Krishnamurti Bhadriraju. The Dravidian Languages

Подождите немного. Документ загружается.

5.4 Case

studies

191

b. Te. viru-gu

Kui vringa vripka vring-i vrik-t-

∗

vrik-pa)

vr¯ı-va vr¯ı-t-

The Kui forms presuppose the non-past forms

∗

wiri-nk :

∗

wiri-nkk beside the original

root

∗

wiri- (with non-past -w-) in vr¯ı-va. However, as shown by vring-i (past intr), the ve-

lar sufﬁx has been incorporated into the stem. The nominal sufﬁxes also exhibit variation,

e.g. Te. wiriwi ‘abundance’; Ta. virivu n. ‘expansion’, virippu n. ‘opening out’; Ma. viric-

cal (<

∗

viri-tt-al) n. ‘split’. Such nominals are based on underlying tense/voice sufﬁxes.

On the basis of the above discussion, we can also provide a morphological account

for the Proto-Dra

vidian reconstructions of

extended stems derived from

the root

∗

tir-‘to

turn, twist’ etc. (section 5.2 above); cf. (12).

(12) Proto-Dravidian root:

∗

tir-:

a. With vowel sufﬁxes:

∗

tir-i- [3246], tir-a-, tir-u-

b. With -L sufﬁxes:

∗

tir-a-

∗

tir-a-y [3245, 3244]

c. With stop sufﬁxes:

Nonpast Past

Intr Tr Intr Tr

∗

tir-u-ku [3251]:

∗

tir-u-kku [3246]

∗

tir-u-ntu :

∗

tir-u-nttu [3257]

∗

tir-a-nku :

∗

tir-a-nkku [3244]

∗

tir-u-mpu :

∗

tir-u-mppu [3246,

3258]

5.4.5 Examples incorporating past

∗

-nt- in the stem

There are, in addition, several examples of fresh verbal bases in different Dravidian

languages, formed by incorporating the Proto-Dravidian past-tense morph

∗

nt.

Subrahmanyam (1971: 204–6) cites the following examples:

(13) a. PD

∗

iru-(

∗

iru-nt- past intr) ‘to exist, to be’ [480]: Kui r¯ınd- ‘to be stable,

steady’ (earlier: Emeneau 1967a: 391).

b. PD

∗

n¯oy (

∗

n¯oy-nt-) ‘to pain’ [3793]: Ku

r. Malt. nunj- ‘to pain’.

c. PSDr

∗

p¯ay (

∗

p¯ay-nt-) ‘to spring, leap’ [4087]: Kui p¯anj- ‘to ﬂy, leap’.

d. PD

∗

zi (

∗

zi-nt-) ‘to squeeze, press’ [4183]: Ka. pi

du, hi

du;Ko

d-; Tu. pu

diyuni;Te.pi

du; Kol., Nk. pi

d- ‘to squeeze, wring’ :

Ta. Ma. Ka. pi

zi. Kui, Ko

da and North Dravidian also have forms from

∗

zi.Tu.pur˜ncuni, pure˜ncuni ‘to squeeze’ are from

∗

z(i)-ntt- and

∗

pi-

z-ay-ntt-, respectively. Note a restricted Proto-Dravidian sandhi rule

∗

z + nt →

∗

192 Wo r

d formation

e. PSD

∗

kuray (

∗

kuray-nt-:

∗

kuray-ntt-) ‘to be reduced in size’ [1851]:

Kota kornj v.i., korc v.t.; PSD

∗

piri (

∗

piri + nt :

∗

piri-ntt) ‘to separate’

[4176]: Kota pirnj v.i., pirc- v.t.

∗

z-V- (pa

z-V-nt-) ‘to ripen’ [4004] (not included in Subrahmanyam’s

list): Te. pa

du ‘to ripen, mature’, n. ‘fruit’; Kol. Nk. pa

d- ‘to become

ripe’; Pa. pa

d ‘a plant matures’; Go. pa

d-, Ko

da pa

d- ‘to ripen’;

Kuvi pa

du ‘ripe fruit’ (<

∗

z (u)-nt-, past intransitive stem); Tu. parn-

duni ‘to ripen’, parnd¨ı ‘ripe fruit’; Pa. pa

r˜n- ‘to ripen’; Oll. pa

rng- ‘to

become grey (of hair)’. Tamil, Malay¯a

lam etc. have pa

zu; cf. Malto

p¯an-, Ku

r. p¯an (pa˜nj¯a) ‘to ripen’, pa˜njka ‘fruits’. PD

∗

z-i-/-u- with

non-past

∗

-nk-, past

∗

-nt- would

explain all the items in this etymolo

gy.

