ELECTRONIC JOURNAL OF VEDIC STUDIES,

     Vol.  5 (1999), issue 1 (September)
===============================================================================

Editor-in-Chief: Michael Witzel,        Harvard University

Managing Editor: Enrica Garzilli,       University of Perugia
Assistant Editor: Makoto Fushimi,       Harvard University

Technical Assistance: Ludovico Magnocavallo, Milano

       Editorial Board:

Madhav Deshpande        University of Michigan, Ann Arbor
Harry Falk              Freie Universitaet Berlin
Yasuke Ikari            Kyoto University
Boris Oguibenine        University of Strasbourg
Asko Parpola            University of Helsinki
---------------------------------------------------------------------------
email: ejvs-list@shore.net
witzel@fas.harvard.edu
www1.shore.net/~india/ejvs

(C) With the authors and the editors                   ISSN 1084-7561
-----------------------------------------------------------------------------



Editor's note


	There has been, for the past few years, a lively discussion, in
India and in some sections of her expatriate communities, but to much
lesser degree in 'western' Indology, of the so-called 'Aryan Invasion'
theory. The outright refusal of any immigration from outside the
subcontinent of speakers of (Indo-)Aryan languages such as Vedic Sanskrit,
Nuristani (Kafiri) or Old Iranian (Avestan) is based on disparate arguments
stemming from observations in the Vedic texts, in archaeology, geology,
astronomy, etc. However, language, if handled according to the well
established and internationally accepted rules of historical and
comparative linguistics that are used in the comparison of all human
languages, is an independent and unbiased yardstick, with which prehistoric
periods can be evaluated. The following paper investigates the various
layers of South Asian languages that have been superimposed upon each other
during the past few millennia. As will be seen, the usual "Aryan"-Dravidian
divide is much too simple a concept to fit reality.
	Just as today, prehistoric South Asia was home to a number of
language families, only some of which have survived as today's families,
i.e. Indo-European (Sanskrit, Hindi etc.), Dravidian (Tamil, etc.), Munda
(Santali, Mundari etc.), Tibeto-Burmese (Newari, Naga, Manipuri, etc.); in
addition there are the remnant, isolated languages such as Burushaski in N.
Pakistan, Kusunda in Central Nepal, Nahali in Central India, Vedda in Sri
Lanka, and Andamanese. Further, it will become obvious that today's
languages, for example Hindi, contain ancient layers that go back beyond
all Aryan, Munda or Dravidian and that are close to the language of the
first immigrating groups of Homo Sapiens sapiens, that entered South Asia
from Africa, via Western Asia, some 30,000 years ago.
	Providing this kind of impartial historical relief to the ongoing
discussion hopefully will steer it into calmer waters.

								M. Witzel



Note on transcription.

Vedic and Sanskrit are transcribed here according to the usual
Kyoto-Harvard system. Old Iranian presents a number of challenges in a 7 or
8 bit system. The Avesta alphabet is represented here as follows.

a A � � a~  a.~  @(schwa)   @@ (long schwa)
e �    		o � i I u U
k x x' xv 	g g' gh
c j
t th d dh  t~
p f  b w
ng ng' ngv n n' n~ m  m~
y' y v r
s z S Z S'  S~
h

In non-MacIntosh machines some of the above Avestan vowels will have
strange results. In other languages, similar conventions are followed, e.g.
Z = sh, velar fricative = gh, a~ nasalised a, etc. Paragraph sign = $.
Note that not all diacritical marks of Burushaski, etc., could be
represented in this 7/8 bit email version.
A web version with better diacritics (using Acrobat Reader, etc. ) will be
available soon at our website.
Due to its length the paper is sent in four parts, a, b, c, d  (as
indicated below)
================================================================================
(C)  COPYRIGHT NOTICE:
A COPY OF THIS ARTICLE MAY BE MADE FOR PRIVATE STUDY. ALL COPIES MADE FOR
WHATEVER PURPOSE MUST INCLUDE THIS COPYRIGHT NOTICE.
THE TEXT MAY NOT BE MODIFIED IN ANY WAY NOR MAY IT BE REPRODUCED IN
ELECTRONIC OR ANY OTHER FORMAT WITHOUT THE WRITTEN PERMISSION OF THE
EDITOR-IN-CHIEF.

THE FOLLOWING MATERIALS WERE FIRST PUBLISHED IN THE ELECTRONIC JOURNAL OF
VEDIC STUDIES.

ALL INQUIRIES ARE TO BE SENT TO THE EDITORS.

ELECTRONIC JOURNAL OF VEDIC STUDIES, Vol. 5 (1999), issue 1 (Sept.)
================================================================================
Michael Witzel
(Harvard University)

				Substrate Languages in Old Indo-Aryan

				(Rgvedic, Middle and Late Vedic)

		<part a>

$ 0. Definitions
$ 1. Greater Panjab
$ 1.1. Rgveda substrate words
$ 1.2. Para-Munda loan words in the Rgveda
$ 1.3. Para-Munda and the Indus language of the Panjab
$ 1.4. Munda and Para-Munda names
$ 1.5. Other Panjab substrates
$ 1.6. Dravidian in the Middle and Late Rgveda

		<part b>

$ 1.7. Greater Sindh
$ 1.8. The languages of Sindh
$ 1.9. The Southern Indus language: Meluhhan
$ 1.10. Further dialect differences
$ 1.11. Dravidian immigration

$ 2. Eastern Panjab and Upper Gangetic Plains
$ 2.1. The Kuru realm
$ 2.2. The substrates of Kuru-Pancala Vedic
$ 2.3. The Para-Munda substrate

		<part c>

$ 2.4. Substrates of the Lower Gangetic Plains and "Language X"
$ 2.5. Tibeto-Burmese
$ 2.6. Other Himalayan Languages
$ 3. Central and South India
$ 4. The Northwest
$ 5. Indo-Iranian substrates from Central Asia and Iran
$ 6. Conclusions.

		<part d>

Abbreviations and Biblography
Annotations

	The languages spoken in the northern part of the Indian
subcontinent in prehistoric times have been discussed throughout most of
this century. This concerns the periods of the Rgveda and of the Indus or
Harappan Civilization (nowadays also called Indus-Sarasvati civilization in
some quarters). Since the Twenties, the area of the newly discovered Indus
civilization has been regarded, beginning with J. Bloch, as having been
populated by Dravidian speakers, while other early 20th century scholars
such as S. Le'vy and J. Przyludski have stressed the Austro-Asiatic (Munda)
substrate of Northern India, -- both are positions that have been
maintained until today (e.g., Burrow, Emeneau, Parpola vs. Kuiper, Hock,
Southworth). The relationship of these languages to the archaic (Vedic)
form of Sanskrit has played a major role in such discussions. Both
Dravidian and Munda have usually been understood as having preceded, as
substrate languages, the introduction of Old Indo-Aryan (Vedic Sanskrit).
Such suppositions will be investigated in this paper, and evidence will be
produced indicating that the linguistic picture of this early period of
South Asia was much more complex -- as complex, indeed, as that of modern
India.

$ 0. Definitions.

