Krishnamurti Bhadriraju. The Dravidian Languages

Подождите немного. Документ загружается.

4.5 Historical

phonolo

gy: consonants

131

the succeeding sonorants is also to be taken into account. Thus k- is more

easily inﬂuenced to alternate with g-, than c- with j- and t- with d-; p->

b- is extremely rare, while

t- >

d- never occurs on account of lack of forms

with

t- in Dravidian (Krishnamurti 1961: 28–9).

Zvelebil’s (1972a) study of initial voiced stops based on DED showed that the ratio

between voiced and voiceless stops in Dravidian was 1:10. The highest rate of voicing

was found in South Dravidian II and Central Dravidian, while in North Dravidian the

ratio was 1:14. The study was useful, but we need to remember that languages studied

have different sized lexicons. It may be rewarding to take a list of two hundred or so

words with initial stops for which cognates are available in most of the languages in all

the subgroups and study the pattern of voicing in these. Emeneau’s study of expressives

(1969a, 1993) shows that in this category of words one can expect voiced stops to be

reconstructed for Proto-Dravidian, at least for some of the small subgroups:

∗

g-: gala gala Ka. Tu.; Te. [1302]

du gu

du Ko. Ka. Tu.; Go.; Kol.

Pa. [1659]

na ga

na Ka. Tu.; Te. [1162]

gama gama Ko. Ka. Tu.; Te.; Ku

r. [1247]

∗

d-: daga daga Ko. Ka. Tu.; Te.; Ku

r. Malt. (<? IA) [2998]

It is also a question of how old these expressions are in the literary languages to make

sure

that they have not culturally diffused during recent times.

Another area to look for is a possible ‘voicing laryngeal’ in early Dravidian. Several

etymologies with widely distributed voiced stops in many languages have unexplained

aspiration associated with stops, both voiced and voiceless, in Naik

ri, a Central Dravidian

language, and in Naiki (Chanda). These are not borrowed items from Indo-Aryan, e.g.

Naik

ri:

(54) gha

di ‘grass, straw’, Kol. ga

di;Te.ga

di, Go. ga

di, ga

du,Ko

da ga

di,

Kuvi ga

dr¯ı|| IA kha

ta-, kha

da- CDIAL 3769 [1158].

(55) gh¯a

li ‘wind, air’, Nk. (Ch.) gh¯ay id., Kol. g¯ali; Ka. Ko

d. Tu. g¯a

li;Te.g¯ali,

Go. g¯al,Ko

da g¯ali, Kuvi g¯ali [1499].

(56) ghel ‘to win’, Kol. gell-; Ta. keli ‘to conquer’, Ko. gel-(ge

d-), To. kelc-,

d. gel-(gedd-), Tu. gelpuni, genduni, Ka. gel-(gedd-); Te. gelucu,Ko

gels-, Kuvi gelh- [1972].

(57) ghummi ‘storage basket for grain’; Te. gummi id. Ta. Ka. Ko

d. forms with

k- mean something else [2117].

(58) ghun

d ‘stone’, Kol. gun

d;Ta.ku

tu ‘ball, anything globular, bullet’, Ko.

dgal‘a huge round stone’, Ka. Tu. gu

du;Te.gu

du ‘rock, bullet,

anything spherical’, gu

dr-ani adj. ‘round’, Go. gon

dra ‘round’ [1695].

132 Phonolo

gy: historical

and comparative

(59) ghurram ‘horse’, Nk. (Ch.) kurmam, Kol. gurram,Pa.gurrol;Te.gu

r(r)

am, Go. gurram,Ko

da gu

ram, Kuvi gurromi [1711b].

SD I:

∗

kut-ir-ay forms do not seem to be phonologically related to the above set.

(60) dh¯av ‘distance’, Kol. davva, daut¯an ‘distant’; Te. dawwu ’distance’, Nahali

dhava ‘distance’ [?446].

The other cognates gi

ven by

DEDR looking to

∗

t-ay do not ﬁt phonologically. A re-

lated form in Nahali suggests another substrate language as the ultimate source. Naiki

(Chanda):

(61) khaj ‘itch’, Pa. kajra;Ko

d. kajji,Tu.gajji, Ka. kajji, gajji ‘scab, itch’; Te.

gajji ‘itch’, also kasi, Go. gajji, gajju,Ko

da gazi,Kuikas;Ku

r. khasr¯a

id., x¯a

sn¯a ‘to scratch for relief’ ||>Pkt. khajj¯u CDIAL 3827 [1104].

