§13.  Absence of Indian influences in Indo-Iranian

	When compared to Eastern IE or to the rest of IE, Avestan and Old
Persian share many innovations with Vedic, which was the initial reason to
set up this group of languages as  a separate branch of IE, IIr. Just as in
biology (taxonomy, the human pedigree, genetics, etc.) or in manuscript
study (setting up of a stemma), the occurrence of common innovations always
indicates that the innovative group has split off from the core group, and
obviously is to be dated later  than the core.
	For example, Vedic ah-am ''I''= Avestan az-@m, az-„m O.Pers.  ad-am
have added the additional morpheme IIr. -am (as in ay-am, iy-am); it was
transferred to the rest of the pronouns: tvam, vayam, yUyam as well. This
feature is not found in other IE languages: Lat., Greek egO, Gothic ik
(Engl. I), O.Slavic azu, jazu; it clearly separates IIr. from the other E.
and W. IE languages.
	While Iranian, at first sight, seems to be more innovative than OIA
in its phonology (s > h, kh > x;  p, t, k + consonant  > f, th, x + cons.,
etc.), it frequently is also more archaic than Vedic. It lacks the many
innovations that characterize Vedic, for example the absolutives in -tvA,
-ya, ntr. pl. in -Ani,  the perf. jaga-u, or the normalization in g- of the
present stems beginning in j/g-: IE gwm-sk'e-ti > IIr.  *ja-s'ca-ti >
Avest. jasaiti :: Vedic gacchati. (Note that j is retained only in
traditional names such as  jamad-agni and in the perfect, ja-gAm-a, etc.)
Importantly, Iranian it misses the generalization of the already Rgvedic
e-perfects, derived from IIr. *sazdai (Avest. hazde) > Vedic sede with many
analogical formations such as mene. Since sound changes are not random and
develop in linear fashion, these innovations must have occurred well after
Vedic had separated from late IIr./pre-Iranian, thus : IE --> E. IE -> IIr
--> Vedic, or Iranian.
	The advocates of the autochthonous  theory, however, would have the
Vedic innovations occur in the Panjab only after the Iranian speakers had
left the subcontinent, while retaining some very archaic features.
(Talageri 2000, against all linguistic evidence, even denies close
relationship of both groups). Some other innovations found both in India
and Iran would have occurred earlier than that while both groups still
lived in the Panjab; still others (found in E. IE, such as in Slavic) would
have occurred at a still earlier, third level, again in the Panjab, while
languages of the fourth level (including Greek, Latin, Germanic, etc.)
would have left the subcontinent even before  this.
	While all of this is possible in a purely theoretical scenario,
there are a number of arguments that render it impossible. Some of them
have been listed by Hock (1999, see above). Others include such items as
the temperate, non-tropical core vocabulary of IE, early IE loans from
Semitic somewhere in the Near East (**wVjn-, IE *woin- 'wine', cf. J.
Nichols 1997: 143), or on a more typological level, the intermediate
position of IE between the Uralic and Kartvelian (W. Caucasian) language
families (Nichols 1997, 1998). As far as the Satem language IIr is
concerned, one can add the early close links of IIr (and, later, early
Iranian) with Uralic in S. Russia and in the Ural and W. Siberian regions,
and the new terminology coined for the horse-drawn chariot (ratha/ratha),
first introduced in the S. Russia/Ural area. This list, which could be
extended, clearly points to the areas north of the Near East, and strongly
militates against the assumption of an Indian homeland of OIA, IIr, and,
worse, of IE (see below).
	How can the autochthonous theory then deal with archaisms found in
Iranian that are not found in Vedic? Such archaisms ought to have been
preserved in Vedic; they must have been forgotten (just like the tree names
mentioned above) all over the subcontinent when the Iranians supposedly
left it. Such collective amnesia, and in addition, one restricted just to
certain archaic items does not make for a good case. It is, again, one of
very special pleading.
	It should also be mentioned in passing, that if  the Iranians
emigrated from India, why we do not find ''Indian bones'' or genes of this
massive emigration in Iran and beyond? Indian skeletons are, as Kennedy
informs us (1995), remarkably different from Near Eastern ones.[N.130]
Again, indigenists would have to argue that only that section of the Panjab
population left westwards which had basically 'non-Indian' physical
characteristics, very special pleading indeed. To adopt an OIT stance
precisely mirroring the Indo-Aryan immigration theory based on 'trickling
in' is not possible as this 'trickling out' would comprise all subfamilies
of IE, from Tocharian to Celtic, and would constitute a much more massive
emigration.
	The IE theory can explain the materials found in the various
languages much more satisfactorily: the Iranian languages simply miss the
Indianization of IE, just as the very conservative Old Icelandic or
Lithuanian escaped the 'Christianization' and 'Europeanization' for a long
time.

