Вы находитесь на странице: 1из 13

TOWARD PROTO-ELAMO-DRAVIDIAN

DAVID W. MCALPIN

Portland, Oregon
It is demonstratedhere that the Dravidian family of languages in South Asia is is basedon cognatewith Elamite,an ancientlanguageof West Asia. This demonstration Dravidianterms. The 57 Elamitewords (mostly verb stems) pairedwith corresponding and interlocking.A beginningis correspondencesare, on the whole, straightforward the phonology of Proto-Elamo-Dravidian.* made in reconstructing

Although numerous attempts have been made to find genetic connections, the Dravidian family of languages has long remained an isolate in diachronic studies. Most of these attempts have been toward Uralic or the various languages of the ancient Near East, plus a smatteringof others.1This paper shows that the Dravidian family is cognate with Achaemenid Elamite, an ancient language from around the Persian Gulf, which has also been an isolate.2 Not only can these two languages be shown as cognate, but enough detail is available to begin reconstructing the protolanguage, which I have tentatively labeled Proto-Elamo-Dravidian (PED). While Elamite is cognate with the Dravidian family as a whole, it is in no sense a Dravidian language as that term is normally used. Thus, the term 'Dravidian' has been left to its current usage, referringto a well-definedgroup of South Asian languages. To the best of my knowledge, only two previous attempts have been made to connect Elamite and Dravidian. The first, by Caldwell ([1913] 1961:65-7), compared Dravidian with the Elamite of the Behistun inscription as part of a general comparison of Dravidian and 'Scythian' languages. The second, by Bork (1925:82-3), compared Elamite with Brahui. Both studies lack sufficient data to be convincing. Without the background work of the Dravidian etymological dictionary (DED, Burrow & Emeneau 1960), and particularly the glossary in Hallock 1969, little more could be said than that the similarities between
* I wish to thank RichardT. Hallockfor his commentson an earlierdraft of this paper,and for his advice and corrections on details of Elamite. I also want to thank Andrew Sihler for his commentsand his encouragement. for the ideas presentedhere. However,I take sole responsibility 1 Dravidian has been connected to the languages and language families listed below, where the numbersrefer to entries in Andronov's bibliography(1966); the list is far from exhaustive. Particularnote should be taken of Burrow 1946 and Tyler 1968. Ural-Altaic(Scythian):67, 68, 110, 547, 697.
Uralic: 95, 255, 548, 551, 639, 658.

Altaic: 376. Korean: 266. Basque: 316-19, 682. Sumerian:659. Hittite: 526. Mitanni: 83. Indo-European:228. Austric: 467. Nubian: 653. Indeterminate: 30, 65, 262, 470. 2 Credit is due Erica Reiner for the germ of the idea that Dravidian and Elamite might be related. 89

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

90

LANGUAGE,

VOLUME 50, NUMBER

1 (1974)

Dravidian and Elamite are suggestive, and no conclusive demonstration would be possible. also called Anzanite or Susian,3 was the language spoken in Elam1. ELAMITE, an ancient kingdom centered on Susa, immediately to the east of Mesopotamia, between the Zagros Mountains and the Persian Gulf. Existence of the kingdom was first recorded about 3000 B.C.,and it continued with variations in status and power until defeated by the Assyrians under Assurbanipal in 640 B.C.The area was later absorbed into the Achaemenid Persian Empire. Elam borrowed heavily from Mesopotamia and was definitely in the sphere of Mesopotamian culture, but it maintained a separate identity throughout.4 The language of Elam has been recorded in threeversions. The first, Proto-Elamite, was a hieroglyphicscript, first used only a few centuries after the beginning of writing in Sumer. This script was in use from 3000 to 2200 B.C.,and has never been deciphered. The second version, Middle Elamite, was written in a cuneiform script derived from Sumer which came into use around 2500 B.C. It was used in the records of the Elamite kingdom at Susa. The third version, Achaemenid Elamite (AE), derives from the special relationship of the Elamites to the Achaemenid Persians.5The Elamites were the imperial recordkeepers, and Elamite functioned as the third language of the empire after Old Persian and Akkadian. AE data come in two sets. The first is from the royal inscriptions6f-well composed, carefully written, containing about 5,000 words. The second set consists of the imperial records from Persepolis,7which are translations of records made for and by the Elamite clerks. Many of these memoranda are literal translations from Old Persian and are often hastily written, so they are not necessarily reliable in individual cases, or for syntax in general. However, the corpus is enormous, about ten times the size of the inscriptions. This gives them a statistical value and depth of lexicon not available from the other sources. From this total mass of data, a relatively large number of presumably native Elamite lexical items emerges, after all Old Persian, Akkadian, or other obvious loanwords are removed. A substantial number of verb stems is especially important. AE was written in a cuneiform syllabary with relatively few (ca. 20) word signs other than the numerals and determinatives.8 Unfortunately these word signs are
3 The major sources on Elamite are: Bork 1925, Paper 1955, Cameron 1948, Hallock 1969, and Hallock'sother works cited in the references,below. 4 Remarksby travelers,as recorded the by geographerIstahri,indicatethat Elamitecontinued to be spoken around Susa until after 1000 A.D. (Bork, 73). 5 When the Assyriansunder AssurbanipalcapturedSusa and most of Elam, the remnantof the ElamiteKingdom(called Anshan)fell to a branchof the Achaemenidclan from neighboring Pars; thus Elam was the first step in the Achaemenids'road to empire.Since it was this Anshan branchof the Achaemenidfamily that founded the empire,Elam was centralto it, and Susa was always a major administrative capital. Later, the Persianbranch of the family (i.e. from Pars), beginningwith Darius, gained the throne after a brief usurpation.It was Darius who built Persepolis primarilyas a ceremonialcapital. 6 The major inscriptionis the great Behistuninscriptionof Darius which is trilingualin Old I, Persian,Elamite,and Akkadian.This inscriptionprovidedthe originalclues for the decipherment of the cuneiformscripts. 7 These have been collectedand translated Cameronand by by Hallock 1969. 8 Paper (4-36) has a detailed discussion of the script and phonology. Hallock (1969:82-6) gives the actual sign lists and transliteration scheme.

