You are on page 1of 37

PROTO-DRAVIDO-URALIAN

Stephen A. Tyler
Dept. of Anthropology and Linguistics
Rice University
September 1986
..
.
Abbreviations:
Abbreviations used in this paper are as follows: Uralian
Fi. Finnish, Vt. vote, Est. Estonian, Lp. Lappish, Md. Mordvin, Vty. Votyak,
Zr. Ziryene, Vg. Voguil, Os. Ostyak, Hu. Hungarian, Yr Yurak, Sk. Selkup,
ot
Km. Kamassian, Yk. Yukagir; Dravidian Ta. Tamil, Ka. Kannadf, Ko. Kota,
'('
To. Toda, Tu. Tulu, Te. Telugu., Kol. Kolami, Pa. Go. Gondi, Kur. Kurukh,
vt
and Br. Other abbreviations include DED (Dravidian Etymological
Dictionary); DEDS (Dravidian Etymological Dictionary: Supplement); DEN
(Dravidian Etymological Notes); FUV (Finno-Ugric Vocabulary); IE (Indo-
European); sg. (singular); pl. (plural)
..
.
-------------------
TABLE 1. Proto-Dravidian Phonology
Consonants
;Bilabial Dental Alveolar Retroflex Palatal Velar
p t
!;1
t c k

N
m n n ? n n ?
1 1
.
3 2
r r
...
v y
Vowels
Front Back
i i u u
e 0 0
a a
..
.
1. sometimes written
2. = ! ! in various reconstructions
3. Dento-alveolar flap
TABLE 2: Proto-Uralian Phonology
Consonants
Bilabial Dental Alveolar Cacuminal
1.
2.
p t

s s

c
m n n ?
.
1
r
v
Vowels
Front Back
il
u
..
.
e 0
a a
These may be reconstructed with contrasting length.
0
Palat-velar fricative, Collinder.
Palatal Velar
k
,.
s
""
c
I
? n
l}
I
1 ?
y
TABLE 3: Uralian-Dravidian Phonological Correspondences
Ur Dr Ur Dr
Ur Dr
a a
k k r 1
a a
" c c
w v
a a ~ t
y y


"
a a c t
..
a e s c
..
a e s t
..
"
a ay s t
..
i a
s c
e e
~ !.
e e y'tj
r
..
e a
7Y
r
i i t t
";'"
i Jj1!f (
t t
"'
i ay
t t
0 0
n n
-
/
0 0
n n
/
N
0 u n n
v
u
I)
n
.
- u. u p p
u 0 m m
u iy m v
u 0
1 1
1 1

1 r
,,
r r
r r
,.
r t

TABLE 4: Proto-Dravido-Uralian Phonology
Consonants
Bilabial Dental Alveolar Retroflex Palatal
1.
2.
3.
4.
5.
p t
m n
1
r
v
..

Front
i
e
i
....
e
Vowels
-
a a
t c
.
,...
n n
1
r
..
Back
-
u u
0 0
All Dravidian alveolars are split from dento-alveolars.
All Ural ian sibilants are variants of c.
All Ural ian palatals are palatalized retroflexes.
All Ural ian cacuminals are palatalized retroflexes.
Ural ian
..
a is a fronted (palatalized) a.
?
Velar
k
y
TABLE 5: Summary of Noun and Verb Inflectional Correspondences and Proto-forms.
Proto-Dravido-Uralian
nominative
genitive/oblique *-nV/*-tV
accusative *-m
locative *-ne/*-te
dative/lative *-kV
present/hortative *-k
past *-i, *-t, *-c
Prato-Dravidian
*-nV/*-tV
*-n/*-m
*-ne/*-te
*-kku/*-ku
*-k
*-i, *-t, *-c
Proto-Uralian
*-n, *-ta/*-di
*-m
*-na/*-na
*-tta/*-tta
*-kV
*-k
/
*-i, *-s, *-c
I.
1
Introduction
Since the publication of "Dravidian and Uralian: the lexical evidence"
(Tyler 1968), which summarized research on extra-Indian Dravidian connections
prior to 1968, a n u m ~ r of important publications have both widened and
deepened our understanding of the relationship between Dravidian and Uralian.
v "
The most important contributions have come from the work of Illic-Svityc
(1971-84) and Marlow (1974). Unfortunately neither of these works has been
"
widely available. Illil-Svityc's work is in Russian and Marlow's is in an
unpublished Ph.D. dissertation. I have not been able systematically to
., v
incorporate Illic-Svityc's work in this paper, and can only indicate here
the general correspondence between his comparisons, reconstructions, and
conclusions and those given by me in 1968 and by Marlow in 1974. For an
account of the Russian work, readers should consult the paper by V.
Shevoroshkin on Nostratic in this volume (pp. ). In general though
.. .,
Illic-Svityc includes Dravidian and Uralian, along with Altaic and Elamite
in h}.s reconstructed Eastern Nostratic "family of languages. His conclu-
sions,are supported by etymological comparisons, phonological reconstruc-
tions, and some morphological observations.
Marlow's work contains the most extensive vocabulary comparison yet
achieved. She lists more than 700 reconstructed Uralian vocabulary items
representing a "core" Uralian vocabulary and suggests Dravidian,cognates for
more than 600 of the Uralian words. She also reconstructs a tentative Proto-
Dravido-Uralian phonology based on systematic sound relationships between
2
reconstructed etyma from both language families. systematic
knowledge of both Dravidian and Uralian languages, utilize
sources not normally available to those whose specialist k is limited
to only one 'of the two language families. As this valuable has not
~ been published and remains largely unknown to
appendix to this paper some representative etymologies from it.
Other work on Dravidian, not directly incorporated here, but forming
part of the growing discourse on wider dravidian connections, includes the
studies of several Japanese scholars on Japanese and Dravidian (see, e.g.
Fujiwara 1981 for details and references). The Japanese-Dravidian etyma
suggested by Fujiwara look convincing, but I cannot really judge the
Japanese evidence. Moreover, most of the Japanese scholars seek to show a
connection directly between Japanese and Dravidian, where, following Menges
(1964, 1969), I would connect Prato-Dravidian and Proto-Altaic, the relation-
ship between Japanese and Dravidian then being mediated through Altaic.
Closer to the historic and contemporary Dravidian locale is Elamite,
and here McAlpin (1973) has provided a fairly convincing case for a genetic
connection between Dravidian and Elamite which will become even more evident

