From the dialects of Old Indo-Aryan to Proto-Indo-Aryan and Proto-Iranian

Preliminary note

The Aryan or Indo-Iranian languages are divided into two basic groups, Indo-Aryan (i.e. languages nowadays mainly spoken in India in its pre-1947 sense of South Asia) and Iranian (i.e. languages nowadays mainly spoken in Iran, also rather in the historical sense of the Persian Empire, which extended to Central Asia and the Indus Valley). As argued below, this dialectal split goes back to the very emergence of Proto-Aryan from Late Proto-Indo-European, far from India and Iran (see section 2.9). Although these terms are anachronistic and even otherwise unsatisfactory, ‘Proto-Indo-Aryan’ (applied to the Mitanni Aryan of Syria by Thomas Burrow in 1973) and the analogical ‘Proto-Iranian’ will be used for extra-Indian and extra-Iranian language forms (whether real or hypothetical) that are ancestral (or, as in the case of Mitanni, parallel) to Old Indo-Aryan and Old Iranian.

I am grateful to the organisers of the symposium for this opportunity to pay respect to the memory of Sir Harold Bailey, with whom I had a cup of tea and a pleasant chat several times, mostly in 1987, when I spent one term in Cambridge. In accordance with the title suggested by the organisers, my paper originally dealt with ‘Vedic and the entry of the Indo-Aryans into South Asia’ (sections 1 and 3). The language of the Mitanni Aryans, which is now fairly generally admitted to represent the Indo-Aryan branch (rather than the Iranian branch or Proto-Aryan), provides one of the few means of dating (approximately) the coming of the Indo-Aryans to South Asia, and as I had some new points to present in this connection, I afterwards extended the paper to include the Mitanni Aryans and related questions concerning the formation of Proto-Indo-Aryan and Proto-Iranian (section 2). My best thanks go to Nicholas Sims-Williams for accepting this extended version of the paper in spite of its length, for several valuable comments and for editing my English; and to Alexander Lubotsky and Jorma Koivulehto for their comments on the possible Proto-Indo-European background of the Vedic suffix -ās-as.

For reasons of space, a large part of the original manuscript for section 2 has been cut out to be presented as a separate study elsewhere. In that paper (Parpóla 2002b) I argue for the common descent of the following peoples, whom I see as representing the earliest wave of Aryans in Central and South Asia: (1) the Aryan élite superstratum which I assume to have first taken over the rule in the BMAC or ‘Bactria and Margiana Archaeological Complex’ and to have expanded from Afghanistan to the Indus and Ganges Valleys ca. 2100–1900 B.C.; (2) the Dāsas, Dasyus and Panis mentioned as their enemies by the R̥gvedic Aryans ca. 1400 B.C.; (3) the Vrâtyas of the Middle Vedic texts, whose militaristic and orgiastic cult anticipates the Sākta Tantrism of later Hinduism; (4) the earliest speakers of the Māgadhī Prakrit of eastern India; (5) the former ‘Kafirs’ of Nuristan in northeastern Afghanistan. An additional excuse for excluding these peoples from a paper primarily dealing with Indo-Aryans is that their language in my opinion originally belonged to the ancestor of the ‘Iranian’ branch, where their closest relatives appear to be (6) the Scythians of the Eurasian steppes, the Sakas of Khotan and the Wakhis of the Pamirs (see section 2.10.2).

Old Indo-Aryan, or Sanskrit in its broader sense, refers to the oldest textually attested forms of Indo-Aryan, i.e. the Aryan speech of India in the sense of South Asia. On a very general level, two successive phases of Old Indo-Aryan are distinguished: (1) Vedic Sanskrit, attested roughly from 1500 to 400 B.C., and (2) Classical Sanskrit, normatively defined by Pāņini in his grammar ca. 400 B.C., which is also the approximate date for the oldest layers of the Sanskrit Epics, the earliest literature in Classical Sanskrit.

Old Vedic, or the language of the R̥gveda, is datable to Late Bronze Age, which in South Asia ends about 1150–1050 B.C. (for this date, cf. Allchin 1995: 39, 330; Erdosy 1995a: 80, 83f.). This is because in contrast to the Atharvaveda, the Rgveda does not seem to contain any reference to iron but only to copper or bronze (cf. Rau 1974: 19–25; Schlerath1997: 819; Witzel1995a: 98). Michael Witzel, who has studied the formation of the Vedic corpus intensively in recent years, correlates the transition to the post-R̥gvedic Khila hymns and the Atharvaveda with the transition from the Bronze to the Iron Age ca. 1150 B.C. and takes this as the starting point for a retrospective chronological calculation of the dates of kings and other people mentioned in the R̥gvedic hymns (see his genealogical chart of early Vedic kings reproduced opposite as Table 1).

On this basis, and also taking into consideration R̥gvedic references to tribal history, Witzel (1995a,b; 1999a,b) proposes dating the R̥gvedic hymns to ca. 1700–1200 B.C., distinguishing three sub-periods:

Early R̥gvedic periodica. 1700–1500 B.C.), associated with the Vedic tribes Yadu and Turvaśa, and Anu and Druhyu. These tribes ‘are regarded as settled in the Panjab at the time of the arrival of the Pūrus and Bharatas’ (Witzel 1995b: 328).

Middle R̥gvedic period (ca. 1500–1350 B.C.), when the main bulk of the hymns was composed, focused on the immigration of the closely related tribes of the Pūrus and Bharatas, notably under the rival kings Trasadasyu and Sudās (see Table 1).

Late R̥gvedic period (ca. 1350–1200 B.C.), during which the Kuru tribe makes its first appearance with King Kuru ś;ravaṇa (descendant of the Pūru chieftain Trasadasyu), to gain supremacy under his descendant Parikṣit (first mentioned in the ‘apocryphal’ Khila hymns of the R̥gveda) (Witzel 1995a; 1999a).

The division of R̥gvedic history into three sub-periods, their correlation with the R̥gvedic tribes specified, and the absolute dates suggested by Witzel all agree well with my own reconstruction, based mainly on an analysis of the evolution of the Vedic religion and on correlation with archaeological cultures. I therefore accept the above historical outline as essentially correct. But I cannot endorse Witzel’s proposals concerning the compositional chronology of the R̥gvedic hymns. Witzel (1999a: overview) assigns to the early period ‘some hymns of books 4, 5, 6’, to the middle period ‘books 2, 4, 5, 6 + 3, 7, 8.1–66, 1.51–191’ and to the late period 8.67–103, 1.1–50, 10.1–84 and 10.85–191. This scheme will be criticised further on and an alternative layering proposed (see section 1.4).

Besides the dating of the hymns, the location of the R̥gvedic hymns is important, not least for discussions of their linguistic substrata, where I find reason for strong disagreement with Witzel (see section 3.2). The river names and other geographic references clearly suggest that the R̥gvedic tribes lived mainly in the Pan-jab and the surrounding areas from Afghanistan in the west to the upper reaches of the Ganges river in the east (cf. Zimmer 1879: 3–48; Hopkins1896: 87f.; Witzel1995a: 93; Scharfe1996: 356–63). As Witzel has emphasised, the majority of the R̥gvedic hymns centre on the immigration of the Pūru and Bharata tribes into the Panjab, particularly at the time of the rival kings Trasadasyu and Sudās; the latter eventually wins the famous ‘battle of the ten kings’ (R̥V 7,18) on the river Paruṣṇī, identified by Yāska with Irāvatī, the modern Ravi, one of the main tributaries of the Indus (cf. Zimmer 1879: 11).

King Sudās belongs to the main (second) period of the R̥gveda. But although he fought in the Panjab, either he himself or his father Divodāsa seems to have arrived there from Afghanistan. Because this fact is important for the reconstruction of the history of the R̥gveda and the immigration of the Indo-Aryans into South Asia (cf. sections 2.4; 2.6; 2.7 and 2.10.2), it is useful briefly to mention the principal evidence for this.

The first three stanzas of R̥V 6,61 make it clear that King Divodāsa was born in Arachosia, in southern Afghanistan. Here River Sarasvatī is said to have given the powerful Divodāsa as a son to Vadhryaśva, who worshipped her with offerings. Sarasvatī has seized from the [inimical] Paṇi [who is rich but does not offer Soma to Indra] his numerous cattle (lit. ‘enjoyment’); like a root-digging (boar) (cf. Hoffmann 1976: 387), she breaks through the back of the mountains with her roaring waves and she slays the Pārāvatas; she is asked to throw down the deva-revilers, every descendant of Br̥saya (lit. the descendants of every Br̥saya) in the possession of magic (māyā). The poet praises the Goddess who has given this riverine tract to his people.

As shown by Alfred Hillebrandt (1927: I, 510–515) and accepted by Hans Henrich Hock (1999b: 165f.) among others, the Sarasvatī mentioned here is not the stream in India (as maintained e.g. by Macdonell and Keith 1912:1, 363) but the river called Harax^vaitī in Avestan and Harauvatiš in Old Persian, situated in southern Afghanistan which was known (in Greek) as the province of Arachosia in the Persian Empire. In Alexander’s time, the satrap of Arachosia was a man called Bαρσαέντης (Arrian, Anabasis 3,8,3; 3,21,1), whose name resembles Br̥saya, while the Pā́rāvata (lit. ‘people from afar’) correspond to the people called Παρουν̑ται, whom Ptolemy (6,17) places in Areia (the province next to Arachosia in Afghanistan).

The implication of the two Sarasvatīs was pointed out by Thomas Burrow in a passage worth quoting at length because it contains other important conclusions as well:

The colonization of North-West India by the Indo-Aryans was an extensive operation, lasting over generations, which could only have been carried out on the strength of an extensive population base immediately outside the sub-continent. That is to say that before these migrations Proto-Indoaryans must have been in occupation of large tracts of eastern Iran and western Afghanistan (such as Bactria, Areia (Haraiva), Arachosia, and Drangiana), which only at a later period came into the possession of the Iranians. One would also not expect that the migrations into India left these countries empty of Proto-Indoaryans, but rather that this was a movement of the surplus population, so that when the Iranians took control of this territory they would find the Proto-Indoaryans settled there, and that in due course of time the latter would be absorbed into and merged with the later-coming Iranians.

A slight trace of the earlier home of the Indo-Iranians has been preserved in some river names which were transferred by them from the old to the new country. This is how the relationship between the the Indian Sarasvatī and Iranian Hara^xvaitī is to be explained. There is not, as Bartholomae suggests, a common inheritance from Proto-Aryan sarasvatī here, since…this region was not the home of the Proto-Aryans. On the contrary Sarasvatī is in the first place the Proto-Indoaryan name of the river in Iran [i.e. in Arachosia], which after the migration was transferred to the river in India. The Iranian name is a loanword from Proto-Indoaryan, with the substitution of h- for s-, occurring also in hindu-…Another case is the river name sarayu-, which was transferred first from Iran (haraiva-/harōyū-) to a river in North-West India, and then again from there to a tributary of the Ganges in eastern India. (Burrow 1973: 126)

Whether Vadhryaśva is the same person as Atithigva, and Divodāsa the son of Vadhryaśva the same as Divodāsa the father of Sudās, as assumed by Witzel (1995b: 319; see Table 1), is not absolutely certain but very likely (cf. Macdonell and Keith 1912:1, 363f.; appendix A in Witzel 1995b: 343–52).

The Middle Vedic period (ca. 1200–600 B.C.) is datable to early Iron Age and mostly pre-urban times. Linguistically and textually four successive subperiods can be distinguished according to literary categories, though roughly we may speak of ‘the Brāhmaṇa period’:

— mantra collections (Paippalāda- and Śaunaka-Atharvaveda; the Khila hymns of the R̥gveda; the Mantra portions of the Black Yajurveda Saṃhitās; and the Kauthuma and Jaiminīya Saṃhitās of the Sāmaveda);

— prose portions of the Saṃhitās of the Black Yajurveda (Saṃhitās of the Kaṭha, Kapiṣṭhala-Kaṭha, Maitrāyaṇīya and Taittinya schools);

— older Brāhmaṇa texts;

— younger Brāhmaṇa texts, Āraṇyakas and older Upaniṣads (the last mentioned being the Jaiminīya-Upaniṣad-Brāhmaṇa, the Br̥had-Āraṇyaka-Upaniṣad and the Chāndogya-Upaniṣad); and

— the oldest Sūtra texts (especially the Baudhāyana- and the Vādhūla-Śrauta-sūtra).

The earliest post-R̥gvedic texts appear to have been composed in the western half of the north Indian plains, from where the Vedic Aryans had been moving both eastwards into the Gangetic Valley and southwards, up to the Vindhya mountains. The ‘Kurukṣetra’ in the Gaṅgā-Yamunā Doab, the domicile of the powerful Kuru tribe, was the centre of Orthodox Vedic culture’ in the Middle Vedic period. The formation of the Vedic canon in the Middle Vedic period is a ‘historical’ process and as such outside the scope of this paper. It is also a very complex event, and I can do no better than refer the reader to the extensive, detailed and insightful synthesis of Michael Witzel (1997a; for the geographical location of the invidividual Vedic schools and texts, see Witzel 1987; 1989: 110–17; 1995a: 93f.).

Late Vedic is the language of the ‘Sūtra period’ (ca. 600–400 B.C.), when most of the Śrautasūtras and Gr̥hyasûtras were composed. However, the creation of the latest parts of the Vedic literature including the Dharmasutras and e.g. the Vaikhānasa-Śrautasūtra continued a good deal after 400 B.C.

It is only in the Sūtra period that the Vedic culture appears to have started moving to south India, where the Old Tamil texts dated to the early centuries of the Christian era refer to the performance of Vedic sacrifices (cf. Parpola 1984, with further references; 2002a).

Pāṇini composed his monumental grammar of Sanskrit in the northern Indus Valley around 400 B.C. (cf. Cardona 1976; Scharfe1977: 88). Among the earliest extensive texts in Classical Sanskrit is Patañjali’s ‘Great commentary’ on Pāṇi-ni’s grammar, datable to ca. 150 B.C. The Mahābhārata began to come into being about 400 B.C., although it is difficult to distinguish with certainty between the different layers of this enormous epic (of some 75,000 verses) that continued to grow until about 400 A.D. Epic Sanskrit is an early variety of Classical Sanskrit which often though not consistently deviates from Pāṇini’s rules. (See Brockington 1998.)

I see the origin of Epic/Classical Sanskrit as a central problem in research relating to the entry of Indo-Aryan speakers into India. It is clear that Epic Sanskrit goes back to bardic traditions that predate even the earliest parts of the Mahābhārata. Specimens of such ‘proto-epic’ poetry are preserved in verses quoted in the Middle Vedic and early Buddhist texts, where they are called either gāthā or śloka (see Horsch 1966; cf. Weber 1891; Renou1954; Witzel1989: 127f.; and below, section 1.5.2). Although Epic Sanskrit naturally has been much influenced by Vedic Sanskrit, it has some well known dialectal features which show that it entered South Asia separately from Old Vedic.

There are numerous dialectal features in Vedic as well as Old and Middle Indo-Aryan in general; recent publications on this topic include the large volume edited by Colette Caillât (1989), containing among other useful papers an extensive survey of the Vedic dialects by Michael Witzel (1989); the comprehensive survey of Middle Indo-Aryan by Oskar von Hinüber (1986; cf. also Norman 1995 and Caillât 1997); and several studies by Georges-Jean Pinault (esp. 1998; seesection 1.4.4). In the following, I would like to focus on just a couple of well-known features that allow important conclusions.

The original dialect of the main bulk of the R̥gveda (Middle R̥gvedic in the above scheme) had some systematic innovations not shared by Classical Sanskrit. Very important among these is rhotacism, by which term we here mean the merger of the Proto-Indo-European liquids *l and *r so that PIE *l was systematically replaced with *r and only this latter phoneme remains. This rhotacism has taken place also in Avestan and Old Persian. Epic/Classical Sanskrit, on the other hand, has in many words preserved the original PIE *l, and there are traces of its preservation in some Iranian dialects as well (cf. e.g. Ossetic lœsœg ‘salmon’; Mayrhofer 1989: 10) and in Nuristani (cf. Nelson 1986). These dialects must have separated from Proto-Aryan before it innovated with rhotacism.

However, words which retain the original PIE *l start to be introduced into the R̥gvedic language in its Late period, obviously from a language that had preserved it. In the youngest parts of the R̥gveda, l is eight times as frequent as in the oldest parts, where it is actually found in very few words, while in Epic and Classical Sanskrit l is on average three times as frequent as in the entire Vedic literature. Many authorities have long ascribed this state of affairs to the mixing of two basic dialects in Old Indo-Aryan. (See Wackernagel 1896:1, xiv, xix and 209–21; Renou 1957: 4 and 51 n. 49; Debrunner 1957: 116–22.)