§14.  Date of Indo-Aryan innovations

	As has been mentioned, the linguistic innovations of Vedic Sanskrit
are supposed by autochthonists to have taken place only after the Iranians
(and other Indo-Europeans) had left the subcontinent (Elst 1999: 122,124
sqq). It is difficult to argue against this kind of  assumption on general
linguistic grounds as language changes cannot easily be tied to certain
areas, unless there is evidence from inscriptions and clearly localizable
texts. However, the distribution of IE dialect features mentioned above
(Hock 1999) makes IE innovations after an Iranian/IE exodus from India
unlikely;[N.131] for, even though the old Satem innovations include Vedic,
they exclude Latin, Greek, Tocharian,  etc.
	Further, a good indicator is found in IE plant and animal names
(''willow'', etc.) and especially in the word for the horse drawn chariot,
Sanskrit ratha, O.Iran. ratha.  This word is attested in the oldest IIr
texts, in the RV and in the Avesta, also with the secondary formation Ved.
rathin-, O.Av. rathI 'the one who has a chariot, charioteer'. Even more
tellingly, it appears in the inherited, archaic compound, with a locative
case ending in its first member, RV rathe-STha, Avest. rathaE-s'ta-
'charioteer' (cf. also savyeSTha 'warrior').
	As the autochthonous theory would have the RV at c. 5000 or,
according to some, before the start of the Indus civilization at 2600 BCE,
the Iranians or other Indo-Europeans should have exported the chariot from
S. Asia at that time. But the chariot is first found in a rather archaic
form ('proto-chariot'), betraying its origin in a ox-drawn wagon (anas,
*weg'h-o- > wagon, veh-icle), at c. 2000 BCE, in Russia and at Sintashta,
W. and E. of the Urals.  As its invention is comparatively late, the
western IE languages retain, not surprisingly, the older meaning of the IE
word, *roth2o-''wheel'' (Lat. rota, Germ. Rad 'wheel'); they simply have
moved away, before this development took place, from the original central
IE region (such as the Ukraine) westwards into Europe.[N.132]
	The indigenist counter-argument could maintain that the newly
introduced chariot spread quickly from the Near East or Central Asia all
over the Iranian and Indian world, with its IIr name, *ratha. It would thus
belong only to a secondary historical level (after that of the earlier
"Panjab Indo-Europeans"). This argument, however, would run into a number
of difficulties: for, strangely, the word in its new meaning of 'chariot'
never reached the neighboring  Proto-Slavic tribes, nor the other European
'emigrants' (Grk. has  ha'rma/harmatos, Latin currus, curriculum, rota) on
the western side of Eurasia while it is known to the close neighbors, the
(Northern) Iranians. Worse, the word and the object are found already in
the RV (supposedly a text of pre-Indus age, 2600 or c. 5000 BCE!), well
before its invention.[N.133] In short, multiple insurmountable
contradictions emerge.
	The word cakra 'wheel' may be a much older adaptation from Sumerian
gil-gul 'wheel' and GIS'gi'gir 'wagon,' to IE *kwe-kwl-o- > IIr. cakra (or,
it is derived from a common origin, Littauer and Crouwel 1996). However,
the newly specialized meaning ratha ''chariot'' is restricted to IIr.; its
archaeological attestation puts PIIr, again, close to the Urals. -- On the
other hand, there are common PIE words  for the cart or four-wheeled wagon
(anas) and its constituent parts, such as and akSa 'axle', ara 'spoke,
pin', nabhya 'nave', yuga 'yoke', razmi, razanA 'reigns', etc.; for details
see EWA, s.v.  They are much older, PIE,  as they refer to the more
primitive technology of solid wheel wagons and carts that was developed in
Mesopotamian in the late 4th millennium.
	In sum, if according to the autochthonous theory, the Iranians had
emigrated westwards well before the RV (2600/5000 BCE), how could both the
Indians (in the Panjab) and  Iranians (from the Ukraine to Xinjiang) have a
common word for the horse drawn chariot as well as a rather ancient word
for the charioteer? Both words must have been present at the time of the
Indo-Iranian parent language. As the linguistic evidence shows, the
technical innovation was already Indo-Iranian (note Proto-IIr. dental *th
that regularly developed to > Ir. interdental th, as in OIran. ratha), and
it must have happened at the place of its invention, in the plains near the
IIr. River rasA (Volga), certainly not in the Panjab.
	Consequently, the occurrence of ratha/ratha in IIr. at c. 2000 BCE
shows that its import was carried out, along with many other IIr. items of
culture and religion, from the S. Russian/Central Asian steppes into the
subcontinent, and not vice versa. This is one of the few clear cases where
we can align linguistic innovation with innovation in material culture,
poetics and myth, and even with archaeological and historical[N.134]
attestation. Therefore, we have to take it very seriously. Anyone of the
various revisionist or autochthonous dating schemes that circumvent this
innovation in technology and language dealing with the horse drawn,
spoke-wheeled chariot at c. 2000 BCE  is doomed to failure.
	Other (theoretically) possible scenarios such as an import, along
with that of the horse (see below), from some (N.) Iranians near the Urals
into the area of the Indo-Aryans who had supposedly remained stationary in
the Panjab, run counter to the archaic formation of the words concerned
(ratheSTha, savyeSTha) and the clearly secondary, inherited form in Iranian
(ratha-), and would amount, again,  to very special pleading.

