By W. B. HENNING
1. 'tyw'r'k
During the last years the problem of the " Tokharians " has been
studied by S. Levi ("Le ' Tokharien '," JA., 1933, i, pp. 1-30), Pelliot
("Tokharien et Koutcheen," JA., 1934, i, pp. 23-106), H. W. Bailey
("Ttaugara," BSOS., viii, pp. 883-921), again Pelliot ("A propos du
' tokharien '," T'oung Poo, xxxii, pp. 264 sqq.), Haloun (" Zur Ue-tsi-
Frage," ZDMG., 91, pp. 243-318), and E. Sieg ("Und dennoch
' Tocharisch'," Sb.P.A.W., 1937, pp. 130-9). Facts mentioned in
these articles are, as a rule, not repeated.
The Sogdian form of the " Tokharian " name, 'tyw'r'k, suggests
the pronunciation atyudre (from 9tyudrak). This apparently derives
from Hyudre and permits the assumption of parallel forms such as
*t9yudr- and *tayuar-. It is possible to take as starting-point Hayudr
instead of *tyuar as vowel-reduction is well attested for the first syllable
of Sogdian words; the entering of a prothetic vowel to ease the
newly created double consonant at the beginning of a word is quite
common in Sogdian as well as in other Iranian languages; here as
in other cases historical spelling often obscures phonetical realities;
e.g. Olr. mrya- " bird" is maya, mya, finally amya in Sogdian:
amya, warranted by Choresmian 'A* I amya, may be written 'mry'
(cf. Gauthiot, Gramm., i, pp. 40 sqq.).
The pronunciation tyudr (tayuar, tayudr) was already attested
by Syriac thwrstn, in the Singanfu inscription, i.e. tayudristdn.
Furthermore, the group -yu- in the middle of the word is confirmed
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
546 W. B. HENNING—
Saka (h = x), Syriac and Chinese (see Pelliot, T'oung Poo, xxxii,
p. 264 n.); ambiguous : only Sogdian.
Quantity of Vowel in Second Syllable : long a. in Arabic, Pahlavi,
Sogdian, Saka, Sanskrit, perhaps even Chinese; ambiguous, either
a or a: Greek, Armenian, Syriac.
Consequently, only those languages are ambiguous which in
writing do not distinguish a and a, x and y ; wheresoever distinction
is possible x a n ( i ® res P- a r e indicated. Therefore, the only correct
forms of the name are txudr (atxudr, taxudr, toxudr) and toxdr
(tuxdr).
It is very difficult indeed to determine the relation between these
two forms. If nothing but the possibilities of phonetical development
within Iranian languages had to be considered, there could be no
hesitation in answering the question : toxdr (tuxdr) is developed from
older taxudr. The change from older ahu, axy- into ox, ux (some-
times oxu as intermediate stage) is abundantly attested for nearly
all Iranian dialects, beginning with Old Iranian languages (cf., e.g.,
Avestan -awuka < -ahva = Skt. -asva; Avestan varmhim, vanhim
< vahvim = Skt. vasvim, etc.). A fine example is offered by the
rendering of Old Persian harahbvatis (corresponding to Skt. SarasvaM)
in the Akkadian and Old Aramaic versions of the BisutQn inscription :
Akk. a-ru-hat-ti and Aram. TllTm (line 39, ed. Cowley, p. 253)
= haruxoti approach medieval Arrux0^ rather than the official
Old Persian form or Avestan haraxvaitl-, Gr. 'Apaxoo-ia. The
intermediate form XopoxodS (Isidor of Kharax), attested as it
is about half a millennium after the phonetically younger form
haruxati, may help to remove the apparent difficulty which lies
in the fact that *txudr occurs only many centuries after Toxapoi.
Nevertheless, this difficulty should not be disregarded: the priority
of the group t(a)xudr (which, in Greek, should be *raxu)poi or
*raxoapoi) cannot be taken for granted until an old example
has been found. Ptolemy mentions Tax°POL with regard to the
northern section of the Yaxartes region, but Toxapoi with reference
to Bactria (see Marquart, Erdnsahr, p. 206 and n. 4): it is tempting
though probably unwise to draw the inference that rax°Pot represents
*taxuar as the original name of the people whilst roxapoi is a later
form developed under the influence of the Iranian language spoken
in Bactria.
On the other hand, it must be examined if there is any possibility
of the reverse development: Hxudr from older *tuxdr. Such a
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
548 W. B. HENNING—
change could have taken place either in the original language of the
Tokharians which is completely unknown, or in the indigenous
language of Bactria, undoubtedly Iranian, probably adopted by the
immigrants, which, however, is equally unknown (cf. Bailey, Ttaugara,
pp. 892 sq.). As we have to deal with unknown languages nothing,
in fact, can be proved or disproved.