For example, the word śloka-, a key term of the Epic tradition, preserves the original *l of the Proto-Indo-European root *k̑leu- ‘to hear’ (Pokorny 1959: 6057; LTV 297–8). The word ślóka- is found in the old family books (R̥V 2–7) only 9 times, but 17 times in the other, younger books (cf. Grassmann 1996: 1432; Lubotsky 1997: II, 1425), while the root is only found in the form śru-, which along with its derivatives occurs many hundreds of times in the R̥gveda (cf. Grassmann 1996: 1419–22, 1425–32; Lubotsky 1997: II, 1419–25). It is not possible to derive words such as śloka from the R̥gvedic language, which lacked the phoneme l altogether. I would like to repeat the following conclusion drawn by Emeneau to underline its importance:

The R̥gvedic dialect, then, is clearly not the direct ancestor of classical Sanskrit. There must have been, even on this much evidence, several closely related dialects in the period of the R̥gveda composition, one of which is the basic dialect of this text, another of which is basically the ancestor of the classical language of some centuries later. (Emeneau 1966: 127)

In the case of rhotacism a phonological feature is involved. Similar dialectal differences between the Vedic and Classical Sanskrit prevail in morphology as well:

[I]t is also clear that the R̥gvedic linguistic norm, even apart from hymns that represent something very close to the classical language, was a mixed dialect, and that one of the elements in the mixture was something near to classical Sanskrit. Only such a hypothesis will explain such mixtures as those of -āsas and -ās, of -ebhis and -āis [in the instrumental plural of the Indo-European o-stems], of the instr. sg. endings -ā and -ena of IE o-stems, of the neuter plurals of Indo-European o-stems in -ā and -āni, and so on…the R̥gvedic language represents a mixture [of dialects]…; one very like, or even identical with, the ancestor of classical Sanskrit probably formed one very important element in the mixture. (Emeneau 1966: 127)

I take up for a closer examination just the two alternative nominative plurals of the masculine -a-stems, -ās and -āsas (in the pause form, -āḥ and -āsaḥ). Vedic -āsas corresponds to Avestan -āŋhō and Median or Old Persian -āha and can therefore be reconstructed for Proto-Aryan.

It has been suggested that Proto-Aryan may have introduced the double plural ending *-ās-as only in the masculine -a-stems in order to keep their nominative plural separate from the nominative plural *-ās of the feminine -ā-stems. (Secondarily, the suffix -āsas is applied in the Veda to a few feminine -ā-stems; see Lanman 1878: 362.) In Avestan, however, -ā̊ 〈 *-ās (cf. the sandhi form -ā̊s-ca) is found only in the nominative plural of the feminine -ā-stems, the nominative plural of the masculine -a-stems being distinguished from it by the ending -ā̆ (-ā-ca), which has a counterpart in Saka, too, and may go back to a Proto-Indo-European collective suffix *-ā (thus Hoffmann 1975: 70, who compares the Latin form loca from locus; Hoffmann actually ascribes this to Sogdian too, but Nicholas Sims-Williams points out that, unlike Khotanese Saka, Sogdian cannot in fact distinguish between the results of *-ā [〉 Khotanese -a] and *-āh [〉 Khotanese -e]: both give -a).

Another explanation for Proto-Aryan -āsas is the analogy of the *-i- and *-u-sterns, which have manifold relationships with the *-a-stems and in which the nominative plural was likewise bisyllabic, -ay-as and -av-as. (Cf. Debrunner and Wackernagel 1930: III, l0lf.)

Excursus: The Proto-Indo-European background of-ās-as

It is possible that Proto-Aryan *-ās-as continues Proto-Indo-European *-ōs-es, for there is evidence for a fully parallel double suffix in Proto-Germanic as well. Gothic -ōs (dagōs ‘days’) and Old Norse -ar (dagar) can be derived from Proto-Germanic *-ōz, which in its turn is the regular development of PIE *-ōs, as PIE post-vocalic word-final *-s became voiced *-z in Proto-Germanic. But such West Germanic forms as Old Saxon dagōs and Old English dómas presuppose a Proto-Germanic reconstruction *-ōs, which cannot be reconciled with the above explanation of the East and North Germanic forms. The Old Saxon and Old English forms imply that their -s comes from a non-final *s in Proto-Germanic. Proto-Germanic *-ōsez (in words having the accent on the thematic vowel) or *-ōzez (in words with the accent on the root) from PIE *-ōs-es would provide an explanation, as the vowel *-e- of these assumed Proto-Germanic suffixes would have been dropped in the prehistory of all Germanic languages. Gothic -ōs, on the other hand, could as well go back to Proto-Germanic *-ōsez or *-ōzez and Old Norse -ar to *-ōzez. (See Bammesberger 1990: 43f.)

Thus it seems that Proto-Germanic and Proto-Aryan both had the same innovation. However, it may have taken place separately in these two branches, which seem to have emerged far from each other and are not otherwise closely related. There is a general agreement that PIE *-ōs is the original nominative (and vocative) plural marker of the masculine *-o-stems, and that it goes back to the combination of the thematic vowel *-o-with the nominative plural marker *-es. This original suffix *-ōs, however, seems to have been preserved only in few languages: besides the Aryan branch, it is known just from the Oscan-Umbrian sub-branch of Italic and from Old Irish; in Hittite the -a-stems (PIE *-o-sterns) form their nominative plural with -aš or -ēš, the latter being thought to be a transfer of the corresponding suffix of the -i-stems (-ēš from PIE *-ey-es). Because the contraction of o + e into ō in the suffix *-ōs resulted in the plural suffix *-es becoming blurred in the *-o-stems (in contrast to the consonantal and *-i- and *-u-stems) in Proto-Indo-European, there seems to have arisen a need to innovate and replace *-ōs with a clearer nominative plural suffix. While this was done by adding the suffix *-es to the original suffix *-ōs and making this case end as in other stems, another early solution was to replace it altogether with the pronominal masculine nominative plural suffix *-oi: this suffix is reflected in Celtic (which also had *-ōs!), Latin, Greek, Phrygian, Baltic, Slavonic, and Tocharian, possibly also in Albanian and Hieroglyphic Luwian. (Schmitt-Brandt 1998: 206–9; cf. also Debrunner and Wackernagel 1930: III, 101f.; Szemerényi 1989: 194, 196; Beekes 1995: 191f.)

Returning to Indo-Aryan, the double plural suffix -āsas ‘contrasts with -ās, which latter is found also in the R̥gveda and is the only form in the classical language. It has been demonstrated that, in the R̥gveda, -āsas is the form proper to its dialect, from metrical considerations, viz. that written -ās many times has to be read -āsas to mend defective meter, while -āsas never has to be read -ās’ (Emeneau 1966: 126f.; for the evidence, see Lanman 1878: 345f.).

Emeneau clearly suggests that -āsas originally was the proper suffix of the nominative plural of the masculine -a-stems in the R̥gvedic dialect. However, this need not have been so; the R̥gvedic dialect could have possessed both -āsas and -ās from early on, making use as it does of their metrically convenient difference in the number of syllables; in fact both forms frequently occur side by side and often replace each other in parallel verses: R̥V 5,59,6a té ajyeṣṭhā́ ákaniṣṭhāsa(ḥ); 5,60,5a ajyeṣṭhā́so ákaniṣṭhâsa été (cf. Lanman 1878: 344f.; Debrunner and Wackernagel 1930: III, 100).

But although -āsas disappeared from the literary texts rather quickly, did it continue to be used in the spoken language of the R̥gvedic people? It survives in the later Middle Indo-Aryan dialects (Pāli and Ašoka’s inscriptions) as -āse; and this is not an isolated phenomenon, for several other features not found in Classical Sanskrit connect later Prakrits with the language of the Rgveda (thus e.g. R̥gvedic -ebhis as the instrumental plural of the -a-stems becomes -ehi in Middle Indo-Aryan [cf. Hinüber 1986: 145], while Classical Sanskrit has only -ais). T. Y. Elizarenkova (1989), while drawing this conclusion, suggests on the basis of some further evidence that the R̥gvedic poets used different language in the hymns and in their everyday speech, which early on started to exhibit Prakritic features.

This question has been studied also by Colette Caillât in an article entitled ‘Vedic and Early Middle Indo-Aryan’ (1997): ‘Since the time of Franke and others, it has generally been accepted that not a few features of Middle Indo Aryan (MIA) are inherited from some form of Vedic rather than classical Sanskrit dialect(s)’. (Caillat 1997: 15)

Among other things, she discusses (pp. 18–21) the MIA suffix -āse ‘generally assumed to be “Eastern” (because of the final -e continuing Sanskrit -aḥ in the East)’, a conclusion which in Caillat’s opinion is ‘not really warranted’, because ‘manuscript tampering cannot be excluded’ and in several cases the Sinhalese tradition has -āso, either as a variant reading or (once) as the sole reading against the -āse of the Burmese tradition (p. 18).

The Pāli commentators and grammarians seem to hesitate about the function of the final -se: should it be viewed as the nom. pl. marker…or as an invariant particle…liable to be added, for the sake of euphony whether to the verb in the 1 pl. (karomase) or to the noun in the nom. pl. (gatāse)? (Caillat 1997: 19)

Analysing the oldest Pāli texts, Caillat (p. 19) finds eight substantives with -āse. There is, however, a much larger number of verbal adjectives, which often function as predicates, being sometimes paired with finite verbs, as in the following old triṣṭubh-jagatī formula, which recurs in many texts:

ye keci buddhaṃ saraṇaṃ gatāse /na te gamissanti apāya(-bhūmi)ṃ //.

But the phrase viyāpaṭāse in the ‘eastern’ pillar edict of Aśoka at Delhi-Topra should be interpreted differently, from viyapaṭā (ā)se, where āse would be an old preterite akin to āsi.

Thus it would appear that the Pāli nom. pl. -āse could have been of mixed origin. This would explain why it retains only part of the Vedic characteristics. On the one hand, the gatāse type might have been in use anciently in the East, and was apparently not confined to verse. On the other hand, it seems that literary texts blend the (more colloquial?) gatāse and the (more artificial?) brāhmaṇāse types. Despite the conventional character of the Pāli usage and many uncertainties, the above investigation again shows the existing gap between Old Indo-Aryan (OIA) and the language of the oldest MIA documents known to us. (Caillat 1997: 20–21.)

Witzel (1989: 212f.) points out that -āse is found in Prakrit only in Pāli gāthās and twice in the Delhi Topra inscription of Aśoka; everywhere else in Prakrit we find the suffix -ā from -ās, since Ardha-Māgadhī -ao seems to be a new formation on the analogy of the consonant stems (cf. Hinüber 1986: 144 [and Caillât 1997: 18 n. 18]), and the Apabhraṃśa vocative plural -aho is explained from the stem -a- + the vocative interjection bho (cf. Hinüber 1986: 146). According to Witzel, -āse has acquired the Eastern nom. pl. -e (cf. also Hinüber 1986: 144); he concludes that it is either a fairly late development or that ‘perhaps it was indeed the famous “first wave” of Indo-Aryan immigration into the East which had perpetuated the spread of the R̥gvedic usage -āsaḥ in the East, where it remained in use, while the Kuru-Pañcāla form -āḥ gained prominence in the rest of the Middle Indian dialects’ (Witzel 1989: 213 n. 281; see fig. 1 below).

In my view, Vedic -āsaḥ may have spread to the east with the Kāṇvas (the Kāṇva school moved rather far east to Kosala, cf. Witzel 1989: 113), and the Kāṇvas are connected with the Yadu-Turvaśa tribes, the ‘early’ R̥gvedic tribes. But theirs would not have been the ‘first wave’ of Aryan immigrants to the east: the ‘Pre-Māgadhī’ (with *-az 〉 -e) may have come there with the Dāsas (on whom cf. section 2.10.2).

In any case, the ancestor of Epic or Classical Sanskrit did not possess the suffix *-āsas, but only the suffix -ās: this must have been the main factor contributing to the rapidity with which the suffix -āsas was ousted by the suffix -ās in the texts. While the numerical ratio between -āsas and -ās in the R̥gveda on the average is 1: 2, in the youngest parts of the R̥gveda -āsas is already rare. In the verses of the (Šaunaka) Atharvaveda which are not mere variants of R̥gvedic verses, -ās occurs 24 times more often than -āsas, while in the prose portions of the Taittirīya Saṃhitā only -ās is found. (Cf. Lanman 1878: 344), xiv; Debrunner and Wackernagel 1930: III, 100.)

In his discussion of the double plural suffix, Witzel states:

The R̥gvedic nom. pl. -āsaḥ is found in Old Avestan as -āŋhō [recte -ā̊ŋhō (ed.)], and in Mede as -āha (as represented in the O. Pers. inscriptions: aniyāha bagāha). The extension by *-as is, therefore, an Indo-Ir. development (*-āsas) which had (partially) affected some of the tribes but not all, notably not those of a later wave of immigrants (e.g., Y. Avestan, O. Pers., Post-R̥gvedic). While the innovation -āsaḥ is found in the RV, the older form -aḥ is found exclusively in post-RV texts (except for archaism and quotations from the RV/mantra language). (Witzel 1989: 212)

[Nicholas Sims-Williams points out that in fact -ā̊ŋhō is Younger Avestan as well as Old Avestan; that Old Persian -āha is ‘Median’ is controversial and unprovable.]

On this basis, Witzel assigns the double suffix -āsas to the tribes of Yadu-Turvaśa and Anu-Druhyu = the first wave immigrants (see fig. 1a), and the simple suffix -ās to the Pūru and Bharata tribes = the second (later) wave of immigrants to the Panjab (cf. Witzel 1989: 212, 233–4, 238). Witzel further seems to have reasoned that the simple plural suffix -ās belonged to the language of the Pūrus and the Bharatas as it was their descendants who founded the Kuru kingdom and the language of the Kurus had only the suffix -ās in the following Middle Vedic period (see fig. 1b).

But although historically the immigration of the Yadu-Turvaśa and Anu-Druhyu tribes into the Panjab preceded that of the Pūrus and Bharatas, the great majority of the R̥gvedic hymns and consequently also their dialect do not reflect the language of those tribes who had come earlier. Witzel himself assigns only a few hymns to this ‘first layer’ of his. He observes: ‘clearly this corpus [of R̥gvedic hymns] was composed primarily by the Pūrus and Bharatas and spans the story of their immigration’ (Witzel 1995b: 328).

If -āsas is characteristic of the main R̥gvedic dialect, then it is the Pūrus and Bharatas to whom the main original dialect of the R̥gveda with its double plural suffix -āsas belongs.

Results of a dialect mixture are already clearly visible in the youngest hymns of the R̥gveda, but this mixture is not so clear yet in the older hymns, especially as some of the dialect features characteristic of Classical Sanskrit have been introduced into the text only much later by the editors of the hymn collection; this is illustrated for example by the earlier mentioned cases where the ending -ās for metrical reasons is to be read -āsas. Therefore, it is likely that the older hymns were composed relatively free from contact with the speakers of the ancestor of Classical Sanskrit, while the youngest hymns were composed after a contact which was quite strong and which in later Vedic texts is even stronger. By the time of the Yajurvedic prose, which is also a new type of text totally unknown from the R̥gvedic tradition, there is no more trace of the R̥gvedic suffix -āsas. It is evident that the youngest hymns of the R̥gveda were composed in the Panjab and the neighbouring regions of the north Indian plains, where all the Middle Vedic texts too have come into being. Thus the dialect mixture discussed above took place after the Pūrus and Bharatas arrived in the Panjab and came into contact with the tribes who had settled there before them. Thus we must conclude that the dialect of some of these earlier arrived tribes should have only the oldest plural suffix -ās.

As noted earlier, I consider the provisional frame that Michael Witzel has set up for the historical processes of the R̥gveda (with three sub-periods involving two major waves of immigration, together with their dating) as essentially correct, but object to Witzel’s suggestions concerning the textual layering. In reply to a question by Jost Gippert at the Second World Vedic Workshop held in Kyoto in 1999, Witzel confirmed that his assignment of specific hymns to the oldest textual layer is not based on any linguistic or metrical peculiarities, but simply on the fact that those hymns speak of the earliest Aryan tribes in South Asia (especially the Yadu and Turvaśa, and the Anu and Druhyu), usually in positive terms. This method is defended as follows:

It is important to note that four of the ‘Five Peoples’ of the RV, the Yadu-Turvaśa and the Anu-Druhyu, do not figure prominently in most of the RV, and if so, the stanzas praising them are composed in standard R̥gvedic, not a hypothetical l-dialect represented in eastern MIA. Only the newcomers, the Pūru and their original sub-tribe, the Bharata, play a major role; most of the books 2–7 have been composed when the Pūru-Bharatas were about to enter or had just entered the Panjab. Whatever had been composed before must have been recast in Pūru-Bharata style (or has been lost). (Witzel 1997b: 263)

But as the early tribes are spoken of even in very ‘young’ hymns, an arbitrary element will inevitably be present in such assignments. The procedure becomes all the more unacceptable when subjective estimations—as to which hymns do and which do not belong to the oldest textual layer—are made the basis for judging the value of loanword occurrences and reconstructing the prehistory of the Indus Valley on that basis (Witzel 1999b,c,d; see section 3.2).

The representation of Proto-Indo-European *l and the suffixes -ās and -āsas in the R̥gveda thus show that although the ‘youngest’ hymns of the R̥gveda are most recent from the point of view of the textual history, i.e. the time of their composition and inclusion in the text collection, from the point of view of dialect formation involving the entry of Indo-Aryan speakers in South Asia at different times they reflect an earlier layer. In the case of the particular linguistic details discussed above, the earlier layer of previously arrived tribes represented by the ‘youngest’ hymns exhibits a more archaic stage of language development, which indicates that the dialect(s) involved had separated from Proto-Aryan at an earlier point of development than the dialect(s) of the ‘older’ hymns.

But is the dialect ancestral to the Epic and Classical Sanskrit (which rises to prominence after the dialect mixture following the Pūru-Bharata immigration) the dialect of the Yadu-Turvaśa and Anu-Druhyu tribes, which according to Witzel represent the first Indo-Aryan immigration? On the basis of textual, religious and archaeological evidence, I am inclined to answer this question negatively and to posit an immigration of Aryan speakers to South Asia that is earlier than that of the Yadu-Turvaśa. I will return to this question after having presented my own identification of the Yadu-Turvaśa layer in the R̥gveda.

In the quotation cited in section 1.4.2, Witzel thinks that whatever poetry the Yadu-Turvaśa tribes composed has either been recast in Pūru-Bharata style or been lost. Although much of the Yadu-Turvaśa poetry is likely to have been both recast and lost, I believe that much of it has also survived: a considerable part of the recasting seems to have been done in its own original style.

