Notes on the historical phonology of Indo-Iranian loanwords in Northwestern Tibetan dialects

Размещено на http://allbest.ru

2

Institute of Oriental Studies, Russian Academy of Sciences

Notes on the historical phonology of Indo-Iranian loanwords in Northwestern Tibetan dialects

Anton Kogan

Moscow

Abstract

Recent research has shown that the vocabulary of certain Northwestern Tibetan dialects contains a significant number of Indo-Iranian loanwords. It is, however, still unclear if these loanwords have been borrowed from a single or from several sources, and whether their presence is the result of substratum or adstratum interference. Likewise, the exact genetic position of the donor-language(s) within the Indo-Iranian group so far remains undetermined.

The study of all these issues should, no doubt, be based on facts of historical phonology. In this article I attempt to identify the most conspicuous historical-phonological features of the Indo-Iranian elements found in Northwestern Tibetan varieties. Furthermore, I make some preliminary conclusions concerning the linguistic geography of the region in the pre- Tibetan period, as well as the direction and relative chronology of Tibetan migrations.

Keywords: historical phonology; language contact; linguistic substratum; Indo-Iranian languages; Dardic languages; Tibetan dialects; Ladakhi language; Balti language; Purik language.

Introduction

A recent etymological study of the vocabulary of certain Northwestern Tibetan varieties, namely Ladakhi, Balti and Purik, has revealed the existence of a significant Indo-Iranian lexical stratum in these dialects (Kogan 2019).

This discovery is to some extent in line with the hypothesis put forth more than a century ago by the German scholar August Hermann Francke, and still popular among Tibetologists, according to which the pre-Tibetan population of Ladakh and adjoining areas was somehow related to the present-day speakers of Dardic languages (Francke 1907).

It should, however, be kept in mind that the overall picture of ethnic and linguistic history of the region is still far from clear. The presence of Indo-Iranian loanwords in the Tibetan dialects of Ladakh and Baltistan raises more questions than it answers.

We do not know if these loanwords have been borrowed from a single source or several sources, whether their existence is the result of substratum or adstratum interference, Based on our present knowledge of language situation in the Tibetan Empire and later northwestern Tibetan kingdoms, the influence of an unknown Indo-Iranian superstrate on Tibetan dialects should be considered improbable. and finally, which branch or branches of Indo-Iranian they represent.

It goes without saying that the answers to these questions must be based on facts of historical phonology.

The only possible way to establish the exact genetic position of the donor language is to compare its historical-phonological peculiarities with those of all the four known branches of the Aryan subfamily, namely Indo-Aryan, Iranian, Dardic and Nuristani.

The strongest indication of borrowing from several related lects is, no doubt, the presence of more than one type of phonological development in the same position of the word, inexplicable by secondary processes like analogical changes.

The choice of substratum or adstratum alternative can hardly be made either without taking into account historical-phonological data. In our case, such a choice is actually a complex task.

The present-day Indo-Iranian neighbors of Northwestern Tibetan dialects are two Dardic languages: Kashmiri and Shina. The Ladakhi and Purik dialects are also in contact with Brokskat, a Dardic language genetically close to Shina. However, as shown in the above-cited article (Kogan 2019), the speakers of this language must be compara-tively recent migrants to their present habitat. There are strong reasons to believe that both of them spread to their present-day speaking areas in the Middle Ages due to ethnic migrations.

The population of Kashmir until the 13 ^th century spoke an Indo-Aryan language, most likely closely related to West Pahari dialects (Kogan 2016), whereas in the now Shina-speaking valleys of Karakoram the language spoken one thousand- plus years ago was probably Burushaski (Jettmar 1975).

It should be born in mind that Indo- Iranian influence on Tibetan varieties under study, whatever its nature, cannot be exactly dated, even though the Tibetan conquest of Ladakh and adjoining areas in the 8^th century A.D. provides a terminus post quem for the process.

In light of this fact, it cannot be ruled out that at least a part of loanwords were borrowed in the 2^nd millennium A.D. directly from Shina and/or Kashmiri.

Another possible source of lexical loans may have been East Iranian Saka dialects spoken before the 11^th century A.D. in the western part of the present-day Xinjiang, immediately to the north-east of Ladakh and Baltistan. It is, however, worth noting that a tentative attempt to find Irainan phonological features in Ladakhi, Balti and Purik words of Indo-Iranian origin made in Kogan 2019 was not successful.

