Você está na página 1de 35


Author(s): H. W. Bailey
Source: Bulletin of the School of Oriental and African Studies, University of London, Vol.
11, No. 4 (1946), pp. 764-797
Published by: Cambridge University Press on behalf of School of Oriental and African
Stable URL: http://www.jstor.org/stable/608590
Accessed: 22-09-2017 06:02 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

Cambridge University Press, School of Oriental and African Studies are collaborating
with JSTOR to digitize, preserve and extend access to Bulletin of the School of Oriental
and African Studies, University of London

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms


1. kha, gnia-, bana- 10. mahala

2. draysa- 11. s-s-s
3. ttriksa 12. t - -l
4. ks 13. bramma
5. Bhadrakalpikdc-str
6. parampula 15. dru-
7. Simplified Consonantal Groups 16. h
8. kremeru 17. ogana, maKa
9. Intervocalic -t-


T. Burrow, Language . . The Language of the Kharo.thi Documents from

Turkestan 1937.
,, Translation. . A Translation of the Kharo.thi Documents from Chinese
Turkestan 1940.
H. Liiders, Textilien . . Textilien in alten Turkistan 1936.
,, Tierkreis . . Zur Geschichte des ostasiatischen Tierkreises 1933.
, Weitere Beitrdge . Weitere Beitrige zur Geschichte und Geographie v
Ostturkestan 1930.
R. Pischel, Grammatik . . Grammatik der Prakritsprachen.
Lexicon Bacot . . . Dictionnaire tib6tain-sanscrit par Tse-ring-ouang-gyal . . . par
J. Bacot 1930.
Zap. Inst. Vost. . . . Zapiski Instituta Vostokovedov.
Zap. Koll. Vost. . . . Zapiski Kollegii Vostokovedenia.
K . . . . . B. Karlgren, Analytic Dictionary of Sino-Japanese.

The following pages are dedicated to the memory of Sir William Jones,
the seeds of whose sowing have borne an abundant crop in the world of Asiatic

T HE word Gandhari has been chosen to head this study as a term sufficiently
wide in its scope to embrace the forms of the one Middle Indian dialect
of the north-west of India, centred in the old Gandhara region, around modern
Peshawar, and which we meet in most varied sources.' Under this name
I propose to include those inscriptions of Asoka which are recorded at
Shahbazgarhi and Mansehra in the Kharosthi script, the vehicle for the remains
of much of this dialect. To be included also are the following sources: the
Buddhist literary text, the Dharmapada found in Khotan, written likewise
in Kharosthi, of which a new reading of the text available in facsimile is given
in BSOAS 11. 488-512; the Kharosthi documents on wood, leather, and silk
from Cad'ota (the Niya site) on the border of the ancient kingdom of Khotan,
which represented the official language of the capital Krorayina, t
K 572, 512 lou-lan < lhu-lan (lou in a series with alternation of k and 1)
of the Shan-shan kingdom, and of one document, no. 661, dated in the reign
of the Khotana maharaya rayatiraya hinajha dheva vijida-simha. With this

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

more copious material must be grouped the scattered traces of the same Middle
Indian dialect in Khotanese, Tibetan, Agnean, Kuchean, the earlier Chinese
Buddhist transliterations, as, in particular, in the Dirghagama of the Dharma-
guptaka sect and the remains in Sogdian, Uigur Turkish, and in Mongol
(in living use), and also in Manchu texts. The modern Dardic languages Sina,
Khowar, Phaliura and others represent the same type of Middle Indian. Much
material in Chinese texts, for most of us hidden and inaccessible, remains to
be gathered and sifted. The preliminary studies of P. Pelliot in Les noms
propres dans les traductions chinoises du Milindapanha (JA 1914. 2. 379-419),
of Fr. Weller in his paper tfber den Aufbau des Pdtikasuttanta (Asia Major
5. 1928), and of E. Waldschmidt in his Bruchstiicke buddhistischer Suitras aus
dem zentralasiatischen Sanskritkanon I, 1932, have hardly realized the
importance of this North-Western Prakrit.
Some of the many problems are touched upon in the following pages.
1. kha, giina-, bana-
The three words kha, giina-, and bana- occur in unpublished Khotanese
documents from Khadalik and Mazar-tagh, east and north of Khotan. The
printing of these texts in a volume of Khotanese texts will not be soon possible,
and accordingly they must be quoted here.
(1) Khadalik 0013 d, 2 a.
1. ganam kha 6 vis'akramtta piha hauda ham ///
2. ganam kusi 2 hamdara prui buda asirya rrahaja
3. gausi byauda kisa 12 [space unwritten]
"Wheat 6 kha, Vis'akamtta gave the price . . wheat 2 kiisa, other ... she
took away. Acarya-Rrahaja (?) received millet 12 kfisa."
The name Visakamtta is shown to be feminine by the accompanying participles
fem. hau.d and bu.da. Visa is the name of the ruling family in Khotan, attested
also as v-j'tta and vTjatta in Khotanese, as vijida in Niya document no. 661,
as bi-ja-ya in Tibetan (see JRAS 1942, 14; F. W. Thomas, BSOAS 11. 519 ff.).
The second part of the name, -kdmtta, would be a regular Khotanese form of a
Sanskrit kdnti (not Prakrit kdnti-) with the usual transference of an -4 stem
to the -a inflexion which survives in Khotanese as -a.
(2) Khadalik ii 3 (wooden tablet).
A 1. [I tti buri va miira hamga phemmasta giqii haudai u aysdamr 10sem
hvamda giqnai 1 bar 1///
2. salya vinausa giQna haudi 3 bana 3 u aysdam kha 4 ausyaki aysdam
kha 2 11 budadatta///
3. fresa giina 1 barn 1 aysdam kha 2 samrgasrai giiii 1 bam 1 u aysdram
kha 2 11 samgaka gqin 1 ba<m> 1///
4. brasurai haudi kha 4 11 mmatti aysdam kha 6 parna 11 suhadatta
giina 2 ba<m> 22 1] brapumnai gilnai 1
1 Either hama or hamai seems a possible reading.
2 Of the two units, one above the other, expressing the 2 the lower has been broken away.
VOL. XI. PART 4. 49

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
766 H. W. BAILEY-

Ba 1. mj mamatti giiina 3 bana 3

2. 1] sucam.dra gfina 2 bana 2
3. 1| budasamga aysdam kha 1
4. 11 sulocam hau1i kha 2 suram 2
5. sufiraka aysdam kha 3
Bb 1. si va ginam pudai
2. samgaka gifnii 2
3. brapurmna i giiin 1
4. sulocam gqnii 3
5. suhadatta 1
6. samigapqin 1
Bc 1. l .si va aysdam pudai
2. budadatta kha parna
3. /// na brapqina kha
4. /// suram kha 3
5. /// rai kha 2
6. /// k kha 3
(3) Khadalik VI 14 b (fragment of a torn document).
1. /// kusa 9 kha 5 ///
2. ganam 2 kusa 13 kha 5
3. gausi kisa 13 kha 5
4. /// <k>isa 15 kha 2
5. ///kha ///
(4) Mazar-tagh a 1.0039
1. /// kha 7 gara kha ///
2. /// tsue drupaiysji2 kha mira 9
3. /// pamdi hamai tsue kf
4. /// 5 ttilava tsue hamai kha
5. /// '>e 3 ysam kha 8 guira 7
Of these documents 2-4 cannot yet be fully interpreted. But evidently
amounts of grain and the like are set down opposite personal names: ganam
'wheat', gausa 'millet', gura 'grapes' occur elsewhere, but aysdam is
unknown; eysam is 'millet' (< *arzana-). Phemmdsta 'to Phema' refers
to the city known in other Khotanese documents and in Chinese and Tibetan
texts (see BSOS 8. 935). The personal names cannot be considered here.
The words of measure are kusa-, kha, bana-, and giuna-. Elsewhere in Khotanese
we find kuisa-' drum ', beside which in Sogdian occurs kws ' drum ' and' vessel ',
and in the related Asi4 are found Digor k'os and Iron k'ais 'bowl, cup'. In
1 The purnii is broken.
2 ganarM and dru are uncertain.
3 Probably the broken aksara is e.
4 On the relationship of Asi (here used for the less satisfactory word ' Ossetic ') and Sogdian,
see Asica, Philological Society's Transactions 1945, pp. 2-3. The Asi word has k' < Old Iranian k
when no aspiration supervened. I would add here that Asi, Iron AsA, is from an older *asia-, or
possibly *Arsia-, but the r has left no trace here.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

another Mazar-tagh document a I 0037, 5 occurs pamjsa kusa tcahau samga

'5 kusa and 4 samga ', and from the medical texts we know that 1 samga
corresponds in Sanskrit to 4 prastha.
The relationship of kha, used in both singular and plural, to kisa resembles
that of khi to milima in the Niya documents: there 20 khi make 1 milima.
The words kha and khi may be related, but the phonetic explanation has not
been found: khi is inflected khiyi, khiyena. T. Burrow (Language 86) proposed
Greek xovs as the origin. Evidently difficulties remain: the %ovs was a liquid
measure, and the Greek vowel ov is hard to reconcile either with khi or kha.
It should be noted that the ratio 1 to 20 is also that employed in similar contexts
in Tibetan documents: 20 bre make 1 khal, used for measuring grain or for
bundles; and in turn bre-bo renders Sanskr. drona (Mahavyutpatti 6766)
and khal, Sanskr. khidr. According to the Dictionnaire thib4tain-latin-franFais
the khal is one modius. In JRAS 1927. 809 we find a document: gro khal
phyed dain giis dain bre bi ' two and a half khal and four bre of wheat '. H. A.
Jaschke's Tibetan-English Dictionary gives bre as the equivalent of about four
The word bana-, inflected with the usual later Khotanese inflexion of a
stem in -ana-: nom., acc.'sing. -am plur. -ana, hence bam, plur. bana, corre-
sponding to the older Khotanese nom. sing. -ana, acc. sing. -anu, nom. plur.
-ana, has been found only in these documents. It may be identified with the
bamna of the Niya document no. 66. That word the editors proposed to connect
with bhana in no. 149 rupya bhana 1, with which T. Burrow (Language 17)
compared 'bhdnda(?) '. No. 66 does not appear in T. Burrow's Translation.
H. Liiders (Textilien 22, note 1) showed that bhana corresponded to Sanskr.
bhdnda in the sense of ' alarmkra, ornament'. Hence we must separate this
bhana from bamna. If bana- is Iranian, it will be bana- < *banda- (as usual
n < nd), which is attested in NPers. band, and earlier in Avestan nivanda-.
It would then mean 'a bundle, packet '.
Khotanese giua- may be traced in the goni ' sack' of the Niya documents,
discussed by H. Liiders (Textilien 6). The n of the Khotanese word is due to
the following i, as is to be noted also in nom. sing. ddraid, acc. sing. draniu
from Sanskr. dhdrani.
2. draysa-
The word draysa- has been found twice in official documents.
1. Ch 00269. 47-49 cu pa gumatt7rai hiye h-vi drraysd bidd ttirkd uha : hiv
rrvi skyesd ye ttu pa bisa cimuda bu.ddmdd 'what then was the royal present
(skyesa-, Tib. skyes "present ") of the Turkish iiga upon the load belonging
to the master (hiye = hTyai, translating Sanskr. svdmin-, BSOS 9. 539) of
Gumattira, all that the Cimuda carried off'. For GumattTra see Khotanese
Names in the New Indian Antiquary, extra series I, p. 1, and for Cimuda,
the Cimul Turks, see BSOS 8. 917 note 1 and JRAS 1939. 87.
2. P 5538 A 34 (an imperial rescript, parau) drryysi-bard stura 'the beasts
(horses) bearing loads '.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
768 H. W. BAILEY-

The meaning is given by the context, and its etymo

are with Khotan. dalys- found in dalsd 2 sing. conj
Sanskr. avaropya (N 76. 42); Pasto lesal 'to load ', Parac
back ', Ormuri daz- ' to load ', Sarikoli 8erz- 'to load ',
seize' (see Morgenstierne, EVP 42; IIFL 1. 251), MidPar
load up' (A. Ghilain, Essai sur la langue parthe 52) < *d
may be either an Old Iran. *draza- or a later metathesi
In the Niya documents no. 329 darsa- 'load', and in s
dars- 'to load' have been recognized to be Iranian
7. 510). The rs, which may represent rs or the voiced
Iranian dialect form *darza- with rz < rz, or a form *d

