Krishnamurti Bhadriraju. The Dravidian Languages

Подождите немного. Документ загружается.

4.4 Historical

phonolo

gy: vowels

111

There are a few items of Pre-Toda

∗

o > u which are considered borrowings from

the neighbouring languages. Pre-Toda

∗

o remains in six items only including the name

of the tribe, To. toz, mox ‘a Toda woman’ (Subrahmanyam 1983: 145–6). All the rules

formulated to explain the developments in Toda belong to the Pre-Toda stage when the

formative vowel was still identiﬁable, since it is the quality of this vowel that seems

to determine a number of changes. We need to recognize a further rule dropping V

between Pre-Toda and Toda:

(xiii) V

> Ø/(C

−−−

There are also other changes like contraction of root and formative syllables, loss

of preconsonantal nasal, etc., which are obvious from the examples given. In almost

all cases where he encountered exceptions, Subrahmanyam has invoked a loanword

explanation without further discussion. In many of such exceptions it can be that the

sound change is still in progress and has not covered the whole lexicon. For instance

∗

remains in the name of

the Todas as

toz [todz], which does not

seem to be a borrowing

from any other language.

4.4.4.2 Centralized vowels in Nilgiri languages (NL)

With the

exception of Kota, all the other

Nilgiri languages have centralized

vowels (see

section 2.1.1). Toda has high and mid central vowels (short and long), ¨ı

¨ı¨o

¨o ; it also has

front rounded ¨u and

u. Zvelebil 1973 (11–12) set up four centralized vowels ¨ı

¨ı, ¨e

¨e ,¨u

u, ¨o

(central unrounded and rounded) for Iru

la, but in 1979 (ch. 4) he adds ¨a and

a; Perialwar

(1979: ch. 2) has ﬁve centralized vowels, short and long, for Iru

la, corresponding to

the ﬁve non-centralized ones, short and long. Ku

rumba dialects (Kapp 1978, 1987;

Zvelebil 1982a, 1988) also have ¨ı and ¨e, short and long. With the exception of Kota,

most of the Nilgiri languages have centralized vowels caused by splits of i and e when

followed by retroﬂexes (or alveolars in some cases); this statement does not completely

deﬁne the environments that centralize vowels in Toda. In all cases, original allophonic

differences became phonemic when the conditioning factors became obscure or lost. This

phenomenon has not inﬂuenced one Nilgiri language, namely Kota. Ko

dagu, which is not

a Nilgiri language, also has centralized vowels. First, examine the following examples:

(37) PD

∗

li ‘parrot’. Ta. Ma. ki

li, Ka. Tu. gi

li, gi

ni,Ko

d. g¨ı

ni; NL: Ko. ki

vaky,To.ki

ly [1584], Ir. k¨ı

li,

AKu., PKu k¨ı

li.

(38) PD

∗

k¯ı

z/ki

z-V- ‘below’. Ta. Ma. k¯ı

z/ki

z-V, OKa. k¯ı

z/ki

z-(V)-, ke

lagu ‘bot-

tom’, Ko

d. k

¨ı

l¨ı, k¨ı-, Tu. k¯ı

l¨ı ; NL: Ko. k¯ı-, k¯ı

rm ‘monsoon clouds’ (from

the western side), To. k¯ı - ‘lower, below’, k

¨ıe, k

¨ıye ‘down, below’ [1619],

AKu. k

¨ıe (k

¨ı-, k

¨ıa- in cpds.).

(39) PSD I

∗

tu ‘eight’. Ta. Ma. e

tu, Ka. e

tu,Ko

d. ¨e

t¨ı,Tu.e

nma; NL: Ko.

t,To.¨o

t [784],

AKu. ¨e

tu.

112 Phonolo

gy: historical

and comparative

(40) PSD

∗

¯e

z-:

∗

z-V- ‘seven’. Ta. Ma. ¯e

z(u)/ e

z-u, OKa. ¯e

zu/ e

z-u, ¯e

l-/e

l-u,

l¨ı,Tu.¯e

l¨ı; NL: Ko. ¯ey, e

l-, e-, To. ¨ow,

o,Ir.

lu ,

AKu.

lu [910].