To sum up, borrowing from an ad- strate language should be considered probable if a particular Indo-Iranian word shows Indo- Aryan, East Dardic (Shina or Kashmiri) The existence of the East Dardic subbranch consisting of Kashmiri, Shina and Kohistani languages was first hypothesized by George Abraham Grierson (1906; 1919). For the latest views on this group and its historical-

phonological features see Kogan 2015; Kogan 2016. or Iranian historical-phonological traits. In the following sections an attempt will be made to identify the most conspicuous historical-phonological features of Indo-Iranian elements found in Northwestern Tibetan dialects. The material studied here as well as lexical comparisons are drawn chiefly from Kogan 2019, where the re-spective etymologies are also discussed. If a word not mentioned in this article is analyzed, its etymology will be discussed in the present text. First, I shall address the development of vowels, syllabic sonorants and some sequences containing vowels and sonorants, then the development of single consonants and consonant clusters. Finally, some preliminary conclusions will be drawn.

Vowels and syllabic sonorants

Vowels (monophthongs and diphthongs). There is no phonological vowel length in Northwestern Tibetan dialects. That is why PII *a and *a are usually reflected in the same way:

1. Ladakhi asur `mustard seed as a spice used in Ladakhi pickle' (Norman 2010: 1073). For ease of reference, etymologies are numbered consecutively. Cf. OIA asun `Sinapis ramosa', Sindhi ahuri `mustard seed', Lahnda фhur, ahьr, ahur, Punjabi ahur, Hindi-Urdu asurп `mustard', Kashmiri asoru `plant, Sinapis ramosa'.

2. Balti basanda `dandelion' (Sprigg 2002: 27). Cf. OIA vasanta- `vernal, pertaining to spring', OIA vasanta-, Shina bazцn, Phalura basand, Bashkarik basan, Torwali basan, Kalasha basun, Khowar bosun, Pashai wahon(d), Gawar-Bati wasand `spring'.

3. Balti bat `boiled mixture of germinated grain flour and ordinary flour (made during Ramzan)' (Sprigg 2002: 27). Cf. OIA bhakta- `food; boiled rice', Khowar bot `evening meal', Kalasha batay `flour taken to be eaten in high pastures' (Trail 1999), Shina bat, Bashkarik batt, Torwali bat, Kashmiri bati, Lahnda, Punjabi bhatt, Hindi-Urdu, Nepali, Gujarati, Marathi bhat `boiled rice', Wakhi bat `wheat flour gruel', Ishkashimi bat, Shughni bat, Munji batdk `ritual food made of wheat flour and butter, cooked in water or milk'. The above-cited East Iranian words are most probably borrowed from some Dardic source.

4. Balti, Purik chal `overflow, spill over', Ladakhi (Leh and Shamskat dialects) chal-ces `to splash, to spill over' (Sprigg 2002: 41; Norman 2010: 297). Cf. OIA ksarati, ksalati `flows, trickles', ksalayati `washes', Proto-Iranian *xsar- `to flow' (> Persian sarпdan `to trickle', (ab)sar `waterfall', Ossetic дxsдrdzдn `waterfall' (< *xsar-cana-), Middle Persian Xsart `the river Jaxartes' (Livshits 2003)), Kashmiri chalun `to wash', char `a sprinkle of water etc. from the fingers'), Gawar-Bati char `rapids in a stream', Phalura, Indus Kohistani char, Bashkarik, Kalasha uchar, Shina char `waterfall'.

5. Ladakhi darak, Purik deraq `stiff, hard' (Norman 2010: 438) < PII *dharaka-. Cf. Kashmiri dor `firm, hard, strong, compact, durable, solid', darun `to become steady (of something in motion), to become firm, to stand steady' (Grierson 1915-1932: 238, 246), Pashai d(h)ar-, Wotapuri dar-, Phalura dhвara- `to remain, stay', Torwali derп `they remained', Khowar dorik `hold back, wait, keep', Hindi, Punjabi dharna `to keep', Gujarati dharvь `to hold, catch', OIA dharayati, dharati `holds, keeps', Av. daraiiehi `(you) hold'.

6. Balti gzar `to flow' (Sprigg 2002: 72), Purik zar `id.', Ladakhi zar-ces, dzar-cas `to drip, to run down, to trickle out' (Norman 2010: 790). Cf. Av. yzar- `to flow', Ossetic gzжlyn `to pour down, drip', OIA jhara- `waterfall', jharп `river', Prakrit jharaп `drips', Hindi jharna `to ooze, trickle away' < PII *gjhar- < PIE *d^hg^wher- (Cheung 2007: 124) or *g^ugh'er- (LIV: 213-214).