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

TOWARD PROTO-ELAMO-DRAVIDIAN

91

used for some of the most common nouns, which as a result are often difficult to interpret phonologically. The considerable Persianization of the texts allows access to the writing system and phonology through names and loanwords. Voicing is definitely not contrastive; and while graphic gemination is contrastive in part, it is not yet certain that there was phonological gemination. In general, true geminates are normally written as such, while true non-geminates may vary. The written gemination contrast is much more consistent for sonorants than for obstruents. Three vowels (a i u) are always contrasted; a fourth (e) often is; and a possible fifth (u) is sometimes indicated, but more probably stands for [w]. Based on internal structuring and comparison with Old Persian and Akkadian, the basic units of the script (presumably also the phonemes) are:9 a i e u (i) (h) y k tp s s c r n I (1) m. The unit h is probably empty in AE; i.e., hV is only an initial variant of V. However, this is not the case for older forms of Elamite, where h is used in the verb morphology: cf. Middle Elamite huttahs 'they did', hutta 'he did' (AE has huttas for both). The units s and c do not seem to be in full contrast; separate signs exist for SACAand for si cI, but otherwise only one sign exists for the pair. I use I to represent the consonant of the sign EL. When they occur in Persian names, [m] and [v] are both written as m in Elamite. The units n and r tend to be interchangeable in some positions, particularly in case endings. It is almost certain that vowel nasalization existed and was not written except for the optional insertion of the
(presumably) original nasal: HI-DU-I.S and HI-IN-DU-IS both exist for 'Hindu', but

Old Persian may be an influence here. Second vowels in two-vowel sequences are usually dummies: DA-ISis written for das, etc. Many final vowels and some interconsonantal vowels are also to be ignored. The morphology is of the agglutinative type, and the structure of the verb is quite simple. Adjectives follow nouns, and verbs occur at the end of sentences. Other relevant details of the morphology will be discussed when compared with Dravidian. Little more can reliably be said about the syntax, because of contamination from Old Persian in AE.
LANGUAGES are indigenous only to South Asia,10 primarily 2. THE DRAVIDIAN

in southern India, where they form a continuous block of over 120,000,000 speakers.11As literary languages they have a long history, with Old Tamil going back at least to the early centuries A.D.,and splitting into Middle Tamil and Old Malayalam around 1000 A.D.Old Kannada is first recordedaround 500 A.D.,and Telugu around 1000. These four are the major Dravidian languages. Tulu and Kodagu are also important but non-literary members of the family. The other languages in the family are spoken by tribal groups, of which Gondi and Kurux are the most important numerically. Diachronically the Dravidian languages are divided into three major groups labeled South, Central, and North. South Dravidian (SDr.) consists of Tamil, Malayalam, Kannada, Kodagu, neighboring tribal languages, and
9 Because of its essentially graphic nature, the transcription of Elamite is normally given in italics. The actual cuneiform signs are cited in small capitals according to Hallock's scheme. Because of their more abstract nature, I normally cite Dravidian and PED between slashes. 10 Abbreviations used below for names of Dravidian languages are: Ta[mil], Ma[layalam], Ka[nnada], Te[lugu], Kur[ux]. 11The basic sources on Dravidian are Burrow & Emeneau 1960, 1968, Caldwell 1913, Kumaraswami Raja 1969, and all of Krishnamurti's and Emeneau's works cited in the references, below.

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

92

LANGUAGE,

VOLUME 50, NUMBER

1 (1974)

probably Tulu. Central Dravidian (CDr.) consists of Telugu, Gondi, and a number of tribal languages of Central India. North Dravidian (NDr.) consists of isolated Brahui and the closely related Kurux and Malto of Eastern India. Brahui, spoken south of Quetta in Pakistan, near the border with Afghanistan, is the least typical of the Dravidian languages. South Dravidian provides the great bulk of Dravidian forms-especially Old Tamil, which has a position much like that of Sanskrit in Indo-European. Studies in Dravidian have probably been overly influenced by Tamil. In general, Dravidian languages are consistently left-branching, SOV, agglutinating languages. Nouns and verbs tend to dominate the lexicon to the exclusion of other form classes. As a result, there are few true adjectives and adverbs; participles and other verbals are used in their place. The phonology of Proto-Dravidian is anomalous in having six contrasting stop positions: labial, dental, alveolar, retroflex, alveopalatal [affricate], and velar. Of these, all except the alveolar and retroflex stops occur initially; no obstruents occur finally. The alveolar is the least attested of the stops, and has disappeared in most Dravidian languages in favor of the five-stop pattern which is the South Asian areal norm: contrastive alveolar stop /r/ occurs today only in some dialects of Tamil, Malayalam, Toda, Kota, and Konda. It is realized as a tap or trill [?] when simple, and either as an affricate [t_r] (Tamil) or stop [tt] (Malayalam) when geminate. Three nasals-labial, alveolar, and retroflex-clearly contrast by themselves in the proto-language; but Old Tamil shows some evidence that dental and alveolar nasals also contrasted, and Malayalam evidence points to the possibility of an initial alveopalatal /fi/. Contrasting alveolar and retroflex laterals exist along with a simple tap /r/ and a coronal approximate /r/ [i]. Semivowels /y/ and /v/ also occur. Proto-Dravidian stops and laterals contrast in gemination when non-initial, but never in voicing.12Consonant clusters are restricted to sonorants followed by stops, and the clusters must be homorganic unless a juncture intervenes. The vowel system has five vowels /a i e u o/, with contrastive length but no nasalization. Vowels in non-initial syllables are less stable, and may fall into subphonemic vowel harmony.13Verb roots tend strongly to the pattern (((C)V)C)V((C)C), where V is any vowel, C is any non-vowel, and CC represents clusters or geminates. A summary of the phonemes and major clusters of Proto-Dravidian is given in Table 1.14
3. CORRESPONDENCERULES AND THE PHONOLOGYOF PED. The following corre-