The above data are important to establish two facts. First, centralization is not a phe-

nomenon of all Nilgiri languages, because Kota has no centralized vowels and a non-

Nilgiri language like Ko

dagu also has centralized vowels. Second, the conditioning

factors of centralization are not the same among different Nilgiri languages. Therefore,

centralization of vowels cannot be cited as an areal feature of all the Nilgiri languages

(Zvelebil 1980).

The Nilgiri languages are all non-palatalizing, unlike Tamil and Malay¯a

lam (see sec-

tion 4.5.1.4, Rule 14b). They must have separated from Pre-Tamil before the rule of

palatalization set in, i.e. before recorded literature evolved in Early Tamil, about the

third century BCE (Emeneau 1989: 135). We notice that palatalization did not occur in

early Tamil before retroﬂexes and also some alveolars, especially

∗

-r (Emeneau 1995:

§14, 18). This means that the front vowels /i ¯ıe¯e/ were retracted and centralized be-

fore retroﬂexes and possibly some alveolar consonants at that point in time, as shown

by applying the principle of phonetic reconstruction inferred from the results of sound

change. Firth (1934) says that even in modern Tamil the front vowels are retracted be-

fore retroﬂex consonants ‘giving the vowels a centralized obscure quality’ (cited by

Emeneau 1994: 194). This phenomenon must have been shared by all those languages

that have developed centralized vowels by the loss of retroﬂex and alveolar consonants

that caused

the allophones. Toda, Iru

la and Ku

rumba must have split from Pre-Tamil af-

ter it developed the centralized allophones of front vowels, but before palatalization took

place. The environments causing centralization of front vowels (following retroﬂexes

and

∗

t) are mostly similar between Ko

dagu and Kurumba. Toda and Iru

la also have

centralized front and back rounded vowels, not shared by Ko

dagu and Ku

rumba. Iru

also has lost initial short vowels before alveolar

∗

r and

∗

l, not found in Toda. They

do not seem to have a common stage of development. Kota must have separated at

a still earlier point in time, i.e. before Pre-Tamil developed centralized allophones of

the front vowels when followed by retroﬂex consonants. We will review this hypoth-

esis in the concluding chapter. The successi

ve splits of Pre-Tamil are represented in

ﬁgure 4.2.

4.4.4.3 Ko

dagu developments

dagu registers the split of Pre-Ko

dagu i to i /¨ı and e to e/¨e; the centralized vowels

occur before retroﬂex consonants and r (<

∗

-t-[-r-]). At a later period, the consonants

that conditioned the original allophones were lost through assimilation to the following

obstruents. After bilabial consonants front vowels become the corresponding back vow-

els which bleed all such instances from centralization. Root vowel

∗

e becomes a when

4.4 Historical

phonolo

gy: vowels

113

Pre-Tamil

Kota Toda Irula Kurumba Kodagu Malayalam Tamil

Figure 4.2 The Pre-Tamil sub-branch of South Dravidian I

followed by retroﬂexes followed by V

, that is [+low] (Emeneau 1970a: 46–9, 1970b:

145–58, repr. 1994: 183–201).

Rule 8a. Lowering of front mid vowel

∗

e > a/# (C

)

−−−

-V

∗

k;V

= [+low]) Pre-Ko

dagu to

dagu change

PSD

∗

l-ay >Ta. ki

lai ‘to dig up’, Ko

d. ka

l-a [1588]; PSD

∗

t-a- >Ta. pi

t-ar ‘nape of

neck’, Ka. pe

da:Ko

d. pa

da ma

de ‘back of head’, Pa. pi

dtel ‘behind’ [4146].

Rule 8b. Retraction and rounding

of front vowels after labials

i¯ı e ¯e >u¯uo¯o /#C

1−−−

-V

= p, m, b[<

∗

w])

PSD

∗

t-u ‘to leave, release’ > Ta. vi

tu (vi

t-): Ko

d. bu

d-(bu

t-), PSD

∗

w¯ı

z-/

∗

z-V- >Ta.

v¯ı

z (-v-, -nt-)/vi

z-u(-v-, -nt-) ‘to fall’: Ko

d. b¯u

l-(b¯uv-, budd-) [5430],

∗

ti ‘box’ >Ta.

