Krishnamurti Bhadriraju. The Dravidian Languages

Подождите немного. Документ загружается.

4.5 Historical

phonolo

gy: consonants

151

sanduni: SD II: Te. can(u)- (ca

t-) ‘to go, to be ﬁt’, cellu ‘to pass, as

time, be current, to be suitable’, calupu v.t. ‘to pass time, to do’, Go. son-,

sond-, han-, hand- ‘to go’, Ko

da son-(soR-), solp- ‘to pass time’, Kui

sal-(sas-), Kuvi hal-(hacc-), Pe. Man

da hal-(hac-); CD: Nk. ca-(ca

d-),

Pa. cen-(cend-), Oll. sen-(sey-, sen

d-), Gad. cen-(cey-); ND: Ku

r. calr-

‘to continue, go on’ [2781].

Here the n-stem seems to have arisen from restructuring on the basis of the past stem.

It is not certain if it goes back to the Proto-Dravidian stage. Krishnamurti (TVB:

§1.96)

has also dealt with

∗

t alternation at a reconstructed stage, e.g. Te. m¯anu ‘to heal’: Ta.

m¯a

ru,Te.kan- ‘to bring forth young’, kandu ‘infant’: Ka. karu ‘young child’, kandu

‘calf’: CD: kar ‘sapling’; ND: Br. xan- ‘to give birth to’ [1411].

4.5.6.3

∗

n- /

∗

South Dravidian I preserves intervocalic -

n. South Dravidian II: Early Telugu inscriptions

(up to the ninth century CE) preserved the contrast between -

n- and -n-, e.g. pa

‘command, work’, ko

ni ‘having taken’ vs. ani ‘having said’, and the merger seems to

have been completed by the eleventh century CE (TVB:

§1.82). Although -

n merged with

-n, certain paradigms of verbs ending in -n(<

∗

n) show in sandhi a change of a dental

to a retroﬂex, e.g. ka

ti-ni ‘I saw’, but kan-i‘having seen’;

da–Kui–Kuvi–Pengo–

Man

da also preserve morphophonemic

n, e.g. u

n- (past u

t-). Ko

da has retained the

contrast between n and

n even at the lexical level. In Central Dravidian Ollari preserves

the contrast of n and

n, while the other members show merger. North Dravidian languages

have completely merged

n with n.

There is no evidence to reconstruct

∗

-˜n-/ -˜n for Proto-Dravidian.

4.5.7 Non-initial consonants: other sonorants

4.5.7.1 Liquids

∗

-r, -

∗

-l,

∗

4.5.7.1.1

∗

-r-/

∗

-r Alveolar and retroﬂex phonemes did not occur word-initially in

Proto-Dravidian. /r/ has remained unchanged in most of the languages. One exception

is Toda in which it has several reﬂexes r,

s, Ø, the conditioning factors of which are

not known. In several items V

r-V

contract to

, e.g. To. ¯u

n ‘pith’ : Ta. uram ‘heart

of tree’ [DED 558], To. t¯u

l-(t¯u

d-) ‘to roll up’ : Ta. curu

l ‘to roll up’ [DED 2211] etc.

In most such cases V

happens to be a non-lo

w vowel. In closed Pre-Toda syllables

The past-tense allomorph is -ti-(<

∗

-tt-i-) elsewhere, e.g. p¯o-ti-ni ‘I went’. If we take the under-

lying root as kan-, there is no phonetic motivation to descriptively derive t →

t/#kan+

−−−

.Ifwe

take the underlying -n as -

n, then the sandhi rule becomes normal. After alveolar -n ending stems

like an- ‘to say’, the tense sufﬁx became

∗

tt-i which later merged with

t-i, e.g.

∗

an-tti >

∗

an-tti >

∗

an-

ti > an-

ti. These two paradigms have structurally merged, obscuring the different historical

routes. Therefore, the surface solution is to say the t →

t operates after (C)Vn- roots to derive

the correct forms.

152 Phonolo

gy: historical

and comparative

-r- preceding a stop consonant was lost, To. ¨ıfoθ :Ta.iru-vatu ‘twenty’ [474], teg- ‘to

fasten a loin cloth’: Ta. ceruku [2778]. Before a non-apical obstruent (i.e. ptk),

∗

r >

e.g. To. a

sky ‘rice’ (<

∗

ar-kki):Ta.arici [215]], To. e

st ‘bull’ (<

∗

er-tt-): Ta. erutu,Ko.

