critical observations
by F. W. Thomas, Oxford
I am pleased that Professor Bailey should have given
attention to my article. Since a friend whom about eight
years ago I pressed to take up the investigation of the India
Office collection of Saka-Khotani 'documents' found him¬
self too much hampered by other researches and ultimately
decided to give up the task, it is satisfactory that in 1934
he was able to hand it on with my concurrence to so compe¬
tent a successor.
A good part of Professor Bailey's paper is addendum,
partly adding new precisions and information relating to old
matters, such as Saka-Khotani I <d, the Chinese represen¬
tation of Indian cerebrals by palatals, or to matters, such
as Chinese ts represented by tc and y represented by hi {hi:),
brought to light in my article. So far as this particularly
concerns the specialists in Saka-Khotani or Chinese, I need
not enter into it. But in regard to the transliteration of the
initial 24 lines of the MS. what I wrote (p. 17) was —
'For the greater part of this introduction, which has not
been found elsewhere, it would not be difficult, with the
aid of the index given below, to supply the Chinese char¬
acters corresponding to the Brähmi aksaras, since the
sense, consisting largely of invocations of Buddhas and
Bodhi-sattvas, is rather apparent. But the procedure
would be partly conjectural, and this, in the case of Chinese
homophones, could not be entirely free from risk; . . .'
The 24 lines do not belong to the Vajra-cchedikä. If
their text should ultimately be identified in some other MS.
from Central Asia — ^and certainly there are in such MSS.
many passages containing similar matter — it would for some
purposes be welcome.
We should, however, be grateful to Professor Bailey
for furnishing a transcription and for turning it to account
in connection with Saka-Khotani phonology: I append a
note of the points in which the transliteration should, I sug¬
gest, be corrected. But he might have been provided with
an additional control, if he had worked out a translation,
as explained above: and he might also have succeeded in
identifjdng the Gaganasambhava of line 7, whom a serious
search by Mr. Höryü Kuno and myself (in 1936) failed to
find mentioned, by that name, elsewhere.
The profKJsed corrections in the readings are as follows: — 1.1,
kaumi kau (for kimni käm; cf. p. 3, where kau is cited); 1.2, tsyai
(for tsye); 1.3, hau : sü (for hau khü); 1.4, insert ~ at end; 1.5, pvavt (conceivably) (for svavi) and Äa:( ?) for hara{ 1), blurred); 1.6, ämmi (for dmmi); 1.8, hvü : hi (for hvü : hi); 1.9, clikim ? (for diki); 1.10, yvi;
(for yvi); 1.12, insert - after kimmi käm; 1.13, svi (for svi); phehi:
(for phehi): 1.17, kvimni (for kvimni); 1.20, hväm: (for hväm); 1.21,
hem : ni (for hem : ni): iahi: (for iahi); 1.24, paia (for paia);ha: (for
^ji).
In these and other cases I neglect the doubts as to i
and i, concerning which see below.
To come now to my own misreadings. It will be seen
that about 12 items in the list relate to cases of tjv: these
I will consider infra. Those corrections which I accept forth¬
with are —
1. 25, tsimni — ^found also in 1. 74— (for tsimni, a rather
blurred i, perhaps due to hesitation on the part of the scribe);
1. 27, k$i (for cvi; the scribe first wrote c); 11. 27, 30, 73, 74,
tü (for khü — and we may be grateful for Professor Bailey's
explanation of / = Chinese ch); 11.30, 38, damni, dammi
(for da!*)\ 1. 33, si — perhaps an error on the part of the
scribe, who elsewhere gives if and se — (for H); 1. 39, 'itnni — a
scribe's error — (for 'imni, which occurs frequently); 1. 47, di
(for dvi); 1. 52, tsi (for tse); 11. 72, 76, 84, ti—t hardly distin¬
guishable from the d of 1. 71 (see Plates I and II in
Dr. Hoebnle's article JRAS 1911, and Plate CXLV in
Sir A. Stein's Serindia — (for khi); 1. 70, svi (for gvi).
Concerning the others I may particularize a little. 1. 38,
1) Professor Balle v hat von diesen Verbesserungen Kenntnis
genonuneii und einige davon angenommen. !>• H.
