Krishnamurti Bhadriraju. The Dravidian Languages

Подождите немного. Документ загружается.

5.2 V

ariability of f

ormative suf

ﬁxes 181

5.2 Variability of formative sufﬁx

A Dravidian Etymological Dictionary of 1984 [DEDR] has scores of entries, which lead

us to reconstruct primary roots as well as extended stems for Proto-Dravidian (mainly

of verbs but also of nouns), e.g.

(1) a.

∗

tir-a-y (-p-/-mp,-nt-) v.i. ‘to roll’; (-pp-/-mpp-, -ntt-) v.t. ‘to roll up’; n.

‘wave, screen’,

∗

tir-a-nku v.i. ‘be curled up’,

∗

tir-a-nkku v.t. ‘to shrivel’

[3244].

∗

tir-a-

l (- p-, -

t-) v.i. ‘to become round’, (-pp-, -

t-) v.t. [3245].

∗

tir-i-(-p-, -nt-) v.i. ‘to turn’; (-pp-, -ntt-) v.t. ‘to turn’;

∗

tir-uku v.i.,

∗

tir-u-kku v.t. ‘to twist’;

∗

tir-u-mpu v.i.,

∗

tir-u-mppu v.t. ‘to twist, turn’

[3246].

d. tir-u-ntu v.i. ‘to be corrected, be repaired’;

∗

tir-u-nttu v.t. ‘to correct,

rectify’ [3251].

The Proto-Dravidian root obviously must be

∗

tir-, meaning ‘turn, roll, twist, change

shape’ → ‘correct’, etc.

The formatives occur in two

layers. The

ﬁrst la

yerisV

= i, a, u;

and the second layer, either a sonorant (L) as in y,

l; or a simple or geminated stop ±

homorganic nasal: P as in

∗

ku;PPasin

∗

kku;NPasin

∗

nku,

∗

ntu,

∗

mpu; NPP as in

∗

nkku,

∗

nttu,

∗

mppu. The bound root

∗

tir- patterns with autonomous roots like

∗

key ‘to do’,

∗

wil

‘to sell’,

∗

cal ‘to go’, etc. which enter into inﬂection without any formative being added.

It is well known that the pairs of sufﬁxes

∗

k :

∗

kk,

∗

nk :

∗

nkk,

∗

nt :

∗

ntt,

∗

mp :

∗

mpp

synchronically encode an intransitive:transitive distinction in both South Dravidian I

and South Dravidian II. But why should there be so many series of sufﬁxes fulﬁlling

the same function? And how do sets of related forms such as those in (1) above arise?

These are the problems that I want to address in this chapter.

The morphophonemic changes

that the roots and suf

ﬁxes undergo when

followed by

different formatives have been discussed in section 4.3.

Through comparison of the cognates of (C)

VC-type stems involving sufﬁxes in L

or P within reconstructed Proto-Dravidian, it is possible, in some cases, to identify an

underlying root of the type

∗

(C)

V-. For instance, the stems

∗

k¯a-y-/

∗

k¯a-,

∗

k¯a-nku (v.i.) :

∗

k¯a-nkku (v.t.),

∗

k¯a-ntu (v.i.) :

∗

k¯a-nttu (v.t.),

∗

k¯a-mpu (v.t.) : k¯a-mppu ‘be(come) hot,

burn, to dry up, etc’ [1458], and

∗

k¯a-

l ‘to burn, ﬂame’ [1500] could possibly be related

to words meaning ‘black, burnt, etc.’, such as

∗

k¯a-r :

∗

kar-V

‘be

scorched, burnt black,

black’ [1278], with

∗

k¯a- as the ultimate root and the semantic development ‘burn’ →

‘burn black’ → ‘black colour’ → ‘coal’, etc.

What follows is a summary of my recent paper on formative sufﬁxes (see Krishnamurti 1997a).

The sufﬁxes in parentheses refer to non-past and past markers, respectively. The numbers in

square brackets refer to the entries in DEDR.

182 Wo r

d formation

5.3 Primary derivative sufﬁxes as earlier inﬂectional sufﬁxes:

the hypothesis

Based on a critical study of many etymologies of the type (1) above, I venture to propose

that primary derivative sufﬁxes arose through the incorporation of inﬂectional sufﬁxes

into the stem, and that this development took place in several stages, largely within

Proto-Dravidian.

