Krishnamurti Bhadriraju. The Dravidian Languages

Подождите немного. Документ загружается.

1.7 Dr

avidian and Indo-Aryan

and provided a more acceptable answer in my ﬁrst survey article (Krishnamurti 1969b:

324–5):

It is the Dravidian languages (particularly South Dravidian) which show

evidence of extensive lexical borrowing but only a few traits of structural

borrowing from Indo-Aryan. On the contrary, Indo-Aryan (particularly

Middle and Modern) shows large scale structural borrowing from Dravid-

ian, but very little lexical borrowing. How can we reconcile these conﬂict-

ing facts in order to work them into a framework of a bilingual situation?

I proposed an answer in a long footnote, as follows, which many subsequent researchers

seem to have missed:

That Middle Indo-Aryan and New Indo-Aryan have been built on a

Dravidian substratum seems to be the only answer. The fact that the invad-

ing Aryans could never have outnumbered the natives, even though they

politicall

y controlled the latter, i

s a valid inference. We may for

mulate the

situation as follows: If the speakers of L

(mother tongue) are constrained

to accept L

(2nd language) as their ‘lingua franca’, then an L

will develop

with the lexicon of L

and with the dominant structural features of L

and

= Dravidian languages, L

= Varieties of Sanskrit, L

= Middle

Indic. This is also true of modern Indian varieties of English, which

have an English (L

) lexicon but a large number of structural features

of Indian languages (L

). Here, of course, the situation is different since

the native languages have not been abandoned. But what is interesting

is that Indian languages have freely ‘borrowed’ words from English but

no structural features; transfer of only structural features excluding the

lexicon is evident when Indians speak English as a second language.

The hypothesis that most of the present New Indo-Aryan speakers should

have been originally Dravidians and also presumably Kolarians (Munda

speakers) was suggested long ago (see Caldwell 1956: 52–61). Quoting

Hodgson, Caldwell says, ‘ ...the North Indian vernaculars had been de-

rived from Sanskrit, not so much by the natural process of corruption

and disintegration as through the overmastering, remoulding power of the

non-Sanskritic elements contained in them’ (p. 53). Emeneau says, ‘In the

case of Sanskrit, however, the Dravidian substratum is easily accessible in

its dozen or more living languages, and in that a Proto-Dravidian can be

worked out, given enough scholars interested in the matter’ (1954: 285);

also see S. K. Chatterji (1957), see particularly, pp. 212–13 in which he

speaks of non-Aryan substratum of Aryan.

42 Intr

oduction

Recently, Thomason and Kaufman (1988: chs. 2, 3) have made a distinction between

‘borrowing’ and ‘interference through shift’. They consider the Dravidian and Indo-

Aryan situation comes under the latter type. Commenting on Emeneau’s (1956) remark

that ‘absorption, not displacement, is the chief mechanism in radical language changes

of the kind we are considering’, they say:

...the two basic assumptions are that Dravidian speakers, shifting in con-

siderable numbers to the language of the Indo-Aryan invaders, imposed

their own habits of (among other things) retroﬂex vs. dental articulation on

Indic as they learned it; and they were numerous enough to inﬂuence Indic

as a whole, through the eventual imitation of their ﬂawed Indic by original

Indic speakers. The interference in Indic from Dravidian is striking in view

of the fact that, as has frequently been observed, there are few old Dravidian

loanwords in Indic. In sharp contrast, Sanskrit inﬂuence on some literary

Dravidian languages has come about through borrowing – native speakers

of Dravidian languages are initiators of the structural changes – and accor-

dingly we expect, and indeed ﬁnd, large numbers of Sanskrit loanwords in

Dravidian. (1988: 39–40)

Apparently they also did not notice that I made a similar observation two decades earlier,

though I used dif

ferent labels, lexical borrowing

vs. structural borrowing.

