avakramati). A similar rule seems to have been operative also in Proto-Dravidian, although with a somewhat restricted scope, i.e. within a morpheme.7 3. Final consonants Dravidian speech habits do not favour final consonants. In modern Tamil all words end in vowels (Schiffman 1999: 4); in Old Dravidian some consonants did occur as 7 Steever 1998: 17–8. This parallellism is far from clear, see Kobayashi (2004: 9, 165–6, 176, 182). 2 finals – from among those allowed only n, m and r could also take this place in Old Indo-Aryan. The loss of word-final consonants is already quite marked in Vedic: normally only a single consonant can remain at the end of a word. But the reappearance of a dropped s after n in sandhi shows that the elision of the last consonant was (at least in this case) a recent phenomenon, not yet complete. E.g. devā s tvám for devān + tvám, earlier *devāns; but abruvan tád, where no s was deleted. Another example is ānaṭ, quoted earlier, showing that the loss of final clusters was more recent than the appearance of retroflex phonemes. So it all happened already in India, under the probable substrate influence of Dravidian. And the remaining single consonants have also lost most of their attributes: no trace of aspiration or voicedness can be found (e.g. *triṣṭúbhs → triṣṭúp), place of articulation very weak (completely lost in the case of m and s). E.g. tásmāj jātā for tásmāt + jātā; índra ca for índras + caś and for m we have the anusvāra ṁ, that means "a nasal without defined articulatory position". This tendency continues so that in the Prakrits final consonants completely disappear (using techniques very similar to those used by Tamil to eliminate them, i.e. elision or adding a vowel, a/u); only nasals leave a trace as nasalisation of the final vowel – as in spoken Tamil to-day. 4. Sibilants Proto-Dravidian had no sibilants at all, but already in Old Tamil the phoneme c was frequently pronounced as s or initially and intervocally (so Sanskrit iva is spelt Civan in Tamil, but the pronunciation is ivan). Proto-Vedic must have had six: s, , ṣ, and their voiced allophones z, ź and ẓ.8 By the age of the RV, the voiced sibilants have all disappeared, leaving different traces: probably first they were replaced by the available Old Dravidian approximants, y, ẕ and w; then y and w merged with the preceding vowel, resulting in Sanskrit ē and ō (e.g. *mazdhā → *maydhā → *măidhā→ mēdhā, cf. Avestan Ahura Mazdā). The retroflex approximant ẕ (normally transcribed as ẓ, ḻ or r), the logical substitute for the retroflex sibilant ẓ, either dropped out lengthening the previous vowel (*ni-sda → *niẓḍa → *niẕḍa → nīḍa, cf. English nest) or it came to be reinterpreted later as r (*dus-ga → *duẓga → *duẕga → durga); this reinterpretation of ẕ is not surprising, as it happened both in many Dravidian languages and also in many borrowings from Tamil to Sanskrit. Then in the Middle Indic languages the sibilants merge into one (in most dialects into s, in Magadhi into ), and even that weakens into an aspiration in clusters. As in Old Tamil: the sibilant realization of the phoneme c can occur only initially and intervocally. The pronunciation of ṣ and ẓ was not necessarily retroflex; it may have been postalveolar. And these sounds were but allophones (by the RUKI-rule) of s and z. So there were only two sibilant phonemes, s and , but the allophones showed distinctly different behavior both in sandhi and in the development of the language. 8 3 5. Consonant clusters In Proto-Dravidian consonant clusters are generally absent, only geminates and nasal + homorganic stop clusters can occur within a morpheme. Although in Old Tamil we find several consonant clusters on morpheme boundaries, but most Sanskrit clusters are impossible (Lehmann 1994: 11, Rajam 1992: 52–112). The Proto-Dravidian pattern is closely followed by all Middle Indic languages: only a single consonant can start a word, and medially only geminates and nasal + homorganic stop clusters remain. Initial consonants representing an old cluster, however, frequently get doubled in Prakrits as second members of compounds. As the previous word necessarily ends in a vowel, the position is now intervocalic, where geminates are acceptable. This phenomenon surprisingly resembles the Tamil sandhi whereby initial stops are doubled after vowels (although here obviously in order to preserve the unvoiced articulation). 6. Voicing In middle Prakritic dialects unvoiced intervocalic unaspirated stops generally become voiced (and the velars typically disappear). This corresponds fairly well with the absence of the voiced/unvoiced phonemic contrast in Tamil, where intervocally all stops become voiced9. The phoneme k, however, remains unvoiced but gets spirantized, and this χ being absent from Indo-Aryan10 might explain its loss in Prakrits. Or rather the fact that this sound (or its Aryan counterpart, h) was used for the Sanskrit aspirates. 