Krishnamurti Bhadriraju. The Dravidian Languages

Подождите немного. Документ загружается.

4.5 Historical

phonolo

gy: consonants

121

sound change had spread to the entire Kanna

da-speaking area (Gai 1946: 13–14). In

Pampabh¯arata, the ﬁrst literary work of mid tenth century, the sound change was not

attested. In uneducated Kanna

da h- started getting lost as early as the beginning of

the thirteenth century and this distinguishes standard from non-standard in Modern

Kanna

da: words involving this change occur in neighbouring languages Tu

lu, Ba

daga,

different Ku

rumba dialects,

Alu Kurumba, B¨e

t¨ıKurumba, J¯enu Kurumba and Sh¯olega.

Examples: OKa. p¯alu > Middle and Mdn. Ka. h¯alu ‘milk’, OKa. p¯u(vu) ‘ﬂower’ >

h¯uvu, OKa. putta ‘anthill’ > hutta (see etymologies (13), (16), (17)).

Alu Ku

rumba has

many forms with Ø corresponding to h- of Mdn Kanna

da, e.g

AKu. ¯og ‘to go’ (< Ka.

h¯ogu < p¯ogu), J¯enu Ku

rumba has both zero and h- forms, e.g. ¨e

nu ‘female, woman’

(< Mdn Ka. e

n < he

nu), h¨e

di ‘pig’

(< OKa. pandi, handi). Some

rumba dialects

also show the h- forms and p- forms beside Ø- forms. It is not possible to conclude

anything since clear data are lacking on word history. Kota and Toda have hardly any

direct borrowings from Kanna

da involving this change; Kota has h¯a

l ‘ruin’ (< Ba

daga)

from Ka. p¯a

z, h¯a

l, ¯a

l ‘ruin’ [4110]. Ba

daga has variants with h-/Ø-, e.g. he

nu/e

(Pilot-Raichoor 1997: 137). Ba

daga also has many more native items with p-(p¯ogu

‘to go’, pui/poi ‘false’, etc.) which make us doubt

if they really migrated to

the Nilgiris

in the sixteenth century a

s proposed (Hockings and Pilot-Raichoor

1992, Foreword

by Emeneau:

viii

–ix), and

not much earlier, i.e. after

the sound change

∗

p > h-began

(tenth century CE) but had not run its full course (fourteenth century CE). Emeneau

also mentions the ‘disquieting statement of Francis’ that Ba

daga has two forms of r

(r and

r), of course not

in the dialects covered by the

authors (ix).

4.5.1.2

∗

t- remains t- (or d-) in all the languages (see ety. (19), (27), (28)); one more basic

item:

(44) PD

∗

talay ‘head, top, hair’. SD I: Ta. talai, Ma. tala,Ko.tal,To.tal ‘head’,

tas

m ‘top’, Ko

d. tale ‘end’, Ka. tale, tala,Tu.tarε; SD II: Te. tala, Go.

tal, tala, tal¯a, talla,Ko

da tala,Kuitlau (pl tl¯aka), Kuvi tr¯ayu (pl tr¯aka);

CD: Kol. Nk. tal,Pa.tel, Oll. tal, Gad. tallu; ND: Malt. tali ‘hair of head’

[3103].

Also compare

∗

tiyam ‘sweet’ (7),

∗

tew-i- ‘to be sated’ (27),

∗

tokal ‘skin’ (28).

4.5.1.3

∗

c- remains in a large number of etymologies in all the languages. It has, however,

undergone certain irregular (incomplete) sound changes,

∗

c- >Ø-/t-/k-. In all South

This form makes me doubt if it is inherited. Ka. nd and not

d corresponds to PD

∗

nd. Does this

indicate that JKu. borrowed this form after h- < p- and before nd- <

∗

nd, or is it only a problem

of recording (Zvelebil 1988)?

122 Phonolo

gy: historical

and comparative

Dravidian I languages it is lost in a number of items, perhaps through two intermediate

stages such as

∗

c- >

∗

s- > h- > Ø-. Since h- was not phonemic in South Dravidian

I it was not recorded. This sound change was shared by South Dravidian I and South

Dravidian II and it is still an ongoing change in south and southeastern Gondi dialects.

The change apparently spread from Telugu to Gondi through Koya and Ma

ria dialects

in which it is a completed regular sound change with Ø corresponding to s and h of the

other dialects. It ceased to operate in the northern and western dialects after a few items

had passed

through the

ﬁrst phase

of change (see Krishnamurti

1998b for a comprehen-

sive treatment of this sound

change in Gondi dialects). This

change, insofar as South

Dravidian I is concerned

,was

ﬁrst treated in great detail

by Meile (1943

–5) and Burrow

(1947), independently. The changes affecting PD

∗

c- were

examined later by Krishna-

murti (1961:

§§1.15–21), Zvelebil (1970b: 109–15) and Subrahmanyam (1983: 317–34).

