Krishnamurti Bhadriraju. The Dravidian Languages
Подождите немного. Документ загружается.
4.5 Historical
phonolo
gy: consonants
171
4.5.8.5
∗
-NPP
Kumaraswami Raja published a paper and then a short monograph (1969a, b), showing
the necessity to reconstruct a sequence of
∗
-NPP for Proto-Dravidian on the basis of
the correspondences in table 4.8, originally observed in the major literary languages.
Table 4.8. Correspondences of Proto-Dravidian
∗
NP and
∗
NPP in
Tamil–Malay¯a
.
lam and Telugu–Kanna
.
da
Proto-Dravidian Tamil–Malay¯a
.
lam Telugu–Kanna
.
da
a.
∗
NP NP [NB] NB/V
−−−
B/
¯
V
n
−−−
(Telugu)
B/
¯
V
−−−
(Kanna
.
da)
b.
∗
PP PP PP/V
−−−
P/
¯
V
−−−
c.
∗
NPP PP NP/V
−−−
In Tamil and Malay¯a
.
lam
∗
NP and
∗
PP remain unchanged irrespective of the length
of the preceding vowel. In Kanna
.
da and Telugu, which developed contrastive voic-
ing from the earliest known period,
∗
NP,
∗
PP occurred as NB, PP after short vowels.
Only after long vowels Kanna
.
da and Telugu had different developments. But correspon-
dence (c) is not taken care of by (a) and (b). There are cognates in which Tamil and
Malay¯a
.
lam show PP and Telugu and Kanna
.
da show NP (contrasting with NB from PD
∗
NP), e.g.
(111) PD
∗
t¯o
.
n-
.
t
.
t-am ‘garden’ from
∗
t¯o
.
n
.
tu ‘to dig’. SD I: Ta. Ma. t¯o
.
n
.
tu ‘to dig’,
t¯o
.
t
.
tam (<
∗
t¯o
.
n
.
t + tam) ‘garden’, Ko. t¯o
.
d-v.,t¯o
.
tm- n., To. tw
¯
¨ı
.
r-v.,tw
¯
¨ı
.
tm-
n., Ko
.
d. t¯o
.
d-v.,t¯o
.
ta n., Ka. t¯o
.
du ‘to dig, take water out of a well’, t¯o
.
n
.
ta,
t¯o
.
ta ‘garden’,Tu. t¯o
.
duni v., t¯o
.
ta n.: SD II: Te. t
˜
¯o
.
du (> Mdn Te. t¯o
.
du)‘to
draw water up, to scoop out’, t
˜
¯o
.
ta (> Mdn Te. t¯o
.
ta) ‘fenced garden’, Ko
.
n
.
da
.
t¯o
.
n
.
ta ‘garden’; CD: Pa.
.
t
˜
¯o
.
d- ‘to draw water from well’, Gad. t¯on
.
d- ‘to bale
out water’ [3549].
This example clearly shows that a voiceless -
.
t after a nasal, i.e. in -
.
n]
.
t, must be derived
from a geminate -
.
t
.
t of Proto-Dravidian. Tamil, Malay¯a
.
lam, Kota, Toda, Ko
.
dagu and Tu
.
lu
do not show a nasal before a voiceless single -
.
t, which meant that they had lost the nasal,
but retained the geminate, represented as -
.
t
.
t in Tamil and Malay¯a
.
lam and as -
.
t in the
others. It appears that Ko
.
dagu and Tu
.
lu borrowed the form from Kanna
.
da after it had lost
.
n before
.
t. Old Telugu and Old Kanna
.
da preserved the nasal + a single voiceless stop. In
other words, in the languages which had already developed voiced–voiceless contrast in
stops, PD
∗
NP:
∗
NPP developed into NB: NP (Kanna
.
da–Telugu in the above example);
172 Phonolo
gy: historical
and comparative
in the languages which had not developed the voicing contrast (particularly Tamil and
Malay¯a
.
lam) the original differentiation of single vs. double stop remained and the pre-
consonantal nasal was lost. Note that no language preserves the sequence
∗
-NPP, since
phonotactically it is not permitted in any of the descendant languages. Kumaraswami
Raja proposes the following correspondences (∼=nasalization of the preceding vowel):
∗
-nkk: Ko. (
ŋ
)k, Ka. Tu. (
.
n)k;Te.
.
nk/∼k; Kol. Nk. (
ŋ
)k, Oll. ∼k;Ku
.
r.