It is generally agreed that the original core, the oldest part, of the R̥gveda-Saṃhitā consists in the ‘family books’, R̥V 2–7, each composed by a particular family of poets, and each arranged according to the addressee deities and metres in a uniform way. In my opinion it is the hymns of these poet families (including the hymns ascribed to them in books 1 [the latter half], 9 and 10) that represent the Pūru-Bharata tribes of the ‘second wave’. The Atri family of book 5, however, has a special relationship with the Kāṇva family of book 8, as well with the Anu tribe that is often mentioned with the Turvaśa and Yadu (cf. Oldenberg 1888: 213f., 220f.). Like the Kāṇvas (see below), the Atris also resided in the northwest: it is only in the Atri hymn R̥V 5,53,9 that the Kābul and Kurum Rivers are mentioned, along with other rivers of northwest Pakistan and Afghanistan (cf. Hopkins 1896: 88). The relationship between these two singer families is underlined by a linguistic isogloss: the Atris of the 5th book use no infinitives formed with the suffix -tu- and the Kāṇvas have at least no infinitives in -tum and -tavái (cf. Wackernagel 1896:1, xiii; Renou 1957: 4; Scharfe1996: 361 n. 14). This isogloss differs from Classical Sanskrit, which has only one infinitive, in -tum.

The first major addition to the core of the family books is book 8 or the ‘Kāṇva book’: the hymns of this book are ascribed to the descendants of Kāṇva (chiefly the hymns 1–66) and of Áṅgiras; also the first half (hymns 1–50) of book 1 belongs to the Kaṇva family. It is the Kāṇvas that of all singer families of the R̥gveda have especially close relations with the Turvaśa and Yadu tribes (cf. Zimmer 1879: 122–4; Oldenberg1888: 220f.; Hoffmann1975: 49). Yet, because so many of them are clearly later than most of the family book hymns, the early claims for a higher antiquity for the Kāṇva hymns have generally been discarded. Nevertheless, it is the tradition as a whole, not the actual hymns, that is important in this context, as was seen by Maurice Bloomfield:

The hymns themselves allude in clearest language to songs of old that were composed by the Rishis of the past. The later poets undertake to compare, more or less boastfully or complacently, their own compositions with those of the ancient masters…Such reports are significant because they show that the Vedic poets were aware of the fact that Rig-Vedic composition stretched over a long period, preceding their own time. The suggestion has also been made that hymns which refer to themselves as ‘new’, as ‘having been patterned after old’, as ‘having been made in the manner in which Atri, Kaņva, Jamadagni, and other worthies made their hymns’, are of recent origin. In the light of the materials which are worked up in this book [i.e., the repetitive formulaic expressions in the R̥gveda], I have grown more sceptical as to our judgement in these matters…it is very often easy to point out signs of relative lateness, but I have yet to find any hymns in the collection which show positive signs of coming from the archetype period, that is to say, from the period when hymns of this sort were first composed…In any case they one and all abound in repetitions. Many hymns of the pragātha collection of Kaṇva and the numerous Kaṇvids are most certainly late clap-trap, but the important role which these hymns play in the Sāma-Veda canon should warn us from condemning the rather banal compositions of the eighth book, because this involves the condemnation of the Sāma-Veda to a late date. (Bloomfield 1916: 20–21; cf. ibid. 639–40)

The Kāṇva hymns differ from the ‘family books’ in many respects. For one thing, the hymns are arranged in a different way; the hymns are predominantly in the pragātha and gāyatrī metres, and have a strophic structure, consisting of dis-tichs (pragāthas) or tristichs (gāyatrī). This same strophic structure is found in book 9, which was put together after the 8th book had been added to the R̥gveda, and which contains the hymns addressed to the Purifying Soma collected from books 2–8. Together the Kāṇva hymns and book 9 constitute the earliest version of the Sāmaveda, being meant to be sung by the Udgātr̥ priest, while the family books were meant to be recited by the Hotr̥, the priest who originally poured the offering libation into fire. The ‘Sāmavedic’ character of these hymns is visible from the occurrence of the root gā- ‘to sing’ and its derivatives (in addition to gāyatrī, pragātha, udgātr̥ already mentioned, cf. also gāyatra-, the name of the most important melody or sāman-, as well as sāma-ga- ‘Sāman singer’ (cf. R̥V 8,98,1 indrāya sāma gāyata…). (cf. Oldenberg 1884.)

Alfred Hillebrandt (1927:1, 448ff.) has noted that poets of the Kāṇva family criticise the habit of other priestly families who make offerings of plain Soma (to Vāyu and Indra) without mixing it with milk or curds, with honey or with barley, so as to make it sweet. The recipients of the Soma mixed with milk or curds are Mitra and Varuṇa, while the Soma mixed with honey is offered to the Aśvins and the Soma mixed with barley to Indra and his team of horses or to the demon Marka (death).

E. W. Hopkins (1896) published a long list of words that occur in the 8th book but not elsewhere in the R̥gveda (pp. 29–52); and similar lists of words occurring in addition to the 8th book only in books I, IX and X (pp. 52–70). He noted that the differences are not just lexical, but include also cultural and geographic background and proper names (pp. 84–8): in contrast to the family books, the Kāṇva hymns repeatedly refer to sheep and agriculture, including ploughing. The only R̥gvedic reference to the Swat River (suvā́stu- ‘providing good habitation’) is in the Kāṇva hymn R̥V 8,19,37b. Karl Hoffmann (1975: 16) observed that besides other linguistic peculiarities the 8th book has an unusually high number of words with retroflex consonants, including the name of its eponym, Kaṇva. Besides, four out of the five occurrences of the word uṣṭra- ‘camel’ are found in the 8th book (the fifth being in the late hymn R̥V 1,138,2), praising gifts of ‘a hundred camels’ etc. that various kings have presented to the singers. This as well as the names of the royal givers have suggested the neighbourhood of Central Asia and Iranian speakers: Caidyá Kaśú (kaśú- is a hapax in Indo-Aryan, while kasu‘small’ is common in the Avesta), Pr̥thuśrávas Kānītá (cf. the Scythian names Kανίτης, Kάνιτος), Dāsá Balbūthá and Tárukṣa, Tiríndira and Párśu (cf. Old Persian Pārsa- 〈 *Pārśva- ‘Persian’) (Hoffmann 1975: 9); see section 2.4.

The evidence suggests that the core area of the Yadu-Turvaśa and the Kāṇva family was the Swat Valley. But some of them must have moved early on to the plains, as they are found later rather far east and south. Reference has already been made to the Kāṇva school in Kosala in the Brāhmaṇa period, and Hartmut Scharfe (1996: 361 n. 14) points out that ‘a short-lived Kāṇva dynasty succeeded the Śuṅgas in Magadha in the middle of the first century B.C. The Yādva ‘descendant of Yadu’, mentioned in R̥V 7,19,8 (along with Turvaśa an enemy of King Atithigva) and in 8,6,46, can be compared to the epic Yādava clan of Mathurā.

Georges-Jean Pinault (1998) in a penetrating article has revealed a most important fact concerning the Kāṇva hymns: they alone in the R̥gveda provide exact parallels to the Mitanni Aryan proper names from the Near East (ca. 1500–1300 B.C.). Eight Mitanni Aryan bahuvrīhi proper names end in -att(h)i ‘having X as his guest’: Biriatti /Priyātt(h)i-/, Mittaratti /Mitrātt(h)i-/, Ašuratti /Asurātt(h)i-/, Mariatti /Maryātt(h)i-/, Šuriatti /Sūryātt(h)i-/, Intaratti /Indrātt(h)i-/, Paratti /Prātt(h)i-/ and Šūatti /Suvatt(h)i-/. The Kāṇva tradition of the R̥gveda knows five parallel proper names ending in -atithi ‘guest’: Médhyātithi-, Médhātithi-, Nīpā́tithi-, Mitrā́tithi- and Devā́atithi- (cf. Pinault 1998: 453–5). Pinault refutes earlier etymologies for atithi- as inadequate. He determines the semantic meaning (‘who stays nearby, who is placed next to [the house of the host]’) from an analysis of R̥gvedic and Homeric verses that describe the attitudes towards, and the treatment of, the guest. The new Indo-European etymology proposed suggests, according to Pinault (1998: 473), three possible developments in Indo-Iranian:

PIE *h₂o-th₂-ti 〉 Proto-Aryan *át^hiti- 〉

(1) *át^hti- 〉 Proto-Iranian = Avestan asti-

(2) Proto-Indo-Aryan *áthiti-, but as a second member of a compound dialec-tally either (2a) *-atti-, which corresponds to the overall language of the R̥gveda, or (2b) *-atthi-, which reflects a deviant Vedic dialect and explains the formation of the Vedic word átithi-.

The connection detected by Pinault between the Kāṇva hymns and Mitanni Aryan strongly supports the hypothesis that the Kāṇva hymns represent the traditional poetry of the Yadu-Turvaśa tribes and therewith the ‘first wave’ of R̥gvedic Indo-Aryans in South Asia.

At the end of the treaty made in 1380 B.C. between the Hittite king Šuppilu-liuma and his later vassal, the future Mitanni king Sātivāja, deities are invoked to see that the treaty is not breached and to punish those who break it. After one hundred and four Hittite and Hurrian divities, four Aryan divinities are invoked: Mitra and Varuṇa, Indra and the Nāsatyas. Paul Thieme (1960) has pointed out that these same divinities are mentioned in the same order in R̥V 10,125,1: ahám mitrā́vâruṇobhā́ bibharmy ahám indrāgnī́ ahám aśvínobhā́; the only difference is that instead of Indra in the treaty, Indra-and-Agni are mentioned here, and that the divine charioteer twins are mentioned by a different name. The name Nāsatya ‘rescuer, savior’ is very commonly used in the R̥gveda (cf. Zeller 1990: 5f., 172); the Avesta (Vidēvdāt 10,9; 19,43) knows (along with Indra and Saurva = Post-R̥gvedic Śarva = Rudra) a single Nā̊rjhaiθiia as a deity condemned by Zarathushtra (cf. Thieme 1960: 315).

The speaker in the above-quoted hymn R̥V 10,125 is the Goddess Vāc ‘Voice, Speech’; in the Brāhmaṇa texts she is identified with the goddess Sarasvatī and the goddess Sāvitrī, the daughter of the sun, and functions as a goddess of fertility and victory, much like the later Hindu goddess Durgā and the Old Iranian goddess Anāhitā (see Parpola 1999). In the R̥gveda, however, Vāc appears personified as a goddess only in the ‘late’ books 1 and 10, with the exception that she is praised as the ‘mistress of the gods’ (rā́ṣṭrīdevā́nām) in the 8th book (R̥V 8, 100, 10–11), another connection with the Kāṇvas.

There is another important point that connects this Mitanni treaty with the Kāṇvas and the ‘late’ books of the R̥gveda. Among the singer families of the R̥gveda, it is especially the Kāṇvas and the Atris—and the Vasiṣṭhas of the 7th book, whose special deity is Varuṇa, also mentioned in the Mitanni treaty—that worship the Aśvins. The R̥gveda contains altogether 54 complete hymns addressed to the Aśvins; 16 of them are found in the 1st book, 12 in the 8th book, 6 in the 5th book and 5 in the late 10th book; of the family books, the 7th book has 8 complete Aśvin hymns, but the other family books only very few: there is 1 hymn in the 2nd book, 1 in the 3rd book, 3 in the 4th book and 2 in the sixth book (cf. Zeller 1990: 1; also 163f. for a detailed listing of hymns, verses and references). Even the Yādavas of Mathurā have preserved their worship of the Aśvins (associated in the Veda with day and night) in their special deities Balade va and Kr̥ṣṇa —brothers of whom one is white and the other black—associated with the horse chariot in the Epic team of Arjuna ‘White’ and his charioteer K̥ṣṇa ‘Black’ (cf. Parpóla 2002a).

We have concluded above that the archaic Old Indo-Aryan dialect encountered by the immigrating Pūru and Bharata tribes in the Panjab was ancestral to Epic and Classical Sanskrit, and, secondly, that it was not the language of the Yadu, Turvaśa and Anu. These Indo-Aryan tribes had arrived earlier and stayed in the northwest, in and around the Swat Valley, although some of them had proceeded further south and east.

There is general agreement that the latest part of the R̥gveda is the 10th book. Linguistically the 10th book provides the clearest evidence of the dialect mixing that took place after the Pūru-Bharata tribes had settled in the Panjab and had been subjected to the substratum influence of the language of its previous inhabitants. At the same time as the features characteristic of the later Epic and Classical Sanskrit (including l) enter the R̥gvedic language, religious themes and cultural concepts that are notably different from the R̥gvedic tradition make their first appearance; the riddles (brahmodya) and ‘speculative hymns’ including the ‘creation hymn’ (R̥V 1,129), the ‘Puruṣa hymn’ about the ‘primeval man’ whose sacrifice and dismemberment brings into being the cosmos and human society divided into four classes (R̥V 10,90), the ‘marriage hymn’ (R̥V 10,85) with its first clear references to calendrical asterisms and the ‘funeral hymns’ (R̥V 10, 14–16), can be mentioned as well known examples.

It is also generally agreed that these latest hymns of the R̥gveda are the earliest specimens of a great mass of oral traditions that was collected—most probably in the eastern Panjab, in Kurukṣetra—in the extensive Atharvaveda-Sarṃhitā during the immediately following Mantra Period (cf. Gonda 1975: 28f.; above, section 1.1.3). The contents of the Atharvavedic hymns, however, cover a much wider range of non-R̥gvedic material than the early samples in R̥V 10: in particular, they contain much ‘magical’ material (‘medical’ and love charms, sorcery and so on) and rituals connected with kingship and war. (cf. Bloomfield 1899: 46f.; Witzel 1997b: 275–83.)

The Atharvaveda is in a ‘later language’, closer to the language of the Brāhmaṇa and Sūtra texts and to Classical Sanskrit. But, as Bloomfield rightly stresses,

For reasons that are nearly always one-sided and subjective, sometimes patently erroneous, the language of the popular or Atharvanic hymns is generally regarded as chronologically later than that of the hieratic hymns [of the main bulk of the R̥gveda]…many linguistic forms that are looked upon as indications of late date are in reality as old, sometimes older, than the entire individual period of the Aryan language in India…A given form is not necessarily of recent origin because it begins to crop out in the tenth book of the RV., appears still more frequently in the popular Vedic collection of the AV., and is the regular form of the post-mantric language; nor, consequently, are hymns necessarily late because they abound in words or forms that are strangers to the diction of the hieratic hymns. (Bloomfield 1899: 46)

Examples of linguistic archaisms in the language of the Atharvaveda (and the late hymns of the R̥V) contrasting with the language of the family books of the R̥gveda are:

AV r, l vs. R̥V r,

AV -ās vs. R̥V -ās, -ās-as;

the instrumental plural AV -ais vs. R̥V -ebhis;

the dual AV -au vs. R̥V -ā;

AV pánthānam, pánthānas vs. R̥V pánthām, pánthās;

the gerund -tvā vs. R̥V -tvī,

AV sarva- vs. R̥V viśva-;

AV svap- vs. R̥V sas-

(cf Bloomfield 1899: 46ff.; Lanman 1878: 340L; 349f.; 441f.; Tikkanen 1987: 252).

Some linguistic forms characteristic to the language of the Atharvaveda and differing from the language of the R̥gveda anticipate Middle Indo-Aryan:

AV karomi, kuru vs. R̥V kr̥ṇomi, kr̥dhi;

AV puruṣa-, pūruṣa- vs. R̥V *pūrṣa-;

AV kim vs. R̥V kad

(cf. Bloomfield 1899: 46; Wackernagel 1896:1, xxviii; Renou 1957: 7,13).

This suggests that the language of the Atharvaveda had already for some time been subjected to the substratum influence of languages spoken by the earlier population of the Indus Valley, particularly Dravidian (cf. section 3.2).

In addition, many words which are common in the R̥V occur either not at all or very rarely in the AV (cf. Wackernagel 1896:1, xxviii; Renou 1957: 13), while the AV in its turn has a considerable number of words not found in the R̥V (cf. Renou 1956: 32–4). In spite of numerous recent publications on the Atharvaveda, especially the Paippalāda recension, Renou’s note (1956: 32 n. 1) about the lack of a satisfactory linguistic study of the Atharvaveda is still valid.

For quite some time, the Atharvavedic tradition was not fully accepted into the Vedic fold, which elaborated the ‘classical’ Śrauta ritual of the ‘threefold knowledge’ of the R̥gveda, Sāmaveda and Yajurveda during the Mantra Period. Its representatives therefore had to struggle to conform and to get accepted; in the late redaction of the AV, ‘popular’ linguistic forms originally belonging to the Atharvavedic dialect and found in the parallel passages of the late books of the R̥gveda are replaced by the corresponding ‘hieratic’ forms. (cf. Bloomfield 1899: 46ff.; Witzel 1997b: 266ff.)

That the Atharvavedic hymns originate from a tradition different from that of books 2–9 of the R̥gveda is betrayed by the innumerable prosodie mistakes committed by their authors in trying to apply the R̥gvedic meters (cf. Bloomfield 1899: 42f.). About the antiquity of the main metre of the Atharvaveda, which approaches the Epic śloka, Bloomfield states:

The Atharvanic and Gr̥hya anustubh may be designated as the popular anuṣṭubh in distinction from the hieratic anuṣṭubh of the soma-hymns in the RV.: a hymn like RV. 10.85 is, of course, in popular anuṣṭubh. Considering the absolute quantitative freedom of the eight-syllable line of the Younger Avesta, we have reason to assume that the popular and freer anuṣṭubh is structurally and chronologically earlier than the better regulated hieratic (somic) anuṣṭubh. (Bloomfield 1899: 41)

[Nicholas Sims-Williams notes that the eight-syllable line of the Younger Avesta is not universally accepted, though he himself is inclined to accept it. On the metrics of the Avesta, see especially Lazard 1984.]