Ma. pe

ti:Ko

d. po

ti ‘box’, PSD

∗

v¯e

l-nt- ‘to want’ >Ta. v¯e

tu ‘to want’: Ko

d. b¯o

d- ‘to

beg’ [5528].

Rule 8c. Centralization of front vowels

i,¯ı , e, ¯e > ¨ı,

¨ı, ¨e,

¨e /#(C

)

−−−

-V

= labial or palatal; C

= Retroﬂex

∗

t ) Pre-Ko

dagu to Ko

dagu.

∗

t-u ‘to put, place’: PSD

∗

t-u > Ta. i

tu (i

t-): Ko

d. ¨ı

d-(¨ı

t-) [442], PD

∗

k¯ı

z ‘below’ >

Ta. k¯ı

z > Ko

d. k

¨ı

l¨ı [1619], PSD

∗

tu ‘eight’ > Ta. e

tu,Ko

d. ¨e

t¨ı [784], PD

∗

t¯et- ‘to be

thorough, to recover’ > Ta. t¯e

ru,Ko

d. t

er-(t

eruv-, t

end-) ‘(man) becomes fully grown’

[3471]. The centralization here took place before

∗

t [-r] merged with r (<

∗

r), because it

does not take place before r <

∗

r or before the reﬂex of geminated

∗

tt. Rule 8c applies

to items that hav

e been left out after Rules 8a and 8b have applied. It is in a bleeding

relation to those rules. They also indicate the chronological proﬁle of the changes.

114 Phonolo

gy: historical

and comparative

4.4.4.4 Kota vowel shift

Kota has no centralized vowels like the other Nilgiri languages. This has been explained

as due to its earlier separation from Pre-Tamil. It has, however, another idiosyncratic

change of the root vowel being harmonized to the formative vowel

∗

-e (<

∗

-ay) which

was later lost (Emeneau 1970a: 49–50, 1969b: 21–34, repr. 1994: 175–82).

Rule 9a. Fronting and raising of formative -ay

ay > e/#(C

) o, u (C

−−−

#(C

= any consonant admissible in the

intervocalic position: (C

= a geminate stop or a nasal + stop; stem-

ﬁnal

∗

-ay becomes -e in Pre-Kota)

Rule 9b. Root-vowel assimilation

∗

u > e, i/#(C

)

−−−

-V

= e from Rule 9a)

Root vowels o, u become e, i respectivel

y, before formative

e in Pre-Kota.

Rule 9c. Loss of the formative vowel

∗

e > Ø / #(C

)e, i (C

−−−

Stem-ﬁnal, formative -e derived from Rule 9a is lost; Pre-Kota to Kota.

Rule 9d. Simpliﬁcation of ﬁnal consonant cluster

> C

/# (C

)e, o

−−−

The ﬁnal cluster resulting from Rule 9c is simpliﬁed, or (C

) is lost between Pre-Kota

and Kota.

This is a three-step sound change: (1) stem-ﬁnal formative -ay becomes -e in Pre-

Kota; (2) radical vowels o, u are fronted to corresponding height, being assimilated

to the following formative vowel -e; (3) the loss of the formative vowel -e and the

simpliﬁcation

of the preceding consonant cluster. The root vowel can

be short or long

and the intervening consonant can be single or a cluster. The

consonant clusters in the

intervening position are PP (geminate stops) or NP nasal + stop.

Examples: PSD

∗

t-ay ‘umbrella’ > Ta. ku

t-ay:Ko.ke

r (<

∗

r-e <

∗

d-e <

∗

ay) [1663]; the relative chronology of what happened to the medial consonant(s) is not

relevant to the sound changes in question; PSD

∗

t-ay ‘stone of fruit’ >Ta. ko

tay:

Ko. ke

t ‘testes’ [2069]; Ta. k¯o

tay ‘west wind’: Ko. k¯e

r ‘SW monsoon’ [2203]; PSD

∗

ol-

a-kk-ay ‘pestle’ > Ta. ulakkai:Ko.elk (also loss of -a- between -l- and -kk-) [672]; PD

∗

kupp-ay ‘heap’ > Ta. kuppai:Ko.kip [1731a]; Ta. pu

tai ‘male organ’: Ko. pi

d ‘vulva’

[4273]; PSD

∗

m¯u

lay ‘brain, marrow’ > Ta. m¯u

lai,Ko.m¯ı

l. [5051].