with the publication of his further work, and with the progress of research
on the relationship between early non-cuneiform Elamite script and the Indus
Valley script. In this context, the recent attempts to translate the Indus
script ar.e relevant since both the Finnish and Russian teams have concluded
that the language of the Indus script is Dravidian. Even though there
is little proof in the way of accepted translation, the Dravidian attribu-
tion is still the most likely one. This might seem to be contradicted by
3
ltfcheuer's (1978) recent attempt to link the Indus script with Sumerian,
but in view of Boisson's (1985) attempt to link Sumerian and Dravidian,
the contradiction dwindles.
Much of this work on the relation of Dravidian to archaic Middle
Eastern languages is still in early stages and is consequently difficult
to evaluate, but in connecti'pn with both archeological and ethnological
evidence, which has for some time implicated these relationships, it is
likely that further research will confirm them. One effect of the Middle
Eastern connection is, of course, to reduce the importance of geographic
1)
distance between the eravidian and Uralian languages as a stumbling block to
"
acceptance of their genetic relationship. The Middle Eastern research, along
with archeological evidence, suggests an unbroken chain of interconnected
cultures stretching from Central Russia to Central India, and we already know
from the case of Indo-European that geography is neither a barrier to, nor a
predictor of genetic linguistic relationships. Whatever the ultimate
significance of this related research, we now have in hand an accumulation
of evidence from independent sources confirming the genetic connection be-
t w e ~ Dravidian and Uralian.
Apart from etymological correspondences and morphological recon-
structions, the following list provides a general background of structural
similarities between the two language families.
(1) Word order is generally S 0 v.
(2) Qualifers precede the qualified, thus adj + N, a d ~ V.
(3) Order of 'Comparison is "standard+ comparator+ adj."
(4) A separate negative conjugation, and a negative auxilliary
4
inflected for person, number, and tense.
(5) Common verbal structure consisting of base + tense/mode +
personal ending.
(6) Separate verb conjugations for "objective" and "subjective"
(7) Personal endings of verbs are derived from personal pronouns.
(8) Common derivational order: root + derivative suffix +
inflection.
(9) Common pattern of affixation; elements joined by suffixing
rather than by prefixes or infixes.
(10) Two correlative noun categories: Dravidian rational ("high
class")/nonrational ("low class"). is equivalent to Uralian animate/
inanimate.
(11) Common nominal structure consisting of base + case suffix or
base + plural + case suffix or base + oblique + post-position.
(12) Cases:
s
both have case system,.,consisting of a set of "core"
cases (nominative, genitive, accusative, dative, and locative) and other
der:f.."Y'ed cases.
(13) Post-positions: other "case" relations are expressed by
post-positions derived from nouns or verbs.
(14) Singular-and plural have the same case forms.
(15) Adjectives and adverbs are scarce and usually derived from
nominal or verbal phrases.
(16) There are no definite and indefinite articles.
(17) Gender systems are either absent or underdeveloped.
5
(18) Echo words are frequent.
(19) A counnon system of enumeration (see below, pp. ... JZ..).
Given these impressive parallels of structure, the large common
vocabulary, and systematic sound correspondences already provided no one
can any longer reasonably conclude that the genetic relationship between
Dravidian and Uralian is only a possibility. I take it as sufficiently
demonstrated and consequently turn to the next order of business, which is
the reconstruction of Dravido-Uralian phonology and morphology.
II. Proto-Dravido-Uralian Phonology
Based on the proto phonology of Dravidian underlying the Dravidian
Etymological Dictionary (Burrow & Emeneau 1961) and on Zvelebil's Compara-
tive Dravidian Phonology (1970) we have the following table of Proto-
Dravidian phonology.
..

Table 1 about here
Based largely on the work of Collinder (1960), Proto-Uralian phonology
is reconstructed as indicated in Table 2.
Table 2 about here
Using the list of cognates published in Tyler (1968), and in Marlow
(1974), we derive the phonological correspondences between
- -- ---- -- --------
-----------
6
Uralian indicated in Table 3.
Table 3 about here
From the correspondences in Table 3, we derive Proto-Dravido-
Uralian phonology as given in Table
Table 4 about here
While much continues to be problematic, particularly in the
Uralian reconstructions, and many details remain to be worked out, we can
still point to some general processes of phonological change. Among them
are: (1) palatization; (2) vowel fronting - sometimes in association with
("
palatalization; (3) tetroflexion.
Palatalization is a process internal to both language families
.,;'
involving k > c, k > kh, k > s before front in Dravidian and probably
n > before front vowels or a labial or velar palatal in the next syllable.
,
c. , ,
In a palatal series ft, s, n, 1 is reconstructed even though
/ , ,
the evidence for s, n, 1 is very contradictory and points to their derivative
states, possibly as the result of contact with Slavic. In Uralian the
.,;' ,; /
(.. C C C. f .t / I "'
probable development is: k > c; > , > ,; c > s; s > s; 1 > 1 ; t > c.
Uralian affricates would all be allophones or derivatives of Proto-Dravido-
Uralian *t, *c, and *k. Where Proto-Uralian does not have the,expected
palatalization the conditioning following syllable is either reduced or
contracted through the loss of an intervening syllable. This would be
---
7
particularly the case with syllables containing a sonorant r or 1. The
.,
pre-palatal s is a positional variant of Proto-Dravido-Uralian *r where it

