zy zyxwv The Reconstruction of Prehistoric South Asian Language Contact FRANKLIN C . SOUTHWORTH South Asia Regional Studies University of Pennsylvania Philadelphia, Pennsylvania 19104 zyxw THE PROBLEM INDO-ARYAN LANGUAGES are spoken by upwards of 500,000,000 people spread over a contiguous area which covers Pakistan, the northern two-thirds of India, and Bangladesh (see map, FIG.1). Within this area there are various groupssome of moderate size, most extremely small - speaking languages of three other families: Dravidian, Austro-Asiatic, and Sino-Tibetan, as well as a few isolated languages (e.g. Nahali in central India, and Burushaski in the mountains of the Hindu Kush). The major literary Dravidian languages are found in the southern part of India, also known as “peninsular India.” An interesting linguistic aspect of the region which has often been pointed out is that most South Asian languages, regardless of their genetic affiliations, share a number of linguistic features, and thus one may appropriately speak of South Asia as a “linguistic area” (see especially Masica 1976). These features include extremely detailed resemblances in word order (SOV in unemphatic declarative sentences), and syntactic processes like the “dative-subject” construction (as in Marathi ma-la avadte “me-to it-pleases,” i.e. ‘I like it’), the extensive use of “conjunctive participles” for conjoining sentences (as in Tamil pd-i paru “having gone look”, i.e. ‘go and look’), and “double causative” verbs (causative of causative - e.g. Hindi banva- ‘to cause to cause to be made’, from ban& ‘make, cause to be made’, from ban- ‘be made’). The history of Indo-Aryan languages commences with the hymns of the Rigveda, widely thought to have been originally composed sometime during the second millenium B C . Though these are primarily ritual texts, they contain a good deal of description of battles between the Aryan gods and an enemy people, variously described as “godless,” “phallus-worshippers,’’ “blackskinned,” “flat-nosed,” and “of obscure speech” (see Deshpande 1979:2). If one takes these accounts literally one might get the impression that the early Aryans, with the help of Indra and the other gods, massacred large numbers of these enemies, destroyed their strongholds, flooded their fields, and enslaved the survivors. Such a notion was indeed prevalent among early western Indologists, and is popular among educated Indians today (especially in the Indo-Aryan-speaking north). One piece of linguistic evidence which has been 201 z 208 ANNALS NEW YORK ACADEMY OF SCIENCES zyxwvutsr zyxw FIGURE 1. Major South Asian languages. held to support this theory of conquest is the word Dasa-, one of the terms for the enemy race, which is found in later Indo-Aryan in the meaning ‘slave’. With the discovery and excavation of major sites of the Indus Valley civilization (now in Pakistan), Sir Mortimer Wheeler believed that he had discovered tangible evidence of the great battles of the Rigveda. The once-great cities of Harappa and Mohenjodaro and the towns and villages associated with them, some of which were fortified, were declared to have been the homes of the Dasa-s, and the invasion of the Aryans was claimed to have been the cause of the abandonment of all of the larger sites in the mid-second millenium B.C. - though the “end of Mohenjo-Daro . . . was rooted in deeper causes of decline” such as possible floods, salination of the soil, or obstructed irrigation (Wheeler 1966:83). This notion received wide acceptance not only among prehistorians, but was also adopted by many members of the general public who, for whatever reasons, were fond of the idea of the Aryans as “conquerors,” charging in with their swift chariots and iron swords to demolish the blackskinned indigenes, who were peaceful, sedentary farmers and herders. When zyxwvutsrq SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT 209 later research showed up the possibility that the Dasa-s might have been Dravidians, the theory became even more popular in northern India. Unfortunately for Wheeler’s theory, the evidence of massacre and destruction was minimal. The score or so of skeletons found in various locations in the upper levels of Harappa, where they were supposedly slaughtered by the Aryan invaders (Wheeler 1966:79-83), could equally well have been stragglers from the abandoned city who were attacked by bands of looters. Recent evidence suggests that a number of these individuals were probably suffering from advanced stages of sickle-cell anemia, or a related anemia (Kennedy 1981:21-2, 1982:291-2), and thus they may have been invalids who were left behind as the city was abandoned. As to the reasons for the abandonment of the sites, current research appears to support the probability that there were multiple causes - including floods, ecological collapse of the agricultural system through overplanting and salinization of the soil, a shift of the main Indus river bed, and perhaps diseases like malaria and smallpox (see relevant papers in Possehl 1982). Much of the destruction of the sites took place later, as over the centuries local residents helped themselves to the nicely baked bricks for use as building materials. Why is a linguist concerned with this controversy? Primarily because the strongest evidence for the dominant role of the Aryans in South Asian politics is the Indo-Aryan languages themselves, which have been the languages of the dominant groups in North India from the Rigveda on, as far as written records can tell us. These languages are generally believed by linguists to have originated outside the South Asian subcontinent. Thus, an invasion theory is apriori not implausible. And even without falling into the trap of equating language with race, nevertheless the overwhelmingly dominant position of speakers of these languages in the northern part of the subcontinent needs to be explained somehow. It is also necessary to explain how, if speakers of these languages were so dominant, their languages absorbed many structural features from the speech of others. It is claimed (and I will examine the evidence for this claim) that these features began entering Indo-Aryan speech from the earliest times. It is also generally recognized that the early Aryan religion was substantially changed in form over time, probably through contact with other peoples; note particularly the increase in the importance of female deities in later Hinduism (Basham 1954:312 ff.). If there was so much linguistic and cultural influence on the OIA speakers by the earlier inhabitants, how is it that the Indo-Aryan languages became the principal forms of communication throughout the northern part of the subcontinent, and where have the languages of the earlier inhabitants gone? This question is best divided into two parts: (1) How did the Indo-Aryan languages become the languages of the elite throughout the northern region, if the Aryans did not in fact disrupt the earlier more advanced civilization, or take control of the area by military conquest? (2) How is it that speakers of Dravidian (and/or whatever languages were spoken there) have disappeared so completely? We need a reasonable account of this, since otherwise we are open to the accusation (as all “substratum” theories are) that we invoke an unattested language when we need it, but then let the speakers conveniently zy zyxwvu zyxwvuts 210 ANNALS NEW YORK ACADEMY OF SCIENCES zy die off when they are no longer useful to our theory. These are the problems which give rise to the present paper. THE METHOD zyxwv zyxwvu The approach to the solution of this problem will make use of a set of techniques which I call “linguistic archaeology.” This includes the usual methods of linguistic reconstruction, along with what we might call “retrospective sociolinguistics”- that is, extrapolating backwards from contemporary sociolinguistic studies in an attempt to reconstruct a prehistoric social context which is appropriate to the reconstructed and attested linguistic data of the period in question. The approach also makes use of work in other fields dealing with prehistory such as archaeology and palaeobotany, wherever it is relevant (and comprehensible to a mere social scientist). This multi-pronged approach is necessary because no single piece of evidence is compelling by itself. In fact, every bit of evidence to be presented here can be (and much of it has already been) explained away by one or another scholar who disagrees with the conclusions. As William Labov has put it, zyx We can never claim to have resolved an historical issue decisively . . . the best . . . we can d o is to develop the most plausible reconstruction of past events . . . (1977:257). I hope that by putting together data from a number of different sources, the totality of the evidence will lead us to the most plausible reconstruction of the sociolinguistic situation in which the old Indo-Aryan language was spoken, beginning in the second millenium B.c.. THE EVIDENCE OF LINGUISTIC CONVERGENCE A number of important papers by Murray Emeneau, along with a few by other scholars, have investigated the origins of some of the shared linguistic features which define the South Asian “linguistic area” (see e.g. Dil 1980). As a Dravidianist and Sanskritist, Emeneau has been able to provide important evidence for the case that several of these features originated in Dravidian, and he