Zeitsohrift d. DMG. Bd. 02 (None Foliir Bd. 171 .'JS
3 9«
ttyai (for tyai). We are here deahng with minutiae; but I
think that the correction must be right, because the form
differs hardly perceptibly from that in the Serindia Plate
CXLV, which uses only tt (not t).
11. 68, 71, 75, kvt for jvi. I cannot distinguish the jv
from that in the jvä read by Professor Konow in 1. 63 of
the Stael-Holstein document and by Professor Bailey in
11. 2 and 6 of the document published in Acta Orientalia,
XIV, 258 sqq. ; but, as I also fail to distinguish it from the
kvä read by Dr. Hoernle in line 3 of the document on
Plate XVII of Manuscript Remains oj Buddhist Literature
and by Professor Konow in 1. 13 of the Stael-Holstein
document, where Kvacü is certain (I have also note of further
occurrence), I think that Chinese etymology must settle the
question.
1. 74, 'tütümni dittography of tu" is a curious case.
There is no dittography, and the second tü does not belong
to 1. 74, being a subscript correction of the 'blurred aksara'
in 1. 73.
1. 40, ha: = Ao:
I. 41, bhüiyü for bhüiyü.
We must thank Professor Bailey for getting rid of the
strange form bhüiyü. But the resultant ha : bhui does not
exhibit 'the rare use of bh for h\ The bhu is an attempt to
represent the sound which appears as vjt in 1. 78.
II. 48 and 49, ttü for tü (in 1. 25 ttü is given). The
reference to M. Leumann's Sakische Handschriftproben is
somewhat ineffective, since Dr. Leumann's plates illustrate
only the formal calligraphic hands of literary texts. The
question of tü and ttü is a little difficult, because those
published alphabets which use tt in place of t show a different
form of the vowel. But I have no objection to accepting
the ttü of Professor Konow and Professor Bailey, which
I have myself given in 1. 57, and I will not recur to the
ancient misere as to tt or nt (as Dr. Hoernle at first read),
although the ttd for d in some alphabets (see Hoernle,
JRAS, 1911, Plates I and II and Fig. 1, and Sir A. Stein's
Plate No. CXLV) is always raising qualms ; cf . Dr. Konow's
paper in the Berlin Academy Sitzungsberichte, 1935, pp. 777
-778, with a form graphically descended (as M. Lfevi re¬
marked) from nd, which he shows to have the value d.
1. 56, ttihi: (for ttihi:). The vowel looks like a com¬
promise between i and i. But we may accept ttihi:, which
in 11. 60, 77 we have given for the same word.
Coming now to the case of vjt in such words as yauvi/
yauti, süvijsüti, kavijkati. Professor Bailey, who himself
in 11. 1-24 gives 11 readings with v and only 2 with t, says
that 'alternation of -t-, -v- and zero was from the first' (but
he hardly means quite so much) 'observed in Khotanese':
it will be found noted by Leumann in Maitreya-samiti
(1919, pp. 36—8) and by Professor Konow in Saka Studies
(1932, pp. 26-7). It is not supposed that the alternate spellings
with t and v represent different pronunciations ; it is thought
that the intervocalic t had some weakened sound (which
Professor Bailey thinks [p. 12] may have been a glottal
stop) and so came to be employed where it had no etymo¬
logical right!). In our case v, which alternates with it,
should likewise denote a glottal stop.
I think that this view is not reasonable and that it is
not in accord with fact.
As to the first of these points: In the MS. we have
v -f- zero and y + zero in the tjrpes —
yau — yauvi, etc.: se — leyi, tseyi, etc.
SÜ — süvi, etc.: si = Jtyi, viyi,
where the final vowel is in all cases = zero. Why then do
we not find the t in place of the y as well as in place of the v ?
Secondly, it is remarkable how constant is the association
of the supposed t with a preceding u or with evidence of
representing a v. Professor Bailey adduces (p. 12) cases
where the t replaces older v (also vice versa, as in cvävaja
= cvätaja, BSOS IX, p. 933). But in the present text all
I) The Khotani poet whose work was brought to light by Pro¬
fessor Konow (Saka Studies, p. 112) was uninspired when he wrote,
'The letter t, the vigour of the things, explains the whole truth'.