3. ttriksa

Niya document no. 565 has the following passage

naksatrami trik.a sarvakarya sahidavya, which I propos
dragon naksatra (period) all grievous affairs must be
take triksa as 'harsh, sharp, grievous', and to identify
ttriksa, which renders Sanskr. karkasa 'harsh' (Ch 0026
ddruna 'cruel' (Suvarnabhdsa 62 v 2) and Tib. rno-p
136 r 1). Khotanese as often presents us with a variety
ttrimksa, treksa, ttrreksa, ttrraiksa, ttraiksa. This w
H. Liiders (Tierkreis 5), who saw indeed that the spell
implied *triksna, but who thought of Sanskr. drdksd 'vine
who proposed 'endurance(?)' (Translation 111), and e
but later 'insatiableness', comparing Pali a-tricchd (Tran
If then we identify Niya triksa with Khotanese ttriksa,
discussion. The Niya word may be explained from a
*trtksna, derived from tTksna by the intrusion of -r-
the suprascript strokes. It has been known since E. J.
(Kharosthi Inscriptions 320-1) that a suprascript stroke
viga, naga, sug'uta (beside sugnuta), sig'aya (beside signa
grineydti, sdtra, sapaka, Msna, sunade (beside snunade),
related foreign words vighna, nagna, krsna, grhnT-, sn
MidPers. dasn, show a nasal. To this list we can add the
in no. 509, that is, the Iranian *gavazna- 'hart' attested
Sogd. y'wzn-, Av. gavasna- (sn < zn), Asi Digor y.auanz
quaz (fem.) 'deer', ZorPahl. gw'zn (GrBund. 96. 1, se
gavazn. The same use of a suprascript stroke is found also
tions in India, where we find ta.sa 'trsndi' (ed. Konow, n
(Kalawan inscription 1. 4, Epigr. Indica 1932) beside
sunhi. The problem is: what was the Gandhari pronunc
1 It should be noted that in the Niya documents z can be indicat
as shown by alternative spellings, such as asimatra, ajhimatra, asimatr
(Khotan. -ysina, Uigur -zyn); busmoyika, bujhmoyika.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

with a nasal following a consonant ? An indication may be afforded by con-

sideration of the treatment of agni 'fire'. In the Kharosthi Dharmapada, where
a similar use of a suprascript stroke in nasal groups is not found, agi represents
an older agni, no nasal being written. T. Burrow has recognized agnisala
'fire-room' in the agisala of two Kharosthi inscriptions (The term agisala in
two Kharosthi inscriptions, volume in honour of Sir Aurel Stein, Journ. Greater
India Society 11 (1944) 13-16). In Khotanese the same Gandhari word occurs
as a loanword in the excellent old text E 23.98 as amggdsdlu, where the nasal
precedes the g. It will have been seen that in Pall sunhd the nasal has been
displaced, as it has also in Pali tanhd ' trsna'. For other Indo-Aryan forms
see J. Bloch, La langue marathe 140. In a later type of Khotanese the m of
ttrimksa is hardly to be trusted as proving a Gandhari nasal, and the Khotan.
word appears more commonly without a nasal. The superior stroke in Niya
triksa should, however, certainly be identified with the stroke used where
originally a nasal had stood: an intermediate stage between *trtksna and
Khotan. ttriksa seems to be intended, which may have been *trimksa.
The postulated *trTksna < tiksna actually occurs in the Buddhist Sanskrit
text of the Tibetan-Sanskrit dictionary of Dirghayur-indrajina (published in
facsimile by J. Bacot, 1930), folio 98 a 3, where trzksnam is rendered by rno-byed
'sharpened '. Unfortunately the case is not cogent, since a similar intrusive -r-
occurs in 91 a 1 hrimsrah, that is, himsra- 'harmful', and 17 b 1 trarksyah
rendered by khyuin 'mythical bird', in Sanskrit tdrksya. The -intrusive -r-
which is found also in New Indo-Aryan is likewise of uncertain date. Thus
Tirahi has triyna 'sharp', West Panjabi trikkhd 'sharp', Kasmiri tryukhu
'acute' (see R. L. Turner, Nepali Dictionary 282; G. Morgenstierne, Acta
Orientalia 12 (1934), 166). Similarly Panjabi has tru.ttha, Sanskr. tusta. Note,
too, the words discussed by G. Morgenstierne in Western Pahari (Acta Orientalia
6 (1928), 305-7).
A second Middle Indian form derived from tTksna- is preserved in Khotan.
trmkha 'peak', from a dialect where ks > kh-, -kkh- (as we find two distinct
dialect forms from bhiksu- in the Kharosthi Dharmapada : Gandhari bhiksavi
(plur.), beside bhikhu). First *trtksna has yielded *trikkha- (as in Panjabi
trikkhdi 'sharp') and the doubled consonant has been replaced by the nasal
and single stop. This replacement is well known, and reference can be made
to R. Pischel (Grammatik p. 203): jappati beside jampati, Sanskr. jalpati; or in
Gandhari the words attested by Khotan. ja.mph- ' to state ' < *japph-, Sanskr.
jalp-, and the proper name Khotan. (N 171.7) kanmphina-, beside (P 2933. 9)
mahdkaphaind, corresponding to Buddhist Sanskr. kapphina- (with a variety
of other forms collected in Gauri Shankar, Sivasvamin's Kapphindbhyudaya
1937, p. xl ff.), and Pali kappina- from kalp-. The Gilgit MS. has kapphina
(JA 1932. 1.35). Khotanese as often has various spellings: trmkha-, ttrraikha-.
The explanation of Sten Konow in Saka Studies 185 of t.rmkha- < tri-.ringa-,
adopted by M. Leumann in the Glossary to the "Didactic Text" E, could
never have seemed satisfactory.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
770 H. W. BAILEY-

Beside these two derivatives, we have in Khotanese a

form ttTksna-.
There remain to discuss the problems of the inferio
and also of the value of the symbol Y ks.

4. ks

The Kharosthi has two signs Y and -. A.-M. Boye

JA 1911. 1. 422-430) pointed out the distinction betwe
evidence that y corresponded to Sanskr. ks, while Y co
He had become aware of the difference as one of the e
documents from Chinese Turkestan, and the distinctio
throughout in the transliteration by the use of ci for
notation was later adopted in the edition. E. Hultz
Asoka inscriptions (Corpus Inscriptionum Indicarum
the distinction and employed ks for Y in aksati, ad
ksamanaye, ksamitavya, moksaye, samksitena, and vra
aksamti or possibly haksamti should be read). T
sponding to Sanskr. vrksa- he, however, read as rucha
Unfortunately it is not possible to verify it on the fac
version is lacking, but in view of Niya vrksa it is p
passage also *ruksani should be read. T. Burrow
of the use of Y in the Niya texts. Sten Konow in h
inscriptions adopted ks (Kharosthi Inscriptions, Corpu
II 2, 1929, ex), and the same ks will be found in the tr
Dharmapada (BSOAS 11. 488-512).
The evidence for the value of Y in Gandhari is s
decisive. A difficulty arises in distinguishing the
Sanskrit words which came in with the religious liter
1. Y corresponds to ks in Sanskrit words. A.-M.
the evidence and the value of the sign was discussed b
Inscriptions 302) in the Niya documents. I put ks
transliterating Kharosthi below.
(1) Kharosthi inscriptions ksema, siharaksita, bhi
(2) Niya documents ksamni, k.sati, k.itra (Sanskr
The value of the suprascript stroke is considered b
(3) Dharmapada ksadi, ksaya, bhiksu, aveksidi.
(4) Coins with three scripts, Kharosthi, Brahmi,
Catalogue of the Coins of the Andhra Dynasty 1908 p
Brahmi ksahardta, Greek IAHAPATA; pp. 72, 7
Brahmi ksatrapa, Greek IAT<PAIIA). The same
in the Kharosthi inscriptions. Conversely, to rend
used in the Greek name PtAoeEvos written philase
of the Coins in the Indian Museum at Calcutta, 30). Bu

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

of noAv'Evos the Greek ks is rendered by a Kharosthi k with diacriti

see below, p. 796.
(5) Seal inscription with Kharosthi sivaraksita, Brahmi sivara
Konow no. 11). It should be noted that the reading of J. Marshall (Ar
Survey, Annual Report 1914-15, 35) for the Brahmi ' sivarachhi
a lapsus: the ks is quite clear.
2. Indian loanwords in Brahmi script in Khotan, Maralbasi, T
Kuci, and glosses in the Uigur Desasvastika (ed. W. Radlov, .T
Bibliotheca Buddhica xii, 1910). The Agnean, Kuchean, and Uigu
hardly permit a distinction between Gandhari and Sanskrit words
(1) Khotan.
The problem is complicated by the necessity of establishing the value of
the Khotanese ks:
(a) Khotan. ks, Gandhari ks, Sanskrit ks:
raksaysa < Gandhari *raksaza-, Sanskr. raksasa-: here the a before ks
and the z exclude Sanskrit as the direct source; bhiksu-samrgga (E) beside
bilsamngga, later bTsaga (whence Agnean pissaik).
(b) Khotan. ks, tc, khy, Sanskr. ch:
ksatra-, ksattra- (E), Sanskr. chattra 'umbrella'; E 8.32 chadd Sanskr.
ksate 'in a wound', P 4099. 253 tcadai (see Sten Konow, Norsk Tidsskrift for
Sprogvidenskap 7. 22f., which I wrongly contested, ZDMG 92. 588, and
accepted, BSOAS 10. 587). Khotan. richa- in the phrase hauda richa- 'the
seven bears' stands beside rakhya in a fragment Kha. 1. 131a <u)rmaysde
hauda rakhya; in the Sanghata-sutra 25 a 1 hauda rimchdnu hilai, and in the
Siddhasara 3 v 5 khu urmaysdi hauda rricham pa jsdte 'when the sun goes
northwards', corresponds to a Sanskr. rksa-, but here Indian Prakrits and
New Indo-Aryan show Maharastri, Ardhamagadhi, gauraseni riccha-, beside
rikkha- (R. Pischel, Grammatik p. 219); Marathi rTms, ris, Hindi rich (see
J. Bloch, La langue marathe 113).
With this replacement of Sanskr. ch by ks must be considered also the use
of Khotan. ts (a digram indicating ts) in the Sanskrit of Khotan for Indian
Sanskrit ch, that is, aspirated t : gatsa-, Sanskr. gaccha-, with further examples
in BSOS 9. 540. Contrast the same Gostanian Sanskr. ks in saiksaya- ' iksaya-',
baiksu 'bhiksu ', beside which also ks > kh in pakhalaya ' praksalaya'.
(c) Khotanese also adapted ks to Iranian pronunciation with fricative
X before s, and hence introduced a more accurate indication of the pronuncia-
tion Xs by writing h-ks (which I had not understood when writing the note in
ZDMG 92. 588). Thus we find P 2025. 253 rahaksaji jsa ' by the raksasas',
Ch 00266. 176, P 2025. 267 rahaksaja, where older Khotanese wrote raksaysa.
Ch 00266. 178 has raksajai. Note also P 2958. 102 tt7sirahi:ksya ' Tisyaraksita ',
P 2958. 59 ttahiksasilai 'Taksasila', P 2958. 147 drahi:ksg '"araksa', P 2898. 7
durabiha:ksd ' durbhiksa ', P 2906. 13 ahd:ksaubhd ' Aksobhya '. To represent
this Iranian Xs the Greek signs X P were used on Kushan coins in APA OXPO,
OAXPO, OXPO (see my Zoroastrian Problems 65).

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
772 772 ~~~~H. W. BAILEY-

(d) Khotan. ks was used to write Chin. ts-, ty'- < d"i- as in ksahdi:, Chin.~
K1187 tso < I'akc, 4* K 811 tsci < dFi, but Khotan. ch was used for Chin.
!6 <cl'j-, ~-, th?- as ksu, chi fzj K 1245 tsu < d"iu; cha A K 1157 ts'a; chi
bt K 1256 ts'u <tsi o (see ZDMG 92. 587).
(2) Maralbagi Iramian (see Sten Konow, Ein neuer Saka-Dialekt 1935).
Anew Briihmi sign was introduced to represent Iranian Xs' as in x'imane X?ane
'in the sixth regnal year ', and xs'eru adj. ' of the country ', Khotan. kstrul.
In contrast Sanskr. ch appears as cch in cchatra, Sanskr. chattra 'umbrella';
and the Iranian chi- 'to go' (ch = th aspirated) corresponds to Khotan. tsu-
(ts = tt' unaspirated).
(3) Tibet.
In Sanskrit words from literary sources Tibetan has ks for Sanskr. ks
But the early Glossary edited by J. Hackin (Formulaire sanscrit-tibitain 98)
has a variety of spellings to express the sound in words having originally ks:
(a) 4dag-khyi-na, 4dag-khri-na ' daksinia' 'a-nag-khya-ra ' anaksara'

(b) lag-t'a ' laksa' lag-s'a-na ' laksania' byi-ro-pag-A'a ' viripiiksa;
u-prag-ha ' upeks'.
(c) na-kha-tra ' naksatra ', rag-kha-sa, ra-kha-sa 'rdksasa' &&ag-khu
(d) byi-ro-pag-6ha 'vir&ipiksa ' (p. 90).
(4) Agni.
ksatri (Sanskr. Jksatriya), daksimn, daksinakc, h4iksipat, r&ksds-, ksa#-, ksdnti,
kiuT'snti (Sanskr. ksiinti), prdtimokds, yakds (Sanskr. yaksa), niksdntrd~ (anskr.
naksatra). The anaptyxis in kf's. andl -kds may be taken to confirm the value
of k as the first component of ks in Agni.
(5) Kuci.
ksatriye (Sanskr. ksatriya). ksiitre, kui7Cdtre (Sanskr. chattra), daksinake, 1ca'4a,
nteksd~tdr. Here ks for oh of Sanskr. chattra resembles Khotan ksattra.
(6) Glosses in Br5hmI script to the Uigur Des'asviistika:
'pygny abhiksna; byrwp'k'sy virupakse; y'ks'y yaksi. Add to this rk'1 S')~
for which no gloss is given, 'rdk'sasa' (reading probably r'k.g's or r'ks'z).
(7) Turkish in BrThm! script:
jihhapath, Sanskr. S'iksdpada (A. von Gabain, Alttiirkische Grammatik, p. 313)
with hh indicating Xs' and hence assimilated to the Turkish x~as in ohs'a- ' to
resemble'" (see BSOS 9. 293, 299), similar to IKhotan. h-ks for ks.
3. Indian loanwords in Sogdian script: Sogdian, Uigur, Mongol, and
Manchu. Distinction between Giindhdri- and Sanskrit cannot always be made.
Indian ks~ is rendered by kN, but with some assimilation to '
(1) Sogdian.
9'kg'ptt Dhuta 41, S'ks'pt P 5. 66, gk.g'pwt P 2. 435 ' siksipada '; pr"-tymwk?
IDhuta 5 ' priitimoksa' pykg'kw YJ 18a, 28a ' bhiksu' (with pseudo-archaic
-'kw = -u); mwks' VJ 42b 'moksa'-; kcn Dhyiina 394, 'kMn P 2. 121 'ksan~a'
rAx, rkkg P 3. 93. ' raksa' yk 'YU~ P 2. 403 ' yaksa '; pkg in a Sanskrit