Hock (1975, 1982, 1996) has persistently questioned the theory of a Dravidian sub-

stratum in Indo-Aryan from pre-historic times. This is questioning over one-and-a-half

centuries of scholarship in comparative studies. While Kuiper has provided evidence for

the integration of some accomplished Dravidians into the Aryan fold (see section 1.7.1),

Hock (1996: 57–8) uses this evidence to suggest that

Rgvedic Aryans and non-Aryans

met as ‘near-equals’. There is no such implication from the way that non-Aryans were

described as ‘dark-skinned’, ‘indistinct speakers’ and ‘godless’. In conformity with his

line of thinking, Hock calls the substratum theory a ‘subversion’ and supporters of it

‘subversionists’. His approach ignores both the histor

y and geography of Aryan and

Dravidian contact, and the fact that the evolution of Middle and Modern Indo-Aryan

has been a slow and unconscious process and is not the consequence of the Dravidian

natives deliberately ‘subverting’ the structure and system of Indo-Aryan. The scenario

with three Dravidian languages scattered at distant points on the northern periphery,

with several islands in central India, and with thick concentration in the south indicates

that most of

the early native Dravidian speakers in the north and centre had merged with

local speech communities within Indo-Aryan. Constraints of space prevent me from

countering his arguments, which sound clearly strained and biased.

1.8 Afﬁnity

between Dravidian

and languag

es outside India

1.8 Afﬁnity between Dravidian and languages outside India

In a short but well-informed paper Austerlitz (1971) has characterized most of the

comparisons made between Dravidian and the other families/languages such as Uralic,

Altaic, African, Basque, Sumerian, etc. as ‘unprofessional’, ‘typological’, ‘wrong’, ‘ex-

periments without intent to convince’, ‘unsystematic’ etc. (1971: 254–6). He suggests

ﬁrst the reconstruction of proto-languages of established families, and then proceed-

ing to compare the proto-languages as a method of making long-range comparisons to

reconstruct macro-families. He says, ‘In so doing we can collate our results with infor-

mation about migrations, paleontology, archaeology, and other ﬁelds and thus attempt

to capture a more realistic picture of the linguistic past of the continents’ (1971: 257–8).

1.8.1 Dravidian and Ural-Altaic

Caldwell hypothesized about the genetic connection between Dravidian and the so-

called ‘Scythian’ or ‘Turanian’ (see fn. 5). He said that the Dravidian languages ‘bear

a special relationship to a particular family included in the group, the Finno-Ugrish’

(1956: 65). During the twentieth century several scholars pursued the idea of a genetic

relationship

between Dravidian and Ural-Altaic.

Burrow, in one of his earlier

papers

(1944), gave a brief history of the theory with 72 etymologies referring to body parts

in Dravidian and Uralian. Zvelebil (1990a: 99–103) has surveyed the history of this

theme with bibliographical references, but he has not cited the line of argumentation

of each contributor. Krishnamurti (1969b: 326, 328–9) has reviewed the arguments of

Karl Bouda (1953, 1956) and Menges (1964). Earlier F. O. Schrader pursued the theory

(1937). Subsequent contributions, according to Zvelebil, include Pentti Aalto (1971),

Karl Menges (1977), S. A. Tyler (1968), and several recent publications of J. Vachek

including two in Arch´ıv Orient´aln´ı (1978, 1987).

Zvelebil says that ‘the most important agreements are in morphology’. This matter has

to be established in terms of identiﬁed morphs with similar function, or else the similari-

ties could be just typological. Zvelebil does mention such identities without any example

(1990a: 101). Pentti Alto (1971: 63–5) points out thatsome case sufﬁxes like Fenno-Ugric

lative

∗

-k(a), Proto-Uralian

∗

-m for accusative sg. compare favourably with Dravidian

dative

∗

-kk(V) and accusative

∗

-Vn. Vacek suggests the possibility of ‘prolonged ancient

contact’ leading to borrowing and diffusion (Zvelebil 1990a: 103). Zvelebil concludes

that the ‘Uralaltaic-Dravidian hypothesis remains the most promising’ (103). In review-

ing Lahovary’s various studies relating Dravidian with ‘peri-Mediterranean’, I called

it a ‘colossal adventure in time and space’ (Krishnamurti 1969b: 329). The method

that most scholars have followed is not comparing one proto-system with the other, but

showing parallels in selected features between some languages of one family and some

of the other famil

44 Intr

oduction

1.8.2 Dravidian and Elamite

Zvelebil (1990a) devotes considerable space to survey the claims made by scholars re-

lating Dravidian to Elamite (104–15) and Japanese (116–22). I also reviewed McAlpin’s

papers (1974, 1975, 1979) in my 1985 survey of comparative Dravidian studies. His

book was published in 1981 and not included in my survey article written in 1980. I

quote the following (1985: 225–6):

(1) Many of the rules formulated by McAlpin lack intrinsic phonetic/

phonological motivation and appear ad hoc, invented to ﬁt the proposed

correspondences: e.g. Proto-Elamo-Dravidian

∗

e > Ø Elamite, when

followed by t, n, which are again

followed by

a; but these remain

undis-

turbed in Dravidian (1974: 93). How does a language develop that kind

of sound change? This rule was dropped a few years later, because the

etymologies were abandoned (see 1979: 184).