spondence rules are based on the pairs of lexical items in the glossary (?4). The correspondences are sufficiently complete and detailed to allow a partial statement of the phonemes of PED. The rules are given in the following format: *A > B: C. Here *A represents the PED phoneme, B the AE graphical unit, and C the Proto12

It would be more accurateto say that Proto-Dravidian had a tense-lax contrastwhich could

be indicated phonologically by length, gemination, manner, and/or voicing, depending on the environment. I follow tradition by indicating this with written gemination. It should be kept in mind that this need not be true phonological gemination, e.g. in /1kk/, /Ikk/. 13 See Krishnamurti 1958 for a discussion of this problem. Since it is sometimes impossible to distinguish {/i/ /e/} and {/u/ /o/} before /a/ in Proto-Dravidian, I have marked these indeterminate vowels with a grave accent, i.e. */i/ and */u/. 14 The NCC cluster pattern is not attested in Dravidian; see Kumaraswami Raja for arguments concerning its validity.

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

TOWARD PROTO-ELAMO-DRAVIDIAN t-t-tt-pp-nt-mp-mpp- -nttmn-m-n- (?) -mm- -nn- (?) p-m

93

-p-

-r- -t-rr- -tt-nr- -nt-nrr- -ntt-n-nn-n

c-c-cc-fic-iiccn- (?)

k-k-kk-ik-irkk-

-n-nn-n

-1-, -11-,-1C-, -1; -1-, -!1-, -1C-, -!;

-r-, -rC-, -r; -r-, -rC-, -r (?); y-, -y-, -yy-, yC-, -y; v-, -v-.
i, i, e, e, a, 5, o, 6, u, u.
TABLE 1.

Proto-Dravidianphonemesand majorclusters,with their distribution.

Dravidian phoneme. The arrowhead (>) is read as 'gives rise to', and the PED environment is indicated by the slash and dash notation along with the common symbols. After each rule, the numbers in parentheses refer to the glossary where the examples are presented. Glossary pairs with numbers under 36 are prime correspondences; pairs with numbers 37-48 are of a secondary nature; and pairs with numbers above 48 are doubtful. A hyphen after a form indicates that it has been abstracted in some way (i.e., it is a root or stem), and the usual asterisk is used before unattested Dravidian forms.
3.1. VOWELS AND SEMIVOWELS.

(1) *e > 0:
*u (2) *a > a: a (3) *i > i

e /I#
u /

ta

(5 [PED *ita], 7, 49a [PED *ena]).

(4) *u > u: u (5) *o > u: oJ

#C

This set of correspondences holds for all the cognate pairs except 28 (see glossaryfor discussion) and 37.

(6) *e > eli: e / #C C

(45, 46, 51, 52, 56).

Two possible exceptions are bera-: paray- (19) and be- : vay- (or pey-?) (40). The /a/ in /paray/ is quite probablya phonologicallyconditioneddevelopmentin Dravidian.Note that cila and pir are grammaticalparticles in AE. This, together with the restrictedoccurrenceof e, probably indicates that Elamitehad a normal stress on the first syllable of a morpheme,as most often in Dravidian (but not always, cf. rule 1).

(7) *a > a: a (8) *i > i : / #(C)VC(C) (9) *u > u: u


include 14, 31 (with unstable unstressedvowel); Examples are numerous. Counter-examples 22, 54 (non-comparablebecause of derivationalmorpheme/-ay/ in Dravidian); and 8, 21, 48 (with differentstem vowels). (10) *Wi] > 0: v [e] I # IV (35a-b, 47, 48).

(11) *w >

: v/ V

(7; cf. rule 47).

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

94

VOLUME LANGUAGE, 50, NUMBER1 (1974) 3.2. CONSONANTS. , V_ # (14, 38, 51, 53). (12) *k > k: k / #V V (possibly with h ?) (24). (13) *k > 0 k 15 / V (14) *kk > kk: kk /VV V(39). (15) *sk > gk: kk/ V V(29). (16) *Uk > k: qk / V. V (35a). V (31). (17) *ykk > kk *:ykk / V

Another possible patternexists with sunki: *cukkay(54), but the connectionis doubtfuland the PED form is unclear.

VrC (15, 17; cf. rule 28). (18) *t > t : 0/ # *t > t : t V /# (16, 39, 56). (19) t > *t / t : V_ V (5, 7, 33, 38). (20) V (1, 20, 35b). (21) *tt > t(t) : tt / V V (15, 31; cf. rule 38). (22) *rt > rt : t /V V (18, 19, 20, 52, 57). (23) *p > p: p / #
The set be-: tay- (or pey-?) (40) may be an exceptionto this rule.

(24) *p > p: v/V V(11, 13b). V (26; cf. rule 16). (25) *mp > p(p) : mp / V V V (21, 22, 23; cf. rules 18 and 28). (26) *S > s : t I # V, V _# (2, 3, 12). (27) *s > s : y V > C where X = any C except li/r 0 # *s s: V, V VX, C (28) / and /1/ (6?, 29, 41, 42, 53; cf. rules 18 and 26).
Note in rules 27 and 28 how often the vowel next to the PED /g/ is long in Dravidian.This may be one origin of Dravidiancontrastivevowel length.