is not derived from palatalization of an underlying k.
Uralian has many vowels with fronted/raised allophones. In Pro,to-
Uralian the main reflex of this process. In other cases, vowel-
initial words will have an onset glide v/y which has the effect of fronting
the vowel. These glides sometimes become permanently affixed and the vowel
becomes fronted. An identical process occurs in Dravidian i/e and u/o
initials. Palatalization of an initial consonant before a front vowel in
Uralian may result in subsequent variants in which the vowel of the first
syllable is raised. This also occurs in Dravidian, but less frequently.
In general, the cacuminal series in Proto-Uralian corresponds to
the retroflex series in Dravidian, with some reduction in the total number
of phonemes in the series in Proto-Dravido-Uralian. To this group we should
add the velar affricate "f . Proto-Uralian cacuminals are all "affricatized"
retroflexes. This process also occurs within Dravidian where the reflexes
of *r are often affricates. Since the Proto-Dravidian retroflexes do not
..
initially, this might indicate a different reconstruction, e.g. from
affricate to stop or sonorant, but since affricatization occurs in Dravidian,
affricatization rather than "de-affricatization" seems more likely. In
addition, Uralian has internal evidence for the existence of retroflexion.
The of retroflexion also involves a problem about consonant
clusters. Both Uralian and Dravidian have frequent consonant clusters or
gemination and there is some evidence to suggest that Dravidian retroflexes
correspond to Uralian consonant clusters, but this is difficult to determine
8
just from the evidence of reconstructed roots.
The major source of problems in the reconstruction, however, apart
from the Proto-Uralian vowels, are the liquids, which generally correspond
as a whole but not in particulars. That is, most r's and l's line up, but
not in a predictable pattern.
III. Morphological Reconstruction
2
This section is intended only as a beginning or introduction to the
reconstruction of Proto-Dravido-Uralian morphology. It presents a recon-
struction of some numerals, the system of enumeration, the pronouns, the
plurals of nouns, the core case suffixes, and some of the tense/mode suf-
fixes of the verb.
NUMERALS:
Dravidian-Uralian numeral correspondences include the numerals
''one" "four" and "ten" and a conunon system of enumeration based on "eight."
o g ~ t e s for the numerals one and four are as follows:
Dravidian Ural ian
"one" DED834 Ta. onru FUV126 Fi. yksi
Ko. od Lp. ok'ta
Te. oka Vty. odik, ok
Pa. &;. Zr. ~ t
DED3024 Ta. nal
..
Fi. nelja "four" . FUV102
Ko.
.-
nang Zr. nol
Kur. nakh Hu.
,
negy
- --------
Not that these forms also agree in consonantal variation. For "one",
IV
both have variants with a nasal followed by *-t- (Cf. Ta. Zr. ot)
and a reduced form of this variant without the nasal (Cf. Ko. od, Vty,.
9
odik). Both also have variants with -k- rather than -nt-/-t-. Similarly,
both have variants of "four" with a liquid (-1-) and with a -k- or -g-.
This may mean that the -k- forms are reduced variants of *on-k- and
*nal-k- where the -t- and -k- have some function as suffixes (see
PP
ICf
The system of enumeration is transparent in the following example:
Te. Fi
one oka yksi
two rendu kaksi
..
three mudu kolme
.
four nalugu
,,
nelja
five ayqu viisi
six aru kuusi
seven edu seitse

eight enntmidi kahdeksan
..
nine tommidi yhdeksan
.
ten padi kymmenen
In both cases it is obvious that the numerals for 8 and 9 are compound
forms. The Uralian system shows two variants. In Sk., for example, 8 is
"
) .
Ca9 (;itty "" 2 J ciif"" "leSS J II kbt 10) J
11
tWO from ten iS eight 0 II
.. '::) ..
and 9 is ukkyr caf kot (ukkyr 1), "one from ten is nine." Yk., on the
3Q.
other hand, has for 8 sidnteet, from 2, and
four is eight." Km. also has the "twice four" form
teet 4; thus, "twice
"''
(sentee'df, from
10
"' 4
sen= 2 and tee'df 4). Yk. 9, however is
1
as expected,in the "minus"
form (haasajUu, = 10). Fi.,