has concluded that the influence of Dravidian on early Indo-Aryan was of such an extent that it can only be explained by positing a “bilingual majority”- i.e. a large mass of people who had learned Sanskrit as a second language, and into which they carried over features from their own native language, thereby influencing even the usage of Aryan native speakers. Emeneau believes that the evidence of such carried-over features can be found in the oldest document of Indo-Aryan, the Rigveda itself, implying extensive bilingualism (and therefore extensive social integration of the Aryans and the earlier inhabitants) from the time the ancient Aryan language was first spoken in the subcontinent. Other scholars have resisted this conclusion, particularly as far as the Rig- SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT 211 zy zy zyxwv zyxwvuts zyxw Veda is concerned. (They are perhaps prepared to allow it for later periods for which the evidence is stronger.) Students of the Vedas point to the extreme hostility toward the D2sas and other non-Aryans which is reflected in the Vedas, as well as the sanctions against the learning of non-Aryan languages or the imitation of non-Aryan modes of behavior. (Of course, the very presence of such sanctions must indicate that the rules were occasionally broken . . .) Linguists have tried to show that the changes in question can be explained as purely normal internal linguistic changes (Hock 1975:98), or that their appearance was subsequent to the compilation of the Rigveda in its earliest form (Deshpande 1978). This controversy remains unresolved at present because neither side is convinced by the arguments of the other. Before looking at the evidence for early language contact, therefore, I would like to discuss some of the main conclusions of a recent study of contemporary cases of linguistic diffusion in South Asia, in order to establish an empirical base from which to extrapolate. Indo-Aryan and Dravidian languages differ considerably in their surface structure, and we find that whenever there is prolonged contact involving ianguages of these two families, certain structures tend to change in predictable directions. I shall only present two brief examples here. (These are not necessarily the best examples, but they are the easiest to present in brief form.) Example A: Copula Deletion The lack of a copula verb in equational sentences in Dravidian languages has often been noted. In those Indo-Aryan languages which are not suspected of having been influenced by Dravidian, deletion of the copular verb is rare or non-existent; thus, the following would be a normal Hindi-Urdu sentence: (Al) ram bimar hai “Ram sick is” = ‘Ram is sick’ Marathi, which is spoken in an area contiguous to two major Dravidian languages (Kannada and Telugu), and which is suspected of having had more intensive contact with Dravidian in the past, shows optional deletion of the non-past copula following an NP or an Adj (but not, for example, following a locative), as in: (A2) ha mazhft bhati @he) “this my brother is” = ‘This is my brother’; (A3) te cangle (ahe) “that good (is)” = ‘That’s good’ In informal speech, this deletion of the copula is frequent enough that the use of the copula carries a somewhat emphatic force. Kasargod Marathi, a form of Marathi spoken deep in Dravidian territory (roughly at the Kannada-Malayalam border), deletes invariably after an N P in the non-past, as in (A4) te kdgu “that who” = ‘Who is that?’ but retains the copula obligatorily after a locative, as in (A5) kutro ti ha “dog here is” = ‘The dog is here’ (Ghatage 1790:125) 212 zyxwvutsrq zy ANNALS NEW YORK ACADEMY OF SCIENCES Note that in Kupwar Marathi, the opposite process has taken place, as indicated by Gumperz and Wilson (1971:158-9): here the local version of Kannada has introduced a copula in equational sentences, apparently under the influence of Marathi. Suarashtri, a language based on a form of Gujarati which has been spoken in the Tamil area for over 500 years, has completely realigned its use of the copula to match exactly the Tamil usage. Saurashtri, like Tamil, has no copula in “categorical” equations such as (A6) raman cokkar menik (“Raman good man”) (Tamil raman nalla manitan) ‘Raman is a good man’, zy but uses the copula preceded by an adverbial marker, again like Tamil, in “noncategorical” equationsa: (A7) raman saukyam kan se (“Raaman well [marker] is”) (Tamil d m a n saukyam-aha irukkiran) ‘Raman is well’ Example B: The Ergative Construction All the modern Indo-Aryan languages have inherited a verbal system in which the past or perfective forms are based on the OIA past participle, and accordingly transitive verbs in the past are construed passively. (In some of the modern languages, particularly the eastern languages such as Bengali, this construction has been lost.) Thus in Hindi-Urdu (which again showsthe “unspoiled” state of Indo-Aryan) the subject of a perfective transitive verb is followed by an agent marker ne, and the verb agrees with the object: (Bl) ram ne mithai kha-i (Ram by sweets eaten-Fem.) (B2) sita ne mithai kha-i (Sita by sweets eaten-Fem.) (B3) ram ne am kha-ya (Ram by mango eaten-Masc.) (B4) sit%ne a m kha-ya (Sita by mango eaten-Masc.) ‘Ram ate sweets’ ‘Sita ate sweets’ zyxwvu zy ‘Ram ate a mango’ ‘Sita ate a mango’ Compare the following sentences with intransitive verbs, which agree with their subjects: (BS) ram a-y2 ‘Ram came’ (Ram came-Masc.) (B6) sita a-i (Sita came-Fem.) ‘Sita came’ zyxwvuts a These examples are from Pandit 1972:13-15. Note that the Tamil verb here is not strictly an “existential” copula, but rather a verb meaning something like ‘be situated in place or time’. (The Malayalam cognate means ‘sit’.) The Tamil “adverbial” suffix element -aha is actually the closest thing to a copula to be found in Tamil; it means something like ‘become’, so that the literal meaning of the Tamil example in (A7) can be rendered as “Raman, having become well, is situated in time”-i.e. ‘ . . . is well at present.’ 213 SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT zy Marathi shows a similar state of affairs in the following cases: (B7) ram2 ne mithai khal-l-i (Ram by sweets eat-past-Fem.) (B8) sita ne mithai khal-l-i (Sita-by sweets eat-past-Fem.) ‘Ram ate sweets’ ‘Sita ate sweets’. but in other contexts the system begins to break down. The agent marker is deleted in all sentences with first and second person subjects, as in the following case: (B9) tu mithai khal-l-i-s? (thou sweets eat-past-Fem.-2sg.) ‘You ate sweets?’ This sentence also illustrates a related phenomenon, namely the addition of secondary subject concord following the object concord. Furthermore, there are a number of syntactically transitive verbs which show only subject agreement, e.g.: (B10) mi caha pya-1-0 (I tea drink-past-Masc. lsg.) (B11) mi caha pya-l-e (I tea drink-past-Fem. lsg.) ‘I drank tea’ (male speaker) ‘I drank tea’ (female speaker) Looking at Kupwar Marathi, a variety of Marathi spoken on the MarathiKannada border, we find that the entire ergative construction is missing, and verb agreement is always with the subject N P (Gumperz and Wilson 1971:157): (B12) (tu) ghodi di-l-a-s (thou mare give-past-Masc.-2sg.) ‘You sold the mare’ zyx zyxwvuts zyxwvut The ergative construction is also completely absent in Saurashtri (Pandit 1972:12-13). Several conclusions can be drawn from these examples. First, they clearly demonstrate the existence of a ‘‘convergence continuum” in which some features, and some languages, show higher degrees of convergence than others. Second, the beginnings of convergence are insidious - in the sense that disease symptoms may be insidious, by not providing obvious symptoms to facilitate a definitive diagnosis. For example, the Marathi variable copula deletion, or even the changes in the ergative construction, are not obviously identifiable in isolation as instances of convergence, but when seen in the context of the continuum described above, they clearly appear to be so -especially when they can be supported by a dozen examples of similar type (for which see Southworth, forthcoming [2]). This point is particularly important because, when we look for evidence of convergence in the past, the texts we find (especially if they are of a formal nature, such as the ancient ritual texts of India) are likely to show us only the tip of the iceberg-i.e the least converged portion of the continuum. In these and other convergence situations it is possible to define, in an appropriate way, three distinct degress of convergence. The minimal situation, which can be calledfirst-degree convergence, is one which involves some surface structure variation with no clear violations of existing rules, as in the 214 zyxwv zy ANNALS N E W YORK ACADEMY OF SCIENCES case of copula deletion in Marathi which involves primarily stylistic variants. Second-degree convergence involves changes in surface structure rules, without fundamental changes in the deep structure of the language. The copula rule in Kasargod Marathi, or the loss of the Marathi agent marker in some cases in surface structure, could be considered examples of this degree of convergence. In third-degree convergence a feature is altered to the point where the converging language is no longer distinct from the model language with respect to that feature, as in the case of the lost ergative construction in Saurashtri and in