38*
the t's which he would read either follow u or are substitutes
for v: of the latter case the only two instances are kati (1. 62),
which is Chinese chiao (older kau) and gati (1. 78), which
is Chinese wo^), both which instances present evidence of a m
or other labial element. But perhaps the most decisive evi¬
dence is provided by the form ha : bhui (the same word as
our gavi) noted above. The scribe could not have more
clearly indicated that what he was trying to represent was
a u\ Similarly we have Tibetan pahve = pao, '■'pau.
It is worth while to note in the Stael-Holstein document
the occurrences of this intervocal t, which in documents is
quite rare. Professor Konow reads it only in the words
Visa'sambhata (11. 7 and 33), Raurata (1. 12), Kautanai
(1.20), Karahattapata (1.30), Ttrüsahüta (11.30-1), Rrüki¬
bayarkäta (1. 31), Puskaka'jaiyäta (1. 47), hiaäta (1. 49),
hamäte (1. 53). Here ViSa'sambhata, denoting a person
whose name certainly was Vijaya^ambhava, must always
have made us uncomfortable; Raurata, if it denotes Loulan,
which is highly doubtful, can give no account of its t; Kau¬
tanai, if it means Kao-ch'ang or Qoco (as suggested by
Mr. CiiAUSON in JRAS, 1931, p. 306), can do so only by
being read as Kawanai (i.e. "/ai or ^ja or = [Kara]khoja);
Karahattapata, unless it is °pawa, contains an unexplained
final dissyllable ; Rrükibayarkäta, certainly Bayirkou"), must
be '^käuxi; Puskaka'jaiyäta is entirely unknown; and as to
htsdta and hamate, explained by Professor Konow as Con¬
junctives of ham and his, it is at any rate not supposed that
the t was pronoimced as such. Some similar forms are noted
infra (p. 601). Ttrüsahüta, where the t follows u, may owe
its non-identification to that reading with t.
This evidence will, I think, convince the impartial
reader that the sign regarded as a < has the value of a w.
But why not treat it simply as a u ? We may note that in
both these texts it occurs all but exclusively in final syllables.
1) From rigak. Jap. ga; but of ^m, Jap. go, with the same meaning.
2) As shown by Prof. Minorsky in BSOS, Vol. VIII, p. 918.
Might we suppose that, since it differs only slightly from the
acknowledged v, it is merely an example of the tendency
in writing, which MS. — and we are here dealing with MS. —
often exhibits, to slur the end of words ? Some persons use
at the end a different s from that used at the beginning,
and in the Arabic-Persian-Urdü script a differentiation is
crystallized.
There has been, I think, a failure to realize in what
manner the aksaras are inscribed. In writing the Tibetan
the scribe does not proceed as we might imagine : his method
is to begin nearly all aksaras with the part 7 or ^ and then
to add the remainder. A scrutiny of our MS. and of others
will show, in some cases by the inking, in others by imper¬
fect junctions at top or bottom, in others by both, that the
practice of the Sa-cu scribes was similar. Might we then
suppose that the signs read as <'s are simply imperfect v's ?
Such a suggestion is refuted by the fact that such 'imper¬
fect v's' vsver occur initially; and the only remaining proba¬
bility might be that the form was systematically modified in
order to represent a special sound, which must have been w.
For we cannot suppose that the name [ Vijaya]8ambhava could
be written [Visa']samb}iata because both v and t (cf. Gagana¬
sambhava in our text, 1. 7) represented a glottal stop.
This brings us to our second point, namely that the sign
used is, in faet, not a t. In MSS. comparable to the present
an uncompounded < is a rarity, being regularly replaced
by tt. But in a compound aksara it is quite common, and
it has the shape of the lower part of tt: see ttihi, tceyi, tseyi
in the annexed illustration. Evidently the uncompounded t
had in this alphabet that same shape. This is not mere in¬
ference: for the illustration shows the uncompounded t in
that form (by the side of the alternative tt), and the same
may be found in several recurrences in the MS. Ch. XLVI.