Etymological *u and *u also merge:

7. Balti, Purik, Ladakhi multuk, multuk `fist' (Norman 2010: 705; Sprigg 2002: 118) < *multak < *mustaka-.<sup>8</sup> Cf. OIA musti-, Av. musti-, Khotanese mustu, Sindhi muthi, Lahnda, Punjabi mutth, Hindi-Urdu, Gujarati, Marathi muth, Nepali muthi, Persian most, Shughni mut, Wakhi most, Shina mut(h), Kashmiri moth, Phalura, Kalasha, Khowar musti, Gawar-Bati mustak, Burushaski (loanword) muc `fist'.

8. Balti mulak, mulu, Purik mulaq `turnip' (Sprigg 2002: 118; Zemp 2018: 944). Cf. Burushaski mulo, Shina muulo `id.', OIA mula- `root', mulaka- `radish', Shina mьlп, Khowar mu\ `root', Pashai muluk, Hindi-Urdu mulп `radish' < PII mula- `root'. For more details on the change *st > lt in Northwestern Tibetan dialects see Kogan 2019. Probably, related to Middle High German mul, German Maul `muzzle' (Mayrhofer 1996: 369). No reflexes of PII initial or medial *i and *п have been attested in our material. There is, however, one instance of the drop of * п in the word-final position (see 1.). As we can see, Balti shows double reflexes of PII word-final *a. This vowel is sometimes preserved (cf. basanda `dandelion') and sometimes dropped (cf., e.g. bat `boiled mixture of germinated grain flour and ordinary flour (made during Ramzan)'). It cannot be ruled out that the difference in reflexes is due to different borrowing sources. In one example, we find an unusual vowel sequence aa in Balti, which may correspond to short a in Old Indo-Aryan as well as in Iranian and Dardic:

9. Balti baan `man or men who sing religious songs and foretell the future' (Sprigg 2002: 24). Cf. Kashmiri wan-, Indus Kohistani ban- `to say', Kalasha bandek `1. to teach; 2. To announce so as to inaugurate, to order or command' (Trail 1999: 27), OIA vandate `praises, worships', Av. vand-, Khotanese van-, Parthian wynd- `to praise, honor, worship'. It is, however, quite possible that the source of the Balti word reflects some lengthened- grade derivative of the above-cited root. Such derivatives are attested in several Dardic languages. Cf., e.g. Wotapuri -ban `sprechend, Sprecher' in alik-ban `Lьgner' (alik `Lьge') with the medial a being the regular continuant of the etymological *a (Buddruss 1960: 20, 87, 92). Before a final nasal the change *a > o takes place:

10. Balti (Skardu dialect) dom `sadness, trouble, difficulty, adversity' (Norman 2010: 493). Cf. Shina damizhar `adversity, trouble' (Bailey 1924), Burushaski dnm(i)jar `trouble, inconven- ience, worry' (Lorimer 1938), The Burushaski word is likely to have been borrowed from Shina. OIA damayati `tames, subdues', Ossetic domyn `to tame; exhaust; demand'.

11. Balti chon `vain' (Sprigg 2002: 43), Purik chon (“tfnn”) `useless, in vain, unfounded' (Zemp 2018: 109), Ladakhi chon `spontaneously, for no reason, gratuitously; in vain, for nothing, to no avail, useless; free, for free, at no cost' (Norman 2010: 310). Cf. Bashkarik chan, Phalura cheniko (e < a in i-umlaut position), Shina chon, Kashmiri chon (< *chanu with o < a in u- umlaut position), Gawar-Bati chлnika (e < a in i-umlaut position), Pashai (Kurangali dialect) chдni (a < a) `empty, void', Burushaski chan (Hunza), can (Yasin) deer (Schьssel), unterbeschдftigt, ohne Arbeit, frei' (Berger 1998: 106). In theory, there is a possibility that the Balti and Purik words have been borrowed from Shina, whereas the Ladakhi word is a loan from Kashmiri. Should this be the case, the Ladakhi example cannot illustrate the aforesaid historical phonological process, because in Kashmiri the change *a > o was caused by м-umlaut rather than the fol-lowing nasal. Semantic data, however, show that separate borrowing of the lexeme under analysis is rather unlikely. In all the three Tibetan varieties, this lexeme has undergone the same semantic development (`empty' > `(in) vain'). Technically, it cannot be ruled out that such a development took place independently in all the dia-lects, but it seems more probable that it occurred either in the donor language before borrowing or after borrowing but before the divergence of Ladakhi, Balti and Purik. In either case, the source language could hardly have been Kashmiri. In the latter language the development of u-umlaut dates back not earlier than to the 17^th century (Ko-gan 2016), i.e. to a period when linguistic Tibetanization of not only Ladakh but also Baltistan was, no doubt, an accomplished fact. This said, direct borrowing from Kashmiri into Balti is scarcely possible, because these two lan-guages, being geographically separated by the Shina-speaking area, are not in close contact with each other.