(29) *c> c: c/ (30) *c > s (31) (32) (33) (34) c

{} {a

(25, 44, 45). (24, 43, 54).

*cc > cc: cc V/V V(30, 34). *nc > ns: nc / V (37). *r > r: r/V V, V_ C (3, 6, 8, 18, 19, 21, 23, 43, 52). *r > rr: r/V V (14, 32, 41). V

Rules 33 and 34, with their examples,are in themselvesclear proof that Elamiteand Dravidian are cognate languages.For two languageseach to have contrastingr's is not common; but that these r's should correspondone-to-one, without counter-example,16 borderson the statistically impossibleunless the languagesare cognate.

(35) (36) (37) (38)

*n > n : n / #V (26, 27, 28, 29). *n > n: n/r (?) /V V (46, 49a). *nn > tnn: n(n) / V V, V. # (25, 42). *rn > rn : n / V V (17, 44; cf. rule 22).

151 use the symbol k for the weak /k/ of SDr. verb morphologywhich disappearsin many positions, see glossaryset 24. 16 However, three pairs are indeterminate.Glossary set 18 is ambiguoussince there are two equally plausible Dravidianentries. In set 14 the Elamiteform is ambiguous,and in set 43 the Dravidianform is ambiguous.

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

TOWARD PROTO-ELAMO-DRAVIDIAN

95

(9; cf. rules 22 and 38). (39) *nr > nr: n + r / V > *nk V nk: nk V /V (49b; cf. rule 16). (40) V (22, 35a-b, 45, 47, 48, 51; note rule 10). (41) *1 > : } / V (42) (43) (44) (45) (46) (47) *1>n: I/V #(27). *11> 11: 1(1)/ V V (16, 50). *1 > : 1 (36, 56, 57). *m > m : m / # V, V V (10, 31, 34; cf. rule 47). V (4; see discussion in glossary). *mm > mm (?) : mm / V *v > m : v / # V (30, 32, 33, 46, 55 [iur< *vur?]; cf. rules 45 and 11).

For rules 45-47, recall that the Elamitegraphicunit m does not contrast [m] and [v], and that its actual phonologicalvalue is ambiguous.

The correspondences are straightforward,and the variations are well within the limits of phonological possibility. If we add /y/, which exists in both AE and Dravidian but has no cognate pairs, then the tentative set of phonemes for PED is completed. This is given in Table 2. The major changes from PED to Elamite are
p-p-mpt-t-ttk-k-cc- -kk- ic- -rk-rkk-k (?) cm-m-mmn-n-nn-

-C-, -t; s-, - s-, -k-, -s; -l-, -ll-, -1;


-r-;,-r-, -rt-, -rn-, -rs-; (y,) -w, (v-?).
a, i, e, u, o

TABLE 2. Phonemesand clustersof Proto-Elamo-Dravidian.

the merger of /m/ and /v/ (perhaps with vowel nasalization taking over the contrast), the loss of /o/, and the loss of nasals before consonants. More detail is difficult, since so little is known of the actual phonology of Elamite. For Proto-Dravidian, contrastive vowel length was developed (retained?), and consonant clusters were simplified. There are several indications that Dravidian developed internal contrastive vowel length. One is that South Dravidian verb roots of the form (C)VCVC rarely have long vowels, while long vowels are fairly common in the form (C)VC. This may be the result of the loss of the middle consonant, with fusion of the adjacent vowels. Another indication is that, in purely Dravidian forms, long vowels are rarely followed by geminate consonants unless a juncture is involved; i.e., (C)VCC, (C)VC, (C)VC, and (C)VC+C are common, but (C)VCC is not. Third, and similarly, some archaic South Dravidian verbs vary in stem-vowel length; Ta. Ma. 'see' has /kan-/ for the present-future base (before vowel) and /kan-/ for the past base (before consonant); see also the discussion of glossary set 24. These data suggest that long vowels and long (i.e. geminate) consonants were in complementary distribution at some point in Dravidian. If we add a compensatory lengthening because of loss of /s/, it is quite likely that the great bulk of Dravidian vowel-length contrasts can be explained as an internal development. The simplification of the consonant clusters led to new

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

96

LANGUAGE,

VOLUME 50, NUMBER

1 (1974)

types of articulations in Dravidian;17 thus */rt/ and */rn/ became /t/ and /ni/ respectively. However, this is not the only source of the Dravidian retroflex series; PED */// gives rise to /1/, and vowel environment seems to influence the distribution of retroflexes. Where PED had prevocalic *t *l after *e or *i, the consonant involved is retroflex in Dravidian: see glossary sets 5, 33, 47, and 48, along with rules 20 and 41. I have no explanation for this odd rule, and it will be interesting to see if more data will verify or contradict the observation. The alveolar series in Dravidian arises from */r/, with */ni/ > /nr/ (see rules 33 and 39). However, it is not clear where Dravidian /rr/ comes from, possibly from */tr/ or */rt/. The origin of /r/ is also not clear, but */rs/ is a possibility (see glossary set 6). Dravidian has merged */w/ and */v/, if they were ever really separate. Other changes are relatively minor, and consideration is limited by our level of knowledge of Proto-Dravidian.
OF COGNATETERMS. The following numbered set of cognate terms 4. GLOSSARY

consists of AE words followed by Dravidian forms. The Elamite is given as normalized by Hallock 1969, except that c is used for z and (u)l is used for EL. Hallock's normalization is very conservative and, in general, uses the most common form as the norm. It should be kept in mind that voicing is not contrastive -i.e., be equals pe, etc.-and that m could be either [m] or [v]. Dubious readings of the Elamite forms are in brackets. The Dravidian form is either a reconstruction based on the information in Burrow & Emeneau 1960 and 1968 (DEDS), or a citation of the best attested form. The reconstructions are preceded by the usual asterisk, while the cited forms are preceded by the standard abbreviation of the language. Where a citation form is a reasonable approximation to the reconstruction, asterisk and language name are separated by a slash: (*/Ta.) indicates that the form attested in Tamil is usable as the Proto-Dravidian form. The Dravidian term is immediately followed by its referencenumber in DED(S). Dravidian orthographic ai has been normalized as /ay/. The ordering is that of the Elamite phonemes given in ?1 above.
(1) atta 'father': */Ta. /attan/ 'father, elder'; */Ta. /attay/ 'woman of rank, father's sister'
(DED 121).18