7. Aspiration Old Dravidian had no aspirates. Already in Vedic in place of some voiced aspirates we find only h; in the Middle Prakrits h takes the place of all intervocalic aspirates. Tamil has an h-sound, the so-called āytam; and in modern pronunciation the letter k is also realized as χ intevocally. On the other hand, most Dravidian languages adopted aspirates from Indo-Aryan, but this lies outside the scope of this paper. 8. Sandhi Caldwell’s law. The general opinion is that it was operative already in Proto-Dravidian (Steever 1998: 15). Mahadevan (2003: 247–51) challenges this suggesting that Old Tamil had only unvoiced stops. But it seems that some very early loanwords to Sanskrit clearly show the modern distribution of allophones: from Tamil kuṇṭam, koṇṭai and nakar we have Sanskrit kuṇḍaḥ, kuṇḍaṁ and nagaraṁ (Burrow–Emeneau 1984, Mayrhofer 1992–96). 10 To be more exact, the jihvāmūlīya can be found in some traditions, but never in this position – only word-finally before an unvoiced guttural. 9 4 Accepting that Dravidian speech habits fundamentally influenced Indo-Aryan, we may suggest explanations for some of the more weird sandhis of Sanskrit. Although most external sandhi rules are fairly natural, some are not: notably the handling of final s. Like m it also loses its special point of articulation (i.e. as a dental), and before unvoiced consonants it can appear as a velar, palatal, retroflex, dental or bilabial spirant: χ, , ṣ, s, φ (although in most texts the velar jihvāmūlīya and the labial upadhmānīya are replaced by the neutral visarga ḥ). Old Dravidian has only one unvoiced spirant, a h-sound, and in Sanskrit the whole series was generally replaced by it (ḥ). As Dravidian h cannot occur in final position, an enunciative vowel is added in pronunciation, a typical Dravidian feature. And this is the visarga of classical Sanskrit, as pronounced by Indians. Now before voiced consonants a similar series of allophones must have existed: , ź, (ẓ), z, ś and before vowels, z11 or ẓ (after the high vowels i and u by the RUKI-rule). As Old Dravidian had three voiced approximants (y, ẕ and w), ẕ was used for the retroflex ẓśfor the rest, y before front (palatal) vowels, w in other contexts. As mentioned above in the section on sibilants, ẕ was reinterpreted as r; and this is already the regular sandhi of -is/-us before voiced sounds. The -ay/-aw resulting from -as could turn into a diphtong, ăi or ău (later monophthongized as ē or ō). Before voiced consonants -ō is the regular sandhi.12 This final ău and ăi (also of different origin, e.g. as Locative ending) have peculiar sandhi behavior before vowels. Naturally we would expect the consonatization of the second element into v and y. This must have been the Old Vedic sandhi; although the received Vedic text hides this, but the scansion shows: I.37,13b sá ha bruvat dhvann ā The end of the line is ˉ ˉ ˉ , but it should be ˘ ˉ ˘ ˉ : sá ha bruvatay ádhvan ā III.9,4c ánv īm avindan nicirāso adrúho The ˉ ˉ ˉ ˘ ˉ ending should be ˉ ˘ ˉ ˘ ˉ : ánv īm avindan nicirāsav adrúho With an initial a, av-a and ay-a later became monophthongized as ō/ē (exactly as it happens in all Prakrits in all positions; perhaps through ava → ŏvŏ → ŏŏ→ ō and aya → ĕyĕ → ĕĕ → ē). The standard spelling with an avagraha or apostrophe (-o ' and -e ') somewhat obscures the process. Before other vowels the series -av V[back] and -ay V[front] could be reinterpreted as a vV[back] and -a yV[front], where the initial v/y could be taken as the automatic, non-phonemic v/y added in many Dravidian languages before initial back/front vowels, and therefore understood as the sequence -a V, i.e. the standard Sanskrit sandhi.13 Not impossibly realized as ź before front vowels (i) And also the reflex of -as in most Prakrits. In the palatalizing Eastern (Magadhan) dialect, where took the place of s, the analoguous result is -ē (-as → -a → -aź → -ay → -ăi → -ē). 13 When -ay is not the result of -as before front vowel, -ay V[back] also occurs and the y behaves similarly, although in this position it could not be taken for the automatic Dravidian y. This might be due to analogy, very strong in Sanskrit between the i and u sounds; or another Dravidan rule may come to 11 12 5 In the foregoing eight paragraphs I hope to have demonstrated that all the major phonetic shifts in the history of Old and Middle Indo-Aryan – the appearance of retroflex phonemes, the gradual loss of diphthongs, syllabic consonants, final consonants, sibilants, consonant clusters, intervocalic aspirates and unvoiced consonants –, all these could be attributed to the influence of a Tamil-like old Dravidian language. With this hypothesis some of the more difficult sandhi phenomena could also be explained. Probably I need not emphasize what such an overwhelming influence of Dravidian would imply for the general understanding of the sources of the development of Indian culture. Ferenc Ruzsa Eötvös Loránd University Dept. of Ancient and Medieval Philosophy 1088 Múzeum krt. 4/i Budapest, HUNGARY [email protected] Bibliography Aufrecht, Theodor. 