158

Two words have been regarded as names of the old language of Qaras'ahr (Karashar) (“I A”), which is almost universally called “Tokharian” now. One, arsi, found in documents written in that language, has been claimed as the indigenous name, used by the speakers of I A themselves; several scholars, however, foremost among them Professor H. W. Bailey, deny that arsi refers to I A at all and insist that it is a foreign word, a Prakrit of Skt. arya-, so that “arsi language” should mean aryabhdsa = Sanskrit. The other, twγry (togry), known from Uigur Turkish colophons to Buddhist books, is responsible for the introduction of the name of “Tokharian”; the late Professor Sten Konow, supported by Professor Bailey and others, rejected the opinion that the language designated as twγry by the Uigurs was the language now named “Tokharian”; in his view, twγry meant an Iranian dialect, probably Khotanese Saka. It is proposed here to re-examine the colophons in which twγry is mentioned, without entering into the problem of arsi for the present.

In an article published ten years ago¹ I endeavored to settle the limits of the area in which the twγry language was spoken by having recourse to a geographical or political term, the “Four-Twγry-Land”, occurring in Sogdian, Uigur, and Middle Persian sources contemporary with the Uigur colophons. It became clear that this “Four-Twγry-Land”, presumably the home-land of the twγry language, lay in Chinese Turkestan, and probable that it lay “near or between Bishbaliq and Kucha” (p. 550), i.e., precisely in the area where the speakers of I A lived; but the evidence remained inconclusive, chiefly because the one passage that in the nature of things could furnish proof, a paragraph in the inscription of Karabalgasun, could not be treated satisfactorily owing to the absence of a competent study on its Chinese version - may I once again appeal to Sinologists to turn their attention to that monument?

1 Argi and the “Tokharians”, BSOS., ix, 545-71.
[341]
159

Two facts that emerged from that article may conveniently be recalled here:
(1) the name of the country that has a right to the name of Tokhar in history, i.e., the region of Balkh, was, at the time to which the Uigur colophons belong, sufficiently distinct from the name of twγry to admit of // no confusion,
(2) the second consonant in Twγry, which was probably pronounced Tuγre, was voiced (y).
The tendency to confuse different names on the strength of mere surface resemblance and with little or no regard to time and space is as prominent in recent contributions to the “Tokharian” problem as it was in the earlier ones; in face of it it becomes ever more important to stress diversity rather than likeness, confine ourselves to facts, and resign speculations. That the Tuγre language was misnamed Tokharian was due to such confusion; the identification, across seven or eight hundred years and five hundred miles or more, of Tuγre with the name of the Θαγοθροι ¹ noticed by Maës (probably end of 1st century AD) is equally purely speculative ².

As neither the colophons nor the passages in which the “Four-Twγry-Land” is mentioned seemed to provide any decisive argument either way (even though the weight of evidence favoured the view that twγry = I A), the problem remained undecided and the flood of ingenious hypotheses unabated ³. It was with a start of surprise that a little while ago I noticed that the chief colophon, a text that has been studied and re-studied by dozens of scholars these last thirty years, did in fact contain proof beyond reasonable doubt ⁴. This was overlooked because F. W. K. Müller, in his reading of the Uigur passage, committed a small error and all other scholars, feeling justified in taking his reading on trust, apparently omitted to check it; for the mistake is plain enough. I hasten to say that no disrespect is intended to the memory of this truly eminent scholar whose learning few, if any, can emulate. Indeed, he excelled in reading the Sogdian-Uigur cursive script, that invention of Ahriman's. Unfortunately, as all who have tried their hand at it can testify, it is impossible to transcribe a few pages in that script without going astray in some point or other. In truth, it was in a minute point that F. W. K. Müller was in error; the effect, however, on later research was considerable.

1 A hypothesis to account for the resemblance was put forward in BSOS., ix, 563 sq. The view now held by Professor Bailey (Trans. Philol. Soc., 1947, 152 sq.) approximates to it. - As Opodva = Sogdian &"rwdn (brufn), the initial of QdyovpOL, Qdyovpov opos, Qoydpa -noXt-S should appear as 8 in Sogdian; for all four names belong to one and the same report. The resemblance is thus only partial.
2 It is to be feared that adventurous spirits will soon discover the hitherto happily overlooked colony which the untiring Yüe-chi (Yuezhi)/Θαγοθροι would appear to have founded in Eastern Balochistan : Parthian Twγrn i.e. Tuγran in the inscription of Shapur I (line 2), beside Twrgstn i.e. Turγastan (instead of Tuγrastan ?) = Τοθρηνη = MPers. Twrstn i.e. Turastan (line 19 = MPers. line 24; also in Pers. I 3 where twrst instead of Twrstn; Man. Parth. Turan-shah, Armenian Turan, Arabic Turan). MPers. Tur- is regularly developed from Tuγr- (by *Tuwr-, cf. murw from murγ, etc.).
3 For a list of relevant works see Bailey, Trans. Philol. Soc., 1947, 138 sq.
4 The point was made in a paper I read to the American Oriental Society at the meeting in New York in April 1946.
[342]
160