Kupwar Marathi, or the situation of the copula verb in Saurashtri. Marathi, which shows a number of examples of first- and second-degree convergence towards neighboring Dravidian languages, but no cases of thirddegree convergence, is not currently in contact with any Dravidian language except at its southern border. Kupwar Marathi, as noted above, is a variety of Marathi spoken in the Marathi-Kannada border area. The other languages mentioned are spoken by small groups of speakers who are entirely surrounded by speakers of Dravidian languages, and have been so for at least several centuries. The majority of these IA speakers are fully fluent in the local Dravidian languages, and in many cases they form a caste and occupational group which has created an occupational niche for itself in the local socio-economicsystem. Thus we might hypothesize that third-degree convergence occurs only in those cases which we can describe as involving the symbiosis of different groups of speakers, whereas lesser degrees of convergence imply either more distant contact (presumably involving features diffusing from a frontier zone), or perhaps convergence dating from an earlier period. In the Marathi case there are good historical as well as linguistic reasons -e.g. the Dravidian element in Marathi place-names (for which see Southworth forthcoming [3]) - for positing a bilingual situation at an earlier period. It is possible that the degree of syntactic convergence which a language exhibits toward another language correlates, at least in an approximate way, with the degree of lexical and phonological convergence. On the basis of a study of numerous cases of convergence in South Asia and elsewhere (Southworth forthcoming [2]), it seems that in general, languages showing first-degree convergence are likely to show lexical borrowings belonging to the surface category of nouns, perhaps including a few other words belonging to other categories of “substantive” words (adjectives and verbs with concrete reference)- whereasfunction words (prepositions/postpositions, adverbs, numerals, conjunctions, auxiliary verbs, pronouns) are found mainly in cases of secondor third-degree convergence. The borrowing of inflectional affixes is usually a symptom of third-degree convergence. On the phonological side, an early stage of syntactical or lexical convergence is often accompanied by changes in phonemic distribution, whereas the importation of new distinctions, if not confined to some small elite subgroup, is indicative of a higher degree of convergence. The loss of native distinctions (e.g. the loss of the distinction between aspirated and unaspirated stops in some IA languages under Dravidian influence) is usually indicative of third-degree convergence. These correlations, zyxw zyx SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT 215 zy though not absolute, may provide some additional clues of use in the evaluation of prehistoric convergence situations. CONVERGENCE IN OLD INDO-ARYAN The above discussion establishes a context for a brief examination of the evidence of Indo-Aryan convergence toward Dravidian which is offered by the Rigveda and other early texts. The features which have been attributed to Dravidian by myself and others include phonological, lexical, and syntactic features. On the phonological side, the most striking feature is the occurrence of the retroflex-dentalcontrast, which occurs from the earliest period (though the frequency of retroflex consonants increases with time from the earliest texts up to the classical period), and which eventually includes sibilants, stops, nasals, and laterals. The retroflex-dental distinction sets the Indo-Aryan languages apart from all other Indo-European languages- including the closely related Iranian languages, which together with Indo-Aryan form the Indo-Iranian subfamily of Indo-European.b Generations of scholars have thought it more than a coincidence that this distinction is found precisely in the same region as the Dravidian languages, for which this must be considered a proto-feature. Hans Hock has however suggested that the sequence of changes which led to the development of this contrast in OIA can be considered internally motivated, i.e. this sequence involves nothing other than normal conditioned sound change, whose working can be observed in words of Indo-European origin (Hock 1975:99 ff.). The sequence of changes appears to have begun with a palatal sibilant [S] (probably [i]before voiced consonants), which had developed from PIE /s/ in certain environments.c This [S/i]developed into retroflex [s/z] in OIA, but remained [s] in Iranian. Subsequently, apical consonants in contact with this [s/z] developed retroflex allophones, which were at first non-distinctive but later became distinctive as a result of other changes: for example, OIA nida‘nest’is derived from a PIE *ni-sd-os (compare Gothic nist, E. nest) through the following series of changes: *-isd- *-izd- *-Ed- *-izd- -izd-id- (the last change involving loss of the [z] with “compensatory lengthening” of the vowel). (See Burrow 1955 for details.) zyxw z zy zyxwvu + + --f + + zyxw Hock (1975:lOl) mentions various exceptions to the statement that a retroflex-dental contrast is lacking in Indo-European; his list includes, in addition to several western European langauges, an eastern Iranian dialect. However, outside of Indo-Aryan there is nothing remotely comparable to the complete series of contrasting stops, nasals, and laterals found in Indo-Aryan. The palatalization occurred following an original /r/, /u/, /k/, or /i/ (the so-called RUKl rule), and affected all the sutem languages, though not in all environments in all languages (see e.g. Arumaa 1976 for Slavic). Note, however, that it is only in Indo-Aryan that this palatal I S / became retroflex. 216 zyxwv zyx zy zyxwvuts zyxwvu zyxw zy zyxwvu ANNALS NEW YORK ACADEMY OF SCIENCES TABLE 1. Indo-Aryan Phonological Features Attributed to Dravidian and Their Chronological Development Vedic Pali (1000 B.c.) (500 B.c.) 1. Retroflex-dental contrast + 2. Assimilation of stops sporadic i + Prakrit : (300 B.c.) : + I i + (increasing in friquency ove{ time) regular i regular I I I I i 3. “Weakening” of stops (ABSENT) sporadic 4. Deaspiration (h+@/C-) (ABSENT) : !regular I sporadic : :regular PDra- vidian feature , I I I , I regular :regular : sporadic :regular I I I !late PDravidian (sporadic) ;: i no aspiration ;; in precontact :: Dravidian , I I I i PDravidian (?3000 B.c.) ” I :regular PDra; vidian feature * I I i i: :: , I I 5. Initial stress “0uter”ZA (?lo00 B.c.) ! (ABSENT) (ABSENT) (ABSENT)/coexisting i :regular PDraI ; with other vidian feature I : patterns i ; I ii If the original change from [ S / i ] to [s/z] was the result of Dravidian influence, how could it be presumed to have happened? In Dravidian, the only sound anywhere in the phonetic neighborhood of the Indo-Iranian [S/i]was a retroflex continuant [R] (normally voiced, but non-distinctively).d Would it be too much to assume that Dravidian speakers trying to reproduce [ S ] or [i]would substitute their retroflex [R] for it? Such a supposition, though plausible enough, does not of course constitute proof that it did happen that way. We must look at the historical context of the change for supporting evidence. The fact that the frequency of retroflex consonants increases through the history of Indo-Aryan suggests that what we see in the Rigveda is the leading edge of a change. TABLE1 shows that this is the first of several changes which took place in Indo-Aryan in a similar way-i.e. increasing in frequency or in regularity at successive historical stages - all bringing the phonology of IndoAryan closer to that of Dravidian. Change 2 in the table involves the assimilation of dissimilar obstruents: e.g. OIA sapta- ‘seven’ Pali Pkt. satta- Panjabi satt, Hindi-Urdu s2t; OIA migra- ‘mixed’ Pali Pkt. missa-. Change 3 involves voicing of intervocalic voiceless obstruents along with spirantization or loss of voiced ones: OIA Suka- ‘parrot’ Pali suka-/Suva-, Pkt. su(g)a-; OIA nija- ‘own’-* Pali nijainiya-, Pkt. nia-; OIA ruta- ‘voice’-* Pali rutahuda-, Pkt. ma-. Change 4 is the deaspiration of aspirated consonants as in OIA + + + +. Bh. Krishnamurti (1958) reconstructs this sound as /*z/ in Proto-Dravidian. I have not used the symbol z in this paper because I wished to avoid giving a misleading appearance of equivalence between the Dravidian retroflex continuant (which 1 have symbolized as [R]and the IndoAryan retroflex sibilant [z]. There is no inherited is]-type sibilant in Dravidian, except occasionally as an allophone of /c/ -e.g. in Tamil in intervocalic position. Note also that early Dravidian has no distinctive voicing at all (see Zvelebil 197078-9). zy zyxw zyxwvu zyxwvu 217 SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT kaphoni- ‘elbow’ Pali kaponi-, OIA *kathika- ‘agreement’--*Pkt. kat(h)ika-, Pali hattha- ‘hand’ (from OIA hasta-) Hindi-Urdu hath, Assamese Bg. Or. M. KO.hat. (In the modern Indo-Aryan languages of the “outer” group, the deaspiration of non-initial stops is very frequent, and in some cases regular. See Southworth forthcoming [4]for a discussion of the “inner” and “outer” groups.) The last change listed in the table involves the shift of the stress accent in a word to the initial syllable. (This is the prevailing pattern in Dravidian, see Zvelebil 1970:40.)While Vedic