As to the indigenous language of Bactria, we are at liberty to assume
a strong affinity with the Sogdian neighbouring language. It is a well-
known fact that there was a tendency in Sogdian to replace any
kind of w-sound (sonant as well as consonant),1 especially where a
Sogdian y (representing Old Iran, h, y, g, y) stood near to it. This
tendency, however, operates in exactly the same direction as in other
Iranian languages though it is much more pronounced : -yu- changes
to -uy_- in the middle 2 of a word, whether a vowel precedes 3 or a
consonant,4 whilst final -ayu results in -uy as in cognate dialects5;
the Sogdian verb patyudy- " to kill", quite similar to dtyuar- in its
grouping of sounds, is patuydy- in later texts ( > patuyay- > ptuyay-
> tuydy in Yaghnobi); hence, in Sogdian we should expect tuydr
for older tyudr.
There are, however, peculiar spellings like Sywt-, syw8-, yywt-,
yyws-, representing older Suyd-, suyS-, yuyt-, yuys-, which seem.to
authorize the assumption of progressive w-metathesis in Sogdian.
It is, of course, possible to look upon these spellings as nothing but
a rather unusual case of inverse writing, based upon the experience
of frequent change of -yu- (written yw) into -uy- (written -wy-);
this view could be supported by the apparently absurd spelling
wc'ywt'k (Dhuta 215) instead of wc'wytk or wyc'wytk " wise " ; other
spellings, wc'ytk, wyc'ytk, wc'yty, wyc'yty clearly show that -'w- in
wc'wytk denotes a short vowel, u or preferably o (see Beichtbuch,
p. 62): it would require a great deal of credulity to believe the
reality of a pronunciation U9cayud- < U9coyd- which the Sogdian
1
"Ein manichaisches Bet- und Beichtbuch" (Abh.P.A.W., 1937, No. 10;
quoted as Beichtbuch), p. 89.—For the remarkable change of u to u in closed syllables
in cases of epenthesis (e.g. nyund- : nyuend-, Beichtbuch, p. 76) compare Saka
mura : mvire, etc.
2
In Parthian even initial xu- results regularly in «X' > c^- Sogdian wywiw " six "
= tixuasu from *xsnasam.
3
auxast- < axuast, iu-j/ay- < iymty-, see Beichtbuch, p. 59.
4
sux&y- < sxuay-, cf. zux$k " pupil" < ixuslc.
5
'u>x = UX " mind " from ax2?> see Beichtbuch, p. 86 ; Man. Sogd./su»x =fasux
" parasang ", Beichtbuch, p. 125, from frasaxu (in Chr.S. fasax, and in Buddh. S.
'ps'ny = vjsamx < frasanha-).
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
ARGI AND THE " TOKHARIANS " 549
The only occurrence of the name " Twyr " which has not yet been
mentioned in the course of the present article seems to be Saka
ttaugara : its main value, I think, lies in that it has provoked Professor
Bailey's ingenious contribution. It occurs in the Saka part of the
Stael Holstein scroll, published along with the Tibetan document
by Professor F. W. Thomas and Professor Sten Konow (Two medieval
documents from Tun-huang, Oslo, 1929), in a list of names (27-31)
that follows another list containing place-names (10-26). In the course
of the last years, it has become increasingly clear * that this list of
place-names, undoubtedly one of the most valuable documents for
the history of Central Asia, enumerates at first places between (about)
Khotan and Liang-tsou 2 and then the towns of Si-tsou,3 from Qomul
in the east to Urumci and Camilbaliq in the west. Several names
mentioned separately at the end of the list are possibly to be placed
to the south of Tun-^uang.4
The second list, however, in which ttaugara stands, contains
names of nations. Professor Bailey, comparing a document from
Sa-tsou which offers similar names in an enumeration of the tribes
that constituted the population of Kan-tsou, drew the inference that
the nations mentioned in the Stael Holstein scroll were to be looked
for also in Kansu. On the other hand, the names which are
common to both lists are comparatively few, namely (1) Tolis
(tulisi ~ ttUdlsa 5), (2) Turk Bayirqu,8 (3) imju, ijuva, imjuva. The
alternative possibility that the list of nations refers to the territory
described before cannot, perhaps, be discarded. As there was no
need for the envoys of the Khotanese king to describe the population
of the Khotan region the list probably relates to the inhabitants of
Kan-su (with Sa-tsou) and Si-tsou.
The list of nations is divided into two sections, the first of which
(27-8) ends with tti tulisi stare " these are Tolis ". The summariza-
.tion of the second section (29 sqq.) to be found in the words tti ttari
1
Besides the edition, cf. Clauson, JBAS., 1931, pp. 297-309; F. W. Thomas,
BSOS., viii, pp. 793 sq., ZDMG., 91, pp. 13-15, 47 sq.; Minorsky, Hudvd al-'Alam,
pp. 271 sq., 292.
2
Gakimamni (14), between Kua-tsou and Su-tsou, is probably 3£ f|J t)-man
(*ngi«>ok-muvn, Karlgren 1318 + 609).