et, Ka. ettu [815]. PSD

∗

r >

s in Toda following a front vowel and before a voiceless

non-apical obstruent, To. pe

sk ‘ﬂying fox’ (<

∗

ver-kk-):Ta.veruku ‘tom-cat’ [5490]. In

the majority of cases PD

∗

r remains in Toda, e.g. To. ¨ır ‘female buffalo’: Ta. erumai

[816], p

¨ır ‘quarrel’: Ta. p¯or [4540]. Some of the South Dravidian II languages have

[

r] corresponding to

∗

r. Brahui has -r/-rr; it is lost in inﬂection before a consonant,

e.g. marr- ‘to obey’ [4722], but baf-, bass- ‘to come’ (in inﬂection) from waru-w-,

war-cc-. In Kanna

da and Telugu root-ﬁnal -r gets assimilated to the following stop con-

sonant(s), voiceless

or voiced, e.g. Te.

wiccu ‘to blossom, to

open, burst

’, wiri ‘ﬂower’,

wirugu ‘to break’, Ka. biccu, bircu id. [5411] (for further details see TVB:

§ 1.101, p. 44,

§§1.173–85, pp. 74–9; Emeneau 1970a: 90–1, 1994: 49–70, Subrahmanyam 1983: 393–

403).

4.5.7.1.2

∗

z-/

∗

z The liquids r and

z are never

geminated in Proto-Dravidian or

older descendant languages.

The phoneme

∗

z, which was apparently pronounced as

a retroﬂex approximant (frictionless continuant), survives in some regional and social

dialects of Tamil and in Malay¯a

lam. In several modern dialects of Tamil and in the rest

of the languages it has merged with many other phonemes including Ø. Since no clear

isoglosses enclosing subgroups are discernible, it appears that the phoneme survived

until recent times in most of the languages. It was retained in the early records of literary

languages and a separate symbol was employed for it in early Telugu and Kanna

inscriptions. The reﬂexes are as follows: SD I: Mdn Ta. (social dialects)

l, y, Ma.

z,Ko.y, r , Ø,To.

s, w,

d, r , y,

l, Ø,Iru

la.

l,Ko

d. y,

l, Ø;Tu.

l (Brahmin dial.),

r (Common dial), OKa.

z, Middle and Mdn Ka.

l, r; SD II: Early Inscriptional Te.

Old and Later Te.

d, r, Go.–Ko

da–Kui–Kuvi–Pe.–Man

r; CD: Kol. Nk. r, Nk. (Ch.)

y, Ø, Pa., Oll. r, Gad.

d; ND: Ku

r.–Malt. Ø, r,

d;Br.r, rr, Ø. Some of the languages

have clearly deﬁnable conditions of split, e.g. Ka. -r[C,

l elsewhere, Telugu C]r-,

elsewhere, etc. The only subgroup that has a regular development is South Dravidian II

(other than Telugu) which has

r as the dominant reﬂex. In South Dravidian I, the widely

shared reﬂex is -

l (see Krishnamurti 1958b, TVB:

§1.124, 2001a: 42–75, Burrow 1968a,

Emeneau 1957b: 51–7, 1970a: 98–9).

(86) PD

∗

z-u ‘to plough, dig up’. SD I: Ta. u

zu, Ma. u

zu,Ko.ug-(u

rt-), To.

sf-(u

st-), Ko

d. ¯u

l-(upp-, utt-), Tu. ¯u

duni; SD II: Te. dunnu (<

∗

z-n-) v.t.,

dukki (<

∗

dukk- <

∗

z-kk-) n. ‘ploughing, tillage’, Go. u

r-,Ko

da–Kui–

Kuvi–Pe.–Man

r¯u-(

r¯u-t -); CD: Kol. Nk. ur-, Pa. u

r-, Gad. ¯u

d-; ND: uy-

(uss-) ‘to plough’, Malt. us- ‘to turn up as pigs do’ [688].

4.5 Historical

phonolo

gy: consonants

153

In South Dravidian II by metathesis and vowel contraction

∗

z comes to be in initial

position as in the above case

∗

z¯u- <

∗

z-u-. In Telugu the resultant initial

z- becomes

d-, which subsequently merged with d-(<

∗

t-), e.g.

dig(g)u/

digu > digu ‘to descend’

∗

z-g-) (see TVB: §§1.124–9, pp. 52–5). It appears that this Proto-Dravidian phoneme

has the least stability among liquids. The fact that it has merged with several existing

phonemes and is eliminated from the inventory of most of the languages hints at two

factors. First, it was a highly marked segment both in its phonetic properties and also in

its status

as a phoneme; second, there

was some kind of typological

pressure to eliminate

it from the normal system.

4.5.7.1.3

∗

-l-/

∗

-l In South Dravidian I Kota changes -li to - j ,

∗

puli ‘tiger’ > Ko. puj,

∗

eli ‘rat’ > Ko. eyj. Toda has -s for

∗

-l. In Tu

lu, -l- becomes -r- intervocalically in

many words irregularl

y, e.g.

∗

talay ‘head’ >Tu. tarε,but

∗

malay ‘forest’ > Tu. malε

(Kekunnaya 1994: 58–9). In Kui described by Winﬁeld

∗

-l>-

d (rather unusual!), but in

C] position -l; other dialects of Kui show -l. Brahui has two lateral phonemes, voiced

/l/ and voiceless / / (written as <l

h>in Bray 1909) corresponding to both

∗

-l and

∗

e.g.