We can leave aside the colophon to the Uigur Dasa-karmapatha-avadana-mala (Sb.P.A.W., 1918, 583; 1931, 678), which tells that that book was translated from the language of ^əwkw Kwys^ən (Okü Küsan) into the twγry language, and from that into Turkish. In view of the geographical position of the regions in which those languages were spoken (Kucha ¹ : Qarashahr? : Turfan) it may seem likely that the chain of translations was from Kuchean into I A, and from I A into Turkish, but it cannot be called impossible that the first translation was from Kuchean into some Iranian dialect ².

(follows non-proofread section)

The chief text, of course, is the most elaborate among the colophons to the Uigur Maitrisimit, Sb. P.A.W., 1916, 414, No. 48, and plate opposite p. 416. Miiller read:

Nakridish ulushta toymts Aryacintri bodisvt ksi acari
Antkak tilint(in) Toγri tilinca yaratmish
Il-baliqda toymish Prtanyaraksit ksi acari
Toxri tilintin Turk tilinca avirmish
Maitri(si}mit nom bitig

“The sacred book Maitreya-samiti,
which the Bodhisattva guru acarya Aryacandra, who was born in the country of Nagaradesa, had composed in the twγry language out of the Indian language, and which the guru acarya Prajnaraksita, who was born in Ilbaliq, translated from the twγry language into the Turkish language.”

The mistake is in the first word, N^əkrydysh. The letters ^ə (Alef) and n are rarely distinguishable in the middle of a word, but always different from each other in initial position: here the first letter is ^ə. Further, while kn, k^ə, kr can often be confused, the three groups are neatly distinct in this manuscript (cf. kr in ykrmy line 2; k^ə in synk^ər lines 1 and 2, ^əntk^ək line 5): here the group is kn. The correct reading of the name of Aryacandra's birthplace, therefore, is ^əəknyδysh.

1 Provided that *tokw Kwys^n had the same value as the simple Kwys^n. Identity of 3itikv} with the Tibetan ^O-sku was suggested long ago (cf. BSOS., ix, 560). I should hesitate to follow Professor Bailey in his proposal to compare ^wkw further with Chinese [] and [] = bcvk (Trans. Philol. Soc., 1947, 147). The only sound that is common to ^tokw and bcok is the k. For a Kuchean word underlying bcok we should expect *p3q or *^q in Uigur. Chin. bcvk is a monosyllable with consonantic initial: Uig. *wkw is a disyllabic with vocalic initial.
2 We know now that translations were made from Kuchean into Sogdian, see my Sogdica, 59 sqq. Moreover, a fragment of the Sogdian version of this particular book could recently be identified. Its title in Sogdian was Ss? fyr>krtyh “The ten good deeds” -The fragment (“Ti a”), which is from the pncmyk prwrt “fifth chiian”, contains part of the story of King Kancanasara (kntns^r in Sogdian). See Muller, Uigurica, Hi, 27 sqq.; I. J. Schmidt, hDstahs-bLun Oder Der Weise und der Thor, ch. i, transl. 4-7.
[343]
161

It is obvious that ^əəknyδysh is Agnidesa “the land of Agni” = Qarashahr (Karashar); the form Agnidesa was not hitherto attested, but the equivalent Agnivisaya occurs, see Luders, Writere Beitrage, Sb. P.A.W. 1930, 29 line 5. Agni is the sanskritized form of the name which originally was Argi, later shortened to Arg ¹. Thus Aryacandra, who composed the Maitrisimit in the twγry language, was a native of Qarashahr (Karashar), where I A was the current language.

This clinches the argument: according to the colophons to the I A version of the Maitrisimit Aryacandra composed that work: according to the Uigur colophons Aryachandra composed that work in the twγry language: Aryacandra's mother tongue was I A: hence I A is Twγry. There is no need to recapitulate here the reasons that led Sieg and Miiller to the same result: they were weighty enough, and after forty years' discussion they stand almost untouched by the laborious arguments proffered by those who wished to attach the name of twγry to some other language; their mainstay was the argumentum e silentio that there might have been in existence a third version of the Maitrisimit.

At the same time we can now finally dispose of the name “Tokharian”. This misnomer has been supported by three reasons, all of them now discredited.
Firstly: the vague resemblance of twγry to the name of Tokharistan; it would be waste of time to discuss this any further.
Secondly: the proximity, in the colophons, of Kwys^ən to Twγry, Kwys^ən having been identified with Kushan; since Haneda's much-cited article we know that Kwys^ən (Koisyn) was the Turkish name of Kucha.
Thirdly: the presumed fact that Aryachandra was a native of Nagara = Jalalabad on the Kabul river, a place which, though very far from lying in Tokharistan, was at least within reasonable distance from it; now we see that he was a native of Agni (Karashar).