See also 48.

This change does not occur before clusters and the historical vowel *i. The etymological a is preserved in these positions:

12. Balti zan-zos, dzan-zos (Turtuk dialect), Purik zan-zos `wife; family' (Norman 2010: 842; Sprigg 2002: 180). This word is most probably a compound. Its second component may be connected with Tibetan tshos `color' (Kogan 2019). Cf. Phalura jeeni `female person' (Liljegren, Haider 2011: 76), Kashmiri zdn', Sindhi, Lahnda, Punjabi janп `woman', Bashkarik jin kar- `to marry', Indus Kohistani zhal `marriage' (Zoller 2005: 202), OIA jani-, Av. faini- `woman, wife'.

See also 2.

The Proto-Indo-Iranian diphthong *au is monophthongized (*au > u):

13. Ladakhi, Purik kulik `lock; key; the joint of the jaw' (Norman 2010: 12) < PII *kaula-, *kaulikд- `curved' (= Proto-Iranian *kaura- > Khotanese kьra- `crooked' (Bailey 1979: 62)). Cf. Ti- rahi koold, Pashai kola, Shumashti kolдnta, Khowar koli, Bashkarik kцl, Torwali kцl, Phalura kь- ulo, Shina kцlu `curved, crooked'³ For the semantic change `crooked' > `key' cf. Greek кАціс,, Latin clдvis, Old Church Slavic kl'ucb (ключь) `key' < PIE *klлu- `Haken, krummes Holz' (Pokorny 1959: 604-605)..

Syllabic sonorants. The Proto-Indo-Iranian syllabic sonorant *r is vocalized in interconsonantal position:

14. Ladakhi kit-ces `to catch, seize, capture' (Norman 2010: 82). Cf. Bashkarik gоta `took, bought', Wotapuri gat `took', Gawar-Bati gьtim `I took' < *grpta- < *grb-ta- (Vedic grbdha-, Av. g or opta- `taken').

15. Ladakhi shen-ces `to squash' (Norman 2010: 986). Cf. OIA smдti `crushes, breaks', Av. a-saro-ta- `not discouraged (lit. `not broken')', Khowar senik `to crush' < PIE *k'erhr- `brechen, zerbrechen (intr.)' (LIV: 327-328).

The vowel e in the latter example may have been a free allophone of i in the donor language. Cf. the free variation of i and ё after s in Kashmiri.

Vowel sequences and sequences with intervocalic sonorants. Secondary vowel sequences which appear due to deletion of intervocalic consonants contract to monophthongs.

*a+*u > o:

16. Ladakhi po-ze `ram, full-grown male sheep' (Norman 2010: 540) < *pau- < PII *pasu- `cattle' (> OIA pasu-, Av. pasu- `id.', Pashto psd, Ossetic fis `sheep'). As can be seen, the development of this sequence differs from that of the Proto-Indo-Iranian diphthong *au. This fact implies that the two sound changes could hardly be simultaneous. The element -ze in the Ladakhi word is historically a suffix found also in Classical Tibetan (Kogan 2019).

*a+*i > e:

17. Ladakhi tshele `thatched roof, shade made of branches and leaves, thatched-roofed house or hut' (Norman 2010: 773), Balti tshele `hut' (Sprigg 2002: 170) < *scaila- < PII *scad-ila-. Cf. OIA chadayati `covers', chadis- `cover, roof', Lahnda, Punjabi chatt, Hindi chat `roof', Kashmiri chey `a kind of thatching grass' (Grierson 1915-1932: 1066), Proto-Iranian *sad- `to cover' (> Pashto psoldl `to adorn' from *upa- or *pati-sad-) < PIE *sk'ed-.

Sequences containing intervocalic sonorants, both historical and secondary, yield monophthong vowels.

*aya > e:

18. Ladakhi shen (Leh dialect), sen (Shamskat dialect) `(wooden) floor' (Norman 2010: 986) < PII *srayana- `leaning, foothold'. Cf. OIA sray-, Av. sray- `to lean'.