(62) kh¯ır ‘line’, Kol. k¯ıra,Pa.g¯ı r, Gad. g¯ıri;Ko

d. g¯ı c- ‘to mak

e a mark by

scratching’, Ka. g¯ı

ru ‘to scratch’, n. ‘a line’, g¯ıku, g¯ıcu v.t., g¯ı

tu n., Tu.

g¯ıruni v.t., g¯ı

t¨ı ‘line, scratch’; Te. g¯ı

ru v.t., g

¯ıku, g¯ıcu, g¯ıyu ‘to scratch’,

¯ıta, g¯ı

tu ‘stroke, line’, Go. ‘line’, Ko

da g¯ı r n., Kui–Kuvi g¯ıra ‘line’; A

number of languages have k- forms: Ta. Ma. Ko. Tu. ||>? Skt. k¯ıraka-‘a

scribe’ [1623].

(63) phar ‘big, elder’, Pa. berto ‘big’, Oll. berit;Te.bebbuli(< per-puli) ‘tiger’,

Go. ber- ‘big’, Kui beri beri inba ‘to swell, increase in size’. Most other

languages have p- [4411].

Why does Naik

ri have aspirated voiced stops in native words, and that, moreover,

where the aspirates correspond to voiced stops in many subgroups, mainly in South

Dravidian and Central Dravidian? It has unaspirated g- (16 items) contrasting with

gh- (10 items); similarly, dh- (3 items): d- (2 items),

dh- (3 items):

d- (8) all of which

appear to be loanwords from Early Telugu, kh- (8 items): k- (c. 80 items), ph- (2 items)

p- (70 items). The cognates in the case of kh- and ph- have no voiced counterparts. The

related words in IA and Nahali (perhaps borrowings from Dravidian) for ety. (54), (60)

and (61) also show aspirated stops. The evidence is preponderantly with gh- correspond-

ing to voiced g- in most other languages. It is better to leave the matter here without any

conclusion, until more ﬁeldwork is done on Naik

ri and Naiki.

4.5.2.1 A quantitative study: voiced or voiceless stops

in Proto-Dravidian?

Methodology I have adopted a quantitative approach to determine if Proto-Dravidian

had a contrast between voiced and

voiceless stops in word-initial position.

The ﬁrst step:

I am indebted to Lincoln Moses, my former research collaborator (see Krishnamurti 1983) and

Statistical Consultant at the Center for Advanced Study at Stanford, who suggested this approach

to me in October 2000. It is a fairly standardized technique.

4.5 Historical

phonolo

gy: consonants

133

using DEDR (1984) as the source, I listed separately the entries with initial stops k/g, c/j,

t/d, p/b that have cognates in all four subgroups. Even if one language in a subgroup has

a clear cognate, I have taken the subgroup as represented. This criterion has naturally

limited the list to a manageable few. DEDR has a total of 5,558 numbered entries. The

number of entries with initial stops and the ones represented in all four subgroups are

as follows:

Initial segment DEDR entries Difference Entries in all subgroups

k/g 1075–2263 1189 37

c/j 2264–2900 637 18

t/d 2995–3567 573 20

p/b 3801–4614 814 44

Total 3213 119

These ﬁgures show that entries beginning with stops constitute 57.81 per cent of the

total of 5,558 entries in DEDR and the ones represented in all subgroups make up 3.7

per cent of this total. I have not considered initial

d (DEDR 2938–94) 57 entries since

none of them is represented in all four subgroups. Note that Proto-Dravidian had no

words beginning with apical (alveolar and retroﬂex) stops.