	$ 0.1. By way of introduction, as few definitions are in order.
First of all, in must be stressed that Vedic, Dravidian and Munda belong to
three different language families (respectively, Indo-European, Dravidian
and Austro-Asiatic). Since this is no longer recognized in some of the more
popular sections of the press and the publishing business, it must be
pointed out that the recognition of basic differences between language
groups (in word formation, declension, conjugation and in syntax) is a well
established item of linguistic science that applies to all human languages
(summaries by Hock 1986, Anttila 1989). One cannot make an exception just
for the subcontinent and claim that South Asian languages are so similar
that they belong to a new linguistic 'family' (S. Kak).
	What South Asian languages indeed have in common are certain
features, especially some of syntax, that are due to long standing
bilingual contacts and that make them appear superficially similar, just
as, for the same reasons, the Balkan languages Rumanian, Bulgarian,
Albanian, and Greek share some peculiarities which make translation between
them easy. Nevertheless, nobody in Europe or elsewhere would deny that they
belong, respectively, to the Romance, Slavic, Albanian and Greek
sub-branches of Indo-European (IE), and it is not maintained that they form
a new 'Balkan family'.
	Of course, the South Asian languages also share a lot of common
cultural vocabulary derived from Sanskrit (sometimes effectively disguised
by the development of the language in question, especially in Tamil), just
as European languages, whether IE, Uralic, Basque or even Turkish share
many Greek and Latin words of culture and science, and more recently, of
technology.
	$ 0.2. Secondly, the materials available for this study have to be
reviewed briefly. Since we cannot yet read the Indus script with any
confidence (Possehl, 1996b, discusses the rationale of some 50 failed
attempts), we have to turn to the Vedic texts first.
	I will concentrate here on evidence from the Vedas as they are
earlier than Drav. texts by at least a thousand years. This also has the
advantage that the oldest linguistic data of the region are used, which is
important because of the quick changes that some of the languages involved
have undergone. The Vedas provide our most ancient sources for the Old
Indo-Aryan variety (IA; OIA = Vedic Sanskrit) of the Indo-Iranian branch
(IIr. = Old Iranian, Nuristani and Old Indo-Aryan) of the Indo-European
language family (IE = Celtic, Germanic, Italic, Slavic, Greek, Hittite,
Tocharian, etc.) that are spoken in the subcontinent. However, these texts
also contain the oldest available attestation for non-Indo-European words
in the subcontinent (Dravidian, Munda, etc.)
	$ 0.3. The Vedas were orally composed (roughly, between 1500-500
BCE) in parts of present day Afghanistan, northern Pakistan and northern
India. To this day, their oral transmission has been exceptionally good, as
is commonly known. They are followed by the early Dravidian sources,
represented by the ancient Tamil "Sangam" (CaGkam) texts of South India
(stemming from the beginning of our era); however, these texts still are
virtually unexplored as far as non-IA and non-Drav. substrates and
adstrates from neighboring languages are concerned. From a slightly earlier
period than the Sangam texts comes the Buddhist Pali canon of (western)
Northern India; it has been composed in an old form of Middle Indo-Aryan
(MIA). The Epic texts (mahAbhArata, rAmAyaNa) were composed by a host of
bards from various parts of northern India in a form of Sanskrit that is
heavily influenced by MIA.
	In order to evaluate the substrate materials, the time frame and
the geographical spread of these texts have to be established first. The
procedures to arrive at a fairly secure dating cannot be discussed here in
any detail; this would take another long paper. It may suffice to point out
(Witzel 1987, 1989, 1995, 1999) that the Rgveda (RV) is a bronze age
(pre-iron age) text of the Greater Panjab that follows the dissolution of
the Indus civilization (at c. 1900 BCE) -- which limits its time frame to
(maximally) c. 1900 -1200 BCE; the latter date is that of the earliest
appearance of iron in the subcontinent. The RV is followed by a number of
other Vedic texts, usually listed as saMhitAs, brAhmaNas, AraNyakas and
upaniSads. Linguistically, however, we have to distinguish five distinct
levels: (1) Rgveda, (2) other saMhitAs (mantra language), (3) yajurveda
saMhitA prose, (4) earlier and later brAhmaNas (incl. AraNyakas and
upaniSads) and (5) the late Vedic sUtras (Witzel 1987, 1997; for
abbreviations of names of texts, their dates and their geographical
location see attached list).
	While the area of the RV, as clearly visible in the mentioning of
the major rivers, is the Greater Panjab (with the inclusion of many areas
of Afghanistan from Sistan/Arachosia to Kabul/Gandhara), its temporal
horizon consists of three stages, roughly datable between c. 1700-1200 BCE
(Witzel 1995, 1999, J. R. Gardner, Thesis Iowa U. 1998, Th. Proferes, Ph.D.
Thesis, Harvard U. 1999). They are:
#  I. the early Rgvedic period (fn.1): c. 1700-1500 BCE: books (maNDala) 4,
5, 6, and maybe book 2, with the early hymns referring to the yadu-turvaza,
anu-druhyu tribes;
#  II. the middle (main) Rgvedic period, c. 1500-1350 BCE: books 3, 7, 8.
1-66 and 1. 51-191; with a focus on the bharata chieftain sudAs and his
ancestors, and his rivals, notably trasadasyu, of the closely related PUru
tribe.
#  III. the late Rgvedic period, c. 1350-1200 BCE:
books 1.1-50, 8.67-103, 10.1-854; 10.85-191: with the descendant of the
pUru chieftain trasadasyu, kuruzravaNa, and the emergence of the
super-tribe of the kuru (under the post-RV parikSit, Witzel 1997).
These levels have been established, not on the basis of linguistic
criteria, but on the basis and by the internal criteria of textual
arrangement, of the 'royal' lineages, and independently from these, those
of the poets (RSis) who composed the hymns. About both groups of persons we
know enough to be able to establish pedigrees which sustain each other.
Applying this framework to the linguistic features found in the various
maNDalas of the Rgveda, we are in store for some surprises.
	$ 0.4. Before coming to this, however, another item must be
discussed briefly, that of the concept of substrates. The RV contains some
300 words, that is roughly 4% of its hieratic vocabulary, that are not
Indo-Aryan (Kuiper 1991). It is possible to establish their non-IA
character by studying their very structure. For, words belonging to a
certain language follow well-established patterns. The word structure of
English (or IE in general) is well known. In English, for example, a word
cannot start with tl- or pt-. Words such as Tlaloc, an Aztec god, are
impossible, and those in pt- are loans from Greek, such as Ptolemy. Whorf's
structural formula of English monosyllablic words (Language, Thought and
Reality, 1956; simplified):
			{ 0, (s+/-) C-ng   + V  + 0, C-h }
allows to predict that English words beginning in ngo-  or ending  -goh are
not possible. If ng- or nk- do occur now, they are late loans from African
languages (e.g., Nkrumah); or, before the influx of Yiddish or German words
into American English, sh + consonant also was not allowed, while we now
have: to shlep or strudel, as opposed to older words such as to slip or to
stride. These examples also show that foreign words can enter a host
language in pronunciations close to their original ones (however, strudel
does not have the German but the American -r-), and that, at the same time,
at they can easily be detected if they violate the original structure of
the language in question.
	IE nouns and verbs have three parts: root (dhAtu), suffix
(pratyaya) and ending, such as dev-a-M zaMs-a-ti "he praises the god." The
root (dhAtu), the part of the word carrying the lexical meaning (dev
"heavenly", SaMs "praise"), is enlarged by suffixes (immediate/primary:
kRt, secondary: taddhita). They are attached (here: -a-) to the root and
are followed by the noun endings (-m) or verb endings (-ti). IE roots
ordinarily have three consonants, and can only have the structure given
below, where ( ) indicates possible appareance; b is very rare in IE; C =
consonant (includes the laryngeal sounds, H  = h1, h2, h3); e = standard IE
vowel (> Skt. a); it can change to o (> Skt. a), E, O (> Skt. A) or
disappear (zero forms); R = resonants, the "semi-vowels" y, r, l, v and m,
n which can also appear as i, R, L, u, a, a; further, s when found at the
beginning of roots, is unstable and can disappear (as in spaz 'spy' :
paz-ya-ti 'he sees'). IE/IA/Vedic roots must conform to the following
formula (Szemere'nyi 1970):

		prefixes +/- {(s) (C) (R) (e) (R) (C/s)} +/- suffixes

	Possible thus are, e.g., Skt. ad (eC), pat (CeC), zrath (CReC),
bandh (CeRC), kR (CR), zru (CRR), kram (CReR), krand (CReRC), i (R), iS
(RC), man (ReR), manth (ReRC), tras (CRes), tvakS (CReKs), stambh (sCeRC),
svap (sReC), sas (ses) etc.; with laryngeals: bhU (CRH), brU (CRRH), IkS
(HRCs), as (Hes), etc. Sounds inside a root are arranged according to the
following order of preference: C/s-R-e, thus : CRe-(Skt. zram...), sRe-
(Skt. srav...) are allowed, but not: RCe-, Rse- (Skt. *Rka..., *usa...).
Not allowed in IE are the following consonant groupings in a root, the
types: bed, bhet, tebh, pep, teurk/tekt (Skt. *bad, bhad, tabh, pap, tork,
takt) This classification of possible roots often allows to classify non-IE
roots and words at a glance.
	The number of primary suffixes is limited to certain types, usually
*Ce, CR, CRe, R, Re, es (Skt. -ta, -ti, -tra, -i, -ya, -as) etc. Secondary
suffixes build up on the primary ones, thus Skt. -u-mant, -a-tAt, -a-mAna,
etc. On the other hand, suffixes such as -Aza, -Ta, -an-da/-a-nda-,
-bUth-a/-bU-th-a (see below) do not exist in IE and IIr. Therefore, the
very structure of many of the 'foreign' and loan words in the RV simply do
not fit the IE structure of those properly belonging to Ved. Sanskrit (just
as Nkrumah, Mfume must be foreign words viz. recent loans in English).
Consequently, RV words such as kInAza, KIkaTa, Pramaganda, BalbUtha, BRbu,
BRsaya are simply not explainable in terms of IE or IIr: the verbal/nominal
roots kIn, kIk, mag, balb, bRs do not exist in IE as only roots of the
format {(cons.) (R) e (R) (cons.)} are allowed and as b is very rare in IE;
further, only S (but not s) is allowed in Vedic after i,u,r,k, and finally,
the suffixes -A-za, -Ta, -an-da/-a-nda-, -bUth-a/-bU-th-a do not exist in
IE/IIr.

	$ 0.5. The structure of RV words has already been studied at some
length by former colleague at Leiden and one of my several great teachers,
F.B.J. Kuiper (1991, cf. 1955). However in this small book, written at the
age of 85, he limited his task to a discussion of their structure and to
pointing out some features which link them to Dravidian and Munda, and, as
he conceded, "maybe to some unknown language(s)." Therefore, he did not
proceed to discuss the Indus language, nor did he study the various levels
of Rgvedic speech beyond the usual division into older (books 2-7, etc.)
and late RV (book 10). However, as soon as we apply the three stage
leveling discussed above, a different picture of the RV and the subsequent
Vedic texts emerges than known so far. To sum up, we can distinguish the
following substrate languages.

- A Central Asian substrate in the oldest Rgvedic;
- RV I: no Dravidian substrate but that of a prefixing Para-Mundic (or
Para-Austroasiatic) language, along with a few hints of Masica's U.P.
Language "X", and some others;
- RV II and III: first influx of Dravidian words;
- Post-RV (YV, AV Mantras <MS, KS, TS, VS, AV, PS> and later Vedic):
continuing influx of the same types of vocabulary into the educated Vedic
speech of the Brahmins; occurence of Proto-Munda names in eastern North
India.
- Other substrates include Proto-Burushaski in the northwest,
Tibeto-Burmese in the Himalayas and in Kosala, Dravidian in Sindh, Gujarat
and Central India, and predecessors of remnants language groups, now found
in isolated pockets of the subcontinent (Kusunda in C. Nepal, pre-Tharu in
S. Nepal/UP, Nahali in C. India, and the pre-Nilgiri and Vedda substrates).

	So far, linguists have concentrated on finding Dravidian and Munda
reflexes, especially in the oldest Veda, the Rgveda (RV). These studies are
summed up conveniently in the etymological dictionaries by M. Mayrhofer
(Indo-Aryan; KEWA, EWA), Th. Burrow - M.B. Emeneau (Dravidian; DED, DEDR),
and in the work of F.B.J. Kuiper (Munda/Austro-Asiatic; 1948, 1955, 1991,
Pinnow 1959). In addition, it has especially been F. Southworth who has
done comparative work on the linguistic history of India (IA, Drav., Munda)
during the past few decades; his book on the subject is eagerly awaited.
	These items will be discussed in some detail below, including a
discussion of the procedures followed as well as some examples for these
substrates. Finally, the conclusions we have to draw from the complex
linguistic picture of Vedic times will be discussed.

