^{page 883 note 1} u is written under the line.

^{page 883 note 2} hvaihu:ra are probably Uighurs, cf. Tib, . Hor, JRAS 1931, 832 Google Scholar, and Chinese

^{page 883 note 3} sūlya seems elsewhere, in a document from the Khotan region, to mean the people of Kāšγar (Tib. su-lig).

^{page 883 note 4} hervī, often in this document, “ any ”.

^{page 883 note 5} āphāje “ investment (?)”. Cf. phaj- in haṃphāj- “ to envelope ”; naṣphaj- occurs in cạ naṣphajāṃde mamī puña avamāta.

^{page 884 note 1} bādūṃna occurs in two other passages of this document: 64 ttī miṃ biśi bādūṃna ārri ttyāṃ pahaisāṃ si … “ thus all the bādūna attributed (iḍa-partic. to ar- in ham-ar-, hamiḍa “ join ” and nāmavarida “ famous ”) the fault to those pahaisa, saying that … .“ 69 u khvai būṃna ttū heri bausta “ and when the bādūna understood that matter “.

^{page 884 note 2} Apparently dittography.

^{page 884 note 3} karattaha perhaps corresponds to karastaha of the first list, pata could represent Old Iran, pati- “ lord”, cf. Khotan Saka spāta, spā. “ general” < *spādapati, corresponding to Tibetan sde-dpon. So read (in place of yāta, yā) spāta, spā in the Ṣacū document 47, 39.

^{page 884 note 4} sāḍimīya is struck out. One might think of Solmi.

^{page 884 note 5} ttrrūki-, ttūrki- of this name is perhaps “ Turk ”, Tib. dru-gu (F. W. Thomas, JRAS. 1931, 816 if.). In this same document Ch. 00269, line 48, we find tturki uha:hivi “ belonging to the chief (?) of the Turks “.

^{page 884 note 6} laicū is almost certain. In Ch. 0048, 5, k0aṃmicū and ṣacū are also mentioned in association. In the same context occurs śvahvā, where one will no doubt recognize śāhvī.

^{page 885 note 1} For s and ś, ḍ and l, cf. siḍathasi, śīlathasa“ Šiltās, Chilās ” in Ch. 1, 0021a, b 15–16 Google Scholar, edited in Ada Orientalia (in the press). They are probably a section of the *tōlis.

^{page 885 note 2} Clauson's very doubtful discussion of this passage setting ttaugara in Tokhāristān need not be considered here (JRAS 1931, 309).Google Scholar

^{page 885 note 3} The occurrence of the one name—of ttaugara thod-kar in the east and of in the west—used of the one people, suffices to prove that this people had brought the name with them, since the name is found in places too remote for it to be possible to suppose the name to be a foreign designation. It is therefore evidently their native name.

^{page 887 note 1} Taisho ed. 2087, p. 945, col. 3.Google Scholar

^{page 887 note 2} It is necessary to treat the problem in somewhat elementary fashion, since Indianists, who have largely interested themselves in this matter, seem rarely to haveunderstood the method of writing in consonantal scripts.

^{page 888 note 1} We have however kr'wr'n beside kwr'ynk in Sogdian for krorayina. Pelliot“s transcription *darwān, Tokh. 31, is naturally not acceptable.

^{page 889 note 1} Transcriptions of F. W. K. M¨ller must be used for linguistic purposes withsome caution. He was evidently satisfied to get a set of graphic correspondences even if he did violence to the phonetic system of the languages. In Iranian his first attempts to render Persian, Parthian and Sogdian were perhaps excusable at the time, but they did not give Iranian forms. For Turkish—a simpler phonetic system‐his method had less evil effects. But the same mechanical results gave for example in place of balïq, if the two dots distinguishing q from γ were absent.

