Krishnamurti Bhadriraju. The Dravidian Languages
Подождите немного. Документ загружается.
4.5 Historical
phonolo
gy: consonants
161
monosyllables. The formative syllable can be -V
2
,V
2
-P, V
2
-L, V
2
-NP, V
2
-PP etc. (see
figure 4.1). I explained the mechanism as follows:
16
Stage II: The second syllable, being heavy (i.e. CVC as opposed to light
CV), carried stress only when V
1
and V
2
were of the same quality or when
V
2
was [+low]. The sequences V
1
and V
2
were:
∗
(C) V
1
C-V
2
...
ii
uu
aa
i/e a
u/o a
When the consonant between V
1
and V
2
was R (a liquid, a phonetic con-
tinuant ([
r] ←/
∗
t/), or a flap r,
.
r (<
∗
r,
∗
.
z,
∗
l,
∗
.
l)), V
1
tended to become
weaker in articulation, being closely followed by more prominent V
2
across
a lighter consonant. This led to the lengthening of V
2
and the loss of V
1
,
producing initial apicals and consonant clusters followed by a long vowel
which was identical with V
1
in quality. There is evidence that, before this
happened, e$a and o$a also became
∗
e$e and
∗
o$o, respectively. Conse-
quently,
∗
mo
.
z-ang- >
∗
mo.z-ong >
∗
m
.
z¯o-ng- ‘to roar ...
Where the qualities of V
1
and V
2
were not identical in quality, V
2
tended
to be weakened and lost, bringing the root-final consonant R into contact
with the suffix consonant, e.g.
∗
mi
.
z-u-ng >
∗
mi
.
z-ng- ‘to swallow’. The
vowel sequences were:
∗
(C) V
1
C-V
2
..
iu
eu/i
au/i
ui
ou/i
Presumably,
∗
i merged with u in V
2
position. The loss of
∗
u in the second
syllable is a much older tendency, found in all subgroups of the Dravidian
16
During my fieldwork on the Ko
.
n
.
da language, I noticed that the dialect of my informant in the
Araku valley did not have initial consonant clusters, but he regularly stressed the second syllable
when the first syllable was short, e.g. sar´as ‘snake’, mar´an ‘tree’, pe
.
r´elzin`ad ‘it is exploding’.
When I visited the Sova village to the northwest, I found the speakers there pronounced these
forms with initial clusters, sr¯asu, mr¯anu, p
.
r¯elzinad. This gave me the insight that it was a single
jump from stressed second syllable to vowel lengthening of the syllable and loss of the first
syllable. Also notice pe
.
r´el- where the second syllable (originally
∗
a) became e before ‘apical
displacement’ (see Krishnamurti 1969a: 23).
162 Phonolo
gy: historical
and comparative
family. Therefore it appears that vowel lengthening operated in the envi-
ronments that preserved V
2.
Stage III: The outputs of Stage II with patterns like (C)VR-C ...were
subjected to assimilation in Subgroup I (i.e. SD I), but to metathesis in
Subgroup 2 (i.e. SD II). The function of metathesis was to preserve the
root syllable from threatened assimilation of the final consonant. Therefore
∗
(C)VR-C ...>(C)RV-C ....
Stage IV: During this stage, the consonant clusters formed through
the earlier processes were again simplified by loss of R in Telugu (see
Krishnamurti 1961,
§1.145) and loss of C in the other languages of Sub-
group 2 (SD II).
The rules of apical displacement
must have started at least
two thousand years ago, and are still going on in some of the languages
of this subgroup, particularly in Kui. (Krishnamurti 1978a: Appendix,
pp. 18–19)
Kolami and Parji of Central Dravidian borrowed forms which had undergone metathesis
in Early Telugu. These forms still show word-initial
.
d- that Telugu merged with d- later,
indicating
the early chronology of their bor
rowing, e.g. Te.
duwwu ‘tiger’ (<
∗
.
duww-
<
∗
.
zu-ww- <
∗
u
.
z-(u)-w-), > lw Kol. (Kinwa
.
t dial)
.
d¯u-, Pa.
.
d¯u- (pl
.
duvul). SD I: Ta.
u
.
zuvai ‘tiger’, SD II: Go.
.
d¯u,
.
duwwal,
.
duw¯al ‘panther’ [692]. Te.
.
dig(g) u > digu
‘to descend, get down’, Go.
.
dig-, Ko
.
n
.
da
.
dig-: Kol.
.
dig-,
.
digg-, dig-, Nk. (Ch.)
.
dig-,
but Pa. i
.
rv-, Oll. i
.
rg-, Gad. i
.
dg-; SD I: Ta. Ma. i
.
zi, Ka. i
.
li [502]. But the Central Dravidian
languages have not borrowed any items with consonant clusters; by this evidence we
believe that these forms have not resulted from a shared innovation but from borrowing.