ŋ
kh,
ŋ
k, ∼x, kk, Malt. (n)q,(n)k,Br.nk; the remaining languages have no
nasal.
∗
-˜ncc: Ko. Ka. Tu. (n)c;Te.nc /∼c, Go. nc; Kol. ns;Ku
.
r. ∼c,Br.nc; the
rest cc, c or s.
∗
-n
.
t
.
t: Ko. Ka
.Ko
.
d. Tu.
(
.
n)
.
t; Te. Go.
(
.
n)
.
t; Kol.
Nk. Pa. Gad. (
.
n)
.
t;Ku
.
r. (
.
n)
.
t;
the rest
.
t
.
t or
.
t.
∗
-ntt: Ka. nt, c, s,Tu.nt/˜nc, ∼t/∼c, t/c; Kol. Gad. nt;Ku
.
r. ∼t. The others
have no evidence of
a nasal.
∗
-ntt: Ko. nt, t, Ka. Tu. (n)t;Te.∼t/∼
.
t/
.
t;Ku
.
r. ∼t; the rest of the languages
have
tt,
.
t or t.
∗
-mpp: Ka. Tu. (m)p; Te. (m) p/∼p; Kol. (m)p;Ku
.
r. (m) p/∼p.
In all the correspondences, Tamil and Malay¯a
.
lam always have geminates with the
loss of the nasal. Phonetically NP is an impossible cluster in Tamil and Malay¯a
.
lam. It
appears that Kota must have split from Pre-Tamil before it lost the nasal, hence it attests
to the presence of NP clusters. Toda must have branched from Pre-Tamil after the loss
of
∗
N; therefore, there
is no single case of a nasal before
a voiceless stop in Toda.
In the
remaining South Dravidian languages, Kanna
.
da and Tu
.
lu show NP sequences regularly;
Ko
.
dagu also shows nasals occasionally. In South Dravidian II Telugu preserves the nasal
before a voiceless stop at some point in its history, followed occasionally by Gondi. The
rest of the languages have lost the nasal before a voiceless stop. In Central Dravidian
most languages show the NP cluster contrasting with NB. I
n North Dravidian Ku
.
rux and
Brahui preserve the nasal before voicelss stops. As in the case of NP, the etymological
boundary may also fall between the root-final N followed by geminate stop suffix (PP).
(112) PD
∗
a
.
n-
.
t
.
t- ‘to adhere, stick’. SD I: Ta. Ma. a
.
t
.
tu v.t. ‘to join, stick’, Ka. a
.
n
.
tu
‘to stick, adhere to’, a
.
nke, a
.
npu ‘smearing’, Tu. a
.
n
.
tuni v.i. ‘to stick’, v.t.
‘to attract’, a
.
n
.
t¨ı ‘gum’; SD II: Te. a
.
n
.
tu v.i., Pe. an
.
d- ‘to adhere’, Kuvi a
.
t-
‘to get stuck’; CD: Kol. a
.
n
.
t- ‘to stick’, a
.
t-, Nk. a
.
t
.
t- [96].
(113) PD
∗
kalanku v.i. ‘to stir’,
∗
kalankku v.t. ‘to agitate, stir’. SD I: Ta. kala
.
nku
v.i., kalakku v.t., Ma. kala˙n˙nuka v.i., kalakkuka, v.t., Ko. kalg- v.i., kalk- v.t.,
To. kalx- v.i., kalk- v.t., Ko
.
d. kala
ŋ
g- v.i., kalak- v.t., Ka. kala
.
nku, kalaku
v.t. ‘to agitate, perturb’, kala
.
du ‘to be shaken’, Tu. kala
.
nkuni v.i. ‘to be
turbid’, kala
.
nk¨ı ‘turbidness’; SD II: Te. kala
n
gu ‘to be stirred up’, kala
n
cu
4.6 Conc
lusion
173
v.t., kala
n
ka n.‘confusion, turbidness’, Kui glahp-(glaht-) ‘to confound’;
ND: Ku
.
r. xalx- ‘to disturb, make muddy’, Malt. qal
.
g- ‘to disturb, as water’
[1303].