The second step was to set up three possible representations of stops initially and

give them numerical values, e.g. k- = 3, k-/g- = 2, g- = 1. I prepared a table with all

languages listed at the top according to subgroups from Tamil to Brahui, as follows:

Tamil, Malay¯a

lam, Ko

dagu, Toda, Ko

ta, Kanna

da and Tu

lu (seven languages: SD I);

Telugu, Gondi, K

da, Kui, Kuvi,

Pengo, Man

da (seven languages:

SD II); Kolami,

Naik

ri/Naiki, Parji, Ollari, Gadaba (ﬁve languages: Central Dravidian); Ku

rux, Malto,

Brahui (three languages: North Dravidian) = twenty-two languages. For each entry in

a row I marked under each language if the cognate has one of the possible reﬂexes of a

given segment, e.g. k-, or k-/g- or g-. In the vertical column at the end of each language,

it would then be possible to list the total number of entries that have different reﬂexes:

k- or k-/g- or g-.

The third step was to multiply each variant with a numerical value already assigned,

say, k 5 × 3 = 15, k/g 5 × 2 = 10, g 5 × 1 = 5; the aggregate value for ﬁfteen entries

is 15 + 10 + 5 = 30. Therefore, the average voicing index is this total value divided

by the number of entries available, i.e. 30/15 = 2. This procedure was repeated for each

language and subgroup; then, for the whole family, the score for a given segment could

be computed. After doing this for the four stops, an average index of the four indexes

could be arrived at by dividing the total index value by four, standing for the four stop

segments. That is the index for the whole family and is taken to represent the status of

the proto-stage.

The fourth step was interpreting the results. The ﬁnal score for all the entries for the

whole family should range between 3 and 1. Any ﬁnal score above 2.5 would suggest

that Proto-Dravidian had only voiceless stops to start with, and voicing has inﬁltrated

134 Phonolo

gy: historical

and comparative

into some languages, owing to internal changes and/or borrowing from other families,

like Indo-Aryan or Munda. If the ﬁnal score was 1.5 or below, it would suggest that

Proto-Dravidian had primarily voiced stops, and voiceless stops must have inﬁltrated

into the system through the same factors. If the score was between 1.51 and 2.49 the

result would be indecisive: it would show that Proto-Dravidian had both voiceless and

voiced stops, with a higher score suggesting more words with initial voiceless stops

and a lower score suggesting more words with voiced stops. Circumstances leading to

developing voicing secondarily are greater and more natural in linguistic change than

the reverse. It is also possible to work out scores for individual entries by language,

subgroup or family and discover the ones that have tilted the scores one way or the

other. Similarly, the average index by language would show which language and which

subgroup registers a lower or higher score in voicing phenomena. It would also be

possible to ﬁnd out historical explanations for such variation. For instance, Telugu has

cognates for 35 out of 37 entries with k-/g-; out of these there are 26 items with k-

(26 × 3 = 78), 7 items with k-/g- (7 × 2 = 14) and 2 items with g-(2× 1 = 2);

the total value for 35 cognates is 78 + 14 + 2 = 94; the average score is 94/35 =

2.685. The average scores for all four segments give the total voicing score for the whole

language.

Before we look at the results, certain clariﬁcations would be in order. I have recon-

structed the proto-phonemes for each entry under velar, palatal, dental and labial, only

to make sure that items belonging to one phoneme are not listed under another. There

are sound changes, which affected the word-initial phonemes in different languages,

e.g. palatalization of

∗

k (section 4.5.1.4) gives either c/s or j in one or more languages

which had this sound change, namely Tamil, Malay¯a

lam and Telugu. If the changed

phoneme is voiced, the proto-phoneme is also counted as voiced, although it is possible

for voicing to have developed after palatalization. Such cases are numerically so rare that

the score will not be affected by them. In South Dravidian I, PD

∗

c- became Ø in some

items, presumably through intermediate

∗

s and

∗

h; therefore, c-/Ø- alternation compar-

atively would suggest the segment as a voiceless one in a given language. Similarly, if

there is s/h/Ø (dialectally in

Gondi), it is taken to represent a voiceless phoneme for

the relevant item (section 4.5.1.3.2). Palatalized velars are not involved in loss. Another

such change is

∗

p- >h- in Early Medieval Kanna

da. DEDR mostly takes only the older

forms with p-, but where only h- occurs corresponding to p- in other languages, it is

also taken as a voiceless stop (section 4.5.1.1). Alternations k-/g-, t-/d-, etc. are set up

even if there is one item derived from the same root which shows a voiced stop, while

several others show a voiceless stop; the relative frequency of the alternation for a given

etymology is not taken into account. Sometimes the alternation could be between social

dialects. Not all circumstances of such alternations are retrievable from mere listings in

DEDR.