At a very early stage within Proto-Dravidian, sonorant sufﬁxes of the L type (l,

l, r,

w, y) were added to (C)

V- or (C)VC-V-stems to form extended intransitive/middle-

voice stems. This assumption is based on the observation that verb stems ending

in sonorant sufﬁxes tend to be intransitive in the descendant languages. Forms with

these sufﬁxes are preserved intact in the literary languages of the south, namely Tamil,

Malay¯a

lam, Kanna

da and Telugu. At a later period, -L, -V-L lost their identity as gram-

matical elements and became incorporated into the preceding stems, as in (1a, b) above

(Krishnamurti 1961: 146–7; Emeneau 1975: 2–3).

Proto-Dravidian also had a very early stage in which P sufﬁxes were added to primary

roots, and later to extended stems with -L and -V-L. Only a subgroup of South Dravidian

consisting of Tamil, Malay¯a

lam, Ko

dagu, Toda, Kota and Ba

daga preserves this stage of

development in verb conjugation, e.g. Ta. a

z-i (-v-, -nt-) v.i. ‘to perish’, a

z-i (-pp-, -tt-)

‘to destroy’ [277]; compare this with a

z-u-ku ‘to rot’, a

z-u-nku ‘to be spoiled’ [284].

These P-sufﬁxes signal both tense and voice. Dental vs. non-dental indicates past vs.

non-past; simple (N)P signals intransitive, and geminate (N)PP, transitive:

(2) Non-past Past

Intransitive

∗

Transitive

∗

mpp

∗

nkk

∗

ntt

The non-past paradigms include present, future, aorist (habitual), inﬁnitive, impera-

tive, negative, etc. Within the non-past, there must have been a morphological contrast

between the labial and velar series, but the contrast tended to be blurred later.

The next stage was the incorporation of tense/voice sufﬁxes into the preceding stems,

with loss of tense meaning but preservation of the voice distinction. The latter is

I made this observation by comparing -(V)L-sufﬁxes of Telugu verbal bases with those of the

other Dravidian languages. I pointed out that the use of -r- as an intransitive reﬂexive marker

occurs synchronically in Ku

rux (kam- ‘to make’: kam-r- ‘to be made’) reﬂecting the Proto-

South Dravidian situation. Since such a construction also occurs in Malto, Emeneau proposed

the reconstruction of -r- as intransitive marker in Proto-Dravidian and suggested that it was

incorporated into verb stems in all Dravidian languages except in Ku

rux and Malto (1975: 2–3;

for Malto, see also Mahapatra 1979: 144–52). I have not gone beyond this point in investigating

the

∗

V-L-sufﬁxes of Proto-Dravidian.

5.3 Primary

derivatives

as earlier in

ﬂectional

suf

ﬁxes 183

preserved mostly in disyllabic and trisyllabic stems, with NP: (N)PP indicating an in-

transitive:transitive alternation in most of South Dravidian. Traces of this alternation

are found in Kanna

da and Tu

lu.

South Central Dravidian also preserves this stage,

with the exception of Telugu and Gondi; and some traces exist in all other subgroups

and languages (Subrahmanyam 1971: 52–4). As a consequence of this change, the con-

trast

∗

NP:

∗

NPP (or its reﬂexes NP: PP, (N)B: (N)P, etc.) has come to signal only an

intransitive:transitive distinction.

The languages which have lost the paired intransitives and transitives of the above type

have added the reﬂexes of

∗

-tt or

∗

-pp as transitive–causative markers, e.g. Te. jaru-gu

‘to slide’ : jaru-pu ‘to move’, tiru-gu v.i.‘to revolve’: tri-ppu v.t.‘to revolve’. It is certain

that these were already in use for certain stems in Proto-Dravidian before the emergence

innovative pairs of intransiti

–transitive markers.

Since several languages across subgroups show the additive morphemes -pp,

-tt (> -cc following a front vowel or -y) as transitivizers, they need to be reconstructed

for Proto-Dravidian (sections 7.3.1–.2). Some of the intransitive bases to which these are

added reﬂect simple Proto-Dravidian roots such as

∗

m¯ey-(-pp-) ‘to graze’,

∗

k¯ay-(-pp-/

-tt-) ‘to burn’,

∗

tiHn-(-tt-) ‘to eat’,

∗

n-(-tt-) ‘to drink’ etc. suggesting the antiquity of

the additive morphemes. After the ﬁnal split into subgroups, it is likely that their use was

extended to most of the verb stems in the languages that had lost the (N)P:(N)PP pattern

marking intransiti

ve:transitive. Another factor

contributing to the extensiv

e use of the

additive elements is their structural parallelism (-mp-p- > -pp, nt-t- > -tt) to the ﬁnal

–CCV of paired intransitive–transitive sets which had originally a much wider gram-

matical function; e.g. Old Tamil na

ta- (past intr. nata-nt-, past tr. nata-tt- <

∗

ta-nt-t-),

Mdn Ta. na

ta- v.i. ‘to walk’ : nata-tt- v.t. ‘to make someone walk’ (only additive without

tense meaning).