Note that NP: NPP of the second syllable originally represented an intransitive–transitive
difference in South Dravidian I and II. It is preserved in Tamil, Malay¯a
.
lam and some
of the Nilgiri languages in South Dravidian I; in Kanna
.
da the transitive is a relic form
not matched by a corresponding intransitive; in Tu
.
lu the form with the suffix traceable
to NPP gives intransiti
ve meaning. In South Dra
vidian II Kui
–Kuvi–Pengo–Man
.
da
generally retain the NB–NP alternation to signal intransitive–transitive difference.
North Dravidian has also lost the original alternation but retains the formal difference.
4.6 Conclusion
Ihav
e classi
fied the
sound changes into shared innov
ations and typologically motiv
ated
changes, which behave differently in subgrouping languages. This is a proposal that I
am making for historical linguists to consider.
4.6.1 Shared innovations
4.6.1.1 South Dravidian
There are two clear phonological innovations shared by all members of the two subgroups
of South Dravidian, South
Dravidian I and South Dravidian
II : (i) The merger of Proto-
Dravidian high vowels
∗
i
∗
u with
∗
e
∗
o in Proto-South Dravidian in the environment [C-a
(see Rule 4a, c, section 4.4.2). The etymologies attesting the origin of long vowels ¯e and
¯o from PD
∗
i/
∗
eC-a,
∗
u/
∗
oC-a in the two subgroups show clearly that the developments
in Central Dravidian and North Dravidian are different (see Rule 2 in section 4.3.2.1
followed by ety. (7), (8), (10); Rule 4 in section 4.4.2.2 and Rule 6 in section 4.4.3
and ety. (31), (32), (34), (36); see also Krishnamurti 1980: 502–3). (ii) The loss of
∗
c
through two intermediate stages of s and h, initially and medially in South
Dravidian
I. This change is also shared by South-Central Dravidian (South Dravidian II) where
s- >h- is found in many of the languages dialectally. In Gondi h- >Ø- is an ongoing
change which is completed in the southern dialect of Koya. Kui–Kuvi–Pengo–Man
.
da
show the rule of s > h spreading dialectally. I have proposed that the South Dravidian
I change had also gone through similar intermediate stages although they were not
directly attested in any of the languages. However, there is ample indirect evidence of
Sanskrit and Prakrit borrowings from Dravidian with s and h attesting to these stages
(see Rule 13, section 4.5.1.3, fn. 5; see ety. (10), (24), (45), (46/104). The formative
suffix -¯a in Old Tamil is also shown to have developed from a prehistoric
∗
-aha <
∗
-asa
through contraction.
(iii) The apical displacement rules by which medial apicals came to
word-initial position characterized South Dravidian II as a shared innovation (Rule 20,
174 Phonolo
gy: historical
and comparative
section 4.5.7.3, ety. (90)–(94)). But we also notice a similar, if not an identical, rule
now spreading in South Dravidian I in a few lexical items by which the medial apicals
come to the initial position by loss of a word-initial short vowel (see section 4.4.4.6).
There is no evidence of this as a genetic phenomenon either in Central or in North
Dravidian. (iv) In both these subgroups PD
∗
t is represented by r in intervocalic position
(sections 4.5.5.3, 4.5.8.4.3). This distinguishes South Dravidian subgroups from Central
Dravidian, where it remains a stop [
∗
d]. Telugu goes with South Dravidian I in merging
r with rin all positions. This is a typologically motivated change in all the languages
with independently recorded histories and not a shared innovation.
South Dravidian II has two exclusive isoglosses: (i) The replacement of
∗
.
z by
.
r occurs
in all except Telugu where it is replaced by
.
d; (ii) the replacement of
∗
.
l
∗
.
n by lnregularly
in Telugu and by l/
.
r (dialectally) and
.
n/n, respectively, in the other members of South
Dravidian II. Kui–Kuvi–Pengo–Man
.
da replace PSD
∗
t [r],
∗
nt [nr],
∗
tt [R = voiceless
alveolar trill] by palatal
affricates
j, nj or s the corresponding voiceless
consonant
(instead of c) (see section 4.5.8.1.3).
Within South Dravidian I, there are isoglosses enclosing smaller subgroups. Tamil
and Malay¯a
.
lam have three shared innovations, (i) PSD
∗
e,
∗
o[C-a> i, u[C-a (Rule 4b,
c, section 4.4.2, ety.(22)–(26)), (ii) the palatalization rule (Rule 14b, section 4.4.1.4,
ety.(37), (38), (49)–(51)), and (iii) the uniform loss of N from PD
∗
NPP clusters (section
4.5.8.5., ety. (111)–(113)).
