Krishnamurti Bhadriraju. The Dravidian Languages
Подождите немного. Документ загружается.
6.2 Gender
and number
: identi
fication
and de
finition 211
Proto-Dravidian. The arguments given in detail in Krishnamurti 1975a are summarized
here:
1. It would be much simpler to motivate innovations which resulted in Types I and III
from II than the reverse. The innovation in South Dravidian I (Type I) is the creation of a
feminine singular category in
∗
awa
.
l by adding the derivative suffix
∗
-a
.
l to the underlying
root
∗
aw-, a process also found in noun morphology, e.g.
∗
mak-antu ‘son’,
∗
mak-a
.
l
‘daughter’ [4616]. In the case of plural the difference is one of semantic extension of
‘men’ > ‘men and women’ > ‘women’, because of
∗
awar being the natural choice to
represent mixed groups of ‘men and women’. We can even say that the first part in
semantic extension took place within Proto-Dravidian itself (see Krishnamurti 1975a:
§12). The later shift
in Types I and III is the use
of
∗
awar to mean
exclusive group(s)
of women. This last change did not take place in Central Dravidian and South Central
Dravidian.
7
On the contrary, it would be difficult to motivate a change in South-Central
Dravidian (South Dravidian
II) and Central Dravidian
from Types I or III to II, i.e.
splitting the meaning of ‘men and women’ by assigning ‘women’ to the non-human
group signalled by
∗
away. What could be the linguistic or sociological contexts in which
such a semantic shift would be induced? Therefore, Type II is a plausible candidate to
represent the Proto-Dravidian gender system.
2. The
creation of
∗
awa
.
l in South Dravidian I has restricted the meaning of
∗
atu to
non-human animate and inanimate in the singular. Similarly the extension of
∗
awar to
human has restricted the meaning of
∗
aw
ay
in
South Dravidian I (plus Telugu)
and North
Dravidian to non-human animate and inanimate.
3. Telugu, being a major literary Dravidian language, has a great deal of give and
take with the other literary languages of South Dravidian, Tamil and Kanna
.
da. This
explains why it was influenced by the semantic shift of the other literary South Dravidian
languages in the case of [+ high] plural, while retaining the inherited system in the
singular. A similar shift independently in Ku
.
rux–Malto only shows the naturalness of
the semantic change involved.
4. South-Central Dravidian (South Dravidian II) and Central Dravidian are genetically
distinct subgroups and their sharing a common
gender system would show that it is a
shared retention and not a shared innovation. Note that PD
∗
awantu (nominative,
∗
awan
oblique) is preserved in South Dravidian II, Central Dravidian and North Dravidian,
7
There are independent and isolated innovations in different languages involving the ‘human
female’ subcategory. Note avek ‘they (women)’ was innovated only by Pengo in South Dravidian II
(table 6.1). In Gadaba the finite verb suffix for m pl, -ar, is found in some sentences as agreeing
with subject nouns meaning ‘women’, instead of the expected -av; Texts II: 98. ¯asmaskil u
.
nkun
k¯edar, 99. ¯or p¯a
.
tel p¯ardar, 98. ‘Women will do planting of seedlings’, 99. ‘They will sing songs’
(Bhaskararao 1980: 75, 83, also cf. 15). In Malto the accusative case inflection is ad-in ‘her’, but
ad-en ‘it’ (acc) from the same pronoun ¯ad ‘she, it’ (Mahapatra 1979: 13). These illustrate the
naturalness of the grammatical change in question.
212 Nominals
being a retention. But South Dravidian I has lost
∗
-tu of the nominative and analogically
restructured the nominative as
∗
awan on the oblique (cf. a similar independent change in
Pengo). In Ko
.
n
.
da also, a similar loss of the final syllable occurs in finite verbs, when they
are not followed by a vowel; e.g. v¯an
ru kitan ‘he did’, but v¯anru kitanr-a? ‘Did he do?’
8
5. In derivational morphology
∗
-ar is the plural of
∗
-ant 3m sg. It would then make
sense to consider
∗
-ar to originally mean ‘men’ before it extended to ‘persons’.
6. The only argument in favour of Type III being the proto is that it is shared by
members of diverse subgroups, Telugu of South Dravidian II and Ku
.
rux–Malto of North
Dravidian. As a general principle this is fine, but it cannot be applied as a rule of thumb.
In that case, the lack of gender in Toda and Brahui should be considered proto since
they are from widely separated branches of the family. Item (4) above uses the same
argument as well as many others to consider Type II as reflecting the Proto-Dravidian
situation. (For further elaboration,
see Krishnamurti 1975a.)
6.2.4 Gender-number marking in finite verbs
The pronominal suffixes in finite verbs, which agree with subject NPs,
are closely related
in form to the demonstrative pronouns and are reconstructible for Proto-Dravidian, i.e.