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

dhdraW~ ' paksa ' (JRAS 1912, 629 if., 1. 5). In y'ys' *yaXs (E. Benveniste,
BSOS 9. 501) kg has been changed to Iraniuan Xs. Manich. Sogd. c5spS
Cx?'P cX?pt s~ik-spada' shows the same x< <ks (W. B. Henning, Ein
manich. Bet- u. Beichtbuch 123). With this kg, Yv) XS contrast Wc, M6y for Sanskr.
cch in 'yt6g'ntk Dhuta 84, 'yt6y'nty Dhuta 133, Sanskr. iccluintika.
(2) Uigur from Sogdlian.
'kg'r ' aksara ', ksan ' ksania', ksanti ' ksiinti ', laksvan ' laksana', ks'atrik
'ksatriya ', raksvaz, raksvas 'rdksasa ', somaki.v'imi ' Somaksema (F. W. K.
Mulfler, Uigurica 1 33: naote that Uiigur cannot distinguish i from e nor u from o).
Here, too, Xs' replaces k.' in c"aXs'apat ai ' the month &aXsapat', corresponcling
to Manich. Sogd. C"x4'r m'Xy. The spellings vary: in Manichean script
6aygapit, caysaput, t'X.apt-fy (Tiirkische Turfan-Texte III 130, 137).
(3) Mongol from Uigur (see Fr. Weller, fiber das Brahmajdla Siltra (2),
Asia Major 9. 431 if. and B. Y. Vladimircov, Mongolica I, Zap. Koll. Vostok. I,
for such Uigur loanwords).
r'ks's ' r5ksasa ' (Kovalevskij 2660); k.'sn ' ksanta' (J. IRahder, Glossary
to the Das'abhilmika-siUtra, 1928, 58). Here, too, 'sX"p't .vaXs'abad s~ikspada'
has the assimilatedcx in place of ks' (Kovalevskij 1451, Fr. Weller, loc. cit.).
(4) Mlanchu from Mongol.
lakgan ' laksan~a '; rah'as ' riksasa ', where the Manchu sign transliterated Ii
renders Chinese X as in A." Xa, in Manchu Ka.
4. Middile Parthian has yX 'yaksa' and byX?y4 (2 sing.) 'you beg' from
'bhiksu' with x< <ks (Andxeas-Henning, Mitteliran. Manichaica III 52, 66).
5. Chinese: early transliteration from Gdndh5iri.
a gIj K 569, 1154 lo-ts'a < la-ts'at, indicating *ra~tsaz, 'riiksasa'
K224, 1153 ie-tsca < ia-ts'a' yaksa 'beside 0 jX K 568, 1153 iie-tjs'a <jiak-ts'a;
liJ fil K 1154, 527 tsca-li < ts'at-lji 'ksatriya'; K17tst'u tick
ts'iuk 'caksuhy' (E. Waldschbmidt, Bruchstiicke p. 178, no. 113). This value of
ts' for ks corresponds to the Khotanese use of ks~ for Chin. t , as noted above,
P. 772.
6. Tndian words in the Chinese-Sanskrit lexiceons (ed. P. C. Bagchi, Deux
lexiques sanscrit-chinois) show :
(a) laksan~a 5 b 3; cuksi ' good ' 17 a 5 ; naks~atra 34 b 5.
(b) chaya'chdyd'6 b 3 kacha 'kak?a '6 b 3 sahina ' laksna '3 a1.
7. New Indo-A-ryan Dardic languages distinguish Old Indian ks from ch.
Thus Sinii 6phl" 'aksa' 6cho 'aksota ', j.ac ' driiksi', pac 'paksa' rdhi
'raksati', 1ho 'liks', mach-i 'maksik ', but chinoiki 'to split', Sanskr. chid-.
K&44mIr has dach 'drxksi', rachun ' raks- ' with ch = ta', but 6he"nun, Sanskr.
chid- with eh = ts'.
8. The following points can therefore be set out in evidence
(1) In Kharosthi script so far no conjunct of k over s has been noted.
(2) ks and ch are kept apart by the use of distinct signs, hence the value
palatal ch is excluded.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
774 H. W. BAILEY-

(3) The existence of a single sign not a conjunc

probably exclude a value ks.
(4) Khotanese has ks usually for Old Indian ks an
ks also for Sanskr. ch in ksattra, Sanskr. chattra. Th
approached ch in sound and so would permit the in
(ch), that is fricative, rather than the occlusive k +
(5) Kuchean ks-, kus- in ksatre, ku'sdtre, Sanskr.
ksattra; but kus- shows a reinterpretation of ks wit
(6) Khotan. ks for Chin. ts indicates that ks may
or ts' (c, ch).
(7) Between ks and ch, a sound ts, ts' (c, ch) woul
the tongue being brought from k to t but not as far as
(8) A sound ts, ts' (c, ch) distinct from c, ch is known
It seems therefore necessary to conclude that Ga
a retroflex unaspirated ts or the aspirate ts' and w
expressed by c or ch in transliteration.
9. In the Niya documents, although some cases of
suprascript stroke occur, as in no. 358 saksi and no.
the stroke is present. To interpret this stroke two other
at the same time, the c and the k. The stroke whic
was formerly present has been treated above in connect
The c, beside which in the Niya documents c witho
stands where Sanskrit had sc, but occurs also in fo
Sanskr. niscaya; paca, Sanskr. pascd ; pacema-kalam
no. 661, Sanskr. pascima-kdla-. In the Buddhist San
corresponds to Sanskr. sasyaTm ca. The foreign word na
in a Sanskrit Brahmi inscription as nascira in the title
Acta Orientalia 18. 37). Its origin is Iranian1: ZorPa
game', whence the Armenian loanword naxcir-k'; M
nh6yhr ' hunting' (BSOAS 10. 949); Sasanian inscripti
'master of the hunt' (BSOS 9. 232); Bud. Sogd. ny
naXcTr 'mountain-goat', WaXi naXcir 'fox'. Here
sponding to Iranian hi, %c, Xs, Sanskr. sc. Another
is discussed below, p. 795. Cf. also in Uigur vrcik
saniscar ' sanaiscara ', Tiirk. Turfan-Texte, 7, pp. 12
In the Dharmapada pacha < pascid shows the as

1 I explain this word by an Old Iranian *naXti-ci6ra- ' origina

to the epithet Zor.Pahl. tam-toXmak (a translation of Av. tamasc
applied to the wild beasts in Zoroastrian terminology (Greater
we have similarly NPers. hu)ir, huiir 'handsome', Zor.Pahl. M
Av. huciOra-; and also in Zor.Pahl., NPers. anfir 'fig' atteste
'ncyr *anclr or *anjtr (see G. R. Rachmati, Zur Heilkunde der
facsimile, 1. 8). This represents an Old Iranian *ana6i0ra- in ref
that the fig did not flower (see B. Laufer, Sino-Iranica 411). Fo
roam' see C. Bartholomae, IF 38.23 f.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

proper name Hariscandra the sc is represented by c in Khotanese

(see below, p. 799). Evidently between sc and ch or c a sound develo
was indicated by c in the Niya documents (beside c). Possibly som
the sibilant survived, perhaps, as the Iranian h6 suggests, in the
before c: hc, till replaced either by the unaspirated c or the aspi
when ch < sc fell together with ch, Sanskr. ch, as in the words j
Khotan. hardcamnd and the Dharmapada pacha < pascat.
If keyi in the Bajaur casket inscription D 1 is rightly identified wit
kascit, the y, that is, y with fricative-stroke, would represent a v
or c to z. For such voiced resultants of original conjunct consonant
p. 778.
Set over Y ks > ts the stroke would similarly indicate the presence of an
aspiration associated with the sibilant, which would show the stage before the
coalescence of t and s to ch. The case is similar with k, which is discussed
below in reference to makta, p. 795. It will be seen there that k corresponds to
Old Indian sk, but that sk could be replaced by k in the Dharmapada without
diacritic. The diacritical mark may then have indicated an accompanying
aspiration before the k. That aspiration was then not preserved in the Dharma-
pada, or at least not marked.
The three signs ks c k therefore indicate sounds developed out of Old Indian
ks, sc, and sk, which had not yet in the Gandhari of the Niya documents (and
for K, of the Kharosthi inscriptions) reached their full developments into c
(ch), c (ch), and k (kh).

5. Bhadrakalpikd-s?tra

Among the Stein MSS. from Tun-huang is one, numbered Ch c. 001, con-
taining the Khotanese version of the Bhadrakalpika-sfutra, which recites the
names of the Thousand Buddhas of the present age. The preface and epilogue
were published with translation by Sten Konow in his Saka Versions of the
Bhadrakalpikdsutra (Norsk Videnskaps-Akademi, 1929). The first seventeen
lines of this preface are found also in the Pelliot MS. numbered P 2949. The
Buddha names are given one by one in a formula in the Indian nominative
singular as follows: namau krrakasumdau nima tathdgatau, with occasionally
buddhau in place of tathdgatau. A facsimile plate of this MS. is given in M. A.
Stein, Serindia, CXLVI.
Most of the Buddha names are in the form of Buddhist Sanskrit words,
but some show Middle Indian forms. The names quoted here are of particular
interest in this connection. Other versions of the names in Chinese, Tibetan,
Mongol, Manchu, and Sanskrit are given in Fr. Weller, Tausend Buddhanamen
des Bhadrakalpa (1928), which was reviewed by J. Nobel in Asia Major 5. 275 ff.
The source of these Sanskrit names is not made clear.
The numbers before the names from the Khotanese text represent the lines
of the MS.; the numbers given with the Sanskrit equivalents refer to the
numbers in Fr. Weller's book.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
776 H. W. BAILEY-

1. s, sy < sy. 235 ti4au, 278, 559 tisyau, 405 tisyo, Sanskr. 26 tisya. The
older sy is found in 336, 477, 559 pusyau, 291 manusyacadrrau. In Kharosthi
inscriptions sy in kasyaviyana (ed. Konow, no. 34) stands beside kasavi'ana
(ibid., no. 33); Niya documents manusa, Dharmapada manuJsa < manusya;
Khotanese in the proper name P 2958. 102 ttUzirahi:ksya ' Tisyaraksita 'and E
25.202 pulsad 'pusya '.
2. rs < s. 274 heterst, 304 hetirsi, 563 hetersau, Sanskr. 109 hitaisin-. Niya
Bud. Sanskrit no. 511 has hitersina (gen. plur.). In a dharani of Ch c. 001, 941
occurs ahitairsina. Without the -r- we have Jataka-stava 25 v 1 hittesi stana
yuIai duska kird 'being Hitaisi thou didst perform a difficult task.' Similarly
rs indicating rg occurs in Khotan. mahairsUd gvTha: rrumr 'buffalo cow-butter',
rendering Tib. mahehi mar, from Sanskr. mahisa-. Agnean (301 a 4) has mahirsadi
'buffaloes' (Schulze-Sieg-Siegling, Tocharische Grammatik p. 107). From
a Tibetan source B. Laufer quoted Sanskr. mahMrsa (Loanwords in Tibetan,
no. 38, T'oung-Pao 1916). The corresponding word Sanskr. mahisT 'queen'
is found in Khotanese Ch 00266. 124 (Sudhana-avaddna) mahaisa (s =z)
'queen' and in the Kharosthi inscription (ed. Konow, no. 15, A 2) mahesi
with the s representing z which is found also in the Miranl inscription from
Turkestan in esa 'this' and in Niya no. 661, 3, 4 masd, 4 niravasiso and 7
3. s < s. 480 asingabuddhau, 509 asagakausau, 652 asarmgadvajau, 531
asamgakTrtau, 526 asamgajau, 836 asamgaratna correspond to names with
Sanskrit asaiga-: asanga-buddha, -kosa, -dhvaja, -kTrti, -ja, -ratna. In two
other Khotanese MSS. we find P 3513, 4 v 3 asamgaraja, 10 v 2 asamgakausa,
S 2471. 62 asagaraja. The Dharmapada has once B 3 saga = Sanskr. sanga-.
R. L. Turner has called my attention to the s of the Sina forms sam 1 sing. pres.,
segas pret., soiki inf. 'to attach' < sajjati, and sdcam, situs, sacoiki 'be
attached' < sajyate. The change s > s in this word is unexplained.
4. sv < sm. 567 svatiprrabhau, Sanskr. 700 smrtiprabha. There is a scribal
error of t for v in 651 statimdrrau, Sanskr. 869 smrtindra. In my transliteration
of P 3513, 14 r 3 I have stvatyupasthinajdmr, that is, smrti-upasthdna-ja-, but
my reading needs to be checked with the MS. The Dharmapada has svadi
< smrti.
5. s, s < t, th, dh. 375, 547 bhaigrasau beside 390 bhglrathau, Sanskr.
bhdgirathi; 628 visinyau, Sanskr. 824 vidhijna; with 723 visityau miswritten
for *visinyau; 652 masuravattrau (tt for kt), Sanskr. madhuravaktra. The
Dharmapada has gasedi where Pali has ghdteti; vanase'a, Pali vanathaja; sagasa
beside sacadha, Sanskr. sainkhydta, and B 11 ' masuru ',2 Sanskr. madhura. It has
also sisila, Sanskr. Sithila, assimilated from *sisila unattested. In C r 7 'gasana'
represents gdthdndm, corresponding to Pali vdca. Niya documents supply
several forms. We find masu, Sanskr. madhu (H. Liiders, Tierkreis 5-6), where
1 A.-M. Boyer, JA 1911. 1. 415, where 8 was not yet understood.
I put within quotation marks readings taken from E. Senart's study (JA 1898) for which
no facsimile is available.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