Sanskrit had a shifting musical accent, this is believed to have given way in later Sanskrit to a positionally determined stress accent, the effects of which can be seen in the lengthening and shortening of certain vowels in modern Indo-Aryan languages. Thus for example, OIA amravataka- ‘hogplum’ would have been accented on the second syllable, leading to a shortening of the first vowel in Pkt. ambataka- and Hindi-Urdu ambara, while the second (accented) vowel remains long; the “outer” group languages however have an initial long vowel (in some cases alternating with short vowel), but shorten the second vowel: Assamese 2mar2, Bengali amra, Oriya am(b)ar%,Marathi ambada. In another case, OIA aksavata- ‘wrestling ground’, which would have been accented on the third syllable, the vowel length is retained in Pali akkhavata-, Pkt. akkhavadaga-, and Hindi-Urdu akhara, but the first vowel is lengthened in Assamese akhara, B. 0. akhara, and variably lengthened in Gujarati akhar/akhar. As noted in the table, this shift of stress accent to the first syllable coexists with other stress patterns in the “outer” languages. The situation in syntax shows a comparable picture. Numerous lengthy articles have been written on this subject, and there is only scope here for a 2 depicts quick glimpse at the type of evidence which can be mustered. TABLE the stages of the so-called “quotative” construction, which appears in sentences reporting direct discourse (DD) with a verb of saying (SAY), with or without an intervening particle. It was noted early in Indo-Iranian studies that Avestan, the oldest form of Iranian, and OIA share a quotative particle, itihiti, of unknown origin, and that Avestan uses uiti before the direct discourse and after the verb of saying, whereas OIA has not only that order but also the order DD + iti + SAY, which happens to be the order of a similar construction in Dravidian. (In Dravidian, the element enr-u can be translated literally as “having-said” or “unquote”. This form is a “conjunctive participle”, a verbal form which is frequently used in South Asian languages to conjoin simple sentences into complex sentences.) + + TABLE 2. Syntactic Convergence: The “Quotative” Construction (Pattern A) Proto-IE Avestan Vedic Later Sanskrit DD DD DD + SAY [earlier?] (ABSENT) + iti + SAY + iti + SAY + enr-u + SAY (Pattern B) SAY ( + PTCL) + DD [later?] SAY ( + uiti) + DD SAY + DD ( + iti) (ABSENT) ______~ Dravidian DD (ABSENT) 218 zy zyxwvutsrq zyx zyxw ANNALS NEW YORK ACADEMY OF SCIENCES From this it has generally been concluded that the Avestan order was the only order used in the Indo-Iranian parent language, and that OIA innovated the other order under Dravidian influence. Hock (1975:107 ff.) has again attempted to throw doubt on this conclusion by pointing out that both orders are found elsewhere in older Indo-European languages; thus it is possible that Proto-Indo-Iranian inherited both orders, and Avestan lost one of them. Once again the evidence for convergence is ambiguous, but once again perhaps the situation can be clarified by looking at the historical context. In the subsequent history of OIA the second order (i.e. SAY + DD + PARTICLE, the order found in Avestan) almost completely disappears. Thus here again we seem to be looking at the leading edge of a change in the earliest Indo-Aryane In fact, as Kuiper (1967) has pointed out, the “conjunctive participle,” to which allusion was made above, shows a parallel development. HISTORICAL IMPLICATIONS Before looking at the lexical evidence, something needs to be said about the historical implications of the above proposal. The developments discussed in the previous section have been presented as a sequence in time- the longer the time, the more convergence takes place. And while this is no doubt true, it is not the whole picture. We must also take into account the probability that, at any given period, some varieties of Indo-Aryan showed more convergence than others. In other words, some of the changes which are found later in the record probably coexisted in fact with earlier stages. Thus, for example, the frequency of retroflex consonants was probably higher in some varieties of informal speech which were contemporaneous with the Vedic hymns, than in the hymns themselves, and perhaps even higher among groups of IndoAryan speakers who were not directly associated with the Vedic rituals. Why should we make this assumption? First, because of the extreme cultural chauvinism of the Vedic texts, with their denigration of everything nonAryan and their invocation of heavenly retribution on those who mispronounce the words of the Vedas or imitate non-Aryan ways, it must be assumed that conscious efforts were made to avoid perceived “Dravidianisms” or other “barbarisms”-especially in the sacred texts. (In fact, the study of Sanskrit etymology is fraught with difficulties because of the many disguised Dravidian and other foreign words, which were “Sanskritized” by the Brahmans when it was not expedient to replace them with inherited Indo-Aryan words.) zy There is a reasonable possibility that the Indo-Iranian “quotative particle” (u)iti has a Dravidian origin. Attention may be called to the Dravidian verb PDr0 *vin-‘hear, listen, ask‘(DED5516), which is reconstructible to the oldest level of Dravidian. The conjunctive particle of the causative of this verb in Tamil would be vitti, meaning ‘having caused to hear’. (I owe the explanation of this form to David McAlpin.) If it seems far-fetched to assume that such a form might be borrowed, we can point to the example of Kasargod Marathi, which has borrowed the equivalent form from Malayalam. (An example is: mi tu ye:-nni muntlo [“I you come-Quotative said”] ‘1 told you to come’-Ghatage 1970:127; cf. the Tamil quotative enru, colloquially -mu.) zyxw zy zyxwvu SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT 219 Furthermore, there is clear evidence that there were speakers of Aryan languages who were not part of the Vedic society. Some of these “outsiders” do not appear in very clear form until the later literature, and there is very little evidence for them in the ancient texts. One of the major lineages in the later Puranic literature is that of the Yadava-s, descendants of a (presumably mythical) Yadu. In a very interesting paper, Rornila Thapar has shown that the places where the Yadavas were said to be located correlate quite closely with the locations of certain early pottery finds (Thapar 1975; see FIG.2). The Purus, the other major lineage, are located in a central area which correlates with the finds of OCP (ochre-colored pottery) and later becomes the habitat of people using PGW (painted gray ware), whereas the Yadavas were said to be located mainly in peripheral areas correlating with the finds of BRW (black-and-red ware). The Yadavas, incidentally, were a somewhat marginal people who practiced cross-cousin marriage, a notorious Dravidian custom, and were apparently not recognized as full-fledged Aryans by those in the central region. The difference in regional distribution displayed in FIGURE2 happens to correlate surprisingly well with a purely linguistic division, suggested in the early years of this century by Sir George Grierson, who compiled the Linguistic Survey of India between 1890 and the 1920s. Grierson dubbed the linguistic varieties of the two regions the “Inner Group” and the “Outer Group” of Indo-Aryan. He offered various pieces of evidence for this division, of which the most cogent was the presence in the “Outer Group” languages of past forms in -I-, which are clearly an innovation. Recent investigation (Southworth forthcoming [4])has unearthed several other features which support this division (including the shift of the stress accent to the initial syllables of words, mentioned above), suggesting that it is probably quite old. The older Indo-Aryan literature, including most of the Prakrits, belongs almost entirely to the “Inner Group,” though one characteristic of the “Outer Group” languages begins to appear in the late Rigveda. (This is the -tavya gerundive, which appears in the modern “Outer” languages as a future or subjunctive form : e.g. the -bin Bengali jab0 ‘I will go’, the -v- in Marathi zave ‘should go’.) FIGURE3 shows the distribution of the past -1- suffixes, the main characteristic of the “Outer Group” languages. Comparing this with FIGURE2, we should bear in mind that these two maps represent a difference in time of nearly 2,000 years - assuming that the distribution of peoples described in the Puranas is based on traditions going back to the end of the Rigvedic period (around 1000 B.c.), and that the modern distribution of the major languages is approximately as it was in about A.D. 1O00, when these languages first began to appear in the historical record. Taking the Purus of FIGURE2 as representing the “Inner Group”, and the Yadavas as representing the “Outer Group,” we would have to assume that both groups extended their domains in the intervening two millenia, and in the case of the “Inner Group”, much of this expansion was into former “Outer Group” territory. In fact, Grierson himself (without any possible knowledge of the correlations pointed out by Thaper) has posited such an expansion on purely linguistic grounds. In spite of this presumed expansion, however, it is striking that the eastern boundary between zyxwvu zyxwvu zy 220 ANNALS NEW YORK ACADEMY