A ﬁnal, analytic, stage is found in all South Dravidian literary languages. New tran-

sitive stems are derived by the addition of different auxiliary verbs to non-ﬁnite forms

of the main verb, as in Te. wirugu v.i. ‘to break’; wiraga go

t- v.t. ‘to break or snap’

(< wirag-an- (inf) + ko

t ‘to beat’). In some cases, the older sufﬁxal structures coexist

with the ﬁnal analytic stage, but with semantic differences; cf. Mdn Te. k¯al-cu ‘to light

(a cigarette), to burn’ beside k¯ala be

t- ‘to burn something down’.

DEDR 169: Ta. am-u-nku v.i. ‘to sink, be pressed down, crushed’; am-ukku (<

∗

am-unkk) v.t.

‘to crush’; Ka. am-ugu, av-ugu v.i. ‘to yield to pressure’; av-u˙nku (beside am-uku etc.) v. t. ‘to

press or hold ﬁrmly’; Tu. av-u˙nk-uni, av-ump-uni ‘to press down’. Tu

lu here offers evidence for

∗

nkk and

∗

mpp of Proto-Dravidian with loss of the tense contrast. DEDR 240: Ta. al-a˙nku v.i.

‘to be, shaken, etc.’, al-akku v.t. ‘to cause to move, shake’, Tu. al-a˙nk-uni v.t. ‘to agitate, wave’;

al-a˙ng-uni has both intransitive and transitive meanings; DEDR 524: Ta. i

ruku v.i. ‘to become

tight’, i

rukku v.t. ‘to tighten’; Ka. iru˙nku v.t. ‘to compress’ (also ira˙nku, iri˙nku presupposing PSD

∗

irunkk- v.t.); so also DEDR 3246: Ka. tirumpu v.t. ‘to cause to go round’.

184 Wo r

d formation

I assume that the ﬁrst two developments took place at various stages within Proto-

Dravidian, with each successive stage having a wider spread – lexical and areal – than the

earlier one, but with all the three stages still synchronically coexisting. Typologically we

notice a progression from synthetic to analytic in this scenario. Note also that extended

stems of two or three syllables are more numerous than monosyllabic roots in the later

stages of Proto-Dravidian.

The next question is to ﬁnd empirical support for the above proposal and to ﬁnd

missing links in the morphological development of extended stems.

5.4 Case studies

The most transparent etymology which gives immediate evidence for how formatives

came into existence is that of PD

∗

o ‘to be appropriate, to

suit

’ [924]:

(3) SD I: Ta. o (-pp-, -tt-) ‘to be suited, appropriate’, oppu (-i

n-) ‘to agree,

assent’, n. ‘consent, uniformity’, opp-am ‘comparison’, oppu-mai ‘like-

ness’, ovvu ‘to be like, congruous’, Ma. okku-ka ‘to be like’, ott-a ‘equal,

consistent’, opp-am ‘equality, harmony’, Ko

d. o (-pp-, -tt-) ‘to be suitable’,

otta-(-a

d-) ‘to consent’, To. up-(-y-) ‘to be pleased, to agree’, Ko. op-

(-y-) ‘to be acceptable’, Ka. oppu ‘to agree with’, oppa ‘ﬁtness’, opp-ike

‘agreeing’, ommu ‘to concur’, Tu

oppiyuni ‘to ag

ree to

’, otto

n¨ı ‘to ag

ree

’,

ombuni ‘to be suitable’;

SD II: Te. oppu ‘to suit, be agreeable’, n. ‘ﬁtness, beauty’, oppa-g-incu ‘to

entrust’, oppu-dala ‘assent’, -yokka genitive sufﬁx meaning ‘belonging to’,

ommu ‘to suit’, ovvu ‘to agree’, Go. app-(< opp-) ‘to be pleasing’, Ko

op-(op-t-) ‘to agree’, Kuvi ¯op-(-it-) ‘to agree, consent’;

CD: Kol. ovvol ‘good’, Pa. op-ip-(op-it-) ‘to give in charge, hand over’;

ND: Ku

r. okk- ‘to ﬁt in well, agree with’.