Khotanese uses mau < *madu; asimatra, asimatra, ajhimatra 'adhimatra';

no. 16 ajhitrodasa varsa 'over thirteen years' (T. Burrow, Translation); no. 17
ajhi 'too much', no. 562 ajhi 'in addition', which represent Sanskr. adhi:
the alternation s, s, jh indicates z. The AAoka inscription Shahbazgarhi I 2
sasumate represents sddhumatd of other versions. The Taxila scroll provides
bosisatva, Sanskr. bodhisattva (ed. Konow, no. 27); the Lahore inscription
bosavaruma, Sanskr. Bodhivarma (ed. Konow, no. 49). See F. W. Thomas,
BSOS 8. 791. H. Liiders further (Acta Orientalia 18. 25) recognized the same
s < dh in the Kharosthi inscription (ed. Konow, no. 27) de-sama, Sanskr.
deyadharma 'presentation'. Corresponding to the Dharmapada sisila we have
in Niya documents sisila beside sitila (T. Burrow, Language 21). This Gandhari
change can be traced in loanwords in other languages. In Khotanese E 23. 161
mdysila corresponds to Sanskr. mithild. Sten Konow (Norsk Tidsskrift for
Sprogvidenskap 13. 211) recognized in Agnean anas Sanskr. anatha. Khotan.
siysa- 'Siti' corresponds to Agnean sisd- (BSOAS 10. 560). Traces
of the same change are found in early Chinese transliterations. Beside spellings
which indicate the th of Mithila, Chinese has 3g R % K 618, 186, 569 mi-i-lo
< mjie-i-ld, in which -i- represents a foreign z (see provisionally BSOAS 10.
909, 919). Further abridged we have also ~ 1 K 1289, 186 u-i < miu-i
representing *muzi(l). See Akanuma, Indo Bukkyo koyumeisijitenr 'Dictionary
of Indian Buddhist Names, s.v. Mithild, and P. C. Bagchi, Journ. Greater
India Society 10 (1943), 7, 37, 42. For the absence of -I note j_ ; K 342, 1265
kia-uei < ka-wi ' kapila ' (E. Waldschmidt, Bruchstiicke 154).
6. ny < nj. 649 anyanau, Sanskr. 865 anjana; 628 visinyau, Sanskr. 824
vidhijia. This is found in the Dharmapada kuiara-, Sanskr. kunjara-' elephant'
and in Khotanese, Suvarnabhdsa 4 v 6 nairamnia, Sanskr. nairaijand (cf. also
jn in sarvamnia-, sarvaiia-, Sanskr. sarvajna-; abhina-, Sanskr. abhijiia-) ; Niya
gamnia, Sanskr. ganja. See also BSOAS 10, 904, 910, 912.
7. d < t. 239 suradau, Sanskr. 34 surata; 395 anihidavddau, Sanskr. 356
anihitavrata; 570 <a)mrdiprrabhau, Sanskr. 706 amrtaprabha ; 546 samdhiddtse
(t wrongly for k), Sanskr. 658 samShitdinga; 552 saubhidatmau, Sanskr. 671
Sobhitakdya; 640 gaudamau, Sanskr. 847 gautama; 694 ndgarudau, Sanskr.
958 ndgaruta; 465, 692 huddrcau, Sanskr. 498, 951 hutdrci; 689 maruddthivau,
Sanskr. 946 marudadhipa (read marutddhipa ).
Gaudamau is particularly interesting. Khar. inscr. no. 10 has gotama. In
the Dharmapada godama is used, and so in early Chinese a & K 490 kii-t'an 1
< kiu-d'dm, Jap. kudon, to which Fr. Weller referred, Uber den Aufbau des
Pdtikasuttanta (Asia Major 5. 108); Manchu has goodama.
8. au < upa. 355, 459 anaumau, Sanskr. 276, 487 anupama; 351 anauma-
rdstrau, Sanskr. 268 anupamars.tra; 476 anaumakirtau, Sanskr, 520 anupama-
kirti; 558 anaumavddhau, Sanskr. 682 anupamavddin-; 622 anaumasri,
Sanskr. 811 anupamasri. Khotanese E 14. 24 augamo, 6. 14 ttamo, 14. 38

1 t'an < d'dm, not in Karlgren: fan-ts'ie ' jaj K 1129, 650 t'u-nan < d'uo-ndm.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
778 H. W. BAILEY-

autame 'likeness ', Sanskr. upami, seems to have been

with the g, t of hiatus.
9. urdda. 516 ur.d1agarbhau, Sanskr. 601 uddragar
Buddhist-Sanskrit than Gandhari, sincb Khotan. uvd
the regular Gandhari loss of -d-, corresponds to Pal
For the interchange of d - r one may refer to H. Liider
karila- (Acta Orientalia 16. 131 ff.).

6. parampula
Here I would briefly call attention to my discussio
nection between the Asi word, Digor falaembulai, Ir
Tualon alfalamblai 'around' and the words in the Ni
parabulade and no. 586 parampulammi, in the Philologic
1945, pp. 9-11. F. W. Thomas's explanation of t
Orientalia 14, 109-111 as derived from Greek 7rapqE
be correct.

7. Simplified Consonant Groups

J. Bloch (L'Indo-Aryen du Veda aux temps modernes 89) had occasion to
write of the geminees simplifiees. He instanced the future in -h- of the Asoka
inscriptions (h < ss < sy); Pall kahdpana beside Sanskr. kSdrspana; Jaina
Prakrit gdya < gdtra, rid < rdtr ; Bengali gd(y) 'body'; Singhalese rl
'night'; and likewise Pall veth-, gauraseni vedh-, Bengali ber- beside Sanskr.
vest-. R. Pischel (Grammatik 209) had quoted also other words.
Isolated cases of this simplification occur in Gandhari sources. From a
source different from Gandhari, where ks became ts (ts', ch) as shown above,
p. 770 ff., there occurs in the Dharmapada aveha and avehino < apekkha < -apeksa
with h < kkh < ks. In the Kharosthi inscription (ed. Konow, no. 11) adha-
'eight' arises from *attha < as.t, as does Marathi dd(h)-. In the Niya documents
bhighu, occurring twice in one document, in the compound bhighu-sa.mgha,
has gh < kkh < ks, Sanskr. bhiksu. The word uta, once ut'ha, 'camel,' Sanskr.
ustra-, has t < tt, but in Khotanese the word has become ula-, indicating the
further change of -t- > -- > -1-. The word padCa on the silk-strip in Sten
Konow, Note on the Inscription on the Silk-strip no. 34: 65 from a mass grave
at Lop-nor, Eastern Turkestan (Swedish Expedition Reports VII: 1) may
indicate the same development of d < t < tt. Sten Konow himself read pata,
but the g is clear, and was adopted by F. W. Thomas (BSOAS 11. 546 ff.).
When, too, in the Niya documents we have aja beside aja 'to-day', Sanskr.
adya, the explanation will be j (= z) <j (=jj) < dy.
The group ntr appears as n in the Dharmapada mana-bhani corresponding
to Pali manta-bhM.n, with mana, Sanskr. mantra, and similarly tani represents
older tantre. Here belongs a Khotanese word which I had failed to recognize
in the Rama text P 2781. 75 (BSOAS 10, 372), namaian 'I invite'. The passage
should be read pisd vd namanii vistdi 'I invite the teachers hither' (-u beside

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

umr arises from older -dam 1 sing. pres.). The same form of the base nimqn-
occurs in Kha 1. 221, 25 nimqnmiidi 'they invited'. The more usual form is
numandr- (E. 13. 1 numandrdte) and nimamdr- (Kha 1. 221, 23 nimadrrddi),
which represents Bud. Sanskr. nimandr- as found in a text from Kuci upani-
mandrayate (H. Liiders, Weitere Beitrage 7, folio 6 v 6), corresponding to Indian
Sanskr. nimantraya-. In Khotanese niman-, a Middle Indian mana- < mantra-
is attested. It should be added that Khotanese knows also mamdrra-' mantra '.
The group ndr is replaced by mn in Khotanese P 2957. 30 hardcamna
'Harigcandra', which appears in other copies of the Sudhana-avaddna as
P 2025. 129 haracanada, Ch 00266. 76-7 haracadra. In the Sanskrit-Tibetan
glossary (ed. J. Hackin, Formulaire sanscrit-tibetain du Xe siecle, p. 17, 1. 63)
the name is written haracantra.l
Simplification of other groups occurs in Khotanese sida- 'good ', asdda-
'bad ', Sanskr. siddha; snada-, Sanskr. sniddha; S 2471. 260 bauda, Sanskr.
buddha, P 5537. 35 buiddksaittrra 'Buddhaksetra ', and the proper name
namaubudd (see BSOAS 10. 923); ssaddtana ' guddhodana'; SSdya-, with
fem. adj. Ssdtlmje ' gakya '. In the same way the explanation of Khotanese
gava is found. It occurs, with the epithet saglja 'made of stone ', in the
Itinerary 2 (Acta Orientalia 14 (1936) 264) 11. 28, 33. An attempt was made
to explain gava as ' hut ', comparing the hgo-ban of a Tibetan document dealing
with Khotan (BSOS 8. 923 note 1). It may more directly be equated with
Mid. Indian gabbha- < Old Indian garbha-, which is used of a 'cell' or a
'chamber' (as in Pali, Mahdvamsa 27. 15). In Brahmi inscriptions H. Liiders
(Epigraphia Indica 10. 215) has gibha < garbha 'cell' and gabbha 'hall'.
In the Dharmapada gabha- is written. Khotan. gava < gabha < gabbha ' cell'
would give an excellent sense. In other words the geminate is simplified but
remains unvoiced: Khotan. anicca-, anica-, Sanskr. anitya-; sakici 'honour-
ing', Dhammapada sakhaca, Sanskr. satkrtya; agapl 'unfit' < akalpika-;
kapdysa 'cotton', Uigur k'p'z, Sanskr. karpdsa-. Khotan. dnata corresponding to
Niya anada and Maralbasi Iranian anidu may be connected with ijiapta- by way
of *anatta-. The treatment j i> n is attested in this word already in the Asoka
inscriptions, Mansehra and Shahbazgarhi anapemi. Niya has also anati
< jniapti- (with n, not n), and anatena 'by order of'.
In Chinese two Indian words are found with Chinese 1 in place of Indian tr:
maitreya and trdyastri.msa.
In the early Chinese transliteration Maitreya is rendered by 3 I K 618, 523
mi-la < mjie-lak, Jap. miroku. Beside this must be set other Central Asian
forms of the name: Khotanese mdttrai (mdtr-, mitr-, maittr-), Sogd. Bud.
m'ytr'k, mytr'y, Manich. Sogd. mytry, Uigur m'ytry, mytry, mytryy, Mid. Parth.
mytrg, Mongol maidari (in modern use in Kalmuk madr, madrrd, G. J. Ramstedt,
1 Though perhaps little probant in a proper name, the c, c of this name indicates what
developed from Old Indian ic, see above p. 774. Uigur has xaricantri.
2 A new edition of this text has been in the hands of the printer since 1942 in Khotanese
Texts II. G. Morgenstierne has some useful notes on the text in Norsk Tidsskrift for Sprogviden-
skap 12. 269 if.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
780 H. W. BAILEY-