Lp. and Zr. have a variant form of
2 before a 10 suffix; kah, gav- and kyk-
s
repectively 2 and -deksan, -deksa,
1\.
-ce, -mys, -aamys, jaamys respectively = 10. They are thus in the "two
..
from ten" form. In these same languages 9 is formed with a 1 (yh-, u-, ov-,
ok- respectively) before the ten forms above, and are thus '.'one from ten =
nine".
In Dravidian 9 is generally in the "from" or "minus" form. For
example, Ta. = 1, patu = 10), Go. unmak (un = 1, mak = 10)
(lf
"one from ten" (DED 862). The num}{al 8 though, is less clear. In South
Dravidian, with the exception of Tu., it appears to be a simplex form
en-, but Tu. and Central Dravidian languages generally have a
compound form of en- or en- plus -rna, -midi, -madi, -mater, -mur-. These
-m- suffixes appear to be "tens" even though none occurs as a free form
with that meaning.
-
elJllla, e.g. would seem to be "two from ten". These
Central Dravidian versions of 8 would thus be comparable to those in
Uralian using 10 as a subtractive base, but Dravidian is probably
.......... - .
DENS 13 Ta. "other", ...... not 4erived from any numeral for 2 (but see
.
"the rest"; Br. elo "the other one", "the second"), though it may be
cognate with *in- (DED 387), inai "pair", "double", "couple", which
would then yield *en -tu where -tu might
.
be cognate with Uralian 4, as in
Yr. teet, Km tee'da, Sk teety, and thus
ceo.
yield a "twiJ( four form in
Dravidian, but this seems far-fetched without further evidence on the
Dravidian side. It is more likely that Dravidian *en- is cognate with
.
*ell- "count", "calculate", "number" (DED 678). It is thus, "the number that
counts". It may be cognate with Hu. nyole 8, but without further evidence
from Uralian, this is only speculative
.---------------------------
11
There is also a connection here with Indo-European. Some reflexes
of IE 8 appear to have a dual suffix, indicating something like "twice four".
These dual forms may point to the existence of an archaic IE method of
counting by "fours" that is probably related to the similar Dravido-
Uralian system of counting by "doubling". Whatever the IE connection, it
is clear that both Dravidian and Uralian have a conunon system in which eight
and its factors and multiplesare the basis. It should be noted that systems
of weights and measures based on eight or sixteen have been current in India
since the Indus still in second
(16 annas to a rupee, "an 8 anna crop", etc.) up to about 1950, when India
changed to a decimal and metric system.
The "ten" suffixes in Zr. -mys, -(aa)mys, (jaa)mys (Cf. kykjamys =
-t-lu. /(.-vt. - rvtiilrfYI C\);
10- 2 = 8, okmys = 10- 1 = Fi. and Est. suffixes -menen, men
(kymmenen = 9, kymm = 9) form a cognate set to which should be added the
Fi.-mant, as in kolmant "third", and the following "tens": Vty. jwamyn (20),
komyn (30),)Kvatymyn (40), vetymyn (50);
, , It
Zr. nela-myn (40), Vg. nal-men (40),
Hu. negy-ven (40), hat-ven (60). These yield an Uralian "ten" form in *min-t.
('I
Coll\:nder reconstructs mono (FUV 133) and adds to the set Fi. moni "many",
Lp. moad'de "more than one", Vty. mynda "as much as", Zr. symyn "so much,
0
so many" and kymyn "how much?" He compares these to IE *mf(n- (Got. manags
"many", "several"; Ir. meneic "frequent", etc.). As a numeral, however,
there is a better correspondence with the Dravidian tens suffix *mant-,
which is probably the source of the Dravidian ten forms noted in connection
t'
with the discussion of eight and nine. These reflexes of (ma, mak,
madi, midi) are usually thought to be homorganic variants of *pan/pak- "ten"
(DED 3236), but this requires a very complex allophony for the initial
*p- of *pan-/pak, to say nothing of the stem allomorphy when *pan-/pak
functions as a tens base. The evidence is more consistent with two
separate tens words/suffixes, one \in *pan-/pak- or more likely, *pan-k
and another in *ma(n)-/man-t-, the two latter being cognate with *man-t-
(DED 3847), which means "group",' "herd", "flock", but is also used as a
numeral classifier suffixed to numerals when counting humans, as in Te.

nalugu mandi manusulu ("four human men") as opposed to nal,gu
("four oxen"). Note here, too, the form-meaning correspondence with IE
0
12
which thus points to an archaic form indicating "collection" in all
three language families.
As for the reconstructed form *pan-k- above, it seems clearly con-
nected to IE *pen-kw "five". The interpretation is that the *-k suffix of
Dravidian *pan-k and the *-kw suffix of IE can both be traced to an archaic
dual or plural (on the dual and plural in Proto-Dravido-Uralian (see below,
pp .20 -I ) . * On this reading, *pan- means "five", and pan-,k is "ten" or
''two fives". It is probably cognate in its archaic reading with words for
"full", "replete" (Cf. DED 3174).
To sum up, Dravidian and Uralian have clear cognates for the
numerals one and four, and probably also for ten (*man-t-). More importantly
..5
they have the same structure of enumeration, both forms for the
first seven numerals and compound, derivative forms for eight and nine.
Enumeration is based on eight, which is either a doubling of four or sub-
tration of two from ten. Both systems also implicate a more remote
connection with IE of the sort postulated in Nostratic.
13
PRONOUNS:
The correspondence between Dravidian and Uralian pronouns is com-
plex, but still apparent. There is, first of all, a general structural
equivalence in the distinctions of person and number, but not in gender.
Uralian does not distinguish between masculine, feminine, and neuter in
the third person, whereas Dravidian distinguishes between masculine on the
one hand and feminine/neuter on the other in the singular third person.
Compare the following paradigms.
Finnish Telugu
Singular Plural Singular Plural
. '
mina me - nenu memu 1st pers.
2nd pers.
..
sina te nuvvu miru
h;; he va:du m/f
-
m. varu
,
3rd pers.
f/n. adu
V't/
f/n aJ1u
Dravidianists do not agree on the reconstruction of gender in Proto-
Dravtdian, but something on the Telugu pattern above has been thought likely,
even though it is doubtful that there was a distinction between masculine
and feminine/neuter in the singular. Note too, that the plural distinction
is not specifically an indication of number as such. It is as much, if not
more, an indication of respect and honor, indicating the difference between
"high class" and "low class" people and things. All the early native
grammars attest to this distinction and note that low class nouns need not
be pluralized, though they can be. It is thus doubtful that these "plurals"
14
are plurals or that they encode a "gender" distinction since both males and
females, as well as some "neuters" can be either high class or low class.
In effect, so called gender wouldbe marked only in the "plural" since that
is where the honorific suffix occurs. would mean that the masculine
singular is a late development inspired .by the imbalance created by
honorifics in the plural.
Dravidian has a distinction between inclusive and exclusive in the
second person plural that seemingly has no direct counterpart in Uralian.
It also has a set of reflexive pronouns (*tan- sg.,/*tam- pl.) which occur
in first and second person and are used as third person pronouns.
Phonological correspondences in pronominal forms are similarly
complex. Collinder reconstructs for first person sg. in Uralian
where Dravidian first person sg. is *yan/*nan. Similarly, Uralian has
*mi/*my in first person pl. while Dravidian has *yam/*nam. Where Uralian has
*m-, Dravidian seemingly has *y-/*n-, but both have an n/m opposition in
singular and plural. That is, Uralian has (FUV 34, 35, 145) where