∗
-ant msg(>
∗
-an in South Dravidian I), -Vr m pl or hum pl,
∗
-at n–msg;inthe
case of non-masculine plural, the widely distributed reconstruction is
∗
-aw in South
Dravidian I, South Dravidian II and Central Dravidian. There are, however, several
exceptions to the proto system. The number of contrasts in verb concord is some-
times less, but never more, than the number of contrasts specified by the demonstrative
pronouns.
Within South Dravidian I, the K¯a
.
nikk¯ara and Erukala dialects of Tamil, Modern
Malay¯a
.
lam, Ko
.
dagu, Ku
rumba and Toda have lost gender distinction in the verb system
either by a total loss of agreement features in personal suffixes (as in Modern Malay¯a
.
lam)
or by extending the distribution of the non-human singular suffix, as in Ko
.
dagu, and
Toda. In Ku
rumba and Kota, the third neuter singular is extended to the plural also. In
Ku
rumba -ad occurs in the singular for ‘he, she, it’ and for the neuter plural ‘they’; -o
occurs corresponding to
∗
Vr with the loss of -r.
In South-Central Dravidian, Gondi and Pengo have extended the velar suffix
∗
-kk,
∗
-nk
of non-person plural to verbs, taken from the noun morphology. Pengo, which has a six-
way distinction in subject pronouns, has a five-way distinction in verb agreement, with
one morpheme -at for feminine singular and non-human singular, true to the subgroup
trait. All other South-Central Dravidian and Central Dravidian languages have a four-way
contrast in verb agreement corresponding to subject pronouns.
8
The independent developments in Ko
.
n
.
da and Pengo point to the naturalness of the loss of the
final unaccented syllable in a free form.
6.2 Gender
and number
: identi
fication
and de
finition 213
Ku
.
rux and Malto use the same form for the singular and plural non-human, as is
the case with Later Tamil and Kota (Subrahmanyam 1971: 401–2). Brahui, which has
retained only number and lost all gender distinction, preserves the traces of Proto-
Dravidian contrasts in form in pronouns and in verb agreement, e.g. ¯od ‘he, she, it’, ¯ofk
‘they’ (<
∗
at and
∗
aw(a) + kk); in the verbs ¯e, as, ¯ar(a), ur occur in the singular and ir,
¯er, ira, or, ur, as in the plural. It is apparent that these can be traced to PD
∗
-ant,
∗
-Vr
on formal grounds. It therefore appears that Brahui originally had a four-way contrast
of gender–number as in the other two North Dravidian languages.
6.2.5 Gender-number marking in nominal derivation
The determination of gender in Dravidian is mostly based on meaning and not on form as
the preceding description shows. There are, however, certain derivational morphemes,
which may denote either lexical (sex) or grammatical gender. But true grammatical
gender is expressed by the anaphoric use of pronouns, or by numeral or verbal agreement.
The markers are
∗
-an-t/
∗
-wan-t (male or masculine singular in Proto-Dravidian),
∗
-a
.
l,
∗
-i(female or feminine singular in South Dravidian, South-Central Dravidian, North
Dravidian),
∗
-ar/
∗
-war (masculine plural/human: Proto-Dravidian);
∗
-(n)k(k)a (South-
Central Dravidian, Brahui),
∗
-
.
l (Central Dravidian, Telugu, Tu
.
lu), or a combination of the
two as
∗
-(n)k(k)a
.
l (common or originally non-human plural: South Dravidian, Central
Dravidian, Tu
.
lu). Cardinal, elicited numerals have neuter (non-person) concord, but,
when they classify persons, the derivative suffix
∗
-war is added to the numeral root
morpheme, e.g. Tamil m¯un
ru ‘three things’, but m¯u-var ‘three persons’. The common
plural form,
∗
-nkkV
.
l, has progressively replaced the human plural and it appears to be
a trend already prevalent in the pre-divided stage of Proto-Dravidian. Even in Ca˙nkam
Tamil we note the use of -ka
.
l added to -ar as a plural marker, e.g. arac-ar-ka
.
l ‘kings’
(Kalittokai 25:3) (see Rajam 1992: 272).
The gender–number marking (lexical in some and grammatical in others) is illustrated
below for most of the languages (largely based on Shanmugam 1971a: 30–103, Zvelebil
1977: 12–16).
South Dravidian:
∗
-an/
∗
-wan,
∗
-a
.
l,
∗
-i;
∗
-ar/
∗
-war,
∗
-k(k)a
.
l.
1. Tamil: ka
.
l-va
n ‘thief’, ka
.
l-vi ‘a female thief ’, mak-an ‘son’, mak-a
.
l ‘daughter’,
t¯o
.
z-i ‘a female friend’, ka
.
l-var ‘thieves’, k¯e
.
l-ir ‘relatives’, ai-var-ka
.
l ‘five persons’
(-ka
.
l added to -var, Classical Tamil); y¯a
.
n
.
tu-ka
.
l ‘years’, mara˙n-ka
.
l ‘trees’, the use
of -ka
.
l to denote plural neuter is extremely rare in Tamil. Modern Tamil loses the
final -
.
l, maran-ka ‘trees’, pasu-kka ‘cows’ etc.