OF SCIENCES z 22 1 SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT zyx zyx Indo-Aryan languages with -1- past other Indo-Aryan languages mn FIGURE 3. “Inner” and “Outer” Indo-Aryan. (After Grierson.) non-Indo-Aryan languages the two groups in the Gangetic plain has remained in the vicinity of the confluence of the Ganga and Yamuna rivers, where we now find the cities of Allahabad and Varanasi (Benares). This region has traditionally been considered the eastern limit of Aryavarta, the sacred land of the ancient Hindus. Putting these pieces of evidence together we have a strong suggestion of differences in language, social structure, ideology, and material culture between the central and peripheral regions - differences which have persisted over at least three millenia. The central region is still the stronghold of orthodox Hinduism, containing the majority of the most ancient and sacred pilgrimage places. And it is of course the peripheral regions which show the greatest influence of non-Aryan languages and cultures. The tension between center and periphery has never been fully resolved. The history of Hindu society has involved repeated absorption of dissident groups and their ideas into the orthodox mainstream, starting with the Buddhists (who wrote in Pali), and including Jains, Ajivikas, and others (who used varieties of Prakrits), and later others writing in early forms of modern Indo-Aryan. As these groups sought to preserve their teachings or to broadcast them to a wider public, the literary forms of their languages began to appear in the historical record. Here as elsewhere, the political process also played a role in producing records of more popular languages. The oldest inscriptions in India, those of King Ashoka in the third century B.c., show a presumably popular form of Prakrit. (Ashoka was a Buddhist, and his choice of Prakrit rather than Sanskrit to publish his decrees and broadcast his fame was in line with the anti-Brahman, anti-Sanskrit policy of the Buddhists of his day; it was also probably the most effective medium for reaching a wider public.) Thus I would suggest that the OIA-Dravidian convergence continuum can be regarded as a space-time continuum in which Sanskrit, the vehicle of main- 222 zy zyxwv ANNALS NEW YORK ACADEMY OF SCIENCES stream orthodox Hinduism from the Vedas to the present day, kept its distance as best it could from the polluting influence of Dravidian and other indigenous languages - and from forms of Indo-Aryan which were heavily influenced by these languages. This continuum can be presumed to have existed from the time when Indo-Aryan was first spoken in the South Asian subcontinent, if not before. We get only occasional glimpses of it in the form of phonological alternations in the earliest texts, while Pali and the Prakrits provide some clearer indications. But it must be remembered that these were also formal literary languages. It is only with the appearance of the true vernaculars of modern Indo-Aryan that it becomes possible to gauge the full extent of Dravidian influence on Indo-Aryan structure. By the time these appear, starting with the tenth century A.D., we see the results of at least 2000 years of convergence. zyxwv zy zyxwvu THE LEXICAL EVIDENCE I turn now to the evidence of lexical borrowing for further clues about the circumstances of early contact. Dravidian Listed below are a few of the lexical items in early Indo-Aryan which can be attributed to Dravidian, grouped into historical periods according to the probable time of borrowing. Here again no single item is compelling proof, but for all the items here the most likely source (in my opinion) is Dravidian, since the terms are all derivable from basic Dravidian elements. (The total number of probably borrowings I have found is: period A, 7; period B, 12; period C, 10; if other possible borrowings are included, the totals would be 10, 13, and 16, respectively.) OIA Lexical Borrowings from Dravidianf zy zyxwv A. PROTO-INDO-IRANIANPERIOD 1. mukhan- ‘face’ (Par. rnuh) 2. tanu- ‘body, person, self (Avestan tanu) (2500-2000 B.c.?) : PDrl *rnunkk- ‘nose’ (DED5024) (PDrl *rnun- ‘in front’, DED5020a) : PDr0 t a d t a n - ‘(one)self‘ (DED3196) f The designations PDr-0, PDr-1, PDr-2, PDr-3 refer to different (presumed) chronological levels of Proto-Dravidian, with PDr-0 the earliest and PDr-3 the latest. Roughly speaking, PDr-0 refers to any item with cognates in Brahui and any other language; PDr-1 refers to items shared by North Dravidian (Kudux and/or Malto) and any other branch (except Brahui); PDr-2 includes items shared by the Kolami-Parji group (Kolami-Naiki-Naikri-Parji-Gadaba) and any South Dravidian language(s); PDr-3 refers to items shared between any language of SDr-I (TamilMalayalam-Toda-Kota-Kodagu-Kannada-Tulu, etc.) and SDr-I1 (Telugu-Gondi-Konda-Kui-KuviPengo-Manda). See Southworth, McAlpin, and Chaudhuri forthcoming for further details. zyxw zy zyxwvu SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT 3. mataci- ‘locust’ (Avestan madaha) 4. n a d a h a d a - ‘reed’ (Iranian nada, Av. nada) 5. d%a- ‘demon, enemy, slave’ (Av. daha) : PDr2 *mituce ‘locust’ (DED4850b) (PDr2 *miti- ‘leap’, DED4850a) : PDr3 *iianal, *nal ‘reed’ (DED2909) : PDrO *taz- ‘be low, descend’ (DED3178) B. RIGVEDIC PERIOD 1. Sava- ‘corpse’ 2. phala- ‘fruit’ 223 (1500-1000 B.C.) : pDrl *cB- ‘die’, *cav-a corpse’ (DED2426) : PDrO *paz- ‘ripen’ *pazam ‘ripe fruit’ 3. yadava- ‘name of a people’ 4. kunda- ‘pot, hole, pit’ (DED4004) : PDr0 *yritu-van ‘goathheep herder’ (PDr0 yatu ‘goatlsheep’, DED5152 + an ‘man’) : PDrl *kuntam ‘cavity, pit’ (DED1669, 2082) (? PDr0 *kuz- ‘cavity’, DED1818) + c. LATER VEDIC PERIOD (1000-500 B.c.) 1. mlecchati ‘speaks an ununderstandable language’; cf. Pali milakkha- ‘non-Aryan’ 2. nagara- ‘town’ : PDr2 *muzi/mizi- ‘say, speak, utter’; 3. khala- ‘threshing floor’ : 5. pattana- ‘town’ : 6. pat- ‘recite, read’ : : *muzankk- ‘make noise, speak’ (DED4989); cf. tamiz ‘Tamil’ (+ *tan mizi ‘own speech’) PDr3 nakar ‘house, palace, town, city’ (DED3568) PDrl kala-m/n ‘open space, threshing floor’ (DED1376) PDr3 patti ‘cow-stall, habitation’; *pattinam/pattanam ‘town, city, seaport’ (DED3868) PDrl patu ‘sing, chant, etc. (DED4065) Two general points need to be made in connection with the above items. First, I have proposed etymologies for two names of early peoples: (1) diisa(the famed “demons” of the Rigveda, and (2) yiidava- (the “non-Vedic” Aryan people mentioned previously. The word yiidava- later comes to mean a herder, or a member of a traditional herding caste. (See further discussion below.) The second point is that this list includes, alongside of numerous nouns, two verbs (items C1 and C6). Significantly perhaps, these verbs both appear in the latest period shown here. From what has been said above, the presence of verbs in a list of borrowed words would indicate a degree of convergence which is characteristic of a stage beyond the initial period of contact among groups. zyxwvuts Other Sources of Lexical Items Apart from words which can be traced to Dravidian, we find a number of words in early OIA which cannot be convincingly derived from either Indo- ANNALS NEW YORK ACADEMY OF SCIENCES 224 European or Dravidian, and which therefore must be presumed to derive from some other (as yet unidentified) language or languages. To this category belong many of the names of indigenous flora and fauna, suggesting that these words may have come into both Indo-Aryan and Dravidian languages from the language@) of earlier inhabitants of the subcontinent. The same applies to a number of words connected with agriculture, such as the words for the plow, for rice, and for wheat listed below. Wheat, along with barley, was the main crop of the Indus Valley civilization, and rice was known in late Indus Valley sites, but only barley (yava-) appears in the Rigveda. Furthermore, neither the Vedic people nor the early Dravidians (as far as can be known) practiced sedentary agriculture, being primarily herders. On the other hand, plowed fields are found even in pre-Indus Valley contexts (La1 1971). This and other evidence clearly indicates that settled farming was well established in preHarappan times. zyxwv Examples of OIA Lexical Borrowings from Other Sources (Unknown) A. PROTO-INDO-IRANIAN PERIOD godhuma- ‘wheat’ (OPers. gantumo) : 1. lahgala- ‘plow’ (Marathi nabgar, etc) 2. Sira- ‘plow’ (1500-1000 B.c.) : PDr2 *fiafikal ‘plow’ (DED2907)s : PDr2 *cer ‘plow, ox team’ (DED2815) (?