^{page 890 note 1} The -y of the Turkish form has probably not the same origin as the -i- in Armenian t“uχari-k“. It seems to be due in both cases to some Iranian form. In Turkish it may be the -i of the nom. sing. Sogdian, , as in yymky Sogd. ymgyy, ymqyy BSOS. viii, 588 Google Scholar, but in Armenian this would not be possible: it would there be ratheran adjectival -ī < -ik. The Turkish form is the same in both Buddhist and Maniehean texts. Markwart, , Festgabe Szinnyei (1927) 67, read toχary.Google Scholar

^{page 890 note 2} Pelliot's interpretation of the Syriac form, Tokh. 48, note 1, ad calc, is unacceptable.

^{page 890 note 3} So the Wei annals, translated by Lévi, S., Le “ Tokharien,” p. 11, state: l'écriture est comme celle des Brahmanes.Google Scholar

^{page 891 note 1} In the description of Sogdiana, the Tang Annals (cap. 221, , p. 1, col. 8) state: which Chavannes, Documents sur les Tou-kiue occidentaux, 134, rendered: (Ces gens) sont habitués à écrire en lignes horizontales.

^{page 892 note 1} The Brāhmī script distinguishes 48 sounds for classical Sanskrit, and in Dialect A there are 10 (if was originally distinct from dha there are 11) additional signs. Documents in Dialect A contain Sanskrit words, and in foreign names also such sounds as h and kh are represented. Still other Brāhmī signs are used to write Barčuq (Maralbashi) Saka and Turkish.

^{page 892 note 2} The first notice of these fragments was given by Coq, Von le, “ Köktiirkischesaus Turfan,“ SBAW 1909, 1049 Google Scholar “ … mehrere gössere Fragmente einer Buchrolle in einer bis heute noch unbekannten semitischen kursivschrift…. “ Müller, F. W. K. added a postscript, p. 1061 Google Scholar: “ Die ” bis heute noch unbekannte semitische kursivschrift“ ist, wie ich inzwischen feststellen konnte, die Schrift der Hephthaliten richtig wohl oder ” weissen Hunnen “.”

The Hephthalites had occupied Tokhāristān about A.D. 468. Presumably they adopted the Tocharian writing, since according to they had nowriting of their own ( Chavannes, , BEFEO. 1903, 404 Google Scholar: dans ce pays, on ne connait pasd'e“criture). The script of these Central Asian fragments was in any case not confined to the Hephthalites. Hiuan Tsang records it also in Šiynān and Śyāmāka, which (cf. Herrmann, in Southern Tibet, viii, 447)Google Scholar was Mastujand Čitrāl. It was therefore premature to call these fragments Hephthalite.

* One must of course in this problem keep in mind the possibility of a phonetic change of to ē which would suit the NPers. form haitāl.

^{page 893 note 1} Muslim authors, Ibn al-Muqaffa“ and Muqaddasī, quoted by Marquart, , Ērānšahr, 88–9, indicate that a dialect of fārsī “ Persian ” was used in Balkh from the eighth century; probably, as Marquart thought, the language of Sasanian Persia had penetrated the city.Google Scholar

^{page 893 note 2} These forms are not Persian, but similar to Sogdian and Khotan Saka.

^{page 893 note 3} If it could be proved that the first vowel of toγara was an o etymologically distinct from ă and ŭ, Iranian would be excluded. But evidence is lacking. Tibetan o, Greek o, Saka au, Armenian o (in to χarastan) support o. Armenian t“uχxari-k“ could be due toan Iranian pronunciation where ŭ and ŏ were not distinguished. But it would still be necessary, even if the first syllable were ŏ, to prove the existence of ŭ also in the language.

^{page 893 note 4} In the sixth century A.D. according to the Chou annals (composed A.D. 630) based on a report of who travelled in 518–522, the Hephthalites were related to the It seems to refer to the time of the report, not to the time of the origin of the (Hephthalites) in Dzungaria (see Herrmann, , Asia Major, ii, 569).Google Scholar Cf. the annals, T“ang, Chavannes, , Documents sur Us Tou-Kiue occidentaux, p. 158.Google Scholar

^{page 894 note 1} ”ry'čy with Sogdian spelling of nt for nd (cf. Sogd. 'sk'nt- Av. 21).