3
Secu (17, 20), see Bailey and Minorsky, BSOS., viii, p. 120.
4
Anitturnga may be the same name as later Handling (s. Bretschneider, Medieval
Res., ii, pp. 218 sq.).
5
Cf. twlyS = Tolis, in Manichsean script (as part of a name, Mahrnamag, line 97).
• See Minorsky, BSOS., viii, p. 918.
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
554 W. B. HENNING—
ttrusahuta is less clear. Professor Bailey, with whom I had the pleasure
of discussing the passage, believes that du is to be read instead of
ttru ; he further suggests taking ttaridusa- for one word and comparing
Tardus, the well-known name of a federation of Turkish tribes, written
ttarddsa in unpublished Saka documents. In that case I should like
to propose to take -huta for an irregular spelling of classical Saka
hvata, and to translate " these are called Tardus ". Three further
names constitute the end of the list.
The two federations, Tolis and Tardus, were, according to
V. Thomsen (Inscriptions, pp. 146 s.), the main constituents of the
Eastern Turks.1 It is the natural course, therefore, to pay attention
to the names of Eastern Turkish tribes in attempting to identify the
names offered by the Saka list. The Chinese lists of fifteen Tolis
tribes and of nine proper Uyyur tribes will prove helpful (enumerated
by Chavannes, Doc, pp. 87 sq., 94; by Schlegel, Karabalgasun,
pp. 1, 8 sq. ; Schlegel has earned the gratitude of non-Sinologists by
indicating the contemporary Chinese pronunciation which alone, even
though sometimes not quite correct, renders identifications possible).
The various names of the Saka list are to be examined
separately:—
(1) Both sections begin with the same word, or with two similar
words, namely (a) ijuva, probably misspelt for imjuva, and (b) imju.
Professor Minorsky compares Turk, incu " appanage " (BSOS., viii,
p. 918) or injii " the tribe belonging to the chief's household "
(Hudud, p. 292). It is indeed very likely that the word is a term for
a tribal division if it is not the name of a tribe. It occurs in the Uyyur
fragment published by Haneda (Mem. Res. Dep. Toyo Bunko, Nr. 6,
p p . 1 - 2 1 ) , A 1 0 s q . : incu urungu sangu'n apa cur o o incu ... As
the names of the Uyyur Manichaeans enumerated in the fragment
are usually preceded by an indication of their origin, either a
geographical term (e.g. kusdnlig) or the name of a tribe (e.g. k'nc'k, see
Bailey, Ttaugara, p. 913, n. 1), it is probable that incu has a similar
meaning ; Haneda's interpretation 2 of the word as " pearl " (p. 7)
is hardly admissible : in Uyyur " pearl " is yincu, not 'yncw.—Cf. also
Tci imjuva, surely the name of a tribe, in the Satsou document published
1
It is not felt necessary to enter into a discussion of the Sie-ien-to (supposedly
Sir Tardus) here.
2
In the same fragment, B 2, nyw?kl'[r] " auditores " is to be read instead of the
mysterious u'xusikla[r]; the cigU Myr "ySy (Mir-ayaSe) is a Sogdian (" Sunday-
wish " ) .
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
ARGI AND THE " TOKHARIANS " 555
script = " Ancient Chinese " ia is attested, e.g. for 4^ (223) and
3|5 (226), see F. W. Thomas, ZDMG., 91, p. 41. It is perhaps to be
considered if this explanation is to be preferred to Professor Thomas's
combination with Emil, Emin (ibid., p. 48), written in Chinese P@
Jg ien-mie (*-ien-miet, 273 + 622) and | f§ ien-mien (*-ien-mian,
624), i.e. exactly Emil, Emin x (also tribal name, Chavannes, Doc,
p. 123 n.).
(11) caraiki, i.e. *cariy, *carix, perhaps even *cariq (on h and h:
see Bailey, BSOS., ix, pp. 297 sqq., esp. p. 300 n.). This tribe, J ,=~ =
v v *
Cariq, Caruq, is to be found later on in Barcuq (Maralbasi), as stated
by Kasyarl (i, 318) who, furthermore, indicates that at his time
(second half of eleventh century) the J^f- were living between
Yayma (in Kasyar) and Cumul (near Manass), i, 28.2
(12) yabuttikari, i.e. *yabutqar or *ydbutkdr, probably the eighth
Uyyur tribe, $j %J ^ iau-u-ko, *iak-miu9t-kdt (568 + 1278 + 73),
Schlegel's Yamukar. The first and the third sign are employed also
for yay-la-qar; it is, however, rather strange to find the same character
used in rendering both ya and yay; no doubt need be felt about ^5
as equivalent of biMi {but) as well as biri (bor).
(13) anakidipabhiitti is a puzzle (-bh- is also unusual). Possibly
two words : ana " other " yilpabut ? Another (equally unsatisfactory)
explanation has been proposed above, p. 557, n. 1.