	Likewise, the many linguistic archaisms in Old Iranian cannot
readily be explained by a supposed  Iranian emigration from India. The Old
Avestan of zarathus'tra frequently is more archaic than the RV and
therefore too archaic to have moved out of India after the composition of
the RV (supposedly, before 2600/5000 BCE). For example, the Avestan
combination within a sentence of  neuter plural nouns  with the singular of
the verb is hardly retained even in the other older IE languages.
Conversely, something not found in Iranian, i.e. the  Rgvedic perfect forms
jabhAra or mene, are a local IA innovation. All of this points to
separation of Proto-Iran. and Proto-OIA at some time before the RV. Also,
it cannot have happened inside S. Asia as the Avesta lacks all those
typically S. Asian words that are local loans into Vedic (§16; Witzel
1999a,b). Incidentally, the lack of S. Asian substrate words in Iranian
(cf. Bryant 1999) also explains why the archaic Iranian traits cannot have
been preserved in the Panjab, side by side with the RV, before the supposed
Iranian move westwards.[N.135]
	One can only conclude that Proto-Iranian (> Avestan, O.Persian)
split off from IIr and thus, from pre-Old IA. (> Vedic, Mitanni IA, etc.)
at an early date, and definitely so while spoken outside the Panjab.
Because of the early split, Old Iranian preserved some archaic features,
while also developing innovations on its own (Iran. x < IIr kh, h < s,
etc.). In sum, Proto-Iranian never was spoken in the Panjab.
	Or, to give another example, according to the autochthonous theory,
Proto-Ir. would have to had to leave the Panjab before the Vedic dialects
of the RV took over (or developed) the so-called retroflex (mUrdhanya)
consonants.


§15.  Absence of retroflexes in Iranian

	While the feature of retroflexion (T, Th, D, Dh, S, N)  is
sporadically found also in some other parts of the world (Hock 1986), such
as in Scandinavia or Australia (innovative in both cases), it is typical
for S. Asia when compared to its neighboring regions, that is Iran,
West/Central Asia, the Himalayas, S.E. Asia.[N.136]
	In the autochthonous scenarios discussed above,  the hypothetical
emigrants from India would have lost the S. Asian ''bending back of their
tongues'' as soon as they crossed the Khyber or Bolan Passes:[N.137] not
even Old Iranian (East Iran. Avestan) has these sounds. But, conversely,
the Baluchi, who originally were a W. Iranian tribe, have acquired
retroflexion -- just in some of their dialects -- only after their arrival
on the borders on the subcontinent, early in the second millennium CE
(Hoffmann 1941, cf. Hock 1996, Hamp 1996). The same happened to other late,
incoming groups such as Parachi, Ormuri (from W. Iran) that are found in E.
Afghanistan, and also to some local Iranian Pamir languages such as Wakhi.
Clearly, retroflexion affects those moving into the E. Iranian
borderland/Indus plain. Importantly, the most widespread  appearance of
retroflexes is among the cluster of Hindukush/Pamir languages, that is the
languages surrounding these mountains in the east (Nuristani/Kafiri,
Burushaski, Dardic and the rest of these northernmost IA languages) as well
as in the north (some of the Iranian Pamir languages: Wakhi, Yigdha,
Sanglechi, Ishkashmi, Khotanese Saka), as detailed by Tikkanen (in Parpola
1994: 166). Retroflexes may also have belonged to a part of the Central
Asian/ Afghanistan substrate of the RV (Witzel 1999a,b). Retroflexion
clearly is a northwestern regional feature that  still is strongest and
most varied in this area.
	Had retroflexion indeed been present in the pre-Iranian or the
Proto-Iranian coeval with the (Rg)Vedic period, its effects should be
visible in Old Iranian, at least in Avestan[N.138] which was spoken in East
Iran, that means in part on the territory of modern Pashto (which has
retroflexes indeed).
	Cases such as IIr *waj'h-tar  > Av.  vas'tar, but > Ved. voDhar-
are clear enough and present perhaps the best testimony for the several
stages of conditioned reflexes in the development from IE to Vedic:  a
change from Ved. voDhar-  --> Avestan vas'tar-  is plainly impossible in
any version of phonetics, as also voDhar- --> IE *wek'h-tor- (as  in Latin
vec-tor): missing consonants as in vo-Dhar- do not suddenly (re-)emerge out
of the blue in other languages, and nota bene: not as a phonetically
changed -s'- in Iranian, as -k- in Latin, or as -k- in Gaulish Vectur-ius,
or as -g- as in Engl. wagon; rather, with the IE theory, they all stem from
<  IE *weg'h-tor- (neglected by Misra 1992).
	The case of voDhar- is pre-conditioned by the development of IE k',
g' > IIr c', j', which  changed to Proto-Iran. and Pre-Vedic s', z', then
to early Vedic retroflex .s.', .z'., which only then could influence the
following consonant (of the -tar suffix), as to deliver the retroflex
'suffix' -Dhar-. At this stage, the same retrograde Sandhi  as seen in
budh+ta > buddha took place (.z'h.-da > .z'.dha),  and only then, the
voiced sibilant .z'. disappeared, normally (as in lih: li.z'.Dha > lIDha)
with compensatory lengthening of the preceding vowel; but, in the
particular environment of voDhar (a.z'. > o, just as az > e)  represented
by o + retroflex consonant (-tar suffix), in short:
IE *weg'h + ter   > IIr * vaj'htar-	 > pre-Ved.  *va.z'.Dhar-[N.139]  >
Ved.    voDhar-
				     	 > pre-Iran. *vas'tar-      	 >
Avest. vas'tar-
In sum, the well-known rules of IE sound changes explain the development
from the root vah (IE *weg'h) without problem, while an OIT theory would
have great difficulty to get from voDhar- to any Avestan,  Latin, English,
etc.,  forms.
	Again, it is important to stress that retroflexes have not occurred
in (Old) Iranian, which has kept the older sound sequences. In addition,
these changes allow a relative and even an absolute dating:  *az'dh > oDh
is parallel to *sazd- > sed, i.e. both are post-Indo-Iranian and even
post-Mitanni; as pointed out above, Mitanni OIA keeps the sequence azd. In
other words, Rgvedic is younger than the Mitanni words preserved at c.
1450-1350 BCE. At any rate, RV -ed- is definitely younger than the Mitanni
forms because the IIr form *sazdai > Ved. sede  (3 sg. perf., cf. Avestan
hazde) 'he has sat' has already spawned a number of analogical formations
in the RV which are not conditioned by -azd-. These are found  even in the
older sections of the  RV:  yam > yem:  yemuH 4.2.14, pac > pec:  pece
4.18.13 etc.[N.140]
	In all the cases detailed above, the retroflex is a late, i.e. a
Vedic innovation that is not shared by Iranian and the other IE languages.
In short, the innovation is rather low down on the 'family pedigree', in
cladistics. Any biologist would classify a similar development in
biological materials as a clear indicator of a late development, as an
innovation, -- in case, one  that separates IA from the rest of IIr and IE.
In other words, Vedic Sanskrit does not represent the oldest form of IE as
autochthonists often claim.