$ 1. Greater Panjab

$ 1.1. Rgveda substrate words.
	The RV reflects the Panjab and its immediate surroundings of c.
1500-1200 BCE., most clearly visible in its river names, extending from the
Kabul River to the yamunA (mod. Jamna) and even the Ganges (gaGgA,
mentioned only twice) and it represents evidence from the three subsequent
historical periods mentioned above. It is important to note that RV level I
has no Dravidian loan words at all (details, below $ 1.6); they begin to
appear only in RV level II and III.
	Instead, we find more some three hundred words from one or more
unknown language(s), especially one working with prefixes. Prefixes are
typical neither for Dravidian nor for Burushaski (cf. Kuiper 1991: 39 sqq.,
53, see below). Note that the "prefixes" of Tibeto-Burm. (Benedict 1972) do
not agree with those of the RV substrate either. Their presence apparently
excludes also another unknown language which occasionally appears in the RV
and more frequently later on with typical gemination of certain consonant
groups (perhaps identical with Masica's "Language X" 1979, see below; cf.
Zide and Zide 1973:15). The prefixes of the RV substrate are, however,
close to, an in part even identical with those of Proto-Munda; taking my
clue from Kuiper (1962: 51,102; cf. now Zide MT II: 96), I will therefore
call this substrate language Para-Munda for the time being.

$ 1.2. Para-Munda loan words in the Rgveda
	We can start with the convenient list of Kuiper (1991), who does
not, however, discuss each of the 383 entries (some 4% of the hieratic RV
vocabulary!) This list has been criticized by Oberlies (1994) who retains
"only" 344-358 words, and minus those that are personal names, 211-250
'foreign' words.(fn. 2) One can, of course, discuss each entry in detail
(something that cannot be done here), but even Oberlies' lowest number
would be significant enough, in a hieratic text composed in the traditional
poetic speech of the Indo-Iranian tradition, to stand out, if not to
surprise. It is a clear indication of a strong substrate and of
amalgamation of IA speakers with the local tradition. In evaluating this
list, it must be said that it is much more difficult to discern
Para-Munda/Austro-Asiatic words, than to establish IA or Dravidian
etymologies, as an etymological dictionary of Munda is still outstanding
(in preparation by David Stampe et al.). Nevertheless, one can, for the
time being, make use of Pinnow's reconstructions of Proto-Munda in his
investigation of Kharia (1959), Bhattacharya's short list (1966: 28-40),
Zide & Zide's discussion of agricultural plants (1973, 1976), and Kuiper's
relevant studies (especially 1955, 1991; his 1948 book is still very
useful, in spite of his own disavowal of it, as a collection of relevant
materials). It must be stressed that neither the commonly found Drav. nor
Munda etymologies are up to the present standard of linguistic analysis,
where both the root and all affixes are explained. This is why most of the
subsequent etymologies have to be regarded as preliminary. (Note that only
a few examples are given below for each category; fuller details will be
included in a forthcoming paper and monograph).
	Among the '300 foreign' words of the RV, those with certain
prefixes are especially apt to be explained from Para-Munda (viz. directly
from Austro-Asiatic). However, "Owing to the typological change that has
taken place in these languages, only some petrified relicts remain" (Kuiper
1991: 39). Typical prefixes in modern Munda are such as p-, k-, m-, ro-,
ra-, ma-, a, @-, u-, ka- (Pinnow 1959:10 sqq.; cf. also the plural suffix
-ki in Kharia, p. 265 $341a, 211 $145c); some of them are indeed attested
in the c. 300 'foreign words' of the RV.
	Of special interest for the RV substrate are the prefixes ka-, ki-,
kI-, ku-, ke-, which relate to persons and animals (Pinnow 1959: 11; cf. p.
265 $341a) and which can be compared, in the rest of Austro-Asiatic, to the
'article' of Khasi (masc. u-, fem. ka-, pl. ki-, cf. Pinnow 1959: 14). The
following words in the RV are important, even if we cannot yet find
etymologies. (In the sequel, Sanskrit suffixes and prefixes are separated
from the substrate word in question).
#  The Prefix ka-:
	kakardu 'wooden stick', EWA I 286 'unclear';
	kapard-in 'with hair knot', Kuiper 1955: 241 sqq.; EWA I 299
'non-IE origin probable'; kabandh-in,
	kavandha 'barrel' Kuiper 1948: 100. EWA I 327 'unclear';
	kAkambIra 'a certain tree', EWA I 334 'unclear'.
#  The prefix ki-:
	kimId-in 'a demon', EWA I 351 'unclear'; cf. zimida, zimidA 'a
demoness', Kuiper 1955: 182;
	kIkaTa 'a tribe' 3.53.14; EWA 'foreign name of unknown origin';
prefix kI- points to Austro-As.; cf. Sant. kaT- 'fierce, cruel', or common
totemic tribal name (like mara-Ta PS : Munda mara� 'peacock' IA matsya
'fish', kunti 'bird') ~ Sant. kaTkom 'crab'? cf. Shafer 1954: 107, 125;
	kIkasA (dual) 'vertebra, rib bone' 10.163.2, EWA I 355 'unclear';
"formation like pi-ppala, etc. and connected with lex. kazeruka..." Kuiper
1955: 147;
	kIja 'implement, spur?', 8.66.3; EWA I 355 'loan word possible';
KEWA I 214 and Kuiper 1955: 161, 165: 'doubtful Drav. etym.' (Burrow, BSOAS
12: 373);
	kInArA dual, 'two ploughmen' 10.106.10; EWA I 356 'probably
artificial for kInAza', rather z/D/r, Kuiper 1948: 6, 38, 1991: 30-33, and
1955: 155f., 1991: 26 on suffixes -Aza/-Ara, (cf. also -na/-ra in
rAspina/rAspira); on z as hyper-Sanskritization for S/r cf. vipAz; Kuiper
1991: 46 on suffix -za; if kInAra- contains a suffix, then probably no
prefix kI-.
	kInAza 'plough man' 4.57.8 (late), AV; Kuiper 1955: 155, 1991: 14,
26, 46 see kInAra; EWA I 356 'unclear'.
	kIlAla 'biestings, a sweet drink'; in AV 4.11.10 next to kInAza;
EWA I 358 'unclear'; discussion, above: Khowar kiLAl, Nuristani kilA' etc.,
Bur. kIlAy, Kuiper 1955: 150f., CDIAL 3181.
	kIsta 'praiser, poet' 1.127.7, 6.67.10, to be read as [kis@tAsaH]
Kuiper 1991:23, 1955:155; the unusual sequence -Is- (see introd.) points to
a loan word (Kuiper 1991:25); EWA I 358 'not clarified'; cf. Kuiper 1991:
20, 23, 25; to be compared with RV zISTa 8.53.4 with var. lect. zISTeSu,
zIrSTeSu, zIrSTresa, Kuiper 1991: 7, 71; this is Sanskritization of
*k'Is@teSu, Witzel 1999; cf. EWA II 644
#  ku-:
	kumAra 'boy, young man', EWA I 368 'not convincingly explained';
cf. CDIAL 3523, 13488; Kuiper 1955: 146f. compares Tel. koma 'young', Tam.
kommai, etc.; cf. zi(M)zu-mAra (see below); but note, in Munda: m@ndra, m@r
'man' (pers. comm. by D. Stampe).
	kurIra 'women's hair dress', 10.92.8, EWA I 371 'unclear', Kuiper
1955: 152, 1991: 14, 29-31 compares Tam. koTu 'horn, coil of hair', DEDR
2200
	kuruGga 8.4.19, name of a chieftain of the turvaza (cf. Kuiper
1991: 6, 17); EWA I 371 'unclear;' however, cf. kuluGga 'antelope', and the
frequent totemistic names of the Munda
	kuliza 'ax', EWA I 374 'not securely explained'; Kuiper 1955: 161,
163 compares Tam. kul~ir 'battle ax'; Skt. kuThara, kuddAla 'hoe', and
Sant., Mundari kutam 'to beat, hammer', Mundari, Ho kutasi 'hammer', Kan.
kuTTu 'to beat, strike, pound'; cf. Kuiper 1991:14; Berger 1963: 419
*kuDiza, from *kodez in Kharia khoNDe�j 'ax', Mundari koNDe�j 'smaller kind
of wood ax', with prefix kon- and Kharia te�j 'to break'