(14) karattahapata, Professor Bailey explains karattaha as Turkish
Qara-tay. This is surely right 3 and could be supported by yijrikim-
nittaki (in the Sa-tsou document, Bailey, Ttaugara, p. 883) which
can be explained only as Turk, yipkin tay " brown (violet) mountains ".
-pata (Bailey, Ttaugara, p. 884 n. 3 : = pati " lord ") is rather curious
here; one might expect a post-position " in " ( = classical Saka
pata1).
(15) sddimiya, struck out in the manuscript, follows upon tti
ttaridusa huta. Professor Bailey has recognized Sulml (Solmi), cf.
1
The combination of ayabiri and emil would require the presupposition of
(erroneous) re-transcription from Chinese (confusion of m and 6)'; the mysterious
anah-.idipabhutli may have to be explained in that way ; this could be re-transcribed
from Chinese, say, *TJ? | | | j t ^ / *-dn-iak-li3p-m}ii9t, representing Turkish
*anyaylibut or the like. A Chinese intermediate form alone explains Saka hvaih-.ura =
uiyur, containing hvai — [5] yuai (xuei), a. Bailey, Ttaugara, p. 883, n. 2.
2
I disregard the iyraq, uyraq, living between Yayma and Cariq.
3
The alternative explanation as name of a nation, suggesting *qara-taypat with
the name of the Wei (t'o-po, *fdk-b'itdt, 1159 + 750; cf. t'u-fa, *t'uk-piwvt, 1131 +
750, s. Haloun, ZDMO., 91, p. 275), is certainly to be rejected.
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
558 W. B. HENNING—
1
The tnCrywS ysyS (line 5, verso line 1) is noteworthy, tm'rywi is to be looked for
in Faryana as is proved by the title x^eS; hence, tm'ryws, Uimaryua" = Tamdxui
near Isfara (Barthold, Turkestan, p. 160).
2
Professor Bailey kindly told me that the equation ts'u-mi = iumul has been
found before by Pelliot, T'oung Pao, xxxii, p. 363.
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
ABGI AND THE " TOKHABIANS " 559
1
The Oyuz tribe jy Hlkdr (KSsyarl, i, p. 57), living far away in the west,
is ruled out; tiihar would be written *ttukari; the spelling ttau[ni]gara leaves no
doubt that the word ended in a, not in r. The presence of such a great number of
Turkish tribes in Kansu, Sa-tsou, and Si-tsou is hard to explain if the date proposed
by the learned editors for the Saka document and the Tibetan document attached
to it (second half of eighth century) is adopted. Though, of course, the infiltration
of Turkish, mainly Tolis, tribes into this region began at an early date, one would not
expect so many of them there before the destruction of the Northern Uyyur empire
(840), which led to the foundation of Uyyur principalities in the territory in question
by emigrant Uyyur tribes. The Khotanese king Visasambhata may have been a
namesake of Vijayasambhava, not the same person. Preservation of Tibetan as
official language at the beginning of the Uyyur reign would not be surprising. The
name of the Sa-tsou king The-po, The-bo, otherwise unknown (according to Professor
Thomas, p. 129), recalls the name of the leader of the Uyyur immigrants, P'ang-
t'3-h (Bretschneider, Res., i, p. 241 ; Minorsky, Hudud, p. 226 n. 2, 264 n. 5); this
name, suggesting (if )jfj is the correct reading) P'ang tegin, a wholly unsatisfactory
name for an Uyyur, is apparently distorted (so as to obtain the well-known title
t'd-to ~ tegin). It is, I think, the same name as T'3-p'ang-l3 (another Turkish chief,
Chavannes, Doc., p. 86); alternatively, the possibility that the correct form of the
name was *T'3-p'ang-t'3-U = T's-p'ang tegin should be considered. In any case,
The-po, The-bo might be a shortening of this name, thus rendering t's-p'ang (it is
probable that p'ang, *b'dng, Karlgren 585, would correspond to *bo or the like in
Central Asia; jf^jp t's is employed also in ^ p 1$) t'3-to, to-lis). If the documents
were written shortly after the Uyyur invasion the interest displayed by the Khotanese
envoys in the population of the neighbouring country would be comprehensible.
I do not know if any chronological inference can be drawn from 11. 41-2 of the Saka
text: "Cam ttutt-u ordered to build the stupa (Bailey, BSOS., viii, p. 119) of
gumattiri," i.e. " the great stupa of Hgum-stir " (F. W. Thomas, Tibetan Literary
Texts and Documents, i, p. 108).
VOL. IX. PART 3. 37
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
560 W. B. HENNING—
ending -e1 < -ak which, of course, did not belong to the original
form of the name. Assumption of a vowel (a) between y and r is not
supported (though not completely excluded) by the spelling; if the
name were Hoyar one would expect at least one example of Hwyr,
especially in Uyyur. If twyr is considered with unbiased mind, by
far the most probable pronunciation appears to be Tuyr.