(2) He set up retroﬂexes as an innovation in Dravidian resulting from

PED

∗

rt (1974: 94). Later he abandoned this rule and set up retroﬂexes

and dentals for PED and said that Elamite merged the retroﬂexes with

dentals (1979: see chart on 184–5). But the following statement in the

body of the article, referring to his updated version of correspondences,

is puzzling: ‘The major additions have been ...the splitting of the Proto-

Elamo-Dravidian dental series into dental and post-dental series reﬂecting

the dental-retroﬂex contrast in Dravidian.’ But the chart shows merger in

Elamite and not

split of PED dental into dental

and retro

ﬂex in Dravidian!

The correspondences between 1974 and 1979 have undergone total change,

which meant that earlier ‘etymologies’ were abandoned and new ones

commissioned. The 1979 correspondences sound more plausible, but the

etymologies are weak; e.g. PDr. ¯ı n ‘to bear young (of animal), to yean’ is

said to be cognate with Elamite ˇsinni ’to arrive’ by positing loss of PED

∗

ˇs in PDr. This semantic connection is perhaps dictated by the author’s

English language background in which people refer to ‘the arrival of a

baby’. But it is extremely odd to attribute this adhoc semantic connection

to PDr. For those who know Dravidian, this meaning shift is extremely

spurious and adhoc.

Zvelebil (1990a) reviews McAlpin’s study in greater detail, but ends saying ‘I am also

convinced that much additional work is to be done and many alterations will have to be

made to remove the genetic cognation in question from the realm of sheer hypothesis

and establish it as a fact acceptable to all’ (115). In his book (1981) McAlpin shows only

81 items, out of a corpus of 5,000 lexical items of Elamite, as possible cognates with

Dravidian. The basic problem is the small size of the database, and the fact that it belongs

1.8 Afﬁnity

between Dravidian

and languag

es outside India

to the sixth century BC, by which time the major Dravidian languages had already split

(mainly Telugu and Tamil). In my recent research in reconstructing a Proto-Dravidian

laryngeal, I noticed two important cognates that have h- in Elamite,

the deictic roots

∗

ah ‘that, not far’,

∗

(h) ih ‘this’ and

∗

huh ‘that (remote)’ (McAlpin 1981: 81–2, A3-6;

also 26,

§ 221.31 (5); Krishnamurti 1997b: 149, fn. 2; also see Zvelebil 1990a: 111);

secondly, the Proto-Dravidian verb

∗

caH- ‘to die’ has a close cognate in Middle Elamite

sa- ‘life to be cut off’, sahri ’death’ (McAlpin reconstructs

∗

cah- for PED, see 1981:

99). We need more cognates of an atypical kind to rule out the possibility of chance.

1.8.3 Dravidian and Japanese

Susumu Ohno has been a persistent

writer on the relationship betw

een Japanese and

Tamil. It was pointed out to him that any relationship with Japanese should go back to an

earlier stage than Tamil (i.e. Proto-Dravidian). He published a book (1980) dealing with

the so-called correspondences between Tamil and Japanese without any reference to the

external history or archeology of the two speech communities to support his hypothesis.

Zvelebil says that there is ‘evidence’ of relationship but no ‘proof’ (1990a: 117). Some

are ‘false cognates’, he says, Ta. c¯ettu ‘red’ [DEDR 1931 c¯e-(<

∗

k¯e-)] with Japanese

sita ‘red’. This change is dateable to the third century BC in Tamil. Zvelebil concludes

saying, ‘there is strikingly close typological afﬁnity between Dravidian (particularly

Modern Tamil) and Japanese in their morpho-syntactic structure’ (1990a: 121–2). The

typological similarities arise from the hypothesis that Japanese was also considered an

offshoot of the Altaic stock.