2. Malay¯a
.
lam: ta
.
t
.
t-an ‘goldsmith’, mak-a
.
l ‘daughter’, kora-tt-i ‘a Korava woman’,
¯ı
.
za-var ‘toddy-tappers’, iru-var ‘two persons’, a
.
t
.
tu-ka
.
l ‘leeches’, mara˙n-˙na
.
l
‘trees’, pa´su-kka
.
l ‘cows’.
214 Nominals
3. Kota: kur
.
d-n ‘blind man’, ka
.
l-i ‘female thief ’, av-r ‘those persons’, ¯a
.
l-g¯u
.
l ‘people’,
n¯ay-g¯u
.
l ‘dogs’, marm-g¯u
.
l ‘trees’.
4. Toda: ko
.
l-
.
n ‘thief’, to
.
lx-t-y ‘woman of goldsmith caste’, m¯ım-i ‘mother-in-law’,
mox-¯am ‘boys’, p¯uf-¯am ‘flowers’.
5. Ko
.
dagu: ka
.
l
.
l-¨en ‘male thief ’, ka
.
l
.
l-i ‘female thief ’, ka
.
l
.
la-(r) ‘thieves’, m¯u-v¨e ‘three
persons’, t¯o
.
l¨en-ga(
.
l ) ‘wolves’, a
.
n
.
na-˙nga(
.
l ) ‘elder brothers’. [Ko
.
dagu loses word-
final liquids.]
6. Kanna
.
da: aras-an ‘king’, mag-an ‘son’, aras-i ‘queen’, mag-a
.
l ‘daughter’, pola-ti
‘an outcaste woman’ (see section 6.2.6 below for -t), aras-ar ‘kings’, ka
.
l
.
l-ar
‘thieves’, magan-dir
9
‘sons’, t¯ay-vir ‘mothers’, mane-ga
.
l ‘houses’, gorav-ar-ka
.
l
‘masters’ (the last in an
inscription of the ninth century;
see Gai 1946: 28),
n¯ay-
gu
.
lu ‘dogs’.
7. Tu
.
lu: kur
.
d-e ‘blind man’, kur
.
d-i ‘blind woman’, ajj-er¨ı ‘grandfathers’, ir-ver¨ı ‘two
persons’, pili-k
.
lu ‘tigers’, p¯u-ku
.
lu ‘flowers’; arasu-
.
lu ‘kings’, pucce-
.
lu ‘cats’.
10
South-Central Dravidian:
∗
-an-t/
∗
-wan-t,
∗
-a
.
l,
∗
-i;
∗
-Vr,
∗
-(n)k(k)V. Here, the derived
female forms do not represent the feminine gender. A female human plural
∗
-si-kk has
developed as an innovation in all these languages, except Telugu.
8. Telugu: tammu-
.
n
.
du ‘younger brother’, k¯o
.
d-alu ‘daughter-in-law’, iru-wuru ‘two
persons’, r¯aju-lu ‘kings’, t¯o
.
ta-lu ‘gardens’. An older
∗
-(n)kka
.
l reanalysed as -(n)kk-
and -V
.
l can be detected in certain forms like mr¯anu ‘tree’: mr¯a
n
-kulu ‘trees’, go
.
du-
gu ‘umbrella’ : go
.
dugu-lu ‘umbrellas’ (< PD
∗
maram,
∗
ko
.
t-V-); (-lu plural from
∗
-
.
lV is an exception in this subgroup.)
9. Gondi: tott-¯or ‘ancestor’ (-¯or <
∗
¯or <
∗
awar <
∗
awa-nru), s¯el¯a
.
r (pl s¯el¯a-hk)
‘younger sister’ (from
∗
s¯el-¯a
.
l-sk), p¯e
.
r-¯ı ‘girl’, kan
.
d-¯ı r ‘boys’, mar-k ‘sons’, ka
.
t-k
‘eyes’, m¯ar¯e-ng ‘plumes’, pi
.
t
.
t¯e-ng ‘birds’.
10. Ko
.
n
.
da: ka
.
tka-yen ‘miserly person’, ka
.
n-i ‘blind woman’, ana-si-r ‘older brothers’,
r¯as-ku ‘kings’, no
res-ku ‘tigers’, b¯ıb-si-k ‘elder sisters’, kome-
ŋ
‘branches’, r¯eto-
ŋ
,
‘crabs’.
9
This particular plural form belongs to a small class preserved in Early Kanna
.
da in an inscrip-
tion of the eighth century (Gai 1946: 28) where the historically correct analysis was magand-ir
(<
∗
mak-ant-). But by reanalysing it on the presumption that the stem was magan-, the gram-
marians treated -dir as the plural suffix. There were also in Old Kanna
.
da avand-ir ‘those per-
sons’, ivand-ir ‘these persons’ in which the stems reflect Proto-Dravidian
∗
awant- and
∗
iwant-
(Ramachandra Rao 1972: 54). In Middle and Modern Kanna
.
da, -ndru has become a productive
human plural suffix, added to kinship terms (see Krishnamurti 1975a: 340, Kushalappa Gowda
1972: 218).