Part of the Atharvaveda is in prose, a new mode of literary expression which becomes predominant in the Brāhmaṇa texts, starting with the Saṃhitās of the Yajurveda. These Middle Vedic texts codify and explain the Vedic ritual corpus, which incorporated many sacrificial rites that had not originally been part of the R̥gvedic Soma cult. Royal rites (rāja-karmāṇi) to be performed by the priest of the royal house (purohita) occupy a very prominent position in the Atharvaveda, and Bloomfield (1899: 74) already saw in such Atharvavedic rites the source of the royal sacrifices of the Śrauta ritual, in particular the horse sacrifice (Aśva medha), the royal consecration (Rājasūya), the Vājapeya (involving a chariot drive) and the Sautrāmaṇī. These rituals which had been performed in South Asia before the arrival of the Aryans associated with the family books of the R̥gveda and the Soma cult were however presented as ‘variants’ of the Soma sacrifice, which was now made an essential part of them, and all ritual acts were given a R̥gvedic stamp by assigning a R̥gvedic mantra to be pronounced at their performance. This resulted in a situation in which

the R̥gveda, in a way, occupies an anomalous position: though chronologically older and the source of most of the formulas contained in the Yajurveda, it was peripheral to what became the main tradition of ritualism into which it was intercalated only at a later date. (Gonda 1975: 88)

Evidently the R̥gvedic mantras originally accompanied the Soma cult, while the royal rituals and other Yajurvedic cults were accompanied by yajus formulae in prose (see Oldenberg 1917).

One important theme in the ‘proto-epic’ gāthās and ślokas of the Middle Vedic texts is the praise of particular kings (cf. Horsch 1966: 251–84); in this they continue the dānastutis of the R̥gveda, mostly found in the eighth or Kāṇva book; the late books 1 and 10 and the Atri book 5 also stand out (cf. Patel 1929: 12–28). Their ritual context can be seen from the following description given of the horse sacrifice in the Śatapatha-Brāhmaṇa (13,1,5,1–6; cf. also Horsch 1966: 420ff.):

1.…when a man attains to distinction, the lute is played to him. Two…lute-players sing (and play) for a year; for that—to wit, the lute—is a form (attribute) of distinction: it is distinction they thus confer upon him…. 3. One of those who sing is a Brāhmaṇa, and the other a Rājanya; for the Brāhmaṇa means priestly office, and the Rājanya noble rank: thus his distinction (social position) comes to be guarded on either side by the priesthood and the nobility.…5. The Brāhmaṇa sings by day, and the Rājanya at night…6. ‘Such sacrifices he offered, — such gifts he gave!’ such (are the topics about which) the Brāhmaṇa sings…‘Such war he waged, — such battle he won!’ such (are the topics about which) the Râjanya sings…Three stanzas the one sings, and three stanzas the other…To both of them he presents a hundred [cows]. (Transl. Eggeling 1900: V, 285–7)

This practice of singing praise songs of three stanzas is undoubtedly connected with the ‘Sāmavedic’ practice of singing songs in tristichs, initiated in the Kāṇva hymns of the R̥gveda (see above, section 1.4.3).

The narrative poetry of the epics and Purāṇas can likewise be traced back to royal rituals described in the Middle Vedic texts (on the following, see also Horsch 1966: 20ff). When the sacrificial horse is set free to run around for one year, and the royal sacrificer and the priests have seated themselves on golden seats,

the Adhvaryu calls upon (the Hotr̥), saying, ‘Hotr̥, recount the beings: raise thou this Sacrificer above the beings!’ Thus called upon, the Hotr̥, being about to tell the Pāriplava Legend, addresses (the Adhvaryu), ‘Adhvaryu!’ — ‘Havai hotar!’ replies the Adhvaryu. 3. ‘King Manu Vaivasvata’, he says; — ‘his people are Men, and they are staying here’; — householders, unlearned in the scriptures, have come thither: it is these he instructs; — ‘The R̥k (verses) are the Veda: this it is’; thus saying, let him go over a hymn of the R̥k, as if reciting it. Masters of lute-players have come thither: these he calls upon, ‘Masters of lute-players’, he says, ‘sing ye of this Sacrificer along with righteous kings of yore!’ and they accordingly sing of him; and in thus singing of him, they make him share the same world with the righteous kings of yore. (SB 13,4,3,2–3, transi. Eggeling 1900: V, 361–3)

On the following mornings, there is a similar procedure, but the contents vary as follows (ŚB 13,4,3,2–14):

Every morning, the Hotar priest after his own short presentation calls on ‘the masters of lute-players’, or rather, ‘harp-players with their retinue’ (vīṇāgaṇaki-naḥ), who ‘sing of this sacrificer along with the righteous kings of yore’. The same ten-day cycle ‘revolves’, i.e. is repeated again and again for a year, as long as the sacrificial horse roams around, and includes performances of magic tricks, etc.

As can be seen, the pāriplava legend of the horse sacrifice implies recitation of what is virtually epic and Purāṇic poetry to uneducated masses during great, long-lasting royal rituals, with the purport of firmly establishing the king’s legitimacy and popularity among his people. In the beginning of the Mahābhārata, it is told that the epic was first recited at the King Janamejaya’s snake sacrifice (sarpa-sattra): this sacrifice has a counterpart in the mythical sarpa-sattra described in Middle Vedic texts (PB 25,15 and BaudhŚS 17,18).

It is often suggested that the silence of the R̥gveda about magic, sorcery and many other Atharvavedic topics is due to social differentiation, which kept the priestly religion apart from the ‘popular’ cults of the lower classes. Such a social stratification no doubt came into being in the course of repeated immigrations of small but powerful groups that gradually subdued earlier populations, not without keeping themselves aloof. But the royal rites of the Atharvaveda alone prove that it is not just a question of lower class religion. There are slight but clear dialectal differences in addition to religious divergences which point to a separate origin of the Atharvavedic traditions. To account for them, we have to posit an Indo-Aryan immigration to South Asia that preceded the coming of the ‘R̥gvedic’ Aryans, even the earliest wave of the Yadu and Turvaśa tribes identified here with the poetic tradition of the Kāṇvas.

Above, several waves of Indo-Aryan speaking immigrants to South Asia have been distinguished on the basis of language, textual characteristics and religion. In the following section an attempt is made to trace these in the archaeological record, which has the potential of providing a more exact frame for historical reconstruction.

All human communities have spoken some (majority) language and made use of animals and artefacts, which within one and the same community have some uniformity. We usually have in historical sources some direct indication of the time and place when and where a given ancient language was spoken, while the temporal and geographical horizons of distinct archaeological cultures can be defined much more exactly. In addition, the position of each language with respect to other languages before and after, and synchronically far and near, is defined by its genetic relationships with related languages and areal contacts with unrelated languages indicated by structural influences and loanwords. Similar vectors exist for the archaeological cultures: each of them has developed out of something and eventually develops into something else, receiving and giving influences from and to other cultures far and near.

I trust that if we aim at a ‘total reconstruction’, trying to correlate all known archaeological cultures of a large enough region with all languages that could have been spoken there, it is possible to establish relatively reliable correlations between these two different kinds of evidence. There usually are some equations where the various parameters overlap to such a degree that a certain conclusion is possible or even unavoidable; these starting points will at once yield certain corollaries within the framework of operation, and these again will define new ones. The main criterion will be how well each piece of the puzzle fits in the total web of relationships. (On this ‘total relationship principle’ cf. Mallory 1998.)

Pitfalls exist, of course: we probably do not know all the languages and cultures involved, for example; and several languages may have been used by people representing one archaeological culture, or vice versa. People representing one archaeological culture may also change their language, adopting a new one, but this takes place only when new people—not necessarily very many—move in, which usually leaves some traces in the archaeological record. In spite of the unavoidably hypothetical nature of the results—which can and must be constantly checked and revised to accord with new evidence—attempting to correlate the textual-linguistic and archaeological data seems to me a useful exercise, because the two kinds of record have preserved different aspects of the past and can supplement each other: for ‘pots and pans’ cannot tell what language their users spoke, but once that language has been identified, it will obtain a more precise distribution in time and space from the mute artefacts.

In the following, I start from the ‘safest’, i.e. latest, identifications in South Asia and the neighbouring regions, and work backwards in time to the Bactria and Margiana Archaeological Complex (BMAC, ca. 2100–1500 B.C.), the ‘filter’ through which the earliest Aryan speakers coming from the pastoral-nomadic cultures of the Eurasian steppes seem to have successively passed on their way to Iran and India.

A luxury ceramic called Painted Grey Ware (PGW)—an élite ware that constitutes only a small percentage of the total pottery of this culture—characterises the relatively uniform archaeological culture of the Panjab, Haryana and Uttar Pradesh between ca. 1100/900 and 400 B.C. (On the PGW and its dating, see Tripathi 1976; Ghosh 1989:1, 107f.; Erdosy 1995a: 80, 83f.) Geographically and temporally the distribution of the PGW without doubt provides the best imaginable archaeological match for the Vedic culture in its core area during the Middle Vedic period. With good reasons, this equation is widely accepted. Both iron and horse are mentioned in Middle Vedic texts and they are present in the culture characterised by the PGW as well.

The Kuru tribe may owe its supremacy in the Middle Vedic period not only to the strength of the R̥gvedic Pūrus and Bharatas whose line it continues but also to new Late/Post-R̥gvedic immigrants, who were not numerous enough to make any noticeable linguistic impact. It appears likely that the PGW culture got both its iron and its grey ware from the probably Iranian-speaking immigrants who arrived in southern Central Asia ca. 1100 B.C. At Pirak III in Pakistani Baluchistan there is evidence around 1100 B.C. of the intrusive appearance of iron in association with a new type of wheel-made grey ware and other artefacts with parallels in southern Central Asia (late phase of the Yaz I complex, see section 2.4), including Afghanistan (especially Tillya Tepe), and in Sind (Jhangar and Chanhu-daro) (cf. Jarrige and Santoni 1979:1, 373–398; Parpola 1988: 264f.).

The name Kuru first appears in the late R̥gvedic hymn 10,32,9, where King Kuruśravaṇa is said to be a descendant of Trasadasyu, the well-known Pūru king. But according to Mahābhārata 1,89,42, the eponymous founder of the Kuru kingdom is the son of Saṃvaraṇa and of Tapatī, the daughter of the sun-god (tataḥ saṃvaraṇāt saurī suṣuve tapatī kurum / rājatve taṃ prajāḥ sarvā dharmajña iti vavrire: the second line is folk-etymologically based on the name Saṃvaraṇa). Kuru’s parents here are almost certainly of Iranian origin: Saṃvaraṇa can be compared with Younger Avestan Hąm.varəti ‘valour, bravery’ (personified as a divinity), while Tapatī compares with the goddess who according to Herodotus (4,59) is called Tαβιτί (usually connected with the (Indo-)Iranian root tap- ‘to burn’, as well as Ossetic tœvd and Modern Persian taft ‘hot’; this etymology, however, is criticised by Vasmer 1971:1, 119). Herodotus also states that Tαβιτί is the chief object of worship by the Scythians, and identifies her with Hestia, the Greek goddess of the family hearth and domestic fire. Hence it may not be a mere coincidence that the name of the Kuru dynasty resembles the name of the founder of the Persian Empire, Kuruš ‘Cyrus’ (a comparison which has been made several times; cf. Hoffmann 1975: 13).

The Medes and Persians are first mentioned in Assyrian sources in texts from 843–836 B.C. (cf. Diakonoff 1985: 61ff.) Cuyler Young (1985) connects these tribes with the Late West Iranian Buff Ware (ca. 900–700 B.C.) and traces this ceramic back to southern Central Asia, where it first appears around 1100 B.C. This date approximately coincides with the appearance of the proper name Kuru in the Veda (see section 2.2.1).

[Roman Ghirshman (1977: 49f.) proposed that the Medes and Persians came from the north over the Caucasus, as the Scythians did according to Herodotus (1,103–4; 4,11–12), but this hypothesis is not considered very convincing; cf. Sundermann 1979: 598; Diakonoff 1985: 50ff.; Francfort 1989:1, 437f.]

The roots of the Medes and Persians are thus traced to southern Central Asia of the late second millennium B.C. From ca. 1500 B.C. to 1000 B.C., practically the whole of southern Central Asia was occupied by the Yaz I related cultural complex. The type-site Yaz I in Margiana (Merv) is related to Anau IVA and other sites in the Kopet Dagh region, Kuchuk Tepe I-II in Uzbekistan on the Amu Darya, Tillya Tepe I-III in northeast Afghanistan, Mundigak V in southwestern Afghanistan and Pirak IB-III in Pakistani Baluchistan. The Yaz I complex is characterised by hand-made pottery, mostly plain, but sometimes painted in red or violet with geometric motifs, especially hanging trianges. This new type of ceramics, which is related to that of the Chust culture in Ferghana (on which see Zadneprovskij 1993), largely replaced the previous ‘Namazga VI’ type wheel-turned pottery that had prevailed during the first half of the second millennium B.C. in southern Central Asia; the proportion of the earlier wheel-made pottery increases from 6% to 14% towards the end of the Yaz I period. Terracotta figurines of horse riders appear at Pirak LA-IB-II, and knife and arrowhead types have parallels in the northern steppe zone. The economy of southern Central Asia however remains essentially similar to that of the preceding Bronze Age, i.e. agriculturally based. The immigrating nomads have become sedentary.

Some lexical items and proper names in R̥gvedic hymns composed in the northwest suggest the presence of Iranian speakers in the neighbourhood at the time of their composition, ca. 14th century B.C. (see section 1.4.3; cf. also Witzel 1980). On the basis of its distribution and date, the Yaz I complex is therefore likely to reflect the coming of the majority of the Iranian tribes to or towards their present habitats. The name of the Pashtuns and their language Paṣ̌to is probably derived from *Parsa- or *Parsu- and can be connected with the Πάρσιοι placed in Afghanistan by Ptolemy (cf. Skjærvø 1989: 384) and with the Párśu of R̥V 8,6,46 (cf. Hoffmann 1975: 9), perhaps even with Pārśva in Pārśvanātha, the name of the 23rd Tīrthankara of the Jainas (ca. 700 B.C.). This Iranian identification of the Yaz I complex is supported by the striking fact that not a single burial has been discovered from this area between ca. 1500–330 B.C., in contrast to Chorasmia and the Chust culture of Ferghana, where graves belonging to the same period are known. Henri-Paul Francfort (1989: I, 437) has interpreted this as an indication that the sedentary population started practising the funerary rites of exposure described in the Avesta, while the nomadic Saka continued practising inhumation. (On Yaz I, see Francfort 1989:1, 430–438.)

If the Yaz I complex is equated with the Iranians, it implies the arrival of Iranian mounted horsemen in southern Central Asia around 1500 B.C. This date is so close to the violent beginning of the Ghalegay V phase in Swat ca. 1400 B.C. (see section 2.6) and to the estimated arrival of the Pūru-Bharata wave of R̥gvedic tribes in the Panjāb, also ca. 1400 B.C. (see section 1.1.1), that a causal relationship is likely. Actually, L. H. Gray (1927: 439) proposed that the ahura-worshipping Iranians drove the deva -worshipping Indo-Aryans out of their former habitat in southern Central Asia to India. Gray’s hypothesis, however, was rejected by Burrow (1973: 133), because the destruction of the Indus Civilisation, which he (mistakenly) ascribed to the R̥gvedic Aryans, took place already ca. 1900 B.C., much too early for any Iranian involvement. In R̥V 6,61 (see section 1.1.2) the poet praises Arachosia as a wonderful place which he does not want to leave (verse 14). The country of Hapta Hindu (i.e. the Panjāb), on the other hand, is presented as a less inviting place associated with excessive heat in Vidēvdāt 1,19. The hypothesis that Iranian horsemen forced the second wave of R̥gvedic Aryans to move to India is in agreement with the fact that riding is known to the R̥gveda, including the Atri book (5,61,2–3), but is mentioned only very few times and has obviously not surpassed chariotry in prestige nor become a common means of moving around by this time (see Falk 1994).

After this excursus on archaeological evidence for the presence of the early Iranians in Iran, in southern Central Asia, and South Asia, we continue with archaeology and the Indo-Aryans. Most parts of the area occupied by the PGW in the Middle Vedic period (see section 2.2) belonged earlier to the Cemetery H culture, dated to ca. 1900–1300 B.C.

The urban phase of the Harappan culture that had been dominating the Indus Valley since ca. 2600 B.C. came to an end by ca. 1900 B.C. In the Panjab and Hary ana it was smoothly succeeded by a village culture named after ‘Cemetery H’ at Harappa, where a new type of burial was introduced. While many Harappan traditions continued without any break, city life vanished, and with it the Indus script. There was however a fundamental change in funerary customs, as extended inhumation in graves was replaced by cremation followed by deposition of the ashes in funerary urns.

The funerary urns of Cemetery H are painted with partly old, partly new motifs associated with the fate of the dead. Thus we find a peacock carrying the dead person to the stars of the sky, the deceased being depicted as lying horizontally on the bird’s back or in its stomach. While the peacock is prominent in the Mature Harappan painted pottery, it is never before depicted like this as a soul-carrier.