There is an extension of the above rules to a few cases even where C
2
is not an apical
consonant (in one or two languages of this subgroup), e.g. Te. adi (noml) ‘she, it’ : d¯an-i-
(obl; <
∗
ad-an-), idi ‘this one’(n-m sg), d¯ıni- (ob
l;
<
∗
id-an-), ¯edi ‘which
one?
’: d¯eni- (ob
l;
<
∗
ed-an-); Ko
.
n
.
da adi (noml) ‘that woman or thing’: dani- (obl), idi (noml) ‘this woman
or thing’ : deni- (obl; <
∗
id-an-), so also in the plural Te. awi : w¯ani, iwi : w¯ıni, ¯ewi: w¯eni-,
etc. Ko
.
n
.
da also has avi: wani, ivi:veni, etc. Similarly corresponding to Te. w
˜
¯
a
.
du: w¯ani-
‘that man’, w
˜
¯ı
.
du : w¯ıni- ‘this man’, ew
˜
a
.
du : ewani-/w¯eni- ‘which man?’, Ko
.
n
.
da has
w¯an
ru: wani, w¯enru : weni-, ayenru : ayeni-. The first two are from PD
∗
awandu/awanru,
∗
iwandu/iwanru. Note that Telugu kept the distal and proximal difference by maintaining
the vowel qualities also in the metathesized forms. The developments in Ko
.
n
.
da were
normal since Ce- of the proximal forms from
∗
iC-a would not make them homophonous
with the interrogative forms, as would have happened in Telugu. The interrogative forms
in Ko
.
n
.
da are not part of the deictic paradigm as in Telugu.
17
17
In Ko
.
n
.
da the interrogative root is ay-: ay-en-/ay-enru ‘which man?’ ay-ed ‘which woman/thing?’,
ay-er ‘who? (m and f)’, ay-ev ‘which? (non-pers)’.
4.5 Historical
phonolo
gy: consonants
163
In Classical Telugu the verb an- ‘to say’ has metathesized doublets an-an ∼ ˜n¯a-n
‘to say’, ana-ka ∼ n¯a-ka ‘without saying’, anaw-˜u
.
du ∼ n¯a-w˜u
.
du ‘after saying’.
No other language has metathesized forms involving -n as C
2
. There is one case
with a metathesized -m- and one with -g-asC
2
in South Dravidian II (excluding
Telugu):
(95) PD
∗
mak-antu ‘male’,
∗
mak-a
.
l ‘female’. SD I: Ta. makan ‘son’ (>m¯an,
m¯o
n), maka
.
l ‘daughter’, makka
.
l ‘children’, Ma. makan, m¯on ‘son’, m¯o
.
l
‘daughter’, makka
.
l ‘children’, Ko. mog ‘child, wife’, To. mox ‘child, son,
male, daughter’, Ko
.
d. m¯ov˙e
n
‘son’, m¯ova ‘daughter’, makka ‘children’, Ka.
maga, magan ‘son, male’, maga
.
l ‘daughter’, makka
.
l ‘children’, Tu. mage,
m¯onu ‘son’, maga
.
lu ‘daughter’, makka
.
l ‘children’; SD II: Te. magan
ru ‘son’
(inscr), maga, moga ‘male’, mag˜a
.
du ‘husband, brave person’, maguwa
‘woman’, Go. miy¯a
.
r ‘daughter’, Ko
.
n
.
da maga ko
.
ro ‘male child’, g¯a
.
ru, g¯alu
‘daughter’, Pe. Man
.
da g¯a
.
r (<
∗
mg¯a
.
l), Man
.
da n¯a-mg¯a
.
r ‘my daughter’; ND:
Malt. maqe ‘boy’, maqi ‘girl’ [4616].
This change is restricted to Ko
.
n
.
da, Pengo and Man
.
da. Another form involving -m-is
Ko
.
n
.
da m¯ul (m¯u
.
t-) ‘to urinate’, m¯ulku ‘urine’, Kui m¯ul-(m¯u
.
t-), Kuvi m¯u
.
nk-v.,mr¯uka,
m
.
n¯uka n., Pe. m¯u
.
nku,Ma
.
n
.
da m¯u
.
nke n. The Central Dravidian and North Dravidian
languages provide evidence for the reconstruction of
∗
um(b)-u
.
l v.; South Dravidian I has
forms beginning with m- in modern dialects: Ta. m¯o
.
l v., Ma. mo
.
l
.
lu n., Ko. mo
.
l n., Ir.
ma
.
l ‘to urinate’ [644]. This is a very intriguing etymology.
4.5.8 Consonant groups -CC, -C-CC
Only three kinds of consonant groups occurred in postvocalic position in Proto-
Dravidian, namely geminates (of obstruents and some non-obstruents) and clusters of
homorganic nasal + stop, homorganic nasal + geminate stop. A single voiceless stop
occurring in postvocalic position (i.e. C
2
of (C
1
)VC
2
-) in any of the Dravidian lan-
guages can be traced back to a geminate stop in Proto-Dravidian. If it was a single stop,
it would appear in all languages as a lenis consonant. The Kui–Kuvi–Pengo–Man
.
da
subgroup does not contrast between single and double consonants. The Nilgiri langua-
ges which lose the final enunciative vowel also do not show the contrast between P
and PP.