10
The use of -
.
lu as a plural marker in the last two examples is found only in Tu
.
lu in South
Dravidian I. This was one of the reasons to class Tu
.
lu with Central Dravidian earlier. Later in
this chapter it is argued that
∗
-
.
l is a retention of a Proto-Dravidian plural sign in Telugu, Tu
.
lu and
the Central Dravidian languages.
6.2 Gender
and number
: identi
fication
and de
finition 215
11. Kui: t¯o
.
r-enju ‘friend’ (-nj <
∗
-nt), gah-ali (pl gah-al-ska) ‘sweeper woman’, d¯ada-
ru ‘elder brothers’, t¯o
.
re-
ŋ
ga ‘male friends’, k¯or-ka ‘buffaloes’, k¯o
.
di-
ŋ
ga ‘cows’.
12. Kuvi: k¯ut-ka (k¯undu sg) ‘mushrooms’, k¯oma-
ŋ
ga ‘branches’, seppu-
ŋ
a ‘shoes’.
13. Pengo:
.
to
.
n
.
d-en ‘brother’, to
.
rnd-el ‘sister’,
.
to
.
n
.
d-ar ‘brothers’, ¯a-cku ‘women’,
kogle-k ‘women’, k¯o
.
di-
ŋ
‘cows’.
14. Man
.
da: n¯a-mga-hke (<
∗
n¯a-maga
.
l-ska) ‘my daughters’, h¯ulpand-el ‘a beautiful
woman’, ka
.
rd-er ‘boys’, kan-ke ‘eyes’, p
.
r¯e-ke ‘bones’, t¯uku-
ŋ
‘feathers’ (Burrow
1976: 42–3).
Central Dravidian:
∗
-an-t,
∗
-a
.
l,
∗
-Vr,
∗
-
.
l,
∗
-(k)kV
.
l; innovated:
∗
-cil,
∗
-til.
15. Kolami: t¯or-en (with loss of final -d, cf. Parji) ‘younger brother’, komm-al ‘daugh-
ter’, m¯as-ur ‘men’, do
ŋ
ga-l ‘thieves’, g¯ar-sil ‘hail stones’, ella-gul ‘houses’.
16. Parji: tol-ed (obl. tol-en) ‘brother’ (-ed <
∗
-ed <
∗
-ad <
∗
-an- d), k¯et-al ‘widow’,
tol-er ‘brothers’, mayi-l ‘husbands’, tulla-kul ‘weavers’, pu
.
rut-il ‘insects’, v¯ar-til
‘roots’, ¯enu-cil ‘elephants’.
17. Ollari: maggi-in
.
d ‘man’, ko
.
r-al ‘son’s wife’, il-er ‘bridegrooms’, ki-l ‘hands’, gar-
sil ‘hail stones’, sir-kil ‘buffalos’, aya-sil ‘wives’, kan¯ır-til ‘tears’, ile-v ‘brides’.
North Dravidian:
∗
-an-t/
∗
-wan-t,
∗
-i,
∗
-Vr/
∗
-wVr;
∗
-kk (the last only in Brahui).
18. Ku
.
rux: ¯al-as ‘man’, ¯al-i ‘woman’, ¯al-ar ‘men’ (neu pl gu
.
th¯ı is apparently not
Dravidian, e.g. man gu
.
th¯ı ‘trees’).
19. Malto: maq-e ‘son’, qal-we ‘thief’, maq-i ‘girl’, mal-er ‘men’.
20. Brahui: b¯a-k ‘mouths’, pu
.
t-¯ak ‘hairs’.
6.2.6 Reconstruction of gender–number suffixes
The suffixes that can be reconstructed for Proto-Dravidian are -an
t for masculine singu-
lar (in derivation, male human singular), -a
.
l,-i for feminine or female human singular
and -Vr for [+ high] plural (either male human or simply human plural, depending on
the language and subgroup). Within Proto-Dravidian itself it appears that
∗
-ant /
∗
-want
and
∗
-ar/
∗
-wVr occurred as allomorphs of masculine singular and masculine plural,
respectively. Ku
.