∗

p¯al ‘milk’ > Br. p¯a



and

∗

t¯e

l ‘scorpion’ > Br. t¯e



; the conditions of split in Brahui

are not clear. The remaining languages represent

∗

l as an alveolar lateral. It has been

pointed out that

∗

l and

∗

r alternate

in Proto-Dravidian (see ety. (1),

(2), (4a), (13), (14),

(24) and (56)).

4.5.7.1.4

∗

l-/

∗

l All South Dravidian I languages preserve the contrast between alve-

olar and retroﬂex laterals /l

l/. Toda develops -

l into a voiceless retroﬂex lateral (in some

environments), Tu

lu merges -

l with -l in the North Common dialect, elsewhere it is

preserved. In South Dravidian II, in Telugu it merged with -l since the tenth century.

In Gondi, Ko

l >

r, l (in different dialects); in Kui -

l merges with -l (<

∗

-l) and

splits as C] l and elsewhere -

d ∼ -l (regionally); Kuvi–Pengo–Man

da have -

r. In Central

Dravidian Naik

ri preserves -

l after non-front vowels, but it has -l after the front vowels;

the rest of the languages have -l.InND

∗

l merges with

∗

-l in Ku

rux and Malto; in

Brahui

∗

l merged with

∗

-l which then split into -l and -



under conditions not yet clear.

(87) PD

∗

lan ‘open place, threshing ﬂoor’. SD I: Ta. ka

lam, ka

lan, Ma.

lam,Ko.ka

lm,To.ko



n (obl ko



lt-), Ko

d. Tu. ka

la, Ka. ka

la, ka

‘threshing ﬂoor, battle ﬂoor’; SD II: Te. kalanu, Go. kalam, ka

r¯a,Ko

kalam, ka

ran,Kuiklai, Kuvi k

r¯anu, Man

da k¯a

ra; CD: Kol. kalave,

r. ka

lave, Nk.(Ch.) kalay,Pa.kali, Oll. kalin; ND: Ku

r. xall- ‘ﬁeld’,

khal¯ı ‘threshing ﬂoor’, Malt. qalu ‘ﬁeld on the hills’ [1376].

In a typologically driven change

∗

n and

∗

l tended to merge with the alv

eolar

∗

n in

all languages outside the South Dravidian I group; this could be due to contact with

154 Phonolo

gy: historical

and comparative

Indo-Aryan languages which lack such contrast. But it must be noted that the merger is

not a shared innovation in these subgroups.

4.5.7.2 Semivowels

∗

4.5.7.2.1

∗

-w-/

∗

-w Intervocalic -w- is derived from root-ﬁnal

∗

-p followed by for-

mative sufﬁxes in most cases, e.g.

∗

cup-:

∗

cuw-ar (ety. 10). Other instances of -w- show

alternation with -m- and -k- in many languages (see TVB:

§1.111 a, b, c). There are

very few roots in Dravidian with an underlying -w, followed by formative sufﬁxes, re-

constructable to PD

∗

-w-. I consider

∗

kew-i ‘ear’ related to

∗

kep- ‘to say’ (see ety. (18)).

Several items that show -w- also seem to derive it from an older -k- (see ety. (27)),

∗

t¯e- ‘to

belch’ (<

∗

tew-V- < tek-V- [3451, 3453]).

n¯ıru, niw-uru ‘ashes’only Telugu

preserves

the -w- form. In To. and Br. - f - corresponds to -w- not necessarily of the root syllable.

In Middle Kanna

da -w- of the sufﬁx syllable became -b-, e.g. iru-war ‘two persons’

> ir-war > irbar > ibbaru.Br.bar-if- ‘to cause to come’, cf. Ta. varu-vi-.

4.5.7.2.2

∗

-y-/

∗

-y Both intervocalic and root-ﬁnal -y alternate in Proto-Dravidian

itself with -c-. However, there are still etymologies in which all the languages show only

-y-. The following is a good example of PD

∗

-y-:

(88) PD

∗

w¯ay/

∗

way-V ‘open space’. SD I: Ta. Ma. vayal,Ko

d. b¯el¨ı, Ka. bayal(u),

Tu. bayil¨ı; SD II: Te. bayalu (lw < Ka.), w¯ewili ‘ﬁeld’, Go. v¯avur, v¯aya,

Kuvi bayalu; CD: Kol. v¯ega

d, Nk. (Ch.) v¯ayur, v¯avur,Pa.v¯aya [5258].

There are more cases of -y- < -c-, e.g. PD

∗

pic-ar >

∗

piy-ar/pey-ar > p¯er ‘name’ (ety.

(31)), PD

∗

kac- ∼

∗

kay- ‘bitter’ (ety. (42)), PD

∗

pac- ∼

∗

pay- ‘yellow-green’ (ety. (80)).