^{page 894 note 2} In view of the statement of Sehwentner, Tocharisch 12, note 1, that the Turkish änätkäk “ istnoch unerklärt”, it should be pointed out that the Turkish 'n'tk'k, ”ntk'k, 'ntka”, is quite simply explained by reference to the Sogdian ”yntk'w “ Indian “ ( Reichelt, , Die soghd. Handschriftenreste d. Brit. Mus., ii, 70, 1. 37).Google Scholar In Sogdian words-”k -”W -”y interchange according to the obsolescent system of nominal inflexion(cf. Gram. sogd. ii, 75).Google Scholar A form *”yntk”k is nom. to the ace. ”yndk”w. It should be vocalized *indukak, a derivative of induk from hindu- (Sogdian does not preserve Old Iranian h-), Mid.Pers. hindūγ, Armen. hnduk. This induk is attested in the plural ”yntkwt induk-t (with trajected w, rather than with Reichelt “ nom. sg. koll. (?) ”).Turkish has modified and rearranged the vowels, a method of adaptation attested in other foreign words, e.g. p”rd”n Skt. pradhāna, symyt Skt. samiti. Pelliot, , T“oung-Pao, 1931, 459 Google Scholar (quoted by Gabain, A. von, SBAW 1935, 169)Google Scholar had not fully understood the Sogdian forms. The Chinese ( Life of Hiuan Tsang, Taisho ed. 2053, p. 227, col. 2,1. 24) approximates to the Sogdian form, since it implies a reading *induka-.Google Scholar

^{page 894 note 3} Sogd. prtnyh, pr'tny’, pr'ttny’ ( Müller-Lenz, , Soghdische Texte, ii, 90–91 Google Scholar, Reichelt, loc. cit., i, Dhuta, 99, 100, 162, 195) transcribes Skt. prajñā. It has hence passed toTurkish. The Sogdian indicates a Central Asian pronunciation dnị for Skt. jñ.Google Scholar

^{page 894 note 4} For the use of -čä ” in ” (beside “ into ”), cf. änätkäk tilinčä ” in the Indian language ” in the title of the Sūtra quoted by Midler, F. W. K., Uigurica ii, 51 note 1.Google Scholar

^{page 895 note 1} Schwentner's ” deutlich', Tocharisch 12, is too optimistic.

^{page 896 note 1} The meaning of the verb ritw-, nominal derivative retwe, in Kuchean ritt- and raitwe, is among those best attested. It translates in both dialects Skt. yog-, yuj- “ to join, compose”. The Turkish yaratmiš “ make, create” in the same context confirms this meaning. It is impossible to justify the use of ” wbersetzen ” either for the verb ritw- or the noun retwe.

^{page 896 note 2} It is equally the practice of colophons in Khotan Saka to omit reference to translation from Sanskrit, although such information may be given at the beginningor in the body of the work.

^{page 896 note 3} “ Kuchean ” for Dialect B is now beyond dispute. Turkish kwys'n (in Sogdian script) and kws'n (in Arabic script) is the name of Kuci (Kuchā). küsän tili” language of Kuci” is conclusive. It should be remembered that kuci is the name of a country (Hiuan Tsang used “ land of Kuci”), not only of a city as Müller and Sieg seem to have imagined, see Schwentner, , Tocharisch, 13–14 Google Scholar. Sanskrit has kaucya “ Kuchean ” for the people of the land of Kuci ( Lüders, , Weitere Beitrdge zur Geschichteund Geographie von Ostturkestan, SBAW 1930, 17). [It is very necessary for Central Asian studies that all Kuchean materials should now soon be made available.]Google Scholar

^{page 897 note 1} The same argument would prove that Tibetan, Chinese and Khotan Saka were identical, because the Sumukha-dhāraṇī is known in all three versions.