(2) as 'herd(s), domesticanimal(s)': */ay/ 'cowherd(caste), cow' (DED 283). (3) aras 'granary' (?), 'large building' (?) [cf. irfa- (6)]: */Ta. /aray/ 'room, chamber,treasury'
(DED 272).

(4) am[la?] 'mother'19: */Ta. /amma/ 'mother' (DED 154). (5) da- 'set, place, deposit; send': */Ma. /it-/ 'place, deposit, put; throw, cast away; give' (DED(S) 375). Cf. */Ta. /itam/ 'place, space, spot' (DED 368). Note rule 1. (6) irsa- 'great', irsana'great thing', irsara'great person' [cf. aras (3)?] : */Ta. /iray/ 'anyone who is great, king, etc.' (DED 448). Cf. */Ta. /eru-/ 'rise, ascend, be high' (DED 723a) [/r/ < /rs/?]
17This process has not yet completelystopped. In Malayalam,which has a very conservative phonology, Sanskrit ks is assimilatedas [c] (i.e. [ts]), contrastingwith the non-retroflex[c]. 18 The word /attan/ is archaic in all Dravidian languages having the term; */Ta. /appan/ 'father' is more usual. 19The Elamiteform occurs in a single unclearreading.The cuneiformsigns LA and MA differ primarilyin the angle of their first stroke, so ma could be a possible reading.Termssuch as this and 'father' (1) are, of course, weak evidence that the languagesare cognate. However, these terms are consistentwith correspondences based on other evidence;it would be surprisingif the languageswere cognate but these terms were not.

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

TOWARD PROTO-ELAMO-DRAVIDIAN

97

(7) dau-'help' : */Ta. /utav-/ 'help, be of use' (DED 522).20 (8) uri- 'believe': */Ta. /uru-/ 'pass in one's mind, think, heed' (DED 610). Cf. */Ta. /unni-/ 'think, consider, meditate' (DED 625). Especially note Brahui hunning(hur-, hutt-) 'look (for), consider' with both /nn/ and /r/ in one verb. (9) unra 'each', un 'one' + ra (?): */onr/ 'one thing, the number one' (DED 834d) [< /on/ 'one' + /t/ 'thing']. Cf. */Ta. /or/ 'one' (DED 834a).21 (10) umi- 'grind (grain)': */Ta. /umi/ 'husk, chaff' (DED 548). (11) hapi- 'apply pressure, press oil' (?): */Ta. /avi-/ 'become repressed,subdued; suppress, repress,extinguish,destroy' (DED 226). (12) hasa- 'count, reckon': */Ka. /aya/ 'measure,extent, boundary'(DED 311). (13a) hi 'this', opp. hupe'that': */Ta. /i(v)-/ [/i/ before C, /iv/ before V] 'this' (DED 351). A fundamentalDravidianformativefound in all Dravidian languages,contrastingwith */Ta. /a(v)-/ 'that' (DED 1) and with */Ta. /u(v)-/ 'near you' (DED 475). (13b) hupe 'that', opp. hi 'this': */Ta. /u(v)/ 'not near, not far, at an intermediatedistance; near you' (DED 475). This term is archaic in most modern Dravidian languages. (14) karri (kariri)'kid': */Ta. /karu/ 'young (of animal), foetus (of animal)' (DED 1074). Cf. */Ta. /kori/ 'sheep' (DED 1799)? (15) tarti- 'conceal, comprise' (?): */at-/ 'shut, close, obstruct, conceal' (DED 73). Note rules 18 and 22. (16) talli- 'write': */Ta. /tall-/ 'push, push forward,push in' (DED 2559).22 (17) turna- 'know': */Ta. /u.nar-/'awaken, be conscious, perceive, know' (DED 518). Note rules 18 and 38. (18) pari- 'go to, issue' : */Ta. /pari-/ 'run away, flow quickly, fly off' (DED 3311). Cf. */Ta. /pari-/ 'run, go out, escape' (DED 3268). (19) bera-'read' : Ta. /paray-/ 'speak, say, tell' (DED 3318). (20) putu 'lamb': */p6tta/ 'young (animal or plant)' (DED 3748); AND */Ta. /pottu/ 'male (of animals)' (DED 3747). (21) sara- 'cut off, divide' : */Ta. /tari-/ 'be cut off, cut, cut down' (DED 2562). Note rule 26. (22) salu 'a high social class', salur 'gentlemen': */Ta. /talay/ 'head' (DED 2529), a basic Dravidianterm. (23) sura- '(to) present' : */Ta. /tura-/ 'leave, forsake, abandon, give up' (DED 2768).23 (24) sa- 'go to, go off': */Ta. /ca(k)-/ 'die' (DED 2002). Cf. */cak-/ 'go, move forward,proceed, happen' (DED 2006).24 (25) cinna (qualifies 'boys') pps. 'infant': */Ta. /cinna/ 'little, small, young' (DED 2135), a basic Dravidianterm. (26) nappi'god' : */Ta. /namp-/ 'believe, trust, long for, confide in' (DED 2975). (27) nan 'day': */Ta. /nal/ 'day' (DED 3025). (28) nu 'thou', numi 'you (pl.)', -ni 'thy' : */ni/ (obl. */nin/) 'thou' (DED 3051); */nim/ (obl.
*/num/) 'you (pl.)' (DED 3055).25
20 There is also a parallelform */uitak-/'help, etc.' which occurs in many languages.Internal reconstructionwould favor /utak-/ as the older form, but /utav-/ seems very widespread.Note rule 1. 21 Since Elamitenormallyused figuresin writingnumerals,relativelyfew Elamitenumbersare known. Of these, none seems to be cognate with Dravidian, except possibly Elamite ki 'one' with Dravidian*/ok(k)/ 'one, whole' (DED 834b). Numbersystemsneed not be stablediachronically; Altaic has repeatedlyreplacedits numbers. 22 This is a striking consideringthat Elamiteused a cuneiformscriptemploying correspondence, a stylus and wet clay. 23 The actual meanings seem to be closer than the glosses suggest; cf. the Elamite hasurasta 'for [provisions]he presented'. 24 In Tamil and Malayalam,/ca/ is a term of disrespect,best translated'kick the bucket'. It is also one of the few verbs showing a variation in vowel length: Ma. cakuka 'to die', cattu 'died'. This is undoubtedlyan archaism. 25Both the Elamiteand Dravidianforms show variationbetween/i/ and /u/ in their forms for