4.5.8.1 -PP
In most of the languages, other than Tamil and Malay¯a
.
lam, a geminate stop is retained
after a Proto-Dravidian short vowel, but it is simplified to P after a long vowel, e.g. Te.
Ka. ¯a
.
du ‘to play’, ¯a
.
ta ‘game’, Ko. ¯a
.
r- v.i., ¯a
.
t- v.t.,To. ¯o
.
d- v.i., ¯o
.
t- v.t., Ko
.
d. ¯a
.
d- v.i., ¯a
.
t- v.t.:
Ta. Ma. ¯a
.
tu v.i. ‘to play’, ¯a
.
t
.
tu v.t., ¯a
.
t
.
t-am (<
∗
¯a
.
t + tam) ‘game, dance’ [347]. In most of
164 Phonolo
gy: historical
and comparative
the languages, the contrast of single vs. double stop developed into voiced vs. voiceless,
particularly in the environment of
¯
V].
4.5.8.1.1
∗
-pp
(96) PD
∗
kupp-V ‘heap’, SD I: Ta. kupp-ai, kupp-am, kupp-al ‘heap’, Ma.
kuppa,Ko.kip ‘rubbish’, To. k¨ıp id., Ko
.
d. kupp¨ı ‘heap of dung’, Ka. kuppe,
guppe ‘heap’, kuppu ‘to heap up’, Tu. kuppε, guppε, kippε; SD II: Te.
kuppa, Go. kupa, kuppa,Ko
.
n
.
da kupa, Kui, Pe. Man
.
da kupa, Kuvi k¯upa,
kuppa; CD: Kol. kupp kal- ‘to gather’, Pa. kuppa ‘stack’, koppa ‘hillock’,
Oll. kupp¯a ‘heap’; ND: Ku
.
r. xop- ‘to form into a pile’, Malt. qop- ‘to heap
up’ [1731a].
Also see ety. (74) and (100) (for South Dravidian I). Here, the Proto-Dravidian root
could be
∗
kup- with a derivational suffix -ay which is also reconstructable for the whole
family.
4.5.8.1.2
∗
-tt
18
(97) PD
∗
pitt- ‘to fart’. SD II: Te. pittu ‘to fart’, Go. pitt-/p¯ıt -v.,p¯ıtu n., Ko
.
n
.
da
p¯ı t - v.i., p¯ıtu n., Kui, Kuvi, Pe. Man
.
da p¯ı t -; ND: Ku
.
r. Malt. p¯ı t - [4167].
It is interesting to note that this has no cognates in South Dravidian I and Central
Dravidian. It could be related to PD
∗
piy/
∗
p¯ı ‘excrement’ [4210] or
∗
pi-n/
∗
pi-t ‘back,
buttocks’ [4205] (TVB:
§1.194, p. 82), represented in all subgroups.
4.5.8.1.3
∗
-tt
(98) PD
∗
watt-/wat-V- ‘to dry up’. SD I: Ta. varru ‘to evaporate’, Ma. varruka,
Ko. vat-, Ko
.
d. batt-, Ka. battu, baccu,Tu.battel¨ı ‘leanness’; SD II: Te.
wa
.
t
.
tu ‘to dry up’, Go. vatt-, Ko
.
n
.
da vaR-, Kui vas-, Pe. Ma. vac-: CD: Kol.
vat-, Pa. vett-, ve
.
t
.
t- ‘to wither’, vetip-(vetit-) v.t. ‘to dry up something’,
Oll. va
.
t-, Gad. va
.
t
.
t- v.i., va
.
tp- v.t. ‘to dry up in sun’; ND: Ku
.
r. batt- ‘liquids
decrease by evaporation’, Malt. bat- ‘(water) to dry up’, Br. b¯arring ‘to
become dry’, b¯arif- ‘to make dry’ [5320].
Most of the languages of South Dravidian also preser
ve the forms traceable to the root
∗
wat-V-. Note that
∗
tt merges with a dental
∗
-tt in Modern Tamil, Kanna
.
da, Tu
.
lu, Ko
.
dagu,
Gondi, Kolami, Parji (dialectally) and Ku
.
rux–Malto; it merges with
∗
.
t
.
t in Telugu, dialec-
tally in Parji, Ollari and Gadaba. It is preserved distinctively as a voiceless alveolar trill
18
The past-tense suffiix -tt- is not considered here. Following a root ending in -y or a front vowel,
the tense allomorph -tt- becomes -cc- in Modern Tamil, Malay¯a
.
lam, Ko
.
dagu, Kota and Toda
(see Subrahmanyam 1983: §26.5.1, pp. 366–70).