One can see that the extended stems o-pp-, o-tt-, o-mp-, o-kk-, o-ww-, etc. have arisen

from a monosyllabic root

∗

o- which still occurs in Tamil and Ko

dagu. The similarity

between the stem extensions and tense morphs is striking.

The etymologies of three frequently used verbs, namely

∗

¯a- ‘to be, happen, become’,

∗

p¯o ‘to go’ [4572] and

∗

l ‘to be’ [697], bear ample evidence to how extended stems

with original inﬂectional morphs have gradually evolved as formatives, having lost their

grammatical meaning:

5.4.1

∗

¯a ‘to be’ [333]

(4) a.

∗

¯a-: Ta. ¯a (past stem ¯an-, ¯ayi-) ‘to be, happen’, ¯a (-pp-, -tt-) v.t. ‘to cause,

bring about’; ¯a n. ‘becoming’, ¯a-m (<

∗

¯a+um) ‘yes’; Tu. ¯a-pini (2neu sg

5.4 Case

studies

185

past ¯a

du); OTe. ¯a (in ¯ayen/ayyen 3 past sufﬁx -en with inserted glide y)

id.; Go. ay- (irregular, 3sg imperfective ¯and; some forms from ¯a), most

dialects have ¯a ‘to be’; Ko

da, Kui, Kuvi, Pengo, Man

da ¯a (¯a-t-) id.

∗

¯aku v.i.:

∗

¯akku v.t.: Ta. ¯aku (¯aku-v-, ¯ak-i) v.i., ¯akku (¯akk-i-) ‘to make’,

¯akk-am n. ‘creation’ (transitive noun), ¯ak-a inf ‘completely’; Ma. ¯aku-

ka v.i., ¯akku-ka v.t., ¯akk-i -kka (<

∗

¯akku-wi-kka- caus inf of tr with loss

of -w-) ‘to cause to make’; Ko. ¯ag v.i. (¯ay-/¯an-; some forms from ¯a-);

¯ak- v.t.; To. ¯ox (¯oy-, ¯on-, ¯o) v.i., ¯ok-(¯oky-) v.t.; Ka. ¯agu (¯an-, ¯ay etc.);

d. ¯ag- (irreg ¯ay-, ¯an-, ¯a) v.i., ¯ak (¯ak-i) v.t.; Te. agu, awu (ay-i ppl,

ay-na adj); Nk. akk- ‘to make’.

∗

¯a- p- v.i. ‘to become’:

∗

¯a-pp- v.t. ‘to make,

etc.

’: Kui ¯ava-(¯a-t-) ‘to

become’, ¯ap-ka (¯ap-ki) plural action stem; Kol. ¯ap-(¯ap-t-) ‘to keep in

a place’; Nk. ¯ap- ‘to keep’. Cf. Ta. ¯a-( pp-) ‘to cause, bring about’.

∗

¯an-/an- v.i.: Tu. (2neu sg past ¯a

du); Kol. an- (irreg, past an

d-, imper

¯an-) ‘to be in place’; Nk. an

d ‘to be’; Br. anning (an-, as-, a-) ‘to be’.

Note that both ¯a- and ¯aku- are used as inﬂectional bases in Old Tamil (PN), as in ¯ak-i

‘if’ beside ¯a-(y)i

n id., ¯ak-um ‘(subj) will

become

’:¯a-m id. (Subramoniam

1962). The

last two forms are also given by Tolk¯appiyam (ﬁrst century AD or BC). It can therefore

be concluded that ¯a- is not simply a contraction of ¯aku- (Israel 1973: 235).

As for the putative sufﬁxes in (4b–d), note that the Old Tamil classics use -k- beside -t-

in aorist, i.e. non-past, constructions (Israel 1973: 145, 193ff.), as in u

n-k-um ‘we drink’,

t-um ‘we see’, varu-t-um ‘we come’ from the roots u

n- ‘to drink’, k¯a

n- ‘to see’ and

varu ‘to come’. (The sufﬁx -um when not following -k-or-t- is said to simultaneously

mark habitual tense and person. In Malay¯a

lam, the suﬁxes -um and -kkum similarly occur

as non-past aorist markers.) Further,

∗

-p- [-w-] occurs as a future-tense marker added

to ¯aku-inOTa.¯aku-pa/-va ‘they will become’ (Israel 1973: 235). That is, it too marks

a non-past structure. Finally the base form in (4d) corresponds to the past stem of the

South Dravidian languages.

