$ 2.4. Substrates of the Lower Gangetic Plains and "Language X".
	Next to the Mundas, there must have been speakers of other
languages, such as Tibeto-Burmese, who have left us names such as kosala,
kauzikI (mod. kosi), perhaps also kAzi and kauzAmbi (mod. kosam), from
Himalayan khu, ku (Witzel 1993). In IA they also have left such words as
the designations for cooked rice IA *cAmala and probably also PS zAli
'rice'.
	In Uttar Pradesh and North Bihar (attested in Middle and Late Vedic
texts, c. 1200-500 BCE) another apparent substrate appears in which the
'foreign' words do not have the typical Para-Munda structure, with the
common prefixes, as described above. Masica (1969) called this unknown
substrate "language X". He had traced it in agricultural terms in Hindi
that could not be identified as IA, Dravidian or Munda (or as late loans
from Persian, S.E. Asia, etc.). Surprisingly some 30% of the terms are of
unknown, language "X" origin, and only 9.5% of the terms are from Drav.,
something that does not point to the identity of the Indus people with a
Drav. speaking population.
	However, only 5.7% of these terms are directly derived from Munda.
Obviously, the pre-IA population of the Gangetic plains had an extensive
agricultural vocabulary that was taken over into all subsequent languages.
F.B.J. Kuiper has pointed out already in 1955: 137-9 (again in 1991: 1)
that many agricultural terms in the RV neither stem from Drav. nor from
Munda but from "an unknown third language" (cf. Zide & Zide 1973: 15). This
stratum should be below that of Para-Munda which is the active language in
the middle and late Vedic texts.
	Again, it has been Kuiper who has pointed the way when he noted
that certain 'foreign' words in the Vedic substrate appear with geminate
consonants and that these are replaced in 'proper' Vedic by two dissimilar
consonants (1991: 67). Examples include: pippala RV (1.164.20,22; 5.54.12,
su- 7.101.5 ) : piSpala AV (in Mss.) 9.9.20,21; 6.109.1,2; su-piSpala MS
1.2.2:11.7, guggulu AV, PS : gulgulu KS, TS, kakkaTa PS 20.51.6, KSAzv. :
katkaTa TS. Kuiper adds many other cases of Vedic words that can be
explained on the basis of words attested later on.
	In RV geminates also occur in 'onomatopoetic' words: akhkhalI-kR
'to speak haltingly' or 'in syllables?', cf. now Nahali akkal-(kAyni) '(to
cry) loudly in anguish' MT II 17, L 33 (kAyni < Skt. kathayati 'to tell'
CDIAL 2703, cf. 38) MT II 17; cf. also jaJjan- RV 8.43.8 etc., ciccika
10.146.2 'a bird'?, and cf. also azvattha 1.135.8 : azvatha a personal
name, a tree, 6.47.24, with unclear etymology, (Kuiper 1991: 61, 68).
	Post-RV, new are: hikkA PS 4.21.2, kakkaTa PS 20.51.6 (MS kakuTha,
TS katkaTa!), KSAzv in YV: kikkiTA KS, TS, kukkuTa VS, pilippilA TS
7.4.18.1, cf. also TS Akkhidant, prakkhidant TS 4.5.9.2, Ajjya 5.2.7.3.
	Especially interesting is the early gemination *dr > ll: kSullaka
AV 2.32.5, TS 2.3.9.3 kSullaka, < kSudra 'small' (a children's word?);
later on, among others, bhalla-akSa ChU4.1.2, bhalla Br., MBh (with variants
phala, phalla! EWA s.v.); JB malla 'a tribe' (in the Indian desert,
Rajasthan; cf. DEDR 4730), etc.
	Though certain geminates, especially in word formation and flexion
(-tt-, -dd-, -nn- etc.), are allowed and common, they hardly ever appear in
the stem of a word (Sandhi cases such as anna, sanna etc. of course
excepted). Until the late BrAhmaNa texts, other geminates, especially bb,
dd, gg, jj, mm, ll, but also kk, pp, etc., are studiously avoided, except
in the few loan words mentioned above (pippala, gulgulu, katkaTa etc.
(Kuiper 1991: 67 sqq.).
	It will be readily seen that Kuiper's seminal observation reflects
a tendency that can be observed throughout the Vedic texts. Geminates,
especially the mediae, apparently were regarded, with the exception of a
few inherited forms such as majj 'to dive under', as 'foreign' or
'barbaric'. They did not agree with the contemporary Vedic (and even my
own) feeling of correct speech (Sprachgefu"hl).
	However, starting with Epic Sanskrit, forms such as galla, malla,
palla, etc. are normal and very common (however, -mm-, perhaps regarded as
Drav.(?) remains rare); such words, in part derive from normal MIA
developments, in part from the substrate.
	This tendency can be sustained by materials from various other
sources. In the language 'X' only a few of Masica's agricultural substrate
words that do not have a clear etymology (1969: 135) contain such
geminates: Hindi kaith < Skt. kapittha CDIAL 2749 (Mbh), piplI/pIplA <
pippala (RV), roTI < *roTTA, roTika 10837 (Bhpr.); karela <
karella/karavella 3061, khAl < khalla 3838-9 (Suzr.); to these one can add
the unattested, reconstructed OIA forms (Turner, CDIAL, see Masica 1969:
136): *alla CDIAL 725, *uDidda 1693, *carassa 4688, *chAcchi 5012, *bAjjara
(see, however, OIA *bAjara, 9201 bAjjara HZS: varjarI!), *balilla 9175,
*maTTara 9724, *suppAra 13482, *sUjji/sOjji 13552. However, these words
have come into NIA via MIA, and that their geminates may go back to a
consonant cluster without geminates (see below, on Turner's reconstructs).
	All of these tendencies are reconfirmed by what we can discern in
the other substrate languages. While there still are but a few cases in the
northwest, the substrates located further east and south all have such
geminates. (Incidentally, the northwest has retained the original, non-
geminate consonant groups, such as -Cr-, to this day, cf. Khowar bhrar,
Balkan Gipsy phral 'brother', W. Panj. bhrA, E. Panj. bh(a)rA : Hindi bhAI,
etc.).
	In the unstudied substrate of the Kathmandu Valley (inscriptions,
467-750 CE, see below), geminates are found in the following place names:
gamme, gullataMga, gollaM, jajje-, dommAna, daGkhuTTA-, bemmA, cf. also
bhumbhukkikA (onomat. with double consonant: < *bhumbhum-ki-kA?); cf. also
village names such as joJjon-diG, tuJ-catcatu, thuMtuM-rI, daNDaG-(guM).
	In the substrate of modern Tharu: e.g. ge~TTI, ghaTTI, TippA (?),
ubbA; cf. also 'onomatopoetic' words such as jhemjhemiyA 'small cymbal or
drum', bhubhui 'white scurf', gula-gula 'mild' (with the usual middle
Vedic, OIA, Tamil, etc. form of the "expressive" and onomatopoetic words:
type kara-kara versus older Vedic bal-bal).
	In modern Nahali (Kuiper 1962: 58 sqq., 1966) the following
substrate words can be found, though apparently various types of consonant
groups are allowed: bekki, beTTo, bokko, coggom, cuTTi, joppo/jappo, kaggo,
kAllen, maikko, oTTi, poyye, unni. Additions to this list can easily be
supplied now from that of A. Mundlay (MT II) which are not obviously from
NIA include 8 aDDo, 91 attu', 182 bekki, 203 beTTo, 221 bijjok, 232
biTThAwi, 255 buddi, etc.
	In the Drav. Nilgiri languages (Zvelebil 1990:63-72) there are a
few isolated geminating words that go back to a pre-Drav. substrate, e.g.
Irula mattu 'lip', Dekkada 'panther', muTT(u)ri 'butterfly', vutta
'crossbar in a house'.
	The Vedda substrate contains the same type of words:: cappi 'bird',
potti 'a kind of bee', panni 'worm' (de Silva 1972: 16).
	Finally by way of appendix, in the isolated Andamanese language
(Aka BIada dialect), a few consonant groups seem to be allowed, but hardly
any geminates are found (Portman 1887): dAkkar-da 'bucket' p.18, kAttada,
badda 'crab' 22, chetta-da 'fruit' 34, tokko dElE kE 'to go along the
coast', chetta-da 'head' 36, sissnga kE 'to hiss' 38, udda 'maimed' 48,
peggi 'many' 48, teggi lik dainga 'noise' 52, teggi lik dainga kE 'to obey'
54, molla-da 'smoke' 72, tekke yAbadO 'straight' 78.