*awa > o:

19. Ladakhi phok `incense or burning juniper leaves or other fragrant burning materials' (Norman 2010: 580) < PII *pawaka- `purifier'. Cf. OIA pavayati `purifies', Persian pak `pure' (< *pawaka-).

*aiya > e:

20. Balti be-kar `court singer and dancer who improvises poems and songs' (Sprigg 2002: 28), Ladakhi be-da `member of the caste that used to be itinerant musicians', be-mo f. (Norman 2010: 640-641) < *baiya- < PII *wad(i)ya- `music, musical instrument', cf. OIA vadya- `musical instrument, music', vadyakara- `musician', vadayati `plays a musical instrument', Punjabi vajja `musical instrument', Hindi-Urdu baja `music', Kashmiri waz- `to sound (of bell, clock etc.)', way- `to play a musical instrument', Shina (Guresi) baz- `to strike (of a gong etc.)', Ossetic wadynj `flute, panpipe' < *vвddnicп.

Vowel assimilation. Vowel assimilation, both progressive and regressive, was noted:

21. Balti men-ze `lump of dough' (Sprigg 2002: 113), Ladakhi men-ze `ball of dough ready to be shaped and cooked, flattened dough ball, or shaped bread ready to be cooked' (Norman 2010: 710). Cf. Burushaski maano grцsserer Teigklumperi (Berger 1998: 272), Shina mino `id.', Brokskat manпli, Dameli man `bread', Bashkarik man, Phalura mandili `very soft bread' (Morgenstierne 1940; 1941), Phalura mand- `to knead', Romany ma(n)ro `bread', Sindhi manп `bread, loaf, food', Punjabi manda `a thick cake', West Pahari (Bhalesi) manni `a large cake', Assamese mar- `to knead (dough)', Maithili mar- `to knead', OIA mrdnati `crushes, kneads, rubs', Av. mцrdnddn `(they) destroy'.

22. Balti monthok `clod of earth' (Sprigg 2002: 118) < *mat-thok? Cf. OIA mrttika- `earth, clay', Hindi-Urdu, Punjabi mittп, Nepali mato, Dameli mathi `id.', Phalura methi, Shina mati `clay'; Burushaski thфko `Klumpen, Ballen (von Zucker, Salz, Butter)'' (Berger 1998: 452), Indus Kohistani thokh `a clod (earth, salt)' (Zoller 2005: 220).

See also 7.

Consonants

Single consonants. The devoicing of the historical initial voiced consonants, characteristic of Central and Upper Ladakhi dialects including Leh, affected several Indo-Iranian loanwords:

23. Ladakhi pat, paddi, patsi `totally, completely' (Norman 2010: 533) < *bad. Cf. OIA bad- dha- `bound', Gujarati bдdhь `whole, entire', West Pahari (Kotgarhi) baddho `all, entire, (pl.) all together', Kashmiri bod `handful'.

24. Ladakhi tul `powdered dung' (Norman 2010: 448) < *dul. Cf. OIA dhuli- `dust, powder', dhulika- `pollen, fog, mist'; Prakrit dhuli, Hindi dhul, Punjabi, Lahnda, Gujarati dhur, Sindhi dhuri, Kumauni dhuli, Bengali dhul, Marathi dhul `dust'; Nepali dhulo `dust, powder'; Tirahi "dщda", Kalasha udhru, Shina ьdu, Phalura duri `dust'; Pashai (Wegali dialect) durп `dust-storm'; Torwali dur `mist'.

See also 14.

The above examples give reason to believe that the process of borrowing may have begun before and finished after the devoicing of the initial mediae (Kogan 2019).

When not devocalized, the etymological initial voiced aspirated obstruents lose aspiration and merge with their plain voiced counterparts. Thus, *bh > b, *dh > d. See 3, 5.

The process of deaspiration, followed by dentalization and assibilation, seems to have affected the Proto-Indo-Iranian affricate *jh (*jh >*j > (d)z): gzar `to flow' < PII *gjhar- (see 6).

25. Another probable instance of this sound change is Ladakhi zanggi `a species of tiny flying biting insect' (Norman 2010: 825). This word is likely to reflect some derivative of PII *jhan- `to strike, kill, injure' (> OIA han-, Av. jan- `id.'). Cf. the semantic development of this root in some Iranian languages, e.g. Talysh zan- `to bite, sting (of insects and snakes)' (Ras- torguyeva, Edelman 2007: 136). If we assume the same development for the Indo-Iranian lect from which the Ladakhi noun has been borrowed, we may consider the word in question to be the reflex of PII *jhan-aka- with the meaning `sting, stinger' to which the adjectivizing suffix -п (< PII *-in/-п) has been added.