There is a large number of etymologies requiring a ﬁnal

∗

-y,PD

∗

kay ‘hand’ (ety. (41)),

∗

key ‘to do’ (ety. (49)),

∗

k¯ay ‘to be hot’ (ety. (52)),

∗

ney ‘oil’ (ety. (43)). In most of

the languages the loss of -y produces compensatory lengthening of a preceding non-low

vowel,

∗

ciy-/c¯ı - ‘to give’ (ety. (45)).

4.5.7.2.3

∗

-H My recent researches showed that Proto-Dravidian had a laryngeal /H/,

which patterned with semivowels [−syll, −cons] (Krishnamurti 1997b). It survived

in a few lexical items in early Tamil as a restricted phoneme called ¯aytam [aydam]. It

occurred as a root-ﬁnal segment before a voiceless stop in the deictic forms such as a

h-tu

‘that’ (neut sg), i

h-tu ‘this’ (neut sg), pa

h-tu ‘ten’; -tu is a neut sg sufﬁx in these three

forms and the roots end in -

h. The properties of ¯aytam, as described by Tolk¯appiyam,

were: (i) it occurred after a short vowel and before a stop (voiceless), and its place

of articulation is like that of the stop (Subrahmanya Sastri 1934: 66). In other words

[

h] assimilates to the following voiceless stop. There is /pattu/ beside /pa

htu/, but the

forms /attu/ and /ittu/ are not attested. There are other constructions like Ta. ap-po

zutu

4.5 Historical

phonolo

gy: consonants

155

‘that time, then’, ip-po

zutu ‘this time, now’, Te. appu

du ‘then’, ippu

du ‘now’, attested

in early literary texts. In addition, all subgroups have ¯a and ¯ı as remote and proximate

demonstrative adjectives which also go back to PD

∗

¯a,

∗

¯ı ; they are free forms which

occur in noun phrases before other nouns bound syntactically as determiners, but not

morphologically. It appears that what is called ¯aytam in Tamil was an /h/ type of sound,

which lengthened the preceding vowel, when the output is a free form or assimilated to a

following voiceless stop in fused compounds; thirdly, it is lost if the following segment

is phoneticall

y voiced in the environment of

[V which was not an enunciati

ve vowel.

Thus we have three developments ¯a, ¯ı ; a

h-, i

h-; or a-, i-; the last in the pronouns a-tu

and i-tu also reconstructable to Proto-Dravidian. Similarly, pa

h- with -

h assimilated in

pat-tu has a variant p¯a-inOTa.o

n-p¯a-n ‘nine’ (−1, +10 = 9), but pat-in- (adjecti

val

form of pattu). Tolk¯appiyam had all four, iru-pa

htu ‘twenty’, e

zu-patu ‘seventy’, o

n-p¯an

‘nine’, pati-i

n-onru ‘eleven’. There are other lexical items (about half a dozen) in which

¯aytam occurred in Old Tamil

with cognates in the other languages

with length of the

preceding vowel or gemination of the following voiceless stop. Apparently the ¯aytam

was fast going out of use and it was preserved in some relic forms like the above.

The Old Tamil ¯aytam apparently was a reﬂex of a PD

∗

H (laryngeal) which seemed

to have had a much wider distribution in Proto-Dravidian but survived in a few frozen

forms in Early Tamil. As far as the three lexical items are concerned there is no doubt

that an h-like element must be responsible for the phonological processes. The remote

and proximate forms in Kuvi–Gondi in South Dravidian II and Ku

rux.–Malto in North

Dravidian have a h- freely varying with

zero in deictic forms, Go.

had/ad ‘that’, hid/id

‘this’, hav/av ‘those’, hiv/iv‘these’; Pengo and Kuvi also have h- forms in the demonstra-

tive bases. Ku

r. asan/hasan ‘there’, iyy¯a/hiyy¯a ‘in this place’, have the sporadic retention

of PD

∗

h. Medieval Kanna

da had aha

nge/h¯a

nge ‘that manner’, iha

nge/h¯ı

nge ‘this man-

ner’ in which the source of -h is not known. It could well be a surviving remnant of a

Proto-Dravidian laryngeal. It surfaced in written records after /h/ became phonemic in

Kanna

da after the sound change p- > h- which started in the tenth century and was com-

pleted by the fourteenth century.

The numeral ‘ten’ has aspirated variants in Kanna

and Telugu from early times, Te. ¯e

mbhadi ‘ﬁfty’ (eighth century; B. Radhakrishna 1971:

249), Ka. ombhattu ‘nine’, tombhattu ‘ninety’ (AD 869, SII xi, Par

t 1, No. 13). The as-

piration of the allomorphs of ‘ten’ continues into the modern languages, Te. mupphay

‘thirty’, nalabhay ‘forty’, ¯ebhay ‘ﬁfty’.