1861–63. Die Hymnen des Ṛigveda. Berlin: Ferd. Dümmler. Basak, Radhagovinda. 1971. The Prākrit Gāthā-sapta atī, Compiled by Sātavāhana King Hāla (Bibliotheca Indica 295). Calcutta: The Asiatic Society. Bryant, Edwin F. 1999. Linguistic substrata and the indigenous Aryan debate. In Johannes Bronkhorst and Madhav M. Deshpande (eds.), Aryan and NonAryan in South Asia: Evidence, Interpretation and Ideology (Harvard Oriental Series Opera Minora 3), 59–84. Cambridge: Department of Sanskrit and Indian Studies, Harvard University. Bubenik, Vit. 1996. The Structure and Development of Middle Indo-Aryan Dialects. Delhi: Motilal Banarsidass. Burrow, Thomas. 1955. The Sanskrit Language. London: Faber & Faber. Burrow, Thomas and Murray Barnson Emeneau. 1984. A Dravidian Etymological Dictionary. 2nd edition. Oxford: Clarendon Press. Deshpande, Madhav M. 1993. Genesis of Ṛgvedic retroflexion: a historical and sociolinguistic investigation. In ——, Sanskrit & Prakrit. Sociolinguistic Issues, 129–96. Delhi: Motilal Banarsidass. Hinüber, Oskar von. 2001. Das ältere Mittelindisch im Überblick (Sitzungsberichte der philosophisch-historischen Klasse 467; Veröffentlichungen der play. „Initial *y in Proto-Dravidian is often lost when it is followed by a low vowel (Burrow 1945, Krishnamurti 2003:143); this reminds us of the anomalous loss of stem-initial /y/ in Skt. prá-üga-, and of the deletion of a final /y/ before a vowel in the Sanskrit sandhi rule /-e V-/ → -a V-, (x67), but the conditioning context there is not limited to low vowels but includes any vowel.” Kobayashi (2004: 177). 6 Kommission für Sprachen und Kulturen Südasiens 20). 2nd edition. Wien: Verlag der Österreichischen Akademie der Wissenschaften. Hock, Hans Henrich. 1996. Pre-Ṛgvedic Convergence Between Indo-Aryan (Sanskrit) and Dravidian? A Survey of the Issues and Controversies. In Jan E.M. Houben (ed.), Ideology and Status of Sanskrit. Contributions to the History of the Sanskrit Language (Brill's Indological Library 13), 17–58. Leiden: E.J. Brill. Jacobi, Hermann. 1886 [1967]. Ausgewählte Erzählungen in Mahārāshṭrī. Zur Einführung in das Studium des Prākṛit. Grammatik – Text – Wörterbuch. Leeipzig: Hirzel [Darmstadt: Wissenschaftliche Buchgesellschaft]. Killingley, Siew-Yue and Dermot Killingley. 1995. Sanskrit (Languages of the World/Materials). München: Lincom Europa. Kobayashi, Masato. 2004. Historical Phonology of Old Indo-Aryan Consonants (Study of Languages and Cultures of Asia and Africa Monograph Series 42). Tokyo: Research Institute for Languages and Cultures of Asia and Africa, Tokyo University of Foreign Studies. Krishnamurti, Bhadriraju. 2003. The Dravidian Languages. Cambridge: Cambridge University Press. Kuiper, Franciscus Bernardus Jacobus. 1974. The Genesis of a Linguistic Area. International Journal of Dravidian Linguistics 3. 133–53. Lehmann, Thomas. 1994. Grammatik des Alttamil, unter besonderer Berücksichtigung der Caṅkam-Texte des Dichters Kapilar. Stuttgart: Franz Steiner Verlag. Mahadevan, Iravathan. 2003. Early Tamil Epigraphy. From the Earliest Times to the Sixth Century A.D. (Harvard Oriental Series 62). Cambridge MA and London: Department of Sanskrit and Indian Studies, Harvard University. Mayrhofer, Manfred. 1992–96. Etymologisches Wörterbuch des Altindoarischen. Heidelberg: Winter. Pischel, Richard. 1981. A Grammar of the Prākrit Languages. Delhi: Motilal Banarsidass. Rajam, V.S. 1992. A Reference Grammar of Classical Tamil Poetry (150 B.C. – prefifth / sixth century A.D.) (Memoirs of The American Philosophical Society 199). Philadelphia: The American Philosophical Society. Ramat, Anna Giacalone and Paolo Ramat (eds.). 1998. The Indo-European Languages (Routledge Language Family Descriptions). London: Routledge. Scharfe, Hartmut. 2006. Indo-Aryan and Dravidian convergence: gerunds and noun composition. In Bertil Tikkanen and Heinrich Hettrich (eds.), Themes and Tasks in Old and Middle Indo-Aryan Linguistics, 197–254. Delhi: Motilal Banarsidass. Schiffman, Harold F. 1999. A Reference Grammar of Spoken Tamil. Cambridge: Cambridge University Press. Steever, Sanford B. 1998. Introduction to the Dravidian Languages. In —— (ed.), The Dravidian Languages (Routledge Language Family Descriptions), 1–39. London: Routledge. 7 Williams, Monier. 1876 [1961]. akuntalā by Kālidāsa (Chowkhamba Sanskrit Series 12). Oxford: Clarendon Press [Varanasi: Chowkhamba Sanskrit Series Office]. Woolner, Alfred C. 1928 [1996]. Introduction to Prakrit. Delhi: Motilal Banarsidass. 8