Given this evidence, we can interpret the data in (4) as follows. Set (a) naturally

contains the original, unmodiﬁed root. Set (b) is based on an extended stem ¯a-k-ofset

(a), which incorporates an old non-past sufﬁx -k- (intransitive) : -kk- (transitive). This

set shows variation between

∗

¯a and

∗

¯ak in inﬂection. Set (c) is based on another non-past

stem of ¯a, viz. ¯a- p-[¯a-w-] : ¯a-pp-. Set (d) is based on the past stem of set (a), namely

¯an-. All the four sets can be derived from the following reconstructed system of early

Proto-Dravidian.

The inﬁnitive ¯aka is ambiguous, since it can be analysed either as ¯a-ka or as ¯ak-a. Classical Tamil

also has the simple inﬁnitive ¯a < ¯a-a with loss of the short a (Agesthialingom 1979: 94).

186 Wo r

d formation

(5)

Non-past Past

Intransitive

∗

¯a-k-

∗

¯a- p-

∗

¯a-(i)n-

Transitive

∗

¯a-kk-

∗

¯a-pp-

∗

¯a-tt-

As we have seen, the formations in -k(k)- and -p(p)-, as well as the one in -(i)n-, have

become generalized verb stems, losing their tense distinctions. In set (c), however,

∗

-pp-

(> - p-) has retained traces of its voice distinction by serving as a transitive marker in

some of the languages. Only the sufﬁx -tt- is retained as an inﬂectional morpheme in

South Central Dra

vidian.

Further support for the reconstructed formations in (5) comes from the distribution

of forms in the various subgroups of Dravidian; cf. (6). None of the extended forms is

limited to just one subgroup, a fact which precludes the assumption that they originated

in the various

descendant subgroups. The

geographical distribution, combined

with the

data in (4), further suggests that the incorporation of sufﬁxes into the stem began within

Proto-Dravidian, for, again, the use of extended forms with loss of their original tense

and/or voice distinctions cuts across the different descendant subgroups.

(6)

∗

¯a SD, SCD

∗

¯ak-:¯akk- SD; Nk. of CD

∗

¯ap-:

∗

¯app- Ta. of SD; Te., Kui of SCD; Kol., Nk. of CD

∗

¯an Tu. of SD; Go. of SCD; Kol., Nk. of CD; Brahui of ND

5.4.2

∗

p¯o ‘to go’ [4572]

A structural

parallel to

∗

¯a in Proto-Dravidian

is found in

∗

p¯o beside

∗

p¯oku ‘to go’

attested only in South Dravidian and South Central Dravidian.

(7) a.

∗

p¯o :Ta.p¯o (non-past: p¯ov-/p¯okuv-/p¯otuv-, past: p¯on-/p¯oyin-; neg p¯ok-);

Ma. p¯oka;Ko

d. p¯o ( p¯op-, p¯ok-, p¯oc-, p¯oy-); Ka. p¯o, p¯o-li, p¯o-lu ‘state

of going, ruin’; Te. p¯o ( p¯o-yi).

∗

p¯o-k-:

∗

p¯o-kk-: Ta. p¯oku (future: p¯oku-v-, neg. p¯ok-), p¯ok-ai n. (intr)

‘departure’; p¯okku (p¯okk-i-) ‘to cause to go, send’; p¯okk-am n. (tr)

‘causing to go; exit, way’; Ma. p¯oka (inf, interpreted as -ka, but his-

torically -k + a), p¯ok-al n. ‘going’, p¯okkuka v.t. ‘to make to go, re-

move’, p¯okk-al n. ‘removing’; Ko. ¯og (¯oy-/¯on-, also ¯o based on

∗

p¯o),

¯ok-c ‘to cause to go’, p¯ok-(p¯ok-y-) ‘to spend time’; To. p

¨ıx (p

¨ı-);

Ka. p¯ogu, h¯ogu;Ko

d. p¯ok-; Te. p¯owu (-w- < -g-), non-past neg p¯o-,

p¯ow- (intr), p¯oka n. ‘going, departure’; Ko

da, Pe. p¯ok (-t-) ‘to send’;

Man

da p¯uk-(-t-) id.