Emeneau (1994: 178–82) has noted that

there are many exceptions to the above sound

changes. Seven out of 26 items which register the changes as proposed have doublets like

4.4 Historical

phonolo

gy: vowels

115

peg, pog ‘smoke’ (<

∗

pokay), mel, mol ‘breast’ (<

∗

molay). Besides, there are 22 nouns

and 9 verbs with unchanged vowels under the conditions of expected change, e.g Ko. ¯ol

‘palm leaf ’: Ta. ¯olai [1070], Ko. ko

lv-(ko

ld-) ‘to rot’: Ta. ku

zai (-v-, -nt-) [1822]. For

some of these Emeneau invoked borrowing from neighbouring languages like Ba

daga,

Kanna

da and Tamil (179). It is possible that the order in which the above sound changes

applied could be different, producing different results. If Rules 9b and 9c started from

different ends of the Kota territory, it is possible that loss applied ﬁrst and blocked assim-

ilation of the root vowel in the overlapping area, leading to what is called rule reordering.

The large number of doublets support this alternative. In the case of verbs -e- <

∗

-ay

occurs medially and its loss must have preceded the assimilation rule (Krishnamurti

1994a: xvi–xvii). It is also possible that the sound change had ceased midway.

4.4.4.5 Tu

lu development

A rule similar to Rule 8b in Ko

dagu also occurred in Tu

lu independently whereby front

unrounded vowels of the root syllable became rounded after bilabial consonants /p b m/

(Emeneau 1994: 199–200; earlier Ramaswami Aiyar 1936: IL 6.385–439):

Rule 10a. Rounding of front vowels after labials

[+V, −back, −rounded] > [+back, +rounded]/ # C

1−−−

= [+labial],

= retroﬂex [+apical, −anterior])

This rule must have been quite native and old to Tu

lu, since it occurs in all regional

and social dialects (Kekunnaya 1994: 50–1). Emeneau thinks that this change could be

basic to Tu

lu and that it spread to Ko

dagu later through areal diffusion, creating a state

of sharing an isogloss (1994: 199–200).

PSD

∗

ti ‘handle’ > Ta. pi

ti, Ka. pi

di:Tu.pu

di; PSD

∗

tu ‘to leave’ > Ta. vi

tu,

Ka. bi

du:Tu.bu

du, PSD

∗

w¯ı

z ‘to fall’ > Ta. v¯ı

z, Ka. b¯ı

lu:Tu.b¯u

lu; PSD

∗

n ‘woman’

>Ta. pe

n, Ka. pe

n, he

nu:Tu.po

nu; PSD

∗

w¯e

tu ‘to wish’ > Ta. v¯e

tu, Ka. b¯e

du ‘to

beg’: Tu. b¯o

du ‘to wish, want’. The sound change is different from that in Ko

dagu in

that a following

∗

t does not behave like a retroﬂex. It appears that the sound change is

now spreading to environments where no retroﬂex is involved as C

, e.g. PSD

∗

wittu

‘seed’ > Ta. wittu:Tu.buttu. In one case a reverse change

∗

u > i/#p– is noticed, which

is dissimilatory, e.g. PSD

∗

puli ‘tiger’ > Ta. puli:Tu.pili, PSD

∗

molay ‘breast’ > Ta.

mulai:Tu.mirε (Kekunnaya 1994: 51).

4.4.4.6 Aphaeresis in Iru

la and Tu

Aphaeresis is the loss of a word-initial vowel before an accented or heavy syllable.