1.8.4 Dravidian and Nostratic

Holger Pedersen, the Danish linguist, as early as 1903, proposed that the Indo-European

languages could be genetically related to Semito-Hamitic/Afro-Asiatic, Uralic, Altaic,

Yukagir and Eskimo and called this macro-family ‘Nostratic’, from Latin noster ‘our’,

meaning ‘our primordial tongue’. Former Soviet linguists, Illiˇc-Svityˇc, Dolgopolsky,

Shevoroshkin, etc., working independently, expanded the hypothesis to include some

other families including Dravidian under Nostratic. There is no agreement even among

the supporters of the hypothesis on what language families make up Nostratic. There has

been a great deal of writing on this hypothesis during the past three decades. Illiˇc-Svityˇc

(b.1934–d.1966) compiled an etymological dictionary of 607 items from these families

with a statement of phonological correspondence before his premature tragic death. The

Nostratic hypothesis has its adherents as well as detractors. Greenberg (2000) proposed

the reconstruction of a Eurasiatic family and provided reconstructions of grammatical

McAlpin has apparently based his reconstructions on my proposal for a Proto-Dravidian laryn-

geal in my 1963 review of Burrow and Emeneau (1961). I made that observation purely on the

basis of internal evidence based on Old Tamil ¯aytam.

46 Intr

oduction

morphemes. He excluded Afro-Asiatic and Dravidian from Eurasiatic, but included

Uralic–Yukaghir, Altaic, Eskimo-Aleut, and Korean–Japanese–Ainu. In a personal com-

munication in October 2000 at Stanford, he said that Dravidian could be a sister of

Eurasiatic and not a daughter.

I will conﬁne myself to examining Dravidian data in relation to Nostratic. Dolgopolsky

(1986) has proposed ‘morphemic stability’ as a criterion for selecting the lexical items

for comparison. He arrived at ﬁfteen items with very high ‘morphemic stability’ cross-

linguistically. He then goes on to cite ‘phonetic similarities’ among the lexical items

representing these ﬁfteen from different families. Unfortunately Dravidian is not given

in this listing, because he does not believe Dravidian to be a member of Nostratic

(Campbell 1998: 110), but Bomhard and Kerns (1994) include it. I give the Proto-

Dravidian reconstructions of these and cite some of the reconstructions of Illiˇc-Svityˇc

from Proto-Nostratic (PN) cited by Campbell (1998: 124–8). The Eastern branch is said

to include Uralic, Altaic and Dravidian.

Dolgopolsky’s items Proto-Dravidian reconstructions [DEDR]

(1) ﬁrst-person marker

∗

y¯a-n/

∗

ya-n- (mine:

∗

yAH -n-) [5160, 5154]

(2) ‘2’

∗

¯ı r /

∗

ir- [474]

(3) second-person marker

∗

n¯ı-n [3684, 3688]

(4) ‘who’ ‘what’

∗

y¯a-w-ant/-at (mine

∗

¯AH -a-) [5151]

(5) ‘tongue’

∗

n¯a-l [3633]

(6) ‘name’

∗

pin-cc-V-r [4410]

(7) ‘eye’

∗

n- [1159a]

(8) ‘heart’

∗

t-V- [1693]

(9) ‘tooth’

∗

pal [3986]

(10) verbal neg (both

∗∗

-aH-/

∗∗

aH-aH

negative proper

and prohibitive)

(11) ‘ﬁnger/toe nail’

∗

ok-Vr,

∗

k¯or- [561]

(12) ‘louse’

∗

p¯en [4449]

(13) ‘tear’ (noun)

∗

n-n¯ır ‘eye-water’ [1159b]

(14) ‘water’

∗

n¯ır [3690]

(15) ‘dead’

∗

caH- ‘to die’ [2426]

The reconstructions from other families do not match most of these. There is no adjec-

tive ‘dead’ in Dravidian. Instead the past par

ticiciple

‘the dead ...’ from the verb ‘to

die’ is used. It is not clear what (13) means. There is a comment ‘no correspondences’