the sign of the singular is -na. The argument advanced here is that mi is
"
the plural marker *-m and is not the pronominal base to which -na is suffixed.
This is confirmed by the distribution of personal endings for verbs in
Uralian, which are derived from the pronouns. Finish, for example, has -n,
first person sg. and -m first person pl. personal endings. Though some-
times reversed, -n and -m are the usual markers of the first person in verbs.
Dravidian, too, uses derived forms of the personal pronouns as personal
endings and has the same distribution: -n first person sg., -m first person
pl.
-1f(n) -en 1
. .
15
There is no direct correspondence between initial consonants in
the pronouns because each family has used different forms of the pronominal
variants that must have been present in their common ancestral source.
Evidence for this variation is common in the daughter languages of both
()..
Dravidian and Urali,n. Note, for example, Dravidian *yan/*nan, *yan/*nam
(DED 4231, 4234), but also Te. emu, nemu, memu; Go. ammat, mammat (first
person pl.); Te. enu, nenu, Go. ana, nanna (first person sg.). In Uralian,
compare Fi. mina, Hu. en, Vg. am (first person sg.). Thus, Proto-Dravido-
Uralian had first person forms in *CVn/*Vn (sg.), and *CVm/*Vm (pl.). Proto-
Dravidian *y- in pronouns was originally probably an onset glide. The
phonological realizations of the initial consonants was either *m- as a
common first person pronominal base or *n-/*m- in which the initial consonant
reduplicated the signs of the singular and plural respectively. The
tentative Proto-Dravido-Uralian reconstruction for first person pronouns is:
(n)-"'
1*'/eY-/*(m)"en- sg. *(m)em pl.
The second person pronouns are problematic in a different way.
Uralian has *tirJ./*tyna sg. and ti/ty pl. (FUV 57, 62) for second person
pronpuns, while Dravidian generally has *(n)in-/*in- sg., pl.
on the pattern of the first person. The second person also has a "high
class" pl. in *nfr-. As in the first person, these pronouns in both lan-
guage families are the source of the second person endings of the verbs,
Uralian generally having forms in *-t- (Collinder 1960:308-10), Dravidian
in *-i(m)/*-t sg. and -ir pl A solution along the lines of the first
person is tempting, but not likely since it would leave -i-/-f: as the
only mark of the second person. A better solution is to relate the Uralian
16
forms in *-t- to the Dravidian reflexive pronouns *tan- sg., tam pl.,
especially since they are also used as pronouns of the second person anyway
(Cf. DED 2582, 2612) as in Ta. tim ''you" (honorific pl.), tinkal id

They can also be used in the third person 'for "he", "she", and "it".
This interpretation fits the fact that ITralian, too, had two words for
"thou", one in *t-, and another in *n- (eollinder 1965:134). Since only
Uralian shows the t/n variation in the non-reflexive pronouns, it seems
likely that Dravidian has specialized its reflexive pronouns from the
*t- pronouns in Proto-Dravido-Uralian. This is corroborated by the occur-
renee of second person sg. personal endings in -t and -t in Central
..
I.C.
Dravidian, and perhaps by endings in -s (<*-t?) in Since the personal
endings of verbs are derived from pronouns these -t suffixes implicate the
existence of a second person pronoun in t- in Dravidian. As indicated above,
it is related to the reflexive pronouns in t-. then,
had second person pronouns in *tin-/*nfn- sg. and *tfm-/*nlm- pl.
The vocalic difference *-e- and *-i- between first and second person
pronouns is probably related to a proximate/remote distinction that also
oper,tes in the third person as well, and in the third personit is doubtful
that the pronouns in either language family can be reconstructed indepen-
dently of the demonstrative bases built on this distinction. In Uralian
it is generally the case that the demonstrative pronouns and the third
person pronouns are either derived from or are directly interchangeable
,
with the demonstratives. Collinder reconstruets *ce/*ci for one set of
,.
demonstratives and ta for another (FUV 56, 62), and speculates that the
latter may have had two forms, one with a back vowel and one with a front
17
vowel marking remote and near respectively. In this group we should also
consider the Uralian reflexive pronoun reconstructed as *ica (FUV 16),
which Collinder notes is probably a compound of *i + ce, the reconstructed
demonstrative. Although nothing is straightforward here, it is clear that
the demonstratives and the third person pronouns are derived from the same
underlying base, something like *ti. The evidence from Hu. az "he",
a..
"she", "that", and ez "he", "she", "this", and the evidence of *icr <
*i- *ce above led to the following reconstruction: back vowel + *-ti =

third person sg. (remote) and remote demonstrative. Front =
third person sg. (proximate) and proximate demonstrative. If this recon-
struction is correct, then the third person sg. pronouns of Uralian corre-
spond completely with the third person sg. pronouns of Dravidian.
Dravidian third person pronouns are also derived from demonstrative
bases (DED 351, 475), thus, for example 1ra. avan is a- (remote demon. base) +