The country " Four Tuyr " covers roughly the same area as the
Tibetan " Drugu-country ", a name of Bisbaliq, Turfan, and Qarasahr,
as has been shown by Professor Thomas's researches (see JRAS., 1931,
pp. 807-827, esp. p. 826). It may be worth considering if dru-gu
(durgu) can be regarded as transcription of tuyr; even if originally
dru-gu had been meant for t'u-kiie, the choice of this term for the
region in question was probably influenced by the indigenous name
tuyr.
It is necessary to touch on the question (though a full review of the
problem cannot be given here) if an original connection between
txudr, the name of Tokharistan, and tuyr, the name of Bisbaliiq-
Qarasahr, is conceivable. It is hardly justifiable to take the presence
of such a connection for granted as is usually done ; on the contrary,
its existence very much needs proving. Until this proof (or disproof)
is forthcoming a considerable element of uncertainty will necessarily
be inherent in all theories on the origin of the " Tokharians ".
As unknown languages, however, are involved, it will be found
impossible to disprove any proposal to unite all the different forms
notwithstanding their indubitable disparity at the time when the
Central Asian documents were written. The scheme shown on the
following page, e.g., would cover all the forms.
Professor Haloun, in his excellent contribution " Zur Ue-tsi-
Frage ", maintains that the " Tokharians " were Ue-tsi everywhere,
in Bactria as well as in the Nan-san.2 He refrains, however, from
1
Quantity undecided. Sogdian loanwords in New Persian suggest e rather than
e (might be, however, influenced by the ordinary Persian ending -a). Chinese tran-
scriptions may be helpful in solving the problem, cf., e.g., H§ J§J p f § , Parthian hasenay
(Waldschmidt-Lentz, Stcllung Jesu, pp. 83, 85 sqq.) : x°-si-n9n9> *ydt-si-n3ng
presupposes sogdianized pronunciation as yasene (yamn'k) ; on the Central Asian
pronunciation of Chin, -ng, see Pelliot, JA., 1912, i, pp. 588 sqq., and particularly
F. W. Thomas, ZDMO., 91, pp. 10 sq. (where, however, Hjj is rendered by dimiii,
i.e. ding).
2
The second part, at least, of this assertion is open to further investigation. Apart
from Oayovpoi in Ptolemy (b. vi, ch. 16), Saka ttwwgara and Tibetan Phod-kar
(etc.), Professor Haloun's demonstration is based mainly on the fact that Kumarajiva
explains Tou-k'n-lo, i.e. transcribed Skt. Tukhara, by /J> ^ J£ siau Ue-tsi " Little
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
562 W. B. HENNING—
West East
*t'yur
j
I I
*t'3yur
*t'6yar
I i r 1
Taxopoi *tx'/dr Boyapa *i'oyr
tuyr
tayyiar
tje-tsi". In Professor Haloun's opinion, this gloss proves that Kumarajiva knew
that the " Little tJe-tsi" of the Nan-san region called themselves " Tokharians "
(p. 280). It will hardly be admitted that this interpretation (contemplated before
by S. Levi, rejected by Pelliot) is the only one possible. Kumarajiva in explaining.
Tou-k'ii-lo should have written (Great) tTe-tsi; instead, he wrote " Little tJe-tsi",
thus misinterpreting the text he was commenting upon in a singular and ridiculous
fashion. It is as well to assume that Kumarajiva, bearing in mind the usual equation
Tukhara = (Great) tje-tsi, contemplated at first writing simply " tJe-tsi", but
substituted " Little tJe-tsI" because those were the only tJe-tsS that he knew of by
his own experience and expected his readers to have heard of. On the Phod-kar
(Thod-kar, Thod-gar), living in north-eastern Tibet, and mentioned many centuries
after the last occurrence of the " Little tle-tsl", see F. W. Thomas, JBAS., 1931,
834 sq. Their habitat is perhaps not yet sufficiently well defined to render their
identification with the Little Ue-tsi an incontestable certainty. If Tib. Dru-gu is =
Tuyr, the Drug-cun in Western Kan-su, " Little Dru-gu " according to Professor
Thomas (pp. 814 sqq.), might be = " Little Tuyr " = " Little Ce-tsi".
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
ARGI AND THE " TOKHAKIANS ; ' 563
" Tokharians " even though it might prove necessary to modify his
proposals in. minor points.