	It can be stated, therefore, that the substrate languages outside
of the extreme northwest indicate broad evidence for original geminates.
Differently from IA (cf. below, on Turner's reconstructions), these words
have not been pushed through the 'filter' of MIA, that means their original
consonants clusters have not been 'simplified' (e.g. kt > tt, kS > kkh,
etc.) Such striving for simpler syllable structure is known from many
languages, e.g. Latin noctem > Italian notte, French nuit [nu"i], or O.Tib.
bgryad > Tib. [yƒ] 'eight', Jpn.-Austro-Thai *krumay > Jpn. kome 'rice'
(Benedict), Kathmandu Valley substrate kicipriciG(-grAma) > Newari
kisipi~Di, etc. Even then, the tendency seems especially strong in S. Asia
and probably has worked on IA from the beginning, as for example in the
early example AV kSullaka < kSudraka. In Drav. various consonant groups are
allowed, including geminates (Zvelebil 1990: 10 sqq.:) e.g., kakku, kaccu,
kaTTu, kattu, kappu, kammu; (cf. also the interchange p- :: -pp-/-v- ::
-p/-u).
	One can therefore put the question whether this old substrate
tendency has already influenced the Para-Munda of the RV. In Munda itself,
such geminates are very rare (cf. Kuiper 1991: 53), and open syllables are
common. However, there is a tendency in the Munda languages to eliminate
consonant groups caused by vowel loss in prefixes (Pinnow 1959: 457); this
does not cause geminates in such cases but is in line with the similar
developments from Old to Middle and New IA (e.g. akSi 'eye' > akkhi > A~kh,
rakta 'colored, red' > ratta > rAt, etc.). One may therefore explain many
of the 'foreign' words with geminates in Vedic and post-Vedic, excluding
Drav. loans, in the same way.
	For the same area that is covered by Masica's language "X", and for N.
India in general, one may also adduce the many words in NIA that are not
attested in Vedic, Classical Skt. or the various MIA languages such as Pali
but that occur only in their NIA form. They have been collected and
reconstructed by V. Turner in his CDIAL. These include the starred forms,
appearing in their reconstructed OIA form, and those words that do not
appear in Ved. but are more or less accidentally attested in late Skt.
texts, and the substrate words dealt with by Turner. They have a typical,
often non-IA structure, including the very common cluster -ND-, -TT-. Their
root structure follows the following pattern. (C = any consonant, @ any
vowel)
*C@kkh, C@g, C@gg, C@cc, C@cch, C@jj, C@Jc, C@T, C@TT, C@NTh, C@D, C@DD,
C@Dg, C@ND, C@dd, C@n, C@pp, C@mp, C@bb, C@mm, C@r, C@rC, C@l, C@ll, C@v,
C@s, C@zz, C@h.
In Turner's CDIAL there are only a few forms such as *Cr@k, Cr@c, Cr@NT,
Cr@ll, Cl@kk; this does not surprise as all reconstructed words have passed
through the filter of MIA and have lost such clusters, -- except in the
extreme northwest (Lahnda and Dardic).
	Double consonants at the end of roots may go back to complicated
clusters that can no longer be reconstructed, for example *C@kkh < **C@kS
(cf. RV kSviGkA, ikSvAku, and compare Ved. clusters such as matkuNa,
matkOTaka, kruJc). Consonant clusters with various realizations in
pronunciation may also be hidden in many Vedic loan words (Kuiper 1991 : 51
sqq., Ved. cases p. 67 sqq.).