The most recent comprehensive treatment of the developments of

∗

c occur in Emeneau

1988 (repr. 1994: 307–85). The irregular changes

may be formulated as follows:

Rule 13. Affricate weakening, loss, irregular merger with dentals and velars

∗

c > (

∗

s >

∗

h) > Ø/ #

−−−

... (South

Dravidian I; Telugu)

The number of items subject to Rule 13a is 14 per cent of the total number of items

(500) that require PD

∗

c-. The rule implies that c > Ø passed through successive stages

of weakening in South Dravidian I also although there is no direct attestation of these

stages since /h/ was not phonemic (Krishnamurti 1961:

§1.53, p. 23, 1998b: 69). The

representation of

∗

c as [s-] in some of the southern languages is well attested, e.g. spoken

Tamil, Kota, Kanna

da, Tu

lu, non-standard Telugu and all languages of South-Central

Dravidian (SD II), Kolami, Naiki and Gadaba of Central Dravidian. Emeneau is inclined

to delink the South Dravidian I change from the scope of

∗

s- > h->Ø- in Gondi dialects

and says that it was a one-step change in South Dravidian I, i.e.

∗

c>Ø (Emeneau 1994:

§16, p. 355). It is difﬁcult to visualize phonetically a jump from

∗

c-toØ-, that too in the

initial position of a root syllable. The sharing of this sound change is one of the crucial

arguments in favour of South Dravidian I and South Dravidian II being sub-branches of

Proto-South Dravidian. At about the same time (second century BCE) Sinhala changed

s to h in non-ﬁnal positions, perhaps prompted by a similar change in the Tamil area

(Masica 1991: 205–7).

In 13 groups all SD I languages and Telugu have lost

∗

c-; the

Early Tamil attests the loss of Sanskrit and Prakrit sibilants ´s,

s, s in loanwords (Burrow 1943:

132–5) more than any other southern language. It is reasonable to assume that, even in loanwords,

s ﬁrst became h as it happened in Sinhala before it became Ø. There is crucial evidence in Tamil

lexical phonology which supports this assumption. Only Tamil has -¯a as V

and no other member

of SD I. In several lexical items where Ta. has -¯a, the other languages point to an older -acV-

[-asu] which, we can suppose, became -ah(V) that resulted in -¯a through contraction. Examine:

(1) PD

∗

car-ac- [sar-asu] > SD I: Ta. ar-¯a ‘snake’ (also ara, aravu), Ma. aravu;SD

II: Te. tr

acu (<

∗

tar-a-ncc-) ‘cobra’, Go. taras, taranj,Ko

da saras, sr¯asu,Kui

4.5 Historical

phonolo

gy: consonants

123

remaining 57 cases have irregular loss beside c- in one or more languages. It is possible

that items which had developed s- variants in social dialects were the ones which suffered

change but it is impossible to retrieve the sociolinguistic aspects involved in the change.

(45)

∗

ciy-/

∗

c¯ı - ‘to give’. South Dravidian I: Ta.¯ı (-v-, -nt-) ‘to give to inferiors’,

¯ıvi ‘gift’, Ka. ¯ı -(itt),Te. icc-(¯ı -, ¯ı y -/iyy-, iww-), ¯ıvi ‘gift’: South Dravidian

II: Go. s¯ı -, h¯ı -, ¯ı -, Ko

da s¯ı -(sit-), Kui s¯ıva (s¯ı t-), (K.) h¯ı -, Kuvi h¯ı-(h¯ı t-),

Pe. h¯ı-(h¯ı t-), s¯ı - (in songs), Man

da h¯ı; CD: Kol. Nk. s¯ı -, Pa. c¯ı - (past ci ˜n-

∗

ciy-nd-), Oll. Gad. s¯ı-(s¯ı d-); ND: Ku

r. ci’ (cicc-), Malt. ciy-(cic-) [2598].

Also see

∗

cup:

∗

cuw-ar ‘salt’ (10),

∗

col-ay ‘ﬁreplace’ (24).

Emeneau considers this sound change as a possible case of lexical diffusion, which

failed to cover all eligible lexical items before it ceased to operate (1994:

§§12–14) (see

table 4.4).

sr¯asu, sr¯acu, Kuvi r¯acu,Pe.r¯ac, Man

da trehe. Pkt. DNM sar¯ahaya- [2359]. The

Proto-Dravidian reconstruction would be

∗

car-a-ncc- >

∗

car-a-cc- >

∗

car-a-c- [car-

as-V]. -as > -ah >-¯a to account for the long -¯a in Ta. The -h element is reﬂected in

De´s¯ın¯amam¯al¯a’s borrowing and also in Man

da trehe. Note that Tamil adds a further

formative sufﬁx -vu whenever it ends in a long vowel in V

position and the long

vowel in the unaccented position gets shortened.