It is not admissible to compare with Ue-tsii a developed form such
as toyar, just as it would be impossible to use the modern Mandarin
pronunciation of ft J£ for' comparative purposes. The only com-
parable form is that which was employed by the Ue-tsi' themselves
at the date when the Chinese transcription was created. The con-
temporary form has, of course, to be reconstructed. It has been
surmised above that the attested forms may derive from *t'yur or
a similar form, e.g. *0yur, *t'yu&r, *6yuar, etc. Chinese ft, in con-
temporary pronunciation **zngiwa8,1 could be a rendering of such a
form, as perfect as can be expected in a name containing a group
of initial consonants foreign to Chinese speech.2 The second sign, *t'ia,
might be regarded as representative of an indigenous plural ending.3
It would be well in accord with Professor Haloun's theories to find
the Tuyr-Ue-tsi in and near Qarasahr. When the main body of
the Ue-tsii, pressed hard by the U-sun, was compelled to leave the
T'ien-san region, some tribes may have stayed, as subjects of the
U-sun (see Haloun, p. 246). According to Professor Haloun, the
U-sun were " das urspriingliche ethnische Substrat der die beiden
' tocharischen ' Dialekte sprechenden Bevolkerung Ostturkestans "
(p. 254). It may be permitted to say that the population of the region
between Kaiyar and Bisbaliq, at the beginning of our era, consisted
mainly of U-sun and a few tle-tsi and Saka tribes (the latter, e.g., in
Barcuq). During the following centuries the remaining Ue-tsi tribes
have been absorbed by the more powerful U-sun; the process of
amalgamation probably was completed at the beginning of the T'ang
reign. At that date, nothing remains of the Ue-tsi but their name,
which, as Tuyr, is used indiscriminately for the same country,
population, and language as the name of the U-sun (Arsin, arsi),*
1
The initial z- appears to be anything but certain. For the T'ang pronunciation,
Saka Icamfturi, name of a tribe (in unpublished Saka documents; Professor Bailey
kindly communicated this name to me), is interesting: apparently = pj ft bung-He,
Icungur (thereby containing ft — gur).
* It would be interesting to learn from Sinologists how in their opinion the Chinese
could have rendered a word *8yur (or the like) in a less ambiguous way.
2
One might even go one step farther and assume an internal tJe-tsi sound change :
postsonantic 8 > r (attested in numerous languages); therefore, *t'yur < older
*6yuhi-, the latter being " attested " by Chinese *zngotia, *zngj,wat-t'ia.
4
It is taken for granted that ar$i, drain, and u-sun represent the same name;
the Chinese had some difficulty in rendering -r at the end of a syllable (on Jjj
see Haloun, pp. 252, 314); if necessary, -r could be explained as in Arg.
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
564 W. B. HENNING—
2. 'rkcyk
As stated above, in the Sogdian " List of nations " 2 'rkcyk follows
the nations of Kasyar, Khotan, and Kuci (Kuca). From the relative
importance of those districts it is at once apparent that the fourth
partner in a series like this refers to Qarasahr.3
'rkcyk, i.e. arkclk, is an adjective derived from 'rk, i.e. Ark (suffix
cik, see Gauthiot-Benveniste, Gramm., ii, p. 96). Ark (or Arg) occurs
as the name of a country: in a Middle Persian Manichaean hymn 4
homage is paid to the religious leader5 'y n'mwrng frwx'
[hjw'b'd shr 'y 'rq " of the famous, blissful, and prosperous country
of Ark ".
Ark (Arg) is mentioned also by Persian geographers and historians.
The Hudud al'Alam describe J j l = Ark, a town in the Toquzoyuz
(Uyyur) country, as " a small town near the river Khulandyiin,
possessing plenty of fruit, except grapes. To it belong seven villages,
and Ark and its districts are said to turn out 20,000 men " (tr.
Minorsky, p. 94). Professor Minorsky, commenting upon this passage
(pp. 273 sq.), has found that Ark is embedded in a series of names
that formed part of an itinerary leading from (Bars^an via) Kuca
1
Attested by Xuei-lin : " Kuca a ete appele aussi tTe-tsI et U-sun," see Pelliot,
JA., 1934, i, p. 90, n. 1.
2
Written in Sogdian script.
3
The names following 'rkcyk in the list, probably pointing to the Turfan region,
are not yet sufficiently clear to offer confirmation.
4
Not yet published (M 297); in Manichsean writing.
5
Name and title missing.
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
ARGI AND THE " TOKHARIANS " 565
V
1
In Sogdian (s. Minorsky, p. 273); also attested in a Manichaean fragment,
where 'hryy knS- — se hand (< 30re) is written, s. Beichtbuch, p. 12 (but drraya
kaiflihe seems to be different). One might think of §J| 2El J $ tsang-san-U'mg (better
than I j j j j mie), " t$a,ng-three-cities," between Ien-k'i (Qarasahr) and Qoco,
see Chavannes, Doc, p. 6.
1
Khulandyun looks distinctly Sogdian (cf. Minorsky, p. 206 and n. 4); -yun
" kind " is rather obvious, but no word *ywS'nt (might be " ""covering ") or *yw(')k'nt
is attested ; yw'r'nt (" right ", ace. to Rosenberg, Izv., 1918, 831, originally " brilliant
glorious, good ") is not satisfactory either; the word might have the same meaning
as ^jjj tan ("insipid, tasteless, watery"), in the T'ang-su name of the Khaidu-gol,
Chavannes, Doc., p. 6.