,...._
(v).._.,an "he", and means "that he" (remote), and contrasts with iva_n "this--he"-
(proximate) from i- (proximate domon. base) + So too, atu, from a-
(remote demon. base) + -tu "it", "she", "this it/she", contrasts with itu,
from . i- (proximate demon. base) + tu "it, she", "this it/she". Here the
..
,,
form indicating "it, she" corresponds with Uralian *ta/*ti (demon.
base). Dravidian also forms the third person pl. pronoun on the demonstra-
tive (e.g. Te. aval, avar, from a + v
N' IV
glide + L arkal).
Since Uralian does not differentiate gender, Dravidian has created a
masculine gender in the sg. third person pronoun or Uralian has lost it.
The former is more likely since the third person verbal endings having
gender distinctions are not reconstructable in all formsof the Dravidian
18
verb paradigm. Only the epicene pl. can be reconstructed (see
Subrahmanyam 1971:414-22). The development of a separate masculine
verbal suffix may have occurred late in Proto-Dravidian, but even this is
doubtful, and it is more likely that the third person verbal endings were:
Sing. Pl.
3rd pers. m *-ri High class
3rd pers. f *-tu
3rd pers. n

Low class
? N ---.1<
As indicated previously, the plural was not just human+ gender + number, but
was "high class" vs. "low class", in which the latter needed no plural, and
in which the high class category might be used for humans, non-humans, males,
females, and neuters, singular or plural. At the Pre-Dravidian stage, the
whole paradigm could probably be collapsed into a single opposition between
high class and low class without any indication of number. The *-k above
would originally have been a dual marker (see discussion below,

and of the *-! plurals in Dravidian, which anyway often used in
association with *-k, as in Ta. and interpreted as a "double plural,"
would have been later innovations connected with the loss of duality and the
inception of a truly plural semantic distinction.
Uralian is even more complicated. It is not possible to reconstruct
a single plural morpheme for the third person, though the incidence of forms
in -k suggests something similar to the Dravidian situation outlined above,
so that these -k plurals are in some way related to the dual *-k. This
would indicate a possible Proto-Dravido-Uralian third person paradigm as
follows:
Sg. (?)
Third Person
*-tu
("he", "she", "it")
Dual
*-k
Pl.(?)
\1
*-ri
19
The pl. would have been restricted to high class "things". The parenthesized
(?) indicates doubt about the semantic interpretation of the words "singular"
and "plural". Note too, that each of those proto forms in pronominal usage
would have been suffixed to appropriate proximate/remote demonstrative bases.
The reconstructed Proto-Dravido-Uralian pronominal paradigm for the
nominative would be (dual omitted):
Sg. Pl.
First person *(n)en/*(m)en *(m)em
Second person *tin/nin *tim/*nim
Third person *-tu *-ri/*-0
..
.
As in the numerals, there are also connections between the Dravido-
Uralian pronouns and IE pronouns. Collinder and others have noted the
similarity between Uralian and IE pronouns, but the evidence presented here
points to a conunon source not just and Uralian, but for Proto-
Dravido-Uralian and Proto-IE, as proposed by the Russian linguists in their
combined Nostratic groups.
The major difficulty in this reconstruction involves not the
20
relationship with IE, but the developments of the dual and the plural.
Uralian languages generally have singular, dual, and plural. Dravidian
languages have no dual. The dual is often lost in Uralian or only
sporadically represented in some daughter languages, and the plural is
highly variable, there being no plural common to all of the case suffixes.
Thus, it is not possible to reconstruct one morpheme as the Proto-Uralian
plural. The most frequent markers of the plural are -t, -i, -k, and -n. The
- I I
dual is reconstructed by Collinder as *-ka/ka + n/n)and *-n. Since those
dual morphemes overlap with the plural markers -k and -n, it is likely that
some of the plurals in-k and-n are derived from the dual, especially
where the dual has disappeared or atrophied. This would leave -t and -c as
candidates plural.
The most common "low-class" Prato-Dravidian plural morphemes are
*-ku, with -n as morphophoneme, and *-lu. Less common are derivatives of

*-t + *-ku "' *-lu, as in Kui/Go. -sk(t+ku), 011. -til, Pa. -cil(t+lu).
These *-t- plurals could be a Central Dravidian innovation, but given their
variable forms that seems unlikely. Moreover, since *-t- takes either *-k
or as suffixes, these forms appear to be compound plurals consisting of
. .
old plurals in *-t- remade with the newer and more productive *-ku *-lu plurals

This leads to the following interpretation. Both Dravidian and Uralian had
-c
a dual in *-k which)when dual distinction was lost (totally in Dravidian,
partially in Uralian))became a plural (-k, -n in Uralian, *-ku in Dravidian).
In Dravidian a new plural suffix (*-lu) was used originally as a suffix to