These forms illustrate the way the stems of set (b) were remade from the root of

set (a) by incorporating the non-past sufﬁxes k:kk into the stem in the intransitive and

5.4 Case

studies

187

transitive respectively. Interestingly, there is a greater regularity in the inﬂection of the

extended stem than that of the original root

∗

p¯o-. Also notice the occurrence of p¯oku in

the

NON-PAST intransitive paradigm of Old Tamil as p¯oku-v-, neg. p¯ok-, as opposed to p¯o

PAST p¯o-y-in, p¯o-n. This distribution preserves a trace of -ku- as a non-past (aorist,

optative) marker, even though the sufﬁx generally has lost its meaning. Ko

da, Kui and

Pengo of South Central Dravidian retain only the transitive forms of the derived stem.

Telugu shows both South Dravidian and South Central Dravidian features.

5.4.3

∗

l ‘to be, have’ :

∗

tu (<

∗

l-ntu) ‘is, are’ [697]

The form

∗

l is attested in Ta. Ma. u

l;Ko.o

l; Ka. u

l, o

l ‘to be, to have’; Ko

d. u

l-; Kui

lohpa (loh-t-) ‘to remain’; the

form

∗

tu is reﬂected in Ta. u

tu ‘is, are’ (existence),

nmai ‘truth’; Ma. u

tu ‘there is, exists’; Ko. o

do (3neu of

∗

l); To. w¨ı

ld- ‘to exist’

(3rd pers w¨ı

d-i); Ka. u

tu (3neu) ‘that is, that exists’, n. ‘existence’; Ko

d. u

d¨ı (3rd pers

l-); Tu. u

du (3neu sg of u

l, pres tense); Te. u

du ‘to exist, live, dwell’, u

du(nu)ﬁn

verb ‘he, she, it, they (neu) are’, u

d-r-u ‘they (hum) are’, where -

d- is a sandhi variant

of the aorist sufﬁx -d- (an alternative form occurs later as u

du-du-ru where u

du is

entirely treated as a root); Br. u

t (pres 1sg), us (pres 2sg), un (1pl), ure (2pl), ur (3pl)

related irregularly to the root anning ‘to be’.

Since both South Dravidian, South Central Dravidian and Brahui of North Dravidian

inherit reﬂexes of

∗

tu, the form must be reconstructed for Proto-Dravidian, as an

inﬂected form of the root

∗

l. Now, Tolk¯appiyam analyses

∗

tu (<

∗

l-ntu) as containing

-tu-, a non-past marker. But in terms of Proto-Dravidian morphophonemics the morph

should have the form

∗

ntu to yield u

tu <

∗

l + ntu; and this

∗

-ntu- is clearly different

from the past-tense marker -tu-. (Ka. u

tu is derivable from

∗

tu, in which case the

underlying sequence is interpretable as t (non-past) + tu (neu sg). In Old Kanna

da the

form is used in the neuter singular and plural (cf. Ramachandra Rao 1972: 257).

Moreover, the form u

tu is nowhere used in the past tense. The incorporated sufﬁx

∗

-ntu- must therefore be the same aorist marker that is retained in Old Tamil (-t-) and

also in Literary Telugu (-d-). In Old Tamil it occurs with only a limited number of verbs

in the ﬁrst plural and the second singular and plural (Glazov 1968: 106). It expresses

an indeﬁnite present–future meaning, as in varu-t-i (KT 91.14) ‘you come’, ¯a

tu-t-um

(Cil.9.63) ‘we will bathe ourselves’.

The use of

∗

tu differs from these other structures with

∗

-ntu- in two signiﬁcant

ways: (i) it is used as a third person in almost all the Dravidian languages, and (ii) it

is not limited to Old Tamil and Literary Telugu but is found in all the subgroups of

Another etymology similar to

∗

tu is

∗

w¯e

tu [DEDR 5528] from

∗

w¯e

l-ntu/

∗

w¯e

n-tu which

has cognates for the composite form in most of the languages of South Dravidian and Cen-

tral Dravidian, and Telugu of South Central Dravidian. Here

∗

(n)tu is a non-past morpheme of

Proto-Dravidian.

188 Wo r

d formation

Dravidian. The latter fact permits us to consider it an inheritance from Proto-Dravidian.