#  Double prefixes in C@r-.
More important, perhaps, are the so-called 'double prefixes' in
Austro-Asiatic, composed of a prefix (e.g. k-) followed by a second prefix
(mostly -n, see Pinnow 1959: 11). The use of k-n- is clear in names of
domesticated animals, in Sora kin-sod 'dog' : Kharia solog 'dog'; Sora
kim-med 'goat' : Remo -me�; k@m-bon 'pig' : Juang bu-tae (see Pinnow 1959:
168, cf. Jpn. buta, Austr. > Sino-Tib. *mba(gh)); Sora ken-sim 'chicken' :
Mundari sim. Such double prefixes seem to be rarer in Munda now than in
Eastern Austro-Asiatic; cf., nevertheless, Kuiper 1991: 94 on zar-varI
'night': za-bala 'variegated'; Kuiper 1948: 38 on kal-, kil-, p. 138 on the
prefix k-, 1948: 49f. 'prefix k@r-, kar-, and gala-'; note Sora kAr-dol
'being hungry' (D. Stampe, oral communication).
	The clearest Vedic example is, perhaps, Ved. jar-tila 'wild sesame'
AV : tila 'sesame' AV, (cf. tilvila 'fertile' RV, Kuiper 1955: 157,
tilpiJja, -I 'infertile sesame' AV, on Sumer. connections s. below). Double
prefixes are typical for the Rgvedic loans, especially formations with
consonant-vowel-r = C@r- (and also C@n-, C@m-), that were adapted in Vedic
with various IA vowels (R, ur, etc., see Kuiper 1991: 42 sqq.; cf. below on
Nepalese substrate words). Examples with C@r (and due to the common Vedic
interchange of r/l, also C@l-) include:
	karaJja name of a demon, EWA I 310 'unclear', cf. the tree name
karaJja, DEDR 1507 Kan., Tel. kAnagu, Konda karaG maran etc.; CDIAL 2785.
	karambha 'gruel', Kuiper 1991: 51 sqq. compares loan words with -b-
> -bh- (Pkt. karamba 'gruel'); -- rather with a prefix kar-
and popular etymology with ambhas- 'water' RV, or ambu 'water' Up., Mbh.
Kuiper 1991: 63; cf. also Kurukh, Malto amm 'water', but also Tamil am, Am
DEDR 187;
	karkandhu later tree name 'Zizyphus Jujuba', but personal name in
RV 1.112.6; EWA I 313 'not clear'; the Drav. word the meaning of karkandhu,
DEDR 475, 2070, 3293;
	khargalA 'owl' 7.104.17 (late), EWA I 448
	kalmalIk-in 'shining' 2.33.8; EWA I 325 'unclear'; however, cf.
kalmASa 'spotted', Kuiper 1948: 38; see below on kilbiSa
Further: kR- [k@r-], see Kuiper 1991: 40 sqq., 23:
	kRpITa 'bush, brush' EWA I 394 'unclear', cf. also kRmuka 'faggot,
wood' KS, CDIAL 3340a; 'unexplained' Kuiper 1955: 160
kRzana 'pearl', Urdhva-, kRzanA-vat, EWA I 396 'not securely explained';
Kuiper 1955: 152 compares kR-zana with other words for 'thick, round', such
as Skt. lex. zAni 'colocynth?'
	khRgala meaning unclear: 'staff, crutch, amulet, armor, brush?'
2.39.4; EWA I 494; cf. khargala 'owl', above, KhArgali PB? -- Kuiper 1948:
49f. 'well-known prefix k@r-, kar-, and gala-'
	kilbiSa 'evil action'; EWA I 354 'not sufficiently clear', Kuiper
1955: 175 compares TS, VS kalmASa 'spotted' and Epic kalmaSa, Pkt. kamaDha
(cf. Pinnow 1959: 379 sqq., Kuiper 1991:36 sqq.), Kuiper 1948: 38, 138 on
prefixes kal-, kil-, kar-; Sant. boDor, bode, murgu�c 'dirty', with
adaptation -S-/D- into Ved. similar to VipAz-/VibAl-/*VipAZ (see below).
	Due to the frequent interchange k[k']/z, (see below) the prefix
zar-/zal- belongs here as well (cf. kar-koTa-ka RVKh ~ zar-koTa AV):
	zarvarI 'night', api-zarvara; EWA II 621 compares *zarvar, zarman
'protection'; Kuiper 1955: 144 u. 1955: 170 compares zambara, karbura,
Kuiper 1991: 30 zabala 'variegated' with simple prefix, as compared with
prefix + infix (''double prefix") in zambara (cf. Kuiper 1948: 136)
	zalmali name of a tree, 'Salmalia malabarica', EWA II 622 'probably
not to be separated from RV 3.53.22 zimbala', CDIAL 12351 (not related
Tib.-Burm. *siG 'tree'); Kuiper 1991: 65 on cases with -lm- for -mm-:
'different dissimilations of *zamma/zimmal'.
	sRJjaya 'name of a person' 6.27.7 (next to turvaza), 4.15.4 (next
to daivavant), sArJjaya 'descendent of S.' 6.47.25; EWA II 743 supposes
connection with sRjaya 'a certain bird' KS, which would agree with the
totemistic names in Munda; cf. Kuiper 1991: 7, on non-IA tribal names in RV
sRbinda name of a demon 8.32.2; EWA II 744 with Kuiper 1991: 40,43 (and
earlier) on names such as ku-surubinda TS, PB, �B, kusur-binda JB and
bainda VS 'member of the tribe of the Binds' (probably also the name of the
Mountain range, post-Vedic vindh-ya), vi-bhindu RV 8.2.41, 1.116.20,
vi-bhindu-ka, vi-bhindu-kIya JB $203; cf. Kuiper 1939 = 1997: 3 sqq., 1955:
182, Witzel 1999).
 	In the same way, the prefixes jar, tar, nar, par, bar, zar, sR =
[j@r, t@r, etc.]: jarAyu, jarUtha (cf. also Ved. jar-tila : tila); taranta,
tarukSa, tRkSi, tRtsu, nAr-miNI, epithet of a fort; nAr-mara, probably the
area of or the chief of UrjayantI; parNaya, parpharI-ka, parzAna;
prakaGkata (next to: kaGkata), prakala, parpharvI, pramaganda (next to:
magadha), pra-skaNva, pharva-ra, phAriva; pRthi, pRthI, pR-dAku [p@r-dak-u]
< Munda da�k 'water'?, barjaha; (cf. also NAr-Sada RV, NAr-vidAla,
NAr-kavinda PS and *ku-bind in: Ved. ku-sur(u)-binda, bainda, vi-bhindu,
vi-bhindu-kI-ya).
	Furthermore, the formations with other vowels that are adaptations
of [-@r] as above in [k@r]: tirindi-ra, turIpa, turphari, turva/turvaza,
turvIti, tUrNAza, sUrmI.
	Instead of C@r, the much more common double prefix of Munda, C@n-,
C@m-, is found as well: kaGkata; zamba, zambara (cf. zabala!), zAmbara,
ziMzapA, ziMzumAra, ziJjAra, zimbala, zimbAta, zimyu. Compare also the
prefixes in C@s-: puSkara, puSya, rAspina, rAspira.
	Kuiper (1991: 39 sqq.) also discusses other prefixes, such as A-,
i-, u-, o-, ni-, bhR-, ma-, sa-, za-, hi-. Among them, the old prefix u-
(o-) would be of special interest; however, is found in the RV only in some
5 or 6 cases.
	A clear case is za-kunti(-kA) RV, za-kunta AV, Ved. za-kunta-ka
'bird', za-kuntalA name of a nymph, Ved. kunti a tribal name, next to the
matsya (IA, 'the Fishes'). The Ved. words belong to Kharia kon-the'd, Sora
on-tid@n, etc.; Korku ti-tid 'a certain bird', Ved. tit-tir-a 'partridge',
Pinnow 1959 160 : 336; cf. however RV za-kuna, za-kuni (Kuiper 1991:44).

$ 1.3. Para-Munda and the Indus language of the Panjab

	In short, Para-Munda prefixes are thus very common in the RV. One
has to agree with Kuiper 1991: 39f: "According to some scholars Munda was
never spoken west of Orissa, Bihar, Madhya Pradesh and eastern
Maharashtra... The obvious occurrence of Old Munda names in the Rigveda
points to the conclusion that this statement should be revised." If (some
of) the words quote above should not go back directly to Proto-Munda, one
may think, especially in the case of the untypical formation C@r, of an
unknown western Austro-Asiatic language, "Para-Munda" (cf. Kuiper 1962: 51,
102).
	If this initial interpretation is correct, several far-reaching
conclusions can be drawn. The very frequency itself of non-Drav. loan words
in the early (as well as in the later) RV is remarkable: it indicates a
much stronger non-Drav. substrate in the Panjab than usually admitted.
Because of the great similarity with Austro-Asiatic formations and because
of some already established (Para-)Munda etymologies (such as za-kunta ~
Kharia kon-the'd, etc., Pinnow 1959 160 : 336), this substrate is likely to
be an early form of western Austro-Asiatic.
	Is the Indus language therefore a kind of Proto-Munda? Against this
may speak first of all, as Kuiper states (1991), that the RV substrate does
not have infixes like Munda. However, -n-infixes can be adduced in
ka-bandha/ka-vandha, kar-kandhu, gandhA-ri, pra-maganda, za-kunti < PMunda
*Sa-kontid, sR-binda and, e.g., in post-RV ku-sur(u)-binda, bainda,
vi-bhindu, vi-bhindu-kI-ya. Yet, the substrate may be a very early form of
Munda (or another variety of Austro-Asiatic) which still used prefixes
actively, just like the eastern Austro-As. languages, e.g. Mon, Khmer, do
even today (cf. also below, on Sumerian). Further, the infixes may have
developed from prefixes which had found their way into the root (Pinnow
1959: 15). Among these, one can include 'double' prefixes such as k@-r-,
S@-r-, p@-r- etc. (Pinnow 1959: 11). If this is correct, then Rgvedic
Proto-Munda represents a very old stage of Austro-Asiatic indeed, something
that does not surprise for a text of c. 1500 BCE.