(2) PD

∗

kal-ac- ‘to quarrel’ > SD I: Ta. kal-¯a-vu ‘to be angry’, kal-¯a-y ‘to quarrel’,

kal-¯a-m ‘war, battle’ (<

∗

kal-ah- <

∗

kal-ac-), Ma. kalacuka ‘be disturbed’, kalacal

‘quarrel’; SD II: Kui glahpa (glah-t-) ‘to confuse’ ||>Skt. kal-aha- ‘quarrel, ﬁght’

[1303; also see Burrow 1948: 371]. Cognates occur in all subgroups, but only the

diagnostic ones are given; the occurrence of -s form in Ma., h- form in Kui, and -h

in Skt. borrowing point to -as > -ah >-¯a.

(3) PD

∗

kan-acV ‘dream, to dream’ > SD I: Ta. kan-¯a (<

∗

kan-ah- <

∗

kan-ac-), kan-avu

‘dream’, v.i. ‘to dream’, Ma. kan-¯avu, kin-¯avu, kan-avu n., kan-avu v.i., Ko. kancn

‘dream’, Ko

d. kenaci, Ka. kanasu, kanasa n.; SD II: Go. kansk, kansk-, kanjk- v.i.

[1407].

(4) PD

∗

l-acV ‘sacred basil’ > SD I: Ta. tu

z-¯ay, tu

l-aci, tu

l-avu, Ma. tu

l-asi,Ko

d. to

asi, Ka. to

l-aci, to

l-asi,Tu.tu

l-asi, tul-asi; SD II: Te. tul-asi; CD: Pa. tul-ca ||>Skt.

tulas¯ı; some of the Modern Dravidian forms could have been reborrowed from Skt.

[3357].

(5) PD

∗

nel-a-nc/

∗

nel-a-ncc-‘moon, moon-light’ > SD I: Ta. nil-¯a (<

∗

nel-ah- <

∗

nel-

ac-<

∗

nel-acc-), nil-avu, Ma. nil-¯a,Ko

d. nel-aci; SD II: Go. nelanj, Kui, Kuvi, Pe.

Man

da l¯enj-(<

∗

nl¯enj < nelanj-), Kui

d¯anju (<

∗

l¯anj- <

∗

l¯enj-); CD: Pa. neli ˜n [nelinj-],

Gad. neli

. This compares well with

∗

car-a-ncc- in (1) above [3754].

(6)

∗

pal-acV/

∗

pan-acV ‘jack fruit tree’ > SD I: Ta. pal-¯a, pil-¯a, pal-avu, Ma. pal-¯avu,

pl¯avu,Ko

d. palaci, Ka. panasa, palasa, palasu; SD II: Te. panasa, Kuvi panha,

nha; CD: Pa. penac, Gad. panis ||>lw. Skt. panasa-, palasa-, phanasa-, phalasa-

[3988].

(7)

∗

t-aca- [ka

t-asV-] ‘male of cattle, heifer’ > SD I: Ta. ka

t-¯a, ka

t-¯ay, ka

t-avu, ki

¯a, ki

t-¯ay, Ma. ka

t-¯a, ki

t-¯a(vu), ka

t-acci ‘young cow’, Ko. ka

rc n¯ag/kurl ‘calf of

buffalo/cow’, Ko

d. ka

d-¨ıci,Tu.ga

d-asu, Ka. ka

d-asu; SD II: Go. k¯a

r-¯a,Ko

nda

r¯alu,Kuigr¯a

du, k

rai, Kuvi

d¯alu; ND: Ku

r. ka

r-¯a ‘young male buffalo’, ka

r-¯ı

124 Phonolo

gy: historical

and comparative

able 4.4.

Number

of lexical

items showing

c-/ Ø-

alternation

DEDR entry nos. c- Ø-

2341, 2391, 2410, 2485, SD I, SD II: Te.

2552, 2559, 2598, 2617,

2674, 2776, 2798, 2826,

2857 (13)

2335 + 157, SD II: other than Te., SD I, SD II: Te.

2342 + 162 + 271 CD

2698 (3)

2684 + 664 (1) SD I SD I, SD II: Te.

(28) one or more of the and Ø, or c or Ø

languages with

doublets in c-

(25) Sporadic loss or

retention of c-

Rule 13b.

∗

c- > t- (Toda – regular)

(Other languages – irregular)

This is a fairly regular sound change treated in detail by Emeneau (1957b; 1988 repr.