5.4.6 Further discussion

Let us at this point take stock of our ﬁndings so far, amplifying them with a few additional

comments.

(i) The verb conjugation with inﬂectional morphemes signalling both tense and voice

is an ancient one, still preserved in the Tamil–Ko

dagu subgroup of South Dravidian.

(ii) The non-past intransitive morphs

∗

-nk,

∗

-mp lost their inﬂectional value and

be-

came incorporated as mere voice markers in early Proto-Dravidian.

(iii) In the Tamil–Ko

dagu subgroup the loss of N in NPP led to its merger with PP in

the non-past as in (14).

(14)

Proto-Dravidian Tamil–Ko

dagu subgroup

Non-past Non-past

Intr Tr Intr Tr

∗

pp p pp

∗

mpp — pp

∗

kk k kk

∗

nkk — kk

The merger of

∗

mpp with pp and of

∗

nkk with kk must have led to the analogical

replacement of the intransitive sufﬁxes

∗

mp and

∗

nk by p and k, respectively, in South

Dravidian (Tamil–Ko

dagu subgroup). This did not happen in the case of

∗

nt because

the past sufﬁx

∗

t had a very low functional yield (only nine Old Tamil verbs take it in

their past

tense), whereas

∗

nt had a much higher frequency, and transitive formation by

geminating the ﬁnal consonant of the tense marker affected the past allomorph

∗

-nt-in

many more cases than

∗

t in the Tamil–Ko

dagu subgroup.

Emeneau (1967a: 366, 388–90) and Subrahmanyam (1971: 95–9) consider the system an inno-

vation within South Dravidian.

5.4 Case

studies

193

(iv) Palatalization of a dental stop (single or double) after

∗

y/i is witnessed in many

languages, requiring its reconstruction for the Tamil–Ko

dagu subgroup, for Pre-Parji,

and for Proto-South Central Dravidian (see the next section), and also perhaps for Ku

rux–

Malto. Apparently, the sound change was developing dialectally within Proto-Dravidian

itself.

5.4.7 Past

∗

-c incorporation in Telugu

Many Old Telugu verbs ending in -cu do not have corresponding palatal formatives

in other language groups, e.g. ¯e

d-cu ‘to weep’, na

da-cu ‘to walk’, nil-ucu ‘to stand’,

ra-cu ‘to forget’. In the conjugation of these verbs, the formative -cu occurs in the

past inﬂection (also in the durative) alternating with w in the non-past, as in nil(i)-ci

‘having stood’, nil(a)-w-an (inf.) ‘to stand’.

I have shown elsewhere (Krishnamurti 1961: §§2.84–7, 162–3; §§2.43–2, 45, 14–15.2)

that the c was incorporated into the base from an old past sufﬁx

∗

-cci which occurs as past

participle morph in the other South Central Dravidian languages (§§7.4.1.5, 7. 7. 1. 2).

Similarly c¯eyu ‘to do’ : c¯e-si ‘having done’ has a -si traceable to

∗

-ci. Because of the

wider occurrence of -i as past participle sufﬁx in verbs like win-i ‘having heard’, ka

t-i

‘having built’ etc. which goes back to PD

∗

-i,

∗

-c- and

∗

-cc- have been reanalysed as

parts of the stems; hence

∗

c¯e-si → c¯es-i,

∗

nil-ci → nilc-i. The -w- variant is a reﬂex

of Proto-Dravidian non-past

∗

-p-[-w-]. I further have shown that a similar incorpora-

tion of the c of

∗

ci into the verb bases can be suspected in many other languages of

South Central and North Dravidian (165). Note that Ku

rux–Malto and Brahui retain

a Proto-Dravidian sufﬁx

∗

cc as past-tense marker, which does not result from palatal-

ization of

∗

tt (165), e.g. Te. p¯aru, par-acu ‘to ﬂee’: Malto par-ce ‘to run away’ [4020;

cf. 3963].

In some cases, where Old Telugu -cu/-ncu were used as transitive markers (alternating

with non-past -pu/-mpu) they could be traced back to the palatalized past sufﬁx

∗

-tt/

∗

-ntt

following a front vowel or y as in

di-nc/

di-mp- ‘to lower’, k

¯acu /k

¯apu (<

∗

k¯ay-ntt/

∗

k¯ay-

mpp) ‘boil’. Later on, -cu was generalized as a transitive marker and extended to stems

of different underlying structure, e.g. k¯alu intr ‘to burn’ : k¯al-cu tr ‘to burn’. The non-

past marker -pu/-mpu was analogically replaced in entire paradigms by the

original

past-transiti

ve marker converted into a transitive suf

ﬁx with the

loss of tense meaning.