$ 1.4. Munda and Para-Munda names
	However, direct contact of the non-Indo-Aryan words in the RV with
predecessors of present day Munda languages is more complex. Some of the
substrate words may, at least in part, have entered the RV through the
intervention of the Indus language (lAGgala etc., see below). Yet, there
also are a few direct correspondences with reconstructed Proto-Munda
(za-kunta < *kon-ti�d) which indicate the archaic character of the
para-Mundic Indus language. For example, the name of pramaganda, the
chieftain of the kIkaTa (RV 3.53.14) who lived south of kurukSetra (cf.
Witzel 1995). Both words are non-Indo-Aryan and they show clear indications
of Mundic character: maganda can be explained as ma-gand with the old, now
unproductive Munda prefix ma- that indicates possession. The word gand may
belong to Munda *gad/gaD, ga-n-d/gaND (Pinnow 1959: 351 $498) that is also
seen in gaNDa-kI, gaGgA (Witzel 1999, if not modeled after the tribal names
aGga, vaGga, see below), W. Nepali gAD (as 'suffix' of river names, Witzel
1993) and apparently also in ma-gadha (with Sanskritization > dh). Kuiper
1991: 43f. (8, 21, 96, also 1955) has explained the prefix pra- [p@r] (cf.
prefixes such as k@r-/S@r-) from Munda, which looks perfectly Indo-Aryan
but in this case certainly is 'foreign' (p@r 'son of'? Kuiper 1991: 43).
The tribe of chief pra-maganda, the kIkaTa, has either the typical 'tribal'
suffix -Ta (see below) or the old Austro-As. plural prefix ki-, (or maybe
both). Cf. further the prefix kI-/ki- in: kInAza/kInAra 'plough man',
kimIdin, kIkasa, etc., all of which may be compared with the Munda prefix
k- for designation of persons (and the plural prefix ki- of Khasi; note
that in RV, k- also applies to items merely connected with humans and
animals).
	Further RV substrate names of persons, tribes and rivers include
some exactly from the areas where Indus people are to be expected: in their
late/post- Indus new settlement area (J. Shaffer 1995: 139) in the eastern
Panjab, in Haryana (kurukSetra), and especially east of there, well into
the Gangetic plains. Even during the middle/late Vedic period, the local
rivers of E. Panjab are still designated by non-Indo-Aryan names: the
famous bharata chieftain sudAs crosses (RV 3.33) the zutudrI and vipAz and
settles on the sarasvatI. They are not explainable from IA:
	zutudrI (Satlej) < *S@-tu-da�? from Munda *tu 'float, drift',
Kharia thu'da� < *tu-da� (da� 'water'), Khasi p@r-tIu 'outflow'; note the
later popular etymology zatadru 'running with a hundred streams'.
	vipAz < *vipaZ/*vibal (cf. also VibAlI RV 4.30.11-12), and note
that the sarasvatI still has a similar name, vaizambhalyA (with many
variants, always a sign of foreign origin, in the brAhmaNa texts: TB
2.5.8.6, -bhalyA, -pAlyA, -balyA ApZS 4.14.4, -bhalyA BhAr.zikSA; cf. also
RV vizpalA?), which is to be derived from something like *viSambaZ/
*viSambAL, probably with the prefix zam/k'am- (as in zam-bara, kam-boja)
from *(vi)-Sam-bAZ, (note the popular etymology from vi-zambala 'having
widespread blankets'). It is likely that during the Indus period, the
original name of the famous Rgvedic river sarasvatI was something like
*vi(Sam)baL/vi(Sam)baZ. If one insists, indeed, on renaming the Indus
(sindhu, Bur. sende) culture, it should be renamed the Harappan or
sende-vibaZ culture.
	The land of tUrghna (TA), north of this region, has no Indo-Aryan
etymology either, and khANDava (TA), with its suspicious cluster -ND- (K.
Hoffmann 1941), south of KurukSetra, is inhabited by the kIkaTa under their
chieftain pra-maganda. Note also, in the same area (kurukSetra), the
appearance of Pinnow's u-suffixes in 'foreign words', e.g. khANDava,
kArapacava, naitandhava (Pinnow 1953-4).
	The Greater Panjab names of gandhAra, kubhA, krumu, kamboja may be
added. -- gandhAri RV, gAndhAra Br., O.Pers. gandAra, Herodotos ganda'rioi,
EWA I 462, cf. Munda *ga(n)d 'river', the river names of the Gangetic
plains, gaNDakI and gaGgA, the gandhina people on its upper course, and
Nep. -gAD in river names. gandhAra is formed with the common suffix -Ara,
-Ala (Witzel 1993, 1999); --kubhA, cf. Skt. kubja 'bent', Kuiper 1948:
42f., Sant. kubja which belongs to Munda Dui�j, k@b-Duj etc. (Pinnow 1959:
21, 91: $108, 249 $ 286 Kharia Dui�j 'bend', Santali k@bDuj 'ugly',
k@bDuju�d 'crooked', p. 435e Santali k@bn�j 'bent', etc.) -- krumu from
Munda *k@-rum 'luke warm'?? cf. Kharia rum 'to burn', Sant. ur-gum 'luke
warm', Mon uj-ruG 'humid, warm'; -- The kamboja (AV, PS) settled in S.E.
Afghanistan (Kandahar); cf. O.Pers. kambujIya (or kambaujIya?) 'Cambyses';
however, their name is transmitted as  Ambautai by Ptolemy (Geography
6.18.3), without the typical prefix). This change in the first syllable is
typical for Munda names (see below aGga : vaGga, kaliGga : teliGga; kulUTa
: UlUTa, etc.) - Mundas that far west cannot be excluded a piori (Kuiper
1991: 39).

 	It may be asked, how far Austro-Asiatic speakers extended westwards
during and before the RV period. Until now, the present distribution of the
Munda languages has led to rather far-going conclusions, for example by
Burrow (1958, cf. Southworth 1979: 200). Starting from the modern
settlement areas of the Mundas in Eastern India (Bihar, Orissa, W. Bengal)
and on the River Tapti (in northwestern Maharastra and Madhya Pradesh) he
regarded it as impossible that the Munda could ever have settled in the
Panjab. Kuiper, however, has been of a different opinion (1955: 140, 1991:
39, see also 1948: 8, cf. Witzel 1980, 1993 on the substrate in Nepal, and
1999 for the Panjab area). The cases discussed above indicate a strong
(Para-)Austro-Asiatic substrate in the Panjab, and there are some hints
which point to Munda influence in the Himalayas (Konow 1905, Witzel 1993,
see below) and even in E. Afghanistan (zambara, Kamboja).
	An important result is that the language of the Indus people, at
least those in the Panjab, must have been Para-Munda or a western form of
Austro-Asiatic. (Even a minimalistic formulation would have to speak of
some three hundred words from one or more unknown languages, especially one
working with prefixes.)

	In view of the recent comparison by the late I. M. Diakonoff of
Munda and Sumerian (MT III, 54-62, but note the criticism by Bengtson MT
III 72 sq., and cf. still differently, Bomhard, MT III 75 sqq.) this
characterization of the pre-IA Panjab acquires special importance (cf.
already Przyludski 1929: 145-149). To follow up, the role of compound nouns
in Sumerian versus old 'prefixes' in Munda would need further
investigation. In this regard, it should be noted that Sumerian has
implosive consonants, just as Munda, Khasi, Khmer, the Himalayan language
Kanauri and the Kathmandu Valley substrate, all of which may point to a
S./S.E. Asian areal feature.
	If Diakonoff's proposal were indeed borne out, the Rgvedic
Para-Munda substrate in the Panjab of c. 1500 BCE would represent an early
link to Sumerian. Notably, Sumerologists, though without any firm reasons
going beyond some vague mythological allusion to more eastern territories
(Dilmun, etc.), think that the Sumerians immigrated from the east, from the
Indus area. If indeed so, the speakers of (Para-)Austro-Asiatic would have
been builders of a number of great civilizations, from Mesopotamia to
Pakistan/India, Burma and Cambodia.
	If a relationship with Munda could not be confirmed by obvious
etymologies, a minimal position would be to define the c. 300 non-Dravidian
loan words as coming from an unknown, prefixing language of the Greater
Panjab, which might be called, for lack of a self-designation, after its
prominent geographical features, the gandhAra-khANDava or perhaps better,
the kubhA-vipAz, or simply, using the archaelogical term, the Harappan
language.
	Finally, in reviewing the evidence of the Rgvedic Para-Munda, it
should be taken into account that Northern and Southern Munda differ from
each other in many respects, the southern version usually being more
archaic (Zide 1969: 414 sq., 423), though much less known. This difference
as well as the shift of Munda from a prefixing language with mono-syllabic
roots to one working, in typical South Asian fashion, with suffixes, may
have been influenced or even may have been due to a north Indian substrate
such as Masica's "Language X".

$ 1.5. Other Panjab substrates

	If the Indus language is a kind of Para-Munda it cannot, however,
be excluded that one or more unknown languages are involved (cf. Zide and
Zide 1973:15) in the Rgvedic substrate. From the older RV onwards, we find
a number of words that cannot be determined as Para-Munda. Examples include
the words with geminates (see below) e.g. pippala RV 5.54.12 and an
undetermined number of the c. 300 'foreign words.' Some of them can be
traced as being loan words from more distant eastern (Austro-As.) or
western (Near Eastern) languages; the path the loans have taken is clear
(see below) in the case of RV lAGgala <-- Indus *langal <-- PMunda *Jan-kel
<-- Austric (Makassar) naGkala; Ved. vrIhi < Indus *vrijhi <-- PMunda (c.
1500 BCE) *@rig/ Tib./Malay (')bras <-- S.E. As. **@�@rij (?); Ved. mayUra
'peacock' <-- Indus *mayur <-- PMunda mara� 'crier' <-- Austr. (Malay) merak
--> Sino-Tib. *raka 'cock'. Note also the various substrates in Burushaski,
Nahali and "Dhimal" (Kiranti languages in E. Nepal) discussed in MT II, III
and by Kuiper 1962: 14 sqq., 40, 42, 46f, 50f., Berger 1959: 79; and cf.
those of the Kathmandu Valley and Tharu (s. below).
	In short, the Panjab is an area of a Pre-Rgvedic, largely
Para-Munda substrate that apparently overlays a still older local level
which may be identical with Masica's "language X" found in the Gangetic
plains (preserved in some Hindi words). In general, the vocabulary of
Para-Munda and "language X" words is limited to local flora and fauna,
agriculture and artisans, to terms of toilette, clothing and household;
dancing and music are particularly prominent, and there are some items of
religion and beliefs as well (Kuiper 1955, 1991). Since no traces of the
supposedly Dravidian "Trader's Language" of the Indus civilization (Parpola
1994) are visible in the RV, the people who spoke this language must either
have disappeared without a trace (cf. below on MeluHHa) or, more likely,
the language of the Panjab was Para-Munda already during the Indus period
(2600-1900 BCE).
	Or, as expressed by Kuiper (1991: 53) in another context: "Burrow
and Emeneau understandably and rightly ignore the Pan-Indic aspects, but
... their dictionary [DEDR], by omitting all references to Munda, sometimes
inevitably creates a false perspective from a Pan-Indic point of view."
	The large number of agricultural words alone (Kuiper 1955) that
have no Dravidian explanation indicates that the language of the Indus
people cannot have been Dravidian (cf. also Southworth 1988: 663). Their
successors, the Indo-Aryans, preferred to tend their cattle and they spoke,
like their brethren in spirit, the Maasai, about their sedentary
non-Indo-Aryan neighbors in southern KurukSetra in this fashion: "what is
the use of cattle among the KIkaTa?" (ki'm te kRNvanti KI'kaTeSu gA'vaH, RV
3.53.14).