(Dolgopolsky 1986: 43). I have taken it to mean ‘a tear drop’ not ‘a cut’. Illiˇc-Svityˇc’s

reconstructions: Uralic

∗

mi (1); in Dravidian -n/-m are sg/pl sufﬁxes in items (1) and

(3), PN

∗∗

to ‘two’ (2), Proto-Uralic

∗

ti ‘thou/you’ (3), PN

∗∗

K’e ‘who’ (4),

∗∗

K’¨a/lH/¨a

‘tongue’(5),

∗∗

nimi ‘name’(6),

∗∗

HuK’a ‘eye’(7),

∗∗

k’ErdV- (8) ‘heart’,

∗∗

p’/alV

1.8 Afﬁnity

between Dravidian

and languag

es outside India

‘tooth’(9),  

∗∗

¨ala (10),

∗∗

p/a/r/¨a ‘nail’(11),

∗∗

t’ ¨ajV ‘louse’(12),

∗∗

wete

(14) ‘water’,

∗∗

m/¨a/rV ‘die’(15). <K> is a cover symbol for any velar consonant; <’>

a glottallized stop. Among these only item (9) ‘tooth’ matches, but the reconstruction

does not explain Indo-European: Lat. dent-, OIA dant-. South Dravidian I has

∗

al as a

negative verb, which looks like the reconstruction in (10). Even then two out of ﬁfteen

(6.5 per cent) can be simply chance resemblances.

Zvelebil (1999) says, ‘There are, from Dravidian perspective, reconstructions which

I would not certainly hesitate to accept’ and gives some speciﬁc reconstructions. The

foregoing is enough to show that the question is still very speculative.

1.8.5 Dravidian and Harappan

There is no decisive support to the recent proposals that Aryans were native to the

Indian subcontinent. The hypothesis that the Indus civilization was Indo-Aryan is also

rejected by knowledgeable historians, because of several factors: the Indus civilization

around Harappa ended before the Aryans entered India; the seals bear evidence of a non-

preﬁxing language; it ‘had an extensive agrarian base and an urban population dependent

on food production in rural areas’ (Thapar 2000:13); there is no evidence that the Indus

valley people (like the Dravidians) knew of the lion and the horse; for the Aryans, the

knowledge of the tiger and the elephant came much later.

Zvelebil gives the history of the hypothesis tracing the Harappa culture to proto-

Dravidian as well as a summary of the present state of the art (1990a: 84–98). The

period of the Indus civilization is broadly placed between 2500 and 1700 BC. There

were over 3,000 seals written in a script with 419 signs. It was likely to be logosyllabic

with some signs (pictograms) representing full words. Russian and Finnish scholars com-

puterized the data and made an effort to relate meaning to the signs in the 1960s. In India

Mahadevan (1977) published a concordance of the signs, with a listing of the sequences

and frequencies of all those that co-occur with a given sign. The Finnish scholars also

prepared another concordance. The writing was right to left, sometimes both right to left

and left to right. Numerous

attempts have been made at decipherment but none of these

has borne fruit. The latest book-length treatment of decipherment of the Indus script is

by Asko Parpola (1994) but it only explains certain hunches with no decipherment.

Possehl (1996) has critically reviewed different attempts at decipherment by thirty-

ﬁve scholars between 1924 and 1992 and concludes: ‘Since there is little basic research

on the script and so little sharing of programmatic visions, it is scarcely a wonder that

the writing system has not yet been understood’ (168).

The subtitles of the above four subsections as well as much of the contents are taken

from Zvelebil’s recent, painstaking study of these problems. Since even by Zvelebil’s

own confession they are all speculative studies, I have not made an original study of

these except where I have given references to works that I have consulted.

Phonology: descriptive

2.1 Introduction

The Dravidian languages retain most of the contrasts of Proto-Dravidian vowels and con-

sonants. Proto-Dravidian had ﬁve vowels /ieaou/with length contrast /¯/. Alternatively,

one can set up ten vowels for Proto-Dravidian, ﬁve short and ﬁve long. Diphthongs [ai

au] can be treated as sequences of a vowel + a semiconsonant, i.e. /ay aw/ patterning

with VC. There were seventeen consonants in Proto-Dravidian: six stops /p t t

tck/,

four nasals /m n

n˜n/, two laterals /l

l/, one ﬂap /r/, one retroﬂex frictionless continuant

z/, three semivowels /w y H/. The last one is a laryngeal, which patterns phonologically

with semivowels. All departures from this system can be traced to two sources: (a) cer-

tain sound changes within the historical period of individual languages or subgroups;

(b) borrowing from contact languages, either of the same family or of a different family.