Iru

la shows evidence of such a loss after it had branched off from Pre-Tamil. This

sound change mainly occurred before Proto-South Dravidian intervocalic

∗

r and

116 Phonolo

gy: historical

and comparative

∗

l. Zvelebil (1979: 68) gives 5 words with r-, 1 with r- and 2 items with l-. Perialwar

(1979: 228–9) has listed 9 different lexemes with r - and 7 with l-, all native. DEDR

index has incorporated 8 items with r-, 1 with

r- and 2 with l-. I have not taken into

account the loanwords from Indo-Aryan, e.g. Ir. ra

du, re

du (the latter lw from Mdn

Ta.?) ‘two’: Ta. ira

tu (>spoken re

du) [474], Ir. r¯avu, r¯apodu ‘night’: Ta. ir-¯a, ir-avu, r¯a

‘night’ [2552], Ir.

rangu ‘to descend’: Ta. ir-anku [516], Ir. rongu ‘to sleep’: Ta. uranku

id. (< PSD

∗

ot-a-nku) [707], Ir. l¯e-(l¯end-) ‘to wander’: Ta. alai (-nt-) ‘to wander’ [240].

Besides the vowel loss, we notice change in vowel quality in re

du, rongu. The form

du, being a doublet beside the normal ra

du, and the representation of r for

r in

orangu make the history of these two items suspicious.

The tendency to lose an initial short vowel is also found, on a limited scale, in all

South Dravidian languages, but the

change apparently is intensi

ﬁed

in Iru

and Tu

lu,

independently. Kota has one form rek ‘wing’ [2591] and Toda has none beginning with

r-orl-. Tu

lu not only has entries with l- (14) and r - (15), but also with

d (11) – indicating

a wider application of the rule of initial short vowel loss. Ko

dagu has three items with

r- [489, 2591, 5169] all of which look like loanwords from Kanna

da and one with l-,

onomatopoetic [5195]. Kanna

da has 18 entries with r-, 6 with

r- and 18 with l-, several

of which seem to be native and not loanwords. Under Tamil there are 7 with r- and one

with l- indicating initial short vowel loss, some time between Middle and Modern Tamil

(three more items with l- are onomatopoetic, like lo

ta lo

ta).

In Tu

lu aphaeresis takes

place dialectally. In the Nor

thern Common dialect the fol-

lowing rule operates (Kekunnaya 1994: 42–4):

Rule 10b. Word-initial short vowel loss before apicals

→ Ø/#

−−−

-V

-(V

= a, e, o; less frequently i, u;C

= a

retroﬂex consonant

l [+apical, −anterior] ; C

= geminate or nasal

+stop sequence), e.g.

Gloss SB SC NB NC

‘to throw’ a

dkk¨ıa

dakk¨ıa

dak¨ı dakk¨ı

‘to measure’ a

lappu a

lapu lappu

‘to dry’ u

ngu u

ngu nu

ngu

In some cases the SC also drops the short vowel:

Gloss SB SC NB NC

‘to stumble’ e

nku

nku/da

nku e

nku da

nku

‘to call’ o

leppu leppu o

lepu leppu

The rule of vowel loss now extends to environments in which non-retroﬂex consonants

occur as C

(Kekunnaya 1994: 44):

4.4 Historical

phonolo

gy: vowels

117

Gloss SB SC NB NC

‘without’ idyantε idyantε, idy¯antε dantε d¯antε

‘fasting’ upasso p¯aso up¯asa p¯asa

The last item, obviously a borrowing from Indo-Aryan (Skt. upav¯asa-), has suffered

vowel loss in the common dialects of south and north. Kekunnaya gives seventeen

examples of vowel loss in all. I discussed this ‘trend’ in South Dravidian I, which could

be interpreted as an example of what Sapir (1921: 172) called ‘drift’. The typological

design underlying

the drift could be to maintain the

CVC structure of the root syllab

while at the same time giving apical consonants an even distribution with the other

consonants by promoting them to the word-initial position (Krishnamurti 1978a: 17, 18

fn. 14). It is probably not possible to say that the South Dravidian II rules of metathesis

and vowe

l contraction are related to the

above rule historically as

a shared innovation.