The changes in Harappan culture that are manifest at Cemetery H suggest outside influence and the agency of sufficiently powerful immigrants, though they need not have been very numerous. The origin of these immigrants remains problematic, because the neighbouring Bactria and Margiana Archaeological Complex (BMAC) in Afghanistan (see section 2.10), which spread to Sind ca. 2000–1900 B.C., practised only inhumation. Stuart Piggott (1950: 230–35) made comparisons with the Late/Post-Harappan Jhukar culture in Sind and the Kulli culture of Baluchistan. Pottery said to be similar to both Cemetery H and Jhukar type pottery has been discovered at the Kuzali phase of the BMAC in southern Uzbekistan (Askarov 1984: 94–5), datable to ca. 1700 B.C. (see section 2.10.3). Cremation as the principal mode of disposing of the dead was introduced in northern Bactria by newcomers from the northern steppes representing the Fedorovo variety of the Andronovo cultural complex; this took place immediately after the Kuzali phase, during the Molali and Bustan phases at the end of the Bronze Age (cf. Avanesova 1995).

In any case, the Cemetery H culture persists over a vast area of the northwestern plains until about 1300 B.C. Six centuries is a long enough period for the non-Aryan language of the Harappans (see section 3.2) to have largely shifted into the Old Indo-Aryan ancestral to Classical Sanskrit, i.e. early ‘Atharvavedic’. The end of the Cemetery H culture around 1300 B.C. coincides with the date proposed by Witzel for the Late R̥gvedic period (ca. 1350–1200 B.C.), by which time the dialect mixing discussed earlier had taken place. This period would effect the transition to the Middle Vedic period, equated above with the PGW culture, which starts ca. 1100 B.C. The oldest Yajurvedic texts, associated with the Kaṭha and Kapiṣṭhala-Kaṭha schools, were composed around the 10th century B.C.: this took place in the Panjab, where the Greek texts place the tribes of καθαι̑οι and καμβισθόλοι. The formation of the Yajurvedic ritual presupposes a longer foregoing development in this region; it is likely to have started from ca. 1600 B.C., when contacts with the Ghalegay IV culture of Swat begin (see section 2.6).

The periodisation of the cultures in the Swat Valley is based on the layers in the rock shelter of Ghalegay. (cf. Stacul 1987: 29f.) The earliest period, Ghalegay I (ca. 3000–2500 B.C.), represents the Northern Neolithic culture known from the surrounding areas and especially from Kashmir, with links to China. The Ghalegay II period (ca. 2500–2000 B.C.) is characterised by Kot Diji style pottery, which attests to the presence of Harappans belonging to the not fully urbanised branch which occupied the northernmost Indus Valley. Scarcity of structural remains and domestic articles has suggested that they did not settle permanently in Swat, but came there temporarily to log pine trees for the big cities in the south. After the disintegration of the Indus Civilisation, the local Neolithic people took over the Swat Valley again for some centuries (Ghalegay III, ca. 2000–1700 B.C.). (cf. Stacul 1987: 115–20.)

The next period, Ghalegay IV (ca. 1700–1400 B.C.), marks the arrival of a new, intrusive culture, called Gandhara Grave culture. (This term includes the fairly similar next phase too). It is among the first in South Asia to have clear evidence for the domesticated horse, in motifs on painted pottery as well as in bone finds. The black-grey, burnished pottery introduced from the beginning of this phase is widespread throughout Swat, and its affinities are with the late phase of the BMAC culture widespread in the surrounding areas at this time (see sections 2.10.5); the graves have been compared with the graveyards of Bishkent and Vakhsh in south Tajikistan (cf. Mandel’štam 1968; Müller-Karpe 1983: 117–19; Lyonnet 1994: 430f.). A little later, from around 1600 B.C., red pottery with motifs such as the peacock and three-branched fig tree appear alongside the grey-burnished ware, attesting to lively contacts with the Cemetery H culture of the Panjab plains. The earlier Northern Neolithic also continues to form a component of the population of Swat during this period. (cf. Stacul 1987: 120–26.)

The next phase, Ghalegay V (ca. 1400–800 B.C.), is introduced with some violence, evidenced by human skulls and other disconnected skeletal remains scattered at random without any traces of burial in a thick barren layer separating phases IV and V at Bīr-koṭ-ghwaṇḍai. (cf. Stacul 1987: 126.) A new type of grey vase with a characteristic button base is introduced in both Swat and Dir; it has ‘strong stylistic and typological analogies with northwest Iran during the period of Hasanlu V’ (Tusa 1979: 691).

Languages of the Dardic or ‘mountain tribe’ group of Neo-Indo-Aryan are spoken in the northwest more or less in the area once occupied by the Gandhara Grave culture plus Kashmir. In their relative isolation in the mountainous periphery they have remained little disturbed and have preserved many archaisms. Particularly significant are the traces of the gerund formed with the suffix -tvī́ (cf. Buddruss 1973: 45), which in Old Indo-Aryan is attested in the R̥gveda alone; also in Middle Indo-Aryan the only Prakrit in which this particular gerund is reflected belongs to this area in the northwest, namely, Gāndhārī with its -ti gerund (cf. Hinüber 1986: 199f.). In the R̥gveda, the gerund in -tvī́ occurs 37 times, being more than half as frequent again as -tvā́—the form that prevails in Classical Sanskrit—which occurs 23 times in the R̥V and with which -tvī́ has been replaced in all the R̥V stanzas repeated in the AV. The gerund in -tvī́ occurs in all books of the R̥V. (cf. Tikkanen 1987: 81f.)

This linguistic feature suggests that a considerable number of R̥gvedic Aryans probably came to South Asia via the northwest. The date for the violent beginning of the Ghalegay V phase, around 1400 B.C., neatly agrees with that calculated for the immigration of the Pūru and Bharata tribes. If these tribes moved to the Panjab via Swat, they could not have stayed there a long time, as King Divodāsa was born in Arachosia, but his son Sudās already fought in the Panjab. (See section 1.1.2.) However, some of the newcomers—the Atris—apparently stayed back there, while the others continued to the plains. It seems that the Kāṇvas of the 8th (and 1st) book are a similar residue of the first wave of the earlier arrived Yadu and Turvaśa tribes, who can thus be connected with the Ghalegay IV period.

Gandhara graves are rectangular pits, often surrounded by a ring of stones. About two thirds of them are interment burials and about one third cremation burials. The oldest graves predominantly consist of urn burials, agreeing with the funerary customs of the Cemetery H culture in the Panjab; the latest phase consists predominantly of inhumations, while in the middle phase both types are about equal in number. (Cf. Müller-Karpe 1983: 29–51.) The funerary customs of the Gandhara Graves can be correlated with those of the Veda. I must postpone an elaboration of this point to another publication, and refer the reader to the comparisons made by Müfler-Karpe (1983: 96–113). Just one very general statement: the R̥gveda knows both inhumation and cremation.

The last phase of the Ghalegay V period belongs to the Iron Age and contains also wheel-turned red and grey ceramic with remarkably thin walls. (Cf. Müller-Karpe 1983: 73f.) This last-mentioned ceramic is very likely connected with the introduction of iron and the Painted Grey Ware to the plains of the Panjab and the Doab in the immediately post-R̥gvedic period (cf. section 2.2.1). The passage from R̥gvedic to Middle Vedic is reflected in a few links between the Ghalegay V and PGW cultures. From the PGW layers at Ahicchatra come cremation urns with perforations near the neck, resembling the perforated ‘eyes’ and ‘mouth’ on the ‘face-urns’ of Ghalegay V; and terracotta human figurines of the Ghalegay V type have been found from the PGW layer at Jakheran, U.P. (cf. Agrawal 1982: 250; Parpóla 1988: 248.)

We have seen above that the main wave of the R̥gvedic Indo-Aryans (represented by the family books 1–7) in all likelihood originated from the regions immediately to the west of the Indus Valley, such as Arachosia (see section 1.1.2), and that this wave is very probably represented in the Ghalegay V culture of the Swat Valley, with strong parallels in pottery with the Button Base Ware of the Hasanlu V period (ca. 1500–1000 B.C.) in northwestern Iran (see section 2.6). There is again reason to quote Burrow:

The Avesta has no ethnic term to denote the Proto-Indoaryans, which is not surprising since both they and the Iranians called themselves Aryans and spoke closely related languages. The opposition between the two sides is always spoken of in religious terms between the Mazdayasnas, worshippers of Ahura Mazdā and followers of Zoroaster on the one hand, and the Daēvayasnas…on the other…Opposition to daēvas and daēvayasnas is ascribed in the Yasts even to ancient and legendary kings such as Yima…In spite of the efforts at conversion on the part of the Mazdayasnians the worship of the daēvas continued to exist for a considerable period, much as later the Mazdayasnian religion itself continued to exist in Islamic times…

The Avestan evidence so far discussed concerning the Proto-Indoaryans has dealt with eastern Iran, but in view of the appearance of Proto-Indoaryans in the Near East, one would expect that they had also a base, or bases, in western Iran from which the invasion of the Huirían country was mounted, just as their migration to India depended on their base in eastern Iran. The geographical horizon of the Avesta is almost exclusively eastern Iranian, but it does have some references which indicate the presence of Proto-Indoaryans in northern central Iran, bringing them within striking distance of the Near East. These are the references which occur from time to time in the Avesta to the Māzanian daēvas. The adjective in question (māzainya-) is derived from *māzana-, the name of a country which happens not to occur as such in the Avesta, but whose location is indicated by the fact that it has always been known to be connected with the country later known as Māzandarān, i.e. the territory between the southern shore of the Caspian sea and the Alburz mountain range. In later tradition this figures prominently as a region hostile to the Iranians and as a notorious home of Dēvs. The presence of daēvas in Māzana indicates the presence of daēva-worshippers, and since we have seen that the daēva-worshippers were the Proto-Indoaryans, we can conclude that the Avestan references to Māzanian daēvas indicate their presence also in this region. (Burrow 1973: 133–134)

Burrow’s conclusion is uncertain, however, for the Zoroastrian equation of māzaniia- with Māzandarān is much later and there is no indication in the Avesta that it is a geographical term (cf. Gnoli 1980: 44–50; Skjærvø 1995: 165).

That Mitanni Aryan belongs to the ‘Indo-Aryan’ branch—as argued by Paul Thieme (1960), Manfred Mayrhofer (1966; 1974) and Thomas Burrow (1973)— has not been seriously doubted in recent years except by Hans Henrich Hock (1999a: 2f.). Hock observes that the strongest argument for this hypothesis is the numeral aika- ‘one’ (in Kikkuli’s Hittite manual for training chariot horses) vs. reflexes of *aiva- in all known Iranian languages. Hock’s claim that even this is no conclusive proof, because lost Iranian languages might have had *aika-, and that Mitanni Aryan may equally well represent Proto-Aryan does not seem reasonable. In any case, the ‘Indo-Aryan’ affinity has been strengthened by Pinault (1998), whose study furthermore relates the Mitanni Aryan specifically to the ‘Kāṇva wave’ of the R̥gveda (see section 1.4.4).

The Kāṇva and Mitanni worship of the Nāsatyas, the divine charioteers, agrees with the Mitanni Aryans’ fame as masters of chariotry. It is well established that the name of the Mitanni king in the treaty with Šuppiluliuma is probably to be read /Sāti-vāja-/, which compares with Vedic vā́ja-sāti- ‘winning of vā́ja’ and related phrases, while three successive Nuzi kings seem to bear related names: /Sāta-vāja-, Vāja-sāta-, Sāta-vāja-/ (cf. Mayrhofer 1974: 25–8). The Vedic texts describe in quite some detail the chariot race, whose prize is called vā́ja- (on this term see now Hintze 2000: 67f., 85–119); in the following I shall try to show that this source material has not been exhausted for the elucidation of the Mitanni evidence and vice versa, thus supplying new evidence for the Indo-Aryan affinity of the Mitanni.

The turning pillar of the Vedic chariot race is called kāṣṭhā-, sthūṇā- or stambha- (Sparreboom 1983: 41). According to Baudhāyana-Śrautasūtra 11,7, it has to be made antaḥ-pakṣas-, which Sparreboom (1983: 42, 59 and 81 n. 27) translates ‘winged’, literally ‘(placed) between (two) wings’, connecting it with the epithet of the chariot-horses in the mantra TS 7,4,20g, ví-paksas-: the correct translation ‘having bird’s wings’ was suggested for ví-pakṣas- in R̥V 1,6,2 by Johanna Narten (1960: 133 n. 38), likewise ‘having a bird’s flight’ for ví-patman-in R̥V 1,180,2. The horses of the Aśvins, too, are winged (patatrín-), and almost always fly, being called freely either aśva- ‘horse’ or ví- ‘bird’ (Zeller 1990: 108). This calls for comparison with the winged horses on Near Eastern seals of the second half of the second millennium B.C. (cf. Collon 1987: 66f.). Particularly important, however, is the seal of the Mitanni king Sauštattar from ca. 1420 B.C. (fig. 2). Sauštattar was an ancestor of King Sāti-vāja. Here we see a pillar ending in a spoked wheel that is flanked by bird-wings.

H. Frankfort (1939: 205–16 and 273–83), while discussing ‘Assur, “sacred tree” and winged disk’ and ‘the seals of Mitanni’, noted that the Mitannians and the Hittites appear to have introduced the originally Egyptian motif of the winged sun-disk as a symbol of imperial power into Asiatic glyptics, but in contrast to Egypt, where the winged sun-disk hovers freely in the sky, in the Mitanni seals it is firmly supported by a pillar, while two pillars are usual on Hittite seals and monuments (fig. 3). Frankfort associated this post with the Vedic conception of a cosmic pillar that supports the sky, but he did not comment on the difference between the Mitanni and Hittite seals. I surmise that the presence of one or two pillars might stem from the use in the race-course of just one turning post (this was the case with Greek hippodromes, where ‘the turning post normally consisted of a single column or a less permanent feature’, cf. Humphrey 1986: 255) or of two (as in the Roman race-course, which had a meta prima and a meta secunda, cf. Humphrey 1986: 26, 37–8, 42–6, 255–9, 323).

Two turning posts are required if the chariot race involves more than one round. The technical terms in Mitanni Aryan contained in Kikkuli’s Hittite handbook for training chariot horses include the word wa-ar-ta-an-na /vartana-/‘round’ as the second member of compounds with a-i-ka- laika-/ ‘one’, ti-e-ra-/ tri/‘three’, pa-an-za- /pañca-ड ‘five’, Ša-at-ta-/ satta-/ or /sapta-/‘seven’, and na-a-/ nava-/‘nine’. (cf. Mayrhofer 1974: 14f.) The Vedic chariot race, however, involved just one round, from the starting point, around the turning post and back (cf. Sparreboom 1983: 41). The Vedic chariot race seems to have as its heavenly model the single daily round of the Aśvins, expressed by the term vartís- ‘going around’, which is used of the Aśvins in 24 verses of the R̥gveda (Zeller 1990: 46 with n. 311). The Aśvins make these rounds in the company of their sister or wife, the daughter of the sun, who brings the first light to the world; for this reason, the Aśvins use a chariot all of whose parts are triple (R̥V 1,34 and Zeller 1990: 100–108). The ‘triple’ chariot of the Aśvins can be compared to the troika used at the Vedic chariot races: in addition to the yoked pair of two horses, there is an extra horse that runs on the left side to help in turning around the pillar at the far end of the race course. (cf. Sparreboom 1983: 36f.)

The wheel at the top of the pillar in Sauštattar’s seal also has a parallel in the Vedic chariot race. At the Vājapeya sacrifice, the king, on starting his victorious race (which is compared to the course of the sun and which is supposed to bring the victor to heaven, cf. Sparreboom 1983: 20, 58; and R̥V 1,169,2, Hoffmann 1975: 223), ascends his chariot near the cātvāla pit, pronouncing the mantra TS 1,7,8a: ‘By the instigation of the god Savitar, through Br̥haspati the vāja-winner, may I win the vāja. ‘At the same time the Brahman priest mounts a chariot-wheel made of udumbara wood and provided with 17 spokes, fixed at the top of an axle that has been raised in the cātvāla pit, the starting point of the race. While ascending the wheel, the Brahman utters the mantra TS 1,7,8b: ‘By the instigation of the god Savitar, through the vāja-winner Br̥haspati, may I climb to the highest vault’. (Cf. BaudhŚS 11,7; Sparreboom 1983: 38f.) While the charioteers run the race, the Brahman performs and wins it symbolically: he sings the vāja-winners’ Sāman and a ritual assistant turns him around clockwise three times (pradakṣi-ṇamāvartayati, Āpastamba-Śrautasūtra 18,4,10; Sparreboom 1983: 46f.). This ritual probably lies behind the Indian conception of the cakravartin ‘ruler of the universe’, literally ‘possessing the turning of the wheel’ (cf. Sparreboom 1983: 157f.; Scharfe 1989: 51–5). It seems to survive in the Buddha’s ‘turning of the wheel of universal law’, dharmacakrapravartaṇa; the notion of dharma is also present in the horse race, where the king’s horse is yoked with the mantra TS l,7,7e invoking the god Savitar to impel the dharma for the king.

According to Baudhāyana-Śrautasūtra 11,7, the turning pillar is a sthūṇā made of udumbara wood. The pillar in the sitting shed of the sacrificial place is also called audumbarī sthūṇā, and it should be vi-śākhā-, ‘having branches that spread in different directions, branched, forked’; this may indicate how the turning pillar was in practice made ví-pakṣas- ‘bird-winged’. That it had the same symbolism is suggested by the fact that in the Epic, Viśākha- is another name for the war-god Skanda (or for his son; cf. Hopkins 1915: 227). Possibly the large bearded male figure with bird’s wings in the Mitanni King Sauštattar’s seal represents the king himself as an incarnation of the victorious war-god Viśākha- (corresponding to the male aspect of the Near Eastern goddess of victory, Ištar-Šawuška, cf. Stein 1994: 297). The Vedic audumbarī pillai with spread branches is to be worshipped with the mantra PB 1,4,9 r̥tadhāmāsi svarjyotiḥ, ‘You are the abode of the universal law, whose light is the sun’ (LŚS 2,2,20 = DŚS 4,2,11). Artatama = /R̥tá-dhāmā/ is attested as the name of a Mitanni king, a contemporary of Pharaoh Thutmosis IV of Egypt (1400–1390 B.C.), probably the son of King Sauštattar and the great-grandfather of Sāti-vāja (cf. Mayrhofer 1974: 22f.; Wilhelm 1994: 294f.).