4.5 Historical
phonolo
gy: consonants
165
/R/ contrasting with /
r/ in Ko
.
n
.
da; in the other members of the subgroup it merges with a
palatal or sibilant c/s. The last one seems to be the only shared innovation. The gradual
elimination of
∗
t in most of the languages has led the six-point stop system to become
five-point like most of the languages in the Indian area. Also study ety. (16) and (77).
4.5.8.1.4
∗
-
.
t
.
t
(99) PD
∗
ko
.
t-∼ ko
.
t
.
t-/ko
.
t-V ‘to beat’. SD I: Ta. Ma. ko
.
t
.
tu ‘to tap, to beat, hammer,
pound’, Ta. ko
.
tu ‘to thrash’, Ko. ko
.
tk-, To. kw¨ı
.
tk-, Ko
.
d. ko
.
t
.
t- ‘to tap’, Ka.
ko
.
t
.
t-a
.
na n. ‘beating the husks of paddy’, Tu. ko
.
dapuni ‘to hammer’, Te.
ko
.
t
.
t- v.t., Go. ko
.
t- ‘cut with axe, strike with horn’, kohk- ‘to pound’, Pe. ko
.
t-
‘to thresh with flail’; CD : Pa. ko
.
t
.
t-, Oll. ko
.
t-; ND: Ku
.
r. xo
.
t
.
t- ‘to smash’,
xo
.
tr- ‘to be broken’, Malt. qo
.
t- ‘to break’, qo
.
tr- ‘to be broken’ [2063].
Some of the languages show -
.
d-(<
∗
-
.
t-) which alternated with
∗
.
t
.
t when followed by a
derivational suffix beginning with a vowel.
4.5.8.1.5
∗
-cc
(100) PD
∗
mucc-∼
∗
muc-V-∼
∗
muy- ‘to cover’. SD I: Ta. muccu ‘to cover’, Ko.
muc-, To. m¨uc-, Ko
.
d. mucc-, Ka. muccu,Tu.muccuni;Te.muccu ‘to close’
(archaic), Go. mucc-, muc- ‘to cover’, Ko
.
n
.
da, Kui mus- ‘to cover’, Kuvi
muh-, muc-, Pe. Man
.
da muc-; CD: Nk. mus-, Nk.(Ch.) muc-; ND: Ku
.
r.
mucc- ‘to close door’, mus
u
g- ‘to wrap up’, Malt. muc- ‘to close’, musg-
‘to pack up’, Br. must ‘shut, closed’ [4915].
Notice again the alternation between c and s (geminate and single) in several languages.
There is an etymologically related set requiring PD
∗
muy/
∗
m¯uy presupposing the weak-
ening of -c to -s and -
∗
y in Proto-Dravidian itself. DEDR lists this group under the same
etymology (also see TVB:
§1.193, p. 82).
4.5.8.1.6
∗
-kk-
(101) PD
∗
wekk- ‘to hiccup’. SD I: Ta. vikku ‘to hiccup’, n. ‘hiccup’, Ma. vikkuka,
v.i., vikku n. ‘impediment in speech’, vikkam ‘stammering’, To. p¨ık-‘to
cough’, Ka. bikku ‘to pant, sob, hiccup’, n. ‘hiccup’, Tu. bikkuni ‘hold
one’s breath’, bikk¨ı n. ‘holding one’s breath, sob’: SD II: Te. vekku, vegacu
‘to hiccup’, vekk-ili ‘hiccup’, Kui vek- ‘to cough’, Kuvi vek- ‘to choke
when drinking or eating’; CD: Kol. veksi ‘hiccup’; ND: Ku
.
r. bekkh- ‘to be
choked’, Malt. beq- [5383].
As stated earlier, Proto-Dravidian stop geminates occur root-finally (with enunciative
-u) where the root is also a free form or when it is followed by derivational suffixes
166 Phonolo
gy: historical
and comparative
beginning with vowels. If a formative suffix follows there is no gemination, e.g. see ety.
(99)–(101).
4.5.8.2 -PP < -R-P, -BB < -R-B
There occur in some languages of South Dravidian I, South Dravidian II and Central
Dravidian double voiceless and voiced stops which have resulted from assimilation of a
root-final consonant (non-nasal sonorant),
∗
r,
∗
r,
∗
l,
∗
.
l,
∗
.
z with a stop which is part of the
derivative suffix. These are originally trisyllabic bases of the type (C
1
)V
1
C
2
-V
2
-PP/-NP,
preserved in Tamil and Malay¯a
.
lam. In such cases V
2
(= u) is lost and C
2
is regressively
assimilated to the following P or B. The resulting disyllabic forms are found largely in
Telugu and Kanna
.
da (and also Tu
.
lu and Ko
.
dagu through contact with Kanna
.
da). The
South Dravidian II and Central Dravidian languages which also attest to the loss of V
2
generally preserve the unassimilated clusters. Examples: Te. ippa Bassia longifolia, Kol.
ippa, Ka. ippe,Tu.ippe; irippe ‘the olive tree’, Ko
.
n
.
da ipa (maran): Pa. irup, irpa, Gad.
irpa, Go. iru, irup, i
.
rup, hirp; Kui, Kuvi irpi;Ta.iruppai ‘mahua tree’, Ma. iruppa [485].