	The adherents of the autochthonous theory would again have to take
recourse to special pleading, arguing that retroflexion occurred only after
the Iranians had supposedly left (i.e., well before the RV, at 4-5000 or
2600 BCE), or while they were living in some area of the Panjab untouched
by this phenomenon. This individual argument is, again, not a priori
impossible. But, it is not admissible on other grounds, such as the
occurrence of local loan words in Vedic. These have been taken from the
Panjab substrate (Witzel 1999a,b) that has unconditioned retroflexes (such
as in vANA, vINA, etc.), and these substrate words are, again, missing in
Iranian.[N.141]
	Retroflexion in Vedic must have been a regional feature, acquired,
just as it was by the Pashtos and the more recently arrived the W. Iranian
Baluchis, at the time of immigration.
	In sum, retroflexion affects all those moving into the E. Iranian
borderland, the Indus plain and the subcontinent. but this does not work
vice versa:  those who move out of India, sooner or later, loose it.
However, if this would be taken as proof of OIT, it does not work at all:
this particular development does not help to explain words such as Ved.
voDhar- which cannot turn into  Iran. vas'tar-, Latin vector, etc.[N.142]
The same conclusion can be reached when studying local Panjab loanwords in
the RV.


§16. Absence of 'Indian' words in Iranian

	As has been underlined several times, the hypothetical emigrants
from the subcontinent would have taken with them a host of ''Indian'' words
-- as the Gypsies (Roma, Sinti) indeed have done. But, we do not find any
typical Old Indian words beyond S. Asia, neither in the closely related  in
Old Iranian, nor in E. or W. IE, except for the usual words of culture
(Wanderwoerter) such as some recent imports into English (orange, tea/chai,
or curry, punch, veranda, bungalow), or the older ones of the type rice,
beryl, hemp, etc.[N.143] One would expect 'emigrant' Indian words such as
those for lion (siMha), tiger (vyAghra AV+, pRdAku AV+, zArdUla MS+,
puNDarIka lex.),[N.144] elephant (gaja Manu+ ibha RV?, kuJjara Mbh.+),
leopard (dvIpin AV+, Ep., citra-ka, etc. lex.), lotus (padma, kamala,
puNDarIka), bamboo (veNu), or some local Indian trees (azvattha, zamI,
bilva, jambu), even if some of them would have been preserved, not for the
original item, but for a similar one (e.g. English [red] squirrel > N.
American [gray] squirrel). Instead of Indian words we find, e.g., for siMha
'lion'  new formations : Iran. s'er, Grk. lIs, Lat. leO(n) (cf. Witzel
1999a,b), and similarly, Gr./Latin ones for 'tiger', 'lotus'.  Many of them
come from a Mediterranean/Near Eastern substrate, but not as expected in
any OIT scenario, from the S. Asian one visible in Vedic.
	In sum, no typical Indian designation for plants or animals made it
beyond the Khyber/Bolan passes. The only clear exception would be the birch
tree, whose IE name *bhRg'ho- is found all the way from India[N.145] to
Europe: Ved. bhUrja KS+, Ir. Pamir dial. furz, Shugni vAwzn < *barznI;
Osset. boers(oe);  Lith. be'rz'as, Serbo-Croat. bre`za; German Birke, Engl.
birch, etc. (cf. §12.6, n.113). The other 'European' trees that are found
in the northwest of the subcontinent, and beyond up to Russia/Urals, are
absent from Sanskrit vocabulary.[N.146]
	This situation has been well explained by the assumption of IE
linguists that these European/Caucasus/Ural tree names were remembered
(sometimes, in the Central Asian steppes and deserts, only in old sayings
or in poetry?) down to the very doorsteps of South Asia in Afghanistan, or
were applied to similar items, but were utterly forgotten in the tropical
S. Asia as there were no similar trees to which these IE names could be
applied. One apparent exception, vetasa, can easily be explained by a
transfer of meaning, from the very pliable (Afghan) 'willow' twigs to the
equally pliable 'reed, cane' (see above).[N.147]
	The autochthonous theory again must introduce the improbable
auxiliary assumption that all such words have been forgotten inside the
subcontinent after, or even as soon as, the Iranians (and other
Indo-Europeans) supposedly had crossed the Suleiman Range and the
Khyber/Bolan passes into Afghanistan and Iran.
	However, many if not most S. Asian plant and animal names have
clear, non-IE  local origins; in other words, they are loan words from the
local S. Asian languages[N.148] (e.g., RV mayUra 'peacock', vrIhi 'rice',
etc.). Others are new formations, built on the basis of IE words, e.g.
'elephant':  hastin (+ mRga) RV 1.64.7, 4.16.13 etc., 'the (wild animal)
with the hand, the elephant', used for words such as Late Ved. gaja, ZB
14.4.1.24 mataGga, Epic nAga, RV(?) ibha.[N.149] Or 'tiger', vyAghra < 'who
tears apart?' (KEWA III 274), 'who smells scents by opening [his jaws]'(?)
EWA II 593,  for VS zArdUla, puNDarIka (lex.), (note also N.Pers. bebr).