	As we can no longer reckon with Dravidian influence on the early RV
(see immediately below), this means that the language of the pre-Rgvedic
Indus civilization, at least in the Panjab, was of (Para-)Austro-Asiatic
nature. This means that all proposals for a decipherment of the Indus
script must start with the c. 300 (Para-)Austro-Asiatic loan words in the
RV and by comparing other Munda and Austro-Asiatic words. (For the Indus
script see Fairservis 1992: 14, Parpola 1994: 137 sqq., Possehl 1996b). The
decipherment has been tried for the past 35 years or so mainly on the basis
of Dravidian. Yet, few Indus inscriptions have been "read" even after all
these years of concerted, computer-aided attempts, and not yet in a fashion
that can be verified independently (cf. a summary of criticism by Zvelebil
1990). Perhaps that is not even attainable, due to the brief nature of the
inscriptions (7 signs on average and hardly more than 20). Yet, Kuiper's
'300 words' could become the Rosetta stone of the Indus script.
	Further, investigations of the South Asiatic linguistic area
(Sprachbund) must be reformulated accordingly, for example the question of
the retroflex sounds, see Tikkanen 1988, and cf. Zvelebil 1990: 71 on the
distinction between true retroflex sounds (domals, 'cerebrals') and
cacuminals. In the RV they cannot go back either to Proto-Drav. influence,
as usually assumed, because they are already found in the older parts of RV
(books 4,5,6) where no Drav. loans are present; they also cannot go back to
Proto-Munda influences because Munda originally had no retroflexes (Pinnow
1959, except for D, see Zide 1969: 414, 423). The clear increase of the
retroflexes in RV books 1, and especially in 10 is remarkable. In the older
RV one can only detect very few cases of not internally conditioned,
original and clearly non-IA retroflexes: RV 6: kevaTa 'hole'; reNu-kakATa;
rANDya, zANDa, (hiraNya-)piNDa (late hymn), RV 4, 5: krIL-; RV 2: zaNDika,
mArtANDa, pipILe? (pID); cf. also jaTha'ra in RV 1,2,3,5,6,9,10. None of
these old words is Dravidian (see below). In short, the people of the
(northern) Indus civilization must have spoken with retroflexes.
	Almost the same situation exists with regard to another item of
suspected substrate influence, the innovation in Vedic of the grammatical
category of absolutives (not found in Old Iranian!, see below). They occur
in RV 4 with 1, RV 6 with 1, RV 2 with 4 cases (a relatively high number in
this short book!); equally, in RV 3 with only 1, RV 7 with 4, RV 8 (kANva
section) with 0, RV 8 (AGgirasa section) with 2, RV 9 with 4; even RV 1
(kANva section) only with 5. - Really innovating are only the late books RV
1 (AGg.) with 34, and RV 10 with 60 forms.

$ 1.6. Dravidian in the Middle and Late Rgveda
	As has been repeatedly mentioned, there are no traces of Dravidian
language in the Panjab until c. 1500 BCE, not even of the supposedly
Dravidian speaking traders and rulers of the Indus civilization; however,
Drav. loan words suddenly appear in the RV texts of level II (books 3, 7,
8.1-66 and 1.51-191) and of level III (books RV 1.1-50, 8.67-103, 10.1-854;
10.85-191). These include personal and tribal names, as well as cultural
terms.
	For comparisons, we are limited to Burrow-Emeneau's DEDR, and a few
lists from old Tamil texts, but scholars usually work directly with Tamil,
Kannada, Telugu (etc.) comparisons; a reconstruction of Proto-Drav. forms
is but rarely given.
	To begin with, many words that have been regarded as Drav., are now
explained as coming from Munda or another substrate language, for example,
mayUra 'peacock' whose correspondence in Munda *ma-ra� still has an
appellative meaning, 'crier'; (PMunda *ra�k 'to cry,' Pinnow 1959: 76 $57).
However, this is not so for the Drav. designation, where 'peacock feather'
is reconstructed at a level earlier than 'peacock' itself. Indeed, many of
the 26 words attested in the RV that Burrow (1945, 1946, 1947, 1947-48,
1955, cf. Southworth 1979 sqq.) originally listed as Drav., as well as
those added by Southworth (1979) and Zvelebil (1990) cannot be regarded as
early Dravidian loans in Vedic.
	Even if one would regard all of them, for argument's sake, as
Dravidian, only kulAya 'nest' 6.15.16, karambha 'gruel' 6.56.1, 6.57.2,
ukha-(cchid) '(lame) in the hip' 4.19.9 occur in early Rgvedic. These words
can, however, no longer be explained as Dravidian:
	karambha 'gruel' CDIAL 14358, no longer in DEDR; Kuiper 1955: 151
Drav. etym. is 'doubtful', EWA I 310 'unclear'; Kuiper 1991: 51 sqq.
compares loan words with -b- > -bh- (Pkt. karamba 'gruel');
	kulAy-in 'nest-like' 6.15.16, cf. kulAyayat- 7.50.1; from Drav.
CDIAL 3340, cf. DEDR 1884 Tam. kuTai, DEDR 1883 Tel. gUDa 'basket', but the
word formation is unclear; further Drav. *-D- > Ved. -l-?; EWA I 373 'not
clear', comparing N.Pers kunAM, East Baluchi kudhAm < kudAman, with the
same problems; 'foreign word' Kuiper 1991: 14.
 	ukha 'pan, hip' in ukha-chid 'breaking the hip, lame' 4.19.9, cf.
MS 4, p. 4.9 ukhA' (dual) 'hips'; DEDR 564 'particular part of upper leg' :
ukkam 'waist' Tulu okka 'hip'; for sound change Drav. k: Ved. kh, s. Kuiper
1991: 36, cf. 1995: 243; however, EWA I 210 compares Latin auxilla 'small
pot', Latin aulla 'pot' (Pokorny 88), yet declares 'not sufficiently
explained'. As RV 4.19 is not seen as a late hymn, this might be the oldest
Drav. loan in Vedic (RV I).