4.5 Historical

phonolo

gy: consonants

135

able 4.7a

oicing index (VI)

of word-initial

stops in South

Dravidian I

and II for

119 entries

Seg Ta Ma. To. Ko. Ko

d Ka. Tu. Te. Go Kon. Kui Kuvi Pe. Man

k/g 3.00 2.97 2.97 2.67 2.93 2.72 2.72 2.69 2.74 2.41 2.57 2.50 2.83 2.77

c/j 3.00 3.00 3.00 2.77 3.00 2.88 2.78 2.84 3.0 3.00 2.70 2.69 3.00 2.66

t/d 3.00 3.00 3.00 2.70 2.70 2.63 2.68 2.68 2.58 2.67 2.50 2.60 2.70 3.00

p/b 3.00 3.00 3.00 3.00 2.90 3.00 2.89 2.90 2.85 2.93 2.92 2.81 3.00 3.00

VI 3.00 2.99 2.99 2.78 2.88 2.81 2.77 2.77 2.83 2.75 2.67 2.65 2.88 2.86

Table 4.7b Voicing index of word-initial stops in Central Dravidian for

119 entries

Segment Kol. Nk. Pa. Oll. Gad. Subgroup

k/g 2.80 2.90 2.80 2.80 2.90 2.84

c/j 2.83 3.00 3.00 3.00 3.00 2.97

t/d 2.90 2.82 2.84 3.00 2.77 2.87

p/b 2.86 2.90 2.74 3.00 2.92 2.89

VI by lang. 2.85 2.90 2.85 2.95 2.89 2.89

Results Tables 4.7a-d give the voicing

index for each segment, language,

subgroup

and the whole family. VI = voicing index by language (average of the four rows under

each column).

From table 4.7a, the voicing index for South Dravidian I is 2.89 and for South Dravidian

II is 2.77. For the whole branch, the voicing index is 2.86. Tamil has the maximum score,

i.e. no voiced stops in word-initial position. The lowest score in South Dravidian I is that

of Tu

lu. In South Dravidian II, Pengo–Man

da show the highest score (fewer words with

initial voicing) and Kui–Kuvi the lowest (more words with initial voicing). On the whole

South Dravidian II has more words with word-initial voicing

than South Dravidian I.

In table 4.7b the voicing index is like that of South Dravidian I. Naiki and the Ollari–

Gadaba subgroup have fewer words with initial voiced stops.

In table 4.7c all members show a high score, suggesting more voiceless stops word-

initially, despite their exposure to Indo-Aryan for over two millennia.

Scores for each of the segments, by subgroup and family, are given in table 4.7d.

The score for the whole family is 2.855 or 2.86, which clearly proves that Proto-

Dravidian had primarily only voiceless stops and it is still the dominant feature in native

lexical items. We also notice that velar and dental series are more prone to voicing

than palatal and labial series. South Dravidian II has a lower score than all other sub-

groups suggesting a higher degree of voicing than other languages. This coincides with

136 Phonolo

gy: historical

and comparative

able 4.7c

oicing index of

word-initial

stops in North

Dravidian for 119 entries

Segment Ku

rux Malto Brahui Subgroup

k/g 2.90 2.90 2.70 2.83

c/j 3.00 2.85 3.00 2.95

t/d 2.75 2.82 2.78 2.78

p/b 2.87 2.82 3.00 2.90

VI 2.88 2.85 2.87 2.865

ble 4.7d

Voicing

index of word-initial stops

in the whole family

Segment SD I SD II CD ND Family

k/g 2.85 2.64 2.84 2.83 2.79

c/j 2.92 2.84 2.97 2.95 2.92

t/d 2.82 2.68 2.87 2.78 2.78

p/b 2.97 2.92 2.89 2.90 2.92

Final VI 2.89 2.77 2.89 2.87 2.855

Zvelebil’s study reported earlier. South Dravidian I still leads in voiceless stops, because

Tamil–Toda have only voiceless stops in word-initial position; Malay¯a

lam has internal

voicing sparingly.