These new formations must have been introduced when the immigrating
speakers of Indo-Aryan (again, not the Iranians!) were first faced with
them in the Greater Panjab.  Indigenists (Talageri 2000, Elst 1999, etc.)
denounce such cases as just one more of the common substitutions based on
poetic or descriptive formations, or as dialect designations which can
happen at any stage in the history of a language (e.g. Vulgar Latin
caballus > French cheval, etc. for older equus). However, such critics once
again overlook the wider complex, the complete absence of original  IE/IA
words for S. Asian plants/animals built with clear IE roots and/or word
structure. The absence of IE/IA words for local plants and animals clearly
militates against any assumption that Pre-IA, Proto-IIr or PIE was the
local language of the Panjab or of Uttar Pradesh during (pre-)Harappan
times.
	This also agrees with the fact that most of the S. Asian loan words
in the Rgveda, excluding some Central Asian imports, are not found in Iran
and beyond.[N.150] These words include Kuiper's (1991) c. 380 'foreign
words' in the RV. Again, not all of them could have been lost as soon as
the hypothetical  IE or Iranian  emigrants crossed over into Iran and
beyond. One would at least expect a few of them in the 'emigrant'
languages. Such Indian words should have survived in the west and could
have acquired a new meaning, such as British Engl. corn  'wheat' > 'maize'
in America. The Gypsies, after all, have kept a large IA vocabulary alive,
over the past 1500 years or so,  during their wanderings all over the Near
East, North Africa and Europe (e.g., phral 'brother', pani 'water', kara'l
'he does').
	No amount of special pleading will convince an independent
(linguistic) observer of a scenario that relies on the total loss of all
typical S. Asian words in Iranian and all the other 'emigrant'
Indo-European languages. Again, Occam's razor requires to scrap the theory
of an 'Aryan' or, worse, an Indo-European emigration from the Panjab to the
West.

§17. IE words in Indo-Iranian; IE Archaisms vs. Indian innovations

	Conversely, and not unexpectedly by now, typical IIr. words
indicating a temperate climate, and with IE root and suffix structure, such
as 'wolf' (vRka: Avest. v@hrka; cf. Lith. vilkas, O.Slav. vl'ku, Alban.
ulk, Grk. lu'kos, Lat. lupus, Gothic wulfs < *wlkwos), 'snow/winter' (hima:
Avest. zim/ziiam, Grk. xiOn 'snow', -khimos,  Lat. hiems, Gaul. Giamon-,
Armen. jiun  'snow', etc.), 'birch tree' (bhUrja, Pamir Dial. furz, Osset.
boers(oe), etc. are found in E. Europe, Greater Iran and on the
northwestern borders of the subcontinent (Kashmir).  However, neither snow
nor birch are typical for the Panjab or Indian plains. It is, again,
theoretically possible that these words belonged to the supposed original
IE/IA vocabulary of the northwestern Himalayas and therefore could have
been transported westward by a hypothetical IE westward emigration. But,
this scenario is contradicted by the evidence of the last section dealing
with all the other IE 'cold climate' words that have not been preserved in
India, not even in the Northwest or in the Himalayas. Therefore, words such
as those for 'wolf' and 'snow' rather indicate linguistic memories of a
colder climate than an export of words to Iran and Europe, such as that for
the high altitude Kashmirian birch tree.
	More importantly, typical Indian grammatical and lexical
innovations are not found among the other Indo-European languages. While
some, stemming from the IIr period, are met with in Old Iranian (pronoun
ah-am 'I', Avest. az@m; Nom.Pl. azvAsa-as, Avest. aspa^onghO, etc.),  the
typical Indian innovations found already in the RV (jabhAra for jahAra,
sede/mene, absolutives, etc.) are not. The first type of innovation is
attributed to the common source language, i.e. Indo-Iranian rather than OIA
influencing the neighboring Old Iranian.[N.151] It would be against all
rules of (IE and non-IE) comparative linguistics to assume that such late,
(low-level, in term of family tree or cladistics) developments  should not
apply just in the single case of Indo-Aryan, and to assume, instead, early
innovation inside India (azvAs-as, ratha, babhru 'mongoose', etc.) that
would have selectively been exported to Iran (of course, minus all *typical
Indian* RV innovations!), innovations that would not have been carried out
in the rest of the Indo-European languages: just too many auxiliary
assumptions!
	The autochthonous theory would, again, have to assume that all such
Indian innovations would have been carried out after the speakers of
Iranian (and/or all other Indo-European languages)  had left the
subcontinent, which is contradicted by absence of typical  Indian words in
other Indo-European languages and in Iranian, and by the absence further
west of Indo-Iranian innovations such as the chariot (*ratha).  Occam's
razor applies again.