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

98

LANGUAGE, VOLUME 50, NUMBER 1 (1974)

(29) nuske-'guard, keep, protect': Ta. /nokk-/ 'look at, see, watch' (DED 3144). (30) macci-'remove, withdraw'26: */vacc-/ 'carry off, seize, rob' (DED 4567). 'engagefor hire, (31) martukkas 'portion of herdpaid to herdsmanfor his services' : */matarJkk-/
put down deposit' (DED 3795).27

(32) marri-'seize, hold, occupy', marrip'artisans': */Ta. /vari-/ 'bind, tie, fasten' (DED 4305). Cf. */Ta. /vari/ 'tax, levy' (DED 4307). (33) mite- 'go forth': */Ta. /vit-/ 'leave, quit, abandon; let go, dispatch, throw' (DED 4419), a basic Dravidianterm. (34) mucci 'vat(s)' (?), 'some sort of storage vessel closed from above' : */Ta. /mucc-/'cover, close' (DED 4025). (35a) laki- 'go across' [cf. -lakka across] : Ta. /vilark-/ 'be transverse;turn aside; fall across, separate' (DED 4450). (35b) lati- 'reserve' (?): Ta. /vilatt-/ 'turn aside, divert,separate'(DED 4450).28

The following entries are of a secondary nature. They provide further evidence to support already established correspondences.
(36) ulma- 'think', ulma = ulma (?): */u!(!-/p-/k-)/ 'think, think on' (DED 600). Cf. */Ta. /ul/ 'interior, inside, mind' (DED 600). (37) unsa- 'receive in exchange': */Kur. /inj(r)-/ 'receive, accept, get' (DED 365; NDr. only). The suffix/r/ is a regularstem formativein Kurux. (38) kuti- 'carry away, bear, uphold', kutira 'bearer': */Ta. /kuti-/ 'jump, leap, escape; trot, jolt, shake while walking' (DED 1419). Also */Ta. /kutiray/ 'horse' (DED 1423).29 (39) tukki- 'cut, engrave' : */tukk-/ 'push, shove, remove' (DED 2689). (40) be- 'create' : */Ta. /vay-/ 'put, place, store up, create' (DED 4565). Cf. */Ta. /pey-/ 'rain; put, place' (DED 3610). (41) sarra- 'collect (?), assemble' : */lar-/'search, collect, gather' (DED 319). Always found compoundedwith */ac-/ 'search,examine,gather' (DED 306), as in Ta. /aray-/ 'search,
etc.'30

(42) sinnu-'come': */Ta. /in-/ 'bear (young),yield, produce'(DED 473). Also as a derivativeon the same stem: */Ta. /int-/ 'gather, come together,join' (DED 458). (43) sari- 'destroy': */car-/ 'tear, cut, split' (DED 1951). The /r/ is ambiguous;it may be /r/. (44) cirna'milk' (?): */Ta. /cinujk-/ 'ooze, issue in drops, drizzle' (DED 2079).
(45) cila 'so, thus, then, as follows' : */Ta. /cel-/ 'go, flow, pass on' (DED 2286).

(46) meni 'then, after, after that' : */ve(ri)n/'back, after, behind' (DED 4518).
(47) li- 'deliver, turn over to, communicate': */Ta. /vili-/ 'say, speak, call, invite' (DED 4460).

Cf. */pili-/ 'call, shout' (DED 3447). 'you'. Since this variationis unusualfor both languages,it mightreflecta variationin PED. Also note that both languagesform the plural in /m/-a very rare plural in both. It seems likely that both Elamiteand Dravidianhave replacedtheir 3rd person pronouns.The 1st person pronouns do not seem to correspond:El. u(n) 'I', nuku 'we'; Dr. */yfn/ (/yan/) 'I', */yam/ (/yam/) 'we
(incl.)', */nam/ (/nam/) 'we (excl.)'