$ 2.5. Tibeto-Burmese
	Still, this is not all as far as the Gangetic plains are concerned.
The eastern section of the North Indian plains (E. Uttar Pradesh and N.
Bihar) provides some indications of Tib.-Burm. settlements. The name of the
Avadh (Oudh) area north of Benares in late Vedic texts is kosala; this form
should not appear in Vedic/Skt.; it should have been *koSala or *kozala (as
is indeed found in the Epics). The word clearly is foreign, and should
belong, together with the slightly more eastern river name kauzikI
(post-Vedic, mod. kosi) to a Tib.-Burmese (TB) language. Such designations
for 'river' are indeed found in eastern Himalayish: R. kosi, many Rai river
names in -ku, -gu, in medieval Newari (kho, khu, khwa; ko 'river' in the
unpublished Newari amarakoza) and modern Newari (khu, khusi 'streamlet,
creak') in and near the Kathmandu Valley, where it is already found in
Licchavi time inscriptions, 467-750 CE, as: cUllaM-khu, theG-khu,
japti-khU, huDi-khU, pi-khu-, vihliM-kho-srota, ripziM-ko-setu. It is
perhaps derived from TB *kluG (details in Witzel 1993).
	Perhaps one may add the name of the tribe around Benares (kAzI)
whose older, Vedic form is kAzi (AV, still regarded as outsiders to whom
one sends one's fever, PS 12.1-2), and its western neighbor, the kUzAmba,
kauzAmbi (the later town kauzAmbI, mod. village of kosam near Allahabad).
R. Shafer (1954) has a host of names, taken from the list of peoples in the
much later mahAbhArata Epic that must be taken with caution (redaction only
c. 500 CE, where even the Huns are included with hUNa, harahUNa, - they
have become a Rajput clan!)
	Indeed, early evidence for mountain tribes which might have been
Tib.-Burm. is found in the Vedic texts all along the Himalayas. These
mountain tribes, probably of Himachal Pradesh and Western Nepal, lived on
the border of the Vedic settlement. They are first encountered in AV (1200
BCE) under the names kirAta, in the western Himalayas where they appear as
herb collecting mountain girls (kairatikA kumarikA PS 16.16.4, ZS 10.4.14.,
kailAta PS 8.2.5). The more eastern text VS 30.16 has them as living in
caves; cf. also the popular form kilAta PB, JB, ZB; (for details see Witzel
1993, 1999, and cf. KEWA I 211, EWA I 352, and also EWA I 311, s.v. KAR,
and Prakrit cilada).
	An alternate form of the name, kIra, may have been retained in
Kashmir, attested in 550/600 CE (bRhatsaMhitA 14.29). Its name is close to
that of the kirAta who are attested in the early inscriptions of Nepal (467
CE sqq.). Hsuan Ts'ang, Hsiyuki (c. 600 CE, cf. T. Funayama 1994: 369),
however, knows of them as kilito (Karlgren 1923, no. 329-527-1006), a
people in Kashmir who had their own king shortly before his time. The -ta/
-Ta suffix is common in many North Indian tribal names (Witzel 1999, cf.
above).
	Since the RV, tribal names are found have the suffix -ta/-Ta
(Witzel 1999), e.g. kIkaTa, bekanATa (certainly a non-IA name: b-, -T-),
maraTa PS 5.21.3, 12.2.1, kirAta AV, PS, AraT(T)a/arATTA BZS (cf. Sumer.
aratta, an Eastern country, Sistan), kulUTa, kulUta (MBh), kulU-ta(ka),
(but also: kolUta, kaulUta, kuluTa, and even ulUTa, ulUta, see Kuiper 1991:
38 (cf. Pinnow 1959: 198f., cf. S. Le'vy, JA 203, 1923, 52 sqq. = Bagchi
1929: 119 sqq.), finally luLu in W. Pahari, CDIAL 3348, with the typical
prefix change of Munda; virATa, a king of the Matsya (Mbh) and a country in
bRhatsaMhitA, Pkt. virADa, mod. Berar.
	However, names in -ta (and -nda) are restricted to the Himalayan
mountains while those with -Ta (and -NDa) occur all over the northern
Indian plains (Witzel 1999). As for the origin of the suffix -Ta, compare
the plural suffix -To in Nahali (Berger 1959, Mundlay MT II, 1996, 5, cf.
Kuiper, 1991: 45 on 'Dravidian' -Ta).
	Beyond this, the early texts do not allow us to decide on the
language and appearance of the kirAta. (The Epic calls them gold-colored).
However, MS and ZB list them with the Asura ('demons') kilAta-akuli.
	Apart from these Vedic sources for (possible) early Tibeto-Burmese,
the earliest datable, and so far not utilized evidence is found in Nepalese
inscriptions (467 CE+) (fn. 16). The inscriptions are in classical
Sanskrit, but contain a host of place names, some personal and tribal
names, and even a number of non-Sanskritic, traditional local names for
government offices which must be considerably older than c. 200 CE.
	A note on the transcription of 'foreign' words in Sanskrit and in
Indian alphabets is in order here. Just as in the case of adaptation of
'foreign words' to the Rgvedic phonetical pattern, the local words of the
Kathmandu Valley had to be adapted to the possibilities of Sanskrit
pronunciation and of spelling them in the Gupta (NAgarI style) alphabet.
#  several vowels are used intermittently: i/e, i/I, u/U/o (also va/o),
R/ri/o [@,o];
#  there is variation in some consonants as well, notably:
d/D (no retroflex!), tt/D, k/kh, b/bh, ll/ l, s/z (no S ?); jJ (common N.
Indian pronunciation: gy?); note aspirated m, n, r |hm, hn, hr|.
Typical is the spelling of the government office zolla/zullI/zulI or of the
name of the town of Bhaktapur in Licchavi inscriptions: khRpuG, khopRG
[kh¤priG], (mA-)kho-, > medieval khvapo, khvapva(M), khvapa, khapva, khopva
[khopa]) > mod. khvapya [khope], (for medieval names see Witzel 1999,
1993). Of importance is a variation (just as in Kanauri) that indicates
implosive consonants: co/cok/cokh. -- For all such variant spellings in the
Licchavi inscriptions, see Witzel 1980: 327, n. 60,69, 72, 74, 75, 87,
1993: 240 sqq., 248, n. 171-3, and 1993, n. 120, 152.
	The actual attribution of the locally spoken language and its
substrate found in the Licchavi inscriptions remains in the balance. It may
be early Newari or a predecessor, the kirAta language of the so-called
kirAta dynasty (see below) that reigned in the valley well before 200 CE
and has left us with names of government offices such as zulli, kuthera. If
it is indeed early Newari, it is a very archaic form, characterized by a
large numbers of initial clusters (Cr-, etc.), which differ even from the
oldest attested Newari texts ( 983 CE.) Such consonant clusters are very
rare in medieval and certainly in modern Newari.
	A clear case for TB is ti 'water'; I have compared (1980 n. 90, n.
94) co(kh)-, bu-, dol/dul, khu, gal/gvala of the Licchavi inscriptions with
mod. New. words: -co 'hill, mountain top', mod. New. cwa, cwak-, cf. Kaike
chwang, Khaling cong; (note also cuk 'mountain range' in Gilyak); -bu,
'land'; O.New. bu/bru, cf. Tamang pU; -gaa '*village'? cf. Mod. New.
"classifier for round objects, part of Kathmandu", O.New. gvala(M), but
note Skt. gola(ka), 'ball, globe'; perhaps cognate with TB (Benedict, 1972:
444) *r-wa / *g-wa; cf. 91 *wal 'round'; -ko 'slope', kwa, kwaa 'down';
pA-kA 'slope of a hill'; cf. Thakali koh-plen. (K. P. Malla has explained
some of such place names as being of Newari origin (1981: 17).
	The long list of substrate names includes (place names not specified):
aziG-ko (area) (ko 'river? or ko 'slope?'), uTTane, uDra, etaG- (village),
kaGku-laM (area) (lam 'road'?), kaDam-priG (area) (priG = pRG),
kampro-yambI, kambIlampra, kAduG- (village), kuthera-(office),
kuhmuM-(area) (see hAhmuG), keTumbATa (name of a KirAta official), kozI
(river), khaDabraMzai, khArevAlga-co (co, cok 'pass'), khuDU-(deity),
khRpuG- (village), khainaSpu (area), kho-pRG- (village), gamme (area),
tuJ-catcatu- (village), thuMtuM-rI- (fortress), daNDaG-guM, dommAna,
panapphu (area), puNDri-(palace), puttI- (river), prayiTTikhA (area),
proGprovAG, brahmuG (office), bhumbhukkikA- (deity), mAp-cok-(office) cf.
-co(k/kh) 'pass', yebraMkhara, rogamAcau (watchman), liG-gvala- (office),
vottarino?, voddi- (province), zulhmuG (office), zolla, zullI, zulI
(office), hasvimavallI- (village), hAhmuG- (place), hnA-guM, hmas-priG-
(village), hnu-priG, hrIm-ko (area), and many more.
All these data have not yet been exploited for Tib.-Burm. linguistics. (For
place names, see Witzel 1980, 1993; for relations between the eastern
Himalayan languages and Munda, s. Kuiper 1962: 42, with Nahali, p. 46f; cf.
Laufer 1916-18, 403 sqq.).
	The Kathmandu Valley, however, seems to have has its own strange
substrate, below this Tib.-Burm. level. It is visible in some place names
which definitely do not look Tib.-Burm. Some of them are characterized by
the geminates studied above: gamme, gullataMga, gollaM, jajje-, dommAna,
daGkhuTTA-, bemmA, cf. also bhumbhukkikA (onomatopoetic with double
consonant < *bhumbhum-ki-kA?).