A good case for voiced stop gemination is Te. taggu ‘to decrease’, Ka. taggu, targu, ta
.
zgu
‘to be, become low’, t¯a
.
z ‘being low’, Tu. tagguni ‘to be humble’, t¯aruni ‘to sink in’, Ko.
tag,To.tog. These forms presuppose an underlying
∗
ta
.
z-unk- from PD
∗
t¯a
.
z ‘to fall, be
low’; cf. Ta. Ma. t¯a
.
z,To.t¯oy, Ka. t¯a
.
z, t¯a
.
l ‘to be low, decrease, decay’, Br. da
.
r- ‘to go down’
[3178]. (For further discussion with examples, see TVB
§§1.173–1.185, pp. 74–9.) The
loss of a high vowel in a heavy syllable, i.e. ip$pa < ir$pa <ir$ppa-<ir$up$pa- seems
to be a typological feature of most of the languages (outside Tamil and Malay¯a
.
lam) to
promote the favourite stem types, (C)
VC-, or (C)VCC-. It is not a shared
innovation.
4.5.8.3 Gemination of sonorants
PD
∗
r and
∗
.
z do not geminate in any language. Although doubling of the other sonorants
like nasals m, n,
.
n, laterals l,
.
l, semivowels w, y is found in different languages, there is
no evidence for their reconstruction in Proto-Dravidian. There are only a few cases of
single vs. double contrast in the case of the two laterals
∗
l and
∗
.
l (see Emeneau 1970a:
94–5). Emeneau cites
∗
nelli Phyllanthus emblica; this occurs with gemination in Tamil,
Malay¯a
.
lam, Ko
.
dagu, Kanna
.
da, Tu
.
lu, Telugu and Parji nelli,butKo
.
n
.
da neli,Kuine
.
di
[3755]. For
∗
.
l
.
l, Emeneau gives PD
∗
pi
.
l
.
lay ‘child’, which occurs with a double consonant
in Tamil, Malay¯a
.
lam, Ko
.
dagu, Kanna
.
da, Tu
.
lu, Telugu, Kolami, Naiki, Gondi, Kuvi,
Ku
.
rux and Brahui. Ko
.
n
.
da and Malto have a single -l-. This can be taken as a clear case
of PD
∗
.
l
.
l (also see TVB:
§§1.198–1.202; Subrahmanyam 1983: §§34–5).
4.5.8.4
∗
-NP
Following open-syllab
led roots (C
1
)V
2
-, or stems that already had the first layer of for-
matives -V
2
as in the type (C
1
)V
1
C
2
-V
2
-, a series of morphs occurred with grammatical
4.5 Historical
phonolo
gy: consonants
167
alternation, namely -NP and -NPP in the case of verbs in which -NP signalled intransitive
+ non-past/past and -NPP, the corresponding transitive + non-past/past. Even within
Proto-Dravidian, the tense meaning was lost first in most of the area and the morphs sig-
nalled only voice, which continued in Tamil–Malay¯a
.
lam–Ko
.
dagu of South Dravidian I
and the Ko
.
n
.
da–Kui–Kuvi–Pengo–Man
.
da group of South Dravidian II. Both the voice
and tense meanings survive in the conjugation of certain verbs in Tamil–Malay¯a
.
lam–
Ko
.
dagu (Krishnamurti 1997a). This problem will be discussed in detail in section 7.3.
In the case of nouns -NP signals a substantive and -NPP an adjective.
It happens that in several cases the etymological boundary is between N and P (see
figure 4.1), i.e. the root is a (C
1
)VC
2
- type where C
2
is N. In that case, the following
formative is
-P. There is no way to separate
this group from the former e
xcept through
comparison of cognates.
In trisyllabic stems reconstructed for Proto-Dravidian, the vowel of the second syllable
(V
2
= u) is lost leading to the assimilation of C
2
(usually
∗
.
z,
∗
l,
∗
.
l) with the following
nasal in many languages of South
Dravidian. I have dealt with
the above types in detail
in TVB:
§§1.160–1.172. Except for Tamil and Malay¯a
.
lam, most other languages lose the
nasal after a long vowel, i.e.
¯
V] NB >
¯
V]B and the sequence -NP [NB] is preserved after
a short vowel. Kota loses the nasal after a long vowel but retains it after a short vowel.