rux xal-b-as ‘thief’, xal-b ‘theft’ and Malto qal-we ‘theft’ would lead
us to relate w-/b-tothe
∗
w of
∗
-want (see Shanmugam 1971a: 107–8). But, alterna-
tively, -
∗
w could be an abstract noun-forming suffix derived from PD
∗
-way also (see
section 5.8), to which the gender marker -as (<
∗
-ant) was added. In any case, I think
that these allomorphs
∗
-ant ∼
∗
-want go to Proto-Dravidian; the second allomorph was
due to reanalysing w- as part of the suffix, separating a-, i-, u- as the deictic roots. I have
identified the deictic roots as
∗
aH-,
∗
iH-,
∗
uH- and proposed a rule
∗
H → w/
—
+ a,
by which
∗
aw-antu,
∗
iw-antu and
∗
uw-antu,
∗
aw-ar,
∗
aw-ay, etc. were derived. This rule
was extended to environments of following semivowels also in Early Tamil: aw-y¯a
nai
216 Nominals
‘that elephant’ (see section 4.5.7.2 c; Krishnamurti 1997b). The laryngeal was assimi-
lated to a following voiceless stop or was dropped before voiced stops:
∗
aH + ka
.
tal →
ak-ka
.
tal ‘that sea’,
∗
aH-t ∼
∗
aH-tu → att- ∼ a-tu [adu]. When it was not bound,
∗
aH
was contracted to
∗
¯a in all subgroups. These developments explain how the analogi-
cal restructuring of
∗
-want and
∗
-war could go to the late stages of Proto-Dravidian
itself.
South Dravidian I had two clear shared innovations: (i) The loss of
∗
t in nominative
singular (see fn. 3 above) and (ii) the creation of awa
.
l as feminine singular. The final
∗
t
(perhaps a nominative marker) is preserved in all other subgroups – South Dravidian II,
Central Dravidian and North Dravidian. Certain other consonants occur before the gender
suffixes, which ha
ve to be taken as stem formati
ves and not part of the gender
markers,
e.g.
∗
-kk-antu in Ta. mutu-kk-an,Te.muduk-˜a
.
du, Pa. Oll. mutt-ak;
∗
-tt-antu/
∗
-tt-i in Ta.
Ma. oru-tt-an ‘one man’, oru-tt-i ‘one woman’, beside oru-va
n, has cognates in Te. ¯or-ti
‘one woman’, Ku
.
r. or-ot, Malt. or-te ‘one man’. With the loss of the
gender marker,
it could be that the stem formatives took over the function of marking gender in Parji,
Gadaba, Ku
.
rux and Malto.
The feminine suffix
∗
-a
.
l/
∗
-¯a
.
l is found in South Dravidian I and South Dravidian II and
Central Dravidian as part of their derivational morphology in several items denoting a
female person, e.g.
∗
mak-a
.
l ‘daughter’ (South Dravidian I, South Dravidian II) [4616],
∗
k¯o
.
z-¯a
.
l ‘daughter-in law’ (South Dravidian II and Central Dravidian) [2149]. An inno-
vation of Central Dravidian is to create three sets of derived nouns by adding gender
suffixes to numerals one to three (see numerals in this chapter).
There were two plural suffixes in Proto-Dravidian, -Vr for [+ high] also called the
epicene plural and
∗
-(n)kkV,
∗
.
l and (n)kkV
.
l for [− high], neuter or non-human plural. The
latter one became generalized as the normal plural suffix in different languages. There
is no dispute about the reconstruction of the first. The 3 hum pl suffix -Vr occurred
as plural marker of masculine/human nouns in all literary languages in their earlier
stages, but it was gradually replaced by the neuter plural which came to be known as
the common plural. Examples: Ta. k¯e
.
l-ir ‘relatives’, Ka. ka
.
l
.
l-¯ı r ‘thieves’, Te. allu-ru
‘sons-in-law’. In the modern languages, the demonstrative pronouns still show -ar,Ta.
Ma. av-ar, Ka. avru,Te.w¯aru (for the other languages see table 6.1). The agreement
marker in the finite verbs is also retained as -Vr,Ta.avaru p¯or¯aru ‘They are going’
(3 pol sg), Te. w¯a
.
l
.
lu we
.
lt¯aru ‘They will go’, w¯aru we
.
lt¯aru ‘He (pol) will go’. The -Vr
is also retained in the plural of human numerals, Ta. iruvar ‘two persons’, Ka. ibbaru
id., OTe. iruwuru, Mdn Te. mu-gguru ‘three persons’ from earlier m¯u-guru < m¯u-wuru.
The languages of South Dravidian II and Central Dravidian also preserve the [+ high]
plural marker in numerals and finite verbs, Ko
.
n
.
da ri
ʔ
er ‘two persons’, v¯ar v¯atar ‘they
came’.
6.3 Cases 217
The neuter plural
∗
-(n)ka
.
l used in South Dravidian I was considered a combination
of two suffixes
∗
-(n)kkV and
∗
-
.
l by Jules Bloch (1954: 10). He says, ‘The guttural alone
is common to the whole family.’ Even by the Early Ca˙nkam period -ka
.
l was becoming a
common plural, e.g. arac-ar-ka
.
l ‘kings’. The variant
∗
-(n)kk occurs in South Dravidian II
and Brahui;
∗
-
.
l occurs as one of the plurals in Central Dravidian and Tu
.
lu. Telugu has
generalized
∗
-
.
l. All languages of South Dravidian I and Central Dravidian have reflexes
of
∗
-(k)ka
.
l. The suffix shared by South Dravidian II and Brahui must be a retention and
not a shared innovation; so also,
∗
-
.
l must be a retention because it is shared by Telugu,
Tu
.
lu and Central Dravidian. The combination of these two into
∗
-(n)kkV
.
l must be an
innovation in South Dravidian I and Central Dravidian, representing a shared isogloss.