5.4.8 Additive

∗

-tt and

∗

-pp as transitive–causative markers

Something similar must have happened in the languages of South Dravidian I. The past

and non-past transitive markers of stems of Tamil Class IV verbs, reconstructible as

∗

-(n)tt- and

∗

-(m)pp-, must have occurred in syntactic contexts where they could be

interpreted as mere

transitive markers added to the corresponding intransitive stems.

Thus R. Kothandaraman identiﬁes three types of Old Tamil stems derived from the

194 Wo r

d formation

conjugation in

∗

nt :

∗

ntt, and these stems reﬂect three Pre-Tamil stages of change in their

syntactic functions; cf. (15). In Type I both sets of forms are used as past participles, in

their original intransitive–transitive function. In Type II the ﬁrst form functions as a past

participle in intransitive, while the second one has become a new verbal base to which

-ttu is added as a transitive marker, having lost its tense meaning. A new past participle is

formed from this form, as in v¯a

z-tt-i from Ta. v¯a

z-ttu ‘to felicitate, bless’, tr of v¯a

z. This

development helps to explain how

∗

tt and

∗

pp came to develop into transitive sufﬁxes

within Proto-Dravidian. In Type III, Pre-Tamil or Proto-South Dravidian totally lost the

inﬂectional

meaning of

∗

nt :

∗

(n)tt, and the stem was remade retaining only the voice

difference. This last type is apparently the latest stage in the evolution of formatives,

still within Proto-Dravidian in the case of certain forms.

(15)

Tamil

Past intr Past tr

Type I

∗

m¯ey ‘graze’ m¯ey-ntu m¯ey-ttu [5093]

Type II

∗

w¯a

z ‘live, ﬂourish’ v¯a

z-ntu v¯a

z-ttu (only tr) ‘to

felicitate, bless’ [5372]

Type III

∗

tir- ‘be changed’ tiru-ntu tiru-ttu (formatives

mark only intr–tr)

(tirunt-i)(tirutt-i-) [3251]

‘be changed’ ‘to rectify’

Extended stems with

∗

nk :

∗

nkk,

∗

mp :

∗

mpp must, then, go back to a much deeper

chronological stratum within Proto-Dravidian, since none of the descendant languages

preserves

∗

(n)k :

∗

(n)kk in the non-past. Telugu provides indirect evidence for

∗

mpp

(non-past tr) presupposing a corresponding

∗

mp (non-past intr). Therefore, wherever

these sufﬁxes are found in extended stems their history is to be traced by extending the

logic or parallel developments of the dental series.

Alveolar and retroﬂex series, wherever they occur (cf. Ta. curu

l ‘to curl up’, past

curu

t (v.i.) : curu

t (v.t.) <

∗

curu

l-nt-:

∗

curu

l-ntt), are to be taken as sandhi variants of

the dental sufﬁx. Note that curu

tu belongs to Type II above; it acts as a remade stem

only in the transitive.

There is no evidence to set up an independent palatal series of (c):˜nc : ˜ncc as forma-

tives for Proto-Dravidian since this series is derivable through palatalization of dentals

following front vowels – a widely inherited dialectal change within Proto-Dravidian.

We have noticed that, when the tense signiﬁcation is lost,

∗

tt and

∗

pp came to be inter-

preted as mere transitive sufﬁxes within Proto-Dravidian. Tamil generally has pp after

roots of Class IV ending in i and y, and tt after others. Other languages have general-

ized one or the other through a combination of phonological and morphological criteria.

What is interesting is that

∗

tt,

∗

pp,

∗

kk in derived nouns carry transitive meanings, while

5.5 Ear

lier studies on

stem formatives

195

∗

t/nt,

∗

k/nk and

∗

p/mp carry intransitive meanings. Consider the following examples, e.g.