Mayrhofer (1974: 25f.) argues that the most genuine form of the royal name Sauštattar is to be read Saušsatattar. This form has so far remained unexplained. Mayrhofer mentions, however, *Su-sthātar- ‘provided with good horse drivers’ as the most likely proposal, noting that such a compound would be semantically suitable as the name of a nobleman; the problem is that this compound is not attested in Vedic or Avestan. Sauštattar’s father’s name is most probably to be read as Parsatattar, which Mayrhofer suspects to represent a protoform of puraḥ-sthātár- ‘one who stands in front, leader’, attested in R̥V 8,46,13. As Sauštattar’s descendants’ names are all connected with Vedic chariotry, I would like to propose rather *Savya-šthātar- ‘one who stands on the left (in the chariot), car-fighter’ (as opposed to the charioteer, who stands on the right in the chariot, cf. TB 1,7,9,1 dvau savyeṣṭha-sārathī́): parallels found in Vedic texts are savya-ṣṭhā́-in AV 8,8,23, savye-sthā́- in TB (see above) and savya-ṣṭhá- in ŚB 5,2,4,9; 5,3,1,8; 5,4,3,17–18 (cf. Sparreboom 1983: 162; and Mayrhofer 1996: II, 716, where these compounds are mentioned, but without reference to the name of the Mitanni king, so I assume that this proposal has not been made before).

Sauštattar’s father’s name could be *Pra-sthātar- ‘pre-eminent chariot-fighter’. Compare prati-prasthātar-, which denotes a priest of the Vedic Śrauta ritual working as a pair of the main Yajurveda priest who is called adhvar-yú-, literally ‘connected with the road(s)’, the word adhvarā- denoting ‘ritual’ but originally ‘road, path’; the term adhvaryu- may originally have meant ‘charioteer’, the one who keeps the chariot on the road. The sacrifice is early on often equated with the chariot (cf. Sparreboom 1983: 24ff.) and the divine sacrificial priest Agni is called rathī́r adhvarā́ṇām ‘charioteer of the rites’ (TS 2,5,9,2–3; 5,4,10,1; ŚB 1,4,2,11; cf. Sparreboom 1983: 17). The ‘adhvaryu-pair’ (= Adhvaryu and Prati-prasthātar, defined as ‘the two leaders/conductors of the sacrifice’, adhvarasya netārau, by Bhaṭṭa Bhāskara in his commentary) is equated with the two Aśvins, the divine charioteers, in the ‘five-hotar’ formula (TĀ 3,3 agnír hótā / aśvínā-dhvaryū́ / tváṣṭāgnī́t / mitrá upavaktā́). Cf. further prājitar- ‘charioteer’ in Br̥hat-kathāślokasarņgraha 10,39; Sparreboom 1983: 163), consisting of pra- ‘fore- (= leading, pre-eminent)’ and *ājitar- ‘race-driver’, which is not independently attested but may be assumed on the basis of the phrase commonly used in Brāhmaṇa texts ājím aj- ‘to drive a chariot race’.

The beginning of Aryan rule in Mitanni can be narrowed to the 16th century B.C. (cf. Mayrhofer 1966: 26ff.). Cuyler Young (1985) has plausibly linked the arrival of the Mitanni Aryans in Syria with the sudden appearance of Early West Iranian Grey Ware in great quantities all along the Elburz mountains, in Azerbaijan and around Lake Urmia around 1500 B.C. Young sees this intrusive ceramic as an evolved form of the Gurgan Grey Ware of the Tepe Hissar IIIC horizon, which continued in an impoverished form at Tureng Tepe until ca. 1600 B.C. (Cf. Parpóla 1998: 127f.)

This reconstruction tallies with the Kāṇva connection of the Mitanni Aryans, which relates them to the Ghalegay IV culture in Swat dated to ca. 1700–1400 B.C. (see section 2.6). The Ghalegay IV culture in turn is connected with the Vakhsh-Bishkent cemeteries in south Tajikistan and with the last, Takhirbaj/ Kuzali-Mollali-Bustan phase of the BMAC culture, ca. 1750–1500 B.C. (see section 2.10.5).

Young’s reconstruction provides a chronologically and typologically acceptable link between the Mitanni Aryans and the Gurgan Grey Ware in the southeast corner of the Caspian Sea. Roman Ghirshman (1977: 6–12) connected the Mitanni Aryans with the Tepe Hissar IIIC phase of the Gurgan Grey Ware, now dated to ca. 1900–1750 B.C. The large number of bronze weapons suggests that this rich culture was ruled by a warring aristocracy. On the basis of an alabaster cylinder seal depicting a horse-drawn chariot found at Tepe Hissar HIB and trumpets made of precious metal which were used for giving signals while training chariot horses and directing chariots in battles, Ghirshman (1977: 14–19, 27–32) suggested that Gurgan was the home of the ‘Proto-Indo-Aryans’, who had come there from the north, who domesticated the strong horse native to the Turkoman steppes and who were the first to yoke the horse to a war chariot. Ghirshman (1977: 32–44) further pointed out that the Tepe Hissar IIIC phase of the Gurgan Grey Ware culture is closely related to the approximately contemporary cultures to the east of Gurgan, in southern Turkmenistan and Bactria, interpreting this as evidence for the movement of the ‘Proto-Indo-Aryans’ to South Asia.

However, later research has suggested that the Tepe Hissar IIIC culture of Gurgan came from Margiana and Bactria rather than vice verse (see section 2.10.3) and that the horse was domesticated and yoked to a war chariot in the steppes of the southern Urals rather than in Gurgan. Before discussing the Bactria and Margiana Archaeological Complex or BMAC (in section 2.10), it is necessary to deal with the problem of the original homeland of the Aryan speakers and the birth of the horse-drawn war chariot.

The following brief overview will be limited to a summary of some main conclusions drawn by Carpelan and Parpóla (2001); for the individual cultures involved cf. also Mallory and Adams 1997. The dates given are based on calibrated radiocarbon measurements whenever these are available.

In all likelihood, the unified Indo-European proto-language was spoken in the Srednij Stög culture (ca. 4500–3600 B.C., according to calibrated radiocarbon datings) in the Ukraine (this conclusion was reached already in J. P. Mallory’s fundamental study of 1989). Its successor in the Ukraine and southern Russia, the Pit Grave (Yamnaya) culture (ca. 3600–2500/2200 B.C.), was probably the community from whose Late Central Proto-Indo-European language descend the Graeco-Armenian and Aryan branches. A language ancestral to Graeco-Armenian is likely to have been spoken in the Catacomb Grave culture (ca. 2500/2200-1900 B.C.) in the Pontic steppes, while the speech of the eastern descendants of the Pit Grave culture evolved into Proto-Aryan. According to the testimony of the numerous and partly very early Aryan loanwords in the Uralic (Finno-Ugric) languages spoken in the forest zone of eastern Europe (see now especially Koivulehto 2001), the Aryan proto-language was dialectally differentiated from the start. It appears that the Poltavka culture (ca. 2500/2200-1900 B.C.) in the treeless steppe between the Volga and Ural rivers was ancestral to the ‘Iranian’ branch, while the Abashevo culture (ca. 2300–1900 B.C.) in the forest steppe from the upper Don to mid-Volga and Kama rivers and further to the southern Urals was ancestral to the ‘Indo-Aryan’ branch.

The Indo-Aryan affinity of the Abashevo culture is strongly suggested by a new etymology for the word *śi(k)šta ‘beeswax’ in the Finno-Ugric languages Mordvin, Mari, Udmurt and Komi spoken in the area of the rivers Oka, (mid-) Volga and Kama in central Russia. This area occupied by the Abashevo culture has throughout history been famous for its honey-forests abounding in oak and lime trees, which did not extend eastwards to Siberia, and Proto-Finno-Ugric has long been considered to have *mete ‘honey’ and *mekše ‘bee’ as loanwords from Early Proto-Aryan *medhu and *mekš(i)-. Proto-Volga-Permic *śi(k)šta ‘beeswax’ has an exact counterpart in Indo-Aryan śiṣṭa- ‘id.’ (cf. especially madhu-śiiṣṭa- in Rāmāyana 5,60,10), past participle of the root śiṣ- ‘to leave (over)’, which is not found in any Iranian language; even within Indo-Aryan, it appears to belong not to the R̥gvedic but to the Atharvavedic dialect. (Cf. Carpelan and Parpóla 2001.)

The Abashevo culture (see Prjaxin and Xalikov 1987) seems to have expanded from the upper Don to areas previously occupied by the Volosovo culture (ca. 3900–2400 B.C.), which was probably Proto-Finno-Ugric-speaking, and then by the Fatyanovo-Balanovo culture (ca. 2500–1800 B.C.), where the élite probably spoke Proto-Baltic and the majority Finno-Ugric; from the earliest historical times (ca. AD 550) Finno-Ugric languages of the Volgaic and Permic branch have been spoken in these regions, so the Aryan élite of the Abashevo culture were linguistically absorbed into the Finno-Ugric majority. Bilingualism, however, seems to have continued for several centuries after 1900 B.C. in the Kama basin (the homeland of the Ugric branch of Uralic) and in western Siberia (the area of the Samoyed branch of Uralic), where the Abashevo culture and its metallurgical tradition may have been continued by the ‘Sejma-Turbino intercultural phenomenon’ (ca. 1900–1500 B.C.) that started to exploit the copper deposits of the Sayan and Altai mountains, and traded metals to Estonia and Finland.

In spite of some dialectal differences, Proto-Aryan seems to have continued as a linguistic unity until the end of the Sintashta-Arkaim culture (ca. 2100–1900 B.C.): both the Poltavka and the Abashevo cultures contributed to the emergence of this rich culture in the southern Urals (see Gening et al. 1992; Zdanovič 1995; 1997). The aristocratic graves of the Sintashta-Arkaim culture have yielded the earliest known horse-drawn chariots (see Anthony and Vinogradov 1995; Anthony 1998), which now quickly spread west (to eastern and central Europe and Mycenaean Greece, cf. Kuz’mina 1994b), east (to Siberia, cf. Kuz’mina 1994a) and south (to southern Central Asia and beyond, cf. sections 2.8.2 and 2.8.4).

The final split of Proto-Aryan seems to have taken place ca. 1900 B.C. with the Ural river as the borderline. The more conservative ‘Proto-Indo-Aryan’ came to be spoken east of the Ural river, in the Andronovo cultural complex (ca. 1900–1300 B.C.), which spread widely throughout the Asiatic steppes of Siberia and Central Asia. (On the Andronovo cultural complex, see Avanesova 1991; Kuz’mina 1994a.) ‘Proto-Iranian’ was at first spoken mainly west of the Ural river, in the early Timber Grave (Srubnaya) cultures (ca. 1900–1500 B.C.) of the east European steppe and forest steppe. These two closely related cultural complexes were thus separated for several centuries, until ca. 1500 B.C.

E. N. Chernykh (1992: 241) has associated the wide spread of ceramics decorated by attaching ‘rollers’ or ‘little walls’ to the body of the pot (valikovaja keramika) with the expansion of Indo-Iranian speakers. In my opinion this spread, which started around 1500 B.C., should rather be interpreted as mirroring the migrations of the ‘Proto-Iranians’, who spread eastwards to Siberia and southwards to Central Asia, into areas until then occupied by ‘Proto-Indo-Aryan’ speakers (cf. Parpola 1998: 134). This type of ceramic is thought to have developed from the Early Timber Grave variant called mnogovalikovaja found mainly in the Ukraine. In Central Asia the valikovaja keramika is found in the intrusive Yaz I culture, which also introduced there mounted nomadism (evidenced by terracotta statuettes of eagle-headed riders) as well as a new mode of disposal of the dead, possibly by exposure. (No graves dating between ca. 1500 and 330 B.C. have been found in Central Asia; see section 2.4.) The spread of the valikovaja keramika, which coincides with the emergence of the classical Timber Ware culture in eastern Europe is a prelude to the astounding cultural unification of the Eurasian steppes from the Danube to Inner and East Asia that was completed by ca. 1000 B.C. Thereafter, for more than a millennium, the predominant language of this huge area was Scythian/Saka in its various dialects.

The classical Timber Grave culture (ca. 1500–1000 B.C.) came into being as a result of the expansion of the Early Timber Grave cultures, which were the immediate successors of the Poltavka culture of the Volga steppes. This expansion went west and northwest, to areas previously occupied by the Catacomb Grave and Corded Ware cultures (cf. Klochko 1994: 163). The substratum influence of these new areas could explain important sound changes which distinguish the Iranian branch from the Indo-Aryan branch. The people using Corded Ware in this area probably spoke a language ancestral to later Baltic and Slavic languages. This substratum language may have caused the deaspiration of voiced aspirated stops, an isogloss linking Iranian with Balto-Slavic and Germanic (cf. Kortlandt 1978: 115; Lubotsky 2001). Lexical contacts between Iranian and Balto-Slavic languages also seem to go back to Proto-Slavic and Pre-Scythian times (see Zaliznjak 1962; 1963; Ivanov 1996). The change of Proto-Indo-European *-t^st-/ *-d^zd- into *-st-/*-zd- is another isogloss uniting Iranian with Balto-Slavic—and in this case also Greek—while in Proto-Indo-Aryan the result was *-tt-/*-dd-.

Around 1900 B.C., the Early Timber Grave culture expanded to areas formerly occupied by the Catacomb Grave culture, whose carriers were probably linguistically related to the later Greeks and Armenians. I would like to suggest that the substratum influence of this area lies behind another sound change distinguishing Iranian languages from Indo-Aryan, namely the change *s 〉 h, mainly inter-vocalically and word-initially before a vowel. This change is also characteristic of Greek and Armenian; in Greek it had taken place by ca. 1500 B.C. when Mycenean Greek is attested (cf. Rix 1992: 76–81). This isogloss connecting Iranian, Greek and Armenian has been considered by Vittore Pisani and Roberto Gusmani to be due to prehistoric contact between these languages (cf. Gusmani 1972: 22f.; cf. also Rix 1992: 8). If it was initiated by a Pre-Proto-Armenian-related substratum in the second quarter of the second millennium B.C., it may have spread to all Proto-Iranian dialects and have still been in operation around 1500 B.C., when the ancestors of the later East and West Iranians are assumed to have arrived in Central Asia and the Indo-Iranian borderlands.

For some time it has been widely thought that this pan-Iranian sound change was much later, dating from the beginning of the first millennium B.C., because ‘the name of Susa (presumably in the form Sūša [n]) was certainly taken over by the Persians in Persis, and it is only after this takeover that the change /s/ 〉 /h/ has caused the name of the country “Elam” to appear in Old Persian as 〈U-v-j-〉 /(H)ūža/’ (Mayrhofer 1989: 7 [translated], with reference to Szemerényi 1966; 1968). However,

the Achaemenid Elamite spelling of the town Susa is šu-šá-(an) and this name is rendered in Old Persian as ç-u-š-a, a spelling which Szemerényi has to explain as due to a second borrowing from Elamite. It is implausible that one and the same Elamite name should have given rise in Old Persian both to u-v-j- and to ç-u-š-a-. More convincing is, therefore, the view of F. de Blois that the name of the land of Elam, OP Hūja-/, and the one of the Elamite capital, OP /çūša/ are two different names. (Hintze 1998: 146)

Almut Hintze (1998: 145f.) further rightly points out that Avestan Harax^vaitī-and Old Persian Harauvatiš versus Vedic Sárasvatī-, and Younger Avestan and Old Persian Hindu- versus Vedic Sśndhu- (and possibly a few other similar cases) imply that the sound change of *s 〉 h was still productive when the Iranians arrived in the neighbourhood of these rivers in the Indo-Iranian borderlands. Here Hintze leans on the authority of Karl Hoffmann, who in 1940 concluded that, as this sound change is pan-Iranian and therefore relatively old, the Iranians must have been acquainted with India rather early, about the time of the R̥gveda (cf. Hoffmann 1975: 14 n. 1). Because also the Iranian languages of the northern steppes including Scythian have undergone this sound change, which could not have been operative for very many centuries, it must have come to Central Asia with the ancestors of all West and East Iranian languages presently spoken in Central Asia and Iran, i.e. most probably around 1500 B.C., when the Yaz I culture appears (cf. Hintze 1998: 149; section 2.4). (For the Neo-Assyrian text K. 252, which would seem to be in conflict with this dating of the sound change, see Hintze 1998: 147f. and below, section 2.10.4).

Afghanistan and the surrounding areas, through which the R̥gvedic Aryans must have passed on their way to South Asia, were previously a blank on the archaeological map as far as the Bronze Age is concerned. The excavations of the past few decades, however, have brought to light what its principal excavator Viktor Sarianidi (1986 and 1998a, with further references) has chosen to call ‘the Bactria and Margiana Archaeological Complex’ = BMAC; an alternative name coined by Henri-Paul Francfort is ‘the Oxus Civilisation’. I find it useful to make a distinction between the early phases before the assumed presence of people representing the northern steppe cultures, and will use the term ‘Oxus Civilisation’ when speaking of these early phases, and ‘BMAC when speaking of the later phases that appear to have had a mainly Aryan-speaking ruling élite. (Hiebert 1994 excludes the post-’urban’ phase from the BMAC, and deals with the peak and later ‘urban’ phases as a single period.) The core area of this archaeological complex are the oases of Bactria (Daulatabad, Dashly and Farukhabad in northern Afghanistan, Sapalli and Dzharkutan in southern Uzbekistan) and Margiana (Namazga, Taip, Gonur, Togolok and Kelleli in southern Turkmenistan). The following developmental phases can be discerned (cf. Hiebert 1994: 29–87; Francfort 1989; 2001; Baghestani 1997: 143–5):

Ca. 2500–2400 B.C. = the formative phase: The ‘Oxus Civilisation’ came into being around 2500 B.C. under the influence of the Namazga IV culture of southwestern Turkmenistan, the Proto-Elamite culture of Iran (cf. Lamberg-Karlovsky 1986), the Early and Mature Harappan cultures of Baluchistan and the Indus Valley, as well as the Kelteminar and Afanasevo cultures of the steppes to the north.