There are several words and grammatical forms showing the same phonological pro-

cesses reminiscent of the behaviour of a laryngeal, but the /

h/ would not surface, since

it was not phonemic in any of the early stages of the literary languages (except for Early

The /h/ in the forms cited for Kanna

da could not be traced to an older

∗

p-. The laryngeal

articulation could be recognized after /h/ started contrasting with /p/ in Medieval Kanna

da.

156 Phonolo

gy: historical

and comparative

Tamil for a brief period). Its effects are, however, seen in the neighbouring phonemes.

The attractive candidates for proposing a laryngeal are (i) the root of the number word

‘three’

∗

muH-/m¯u-; (ii) eight irregular verbs with aberrant phonology, namely

∗

caH- ‘to

die’,

∗

taH-r- ‘to bring’,

∗

waH-r- ‘to come’,

∗

aHn- ‘to say’,

∗

tiHn- ‘to eat’,

∗

n- ‘to eat,

drink’,

∗

weHn- ‘to hear’,

∗

kaH

n- ‘to see’. In all these, the free forms (imperatives, e.g.

w¯a, t¯a and derived nouns, t¯ı n, ¯u

n) have long vowels and the bound forms in inﬂection

have short (sometimes long) vowels, a result, it can be claimed, of the loss of the laryngeal

before a voiced consonant. The replacement of H by y later explains all the qualitative

changes in vowels; (iii) twenty-one verbs and derived nouns which are related by a quan-

titative change in the vowels, i.e. long vowels in free forms (nouns) and short vowels in

inﬂected verbs, e.g.

∗

tu ‘to perish’:

∗

k¯e

tu ‘evil’; (iv

) personal pronouns in which the

nominatives (free forms) have long vowels and the oblique stems (bound forms) have

short vowels,

∗

y¯an/

∗

yan- ‘I’, etc; (v) the negative morpheme in Proto-Dravidian

∗

-aHa-

on the basis of -¯a-, -ay-, -a-, -Ø-, -wa-,

V- in different subgroups (see for details

Krishnamurti 1997b:

§§4–5). Only a few examples involving the laryngeal are given here:

(89/48) PD

∗

cah- ‘to die’. SD I: Ta. c¯a-(c¯av-, cett- <

∗

ca-tt-) v.i., Ma. c¯a-(catt-),

with -y<

∗

-H,To.soy-(soyt-), Ko. c¯av ‘death’, Ko

d. c¯av¨ı/c¯a

l-(c¯av-, catt-)

‘to die’, c¯avu ‘corpse’, Ka. s¯ay-(satt-) v.i., s¯avu n., Tu. sai-pini, tai-pini

v.i., s¯avu, t¯avu n.; SD II: Te. ca-cc-(<caH-cc-) v.i., c¯awu n., Go. sai-, s¯ay-,

s¯a-, h¯a-, Ko

da, Kui s¯a-(s¯at-), Kuvi hai-(h¯at-), Pe. Man

da h¯a-(h¯at-); CD

∗

cay-/

∗

c¯ay:Pa.cay-(ca-˜n- <

∗

cay-nj-<

∗

cay-nd-), Oll. say-, Gad. cay-;

ND:

∗

caH- >

∗

ceH- >

∗

keH-: Ku

r. kh¯e-, k¯e-(kecc-), Malt. key-(kec-), Br.

kah-(kask-, neg. kas-) ‘to die’, kas-if- ‘to kill’ [2425].

Variations in vowel length and quality, the appearance of -y in many languages and

the preservation of -h in Brahui support the setting up of a laryngeal /H/ in the root in

Proto-Dravidian. Also see

∗

aHn- ‘to say’,

∗

waH-r- ‘to come’ (ety. (5), (72)). For more

examples see Krishnamurti (1997b).

For the negative morph at the Proto-Dravidian

stage, I reconstructed

∗

-aHaH which would develop to -¯ay/-ay (<

∗

-¯aH<

∗∗

-aH-aH); the

occurrence of -¯ay in Malay¯a

lam and -ay in Tu

lu and Parji are thus explained. Loss of

-y before a consonant is common in Dravidian. The resultant short -a- is lost in some

languages either after long vowel stems or in the unaccented position in verb conjugation,

e.g. Te. r¯a + a + ka →r¯a-ka. SCD (SD II) -wa (>-

a-) from -

∗

HaH is also explainable

by the rule

∗

H →w/

−−−

+a which is needed, anyway, to explain PD

∗

aw-antu ‘he’,

∗

aw-ar

‘they (persons)’, aw-ay ‘they (non-persons)’ and PSD I

∗

aw-a

l ‘she’.