^{page 897 note 2} It is well to remember that the author Āryacandra was from Nagaradeśa, to the south of Tokhāristān.

^{page 897 note 3} We have to remember that in 1933 near Samarkand a document in unknown script was found. It is stated to be written from right to left, the letters not being joined, see Sogdiiskii; Sbornik, Academy of Sciences, Leningrad, 1934, p. 37, No. 15.Google Scholar We have also the long list of names in the colophon of the Gilgit Sanskrit MS. (see Levi, S., JA 1932, 1, 45 ff.), such as khukhuthūla, khukhuphaṇa, utruphaṇa, lerapukhra, lerakṣiṇa. These are evidently not Turkish which does not know initial l- (or r-).Google Scholar

^{page 897 note 4} A fact recognized also by Müller, and Sieg, , loc. cit., SBAW 1916, 410 note 2Google Scholar, who state that käṣṣi, wasaṃpāt, piṃtwāt, kaṣār, len paryān and rājagri are known also in Kuchean.

^{page 898 note 1} The history of Agni (later called by the half-Turkish half-Persian name Qarašahr)is given, somewhat too briefly, by Levi, S., Le “ Tokharien ”, JA 1933, 1 Google Scholar, 8 ff.It is interesting to recall that who died before A.D. 345, seems to have been a sovereign with power extending to Krorayina (see Chavannes in Stein, , Ancient Khotan, 537, 543 Google Scholar, Thomas, F. W., Acta Orient. 1934, 49).Google Scholar People of Agni were also dispersed in Kansu and the Qomul region ( Pelliot, , T“oung Pao, 1931, 496 Google Scholar, and Giles, , BSOS vi, 844: “the Lung (Dragon) tribe ”).Google Scholar

^{page 898 note 2} is used to transcribe Skt. u in udyāna

^{page 899 note 1} A letter from Fr. von Gabain has shown this hope to be unlikely of fulfilment.

^{page 901 note 1} It is almost amusing that the discovery of Kuchean Texts in the Kingdom of Agni (they were found also in the Turfan region and in Tun-huang, , Stein, , Serindia,ii, 915) has been used as an argument that Kuchean was also the indigenous languageof the often hostile Agneans.Google Scholar

^{page 901 note 2} Tocharische Sprachreste, introd. v, “ nur in Büchern ”, Pelliot seems to have gone farther, Tokh. 63: quant au “ dialecte A ”, il n'est représenté que par des manuscrits d'un caractère littéraire.

^{page 902 note 1} After Siegling, apud Lüders, loc cit., 26, where, however, probably by an oversight, stands “ Skandha, Kumāra ” as if they were two different gods. Khotan Saka uses skaṃndhä Skt. skanda-lcumara.

^{page 902 note 2} The naivāsika is known also in Turkish. Müller, , Uigurica, ii, 83 Google Scholar, has niwasiki, p. 80, naivaziki; in the Uighur-Chinese glossary n'yb'syky naiwasiki is explained by ” good genius ”. The Mahāvyutpatti has naivāsika explained as “ inhabitant ”.Google Scholar

^{page 903 note 1} The exhortation (päklyossu = “let it hear ”) is on the model of the Sanskrit rubric śṛṇotv āryasaṃghaḥ.

^{page 903 note 2} No. 414 contains prescriptions for the poṣatha-pravāraṇā of the bhikṣuṇīs (nuns). The formulae to be uttered are in Sanskrit, but the instructions are in Dialect A. Here too we shall see the indigenous language used to explain the sacred language. No evidence exists nor is there probability to make credible the existence of two sacred languages in this one country.

^{page 904 note 1} The name karacharien proposed by Lévi, , JA 1913, 2, 380 Google Scholar, and adapted to Karashahrian by Mironow, , Rocz. Orient. 6 (1928), 89 ff., is taken from too late a period to be acceptable. For Agnean we have the warrant of Skt. agneya and the contemporary name of the country itself.Google Scholar

^{page 904 note 2} On the model of Indo-Iranian.