6.2.7 Conclusion
From a synchronic and diachronic study of gender–number in Dravidian, the following
general observations can be made:
1. Reduction in gender–number distinctions tends to be a typologically motivated and
not a genetically inherited
change; therefore, it does
not serve always as a strong
basis of subgrouping.
2. The number of gender–number contrasts in governed positions (e.g. finite verbs) is
never larger than in governing positions (Subject NPs).
3. Category simplification (neutralization) takes place more often in governed positions
than in governing positions.
4. In neutralization, there are cases of suspension of (a) both number and gender
(Malay¯a
.
lam in verbs) or (b) suspension of gender but retention of number (Kanna
.
da
dialects, Gowda and Southern Havyaka in verbs, Toda and Brahui in pronouns), but
no language retains gender alone totally suspending number (also see Greenberg
1963: 95, Universals 36 and 37).
5. In neutralization, it is more often the unmarked cate
gories (singular in number and
non-masculine in gender) that extend their ranges of usage than the marked ones.
Thus non-masculine
∗
atu (sg) replaces
∗
away (pl), and these two replace masculine
∗
awantu and
∗
awar, respectively, and not vice versa. There is one exception to this
statement
in the H
¯alakki
dialect of Kanna
.
da
(see table 6.1, Kanna
.
da).
6. In gender–number reconstruction, contrasts in subject pronouns are more basic and
primary than agreement features in verbs.
6.3 Cases
Case relations in Dravidian are expressed either by bound morphemes or by grammati-
calized nouns or verbs, called postpositions. We will begin with case markers, which are
reconstructible, and then deal with postpositions of individual languages. The nominative
218 Nominals
case is unmarked. The noun is used in its elicitation form in the nominative singular;
in the plural it is the noun with the plural suffix. A final consonant or syllable found in
the nominative is sometimes missing in non-nominative cases, e.g. Ta. mara-m/mara-
n
‘tree’: Ta. mara-tt-ai ‘tree’ (acc), mara-tt-in- (gen) ‘of a tree’, Ka. mara ‘tree’, mara-da-
‘of a tree’. Here, one can technically consider -m as the nominative singular, but it does
not make the analysis any simpler. The plural forms retain the final -m:Ta.mara˙n-ka
.
l
(nom pl) ‘trees’, mara˙n-ka
.
l-ai (acc pl) ‘trees’. Therefore, it would be much simpler to
propose Ø marking in the
nominative and posit loss or addition
of any element to the
stem when it occurs
with other cases.
The demonstrativ
e and interrogative masculine
forms based on the deictic roots
as
well as certain masculine nouns end in
∗
-ntu in nominati
ve singular in Proto-Dravidian,
e.g.
∗
awa-ntu ‘he’,
∗
maka-ntu ‘son’. The oblique stem occurs without the final
∗
-tu in all
the languages. It is, therefore, technically correct to consider
∗
-tu as nominative singular
which is replaced by
∗
-r in the plural. In the languages of South Dravidian I, the final
∗
-tu was lost as a shared innovation in the nominative singular, i.e. the nominative was
restructured on the analogy of the oblique base ending in -n, e.g
∗
awan-. With the
exception of these two cases, the nominative case does not carry any marking.
6.3.1 The oblique stem
The non-nominative cases are called the oblique cases. Some noun stems add certain
suffixes or augments to form the oblique stem. These elements were called ‘inflectional
increments’ by Caldwell. It is difficult to predict or define the phonological or semantic
properties of stems that ha
ve different forms between the
nominative and the oblique
cases. The ‘inflectional increments’ are called c¯ariyai (‘signs, markers’) by Tolk¯appiyam
and upavibhakti (co-cases) in traditional Telugu grammars. In modern descriptions they
are called ‘empty morphs’, ‘link morphs’ or ‘augments’. Their function is to make the
stem eligible to receive case markers. I call them augments, which is a neutral term. They
have no semantic content. But in several languages, the oblique stem is identical in form
with the genitive case form, which is used as an attribute to a following noun. In Telugu,
an oblique stem can either take a case suffix or syntactically function as an adnominal,
e.g. kannu ‘eye’, obl ka
.
n-
.
ti, with case marking ka
.
n
.
ti-ki ‘to the eye’, ka
.
n
.
ti-t¯o ‘with the
eye’; syntactically, ka
.
n
.
ti
1
p¯apa
2
‘the pupil
2
of the eye
1
’. The following augments can be
reconstructed for individual subgroups or the whole family, as the case may be.