4.4.5 Vowel changes in Central Dravidian: Parji

In Pre-Parji PCD

∗

¯a became e, ¯e in radical syllables (V

) when followed by an alveolar

consonant, PCD

∗

nd,

∗

tt,

∗

l deriving from PD

∗

l (Krishnamurti

1978b):

Rule 11. Low vowel fronting and raising before apicals in Pre-Parji

[+V, +low] > [−back, −low, −high]/# (C

)

−−−

= alveolars

[+anterior, +apical])

∗

¯at- ‘to cool’. PCD

∗

¯ad-/

∗

¯ar:Pa.¯ed,NE¯e

d ‘to cool, cool off’, Kol. Nk. ¯ar- ‘to dry up’;

PSD

∗

¯at- u [¯ar-u] id. [404], PD

∗

panti [pandi] ‘pig’ > PCD

∗

pandi:Pa.pend, (NE) pen

Oll. pan

d, Gad. pan

du; PSD

∗

panti [pandi/panri] [4039], PD

∗

man-(man-t-) ‘to be’,

PCD

∗

man-(man-d-, man-tt-): Pa. men-(mend-, mett-) ‘to stay’, Oll. man-, Gad. man-,

Nk. man-; PSD man-(man-

t-/man-tt-); PND

∗

man (n)- [4778]. There are forty-seven

etymologies in which the rule operates. This shows that, until the separation of the Parji–

Ollari–Gadaba subgroup, Proto-Dravidian alveolar stop sequences remained intact. Only

after Rule 11 did the rule of dialect split take place in Parji, which merged the Pre-Parji

alveolar stop with dental and retroﬂex stops respectively in different regional dialects.

Exceptions to the rule are motivated by grammatical criteria, namely pronouns ending

in -n,-r or are clearly identiﬁable loanwords from Indo-Aryan. True exceptions, which

are less than a dozen, have been discussed in detail in Krishnamurti (1978b: 258–9).

The vowel raising rule does not apply to a and ¯a, preceding l, n which derived from

∗

n later (for detailed discussion, see Krishnamurti 1978b).

In ﬁve lexical items we notice a change of

∗

¯e to a¯a in Parji before

∗

r and

∗

l; none

of the items which are the output of Rule 7 is involved in this, e.g. Pa. v¯ar ‘root’ <

∗

w¯er

118 Phonolo

gy: historical

and comparative

[DEDR 5535], Oll. Gad. Kol. Nk. v¯er id. For some of these items only the southern

dialect shows the low vowel form (Krishnamurti 1978b: 259).

4.4.6 Vowel changes in North Dravidian: Brahui

Brahui has inherited Proto-Dravidian high and low vowels, short and long /a ¯ai¯ıu¯u/.

Among the mid vowels, Bray gives short and long e¯e, but only ¯o. Emeneau (1962d:

§2.1)

has established that short o occurs only in unaccented non-root syllables and that there

is no contrast between e and ¯e. Therefore, he sets up only long mid vowels phonemically

/¯e¯o/, accounting for this situation because of the inﬂuence of neighbouring Indo-Aryan

and Iranian languages. Emeneau

shows that PD

∗

e is replaced by a and i , and

∗

o by ¯o,

u and a in root syllables. The conditions of the split are not clear. In a recent article

(1997: 441–3), he has provided evidence for PD

∗

e being represented by ¯e in root

syllables, which ﬁlls in a phonological gap. Examples: PND

∗

e > Br. a, e.g. hal ‘rat’

∗

el-i[833], kah- ‘to die’ (< PND

∗

keh- < PD

∗

caH-∼

∗

ceH-‘to die’); in this case only

Brahui preserves the Proto-Dravidian laryngeal as /h/: PND

∗

e > Br. i , e.g. pir ‘to swell’

∗

per-V- [4411], mir- ‘to plaster’ <

∗

z-V- [5082]. Emeneau gives 15 instances of

∗

e > a and 7 of

∗

e > i; from new material PND

∗

e > ¯e in Br. b¯egh ‘knead’: PD

∗

mel-k-,

cf. Ta. melku ‘to become soft (by

soaking, etc.)