There is now reasonable evidence to believe that South Dravidian I lost PD

∗

c- through

the intermediate stages of

∗

s- and h- and the missing phonetic links were not recorded

because the laryngeal

articulation ceased to operate in most of the literary languages

by the early CE. Some islands had remained which gave the evidence of h- in deictic

4.5 Historical

phonolo

gy: consonants

157

bases. But loanwords from Dravidian into Sanskrit and Prakrits show evidence of -s-

and -h-inka

t¯aha- ‘heifer’, kalaha- ‘strife, quarrel’, tulas¯ı- ‘basil plant’, sar¯ahaya-‘a

snake’ (the last one a folk etymology because of -haya ‘horse’ from Dravidian

∗

saraha-

∗

saras-) (see section 4.5.1.3, fn. 5 above).

4.5.7.3 Initial apicals through metathesis and vowel contraction

Non-nasal apical consonants (

∗

t [r],

∗

t [

d], l,

l, r,

z),

which did not occur word-initially

in Proto-Dravidian, shifted to the initial position in the languages of South-Central

Dravidian (South Dravidian II) by a phonological change. In radical syllables which

had V

R-V

-, R (R = non-nasal apical

consonant) comes to the initial

position by

this sound change. If the Proto-Dravidian base was (C

-V

- this sound change

has resulted in creating consonant clusters (with R as the second member) which were

also not allowed in Proto-Dravidian. A recent discussion of this sound change occurs

in Krishnamurti (1978a) within the framework of lexical diffusion. Earlier discussion

with bibliographical references occurs in Krishnamurti (1961: 51–7), Emeneau (1970a:

33–6), Zvelebil (1970b) and Subrahmanyam (1983:

§16, pp. 223–48).

Rule 20. Apical displacement (Proto-South Dravidian II)

a. V

R-V

- > R

-(V

= V

,orV

= [+low])



R-V

-(CC)- > RV

- Ø-CC- (V

= [+high] and V

= V

)

b. C

R-V

- > C

-(V

= V

,orV

= [+low])



.(C

R-V

-CC- > (C

)RV

- Ø-CC- (V

= [+high] and V

= V

)

Rules 20a, a



shift medial apicals to word-initial position with compensatory length-

ening of the root vowel (V

) in 20a. Rules 20b, b



create consonant clusters with an

apical R as the second member. Rules 20a, b have been treated in section 4.3.3 under

vowels. After V

R- metathesize to RV

-V

and V

contract to produce a long V

by Rules 20a, b. Here both the vowels are of the same quality, or V

is a low vowel -a,

while V

can be

∗

o or

∗

a. It may be recalled that Proto-Dravidian

∗

u merged

with

∗

o, respectively, before [

∗

C-a in Proto-South Dravidian, the undivided stage of

South Dravidian I and South Dravidian II (see section 4.4.2, Rule 4a, b). Rules 20a



suggest loss of V

and simple metathesis of V

RtoRV

. Here the root and sufﬁx vowels

-V

) are not of the same quality and V

is a high vowel /i u/; mid vowels /e o/ do not

occur as V

These changes were inherited at the undivided stage of Proto-South Dravidian II and

they continued into the modern languages even after successive breaks. Consequently, the

segments represented by C,R, V

and V

are different at different stages. In Telugu in the

case of Rules 20a, b, R is any apical consonant speciﬁed above. But for Rules 20a



Retroﬂex nasal [

n] also did not occur word-initially, but it is not involved in this sound change.

158 Phonolo

gy: historical

and comparative

hastobeoneof/ptkbdgmsw/andRis/t[r] r

z/; the retroﬂex stop and laterals /

l/ are excluded as R. For the remaining ﬁve languages of this subgroup Rules 20a, a



are

the same. But in the case of 20b, b



adds n to the consonants speciﬁed for Telugu, and

R includes all apicals including the laterals and

d. The change is generalized by enlarging

the environments. The consonant clusters formed by Rules 20 b, b



got simpliﬁed later

by the following rules:

Rule 21. Initial cluster simpliﬁcation

a. CR- > C-/#

−−−

(Telugu)

b. CR- > R-/#

−−−

(Gondi–Ko

da–Kui–Kuvi–Pengo–Man

da)

Gondi has evidence of the operation of Rules 20a and a



; it has indirect evidence of

Rules 20b and b



having operated in some dialects, because there are examples which

resulted

from the application of Rule

21b for which the input is 20b

and b



.Ko

da also

has examples like Gondi for Rules 20a and a



and 21b which presuppose that at one time

it also had 20b and b



. Rule 21b operates where C

is /n w s/.