^{page 905 note 1} Indian loanwords in Kuchean ( Mironow, , Kuchean Studies, i, Rocz. Orient. 1928 Google Scholar; Woolner, , Sanskrit names of drugs in Kuchean, JRAS 1925)CrossRefGoogle Scholar and in Agnean (Tocharische Grammatik, passim) have already attracted a large amount of attention. Other loanwords have received occasional notice (bibliography in Schwentner, , Tocharisch 46).Google Scholar

^{page 905 note 2} Cf. the passage, 64 b 2:

sne-kāraṃ saṃ lyalypu ñkät

sne-pältikāñ cem ñaktañ kus ne cami

The context of 222 a 2 (description of an evil period of time) makes it likely that here sne-pältik may be parallel to sne-käruṃ ” merciless ”. In form pältik is, according to Toch. Gram. p. 13, ” ganz unklar ”.

^{page 905 note 3} In Krorayina the ṣoṭhaṃgha was a tax-collector.

^{page 905 note 4} Toch. Gram. p. 13, curiously compares NPers. kad-χudā (so to read). MidPers. uses ktkhwt'y “ master of the house, governor of a province ”, but the second component is indispensable.

^{page 905 note 5} The contexts are not so colourless as they seemed to Lévi, , Le “ Tokharien ”, I. 6.Google Scholar

^{page 905 note 6} The variant in 251b has ṣokyākāl.

^{page 906 note 1} For *ārśa- one would expect in other texts a spelling Cf. (if they are not due to literary pronunciation of Sanskrit) Al Bairūnī's and Abū ”1-Qāsim Sā“id b. Aḥmad b. Sā“id's for Āryabhaṭa quoted by Ferrand, Gabriel, BSOS vi (1931), 336, note 4.Google Scholar

^{page 906 note 2} Single ś expresses ź in the older Khotan Saka.

^{page 907 note 1} In Khotan Saka itself rś can indicate rź. If the Agnean word were direct from a Prakrit rź the change to rś would be due to the same tendency in Agnean, as in the Niya documents, to replace voiced by unvoiced consonants.

^{page 908 note 1} Skt. āryadeśa is doubtless “ Central India” in Rājataraṇgiṇī, i, 315 Google Scholar, where it is used in reference to the raids of Mihirakula, and the adj. āryadesya, ibid., vi,89, where a college (maṭha) for students from Āryadeśa is mentioned. In the proverb ( Böhtlingk, , Indische Sprüche 1025)Google Scholar āryadeśa-kula means ” a family of India”, where Böhtlingk rendered literally “ im Lande der Ârja ”. The Tibetan titles of two medical texts in the Tanjur contain the words hphags-yul, which Cordier rendered by āryadeśa ( Catalogue du fonds tibétain, iii, p. 502).Google Scholar Mdo-hgrel 151, No. 5, was composed by hphags-yul sman-pa danadasa “ the physician Dānadāsa of Phataha in Āryadeśa ”, which is explained by the statement that Phataha is a place (yul-gru) in rgya-gar dbus-hgyur ” the central part of India ”; No. 7 was written by hphags-yul dbus-hgyur mathurahi rgyal-rigs kyi sman-pa ” the physician Raghunātha, a Kṣatriya, of Mathurā in the central part of Āryadeśa ”. [Cordier gives in the Catalogue Magadha for dbus-hgyur, although he had earlier, BEFHO 1903, 628 Google Scholar, rendered by “ l'Inde centrale (rgya-gar dbus)” and “ le médecin Raghunātha, de Mathurā, dans l'Inde centrale ”. If Magadha is right, the Tibetan knowledge of Indian geography is inexact.] hphags-yul is here equated with rgya-gar “ India ”. In a letter of 19.12.1935, Professor F. W. Thomas informed me that hphags-pahi yul occurs in the Ladakh Rgyal-rabs, ed. Francke, A. H., p. 25 Google Scholar, 1. 3; and in the sub-title of the Dbag-bsam-ljon- , whose author was born in A.D. 1702, ed. S. C. Das, p. 1, we have hphags-yul rgya-nag bod ” India China Tibet”. We shall probably prefer to render hphags-yul by āryadeśa rather than by the āryāvarta adopted by S. C. Das in his index.Professor Thomas has also pointed out that in the dictionary , hphags-yul is rendered by su (read: pu)ṇya-deśa and madhyadeśa.