Voicing can be accounted for through borrowing from Indo-Aryan or Munda, and also

through voicing assimilation within native words, which I explained earlier as extension

of the feature of voicing from non-initial segments to the initial segment, e.g. Ta. Ma.

tiruttu, Ka. tirdu, tiddu (<

∗

tir-t- <

∗

tir-utt-) ‘correct’, Ko

d. tidd-, Tu. tird-/tidd-, but Te.

diddu [3251]. I give below the lexical items that have only voiced stops in more than one

subgroup; these naturally exclude Tamil, which represents even initial voiced stops as

voiceless because of the limitation

of its orthography. In the whole data there are three

onomatopoetic items and two have decidedly voiced initials across subgroups, namely

∗

d V gu

d V ‘rumbling noise, thunder’ [1659],

∗

gurV gurV ‘snore, growl’ [1852],

∗

tV ca

tV ‘cracking sound’ [2796]. The last one has no voiced stop representation

in any language. All the three occur in Indo-Aryan also and CDIAL has Middle Indic

reconstructions:

(64) a.

∗

d V gu

d V onom. ‘gurgling or rattling sound’. SD I: Ta. Ma. ku

tuku

tu,

To. ku

dx- v.i. ‘stomach rumbles’, Ko gu

rg, gu

rgu

r in- (in- ‘say’), Ka.

dugu

du onom., gu

dugu n. ‘thunder, roar’, Tu. gu

dugu

du; SD II:

Te. gu

dagu

da onom. of drinking fast, Go. gu

rnj- ‘to thunder’, Ko

rugu

ru ‘running fast’, Kui

r¯u ‘thunder’, Kuvi g

nu- ‘to thunder’;

4.5 Historical

phonolo

gy: consonants

137

CD: Kol. gu

dm- ‘make noise’, Pa. gu

ri, gu

dr- ‘to thunder’; ND: Ku

rgu

rur- v.i. ‘make noise like thunder’||CDIAL 4180 cites IA etymolo-

gies and reconstructs for IA

∗

dugu

du- [1659].

∗

gurV gurV onom. ‘sound of snoring, growling’. SD I: Ta. kurukuruppu,

rukk-ai ‘snoring’, Ma. kurukurukka ‘breathe hard’, kurkku ‘a snore’,

Ka. gu

ruguru ‘sound of snoring’, gu/oruku, gorke ‘snoring’, Tu. guran`e

‘snarling of dog’, gurru kore- ‘to snore’; SD II: Te. gu

raka, gurru

‘snoring’, Go. gurr-‘to snore’, Ko

da g¯or- v.i ‘snore’, g¯or-u

n ‘snoring’,

Kui

dr¯oka (<

∗

gr¯o- <

∗

gor-V-) ‘snoring’, Kuvi

druk-, gurr- v.i. ‘snore’;

CD: Kol. gurga

dil- ‘dog growls’, Pa. gurr- ‘to hiss’, gurj- ‘to squeak’;

ND: Ku

r. gurr- ‘roar as tiger, shout’, Br. gh

urr- ‘to gro

’ ||

∗

guragur

CDIAL 4486, Skt. ghuraghur¯a-yate ‘it growls’ 4489 [1852].

At this point we do not know the historic source of such onomatopoetic expressions,

in the absence of comparative and etymological data from Munda languages. Since

onomatopoetic expressions tend to fall outside the normal phonological system of a

language (see Emeneau 1993: 83), we cannot generalize that voiced and voiceless stops

contrasted in Proto-Dravidian and the contrast is conﬁned only to onomatopoeia. We

can safely conclude from the quantitative study that Proto-Dravidian had only voiceless

stops and voicing originated in the descendant languages through internal changes and

borrowing. We have shown that inﬁltration of voicing in the whole family stands at .16

out of 3.00, i.e. .005 per cent. It is remarkable that the proto-feature is so persistent even

after about 3,000 years of breakup of Proto-Dravidian.

4.5.3 Initial consonants: nasals

4.5.3.1

∗

m- remains in all languages, except Brahui in which it splits into m- and b-; b-

when followed by front vowels, and m- elsewhere. This split is parallel to the split of

∗

into d- and n- (Krishnamurti 1969c).