Centralized vowels occur in Toda, Ko
.
dagu and Ku
rumba,
but not in Kota. I have
proposed that the phonetic basis of this, namely retracted tongue position before retroflex
consonants, was an inherited feature from Pre-Tamil (section 4.4.4.2) and it is not an
areal feature of the Nilgiri languages, since Kota and Ba
.
daga do not have it. We must
still examine if there is strong morphological evidence for putting Toda and Kota as
branches from a single node.
One clear isogloss encloses Kanna
.
da, Ko
.
dagu, Ku
rumba and Tu
.
lu, i.e.
∗
w- > b- (Rule
18a, section 4.5.4.1, ety. (14), (71), (88), (101)). A rule rounding front vowels after labial
consonants is found in Tu
.
lu and Ko
.
dagu, see Rules 8b (section 4.4.4.3) and 9 (section
4.4.4.5). Kanna
.
da change p- >h- >Ø- (Rule 12) is found in Ba
.
daga only as a shared
innovation, according to some scholars (section 4.5.1.1, ety. (16), (17), (29), (31), (82)).
It is found in many of the neighbouring languages, Ko
.
dagu, Tu
.
lu, Ku
rumba, Kota etc.
through contact and borrowing.
4.6.1.2 Central Dravidian
It is important to know that the changes shared by South Dravidian I and South Dravidian
II are not shared by Central Dravidian. That is why South Dravidian II which was
originally grouped with Central Dravidian has been separated from it in my writings
from 1974 onwards. A shared innovation of Central Dravidian languages is the treatment
4.6 Conc
lusion
175
of
∗
t [d] intervocalically. It is represented by -d- in Kolami, Naiki and Parji and by -y-
(<-d-) in Ollari and Gadaba. The NE dialect of Parji shows it as a retroflex stop -
.
d-.
Kolami and Naiki also show -r- which look like borrowings from Telugu. The treatment
of a single
∗
t as [r] in the intervocalic position is a shared phonetic phenomenon of
Proto-South Dravidian (section 4.5.5.3). The Central Dravidian languages represent this
as a stop [d] in all positions.
(114) PD
∗
k¯et/
∗
ket-V- ‘to winnow’,
∗
k¯ett- n. ‘winnowing basket’. South Dravid-
ianI:Ta.c¯e
.
t
.
tai ‘winnowing basket’ (lw <Te.), Ma. c¯e
ruka v., Ko . k¯er-v.,
To. k
¯
¨
o
r- v., Ka. k¯eru v.; SD II: Te. cerugu (<
∗
cer-ugu) v.t., c¯e
.
ta ‘winnowing
basket’ (<
∗
c¯etta- <
∗
k¯ett-), ND: Ku
.
r. k
˜
¯es-, Malt. k¯es- : CD: Kol. Nk. k¯ed-
v., k¯et- n. ‘winnowing basket’, Pa. k¯ed-, k¯e
.
d-v.,k¯eti, k¯e
.
ti n., Oll. k¯ey v. ,
k¯eti, k¯e
.
tin n., Gad. k¯ey v. , k¯e
.
ten n. [2019].
The other South Dravidian II languages do not have the verb (Pengo has a doubtful
verb j¯ec- ‘to winnow’), but have the derived noun borrowed from Pre-Telugu after
palatalization in the form
∗
c¯ett-V: Go. s¯eti, h¯eti,Ko
.
n
.
da s¯eRi,Kuis¯esi, Kuvi Pe. Man
.
da
h¯eci ‘winnowing basket’. The absence of a verb corresponding to the derived noun and
the fact that ‘winnowing baskets’ made by Telugus are used in South Dravidian II tribal
areas supports early borrowing of the Pre-Telugu word into this subgroup. Also note
that Tamil has a noun borrowed from Telugu and no verb; the other South Dravidian I
languages also have no derived noun. Only the Central Dravidian languages have both
the verb and the noun with different phonology.
20
An isogloss which binds a subgroup of Central Dravidian is the loss of word-initial
n-; this is found in the whole subgroup in two items, e.g.
∗
¯ı n /
∗
in-‘you’(<
∗
n¯ı n /
∗
nin-),
but it is regular in the case of Kolami–Naiki (Suvarchala 1992: 20). We cannot attribute
this to this subgroup since n- loss is sporadically found in other languages also (see TVB:
§1.218–220, pp. 91–2).