both the dual *-k (*ku-lu) and the older plural *-t-(*t-lu) or was inde-
.
pendently suffixed directly to the noun base. On this reading, Dravidian
21
has lost the dual distinction, but has used the Proto-Dravido-Uralian dual
suffix *k- as a generalized plural, leaving evidence of its past dual
functions in the compound forms derived from *ku + lu. The Dravidian
*-t- plurals would either be derived from a common Proto-Dravido-Uralian
*-t used perhaps for low-class nouns or it is somehow related to the
high-class plural *-ri.
CASE:
Both Dravidian and Uralian have case systems in which case morphemes
are suffixed to noun bases. Both have a set of "core" cases that daughter
0 -
languages expand by various processes. It is doubtful
that either language family has a reconstructible system of cases beyond
those of a nominative suffix in both Uralian and Dravidian), a
genitive, and an accusative. To these one might add a locative or
locative/instrumental and a dative or dative/lative, even though these are
probably either derived from other nouns or are secondary forms suffixed to
the genitive.
Collinder reconstructs *-n as the proto-suffix of the genitive in
Uralian. Dravidian has -nV or -tV in most daughter languages. Where -tV
is not the form of the genitive itself, it is reported as an oblique suffix
Q
to which other suffixes and may be added. In other instances
-nV and -tV appear in what is likely a combined form -ntV. Several inter-
pretations are possible here, but one is that neither -nV nor -tV is a
genitive at all, but both are instead oblique bases to which other suffixes
and pjst-positions are added. For the genitive, possession would be indicated
22
either by the oblique itself or by an adjective suffix corresponding to
the -V in -nV and -tV (probably *a). This may well have been the Pre-
Dravidian situation. It may also have been the circumstance in Pre-
Uralian, for there, too, Collinder notes the occurrence of genitives in
,.
-t in Ket, 'but more important is the occurrence of *ta-/*ta as a
partitive ot separative case suffix. Since most of the functions served by
this suffix in Uralian languages are served by the genitive in Dravidian,
0
this strngFlY suggests that the Uralian genitive and partitive are related
I..!/
to one another and have the same morphological representation as the
genitive/oblique in Dravidian, i.e. -n and -t. We can thus set up common
Proto-Dravido-Uralian "genitives" in *-nV- and *-tV, which originally
distributed respectively to "high class" and "low class" nouns, and which
_ base_s..-the genitive notion being conveyed either
by these bases themselves or by an adjective suffix, probably *-a.
Originally, these oblique bases would have been the bases for all other
"cases" (except the nominative and accusative) and
Note that phonologically similar suffixes function adjectivally
in inflection, being the commonest forms of the past and non-past
verbal adjectives or participles. They are thus often indistinguishable
from tense suffixes, which probably means that at some remote time tense
suffixes were derived from those same forms, but this is speculative and
outside the discussion here.
Collinder reconstructs *-m as the Proto-Uralian accusative suffix.
For Prato-Dravidian, the accusative is *-ae, *-e, *-n, *-m, with.'the latter
being the form used in early Tamil script. Since the vowel accusatives are
--------------------
---
rv:
23
(.
probably de-nasalized (<e.g. *-aA), the form of the Proto-Dravidian
accusative was probably *-n/*-m, but since the *-n forms overlap with the
e C(
genitive *-n, there would be a for *-n to be dropped or strengthened
in some way. Collinder notes this same overlap between accusative and
genitive in Uralian where -n- has become the form of both genitive and
accusative in several languages. This points then, to the reconstruction of
a Proto-Dravido-Uralian accusative suffix in *-m.
Both Dravidian and Uralian have a variety of locative expressions,
'
but Collinder reconstructs *-na/*-na as the Proto-Uralian locative suffix.
In Dravidian languages many locatives derive from *ulle "to be inside" but
there are, particularly in Central Dravidian)locatives or locative-
instrumental suffixes in -ne/-ine and -te/-te, where the locative is prob-
ably expressed by the vowel -e, and the consonantal part is the oblique
suffix -n/-t. Despite the variability here, it seems likely that both
Dravidian and Uralian have reflexes either of a Proto-Dravido-Uralian loca-
tive suffix in *-ne or both show the same development of a case suffix
from the oblique base. Note too, that many Finno-Ugric languages use a

suffix as a locative, paralleling the distribution of n/t
.
locatives in Dravidian. Consequently, we can either set up *-ne/*-te or *-e
as the locative with the proviso that the latter be suffixed to the
oblique *-n/*-t.
Proto-Dravidian has a dative suffix *-kku/*-ku, which is sometimes
suffixed to nouns in the same way as the accusative and the genitive, but in
other instances is suffixed to the oblique base. Collinder reconstructs
*-kV as a lative suffix. Its reflexes in the daughter languages express
24
several "dative" functions, e.g. translative, lative, and motion toward
or away from a source or goal. There is then, no difficulty in supposing
that this Proto-Uralian *-kV lative corresponds to the Prato-Dravidian
-rlU--
dative and we can reconstruct *-kV dative.
'('" 0
Other cases can doubtlessly be reconstructed,
but for present purposes it is sufficient to show an over-all correspondence
between Prato-Dravidian and Proto-Uralian in the manner of forming gram-
matical constructions involving case relations specific correspondences
between the phonological realizations of the morphemes representing the
most important case relations, v,z., the nominative, genitive, accusative,
locative, and dative.
VERB FLEXION:
Dravidian and Uralian have a common verb structure consisting of a
verbal base + tense/mode suffixes + personal ending suffixes. As indicated
in the discussion of pronouns, the personal suffixes of verbs are derived
from the pronouns both in Dravidian and in Uralian. Both agree in having
a tense opposition between the past and the non-past. They also
agree.in the opposition of first and second person to third person in the
sense that the latter either has no special suffix or that the singular/
plural distinction is less differentiated.
Among the morphemic realizations of the non-past in Uralian, the
most important is reconstructed by Collinder as *-k. This morpheme had two
functions, one as a characteristic of the present tense, and the other as
a mood charac-teristic of the imperative. Dravidian, too, uses various
25
forms of a *-k suffix in forming either a present tense or a present-
future tense. As in Uralian, this same suffix is also often used in the
formation of the hortative/imperative morpheme.
/ /
Collinder reconstructs *-i "' *-Ji, and *s "' *c as morphemes of the
Uralian preterite. Among the Dravidian past tense suffixes reconstructed by
Subrahmanyam (1971:189-90), are: *-i, *-cc, *-t, and *-tt. Since Dravidian
*t frequently corresponds to Uralian * and *{, it is likely that some of
/ ;'
the Uralian *-s, *-c morphemes are reflexes of Proto-Dravido-Uralian *t,
and we can thus provisionally establish *-i, *-t, and *-c as Proto-Dravido-
Uralian past tense morphemes, even though these correspondences will need
to be checked in a closer study of verbal roots in Uralian.
Dravidian and Uralian also have, as a proto-feature, special verb
inflections involving the differentiation-of first and second person on the
one hand from third person on the other. In Uralian this is the distinc-
tion between the "objective" and "subjective" conjugations of the verb. In
the objective conjugation both the subject and the direct object are indi-
cated at the same time. When the finite verb has a definite direct object
and the subject is in the first or second person, the personal ending of
.
the v ~ r indicates the number of the direct object by means of a suffix
interposed between the tense suffix arid. the personal ending, thus: base +
tense + object suffix + person suffix. This is opposed to the subjective
conjugation which omits the direct object suffix from the flexion. In
other instances, the objective has a form of the personal suffix different
from that of the subjective conjugation. This suffix indicates both the
subject and the direct object.
26
These distinctions are paralleled in Dravidian by a similar special
conjugation. One conjugation is used only when the object of the verb
is the third person, but when the object {direct or indirect) is the first
or second person all of the verb tenses are formed on a special base con-
sisting of a verbal root plus a non-tense/mode suffix. The tense suffixes
and the personal suffixes are then added to this augmented base.
Although the morphological structure of the Dravidian special con-
jugation does not correspond to the Uralian objective, it is clear that
their functions correspond. Since these special conjugations tend to be only
sporadically represented in each language family, but sufficiently so as to
be reconstructed as part of their respective proto-systems, there can be
little doubt that they are archaic formations and that they represent a
kind of specialization that, though it may not be unique, is certainly more
likely to be part of a common heritage than a chance innovation among unre-
lated languages.
Finally, Dravidian and Uralian have both an affirmative and a nega-
tive conjugation. Again, as in the case of the special conjugation above,
despjte differences in formation, there is an over-all similarity which
extends in this case to a common negative morpheme {*ci-11- in Dravidian
and *ell- in Uralian) pointing to a Proto-Dravido-Uralian negative base
*cill-. In both languages this negative is inflected for person and number,
and enters into all the other verb functions {conditional, imperative,
adjective, etc.), but is often only partially inflected for tense, using in-
stead periphrastic forms.
A summary statement of the comparative data for Prato-Dravidian and
----
27
Proto-Uralian is given in Table 5.
Table 5 about here
CONCLUSION:
In sum, there is a close correspondence between the verbal systems
of the two languages which includes not only correlations of form such as
the four-way classification of affirmative vs. negative and objective vs.
subjective, the structure of the verbal base, the order of suffixing, and
the kind of suffix, but also many morphological realizations of these formal
categoc
So too, with the nominal system. Both languages have a common
structure involving the order of suffixed nominal elements consisting of
0
case suffixes and pkst-positions. These cases have corresponding phonologi-
cal realizations, and there is also a correspondence for the plural. The form
and realization of the pronominal systems are correlative, as are their
numeral systems, at least in part. Together with the phonological and
correspondences worked out in other publications, we now have
the beginnings of a reconstructed Proto-Dravido-Uralian language. Many
details, of course, remain to be worked out, but the general character of
the reconstruction is evident.
Footnotes
1. A shorter version of this paper was read at the Rice University