Rule 16a. Nasal split in Brahui

m > b/#

−−−

[+V, −back] (Brahui)

(65) PD

∗

m¯e, m¯e-l ‘above, high, excellence’. SD I: Ta. m¯e, m¯el n. ‘height, high

place, above, sky’, Ma. m¯e, m¯en,Ko.m¯e,To.m¯el ‘up’, Ko

d. m¯e, m¯em-

adj, Ka. m¯e, m¯egu n., m¯em- adj., Tu. m¯el¨ı ‘upper part’; SD II: Te. m¯elu

‘high, excellent, good’, Go. m¯elta ‘good’; ND: Br. b¯e ‘up, over, on (verbal

preﬁx)’, b¯e-ha

rsing ‘to turn over’ [5086].

(66) PD

∗

m¯ey ‘to graze’. SD I: Ta. m¯ey ‘to graze’, Ma. m¯ey,Ko.m¯ey,To.m¯ıy ,

d. m¯ey, Ka. m¯e-, m¯eyu (past m¯edu), Tu. m¯eyuni v.i., m¯epuni v.t.; SD II:

138 Phonolo

gy: historical

and comparative

Te. m¯eyu v.i, m¯epu v.t., Go. m¯ey-, Ko

da m¯ey-, Kuvi m¯ey, m¯e-; CD: Kol.

Nk. m¯ı - v.i., m¯ı p- v.t. Pa. m¯ey- v.i., m¯ep- v.t.; ND: Ku

r. men- v.i, menta- v.t.,

Malt. min- v.i., Br. bei ‘grass ﬁt for grazing’ [5093].

Krishnamurti (1969c) examined fourteen etymologies to establish a split of n- into

d-/n- with one exception and m- > b-/m- also with one exception. The authors of DEDR

have accepted the above rules. For retention of

∗

m- in different

subgroups, see

∗

maram

‘tree’ (35).

4.5.3.2

∗

n- remains unchanged in all languages except Brahui before all vowels

(Krishnamurti 1961:

§ 1.24; 1969c). Old Tamil distinguished between dental and alve-

olar nasals; word-initially and before a dental stop, dental [n] occurred, in all other

non-initial positions (intervocalically, in gemination, before

∗

t, and ﬁnally) the alveolar

nasal [

n] occurred. However, according to Tolk¯appiyam there were few contrasts be-

tween the two in non-initial positions, e.g. verin ‘back’: var-i

n ‘if one comes’, both from

Cankam texts; there are text variants to the contrary in some cases (Shanmugam 1971b:

32–3, also fn. 11). Malay¯a

lam developed contrast secondarily in geminates,

∗

nt,

∗

[nd/n

d] >nn (dental),

∗

nn [nn] > nn, e.g. punna ‘mast-wood’ vs. kunnu ‘hill’ (<

∗

kuntu)

(see Subrahmanyam 1983: 380). It appears that the nasal+ alveolar sequence merged

with nasal +dental and both these developed into a geminate dental stop in Malay¯a

lam. In

any case there is no comparative evidence to set up two n-phonemes for Proto-Dravidian.

As in Tamil there could have been a phonetic difference between initial dental [n] and

non-initial alveolar [n

] in Proto-Dravidian also, because /n/ occurs word-initially in

Proto-Dravidian reconstructions unlike the alveolar phonemes /

t l r/. For this reason /n/

is put in the dental column in the chart (see table 4.2).

Rule 16b.

∗

n > d/#

−−−

[+V, −back] (Brahui)

(67) PD

∗

n¯ı r ‘water’. SD I: Ta. Ma. Ko To. n¯ı r,Ta.¯ı r ‘moisture’, Ma. ¯ırppam

‘dampness’, Ko

d. Tu. n¯ır¨ı, Ka. n¯ı r , n¯ıru; SD II: Te. n¯ıru, Go. n¯ı r,Ko

n¯ır masu ‘dew’, Kui n¯ıru ‘juice, sap, essence’, CD: Kol. Nk. ¯ı r , Pa. Oll.

Gad. n¯ı r; ND: Br. d¯ı r ‘water, ﬂood-water’ || >Skt. n¯ı r ‘water, juice, liquor’

[3690].