$ 2.6. Other Himalayan Languages
	D. D. Sharma, Old-Indo-Aryan element in Kinnauri (in: R.K. Sharma
et al. (eds.), Dr. B. R. Sharma felicitation Volume, Tirupati 1986,
149-155) describes older elements in the kOchI dialect, spoken in the
western part of the former state of Bashahr, along the upper Satlej River.
The vocabulary given by Sharma, however, shows traces of OIA, MIA and NIA
-- as might have been expected. One curious feature of L.Kin. is the
division of nouns in animate (suffix -s) and inanimate (suffix -G) which he
compares to that of the Munda languages, while he links the endings to OIA
masc. -s, neuter -m.
	However, his materials represent a mixture of OIA, MIA and NIA
forms that have to be separated. Typically, we find OIA kvath 'to boil'
preserved as kwath or grAma 'village' as grAma-G (as opposed to NIA
gau~/gao~ etc.); next, forms which represent a MIA stage such as sappa-s
'snake' < sarpa, and NIA forms such as bAyA 'brother' < bhrAtA, tau 'heat'
< tApa, dauya-G 'curds' < dadhi, ana-G 'food' < anna, or mAmA 'maternal
uncle'. There are several cases of "gAndhArI metathesis" as well: trAma-G
'copper' < tAmra, cf. grota-N 'cow urine' < gomUtra etc.
	The case is of interest as it shows, just as that of early
Burushaski, the interaction of plains and mountain people (cf. also, below,
on Bangani). The present case also provides some indication of the early
date of such interaction between IA and TB speakers; this may be reflected
even in AV, if the kirAta indeed are TB speakers, and if the name has not
been passed on from an unknown earlier population (cf. the Kashmiri pizAca,
nAga traditions, above) to TB speakers.
	However that may be, from at least 1100 CE onwards, we see an
increasing Aryanization of the western Himalayas and W. Nepal with the
spread of the khaza tribe (found already in Manu's law book); by 1150 CE
they are still mentioned in the rAjataraGgiNI as settling southwest of the
Kashmir Valley. khas kurA is the self-designation of what was called the
"language of the Gurkhas" (in Newari called khaMy < khas); they have
substituted the name Nepali only in this century. By 1150 CE they had
established the W. Nepal/C. Tibetan Malla kingdom; by 1769 they had
conquered the Kathmandu Valley; and by 1900 they had settled, mixed with
Gurung, Magar, and other TB tribes speaking Nepali as lingua franca, in
Darjeeling, Sikkim, S. Bhutan and some parts of Assam. This movement is
indicated by their renaming of river names all across the Himalayas (Witzel
1993).
	Some part of the Himalayas may also have been occupied by the
pre-Tibetan language of W. and Central Tibet, Zhang Zhung. (See the list of
Zhang Zhung words, Thomas 1933, C. Beckwith, The Tibetan Empire in Central
Asia. Princeton University Press: 1987. The history of the settlement of
the Himalayas is far from clear. (For some details, based especially on
hydronomy, see Witzel 1993, and cf. now van Driem
http://iias.leidenuniv.nl/host/himalaya/). For example, the thAmi tribe who
live higher up in the tAma kosi valley east of Kathmandu belong, as their
language shows according to Shafer (1964: 3 n.1), to the Western Himalayish
group of the Bodic division of Tibeto-Burmese (Kanauri, etc.). Indeed, the
thAmi claim to have immigrated from Humla in northwest Nepal. This is one
indication among others (Witzel 1993) that there was a west-east flow of
population and languages, similar to the much later one of the Nepali
speaking Khas tribe.
	The intriguing question of Bangani has not been entirely resolved.
Bangani is spoken just east of Kinnauri, in the western-most tip of
Garhwal, Uttar Pradesh. Zoller (1988,1989) has reported a non-IA substrate
in this otherwise typical NIA language found high up in the western
Himalayas. Surprisingly, this substrate is a strange western variety of IE
with words such as ogno~ 'unborn' (not Skt. a-ja) and goNo 'give birth'
(not Skt. jan), kotro 'fight' (not Skt. zatru), dokro 'tear' (not Skt.
azru); the initial d- is W. IE, cf. Greek dakru, Engl. tear, as opposed to
E. IE : Skt. azru, Avest. asru, Lithuanian aSara. This claim has been
disputed by G. van Driem (1996, 1997), but has been sustained by research
carried out in Bangan by Anvita Abbi of Delhi University (see H.H. Hock [On
Bangani] http://www-personal.umich.edu/~pehook/bangani.html, with further
discussion). Anvita Abbi recognizes three layers in Bangani: words of the
type dokro, lokto, gosti, the general NIA Pahari level, and recent loans
from Hindi, etc.
	In principle, bands or tribes who have 'lost their way' and turn up
in unexpected areas are not altogether unknown. Tokharian, the easternmost
IE language, has western characteristics (ka"nt, ka"nte '100'), and the
North Iranian Alani, ancestors of the Ossetes, traveled all the way through
Central Europe, Spain and North Africa with the Germanic Vandals, to settle
in Tunisia.
	Tib.-Burm. is, however, not the first language in the Central
Himalayas. In Nepal it has been preceded by the isolate of Kusunda,
genetically unrelated to other language families just as Burushaski (see
below). Kusunda has recently been treated at length in MT II and III (cf.
Shafer, 1966 : 145; 1954 :10 sqq. The language is reported to have died out
by now. It is important to point out the difference between Hodgson's
(1848, 1880) and Reinhard's (1969, 1970) Kusunda, a point also mentioned by
P. Whitehouse MT III : 31; however, these differences extend beyond the
grammatical forms cited to the basic vocabulary, e.g. gipan 'hand'
H(odgson) : Aibi R(einhard); ing gai 'star/night' H : sA'nAm R (cf. ing,
ing ying 'sun'); jum 'moon' H : niho' R; cf. also smaller variations: toho
'tooth' H : uhu R; gitAn 'skin' H gitat R. It goes without saying that, for
a thorough investigation of Kusunda, the loans it has received from Nepali
and some of the neighboring TB languages such as (Kham-)Magari, Gurung,
Chepang, Newari, etc. must be taken into account, and that its relation to
the nearby substrate in Tharu (and Masica's "Language X") needs to be
evaluated.
	In passing, the old theory of a Munda substrate in the Himalayas
should be revisited. It goes back to S. Konow, On some facts connected with
the Tibeto-Burman dialect spoken in Kanawar, ZDMG 59, 1905, 117-125. This
has been denied by P.K. Benedict, Conspectus, p. 7, n. 23, by J. J. Bauman
(1975) Pronouns and Pronominal Morphology in Tibeto-Burman; and G. van
Driem 1992a, 1993b, 1993f, 1993g, 1994b, 1995a, 1997c, Rutgers 1993, Turin
1998 (see website :
http://iias.leidenuniv.nl/host/himalaya/individ/kirmor.html).
	Nevertheless, it must be remembered that the name of the R. gaNDakI
can be traced back to Munda. It is found all over Central Nepal, where the
major rivers are called "the seven gaNDaki". How far into the Nepalese
hills did the settlements of a Munda speaking people reach? Even in
exclusively Nepali speaking W. Nepal, the common hydronomical 'suffix' gAD
denoting 'river' may be connected with the Munda word da'k, ganda'k (Witzel
1993, 1999; further materials in Kuiper 1962: 10, with lit.; and already B.
H. Hodgson, Comparative vocabulary of the languages of the broken tribes of
Nepal, in: Miscellaneous Essays related to Indian Subjects, Vol. I p. 161
sqq., London 1880; cf. On the Che'pa'ng and Ku'su'nda Tribes of Nepa'l,
JASB XVII/2, 1848, p. 650 sqq.).
	A further hint may be provided by the implosives found in the
substrate of the Kathmandu Valley (cokh/cok/co, see above) and in Kanauri
(see Grierson, LSI on Kanauri). We may see here an areal feature of
implosives that has influenced both the Tib.-Burm. languages in Kinaur
(Kanauri) in the western Himalaya and in the Kathmandu Valley. Apart from
Munda and Sindhi, this feature is otherwise not found in S. Asia. There are
indications in the eastern Himalayas of a pre-TB population (Witzel 1993).
Even today, the Munda languages Satar and Santali are actually spoken in
the extreme south-east of Nepal (probably, like the Kurukh, recent
imports). Other Munda speakers are, after all, found south of the Ganges,
only about a hundred miles south of Eastern Nepal.
	Finally, there are the various Tharu tribes who live in the
foothills of the Himalayas, from the rAmgaGgA river in U.P. (India) to the
eastern border of Nepal, and in some bordering hill tracts, such as in the
rAptI Valley (Chitawan, just 50 miles SW of Kathmandu). They practice
slash-and-burn agriculture and nowadays speak a form of one of the
neighboring NIA languages, just like the Nahali or Vedda (see below);
however, I believe that we can find, again, a so far unstudied substrate
from a pre-IA, Pre-Munda language.
	Although often referred to as an archaic, remnant group, they have
been little studied (cf. the bibliography in Leal 1972). Some of the
vocabulary looks TB: for example TB ti- 'water' in Tharu suitI 'small
river.' (For -ti in Himalayan river names, see Witzel 1993). And indeed, D.
N. Majumdar, The Fortunes of Primitive Tribes, Lucknow 1944 reports blood
group types 'predominantly Mongoloid.' This is now supported by recent,
more advanced genetic studies. The Tharu are very isolated within S. Asia
(L. Cavalli-Sforza 1994: 84, 239 with fig. 4.14.1). As for the suspected
substrate, D. Leal, Chitwan Tharu Phonemic Summary. Kirtipur Summer Inst.
of Linguistics 1972, provides an example of the influence of their original
non-NIA language, i.e. the difficulty the Chitaun Tharu have to pronounce
aspirated mediae (bh > b@h; cf. above, on the Kathmandu Valley substrate).
	The Tharu word list in S. M. Joshi (ed.) paryAcavAcI zabda koz,
Kathmandu : nepAl rAjakIya prajJA-pratiSThAn VS 2030 (1974) contains lists
of 2914 words, most of which are close to Bhojpuri and Nepali; there are,
however, a number of words (cf. Witzel 1999, n. 43) which are neither
related to the surrounding IA languages nor to the nearby TB ones (Magar,
Chepang, Newari, Tamang) such as: ubbA 'small box,' koGhilA 'tiger', khUdI
'sugar cane', gukhA 'shaman', gulagula 'mild', gÆTTI 'splinter',
jhemjhemiyA 'small cymbal or drum', TippA 'mountain top' (probably NIA), ta
'small', tIra 'afterbirth', tIlvA 'whore house', nimak 'salt', bhubhui
'white scurf', yedi 'brick'. But the agricultural terms are NIA: bAjrA
'millet', dhAn 'rice', makai 'maize', gehUM 'wheat', as well as most of
their basic vocabulary.

	All these cases indicate that we probably can discover more
substrates if more work along these lines would be done. But we lack
etymological dictionaries for most NIA languages (apart from Turner's great
work, CDIAL), not to speak of Munda (in preparation by D. Stampe et al.)
and TB; (see, however, those on the internet: Starostin et al., accessible
from: http://starling.rinet.ru/). For example, it may very well be that the
Bihari languages have more Tib.-Burmese substrate words. There is, after
all, cAmal 'cooked rice' in Nepali, cAwal in Hindi, etc. which can be
connected with TB *dza 'to eat', Newari jA 'cooked rice, etc.' Yet, nobody
in Indian Studies is looking for such substrate material.