’ [5078], with loss of l before k and

lengthening of the preceding vowel (1997: 442). There are 13 certain cases of retention

of PD

∗

¯e as Br. ¯e (Emeneau 1962d:

§§ 2.13–21). PND

∗

o > Br. ¯o (4 cases), e.g. t¯or-

(t¯on-, t¯o-) ‘to hold, keep’ <

∗

t-V- [3480], PND

∗

o > Br. u (5 cases), e.g. cu

t ‘a drop’,

ting- ‘to drip’ <

∗

t-: Ta. co

tu ‘to fall in drops’, n. ‘a drop’ [2835],

∗

o > a (2 cases),

Br. xall- ‘to strike, kill’ <

∗

kol- ‘to kill’ [2132]. There are 8 clear cases of

∗

¯o > Br. ¯o,

e.g. Br. t¯ola ‘jackal’, Ka. t¯o

la,Tu.t¯o

l¨ı [3548] (see Emeneau 1962d:

§§ 2.22–7).

4.4.7 Diphthongs ai and au

Literary languages have treated ai and au as diphthongs rhyming with long vowels

metrically. But in terms of reconstruction, roots and extended stems ending in ai and au

are structurally parallel to (C

-(V

) where C

is L = nasal or liquid, e.g.

∗

kay

‘hand’ parallels

∗

man ‘to live’,

∗

cal ‘to go’,

∗

kal ‘stone’ and

∗

ur-ay ‘to rub’ parallels

∗

tir-u

l ‘to turn’,

∗

mar-am ‘tree’, etc. In many cases, root-ﬁnal -y comes from an older

∗

-c,

∗

wey- ‘to be hot’ <

∗

wec-. The traditional

grammarians of the literary languages rightly

treat the non-syllabic vowels i and u as equivalent to y and w, respectively (Krishnamurti

1961:

§§1.121–2, 119–21). Proto-Dravidian roots ending in aw are rare.

∗

ay has different developments in root and formative syllables. The languages of

South Dravidian I preserve

∗

ay-; in South Dravidian II Te. has ay-/ey->¯e- and Gondi has

ay-; most of the languages of South Dravidian II, Central Dravidian and North Dravidian

develop it to ey/¯e and iy/¯ı with loss of -y before consonants, e.g.

4.5 Historical

phonolo

gy: consonants

119

(41) PD

∗

kay ‘hand’. SD I: Ta. kai, Ma. kai, kayyi,Ko.kay,To.koy,Ko

d. kay,

Ka. kayi, kayyi, kayyi, key,Tu.kai; SD II: Te. c¯eyi; in classical texts, also

kai- in compounds and k¯elu ‘hand’, Go. kay,Ko

da kiyu (pl kiku), Kui

kaju, kagu (pl kaska, kaka), Kuvi keyu (pl keska), Pe. key, Man

da kiy;

CD: Kol. key, kiy, k¯ı (dial), Nk. k¯ı ,Pa.key, Oll. ki (pl kil), Gad. kiy, kiyy¯u

(pl kiyyl, kiyk¯ıl); ND: Ku

r. xekkh¯a, Malt. qeqe [2023].

(42) PD

∗

kac->

∗

kay- ‘to be bitter’, ‘bitterness’. SD I: Ta. kai, kay-a, kac-a ‘to

be bitter’, kac-appu, kay-appu, kacc-al, kai-ppu ‘bitterness’, Ma. kai-kka,

kas-akka v.i., kaippu ‘bitterness’, Ko. kac- v., To. koy-, v., Ko

d. kay-v.,

Ka. kay, kayi, kayyi, kaypu n., Tu. kaipε, kaipe n.; SD II: c

¯edu ‘bitterness’,

kasu- ‘unripe’, Go. kay-, kaitt-v.,kait¯a, keht¯a adj, Kuvi kam-beli adj, Pe.

ke- v., Man

da kem-bel adj; CD: Nk. kayek ‘unripe’, Pa. k¯ep-(k¯et-) v., Gad.

keym-bur, k¯em-bur adj [1249].