^{page 909 note 1} Sieg, translated the passage in SBAW 1918, v. infra, but the context had been misunderstood. It is defined clearly by the fragmentary conversation with the Great King (“ Mahārāja ”). The poet has been enjoined to make a Kāvya. All must therefore accord with that fact. Sieg's version was distorted by mistranslation of ritw- and retwe, and his unsupported interpretation of ārśi.Google Scholar

^{page 911 note 1} Or read kulis as gen. sing., and translate “ Let it be deemed the power of — (kulis), the evil power of the Kleśas ”.

^{page 912 note 1} ” Sanskrit” is attested in the seventh century A.D. outside India, corresponding to Chinese in the colophon of the Sanskrit-Chinese dictionary of (Turkish kytsy, I-Tsing, born 635), entitled “ Book of a thousand Sanskrit characters ”. The colophon reads: cīnā akṣara sahasra mālo ārya bhāṣa smapta that is probably: āryabhāṣacīnākṣara-sahasramālā samāptā See Bagchi, , Deux lexiques sanscrit-chinois, i, 1929, pp. 217–18 and 330 (= Taisho ed. 2133).Google Scholar

^{page 912 note 2} Involving also a violent misrendering of ritw- and retwe.

^{page 912 note 3} Speculations on these Asiani are at present of little use. We cannot be sure from the Greek and Latin texts whether the Asiani were kings of the Toehari before or after their settlement in Tokhāristān. The name seems to have a suffix -āna-, which is familiar in forming adjectives in Iranian. It might mean that the Asiani were Iranian or that it is a name bestowed by Iranians or a name which hadreached the Greeks through Iranian channels. (Theories are offered by Charpentier, , ZDMG 71, 347 ff.)Google Scholar

^{page 912 note 4} The most recent attempt linguistically by N. Fukushima, On the Designation-Problem of the, so-called Tokharian Language, Memorial volume dedicated to Katsuji Fujioka, 1935.

^{page 913 note 1} Similarly a stage of Iranian older than the Khotan Saka texts is attested by the Niya Kharoṣṭhī Documents, where rodana ” madder ” contrasts with Khotan Saka rrūnai < *raudanaka-, WBalōčī rōdan ( Burrow, , BSOS vii (1935), 787).Google Scholar In the case of Khotana, the change can be seen in older Saka hvatana, later Saka hvaṃna. The Barčuq Saka texts have 6, 6, and 8 b 6. Both words probably refer directly to the inhabitants of Khotan. I cannot suppose with Konow ( Ein neuerSaka-Dialekt, pp. 30–31)Google Scholar that the writers were using the word of themselves. The contexts are not quite clear, but this view at least seems excluded. I suspect that kãñcake of 8 a 3 is connected with the name kanĵākī (the word is attested as knĵ'k and knĵk-, implying at least a Turkish pronunciation känĵäk) applied to the non-Turkish people near Kāšγar. Kāšγarī (i, 31, lines 5–6) remarks: yutakallamu fīhā bi ”lkanĵākiyyati; “ and the districts (rustāq) belonging to Kāšγar speak in Kanĵākī”. I am indebted for the reference to V. Minorsky. We should probably recognize the same name in the Turkish k”nč”k *känčäk of the Uighur document published by Haneda, , Toyo Bunko, Memoirs vi (1932), p. 3, 1. 2.Google Scholar