26. This suffix was widely used, e.g. in Old Indo-Aryan, to form adjectives with a possessive sense, often prone to nominalization: kesin-, Nom Sg kesп `longhaired' (cf. kesa- `hair'), paksin- Nom Sg paksп `winged; bird' (cf. paksa- `wing'), hastin- Nom Sg hastп `possessing a hand; elephant' (cf. hasta- `hand').

27. The original meaning of the source form of the Ladakhi word should thus have been “possessing a sting”.

The elision of the intercon- sonantal short a (*jhanaka- > *zang-) is not uncommon in some Dardic languages (cf., e.g. Pashai suqg `dog' < *sunaka-, sa^g `earth' < *samaka-). This sound change must have resulted in the formation of the secondary cluster ng, which in turn must have precluded the change a > o in the initial syllable.

Other etymological palatal affricates are also dentalized word-initially (*c > c (=ts),¹⁵ *j > (d)z): Balti, Purik and Ladakhi words are transcribed using the standard Tibetological transcription system, where c and ch stand for the voiceless palatal and voiceless palatal aspirated affricate respectively, while their den-tal counterparts are conveyed by ts and tsh. In Dardic and Iranian examples, as well as in Proto-Indo-Iranian re-constructions, c, ch mark the palatal affricates, while c, ch are used for the dental ones.

28. Ladakhi tsapik `a little, a bit, a little while' (Norman 2010: 731) < PII *cap- `to catch, snatch, pick, pinch', cf. Shina cap-, Indus Kohistani cap-, Gawar-Bati cep- `to bite', Pashai cip- `to bite off', Kashmiri cop `a bite', Balochi camp- `to snatch', Ossetic cжfsyn `to stick, glue', Munji cob- `to pluck'.

See also 12, 17.

The initial palatal affricates in turn usually correspond to retroflex affricates or their regular reflexes in the Dardic languages (see 4, 11).

The two above types of correspondences may have resulted from a kind of affricate shift involving the dentalization of the historical palatal affricates and the concomitant palatalization of the retroflex ones.

A similar historical-phonological process has affected the affricate system of Kashmiri (Kogan 2016).

It is worth noting that the initial palatal affricates in Burushaski loanwords are sometimes preserved as such:

29. Ladakhi cancil, chancil `the green outer shell or fruit of walnut' (Norman 2010: 267). Cf. Burushaski chanjil (Nagir, Hunza), canjil (Yasin) In Berger's notation c stands for the voiceless palatal affricate. `die grьne дussere Schale der Walnuss, Hдutchen zwischen den Teilen des Walnusskerns', Shina chaciil `id.' (Berger 1998: 96).

30. Balti cangti `drop', Purik, Ladakhi (Shamskat and Leh dialects) cangti `leak in a roof' (Norman 2010: 267). Cf. Burushaski Chan man-, Shina chachan- `(Wasser) sickern, herabfliesseri (Berger 1998: 97).

31. Balti cha `millet' (Sprigg 2002: 41), Ladakhi (Shamskat dialect) cha `a variety of millet, a cereal grain which was grown in Ladakh in the past, but very little now' (Norman 2010: 290). Cf. Burushaski cha (Nagir, Hunza), ca (Yasin) `Hirse, Kolbenhirse, Setaria italica' (Berger 1998: 95).

Etymological intervocalic stops, both voiceless and voiced, are dropped in a number of cases:

32. Balti (Skardu dialect) spa `taste, tasty (neutral), (sexual) enjoyment', Purik, Ladakhi (Nubra dialect) spa `taste, flavour' (Norman 2010: 542; Sprigg 2002: 159) < PII * swada-. Cf. OIA svada- `taste', Iranian *x<sup>w</sup>ada- (> Persian xva `good taste', Balochi wad `salt'), Shina ispavu `tasty'.

33. Ladakhi perak `Ladakhi women's head-dress, covered with turquoise and coral' (Norman 2010: 539) < PII *paridhaka-. Cf. OIA paridha- `to put on (clothes)', Sindhi paharanu, Nepali pairanu, Hindi pahirna `to put on, wear', Kashmiri pOrun `to put on; to adorn, ornament', Khowar purduik `to cover oneself, put on a cloak'.

34. Ladakhi shanti `a leafy vegetable' (Norman 2010: 975) < PII *saka- `green vegetable'. Cf.

OIA saka-,Shina, Indus Kohistanisa, Phaluraso, Bashkariksa, Kashmirihakh, Kalasha sak,

Khowar sax, Waigali, Dameli ca `id.', Indus Kohistani sari `a green vegetable with round leaves' (Zoller 2005: 378) < PIE *k'eko- (Mayrhofer 1996: 628).