The ﬁve-vowel system is fairly stable in all subgroups; new centralized vowels have

been added

, through splits, to the list of

phonemes of several of the Nilgiri

languages,

namely Toda, Iru

la, Ku

rumba of South Dravidian I. Long vowels are less subject to

change than the short vowels. Four of the stops have remained stable in the whole fam-

ily: /p t

t k/. The affricate /c/ gets variable treatment in different subgroups, represented

as [ts c s]. The alveolar /t

/ also has undergone change and has been eliminated as a

distinctive unit outside South Dra

vidian. Voiced or lenis allophones of stop phonemes

have developed within the native element in all languages and subsequently they became

phonemic, except in Old Tamil and Malay¯a

lam. For all other subgroups, both voiceless

and voiced stops need to be reconstructed. Among nasals /m n/, and among liquids

/r l/ have been the most stable. Retroﬂex /

n/ have been gradually replaced by /l n/

in all but the languages of South Dravidian I, while the frictionless continuant /

z/, a

highl

y marked segment, has vanished in most languages, having

merged with different

phonemes in different languages; it has lost its phonemic identity, except in Tamil (di-

alectally) and Malay¯a

lam. Aspirated stops and affricates, voiceless and voiced, have been

added to the inventories of Kanna

da, Telugu and Malay¯a

lam from the beginning of lit-

erature in these languages, owing to extensive borrowing from Sanskrit and the Prakrits

(see sections 10.2–3). Even some non-literary languages (particularly Kolami, Naiki,

2.2 V

owels

rux) have added aspirated stops through bilingualism in surrounding Modern Indo-

Aryan languages. The Indo-Aryan-like completeness in the stop series in ﬁve points of

articulation – labial, dental, retroﬂex, palatal and velar – could have been a typologically

based reason why alveolar

∗

t was eliminated from the system, since there could not have

been aspirated series involving it. Also note that the alveolar

∗

t is the most highly marked

segment in the series dental–alveolar–retroﬂex among the Dravidian languages and the

languages of the South Asian area.

Alveolar and retroﬂex consonants did not begin a word in Proto-Dravidian and this fea-

ture is preserved in all except in one subgroup (South Dravidian II). In South Dravidian I,

during recent times, words beginning with r and l appear through loss of the ﬁrst syllable,

e.g. Ta. Ma. ra

du (<

∗

ira

tu) ‘two’; in Ir

la and T

lu this

change has spread to several

lexical items, creating words with initial r,

d and l. Consonant clusters did not occur

word-initially in Proto-Dravidian. Even here only one subgroup (South Dravidian II)

makes a departure. The Proto-Dravidian laryngeal /H/ was preserved in a few cases

in early Tamil as a distinct sound, called ¯aytam [aydam] (section 4.5.7.2 (c)). Its

presence is inferred from the lengthening of a preceding short vowel in most cases

(Krishnamurti 1997b; see

section 4.5.7.2c). The descripti

ve and typological changes in

vowels and consonants in different subgroups and languages are dealt with in greater

detail below.

2.2 Vowels

2.2.1 South Dravidian (SD I)

The inherited system of ten vowels /ieaou¯ı¯e¯a¯o¯u/ is retained intact in Tamil and

Malay¯a

lam, both old and modern. Ko

dagu and most of the Nilgiri languages (Toda,

Iru

la, Ku

rumba, etc.) have developed the centralized vowels ¨ı ¨ı ¨e ¨e when followed by

retroﬂex (sometimes alveolar) consonants. These apparently were allophonic to start

with but became phonemic with the obliteration or modiﬁcation of the conditioning

environments.