Symposium on Genetic Classification of Languages. It has benefitted
from discussion with various symposium participants.
2. For all Uralian morpheme reconstructions I have relied throughout
on Collinder (1960, esp. pp. 220-51; 1965, esp. pp. 122-35).
Dravidian morpheme reconstructions are my own, but for an excellent
account of Prato-Dravidian verb flexion and its morphology, see
Subrahmanyam (1971). On Dravidian cases, see the papers in
Agesthialingom and Gowda (1976).
BIBLIOGRAPHY
Agesthialingom, S., & K. Kushalappa Gowda (eds.).
1976 Dravidian Case System. Annamalainagar: Annamalai University.
Boisson, Claude
1985 Quelques Ressemblances Lexicales entre Elamite,
et Dravidien Ms.
Burrow, Thomas, & M. B. Emeneau
1961 A Dravidian Etymological Dictionary. Oxford: Clarendon Press.
1968 A Dravidian Etymological Dictionary: Supplement. Oxford:
Clarendon Press.
Collinder, Bjorn
1955 Fenno-Ugric Vocabulary. Stockholm: Almqv.ist & Wicksell.
----
1957 Survey of the Oralie Languages. Stockholm: Almqvist &
Wicksell.
----
1960 Comparative Grammar of the Oralie Languages. Stockholm:
Almqvist & Wicksell.
1965 An Introduction to the Oralie Languages. Berkeley: University
----

of California Press.
Fujiwara, Akira.
0
1981 The Japanese-Dravidian Vocabulary of and Fauna. Bulletin
of the International Institute for Linguistic Sciences. Kyoto
Sangyo University. 2:4:73-97.
. '
v ,/
Illic-Svityc, V. M.
1971-84
v
Opyt Sravnenija Nostraticeskix Jazykov (semito-xmitskij,
kartvel'skij, indoevropeskij, ural'skij, dravidijskij,
al taj skij ) Moskva: "Nauka"
McAlpin, David W.
1973 Toward Proto-Elamo-Dravidian. Language 50:1:89-101.
Marlow, Johanna Pudas
1974 More on the Uralo-Dravidian Relationship: A Comparison of
Uralic and Dravidian Etymological Vocabularies. Unpublished
Ph.D. dissertation. University of Texas: Austin, Texas.
Menges, Karl H.
1964 Altajisch und Dravidisch. Orbis 13:66-103.
----
1969 The Dravido-Altaic relationship. Journal of Tamil Studies.
1:35-39.
Michener, John E.
1978 Studies in the Indus Valley Inscriptions. New Delhi:
Oxford & IBH.
P. S.
1971 Dravidian Verb Morphology: A Comparative Study. Anna-
malainagar: Annamalai University
Tyler, Stephen A.
1968 Dravidian and Uralian: the lexical evidence. Language
44:798-812.
Kamil
1970 Comparative Dravidian Phonology. The Hague: Mouton.
---- --- ------

You might also like