Ta. a

zi (-v-, -nt-) ‘to perish’, (-pp-, -tt-) ‘to destroy’ → Verbal nouns: a

zi-mpu ‘evil deed’,

zi-vu ‘destruction’ (intr.), a

zi-ppu ‘destruction’ (tr.); Ma. a

zi-yuka ‘to be expended,

etc.’ → a

zivu (intr.) ‘expense’, a

zi-cc-al (tr.) ‘expenditure’; Inscr. Te.

zaccu (<

∗

zi-cc- <

∗

zi-ntt-,

∗

zi-mpp-) v.t. ‘to destroy’ →

da-pp-i ‘destruction’ [277]; Ta. ari ‘to know’ →

ri-vi (intr.) ‘knowledge’, arivai ‘wisdom’ [314] (see section 4.5.8.5).

No Dravidian language preserves PD

∗

NPP as such, but Kumaraswami Raja (1969b)

has conclusively shown that such a reconstruction is warranted by the correspondence

NP:PP

in different Dravidian languages,

such as Ka.

tu ‘eight’: T

a., Ma.

tu (<

∗

tu), Ko. e

t,To.¨o

t, Kod. ¨e

t [847]. The NP sequences are generally attested in Telugu and

Kanna

and occasionally in North Dra

vidian, where voiced and voiceless

stops contrast

after homorganic nasal; cf. e.g. Ka. tirumpu ‘to cause to go round’ : Ta. Ma. tiruppu id.

∗

tir-umpp-); Te. pe

ti ‘female of animal’ : Ta. pe

tai, Ma. pe

ta (<

∗

t-ay). The

solution provided by

Kumaraswami Raja has a far-reaching

effect on our understanding

of a number of problems of comparative Dravidian morphology.

5.5 Earlier studies on stem formatives

After Caldwell there was not any serious study for a long time on how stem forma-

tives evolved in Dravidian. Krishnamurti has given the comparative distribution of

-V-L, -V-(N)P, and V-(N)PP among different subgroups. He proposed that disyllabic

stems in -V + liquid r, l,

l tend to be mostly intransitive thereby giving these an erst-

while grammatical function still preserved synchronically in Ku

rux and Malto (see

section 5.3). He also suggested that a former past morph

∗

-ci/

∗

-cci got incorporated into

certain stems through reanalysis as formative

∗

cc + past i. Emeneau ([1967b/1994:

99–100) considered that

∗

NP :

∗

NPP found in SD I minus Kanna

da and Tu

lu was an

innovation in South Dravidian. But I have given arguments why this type of inﬂection

must be a retention and not an innovation (Krishnamurti 1994a: xxi). The main argu-

ment was that it was unusual for innovating a combination of such complex grammatical

features as tense and voice in the same markers rather than losing one of these at a time

by a process of simpliﬁcation. One set of languages has retained voice and lost tense, i.e.

most of SD I and the Kui–Man

da subgroup of SD II. Loss of both these categories occurs

in Kanna

da, Tu

lu and Telugu, which have a number of lexical items without matching

intransitive–transitive pairs. The progression of change was tense–voice > voice > Ø

(Krishnamur

ti 1994a: xxiv).

In another article, Emeneau (1975, repr. 1994) discusses at length the formation of

plural action stems in Kui–Kuvi with -k-, -p-, -v-, -b- and reconstructs this phenomenon

back to Proto-Dravidian and treats it as a retention in Kui–Kuvi, etc. I gave reasons to

treat these as only relics of Proto-Dravidian non-past markers that have acquired a new

meaning in a subgroup consisting of Kui–Kuvi, with traces in Ko

da–Pengo–Man

196 Wo r

d formation

(Krishnamurti 1994a: xxii–xxiv, [1997a: §7.2] 2001: 302–4). Winﬁeld (1928) classiﬁes

verbs into four conjugations in terms of the augment they take in the formation of the

present participle, namely 1st: -k, 2nd -pa, 3rd -v, 4th -b. When the same augments are

added uniformly to all bases, they denote plural action, but are inﬂected like Class 1. This

transition provides the clue for these consonantal elements (which derive from Proto-

Dravidian non-past markers) signalling a plural action in Kui and being frequentative

markers in the other members of this subgroup.

5.6 Stem formatives in nouns

If all instances of

∗

-(V)-NP,

∗

-(V)-NPP added to verbal roots originally referred to

tense and voice, which meanings they had lost in the course of evolution, how can we

interpret nouns which cannot be traced to any extant verbs, e.g.

∗

mar-u-ntu ‘medicine’?