	To go into some further detail, the many archaisms in Old Iranian
cannot readily be explained by an Iranian emigration from India: First of
all when and where should this have happened? SW and Central Southern Iran
was occupied by the Elamians, the western parts were settled by W. Iranians
only after c. 1000 BCE (cf. Hintze 1998) and were settled by non-IE peoples
before. About E. Iran/Afghanistan we have only stray Mesopotamian, copious
archaeological and a few isolated Vedic sources. They point to non-IE
settlements as well: in S. Iran,  Elamian up to Bampur, Meluhhan east of it
in Baluchistan/Sindh, and Arattan north of it in Sistan, while the northern
fringe was occupied by the Bactria-Margiana substratum that is visible in
Indo-Iranian (Witzel 1999a,b).
	If the Iranians had moved out from the Panjab at an ''early date'',
they would have missed, the supposed 'Panjab innovation' of the use of the
(domesticated) horse (already  Indo-European: Latin equus, etc.), and
especially the later one of the horse-drawn chariot (IIr. ratha).  If, on
the other hand, they had moved out a little later, say, after the Mitanni
Indo-Aryans, all of this would have come too late to account for the
non-appearance of Iranian tribes in the RV which has only some
(pre-)Iranian looking names (Witzel 1999),   camels (RV 8) and some Afghani
rivers (gomatI in the Suleiman Range, sarayu in Herat, sarasvatI in
Arachosia). We cannot make the Iranians move from India to Iran, say, at
5000 or 2600 BCE, then to introduce the innovation of horse pastoralism
(not present in the subcontinent then!), and then let them take part, at c.
2000 BCE, in the innovation of the already IIr horse drawn chariot (*ratha,
§ 12.6, §21).
	In addition, Old Iranian in general is too archaic to have moved
out of India after the composition of the RV: while Old Avestan (of
zarathus'tra) has, to be sure, many forms which correspond to Rgvedic ones,
much of his language is even more archaic: as has been mentioned, the
retention of the use of neuter plural with singular of the verb is
something that has elsewhere been retained in Hittite; the old nom. pl.
masc. in -As = Avest.  -As-, -A- is found in the RV next to the innovation
devAs-as; an archaism in the perfect stem which appears in the RV such as
babhr- (Avest. bawr-) next to the new formation RV jabhr-; archaisms in
names such as jamad-agni (= Avest. jimat~) next to the innovative  RV
gamad,  etc.
	All of this points to a time of separation of IA and Iranian before
the RV and thus, not inside India. The hypothetical argument that these
traits were preserved in the Panjab side by side with the RV does not hold,
for Iranian does not show any typical  Indian elements (see above).[N.152]
	If the Iranians had indeed left the Panjab before the RV, serious
chronological difficulties would arise, whether we were to accept the
autochthonous theory of the RV well before  the Indus civ. (2600/5000 BCE)
or whether we accept the traditional Indologist's dating of the RV sometime
in the 2nd mill. BCE.  In all these cases, Iranian is far too archaic to
have been a close neighbor, in the Panjab,  of the Rgvedic dialects.
Further, it lacks any indication of Indian influence on its grammar and
vocabulary  (see above).
	One can only conclude that Old Iranian, including Avestan,  split
off from (Proto-)Old Indo-Aryan (Vedic, etc.) at an early date, preserved
some archaic features while developing innovations on its own (s > h, kh >
x, j'n > sn,  etc.) and that it was never in early close contact with the
Panjab and its substrate languages. Such close contact would also have
effected the one typical phonetical development that the  Iranians actually
'escaped' before the Vedic dialects of the RV adopted or developed it,  the
retroflex sounds (see above §15).