$ 3. Central and South India.
	Turning further South, the language isolate Nahali is spoken on the
upper TaptI river on the border of Maharashtra and Madhya Pradesh. To be
more specific, Nahali nowadays is a NIA language, but it shows below this,
at successively lower levels, a Dravidian, a Munda and an isolated level
which comprises some 24% of its vocabulary (Kuiper 1962: 51, 1966). The
speakers of modern Nahali, to be short, the Nahals are the remnants of the
first Indian population. At least, they have preserved the remnants of the
earliest language spoken in India that we can ascertain so far. Future
comparisons may lead us beyond that, for example the proposed comparisons
between Nahali and Ainu, or between Andamanese and Papua (Indo-Pacific).
	Nahali has been extensively treated in this macro-comparative way
in MT II and III. As has been first seen by Shafer and Kuiper, Nahali has
connections with Ainu, etc. (for which now see MT II), and thus represents
remnants of the earliest substratum of modern homo sapiens sapiens that
moved from the Near East all the way to E. Asia (and S.E. Asia, Australia).
However, it must be noted that the retroflex sounds in Australian are a
relative new development as well and cannot be the cause of their (almost)
Pan-South Asian prevalence in prehistoric times.
	Berger (1959) was of the opinion that the Nahals were identical
with the well known niSAda of the Chambal, Malwa and Bandelkhand areas. He
discussed their mythology as found in the Mahabharata; however the niSAda
are found already in the Middle Vedic texts. The nihAl or nAhal are also
found (Berger 1959: 35) in many medieval texts, such as in Hemacandra's
Grammar (c. 1200 CE) as lAhala; in Padma Pur. nAhalaka, together with the
bhilla, as mountain/jungle tribe; in puSpadanta's harivaMzapurANa as
nAhala, synomym of bhilla, savara (another jungle tribe : modern Saora);
also in vikarmaGkadevacaritra of bilhaNa (c. 1150 CE), and in rAjazekhara's
drama bAlarAmAyaNa (on the R. narmadA). Berger wanted to identify them with
the DahAla as well; they are found in inscriptions of the kalacuri dynasty
of tripurI and in Albiruni (1030 CE). All of their territories are c. 400
km away from the modern eastern Nahalis near Nimar.
	He thus derived Nahal/Nihal from a form such as *nezad reflected by
Ved. NiSAda. Indeed, the word is found in early post-RV texts: KS, MS, and
with the typical sound changes in 'foreign' words: NiSAda : *NiSidha : ZB
NaDa NaiSidha, (apparently the Vedic 'ancestor' of the Epic Nala NaiSadha :
*NiSadha); thus  d: dh (as in magadha : pra-magandha, etc.). The name
certainly is a popular etymology (however, the modern self-designation of
the Nahals is kalTo, du. kalTih-Tel, pl. kaliTTa; < stem *kaliT-o, s.
Kuiper 1962: 82, 17, 27, Mundlay MT II 5-7, no. 858 kalTo, pl. kolTa).
	The niSAda are described in Vedic texts (first MS 2.9.5 =KS 17.13,
TS 4.5.4.2, VS 16.27) as being ¸neither wilderness (araNya) nor settlement
(grAma);› who are ¸given over to the earth:› (asyAm eva parIttAH), next to
jana '(foreign) tribe' PB, other non-Brahmins (JB), and samAnajana ¸one's
own people› (cf. PB 16.6.7-9); cf. also KB 25.15, LZS 8.2.8 on temporary
residence in a naiSAda settlement. Similarly, MS 2.9.5 describes the
niSAda, among Rudra's names and his people, together with hunters and other
low caste people (=KS 17.13, TS 4.5.4.2, VS 16.27); -- AB 8.11 as robbers
in the wilderness; similarly the dasyu JB 2.423:$168, where the text
insists on kSatriya accompaniment during travel, necessary to keep the
dasyu at bay and turn them ¸sweet (madhu)›, cf. AB 8.11 where the dasyu rob
a wealthy man or a caravan in the wilderness. Acculturation is seen at MS
2.2.4, where their chief (sthapati) is allowed to offer sacrifices, cf. KZS
1.1.12. The inclusion of the headman of the niSAda reflects the well-known
process of upward social movement, called ¸Sanskritization.› (Witzel 1997)
	Their Vedic designation obviously is a popular etymology "those who
sit at home." However, they are more frequently described as robbers (still
a favorite occupation of the Nahals in early British times) -- against whom
one had to guard when traveling through uninhabited territory. Their
chieftains (sthapati), however, were allowed into the Aryan fold and could
perform solemn Vedic sacrifices, clearly an early form of Sanskritization.
	It may very well be that Rajasthani has a strong Bhili (and Nahali)
substrate; Koppers (1948: 23, Kuiper 1962, 1966, 1991) and Shafer (1940,
1954: 10) thought that the Bhils once spoke Nahali as well. The Bhils are
now widely spread between the arAvaLA (Aravalli) Mountains, the Vindhya
Mts. and the Tapti River (Khandesh area); they now speak Gujarati-like IA.
	In the Vindhyas we find a number of north and central Dravidian
languages. However, both North Dravidian languages, Kurukh (Oraon, on the
borders of Bihar/Orissa/Madhya Pradesh; the settlement in Nepal and Assam
is recent) and Malto (on the bend of the Ganges in S.E. Bihar) are
late-comers to Munda territory as many loans from Munda languages indicate.
On the other hand, the third north Drav. language, Brahui, spoken in
Baluchistan has returned to E. Iran only a few hundred years ago (Elfenbein
1987); it has no older Iranian loans (from Avestan or Pashto, just from
their symbiotic neighbors, the Baluch).
	In the Vindhya Mountains we find such names as the following: the
Vidarbha people, in the area around Nagpur, (the mod. barhAD, Berar <
virATa, Mbh) are mentioned (JB), along with their fierce mAcala dogs 'that
kill even tigers' (note that this is an area with early iron and horses).
vidarbha seems to be a popular etymology vi-darbha 'with widely spread
darbha (grass)', especially if connected with Munda da�b 'to thatch'
(Pinnow 1959: 69), cf. vi-bhindu in the Gangetic plains (above). The name
of the vibhindus is related to that of the bainda tribe (derived from
*bind) that still survives in the Vindhyas today, and names such as
ku-sur(u)-binda (above). The very name of the Vindhya (post-Vedic) can be
related, with typical Sanskritizing interchange of d : dh, as in
pra-maganda : magadha, (above). East of these mountains, we have the
kaliGga (cf. triliGga south of Orissa) and aGga, vaGga. All of these are
names that hardly have a Drav. etymology, but which look Austro-Asiatic
because of their prefix changes.
	However, all around Vidarbha, the first Drav. river names are met
with : the pUrNA (< *pEN) west of it, the vEn-gaGgA east of it, and the
pain-gaGgA south of it. They all are adaptations of a Drav. term for
rivers, DEDR 4160a *pEN-: *peN-V- 'to twine, twist'. It seems that the area
which still has a Munda name in the Vedic middle period (vidarbha) has also
received a Dravidian overlay. This is confirmed by Drav. place names in
-oli in Maharastra and in -palli, -valli, -pal in Bastar, just east of the
Vidarbha area (now southernmost Madhya Pradesh) where they range from 21%
in the south to only 0-4% as one approaches the Raypur plains. The south
and southwest of Bastar is occupied by the Drav. Gonds, all other regions
by Chattisgarhi Hindi speakers. (For an overview of studies in (South)
Indian place names see the paper by M.N. Nampoothiry, Indian Toponymy. A
critical evaluation of the work done in this field in India with a
bibliography in: Puthusseri Ramachandran and K. Nachimuthu (eds.)
Perspectives in Place Name Studies : Proceedings of the National Seminar on
South Indian Place Names, Held at Trivandrum on 21-23 June 1985. A
Festschrift to Prof. V.I. Subramoniam, On His Sixtieth Birth Day.
Trivandrum: Place Name Society, 1987. p. 1-47, --- including a good
bibliography, also of unpublished Indian theses).
	The South is frequently supposed to have been Dravidian from times
immemorial. However, in the refuge area of Nilgiris with their isolated
Drav. tribes (Toda, etc.), we find a substrate, see Zvelebil 1990, 63-70.
Isolated words indicating this pre-Drav. substrate (Zvelebil 1990: 69f.,
Zvelebil 1979: 71f.) include the following Irula words mattu 'lip',
Do"kene, dekene, Dekena, Dekkada 'panther', ovarakaGku, OrakaGku, OraGgeku,
OraGge, Orapodu 'tomorrow' (unless DEDR 707 Tam. uR2aGku 'to sleep'),
buNDri 'grass hopper' (unless DEDR 4169), muTT(u)ri 'butterfly' (unless
DEDR 4850 miTL 'locust'), vutta 'crossbar in a house'. These instances
should encourage Drav. specialists to look for substrates in Tamil, Telugu,
Kannada, etc. However, just like the propagators of indigenous "Aryans" in
the North, Dravidians of the South frequently think that they are
autochthonous.
	In Sri Lanka, the remnant population of the Vedda now speaks
Sinhala. (De Silva, M.W. Sugathapala, Vedda language of Ceylon; texts and
lexicon. Mu"nchener Studien zur Sprachwissenschaft. Beiheft n.F. 7.
Mu"nchen: R. Kitzinger, 1972). The substrate that they may have preserved
is in urgent need of thorough study, carried out by comparing Pali, Sinhala
and Tamil words. Some typical words, interestingly many with geminates,
that cannot be linked either to Sinhala or to Tamil are: cappi 'bird',
munDi 'monitor lizard', potti 'a kind of bee', panni 'worm', rukula 'home,
cavity' (see de Silva 1972 : 16; his vocabulary, pp. 69-96, does not
contain etymologies).
	Finally there is Andamanese, but unlike the Austro-Asiatic
Nicobarese, so isolated that it can only be compared in long-range fashion,
with other Australo-Pacific languages.