Toda regularly loses the preconsonantal nasal irrespective of the length of the preceding
vowel. In Malay¯a
.
lam where NP stands for
∗
-nk,
∗
-˜nc,
∗
-nt there is progressive assimilation
as -˙n˙n,-˜n˜n and -nn suggesting that the postnasal stops were voiced allophonically at an
earlier stage. The alveolar sequence
∗
-nt also produces a dental
geminate suggesting an
earlier merger of
∗
-nt with
∗
-nt. The retroflex and labial sequences (-
.
n
.
d,-mb) are not
assimilated. Assimilation was attested in the inscriptions even by the tenth century. The
alveolar sequence becoming /nn/ was attested in the eleventh and twelfth centuries. The
sequence /nt/ is always assimilated when it is a past-tense marker. Formative /nt/ shows
assimilation after (C)VC-V and not after the radical vowel, i.e. (C)
V-, (C)V-. In the case
of velar and palatal sequences assimilation after the radical vowel is much less frequent
than after the formative vowel. About 75 per cent of the velar sequences and 50 per
cent of the palatal sequences show progressive assimilation. In compounds there is no
nasal assimilation, if the nasal and stop are separated by a word boundary, e.g mu
.
za
.
n-k¯al
‘knee’, mu
.
za
.
n-kai ‘elbow’. The assimilation is said to be a sound change still in progress
(Subrahmanyam 1983:
§22.5, pp. 309–12, V. I. Subramoniam 1972, N. Kumaraswami
Raja 1980). In the case of Old Telugu after a long vowel, the preconsonantal nasal was
replaced by nasalization of the preceding vowel; the nasalized vowels became oral in
Middle and Modern Telugu, e.g. PSD
∗
¯unku ‘to swing’ > OTe.
˜
¯ugu > Mdn. Te. ¯ugu.
4.5.8.4.1 -mp [mb] Kanna
.
da loses the nasal after a long vowel, but retains the se-
quence -mb after a short vowel. Ko
.
dagu and Tu
.
lu retain -mb. In Telugu and the other
168 Phonolo
gy: historical
and comparative
languages of South Dravidian II and in the Central Dravidian languages, there was pro-
gressive assimilation of -mb to -mm after a short vowel and -m after a long vowel. A
similar change occurs optionally in Tu
.
lu also after a short vowel, but they could as well
be independent changes. In North Dravidian the sequence is retained, but there are no
clear etymologies.
(102) PD
∗
nampu (<
∗
nay-mp-) ‘to long for, trust’. SD I: Ta. nampu, Ma. nam-
puka,Ko.namb-, To. nob,Ko
.
d. namb-, Tu. nambuni, nammuni;SDII:
Te. nammu,Ko
.
n
.
da nami-, Kui nam-, Kuvi namm-/nam- [3600]. Cf. Ta.
naya ‘to desire’, naccu ‘to be agreeable’, Te. nayamu ‘good’, naccu ‘to be
agreeable’ etc. [3602, 3576].
(103) PD
∗
p¯ampu ‘snake’. SD I: Ta. p¯ampu n. adj p¯appu, Ma. p¯ampu,Ko.p¯ab,
To. p¯ob,Ko
.
d. p¯amb¨ı, Ka. p¯avu, h¯avu,Tu.h¯avu (<lw. Ka.); SD II: Te. p¯amu
n., p¯apa- adj; CD: Kol., Nk. Nk. (Ch.) p¯am (lw < Te.), Pa. Oll. b¯am, Gad.
b¯amu, b¯amb [4085]. Te. n¯omu ‘a religious vow’ is from
∗
n¯on-pu, cf. Ta.
n¯o
n ‘to practise austerities’, Ta. Ma. n¯onpu, n¯ompu ‘ceremonial fasting’,
Ka. n¯on v., n¯ompu n., Tu. n¯ombu ‘fasting’ [3800].
4.5.8.4.2
∗
-nt [nd]
(104/46) PD
∗
c¯ıntu ‘date palm’,
∗
c¯ıntt- adj. SD I: Ta.¯ıntu, ¯ı ˜ncu ‘date-palm, Phoenix
farnifera’, ¯ıccam pa
nai ‘wild date-palm’, Ma. ¯ıntal, Ka. ¯ıcal,Tu.¯ı ˜ncil¨ı,
¯ıcil¨ı (<
∗
¯ıntt-), ¯ınd¨ı ‘sago-palm’; SD II:Te.
˜
¯ıta,
˜
¯ıdu, Go. s¯ındi, h¯ındi,
¯ındi,Ko
.
n
.
da s¯ıtel,Kuis¯ıta, Kuvi s¯ındi; CD: Pa. c¯ınd, Gad. s¯ındi; ND: Ku
.
r.
k
˜
¯ıd¯a ‘palm tree’. || Skt. hint¯a
.
la ‘marshy date tree’, Pkt. sind¯ı ‘date palm’
[2617, CDIAL 14093].
This is an important etymology for two reasons: (i) When Proto-Dravidian
∗
c- was lost
in the southern group (SD I), it passed
through two intermediate stages
∗
s- and
∗
h-; both
these stages are preserved in Pkt. and Skt. loanwords;
∗
ntt changes to
∗
ncc- following a
palatal vowel in the languages which show a palatal. Tamil and Malay¯a
.
lam lose the nasal
before a geminated stop (see section 4.5.8.5). (ii) It attests the change of Proto-Dravidian
∗
c-tok- in North Dravidian.