Kalmiikisches W6rterbuch 259), Manchu maidari, A

maitrdk, maitreye. The Pall metteyya with tt < tr i
the final syllable, it is to be noted that in the Gandhar
-aya, -eya may be replaced by -aga, -ega with g represe
(= y): so we have aprameya beside apramega; rajada
ing', beside rajadareya, rajadhdraja, rajadhdreya
(see T. Burrow, Translation 125). Similarly y, writte
Dharmapada udaka, Sanskr. udaya.1 Hence the Gind
could have been *metreya-, *metraya-, *metrega, *met
is attested in Khar. insc. no. 68. From the form *me
Agnean metrak, and Kuchean maitrlk may be di
maittrai would more easily come from *maitraya- o
desired represent *metraia. It should be added that
and -e appear beside -aga and -age; and the Tndian
sented in Agnean by -ak and in Kuchean by -ak
daksinke; riSak beside rsdke.
If now we compare Chin. *mjie-lak, Jap. miro
further change of tr > dr > I will be noted.
The second Indian word trdyastrimaa is common in B
appears in the various Central Asian languages. Both
Indian form are met with. Here only the Middle In
Khotan. tt0vatriTa (Suvarnabhdsa 4 v 7, see BSOAS 1
Kuchean tapattris. Chinese has Iln i]J tau2-li < tdu
absence of the final sibilant -s(a), that is Z, is consi
we may note the 1 corresponding to -tr-, probably
change tr > dr > 1.
For this treatment of dr > 1 we have Niya cula4a-
occurs in the Khotanese name of one of the sixteen
ciida-pantho sthiru, S 2471. 77 c.dapattai sthiri, Ch
The word is well attested elsewhere: Sanskr. ksudra
cula (see R. Pischel, Grammatik 223; R. L. Turner,
chulyahd). Similarly Old Indian bhadra-' good', Bhad
occur beside bhalla-, Bhalla, Malla (see J. Przyluski,
1. 9). J. Bloch (La langue marathe 144) lists Marathi
Sanskr. drdraka; ol, Sanskr. udra, and others. In Kh
stava 5 r 4; beside variant spellings elsewhere: l
has kept dr, but Pali has ludra, rudda, and ruila (see
Gott. Ges. Wiss. 1898, 1-2).
In contrast to these two Chinese words the word
iJ] fq1 K 1154, 527 ts'a-li < ts'at-lji with t-l for tr.
1 To represent y the Kharothi inscriptions have k, as in V
Epigraphia Indica 34) = *viSaya? or k, that is, k with fricative-s
bhagavato (ed. Konow, no. 15, A 12) and probably also in diva
Niya documents have -kala, -gala, -yala 'time' and the Dharm
2 For tau < tdu, not in Karlgren, the fan-ts'ie is 4 4 K 1187

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

8. kremeru

The word kremeru is found in two Niya documents, nos. 318 and 660.
H. Liiders in discussing the textiles of Turkestan (Textilien 35) did not explain
it and in T. Burrow's Translation it remains untranslated. No. 660, printed
here to simplify discussion, presents us with a list of textiles sent, received or
bought, preceded by a list of personal names. [No facsimile is available and
the reading of n - n cannot be checked. The editors have failed to distinguish
them even when they are quite clear, as e.g. in no. 714.]
Niya 660
Columnl A
1. .i puna khvaniyade gadagalade pata pamdura nikastamti
2. pusgariyade rayaga pata 1 presitamti
3. cetrakirti sanapru 1 gimnita
4. rathapala palaga-varna 1 gimnita
5. daruge pata 1 gimnita
6. micga'e palaga-varna nutamna krita
7. kapota'e bamdhitaga palaga-varna 1 gimnita
8. pumiiasena 4 3 pataamca gimnita
9. mogayasa bamdhaga kremeru nutamna kritamti
Column B
1. parvatiye patamca 2 gimnitamti
2. namilga'e sanapru pata 1 gimnita
Translation :-
1. ... again from (= after) the time of his leaving the Khvani were issue
silk-rolls, yellow
2. from Pusgari(ya-) they sent 1 red silk-roll
3. Cetrakirti took one vermilion <silk-roll)
4. Rathapala took one <silk-roll) variegated
5. Daruge took one silk-roll
6. Micga'e bought <one silk-roll) varigated new
7. Kapota'e took one <silk-roll> bound up (?), variegated
8. Pumniasena took seven silk-rolls

9. Of Mogaya they bought <silk-rolls) bound up (?), crimson, new

Column B
1. Men of Parvata took two silk-rolls
2. Namilga'e took one vermilion silk-roll
As the translation tries to indicate, each entry seems to refer to a pata
'silk-roll' whether the word is expressed or not; so sanapru A 3 is without,
sanapru B 2 is associated with pata. Clearly several adjectives of colour are
employed. In A 1 parmlura should in spite of the md be equated with Sanskr.
pdndura 'white, yellow'. Since n replaces nd in the Dharmapada as in pina,
and we have bhana (? read bhana) in the Niya document 149, Sanskr. bhadnda-,
see above, p. 767, the word here with md may have passed through another
VOL. XI. PART 4. 50

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
782 H. W. BAILEY-

language in which .d was replaced by md (as for ex

and md interchange: hvandi, hvamd ' men '. Cf. also
pimtwdt, but Agnean also maE.dl-, pdndukampal). The r
here was identified with rayaka ' royal' in T. Burro
be connected with rdga ' redness '. A 3 sanapru and
'vermilion' and to be recognized as yet another form of
in Old Persian sinkabrus. A 4,6 pala4a-varna 'of palaga c
word connected with the Mid. Iranian pal- of ZorPahl.,
falan', farang), Syriac loanword prng, which means 'spe
the word musk: palang-musk 'speckled basil' (see B.
Pasto has prang 'leprosy' (Morgenstierne, EVP 59); a
paesa- 'leprous', ZorPahl. NPers. pes beside NPers
variegated ', the word probably first referred to a
find further Pasto prang 'panther ', prangai 'leo
palang, Sogd. (P 3. 137; 13. 26) pwrSnk-, and also
words in Greek rrapSos, 7rcap8aAts, -os and Sansk
parond and Khowar pardum. It is evident that Iran
'variegated '. In Khotanese I < rd is regular: h
would suit here. Thence Niya palac a 'variegated'. T
could be Iranian (Avestan has varanah- ' colour ', Kh
< *varnaka), and if so borrowed together with palag
the Indian varna-. A 7 bamdhitaga and A 9 bamdhag
a ' bundle ' or' roll' (unless the silk was prepared for us
In kremeru I see also a colour name, and hence, it wo
the word for ' crimson, red ' that we know in ZorPahl.
Greek KEptK<p>- (see BSOS 6. 946), Armen. karmir,
krm'yr VJ 838, New Sogd. (Yaynabi) kimer ' red ' (S. I.
sogdijskie sootvetstvia, Zap. Inst. Vostok. VI, 1937,
of a raksasa in the Mahabharata Kirmira seems to be this same word. If
then kremeru is a modification of this word, its form requires a brief explanation
In the first syllable -re- replaces an older -ar- or -ir- by a metathesis particularly
common with syllables containing -r-. In the Niya documents we have trubhik.a,
Sanskr. durbhiksa. In Khotanese we have, Jdtaka-stava 26 r 4 dsrTvd&,
Sanskr. dsTrvada-, and in purely Iranian words grama-' hot ', Old Iran. *garna-;
ssandrdmatd- 'the goddess gri' < *svantd drmati-. See also on dru- <dur-
below, p. 791. The -e- of the second syllable can be compared with the e of
peta, Sanskr. pTta-, in the same Niya document no. 318, and the similar e,
Sanskr. j, in the Dharmapada balaneku, Sanskr. baldnmka-; neda-, Sanskr.
nt.da-, ksera-, Sanskr. ksTra-. Less reliable are the spellings of the Chinese-
Sanskrit lexicon (ed. P. C. Bagchi, Deux lexiques sanscrit-chinois I, p. 263)
seta-, Sanskr. s~ta-; p. 255 degha, Sanskr. dTrgha-. There remains the final -u:
two explanations seem possible, it may be the inflexional ending -u < -am
of an Iranian word or may contain the same suffix -iu found in the Khotanese
adjective ksTri ' of the country ', or the corresponding Maralbasi Iranian

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

XSeri-. The phrase in the Niya document no. 318 with kremeru is kremeru-
paliyarnaga prahuni 'a garment of crimson colour', if paliyarna represents
an Old Iranian *parivarna-, but possibly here, too, a derivative of *parda-
'variegated' is concealed, and it should be understood as *paliya-varnaga
' of variegated colour'.
9. Intervocalic -t-

The writing of intervocalic -t-, beside g, v, y where earlier k, g, y had existed

is well known in the various stages of Khotanese (see BSOAS 10. 572-3;
Sten Konow, Saka Studies 26, Khotansakische Grammatik 26). In the excellently
written manuscript E dtama represents (1) Sanskr. dgama- and (2) an Old
Iranian *dkama-; and dtdsia represents dkdsa, whence later dvasa and dsa
were made. The co-existence of the older Khotanese spelling with intervocalic
-t- where Old Iranian had had -t- (as in pitar- 'father') and a newer system of
spelling when Old Iranian -t- was no longer spoken in Khotan could have led
to a recognition of an unpronounced -t- in the orthography whence it might
be extended to places where -t- had never existed, but where either a hiatus
or a glide consonant (y, w) had been developed.
But traces of a similar -t- occur outside Khotanese. In Uigur m't'r *matar
or *madar (F. W. K. Miiller, Uigurica 3. 92) corresponds to Sanskr. makara-,
and in Agnean and Kuchean mdtir (Schulze-Sieg-Siegling, Tocharische
Grammatik p. 62) is the same word. It has survived in Mongol m't'r, whence in
modern Mongol, Kalmuk has matr (G. J. Ramstedt, Kalm. Worterbuch 258);
and in Manchu madari 'monster'. So far in Khotanese only the spelling
magara- has been found. A second word in Uigur is rytybyd *ritiwed (or -wid ?)
representing Sanskr. rgveda in the translation of Hiian-tsang (A. von Gabain,
Briefe der uigurischen Hiien-tsang-Biographie 413).
Ought a solution to be sought, not in Khotanese, but in Gandhari ? A
favourable ground for the development of this use of intervocalic -t- would
be provided if two uses existed side by side in Gandhari: (1) intervocalic -t-
written and pronounced, and (2) intervocalic -t- replaced by a written alif
and no longer pronounced as t (a similar state is familiar in England where,
e.g., the word 'matter' [maetr] or [mat'jr] of one speaker stands beside the
[maer] of another). From this uncertainty the intrusion of -t- into written
or spoken words where -t- had not previously existed would represent a case
of pseudo-archaism. In Gandhari the first condition seems to have existed.
Thus while intervocalic -t- is normally written in Kharosthi inscriptions, and in
the Dharmapada, and in the Niya documents t and d are used indifferently,
certain cases with alif in place-of -t- occur. The inscription edited by T. Burrow
(Journal Greater India Society 11 (1944) 15) from a lamp found in Swat has
sagami ca'udisami 'in the sangha of the four quarters ', where ca'u- represents
catur-;. in other inscriptions catudi&e, caturdiJe, cadudiJe occur. The same
word for ' four ' in the Dharmapada appears as ca'uri, corresponding to Sanskr.
caturas, and as ca'ura three times in the Niya documents. In the Kalawan

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
784 H. W.; BAILEY-

inscription, 1. 2 (edited by Sten Konow in Epigraphia

with unmistakable 'i, has pra'i- from older prati- w
inscriptions. On the Bajaur casket D 3 (Epigraphia Ind
vite, with the t with fricative-stroke (= 8) in prati-,
first d < t in cadudise quoted above. In another inscri
nirya'ide corresponds to Sanskr. nirydtita-. Here an
attested in the Bud. Sanskrit of Niya document no. 511
with -d- in place of -t-, while Pali has both niyyate- an
development of this word is shown by the Khot
occurring in the Suramgama-samddhi-sutra 138
Khotanese, elsewhere in later spelling narasa/-, in w
an older -d-, as is usual, see below, p. 794. In the Dh
avha'i, Sanskr. dbhdti.
Such an intervocalic -t- can be traced further in Chinese transliterations.
From the Dharmaguptaka text (E. Waldschmidt, loc. cit. 176, 233) we have
51 gg K 647, 1015 na-t'ou < nd-d'au representing Gandhari, *nddo or *na'o,
Sanskr. ndgo. In the Dharmaguptaka text (E. Waldschmidt, Bruchstiicke
176, 180, 188) however where the Gandhari word corresponding to Sanskr.
maharddhika- written a1 O p tp K 593, 996, 890, 342 mo-t'ien-t'i-k'ie <
mua-t'ien'ie-'iei-g'ia, | | I K 985, 1330 mo-t'ien-t'i-ii < mud-t'ien-d'iei-zo,
| | | K 224 mo-t'ien-t'i-ie < mud-t'ien-d'iei-ia would yield Gandhari
*matedhiga, -iya, we must accept the West Chinese dialect pronunciation of
X hien < xien, which Professor G. Haloun has mentioned to me (see also T.
Watters, On Yuan Chwang's Travels in India i, 135), and hence read *mahe-
dhiga. The -n of t'ien need not indicate an n in the Indian original, as we
can see also in e X ^ K 683, 551, 47 pa-lien-fu < pa-ladn-piuat' Pataliputra'
through a Gandhari form *pdliputr(a), but here Greek spellings show HaAfldoOpa
and HaA't4poOpa. Below, p. 787, we have *zian-d'au beside *zia-dau.
The Kalmuk matr shows that at some stage the -t- replacing the -k- of
Sanskr. makara was pronounced as a dental.