Now, its third-person use is in complementary distribution with the ﬁrst- and second-

person use of non-past

∗

-ntu- in Old Tamil and Literary Telugu. It therefore provides the

‘missing link’ that permits us to reconstruct a complete Proto-Dravidian paradigm for

forms containing the sufﬁx

∗

-ntu- and thus permits us to show that the Old Tamil and

Literary Telugu forms are not just regional innovations.

Notice ﬁnally that the Telugu verb u

du occurs in both past and non-past paradigms

d-i-ri ‘they (3hum) were’, u

du-du-ru ‘they are/will be’), thereby losing the original

signiﬁcation of

d <

∗

nt as a non-past marker. (The only exception is the third-person

human plural u

d-ru ‘they are’ of Classical Telugu, which is now archaic.) Here, then,

we ﬁnd the same phenomenon of ‘tense loss’ as with incorporated markers such as

∗

-k(k)- discussed in sections 5.4.1–2.

5.4.4 Incorporation of tense–voice sufﬁxes as stem formatives

In all the South Dravidian languages, except Kanna

da, there is a verbal conjugation

(Tamil Lexicon Class IV) in which tense and voice are combined in the same morphs.

Roots which generally end in i , u, y (incl ay), r and

z belong to this class. The conjuga-

tion is characterized by the following original sufﬁxes:

(8)

Non-past Past

Intr Tr Intr Tr

Aorist (k)-um kk-um nt tt

Future p [-w-] pp

In the following discussion we will see how these tense–voice sufﬁxes were incorpo-

rated to form extended stems in different South Dravidian, South Central and Central

Dravidian languages. Let us begin with an examination of reﬂexes of the verb

∗

z-i ‘to

descend’ [502].

(9)

Non-past Past

Intr Tr Intr Tr

Ta. i

zi-v- i

zi-pp- i

zi-nt- i

zi-tt- (> i

zi-cc-)

Ma. i

zi-v- i

zi-pp- i

zi-˜n˜n- i

zi-cc-

d. ¨ı

li-v- ¨ı

li- p- (dial.

¨ıp-) ¨ı

li-nj- (¨ı

l¨ıp-i-)

To. ¯ıx-(<

∗

z-g-) ¯ık-(<

∗

z-kk-) (¯ıx-y-) (¯ık-y-)

Ka. i

li-v- i

li-p- i

li-d- i

li-s-i

Pa. i

r-v-(<

∗

r-g) i

r-k-ip- i

r-k-it-

An alternative analysis of u

tu as containing the third singular neuter sufﬁx -tu must be rejected,

since it would not permit us to account for its tense meaning (present–future) and since the

attested forms are not conﬁned to the neuter singular.

5.4 Case

studies

189

Oll.Gad. i

r-g- i

rig-p- i

rig-t-

Te.

diggu/

digu

∗

z-g)

di-mp- [

dindu]

di-nc-

(arch.)

Kui d¯ı-v- d¯ı-p- d¯ı-t- d¯ı-p-t-

di-g-

di-p-(

dig-it-)

di-p-t-

∗

zi-(m)p-

∗

zi-mpp-

∗

zi-nt-

∗

zi-ntt-

/-(n)k-

Note that in the non-past there is evidence for both a labial and a velar sufﬁx as seen in

Toda, Kanna

da (i

zaku v.t.), Parji, Gadaba, Telugu and Ko

da. The Toda stem is remade

in the intransiti

ve and transitive by

incorporating the erstwhile non-past

morphs

∗

kk (or

∗

nk:

∗

nkk). So also in Parji and Ollari i

rv-/i

rg- have become the basic stems

without any tense meaning attached to their v/g. In Parji the intransitive:transitive

alternation is retained as g:k, whereas it is only signalled by sufﬁxation in Ollari. In

Kui,

too, the -

v-:-p- alter

nation signi

ﬁes intransiti

ve and transitive in

the non-past

(inﬁnitive).