3
See Bailey, BSOS., viii, p. 120.
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
566 W. B. HENNING—
1
It gives me pleasure to state that years ago Professor Schaeder, in the course of
a conversation, proposed to regard 'rlc ('rle-cyq) as a geographical name, not as an
appellative " castle " (as Muller thought).
2
It is, however, at least equally possible, if not more probable, that his much
more powerful namesake (808-821) is meant.
3
Grammatically, the geographical terms are, all of them, adjectives ; the Persian
(xwaSay) and Sogdian (xwataw) titles cannot, naturally, be regarded as official
designations of the rulers in question; only under (3) are genuine titles given.
4
It is hard to explain the prothetic vowel which if not articulated would not be
written in Manichaean script though possibly it would in Sogdian script; cf., e.g.,
Buddh. S. 'kwty as against Man. S. kwty (kuti), " dog " ; 'kwty suggests a development:
kuti > kdtl > kti > dkti (rather dubious). Derivation of 'kwcyk from *'kw =
Turkish Oqu (ktisan), see above, p. 560, is very unlikely.
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
ARGI AND THE " TOKHARIANS " 567
(3 a) is the " kds-ism king " (not " the king of *KasI "), i.e. " the king
of Kds = Kasyar ", Chin, -fj fp k'ia-sa. (3 /?) refers to Aqsu, and
(5) to Uc (see below). All the important principalities between Kasyar
and Bisbaliq are mentioned, only Qarasahr appears to be missing:
it is, in fact, clear from this passage alone that Ark (Arg) is Qarasahr.
Accordingly, the enumeration comprises :—
(1) Bisbaliq
(2) Qoco
(3) Kuca
(a) Kasyar
(£) Aqsu
(4) Qarasahr
(5) Uc.
The principalities are not arranged in a purely geographical order,
but perhaps according to their relative political importance; yet it
would be rash to come to the conclusion that, at the date of the text,
Kasyar and Aqsu were politically dependent on Kuca whilst Uc had
retained its independence; it is quite likely that more attention was
paid to ecclesiastical than to political divisions.
It seems desirable to study more closely the names of Aqsu and
Uc attested in the Mahrnamag. The identity of prw'nc with Aqsu,
proposed by Professor Minorsky (Hudud, p. 482), will hardly be con-
tested. The feminine ending of prw'nc suggests a fuller form *prw'nc
Jcand, of which prw'nc is an abbreviation ; the masculine form corre-
sponding to prw'nc is either *parwdnak or *parwak (to be pronounced :
*parwdne and *parwe); for the latter cf. lease < kdsak " kachgarien "
beside kdsdnc " kachgarienne " (q's'nc, Mahrnamag, line 146, as part
of a name); again, *parwdnaJc derives from *parwdn or *paru, whilst
*parwah admits only *paru as name of the country ; the initial p-
possibly expresses b- if the scribe of the Mahrnamag (who changed his
orthographical principles every other line) followed Sogdian ortho-
graphy when writing this word. No certainty can be gaiined either
way : Parudn/Barudn and Paru/Baru are equally likely.
The various names of Aqsu have been studied by Pelliot, T'oung
Poo, xxii, pp. 128-130, and by Minorsky, Hudud, pp. 294 sqq.
Paru/Baru could be supported by Skt. Bharuka, Hiian-tsang's Po-lu-
kia (seeLuders,<S*.P.,4.tP., 1922, p. 258); Pelliot (p. 129) has adduced
5Jc )j& mo-lu, *mudt-luk (Karlgren 636 + 574). Final -k(a) is due to
sanskritization.
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
568 W. B. HENNING—
On the different names employed for Uc, see Pelliot, Toung Poo,
xxii, pp. 130-2 ; Minorsky, Hudud, pp. 293 sqq. The present-day
name occurs from the end of the tenth century (viz. in the Hudud
al-'Alam); it is frequently mentioned by Kasyari; cf. also ^ Ifc
uo-ts'i on Bretschneider's map ( = Uc). This form may be due to
turkicization ; before the Turkish immigration the name was *0cu,
Ocu, Ucu, or the like : some such name has yielded 'wcwr = Ociir,
Ocur, ticur; for the final r cf. Marco Polo's Succiur (Succuir), i.e.
Sukcur < J$i>)Wsiuk-t'sidu,1 or even modern Kucdr < Kucd. On
the other hand, it is generally assumed that in the T'ang period the
name of Uc ended in -k; Skt. h[e]cyuka (restored by Pelliot)
as well as its Chinese transcription Jj| JU JJg hi-tsou-kia (*yiei-t'si9u-ka)
render no proof in either direction, as the ending may be sanskritized.