There are six certain etymologies, Br. ditar ‘blood’ <

∗

ney-tt-Vr[ DED 3798], Br. dar¯o

‘yesterday’ <

∗

ner-V- (Br. a- <

∗

e-) [3758], Br. d¯e ‘sun’ <

∗

n¯er- [3744], etc. There is a

much larger number

preserving

∗

n- before non-front vowels. The only exception to the

split rule is second sg pron. n¯ı in Brahui which could have been analogically restructured

4.5 Historical

phonolo

gy: consonants

139

on n- allomorphs of the paradigm, n-¯a gen sg, num- pl nom. and obl. (Krishnamurti

1969c). For other cases of retention of

∗

n- see ety. (4b), (25) and (43).

There is a sporadic loss of

∗

n- in various languages (Krishnamurti 1961: §§1.32,

1.218–20, pp. 17, 91–2), but its loss is more common in the Kolami–Naiki subgroup

(see (67)). It is almost a regular sound change in Naiki (Suvarchala 1992: 20).

4.5.3.3

∗

˜n-

Proto-Dravidian

∗

˜n occurred only word-initially mainly before

¯a and

¯e (short or long). It

is phonemically preserved in Old Tamil, Old and Modern Malay¯a

lam, Ko

dagu and Tu

lu.

Not all items that require the reconstruction of

∗

˜n are preserved in all these languages.

Tamil and Malay¯a

lam give evidence

of contrast between this and the

other nasals. This

phoneme has merged with n-(<

∗

n-) in almost all the languages (including Middle and

Modern Tamil). Some relic forms after merger are retained in Malay¯a

lam, Ko

dagu,

lu and Havyaka Kanna

da (Subrahmanyam 1983: 375–6). Burrow (1946: 603–16)

has discussed this problem in great detail, followed by Krishnamurti (1961:

§§1.25–32,

pp. 13–17), Emeneau (1970a), Zvelebil (1970b) and Subrahmanyam (1983). Burrow

giv

es 12 Tamil examples for

∗

˜n¯a-, of which in 5 cases Malay¯a

lam has ˜n- but Tamil has

n-, e.g. Ta. n¯a

ru ‘young plant’: Ma. ˜n¯aru; he cites 18 cases of Ta. ˜na- corresponding to

˜na-, ˜ne- in Malay¯a

lam and suggests that original a and e merged with e after ˜n in Old

Tamil. Then he gives examples of alternation of na-/ne- in Kanna

da and Telugu and

derives these from an original ˜n- which no language preserves. He contrasts these with

etymolo

gies pointing to unchanged

∗

na-or

∗

ne- (607–8). From Burrow’s treatment of

the problem we note three facts: (1) neutralization of contrast between

¯a and

¯e after

∗

˜n-;

(2) because of neutralization, the varying vowel qualities occurring after n- (as na-/ne-)

do not necessarily indicate the original qualities of vowels before such neutralization

occurred in Proto-Dravidian, if such a stage really existed; and (3) that the merger of

∗

˜n-

with

∗

n- started very early in Proto-South Dravidian and was accomplished in SD II by

the time it branched off. In South Dravidian I the merger was completed in Kanna

da;

while the change was still in progress, Tu

lu split ﬁrst, followed by Pre-Tamil from which

dagu and then Malay¯a

lam split later. Therefore, all these languages show the retention

∗

˜n- in different degrees. In Central Dravidian and North Dravidian, we do not know

when PD

∗

˜n- merged with the reﬂexes derived from

∗

n-. Where there is ˜n- in Ta. Ma.

and Ø- in the other languages, it is possible that

∗

˜n- is secondary in Ta. Ma. (<

∗

y-) (see

for examples Krishnamurti 1961:

§1.31, p. 17), e.g. Ta. ˜na

tu :Te.e

dri ‘crab’ (also Ø

in Ka. Go. and all Central Dravidian languages).