See also 17, 20, 48.

Intervocalic voiceless stops, however, are sometimes preserved or sonorized. Such instances require a separate analysis. The retention of intervocalic k is observed exclusively in probable reflexes of the reconstructed formations with the suffix *-k-, i.e. masculine nouns in *- aka- or feminine nouns in *-ika-:

35. Ladakhi kuruk `donkey foal; foal' (Norman 2010: 12), Balti bong-kuru `donkey colt' (Sprigg 2002: 30) < *kurtaka-. Cf. Kalasha kuak (< *kurak) `child (male or female); the offspring of a human or animal' (Trail 1999), Shumashti kur, Dameli kurn `child', Ashkun kuro `child, foetus', Kati kru, kuruk `young of animals', Prasun kyьru `young of animals, child' < *kurtaka-, Kurdish kurr `son', Middle Persian kurrak, Persian kurra `foal' < *kurna(ka)-.

See also 5, 7, 8, 13, 19, 31, 48.

All such cases can be easily explained, if we assume that the suffix *-k- retained its productivity in the donor language for a long time and was perceived as a separate morpheme when intervocalic consonants were dropped. In this connection, an interesting fact is that this formant has probably been suffixed to some roots of Burushaski origin:

36. Balti kulak `meal (quickly made mixture of buttermilk and flour)' (Sprigg 2002: 94), Purik kholak `a certain dish', Ladakhi (Leh, Shamskat and Nubra dialects) kholak, qholak `ready-to-eat dough of roasted flour' (Norman 2010: 117). Cf. Burushaski (d)-q(h)ul-an- (Hunza, Nagir), d-xul-an- (Yasin) `(Teig) kneten' (Berger 1998: 357).

The sonorization of the historical intervocalic t can be observed after a non-etymological nasal, or perhaps nasalization:

37. Balti rindi `lead, bullet' (Sprigg 2002: 139), Ladakhi rindi `bullet; lead (metal)' (Norman 2010: 911) < PII *nti- `flowing, melting, (metal) casting' Cf. Bashkarik nd, Torwali zit (z < r) `brass', Shina nil `brass, bronze, copper', Gawar-Bati rоt `copper', OIA rпti- `stream; yellow brass, bell-metal'.

The development of secondary nasals and nasalization is a well-attested phenomenon in a number of Dardic, Indo-Aryan and East Iranian languages.

In the language under study rhotacism, i.e. the change of PIE *l to r, seems not to have been a regular process in intervocalic position. Like in Nuristani, Dardic and most Indo-Aryan dialects, the distribution of reflexes of the two sonorants is not always clear. Possible, albeit not always provable cases of retention of PIE *l as well as lambdacism (the change of PIE *r to l) have been attested:

38. Ladakhi puli, polo `Ladakhi biscuits of a particular type' (Norman 2010: 541). Cf. OIA pura- `cake', pauli- `a cake of scorched grain and ghee', Sindhi, Punjabi, Hindi, Kumauni purп, Gujarati, Marathi purп `fried cake', Kashmiri pur `a kind of cake fried in ghee' < PIE *pur- `corn, wheat' (Gamkrelidze, Ivanov 1984: 657).

See also 4,¹⁷ 8.

PII *s, both prevocalic (word-initial) and intervocalic, is preserved:

39. Ladakhi sale `knitting needle' (Norman 2010: 1005). Cf. Phalura sileni `needle', OIA sпvyati `sews', Khotanese hпya `sewn stuffs', Ossetic x^wiyin `to sew'.¹⁸

See also 2.

PII initial *s (< Proto-Aryan¹⁹ *c < PIE *k') is retained:

40. Balti shang `wisdom, sense' (Sprigg 2002: 151)<sup>20</sup>, Purik say `consciousness' (Zemp 2018: 931), Ladakhi shang `alertness, awareness, caution, prudence' (Norman 2010: 553). Cf. Burushaski (loanword) say `awake, aware; care, heed, attention' (Lorimer 1938: 322), Shina son `care, anxiety; awake, alert' (Bailey 1924), Khowar sang `fear, suspicion' (Morgenstierne 1973), OIA sanka- `fear, distrust', sankate `is afraid, distrusts' < PIE *k'enk- 4n der Schwebe sein, hдngen (intr.)' (LIV: 325).