1994:

§§18–23). In the latter article, Emeneau has added more examples of the operation

of Rule 13b and explained the exceptions. The irregular development of t- from

∗

c-was

‘young female buffalo’ (ﬁnal vowels are gender markers borrowed from IA), Br.

xar¯as ‘bull’ || > Skt. ka

t¯aha- ‘young female buffalo’ [1123; also Burrow 1948: 368].

Note the occurrence of -cV/-sV in SD and Brahui and Sanskrit loanword with -ha.

In all these cases and similar ones, Old Tamil formative -¯a which is unique can be interpreted to

have developed from the contraction of an older

∗

-ah <

∗

-as. These examples provide the missing

phonetic links otherwise shrouded in history, because -h was subphonemic. The laryngeal

∗

had already gone out of the Tamil speech except in some relic forms with restricted distribution.

All this evidence supports the view that the change s > h > Ø occurring in South Dravidian I

had spread to Gondi dialects in South Dravidian II from Telugu. The sound change is still running

its course dialectally in South Dravidian II. Also notice that s > h occurs dialectally in the other

members of South Dravidian II also, namely Kui, Kuvi, Pengo and Man

da. Burrow (1947: 133)

cited items (4) and (6) as instances of Sanskrit words losing -s > -Ø in early Tamil. In DEDR

these are included as native groups indicating Sanskrit as the borrower with cross references to

CDIAL entries. Further Burrow mentioned categorically the loss of Sanskrit sibilants, but only a

few items with Sanskrit affricate c- (p. 134). The examples given by Burrow showed the loss of

sibilants irrespective of their position in a word. Another point which receives indirect support

from these examples is the possible merger of the Proto-Dravidian laryngeal

∗

H with

∗

h <

∗

leading to parallel phonetic developments (see Krishnamurti 1997b).

4.5 Historical

phonolo

gy: consonants

125

able 4.5.

Langua

ges sho

wing

c-/t- alternation

c- t-

SD I: Ta. Ma. Ko. Ka. Tu. Ta. To. Ko. Ko

d. Ka.Tu.

SD II: Te.Go, Ko

da, Kui, Kuvi, Te. Kui

Man

CD: Kol. Nk. Pa. Gad Kol. Nk. Pa. Gad.

ND: Ku

r. Br.

Table 4.6. Languages showing c-/t-/Ø- alternation

c- t- Ø-

SD I: Ta. Ma. Ko. Ka. Ta. Ma. Ko. To. Ko

d. Ka. Ta. Ma. Ko. Ir.

AKu.,

d. Ko

d. Ka.Tu.

SD II: Te. Go. Ko

da, Kui, Te. Go.Ko

da, Kui, Te. Go.

Kuvi, Man

da Man

CD: Kol. Gad. Gad. Pa.

ND: Ku

r. Malt. – –

mentioned earlier by Burrow (1947: 142) and Krishnamurti (1961: 12–13, 89–90). Eme-

neau has surveyed all the relevant lexical items and illustrated many examples for the

irregular change of

∗

c- > t- in man

y Dravidian languages which defy

generalization

(

§§21–3). The oppositions shown in table 4.5 are noticed, based on Emeneau’s data and

discussion. Tables 4.5 and 4.6 summarize Emeneau’s analysis of irregular reﬂexes in-

volving zero and t -, beside retention of PD

∗

c- in different subgroups. The gaps in the

tables like the absence of t - forms in North Dravidian seem to be signiﬁcant since a

whole subgroup does not show a dental reﬂex for PD

∗

c-. It is clear that Ø is conﬁned to

South Dravidian I spreading into Telugu and more recently Gondi of South Dravidian II.

Krishnamurti (1961:

§1.90, p. 40) proposed that

∗

c- > t- would have preceded in point

of time

∗

c-(s- > h-) >Ø-, since an occlusive t- was more likely to arise from the affricate

in Proto-Dravidian than when it became a sibilant at least in a part of the lexicon that

was affected by other changes leading to its loss.

Rule 13c.

∗

c > k/#

−−−

[+syllabic, −low] (North Dravidian)

Emeneau (1961b: 371–88; repr. 1994:1–15) proposed a rule that PD

∗

c- becomes a

velar voiceless stop k- in North Dravidian before u, ¯u and e, ¯e; he gave four cases of k [u,

¯u and four for k [e, ¯e. Actually the instances involving the mid vowels were not as clear

as the ones with high back vowels. In 1988 (repr. 1994) he noticed that besides North

126 Phonolo

gy: historical

and comparative

Dravidian several other Dravidian languages also showed k- sporadically. He called this

sound change ‘sporadic’ and irregular and not speciﬁc to North Dravidian only. I have

restated Rule 13c by including all high and mid vowels in the environment in view of

ety. (46) below. Some of the etymologies involved follow (see Emeneau 1994:

§§28, 29):

(46) PD

∗

c¯ınt- ‘date-palm’. SD I: Ta. ¯ıntu, ¯ı ˜ncal, Ma. ¯ıntal, Ka. ¯ıcal (<

∗

¯ı n -ccal

∗

¯ıntt-al); SCD (SD II): Te.