In some of the languages, trisyllabic bases become disyllabic with the loss of -u-
before -nt-:
19
(105) PD
∗
maruntu ‘medicine’, adj
∗
maruntt-. SD I:
Ta.
maruntu n., adj marutt-,
marutta
n ‘physician’, Ma. marunnu,Ko.mad,To.mad,Ko
.
d. madd¨ı, Ka.
19
The past-tense
∗
-nt- is also found in SD I languages and in Central Dravidian (see Emeneau
1967, Subrahmanyam 1983: §26.5). This sequence is palatalized to -nj- after front vowels in
South Dravidian I. In Central Dravidian Parji also independently palatalizes -nd-to-nj- but this
is a recent development.
4.5 Historical
phonolo
gy: consonants
169
mardu, maddu,Tu. mard¨ı; SD II: Te. ma-ndu ‘medicine’, Go. mat, matu,
matta; CD: Pa. merud (pl merdul), Oll. mardil, Gad. marid; ND: Ku
.
r.
mandar, Malt. mandru [4719].
In
∗
c¯ıntu ‘to blow the nose’ which is from
∗
c¯ı m -t- based on comparative evidence; the
root is preserved in Ka. s¯ı n -, Tu. s¯ımp¨ı ‘to blow the nose’, Kor. c¯ımpu [2618].
4.5.8.4.3
∗
-nt- [nd/nr] See ety. (3)
∗
en-tu ‘sun’, (4b)
∗
nin-t- ‘to be full’, (70)
∗
˜n¯entu
‘time, day’. In a large number of etymologies it seems that
∗
-nt arises in sandhi by
combining a root-final alveolar nasal with a following dental formative suffix. In this
sequence the voiced alveolar stop
∗
d/
∗
r merges with a dental, a retroflex or a palatal stop.
Mer
ger with a palatal voiced stop
occurs in Kui
–K
uvi
–P
engo
–Man
.
d
a, which is a shared
innovation. PD
∗
nt represented as OTa. -nr becomes
.
n
.
n in Modern Tamil (o
.
n
.
n¨ı ‘one’
from older on
ru); in Malay¯a
.
lam it develops to a dental geminate -nn- (apparently through
an intermediate merger stage of -nd-); Kota and Toda preserve
d with the loss of the
nasal; in Old Telugu it was recorded as -n
r in inscriptions and merged with -
.
n
.
d later;
Gondi has merged it with a dental or retroflex in different dialects; in Central Dravidian,
Parji (dialectally) and Gadaba represent this as -
.
n
.
d. In Kanna
.
da, Ko
.
dagu, Kolami, Naiki
and Ku
.
rux–Malto,
∗
d merges with a voiced dental stop -d. In the Kui–Man
.
da subgroup
and in Tu
.
lu,
∗
-nr merges with -nj. Except for the Kui–Man
.
da subgroup, none of the
above developments can be called a shared innovation. This means that the cluster was
preserved until all the subgroups were split into independent languages and the phoneme
got eliminated through mergers later, owing to the typological pressure of converting
the six-point stop system to a five-point one like the other languages in the area.
(106) PD
∗
on-tu ‘one’. SD I: Ta. onru (>o
.
n
.
nu), Ma. onnu,Ko.od,-ond,To.
w¨ı
d,Ko
.
d. ond¨ı, Ka. ondu,Tu.o ˜nji: SD II: Te. o
.
n
.
du, Go. u
.
n
.
d¯ı , und¯ı,Ko
.
n
.
da
un
ri; ND: Ku
.
r. ¯on, ¯ond, Malt. -ond,Br.asi
.
t [990d].
4.5.8.4.4
∗
-
.
n
.
t[
.
n
.
d]
(107) PD
∗
ka
.
n-
.
tV ‘warrior’. SD I: Ta. ka
.
n
.
tan ‘warrior, husband’, ka
.
n-avan
‘husband’, Ma. ka
.
n
.
tan ‘the male,
of a cat
’, ka
.
navan ‘husband’, Ko. ga
.
n
.
d
‘male’, To. ko
.
d
.
n ‘Ba
.
daga husband’, Ko
.
d. ka
.
n
.
d¨e ‘male (of dogs and other
animals)’, Ka. ga
.
n
.
du ‘bravery’, ga
.
n
.
da ‘strong person, husband’, Tu. ga
.
n
.
du
‘male, stout’, ka
.
n
.
da
.
ni ‘husband’; SD II: Te. ga
.
n
.
du ‘bravery’, ga
.
n
.
du-billi
‘male cat’; CD: Nk. ga
.
rek ‘man, male’; ND: Malto ge
.
n
.
da ‘male’ ||>Skt.
ga
.
n
.
da-, ga
.
n
.
dira- ‘hero’ [1173].
See ety. (91) for
∗
.
n
.
t in the second syllable.