§18. Absence of Indian influence in Mitanni-Indo-Aryan

The same scenario as discussed so far is indicated by the IA loan words in
the Hurrite language of the Mitanni realm in northern Iraq/Syria (c.
1460-1330 BCE). Again, if there was an (early) emigration out of India by
(Vedic) Indo-Aryans it would be surprising that even the Mitanni documents
do not show typical South Asian influence.[N.153]
	Rather, is obvious that the remnants of early IA in Mitanni belong
to a pre-Rgvedic stage of IA, as is seen in the preservation of  IIr -zdh-
> Ved. -edh-, in priyamazdha (bi-ir-ia-ma-as'-da[N.154]) : Ved. priyamedha
: Avest. -mazdA. These texts also still have IIr  ai > Ved. e (aika : eka
in aikavartana). Another early item is the retention of IIr. j'h > Ved. h
in vas'ana(s')s'aya 'of the race track' = [vaz'hanasya] cf. Ved. vAhana-
(EWA  II 536, Diakonoff 1971: 80, Hock 1999: 2); they also share the
Rgvedic (and Avestan) preference for r (pinkara for piGgala, parita for
palita). Importantly, Mitanni-IA  has no trace of retroflexion.
	How could all of this be possible if one supposes an emigration
from India, in some cases (Misra 1992) even after the supposed date of the
RV (5000 BCE)? The RV is, after all, a text that already has all these
features.
	The Mitanni loan words (Mayrhofer 1979, EWA III 569 sqq.) from
Pre-Vedic OIA share the typical IIr innovations, such as the new Asura gods
varuNa (EWA  II 515 a-ru-na, u'-ru-wa-na, not found in Iran) and mitra
(Avest. mithra, Mitanni mi-it-ra), and indra (Mit. in-da-ra/in-tar, Avest.
iNdra)[N.155] who is marginalized in Iran, and the nAsatya
(na-s'a-ti-ya-an-na = azvin, Avest. na^onghaithiia).[N.156] These
innovations also include the  new the concept of Rta (Iran. arta, in very
late Avest. pronunciation = aS~a), contained in names such as artasmara
(ar-ta-as'-s'u-ma-ra), artadhAman (ar-ta-ta-a-ma),[N.157] and perhaps also
the newly introduced ritual drink, sauma, IIr *sauma (Ved. soma, Avest.
haoma, EWA II 749). The Mitanni sources  show extensive use of the
domesticated horse (as'uua, cf. names for horse colors[N.158]), the chariot
(rattas') and chariot racing (a-i-ka-, ti-e-ra-, pa-an-za-, s'a-at-ta-,
na-a-[w]a-wa-ar-ta-an-na= [aika-, tri-, panca-, satta- (see n.160),
nava-vartana]; tus'ratta/tuis'eratta = RV tveSaratha).
	To see in these names a post-RV form of OIA, a Prakrit  (Misra
1992, Elst 1999:183),[N.159] is therefore misguided and based on
insufficient knowledge of near Eastern languages. Misra's 'prAkRtic
influences' in Mitanni IA are due to the peculiarities of the cuneiform
writing system and to the Mitanni form of the Hurrite language. It has been
asserted for long that satta in satta-vartana 'seven turns' has been
influenced by Hurrite s'inti 'seven' (J. Friedrich 1940, cf. Cowgill 1986:
23, Diakonoff 1971: 81; this is under discussion again,[N.160] but clearly
a Hurrite development); however, the words starting with b- such as bi- did
not receive their b- from a MIA pronunciation of vi,[N.161] as Misra
maintains, but are due to the fact that Mitanni does not allow initial v-
(Diakonoff 1971: 30, 45). In sum, the Mitanni IA words are not Prakritic
but (pre-)Rgvedic.
 	On the other hand, the Mitanni texts clearly indicate  typical OIA
(Vedic) linguistic innovation: aika-vartana (a-i-ka-ua-ar-ta-an-na)[N.162]
instead of Ir. aiva- or general IE *oino- > *aina-), and yet, the
vocabulary does not yet show signs of typical South Asian influence: for
example, there is no retroflexation in mani-nnu, Avest. maini, Elam O.P.
*bara-mani, and Latin monIle. But retroflexation is precisely what is found
once OIA enters South Asia: RV maNi 'jewel'.[N.163] Finally Mitanni IA has
no typical South Asian loan words  such as ANi  'lynch pin'.
	In sum, Mitanni-IA is older than the RV, cannot have come from the
Panjab but must have been spoken in the north-eastern border areas of
Mesopotamia where it influenced the  Hurrite language of the Mitanni that
belongs, just like its later relative Urartu, to the Caucasus group of
languages.
	Indeed, some of the rather indirect IA influx into the Near East
may have been earlier than the one visible in Mitanni. The Kassite
conquerors of Mesopotamia (c. 1677-1152 BCE) have a sun god
s'uriias',[N.164] perhaps also the marut and maybe even bhaga (bugas'?), as
well as the personal name abirat(t)as' (abhiratha); but otherwise, the
vocabulary of their largely unknown language hardly shows any IA influence,
not even in their many designations for the horse and horse names[N.165]
(Balkan 1954).[N.166]