Disyllabic and trisyllabic nouns of this kind can mostly be related to verbs, and we also

showed the meaning of transitivity surviving in derived nouns. Since the incorporation

of the tense–voice morphs into the stem was a process which was going on in the Proto

language from an indeﬁnite period, the absence of formal marking between verbal and

nominal bases could lead to the formation of certain nominal stems from other nouns

following the same phonological pattern, e.g. mar-u-ntu from mar-am ‘tree’. The best

we can do at this point is to call it a ‘phonological analogy’. While going through the

DEDR one does not come across a large number of such underivable stems of two or

three syllables. What is proposed here is that, at the level of roots, both nouns and verbs

are distinct grammatically, but with identical phonological structure, e.g.

∗

pal ‘tooth’,

∗

p¯al ‘milk’,

∗

kal ‘stone’,

∗

k¯al ‘leg’; in some cases, it appears that the noun and verb are

semantically and phonologically related

∗

n ‘eye’:

∗

k¯a

n [kaH

n] ‘to see’,

∗

kay ‘hand’,

∗

key ‘to do’, etc. This kind of pattern equivalence between nominal and verbal stems of

different sizes is the basis of what I am calling ‘phonological analogy’, which must have

given rise to forms like mar-u-ntu ‘medicine’ at a deeper chronological stratum within

Proto-Dravidian. The absence of a large number of such nominal stems of two or three

syllables with

∗

-NP,

∗

-NPP as the ﬁnal syllable, not related to any extant verbs, is the

basis of the authenticity of the above arguments.

The process of gemination of a ﬁnal P (P → PP) had two functions in Proto-Dravidian:

(i) it formed the transitive stem; (ii) it also derived a nominal stem from adverbs or

adjectives. Consequently, we ﬁnd the transitive stems to have an identical phonological

structure to derived nominals in a considerable number of cases. It is this coincidence

in structure that led Caldwell to say that NP: PP (our

∗

PP,

∗

NP:

∗

NPP) originally

marked nouns and secondarily they became verbal. The opposite is the truth, which is

borne out by the following arguments and examples.

(i) Intransitives ending with NP do not function as nominals, but transitive-like stems

function also like nominals.

5.7 Phonolo

gical chang

es in Proto-Dr

avidian roots

197

(ii) The intransitive nominals are derived by adding noun-forming sufﬁxes to intran-

sitive verb stems, whereas the transitive stems can occur as nominals without further

additions.

(iii) The transitive stems can also take other noun-forming afﬁxes. In that case, it is a

parallel process of noun-formation by adding derivative morphs to transitive stems.

(iv) That the transitive stems look identical to nominal stems is a structural coincidence

and not a semantic or grammatical shift. Primarily the stems are verbal.

5.7 Phonological changes in Proto-Dravidian roots

A proto-Dravidian root can be either free or bound. Free roots can be free forms or words;

in the case of verbs, most verb roots may also occur without any additions as minimal

utterances in the imperati

ve sg. The degree of freedom

from high to low is indicated by

the canonical forms as (C)

VC< (C)VC < (C)

V <(C)V-; the last one is always bound

(see TVB: §2.113, 2.114; pp. 173–4). Numerically roots which are free forms are many

more than the bound ones (see section 5.1).

The syllable structure of the Proto-Dravidian roots and stems is best preserved in Early

Tamil, and to some extent in Malay¯a

lam, a west coast dialect of Tamil which became

an independent language by about the ninth century. The reconstructed roots undergo a

number of changes, which

have been treated under different

headings in chapter 4. Some

alternations go to Proto-Dravidian itself, namely consonantal changes in section 4.3;

contraction of (C

-V

> (C)

-, e.g. pey-ar > p¯e-r ‘name’,

∗

kic-ampu >

∗

kiy-

ampu > k¯ı-mpu/k¯e-mpu ‘a yam’ in section 4.3.2.1; a long root vowel becoming short,

when followed by a formative V (=V

∗

¯ır-/

∗

ir-V- ‘two’ (adj), in section 4.3.2.2. These

need not be repeated here.

There is a set of sound changes that introduced quantitative changes in the root

syllable in different subgroups and individual languages. These have been described

both comparatively and typologically by Krishnamurti earlier (1955). I summarize the

ﬁndings as follows:

1. (C)V

y/w/k-V

> (C)

-. This is the same as the Proto-Dravidian rule mentioned

above, except that it continues into the historical period of individual languages. It

is therefore (see Krishnamurti 1955:§8) not reconstructed for a distant past, but for a

much later time frame, e.g. T

¯ur-cu ‘to breathe’, ¯ur-pu ‘breath’ <

∗

uy-r-cc-,

∗

uy-ir-pp-;

the older stage has -c- [645]; only Telugu shows contraction. Toda has contraction of

syllables with the loss of a liquid

∗

r or

∗

2. Loss of a semivowel y/w as C

can lead to compensatory lengthening of the pre-

ceding vowel when followed by

∗

-NP/-NPP. In other words, a root of (C)VC- type

Caldwell maintained that ‘stability in the root-vowels is the rule, and change is the exception’

(1956: 217). He did point out some of the changes which have been handled in recent years

either by morphophonemic changes within Proto-Dravidian or by positing a laryngeal.