4.6.1.3 North Dravidian
The shared innovations are
∗
k becoming x/q before all vowels except the high front
ones (Rule 15,
section 4.5.1.4) and the velarization of PD
∗
c before high vowels (Rule
13c, section 4.5.1.3, ety. (46) ff.). Another shared innovation is
∗
w- >b- perhaps under
the influence of the neighbouring Eastern Indo-Aryan languages (Rule 18b). This sound
change has implications for the original home of Brahui. Western Indo-Aryan (Kashmiri,
Sindhi, Lahanda, Panjabi, Gujarati, Marathi) did not share the East IA change of
∗
v- > b-.
20
Zvelebil has treated the forms with s/h in South-Central Dravidian as a case of irregular palatal-
ization (1970b: 117–18). ‘This replacement seems however irregular and dependent upon some
additional factors...’ (118).
176 Phonolo
gy: historical
and comparative
This makes it likely that the speakers of Brahui left east India where they were with
Ku
.
rux–Malto till after the operation of this sound change around the eighth century AD.
Many scholars think that Brahui is the first column of speakers left behind when the
Proto-Dravidian speakers entered India from the northwest. This is highly speculative,
since Brahui does not preserve any archaic features. The oldest Dravidian features in
phonology and grammar are preserved in the southern group of languages.
4.6.2 Changes with typological goals
A number of sound changes have occurred or are occurring in contiguous languages
at different times, producing a final result, which, if we looked back after many years,
would give the impression that they were shared innovations. These are different from the
sound changes discussed in section 4.6.1 in several respects: (1) they do not have a fixed,
definab
le time frame, except that the
y are all post-Proto-Dravidian;
(2) there is evidence
that they have been occurring in different languages at different times; some are on-going;
(3) they cut across the subgroups set up on the basis of shared innovations; (4) it seems
possible that their spread can be defined in terms of broad geographical regions. For these
reasons they are considered typologically motivated sound changes, since each of these
can be shown to have affected the phonological systems of the concerned languages in
creating greater internal symmetry and cohesion (see Krishnamurti 1998a: 75–8).
l. Proto-Dravidian root-final
∗
-ay, developed to -ey and then to
˘
¯e and
˘
¯ı in almost all
South-Central Dravidian, Central Dravidian and North Dravidian languages (section
4.4.7). This change (monophthongization) has eliminated diphthongs in most of the
languages.
2. PD
∗
y- is lost in all languages except Old Tamil; Modern Tamil and Malay¯a
.
lam also
do not have it. The vowel following
∗
y was
∗
˘
¯ representing neutralization of
˘
¯a and
˘
¯e. This
vowel was represented as e/¯a in South Dravidian I, as ¯e/¯a in South-Central Dravidian,
as a/¯a in Central Dravidian and as e/¯e in North Dravidian. Similarly PD
∗
˜n- merged
with n- in all languages except in Malay¯a
.
lam (with some relics in Ko
.
dagu and Tu
.
lu)
which have retained some items with
∗
˜n and not all. There are a few lexical items with
∗
˜n
followed by i and
˘
¯o, and not by the other vowels,¯ı and
˘
¯u (see Rule 17 and section 4.5.3.3).
The restricted distribution
of these phonemes, compared with the other consonants
that
occur initially in Proto-Dravidian, and their low functional load must have been the
typological factors leading to their eventual elimination through merger in most of the
languages.
3. PD
∗
.
z is the most characteristic of Proto-Dravidian phonemes which is not found
in many languages of the world. It is a retroflex frictionless continuant (approximant).
It is lost in all languages consequent on splits and mergers during the historic times,
except in some regional and social dialects of Tamil and Malay¯a
.
lam. In Parji of Central
Dravidian, it developed to
.
r distinctively. It merges in different languages with
.
d
.
rl
.
lc
4.6 Conc
lusion
177
ryˇs
.
sØ. No clear isoglosses seem possible except for a subgroup of South Dravidian
II, i.e. Ko
.
n
.
da–Gondi–Kui–Kuvi–Pengo–Man
.
da in which it developed to
.
r (see section
4.5.7.1.3).