$ 4. The Northwest.

	We now return to a region for which we have larger amount of early
sources, the Greater Panjab, the area of the first Indo-Aryan influx into
the subcontinent as reflected by the hymns of the RV. As has been pointed
out, the Rgvedic area is characterized by an almost total substitution of
local, pre-IA river names by those of IA type, such as gomatI 'the one
having cows' (mod. Gomal), mehatnu ' the one full of fluid", asiknI 'the
black one' (now Chenab). Tribal names, include next to typical IA ones
(druhyu 'the cheaters', bharata 'the ones who carry (sacred fire?),' many
that have no plausible IA etymologies, such as: the gandhAri tribe of
gandhAra, the area between Kabul and Islamabad in Pakistan; zambara, a
mountain chieftain; vayiyu and prayiyu (chieftains on the suvAstu, modern
Swat); mauja-vant, a Himalayan peak. This kind of evidence indicates the
typical picture of an intrusive element, the IA, overlaying a previous
population. Unlike Northern America for example, only a few pre-IA river
names have survived, such as: kubhA (mod. Kabul river), krumu (mod.
Kurram), and maybe even the sindhu (Indus); these have no clear or only
doubtful IA/IE etymologies (see below).
	North of this area, at the northern bend of the Indus
(Baltistan/Hunza), the language isolate Burushaski is spoken whose
prehistory is unknown (cf. now MT II, III). However, the language and the
tribal name are indirectly attested in this general area ever since the RV:
*m/bruZa (mod. buruSo) > Ved. mUja-vant, Avestan muZa (see below). Indeed,
already the RV contains a few words which are still preserved in Bur., such
as Bur. kilAy, Ved. kIlAla- 'biestings, a sweet drink' RV 10.91.14, (note
AV 4.11.10 next to the loan word kInAza, see above); kIlAla cannot have a
IA etymology (EWA I 358 'unclear'); continuants are found in the Dardic
branch of IA (Khowar kiLAl), and in Nuristani (kilA' etc.), as well as in
later Skt. kilATa 'cheese', cf. DEDR 1580 Tam. kil~AaN2 'curd'); for
details see Kuiper 1955: 150f., Turner, CDIAL 3181, Tikkanen 1988. Further,
the following words, mES 'skinbag', CDIAL 10343 < Ved. *maiSiya 'ovine',
meSa 'ram' RV; gur 'wheat' pl. guriG/gureG < *ghorum, gurga'n 'winter
wheat', cf. Ved. godhUma; bras 'rice', different from briu' 'rice (< Shina
briu' ), cf. Ved. vrIhi; bus 'sheaf', CDIAL 8298, cf. Ved. busa, bRsI
'chaff' (cf. Pinnow 1959: 39); ku(h)a' (Berger ghua') 'new moon', cf. Ved
kuhU 'deity of new moon'; ghupas (Berger gupa's) 'cotton', cf. Ved.
karpAsa, Kashm. kapas; baluqa 'stone' (in a game), cf. ba'ltaS 'stone
thrown at someone', cf. Ved. parazu '(stone) ax', Greek pe'lekus, see EWA
II, 214; baG 'resin of trees', baG ~ IIr bhaGga 'hemp, cannabis', cf.
Khowar boG. Most of the words from IA languages in Turner's CDIAL that have
Bur. correspondences are, however, late loan words from the neighboring
Dardic languages, especially from Shina and Khowar (cf. Lorimer 1937,
Berger 1959, 1998).
	Importantly, in Proto-Burushaski (or in its early loans from the
lowlands) and the pre-Vedic Indus language there is, as treated in $ 1.10,
there is interchange of k/z, and retention of -an- (not > -o-): Bur. kIlAy
: Ved. kIlAla, but Son 'blind one-eyed' : Ved. kANa; ghoro (Berger ghuro')
'stone, pebbles', cf. Ved. zar-kara, cf. also (Witzel 1999) ghoqares,
Berger gho'kura.c. 'raven', Ved. kAka; Bur. ghazu' 'onion', cf. Ved.
lazuna, Shina kazu; ghon, Berger ghu'un 'quail', cf. (?) Ved. laba. It has
indeed occasionally been maintained that Burushaski extended into the
Panjab in earlier times (L. Schmid 1981, Tikkanen 1988), but the Vedic
evidence does not support this. We cannot be sure exactly how far Rgvedic
geographical knowledge extended northwards, and how much practical
interaction existed between RV and Proto-Burusho people. Yet, the RV knows
of some small right side contributory rivers of the Indus that are located
north of the confluence with the Kabul River; they have IA names: RV
10.75.6. tRSTAmA '< tRS 'the rough, (or) the dried up (river)', susartu
'the one running well', rasA 'the one full of sap', zvetI 'the white one'.
	While it is questionable how far south Burushaski territory
extended at this early time, some of the loan words mentioned above
indicate that there was early contact. That extends perhaps also to
medicinal and other herbs (cf. below on KirAta), for it may be that the
name of the BuruSo is reflected by the RV mountain name mauja-vant "having
mUja (people)", cf. the east Iranian equivalent, Avestan muZa. This is the
mountain where the best Soma, a hallucinogenic plant, comes from. The RV
and E. Iranian (Avestan) forms look like adaptations of the local
self-designation, *mruZa, Vedic mUja-, Avest. muZa, and are attested since
the middle of the first millennium in early Tib. bru-Za, Sanskritized
puruSa (von Hinu"ber 1989, 1980), local 10th cent. inscriptions prUzava
(Jettmar 1989: xxxvii), mod. Bur. buruSo.
	Phonetic reflexes of Bur. have been seen (Tikkanen 1988) in the
Vedic (and Dravidian) retroflex consonants that have otherwise found a
number of explanations, from a Dravidian substrate to an internal East
Iranian and Vedic development. The occurrence of these sounds clearly
reflects an areal feature that is strongest in the Northwest, but extends
all the way to Tamil in the South, and has also influenced Munda to some
extent. Below, it will shown that it is an ancient feature of the Indus
language as well, and that it must not be traced back to Bur. influence,
which seems to have been limited, even in Rgvedic times, to the upper Indus
valley.
	Some early syntactic influence by Burushaski on Vedic in the
formation of the Absolutive has been assumed by Tikkanen (1988); it is
found already in earliest RV but only as past verbal adverb/conjunctive
participle. This clearly S. Asian feature, unknown in the sister language
of Vedic, Old Iranian, is also found in various degrees in Drav. and Munda,
and may have been an early regional feature whose ultimate origin remains
unclear (cf. Witzel 1999)

	Another modern language in the same area is Khowar which belongs,
along with Kashmiri, Swati, etc. to the Dardic branch of IA. In its
phonetics and vocabulary, however, it shows a strong local substrate,
similar to Burushaski. Unique for Khowar, however, is a particular
substrate whose origin remains unclear so far. It seems that the Khowars
are a late immigrant group who have taken over a Dardic language.
Substrate(?) words in Khowar which are neither IA nor Burushaski include
(Kuiper 1962: 11, cf. Morgenstierne 1947: 6, Lorimer 1935 : xxi): ghec
'eye', ap�ak 'mouth', krem 'back', camoTh 'finger', iskI 'heel', askAr
'lungs'. Kuiper (1962: 14) compares ghec 'eye' with Bur. ghai(c)-, gh'i-,
ghe-ic- 'to appear, seem, be visible', and with g'e- 'to look, seem,
appear', da-g'e- 'to peer' of the Munda language Sora and with Parengi gi-
'to see'. (Differently, Morgenstierne, FS Belvalkar, 2nd section p. 91.)
For Bur. loans in Dardic and in Nuristani see Tikkanen 1988: 305 (cumar
'iron', ju 'apricot', etc.), cf. Fussman 1972 II, 37 sqq.; Lorimer 1938:
95, Morgenstierne 1935: xxi sqq., 1947: 92 sqq.; Schmidt 1981, Berger 1998.