Ca. 2400–2100 B.C. (Early Namazga V Period) = the ‘early urban’ phase: The ‘Oxus Civilisation’ acquired amongst other things a distinctive iconography and seal type (compartmented metal stamp seal) under the strong influence of southeast Iran (Tepe Yahya, Shahdad and Shahr-i Sokhta), Baluchistan (Shahi Tump) and southwest Turkmenistan (Altyn Depe). After the Indus Civilisation founded the settlement of Shortughai in Badakhshan in eastern Bactria, its influence increased.

Ca. 2100–1900 B.C. (Late Namazga Vperiod) = the ‘peak urban’ phase: The BMAC is particularly rich and dynamic during this and the following phases. The grave goods include a large number of different weapons including swords, hatchets and maces, and impressive symbols of power such as beautifully decorated ritual axes and long stone sceptres. Horses are depicted on axe- and mace-heads, but as these are without archaeological context they may date from the ‘late urban’ phase. Wine cups of gold and silver depict men drinking wine, ploughing with teams of oxen, hunting game and driving ox-pulled wagons. (cf. Amiet 1986: 201f., 326ff.) The BMAC had contact with people coming from the northern steppes. The BMAC’s expansion through smaller groups to Gurgan in northern Iran, to Shahdad in Kerman, to Seistan and to Pakistani Baluchistan is attested by the import of the entire cultural complex including burials into these areas, while the unidirectionality of this movement is suggested by the absence of goods imported from these regions to Bactria and Margiana (cf. Hiebert and Lamberg-Karlovsky 1992; Hiebert 1995). There is evidence of lively interaction with the late Mature phase of the Indus Civilisation (cf. Jarrige and Quivron 1999; Parpóla 2001). See section 2.10.2.

Ca. 1900–1750 B.C. = the ‘late urban’ phase: This phase is distinguished from the preceding mainly through the presence of the horse and the camel. The BMAC appears to have gained great mobility, and to have spread widely both west and east. It was in trading contact with the Assyrians of Cappadocia and Syria, and typical BMAC seals have been found as far east as the Ordos region in China. Residues of drinks in cultic vessels provide interesting clues to religion. See sections 2.10.3-4.

Ca. 1750–1500 B.C. = the impoverished ‘post-urban’ phase of the BMAC: The rich mature ‘urban’ phase of the BMAC came to a sudden end ca. 1750 B.C. It was followed by what is called the Takhirbaj period in Margiana, the Kuzali, Mollali and Bustan periods in northern Bactria, and the Namazga VI period in the Kopet Dagh region. In this impoverished phase, the small finds distinctive of the mature phase of the BMAC such as seals and axe-heads disappear and architecture is continued only to a limited extent. See section 2.10.5.

It is a fact that neither Andronovo pottery nor barrows typical of the steppe culture are found south of the line defined by the Kopet Dagh, Hindukush and Pamir mountains. (cf. Francfort 2001.) Yet Aryan languages originally spoken in the northern steppes had found their way to Syria and South Asia by the middle of the second millennium. The expansion of the BMAC towards these directions in the west (Gurgan) and in the east (the Bolan Pass from Baluchistan to the Indus Valley) provides an archaeological counterpart to this linguistic spread from the steppes. This implies that an élite of Aryan-speaking pastoralist-warriors coming from the steppes took over the rule of the BMAC, apparently relatively peacefully, thus providing it with an effective hierarchical leadership; this in turn gives a plausible explanation for the increased dynamism and aggressiveness of the BMAC in the ‘urban’ phase. Apart from the increase in weapons and luxury objects, and the introduction of the horse and camel, no major change takes place in comparison to the earlier phase of the Oxus Civilisation. Much the same thing happened in the Mitanni kingdom, where the majority of the population remained Hurrian-speaking: the new Aryan rulers adopted the local culture wholesale, while introducing the horse-drawn chariots and making Mitanni one of the strongest powers in the Near East. (Cf. Parpola 1988; Mallory 1998.)

Although I now think that the R̥gvedic king Divodāsa’s battles against the Dāsas, Dasyus and Panis took place only around 1400 B.C., in eastern Afghanistan (on the way from Seistan to Swat), I continue to identify the earliest Aryan rulers of the BMAC (in the ‘early urban’ phase) with the Dāsas (cf. Parpola 1988: 216ff.). This is argued more extensively in another publication (Parpola 2002b), along with presentation of more material on the ethnic identity of the Dāsas, which has now become clearer to me: they appear to be a vanguard of the ‘Iranian’ branch related to the later Scythians, the Khotanese Saka and the Wakhis of the Pamirs, and to have come from the lower Volga steppes (cf. the preliminary note, section 2.9.1 and below). I base this conclusion partly on L. T. P’jankova’s identification of the earliest steppe ceramics recovered from the floors of the central part of the fortress of Togolok 1 in Margiana as related to the Poltavka and early Timber Grave ceramics of the lower Volga and the northern Caspian Sea regions (cf. P’jankova 1993a: 115–17; Parpola 1998: 123f.; Hiebert 1994: 69).

The ethnic names Dāsa (in the R̥gveda) and Daha (in Old Iranian sources) have an appellative counterpart only in Khotanese Saka (daha ‘man’, this meaning being often the basis of ethnic names) and its cognates in Wakhi and possibly Ossetic: these are all East Iranian languages (cf. Bailey 1960). The Dāsa language however seems to have separated from (Pre-)Proto-Iranian so early that many of the characteristically ‘Iranian’ linguistic features had not yet come into being: the ethnic name itself shows that it has escaped the change *s 〉 h, dated above to ca. 1700–1500 B.C. (see section 2.9.5). Further, I present evidence suggesting that the Dāsas were ancestors of the later ‘Kafirs’ of Nuristan (on whom see Jettmar 1975; Edelberg and Jones 1979; Nelson 1986; Degener 2002); the Dāsas also seem to have been the ‘Proto-Māgadhī’-speaking originators of the ‘proto-Śākta-Tantric’ rituals that are associated with warring bands called Vrātyas in Vedic texts. (On this latter point, see now Parpola 1999; on the Vrātyas, see Falk 1986.)

By the ‘late urban’ phase the BMAC appears to have gained great mobility and to have spread widely both west and east. Probably a fresh wave of Aryans coming from the north took over the rule of the BMAC. So far only camel and donkey bones have been securely identified in the context of the BMAC in Margiana (cf. Hiebert 1994: 135), but bones of both the horse and the donkey come from northern Bactria (cf. Askarov 1993: 69). The domesticated horse belonged to the equipment of the Andronovo pastoralists attested in southern Central Asia at this time by their ceramics. The horse and the camel are represented in the beautiful weapon-sceptres of the BMAC. Along with the camel, the horse also appears as a new motif on a number of stamp and cylinder seals from Margiana (cf. Sarianidi 1998b: 47 and nos. 1395, 1397–9, 1401, 1441–2, 1444–5, 1482). The two-wheeled horse-drawn chariot is depicted on a cylinder seal at Tepe Hissar in Gurgan, where the BMAC expands ca. 1900 B.C., and its use in battle is suggested by the BMAC metal trumpets (cf. Ghirshman 1977).

Ceramics of the ‘Neo-Petrovka’ variant of the Andronovo culture have recently been found at several places in Tajikistan: at the metallurgists’ settlement of Tugai near Samarkand (Sarazm III or IV period, ca. 2300–1700 B.C.), at the nearby Zardcha Khalifa burial (with Sintashta-Arkaim type cheek-pieces for the horse) in the upper reaches of Zeravshan and at the necropolis of Dzharkutan. In all these cases their context is the Sapalli phase of the northern Bactrian variant of the BMAC culture (cf. Masson 1999 and Kuz’mina 1999, both citing primary reports unaccessible to me).

Among the most significant finds of the BMAC in northern Bactria are the maṇḍala- like square fortress-palace of Sapalli-Tepa, which greatly resembles its parallel at Dashly-3 in southern Bactria; the fire-temple of Dzharkutan; and the remains of grapes/raisins and jujube fruits (still locally used for the fermentation of alcoholic drinks) found in jars given as grave goods. (cf. Askarov 1993.) Significantly, three different names of the jujube are mentioned in the Brāhmaṇa texts, always in the context of the Sautrāmaṇī ritual, in which the cultic drink is surā. The beer-like surā is associated with the demon Namuci, while Indra became sick from drinking it (cf. e.g. ŚB 5,5,4,10; 12,7,1,3; 12,7,2,9; 12,9,1,8). Ceramics of the slightly later Kuzali phase of the BMAC in northern Bactria have been compared with the Cemetery H culture of the Panjab, which seems to represent the ‘Atharvavedic’ tradition of Indo-Aryan (cf. section 2.5).

There appears to have been a cultic difference in the BMAC variant of Margiana. In spite of the negative results of tests (cf. Nyberg 1995: 400 and my strong reservations in Parpóla 1998a: 126f.), there seems to be evidence for the ritual use of the ephedra plant in Margiana in the form of stems and pollen grains preserved in layers of organic remains from the bottom of ceramic vessels. In one vessel, seven successive layers could be analysed, and the earliest layers contained besides ephedra also hemp. These vessels come from rooms in the centre of the fortresses of Gonur-1 and Togolok-21 which are set apart by several peculiar features suggesting a cultic preparation of liquids. (Cf. Meyer-Melikyan and Avetov in Sarianidi 1998a; Sarianidi 1998a: 95–8; also Hiebert 1994: 123–6; Parpóla 1988: 236f.). Plants of the genus ephedra are the strongest candidates for being the original *sauma plant of the Indo-Iranians, the soma of the Veda and the haoma of the Avesta (cf. Falk 1989; Nyberg 1995).

The R̥gvedic hymns repeatedly emphasise that their enemies, the Dāsas, did not press Soma or worship Indra. If the Dāsas were the élite of the foregoing ‘peak urban’ phase of the BMAC and they were related to the later Scythians, as I think they were, perhaps the hemp found in the earliest layers of the cultic vessel along with the ephedra is a survival from the preceding period. The Scythians drank wine (Herodotus 4,66 and 70), but they also used hemp(κάvvαβις), throwing it upon hot stones so that it gave out a vapour: this hemp-vapour served the Scythians instead of a water-bath (Herodotus 4,74–5).

Graves of the Loulan alias Qäwrighul culture of the Tarim Basin in eastern Sinkiang (ca. 2000–1700 B.C.) have yielded well preserved mummies now kept at the Urumchi Museum. Most of these mummies were accompanied by carefully bundled twiglets of ephedra; they were usually placed in pouches formed by tying up a part of the mummy’s shroud. (Cf. Parpóla 1998a: 127; Barber 1999: 159). Elizabeth Barber (1999: 148–67) has argued that these earliest cultivators of the Tarim Basin oases came along the northern edge of Tien Shan from the BMAC. Additional evidence for this, which she does not mention, is the compartmented metal seal discovered by Sir Aurel Stein at Kucha in the Tarim Basin, where Stein also came across a Harappan type ‘etched’ carnelian bead; the Kucha seal has a close parallel at Gonur-1 in Margiana and at Shahdad in Kerman, Iran (cf. Baghestani 1997: 8, 146 and 365f., no. 606).

The extraordinary mobility of the BMAC people towards the east along the later Silk Road is attested by the compartmented metal seals purchased in large numbers by Christian missionaries in the 1930s in the Ordos region of China. While nothing comparable is found in the local Chinese cultures, the motifs of the Ordos seals have close parallels among the BMAC seals; the connection with China is supported by the discovery of silk remains in the BMAC graves of Sapalli-Tepa in Uzbekistan (cf. Amiet 1977: 119f.; 1986: 192, 199, 320, figs. 187–8; Biscione 1985; Parpóla 1988: 234, 292; Baghestani 1997: 6–8, 145f., 400–403). It is no wonder that the BMAC art became one of the important bases of the Eurasian animal style (cf. Brentjes 1982: 59f.).

The Proto-Indo-Aryan takeover of the Mitanni kingdom at the end of the ‘post-urban’ period of the BMAC in the 16th century B.C. (cf. section 2.10.5) was preceded by intensive economic and cultural contacts of the BMAC with the Assyrians in Cappadocia and Syria during the ‘late urban’ phase. According to the cuneiform archives found at an Assyrian trading colony near Kültepe in Cappadocia, these Assyrian merchants imported tin from the east, possibly from northwestern Afghanistan, and textiles from the south (Syria), bartering these goods for silver and gold in Anatolia. This trade, which flourished for seventy years in 1920–1850 B.C., affected all areas involved and the contacts are reflected in the seals of the various regions (cf. Collón 1987: 41ff.). From about 1850 B.C. Syria was in close contact with Egypt, and many Egyptian motifs started to appear in Syrian glyptics, among them the ‘twist’ and the winged solar disk. (Cf. Collón 1987: 52ff.) In Margiana,

foreign contacts led to the development of stamp-cylinders for sealing, alongside the local two-sided stamp-seals…That contact with the sealing practices of the west existed is demonstrated by the discovery of two seal impressions on a potsherd from Taip-depe (599–600): these are patently inspired by contemporary Syrian seals…precisely at a period when Assyria was involved in a flourishing trade with Turkey via Syria and was receiving tin from the east. (Collón 1987: 142)

At present nearly a hundred cylinder seals are known from the BMAC (see Sarianidi 1998b: 260–73, 320–23), though only those from Margiana come from excavated contexts, i.e. the ‘late urban’ period (ca. 1900–1750 B.C.). While certain motifs of the BMAC seals undoubtedly come from Syria (see also Amiet 1986: 190), at least one new motif on Syrian seals undoubtedly comes from Bactria and Margiana: this is the two-humped Bactrian camel, which, along with the horse, is a new motif in the ‘late urban’ phase of the BMAC as well (cf. Collón 1987: 161 no. 738; Parpóla 1988: 233 n. 274; Sarianidi 1998b: 41f.; nos. 108–11, 888, 916, 919, 1633–5, 1770).

The iconographie motifs of the seals are important religious symbols and their adoption implies ideological convergence. The Assyrians with whom the BMAC people were in contact from the 20th century B.C. onwards later became vassals of the Mitanni Aryans; one may therefore expect a strong Assyrian influence on the religion of the BMAC and the Mitanni Aryans (using Mitanni-style seals). When the Assyrians regained their power after the decline of the Mitanni, their national god Aššur was represented in art by the winged solar disk, which had been a central symbol in Mitanni glyptics as well. In Mitanni glyptics the winged solar disk is often associated not only with a supporting pillar, which can be connected with the cosmic pillar as well as with the turning pillar of the horse race (which has a heavenly model) (see section 2.8.2), but also with the ‘tree of life’, as often in Middle and Neo-Assyrian seals (cf. Frankfort 1939: 205ff.; Collon 1987: 65f.). Frankfort (1939: 204f.) inquired into the meaning of ‘the highly artificial “sacred tree” which usually appears as the source of power round which the figures are symmetrically grouped’ and noted that ‘it occupies—alone [on a particular seal] and together with the king [on another seal]—the place taken by a god [on a third seal]’, concluding that it was ‘likely that the “sacred tree” on Assyrian seals represents the national god, Assur’.

This ‘Tree of Life’ has indeed turned out to be a key to the Assyrian religion. Its symbolism was preserved up to mediaeval times in Jewish mysticism:

The Tree of Life of Kabbalah is a multi-layered symbol in which the metaphysic structure of the universe (macrocosm) and the model of the perfect man (microcosm) converge as the ‘image’ of God. It is composed of ten divine powers called sefirot (‘[primordial] numbers’, lit., ‘countings’), defined as aspects or attributes of God and systematically associated with parts of his ‘body’, so as to constitute an anthropomorphic whole. It thus effectively depicts God as the ‘sum total’ of his divine powers, ‘gods’…the tree with its entire associated doctrinal apparatus can be shown to be based on a Mesopotamian model perfected in Assyria in the second millennium B.C. That this model could be made an integral part of Jewish religious thought underlines the basic similarity of the Assyrian and biblical concepts of God. (S. Parpola 1997: xxiii)

The sefirot numbers correspond to the sacred numbers connected with the great gods of the Mesopotamian pantheon, yielding a powerful ‘magic diagram’ (see S. Parpóla 1993). The cuneiform texts present also the ‘great gods’ of the Babylonian pantheon as functions, powers and attributes of the Babylonian national god Marduk: as his ‘kingship, might, wisdom, victory, strength, counsel, judgement’ etc. (cf. S. Parpóla 1997: lxxxii n. 23).