Old and Middle Telugu has evidence of all the rules that are applicable. In Modern

Telugu, Gondi and Ko

da, the above rules have ceased to operate. But in the subgroup

Kui–Kuvi–Pengo–Man

da the sound changes, which apparently have had a history of

over 2,500 years, are still in progress. This is evident from the fact that free variation

between unchanged and changed forms is attested in the dialects of each of these (see

Krishnamurti 1978a:

2.5, p.10). The reﬂexes of non-nasal apicals after the operation

of the sound changes are given in Krishnamurti 1978a: table 10.1. Only Telugu data are

given below:

−−−

∗

t- [-r-] r- (>r-) r-

∗

t- [-

d-]

d- (>d-)

−−−

∗

r- r- r-

∗

z- (>

d- >d-)

z- (>r-)

∗

l- l-

−−−

∗

l- l-

−−−

Word-initially, Early Telugu kept the three-way contrast r:

d : r . In Middle and Modern

Telugu,

r- merged with r-(<

∗

r) and

d- with d-(<

∗

t-). In the postconsonantal position,

both

r and

z merged with r (<

∗

r). By Rule 21a, Cr- became C- with the loss of r-.

The following examples illustrate the rules with special reference to the developments

of medial apicals after they have shifted to the initial syllable:

(90) PD

∗

k¯o

z-/

∗

z-V- ‘young, tender, fresh’. SD I: Ta. ko

zuntu ‘tender, young’,

zantai ‘infant’, ku

zai ‘to sprout’ (<

∗

z-V-), Ma. ko

zunnu ‘tender shoot’,

To. kw¨ız, ‘twig’, Ka. ko

nasu ‘young one of wild beasts’, Tu. korε ‘weak,

4.5 Historical

phonolo

gy: consonants

159

small’, Koraga korayi, kori ‘husband’; SD II: Te. ko

duku ‘son’, k¯o

d-alu

‘daughter-in-law’, krotta, kro- ‘new’ (<

∗

z-tt), Mdn Te. kotta ‘new’, Go.

rs- ‘to sprout’, ko

ri-¯a

r ‘daughter-in law, younger brother’s wife’, Ko

ro ‘child’, ko

rya, ko

resi ‘daughter-in-law’, Kui k¯o

ru ‘new shoot’, ko

rgi

‘fresh, new’, ku

ra, k

rua, k

ruha ‘husband’, Kuvi ku

ria ‘daughter-in-law’,

r¯ogi ‘young, immature’, Pe. k

rogi ‘fresh’, ko

riy g¯a

r ‘son’s wife’, Man

riya g¯a

r id.; CD: Kol. koral, Nk. kora

l ‘daughter-in-law, bride’, Pa.

r ‘very young’, ko

rc- ‘to sprout’, ko

rol ‘bride’, Oll. ko

ral ‘son’s wife’,

Gad. ko

dus-, ko

dc- ‘to sprout’; ND: x¯or- ‘to shoot out new leaves’, korr¯a

‘fresh’, Malt. q¯oro ‘infant’, Br. xarr- ‘to sprout’, xarruni ‘greenness, wife’

[2149].

In South Dravidian I Kanna

da has no cognate with -

l but -

l/-

n alternate in Old Kanna

da.

lu, Koraga have -r for -

z as expected. In South Dravidian II

∗

z is represented by -

whether in intervocalic or postconsonantal position. North Dravidian has -r for

∗

(91) PD

∗

¯ı r /ir-V

‘tw

’. S

DI:Ta.

tu ‘tw

’(>Middle

and Mdn Ta.

tu),

¯ı r -/iru- adj ra

tu; ¯ı r -/iru adj, Ko. ir- adj, To. ¯e

d ‘two’, ¨ı-, ¯ı r - adj, Ir. ra

du,

du,Ko

d. da

d¨ı ‘two’, ¯ı r -/iru- adj, Ka. era

du ‘two’, iru- adj: SD II: Te.

du ‘two’,¯ı r /iru- adj, Go. ran

d ‘two’, rahk rahk ‘two each’, ir- adj, Ko

di ‘two’, ri- adj, ri

er ‘two men’, rine

d ‘two days’, Kui r¯ı n

de ‘two’, r¯ı -

adj, Kuvi ri

di ‘two’, r¯ıari ‘two men’, Pe. rin

dek ‘two women’, rin

‘two

things’, rikar ‘two men’, Man

da ri ‘two’, rikar ‘two men’; CD: Kol. indi

‘two things’, ¯ı r a l ‘two women’, Nk. (Ch.) ern

di ‘two things’, ir- adj, Pa.

du ‘two things’, Oll. in

di, Gad. i

dig ‘two’; ND: Ku

r. ¯e

d ‘two’, irb- ‘two

persons’, Malt. iwr (<

∗

irw-) ‘two persons’, Br. ira

t ‘two’, ir¯a adj [474].

In South Dravidian I, we notice a recent trend of short vowel loss before alveolar r-

(aphaeresis) in a few lexical items: Tamil, Malay¯a

lam, Kanna

da, Ko

dagu and Iru

la show

such forms. Te. re

du was perhaps from an older unattested

∗

r¯e

du (<

∗

er-a

d-) with

shortening of r¯e-tore- before a consonant cluster. The other South Dravidian II (SD II)

languages had

∗

r¯ı n

d- from

∗

ir-i-n

d /

∗

ir-u-

d with shortening of¯ı to i before a consonant

cluster in some of the languages. In the case of Ko

da ru

di (<

∗

di), the front vowel

is retracted because of the following retroﬂex consonant-cluster. In North Dravidian

rux ¯e is a little puzzling, since it does not derive from Proto-South Dravidian

∗

era

d-.