This example shows that

∗

-c ∼

∗

-y was part of Proto-Dravidian variation since it is

reﬂected in all the subgroups. These contrast with radical

∗

ey in South Dravidian I and

the languages that maintain

∗

ay/

∗

ey contrast, Telugu, Gondi etc.:

(43) PD

∗

ney ‘oil, ghee’. SD I: Ta. Ma. ney,Ko.nay,To.n¨ıy,Ko

d. ney, Ka.

ney, n¯ey,Tu.neyi, n¯eyi; SD II: Te. neyyi, n¯eyi, Go. ney, n¯ı y, n¯ı,Ko

da niyu,

Kui n¯ıju, Kuvi n¯ıyu, Man

da ney; CD:

Kol. Nk. Pa. Oll.

ney, Gad. ney(yu);

ND: K

r. n

¯et¯a ‘grease,

fat

’, Malt. ne

nya ‘fat of animal’ [3746].

The formative-ay becomes -a in Malay¯a

lam and Telugu. Kanna

da, Pre-Kota, Ko

dagu and

lu change it to -e. The rest of the languages lose formative vowels (see Krishnamurti

1961:

§1.285, p. 121; ety. (23), (24), (26) and section 4.4.4.4).

4.5 Historical phonology: consonants

The Proto-Dravidian consonants are presented in a chart in table 4.2. Only nine con-

sonants occur word-initially (C

), namely four stops /p t c k/, three nasals /m n ˜n/ and

two semivowels /w y/. A phonologically signiﬁcant statement is that apical consonants

(alveolars and retroﬂexes) are excluded from word-initial position in Proto-Dravidian.

All consonants except

∗

˜n occur in intervocalic position, i.e. as C

in bound Proto-

Dravidian root syllab

les. Consonant-clusters which follow V

are PP or NP [NB]; in

the latter case, the ﬁrst segment (N) can also be a part of the root, established on

comparative grounds. There is also evidence for the occurrence of NP-P/N-PP with a

morphological boundary between NP and P, or N and PP. In the case of words end-

ing in stops a non-morphemic /u/ is added at the end. Single and double obstruents

contrast in Proto-Dravidian. In the case of sonorants, there is marginal evidence to

propose contrast between single and double laterals

∗

l and

∗

l (Emeneau 1970a:7,121;

120 Phonolo

gy: historical

and comparative

able 4.3.

Distribution

of Pr

oto-Dravidian

phonemes

Stop #

−−−

(V)# V

−−−

(V)# V

−−−

(V)#

−−−

(V)#

p- -w- -pp -mb -mp-p -p

t- -d- -tt -nd -nt-t -t

–-

d- /-r- -tt-nd-nt-t -t

–-

d- -

c- -s- -cc -˜nj -˜nc-c -c

k- -g- -kk- -ng -nk-k -k

Nasal

m- -m- – -m

n- -n- – -n

–-

n- – -

˜n- – – ?-˜n

Lateral

– -l- ?-ll- -l

–-

l- ?-

l- -

Approximant

–-r- – -r

–-

z- – -

Semivowel

y- -y- – -y

w- -w- – -w

– -H- [P -H

Subrahmanyam 1983:49);

3 ∗

∗

z occur only singly both in Proto-Dravidian and in the

descendant subgroups. Single and double sonorants contrast synchronically in many

Dravidian languages, but the contrast is not traceable to Proto-Dravidian. Table 4.3

shows the segments and sequences that can be reconstructed for Proto-Dravidian. Pho-

netic representations at the Proto-Dravidian stage are given below. The symbols in the

ﬁrst column can be taken to represent both phonemic and phonetic values.

4.5.1 Initial consonants: voiceless stops

4.5.1.1

∗

Rule 12. Laryngealization of bilabial stop

∗

p > h/#

−−−

...(Middle Kanna

da)

Within the historic period of Kanna

da, PSD

∗

p- became h- around the tenth century

AD. The change was completed by the fourteenth century, but both p- forms and

h-forms coexisted, with the unchanged forms slowly declining.

In 931 CE only two

words occur with h- in Kanna

da inscriptions and by the end of the tenth century the

Zvelebil (1970b: 77) lists single and double l, y, v but he does not give any examples for the

double consonants in Proto-Dravidian.