For pairs 30-34, recall that the Elamite graphic unit m does not distinguish[m] from [v]. This is the longest fit of any cognate pair. In fact, the Elamitewould almost exactlyfit with a possible Tamil noun derivedfrom the verb stem, viz. *matakkay. 28 The Tamil forms in 35a-b seem to be derivativesof a stem */vila(C)-/'be across'. Note rule
27

26

10.
29

In Dravidian, secondary meanings have come to predominate. The main semantic connection

to AE is the second group of meanings, referringto movementsin carryinga headload. The Dravidian for 'horse' is actually too close to the Elamite for 'bearer', and is quite possibly a loanwordinto Dravidian. 30 That this is originallya compoundverb is indicatedby the fact that it is the only SDr. verb in its conjugationclass with two long vowels that is not an obvious compound. Otherwise,for this class long vowels are restrictedto monosyllables.

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

TOWARD PROTO-ELAMO-DRAVIDIAN

99

(48) lilu- 'come forth': Te. /velalu-/ 'go/come out' < */vel-/ 'be open, public' (DED 4526). Cf. */Ta. /veli/ 'be public; outside' (DED 4526).

The following entries have cognate pairs which are of a more dubious nature. Correspondences depending on their evidence alone are to be considered provisional.
(49a) na- 'say': */Ta. /en-/ 'say, utter, express,think' (DED 737).31 (49b) anka (or naka) 'if, when' [Conditionalof na-? Metathesis?Two ver6s?] : Ma. /enkil/ 'if', conditional of /en/. The conditional in /-kil/ is an archaic SDr. pattern kept in this frozen form. (50) ulla- '(transportand) deliver' (i.e. cause to be in a place) : */Ta. /ul-/ 'be, exist in a place' (DED 599), a fundamentalDravidianterm. (51) kellira 'commander, admiral' < kelu-(?) 'govern' (?): */kel-/ 'conquer, overcome, succeed' (DED 1641). (52) pir 'in addition, together': */Ta. /perukk-/'gather, pick up, glean' (DED 3623). (53) sak 'counterpart(payment),equivalency' : */Ta. /ak- (a-, an-)/ 'become,verbof identity,be' (DED 282), a fundamentalDravidianterm. (54) sunki 'king': */cukkay/'star, dot' (DED 2175). (55) mur'where, place (?)': */Ta. /Qr/'village, town, city' (DED 643). (56) telte 'a kind of fruit': */Ka. /tel/ 'thin, fine, delicate' (DED 2826). > */tel+tu/ 'delicate one'. (57) pul 'a kindof fruit' : */Ta. /puli/' sourness,tartness'(DED 3546);also Ta. /puli/ 'tamarind'. 5. COMPARISON OF MORPHOLOGIES. The basic morphology of Elamite is fairly well worked out, the principal remaining problem being that the meaning of the various verb forms is often ambiguous. The modern Dravidian languages have fairly simple morphologies. However, the major effort in historical Dravidian studies has gone into the phonology. As a result, our knowledge of ProtoDravidian morphology is spotty at best, although the morphology of Proto-SouthDravidian is fairly well worked out.32 For these reasons, the following discussion of morphology is uneven in quality and usually tentative in nature. The verb in Proto-South-Dravidian had only two basic forms contrasting in tense (or aspect?), traditionally labeled past and future. Another fundamental pattern contrasted the finite verb (i.e. final, with personal verb endings) with the non-finite (i.e. non-final, without personal verb endings). There was also a large set of participles and other verbals, and a separate negative verb conjugation (or negative verb?) A very important derivational pattern throughout Dravidian is the causative series,33 where the causative of a verb stem is formed by doubling the final consonant or by an augment, e.g. Ta. /matajk-/ 'be folded', /matakk-/ < */matarkk-/) 'fold', and /var-/ 'go', /varutt-/ 'cause to go'.
31 The verb /en/ is an absolutelyfundamentalterm in Dravidian,having many syntacticuses. It is the similarityof use, as much as the phonology, that indicatesthis pair as cognate. If metathesis is involved, it is a uniquecase; but vowel deletion seems more likely (see rule 1). 32 See 1971 for an excellent survey of Dravidianverb morphology.However, Subrahmanyam his conclusions may be too heavily oriented toward South Dravidian. See also Emeneau 1967 for a detailedcomparativestudy of South Dravidianverb morphology. 33This causativeseries can go as far as four degreesmorphologicallyin Modem Malayalam: var- 'come', varutt-,varuttikk-, and varuttippikk-. The term 'intransitive'is restrictedto the first verb root, since the first causativeof an intransitiveroot becomesa transitive-'to cause to be folded' equals 'to fold'. Manyverb roots in Dravidianare intransitive and agentless,and are thus translatedas passives.Thereare probablyno true passive formationsin strictlyDravidianstyles.

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

100

LANGUAGE, VOLUME 50, NUMBER 1 (1974)

The Elamite verb also had a very sparse morphology.34 The main contrast seems to have been between transitive and non-transitive forms. There was also a present-future distinction which was not morphologically involved in the transitive-non-transitive contrast. Each of the above forms had a secondary form with the element -ma- added as an auxiliary. Personal endings were used throughout. In addition there were 'final' and 'continuative' forms, besides the imperatives and verbals. Nothing in the morphological forms of South Dravidian and Elamite seem to correspond directly except possibly the Elamite present-future in -n(Conj. III) corresponding with the Old Tamil future (i.e. non-past) in -un/-um. Also, the 3sg. ending in Conj. III -n-ra- seems to correspond to the Dravidian personal ending */-anr/ (see below). In general, the verbal morphologies of the two languages seem capable of corresponding, but too little is known of the history of either language for anything definite to be stated. The situation with the nouns is somewhat clearer. Dravidian nouns have a simple agglutinative structure, noun (+ plural) + case. The plural marker is varied, with /-r/, /-kal/, and /-lu/ being the most common in South and Central Dravidian. Only four formal cases in Dravidian do not have an obvious internal development: the nominative, accusative, dative, and genitive. Nouns often have two stems, one for the nominative and vocative and another (called the oblique) for all other marked cases, but the details vary from language to language. The Elamite noun has a similar agglutinative structure, the usual plural marker being -_p.35Four of its simple cases-nominative, accusative, allative,36 and genitive-correspond to the basic Dravidian cases above. These are compared with their Dravidian counterparts in Table 3.37 Another striking correspondencebetween Elamite and Dravidian is in the 2nd person pronouns (see glossary set 28).
NOM. Acc. DAT. GEN. TAMIL KANNADA TELUGU BRAHUI ELAMITE 0 0 0 0 0 -ay -an/-am -n/-r/0 -nu)e -ku -(n)e -ke -aku (ALL.) -ki/-ka -a -a -na -in/-a -na/-ni TABLE 3. Comparisonof basic cases in Dravidianand Elamite.