0015 (fol. 7) photographed on Plate CL of Serindia and on
Plate VI in Dr. Hoeenle's article*). The form of t thus
1) A MS. in this same alphabet contains a Sanskrit verse with
the same t (in dauta = düta) repeated, as well as tt.
definitely taught and used excludes the sign in the supposed
ti of our illustration from even a graphic signification of t.
Moreover, the numerous MS. alphabets seen in MSS. from
Sa-cu show a complete unawareness of the existence of
the sign.
What then has been the temptation to read t % Probably
its moderate resemblance to an older t of literary and other
MSS. Concerning this form Dr. Hoernle wrote (JRAS,
1911, p. 468).
'the signs for n and t are easily distinguishable, for t is
written with a very elongated left limb, while n has two
short and equal limbs'
and he refers to his Plate VI. The form does occur in several
documents (see, for the example, Plate XVII of Dr. Hoeenle's
Manuscript Remains . . .) having an alphabet rather similar
to that of our MS.; and it tends to preserve its shape in
composition. The document illustrated by Hoeenle's Plate
comes from the Khotan area, not from Sa-cu. Its uncom¬
pounded t does not resemble the uncompounded Sa-cu t (re¬
placed in our MS. by tt, but present in other MSS.), nor is
its compounded t the compounded Sa-cu t of our tt, tc, ts:
if our MS. had st, its t would not be that of the st in the
document. The t does not correspond to the t which it is
proposed to read in our MS.
We must, however, take account of two other MSS.
The first of these, reproduced in Plate CXLVI of Serindia,
has for the t of tathägata, etc., a form considerably resembling
the quasi-< of our MS. But, firstly, it will be seen upon
inspection of the script, which is rather calbgraphic, that
t is only a stylized form of the t with 'elongated left limb'
and that it is not formed in the same way as the supposed t
of our MS. ; and this is proved by the shape of the nt (lines 10,
14), pt (1. 13), st in stäm (col. 2, line 2), stä (col. 2, line 2),
and bj' a proportion of the occurrences of the t itself. More¬
over, the different character of the MS. is shown by its using
the old Saka-Khotani d.
The other MS. is a carefully written document in the
alphabet of our MS. : it uses tt, tc, tty, <s as in our MS. and
in the same forms: hence its implicit t is that of our MS.
Now we find the quasi -< occurring in —
hamäte (3 times), jsate (1), inäta (1), minäte (2), nisäte (1),
brräta (1) and brrätari (3), tcauta (1) and tcauti (1), Mtala
tsai (1).
Are these not conclusive proof of the <- value (graphic) of
the sign ? This prima facie view is, however, at once rendered
questionable by the circumstance that in this long document
there are so few t's and that out of these 15 eight are verbal
forms of the type of those found with the quasi-< in the Stael-
Holstein document {hamäte, hisäta), while in 4 more we have
the word for 'brother' {Jbräta), wherein the t had doubtless
long been dead ; see also the forms bräre, brärä, brärenu of
Professor Konow's 'Ein neuer Saka-Dialekt' (Index and
Introduction, p. 11). This very definitely proves that the
alphabetic system of the MS. is identical with that of the
Stael-Holstein document and that its quasi-< js in no way
accidental.
But does it not also definitely prove that the theory of
a decayed t, as a glottal stop, is correct after all ? That might
be so, but that the scribe has given away the whole case by
writing hitala tsai in a second occurrence as hivala tsä. For
him also, therefore, the quasi-< was equivalent to v. And
then we go on to observe that the forms tcauta and tcauti
again are instances of the sign following a u; and, turning
to Professor Bailey's paper, we find häva (and hä) 'root' by
the side of bäti and bäta. All this can only, it seems, mean
that the quasi-<, irrespective of its origin, bore the pronuncia¬
tion w, in Khotani as well as undeniably in the Chinese.
Beyond all question the idea of an alphabetic w was familiar
in äa-cu. For there, in the religious and scribal orders, the
'Tokhari' and Tibetan languages were familiar, and each of
them had for w a special sign.