	The neighboring area, Kashmir, is of great interest. Its prehistory is 
little known. In the Neolithic, there were relations with Central Asia and 
China, but the influence of the Indus civilization (2600-1900 BCE) is strong 
and long-lasting; of course, this does not tell us anything about the 
language(s) spoken then. Unfortunately, the Vedic texts, which know of the 
neighboring Indus valley do not mention Kashmir by name. It is first 
mentioned by the grammarian PataJjali (150 BCE). The native Kashmiri texts
(rAjataraGgiNI, nIlamata purANa, cf. Witzel 1994, Tikkanen 1988, L. Schmid
1981), however, know of the previous populations, the pizAca 'ghouls' and
the nAga 'snakes' (that can change into human shape at will). These are
common names for 'aboriginals'; cf. the Tib.-Burm. Naga tribe on the
Burmese border. Yet, these designations may retain some historical memory.
The chief of the pizAca is called nikumbha (nikumba in MilindapaJho), and
the NAgas have such 'foreign' names such as karkoTa, aTa, baDi, bahabaka,
cATara, cikura, Ccukkaka, etc. The list of some 600 Kashmir nAga names in
the local nIlamatapurANa contains many such non-Sanskritic names; they have
not been studied (see Witzel, in press).
 	 Just as in Northern India and Nepal, most river and place names in
Kashmir have been Sanskritized; note, however, the river and place names:
ledarI, a river in the SE of the Valley (also in the place name levAra <
ledarI-agrahAra); -muSa, a 'suffix' in the names of several villages:
khonamuSa (mod. Khunamoh), katImuSa, (mod. Kaimoh, next to lati-kA), rAmuSa
(mod. Ramuh); also, the paJcAla-dhAra mountain, (mod. (pIr) pantsAl range,
south of the Valley), may reflect an old name, cf. the Ved. tribal name
paJcAla, and Grierson, Dict. of Kashmiri III : 744; cf. Nepali himAl
'Himalaya range', CDIAL 14104. Such names have not been studied in detail
(cf., however, L. Schmidt 1981, Witzel 1993).
	Like all other Indian languages, the Kashmiri language itself has
not been thoroughly scrutinized for more substrate materials, cf., however,
the report by L. Schmidt (1981), who assumes that 25% of the vocabulary and
toponymy belong to a pre-IA substrate. A. Parpola (Tikkanen 1988: 305)
thinks of a Proto-Tib. or Sinitic substrate. However, the peculiar
phonology of Kashmiri (and Dardic in general) sustains the assumption of a
strong northwestern substrate influence.
	In the northwest another IIr. language which shares some regional
peculiarities with Dardic, is spoken: Nuristani or Kafiri, as it was
formerly called, is (differently from the older handbooks which lump it
together with the Dardic branch of IA) a third branch of the Indo-Iranians
(G. Morgenstierne, Irano-Dardica. Wiesbaden 1973). It has survived in the
mountains of East Afghanistan and in neighboring Chitral (N.W. Pakistan).
The Kalasha (Chitral) subgroup have even preserved their ancient non-Hindu
and non-Iranian religion. Nuristani has preserved such sounds as IIr. c'
that has been changed even in the RV > z (c. 1500 BCE) and in Old Iranian >
s. It has transmitted at least one loan word into Vedic, Nur. *kat'S'a >
Ved. kAca 'shining piece of jewelry' (K. Hoffmann 1986, EWA I 335).

	Finally, one must be open to assume the influence of other
substrate languages in the Hindukush/Pamir areas. There are local personal
names such as RV zambara kaulitara and his father *kulitara who are 'in the
mountains', prayiyu and vayiyu in Swat; names of demons (as always,
intentionally confused with those of real, human enemies) such as cumuri,
namuci, uraNa, arbuda, pipru, zambara; tribal names such as gandhAri,
dRbhIka(?), varc-in(?); river names such as  kubhA, krumu, sindhu(?). Note
also that the Avesta (Videvdad 1) speaks about some of these areas, notably
var@na (varNu) as an-airiia "non-Aryan".


$ 5. Indo-Iranian substrates from Central Asia and Iran

	Beyond this area, Central Asia must have been the source of a host
of unstudied words in Proto-IIr., which are found both in IA and Old
Iranian but which do not have an IE etymology and must represent the
language of the Bactria-Margiana region (BMAC culture 2100-1900 BCE), or
other Central Asian substrate(s). They include plants, animals, and
material culture; their concentration in the area of brick-built settlement
and agriculture as well as some newly introduced animals should be noted.
Such words, as found in Ved. /Avestan, include:
	# uSTra / uStra 'camel', middle and new Akkadian udru "Bactrian
camel" is a loan from Iran, see EWA I 238, KEWA III 652, cf. Diakonoff in
JAOS 105, 1985, 600; the camel was introduced into the BMAC area from
Central Asia only in the late 3rd mill. BCE.
	# khara / xara 'donkey', cf. Toch.B ker-ca-po < *karca-bha?, with
the common Indian animal suffix -bha (as in garda-bha, zara-bha, RSa-bha);
the word ultimately may be a late 3rd mill. Near Eastern loan, cf. Akkad.
(Mari) HArum, ajarum 'male donkey', EWA I 447. Note also the overlap with
Dravidian (denied by EWA 473): Drav. *garda > Tamil kal_utai, etc., one of
the few possible links of a Central Asian substrate with Dravidian (and
with Vedic);
	# iSTi, iSTikA / iStiia 'brick', z@mOiStuua 'clay brick'; OP. iSti,
MP., NP. xiSt; cf. Toch. izcem 'clay'? Clay bricks are unknown in northern
Central Asia (Kazakhtan), the putative homeland of IIr (except for their
sudden appearance in the Sintashta Culture east of the Urals, c. 2000 BCE,
for which a link with the BMAC has been supposed);
	# sthUna / stUnA, stunA, OP. stUnA 'pillar', unless it belongs to
Ved. sthUra 'tall, thick', Avest. -stura, Khot. stura (thus EWA II 768);