Of course, in a religion of this type, the borderline between the surface level (polytheism) and the deeper level (monotheism) is subtle, and the distinction between the two was certainly often lost in practice. Examples are not lacking in Assyrian texts and iconography where Aššur appears as if he were just one god among many— granted, the most exalted one, but still on the same level with other gods. In the prophecies, state cult, and royal inscriptions, however, he is always strictly set apart from his emanations. In accordance with his special status, he is represented as a winged disk hovering over the ‘Tree of Life’. His fundamental unity with his powers is, however, made clear by his seal, where he is said to ‘hold a cosmic bond binding together the great heavens and the Igigi and Anunnaki gods’. (S. Parpola 1997: xxv)

In Achaemenid art, the winged solar disk, one of the principal symbols of Aššur, is taken over by Ahura Mazdā. The conception of Ahur a Mazdā as a monotheistic god with the ‘holy immortals’ representing his qualities or powers is also strikingly similar to that of Aššur. This conception was not a creation of Zarathushtra; as Mary Boyce puts it, ‘the mould was already old in which Zoroaster cast his new doctrines’:

That divine attributes should be isolated, and then invoked and worshipped as independent beings, was a characteristic of the pagan Iranian religion, as we have seen in the case of Mithra, surrounded as he is by Friendship, Obedience, Justice, Courage and Divine Grace. (Boyce 1979: 23)

The ‘holy immortals’ of Zarathushtra correspond to the Ādityas of the Veda: these ‘personalized powers…represented the principles on which human action depended’ (Brereton 1981: ix). Two of the Ādityas, Mitra and Varuṇa (‘Alliance’ and ‘Commandment’ according to Brereton 1981), are invoked in the Mitanni treaty. In addition, the Ādityas comprised such abstract deities as Aryaman ‘Civility’, Bhaga ‘Fortune’, Aṃśa ‘Share’ and Dakṣa ‘Ability’; and, according to Karl Hoffmann (1992: 715–32), also Amartya Gaya ‘Immortal Life’ (cf. R̥V 10,64), which has a counterpart in the Iranian Gayōmart ‘mortal life’, from whom mankind was born.

The worship of deities called Asura ‘Lord’ (= Agni) and *Mazdhā ‘Wisdom’ by the Mitanni Aryans (ca. 1500–1300 B.C.) and by the R̥gvedic Aryans of the oldest layer (ca. 1700–1500 B.C.: the Yadu and Turvaśa tribes associated with the Kāṇvas) respectively is proved by the proper names Ašuratti = / Asurātthi / ‘who has the Lord as his guest’ (Mitanni Aryan from AlalahȾ) and R̥gvedic (Kāṇva) Médhātithi ‘who has the (god) Wisdom as his guest’ (cf. Pinault 1998: 453f.). On the other hand, such personalised principles governing the human social life are completely missing from the catalogue of Scythian deities supplied by Herodotus (4,59). I would like to propose that these abstract deities involving a highly sophisticated conception of divinity were adopted into the ‘Proto-Indo-Aryan’ pantheon of the BMAC from the Assyrians, whose religion is based on the millennial tradition of the Mesopotamian city civilisation. In Assyria, the worship of Aššur as a monotheistic God was associated with mysticism and an ecstatic cult with prophecies; these can be traced back to the religion of the Sumerians and Akkadians in the third millennium B.C. (See S. Parpóla 1997.) Even the choice of the word asura ‘lord’ as the designation of the highest divinity may have been influenced by its similarity to the name of Aššur.

In this connection I would like to take up a piece of evidence, which has been quoted as being in conflict with the early dating of the Iranian sound change *s 〉 h (see section 2.9.5). A Neo-Assyrian ritual text (III R 66 = K. 252 published with a Dutch translation and notes by Frankena 1953: 1–22) and possibly dating from the time of Sargon II (who ruled 721–705 B.C.) mentions, beside Elamite gods, a divine name or two not otherwise known in the Assyrian and Babylonian pantheon, namely ^Das-sa-ra ^D ma-za-áš/Assara Mazaš/, interpreted as rendering Ahura Mazdā. The dating is complicated by the fact that the text with its list of gods is an update of a tradition reaching back to the Middle Assyrian period (ca. 1500–1000 B.C.). As Hintze (1998: 147f.) points out, the text is intriguing, as it ‘might attest an unchanged intervocalic s in Iranian’ and provide ‘a terminus ante quern for the life time of the prophet’. However, she admits that the Gathas attest to the completion of the sound change *s 〉 h by Zarathushtra’s time, and that ‘it may be equally possible that the god name Ahura Mazdā existed in the pre-Zoroastrian Proto-Iranian religion as the name of one among several Asuras’: she quotes Johanna Narten (1982: 66), who has shown that the Avestan invocation mazdā̊scā ahurā̊ŋhō ‘the Mazdā and the (other) Ahuras’ was used secondarily as a petrified formula already at the time when Zarathushtra composed his Gathas.

But when did Zarathushtra compose his Gathas? The ‘Proto-Indo-Aryan’ poetry must have continued also west of the Hindukush after the separation of the R̥gvedic Aryans proper, changing its language into Old Iranian after the coming of the Iranians to these regions (cf. Burrow 1973 quoted in section 1.1.2); this is suggested by the fact that comparison with the Avestan Gathas makes it possible to reconstruct the original wording of R̥gvedic hymns that were altered by later diasceuasts in South Asia (cf. Pirart 1989). It is a fact that the R̥gveda and the Avesta share many poetic phrases (cf. Schlerath 1968; Gercenberg 1972: 90–127).

One formulaic phrase used by Zarathushtra in Yasna 44 (‘O Lord, what I ask you, answer me truly’) resembles an Assyrian extispicy formula (‘O Sun, great lord, what I ask you, answer me with a true yes’). Zarathushtra could have learnt such a formula—and the esoteric monotheistic religion of the Assyrians that he tried to restore among the Iranians—if he was among the Median princes who were brought up by the Assyrians to train them for state service as high officials and to brainwash them to loyalty to Assyria. (S. Parpola, forthcoming; cf. Panaino 2000: 41.) If this indeed was the case, it would most probably have taken place in the eight century B.C. If one compares Old Avestan linguistically with Old R̥gvedic (ca. 1350 B.C.) on the one hand and with Old Persian (ca. 520 B.C.) on the other, this seems a reasonable dating.

The desiccation that occurred around ca. 1750 B.C. must have been a major cause of the collapse of the BMAC at the beginning of its ‘post-urban’ phase, but the arrival of a new wave of Indo-Aryan speaking steppe nomads is likely to have contributed to this development (cf. also Biscione 1977):

At Gonur south, the building complex appears to have been abandoned suddenly at the end of [the ‘urban’ phase of] the BMAC. Valuable grinding stones are found in the upper levels, unbaked ceramics lay near pottery kilns, and storage areas have large khoms still in place. Similar situations appear to be the case at Togolok 1 and Togolok 21, where there is no occupation following the Period 2…occupations. (Hiebert 1994: 129)

In Margiana, the steppe contacts of this phase have shifted to the northwest, the ceramics indicating the presence of Andronovo pastoralists coming from the Tazabagyab culture of Chorasmia (cf. Hiebert 1994: 70). In Tajikistan, Uzbekistan and northern Afghanistan, imported steppe ceramics represent the Fedorovo variant of the Andronovo cultures (cf. Kuz’mina, in press).

The date 1750 B.C. is close to that posited for the first wave of R̥gvedic Aryans, most likely represented by the Vakhsh/Bishkent culture of Tajikistan (cf. P’jankova 1986) and the Ghalegay IV culture of Swat (ca. 1700–1400 B.C., see sections 2.6; 2.10.3). Some of the R̥gvedic Aryans are likely to have belonged to the newly arrived Andronovians, who probably also pressed *Sauma, and worshipped especially Indra, a less ‘civilised’ god of war, while the Ādityas and the Aśvins probably came from the religion of the ‘late urban’ phase of the BMAC.

The impoverished phase of the BMAC continued until ca. 1500 B.C., when the arrival of (in all likelihood Iranian-speaking) mounted pastoralists initiated the Yaz I culture (see sections 2.4 and 2.9.4-5) and probably caused Divodāsa and other R̥gvedic kings of the ‘second wave’ to move to India (see sections 1.1.3; 2.4; 2.6).

Alexander Lubotsky (2001) has collected ‘all Sanskrit etyma which have Iranian correspondences, but lack clear cognates outside Indo-Iranian’, numbering in all ‘some 120’. Lubotsky concludes that some of these words may have been borrowed from Uralic languages but that the major portion is likely to come from the language of the BMAC, which he considers non-Indo-European. One criterion is the non-PIE word structure that many of these loanwords have, being trisyllabic with long middle syllable.

According to Lubotsky, ‘the Indo-Aryans were presumably the first who came in contact with foreign tribes’ in Central Asia ‘and sometimes “passed on” loanwords to the Iranians.’ Lubotsky also points out that these Indo-Iranian loanwords that are likely to have been borrowed mainly from the language of the BMAC are phonologically and morphologically similar to many loanwords found in Sanskrit alone. This ‘indicates that, to put it carefully, a substratum of Indo-Iranian and a substratum of Indo-Aryan represent the same language, or, at any rate, two dialects of the same language’. While I subscribe to this important conclusion, I do not think the explanation offered by Lubotsky is the only possible one to account for this fact. Lubotsky finds himself bound to assume that ‘the language of the original population of the towns of Central Asia, where Indo-Iranians must have arrived in the second millennium B CE, on the one hand, and the language spoken in Punjab, the homeland of the Indo-Aryans, on the other, were intimately related’. In my opinion, it is also possible that Indo-Aryan has simply preserved many more Central Asian loanwords than Iranian, which according to the model proposed earlier in this paper and also by Lubotsky himself, largely received them secondarily from Indo-Aryan, i.e. that branch of Indo-Aryan which remained in Central Asia, Iran and Afghanistan.

While I find it quite likely that the Proto-Indo-Aryans still had contact with the original language of the BMAC (i.e. the language of the ‘Oxus Civilisation’ as defined above in 2.10.1), it may have become largely assimilated into Indo-Aryan by the time the Iranian speakers arrived. It seems probable to me, however, that most of the BMAC loanwords came to Indo-Aryan (including Vedic as well as Classical Sanskrit) from the language of the Dāsas, the Aryan elite of the earliest ‘mature’ phase of the BMAC (cf. section 2.10.2).

I would like to draw attention to one peculiarity of the hypothetical ‘Oxus’ language not specifically mentioned by Lubotsky. The vocalic r̥ occurs in the first syllable, even when it is accented, in several of Lubotsky’s loanwords, such as r̥bī́sa- n. Oven’ and śr̥gālá- m. ‘jackal’ (with Lubotsky’s trisyllabic structure with long middle syllable), *gr̥da- ‘penis’ (in which Lubotsky draws attention to the two unaspirated voiced stops in the same syllable, a root structure impossible for an Indo-European word) and *r̥ši-‘seer’ (with aberrant initial accentuation in Sanskrit ŕ̥ṣi- noted by Lubotsky).

Moreover, the initial syllable containing r̥ is characteristic of a large number of proper names recorded by the R̥gveda for the Dāsas, Dasyus and Panis: Dŕ̥bhīka, Śŕ̥binda, Bŕ̥saya (with the dental s after r̥, as also in br̥sī-f.‘roll of twisted grass, cushion, seat of a religious student or an ascetic’), Br̥rbú (cf. bŕ̥būka- ‘thick’?), *Pr̥ní (the assumed protoform of Paṇí), Pipru Mr̥gaya (Lubotsky lists *mr̥gá- as a likely loan-word).

Bŕ̥saya and Dŕ̥bhīka are particularly important, because they can be located in the regions west of the Indus Valley with the help of Greek sources: Bŕ̥saya = Bαρσαέντης (see section 1.1.2); Dŕ̥bhīka = Δέρβικες, who according to Strabo (11,8,8; 11,9,1) lived as nomads east of the Caspian Sea (they slaughtered people even for slight offences and worshipped Mother Earth with sacrifices of male victims, 11,11,8), to be compared also to Avestan Driβika (Vidēvdāt 1,8), the name of an inimical people in the country of Harōiva (cf. Parpola 1988: 257f.).

This vocalic ŕ̥ in the first syllable need not be interpreted with its Vedic phonetic value. It could stand for example for ěr (cf. Δέρβικες!) or for ŏr, with short vowels that do not exist in Vedic (Bertil Tikkanen, personal communication).

Another peculiarity in the above BMAC/Dāsa words is the phoneme b (cf. r̥bī́sa-, Śŕ̥binda, Br̥bú), which in Indo-Aryan rarely goes back to Proto-Indo-European *b? but occurs in many words of probably foreign origin (cf. Wackernagel 1896: I, 181–4). Among these words are further Dāsa names, Balbūthā and Ilībíśa. These again have the phoneme l not found in the main dialect of the R̥gveda. Both b and l occur in kílbiṣa- ‘injury, transgression; injustice, sin, guilt’, which Watkins (1995: 398f.) has shown to be a borrowing from the Dāsa language (cf. R̥V 5,34,4 and 6), originally denoting some hostile behaviour or action taken over from the enemy language.

A few other probable BMAC loanwords outside the Indo-Iranian material discussed by Lubotsky may also be mentioned. The following words have the trisyllabic structure with long middle syllable as well as the phoneme l:

kúlāla- m. ‘potter’ (VS), kaulāl´- m. ‘potter’ (VS, ŚB), n. ‘potter’s ware’ (ĀśvGS); this word is also found in the northwest, in the Dardic languages Pashai and Kashmiri, as well as in Iranian Parachi (cf. CDI AL 3341). The potter was not an Ārya but a Śūdra and vessels turned on a wheel belonged to demons (asuryà-); all clay vessels needed in the Vedic ritual had to be made by hand only by an Ārya (cf. Rau 1972: 13).

palālī́- f. ‘stalk, straw’ (AV), pílāla- n. ‘stalk, straw’ (Manu), ‘millet straw’ (Suśruta), widely attested in Middle and Neo-Indo-Aryan (cf. CDIAL 7958).

kīlā́la- ‘some kind of milk product’ (Kuiper 1991: 14) (R̥V 10,91,14 kīlāla-pá-; seven times in the AV), Classical Sanskrit kilāṭa- (Suśruta); of modern languages, the word is found almost exclusively in Kafiri and Dardic, where it mostly means ‘fresh cheese’; otherwise it survives only in Sindhi kiroṭu, kirūṭu ‘cheese made from skimmed milk’ (see CDIAL 3181; Fussman 1972: II, 175–7). Although the word is found also in Burushaski as kīlāy (and is therefore often considered to be a loanword from Burushaski, e.g. by Witzel 1999d) and in Dravidian (Tamil kiḻaāṉ ‘curds’, etc., see DEDR 1580), an Oxus’ origin seems preferable to me.

In 1991 F.B.J. Kuiper published a useful booklet on the ‘foreign’ linguistic and cultural elements in the R̥gveda, culminating in a list of 383 ‘foreign words’ (pp. 90–93). Rahul Peter Das (1995) has published a long and useful review, though it is unjust in being one-sidedly negative (cf. also Kuiper 1995).

Basing himself on Kuiper’s work, Michael Witzel (1999b; d) argues that the great majority of the foreign words in the Rgveda is derived from Austro-Asiatic, more precisely from a hypothetical language that he calls ‘Para-Munda’. According to Witzel, the amount of ‘Para-Munda’ loanwords in the R̥gveda, borrowed when the R̥gvedic Aryans arrived in the Panjab, indicates that it was the language of the Indus Civilisation. However, hardly a single concrete etymology presented has a wider support; one of Witzel’s main sources is Kuiper’s Proto-Munda words in Sanskrit (1948), denounced by most experts including Kuiper himself (he did not even refer to it in Kuiper 1991); general comparisons of R̥gvedic words beginning with ka-, ku- etc. with hypothetical Austro-Asiatic prefixes of the same shape do not qualify as any sort of proof. Witzel’s hypothesis of an Austro-Asiatic homeland in the Panjab is aptly criticised by George van Driem:

The Punjab is not only far away from the geographical centre of gravity of modern Austro-Asiatic language communities, the Punjab in the far northwest is beyond the range of any modern or historically attested Austro-Asiatic language community. (Driem 1999: 74)

According to Driem, the Austro-Asiatic homeland lay in the northeast, along the Brahmaputra and around the Bay of Bengal.

In his recent M.A. thesis (1998), Bryan Wells claims that there are fundamental differences between Indus inscriptions coming from the Panjab and Sindh. Partly on this basis, Witzel suggests that the Indus people spoke two different languages, ‘Para-Munda’ in the Panjab and Dravidian in Sindh. This is hardly correct, however, for from the identity of the sign sequences at Harappa and at Mohenjo-daro it is quite certain that the vast majority of the Indus inscriptions in the Indus Valley has only one language.

Elsewhere I have argued at great length that the written and majority language of the Indus Civilisation belonged to the Dravidian family (Parpóla 1994) and there is no need for repetition here. (For a recent critical assessment of the question ‘Dravidian substratum influence or convergence?’, see Hock 1996.) Witzel attempts to discredit the presence of Dravidian speakers in the Panjab by claiming that there are no Dravidian loanwords in the earliest of his three textual layers of the R̥gveda; Dravidian loanwords appear only in the second and third layer. (In his detailed examination of the material, however, he himself admits that the likelihood of one Dravidian loanword even in this small layer is great: this is ukha- ‘hip’ in R̥V 4,19,9: cf. DEDR 564.) I hope to have shown (in section 1.4) that Witzel’s ‘earliest layer’ has no real foundation, and that the latest hymns of the R̥gveda actually reflect the earliest layer of Vedic Aryans in South Asia.

A large part of the ‘foreign words’ of the R̥gveda, claimed to be ‘Para-Munda’ by Witzel, are likely to come from the BMAC language, which seems to have been introduced to the Indus Valley around 2000 B.C. (cf. section 2.10.2). The peacock is native to South Asia, but the structure of the word (trisyllabic with long middle syllable) suggests that mayū́ra- is probably an ‘Oxus’ word, whatever its ultimate origin: both Dravidian and Austro-Asiatic etymologies have been proposed (cf. DEDR 4642 and CDIAL 9865).