The assumption of final -k depends wholly on the interpretation to be
given to ^flj)£(T'ang historiographers), u'-tsou (tew); the contemporary
pronunciation, according to Karlgren (1317 + 163), was jiu-t'siuk
or jiu-t'sisu,2 and therefore permits a sonantic final. Not decisive
is the ancient name (.Xan period) Uan-su (*-u9n-siuk, A.D. 600) even
if it should be connected with the other forms ; the objections that
can be raised against Barthold's equation of Uan-su with J_j^>.
(Hudud and Gardezi) are numerous.3
It may be permitted to turn back once more to Ark. The present-
day name, Qarasahr, seems to be of fairly recent origin. At the time
of Timur, the region was called dalis,* and this name was still
employed at the beginning of the seventeenth century (Benedict
Goes, Cialis, see Bretschneider, Med. Res., register s.v. Chalish). Ark
(Arg) is attested for the time from about A.D. 800 to 1050 (Gardezi).
There is a gap between Gardezl's time and Timur's : it is hard to
believe that for several centuries a district (and town) of the standing
of Qarasahr should never have been mentioned; it is particularly
incredible that Qarasahr should have been omitted by Kasyari;
1
On Su-t?ou, see Pelliot, JA., 1912, i, pp. 591 sqq.
2
jiu-YM?!1 • 'w\cwr = siuk-]t'si»u : suk]cur.
3
Not only the finals disagree (even * J j * L , Minorsky, p. 294, would be dubious
if the presence of a final -Ic can be disproved), but also the initials (that of the Chinese
name being a sonant) as well as the middle consonants (-»c- as against -n-s-; more-
over a Chinese final -n in the JTan-period does not necessarily represent foreign -n;
in the present case = r, according to Pelliot); furthermore, only contemporary
Chinese names are comparable.
4
Turkish, ace. to Pelliot, T'oung Pao, xxxii, p. 265 (cf. ialis " wrestling"
[Kasyari] ?).
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
570 W. B. HENNING—
for this reason alone, the identity of Sulmi* with Qarasahr (proposed
above, p. 558) should be considered.2
Thefirstmention of Ark coincides roughly with the end of the period
in which Ien-k'i (Qarasahr) though continuously troubled by much
more powerful neighbours enjoyed a status of independence; the
old name Ien-k'i, Skt. Agni, has been studied in full by Pelliot,
T'oung Pao, xxxii, pp. 266 sqq. It seems probable that Ark (Arg)
is a rather unexpected form of the same name. Professor Sieg has
recently discovered the indigenous name, Kuchean akene ypoy;
furthermore, he has assumed that akene is an adjective derived from
Kuchean dke (A ok) " end ", so that akene ypoy would mean " the
country situated towards the end", i.e., " the border country"
(Sb.P.A.W., 1937, pp. 130 sq.). If, therefore, akene, which is at the
base of transcriptions such as A-k'i-ni, agni, etc., is to be regarded
as a case of ellipsis, the possibility that the substantive dk/dke from
which akene derives was also employed as the name of the country
cannot be discarded. This name *dk, otherwise unattested, could be
represented by Ark.
There are two ways of explaining the medial /. Either it is a later
addition, or it belonged to the original form and has been dropped
in Kuchean (and in A). Both explanations would presuppose indistinct
articulation of a post-sonantic r up to complete disappearance (as,
e.g. in southern English), particularly in ante-consonantic position ;
in languages with insufficient orthographical tradition, especially in
cases of oral tradition, wrong forms inevitably follow. Some Central
Asian cases of an unetymological r may be mentioned : Vcur, *Sukcur,
Kucdr (see above); tamdryus : tamdyus (see p. 558, n. 1); Kuchean
yurpdske : (reconstructed) *yavaske, *bagazi : *burgayi (both recon-
structed) 3 ; Centra] Asian karpisaya : kapisa, korttana = Khotan *;
Sogdian fiaryar : vihdra, skord < skoO, yum < \un.h These cases
are not confined to Turkish languages (Pelliot, JA., 1934, i,
pp. 30 sq.).
Nothing, in fact, conflicts with the proposal to regard Ark as an
incorrect form of *Ak. On the whole, however, the alternative
explanation is more attractive ; for, if Ark (or Arg, indistinguishable
1
Turkish ? Cf. sumli-maq " to gibber " (Kasyari) ?
2
But cf. Minorsky, Hudud, pp. 276, 497, on cXf
3
See Pelliot, JA., 1934, i, p. 91 n. 2.
4
In "Fan Yu Tsa Ming " : Bagchi, Deux Lexiques, i, p. 295.
5
See Beichtbuch, pp. 88 sq.^Buddh. S. 'ps'ny, ps'ny (= gfmy x, foayx> n o t
*afsang), mentioned by Pelliot, JA., 1934, i, pp. 30 sq., is different (fra- > fa).
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X
ARGI AND THE " TOKHARIANS " 571
Downloaded from https://www.cambridge.org/core. Western Sydney University Library, on 07 Jan 2019 at 11:05:13, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0041977X0007837X