The merger is described by the following rule:

Rule 17. Palatal nasal reduction

∗

˜n >

∗

n/#

−−−

¯a

¯e)

140 Phonolo

gy: historical

and comparative

DEDR has 37 etymologies listed under ˜n- (2901–2937). A re-examination of these shows

that both Tamil and Malay¯a

lam provide evidence for ˜n- > n- and also variation between

¯a

and

¯e, during the historic period, Ta. ˜n¯a-, ˜ni-: Ma. n¯a-, ni- [2918, 2921, 2922], Ta. n¯a-:

Ma. ˜n¯a- [2906, 2909, 2911, 2914], Ta. ne- : Ma. ˜ne- [2925, 2928, 2929, 2933, 2934,

2935], Ta. na- : Ma. ˜na-/na- [2903, 2905], Ta. no- : Ma. ˜no- [2936]. Beside these there

are other correspondences: Ta. ˜na-/˜ne- : Ma. ˜na-/na- [2904], Ta. ˜n¯a- : Ma. ˜n¯a- [2908,

2910, 2912, 2913, 2915, 2919, 2920]; there is one case where Ta. ˜n¯a-: Ma. ˜n¯e- [2917].

Shanmugam (1971b: 37–8) gives illustrations for the merger of ˜n- with n- in Pre-Tamil.

When we are dealing with this problem, we need to consider the period (pre-ninth

century CE) when Tamil and Malay¯a

lam have to be taken as dialects of the same lan-

guage, rather than as two independent languages. In that case, it appears that the vocalic

variation

¯a and

¯e is shared by both of them as also the replacement of ˜n by n. It appears

in Proto-Dravidian

∗

˜n was probably followed by all the vowels, but during the historical

period we ﬁnd mainly

¯a,

¯e, i and

¯o, but not¯ı and

¯u; this distribution seems

phonologically

defective and unmotivated. It could be that

∗

˜n- was frozen midway of the change (after

its merger with

∗

n- was completed before the other vowels) in a few South Dravidian I

languages. What is difﬁcult to ascertain is if the vowel qualities were distinct in some

subgroups, mainly South Dravidian II and Central Dravidian, when

∗

˜n- coalesced with

n- in these languages.

Notice, Ta. Ma. ˜n¯a-: n¯a- in other languages in 2906, 2909, 2914, 2918, but Ta. Ma.

∗

˜n¯a-:n¯e-i

n other languages in 2908, 2910

–2913,

2915, 2919, 2920; Ta.

˜n¯a-,

Ma.

˜n¯e-

all others n¯a-/n¯e- (2907). In some South Dravidian II languages the vowel is represented

as ¯o- before retroﬂex consonants, suggesting a change of ¯e- > ¯o- through retraction.

(68) PD

∗

˜n¯a

n/˜n¯e

n ‘string, cord’. SD I: Ta. Ma. ˜n¯a

n,Ta.n¯a

n,To.n

n, Tu. Ka.

n¯e

nu;Te.n¯anu ‘a necklace’, Go. n¯o

ne,Kuin¯o

nu, Kuvi n¯o

no ‘rope’ [2908].

(69) PD

∗

˜n¯atu ‘to emit smell’ > SD I: Ta. ˜n¯aru, n¯aru, Ma. n¯aruka ‘to stink’,

Ko. n¯ar ‘to smell bad’, Ko

d. n¯ar,Tu.n¯ad-uni v.i., Ka. n¯a

ru; CD: Pa. ned-

(nett-) ‘to smell’ [2918].

(70) PD

∗

˜n¯entu ‘time, day’. SD I: Ta. ˜n¯anru, n¯anru, Ma. ˜n¯annu; SD II: Te. n

¯e

‘today’, Go. n¯en

d,Ko

da n¯en

ru, Kuvi n¯ecu, ninju,Pe.n¯enje

n, Man

n¯enj(e) ‘today’ [2920].

The question is, if (68) and (70) point to PD

∗

¯e and (69) to PD

∗

¯a, how does one

explain the different qualities of vowels in South Dravidian I and South Dravidian II?

There is one precious etymology that Burrow has ignored and the other writers have

also not considered. Malay¯a

lam has

∗

˜n¯an/˜nan- ‘I’ corresponding to n¯an/nan- in South

Dravidian I, n¯en/nan-, n¯a- beside n¯an/n¯a- in South Dravidian II. I have explained this

problem elsewhere (section 6.4.1.1). It shows that the domain of variation between low

and mid front vowels is mainly South Dravidian I and South Dravidian II. They are