10. mahala

The unexpected form of Agnean mahur, Kuchean mahur 'crown' with

medial -h- in contrast to the Sanskrit k of makuta, mukuta, from which by way
of Gandhari the word probably reached Agni and Kuci, naturally caused some
hesitation in identifying the words. I was able (BSOS 8. 914; 10. 902) to
quote a Khotanese murkhuta. It has still only been found in one passage.
This, too, has its peculiar form with kh < k, and the intrusive -r- which changed
the rhythm of the word, as Khotan. mursala-, mdrsala- ' hammer ' has modified
Sanskr. musala-. The Gandhari form to explain Agnean mahur is *mahud'a,
and this can probably be traced among the Indian words which have passed
into Manchu, presumably by way of Mongol, Uigur, and Sogdian. In Manchu
hnahala ' crown ' in the phrasefucihingge mahala' Buddha's crown ' corresponds
to Chin. j kuan 'crown', Mongol tydym 'diadem' (from Sogd. SySym

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

< 8ctarSla), and Tib. dbu-rgyan ' head ornament' in the chapter of Buddhist
words of the Four-Language Mirror, the Duiin hacin-i hergen kam6iha buleku
bithe. It should be noted that the sign transliterated h in Manchu is X (unvoiced
fricative). I. Zakharov (Mancizursko-russkij slovar) renders mahala by gapka
' crown'. The form mahala has been developed from *mahula just as Manchu
subargan 'sepulchre' corresponds to Mongol suburgan, Uigur suburyan, on
which word one can now see W. B. Henning, Two Central Asian Words, Philo-
logical Society's Transactions 1945, pp. 157 ff. The second syllable contains
an unstable vowel here, as in Mongol and Turkish, to which P. Pelliot has had
occasion to draw attention (T'oung-Pao 37, 91, note 5; 96 note 2; 99 note,
and 109).
The further fate of mahala in Manchu must be noted. A denominative verb
mahala-la- 'to wear a crown' was created. But also by cutting off -la as if it
were their own common suffix and adding -tu, -tun they made mahatu, mahatun
'hat', as they added -tu in temgetu ' mark, seal' from Turkish tamya; the
final n is unstable. In rendering the New Testament into Manchu the translators
used mahatun to translate Sta3bS~La.

11. s-s--s

1. Khotanese viysama- 'unsuitable' corresponds i

visama-. In the Siddhasara 7 v 2 occurs prara ttydm s
140 v 3 rani-bzin-gyis mi mthun-pa yin yan 'the char
able'; 145 v 3 khu hv.da viysama hame, Tib. 210 v 6
zos-pa dafn 'when the food is unsuitable'; 8 r 3 viysam
snoms-pa 'unequal heat'; P 2897. 37 vgysama salt
visama-vars.h of the Suvarnabhssa (ed. J. Nobel 94); C
hvldina khi.stana 'with improper food and drink'. Th
to be derived from a Gandhari form *visama- with s
in Khotanese is prasama-' proper ', attested in the Siddha
hi arve jsa 'with suitable medicament'; 152 v 4 pre
prisamai, Tib. mthun mthun' suitable to each case '. A ne
occurs in 1 bis r 1 aprrasama arve 'unsuitable m
aprisamdm hvaddm khastdm jsa 'from unsuitable
pra- can correspond to either pra- or prati- of Old In
original here, the word is the same as pratdsama- E 2
For those who cannot venture to go to the Vedas
of the Niya documents (see F. W. Thomas, BSOAS 11.
'unsuitable, improper' seems to be the meaning, a co
prasama- seems reasonable. T. Burrow (Language 78)
Would it not be better to derive avasama from a- an
(or pratisama-), with the p treated as intervocalic
seems to be found in Niya picav-, and in the Dharma
prana, Sanskr. prdna-, and payesidi, Sanskr. pracesy

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
786 H. W. BAILEY-

usually preserved.1 In Khotanese pr- is replaced by vr-

rrusta-vrahauna ' red-garbed ', beside prahona- ' garm
2. Khotan. ttriiysa, truysa (Siddhasdra 18 r 3) 'gou
gru-sa (quoted earlier in BSOS 8. 139). Sanskrit has
In the Sanskrit medical text Ch ii 003 ' Jvaka-pusta
55 r 2 ttraplAsindm corresponds to Khotan. ttriiysi.
Sanskrit-Chinese Lexicon 56 a 2 has Sanskr. trapusa with
also in Lexicon Bacot trayusam, Tib. ga-gon, where
for p rather than a Middle Indian form. In India the
from Jaina texts by Ratnacandraji in his Ardhamag
gru-sa, since gr and dr interchange in Tibetan spelli
Khotanese truiysa- from the same Gandhari *tra'uza
to i. P. Pelliot has recently written of the word and
Mongol (T'oung-Pao 37. 86-90).
The name of the merchant Trapusa, with his com
Buddhist legend, who came from North-West India,
Asian texts. In Khotanese the pair occur in E 25.271
Ttraviysa and Vali- (or Valiya-) '. It is from a form
Chinese has taken X 5 K 890, 1309 t'i-uei < d'iei-jwe
interpreted as *t(r)iviz-. His companion is t uiJ K
Jap. hari. Uigur has d'rpwz-y p'ly-ky. Here is
in early Chinese transliteration. flJ iJ tau-li < t
p. 780, as taken from a Gandhari form corresp
ttdvatrTsa-. Here *tduSzz- has survived as tau-li without distinct indication
of final z for which a final -i would have sufficed. This case arises
in Fpi g K 1, 1307 a-uei < d-ngjwei, Ak S K 210, 457 iang-kuei < iang-g
'Asa foetida' from Khotan. amgusdd, whence also Kuchean anhkwas an
Uigur 'nk'pws (BSOS 8. 913). The Chinese can be interpreted as *an
with z expressed by final -i. Two other cases of final z and 8 should be a
here. ~ fF K 721, 569, 649 po-lo-nai < pud-ld-n4i represents a Gand
*vdrdnaz(Q) with -z expressed by -z (which at an earlier period was a den
E- K 766, 1303 san-mei < sdm-mudi 'samadhi, concentration' repre
a Gandhari form *samtd(i). Khotanese s < st in P 4099. 162 sagrrahdvdsa, P 2
39 sagrrdvasyau (instr. plur.), Ch 00270. 4 sagravasau beside E samgrahda
Sanskr. samgrahavastu, suggests a way by which the Chinese translitera
* 4t K 863, 1308 sa-uei < sia-jaii ' gravasti' and jlp , 5 ; K 342, 130
569, 1265A kia-uei-lo-uei < ka-wi-ld-jdwi 'Kapilavastu' have arisen thro
Gandhari forms *s(r)dvas- and *kavilavas-.
12. t-8-I

In the Dirghagama of the Dharmaguptaka sch

Bruchstiicke 166) * B K 863, 569 sa-lo < sja-ld c
of the Bud. Sanskr. Sdtdgiri. Here I is written for
1 Cf. also Khotanese pema 'image' < pratimd, BSOAS

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

occurrences of this I can be seen, loc. cit. pp. 178, 231-3. In the Sangtti-sutra
(studied by S. Behrsing, Das Chung-tsi-king des chinesischen Dirghigama
1930, 30-1) the name of the Mango (amba) grove is given in Chinese as x g-]
K 1015 sa-t'ou < ia z-d'au and j [ K 968, 1015 ts'an-t'ou < zidn-d'au. In the
later Chinese from a Sanskrit text the name is E: f g K 1185, 577, 342
tsa-lu-kia < tsidt-luo-ka with the unaspirated ts for the voiced Indian j. The
Tibetan is known from W. W. Rockhill's Life of the Buddha 133, in the
transliterated Buddhist Sanskrit jaluka. The earlier Chinese transliteration
would then have intended *z'auy. Here, too, we find 8 -1. Similarly in the
word for paper, the t- 1 (or r)- 8 series is found. Bud. Sanskrit kdgata
is Tendered by Tibetan sog-gu ' paper ' (Lexicon Bacot 180 b 2) and djiii-kigata
by bkah-sog (ibid. 6 a 2). In the Chinese-Sanskrit lexicon of Li-yen (ed. P. C.
Bagchi, Deux lexiques sanscrit-chinois I p. 287, with P. Pelliot's note) kakari
corresponds to I-tsing's kdkali 'paper'. Other forms are well known: Sogd.
Manich. q'y8' (W. B. Henning, Ein Manich. Bet- u. Beichtbuch 65); Bud.
k'^y'kh (E. Benveniste, Textes sogdiens, P 2. 951, p. 181), Uigur k'gd' and
k'g'd' (Turk. Turfan-Texte 7. 35), NPers. kdyas and Mongol qagudasu, -n
'leaf of paper ' with the common Mongol singular suffix -sun (Kovalevskij 744).

13. bramma

In the transliteration (printed in BSOAS 11. 488-498) of the text of the

Dharmapada from the published facsimiles I have adopted the doubled mm
in the words bramma (beside brama) and sammasadi. The same aksara is used in
the word gammira (beside gamira) as printed by E. Senart (JA 1898. 2. 235)
which may be used here, although no photograph is available. The reasons
for preferring mm to the mh for which E. Leumann argued in Die Ligatur MH
in der Kharosthz-Handschrift des Dhammapada (Album Kern, 1903) and which
Sten Konow also adopted (Kharosthi Inscriptions cxi) fall to be given here.
It happens that the words brahman- and brdhmana- occur so frequently in
Buddhist texts that the languages of Central Asia show not only Sanskrit
forms of these words but also forms from Gandhari. But the quest is fortunately
able to start from the Asoka inscriptions.
The doubling of the m in Kharosthi script is attested by the word samma-,
corresponding to Sanskrit samyak, in the Asoka inscription Shahbazgarhi 9. 19;
13.5. Mansehra has samya-. The aksara so employed, m written below m, is
indistinguishable from mam, that is, with the subscript m which corresponds to
anusvara of the Brahmi script. The constant use of this subscript m
(= anusvara) may have made its further use seem less remarkable. But
otherwise doubled letters are rare in Kharosthi. The editors of the Niya docu-
ments have adopted .tt in one word no. 143 moti[.ttom] and once ddh in no. 75
mamdhuddhidhasa. Of neither is a facsimile to hand at present. To express in
Kharosthi used the device of mn: amna, Sanskr. anya, and for nn could use mn.

1 For sa < zza, not in Karlgren, see BSOAS 10. 914, note 1.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
788 H. W. BAILEY-

Here, too, in Niya mam coincided in form with mma (se

Rapson, Kharosthi Inscriptions 300). It would ha
methodical to have used m, not m, everywhere for this
could also use mma, that is, m subscript to the pre
by ma.
In the Dharmapada there can be no hesitation in identifying the sign here
in question in bramma with the sign employed in samma in the Asoka
inscription, and the Niya sign for mam or mma. Graphically they intend the
same sign. E. Leumann's arguments (loc. cit.) lack basis, since he had no
realization (at that time hardly possible) of the Gandhari type of Prakrit.
In particular his attempt to trace his samhas- to a form with *smrs- (though
followed hesitatingly by Sten Konow, Kharosthi Inscriptions cxi) was
The quest must be differently commenced: what evidence is there for
the Gandhari form of the Sanskrit words brahman- and brahmana- 2 The
material will be seen to be fairly abundant. It will be noticed that GandharI
are used beside Sanskrit forms.
1. Khotanese (1) from Gandhari, E 1.35 bramananu (gen. plur.); E
brammana-, E 7.20 brdmana-bv&mate; brrammanm, brra.mma (P 2801. 8, nom.
sing.), brramana (plur.); brammanassai ' son of a brahmana ', brrammana-ruvai
'having the form of a brahmana'; (2) from Sanskrit (see the forms given
BSOAS 10. 914) brrthmau, brahmdne, brrahmdm:, bramh.m, brrahamadnna
(gen. plur.); and in proper names brahmadattd, brahmadeva. E has brahmd
(5.11), brahmalova 'Brahma-loka', brahmana-. In the Bhadrakalpikd-sutra
occur Buddha names Brrihmadevau, Brrdhmanau.
2. Agnean (1) from Gandhari bram, pram; bramani, brdmam, prrmam,
brdmninc- (fem.), brdmnune (abstract); (2) from Sanskrit brahme, brahmarsi,
3. Kuchean (1) from Gandhari bran; (2) from Sanskrit brahmane,
brahmasvar; brahmannai, brdhman?avargd.
4. Tibetan bram- in bram-ze 'brahmana'; bram-ma, hbar-ma 'Brahma'
in the Tibetan-Sanskrit Glossary (Formulaire sanscrit-tibetain, ed. J. Hackin
5. Sogdian pr"mn, plur. pr'mnt 'brahmana '. The divine name Brahma
is translated by 'zrw'.
6. Uigur bramnlar (plur.), bramanlar (F. W. K. Miiller, Uigurica II 19, 25),
beside Sanskrit forms brXmadati (ibid. II 22); brXma-cadiski-sudur (id. Zwei
Pfahlinschriften aus den Turfanfunden 33).
7. Mongol biraman, birman (B. Y. Vladimircov, Mongolica I 317, Zap.
Koll. Vost. I) Kalmuk birmn (G. J. Ramstedt, Kalm. Worterbuch 46), beside
braxamadati 'Brahmadatta' (Fr. Weller, Uber das Brahmajala-sutra, Asia
Major 9. 432).
8. Manchu biraman ' brahmana '.
9. Chinese i K 18 fan < b'wvm, transliterated into Tibetan script as