¯ıta (<

∗

¯ı n -tt-), Go. s¯ındi, h¯ındi, ¯ındi,Ko

s¯ıntel,Kuis¯ıta, Kuvi s¯ındi; CD: Pa. Gad. s¯ındi:NDKu

r. k

¯ınd¯a [2617].

∗

t-u ‘to burn’: SD I and II, CD have cognates: Ku

r. Malt. ku

r- [2654]; PD

∗

cur-V

‘to shrink, shrivel’. South Dravidian I, Te. and Central Dravidian have cognates: ND:

kurr- [2687]; PD

∗

cum-V- ‘to carry on head’. South Dravidian I, South Dravidian II,

Central

Dravidian have cognates:

ND: Ku

Malt.

kum-,

Br.

kub¯en ‘hea

vy’ [2677]; PD

∗

c¯ur/

∗

cur-V- ‘to curl, roll up’. South Dravidian I, South Dravidian II, Central Dravidian

have cognates: ND: Ku

r. k¯ur, Malt. ku

rg-, Br. k¯ur-[2684]; South Dravidian I and Te. also

have forms traceable to

∗

kur-u

l with the same meaning in

[1794]. The North Dravidian

forms can match both the etymologies. For the forms with

∗

c [

¯e Emeneau

gives

9 examples, not all of which are clearly reconstructable to

∗

¯e, e.g.

∗

cer- ‘to insert’. South

Dravidian I and South Dravidian II have cognates; ND: Ku

r. xerr-, Malt. qer-. The North

Dravidian consonant points to PD

∗

k- rather than

∗

c-, but Ka. has no k-. Emeneau has

given k- forms for

∗

c- sporadically in Toda [2599], Parji [2484], suggesting c-> Ø-,

t-, k- [2591] etc. Since there is no pattern in these changes, one cannot be certain of

the cognates grouped under the same entry. For instance, I would separate the words

meaning ‘wing’ in the following etymology into two groups:

(47) a. PD

∗

cet-ank-/-ankk- ‘wing’. SD I: Ta. cirai, ciraku, irai, iraku, irakkai,

Ma. ci

raku,To.tergy, Ka. eranke, erake, rakke, rekke,Ko

d. terake,Tu.

edi˙nke, re ˙nke; SD II: Te. e

raka, rekka (some South Dravidian II and

Central Dravidian languages have borrowed from Telugu).

b. PD

∗

ket-V- ‘wing, feather’: Ko. kera (

)l,Tu.kedu

nke ‘tip of wing’, kedi

‘feather’, Go. ger

(g) [2591].

In any case there are clear etymologies suggesting PD

∗

c > ND k- before non-low

vowels. A very valuable etymology pointing to

∗

c > k [ay ∼ ey is the group meaning

‘die’ which seems to involve a Proto-Dravidian laryngeal

∗

H because it is preserved

as a laryngeal fricative in Brahui and also because of the aberrant vocalism in different

languages (Krishnamurti 1997b: 152):

(48) PD

∗

caH- ‘to die’, PSD

∗

caH-/

∗

c¯a-. SD I: Ta. c¯a (non-past c¯a-v-, past

ce-tt-<

∗

ca-tt), Ma. c¯a-(catt-); with

∗

H > y,To.soy-(sot-) (borrowing

4.5 Historical

phonolo

gy: consonants

127

from Badaga), Ka. s¯ay-(satt-), Tu. sai- ‘to die’; all the South Dravidian

languages also have a noun in

∗

c¯aw (<

∗

caH-w-) ‘death’; SD II: Te. cacc-

(past cacc- <

∗

caH-cci, non-past [imper/inf] c¯a-/c¯aw-, by contraction of

∗

aH to ¯a) ‘to die’, ca-mpu (<

∗

caH-mpp-) v.t. ‘to kill’, c¯awu n. ‘death’, Go.

sai-, s¯ay-, s¯a-, h¯a-, Ko

da s¯a-(s¯a-t-), Kui s¯a-(s¯a-t-), Kuvi hai-, h¯a-(h¯a-t-),

Pengo–Man

da h¯a-(h¯a-t-). PCD

∗

cay-/

∗

c¯ay-: Pa. cay-(ca-˜n- < cay-nj-

∗

cay-nd-), Oll. say-, Gad. cay- ‘to die’. ND

∗

keH-,

∗

key-: Ku

r. kh¯e-∼

∗

k¯e

(ke-cc-), Malt. key-(kec-) ‘to die’, keype n. ‘death’, Br. kah-(kask-, neg,

kas-) ‘to die’, kas-if- ‘to kill’ [2425].