198 Wo r

d formation

becomes (C)

V, e.g. Te. t

ugu (<

∗

tuy-nk-), Ta. t¯u-˙nku, Ma. t¯u-˙n˙nuka, Ka. t¯u-gu,Tu.t¯u-

˙nkuni [3376a,b]; in some cases the root-ﬁnal -y is lost without compensatory lengthening,

Te. bo-nku ‘to lie’ (<

∗

poy-nkk-), Ta. Ma. poy ‘to lie’, po-kkam ‘falsehood’ (<

∗

poy-

nkk-am) [4531, 4559 no need to put in two entries;

∗

poy- is from older

∗

poc-] (see

Krishnamurti 1955: §§16,17).

3. PD

∗

Vy/

∗

Vy are both represented as

Vy in Old Telugu in the case of both verbs

and nouns. In other words, contrast of length is lost before -y in root syllables in Telugu

(Krishnamurti 1955: §14, fn. 20, Subrahmanyam 1970a).

4. Metathesis and vow

el contraction led to disyllabic

and trisyllabic bases becoming

monosyllabic and disyllabic, respectively, in SD II languages (see section 4.4.3),

∗

mar-

an > mr¯a-n in Old Telugu, Ko

da, Kui–Kuvi (see ety. 35), Te. kro-tta <

∗

z-utt-V [2149]

(see section 4.5.7.3). (Krishnamurti 1955: §§18–20, TVB: §§1.121–59.)

5. In PD trisyllabic stems, V

is lost, leading to three developments: (i) assimilation

of the root-ﬁnal liquid to a following -PP/-B. This change will give rise to disyllabic

stems, e.g.

Te.

ceppu ‘sandal’ (<

∗

cer-pp- <

∗

ker-pp-); cf. Ta. cer-uppu, Ma. cerippu,

Ko. kevr (<

∗

ker-v-), To. kerf; CD: Kol. Nk. kerri, Nk. (Ch.) kerri, kerig; ND: Ku

kharp¯a.; Gondi and Parji borrowed the word from Pre-Telugu before assimilation of the

consonants: Go. serpum, sarpum, etc. Pa. cerup, cerpu [1963]. The syllable reduction

with the loss of the unaccented vowel runs through all subgroups [1963]. Depending

on the pre-assimilation stage, Telugu and Kanna

da developed geminate voiced stops

also as a result of this change, Te. taggu ‘to be reduced’, Ka. taggu, targu, ta

zgu id.

∗

z-u-nk-, with loss of -u-, -n-, and voicing of k to g); cf. Ta. Ma. t¯a

z ‘to be lowered’

[3178], see section 4.5.8.3 (for a detailed treatment, see Krishnamurti 1955:§§7–10, TVB

§§1.173–85); (ii) metathesis of (C)VL- to (C)LV- in SD II languages, Te. brungu ‘to be

immersed’ (<

∗

zu-ng- <

∗

z-ng- <

∗

z-u-nk-), Kui b

dga id. [4993]; (iii) loss of

the root-ﬁnal liquid, cf. Ta. mu˙nku, Ma. mu˙n˙nuka ‘to sink, plunge’ (<

∗

z-nk- <

∗

u-nk-), Kol. Nk. mu

g-, Kui, Pe. Man

da munj- [4993].

6. Some grammatical classes have alternation between short and long vowels: (i) verbal

roots hav

e short vowels and derived nouns have long vowels, e.g.

∗

t-u v.i. ‘to perish,

to be spoiled’:

∗

k¯e

tu ‘damage, evil’ [1942]; (ii) in personal pronouns the nominatives

have long vowels, but the oblique stems have short vowels, e.g.