Modern Tamil and Malay¯a

lam have added /

æ/ (lower mid front vowel) through bor-

rowing from English words such as ‘bank’. Modern Tamil also has nasalized vowels

developing from word-ﬁnal Vm/n sequences, but ﬁnal m/n are retained when a vowel fol-

lows, e.g. av ˜a/avan- ‘he’ (the nasalized vowel is fronted dialectally [ε

]), mar ˜a/maram-

‘wood’ (following vow

el rounding [o

/ɔ

], dialectally). Most words in Modern Tamil end

in vowels. Words beginning with a front vowel have a [y] onglide in speech; similarly, a

rounded vowel has a [w] onglide (Annamalai and Steever 1998: 102, Asher and Kumari

1997: 406, Schiffman and Eastman 1975: xix). A short /u/ following a word-ﬁnal stop

(in Modern Tamil any consonant) is phonetically a back unrounded vowel [¨ı] which was

called the ‘enunciative vowel’ (Bright 1975:

§§ 3ff.). This is regularly lost when followed

by a vowel across a morph or word boundary, n¯a

tu ‘country’ + i

n → n¯a

t-in-, obl of n¯a

tu.

50 Phonolo

gy: descriptive

The only exception is stems of the CVP-u type (P = voiceless stop; Kumaraswami Raja

1975).

Toda has the following vowels in addition to the ten-vowel system: ¨u

¨u ¨ı ¨ı (front

rounded and back unrounded vowels) and ¨o ¨o:, mid central rounded vowels (Emeneau

1957: 21, 1984:7). Shalev et al. (1994) conﬁrm the phonemic analysis of Emeneau and

set up sixteen vowels for Toda, despite the gap of sixty years in the collection of ﬁeld data.

They consider, on acoustic grounds, that ¨ı

¨ı are ‘high-mid central unrounded vowels’

and ¨o

¨o, ‘central, mid rounded vowels’. They add [æ], which Emeneau treated as an allo-

phone of /e/. They have further discussed the variation in allophonic quality and duration

of vowels before voiceless and voiced consonants: ‘the V/V ratio of 50% is maintained

even when other

factors (the voicing value

of the following consonant) affect

vowel dura-

tion, which indicates that this ratio is the cue for quantity’ (Shalev et al. 1994: 28). They

say that no other Dravidian language has ¨ı ¨u ¨o (central high unrounded, front high and mid

rounded vowels) as described here, which are unique to Toda with restricted distribution

(1994: 29–30). This is not true since Iru

la also has similar vowels (Zvelebil 1973: 11–12).

Unlike the other Nilgiri languages (Emeneau 1944, vol.1: 15ff.), Kota has no central-

ized vowels and it is puzzling

since its closest sister Toda

has them.

Iru

la has ¨ı ¨e ¨u ¨o, short and long. According to Zvelebil (1971a: 116–17), these are

‘central (unrounded and rounded) vowels, respectively high and mid’. He says, ‘the

feature “rounded: unrounded” is not entirely clear so far. What is clear is the position

of ¨u and ¨o which are further back than ¨ı and ¨e, respectively’. He added ¨a

¨a to

his earlier

(1971a, 1973) inventory of Iru

la phonemes in The Irula (

rla) Language, Part II (1979:

23), e.g. p ¨a

ndi ‘pig’, ¨ane ‘elephant’. Perialwar (1978b: ch. 2) sets up ﬁve centralized

vowels corresponding to the inherited ﬁve non-centralized vowels. He adds ¨a to the list

of Zvelebil. He says that the feature of rounding is present in ¨u and

¨u. It is not certain if

the centralized ¨a and

¨a are phonetically [ə] shwa-like. Clarity of phonetic deﬁnition is

lacking in the descriptions of Zvelebil, Kapp and Perialwar. Difﬂoth (1968 [1976]: 129)

describes ¨u as ‘a centralized rounded back vowel’ and he calls it the enunciative vowel.

In Tamil the enunciative vowel is the unrounded back vowel [¨ı].

Zvelebil gives the following examples,

to which I have added the Tamil cognates and

references to DEDR:

piy ¨e ‘crime’: Ta. pi

zai ‘fault’ [4187]

¨ıye/k ¨ı ¨e- ‘down, below’: Ta. k¯ı

z ‘below’ [1619]

v ¨uyu ‘to fall’: Ta. vi

zu ‘to fall’ [5430]

¨uy ‘to surround’: Ta. m¯u

z [5030]

¨e

tu ‘eight’: Ta. e

tu ‘eight’ [784]

¨ekka ‘to hear’: Ta. k¯e

l ‘to hear, to ask’ [2017]

k ¨ottu ‘neck’: Ta. ka

zuttu ‘neck’ [1366]

¨o

lu ‘seven’: Ta. ¯e

z ‘seven’ [910]