	Only cases in the middle and late RV remain: In the early RV
(2,4,5,6) possible Drav. words are found only in some additional, late
hymns (insertion after the initial collection of the RV, c. 1200 BCE, s.
Witzel 1995):
#  -phala 4.57.6 'fruit' DEDR 4004, Tam. pal~u 'to ripen', pal~am 'ripe
fruit', etc., see Zvelebil 1990: 78 with literature, Parpola 1994: 168;
CDIAL 9051, 9057; EWA II 201 doubts Drav. origin, and derives it from IA
phal/r 'to coagulate, condense', but finds 'origin of IA *phal/phar not
explained'; that means, a Middle RV loan from Drav. remains possible, or
from Munda: Sant. piTiri 'swelling of glands as in mumps', Sora pEl 'to
swell, grow in bulk (seeds)'; Kuiper 1948, 163, compares Kharia poTki 'to
sprout', potri 'pregnant', etc., cf. 1955: 144, 158, 183; Pinnow 1959:173,
$ 378.
#  phAla 'plough share' 4.57.8, Turner, CDIAL 9072, connects phalati, Iran.
*spAra, and thinks that it has been influenced later on by Drav./Munda; not
in DED(R); EWA compares N.Pers. supAr, Pashto spAra, ISkaSmi uspir <
*spa/Arya?
#  -piNDa 6.47.23 'ball, dumpling'; the many divergent NIA forms speak for
a loan word, see CDIAL 8168 and add.; Drav., Burrow 1946: 23; Munda, Kuiper
1948: 142, 162, cf. 1991: 14; DEDR 4162 Tam. piNTi, Konda piNDi etc.
'flour'? - EWA II 128 'unexplained'; cf. also K. Hoffmann, Diss. 1941: 380
sqq. and perhaps Armenian pind 'compact, firm' < Iran. (< Ved.?)
	In middle RV (3,7,8):
#  kuNAru 3.30.8 'lame in the arm?', or name of a person, see EWA I 362
'unclear'; however, compare Drav.: Kan. kuNTa 'cripple', Mal. kuNTan
'cripple', etc., CDIAL 3259-60, DEDR 1688
#  mayUra 3.45.1 DEDR 4642, 'peacock' PS, mayUrI RV 1.191.14, mayUra-roman
RV 3.45.1, mayUra-zepya RV 8.1.25; generally regarded as Drav.: DEDR 4642
Tam. maJJai, mayil; northern Kasaba dialect of Irula muyiru, Tulu mairu,
Konda mrIlu, miril, (*mayil/mayir, see Zvelebil 1990: 77, with discussion
and lit.). However, originally from Munda: PMunda *mara� 'crier', Kharia
mara�, Santali, Mundari, Ho mara�, Kurku mara, Sora mArAn 'peacock, Pavo
cristatus', see Pinnow 1959: 205 $90; cf. also Skt. marUka (lex.) 'peacock,
deer, frog, Curcuma Zerumbet', and Khotanese Saka murAsa 'peacock' (EWA II
317, KEWA II 587, CDIAL 9865, add. 9865, DEDR 4642, Bagchi 1929: 131,
Southworth 1979: 191 sqq., 200, cf. Zvelebil 1990: 77, Hock 1975: 86). The
rare tribal name mara-Ta PS 5.2.1, 12.2.1 (Witzel 1999) belongs here; the
maraTa probably lived south of the Ganges and north of the Vindhya.
	The above may indicate that the Dravida entered into contact with
some groups of Munda speakers fairly early (before the Middle RV); however,
just as in the Vedic case, one or two intervening language(s) (*mayil /
*mayur) must delivered the word to Drav. and Vedic, for example the
"Language X" or a Northern and Southern Indus language; in the south, this
must have occurred before Sindh was practically deserted in the post-Indus
phase (Allchin 1995: 31 sqq.). The Ved. form mayUra may have been
influenced by mAyu 'bleating'.
#  phala 3.45.4 see above
#  kANa 7.50.1 'one-eyed' EWA I 336 'unclear'; cf. Avest. kar@na 'deaf' :
kar@na 'ear' and cf. DEDR 1159 Tam. kaN 'eye' and DEDR 1443 kAN 'to see',
both now without reference to Skt.; Zvelebil 1990: 79 compares DEDR 1159
and finds, 'rather speculative', the Drav. negative suffix -a/-A; cf.
Kuiper 1991: 79. --However, cf. Burushaski zon, zOn 'blind' (see above,
with northwestern interchange of Ved. z/k, Witzel 1999); note also that
kANa is found as hapax RV 10.155.1 next to 'mountain', a 'foreign' name and
an onomatopoetic: giriM gaccha, zirimbiTha, budbud- (cf. Santali buDu�c
buDu�c 'to bubble up').
#  kulpha 7.50.2 'ankle', CDIAL 4216, from Drav.; cf. DEDR 1829 kuLampu
'hoof'?; EWA I 376 'completely unclear', Kuiper 1955:148 loan word because
of AV gulpha and points (1991: 35) to variant forms in Ved. (gulpha) and
MIA (gopphaka, guppha, goMpha).
#  daNDa 7.33.6 (late hymn) 'stick', DEDR 3048 Mal. taNTa 'forearm, arm',
Tel. daNDa, etc., cf. DEDR 3051, CDIAL 6128; Munda, Kuiper 1948: 76: Sant.
DaNTa 'thick stick, club', Da(N)TiTit 'stem (of mushrooms)', DaNDi 'stick,
staff, stalk', cf. Mundari DANdi 'small stick'; EWA I 691 'not explained'
#  kuNDa- 'vessel' 8.17.13 can be compared with Avest. kunda/-I, kundiZA,
the name of demons ('pot-bellied'); Dravid., DEDR 1669 Tam. kuTTam
'deepness, pond', Tel. kuNTa, kuNDu, Kur. xoNDxA etc., DEDR 2082; Kuiper
1948: 76 Drav., 1991:14 'foreign'; CDIAL 3265; EWA I 363 points to the
difference in meaning between Drav. and Ved. and concludes 'unclear,
perhaps loan word'
#  mayUra 8.1.25, see above
#  naLa 8.1.33 'reed', naDa/nala/nada, EWA II 7 from IIr. *nada (Nuristani
nO < *nada, Parth. nad 'flute', N.Pers. nAy 'flute') < IE *nedo (Hitt. nata
'reed', Armenian net), however without actual explanation of the variation
*d > D; DEDR 3610 compares, strangely, Tam. nal 'good' with the Skt. name
Nala, idem Zvelebil 1990: 82; however, Nala is found in Vedic, ZB 2.3.2.1-2
as NaDa NaiSidha, and in Mbh. as Nala NaiSadha, the king of the (probable)
Munda tribe of the NiSidha/NiSadha = Ved. NiSAda (MS, VS, see below); cf.
Kuiper 1991: 33 on D/d, and p. 19 nALI 10.135.7 'flute, pipe' (cf. 1948:
82).
#  kANuka 8.77.4; (poet: kurusuti kANva) next to saras 'pond'; unclear in
meaning and etym., EWA I 336; Kuiper 1991 as foreign.
	In late RV (1, 10):
#  ulUkhala 1.28 'mortar' DEDR 672 Tam. ulukkai, Kan. olake, KoDagu oLake,
and Kota oLka, oLkal kal '(stone) mortar', Malto loRa 'stone to grind
spices' (S. Palaniappan, by letter); EWA I 231 'problematic'; cf. Zvelebil
1990: 79 with lit., Kuiper 1991: 14, 41 'still unexplained', compares loan
words with prefix u-; any connection with khala 'threshing floor' RV
10.48.7?
#  vriz 1.144.5 'finger', DEDR 5409 Tam. viral, Go. wirinj, now without
reference to Skt. vriz; EWA II 597 from IA *vrez 'to bend', Avest. uruuvaEs
'to bend, curve'
#  bila 1.11.5, 1.32.11 'hole, cave' CDIAL 9245 'Dravid.'; DED 4459 = DEDR
5432 now without reference to Skt., cf. also DEDR 4194; Kuiper 1991:14
'foreign', EWA II 225 'not clear'
#  a-phalA 10.71.5 'without fruit', see above;
#  phal-inI 10.97.15 'having fruits', see above;
#  mayUra 1.191.14, see above;
#  piNDa 1.162.19, see above
#  kUTa 10.102.4 'hammer' DEDR 1651, 1655, 1883, app. 29; previously
explained by Burrow as Drav., later explained by him as IE (German hau-en),
but see EWA I 384 'unclear'
#  phAla 10.117.7 'plough share', see above
#  phala 10.146.5 'fruit', see above
#  kANa 10.155.1, see above
#  kaTu(ka) 10.85.34 'pungent'; CDIAL compares khaTTa 'pungent'; EWA I 290
Lithuanian kartu`s 'bitter'? or DEDR 1135 Tam. kaTu 'to pain; pungent;
cruel, harsh, bitterness', Kurukh xaRxa 'bitter', Malto qaRqe 'bitter',
Brahui xarEn 'bitter' etc.
	Finally, bala RV 1,3,5,6,7,9,10 'strength, force'; EWA compares
Latin de-bilis etc., IE *belo-, which is otherwise not found in IIr.
(perhaps in Osset./Sarmatian); see, however, Kuiper 1990: 90 on the rare IE
(initial) b-, and on the impossibility of an IE etymology; cf. CDIAL 9161;
now, against Drav. origin Burrow, see EWA II 215; cf., nevertheless, DEDR
5276 Tam. val 'strong', Kurukh balE 'with the help of', Brahui balun 'big'.