The discovery of Aryachandra's true country of origin throws light on other hitherto obscure points. Why, for example, is this bearer of the grandiloquent title of Bodhisattva unknown to the history of Buddhism? If he lived in an outlying province of India or in Tokharistan, why is his Maitreyasamiti, a unicum in the history of Buddhist literature, not so much as mentioned in the Chinese Tripitaka and the great Tibetan collections, although it was read and studied in such out-of-the-way places as Agni (Karashar) and Turfan? The answer to either question is clear now : Aryacandra was a highly respected teacher in an insignificant little town away from the mainstream of Buddhist culture; his fame did not go far beyond the immediate neighborhood of Agni (Karashar); that he was given the title of Bodhisattva is nothing but a piece of local patriotism ¹.

1 The -r- in the name is attested in sources of so different a character as the Niya documents, Khotanese Saka, Manichaean Middle Persian and Sogdian, Persian geographers, and let us add, Chinese historians; for that Yen-ch'i transcribes Argi (or *Argi) is plain (see BSOS., ix, 571). As far as I know, there is no warrant for *Angi (Bailey loc. cit. 127). All other spellings derive from the sanskritized Agni; so also Kuchean akene in which I now see an adaptation of agne, a(g)nye in the Murtuq document.
[344]
162

1 I regret to say that I cannot attach importance, for the study of the term Twγry, to the names of Ttaugara and Ttaudagar- on which Professor Bailey puts such value. The former, Stael Holstein scroll 29, was explained as the name of a Turkish tribe, Ttau(m)gara = Tongra, in BSOS., ix, 553-9. Professor Bailey, who at first accepted this explanation, has now (Trans. Philol. Soc., 1947, 148) returned to his earlier identification with Θαγαρα and thinks that my opinion was based on two fallacious arguments: that tribes mentioned in the Khotanese scroll should be presumed to be of Turkish origin, and that all forms of the names extant in the 8th to 10th centuries in Kansu had already been documented. I did not make the second assumption, which, as far as I can see, forms no part of the argument I put forward; I am fully alive to the defectiveness of our material. As to the first assumption, it seems to me a reasonable one. If the date I suggested for the scroll (middle of the 9th century, BSOS., ix, 55Qn.) is accepted - I do not think Professor Bailey disputes it - , it means that the scroll was written shortly after the great invasion of Kan-su by Uigur and other Turkish tribes in consequence of the destruction of the Uigur empire by the Qιrqιz. The authors of the scroll aim at giving a picture of the conditions that resulted from this invasion; their interest centers on these Turkish tribes whose mere names were new to their master in Khotan. However, while this assumption may be a reasonable one, my opinion was based not, or not chiefly, on it, but on the wording of the passage in the scroll. The text there says that the Sikari, Ttaugara, Ayabiri, Caraih:i Yabuttikari, etc., “are called Tardus” (BSOS., ix, 554; cf. Konow, AO., xx, 138, 154) (Tardush = western  or right wing, half). A previous list of tribal names ends with the words “these are Tolis” (Tolis = eastern  or left wing, half). Both Tardus and Tolis are well-known federations of Turkish tribes. The Sikari, Ayabiri, Caraihii, and Yabuttikari have been successfully identified as Turkish tribes (BSOS., ix, 556-7). Should we not trust the authors of the scroll and accept their statement that the Ttaugara, too were a Tardus, i.e., a Turkish, tribe? No explanation can be accepted which accounts neither for the final -a in Ttaugara (a stable ~a, not a pl. ending), nor for the intervocalic -g- which here invariably indicates a palatal g, never a y; unemended, Ttaugara represents *tögere (*toγar would be spelt *Ttauh:ari), The emended form, Ttau(m)gara fulfils all conditions.

As to Ttaudagar- (attested is ttauddgardm, gen. pl., P 2741, 78, see Bailey, BSOAS., xii, 321, 323), I should hesitate to follow Professor Bailey in his identification of it with Ttaugara, from which it differs in two important points:
1. the insertion of da, which is not found in any other Khotanese word so far made known,
2. the ending (-gara should have -garatam or -garavam in the gen. pl., cf. sacupatdm P 2741, 33, 64, augavdm ibidem 66, and similar spellings). At first I thought to find here the [] To-lan-ko (Ta-lam-kat), an Uigur tribe, that could suitably be mentioned in a text as crowded with Turkish names as P 2741 undoubtedly is; but this would demand a minor concession (misspelling of *Ttaudamgardm - the difference is very slight). On re-reading the passage, I see (taking into account Professor Bailey's discovery, kindly communicated by him, that vi ha - Skt. adi) that it is not certain as yet that Ttaudagar' is a tribal name (“two thousand troops came in front of the Ttauddgar-s etc.”, pyamtsa as postposition as in lines 17 and 129 in the same document); it may be the name of a locality.