Two noun formatives are also noteworthy. The way to form an abstract noun from another noun in Elamite is to add -me(-mi): sunki 'king', sunkime 'kingship, kingdom'. This is obviously related to Ta. /-may/, a formative of abstract nouns; e.g. alavalay 'babbler', alavalaymay 'babbling'. Elamite personal nouns are regularly formed in -ra: kellira 'commander' (51), kutira 'bearer' (38). This probably corresponds to the Dravidian 3rd person formative */-anr/ (Ta. -an, Te. -.duetc.), which is used both as a noun formative and as a personal verb ending. 6. No attempt has been made to be exhaustive in the search for Elamite-Dravidian correspondences, although the great majority of possible Elamite words were
possible corresponds plural /-v/ personneuterpronounsof Tamil and Malayalam. 36 The simple dative has no ending in Elamite.The allativeindicates'motion toward', as does the dative in Dravidian.(Information concerningthe Elamitenouns is taken from Paper,69-90.) 37 The Dr. accusative ending is usuallythe same as the nominativefor inanimates.
34 Informationon the Elamiteverb is abstractedfrom Hallock 1959. 35 It is that this to the in in the 3rd

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

TOWARD PROTO-ELAMO-DRAVIDIAN

101

investigated. Of about 300 good AE lexical items, i.e. those with a statable phonological form and a usable meaning, about 60 correspondences were established, with about 10 more of a very dubious nature. This list of 300 included meanings such as 'a kind of tool' and 'a kind of fruit', even though it is almost impossible to establish a good correspondence for such a word. Other Elamite words are available, but with very ambiguous meanings, if any. This paper is obviously the first step in a continuing process. There are undoubtedly errors in some of the forms of Elamite, and particularly of Proto-Dravidian as presented here. However, the major thesis that Elamite and Dravidian are cognate does not depend on a few etyma, but on the mass of data. The correspondences interlock too well for the thesis to be doubted.
REFERENCES M. 1966. Materials for a bibliography of Dravidian linguistics. Kuala ANDRONOV, Lumpur: Dept. of Indian Studies, University of Malaya. BORK,FERDINAND. 1925. Elam, B: Sprache. Reallexikon der Vorgeschichte 3.70-83. BURROW, T. 1946. Dravidian studies IV: the body in Dravidian and Uralian. Bulletin of the School of Oriental and African Studies 11.328-56. 1960. A Dravidian etymological dictionary. Oxford: Clarendon -, and M. B. EMENEAU. Press. 1968. A Dravidian etymological dictionary: Supplement. Oxford: Clarendon Press. REV. ROBERT. 1913. A comparative grammar of the Dravidian or SouthCALDWELL, Indian family of languages. 3rd ed. London. (Reprinted, Madras: University of Madras, 1956, 1961). G. 1948. Persepolis Treasury tablets. (Oriental Institute publications, GEORGE CAMERON, 65). Chicago: University of Chicago Press. M. B. 1962. Brahui and Dravidian comparative grammar. (University of EMENEAU, California publications in linguistics, 27.) Berkeley & Los Angeles: University of California Press. - . 1967. The South Dravidian languages. JAOS 87.365-413. T. 1958. Notes on Achaemenid Elamite. Journal of Near Eastern RICHARD HALLOCK, Studies 17.256-62. . 1959. The finite verb in Achaemenid Elamite. JNES 18.1-19. -- . 1960. A new look at the Persepolis Treasury tablets. JNES 19.90-100. . 1962. The pronominal suffixes in Achaemenid Elamite. JNES 21.53-6. . 1969. Persepolis Fortification tablets. (Oriental Institute publications, 92.) Chicago: University of Chicago Press. BH. 1958. Alterations i/e and u/o in South Dravidian. Lg. 34.458-68. KRISHNAMURTI, --. 1961. Telugu verbal bases. (University of California publications in linguistics, 24.) Berkeley & Los Angeles: University of California Press. - . 1969. Comparative Dravidian studies. Linguistics in South Asia (Current trends in linguistics, 5), ed. by T. Sebeok, 309-33. The Hague: Mouton. KUMARASWAMI RAJA, N. 1969. Post-nasal voiceless plosives in Dravidian. (Dept. of Linguistics, publ. 18.) Annamalainagar: Annamalai University. HERBERT H. 1955. The phonology and morphology of Royal Achaemenid Elamite. PAPER, Ann Arbor: University of Michigan Press. P. S. 1971. Dravidian verb morphology: a comparative study. (Dept. SUBRAHMANYAM, of Linguistics, publ. 24). Annamalainagar: Annamalai University. A. 1968. Dravidian and Uralian: the lexical evidence. Lg. 44.798-812. TYLER, STEPHEN 30 October 1972.] [Received

This content downloaded from 117.193.185.109 on Wed, 19 Jun 2013 11:26:21 AM All use subject to JSTOR Terms and Conditions

Вам также может понравиться