	If one now thinks through the implications of the autochthonous
theory again, the ancestors of the Mitanni Indo-Aryans would have left
India very early indeed (well before their favorite date of the RV,
2600/5000 BCE, and well before 1900 BCE, the supposed date of the brAhmaNa
texts, Kak 1994). They would have done so with the Rgvedic dialect features
(ai > e, zdh > edh) not yet in place, and without any of the alleged MIA
forms of Misra (satta, etc.), but with the typical OIA and IIr terms for
horses and chariot racing (before their invention and introduction into
South Asia)! They would have lingered somewhere in N.W. Iran to emerge
around 1400 BCE as Hurrianized Mitanni-IA, with some remnant IA words and
some terms of IA religion. But they would have done so without any of the
local South Asian innovations[N.167] (no retroflex in mani-, no -edh-, -h-,
etc.) that are already found in the RV, and also without any particularly
Indian words (lion, tiger, peacock, lotus, lynch pin ANi) all of which
would have been 'selectively' forgotten while only typical IA and IE words
were remembered. In short, a string of contradictions and improbabilities.
Occam's razor applies again.
	Similarly, the parna (Gr. parnoi, Ved. paNi) and dasa/dAsa ~ Avest.
(az'i) dahAka, ~Ved. dAsa ahIzu, Lat. dahi, Grk. daai, Avest. da^ongha (::
airiia, cf. dahae :: arii), would have escaped their Panjab IA enemies (RV
dasa, dasyu, paNi :: ari, arya, Arya) northwards  in order to settle at the
northern fringes of Iran well before the time of the  RV, e.g., as the
parna, still without retroflexion and accompanying loss of -r-.
Unfortunately for the autochthonous theory,  these N. Ir. tribes occur
already in the RV, significantly  not as real life but as mythical enemies,
and now with retroflexion. Significantly, all while the same authors who
composed the RV hymns are supposed by the indigenist and revisionist
writers not to remember anything beyond the Panjab. Again, multiple
contradictions: Occam's razor applies.

***

Summary: Linguistics

	In sum, all of the linguistic data and the multitude of possible
autochthonous scenarios based on them lead to the same kinds of culs de sac
or Holzwege.
	There is no evidence at all for the development of IE, IIr, and
even of pre-OIA/Vedic inside the subcontinent. It is contradicted, among
other items, by the Iranian and Mitanni evidence. An emigration of the
Iranians and other Indo-Europeans[N.168] from the subcontinent, as supposed
by adherents of the autochthonous theories, is excluded by the linguistic
evidence at large.
	To maintain an Indian homeland of IE, IIr, and Pre-OIA requires
multiple special pleading of a sort and magnitude that no biologist,
astronomer or physicist would tolerate. Simply put, why should we allow
special, linguistic pleading just in the case of India? There is nothing in
the development of human language in India that intrinsically differs from
the rest of the world. Occam's razor applies.

	So far, most of  the linguistic evidence presented in the previous
sections has been neglected by advocates of the autochthonous
theory,[N.169] and if it has been marshaled at all, it has been done so ad
hoc, even by the lone, autochthonously minded Indian historical linguist,
S.S. Misra. His rewriting of IE linguistics remains incidental and
idiosyncratic, and it results in multiple contradictions, just as the rest
of the theory. The autochthonists must do a lot of  homework and try to
contradict the linguistic data discussed above (detailed in § 13-18) before
they can hope to have any impact on linguistic discussions.

	Conversely, the data derived from linguistic study are consistent
throughout: they clearly indicate that an Eastern IE language, the Vedic
branch of IIr, has been Indianized and has grammatically innovated after
its arrival in the Panjab, while Iranian has escaped this influence as it
did not  enter the subcontinent then. Exactly how the IA language and the
IA spiritual and material culture of the archaeologically still little
traced Indo-Aryan speaking tribes was introduced, that is still an open and
very much debated question. It can be traced securely, so far, only in the
evidence coming from the texts (horses, chariots, religion, ritual,
poetics, etc.) and from the features of the language itself that have been
discussed here at length. Possibly, genetic evidence, especially that
deriving from studies of the male Y chromosome, may add to the picture in
the near future.

	In the sequel, the evidence from texts, archaeology, and some
natural ("hard") sciences will be adduced. This is perhaps the right place
to point out that these fields of scholarship proceed in their own fashion
and with various methodologies, and that the data obtained from all these
fields have their own characteristics. It is not always the case, for
example, that evidence from archaeology can flawlessly be matched with
linguistic or genetic data. The nature of evidence in these fields often is
too disparate. Some scholars (such as the archaeologist Shaffer) actually
refuse to take into account anything that is outside archaeology,
especially the "tyrannical" linguistics. This is of course not quite true,
as palaeontology is tacitly accepted. Second, it must be pointed out that
many of these fields, such as archaeology, provide "hard" evidence, but
then interpret their data in various ways, just as it occurs in the other
humanities. The same is true also, e.g. for studies of palaeo-climate. The
distinction between the 'hard sciences' and the humanities is not as strict
as is often made out.
	Nevertheless, we should keep looking for overlaps in evidence and
draw our own, often preliminary conclusions, -- preliminary as several if
not all of the fields involved are in constant development.

(continues  with sections  d- g)

========================================================
Michael Witzel
Department of Sanskrit & Indian Studies, Harvard University
2 Divinity Avenue, Cambridge MA 02138, USA

ph. 1- 617-496 2990 (also messages)
home page:  http://www.people.fas.harvard.edu/~witzel/mwpage.htm

Elect. Journ. of Vedic Studies: http://nautilus.shore.net/~india/ejvs/