	# yavyA /O.P. yauviyA 'channel', > MP., NP. jO, jOy 'stream,
channel', Parachi ZI 'rivulet', EWA II 405; both words, typical for loans,
do not go back to exactly the same source;
	#  godhUma / gantuma 'wheat' from a Near Eastern language, cf.
Semitic *HnT, Hitt. kant (EWA 499) and Egyptian xnd;
	#  parSa / parSa 'sheaf', see EWA II 101;
	# bIja / OIran. *bIza (in names), 'seed, semen', Buddh. Sogdian
byz'k, Parachi bIz 'grains';
	# zaNa / kana- 'hemp', MP. San 'hemp', Khot. kaMha, Osset. gŽn,
gŽnŽ, Russ. Church Sl. konoplja, Gr. ka'nnabis, itself a loan from
Scythian, as also also Old High German hanaf, Dutch hennep < *kanap;
	# bhaGga / banga 'hemp, hashish', if the word does not belong to
bhaJj 'to break';
	# *sinSap 'mustard': Ved. saSarpa 'mustard', Khot. zzazvAna,
Parthian SyfS-d'n, Sogdian SywSp-dhn, MP. span-dAn 'mustard seed'; Greek
si'napi; < pre-Iran. *sinSapa < **sinsap (Henning s1ens2ap); cf. also:
Malay sawi, s@sawi, or Austro-As. *sapi, sV(r)-sapi; further EWA 712, 727:
ziMza'pA RV+ 'Dalbergia sissoo' NP. SISam, Pashto S@wa < *zISampA, CDIAL
12424), Elam. Se-iS-Sa'-ba-ut = /SeSSap/;
	# kazyapa / kasiiapa 'turtle', Sogdian kySph, NP. kaSaf, kaS(a)p
'tortoise'; cf. Kashaf RUd, a river in Turkmenistan and Khorasan;
	# pard/pandh 'spotted animal, panther' : Ved. pRdAku 'snake' RV,
pRdakU AV, pRdAkhu BZS (EWA II 163), with Para-Munda prefix p@r?; Khowar
purdu`m < *pRdhUma? KEWA II 335, CDIAL 8362; Bur. (Yasin) phu'rdum 'adder,
snake'; later Skt. 'tiger, panther'; NP. palang 'leopard' < O.Iran. *pard-,
Greek pa'rdalis, pa'rdos, le'o-pardos 'leopard' (EWA II 163), all < **pard
'spotted, wild animal?'; Henning reconstructs **parth (but note Greek
pa'nthEr), which may have been close to the Central Asian form;
	# *kar(t)ka 'rhinoceros', Ved. khaDga 'rhinoceros' MS+, EWA 443,
cf. N.P. karka-dAn, Arab. karkaddan, Aelianus karta'zOnos (*kargazOnos)
'Indian rhinoceros', all from a pre-Aryan source; however, cf. Kuiper 1948:
136 sqq.
	# bheSaja / baEsaziia 'healing'; IIr *bhiS-aj > Ved. bhiS-aj; the
root *bhiS may be a loan word (cf. EWA s.v.);
	# vInA 'lute': Ved. vINA Khot. bIna 'harp, lute', Sogdian wyn'
'lute', MP. win 'lute', Armen. vin 'lute', unless loans from India, cf. EWA
II 568;
	# *kapauta 'blue': Ved. kapota 'pigeon', O.P. kapauta 'blue'; Khot.
kavUta 'blue', MP. kabOd 'grey-blue', kabOtar 'pigeon'; EWA I 303, Kuiper
1991;
	# *kadru 'brown': Ved. kadru 'red-brown', kadrU 'a snake deity',
Avest. kadruua.aspa 'with brown horses, NP. kahar 'light brown';
The following words may be of still older origin and may have been taken
over either in E. Europe or in Northern Central Asia:
	# *medh/melit 'sweet, honey': IE. *medhu 'sweet' is found in Ved.
madhu 'sweet, honey, mead', Avest. madhu, Sogd. mdhw 'wine', (cf. Bur. mel
'wine, from grapes'), Toch. B mit 'honey', Gr. me'thu 'wine' etc.; it has
spread to Uralic *mese, mete; Finnish mete, Hungarian me'z 'honey', Chin.
mi < *miet, Sino-Korean mil, Jpn. mitsu < *mit(u); Iran. *madhu > Turkish,
Mongolian bal 'honey'; Arabic mAdI?, and to > Toch. B mot 'intoxicating
drink'. --- From another source **melit, Greek me'lit-, Hitt. milit, Latin
mel, mell-, Gothic milith; in Nostratic (Illich-Svitych, Opyt II, Moskva
1976 : 38sq.) both forms are united under *majLa > *Ural. majdh'a', Drav.
maTT, miTT, Altaic /m/ala, bala; cf. also, still further afield, in
Polynesia: Samoan meli, Hawaiian mele, meli; mele, melemele 'yellow', Maori
miere; Tongan melie 'sweetness, sweet, delicious', Rarotongan meli 'honey',
Mangareva mere 'honey'.
	# *sengha/singha 'lion' : Ved. siMha 'lion' < * sinj'ha < *sing'ha
differs from Proto- Iran. *sarg: Khoresmian sargh, Parthian Sarg, Khot.
sarau; Henning reconstructs **s1eNgha; -- loans into nearby languages, such
as Toch. A ziza"k, B zecake 'lion'; Tib. seGge, Chin. *suan-ngei (Henning,
EWA), note, however, Karlgren 1923, no. 893 Arch. Chin. *,Si, Jpn. *si >
shi(-shi); cf. perhaps Armenian inc, inj EWA II 727, KEWA III 447; the
western IE languages have received the 'lion' word from a different source,
Gr. lIs, leon(t)-, Lat. leon-.
	 In short, western and central Iran must have been inhabited by
(archaeologically well attested) peoples of non-IIr speech. However, their
languages have left few remains in Iranian. Apparently, Elamian was spoken
up to simaSki (Kerman/Bandar Abbas area), while aratta (Sistan) and marhaSi
(W. Baluchistan, Bampur region) apparently had other language(s), (Vallat
1980). All of these data need to be studied in greater detail, especially
the early IIr substrate language(s).

$ 6 Conclusions.
	In short, the early linguistic picture of South Asia in the second
and first millennium BCE, during the Indus and Vedic periods, is as complex
as, or even more so than its modern counterpart. The materials adduced
above also indicate that, even with the addition of the modern descendants
of Proto-Burushaski, -Nahali and -Kusunda, we have to reckon with, and make
use of a number of substrate words from such languages as Masica's
"Language X", Tharu, the Kathmandu Valley, or the Panjab and the Sindh
varieties of the Indus language. It must be underlined, that except for the
few items pointed out for the Vedda and Nilgiri languages, the prehistoric
linguistic situation of South India (before Dravidian) is entirely unclear:
in this respect, a lot of spade work needs to be done by Dravidian
specialists; the same applies to Munda and the eastern and central parts of
India; yet, just as in the modern North Indian languages, no progress has
been made in this respect over the past few decades.
	The few available etymological dictionaries do not provide detailed
information about the historical and geographical spread of the words
discussed, though Mayrhofer's EWA now gives an idea at least of the
historical levels, but hardly of the geographical spread. DEDR does not
have any such information yet, and we need to check the on-line dictionary
at Cologne
(http://www.uni-koeln.de/phil-fak/indologie/tamil/otl_search.html); and the
KWIC Concordance of Classical Tamil texts
(http://www.uni-koeln.de/cgi-bin/SFgate). A Munda etymological dictionary
is still under preparation.
	In addition, the ancient Vedic and Tamil texts still hold out a lot
of important and interesting data. We would profit very much from detailed
historical grammar of Tamil and a study of substrates in Tamil (and the
other Dravidian languages).
	The data discussed above indicate that we have to reckon with a
number of layers of languages (and the populations which used them). The
situation is best illustrated by Nahali (see above) with its subsequent
layers of Proto-Nahali, Munda, Dravidian and NIA. If Hindi was studied in
the same way, we would find similar layers of Masica's "Language X",
Para-Munda, Old IA (with influences from the Indus language, and
Proto-Drav., -Munda, -Tibeto-Burmese), early Persian (dipi/lipi 'script')
and Greek (yavana 'Greek', suruGgA 'subterranean channel', but cf. Kuiper
1997: 186-190) loans, a continuous stream of Sanskrit loan words, medieval
loans from Arabic, Turkish, Mongolian and Persian, as well as the more
recent English loan words and Neo-Sanskrit words such as dUrdarzan
'television'.
	Especially, the etymology of Panjabi and Sindhi words should be
taken up, finally, in order to delineate the linguistic history of these
areas that are so critical for the immigration and acculturation of IA and
Drav. speakers. A thorough study of the (usually very conservative) river
names, not just of the major rivers mentioned above but even of small
creeks, as has been done in Europe during this century, would substantially
aid in this undertaking. Names of settlements change much more easily but
should not be neglected either.
	In comparison with the linguistic history of the nearby East
Iranian languages (especially Pashto), this kind of investigation would aid
substantially in determining the history of human settlement in South Asia
and would be a major contribution to the ongoing debate about the "Aryan
invasion" or, rather, the trickling in, immigration and amalgamation of
speakers of IA (as well as Dravidian) languages. Once the data derived from
archaeology and genetics are added, a much clearer picture of the
settlement of South Asia will finally emerge that will put much of the
current speculation to rest.




===========================================================================
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