∗

t¯an /

∗

tan- ‘self ’; (iii)

several irregular verbs have such length alternation:

∗

war-/

∗

w¯a- ‘to come’,

∗

tar-/

∗

t¯a-

‘to give to the 1st or 2nd person’ etc. Caldwell mentioned some of these but

could not

provide any solution (Caldwell 1956: 210–17). Until recently, multiple reconstruction

of roots was the solution adopted by Dravidian comparativists. Recently Krishnamurti

explained all such irregular alternations in terms of a reconstructed laryngeal

∗

H (see

Krishnamurti 1997b; for a summary see section 4.5.7.2.3). A laryngeal

∗

H was lost and

it simultaneously lengthened a preceding vowel when the output was a free form, but

was lost systematically when the output was a bound form, thus PD

∗∗

taHn >

∗

t¯an/tan-

nom/obl,

∗∗

keH

t-u >

∗

k¯e

t-u/

∗

t-u- n./v.,

∗∗

waH-r- >

∗

w¯a/

∗

wa- imper./past. Early Tamil

5.8 Derivational

suf

ﬁxes 199

preserved a few relics of a PD laryngeal in deictic bases: a

h- ‘remote’, i

h- ‘proximate’,

the number word ‘ten’ pa

h-tu and about half a dozen lexical items of which three have

cognates with lengthened vowels (see Krishnamurti 1997d: 147–50).

5.8 Derivational sufﬁxes

Each of the Dravidian languages has a stock of noun-forming and verb-forming afﬁxes,

which will be dealt with in the respective chapters. Derivational sufﬁxes which are

reconstructible are not many and are given below.

5.8.1 Deverbal nominals

(a) By adding

∗

-ay to monosyllabic v

erb roots,

∗

wil ‘to sell’:

∗

wil-ay ‘price’ [5221: SD I

and Te.], Ta. vila-ai, Ma. vil-a, Ka. bel-e,Te.wel-a ‘price’;

∗

t- ‘to tie, bind’:

∗

t-ay

‘dam’ [1147: SD I, SD II, CD, ND]; Ta. naku ‘to laugh’: nak-ai n., Ka. nagu ‘to laugh’:

nag-e ‘laughter’ [3569].

(b) By geminating the ﬁnal stop of the root in disyllabic stems or the formative of

stems of two or more syllables, e.g. root-ﬁnal stop:

∗

¯a

tu ‘to play’:

∗

¯a

t-u ‘playing, a

game’ [347],

∗

a˜ncu ‘to fear’: accu (<

∗

a˜ncc-) n. ‘fear’ [55]; also by adding -am to the

stem with the ﬁnal stop doubled: ¯a

t-am ‘game, dancing’, acc-am (<

∗

a˜ncc-am) n. ‘fear’

(Ta.–Ma.).

∗

t-u ‘to

perish’ :

∗

t-al ‘evil’ (SD I);

∗

k¯u

tu v.i. ‘to join’, k¯u

tu ‘to join’ v.t., n. ‘mixture’ (see (a)):

k¯u

t-al ‘joining (intr.)’, k¯u

t-al ‘uniting’ (tr.), also

∗

k¯u

t-am ‘union’, see (SD I and Te.)

[1882];

∗

enku ‘to remain, be left over’, + al →

∗

enk-al ‘left-over food’: Ta. e˜ncu ‘to

remain’, e˜nc-al ‘defect’, eccam (<

∗

e˜ncc-am), Ma. ecc-il, Ka. enj-al,Ko

d. ecci (l-loss);

SD II: Te. eng-ili n. ‘left-over food’, Pe. Ku

r. Malt. have verbs derived from

∗

enk-

[780].

(d) By adding

∗

-t-al/

∗

-tt-al which consists of two noun-forming elements

∗

-t/-tt + al;

also similarly

∗

-t-am. These are added to roots ending

∗

t, e.g.

∗

¯o

tu v.i. ‘to run’,

∗

¯o

v.t. ‘to make to run’: Ta. ¯o

tu, ¯o

t-am ‘running’ (<

∗

¯o

t + t-, ¯o

t + t-am) (SD I and II). An

agentive noun is derived by adding -i to transitive stems, e.g. ¯o

t-i ‘captain of a ship’ (Ta.

Ma.); Ka. k¯u

du ‘to join’: k¯u

t-am ‘union’ (<

∗

k¯u

t-t-am), p¯a

du ‘to sing’: p¯a

t-am ‘song’.

(e) By geminating

the post-nasal stop of a formative in stems of two or more syllab

les.

Some languages lose the preconsonantal nasal (PP <

∗

NPP) and others degeminate the

voiceless stop (NP <

∗

NPP); Ta. may-a˙nku ‘to be confused’: may-akku, may-akk-am

‘confusion’ (<

∗

may-an-kk-), Ta. a˜ncu ‘to fear’ : accu ‘fear’ (<

∗

a˜ncc-) (SD I and Te.).