^{page 913 note 2} Laufer, , Sino-Iranica 361, who had first seen the connection between the Kuchean and the Chinese words, quoted the Kuchean without the final -s.Google Scholar

^{page 913 note 3} According to born A.D. 680, quoted by Lévi, , JA 1915, i, 89: C'est seulement en arrivant dans le région de yu-t‘ien (Khotan) qu'on en voit.Google Scholar

^{page 914 note 1} Similarly r for Prakrit l < ḍ = Skt. ṭ regularly: kori- Skt. koṭi, Sogd. Turk, kwldy, kwlty, Khotan Saka kūla; Kuchean kākori, kākoṭi, Khotan Saka kākaula, kākoṭä, Tib. kakola, Skt. kākoṭī, kākolī. Cf. also makara “ monkey ” given by the Kuchean šaman in the Fan-yu tsa-ming, ed. Bagchi, , p. 297 Google Scholar, as Skt. [elsewhere markata], Khotan Saka makala; Skt. l is replaced by r in many other words (nira = nīla, vipuriya = vipulya, ruka = loka) of thisvocabulary.

^{page 914 note 2} f is certain. Buddhist Sogdian does not always distinguish p and f (which is possible by the alternative use of p and β) but Manichean Sogdian has both p and f and in this word gives farn, cf. Oss. farn “ luck ”.

^{page 915 note 1} Etymologically Engl. sparrow has been compared with Agnean ṣpār- [” sparrow ” is Skt. caṬaka]. Would such a meaning suit here ?

^{page 916 note 1} Reuter's, suggestion, studia Orientalia (offered to K. Tallqvist), 1925, 232–4.Google Scholar, that ts in Dialect A might represent the fricative θ would introduce an isolated fricative into the language. Reuter himself recognized that no positive proof was to hand, buthe suggested three pieces of indirect evidence. (1) ts in aptsar- Skt. apsaras, and saṃtsāra Skt. saṃsara. Both these words, a fact probably not known to Reuter, occur also in Khotan Saka with ts, avätsara, saṃtsāra (the latter also in Kharoṣthī, , BSOS viii, 423, 427).Google Scholar The ts may be due to a Prakrit form with ts or direct from Khotan Saka. Since Khotan Saka uses th = θ, it is definite proof against Reuter'a suggestion of an interchange in Dialect A of s and θ in these two words. (2) Interchange of ts, tsts, tts, tss and ss. A value ts is equally comprehensible in these alternations. It isalso necessary to remember that ts > s may mark the later stage of the language. (3) ts is treated as a single consonant and may be written doubled. Reuter assumed that this excluded the value ts. There is an obvious error here in supposing that thespeakers of Dialect A thought of sounds according to modern phonetic analysis. The case of c = tš indicates a different point of view. This c = tš was considered as a single consonant and was written doubled in native words as kucc-aśśi ” what indeed ” and mäccek “ ipsi” ( Toch. Gram., p. 180, 192)Google Scholar and in Indian (Prakrit or Sanskrit) words, as in viccā-ṣiṃ adj. to Skt. vidyā ( Toch. Gram., p. 54)Google Scholar, as also cch in ucchiṣt and murcchäntu. That is, ttš is written cc = tštš. The adoption of ts for a sound felt to be simple filled a gap in the Brāhmī alphabet. Hence tsts means tts, as cc = tštš means ttš, with which the alternative spelling tts agrees. The Chinese transcription of the name of Kuci as kutsi shows that ts was known there, and in Agnean yāmutsi ts represents Chinese ts. It must be note d also that in Khotan Saka kh th ph are used for fricatives (χ θ f), and that for Turkish χ Dialect A also uses hkh (and hk) (Toch. Sprachr. introd. xii, where χatun should be read for qatun). Dialect A was therefore aware of the convenience of this use of the Brāhm¯ aspirates. To express θ, th would be expected according to system in Dialect A. We may note also the proposed comparisons of the Niya Kharoṣthī Document kitsayitsa (a title, possibly ” elder ”) with Kuchean ktsaitsaññe “ age ” and of aṃkratsa with Kuchean aknātse, Agnean āknats (= Skt. ”ignorant, young”), JRAS 1935, 672–3.Google Scholar In these documentsts ts used for Skt. ts as in saṃvatsari. Here too we find th chosen to represent Iranian θ in thavaṃnae, thavaṃnamae ( BSOS vii (1934), 512)Google Scholar, Khotan Saka thauna “ cloth ”.* There is therefore no reason to conjecture that ts is θ in Dialect A. The proof against such a theory is positive.