4. PD
∗
.
n
∗
.
l (retroflex sonorants) became deretroflexed as dental/alveolar nlin South
Dravidian II, Central Dravidian and North Dravidian. These phonemes are, however,
internally reconstructable in certain languages of South-Central Dravidian and Central
Dravidian. All South Dravidian I languages preserve them. Retroflex
∗
.
l is allophonically
preserved in Naik
.
ri after non-front vowels; before front vowels
∗
l and
∗
.
l merged into /l/;
∗
.
n is preserved in Ko
.
n
.
da. The merger of the two retroflex sonorants with alveolars is
still
an ongoing change, and it is a sweeping one too. The languages in which the changes have
occurred during the recent past border on Indo-Aryan languages which have only l and n
(except Marathi which probably has retroflex
.
l and
.
n being a geographical neighbour of
Kanna
.
da which preserves the
contrast between
l:
.
l and n:
.
n) (see sections 4.5.7.1d, 4.5.6).
5. PD /
∗
t
∗
tt
∗
nt/ are preserved only in Toda, Iru
.
la and Kurumba, and to some extent as
/tt
/ and /nd/ in Malay¯a
.
lam in South Dravidian I; they are preserved as /rRnr/ in Ko
.
n
.
da
of South Dravidian II. PD
∗
t [r] merged with the flap /r/ in all the literary languages and
in others that inherited the [
r] allophone in South Dravidian I. In South Dravidian II in
the subgroup Kui–Kuvi–Pengo–Man
.
da, it merged with j; so also,
∗
-nd became -nj and
∗
tt [R] merged with c/s. In Central Dravidian
∗
t [d] merged with /d/. In the languages of
all subgroups (except some Nilgiri languages and the subgroup of South Dravidian II
mentioned above), the sequences /
ttnt/
have merged with dental /tt nd/
or retro
flex
/
.
t
.
t
.
n
.
d/
.
Some of these changes are datable within the literary languages. No common historical
stage can be postulated. These chains of sound changes have led to the six-point stop
system becoming five-point as it is in most of the languages of the Indian linguistic area
(see sections 4.5.5.3, 4.5.8.1.3, 4.5.8.4.3).
6. Two syllable types have become normalized in all languages except Tamil and
Malay¯a
.
lam, namely (C)VCCV or (C)
¯
VCV. A number of phonological changes have
occurred leading to this typological goal. Such a shift is also evidenced in Indo-Aryan
(Krishnamurti 1991a). PP >P, N P >B/#(C)
¯
V
−−−
is part of this strategy. Loss of a high
vowel /i u/ in the medial (unaccented) syllable has led to the creation of disyllabic forms
from underlying trisyllabic
ones, e.g.
∗
mar-u-ntu ‘medicine’: Te. mandu, Ka. maddu,
mardu,Pa.merd-(<
∗
mar-nt-).
7. Consequent on changes in canonical shapes, obstruent voicing became phonemic
in almost all the languages except Tamil, Malay¯a
.
lam and Toda, as follows:
PD
∗
P-
∗
-P-
∗
-PP-
∗
-NP-
∗
-NPP- >
post-PD ? -B- -P- -NB- -NP-
The initial position is filled by secondary voicing and through borrowings from Indo-
Aryan. The older single vs. double contrast became voiced vs. voiceless (see general
comments in sections 4.5.1, 4.5.2, 4.5.5, 4.5.8.1, 4.5.8.4, 4.5.8.5).
178 Phonolo
gy: historical
and comparative
4.6.2.1 Observations
(a) Most of the historically identifiable shared innovations have exceptions. The ty-
pologically motivated ones are extremely regular.
(b) It is likely that shared innovations generally spread through lexical diffusion.
Typologically triggered sound changes show the Neogrammarian regularity and they are
exceptionless. It is not clear if they imply a different mechanism.
(c) What are the consequences of recognizing two types of sound change for the
comparative method and reconstruction?
What is proposed here is that certain sound changes are motivated or caused by
system-internal pressures and such changes tend to be very regular compared to those
which are caused by sporadic shifts in the speech habits of speakers. For instance, in
Dravidian, the palatalization of a velar before palatal vowels is one such, although it
is phonetically conditioned and quite common in the languages of the world. There
is nothing system-internal to initiate this change, although it may have brought about
changes in the distribution of certain phonemes. In contrast to this, the replacement of
∗
-t-[-r-/-d-] by r/d/ j ,of
∗
tt by tt/t ∼
.
t
.
t/
.
t∼ c/s and of nt [nd/nr] by nd/
.
n
.
d/nj in different
languages is typologically motivated. Hence, its spread has been sweeping and there
are hardly any exceptions after its operation (for further discussion, see Krishnamurti
1998a).
This
proposal must be examined seriousl
y by historical linguists.