(f) By lengthening the root-vowel, sometimes in combination with other processes,

e.g.

∗

tu ‘to perish, be spoiled’:

∗

ta,

∗

k¯e

tu,

∗

t-al,

∗

tu-ti ‘ruin, loss, damage’

(mainly SD I and Te.).

da has unexpected palatalization of the verb, enz- ‘to be left over’, es- v.t. ‘to leave or save

food’.

200 Wo r

d formation

(g) -am added to an intransitive or transitive verb stem,

∗

c¯o

t- ‘to run’,

∗

c¯o

t-am ‘boat’

[2861] > Ta. Ma. ¯o

t-am, Ka. Tu. ¯o

d-a, SD II: Te. ¯o

d-a;(>lw Go. Pa. ¯o

d-a).

Ka. s¯ol

‘to be defeated’: s¯ol-am ‘defeat’.

(h) Multiple noun formatives: am + t + am → antam, opp-antam ‘agreement, con-

tract’ from oppu ‘to agree’, in Ta. Ma. Ka. Te. opp-andam id. Note that oppu is a

reanalysed stem from the root

∗

o- incorporating the non-past morph -pp, which occurs

in all subgroups as such [924]. It also functions as a nominal stem o-ppu ‘uniformity,

suitability’; with the addition of -am, opp-am ‘comparison, likeness’ (SD I and Te.).

With a different non-past sufﬁx incorporated: o-kk- (cf. Ma. okkuka ‘to be like’, Ku

okk- ‘to tally, agree with’),

∗

okk-al ‘relatives, kinsmen’ (SD I) [925]. Another complex

noun formativ

eis

∗

t + al + ay →

∗

talay, cf. T

oppu-dala ‘agreement’, Ka. tavu ‘to

decrease’: tavu-dale ‘destruction’ (see section 5.6 (3)).

(i)

∗

-(i)kay. Ka. b¯ey ‘to burn’: b¯e-ge ‘ﬁre, ﬂame’, pa

nu ‘to make’: pa

n-ige ‘deco-

ration’, tu

du ‘to wear’(<

∗

du):to

d-ige ‘ornament to wear’

(j)

∗

-(i)kk-ay: Ka. alasu ‘to be weary’: alas-ike ‘weariness’, ir- ‘to be’: ir-ke ‘an abode’,

agal ‘to be separated’: agal-ke ‘separation’; Te. k¯oru ‘to wish’ : k¯or-(i )ke ‘a wish’, p¯unu

‘to undertake’, p¯un-(i )ke, p¯un-(i)ki ‘perseverance’, manu ‘to live’: man-iki ‘living’.

5.8.2 Deadjectival nominals

(k)

∗

-may.Ta.p¯er/per-u adj ‘big’, peru-mai ‘abundance’, Ka. per-me ‘increase, great-

ness’, hem-me ‘pride, insolence’; Te. p¯er-mi ‘greatness, superiority’ [4411].

5.9 Compounds

Here we deal only with reconstructible compounds, sequences of at least two roots, which

function as words in Proto-Dravidian, or at least in one of the subgroups. Consequently,

compound-like constructions, which are not attested by at least two languages, have

been eliminated. All the items are taken from DEDR. Several of these present problems

of analysis, both phonological and semantic. However, we can still see the patterns that

were established in the family before the individual languages emerged and enlarged the

inventory.

Emeneau, in an unpublished paper,

has discussed two items: 5496b we

ney ‘butter’

(

∗

l +

∗

ney, lit. ‘white ghee/oil’) and 4460a

∗

poC-ku

z-V ‘navel’ (‘hollow of stomach’),

of which the second member is 1818

∗

k¯u

∗

z-V ‘pit, hollow’, and the ﬁrst part has

been identiﬁed as 4460b

∗

t- ‘navel’.

He has also dealt with six items mainly conﬁned

to individual languages.

cf. Skt. ho

da ‘boat, rafter’. Note Skt. h- preserves the intermediate stage in the sound change

c- >

∗

s- > h- > ø [1039]. The other ‘run’ word in [1041] is cognate with the other two.

I am indebted to Professor Emeneau for loaning me a copy of the paper entitled ‘Some Dravidian

noun compounds’.

Most languages point to a reconstruction with two voiced stops or at least one,

∗

d-/

∗

d-;

only Malay¯a

lam shows p-. If the ﬁrst part should mean ‘stomach’, 4494

∗

tV ‘stomach’ would