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

barn (JRAS 1927, 294; 1929. 49), Jap. bon, represents brahma. P. Pelliot
treated of this word (T'oung-Pao 25,1928,455), where he pointed to the alterna-
tion of f and I in the phonetic group as a reason to interpret the initial as
older br-. We may adopt *bram as the syllable the Chinese aimed at repro-
ducing. It occurs also in compounds ~ X X K 18, 342, 186fan-kia-i < b'iwm-
ka-i, Jap. bonka-i ' brahmakayika-'. Brahmana is represented by X g p
K 753, 569, 609 p'o-lo-man < b'ud-ld-muan, Jap. baramon. From the Sanskrit
forms with X, y are found g ~ ~ l e K 753, 569, 342, 593, 674 p'o-lo-
xo-mo-na < b'ua-la-ya-mua-nja. Similarly Brahmadatta is found as I g I
j1 X ; K 750, 569, 71, 593, 956, 1006 po-lo-xa-mo-ta-to < b'uat-ld-dp-mud-
d'at-td, Jap. barakamadatuta, but t # jt fan-mo-ta < b'wvm-mua-d'at.
It will be seen that these various languages had no difficulty in expressing
the h if they wished and as they did in Sanskrit words, and their agreement on
the absence of h will prove the absence of h in the Gandhari form, for which
we should therefore expect bramma. The result agrees with the form bramana
of the Asoka Shahbazgarhi inscription and with the word bramanena in the
Kharosthi inscription (ed. Konow, no. 156, where no trace of either subscript
h or m can be seen) and with the name bramadata- in the inscription (ed. Konow,
no. 37). Niya 514 has brammana or bramamna.
For the word sammasadi it is clear that the source is sammrs-, where no
trace of h can be detected, and gammira represents an older *gambira-
< gambhTra-, with mm < mb. Khotanese P 2782.4 has gambira.
14. Medial -a- and final -a
Final -a in Gandhari was, it seems, either no longer pronounced or tending
to disappear, as a spelling like Niya nice, Sanskr. niscaya, makes plausible
(see the discussion by H. Liiders, Acta Orientalia 18. 26 ff.). In later Khotanese
nouns ending in -ana- and -ama- have a nom. sing. in -am and a plural -ana
(or -na and -ama). The anusvara of the nom. sing. -am may be omitted, see
BSOAS 10. 902. We have then dysam 'seat' beside baisa aysna ' all seats'
(P 3513, 79 v 3); dvamr 'desire ', plur. dvama (older dtama-). In the case of
final -n, -m the Khotanese could use the anusvara, but in other cases they
continued to write the full syllable, and it is impossible to prove whether the
final vowel was pronounced or not. The plur. -a survived, however, as -a.
This resembles the Pasto stage, where masc. sur < *suXrah 'red' contrasts
with fem. sra < *suXra.
Evidence for the apocope of medial Indian -a- is more abundantly to hand.
In the Niya document no. 415 samnera ' a novice ' (T. Burrow, Translation 83)
stands beside samanera; in Kuchean sanmir there has been further change
of nm < mn. Agnean has sdmner; Chinese 'i 3 K 846, 618 sa-mi
< sa-mjie, from which B. Y. Vladimircov derived Uigur and Mongol sabi
disciple (Mongolica I 314, Zap. Koll. Vost. I). A name such as Niya 'iparasma,l
1 I use 1' for the sign in E. J. Rapson's table no. 221, which the editors read lp and T. Burrow
has rendered by ly. In kal'ana-, Sanskr. kalydata-, 1' occurs where in Khotanese kaddna- was

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
790 H. W. BAILEY-

livarajhma beside livrasma may be interpreted in the

many examples are found (possibly at times due to a
itself). In Ch c. 001 occur 893 marna-, gen. pl. mar
'death', 906 drniid 'dharam ', older ddrania, 932
ddriind pata ' verses made of dharanis', 925 dvarna
cdrne in a Sanskrit dharani, 101 mahfrgebhya ' mahora
Sitdtapatrd-dhdranm. P 2026. 75 sarnd, 72 sarand; P
'padmasana-', ibid. s~hiysna 'si.mhasana-', 79 v 3
80 r 1 gaurva ' gaurava '. Further in dsa ' sky ', bes
dgaia- ' akasa-'. Note also Khotanese pindvd 'pin
is lost in Ch c. 001, 979 parhdrd ' parihra ', 985 par
Chinese shows similar loss of a medial syllable, but he
for monosyllables may have favoured the spellings
1167 p'u-sa < b'uo-sdt ' bodhisattva '; t , K 744,
(s < older s-) ' Bimbisara'. Loss of final -a is sho
tsi < tsi (ts unaspirated for Indian j) 'Jeta ', for
jiya- (E 21.9 jiya-vana-), jiva rrespura (P 5538 b
rraispurd (S 2471 b 2), ji rraispurai (S 2471. 93); sim
< nd-kiidt ' nagara'; and f K 47 fo < b'uat 'B
In the Sanskrit of Khotan -ipakarna correspon
Khotanese text' upakarana ' (BSOS 9. 531). Exception
In this same bilingual (BSOS 9. 522) Sanskr. sagar
sponds to Khotan. sakhyairma. The older Khotanese
shortened penultimate. Bud. Sogd. has snkr'm.
The consonantal script of Sogdians and Uigurs did
Indian -a- consistently, hence dhira.n appears in Sogdian
in Uigur (given as tarni, darni in the Analytischer
to Mongol t'rny (Kovalevskij 1675), and thence to liv
Ramstedt, Kalm. Worterbuch 387).
The name Ndgasena became in the Chinese Miland
797 na-sien < nd-sien, with nd < ndga, as in Kho
older ndga-, ndta- (BSOS 8. 929). In the Niya docume
In the name of the Dharmaguptaka school S ~ ~
< d'dm 2-miu-tak the au < ayu, a'u has passed to u, a
in Khotan. truysa < *tra'uza-, Sanskr. trapusa-. In th
(ed. Konow, no. 45) H. Liiders has identified dharma
(Acta Orientalia 18. 17) where -a'u- is kept. Khotanes
'Krakusanda' and (E 23.287) krrosumni ' Krakus
-ayu-. Similarly for the name Asvagupta in the Mila
[ El[ K 753, 1346 o-po-iie < 'at4-pua-jvt < *aspa-

1 kie < kiiit, not in Karlgren: fan-ts'ie ) a K 486, 73 ki-

2 t'an < d'am, not in Karlgren: fan-ts'ie M * K 1129, 65
3 The -e'a- resembles the -eya- beside -aga- in Niya documen
4 o < *at, not in Karlgren: fan-ts'ie ,j - K 1288, 57 u-Xo <

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

-a'u-, -ayu- has survived (see P. Pelliot, Les noms propres dans les traductions
chinoises du Milindapanha, JA 1914. 2. 390 ff.).

15. dru-

The Dharmapada has regularly dru- in place of older dur- and dus-: druga,
Sanskr. durga-; drukida, Sanskr. duskrta-; drupamoksu, Sanskr. duspramoksa-;
drumedha, drumedhino, Sanskr. durmedhas-, -in-; drulavha-, Sanskr. durlabha-.
E. Senart (JA 1898) has also A viii ' druracha ', A viii ' drunivarana ', B 8
' drugatio ', B 9 ' savadugatio ', C frag xxii ' drusila ', of which no facsimile is
available. The Niya documents have preserved the traditional dur- in durbala,
durldpa (Sanskrit durldbha), but that dru- was known is indicated by trubhiksa
beside tumbhiksa, Sanskrit durbhiksa-. Khotanese has the Indian durbhiksa
in the forms E durbhiksa-, P 2741. 100 durbiksa, Ch 1. 0021b, b 52 ddrabaiksa,
S 2471. 279 daurabaiksa, P 2898. 7 dirabiha:ksd, and the adjective Suvarna-
bhdsa *27 v 6 durbiksinai. But the form drubhiks;nai (Kha 1. 214a, b 4) is also
found. The original Iranian dus-, duz- is preserved as dus- and dusa- in Khotanese
duspya 'weak ', dusiyuda- 'ill-done ', whence this prefix could be used before
Indian words dusas;la, dusacaidye ' ill-thought '. In modern Dardic the Phaliira
dialect has drubalu 'lean', drug ' ravine'. We find also dhrigo 'long', as we
have driga in the Dharmapada. See G. Morgenstierne, Notes on Phalu.ra
(Norsk Videnskaps-Akademi 1941) pp. 33, 34. T. Burrow (Dialectical Position
of the Niya Prakrit, BSOS 8. 435) has drawn attention to traces of dru- in
Torwali. Traditional knowledge of dur- probably prevented the writing of
dru- even when that had become the pronunciation.

16. h

There are indications that in Gandhari h became unstable and often either
ceased to be pronounced or was emphasized as X. The word for ' householder ',
exceedingly common in Buddhist texts, is translated in the Khotanese text
P 2787. 136 bisgddrai (from bTsa ' house' and ddr- ' to hold '), but more usually
the Indian technical word was adopted. This word in the Gandhari of the
Asoka inscription Shahbazgarhi 12.1 is grahatha-, from Old Indian grhastha-
'remaining in the house'. In the Niya document no. 489 grihasta occurs.
The Chinese knew it as |P j ti K 569, 846, 223 o-lo-so-t'a < ngd I-la-sa-t'd,
beside the reading of the Brahmi garasatha in Li-yen's Sanskrit-Chinese lexicon
(ed. P. C. Bagchi 30 a 3): intended is an Indian *grastha < *grahastha. Since
in Gandhari th, t'h took the place of sth-, -sth-, as in the Dharmapada dhamat'ho,
avit'hido, t'hanehi, and Khotan. vathdyaa-, later vaksdyaa-, Sanskr. upasthdyaka,
beside Niya vat'hayaga, the expected Gandhari form of grhastha would be
*grihat'ha- or *grahat'ha-. The absence of r from the group gr- would, however,
also be possible in Gandhari, and from a *gahat'ha- (corresponding to a Central
Prakrit gaha.ttha, as attested in Pali) the Khotanese took their gdthaa-, later
gdksaa-, with -i- <-aha-. It would also be possible that a Central Indian
1 o < ng, not in Karlgren: fan-ts'ie x fiJ K 1280, 414 u-xo < nguo-yd.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
792 H. W. BAILEY-

Prakrit form of a word of the religious vocabulary

Gandhari, as we find bhikhu in the Dharmapada besi
modifications the word is found in Sogdian k'rt'k, k
Kuchean kattdke (see BSOS 8. 905), for I need not d
A. van Windekens' Lexique etymologique des dialectes
Indo-European enthusiasm has led to an underestim
Once, in Ch 1. 0021b, b 31 Khotanese has md < ma
the usual mahaydna-, and contrasts with Chinese
mo-xo-ien < mud-xa-ain with X for h, whereby it is
Chinese had no h before a. It will be recalled that rem
mahd- was replaced by md- in Muslim books. Ma
corresponding to a Sanskr. mahdacna, see V. Minorsk
Turks and India (1942) 18. Khotanese has P 2896.
sagrrdsyau (both instr. plur.) beside P 4099. 162 sagrrah
'samgrahavastu'. Agnean wydr 'vihara' recalls the n
vyarivala in Niya document no. 415 (T. Burrow, Lang
Beside this we have in other Niya documents viharav
pala '. Chinese has among others the spelling M XN J
< b'ji-xd-ld, Jap. bikara with h replaced by X. Bu
Uigur brX'r, has an intrusive r and X for h. For th
documents T. Burrow (Language 10) has gathered so
beside danagrahana 'giving and taking', goma, g
godhiuma, paropimtsdmani (Bud. Sanskrit) from pa
i < ahi. Intrusive h is found in prihito ' prita ' and
initially: hastama interchanges with astama, Kho
The Dharmapada has intrusive h in parvahi'a 'pravr
written for the very similar sign for alif. The word
5673)' birthday feast' is found in Khotanese P 2958.
P 2025. 98, Ch 00266. 55 jauttamaka, P 2928. 12 j
'caravan leader 'Agnean has sdrthavdk, sdrthavahe, Uigu
sdtdvdiya- (later sdtTka-), Mid. Parth. s'rtw' (Andreas-H
chaica III 61). The word mahi- 'earth' is represented b
35 . 1 K 618, 846, 773 mi-sa-se < mjie-sa-sak ' Ma
but in Khotanese in a dharani Ch c. 001, 898 (Su
makhisvara with kh for h (and i for e) where the co
has mahesvara. This same spelling makhi 'mahi'
in the Niya document 511 (Bud. Sanskrit) in the ver
imdram vivrdhi abhivarsatu makhi
ude.mtu sasya ca jayaya pdrthiva
that is, 'May Indra rain increase on the earth, may the crop grow and the
king go forth to victory.' The editors, thinking of no Prakrit form, equated
makhi with Sanskr. makhe (loc. sing.) 'sacrifice' and T. Burrow (Translation
1 This recalls the r in Khotanese murkhuta-, mursala- from Indian mukuta-, musala-, see
above, p. 784.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

101) rendered by 'lord of sacrifice ', evidently intending makhin-, but a com-
parison with the well-known verses at the end of dramas assures the translation
' earth'. Thus the Ratnavali and Priyadarsika of Harsadeva close with the
stanza beginning:
urv7m udddma-sasydm janayatu visrjan vdsavo vrstim istdm
'May Vasava (Indra), pouring out desired rain, make the earth to abound
in grain.'
In Bud. Sogdiany (= X) represents Indian h: my"m'yh' Mahamaya ', my'ys/r
'Mahesvara ', my'pwy ' mahabodhi'. In Uigur according to the following
vowel either k or X is written: mkyntr'syny 'Mahendrasena'; mx'klp
'mahakalpa '.
17. ogana, maka
These two words occur side by side in the Niya documents nos. 713, 714.
713, 8-9 avanemci pati maka ogana tanu tanu gothade nikhalemi taha na
dharma asti avanemci bhuma amne krisamti tanu gothade asmaga maka ogana
nikhalemi 'I produce the tax of the avana,1 maka ogana from each farm of
mine. Tle law (dharma) is not thus : others plough the avana ground; I produce
our maka ogana from my own farm'.
714, 3 pati cimtidaga ghrida pasava kosava arnavaji thavasta'e raji nammata'e
camdri kammamta 1 amna maka ogana crorma amna suga ekamamta.
In this document several items remain unexplained. So far as it can be
explained, it may be rendered : ' tax calculated: butter, small cattle, woollen
cover, woollen cloth (?), carpet-cloth, . . ., felt, trousers of cloth, one, other
maka ogana, veil, and the whole of the other yarn (?) '.
In this translation I offer certain new interpretations. In arnavaji ' woollen
cloth (?) ' I conjecture an Iranian *varna- or *varnava- 'wool', to Av. vardna-,
corresponding to Sanskr. itrnd. The Indian word kambala 'blanket', common
in Khotanese texts as a loan-word, is absent from the Niya documents. Does
arnavaji represent the kambala cloth ? In cdmdri kammamta here and in
camdri kammamta in no. 272, Obv. 8, I see 'trousers made of cidar-cloth'.
That is to treat camdri as an adjectival derivative in -i to the Iranian *cddar-
attested in NPers. cddar 'cloth, mantle ' from the base cad-, cand- 'to cover'
with a suffix -ar- (or -ara-), such as we find, for example, in Avestan zdvar-
strength '. With kammamta I compare Khotanese kaumadai, kammadd (BSOS
9. 532). This kaumadai translates Gostanian Sanskrit suthamna, which corre-
sponds to the Mahavyutpatti 5849 sunthand ' trousers ', Tib. dor-ma (H. Liiders,
Textilien 22). The Lexicon Bacot 81 b 1 has sunthdnam and nicola, explained
by dor-ma.