That the Saka-Khotani spoken language, like Pro¬
fessor Konow's 'New Saka Dialect', possessed a t« is evident
from the names such as Winesa, recorded in Tibetan script.
It seems as if in the transcription of our MS., and of others
with similar employment of the sign in question, the t should
be replaced by w^).
Concerning the final i I had written —
Tt is possible that a speciality of this final i is indicated
in the actual script, where . . .; and we are tempted to
represent the vowel in this use by a special sign, say t;
but the differential form is not always obvious, and, as no
misunderstanding is caused by the use of the ordinary i,
it seems preferable to abstain.'
In accord with this suggestion, first announced at the Bonn
Orientalisten Tag in 1936 (Report, p. 9), Professor Bailey
has now, so far as I can judge, made a faithful, and for the
most part successful, attempt to distinguish ocularly between
the two forms. But we should not forget that we have before
us not a print, but a MS. If we had to transcribe into, say,
Sanskrit or Saka-Khotani a letter or a MS. written by one
of our correspondents, we should follow the certain intention
of the writer, even if his e's sometimes had the appearance
of t's or vice versa. This is a normalizing procedure, which
Professor Bailey has sedulously avoided, accepting incon¬
sistencies such as are exhibited in the illustration. I note
siysi (1.3), tsiysi (11.9-10), thayi (1.9), tseyi (1.12),
khiyi (1. 10), siysi (1. 16), ayi (1. 18), tsyeyi (1. 14).
But even this heroic procedure will not preclude a reasoned
dispute in regard to some of the instances : and, when we see
that in cases such as cimni, sauvi and so forth Professor
Bailey has always found i and not i, we suspect even a
certain intrusion of non-ocular considerations. Anyone who
will scan the photographs will see that mere ocular test
might produce a different result with a fresh reader.
Nevertheless, the difference in the intended form is real.
1) In the Uttara-tarUra fragment edited by H. W. BAiLEy and
E. H. Johnston in BSOS, VIII pp. 77 sqq., there is a single
instance, in the word hvata, aU other t'« being represented by tt.
This hvata will be parallel, in respect of the t, to the hamäte, etc.,
noted above.
I would add that it exists in the Stael-Holstein document and
possibly also in the documents edited by Profe,ssor Bailey
and even in all such documents hitherto edited or shown in
Plates. There remains for specialist scholars the delicate
task of dealing with the occurrences.
Professor Bailey proposes to derive the sign in question
from the old Saka-Khotani ä, an idea which was in my mind
also. It may be asked, why then not represent it by ö or e
(like the silent final e in English) ? No doubt the specialists
could give good reasons for preferring i. One which has
come under my observation may be cited. A Saka-Khotani
MS. from Sa-cu mentions some Indian place-names, among
which are ' Pätaliputtri' and Ttahiksastlai (= Taksasilä).
The second name, with the hi for or y which we have in
our Chinese text, is written with an indubitable i.
As regards my reading syom, and not syau, in some
occurrences of a word which also appears as syä and syäm,
I apprehend that in the script -om would be indistinguishable
from -au. Beyond all question the dominant pronunciation
at äa-cu was with o, which I cited also from Tibetan tran¬
scription. Since what in Saka-Khotani texts has been re¬
presented as au has too many values (Professor Bailey,
pp. 3^), perhaps it may be worth while to consider whether
here also a differentiation can be effected.
As a postscript to this discussion of critical matters I
should like to add an observation which may be new.
In a Saka-Khotani MS. cited above we have a certain
example of the guttural n, clearly distinguished from j. It
occurs in the repeated phrase (an exercise in dating docu¬
ments) —
thina (sometimes thyina) hint nämye, ksuni asi salya (or
°yai) munaji mästa simyävija pidaki ttyi prracina cä.