4.5.1.3.1

∗

c- and

∗

t-mergerinTu

lu PD

∗

c- and

∗

t- merged into

∗

c- in ‘Proto-Non-

Brahmin Tu

lu’ which later changed

into

c-, s-, h- and Ø- in different

lu dialects,

the exact demarcation of which is not known (Subrahmanyam 1983: 321–2, based on

Shankar Bhat’s observations). Kekunnaya (1994: 52–3) says that PD

∗

c-,

∗

t- = Tu. s-, t-in

North Brahmin and South Brahmin dialects. PD

∗

c-,

∗

t- >

∗

c- in the Common (non-

Brahmin) dialects leading to the representation of this

∗

c-ass- in SC and as t-

in NC:

SB NB Proto-C SC NC

∗

c-Ta.cappu ‘leaves’ sappu sappu sappu tappu

∗

t-Ta.talai ‘head’ tarε tarε sarε tarε



∗

The result of

∗

t-,

∗

c- >

∗

c- is further represented

c- in the Harijan dialect

(south-

west and south-central) and as h- in the Jain dialect; s-/h- vary in south-central B

and C dialects and it is represented as Ø- in southeast Common and Harijan dialects.

Interdialectal borrowing makes this picture further confusing, replacing s- of southern

dialects by t- of northern dialects (see Emeneau 1994: 370–2).

4.5.1.3.2

∗

c- > s- > h- > Ø in Gondi dialects PD

∗

c- is represented in different

Gondi dialects as s-, h-, Ø-. There are three items that have c- instead of s- in Adilabad

and Yeotmal districts [2391, 2677, 2865] and it could be due to the inﬂuence of Marathi.

In most of the items even these western dialects have only s-. The western and northern

dialects represent Proto-Gondi

∗

s-bys-, the farther eastern and southern dialects sho

Ø, and the middle dialects have h-. Though this sound change is still in progress, it is

a completed sound change in the southern dialects of Hill-Ma

ria, other Gondi dialects

of Bastar, and Koya of Andhra Pradesh. This change has been treated by Krishnamurti

(1998b) to establish that lexical diffusion can result in a regular sound change. The

second aspect is that the Gondi change is a continuation of the change of

∗

c > Øin

SD I and Telugu and it provides the missing phonetic links in the ﬁnal output, i.e.

∗

c- >

[s- > h->]Ø.

128 Phonolo

gy: historical

and comparative

4.5.1.4

∗

Burrow (1943) has dealt with this problem systematically followed by Krishnamurti

(1961:

§§1.17–21, pp. 10–11). Burrow’s statements can be formulated as a set of rules:

Rule 14. Palatalization of velars

∗

k > c/#

−−−

[+V, −back] (Telugu)

∗

k > c/#

−−−

](V

= [−back], C

= retroﬂex consonant) (Pre-Tamil)

The sound change occurred in Telugu and Tamil–Malay¯a

lam, independently, because the

environments are different. Tamil and Malay¯a

lam palatalize, if the

∗

k- is not followed by

a retroﬂex consonant in the next syllable. Telugu has no such constraint. Before Proto-

Dravidian front vowels as well as derived front vowels, Telugu palatalizes the velar

∗

k-.

For instance,

∗

ay > ¯e causes palatalization in Telugu and not in Tamil and Malay¯a

lam

(Burrow 1943: 128), e.g. Te. c

¯eyi ‘hand’ <

∗

key <

∗

kay (ety. (41)), c¯enu ‘ﬁeld’ (< key-m-

< kay-m-) [1958], c

¯edu ‘bitter’ (<

∗

key-nd- <

∗

kay-nd-), see (42). Krishnamur

ti (1961:

§1.18,

p. 10) showed that palatalization

occurred after another sound

change, namely

metathesis and vowel contraction, which blocked palatalization by removing the front

vowel from the immediate environment of

∗

k-. Forms such as krinda (<

∗

z-nd-) ‘below’

do not show palatalization; similarly, a number of words beginning with gi-, g¯ı -, ge-, g¯e-

are also not palatalized, giccu ‘to pinch’, g

¯ıku ‘to scribble’, g¯ı

ru ‘to scratch’, ge

tu ‘to

push out by neck’ etc. (Krishnamurti 1961:

§1.59, p. 25). Obstruent voicing could have

preceded palatalization and, therefore, inhibited its occurrence. E. Annamalai (1968) has

provided solid evidence for the non-operation of the palatalization rule in echo-words.