+ PDr0 *Eer- ‘join, approach’, DED2814) C. LATER VEDIC PERIOD vrihi- ‘rice’ (2500-2000 B.c.?) SDr *kbti, kbtumpai, Br. xolum (DBIA123) B. RIGVEDIC PERIOD : zy (1000-500 B.C.) PDr2 vari(ci) ‘rice’ (DED215, 4639, 5265) (+- PDrl *varinci ‘seed’) cf. ProtoMunda *arig ‘Setaria Italica’ THE EARLY SOCIOLINGUISTIC SITUATION RECONSTRUCTED In trying to pull these various pieces of evidence together to create a coherent picture of the early sociolinguistic situation, I will divide the following discussion into three sections corresponding to three distinct prehistoric periods: (A) the pre-Harappan or pre-Indus Valley Civilization period (before 2500 The ultimate source of this form is generally thought to be Austro-Asiatic (cf. Khasi lynkor, Santali nahel). The Dravidian form, PDr2 *fiankal, looks very much like a folk etymology, i.e. a compound of PDrl *firil ‘earth’ (DED2907) and PDr0 *kal ‘stone’ (DED1298). (Cf. PDr2 *fiankliz ‘earthworm’, DED2906). The OIA hal- ‘plow’is also possibly derived from the PDr0 *kal ‘stone’. zyxwvutsrqp zy zyxwvu 225 SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT (B) the period of Harappan urbanism (approximately 2500-1700 B.c.), and (C) the post-Harappan period, which includes the early Vedic period. B.c.), A. The Pre-Harappan Period zy zyx Walter Fairservis, an archaeologist who has worked on the Harappan civilization and its predecessors, as well as on the decipherment of the Harappan script, has suggested that there were three identifiable culturally distinct groups in western India in the period around 2500 B.c., shortly before the development of Harappa and Mohenjo-Daro into full-fledged cities (see FIG.4). These groups were: (1) “a hunting-gathering-herding group in Gujarat and perhaps southern Rajasthan”; (2) “village-farming groups along the Indus River extending east to Malwa and up the Ghaggar River”; and (3) three village-farming groups in the hill and adjacent country of eastern Baluchistan and southern Afghanistan” (Fairservis 1976:3). Deducing linguistic relationships from cultural relationships is a notoriously risky business, but certain possibilities can be suggested. First of all, because of the disparity of technological development between group (I) and the others, it seems likely (as Fairservis suggests) that the language of (1) was distinct from that of the other two groups - though one can say nothing about their ultimate relatedness or unrelatedness. I will designate this group and their language as simply “tribal”, a term used frequently in modern South Asia for peoples enjoying a pre-industrial economy. Regarding the relationships between groups (2) and (3), the following con4, existed in three siderations are relevant. Group (3), as depicted in FIGURE I zyxwvuts FIGURE 4. Pre-Harappan cultures in western India ca. 2500 B.C. (After Fairservis.) *(The sites of Mohen-jo-Daro and Harappa become the major cities of the lndus Valley in the next period. Lothal is a later site.) 226 ANNALS NEW YORK ACADEMY OF SCIENCES zy zyxwvu zy regional variants, which shared numerous common features (such as the use of bangles, figurines of females and cattle, cereal cultivation, water control) while differing in details of cultural style (vessel forms, decorative motifs, methods of burial, or emphasis on cattle vs. sheep/goat husbandry). All of these have “stylistic ties, artifactually at least”, with group (2), and one group (3C in FIG.4)has particularly strong links which continue into the next period, that of the Indus Valley civilization (Fairservis 1976:3). I will discuss the further implications of these relationships after considering the next period. B. The Period of the Harappan Civilization (2500-1700 B.c.) The Harappan urbanism seems to have been a rather rapid development: within a space of about 200 years, the large cities of Mohen-jo-Daro and Harappa sprang up from earlier clusters of small villages and some larger villages. In Fairservis’s words, “Its cultural content shows close ties to the hill borderlands” (group 3), but particularly to subgroup 3C (1976:7). It is “imposed upon earlier village farming groups in the southern Indus Valley” (1976:3), i.e. group 2. Fairservis notes further that “There is a high degree of conformity in settlement plan and artifact style in all Harappan sites of this period (ca. 2000 B.c.)” (1976:3), which would suggest that there probably existed a lingua franca used throughout the Indus Valley civilization at this time. We may also note the highly uniform system of weights and measures used in all the major sites. On the other hand, it has been pointed out by Fentress that there were differences between the northern metropolis of Harappa (on the edge of the Panjab) and the southern city of Mohen-jo-Daro on the Indusdifferences which point to divergences in ideology rather than in material culture, such as differences in size and structure of temples, in the numbers and locations of mother goddess figurines, etc. (Fentress 1976). Thus it also seems probable that these different local traditions involved at least highly divergent dialects, and perhaps distinct languages (whether of the same or different families), alongside the aforementioned lingua franca. It was noted above that numerous words for local fauna and flora, including a number of words for food crops, cannot be definitively traced either in Dravidian or in Indo-Aryan, and must be presumed to have originated in some earlier language, which I will now designate by the term “pre-Harappan”. It seems reasonable to conclude, on the basis of the evidence given above of lexical diffusion between Indo-Aryan and Dravidian, that speakers of both these languages were present in the subcontinent in the period under discussion. Leaving the Indo-Aryans for the moment, since they are by all accounts the latest arrivals, what are the possible ways of assigning the Dravidian and the pre-Harappan languages to the groups identified by Fairservis? Fairservis himself proposes that the “tribal” languages of Gujarat (group 1) are the source of the botanical and agricultural vocabulary I have mentioned, and that groups (2) and (3) are different languages belonging to the Dravidian family. I would oppose this on the following grounds. First, contact between group (1) and the Indus Valley people is demonstrated only in SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT 221 zy the late phase of the Indus Valley civilization in the southern site of Lothal (see Possehl 198072 ff.). It is possible that rice, which has been found at Lothal but not in the Indus Valley sites, was first discovered by the Indus Valley people in this area. But this is unlikely to have been the source of the words for other flora and fauna such as the date, sesame, bamboo, or the pipal tree, which were well known in the Indus Valley. Secondly, the words for plow mentioned above tie in with the archaeological discovery of a plowed field in a preHarappan context of group (2). As noted below, the word for plow appears to have cognates in Munda and other Austro-Asiatic languages (such as Khasi in northeastern India, and Thai in southeast Asia), and this may be its ultimate source. Hock (1975236 ff.) also points out, citing evidence from Sylvain Levi, that a number of non-Indo-Aryan tribal names appear to be derived from Munda, and some of these occur in Rig-Vedic. But Munda also fails to qualify as a source of most of the botanical terms mentioned above, as pointed out in Southworth 1976. I would therefore suggest a second possibility, namely the identification of group (2) as pre-Harappan and group (3) as Dravidian. The evidence for the second part of this identification includes the present location of the Brahui language in the territory occupied by subgroups (3A) and (3B), as well as the links between group (3) as a whole and earlier farming villages in Iran to the west - tying in with the possibility (recently demonstrated by McAlpin 1982) that Dravidian is related to Elamite, an ancient language of western Iran. There is some evidence to support the notion of Dravidian as a major language of the Indus Valley civilization. I have pointed out elsewhere that the inventory of food plants reconstructible for PDr-1 show a fairly good fit with those known from the Indus Valley sites (Southworth forthcoming [ 5 ] ) . Two specific items are interesting from this point of view: (1) Sesame. Sesame seeds were found at Harappa, one of the major sites of the Indus Valley civilization (Vats 1940). Sesame was important both as a food and an ingredient of religious ceremonies in ancient India, and is still so today. Bedigian (1985) has drawn attention to the similarity between one of the Mesopotamian words for sesame (Akkadian ellu) and the South Dravidian word ellu (+el). Even assuming a connection between the two, it is impossible to determine without further evidence which was the source, or whether there was perhaps some third source. Nonetheless, since there is a strong likelihood that Mesopotamia first acquired sesame through trade with the Indus Valley, the resemblance between the words for sesame does provide some support for a relationship between the Indus Valley and speakers of Dravidian languages. zyxwvut zy zyxwvutsr zyx The OIA word for sesame is tila-, from which the modern 1A words are derived. The word for ‘oil’ in OIA is taila- (meaning ‘of tila-’), and is the source of the word for oil in most of the NIA languages-e.g. Hindi-Urdu tel. (Both OIA tila- and taila- occur in the Atharvaveda, which is assigned to the late Vedic period.) It has long been suspected that there is some relationship between the 1A words and the Dravidian el(1u) ‘sesame’(attested only in SDr.), and Burrow suggested that the original form was perhaps a Dravidian *cel, given that PDr. initial *c- is often lost in South Indian languages, and in some cases is replaced by *t- (Burrow 1947:142-3). 228 ANNALS NEW YORK ACADEMY OF SCIENCES zy (2) The word for ‘date’ is reconstructible to the PDr-1 level as *kintu or *cintu; the PDr-2 form would have been *cintu (pronounced [ci:ndu] or [skndu]). A similar form, mu-kindu ‘wild date palm’, is reconstructed for proto-Bantu by Nurse (1983:142). The source of these words is unknown, but both the Indus Valley and West Asia are possibilities.’ Dates were known in the Indus Valley (Vats 1940:547), and the region of Sindh on the lower Indus is well-known for this fruit. An interesting possibility is that the very name for this region, and the original name of the Indus itself (OIA sindhu-), is connected with this Dravidian word for ‘date’- whether the fruit was named after the region (“fruit of Sindh”), or the region after the fruit (“land of dates”). Apart from this evidence, the suggestion of Dravidian as a major language of the Harappan civilization is certainly more plausible a priori than recent proposals by S . R. Rao (1973) and K. D. Sethna (1980, 1981) that the authors of the Indus Valley civilization were the Vedic Aryans. The Vedas contain a number of indirect references to items and concepts of daily life, but nothing which could be interpreted as indicating that they lived in cities-quite the contrary, in fact. The references to the destruction of fortresses, which Wheeler interpreted as the destruction of the walled Harappan cities, may possibly refer to raids on some of the walled villages in the urban hinterlands-but even in that case, such references clearly put the Vedic Aryans outside of the already existing “fortresses”. Sethna’s attempt to explain the lack of words for certain objects (such as cotton) in Vedic by placing the Vedas in the preHarappan period does not help, since one would then expect to find references to the Harappan cities in the later literature. I would propose another resolution of this problem. If the proper name Yadava- is derived from a Dravidian *y%tuvan,as suggested above, this would imply considerable contact between Indo-Aryan and Dravidian speakers in the context of animal herdingj I suggest that the major focus of this contact zyx zyx zyxwvutsr zyxwv zyx -. The Akkadian ellu clearly has nothing to do with a PDr. *cel(lu), since the Dravidian changes *c- 0 and *c- t- are restricted to South India, and must have taken place a millenium or so later than the Akkadian records. At this stage the connection between Akkadian and Dravidian looks more plausible than that between Dravidian and OIA: if OIA tila- is connected with the Dravidian word, the relationship is probably not the one suggested by Burrow. (Kuiper has suggested a possible Munda derivation of OIA tila-: 1955:157.) -+ ‘ “The earliest record is a sample of stones from Egypt dated about 4500 B.C. Traces from the third rnillenium B.C. from western India, records of the Sumerian civilization . . . jointly support Chevalier’s . . . suggestion that domestication occurred simultaneously in different places in the area between the Indus and the Atlantic” (J. H. M. Oudejans, in Simrnonds 1979:230). Following are two additional pieces of evidence in support of the thesis that Dravidian and Indo-Aryan speakers interacted in herding contexts in prehistoric times: (I) PDr2 *cal- (DED2781) means ‘go, pass, occur, etc.’, with meanings very similar to OIA calati, which (along with its doublet carati) is presumably of Proto-Indo-European origin. Thus, if the resemblance is due to borrowing, it would have to be presumed that the source was IndoAryan and the borrowing language Dravidian. Of more interest in the present context is the existence of a second Dravidian verb, PDr2 *car- (DED2362) meaning ‘pursue (men, game, etc.), drive, chase’ (cf. also Kolarni ser- ‘go’, serp- ‘let (cattle) get lost’ under DED2781). The distribution of meanings between the two verbs parallels closely that of the (modern and earlier) Indo- SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT zy zy 229 (though not necessarily the earliest phase of it) was probably associated with the Indus Valley civilization. The cities would have required large quantities of milk and meat, as well as other products such as wool and possibly skins. Nearby villages could probably not have provided these items in sufficient quantities, whereas pastoralists in the hinterland could have done so - and could have performed other functions as well (e.g. trading and transport of goods between the cities). Possehl(l979) has suggested that, despite the paucity of concrete evidence for their presence, there are good structural reasons for assuming that pastoralists played such a role in the Indus Valley civilization. I suggest that this role was likely to have been filled, at least in the first instance, by Dravidian-speaking *yatu-van, who would have been so designated by themselves and possibly by the inhabitants of the cities (if they were Dravidian speakers). With the arrival of “outer group” Indo-Aryans in the region, some Indo-Aryan speakers may have come to be designated by the same term, and perhaps eventually came to apply to themselves their own version of it (yiidava-). Vedic Aryans might have applied the term indifferently to Dravidian-speaking and Indo-Aryan-speaking herders, whom they considered equally outside the Aryan pale. Coming from similar (i.e. nomadic or semi-nomadic) antecedents, both groups of herders would have shared certain social institutions, such as the size of the herding group, and perhaps the custom of communal ownership of the herd. If, on the other hand, crosscousin marriage was adopted by Indo-Aryan-speaking Yadavas from their Dravidian-speaking counterparts, this would seem to suggest fairly extensive intermarriage between the two groups, which could well lead to generalized bilingualism (cf. the situation among Brahui and Baloch described by Emeneau in Dil 1980:335). There was however a crucial difference between the two groups: the IndoAryan speakers, coming from Central Asia, knew and used the domesticated horse, and were thus (at least initially) far more mobile than other groups. Furthermore, since the subcontinent has never been a good breeding area for horses, contacts with Central Asia were necessary in order to maintain a continued supply. The implications of this point will be spelled out further below. It would seem, then, that the growth of the Harappan urbanism was related to an interaction between groups (2) and (3), in which subgroup (3C) imposed its cultural forms - and presumably its language - on villages of group (2). Thus we can regard the linguistic composition of the Harappan cities, towns, zy Aryan verbs: the verbs with -I- refer to general movement, whereas those with -r- refer to movement of animals (driving, grazing, hunting). (2) OIA gotra- in the Rigveda means ‘cowpen, enclosure for cattle’; later it comes to mean ‘family, clan, lineage’. The parallelism with the semantic development of Dravidian patti ‘cowstall’ pattinam ‘town, city, seaport’ is striking, and suggests the identification, at an earlier period, of the cowpen with the house or habitation area, perhaps involving communal ownership of the herd. Archaeologically, this culture complex can perhaps be linked to Allchin’s “cattle keepers” of South India (Allchin 1963), as well as to modern Indo-Aryan (Dardic)-speaking transhumants on the upper Indus: for example, among Shina transhumants the word goth (from Skt. go-sta- ‘cattle stand’) means both ‘house/home’ and ‘place where the cattle herd is localed‘ (Dr. Ruth L. Schmidt, personal communication). -. 230 ANNALS NEW YORK ACADEMY OF SCIENCES and villages as composed of a “pre-Harappan” base, with important segments of the population speaking Dravidian, and (at a somewhat later period) IndoAryan. Given the circumstances discussed above, the lingua franca of the civilization as a whole (if there was a single one) was most likely a form of Dravidian which contained both pre-Harappan and Indo-Aryan lexical elements. What of the Vedic Aryans? According to the accepted chronology they would have made their appearance in the northern borderlands only towards the end of the Harappan period (though there is no compelling reason for excluding an earlier arrival). As I have suggested above, their contact with the Harappan cities was probably limited to skirmishes on the outskirts apart from indirect contact mediated by groups like the Yadava herders. zy zy zy zyxw C. Post-Harappan Period (1700-500 B.c.) The next period witnesses a series of eastward movements. A progressive movement toward the east and south was already underway during the late Harappan period, as indicated by the establishment of sites like Lothal (in Gujarat) and Kalibangan (in Rajasthan). The Vedic Aryans during this same period were probably moving eastward along the Himalayan foothills. The movement of the “Outer Group” Aryans probably preceded them, though their movement may have followed (in part) a more southerly pathway. The inhabitants of the Indus Valley gradually abandoned the urban way of life and reverted to village farming in many areas, though recent discoveries indicate that smaller cities continued for a much longer period than believed earlier. Towards the middle of the first millenium B.c., new cities rose in the Gangetic plain in a context associated with the Buddhist and early Hindu civilizations. I suggest that it was during this period of about 1000 years, when the old urban structures had collapsed (at least in the Indus Valley) and much of the population was on the move in small groups, that the Indo-Aryan languages (primarily those of the “Outer Group”) became promoted into regional lingua francas - perhaps mainly because of the greater mobility of their speakers, which by this time would have permitted them to establish far-flung networks across the subcontinent. Thus Indo-Aryan languages