5
Word formation: roots, stems,
formatives, derivational suffixes
and nominal compounds
5.1 Structure of roots and formatives
It is beyond doubt that Proto-Dravidian roots (verbal and nominal) were all monosyllabic
with the canonical shape (C)
˘
¯
V(C) (see section 4.2). As Caldwell has already mentioned
(1956: 206ff.):
We find in these languages groups of related words, the first syllables
of which are nearly or wholly identical, whilst their second syllables are
different in each instance, and in consequence of this difference produce
the required degree of diversity in the signification of each member of the
group.
Regarding the elements that follow the root, Caldwell says:
The specialising particle, which was probably a separable suffix, formative,
or postposition at first, has become by degrees a component part of the
word; and this word, so compounded, constitutes the base to which all
formatives, properly so called, and all inflexional particles are appended.
(1956: 206)
Caldwell recognized the underlying root of a family of words as the invariable element;
he calls the variable elements, mainly of the -L/-VL type (see section 4.2), ‘particles of
specialization’. The stems formed by the addition of these enter into inflection as verbs
or nouns, etc. The so-called ‘specialising particles’ diversify the meaning of the root
but no clear meaning can be assigned to each of them. He also calls them ‘formative
additions’ (1956: 203).
Caldwell divides roots into those which occur either as verbs or as nouns, and nouns,
which cannot be derived from any underlying verbs, e.g.
∗
k¯al ‘leg’,
∗
kal ‘stone’ (1956:
196). Suffixes of NP: NPP type (see section 4.2), which signal intransitive–transitive
meaning, were ‘signs of verbal nouns’, according to Caldwell (1956: 196). The vowels,
which occurred between the roots and these ‘clashing consonants’, were inserted
‘euphonically’ (1956: 196). In our terminology, he considers V
2
as an epenthetic vowel.
180 Wo r
d formation
He gives examples both from verb stems and noun stems that have what we now con-
sider
∗
-V-L,
∗
-V-NP, and
∗
-V-NPP. Caldwell also calls attention to a parallel phonological
process in Tamil, i.e. gemination of the final obstruent in forming transitive verbs from
intransitive, and in forming adjectives out of nouns, e.g. Ta. tir-u-ntu v.i. ‘to change’,
tir-u-ttu v.t. ‘to correct, rectify’ (<
∗
tir-u-ntt-); mar-u-ntu ‘medicine’, adj. mar-u-ttu
(<
∗
mar-u-ntt-). Caldwell has not given any supporting arguments for considering the
formatives, i.e. L, P, NP, PP and NPP, as ‘signs of verbal nouns which secondarily became
verbs’ (198). Actually it appears that the opposite is the fact. There is more justification
to treat the formatives as original tense and voice suffixes, which later lost the tense
meaning first, but retained transitivity. In some languages even the latter meaning is lost
and they have become simple formatives without any grammatical and semantic content.
Caldwell, following Gundert, further tried to derive some of the -VL suffixes from in-
dependent roots like
∗
il ‘house’,
∗
u
.
l ‘to be’, but this was a mere guess and not supported
by comparative evidence
(see Krishnamurti 1961:
§2.32, p.145). He also thought that it
was ‘euphony only that determined which of the consonants g, ´s,
.
d, d,orb should be
affixed as a formative to any particular verb or noun’. Even this is speculation as can be
seen from below.
Table 5.1. Canonical shapes and number of root morphs in
Proto-Dravidian
Canonical form of the root Number of root morphs
1.
˘
V- (5 short vowels) = 5
2.
˘
VC (5 short vowels × 16 consonants
1
) = 80
3. C
˘
V - (6 consonants
2
× 5 short vowels) 30
(3 consonants
3
× 3 short vowels) 9 = 39
4. C
˘
VC (39 × 16 consonants) = 624
Total = 748
5. 1–4 with long vowels = 748
Grand total = 1,496
The total number of possible (not necessarily actual) and reconstructible roots for
Proto-Dravidian would be 1,496 on the basis of morpheme structure conditions, dictated
by the phonotactics described in section 4.2 as per the details given in table 5.1.
Any chosen Proto-Dravidian root, therefore, has to be one or the other of this total
number.
1
All consonants except ˜n.
2
p, t, c, k, m, n.
3
y,˜n, w occur only before unrounded vowels /i e a/. We are still not certain about the total
distribution of the laryngeal
∗
H in Proto-Dravidian.