	The same is the case with some words that have later on been added
and discussed (Sanskrit Index of the DEDR, p. 559-763) and elsewhere.
Most of them are too late in attestation to be of interest here.
In DEDR we find: 
	Early RV: phalgu 'minute weak' 4.5.14, kalaza 'vessel' 4.27.5,
6.69.2, 3.32.15, 7.69.6; and later: taDit 'flash' 2.23.9 (late addition),
1.94.7 phAla 'plough share' 4.57.8 (late); -- middle RV: ukhA 3.53 'pan,
hip' (late addition), kavaSa 'straddle legged', a personal name 7.18.12,
kUla 'slope, bank' 8.47.11. -- late RV: ukhA 'pan, hip' 1.162.13,15; khala
'treshing floor' 10.48.7. Of these, only phalgu 'minute weak' (RV 4)
remains as a possible early loan into IA, if it indeed belongs to DEDR
4562, Tam. pollu 'empty husk of grain'. Again, all other words regarded as
Dravidian appear only in the middle and especially in the in later RV.
	Southworth (1990, 1995) adds the following examples of early
contact between Drav. and Indo-Ar., however, without ordering the texts
historically.
#  car-, carati RV : Tamil cel 'to go, flow, pass, be suitable' (already
PerunkuN2r2Ur KilAR2, c. 160-200 CE (Zvelebil); DEDR 2781 "probably from
IA", CDIAL 4715; the word is IA, derived without problems from IE *kwel(h);
perhaps accidental agreement with Drav. cel.
#  mAyA 'confusion, wonderment, awe' RV (found in all of RV, just as
mAy-in, mayA-vat, mayA-vin), = Avest. mAiiA 'awful power' :: Tam. maya-
'mistake, misunderstand'; mayakku- 'bewilder, confuse, intoxicate, alcohol'
etc.; DEDR 4706 without comparison with Skt.; the Skt. and Drav. meanings
do not agree; also, because attested that early in the RV and Iran., Drav.
origin (only Middle-RV Drav. influence!) is unlikely -- unless it would
have taken place in Iran (Southworth 1979: 196f.: "high degree of contact
... at the earliest period for which we have records and possibly before");
however, see below, on tanU.
	Southworth 1979: 203, 228 f., 1990: 222-3, 1995 reconstructs as
further indication of early contact between Drav. and Indo-Ar. in Iran, a
word *tanu 'self', Tamil tAN2/taN2 'oneself', tanU RV 'body, self/oneself',
for this meaning see now J. R. Gardner, PhD diss., U. of Iowa 1998. The
variation in vowel length in the Drav. pronoun (Tam. tAN2/taN2 'oneself')
is old (Krishnamurti 1968). However, next to the RV instances, there is
Avest. tanU 'body, self', O.Pers. tanU 'body', all have no clear IE
etymology. Pokorny 1959: 1065, 1069 derives them from IE *ten 'to stretch',
in other IE languages the meaning mostly is 'thin'; EWA II 622 connects
tan-U '*Ausdehnung, ausgespannte Hu"lle' with tan.
	The comparison of the IIr. and Drav. words would presuppose a very
close relationship between Drav. and (pre-)Indo-Ar. tribes indeed, as
pronouns are not taken over easily. Such early Drav.-IA relationships are
not found otherwise: there are no early loans in designations of material
culture, e.g. pastoralist terms in Vedic/Drav.: horse: azva : ivuli,
kutira, cow: gau- : A(N2), sheep: avi : (y)ATu, koR2i, goat : aja : (y)ATu,
koR2i, dog: zvan : nAy, nAi. This would rather point against a neighborly
relationship of both languages in any pre-South Asian context
#  garda-bha 'donkey' RV, late, only 1.23.5, in the appendix hymn 3.53.23
next to rAsa-bha 'donkey'!, RV vAlakhilya 8.56.3 :: Tam. kal~utai, Gondi
gARdi, etc., to which DEDR 1364 compares Skt. gardabha; CDIAL 4054; EWA I
473 cf. gard 'to cry shout', not from Drav.
#  pizAca, pizAcI AV, pizAci- 'demon' RV, late: 1.133.5 :: Tam. pEy-
'devil, goblin, madness' DEDR 4468, without comparison with Skt., and
without suffixing -zAci-, only: pEytti, pEycci, pEcci 'demoness'.
#  zava (not in RV, diff. Southworth 1979: 197), only PS+ : Tam. cA- 'to
die' (Kural), Ko. ca-v- 'corpse' DEDR 2426 compares Skt. zava; EWA II
derives zava from zav 'to swell' AVP; CDIAL 12356 not from Drav. As the
word is early in Drav., perhaps accidental look-alike.
#  paThati 'to recite' RVKh., TA, Up. : Tam. pATu 'sing, chant', pATTu
'song', attested already in PerunkuN2R2Ur KilAR2, DEDR 4065 without
reference to IA; EWA II 69; CDIAL 7712 < *pRthati; Drav. <-- Indo-Ar.,
Burrow-Emeneau 1962: 46, no. 242. Rather to be derived from MIA pupil's
slang Ved. prath 'to spread out (a text, in recitation)'?; compare the
frequent loan words in the context of Vedic teaching and learning: maNDala,
kaNDa, kANda, prapAThaka, paTala, daNDa, MIA: orimikA 'a section of KS' etc.
#  nagara 'town' TA, but cf. already nagar-in JB :: Tam. nakar 'house
abode, town, city'; cf. EWA II 5, CDIAL 6924; DEDR 3568 IA --> Tam. nakar
'house, town, etc.' But why nakar from Skt.? There is no IA etymon, nor is
there one in Drav. and Munda. Drav. for settlements: DEDR 3568 nakar
'house, town', 1655 kuTi 'home', 3868 paTTi 'cow stall, village', 5393
viTu(ti) 'temporal residence', 2007 cEri 'street, village', 752 Ur
'village', 4362 pUNTi 'town, village', 4047 pAkkam 'seaside village', 4646
maTappam 'agricultural town', 807 eyil 'fortress'; 4064 pATi 'town', 4112
pAl~i 'temple, town', 4555 Kan. pol~al 'town', 5549 vai, 3911 pati, 2814
cEr; 3638 nATu 'open country' (opp. nakaram); -- cf. also Skt. haTTa
'market' ~ Santali, Mundari, Ho hatu, Korwa watu < PMunda *watu Pinnow
1959: 79 $ 69.-- In short, the word may be a loan from the southern Indus
language or one from the Malwa area.
	Thus, the words added by Southworth are post-Rgvedic (zava,
paThati, nagara), or they are attested in relatively late RV sections
(gardabha, pizAci), or they are of dubious nature (car, mAyA, tanU).
Therefore, it is not possible to suppose, with Southworth, an early close
contact, even in Iran, and on all levels of society, of Dravidas and
Indo-Aryans. Rather, one has to agree with Kuiper, who stresses the very
hesitant acceptance of non-Indo-Aryan words and forms in the high level,
poetic language of the RV. The words collected by Southworth in his second
list (not discussed here) can have been taken over into Drav. at any time
after the RV, e.g. accu 'axle' < akSa RV.
	Furthermore, most of the c. 800 words in the list provided by DEDR,
p. 759-764 are attested only in the Epics or in class. Skt. Of the c. 61
words listed in the appendix of DEDR which are supposed to come from
Indo-Aryan, only a few can be regarded as (possible) early loans; they all
should be checked in early Tamil before something that even approaches a
final decision can be made.
	Finally, among the words in Zvelebil's recent list (1990: 77-82) of
22 "early" Drav. loans into Skt., most have already been discussed above;
yet, none of them nor the ones newly mentioned are Rgvedic: 8. bilva 'Aegle
marmelos, Bel tree' AV, 10. kuNapa 'corpse' AV, 11. kurkura 'dog' AV, 12.
arka 'Calatropis gigantea', ZB, 12a. candana 'sandal wood, paste' Nirukta,
13. kavaca 'armor' PS, ZB, kavacin AV, 13a. jaTA 'matted hair' GS, 13b.
mAlA 'flower necklace', GS, mAlya RVKh, 13c. eDa 'sheep' KZS, eDaka JB,
aiDaka ZB. The rest of the words are only post-Vedic.
	Zvelebil's summary is: "as Emeneau (1971) writes, 'We end, then
with a small, but precious handful of Vedic forms for which Dr. etymologies
are certain and acceptable as may be expected in this field of areal
linguistics, adding, though that no chronology of the borrowings is
possible" (Zvelebil 1990: 81; similarly Parpola 1994: 168.) According to
what has been said above, this has to be modified drastically: Rgvedic
loans from Drav. are visible, but they also are now datable only to middle
and late Rgvedic (in the Greater Panjab), and they can both the localized
and dated for the Post-Rgvedic texts (Witzel 1987, 1989).
	Of all the words mentioned so far that have been regarded as Drav.,
only the following few are possible for the early RV :
 ukha[-chid] 'hip[-breaking]' 4.19.9; phalgu `minute weak' 4.5.14, 
 ANi 'lynch pin' 5.43.8, (whose ultimate source is unclear, and, 
 very tentatively, bala 'force' 5.57.6, 5.30.9, probably from IE, 
 cf. Latin de-bilis).
Whether this is enough to ensure the presence of (even a small number of )
speakers of Dravidian in the Panjab during early RV times may remain in the
balance. From the middle RV come: kavaSa 'straddle legged', (a personal
name) 7.18.12, kUla 'slope, bank' 8.47.11 and perhaps also kuNDa 'vessel'
8.17.13.
	Consequently, if more of the middle and late RV words mentioned
above are accepted as Drav. and even if some of the words excluded above
for the early RV should be accepted, this would not change the general
picture: There is very little Dravidian, but there are about 300 words of
the Indus substrate. For it cannot be said, conversely, that there were,
during the older and middle RV, clear indications (or: "a precious
handful", Zvelebil, Emeneau) of a strong Drav. substrate in the Panjab. At best, one
can speak of a few very isolated cases which have been taken over into the
RV; clearly this indicates an adstrate rather than a substrate.
	This result is important for the time of the immigration of
speakers of Dravidian into the Panjab and it specifically underlines that
the Indo-Aryans did not at once get into contact with speakers of Drav. but
only much later, when the tribes speaking IA were already living in the
Panjab and on the sarasvatI and yamunA. Apparently, Dravidian speakers
began influencing the Panjab only at this moment in time (cf. Allchin 1995:
31 sqq., see above). Consequently, all linguistic and cultural
deliberations based on the early presence of the Drav. in the area of
speakers of IA, are void or they have to be reinvestigated.
	It cannot be argued that the immigration of the Dravidians into the
Panjab should have taken place earlier than discussed above, for the simple
reason that Drav. words do not exist in that early period; the same is the
case if only the upper class such as traders (cf. vaNij 'trader?' RV
1.112.11, 5.45.6, AV, (pra-)vANa 'trade?' 4.24.9, see Kuiper 1955: 168) and
administrators of the Indus Civilization was composed of Dravidian speakers
(Parpola 1994, Fairservis in: Southworth, 1979: 208, 228; contra, Hock
1975: 87f., cf. Southworth 1992: 663), and that in consequence, the Indus
inscriptions should be read as Dravidian. In this case, one would expect,
after some 400-700 years of the flourishing of the Indus civilization,
cases of bilingualism. Conseqently, much more Drav. influence should have
been retained than visible in the few (late) words found in the c. 380
'foreign' words. One would expect at least a few important loan words from
the fields of trade, handicraft or state organization (at least, from the
post-Indus, village level type cultures). This, again, is not the case.
PaNi '(rich) foreigner, demon' cannot be connected with 'trader' inside the
RV, and paN 'to barter' appears first only in (post-Rgvedic) KS, pra-paNa
'trade' AV, prati-paNa 'exchange' (see EWA II 69; DEDR 3884 does not help:
paN 'work, service', paNikkaN2 'carpenter'; cf. Kuiper 1955: 168, on vANa,
vaNij.) In addition, there are not many designations of RV artisans, except
for IA takSan 'carpenter', etc. (see below). Even if Drav. had been the
traders' language, one would be at loss to answer the question why Drav.
influence is only seen in the middle and late RV as well as later one
(AV+).
	Summing up, early Dravidian influence in the Panjab can be
excluded, but must be explained for the following middle and later RV
periods. This is best done by the scenario mentioned above: middle and
later RV immigration of Drav. speakers from Sindh. Incidentally, it must be
noted that in all of the RV, there are no typical Drav. words for
agriculture which should be expected if the Indus people of the Panjab had
been speakers of Dravidian. This agrees with the reconstruction of
Fairservis (1995), Southworth (1979, 1988, 1990: 663, and McAlpin (1979) of
early Dravidian: an originally pastoral society that acquired agriculture
only in South Asia. All of this indicates that we have to take a closer
look at the regions bordering the Panjab in the South, especially Sindh.


===========================================================================
Michael Witzel                          Elect. Journ. of Vedic Studies
Harvard University                  www1.shore.net/~india/ejvs
---------------------------------------------------------------------------
my direct line (also for messages) :  617- 496 2990
home page:     www.fas.harvard.edu/~witzel/mwpage.htm