In a recent article Emeneau (1995) has illustrated the non-operation of palatalization in

several forms which had an alveolar as C

in Pre-Tamil. Therefore, the environment in

Rule 14b can be expanded to include alveolars, i.e. C

= [+apical]. Examples:

(49) PD

∗

key ‘to do, make, create’. SD I: Ta. Ma. cey,To.k¨ıy,Ko.key, gey,Ko

key,Tu.geyi-, gai-; SD II: Te. c¯ey-, Go. kiy-, k¯ı -, Ko

da ki-, Kui ki-, gi-,

Kuvi k¯ı -, Pe. ki-, Man

da ki-; CD: Pa. Oll. Gad. key-; ND: Br. k¯e- [1957].

(50) PD

∗

t-u ‘to perish, decay, be spoiled’. SD: Ta. Ma. ke

t-u,Ko.ke

r-, To.

k¨o

r-, Ko

d. k¨e

d- ‘to be spoiled’, Ka. ke

du, ki

du (ke

t-); SD II: Te. ce

[1942].

∗

l-i ‘parrot’. SD I: Ta.

Ma.

li, To. Ko. ki

ly,Ko

d. g¨ı

ni, Ka. Tu. gi

li, gi

ni;SDII:

Te. ciluka; (Go. si

r¯ı ,Ko

da si

ra,Pe.hi

ra are plausibly loanwords from Pre-Telugu in

the form

∗

li >

∗

li); CD: Pa. kil, Gad. killi

[1584, see (37)].

(51) ?PD

∗

kic-V kic-V ‘chirping, squeaking’ (onomatopoetic expression). SD

I: Ta. Ma. kiccu kiccu, k¯ıccu, Ka. kica kica,Tu.kicikici n. ‘screaming’, k¯ı c -

‘to squeak’; SD II: Te. kicakica adv., k¯ıcur¯ayi ‘an insect making noise’.

4.5 Historical

phonolo

gy: consonants

129

Etymologies (49) [earlier see (8) and (18)] illustrate Rule 14a, (37) and (50), Rule 14b.

Ety. (51) exempliﬁes the non-occurrence of palatalization in expressives.

Emeneau (1995:

§§5–17) has noticed that in Tamil and Malay¯a

lam even alveolars

t ttnt l n r/ have blocked palatalization in a considerable number of cases, if not all. He

gives ﬁve examples with

r, one with n and two with l as C

,Ta.kir¯avu ‘to cry’ [1590],

rukku, k¯enam ‘craziness’ [1596, 2021], kil ‘to be able’[1570], etc. There are cases

for non-palatalization in Telugu, but these look like early borrowings from Kanna

da in

the literary texts: kelasamu ‘work’ [1970], kelanu ‘side’ [1969] (cf. Ka. kelasa, kela);

several expressives are anyway exceptions to palatalization: Ta. kila kila [1575], Ta. ki

ra [1593]; for another expressive Emeneau reconstructs a voiced stop for PSD

∗

giru

ru [1595] ‘go round

and round

’ with co

gnates from Kanna

da, Tu

lu and T

elugu. Further,

Emeneau cites six forms where the palatalization rule is not blocked even when C

is an

alveolar, ce

ri ‘be tight, crowded’, ceru ‘to control’ [1980], cirukku ‘be angry with’ [1597]

with cognates from Kanna

da, Telugu, Kui and

Brahui which, according to Emeneau,

are

related to

∗

kin- ‘to be angry’: Ta. cina v.i. ‘be angry’, Ma. cinam ‘anger’, Ka. kinisu,Te.

kiniyu ‘be angry’ [1600], Ta. ce

ru (cerr-) ‘to kill’ [1981], Ta. cil-/cir- ‘small’(<

∗

kil-/

∗

kit-)

[1571, 1594] with cognates mainly from SD I, Ta. cil ‘sound, noise’ (<

∗

kil-) [1574], Ta.

cil ‘small piece’ [1577], Ta. c¯e

ru ‘mud’ (<

∗

kec-at-) [2020], Ta. c¯enai ‘yam’ (<

∗

k¯en-)

[2022]. In two entries Ma. ki

r- lacks palatalization [1591,1562]. Two more cases where

∗

-r seems to block palatalization are Ta. k¯ıri ‘mongoose’ [1614], k¯ırai ‘greens’ [1617].

It is important to note that most of the instances where palatalization is blocked

seem to be conﬁned to South Dravidian I. Emeneau’s paper gives one more reason for

alveolars and retroﬂexes to be grouped as a natural class at the proto-level (see sections

4.2, 4.5 above). In the case of Ko

dagu and the Nilgiri languages we have noticed that

vowel-centralization is caused by the following retroﬂexes and alveolars, mainly

∗

(Krishnamurti 1975b; Emeneau 1995: 407–8).