Note that this etymology has interesting semantic extensions reﬂected in the whole family,

namely to sprout > (a) fresh, tender, young > (b1) new, immature > (b2) child, bride (kinship

terms: wife, husband, son, daughter, daughter-in-law, brother’s wife, etc.). Central India is rich

in the shift to kinship terms. Telugu here has both -

d- and C]r-, i.e. ko

duku ‘son’, k¯o

dalu

‘daughter-in-law’, krotta ‘new’ with simpliﬁcation of kr-tok- in Modern Telugu.

160 Phonolo

gy: historical

and comparative

It must be an independent and sporadic vowel change. All languages preserve the root

syllable ¯ı r /ir-V-. This etymology supports the assumption that metathesis took place at

the undivided stage of these languages. Also note ety. (32), (33), (34), (35) and (38),

which provide evidence of the operation of the above rules. In the whole subgroup

(minus Telugu) the sound change included, among the initial consonants, the nasal n-

and among the members of R, the retroﬂex stop

∗

d and the lateral phonemes

∗

l., e.g.

(92) PD

∗

t-ac- ‘male of a domestic animal’. SD I: Ta. ka

t¯a, ka

tavu, ki

t¯a ‘male

of sheep, goat, bull’, Ma. ka

t¯a, ki

t¯a, ka

tacci,Ko.ka

rc,Ko

d. ka

d¨ıci, Ka.

dasu,Tu.ga

dasu; SD II: Go. k¯a

r¯a,Ko

da (dial.) g

r¯alu,Kuigr¯a

du,

dr¯a

du, k

rai ‘young female buffalo’, Kuvi

d¯alu; ND: Ku

r. ka

r¯a ‘young

male buffalo’, ka

r¯ı ‘young female buffalo’ (the ﬁnal vowels are gender

markers), Br. xar¯as ‘bull’ [1123]; also section 4.5.1.3, fn. 5.

Note the sporadic change of g- >

d- in Kui. Also notice the loss of the initial member of

the cluster in Kuvi. Similarly

∗

caracc-/

∗

carancc- ‘snake’ >Te. tr

acu, t¯acu ‘cobra’, Go.

taras, taranj,Ko

da saras, sr¯asu,Kuisr¯asu, sr¯acu, Kuvi r¯acu,Pe.r¯ac, Man

da trehe

[2359].

(93) PSD II

∗

nela-nj- >

∗

nl¯enj- >l¯enj- ‘moon’ : Go. nelenj-, lenj-, Kui

d¯anju

(¯a <

∗

¯e), Kuvi l¯enju, Pe. Man

da l¯enj-; SD I

∗

nel-aH- <

∗

nel-ac-: Ta. Ma.

nil-¯a, nil-avu,Ko

d. nelaci;CD

∗

nel-i

-: Pa. neli ˜n (pl. nel ˜nil), Oll. neli

Gad. nel¯a (pl. nel

¯ı l ) [3754].

(94) PD

∗

tit-V- ‘to open’. SD I: Ta. tira (tirant-) ‘to open (as a door, one’s eyes)’

Ma. tu

rakka,Ko.terv-(terd-), To. ter-(terθ-), Ko

d. tora (torand-), Tu.

jappuni ‘to open’, terapu ‘space’; SD II: Te. te

racu ‘to open’, Go. tar¯ı-,

terr-, te

r-, Ko

da re- (dial. tere-) ‘to open’, Kui d¯apa (d¯at-; ¯a <

∗

¯e), Kuvi

de-, Pe. Man

da j¯e-(<

∗

r¯e-); ND: Ku

r. tisg- ‘to open’, Malt. tisg- ‘to lift the

latch’ [3259].

In a quantitative study (Krishnamurti 1978a: 9), it was shown that only 16 per cent of the

eligible lexical items were affected by the above changes in Gondi, 21 per cent in Ko

da,

but 72 per cent in Kui, 61 per cent in Kuvi, 63 per cent in Pengo and 65 per cent in

Man

da. The sound changes spread through the mechanism of lexical diffusion. Telugu

was not included in the study because it has a much larger lexicon than the other non-

literary languages and would not have a parity for comparison. The sound changes have

ceased to operate in two of the seven languages, namely Telugu and Gondi. But in Ko

dialectally consonant clusters are formed, e.g. maran ‘tree’ (Araku dialect): mr¯anu (Sova

dialect). In the remaining four languages the sound changes are still ongoing.

There has been discussion of the phonetic mechanism underlying apical displacement.

It is crucial to note that this sound change occurs in polysyllabic stems and not in