The Telugu forms are diagnostic for the reconstruction of the proto-sufﬁxes. A new

verbal base

dindu (<

∗

z-nd-; cf. Ta. i

zi-nt-) v.i. ‘to sink, fall, droop, die’ is created in-

corporating the Proto-Dravidian intransitive past tense morph -nt- (Krishnamurti 1961:

52). The intransitive form

diggu/

digu is from

∗

zi-gg- <

∗

z-g- (see Krishnamurti 1961:

§1.124; 52, 53) through metathesis, gemination of the voiced stop after (C)V-, and sub-

sequent degemination. The comparison with Parji, Ollari and Toda shows an underlying

velar g (< -k-) which was apparently a non-past marker. There are two transitive forms

in Telugu:

di-nc-, which occurs as a base before past-tense sufﬁxes and

di-mp-, which

occurs before non-past sufﬁxes in Classical Telugu. Owing to the retention of the nasal,

these forms are traceable to PD

∗

z (i)-ncc-(<

∗

zi-ntt-) for the past tense base and

∗

(i)-mpp- for the non-past base.

Tamil, Malay¯a

lam and Ko

dagu attest the past-tense form i

zi-cc- from i

zi-tt-. The nasal

∗

ntt (transitive past, corresponding to

∗

nt intransitive past) is lost here in accordance

with normal sound change (see section 5.3 above). The Telugu sufﬁx retains the original

alternation of

∗

nt :

∗

ntt. Again,

di-mp- (<

∗

zi-mpp-) ‘to make one dismount’ presupposes

an intransitive

∗

-mp as opposed to transitive

∗

-mpp in the non-past. This piece of evidence

allows us to reconstruct

∗

mp beside

∗

p as non-past

intransitive marker within Proto-

Dravidian; however, while we have evidence for

∗

k there is none for

∗

nk in the intransitive

non-past.

I propose that

∗

nk and

∗

mp were incorporated into stems as (derivative) voice sufﬁxes

carrying intransitive meaning at a deeper chronological stratum of Proto-Dravidian.

The morphs

∗

p and

∗

k must have replaced

∗

mp and

∗

nk respectively in all non-past

190 Wo r

d formation

environments. Note that Old Kanna

da (Pampa Bharata; cf. Ramachandra Rao 1972:

Index) has i

li-s- and i

li- p- as past and non-past bases requiring underlying i

zi-cc-,

zi-pp- as found in the Tamil–Ko

dagu subgroup (with -cc- > -c- > -s-; -pp- > - p-

in Kanna

da). This illustration shows that the tense/voice conjugation, which is still re-

tained intact in the Tamil–Ko

dagu subgroup of South Dravidian, is reconstructible for

Proto-Dravidian. In the other members of South Dravidian, Central Dravidian and South

Central Dravidian, these morphemes have lost their tense meaning, becoming mere for-

mative sufﬁxes. The erstwhile tense marking can, however, be reconstructed from the

distribution of these morphemes in verb paradigms as illustrated above.

The past sufﬁx -ttu and the non-past sufﬁx -ppu, when delinked from their unlikely

looking relatives -nt (past intr) and - p,-mp- (non-past intr), have come to be segmented as

pure transitive markers which have become additive morphemes in a later chronological

stratum of Proto-Dravidian.

Cf. Ta. i

zi-ttu (i

zi-tt-i), i

zi-ccu ‘to lower’; i

zi-vu, i

zi-pu n.

‘inferiority’ (intr), i

zi-ppu n. ‘contemptuous treatment’ (tr).

The above discussion leads us to set up the following reconstructions within Proto-

Dravidian.

(10)

Non-past Past

Intr Tr Intr Tr

∗

zi-p- :

∗

zi-mp-

∗

zi-pp- :

∗

zi-mpp-

∗

zi-nt-

∗

zi-ntt-

∗

zi-k- :

∗

z-nk-

∗

zi-kk- :

∗

zi-nkk-

In this particular example evidence is lacking for the sufﬁx nk:nkk, but such evidence

found in various other etymolo

gies where the suf

ﬁx

is seen to have been completel

incorporated into the base in early Proto-Dravidian. Consider for instance

∗

wir-i-‘to

extend, open, split, crack, burst’ [5411]. The data in (11a) present a similar picture

to those in (10); but as (11b) shows, for this stem, some languages presuppose a velar

non-past sufﬁx, instead of a labial.

(11)

Non-past Past

Intr Tr Intr Tr

a. Ta. viri-v- viri-pp- viri-nt- viri-tt-

Ma. viri-v- viri-pp- viri-˜n˜n viri-cc-

d. biri-v- biri-p- biri-˜nj- biri-c-

To. pir-s- pir-c-

(< piry + θ -) (< piry + t-)

Ka. biri-v- biri-d-

As stated earlier (section 5.3), I am not ruling out the other possibility of

∗

-tt,

∗

-pp being also

original and ancient.