(108) PD
∗
pi
.
n
.
t- <
∗∗
pi
.
z-nt- ‘to squeeze’. SD I: Ta. pi
.
n
.
ti ‘what is squeezed’, Ko
.
d.
pu
.
n
.
d-, Ka. pi
.
n
.
du, hi
.
n
.
du ‘to squeeze out’, Tu. pur˜ncuni, pure˜ncuni ‘to
170 Phonolo
gy: historical
and comparative
squeeze, as a lemon’, pu
.
n
.
diyuni, ‘to wring a wet garment’, pi
.
n
.
d, pu
.
n
.
di
‘oilcake’; SD II: Te. pi
.
n
.
du ‘to squeeze, milk, wring’, pi
.
n
.
di ‘oilcake’; CD:
Kol. pin
.
d-, p¯ı n
.
d-, Nk. p¯ı
.
n
.
d- [4183].
Languages belonging to all subgroups also have cognates for the same meaning from
∗
pi
.
zi [4183]. It must be admitted that there are hardly any cases where -
.
n
.
t is a primary
suffix, i.e. where it is not derived from sandhi between the root final
∗
.
n with a following -t
(P-suffix), or from
∗
-
.
z + nt- >-
.
n-
.
t-. Other cases include
∗
pa
.
n
.
tu ‘fruit’ <
∗
pa
.
z-nt- [4004],
u
.
n
.
tu ‘to be’ <
∗
u
.
l-nt- [697].
4.5.8.4.5
∗
-˜nc [˜nj]
(109) PD
∗
tu ˜nc- ‘to sleep’. SD I: Ta. Ma. tu ˜ncu (cf. tuyil ‘to sleep’, n. ‘sleep’):
SD II: Go. sunj-, hunj-, unj-, Ko
.
n
.
da sunz-, sus-, Kui sunj-, Kuvi h¯unj-, Pe.
Man
.
da hunj-; CD: Pa. tu ˜n-, cu ˜n- ‘to sleep’, tu ˜n-ip- ‘to put to sleep’, Oll.
tu ˜n-, tuy
ŋ
g-, Gad. tu
ŋ
-; ND: ?Br. t¯ul- ‘to sit’ [3291].
If
∗
tuy- was the underlying root
(past:
∗
tuy-nt-) we need to posit palatalization
for Proto-
Dravidian itself as in
∗
tu(y)-˜nc- plausib
ly from
∗
cu(y)-˜nc- <
∗
cuy-ntt-. There
are long
vowel forms which look to
∗
cuy-/
∗
tuy- as the ultimate root: To. t¨us, Gad. (P) tuy
ŋ
g-, Kui
s¯us-, Br. t¯ul-. By putting all these in the same group, DEDR seems to suggest this. There
is another related group of etyma,
∗
uy-al,
∗
tuy-al ‘to wave, swing’, ¯u-nku ‘to swing’,
[731, 3376b],
∗
t¯u-nku ‘to shake, swing, sleep’ [3376a], ultimately traceable to the root
∗
cuy- on account of t-/Ø- alternation. Other forms include
∗
a ˜ncu ‘to fear’ [55],
∗
na ˜ncu
‘poison’, na ˜ncc- adj [3580],
∗
ne ˜ncu ‘heart, core’ [3736].
4.5.8.4.6
∗
-nk [ng] This sequence can be primary derivable from PD
∗
-nk or from a
combination of root-final -
.
n or -m with a following stop suffix
∗
-k.
(110) PD
∗
ponku ‘to boil’. SD I: Ta. po
.
nku ‘to boil, bubble up’, Ma. po˙n˙nuka,Ko.
pog- ‘to boil over’, po
ŋ
g- ‘to increase’, To. p¨ıg-, Ko
.
d. po
ŋŋ
-, Tu. bo
.
nguni
‘to be distended’, ponga
.
dε ‘proud flesh’; SD II: Te. pongu ‘to boil’, n.
‘boiling, joy, pride’, Go. p¯o
ŋ
- ‘to swell’, Ko
.
n
.
da po
ŋ
i- ‘to swell up’, Kuvi
po
ŋ
g- ‘to swell’; CD: Kol. po
ŋ
g- ‘to boil over’, Nk. po
ŋ
g- ‘to expand’;
ND: Ku
.
r. p¯ux- ‘to swell, as rice in water’, pokpokr- ‘to puff out’, Malt.
po
.
gole- ‘to sw
ell
’, po
ŋ
gj- ‘to be increased’, p¯u
.
gr- ‘to be swollen’ [4469].
Note that Malay¯a
.
lam and Ko
.
dagu show progressive assimilation of
∗
nk to ˙n˙n, suggesting
that [
ŋg] as an intermediate stage was part of the PSD phonetic state.
From the foregoing description the
∗
-NP sequence is preserved in most of the lan-
guages as -NB after
a radical short vowel. Kota optionally and Toda regularly drop the
nasal but retain the voiced stop even after a radical short vowel.