6.3.1.1
∗
-tt-
South Dravidian I The most widely represented augment is
∗
-tt- which occurred in
Pre-Tamil after nouns of (C)VCVm/
n type. It replaces the final nasal. The reflexes of
this can be found
in Tamil, Malay
¯a
.
lam, Iru
.
la, Ko
.
dagu, Toda and K
ota. Kanna
.
da and Tu
.
lu
have -d- correspondingly:
6.3 Cases 219
Old Tamil: Neuter nouns ending in -a] m replace -m by -tt- as an augment, e.g. mara-
m/-
n ‘tree’ mara-tt- in all cases, except the sociative -o
.
tu in Old Tamil, mara-ttu-kku
‘to a tree’, mara-tt-il ‘in a tree’, but k¯alam-o
.
tu ‘with time’ (Shanmugam 1971a: 201–3).
Also stems of three morae (C)
¯
VCV/(C)VCVCV, in which the final syllable is -
.
tu/-
ru, add
the augment -tt-, k¯a
.
tu ‘forest’: k¯a
.
t
.
t-il ‘in the forest’ (<
∗
k¯a
.
t-tt-), ka
.
liru ‘male elephant’,
ka
.
li
r r-o
.
tu ‘with a male elephant’. These two types are further generalized in Modern
Tamil; even loanwords ending in -m like Skt. d¯uram ‘distance’ and Eng. sis
.
tem (system)
replace the final -m by -tt- before case suffixes. Modern Tamil also has the obliques of
stems ending in -
.
tu and -ru (<
ru)ask¯a
.
t
.
t-, ¯att-(< ¯ar r), respectively (Schiffman 1999:
25–6). Note that Old Tamil
tt becomes tt in Modern Tamil.
Malay¯a
.
lam has a parallel
pattern; neuter nouns ending
in -
am and stems in final -
.
tu/-
tu
add -tt as augment. Final -m/-n is dropped and sandhi will result in the gemination of
retroflex and alveolar voiceless stops as in Tamil, mara-tt-¯al (instr), k¯a
.
t
.
t-¯al ‘by the forest’,
c¯o
r r-il ‘in the rice’. Modern Malay¯a
.
lam obligatorily takes two augments -tt- and -in-in
dative and genitive cases, mara-tt-in ‘of the tree’ (Asher and Kumari 1997: 191–4).
Iru
.
la has -tt- for stems ending in -am, mara-tt-e ‘tree’ (acc), kelaca-t-ke ‘for work’
(dat); -tt- is also extended to
human nouns, e.g.
r¯aman-itt-i ‘in Raman’, ava
.
l-itt-i ‘in
her’ (Perialwar 1978b: 7–9).
Ko
.
dagu adds -t- after nouns ending in -am, e.g. maram ‘tree’: mara-t-na/ mara-t¨ı-na
(acc), mara-t-i ˜nji (abl), mara-t-l¨ı (loc).
Toda has -t- as oblique marker which replaces a stem-final nasal (-n,-
.
n,-m), e.g. m¯e
.
n
‘tree’: m¯e
.
n-t-k ‘to a tree’, neln- ‘ground’: nelt-, ¯ırm- ‘dampness’: ¯ırt-; also in the class
of forms that had an underlying -
.
t/-
t, kw¨ı
.
r ‘horn’: kw¨ı
.
t-, ¯ar ‘way, road’: ¯at-. The same
marker is extended to other classes, nes
of ‘moonlight’: nesot-; -t,-d,-θ are all added to
stems in the genitive case (Emeneau 1984: 70–6).
Kota has -t in inherited stems in final -m,-
.
r,-r, e.g. mar-m ‘tree’: mar-t-n ‘tree’ (acc),
mol-m ‘hare’: molm-t-k ‘to the hare’,
∗
n¯a
.
r-t- →n¯a
.
t- ‘country’,
∗
v¯er-t- → v¯et- ‘root’.
Kanna
.
da has -d-, perhaps a weakened variant of -tt- in the class of neuter stems ending
in -am, mara-d-a (gen) ‘of the tree’, mara-d-inda (instr–abl) ‘by the tree’, in genitive,
instrumental–ablative and locative cases. Ba
.
daga also has -d in a-ending stems, mara-d-¯o
‘in the tree’, h¯ava-d-enda ‘by the snake’.
Tu
.
lu has -ta and -da as genitive markers in complementary classes of stems, e.g. p¯u-ta
‘flower’ (gen), ka ˜nji-da ‘calf’ (gen); the ablative case marker is given as -tt¨ı.Itisnot
clear if the dental element is to be interpreted as an augment or part of the case suffixes.