Rule

15.

Spir

antization/retraction of the velar (North Dravidian)

∗

k > x, q/#

−−−

= All but the high front vowels i¯ı[+V, −[+high,

−back]]).

When Burrow (1943:

132

–9) formulated this

change he said that only before high front

vowels i,¯ı ,

∗

k- remained k-; before all other vowels it became /x/ (spelt as kh-) in Ku

rux

and Brahui, and as q- in Malto

in North Dra

vidian. He discussed this change in great

detail and also tried to explain the few exceptions to the rule framed by him. He gave

24 items where the rule operated and 8 where

∗

k- remained unchanged before high

Originally it was thought that Malto q was also a postvelar spirant, but Burrow and Bhattacharya

in a ﬁeld trip in 1957–8 discovered that q is a uvular stop and not a spirant (see Emeneau 1994:

14). Mahapatra (1979: 25) conﬁrms this fact.

130 Phonolo

gy: historical

and comparative

front vowels. Pfeiffer (1972: 63–7) demonstrated that the rule was generalized in Ku

rux

and Malto extending the environments to all high vowels, i.e.

∗

k remains k- before all

high vowels, but Brahui still retains the original environment (Emeneau 1994: 14, fn.1).

Examples for this rule can be seen in ety. (8), (11), (12), (18), (38) and (41), to which

the following may be added:

(52) PD

∗

k¯ay ‘to grow hot’. SD I: Ta. k¯ay, Ma. k¯ayuka,Ko.k¯ay,To.k¯oy,Ko

k¯ay,Tu.k¯ay, Ka. k¯ay; SD II: Te. k

agu v.i., k

acu v.t., Go. k¯as- v.i., k¯ap v.t.,

da k¯ay-, v.i., k¯ap- v.t., Kui k¯ag-, k¯and- v.i., k¯as- v.t., Kuvi k¯ay-, k¯ad-

v.i., k¯ap- v.t. Pe. Man

da k¯ay v.i.; CD: Kol. Nk. k¯ang- v.i., k¯ap- v.t., Pa. k¯ap-

v.t., Oll. k¯ayp- v.t.; ND: Ku

r. x¯ay-, Malt. q¯ay- v.i., Br. x¯axar ‘ﬁre’ [1458].

(53) PD

∗

k¯u-r- ‘to sleep’. SD I: Tu. k¯ur-uni; SD II: Te. k¯uru v.i., k¯urku ‘a nap’,

Go. k¯urk- ‘to doze’; CD: Pa. Gad. k¯urk- ‘to nod in sleep’; ND: Ku

r. Malt.

k¯ug- ‘to doze’ [1902].

4.5.2 Initial consonants: voiced stops

Burrow (1938) discussed the question of initial voiced stops, comparing etymologies in-

volving this feature in the literary languages of South India, mainly Telugu and Kanna

da.

He concluded that voicing was secondary in Dravidian and the Proto-Dravidian con-

dition is preserved in Old Tamil and Malay¯a

lam. He was supported with further argu-

ments by Krishnamurti (1961:

§§1.55–9, 1.70–3, pp. 24–5, 28–9). Zvelebil (1970b),

Emeneau (1970a) and

Subrahmanyam (1983) have endorsed

this position but have not

added anything new. Krishnamurti gives the environments which seem to promote sec-

ondary/sporadic voicing in word-initial obstruents, in so far as Telugu is concerned, but

the observations apply to a number of other languages also in different etymological

groups: (1) when the root syllable ends in a phonetic sonorant as opposed to a voiceless

obstruent, T

gillu ‘to pluck’, jarugu ‘to slide’, etc.; (2) w

here a sonorant occurs as the

second member of a word-initial cluster which converts an initial obstruent to a voiced

one, e.g. Te. kruccu/ gruccu ‘to pierce’, etc.; (3) when a radical liquid is assimilated to a

voiced obstruent, e.g. Te. diddu ‘to rectify’ (<

∗

tirdu); (4) where a root-ﬁnal semivowel

or liquid is lost before a nasal–stop combination, Te. bonku ‘to lie’ (< poy-nkk-), also

where a root-ﬁnal nasal is joined to an obstruent, e.g. janku ‘to fear’ (< ca

n-kk-).

While the general principle seems to be the extension to initial voice-

less stop of voicing occurring in the succeeding syllables, there are also

innumerable cases where initial voicing is not produced under the same cir-

cumstances. However, voicing is more wide-spread in constructions, which

are capable of being analysed (in PDr.) into a base plus a derivative sufﬁx

than in unsufﬁxed stems as kappu, tappu, tannu, etc. Again the relative

susceptibility of the various voiceless stops to the assimilative inﬂuence of