* A word probably known also in Kuci since in the Kuchean Fan-yu tsa-ming (ed. Bagchi, , pp. 48 and 279, No. 537)Google Scholar thacaṇa “ cloth ” is probably a misreading of thavaṇa. Here too th represents Iranian θ.

^{page 917 note 1} It is hoped to take up this problem later. We have to recognize at least a cultural connection between Krorayina and the northern cities. To this cultural unity belong the three titles: (1) gauśura in a Sanskrit document from Kuci ( Lüders, , Zur Geschichte und Geographie Ostturkestans, SBAW 1922)Google Scholar, corresponding to guśura of the Kharoṣṭhī documents, (2) ṣoṭhaṃgha in Agnean (3) cazba in the Barčuq Saka, corresponding to Krorayina cojhbo. Three names may indicate even ethnic connections. In the Ṣacū document edited by Konow we have in the region of Tturpaṃni (Turpan, Turfan) the people Argīña in the phrase Argīñvā bisā kaṃtha “ the town among the Argīña ”. With this may be compared the name Argiya in the Niya Kharoṣṭhī documents. A personal name in the Niya documents cimola probably appears in Khotan Saka, Ch. 00269, as the name of a people cimūḍa associated with the hvaihu:ra. It is possible to connect these with the čumul of Kāγarī, a people near Bišbaliq. It is admittedly impossible to prove they were not originally Turks, but it is possible to see in them a people whom the Turks had absorbed. The third name is acuñi. In the Niya Kharoṣṭhị documents occurs acuñiya aṃcuñi acu[ñ]i as a personalname. The same name is attested as the name of a king of Kuci, in Chinese quoted by Lévi, , Le “ Tokharien”, 22–3.Google Scholar Lévi proposed to equate *ačuñi with Skt. arjuna, a phonetic equation which naturally seemed doubtful to Pelliot, , Tokh. 72 Google Scholar, note 1. We may keep in mind also the still uncertain Niya ogu and of the Turkish colophon. There is also a possibility that the Niya Kharoṣṭhī name kaṃjaka is connected with kaněak, see above, p. 913. In vocabulary we may, beside the reference to Burrow, , JRAS 1935, 667 ff.Google Scholar, note also Agnean ṣlyok ” strophe” beside Niya Kharoṣṭhī ṣilȳoka, ṣilȳoga “ urkundliches Schriftstuck ” according to Lüders, , BSOS viii (1936), 654.Google Scholar It may further be indicated that a section of the would probably solve the problem.

^{page 921 note 1} The name in Turkish, in Chinese, Greek , similarly offers βγ and kb“ We should probably prefer to trust the form in Sogdian script.

^{page 921 note 2} Three unimportant points may be indicated here. On p. 261 the reference to Switzerland is due to a misunderstanding of the phrase “ identical language ”: Swiss ksī “ been ” is sufficiently different from German gewesen. On p. 264 the earlier explanation of Syriae is repeated without reference to the essential difficulty, the absence of w in the first syllable. On the same page, lines 3–4 are not strictly correct, since neither of the two erroneous forms (due to an error in copying, since no originality was sought in the Chinese conjectural reconstructions) entered into the following discussion, and t“uoχuâlâ was attested by the third Chinese transcription.