Note that it recurs in two cases. Correspondingly, the neuter demonstrative nouns have
-ta and -t
i, ay-t-a ‘that’ (gen), ay-t
i (abl). After human nouns the genitive is -na, which
shows that -a can be treated as the genitive suffix and -t- and -n- as augments (Bhat
1967: 85–7). Koraga has -tt/-t in non-human stems in all three dialects, e.g. mara-tt-a
(acc), erd¨ı ‘ox’: erd¨ı-ta- (Bhat 1971: 7, 2, 40).
220 Nominals
South Dravidian
II
Telugu has -t-i (<
∗
-tt-i) as oblique-genitive marker
11
in stems of
three morae ending in
∗
-
.
tu [-
.
du] and
∗
-tu [-ru], n¯a
.
du ‘country’: n¯a
.
ti-, ¯eru ‘river, stream’:
¯e
.
ti-ki ‘to the river’; disyllabic stems in final
¯
Vyi/
˘
V
ru replace the final syllable by -ti, e.g.
c¯eyi ‘hand’: c¯e-ti-, pa
ru (also spelt parru) ‘village name suffix, low land’:
∗
par-ti/par-ti.
These processes and forms also continue in Modern Telugu, except for the merger of
OTe.
r with r (<
∗
r).
In Gondi, singular nouns of the
non-human category in
final -a or a sonorant consonant
-n,-r,-
.
r,-m take the -t augment, l¯on ∼ r¯on ‘house’ (<
∗
.
l¯on/
.
r¯on <
∗
o
.
l-an): r¯o-t-¯al ‘from
the house’, mara(n) ‘tree’: mara-t-, n¯ar ‘village’:
∗
n¯ar-t- → n¯a
.
t-[¯e ‘in the village’ (here
r is from Pre-Gondi
∗
r). Monosyllabic stems in final liquids and semivowels add -d as
the oblique marker, e.g. kay ‘hand: kay-d-, k¯al ‘leg’: k¯al-d-¯e ‘with the leg’, n¯el ‘field’:
n¯ed-. There are distributional differences between dialects (Rao 1987b: 139–48).
In Ko
.
n
.
da non-masculine stems ending in vowels or nasals take -di as the oblique
formative, e.g. ayli ‘girl’: ayli-di-
ŋ
(acc–dat), guram ‘horse’: guram-di-
ŋ
(acc–dat.).
This suffix can be interpreted as a sequence of -d- (weakened form of -tt-) plus -i (the
genitive marker). In a few irregular stems an underlying -d-or-t- as augment is needed
to explain the locative forms, e.g. ilu ‘house’: in
ro (<
∗
il-n-d-o) ‘in the house’, n¯aru
‘village’: n¯a
.
to (< n¯a
.
t-
.
to<
∗
n¯a
.
t-t-o) ‘in the
village
’.
12
There
are a few exceptions to
this with a zero oblique suffix, go
.
reli ‘axe’: go
.
reli-
ŋ
(acc–dat). a few loanwords in final
-m take -ti instead, d¯esem ‘country’: d¯esem-ti- (obl–gen). For details, see Krishnamurti
1969a: 248–53.
Kui has the augment -tin-(<
∗
tt-i + n, a sequence of three augments) in neuter stems
of the type, (C)
¯
VC, e.g. k¯oru ‘buffalo’: k¯oru-tin-gi (dat), k¯or-ka (pl): k¯orka-tin-gi (dat).
‘The genitive is the same as the inflectional base’ (Winfield 1928: 24, 28).
In Kuvi -ta- occurs as the oblique marker in non-human nouns in the singular and
plural before the dative suffix, e.g. ilu ‘house’: ilu-ta-ki (dat), ilka-ta-ki (pl dat); -t-
occurs in singular nouns before accusative and locative cases, e.g. ilu-t-i ‘house’ (acc),
ilu-t-a (loc). In a different dialect, owing to prehistoric sandhi, some stems have irregular
obliques in the locative, e.g. ilu ‘house’: ijo- (loc), n¯ayu ‘village’: n¯ajo- (loc). The genitive
is -t-i (Israel: 1979: 61–4; see the Ko
.
n
.
da forms above).
In Pengo
and Man
.
da -t- is added
as an augment to non-human nouns, Pe.
mar ‘tree’:
ma(r)-t-i
ŋ
(acc), ma(r)-t-o (loc), ma(r)-t-i (gen), mar-(t)-a
ŋ
(instr–loc); in the plural
key-ku-t-a
ŋ
‘from the hands’; with sandhi in n¯az ‘village’: n¯a
.
t-i (gen), n¯a
.
t-i
ŋ
(acc–
dat), n¯a
.
t-o (loc); the -d- variant occurs in monosyllabic stems which do not end in
voiceless consonants, e.g. k¯al ‘leg’: k¯al-di (gen), k¯al-d-i
ŋ
(acc–dat). Man
.
da has vay-ti-k
11
Actually this is a combination of two augments, tt + i. For -i see below.
12
It appears that the change of PD
∗
-
.
t- > -r- in South Dravidian II must be a recent change in the
nominative after the paradigm was formed on the
.
t-ending base (see section 4.5.5.3–4).