In one instance there is added —
ahämgulasaihvada ttyi kana ahämhvada cathalyai ri,
concerning which I need not mention any conjectures. In
the first phrase, a formal opening which is found also in
several of the Hoernle documents (see JASB, 1897,
Plates XII, XIII; 1902, Plate VI), with such variants as cu
for the final cä, we recognize at once, with the help of
Professor Bailey's paper in BSOS IX (1938) p. 541, the,
'Horse' year {asi salya) and the month Mufiaji: simyavija
I must leave to others. That thinahini is a Chinese Nien-hao
date, similar to the cm Mna given by Professor Bailey {op. cit.
p. 541), being equivalent to a t'ien-heng {hsingi), is clear
from a similar passage in the calligraphic MS., which reads —
ttä thyemni cümni sü hye: bädi tcürmye ksumni asi salya
padauysye mäiti nämmye hadai Kvamnina ämna räsiya
stäm dü yude.
Here I do not venture to tackle sü hye (Is it puhye 'fifth' ?);
but tcürmye may be the ordinal of tcahaura 'four' and bädi
means 'time'. The räJi Kvamnina might perhaps be a Kämna.
So we have the fourth ksumna or ksuna of the T'ien-chun
period, the Horse year, the first month, the 'name' ('ninth',
rvaumye'i) day and the Känma(?) räü^.
The occurrence of ksuna in connection with a Chinese
nien-hao date reminds us of the old difficulty in regard to
the vocalization of the Sanskrit-Prakrit word chuna = ksana,
used in the dating of Kusana inscriptions. But in 'Tokhari B'
documents also we have datings in ksum years (see M.
Levi's paper in Journal Asiatique, XI. II (1913), pp. 311 sqq.,
esp. pp. 312-8). Accordingly, in Kusana India, in Khotan
(Stael-Holstein document), in Kucä, in Shan-shan Kharosthi
(see Index) and in Sa-cu we have evidence of a word kfum
used in dating; and a probable explanation of this wide¬
spread use is that the word is affected by a Chinese term
{ksum, chung, j^, 'cycle', 'year') established by the Chinese
rule in Turkestan.
November 1938.
1) Professor Bailey would read thyemni tcümni, T'ien-'tsun'
and translate aühye 'auspicious', and then continue 'from the Kvani we
have condescended pastämdu) to make a ruling'. This is preferable.
(1) Since I understand that the alternation of t and v
causes some difficulty to those not immediately concerned
with the Khotanese language, the matter, which had seemed
rather obvious, may perhaps be made more clear. The ex¬
planation seems to be simple. Additional evidence from an¬
other source the Siddhasära (Ch ii 002), now available in fac¬
simile in Codices Khotanenses (Monumenta Linguarum Asiae
Maioris, vol. II), will assist here. The scribe has used the
one identical ak§ara ta in the following ways (the words are
selected out of a large number): —
i. with Old Iranian etymological value and in Sanskrit
words, hauta 6 v 1, hota 1 bis r 3 (beside hauva, hova 7 r 5),
bäta 'wind' 4 r 2, bäta 'root' 129 v 2 (beside häva, bä 129 v 5),
hodäta 'seventy' 7 r 2, datäni 'wild beasts' 144 v ^, mahäh-
hüta 3 r 1.
ii. initial, tavadye 20 t 2 = ttavamdye pa.ssim, tümgurn
101 r .5, 124 V 1 = ttümgara, tikyäm 129 r 4 = ttikyäm.
iii. medial, interchanging with v and zero, hauta, hauva..
bäta, bäva, bä as above, detadärä 139 r 4, devadärä 139 r .').
iv. medial, interchanging with tt, thyauta 132 v 1, 152 r 4
= thyautta 128 r 2 et passim, tta-ta 152 v 3 = <ta tta pa.ssim.
V. interpolated, hamyata Uka 155 v 1 = ham,yu iJka,
hemnäta chava 150 r 3, as dätäksinya quoted in Konow, Saka
Studies 27.
vi. in conjunct groups, ciräthiihtui 105 v 2 beside cirat-
taihttai 101 r 5 (translating Skt. bhüniwha, cf. Skt. ciratikta),
nahta:makala 143 v 4, trold 104 r 4 besiehe ttrrolä 107 r 2,
triphalä 14 r 2, citrai 142 r 4, ätmagaupta 10 r 1, paste 132 v 5,
güsta 134 v 4, habair.std 109 r 4.
An identical aksara formed in the same way wilh tiie
value ta is found in many other documents, so on plnte