See also 15, 32. Possibly, PII *ksar- (>Balti, Purik, Ladakhi chal `overflow, spill over', see 4) is related to PII *gjhar-, reflect-ing another variant of the same Proto-Indo-European root (Rastorgueva, Edelman 2007). Burushaski sel `Nadel, Stecknadel' (Berger 1998b: 377) may have been borrowed from the same Indo- Iranian source as the Ladakhi word. The terms “Aryan” and “Indo-Iranian” are not used as synonymous in the present work. Following D.I. Edelman (1992), we believe that the split of the Aryan unity began with the separation of the Nuristani branch. The remaining three branches, i.e. Indo-Aryan, Iranian and Dardic, form the Indo-Iranian unity, whose disintegra-tion took place at a later date. Thus, Indo-Iranian is considered a lower-level taxon within the Aryan subfamily. In traditional Tibetological transcription sh stands for voiceless palatal sibilant.

Preliminary conclusions

The above analysis has revealed a number of facts that are directly relevant to the questions formulated at the beginning of the present work. First of all, four apparent cases of nonuniform phonological development in the same position have been detected. These are:

1) PII medial *a or *a > aa in Balti baan `man or men who sing religious songs and foretell the future' but > a elsewhere;

2) PII final *a preserved in Balti basanda `dandelion' but dropped elsewhere;

3) the initial r of PII *-rt- has moved to the word-initial position in Balti rkat `to cut down with a sword', whereas in Ladakhi kuruk `donkey foal; foal', Balti bong-kuru `donkey colt' the same intervocalic cluster evolves into r (perhaps, through retroflex r);

4) PII intervocalic *s is probably preserved in Balti lashi `a resinous wood used as a candle or torch because it burns slowly' but dropped in Ladakhi po-ze `ram, full-grown male sheep'.

It is easy to see that all the above cases involve words peculiar to Balti. As shown in Kogan 2019, the majority of Indo-Iranian and Burushaski loanwords in Northwestern Tibetan fall into two groups: those characteristic of all the dialects, and those found only in Muslim varieties, i.e. in Balti and/or Purik. All the attested instances of seemingly unmotivated irregularities in historical phonology belong to the latter category. This fact seems to be explicable, if we bear in mind that the route of Tibetan migration to the present-day Muslim regions must have inevitably passed through Ladakh. The Indo-Iranian speakers of this or adjoining areas must have been the first people whom the Tibetans had encountered in the course of their northwestward movement. It was the contact with these people that had most probably resulted in the acquisition of loan vocabulary, now shared by Ladakhi, Purik and Balti. Later on, a certain part of Tibetan speakers migrated from Ladakh further northwest to Baltistan, where a number of lexical items from local or neighboring Indo-Iranian dialects could have been borrowed. In other words, Muslim dialects seem to possess at least two chronologically different Indo- Iranian lexical strata. A remarkable fact is that the earliest stratum, common with Ladakhi, does not show any historical-phonological irregularities. It means that there is so far no indication that loanwords belonging to this stratum have been borrowed from more than one source. This is why we consider it reasonable to accept the single-source hypothesis as a working one.

Although the lexical data analyzed above are rather scanty, they nevertheless do give us some hints as to the genetic position of the donor language. In my previous paper it was noted that no unquestionable examples of typical Indo-Aryan phonological developments had been so far attested in the material (Kogan 2019). Now it can be added that certain historical- phonological features, discovered during our study, make it virtually impossible to classify the source-language of at least some of the Indo-Iranian loans as Indo-Aryan. Here undoubtedly belong the deaspiration of PII *jh (< PIE *gh and *g^wh in the palatalizing position) with the subsequent change j > z,²⁸ and the distinction, at least in word-initial position, between the reflexes of PII *ks and PIE *tk'.²⁹ On the other hand, the retention of the prevocalic (word-initial) and intervocalic PII *s suggests that the language in question was not Iranian, since, as is widely known, in Iranian the lenition *s > h occurred in both positions. The Nuristani origin of the above loanwords is hardly probable either, since no instance of a regular Nuristani development of the PIE palatal *k' into the dental affricate c has been noted. Instead, as we have shown, the change *k' > s has taken place. Apart from this, in Nuristani etymological voiceless See the note on Ladakhi zanggi `a species of tiny flying biting insect'. As noted above, in Indo-Iranian this cluster seems to have merged with PIE *k's. Its Proto-Indo-Iranian re-flex may perhaps be reconstructed as *cs. aspirates have been deaspirated with the result that the opposition in aspiration ceased to exist for consonants, whereas our data contain quite a number of words with voiceless aspirated stops and affricates.