became the most viable forms of intergroup and inter-regional communication. But it is important to remember that -whatever ideological commitment these Indo-Aryan speakers and their descendants had toward Aryan ethnic “purity”- they already represented at this period a cultural and linguistic amalgamation of two or more different ethnic streams. With this development the stage was set for the consolidation of Aryanspeaking kingdoms throughout northern India, and the struggle between orthodox Brahmanical Hinduism (propagated through Sanskrit) and the various dissident groups who used more “evolved” forms of Indo-Aryan. What happened to the speakers of Dravidian languages? They have not totally disappeared from the north, of course, since Dravidian languages like Kudux and Malto are still spoken within Indo-Aryan territory. But in the absence of historical knowledge we cannot say whether these groups are sur- SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT zyxw zy 231 vivors of earlier larger populations in the areas where they are now found, or possibly moved into the region from elsewhere (though it should be noted that they are spoken in areas where ancient Yadavas and users of BRW pottery were found, see above). However, the process of Indo-Aryanization of Dravidian-speaking areas can be seen even in historical times in Maharashtra, where (according to Deshpande 1979:102)some of the southern kingdoms were Dravidian-speaking at an earlier period, but under the influence of the northern kingdoms changed their official language to Marathi. Even without the existence of such a powerful influence we can see dramatic shifts of language among the Brahui, who were at one time the dominant people of their region, but many of whom now (even among the former royal families) can no longer even speak Brahui at all (Emeneau, in Dil 1980:355). The situation is clearly more complex than the above examples suggest. Gumperz and Wilson, on the basis of their study of multilingualism on the Kannada-Marathi border, have suggested that the retention of one’s home language is the norm in a society which values “the ethnic separateness of home life” (1971:153). I would suggest that this depends very much on circumstances. In looking at cases of immigrant groups in different parts of India, we find that when there is something to gain in terms of status and power by discontinuing the use of the home language, many individuals and many groups appear ready to take this step. Thus groups like the Kudumbi of Kerala, whose background prevents them from being considered “backward castes” of Kerala and qualifying for government aid, are prepared to drop their original language along with their separate identity. At the other end of the social scale, powerful local groups are prepared to link themselves with more powerful neighbors by the same method. In the middle of the socio-economiccontinuum, on the other hand, groups like the Saurashtri of the Tamil country, who have established a niche for themselves and are not particularly mobile as a group, have no reason to drop their ancestral language. The foregoing account is necessarily speculative, since it deals with a period long before written history and must make use of whatever stray bits of evidence can be found. For the skeptical, it is a simple matter to tear such an account down by refuting various parts of it piecemeal. If we hope to get any understanding of these prehistoric times, however, what is needed is to muster all the available evidence from all the relevant disciplines, and to see how each piece fits into the total context. This is what this paper has attempted to do, and I hope that it may stimulate further -and ultimately more successful attempts in the same direction. zyxw zyxwvutsr zyxwv REFERENCES ALLCHIN, F. R. 1963 Neolithic cattle-keepers of South India: astudy of the Deccan ashmounds. Cambridge: Cambridge University Press. ARUMAA, P. Urslavische Grammatik, II. Konsonantismus:42-6. Heidelberg; Carl Winter. 1976 232 zy zyxwvutsrq zyxw ANNALS NEW YORK ACADEMY OF SCIENCES BASHAM,A. L. 1954 The wonder that was India. New York: Grove Press, Inc. BEDIGIAN, DOROTHEA 1985 Evidence for cultivation of sesame in the ancient world. Economic Botany (submitted). BURROW,T. 1947 Dravidian studies vi. Bullefin of the School of Oriental & African Studies 12. 1955 The Sanskrit language. London: Faber & Faber. B. EMENEAU BURROW,T. AND MURRAY 1984 A Dravidian etymological dictionary (revised edition). Oxford: Oxford University Press. DBIA = Emeneau & Burrow 1962. DED = Burrow & Emeneau 1984. DESHPANDE,MADHAVM. 1978 Genesis of Rigvedic retroflexion: A historical and sociolinguistic investigation. In Despande and Hook 1978. 235-306. 1979 Sociolinguistic attifudes in India: An historical reconstruction. Ann Arbor, MI: Karoma Publishers. DESHPANDE,MADHAVM. AND PETER E. HOOK,EDS. '978 Aryan andnowAryan in India. Ann Arbor: The University of Michigan Center for South and Southeast Asian Studies. DIL, ANWARS., Ed. Languages and linguistic area: Essays by Murray B. Emeneau. Stanford, CA: 1980 Stanford University Press. EMENEAU, M. B. AND T. BURROW Dravidian borrowingsfrom Indo-Aryan. Berkeley: University of California Press. 1962 FAIRSERVIS, WALTERA. 1976 (ms.) Archaeology and linguistics: an Indo-Pakistan model. Paper presented at the Symposium o n Linguistic Archaeology at the Annual Meeting of the American Anthropological Association, Houston, Texas. FENTRESS,MARCIAA. Resource access, exchange systems, and regional interaction in the Indus Valley: 1976 An investigation of archaeological variability af Harappa and Moenjodaro. Doctoral dissertation, University of Pennsylvania, Philadelphia. GHATAGE,A. M. Marati of Kasargod. State Board for Literature and Culture, Bombay. 1970 GRIERSON, SIR GEORGE Linguistic survey of India. Calcutta. 1903-1928 GUMPERZ,JOHNAND ROBERTWILSON Creolization without pidginization: A case from the Indo-Aryan/ Dravidian 1971 border. In Pidginization and creolization of languages, edited by Dell Hymes. Cambridge: Cambridge University Press, pp. 151-67. HOCK,HANSHEIRICH 1975 Substratum influence on (Rig Vedic) Sanskrit? Sfudies in fhelinguisticsciences No. 2. Urbana: University of Illinois. KENNEDY, KENNETHA . R. 1981 Skeletal biology: When bones tell tales. Archaeology. January-Febru- ary:17-24. Skulls, Aryans and flowing drains: The interface of archaeology and skeletal 1982 biology in the study of the Harappan civilization. In Haruppan Civilization, edited by G. Possehl. New Delhi: Oxford & IBH Publishing Co., pp. 289-295. zyxwvu zyxwvut zyxw zy zyxwvu zyxwvuts SOUTHWORTH: SOUTH ASIAN LANGUAGE CONTACT 233 KRISHNAMURTI, BH. 1958 Proto-Dravidian *z. Indian Linguistics, Turner Jubilee Volume 1:259-93. KUIPER,F. B. J. Studia indologica. Festschrift fur K Kirfel. Bonner Orientalische Studien N. 1955 s. 3. 1967 The genesis of a linguistic area. Indo-Iranian Journal. 1081-102. LABOV,WILLIAM 1977 On the use of the present to explain the past. Linguistics at the crossroads. Liviana Editrice, Jupiter Press, pp. 226-261. LAI, B. B. Perhaps the earliest ploughed field so far excavated anywhere in the world. 1971 Purafattva (New Delhi). MASICA, COLIN P. Defining a linguistic area: South Asia. Chicago: University of Chicago Press. 1976 NURSE,DEREK A linguistic reconsideration of Swahili origins. Azania XVIII:127-50. 1983 PANDIT,PRABODH B. India as a sociolinguistic area. Ganeshkhind, Poona-7: University of Poona. 1972 POSSEHL, GREGORY 1979 Patterns of Harappan settlement and subsistence. South Asian Archaeology 1980 Indus civilization in Saurashtra. Delhi: B. R. Publishing Corporation. ED. Harappan civilization. New Delhi: Oxford & IBH Publishing Co. 1982 RAO, s. R . Lothal and the Indus civilization. Bombay: Asia Publishing House. 1973 K. D. SETHNA, 1980 The problem of Aryan origins. Calcutta: S & S Publishers. Karpasa [cotton] in prehistoric India. New Delhi: Biblia Impex. 1981 SIMMONDS, N. W., ED. Evolution of crop plants. London: Longman. (reprinted 1979) 1976 SOUTHWORTH, FRANKLIN C. Cereals in South Asian prehistory: a look at the linguistic evidence. Ecological 1976 backgrounds of South Asian prehistory, edited by K. Kennedy and G. Possehl. Ithaca, NY: South Asia Program, Cornell University. Forthcoming [I] Linguistic archaeology of the South Asian subcontinent Forthcoming [2] Towards a typology of linguistic convergence (with special reference to South Asian languages). See [l] above. Forthcoming [3] Maharashtrian place names and the question of a Dravidian substratum. See [I] above. Forthcoming [4] Grierson revisited: The “inner” and “outer” groups of Indo-Aryan. See [l] above. Forthcoming [ 5 ] Ancient economic plants of South Asia: Linguistic archaeology and early agriculture. To appear in a Festschrift for Edgar Polome. Austin: University of Texas Press. DAVIDW. MCALPINAND SHUBHA CHAUDHURI. Forthcoming. Proto-Dravidian cultural vocabulary. zyxwvu zyxwvuts 234 zyxwvutsrq zyx zy ANNALS NEW YORK ACADEMY OF SCIENCES THAPAR,ROMILA Puranic lineages and archaeological cultures. Purafatfva8:86-98 (reprinted in 1975 R. Thapar, Ancienf Indian social history. New Delhi: Orient Longman, 1978, pp. 240-267. VATS,M. S. Excavations at Harappa ( 2 vols.). Delhi: Government of India. 1940 WHEELER,SIR MORT~MER Civilization of the lndus Valley and beyond. London: Thames and Hudson. 1966 ZVELEBIL,KAMIL 1970 Comparative Dravidian phonology. The Hague: Mouton.