1 The interpretation of avana is hardly yet settled. F. W. Thomas has again urged its identifica-
tion with Indian apaa a' market ' (BSOAS 11. 531 f.). The alternative explanation from an Iranian
*dvahana- 'dwelling-place' cannot be excluded linguistically, although one would wish to find
the Iranian word in some contiguous source (its existence in Old Iranian is attested by Old
Pers. avahana-). With bhuma ' land ' either meaning would suit ' land of the market-town ' or
'land of the settlement'. Of the two the more precise dpana might be preferred.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
794 H. W. BAILEY-

For modern representatives of sunthdna see G. Morge

12. 187, who has Tirahi satdn ' trousers'. The Khota
as kabalija 'made of kambala, blanket-cloth'. I see th
Obv. 6 karci kamuta (so read) and for karci I would
Zoroastrian Problems 8), assuming an intermediate I
replaced -t before i (Asi kaerc 'fur-coat' may be ulti
would give f trousers made of karci material'. In cir
I see an Iranian *cirorma from *cihravarma-, whic
Iranian *iOra-varma- 'face-covering, veil' (cf. the Ir
'coverlet' in Armenian < Iranian *varmaka-, quot
Transactions 1945, p. 28). With these words is closely
rotam, in which I think we should see an Iranian *r
have been an Old Iranian *rauda-banda- 'face-bi
ruy-band 'veil'). The earlier proposal (see T. Burr
I also prematurely adopted in BSOS 8. 913) to identif
rodan, Khotanese rrinai, must then be given up.
It will be seen that I have not been able to follow
again proposed (The Early Population of Lou-lan-Sha
India Society 1944, 64) to take cdmdri kammamta a
which was obtained at night while the moonlight w
presumably have meant 'connected with moonlight
as an epithet replaced some word meaning 'jade-sto
I have read 1 in place of the editors' na after kamm
facsimile that it is not n, as a comparison with the v
agetana yatmana in 1. 5 shows (M. A. Stein, Innermo
The word sud'a is still not clear to me. It is perh
The ogana forms part of the tax (paCi) as a farm p
that g can replace y in these Niya documents (see t
p. 780, apramega, Sanskr. aprameya-; samaga, Sansk
*oyana- would be the Gandhari form of Sanskr. oda
of intervocalic -d- is shown in Khotanese (E 17.43) ggoy
gauyini, Chinese early transliteration X #E J K
< kiu-a-nji, in which kii < kiu is a common spelling
9. 541), Bud. Sanskr. goddinya, and Bud. Sogd. k'wy'ny (
Handschriftenreste, Dhydna-Text 115). Beside this th
K 490, 342, 659 kii-k'ie-ni < kiu-g'ia-nji, that is, a G
y < y < d. In Khotanese the name Bud. Sanskr. g
sjiddutana-. The replacement of -d- by -y- is foun
'to present', Bud. Sanskr. acchddaya-, Pali acchdd
-d- is also absent in Khotan. praysd, older (E) pray
'favour'; Khotan. uvdra-< *u'ara-, Sanskr. uddra-
above, p. 778). In Khotan. vna ' pain' (older vedand-,
1 < eya- < -eda-. In the Tibetan document edited by
1928. 575, nor phyugs dain nas bra[s] we have prob

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

with 'rice' for hbras rather than 'fruit', which is expressed by hbras-bu when
standing alone, though as part of a compound hbras may be ' rice ' or ' fruit':
hbras-ian ' failure of fruit', rgyu-hbras ' cause and effect ', but contrast hbras
dkar 'white rice ', hbras-chan 'rice-beer '. Rice was recorded by the Chinese
in Kuci, Khotan, and adjacent regions (see B. Laufer, Sino-Iranica 372 if.,
and for modern times see H. Yule, Marco Polo I 189, ed. H. Cordier). The
Chinese military colonies had wheat, millet, and rice (E. Chavannes, Les
documents chinois decouverts par Aurel Stein, 1913, xiv).
The second word makEa occurs also in another Niya document 505, 1-2
ekhara mogiya uthita uta trodasa parikreyammi seyita tsugenamma pacevara
satu milima 2 khi 10 41 maka khi 4 1 kavasi 1 pacevara pimda milima 3 chata4a
1 katari 1 tena tsugenamma gid'a ' Ekhara Mogiya arose, he took a camel at
a hire of thirteen.1 Tsugenarnma: provisions, meal 2 milima, 15 khi; maka
5 khi; one upper garment (kavacikd); the whole provisions 3 milima; one
garment, one knife: by this Tsugenamma were taken'.
In no. 505 makla is part of the pacevara 'provisions for a journey ', as is
shown by calculation of the total value of the pacevara: 2 milima, 15 khi,
5 khi = 3 milima, since 20 khi make 1 milima. In no. 714 mak'a is part of the
tax and in no. 713 is produced from a farm.
A note on pacevara is first called for. T. Burrow (Language 102) envisaged
a connection with Bud. Sogd. ps"fr ' provisions ', and for the first part proposed
*paOya-, comparing Sanskr. pdtheya. We may add a similar derivative in
Avestan pdOmainya-. Manich. Sogd. ps'fr, pys"fr in Sogdian script is given
by W. B. Henning (Ein Manich. Bet- u. Beichtbuch 63), where a connection with
Avestanpiwd- ' food' is conjectured. The Mahavyutpatti 7182 has pathyadana-
(pathyo-), Tib. lam-rgyags 'provisions for the road'. The same word appears
as sathyadanam in the Lexicon Bacot 173 b 2 with a common error of s for p.
For padevara we may assume an Iranian *paOyabara-' to be carried on the road',
whence with d modified by the i to e would arise *pahzevar, *pahsevar,2 surviving
in the Niya documents as pacevara with c = hc or hs, see above, p. 774. Sogdian
ps",3r, pys"r may equally represent *paOydbara-, but with d maintained.
Modification of ia > e is, however, found in Sogdian. It is possible to cite
VimalakTrti-nirdesa-sutra 41 k's'yp ' Kasyapa', and the Sogdian -'y- in the
imperfect of verbs where the augment followed -i-, as in pt'yyws *pateyos
< *pati-a-gausat 'he heard' (see E. Benveniste, Grammaire sogd. II 28;
H. Reichelt, Studia indo-iranica 248 if.).
This may be the place also to propose an interpretation of the -e- of Agnean
kurekar and the similar Uigur kwryk'r. The original Indian is ki.tdgdra-, which
is found with two meanings (1) ' pinnacled building ', (2) ' bier '. If we assume
an intermediate Gandhari stage *ka.idydr(a)-, it seems possible that the li
1trodasa implies muli milima, as in no. 762 D 1 aspasa muli milima 4 3 khi 10.
2 The replacement of i after a voiceless fricative by the fricative s is the same that is noted
from Gandhari in Chinese ];] K 1212, 819 si-ts'i-sou < sak-sie-sieu representing *hdkhesu
< sakyesu (E. Waldschmidt, BruchstUcke 154). For .si < 8sak, not in Karlgren, the fan-ts'ie is
] K 856, 1223 sang-tsi < siang-tsiak.

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms
796 H. W. BAILEY-

or ri which had arisen from -t- may have induced the

has kulatdra- and kuldra-, corresponding in the S
khan-pa br6egs-pa ' pinnacled house '. In Uigur kwr
with dv 'house' (F. W. K. Miiller, Uigurica III 71,
Manichean Sogdian have kwn'k'r (W. B. Hennin
sense of 'platform' or 'bier ', as we find kiitdgdra
20.36 sovanna-kfutdgramhi 'on a golden bier'.
the identity of kwn'k'r with kttadgara- inevitable,
a replacement of r- r by n - r. R. L. Turner has q
Indo-Aryan Singhalese kuldra-ge' a kiutagara house
Turning now to maka,'we find the distinct Kharosth
mark set over the ordinary k. In the Niya documen
proper names kumsena, kumtsena beside kumnase
vukimte; and also in the words pake, avalika (E. J.
tions 301). Outside the Niya documents the sign i
inscriptions (ed. Konow, no. 80 and no. 65): no.
corresponds to Sanskr. samskdra-; no. 80, D 3 kamd
skandha-. These two words appear in the Dharm
For the absence of aspiration in this kana we sho
inscriptions (ed. Konow) no. 24 pukarani, and no. 44
'pool' beside the Niya document no. 383 pusgar
(T. Burrow, Translation 78) and Dharmapada pu
Pali has pokkhara.
To this must be added the k with diacritical mark written above on the
Polyxenos coin mentioned by R. B. Whitehead in Numismatic Chronicle
xx (1940) 19-20. I have seen only a hand copy, in which the diacritic is a dot,
but in value it may be the same as the dash of this inscriptional and Niya k.
It would then be better to read palikinasa with k corresponding to Greek ks.
In k we have an attempt to express a sound resulting from an older group,
in this case sk. We have then three stages of an evolution: Old Indian sk,
Kharosthi inscriptions k, Niya documents k and k, and Dharmapada k. The
value of aspirate kh would seem to be excluded by the k of Niya and the
Dharmapada. Possibly some trace of the original sibilant survived before or
after the k (with which it would be possible to compare the treatment of the
nasal in groups as indicated above, p. 769). Since we have also k < sk in
Gandhari, it would appear that both sk and sk passed through some intermediate
stage to k.
If now for maka we conjecture an older stage *maska- or *maska-, it must
be asked if the Sogdian word mask- is related. In the older Sogdian source
the letter no. 3 (H. Reichelt, Soghdische Handschriftenreste II 22) 1. 10 we have
m'tskh; Bud. Sogd. m'st'kh, m'skh, mskh; Manich. Sogd. m'sk (see W. B.
Henning, Ein Manich. Bet- u. Beichtbuch 91; E. Benveniste, Textes sogdiens
on P 2. 251). The word has been rendered 'basis, subsistance '. In the Sitra
of Causes and Effects 36-7 occurs rtyms 'sty ZKZY ZKw zw'n 't m'st'kh wn'nt

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms

d'wn wyspn'c
by 'puis
by 'puis
il y ailceux
y a qui
qui 6tablissent
leur vieleur vie
et leur
en en
'. Here
'. Here
to Chin.to Chin.
t .ts'ai
P 2.
P 2.
is what
is what
is furnished
is furnished
to theto
the butcher,
the basic
his meat.
his meat.
The word
The word
with its
laterits later
forms seems
be be
a derivative
a derivative
of *mSta-
of *mSta-
in ZorPahl.
in ZorPahl.
m'tg *mdtak
m'tg *mdtak
' basis
basis ',',m'tkwr
*matakavar ' principal
' principal
', NPers.
', NPers.
' basis'.' basis'.
In connection
In connection
with food
foodthe theword
in the
in the
< Mid. Iran.
< Mid. Iran.
of of
', found
', found
in ZorPahl.
in ZorPahl.
Problems 100).
Niya maxa
corresponds to Sogdian
to Sogdian
as applied
as applied
to food.
to Since
food.inSince in
the Sogdian
textP P 2. 2.
is brought
is brought
into into
with the
the butcher,
the reference
referencemay may bebeto to
as food,
as food,
and hence
and hence
in theinNiya
the documents
Niya documents
be be
specialized,asasa aforeign
foreign word,
The Old
Old Indian
word for
for 'meat'
mdmsa was also used in the form mimtsa 'flesh'.


P. 491: Ac7a omit the dot under pra'ujadi.

P. 497: 0 9 sanamena with s.
0 13 brammano, not mh, as in the Glossary.
P. 511: s.v. sabrayana.na: omit dot under the s of the Pall word.
s.v. samidhi: read samrddhi.

VOL. XI. PART 4. 51

This content downloaded from on Fri, 22 Sep 2017 06:02:09 UTC
All use subject to http://about.jstor.org/terms